159 files changed, 63225 insertions, 0 deletions

diff --git a/libjpegtwrp/Android.mk b/libjpegtwrp/Android.mk
new file mode 100755
index 000000000..de852cb03
--- /dev/null
+++ b/libjpegtwrp/Android.mk
@@ -0,0 +1,79 @@
+LOCAL_PATH:= $(call my-dir)
+include $(CLEAR_VARS)
+
+LOCAL_ARM_MODE := arm
+
+# Set ANDROID_JPEG_USE_VENUM to true to enable VeNum optimizations
+ANDROID_JPEG_USE_VENUM := true
+
+# Disable VeNum optimizations if they are not supported on the build target
+ifneq ($(ARCH_ARM_HAVE_VFP),true)
+ANDROID_JPEG_USE_VENUM := false
+else
+ifneq ($(ARCH_ARM_HAVE_NEON),true)
+ANDROID_JPEG_USE_VENUM := false
+endif
+endif
+
+LOCAL_SRC_FILES := \
+	jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c \
+	jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c \
+	jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c \
+	jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c \
+	jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c \
+	jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c \
+	jfdctint.c jidctflt.c jquant1.c \
+	jquant2.c jutils.c jmemmgr.c \
+
+# use ashmem as libjpeg decoder's backing store
+LOCAL_CFLAGS += -DUSE_ANDROID_ASHMEM
+LOCAL_SRC_FILES += \
+	jmem-ashmem.c
+
+# the original android memory manager.
+# use sdcard as libjpeg decoder's backing store
+#LOCAL_SRC_FILES += \
+#	jmem-android.c
+
+
+# the assembler is only for the ARM version, don't break the Linux sim
+ifneq ($(TARGET_ARCH),arm)
+ANDROID_JPEG_NO_ASSEMBLER := true
+endif
+
+ifeq ($(strip $(ANDROID_JPEG_NO_ASSEMBLER)),true)
+LOCAL_SRC_FILES += jidctint.c jidctfst.c jidctred.c
+else
+ifeq ($(ANDROID_JPEG_USE_VENUM),true)
+LOCAL_SRC_FILES += jidctvenum.c
+LOCAL_SRC_FILES += asm/armv7/jdcolor-armv7.S
+LOCAL_SRC_FILES += asm/armv7/jdcolor-android-armv7.S
+LOCAL_SRC_FILES += asm/armv7/jdidct-armv7.S
+LOCAL_CFLAGS    += -DANDROID_JPEG_USE_VENUM
+else # ANDROID_JPEG_USE_VENUM, false
+LOCAL_SRC_FILES += jidctint.c jidctred.c jidctfst.c armv6_idct.S
+LOCAL_CFLAGS    += -DANDROID_ARMV6_IDCT
+endif # ANDROID_JPEG_USE_VENUM
+endif
+
+LOCAL_CFLAGS += -DAVOID_TABLES
+LOCAL_CFLAGS += -O3 -fstrict-aliasing -fprefetch-loop-arrays
+
+# enable tile based decode
+LOCAL_CFLAGS += -DANDROID_TILE_BASED_DECODE
+
+ifdef NEEDS_ARM_ERRATA_754319_754320
+asm_flags := \
+	--defsym NEEDS_ARM_ERRATA_754319_754320_ASM=1
+
+LOCAL_CFLAGS+= \
+	$(foreach f,$(asm_flags),-Wa,"$(f)")
+endif
+
+LOCAL_MODULE:= libjpegtwrp
+
+LOCAL_SHARED_LIBRARIES := \
+	libcutils
+
+include $(BUILD_STATIC_LIBRARY)
+
diff --git a/libjpegtwrp/CleanSpec.mk b/libjpegtwrp/CleanSpec.mk
new file mode 100644
index 000000000..b84e1b65e
--- /dev/null
+++ b/libjpegtwrp/CleanSpec.mk
@@ -0,0 +1,49 @@
+# Copyright (C) 2007 The Android Open Source Project
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+# If you don't need to do a full clean build but would like to touch
+# a file or delete some intermediate files, add a clean step to the end
+# of the list.  These steps will only be run once, if they haven't been
+# run before.
+#
+# E.g.:
+#     $(call add-clean-step, touch -c external/sqlite/sqlite3.h)
+#     $(call add-clean-step, rm -rf $(PRODUCT_OUT)/obj/STATIC_LIBRARIES/libz_intermediates)
+#
+# Always use "touch -c" and "rm -f" or "rm -rf" to gracefully deal with
+# files that are missing or have been moved.
+#
+# Use $(PRODUCT_OUT) to get to the "out/target/product/blah/" directory.
+# Use $(OUT_DIR) to refer to the "out" directory.
+#
+# If you need to re-do something that's already mentioned, just copy
+# the command and add it to the bottom of the list.  E.g., if a change
+# that you made last week required touching a file and a change you
+# made today requires touching the same file, just copy the old
+# touch step and add it to the end of the list.
+#
+# ************************************************
+# NEWER CLEAN STEPS MUST BE AT THE END OF THE LIST
+# ************************************************
+
+# For example:
+#$(call add-clean-step, rm -rf $(OUT_DIR)/target/common/obj/APPS/AndroidTests_intermediates)
+#$(call add-clean-step, rm -rf $(OUT_DIR)/target/common/obj/JAVA_LIBRARIES/core_intermediates)
+#$(call add-clean-step, find $(OUT_DIR) -type f -name "IGTalkSession*" -print0 | xargs -0 rm -f)
+#$(call add-clean-step, rm -rf $(PRODUCT_OUT)/data/*)
+
+# ************************************************
+# NEWER CLEAN STEPS MUST BE AT THE END OF THE LIST
+# ************************************************
diff --git a/libjpegtwrp/MODULE_LICENSE_BSD_LIKE b/libjpegtwrp/MODULE_LICENSE_BSD_LIKE
new file mode 100644
index 000000000..e69de29bb
--- /dev/null
+++ b/libjpegtwrp/MODULE_LICENSE_BSD_LIKE
diff --git a/libjpegtwrp/NOTICE b/libjpegtwrp/NOTICE
new file mode 100644
index 000000000..007625fec
--- /dev/null
+++ b/libjpegtwrp/NOTICE
@@ -0,0 +1,38 @@
+This software is based in part on the work of the Independent JPEG Group.
+
+----------------------
+
+The authors make NO WARRANTY or representation, either express or implied,
+with respect to this software, its quality, accuracy, merchantability, or
+fitness for a particular purpose.  This software is provided "AS IS", and you,
+its user, assume the entire risk as to its quality and accuracy.
+
+This software is copyright (C) 1991-1998, Thomas G. Lane.
+All Rights Reserved except as specified below.
+
+Permission is hereby granted to use, copy, modify, and distribute this
+software (or portions thereof) for any purpose, without fee, subject to these
+conditions:
+(1) If any part of the source code for this software is distributed, then this
+README file must be included, with this copyright and no-warranty notice
+unaltered; and any additions, deletions, or changes to the original files
+must be clearly indicated in accompanying documentation.
+(2) If only executable code is distributed, then the accompanying
+documentation must state that "this software is based in part on the work of
+the Independent JPEG Group".
+(3) Permission for use of this software is granted only if the user accepts
+full responsibility for any undesirable consequences; the authors accept
+NO LIABILITY for damages of any kind.
+
+These conditions apply to any software derived from or based on the IJG code,
+not just to the unmodified library.  If you use our work, you ought to
+acknowledge us.
+
+Permission is NOT granted for the use of any IJG author's name or company name
+in advertising or publicity relating to this software or products derived from
+it.  This software may be referred to only as "the Independent JPEG Group's
+software".
+
+We specifically permit and encourage the use of this software as the basis of
+commercial products, provided that all warranty or liability claims are
+assumed by the product vendor.
diff --git a/libjpegtwrp/README b/libjpegtwrp/README
new file mode 100644
index 000000000..86cc20669
--- /dev/null
+++ b/libjpegtwrp/README
@@ -0,0 +1,385 @@
+The Independent JPEG Group's JPEG software
+==========================================
+
+README for release 6b of 27-Mar-1998
+====================================
+
+This distribution contains the sixth public release of the Independent JPEG
+Group's free JPEG software.  You are welcome to redistribute this software and
+to use it for any purpose, subject to the conditions under LEGAL ISSUES, below.
+
+Serious users of this software (particularly those incorporating it into
+larger programs) should contact IJG at jpeg-info@uunet.uu.net to be added to
+our electronic mailing list.  Mailing list members are notified of updates
+and have a chance to participate in technical discussions, etc.
+
+This software is the work of Tom Lane, Philip Gladstone, Jim Boucher,
+Lee Crocker, Julian Minguillon, Luis Ortiz, George Phillips, Davide Rossi,
+Guido Vollbeding, Ge' Weijers, and other members of the Independent JPEG
+Group.
+
+IJG is not affiliated with the official ISO JPEG standards committee.
+
+
+DOCUMENTATION ROADMAP
+=====================
+
+This file contains the following sections:
+
+OVERVIEW            General description of JPEG and the IJG software.
+LEGAL ISSUES        Copyright, lack of warranty, terms of distribution.
+REFERENCES          Where to learn more about JPEG.
+ARCHIVE LOCATIONS   Where to find newer versions of this software.
+RELATED SOFTWARE    Other stuff you should get.
+FILE FORMAT WARS    Software *not* to get.
+TO DO               Plans for future IJG releases.
+
+Other documentation files in the distribution are:
+
+User documentation:
+  install.doc       How to configure and install the IJG software.
+  usage.doc         Usage instructions for cjpeg, djpeg, jpegtran,
+                    rdjpgcom, and wrjpgcom.
+  *.1               Unix-style man pages for programs (same info as usage.doc).
+  wizard.doc        Advanced usage instructions for JPEG wizards only.
+  change.log        Version-to-version change highlights.
+Programmer and internal documentation:
+  libjpeg.doc       How to use the JPEG library in your own programs.
+  example.c         Sample code for calling the JPEG library.
+  structure.doc     Overview of the JPEG library's internal structure.
+  filelist.doc      Road map of IJG files.
+  coderules.doc     Coding style rules --- please read if you contribute code.
+
+Please read at least the files install.doc and usage.doc.  Useful information
+can also be found in the JPEG FAQ (Frequently Asked Questions) article.  See
+ARCHIVE LOCATIONS below to find out where to obtain the FAQ article.
+
+If you want to understand how the JPEG code works, we suggest reading one or
+more of the REFERENCES, then looking at the documentation files (in roughly
+the order listed) before diving into the code.
+
+
+OVERVIEW
+========
+
+This package contains C software to implement JPEG image compression and
+decompression.  JPEG (pronounced "jay-peg") is a standardized compression
+method for full-color and gray-scale images.  JPEG is intended for compressing
+"real-world" scenes; line drawings, cartoons and other non-realistic images
+are not its strong suit.  JPEG is lossy, meaning that the output image is not
+exactly identical to the input image.  Hence you must not use JPEG if you
+have to have identical output bits.  However, on typical photographic images,
+very good compression levels can be obtained with no visible change, and
+remarkably high compression levels are possible if you can tolerate a
+low-quality image.  For more details, see the references, or just experiment
+with various compression settings.
+
+This software implements JPEG baseline, extended-sequential, and progressive
+compression processes.  Provision is made for supporting all variants of these
+processes, although some uncommon parameter settings aren't implemented yet.
+For legal reasons, we are not distributing code for the arithmetic-coding
+variants of JPEG; see LEGAL ISSUES.  We have made no provision for supporting
+the hierarchical or lossless processes defined in the standard.
+
+We provide a set of library routines for reading and writing JPEG image files,
+plus two sample applications "cjpeg" and "djpeg", which use the library to
+perform conversion between JPEG and some other popular image file formats.
+The library is intended to be reused in other applications.
+
+In order to support file conversion and viewing software, we have included
+considerable functionality beyond the bare JPEG coding/decoding capability;
+for example, the color quantization modules are not strictly part of JPEG
+decoding, but they are essential for output to colormapped file formats or
+colormapped displays.  These extra functions can be compiled out of the
+library if not required for a particular application.  We have also included
+"jpegtran", a utility for lossless transcoding between different JPEG
+processes, and "rdjpgcom" and "wrjpgcom", two simple applications for
+inserting and extracting textual comments in JFIF files.
+
+The emphasis in designing this software has been on achieving portability and
+flexibility, while also making it fast enough to be useful.  In particular,
+the software is not intended to be read as a tutorial on JPEG.  (See the
+REFERENCES section for introductory material.)  Rather, it is intended to
+be reliable, portable, industrial-strength code.  We do not claim to have
+achieved that goal in every aspect of the software, but we strive for it.
+
+We welcome the use of this software as a component of commercial products.
+No royalty is required, but we do ask for an acknowledgement in product
+documentation, as described under LEGAL ISSUES.
+
+
+LEGAL ISSUES
+============
+
+In plain English:
+
+1. We don't promise that this software works.  (But if you find any bugs,
+   please let us know!)
+2. You can use this software for whatever you want.  You don't have to pay us.
+3. You may not pretend that you wrote this software.  If you use it in a
+   program, you must acknowledge somewhere in your documentation that
+   you've used the IJG code.
+
+In legalese:
+
+The authors make NO WARRANTY or representation, either express or implied,
+with respect to this software, its quality, accuracy, merchantability, or
+fitness for a particular purpose.  This software is provided "AS IS", and you,
+its user, assume the entire risk as to its quality and accuracy.
+
+This software is copyright (C) 1991-1998, Thomas G. Lane.
+All Rights Reserved except as specified below.
+
+Permission is hereby granted to use, copy, modify, and distribute this
+software (or portions thereof) for any purpose, without fee, subject to these
+conditions:
+(1) If any part of the source code for this software is distributed, then this
+README file must be included, with this copyright and no-warranty notice
+unaltered; and any additions, deletions, or changes to the original files
+must be clearly indicated in accompanying documentation.
+(2) If only executable code is distributed, then the accompanying
+documentation must state that "this software is based in part on the work of
+the Independent JPEG Group".
+(3) Permission for use of this software is granted only if the user accepts
+full responsibility for any undesirable consequences; the authors accept
+NO LIABILITY for damages of any kind.
+
+These conditions apply to any software derived from or based on the IJG code,
+not just to the unmodified library.  If you use our work, you ought to
+acknowledge us.
+
+Permission is NOT granted for the use of any IJG author's name or company name
+in advertising or publicity relating to this software or products derived from
+it.  This software may be referred to only as "the Independent JPEG Group's
+software".
+
+We specifically permit and encourage the use of this software as the basis of
+commercial products, provided that all warranty or liability claims are
+assumed by the product vendor.
+
+
+ansi2knr.c is included in this distribution by permission of L. Peter Deutsch,
+sole proprietor of its copyright holder, Aladdin Enterprises of Menlo Park, CA.
+ansi2knr.c is NOT covered by the above copyright and conditions, but instead
+by the usual distribution terms of the Free Software Foundation; principally,
+that you must include source code if you redistribute it.  (See the file
+ansi2knr.c for full details.)  However, since ansi2knr.c is not needed as part
+of any program generated from the IJG code, this does not limit you more than
+the foregoing paragraphs do.
+
+The Unix configuration script "configure" was produced with GNU Autoconf.
+It is copyright by the Free Software Foundation but is freely distributable.
+The same holds for its supporting scripts (config.guess, config.sub,
+ltconfig, ltmain.sh).  Another support script, install-sh, is copyright
+by M.I.T. but is also freely distributable.
+
+It appears that the arithmetic coding option of the JPEG spec is covered by
+patents owned by IBM, AT&T, and Mitsubishi.  Hence arithmetic coding cannot
+legally be used without obtaining one or more licenses.  For this reason,
+support for arithmetic coding has been removed from the free JPEG software.
+(Since arithmetic coding provides only a marginal gain over the unpatented
+Huffman mode, it is unlikely that very many implementations will support it.)
+So far as we are aware, there are no patent restrictions on the remaining
+code.
+
+The IJG distribution formerly included code to read and write GIF files.
+To avoid entanglement with the Unisys LZW patent, GIF reading support has
+been removed altogether, and the GIF writer has been simplified to produce
+"uncompressed GIFs".  This technique does not use the LZW algorithm; the
+resulting GIF files are larger than usual, but are readable by all standard
+GIF decoders.
+
+We are required to state that
+    "The Graphics Interchange Format(c) is the Copyright property of
+    CompuServe Incorporated.  GIF(sm) is a Service Mark property of
+    CompuServe Incorporated."
+
+
+REFERENCES
+==========
+
+We highly recommend reading one or more of these references before trying to
+understand the innards of the JPEG software.
+
+The best short technical introduction to the JPEG compression algorithm is
+	Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+	Communications of the ACM, April 1991 (vol. 34 no. 4), pp. 30-44.
+(Adjacent articles in that issue discuss MPEG motion picture compression,
+applications of JPEG, and related topics.)  If you don't have the CACM issue
+handy, a PostScript file containing a revised version of Wallace's article is
+available at ftp://ftp.uu.net/graphics/jpeg/wallace.ps.gz.  The file (actually
+a preprint for an article that appeared in IEEE Trans. Consumer Electronics)
+omits the sample images that appeared in CACM, but it includes corrections
+and some added material.  Note: the Wallace article is copyright ACM and IEEE,
+and it may not be used for commercial purposes.
+
+A somewhat less technical, more leisurely introduction to JPEG can be found in
+"The Data Compression Book" by Mark Nelson and Jean-loup Gailly, published by
+M&T Books (New York), 2nd ed. 1996, ISBN 1-55851-434-1.  This book provides
+good explanations and example C code for a multitude of compression methods
+including JPEG.  It is an excellent source if you are comfortable reading C
+code but don't know much about data compression in general.  The book's JPEG
+sample code is far from industrial-strength, but when you are ready to look
+at a full implementation, you've got one here...
+
+The best full description of JPEG is the textbook "JPEG Still Image Data
+Compression Standard" by William B. Pennebaker and Joan L. Mitchell, published
+by Van Nostrand Reinhold, 1993, ISBN 0-442-01272-1.  Price US$59.95, 638 pp.
+The book includes the complete text of the ISO JPEG standards (DIS 10918-1
+and draft DIS 10918-2).  This is by far the most complete exposition of JPEG
+in existence, and we highly recommend it.
+
+The JPEG standard itself is not available electronically; you must order a
+paper copy through ISO or ITU.  (Unless you feel a need to own a certified
+official copy, we recommend buying the Pennebaker and Mitchell book instead;
+it's much cheaper and includes a great deal of useful explanatory material.)
+In the USA, copies of the standard may be ordered from ANSI Sales at (212)
+642-4900, or from Global Engineering Documents at (800) 854-7179.  (ANSI
+doesn't take credit card orders, but Global does.)  It's not cheap: as of
+1992, ANSI was charging $95 for Part 1 and $47 for Part 2, plus 7%
+shipping/handling.  The standard is divided into two parts, Part 1 being the
+actual specification, while Part 2 covers compliance testing methods.  Part 1
+is titled "Digital Compression and Coding of Continuous-tone Still Images,
+Part 1: Requirements and guidelines" and has document numbers ISO/IEC IS
+10918-1, ITU-T T.81.  Part 2 is titled "Digital Compression and Coding of
+Continuous-tone Still Images, Part 2: Compliance testing" and has document
+numbers ISO/IEC IS 10918-2, ITU-T T.83.
+
+Some extensions to the original JPEG standard are defined in JPEG Part 3,
+a newer ISO standard numbered ISO/IEC IS 10918-3 and ITU-T T.84.  IJG
+currently does not support any Part 3 extensions.
+
+The JPEG standard does not specify all details of an interchangeable file
+format.  For the omitted details we follow the "JFIF" conventions, revision
+1.02.  A copy of the JFIF spec is available from:
+	Literature Department
+	C-Cube Microsystems, Inc.
+	1778 McCarthy Blvd.
+	Milpitas, CA 95035
+	phone (408) 944-6300,  fax (408) 944-6314
+A PostScript version of this document is available by FTP at
+ftp://ftp.uu.net/graphics/jpeg/jfif.ps.gz.  There is also a plain text
+version at ftp://ftp.uu.net/graphics/jpeg/jfif.txt.gz, but it is missing
+the figures.
+
+The TIFF 6.0 file format specification can be obtained by FTP from
+ftp://ftp.sgi.com/graphics/tiff/TIFF6.ps.gz.  The JPEG incorporation scheme
+found in the TIFF 6.0 spec of 3-June-92 has a number of serious problems.
+IJG does not recommend use of the TIFF 6.0 design (TIFF Compression tag 6).
+Instead, we recommend the JPEG design proposed by TIFF Technical Note #2
+(Compression tag 7).  Copies of this Note can be obtained from ftp.sgi.com or
+from ftp://ftp.uu.net/graphics/jpeg/.  It is expected that the next revision
+of the TIFF spec will replace the 6.0 JPEG design with the Note's design.
+Although IJG's own code does not support TIFF/JPEG, the free libtiff library
+uses our library to implement TIFF/JPEG per the Note.  libtiff is available
+from ftp://ftp.sgi.com/graphics/tiff/.
+
+
+ARCHIVE LOCATIONS
+=================
+
+The "official" archive site for this software is ftp.uu.net (Internet
+address 192.48.96.9).  The most recent released version can always be found
+there in directory graphics/jpeg.  This particular version will be archived
+as ftp://ftp.uu.net/graphics/jpeg/jpegsrc.v6b.tar.gz.  If you don't have
+direct Internet access, UUNET's archives are also available via UUCP; contact
+help@uunet.uu.net for information on retrieving files that way.
+
+Numerous Internet sites maintain copies of the UUNET files.  However, only
+ftp.uu.net is guaranteed to have the latest official version.
+
+You can also obtain this software in DOS-compatible "zip" archive format from
+the SimTel archives (ftp://ftp.simtel.net/pub/simtelnet/msdos/graphics/), or
+on CompuServe in the Graphics Support forum (GO CIS:GRAPHSUP), library 12
+"JPEG Tools".  Again, these versions may sometimes lag behind the ftp.uu.net
+release.
+
+The JPEG FAQ (Frequently Asked Questions) article is a useful source of
+general information about JPEG.  It is updated constantly and therefore is
+not included in this distribution.  The FAQ is posted every two weeks to
+Usenet newsgroups comp.graphics.misc, news.answers, and other groups.
+It is available on the World Wide Web at http://www.faqs.org/faqs/jpeg-faq/
+and other news.answers archive sites, including the official news.answers
+archive at rtfm.mit.edu: ftp://rtfm.mit.edu/pub/usenet/news.answers/jpeg-faq/.
+If you don't have Web or FTP access, send e-mail to mail-server@rtfm.mit.edu
+with body
+	send usenet/news.answers/jpeg-faq/part1
+	send usenet/news.answers/jpeg-faq/part2
+
+
+RELATED SOFTWARE
+================
+
+Numerous viewing and image manipulation programs now support JPEG.  (Quite a
+few of them use this library to do so.)  The JPEG FAQ described above lists
+some of the more popular free and shareware viewers, and tells where to
+obtain them on Internet.
+
+If you are on a Unix machine, we highly recommend Jef Poskanzer's free
+PBMPLUS software, which provides many useful operations on PPM-format image
+files.  In particular, it can convert PPM images to and from a wide range of
+other formats, thus making cjpeg/djpeg considerably more useful.  The latest
+version is distributed by the NetPBM group, and is available from numerous
+sites, notably ftp://wuarchive.wustl.edu/graphics/graphics/packages/NetPBM/.
+Unfortunately PBMPLUS/NETPBM is not nearly as portable as the IJG software is;
+you are likely to have difficulty making it work on any non-Unix machine.
+
+A different free JPEG implementation, written by the PVRG group at Stanford,
+is available from ftp://havefun.stanford.edu/pub/jpeg/.  This program
+is designed for research and experimentation rather than production use;
+it is slower, harder to use, and less portable than the IJG code, but it
+is easier to read and modify.  Also, the PVRG code supports lossless JPEG,
+which we do not.  (On the other hand, it doesn't do progressive JPEG.)
+
+
+FILE FORMAT WARS
+================
+
+Some JPEG programs produce files that are not compatible with our library.
+The root of the problem is that the ISO JPEG committee failed to specify a
+concrete file format.  Some vendors "filled in the blanks" on their own,
+creating proprietary formats that no one else could read.  (For example, none
+of the early commercial JPEG implementations for the Macintosh were able to
+exchange compressed files.)
+
+The file format we have adopted is called JFIF (see REFERENCES).  This format
+has been agreed to by a number of major commercial JPEG vendors, and it has
+become the de facto standard.  JFIF is a minimal or "low end" representation.
+We recommend the use of TIFF/JPEG (TIFF revision 6.0 as modified by TIFF
+Technical Note #2) for "high end" applications that need to record a lot of
+additional data about an image.  TIFF/JPEG is fairly new and not yet widely
+supported, unfortunately.
+
+The upcoming JPEG Part 3 standard defines a file format called SPIFF.
+SPIFF is interoperable with JFIF, in the sense that most JFIF decoders should
+be able to read the most common variant of SPIFF.  SPIFF has some technical
+advantages over JFIF, but its major claim to fame is simply that it is an
+official standard rather than an informal one.  At this point it is unclear
+whether SPIFF will supersede JFIF or whether JFIF will remain the de-facto
+standard.  IJG intends to support SPIFF once the standard is frozen, but we
+have not decided whether it should become our default output format or not.
+(In any case, our decoder will remain capable of reading JFIF indefinitely.)
+
+Various proprietary file formats incorporating JPEG compression also exist.
+We have little or no sympathy for the existence of these formats.  Indeed,
+one of the original reasons for developing this free software was to help
+force convergence on common, open format standards for JPEG files.  Don't
+use a proprietary file format!
+
+
+TO DO
+=====
+
+The major thrust for v7 will probably be improvement of visual quality.
+The current method for scaling the quantization tables is known not to be
+very good at low Q values.  We also intend to investigate block boundary
+smoothing, "poor man's variable quantization", and other means of improving
+quality-vs-file-size performance without sacrificing compatibility.
+
+In future versions, we are considering supporting some of the upcoming JPEG
+Part 3 extensions --- principally, variable quantization and the SPIFF file
+format.
+
+As always, speeding things up is of great interest.
+
+Please send bug reports, offers of help, etc. to jpeg-info@uunet.uu.net.
diff --git a/libjpegtwrp/README-VeNum b/libjpegtwrp/README-VeNum
new file mode 100644
index 000000000..ee51449c7
--- /dev/null
+++ b/libjpegtwrp/README-VeNum
@@ -0,0 +1,20 @@
+README-VeNum
+Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum
+=================================================================
+
+Consumers are increasingly browsing Web based photo galleries incorporating
+high-resolution JPEG images with their mobile devices. These images are
+decoded by the mobile device, and typically also scaled down to fit the user's
+zoom selection on the device's browser.
+
+In order to improve both decode and download times, Qualcomm Innovation Center
+has optimized the JPEG library found on many common OS platforms such as Web
+OS, Android, and Chrome OS. Our team re-implemented several routines to utilize the
+DSP-like SIMD capabilities of the ARM NEON instruction set. These were then
+tuned and tested on Qualcomm's Snapdragon platform which implements the VeNum
+implementation of these same instructions.
+
+The specific areas of focus cover VeNum/NEON acceleration of Inverse Discrete
+Cosine Transform (iDCT) for 8x8, 4x4, 2x2, and 1x1 block sizes and YUV to RGB
+color space conversion. This resulted in a range of 18-32% improvement in JPEG
+decode and downscale times for images greater than 2Mpixels.
diff --git a/libjpegtwrp/ThirdPartyProject.prop b/libjpegtwrp/ThirdPartyProject.prop
new file mode 100644
index 000000000..e88cc6355
--- /dev/null
+++ b/libjpegtwrp/ThirdPartyProject.prop
@@ -0,0 +1,9 @@
+# Copyright 2010 Google Inc. All Rights Reserved.
+#Fri Jul 16 10:03:09 PDT 2010
+currentVersion=8a
+version=6b
+isNative=true
+name=jpeg
+keywords=jpeg
+onDevice=true
+homepage=http\://www.ijg.org/
diff --git a/libjpegtwrp/ansi2knr.1 b/libjpegtwrp/ansi2knr.1
new file mode 100644
index 000000000..f9ee5a631
--- /dev/null
+++ b/libjpegtwrp/ansi2knr.1
@@ -0,0 +1,36 @@
+.TH ANSI2KNR 1 "19 Jan 1996"
+.SH NAME
+ansi2knr \- convert ANSI C to Kernighan & Ritchie C
+.SH SYNOPSIS
+.I ansi2knr
+[--varargs] input_file [output_file]
+.SH DESCRIPTION
+If no output_file is supplied, output goes to stdout.
+.br
+There are no error messages.
+.sp
+.I ansi2knr
+recognizes function definitions by seeing a non-keyword identifier at the left
+margin, followed by a left parenthesis, with a right parenthesis as the last
+character on the line, and with a left brace as the first token on the
+following line (ignoring possible intervening comments).  It will recognize a
+multi-line header provided that no intervening line ends with a left or right
+brace or a semicolon.  These algorithms ignore whitespace and comments, except
+that the function name must be the first thing on the line.
+.sp
+The following constructs will confuse it:
+.br
+     - Any other construct that starts at the left margin and follows the
+above syntax (such as a macro or function call).
+.br
+     - Some macros that tinker with the syntax of the function header.
+.sp
+The --varargs switch is obsolete, and is recognized only for
+backwards compatibility.  The present version of
+.I ansi2knr
+will always attempt to convert a ... argument to va_alist and va_dcl.
+.SH AUTHOR
+L. Peter Deutsch <ghost@aladdin.com> wrote the original ansi2knr and
+continues to maintain the current version; most of the code in the current
+version is his work.  ansi2knr also includes contributions by Francois
+Pinard <pinard@iro.umontreal.ca> and Jim Avera <jima@netcom.com>.
diff --git a/libjpegtwrp/ansi2knr.c b/libjpegtwrp/ansi2knr.c
new file mode 100644
index 000000000..4e05fc2d3
--- /dev/null
+++ b/libjpegtwrp/ansi2knr.c
@@ -0,0 +1,693 @@
+/* ansi2knr.c */
+/* Convert ANSI C function definitions to K&R ("traditional C") syntax */
+
+/*
+ansi2knr is distributed in the hope that it will be useful, but WITHOUT ANY
+WARRANTY.  No author or distributor accepts responsibility to anyone for the
+consequences of using it or for whether it serves any particular purpose or
+works at all, unless he says so in writing.  Refer to the GNU General Public
+License (the "GPL") for full details.
+
+Everyone is granted permission to copy, modify and redistribute ansi2knr,
+but only under the conditions described in the GPL.  A copy of this license
+is supposed to have been given to you along with ansi2knr so you can know
+your rights and responsibilities.  It should be in a file named COPYLEFT.
+[In the IJG distribution, the GPL appears below, not in a separate file.]
+Among other things, the copyright notice and this notice must be preserved
+on all copies.
+
+We explicitly state here what we believe is already implied by the GPL: if
+the ansi2knr program is distributed as a separate set of sources and a
+separate executable file which are aggregated on a storage medium together
+with another program, this in itself does not bring the other program under
+the GPL, nor does the mere fact that such a program or the procedures for
+constructing it invoke the ansi2knr executable bring any other part of the
+program under the GPL.
+*/
+
+/*
+---------- Here is the GNU GPL file COPYLEFT, referred to above ----------
+----- These terms do NOT apply to the JPEG software itself; see README ------
+
+		    GHOSTSCRIPT GENERAL PUBLIC LICENSE
+		    (Clarified 11 Feb 1988)
+
+ Copyright (C) 1988 Richard M. Stallman
+ Everyone is permitted to copy and distribute verbatim copies of this
+ license, but changing it is not allowed.  You can also use this wording
+ to make the terms for other programs.
+
+  The license agreements of most software companies keep you at the
+mercy of those companies.  By contrast, our general public license is
+intended to give everyone the right to share Ghostscript.  To make sure
+that you get the rights we want you to have, we need to make
+restrictions that forbid anyone to deny you these rights or to ask you
+to surrender the rights.  Hence this license agreement.
+
+  Specifically, we want to make sure that you have the right to give
+away copies of Ghostscript, that you receive source code or else can get
+it if you want it, that you can change Ghostscript or use pieces of it
+in new free programs, and that you know you can do these things.
+
+  To make sure that everyone has such rights, we have to forbid you to
+deprive anyone else of these rights.  For example, if you distribute
+copies of Ghostscript, you must give the recipients all the rights that
+you have.  You must make sure that they, too, receive or can get the
+source code.  And you must tell them their rights.
+
+  Also, for our own protection, we must make certain that everyone finds
+out that there is no warranty for Ghostscript.  If Ghostscript is
+modified by someone else and passed on, we want its recipients to know
+that what they have is not what we distributed, so that any problems
+introduced by others will not reflect on our reputation.
+
+  Therefore we (Richard M. Stallman and the Free Software Foundation,
+Inc.) make the following terms which say what you must do to be allowed
+to distribute or change Ghostscript.
+
+
+			COPYING POLICIES
+
+  1. You may copy and distribute verbatim copies of Ghostscript source
+code as you receive it, in any medium, provided that you conspicuously
+and appropriately publish on each copy a valid copyright and license
+notice "Copyright (C) 1989 Aladdin Enterprises.  All rights reserved.
+Distributed by Free Software Foundation, Inc." (or with whatever year is
+appropriate); keep intact the notices on all files that refer to this
+License Agreement and to the absence of any warranty; and give any other
+recipients of the Ghostscript program a copy of this License Agreement
+along with the program.  You may charge a distribution fee for the
+physical act of transferring a copy.
+
+  2. You may modify your copy or copies of Ghostscript or any portion of
+it, and copy and distribute such modifications under the terms of
+Paragraph 1 above, provided that you also do the following:
+
+    a) cause the modified files to carry prominent notices stating
+    that you changed the files and the date of any change; and
+
+    b) cause the whole of any work that you distribute or publish,
+    that in whole or in part contains or is a derivative of Ghostscript
+    or any part thereof, to be licensed at no charge to all third
+    parties on terms identical to those contained in this License
+    Agreement (except that you may choose to grant more extensive
+    warranty protection to some or all third parties, at your option).
+
+    c) You may charge a distribution fee for the physical act of
+    transferring a copy, and you may at your option offer warranty
+    protection in exchange for a fee.
+
+Mere aggregation of another unrelated program with this program (or its
+derivative) on a volume of a storage or distribution medium does not bring
+the other program under the scope of these terms.
+
+  3. You may copy and distribute Ghostscript (or a portion or derivative
+of it, under Paragraph 2) in object code or executable form under the
+terms of Paragraphs 1 and 2 above provided that you also do one of the
+following:
+
+    a) accompany it with the complete corresponding machine-readable
+    source code, which must be distributed under the terms of
+    Paragraphs 1 and 2 above; or,
+
+    b) accompany it with a written offer, valid for at least three
+    years, to give any third party free (except for a nominal
+    shipping charge) a complete machine-readable copy of the
+    corresponding source code, to be distributed under the terms of
+    Paragraphs 1 and 2 above; or,
+
+    c) accompany it with the information you received as to where the
+    corresponding source code may be obtained.  (This alternative is
+    allowed only for noncommercial distribution and only if you
+    received the program in object code or executable form alone.)
+
+For an executable file, complete source code means all the source code for
+all modules it contains; but, as a special exception, it need not include
+source code for modules which are standard libraries that accompany the
+operating system on which the executable file runs.
+
+  4. You may not copy, sublicense, distribute or transfer Ghostscript
+except as expressly provided under this License Agreement.  Any attempt
+otherwise to copy, sublicense, distribute or transfer Ghostscript is
+void and your rights to use the program under this License agreement
+shall be automatically terminated.  However, parties who have received
+computer software programs from you with this License Agreement will not
+have their licenses terminated so long as such parties remain in full
+compliance.
+
+  5. If you wish to incorporate parts of Ghostscript into other free
+programs whose distribution conditions are different, write to the Free
+Software Foundation at 675 Mass Ave, Cambridge, MA 02139.  We have not
+yet worked out a simple rule that can be stated here, but we will often
+permit this.  We will be guided by the two goals of preserving the free
+status of all derivatives of our free software and of promoting the
+sharing and reuse of software.
+
+Your comments and suggestions about our licensing policies and our
+software are welcome!  Please contact the Free Software Foundation,
+Inc., 675 Mass Ave, Cambridge, MA 02139, or call (617) 876-3296.
+
+		       NO WARRANTY
+
+  BECAUSE GHOSTSCRIPT IS LICENSED FREE OF CHARGE, WE PROVIDE ABSOLUTELY
+NO WARRANTY, TO THE EXTENT PERMITTED BY APPLICABLE STATE LAW.  EXCEPT
+WHEN OTHERWISE STATED IN WRITING, FREE SOFTWARE FOUNDATION, INC, RICHARD
+M. STALLMAN, ALADDIN ENTERPRISES, L. PETER DEUTSCH, AND/OR OTHER PARTIES
+PROVIDE GHOSTSCRIPT "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER
+EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.  THE
+ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF GHOSTSCRIPT IS WITH
+YOU.  SHOULD GHOSTSCRIPT PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL
+NECESSARY SERVICING, REPAIR OR CORRECTION.
+
+  IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW WILL RICHARD M.
+STALLMAN, THE FREE SOFTWARE FOUNDATION, INC., L. PETER DEUTSCH, ALADDIN
+ENTERPRISES, AND/OR ANY OTHER PARTY WHO MAY MODIFY AND REDISTRIBUTE
+GHOSTSCRIPT AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING
+ANY LOST PROFITS, LOST MONIES, OR OTHER SPECIAL, INCIDENTAL OR
+CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE
+(INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING RENDERED
+INACCURATE OR LOSSES SUSTAINED BY THIRD PARTIES OR A FAILURE OF THE
+PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS) GHOSTSCRIPT, EVEN IF YOU
+HAVE BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES, OR FOR ANY CLAIM
+BY ANY OTHER PARTY.
+
+-------------------- End of file COPYLEFT ------------------------------
+*/
+
+/*
+ * Usage:
+	ansi2knr input_file [output_file]
+ * If no output_file is supplied, output goes to stdout.
+ * There are no error messages.
+ *
+ * ansi2knr recognizes function definitions by seeing a non-keyword
+ * identifier at the left margin, followed by a left parenthesis,
+ * with a right parenthesis as the last character on the line,
+ * and with a left brace as the first token on the following line
+ * (ignoring possible intervening comments).
+ * It will recognize a multi-line header provided that no intervening
+ * line ends with a left or right brace or a semicolon.
+ * These algorithms ignore whitespace and comments, except that
+ * the function name must be the first thing on the line.
+ * The following constructs will confuse it:
+ *	- Any other construct that starts at the left margin and
+ *	    follows the above syntax (such as a macro or function call).
+ *	- Some macros that tinker with the syntax of the function header.
+ */
+
+/*
+ * The original and principal author of ansi2knr is L. Peter Deutsch
+ * <ghost@aladdin.com>.  Other authors are noted in the change history
+ * that follows (in reverse chronological order):
+	lpd 96-01-21 added code to cope with not HAVE_CONFIG_H and with
+		compilers that don't understand void, as suggested by
+		Tom Lane
+	lpd 96-01-15 changed to require that the first non-comment token
+		on the line following a function header be a left brace,
+		to reduce sensitivity to macros, as suggested by Tom Lane
+		<tgl@sss.pgh.pa.us>
+	lpd 95-06-22 removed #ifndefs whose sole purpose was to define
+		undefined preprocessor symbols as 0; changed all #ifdefs
+		for configuration symbols to #ifs
+	lpd 95-04-05 changed copyright notice to make it clear that
+		including ansi2knr in a program does not bring the entire
+		program under the GPL
+	lpd 94-12-18 added conditionals for systems where ctype macros
+		don't handle 8-bit characters properly, suggested by
+		Francois Pinard <pinard@iro.umontreal.ca>;
+		removed --varargs switch (this is now the default)
+	lpd 94-10-10 removed CONFIG_BROKETS conditional
+	lpd 94-07-16 added some conditionals to help GNU `configure',
+		suggested by Francois Pinard <pinard@iro.umontreal.ca>;
+		properly erase prototype args in function parameters,
+		contributed by Jim Avera <jima@netcom.com>;
+		correct error in writeblanks (it shouldn't erase EOLs)
+	lpd 89-xx-xx original version
+ */
+
+/* Most of the conditionals here are to make ansi2knr work with */
+/* or without the GNU configure machinery. */
+
+#if HAVE_CONFIG_H
+# include <config.h>
+#endif
+
+#include <stdio.h>
+#include <ctype.h>
+
+#if HAVE_CONFIG_H
+
+/*
+   For properly autoconfiguring ansi2knr, use AC_CONFIG_HEADER(config.h).
+   This will define HAVE_CONFIG_H and so, activate the following lines.
+ */
+
+# if STDC_HEADERS || HAVE_STRING_H
+#  include <string.h>
+# else
+#  include <strings.h>
+# endif
+
+#else /* not HAVE_CONFIG_H */
+
+/* Otherwise do it the hard way */
+
+# ifdef BSD
+#  include <strings.h>
+# else
+#  ifdef VMS
+    extern int strlen(), strncmp();
+#  else
+#   include <string.h>
+#  endif
+# endif
+
+#endif /* not HAVE_CONFIG_H */
+
+#if STDC_HEADERS
+# include <stdlib.h>
+#else
+/*
+   malloc and free should be declared in stdlib.h,
+   but if you've got a K&R compiler, they probably aren't.
+ */
+# ifdef MSDOS
+#  include <malloc.h>
+# else
+#  ifdef VMS
+     extern char *malloc();
+     extern void free();
+#  else
+     extern char *malloc();
+     extern int free();
+#  endif
+# endif
+
+#endif
+
+/*
+ * The ctype macros don't always handle 8-bit characters correctly.
+ * Compensate for this here.
+ */
+#ifdef isascii
+#  undef HAVE_ISASCII		/* just in case */
+#  define HAVE_ISASCII 1
+#else
+#endif
+#if STDC_HEADERS || !HAVE_ISASCII
+#  define is_ascii(c) 1
+#else
+#  define is_ascii(c) isascii(c)
+#endif
+
+#define is_space(c) (is_ascii(c) && isspace(c))
+#define is_alpha(c) (is_ascii(c) && isalpha(c))
+#define is_alnum(c) (is_ascii(c) && isalnum(c))
+
+/* Scanning macros */
+#define isidchar(ch) (is_alnum(ch) || (ch) == '_')
+#define isidfirstchar(ch) (is_alpha(ch) || (ch) == '_')
+
+/* Forward references */
+char *skipspace();
+int writeblanks();
+int test1();
+int convert1();
+
+/* The main program */
+int
+main(argc, argv)
+    int argc;
+    char *argv[];
+{	FILE *in, *out;
+#define bufsize 5000			/* arbitrary size */
+	char *buf;
+	char *line;
+	char *more;
+	/*
+	 * In previous versions, ansi2knr recognized a --varargs switch.
+	 * If this switch was supplied, ansi2knr would attempt to convert
+	 * a ... argument to va_alist and va_dcl; if this switch was not
+	 * supplied, ansi2knr would simply drop any such arguments.
+	 * Now, ansi2knr always does this conversion, and we only
+	 * check for this switch for backward compatibility.
+	 */
+	int convert_varargs = 1;
+
+	if ( argc > 1 && argv[1][0] == '-' )
+	  {	if ( !strcmp(argv[1], "--varargs") )
+		  {	convert_varargs = 1;
+			argc--;
+			argv++;
+		  }
+		else
+		  {	fprintf(stderr, "Unrecognized switch: %s\n", argv[1]);
+			exit(1);
+		  }
+	  }
+	switch ( argc )
+	   {
+	default:
+		printf("Usage: ansi2knr input_file [output_file]\n");
+		exit(0);
+	case 2:
+		out = stdout;
+		break;
+	case 3:
+		out = fopen(argv[2], "w");
+		if ( out == NULL )
+		   {	fprintf(stderr, "Cannot open output file %s\n", argv[2]);
+			exit(1);
+		   }
+	   }
+	in = fopen(argv[1], "r");
+	if ( in == NULL )
+	   {	fprintf(stderr, "Cannot open input file %s\n", argv[1]);
+		exit(1);
+	   }
+	fprintf(out, "#line 1 \"%s\"\n", argv[1]);
+	buf = malloc(bufsize);
+	line = buf;
+	while ( fgets(line, (unsigned)(buf + bufsize - line), in) != NULL )
+	   {
+test:		line += strlen(line);
+		switch ( test1(buf) )
+		   {
+		case 2:			/* a function header */
+			convert1(buf, out, 1, convert_varargs);
+			break;
+		case 1:			/* a function */
+			/* Check for a { at the start of the next line. */
+			more = ++line;
+f:			if ( line >= buf + (bufsize - 1) ) /* overflow check */
+			  goto wl;
+			if ( fgets(line, (unsigned)(buf + bufsize - line), in) == NULL )
+			  goto wl;
+			switch ( *skipspace(more, 1) )
+			  {
+			  case '{':
+			    /* Definitely a function header. */
+			    convert1(buf, out, 0, convert_varargs);
+			    fputs(more, out);
+			    break;
+			  case 0:
+			    /* The next line was blank or a comment: */
+			    /* keep scanning for a non-comment. */
+			    line += strlen(line);
+			    goto f;
+			  default:
+			    /* buf isn't a function header, but */
+			    /* more might be. */
+			    fputs(buf, out);
+			    strcpy(buf, more);
+			    line = buf;
+			    goto test;
+			  }
+			break;
+		case -1:		/* maybe the start of a function */
+			if ( line != buf + (bufsize - 1) ) /* overflow check */
+			  continue;
+			/* falls through */
+		default:		/* not a function */
+wl:			fputs(buf, out);
+			break;
+		   }
+		line = buf;
+	   }
+	if ( line != buf )
+	  fputs(buf, out);
+	free(buf);
+	fclose(out);
+	fclose(in);
+	return 0;
+}
+
+/* Skip over space and comments, in either direction. */
+char *
+skipspace(p, dir)
+    register char *p;
+    register int dir;			/* 1 for forward, -1 for backward */
+{	for ( ; ; )
+	   {	while ( is_space(*p) )
+		  p += dir;
+		if ( !(*p == '/' && p[dir] == '*') )
+		  break;
+		p += dir;  p += dir;
+		while ( !(*p == '*' && p[dir] == '/') )
+		   {	if ( *p == 0 )
+			  return p;	/* multi-line comment?? */
+			p += dir;
+		   }
+		p += dir;  p += dir;
+	   }
+	return p;
+}
+
+/*
+ * Write blanks over part of a string.
+ * Don't overwrite end-of-line characters.
+ */
+int
+writeblanks(start, end)
+    char *start;
+    char *end;
+{	char *p;
+	for ( p = start; p < end; p++ )
+	  if ( *p != '\r' && *p != '\n' )
+	    *p = ' ';
+	return 0;
+}
+
+/*
+ * Test whether the string in buf is a function definition.
+ * The string may contain and/or end with a newline.
+ * Return as follows:
+ *	0 - definitely not a function definition;
+ *	1 - definitely a function definition;
+ *	2 - definitely a function prototype (NOT USED);
+ *	-1 - may be the beginning of a function definition,
+ *		append another line and look again.
+ * The reason we don't attempt to convert function prototypes is that
+ * Ghostscript's declaration-generating macros look too much like
+ * prototypes, and confuse the algorithms.
+ */
+int
+test1(buf)
+    char *buf;
+{	register char *p = buf;
+	char *bend;
+	char *endfn;
+	int contin;
+
+	if ( !isidfirstchar(*p) )
+	  return 0;		/* no name at left margin */
+	bend = skipspace(buf + strlen(buf) - 1, -1);
+	switch ( *bend )
+	   {
+	   case ';': contin = 0 /*2*/; break;
+	   case ')': contin = 1; break;
+	   case '{': return 0;		/* not a function */
+	   case '}': return 0;		/* not a function */
+	   default: contin = -1;
+	   }
+	while ( isidchar(*p) )
+	  p++;
+	endfn = p;
+	p = skipspace(p, 1);
+	if ( *p++ != '(' )
+	  return 0;		/* not a function */
+	p = skipspace(p, 1);
+	if ( *p == ')' )
+	  return 0;		/* no parameters */
+	/* Check that the apparent function name isn't a keyword. */
+	/* We only need to check for keywords that could be followed */
+	/* by a left parenthesis (which, unfortunately, is most of them). */
+	   {	static char *words[] =
+		   {	"asm", "auto", "case", "char", "const", "double",
+			"extern", "float", "for", "if", "int", "long",
+			"register", "return", "short", "signed", "sizeof",
+			"static", "switch", "typedef", "unsigned",
+			"void", "volatile", "while", 0
+		   };
+		char **key = words;
+		char *kp;
+		int len = endfn - buf;
+
+		while ( (kp = *key) != 0 )
+		   {	if ( strlen(kp) == len && !strncmp(kp, buf, len) )
+			  return 0;	/* name is a keyword */
+			key++;
+		   }
+	   }
+	return contin;
+}
+
+/* Convert a recognized function definition or header to K&R syntax. */
+int
+convert1(buf, out, header, convert_varargs)
+    char *buf;
+    FILE *out;
+    int header;			/* Boolean */
+    int convert_varargs;	/* Boolean */
+{	char *endfn;
+	register char *p;
+	char **breaks;
+	unsigned num_breaks = 2;	/* for testing */
+	char **btop;
+	char **bp;
+	char **ap;
+	char *vararg = 0;
+
+	/* Pre-ANSI implementations don't agree on whether strchr */
+	/* is called strchr or index, so we open-code it here. */
+	for ( endfn = buf; *(endfn++) != '('; )
+	  ;
+top:	p = endfn;
+	breaks = (char **)malloc(sizeof(char *) * num_breaks * 2);
+	if ( breaks == 0 )
+	   {	/* Couldn't allocate break table, give up */
+		fprintf(stderr, "Unable to allocate break table!\n");
+		fputs(buf, out);
+		return -1;
+	   }
+	btop = breaks + num_breaks * 2 - 2;
+	bp = breaks;
+	/* Parse the argument list */
+	do
+	   {	int level = 0;
+		char *lp = NULL;
+		char *rp;
+		char *end = NULL;
+
+		if ( bp >= btop )
+		   {	/* Filled up break table. */
+			/* Allocate a bigger one and start over. */
+			free((char *)breaks);
+			num_breaks <<= 1;
+			goto top;
+		   }
+		*bp++ = p;
+		/* Find the end of the argument */
+		for ( ; end == NULL; p++ )
+		   {	switch(*p)
+			   {
+			   case ',':
+				if ( !level ) end = p;
+				break;
+			   case '(':
+				if ( !level ) lp = p;
+				level++;
+				break;
+			   case ')':
+				if ( --level < 0 ) end = p;
+				else rp = p;
+				break;
+			   case '/':
+				p = skipspace(p, 1) - 1;
+				break;
+			   default:
+				;
+			   }
+		   }
+		/* Erase any embedded prototype parameters. */
+		if ( lp )
+		  writeblanks(lp + 1, rp);
+		p--;			/* back up over terminator */
+		/* Find the name being declared. */
+		/* This is complicated because of procedure and */
+		/* array modifiers. */
+		for ( ; ; )
+		   {	p = skipspace(p - 1, -1);
+			switch ( *p )
+			   {
+			   case ']':	/* skip array dimension(s) */
+			   case ')':	/* skip procedure args OR name */
+			   {	int level = 1;
+				while ( level )
+				 switch ( *--p )
+				   {
+				   case ']': case ')': level++; break;
+				   case '[': case '(': level--; break;
+				   case '/': p = skipspace(p, -1) + 1; break;
+				   default: ;
+				   }
+			   }
+				if ( *p == '(' && *skipspace(p + 1, 1) == '*' )
+				   {	/* We found the name being declared */
+					while ( !isidfirstchar(*p) )
+					  p = skipspace(p, 1) + 1;
+					goto found;
+				   }
+				break;
+			   default:
+				goto found;
+			   }
+		   }
+found:		if ( *p == '.' && p[-1] == '.' && p[-2] == '.' )
+		  {	if ( convert_varargs )
+			  {	*bp++ = "va_alist";
+				vararg = p-2;
+			  }
+			else
+			  {	p++;
+				if ( bp == breaks + 1 )	/* sole argument */
+				  writeblanks(breaks[0], p);
+				else
+				  writeblanks(bp[-1] - 1, p);
+				bp--;
+			  }
+		   }
+		else
+		   {	while ( isidchar(*p) ) p--;
+			*bp++ = p+1;
+		   }
+		p = end;
+	   }
+	while ( *p++ == ',' );
+	*bp = p;
+	/* Make a special check for 'void' arglist */
+	if ( bp == breaks+2 )
+	   {	p = skipspace(breaks[0], 1);
+		if ( !strncmp(p, "void", 4) )
+		   {	p = skipspace(p+4, 1);
+			if ( p == breaks[2] - 1 )
+			   {	bp = breaks;	/* yup, pretend arglist is empty */
+				writeblanks(breaks[0], p + 1);
+			   }
+		   }
+	   }
+	/* Put out the function name and left parenthesis. */
+	p = buf;
+	while ( p != endfn ) putc(*p, out), p++;
+	/* Put out the declaration. */
+	if ( header )
+	  {	fputs(");", out);
+		for ( p = breaks[0]; *p; p++ )
+		  if ( *p == '\r' || *p == '\n' )
+		    putc(*p, out);
+	  }
+	else
+	  {	for ( ap = breaks+1; ap < bp; ap += 2 )
+		  {	p = *ap;
+			while ( isidchar(*p) )
+			  putc(*p, out), p++;
+			if ( ap < bp - 1 )
+			  fputs(", ", out);
+		  }
+		fputs(")  ", out);
+		/* Put out the argument declarations */
+		for ( ap = breaks+2; ap <= bp; ap += 2 )
+		  (*ap)[-1] = ';';
+		if ( vararg != 0 )
+		  {	*vararg = 0;
+			fputs(breaks[0], out);		/* any prior args */
+			fputs("va_dcl", out);		/* the final arg */
+			fputs(bp[0], out);
+		  }
+		else
+		  fputs(breaks[0], out);
+	  }
+	free((char *)breaks);
+	return 0;
+}
diff --git a/libjpegtwrp/armv6_idct.S b/libjpegtwrp/armv6_idct.S
new file mode 100644
index 000000000..18e4e8a18
--- /dev/null
+++ b/libjpegtwrp/armv6_idct.S
@@ -0,0 +1,366 @@
+/*
+ * Copyright (C) 2010 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/*
+ * This is a fast-and-accurate implementation of inverse Discrete Cosine
+ * Transform (IDCT) for ARMv6+. It also performs dequantization of the input
+ * coefficients just like other methods.
+ *
+ * This implementation is based on the scaled 1-D DCT algorithm proposed by
+ * Arai, Agui, and Nakajima. The following code is based on the figure 4-8
+ * on page 52 of the JPEG textbook by Pennebaker and Mitchell. Coefficients
+ * are (almost) directly mapped into registers.
+ *
+ * The accuracy is achieved by using SMULWy and SMLAWy instructions. Both
+ * multiply 32 bits by 16 bits and store the top 32 bits of the result. It
+ * makes 32-bit fixed-point arithmetic possible without overflow. That is
+ * why jpeg_idct_ifast(), which is written in C, cannot be improved.
+ *
+ * More tricks are used to gain more speed. First of all, we use as many
+ * registers as possible. ARM processor has 16 registers including sp (r13)
+ * and pc (r15), so only 14 registers can be used without limitations. In
+ * general, we let r0 to r7 hold the coefficients; r10 and r11 hold four
+ * 16-bit constants; r12 and r14 hold two of the four arguments; and r8 hold
+ * intermediate value. In the second pass, r9 is the loop counter. In the
+ * first pass, r8 to r11 are used to hold quantization values, so the loop
+ * counter is held by sp. Yes, the stack pointer. Since it must be aligned
+ * to 4-byte boundary all the time, we align it to 32-byte boundary and use
+ * bit 3 to bit 5. As the result, we actually use 14.1 registers. :-)
+ *
+ * Second, we rearrange quantization values to access them sequentially. The
+ * table is first transposed, and the new columns are placed in the order of
+ * 7, 5, 1, 3, 0, 2, 4, 6. Thus we can use LDMDB to load four values at a
+ * time. Rearranging coefficients also helps, but that requires to change a
+ * dozen of files, which seems not worth it. In addition, we choose to scale
+ * up quantization values by 13 bits, so the coefficients are scaled up by
+ * 16 bits after both passes. Then we can pack and saturate them two at a
+ * time using PKHTB and USAT16 instructions.
+ *
+ * Third, we reorder the instructions to avoid bubbles in the pipeline. This
+ * is done by hand accroding to the cycle timings and the interlock behavior
+ * described in the technical reference manual of ARM1136JF-S. We also take
+ * advantage of dual issue processors by interleaving instructions with
+ * dependencies. It has been benchmarked on four devices and all the results
+ * showed distinguishable improvements. Note that PLD instructions actually
+ * slow things down, so they are removed at the last minute. In the future,
+ * this might be futher improved using a system profiler.
+ */
+
+#ifdef __arm__
+#include <machine/cpu-features.h>
+#endif
+
+#if __ARM_ARCH__ >= 6
+
+// void armv6_idct(short *coefs, int *quans, unsigned char *rows, int col)
+    .arm
+    .text
+    .align
+    .global armv6_idct
+    .func   armv6_idct
+
+armv6_idct:
+    // Push everything except sp (r13) and pc (r15).
+    stmdb   sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r12, r14}
+
+    // r12 = quans, r14 = coefs.
+    sub     r4, sp, #236
+    bic     sp, r4, #31
+    add     r5, sp, #224
+    add     r12, r1, #256
+    stm     r5, {r2, r3, r4}
+    add     r14, r0, #16
+
+pass1_head:
+    // Load quantization values. (q[0, 2, 4, 6])
+    ldmdb   r12!, {r8, r9, r10, r11}
+
+    // Load coefficients. (c[4, 1, 2, 3, 0, 5, 6, 7])
+    ldrsh   r4, [r14, #-2] !
+    ldrsh   r1, [r14, #16]
+    ldrsh   r2, [r14, #32]
+    ldrsh   r3, [r14, #48]
+    ldrsh   r0, [r14, #64]
+    ldrsh   r5, [r14, #80]
+    ldrsh   r6, [r14, #96]
+    ldrsh   r7, [r14, #112]
+
+    // r4 = q[0] * c[0];
+    mul     r4, r8, r4
+
+    // Check if ACs are all zero.
+    cmp     r0, #0
+    orreqs  r8, r1, r2
+    orreqs  r8, r3, r5
+    orreqs  r8, r6, r7
+    beq     pass1_zero
+
+    // Step 1: Dequantizations.
+
+    // r2 = q[2] * c[2];
+    // r0 = q[4] * c[4] + r4;
+    // r6 = q[6] * c[6] + r2;
+    mul     r2, r9, r2
+    mla     r0, r10, r0, r4
+    mla     r6, r11, r6, r2
+
+    // Load quantization values. (q[7, 5, 1, 3])
+    ldmdb   r12!, {r8, r9, r10, r11}
+
+    // r4 = r4 * 2 - r0 = -(r0 - r4 * 2);
+    // r2 = r2 * 2 - r6 = -(r6 - r2 * 2);
+    rsb     r4, r0, r4, lsl #1
+    rsb     r2, r6, r2, lsl #1
+
+    // r7 = q[7] * c[7];
+    // r5 = q[5] * c[5];
+    // r1 = q[1] * c[1] + r7;
+    // r3 = q[3] * c[3] + r5;
+    mul     r7, r8, r7
+    mul     r5, r9, r5
+    mla     r1, r10, r1, r7
+    mla     r3, r11, r3, r5
+
+    // Load constants.
+    ldrd    r10, constants
+
+    // Step 2: Rotations and Butterflies.
+
+    // r7 = r1 - r7 * 2;
+    // r1 = r1 - r3;
+    // r5 = r5 * 2 - r3 = -(r3 - r5 * 2);
+    // r3 = r1 + r3 * 2;
+    // r8 = r5 + r7;
+    sub     r7, r1, r7, lsl #1
+    sub     r1, r1, r3
+    rsb     r5, r3, r5, lsl #1
+    add     r3, r1, r3, lsl #1
+    add     r8, r5, r7
+
+    // r2 = r2 * 1.41421 = r2 * 27146 / 65536 + r2;
+    // r8 = r8 * 1.84776 / 8 = r8 * 15137 / 65536;
+    // r1 = r1 * 1.41421 = r1 * 27146 / 65536 + r1;
+    smlawt  r2, r2, r10, r2
+    smulwb  r8, r8, r10
+    smlawt  r1, r1, r10, r1
+
+    // r0 = r0 + r6;
+    // r2 = r2 - r6;
+    // r6 = r0 - r6 * 2;
+    add     r0, r0, r6
+    sub     r2, r2, r6
+    sub     r6, r0, r6, lsl #1
+
+    // r5 = r5 * -2.61313 / 8 + r8 = r5 * -21407 / 65536 + r8;
+    // r8 = r7 * -1.08239 / 8 + r8 = r7 * -8867 / 65536 + r8;
+    smlawt  r5, r5, r11, r8
+    smlawb  r8, r7, r11, r8
+
+    // r4 = r4 + r2;
+    // r0 = r0 + r3;
+    // r2 = r4 - r2 * 2;
+    add     r4, r4, r2
+    add     r0, r0, r3
+    sub     r2, r4, r2, lsl #1
+
+    // r7 = r5 * 8 - r3 = -(r3 - r5 * 8);
+    // r3 = r0 - r3 * 2;
+    // r1 = r1 - r7;
+    // r4 = r4 + r7;
+    // r5 = r8 * 8 - r1 = -(r1 - r8 * 8);
+    // r7 = r4 - r7 * 2;
+    rsb     r7, r3, r5, lsl #3
+    sub     r3, r0, r3, lsl #1
+    sub     r1, r1, r7
+    add     r4, r4, r7
+    rsb     r5, r1, r8, lsl #3
+    sub     r7, r4, r7, lsl #1
+
+    // r2 = r2 + r1;
+    // r6 = r6 + r5;
+    // r1 = r2 - r1 * 2;
+    // r5 = r6 - r5 * 2;
+    add     r2, r2, r1
+    add     r6, r6, r5
+    sub     r1, r2, r1, lsl #1
+    sub     r5, r6, r5, lsl #1
+
+    // Step 3: Reorder and Save.
+
+    str     r0, [sp, #-4] !
+    str     r4, [sp, #32]
+    str     r2, [sp, #64]
+    str     r6, [sp, #96]
+    str     r5, [sp, #128]
+    str     r1, [sp, #160]
+    str     r7, [sp, #192]
+    str     r3, [sp, #224]
+    b       pass1_tail
+
+    // Precomputed 16-bit constants: 27146, 15137, -21407, -8867.
+    // Put them in the middle since LDRD only accepts offsets from -255 to 255.
+    .align  3
+constants:
+    .word   0x6a0a3b21
+    .word   0xac61dd5d
+
+pass1_zero:
+    str     r4, [sp, #-4] !
+    str     r4, [sp, #32]
+    str     r4, [sp, #64]
+    str     r4, [sp, #96]
+    str     r4, [sp, #128]
+    str     r4, [sp, #160]
+    str     r4, [sp, #192]
+    str     r4, [sp, #224]
+    sub     r12, r12, #16
+
+pass1_tail:
+    ands    r9, sp, #31
+    bne     pass1_head
+
+    // r12 = rows, r14 = col.
+    ldr     r12, [sp, #256]
+    ldr     r14, [sp, #260]
+
+    // Load constants.
+    ldrd    r10, constants
+
+pass2_head:
+    // Load coefficients. (c[0, 1, 2, 3, 4, 5, 6, 7])
+    ldmia   sp!, {r0, r1, r2, r3, r4, r5, r6, r7}
+
+    // r0 = r0 + 0x00808000;
+    add     r0, r0, #0x00800000
+    add     r0, r0, #0x00008000
+
+    // Step 1: Analog to the first pass.
+
+    // r0 = r0 + r4;
+    // r6 = r6 + r2;
+    add     r0, r0, r4
+    add     r6, r6, r2
+
+    // r4 = r0 - r4 * 2;
+    // r2 = r2 * 2 - r6 = -(r6 - r2 * 2);
+    sub     r4, r0, r4, lsl #1
+    rsb     r2, r6, r2, lsl #1
+
+    // r1 = r1 + r7;
+    // r3 = r3 + r5;
+    add     r1, r1, r7
+    add     r3, r3, r5
+
+    // Step 2: Rotations and Butterflies.
+
+    // r7 = r1 - r7 * 2;
+    // r1 = r1 - r3;
+    // r5 = r5 * 2 - r3 = -(r3 - r5 * 2);
+    // r3 = r1 + r3 * 2;
+    // r8 = r5 + r7;
+    sub     r7, r1, r7, lsl #1
+    sub     r1, r1, r3
+    rsb     r5, r3, r5, lsl #1
+    add     r3, r1, r3, lsl #1
+    add     r8, r5, r7
+
+    // r2 = r2 * 1.41421 = r2 * 27146 / 65536 + r2;
+    // r8 = r8 * 1.84776 / 8 = r8 * 15137 / 65536;
+    // r1 = r1 * 1.41421 = r1 * 27146 / 65536 + r1;
+    smlawt  r2, r2, r10, r2
+    smulwb  r8, r8, r10
+    smlawt  r1, r1, r10, r1
+
+    // r0 = r0 + r6;
+    // r2 = r2 - r6;
+    // r6 = r0 - r6 * 2;
+    add     r0, r0, r6
+    sub     r2, r2, r6
+    sub     r6, r0, r6, lsl #1
+
+    // r5 = r5 * -2.61313 / 8 + r8 = r5 * -21407 / 65536 + r8;
+    // r8 = r7 * -1.08239 / 8 + r8 = r7 * -8867 / 65536 + r8;
+    smlawt  r5, r5, r11, r8
+    smlawb  r8, r7, r11, r8
+
+    // r4 = r4 + r2;
+    // r0 = r0 + r3;
+    // r2 = r4 - r2 * 2;
+    add     r4, r4, r2
+    add     r0, r0, r3
+    sub     r2, r4, r2, lsl #1
+
+    // r7 = r5 * 8 - r3 = -(r3 - r5 * 8);
+    // r3 = r0 - r3 * 2;
+    // r1 = r1 - r7;
+    // r4 = r4 + r7;
+    // r5 = r8 * 8 - r1 = -(r1 - r8 * 8);
+    // r7 = r4 - r7 * 2;
+    rsb     r7, r3, r5, lsl #3
+    sub     r3, r0, r3, lsl #1
+    sub     r1, r1, r7
+    add     r4, r4, r7
+    rsb     r5, r1, r8, lsl #3
+    sub     r7, r4, r7, lsl #1
+
+    // r2 = r2 + r1;
+    // r6 = r6 + r5;
+    // r1 = r2 - r1 * 2;
+    // r5 = r6 - r5 * 2;
+    add     r2, r2, r1
+    add     r6, r6, r5
+    sub     r1, r2, r1, lsl #1
+    sub     r5, r6, r5, lsl #1
+
+    // Step 3: Reorder and Save.
+
+    // Load output pointer.
+    ldr     r8, [r12], #4
+
+    // For little endian: r6, r2, r4, r0, r3, r7, r1, r5.
+    pkhtb   r6, r6, r4, asr #16
+    pkhtb   r2, r2, r0, asr #16
+    pkhtb   r3, r3, r1, asr #16
+    pkhtb   r7, r7, r5, asr #16
+    usat16  r6, #8, r6
+    usat16  r2, #8, r2
+    usat16  r3, #8, r3
+    usat16  r7, #8, r7
+    orr     r0, r2, r6, lsl #8
+    orr     r1, r7, r3, lsl #8
+
+#ifdef __ARMEB__
+    // Reverse bytes for big endian.
+    rev     r0, r0
+    rev     r1, r1
+#endif
+
+    // Use STR instead of STRD to support unaligned access.
+    str     r0, [r8, r14] !
+    str     r1, [r8, #4]
+
+pass2_tail:
+    adds    r9, r9, #0x10000000
+    bpl     pass2_head
+
+    ldr     sp, [sp, #8]
+    add     sp, sp, #236
+
+    ldmia   sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r12, r14}
+    bx      lr
+    .endfunc
+
+#endif
diff --git a/libjpegtwrp/asm/armv7/jdcolor-android-armv7.S b/libjpegtwrp/asm/armv7/jdcolor-android-armv7.S
new file mode 100644
index 000000000..95bd4bf78
--- /dev/null
+++ b/libjpegtwrp/asm/armv7/jdcolor-android-armv7.S
@@ -0,0 +1,1223 @@
+/*------------------------------------------------------------------------
+* jdcolor-android-armv7.S
+*
+*  Copyright (c) 2010, Code Aurora Forum. All rights reserved.
+*
+*  Redistribution and use in source and binary forms, with or without
+*  modification, are permitted provided that the following conditions are
+*  met:
+*      * Redistributions of source code must retain the above copyright
+*        notice, this list of conditions and the following disclaimer.
+*      * Redistributions in binary form must reproduce the above
+*        copyright notice, this list of conditions and the following
+*        disclaimer in the documentation and/or other materials provided
+*        with the distribution.
+*      * Neither the name of Code Aurora Forum, Inc. nor the names of its
+*        contributors may be used to endorse or promote products derived
+*        from this software without specific prior written permission.
+*
+*  THIS SOFTWARE IS PROVIDED "AS IS" AND ANY EXPRESS OR IMPLIED
+*  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+*  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT
+*  ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS
+*  BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+*  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+*  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
+*  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
+*  WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
+*  OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN
+*  IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*--------------------------------------------------------------------------
+
+*--------------------------------------------------------------------------
+*                         FUNCTION LIST
+*--------------------------------------------------------------------------
+*
+* - yvup2rgb565_venum
+* - yyvup2rgb565_venum
+* - yvup2abgr8888_venum
+* - yyvup2abgr8888_venum
+*
+*--------------------------------------------------------------------------
+*/
+
+    .section yvu_plain_to_rgb_android, "x"  @ AREA
+    .text                              @ |.text|, CODE, READONLY
+    .align 2
+    .code  32                          @ CODE32
+
+/*-----------------------------------------------------------------------------
+ *   ARM Registers
+ * ---------------------------------------------------------------------------- */
+p_y       .req r0
+p_cr      .req r1
+p_cb      .req r2
+p_rgb     .req r3
+p_bgr     .req r3
+length    .req r12
+
+    .global yvup2rgb565_venum
+    .global yyvup2rgb565_venum
+    .global yvup2abgr8888_venum
+    .global yyvup2abgr8888_venum
+
+@ coefficients in color conversion matrix multiplication
+.equ COEFF_Y,          256             @ contribution of Y
+.equ COEFF_V_RED,      359             @ contribution of V for red
+.equ COEFF_U_GREEN,    -88             @ contribution of U for green
+.equ COEFF_V_GREEN,   -183             @ contribution of V for green
+.equ COEFF_U_BLUE,     454             @ contribution of U for blue
+
+@ Clamping constants 0x0 and 0xFF
+.equ COEFF_0,          0
+.equ COEFF_255,        255
+
+@ Bias coefficients for red, green and blue
+.equ COEFF_BIAS_R,   -45824            @ Red   bias =     -359*128 + 128
+.equ COEFF_BIAS_G,    34816            @ Green bias = (88+183)*128 + 128
+.equ COEFF_BIAS_B,   -57984            @ Blue  bias =     -454*128 + 128
+
+
+/*--------------------------------------------------------------------------
+* FUNCTION     : yvup2rgb565_venum
+*--------------------------------------------------------------------------
+* DESCRIPTION  : Perform YVU planar to RGB565 conversion.
+*--------------------------------------------------------------------------
+* C PROTOTYPE  : void yvup2rgb565_venum(uint8_t  *p_y,
+*                                       uint8_t  *p_cr,
+*                                       uint8_t  *p_cb,
+*                                       uint8_t  *p_rgb565,
+*                                       uint32_t  length)
+*--------------------------------------------------------------------------
+* REG INPUT    : R0: uint8_t  *p_y
+*                      pointer to the input Y Line
+*                R1: uint8_t  *p_cr
+*                      pointer to the input Cr Line
+*                R2: uint8_t  *p_cb
+*                      pointer to the input Cb Line
+*                R3: uint8_t  *p_rgb565
+*                      pointer to the output RGB Line
+*                R12: uint32_t  length
+*                      width of Line
+*--------------------------------------------------------------------------
+* STACK ARG    : None
+*--------------------------------------------------------------------------
+* REG OUTPUT   : None
+*--------------------------------------------------------------------------
+* MEM INPUT    : p_y      - a line of Y pixels
+*                p_cr     - a line of Cr pixels
+*                p_cb     - a line of Cb pixels
+*                length   - the width of the input line
+*--------------------------------------------------------------------------
+* MEM OUTPUT   : p_rgb565 - the converted rgb pixels
+*--------------------------------------------------------------------------
+* REG AFFECTED : ARM:  R0-R4, R12
+*                NEON: Q0-Q15
+*--------------------------------------------------------------------------
+* STACK USAGE  : none
+*--------------------------------------------------------------------------
+* CYCLES       : none
+*
+*--------------------------------------------------------------------------
+* NOTES        :
+*--------------------------------------------------------------------------
+*/
+.type yvup2rgb565_venum, %function
+yvup2rgb565_venum:
+    /*-------------------------------------------------------------------------
+     *  Store stack registers
+     * ------------------------------------------------------------------------ */
+    STMFD SP!, {LR}
+
+    VPUSH {D8-D15}
+
+    PLD [R0, R3]                       @ preload luma line
+
+    ADR   R12, constants
+
+    VLD1.S16  {D6, D7}, [R12]!         @ D6, D7: 359 |  -88 | -183 | 454 | 256 | 0 | 255 | 0
+    VLD1.S32  {D30, D31}, [R12]        @ Q15   :  -45824    |    34816   |  -57984 |     X
+
+    /*-------------------------------------------------------------------------
+     *  Load the 5th parameter via stack
+     *  R0 ~ R3 are used to pass the first 4 parameters, the 5th and above
+     *  parameters are passed via stack
+     * ------------------------------------------------------------------------ */
+    LDR R12, [SP, #68]                 @ LR is pushed into the stack so SP is
+                                       @ decreased by 4,
+                                       @ D8-D15 are also pushed into the stack
+                                       @ so SP is decreased by
+                                       @ 8-byte/D-Register * 8 D-Registers = 64,
+                                       @ so SP needs to be increased by 64+4=68
+                                       @ to get the value that was first pushed
+                                       @ into stack (the 5th parameter passed in
+                                       @ throught stack)
+
+    /*-------------------------------------------------------------------------
+     *  Load clamping parameters to duplicate vector elements
+     * ------------------------------------------------------------------------ */
+    VDUP.S16  Q4,  D7[1]               @ Q4:  0  |  0  |  0  |  0  |  0  |  0  |  0  |  0
+    VDUP.S16  Q5,  D7[2]               @ Q5: 255 | 255 | 255 | 255 | 255 | 255 | 255 | 255
+
+    /*-------------------------------------------------------------------------
+     *  Read bias
+     * ------------------------------------------------------------------------ */
+    VDUP.S32  Q0,   D30[0]             @ Q0:  -45824 | -45824 | -45824 | -45824
+    VDUP.S32  Q1,   D30[1]             @ Q1:   34816 |  34816 |  34816 |  34816
+    VDUP.S32  Q2,   D31[0]             @ Q2:  -70688 | -70688 | -70688 | -70688
+
+
+    /*-------------------------------------------------------------------------
+     *  The main loop
+     * ------------------------------------------------------------------------ */
+loop_yvup2rgb565:
+
+    /*-------------------------------------------------------------------------
+     *  Load input from Y, V and U
+     *  D12  : Y0  Y1  Y2  Y3  Y4  Y5  Y6  Y7
+     *  D14  : V0  V1  V2  V3  V4  V5  V6  V7
+     *  D15  : U0  U1  U2  U3  U4  U5  U6  U7
+     * ------------------------------------------------------------------------ */
+    VLD1.U8  {D12},  [p_y]!            @ Load 8 Y  elements (uint8) to D12
+    VLD1.U8  {D14},  [p_cr]!           @ Load 8 Cr elements (uint8) to D14
+    VLD1.U8  {D15},  [p_cb]!           @ Load 8 Cb elements (uint8) to D15
+
+    /*-------------------------------------------------------------------------
+     *  Expand uint8 value to uint16
+     *  D18, D19: Y0 Y1 Y2 Y3 Y4 Y5 Y6 Y7
+     *  D20, D21: V0 V1 V2 V3 V4 V5 V6 V7
+     *  D22, D23: U0 U1 U2 U3 U4 U5 U6 U7
+     * ------------------------------------------------------------------------ */
+    VMOVL.U8 Q9,  D12
+    VMOVL.U8 Q10, D14
+    VMOVL.U8 Q11, D15
+
+    /*-------------------------------------------------------------------------
+     *  Multiply contribution from chrominance, results are in 32-bit
+     * ------------------------------------------------------------------------ */
+    VMULL.S16  Q12, D20, D6[0]         @ Q12:  359*(V0,V1,V2,V3)     Red
+    VMULL.S16  Q13, D22, D6[1]         @ Q13:  -88*(U0,U1,U2,U3)     Green
+    VMLAL.S16  Q13, D20, D6[2]         @ Q13:  -88*(U0,U1,U2,U3) - 183*(V0,V1,V2,V3)
+    VMULL.S16  Q14, D22, D6[3]         @ Q14:  454*(U0,U1,U2,U3)     Blue
+
+    /*-------------------------------------------------------------------------
+     *  Add bias
+     * ------------------------------------------------------------------------ */
+    VADD.S32  Q12, Q0                  @ Q12 add Red   bias -45824
+    VADD.S32  Q13, Q1                  @ Q13 add Green bias  34816
+    VADD.S32  Q14, Q2                  @ Q14 add Blue  bias -57984
+
+    /*-------------------------------------------------------------------------
+     *  Calculate Red, Green, Blue
+     * ------------------------------------------------------------------------ */
+    VMLAL.S16  Q12, D18, D7[0]         @ Q12: R0, R1, R2, R3 in 32-bit Q8 format
+    VMLAL.S16  Q13, D18, D7[0]         @ Q13: G0, G1, G2, G3 in 32-bit Q8 format
+    VMLAL.S16  Q14, D18, D7[0]         @ Q14: B0, B1, B2, B3 in 32-bit Q8 format
+
+    /*-------------------------------------------------------------------------
+     *  Right shift eight bits with rounding
+     * ------------------------------------------------------------------------ */
+    VSHRN.S32   D18 , Q12, #8          @ D18: R0, R1, R2, R3 in 16-bit Q0 format
+    VSHRN.S32   D20 , Q13, #8          @ D20: G0, G1, G2, G3 in 16-bit Q0 format
+    VSHRN.S32   D22,  Q14, #8          @ D22: B0, B1, B2, B3 in 16-bit Q0 format
+
+    /*-------------------------------------------------------------------------
+     *  Done with the first 4 elements, continue on the next 4 elements
+     * ------------------------------------------------------------------------ */
+
+    /*-------------------------------------------------------------------------
+     *  Multiply contribution from chrominance, results are in 32-bit
+     * ------------------------------------------------------------------------ */
+    VMULL.S16  Q12, D21, D6[0]         @ Q12:  359*(V0,V1,V2,V3)     Red
+    VMULL.S16  Q13, D23, D6[1]         @ Q13:  -88*(U0,U1,U2,U3)     Green
+    VMLAL.S16  Q13, D21, D6[2]         @ Q13:  -88*(U0,U1,U2,U3) - 183*(V0,V1,V2,V3)
+    VMULL.S16  Q14, D23, D6[3]         @ Q14:  454*(U0,U1,U2,U3)     Blue
+
+    /*-------------------------------------------------------------------------
+     *  Add bias
+     * ------------------------------------------------------------------------ */
+    VADD.S32  Q12, Q0                  @ Q12 add Red   bias -45824
+    VADD.S32  Q13, Q1                  @ Q13 add Green bias  34816
+    VADD.S32  Q14, Q2                  @ Q14 add Blue  bias -57984
+
+    /*-------------------------------------------------------------------------
+     *  Calculate Red, Green, Blue
+     * ------------------------------------------------------------------------ */
+    VMLAL.S16  Q12, D19, D7[0]         @ Q12: R0, R1, R2, R3 in 32-bit Q8 format
+    VMLAL.S16  Q13, D19, D7[0]         @ Q13: G0, G1, G2, G3 in 32-bit Q8 format
+    VMLAL.S16  Q14, D19, D7[0]         @ Q14: B0, B1, B2, B3 in 32-bit Q8 format
+
+    /*-------------------------------------------------------------------------
+     *  Right shift eight bits with rounding
+     * ------------------------------------------------------------------------ */
+    VSHRN.S32   D19 , Q12, #8          @ D18: R0, R1, R2, R3 in 16-bit Q0 format
+    VSHRN.S32   D21 , Q13, #8          @ D20: G0, G1, G2, G3 in 16-bit Q0 format
+    VSHRN.S32   D23,  Q14, #8          @ D22: B0, B1, B2, B3 in 16-bit Q0 format
+
+    /*-------------------------------------------------------------------------
+     *  Clamp the value to be within [0~255]
+     * ------------------------------------------------------------------------ */
+    VMAX.S16  Q9, Q9, Q4               @ if Q9 <   0, Q9 =   0
+    VMIN.S16  Q9, Q9, Q5               @ if Q9 > 255, Q9 = 255
+    VQMOVUN.S16  D28, Q9               @ store Red to D28, narrow the value from int16 to int8
+
+    VMAX.S16  Q10, Q10, Q4             @ if Q10 <   0, Q10 =   0
+    VMIN.S16  Q10, Q10, Q5             @ if Q10 > 255, Q10 = 255
+    VQMOVUN.S16   D27, Q10             @ store Green to D27, narrow the value from int16 to int8
+
+    VMAX.S16  Q11, Q11, Q4             @ if Q11 <   0, Q11 =   0
+    VMIN.S16  Q11, Q11, Q5             @ if Q11 > 255, Q11 = 255
+    VQMOVUN.S16   D26, Q11             @ store Blue to D26, narrow the value from int16 to int8.
+
+    /*-------------------------------------------------------------------------
+     *  D27:  3 bits of Green + 5 bits of Blue
+     *  D28:  5 bits of Red   + 3 bits of Green
+     * ------------------------------------------------------------------------ */
+    VSRI.8   D28, D27, #5              @ right shift G by 5 and insert to R
+    VSHL.U8  D27, D27, #3              @ left  shift G by 3
+    VSRI.8   D27, D26, #3              @ right shift B by 3 and insert to G
+
+    SUBS length, length, #8            @ check if the length is less than 8
+
+    BMI  trailing_yvup2rgb565          @ jump to trailing processing if remaining length is less than 8
+
+    VST2.U8  {D27, D28}, [p_rgb]!      @ vector store Red, Green, Blue to destination
+                                       @ Blue at LSB
+
+    BHI loop_yvup2rgb565               @ loop if more than 8 pixels left
+
+    BEQ  end_yvup2rgb565               @ done if exactly 8 pixel processed in the loop
+
+
+trailing_yvup2rgb565:
+    /*-------------------------------------------------------------------------
+     *  There are from 1 ~ 7 pixels left in the trailing part.
+     *  First adding 7 to the length so the length would be from 0 ~ 6.
+     *  eg: 1 pixel left in the trailing part, so 1-8+7 = 0.
+     *  Then save 1 pixel unconditionally since at least 1 pixels left in the
+     *  trailing part.
+     * ------------------------------------------------------------------------ */
+    ADDS length, length, #7            @ there are 7 or less in the trailing part
+
+    VST2.U8 {D27[0], D28[0]}, [p_rgb]! @ at least 1 pixel left in the trailing part
+    BEQ  end_yvup2rgb565               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST2.U8 {D27[1], D28[1]}, [p_rgb]! @ store one more pixel
+    BEQ  end_yvup2rgb565               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST2.U8 {D27[2], D28[2]}, [p_rgb]! @ store one more pixel
+    BEQ  end_yvup2rgb565               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST2.U8 {D27[3], D28[3]}, [p_rgb]! @ store one more pixel
+    BEQ  end_yvup2rgb565               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST2.U8 {D27[4], D28[4]}, [p_rgb]! @ store one more pixel
+    BEQ  end_yvup2rgb565               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST2.U8 {D27[5], D28[5]}, [p_rgb]! @ store one more pixel
+    BEQ  end_yvup2rgb565               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST2.U8 {D27[6], D28[6]}, [p_rgb]! @ store one more pixel
+
+end_yvup2rgb565:
+    VPOP  {D8-D15}
+    LDMFD SP!, {PC}
+
+                                       @ end of yvup2rgb565
+
+
+/*--------------------------------------------------------------------------
+* FUNCTION     : yyvup2rgb565_venum
+*--------------------------------------------------------------------------
+* DESCRIPTION  : Perform YYVU planar to RGB565 conversion.
+*--------------------------------------------------------------------------
+* C PROTOTYPE  : void yyvup2rgb565_venum(uint8_t  *p_y,
+*                                        uint8_t  *p_cr,
+*                                        uint8_t  *p_cb,
+*                                        uint8_t  *p_rgb565,
+*                                        uint32_t  length)
+*--------------------------------------------------------------------------
+* REG INPUT    : R0: uint8_t  *p_y
+*                      pointer to the input Y Line
+*                R1: uint8_t  *p_cr
+*                      pointer to the input Cr Line
+*                R2: uint8_t  *p_cb
+*                      pointer to the input Cb Line
+*                R3: uint8_t  *p_rgb565
+*                      pointer to the output RGB Line
+*                R12: uint32_t  length
+*                      width of Line
+*--------------------------------------------------------------------------
+* STACK ARG    : None
+*--------------------------------------------------------------------------
+* REG OUTPUT   : None
+*--------------------------------------------------------------------------
+* MEM INPUT    : p_y      - a line of Y pixels
+*                p_cr     - a line of Cr pixels
+*                p_cb     - a line of Cb pixels
+*                length   - the width of the input line
+*--------------------------------------------------------------------------
+* MEM OUTPUT   : p_rgb565 - the converted rgb pixels
+*--------------------------------------------------------------------------
+* REG AFFECTED : ARM:  R0-R4, R12
+*                NEON: Q0-Q15
+*--------------------------------------------------------------------------
+* STACK USAGE  : none
+*--------------------------------------------------------------------------
+* CYCLES       : none
+*
+*--------------------------------------------------------------------------
+* NOTES        :
+*--------------------------------------------------------------------------
+*/
+.type yyvup2rgb565_venum, %function
+yyvup2rgb565_venum:
+    /*-------------------------------------------------------------------------
+     *  Store stack registers
+     * ------------------------------------------------------------------------ */
+    STMFD SP!, {LR}
+
+    VPUSH {D8-D15}
+
+    PLD [R0, R3]                       @ preload luma line
+
+    ADR   R12, constants
+
+    VLD1.S16  {D6, D7}, [R12]!         @ D6, D7: 359 |  -88 | -183 | 454 | 256 | 0 | 255 | 0
+    VLD1.S32  {D30, D31}, [R12]        @ Q15   :  -45824    |    34816   |  -57984 |     X
+
+    /*-------------------------------------------------------------------------
+     *  Load the 5th parameter via stack
+     *  R0 ~ R3 are used to pass the first 4 parameters, the 5th and above
+     *  parameters are passed via stack
+     * ------------------------------------------------------------------------ */
+    LDR R12, [SP, #68]                 @ LR is pushed into the stack so SP is
+                                       @ decreased by 4,
+                                       @ D8-D15 are also pushed into the stack
+                                       @ so SP is decreased by
+                                       @ 8-byte/D-Register * 8 D-Registers = 64,
+                                       @ so SP needs to be increased by 64+4=68
+                                       @ to get the value that was first pushed
+                                       @ into stack (the 5th parameter passed in
+                                       @ throught stack)
+
+    /*-------------------------------------------------------------------------
+     *  Load clamping parameters to duplicate vector elements
+     * ------------------------------------------------------------------------ */
+    VDUP.S16  Q4,  D7[1]               @ Q4:  0  |  0  |  0  |  0  |  0  |  0  |  0  |  0
+    VDUP.S16  Q5,  D7[2]               @ Q5: 255 | 255 | 255 | 255 | 255 | 255 | 255 | 255
+
+    /*-------------------------------------------------------------------------
+     *  Read bias
+     * ------------------------------------------------------------------------ */
+    VDUP.S32  Q0,   D30[0]             @ Q0:  -45824 | -45824 | -45824 | -45824
+    VDUP.S32  Q1,   D30[1]             @ Q1:   34816 |  34816 |  34816 |  34816
+    VDUP.S32  Q2,   D31[0]             @ Q2:  -70688 | -70688 | -70688 | -70688
+
+
+    /*-------------------------------------------------------------------------
+     *  The main loop
+     * ------------------------------------------------------------------------ */
+loop_yyvup2rgb565:
+
+    /*-------------------------------------------------------------------------
+     *  Load input from Y, V and U
+     *  D12, D13: Y0 Y2 Y4 Y6 Y8 Y10 Y12 Y14, Y1 Y3 Y5 Y7 Y9 Y11 Y13 Y15
+     *  D14     : V0 V1 V2 V3 V4 V5  V6  V7
+     *  D15     : U0 U1 U2 U3 U4 U5  U6  U7
+     * ------------------------------------------------------------------------ */
+    VLD2.U8  {D12,D13}, [p_y]!         @ Load 16 Luma elements (uint8) to D12, D13
+    VLD1.U8  {D14},     [p_cr]!        @ Load 8 Cr elements (uint8) to D14
+    VLD1.U8  {D15},     [p_cb]!        @ Load 8 Cb elements (uint8) to D15
+
+    /*-------------------------------------------------------------------------
+     *  Expand uint8 value to uint16
+     *  D24, D25: Y0 Y2 Y4 Y6 Y8 Y10 Y12 Y14
+     *  D26, D27: Y1 Y3 Y5 Y7 Y9 Y11 Y13 Y15
+     *  D28, D29: V0 V1 V2 V3 V4 V5  V6  V7
+     *  D30, D31: U0 U1 U2 U3 U4 U5  U6  U7
+     * ------------------------------------------------------------------------ */
+    VMOVL.U8 Q12, D12
+    VMOVL.U8 Q13, D13
+    VMOVL.U8 Q14, D14
+    VMOVL.U8 Q15, D15
+
+    /*-------------------------------------------------------------------------
+     *  Multiply contribution from chrominance, results are in 32-bit
+     * ------------------------------------------------------------------------ */
+    VMULL.S16  Q6, D28, D6[0]          @ Q6:  359*(V0,V1,V2,V3)     Red
+    VMULL.S16  Q7, D30, D6[1]          @ Q7:  -88*(U0,U1,U2,U3)     Green
+    VMLAL.S16  Q7, D28, D6[2]          @ q7:  -88*(U0,U1,U2,U3) - 183*(V0,V1,V2,V3)
+    VMULL.S16  Q8, D30, D6[3]          @ q8:  454*(U0,U1,U2,U3)     Blue
+
+    /*-------------------------------------------------------------------------
+     *  Add bias
+     * ------------------------------------------------------------------------ */
+    VADD.S32  Q6, Q0                   @ Q6 add Red   bias -45824
+    VADD.S32  Q7, Q1                   @ Q7 add Green bias  34816
+    VADD.S32  Q8, Q2                   @ Q8 add Blue  bias -57984
+
+    /*-------------------------------------------------------------------------
+     *  Calculate Red, Green, Blue
+     * ------------------------------------------------------------------------ */
+    VMOV.S32   Q9, Q6
+    VMLAL.S16  Q6, D24, D7[0]          @ Q6: R0, R2, R4, R6 in 32-bit Q8 format
+    VMLAL.S16  Q9, D26, D7[0]          @ Q9: R1, R3, R5, R7 in 32-bit Q8 format
+
+    VMOV.S32   Q10, Q7
+    VMLAL.S16  Q7,  D24, D7[0]         @ Q7:  G0, G2, G4, G6 in 32-bit Q8 format
+    VMLAL.S16  Q10, D26, D7[0]         @ Q10: G1, G3, G5, G7 in 32-bit Q8 format
+
+    VMOV.S32   Q11, Q8
+    VMLAL.S16  Q8,  D24, D7[0]         @ Q8:  B0, B2, B4, B6 in 32-bit Q8 format
+    VMLAL.S16  Q11, D26, D7[0]         @ Q11: B1, B3, B5, B7 in 32-bit Q8 format
+
+    /*-------------------------------------------------------------------------
+     *  Right shift eight bits with rounding
+     * ------------------------------------------------------------------------ */
+    VSHRN.S32   D12, Q6,  #8           @ D12: R0 R2 R4 R6 in 16-bit Q0 format
+    VSHRN.S32   D13, Q9,  #8           @ D13: R1 R3 R5 R7 in 16-bit Q0 format
+    VZIP.16     D12, D13               @ Q6 : R0 R1 R2 R3 R4 R5 R6 R7
+
+    VSHRN.S32   D18, Q7,  #8           @ D18: G0 G2 G4 G6 in 16-bit Q0 format
+    VSHRN.S32   D19, Q10, #8           @ D19: G1 G3 G5 G7 in 16-bit Q0 format
+    VZIP.16     D18, D19               @ Q9 : G0 G1 G2 G3 G4 G5 G6 G7
+
+    VSHRN.S32   D20, Q8,  #8           @ D20: B0 B2 B4 B6 in 16-bit Q0 format
+    VSHRN.S32   D21, Q11, #8           @ D21: B1 B3 B5 B7 in 16-bit Q0 format
+    VZIP.16     D20, D21               @ Q10: B0 B1 B2 B3 B4 B5 B6 B7
+
+    /*-------------------------------------------------------------------------
+     *  Clamp the value to be within [0~255]
+     * ------------------------------------------------------------------------ */
+    VMAX.S16  Q6, Q6, Q4               @ if Q6 <   0, Q6 =   0
+    VMIN.S16  Q6, Q6, Q5               @ if Q6 > 255, Q6 = 255
+    VQMOVUN.S16  D23, Q6               @ store Red to D23, narrow the value from int16 to int8
+
+    VMAX.S16  Q9, Q9, Q4               @ if Q9 <   0, Q9 =   0
+    VMIN.S16  Q9, Q9, Q5               @ if Q9 > 255, Q9 = 255
+    VQMOVUN.S16  D22, Q9               @ store Green to D22, narrow the value from int16 to int8
+
+    VMAX.S16  Q10, Q10, Q4             @ if Q10 <   0, Q10 =   0
+    VMIN.S16  Q10, Q10, Q5             @ if Q10 > 255, Q10 = 255
+    VQMOVUN.S16   D21, Q10             @ store Blue to D21, narrow the value from int16 to int8
+
+    /*-------------------------------------------------------------------------
+     *  D22:  3 bits of Green + 5 bits of Blue
+     *  D23:  5 bits of Red   + 3 bits of Green
+     * ------------------------------------------------------------------------ */
+    VSRI.8   D23, D22, #5              @ right shift G by 5 and insert to R
+    VSHL.U8  D22, D22, #3              @ left shift G by 3
+    VSRI.8   D22, D21, #3              @ right shift B by 3 and insert to G
+
+    SUBS length, length, #8            @ check if the length is less than 8
+
+    BMI  trailing_yyvup2rgb565         @ jump to trailing processing if remaining length is less than 8
+
+    VST2.U8  {D22,D23}, [p_rgb]!       @ vector store Red, Green, Blue to destination
+                                       @ Blue at LSB
+
+    BEQ  end_yyvup2rgb565              @ done if exactly 8 pixel processed in the loop
+
+
+    /*-------------------------------------------------------------------------
+     *  Done with the first 8 elements, continue on the next 8 elements
+     * ------------------------------------------------------------------------ */
+
+    /*-------------------------------------------------------------------------
+     *  Multiply contribution from chrominance, results are in 32-bit
+     * ------------------------------------------------------------------------ */
+    VMULL.S16  Q6, D29, D6[0]          @ Q6: 359*(V4,V5,V6,V7)       Red
+    VMULL.S16  Q7, D31, D6[1]          @ Q7: -88*(U4,U5,U6,U7)      Green
+    VMLAL.S16  Q7, D29, D6[2]          @ Q7: -88*(U4,U5,U6,U7) - 183*(V4,V5,V6,V7)
+    VMULL.S16  Q8, D31, D6[3]          @ Q8: 454*(U4,U5,U6,U7)       Blue
+
+    /*-------------------------------------------------------------------------
+     *  Add bias
+     * ------------------------------------------------------------------------ */
+    VADD.S32  Q6, Q0                   @ Q6 add Red   bias -45824
+    VADD.S32  Q7, Q1                   @ Q7 add Green bias  34816
+    VADD.S32  Q8, Q2                   @ Q8 add Blue  bias -57984
+
+    /*-------------------------------------------------------------------------
+     *  Calculate Red, Green, Blue
+     * ------------------------------------------------------------------------ */
+    VMOV.S32   Q9, Q6
+    VMLAL.S16  Q6, D25, D7[0]          @ Q6: R8 R10 R12 R14 in 32-bit Q8 format
+    VMLAL.S16  Q9, D27, D7[0]          @ Q9: R9 R11 R13 R15 in 32-bit Q8 format
+
+    VMOV.S32   Q10, Q7
+    VMLAL.S16  Q7,  D25, D7[0]         @ Q7: G0, G2, G4, G6 in 32-bit Q8 format
+    VMLAL.S16  Q10, D27, D7[0]         @ Q10 : G1, G3, G5, G7 in 32-bit Q8 format
+
+    VMOV.S32   Q11, Q8
+    VMLAL.S16  Q8,  D25, D7[0]         @ Q8: B0, B2, B4, B6 in 32-bit Q8 format
+    VMLAL.S16  Q11, D27, D7[0]         @ Q11 : B1, B3, B5, B7 in 32-bit Q8 format
+
+    /*-------------------------------------------------------------------------
+     *  Right shift eight bits with rounding
+     * ------------------------------------------------------------------------ */
+    VSHRN.S32   D12, Q6,  #8           @ D12: R8 R10 R12 R14 in 16-bit Q0 format
+    VSHRN.S32   D13, Q9,  #8           @ D13: R9 R11 R13 R15 in 16-bit Q0 format
+    VZIP.16     D12, D13               @ Q6: R8 R9 R10 R11 R12 R13 R14 R15
+
+    VSHRN.S32   D18, Q7,  #8           @ D18: G8 G10 G12 G14 in 16-bit Q0 format
+    VSHRN.S32   D19, Q10, #8           @ D19: G9 G11 G13 G15 in 16-bit Q0 format
+    VZIP.16     D18, D19               @ Q9:  G8 G9 G10 G11 G12 G13 G14 G15
+
+    VSHRN.S32   D20, Q8,  #8           @ D20: B8 B10 B12 B14 in 16-bit Q0 format
+    VSHRN.S32   D21, Q11, #8           @ D21: B9 B11 B13 B15 in 16-bit Q0 format
+    VZIP.16     D20, D21               @ Q10: B8 B9 B10 B11 B12 B13 B14 B15
+
+    /*-------------------------------------------------------------------------
+     *  Clamp the value to be within [0~255]
+     * ------------------------------------------------------------------------ */
+    VMAX.S16  Q6, Q6, Q4               @ if Q6 <   0, Q6 =   0
+    VMIN.S16  Q6, Q6, Q5               @ if Q6 > 255, Q6 = 255
+    VQMOVUN.S16  D23, Q6               @ store Red to D23, narrow the value from int16 to int8
+
+    VMAX.S16  Q9, Q9, Q4               @ if Q9 <   0, Q9 =   0
+    VMIN.S16  Q9, Q9, Q5               @ if Q9 > 255, Q9 = 255
+    VQMOVUN.S16  D22, Q9               @ store Green to D22, narrow the value from int16 to int8
+
+    VMAX.S16  Q10, Q10, Q4             @ if Q10 <   0, Q10 =   0
+    VMIN.S16  Q10, Q10, Q5             @ if Q10 > 255, Q10 = 255
+    VQMOVUN.S16   D21, Q10             @ store Blue to D21, narrow the value from int16 to int8
+
+    /*-------------------------------------------------------------------------
+     *  D22:  3 bits of Green + 5 bits of Blue
+     *  D23:  5 bits of Red   + 3 bits of Green
+     * ------------------------------------------------------------------------ */
+    VSRI.8   D23, D22, #5              @ right shift G by 5 and insert to R
+    VSHL.U8  D22, D22, #3              @ left shift G by 3
+    VSRI.8   D22, D21, #3              @ right shift B by 3 and insert to G
+
+    SUBS length, length, #8            @ check if the length is less than 8
+
+    BMI  trailing_yyvup2rgb565         @ jump to trailing processing if remaining length is less than 8
+
+    VST2.U8  {D22,D23}, [p_rgb]!       @ vector store Red, Green, Blue to destination
+                                       @ Blue at LSB
+
+    BHI loop_yyvup2rgb565              @ loop if more than 8 pixels left
+
+    BEQ  end_yyvup2rgb565              @ done if exactly 8 pixel processed in the loop
+
+
+trailing_yyvup2rgb565:
+    /*-------------------------------------------------------------------------
+     *  There are from 1 ~ 7 pixels left in the trailing part.
+     *  First adding 7 to the length so the length would be from 0 ~ 6.
+     *  eg: 1 pixel left in the trailing part, so 1-8+7 = 0.
+     *  Then save 1 pixel unconditionally since at least 1 pixels left in the
+     *  trailing part.
+     * ------------------------------------------------------------------------ */
+    ADDS length, length, #7            @ there are 7 or less in the trailing part
+
+    VST2.U8 {D22[0],D23[0]}, [p_rgb]!  @ at least 1 pixel left in the trailing part
+    BEQ end_yyvup2rgb565               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST2.U8 {D22[1],D23[1]}, [p_rgb]!  @ store one more pixel
+    BEQ end_yyvup2rgb565               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST2.U8 {D22[2],D23[2]}, [p_rgb]!  @ store one more pixel
+    BEQ end_yyvup2rgb565               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST2.U8 {D22[3],D23[3]}, [p_rgb]!  @ store one more pixel
+    BEQ end_yyvup2rgb565               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST2.U8 {D22[4],D23[4]}, [p_rgb]!  @ store one more pixel
+    BEQ end_yyvup2rgb565               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST2.U8 {D22[5],D23[5]}, [p_rgb]!  @ store one more pixel
+    BEQ end_yyvup2rgb565               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST2.U8 {D22[6],D23[6]}, [p_rgb]!  @ store one more pixel
+
+end_yyvup2rgb565:
+    VPOP  {D8-D15}
+    LDMFD SP!, {PC}
+
+                                       @ end of yyvup2rgb565
+
+constants:
+    .hword (COEFF_V_RED),  (COEFF_U_GREEN), (COEFF_V_GREEN), (COEFF_U_BLUE) @   359  | -88   |  -183  | 454
+    .hword (COEFF_Y),      (COEFF_0),       (COEFF_255)    , (COEFF_0)      @   256  |   0   |   255  |  0
+    .word  (COEFF_BIAS_R), (COEFF_BIAS_G),  (COEFF_BIAS_B)                  @ -45824 | 34816 | -57984 |  X
+
+/*--------------------------------------------------------------------------
+* FUNCTION     : yvup2abgr8888_venum
+*--------------------------------------------------------------------------
+* DESCRIPTION  : Perform YVU planar to ABGR8888 conversion.
+*--------------------------------------------------------------------------
+* C PROTOTYPE  : void yvup2abgr8888_venum(uint8_t  *p_y,
+*                                         uint8_t  *p_cr,
+*                                         uint8_t  *p_cb,
+*                                         uint8_t  *p_abgr8888,
+*                                         uint32_t  length)
+*--------------------------------------------------------------------------
+* REG INPUT    : R0: uint8_t  *p_y
+*                      pointer to the input Y Line
+*                R1: uint8_t  *p_cr
+*                      pointer to the input Cr Line
+*                R2: uint8_t  *p_cb
+*                      pointer to the input Cb Line
+*                R3: uint8_t  *p_abgr8888
+*                      pointer to the output ABGR Line
+*                R12: uint32_t  length
+*                      width of Line
+*--------------------------------------------------------------------------
+* STACK ARG    : None
+*--------------------------------------------------------------------------
+* REG OUTPUT   : None
+*--------------------------------------------------------------------------
+* MEM INPUT    : p_y      - a line of Y pixels
+*                p_cr     - a line of Cr pixels
+*                p_cb     - a line of Cb pixels
+*                length   - the width of the input line
+*--------------------------------------------------------------------------
+* MEM OUTPUT   : p_abgr8888 - the converted ABGR pixels
+*--------------------------------------------------------------------------
+* REG AFFECTED : ARM:  R0-R4, R12
+*                NEON: Q0-Q15
+*--------------------------------------------------------------------------
+* STACK USAGE  : none
+*--------------------------------------------------------------------------
+* CYCLES       : none
+*
+*--------------------------------------------------------------------------
+* NOTES        :
+*--------------------------------------------------------------------------
+*/
+.type yvup2abgr8888_venum, %function
+yvup2abgr8888_venum:
+    /*-------------------------------------------------------------------------
+     *  Store stack registers
+     * ------------------------------------------------------------------------ */
+    STMFD SP!, {LR}
+
+    VPUSH {D8-D15}
+
+    PLD [R0, R3]                       @ preload luma line
+
+    ADR   R12, constants
+
+    VLD1.S16  {D6, D7}, [R12]!         @ D6, D7: 359 |  -88 | -183 | 454 | 256 | 0 | 255 | 0
+    VLD1.S32  {D30, D31}, [R12]        @ Q15   :  -45824    |    34816   |  -57984 |     X
+
+    /*-------------------------------------------------------------------------
+     *  Load the 5th parameter via stack
+     *  R0 ~ R3 are used to pass the first 4 parameters, the 5th and above
+     *  parameters are passed via stack
+     * ------------------------------------------------------------------------ */
+    LDR R12, [SP, #68]                 @ LR is pushed into the stack so SP is
+                                       @ decreased by 4,
+                                       @ D8-D15 are also pushed into the stack
+                                       @ so SP is decreased by
+                                       @ 8-byte/D-Register * 8 D-Registers = 64,
+                                       @ so SP needs to be increased by 64+4=68
+                                       @ to get the value that was first pushed
+                                       @ into stack (the 5th parameter passed in
+                                       @ throught stack)
+
+    /*-------------------------------------------------------------------------
+     *  Load clamping parameters to duplicate vector elements
+     * ------------------------------------------------------------------------ */
+    VDUP.S16  Q4,  D7[1]               @ Q4:  0  |  0  |  0  |  0  |  0  |  0  |  0  |  0
+    VDUP.S16  Q5,  D7[2]               @ Q5: 255 | 255 | 255 | 255 | 255 | 255 | 255 | 255
+
+    /*-------------------------------------------------------------------------
+     *  Read bias
+     * ------------------------------------------------------------------------ */
+    VDUP.S32  Q0,   D30[0]             @ Q0:  -45824 | -45824 | -45824 | -45824
+    VDUP.S32  Q1,   D30[1]             @ Q1:   34816 |  34816 |  34816 |  34816
+    VDUP.S32  Q2,   D31[0]             @ Q2:  -70688 | -70688 | -70688 | -70688
+
+
+    /*-------------------------------------------------------------------------
+     *  The main loop
+     * ------------------------------------------------------------------------ */
+loop_yvup2abgr:
+
+    /*-------------------------------------------------------------------------
+     *  Load input from Y, V and U
+     *  D12  : Y0  Y1  Y2  Y3  Y4  Y5  Y6  Y7
+     *  D14  : V0  V1  V2  V3  V4  V5  V6  V7
+     *  D15  : U0  U1  U2  U3  U4  U5  U6  U7
+     * ------------------------------------------------------------------------ */
+    VLD1.U8  {D12},  [p_y]!            @ Load 8 Luma elements (uint8) to D12
+    VLD1.U8  {D14},  [p_cr]!           @ Load 8 Cr elements (uint8) to D14
+    VLD1.U8  {D15},  [p_cb]!           @ Load 8 Cb elements (uint8) to D15
+
+    /*-------------------------------------------------------------------------
+     *  Expand uint8 value to uint16
+     *  D18, D19: Y0 Y1 Y2 Y3 Y4 Y5 Y6 Y7
+     *  D20, D21: V0 V1 V2 V3 V4 V5 V6 V7
+     *  D22, D23: U0 U1 U2 U3 U4 U5 U6 U7
+     * ------------------------------------------------------------------------ */
+    VMOVL.U8 Q9,  D12
+    VMOVL.U8 Q10, D14
+    VMOVL.U8 Q11, D15
+
+    /*-------------------------------------------------------------------------
+     *  Multiply contribution from chrominance, results are in 32-bit
+     * ------------------------------------------------------------------------ */
+    VMULL.S16  Q12, D20, D6[0]         @ Q12:  359*(V0,V1,V2,V3)     Red
+    VMULL.S16  Q13, D22, D6[1]         @ Q13:  -88*(U0,U1,U2,U3)     Green
+    VMLAL.S16  Q13, D20, D6[2]         @ Q13:  -88*(U0,U1,U2,U3) - 183*(V0,V1,V2,V3)
+    VMULL.S16  Q14, D22, D6[3]         @ Q14:  454*(U0,U1,U2,U3)     Blue
+
+    /*-------------------------------------------------------------------------
+     *  Add bias
+     * ------------------------------------------------------------------------ */
+    VADD.S32  Q12, Q0                  @ Q12 add Red   bias -45824
+    VADD.S32  Q13, Q1                  @ Q13 add Green bias  34816
+    VADD.S32  Q14, Q2                  @ Q14 add Blue  bias -57984
+
+    /*-------------------------------------------------------------------------
+     *  Calculate Red, Green, Blue
+     * ------------------------------------------------------------------------ */
+    VMLAL.S16  Q12, D18, D7[0]         @ Q12: R0, R1, R2, R3 in 32-bit Q8 format
+    VMLAL.S16  Q13, D18, D7[0]         @ Q13: G0, G1, G2, G3 in 32-bit Q8 format
+    VMLAL.S16  Q14, D18, D7[0]         @ Q14: B0, B1, B2, B3 in 32-bit Q8 format
+
+    /*-------------------------------------------------------------------------
+     *  Right shift eight bits with rounding
+     * ------------------------------------------------------------------------ */
+    VSHRN.S32   D18 , Q12, #8          @ D18: R0, R1, R2, R3 in 16-bit Q0 format
+    VSHRN.S32   D20 , Q13, #8          @ D20: G0, G1, G2, G3 in 16-bit Q0 format
+    VSHRN.S32   D22,  Q14, #8          @ D22: B0, B1, B2, B3 in 16-bit Q0 format
+
+    /*-------------------------------------------------------------------------
+     *  Done with the first 4 elements, continue on the next 4 elements
+     * ------------------------------------------------------------------------ */
+
+    /*-------------------------------------------------------------------------
+     *  Multiply contribution from chrominance, results are in 32-bit
+     * ------------------------------------------------------------------------ */
+    VMULL.S16  Q12, D21, D6[0]         @ Q12:  359*(V0,V1,V2,V3)     Red
+    VMULL.S16  Q13, D23, D6[1]         @ Q13: -88*(U0,U1,U2,U3)     Green
+    VMLAL.S16  Q13, D21, D6[2]         @ Q13: -88*(U0,U1,U2,U3) - 183*(V0,V1,V2,V3)
+    VMULL.S16  Q14, D23, D6[3]         @ Q14:  454*(U0,U1,U2,U3)     Blue
+
+    /*-------------------------------------------------------------------------
+     *  Add bias
+     * ------------------------------------------------------------------------ */
+    VADD.S32  Q12, Q0                  @ Q12 add Red   bias -45824
+    VADD.S32  Q13, Q1                  @ Q13 add Green bias  34816
+    VADD.S32  Q14, Q2                  @ Q14 add Blue  bias -57984
+
+    /*-------------------------------------------------------------------------
+     *  Calculate Red, Green, Blue
+     * ------------------------------------------------------------------------ */
+    VMLAL.S16  Q12, D19, D7[0]         @ Q12: R0, R1, R2, R3 in 32-bit Q8 format
+    VMLAL.S16  Q13, D19, D7[0]         @ Q13: G0, G1, G2, G3 in 32-bit Q8 format
+    VMLAL.S16  Q14, D19, D7[0]         @ Q14: B0, B1, B2, B3 in 32-bit Q8 format
+
+    /*-------------------------------------------------------------------------
+     *  Right shift eight bits with rounding
+     * ------------------------------------------------------------------------ */
+    VSHRN.S32   D19 , Q12, #8          @ D18: R0, R1, R2, R3 in 16-bit Q0 format
+    VSHRN.S32   D21 , Q13, #8          @ D20: G0, G1, G2, G3 in 16-bit Q0 format
+    VSHRN.S32   D23,  Q14, #8          @ D22: B0, B1, B2, B3 in 16-bit Q0 format
+
+    /*-------------------------------------------------------------------------
+     *  Clamp the value to be within [0~255]
+     * ------------------------------------------------------------------------ */
+    VMAX.S16  Q11, Q11, Q4             @ if Q11 <   0, Q11 =   0
+    VMIN.S16  Q11, Q11, Q5             @ if Q11 > 255, Q11 = 255
+    VQMOVUN.S16   D28, Q11             @ store Blue to D28, narrow the value from int16 to int8
+
+    VMAX.S16  Q10, Q10, Q4             @ if Q10 <   0, Q10 =   0
+    VMIN.S16  Q10, Q10, Q5             @ if Q10 > 255, Q10 = 255
+    VQMOVUN.S16   D27, Q10             @ store Green to D27, narrow the value from int16 to int8
+
+    VMAX.S16    Q9, Q9, Q4             @ if Q9 <   0, Q9 =   0
+    VMIN.S16    Q9, Q9, Q5             @ if Q9 > 255, Q9 = 255
+    VQMOVUN.S16    D26, Q9             @ store Red to D26, narrow the value from int16 to int8
+
+    /*-------------------------------------------------------------------------
+     *  abgr format with leading 0xFF byte
+     * ------------------------------------------------------------------------ */
+    VMOVN.I16  D29, Q5                 @ D29:  255 | 255 | 255 | 255 | 255 | 255 | 255 | 255
+
+    SUBS length, length, #8            @ check if the length is less than 8
+
+    BMI  trailing_yvup2abgr            @ jump to trailing processing if remaining length is less than 8
+
+    VST4.U8  {D26,D27,D28,D29}, [p_bgr]!   @ vector store Red, Green, Blue to destination
+                                       @ Blue at LSB
+
+    BHI loop_yvup2abgr                 @ loop if more than 8 pixels left
+
+    BEQ  end_yvup2abgr                 @ done if exactly 8 pixel processed in the loop
+
+
+trailing_yvup2abgr:
+    /*-------------------------------------------------------------------------
+     *  There are from 1 ~ 7 pixels left in the trailing part.
+     *  First adding 7 to the length so the length would be from 0 ~ 6.
+     *  eg: 1 pixel left in the trailing part, so 1-8+7 = 0.
+     *  Then save 1 pixel unconditionally since at least 1 pixels left in the
+     *  trailing part.
+     * ------------------------------------------------------------------------ */
+    ADDS length, length, #7            @ there are 7 or less in the trailing part
+
+    VST4.U8 {D26[0], D27[0], D28[0], D29[0]}, [p_bgr]! @ at least 1 pixel left in the trailing part
+    BEQ  end_yvup2abgr                 @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST4.U8 {D26[1], D27[1], D28[1], D29[1]}, [p_bgr]!  @ store one more pixel
+    BEQ  end_yvup2abgr                 @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST4.U8 {D26[2], D27[2], D28[2], D29[2]}, [p_bgr]!  @ store one more pixel
+    BEQ  end_yvup2abgr                 @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST4.U8 {D26[3], D27[3], D28[3], D29[3]}, [p_bgr]!  @ store one more pixel
+    BEQ  end_yvup2abgr                 @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST4.U8 {D26[4], D27[4], D28[4], D29[4]}, [p_bgr]!  @ store one more pixel
+    BEQ  end_yvup2abgr                 @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST4.U8 {D26[5], D27[5], D28[5], D29[5]}, [p_bgr]!  @ store one more pixel
+    BEQ  end_yvup2abgr                 @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST4.U8 {D26[6], D27[6], D28[6], D29[6]}, [p_bgr]! @ store one more pixel
+
+end_yvup2abgr:
+    VPOP  {D8-D15}
+    LDMFD SP!, {PC}
+                                       @ end of yvup2abgr
+
+/*--------------------------------------------------------------------------
+* FUNCTION     : yyvup2abgr8888_venum
+*--------------------------------------------------------------------------
+* DESCRIPTION  : Perform YYVU planar to ABGR8888 conversion.
+*--------------------------------------------------------------------------
+* C PROTOTYPE  : void yyvup2abgr8888_venum(uint8_t  *p_y,
+*                                          uint8_t  *p_cr,
+*                                          uint8_t  *p_cb,
+*                                          uint8_t  *p_abgr8888,
+*                                          uint32_t  length)
+*--------------------------------------------------------------------------
+* REG INPUT    : R0: uint8_t  *p_y
+*                      pointer to the input Y Line
+*                R1: uint8_t  *p_cr
+*                      pointer to the input Cr Line
+*                R2: uint8_t  *p_cb
+*                      pointer to the input Cb Line
+*                R3: uint8_t  *p_abgr8888
+*                      pointer to the output ABGR Line
+*                R12: uint32_t  length
+*                      width of Line
+*--------------------------------------------------------------------------
+* STACK ARG    : None
+*--------------------------------------------------------------------------
+* REG OUTPUT   : None
+*--------------------------------------------------------------------------
+* MEM INPUT    : p_y      - a line of Y pixels
+*                p_cr     - a line of Cr pixels
+*                p_cb     - a line of Cb pixels
+*                length   - the width of the input line
+*--------------------------------------------------------------------------
+* MEM OUTPUT   : p_abgr8888 - the converted ABGR pixels
+*--------------------------------------------------------------------------
+* REG AFFECTED : ARM:  R0-R4, R12
+*                NEON: Q0-Q15
+*--------------------------------------------------------------------------
+* STACK USAGE  : none
+*--------------------------------------------------------------------------
+* CYCLES       : none
+*
+*--------------------------------------------------------------------------
+* NOTES        :
+*--------------------------------------------------------------------------
+*/
+.type yyvup2abgr8888_venum, %function
+yyvup2abgr8888_venum:
+    /*-------------------------------------------------------------------------
+     *  Store stack registers
+     * ------------------------------------------------------------------------ */
+    STMFD SP!, {LR}
+
+    VPUSH {D8-D15}
+
+    PLD [R0, R3]                       @ preload luma line
+
+    ADR   R12, constants
+
+    VLD1.S16  {D6, D7}, [R12]!         @ D6, D7: 359 |  -88 | -183 | 454 | 256 | 0 | 255 | 0
+    VLD1.S32  {D30, D31}, [R12]        @ Q15   :  -45824    |    34816   |  -57984 |     X
+
+    /*-------------------------------------------------------------------------
+     *  Load the 5th parameter via stack
+     *  R0 ~ R3 are used to pass the first 4 parameters, the 5th and above
+     *  parameters are passed via stack
+     * ------------------------------------------------------------------------ */
+    LDR R12, [SP, #68]                 @ LR is pushed into the stack so SP is
+                                       @ decreased by 4,
+                                       @ D8-D15 are also pushed into the stack
+                                       @ so SP is decreased by
+                                       @ 8-byte/D-Register * 8 D-Registers = 64,
+                                       @ so SP needs to be increased by 64+4=68
+                                       @ to get the value that was first pushed
+                                       @ into stack (the 5th parameter passed in
+                                       @ throught stack)
+
+    /*-------------------------------------------------------------------------
+     *  Load clamping parameters to duplicate vector elements
+     * ------------------------------------------------------------------------ */
+    VDUP.S16  Q4,  D7[1]               @ Q4:  0  |  0  |  0  |  0  |  0  |  0  |  0  |  0
+    VDUP.S16  Q5,  D7[2]               @ Q5: 255 | 255 | 255 | 255 | 255 | 255 | 255 | 255
+
+    /*-------------------------------------------------------------------------
+     *  Read bias
+     * ------------------------------------------------------------------------ */
+    VDUP.S32  Q0,   D30[0]             @ Q0:  -45824 | -45824 | -45824 | -45824
+    VDUP.S32  Q1,   D30[1]             @ Q1:   34816 |  34816 |  34816 |  34816
+    VDUP.S32  Q2,   D31[0]             @ Q2:  -70688 | -70688 | -70688 | -70688
+
+
+    /*-------------------------------------------------------------------------
+     *  The main loop
+     * ------------------------------------------------------------------------ */
+loop_yyvup2abgr:
+
+    /*-------------------------------------------------------------------------
+     *  Load input from Y, V and U
+     *  D12, D13: Y0 Y2 Y4 Y6 Y8 Y10 Y12 Y14, Y1 Y3 Y5 Y7 Y9 Y11 Y13 Y15
+     *  D14  : V0  V1  V2  V3  V4  V5  V6  V7
+     *  D15  : U0  U1  U2  U3  U4  U5  U6  U7
+     * ------------------------------------------------------------------------ */
+    VLD2.U8  {D12,D13}, [p_y]!         @ Load 16 Luma elements (uint8) to D12, D13
+    VLD1.U8  {D14},  [p_cr]!           @ Load 8 Cr elements (uint8) to D14
+    VLD1.U8  {D15},  [p_cb]!           @ Load 8 Cb elements (uint8) to D15
+
+    /*-------------------------------------------------------------------------
+     *  Expand uint8 value to uint16
+     *  D24, D25: Y0 Y2 Y4 Y6 Y8 Y10 Y12 Y14
+     *  D26, D27: Y1 Y3 Y5 Y7 Y9 Y11 Y13 Y15
+     *  D28, D29: V0 V1 V2 V3 V4 V5  V6  V7
+     *  D30, D31: U0 U1 U2 U3 U4 U5  U6  U7
+     * ------------------------------------------------------------------------ */
+    VMOVL.U8 Q12, D12
+    VMOVL.U8 Q13, D13
+    VMOVL.U8 Q14, D14
+    VMOVL.U8 Q15, D15
+
+    /*-------------------------------------------------------------------------
+     *  Multiply contribution from chrominance, results are in 32-bit
+     * ------------------------------------------------------------------------ */
+    VMULL.S16  Q6, D28, D6[0]          @ Q6:  359*(V0,V1,V2,V3)     Red
+    VMULL.S16  Q7, D30, D6[1]          @ Q7: -88*(U0,U1,U2,U3)     Green
+    VMLAL.S16  Q7, D28, D6[2]          @ Q7: -88*(U0,U1,U2,U3) - 183*(V0,V1,V2,V3)
+    VMULL.S16  Q8, D30, D6[3]          @ Q8:  454*(U0,U1,U2,U3)     Blue
+
+    /*-------------------------------------------------------------------------
+     *  Add bias
+     * ------------------------------------------------------------------------ */
+    VADD.S32  Q6, Q0                   @ Q6 add Red   bias -45824
+    VADD.S32  Q7, Q1                   @ Q7 add Green bias  34816
+    VADD.S32  Q8, Q2                   @ Q8 add Blue  bias -57984
+
+    /*-------------------------------------------------------------------------
+     *  Calculate Red, Green, Blue
+     * ------------------------------------------------------------------------ */
+    VMOV.S32   Q9, Q6
+    VMLAL.S16  Q6, D24, D7[0]          @ Q6: R0, R2, R4, R6 in 32-bit Q8 format
+    VMLAL.S16  Q9, D26, D7[0]          @ Q9: R1, R3, R5, R7 in 32-bit Q8 format
+
+    VMOV.S32   Q10, Q7
+    VMLAL.S16  Q7,  D24, D7[0]         @ Q7:  G0, G2, G4, G6 in 32-bit Q8 format
+    VMLAL.S16  Q10, D26, D7[0]         @ Q10: G1, G3, G5, G7 in 32-bit Q8 format
+
+    VMOV.S32   Q11, Q8
+    VMLAL.S16  Q8,  D24, D7[0]         @ Q8:  B0, B2, B4, B6 in 32-bit Q8 format
+    VMLAL.S16  Q11, D26, D7[0]         @ Q11: B1, B3, B5, B7 in 32-bit Q8 format
+
+    /*-------------------------------------------------------------------------
+     *  Right shift eight bits with rounding
+     * ------------------------------------------------------------------------ */
+    VSHRN.S32   D12, Q6,  #8           @ D12: R0 R2 R4 R6 in 16-bit Q0 format
+    VSHRN.S32   D13, Q9,  #8           @ D13: R1 R3 R5 R7 in 16-bit Q0 format
+    VZIP.16     D12, D13               @ Q6 : R0 R1 R2 R3 R4 R5 R6 R7
+
+    VSHRN.S32   D18, Q7,  #8           @ D18: G0 G2 G4 G6 in 16-bit Q0 format
+    VSHRN.S32   D19, Q10, #8           @ D19: G1 G3 G5 G7 in 16-bit Q0 format
+    VZIP.16     D18, D19               @ Q9 : G0 G1 G2 G3 G4 G5 G6 G7
+
+    VSHRN.S32   D20, Q8,  #8           @ D20: B0 B2 B4 B6 in 16-bit Q0 format
+    VSHRN.S32   D21, Q11, #8           @ D21: B1 B3 B5 B7 in 16-bit Q0 format
+    VZIP.16     D20, D21               @ Q10: B0 B1 B2 B3 B4 B5 B6 B7
+
+    /*-------------------------------------------------------------------------
+     *  Clamp the value to be within [0~255]
+     * ------------------------------------------------------------------------ */
+    VMAX.S16  Q10, Q10, Q4             @ if Q10 <   0, Q10 =   0
+    VMIN.S16  Q10, Q10, Q5             @ if Q10 > 255, Q10 = 255
+    VQMOVUN.S16   D23, Q10             @ store Blue to D23, narrow the value from int16 to int8
+
+    VMAX.S16  Q9, Q9, Q4               @ if Q9 <   0, Q9 =   0
+    VMIN.S16  Q9, Q9, Q5               @ if Q9 > 255, Q9 = 255
+    VQMOVUN.S16  D22, Q9               @ store Green to D22, narrow the value from int16 to int8
+
+    VMAX.S16  Q6, Q6, Q4               @ if Q6 <   0, Q6 =   0
+    VMIN.S16  Q6, Q6, Q5               @ if Q6 > 255, Q6 = 255
+    VQMOVUN.S16  D21, Q6               @ store Red to D21, narrow the value from int16 to int8
+
+    /*-------------------------------------------------------------------------
+     *  abgr format with leading 0xFF byte
+     * ------------------------------------------------------------------------ */
+    VMOVN.I16  D24, Q5                 @ D24:  255 | 255 | 255 | 255 | 255 | 255 | 255 | 255
+
+    SUBS length, length, #8            @ check if the length is less than 8
+
+    BMI  trailing_yyvup2abgr           @ jump to trailing processing if remaining length is less than 8
+
+    VST4.U8  {D21,D22,D23,D24}, [p_bgr]!   @ vector store Blue, Green, Red to destination
+                                       @ Red at LSB
+
+    BEQ  end_yyvup2abgr                @ done if exactly 8 pixel processed in the loop
+
+
+    /*-------------------------------------------------------------------------
+     *  Done with the first 8 elements, continue on the next 8 elements
+     * ------------------------------------------------------------------------ */
+
+    /*-------------------------------------------------------------------------
+     *  Multiply contribution from chrominance, results are in 32-bit
+     * ------------------------------------------------------------------------ */
+    VMULL.S16  Q6, D29, D6[0]          @ Q6: 359*(V4,V5,V6,V7)       Red
+    VMULL.S16  Q7, D31, D6[1]          @ Q7: -88*(U4,U5,U6,U7)      Green
+    VMLAL.S16  Q7, D29, D6[2]          @ Q7: -88*(U4,U5,U6,U7) - 183*(V4,V5,V6,V7)
+    VMULL.S16  Q8, D31, D6[3]          @ Q8: 454*(U4,U5,U6,U7)       Blue
+
+    /*-------------------------------------------------------------------------
+     *  Add bias
+     * ------------------------------------------------------------------------ */
+    VADD.S32  Q6, Q0                   @ Q6 add Red   bias -45824
+    VADD.S32  Q7, Q1                   @ Q7 add Green bias  34816
+    VADD.S32  Q8, Q2                   @ Q8 add Blue  bias -57984
+
+    /*-------------------------------------------------------------------------
+     *  Calculate Red, Green, Blue
+     * ------------------------------------------------------------------------ */
+    VMOV.S32   Q9, Q6
+    VMLAL.S16  Q6, D25, D7[0]          @ Q6: R8 R10 R12 R14 in 32-bit Q8 format
+    VMLAL.S16  Q9, D27, D7[0]          @ Q9: R9 R11 R13 R15 in 32-bit Q8 format
+
+    VMOV.S32   Q10, Q7
+    VMLAL.S16  Q7,  D25, D7[0]         @ Q7: G0, G2, G4, G6 in 32-bit Q8 format
+    VMLAL.S16  Q10, D27, D7[0]         @ Q10 : G1, G3, G5, G7 in 32-bit Q8 format
+
+    VMOV.S32   Q11, Q8
+    VMLAL.S16  Q8,  D25, D7[0]         @ Q8: B0, B2, B4, B6 in 32-bit Q8 format
+    VMLAL.S16  Q11, D27, D7[0]         @ Q11 : B1, B3, B5, B7 in 32-bit Q8 format
+
+    /*-------------------------------------------------------------------------
+     *  Right shift eight bits with rounding
+     * ------------------------------------------------------------------------ */
+    VSHRN.S32   D12, Q6,  #8           @ D12: R8 R10 R12 R14 in 16-bit Q0 format
+    VSHRN.S32   D13, Q9,  #8           @ D13: R9 R11 R13 R15 in 16-bit Q0 format
+    VZIP.16     D12, D13               @ Q6: R8 R9 R10 R11 R12 R13 R14 R15
+
+    VSHRN.S32   D18, Q7,  #8           @ D18: G8 G10 G12 G14 in 16-bit Q0 format
+    VSHRN.S32   D19, Q10, #8           @ D19: G9 G11 G13 G15 in 16-bit Q0 format
+    VZIP.16     D18, D19               @ Q9:  G8 G9 G10 G11 G12 G13 G14 G15
+
+    VSHRN.S32   D20, Q8,  #8           @ D20: B8 B10 B12 B14 in 16-bit Q0 format
+    VSHRN.S32   D21, Q11, #8           @ D21: B9 B11 B13 B15 in 16-bit Q0 format
+    VZIP.16     D20, D21               @ Q10: B8 B9 B10 B11 B12 B13 B14 B15
+
+    /*-------------------------------------------------------------------------
+     *  Clamp the value to be within [0~255]
+     * ------------------------------------------------------------------------ */
+    VMAX.S16  Q10, Q10, Q4             @ if Q10 <   0, Q10 =   0
+    VMIN.S16  Q10, Q10, Q5             @ if Q10 > 255, Q10 = 255
+    VQMOVUN.S16   D23, Q10             @ store Blue to D23, narrow the value from int16 to int8
+
+    VMAX.S16  Q9, Q9, Q4               @ if Q9 <   0, Q9 =   0
+    VMIN.S16  Q9, Q9, Q5               @ if Q9 > 255, Q9 = 255
+    VQMOVUN.S16  D22, Q9               @ store Green to D22, narrow the value from int16 to int8
+
+    VMAX.S16  Q6, Q6, Q4               @ if Q6 <   0, Q6 =   0
+    VMIN.S16  Q6, Q6, Q5               @ if Q6 > 255, Q6 = 255
+    VQMOVUN.S16  D21, Q6               @ store Red to D21, narrow the value from int16 to int8
+
+    /*-------------------------------------------------------------------------
+     *  abgr format with leading 0xFF byte
+     * ------------------------------------------------------------------------ */
+    VMOVN.I16  D24, Q5                 @ D24:  255 | 255 | 255 | 255 | 255 | 255 | 255 | 255
+
+    SUBS length, length, #8            @ check if the length is less than 8
+
+    BMI  trailing_yyvup2abgr           @ jump to trailing processing if remaining length is less than 8
+
+    VST4.U8  {D21,D22,D23,D24}, [p_bgr]!   @ vector store Blue, Green, Red to destination
+                                       @ Red at LSB
+
+    BHI loop_yyvup2abgr                @ loop if more than 8 pixels left
+
+    BEQ  end_yyvup2abgr                @ done if exactly 8 pixel processed in the loop
+
+
+trailing_yyvup2abgr:
+    /*-------------------------------------------------------------------------
+     *  There are from 1 ~ 7 pixels left in the trailing part.
+     *  First adding 7 to the length so the length would be from 0 ~ 6.
+     *  eg: 1 pixel left in the trailing part, so 1-8+7 = 0.
+     *  Then save 1 pixel unconditionally since at least 1 pixels left in the
+     *  trailing part.
+     * ------------------------------------------------------------------------ */
+    ADDS length, length, #7            @ there are 7 or less in the trailing part
+
+    VST4.U8 {D21[0],D22[0],D23[0],D24[0]}, [p_bgr]! @ at least 1 pixel left in the trailing part
+    BEQ end_yyvup2abgr                 @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST4.U8 {D21[1],D22[1],D23[1],D24[1]}, [p_bgr]!  @ store one more pixel
+    BEQ end_yyvup2abgr                 @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST4.U8 {D21[2],D22[2],D23[2],D24[2]}, [p_bgr]!  @ store one more pixel
+    BEQ end_yyvup2abgr                 @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST4.U8 {D21[3],D22[3],D23[3],D24[3]}, [p_bgr]!  @ store one more pixel
+    BEQ end_yyvup2abgr                 @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST4.U8 {D21[4],D22[4],D23[4],D24[4]}, [p_bgr]!  @ store one more pixel
+    BEQ end_yyvup2abgr                 @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST4.U8 {D21[5],D22[5],D23[5],D24[5]}, [p_bgr]!  @ store one more pixel
+    BEQ end_yyvup2abgr                 @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST4.U8 {D21[6],D22[6],D23[6],D24[6]}, [p_bgr]!  @ store one more pixel
+
+end_yyvup2abgr:
+    VPOP  {D8-D15}
+    LDMFD SP!, {PC}
+                                       @ end of yyvup2abgr
+
+.end
+
diff --git a/libjpegtwrp/asm/armv7/jdcolor-armv7.S b/libjpegtwrp/asm/armv7/jdcolor-armv7.S
new file mode 100644
index 000000000..b2da6d5d0
--- /dev/null
+++ b/libjpegtwrp/asm/armv7/jdcolor-armv7.S
@@ -0,0 +1,632 @@
+/*------------------------------------------------------------------------
+* jdcolor-armv7.S
+*
+*  Copyright (c) 2010, Code Aurora Forum. All rights reserved.
+*
+*  Redistribution and use in source and binary forms, with or without
+*  modification, are permitted provided that the following conditions are
+*  met:
+*      * Redistributions of source code must retain the above copyright
+*        notice, this list of conditions and the following disclaimer.
+*      * Redistributions in binary form must reproduce the above
+*        copyright notice, this list of conditions and the following
+*        disclaimer in the documentation and/or other materials provided
+*        with the distribution.
+*      * Neither the name of Code Aurora Forum, Inc. nor the names of its
+*        contributors may be used to endorse or promote products derived
+*        from this software without specific prior written permission.
+*
+*  THIS SOFTWARE IS PROVIDED "AS IS" AND ANY EXPRESS OR IMPLIED
+*  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+*  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT
+*  ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS
+*  BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+*  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+*  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
+*  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
+*  WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
+*  OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN
+*  IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*--------------------------------------------------------------------------
+
+*--------------------------------------------------------------------------
+*                         FUNCTION LIST
+*--------------------------------------------------------------------------
+*
+* - yvup2bgr888_venum
+* - yyvup2bgr888_venum
+*
+*--------------------------------------------------------------------------
+*/
+
+    .section yvu_plain_to_bgr, "x"     @ AREA
+    .text                              @ |.text|, CODE, READONLY
+    .align 2
+    .code  32                          @ CODE32
+
+/*-----------------------------------------------------------------------------
+ *   ARM Registers
+ * ---------------------------------------------------------------------------- */
+p_y       .req r0
+p_cr      .req r1
+p_cb      .req r2
+p_rgb     .req r3
+p_bgr     .req r3
+length    .req r12
+
+    .global yvup2bgr888_venum
+    .global yyvup2bgr888_venum
+
+@ coefficients in color conversion matrix multiplication
+.equ COEFF_Y,          256             @ contribution of Y
+.equ COEFF_V_RED,      359             @ contribution of V for red
+.equ COEFF_U_GREEN,    -88             @ contribution of U for green
+.equ COEFF_V_GREEN,   -183             @ contribution of V for green
+.equ COEFF_U_BLUE,     454             @ contribution of U for blue
+
+@ Clamping constants 0x0 and 0xFF
+.equ COEFF_0,          0
+.equ COEFF_255,        255
+
+@ Bias coefficients for red, green and blue
+.equ COEFF_BIAS_R,   -45824            @ Red   bias =     -359*128 + 128
+.equ COEFF_BIAS_G,    34816            @ Green bias = (88+183)*128 + 128
+.equ COEFF_BIAS_B,   -57984            @ Blue  bias =     -454*128 + 128
+
+constants:
+    .hword (COEFF_V_RED),  (COEFF_U_GREEN), (COEFF_V_GREEN), (COEFF_U_BLUE) @   359  | -88   |  -183  | 454
+    .hword (COEFF_Y),      (COEFF_0),       (COEFF_255)    , (COEFF_0)      @   256  |   0   |   255  |  0
+    .word  (COEFF_BIAS_R), (COEFF_BIAS_G),  (COEFF_BIAS_B)                  @ -45824 | 34816 | -57984 |  X
+
+/*--------------------------------------------------------------------------
+* FUNCTION     : yvup2bgr888_venum
+*--------------------------------------------------------------------------
+* DESCRIPTION  : Perform YVU planar to BGR888 conversion.
+*--------------------------------------------------------------------------
+* C PROTOTYPE  : void yvup2bgr888_venum(uint8_t  *p_y,
+*                                       uint8_t  *p_cr,
+*                                       uint8_t  *p_cb,
+*                                       uint8_t  *p_bgr888,
+*                                       uint32_t  length)
+*--------------------------------------------------------------------------
+* REG INPUT    : R0: uint8_t  *p_y
+*                      pointer to the input Y Line
+*                R1: uint8_t  *p_cr
+*                      pointer to the input Cr Line
+*                R2: uint8_t  *p_cb
+*                      pointer to the input Cb Line
+*                R3: uint8_t  *p_bgr888
+*                      pointer to the output BGR Line
+*                R12: uint32_t  length
+*                      width of Line
+*--------------------------------------------------------------------------
+* STACK ARG    : None
+*--------------------------------------------------------------------------
+* REG OUTPUT   : None
+*--------------------------------------------------------------------------
+* MEM INPUT    : p_y      - a line of Y pixels
+*                p_cr     - a line of Cr pixels
+*                p_cb     - a line of Cb pixels
+*                length   - the width of the input line
+*--------------------------------------------------------------------------
+* MEM OUTPUT   : p_bgr888 - the converted bgr pixels
+*--------------------------------------------------------------------------
+* REG AFFECTED : ARM:  R0-R4, R12
+*                NEON: Q0-Q15
+*--------------------------------------------------------------------------
+* STACK USAGE  : none
+*--------------------------------------------------------------------------
+* CYCLES       : none
+*
+*--------------------------------------------------------------------------
+* NOTES        :
+*--------------------------------------------------------------------------
+*/
+.type yvup2bgr888_venum, %function
+yvup2bgr888_venum:
+
+    /*-------------------------------------------------------------------------
+     *  Store stack registers
+     * ------------------------------------------------------------------------ */
+    STMFD SP!, {LR}
+
+    VPUSH {D8-D15}
+
+    PLD [R0, R3]                      @ preload luma line
+
+    ADR   R12, constants
+
+    VLD1.S16  {D6, D7}, [R12]!        @ D6, D7: 359 | -88 | -183 | 454 | 256 | 0 | 255 | 0
+    VLD1.S32  {D30, D31}, [R12]       @ Q15   :  -45824   |    34816   |  -57984 |     X
+
+    /*-------------------------------------------------------------------------
+     *  Load the 5th parameter via stack
+     *  R0 ~ R3 are used to pass the first 4 parameters, the 5th and above
+     *  parameters are passed via stack
+     * ------------------------------------------------------------------------ */
+    LDR R12, [SP, #68]                 @ LR is pushed into the stack so SP is
+                                       @ decreased by 4,
+                                       @ D8-D15 are also pushed into the stack
+                                       @ so SP is decreased by
+                                       @ 8-byte/D-Register * 8 D-Registers = 64,
+                                       @ so SP needs to be increased by 64+4=68
+                                       @ to get the value that was first pushed
+                                       @ into stack (the 5th parameter passed in
+                                       @ throught stack)
+
+    /*-------------------------------------------------------------------------
+     *  Load clamping parameters to duplicate vector elements
+     * ------------------------------------------------------------------------ */
+    VDUP.S16  Q4,  D7[1]              @ Q4:  0  |  0  |  0  |  0  |  0  |  0  |  0  |  0
+    VDUP.S16  Q5,  D7[2]              @ Q5: 255 | 255 | 255 | 255 | 255 | 255 | 255 | 255
+
+    /*-------------------------------------------------------------------------
+     *  Read bias
+     * ------------------------------------------------------------------------ */
+    VDUP.S32  Q0,   D30[0]            @ Q0:  -45824 | -45824 | -45824 | -45824
+    VDUP.S32  Q1,   D30[1]            @ Q1:   34816 |  34816 |  34816 |  34816
+    VDUP.S32  Q2,   D31[0]            @ Q2:  -57984 | -57984 | -57984 | -57984
+
+
+    /*-------------------------------------------------------------------------
+     *  The main loop
+     * ------------------------------------------------------------------------ */
+loop_yvup2bgr888:
+
+    /*-------------------------------------------------------------------------
+     *  Load input from Y, V and U
+     *  D12  : Y0  Y1  Y2  Y3  Y4  Y5  Y6  Y7
+     *  D14  : V0  V1  V2  V3  V4  V5  V6  V7
+     *  D15  : U0  U1  U2  U3  U4  U5  U6  U7
+     * ------------------------------------------------------------------------ */
+    VLD1.U8  {D12},  [p_y]!           @ Load 8 Luma elements (uint8) to D12
+    VLD1.U8  {D14},  [p_cr]!          @ Load 8 Cr elements (uint8) to D14
+    VLD1.U8  {D15},  [p_cb]!          @ Load 8 Cb elements (uint8) to D15
+
+    /*-------------------------------------------------------------------------
+     *  Expand uint8 value to uint16
+     *  D18, D19: Y0 Y1 Y2 Y3 Y4 Y5 Y6 Y7
+     *  D20, D21: V0 V1 V2 V3 V4 V5 V6 V7
+     *  D22, D23: U0 U1 U2 U3 U4 U5 U6 U7
+     * ------------------------------------------------------------------------ */
+    VMOVL.U8 Q9,  D12
+    VMOVL.U8 Q10, D14
+    VMOVL.U8 Q11, D15
+
+    /*-------------------------------------------------------------------------
+     *  Multiply contribution from chrominance, results are in 32-bit
+     * ------------------------------------------------------------------------ */
+    VMULL.S16  Q12, D20, D6[0]        @ Q12:  359*(V0,V1,V2,V3)     Red
+    VMULL.S16  Q13, D22, D6[1]        @ Q13:  -88*(U0,U1,U2,U3)     Green
+    VMLAL.S16  Q13, D20, D6[2]        @ Q13:  -88*(U0,U1,U2,U3) - 183*(V0,V1,V2,V3)
+    VMULL.S16  Q14, D22, D6[3]        @ Q14:  454*(U0,U1,U2,U3)     Blue
+
+    /*-------------------------------------------------------------------------
+     *  Add bias
+     * ------------------------------------------------------------------------ */
+    VADD.S32  Q12, Q0                 @ Q12 add Red   bias -45824
+    VADD.S32  Q13, Q1                 @ Q13 add Green bias  34816
+    VADD.S32  Q14, Q2                 @ Q14 add Blue  bias -57984
+
+    /*-------------------------------------------------------------------------
+     *  Calculate Red, Green, Blue
+     * ------------------------------------------------------------------------ */
+    VMLAL.S16  Q12, D18, D7[0]        @ Q12: R0, R1, R2, R3 in 32-bit Q8 format
+    VMLAL.S16  Q13, D18, D7[0]        @ Q13: G0, G1, G2, G3 in 32-bit Q8 format
+    VMLAL.S16  Q14, D18, D7[0]        @ Q14: B0, B1, B2, B3 in 32-bit Q8 format
+
+    /*-------------------------------------------------------------------------
+     *  Right shift eight bits with rounding
+     * ------------------------------------------------------------------------ */
+    VSHRN.S32   D18 , Q12, #8         @ D18: R0, R1, R2, R3 in 16-bit Q0 format
+    VSHRN.S32   D20 , Q13, #8         @ D20: G0, G1, G2, G3 in 16-bit Q0 format
+    VSHRN.S32   D22,  Q14, #8         @ D22: B0, B1, B2, B3 in 16-bit Q0 format
+
+    /*-------------------------------------------------------------------------
+     *  Done with the first 4 elements, continue on the next 4 elements
+     * ------------------------------------------------------------------------ */
+
+    /*-------------------------------------------------------------------------
+     *  Multiply contribution from chrominance, results are in 32-bit
+     * ------------------------------------------------------------------------ */
+    VMULL.S16  Q12, D21, D6[0]        @ Q12:  359*(V0,V1,V2,V3)     Red
+    VMULL.S16  Q13, D23, D6[1]        @ Q13:  -88*(U0,U1,U2,U3)     Green
+    VMLAL.S16  Q13, D21, D6[2]        @ Q13:  -88*(U0,U1,U2,U3) - 183*(V0,V1,V2,V3)
+    VMULL.S16  Q14, D23, D6[3]        @ Q14:  454*(U0,U1,U2,U3)     Blue
+
+    /*-------------------------------------------------------------------------
+     *  Add bias
+     * ------------------------------------------------------------------------ */
+    VADD.S32  Q12, Q0                 @ Q12 add Red   bias -45824
+    VADD.S32  Q13, Q1                 @ Q13 add Green bias  34816
+    VADD.S32  Q14, Q2                 @ Q14 add Blue  bias -57984
+
+    /*-------------------------------------------------------------------------
+     *  Calculate Red, Green, Blue
+     * ------------------------------------------------------------------------ */
+    VMLAL.S16  Q12, D19, D7[0]        @ Q12: R0, R1, R2, R3 in 32-bit Q8 format
+    VMLAL.S16  Q13, D19, D7[0]        @ Q13: G0, G1, G2, G3 in 32-bit Q8 format
+    VMLAL.S16  Q14, D19, D7[0]        @ Q14: B0, B1, B2, B3 in 32-bit Q8 format
+
+    /*-------------------------------------------------------------------------
+     *  Right shift eight bits with rounding
+     * ------------------------------------------------------------------------ */
+    VSHRN.S32   D19 , Q12, #8         @ D18: R0, R1, R2, R3 in 16-bit Q0 format
+    VSHRN.S32   D21 , Q13, #8         @ D20: G0, G1, G2, G3 in 16-bit Q0 format
+    VSHRN.S32   D23,  Q14, #8         @ D22: B0, B1, B2, B3 in 16-bit Q0 format
+
+    /*-------------------------------------------------------------------------
+     *  Clamp the value to be within [0~255]
+     * ------------------------------------------------------------------------ */
+    VMAX.S16  Q11, Q11, Q4            @ if Q11 <   0, Q11 =   0
+    VMIN.S16  Q11, Q11, Q5            @ if Q11 > 255, Q11 = 255
+    VQMOVUN.S16   D28, Q11            @ store Blue to D28, narrow the value from int16 to int8
+
+    VMAX.S16  Q10, Q10, Q4            @ if Q10 <   0, Q10 =   0
+    VMIN.S16  Q10, Q10, Q5            @ if Q10 > 255, Q10 = 255
+    VQMOVUN.S16   D27, Q10            @ store Green to D27, narrow the value from int16 to int8
+
+    VMAX.S16    Q9, Q9, Q4            @ if Q9 <   0, Q9 =   0
+    VMIN.S16    Q9, Q9, Q5            @ if Q9 > 255, Q9 = 255
+    VQMOVUN.S16    D26, Q9            @ store Red to D26, narrow the value from int16 to int8.
+
+    SUBS length, length, #8           @ check if the length is less than 8
+
+    BMI  trailing_yvup2bgr888         @ jump to trailing processing if remaining length is less than 8
+
+    VST3.U8  {D26,D27,D28}, [p_bgr]!  @ vector store Red, Green, Blue to destination
+                                      @ Blue at LSB
+
+    BHI loop_yvup2bgr888              @ loop if more than 8 pixels left
+
+    BEQ  end_yvup2bgr888              @ done if exactly 8 pixel processed in the loop
+
+
+trailing_yvup2bgr888:
+    /*-------------------------------------------------------------------------
+     *  There are from 1 ~ 7 pixels left in the trailing part.
+     *  First adding 7 to the length so the length would be from 0 ~ 6.
+     *  eg: 1 pixel left in the trailing part, so 1-8+7 = 0.
+     *  Then save 1 pixel unconditionally since at least 1 pixels left in the
+     *  trailing part.
+     * ------------------------------------------------------------------------ */
+    ADDS length, length, #7           @ there are 7 or less in the trailing part
+
+    VST3.U8 {D26[0], D27[0], D28[0]}, [p_bgr]! @ at least 1 pixel left in the trailing part
+    BEQ  end_yvup2bgr888                       @ done if 0 pixel left
+
+    SUBS length, length, #1           @ update length counter
+    VST3.U8 {D26[1], D27[1], D28[1]}, [p_bgr]!  @ store one more pixel
+    BEQ  end_yvup2bgr888                        @ done if 0 pixel left
+
+    SUBS length, length, #1           @ update length counter
+    VST3.U8 {D26[2], D27[2], D28[2]}, [p_bgr]!  @ store one more pixel
+    BEQ  end_yvup2bgr888                        @ done if 0 pixel left
+
+    SUBS length, length, #1           @ update length counter
+    VST3.U8 {D26[3], D27[3], D28[3]}, [p_bgr]!  @ store one more pixel
+    BEQ  end_yvup2bgr888                        @ done if 0 pixel left
+
+    SUBS length, length, #1           @ update length counter
+    VST3.U8 {D26[4], D27[4], D28[4]}, [p_bgr]!  @ store one more pixel
+    BEQ  end_yvup2bgr888                        @ done if 0 pixel left
+
+    SUBS length, length, #1           @ update length counter
+    VST3.U8 {D26[5], D27[5], D28[5]}, [p_bgr]!  @ store one more pixel
+    BEQ  end_yvup2bgr888                        @ done if 0 pixel left
+
+    SUBS length, length, #1           @ update length counter
+    VST3.U8 {D26[6], D27[6], D28[6]}, [p_bgr]!  @ store one more pixel
+
+end_yvup2bgr888:
+    VPOP  {D8-D15}
+    LDMFD SP!, {PC}
+
+                                      @ end of yvup2bgr888
+
+
+/*-------------------------------------------------------------------------
+* FUNCTION     : yyvup2bgr888_venum
+*--------------------------------------------------------------------------
+* DESCRIPTION  : Perform YYVU planar to BGR888 conversion.
+*--------------------------------------------------------------------------
+* C PROTOTYPE  : void yyvup2bgr888_venum(uint8_t  *p_y,
+*                                        uint8_t  *p_cr,
+*                                        uint8_t  *p_cb,
+*                                        uint8_t  *p_bgr888,
+*                                        uint32_t  length)
+*--------------------------------------------------------------------------
+* REG INPUT    : R0: uint8_t  *p_y
+*                      pointer to the input Y Line
+*                R1: uint8_t  *p_cr
+*                      pointer to the input Cr Line
+*                R2: uint8_t  *p_cb
+*                      pointer to the input Cb Line
+*                R3: uint8_t  *p_bgr888
+*                      pointer to the output BGR Line
+*                R12: uint32_t  length
+*                      width of Line
+*--------------------------------------------------------------------------
+* STACK ARG    : None
+*--------------------------------------------------------------------------
+* REG OUTPUT   : None
+*--------------------------------------------------------------------------
+* MEM INPUT    : p_y      - a line of Y pixels
+*                p_cr     - a line of Cr pixels
+*                p_cb     - a line of Cb pixels
+*                length   - the width of the input line
+*--------------------------------------------------------------------------
+* MEM OUTPUT   : p_bgr888 - the converted bgr pixels
+*--------------------------------------------------------------------------
+* REG AFFECTED : ARM:  R0-R4, R12
+*                NEON: Q0-Q15
+*--------------------------------------------------------------------------
+* STACK USAGE  : none
+*--------------------------------------------------------------------------
+* CYCLES       : none
+*
+*--------------------------------------------------------------------------
+* NOTES        :
+*--------------------------------------------------------------------------
+*/
+.type yyvup2bgr888_venum, %function
+yyvup2bgr888_venum:
+    /*-------------------------------------------------------------------------
+     *  Store stack registers
+     * ------------------------------------------------------------------------ */
+    STMFD SP!, {LR}
+
+    VPUSH {D8-D15}
+
+    PLD [R0, R3]                       @ preload luma line
+
+    ADR   R12, constants
+
+    VLD1.S16  {D6, D7}, [R12]!         @ D6, D7: 359 | -88 | -183 | 454 | 256 | 0 | 255 | 0
+    VLD1.S32  {D30, D31}, [R12]        @ Q15   :  -45824   |    34816   |  -57984 |     X
+
+    /*-------------------------------------------------------------------------
+     *  Load the 5th parameter via stack
+     *  R0 ~ R3 are used to pass the first 4 parameters, the 5th and above
+     *  parameters are passed via stack
+     * ------------------------------------------------------------------------ */
+    LDR R12, [SP, #68]                 @ LR is pushed into the stack so SP is
+                                       @ decreased by 4,
+                                       @ D8-D15 are also pushed into the stack
+                                       @ so SP is decreased by
+                                       @ 8-byte/D-Register * 8 D-Registers = 64,
+                                       @ so SP needs to be increased by 64+4=68
+                                       @ to get the value that was first pushed
+                                       @ into stack (the 5th parameter passed in
+                                       @ throught stack)
+
+    /*-------------------------------------------------------------------------
+     *  Load clamping parameters to duplicate vector elements
+     * ------------------------------------------------------------------------ */
+    VDUP.S16  Q4,  D7[1]               @ Q4:  0  |  0  |  0  |  0  |  0  |  0  |  0  |  0
+    VDUP.S16  Q5,  D7[2]               @ Q5: 255 | 255 | 255 | 255 | 255 | 255 | 255 | 255
+
+    /*-------------------------------------------------------------------------
+     *  Read bias
+     * ------------------------------------------------------------------------ */
+    VDUP.S32  Q0,   D30[0]             @ Q0:  -45824 | -45824 | -45824 | -45824
+    VDUP.S32  Q1,   D30[1]             @ Q1:   34816 |  34816 |  34816 |  34816
+    VDUP.S32  Q2,   D31[0]             @ Q2:  -70688 | -70688 | -70688 | -70688
+
+
+    /*-------------------------------------------------------------------------
+     *  The main loop
+     * ------------------------------------------------------------------------ */
+loop_yyvup2bgr888:
+
+    /*-------------------------------------------------------------------------
+     *  Load input from Y, V and U
+     *  D12, D13: Y0 Y2 Y4 Y6 Y8 Y10 Y12 Y14, Y1 Y3 Y5 Y7 Y9 Y11 Y13 Y15
+     *  D14  : V0  V1  V2  V3  V4  V5  V6  V7
+     *  D15  : U0  U1  U2  U3  U4  U5  U6  U7
+     * ------------------------------------------------------------------------ */
+    VLD2.U8  {D12,D13}, [p_y]!         @ Load 16 Luma elements (uint8) to D12, D13
+    VLD1.U8  {D14},  [p_cr]!           @ Load 8 Cr elements (uint8) to D14
+    VLD1.U8  {D15},  [p_cb]!           @ Load 8 Cb elements (uint8) to D15
+
+    /*-------------------------------------------------------------------------
+     *  Expand uint8 value to uint16
+     *  D24, D25: Y0 Y2 Y4 Y6 Y8 Y10 Y12 Y14
+     *  D26, D27: Y1 Y3 Y5 Y7 Y9 Y11 Y13 Y15
+     *  D28, D29: V0 V1 V2 V3 V4 V5  V6  V7
+     *  D30, D31: U0 U1 U2 U3 U4 U5  U6  U7
+     * ------------------------------------------------------------------------ */
+    VMOVL.U8 Q12, D12
+    VMOVL.U8 Q13, D13
+    VMOVL.U8 Q14, D14
+    VMOVL.U8 Q15, D15
+
+    /*-------------------------------------------------------------------------
+     *  Multiply contribution from chrominance, results are in 32-bit
+     * ------------------------------------------------------------------------ */
+    VMULL.S16  Q6, D28, D6[0]          @ Q6:  359*(V0,V1,V2,V3)     Red
+    VMULL.S16  Q7, D30, D6[1]          @ Q7: -88*(U0,U1,U2,U3)     Green
+    VMLAL.S16  Q7, D28, D6[2]          @ q7: -88*(U0,U1,U2,U3) - 183*(V0,V1,V2,V3)
+    VMULL.S16  Q8, D30, D6[3]          @ q8:  454*(U0,U1,U2,U3)     Blue
+
+    /*-------------------------------------------------------------------------
+     *  Add bias
+     * ------------------------------------------------------------------------ */
+    VADD.S32  Q6, Q0                   @ Q6 add Red   bias -45824
+    VADD.S32  Q7, Q1                   @ Q7 add Green bias  34816
+    VADD.S32  Q8, Q2                   @ Q8 add Blue  bias -57984
+
+    /*-------------------------------------------------------------------------
+     *  Calculate Red, Green, Blue
+     * ------------------------------------------------------------------------ */
+    VMOV.S32   Q9, Q6
+    VMLAL.S16  Q6, D24, D7[0]          @ Q6: R0, R2, R4, R6 in 32-bit Q8 format
+    VMLAL.S16  Q9, D26, D7[0]          @ Q9: R1, R3, R5, R7 in 32-bit Q8 format
+
+    VMOV.S32   Q10, Q7
+    VMLAL.S16  Q7,  D24, D7[0]         @ Q7:  G0, G2, G4, G6 in 32-bit Q8 format
+    VMLAL.S16  Q10, D26, D7[0]         @ Q10: G1, G3, G5, G7 in 32-bit Q8 format
+
+    VMOV.S32   Q11, Q8
+    VMLAL.S16  Q8,  D24, D7[0]         @ Q8:  B0, B2, B4, B6 in 32-bit Q8 format
+    VMLAL.S16  Q11, D26, D7[0]         @ Q11: B1, B3, B5, B7 in 32-bit Q8 format
+
+    /*-------------------------------------------------------------------------
+     *  Right shift eight bits with rounding
+     * ------------------------------------------------------------------------ */
+    VSHRN.S32   D12, Q6,  #8           @ D12: R0 R2 R4 R6 in 16-bit Q0 format
+    VSHRN.S32   D13, Q9,  #8           @ D13: R1 R3 R5 R7 in 16-bit Q0 format
+    VZIP.16     D12, D13               @ Q6 : R0 R1 R2 R3 R4 R5 R6 R7
+
+    VSHRN.S32   D18, Q7,  #8           @ D18: G0 G2 G4 G6 in 16-bit Q0 format
+    VSHRN.S32   D19, Q10, #8           @ D19: G1 G3 G5 G7 in 16-bit Q0 format
+    VZIP.16     D18, D19               @ Q9 : G0 G1 G2 G3 G4 G5 G6 G7
+
+    VSHRN.S32   D20, Q8,  #8           @ D20: B0 B2 B4 B6 in 16-bit Q0 format
+    VSHRN.S32   D21, Q11, #8           @ D21: B1 B3 B5 B7 in 16-bit Q0 format
+    VZIP.16     D20, D21               @ Q10: B0 B1 B2 B3 B4 B5 B6 B7
+
+    /*-------------------------------------------------------------------------
+     *  Clamp the value to be within [0~255]
+     * ------------------------------------------------------------------------ */
+    VMAX.S16  Q10, Q10, Q4             @ if Q10 <   0, Q10 =   0
+    VMIN.S16  Q10, Q10, Q5             @ if Q10 > 255, Q10 = 255
+    VQMOVUN.S16   D23, Q10             @ store Blue to D23, narrow the value from int16 to int8
+
+    VMAX.S16  Q9, Q9, Q4               @ if Q9 <   0, Q9 =   0
+    VMIN.S16  Q9, Q9, Q5               @ if Q9 > 255, Q9 = 255
+    VQMOVUN.S16  D22, Q9               @ store Green to D22, narrow the value from int16 to int8
+
+    VMAX.S16  Q6, Q6, Q4               @ if Q6 <   0, Q6 =   0
+    VMIN.S16  Q6, Q6, Q5               @ if Q6 > 255, Q6 = 255
+    VQMOVUN.S16  D21, Q6               @ store Red to D21, narrow the value from int16 to int8
+
+    SUBS length, length, #8            @ check if the length is less than 8
+
+    BMI  trailing_yyvup2bgr888         @ jump to trailing processing if remaining length is less than 8
+
+    VST3.U8  {D21,D22,D23}, [p_bgr]!   @ vector store Blue, Green, Red to destination
+                                       @ Red at LSB
+
+    BEQ  end_yyvup2bgr888              @ done if exactly 8 pixel processed in the loop
+
+    /*-------------------------------------------------------------------------
+     *  Done with the first 8 elements, continue on the next 8 elements
+     * ------------------------------------------------------------------------ */
+
+    /*-------------------------------------------------------------------------
+     *  Multiply contribution from chrominance, results are in 32-bit
+     * ------------------------------------------------------------------------ */
+    VMULL.S16  Q6, D29, D6[0]          @ Q6: 359*(V4,V5,V6,V7)       Red
+    VMULL.S16  Q7, D31, D6[1]          @ Q7: -88*(U4,U5,U6,U7)      Green
+    VMLAL.S16  Q7, D29, D6[2]          @ Q7: -88*(U4,U5,U6,U7) - 183*(V4,V5,V6,V7)
+    VMULL.S16  Q8, D31, D6[3]          @ Q8: 454*(U4,U5,U6,U7)       Blue
+
+    /*-------------------------------------------------------------------------
+     *  Add bias
+     * ------------------------------------------------------------------------ */
+    VADD.S32  Q6, Q0                   @ Q6 add Red   bias -45824
+    VADD.S32  Q7, Q1                   @ Q7 add Green bias  34816
+    VADD.S32  Q8, Q2                   @ Q8 add Blue  bias -70688
+
+    /*-------------------------------------------------------------------------
+     *  Calculate Red, Green, Blue
+     * ------------------------------------------------------------------------ */
+    VMOV.S32   Q9, Q6
+    VMLAL.S16  Q6, D25, D7[0]          @ Q6: R8 R10 R12 R14 in 32-bit Q8 format
+    VMLAL.S16  Q9, D27, D7[0]          @ Q9: R9 R11 R13 R15 in 32-bit Q8 format
+
+    VMOV.S32   Q10, Q7
+    VMLAL.S16  Q7,  D25, D7[0]         @ Q7: G0, G2, G4, G6 in 32-bit Q8 format
+    VMLAL.S16  Q10, D27, D7[0]         @ Q10 : G1, G3, G5, G7 in 32-bit Q8 format
+
+    VMOV.S32   Q11, Q8
+    VMLAL.S16  Q8,  D25, D7[0]         @ Q8: B0, B2, B4, B6 in 32-bit Q8 format
+    VMLAL.S16  Q11, D27, D7[0]         @ Q11 : B1, B3, B5, B7 in 32-bit Q8 format
+
+    /*-------------------------------------------------------------------------
+     *  Right shift eight bits with rounding
+     * ------------------------------------------------------------------------ */
+    VSHRN.S32   D12, Q6,  #8           @ D12: R8 R10 R12 R14 in 16-bit Q0 format
+    VSHRN.S32   D13, Q9,  #8           @ D13: R9 R11 R13 R15 in 16-bit Q0 format
+    VZIP.16     D12, D13               @ Q6: R8 R9 R10 R11 R12 R13 R14 R15
+
+    VSHRN.S32   D18, Q7,  #8           @ D18: G8 G10 G12 G14 in 16-bit Q0 format
+    VSHRN.S32   D19, Q10, #8           @ D19: G9 G11 G13 G15 in 16-bit Q0 format
+    VZIP.16     D18, D19               @ Q9:  G8 G9 G10 G11 G12 G13 G14 G15
+
+    VSHRN.S32   D20, Q8,  #8           @ D20: B8 B10 B12 B14 in 16-bit Q0 format
+    VSHRN.S32   D21, Q11, #8           @ D21: B9 B11 B13 B15 in 16-bit Q0 format
+    VZIP.16     D20, D21               @ Q10: B8 B9 B10 B11 B12 B13 B14 B15
+
+    /*-------------------------------------------------------------------------
+     *  Clamp the value to be within [0~255]
+     * ------------------------------------------------------------------------ */
+    VMAX.S16  Q10, Q10, Q4             @ if Q10 <   0, Q10 =   0
+    VMIN.S16  Q10, Q10, Q5             @ if Q10 > 255, Q10 = 255
+    VQMOVUN.S16   D23, Q10             @ store Blue to D23, narrow the value from int16 to int8
+
+    VMAX.S16  Q9, Q9, Q4               @ if Q9 <   0, Q9 =   0
+    VMIN.S16  Q9, Q9, Q5               @ if Q9 > 255, Q9 = 255
+    VQMOVUN.S16  D22, Q9               @ store Green to D22, narrow the value from int16 to int8
+
+    VMAX.S16  Q6, Q6, Q4               @ if Q6 <   0, Q6 =   0
+    VMIN.S16  Q6, Q6, Q5               @ if Q6 > 255, Q6 = 255
+    VQMOVUN.S16  D21, Q6               @ store Red to D21, narrow the value from int16 to int8
+
+
+    SUBS length, length, #8            @ check if the length is less than 8
+
+    BMI  trailing_yyvup2bgr888         @ jump to trailing processing if remaining length is less than 8
+
+    VST3.U8  {D21,D22,D23}, [p_bgr]!   @ vector store Blue, Green, Red to destination
+                                       @ Red at LSB
+
+    BHI loop_yyvup2bgr888              @ loop if more than 8 pixels left
+
+    BEQ  end_yyvup2bgr888              @ done if exactly 8 pixel processed in the loop
+
+
+trailing_yyvup2bgr888:
+    /*-------------------------------------------------------------------------
+     *  There are from 1 ~ 7 pixels left in the trailing part.
+     *  First adding 7 to the length so the length would be from 0 ~ 6.
+     *  eg: 1 pixel left in the trailing part, so 1-8+7 = 0.
+     *  Then save 1 pixel unconditionally since at least 1 pixels left in the
+     *  trailing part.
+     * ------------------------------------------------------------------------ */
+    ADDS length, length, #7            @ there are 7 or less in the trailing part
+
+    VST3.U8 {D21[0],D22[0],D23[0]}, [p_bgr]! @ at least 1 pixel left in the trailing part
+    BEQ end_yyvup2bgr888               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST3.U8 {D21[1],D22[1],D23[1]}, [p_bgr]!  @ store one more pixel
+    BEQ end_yyvup2bgr888               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST3.U8 {D21[2],D22[2],D23[2]}, [p_bgr]!  @ store one more pixel
+    BEQ end_yyvup2bgr888               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST3.U8 {D21[3],D22[3],D23[3]}, [p_bgr]!  @ store one more pixel
+    BEQ end_yyvup2bgr888               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST3.U8 {D21[4],D22[4],D23[4]}, [p_bgr]!  @ store one more pixel
+    BEQ end_yyvup2bgr888               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST3.U8 {D21[5],D22[5],D23[5]}, [p_bgr]!  @ store one more pixel
+    BEQ end_yyvup2bgr888               @ done if 0 pixel left
+
+    SUBS length, length, #1            @ update length counter
+    VST3.U8 {D21[6],D22[6],D23[6]}, [p_bgr]!  @ store one more pixel
+
+end_yyvup2bgr888:
+    VPOP  {D8-D15}
+    LDMFD SP!, {PC}
+
+                                       @ end of yyvup2bgr888
+
+.end
diff --git a/libjpegtwrp/asm/armv7/jdidct-armv7.S b/libjpegtwrp/asm/armv7/jdidct-armv7.S
new file mode 100644
index 000000000..d61e219f7
--- /dev/null
+++ b/libjpegtwrp/asm/armv7/jdidct-armv7.S
@@ -0,0 +1,762 @@
+/*=========================================================================
+* jdidct-armv7.s
+*
+*  Copyright (c) 2010, Code Aurora Forum. All rights reserved.
+*
+*  Redistribution and use in source and binary forms, with or without
+*  modification, are permitted provided that the following conditions are
+*  met:
+*      * Redistributions of source code must retain the above copyright
+*        notice, this list of conditions and the following disclaimer.
+*      * Redistributions in binary form must reproduce the above
+*        copyright notice, this list of conditions and the following
+*        disclaimer in the documentation and/or other materials provided
+*        with the distribution.
+*      * Neither the name of Code Aurora Forum, Inc. nor the names of its
+*        contributors may be used to endorse or promote products derived
+*        from this software without specific prior written permission.
+*
+*  THIS SOFTWARE IS PROVIDED "AS IS" AND ANY EXPRESS OR IMPLIED
+*  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+*  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT
+*  ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS
+*  BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+*  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+*  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
+*  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
+*  WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
+*  OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN
+*  IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*==========================================================================
+
+*==========================================================================
+*                         FUNCTION LIST
+*--------------------------------------------------------------------------
+* - idct_1x1_venum
+* - idct_2x2_venum
+* - idct_4x4_venum
+* - idct_8x8_venum
+*
+*==========================================================================
+*/
+
+@==========================================================================
+@ MACRO DEFINITION
+@==========================================================================
+    .macro Transpose8x8
+        @==================================================================
+        @ Transpose an 8 x 8 x 16 bit matrix in place
+        @ Input: q8 to q15
+        @ Output: q8 to q15
+        @ Registers used: q8 to q15
+        @ Assumptions: 8 x 8 x 16 bit data
+        @==================================================================
+
+        vswp d17, d24                  @q8, q12
+        vswp d23, d30                  @q11, q15
+        vswp d21, d28                  @q10, q14
+        vswp d19, d26                  @q9, q13
+
+        vtrn.32 q8,  q10
+        vtrn.32 q9,  q11
+        vtrn.32 q12, q14
+        vtrn.32 q13, q15
+
+        vtrn.16 q8,  q9
+        vtrn.16 q10, q11
+        vtrn.16 q12, q13
+        vtrn.16 q14, q15
+    .endm
+
+    .macro IDCT1D
+        @==================================================================
+        @ One dimensional 64 element inverse DCT
+        @ Input: q8 to q15 loaded with data
+        @        q0 loaded with constants
+        @ Output: q8 to q15
+        @ Registers used: q0, q4 to q15
+        @ Assumptions: 16 bit data, first elements in least significant
+        @ halfwords
+        @==================================================================
+
+        @1st stage
+        vqrdmulh.s16 q4,  q15, d0[2]   @q4 = a1*vx7
+        vqrdmulh.s16 q5,  q9,  d0[2]   @q5 = a1*vx1
+        vqrdmulh.s16 q6,  q13, d0[3]   @q6 = a2*vx5
+        vqrdmulh.s16 q7,  q11, d1[1]   @q7 = ma2*vx3
+        vqrdmulh.s16 q2,  q14, d0[1]   @q6 = a0*vx6
+        vqrdmulh.s16 q3,  q10, d0[1]   @q7 = a0*vx2
+        vqadd.s16   q9,  q4,  q9       @q9 = t1 = a1*vx7 + vx1
+        vqsub.s16   q5,  q5,  q15      @q5 = t8 = a1*vx1 - vx7
+        vqadd.s16   q15, q6,  q11      @q15 = t7 = a2*vx5 + vx3
+        vqadd.s16   q11, q7,  q13      @q11 = t3 = ma2*vx3 + vx5
+
+        @2nd stage
+        vqadd.s16   q13, q8,  q12      @q13 = t5 = vx0 + vx4
+        vqsub.s16   q8,  q8,  q12      @q8 = t0 = vx0 - vx4
+        vqadd.s16   q10, q2,  q10      @q10 = t2 = a0*vx6 + vx2
+        vqsub.s16   q12, q3,  q14      @q12 = t4 = a0*vx2 - vx6
+        vqadd.s16   q14, q5,  q11      @q14 = t6 = t8 + t3
+        vqsub.s16   q11, q5,  q11      @q11 = t3 = t8 - t3
+        vqsub.s16   q5,  q9,  q15      @q5 = t8 = t1 - t7
+        vqadd.s16   q9,  q9,  q15      @q9 = t1 = t1 + t7
+
+        @3rd stage
+        vqadd.s16   q15, q13, q10      @q15 = t7 = t5 + t2
+        vqsub.s16   q10, q13, q10      @q10 = t2 = t5 - t2
+        vqadd.s16   q13, q8,  q12      @q13 = t5 = t0 + t4
+        vqsub.s16   q7,  q8,  q12      @q7 = t0 = t0 - t4
+        vqsub.s16   q12, q5,  q11      @q12 = t4 = t8 - t3
+        vqadd.s16   q11, q5,  q11      @q11 = t3 = t8 + t3
+
+        @4th stage
+        vqadd.s16   q8,  q15, q9       @q8 = vy0 = t7 + t1
+        vqsub.s16   q15, q15, q9       @q15 = vy7 = t7 - t1
+        vqrdmulh.s16 q6,  q12, d0[0]   @q6 = c4*t4
+        vqrdmulh.s16 q4,  q11, d0[0]   @q4 = c4*t3
+        vqsub.s16   q12, q10, q14      @q12 = vy4 = t2 - t6
+        vqadd.s16   q11, q10, q14      @q11 = vy3 = t2 + t6
+        vqadd.s16   q10, q7,  q6       @q10 = vy2 = t0 + c4*t4
+        vqsub.s16   q14, q13, q4       @q14 = vy6 = t5 - c4*t3
+        vqadd.s16   q9,  q13, q4       @q9 = vy1 = t5 + c4*t3
+        vqsub.s16   q13, q7,  q6       @q13 = vy5 = t0 - c4*t4
+    .endm
+
+    .macro PART1
+        @==================================================================
+        @ Load input input data from memory and shift
+        @==================================================================
+        vld1.16   {d16, d17},[r0]!     @q8 =row0
+        vqshl.s16  q8,  q8,  #4        @Input data too big?!!
+                                       @Maximum MPEG input is 2047/-2048.
+        vld1.16   {d18, d19},[r0]!     @q9 =row1
+        vqshl.s16  q9,  q9,  #4        @Shift 1 instead of 4
+
+        vld1.16   {d20, d21},[r0]!     @q10=row2
+        vqshl.s16  q10, q10, #4
+
+        vld1.16   {d22, d23},[r0]!     @q11=row3
+        vqshl.s16  q11, q11, #4
+
+        vld1.16   {d24, d25},[r0]!     @q12=row4
+        vqshl.s16  q12, q12, #4
+
+        vld1.16   {d26, d27},[r0]!     @q13=row5
+        vqshl.s16  q13, q13, #4
+        vld1.16   {d28, d29},[r0]!     @q14=row6
+        vqshl.s16  q14, q14, #4
+        vld1.16   {d30, d31},[r0]!     @q15=row7
+        vqshl.s16  q15, q15, #4
+
+        @==================================================================
+        @ refresh the constants that was clobbered last time through IDCT1D
+        @==================================================================
+        vld1.16   {d4, d5},[r7]        @q2 =constants[2]
+        vld1.16   {d6, d7},[r8]        @q3 =constants[3]
+        vld1.16   {d8, d9},[r9]        @q4 =constants[4]
+    .endm
+
+    .macro PART2
+        @==================================================================
+        @ Prescale the input
+        @==================================================================
+        vqrdmulh.s16 q12, q12, q1      @q12=row4 * constants[1] = vx4
+        vqrdmulh.s16 q15, q15, q2      @q15=row7 * constants[2] = vx7
+        vqrdmulh.s16 q9,  q9,  q2      @q9 =row1 * constants[2] = vx1
+        vqrdmulh.s16 q13, q13, q4      @q13=row5 * constants[4] = vx5
+        vqrdmulh.s16 q11, q11, q4      @q11=row3 * constants[4] = vx3
+        vqrdmulh.s16 q14, q14, q3      @q14=row6 * constants[3] = vx6
+        vqrdmulh.s16 q10, q10, q3      @q10=row2 * constants[3] = vx2
+        vqrdmulh.s16 q8,  q8,  q1      @q8 =row0 * constants[1] = vx0
+
+        @==================================================================
+        @ At thsi point, the input 8x8 x 16 bit coefficients are
+        @ transposed, prescaled, and loaded in q8 to q15
+        @ q0 loaded with scalar constants
+        @ Perform 1D IDCT
+        @==================================================================
+        IDCT1D                         @perform 1d idct
+
+        @==================================================================
+        @ Transpose the intermediate results to get read for vertical
+        @ transformation
+        @==================================================================
+        vswp d17, d24                  @q8, q12
+        vswp d23, d30                  @q11, q15
+        vswp d21, d28                  @q10, q14
+        vswp d19, d26                  @q9, q13
+
+        @==================================================================
+        @ Load the bias
+        @==================================================================
+        vdup.32 q4, d1[1]              @a cycle is saved by loading
+                                       @the bias at this point
+
+        @==================================================================
+        @ Finish the transposition
+        @==================================================================
+        vtrn.32 q8,  q10
+        vtrn.32 q9,  q11
+        vtrn.32 q12, q14
+        vtrn.32 q13, q15
+        vtrn.16 q8,  q9
+        vtrn.16 q10, q11
+        vtrn.16 q12, q13
+        vtrn.16 q14, q15
+
+        @==================================================================
+        @ Add bias
+        @==================================================================
+        vqadd.s16 q8, q8, q4
+
+        @==================================================================
+        @ IDCT 2nd half
+        @==================================================================
+        IDCT1D                         @perform 1d dct
+
+        @==================================================================
+        @ Scale and clamp the output to correct range and save to memory
+        @ 1. scale to 8bits by right shift 6
+        @ 2. clamp output to [0, 255] by min/max
+        @ 3. use multiple store. Each store will save one row of output.
+        @    The st queue size is 4, so do no more than 4 str in sequence.
+        @==================================================================
+        ldr       r5, =constants+5*16  @constants[5],
+        vld1.16   d10, [r5]            @load clamping parameters
+        vdup.s16  q6,  d10[0]          @q6=[0000000000000000]
+        vdup.s16  q7,  d10[1]          @q7=[FFFFFFFFFFFFFFFF]
+
+        @Save the results
+        vshr.s16  q8,  q8,  #6         @q8 = vy0
+        vmax.s16  q8,  q8,  q6         @clamp >0
+        vmin.s16  q8,  q8,  q7         @clamp <255
+
+        vshr.s16  q9,  q9,  #6         @q9 = vy1
+        vmax.s16  q9,  q9,  q6         @clamp >0
+        vmin.s16  q9,  q9,  q7         @clamp <255
+
+        vshr.s16  q10, q10, #6         @q10 = vy2
+        vmax.s16  q10, q10, q6         @clamp >0
+        vmin.s16  q10, q10, q7         @clamp <255
+
+        vshr.s16  q11, q11, #6         @q11 = vy3
+        vmax.s16  q11, q11, q6         @clamp >0
+        vmin.s16  q11, q11, q7         @clamp <255
+
+        vst1.16  {d16, d17},[r1],r2    @q8 =row0
+        vst1.16  {d18, d19},[r1],r2    @q9 =row1
+        vst1.16  {d20, d21},[r1],r2    @q10=row2
+        vst1.16  {d22, d23},[r1],r2    @q11=row3
+
+        vshr.s16  q12, q12, #6         @q12 = vy4
+        vmax.s16  q12, q12, q6         @clamp >0
+        vmin.s16  q12, q12, q7         @clamp <255
+
+        vshr.s16  q13, q13, #6         @q13 = vy5
+        vmax.s16  q13, q13, q6         @clamp >0
+        vmin.s16  q13, q13, q7         @clamp <255
+
+        vshr.s16  q14, q14, #6         @q14 = vy6
+        vmax.s16  q14, q14, q6         @clamp >0
+        vmin.s16  q14, q14, q7         @clamp <255
+
+        vshr.s16  q15, q15, #6         @q15 = vy7
+        vmax.s16  q15, q15, q6         @clamp >0
+        vmin.s16  q15, q15, q7         @clamp <255
+
+        vst1.16  {d24, d25},[r1],r2    @q12=row4
+        vst1.16  {d26, d27},[r1],r2    @q13=row5
+        vst1.16  {d28, d29},[r1],r2    @q14=row6
+        vst1.16  {d30, d31},[r1]       @q15=row7
+    .endm
+
+    .macro BIG_BODY_TRANSPOSE_INPUT
+        @==================================================================
+        @ Main body of idct
+        @==================================================================
+        PART1
+        Transpose8x8
+        PART2
+    .endm
+
+    .macro IDCT_ENTRY
+        @==================================================================
+        @ Load the locations of the constants
+        @==================================================================
+        ldr  r5,  =constants+0*16      @constants[0]
+        ldr  r6,  =constants+1*16      @constants[1]
+        ldr  r7,  =constants+2*16      @constants[2]
+        ldr  r8,  =constants+3*16      @constants[3]
+        ldr  r9,  =constants+4*16      @constants[4]
+
+        @==================================================================
+        @ Load the coefficients
+        @ only some input coefficients are load due to register constrain
+        @==================================================================
+        vld1.16   {d0, d1},[r5]        @q0 =constants[0] (scalars)
+        vld1.16   {d2, d3},[r6]        @q1 =constants[1]
+    .endm
+@==========================================================================
+@ END of MACRO DEFINITION
+@==========================================================================
+
+
+    .section idct_func, "x"            @ ARE
+    .text                              @ idct_func, CODE, READONLY
+    .align 2
+    .code 32                           @ CODE32
+
+@==========================================================================
+@ Main Routine
+@==========================================================================
+
+    .global idct_1x1_venum
+    .global idct_2x2_venum
+    .global idct_4x4_venum
+    .global idct_8x8_venum
+
+@==========================================================================
+@ FUNCTION     : idct_1x1_venum
+@--------------------------------------------------------------------------
+@ DISCRIPTION  : ARM optimization of one 1x1 block iDCT
+@--------------------------------------------------------------------------
+@ C PROTOTYPE  : void idct_1x1_venum(int16 * input,
+@                                    int16 * output,
+@                                    int32 stride)
+@--------------------------------------------------------------------------
+@ REG INPUT    : R0 pointer to input (int16)
+@                R1 pointer to output (int16)
+@                R2 block stride
+@--------------------------------------------------------------------------
+@ STACK ARG    : None
+@--------------------------------------------------------------------------
+@ MEM INPUT    : None
+@--------------------------------------------------------------------------
+@ REG OUTPUT   : None
+@--------------------------------------------------------------------------
+@ MEM OUTPUT   : None
+@--------------------------------------------------------------------------
+@ REG AFFECTED : R0 - R2
+@--------------------------------------------------------------------------
+@ STACK USAGE  : none
+@--------------------------------------------------------------------------
+@ CYCLES       : 17 cycles
+@--------------------------------------------------------------------------
+@ NOTES        :
+@ This idct_1x1_venum code was developed with ARM instruction set.
+@
+@ ARM REGISTER ALLOCATION
+@ =========================================================================
+@ r0  : pointer to input data
+@ r1  : pointer to output area
+@ r2  : stride in the output buffer
+@==========================================================================
+.type idct_1x1_venum, %function
+idct_1x1_venum:
+
+    ldrsh   r3, [r0]                   @ Load signed half word (int16)
+    ldr     r2, =1028                  @ 1028 = 4 + 128 << 3
+                                       @ 4 for rounding, 128 for offset
+    add     r2, r3, r2
+    asrs    r2, r2, #3                 @ Divide by 8, and set status bit
+    movmi   r2, #0                     @ Clamp to be greater than 0
+    cmp     r2, #255
+    movgt   r2, #255                   @ Clamp to be less than 255
+    str     r2, [r1]                   @ Save output
+    bx      lr                         @ Return to caller
+
+                                       @ end of idct_1x1_venum
+
+
+@==========================================================================
+@ FUNCTION     : idct_2x2_venum
+@--------------------------------------------------------------------------
+@ DISCRIPTION  : VeNum optimization of one 2x2 block iDCT
+@--------------------------------------------------------------------------
+@ C PROTOTYPE  : void idct_2x2_venum(int16 * input,
+@                                    int16 * output,
+@                                    int32 stride)
+@--------------------------------------------------------------------------
+@ REG INPUT    : R0 pointer to input (int16)
+@                R1 pointer to output (int16)
+@                R2 block stride
+@--------------------------------------------------------------------------
+@ STACK ARG    : None
+@--------------------------------------------------------------------------
+@ MEM INPUT    : None
+@--------------------------------------------------------------------------
+@ REG OUTPUT   : None
+@--------------------------------------------------------------------------
+@ MEM OUTPUT   : None
+@--------------------------------------------------------------------------
+@ REG AFFECTED : R0 - R2
+@--------------------------------------------------------------------------
+@ STACK USAGE  : none
+@--------------------------------------------------------------------------
+@ CYCLES       : 27 cycles
+@--------------------------------------------------------------------------
+@ NOTES        : Output buffer must be an 8x8 16-bit buffer
+@
+@ ARM REGISTER ALLOCATION
+@ ==========================================
+@ r0  : pointer to input data
+@ r1  : pointer to output area
+@ r2  : stride in the output buffer
+@ -------------------------------------------
+@
+@ VENUM REGISTER ALLOCATION
+@ =================================================
+@ q0     : output x0 - x4
+@ q1     : not used
+@ q2     : not used
+@ q3     : not used
+@ q4     : not used
+@ q5     : not used
+@ q6     : not used
+@ q7     : not used
+@ q8     : input y0 - y4
+@ q9     : intermediate value
+@ q10    : intermediate value
+@ q11    : offset value
+@ q12    : clamp value
+@ q13    : not used
+@ q14    : not used
+@ q15    : not used
+@==========================================================================
+.type idct_2x2_venum, %function
+idct_2x2_venum:
+
+    vld4.32    {d16, d17, d18, d19}, [r0]
+                                       @  d16: y0 | y1 | y2 | y3  (LSB | MSB)
+
+    vtrn.32    d16, d17                @  d16: y0 | y1 | X | X
+                                       @  d17: y2 | y3 | X | X
+
+    vqadd.s16  d18, d16, d17           @ d18: y0+y2 | y1+y3 | X | X   q: saturated
+    vqsub.s16  d19, d16, d17           @ d19: y0-y2 | y1-y3 | X | X   q: saturated
+
+    vtrn.16    d18, d19                @ d18: y0+y2 | y0-y2 | X | X
+                                       @ d19: y1+y3 | y1-y3 | X | X
+
+    vqadd.s16  d20, d18, d19           @ d20: (y0+y2)+(y1+y3) | (y0-y2)+(y1-y3)
+                                       @       x0 | x2 | X | X
+    vqsub.s16  d21, d18, d19           @ d21: (y0+y2)-(y1+y3) | (y0-y2)-(y1-y3)
+                                       @       x1 | x3 | X | X
+
+    vtrn.16    d20, d21                @ d20:  x0 | x1 | X | X
+                                       @ d21:  x2 | x3 | X | X
+
+    vrshr.s16  q10, q10, #3               @ Divide by 8
+
+    vmov.i16   q11, #128               @ q11 = 128|128|128|128|128|128|128|128
+    vqadd.s16  q0, q10, q11            @ Add offset to make output in [0,255]
+
+    vmov.i16   q12, #0                   @ q12 = [0000000000000000]
+    vmov.i16   q13, #255               @ q13 = [FFFFFFFFFFFFFFFF] (hex)
+
+    vmax.s16   q0, q0, q12             @ Clamp > 0
+    vmin.s16   q0, q0, q13             @ Clamp < 255
+
+    vstr       d0, [r1]                @ Store  x0 | x1 | X | X
+                                       @ Potential out of boundary issue
+    add        r1, r1, r2              @ Add the offset to the output pointer
+    vstr       d1, [r1]                @ Store  x2 | x3 | X | X
+                                       @ Potential out of boundary issue
+    bx         lr                      @ Return to caller
+
+                                       @ end of idct_2x2_venum
+
+
+@==========================================================================
+@ FUNCTION     : idct_4x4_venum
+@--------------------------------------------------------------------------
+@ DISCRIPTION  : VeNum optimization of one 4x4 block iDCT
+@--------------------------------------------------------------------------
+@ C PROTOTYPE  : void idct_4x4_venum(int16 * input,
+@                                    int16 * output,
+@                                    int32 stride)
+@--------------------------------------------------------------------------
+@ REG INPUT    : R0 pointer to input (int16)
+@                R1 pointer to output (int16)
+@                R2 block stride
+@--------------------------------------------------------------------------
+@ STACK ARG    : None
+@--------------------------------------------------------------------------
+@ MEM INPUT    : None
+@--------------------------------------------------------------------------
+@ REG OUTPUT   : None
+@--------------------------------------------------------------------------
+@ MEM OUTPUT   : None
+@--------------------------------------------------------------------------
+@ REG AFFECTED : R0 - R3, R12
+@--------------------------------------------------------------------------
+@ STACK USAGE  : none
+@--------------------------------------------------------------------------
+@ CYCLES       : 56 cycles
+@--------------------------------------------------------------------------
+@ NOTES        :
+@
+@ ARM REGISTER ALLOCATION
+@ ==========================================
+@ r0  : pointer to input data
+@ r1  : pointer to output area
+@ r2  : stride in the output buffer
+@ r3  : pointer to the coefficient set
+@ r12 : pointer to the coefficient set
+@ -------------------------------------------
+@
+@ VENUM REGISTER ALLOCATION
+@ =================================================
+@ q0     : coefficients[0]
+@ q1     : coefficients[1]
+@ q2     : coefficients[2]
+@ q3     : coefficients[3]
+@ q4     : not used
+@ q5     : not used
+@ q6     : not used
+@ q7     : not used
+@ q8     : input y0 - y7
+@ q9     : input y8 - y15
+@ q10    : intermediate value
+@ q11    : intermediate value
+@ q12    : intermediate value
+@ q13    : intermediate value
+@ q14    : intermediate value
+@ q15    : intermediate value
+@==========================================================================
+.type idct_4x4_venum, %function
+idct_4x4_venum:
+
+        @ Load the locations of the first 2 sets of coefficients
+        ldr  r3,   =coefficient+0*16   @ coefficient[0]
+        ldr  r12,  =coefficient+1*16   @ coefficient[1]
+
+        @ Load the first 2 sets of coefficients
+        vld1.16  {d0, d1},[r3]         @ q0 = C4 | C2 | C4 | C6 | C4 | C2 | C4 | C6
+        vld1.16  {d2, d3},[r12]        @ q1 = C4 | C6 | C4 | C2 | C4 | C6 | C4 | C2
+
+        @ Load the locations of the second 2 sets of coefficients
+        ldr  r3,   =coefficient+2*16   @ coefficient[2]
+        ldr  r12,  =coefficient+3*16   @ coefficient[3]
+
+        @ Load the second 2 sets of coefficients
+        vld1.16  {d4, d5},[r3]         @ q2 = C4 | C4 | C4 | C4 | C2 | C2 | C2 | C2
+        vld1.16  {d6, d7},[r12]        @ q3 = C4 | C4 | C4 | C4 | C6 | C6 | C6 | C6
+
+        @ Load the input values
+        vld1.16  {d16}, [r0], r2       @ d16:   y0  | y1  | y2  | y3  (LSB | MSB)
+        vld1.16  {d17}, [r0], r2       @ d17:   y4  | y5  | y6  | y7  (LSB | MSB)
+        vld1.16  {d18}, [r0], r2       @ d18:   y8  | y9  | y10 | y11 (LSB | MSB)
+        vld1.16  {d19}, [r0], r2       @ d19:   y12 | y13 | y14 | y15 (LSB | MSB)
+
+        @ Apply iDCT Horizonally
+
+        @ q8: y0 |y1 |y2 |y3 |y4 |y5 |y6 |y7
+        @ q9: y8 |y9 |y10|y11|y12|y13|y14|y15
+
+        @======================================================================
+        @ vqrdmulh doubles the result and save the high 16 bits of the result,
+        @ this is equivalent to right shift by 15 bits.
+        @ since coefficients are in Q15 format, it contradicts with the right
+        @ shift 15 here, so the final result is in Q0 format
+        @
+        @ vqrdmulh will also round the result
+        @======================================================================
+
+        vqrdmulh.s16  q10, q8, q0      @ q10: C4*y0  | C2*y1  | C4*y2  | C6*y3  | C4*y4  | C2*y5  | C4*y6  | C6*y7
+        vqrdmulh.s16  q11, q8, q1      @ q11: C4*y0  | C6*y1  | C4*y2  | C2*y3  | C4*y4  | C6*y5  | C4*y6  | C2*y7
+
+        vqrdmulh.s16  q12, q9, q0      @ q12: C4*y8  | C2*y9  | C4*y10 | C6*y11 | C4*y12 | C2*y13 | C4*y14 | C6*y15
+        vqrdmulh.s16  q13, q9, q1      @ q13: C4*y8  | C6*y9  | C4*y10 | C2*y11 | C4*y12 | C6*y13 | C4*y14 | C2*y15
+
+        vtrn.32       q10, q12         @ q10: C4*y0  | C2*y1  | C4*y8  | C2*y9  | C4*y4  | C2*y5  | C4*y12 | C2*y13
+                                       @ q12: C4*y2  | C6*y3  | C4*y10 | C6*y11 | C4*y6  | C6*y7  | C4*y14 | C6*y15
+
+        vtrn.32       q11, q13         @ q11: C4*y0  | C6*y1  | C4*y8  | C6*y9  | C4*y4  | C6*y5  | C4*y12 | C6*y13
+                                       @ q13: C4*y2  | C2*y3  | C4*y10 | C2*y11 | C4*y6  | C2*y7  | C4*y14 | C2*y15
+
+        vqadd.s16     q14, q10, q12    @ q14: C4*y0 + C4*y2 | C2*y1 + C6*y3 | C4*y8 + C4*y10 | C2*y9 + C6*y11 | C4*y4 + C4*y6 | C2*y5 + C6*y7 | C4*y12 + C4*y14 | C2*y13 + C6*y15
+                                       @       S0 | S2 | S8 | S10 | S4 | S6 | S12 | S14
+
+        vqsub.s16     q15, q11, q13    @ q15: C4*y0 - C4*y2 | C6*y1 - C2*y3 | C4*y8 - C4*y10 | C6*y9 - C2*y11 | C4*y4 - C4*y6 | C6*y5 - C2*y7 | C4*y12 - C4*y14 | C6*y13 - C2*y15
+                                       @       S1 | S3 | S9 | S11 | S5 | S7 | S13 | S15
+
+        vtrn.16       q14, q15         @ q14: S0 | S1 | S8  | S9  | S4 | S5 | S12 | S13
+                                       @ q15: S2 | S3 | S10 | S11 | S6 | S7 | S14 | S15
+
+        vqadd.s16     q8, q14, q15     @ q8:  Z0 | Z1 | Z8  | Z9  | Z4 | Z5 | Z12 | Z13
+        vqsub.s16     q9, q14, q15     @ q9:  Z3 | Z2 | Z11 | Z10 | Z7 | Z6 | Z15 | Z14
+        vrev32.16     q9, q9           @ q9:  Z2 | Z3 | Z10 | Z11 | Z6 | Z7 | Z14 | Z15
+
+
+        @ Apply iDCT Vertically
+
+        vtrn.32       q8, q9           @ q8:  Z0 | Z1 | Z2  | Z3  | Z4  | Z5  | Z6  | Z7
+                                       @ q9:  Z8 | Z9 | Z10 | Z11 | Z12 | Z13 | Z14 | Z15
+
+
+        vqrdmulh.s16  q10, q8, q2      @ q10: C4*Z0 | C4*Z1 | C4*Z2 | C4*Z3 | C2*Z4 | C2*Z5 | C2*Z6 | C2*Z7
+        vqrdmulh.s16  q11, q8, q3      @ q11: C4*Z0 | C4*Z1 | C4*Z2 | C4*Z3 | C6*Z4 | C6*Z5 | C6*Z6 | C6*Z7
+
+        vqrdmulh.s16  q12, q9, q2      @ q12: C4*Z8 | C4*Z9 | C4*Z10 | C4*Z11 | C2*Z12 | C2*Z13 | C2*Z14 | C2*Z15
+        vqrdmulh.s16  q13, q9, q3      @ q13: C4*Z8 | C4*Z9 | C4*Z10 | C4*Z11 | C6*Z12 | C6*Z13 | C6*Z14 | C6*Z15
+
+        vqadd.s16     q14, q10, q13    @ q14: C4*Z0+C4*Z8 | C4*Z1+C4*Z9 | C4*Z2+C4*Z10 | C4*Z3+C4*Z11 | C2*Z4+C6*Z12 | C2*Z5+C6*Z13 | C2*Z6+C6*Z14 | C2*Z7+C6*Z15
+                                       @      s0 | s4 | s8 | s12 | s2 | s6 | s10 | s14
+
+        vqsub.s16     q15, q11, q12    @ q15: C4*Z0-C4*Z8 | C4*Z1-C4*Z9 | C4*Z2-C4*Z10 | C4*Z3-C4*Z11 | C6*Z4-C2*Z12 | C6*Z5-C2*Z13 | C6*Z6-C2*Z14 | C6*Z7-C2*Z15
+                                       @      s1 | s5 | s9 | s13 | s3 | s7 | s11 | s15
+
+        vswp          d29, d30         @ q14: s0 | s4 | s8  | s12 | s1 | s5 | s9  | s13
+                                       @ q15: s2 | s6 | s10 | s14 | s3 | s7 | s11 | s15
+
+        vqadd.s16     q8, q14, q15     @ q8:  x0 | x4 | x8  | x12 | x1 | x5 | x9 | x13
+        vqsub.s16     q9, q14, q15     @ q9:  x3 | x7 | x11 | x15 | x2 | x6 | x10 | x14
+
+        vmov.i16      q10, #0           @ q10=[0000000000000000]
+        vmov.i16      q11, #255        @ q11=[FFFFFFFFFFFFFFFF] (hex)
+
+        vmov.i16      q0, #128         @ q0 = 128|128|128|128|128|128|128|128
+
+        vqadd.s16     q8, q8, q0       @ Add the offset
+        vqadd.s16     q9, q9, q0       @ Add the offset
+
+        vmax.s16      q8, q8, q10      @ clamp > 0
+        vmin.s16      q8, q8, q11      @ clamp < 255
+
+        vmax.s16      q9, q9, q10      @ clamp > 0
+        vmin.s16      q9, q9, q11      @ clamp < 255
+
+        vst1.16       {d16}, [r1], r2  @  d16:   x0 | x1  | x2  | x3  (LSB | MSB)
+        vst1.16       {d17}, [r1], r2  @  d17:   x4 | x5  | x6  | x7  (LSB | MSB)
+        vst1.16       {d19}, [r1], r2  @  d18:   x8 | x9  | x10 | x11 (LSB | MSB)
+        vst1.16       {d18}, [r1], r2  @  d19:   x12| x13 | x14 | x15 (LSB | MSB)
+
+        bx         lr                  @ Return to caller
+
+                                       @ end of idct_4x4_venum
+
+@==========================================================================
+@ FUNCTION     : idct_8x8_venum
+@--------------------------------------------------------------------------
+@ DISCRIPTION  : VeNum optimization of one 8x8 block iDCT
+@--------------------------------------------------------------------------
+@ C PROTOTYPE  : void idct_8x8_venum(int16 * input,
+@                                    int16 * output,
+@                                    int32 stride)
+@--------------------------------------------------------------------------
+@ REG INPUT    : R0 pointer to input (int16)
+@                R1 pointer to output (int16)
+@                R2 block stride
+@--------------------------------------------------------------------------
+@ STACK ARG    : None
+@--------------------------------------------------------------------------
+@ MEM INPUT    : None
+@--------------------------------------------------------------------------
+@ REG OUTPUT   : None
+@--------------------------------------------------------------------------
+@ MEM OUTPUT   : None
+@--------------------------------------------------------------------------
+@ REG AFFECTED : R0 - R9
+@--------------------------------------------------------------------------
+@ STACK USAGE  : none
+@--------------------------------------------------------------------------
+@ CYCLES       : 177 cycles
+@--------------------------------------------------------------------------
+@ NOTES        :
+@
+@ It was tested to be IEEE 1180 compliant. Since IEEE 1180 compliance is more stringent
+@ than MPEG-4 compliance, this version is also MPEG-4 compliant.
+@
+@ CODE STRUCTURE:
+@ (i)   Macros for transposing an 8x8 matrix and for configuring the VFP unit are defined.
+@ (ii)  Macro for IDCT in one dimension is defined as four stages
+@ (iii) The two dimensional code begins
+@ (iv)  constants are defined in the area DataArea
+@
+@ PROGRAM FLOW:
+@
+@ The VFP is configured
+@ The parameters to IDCT are loaded
+@ the coefficients are loaded
+@ loop:
+@    decrement loop counter
+@    The first input Matrix is loaded and pre-scaled
+@    The input is prescaled using the constants
+@    IDCT is performed in one dimension on the 8 columns
+@    The matrix is transposed
+@    A bias is loaded an added to the matrix
+@    IDCT is performed in one dimension on the 8 rows
+@    The matrix is post-scaled
+@    The matrix is saved
+@    test loop counter and loop if greater than zero
+@ stop
+@
+@
+@ ARM REGISTER ALLOCATION
+@ ==========================================
+@ r0 : pointer to input data
+@ r1 : pointer to output are
+@ r2 : stride in the output buffer
+@ r3 :
+@ r4 :
+@ r5 : pointer to constants[0] [5]
+@ r6 : pointer to constants[1]
+@ r7 : pointer to constants[2]
+@ r8 : pointer to constants[3]
+@ r9 : pointer to constants[4]
+@ -------------------------------------------
+@
+@ VENUM REGISTER ALLOCATION
+@ =================================================
+@ q0     : constants[0]
+@ q1     : constants[1]
+@ q2     : constants[2], IDCT1D in-place scratch
+@ q3     : constants[3], IDCT1D in-place scratch
+@ q4     : constants[4], IDCT1D in-place scratch, and bias compensation
+@ q5     :               IDCT1D in-place scratch
+@ q6     :               IDCT1D in-place scratch
+@ q7     :               IDCT1D in-place scratch
+@ q8     : Matrix[0]     IDCT1D in-place scratch
+@ q9     : Matrix[1]     IDCT1D in-place scratch
+@ q10    : Matrix[2]     IDCT1D in-place scratch
+@ q11    : Matrix[3]     IDCT1D in-place scratch
+@ q12    : Matrix[4]     IDCT1D in-place scratch
+@ q13    : Matrix[5]     IDCT1D in-place scratch
+@ q14    : Matrix[6]     IDCT1D in-place scratch
+@ q15    : Matrix[7]     IDCT1D in-place scratch
+@==========================================================================
+.type idct_8x8_venum, %function
+idct_8x8_venum:
+
+        push {r5-r9}
+        vpush {d8-d15}
+        IDCT_ENTRY
+        BIG_BODY_TRANSPOSE_INPUT
+        vpop {d8-d15}
+        pop  {r5-r9}
+        bx   lr
+                                       @ end of idct_8x8_venum
+
+@==========================================================================
+@ Constants Definition AREA: define idct kernel, bias
+@==========================================================================
+    .section ro_data_area              @ AREA  RODataArea
+    .data                              @ DATA, READONLY
+    .align 5                           @ ALIGN=5
+
+constants:
+        .hword  23170, 13573, 6518,  21895, -23170, -21895, 8223,  8224
+        .hword  16384, 22725, 21407, 19266, 16384,  19266,  21407, 22725
+        .hword  22725, 31521, 29692, 26722, 22725,  26722,  29692, 31521
+        .hword  21407, 29692, 27969, 25172, 21407,  25172,  27969, 29692
+        .hword  19266, 26722, 25172, 22654, 19266,  22654,  25172, 26722
+        .hword      0,   255,     0,     0
+
+coefficient:                           @ These are the coefficent used by 4x4 iDCT in Q15 format
+        .hword 11585, 15137,  11585,  6270, 11585, 15137,  11585,  6270  @ C4, C2, C4, C6, C4, C2, C4, C6 /2
+        .hword 11585,  6270,  11585, 15137, 11585,  6270,  11585, 15137  @ C4, C6, C4, C2, C4, C6, C4, C2 /2
+        .hword 11585, 11585,  11585, 11585, 15137, 15137,  15137, 15137  @ C4, C4, C4, C4, C2, C2, C2, C2 /2
+        .hword 11585, 11585,  11585, 11585,  6270,  6270,   6270,  6270  @ C4, C4, C4, C4, C6, C6, C6, C6 /2
+
+.end
diff --git a/libjpegtwrp/cderror.h b/libjpegtwrp/cderror.h
new file mode 100644
index 000000000..70435e161
--- /dev/null
+++ b/libjpegtwrp/cderror.h
@@ -0,0 +1,132 @@
+/*
+ * cderror.h
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file defines the error and message codes for the cjpeg/djpeg
+ * applications.  These strings are not needed as part of the JPEG library
+ * proper.
+ * Edit this file to add new codes, or to translate the message strings to
+ * some other language.
+ */
+
+/*
+ * To define the enum list of message codes, include this file without
+ * defining macro JMESSAGE.  To create a message string table, include it
+ * again with a suitable JMESSAGE definition (see jerror.c for an example).
+ */
+#ifndef JMESSAGE
+#ifndef CDERROR_H
+#define CDERROR_H
+/* First time through, define the enum list */
+#define JMAKE_ENUM_LIST
+#else
+/* Repeated inclusions of this file are no-ops unless JMESSAGE is defined */
+#define JMESSAGE(code,string)
+#endif /* CDERROR_H */
+#endif /* JMESSAGE */
+
+#ifdef JMAKE_ENUM_LIST
+
+typedef enum {
+
+#define JMESSAGE(code,string)	code ,
+
+#endif /* JMAKE_ENUM_LIST */
+
+JMESSAGE(JMSG_FIRSTADDONCODE=1000, NULL) /* Must be first entry! */
+
+#ifdef BMP_SUPPORTED
+JMESSAGE(JERR_BMP_BADCMAP, "Unsupported BMP colormap format")
+JMESSAGE(JERR_BMP_BADDEPTH, "Only 8- and 24-bit BMP files are supported")
+JMESSAGE(JERR_BMP_BADHEADER, "Invalid BMP file: bad header length")
+JMESSAGE(JERR_BMP_BADPLANES, "Invalid BMP file: biPlanes not equal to 1")
+JMESSAGE(JERR_BMP_COLORSPACE, "BMP output must be grayscale or RGB")
+JMESSAGE(JERR_BMP_COMPRESSED, "Sorry, compressed BMPs not yet supported")
+JMESSAGE(JERR_BMP_NOT, "Not a BMP file - does not start with BM")
+JMESSAGE(JTRC_BMP, "%ux%u 24-bit BMP image")
+JMESSAGE(JTRC_BMP_MAPPED, "%ux%u 8-bit colormapped BMP image")
+JMESSAGE(JTRC_BMP_OS2, "%ux%u 24-bit OS2 BMP image")
+JMESSAGE(JTRC_BMP_OS2_MAPPED, "%ux%u 8-bit colormapped OS2 BMP image")
+#endif /* BMP_SUPPORTED */
+
+#ifdef GIF_SUPPORTED
+JMESSAGE(JERR_GIF_BUG, "GIF output got confused")
+JMESSAGE(JERR_GIF_CODESIZE, "Bogus GIF codesize %d")
+JMESSAGE(JERR_GIF_COLORSPACE, "GIF output must be grayscale or RGB")
+JMESSAGE(JERR_GIF_IMAGENOTFOUND, "Too few images in GIF file")
+JMESSAGE(JERR_GIF_NOT, "Not a GIF file")
+JMESSAGE(JTRC_GIF, "%ux%ux%d GIF image")
+JMESSAGE(JTRC_GIF_BADVERSION,
+	 "Warning: unexpected GIF version number '%c%c%c'")
+JMESSAGE(JTRC_GIF_EXTENSION, "Ignoring GIF extension block of type 0x%02x")
+JMESSAGE(JTRC_GIF_NONSQUARE, "Caution: nonsquare pixels in input")
+JMESSAGE(JWRN_GIF_BADDATA, "Corrupt data in GIF file")
+JMESSAGE(JWRN_GIF_CHAR, "Bogus char 0x%02x in GIF file, ignoring")
+JMESSAGE(JWRN_GIF_ENDCODE, "Premature end of GIF image")
+JMESSAGE(JWRN_GIF_NOMOREDATA, "Ran out of GIF bits")
+#endif /* GIF_SUPPORTED */
+
+#ifdef PPM_SUPPORTED
+JMESSAGE(JERR_PPM_COLORSPACE, "PPM output must be grayscale or RGB")
+JMESSAGE(JERR_PPM_NONNUMERIC, "Nonnumeric data in PPM file")
+JMESSAGE(JERR_PPM_NOT, "Not a PPM/PGM file")
+JMESSAGE(JTRC_PGM, "%ux%u PGM image")
+JMESSAGE(JTRC_PGM_TEXT, "%ux%u text PGM image")
+JMESSAGE(JTRC_PPM, "%ux%u PPM image")
+JMESSAGE(JTRC_PPM_TEXT, "%ux%u text PPM image")
+#endif /* PPM_SUPPORTED */
+
+#ifdef RLE_SUPPORTED
+JMESSAGE(JERR_RLE_BADERROR, "Bogus error code from RLE library")
+JMESSAGE(JERR_RLE_COLORSPACE, "RLE output must be grayscale or RGB")
+JMESSAGE(JERR_RLE_DIMENSIONS, "Image dimensions (%ux%u) too large for RLE")
+JMESSAGE(JERR_RLE_EMPTY, "Empty RLE file")
+JMESSAGE(JERR_RLE_EOF, "Premature EOF in RLE header")
+JMESSAGE(JERR_RLE_MEM, "Insufficient memory for RLE header")
+JMESSAGE(JERR_RLE_NOT, "Not an RLE file")
+JMESSAGE(JERR_RLE_TOOMANYCHANNELS, "Cannot handle %d output channels for RLE")
+JMESSAGE(JERR_RLE_UNSUPPORTED, "Cannot handle this RLE setup")
+JMESSAGE(JTRC_RLE, "%ux%u full-color RLE file")
+JMESSAGE(JTRC_RLE_FULLMAP, "%ux%u full-color RLE file with map of length %d")
+JMESSAGE(JTRC_RLE_GRAY, "%ux%u grayscale RLE file")
+JMESSAGE(JTRC_RLE_MAPGRAY, "%ux%u grayscale RLE file with map of length %d")
+JMESSAGE(JTRC_RLE_MAPPED, "%ux%u colormapped RLE file with map of length %d")
+#endif /* RLE_SUPPORTED */
+
+#ifdef TARGA_SUPPORTED
+JMESSAGE(JERR_TGA_BADCMAP, "Unsupported Targa colormap format")
+JMESSAGE(JERR_TGA_BADPARMS, "Invalid or unsupported Targa file")
+JMESSAGE(JERR_TGA_COLORSPACE, "Targa output must be grayscale or RGB")
+JMESSAGE(JTRC_TGA, "%ux%u RGB Targa image")
+JMESSAGE(JTRC_TGA_GRAY, "%ux%u grayscale Targa image")
+JMESSAGE(JTRC_TGA_MAPPED, "%ux%u colormapped Targa image")
+#else
+JMESSAGE(JERR_TGA_NOTCOMP, "Targa support was not compiled")
+#endif /* TARGA_SUPPORTED */
+
+JMESSAGE(JERR_BAD_CMAP_FILE,
+	 "Color map file is invalid or of unsupported format")
+JMESSAGE(JERR_TOO_MANY_COLORS,
+	 "Output file format cannot handle %d colormap entries")
+JMESSAGE(JERR_UNGETC_FAILED, "ungetc failed")
+#ifdef TARGA_SUPPORTED
+JMESSAGE(JERR_UNKNOWN_FORMAT,
+	 "Unrecognized input file format --- perhaps you need -targa")
+#else
+JMESSAGE(JERR_UNKNOWN_FORMAT, "Unrecognized input file format")
+#endif
+JMESSAGE(JERR_UNSUPPORTED_FORMAT, "Unsupported output file format")
+
+#ifdef JMAKE_ENUM_LIST
+
+  JMSG_LASTADDONCODE
+} ADDON_MESSAGE_CODE;
+
+#undef JMAKE_ENUM_LIST
+#endif /* JMAKE_ENUM_LIST */
+
+/* Zap JMESSAGE macro so that future re-inclusions do nothing by default */
+#undef JMESSAGE
diff --git a/libjpegtwrp/cdjpeg.c b/libjpegtwrp/cdjpeg.c
new file mode 100644
index 000000000..b6250ff97
--- /dev/null
+++ b/libjpegtwrp/cdjpeg.c
@@ -0,0 +1,181 @@
+/*
+ * cdjpeg.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains common support routines used by the IJG application
+ * programs (cjpeg, djpeg, jpegtran).
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+#include <ctype.h>		/* to declare isupper(), tolower() */
+#ifdef NEED_SIGNAL_CATCHER
+#include <signal.h>		/* to declare signal() */
+#endif
+#ifdef USE_SETMODE
+#include <fcntl.h>		/* to declare setmode()'s parameter macros */
+/* If you have setmode() but not <io.h>, just delete this line: */
+#include <io.h>			/* to declare setmode() */
+#endif
+
+
+/*
+ * Signal catcher to ensure that temporary files are removed before aborting.
+ * NB: for Amiga Manx C this is actually a global routine named _abort();
+ * we put "#define signal_catcher _abort" in jconfig.h.  Talk about bogus...
+ */
+
+#ifdef NEED_SIGNAL_CATCHER
+
+static j_common_ptr sig_cinfo;
+
+void				/* must be global for Manx C */
+signal_catcher (int signum)
+{
+  if (sig_cinfo != NULL) {
+    if (sig_cinfo->err != NULL) /* turn off trace output */
+      sig_cinfo->err->trace_level = 0;
+    jpeg_destroy(sig_cinfo);	/* clean up memory allocation & temp files */
+  }
+  exit(EXIT_FAILURE);
+}
+
+
+GLOBAL(void)
+enable_signal_catcher (j_common_ptr cinfo)
+{
+  sig_cinfo = cinfo;
+#ifdef SIGINT			/* not all systems have SIGINT */
+  signal(SIGINT, signal_catcher);
+#endif
+#ifdef SIGTERM			/* not all systems have SIGTERM */
+  signal(SIGTERM, signal_catcher);
+#endif
+}
+
+#endif
+
+
+/*
+ * Optional progress monitor: display a percent-done figure on stderr.
+ */
+
+#ifdef PROGRESS_REPORT
+
+METHODDEF(void)
+progress_monitor (j_common_ptr cinfo)
+{
+  cd_progress_ptr prog = (cd_progress_ptr) cinfo->progress;
+  int total_passes = prog->pub.total_passes + prog->total_extra_passes;
+  int percent_done = (int) (prog->pub.pass_counter*100L/prog->pub.pass_limit);
+
+  if (percent_done != prog->percent_done) {
+    prog->percent_done = percent_done;
+    if (total_passes > 1) {
+      fprintf(stderr, "\rPass %d/%d: %3d%% ",
+	      prog->pub.completed_passes + prog->completed_extra_passes + 1,
+	      total_passes, percent_done);
+    } else {
+      fprintf(stderr, "\r %3d%% ", percent_done);
+    }
+    fflush(stderr);
+  }
+}
+
+
+GLOBAL(void)
+start_progress_monitor (j_common_ptr cinfo, cd_progress_ptr progress)
+{
+  /* Enable progress display, unless trace output is on */
+  if (cinfo->err->trace_level == 0) {
+    progress->pub.progress_monitor = progress_monitor;
+    progress->completed_extra_passes = 0;
+    progress->total_extra_passes = 0;
+    progress->percent_done = -1;
+    cinfo->progress = &progress->pub;
+  }
+}
+
+
+GLOBAL(void)
+end_progress_monitor (j_common_ptr cinfo)
+{
+  /* Clear away progress display */
+  if (cinfo->err->trace_level == 0) {
+    fprintf(stderr, "\r                \r");
+    fflush(stderr);
+  }
+}
+
+#endif
+
+
+/*
+ * Case-insensitive matching of possibly-abbreviated keyword switches.
+ * keyword is the constant keyword (must be lower case already),
+ * minchars is length of minimum legal abbreviation.
+ */
+
+GLOBAL(boolean)
+keymatch (char * arg, const char * keyword, int minchars)
+{
+  register int ca, ck;
+  register int nmatched = 0;
+
+  while ((ca = *arg++) != '\0') {
+    if ((ck = *keyword++) == '\0')
+      return FALSE;		/* arg longer than keyword, no good */
+    if (isupper(ca))		/* force arg to lcase (assume ck is already) */
+      ca = tolower(ca);
+    if (ca != ck)
+      return FALSE;		/* no good */
+    nmatched++;			/* count matched characters */
+  }
+  /* reached end of argument; fail if it's too short for unique abbrev */
+  if (nmatched < minchars)
+    return FALSE;
+  return TRUE;			/* A-OK */
+}
+
+
+/*
+ * Routines to establish binary I/O mode for stdin and stdout.
+ * Non-Unix systems often require some hacking to get out of text mode.
+ */
+
+GLOBAL(FILE *)
+read_stdin (void)
+{
+  FILE * input_file = stdin;
+
+#ifdef USE_SETMODE		/* need to hack file mode? */
+  setmode(fileno(stdin), O_BINARY);
+#endif
+#ifdef USE_FDOPEN		/* need to re-open in binary mode? */
+  if ((input_file = fdopen(fileno(stdin), READ_BINARY)) == NULL) {
+    fprintf(stderr, "Cannot reopen stdin\n");
+    exit(EXIT_FAILURE);
+  }
+#endif
+  return input_file;
+}
+
+
+GLOBAL(FILE *)
+write_stdout (void)
+{
+  FILE * output_file = stdout;
+
+#ifdef USE_SETMODE		/* need to hack file mode? */
+  setmode(fileno(stdout), O_BINARY);
+#endif
+#ifdef USE_FDOPEN		/* need to re-open in binary mode? */
+  if ((output_file = fdopen(fileno(stdout), WRITE_BINARY)) == NULL) {
+    fprintf(stderr, "Cannot reopen stdout\n");
+    exit(EXIT_FAILURE);
+  }
+#endif
+  return output_file;
+}
diff --git a/libjpegtwrp/cdjpeg.h b/libjpegtwrp/cdjpeg.h
new file mode 100644
index 000000000..2b387b6e5
--- /dev/null
+++ b/libjpegtwrp/cdjpeg.h
@@ -0,0 +1,184 @@
+/*
+ * cdjpeg.h
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains common declarations for the sample applications
+ * cjpeg and djpeg.  It is NOT used by the core JPEG library.
+ */
+
+#define JPEG_CJPEG_DJPEG	/* define proper options in jconfig.h */
+#define JPEG_INTERNAL_OPTIONS	/* cjpeg.c,djpeg.c need to see xxx_SUPPORTED */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"		/* get library error codes too */
+#include "cderror.h"		/* get application-specific error codes */
+
+
+/*
+ * Object interface for cjpeg's source file decoding modules
+ */
+
+typedef struct cjpeg_source_struct * cjpeg_source_ptr;
+
+struct cjpeg_source_struct {
+  JMETHOD(void, start_input, (j_compress_ptr cinfo,
+			      cjpeg_source_ptr sinfo));
+  JMETHOD(JDIMENSION, get_pixel_rows, (j_compress_ptr cinfo,
+				       cjpeg_source_ptr sinfo));
+  JMETHOD(void, finish_input, (j_compress_ptr cinfo,
+			       cjpeg_source_ptr sinfo));
+
+  FILE *input_file;
+
+  JSAMPARRAY buffer;
+  JDIMENSION buffer_height;
+};
+
+
+/*
+ * Object interface for djpeg's output file encoding modules
+ */
+
+typedef struct djpeg_dest_struct * djpeg_dest_ptr;
+
+struct djpeg_dest_struct {
+  /* start_output is called after jpeg_start_decompress finishes.
+   * The color map will be ready at this time, if one is needed.
+   */
+  JMETHOD(void, start_output, (j_decompress_ptr cinfo,
+			       djpeg_dest_ptr dinfo));
+  /* Emit the specified number of pixel rows from the buffer. */
+  JMETHOD(void, put_pixel_rows, (j_decompress_ptr cinfo,
+				 djpeg_dest_ptr dinfo,
+				 JDIMENSION rows_supplied));
+  /* Finish up at the end of the image. */
+  JMETHOD(void, finish_output, (j_decompress_ptr cinfo,
+				djpeg_dest_ptr dinfo));
+
+  /* Target file spec; filled in by djpeg.c after object is created. */
+  FILE * output_file;
+
+  /* Output pixel-row buffer.  Created by module init or start_output.
+   * Width is cinfo->output_width * cinfo->output_components;
+   * height is buffer_height.
+   */
+  JSAMPARRAY buffer;
+  JDIMENSION buffer_height;
+};
+
+
+/*
+ * cjpeg/djpeg may need to perform extra passes to convert to or from
+ * the source/destination file format.  The JPEG library does not know
+ * about these passes, but we'd like them to be counted by the progress
+ * monitor.  We use an expanded progress monitor object to hold the
+ * additional pass count.
+ */
+
+struct cdjpeg_progress_mgr {
+  struct jpeg_progress_mgr pub;	/* fields known to JPEG library */
+  int completed_extra_passes;	/* extra passes completed */
+  int total_extra_passes;	/* total extra */
+  /* last printed percentage stored here to avoid multiple printouts */
+  int percent_done;
+};
+
+typedef struct cdjpeg_progress_mgr * cd_progress_ptr;
+
+
+/* Short forms of external names for systems with brain-damaged linkers. */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jinit_read_bmp		jIRdBMP
+#define jinit_write_bmp		jIWrBMP
+#define jinit_read_gif		jIRdGIF
+#define jinit_write_gif		jIWrGIF
+#define jinit_read_ppm		jIRdPPM
+#define jinit_write_ppm		jIWrPPM
+#define jinit_read_rle		jIRdRLE
+#define jinit_write_rle		jIWrRLE
+#define jinit_read_targa	jIRdTarga
+#define jinit_write_targa	jIWrTarga
+#define read_quant_tables	RdQTables
+#define read_scan_script	RdScnScript
+#define set_quant_slots		SetQSlots
+#define set_sample_factors	SetSFacts
+#define read_color_map		RdCMap
+#define enable_signal_catcher	EnSigCatcher
+#define start_progress_monitor	StProgMon
+#define end_progress_monitor	EnProgMon
+#define read_stdin		RdStdin
+#define write_stdout		WrStdout
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+/* Module selection routines for I/O modules. */
+
+EXTERN(cjpeg_source_ptr) jinit_read_bmp JPP((j_compress_ptr cinfo));
+EXTERN(djpeg_dest_ptr) jinit_write_bmp JPP((j_decompress_ptr cinfo,
+					    boolean is_os2));
+EXTERN(cjpeg_source_ptr) jinit_read_gif JPP((j_compress_ptr cinfo));
+EXTERN(djpeg_dest_ptr) jinit_write_gif JPP((j_decompress_ptr cinfo));
+EXTERN(cjpeg_source_ptr) jinit_read_ppm JPP((j_compress_ptr cinfo));
+EXTERN(djpeg_dest_ptr) jinit_write_ppm JPP((j_decompress_ptr cinfo));
+EXTERN(cjpeg_source_ptr) jinit_read_rle JPP((j_compress_ptr cinfo));
+EXTERN(djpeg_dest_ptr) jinit_write_rle JPP((j_decompress_ptr cinfo));
+EXTERN(cjpeg_source_ptr) jinit_read_targa JPP((j_compress_ptr cinfo));
+EXTERN(djpeg_dest_ptr) jinit_write_targa JPP((j_decompress_ptr cinfo));
+
+/* cjpeg support routines (in rdswitch.c) */
+
+EXTERN(boolean) read_quant_tables JPP((j_compress_ptr cinfo, char * filename,
+				    int scale_factor, boolean force_baseline));
+EXTERN(boolean) read_scan_script JPP((j_compress_ptr cinfo, char * filename));
+EXTERN(boolean) set_quant_slots JPP((j_compress_ptr cinfo, char *arg));
+EXTERN(boolean) set_sample_factors JPP((j_compress_ptr cinfo, char *arg));
+
+/* djpeg support routines (in rdcolmap.c) */
+
+EXTERN(void) read_color_map JPP((j_decompress_ptr cinfo, FILE * infile));
+
+/* common support routines (in cdjpeg.c) */
+
+EXTERN(void) enable_signal_catcher JPP((j_common_ptr cinfo));
+EXTERN(void) start_progress_monitor JPP((j_common_ptr cinfo,
+					 cd_progress_ptr progress));
+EXTERN(void) end_progress_monitor JPP((j_common_ptr cinfo));
+EXTERN(boolean) keymatch JPP((char * arg, const char * keyword, int minchars));
+EXTERN(FILE *) read_stdin JPP((void));
+EXTERN(FILE *) write_stdout JPP((void));
+
+/* miscellaneous useful macros */
+
+#ifdef DONT_USE_B_MODE		/* define mode parameters for fopen() */
+#define READ_BINARY	"r"
+#define WRITE_BINARY	"w"
+#else
+#ifdef VMS			/* VMS is very nonstandard */
+#define READ_BINARY	"rb", "ctx=stm"
+#define WRITE_BINARY	"wb", "ctx=stm"
+#else				/* standard ANSI-compliant case */
+#define READ_BINARY	"rb"
+#define WRITE_BINARY	"wb"
+#endif
+#endif
+
+#ifndef EXIT_FAILURE		/* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+#ifndef EXIT_SUCCESS
+#ifdef VMS
+#define EXIT_SUCCESS  1		/* VMS is very nonstandard */
+#else
+#define EXIT_SUCCESS  0
+#endif
+#endif
+#ifndef EXIT_WARNING
+#ifdef VMS
+#define EXIT_WARNING  1		/* VMS is very nonstandard */
+#else
+#define EXIT_WARNING  2
+#endif
+#endif
diff --git a/libjpegtwrp/change.log b/libjpegtwrp/change.log
new file mode 100644
index 000000000..74102c0db
--- /dev/null
+++ b/libjpegtwrp/change.log
@@ -0,0 +1,217 @@
+CHANGE LOG for Independent JPEG Group's JPEG software
+
+
+Version 6b  27-Mar-1998
+-----------------------
+
+jpegtran has new features for lossless image transformations (rotation
+and flipping) as well as "lossless" reduction to grayscale.
+
+jpegtran now copies comments by default; it has a -copy switch to enable
+copying all APPn blocks as well, or to suppress comments.  (Formerly it
+always suppressed comments and APPn blocks.)  jpegtran now also preserves
+JFIF version and resolution information.
+
+New decompressor library feature: COM and APPn markers found in the input
+file can be saved in memory for later use by the application.  (Before,
+you had to code this up yourself with a custom marker processor.)
+
+There is an unused field "void * client_data" now in compress and decompress
+parameter structs; this may be useful in some applications.
+
+JFIF version number information is now saved by the decoder and accepted by
+the encoder.  jpegtran uses this to copy the source file's version number,
+to ensure "jpegtran -copy all" won't create bogus files that contain JFXX
+extensions but claim to be version 1.01.  Applications that generate their
+own JFXX extension markers also (finally) have a supported way to cause the
+encoder to emit JFIF version number 1.02.
+
+djpeg's trace mode reports JFIF 1.02 thumbnail images as such, rather
+than as unknown APP0 markers.
+
+In -verbose mode, djpeg and rdjpgcom will try to print the contents of
+APP12 markers as text.  Some digital cameras store useful text information
+in APP12 markers.
+
+Handling of truncated data streams is more robust: blocks beyond the one in
+which the error occurs will be output as uniform gray, or left unchanged
+if decoding a progressive JPEG.  The appearance no longer depends on the
+Huffman tables being used.
+
+Huffman tables are checked for validity much more carefully than before.
+
+To avoid the Unisys LZW patent, djpeg's GIF output capability has been
+changed to produce "uncompressed GIFs", and cjpeg's GIF input capability
+has been removed altogether.  We're not happy about it either, but there
+seems to be no good alternative.
+
+The configure script now supports building libjpeg as a shared library
+on many flavors of Unix (all the ones that GNU libtool knows how to
+build shared libraries for).  Use "./configure --enable-shared" to
+try this out.
+
+New jconfig file and makefiles for Microsoft Visual C++ and Developer Studio.
+Also, a jconfig file and a build script for Metrowerks CodeWarrior
+on Apple Macintosh.  makefile.dj has been updated for DJGPP v2, and there
+are miscellaneous other minor improvements in the makefiles.
+
+jmemmac.c now knows how to create temporary files following Mac System 7
+conventions.
+
+djpeg's -map switch is now able to read raw-format PPM files reliably.
+
+cjpeg -progressive -restart no longer generates any unnecessary DRI markers.
+
+Multiple calls to jpeg_simple_progression for a single JPEG object
+no longer leak memory.
+
+
+Version 6a  7-Feb-96
+--------------------
+
+Library initialization sequence modified to detect version mismatches
+and struct field packing mismatches between library and calling application.
+This change requires applications to be recompiled, but does not require
+any application source code change.
+
+All routine declarations changed to the style "GLOBAL(type) name ...",
+that is, GLOBAL, LOCAL, METHODDEF, EXTERN are now macros taking the
+routine's return type as an argument.  This makes it possible to add
+Microsoft-style linkage keywords to all the routines by changing just
+these macros.  Note that any application code that was using these macros
+will have to be changed.
+
+DCT coefficient quantization tables are now stored in normal array order
+rather than zigzag order.  Application code that calls jpeg_add_quant_table,
+or otherwise manipulates quantization tables directly, will need to be
+changed.  If you need to make such code work with either older or newer
+versions of the library, a test like "#if JPEG_LIB_VERSION >= 61" is
+recommended.
+
+djpeg's trace capability now dumps DQT tables in natural order, not zigzag
+order.  This allows the trace output to be made into a "-qtables" file
+more easily.
+
+New system-dependent memory manager module for use on Apple Macintosh.
+
+Fix bug in cjpeg's -smooth option: last one or two scanlines would be
+duplicates of the prior line unless the image height mod 16 was 1 or 2.
+
+Repair minor problems in VMS, BCC, MC6 makefiles.
+
+New configure script based on latest GNU Autoconf.
+
+Correct the list of include files needed by MetroWerks C for ccommand().
+
+Numerous small documentation updates.
+
+
+Version 6  2-Aug-95
+-------------------
+
+Progressive JPEG support: library can read and write full progressive JPEG
+files.  A "buffered image" mode supports incremental decoding for on-the-fly
+display of progressive images.  Simply recompiling an existing IJG-v5-based
+decoder with v6 should allow it to read progressive files, though of course
+without any special progressive display.
+
+New "jpegtran" application performs lossless transcoding between different
+JPEG formats; primarily, it can be used to convert baseline to progressive
+JPEG and vice versa.  In support of jpegtran, the library now allows lossless
+reading and writing of JPEG files as DCT coefficient arrays.  This ability
+may be of use in other applications.
+
+Notes for programmers:
+* We changed jpeg_start_decompress() to be able to suspend; this makes all
+decoding modes available to suspending-input applications.  However,
+existing applications that use suspending input will need to be changed
+to check the return value from jpeg_start_decompress().  You don't need to
+do anything if you don't use a suspending data source.
+* We changed the interface to the virtual array routines: access_virt_array
+routines now take a count of the number of rows to access this time.  The
+last parameter to request_virt_array routines is now interpreted as the
+maximum number of rows that may be accessed at once, but not necessarily
+the height of every access.
+
+
+Version 5b  15-Mar-95
+---------------------
+
+Correct bugs with grayscale images having v_samp_factor > 1.
+
+jpeg_write_raw_data() now supports output suspension.
+
+Correct bugs in "configure" script for case of compiling in
+a directory other than the one containing the source files.
+
+Repair bug in jquant1.c: sometimes didn't use as many colors as it could.
+
+Borland C makefile and jconfig file work under either MS-DOS or OS/2.
+
+Miscellaneous improvements to documentation.
+
+
+Version 5a  7-Dec-94
+--------------------
+
+Changed color conversion roundoff behavior so that grayscale values are
+represented exactly.  (This causes test image files to change.)
+
+Make ordered dither use 16x16 instead of 4x4 pattern for a small quality
+improvement.
+
+New configure script based on latest GNU Autoconf.
+Fix configure script to handle CFLAGS correctly.
+Rename *.auto files to *.cfg, so that configure script still works if
+file names have been truncated for DOS.
+
+Fix bug in rdbmp.c: didn't allow for extra data between header and image.
+
+Modify rdppm.c/wrppm.c to handle 2-byte raw PPM/PGM formats for 12-bit data.
+
+Fix several bugs in rdrle.c.
+
+NEED_SHORT_EXTERNAL_NAMES option was broken.
+
+Revise jerror.h/jerror.c for more flexibility in message table.
+
+Repair oversight in jmemname.c NO_MKTEMP case: file could be there
+but unreadable.
+
+
+Version 5  24-Sep-94
+--------------------
+
+Version 5 represents a nearly complete redesign and rewrite of the IJG
+software.  Major user-visible changes include:
+  * Automatic configuration simplifies installation for most Unix systems.
+  * A range of speed vs. image quality tradeoffs are supported.
+    This includes resizing of an image during decompression: scaling down
+    by a factor of 1/2, 1/4, or 1/8 is handled very efficiently.
+  * New programs rdjpgcom and wrjpgcom allow insertion and extraction
+    of text comments in a JPEG file.
+
+The application programmer's interface to the library has changed completely.
+Notable improvements include:
+  * We have eliminated the use of callback routines for handling the
+    uncompressed image data.  The application now sees the library as a
+    set of routines that it calls to read or write image data on a
+    scanline-by-scanline basis.
+  * The application image data is represented in a conventional interleaved-
+    pixel format, rather than as a separate array for each color channel.
+    This can save a copying step in many programs.
+  * The handling of compressed data has been cleaned up: the application can
+    supply routines to source or sink the compressed data.  It is possible to
+    suspend processing on source/sink buffer overrun, although this is not
+    supported in all operating modes.
+  * All static state has been eliminated from the library, so that multiple
+    instances of compression or decompression can be active concurrently.
+  * JPEG abbreviated datastream formats are supported, ie, quantization and
+    Huffman tables can be stored separately from the image data.
+  * And not only that, but the documentation of the library has improved
+    considerably!
+
+
+The last widely used release before the version 5 rewrite was version 4A of
+18-Feb-93.  Change logs before that point have been discarded, since they
+are not of much interest after the rewrite.
diff --git a/libjpegtwrp/cjpeg.1 b/libjpegtwrp/cjpeg.1
new file mode 100644
index 000000000..d175a961c
--- /dev/null
+++ b/libjpegtwrp/cjpeg.1
@@ -0,0 +1,292 @@
+.TH CJPEG 1 "20 March 1998"
+.SH NAME
+cjpeg \- compress an image file to a JPEG file
+.SH SYNOPSIS
+.B cjpeg
+[
+.I options
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B cjpeg
+compresses the named image file, or the standard input if no file is
+named, and produces a JPEG/JFIF file on the standard output.
+The currently supported input file formats are: PPM (PBMPLUS color
+format), PGM (PBMPLUS gray-scale format), BMP, Targa, and RLE (Utah Raster
+Toolkit format).  (RLE is supported only if the URT library is available.)
+.SH OPTIONS
+All switch names may be abbreviated; for example,
+.B \-grayscale
+may be written
+.B \-gray
+or
+.BR \-gr .
+Most of the "basic" switches can be abbreviated to as little as one letter.
+Upper and lower case are equivalent (thus
+.B \-BMP
+is the same as
+.BR \-bmp ).
+British spellings are also accepted (e.g.,
+.BR \-greyscale ),
+though for brevity these are not mentioned below.
+.PP
+The basic switches are:
+.TP
+.BI \-quality " N"
+Scale quantization tables to adjust image quality.  Quality is 0 (worst) to
+100 (best); default is 75.  (See below for more info.)
+.TP
+.B \-grayscale
+Create monochrome JPEG file from color input.  Be sure to use this switch when
+compressing a grayscale BMP file, because
+.B cjpeg
+isn't bright enough to notice whether a BMP file uses only shades of gray.
+By saying
+.BR \-grayscale ,
+you'll get a smaller JPEG file that takes less time to process.
+.TP
+.B \-optimize
+Perform optimization of entropy encoding parameters.  Without this, default
+encoding parameters are used.
+.B \-optimize
+usually makes the JPEG file a little smaller, but
+.B cjpeg
+runs somewhat slower and needs much more memory.  Image quality and speed of
+decompression are unaffected by
+.BR \-optimize .
+.TP
+.B \-progressive
+Create progressive JPEG file (see below).
+.TP
+.B \-targa
+Input file is Targa format.  Targa files that contain an "identification"
+field will not be automatically recognized by
+.BR cjpeg ;
+for such files you must specify
+.B \-targa
+to make
+.B cjpeg
+treat the input as Targa format.
+For most Targa files, you won't need this switch.
+.PP
+The
+.B \-quality
+switch lets you trade off compressed file size against quality of the
+reconstructed image: the higher the quality setting, the larger the JPEG file,
+and the closer the output image will be to the original input.  Normally you
+want to use the lowest quality setting (smallest file) that decompresses into
+something visually indistinguishable from the original image.  For this
+purpose the quality setting should be between 50 and 95; the default of 75 is
+often about right.  If you see defects at
+.B \-quality
+75, then go up 5 or 10 counts at a time until you are happy with the output
+image.  (The optimal setting will vary from one image to another.)
+.PP
+.B \-quality
+100 will generate a quantization table of all 1's, minimizing loss in the
+quantization step (but there is still information loss in subsampling, as well
+as roundoff error).  This setting is mainly of interest for experimental
+purposes.  Quality values above about 95 are
+.B not
+recommended for normal use; the compressed file size goes up dramatically for
+hardly any gain in output image quality.
+.PP
+In the other direction, quality values below 50 will produce very small files
+of low image quality.  Settings around 5 to 10 might be useful in preparing an
+index of a large image library, for example.  Try
+.B \-quality
+2 (or so) for some amusing Cubist effects.  (Note: quality
+values below about 25 generate 2-byte quantization tables, which are
+considered optional in the JPEG standard.
+.B cjpeg
+emits a warning message when you give such a quality value, because some
+other JPEG programs may be unable to decode the resulting file.  Use
+.B \-baseline
+if you need to ensure compatibility at low quality values.)
+.PP
+The
+.B \-progressive
+switch creates a "progressive JPEG" file.  In this type of JPEG file, the data
+is stored in multiple scans of increasing quality.  If the file is being
+transmitted over a slow communications link, the decoder can use the first
+scan to display a low-quality image very quickly, and can then improve the
+display with each subsequent scan.  The final image is exactly equivalent to a
+standard JPEG file of the same quality setting, and the total file size is
+about the same --- often a little smaller.
+.B Caution:
+progressive JPEG is not yet widely implemented, so many decoders will be
+unable to view a progressive JPEG file at all.
+.PP
+Switches for advanced users:
+.TP
+.B \-dct int
+Use integer DCT method (default).
+.TP
+.B \-dct fast
+Use fast integer DCT (less accurate).
+.TP
+.B \-dct float
+Use floating-point DCT method.
+The float method is very slightly more accurate than the int method, but is
+much slower unless your machine has very fast floating-point hardware.  Also
+note that results of the floating-point method may vary slightly across
+machines, while the integer methods should give the same results everywhere.
+The fast integer method is much less accurate than the other two.
+.TP
+.BI \-restart " N"
+Emit a JPEG restart marker every N MCU rows, or every N MCU blocks if "B" is
+attached to the number.
+.B \-restart 0
+(the default) means no restart markers.
+.TP
+.BI \-smooth " N"
+Smooth the input image to eliminate dithering noise.  N, ranging from 1 to
+100, indicates the strength of smoothing.  0 (the default) means no smoothing.
+.TP
+.BI \-maxmemory " N"
+Set limit for amount of memory to use in processing large images.  Value is
+in thousands of bytes, or millions of bytes if "M" is attached to the
+number.  For example,
+.B \-max 4m
+selects 4000000 bytes.  If more space is needed, temporary files will be used.
+.TP
+.BI \-outfile " name"
+Send output image to the named file, not to standard output.
+.TP
+.B \-verbose
+Enable debug printout.  More
+.BR \-v 's
+give more output.  Also, version information is printed at startup.
+.TP
+.B \-debug
+Same as
+.BR \-verbose .
+.PP
+The
+.B \-restart
+option inserts extra markers that allow a JPEG decoder to resynchronize after
+a transmission error.  Without restart markers, any damage to a compressed
+file will usually ruin the image from the point of the error to the end of the
+image; with restart markers, the damage is usually confined to the portion of
+the image up to the next restart marker.  Of course, the restart markers
+occupy extra space.  We recommend
+.B \-restart 1
+for images that will be transmitted across unreliable networks such as Usenet.
+.PP
+The
+.B \-smooth
+option filters the input to eliminate fine-scale noise.  This is often useful
+when converting dithered images to JPEG: a moderate smoothing factor of 10 to
+50 gets rid of dithering patterns in the input file, resulting in a smaller
+JPEG file and a better-looking image.  Too large a smoothing factor will
+visibly blur the image, however.
+.PP
+Switches for wizards:
+.TP
+.B \-baseline
+Force baseline-compatible quantization tables to be generated.  This clamps
+quantization values to 8 bits even at low quality settings.  (This switch is
+poorly named, since it does not ensure that the output is actually baseline
+JPEG.  For example, you can use
+.B \-baseline
+and
+.B \-progressive
+together.)
+.TP
+.BI \-qtables " file"
+Use the quantization tables given in the specified text file.
+.TP
+.BI \-qslots " N[,...]"
+Select which quantization table to use for each color component.
+.TP
+.BI \-sample " HxV[,...]"
+Set JPEG sampling factors for each color component.
+.TP
+.BI \-scans " file"
+Use the scan script given in the specified text file.
+.PP
+The "wizard" switches are intended for experimentation with JPEG.  If you
+don't know what you are doing, \fBdon't use them\fR.  These switches are
+documented further in the file wizard.doc.
+.SH EXAMPLES
+.LP
+This example compresses the PPM file foo.ppm with a quality factor of
+60 and saves the output as foo.jpg:
+.IP
+.B cjpeg \-quality
+.I 60 foo.ppm
+.B >
+.I foo.jpg
+.SH HINTS
+Color GIF files are not the ideal input for JPEG; JPEG is really intended for
+compressing full-color (24-bit) images.  In particular, don't try to convert
+cartoons, line drawings, and other images that have only a few distinct
+colors.  GIF works great on these, JPEG does not.  If you want to convert a
+GIF to JPEG, you should experiment with
+.BR cjpeg 's
+.B \-quality
+and
+.B \-smooth
+options to get a satisfactory conversion.
+.B \-smooth 10
+or so is often helpful.
+.PP
+Avoid running an image through a series of JPEG compression/decompression
+cycles.  Image quality loss will accumulate; after ten or so cycles the image
+may be noticeably worse than it was after one cycle.  It's best to use a
+lossless format while manipulating an image, then convert to JPEG format when
+you are ready to file the image away.
+.PP
+The
+.B \-optimize
+option to
+.B cjpeg
+is worth using when you are making a "final" version for posting or archiving.
+It's also a win when you are using low quality settings to make very small
+JPEG files; the percentage improvement is often a lot more than it is on
+larger files.  (At present,
+.B \-optimize
+mode is always selected when generating progressive JPEG files.)
+.SH ENVIRONMENT
+.TP
+.B JPEGMEM
+If this environment variable is set, its value is the default memory limit.
+The value is specified as described for the
+.B \-maxmemory
+switch.
+.B JPEGMEM
+overrides the default value specified when the program was compiled, and
+itself is overridden by an explicit
+.BR \-maxmemory .
+.SH SEE ALSO
+.BR djpeg (1),
+.BR jpegtran (1),
+.BR rdjpgcom (1),
+.BR wrjpgcom (1)
+.br
+.BR ppm (5),
+.BR pgm (5)
+.br
+Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
+.SH AUTHOR
+Independent JPEG Group
+.SH BUGS
+Arithmetic coding is not supported for legal reasons.
+.PP
+GIF input files are no longer supported, to avoid the Unisys LZW patent.
+Use a Unisys-licensed program if you need to read a GIF file.  (Conversion
+of GIF files to JPEG is usually a bad idea anyway.)
+.PP
+Not all variants of BMP and Targa file formats are supported.
+.PP
+The
+.B \-targa
+switch is not a bug, it's a feature.  (It would be a bug if the Targa format
+designers had not been clueless.)
+.PP
+Still not as fast as we'd like.
diff --git a/libjpegtwrp/cjpeg.c b/libjpegtwrp/cjpeg.c
new file mode 100644
index 000000000..f2a929f0c
--- /dev/null
+++ b/libjpegtwrp/cjpeg.c
@@ -0,0 +1,606 @@
+/*
+ * cjpeg.c
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a command-line user interface for the JPEG compressor.
+ * It should work on any system with Unix- or MS-DOS-style command lines.
+ *
+ * Two different command line styles are permitted, depending on the
+ * compile-time switch TWO_FILE_COMMANDLINE:
+ *	cjpeg [options]  inputfile outputfile
+ *	cjpeg [options]  [inputfile]
+ * In the second style, output is always to standard output, which you'd
+ * normally redirect to a file or pipe to some other program.  Input is
+ * either from a named file or from standard input (typically redirected).
+ * The second style is convenient on Unix but is unhelpful on systems that
+ * don't support pipes.  Also, you MUST use the first style if your system
+ * doesn't do binary I/O to stdin/stdout.
+ * To simplify script writing, the "-outfile" switch is provided.  The syntax
+ *	cjpeg [options]  -outfile outputfile  inputfile
+ * works regardless of which command line style is used.
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+#include "jversion.h"		/* for version message */
+
+#ifdef USE_CCOMMAND		/* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>		/* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>		/* Think declares it here */
+#endif
+#endif
+
+
+/* Create the add-on message string table. */
+
+#define JMESSAGE(code,string)	string ,
+
+static const char * const cdjpeg_message_table[] = {
+#include "cderror.h"
+  NULL
+};
+
+
+/*
+ * This routine determines what format the input file is,
+ * and selects the appropriate input-reading module.
+ *
+ * To determine which family of input formats the file belongs to,
+ * we may look only at the first byte of the file, since C does not
+ * guarantee that more than one character can be pushed back with ungetc.
+ * Looking at additional bytes would require one of these approaches:
+ *     1) assume we can fseek() the input file (fails for piped input);
+ *     2) assume we can push back more than one character (works in
+ *        some C implementations, but unportable);
+ *     3) provide our own buffering (breaks input readers that want to use
+ *        stdio directly, such as the RLE library);
+ * or  4) don't put back the data, and modify the input_init methods to assume
+ *        they start reading after the start of file (also breaks RLE library).
+ * #1 is attractive for MS-DOS but is untenable on Unix.
+ *
+ * The most portable solution for file types that can't be identified by their
+ * first byte is to make the user tell us what they are.  This is also the
+ * only approach for "raw" file types that contain only arbitrary values.
+ * We presently apply this method for Targa files.  Most of the time Targa
+ * files start with 0x00, so we recognize that case.  Potentially, however,
+ * a Targa file could start with any byte value (byte 0 is the length of the
+ * seldom-used ID field), so we provide a switch to force Targa input mode.
+ */
+
+static boolean is_targa;	/* records user -targa switch */
+
+
+LOCAL(cjpeg_source_ptr)
+select_file_type (j_compress_ptr cinfo, FILE * infile)
+{
+  int c;
+
+  if (is_targa) {
+#ifdef TARGA_SUPPORTED
+    return jinit_read_targa(cinfo);
+#else
+    ERREXIT(cinfo, JERR_TGA_NOTCOMP);
+#endif
+  }
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(cinfo, JERR_INPUT_EMPTY);
+  if (ungetc(c, infile) == EOF)
+    ERREXIT(cinfo, JERR_UNGETC_FAILED);
+
+  switch (c) {
+#ifdef BMP_SUPPORTED
+  case 'B':
+    return jinit_read_bmp(cinfo);
+#endif
+#ifdef GIF_SUPPORTED
+  case 'G':
+    return jinit_read_gif(cinfo);
+#endif
+#ifdef PPM_SUPPORTED
+  case 'P':
+    return jinit_read_ppm(cinfo);
+#endif
+#ifdef RLE_SUPPORTED
+  case 'R':
+    return jinit_read_rle(cinfo);
+#endif
+#ifdef TARGA_SUPPORTED
+  case 0x00:
+    return jinit_read_targa(cinfo);
+#endif
+  default:
+    ERREXIT(cinfo, JERR_UNKNOWN_FORMAT);
+    break;
+  }
+
+  return NULL;			/* suppress compiler warnings */
+}
+
+
+/*
+ * Argument-parsing code.
+ * The switch parser is designed to be useful with DOS-style command line
+ * syntax, ie, intermixed switches and file names, where only the switches
+ * to the left of a given file name affect processing of that file.
+ * The main program in this file doesn't actually use this capability...
+ */
+
+
+static const char * progname;	/* program name for error messages */
+static char * outfilename;	/* for -outfile switch */
+
+
+LOCAL(void)
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "usage: %s [switches] ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "[inputfile]\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -quality N     Compression quality (0..100; 5-95 is useful range)\n");
+  fprintf(stderr, "  -grayscale     Create monochrome JPEG file\n");
+#ifdef ENTROPY_OPT_SUPPORTED
+  fprintf(stderr, "  -optimize      Optimize Huffman table (smaller file, but slow compression)\n");
+#endif
+#ifdef C_PROGRESSIVE_SUPPORTED
+  fprintf(stderr, "  -progressive   Create progressive JPEG file\n");
+#endif
+#ifdef TARGA_SUPPORTED
+  fprintf(stderr, "  -targa         Input file is Targa format (usually not needed)\n");
+#endif
+  fprintf(stderr, "Switches for advanced users:\n");
+#ifdef DCT_ISLOW_SUPPORTED
+  fprintf(stderr, "  -dct int       Use integer DCT method%s\n",
+	  (JDCT_DEFAULT == JDCT_ISLOW ? " (default)" : ""));
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+  fprintf(stderr, "  -dct fast      Use fast integer DCT (less accurate)%s\n",
+	  (JDCT_DEFAULT == JDCT_IFAST ? " (default)" : ""));
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  fprintf(stderr, "  -dct float     Use floating-point DCT method%s\n",
+	  (JDCT_DEFAULT == JDCT_FLOAT ? " (default)" : ""));
+#endif
+  fprintf(stderr, "  -restart N     Set restart interval in rows, or in blocks with B\n");
+#ifdef INPUT_SMOOTHING_SUPPORTED
+  fprintf(stderr, "  -smooth N      Smooth dithered input (N=1..100 is strength)\n");
+#endif
+  fprintf(stderr, "  -maxmemory N   Maximum memory to use (in kbytes)\n");
+  fprintf(stderr, "  -outfile name  Specify name for output file\n");
+  fprintf(stderr, "  -verbose  or  -debug   Emit debug output\n");
+  fprintf(stderr, "Switches for wizards:\n");
+#ifdef C_ARITH_CODING_SUPPORTED
+  fprintf(stderr, "  -arithmetic    Use arithmetic coding\n");
+#endif
+  fprintf(stderr, "  -baseline      Force baseline quantization tables\n");
+  fprintf(stderr, "  -qtables file  Use quantization tables given in file\n");
+  fprintf(stderr, "  -qslots N[,...]    Set component quantization tables\n");
+  fprintf(stderr, "  -sample HxV[,...]  Set component sampling factors\n");
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+  fprintf(stderr, "  -scans file    Create multi-scan JPEG per script file\n");
+#endif
+  exit(EXIT_FAILURE);
+}
+
+
+LOCAL(int)
+parse_switches (j_compress_ptr cinfo, int argc, char **argv,
+		int last_file_arg_seen, boolean for_real)
+/* Parse optional switches.
+ * Returns argv[] index of first file-name argument (== argc if none).
+ * Any file names with indexes <= last_file_arg_seen are ignored;
+ * they have presumably been processed in a previous iteration.
+ * (Pass 0 for last_file_arg_seen on the first or only iteration.)
+ * for_real is FALSE on the first (dummy) pass; we may skip any expensive
+ * processing.
+ */
+{
+  int argn;
+  char * arg;
+  int quality;			/* -quality parameter */
+  int q_scale_factor;		/* scaling percentage for -qtables */
+  boolean force_baseline;
+  boolean simple_progressive;
+  char * qtablefile = NULL;	/* saves -qtables filename if any */
+  char * qslotsarg = NULL;	/* saves -qslots parm if any */
+  char * samplearg = NULL;	/* saves -sample parm if any */
+  char * scansarg = NULL;	/* saves -scans parm if any */
+
+  /* Set up default JPEG parameters. */
+  /* Note that default -quality level need not, and does not,
+   * match the default scaling for an explicit -qtables argument.
+   */
+  quality = 75;			/* default -quality value */
+  q_scale_factor = 100;		/* default to no scaling for -qtables */
+  force_baseline = FALSE;	/* by default, allow 16-bit quantizers */
+  simple_progressive = FALSE;
+  is_targa = FALSE;
+  outfilename = NULL;
+  cinfo->err->trace_level = 0;
+
+  /* Scan command line options, adjust parameters */
+
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (*arg != '-') {
+      /* Not a switch, must be a file name argument */
+      if (argn <= last_file_arg_seen) {
+	outfilename = NULL;	/* -outfile applies to just one input file */
+	continue;		/* ignore this name if previously processed */
+      }
+      break;			/* else done parsing switches */
+    }
+    arg++;			/* advance past switch marker character */
+
+    if (keymatch(arg, "arithmetic", 1)) {
+      /* Use arithmetic coding. */
+#ifdef C_ARITH_CODING_SUPPORTED
+      cinfo->arith_code = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, arithmetic coding not supported\n",
+	      progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "baseline", 1)) {
+      /* Force baseline-compatible output (8-bit quantizer values). */
+      force_baseline = TRUE;
+
+    } else if (keymatch(arg, "dct", 2)) {
+      /* Select DCT algorithm. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (keymatch(argv[argn], "int", 1)) {
+	cinfo->dct_method = JDCT_ISLOW;
+      } else if (keymatch(argv[argn], "fast", 2)) {
+	cinfo->dct_method = JDCT_IFAST;
+      } else if (keymatch(argv[argn], "float", 2)) {
+	cinfo->dct_method = JDCT_FLOAT;
+      } else
+	usage();
+
+    } else if (keymatch(arg, "debug", 1) || keymatch(arg, "verbose", 1)) {
+      /* Enable debug printouts. */
+      /* On first -d, print version identification */
+      static boolean printed_version = FALSE;
+
+      if (! printed_version) {
+	fprintf(stderr, "Independent JPEG Group's CJPEG, version %s\n%s\n",
+		JVERSION, JCOPYRIGHT);
+	printed_version = TRUE;
+      }
+      cinfo->err->trace_level++;
+
+    } else if (keymatch(arg, "grayscale", 2) || keymatch(arg, "greyscale",2)) {
+      /* Force a monochrome JPEG file to be generated. */
+      jpeg_set_colorspace(cinfo, JCS_GRAYSCALE);
+
+    } else if (keymatch(arg, "maxmemory", 3)) {
+      /* Maximum memory in Kb (or Mb with 'm'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+	usage();
+      if (ch == 'm' || ch == 'M')
+	lval *= 1000L;
+      cinfo->mem->max_memory_to_use = lval * 1000L;
+
+    } else if (keymatch(arg, "optimize", 1) || keymatch(arg, "optimise", 1)) {
+      /* Enable entropy parm optimization. */
+#ifdef ENTROPY_OPT_SUPPORTED
+      cinfo->optimize_coding = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, entropy optimization was not compiled\n",
+	      progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "outfile", 4)) {
+      /* Set output file name. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      outfilename = argv[argn];	/* save it away for later use */
+
+    } else if (keymatch(arg, "progressive", 1)) {
+      /* Select simple progressive mode. */
+#ifdef C_PROGRESSIVE_SUPPORTED
+      simple_progressive = TRUE;
+      /* We must postpone execution until num_components is known. */
+#else
+      fprintf(stderr, "%s: sorry, progressive output was not compiled\n",
+	      progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "quality", 1)) {
+      /* Quality factor (quantization table scaling factor). */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (sscanf(argv[argn], "%d", &quality) != 1)
+	usage();
+      /* Change scale factor in case -qtables is present. */
+      q_scale_factor = jpeg_quality_scaling(quality);
+
+    } else if (keymatch(arg, "qslots", 2)) {
+      /* Quantization table slot numbers. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      qslotsarg = argv[argn];
+      /* Must delay setting qslots until after we have processed any
+       * colorspace-determining switches, since jpeg_set_colorspace sets
+       * default quant table numbers.
+       */
+
+    } else if (keymatch(arg, "qtables", 2)) {
+      /* Quantization tables fetched from file. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      qtablefile = argv[argn];
+      /* We postpone actually reading the file in case -quality comes later. */
+
+    } else if (keymatch(arg, "restart", 1)) {
+      /* Restart interval in MCU rows (or in MCUs with 'b'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+	usage();
+      if (lval < 0 || lval > 65535L)
+	usage();
+      if (ch == 'b' || ch == 'B') {
+	cinfo->restart_interval = (unsigned int) lval;
+	cinfo->restart_in_rows = 0; /* else prior '-restart n' overrides me */
+      } else {
+	cinfo->restart_in_rows = (int) lval;
+	/* restart_interval will be computed during startup */
+      }
+
+    } else if (keymatch(arg, "sample", 2)) {
+      /* Set sampling factors. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      samplearg = argv[argn];
+      /* Must delay setting sample factors until after we have processed any
+       * colorspace-determining switches, since jpeg_set_colorspace sets
+       * default sampling factors.
+       */
+
+    } else if (keymatch(arg, "scans", 2)) {
+      /* Set scan script. */
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      scansarg = argv[argn];
+      /* We must postpone reading the file in case -progressive appears. */
+#else
+      fprintf(stderr, "%s: sorry, multi-scan output was not compiled\n",
+	      progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "smooth", 2)) {
+      /* Set input smoothing factor. */
+      int val;
+
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (sscanf(argv[argn], "%d", &val) != 1)
+	usage();
+      if (val < 0 || val > 100)
+	usage();
+      cinfo->smoothing_factor = val;
+
+    } else if (keymatch(arg, "targa", 1)) {
+      /* Input file is Targa format. */
+      is_targa = TRUE;
+
+    } else {
+      usage();			/* bogus switch */
+    }
+  }
+
+  /* Post-switch-scanning cleanup */
+
+  if (for_real) {
+
+    /* Set quantization tables for selected quality. */
+    /* Some or all may be overridden if -qtables is present. */
+    jpeg_set_quality(cinfo, quality, force_baseline);
+
+    if (qtablefile != NULL)	/* process -qtables if it was present */
+      if (! read_quant_tables(cinfo, qtablefile,
+			      q_scale_factor, force_baseline))
+	usage();
+
+    if (qslotsarg != NULL)	/* process -qslots if it was present */
+      if (! set_quant_slots(cinfo, qslotsarg))
+	usage();
+
+    if (samplearg != NULL)	/* process -sample if it was present */
+      if (! set_sample_factors(cinfo, samplearg))
+	usage();
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+    if (simple_progressive)	/* process -progressive; -scans can override */
+      jpeg_simple_progression(cinfo);
+#endif
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+    if (scansarg != NULL)	/* process -scans if it was present */
+      if (! read_scan_script(cinfo, scansarg))
+	usage();
+#endif
+  }
+
+  return argn;			/* return index of next arg (file name) */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  struct jpeg_compress_struct cinfo;
+  struct jpeg_error_mgr jerr;
+#ifdef PROGRESS_REPORT
+  struct cdjpeg_progress_mgr progress;
+#endif
+  int file_index;
+  cjpeg_source_ptr src_mgr;
+  FILE * input_file;
+  FILE * output_file;
+  JDIMENSION num_scanlines;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "cjpeg";		/* in case C library doesn't provide it */
+
+  /* Initialize the JPEG compression object with default error handling. */
+  cinfo.err = jpeg_std_error(&jerr);
+  jpeg_create_compress(&cinfo);
+  /* Add some application-specific error messages (from cderror.h) */
+  jerr.addon_message_table = cdjpeg_message_table;
+  jerr.first_addon_message = JMSG_FIRSTADDONCODE;
+  jerr.last_addon_message = JMSG_LASTADDONCODE;
+
+  /* Now safe to enable signal catcher. */
+#ifdef NEED_SIGNAL_CATCHER
+  enable_signal_catcher((j_common_ptr) &cinfo);
+#endif
+
+  /* Initialize JPEG parameters.
+   * Much of this may be overridden later.
+   * In particular, we don't yet know the input file's color space,
+   * but we need to provide some value for jpeg_set_defaults() to work.
+   */
+
+  cinfo.in_color_space = JCS_RGB; /* arbitrary guess */
+  jpeg_set_defaults(&cinfo);
+
+  /* Scan command line to find file names.
+   * It is convenient to use just one switch-parsing routine, but the switch
+   * values read here are ignored; we will rescan the switches after opening
+   * the input file.
+   */
+
+  file_index = parse_switches(&cinfo, argc, argv, 0, FALSE);
+
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have either -outfile switch or explicit output file name */
+  if (outfilename == NULL) {
+    if (file_index != argc-2) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+	      progname);
+      usage();
+    }
+    outfilename = argv[file_index+1];
+  } else {
+    if (file_index != argc-1) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+	      progname);
+      usage();
+    }
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (file_index < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Open the input file. */
+  if (file_index < argc) {
+    if ((input_file = fopen(argv[file_index], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[file_index]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+    input_file = read_stdin();
+  }
+
+  /* Open the output file. */
+  if (outfilename != NULL) {
+    if ((output_file = fopen(outfilename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, outfilename);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default output file is stdout */
+    output_file = write_stdout();
+  }
+
+#ifdef PROGRESS_REPORT
+  start_progress_monitor((j_common_ptr) &cinfo, &progress);
+#endif
+
+  /* Figure out the input file format, and set up to read it. */
+  src_mgr = select_file_type(&cinfo, input_file);
+  src_mgr->input_file = input_file;
+
+  /* Read the input file header to obtain file size & colorspace. */
+  (*src_mgr->start_input) (&cinfo, src_mgr);
+
+  /* Now that we know input colorspace, fix colorspace-dependent defaults */
+  jpeg_default_colorspace(&cinfo);
+
+  /* Adjust default compression parameters by re-parsing the options */
+  file_index = parse_switches(&cinfo, argc, argv, 0, TRUE);
+
+  /* Specify data destination for compression */
+  jpeg_stdio_dest(&cinfo, output_file);
+
+  /* Start compressor */
+  jpeg_start_compress(&cinfo, TRUE);
+
+  /* Process data */
+  while (cinfo.next_scanline < cinfo.image_height) {
+    num_scanlines = (*src_mgr->get_pixel_rows) (&cinfo, src_mgr);
+    (void) jpeg_write_scanlines(&cinfo, src_mgr->buffer, num_scanlines);
+  }
+
+  /* Finish compression and release memory */
+  (*src_mgr->finish_input) (&cinfo, src_mgr);
+  jpeg_finish_compress(&cinfo);
+  jpeg_destroy_compress(&cinfo);
+
+  /* Close files, if we opened them */
+  if (input_file != stdin)
+    fclose(input_file);
+  if (output_file != stdout)
+    fclose(output_file);
+
+#ifdef PROGRESS_REPORT
+  end_progress_monitor((j_common_ptr) &cinfo);
+#endif
+
+  /* All done. */
+  exit(jerr.num_warnings ? EXIT_WARNING : EXIT_SUCCESS);
+  return 0;			/* suppress no-return-value warnings */
+}
diff --git a/libjpegtwrp/ckconfig.c b/libjpegtwrp/ckconfig.c
new file mode 100644
index 000000000..34baf795b
--- /dev/null
+++ b/libjpegtwrp/ckconfig.c
@@ -0,0 +1,402 @@
+/*
+ * ckconfig.c
+ *
+ * Copyright (C) 1991-1994, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ */
+
+/*
+ * This program is intended to help you determine how to configure the JPEG
+ * software for installation on a particular system.  The idea is to try to
+ * compile and execute this program.  If your compiler fails to compile the
+ * program, make changes as indicated in the comments below.  Once you can
+ * compile the program, run it, and it will produce a "jconfig.h" file for
+ * your system.
+ *
+ * As a general rule, each time you try to compile this program,
+ * pay attention only to the *first* error message you get from the compiler.
+ * Many C compilers will issue lots of spurious error messages once they
+ * have gotten confused.  Go to the line indicated in the first error message,
+ * and read the comments preceding that line to see what to change.
+ *
+ * Almost all of the edits you may need to make to this program consist of
+ * changing a line that reads "#define SOME_SYMBOL" to "#undef SOME_SYMBOL",
+ * or vice versa.  This is called defining or undefining that symbol.
+ */
+
+
+/* First we must see if your system has the include files we need.
+ * We start out with the assumption that your system has all the ANSI-standard
+ * include files.  If you get any error trying to include one of these files,
+ * undefine the corresponding HAVE_xxx symbol.
+ */
+
+#define HAVE_STDDEF_H		/* replace 'define' by 'undef' if error here */
+#ifdef HAVE_STDDEF_H		/* next line will be skipped if you undef... */
+#include <stddef.h>
+#endif
+
+#define HAVE_STDLIB_H		/* same thing for stdlib.h */
+#ifdef HAVE_STDLIB_H
+#include <stdlib.h>
+#endif
+
+#include <stdio.h>		/* If you ain't got this, you ain't got C. */
+
+/* We have to see if your string functions are defined by
+ * strings.h (old BSD convention) or string.h (everybody else).
+ * We try the non-BSD convention first; define NEED_BSD_STRINGS
+ * if the compiler says it can't find string.h.
+ */
+
+#undef NEED_BSD_STRINGS
+
+#ifdef NEED_BSD_STRINGS
+#include <strings.h>
+#else
+#include <string.h>
+#endif
+
+/* On some systems (especially older Unix machines), type size_t is
+ * defined only in the include file <sys/types.h>.  If you get a failure
+ * on the size_t test below, try defining NEED_SYS_TYPES_H.
+ */
+
+#undef NEED_SYS_TYPES_H		/* start by assuming we don't need it */
+#ifdef NEED_SYS_TYPES_H
+#include <sys/types.h>
+#endif
+
+
+/* Usually type size_t is defined in one of the include files we've included
+ * above.  If not, you'll get an error on the "typedef size_t my_size_t;" line.
+ * In that case, first try defining NEED_SYS_TYPES_H just above.
+ * If that doesn't work, you'll have to search through your system library
+ * to figure out which include file defines "size_t".  Look for a line that
+ * says "typedef something-or-other size_t;".  Then, change the line below
+ * that says "#include <someincludefile.h>" to instead include the file
+ * you found size_t in, and define NEED_SPECIAL_INCLUDE.  If you can't find
+ * type size_t anywhere, try replacing "#include <someincludefile.h>" with
+ * "typedef unsigned int size_t;".
+ */
+
+#undef NEED_SPECIAL_INCLUDE	/* assume we DON'T need it, for starters */
+
+#ifdef NEED_SPECIAL_INCLUDE
+#include <someincludefile.h>
+#endif
+
+typedef size_t my_size_t;	/* The payoff: do we have size_t now? */
+
+
+/* The next question is whether your compiler supports ANSI-style function
+ * prototypes.  You need to know this in order to choose between using
+ * makefile.ansi and using makefile.unix.
+ * The #define line below is set to assume you have ANSI function prototypes.
+ * If you get an error in this group of lines, undefine HAVE_PROTOTYPES.
+ */
+
+#define HAVE_PROTOTYPES
+
+#ifdef HAVE_PROTOTYPES
+int testfunction (int arg1, int * arg2); /* check prototypes */
+
+struct methods_struct {		/* check method-pointer declarations */
+  int (*error_exit) (char *msgtext);
+  int (*trace_message) (char *msgtext);
+  int (*another_method) (void);
+};
+
+int testfunction (int arg1, int * arg2) /* check definitions */
+{
+  return arg2[arg1];
+}
+
+int test2function (void)	/* check void arg list */
+{
+  return 0;
+}
+#endif
+
+
+/* Now we want to find out if your compiler knows what "unsigned char" means.
+ * If you get an error on the "unsigned char un_char;" line,
+ * then undefine HAVE_UNSIGNED_CHAR.
+ */
+
+#define HAVE_UNSIGNED_CHAR
+
+#ifdef HAVE_UNSIGNED_CHAR
+unsigned char un_char;
+#endif
+
+
+/* Now we want to find out if your compiler knows what "unsigned short" means.
+ * If you get an error on the "unsigned short un_short;" line,
+ * then undefine HAVE_UNSIGNED_SHORT.
+ */
+
+#define HAVE_UNSIGNED_SHORT
+
+#ifdef HAVE_UNSIGNED_SHORT
+unsigned short un_short;
+#endif
+
+
+/* Now we want to find out if your compiler understands type "void".
+ * If you get an error anywhere in here, undefine HAVE_VOID.
+ */
+
+#define HAVE_VOID
+
+#ifdef HAVE_VOID
+/* Caution: a C++ compiler will insist on complete prototypes */
+typedef void * void_ptr;	/* check void * */
+#ifdef HAVE_PROTOTYPES		/* check ptr to function returning void */
+typedef void (*void_func) (int a, int b);
+#else
+typedef void (*void_func) ();
+#endif
+
+#ifdef HAVE_PROTOTYPES		/* check void function result */
+void test3function (void_ptr arg1, void_func arg2)
+#else
+void test3function (arg1, arg2)
+     void_ptr arg1;
+     void_func arg2;
+#endif
+{
+  char * locptr = (char *) arg1; /* check casting to and from void * */
+  arg1 = (void *) locptr;
+  (*arg2) (1, 2);		/* check call of fcn returning void */
+}
+#endif
+
+
+/* Now we want to find out if your compiler knows what "const" means.
+ * If you get an error here, undefine HAVE_CONST.
+ */
+
+#define HAVE_CONST
+
+#ifdef HAVE_CONST
+static const int carray[3] = {1, 2, 3};
+
+#ifdef HAVE_PROTOTYPES
+int test4function (const int arg1)
+#else
+int test4function (arg1)
+     const int arg1;
+#endif
+{
+  return carray[arg1];
+}
+#endif
+
+
+/* If you get an error or warning about this structure definition,
+ * define INCOMPLETE_TYPES_BROKEN.
+ */
+
+#undef INCOMPLETE_TYPES_BROKEN
+
+#ifndef INCOMPLETE_TYPES_BROKEN
+typedef struct undefined_structure * undef_struct_ptr;
+#endif
+
+
+/* If you get an error about duplicate names,
+ * define NEED_SHORT_EXTERNAL_NAMES.
+ */
+
+#undef NEED_SHORT_EXTERNAL_NAMES
+
+#ifndef NEED_SHORT_EXTERNAL_NAMES
+
+int possibly_duplicate_function ()
+{
+  return 0;
+}
+
+int possibly_dupli_function ()
+{
+  return 1;
+}
+
+#endif
+
+
+
+/************************************************************************
+ *  OK, that's it.  You should not have to change anything beyond this
+ *  point in order to compile and execute this program.  (You might get
+ *  some warnings, but you can ignore them.)
+ *  When you run the program, it will make a couple more tests that it
+ *  can do automatically, and then it will create jconfig.h and print out
+ *  any additional suggestions it has.
+ ************************************************************************
+ */
+
+
+#ifdef HAVE_PROTOTYPES
+int is_char_signed (int arg)
+#else
+int is_char_signed (arg)
+     int arg;
+#endif
+{
+  if (arg == 189) {		/* expected result for unsigned char */
+    return 0;			/* type char is unsigned */
+  }
+  else if (arg != -67) {	/* expected result for signed char */
+    printf("Hmm, it seems 'char' is not eight bits wide on your machine.\n");
+    printf("I fear the JPEG software will not work at all.\n\n");
+  }
+  return 1;			/* assume char is signed otherwise */
+}
+
+
+#ifdef HAVE_PROTOTYPES
+int is_shifting_signed (long arg)
+#else
+int is_shifting_signed (arg)
+     long arg;
+#endif
+/* See whether right-shift on a long is signed or not. */
+{
+  long res = arg >> 4;
+
+  if (res == -0x7F7E80CL) {	/* expected result for signed shift */
+    return 1;			/* right shift is signed */
+  }
+  /* see if unsigned-shift hack will fix it. */
+  /* we can't just test exact value since it depends on width of long... */
+  res |= (~0L) << (32-4);
+  if (res == -0x7F7E80CL) {	/* expected result now? */
+    return 0;			/* right shift is unsigned */
+  }
+  printf("Right shift isn't acting as I expect it to.\n");
+  printf("I fear the JPEG software will not work at all.\n\n");
+  return 0;			/* try it with unsigned anyway */
+}
+
+
+#ifdef HAVE_PROTOTYPES
+int main (int argc, char ** argv)
+#else
+int main (argc, argv)
+     int argc;
+     char ** argv;
+#endif
+{
+  char signed_char_check = (char) (-67);
+  FILE *outfile;
+
+  /* Attempt to write jconfig.h */
+  if ((outfile = fopen("jconfig.h", "w")) == NULL) {
+    printf("Failed to write jconfig.h\n");
+    return 1;
+  }
+
+  /* Write out all the info */
+  fprintf(outfile, "/* jconfig.h --- generated by ckconfig.c */\n");
+  fprintf(outfile, "/* see jconfig.doc for explanations */\n\n");
+#ifdef HAVE_PROTOTYPES
+  fprintf(outfile, "#define HAVE_PROTOTYPES\n");
+#else
+  fprintf(outfile, "#undef HAVE_PROTOTYPES\n");
+#endif
+#ifdef HAVE_UNSIGNED_CHAR
+  fprintf(outfile, "#define HAVE_UNSIGNED_CHAR\n");
+#else
+  fprintf(outfile, "#undef HAVE_UNSIGNED_CHAR\n");
+#endif
+#ifdef HAVE_UNSIGNED_SHORT
+  fprintf(outfile, "#define HAVE_UNSIGNED_SHORT\n");
+#else
+  fprintf(outfile, "#undef HAVE_UNSIGNED_SHORT\n");
+#endif
+#ifdef HAVE_VOID
+  fprintf(outfile, "/* #define void char */\n");
+#else
+  fprintf(outfile, "#define void char\n");
+#endif
+#ifdef HAVE_CONST
+  fprintf(outfile, "/* #define const */\n");
+#else
+  fprintf(outfile, "#define const\n");
+#endif
+  if (is_char_signed((int) signed_char_check))
+    fprintf(outfile, "#undef CHAR_IS_UNSIGNED\n");
+  else
+    fprintf(outfile, "#define CHAR_IS_UNSIGNED\n");
+#ifdef HAVE_STDDEF_H
+  fprintf(outfile, "#define HAVE_STDDEF_H\n");
+#else
+  fprintf(outfile, "#undef HAVE_STDDEF_H\n");
+#endif
+#ifdef HAVE_STDLIB_H
+  fprintf(outfile, "#define HAVE_STDLIB_H\n");
+#else
+  fprintf(outfile, "#undef HAVE_STDLIB_H\n");
+#endif
+#ifdef NEED_BSD_STRINGS
+  fprintf(outfile, "#define NEED_BSD_STRINGS\n");
+#else
+  fprintf(outfile, "#undef NEED_BSD_STRINGS\n");
+#endif
+#ifdef NEED_SYS_TYPES_H
+  fprintf(outfile, "#define NEED_SYS_TYPES_H\n");
+#else
+  fprintf(outfile, "#undef NEED_SYS_TYPES_H\n");
+#endif
+  fprintf(outfile, "#undef NEED_FAR_POINTERS\n");
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+  fprintf(outfile, "#define NEED_SHORT_EXTERNAL_NAMES\n");
+#else
+  fprintf(outfile, "#undef NEED_SHORT_EXTERNAL_NAMES\n");
+#endif
+#ifdef INCOMPLETE_TYPES_BROKEN
+  fprintf(outfile, "#define INCOMPLETE_TYPES_BROKEN\n");
+#else
+  fprintf(outfile, "#undef INCOMPLETE_TYPES_BROKEN\n");
+#endif
+  fprintf(outfile, "\n#ifdef JPEG_INTERNALS\n\n");
+  if (is_shifting_signed(-0x7F7E80B1L))
+    fprintf(outfile, "#undef RIGHT_SHIFT_IS_UNSIGNED\n");
+  else
+    fprintf(outfile, "#define RIGHT_SHIFT_IS_UNSIGNED\n");
+  fprintf(outfile, "\n#endif /* JPEG_INTERNALS */\n");
+  fprintf(outfile, "\n#ifdef JPEG_CJPEG_DJPEG\n\n");
+  fprintf(outfile, "#define BMP_SUPPORTED		/* BMP image file format */\n");
+  fprintf(outfile, "#define GIF_SUPPORTED		/* GIF image file format */\n");
+  fprintf(outfile, "#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */\n");
+  fprintf(outfile, "#undef RLE_SUPPORTED		/* Utah RLE image file format */\n");
+  fprintf(outfile, "#define TARGA_SUPPORTED		/* Targa image file format */\n\n");
+  fprintf(outfile, "#undef TWO_FILE_COMMANDLINE	/* You may need this on non-Unix systems */\n");
+  fprintf(outfile, "#undef NEED_SIGNAL_CATCHER	/* Define this if you use jmemname.c */\n");
+  fprintf(outfile, "#undef DONT_USE_B_MODE\n");
+  fprintf(outfile, "/* #define PROGRESS_REPORT */	/* optional */\n");
+  fprintf(outfile, "\n#endif /* JPEG_CJPEG_DJPEG */\n");
+
+  /* Close the jconfig.h file */
+  fclose(outfile);
+
+  /* User report */
+  printf("Configuration check for Independent JPEG Group's software done.\n");
+  printf("\nI have written the jconfig.h file for you.\n\n");
+#ifdef HAVE_PROTOTYPES
+  printf("You should use makefile.ansi as the starting point for your Makefile.\n");
+#else
+  printf("You should use makefile.unix as the starting point for your Makefile.\n");
+#endif
+
+#ifdef NEED_SPECIAL_INCLUDE
+  printf("\nYou'll need to change jconfig.h to include the system include file\n");
+  printf("that you found type size_t in, or add a direct definition of type\n");
+  printf("size_t if that's what you used.  Just add it to the end.\n");
+#endif
+
+  return 0;
+}
diff --git a/libjpegtwrp/coderules.doc b/libjpegtwrp/coderules.doc
new file mode 100644
index 000000000..0ab5d9bd3
--- /dev/null
+++ b/libjpegtwrp/coderules.doc
@@ -0,0 +1,118 @@
+IJG JPEG LIBRARY:  CODING RULES
+
+Copyright (C) 1991-1996, Thomas G. Lane.
+This file is part of the Independent JPEG Group's software.
+For conditions of distribution and use, see the accompanying README file.
+
+
+Since numerous people will be contributing code and bug fixes, it's important
+to establish a common coding style.  The goal of using similar coding styles
+is much more important than the details of just what that style is.
+
+In general we follow the recommendations of "Recommended C Style and Coding
+Standards" revision 6.1 (Cannon et al. as modified by Spencer, Keppel and
+Brader).  This document is available in the IJG FTP archive (see
+jpeg/doc/cstyle.ms.tbl.Z, or cstyle.txt.Z for those without nroff/tbl).
+
+Block comments should be laid out thusly:
+
+/*
+ *  Block comments in this style.
+ */
+
+We indent statements in K&R style, e.g.,
+	if (test) {
+	  then-part;
+	} else {
+	  else-part;
+	}
+with two spaces per indentation level.  (This indentation convention is
+handled automatically by GNU Emacs and many other text editors.)
+
+Multi-word names should be written in lower case with underscores, e.g.,
+multi_word_name (not multiWordName).  Preprocessor symbols and enum constants
+are similar but upper case (MULTI_WORD_NAME).  Names should be unique within
+the first fifteen characters.  (On some older systems, global names must be
+unique within six characters.  We accommodate this without cluttering the
+source code by using macros to substitute shorter names.)
+
+We use function prototypes everywhere; we rely on automatic source code
+transformation to feed prototype-less C compilers.  Transformation is done
+by the simple and portable tool 'ansi2knr.c' (courtesy of Ghostscript).
+ansi2knr is not very bright, so it imposes a format requirement on function
+declarations: the function name MUST BEGIN IN COLUMN 1.  Thus all functions
+should be written in the following style:
+
+LOCAL(int *)
+function_name (int a, char *b)
+{
+    code...
+}
+
+Note that each function definition must begin with GLOBAL(type), LOCAL(type),
+or METHODDEF(type).  These macros expand to "static type" or just "type" as
+appropriate.  They provide a readable indication of the routine's usage and
+can readily be changed for special needs.  (For instance, special linkage
+keywords can be inserted for use in Windows DLLs.)
+
+ansi2knr does not transform method declarations (function pointers in
+structs).  We handle these with a macro JMETHOD, defined as
+	#ifdef HAVE_PROTOTYPES
+	#define JMETHOD(type,methodname,arglist)  type (*methodname) arglist
+	#else
+	#define JMETHOD(type,methodname,arglist)  type (*methodname) ()
+	#endif
+which is used like this:
+	struct function_pointers {
+	  JMETHOD(void, init_entropy_encoder, (int somearg, jparms *jp));
+	  JMETHOD(void, term_entropy_encoder, (void));
+	};
+Note the set of parentheses surrounding the parameter list.
+
+A similar solution is used for forward and external function declarations
+(see the EXTERN and JPP macros).
+
+If the code is to work on non-ANSI compilers, we cannot rely on a prototype
+declaration to coerce actual parameters into the right types.  Therefore, use
+explicit casts on actual parameters whenever the actual parameter type is not
+identical to the formal parameter.  Beware of implicit conversions to "int".
+
+It seems there are some non-ANSI compilers in which the sizeof() operator
+is defined to return int, yet size_t is defined as long.  Needless to say,
+this is brain-damaged.  Always use the SIZEOF() macro in place of sizeof(),
+so that the result is guaranteed to be of type size_t.
+
+
+The JPEG library is intended to be used within larger programs.  Furthermore,
+we want it to be reentrant so that it can be used by applications that process
+multiple images concurrently.  The following rules support these requirements:
+
+1. Avoid direct use of file I/O, "malloc", error report printouts, etc;
+pass these through the common routines provided.
+
+2. Minimize global namespace pollution.  Functions should be declared static
+wherever possible.  (Note that our method-based calling conventions help this
+a lot: in many modules only the initialization function will ever need to be
+called directly, so only that function need be externally visible.)  All
+global function names should begin with "jpeg_", and should have an
+abbreviated name (unique in the first six characters) substituted by macro
+when NEED_SHORT_EXTERNAL_NAMES is set.
+
+3. Don't use global variables; anything that must be used in another module
+should be in the common data structures.
+
+4. Don't use static variables except for read-only constant tables.  Variables
+that should be private to a module can be placed into private structures (see
+the system architecture document, structure.doc).
+
+5. Source file names should begin with "j" for files that are part of the
+library proper; source files that are not part of the library, such as cjpeg.c
+and djpeg.c, do not begin with "j".  Keep source file names to eight
+characters (plus ".c" or ".h", etc) to make life easy for MS-DOSers.  Keep
+compression and decompression code in separate source files --- some
+applications may want only one half of the library.
+
+Note: these rules (particularly #4) are not followed religiously in the
+modules that are used in cjpeg/djpeg but are not part of the JPEG library
+proper.  Those modules are not really intended to be used in other
+applications.
diff --git a/libjpegtwrp/config.guess b/libjpegtwrp/config.guess
new file mode 100755
index 000000000..413ed41c0
--- /dev/null
+++ b/libjpegtwrp/config.guess
@@ -0,0 +1,883 @@
+#! /bin/sh
+# Attempt to guess a canonical system name.
+#   Copyright (C) 1992, 93, 94, 95, 96, 1997 Free Software Foundation, Inc.
+#
+# This file is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+# General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+#
+# As a special exception to the GNU General Public License, if you
+# distribute this file as part of a program that contains a
+# configuration script generated by Autoconf, you may include it under
+# the same distribution terms that you use for the rest of that program.
+
+# Written by Per Bothner <bothner@cygnus.com>.
+# The master version of this file is at the FSF in /home/gd/gnu/lib.
+#
+# This script attempts to guess a canonical system name similar to
+# config.sub.  If it succeeds, it prints the system name on stdout, and
+# exits with 0.  Otherwise, it exits with 1.
+#
+# The plan is that this can be called by configure scripts if you
+# don't specify an explicit system type (host/target name).
+#
+# Only a few systems have been added to this list; please add others
+# (but try to keep the structure clean).
+#
+
+# This is needed to find uname on a Pyramid OSx when run in the BSD universe.
+# (ghazi@noc.rutgers.edu 8/24/94.)
+if (test -f /.attbin/uname) >/dev/null 2>&1 ; then
+	PATH=$PATH:/.attbin ; export PATH
+fi
+
+UNAME_MACHINE=`(uname -m) 2>/dev/null` || UNAME_MACHINE=unknown
+UNAME_RELEASE=`(uname -r) 2>/dev/null` || UNAME_RELEASE=unknown
+UNAME_SYSTEM=`(uname -s) 2>/dev/null` || UNAME_SYSTEM=unknown
+UNAME_VERSION=`(uname -v) 2>/dev/null` || UNAME_VERSION=unknown
+
+trap 'rm -f dummy.c dummy.o dummy; exit 1' 1 2 15
+
+# Note: order is significant - the case branches are not exclusive.
+
+case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in
+    alpha:OSF1:*:*)
+	if test $UNAME_RELEASE = "V4.0"; then
+		UNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $3}'`
+	fi
+	# A Vn.n version is a released version.
+	# A Tn.n version is a released field test version.
+	# A Xn.n version is an unreleased experimental baselevel.
+	# 1.2 uses "1.2" for uname -r.
+	cat <<EOF >dummy.s
+	.globl main
+	.ent main
+main:
+	.frame \$30,0,\$26,0
+	.prologue 0
+	.long 0x47e03d80 # implver $0
+	lda \$2,259
+	.long 0x47e20c21 # amask $2,$1
+	srl \$1,8,\$2
+	sll \$2,2,\$2
+	sll \$0,3,\$0
+	addl \$1,\$0,\$0
+	addl \$2,\$0,\$0
+	ret \$31,(\$26),1
+	.end main
+EOF
+	${CC-cc} dummy.s -o dummy 2>/dev/null
+	if test "$?" = 0 ; then
+		./dummy
+		case "$?" in
+			7)
+				UNAME_MACHINE="alpha"
+				;;
+			15)
+				UNAME_MACHINE="alphaev5"
+				;;
+			14)
+				UNAME_MACHINE="alphaev56"
+				;;
+			10)
+				UNAME_MACHINE="alphapca56"
+				;;
+			16)
+				UNAME_MACHINE="alphaev6"
+				;;
+		esac
+	fi
+	rm -f dummy.s dummy
+	echo ${UNAME_MACHINE}-dec-osf`echo ${UNAME_RELEASE} | sed -e 's/^[VTX]//' | tr [[A-Z]] [[a-z]]`
+	exit 0 ;;
+    21064:Windows_NT:50:3)
+	echo alpha-dec-winnt3.5
+	exit 0 ;;
+    Amiga*:UNIX_System_V:4.0:*)
+	echo m68k-cbm-sysv4
+	exit 0;;
+    amiga:NetBSD:*:*)
+      echo m68k-cbm-netbsd${UNAME_RELEASE}
+      exit 0 ;;
+    amiga:OpenBSD:*:*)
+	echo m68k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    arc64:OpenBSD:*:*)
+	echo mips64el-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    arc:OpenBSD:*:*)
+	echo mipsel-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    hkmips:OpenBSD:*:*)
+	echo mips-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    pmax:OpenBSD:*:*)
+	echo mipsel-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    sgi:OpenBSD:*:*)
+	echo mips-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    wgrisc:OpenBSD:*:*)
+	echo mipsel-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    arm:RISC*:1.[012]*:*|arm:riscix:1.[012]*:*)
+	echo arm-acorn-riscix${UNAME_RELEASE}
+	exit 0;;
+    arm32:NetBSD:*:*)
+	echo arm-unknown-netbsd`echo ${UNAME_RELEASE}|sed -e 's/[-_].*/\./'`
+	exit 0 ;;
+    SR2?01:HI-UX/MPP:*:*)
+	echo hppa1.1-hitachi-hiuxmpp
+	exit 0;;
+    Pyramid*:OSx*:*:*|MIS*:OSx*:*:*)
+	# akee@wpdis03.wpafb.af.mil (Earle F. Ake) contributed MIS and NILE.
+	if test "`(/bin/universe) 2>/dev/null`" = att ; then
+		echo pyramid-pyramid-sysv3
+	else
+		echo pyramid-pyramid-bsd
+	fi
+	exit 0 ;;
+    NILE:*:*:dcosx)
+	echo pyramid-pyramid-svr4
+	exit 0 ;;
+    sun4*:SunOS:5.*:* | tadpole*:SunOS:5.*:*)
+	echo sparc-sun-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
+	exit 0 ;;
+    i86pc:SunOS:5.*:*)
+	echo i386-pc-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
+	exit 0 ;;
+    sun4*:SunOS:6*:*)
+	# According to config.sub, this is the proper way to canonicalize
+	# SunOS6.  Hard to guess exactly what SunOS6 will be like, but
+	# it's likely to be more like Solaris than SunOS4.
+	echo sparc-sun-solaris3`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
+	exit 0 ;;
+    sun4*:SunOS:*:*)
+	case "`/usr/bin/arch -k`" in
+	    Series*|S4*)
+		UNAME_RELEASE=`uname -v`
+		;;
+	esac
+	# Japanese Language versions have a version number like `4.1.3-JL'.
+	echo sparc-sun-sunos`echo ${UNAME_RELEASE}|sed -e 's/-/_/'`
+	exit 0 ;;
+    sun3*:SunOS:*:*)
+	echo m68k-sun-sunos${UNAME_RELEASE}
+	exit 0 ;;
+    sun*:*:4.2BSD:*)
+	UNAME_RELEASE=`(head -1 /etc/motd | awk '{print substr($5,1,3)}') 2>/dev/null`
+	test "x${UNAME_RELEASE}" = "x" && UNAME_RELEASE=3
+	case "`/bin/arch`" in
+	    sun3)
+		echo m68k-sun-sunos${UNAME_RELEASE}
+		;;
+	    sun4)
+		echo sparc-sun-sunos${UNAME_RELEASE}
+		;;
+	esac
+	exit 0 ;;
+    aushp:SunOS:*:*)
+	echo sparc-auspex-sunos${UNAME_RELEASE}
+	exit 0 ;;
+    atari*:NetBSD:*:*)
+	echo m68k-atari-netbsd${UNAME_RELEASE}
+	exit 0 ;;
+    atari*:OpenBSD:*:*)
+	echo m68k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    sun3*:NetBSD:*:*)
+	echo m68k-sun-netbsd${UNAME_RELEASE}
+	exit 0 ;;
+    sun3*:OpenBSD:*:*)
+	echo m68k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    mac68k:NetBSD:*:*)
+	echo m68k-apple-netbsd${UNAME_RELEASE}
+	exit 0 ;;
+    mac68k:OpenBSD:*:*)
+	echo m68k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    mvme68k:OpenBSD:*:*)
+	echo m68k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    mvme88k:OpenBSD:*:*)
+	echo m88k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    powerpc:machten:*:*)
+	echo powerpc-apple-machten${UNAME_RELEASE}
+	exit 0 ;;
+    RISC*:Mach:*:*)
+	echo mips-dec-mach_bsd4.3
+	exit 0 ;;
+    RISC*:ULTRIX:*:*)
+	echo mips-dec-ultrix${UNAME_RELEASE}
+	exit 0 ;;
+    VAX*:ULTRIX*:*:*)
+	echo vax-dec-ultrix${UNAME_RELEASE}
+	exit 0 ;;
+    2020:CLIX:*:*)
+	echo clipper-intergraph-clix${UNAME_RELEASE}
+	exit 0 ;;
+    mips:*:*:UMIPS | mips:*:*:RISCos)
+	sed 's/^	//' << EOF >dummy.c
+	int main (argc, argv) int argc; char **argv; {
+	#if defined (host_mips) && defined (MIPSEB)
+	#if defined (SYSTYPE_SYSV)
+	  printf ("mips-mips-riscos%ssysv\n", argv[1]); exit (0);
+	#endif
+	#if defined (SYSTYPE_SVR4)
+	  printf ("mips-mips-riscos%ssvr4\n", argv[1]); exit (0);
+	#endif
+	#if defined (SYSTYPE_BSD43) || defined(SYSTYPE_BSD)
+	  printf ("mips-mips-riscos%sbsd\n", argv[1]); exit (0);
+	#endif
+	#endif
+	  exit (-1);
+	}
+EOF
+	${CC-cc} dummy.c -o dummy \
+	  && ./dummy `echo "${UNAME_RELEASE}" | sed -n 's/\([0-9]*\).*/\1/p'` \
+	  && rm dummy.c dummy && exit 0
+	rm -f dummy.c dummy
+	echo mips-mips-riscos${UNAME_RELEASE}
+	exit 0 ;;
+    Night_Hawk:Power_UNIX:*:*)
+	echo powerpc-harris-powerunix
+	exit 0 ;;
+    m88k:CX/UX:7*:*)
+	echo m88k-harris-cxux7
+	exit 0 ;;
+    m88k:*:4*:R4*)
+	echo m88k-motorola-sysv4
+	exit 0 ;;
+    m88k:*:3*:R3*)
+	echo m88k-motorola-sysv3
+	exit 0 ;;
+    AViiON:dgux:*:*)
+        # DG/UX returns AViiON for all architectures
+        UNAME_PROCESSOR=`/usr/bin/uname -p`
+        if [ $UNAME_PROCESSOR = mc88100 -o $UNAME_PROCESSOR = mc88110 ] ; then
+	if [ ${TARGET_BINARY_INTERFACE}x = m88kdguxelfx \
+	     -o ${TARGET_BINARY_INTERFACE}x = x ] ; then
+		echo m88k-dg-dgux${UNAME_RELEASE}
+	else
+		echo m88k-dg-dguxbcs${UNAME_RELEASE}
+	fi
+        else echo i586-dg-dgux${UNAME_RELEASE}
+        fi
+ 	exit 0 ;;
+    M88*:DolphinOS:*:*)	# DolphinOS (SVR3)
+	echo m88k-dolphin-sysv3
+	exit 0 ;;
+    M88*:*:R3*:*)
+	# Delta 88k system running SVR3
+	echo m88k-motorola-sysv3
+	exit 0 ;;
+    XD88*:*:*:*) # Tektronix XD88 system running UTekV (SVR3)
+	echo m88k-tektronix-sysv3
+	exit 0 ;;
+    Tek43[0-9][0-9]:UTek:*:*) # Tektronix 4300 system running UTek (BSD)
+	echo m68k-tektronix-bsd
+	exit 0 ;;
+    *:IRIX*:*:*)
+	echo mips-sgi-irix`echo ${UNAME_RELEASE}|sed -e 's/-/_/g'`
+	exit 0 ;;
+    ????????:AIX?:[12].1:2)   # AIX 2.2.1 or AIX 2.1.1 is RT/PC AIX.
+	echo romp-ibm-aix      # uname -m gives an 8 hex-code CPU id
+	exit 0 ;;              # Note that: echo "'`uname -s`'" gives 'AIX '
+    i?86:AIX:*:*)
+	echo i386-ibm-aix
+	exit 0 ;;
+    *:AIX:2:3)
+	if grep bos325 /usr/include/stdio.h >/dev/null 2>&1; then
+		sed 's/^		//' << EOF >dummy.c
+		#include <sys/systemcfg.h>
+
+		main()
+			{
+			if (!__power_pc())
+				exit(1);
+			puts("powerpc-ibm-aix3.2.5");
+			exit(0);
+			}
+EOF
+		${CC-cc} dummy.c -o dummy && ./dummy && rm dummy.c dummy && exit 0
+		rm -f dummy.c dummy
+		echo rs6000-ibm-aix3.2.5
+	elif grep bos324 /usr/include/stdio.h >/dev/null 2>&1; then
+		echo rs6000-ibm-aix3.2.4
+	else
+		echo rs6000-ibm-aix3.2
+	fi
+	exit 0 ;;
+    *:AIX:*:4)
+	if /usr/sbin/lsattr -EHl proc0 | grep POWER >/dev/null 2>&1; then
+		IBM_ARCH=rs6000
+	else
+		IBM_ARCH=powerpc
+	fi
+	if [ -x /usr/bin/oslevel ] ; then
+		IBM_REV=`/usr/bin/oslevel`
+	else
+		IBM_REV=4.${UNAME_RELEASE}
+	fi
+	echo ${IBM_ARCH}-ibm-aix${IBM_REV}
+	exit 0 ;;
+    *:AIX:*:*)
+	echo rs6000-ibm-aix
+	exit 0 ;;
+    ibmrt:4.4BSD:*|romp-ibm:BSD:*)
+	echo romp-ibm-bsd4.4
+	exit 0 ;;
+    ibmrt:*BSD:*|romp-ibm:BSD:*)            # covers RT/PC NetBSD and
+	echo romp-ibm-bsd${UNAME_RELEASE}   # 4.3 with uname added to
+	exit 0 ;;                           # report: romp-ibm BSD 4.3
+    *:BOSX:*:*)
+	echo rs6000-bull-bosx
+	exit 0 ;;
+    DPX/2?00:B.O.S.:*:*)
+	echo m68k-bull-sysv3
+	exit 0 ;;
+    9000/[34]??:4.3bsd:1.*:*)
+	echo m68k-hp-bsd
+	exit 0 ;;
+    hp300:4.4BSD:*:* | 9000/[34]??:4.3bsd:2.*:*)
+	echo m68k-hp-bsd4.4
+	exit 0 ;;
+    9000/[3478]??:HP-UX:*:*)
+	case "${UNAME_MACHINE}" in
+	    9000/31? )            HP_ARCH=m68000 ;;
+	    9000/[34]?? )         HP_ARCH=m68k ;;
+	    9000/7?? | 9000/8?[1679] ) HP_ARCH=hppa1.1 ;;
+	    9000/8?? )            HP_ARCH=hppa1.0 ;;
+	esac
+	HPUX_REV=`echo ${UNAME_RELEASE}|sed -e 's/[^.]*.[0B]*//'`
+	echo ${HP_ARCH}-hp-hpux${HPUX_REV}
+	exit 0 ;;
+    3050*:HI-UX:*:*)
+	sed 's/^	//' << EOF >dummy.c
+	#include <unistd.h>
+	int
+	main ()
+	{
+	  long cpu = sysconf (_SC_CPU_VERSION);
+	  /* The order matters, because CPU_IS_HP_MC68K erroneously returns
+	     true for CPU_PA_RISC1_0.  CPU_IS_PA_RISC returns correct
+	     results, however.  */
+	  if (CPU_IS_PA_RISC (cpu))
+	    {
+	      switch (cpu)
+		{
+		  case CPU_PA_RISC1_0: puts ("hppa1.0-hitachi-hiuxwe2"); break;
+		  case CPU_PA_RISC1_1: puts ("hppa1.1-hitachi-hiuxwe2"); break;
+		  case CPU_PA_RISC2_0: puts ("hppa2.0-hitachi-hiuxwe2"); break;
+		  default: puts ("hppa-hitachi-hiuxwe2"); break;
+		}
+	    }
+	  else if (CPU_IS_HP_MC68K (cpu))
+	    puts ("m68k-hitachi-hiuxwe2");
+	  else puts ("unknown-hitachi-hiuxwe2");
+	  exit (0);
+	}
+EOF
+	${CC-cc} dummy.c -o dummy && ./dummy && rm dummy.c dummy && exit 0
+	rm -f dummy.c dummy
+	echo unknown-hitachi-hiuxwe2
+	exit 0 ;;
+    9000/7??:4.3bsd:*:* | 9000/8?[79]:4.3bsd:*:* )
+	echo hppa1.1-hp-bsd
+	exit 0 ;;
+    9000/8??:4.3bsd:*:*)
+	echo hppa1.0-hp-bsd
+	exit 0 ;;
+    hp7??:OSF1:*:* | hp8?[79]:OSF1:*:* )
+	echo hppa1.1-hp-osf
+	exit 0 ;;
+    hp8??:OSF1:*:*)
+	echo hppa1.0-hp-osf
+	exit 0 ;;
+    i?86:OSF1:*:*)
+	if [ -x /usr/sbin/sysversion ] ; then
+	    echo ${UNAME_MACHINE}-unknown-osf1mk
+	else
+	    echo ${UNAME_MACHINE}-unknown-osf1
+	fi
+	exit 0 ;;
+    parisc*:Lites*:*:*)
+	echo hppa1.1-hp-lites
+	exit 0 ;;
+    C1*:ConvexOS:*:* | convex:ConvexOS:C1*:*)
+	echo c1-convex-bsd
+        exit 0 ;;
+    C2*:ConvexOS:*:* | convex:ConvexOS:C2*:*)
+	if getsysinfo -f scalar_acc
+	then echo c32-convex-bsd
+	else echo c2-convex-bsd
+	fi
+        exit 0 ;;
+    C34*:ConvexOS:*:* | convex:ConvexOS:C34*:*)
+	echo c34-convex-bsd
+        exit 0 ;;
+    C38*:ConvexOS:*:* | convex:ConvexOS:C38*:*)
+	echo c38-convex-bsd
+        exit 0 ;;
+    C4*:ConvexOS:*:* | convex:ConvexOS:C4*:*)
+	echo c4-convex-bsd
+        exit 0 ;;
+    CRAY*X-MP:*:*:*)
+	echo xmp-cray-unicos
+        exit 0 ;;
+    CRAY*Y-MP:*:*:*)
+	echo ymp-cray-unicos${UNAME_RELEASE}
+	exit 0 ;;
+    CRAY*[A-Z]90:*:*:*)
+	echo ${UNAME_MACHINE}-cray-unicos${UNAME_RELEASE} \
+	| sed -e 's/CRAY.*\([A-Z]90\)/\1/' \
+	      -e y/ABCDEFGHIJKLMNOPQRSTUVWXYZ/abcdefghijklmnopqrstuvwxyz/
+	exit 0 ;;
+    CRAY*TS:*:*:*)
+	echo t90-cray-unicos${UNAME_RELEASE}
+	exit 0 ;;
+    CRAY-2:*:*:*)
+	echo cray2-cray-unicos
+        exit 0 ;;
+    F300:UNIX_System_V:*:*)
+        FUJITSU_SYS=`uname -p | tr [A-Z] [a-z] | sed -e 's/\///'`
+        FUJITSU_REL=`echo ${UNAME_RELEASE} | sed -e 's/ /_/'`
+        echo "f300-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}"
+        exit 0 ;;
+    F301:UNIX_System_V:*:*)
+       echo f301-fujitsu-uxpv`echo $UNAME_RELEASE | sed 's/ .*//'`
+       exit 0 ;;
+    hp3[0-9][05]:NetBSD:*:*)
+	echo m68k-hp-netbsd${UNAME_RELEASE}
+	exit 0 ;;
+    hp300:OpenBSD:*:*)
+	echo m68k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    i?86:BSD/386:*:* | *:BSD/OS:*:*)
+	echo ${UNAME_MACHINE}-pc-bsdi${UNAME_RELEASE}
+	exit 0 ;;
+    *:FreeBSD:*:*)
+	echo ${UNAME_MACHINE}-unknown-freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'`
+	exit 0 ;;
+    *:NetBSD:*:*)
+	echo ${UNAME_MACHINE}-unknown-netbsd`echo ${UNAME_RELEASE}|sed -e 's/[-_].*/\./'`
+	exit 0 ;;
+    *:OpenBSD:*:*)
+	echo ${UNAME_MACHINE}-unknown-openbsd`echo ${UNAME_RELEASE}|sed -e 's/[-_].*/\./'`
+	exit 0 ;;
+    i*:CYGWIN*:*)
+	echo i386-pc-cygwin32
+	exit 0 ;;
+    i*:MINGW*:*)
+	echo i386-pc-mingw32
+	exit 0 ;;
+    p*:CYGWIN*:*)
+	echo powerpcle-unknown-cygwin32
+	exit 0 ;;
+    prep*:SunOS:5.*:*)
+	echo powerpcle-unknown-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
+	exit 0 ;;
+    *:GNU:*:*)
+	echo `echo ${UNAME_MACHINE}|sed -e 's,[-/].*$,,'`-unknown-gnu`echo ${UNAME_RELEASE}|sed -e 's,/.*$,,'`
+	exit 0 ;;
+    *:Linux:*:*)
+	# The BFD linker knows what the default object file format is, so
+	# first see if it will tell us.
+	ld_help_string=`ld --help 2>&1`
+	ld_supported_emulations=`echo $ld_help_string \
+			 | sed -ne '/supported emulations:/!d
+				    s/[ 	][ 	]*/ /g
+				    s/.*supported emulations: *//
+				    s/ .*//
+				    p'`
+        case "$ld_supported_emulations" in
+	  i?86linux)  echo "${UNAME_MACHINE}-pc-linux-gnuaout"      ; exit 0 ;;
+	  i?86coff)   echo "${UNAME_MACHINE}-pc-linux-gnucoff"      ; exit 0 ;;
+	  sparclinux) echo "${UNAME_MACHINE}-unknown-linux-gnuaout" ; exit 0 ;;
+	  m68klinux)  echo "${UNAME_MACHINE}-unknown-linux-gnuaout" ; exit 0 ;;
+	  elf32ppc)   echo "powerpc-unknown-linux-gnu"              ; exit 0 ;;
+	esac
+
+	if test "${UNAME_MACHINE}" = "alpha" ; then
+		sed 's/^	//'  <<EOF >dummy.s
+		.globl main
+		.ent main
+	main:
+		.frame \$30,0,\$26,0
+		.prologue 0
+		.long 0x47e03d80 # implver $0
+		lda \$2,259
+		.long 0x47e20c21 # amask $2,$1
+		srl \$1,8,\$2
+		sll \$2,2,\$2
+		sll \$0,3,\$0
+		addl \$1,\$0,\$0
+		addl \$2,\$0,\$0
+		ret \$31,(\$26),1
+		.end main
+EOF
+		LIBC=""
+		${CC-cc} dummy.s -o dummy 2>/dev/null
+		if test "$?" = 0 ; then
+			./dummy
+			case "$?" in
+			7)
+				UNAME_MACHINE="alpha"
+				;;
+			15)
+				UNAME_MACHINE="alphaev5"
+				;;
+			14)
+				UNAME_MACHINE="alphaev56"
+				;;
+			10)
+				UNAME_MACHINE="alphapca56"
+				;;
+			16)
+				UNAME_MACHINE="alphaev6"
+				;;
+			esac	
+
+			objdump --private-headers dummy | \
+			  grep ld.so.1 > /dev/null
+			if test "$?" = 0 ; then
+				LIBC="libc1"
+			fi
+		fi	
+		rm -f dummy.s dummy
+		echo ${UNAME_MACHINE}-unknown-linux-gnu${LIBC} ; exit 0
+	elif test "${UNAME_MACHINE}" = "mips" ; then
+	  cat >dummy.c <<EOF
+main(argc, argv)
+     int argc;
+     char *argv[];
+{
+#ifdef __MIPSEB__
+  printf ("%s-unknown-linux-gnu\n", argv[1]);
+#endif
+#ifdef __MIPSEL__
+  printf ("%sel-unknown-linux-gnu\n", argv[1]);
+#endif
+  return 0;
+}
+EOF
+	  ${CC-cc} dummy.c -o dummy 2>/dev/null && ./dummy "${UNAME_MACHINE}" && rm dummy.c dummy && exit 0
+	  rm -f dummy.c dummy
+	else
+	  # Either a pre-BFD a.out linker (linux-gnuoldld)
+	  # or one that does not give us useful --help.
+	  # GCC wants to distinguish between linux-gnuoldld and linux-gnuaout.
+	  # If ld does not provide *any* "supported emulations:"
+	  # that means it is gnuoldld.
+	  echo "$ld_help_string" | grep >/dev/null 2>&1 "supported emulations:"
+	  test $? != 0 && echo "${UNAME_MACHINE}-pc-linux-gnuoldld" && exit 0
+
+	  case "${UNAME_MACHINE}" in
+	  i?86)
+	    VENDOR=pc;
+	    ;;
+	  *)
+	    VENDOR=unknown;
+	    ;;
+	  esac
+	  # Determine whether the default compiler is a.out or elf
+	  cat >dummy.c <<EOF
+#include <features.h>
+main(argc, argv)
+     int argc;
+     char *argv[];
+{
+#ifdef __ELF__
+# ifdef __GLIBC__
+#  if __GLIBC__ >= 2
+    printf ("%s-${VENDOR}-linux-gnu\n", argv[1]);
+#  else
+    printf ("%s-${VENDOR}-linux-gnulibc1\n", argv[1]);
+#  endif
+# else
+   printf ("%s-${VENDOR}-linux-gnulibc1\n", argv[1]);
+# endif
+#else
+  printf ("%s-${VENDOR}-linux-gnuaout\n", argv[1]);
+#endif
+  return 0;
+}
+EOF
+	  ${CC-cc} dummy.c -o dummy 2>/dev/null && ./dummy "${UNAME_MACHINE}" && rm dummy.c dummy && exit 0
+	  rm -f dummy.c dummy
+	fi ;;
+# ptx 4.0 does uname -s correctly, with DYNIX/ptx in there.  earlier versions
+# are messed up and put the nodename in both sysname and nodename.
+    i?86:DYNIX/ptx:4*:*)
+	echo i386-sequent-sysv4
+	exit 0 ;;
+    i?86:UNIX_SV:4.2MP:2.*)
+        # Unixware is an offshoot of SVR4, but it has its own version
+        # number series starting with 2...
+        # I am not positive that other SVR4 systems won't match this,
+	# I just have to hope.  -- rms.
+        # Use sysv4.2uw... so that sysv4* matches it.
+	echo ${UNAME_MACHINE}-pc-sysv4.2uw${UNAME_VERSION}
+	exit 0 ;;
+    i?86:*:4.*:* | i?86:SYSTEM_V:4.*:*)
+	if grep Novell /usr/include/link.h >/dev/null 2>/dev/null; then
+		echo ${UNAME_MACHINE}-univel-sysv${UNAME_RELEASE}
+	else
+		echo ${UNAME_MACHINE}-pc-sysv${UNAME_RELEASE}
+	fi
+	exit 0 ;;
+    i?86:*:3.2:*)
+	if test -f /usr/options/cb.name; then
+		UNAME_REL=`sed -n 's/.*Version //p' </usr/options/cb.name`
+		echo ${UNAME_MACHINE}-pc-isc$UNAME_REL
+	elif /bin/uname -X 2>/dev/null >/dev/null ; then
+		UNAME_REL=`(/bin/uname -X|egrep Release|sed -e 's/.*= //')`
+		(/bin/uname -X|egrep i80486 >/dev/null) && UNAME_MACHINE=i486
+		(/bin/uname -X|egrep '^Machine.*Pentium' >/dev/null) \
+			&& UNAME_MACHINE=i586
+		echo ${UNAME_MACHINE}-pc-sco$UNAME_REL
+	else
+		echo ${UNAME_MACHINE}-pc-sysv32
+	fi
+	exit 0 ;;
+    pc:*:*:*)
+        # uname -m prints for DJGPP always 'pc', but it prints nothing about
+        # the processor, so we play safe by assuming i386.
+	echo i386-pc-msdosdjgpp
+        exit 0 ;;
+    Intel:Mach:3*:*)
+	echo i386-pc-mach3
+	exit 0 ;;
+    paragon:*:*:*)
+	echo i860-intel-osf1
+	exit 0 ;;
+    i860:*:4.*:*) # i860-SVR4
+	if grep Stardent /usr/include/sys/uadmin.h >/dev/null 2>&1 ; then
+	  echo i860-stardent-sysv${UNAME_RELEASE} # Stardent Vistra i860-SVR4
+	else # Add other i860-SVR4 vendors below as they are discovered.
+	  echo i860-unknown-sysv${UNAME_RELEASE}  # Unknown i860-SVR4
+	fi
+	exit 0 ;;
+    mini*:CTIX:SYS*5:*)
+	# "miniframe"
+	echo m68010-convergent-sysv
+	exit 0 ;;
+    M68*:*:R3V[567]*:*)
+	test -r /sysV68 && echo 'm68k-motorola-sysv' && exit 0 ;;
+    3[34]??:*:4.0:3.0 | 3[34]??,*:*:4.0:3.0 | 4850:*:4.0:3.0)
+	OS_REL=''
+	test -r /etc/.relid \
+	&& OS_REL=.`sed -n 's/[^ ]* [^ ]* \([0-9][0-9]\).*/\1/p' < /etc/.relid`
+	/bin/uname -p 2>/dev/null | grep 86 >/dev/null \
+	  && echo i486-ncr-sysv4.3${OS_REL} && exit 0
+	/bin/uname -p 2>/dev/null | /bin/grep entium >/dev/null \
+	  && echo i586-ncr-sysv4.3${OS_REL} && exit 0 ;;
+    3[34]??:*:4.0:* | 3[34]??,*:*:4.0:*)
+        /bin/uname -p 2>/dev/null | grep 86 >/dev/null \
+          && echo i486-ncr-sysv4 && exit 0 ;;
+    m68*:LynxOS:2.*:*)
+	echo m68k-unknown-lynxos${UNAME_RELEASE}
+	exit 0 ;;
+    mc68030:UNIX_System_V:4.*:*)
+	echo m68k-atari-sysv4
+	exit 0 ;;
+    i?86:LynxOS:2.*:*)
+	echo i386-unknown-lynxos${UNAME_RELEASE}
+	exit 0 ;;
+    TSUNAMI:LynxOS:2.*:*)
+	echo sparc-unknown-lynxos${UNAME_RELEASE}
+	exit 0 ;;
+    rs6000:LynxOS:2.*:* | PowerPC:LynxOS:2.*:*)
+	echo rs6000-unknown-lynxos${UNAME_RELEASE}
+	exit 0 ;;
+    SM[BE]S:UNIX_SV:*:*)
+	echo mips-dde-sysv${UNAME_RELEASE}
+	exit 0 ;;
+    RM*:SINIX-*:*:*)
+	echo mips-sni-sysv4
+	exit 0 ;;
+    *:SINIX-*:*:*)
+	if uname -p 2>/dev/null >/dev/null ; then
+		UNAME_MACHINE=`(uname -p) 2>/dev/null`
+		echo ${UNAME_MACHINE}-sni-sysv4
+	else
+		echo ns32k-sni-sysv
+	fi
+	exit 0 ;;
+    PENTIUM:CPunix:4.0*:*) # Unisys `ClearPath HMP IX 4000' SVR4/MP effort
+                           # says <Richard.M.Bartel@ccMail.Census.GOV>
+        echo i586-unisys-sysv4
+        exit 0 ;;
+    *:UNIX_System_V:4*:FTX*)
+	# From Gerald Hewes <hewes@openmarket.com>.
+	# How about differentiating between stratus architectures? -djm
+	echo hppa1.1-stratus-sysv4
+	exit 0 ;;
+    *:*:*:FTX*)
+	# From seanf@swdc.stratus.com.
+	echo i860-stratus-sysv4
+	exit 0 ;;
+    mc68*:A/UX:*:*)
+	echo m68k-apple-aux${UNAME_RELEASE}
+	exit 0 ;;
+    news*:NEWS-OS:*:6*)
+	echo mips-sony-newsos6
+	exit 0 ;;
+    R3000:*System_V*:*:* | R4000:UNIX_SYSV:*:*)
+	if [ -d /usr/nec ]; then
+	        echo mips-nec-sysv${UNAME_RELEASE}
+	else
+	        echo mips-unknown-sysv${UNAME_RELEASE}
+	fi
+        exit 0 ;;
+esac
+
+#echo '(No uname command or uname output not recognized.)' 1>&2
+#echo "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" 1>&2
+
+cat >dummy.c <<EOF
+#ifdef _SEQUENT_
+# include <sys/types.h>
+# include <sys/utsname.h>
+#endif
+main ()
+{
+#if defined (sony)
+#if defined (MIPSEB)
+  /* BFD wants "bsd" instead of "newsos".  Perhaps BFD should be changed,
+     I don't know....  */
+  printf ("mips-sony-bsd\n"); exit (0);
+#else
+#include <sys/param.h>
+  printf ("m68k-sony-newsos%s\n",
+#ifdef NEWSOS4
+          "4"
+#else
+	  ""
+#endif
+         ); exit (0);
+#endif
+#endif
+
+#if defined (__arm) && defined (__acorn) && defined (__unix)
+  printf ("arm-acorn-riscix"); exit (0);
+#endif
+
+#if defined (hp300) && !defined (hpux)
+  printf ("m68k-hp-bsd\n"); exit (0);
+#endif
+
+#if defined (NeXT)
+#if !defined (__ARCHITECTURE__)
+#define __ARCHITECTURE__ "m68k"
+#endif
+  int version;
+  version=`(hostinfo | sed -n 's/.*NeXT Mach \([0-9]*\).*/\1/p') 2>/dev/null`;
+  printf ("%s-next-nextstep%d\n", __ARCHITECTURE__, version);
+  exit (0);
+#endif
+
+#if defined (MULTIMAX) || defined (n16)
+#if defined (UMAXV)
+  printf ("ns32k-encore-sysv\n"); exit (0);
+#else
+#if defined (CMU)
+  printf ("ns32k-encore-mach\n"); exit (0);
+#else
+  printf ("ns32k-encore-bsd\n"); exit (0);
+#endif
+#endif
+#endif
+
+#if defined (__386BSD__)
+  printf ("i386-pc-bsd\n"); exit (0);
+#endif
+
+#if defined (sequent)
+#if defined (i386)
+  printf ("i386-sequent-dynix\n"); exit (0);
+#endif
+#if defined (ns32000)
+  printf ("ns32k-sequent-dynix\n"); exit (0);
+#endif
+#endif
+
+#if defined (_SEQUENT_)
+    struct utsname un;
+
+    uname(&un);
+
+    if (strncmp(un.version, "V2", 2) == 0) {
+	printf ("i386-sequent-ptx2\n"); exit (0);
+    }
+    if (strncmp(un.version, "V1", 2) == 0) { /* XXX is V1 correct? */
+	printf ("i386-sequent-ptx1\n"); exit (0);
+    }
+    printf ("i386-sequent-ptx\n"); exit (0);
+
+#endif
+
+#if defined (vax)
+#if !defined (ultrix)
+  printf ("vax-dec-bsd\n"); exit (0);
+#else
+  printf ("vax-dec-ultrix\n"); exit (0);
+#endif
+#endif
+
+#if defined (alliant) && defined (i860)
+  printf ("i860-alliant-bsd\n"); exit (0);
+#endif
+
+  exit (1);
+}
+EOF
+
+${CC-cc} dummy.c -o dummy 2>/dev/null && ./dummy && rm dummy.c dummy && exit 0
+rm -f dummy.c dummy
+
+# Apollos put the system type in the environment.
+
+test -d /usr/apollo && { echo ${ISP}-apollo-${SYSTYPE}; exit 0; }
+
+# Convex versions that predate uname can use getsysinfo(1)
+
+if [ -x /usr/convex/getsysinfo ]
+then
+    case `getsysinfo -f cpu_type` in
+    c1*)
+	echo c1-convex-bsd
+	exit 0 ;;
+    c2*)
+	if getsysinfo -f scalar_acc
+	then echo c32-convex-bsd
+	else echo c2-convex-bsd
+	fi
+	exit 0 ;;
+    c34*)
+	echo c34-convex-bsd
+	exit 0 ;;
+    c38*)
+	echo c38-convex-bsd
+	exit 0 ;;
+    c4*)
+	echo c4-convex-bsd
+	exit 0 ;;
+    esac
+fi
+
+#echo '(Unable to guess system type)' 1>&2
+
+exit 1
diff --git a/libjpegtwrp/config.sub b/libjpegtwrp/config.sub
new file mode 100755
index 000000000..213a6d47d
--- /dev/null
+++ b/libjpegtwrp/config.sub
@@ -0,0 +1,954 @@
+#! /bin/sh
+# Configuration validation subroutine script, version 1.1.
+#   Copyright (C) 1991, 92, 93, 94, 95, 96, 1997 Free Software Foundation, Inc.
+# This file is (in principle) common to ALL GNU software.
+# The presence of a machine in this file suggests that SOME GNU software
+# can handle that machine.  It does not imply ALL GNU software can.
+#
+# This file is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place - Suite 330,
+# Boston, MA 02111-1307, USA.
+
+# As a special exception to the GNU General Public License, if you
+# distribute this file as part of a program that contains a
+# configuration script generated by Autoconf, you may include it under
+# the same distribution terms that you use for the rest of that program.
+
+# Configuration subroutine to validate and canonicalize a configuration type.
+# Supply the specified configuration type as an argument.
+# If it is invalid, we print an error message on stderr and exit with code 1.
+# Otherwise, we print the canonical config type on stdout and succeed.
+
+# This file is supposed to be the same for all GNU packages
+# and recognize all the CPU types, system types and aliases
+# that are meaningful with *any* GNU software.
+# Each package is responsible for reporting which valid configurations
+# it does not support.  The user should be able to distinguish
+# a failure to support a valid configuration from a meaningless
+# configuration.
+
+# The goal of this file is to map all the various variations of a given
+# machine specification into a single specification in the form:
+#	CPU_TYPE-MANUFACTURER-OPERATING_SYSTEM
+# or in some cases, the newer four-part form:
+#	CPU_TYPE-MANUFACTURER-KERNEL-OPERATING_SYSTEM
+# It is wrong to echo any other type of specification.
+
+if [ x$1 = x ]
+then
+	echo Configuration name missing. 1>&2
+	echo "Usage: $0 CPU-MFR-OPSYS" 1>&2
+	echo "or     $0 ALIAS" 1>&2
+	echo where ALIAS is a recognized configuration type. 1>&2
+	exit 1
+fi
+
+# First pass through any local machine types.
+case $1 in
+	*local*)
+		echo $1
+		exit 0
+		;;
+	*)
+	;;
+esac
+
+# Separate what the user gave into CPU-COMPANY and OS or KERNEL-OS (if any).
+# Here we must recognize all the valid KERNEL-OS combinations.
+maybe_os=`echo $1 | sed 's/^\(.*\)-\([^-]*-[^-]*\)$/\2/'`
+case $maybe_os in
+  linux-gnu*)
+    os=-$maybe_os
+    basic_machine=`echo $1 | sed 's/^\(.*\)-\([^-]*-[^-]*\)$/\1/'`
+    ;;
+  *)
+    basic_machine=`echo $1 | sed 's/-[^-]*$//'`
+    if [ $basic_machine != $1 ]
+    then os=`echo $1 | sed 's/.*-/-/'`
+    else os=; fi
+    ;;
+esac
+
+### Let's recognize common machines as not being operating systems so
+### that things like config.sub decstation-3100 work.  We also
+### recognize some manufacturers as not being operating systems, so we
+### can provide default operating systems below.
+case $os in
+	-sun*os*)
+		# Prevent following clause from handling this invalid input.
+		;;
+	-dec* | -mips* | -sequent* | -encore* | -pc532* | -sgi* | -sony* | \
+	-att* | -7300* | -3300* | -delta* | -motorola* | -sun[234]* | \
+	-unicom* | -ibm* | -next | -hp | -isi* | -apollo | -altos* | \
+	-convergent* | -ncr* | -news | -32* | -3600* | -3100* | -hitachi* |\
+	-c[123]* | -convex* | -sun | -crds | -omron* | -dg | -ultra | -tti* | \
+	-harris | -dolphin | -highlevel | -gould | -cbm | -ns | -masscomp | \
+	-apple)
+		os=
+		basic_machine=$1
+		;;
+	-hiux*)
+		os=-hiuxwe2
+		;;
+	-sco5)
+		os=sco3.2v5
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-sco4)
+		os=-sco3.2v4
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-sco3.2.[4-9]*)
+		os=`echo $os | sed -e 's/sco3.2./sco3.2v/'`
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-sco3.2v[4-9]*)
+		# Don't forget version if it is 3.2v4 or newer.
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-sco*)
+		os=-sco3.2v2
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-isc)
+		os=-isc2.2
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-clix*)
+		basic_machine=clipper-intergraph
+		;;
+	-isc*)
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-lynx*)
+		os=-lynxos
+		;;
+	-ptx*)
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-sequent/'`
+		;;
+	-windowsnt*)
+		os=`echo $os | sed -e 's/windowsnt/winnt/'`
+		;;
+	-psos*)
+		os=-psos
+		;;
+esac
+
+# Decode aliases for certain CPU-COMPANY combinations.
+case $basic_machine in
+	# Recognize the basic CPU types without company name.
+	# Some are omitted here because they have special meanings below.
+	tahoe | i860 | m32r | m68k | m68000 | m88k | ns32k | arc | arm \
+		| arme[lb] | pyramid | mn10200 | mn10300 \
+		| tron | a29k | 580 | i960 | h8300 | hppa | hppa1.0 | hppa1.1 \
+		| alpha | alphaev5 | alphaev56 | we32k | ns16k | clipper \
+		| i370 | sh | powerpc | powerpcle | 1750a | dsp16xx | pdp11 \
+		| mips64 | mipsel | mips64el | mips64orion | mips64orionel \
+		| mipstx39 | mipstx39el \
+		| sparc | sparclet | sparclite | sparc64 | v850)
+		basic_machine=$basic_machine-unknown
+		;;
+	# We use `pc' rather than `unknown'
+	# because (1) that's what they normally are, and
+	# (2) the word "unknown" tends to confuse beginning users.
+	i[3456]86)
+	  basic_machine=$basic_machine-pc
+	  ;;
+	# Object if more than one company name word.
+	*-*-*)
+		echo Invalid configuration \`$1\': machine \`$basic_machine\' not recognized 1>&2
+		exit 1
+		;;
+	# Recognize the basic CPU types with company name.
+	vax-* | tahoe-* | i[3456]86-* | i860-* | m32r-* | m68k-* | m68000-* \
+	      | m88k-* | sparc-* | ns32k-* | fx80-* | arc-* | arm-* | c[123]* \
+	      | mips-* | pyramid-* | tron-* | a29k-* | romp-* | rs6000-* \
+	      | power-* | none-* | 580-* | cray2-* | h8300-* | i960-* \
+	      | xmp-* | ymp-* | hppa-* | hppa1.0-* | hppa1.1-* \
+	      | alpha-* | alphaev5-* | alphaev56-* | we32k-* | cydra-* \
+	      | ns16k-* | pn-* | np1-* | xps100-* | clipper-* | orion-* \
+	      | sparclite-* | pdp11-* | sh-* | powerpc-* | powerpcle-* \
+	      | sparc64-* | mips64-* | mipsel-* \
+	      | mips64el-* | mips64orion-* | mips64orionel-*  \
+	      | mipstx39-* | mipstx39el-* \
+	      | f301-*)
+		;;
+	# Recognize the various machine names and aliases which stand
+	# for a CPU type and a company and sometimes even an OS.
+	3b1 | 7300 | 7300-att | att-7300 | pc7300 | safari | unixpc)
+		basic_machine=m68000-att
+		;;
+	3b*)
+		basic_machine=we32k-att
+		;;
+	alliant | fx80)
+		basic_machine=fx80-alliant
+		;;
+	altos | altos3068)
+		basic_machine=m68k-altos
+		;;
+	am29k)
+		basic_machine=a29k-none
+		os=-bsd
+		;;
+	amdahl)
+		basic_machine=580-amdahl
+		os=-sysv
+		;;
+	amiga | amiga-*)
+		basic_machine=m68k-cbm
+		;;
+	amigaos | amigados)
+		basic_machine=m68k-cbm
+		os=-amigaos
+		;;
+	amigaunix | amix)
+		basic_machine=m68k-cbm
+		os=-sysv4
+		;;
+	apollo68)
+		basic_machine=m68k-apollo
+		os=-sysv
+		;;
+	aux)
+		basic_machine=m68k-apple
+		os=-aux
+		;;
+	balance)
+		basic_machine=ns32k-sequent
+		os=-dynix
+		;;
+	convex-c1)
+		basic_machine=c1-convex
+		os=-bsd
+		;;
+	convex-c2)
+		basic_machine=c2-convex
+		os=-bsd
+		;;
+	convex-c32)
+		basic_machine=c32-convex
+		os=-bsd
+		;;
+	convex-c34)
+		basic_machine=c34-convex
+		os=-bsd
+		;;
+	convex-c38)
+		basic_machine=c38-convex
+		os=-bsd
+		;;
+	cray | ymp)
+		basic_machine=ymp-cray
+		os=-unicos
+		;;
+	cray2)
+		basic_machine=cray2-cray
+		os=-unicos
+		;;
+	[ctj]90-cray)
+		basic_machine=c90-cray
+		os=-unicos
+		;;
+	crds | unos)
+		basic_machine=m68k-crds
+		;;
+	da30 | da30-*)
+		basic_machine=m68k-da30
+		;;
+	decstation | decstation-3100 | pmax | pmax-* | pmin | dec3100 | decstatn)
+		basic_machine=mips-dec
+		;;
+	delta | 3300 | motorola-3300 | motorola-delta \
+	      | 3300-motorola | delta-motorola)
+		basic_machine=m68k-motorola
+		;;
+	delta88)
+		basic_machine=m88k-motorola
+		os=-sysv3
+		;;
+	dpx20 | dpx20-*)
+		basic_machine=rs6000-bull
+		os=-bosx
+		;;
+	dpx2* | dpx2*-bull)
+		basic_machine=m68k-bull
+		os=-sysv3
+		;;
+	ebmon29k)
+		basic_machine=a29k-amd
+		os=-ebmon
+		;;
+	elxsi)
+		basic_machine=elxsi-elxsi
+		os=-bsd
+		;;
+	encore | umax | mmax)
+		basic_machine=ns32k-encore
+		;;
+	fx2800)
+		basic_machine=i860-alliant
+		;;
+	genix)
+		basic_machine=ns32k-ns
+		;;
+	gmicro)
+		basic_machine=tron-gmicro
+		os=-sysv
+		;;
+	h3050r* | hiux*)
+		basic_machine=hppa1.1-hitachi
+		os=-hiuxwe2
+		;;
+	h8300hms)
+		basic_machine=h8300-hitachi
+		os=-hms
+		;;
+	harris)
+		basic_machine=m88k-harris
+		os=-sysv3
+		;;
+	hp300-*)
+		basic_machine=m68k-hp
+		;;
+	hp300bsd)
+		basic_machine=m68k-hp
+		os=-bsd
+		;;
+	hp300hpux)
+		basic_machine=m68k-hp
+		os=-hpux
+		;;
+	hp9k2[0-9][0-9] | hp9k31[0-9])
+		basic_machine=m68000-hp
+		;;
+	hp9k3[2-9][0-9])
+		basic_machine=m68k-hp
+		;;
+	hp9k7[0-9][0-9] | hp7[0-9][0-9] | hp9k8[0-9]7 | hp8[0-9]7)
+		basic_machine=hppa1.1-hp
+		;;
+	hp9k8[0-9][0-9] | hp8[0-9][0-9])
+		basic_machine=hppa1.0-hp
+		;;
+	hppa-next)
+		os=-nextstep3
+		;;
+	i370-ibm* | ibm*)
+		basic_machine=i370-ibm
+		os=-mvs
+		;;
+# I'm not sure what "Sysv32" means.  Should this be sysv3.2?
+	i[3456]86v32)
+		basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'`
+		os=-sysv32
+		;;
+	i[3456]86v4*)
+		basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'`
+		os=-sysv4
+		;;
+	i[3456]86v)
+		basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'`
+		os=-sysv
+		;;
+	i[3456]86sol2)
+		basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'`
+		os=-solaris2
+		;;
+	iris | iris4d)
+		basic_machine=mips-sgi
+		case $os in
+		    -irix*)
+			;;
+		    *)
+			os=-irix4
+			;;
+		esac
+		;;
+	isi68 | isi)
+		basic_machine=m68k-isi
+		os=-sysv
+		;;
+	m88k-omron*)
+		basic_machine=m88k-omron
+		;;
+	magnum | m3230)
+		basic_machine=mips-mips
+		os=-sysv
+		;;
+	merlin)
+		basic_machine=ns32k-utek
+		os=-sysv
+		;;
+	miniframe)
+		basic_machine=m68000-convergent
+		;;
+	mipsel*-linux*)
+		basic_machine=mipsel-unknown
+		os=-linux-gnu
+		;;
+	mips*-linux*)
+		basic_machine=mips-unknown
+		os=-linux-gnu
+		;;
+	mips3*-*)
+		basic_machine=`echo $basic_machine | sed -e 's/mips3/mips64/'`
+		;;
+	mips3*)
+		basic_machine=`echo $basic_machine | sed -e 's/mips3/mips64/'`-unknown
+		;;
+	ncr3000)
+		basic_machine=i486-ncr
+		os=-sysv4
+		;;
+	news | news700 | news800 | news900)
+		basic_machine=m68k-sony
+		os=-newsos
+		;;
+	news1000)
+		basic_machine=m68030-sony
+		os=-newsos
+		;;
+	news-3600 | risc-news)
+		basic_machine=mips-sony
+		os=-newsos
+		;;
+	next | m*-next )
+		basic_machine=m68k-next
+		case $os in
+		    -nextstep* )
+			;;
+		    -ns2*)
+		      os=-nextstep2
+			;;
+		    *)
+		      os=-nextstep3
+			;;
+		esac
+		;;
+	nh3000)
+		basic_machine=m68k-harris
+		os=-cxux
+		;;
+	nh[45]000)
+		basic_machine=m88k-harris
+		os=-cxux
+		;;
+	nindy960)
+		basic_machine=i960-intel
+		os=-nindy
+		;;
+	np1)
+		basic_machine=np1-gould
+		;;
+	pa-hitachi)
+		basic_machine=hppa1.1-hitachi
+		os=-hiuxwe2
+		;;
+	paragon)
+		basic_machine=i860-intel
+		os=-osf
+		;;
+	pbd)
+		basic_machine=sparc-tti
+		;;
+	pbb)
+		basic_machine=m68k-tti
+		;;
+        pc532 | pc532-*)
+		basic_machine=ns32k-pc532
+		;;
+	pentium | p5)
+		basic_machine=i586-intel
+		;;
+	pentiumpro | p6)
+		basic_machine=i686-intel
+		;;
+	pentium-* | p5-*)
+		basic_machine=i586-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	pentiumpro-* | p6-*)
+		basic_machine=i686-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	k5)
+		# We don't have specific support for AMD's K5 yet, so just call it a Pentium
+		basic_machine=i586-amd
+		;;
+	nexen)
+		# We don't have specific support for Nexgen yet, so just call it a Pentium
+		basic_machine=i586-nexgen
+		;;
+	pn)
+		basic_machine=pn-gould
+		;;
+	power)	basic_machine=rs6000-ibm
+		;;
+	ppc)	basic_machine=powerpc-unknown
+	        ;;
+	ppc-*)	basic_machine=powerpc-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	ppcle | powerpclittle | ppc-le | powerpc-little)
+		basic_machine=powerpcle-unknown
+	        ;;
+	ppcle-* | powerpclittle-*)
+		basic_machine=powerpcle-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	ps2)
+		basic_machine=i386-ibm
+		;;
+	rm[46]00)
+		basic_machine=mips-siemens
+		;;
+	rtpc | rtpc-*)
+		basic_machine=romp-ibm
+		;;
+	sequent)
+		basic_machine=i386-sequent
+		;;
+	sh)
+		basic_machine=sh-hitachi
+		os=-hms
+		;;
+	sps7)
+		basic_machine=m68k-bull
+		os=-sysv2
+		;;
+	spur)
+		basic_machine=spur-unknown
+		;;
+	sun2)
+		basic_machine=m68000-sun
+		;;
+	sun2os3)
+		basic_machine=m68000-sun
+		os=-sunos3
+		;;
+	sun2os4)
+		basic_machine=m68000-sun
+		os=-sunos4
+		;;
+	sun3os3)
+		basic_machine=m68k-sun
+		os=-sunos3
+		;;
+	sun3os4)
+		basic_machine=m68k-sun
+		os=-sunos4
+		;;
+	sun4os3)
+		basic_machine=sparc-sun
+		os=-sunos3
+		;;
+	sun4os4)
+		basic_machine=sparc-sun
+		os=-sunos4
+		;;
+	sun4sol2)
+		basic_machine=sparc-sun
+		os=-solaris2
+		;;
+	sun3 | sun3-*)
+		basic_machine=m68k-sun
+		;;
+	sun4)
+		basic_machine=sparc-sun
+		;;
+	sun386 | sun386i | roadrunner)
+		basic_machine=i386-sun
+		;;
+	symmetry)
+		basic_machine=i386-sequent
+		os=-dynix
+		;;
+	tx39)
+		basic_machine=mipstx39-unknown
+		;;
+	tx39el)
+		basic_machine=mipstx39el-unknown
+		;;
+	tower | tower-32)
+		basic_machine=m68k-ncr
+		;;
+	udi29k)
+		basic_machine=a29k-amd
+		os=-udi
+		;;
+	ultra3)
+		basic_machine=a29k-nyu
+		os=-sym1
+		;;
+	vaxv)
+		basic_machine=vax-dec
+		os=-sysv
+		;;
+	vms)
+		basic_machine=vax-dec
+		os=-vms
+		;;
+	vpp*|vx|vx-*)
+               basic_machine=f301-fujitsu
+               ;;
+	vxworks960)
+		basic_machine=i960-wrs
+		os=-vxworks
+		;;
+	vxworks68)
+		basic_machine=m68k-wrs
+		os=-vxworks
+		;;
+	vxworks29k)
+		basic_machine=a29k-wrs
+		os=-vxworks
+		;;
+	xmp)
+		basic_machine=xmp-cray
+		os=-unicos
+		;;
+        xps | xps100)
+		basic_machine=xps100-honeywell
+		;;
+	none)
+		basic_machine=none-none
+		os=-none
+		;;
+
+# Here we handle the default manufacturer of certain CPU types.  It is in
+# some cases the only manufacturer, in others, it is the most popular.
+	mips)
+		if [ x$os = x-linux-gnu ]; then
+			basic_machine=mips-unknown
+		else
+			basic_machine=mips-mips
+		fi
+		;;
+	romp)
+		basic_machine=romp-ibm
+		;;
+	rs6000)
+		basic_machine=rs6000-ibm
+		;;
+	vax)
+		basic_machine=vax-dec
+		;;
+	pdp11)
+		basic_machine=pdp11-dec
+		;;
+	we32k)
+		basic_machine=we32k-att
+		;;
+	sparc)
+		basic_machine=sparc-sun
+		;;
+        cydra)
+		basic_machine=cydra-cydrome
+		;;
+	orion)
+		basic_machine=orion-highlevel
+		;;
+	orion105)
+		basic_machine=clipper-highlevel
+		;;
+	*)
+		echo Invalid configuration \`$1\': machine \`$basic_machine\' not recognized 1>&2
+		exit 1
+		;;
+esac
+
+# Here we canonicalize certain aliases for manufacturers.
+case $basic_machine in
+	*-digital*)
+		basic_machine=`echo $basic_machine | sed 's/digital.*/dec/'`
+		;;
+	*-commodore*)
+		basic_machine=`echo $basic_machine | sed 's/commodore.*/cbm/'`
+		;;
+	*)
+		;;
+esac
+
+# Decode manufacturer-specific aliases for certain operating systems.
+
+if [ x"$os" != x"" ]
+then
+case $os in
+        # First match some system type aliases
+        # that might get confused with valid system types.
+	# -solaris* is a basic system type, with this one exception.
+	-solaris1 | -solaris1.*)
+		os=`echo $os | sed -e 's|solaris1|sunos4|'`
+		;;
+	-solaris)
+		os=-solaris2
+		;;
+	-svr4*)
+		os=-sysv4
+		;;
+	-unixware*)
+		os=-sysv4.2uw
+		;;
+	-gnu/linux*)
+		os=`echo $os | sed -e 's|gnu/linux|linux-gnu|'`
+		;;
+	# First accept the basic system types.
+	# The portable systems comes first.
+	# Each alternative MUST END IN A *, to match a version number.
+	# -sysv* is not here because it comes later, after sysvr4.
+	-gnu* | -bsd* | -mach* | -minix* | -genix* | -ultrix* | -irix* \
+	      | -*vms* | -sco* | -esix* | -isc* | -aix* | -sunos | -sunos[34]*\
+	      | -hpux* | -unos* | -osf* | -luna* | -dgux* | -solaris* | -sym* \
+	      | -amigaos* | -amigados* | -msdos* | -newsos* | -unicos* | -aof* \
+	      | -aos* \
+	      | -nindy* | -vxsim* | -vxworks* | -ebmon* | -hms* | -mvs* \
+	      | -clix* | -riscos* | -uniplus* | -iris* | -rtu* | -xenix* \
+	      | -hiux* | -386bsd* | -netbsd* | -openbsd* | -freebsd* | -riscix* \
+	      | -lynxos* | -bosx* | -nextstep* | -cxux* | -aout* | -elf* \
+	      | -ptx* | -coff* | -ecoff* | -winnt* | -domain* | -vsta* \
+	      | -udi* | -eabi* | -lites* | -ieee* | -go32* | -aux* \
+	      | -cygwin32* | -pe* | -psos* | -moss* | -proelf* | -rtems* \
+	      | -mingw32* | -linux-gnu* | -uxpv*)
+	# Remember, each alternative MUST END IN *, to match a version number.
+		;;
+	-linux*)
+		os=`echo $os | sed -e 's|linux|linux-gnu|'`
+		;;
+	-sunos5*)
+		os=`echo $os | sed -e 's|sunos5|solaris2|'`
+		;;
+	-sunos6*)
+		os=`echo $os | sed -e 's|sunos6|solaris3|'`
+		;;
+	-osfrose*)
+		os=-osfrose
+		;;
+	-osf*)
+		os=-osf
+		;;
+	-utek*)
+		os=-bsd
+		;;
+	-dynix*)
+		os=-bsd
+		;;
+	-acis*)
+		os=-aos
+		;;
+	-ctix* | -uts*)
+		os=-sysv
+		;;
+	-ns2 )
+	        os=-nextstep2
+		;;
+	# Preserve the version number of sinix5.
+	-sinix5.*)
+		os=`echo $os | sed -e 's|sinix|sysv|'`
+		;;
+	-sinix*)
+		os=-sysv4
+		;;
+	-triton*)
+		os=-sysv3
+		;;
+	-oss*)
+		os=-sysv3
+		;;
+	-svr4)
+		os=-sysv4
+		;;
+	-svr3)
+		os=-sysv3
+		;;
+	-sysvr4)
+		os=-sysv4
+		;;
+	# This must come after -sysvr4.
+	-sysv*)
+		;;
+	-xenix)
+		os=-xenix
+		;;
+	-none)
+		;;
+	*)
+		# Get rid of the `-' at the beginning of $os.
+		os=`echo $os | sed 's/[^-]*-//'`
+		echo Invalid configuration \`$1\': system \`$os\' not recognized 1>&2
+		exit 1
+		;;
+esac
+else
+
+# Here we handle the default operating systems that come with various machines.
+# The value should be what the vendor currently ships out the door with their
+# machine or put another way, the most popular os provided with the machine.
+
+# Note that if you're going to try to match "-MANUFACTURER" here (say,
+# "-sun"), then you have to tell the case statement up towards the top
+# that MANUFACTURER isn't an operating system.  Otherwise, code above
+# will signal an error saying that MANUFACTURER isn't an operating
+# system, and we'll never get to this point.
+
+case $basic_machine in
+	*-acorn)
+		os=-riscix1.2
+		;;
+	arm*-semi)
+		os=-aout
+		;;
+        pdp11-*)
+		os=-none
+		;;
+	*-dec | vax-*)
+		os=-ultrix4.2
+		;;
+	m68*-apollo)
+		os=-domain
+		;;
+	i386-sun)
+		os=-sunos4.0.2
+		;;
+	m68000-sun)
+		os=-sunos3
+		# This also exists in the configure program, but was not the
+		# default.
+		# os=-sunos4
+		;;
+	*-tti)	# must be before sparc entry or we get the wrong os.
+		os=-sysv3
+		;;
+	sparc-* | *-sun)
+		os=-sunos4.1.1
+		;;
+	*-ibm)
+		os=-aix
+		;;
+	*-hp)
+		os=-hpux
+		;;
+	*-hitachi)
+		os=-hiux
+		;;
+	i860-* | *-att | *-ncr | *-altos | *-motorola | *-convergent)
+		os=-sysv
+		;;
+	*-cbm)
+		os=-amigaos
+		;;
+	*-dg)
+		os=-dgux
+		;;
+	*-dolphin)
+		os=-sysv3
+		;;
+	m68k-ccur)
+		os=-rtu
+		;;
+	m88k-omron*)
+		os=-luna
+		;;
+	*-next )
+		os=-nextstep
+		;;
+	*-sequent)
+		os=-ptx
+		;;
+	*-crds)
+		os=-unos
+		;;
+	*-ns)
+		os=-genix
+		;;
+	i370-*)
+		os=-mvs
+		;;
+	*-next)
+		os=-nextstep3
+		;;
+        *-gould)
+		os=-sysv
+		;;
+        *-highlevel)
+		os=-bsd
+		;;
+	*-encore)
+		os=-bsd
+		;;
+        *-sgi)
+		os=-irix
+		;;
+        *-siemens)
+		os=-sysv4
+		;;
+	*-masscomp)
+		os=-rtu
+		;;
+	f301-fujitsu)
+		os=-uxpv
+		;;
+	*)
+		os=-none
+		;;
+esac
+fi
+
+# Here we handle the case where we know the os, and the CPU type, but not the
+# manufacturer.  We pick the logical manufacturer.
+vendor=unknown
+case $basic_machine in
+	*-unknown)
+		case $os in
+			-riscix*)
+				vendor=acorn
+				;;
+			-sunos*)
+				vendor=sun
+				;;
+			-aix*)
+				vendor=ibm
+				;;
+			-hpux*)
+				vendor=hp
+				;;
+			-hiux*)
+				vendor=hitachi
+				;;
+			-unos*)
+				vendor=crds
+				;;
+			-dgux*)
+				vendor=dg
+				;;
+			-luna*)
+				vendor=omron
+				;;
+			-genix*)
+				vendor=ns
+				;;
+			-mvs*)
+				vendor=ibm
+				;;
+			-ptx*)
+				vendor=sequent
+				;;
+			-vxsim* | -vxworks*)
+				vendor=wrs
+				;;
+			-aux*)
+				vendor=apple
+				;;
+		esac
+		basic_machine=`echo $basic_machine | sed "s/unknown/$vendor/"`
+		;;
+esac
+
+echo $basic_machine$os
diff --git a/libjpegtwrp/configure b/libjpegtwrp/configure
new file mode 100755
index 000000000..35c9db5ca
--- /dev/null
+++ b/libjpegtwrp/configure
@@ -0,0 +1,2011 @@
+#! /bin/sh
+
+# Guess values for system-dependent variables and create Makefiles.
+# Generated automatically using autoconf version 2.12 
+# Copyright (C) 1992, 93, 94, 95, 96 Free Software Foundation, Inc.
+#
+# This configure script is free software; the Free Software Foundation
+# gives unlimited permission to copy, distribute and modify it.
+
+# Defaults:
+ac_help=
+ac_default_prefix=/usr/local
+# Any additions from configure.in:
+ac_help="$ac_help
+  --enable-shared         build shared library using GNU libtool"
+ac_help="$ac_help
+  --enable-static         build static library using GNU libtool"
+ac_help="$ac_help
+  --enable-maxmem[=N]     enable use of temp files, set max mem usage to N MB"
+ac_help="$ac_help
+"
+
+# Initialize some variables set by options.
+# The variables have the same names as the options, with
+# dashes changed to underlines.
+build=NONE
+cache_file=./config.cache
+exec_prefix=NONE
+host=NONE
+no_create=
+nonopt=NONE
+no_recursion=
+prefix=NONE
+program_prefix=NONE
+program_suffix=NONE
+program_transform_name=s,x,x,
+silent=
+site=
+srcdir=
+target=NONE
+verbose=
+x_includes=NONE
+x_libraries=NONE
+bindir='${exec_prefix}/bin'
+sbindir='${exec_prefix}/sbin'
+libexecdir='${exec_prefix}/libexec'
+datadir='${prefix}/share'
+sysconfdir='${prefix}/etc'
+sharedstatedir='${prefix}/com'
+localstatedir='${prefix}/var'
+libdir='${exec_prefix}/lib'
+includedir='${prefix}/include'
+oldincludedir='/usr/include'
+infodir='${prefix}/info'
+mandir='${prefix}/man'
+
+# Initialize some other variables.
+subdirs=
+MFLAGS= MAKEFLAGS=
+# Maximum number of lines to put in a shell here document.
+ac_max_here_lines=12
+
+ac_prev=
+for ac_option
+do
+
+  # If the previous option needs an argument, assign it.
+  if test -n "$ac_prev"; then
+    eval "$ac_prev=\$ac_option"
+    ac_prev=
+    continue
+  fi
+
+  case "$ac_option" in
+  -*=*) ac_optarg=`echo "$ac_option" | sed 's/[-_a-zA-Z0-9]*=//'` ;;
+  *) ac_optarg= ;;
+  esac
+
+  # Accept the important Cygnus configure options, so we can diagnose typos.
+
+  case "$ac_option" in
+
+  -bindir | --bindir | --bindi | --bind | --bin | --bi)
+    ac_prev=bindir ;;
+  -bindir=* | --bindir=* | --bindi=* | --bind=* | --bin=* | --bi=*)
+    bindir="$ac_optarg" ;;
+
+  -build | --build | --buil | --bui | --bu)
+    ac_prev=build ;;
+  -build=* | --build=* | --buil=* | --bui=* | --bu=*)
+    build="$ac_optarg" ;;
+
+  -cache-file | --cache-file | --cache-fil | --cache-fi \
+  | --cache-f | --cache- | --cache | --cach | --cac | --ca | --c)
+    ac_prev=cache_file ;;
+  -cache-file=* | --cache-file=* | --cache-fil=* | --cache-fi=* \
+  | --cache-f=* | --cache-=* | --cache=* | --cach=* | --cac=* | --ca=* | --c=*)
+    cache_file="$ac_optarg" ;;
+
+  -datadir | --datadir | --datadi | --datad | --data | --dat | --da)
+    ac_prev=datadir ;;
+  -datadir=* | --datadir=* | --datadi=* | --datad=* | --data=* | --dat=* \
+  | --da=*)
+    datadir="$ac_optarg" ;;
+
+  -disable-* | --disable-*)
+    ac_feature=`echo $ac_option|sed -e 's/-*disable-//'`
+    # Reject names that are not valid shell variable names.
+    if test -n "`echo $ac_feature| sed 's/[-a-zA-Z0-9_]//g'`"; then
+      { echo "configure: error: $ac_feature: invalid feature name" 1>&2; exit 1; }
+    fi
+    ac_feature=`echo $ac_feature| sed 's/-/_/g'`
+    eval "enable_${ac_feature}=no" ;;
+
+  -enable-* | --enable-*)
+    ac_feature=`echo $ac_option|sed -e 's/-*enable-//' -e 's/=.*//'`
+    # Reject names that are not valid shell variable names.
+    if test -n "`echo $ac_feature| sed 's/[-_a-zA-Z0-9]//g'`"; then
+      { echo "configure: error: $ac_feature: invalid feature name" 1>&2; exit 1; }
+    fi
+    ac_feature=`echo $ac_feature| sed 's/-/_/g'`
+    case "$ac_option" in
+      *=*) ;;
+      *) ac_optarg=yes ;;
+    esac
+    eval "enable_${ac_feature}='$ac_optarg'" ;;
+
+  -exec-prefix | --exec_prefix | --exec-prefix | --exec-prefi \
+  | --exec-pref | --exec-pre | --exec-pr | --exec-p | --exec- \
+  | --exec | --exe | --ex)
+    ac_prev=exec_prefix ;;
+  -exec-prefix=* | --exec_prefix=* | --exec-prefix=* | --exec-prefi=* \
+  | --exec-pref=* | --exec-pre=* | --exec-pr=* | --exec-p=* | --exec-=* \
+  | --exec=* | --exe=* | --ex=*)
+    exec_prefix="$ac_optarg" ;;
+
+  -gas | --gas | --ga | --g)
+    # Obsolete; use --with-gas.
+    with_gas=yes ;;
+
+  -help | --help | --hel | --he)
+    # Omit some internal or obsolete options to make the list less imposing.
+    # This message is too long to be a string in the A/UX 3.1 sh.
+    cat << EOF
+Usage: configure [options] [host]
+Options: [defaults in brackets after descriptions]
+Configuration:
+  --cache-file=FILE       cache test results in FILE
+  --help                  print this message
+  --no-create             do not create output files
+  --quiet, --silent       do not print \`checking...' messages
+  --version               print the version of autoconf that created configure
+Directory and file names:
+  --prefix=PREFIX         install architecture-independent files in PREFIX
+                          [$ac_default_prefix]
+  --exec-prefix=EPREFIX   install architecture-dependent files in EPREFIX
+                          [same as prefix]
+  --bindir=DIR            user executables in DIR [EPREFIX/bin]
+  --sbindir=DIR           system admin executables in DIR [EPREFIX/sbin]
+  --libexecdir=DIR        program executables in DIR [EPREFIX/libexec]
+  --datadir=DIR           read-only architecture-independent data in DIR
+                          [PREFIX/share]
+  --sysconfdir=DIR        read-only single-machine data in DIR [PREFIX/etc]
+  --sharedstatedir=DIR    modifiable architecture-independent data in DIR
+                          [PREFIX/com]
+  --localstatedir=DIR     modifiable single-machine data in DIR [PREFIX/var]
+  --libdir=DIR            object code libraries in DIR [EPREFIX/lib]
+  --includedir=DIR        C header files in DIR [PREFIX/include]
+  --oldincludedir=DIR     C header files for non-gcc in DIR [/usr/include]
+  --infodir=DIR           info documentation in DIR [PREFIX/info]
+  --mandir=DIR            man documentation in DIR [PREFIX/man]
+  --srcdir=DIR            find the sources in DIR [configure dir or ..]
+  --program-prefix=PREFIX prepend PREFIX to installed program names
+  --program-suffix=SUFFIX append SUFFIX to installed program names
+  --program-transform-name=PROGRAM
+                          run sed PROGRAM on installed program names
+EOF
+    cat << EOF
+Host type:
+  --build=BUILD           configure for building on BUILD [BUILD=HOST]
+  --host=HOST             configure for HOST [guessed]
+  --target=TARGET         configure for TARGET [TARGET=HOST]
+Features and packages:
+  --disable-FEATURE       do not include FEATURE (same as --enable-FEATURE=no)
+  --enable-FEATURE[=ARG]  include FEATURE [ARG=yes]
+  --with-PACKAGE[=ARG]    use PACKAGE [ARG=yes]
+  --without-PACKAGE       do not use PACKAGE (same as --with-PACKAGE=no)
+  --x-includes=DIR        X include files are in DIR
+  --x-libraries=DIR       X library files are in DIR
+EOF
+    if test -n "$ac_help"; then
+      echo "--enable and --with options recognized:$ac_help"
+    fi
+    exit 0 ;;
+
+  -host | --host | --hos | --ho)
+    ac_prev=host ;;
+  -host=* | --host=* | --hos=* | --ho=*)
+    host="$ac_optarg" ;;
+
+  -includedir | --includedir | --includedi | --included | --include \
+  | --includ | --inclu | --incl | --inc)
+    ac_prev=includedir ;;
+  -includedir=* | --includedir=* | --includedi=* | --included=* | --include=* \
+  | --includ=* | --inclu=* | --incl=* | --inc=*)
+    includedir="$ac_optarg" ;;
+
+  -infodir | --infodir | --infodi | --infod | --info | --inf)
+    ac_prev=infodir ;;
+  -infodir=* | --infodir=* | --infodi=* | --infod=* | --info=* | --inf=*)
+    infodir="$ac_optarg" ;;
+
+  -libdir | --libdir | --libdi | --libd)
+    ac_prev=libdir ;;
+  -libdir=* | --libdir=* | --libdi=* | --libd=*)
+    libdir="$ac_optarg" ;;
+
+  -libexecdir | --libexecdir | --libexecdi | --libexecd | --libexec \
+  | --libexe | --libex | --libe)
+    ac_prev=libexecdir ;;
+  -libexecdir=* | --libexecdir=* | --libexecdi=* | --libexecd=* | --libexec=* \
+  | --libexe=* | --libex=* | --libe=*)
+    libexecdir="$ac_optarg" ;;
+
+  -localstatedir | --localstatedir | --localstatedi | --localstated \
+  | --localstate | --localstat | --localsta | --localst \
+  | --locals | --local | --loca | --loc | --lo)
+    ac_prev=localstatedir ;;
+  -localstatedir=* | --localstatedir=* | --localstatedi=* | --localstated=* \
+  | --localstate=* | --localstat=* | --localsta=* | --localst=* \
+  | --locals=* | --local=* | --loca=* | --loc=* | --lo=*)
+    localstatedir="$ac_optarg" ;;
+
+  -mandir | --mandir | --mandi | --mand | --man | --ma | --m)
+    ac_prev=mandir ;;
+  -mandir=* | --mandir=* | --mandi=* | --mand=* | --man=* | --ma=* | --m=*)
+    mandir="$ac_optarg" ;;
+
+  -nfp | --nfp | --nf)
+    # Obsolete; use --without-fp.
+    with_fp=no ;;
+
+  -no-create | --no-create | --no-creat | --no-crea | --no-cre \
+  | --no-cr | --no-c)
+    no_create=yes ;;
+
+  -no-recursion | --no-recursion | --no-recursio | --no-recursi \
+  | --no-recurs | --no-recur | --no-recu | --no-rec | --no-re | --no-r)
+    no_recursion=yes ;;
+
+  -oldincludedir | --oldincludedir | --oldincludedi | --oldincluded \
+  | --oldinclude | --oldinclud | --oldinclu | --oldincl | --oldinc \
+  | --oldin | --oldi | --old | --ol | --o)
+    ac_prev=oldincludedir ;;
+  -oldincludedir=* | --oldincludedir=* | --oldincludedi=* | --oldincluded=* \
+  | --oldinclude=* | --oldinclud=* | --oldinclu=* | --oldincl=* | --oldinc=* \
+  | --oldin=* | --oldi=* | --old=* | --ol=* | --o=*)
+    oldincludedir="$ac_optarg" ;;
+
+  -prefix | --prefix | --prefi | --pref | --pre | --pr | --p)
+    ac_prev=prefix ;;
+  -prefix=* | --prefix=* | --prefi=* | --pref=* | --pre=* | --pr=* | --p=*)
+    prefix="$ac_optarg" ;;
+
+  -program-prefix | --program-prefix | --program-prefi | --program-pref \
+  | --program-pre | --program-pr | --program-p)
+    ac_prev=program_prefix ;;
+  -program-prefix=* | --program-prefix=* | --program-prefi=* \
+  | --program-pref=* | --program-pre=* | --program-pr=* | --program-p=*)
+    program_prefix="$ac_optarg" ;;
+
+  -program-suffix | --program-suffix | --program-suffi | --program-suff \
+  | --program-suf | --program-su | --program-s)
+    ac_prev=program_suffix ;;
+  -program-suffix=* | --program-suffix=* | --program-suffi=* \
+  | --program-suff=* | --program-suf=* | --program-su=* | --program-s=*)
+    program_suffix="$ac_optarg" ;;
+
+  -program-transform-name | --program-transform-name \
+  | --program-transform-nam | --program-transform-na \
+  | --program-transform-n | --program-transform- \
+  | --program-transform | --program-transfor \
+  | --program-transfo | --program-transf \
+  | --program-trans | --program-tran \
+  | --progr-tra | --program-tr | --program-t)
+    ac_prev=program_transform_name ;;
+  -program-transform-name=* | --program-transform-name=* \
+  | --program-transform-nam=* | --program-transform-na=* \
+  | --program-transform-n=* | --program-transform-=* \
+  | --program-transform=* | --program-transfor=* \
+  | --program-transfo=* | --program-transf=* \
+  | --program-trans=* | --program-tran=* \
+  | --progr-tra=* | --program-tr=* | --program-t=*)
+    program_transform_name="$ac_optarg" ;;
+
+  -q | -quiet | --quiet | --quie | --qui | --qu | --q \
+  | -silent | --silent | --silen | --sile | --sil)
+    silent=yes ;;
+
+  -sbindir | --sbindir | --sbindi | --sbind | --sbin | --sbi | --sb)
+    ac_prev=sbindir ;;
+  -sbindir=* | --sbindir=* | --sbindi=* | --sbind=* | --sbin=* \
+  | --sbi=* | --sb=*)
+    sbindir="$ac_optarg" ;;
+
+  -sharedstatedir | --sharedstatedir | --sharedstatedi \
+  | --sharedstated | --sharedstate | --sharedstat | --sharedsta \
+  | --sharedst | --shareds | --shared | --share | --shar \
+  | --sha | --sh)
+    ac_prev=sharedstatedir ;;
+  -sharedstatedir=* | --sharedstatedir=* | --sharedstatedi=* \
+  | --sharedstated=* | --sharedstate=* | --sharedstat=* | --sharedsta=* \
+  | --sharedst=* | --shareds=* | --shared=* | --share=* | --shar=* \
+  | --sha=* | --sh=*)
+    sharedstatedir="$ac_optarg" ;;
+
+  -site | --site | --sit)
+    ac_prev=site ;;
+  -site=* | --site=* | --sit=*)
+    site="$ac_optarg" ;;
+
+  -srcdir | --srcdir | --srcdi | --srcd | --src | --sr)
+    ac_prev=srcdir ;;
+  -srcdir=* | --srcdir=* | --srcdi=* | --srcd=* | --src=* | --sr=*)
+    srcdir="$ac_optarg" ;;
+
+  -sysconfdir | --sysconfdir | --sysconfdi | --sysconfd | --sysconf \
+  | --syscon | --sysco | --sysc | --sys | --sy)
+    ac_prev=sysconfdir ;;
+  -sysconfdir=* | --sysconfdir=* | --sysconfdi=* | --sysconfd=* | --sysconf=* \
+  | --syscon=* | --sysco=* | --sysc=* | --sys=* | --sy=*)
+    sysconfdir="$ac_optarg" ;;
+
+  -target | --target | --targe | --targ | --tar | --ta | --t)
+    ac_prev=target ;;
+  -target=* | --target=* | --targe=* | --targ=* | --tar=* | --ta=* | --t=*)
+    target="$ac_optarg" ;;
+
+  -v | -verbose | --verbose | --verbos | --verbo | --verb)
+    verbose=yes ;;
+
+  -version | --version | --versio | --versi | --vers)
+    echo "configure generated by autoconf version 2.12"
+    exit 0 ;;
+
+  -with-* | --with-*)
+    ac_package=`echo $ac_option|sed -e 's/-*with-//' -e 's/=.*//'`
+    # Reject names that are not valid shell variable names.
+    if test -n "`echo $ac_package| sed 's/[-_a-zA-Z0-9]//g'`"; then
+      { echo "configure: error: $ac_package: invalid package name" 1>&2; exit 1; }
+    fi
+    ac_package=`echo $ac_package| sed 's/-/_/g'`
+    case "$ac_option" in
+      *=*) ;;
+      *) ac_optarg=yes ;;
+    esac
+    eval "with_${ac_package}='$ac_optarg'" ;;
+
+  -without-* | --without-*)
+    ac_package=`echo $ac_option|sed -e 's/-*without-//'`
+    # Reject names that are not valid shell variable names.
+    if test -n "`echo $ac_package| sed 's/[-a-zA-Z0-9_]//g'`"; then
+      { echo "configure: error: $ac_package: invalid package name" 1>&2; exit 1; }
+    fi
+    ac_package=`echo $ac_package| sed 's/-/_/g'`
+    eval "with_${ac_package}=no" ;;
+
+  --x)
+    # Obsolete; use --with-x.
+    with_x=yes ;;
+
+  -x-includes | --x-includes | --x-include | --x-includ | --x-inclu \
+  | --x-incl | --x-inc | --x-in | --x-i)
+    ac_prev=x_includes ;;
+  -x-includes=* | --x-includes=* | --x-include=* | --x-includ=* | --x-inclu=* \
+  | --x-incl=* | --x-inc=* | --x-in=* | --x-i=*)
+    x_includes="$ac_optarg" ;;
+
+  -x-libraries | --x-libraries | --x-librarie | --x-librari \
+  | --x-librar | --x-libra | --x-libr | --x-lib | --x-li | --x-l)
+    ac_prev=x_libraries ;;
+  -x-libraries=* | --x-libraries=* | --x-librarie=* | --x-librari=* \
+  | --x-librar=* | --x-libra=* | --x-libr=* | --x-lib=* | --x-li=* | --x-l=*)
+    x_libraries="$ac_optarg" ;;
+
+  -*) { echo "configure: error: $ac_option: invalid option; use --help to show usage" 1>&2; exit 1; }
+    ;;
+
+  *=*)
+    varname=`echo "$ac_option"|sed -e 's/=.*//'`
+    # Reject names that aren't valid shell variable names.
+    if test -n "`echo $varname| sed 's/[a-zA-Z0-9_]//g'`"; then
+      { echo "configure: error: $varname: invalid shell variable name" 1>&2; exit 1; }
+    fi
+    val="`echo "$ac_option"|sed 's/[^=]*=//'`"
+    test -n "$verbose" && echo "	setting shell variable $varname to $val"
+    eval "$varname='$val'"
+    eval "export $varname" ;;
+
+  *)
+    if test -n "`echo $ac_option| sed 's/[-a-z0-9.]//g'`"; then
+      echo "configure: warning: $ac_option: invalid host type" 1>&2
+    fi
+    if test "x$nonopt" != xNONE; then
+      { echo "configure: error: can only configure for one host and one target at a time" 1>&2; exit 1; }
+    fi
+    nonopt="$ac_option"
+    ;;
+
+  esac
+done
+
+if test -n "$ac_prev"; then
+  { echo "configure: error: missing argument to --`echo $ac_prev | sed 's/_/-/g'`" 1>&2; exit 1; }
+fi
+
+trap 'rm -fr conftest* confdefs* core core.* *.core $ac_clean_files; exit 1' 1 2 15
+
+# File descriptor usage:
+# 0 standard input
+# 1 file creation
+# 2 errors and warnings
+# 3 some systems may open it to /dev/tty
+# 4 used on the Kubota Titan
+# 6 checking for... messages and results
+# 5 compiler messages saved in config.log
+if test "$silent" = yes; then
+  exec 6>/dev/null
+else
+  exec 6>&1
+fi
+exec 5>./config.log
+
+echo "\
+This file contains any messages produced by compilers while
+running configure, to aid debugging if configure makes a mistake.
+" 1>&5
+
+# Strip out --no-create and --no-recursion so they do not pile up.
+# Also quote any args containing shell metacharacters.
+ac_configure_args=
+for ac_arg
+do
+  case "$ac_arg" in
+  -no-create | --no-create | --no-creat | --no-crea | --no-cre \
+  | --no-cr | --no-c) ;;
+  -no-recursion | --no-recursion | --no-recursio | --no-recursi \
+  | --no-recurs | --no-recur | --no-recu | --no-rec | --no-re | --no-r) ;;
+  *" "*|*"	"*|*[\[\]\~\#\$\^\&\*\(\)\{\}\\\|\;\<\>\?]*)
+  ac_configure_args="$ac_configure_args '$ac_arg'" ;;
+  *) ac_configure_args="$ac_configure_args $ac_arg" ;;
+  esac
+done
+
+# NLS nuisances.
+# Only set these to C if already set.  These must not be set unconditionally
+# because not all systems understand e.g. LANG=C (notably SCO).
+# Fixing LC_MESSAGES prevents Solaris sh from translating var values in `set'!
+# Non-C LC_CTYPE values break the ctype check.
+if test "${LANG+set}"   = set; then LANG=C;   export LANG;   fi
+if test "${LC_ALL+set}" = set; then LC_ALL=C; export LC_ALL; fi
+if test "${LC_MESSAGES+set}" = set; then LC_MESSAGES=C; export LC_MESSAGES; fi
+if test "${LC_CTYPE+set}"    = set; then LC_CTYPE=C;    export LC_CTYPE;    fi
+
+# confdefs.h avoids OS command line length limits that DEFS can exceed.
+rm -rf conftest* confdefs.h
+# AIX cpp loses on an empty file, so make sure it contains at least a newline.
+echo > confdefs.h
+
+# A filename unique to this package, relative to the directory that
+# configure is in, which we can look for to find out if srcdir is correct.
+ac_unique_file=jcmaster.c
+
+# Find the source files, if location was not specified.
+if test -z "$srcdir"; then
+  ac_srcdir_defaulted=yes
+  # Try the directory containing this script, then its parent.
+  ac_prog=$0
+  ac_confdir=`echo $ac_prog|sed 's%/[^/][^/]*$%%'`
+  test "x$ac_confdir" = "x$ac_prog" && ac_confdir=.
+  srcdir=$ac_confdir
+  if test ! -r $srcdir/$ac_unique_file; then
+    srcdir=..
+  fi
+else
+  ac_srcdir_defaulted=no
+fi
+if test ! -r $srcdir/$ac_unique_file; then
+  if test "$ac_srcdir_defaulted" = yes; then
+    { echo "configure: error: can not find sources in $ac_confdir or .." 1>&2; exit 1; }
+  else
+    { echo "configure: error: can not find sources in $srcdir" 1>&2; exit 1; }
+  fi
+fi
+srcdir=`echo "${srcdir}" | sed 's%\([^/]\)/*$%\1%'`
+
+# Prefer explicitly selected file to automatically selected ones.
+if test -z "$CONFIG_SITE"; then
+  if test "x$prefix" != xNONE; then
+    CONFIG_SITE="$prefix/share/config.site $prefix/etc/config.site"
+  else
+    CONFIG_SITE="$ac_default_prefix/share/config.site $ac_default_prefix/etc/config.site"
+  fi
+fi
+for ac_site_file in $CONFIG_SITE; do
+  if test -r "$ac_site_file"; then
+    echo "loading site script $ac_site_file"
+    . "$ac_site_file"
+  fi
+done
+
+
+ac_ext=c
+# CFLAGS is not in ac_cpp because -g, -O, etc. are not valid cpp options.
+ac_cpp='$CPP $CPPFLAGS'
+ac_compile='${CC-cc} -c $CFLAGS $CPPFLAGS conftest.$ac_ext 1>&5'
+ac_link='${CC-cc} -o conftest $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS 1>&5'
+cross_compiling=$ac_cv_prog_cc_cross
+
+if (echo "testing\c"; echo 1,2,3) | grep c >/dev/null; then
+  # Stardent Vistra SVR4 grep lacks -e, says ghazi@caip.rutgers.edu.
+  if (echo -n testing; echo 1,2,3) | sed s/-n/xn/ | grep xn >/dev/null; then
+    ac_n= ac_c='
+' ac_t='	'
+  else
+    ac_n=-n ac_c= ac_t=
+  fi
+else
+  ac_n= ac_c='\c' ac_t=
+fi
+
+
+
+# Extract the first word of "gcc", so it can be a program name with args.
+set dummy gcc; ac_word=$2
+echo $ac_n "checking for $ac_word""... $ac_c" 1>&6
+echo "configure:538: checking for $ac_word" >&5
+if eval "test \"`echo '$''{'ac_cv_prog_CC'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+  if test -n "$CC"; then
+  ac_cv_prog_CC="$CC" # Let the user override the test.
+else
+  IFS="${IFS= 	}"; ac_save_ifs="$IFS"; IFS="${IFS}:"
+  for ac_dir in $PATH; do
+    test -z "$ac_dir" && ac_dir=.
+    if test -f $ac_dir/$ac_word; then
+      ac_cv_prog_CC="gcc"
+      break
+    fi
+  done
+  IFS="$ac_save_ifs"
+fi
+fi
+CC="$ac_cv_prog_CC"
+if test -n "$CC"; then
+  echo "$ac_t""$CC" 1>&6
+else
+  echo "$ac_t""no" 1>&6
+fi
+
+if test -z "$CC"; then
+  # Extract the first word of "cc", so it can be a program name with args.
+set dummy cc; ac_word=$2
+echo $ac_n "checking for $ac_word""... $ac_c" 1>&6
+echo "configure:567: checking for $ac_word" >&5
+if eval "test \"`echo '$''{'ac_cv_prog_CC'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+  if test -n "$CC"; then
+  ac_cv_prog_CC="$CC" # Let the user override the test.
+else
+  IFS="${IFS= 	}"; ac_save_ifs="$IFS"; IFS="${IFS}:"
+  ac_prog_rejected=no
+  for ac_dir in $PATH; do
+    test -z "$ac_dir" && ac_dir=.
+    if test -f $ac_dir/$ac_word; then
+      if test "$ac_dir/$ac_word" = "/usr/ucb/cc"; then
+        ac_prog_rejected=yes
+	continue
+      fi
+      ac_cv_prog_CC="cc"
+      break
+    fi
+  done
+  IFS="$ac_save_ifs"
+if test $ac_prog_rejected = yes; then
+  # We found a bogon in the path, so make sure we never use it.
+  set dummy $ac_cv_prog_CC
+  shift
+  if test $# -gt 0; then
+    # We chose a different compiler from the bogus one.
+    # However, it has the same basename, so the bogon will be chosen
+    # first if we set CC to just the basename; use the full file name.
+    shift
+    set dummy "$ac_dir/$ac_word" "$@"
+    shift
+    ac_cv_prog_CC="$@"
+  fi
+fi
+fi
+fi
+CC="$ac_cv_prog_CC"
+if test -n "$CC"; then
+  echo "$ac_t""$CC" 1>&6
+else
+  echo "$ac_t""no" 1>&6
+fi
+
+  test -z "$CC" && { echo "configure: error: no acceptable cc found in \$PATH" 1>&2; exit 1; }
+fi
+
+echo $ac_n "checking whether the C compiler ($CC $CFLAGS $LDFLAGS) works""... $ac_c" 1>&6
+echo "configure:615: checking whether the C compiler ($CC $CFLAGS $LDFLAGS) works" >&5
+
+ac_ext=c
+# CFLAGS is not in ac_cpp because -g, -O, etc. are not valid cpp options.
+ac_cpp='$CPP $CPPFLAGS'
+ac_compile='${CC-cc} -c $CFLAGS $CPPFLAGS conftest.$ac_ext 1>&5'
+ac_link='${CC-cc} -o conftest $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS 1>&5'
+cross_compiling=$ac_cv_prog_cc_cross
+
+cat > conftest.$ac_ext <<EOF
+#line 625 "configure"
+#include "confdefs.h"
+main(){return(0);}
+EOF
+if { (eval echo configure:629: \"$ac_link\") 1>&5; (eval $ac_link) 2>&5; } && test -s conftest; then
+  ac_cv_prog_cc_works=yes
+  # If we can't run a trivial program, we are probably using a cross compiler.
+  if (./conftest; exit) 2>/dev/null; then
+    ac_cv_prog_cc_cross=no
+  else
+    ac_cv_prog_cc_cross=yes
+  fi
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  ac_cv_prog_cc_works=no
+fi
+rm -fr conftest*
+
+echo "$ac_t""$ac_cv_prog_cc_works" 1>&6
+if test $ac_cv_prog_cc_works = no; then
+  { echo "configure: error: installation or configuration problem: C compiler cannot create executables." 1>&2; exit 1; }
+fi
+echo $ac_n "checking whether the C compiler ($CC $CFLAGS $LDFLAGS) is a cross-compiler""... $ac_c" 1>&6
+echo "configure:649: checking whether the C compiler ($CC $CFLAGS $LDFLAGS) is a cross-compiler" >&5
+echo "$ac_t""$ac_cv_prog_cc_cross" 1>&6
+cross_compiling=$ac_cv_prog_cc_cross
+
+echo $ac_n "checking whether we are using GNU C""... $ac_c" 1>&6
+echo "configure:654: checking whether we are using GNU C" >&5
+if eval "test \"`echo '$''{'ac_cv_prog_gcc'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+  cat > conftest.c <<EOF
+#ifdef __GNUC__
+  yes;
+#endif
+EOF
+if { ac_try='${CC-cc} -E conftest.c'; { (eval echo configure:663: \"$ac_try\") 1>&5; (eval $ac_try) 2>&5; }; } | egrep yes >/dev/null 2>&1; then
+  ac_cv_prog_gcc=yes
+else
+  ac_cv_prog_gcc=no
+fi
+fi
+
+echo "$ac_t""$ac_cv_prog_gcc" 1>&6
+
+if test $ac_cv_prog_gcc = yes; then
+  GCC=yes
+  test "${CFLAGS+set}" = set || CFLAGS="-O2"
+else
+  GCC=
+  test "${CFLAGS+set}" = set || CFLAGS="-O"
+fi
+
+echo $ac_n "checking how to run the C preprocessor""... $ac_c" 1>&6
+echo "configure:681: checking how to run the C preprocessor" >&5
+# On Suns, sometimes $CPP names a directory.
+if test -n "$CPP" && test -d "$CPP"; then
+  CPP=
+fi
+if test -z "$CPP"; then
+if eval "test \"`echo '$''{'ac_cv_prog_CPP'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+    # This must be in double quotes, not single quotes, because CPP may get
+  # substituted into the Makefile and "${CC-cc}" will confuse make.
+  CPP="${CC-cc} -E"
+  # On the NeXT, cc -E runs the code through the compiler's parser,
+  # not just through cpp.
+  cat > conftest.$ac_ext <<EOF
+#line 696 "configure"
+#include "confdefs.h"
+#include <assert.h>
+Syntax Error
+EOF
+ac_try="$ac_cpp conftest.$ac_ext >/dev/null 2>conftest.out"
+{ (eval echo configure:702: \"$ac_try\") 1>&5; (eval $ac_try) 2>&5; }
+ac_err=`grep -v '^ *+' conftest.out`
+if test -z "$ac_err"; then
+  :
+else
+  echo "$ac_err" >&5
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  CPP="${CC-cc} -E -traditional-cpp"
+  cat > conftest.$ac_ext <<EOF
+#line 713 "configure"
+#include "confdefs.h"
+#include <assert.h>
+Syntax Error
+EOF
+ac_try="$ac_cpp conftest.$ac_ext >/dev/null 2>conftest.out"
+{ (eval echo configure:719: \"$ac_try\") 1>&5; (eval $ac_try) 2>&5; }
+ac_err=`grep -v '^ *+' conftest.out`
+if test -z "$ac_err"; then
+  :
+else
+  echo "$ac_err" >&5
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  CPP=/lib/cpp
+fi
+rm -f conftest*
+fi
+rm -f conftest*
+  ac_cv_prog_CPP="$CPP"
+fi
+  CPP="$ac_cv_prog_CPP"
+else
+  ac_cv_prog_CPP="$CPP"
+fi
+echo "$ac_t""$CPP" 1>&6
+
+echo $ac_n "checking for function prototypes""... $ac_c" 1>&6
+echo "configure:742: checking for function prototypes" >&5
+if eval "test \"`echo '$''{'ijg_cv_have_prototypes'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+  cat > conftest.$ac_ext <<EOF
+#line 747 "configure"
+#include "confdefs.h"
+
+int testfunction (int arg1, int * arg2); /* check prototypes */
+struct methods_struct {		/* check method-pointer declarations */
+  int (*error_exit) (char *msgtext);
+  int (*trace_message) (char *msgtext);
+  int (*another_method) (void);
+};
+int testfunction (int arg1, int * arg2) /* check definitions */
+{ return arg2[arg1]; }
+int test2function (void)	/* check void arg list */
+{ return 0; }
+
+int main() {
+ 
+; return 0; }
+EOF
+if { (eval echo configure:765: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then
+  rm -rf conftest*
+  ijg_cv_have_prototypes=yes
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  ijg_cv_have_prototypes=no
+fi
+rm -f conftest*
+fi
+
+echo "$ac_t""$ijg_cv_have_prototypes" 1>&6
+if test $ijg_cv_have_prototypes = yes; then
+  cat >> confdefs.h <<\EOF
+#define HAVE_PROTOTYPES 
+EOF
+
+else
+  echo Your compiler does not seem to know about function prototypes.
+  echo Perhaps it needs a special switch to enable ANSI C mode.
+  echo If so, we recommend running configure like this:
+  echo "   ./configure  CC='cc -switch'"
+  echo where -switch is the proper switch.
+fi
+ac_safe=`echo "stddef.h" | sed 'y%./+-%__p_%'`
+echo $ac_n "checking for stddef.h""... $ac_c" 1>&6
+echo "configure:792: checking for stddef.h" >&5
+if eval "test \"`echo '$''{'ac_cv_header_$ac_safe'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+  cat > conftest.$ac_ext <<EOF
+#line 797 "configure"
+#include "confdefs.h"
+#include <stddef.h>
+EOF
+ac_try="$ac_cpp conftest.$ac_ext >/dev/null 2>conftest.out"
+{ (eval echo configure:802: \"$ac_try\") 1>&5; (eval $ac_try) 2>&5; }
+ac_err=`grep -v '^ *+' conftest.out`
+if test -z "$ac_err"; then
+  rm -rf conftest*
+  eval "ac_cv_header_$ac_safe=yes"
+else
+  echo "$ac_err" >&5
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  eval "ac_cv_header_$ac_safe=no"
+fi
+rm -f conftest*
+fi
+if eval "test \"`echo '$ac_cv_header_'$ac_safe`\" = yes"; then
+  echo "$ac_t""yes" 1>&6
+  cat >> confdefs.h <<\EOF
+#define HAVE_STDDEF_H 
+EOF
+
+else
+  echo "$ac_t""no" 1>&6
+fi
+
+ac_safe=`echo "stdlib.h" | sed 'y%./+-%__p_%'`
+echo $ac_n "checking for stdlib.h""... $ac_c" 1>&6
+echo "configure:828: checking for stdlib.h" >&5
+if eval "test \"`echo '$''{'ac_cv_header_$ac_safe'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+  cat > conftest.$ac_ext <<EOF
+#line 833 "configure"
+#include "confdefs.h"
+#include <stdlib.h>
+EOF
+ac_try="$ac_cpp conftest.$ac_ext >/dev/null 2>conftest.out"
+{ (eval echo configure:838: \"$ac_try\") 1>&5; (eval $ac_try) 2>&5; }
+ac_err=`grep -v '^ *+' conftest.out`
+if test -z "$ac_err"; then
+  rm -rf conftest*
+  eval "ac_cv_header_$ac_safe=yes"
+else
+  echo "$ac_err" >&5
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  eval "ac_cv_header_$ac_safe=no"
+fi
+rm -f conftest*
+fi
+if eval "test \"`echo '$ac_cv_header_'$ac_safe`\" = yes"; then
+  echo "$ac_t""yes" 1>&6
+  cat >> confdefs.h <<\EOF
+#define HAVE_STDLIB_H 
+EOF
+
+else
+  echo "$ac_t""no" 1>&6
+fi
+
+ac_safe=`echo "string.h" | sed 'y%./+-%__p_%'`
+echo $ac_n "checking for string.h""... $ac_c" 1>&6
+echo "configure:864: checking for string.h" >&5
+if eval "test \"`echo '$''{'ac_cv_header_$ac_safe'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+  cat > conftest.$ac_ext <<EOF
+#line 869 "configure"
+#include "confdefs.h"
+#include <string.h>
+EOF
+ac_try="$ac_cpp conftest.$ac_ext >/dev/null 2>conftest.out"
+{ (eval echo configure:874: \"$ac_try\") 1>&5; (eval $ac_try) 2>&5; }
+ac_err=`grep -v '^ *+' conftest.out`
+if test -z "$ac_err"; then
+  rm -rf conftest*
+  eval "ac_cv_header_$ac_safe=yes"
+else
+  echo "$ac_err" >&5
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  eval "ac_cv_header_$ac_safe=no"
+fi
+rm -f conftest*
+fi
+if eval "test \"`echo '$ac_cv_header_'$ac_safe`\" = yes"; then
+  echo "$ac_t""yes" 1>&6
+  :
+else
+  echo "$ac_t""no" 1>&6
+cat >> confdefs.h <<\EOF
+#define NEED_BSD_STRINGS 
+EOF
+
+fi
+
+echo $ac_n "checking for size_t""... $ac_c" 1>&6
+echo "configure:900: checking for size_t" >&5
+cat > conftest.$ac_ext <<EOF
+#line 902 "configure"
+#include "confdefs.h"
+
+#ifdef HAVE_STDDEF_H
+#include <stddef.h>
+#endif
+#ifdef HAVE_STDLIB_H
+#include <stdlib.h>
+#endif
+#include <stdio.h>
+#ifdef NEED_BSD_STRINGS
+#include <strings.h>
+#else
+#include <string.h>
+#endif
+typedef size_t my_size_t;
+
+int main() {
+ my_size_t foovar; 
+; return 0; }
+EOF
+if { (eval echo configure:923: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then
+  rm -rf conftest*
+  ijg_size_t_ok=yes
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  ijg_size_t_ok="not ANSI, perhaps it is in sys/types.h"
+fi
+rm -f conftest*
+echo "$ac_t""$ijg_size_t_ok" 1>&6
+if test "$ijg_size_t_ok" != yes; then
+ac_safe=`echo "sys/types.h" | sed 'y%./+-%__p_%'`
+echo $ac_n "checking for sys/types.h""... $ac_c" 1>&6
+echo "configure:937: checking for sys/types.h" >&5
+if eval "test \"`echo '$''{'ac_cv_header_$ac_safe'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+  cat > conftest.$ac_ext <<EOF
+#line 942 "configure"
+#include "confdefs.h"
+#include <sys/types.h>
+EOF
+ac_try="$ac_cpp conftest.$ac_ext >/dev/null 2>conftest.out"
+{ (eval echo configure:947: \"$ac_try\") 1>&5; (eval $ac_try) 2>&5; }
+ac_err=`grep -v '^ *+' conftest.out`
+if test -z "$ac_err"; then
+  rm -rf conftest*
+  eval "ac_cv_header_$ac_safe=yes"
+else
+  echo "$ac_err" >&5
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  eval "ac_cv_header_$ac_safe=no"
+fi
+rm -f conftest*
+fi
+if eval "test \"`echo '$ac_cv_header_'$ac_safe`\" = yes"; then
+  echo "$ac_t""yes" 1>&6
+  cat >> confdefs.h <<\EOF
+#define NEED_SYS_TYPES_H 
+EOF
+
+cat > conftest.$ac_ext <<EOF
+#line 968 "configure"
+#include "confdefs.h"
+#include <sys/types.h>
+EOF
+if (eval "$ac_cpp conftest.$ac_ext") 2>&5 |
+  egrep "size_t" >/dev/null 2>&1; then
+  rm -rf conftest*
+  ijg_size_t_ok="size_t is in sys/types.h"
+else
+  rm -rf conftest*
+  ijg_size_t_ok=no
+fi
+rm -f conftest*
+
+else
+  echo "$ac_t""no" 1>&6
+ijg_size_t_ok=no
+fi
+
+echo "$ac_t""$ijg_size_t_ok" 1>&6
+if test "$ijg_size_t_ok" = no; then
+  echo Type size_t is not defined in any of the usual places.
+  echo Try putting '"typedef unsigned int size_t;"' in jconfig.h.
+fi
+fi
+echo $ac_n "checking for type unsigned char""... $ac_c" 1>&6
+echo "configure:994: checking for type unsigned char" >&5
+cat > conftest.$ac_ext <<EOF
+#line 996 "configure"
+#include "confdefs.h"
+
+int main() {
+ unsigned char un_char; 
+; return 0; }
+EOF
+if { (eval echo configure:1003: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then
+  rm -rf conftest*
+  echo "$ac_t""yes" 1>&6
+cat >> confdefs.h <<\EOF
+#define HAVE_UNSIGNED_CHAR 
+EOF
+
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  echo "$ac_t""no" 1>&6
+fi
+rm -f conftest*
+echo $ac_n "checking for type unsigned short""... $ac_c" 1>&6
+echo "configure:1018: checking for type unsigned short" >&5
+cat > conftest.$ac_ext <<EOF
+#line 1020 "configure"
+#include "confdefs.h"
+
+int main() {
+ unsigned short un_short; 
+; return 0; }
+EOF
+if { (eval echo configure:1027: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then
+  rm -rf conftest*
+  echo "$ac_t""yes" 1>&6
+cat >> confdefs.h <<\EOF
+#define HAVE_UNSIGNED_SHORT 
+EOF
+
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  echo "$ac_t""no" 1>&6
+fi
+rm -f conftest*
+echo $ac_n "checking for type void""... $ac_c" 1>&6
+echo "configure:1042: checking for type void" >&5
+cat > conftest.$ac_ext <<EOF
+#line 1044 "configure"
+#include "confdefs.h"
+
+/* Caution: a C++ compiler will insist on valid prototypes */
+typedef void * void_ptr;	/* check void * */
+#ifdef HAVE_PROTOTYPES		/* check ptr to function returning void */
+typedef void (*void_func) (int a, int b);
+#else
+typedef void (*void_func) ();
+#endif
+
+#ifdef HAVE_PROTOTYPES		/* check void function result */
+void test3function (void_ptr arg1, void_func arg2)
+#else
+void test3function (arg1, arg2)
+     void_ptr arg1;
+     void_func arg2;
+#endif
+{
+  char * locptr = (char *) arg1; /* check casting to and from void * */
+  arg1 = (void *) locptr;
+  (*arg2) (1, 2);		/* check call of fcn returning void */
+}
+
+int main() {
+ 
+; return 0; }
+EOF
+if { (eval echo configure:1072: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then
+  rm -rf conftest*
+  echo "$ac_t""yes" 1>&6
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  echo "$ac_t""no" 1>&6
+cat >> confdefs.h <<\EOF
+#define void char
+EOF
+
+fi
+rm -f conftest*
+
+echo $ac_n "checking for working const""... $ac_c" 1>&6
+echo "configure:1088: checking for working const" >&5
+if eval "test \"`echo '$''{'ac_cv_c_const'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+  cat > conftest.$ac_ext <<EOF
+#line 1093 "configure"
+#include "confdefs.h"
+
+int main() {
+
+/* Ultrix mips cc rejects this.  */
+typedef int charset[2]; const charset x;
+/* SunOS 4.1.1 cc rejects this.  */
+char const *const *ccp;
+char **p;
+/* NEC SVR4.0.2 mips cc rejects this.  */
+struct point {int x, y;};
+static struct point const zero = {0,0};
+/* AIX XL C 1.02.0.0 rejects this.
+   It does not let you subtract one const X* pointer from another in an arm
+   of an if-expression whose if-part is not a constant expression */
+const char *g = "string";
+ccp = &g + (g ? g-g : 0);
+/* HPUX 7.0 cc rejects these. */
+++ccp;
+p = (char**) ccp;
+ccp = (char const *const *) p;
+{ /* SCO 3.2v4 cc rejects this.  */
+  char *t;
+  char const *s = 0 ? (char *) 0 : (char const *) 0;
+
+  *t++ = 0;
+}
+{ /* Someone thinks the Sun supposedly-ANSI compiler will reject this.  */
+  int x[] = {25, 17};
+  const int *foo = &x[0];
+  ++foo;
+}
+{ /* Sun SC1.0 ANSI compiler rejects this -- but not the above. */
+  typedef const int *iptr;
+  iptr p = 0;
+  ++p;
+}
+{ /* AIX XL C 1.02.0.0 rejects this saying
+     "k.c", line 2.27: 1506-025 (S) Operand must be a modifiable lvalue. */
+  struct s { int j; const int *ap[3]; };
+  struct s *b; b->j = 5;
+}
+{ /* ULTRIX-32 V3.1 (Rev 9) vcc rejects this */
+  const int foo = 10;
+}
+
+; return 0; }
+EOF
+if { (eval echo configure:1142: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then
+  rm -rf conftest*
+  ac_cv_c_const=yes
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  ac_cv_c_const=no
+fi
+rm -f conftest*
+fi
+
+echo "$ac_t""$ac_cv_c_const" 1>&6
+if test $ac_cv_c_const = no; then
+  cat >> confdefs.h <<\EOF
+#define const 
+EOF
+
+fi
+
+echo $ac_n "checking for inline""... $ac_c" 1>&6
+echo "configure:1163: checking for inline" >&5
+ijg_cv_inline=""
+cat > conftest.$ac_ext <<EOF
+#line 1166 "configure"
+#include "confdefs.h"
+
+int main() {
+} __inline__ int foo() { return 0; }
+int bar() { return foo();
+; return 0; }
+EOF
+if { (eval echo configure:1174: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then
+  rm -rf conftest*
+  ijg_cv_inline="__inline__"
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  cat > conftest.$ac_ext <<EOF
+#line 1182 "configure"
+#include "confdefs.h"
+
+int main() {
+} __inline int foo() { return 0; }
+int bar() { return foo();
+; return 0; }
+EOF
+if { (eval echo configure:1190: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then
+  rm -rf conftest*
+  ijg_cv_inline="__inline"
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  cat > conftest.$ac_ext <<EOF
+#line 1198 "configure"
+#include "confdefs.h"
+
+int main() {
+} inline int foo() { return 0; }
+int bar() { return foo();
+; return 0; }
+EOF
+if { (eval echo configure:1206: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then
+  rm -rf conftest*
+  ijg_cv_inline="inline"
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+fi
+rm -f conftest*
+fi
+rm -f conftest*
+fi
+rm -f conftest*
+echo "$ac_t""$ijg_cv_inline" 1>&6
+cat >> confdefs.h <<EOF
+#define INLINE $ijg_cv_inline
+EOF
+
+echo $ac_n "checking for broken incomplete types""... $ac_c" 1>&6
+echo "configure:1224: checking for broken incomplete types" >&5
+cat > conftest.$ac_ext <<EOF
+#line 1226 "configure"
+#include "confdefs.h"
+ typedef struct undefined_structure * undef_struct_ptr; 
+int main() {
+
+; return 0; }
+EOF
+if { (eval echo configure:1233: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then
+  rm -rf conftest*
+  echo "$ac_t""ok" 1>&6
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  echo "$ac_t""broken" 1>&6
+cat >> confdefs.h <<\EOF
+#define INCOMPLETE_TYPES_BROKEN 
+EOF
+
+fi
+rm -f conftest*
+echo $ac_n "checking for short external names""... $ac_c" 1>&6
+echo "configure:1248: checking for short external names" >&5
+cat > conftest.$ac_ext <<EOF
+#line 1250 "configure"
+#include "confdefs.h"
+
+int possibly_duplicate_function () { return 0; }
+int possibly_dupli_function () { return 1; }
+
+int main() {
+ 
+; return 0; }
+EOF
+if { (eval echo configure:1260: \"$ac_link\") 1>&5; (eval $ac_link) 2>&5; } && test -s conftest; then
+  rm -rf conftest*
+  echo "$ac_t""ok" 1>&6
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  echo "$ac_t""short" 1>&6
+cat >> confdefs.h <<\EOF
+#define NEED_SHORT_EXTERNAL_NAMES 
+EOF
+
+fi
+rm -f conftest*
+echo $ac_n "checking to see if char is signed""... $ac_c" 1>&6
+echo "configure:1275: checking to see if char is signed" >&5
+if test "$cross_compiling" = yes; then
+  echo Assuming that char is signed on target machine.
+echo If it is unsigned, this will be a little bit inefficient.
+
+else
+  cat > conftest.$ac_ext <<EOF
+#line 1282 "configure"
+#include "confdefs.h"
+
+#ifdef HAVE_PROTOTYPES
+int is_char_signed (int arg)
+#else
+int is_char_signed (arg)
+     int arg;
+#endif
+{
+  if (arg == 189) {		/* expected result for unsigned char */
+    return 0;			/* type char is unsigned */
+  }
+  else if (arg != -67) {	/* expected result for signed char */
+    printf("Hmm, it seems 'char' is not eight bits wide on your machine.\n");
+    printf("I fear the JPEG software will not work at all.\n\n");
+  }
+  return 1;			/* assume char is signed otherwise */
+}
+char signed_char_check = (char) (-67);
+main() {
+  exit(is_char_signed((int) signed_char_check));
+}
+EOF
+if { (eval echo configure:1306: \"$ac_link\") 1>&5; (eval $ac_link) 2>&5; } && test -s conftest && (./conftest; exit) 2>/dev/null
+then
+  echo "$ac_t""no" 1>&6
+cat >> confdefs.h <<\EOF
+#define CHAR_IS_UNSIGNED 
+EOF
+
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -fr conftest*
+  echo "$ac_t""yes" 1>&6
+fi
+rm -fr conftest*
+fi
+
+echo $ac_n "checking to see if right shift is signed""... $ac_c" 1>&6
+echo "configure:1323: checking to see if right shift is signed" >&5
+if test "$cross_compiling" = yes; then
+  echo "$ac_t""Assuming that right shift is signed on target machine." 1>&6
+else
+  cat > conftest.$ac_ext <<EOF
+#line 1328 "configure"
+#include "confdefs.h"
+
+#ifdef HAVE_PROTOTYPES
+int is_shifting_signed (long arg)
+#else
+int is_shifting_signed (arg)
+     long arg;
+#endif
+/* See whether right-shift on a long is signed or not. */
+{
+  long res = arg >> 4;
+
+  if (res == -0x7F7E80CL) {	/* expected result for signed shift */
+    return 1;			/* right shift is signed */
+  }
+  /* see if unsigned-shift hack will fix it. */
+  /* we can't just test exact value since it depends on width of long... */
+  res |= (~0L) << (32-4);
+  if (res == -0x7F7E80CL) {	/* expected result now? */
+    return 0;			/* right shift is unsigned */
+  }
+  printf("Right shift isn't acting as I expect it to.\n");
+  printf("I fear the JPEG software will not work at all.\n\n");
+  return 0;			/* try it with unsigned anyway */
+}
+main() {
+  exit(is_shifting_signed(-0x7F7E80B1L));
+}
+EOF
+if { (eval echo configure:1358: \"$ac_link\") 1>&5; (eval $ac_link) 2>&5; } && test -s conftest && (./conftest; exit) 2>/dev/null
+then
+  echo "$ac_t""no" 1>&6
+cat >> confdefs.h <<\EOF
+#define RIGHT_SHIFT_IS_UNSIGNED 
+EOF
+
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -fr conftest*
+  echo "$ac_t""yes" 1>&6
+fi
+rm -fr conftest*
+fi
+
+echo $ac_n "checking to see if fopen accepts b spec""... $ac_c" 1>&6
+echo "configure:1375: checking to see if fopen accepts b spec" >&5
+if test "$cross_compiling" = yes; then
+  echo "$ac_t""Assuming that it does." 1>&6
+else
+  cat > conftest.$ac_ext <<EOF
+#line 1380 "configure"
+#include "confdefs.h"
+
+#include <stdio.h>
+main() {
+  if (fopen("conftestdata", "wb") != NULL)
+    exit(0);
+  exit(1);
+}
+EOF
+if { (eval echo configure:1390: \"$ac_link\") 1>&5; (eval $ac_link) 2>&5; } && test -s conftest && (./conftest; exit) 2>/dev/null
+then
+  echo "$ac_t""yes" 1>&6
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -fr conftest*
+  echo "$ac_t""no" 1>&6
+cat >> confdefs.h <<\EOF
+#define DONT_USE_B_MODE 
+EOF
+
+fi
+rm -fr conftest*
+fi
+
+ac_aux_dir=
+for ac_dir in $srcdir $srcdir/.. $srcdir/../..; do
+  if test -f $ac_dir/install-sh; then
+    ac_aux_dir=$ac_dir
+    ac_install_sh="$ac_aux_dir/install-sh -c"
+    break
+  elif test -f $ac_dir/install.sh; then
+    ac_aux_dir=$ac_dir
+    ac_install_sh="$ac_aux_dir/install.sh -c"
+    break
+  fi
+done
+if test -z "$ac_aux_dir"; then
+  { echo "configure: error: can not find install-sh or install.sh in $srcdir $srcdir/.. $srcdir/../.." 1>&2; exit 1; }
+fi
+ac_config_guess=$ac_aux_dir/config.guess
+ac_config_sub=$ac_aux_dir/config.sub
+ac_configure=$ac_aux_dir/configure # This should be Cygnus configure.
+
+# Find a good install program.  We prefer a C program (faster),
+# so one script is as good as another.  But avoid the broken or
+# incompatible versions:
+# SysV /etc/install, /usr/sbin/install
+# SunOS /usr/etc/install
+# IRIX /sbin/install
+# AIX /bin/install
+# AFS /usr/afsws/bin/install, which mishandles nonexistent args
+# SVR4 /usr/ucb/install, which tries to use the nonexistent group "staff"
+# ./install, which can be erroneously created by make from ./install.sh.
+echo $ac_n "checking for a BSD compatible install""... $ac_c" 1>&6
+echo "configure:1436: checking for a BSD compatible install" >&5
+if test -z "$INSTALL"; then
+if eval "test \"`echo '$''{'ac_cv_path_install'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+    IFS="${IFS= 	}"; ac_save_IFS="$IFS"; IFS="${IFS}:"
+  for ac_dir in $PATH; do
+    # Account for people who put trailing slashes in PATH elements.
+    case "$ac_dir/" in
+    /|./|.//|/etc/*|/usr/sbin/*|/usr/etc/*|/sbin/*|/usr/afsws/bin/*|/usr/ucb/*) ;;
+    *)
+      # OSF1 and SCO ODT 3.0 have their own names for install.
+      for ac_prog in ginstall installbsd scoinst install; do
+        if test -f $ac_dir/$ac_prog; then
+	  if test $ac_prog = install &&
+            grep dspmsg $ac_dir/$ac_prog >/dev/null 2>&1; then
+	    # AIX install.  It has an incompatible calling convention.
+	    # OSF/1 installbsd also uses dspmsg, but is usable.
+	    :
+	  else
+	    ac_cv_path_install="$ac_dir/$ac_prog -c"
+	    break 2
+	  fi
+	fi
+      done
+      ;;
+    esac
+  done
+  IFS="$ac_save_IFS"
+
+fi
+  if test "${ac_cv_path_install+set}" = set; then
+    INSTALL="$ac_cv_path_install"
+  else
+    # As a last resort, use the slow shell script.  We don't cache a
+    # path for INSTALL within a source directory, because that will
+    # break other packages using the cache if that directory is
+    # removed, or if the path is relative.
+    INSTALL="$ac_install_sh"
+  fi
+fi
+echo "$ac_t""$INSTALL" 1>&6
+
+# Use test -z because SunOS4 sh mishandles braces in ${var-val}.
+# It thinks the first close brace ends the variable substitution.
+test -z "$INSTALL_PROGRAM" && INSTALL_PROGRAM='${INSTALL}'
+
+test -z "$INSTALL_DATA" && INSTALL_DATA='${INSTALL} -m 644'
+
+# Extract the first word of "ranlib", so it can be a program name with args.
+set dummy ranlib; ac_word=$2
+echo $ac_n "checking for $ac_word""... $ac_c" 1>&6
+echo "configure:1488: checking for $ac_word" >&5
+if eval "test \"`echo '$''{'ac_cv_prog_RANLIB'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+  if test -n "$RANLIB"; then
+  ac_cv_prog_RANLIB="$RANLIB" # Let the user override the test.
+else
+  IFS="${IFS= 	}"; ac_save_ifs="$IFS"; IFS="${IFS}:"
+  for ac_dir in $PATH; do
+    test -z "$ac_dir" && ac_dir=.
+    if test -f $ac_dir/$ac_word; then
+      ac_cv_prog_RANLIB="ranlib"
+      break
+    fi
+  done
+  IFS="$ac_save_ifs"
+  test -z "$ac_cv_prog_RANLIB" && ac_cv_prog_RANLIB=":"
+fi
+fi
+RANLIB="$ac_cv_prog_RANLIB"
+if test -n "$RANLIB"; then
+  echo "$ac_t""$RANLIB" 1>&6
+else
+  echo "$ac_t""no" 1>&6
+fi
+
+
+# Decide whether to use libtool,
+# and if so whether to build shared, static, or both flavors of library.
+LTSHARED="no"
+# Check whether --enable-shared or --disable-shared was given.
+if test "${enable_shared+set}" = set; then
+  enableval="$enable_shared"
+  LTSHARED="$enableval"
+fi
+
+LTSTATIC="no"
+# Check whether --enable-static or --disable-static was given.
+if test "${enable_static+set}" = set; then
+  enableval="$enable_static"
+  LTSTATIC="$enableval"
+fi
+
+if test "x$LTSHARED" != xno  -o  "x$LTSTATIC" != xno; then
+  USELIBTOOL="yes"
+  LIBTOOL="./libtool"
+  O="lo"
+  A="la"
+  LN='$(LIBTOOL) --mode=link $(CC)'
+  INSTALL_LIB='$(LIBTOOL) --mode=install ${INSTALL}'
+  INSTALL_PROGRAM="\$(LIBTOOL) --mode=install $INSTALL_PROGRAM"
+else
+  USELIBTOOL="no"
+  LIBTOOL=""
+  O="o"
+  A="a"
+  LN='$(CC)'
+  INSTALL_LIB="$INSTALL_DATA"
+fi
+
+
+
+
+
+
+# Configure libtool if needed.
+if test $USELIBTOOL = yes; then
+  disable_shared=
+  disable_static=
+  if test "x$LTSHARED" = xno; then
+    disable_shared="--disable-shared"
+  fi
+  if test "x$LTSTATIC" = xno; then
+    disable_static="--disable-static"
+  fi
+  $srcdir/ltconfig $disable_shared $disable_static $srcdir/ltmain.sh
+fi
+
+# Select memory manager depending on user input.
+# If no "-enable-maxmem", use jmemnobs
+MEMORYMGR='jmemnobs.$(O)'
+MAXMEM="no"
+# Check whether --enable-maxmem or --disable-maxmem was given.
+if test "${enable_maxmem+set}" = set; then
+  enableval="$enable_maxmem"
+  MAXMEM="$enableval"
+fi
+
+# support --with-maxmem for backwards compatibility with IJG V5.
+# Check whether --with-maxmem or --without-maxmem was given.
+if test "${with_maxmem+set}" = set; then
+  withval="$with_maxmem"
+  MAXMEM="$withval"
+fi
+
+if test "x$MAXMEM" = xyes; then
+  MAXMEM=1
+fi
+if test "x$MAXMEM" != xno; then
+  if test -n "`echo $MAXMEM | sed 's/[0-9]//g'`"; then
+    { echo "configure: error: non-numeric argument to --enable-maxmem" 1>&2; exit 1; }
+  fi
+  DEFAULTMAXMEM=`expr $MAXMEM \* 1048576`
+cat >> confdefs.h <<EOF
+#define DEFAULT_MAX_MEM ${DEFAULTMAXMEM}
+EOF
+
+echo $ac_n "checking for 'tmpfile()'""... $ac_c" 1>&6
+echo "configure:1596: checking for 'tmpfile()'" >&5
+cat > conftest.$ac_ext <<EOF
+#line 1598 "configure"
+#include "confdefs.h"
+#include <stdio.h>
+int main() {
+ FILE * tfile = tmpfile(); 
+; return 0; }
+EOF
+if { (eval echo configure:1605: \"$ac_link\") 1>&5; (eval $ac_link) 2>&5; } && test -s conftest; then
+  rm -rf conftest*
+  echo "$ac_t""yes" 1>&6
+MEMORYMGR='jmemansi.$(O)'
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  echo "$ac_t""no" 1>&6
+MEMORYMGR='jmemname.$(O)'
+cat >> confdefs.h <<\EOF
+#define NEED_SIGNAL_CATCHER 
+EOF
+
+echo $ac_n "checking for 'mktemp()'""... $ac_c" 1>&6
+echo "configure:1620: checking for 'mktemp()'" >&5
+cat > conftest.$ac_ext <<EOF
+#line 1622 "configure"
+#include "confdefs.h"
+
+int main() {
+ char fname[80]; mktemp(fname); 
+; return 0; }
+EOF
+if { (eval echo configure:1629: \"$ac_link\") 1>&5; (eval $ac_link) 2>&5; } && test -s conftest; then
+  rm -rf conftest*
+  echo "$ac_t""yes" 1>&6
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  echo "$ac_t""no" 1>&6
+cat >> confdefs.h <<\EOF
+#define NO_MKTEMP 
+EOF
+
+fi
+rm -f conftest*
+fi
+rm -f conftest*
+fi
+
+
+# Extract the library version ID from jpeglib.h.
+echo $ac_n "checking libjpeg version number""... $ac_c" 1>&6
+echo "configure:1650: checking libjpeg version number" >&5
+JPEG_LIB_VERSION=`sed -e '/^#define JPEG_LIB_VERSION/!d' -e 's/^[^0-9]*\([0-9][0-9]*\).*$/\1/' $srcdir/jpeglib.h`
+echo "$ac_t""$JPEG_LIB_VERSION" 1>&6
+
+
+# Prepare to massage makefile.cfg correctly.
+if test $ijg_cv_have_prototypes = yes; then
+  A2K_DEPS=""
+  COM_A2K="# "
+else
+  A2K_DEPS="ansi2knr"
+  COM_A2K=""
+fi
+
+
+# ansi2knr needs -DBSD if string.h is missing
+if test $ac_cv_header_string_h = no; then
+  ANSI2KNRFLAGS="-DBSD"
+else
+  ANSI2KNRFLAGS=""
+fi
+
+# Substitutions to enable or disable libtool-related stuff
+if test $USELIBTOOL = yes -a $ijg_cv_have_prototypes = yes; then
+  COM_LT=""
+else
+  COM_LT="# "
+fi
+
+if test "x$LTSHARED" != xno; then
+  FORCE_INSTALL_LIB="install-lib"
+else
+  FORCE_INSTALL_LIB=""
+fi
+
+# Set up -I directives
+if test "x$srcdir" = x.; then
+  INCLUDEFLAGS='-I$(srcdir)'
+else
+  INCLUDEFLAGS='-I. -I$(srcdir)'
+fi
+
+trap '' 1 2 15
+
+trap 'rm -fr conftest* confdefs* core core.* *.core $ac_clean_files; exit 1' 1 2 15
+
+test "x$prefix" = xNONE && prefix=$ac_default_prefix
+# Let make expand exec_prefix.
+test "x$exec_prefix" = xNONE && exec_prefix='${prefix}'
+
+# Any assignment to VPATH causes Sun make to only execute
+# the first set of double-colon rules, so remove it if not needed.
+# If there is a colon in the path, we need to keep it.
+if test "x$srcdir" = x.; then
+  ac_vpsub='/^[ 	]*VPATH[ 	]*=[^:]*$/d'
+fi
+
+trap 'rm -f $CONFIG_STATUS conftest*; exit 1' 1 2 15
+
+DEFS=-DHAVE_CONFIG_H
+
+# Without the "./", some shells look in PATH for config.status.
+: ${CONFIG_STATUS=./config.status}
+
+echo creating $CONFIG_STATUS
+rm -f $CONFIG_STATUS
+cat > $CONFIG_STATUS <<EOF
+#! /bin/sh
+# Generated automatically by configure.
+# Run this file to recreate the current configuration.
+# This directory was configured as follows,
+# on host `(hostname || uname -n) 2>/dev/null | sed 1q`:
+#
+# $0 $ac_configure_args
+#
+# Compiler output produced by configure, useful for debugging
+# configure, is in ./config.log if it exists.
+
+ac_cs_usage="Usage: $CONFIG_STATUS [--recheck] [--version] [--help]"
+for ac_option
+do
+  case "\$ac_option" in
+  -recheck | --recheck | --rechec | --reche | --rech | --rec | --re | --r)
+    echo "running \${CONFIG_SHELL-/bin/sh} $0 $ac_configure_args --no-create --no-recursion"
+    exec \${CONFIG_SHELL-/bin/sh} $0 $ac_configure_args --no-create --no-recursion ;;
+  -version | --version | --versio | --versi | --vers | --ver | --ve | --v)
+    echo "$CONFIG_STATUS generated by autoconf version 2.12"
+    exit 0 ;;
+  -help | --help | --hel | --he | --h)
+    echo "\$ac_cs_usage"; exit 0 ;;
+  *) echo "\$ac_cs_usage"; exit 1 ;;
+  esac
+done
+
+ac_given_srcdir=$srcdir
+ac_given_INSTALL="$INSTALL"
+
+trap 'rm -fr `echo "Makefile:makefile.cfg jconfig.h:jconfig.cfg" | sed "s/:[^ ]*//g"` conftest*; exit 1' 1 2 15
+EOF
+cat >> $CONFIG_STATUS <<EOF
+
+# Protect against being on the right side of a sed subst in config.status.
+sed 's/%@/@@/; s/@%/@@/; s/%g\$/@g/; /@g\$/s/[\\\\&%]/\\\\&/g;
+ s/@@/%@/; s/@@/@%/; s/@g\$/%g/' > conftest.subs <<\\CEOF
+$ac_vpsub
+$extrasub
+s%@CFLAGS@%$CFLAGS%g
+s%@CPPFLAGS@%$CPPFLAGS%g
+s%@CXXFLAGS@%$CXXFLAGS%g
+s%@DEFS@%$DEFS%g
+s%@LDFLAGS@%$LDFLAGS%g
+s%@LIBS@%$LIBS%g
+s%@exec_prefix@%$exec_prefix%g
+s%@prefix@%$prefix%g
+s%@program_transform_name@%$program_transform_name%g
+s%@bindir@%$bindir%g
+s%@sbindir@%$sbindir%g
+s%@libexecdir@%$libexecdir%g
+s%@datadir@%$datadir%g
+s%@sysconfdir@%$sysconfdir%g
+s%@sharedstatedir@%$sharedstatedir%g
+s%@localstatedir@%$localstatedir%g
+s%@libdir@%$libdir%g
+s%@includedir@%$includedir%g
+s%@oldincludedir@%$oldincludedir%g
+s%@infodir@%$infodir%g
+s%@mandir@%$mandir%g
+s%@CC@%$CC%g
+s%@CPP@%$CPP%g
+s%@INSTALL_PROGRAM@%$INSTALL_PROGRAM%g
+s%@INSTALL_DATA@%$INSTALL_DATA%g
+s%@RANLIB@%$RANLIB%g
+s%@LIBTOOL@%$LIBTOOL%g
+s%@O@%$O%g
+s%@A@%$A%g
+s%@LN@%$LN%g
+s%@INSTALL_LIB@%$INSTALL_LIB%g
+s%@MEMORYMGR@%$MEMORYMGR%g
+s%@JPEG_LIB_VERSION@%$JPEG_LIB_VERSION%g
+s%@A2K_DEPS@%$A2K_DEPS%g
+s%@COM_A2K@%$COM_A2K%g
+s%@ANSI2KNRFLAGS@%$ANSI2KNRFLAGS%g
+s%@COM_LT@%$COM_LT%g
+s%@FORCE_INSTALL_LIB@%$FORCE_INSTALL_LIB%g
+s%@INCLUDEFLAGS@%$INCLUDEFLAGS%g
+
+CEOF
+EOF
+
+cat >> $CONFIG_STATUS <<\EOF
+
+# Split the substitutions into bite-sized pieces for seds with
+# small command number limits, like on Digital OSF/1 and HP-UX.
+ac_max_sed_cmds=90 # Maximum number of lines to put in a sed script.
+ac_file=1 # Number of current file.
+ac_beg=1 # First line for current file.
+ac_end=$ac_max_sed_cmds # Line after last line for current file.
+ac_more_lines=:
+ac_sed_cmds=""
+while $ac_more_lines; do
+  if test $ac_beg -gt 1; then
+    sed "1,${ac_beg}d; ${ac_end}q" conftest.subs > conftest.s$ac_file
+  else
+    sed "${ac_end}q" conftest.subs > conftest.s$ac_file
+  fi
+  if test ! -s conftest.s$ac_file; then
+    ac_more_lines=false
+    rm -f conftest.s$ac_file
+  else
+    if test -z "$ac_sed_cmds"; then
+      ac_sed_cmds="sed -f conftest.s$ac_file"
+    else
+      ac_sed_cmds="$ac_sed_cmds | sed -f conftest.s$ac_file"
+    fi
+    ac_file=`expr $ac_file + 1`
+    ac_beg=$ac_end
+    ac_end=`expr $ac_end + $ac_max_sed_cmds`
+  fi
+done
+if test -z "$ac_sed_cmds"; then
+  ac_sed_cmds=cat
+fi
+EOF
+
+cat >> $CONFIG_STATUS <<EOF
+
+CONFIG_FILES=\${CONFIG_FILES-"Makefile:makefile.cfg"}
+EOF
+cat >> $CONFIG_STATUS <<\EOF
+for ac_file in .. $CONFIG_FILES; do if test "x$ac_file" != x..; then
+  # Support "outfile[:infile[:infile...]]", defaulting infile="outfile.in".
+  case "$ac_file" in
+  *:*) ac_file_in=`echo "$ac_file"|sed 's%[^:]*:%%'`
+       ac_file=`echo "$ac_file"|sed 's%:.*%%'` ;;
+  *) ac_file_in="${ac_file}.in" ;;
+  esac
+
+  # Adjust a relative srcdir, top_srcdir, and INSTALL for subdirectories.
+
+  # Remove last slash and all that follows it.  Not all systems have dirname.
+  ac_dir=`echo $ac_file|sed 's%/[^/][^/]*$%%'`
+  if test "$ac_dir" != "$ac_file" && test "$ac_dir" != .; then
+    # The file is in a subdirectory.
+    test ! -d "$ac_dir" && mkdir "$ac_dir"
+    ac_dir_suffix="/`echo $ac_dir|sed 's%^\./%%'`"
+    # A "../" for each directory in $ac_dir_suffix.
+    ac_dots=`echo $ac_dir_suffix|sed 's%/[^/]*%../%g'`
+  else
+    ac_dir_suffix= ac_dots=
+  fi
+
+  case "$ac_given_srcdir" in
+  .)  srcdir=.
+      if test -z "$ac_dots"; then top_srcdir=.
+      else top_srcdir=`echo $ac_dots|sed 's%/$%%'`; fi ;;
+  /*) srcdir="$ac_given_srcdir$ac_dir_suffix"; top_srcdir="$ac_given_srcdir" ;;
+  *) # Relative path.
+    srcdir="$ac_dots$ac_given_srcdir$ac_dir_suffix"
+    top_srcdir="$ac_dots$ac_given_srcdir" ;;
+  esac
+
+  case "$ac_given_INSTALL" in
+  [/$]*) INSTALL="$ac_given_INSTALL" ;;
+  *) INSTALL="$ac_dots$ac_given_INSTALL" ;;
+  esac
+
+  echo creating "$ac_file"
+  rm -f "$ac_file"
+  configure_input="Generated automatically from `echo $ac_file_in|sed 's%.*/%%'` by configure."
+  case "$ac_file" in
+  *Makefile*) ac_comsub="1i\\
+# $configure_input" ;;
+  *) ac_comsub= ;;
+  esac
+
+  ac_file_inputs=`echo $ac_file_in|sed -e "s%^%$ac_given_srcdir/%" -e "s%:% $ac_given_srcdir/%g"`
+  sed -e "$ac_comsub
+s%@configure_input@%$configure_input%g
+s%@srcdir@%$srcdir%g
+s%@top_srcdir@%$top_srcdir%g
+s%@INSTALL@%$INSTALL%g
+" $ac_file_inputs | (eval "$ac_sed_cmds") > $ac_file
+fi; done
+rm -f conftest.s*
+
+# These sed commands are passed to sed as "A NAME B NAME C VALUE D", where
+# NAME is the cpp macro being defined and VALUE is the value it is being given.
+#
+# ac_d sets the value in "#define NAME VALUE" lines.
+ac_dA='s%^\([ 	]*\)#\([ 	]*define[ 	][ 	]*\)'
+ac_dB='\([ 	][ 	]*\)[^ 	]*%\1#\2'
+ac_dC='\3'
+ac_dD='%g'
+# ac_u turns "#undef NAME" with trailing blanks into "#define NAME VALUE".
+ac_uA='s%^\([ 	]*\)#\([ 	]*\)undef\([ 	][ 	]*\)'
+ac_uB='\([ 	]\)%\1#\2define\3'
+ac_uC=' '
+ac_uD='\4%g'
+# ac_e turns "#undef NAME" without trailing blanks into "#define NAME VALUE".
+ac_eA='s%^\([ 	]*\)#\([ 	]*\)undef\([ 	][ 	]*\)'
+ac_eB='$%\1#\2define\3'
+ac_eC=' '
+ac_eD='%g'
+
+if test "${CONFIG_HEADERS+set}" != set; then
+EOF
+cat >> $CONFIG_STATUS <<EOF
+  CONFIG_HEADERS="jconfig.h:jconfig.cfg"
+EOF
+cat >> $CONFIG_STATUS <<\EOF
+fi
+for ac_file in .. $CONFIG_HEADERS; do if test "x$ac_file" != x..; then
+  # Support "outfile[:infile[:infile...]]", defaulting infile="outfile.in".
+  case "$ac_file" in
+  *:*) ac_file_in=`echo "$ac_file"|sed 's%[^:]*:%%'`
+       ac_file=`echo "$ac_file"|sed 's%:.*%%'` ;;
+  *) ac_file_in="${ac_file}.in" ;;
+  esac
+
+  echo creating $ac_file
+
+  rm -f conftest.frag conftest.in conftest.out
+  ac_file_inputs=`echo $ac_file_in|sed -e "s%^%$ac_given_srcdir/%" -e "s%:% $ac_given_srcdir/%g"`
+  cat $ac_file_inputs > conftest.in
+
+EOF
+
+# Transform confdefs.h into a sed script conftest.vals that substitutes
+# the proper values into config.h.in to produce config.h.  And first:
+# Protect against being on the right side of a sed subst in config.status.
+# Protect against being in an unquoted here document in config.status.
+rm -f conftest.vals
+cat > conftest.hdr <<\EOF
+s/[\\&%]/\\&/g
+s%[\\$`]%\\&%g
+s%#define \([A-Za-z_][A-Za-z0-9_]*\) *\(.*\)%${ac_dA}\1${ac_dB}\1${ac_dC}\2${ac_dD}%gp
+s%ac_d%ac_u%gp
+s%ac_u%ac_e%gp
+EOF
+sed -n -f conftest.hdr confdefs.h > conftest.vals
+rm -f conftest.hdr
+
+# This sed command replaces #undef with comments.  This is necessary, for
+# example, in the case of _POSIX_SOURCE, which is predefined and required
+# on some systems where configure will not decide to define it.
+cat >> conftest.vals <<\EOF
+EOF
+
+# Break up conftest.vals because some shells have a limit on
+# the size of here documents, and old seds have small limits too.
+
+rm -f conftest.tail
+while :
+do
+  ac_lines=`grep -c . conftest.vals`
+  # grep -c gives empty output for an empty file on some AIX systems.
+  if test -z "$ac_lines" || test "$ac_lines" -eq 0; then break; fi
+  # Write a limited-size here document to conftest.frag.
+  echo '  cat > conftest.frag <<CEOF' >> $CONFIG_STATUS
+  sed ${ac_max_here_lines}q conftest.vals >> $CONFIG_STATUS
+  echo 'CEOF
+  sed -f conftest.frag conftest.in > conftest.out
+  rm -f conftest.in
+  mv conftest.out conftest.in
+' >> $CONFIG_STATUS
+  sed 1,${ac_max_here_lines}d conftest.vals > conftest.tail
+  rm -f conftest.vals
+  mv conftest.tail conftest.vals
+done
+rm -f conftest.vals
+
+cat >> $CONFIG_STATUS <<\EOF
+  rm -f conftest.frag conftest.h
+  echo "/* $ac_file.  Generated automatically by configure.  */" > conftest.h
+  cat conftest.in >> conftest.h
+  rm -f conftest.in
+  if cmp -s $ac_file conftest.h 2>/dev/null; then
+    echo "$ac_file is unchanged"
+    rm -f conftest.h
+  else
+    # Remove last slash and all that follows it.  Not all systems have dirname.
+      ac_dir=`echo $ac_file|sed 's%/[^/][^/]*$%%'`
+      if test "$ac_dir" != "$ac_file" && test "$ac_dir" != .; then
+      # The file is in a subdirectory.
+      test ! -d "$ac_dir" && mkdir "$ac_dir"
+    fi
+    rm -f $ac_file
+    mv conftest.h $ac_file
+  fi
+fi; done
+
+EOF
+cat >> $CONFIG_STATUS <<EOF
+
+EOF
+cat >> $CONFIG_STATUS <<\EOF
+
+exit 0
+EOF
+chmod +x $CONFIG_STATUS
+rm -fr confdefs* $ac_clean_files
+test "$no_create" = yes || ${CONFIG_SHELL-/bin/sh} $CONFIG_STATUS || exit 1
+
diff --git a/libjpegtwrp/djpeg.1 b/libjpegtwrp/djpeg.1
new file mode 100644
index 000000000..11beb6a51
--- /dev/null
+++ b/libjpegtwrp/djpeg.1
@@ -0,0 +1,253 @@
+.TH DJPEG 1 "22 August 1997"
+.SH NAME
+djpeg \- decompress a JPEG file to an image file
+.SH SYNOPSIS
+.B djpeg
+[
+.I options
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B djpeg
+decompresses the named JPEG file, or the standard input if no file is named,
+and produces an image file on the standard output.  PBMPLUS (PPM/PGM), BMP,
+GIF, Targa, or RLE (Utah Raster Toolkit) output format can be selected.
+(RLE is supported only if the URT library is available.)
+.SH OPTIONS
+All switch names may be abbreviated; for example,
+.B \-grayscale
+may be written
+.B \-gray
+or
+.BR \-gr .
+Most of the "basic" switches can be abbreviated to as little as one letter.
+Upper and lower case are equivalent (thus
+.B \-BMP
+is the same as
+.BR \-bmp ).
+British spellings are also accepted (e.g.,
+.BR \-greyscale ),
+though for brevity these are not mentioned below.
+.PP
+The basic switches are:
+.TP
+.BI \-colors " N"
+Reduce image to at most N colors.  This reduces the number of colors used in
+the output image, so that it can be displayed on a colormapped display or
+stored in a colormapped file format.  For example, if you have an 8-bit
+display, you'd need to reduce to 256 or fewer colors.
+.TP
+.BI \-quantize " N"
+Same as
+.BR \-colors .
+.B \-colors
+is the recommended name,
+.B \-quantize
+is provided only for backwards compatibility.
+.TP
+.B \-fast
+Select recommended processing options for fast, low quality output.  (The
+default options are chosen for highest quality output.)  Currently, this is
+equivalent to \fB\-dct fast \-nosmooth \-onepass \-dither ordered\fR.
+.TP
+.B \-grayscale
+Force gray-scale output even if JPEG file is color.  Useful for viewing on
+monochrome displays; also,
+.B djpeg
+runs noticeably faster in this mode.
+.TP
+.BI \-scale " M/N"
+Scale the output image by a factor M/N.  Currently the scale factor must be
+1/1, 1/2, 1/4, or 1/8.  Scaling is handy if the image is larger than your
+screen; also,
+.B djpeg
+runs much faster when scaling down the output.
+.TP
+.B \-bmp
+Select BMP output format (Windows flavor).  8-bit colormapped format is
+emitted if
+.B \-colors
+or
+.B \-grayscale
+is specified, or if the JPEG file is gray-scale; otherwise, 24-bit full-color
+format is emitted.
+.TP
+.B \-gif
+Select GIF output format.  Since GIF does not support more than 256 colors,
+.B \-colors 256
+is assumed (unless you specify a smaller number of colors).
+.TP
+.B \-os2
+Select BMP output format (OS/2 1.x flavor).  8-bit colormapped format is
+emitted if
+.B \-colors
+or
+.B \-grayscale
+is specified, or if the JPEG file is gray-scale; otherwise, 24-bit full-color
+format is emitted.
+.TP
+.B \-pnm
+Select PBMPLUS (PPM/PGM) output format (this is the default format).
+PGM is emitted if the JPEG file is gray-scale or if
+.B \-grayscale
+is specified; otherwise PPM is emitted.
+.TP
+.B \-rle
+Select RLE output format.  (Requires URT library.)
+.TP
+.B \-targa
+Select Targa output format.  Gray-scale format is emitted if the JPEG file is
+gray-scale or if
+.B \-grayscale
+is specified; otherwise, colormapped format is emitted if
+.B \-colors
+is specified; otherwise, 24-bit full-color format is emitted.
+.PP
+Switches for advanced users:
+.TP
+.B \-dct int
+Use integer DCT method (default).
+.TP
+.B \-dct fast
+Use fast integer DCT (less accurate).
+.TP
+.B \-dct float
+Use floating-point DCT method.
+The float method is very slightly more accurate than the int method, but is
+much slower unless your machine has very fast floating-point hardware.  Also
+note that results of the floating-point method may vary slightly across
+machines, while the integer methods should give the same results everywhere.
+The fast integer method is much less accurate than the other two.
+.TP
+.B \-dither fs
+Use Floyd-Steinberg dithering in color quantization.
+.TP
+.B \-dither ordered
+Use ordered dithering in color quantization.
+.TP
+.B \-dither none
+Do not use dithering in color quantization.
+By default, Floyd-Steinberg dithering is applied when quantizing colors; this
+is slow but usually produces the best results.  Ordered dither is a compromise
+between speed and quality; no dithering is fast but usually looks awful.  Note
+that these switches have no effect unless color quantization is being done.
+Ordered dither is only available in
+.B \-onepass
+mode.
+.TP
+.BI \-map " file"
+Quantize to the colors used in the specified image file.  This is useful for
+producing multiple files with identical color maps, or for forcing a
+predefined set of colors to be used.  The
+.I file
+must be a GIF or PPM file. This option overrides
+.B \-colors
+and
+.BR \-onepass .
+.TP
+.B \-nosmooth
+Use a faster, lower-quality upsampling routine.
+.TP
+.B \-onepass
+Use one-pass instead of two-pass color quantization.  The one-pass method is
+faster and needs less memory, but it produces a lower-quality image.
+.B \-onepass
+is ignored unless you also say
+.B \-colors
+.IR N .
+Also, the one-pass method is always used for gray-scale output (the two-pass
+method is no improvement then).
+.TP
+.BI \-maxmemory " N"
+Set limit for amount of memory to use in processing large images.  Value is
+in thousands of bytes, or millions of bytes if "M" is attached to the
+number.  For example,
+.B \-max 4m
+selects 4000000 bytes.  If more space is needed, temporary files will be used.
+.TP
+.BI \-outfile " name"
+Send output image to the named file, not to standard output.
+.TP
+.B \-verbose
+Enable debug printout.  More
+.BR \-v 's
+give more output.  Also, version information is printed at startup.
+.TP
+.B \-debug
+Same as
+.BR \-verbose .
+.SH EXAMPLES
+.LP
+This example decompresses the JPEG file foo.jpg, quantizes it to
+256 colors, and saves the output in 8-bit BMP format in foo.bmp:
+.IP
+.B djpeg \-colors 256 \-bmp
+.I foo.jpg
+.B >
+.I foo.bmp
+.SH HINTS
+To get a quick preview of an image, use the
+.B \-grayscale
+and/or
+.B \-scale
+switches.
+.B \-grayscale \-scale 1/8
+is the fastest case.
+.PP
+Several options are available that trade off image quality to gain speed.
+.B \-fast
+turns on the recommended settings.
+.PP
+.B \-dct fast
+and/or
+.B \-nosmooth
+gain speed at a small sacrifice in quality.
+When producing a color-quantized image,
+.B \-onepass \-dither ordered
+is fast but much lower quality than the default behavior.
+.B \-dither none
+may give acceptable results in two-pass mode, but is seldom tolerable in
+one-pass mode.
+.PP
+If you are fortunate enough to have very fast floating point hardware,
+\fB\-dct float\fR may be even faster than \fB\-dct fast\fR.  But on most
+machines \fB\-dct float\fR is slower than \fB\-dct int\fR; in this case it is
+not worth using, because its theoretical accuracy advantage is too small to be
+significant in practice.
+.SH ENVIRONMENT
+.TP
+.B JPEGMEM
+If this environment variable is set, its value is the default memory limit.
+The value is specified as described for the
+.B \-maxmemory
+switch.
+.B JPEGMEM
+overrides the default value specified when the program was compiled, and
+itself is overridden by an explicit
+.BR \-maxmemory .
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR jpegtran (1),
+.BR rdjpgcom (1),
+.BR wrjpgcom (1)
+.br
+.BR ppm (5),
+.BR pgm (5)
+.br
+Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
+.SH AUTHOR
+Independent JPEG Group
+.SH BUGS
+Arithmetic coding is not supported for legal reasons.
+.PP
+To avoid the Unisys LZW patent,
+.B djpeg
+produces uncompressed GIF files.  These are larger than they should be, but
+are readable by standard GIF decoders.
+.PP
+Still not as fast as we'd like.
diff --git a/libjpegtwrp/djpeg.c b/libjpegtwrp/djpeg.c
new file mode 100644
index 000000000..e099e90ae
--- /dev/null
+++ b/libjpegtwrp/djpeg.c
@@ -0,0 +1,616 @@
+/*
+ * djpeg.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a command-line user interface for the JPEG decompressor.
+ * It should work on any system with Unix- or MS-DOS-style command lines.
+ *
+ * Two different command line styles are permitted, depending on the
+ * compile-time switch TWO_FILE_COMMANDLINE:
+ *	djpeg [options]  inputfile outputfile
+ *	djpeg [options]  [inputfile]
+ * In the second style, output is always to standard output, which you'd
+ * normally redirect to a file or pipe to some other program.  Input is
+ * either from a named file or from standard input (typically redirected).
+ * The second style is convenient on Unix but is unhelpful on systems that
+ * don't support pipes.  Also, you MUST use the first style if your system
+ * doesn't do binary I/O to stdin/stdout.
+ * To simplify script writing, the "-outfile" switch is provided.  The syntax
+ *	djpeg [options]  -outfile outputfile  inputfile
+ * works regardless of which command line style is used.
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+#include "jversion.h"		/* for version message */
+
+#include <ctype.h>		/* to declare isprint() */
+
+#ifdef USE_CCOMMAND		/* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>		/* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>		/* Think declares it here */
+#endif
+#endif
+
+
+/* Create the add-on message string table. */
+
+#define JMESSAGE(code,string)	string ,
+
+static const char * const cdjpeg_message_table[] = {
+#include "cderror.h"
+  NULL
+};
+
+
+/*
+ * This list defines the known output image formats
+ * (not all of which need be supported by a given version).
+ * You can change the default output format by defining DEFAULT_FMT;
+ * indeed, you had better do so if you undefine PPM_SUPPORTED.
+ */
+
+typedef enum {
+	FMT_BMP,		/* BMP format (Windows flavor) */
+	FMT_GIF,		/* GIF format */
+	FMT_OS2,		/* BMP format (OS/2 flavor) */
+	FMT_PPM,		/* PPM/PGM (PBMPLUS formats) */
+	FMT_RLE,		/* RLE format */
+	FMT_TARGA,		/* Targa format */
+	FMT_TIFF		/* TIFF format */
+} IMAGE_FORMATS;
+
+#ifndef DEFAULT_FMT		/* so can override from CFLAGS in Makefile */
+#define DEFAULT_FMT	FMT_PPM
+#endif
+
+static IMAGE_FORMATS requested_fmt;
+
+
+/*
+ * Argument-parsing code.
+ * The switch parser is designed to be useful with DOS-style command line
+ * syntax, ie, intermixed switches and file names, where only the switches
+ * to the left of a given file name affect processing of that file.
+ * The main program in this file doesn't actually use this capability...
+ */
+
+
+static const char * progname;	/* program name for error messages */
+static char * outfilename;	/* for -outfile switch */
+
+
+LOCAL(void)
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "usage: %s [switches] ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "[inputfile]\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -colors N      Reduce image to no more than N colors\n");
+  fprintf(stderr, "  -fast          Fast, low-quality processing\n");
+  fprintf(stderr, "  -grayscale     Force grayscale output\n");
+#ifdef IDCT_SCALING_SUPPORTED
+  fprintf(stderr, "  -scale M/N     Scale output image by fraction M/N, eg, 1/8\n");
+#endif
+#ifdef BMP_SUPPORTED
+  fprintf(stderr, "  -bmp           Select BMP output format (Windows style)%s\n",
+	  (DEFAULT_FMT == FMT_BMP ? " (default)" : ""));
+#endif
+#ifdef GIF_SUPPORTED
+  fprintf(stderr, "  -gif           Select GIF output format%s\n",
+	  (DEFAULT_FMT == FMT_GIF ? " (default)" : ""));
+#endif
+#ifdef BMP_SUPPORTED
+  fprintf(stderr, "  -os2           Select BMP output format (OS/2 style)%s\n",
+	  (DEFAULT_FMT == FMT_OS2 ? " (default)" : ""));
+#endif
+#ifdef PPM_SUPPORTED
+  fprintf(stderr, "  -pnm           Select PBMPLUS (PPM/PGM) output format%s\n",
+	  (DEFAULT_FMT == FMT_PPM ? " (default)" : ""));
+#endif
+#ifdef RLE_SUPPORTED
+  fprintf(stderr, "  -rle           Select Utah RLE output format%s\n",
+	  (DEFAULT_FMT == FMT_RLE ? " (default)" : ""));
+#endif
+#ifdef TARGA_SUPPORTED
+  fprintf(stderr, "  -targa         Select Targa output format%s\n",
+	  (DEFAULT_FMT == FMT_TARGA ? " (default)" : ""));
+#endif
+  fprintf(stderr, "Switches for advanced users:\n");
+#ifdef DCT_ISLOW_SUPPORTED
+  fprintf(stderr, "  -dct int       Use integer DCT method%s\n",
+	  (JDCT_DEFAULT == JDCT_ISLOW ? " (default)" : ""));
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+  fprintf(stderr, "  -dct fast      Use fast integer DCT (less accurate)%s\n",
+	  (JDCT_DEFAULT == JDCT_IFAST ? " (default)" : ""));
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  fprintf(stderr, "  -dct float     Use floating-point DCT method%s\n",
+	  (JDCT_DEFAULT == JDCT_FLOAT ? " (default)" : ""));
+#endif
+  fprintf(stderr, "  -dither fs     Use F-S dithering (default)\n");
+  fprintf(stderr, "  -dither none   Don't use dithering in quantization\n");
+  fprintf(stderr, "  -dither ordered  Use ordered dither (medium speed, quality)\n");
+#ifdef QUANT_2PASS_SUPPORTED
+  fprintf(stderr, "  -map FILE      Map to colors used in named image file\n");
+#endif
+  fprintf(stderr, "  -nosmooth      Don't use high-quality upsampling\n");
+#ifdef QUANT_1PASS_SUPPORTED
+  fprintf(stderr, "  -onepass       Use 1-pass quantization (fast, low quality)\n");
+#endif
+  fprintf(stderr, "  -maxmemory N   Maximum memory to use (in kbytes)\n");
+  fprintf(stderr, "  -outfile name  Specify name for output file\n");
+  fprintf(stderr, "  -verbose  or  -debug   Emit debug output\n");
+  exit(EXIT_FAILURE);
+}
+
+
+LOCAL(int)
+parse_switches (j_decompress_ptr cinfo, int argc, char **argv,
+		int last_file_arg_seen, boolean for_real)
+/* Parse optional switches.
+ * Returns argv[] index of first file-name argument (== argc if none).
+ * Any file names with indexes <= last_file_arg_seen are ignored;
+ * they have presumably been processed in a previous iteration.
+ * (Pass 0 for last_file_arg_seen on the first or only iteration.)
+ * for_real is FALSE on the first (dummy) pass; we may skip any expensive
+ * processing.
+ */
+{
+  int argn;
+  char * arg;
+
+  /* Set up default JPEG parameters. */
+  requested_fmt = DEFAULT_FMT;	/* set default output file format */
+  outfilename = NULL;
+  cinfo->err->trace_level = 0;
+
+  /* Scan command line options, adjust parameters */
+
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (*arg != '-') {
+      /* Not a switch, must be a file name argument */
+      if (argn <= last_file_arg_seen) {
+	outfilename = NULL;	/* -outfile applies to just one input file */
+	continue;		/* ignore this name if previously processed */
+      }
+      break;			/* else done parsing switches */
+    }
+    arg++;			/* advance past switch marker character */
+
+    if (keymatch(arg, "bmp", 1)) {
+      /* BMP output format. */
+      requested_fmt = FMT_BMP;
+
+    } else if (keymatch(arg, "colors", 1) || keymatch(arg, "colours", 1) ||
+	       keymatch(arg, "quantize", 1) || keymatch(arg, "quantise", 1)) {
+      /* Do color quantization. */
+      int val;
+
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (sscanf(argv[argn], "%d", &val) != 1)
+	usage();
+      cinfo->desired_number_of_colors = val;
+      cinfo->quantize_colors = TRUE;
+
+    } else if (keymatch(arg, "dct", 2)) {
+      /* Select IDCT algorithm. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (keymatch(argv[argn], "int", 1)) {
+	cinfo->dct_method = JDCT_ISLOW;
+      } else if (keymatch(argv[argn], "fast", 2)) {
+	cinfo->dct_method = JDCT_IFAST;
+      } else if (keymatch(argv[argn], "float", 2)) {
+	cinfo->dct_method = JDCT_FLOAT;
+      } else
+	usage();
+
+    } else if (keymatch(arg, "dither", 2)) {
+      /* Select dithering algorithm. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (keymatch(argv[argn], "fs", 2)) {
+	cinfo->dither_mode = JDITHER_FS;
+      } else if (keymatch(argv[argn], "none", 2)) {
+	cinfo->dither_mode = JDITHER_NONE;
+      } else if (keymatch(argv[argn], "ordered", 2)) {
+	cinfo->dither_mode = JDITHER_ORDERED;
+      } else
+	usage();
+
+    } else if (keymatch(arg, "debug", 1) || keymatch(arg, "verbose", 1)) {
+      /* Enable debug printouts. */
+      /* On first -d, print version identification */
+      static boolean printed_version = FALSE;
+
+      if (! printed_version) {
+	fprintf(stderr, "Independent JPEG Group's DJPEG, version %s\n%s\n",
+		JVERSION, JCOPYRIGHT);
+	printed_version = TRUE;
+      }
+      cinfo->err->trace_level++;
+
+    } else if (keymatch(arg, "fast", 1)) {
+      /* Select recommended processing options for quick-and-dirty output. */
+      cinfo->two_pass_quantize = FALSE;
+      cinfo->dither_mode = JDITHER_ORDERED;
+      if (! cinfo->quantize_colors) /* don't override an earlier -colors */
+	cinfo->desired_number_of_colors = 216;
+      cinfo->dct_method = JDCT_FASTEST;
+      cinfo->do_fancy_upsampling = FALSE;
+
+    } else if (keymatch(arg, "gif", 1)) {
+      /* GIF output format. */
+      requested_fmt = FMT_GIF;
+
+    } else if (keymatch(arg, "grayscale", 2) || keymatch(arg, "greyscale",2)) {
+      /* Force monochrome output. */
+      cinfo->out_color_space = JCS_GRAYSCALE;
+
+    } else if (keymatch(arg, "map", 3)) {
+      /* Quantize to a color map taken from an input file. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (for_real) {		/* too expensive to do twice! */
+#ifdef QUANT_2PASS_SUPPORTED	/* otherwise can't quantize to supplied map */
+	FILE * mapfile;
+
+	if ((mapfile = fopen(argv[argn], READ_BINARY)) == NULL) {
+	  fprintf(stderr, "%s: can't open %s\n", progname, argv[argn]);
+	  exit(EXIT_FAILURE);
+	}
+	read_color_map(cinfo, mapfile);
+	fclose(mapfile);
+	cinfo->quantize_colors = TRUE;
+#else
+	ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+      }
+
+    } else if (keymatch(arg, "maxmemory", 3)) {
+      /* Maximum memory in Kb (or Mb with 'm'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+	usage();
+      if (ch == 'm' || ch == 'M')
+	lval *= 1000L;
+      cinfo->mem->max_memory_to_use = lval * 1000L;
+
+    } else if (keymatch(arg, "nosmooth", 3)) {
+      /* Suppress fancy upsampling */
+      cinfo->do_fancy_upsampling = FALSE;
+
+    } else if (keymatch(arg, "onepass", 3)) {
+      /* Use fast one-pass quantization. */
+      cinfo->two_pass_quantize = FALSE;
+
+    } else if (keymatch(arg, "os2", 3)) {
+      /* BMP output format (OS/2 flavor). */
+      requested_fmt = FMT_OS2;
+
+    } else if (keymatch(arg, "outfile", 4)) {
+      /* Set output file name. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      outfilename = argv[argn];	/* save it away for later use */
+
+    } else if (keymatch(arg, "pnm", 1) || keymatch(arg, "ppm", 1)) {
+      /* PPM/PGM output format. */
+      requested_fmt = FMT_PPM;
+
+    } else if (keymatch(arg, "rle", 1)) {
+      /* RLE output format. */
+      requested_fmt = FMT_RLE;
+
+    } else if (keymatch(arg, "scale", 1)) {
+      /* Scale the output image by a fraction M/N. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (sscanf(argv[argn], "%d/%d",
+		 &cinfo->scale_num, &cinfo->scale_denom) != 2)
+	usage();
+
+    } else if (keymatch(arg, "targa", 1)) {
+      /* Targa output format. */
+      requested_fmt = FMT_TARGA;
+
+    } else {
+      usage();			/* bogus switch */
+    }
+  }
+
+  return argn;			/* return index of next arg (file name) */
+}
+
+
+/*
+ * Marker processor for COM and interesting APPn markers.
+ * This replaces the library's built-in processor, which just skips the marker.
+ * We want to print out the marker as text, to the extent possible.
+ * Note this code relies on a non-suspending data source.
+ */
+
+LOCAL(unsigned int)
+jpeg_getc (j_decompress_ptr cinfo)
+/* Read next byte */
+{
+  struct jpeg_source_mgr * datasrc = cinfo->src;
+
+  if (datasrc->bytes_in_buffer == 0) {
+    if (! (*datasrc->fill_input_buffer) (cinfo))
+      ERREXIT(cinfo, JERR_CANT_SUSPEND);
+  }
+  datasrc->bytes_in_buffer--;
+  return GETJOCTET(*datasrc->next_input_byte++);
+}
+
+
+METHODDEF(boolean)
+print_text_marker (j_decompress_ptr cinfo)
+{
+  boolean traceit = (cinfo->err->trace_level >= 1);
+  INT32 length;
+  unsigned int ch;
+  unsigned int lastch = 0;
+
+  length = jpeg_getc(cinfo) << 8;
+  length += jpeg_getc(cinfo);
+  length -= 2;			/* discount the length word itself */
+
+  if (traceit) {
+    if (cinfo->unread_marker == JPEG_COM)
+      fprintf(stderr, "Comment, length %ld:\n", (long) length);
+    else			/* assume it is an APPn otherwise */
+      fprintf(stderr, "APP%d, length %ld:\n",
+	      cinfo->unread_marker - JPEG_APP0, (long) length);
+  }
+
+  while (--length >= 0) {
+    ch = jpeg_getc(cinfo);
+    if (traceit) {
+      /* Emit the character in a readable form.
+       * Nonprintables are converted to \nnn form,
+       * while \ is converted to \\.
+       * Newlines in CR, CR/LF, or LF form will be printed as one newline.
+       */
+      if (ch == '\r') {
+	fprintf(stderr, "\n");
+      } else if (ch == '\n') {
+	if (lastch != '\r')
+	  fprintf(stderr, "\n");
+      } else if (ch == '\\') {
+	fprintf(stderr, "\\\\");
+      } else if (isprint(ch)) {
+	putc(ch, stderr);
+      } else {
+	fprintf(stderr, "\\%03o", ch);
+      }
+      lastch = ch;
+    }
+  }
+
+  if (traceit)
+    fprintf(stderr, "\n");
+
+  return TRUE;
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  struct jpeg_decompress_struct cinfo;
+  struct jpeg_error_mgr jerr;
+#ifdef PROGRESS_REPORT
+  struct cdjpeg_progress_mgr progress;
+#endif
+  int file_index;
+  djpeg_dest_ptr dest_mgr = NULL;
+  FILE * input_file;
+  FILE * output_file;
+  JDIMENSION num_scanlines;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "djpeg";		/* in case C library doesn't provide it */
+
+  /* Initialize the JPEG decompression object with default error handling. */
+  cinfo.err = jpeg_std_error(&jerr);
+  jpeg_create_decompress(&cinfo);
+  /* Add some application-specific error messages (from cderror.h) */
+  jerr.addon_message_table = cdjpeg_message_table;
+  jerr.first_addon_message = JMSG_FIRSTADDONCODE;
+  jerr.last_addon_message = JMSG_LASTADDONCODE;
+
+  /* Insert custom marker processor for COM and APP12.
+   * APP12 is used by some digital camera makers for textual info,
+   * so we provide the ability to display it as text.
+   * If you like, additional APPn marker types can be selected for display,
+   * but don't try to override APP0 or APP14 this way (see libjpeg.doc).
+   */
+  jpeg_set_marker_processor(&cinfo, JPEG_COM, print_text_marker);
+  jpeg_set_marker_processor(&cinfo, JPEG_APP0+12, print_text_marker);
+
+  /* Now safe to enable signal catcher. */
+#ifdef NEED_SIGNAL_CATCHER
+  enable_signal_catcher((j_common_ptr) &cinfo);
+#endif
+
+  /* Scan command line to find file names. */
+  /* It is convenient to use just one switch-parsing routine, but the switch
+   * values read here are ignored; we will rescan the switches after opening
+   * the input file.
+   * (Exception: tracing level set here controls verbosity for COM markers
+   * found during jpeg_read_header...)
+   */
+
+  file_index = parse_switches(&cinfo, argc, argv, 0, FALSE);
+
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have either -outfile switch or explicit output file name */
+  if (outfilename == NULL) {
+    if (file_index != argc-2) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+	      progname);
+      usage();
+    }
+    outfilename = argv[file_index+1];
+  } else {
+    if (file_index != argc-1) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+	      progname);
+      usage();
+    }
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (file_index < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Open the input file. */
+  if (file_index < argc) {
+    if ((input_file = fopen(argv[file_index], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[file_index]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+    input_file = read_stdin();
+  }
+
+  /* Open the output file. */
+  if (outfilename != NULL) {
+    if ((output_file = fopen(outfilename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, outfilename);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default output file is stdout */
+    output_file = write_stdout();
+  }
+
+#ifdef PROGRESS_REPORT
+  start_progress_monitor((j_common_ptr) &cinfo, &progress);
+#endif
+
+  /* Specify data source for decompression */
+  jpeg_stdio_src(&cinfo, input_file);
+
+  /* Read file header, set default decompression parameters */
+  (void) jpeg_read_header(&cinfo, TRUE);
+
+  /* Adjust default decompression parameters by re-parsing the options */
+  file_index = parse_switches(&cinfo, argc, argv, 0, TRUE);
+
+  /* Initialize the output module now to let it override any crucial
+   * option settings (for instance, GIF wants to force color quantization).
+   */
+  switch (requested_fmt) {
+#ifdef BMP_SUPPORTED
+  case FMT_BMP:
+    dest_mgr = jinit_write_bmp(&cinfo, FALSE);
+    break;
+  case FMT_OS2:
+    dest_mgr = jinit_write_bmp(&cinfo, TRUE);
+    break;
+#endif
+#ifdef GIF_SUPPORTED
+  case FMT_GIF:
+    dest_mgr = jinit_write_gif(&cinfo);
+    break;
+#endif
+#ifdef PPM_SUPPORTED
+  case FMT_PPM:
+    dest_mgr = jinit_write_ppm(&cinfo);
+    break;
+#endif
+#ifdef RLE_SUPPORTED
+  case FMT_RLE:
+    dest_mgr = jinit_write_rle(&cinfo);
+    break;
+#endif
+#ifdef TARGA_SUPPORTED
+  case FMT_TARGA:
+    dest_mgr = jinit_write_targa(&cinfo);
+    break;
+#endif
+  default:
+    ERREXIT(&cinfo, JERR_UNSUPPORTED_FORMAT);
+    break;
+  }
+  dest_mgr->output_file = output_file;
+
+  /* Start decompressor */
+  (void) jpeg_start_decompress(&cinfo);
+
+  /* Write output file header */
+  (*dest_mgr->start_output) (&cinfo, dest_mgr);
+
+  /* Process data */
+  while (cinfo.output_scanline < cinfo.output_height) {
+    num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+					dest_mgr->buffer_height);
+    (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+  }
+
+#ifdef PROGRESS_REPORT
+  /* Hack: count final pass as done in case finish_output does an extra pass.
+   * The library won't have updated completed_passes.
+   */
+  progress.pub.completed_passes = progress.pub.total_passes;
+#endif
+
+  /* Finish decompression and release memory.
+   * I must do it in this order because output module has allocated memory
+   * of lifespan JPOOL_IMAGE; it needs to finish before releasing memory.
+   */
+  (*dest_mgr->finish_output) (&cinfo, dest_mgr);
+  (void) jpeg_finish_decompress(&cinfo);
+  jpeg_destroy_decompress(&cinfo);
+
+  /* Close files, if we opened them */
+  if (input_file != stdin)
+    fclose(input_file);
+  if (output_file != stdout)
+    fclose(output_file);
+
+#ifdef PROGRESS_REPORT
+  end_progress_monitor((j_common_ptr) &cinfo);
+#endif
+
+  /* All done. */
+  exit(jerr.num_warnings ? EXIT_WARNING : EXIT_SUCCESS);
+  return 0;			/* suppress no-return-value warnings */
+}
diff --git a/libjpegtwrp/example.c b/libjpegtwrp/example.c
new file mode 100644
index 000000000..7fc354f04
--- /dev/null
+++ b/libjpegtwrp/example.c
@@ -0,0 +1,433 @@
+/*
+ * example.c
+ *
+ * This file illustrates how to use the IJG code as a subroutine library
+ * to read or write JPEG image files.  You should look at this code in
+ * conjunction with the documentation file libjpeg.doc.
+ *
+ * This code will not do anything useful as-is, but it may be helpful as a
+ * skeleton for constructing routines that call the JPEG library.  
+ *
+ * We present these routines in the same coding style used in the JPEG code
+ * (ANSI function definitions, etc); but you are of course free to code your
+ * routines in a different style if you prefer.
+ */
+
+#include <stdio.h>
+
+/*
+ * Include file for users of JPEG library.
+ * You will need to have included system headers that define at least
+ * the typedefs FILE and size_t before you can include jpeglib.h.
+ * (stdio.h is sufficient on ANSI-conforming systems.)
+ * You may also wish to include "jerror.h".
+ */
+
+#include "jpeglib.h"
+
+/*
+ * <setjmp.h> is used for the optional error recovery mechanism shown in
+ * the second part of the example.
+ */
+
+#include <setjmp.h>
+
+
+
+/******************** JPEG COMPRESSION SAMPLE INTERFACE *******************/
+
+/* This half of the example shows how to feed data into the JPEG compressor.
+ * We present a minimal version that does not worry about refinements such
+ * as error recovery (the JPEG code will just exit() if it gets an error).
+ */
+
+
+/*
+ * IMAGE DATA FORMATS:
+ *
+ * The standard input image format is a rectangular array of pixels, with
+ * each pixel having the same number of "component" values (color channels).
+ * Each pixel row is an array of JSAMPLEs (which typically are unsigned chars).
+ * If you are working with color data, then the color values for each pixel
+ * must be adjacent in the row; for example, R,G,B,R,G,B,R,G,B,... for 24-bit
+ * RGB color.
+ *
+ * For this example, we'll assume that this data structure matches the way
+ * our application has stored the image in memory, so we can just pass a
+ * pointer to our image buffer.  In particular, let's say that the image is
+ * RGB color and is described by:
+ */
+
+extern JSAMPLE * image_buffer;	/* Points to large array of R,G,B-order data */
+extern int image_height;	/* Number of rows in image */
+extern int image_width;		/* Number of columns in image */
+
+
+/*
+ * Sample routine for JPEG compression.  We assume that the target file name
+ * and a compression quality factor are passed in.
+ */
+
+GLOBAL(void)
+write_JPEG_file (char * filename, int quality)
+{
+  /* This struct contains the JPEG compression parameters and pointers to
+   * working space (which is allocated as needed by the JPEG library).
+   * It is possible to have several such structures, representing multiple
+   * compression/decompression processes, in existence at once.  We refer
+   * to any one struct (and its associated working data) as a "JPEG object".
+   */
+  struct jpeg_compress_struct cinfo;
+  /* This struct represents a JPEG error handler.  It is declared separately
+   * because applications often want to supply a specialized error handler
+   * (see the second half of this file for an example).  But here we just
+   * take the easy way out and use the standard error handler, which will
+   * print a message on stderr and call exit() if compression fails.
+   * Note that this struct must live as long as the main JPEG parameter
+   * struct, to avoid dangling-pointer problems.
+   */
+  struct jpeg_error_mgr jerr;
+  /* More stuff */
+  FILE * outfile;		/* target file */
+  JSAMPROW row_pointer[1];	/* pointer to JSAMPLE row[s] */
+  int row_stride;		/* physical row width in image buffer */
+
+  /* Step 1: allocate and initialize JPEG compression object */
+
+  /* We have to set up the error handler first, in case the initialization
+   * step fails.  (Unlikely, but it could happen if you are out of memory.)
+   * This routine fills in the contents of struct jerr, and returns jerr's
+   * address which we place into the link field in cinfo.
+   */
+  cinfo.err = jpeg_std_error(&jerr);
+  /* Now we can initialize the JPEG compression object. */
+  jpeg_create_compress(&cinfo);
+
+  /* Step 2: specify data destination (eg, a file) */
+  /* Note: steps 2 and 3 can be done in either order. */
+
+  /* Here we use the library-supplied code to send compressed data to a
+   * stdio stream.  You can also write your own code to do something else.
+   * VERY IMPORTANT: use "b" option to fopen() if you are on a machine that
+   * requires it in order to write binary files.
+   */
+  if ((outfile = fopen(filename, "wb")) == NULL) {
+    fprintf(stderr, "can't open %s\n", filename);
+    exit(1);
+  }
+  jpeg_stdio_dest(&cinfo, outfile);
+
+  /* Step 3: set parameters for compression */
+
+  /* First we supply a description of the input image.
+   * Four fields of the cinfo struct must be filled in:
+   */
+  cinfo.image_width = image_width; 	/* image width and height, in pixels */
+  cinfo.image_height = image_height;
+  cinfo.input_components = 3;		/* # of color components per pixel */
+  cinfo.in_color_space = JCS_RGB; 	/* colorspace of input image */
+  /* Now use the library's routine to set default compression parameters.
+   * (You must set at least cinfo.in_color_space before calling this,
+   * since the defaults depend on the source color space.)
+   */
+  jpeg_set_defaults(&cinfo);
+  /* Now you can set any non-default parameters you wish to.
+   * Here we just illustrate the use of quality (quantization table) scaling:
+   */
+  jpeg_set_quality(&cinfo, quality, TRUE /* limit to baseline-JPEG values */);
+
+  /* Step 4: Start compressor */
+
+  /* TRUE ensures that we will write a complete interchange-JPEG file.
+   * Pass TRUE unless you are very sure of what you're doing.
+   */
+  jpeg_start_compress(&cinfo, TRUE);
+
+  /* Step 5: while (scan lines remain to be written) */
+  /*           jpeg_write_scanlines(...); */
+
+  /* Here we use the library's state variable cinfo.next_scanline as the
+   * loop counter, so that we don't have to keep track ourselves.
+   * To keep things simple, we pass one scanline per call; you can pass
+   * more if you wish, though.
+   */
+  row_stride = image_width * 3;	/* JSAMPLEs per row in image_buffer */
+
+  while (cinfo.next_scanline < cinfo.image_height) {
+    /* jpeg_write_scanlines expects an array of pointers to scanlines.
+     * Here the array is only one element long, but you could pass
+     * more than one scanline at a time if that's more convenient.
+     */
+    row_pointer[0] = & image_buffer[cinfo.next_scanline * row_stride];
+    (void) jpeg_write_scanlines(&cinfo, row_pointer, 1);
+  }
+
+  /* Step 6: Finish compression */
+
+  jpeg_finish_compress(&cinfo);
+  /* After finish_compress, we can close the output file. */
+  fclose(outfile);
+
+  /* Step 7: release JPEG compression object */
+
+  /* This is an important step since it will release a good deal of memory. */
+  jpeg_destroy_compress(&cinfo);
+
+  /* And we're done! */
+}
+
+
+/*
+ * SOME FINE POINTS:
+ *
+ * In the above loop, we ignored the return value of jpeg_write_scanlines,
+ * which is the number of scanlines actually written.  We could get away
+ * with this because we were only relying on the value of cinfo.next_scanline,
+ * which will be incremented correctly.  If you maintain additional loop
+ * variables then you should be careful to increment them properly.
+ * Actually, for output to a stdio stream you needn't worry, because
+ * then jpeg_write_scanlines will write all the lines passed (or else exit
+ * with a fatal error).  Partial writes can only occur if you use a data
+ * destination module that can demand suspension of the compressor.
+ * (If you don't know what that's for, you don't need it.)
+ *
+ * If the compressor requires full-image buffers (for entropy-coding
+ * optimization or a multi-scan JPEG file), it will create temporary
+ * files for anything that doesn't fit within the maximum-memory setting.
+ * (Note that temp files are NOT needed if you use the default parameters.)
+ * On some systems you may need to set up a signal handler to ensure that
+ * temporary files are deleted if the program is interrupted.  See libjpeg.doc.
+ *
+ * Scanlines MUST be supplied in top-to-bottom order if you want your JPEG
+ * files to be compatible with everyone else's.  If you cannot readily read
+ * your data in that order, you'll need an intermediate array to hold the
+ * image.  See rdtarga.c or rdbmp.c for examples of handling bottom-to-top
+ * source data using the JPEG code's internal virtual-array mechanisms.
+ */
+
+
+
+/******************** JPEG DECOMPRESSION SAMPLE INTERFACE *******************/
+
+/* This half of the example shows how to read data from the JPEG decompressor.
+ * It's a bit more refined than the above, in that we show:
+ *   (a) how to modify the JPEG library's standard error-reporting behavior;
+ *   (b) how to allocate workspace using the library's memory manager.
+ *
+ * Just to make this example a little different from the first one, we'll
+ * assume that we do not intend to put the whole image into an in-memory
+ * buffer, but to send it line-by-line someplace else.  We need a one-
+ * scanline-high JSAMPLE array as a work buffer, and we will let the JPEG
+ * memory manager allocate it for us.  This approach is actually quite useful
+ * because we don't need to remember to deallocate the buffer separately: it
+ * will go away automatically when the JPEG object is cleaned up.
+ */
+
+
+/*
+ * ERROR HANDLING:
+ *
+ * The JPEG library's standard error handler (jerror.c) is divided into
+ * several "methods" which you can override individually.  This lets you
+ * adjust the behavior without duplicating a lot of code, which you might
+ * have to update with each future release.
+ *
+ * Our example here shows how to override the "error_exit" method so that
+ * control is returned to the library's caller when a fatal error occurs,
+ * rather than calling exit() as the standard error_exit method does.
+ *
+ * We use C's setjmp/longjmp facility to return control.  This means that the
+ * routine which calls the JPEG library must first execute a setjmp() call to
+ * establish the return point.  We want the replacement error_exit to do a
+ * longjmp().  But we need to make the setjmp buffer accessible to the
+ * error_exit routine.  To do this, we make a private extension of the
+ * standard JPEG error handler object.  (If we were using C++, we'd say we
+ * were making a subclass of the regular error handler.)
+ *
+ * Here's the extended error handler struct:
+ */
+
+struct my_error_mgr {
+  struct jpeg_error_mgr pub;	/* "public" fields */
+
+  jmp_buf setjmp_buffer;	/* for return to caller */
+};
+
+typedef struct my_error_mgr * my_error_ptr;
+
+/*
+ * Here's the routine that will replace the standard error_exit method:
+ */
+
+METHODDEF(void)
+my_error_exit (j_common_ptr cinfo)
+{
+  /* cinfo->err really points to a my_error_mgr struct, so coerce pointer */
+  my_error_ptr myerr = (my_error_ptr) cinfo->err;
+
+  /* Always display the message. */
+  /* We could postpone this until after returning, if we chose. */
+  (*cinfo->err->output_message) (cinfo);
+
+  /* Return control to the setjmp point */
+  longjmp(myerr->setjmp_buffer, 1);
+}
+
+
+/*
+ * Sample routine for JPEG decompression.  We assume that the source file name
+ * is passed in.  We want to return 1 on success, 0 on error.
+ */
+
+
+GLOBAL(int)
+read_JPEG_file (char * filename)
+{
+  /* This struct contains the JPEG decompression parameters and pointers to
+   * working space (which is allocated as needed by the JPEG library).
+   */
+  struct jpeg_decompress_struct cinfo;
+  /* We use our private extension JPEG error handler.
+   * Note that this struct must live as long as the main JPEG parameter
+   * struct, to avoid dangling-pointer problems.
+   */
+  struct my_error_mgr jerr;
+  /* More stuff */
+  FILE * infile;		/* source file */
+  JSAMPARRAY buffer;		/* Output row buffer */
+  int row_stride;		/* physical row width in output buffer */
+
+  /* In this example we want to open the input file before doing anything else,
+   * so that the setjmp() error recovery below can assume the file is open.
+   * VERY IMPORTANT: use "b" option to fopen() if you are on a machine that
+   * requires it in order to read binary files.
+   */
+
+  if ((infile = fopen(filename, "rb")) == NULL) {
+    fprintf(stderr, "can't open %s\n", filename);
+    return 0;
+  }
+
+  /* Step 1: allocate and initialize JPEG decompression object */
+
+  /* We set up the normal JPEG error routines, then override error_exit. */
+  cinfo.err = jpeg_std_error(&jerr.pub);
+  jerr.pub.error_exit = my_error_exit;
+  /* Establish the setjmp return context for my_error_exit to use. */
+  if (setjmp(jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error.
+     * We need to clean up the JPEG object, close the input file, and return.
+     */
+    jpeg_destroy_decompress(&cinfo);
+    fclose(infile);
+    return 0;
+  }
+  /* Now we can initialize the JPEG decompression object. */
+  jpeg_create_decompress(&cinfo);
+
+  /* Step 2: specify data source (eg, a file) */
+
+  jpeg_stdio_src(&cinfo, infile);
+
+  /* Step 3: read file parameters with jpeg_read_header() */
+
+  (void) jpeg_read_header(&cinfo, TRUE);
+  /* We can ignore the return value from jpeg_read_header since
+   *   (a) suspension is not possible with the stdio data source, and
+   *   (b) we passed TRUE to reject a tables-only JPEG file as an error.
+   * See libjpeg.doc for more info.
+   */
+
+  /* Step 4: set parameters for decompression */
+
+  /* In this example, we don't need to change any of the defaults set by
+   * jpeg_read_header(), so we do nothing here.
+   */
+
+  /* Step 5: Start decompressor */
+
+  (void) jpeg_start_decompress(&cinfo);
+  /* We can ignore the return value since suspension is not possible
+   * with the stdio data source.
+   */
+
+  /* We may need to do some setup of our own at this point before reading
+   * the data.  After jpeg_start_decompress() we have the correct scaled
+   * output image dimensions available, as well as the output colormap
+   * if we asked for color quantization.
+   * In this example, we need to make an output work buffer of the right size.
+   */ 
+  /* JSAMPLEs per row in output buffer */
+  row_stride = cinfo.output_width * cinfo.output_components;
+  /* Make a one-row-high sample array that will go away when done with image */
+  buffer = (*cinfo.mem->alloc_sarray)
+		((j_common_ptr) &cinfo, JPOOL_IMAGE, row_stride, 1);
+
+  /* Step 6: while (scan lines remain to be read) */
+  /*           jpeg_read_scanlines(...); */
+
+  /* Here we use the library's state variable cinfo.output_scanline as the
+   * loop counter, so that we don't have to keep track ourselves.
+   */
+  while (cinfo.output_scanline < cinfo.output_height) {
+    /* jpeg_read_scanlines expects an array of pointers to scanlines.
+     * Here the array is only one element long, but you could ask for
+     * more than one scanline at a time if that's more convenient.
+     */
+    (void) jpeg_read_scanlines(&cinfo, buffer, 1);
+    /* Assume put_scanline_someplace wants a pointer and sample count. */
+    put_scanline_someplace(buffer[0], row_stride);
+  }
+
+  /* Step 7: Finish decompression */
+
+  (void) jpeg_finish_decompress(&cinfo);
+  /* We can ignore the return value since suspension is not possible
+   * with the stdio data source.
+   */
+
+  /* Step 8: Release JPEG decompression object */
+
+  /* This is an important step since it will release a good deal of memory. */
+  jpeg_destroy_decompress(&cinfo);
+
+  /* After finish_decompress, we can close the input file.
+   * Here we postpone it until after no more JPEG errors are possible,
+   * so as to simplify the setjmp error logic above.  (Actually, I don't
+   * think that jpeg_destroy can do an error exit, but why assume anything...)
+   */
+  fclose(infile);
+
+  /* At this point you may want to check to see whether any corrupt-data
+   * warnings occurred (test whether jerr.pub.num_warnings is nonzero).
+   */
+
+  /* And we're done! */
+  return 1;
+}
+
+
+/*
+ * SOME FINE POINTS:
+ *
+ * In the above code, we ignored the return value of jpeg_read_scanlines,
+ * which is the number of scanlines actually read.  We could get away with
+ * this because we asked for only one line at a time and we weren't using
+ * a suspending data source.  See libjpeg.doc for more info.
+ *
+ * We cheated a bit by calling alloc_sarray() after jpeg_start_decompress();
+ * we should have done it beforehand to ensure that the space would be
+ * counted against the JPEG max_memory setting.  In some systems the above
+ * code would risk an out-of-memory error.  However, in general we don't
+ * know the output image dimensions before jpeg_start_decompress(), unless we
+ * call jpeg_calc_output_dimensions().  See libjpeg.doc for more about this.
+ *
+ * Scanlines are returned in the same order as they appear in the JPEG file,
+ * which is standardly top-to-bottom.  If you must emit data bottom-to-top,
+ * you can use one of the virtual arrays provided by the JPEG memory manager
+ * to invert the data.  See wrbmp.c for an example.
+ *
+ * As with compression, some operating modes may require temporary files.
+ * On some systems you may need to set up a signal handler to ensure that
+ * temporary files are deleted if the program is interrupted.  See libjpeg.doc.
+ */
diff --git a/libjpegtwrp/filelist.doc b/libjpegtwrp/filelist.doc
new file mode 100644
index 000000000..e14982ca5
--- /dev/null
+++ b/libjpegtwrp/filelist.doc
@@ -0,0 +1,210 @@
+IJG JPEG LIBRARY:  FILE LIST
+
+Copyright (C) 1994-1998, Thomas G. Lane.
+This file is part of the Independent JPEG Group's software.
+For conditions of distribution and use, see the accompanying README file.
+
+
+Here is a road map to the files in the IJG JPEG distribution.  The
+distribution includes the JPEG library proper, plus two application
+programs ("cjpeg" and "djpeg") which use the library to convert JPEG
+files to and from some other popular image formats.  A third application
+"jpegtran" uses the library to do lossless conversion between different
+variants of JPEG.  There are also two stand-alone applications,
+"rdjpgcom" and "wrjpgcom".
+
+
+THE JPEG LIBRARY
+================
+
+Include files:
+
+jpeglib.h	JPEG library's exported data and function declarations.
+jconfig.h	Configuration declarations.  Note: this file is not present
+		in the distribution; it is generated during installation.
+jmorecfg.h	Additional configuration declarations; need not be changed
+		for a standard installation.
+jerror.h	Declares JPEG library's error and trace message codes.
+jinclude.h	Central include file used by all IJG .c files to reference
+		system include files.
+jpegint.h	JPEG library's internal data structures.
+jchuff.h	Private declarations for Huffman encoder modules.
+jdhuff.h	Private declarations for Huffman decoder modules.
+jdct.h		Private declarations for forward & reverse DCT subsystems.
+jmemsys.h	Private declarations for memory management subsystem.
+jversion.h	Version information.
+
+Applications using the library should include jpeglib.h (which in turn
+includes jconfig.h and jmorecfg.h).  Optionally, jerror.h may be included
+if the application needs to reference individual JPEG error codes.  The
+other include files are intended for internal use and would not normally
+be included by an application program.  (cjpeg/djpeg/etc do use jinclude.h,
+since its function is to improve portability of the whole IJG distribution.
+Most other applications will directly include the system include files they
+want, and hence won't need jinclude.h.)
+
+
+C source code files:
+
+These files contain most of the functions intended to be called directly by
+an application program:
+
+jcapimin.c	Application program interface: core routines for compression.
+jcapistd.c	Application program interface: standard compression.
+jdapimin.c	Application program interface: core routines for decompression.
+jdapistd.c	Application program interface: standard decompression.
+jcomapi.c	Application program interface routines common to compression
+		and decompression.
+jcparam.c	Compression parameter setting helper routines.
+jctrans.c	API and library routines for transcoding compression.
+jdtrans.c	API and library routines for transcoding decompression.
+
+Compression side of the library:
+
+jcinit.c	Initialization: determines which other modules to use.
+jcmaster.c	Master control: setup and inter-pass sequencing logic.
+jcmainct.c	Main buffer controller (preprocessor => JPEG compressor).
+jcprepct.c	Preprocessor buffer controller.
+jccoefct.c	Buffer controller for DCT coefficient buffer.
+jccolor.c	Color space conversion.
+jcsample.c	Downsampling.
+jcdctmgr.c	DCT manager (DCT implementation selection & control).
+jfdctint.c	Forward DCT using slow-but-accurate integer method.
+jfdctfst.c	Forward DCT using faster, less accurate integer method.
+jfdctflt.c	Forward DCT using floating-point arithmetic.
+jchuff.c	Huffman entropy coding for sequential JPEG.
+jcphuff.c	Huffman entropy coding for progressive JPEG.
+jcmarker.c	JPEG marker writing.
+jdatadst.c	Data destination manager for stdio output.
+
+Decompression side of the library:
+
+jdmaster.c	Master control: determines which other modules to use.
+jdinput.c	Input controller: controls input processing modules.
+jdmainct.c	Main buffer controller (JPEG decompressor => postprocessor).
+jdcoefct.c	Buffer controller for DCT coefficient buffer.
+jdpostct.c	Postprocessor buffer controller.
+jdmarker.c	JPEG marker reading.
+jdhuff.c	Huffman entropy decoding for sequential JPEG.
+jdphuff.c	Huffman entropy decoding for progressive JPEG.
+jddctmgr.c	IDCT manager (IDCT implementation selection & control).
+jidctint.c	Inverse DCT using slow-but-accurate integer method.
+jidctfst.c	Inverse DCT using faster, less accurate integer method.
+jidctflt.c	Inverse DCT using floating-point arithmetic.
+jidctred.c	Inverse DCTs with reduced-size outputs.
+jdsample.c	Upsampling.
+jdcolor.c	Color space conversion.
+jdmerge.c	Merged upsampling/color conversion (faster, lower quality).
+jquant1.c	One-pass color quantization using a fixed-spacing colormap.
+jquant2.c	Two-pass color quantization using a custom-generated colormap.
+		Also handles one-pass quantization to an externally given map.
+jdatasrc.c	Data source manager for stdio input.
+
+Support files for both compression and decompression:
+
+jerror.c	Standard error handling routines (application replaceable).
+jmemmgr.c	System-independent (more or less) memory management code.
+jutils.c	Miscellaneous utility routines.
+
+jmemmgr.c relies on a system-dependent memory management module.  The IJG
+distribution includes the following implementations of the system-dependent
+module:
+
+jmemnobs.c	"No backing store": assumes adequate virtual memory exists.
+jmemansi.c	Makes temporary files with ANSI-standard routine tmpfile().
+jmemname.c	Makes temporary files with program-generated file names.
+jmemdos.c	Custom implementation for MS-DOS (16-bit environment only):
+		can use extended and expanded memory as well as temp files.
+jmemmac.c	Custom implementation for Apple Macintosh.
+
+Exactly one of the system-dependent modules should be configured into an
+installed JPEG library (see install.doc for hints about which one to use).
+On unusual systems you may find it worthwhile to make a special
+system-dependent memory manager.
+
+
+Non-C source code files:
+
+jmemdosa.asm	80x86 assembly code support for jmemdos.c; used only in
+		MS-DOS-specific configurations of the JPEG library.
+
+
+CJPEG/DJPEG/JPEGTRAN
+====================
+
+Include files:
+
+cdjpeg.h	Declarations shared by cjpeg/djpeg/jpegtran modules.
+cderror.h	Additional error and trace message codes for cjpeg et al.
+transupp.h	Declarations for jpegtran support routines in transupp.c.
+
+C source code files:
+
+cjpeg.c		Main program for cjpeg.
+djpeg.c		Main program for djpeg.
+jpegtran.c	Main program for jpegtran.
+cdjpeg.c	Utility routines used by all three programs.
+rdcolmap.c	Code to read a colormap file for djpeg's "-map" switch.
+rdswitch.c	Code to process some of cjpeg's more complex switches.
+		Also used by jpegtran.
+transupp.c	Support code for jpegtran: lossless image manipulations.
+
+Image file reader modules for cjpeg:
+
+rdbmp.c		BMP file input.
+rdgif.c		GIF file input (now just a stub).
+rdppm.c		PPM/PGM file input.
+rdrle.c		Utah RLE file input.
+rdtarga.c	Targa file input.
+
+Image file writer modules for djpeg:
+
+wrbmp.c		BMP file output.
+wrgif.c		GIF file output (a mere shadow of its former self).
+wrppm.c		PPM/PGM file output.
+wrrle.c		Utah RLE file output.
+wrtarga.c	Targa file output.
+
+
+RDJPGCOM/WRJPGCOM
+=================
+
+C source code files:
+
+rdjpgcom.c	Stand-alone rdjpgcom application.
+wrjpgcom.c	Stand-alone wrjpgcom application.
+
+These programs do not depend on the IJG library.  They do use
+jconfig.h and jinclude.h, only to improve portability.
+
+
+ADDITIONAL FILES
+================
+
+Documentation (see README for a guide to the documentation files):
+
+README		Master documentation file.
+*.doc		Other documentation files.
+*.1		Documentation in Unix man page format.
+change.log	Version-to-version change highlights.
+example.c	Sample code for calling JPEG library.
+
+Configuration/installation files and programs (see install.doc for more info):
+
+configure	Unix shell script to perform automatic configuration.
+ltconfig	Support scripts for configure (from GNU libtool).
+ltmain.sh
+config.guess
+config.sub
+install-sh	Install shell script for those Unix systems lacking one.
+ckconfig.c	Program to generate jconfig.h on non-Unix systems.
+jconfig.doc	Template for making jconfig.h by hand.
+makefile.*	Sample makefiles for particular systems.
+jconfig.*	Sample jconfig.h for particular systems.
+ansi2knr.c	De-ANSIfier for pre-ANSI C compilers (courtesy of
+		L. Peter Deutsch and Aladdin Enterprises).
+
+Test files (see install.doc for test procedure):
+
+test*.*		Source and comparison files for confidence test.
+		These are binary image files, NOT text files.
diff --git a/libjpegtwrp/install-sh b/libjpegtwrp/install-sh
new file mode 100755
index 000000000..e8436696c
--- /dev/null
+++ b/libjpegtwrp/install-sh
@@ -0,0 +1,250 @@
+#!/bin/sh
+#
+# install - install a program, script, or datafile
+# This comes from X11R5 (mit/util/scripts/install.sh).
+#
+# Copyright 1991 by the Massachusetts Institute of Technology
+#
+# Permission to use, copy, modify, distribute, and sell this software and its
+# documentation for any purpose is hereby granted without fee, provided that
+# the above copyright notice appear in all copies and that both that
+# copyright notice and this permission notice appear in supporting
+# documentation, and that the name of M.I.T. not be used in advertising or
+# publicity pertaining to distribution of the software without specific,
+# written prior permission.  M.I.T. makes no representations about the
+# suitability of this software for any purpose.  It is provided "as is"
+# without express or implied warranty.
+#
+# Calling this script install-sh is preferred over install.sh, to prevent
+# `make' implicit rules from creating a file called install from it
+# when there is no Makefile.
+#
+# This script is compatible with the BSD install script, but was written
+# from scratch.  It can only install one file at a time, a restriction
+# shared with many OS's install programs.
+
+
+# set DOITPROG to echo to test this script
+
+# Don't use :- since 4.3BSD and earlier shells don't like it.
+doit="${DOITPROG-}"
+
+
+# put in absolute paths if you don't have them in your path; or use env. vars.
+
+mvprog="${MVPROG-mv}"
+cpprog="${CPPROG-cp}"
+chmodprog="${CHMODPROG-chmod}"
+chownprog="${CHOWNPROG-chown}"
+chgrpprog="${CHGRPPROG-chgrp}"
+stripprog="${STRIPPROG-strip}"
+rmprog="${RMPROG-rm}"
+mkdirprog="${MKDIRPROG-mkdir}"
+
+transformbasename=""
+transform_arg=""
+instcmd="$mvprog"
+chmodcmd="$chmodprog 0755"
+chowncmd=""
+chgrpcmd=""
+stripcmd=""
+rmcmd="$rmprog -f"
+mvcmd="$mvprog"
+src=""
+dst=""
+dir_arg=""
+
+while [ x"$1" != x ]; do
+    case $1 in
+	-c) instcmd="$cpprog"
+	    shift
+	    continue;;
+
+	-d) dir_arg=true
+	    shift
+	    continue;;
+
+	-m) chmodcmd="$chmodprog $2"
+	    shift
+	    shift
+	    continue;;
+
+	-o) chowncmd="$chownprog $2"
+	    shift
+	    shift
+	    continue;;
+
+	-g) chgrpcmd="$chgrpprog $2"
+	    shift
+	    shift
+	    continue;;
+
+	-s) stripcmd="$stripprog"
+	    shift
+	    continue;;
+
+	-t=*) transformarg=`echo $1 | sed 's/-t=//'`
+	    shift
+	    continue;;
+
+	-b=*) transformbasename=`echo $1 | sed 's/-b=//'`
+	    shift
+	    continue;;
+
+	*)  if [ x"$src" = x ]
+	    then
+		src=$1
+	    else
+		# this colon is to work around a 386BSD /bin/sh bug
+		:
+		dst=$1
+	    fi
+	    shift
+	    continue;;
+    esac
+done
+
+if [ x"$src" = x ]
+then
+	echo "install:	no input file specified"
+	exit 1
+else
+	true
+fi
+
+if [ x"$dir_arg" != x ]; then
+	dst=$src
+	src=""
+	
+	if [ -d $dst ]; then
+		instcmd=:
+	else
+		instcmd=mkdir
+	fi
+else
+
+# Waiting for this to be detected by the "$instcmd $src $dsttmp" command
+# might cause directories to be created, which would be especially bad 
+# if $src (and thus $dsttmp) contains '*'.
+
+	if [ -f $src -o -d $src ]
+	then
+		true
+	else
+		echo "install:  $src does not exist"
+		exit 1
+	fi
+	
+	if [ x"$dst" = x ]
+	then
+		echo "install:	no destination specified"
+		exit 1
+	else
+		true
+	fi
+
+# If destination is a directory, append the input filename; if your system
+# does not like double slashes in filenames, you may need to add some logic
+
+	if [ -d $dst ]
+	then
+		dst="$dst"/`basename $src`
+	else
+		true
+	fi
+fi
+
+## this sed command emulates the dirname command
+dstdir=`echo $dst | sed -e 's,[^/]*$,,;s,/$,,;s,^$,.,'`
+
+# Make sure that the destination directory exists.
+#  this part is taken from Noah Friedman's mkinstalldirs script
+
+# Skip lots of stat calls in the usual case.
+if [ ! -d "$dstdir" ]; then
+defaultIFS='	
+'
+IFS="${IFS-${defaultIFS}}"
+
+oIFS="${IFS}"
+# Some sh's can't handle IFS=/ for some reason.
+IFS='%'
+set - `echo ${dstdir} | sed -e 's@/@%@g' -e 's@^%@/@'`
+IFS="${oIFS}"
+
+pathcomp=''
+
+while [ $# -ne 0 ] ; do
+	pathcomp="${pathcomp}${1}"
+	shift
+
+	if [ ! -d "${pathcomp}" ] ;
+        then
+		$mkdirprog "${pathcomp}"
+	else
+		true
+	fi
+
+	pathcomp="${pathcomp}/"
+done
+fi
+
+if [ x"$dir_arg" != x ]
+then
+	$doit $instcmd $dst &&
+
+	if [ x"$chowncmd" != x ]; then $doit $chowncmd $dst; else true ; fi &&
+	if [ x"$chgrpcmd" != x ]; then $doit $chgrpcmd $dst; else true ; fi &&
+	if [ x"$stripcmd" != x ]; then $doit $stripcmd $dst; else true ; fi &&
+	if [ x"$chmodcmd" != x ]; then $doit $chmodcmd $dst; else true ; fi
+else
+
+# If we're going to rename the final executable, determine the name now.
+
+	if [ x"$transformarg" = x ] 
+	then
+		dstfile=`basename $dst`
+	else
+		dstfile=`basename $dst $transformbasename | 
+			sed $transformarg`$transformbasename
+	fi
+
+# don't allow the sed command to completely eliminate the filename
+
+	if [ x"$dstfile" = x ] 
+	then
+		dstfile=`basename $dst`
+	else
+		true
+	fi
+
+# Make a temp file name in the proper directory.
+
+	dsttmp=$dstdir/#inst.$$#
+
+# Move or copy the file name to the temp name
+
+	$doit $instcmd $src $dsttmp &&
+
+	trap "rm -f ${dsttmp}" 0 &&
+
+# and set any options; do chmod last to preserve setuid bits
+
+# If any of these fail, we abort the whole thing.  If we want to
+# ignore errors from any of these, just make sure not to ignore
+# errors from the above "$doit $instcmd $src $dsttmp" command.
+
+	if [ x"$chowncmd" != x ]; then $doit $chowncmd $dsttmp; else true;fi &&
+	if [ x"$chgrpcmd" != x ]; then $doit $chgrpcmd $dsttmp; else true;fi &&
+	if [ x"$stripcmd" != x ]; then $doit $stripcmd $dsttmp; else true;fi &&
+	if [ x"$chmodcmd" != x ]; then $doit $chmodcmd $dsttmp; else true;fi &&
+
+# Now rename the file to the real destination.
+
+	$doit $rmcmd -f $dstdir/$dstfile &&
+	$doit $mvcmd $dsttmp $dstdir/$dstfile 
+
+fi &&
+
+
+exit 0
diff --git a/libjpegtwrp/install.doc b/libjpegtwrp/install.doc
new file mode 100644
index 000000000..3702b986b
--- /dev/null
+++ b/libjpegtwrp/install.doc
@@ -0,0 +1,1063 @@
+INSTALLATION INSTRUCTIONS for the Independent JPEG Group's JPEG software
+
+Copyright (C) 1991-1998, Thomas G. Lane.
+This file is part of the Independent JPEG Group's software.
+For conditions of distribution and use, see the accompanying README file.
+
+
+This file explains how to configure and install the IJG software.  We have
+tried to make this software extremely portable and flexible, so that it can be
+adapted to almost any environment.  The downside of this decision is that the
+installation process is complicated.  We have provided shortcuts to simplify
+the task on common systems.  But in any case, you will need at least a little
+familiarity with C programming and program build procedures for your system.
+
+If you are only using this software as part of a larger program, the larger
+program's installation procedure may take care of configuring the IJG code.
+For example, Ghostscript's installation script will configure the IJG code.
+You don't need to read this file if you just want to compile Ghostscript.
+
+If you are on a Unix machine, you may not need to read this file at all.
+Try doing
+	./configure
+	make
+	make test
+If that doesn't complain, do
+	make install
+(better do "make -n install" first to see if the makefile will put the files
+where you want them).  Read further if you run into snags or want to customize
+the code for your system.
+
+
+TABLE OF CONTENTS
+-----------------
+
+Before you start
+Configuring the software:
+	using the automatic "configure" script
+	using one of the supplied jconfig and makefile files
+	by hand
+Building the software
+Testing the software
+Installing the software
+Optional stuff
+Optimization
+Hints for specific systems
+
+
+BEFORE YOU START
+================
+
+Before installing the software you must unpack the distributed source code.
+Since you are reading this file, you have probably already succeeded in this
+task.  However, there is a potential for error if you needed to convert the
+files to the local standard text file format (for example, if you are on
+MS-DOS you may have converted LF end-of-line to CR/LF).  You must apply
+such conversion to all the files EXCEPT those whose names begin with "test".
+The test files contain binary data; if you change them in any way then the
+self-test will give bad results.
+
+Please check the last section of this file to see if there are hints for the
+specific machine or compiler you are using.
+
+
+CONFIGURING THE SOFTWARE
+========================
+
+To configure the IJG code for your system, you need to create two files:
+  * jconfig.h: contains values for system-dependent #define symbols.
+  * Makefile: controls the compilation process.
+(On a non-Unix machine, you may create "project files" or some other
+substitute for a Makefile.  jconfig.h is needed in any environment.)
+
+We provide three different ways to generate these files:
+  * On a Unix system, you can just run the "configure" script.
+  * We provide sample jconfig files and makefiles for popular machines;
+    if your machine matches one of the samples, just copy the right sample
+    files to jconfig.h and Makefile.
+  * If all else fails, read the instructions below and make your own files.
+
+
+Configuring the software using the automatic "configure" script
+---------------------------------------------------------------
+
+If you are on a Unix machine, you can just type
+	./configure
+and let the configure script construct appropriate configuration files.
+If you're using "csh" on an old version of System V, you might need to type
+	sh configure
+instead to prevent csh from trying to execute configure itself.
+Expect configure to run for a few minutes, particularly on slower machines;
+it works by compiling a series of test programs.
+
+Configure was created with GNU Autoconf and it follows the usual conventions
+for GNU configure scripts.  It makes a few assumptions that you may want to
+override.  You can do this by providing optional switches to configure:
+
+* If you want to build libjpeg as a shared library, say
+	./configure --enable-shared
+To get both shared and static libraries, say
+	./configure --enable-shared --enable-static
+Note that these switches invoke GNU libtool to take care of system-dependent
+shared library building methods.  If things don't work this way, please try
+running configure without either switch; that should build a static library
+without using libtool.  If that works, your problem is probably with libtool
+not with the IJG code.  libtool is fairly new and doesn't support all flavors
+of Unix yet.  (You might be able to find a newer version of libtool than the
+one included with libjpeg; see ftp.gnu.org.  Report libtool problems to
+bug-libtool@gnu.org.)
+
+* Configure will use gcc (GNU C compiler) if it's available, otherwise cc.
+To force a particular compiler to be selected, use the CC option, for example
+	./configure CC='cc'
+The same method can be used to include any unusual compiler switches.
+For example, on HP-UX you probably want to say
+	./configure CC='cc -Aa'
+to get HP's compiler to run in ANSI mode.
+
+* The default CFLAGS setting is "-O" for non-gcc compilers, "-O2" for gcc.
+You can override this by saying, for example,
+	./configure CFLAGS='-g'
+if you want to compile with debugging support.
+
+* Configure will set up the makefile so that "make install" will install files
+into /usr/local/bin, /usr/local/man, etc.  You can specify an installation
+prefix other than "/usr/local" by giving configure the option "--prefix=PATH".
+
+* If you don't have a lot of swap space, you may need to enable the IJG
+software's internal virtual memory mechanism.  To do this, give the option
+"--enable-maxmem=N" where N is the default maxmemory limit in megabytes.
+This is discussed in more detail under "Selecting a memory manager", below.
+You probably don't need to worry about this on reasonably-sized Unix machines,
+unless you plan to process very large images.
+
+Configure has some other features that are useful if you are cross-compiling
+or working in a network of multiple machine types; but if you need those
+features, you probably already know how to use them.
+
+
+Configuring the software using one of the supplied jconfig and makefile files
+-----------------------------------------------------------------------------
+
+If you have one of these systems, you can just use the provided configuration
+files:
+
+Makefile	jconfig file	System and/or compiler
+
+makefile.manx	jconfig.manx	Amiga, Manx Aztec C
+makefile.sas	jconfig.sas	Amiga, SAS C
+makeproj.mac	jconfig.mac	Apple Macintosh, Metrowerks CodeWarrior
+mak*jpeg.st	jconfig.st	Atari ST/STE/TT, Pure C or Turbo C
+makefile.bcc	jconfig.bcc	MS-DOS or OS/2, Borland C
+makefile.dj	jconfig.dj	MS-DOS, DJGPP (Delorie's port of GNU C)
+makefile.mc6	jconfig.mc6	MS-DOS, Microsoft C (16-bit only)
+makefile.wat	jconfig.wat	MS-DOS, OS/2, or Windows NT, Watcom C
+makefile.vc	jconfig.vc	Windows NT/95, MS Visual C++
+make*.ds	jconfig.vc	Windows NT/95, MS Developer Studio
+makefile.mms	jconfig.vms	Digital VMS, with MMS software
+makefile.vms	jconfig.vms	Digital VMS, without MMS software
+
+Copy the proper jconfig file to jconfig.h and the makefile to Makefile (or
+whatever your system uses as the standard makefile name).  For more info see
+the appropriate system-specific hints section near the end of this file.
+
+
+Configuring the software by hand
+--------------------------------
+
+First, generate a jconfig.h file.  If you are moderately familiar with C,
+the comments in jconfig.doc should be enough information to do this; just
+copy jconfig.doc to jconfig.h and edit it appropriately.  Otherwise, you may
+prefer to use the ckconfig.c program.  You will need to compile and execute
+ckconfig.c by hand --- we hope you know at least enough to do that.
+ckconfig.c may not compile the first try (in fact, the whole idea is for it
+to fail if anything is going to).  If you get compile errors, fix them by
+editing ckconfig.c according to the directions given in ckconfig.c.  Once
+you get it to run, it will write a suitable jconfig.h file, and will also
+print out some advice about which makefile to use.
+
+You may also want to look at the canned jconfig files, if there is one for a
+system similar to yours.
+
+Second, select a makefile and copy it to Makefile (or whatever your system
+uses as the standard makefile name).  The most generic makefiles we provide
+are
+	makefile.ansi:	if your C compiler supports function prototypes
+	makefile.unix:	if not.
+(You have function prototypes if ckconfig.c put "#define HAVE_PROTOTYPES"
+in jconfig.h.)  You may want to start from one of the other makefiles if
+there is one for a system similar to yours.
+
+Look over the selected Makefile and adjust options as needed.  In particular
+you may want to change the CC and CFLAGS definitions.  For instance, if you
+are using GCC, set CC=gcc.  If you had to use any compiler switches to get
+ckconfig.c to work, make sure the same switches are in CFLAGS.
+
+If you are on a system that doesn't use makefiles, you'll need to set up
+project files (or whatever you do use) to compile all the source files and
+link them into executable files cjpeg, djpeg, jpegtran, rdjpgcom, and wrjpgcom.
+See the file lists in any of the makefiles to find out which files go into
+each program.  Note that the provided makefiles all make a "library" file
+libjpeg first, but you don't have to do that if you don't want to; the file
+lists identify which source files are actually needed for compression,
+decompression, or both.  As a last resort, you can make a batch script that
+just compiles everything and links it all together; makefile.vms is an example
+of this (it's for VMS systems that have no make-like utility).
+
+Here are comments about some specific configuration decisions you'll
+need to make:
+
+Command line style
+------------------
+
+These programs can use a Unix-like command line style which supports
+redirection and piping, like this:
+	cjpeg inputfile >outputfile
+	cjpeg <inputfile >outputfile
+	source program | cjpeg >outputfile
+The simpler "two file" command line style is just
+	cjpeg inputfile outputfile
+You may prefer the two-file style, particularly if you don't have pipes.
+
+You MUST use two-file style on any system that doesn't cope well with binary
+data fed through stdin/stdout; this is true for some MS-DOS compilers, for
+example.  If you're not on a Unix system, it's safest to assume you need
+two-file style.  (But if your compiler provides either the Posix-standard
+fdopen() library routine or a Microsoft-compatible setmode() routine, you
+can safely use the Unix command line style, by defining USE_FDOPEN or
+USE_SETMODE respectively.)
+
+To use the two-file style, make jconfig.h say "#define TWO_FILE_COMMANDLINE".
+
+Selecting a memory manager
+--------------------------
+
+The IJG code is capable of working on images that are too big to fit in main
+memory; data is swapped out to temporary files as necessary.  However, the
+code to do this is rather system-dependent.  We provide five different
+memory managers:
+
+* jmemansi.c	This version uses the ANSI-standard library routine tmpfile(),
+		which not all non-ANSI systems have.  On some systems
+		tmpfile() may put the temporary file in a non-optimal
+		location; if you don't like what it does, use jmemname.c.
+
+* jmemname.c	This version creates named temporary files.  For anything
+		except a Unix machine, you'll need to configure the
+		select_file_name() routine appropriately; see the comments
+		near the head of jmemname.c.  If you use this version, define
+		NEED_SIGNAL_CATCHER in jconfig.h to make sure the temp files
+		are removed if the program is aborted.
+
+* jmemnobs.c	(That stands for No Backing Store :-).)  This will compile on
+		almost any system, but it assumes you have enough main memory
+		or virtual memory to hold the biggest images you work with.
+
+* jmemdos.c	This should be used with most 16-bit MS-DOS compilers.
+		See the system-specific notes about MS-DOS for more info.
+		IMPORTANT: if you use this, define USE_MSDOS_MEMMGR in
+		jconfig.h, and include the assembly file jmemdosa.asm in the
+		programs.  The supplied makefiles and jconfig files for
+		16-bit MS-DOS compilers already do both.
+
+* jmemmac.c	Custom version for Apple Macintosh; see the system-specific
+		notes for Macintosh for more info.
+
+To use a particular memory manager, change the SYSDEPMEM variable in your
+makefile to equal the corresponding object file name (for example, jmemansi.o
+or jmemansi.obj for jmemansi.c).
+
+If you have plenty of (real or virtual) main memory, just use jmemnobs.c.
+"Plenty" means about ten bytes for every pixel in the largest images
+you plan to process, so a lot of systems don't meet this criterion.
+If yours doesn't, try jmemansi.c first.  If that doesn't compile, you'll have
+to use jmemname.c; be sure to adjust select_file_name() for local conditions.
+You may also need to change unlink() to remove() in close_backing_store().
+
+Except with jmemnobs.c or jmemmac.c, you need to adjust the DEFAULT_MAX_MEM
+setting to a reasonable value for your system (either by adding a #define for
+DEFAULT_MAX_MEM to jconfig.h, or by adding a -D switch to the Makefile).
+This value limits the amount of data space the program will attempt to
+allocate.  Code and static data space isn't counted, so the actual memory
+needs for cjpeg or djpeg are typically 100 to 150Kb more than the max-memory
+setting.  Larger max-memory settings reduce the amount of I/O needed to
+process a large image, but too large a value can result in "insufficient
+memory" failures.  On most Unix machines (and other systems with virtual
+memory), just set DEFAULT_MAX_MEM to several million and forget it.  At the
+other end of the spectrum, for MS-DOS machines you probably can't go much
+above 300K to 400K.  (On MS-DOS the value refers to conventional memory only.
+Extended/expanded memory is handled separately by jmemdos.c.)
+
+
+BUILDING THE SOFTWARE
+=====================
+
+Now you should be able to compile the software.  Just say "make" (or
+whatever's necessary to start the compilation).  Have a cup of coffee.
+
+Here are some things that could go wrong:
+
+If your compiler complains about undefined structures, you should be able to
+shut it up by putting "#define INCOMPLETE_TYPES_BROKEN" in jconfig.h.
+
+If you have trouble with missing system include files or inclusion of the
+wrong ones, read jinclude.h.  This shouldn't happen if you used configure
+or ckconfig.c to set up jconfig.h.
+
+There are a fair number of routines that do not use all of their parameters;
+some compilers will issue warnings about this, which you can ignore.  There
+are also a few configuration checks that may give "unreachable code" warnings.
+Any other warning deserves investigation.
+
+If you don't have a getenv() library routine, define NO_GETENV.
+
+Also see the system-specific hints, below.
+
+
+TESTING THE SOFTWARE
+====================
+
+As a quick test of functionality we've included a small sample image in
+several forms:
+	testorig.jpg	Starting point for the djpeg tests.
+	testimg.ppm	The output of djpeg testorig.jpg
+	testimg.bmp	The output of djpeg -bmp -colors 256 testorig.jpg
+	testimg.jpg	The output of cjpeg testimg.ppm
+	testprog.jpg	Progressive-mode equivalent of testorig.jpg.
+	testimgp.jpg	The output of cjpeg -progressive -optimize testimg.ppm
+(The first- and second-generation .jpg files aren't identical since JPEG is
+lossy.)  If you can generate duplicates of the testimg* files then you
+probably have working programs.
+
+With most of the makefiles, "make test" will perform the necessary
+comparisons.
+
+If you're using a makefile that doesn't provide the test option, run djpeg
+and cjpeg by hand and compare the output files to testimg* with whatever
+binary file comparison tool you have.  The files should be bit-for-bit
+identical.
+
+If the programs complain "MAX_ALLOC_CHUNK is wrong, please fix", then you
+need to reduce MAX_ALLOC_CHUNK to a value that fits in type size_t.
+Try adding "#define MAX_ALLOC_CHUNK 65520L" to jconfig.h.  A less likely
+configuration error is "ALIGN_TYPE is wrong, please fix": defining ALIGN_TYPE
+as long should take care of that one.
+
+If the cjpeg test run fails with "Missing Huffman code table entry", it's a
+good bet that you needed to define RIGHT_SHIFT_IS_UNSIGNED.  Go back to the
+configuration step and run ckconfig.c.  (This is a good plan for any other
+test failure, too.)
+
+If you are using Unix (one-file) command line style on a non-Unix system,
+it's a good idea to check that binary I/O through stdin/stdout actually
+works.  You should get the same results from "djpeg <testorig.jpg >out.ppm"
+as from "djpeg -outfile out.ppm testorig.jpg".  Note that the makefiles all
+use the latter style and therefore do not exercise stdin/stdout!  If this
+check fails, try recompiling with USE_SETMODE or USE_FDOPEN defined.
+If it still doesn't work, better use two-file style.
+
+If you chose a memory manager other than jmemnobs.c, you should test that
+temporary-file usage works.  Try "djpeg -bmp -colors 256 -max 0 testorig.jpg"
+and make sure its output matches testimg.bmp.  If you have any really large
+images handy, try compressing them with -optimize and/or decompressing with
+-colors 256 to make sure your DEFAULT_MAX_MEM setting is not too large.
+
+NOTE: this is far from an exhaustive test of the JPEG software; some modules,
+such as 1-pass color quantization, are not exercised at all.  It's just a
+quick test to give you some confidence that you haven't missed something
+major.
+
+
+INSTALLING THE SOFTWARE
+=======================
+
+Once you're done with the above steps, you can install the software by
+copying the executable files (cjpeg, djpeg, jpegtran, rdjpgcom, and wrjpgcom)
+to wherever you normally install programs.  On Unix systems, you'll also want
+to put the man pages (cjpeg.1, djpeg.1, jpegtran.1, rdjpgcom.1, wrjpgcom.1)
+in the man-page directory.  The pre-fab makefiles don't support this step
+since there's such a wide variety of installation procedures on different
+systems.
+
+If you generated a Makefile with the "configure" script, you can just say
+	make install
+to install the programs and their man pages into the standard places.
+(You'll probably need to be root to do this.)  We recommend first saying
+	make -n install
+to see where configure thought the files should go.  You may need to edit
+the Makefile, particularly if your system's conventions for man page
+filenames don't match what configure expects.
+
+If you want to install the IJG library itself, for use in compiling other
+programs besides ours, then you need to put the four include files
+	jpeglib.h jerror.h jconfig.h jmorecfg.h
+into your include-file directory, and put the library file libjpeg.a
+(extension may vary depending on system) wherever library files go.
+If you generated a Makefile with "configure", it will do what it thinks
+is the right thing if you say
+	make install-lib
+
+
+OPTIONAL STUFF
+==============
+
+Progress monitor:
+
+If you like, you can #define PROGRESS_REPORT (in jconfig.h) to enable display
+of percent-done progress reports.  The routine provided in cdjpeg.c merely
+prints percentages to stderr, but you can customize it to do something
+fancier.
+
+Utah RLE file format support:
+
+We distribute the software with support for RLE image files (Utah Raster
+Toolkit format) disabled, because the RLE support won't compile without the
+Utah library.  If you have URT version 3.1 or later, you can enable RLE
+support as follows:
+	1.  #define RLE_SUPPORTED in jconfig.h.
+	2.  Add a -I option to CFLAGS in the Makefile for the directory
+	    containing the URT .h files (typically the "include"
+	    subdirectory of the URT distribution).
+	3.  Add -L... -lrle to LDLIBS in the Makefile, where ... specifies
+	    the directory containing the URT "librle.a" file (typically the
+	    "lib" subdirectory of the URT distribution).
+
+Support for 12-bit-deep pixel data:
+
+The JPEG standard allows either 8-bit or 12-bit data precision.  (For color,
+this means 8 or 12 bits per channel, of course.)  If you need to work with
+deeper than 8-bit data, you can compile the IJG code for 12-bit operation.
+To do so:
+  1. In jmorecfg.h, define BITS_IN_JSAMPLE as 12 rather than 8.
+  2. In jconfig.h, undefine BMP_SUPPORTED, RLE_SUPPORTED, and TARGA_SUPPORTED,
+     because the code for those formats doesn't handle 12-bit data and won't
+     even compile.  (The PPM code does work, as explained below.  The GIF
+     code works too; it scales 8-bit GIF data to and from 12-bit depth
+     automatically.)
+  3. Compile.  Don't expect "make test" to pass, since the supplied test
+     files are for 8-bit data.
+
+Currently, 12-bit support does not work on 16-bit-int machines.
+
+Note that a 12-bit version will not read 8-bit JPEG files, nor vice versa;
+so you'll want to keep around a regular 8-bit compilation as well.
+(Run-time selection of data depth, to allow a single copy that does both,
+is possible but would probably slow things down considerably; it's very low
+on our to-do list.)
+
+The PPM reader (rdppm.c) can read 12-bit data from either text-format or
+binary-format PPM and PGM files.  Binary-format PPM/PGM files which have a
+maxval greater than 255 are assumed to use 2 bytes per sample, LSB first
+(little-endian order).  As of early 1995, 2-byte binary format is not
+officially supported by the PBMPLUS library, but it is expected that a
+future release of PBMPLUS will support it.  Note that the PPM reader will
+read files of any maxval regardless of the BITS_IN_JSAMPLE setting; incoming
+data is automatically rescaled to either maxval=255 or maxval=4095 as
+appropriate for the cjpeg bit depth.
+
+The PPM writer (wrppm.c) will normally write 2-byte binary PPM or PGM
+format, maxval 4095, when compiled with BITS_IN_JSAMPLE=12.  Since this
+format is not yet widely supported, you can disable it by compiling wrppm.c
+with PPM_NORAWWORD defined; then the data is scaled down to 8 bits to make a
+standard 1-byte/sample PPM or PGM file.  (Yes, this means still another copy
+of djpeg to keep around.  But hopefully you won't need it for very long.
+Poskanzer's supposed to get that new PBMPLUS release out Real Soon Now.)
+
+Of course, if you are working with 12-bit data, you probably have it stored
+in some other, nonstandard format.  In that case you'll probably want to
+write your own I/O modules to read and write your format.
+
+Note that a 12-bit version of cjpeg always runs in "-optimize" mode, in
+order to generate valid Huffman tables.  This is necessary because our
+default Huffman tables only cover 8-bit data.
+
+Removing code:
+
+If you need to make a smaller version of the JPEG software, some optional
+functions can be removed at compile time.  See the xxx_SUPPORTED #defines in
+jconfig.h and jmorecfg.h.  If at all possible, we recommend that you leave in
+decoder support for all valid JPEG files, to ensure that you can read anyone's
+output.  Taking out support for image file formats that you don't use is the
+most painless way to make the programs smaller.  Another possibility is to
+remove some of the DCT methods: in particular, the "IFAST" method may not be
+enough faster than the others to be worth keeping on your machine.  (If you
+do remove ISLOW or IFAST, be sure to redefine JDCT_DEFAULT or JDCT_FASTEST
+to a supported method, by adding a #define in jconfig.h.)
+
+
+OPTIMIZATION
+============
+
+Unless you own a Cray, you'll probably be interested in making the JPEG
+software go as fast as possible.  This section covers some machine-dependent
+optimizations you may want to try.  We suggest that before trying any of
+this, you first get the basic installation to pass the self-test step.
+Repeat the self-test after any optimization to make sure that you haven't
+broken anything.
+
+The integer DCT routines perform a lot of multiplications.  These
+multiplications must yield 32-bit results, but none of their input values
+are more than 16 bits wide.  On many machines, notably the 680x0 and 80x86
+CPUs, a 16x16=>32 bit multiply instruction is faster than a full 32x32=>32
+bit multiply.  Unfortunately there is no portable way to specify such a
+multiplication in C, but some compilers can generate one when you use the
+right combination of casts.  See the MULTIPLYxxx macro definitions in
+jdct.h.  If your compiler makes "int" be 32 bits and "short" be 16 bits,
+defining SHORTxSHORT_32 is fairly likely to work.  When experimenting with
+alternate definitions, be sure to test not only whether the code still works
+(use the self-test), but also whether it is actually faster --- on some
+compilers, alternate definitions may compute the right answer, yet be slower
+than the default.  Timing cjpeg on a large PGM (grayscale) input file is the
+best way to check this, as the DCT will be the largest fraction of the runtime
+in that mode.  (Note: some of the distributed compiler-specific jconfig files
+already contain #define switches to select appropriate MULTIPLYxxx
+definitions.)
+
+If your machine has sufficiently fast floating point hardware, you may find
+that the float DCT method is faster than the integer DCT methods, even
+after tweaking the integer multiply macros.  In that case you may want to
+make the float DCT be the default method.  (The only objection to this is
+that float DCT results may vary slightly across machines.)  To do that, add
+"#define JDCT_DEFAULT JDCT_FLOAT" to jconfig.h.  Even if you don't change
+the default, you should redefine JDCT_FASTEST, which is the method selected
+by djpeg's -fast switch.  Don't forget to update the documentation files
+(usage.doc and/or cjpeg.1, djpeg.1) to agree with what you've done.
+
+If access to "short" arrays is slow on your machine, it may be a win to
+define type JCOEF as int rather than short.  This will cost a good deal of
+memory though, particularly in some multi-pass modes, so don't do it unless
+you have memory to burn and short is REALLY slow.
+
+If your compiler can compile function calls in-line, make sure the INLINE
+macro in jmorecfg.h is defined as the keyword that marks a function
+inline-able.  Some compilers have a switch that tells the compiler to inline
+any function it thinks is profitable (e.g., -finline-functions for gcc).
+Enabling such a switch is likely to make the compiled code bigger but faster.
+
+In general, it's worth trying the maximum optimization level of your compiler,
+and experimenting with any optional optimizations such as loop unrolling.
+(Unfortunately, far too many compilers have optimizer bugs ... be prepared to
+back off if the code fails self-test.)  If you do any experimentation along
+these lines, please report the optimal settings to jpeg-info@uunet.uu.net so
+we can mention them in future releases.  Be sure to specify your machine and
+compiler version.
+
+
+HINTS FOR SPECIFIC SYSTEMS
+==========================
+
+We welcome reports on changes needed for systems not mentioned here.  Submit
+'em to jpeg-info@uunet.uu.net.  Also, if configure or ckconfig.c is wrong
+about how to configure the JPEG software for your system, please let us know.
+
+
+Acorn RISC OS:
+
+(Thanks to Simon Middleton for these hints on compiling with Desktop C.)
+After renaming the files according to Acorn conventions, take a copy of
+makefile.ansi, change all occurrences of 'libjpeg.a' to 'libjpeg.o' and
+change these definitions as indicated:
+
+CFLAGS= -throwback -IC: -Wn
+LDLIBS=C:o.Stubs
+SYSDEPMEM=jmemansi.o
+LN=Link
+AR=LibFile -c -o
+
+Also add a new line '.c.o:; $(cc) $< $(cflags) -c -o $@'.  Remove the
+lines '$(RM) libjpeg.o' and '$(AR2) libjpeg.o' and the 'jconfig.h'
+dependency section.
+
+Copy jconfig.doc to jconfig.h.  Edit jconfig.h to define TWO_FILE_COMMANDLINE
+and CHAR_IS_UNSIGNED.
+
+Run the makefile using !AMU not !Make.  If you want to use the 'clean' and
+'test' makefile entries then you will have to fiddle with the syntax a bit
+and rename the test files.
+
+
+Amiga:
+
+SAS C 6.50 reportedly is too buggy to compile the IJG code properly.
+A patch to update to 6.51 is available from SAS or AmiNet FTP sites.
+
+The supplied config files are set up to use jmemname.c as the memory
+manager, with temporary files being created on the device named by
+"JPEGTMP:".
+
+
+Atari ST/STE/TT:
+ 
+Copy the project files makcjpeg.st, makdjpeg.st, maktjpeg.st, and makljpeg.st
+to cjpeg.prj, djpeg.prj, jpegtran.prj, and libjpeg.prj respectively.  The
+project files should work as-is with Pure C.  For Turbo C, change library
+filenames "pc..." to "tc..." in each project file.  Note that libjpeg.prj
+selects jmemansi.c as the recommended memory manager.  You'll probably want to
+adjust the DEFAULT_MAX_MEM setting --- you want it to be a couple hundred K
+less than your normal free memory.  Put "#define DEFAULT_MAX_MEM nnnn" into
+jconfig.h to do this.
+
+To use the 68881/68882 coprocessor for the floating point DCT, add the
+compiler option "-8" to the project files and replace pcfltlib.lib with
+pc881lib.lib in cjpeg.prj and djpeg.prj.  Or if you don't have a
+coprocessor, you may prefer to remove the float DCT code by undefining
+DCT_FLOAT_SUPPORTED in jmorecfg.h (since without a coprocessor, the float
+code will be too slow to be useful).  In that case, you can delete
+pcfltlib.lib from the project files.
+
+Note that you must make libjpeg.lib before making cjpeg.ttp, djpeg.ttp,
+or jpegtran.ttp.  You'll have to perform the self-test by hand.
+
+We haven't bothered to include project files for rdjpgcom and wrjpgcom.
+Those source files should just be compiled by themselves; they don't
+depend on the JPEG library.
+
+There is a bug in some older versions of the Turbo C library which causes the
+space used by temporary files created with "tmpfile()" not to be freed after
+an abnormal program exit.  If you check your disk afterwards, you will find
+cluster chains that are allocated but not used by a file.  This should not
+happen in cjpeg/djpeg/jpegtran, since we enable a signal catcher to explicitly
+close temp files before exiting.  But if you use the JPEG library with your
+own code, be sure to supply a signal catcher, or else use a different
+system-dependent memory manager.
+
+
+Cray:
+
+Should you be so fortunate as to be running JPEG on a Cray YMP, there is a
+compiler bug in old versions of Cray's Standard C (prior to 3.1).  If you
+still have an old compiler, you'll need to insert a line reading
+"#pragma novector" just before the loop	
+    for (i = 1; i <= (int) htbl->bits[l]; i++)
+      huffsize[p++] = (char) l;
+in fix_huff_tbl (in V5beta1, line 204 of jchuff.c and line 176 of jdhuff.c).
+[This bug may or may not still occur with the current IJG code, but it's
+probably a dead issue anyway...]
+
+
+HP-UX:
+
+If you have HP-UX 7.05 or later with the "software development" C compiler,
+you should run the compiler in ANSI mode.  If using the configure script,
+say
+	./configure CC='cc -Aa'
+(or -Ae if you prefer).  If configuring by hand, use makefile.ansi and add
+"-Aa" to the CFLAGS line in the makefile.
+
+If you have a pre-7.05 system, or if you are using the non-ANSI C compiler
+delivered with a minimum HP-UX system, then you must use makefile.unix
+(and do NOT add -Aa); or just run configure without the CC option.
+
+On HP 9000 series 800 machines, the HP C compiler is buggy in revisions prior
+to A.08.07.  If you get complaints about "not a typedef name", you'll have to
+use makefile.unix, or run configure without the CC option.
+
+
+Macintosh, generic comments:
+
+The supplied user-interface files (cjpeg.c, djpeg.c, etc) are set up to
+provide a Unix-style command line interface.  You can use this interface on
+the Mac by means of the ccommand() library routine provided by Metrowerks
+CodeWarrior or Think C.  This is only appropriate for testing the library,
+however; to make a user-friendly equivalent of cjpeg/djpeg you'd really want
+to develop a Mac-style user interface.  There isn't a complete example
+available at the moment, but there are some helpful starting points:
+1. Sam Bushell's free "To JPEG" applet provides drag-and-drop conversion to
+JPEG under System 7 and later.  This only illustrates how to use the
+compression half of the library, but it does a very nice job of that part.
+The CodeWarrior source code is available from http://www.pobox.com/~jsam.
+2. Jim Brunner prepared a Mac-style user interface for both compression and
+decompression.  Unfortunately, it hasn't been updated since IJG v4, and
+the library's API has changed considerably since then.  Still it may be of
+some help, particularly as a guide to compiling the IJG code under Think C.
+Jim's code is available from the Info-Mac archives, at sumex-aim.stanford.edu
+or mirrors thereof; see file /info-mac/dev/src/jpeg-convert-c.hqx.
+
+jmemmac.c is the recommended memory manager back end for Macintosh.  It uses
+NewPtr/DisposePtr instead of malloc/free, and has a Mac-specific
+implementation of jpeg_mem_available().  It also creates temporary files that
+follow Mac conventions.  (That part of the code relies on System-7-or-later OS
+functions.  See the comments in jmemmac.c if you need to run it on System 6.)
+NOTE that USE_MAC_MEMMGR must be defined in jconfig.h to use jmemmac.c.
+
+You can also use jmemnobs.c, if you don't care about handling images larger
+than available memory.  If you use any memory manager back end other than
+jmemmac.c, we recommend replacing "malloc" and "free" by "NewPtr" and
+"DisposePtr", because Mac C libraries often have peculiar implementations of
+malloc/free.  (For instance, free() may not return the freed space to the
+Mac Memory Manager.  This is undesirable for the IJG code because jmemmgr.c
+already clumps space requests.)
+
+
+Macintosh, Metrowerks CodeWarrior:
+
+The Unix-command-line-style interface can be used by defining USE_CCOMMAND.
+You'll also need to define TWO_FILE_COMMANDLINE to avoid stdin/stdout.
+This means that when using the cjpeg/djpeg programs, you'll have to type the
+input and output file names in the "Arguments" text-edit box, rather than
+using the file radio buttons.  (Perhaps USE_FDOPEN or USE_SETMODE would
+eliminate the problem, but I haven't heard from anyone who's tried it.)
+
+On 680x0 Macs, Metrowerks defines type "double" as a 10-byte IEEE extended
+float.  jmemmgr.c won't like this: it wants sizeof(ALIGN_TYPE) to be a power
+of 2.  Add "#define ALIGN_TYPE long" to jconfig.h to eliminate the complaint.
+
+The supplied configuration file jconfig.mac can be used for your jconfig.h;
+it includes all the recommended symbol definitions.  If you have AppleScript
+installed, you can run the supplied script makeproj.mac to create CodeWarrior
+project files for the library and the testbed applications, then build the
+library and applications.  (Thanks to Dan Sears and Don Agro for this nifty
+hack, which saves us from trying to maintain CodeWarrior project files as part
+of the IJG distribution...)
+
+
+Macintosh, Think C:
+
+The documentation in Jim Brunner's "JPEG Convert" source code (see above)
+includes detailed build instructions for Think C; it's probably somewhat
+out of date for the current release, but may be helpful.
+
+If you want to build the minimal command line version, proceed as follows.
+You'll have to prepare project files for the programs; we don't include any
+in the distribution since they are not text files.  Use the file lists in
+any of the supplied makefiles as a guide.  Also add the ANSI and Unix C
+libraries in a separate segment.  You may need to divide the JPEG files into
+more than one segment; we recommend dividing compression and decompression
+modules.  Define USE_CCOMMAND in jconfig.h so that the ccommand() routine is
+called.  You must also define TWO_FILE_COMMANDLINE because stdin/stdout
+don't handle binary data correctly.
+
+On 680x0 Macs, Think C defines type "double" as a 12-byte IEEE extended float.
+jmemmgr.c won't like this: it wants sizeof(ALIGN_TYPE) to be a power of 2.
+Add "#define ALIGN_TYPE long" to jconfig.h to eliminate the complaint.
+
+jconfig.mac should work as a jconfig.h configuration file for Think C,
+but the makeproj.mac AppleScript script is specific to CodeWarrior.  Sorry.
+
+
+MIPS R3000:
+
+MIPS's cc version 1.31 has a rather nasty optimization bug.  Don't use -O
+if you have that compiler version.  (Use "cc -V" to check the version.)
+Note that the R3000 chip is found in workstations from DEC and others.
+
+
+MS-DOS, generic comments for 16-bit compilers:
+
+The IJG code is designed to work well in 80x86 "small" or "medium" memory
+models (i.e., data pointers are 16 bits unless explicitly declared "far";
+code pointers can be either size).  You may be able to use small model to
+compile cjpeg or djpeg by itself, but you will probably have to use medium
+model for any larger application.  This won't make much difference in
+performance.  You *will* take a noticeable performance hit if you use a
+large-data memory model, and you should avoid "huge" model if at all
+possible.  Be sure that NEED_FAR_POINTERS is defined in jconfig.h if you use
+a small-data memory model; be sure it is NOT defined if you use a large-data
+model.  (The supplied makefiles and jconfig files for Borland and Microsoft C
+compile in medium model and define NEED_FAR_POINTERS.)
+
+The DOS-specific memory manager, jmemdos.c, should be used if possible.
+It needs some assembly-code routines which are in jmemdosa.asm; make sure
+your makefile assembles that file and includes it in the library.  If you
+don't have a suitable assembler, you can get pre-assembled object files for
+jmemdosa by FTP from ftp.uu.net:/graphics/jpeg/jdosaobj.zip.  (DOS-oriented
+distributions of the IJG source code often include these object files.)
+
+When using jmemdos.c, jconfig.h must define USE_MSDOS_MEMMGR and must set
+MAX_ALLOC_CHUNK to less than 64K (65520L is a typical value).  If your
+C library's far-heap malloc() can't allocate blocks that large, reduce
+MAX_ALLOC_CHUNK to whatever it can handle.
+
+If you can't use jmemdos.c for some reason --- for example, because you
+don't have an assembler to assemble jmemdosa.asm --- you'll have to fall
+back to jmemansi.c or jmemname.c.  You'll probably still need to set
+MAX_ALLOC_CHUNK in jconfig.h, because most DOS C libraries won't malloc()
+more than 64K at a time.  IMPORTANT: if you use jmemansi.c or jmemname.c,
+you will have to compile in a large-data memory model in order to get the
+right stdio library.  Too bad.
+
+wrjpgcom needs to be compiled in large model, because it malloc()s a 64KB
+work area to hold the comment text.  If your C library's malloc can't
+handle that, reduce MAX_COM_LENGTH as necessary in wrjpgcom.c.
+
+Most MS-DOS compilers treat stdin/stdout as text files, so you must use
+two-file command line style.  But if your compiler has either fdopen() or
+setmode(), you can use one-file style if you like.  To do this, define
+USE_SETMODE or USE_FDOPEN so that stdin/stdout will be set to binary mode.
+(USE_SETMODE seems to work with more DOS compilers than USE_FDOPEN.)  You
+should test that I/O through stdin/stdout produces the same results as I/O
+to explicitly named files... the "make test" procedures in the supplied
+makefiles do NOT use stdin/stdout.
+
+
+MS-DOS, generic comments for 32-bit compilers:
+
+None of the above comments about memory models apply if you are using a
+32-bit flat-memory-space environment, such as DJGPP or Watcom C.  (And you
+should use one if you have it, as performance will be much better than
+8086-compatible code!)  For flat-memory-space compilers, do NOT define
+NEED_FAR_POINTERS, and do NOT use jmemdos.c.  Use jmemnobs.c if the
+environment supplies adequate virtual memory, otherwise use jmemansi.c or
+jmemname.c.
+
+You'll still need to be careful about binary I/O through stdin/stdout.
+See the last paragraph of the previous section.
+
+
+MS-DOS, Borland C:
+
+Be sure to convert all the source files to DOS text format (CR/LF newlines).
+Although Borland C will often work OK with unmodified Unix (LF newlines)
+source files, sometimes it will give bogus compile errors.
+"Illegal character '#'" is the most common such error.  (This is true with
+Borland C 3.1, but perhaps is fixed in newer releases.)
+
+If you want one-file command line style, just undefine TWO_FILE_COMMANDLINE.
+jconfig.bcc already includes #define USE_SETMODE to make this work.
+(fdopen does not work correctly.)
+
+
+MS-DOS, Microsoft C:
+
+makefile.mc6 works with Microsoft C, DOS Visual C++, etc.  It should only
+be used if you want to build a 16-bit (small or medium memory model) program.
+
+If you want one-file command line style, just undefine TWO_FILE_COMMANDLINE.
+jconfig.mc6 already includes #define USE_SETMODE to make this work.
+(fdopen does not work correctly.)
+
+Note that this makefile assumes that the working copy of itself is called
+"makefile".  If you want to call it something else, say "makefile.mak",
+be sure to adjust the dependency line that reads "$(RFILE) : makefile".
+Otherwise the make will fail because it doesn't know how to create "makefile".
+Worse, some releases of Microsoft's make utilities give an incorrect error
+message in this situation.
+
+Old versions of MS C fail with an "out of macro expansion space" error
+because they can't cope with the macro TRACEMS8 (defined in jerror.h).
+If this happens to you, the easiest solution is to change TRACEMS8 to
+expand to nothing.  You'll lose the ability to dump out JPEG coefficient
+tables with djpeg -debug -debug, but at least you can compile.
+
+Original MS C 6.0 is very buggy; it compiles incorrect code unless you turn
+off optimization entirely (remove -O from CFLAGS).  6.00A is better, but it
+still generates bad code if you enable loop optimizations (-Ol or -Ox).
+
+MS C 8.0 crashes when compiling jquant1.c with optimization switch /Oo ...
+which is on by default.  To work around this bug, compile that one file
+with /Oo-.
+
+
+Microsoft Windows (all versions), generic comments:
+
+Some Windows system include files define typedef boolean as "unsigned char".
+The IJG code also defines typedef boolean, but we make it "int" by default.
+This doesn't affect the IJG programs because we don't import those Windows
+include files.  But if you use the JPEG library in your own program, and some
+of your program's files import one definition of boolean while some import the
+other, you can get all sorts of mysterious problems.  A good preventive step
+is to make the IJG library use "unsigned char" for boolean.  To do that,
+add something like this to your jconfig.h file:
+	/* Define "boolean" as unsigned char, not int, per Windows custom */
+	#ifndef __RPCNDR_H__	/* don't conflict if rpcndr.h already read */
+	typedef unsigned char boolean;
+	#endif
+	#define HAVE_BOOLEAN	/* prevent jmorecfg.h from redefining it */
+(This is already in jconfig.vc, by the way.)
+
+windef.h contains the declarations
+	#define far
+	#define FAR far
+Since jmorecfg.h tries to define FAR as empty, you may get a compiler
+warning if you include both jpeglib.h and windef.h (which windows.h
+includes).  To suppress the warning, you can put "#ifndef FAR"/"#endif"
+around the line "#define FAR" in jmorecfg.h.
+
+When using the library in a Windows application, you will almost certainly
+want to modify or replace the error handler module jerror.c, since our
+default error handler does a couple of inappropriate things:
+  1. it tries to write error and warning messages on stderr;
+  2. in event of a fatal error, it exits by calling exit().
+
+A simple stopgap solution for problem 1 is to replace the line
+	fprintf(stderr, "%s\n", buffer);
+(in output_message in jerror.c) with
+	MessageBox(GetActiveWindow(),buffer,"JPEG Error",MB_OK|MB_ICONERROR);
+It's highly recommended that you at least do that much, since otherwise
+error messages will disappear into nowhere.  (Beginning with IJG v6b, this
+code is already present in jerror.c; just define USE_WINDOWS_MESSAGEBOX in
+jconfig.h to enable it.)
+
+The proper solution for problem 2 is to return control to your calling
+application after a library error.  This can be done with the setjmp/longjmp
+technique discussed in libjpeg.doc and illustrated in example.c.  (NOTE:
+some older Windows C compilers provide versions of setjmp/longjmp that
+don't actually work under Windows.  You may need to use the Windows system
+functions Catch and Throw instead.)
+
+The recommended memory manager under Windows is jmemnobs.c; in other words,
+let Windows do any virtual memory management needed.  You should NOT use
+jmemdos.c nor jmemdosa.asm under Windows.
+
+For Windows 3.1, we recommend compiling in medium or large memory model;
+for newer Windows versions, use a 32-bit flat memory model.  (See the MS-DOS
+sections above for more info about memory models.)  In the 16-bit memory
+models only, you'll need to put
+	#define MAX_ALLOC_CHUNK 65520L	/* Maximum request to malloc() */
+into jconfig.h to limit allocation chunks to 64Kb.  (Without that, you'd
+have to use huge memory model, which slows things down unnecessarily.)
+jmemnobs.c works without modification in large or flat memory models, but to
+use medium model, you need to modify its jpeg_get_large and jpeg_free_large
+routines to allocate far memory.  In any case, you might like to replace
+its calls to malloc and free with direct calls on Windows memory allocation
+functions.
+
+You may also want to modify jdatasrc.c and jdatadst.c to use Windows file
+operations rather than fread/fwrite.  This is only necessary if your C
+compiler doesn't provide a competent implementation of C stdio functions.
+
+You might want to tweak the RGB_xxx macros in jmorecfg.h so that the library
+will accept or deliver color pixels in BGR sample order, not RGB; BGR order
+is usually more convenient under Windows.  Note that this change will break
+the sample applications cjpeg/djpeg, but the library itself works fine.
+
+
+Many people want to convert the IJG library into a DLL.  This is reasonably
+straightforward, but watch out for the following:
+
+  1. Don't try to compile as a DLL in small or medium memory model; use
+large model, or even better, 32-bit flat model.  Many places in the IJG code
+assume the address of a local variable is an ordinary (not FAR) pointer;
+that isn't true in a medium-model DLL.
+
+  2. Microsoft C cannot pass file pointers between applications and DLLs.
+(See Microsoft Knowledge Base, PSS ID Number Q50336.)  So jdatasrc.c and
+jdatadst.c don't work if you open a file in your application and then pass
+the pointer to the DLL.  One workaround is to make jdatasrc.c/jdatadst.c
+part of your main application rather than part of the DLL.
+
+  3. You'll probably need to modify the macros GLOBAL() and EXTERN() to
+attach suitable linkage keywords to the exported routine names.  Similarly,
+you'll want to modify METHODDEF() and JMETHOD() to ensure function pointers
+are declared in a way that lets application routines be called back through
+the function pointers.  These macros are in jmorecfg.h.  Typical definitions
+for a 16-bit DLL are:
+	#define GLOBAL(type)		type _far _pascal _loadds _export
+	#define EXTERN(type)		extern type _far _pascal _loadds
+	#define METHODDEF(type)		static type _far _pascal
+	#define JMETHOD(type,methodname,arglist)  \
+		type (_far _pascal *methodname) arglist
+For a 32-bit DLL you may want something like
+	#define GLOBAL(type)		__declspec(dllexport) type
+	#define EXTERN(type)		extern __declspec(dllexport) type
+Although not all the GLOBAL routines are actually intended to be called by
+the application, the performance cost of making them all DLL entry points is
+negligible.
+
+The unmodified IJG library presents a very C-specific application interface,
+so the resulting DLL is only usable from C or C++ applications.  There has
+been some talk of writing wrapper code that would present a simpler interface
+usable from other languages, such as Visual Basic.  This is on our to-do list
+but hasn't been very high priority --- any volunteers out there?
+
+
+Microsoft Windows, Borland C:
+
+The provided jconfig.bcc should work OK in a 32-bit Windows environment,
+but you'll need to tweak it in a 16-bit environment (you'd need to define
+NEED_FAR_POINTERS and MAX_ALLOC_CHUNK).  Beware that makefile.bcc will need
+alteration if you want to use it for Windows --- in particular, you should
+use jmemnobs.c not jmemdos.c under Windows.
+
+Borland C++ 4.5 fails with an internal compiler error when trying to compile
+jdmerge.c in 32-bit mode.  If enough people complain, perhaps Borland will fix
+it.  In the meantime, the simplest known workaround is to add a redundant
+definition of the variable range_limit in h2v1_merged_upsample(), at the head
+of the block that handles odd image width (about line 268 in v6 jdmerge.c):
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    register JSAMPLE * range_limit = cinfo->sample_range_limit; /* ADD THIS */
+    cb = GETJSAMPLE(*inptr1);
+Pretty bizarre, especially since the very similar routine h2v2_merged_upsample
+doesn't trigger the bug.
+Recent reports suggest that this bug does not occur with "bcc32a" (the
+Pentium-optimized version of the compiler).
+
+Another report from a user of Borland C 4.5 was that incorrect code (leading
+to a color shift in processed images) was produced if any of the following
+optimization switch combinations were used: 
+	-Ot -Og
+	-Ot -Op
+	-Ot -Om
+So try backing off on optimization if you see such a problem.  (Are there
+several different releases all numbered "4.5"??)
+
+
+Microsoft Windows, Microsoft Visual C++:
+
+jconfig.vc should work OK with any Microsoft compiler for a 32-bit memory
+model.  makefile.vc is intended for command-line use.  (If you are using
+the Developer Studio environment, you may prefer the DevStudio project
+files; see below.)
+
+Some users feel that it's easier to call the library from C++ code if you
+force VC++ to treat the library as C++ code, which you can do by renaming
+all the *.c files to *.cpp (and adjusting the makefile to match).  This
+avoids the need to put extern "C" { ... } around #include "jpeglib.h" in
+your C++ application.
+
+
+Microsoft Windows, Microsoft Developer Studio:
+
+We include makefiles that should work as project files in DevStudio 4.2 or
+later.  There is a library makefile that builds the IJG library as a static
+Win32 library, and an application makefile that builds the sample applications
+as Win32 console applications.  (Even if you only want the library, we
+recommend building the applications so that you can run the self-test.)
+
+To use:
+1. Copy jconfig.vc to jconfig.h, makelib.ds to jpeg.mak, and
+   makeapps.ds to apps.mak.  (Note that the renaming is critical!)
+2. Click on the .mak files to construct project workspaces.
+   (If you are using DevStudio more recent than 4.2, you'll probably
+   get a message saying that the makefiles are being updated.)
+3. Build the library project, then the applications project.
+4. Move the application .exe files from `app`\Release to an
+   appropriate location on your path.
+5. To perform the self-test, execute the command line
+	NMAKE /f makefile.vc  test
+
+
+OS/2, Borland C++:
+
+Watch out for optimization bugs in older Borland compilers; you may need
+to back off the optimization switch settings.  See the comments in
+makefile.bcc.
+
+
+SGI:
+
+On some SGI systems, you may need to set "AR2= ar -ts" in the Makefile.
+If you are using configure, you can do this by saying
+	./configure RANLIB='ar -ts'
+This change is not needed on all SGIs.  Use it only if the make fails at the
+stage of linking the completed programs.
+
+On the MIPS R4000 architecture (Indy, etc.), the compiler option "-mips2"
+reportedly speeds up the float DCT method substantially, enough to make it
+faster than the default int method (but still slower than the fast int
+method).  If you use -mips2, you may want to alter the default DCT method to
+be float.  To do this, put "#define JDCT_DEFAULT JDCT_FLOAT" in jconfig.h.
+
+
+VMS:
+
+On an Alpha/VMS system with MMS, be sure to use the "/Marco=Alpha=1"
+qualifier with MMS when building the JPEG package.
+
+VAX/VMS v5.5-1 may have problems with the test step of the build procedure
+reporting differences when it compares the original and test images.  If the
+error points to the last block of the files, it is most likely bogus and may
+be safely ignored.  It seems to be because the files are Stream_LF and
+Backup/Compare has difficulty with the (presumably) null padded files.
+This problem was not observed on VAX/VMS v6.1 or AXP/VMS v6.1.
diff --git a/libjpegtwrp/jcapimin.c b/libjpegtwrp/jcapimin.c
new file mode 100644
index 000000000..54fb8c58c
--- /dev/null
+++ b/libjpegtwrp/jcapimin.c
@@ -0,0 +1,280 @@
+/*
+ * jcapimin.c
+ *
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains application interface code for the compression half
+ * of the JPEG library.  These are the "minimum" API routines that may be
+ * needed in either the normal full-compression case or the transcoding-only
+ * case.
+ *
+ * Most of the routines intended to be called directly by an application
+ * are in this file or in jcapistd.c.  But also see jcparam.c for
+ * parameter-setup helper routines, jcomapi.c for routines shared by
+ * compression and decompression, and jctrans.c for the transcoding case.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Initialization of a JPEG compression object.
+ * The error manager must already be set up (in case memory manager fails).
+ */
+
+GLOBAL(void)
+jpeg_CreateCompress (j_compress_ptr cinfo, int version, size_t structsize)
+{
+  int i;
+
+  /* Guard against version mismatches between library and caller. */
+  cinfo->mem = NULL;		/* so jpeg_destroy knows mem mgr not called */
+  if (version != JPEG_LIB_VERSION)
+    ERREXIT2(cinfo, JERR_BAD_LIB_VERSION, JPEG_LIB_VERSION, version);
+  if (structsize != SIZEOF(struct jpeg_compress_struct))
+    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE, 
+	     (int) SIZEOF(struct jpeg_compress_struct), (int) structsize);
+
+  /* For debugging purposes, we zero the whole master structure.
+   * But the application has already set the err pointer, and may have set
+   * client_data, so we have to save and restore those fields.
+   * Note: if application hasn't set client_data, tools like Purify may
+   * complain here.
+   */
+  {
+    struct jpeg_error_mgr * err = cinfo->err;
+    void * client_data = cinfo->client_data; /* ignore Purify complaint here */
+    MEMZERO(cinfo, SIZEOF(struct jpeg_compress_struct));
+    cinfo->err = err;
+    cinfo->client_data = client_data;
+  }
+  cinfo->is_decompressor = FALSE;
+
+  /* Initialize a memory manager instance for this object */
+  jinit_memory_mgr((j_common_ptr) cinfo);
+
+  /* Zero out pointers to permanent structures. */
+  cinfo->progress = NULL;
+  cinfo->dest = NULL;
+
+  cinfo->comp_info = NULL;
+
+  for (i = 0; i < NUM_QUANT_TBLS; i++)
+    cinfo->quant_tbl_ptrs[i] = NULL;
+
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    cinfo->dc_huff_tbl_ptrs[i] = NULL;
+    cinfo->ac_huff_tbl_ptrs[i] = NULL;
+  }
+
+  cinfo->script_space = NULL;
+
+  cinfo->input_gamma = 1.0;	/* in case application forgets */
+
+  /* OK, I'm ready */
+  cinfo->global_state = CSTATE_START;
+}
+
+
+/*
+ * Destruction of a JPEG compression object
+ */
+
+GLOBAL(void)
+jpeg_destroy_compress (j_compress_ptr cinfo)
+{
+  jpeg_destroy((j_common_ptr) cinfo); /* use common routine */
+}
+
+
+/*
+ * Abort processing of a JPEG compression operation,
+ * but don't destroy the object itself.
+ */
+
+GLOBAL(void)
+jpeg_abort_compress (j_compress_ptr cinfo)
+{
+  jpeg_abort((j_common_ptr) cinfo); /* use common routine */
+}
+
+
+/*
+ * Forcibly suppress or un-suppress all quantization and Huffman tables.
+ * Marks all currently defined tables as already written (if suppress)
+ * or not written (if !suppress).  This will control whether they get emitted
+ * by a subsequent jpeg_start_compress call.
+ *
+ * This routine is exported for use by applications that want to produce
+ * abbreviated JPEG datastreams.  It logically belongs in jcparam.c, but
+ * since it is called by jpeg_start_compress, we put it here --- otherwise
+ * jcparam.o would be linked whether the application used it or not.
+ */
+
+GLOBAL(void)
+jpeg_suppress_tables (j_compress_ptr cinfo, boolean suppress)
+{
+  int i;
+  JQUANT_TBL * qtbl;
+  JHUFF_TBL * htbl;
+
+  for (i = 0; i < NUM_QUANT_TBLS; i++) {
+    if ((qtbl = cinfo->quant_tbl_ptrs[i]) != NULL)
+      qtbl->sent_table = suppress;
+  }
+
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    if ((htbl = cinfo->dc_huff_tbl_ptrs[i]) != NULL)
+      htbl->sent_table = suppress;
+    if ((htbl = cinfo->ac_huff_tbl_ptrs[i]) != NULL)
+      htbl->sent_table = suppress;
+  }
+}
+
+
+/*
+ * Finish JPEG compression.
+ *
+ * If a multipass operating mode was selected, this may do a great deal of
+ * work including most of the actual output.
+ */
+
+GLOBAL(void)
+jpeg_finish_compress (j_compress_ptr cinfo)
+{
+  JDIMENSION iMCU_row;
+
+  if (cinfo->global_state == CSTATE_SCANNING ||
+      cinfo->global_state == CSTATE_RAW_OK) {
+    /* Terminate first pass */
+    if (cinfo->next_scanline < cinfo->image_height)
+      ERREXIT(cinfo, JERR_TOO_LITTLE_DATA);
+    (*cinfo->master->finish_pass) (cinfo);
+  } else if (cinfo->global_state != CSTATE_WRCOEFS)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Perform any remaining passes */
+  while (! cinfo->master->is_last_pass) {
+    (*cinfo->master->prepare_for_pass) (cinfo);
+    for (iMCU_row = 0; iMCU_row < cinfo->total_iMCU_rows; iMCU_row++) {
+      if (cinfo->progress != NULL) {
+	cinfo->progress->pass_counter = (long) iMCU_row;
+	cinfo->progress->pass_limit = (long) cinfo->total_iMCU_rows;
+	(*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+      }
+      /* We bypass the main controller and invoke coef controller directly;
+       * all work is being done from the coefficient buffer.
+       */
+      if (! (*cinfo->coef->compress_data) (cinfo, (JSAMPIMAGE) NULL))
+	ERREXIT(cinfo, JERR_CANT_SUSPEND);
+    }
+    (*cinfo->master->finish_pass) (cinfo);
+  }
+  /* Write EOI, do final cleanup */
+  (*cinfo->marker->write_file_trailer) (cinfo);
+  (*cinfo->dest->term_destination) (cinfo);
+  /* We can use jpeg_abort to release memory and reset global_state */
+  jpeg_abort((j_common_ptr) cinfo);
+}
+
+
+/*
+ * Write a special marker.
+ * This is only recommended for writing COM or APPn markers.
+ * Must be called after jpeg_start_compress() and before
+ * first call to jpeg_write_scanlines() or jpeg_write_raw_data().
+ */
+
+GLOBAL(void)
+jpeg_write_marker (j_compress_ptr cinfo, int marker,
+		   const JOCTET *dataptr, unsigned int datalen)
+{
+  JMETHOD(void, write_marker_byte, (j_compress_ptr info, int val));
+
+  if (cinfo->next_scanline != 0 ||
+      (cinfo->global_state != CSTATE_SCANNING &&
+       cinfo->global_state != CSTATE_RAW_OK &&
+       cinfo->global_state != CSTATE_WRCOEFS))
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  (*cinfo->marker->write_marker_header) (cinfo, marker, datalen);
+  write_marker_byte = cinfo->marker->write_marker_byte;	/* copy for speed */
+  while (datalen--) {
+    (*write_marker_byte) (cinfo, *dataptr);
+    dataptr++;
+  }
+}
+
+/* Same, but piecemeal. */
+
+GLOBAL(void)
+jpeg_write_m_header (j_compress_ptr cinfo, int marker, unsigned int datalen)
+{
+  if (cinfo->next_scanline != 0 ||
+      (cinfo->global_state != CSTATE_SCANNING &&
+       cinfo->global_state != CSTATE_RAW_OK &&
+       cinfo->global_state != CSTATE_WRCOEFS))
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  (*cinfo->marker->write_marker_header) (cinfo, marker, datalen);
+}
+
+GLOBAL(void)
+jpeg_write_m_byte (j_compress_ptr cinfo, int val)
+{
+  (*cinfo->marker->write_marker_byte) (cinfo, val);
+}
+
+
+/*
+ * Alternate compression function: just write an abbreviated table file.
+ * Before calling this, all parameters and a data destination must be set up.
+ *
+ * To produce a pair of files containing abbreviated tables and abbreviated
+ * image data, one would proceed as follows:
+ *
+ *		initialize JPEG object
+ *		set JPEG parameters
+ *		set destination to table file
+ *		jpeg_write_tables(cinfo);
+ *		set destination to image file
+ *		jpeg_start_compress(cinfo, FALSE);
+ *		write data...
+ *		jpeg_finish_compress(cinfo);
+ *
+ * jpeg_write_tables has the side effect of marking all tables written
+ * (same as jpeg_suppress_tables(..., TRUE)).  Thus a subsequent start_compress
+ * will not re-emit the tables unless it is passed write_all_tables=TRUE.
+ */
+
+GLOBAL(void)
+jpeg_write_tables (j_compress_ptr cinfo)
+{
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* (Re)initialize error mgr and destination modules */
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr) cinfo);
+  (*cinfo->dest->init_destination) (cinfo);
+  /* Initialize the marker writer ... bit of a crock to do it here. */
+  jinit_marker_writer(cinfo);
+  /* Write them tables! */
+  (*cinfo->marker->write_tables_only) (cinfo);
+  /* And clean up. */
+  (*cinfo->dest->term_destination) (cinfo);
+  /*
+   * In library releases up through v6a, we called jpeg_abort() here to free
+   * any working memory allocated by the destination manager and marker
+   * writer.  Some applications had a problem with that: they allocated space
+   * of their own from the library memory manager, and didn't want it to go
+   * away during write_tables.  So now we do nothing.  This will cause a
+   * memory leak if an app calls write_tables repeatedly without doing a full
+   * compression cycle or otherwise resetting the JPEG object.  However, that
+   * seems less bad than unexpectedly freeing memory in the normal case.
+   * An app that prefers the old behavior can call jpeg_abort for itself after
+   * each call to jpeg_write_tables().
+   */
+}
diff --git a/libjpegtwrp/jcapistd.c b/libjpegtwrp/jcapistd.c
new file mode 100644
index 000000000..c0320b1b1
--- /dev/null
+++ b/libjpegtwrp/jcapistd.c
@@ -0,0 +1,161 @@
+/*
+ * jcapistd.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains application interface code for the compression half
+ * of the JPEG library.  These are the "standard" API routines that are
+ * used in the normal full-compression case.  They are not used by a
+ * transcoding-only application.  Note that if an application links in
+ * jpeg_start_compress, it will end up linking in the entire compressor.
+ * We thus must separate this file from jcapimin.c to avoid linking the
+ * whole compression library into a transcoder.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Compression initialization.
+ * Before calling this, all parameters and a data destination must be set up.
+ *
+ * We require a write_all_tables parameter as a failsafe check when writing
+ * multiple datastreams from the same compression object.  Since prior runs
+ * will have left all the tables marked sent_table=TRUE, a subsequent run
+ * would emit an abbreviated stream (no tables) by default.  This may be what
+ * is wanted, but for safety's sake it should not be the default behavior:
+ * programmers should have to make a deliberate choice to emit abbreviated
+ * images.  Therefore the documentation and examples should encourage people
+ * to pass write_all_tables=TRUE; then it will take active thought to do the
+ * wrong thing.
+ */
+
+GLOBAL(void)
+jpeg_start_compress (j_compress_ptr cinfo, boolean write_all_tables)
+{
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  if (write_all_tables)
+    jpeg_suppress_tables(cinfo, FALSE);	/* mark all tables to be written */
+
+  /* (Re)initialize error mgr and destination modules */
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr) cinfo);
+  (*cinfo->dest->init_destination) (cinfo);
+  /* Perform master selection of active modules */
+  jinit_compress_master(cinfo);
+  /* Set up for the first pass */
+  (*cinfo->master->prepare_for_pass) (cinfo);
+  /* Ready for application to drive first pass through jpeg_write_scanlines
+   * or jpeg_write_raw_data.
+   */
+  cinfo->next_scanline = 0;
+  cinfo->global_state = (cinfo->raw_data_in ? CSTATE_RAW_OK : CSTATE_SCANNING);
+}
+
+
+/*
+ * Write some scanlines of data to the JPEG compressor.
+ *
+ * The return value will be the number of lines actually written.
+ * This should be less than the supplied num_lines only in case that
+ * the data destination module has requested suspension of the compressor,
+ * or if more than image_height scanlines are passed in.
+ *
+ * Note: we warn about excess calls to jpeg_write_scanlines() since
+ * this likely signals an application programmer error.  However,
+ * excess scanlines passed in the last valid call are *silently* ignored,
+ * so that the application need not adjust num_lines for end-of-image
+ * when using a multiple-scanline buffer.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_write_scanlines (j_compress_ptr cinfo, JSAMPARRAY scanlines,
+		      JDIMENSION num_lines)
+{
+  JDIMENSION row_ctr, rows_left;
+
+  if (cinfo->global_state != CSTATE_SCANNING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  if (cinfo->next_scanline >= cinfo->image_height)
+    WARNMS(cinfo, JWRN_TOO_MUCH_DATA);
+
+  /* Call progress monitor hook if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->pass_counter = (long) cinfo->next_scanline;
+    cinfo->progress->pass_limit = (long) cinfo->image_height;
+    (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+  }
+
+  /* Give master control module another chance if this is first call to
+   * jpeg_write_scanlines.  This lets output of the frame/scan headers be
+   * delayed so that application can write COM, etc, markers between
+   * jpeg_start_compress and jpeg_write_scanlines.
+   */
+  if (cinfo->master->call_pass_startup)
+    (*cinfo->master->pass_startup) (cinfo);
+
+  /* Ignore any extra scanlines at bottom of image. */
+  rows_left = cinfo->image_height - cinfo->next_scanline;
+  if (num_lines > rows_left)
+    num_lines = rows_left;
+
+  row_ctr = 0;
+  (*cinfo->main->process_data) (cinfo, scanlines, &row_ctr, num_lines);
+  cinfo->next_scanline += row_ctr;
+  return row_ctr;
+}
+
+
+/*
+ * Alternate entry point to write raw data.
+ * Processes exactly one iMCU row per call, unless suspended.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_write_raw_data (j_compress_ptr cinfo, JSAMPIMAGE data,
+		     JDIMENSION num_lines)
+{
+  JDIMENSION lines_per_iMCU_row;
+
+  if (cinfo->global_state != CSTATE_RAW_OK)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  if (cinfo->next_scanline >= cinfo->image_height) {
+    WARNMS(cinfo, JWRN_TOO_MUCH_DATA);
+    return 0;
+  }
+
+  /* Call progress monitor hook if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->pass_counter = (long) cinfo->next_scanline;
+    cinfo->progress->pass_limit = (long) cinfo->image_height;
+    (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+  }
+
+  /* Give master control module another chance if this is first call to
+   * jpeg_write_raw_data.  This lets output of the frame/scan headers be
+   * delayed so that application can write COM, etc, markers between
+   * jpeg_start_compress and jpeg_write_raw_data.
+   */
+  if (cinfo->master->call_pass_startup)
+    (*cinfo->master->pass_startup) (cinfo);
+
+  /* Verify that at least one iMCU row has been passed. */
+  lines_per_iMCU_row = cinfo->max_v_samp_factor * DCTSIZE;
+  if (num_lines < lines_per_iMCU_row)
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+
+  /* Directly compress the row. */
+  if (! (*cinfo->coef->compress_data) (cinfo, data)) {
+    /* If compressor did not consume the whole row, suspend processing. */
+    return 0;
+  }
+
+  /* OK, we processed one iMCU row. */
+  cinfo->next_scanline += lines_per_iMCU_row;
+  return lines_per_iMCU_row;
+}
diff --git a/libjpegtwrp/jccoefct.c b/libjpegtwrp/jccoefct.c
new file mode 100644
index 000000000..1963ddb61
--- /dev/null
+++ b/libjpegtwrp/jccoefct.c
@@ -0,0 +1,449 @@
+/*
+ * jccoefct.c
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the coefficient buffer controller for compression.
+ * This controller is the top level of the JPEG compressor proper.
+ * The coefficient buffer lies between forward-DCT and entropy encoding steps.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* We use a full-image coefficient buffer when doing Huffman optimization,
+ * and also for writing multiple-scan JPEG files.  In all cases, the DCT
+ * step is run during the first pass, and subsequent passes need only read
+ * the buffered coefficients.
+ */
+#ifdef ENTROPY_OPT_SUPPORTED
+#define FULL_COEF_BUFFER_SUPPORTED
+#else
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+#define FULL_COEF_BUFFER_SUPPORTED
+#endif
+#endif
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_c_coef_controller pub; /* public fields */
+
+  JDIMENSION iMCU_row_num;	/* iMCU row # within image */
+  JDIMENSION mcu_ctr;		/* counts MCUs processed in current row */
+  int MCU_vert_offset;		/* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;	/* number of such rows needed */
+
+  /* For single-pass compression, it's sufficient to buffer just one MCU
+   * (although this may prove a bit slow in practice).  We allocate a
+   * workspace of C_MAX_BLOCKS_IN_MCU coefficient blocks, and reuse it for each
+   * MCU constructed and sent.  (On 80x86, the workspace is FAR even though
+   * it's not really very big; this is to keep the module interfaces unchanged
+   * when a large coefficient buffer is necessary.)
+   * In multi-pass modes, this array points to the current MCU's blocks
+   * within the virtual arrays.
+   */
+  JBLOCKROW MCU_buffer[C_MAX_BLOCKS_IN_MCU];
+
+  /* In multi-pass modes, we need a virtual block array for each component. */
+  jvirt_barray_ptr whole_image[MAX_COMPONENTS];
+} my_coef_controller;
+
+typedef my_coef_controller * my_coef_ptr;
+
+
+/* Forward declarations */
+METHODDEF(boolean) compress_data
+    JPP((j_compress_ptr cinfo, JSAMPIMAGE input_buf));
+#ifdef FULL_COEF_BUFFER_SUPPORTED
+METHODDEF(boolean) compress_first_pass
+    JPP((j_compress_ptr cinfo, JSAMPIMAGE input_buf));
+METHODDEF(boolean) compress_output
+    JPP((j_compress_ptr cinfo, JSAMPIMAGE input_buf));
+#endif
+
+
+LOCAL(void)
+start_iMCU_row (j_compress_ptr cinfo)
+/* Reset within-iMCU-row counters for a new row */
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  /* In an interleaved scan, an MCU row is the same as an iMCU row.
+   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
+   * But at the bottom of the image, process only what's left.
+   */
+  if (cinfo->comps_in_scan > 1) {
+    coef->MCU_rows_per_iMCU_row = 1;
+  } else {
+    if (coef->iMCU_row_num < (cinfo->total_iMCU_rows-1))
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->v_samp_factor;
+    else
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->last_row_height;
+  }
+
+  coef->mcu_ctr = 0;
+  coef->MCU_vert_offset = 0;
+}
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_coef (j_compress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  coef->iMCU_row_num = 0;
+  start_iMCU_row(cinfo);
+
+  switch (pass_mode) {
+  case JBUF_PASS_THRU:
+    if (coef->whole_image[0] != NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    coef->pub.compress_data = compress_data;
+    break;
+#ifdef FULL_COEF_BUFFER_SUPPORTED
+  case JBUF_SAVE_AND_PASS:
+    if (coef->whole_image[0] == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    coef->pub.compress_data = compress_first_pass;
+    break;
+  case JBUF_CRANK_DEST:
+    if (coef->whole_image[0] == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    coef->pub.compress_data = compress_output;
+    break;
+#endif
+  default:
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    break;
+  }
+}
+
+
+/*
+ * Process some data in the single-pass case.
+ * We process the equivalent of one fully interleaved MCU row ("iMCU" row)
+ * per call, ie, v_samp_factor block rows for each component in the image.
+ * Returns TRUE if the iMCU row is completed, FALSE if suspended.
+ *
+ * NB: input_buf contains a plane for each component in image,
+ * which we index according to the component's SOF position.
+ */
+
+METHODDEF(boolean)
+compress_data (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  int blkn, bi, ci, yindex, yoffset, blockcnt;
+  JDIMENSION ypos, xpos;
+  jpeg_component_info *compptr;
+
+  /* Loop to write as much as one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->mcu_ctr; MCU_col_num <= last_MCU_col;
+	 MCU_col_num++) {
+      /* Determine where data comes from in input_buf and do the DCT thing.
+       * Each call on forward_DCT processes a horizontal row of DCT blocks
+       * as wide as an MCU; we rely on having allocated the MCU_buffer[] blocks
+       * sequentially.  Dummy blocks at the right or bottom edge are filled in
+       * specially.  The data in them does not matter for image reconstruction,
+       * so we fill them with values that will encode to the smallest amount of
+       * data, viz: all zeroes in the AC entries, DC entries equal to previous
+       * block's DC value.  (Thanks to Thomas Kinsman for this idea.)
+       */
+      blkn = 0;
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+	compptr = cinfo->cur_comp_info[ci];
+	blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
+						: compptr->last_col_width;
+	xpos = MCU_col_num * compptr->MCU_sample_width;
+	ypos = yoffset * DCTSIZE; /* ypos == (yoffset+yindex) * DCTSIZE */
+	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+	  if (coef->iMCU_row_num < last_iMCU_row ||
+	      yoffset+yindex < compptr->last_row_height) {
+	    (*cinfo->fdct->forward_DCT) (cinfo, compptr,
+					 input_buf[compptr->component_index],
+					 coef->MCU_buffer[blkn],
+					 ypos, xpos, (JDIMENSION) blockcnt);
+	    if (blockcnt < compptr->MCU_width) {
+	      /* Create some dummy blocks at the right edge of the image. */
+	      jzero_far((void FAR *) coef->MCU_buffer[blkn + blockcnt],
+			(compptr->MCU_width - blockcnt) * SIZEOF(JBLOCK));
+	      for (bi = blockcnt; bi < compptr->MCU_width; bi++) {
+		coef->MCU_buffer[blkn+bi][0][0] = coef->MCU_buffer[blkn+bi-1][0][0];
+	      }
+	    }
+	  } else {
+	    /* Create a row of dummy blocks at the bottom of the image. */
+	    jzero_far((void FAR *) coef->MCU_buffer[blkn],
+		      compptr->MCU_width * SIZEOF(JBLOCK));
+	    for (bi = 0; bi < compptr->MCU_width; bi++) {
+	      coef->MCU_buffer[blkn+bi][0][0] = coef->MCU_buffer[blkn-1][0][0];
+	    }
+	  }
+	  blkn += compptr->MCU_width;
+	  ypos += DCTSIZE;
+	}
+      }
+      /* Try to write the MCU.  In event of a suspension failure, we will
+       * re-DCT the MCU on restart (a bit inefficient, could be fixed...)
+       */
+      if (! (*cinfo->entropy->encode_mcu) (cinfo, coef->MCU_buffer)) {
+	/* Suspension forced; update state counters and exit */
+	coef->MCU_vert_offset = yoffset;
+	coef->mcu_ctr = MCU_col_num;
+	return FALSE;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->mcu_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  coef->iMCU_row_num++;
+  start_iMCU_row(cinfo);
+  return TRUE;
+}
+
+
+#ifdef FULL_COEF_BUFFER_SUPPORTED
+
+/*
+ * Process some data in the first pass of a multi-pass case.
+ * We process the equivalent of one fully interleaved MCU row ("iMCU" row)
+ * per call, ie, v_samp_factor block rows for each component in the image.
+ * This amount of data is read from the source buffer, DCT'd and quantized,
+ * and saved into the virtual arrays.  We also generate suitable dummy blocks
+ * as needed at the right and lower edges.  (The dummy blocks are constructed
+ * in the virtual arrays, which have been padded appropriately.)  This makes
+ * it possible for subsequent passes not to worry about real vs. dummy blocks.
+ *
+ * We must also emit the data to the entropy encoder.  This is conveniently
+ * done by calling compress_output() after we've loaded the current strip
+ * of the virtual arrays.
+ *
+ * NB: input_buf contains a plane for each component in image.  All
+ * components are DCT'd and loaded into the virtual arrays in this pass.
+ * However, it may be that only a subset of the components are emitted to
+ * the entropy encoder during this first pass; be careful about looking
+ * at the scan-dependent variables (MCU dimensions, etc).
+ */
+
+METHODDEF(boolean)
+compress_first_pass (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  JDIMENSION blocks_across, MCUs_across, MCUindex;
+  int bi, ci, h_samp_factor, block_row, block_rows, ndummy;
+  JCOEF lastDC;
+  jpeg_component_info *compptr;
+  JBLOCKARRAY buffer;
+  JBLOCKROW thisblockrow, lastblockrow;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Align the virtual buffer for this component. */
+    buffer = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr) cinfo, coef->whole_image[ci],
+       coef->iMCU_row_num * compptr->v_samp_factor,
+       (JDIMENSION) compptr->v_samp_factor, TRUE);
+    /* Count non-dummy DCT block rows in this iMCU row. */
+    if (coef->iMCU_row_num < last_iMCU_row)
+      block_rows = compptr->v_samp_factor;
+    else {
+      /* NB: can't use last_row_height here, since may not be set! */
+      block_rows = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
+      if (block_rows == 0) block_rows = compptr->v_samp_factor;
+    }
+    blocks_across = compptr->width_in_blocks;
+    h_samp_factor = compptr->h_samp_factor;
+    /* Count number of dummy blocks to be added at the right margin. */
+    ndummy = (int) (blocks_across % h_samp_factor);
+    if (ndummy > 0)
+      ndummy = h_samp_factor - ndummy;
+    /* Perform DCT for all non-dummy blocks in this iMCU row.  Each call
+     * on forward_DCT processes a complete horizontal row of DCT blocks.
+     */
+    for (block_row = 0; block_row < block_rows; block_row++) {
+      thisblockrow = buffer[block_row];
+      (*cinfo->fdct->forward_DCT) (cinfo, compptr,
+				   input_buf[ci], thisblockrow,
+				   (JDIMENSION) (block_row * DCTSIZE),
+				   (JDIMENSION) 0, blocks_across);
+      if (ndummy > 0) {
+	/* Create dummy blocks at the right edge of the image. */
+	thisblockrow += blocks_across; /* => first dummy block */
+	jzero_far((void FAR *) thisblockrow, ndummy * SIZEOF(JBLOCK));
+	lastDC = thisblockrow[-1][0];
+	for (bi = 0; bi < ndummy; bi++) {
+	  thisblockrow[bi][0] = lastDC;
+	}
+      }
+    }
+    /* If at end of image, create dummy block rows as needed.
+     * The tricky part here is that within each MCU, we want the DC values
+     * of the dummy blocks to match the last real block's DC value.
+     * This squeezes a few more bytes out of the resulting file...
+     */
+    if (coef->iMCU_row_num == last_iMCU_row) {
+      blocks_across += ndummy;	/* include lower right corner */
+      MCUs_across = blocks_across / h_samp_factor;
+      for (block_row = block_rows; block_row < compptr->v_samp_factor;
+	   block_row++) {
+	thisblockrow = buffer[block_row];
+	lastblockrow = buffer[block_row-1];
+	jzero_far((void FAR *) thisblockrow,
+		  (size_t) (blocks_across * SIZEOF(JBLOCK)));
+	for (MCUindex = 0; MCUindex < MCUs_across; MCUindex++) {
+	  lastDC = lastblockrow[h_samp_factor-1][0];
+	  for (bi = 0; bi < h_samp_factor; bi++) {
+	    thisblockrow[bi][0] = lastDC;
+	  }
+	  thisblockrow += h_samp_factor; /* advance to next MCU in row */
+	  lastblockrow += h_samp_factor;
+	}
+      }
+    }
+  }
+  /* NB: compress_output will increment iMCU_row_num if successful.
+   * A suspension return will result in redoing all the work above next time.
+   */
+
+  /* Emit data to the entropy encoder, sharing code with subsequent passes */
+  return compress_output(cinfo, input_buf);
+}
+
+
+/*
+ * Process some data in subsequent passes of a multi-pass case.
+ * We process the equivalent of one fully interleaved MCU row ("iMCU" row)
+ * per call, ie, v_samp_factor block rows for each component in the scan.
+ * The data is obtained from the virtual arrays and fed to the entropy coder.
+ * Returns TRUE if the iMCU row is completed, FALSE if suspended.
+ *
+ * NB: input_buf is ignored; it is likely to be a NULL pointer.
+ */
+
+METHODDEF(boolean)
+compress_output (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  int blkn, ci, xindex, yindex, yoffset;
+  JDIMENSION start_col;
+  JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN];
+  JBLOCKROW buffer_ptr;
+  jpeg_component_info *compptr;
+
+  /* Align the virtual buffers for the components used in this scan.
+   * NB: during first pass, this is safe only because the buffers will
+   * already be aligned properly, so jmemmgr.c won't need to do any I/O.
+   */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    buffer[ci] = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr) cinfo, coef->whole_image[compptr->component_index],
+       coef->iMCU_row_num * compptr->v_samp_factor,
+       (JDIMENSION) compptr->v_samp_factor, FALSE);
+  }
+
+  /* Loop to process one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->mcu_ctr; MCU_col_num < cinfo->MCUs_per_row;
+	 MCU_col_num++) {
+      /* Construct list of pointers to DCT blocks belonging to this MCU */
+      blkn = 0;			/* index of current DCT block within MCU */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+	compptr = cinfo->cur_comp_info[ci];
+	start_col = MCU_col_num * compptr->MCU_width;
+	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+	  buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+	  for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
+	    coef->MCU_buffer[blkn++] = buffer_ptr++;
+	  }
+	}
+      }
+      /* Try to write the MCU. */
+      if (! (*cinfo->entropy->encode_mcu) (cinfo, coef->MCU_buffer)) {
+	/* Suspension forced; update state counters and exit */
+	coef->MCU_vert_offset = yoffset;
+	coef->mcu_ctr = MCU_col_num;
+	return FALSE;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->mcu_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  coef->iMCU_row_num++;
+  start_iMCU_row(cinfo);
+  return TRUE;
+}
+
+#endif /* FULL_COEF_BUFFER_SUPPORTED */
+
+
+/*
+ * Initialize coefficient buffer controller.
+ */
+
+GLOBAL(void)
+jinit_c_coef_controller (j_compress_ptr cinfo, boolean need_full_buffer)
+{
+  my_coef_ptr coef;
+
+  coef = (my_coef_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_coef_controller));
+  cinfo->coef = (struct jpeg_c_coef_controller *) coef;
+  coef->pub.start_pass = start_pass_coef;
+
+  /* Create the coefficient buffer. */
+  if (need_full_buffer) {
+#ifdef FULL_COEF_BUFFER_SUPPORTED
+    /* Allocate a full-image virtual array for each component, */
+    /* padded to a multiple of samp_factor DCT blocks in each direction. */
+    int ci;
+    jpeg_component_info *compptr;
+
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      coef->whole_image[ci] = (*cinfo->mem->request_virt_barray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+	 (JDIMENSION) jround_up((long) compptr->width_in_blocks,
+				(long) compptr->h_samp_factor),
+	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
+				(long) compptr->v_samp_factor),
+	 (JDIMENSION) compptr->v_samp_factor);
+    }
+#else
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+#endif
+  } else {
+    /* We only need a single-MCU buffer. */
+    JBLOCKROW buffer;
+    int i;
+
+    buffer = (JBLOCKROW)
+      (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  C_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+    for (i = 0; i < C_MAX_BLOCKS_IN_MCU; i++) {
+      coef->MCU_buffer[i] = buffer + i;
+    }
+    coef->whole_image[0] = NULL; /* flag for no virtual arrays */
+  }
+}
diff --git a/libjpegtwrp/jccolor.c b/libjpegtwrp/jccolor.c
new file mode 100644
index 000000000..57a76c36b
--- /dev/null
+++ b/libjpegtwrp/jccolor.c
@@ -0,0 +1,527 @@
+/*
+ * jccolor.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains input colorspace conversion routines.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+// this enables unrolling null_convert's loop, and reading/write ints for speed
+#define ENABLE_ANDROID_NULL_CONVERT
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_color_converter pub; /* public fields */
+
+  /* Private state for RGB->YCC conversion */
+  INT32 * rgb_ycc_tab;		/* => table for RGB to YCbCr conversion */
+} my_color_converter;
+
+typedef my_color_converter * my_cconvert_ptr;
+
+
+/**************** RGB -> YCbCr conversion: most common case **************/
+
+/*
+ * YCbCr is defined per CCIR 601-1, except that Cb and Cr are
+ * normalized to the range 0..MAXJSAMPLE rather than -0.5 .. 0.5.
+ * The conversion equations to be implemented are therefore
+ *	Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+ *	Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + CENTERJSAMPLE
+ *	Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B  + CENTERJSAMPLE
+ * (These numbers are derived from TIFF 6.0 section 21, dated 3-June-92.)
+ * Note: older versions of the IJG code used a zero offset of MAXJSAMPLE/2,
+ * rather than CENTERJSAMPLE, for Cb and Cr.  This gave equal positive and
+ * negative swings for Cb/Cr, but meant that grayscale values (Cb=Cr=0)
+ * were not represented exactly.  Now we sacrifice exact representation of
+ * maximum red and maximum blue in order to get exact grayscales.
+ *
+ * To avoid floating-point arithmetic, we represent the fractional constants
+ * as integers scaled up by 2^16 (about 4 digits precision); we have to divide
+ * the products by 2^16, with appropriate rounding, to get the correct answer.
+ *
+ * For even more speed, we avoid doing any multiplications in the inner loop
+ * by precalculating the constants times R,G,B for all possible values.
+ * For 8-bit JSAMPLEs this is very reasonable (only 256 entries per table);
+ * for 12-bit samples it is still acceptable.  It's not very reasonable for
+ * 16-bit samples, but if you want lossless storage you shouldn't be changing
+ * colorspace anyway.
+ * The CENTERJSAMPLE offsets and the rounding fudge-factor of 0.5 are included
+ * in the tables to save adding them separately in the inner loop.
+ */
+
+#define SCALEBITS	16	/* speediest right-shift on some machines */
+#define CBCR_OFFSET	((INT32) CENTERJSAMPLE << SCALEBITS)
+#define ONE_HALF	((INT32) 1 << (SCALEBITS-1))
+#define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
+
+/* We allocate one big table and divide it up into eight parts, instead of
+ * doing eight alloc_small requests.  This lets us use a single table base
+ * address, which can be held in a register in the inner loops on many
+ * machines (more than can hold all eight addresses, anyway).
+ */
+
+#define R_Y_OFF		0			/* offset to R => Y section */
+#define G_Y_OFF		(1*(MAXJSAMPLE+1))	/* offset to G => Y section */
+#define B_Y_OFF		(2*(MAXJSAMPLE+1))	/* etc. */
+#define R_CB_OFF	(3*(MAXJSAMPLE+1))
+#define G_CB_OFF	(4*(MAXJSAMPLE+1))
+#define B_CB_OFF	(5*(MAXJSAMPLE+1))
+#define R_CR_OFF	B_CB_OFF		/* B=>Cb, R=>Cr are the same */
+#define G_CR_OFF	(6*(MAXJSAMPLE+1))
+#define B_CR_OFF	(7*(MAXJSAMPLE+1))
+#define TABLE_SIZE	(8*(MAXJSAMPLE+1))
+
+
+/*
+ * Initialize for RGB->YCC colorspace conversion.
+ */
+
+METHODDEF(void)
+rgb_ycc_start (j_compress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  INT32 * rgb_ycc_tab;
+  INT32 i;
+
+  /* Allocate and fill in the conversion tables. */
+  cconvert->rgb_ycc_tab = rgb_ycc_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(TABLE_SIZE * SIZEOF(INT32)));
+
+  for (i = 0; i <= MAXJSAMPLE; i++) {
+    rgb_ycc_tab[i+R_Y_OFF] = FIX(0.29900) * i;
+    rgb_ycc_tab[i+G_Y_OFF] = FIX(0.58700) * i;
+    rgb_ycc_tab[i+B_Y_OFF] = FIX(0.11400) * i     + ONE_HALF;
+    rgb_ycc_tab[i+R_CB_OFF] = (-FIX(0.16874)) * i;
+    rgb_ycc_tab[i+G_CB_OFF] = (-FIX(0.33126)) * i;
+    /* We use a rounding fudge-factor of 0.5-epsilon for Cb and Cr.
+     * This ensures that the maximum output will round to MAXJSAMPLE
+     * not MAXJSAMPLE+1, and thus that we don't have to range-limit.
+     */
+    rgb_ycc_tab[i+B_CB_OFF] = FIX(0.50000) * i    + CBCR_OFFSET + ONE_HALF-1;
+/*  B=>Cb and R=>Cr tables are the same
+    rgb_ycc_tab[i+R_CR_OFF] = FIX(0.50000) * i    + CBCR_OFFSET + ONE_HALF-1;
+*/
+    rgb_ycc_tab[i+G_CR_OFF] = (-FIX(0.41869)) * i;
+    rgb_ycc_tab[i+B_CR_OFF] = (-FIX(0.08131)) * i;
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ *
+ * Note that we change from the application's interleaved-pixel format
+ * to our internal noninterleaved, one-plane-per-component format.
+ * The input buffer is therefore three times as wide as the output buffer.
+ *
+ * A starting row offset is provided only for the output buffer.  The caller
+ * can easily adjust the passed input_buf value to accommodate any row
+ * offset required on that side.
+ */
+
+METHODDEF(void)
+rgb_ycc_convert (j_compress_ptr cinfo,
+		 JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+		 JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr0, outptr1, outptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf[0][output_row];
+    outptr1 = output_buf[1][output_row];
+    outptr2 = output_buf[2][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = GETJSAMPLE(inptr[RGB_RED]);
+      g = GETJSAMPLE(inptr[RGB_GREEN]);
+      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      inptr += RGB_PIXELSIZE;
+      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
+       * must be too; we do not need an explicit range-limiting operation.
+       * Hence the value being shifted is never negative, and we don't
+       * need the general RIGHT_SHIFT macro.
+       */
+      /* Y */
+      outptr0[col] = (JSAMPLE)
+		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+		 >> SCALEBITS);
+      /* Cb */
+      outptr1[col] = (JSAMPLE)
+		((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
+		 >> SCALEBITS);
+      /* Cr */
+      outptr2[col] = (JSAMPLE)
+		((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
+		 >> SCALEBITS);
+    }
+  }
+}
+
+
+/**************** Cases other than RGB -> YCbCr **************/
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles RGB->grayscale conversion, which is the same
+ * as the RGB->Y portion of RGB->YCbCr.
+ * We assume rgb_ycc_start has been called (we only use the Y tables).
+ */
+
+METHODDEF(void)
+rgb_gray_convert (j_compress_ptr cinfo,
+		  JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+		  JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf[0][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = GETJSAMPLE(inptr[RGB_RED]);
+      g = GETJSAMPLE(inptr[RGB_GREEN]);
+      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      inptr += RGB_PIXELSIZE;
+      /* Y */
+      outptr[col] = (JSAMPLE)
+		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+		 >> SCALEBITS);
+    }
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles Adobe-style CMYK->YCCK conversion,
+ * where we convert R=1-C, G=1-M, and B=1-Y to YCbCr using the same
+ * conversion as above, while passing K (black) unchanged.
+ * We assume rgb_ycc_start has been called.
+ */
+
+METHODDEF(void)
+cmyk_ycck_convert (j_compress_ptr cinfo,
+		   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+		   JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr0, outptr1, outptr2, outptr3;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf[0][output_row];
+    outptr1 = output_buf[1][output_row];
+    outptr2 = output_buf[2][output_row];
+    outptr3 = output_buf[3][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = MAXJSAMPLE - GETJSAMPLE(inptr[0]);
+      g = MAXJSAMPLE - GETJSAMPLE(inptr[1]);
+      b = MAXJSAMPLE - GETJSAMPLE(inptr[2]);
+      /* K passes through as-is */
+      outptr3[col] = inptr[3];	/* don't need GETJSAMPLE here */
+      inptr += 4;
+      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
+       * must be too; we do not need an explicit range-limiting operation.
+       * Hence the value being shifted is never negative, and we don't
+       * need the general RIGHT_SHIFT macro.
+       */
+      /* Y */
+      outptr0[col] = (JSAMPLE)
+		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+		 >> SCALEBITS);
+      /* Cb */
+      outptr1[col] = (JSAMPLE)
+		((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
+		 >> SCALEBITS);
+      /* Cr */
+      outptr2[col] = (JSAMPLE)
+		((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
+		 >> SCALEBITS);
+    }
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles grayscale output with no conversion.
+ * The source can be either plain grayscale or YCbCr (since Y == gray).
+ */
+
+METHODDEF(void)
+grayscale_convert (j_compress_ptr cinfo,
+		   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+		   JDIMENSION output_row, int num_rows)
+{
+  register JSAMPROW inptr;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+  int instride = cinfo->input_components;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf[0][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      outptr[col] = inptr[0];	/* don't need GETJSAMPLE() here */
+      inptr += instride;
+    }
+  }
+}
+
+#ifdef ENABLE_ANDROID_NULL_CONVERT
+
+typedef unsigned long UINT32;
+
+#define B0(n)   ((n) & 0xFF)
+#define B1(n)   (((n) >> 8) & 0xFF)
+#define B2(n)   (((n) >> 16) & 0xFF)
+#define B3(n)   ((n) >> 24)
+
+#define PACK(a, b, c, d)    ((a) | ((b) << 8) | ((c) << 16) | ((d) << 24))
+
+static int ptr_is_quad(const void* p)
+{
+    return (((const char*)p - (const char*)0) & 3) == 0;
+}
+
+static void copyquads(const UINT32 in[], UINT32 out0[], UINT32 out1[], UINT32 out2[], int col4)
+{
+    do {
+        UINT32 src0 = *in++;
+        UINT32 src1 = *in++;
+        UINT32 src2 = *in++;
+        // LEndian
+        *out0++ = PACK(B0(src0), B3(src0), B2(src1), B1(src2));
+        *out1++ = PACK(B1(src0), B0(src1), B3(src1), B2(src2));
+        *out2++ = PACK(B2(src0), B1(src1), B0(src2), B3(src2));
+    } while (--col4 != 0);
+}
+
+#endif
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles multi-component colorspaces without conversion.
+ * We assume input_components == num_components.
+ */
+
+METHODDEF(void)
+null_convert (j_compress_ptr cinfo,
+	      JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+	      JDIMENSION output_row, int num_rows)
+{
+  register JSAMPROW inptr;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  register int ci;
+  int nc = cinfo->num_components;
+  JDIMENSION num_cols = cinfo->image_width;
+
+#ifdef ENABLE_ANDROID_NULL_CONVERT
+    if (1 == num_rows && 3 == nc && num_cols > 0) {
+        JSAMPROW inptr = *input_buf;
+        JSAMPROW outptr0 = output_buf[0][output_row];
+        JSAMPROW outptr1 = output_buf[1][output_row];
+        JSAMPROW outptr2 = output_buf[2][output_row];
+        
+        int col = num_cols;
+        int col4 = col >> 2;
+        if (col4 > 0 && ptr_is_quad(inptr) && ptr_is_quad(outptr0) &&
+                        ptr_is_quad(outptr1) && ptr_is_quad(outptr2)) {
+            
+            const UINT32* in = (const UINT32*)inptr;
+            UINT32* out0 = (UINT32*)outptr0;
+            UINT32* out1 = (UINT32*)outptr1;
+            UINT32* out2 = (UINT32*)outptr2;
+            copyquads(in, out0, out1, out2, col4);
+            col &= 3;
+            if (0 == col)
+                return;
+            col4 <<= 2;
+            inptr += col4 * 3;  /* we read this 3 times per in copyquads */
+            outptr0 += col4;
+            outptr1 += col4;
+            outptr2 += col4;
+            /* fall through to while-loop */
+        }
+        do {
+            *outptr0++ = *inptr++;
+            *outptr1++ = *inptr++;
+            *outptr2++ = *inptr++;
+        } while (--col != 0);
+        return;
+    }
+SLOW:
+#endif
+  while (--num_rows >= 0) {
+    /* It seems fastest to make a separate pass for each component. */
+    for (ci = 0; ci < nc; ci++) {
+      inptr = *input_buf;
+      outptr = output_buf[ci][output_row];
+      for (col = 0; col < num_cols; col++) {
+	outptr[col] = inptr[ci]; /* don't need GETJSAMPLE() here */
+	inptr += nc;
+      }
+    }
+    input_buf++;
+    output_row++;
+  }
+}
+
+
+/*
+ * Empty method for start_pass.
+ */
+
+METHODDEF(void)
+null_method (j_compress_ptr cinfo)
+{
+  /* no work needed */
+}
+
+
+/*
+ * Module initialization routine for input colorspace conversion.
+ */
+
+GLOBAL(void)
+jinit_color_converter (j_compress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert;
+
+  cconvert = (my_cconvert_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_color_converter));
+  cinfo->cconvert = (struct jpeg_color_converter *) cconvert;
+  /* set start_pass to null method until we find out differently */
+  cconvert->pub.start_pass = null_method;
+
+  /* Make sure input_components agrees with in_color_space */
+  switch (cinfo->in_color_space) {
+  case JCS_GRAYSCALE:
+    if (cinfo->input_components != 1)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+
+  case JCS_RGB:
+#if RGB_PIXELSIZE != 3
+    if (cinfo->input_components != RGB_PIXELSIZE)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+#endif /* else share code with YCbCr */
+
+  case JCS_YCbCr:
+    if (cinfo->input_components != 3)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+
+  case JCS_CMYK:
+  case JCS_YCCK:
+    if (cinfo->input_components != 4)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+
+  default:			/* JCS_UNKNOWN can be anything */
+    if (cinfo->input_components < 1)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+  }
+
+  /* Check num_components, set conversion method based on requested space */
+  switch (cinfo->jpeg_color_space) {
+  case JCS_GRAYSCALE:
+    if (cinfo->num_components != 1)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_GRAYSCALE)
+      cconvert->pub.color_convert = grayscale_convert;
+    else if (cinfo->in_color_space == JCS_RGB) {
+      cconvert->pub.start_pass = rgb_ycc_start;
+      cconvert->pub.color_convert = rgb_gray_convert;
+    } else if (cinfo->in_color_space == JCS_YCbCr)
+      cconvert->pub.color_convert = grayscale_convert;
+    else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_RGB:
+    if (cinfo->num_components != 3)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_RGB && RGB_PIXELSIZE == 3)
+      cconvert->pub.color_convert = null_convert;
+    else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_YCbCr:
+    if (cinfo->num_components != 3)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_RGB) {
+      cconvert->pub.start_pass = rgb_ycc_start;
+      cconvert->pub.color_convert = rgb_ycc_convert;
+    } else if (cinfo->in_color_space == JCS_YCbCr)
+      cconvert->pub.color_convert = null_convert;
+    else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_CMYK:
+    if (cinfo->num_components != 4)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_CMYK)
+      cconvert->pub.color_convert = null_convert;
+    else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_YCCK:
+    if (cinfo->num_components != 4)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_CMYK) {
+      cconvert->pub.start_pass = rgb_ycc_start;
+      cconvert->pub.color_convert = cmyk_ycck_convert;
+    } else if (cinfo->in_color_space == JCS_YCCK)
+      cconvert->pub.color_convert = null_convert;
+    else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  default:			/* allow null conversion of JCS_UNKNOWN */
+    if (cinfo->jpeg_color_space != cinfo->in_color_space ||
+	cinfo->num_components != cinfo->input_components)
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    cconvert->pub.color_convert = null_convert;
+    break;
+  }
+}
diff --git a/libjpegtwrp/jcdctmgr.c b/libjpegtwrp/jcdctmgr.c
new file mode 100644
index 000000000..61fa79b9e
--- /dev/null
+++ b/libjpegtwrp/jcdctmgr.c
@@ -0,0 +1,387 @@
+/*
+ * jcdctmgr.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the forward-DCT management logic.
+ * This code selects a particular DCT implementation to be used,
+ * and it performs related housekeeping chores including coefficient
+ * quantization.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+
+/* Private subobject for this module */
+
+typedef struct {
+  struct jpeg_forward_dct pub;	/* public fields */
+
+  /* Pointer to the DCT routine actually in use */
+  forward_DCT_method_ptr do_dct;
+
+  /* The actual post-DCT divisors --- not identical to the quant table
+   * entries, because of scaling (especially for an unnormalized DCT).
+   * Each table is given in normal array order.
+   */
+  DCTELEM * divisors[NUM_QUANT_TBLS];
+
+#ifdef DCT_FLOAT_SUPPORTED
+  /* Same as above for the floating-point case. */
+  float_DCT_method_ptr do_float_dct;
+  FAST_FLOAT * float_divisors[NUM_QUANT_TBLS];
+#endif
+} my_fdct_controller;
+
+typedef my_fdct_controller * my_fdct_ptr;
+
+
+/*
+ * Initialize for a processing pass.
+ * Verify that all referenced Q-tables are present, and set up
+ * the divisor table for each one.
+ * In the current implementation, DCT of all components is done during
+ * the first pass, even if only some components will be output in the
+ * first scan.  Hence all components should be examined here.
+ */
+
+METHODDEF(void)
+start_pass_fdctmgr (j_compress_ptr cinfo)
+{
+  my_fdct_ptr fdct = (my_fdct_ptr) cinfo->fdct;
+  int ci, qtblno, i;
+  jpeg_component_info *compptr;
+  JQUANT_TBL * qtbl;
+  DCTELEM * dtbl;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    qtblno = compptr->quant_tbl_no;
+    /* Make sure specified quantization table is present */
+    if (qtblno < 0 || qtblno >= NUM_QUANT_TBLS ||
+	cinfo->quant_tbl_ptrs[qtblno] == NULL)
+      ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, qtblno);
+    qtbl = cinfo->quant_tbl_ptrs[qtblno];
+    /* Compute divisors for this quant table */
+    /* We may do this more than once for same table, but it's not a big deal */
+    switch (cinfo->dct_method) {
+#ifdef DCT_ISLOW_SUPPORTED
+    case JDCT_ISLOW:
+      /* For LL&M IDCT method, divisors are equal to raw quantization
+       * coefficients multiplied by 8 (to counteract scaling).
+       */
+      if (fdct->divisors[qtblno] == NULL) {
+	fdct->divisors[qtblno] = (DCTELEM *)
+	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				      DCTSIZE2 * SIZEOF(DCTELEM));
+      }
+      dtbl = fdct->divisors[qtblno];
+      for (i = 0; i < DCTSIZE2; i++) {
+	dtbl[i] = ((DCTELEM) qtbl->quantval[i]) << 3;
+      }
+      break;
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+    case JDCT_IFAST:
+      {
+	/* For AA&N IDCT method, divisors are equal to quantization
+	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
+	 *   scalefactor[0] = 1
+	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+	 * We apply a further scale factor of 8.
+	 */
+#define CONST_BITS 14
+	static const INT16 aanscales[DCTSIZE2] = {
+	  /* precomputed values scaled up by 14 bits */
+	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+	  22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
+	  21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
+	  19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
+	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+	  12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
+	   8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
+	   4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
+	};
+	SHIFT_TEMPS
+
+	if (fdct->divisors[qtblno] == NULL) {
+	  fdct->divisors[qtblno] = (DCTELEM *)
+	    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+					DCTSIZE2 * SIZEOF(DCTELEM));
+	}
+	dtbl = fdct->divisors[qtblno];
+	for (i = 0; i < DCTSIZE2; i++) {
+	  dtbl[i] = (DCTELEM)
+	    DESCALE(MULTIPLY16V16((INT32) qtbl->quantval[i],
+				  (INT32) aanscales[i]),
+		    CONST_BITS-3);
+	}
+      }
+      break;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+    case JDCT_FLOAT:
+      {
+	/* For float AA&N IDCT method, divisors are equal to quantization
+	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
+	 *   scalefactor[0] = 1
+	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+	 * We apply a further scale factor of 8.
+	 * What's actually stored is 1/divisor so that the inner loop can
+	 * use a multiplication rather than a division.
+	 */
+	FAST_FLOAT * fdtbl;
+	int row, col;
+	static const double aanscalefactor[DCTSIZE] = {
+	  1.0, 1.387039845, 1.306562965, 1.175875602,
+	  1.0, 0.785694958, 0.541196100, 0.275899379
+	};
+
+	if (fdct->float_divisors[qtblno] == NULL) {
+	  fdct->float_divisors[qtblno] = (FAST_FLOAT *)
+	    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+					DCTSIZE2 * SIZEOF(FAST_FLOAT));
+	}
+	fdtbl = fdct->float_divisors[qtblno];
+	i = 0;
+	for (row = 0; row < DCTSIZE; row++) {
+	  for (col = 0; col < DCTSIZE; col++) {
+	    fdtbl[i] = (FAST_FLOAT)
+	      (1.0 / (((double) qtbl->quantval[i] *
+		       aanscalefactor[row] * aanscalefactor[col] * 8.0)));
+	    i++;
+	  }
+	}
+      }
+      break;
+#endif
+    default:
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+      break;
+    }
+  }
+}
+
+
+/*
+ * Perform forward DCT on one or more blocks of a component.
+ *
+ * The input samples are taken from the sample_data[] array starting at
+ * position start_row/start_col, and moving to the right for any additional
+ * blocks. The quantized coefficients are returned in coef_blocks[].
+ */
+
+METHODDEF(void)
+forward_DCT (j_compress_ptr cinfo, jpeg_component_info * compptr,
+	     JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+	     JDIMENSION start_row, JDIMENSION start_col,
+	     JDIMENSION num_blocks)
+/* This version is used for integer DCT implementations. */
+{
+  /* This routine is heavily used, so it's worth coding it tightly. */
+  my_fdct_ptr fdct = (my_fdct_ptr) cinfo->fdct;
+  forward_DCT_method_ptr do_dct = fdct->do_dct;
+  DCTELEM * divisors = fdct->divisors[compptr->quant_tbl_no];
+  DCTELEM workspace[DCTSIZE2];	/* work area for FDCT subroutine */
+  JDIMENSION bi;
+
+  sample_data += start_row;	/* fold in the vertical offset once */
+
+  for (bi = 0; bi < num_blocks; bi++, start_col += DCTSIZE) {
+    /* Load data into workspace, applying unsigned->signed conversion */
+    { register DCTELEM *workspaceptr;
+      register JSAMPROW elemptr;
+      register int elemr;
+
+      workspaceptr = workspace;
+      for (elemr = 0; elemr < DCTSIZE; elemr++) {
+	elemptr = sample_data[elemr] + start_col;
+#if DCTSIZE == 8		/* unroll the inner loop */
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+#else
+	{ register int elemc;
+	  for (elemc = DCTSIZE; elemc > 0; elemc--) {
+	    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	  }
+	}
+#endif
+      }
+    }
+
+    /* Perform the DCT */
+    (*do_dct) (workspace);
+
+    /* Quantize/descale the coefficients, and store into coef_blocks[] */
+    { register DCTELEM temp, qval;
+      register int i;
+      register JCOEFPTR output_ptr = coef_blocks[bi];
+
+      for (i = 0; i < DCTSIZE2; i++) {
+	qval = divisors[i];
+	temp = workspace[i];
+	/* Divide the coefficient value by qval, ensuring proper rounding.
+	 * Since C does not specify the direction of rounding for negative
+	 * quotients, we have to force the dividend positive for portability.
+	 *
+	 * In most files, at least half of the output values will be zero
+	 * (at default quantization settings, more like three-quarters...)
+	 * so we should ensure that this case is fast.  On many machines,
+	 * a comparison is enough cheaper than a divide to make a special test
+	 * a win.  Since both inputs will be nonnegative, we need only test
+	 * for a < b to discover whether a/b is 0.
+	 * If your machine's division is fast enough, define FAST_DIVIDE.
+	 */
+#ifdef FAST_DIVIDE
+#define DIVIDE_BY(a,b)	a /= b
+#else
+#define DIVIDE_BY(a,b)	if (a >= b) a /= b; else a = 0
+#endif
+	if (temp < 0) {
+	  temp = -temp;
+	  temp += qval>>1;	/* for rounding */
+	  DIVIDE_BY(temp, qval);
+	  temp = -temp;
+	} else {
+	  temp += qval>>1;	/* for rounding */
+	  DIVIDE_BY(temp, qval);
+	}
+	output_ptr[i] = (JCOEF) temp;
+      }
+    }
+  }
+}
+
+
+#ifdef DCT_FLOAT_SUPPORTED
+
+METHODDEF(void)
+forward_DCT_float (j_compress_ptr cinfo, jpeg_component_info * compptr,
+		   JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+		   JDIMENSION start_row, JDIMENSION start_col,
+		   JDIMENSION num_blocks)
+/* This version is used for floating-point DCT implementations. */
+{
+  /* This routine is heavily used, so it's worth coding it tightly. */
+  my_fdct_ptr fdct = (my_fdct_ptr) cinfo->fdct;
+  float_DCT_method_ptr do_dct = fdct->do_float_dct;
+  FAST_FLOAT * divisors = fdct->float_divisors[compptr->quant_tbl_no];
+  FAST_FLOAT workspace[DCTSIZE2]; /* work area for FDCT subroutine */
+  JDIMENSION bi;
+
+  sample_data += start_row;	/* fold in the vertical offset once */
+
+  for (bi = 0; bi < num_blocks; bi++, start_col += DCTSIZE) {
+    /* Load data into workspace, applying unsigned->signed conversion */
+    { register FAST_FLOAT *workspaceptr;
+      register JSAMPROW elemptr;
+      register int elemr;
+
+      workspaceptr = workspace;
+      for (elemr = 0; elemr < DCTSIZE; elemr++) {
+	elemptr = sample_data[elemr] + start_col;
+#if DCTSIZE == 8		/* unroll the inner loop */
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+#else
+	{ register int elemc;
+	  for (elemc = DCTSIZE; elemc > 0; elemc--) {
+	    *workspaceptr++ = (FAST_FLOAT)
+	      (GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	  }
+	}
+#endif
+      }
+    }
+
+    /* Perform the DCT */
+    (*do_dct) (workspace);
+
+    /* Quantize/descale the coefficients, and store into coef_blocks[] */
+    { register FAST_FLOAT temp;
+      register int i;
+      register JCOEFPTR output_ptr = coef_blocks[bi];
+
+      for (i = 0; i < DCTSIZE2; i++) {
+	/* Apply the quantization and scaling factor */
+	temp = workspace[i] * divisors[i];
+	/* Round to nearest integer.
+	 * Since C does not specify the direction of rounding for negative
+	 * quotients, we have to force the dividend positive for portability.
+	 * The maximum coefficient size is +-16K (for 12-bit data), so this
+	 * code should work for either 16-bit or 32-bit ints.
+	 */
+	output_ptr[i] = (JCOEF) ((int) (temp + (FAST_FLOAT) 16384.5) - 16384);
+      }
+    }
+  }
+}
+
+#endif /* DCT_FLOAT_SUPPORTED */
+
+
+/*
+ * Initialize FDCT manager.
+ */
+
+GLOBAL(void)
+jinit_forward_dct (j_compress_ptr cinfo)
+{
+  my_fdct_ptr fdct;
+  int i;
+
+  fdct = (my_fdct_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_fdct_controller));
+  cinfo->fdct = (struct jpeg_forward_dct *) fdct;
+  fdct->pub.start_pass = start_pass_fdctmgr;
+
+  switch (cinfo->dct_method) {
+#ifdef DCT_ISLOW_SUPPORTED
+  case JDCT_ISLOW:
+    fdct->pub.forward_DCT = forward_DCT;
+    fdct->do_dct = jpeg_fdct_islow;
+    break;
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+  case JDCT_IFAST:
+    fdct->pub.forward_DCT = forward_DCT;
+    fdct->do_dct = jpeg_fdct_ifast;
+    break;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  case JDCT_FLOAT:
+    fdct->pub.forward_DCT = forward_DCT_float;
+    fdct->do_float_dct = jpeg_fdct_float;
+    break;
+#endif
+  default:
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+    break;
+  }
+
+  /* Mark divisor tables unallocated */
+  for (i = 0; i < NUM_QUANT_TBLS; i++) {
+    fdct->divisors[i] = NULL;
+#ifdef DCT_FLOAT_SUPPORTED
+    fdct->float_divisors[i] = NULL;
+#endif
+  }
+}
diff --git a/libjpegtwrp/jchuff.c b/libjpegtwrp/jchuff.c
new file mode 100644
index 000000000..f23525054
--- /dev/null
+++ b/libjpegtwrp/jchuff.c
@@ -0,0 +1,909 @@
+/*
+ * jchuff.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains Huffman entropy encoding routines.
+ *
+ * Much of the complexity here has to do with supporting output suspension.
+ * If the data destination module demands suspension, we want to be able to
+ * back up to the start of the current MCU.  To do this, we copy state
+ * variables into local working storage, and update them back to the
+ * permanent JPEG objects only upon successful completion of an MCU.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jchuff.h"		/* Declarations shared with jcphuff.c */
+
+
+/* Expanded entropy encoder object for Huffman encoding.
+ *
+ * The savable_state subrecord contains fields that change within an MCU,
+ * but must not be updated permanently until we complete the MCU.
+ */
+
+typedef struct {
+  INT32 put_buffer;		/* current bit-accumulation buffer */
+  int put_bits;			/* # of bits now in it */
+  int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
+} savable_state;
+
+/* This macro is to work around compilers with missing or broken
+ * structure assignment.  You'll need to fix this code if you have
+ * such a compiler and you change MAX_COMPS_IN_SCAN.
+ */
+
+#ifndef NO_STRUCT_ASSIGN
+#define ASSIGN_STATE(dest,src)  ((dest) = (src))
+#else
+#if MAX_COMPS_IN_SCAN == 4
+#define ASSIGN_STATE(dest,src)  \
+	((dest).put_buffer = (src).put_buffer, \
+	 (dest).put_bits = (src).put_bits, \
+	 (dest).last_dc_val[0] = (src).last_dc_val[0], \
+	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
+	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
+	 (dest).last_dc_val[3] = (src).last_dc_val[3])
+#endif
+#endif
+
+
+typedef struct {
+  struct jpeg_entropy_encoder pub; /* public fields */
+
+  savable_state saved;		/* Bit buffer & DC state at start of MCU */
+
+  /* These fields are NOT loaded into local working state. */
+  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+  int next_restart_num;		/* next restart number to write (0-7) */
+
+  /* Pointers to derived tables (these workspaces have image lifespan) */
+  c_derived_tbl * dc_derived_tbls[NUM_HUFF_TBLS];
+  c_derived_tbl * ac_derived_tbls[NUM_HUFF_TBLS];
+
+#ifdef ENTROPY_OPT_SUPPORTED	/* Statistics tables for optimization */
+  long * dc_count_ptrs[NUM_HUFF_TBLS];
+  long * ac_count_ptrs[NUM_HUFF_TBLS];
+#endif
+} huff_entropy_encoder;
+
+typedef huff_entropy_encoder * huff_entropy_ptr;
+
+/* Working state while writing an MCU.
+ * This struct contains all the fields that are needed by subroutines.
+ */
+
+typedef struct {
+  JOCTET * next_output_byte;	/* => next byte to write in buffer */
+  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
+  savable_state cur;		/* Current bit buffer & DC state */
+  j_compress_ptr cinfo;		/* dump_buffer needs access to this */
+} working_state;
+
+
+/* Forward declarations */
+METHODDEF(boolean) encode_mcu_huff JPP((j_compress_ptr cinfo,
+					JBLOCKROW *MCU_data));
+METHODDEF(void) finish_pass_huff JPP((j_compress_ptr cinfo));
+#ifdef ENTROPY_OPT_SUPPORTED
+METHODDEF(boolean) encode_mcu_gather JPP((j_compress_ptr cinfo,
+					  JBLOCKROW *MCU_data));
+METHODDEF(void) finish_pass_gather JPP((j_compress_ptr cinfo));
+#endif
+
+
+/*
+ * Initialize for a Huffman-compressed scan.
+ * If gather_statistics is TRUE, we do not output anything during the scan,
+ * just count the Huffman symbols used and generate Huffman code tables.
+ */
+
+METHODDEF(void)
+start_pass_huff (j_compress_ptr cinfo, boolean gather_statistics)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int ci, dctbl, actbl;
+  jpeg_component_info * compptr;
+
+  if (gather_statistics) {
+#ifdef ENTROPY_OPT_SUPPORTED
+    entropy->pub.encode_mcu = encode_mcu_gather;
+    entropy->pub.finish_pass = finish_pass_gather;
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    entropy->pub.encode_mcu = encode_mcu_huff;
+    entropy->pub.finish_pass = finish_pass_huff;
+  }
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    dctbl = compptr->dc_tbl_no;
+    actbl = compptr->ac_tbl_no;
+    if (gather_statistics) {
+#ifdef ENTROPY_OPT_SUPPORTED
+      /* Check for invalid table indexes */
+      /* (make_c_derived_tbl does this in the other path) */
+      if (dctbl < 0 || dctbl >= NUM_HUFF_TBLS)
+	ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, dctbl);
+      if (actbl < 0 || actbl >= NUM_HUFF_TBLS)
+	ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, actbl);
+      /* Allocate and zero the statistics tables */
+      /* Note that jpeg_gen_optimal_table expects 257 entries in each table! */
+      if (entropy->dc_count_ptrs[dctbl] == NULL)
+	entropy->dc_count_ptrs[dctbl] = (long *)
+	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				      257 * SIZEOF(long));
+      MEMZERO(entropy->dc_count_ptrs[dctbl], 257 * SIZEOF(long));
+      if (entropy->ac_count_ptrs[actbl] == NULL)
+	entropy->ac_count_ptrs[actbl] = (long *)
+	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				      257 * SIZEOF(long));
+      MEMZERO(entropy->ac_count_ptrs[actbl], 257 * SIZEOF(long));
+#endif
+    } else {
+      /* Compute derived values for Huffman tables */
+      /* We may do this more than once for a table, but it's not expensive */
+      jpeg_make_c_derived_tbl(cinfo, TRUE, dctbl,
+			      & entropy->dc_derived_tbls[dctbl]);
+      jpeg_make_c_derived_tbl(cinfo, FALSE, actbl,
+			      & entropy->ac_derived_tbls[actbl]);
+    }
+    /* Initialize DC predictions to 0 */
+    entropy->saved.last_dc_val[ci] = 0;
+  }
+
+  /* Initialize bit buffer to empty */
+  entropy->saved.put_buffer = 0;
+  entropy->saved.put_bits = 0;
+
+  /* Initialize restart stuff */
+  entropy->restarts_to_go = cinfo->restart_interval;
+  entropy->next_restart_num = 0;
+}
+
+
+/*
+ * Compute the derived values for a Huffman table.
+ * This routine also performs some validation checks on the table.
+ *
+ * Note this is also used by jcphuff.c.
+ */
+
+GLOBAL(void)
+jpeg_make_c_derived_tbl (j_compress_ptr cinfo, boolean isDC, int tblno,
+			 c_derived_tbl ** pdtbl)
+{
+  JHUFF_TBL *htbl;
+  c_derived_tbl *dtbl;
+  int p, i, l, lastp, si, maxsymbol;
+  char huffsize[257];
+  unsigned int huffcode[257];
+  unsigned int code;
+
+  /* Note that huffsize[] and huffcode[] are filled in code-length order,
+   * paralleling the order of the symbols themselves in htbl->huffval[].
+   */
+
+  /* Find the input Huffman table */
+  if (tblno < 0 || tblno >= NUM_HUFF_TBLS)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
+  htbl =
+    isDC ? cinfo->dc_huff_tbl_ptrs[tblno] : cinfo->ac_huff_tbl_ptrs[tblno];
+  if (htbl == NULL)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
+
+  /* Allocate a workspace if we haven't already done so. */
+  if (*pdtbl == NULL)
+    *pdtbl = (c_derived_tbl *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(c_derived_tbl));
+  dtbl = *pdtbl;
+  
+  /* Figure C.1: make table of Huffman code length for each symbol */
+
+  p = 0;
+  for (l = 1; l <= 16; l++) {
+    i = (int) htbl->bits[l];
+    if (i < 0 || p + i > 256)	/* protect against table overrun */
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    while (i--)
+      huffsize[p++] = (char) l;
+  }
+  huffsize[p] = 0;
+  lastp = p;
+  
+  /* Figure C.2: generate the codes themselves */
+  /* We also validate that the counts represent a legal Huffman code tree. */
+
+  code = 0;
+  si = huffsize[0];
+  p = 0;
+  while (huffsize[p]) {
+    while (((int) huffsize[p]) == si) {
+      huffcode[p++] = code;
+      code++;
+    }
+    /* code is now 1 more than the last code used for codelength si; but
+     * it must still fit in si bits, since no code is allowed to be all ones.
+     */
+    if (((INT32) code) >= (((INT32) 1) << si))
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    code <<= 1;
+    si++;
+  }
+  
+  /* Figure C.3: generate encoding tables */
+  /* These are code and size indexed by symbol value */
+
+  /* Set all codeless symbols to have code length 0;
+   * this lets us detect duplicate VAL entries here, and later
+   * allows emit_bits to detect any attempt to emit such symbols.
+   */
+  MEMZERO(dtbl->ehufsi, SIZEOF(dtbl->ehufsi));
+
+  /* This is also a convenient place to check for out-of-range
+   * and duplicated VAL entries.  We allow 0..255 for AC symbols
+   * but only 0..15 for DC.  (We could constrain them further
+   * based on data depth and mode, but this seems enough.)
+   */
+  maxsymbol = isDC ? 15 : 255;
+
+  for (p = 0; p < lastp; p++) {
+    i = htbl->huffval[p];
+    if (i < 0 || i > maxsymbol || dtbl->ehufsi[i])
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    dtbl->ehufco[i] = huffcode[p];
+    dtbl->ehufsi[i] = huffsize[p];
+  }
+}
+
+
+/* Outputting bytes to the file */
+
+/* Emit a byte, taking 'action' if must suspend. */
+#define emit_byte(state,val,action)  \
+	{ *(state)->next_output_byte++ = (JOCTET) (val);  \
+	  if (--(state)->free_in_buffer == 0)  \
+	    if (! dump_buffer(state))  \
+	      { action; } }
+
+
+LOCAL(boolean)
+dump_buffer (working_state * state)
+/* Empty the output buffer; return TRUE if successful, FALSE if must suspend */
+{
+  struct jpeg_destination_mgr * dest = state->cinfo->dest;
+
+  if (! (*dest->empty_output_buffer) (state->cinfo))
+    return FALSE;
+  /* After a successful buffer dump, must reset buffer pointers */
+  state->next_output_byte = dest->next_output_byte;
+  state->free_in_buffer = dest->free_in_buffer;
+  return TRUE;
+}
+
+
+/* Outputting bits to the file */
+
+/* Only the right 24 bits of put_buffer are used; the valid bits are
+ * left-justified in this part.  At most 16 bits can be passed to emit_bits
+ * in one call, and we never retain more than 7 bits in put_buffer
+ * between calls, so 24 bits are sufficient.
+ */
+
+INLINE
+LOCAL(boolean)
+emit_bits (working_state * state, unsigned int code, int size)
+/* Emit some bits; return TRUE if successful, FALSE if must suspend */
+{
+  /* This routine is heavily used, so it's worth coding tightly. */
+  register INT32 put_buffer = (INT32) code;
+  register int put_bits = state->cur.put_bits;
+
+  /* if size is 0, caller used an invalid Huffman table entry */
+  if (size == 0)
+    ERREXIT(state->cinfo, JERR_HUFF_MISSING_CODE);
+
+  put_buffer &= (((INT32) 1)<<size) - 1; /* mask off any extra bits in code */
+  
+  put_bits += size;		/* new number of bits in buffer */
+  
+  put_buffer <<= 24 - put_bits; /* align incoming bits */
+
+  put_buffer |= state->cur.put_buffer; /* and merge with old buffer contents */
+  
+  while (put_bits >= 8) {
+    int c = (int) ((put_buffer >> 16) & 0xFF);
+    
+    emit_byte(state, c, return FALSE);
+    if (c == 0xFF) {		/* need to stuff a zero byte? */
+      emit_byte(state, 0, return FALSE);
+    }
+    put_buffer <<= 8;
+    put_bits -= 8;
+  }
+
+  state->cur.put_buffer = put_buffer; /* update state variables */
+  state->cur.put_bits = put_bits;
+
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+flush_bits (working_state * state)
+{
+  if (! emit_bits(state, 0x7F, 7)) /* fill any partial byte with ones */
+    return FALSE;
+  state->cur.put_buffer = 0;	/* and reset bit-buffer to empty */
+  state->cur.put_bits = 0;
+  return TRUE;
+}
+
+
+/* Encode a single block's worth of coefficients */
+
+LOCAL(boolean)
+encode_one_block (working_state * state, JCOEFPTR block, int last_dc_val,
+		  c_derived_tbl *dctbl, c_derived_tbl *actbl)
+{
+  register int temp, temp2;
+  register int nbits;
+  register int k, r, i;
+  
+  /* Encode the DC coefficient difference per section F.1.2.1 */
+  
+  temp = temp2 = block[0] - last_dc_val;
+
+  if (temp < 0) {
+    temp = -temp;		/* temp is abs value of input */
+    /* For a negative input, want temp2 = bitwise complement of abs(input) */
+    /* This code assumes we are on a two's complement machine */
+    temp2--;
+  }
+  
+  /* Find the number of bits needed for the magnitude of the coefficient */
+  nbits = 0;
+  while (temp) {
+    nbits++;
+    temp >>= 1;
+  }
+  /* Check for out-of-range coefficient values.
+   * Since we're encoding a difference, the range limit is twice as much.
+   */
+  if (nbits > MAX_COEF_BITS+1)
+    ERREXIT(state->cinfo, JERR_BAD_DCT_COEF);
+  
+  /* Emit the Huffman-coded symbol for the number of bits */
+  if (! emit_bits(state, dctbl->ehufco[nbits], dctbl->ehufsi[nbits]))
+    return FALSE;
+
+  /* Emit that number of bits of the value, if positive, */
+  /* or the complement of its magnitude, if negative. */
+  if (nbits)			/* emit_bits rejects calls with size 0 */
+    if (! emit_bits(state, (unsigned int) temp2, nbits))
+      return FALSE;
+
+  /* Encode the AC coefficients per section F.1.2.2 */
+  
+  r = 0;			/* r = run length of zeros */
+  
+  for (k = 1; k < DCTSIZE2; k++) {
+    if ((temp = block[jpeg_natural_order[k]]) == 0) {
+      r++;
+    } else {
+      /* if run length > 15, must emit special run-length-16 codes (0xF0) */
+      while (r > 15) {
+	if (! emit_bits(state, actbl->ehufco[0xF0], actbl->ehufsi[0xF0]))
+	  return FALSE;
+	r -= 16;
+      }
+
+      temp2 = temp;
+      if (temp < 0) {
+	temp = -temp;		/* temp is abs value of input */
+	/* This code assumes we are on a two's complement machine */
+	temp2--;
+      }
+      
+      /* Find the number of bits needed for the magnitude of the coefficient */
+      nbits = 1;		/* there must be at least one 1 bit */
+      while ((temp >>= 1))
+	nbits++;
+      /* Check for out-of-range coefficient values */
+      if (nbits > MAX_COEF_BITS)
+	ERREXIT(state->cinfo, JERR_BAD_DCT_COEF);
+      
+      /* Emit Huffman symbol for run length / number of bits */
+      i = (r << 4) + nbits;
+      if (! emit_bits(state, actbl->ehufco[i], actbl->ehufsi[i]))
+	return FALSE;
+
+      /* Emit that number of bits of the value, if positive, */
+      /* or the complement of its magnitude, if negative. */
+      if (! emit_bits(state, (unsigned int) temp2, nbits))
+	return FALSE;
+      
+      r = 0;
+    }
+  }
+
+  /* If the last coef(s) were zero, emit an end-of-block code */
+  if (r > 0)
+    if (! emit_bits(state, actbl->ehufco[0], actbl->ehufsi[0]))
+      return FALSE;
+
+  return TRUE;
+}
+
+
+/*
+ * Emit a restart marker & resynchronize predictions.
+ */
+
+LOCAL(boolean)
+emit_restart (working_state * state, int restart_num)
+{
+  int ci;
+
+  if (! flush_bits(state))
+    return FALSE;
+
+  emit_byte(state, 0xFF, return FALSE);
+  emit_byte(state, JPEG_RST0 + restart_num, return FALSE);
+
+  /* Re-initialize DC predictions to 0 */
+  for (ci = 0; ci < state->cinfo->comps_in_scan; ci++)
+    state->cur.last_dc_val[ci] = 0;
+
+  /* The restart counter is not updated until we successfully write the MCU. */
+
+  return TRUE;
+}
+
+
+/*
+ * Encode and output one MCU's worth of Huffman-compressed coefficients.
+ */
+
+METHODDEF(boolean)
+encode_mcu_huff (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  working_state state;
+  int blkn, ci;
+  jpeg_component_info * compptr;
+
+  /* Load up working state */
+  state.next_output_byte = cinfo->dest->next_output_byte;
+  state.free_in_buffer = cinfo->dest->free_in_buffer;
+  ASSIGN_STATE(state.cur, entropy->saved);
+  state.cinfo = cinfo;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! emit_restart(&state, entropy->next_restart_num))
+	return FALSE;
+  }
+
+  /* Encode the MCU data blocks */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    ci = cinfo->MCU_membership[blkn];
+    compptr = cinfo->cur_comp_info[ci];
+    if (! encode_one_block(&state,
+			   MCU_data[blkn][0], state.cur.last_dc_val[ci],
+			   entropy->dc_derived_tbls[compptr->dc_tbl_no],
+			   entropy->ac_derived_tbls[compptr->ac_tbl_no]))
+      return FALSE;
+    /* Update last_dc_val */
+    state.cur.last_dc_val[ci] = MCU_data[blkn][0][0];
+  }
+
+  /* Completed MCU, so update state */
+  cinfo->dest->next_output_byte = state.next_output_byte;
+  cinfo->dest->free_in_buffer = state.free_in_buffer;
+  ASSIGN_STATE(entropy->saved, state.cur);
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Finish up at the end of a Huffman-compressed scan.
+ */
+
+METHODDEF(void)
+finish_pass_huff (j_compress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  working_state state;
+
+  /* Load up working state ... flush_bits needs it */
+  state.next_output_byte = cinfo->dest->next_output_byte;
+  state.free_in_buffer = cinfo->dest->free_in_buffer;
+  ASSIGN_STATE(state.cur, entropy->saved);
+  state.cinfo = cinfo;
+
+  /* Flush out the last data */
+  if (! flush_bits(&state))
+    ERREXIT(cinfo, JERR_CANT_SUSPEND);
+
+  /* Update state */
+  cinfo->dest->next_output_byte = state.next_output_byte;
+  cinfo->dest->free_in_buffer = state.free_in_buffer;
+  ASSIGN_STATE(entropy->saved, state.cur);
+}
+
+
+/*
+ * Huffman coding optimization.
+ *
+ * We first scan the supplied data and count the number of uses of each symbol
+ * that is to be Huffman-coded. (This process MUST agree with the code above.)
+ * Then we build a Huffman coding tree for the observed counts.
+ * Symbols which are not needed at all for the particular image are not
+ * assigned any code, which saves space in the DHT marker as well as in
+ * the compressed data.
+ */
+
+#ifdef ENTROPY_OPT_SUPPORTED
+
+
+/* Process a single block's worth of coefficients */
+
+LOCAL(void)
+htest_one_block (j_compress_ptr cinfo, JCOEFPTR block, int last_dc_val,
+		 long dc_counts[], long ac_counts[])
+{
+  register int temp;
+  register int nbits;
+  register int k, r;
+  
+  /* Encode the DC coefficient difference per section F.1.2.1 */
+  
+  temp = block[0] - last_dc_val;
+  if (temp < 0)
+    temp = -temp;
+  
+  /* Find the number of bits needed for the magnitude of the coefficient */
+  nbits = 0;
+  while (temp) {
+    nbits++;
+    temp >>= 1;
+  }
+  /* Check for out-of-range coefficient values.
+   * Since we're encoding a difference, the range limit is twice as much.
+   */
+  if (nbits > MAX_COEF_BITS+1)
+    ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+
+  /* Count the Huffman symbol for the number of bits */
+  dc_counts[nbits]++;
+  
+  /* Encode the AC coefficients per section F.1.2.2 */
+  
+  r = 0;			/* r = run length of zeros */
+  
+  for (k = 1; k < DCTSIZE2; k++) {
+    if ((temp = block[jpeg_natural_order[k]]) == 0) {
+      r++;
+    } else {
+      /* if run length > 15, must emit special run-length-16 codes (0xF0) */
+      while (r > 15) {
+	ac_counts[0xF0]++;
+	r -= 16;
+      }
+      
+      /* Find the number of bits needed for the magnitude of the coefficient */
+      if (temp < 0)
+	temp = -temp;
+      
+      /* Find the number of bits needed for the magnitude of the coefficient */
+      nbits = 1;		/* there must be at least one 1 bit */
+      while ((temp >>= 1))
+	nbits++;
+      /* Check for out-of-range coefficient values */
+      if (nbits > MAX_COEF_BITS)
+	ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+      
+      /* Count Huffman symbol for run length / number of bits */
+      ac_counts[(r << 4) + nbits]++;
+      
+      r = 0;
+    }
+  }
+
+  /* If the last coef(s) were zero, emit an end-of-block code */
+  if (r > 0)
+    ac_counts[0]++;
+}
+
+
+/*
+ * Trial-encode one MCU's worth of Huffman-compressed coefficients.
+ * No data is actually output, so no suspension return is possible.
+ */
+
+METHODDEF(boolean)
+encode_mcu_gather (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int blkn, ci;
+  jpeg_component_info * compptr;
+
+  /* Take care of restart intervals if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      /* Re-initialize DC predictions to 0 */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++)
+	entropy->saved.last_dc_val[ci] = 0;
+      /* Update restart state */
+      entropy->restarts_to_go = cinfo->restart_interval;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    ci = cinfo->MCU_membership[blkn];
+    compptr = cinfo->cur_comp_info[ci];
+    htest_one_block(cinfo, MCU_data[blkn][0], entropy->saved.last_dc_val[ci],
+		    entropy->dc_count_ptrs[compptr->dc_tbl_no],
+		    entropy->ac_count_ptrs[compptr->ac_tbl_no]);
+    entropy->saved.last_dc_val[ci] = MCU_data[blkn][0][0];
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Generate the best Huffman code table for the given counts, fill htbl.
+ * Note this is also used by jcphuff.c.
+ *
+ * The JPEG standard requires that no symbol be assigned a codeword of all
+ * one bits (so that padding bits added at the end of a compressed segment
+ * can't look like a valid code).  Because of the canonical ordering of
+ * codewords, this just means that there must be an unused slot in the
+ * longest codeword length category.  Section K.2 of the JPEG spec suggests
+ * reserving such a slot by pretending that symbol 256 is a valid symbol
+ * with count 1.  In theory that's not optimal; giving it count zero but
+ * including it in the symbol set anyway should give a better Huffman code.
+ * But the theoretically better code actually seems to come out worse in
+ * practice, because it produces more all-ones bytes (which incur stuffed
+ * zero bytes in the final file).  In any case the difference is tiny.
+ *
+ * The JPEG standard requires Huffman codes to be no more than 16 bits long.
+ * If some symbols have a very small but nonzero probability, the Huffman tree
+ * must be adjusted to meet the code length restriction.  We currently use
+ * the adjustment method suggested in JPEG section K.2.  This method is *not*
+ * optimal; it may not choose the best possible limited-length code.  But
+ * typically only very-low-frequency symbols will be given less-than-optimal
+ * lengths, so the code is almost optimal.  Experimental comparisons against
+ * an optimal limited-length-code algorithm indicate that the difference is
+ * microscopic --- usually less than a hundredth of a percent of total size.
+ * So the extra complexity of an optimal algorithm doesn't seem worthwhile.
+ */
+
+GLOBAL(void)
+jpeg_gen_optimal_table (j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[])
+{
+#define MAX_CLEN 32		/* assumed maximum initial code length */
+  UINT8 bits[MAX_CLEN+1];	/* bits[k] = # of symbols with code length k */
+  int codesize[257];		/* codesize[k] = code length of symbol k */
+  int others[257];		/* next symbol in current branch of tree */
+  int c1, c2;
+  int p, i, j;
+  long v;
+
+  /* This algorithm is explained in section K.2 of the JPEG standard */
+
+  MEMZERO(bits, SIZEOF(bits));
+  MEMZERO(codesize, SIZEOF(codesize));
+  for (i = 0; i < 257; i++)
+    others[i] = -1;		/* init links to empty */
+  
+  freq[256] = 1;		/* make sure 256 has a nonzero count */
+  /* Including the pseudo-symbol 256 in the Huffman procedure guarantees
+   * that no real symbol is given code-value of all ones, because 256
+   * will be placed last in the largest codeword category.
+   */
+
+  /* Huffman's basic algorithm to assign optimal code lengths to symbols */
+
+  for (;;) {
+    /* Find the smallest nonzero frequency, set c1 = its symbol */
+    /* In case of ties, take the larger symbol number */
+    c1 = -1;
+    v = 1000000000L;
+    for (i = 0; i <= 256; i++) {
+      if (freq[i] && freq[i] <= v) {
+	v = freq[i];
+	c1 = i;
+      }
+    }
+
+    /* Find the next smallest nonzero frequency, set c2 = its symbol */
+    /* In case of ties, take the larger symbol number */
+    c2 = -1;
+    v = 1000000000L;
+    for (i = 0; i <= 256; i++) {
+      if (freq[i] && freq[i] <= v && i != c1) {
+	v = freq[i];
+	c2 = i;
+      }
+    }
+
+    /* Done if we've merged everything into one frequency */
+    if (c2 < 0)
+      break;
+    
+    /* Else merge the two counts/trees */
+    freq[c1] += freq[c2];
+    freq[c2] = 0;
+
+    /* Increment the codesize of everything in c1's tree branch */
+    codesize[c1]++;
+    while (others[c1] >= 0) {
+      c1 = others[c1];
+      codesize[c1]++;
+    }
+    
+    others[c1] = c2;		/* chain c2 onto c1's tree branch */
+    
+    /* Increment the codesize of everything in c2's tree branch */
+    codesize[c2]++;
+    while (others[c2] >= 0) {
+      c2 = others[c2];
+      codesize[c2]++;
+    }
+  }
+
+  /* Now count the number of symbols of each code length */
+  for (i = 0; i <= 256; i++) {
+    if (codesize[i]) {
+      /* The JPEG standard seems to think that this can't happen, */
+      /* but I'm paranoid... */
+      if (codesize[i] > MAX_CLEN)
+	ERREXIT(cinfo, JERR_HUFF_CLEN_OVERFLOW);
+
+      bits[codesize[i]]++;
+    }
+  }
+
+  /* JPEG doesn't allow symbols with code lengths over 16 bits, so if the pure
+   * Huffman procedure assigned any such lengths, we must adjust the coding.
+   * Here is what the JPEG spec says about how this next bit works:
+   * Since symbols are paired for the longest Huffman code, the symbols are
+   * removed from this length category two at a time.  The prefix for the pair
+   * (which is one bit shorter) is allocated to one of the pair; then,
+   * skipping the BITS entry for that prefix length, a code word from the next
+   * shortest nonzero BITS entry is converted into a prefix for two code words
+   * one bit longer.
+   */
+  
+  for (i = MAX_CLEN; i > 16; i--) {
+    while (bits[i] > 0) {
+      j = i - 2;		/* find length of new prefix to be used */
+      while (bits[j] == 0)
+	j--;
+      
+      bits[i] -= 2;		/* remove two symbols */
+      bits[i-1]++;		/* one goes in this length */
+      bits[j+1] += 2;		/* two new symbols in this length */
+      bits[j]--;		/* symbol of this length is now a prefix */
+    }
+  }
+
+  /* Remove the count for the pseudo-symbol 256 from the largest codelength */
+  while (bits[i] == 0)		/* find largest codelength still in use */
+    i--;
+  bits[i]--;
+  
+  /* Return final symbol counts (only for lengths 0..16) */
+  MEMCOPY(htbl->bits, bits, SIZEOF(htbl->bits));
+  
+  /* Return a list of the symbols sorted by code length */
+  /* It's not real clear to me why we don't need to consider the codelength
+   * changes made above, but the JPEG spec seems to think this works.
+   */
+  p = 0;
+  for (i = 1; i <= MAX_CLEN; i++) {
+    for (j = 0; j <= 255; j++) {
+      if (codesize[j] == i) {
+	htbl->huffval[p] = (UINT8) j;
+	p++;
+      }
+    }
+  }
+
+  /* Set sent_table FALSE so updated table will be written to JPEG file. */
+  htbl->sent_table = FALSE;
+}
+
+
+/*
+ * Finish up a statistics-gathering pass and create the new Huffman tables.
+ */
+
+METHODDEF(void)
+finish_pass_gather (j_compress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int ci, dctbl, actbl;
+  jpeg_component_info * compptr;
+  JHUFF_TBL **htblptr;
+  boolean did_dc[NUM_HUFF_TBLS];
+  boolean did_ac[NUM_HUFF_TBLS];
+
+  /* It's important not to apply jpeg_gen_optimal_table more than once
+   * per table, because it clobbers the input frequency counts!
+   */
+  MEMZERO(did_dc, SIZEOF(did_dc));
+  MEMZERO(did_ac, SIZEOF(did_ac));
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    dctbl = compptr->dc_tbl_no;
+    actbl = compptr->ac_tbl_no;
+    if (! did_dc[dctbl]) {
+      htblptr = & cinfo->dc_huff_tbl_ptrs[dctbl];
+      if (*htblptr == NULL)
+	*htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+      jpeg_gen_optimal_table(cinfo, *htblptr, entropy->dc_count_ptrs[dctbl]);
+      did_dc[dctbl] = TRUE;
+    }
+    if (! did_ac[actbl]) {
+      htblptr = & cinfo->ac_huff_tbl_ptrs[actbl];
+      if (*htblptr == NULL)
+	*htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+      jpeg_gen_optimal_table(cinfo, *htblptr, entropy->ac_count_ptrs[actbl]);
+      did_ac[actbl] = TRUE;
+    }
+  }
+}
+
+
+#endif /* ENTROPY_OPT_SUPPORTED */
+
+
+/*
+ * Module initialization routine for Huffman entropy encoding.
+ */
+
+GLOBAL(void)
+jinit_huff_encoder (j_compress_ptr cinfo)
+{
+  huff_entropy_ptr entropy;
+  int i;
+
+  entropy = (huff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(huff_entropy_encoder));
+  cinfo->entropy = (struct jpeg_entropy_encoder *) entropy;
+  entropy->pub.start_pass = start_pass_huff;
+
+  /* Mark tables unallocated */
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    entropy->dc_derived_tbls[i] = entropy->ac_derived_tbls[i] = NULL;
+#ifdef ENTROPY_OPT_SUPPORTED
+    entropy->dc_count_ptrs[i] = entropy->ac_count_ptrs[i] = NULL;
+#endif
+  }
+}
diff --git a/libjpegtwrp/jchuff.h b/libjpegtwrp/jchuff.h
new file mode 100644
index 000000000..a9599fc1e
--- /dev/null
+++ b/libjpegtwrp/jchuff.h
@@ -0,0 +1,47 @@
+/*
+ * jchuff.h
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains declarations for Huffman entropy encoding routines
+ * that are shared between the sequential encoder (jchuff.c) and the
+ * progressive encoder (jcphuff.c).  No other modules need to see these.
+ */
+
+/* The legal range of a DCT coefficient is
+ *  -1024 .. +1023  for 8-bit data;
+ * -16384 .. +16383 for 12-bit data.
+ * Hence the magnitude should always fit in 10 or 14 bits respectively.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define MAX_COEF_BITS 10
+#else
+#define MAX_COEF_BITS 14
+#endif
+
+/* Derived data constructed for each Huffman table */
+
+typedef struct {
+  unsigned int ehufco[256];	/* code for each symbol */
+  char ehufsi[256];		/* length of code for each symbol */
+  /* If no code has been allocated for a symbol S, ehufsi[S] contains 0 */
+} c_derived_tbl;
+
+/* Short forms of external names for systems with brain-damaged linkers. */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jpeg_make_c_derived_tbl	jMkCDerived
+#define jpeg_gen_optimal_table	jGenOptTbl
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+/* Expand a Huffman table definition into the derived format */
+EXTERN(void) jpeg_make_c_derived_tbl
+	JPP((j_compress_ptr cinfo, boolean isDC, int tblno,
+	     c_derived_tbl ** pdtbl));
+
+/* Generate an optimal table definition given the specified counts */
+EXTERN(void) jpeg_gen_optimal_table
+	JPP((j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[]));
diff --git a/libjpegtwrp/jcinit.c b/libjpegtwrp/jcinit.c
new file mode 100644
index 000000000..5efffe331
--- /dev/null
+++ b/libjpegtwrp/jcinit.c
@@ -0,0 +1,72 @@
+/*
+ * jcinit.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains initialization logic for the JPEG compressor.
+ * This routine is in charge of selecting the modules to be executed and
+ * making an initialization call to each one.
+ *
+ * Logically, this code belongs in jcmaster.c.  It's split out because
+ * linking this routine implies linking the entire compression library.
+ * For a transcoding-only application, we want to be able to use jcmaster.c
+ * without linking in the whole library.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Master selection of compression modules.
+ * This is done once at the start of processing an image.  We determine
+ * which modules will be used and give them appropriate initialization calls.
+ */
+
+GLOBAL(void)
+jinit_compress_master (j_compress_ptr cinfo)
+{
+  /* Initialize master control (includes parameter checking/processing) */
+  jinit_c_master_control(cinfo, FALSE /* full compression */);
+
+  /* Preprocessing */
+  if (! cinfo->raw_data_in) {
+    jinit_color_converter(cinfo);
+    jinit_downsampler(cinfo);
+    jinit_c_prep_controller(cinfo, FALSE /* never need full buffer here */);
+  }
+  /* Forward DCT */
+  jinit_forward_dct(cinfo);
+  /* Entropy encoding: either Huffman or arithmetic coding. */
+  if (cinfo->arith_code) {
+    ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+  } else {
+    if (cinfo->progressive_mode) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+      jinit_phuff_encoder(cinfo);
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else
+      jinit_huff_encoder(cinfo);
+  }
+
+  /* Need a full-image coefficient buffer in any multi-pass mode. */
+  jinit_c_coef_controller(cinfo,
+		(boolean) (cinfo->num_scans > 1 || cinfo->optimize_coding));
+  jinit_c_main_controller(cinfo, FALSE /* never need full buffer here */);
+
+  jinit_marker_writer(cinfo);
+
+  /* We can now tell the memory manager to allocate virtual arrays. */
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr) cinfo);
+
+  /* Write the datastream header (SOI) immediately.
+   * Frame and scan headers are postponed till later.
+   * This lets application insert special markers after the SOI.
+   */
+  (*cinfo->marker->write_file_header) (cinfo);
+}
diff --git a/libjpegtwrp/jcmainct.c b/libjpegtwrp/jcmainct.c
new file mode 100644
index 000000000..5b7ff21dc
--- /dev/null
+++ b/libjpegtwrp/jcmainct.c
@@ -0,0 +1,293 @@
+/*
+ * jcmainct.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the main buffer controller for compression.
+ * The main buffer lies between the pre-processor and the JPEG
+ * compressor proper; it holds downsampled data in the JPEG colorspace.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Note: currently, there is no operating mode in which a full-image buffer
+ * is needed at this step.  If there were, that mode could not be used with
+ * "raw data" input, since this module is bypassed in that case.  However,
+ * we've left the code here for possible use in special applications.
+ */
+#undef FULL_MAIN_BUFFER_SUPPORTED
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_c_main_controller pub; /* public fields */
+
+  JDIMENSION cur_iMCU_row;	/* number of current iMCU row */
+  JDIMENSION rowgroup_ctr;	/* counts row groups received in iMCU row */
+  boolean suspended;		/* remember if we suspended output */
+  J_BUF_MODE pass_mode;		/* current operating mode */
+
+  /* If using just a strip buffer, this points to the entire set of buffers
+   * (we allocate one for each component).  In the full-image case, this
+   * points to the currently accessible strips of the virtual arrays.
+   */
+  JSAMPARRAY buffer[MAX_COMPONENTS];
+
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+  /* If using full-image storage, this array holds pointers to virtual-array
+   * control blocks for each component.  Unused if not full-image storage.
+   */
+  jvirt_sarray_ptr whole_image[MAX_COMPONENTS];
+#endif
+} my_main_controller;
+
+typedef my_main_controller * my_main_ptr;
+
+
+/* Forward declarations */
+METHODDEF(void) process_data_simple_main
+	JPP((j_compress_ptr cinfo, JSAMPARRAY input_buf,
+	     JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail));
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+METHODDEF(void) process_data_buffer_main
+	JPP((j_compress_ptr cinfo, JSAMPARRAY input_buf,
+	     JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail));
+#endif
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_main (j_compress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+
+  /* Do nothing in raw-data mode. */
+  if (cinfo->raw_data_in)
+    return;
+
+  main_ptr->cur_iMCU_row = 0;	/* initialize counters */
+  main_ptr->rowgroup_ctr = 0;
+  main_ptr->suspended = FALSE;
+  main_ptr->pass_mode = pass_mode;	/* save mode for use by process_data */
+
+  switch (pass_mode) {
+  case JBUF_PASS_THRU:
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+    if (main_ptr->whole_image[0] != NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+#endif
+    main_ptr->pub.process_data = process_data_simple_main;
+    break;
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+  case JBUF_SAVE_SOURCE:
+  case JBUF_CRANK_DEST:
+  case JBUF_SAVE_AND_PASS:
+    if (main_ptr->whole_image[0] == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    main_ptr->pub.process_data = process_data_buffer_main;
+    break;
+#endif
+  default:
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    break;
+  }
+}
+
+
+/*
+ * Process some data.
+ * This routine handles the simple pass-through mode,
+ * where we have only a strip buffer.
+ */
+
+METHODDEF(void)
+process_data_simple_main (j_compress_ptr cinfo,
+			  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+			  JDIMENSION in_rows_avail)
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+
+  while (main_ptr->cur_iMCU_row < cinfo->total_iMCU_rows) {
+    /* Read input data if we haven't filled the main buffer yet */
+    if (main_ptr->rowgroup_ctr < DCTSIZE)
+      (*cinfo->prep->pre_process_data) (cinfo,
+					input_buf, in_row_ctr, in_rows_avail,
+					main_ptr->buffer, &main_ptr->rowgroup_ctr,
+					(JDIMENSION) DCTSIZE);
+
+    /* If we don't have a full iMCU row buffered, return to application for
+     * more data.  Note that preprocessor will always pad to fill the iMCU row
+     * at the bottom of the image.
+     */
+    if (main_ptr->rowgroup_ctr != DCTSIZE)
+      return;
+
+    /* Send the completed row to the compressor */
+    if (! (*cinfo->coef->compress_data) (cinfo, main_ptr->buffer)) {
+      /* If compressor did not consume the whole row, then we must need to
+       * suspend processing and return to the application.  In this situation
+       * we pretend we didn't yet consume the last input row; otherwise, if
+       * it happened to be the last row of the image, the application would
+       * think we were done.
+       */
+      if (! main_ptr->suspended) {
+	(*in_row_ctr)--;
+	main_ptr->suspended = TRUE;
+      }
+      return;
+    }
+    /* We did finish the row.  Undo our little suspension hack if a previous
+     * call suspended; then mark the main buffer empty.
+     */
+    if (main_ptr->suspended) {
+      (*in_row_ctr)++;
+      main_ptr->suspended = FALSE;
+    }
+    main_ptr->rowgroup_ctr = 0;
+    main_ptr->cur_iMCU_row++;
+  }
+}
+
+
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+
+/*
+ * Process some data.
+ * This routine handles all of the modes that use a full-size buffer.
+ */
+
+METHODDEF(void)
+process_data_buffer_main (j_compress_ptr cinfo,
+			  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+			  JDIMENSION in_rows_avail)
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+  int ci;
+  jpeg_component_info *compptr;
+  boolean writing = (main_ptr->pass_mode != JBUF_CRANK_DEST);
+
+  while (main_ptr->cur_iMCU_row < cinfo->total_iMCU_rows) {
+    /* Realign the virtual buffers if at the start of an iMCU row. */
+    if (main_ptr->rowgroup_ctr == 0) {
+      for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	   ci++, compptr++) {
+	main_ptr->buffer[ci] = (*cinfo->mem->access_virt_sarray)
+	  ((j_common_ptr) cinfo, main_ptr->whole_image[ci],
+	   main_ptr->cur_iMCU_row * (compptr->v_samp_factor * DCTSIZE),
+	   (JDIMENSION) (compptr->v_samp_factor * DCTSIZE), writing);
+      }
+      /* In a read pass, pretend we just read some source data. */
+      if (! writing) {
+	*in_row_ctr += cinfo->max_v_samp_factor * DCTSIZE;
+	main_ptr->rowgroup_ctr = DCTSIZE;
+      }
+    }
+
+    /* If a write pass, read input data until the current iMCU row is full. */
+    /* Note: preprocessor will pad if necessary to fill the last iMCU row. */
+    if (writing) {
+      (*cinfo->prep->pre_process_data) (cinfo,
+					input_buf, in_row_ctr, in_rows_avail,
+					main_ptr->buffer, &main_ptr->rowgroup_ctr,
+					(JDIMENSION) DCTSIZE);
+      /* Return to application if we need more data to fill the iMCU row. */
+      if (main_ptr->rowgroup_ctr < DCTSIZE)
+	return;
+    }
+
+    /* Emit data, unless this is a sink-only pass. */
+    if (main_ptr->pass_mode != JBUF_SAVE_SOURCE) {
+      if (! (*cinfo->coef->compress_data) (cinfo, main_ptr->buffer)) {
+	/* If compressor did not consume the whole row, then we must need to
+	 * suspend processing and return to the application.  In this situation
+	 * we pretend we didn't yet consume the last input row; otherwise, if
+	 * it happened to be the last row of the image, the application would
+	 * think we were done.
+	 */
+	if (! main_ptr->suspended) {
+	  (*in_row_ctr)--;
+	  main_ptr->suspended = TRUE;
+	}
+	return;
+      }
+      /* We did finish the row.  Undo our little suspension hack if a previous
+       * call suspended; then mark the main buffer empty.
+       */
+      if (main_ptr->suspended) {
+	(*in_row_ctr)++;
+	main_ptr->suspended = FALSE;
+      }
+    }
+
+    /* If get here, we are done with this iMCU row.  Mark buffer empty. */
+    main_ptr->rowgroup_ctr = 0;
+    main_ptr->cur_iMCU_row++;
+  }
+}
+
+#endif /* FULL_MAIN_BUFFER_SUPPORTED */
+
+
+/*
+ * Initialize main buffer controller.
+ */
+
+GLOBAL(void)
+jinit_c_main_controller (j_compress_ptr cinfo, boolean need_full_buffer)
+{
+  my_main_ptr main_ptr;
+  int ci;
+  jpeg_component_info *compptr;
+
+  main_ptr = (my_main_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_main_controller));
+  cinfo->main = (struct jpeg_c_main_controller *) main_ptr;
+  main_ptr->pub.start_pass = start_pass_main;
+
+  /* We don't need to create a buffer in raw-data mode. */
+  if (cinfo->raw_data_in)
+    return;
+
+  /* Create the buffer.  It holds downsampled data, so each component
+   * may be of a different size.
+   */
+  if (need_full_buffer) {
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+    /* Allocate a full-image virtual array for each component */
+    /* Note we pad the bottom to a multiple of the iMCU height */
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      main_ptr->whole_image[ci] = (*cinfo->mem->request_virt_sarray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+	 compptr->width_in_blocks * DCTSIZE,
+	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
+				(long) compptr->v_samp_factor) * DCTSIZE,
+	 (JDIMENSION) (compptr->v_samp_factor * DCTSIZE));
+    }
+#else
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+#endif
+  } else {
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+    main_ptr->whole_image[0] = NULL; /* flag for no virtual arrays */
+#endif
+    /* Allocate a strip buffer for each component */
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      main_ptr->buffer[ci] = (*cinfo->mem->alloc_sarray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE,
+	 compptr->width_in_blocks * DCTSIZE,
+	 (JDIMENSION) (compptr->v_samp_factor * DCTSIZE));
+    }
+  }
+}
diff --git a/libjpegtwrp/jcmarker.c b/libjpegtwrp/jcmarker.c
new file mode 100644
index 000000000..3d1e6c6d5
--- /dev/null
+++ b/libjpegtwrp/jcmarker.c
@@ -0,0 +1,664 @@
+/*
+ * jcmarker.c
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to write JPEG datastream markers.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+typedef enum {			/* JPEG marker codes */
+  M_SOF0  = 0xc0,
+  M_SOF1  = 0xc1,
+  M_SOF2  = 0xc2,
+  M_SOF3  = 0xc3,
+  
+  M_SOF5  = 0xc5,
+  M_SOF6  = 0xc6,
+  M_SOF7  = 0xc7,
+  
+  M_JPG   = 0xc8,
+  M_SOF9  = 0xc9,
+  M_SOF10 = 0xca,
+  M_SOF11 = 0xcb,
+  
+  M_SOF13 = 0xcd,
+  M_SOF14 = 0xce,
+  M_SOF15 = 0xcf,
+  
+  M_DHT   = 0xc4,
+  
+  M_DAC   = 0xcc,
+  
+  M_RST0  = 0xd0,
+  M_RST1  = 0xd1,
+  M_RST2  = 0xd2,
+  M_RST3  = 0xd3,
+  M_RST4  = 0xd4,
+  M_RST5  = 0xd5,
+  M_RST6  = 0xd6,
+  M_RST7  = 0xd7,
+  
+  M_SOI   = 0xd8,
+  M_EOI   = 0xd9,
+  M_SOS   = 0xda,
+  M_DQT   = 0xdb,
+  M_DNL   = 0xdc,
+  M_DRI   = 0xdd,
+  M_DHP   = 0xde,
+  M_EXP   = 0xdf,
+  
+  M_APP0  = 0xe0,
+  M_APP1  = 0xe1,
+  M_APP2  = 0xe2,
+  M_APP3  = 0xe3,
+  M_APP4  = 0xe4,
+  M_APP5  = 0xe5,
+  M_APP6  = 0xe6,
+  M_APP7  = 0xe7,
+  M_APP8  = 0xe8,
+  M_APP9  = 0xe9,
+  M_APP10 = 0xea,
+  M_APP11 = 0xeb,
+  M_APP12 = 0xec,
+  M_APP13 = 0xed,
+  M_APP14 = 0xee,
+  M_APP15 = 0xef,
+  
+  M_JPG0  = 0xf0,
+  M_JPG13 = 0xfd,
+  M_COM   = 0xfe,
+  
+  M_TEM   = 0x01,
+  
+  M_ERROR = 0x100
+} JPEG_MARKER;
+
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_marker_writer pub; /* public fields */
+
+  unsigned int last_restart_interval; /* last DRI value emitted; 0 after SOI */
+} my_marker_writer;
+
+typedef my_marker_writer * my_marker_ptr;
+
+
+/*
+ * Basic output routines.
+ *
+ * Note that we do not support suspension while writing a marker.
+ * Therefore, an application using suspension must ensure that there is
+ * enough buffer space for the initial markers (typ. 600-700 bytes) before
+ * calling jpeg_start_compress, and enough space to write the trailing EOI
+ * (a few bytes) before calling jpeg_finish_compress.  Multipass compression
+ * modes are not supported at all with suspension, so those two are the only
+ * points where markers will be written.
+ */
+
+LOCAL(void)
+emit_byte (j_compress_ptr cinfo, int val)
+/* Emit a byte */
+{
+  struct jpeg_destination_mgr * dest = cinfo->dest;
+
+  *(dest->next_output_byte)++ = (JOCTET) val;
+  if (--dest->free_in_buffer == 0) {
+    if (! (*dest->empty_output_buffer) (cinfo))
+      ERREXIT(cinfo, JERR_CANT_SUSPEND);
+  }
+}
+
+
+LOCAL(void)
+emit_marker (j_compress_ptr cinfo, JPEG_MARKER mark)
+/* Emit a marker code */
+{
+  emit_byte(cinfo, 0xFF);
+  emit_byte(cinfo, (int) mark);
+}
+
+
+LOCAL(void)
+emit_2bytes (j_compress_ptr cinfo, int value)
+/* Emit a 2-byte integer; these are always MSB first in JPEG files */
+{
+  emit_byte(cinfo, (value >> 8) & 0xFF);
+  emit_byte(cinfo, value & 0xFF);
+}
+
+
+/*
+ * Routines to write specific marker types.
+ */
+
+LOCAL(int)
+emit_dqt (j_compress_ptr cinfo, int index)
+/* Emit a DQT marker */
+/* Returns the precision used (0 = 8bits, 1 = 16bits) for baseline checking */
+{
+  JQUANT_TBL * qtbl = cinfo->quant_tbl_ptrs[index];
+  int prec;
+  int i;
+
+  if (qtbl == NULL)
+    ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, index);
+
+  prec = 0;
+  for (i = 0; i < DCTSIZE2; i++) {
+    if (qtbl->quantval[i] > 255)
+      prec = 1;
+  }
+
+  if (! qtbl->sent_table) {
+    emit_marker(cinfo, M_DQT);
+
+    emit_2bytes(cinfo, prec ? DCTSIZE2*2 + 1 + 2 : DCTSIZE2 + 1 + 2);
+
+    emit_byte(cinfo, index + (prec<<4));
+
+    for (i = 0; i < DCTSIZE2; i++) {
+      /* The table entries must be emitted in zigzag order. */
+      unsigned int qval = qtbl->quantval[jpeg_natural_order[i]];
+      if (prec)
+	emit_byte(cinfo, (int) (qval >> 8));
+      emit_byte(cinfo, (int) (qval & 0xFF));
+    }
+
+    qtbl->sent_table = TRUE;
+  }
+
+  return prec;
+}
+
+
+LOCAL(void)
+emit_dht (j_compress_ptr cinfo, int index, boolean is_ac)
+/* Emit a DHT marker */
+{
+  JHUFF_TBL * htbl;
+  int length, i;
+  
+  if (is_ac) {
+    htbl = cinfo->ac_huff_tbl_ptrs[index];
+    index += 0x10;		/* output index has AC bit set */
+  } else {
+    htbl = cinfo->dc_huff_tbl_ptrs[index];
+  }
+
+  if (htbl == NULL)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, index);
+  
+  if (! htbl->sent_table) {
+    emit_marker(cinfo, M_DHT);
+    
+    length = 0;
+    for (i = 1; i <= 16; i++)
+      length += htbl->bits[i];
+    
+    emit_2bytes(cinfo, length + 2 + 1 + 16);
+    emit_byte(cinfo, index);
+    
+    for (i = 1; i <= 16; i++)
+      emit_byte(cinfo, htbl->bits[i]);
+    
+    for (i = 0; i < length; i++)
+      emit_byte(cinfo, htbl->huffval[i]);
+    
+    htbl->sent_table = TRUE;
+  }
+}
+
+
+LOCAL(void)
+emit_dac (j_compress_ptr cinfo)
+/* Emit a DAC marker */
+/* Since the useful info is so small, we want to emit all the tables in */
+/* one DAC marker.  Therefore this routine does its own scan of the table. */
+{
+#ifdef C_ARITH_CODING_SUPPORTED
+  char dc_in_use[NUM_ARITH_TBLS];
+  char ac_in_use[NUM_ARITH_TBLS];
+  int length, i;
+  jpeg_component_info *compptr;
+  
+  for (i = 0; i < NUM_ARITH_TBLS; i++)
+    dc_in_use[i] = ac_in_use[i] = 0;
+  
+  for (i = 0; i < cinfo->comps_in_scan; i++) {
+    compptr = cinfo->cur_comp_info[i];
+    dc_in_use[compptr->dc_tbl_no] = 1;
+    ac_in_use[compptr->ac_tbl_no] = 1;
+  }
+  
+  length = 0;
+  for (i = 0; i < NUM_ARITH_TBLS; i++)
+    length += dc_in_use[i] + ac_in_use[i];
+  
+  emit_marker(cinfo, M_DAC);
+  
+  emit_2bytes(cinfo, length*2 + 2);
+  
+  for (i = 0; i < NUM_ARITH_TBLS; i++) {
+    if (dc_in_use[i]) {
+      emit_byte(cinfo, i);
+      emit_byte(cinfo, cinfo->arith_dc_L[i] + (cinfo->arith_dc_U[i]<<4));
+    }
+    if (ac_in_use[i]) {
+      emit_byte(cinfo, i + 0x10);
+      emit_byte(cinfo, cinfo->arith_ac_K[i]);
+    }
+  }
+#endif /* C_ARITH_CODING_SUPPORTED */
+}
+
+
+LOCAL(void)
+emit_dri (j_compress_ptr cinfo)
+/* Emit a DRI marker */
+{
+  emit_marker(cinfo, M_DRI);
+  
+  emit_2bytes(cinfo, 4);	/* fixed length */
+
+  emit_2bytes(cinfo, (int) cinfo->restart_interval);
+}
+
+
+LOCAL(void)
+emit_sof (j_compress_ptr cinfo, JPEG_MARKER code)
+/* Emit a SOF marker */
+{
+  int ci;
+  jpeg_component_info *compptr;
+  
+  emit_marker(cinfo, code);
+  
+  emit_2bytes(cinfo, 3 * cinfo->num_components + 2 + 5 + 1); /* length */
+
+  /* Make sure image isn't bigger than SOF field can handle */
+  if ((long) cinfo->image_height > 65535L ||
+      (long) cinfo->image_width > 65535L)
+    ERREXIT1(cinfo, JERR_IMAGE_TOO_BIG, (unsigned int) 65535);
+
+  emit_byte(cinfo, cinfo->data_precision);
+  emit_2bytes(cinfo, (int) cinfo->image_height);
+  emit_2bytes(cinfo, (int) cinfo->image_width);
+
+  emit_byte(cinfo, cinfo->num_components);
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    emit_byte(cinfo, compptr->component_id);
+    emit_byte(cinfo, (compptr->h_samp_factor << 4) + compptr->v_samp_factor);
+    emit_byte(cinfo, compptr->quant_tbl_no);
+  }
+}
+
+
+LOCAL(void)
+emit_sos (j_compress_ptr cinfo)
+/* Emit a SOS marker */
+{
+  int i, td, ta;
+  jpeg_component_info *compptr;
+  
+  emit_marker(cinfo, M_SOS);
+  
+  emit_2bytes(cinfo, 2 * cinfo->comps_in_scan + 2 + 1 + 3); /* length */
+  
+  emit_byte(cinfo, cinfo->comps_in_scan);
+  
+  for (i = 0; i < cinfo->comps_in_scan; i++) {
+    compptr = cinfo->cur_comp_info[i];
+    emit_byte(cinfo, compptr->component_id);
+    td = compptr->dc_tbl_no;
+    ta = compptr->ac_tbl_no;
+    if (cinfo->progressive_mode) {
+      /* Progressive mode: only DC or only AC tables are used in one scan;
+       * furthermore, Huffman coding of DC refinement uses no table at all.
+       * We emit 0 for unused field(s); this is recommended by the P&M text
+       * but does not seem to be specified in the standard.
+       */
+      if (cinfo->Ss == 0) {
+	ta = 0;			/* DC scan */
+	if (cinfo->Ah != 0 && !cinfo->arith_code)
+	  td = 0;		/* no DC table either */
+      } else {
+	td = 0;			/* AC scan */
+      }
+    }
+    emit_byte(cinfo, (td << 4) + ta);
+  }
+
+  emit_byte(cinfo, cinfo->Ss);
+  emit_byte(cinfo, cinfo->Se);
+  emit_byte(cinfo, (cinfo->Ah << 4) + cinfo->Al);
+}
+
+
+LOCAL(void)
+emit_jfif_app0 (j_compress_ptr cinfo)
+/* Emit a JFIF-compliant APP0 marker */
+{
+  /*
+   * Length of APP0 block	(2 bytes)
+   * Block ID			(4 bytes - ASCII "JFIF")
+   * Zero byte			(1 byte to terminate the ID string)
+   * Version Major, Minor	(2 bytes - major first)
+   * Units			(1 byte - 0x00 = none, 0x01 = inch, 0x02 = cm)
+   * Xdpu			(2 bytes - dots per unit horizontal)
+   * Ydpu			(2 bytes - dots per unit vertical)
+   * Thumbnail X size		(1 byte)
+   * Thumbnail Y size		(1 byte)
+   */
+  
+  emit_marker(cinfo, M_APP0);
+  
+  emit_2bytes(cinfo, 2 + 4 + 1 + 2 + 1 + 2 + 2 + 1 + 1); /* length */
+
+  emit_byte(cinfo, 0x4A);	/* Identifier: ASCII "JFIF" */
+  emit_byte(cinfo, 0x46);
+  emit_byte(cinfo, 0x49);
+  emit_byte(cinfo, 0x46);
+  emit_byte(cinfo, 0);
+  emit_byte(cinfo, cinfo->JFIF_major_version); /* Version fields */
+  emit_byte(cinfo, cinfo->JFIF_minor_version);
+  emit_byte(cinfo, cinfo->density_unit); /* Pixel size information */
+  emit_2bytes(cinfo, (int) cinfo->X_density);
+  emit_2bytes(cinfo, (int) cinfo->Y_density);
+  emit_byte(cinfo, 0);		/* No thumbnail image */
+  emit_byte(cinfo, 0);
+}
+
+
+LOCAL(void)
+emit_adobe_app14 (j_compress_ptr cinfo)
+/* Emit an Adobe APP14 marker */
+{
+  /*
+   * Length of APP14 block	(2 bytes)
+   * Block ID			(5 bytes - ASCII "Adobe")
+   * Version Number		(2 bytes - currently 100)
+   * Flags0			(2 bytes - currently 0)
+   * Flags1			(2 bytes - currently 0)
+   * Color transform		(1 byte)
+   *
+   * Although Adobe TN 5116 mentions Version = 101, all the Adobe files
+   * now in circulation seem to use Version = 100, so that's what we write.
+   *
+   * We write the color transform byte as 1 if the JPEG color space is
+   * YCbCr, 2 if it's YCCK, 0 otherwise.  Adobe's definition has to do with
+   * whether the encoder performed a transformation, which is pretty useless.
+   */
+  
+  emit_marker(cinfo, M_APP14);
+  
+  emit_2bytes(cinfo, 2 + 5 + 2 + 2 + 2 + 1); /* length */
+
+  emit_byte(cinfo, 0x41);	/* Identifier: ASCII "Adobe" */
+  emit_byte(cinfo, 0x64);
+  emit_byte(cinfo, 0x6F);
+  emit_byte(cinfo, 0x62);
+  emit_byte(cinfo, 0x65);
+  emit_2bytes(cinfo, 100);	/* Version */
+  emit_2bytes(cinfo, 0);	/* Flags0 */
+  emit_2bytes(cinfo, 0);	/* Flags1 */
+  switch (cinfo->jpeg_color_space) {
+  case JCS_YCbCr:
+    emit_byte(cinfo, 1);	/* Color transform = 1 */
+    break;
+  case JCS_YCCK:
+    emit_byte(cinfo, 2);	/* Color transform = 2 */
+    break;
+  default:
+    emit_byte(cinfo, 0);	/* Color transform = 0 */
+    break;
+  }
+}
+
+
+/*
+ * These routines allow writing an arbitrary marker with parameters.
+ * The only intended use is to emit COM or APPn markers after calling
+ * write_file_header and before calling write_frame_header.
+ * Other uses are not guaranteed to produce desirable results.
+ * Counting the parameter bytes properly is the caller's responsibility.
+ */
+
+METHODDEF(void)
+write_marker_header (j_compress_ptr cinfo, int marker, unsigned int datalen)
+/* Emit an arbitrary marker header */
+{
+  if (datalen > (unsigned int) 65533)		/* safety check */
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  emit_marker(cinfo, (JPEG_MARKER) marker);
+
+  emit_2bytes(cinfo, (int) (datalen + 2));	/* total length */
+}
+
+METHODDEF(void)
+write_marker_byte (j_compress_ptr cinfo, int val)
+/* Emit one byte of marker parameters following write_marker_header */
+{
+  emit_byte(cinfo, val);
+}
+
+
+/*
+ * Write datastream header.
+ * This consists of an SOI and optional APPn markers.
+ * We recommend use of the JFIF marker, but not the Adobe marker,
+ * when using YCbCr or grayscale data.  The JFIF marker should NOT
+ * be used for any other JPEG colorspace.  The Adobe marker is helpful
+ * to distinguish RGB, CMYK, and YCCK colorspaces.
+ * Note that an application can write additional header markers after
+ * jpeg_start_compress returns.
+ */
+
+METHODDEF(void)
+write_file_header (j_compress_ptr cinfo)
+{
+  my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
+
+  emit_marker(cinfo, M_SOI);	/* first the SOI */
+
+  /* SOI is defined to reset restart interval to 0 */
+  marker->last_restart_interval = 0;
+
+  if (cinfo->write_JFIF_header)	/* next an optional JFIF APP0 */
+    emit_jfif_app0(cinfo);
+  if (cinfo->write_Adobe_marker) /* next an optional Adobe APP14 */
+    emit_adobe_app14(cinfo);
+}
+
+
+/*
+ * Write frame header.
+ * This consists of DQT and SOFn markers.
+ * Note that we do not emit the SOF until we have emitted the DQT(s).
+ * This avoids compatibility problems with incorrect implementations that
+ * try to error-check the quant table numbers as soon as they see the SOF.
+ */
+
+METHODDEF(void)
+write_frame_header (j_compress_ptr cinfo)
+{
+  int ci, prec;
+  boolean is_baseline;
+  jpeg_component_info *compptr;
+  
+  /* Emit DQT for each quantization table.
+   * Note that emit_dqt() suppresses any duplicate tables.
+   */
+  prec = 0;
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    prec += emit_dqt(cinfo, compptr->quant_tbl_no);
+  }
+  /* now prec is nonzero iff there are any 16-bit quant tables. */
+
+  /* Check for a non-baseline specification.
+   * Note we assume that Huffman table numbers won't be changed later.
+   */
+  if (cinfo->arith_code || cinfo->progressive_mode ||
+      cinfo->data_precision != 8) {
+    is_baseline = FALSE;
+  } else {
+    is_baseline = TRUE;
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      if (compptr->dc_tbl_no > 1 || compptr->ac_tbl_no > 1)
+	is_baseline = FALSE;
+    }
+    if (prec && is_baseline) {
+      is_baseline = FALSE;
+      /* If it's baseline except for quantizer size, warn the user */
+      TRACEMS(cinfo, 0, JTRC_16BIT_TABLES);
+    }
+  }
+
+  /* Emit the proper SOF marker */
+  if (cinfo->arith_code) {
+    emit_sof(cinfo, M_SOF9);	/* SOF code for arithmetic coding */
+  } else {
+    if (cinfo->progressive_mode)
+      emit_sof(cinfo, M_SOF2);	/* SOF code for progressive Huffman */
+    else if (is_baseline)
+      emit_sof(cinfo, M_SOF0);	/* SOF code for baseline implementation */
+    else
+      emit_sof(cinfo, M_SOF1);	/* SOF code for non-baseline Huffman file */
+  }
+}
+
+
+/*
+ * Write scan header.
+ * This consists of DHT or DAC markers, optional DRI, and SOS.
+ * Compressed data will be written following the SOS.
+ */
+
+METHODDEF(void)
+write_scan_header (j_compress_ptr cinfo)
+{
+  my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
+  int i;
+  jpeg_component_info *compptr;
+
+  if (cinfo->arith_code) {
+    /* Emit arith conditioning info.  We may have some duplication
+     * if the file has multiple scans, but it's so small it's hardly
+     * worth worrying about.
+     */
+    emit_dac(cinfo);
+  } else {
+    /* Emit Huffman tables.
+     * Note that emit_dht() suppresses any duplicate tables.
+     */
+    for (i = 0; i < cinfo->comps_in_scan; i++) {
+      compptr = cinfo->cur_comp_info[i];
+      if (cinfo->progressive_mode) {
+	/* Progressive mode: only DC or only AC tables are used in one scan */
+	if (cinfo->Ss == 0) {
+	  if (cinfo->Ah == 0)	/* DC needs no table for refinement scan */
+	    emit_dht(cinfo, compptr->dc_tbl_no, FALSE);
+	} else {
+	  emit_dht(cinfo, compptr->ac_tbl_no, TRUE);
+	}
+      } else {
+	/* Sequential mode: need both DC and AC tables */
+	emit_dht(cinfo, compptr->dc_tbl_no, FALSE);
+	emit_dht(cinfo, compptr->ac_tbl_no, TRUE);
+      }
+    }
+  }
+
+  /* Emit DRI if required --- note that DRI value could change for each scan.
+   * We avoid wasting space with unnecessary DRIs, however.
+   */
+  if (cinfo->restart_interval != marker->last_restart_interval) {
+    emit_dri(cinfo);
+    marker->last_restart_interval = cinfo->restart_interval;
+  }
+
+  emit_sos(cinfo);
+}
+
+
+/*
+ * Write datastream trailer.
+ */
+
+METHODDEF(void)
+write_file_trailer (j_compress_ptr cinfo)
+{
+  emit_marker(cinfo, M_EOI);
+}
+
+
+/*
+ * Write an abbreviated table-specification datastream.
+ * This consists of SOI, DQT and DHT tables, and EOI.
+ * Any table that is defined and not marked sent_table = TRUE will be
+ * emitted.  Note that all tables will be marked sent_table = TRUE at exit.
+ */
+
+METHODDEF(void)
+write_tables_only (j_compress_ptr cinfo)
+{
+  int i;
+
+  emit_marker(cinfo, M_SOI);
+
+  for (i = 0; i < NUM_QUANT_TBLS; i++) {
+    if (cinfo->quant_tbl_ptrs[i] != NULL)
+      (void) emit_dqt(cinfo, i);
+  }
+
+  if (! cinfo->arith_code) {
+    for (i = 0; i < NUM_HUFF_TBLS; i++) {
+      if (cinfo->dc_huff_tbl_ptrs[i] != NULL)
+	emit_dht(cinfo, i, FALSE);
+      if (cinfo->ac_huff_tbl_ptrs[i] != NULL)
+	emit_dht(cinfo, i, TRUE);
+    }
+  }
+
+  emit_marker(cinfo, M_EOI);
+}
+
+
+/*
+ * Initialize the marker writer module.
+ */
+
+GLOBAL(void)
+jinit_marker_writer (j_compress_ptr cinfo)
+{
+  my_marker_ptr marker;
+
+  /* Create the subobject */
+  marker = (my_marker_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_marker_writer));
+  cinfo->marker = (struct jpeg_marker_writer *) marker;
+  /* Initialize method pointers */
+  marker->pub.write_file_header = write_file_header;
+  marker->pub.write_frame_header = write_frame_header;
+  marker->pub.write_scan_header = write_scan_header;
+  marker->pub.write_file_trailer = write_file_trailer;
+  marker->pub.write_tables_only = write_tables_only;
+  marker->pub.write_marker_header = write_marker_header;
+  marker->pub.write_marker_byte = write_marker_byte;
+  /* Initialize private state */
+  marker->last_restart_interval = 0;
+}
diff --git a/libjpegtwrp/jcmaster.c b/libjpegtwrp/jcmaster.c
new file mode 100644
index 000000000..aab4020b8
--- /dev/null
+++ b/libjpegtwrp/jcmaster.c
@@ -0,0 +1,590 @@
+/*
+ * jcmaster.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains master control logic for the JPEG compressor.
+ * These routines are concerned with parameter validation, initial setup,
+ * and inter-pass control (determining the number of passes and the work 
+ * to be done in each pass).
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Private state */
+
+typedef enum {
+	main_pass,		/* input data, also do first output step */
+	huff_opt_pass,		/* Huffman code optimization pass */
+	output_pass		/* data output pass */
+} c_pass_type;
+
+typedef struct {
+  struct jpeg_comp_master pub;	/* public fields */
+
+  c_pass_type pass_type;	/* the type of the current pass */
+
+  int pass_number;		/* # of passes completed */
+  int total_passes;		/* total # of passes needed */
+
+  int scan_number;		/* current index in scan_info[] */
+} my_comp_master;
+
+typedef my_comp_master * my_master_ptr;
+
+
+/*
+ * Support routines that do various essential calculations.
+ */
+
+LOCAL(void)
+initial_setup (j_compress_ptr cinfo)
+/* Do computations that are needed before master selection phase */
+{
+  int ci;
+  jpeg_component_info *compptr;
+  long samplesperrow;
+  JDIMENSION jd_samplesperrow;
+
+  /* Sanity check on image dimensions */
+  if (cinfo->image_height <= 0 || cinfo->image_width <= 0
+      || cinfo->num_components <= 0 || cinfo->input_components <= 0)
+    ERREXIT(cinfo, JERR_EMPTY_IMAGE);
+
+  /* Make sure image isn't bigger than I can handle */
+  if ((long) cinfo->image_height > (long) JPEG_MAX_DIMENSION ||
+      (long) cinfo->image_width > (long) JPEG_MAX_DIMENSION)
+    ERREXIT1(cinfo, JERR_IMAGE_TOO_BIG, (unsigned int) JPEG_MAX_DIMENSION);
+
+  /* Width of an input scanline must be representable as JDIMENSION. */
+  samplesperrow = (long) cinfo->image_width * (long) cinfo->input_components;
+  jd_samplesperrow = (JDIMENSION) samplesperrow;
+  if ((long) jd_samplesperrow != samplesperrow)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+
+  /* For now, precision must match compiled-in value... */
+  if (cinfo->data_precision != BITS_IN_JSAMPLE)
+    ERREXIT1(cinfo, JERR_BAD_PRECISION, cinfo->data_precision);
+
+  /* Check that number of components won't exceed internal array sizes */
+  if (cinfo->num_components > MAX_COMPONENTS)
+    ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
+	     MAX_COMPONENTS);
+
+  /* Compute maximum sampling factors; check factor validity */
+  cinfo->max_h_samp_factor = 1;
+  cinfo->max_v_samp_factor = 1;
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    if (compptr->h_samp_factor<=0 || compptr->h_samp_factor>MAX_SAMP_FACTOR ||
+	compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR)
+      ERREXIT(cinfo, JERR_BAD_SAMPLING);
+    cinfo->max_h_samp_factor = MAX(cinfo->max_h_samp_factor,
+				   compptr->h_samp_factor);
+    cinfo->max_v_samp_factor = MAX(cinfo->max_v_samp_factor,
+				   compptr->v_samp_factor);
+  }
+
+  /* Compute dimensions of components */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Fill in the correct component_index value; don't rely on application */
+    compptr->component_index = ci;
+    /* For compression, we never do DCT scaling. */
+    compptr->DCT_scaled_size = DCTSIZE;
+    /* Size in DCT blocks */
+    compptr->width_in_blocks = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor,
+		    (long) (cinfo->max_h_samp_factor * DCTSIZE));
+    compptr->height_in_blocks = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor,
+		    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+    /* Size in samples */
+    compptr->downsampled_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor,
+		    (long) cinfo->max_h_samp_factor);
+    compptr->downsampled_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor,
+		    (long) cinfo->max_v_samp_factor);
+    /* Mark component needed (this flag isn't actually used for compression) */
+    compptr->component_needed = TRUE;
+  }
+
+  /* Compute number of fully interleaved MCU rows (number of times that
+   * main controller will call coefficient controller).
+   */
+  cinfo->total_iMCU_rows = (JDIMENSION)
+    jdiv_round_up((long) cinfo->image_height,
+		  (long) (cinfo->max_v_samp_factor*DCTSIZE));
+}
+
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+
+LOCAL(void)
+validate_script (j_compress_ptr cinfo)
+/* Verify that the scan script in cinfo->scan_info[] is valid; also
+ * determine whether it uses progressive JPEG, and set cinfo->progressive_mode.
+ */
+{
+  const jpeg_scan_info * scanptr;
+  int scanno, ncomps, ci, coefi, thisi;
+  int Ss, Se, Ah, Al;
+  boolean component_sent[MAX_COMPONENTS];
+#ifdef C_PROGRESSIVE_SUPPORTED
+  int * last_bitpos_ptr;
+  int last_bitpos[MAX_COMPONENTS][DCTSIZE2];
+  /* -1 until that coefficient has been seen; then last Al for it */
+#endif
+
+  if (cinfo->num_scans <= 0)
+    ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, 0);
+
+  /* For sequential JPEG, all scans must have Ss=0, Se=DCTSIZE2-1;
+   * for progressive JPEG, no scan can have this.
+   */
+  scanptr = cinfo->scan_info;
+  if (scanptr->Ss != 0 || scanptr->Se != DCTSIZE2-1) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+    cinfo->progressive_mode = TRUE;
+    last_bitpos_ptr = & last_bitpos[0][0];
+    for (ci = 0; ci < cinfo->num_components; ci++) 
+      for (coefi = 0; coefi < DCTSIZE2; coefi++)
+	*last_bitpos_ptr++ = -1;
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    cinfo->progressive_mode = FALSE;
+    for (ci = 0; ci < cinfo->num_components; ci++) 
+      component_sent[ci] = FALSE;
+  }
+
+  for (scanno = 1; scanno <= cinfo->num_scans; scanptr++, scanno++) {
+    /* Validate component indexes */
+    ncomps = scanptr->comps_in_scan;
+    if (ncomps <= 0 || ncomps > MAX_COMPS_IN_SCAN)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, ncomps, MAX_COMPS_IN_SCAN);
+    for (ci = 0; ci < ncomps; ci++) {
+      thisi = scanptr->component_index[ci];
+      if (thisi < 0 || thisi >= cinfo->num_components)
+	ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+      /* Components must appear in SOF order within each scan */
+      if (ci > 0 && thisi <= scanptr->component_index[ci-1])
+	ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+    }
+    /* Validate progression parameters */
+    Ss = scanptr->Ss;
+    Se = scanptr->Se;
+    Ah = scanptr->Ah;
+    Al = scanptr->Al;
+    if (cinfo->progressive_mode) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+      /* The JPEG spec simply gives the ranges 0..13 for Ah and Al, but that
+       * seems wrong: the upper bound ought to depend on data precision.
+       * Perhaps they really meant 0..N+1 for N-bit precision.
+       * Here we allow 0..10 for 8-bit data; Al larger than 10 results in
+       * out-of-range reconstructed DC values during the first DC scan,
+       * which might cause problems for some decoders.
+       */
+#if BITS_IN_JSAMPLE == 8
+#define MAX_AH_AL 10
+#else
+#define MAX_AH_AL 13
+#endif
+      if (Ss < 0 || Ss >= DCTSIZE2 || Se < Ss || Se >= DCTSIZE2 ||
+	  Ah < 0 || Ah > MAX_AH_AL || Al < 0 || Al > MAX_AH_AL)
+	ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+      if (Ss == 0) {
+	if (Se != 0)		/* DC and AC together not OK */
+	  ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+      } else {
+	if (ncomps != 1)	/* AC scans must be for only one component */
+	  ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+      }
+      for (ci = 0; ci < ncomps; ci++) {
+	last_bitpos_ptr = & last_bitpos[scanptr->component_index[ci]][0];
+	if (Ss != 0 && last_bitpos_ptr[0] < 0) /* AC without prior DC scan */
+	  ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+	for (coefi = Ss; coefi <= Se; coefi++) {
+	  if (last_bitpos_ptr[coefi] < 0) {
+	    /* first scan of this coefficient */
+	    if (Ah != 0)
+	      ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+	  } else {
+	    /* not first scan */
+	    if (Ah != last_bitpos_ptr[coefi] || Al != Ah-1)
+	      ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+	  }
+	  last_bitpos_ptr[coefi] = Al;
+	}
+      }
+#endif
+    } else {
+      /* For sequential JPEG, all progression parameters must be these: */
+      if (Ss != 0 || Se != DCTSIZE2-1 || Ah != 0 || Al != 0)
+	ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+      /* Make sure components are not sent twice */
+      for (ci = 0; ci < ncomps; ci++) {
+	thisi = scanptr->component_index[ci];
+	if (component_sent[thisi])
+	  ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+	component_sent[thisi] = TRUE;
+      }
+    }
+  }
+
+  /* Now verify that everything got sent. */
+  if (cinfo->progressive_mode) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+    /* For progressive mode, we only check that at least some DC data
+     * got sent for each component; the spec does not require that all bits
+     * of all coefficients be transmitted.  Would it be wiser to enforce
+     * transmission of all coefficient bits??
+     */
+    for (ci = 0; ci < cinfo->num_components; ci++) {
+      if (last_bitpos[ci][0] < 0)
+	ERREXIT(cinfo, JERR_MISSING_DATA);
+    }
+#endif
+  } else {
+    for (ci = 0; ci < cinfo->num_components; ci++) {
+      if (! component_sent[ci])
+	ERREXIT(cinfo, JERR_MISSING_DATA);
+    }
+  }
+}
+
+#endif /* C_MULTISCAN_FILES_SUPPORTED */
+
+
+LOCAL(void)
+select_scan_parameters (j_compress_ptr cinfo)
+/* Set up the scan parameters for the current scan */
+{
+  int ci;
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+  if (cinfo->scan_info != NULL) {
+    /* Prepare for current scan --- the script is already validated */
+    my_master_ptr master = (my_master_ptr) cinfo->master;
+    const jpeg_scan_info * scanptr = cinfo->scan_info + master->scan_number;
+
+    cinfo->comps_in_scan = scanptr->comps_in_scan;
+    for (ci = 0; ci < scanptr->comps_in_scan; ci++) {
+      cinfo->cur_comp_info[ci] =
+	&cinfo->comp_info[scanptr->component_index[ci]];
+    }
+    cinfo->Ss = scanptr->Ss;
+    cinfo->Se = scanptr->Se;
+    cinfo->Ah = scanptr->Ah;
+    cinfo->Al = scanptr->Al;
+  }
+  else
+#endif
+  {
+    /* Prepare for single sequential-JPEG scan containing all components */
+    if (cinfo->num_components > MAX_COMPS_IN_SCAN)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
+	       MAX_COMPS_IN_SCAN);
+    cinfo->comps_in_scan = cinfo->num_components;
+    for (ci = 0; ci < cinfo->num_components; ci++) {
+      cinfo->cur_comp_info[ci] = &cinfo->comp_info[ci];
+    }
+    cinfo->Ss = 0;
+    cinfo->Se = DCTSIZE2-1;
+    cinfo->Ah = 0;
+    cinfo->Al = 0;
+  }
+}
+
+
+LOCAL(void)
+per_scan_setup (j_compress_ptr cinfo)
+/* Do computations that are needed before processing a JPEG scan */
+/* cinfo->comps_in_scan and cinfo->cur_comp_info[] are already set */
+{
+  int ci, mcublks, tmp;
+  jpeg_component_info *compptr;
+  
+  if (cinfo->comps_in_scan == 1) {
+    
+    /* Noninterleaved (single-component) scan */
+    compptr = cinfo->cur_comp_info[0];
+    
+    /* Overall image size in MCUs */
+    cinfo->MCUs_per_row = compptr->width_in_blocks;
+    cinfo->MCU_rows_in_scan = compptr->height_in_blocks;
+    
+    /* For noninterleaved scan, always one block per MCU */
+    compptr->MCU_width = 1;
+    compptr->MCU_height = 1;
+    compptr->MCU_blocks = 1;
+    compptr->MCU_sample_width = DCTSIZE;
+    compptr->last_col_width = 1;
+    /* For noninterleaved scans, it is convenient to define last_row_height
+     * as the number of block rows present in the last iMCU row.
+     */
+    tmp = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
+    if (tmp == 0) tmp = compptr->v_samp_factor;
+    compptr->last_row_height = tmp;
+    
+    /* Prepare array describing MCU composition */
+    cinfo->blocks_in_MCU = 1;
+    cinfo->MCU_membership[0] = 0;
+    
+  } else {
+    
+    /* Interleaved (multi-component) scan */
+    if (cinfo->comps_in_scan <= 0 || cinfo->comps_in_scan > MAX_COMPS_IN_SCAN)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->comps_in_scan,
+	       MAX_COMPS_IN_SCAN);
+    
+    /* Overall image size in MCUs */
+    cinfo->MCUs_per_row = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width,
+		    (long) (cinfo->max_h_samp_factor*DCTSIZE));
+    cinfo->MCU_rows_in_scan = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height,
+		    (long) (cinfo->max_v_samp_factor*DCTSIZE));
+    
+    cinfo->blocks_in_MCU = 0;
+    
+    for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+      compptr = cinfo->cur_comp_info[ci];
+      /* Sampling factors give # of blocks of component in each MCU */
+      compptr->MCU_width = compptr->h_samp_factor;
+      compptr->MCU_height = compptr->v_samp_factor;
+      compptr->MCU_blocks = compptr->MCU_width * compptr->MCU_height;
+      compptr->MCU_sample_width = compptr->MCU_width * DCTSIZE;
+      /* Figure number of non-dummy blocks in last MCU column & row */
+      tmp = (int) (compptr->width_in_blocks % compptr->MCU_width);
+      if (tmp == 0) tmp = compptr->MCU_width;
+      compptr->last_col_width = tmp;
+      tmp = (int) (compptr->height_in_blocks % compptr->MCU_height);
+      if (tmp == 0) tmp = compptr->MCU_height;
+      compptr->last_row_height = tmp;
+      /* Prepare array describing MCU composition */
+      mcublks = compptr->MCU_blocks;
+      if (cinfo->blocks_in_MCU + mcublks > C_MAX_BLOCKS_IN_MCU)
+	ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
+      while (mcublks-- > 0) {
+	cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci;
+      }
+    }
+    
+  }
+
+  /* Convert restart specified in rows to actual MCU count. */
+  /* Note that count must fit in 16 bits, so we provide limiting. */
+  if (cinfo->restart_in_rows > 0) {
+    long nominal = (long) cinfo->restart_in_rows * (long) cinfo->MCUs_per_row;
+    cinfo->restart_interval = (unsigned int) MIN(nominal, 65535L);
+  }
+}
+
+
+/*
+ * Per-pass setup.
+ * This is called at the beginning of each pass.  We determine which modules
+ * will be active during this pass and give them appropriate start_pass calls.
+ * We also set is_last_pass to indicate whether any more passes will be
+ * required.
+ */
+
+METHODDEF(void)
+prepare_for_pass (j_compress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr) cinfo->master;
+
+  switch (master->pass_type) {
+  case main_pass:
+    /* Initial pass: will collect input data, and do either Huffman
+     * optimization or data output for the first scan.
+     */
+    select_scan_parameters(cinfo);
+    per_scan_setup(cinfo);
+    if (! cinfo->raw_data_in) {
+      (*cinfo->cconvert->start_pass) (cinfo);
+      (*cinfo->downsample->start_pass) (cinfo);
+      (*cinfo->prep->start_pass) (cinfo, JBUF_PASS_THRU);
+    }
+    (*cinfo->fdct->start_pass) (cinfo);
+    (*cinfo->entropy->start_pass) (cinfo, cinfo->optimize_coding);
+    (*cinfo->coef->start_pass) (cinfo,
+				(master->total_passes > 1 ?
+				 JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
+    (*cinfo->main->start_pass) (cinfo, JBUF_PASS_THRU);
+    if (cinfo->optimize_coding) {
+      /* No immediate data output; postpone writing frame/scan headers */
+      master->pub.call_pass_startup = FALSE;
+    } else {
+      /* Will write frame/scan headers at first jpeg_write_scanlines call */
+      master->pub.call_pass_startup = TRUE;
+    }
+    break;
+#ifdef ENTROPY_OPT_SUPPORTED
+  case huff_opt_pass:
+    /* Do Huffman optimization for a scan after the first one. */
+    select_scan_parameters(cinfo);
+    per_scan_setup(cinfo);
+    if (cinfo->Ss != 0 || cinfo->Ah == 0 || cinfo->arith_code) {
+      (*cinfo->entropy->start_pass) (cinfo, TRUE);
+      (*cinfo->coef->start_pass) (cinfo, JBUF_CRANK_DEST);
+      master->pub.call_pass_startup = FALSE;
+      break;
+    }
+    /* Special case: Huffman DC refinement scans need no Huffman table
+     * and therefore we can skip the optimization pass for them.
+     */
+    master->pass_type = output_pass;
+    master->pass_number++;
+    /*FALLTHROUGH*/
+#endif
+  case output_pass:
+    /* Do a data-output pass. */
+    /* We need not repeat per-scan setup if prior optimization pass did it. */
+    if (! cinfo->optimize_coding) {
+      select_scan_parameters(cinfo);
+      per_scan_setup(cinfo);
+    }
+    (*cinfo->entropy->start_pass) (cinfo, FALSE);
+    (*cinfo->coef->start_pass) (cinfo, JBUF_CRANK_DEST);
+    /* We emit frame/scan headers now */
+    if (master->scan_number == 0)
+      (*cinfo->marker->write_frame_header) (cinfo);
+    (*cinfo->marker->write_scan_header) (cinfo);
+    master->pub.call_pass_startup = FALSE;
+    break;
+  default:
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+  }
+
+  master->pub.is_last_pass = (master->pass_number == master->total_passes-1);
+
+  /* Set up progress monitor's pass info if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->completed_passes = master->pass_number;
+    cinfo->progress->total_passes = master->total_passes;
+  }
+}
+
+
+/*
+ * Special start-of-pass hook.
+ * This is called by jpeg_write_scanlines if call_pass_startup is TRUE.
+ * In single-pass processing, we need this hook because we don't want to
+ * write frame/scan headers during jpeg_start_compress; we want to let the
+ * application write COM markers etc. between jpeg_start_compress and the
+ * jpeg_write_scanlines loop.
+ * In multi-pass processing, this routine is not used.
+ */
+
+METHODDEF(void)
+pass_startup (j_compress_ptr cinfo)
+{
+  cinfo->master->call_pass_startup = FALSE; /* reset flag so call only once */
+
+  (*cinfo->marker->write_frame_header) (cinfo);
+  (*cinfo->marker->write_scan_header) (cinfo);
+}
+
+
+/*
+ * Finish up at end of pass.
+ */
+
+METHODDEF(void)
+finish_pass_master (j_compress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr) cinfo->master;
+
+  /* The entropy coder always needs an end-of-pass call,
+   * either to analyze statistics or to flush its output buffer.
+   */
+  (*cinfo->entropy->finish_pass) (cinfo);
+
+  /* Update state for next pass */
+  switch (master->pass_type) {
+  case main_pass:
+    /* next pass is either output of scan 0 (after optimization)
+     * or output of scan 1 (if no optimization).
+     */
+    master->pass_type = output_pass;
+    if (! cinfo->optimize_coding)
+      master->scan_number++;
+    break;
+  case huff_opt_pass:
+    /* next pass is always output of current scan */
+    master->pass_type = output_pass;
+    break;
+  case output_pass:
+    /* next pass is either optimization or output of next scan */
+    if (cinfo->optimize_coding)
+      master->pass_type = huff_opt_pass;
+    master->scan_number++;
+    break;
+  }
+
+  master->pass_number++;
+}
+
+
+/*
+ * Initialize master compression control.
+ */
+
+GLOBAL(void)
+jinit_c_master_control (j_compress_ptr cinfo, boolean transcode_only)
+{
+  my_master_ptr master;
+
+  master = (my_master_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(my_comp_master));
+  cinfo->master = (struct jpeg_comp_master *) master;
+  master->pub.prepare_for_pass = prepare_for_pass;
+  master->pub.pass_startup = pass_startup;
+  master->pub.finish_pass = finish_pass_master;
+  master->pub.is_last_pass = FALSE;
+
+  /* Validate parameters, determine derived values */
+  initial_setup(cinfo);
+
+  if (cinfo->scan_info != NULL) {
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+    validate_script(cinfo);
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    cinfo->progressive_mode = FALSE;
+    cinfo->num_scans = 1;
+  }
+
+  if (cinfo->progressive_mode)	/*  TEMPORARY HACK ??? */
+    cinfo->optimize_coding = TRUE; /* assume default tables no good for progressive mode */
+
+  /* Initialize my private state */
+  if (transcode_only) {
+    /* no main pass in transcoding */
+    if (cinfo->optimize_coding)
+      master->pass_type = huff_opt_pass;
+    else
+      master->pass_type = output_pass;
+  } else {
+    /* for normal compression, first pass is always this type: */
+    master->pass_type = main_pass;
+  }
+  master->scan_number = 0;
+  master->pass_number = 0;
+  if (cinfo->optimize_coding)
+    master->total_passes = cinfo->num_scans * 2;
+  else
+    master->total_passes = cinfo->num_scans;
+}
diff --git a/libjpegtwrp/jcomapi.c b/libjpegtwrp/jcomapi.c
new file mode 100644
index 000000000..9b1fa7568
--- /dev/null
+++ b/libjpegtwrp/jcomapi.c
@@ -0,0 +1,106 @@
+/*
+ * jcomapi.c
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains application interface routines that are used for both
+ * compression and decompression.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Abort processing of a JPEG compression or decompression operation,
+ * but don't destroy the object itself.
+ *
+ * For this, we merely clean up all the nonpermanent memory pools.
+ * Note that temp files (virtual arrays) are not allowed to belong to
+ * the permanent pool, so we will be able to close all temp files here.
+ * Closing a data source or destination, if necessary, is the application's
+ * responsibility.
+ */
+
+GLOBAL(void)
+jpeg_abort (j_common_ptr cinfo)
+{
+  int pool;
+
+  /* Do nothing if called on a not-initialized or destroyed JPEG object. */
+  if (cinfo->mem == NULL)
+    return;
+
+  /* Releasing pools in reverse order might help avoid fragmentation
+   * with some (brain-damaged) malloc libraries.
+   */
+  for (pool = JPOOL_NUMPOOLS-1; pool > JPOOL_PERMANENT; pool--) {
+    (*cinfo->mem->free_pool) (cinfo, pool);
+  }
+
+  /* Reset overall state for possible reuse of object */
+  if (cinfo->is_decompressor) {
+    cinfo->global_state = DSTATE_START;
+    /* Try to keep application from accessing now-deleted marker list.
+     * A bit kludgy to do it here, but this is the most central place.
+     */
+    ((j_decompress_ptr) cinfo)->marker_list = NULL;
+  } else {
+    cinfo->global_state = CSTATE_START;
+  }
+}
+
+
+/*
+ * Destruction of a JPEG object.
+ *
+ * Everything gets deallocated except the master jpeg_compress_struct itself
+ * and the error manager struct.  Both of these are supplied by the application
+ * and must be freed, if necessary, by the application.  (Often they are on
+ * the stack and so don't need to be freed anyway.)
+ * Closing a data source or destination, if necessary, is the application's
+ * responsibility.
+ */
+
+GLOBAL(void)
+jpeg_destroy (j_common_ptr cinfo)
+{
+  /* We need only tell the memory manager to release everything. */
+  /* NB: mem pointer is NULL if memory mgr failed to initialize. */
+  if (cinfo->mem != NULL)
+    (*cinfo->mem->self_destruct) (cinfo);
+  cinfo->mem = NULL;		/* be safe if jpeg_destroy is called twice */
+  cinfo->global_state = 0;	/* mark it destroyed */
+}
+
+
+/*
+ * Convenience routines for allocating quantization and Huffman tables.
+ * (Would jutils.c be a more reasonable place to put these?)
+ */
+
+GLOBAL(JQUANT_TBL *)
+jpeg_alloc_quant_table (j_common_ptr cinfo)
+{
+  JQUANT_TBL *tbl;
+
+  tbl = (JQUANT_TBL *)
+    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, SIZEOF(JQUANT_TBL));
+  tbl->sent_table = FALSE;	/* make sure this is false in any new table */
+  return tbl;
+}
+
+
+GLOBAL(JHUFF_TBL *)
+jpeg_alloc_huff_table (j_common_ptr cinfo)
+{
+  JHUFF_TBL *tbl;
+
+  tbl = (JHUFF_TBL *)
+    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, SIZEOF(JHUFF_TBL));
+  tbl->sent_table = FALSE;	/* make sure this is false in any new table */
+  return tbl;
+}
diff --git a/libjpegtwrp/jconfig.bcc b/libjpegtwrp/jconfig.bcc
new file mode 100644
index 000000000..c6c53ff63
--- /dev/null
+++ b/libjpegtwrp/jconfig.bcc
@@ -0,0 +1,48 @@
+/* jconfig.bcc --- jconfig.h for Borland C (Turbo C) on MS-DOS or OS/2. */
+/* see jconfig.doc for explanations */
+
+#define HAVE_PROTOTYPES
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+/* #define void char */
+/* #define const */
+#undef CHAR_IS_UNSIGNED
+#define HAVE_STDDEF_H
+#define HAVE_STDLIB_H
+#undef NEED_BSD_STRINGS
+#undef NEED_SYS_TYPES_H
+#ifdef __MSDOS__
+#define NEED_FAR_POINTERS	/* for small or medium memory model */
+#endif
+#undef NEED_SHORT_EXTERNAL_NAMES
+#undef INCOMPLETE_TYPES_BROKEN	/* this assumes you have -w-stu in CFLAGS */
+
+#ifdef JPEG_INTERNALS
+
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+#ifdef __MSDOS__
+#define USE_MSDOS_MEMMGR	/* Define this if you use jmemdos.c */
+#define MAX_ALLOC_CHUNK 65520L	/* Maximum request to malloc() */
+#define USE_FMEM		/* Borland has _fmemcpy() and _fmemset() */
+#endif
+
+#endif /* JPEG_INTERNALS */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+#define TWO_FILE_COMMANDLINE
+#define USE_SETMODE		/* Borland has setmode() */
+#ifdef __MSDOS__
+#define NEED_SIGNAL_CATCHER	/* Define this if you use jmemdos.c */
+#endif
+#undef DONT_USE_B_MODE
+#undef PROGRESS_REPORT		/* optional */
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jconfig.cfg b/libjpegtwrp/jconfig.cfg
new file mode 100644
index 000000000..36a04fa84
--- /dev/null
+++ b/libjpegtwrp/jconfig.cfg
@@ -0,0 +1,44 @@
+/* jconfig.cfg --- source file edited by configure script */
+/* see jconfig.doc for explanations */
+
+#undef HAVE_PROTOTYPES
+#undef HAVE_UNSIGNED_CHAR
+#undef HAVE_UNSIGNED_SHORT
+#undef void
+#undef const
+#undef CHAR_IS_UNSIGNED
+#undef HAVE_STDDEF_H
+#undef HAVE_STDLIB_H
+#undef NEED_BSD_STRINGS
+#undef NEED_SYS_TYPES_H
+#undef NEED_FAR_POINTERS
+#undef NEED_SHORT_EXTERNAL_NAMES
+/* Define this if you get warnings about undefined structures. */
+#undef INCOMPLETE_TYPES_BROKEN
+
+#ifdef JPEG_INTERNALS
+
+#undef RIGHT_SHIFT_IS_UNSIGNED
+#undef INLINE
+/* These are for configuring the JPEG memory manager. */
+#undef DEFAULT_MAX_MEM
+#undef NO_MKTEMP
+
+#endif /* JPEG_INTERNALS */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+#undef TWO_FILE_COMMANDLINE
+#undef NEED_SIGNAL_CATCHER
+#undef DONT_USE_B_MODE
+
+/* Define this if you want percent-done progress reports from cjpeg/djpeg. */
+#undef PROGRESS_REPORT
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jconfig.dj b/libjpegtwrp/jconfig.dj
new file mode 100644
index 000000000..f759a9dbd
--- /dev/null
+++ b/libjpegtwrp/jconfig.dj
@@ -0,0 +1,38 @@
+/* jconfig.dj --- jconfig.h for DJGPP (Delorie's GNU C port) on MS-DOS. */
+/* see jconfig.doc for explanations */
+
+#define HAVE_PROTOTYPES
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+/* #define void char */
+/* #define const */
+#undef CHAR_IS_UNSIGNED
+#define HAVE_STDDEF_H
+#define HAVE_STDLIB_H
+#undef NEED_BSD_STRINGS
+#undef NEED_SYS_TYPES_H
+#undef NEED_FAR_POINTERS	/* DJGPP uses flat 32-bit addressing */
+#undef NEED_SHORT_EXTERNAL_NAMES
+#undef INCOMPLETE_TYPES_BROKEN
+
+#ifdef JPEG_INTERNALS
+
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+#endif /* JPEG_INTERNALS */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+#undef TWO_FILE_COMMANDLINE	/* optional */
+#define USE_SETMODE		/* Needed to make one-file style work in DJGPP */
+#undef NEED_SIGNAL_CATCHER	/* Define this if you use jmemname.c */
+#undef DONT_USE_B_MODE
+#undef PROGRESS_REPORT		/* optional */
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jconfig.doc b/libjpegtwrp/jconfig.doc
new file mode 100644
index 000000000..c18d1c064
--- /dev/null
+++ b/libjpegtwrp/jconfig.doc
@@ -0,0 +1,155 @@
+/*
+ * jconfig.doc
+ *
+ * Copyright (C) 1991-1994, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file documents the configuration options that are required to
+ * customize the JPEG software for a particular system.
+ *
+ * The actual configuration options for a particular installation are stored
+ * in jconfig.h.  On many machines, jconfig.h can be generated automatically
+ * or copied from one of the "canned" jconfig files that we supply.  But if
+ * you need to generate a jconfig.h file by hand, this file tells you how.
+ *
+ * DO NOT EDIT THIS FILE --- IT WON'T ACCOMPLISH ANYTHING.
+ * EDIT A COPY NAMED JCONFIG.H.
+ */
+
+
+/*
+ * These symbols indicate the properties of your machine or compiler.
+ * #define the symbol if yes, #undef it if no.
+ */
+
+/* Does your compiler support function prototypes?
+ * (If not, you also need to use ansi2knr, see install.doc)
+ */
+#define HAVE_PROTOTYPES
+
+/* Does your compiler support the declaration "unsigned char" ?
+ * How about "unsigned short" ?
+ */
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+
+/* Define "void" as "char" if your compiler doesn't know about type void.
+ * NOTE: be sure to define void such that "void *" represents the most general
+ * pointer type, e.g., that returned by malloc().
+ */
+/* #define void char */
+
+/* Define "const" as empty if your compiler doesn't know the "const" keyword.
+ */
+/* #define const */
+
+/* Define this if an ordinary "char" type is unsigned.
+ * If you're not sure, leaving it undefined will work at some cost in speed.
+ * If you defined HAVE_UNSIGNED_CHAR then the speed difference is minimal.
+ */
+#undef CHAR_IS_UNSIGNED
+
+/* Define this if your system has an ANSI-conforming <stddef.h> file.
+ */
+#define HAVE_STDDEF_H
+
+/* Define this if your system has an ANSI-conforming <stdlib.h> file.
+ */
+#define HAVE_STDLIB_H
+
+/* Define this if your system does not have an ANSI/SysV <string.h>,
+ * but does have a BSD-style <strings.h>.
+ */
+#undef NEED_BSD_STRINGS
+
+/* Define this if your system does not provide typedef size_t in any of the
+ * ANSI-standard places (stddef.h, stdlib.h, or stdio.h), but places it in
+ * <sys/types.h> instead.
+ */
+#undef NEED_SYS_TYPES_H
+
+/* For 80x86 machines, you need to define NEED_FAR_POINTERS,
+ * unless you are using a large-data memory model or 80386 flat-memory mode.
+ * On less brain-damaged CPUs this symbol must not be defined.
+ * (Defining this symbol causes large data structures to be referenced through
+ * "far" pointers and to be allocated with a special version of malloc.)
+ */
+#undef NEED_FAR_POINTERS
+
+/* Define this if your linker needs global names to be unique in less
+ * than the first 15 characters.
+ */
+#undef NEED_SHORT_EXTERNAL_NAMES
+
+/* Although a real ANSI C compiler can deal perfectly well with pointers to
+ * unspecified structures (see "incomplete types" in the spec), a few pre-ANSI
+ * and pseudo-ANSI compilers get confused.  To keep one of these bozos happy,
+ * define INCOMPLETE_TYPES_BROKEN.  This is not recommended unless you
+ * actually get "missing structure definition" warnings or errors while
+ * compiling the JPEG code.
+ */
+#undef INCOMPLETE_TYPES_BROKEN
+
+
+/*
+ * The following options affect code selection within the JPEG library,
+ * but they don't need to be visible to applications using the library.
+ * To minimize application namespace pollution, the symbols won't be
+ * defined unless JPEG_INTERNALS has been defined.
+ */
+
+#ifdef JPEG_INTERNALS
+
+/* Define this if your compiler implements ">>" on signed values as a logical
+ * (unsigned) shift; leave it undefined if ">>" is a signed (arithmetic) shift,
+ * which is the normal and rational definition.
+ */
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+
+#endif /* JPEG_INTERNALS */
+
+
+/*
+ * The remaining options do not affect the JPEG library proper,
+ * but only the sample applications cjpeg/djpeg (see cjpeg.c, djpeg.c).
+ * Other applications can ignore these.
+ */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+/* These defines indicate which image (non-JPEG) file formats are allowed. */
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+/* Define this if you want to name both input and output files on the command
+ * line, rather than using stdout and optionally stdin.  You MUST do this if
+ * your system can't cope with binary I/O to stdin/stdout.  See comments at
+ * head of cjpeg.c or djpeg.c.
+ */
+#undef TWO_FILE_COMMANDLINE
+
+/* Define this if your system needs explicit cleanup of temporary files.
+ * This is crucial under MS-DOS, where the temporary "files" may be areas
+ * of extended memory; on most other systems it's not as important.
+ */
+#undef NEED_SIGNAL_CATCHER
+
+/* By default, we open image files with fopen(...,"rb") or fopen(...,"wb").
+ * This is necessary on systems that distinguish text files from binary files,
+ * and is harmless on most systems that don't.  If you have one of the rare
+ * systems that complains about the "b" spec, define this symbol.
+ */
+#undef DONT_USE_B_MODE
+
+/* Define this if you want percent-done progress reports from cjpeg/djpeg.
+ */
+#undef PROGRESS_REPORT
+
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jconfig.h b/libjpegtwrp/jconfig.h
new file mode 100644
index 000000000..15a98177b
--- /dev/null
+++ b/libjpegtwrp/jconfig.h
@@ -0,0 +1,156 @@
+/* android jconfig.h */
+/*
+ * jconfig.doc
+ *
+ * Copyright (C) 1991-1994, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file documents the configuration options that are required to
+ * customize the JPEG software for a particular system.
+ *
+ * The actual configuration options for a particular installation are stored
+ * in jconfig.h.  On many machines, jconfig.h can be generated automatically
+ * or copied from one of the "canned" jconfig files that we supply.  But if
+ * you need to generate a jconfig.h file by hand, this file tells you how.
+ *
+ * DO NOT EDIT THIS FILE --- IT WON'T ACCOMPLISH ANYTHING.
+ * EDIT A COPY NAMED JCONFIG.H.
+ */
+
+
+/*
+ * These symbols indicate the properties of your machine or compiler.
+ * #define the symbol if yes, #undef it if no.
+ */
+
+/* Does your compiler support function prototypes?
+ * (If not, you also need to use ansi2knr, see install.doc)
+ */
+#define HAVE_PROTOTYPES
+
+/* Does your compiler support the declaration "unsigned char" ?
+ * How about "unsigned short" ?
+ */
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+
+/* Define "void" as "char" if your compiler doesn't know about type void.
+ * NOTE: be sure to define void such that "void *" represents the most general
+ * pointer type, e.g., that returned by malloc().
+ */
+/* #define void char */
+
+/* Define "const" as empty if your compiler doesn't know the "const" keyword.
+ */
+/* #define const */
+
+/* Define this if an ordinary "char" type is unsigned.
+ * If you're not sure, leaving it undefined will work at some cost in speed.
+ * If you defined HAVE_UNSIGNED_CHAR then the speed difference is minimal.
+ */
+#undef CHAR_IS_UNSIGNED
+
+/* Define this if your system has an ANSI-conforming <stddef.h> file.
+ */
+#define HAVE_STDDEF_H
+
+/* Define this if your system has an ANSI-conforming <stdlib.h> file.
+ */
+#define HAVE_STDLIB_H
+
+/* Define this if your system does not have an ANSI/SysV <string.h>,
+ * but does have a BSD-style <strings.h>.
+ */
+#undef NEED_BSD_STRINGS
+
+/* Define this if your system does not provide typedef size_t in any of the
+ * ANSI-standard places (stddef.h, stdlib.h, or stdio.h), but places it in
+ * <sys/types.h> instead.
+ */
+#undef NEED_SYS_TYPES_H
+
+/* For 80x86 machines, you need to define NEED_FAR_POINTERS,
+ * unless you are using a large-data memory model or 80386 flat-memory mode.
+ * On less brain-damaged CPUs this symbol must not be defined.
+ * (Defining this symbol causes large data structures to be referenced through
+ * "far" pointers and to be allocated with a special version of malloc.)
+ */
+#undef NEED_FAR_POINTERS
+
+/* Define this if your linker needs global names to be unique in less
+ * than the first 15 characters.
+ */
+#undef NEED_SHORT_EXTERNAL_NAMES
+
+/* Although a real ANSI C compiler can deal perfectly well with pointers to
+ * unspecified structures (see "incomplete types" in the spec), a few pre-ANSI
+ * and pseudo-ANSI compilers get confused.  To keep one of these bozos happy,
+ * define INCOMPLETE_TYPES_BROKEN.  This is not recommended unless you
+ * actually get "missing structure definition" warnings or errors while
+ * compiling the JPEG code.
+ */
+#undef INCOMPLETE_TYPES_BROKEN
+
+
+/*
+ * The following options affect code selection within the JPEG library,
+ * but they don't need to be visible to applications using the library.
+ * To minimize application namespace pollution, the symbols won't be
+ * defined unless JPEG_INTERNALS has been defined.
+ */
+
+#ifdef JPEG_INTERNALS
+
+/* Define this if your compiler implements ">>" on signed values as a logical
+ * (unsigned) shift; leave it undefined if ">>" is a signed (arithmetic) shift,
+ * which is the normal and rational definition.
+ */
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+
+#endif /* JPEG_INTERNALS */
+
+
+/*
+ * The remaining options do not affect the JPEG library proper,
+ * but only the sample applications cjpeg/djpeg (see cjpeg.c, djpeg.c).
+ * Other applications can ignore these.
+ */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+/* These defines indicate which image (non-JPEG) file formats are allowed. */
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+/* Define this if you want to name both input and output files on the command
+ * line, rather than using stdout and optionally stdin.  You MUST do this if
+ * your system can't cope with binary I/O to stdin/stdout.  See comments at
+ * head of cjpeg.c or djpeg.c.
+ */
+#undef TWO_FILE_COMMANDLINE
+
+/* Define this if your system needs explicit cleanup of temporary files.
+ * This is crucial under MS-DOS, where the temporary "files" may be areas
+ * of extended memory; on most other systems it's not as important.
+ */
+#undef NEED_SIGNAL_CATCHER
+
+/* By default, we open image files with fopen(...,"rb") or fopen(...,"wb").
+ * This is necessary on systems that distinguish text files from binary files,
+ * and is harmless on most systems that don't.  If you have one of the rare
+ * systems that complains about the "b" spec, define this symbol.
+ */
+#undef DONT_USE_B_MODE
+
+/* Define this if you want percent-done progress reports from cjpeg/djpeg.
+ */
+#undef PROGRESS_REPORT
+
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jconfig.mac b/libjpegtwrp/jconfig.mac
new file mode 100644
index 000000000..0de3efe24
--- /dev/null
+++ b/libjpegtwrp/jconfig.mac
@@ -0,0 +1,43 @@
+/* jconfig.mac --- jconfig.h for CodeWarrior on Apple Macintosh */
+/* see jconfig.doc for explanations */
+
+#define HAVE_PROTOTYPES
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+/* #define void char */
+/* #define const */
+#undef CHAR_IS_UNSIGNED
+#define HAVE_STDDEF_H
+#define HAVE_STDLIB_H
+#undef NEED_BSD_STRINGS
+#undef NEED_SYS_TYPES_H
+#undef NEED_FAR_POINTERS
+#undef NEED_SHORT_EXTERNAL_NAMES
+#undef INCOMPLETE_TYPES_BROKEN
+
+#ifdef JPEG_INTERNALS
+
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+#define USE_MAC_MEMMGR		/* Define this if you use jmemmac.c */
+
+#define ALIGN_TYPE long		/* Needed for 680x0 Macs */
+
+#endif /* JPEG_INTERNALS */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+#define USE_CCOMMAND		/* Command line reader for Macintosh */
+#define TWO_FILE_COMMANDLINE	/* Binary I/O thru stdin/stdout doesn't work */
+
+#undef NEED_SIGNAL_CATCHER
+#undef DONT_USE_B_MODE
+#undef PROGRESS_REPORT		/* optional */
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jconfig.manx b/libjpegtwrp/jconfig.manx
new file mode 100644
index 000000000..6dd0d008e
--- /dev/null
+++ b/libjpegtwrp/jconfig.manx
@@ -0,0 +1,43 @@
+/* jconfig.manx --- jconfig.h for Amiga systems using Manx Aztec C ver 5.x. */
+/* see jconfig.doc for explanations */
+
+#define HAVE_PROTOTYPES
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+/* #define void char */
+/* #define const */
+#undef CHAR_IS_UNSIGNED
+#define HAVE_STDDEF_H
+#define HAVE_STDLIB_H
+#undef NEED_BSD_STRINGS
+#undef NEED_SYS_TYPES_H
+#undef NEED_FAR_POINTERS
+#undef NEED_SHORT_EXTERNAL_NAMES
+#undef INCOMPLETE_TYPES_BROKEN
+
+#ifdef JPEG_INTERNALS
+
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+#define TEMP_DIRECTORY "JPEGTMP:"	/* recommended setting for Amiga */
+
+#define SHORTxSHORT_32		/* produces better DCT code with Aztec C */
+
+#endif /* JPEG_INTERNALS */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+#define TWO_FILE_COMMANDLINE
+#define NEED_SIGNAL_CATCHER
+#undef DONT_USE_B_MODE
+#undef PROGRESS_REPORT		/* optional */
+
+#define signal_catcher _abort	/* hack for Aztec C naming requirements */
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jconfig.mc6 b/libjpegtwrp/jconfig.mc6
new file mode 100644
index 000000000..c55082df4
--- /dev/null
+++ b/libjpegtwrp/jconfig.mc6
@@ -0,0 +1,52 @@
+/* jconfig.mc6 --- jconfig.h for Microsoft C on MS-DOS, version 6.00A & up. */
+/* see jconfig.doc for explanations */
+
+#define HAVE_PROTOTYPES
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+/* #define void char */
+/* #define const */
+#undef CHAR_IS_UNSIGNED
+#define HAVE_STDDEF_H
+#define HAVE_STDLIB_H
+#undef NEED_BSD_STRINGS
+#undef NEED_SYS_TYPES_H
+#define NEED_FAR_POINTERS	/* for small or medium memory model */
+#undef NEED_SHORT_EXTERNAL_NAMES
+#undef INCOMPLETE_TYPES_BROKEN
+
+#ifdef JPEG_INTERNALS
+
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+#define USE_MSDOS_MEMMGR	/* Define this if you use jmemdos.c */
+
+#define MAX_ALLOC_CHUNK 65520L	/* Maximum request to malloc() */
+
+#define USE_FMEM		/* Microsoft has _fmemcpy() and _fmemset() */
+
+#define NEED_FHEAPMIN		/* far heap management routines are broken */
+
+#define SHORTxLCONST_32		/* enable compiler-specific DCT optimization */
+/* Note: the above define is known to improve the code with Microsoft C 6.00A.
+ * I do not know whether it is good for later compiler versions.
+ * Please report any info on this point to jpeg-info@uunet.uu.net.
+ */
+
+#endif /* JPEG_INTERNALS */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+#define TWO_FILE_COMMANDLINE
+#define USE_SETMODE		/* Microsoft has setmode() */
+#define NEED_SIGNAL_CATCHER	/* Define this if you use jmemdos.c */
+#undef DONT_USE_B_MODE
+#undef PROGRESS_REPORT		/* optional */
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jconfig.sas b/libjpegtwrp/jconfig.sas
new file mode 100644
index 000000000..efdac2229
--- /dev/null
+++ b/libjpegtwrp/jconfig.sas
@@ -0,0 +1,43 @@
+/* jconfig.sas --- jconfig.h for Amiga systems using SAS C 6.0 and up. */
+/* see jconfig.doc for explanations */
+
+#define HAVE_PROTOTYPES
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+/* #define void char */
+/* #define const */
+#undef CHAR_IS_UNSIGNED
+#define HAVE_STDDEF_H
+#define HAVE_STDLIB_H
+#undef NEED_BSD_STRINGS
+#undef NEED_SYS_TYPES_H
+#undef NEED_FAR_POINTERS
+#undef NEED_SHORT_EXTERNAL_NAMES
+#undef INCOMPLETE_TYPES_BROKEN
+
+#ifdef JPEG_INTERNALS
+
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+#define TEMP_DIRECTORY "JPEGTMP:"	/* recommended setting for Amiga */
+
+#define NO_MKTEMP		/* SAS C doesn't have mktemp() */
+
+#define SHORTxSHORT_32		/* produces better DCT code with SAS C */
+
+#endif /* JPEG_INTERNALS */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+#define TWO_FILE_COMMANDLINE
+#define NEED_SIGNAL_CATCHER
+#undef DONT_USE_B_MODE
+#undef PROGRESS_REPORT		/* optional */
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jconfig.st b/libjpegtwrp/jconfig.st
new file mode 100644
index 000000000..4421b7a1a
--- /dev/null
+++ b/libjpegtwrp/jconfig.st
@@ -0,0 +1,42 @@
+/* jconfig.st --- jconfig.h for Atari ST/STE/TT using Pure C or Turbo C. */
+/* see jconfig.doc for explanations */
+
+#define HAVE_PROTOTYPES
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+/* #define void char */
+/* #define const */
+#undef CHAR_IS_UNSIGNED
+#define HAVE_STDDEF_H
+#define HAVE_STDLIB_H
+#undef NEED_BSD_STRINGS
+#undef NEED_SYS_TYPES_H
+#undef NEED_FAR_POINTERS
+#undef NEED_SHORT_EXTERNAL_NAMES
+#define INCOMPLETE_TYPES_BROKEN	/* suppress undefined-structure warnings */
+
+#ifdef JPEG_INTERNALS
+
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+#define ALIGN_TYPE  long	/* apparently double is a weird size? */
+
+#endif /* JPEG_INTERNALS */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+#define TWO_FILE_COMMANDLINE	/* optional -- undef if you like Unix style */
+/* Note: if you undef TWO_FILE_COMMANDLINE, you may need to define
+ * USE_SETMODE.  Some Atari compilers require it, some do not.
+ */
+#define NEED_SIGNAL_CATCHER	/* needed if you use jmemname.c */
+#undef DONT_USE_B_MODE
+#undef PROGRESS_REPORT		/* optional */
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jconfig.vc b/libjpegtwrp/jconfig.vc
new file mode 100644
index 000000000..7e291c75b
--- /dev/null
+++ b/libjpegtwrp/jconfig.vc
@@ -0,0 +1,45 @@
+/* jconfig.vc --- jconfig.h for Microsoft Visual C++ on Windows 95 or NT. */
+/* see jconfig.doc for explanations */
+
+#define HAVE_PROTOTYPES
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+/* #define void char */
+/* #define const */
+#undef CHAR_IS_UNSIGNED
+#define HAVE_STDDEF_H
+#define HAVE_STDLIB_H
+#undef NEED_BSD_STRINGS
+#undef NEED_SYS_TYPES_H
+#undef NEED_FAR_POINTERS	/* we presume a 32-bit flat memory model */
+#undef NEED_SHORT_EXTERNAL_NAMES
+#undef INCOMPLETE_TYPES_BROKEN
+
+/* Define "boolean" as unsigned char, not int, per Windows custom */
+#ifndef __RPCNDR_H__		/* don't conflict if rpcndr.h already read */
+typedef unsigned char boolean;
+#endif
+#define HAVE_BOOLEAN		/* prevent jmorecfg.h from redefining it */
+
+
+#ifdef JPEG_INTERNALS
+
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+#endif /* JPEG_INTERNALS */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+#define TWO_FILE_COMMANDLINE	/* optional */
+#define USE_SETMODE		/* Microsoft has setmode() */
+#undef NEED_SIGNAL_CATCHER
+#undef DONT_USE_B_MODE
+#undef PROGRESS_REPORT		/* optional */
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jconfig.vms b/libjpegtwrp/jconfig.vms
new file mode 100644
index 000000000..55a6ffba5
--- /dev/null
+++ b/libjpegtwrp/jconfig.vms
@@ -0,0 +1,37 @@
+/* jconfig.vms --- jconfig.h for use on Digital VMS. */
+/* see jconfig.doc for explanations */
+
+#define HAVE_PROTOTYPES
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+/* #define void char */
+/* #define const */
+#undef CHAR_IS_UNSIGNED
+#define HAVE_STDDEF_H
+#define HAVE_STDLIB_H
+#undef NEED_BSD_STRINGS
+#undef NEED_SYS_TYPES_H
+#undef NEED_FAR_POINTERS
+#undef NEED_SHORT_EXTERNAL_NAMES
+#undef INCOMPLETE_TYPES_BROKEN
+
+#ifdef JPEG_INTERNALS
+
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+#endif /* JPEG_INTERNALS */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+#define TWO_FILE_COMMANDLINE	/* Needed on VMS */
+#undef NEED_SIGNAL_CATCHER
+#undef DONT_USE_B_MODE
+#undef PROGRESS_REPORT		/* optional */
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jconfig.wat b/libjpegtwrp/jconfig.wat
new file mode 100644
index 000000000..6cc545bae
--- /dev/null
+++ b/libjpegtwrp/jconfig.wat
@@ -0,0 +1,38 @@
+/* jconfig.wat --- jconfig.h for Watcom C/C++ on MS-DOS or OS/2. */
+/* see jconfig.doc for explanations */
+
+#define HAVE_PROTOTYPES
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+/* #define void char */
+/* #define const */
+#define CHAR_IS_UNSIGNED
+#define HAVE_STDDEF_H
+#define HAVE_STDLIB_H
+#undef NEED_BSD_STRINGS
+#undef NEED_SYS_TYPES_H
+#undef NEED_FAR_POINTERS	/* Watcom uses flat 32-bit addressing */
+#undef NEED_SHORT_EXTERNAL_NAMES
+#undef INCOMPLETE_TYPES_BROKEN
+
+#ifdef JPEG_INTERNALS
+
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+#endif /* JPEG_INTERNALS */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+#undef TWO_FILE_COMMANDLINE	/* optional */
+#define USE_SETMODE		/* Needed to make one-file style work in Watcom */
+#undef NEED_SIGNAL_CATCHER	/* Define this if you use jmemname.c */
+#undef DONT_USE_B_MODE
+#undef PROGRESS_REPORT		/* optional */
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/libjpegtwrp/jcparam.c b/libjpegtwrp/jcparam.c
new file mode 100644
index 000000000..a8dcf6154
--- /dev/null
+++ b/libjpegtwrp/jcparam.c
@@ -0,0 +1,614 @@
+/*
+ * jcparam.c
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains optional default-setting code for the JPEG compressor.
+ * Applications do not have to use this file, but those that don't use it
+ * must know a lot more about the innards of the JPEG code.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Quantization table setup routines
+ */
+
+GLOBAL(void)
+jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
+		      const unsigned int *basic_table,
+		      int scale_factor, boolean force_baseline)
+/* Define a quantization table equal to the basic_table times
+ * a scale factor (given as a percentage).
+ * If force_baseline is TRUE, the computed quantization table entries
+ * are limited to 1..255 for JPEG baseline compatibility.
+ */
+{
+  JQUANT_TBL ** qtblptr;
+  int i;
+  long temp;
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  if (which_tbl < 0 || which_tbl >= NUM_QUANT_TBLS)
+    ERREXIT1(cinfo, JERR_DQT_INDEX, which_tbl);
+
+  qtblptr = & cinfo->quant_tbl_ptrs[which_tbl];
+
+  if (*qtblptr == NULL)
+    *qtblptr = jpeg_alloc_quant_table((j_common_ptr) cinfo);
+
+  for (i = 0; i < DCTSIZE2; i++) {
+    temp = ((long) basic_table[i] * scale_factor + 50L) / 100L;
+    /* limit the values to the valid range */
+    if (temp <= 0L) temp = 1L;
+    if (temp > 32767L) temp = 32767L; /* max quantizer needed for 12 bits */
+    if (force_baseline && temp > 255L)
+      temp = 255L;		/* limit to baseline range if requested */
+    (*qtblptr)->quantval[i] = (UINT16) temp;
+  }
+
+  /* Initialize sent_table FALSE so table will be written to JPEG file. */
+  (*qtblptr)->sent_table = FALSE;
+}
+
+
+GLOBAL(void)
+jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
+			 boolean force_baseline)
+/* Set or change the 'quality' (quantization) setting, using default tables
+ * and a straight percentage-scaling quality scale.  In most cases it's better
+ * to use jpeg_set_quality (below); this entry point is provided for
+ * applications that insist on a linear percentage scaling.
+ */
+{
+  /* These are the sample quantization tables given in JPEG spec section K.1.
+   * The spec says that the values given produce "good" quality, and
+   * when divided by 2, "very good" quality.
+   */
+  static const unsigned int std_luminance_quant_tbl[DCTSIZE2] = {
+    16,  11,  10,  16,  24,  40,  51,  61,
+    12,  12,  14,  19,  26,  58,  60,  55,
+    14,  13,  16,  24,  40,  57,  69,  56,
+    14,  17,  22,  29,  51,  87,  80,  62,
+    18,  22,  37,  56,  68, 109, 103,  77,
+    24,  35,  55,  64,  81, 104, 113,  92,
+    49,  64,  78,  87, 103, 121, 120, 101,
+    72,  92,  95,  98, 112, 100, 103,  99
+  };
+  static const unsigned int std_chrominance_quant_tbl[DCTSIZE2] = {
+    17,  18,  24,  47,  99,  99,  99,  99,
+    18,  21,  26,  66,  99,  99,  99,  99,
+    24,  26,  56,  99,  99,  99,  99,  99,
+    47,  66,  99,  99,  99,  99,  99,  99,
+    99,  99,  99,  99,  99,  99,  99,  99,
+    99,  99,  99,  99,  99,  99,  99,  99,
+    99,  99,  99,  99,  99,  99,  99,  99,
+    99,  99,  99,  99,  99,  99,  99,  99
+  };
+
+  /* Set up two quantization tables using the specified scaling */
+  jpeg_add_quant_table(cinfo, 0, std_luminance_quant_tbl,
+		       scale_factor, force_baseline);
+  jpeg_add_quant_table(cinfo, 1, std_chrominance_quant_tbl,
+		       scale_factor, force_baseline);
+}
+
+
+GLOBAL(int)
+jpeg_quality_scaling (int quality)
+/* Convert a user-specified quality rating to a percentage scaling factor
+ * for an underlying quantization table, using our recommended scaling curve.
+ * The input 'quality' factor should be 0 (terrible) to 100 (very good).
+ */
+{
+  /* Safety limit on quality factor.  Convert 0 to 1 to avoid zero divide. */
+  if (quality <= 0) quality = 1;
+  if (quality > 100) quality = 100;
+
+  /* The basic table is used as-is (scaling 100) for a quality of 50.
+   * Qualities 50..100 are converted to scaling percentage 200 - 2*Q;
+   * note that at Q=100 the scaling is 0, which will cause jpeg_add_quant_table
+   * to make all the table entries 1 (hence, minimum quantization loss).
+   * Qualities 1..50 are converted to scaling percentage 5000/Q.
+   */
+  if (quality < 50)
+    quality = 5000 / quality;
+  else
+    quality = 200 - quality*2;
+
+  return quality;
+}
+
+
+GLOBAL(void)
+jpeg_set_quality (j_compress_ptr cinfo, int quality, boolean force_baseline)
+/* Set or change the 'quality' (quantization) setting, using default tables.
+ * This is the standard quality-adjusting entry point for typical user
+ * interfaces; only those who want detailed control over quantization tables
+ * would use the preceding three routines directly.
+ */
+{
+  /* Convert user 0-100 rating to percentage scaling */
+  quality = jpeg_quality_scaling(quality);
+
+  /* Set up standard quality tables */
+  jpeg_set_linear_quality(cinfo, quality, force_baseline);
+}
+
+
+/*
+ * Huffman table setup routines
+ */
+
+LOCAL(void)
+add_huff_table (j_compress_ptr cinfo,
+		JHUFF_TBL **htblptr, const UINT8 *bits, const UINT8 *val)
+/* Define a Huffman table */
+{
+  int nsymbols, len;
+
+  if (*htblptr == NULL)
+    *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+
+  /* Copy the number-of-symbols-of-each-code-length counts */
+  MEMCOPY((*htblptr)->bits, bits, SIZEOF((*htblptr)->bits));
+
+  /* Validate the counts.  We do this here mainly so we can copy the right
+   * number of symbols from the val[] array, without risking marching off
+   * the end of memory.  jchuff.c will do a more thorough test later.
+   */
+  nsymbols = 0;
+  for (len = 1; len <= 16; len++)
+    nsymbols += bits[len];
+  if (nsymbols < 1 || nsymbols > 256)
+    ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+
+  MEMCOPY((*htblptr)->huffval, val, nsymbols * SIZEOF(UINT8));
+
+  /* Initialize sent_table FALSE so table will be written to JPEG file. */
+  (*htblptr)->sent_table = FALSE;
+}
+
+
+LOCAL(void)
+std_huff_tables (j_compress_ptr cinfo)
+/* Set up the standard Huffman tables (cf. JPEG standard section K.3) */
+/* IMPORTANT: these are only valid for 8-bit data precision! */
+{
+  static const UINT8 bits_dc_luminance[17] =
+    { /* 0-base */ 0, 0, 1, 5, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0 };
+  static const UINT8 val_dc_luminance[] =
+    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
+  
+  static const UINT8 bits_dc_chrominance[17] =
+    { /* 0-base */ 0, 0, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 };
+  static const UINT8 val_dc_chrominance[] =
+    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
+  
+  static const UINT8 bits_ac_luminance[17] =
+    { /* 0-base */ 0, 0, 2, 1, 3, 3, 2, 4, 3, 5, 5, 4, 4, 0, 0, 1, 0x7d };
+  static const UINT8 val_ac_luminance[] =
+    { 0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12,
+      0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07,
+      0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xa1, 0x08,
+      0x23, 0x42, 0xb1, 0xc1, 0x15, 0x52, 0xd1, 0xf0,
+      0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0a, 0x16,
+      0x17, 0x18, 0x19, 0x1a, 0x25, 0x26, 0x27, 0x28,
+      0x29, 0x2a, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
+      0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
+      0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
+      0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
+      0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
+      0x7a, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
+      0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
+      0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
+      0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6,
+      0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, 0xc4, 0xc5,
+      0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, 0xd3, 0xd4,
+      0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xe1, 0xe2,
+      0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea,
+      0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+      0xf9, 0xfa };
+  
+  static const UINT8 bits_ac_chrominance[17] =
+    { /* 0-base */ 0, 0, 2, 1, 2, 4, 4, 3, 4, 7, 5, 4, 4, 0, 1, 2, 0x77 };
+  static const UINT8 val_ac_chrominance[] =
+    { 0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21,
+      0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71,
+      0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91,
+      0xa1, 0xb1, 0xc1, 0x09, 0x23, 0x33, 0x52, 0xf0,
+      0x15, 0x62, 0x72, 0xd1, 0x0a, 0x16, 0x24, 0x34,
+      0xe1, 0x25, 0xf1, 0x17, 0x18, 0x19, 0x1a, 0x26,
+      0x27, 0x28, 0x29, 0x2a, 0x35, 0x36, 0x37, 0x38,
+      0x39, 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48,
+      0x49, 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
+      0x59, 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
+      0x69, 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78,
+      0x79, 0x7a, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
+      0x88, 0x89, 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96,
+      0x97, 0x98, 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5,
+      0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4,
+      0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3,
+      0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2,
+      0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda,
+      0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9,
+      0xea, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+      0xf9, 0xfa };
+  
+  add_huff_table(cinfo, &cinfo->dc_huff_tbl_ptrs[0],
+		 bits_dc_luminance, val_dc_luminance);
+  add_huff_table(cinfo, &cinfo->ac_huff_tbl_ptrs[0],
+		 bits_ac_luminance, val_ac_luminance);
+  add_huff_table(cinfo, &cinfo->dc_huff_tbl_ptrs[1],
+		 bits_dc_chrominance, val_dc_chrominance);
+  add_huff_table(cinfo, &cinfo->ac_huff_tbl_ptrs[1],
+		 bits_ac_chrominance, val_ac_chrominance);
+}
+
+
+/*
+ * Default parameter setup for compression.
+ *
+ * Applications that don't choose to use this routine must do their
+ * own setup of all these parameters.  Alternately, you can call this
+ * to establish defaults and then alter parameters selectively.  This
+ * is the recommended approach since, if we add any new parameters,
+ * your code will still work (they'll be set to reasonable defaults).
+ */
+
+GLOBAL(void)
+jpeg_set_defaults (j_compress_ptr cinfo)
+{
+  int i;
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* Allocate comp_info array large enough for maximum component count.
+   * Array is made permanent in case application wants to compress
+   * multiple images at same param settings.
+   */
+  if (cinfo->comp_info == NULL)
+    cinfo->comp_info = (jpeg_component_info *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  MAX_COMPONENTS * SIZEOF(jpeg_component_info));
+
+  /* Initialize everything not dependent on the color space */
+
+  cinfo->data_precision = BITS_IN_JSAMPLE;
+  /* Set up two quantization tables using default quality of 75 */
+  jpeg_set_quality(cinfo, 75, TRUE);
+  /* Set up two Huffman tables */
+  std_huff_tables(cinfo);
+
+  /* Initialize default arithmetic coding conditioning */
+  for (i = 0; i < NUM_ARITH_TBLS; i++) {
+    cinfo->arith_dc_L[i] = 0;
+    cinfo->arith_dc_U[i] = 1;
+    cinfo->arith_ac_K[i] = 5;
+  }
+
+  /* Default is no multiple-scan output */
+  cinfo->scan_info = NULL;
+  cinfo->num_scans = 0;
+
+  /* Expect normal source image, not raw downsampled data */
+  cinfo->raw_data_in = FALSE;
+
+  /* Use Huffman coding, not arithmetic coding, by default */
+  cinfo->arith_code = FALSE;
+
+  /* By default, don't do extra passes to optimize entropy coding */
+  cinfo->optimize_coding = FALSE;
+  /* The standard Huffman tables are only valid for 8-bit data precision.
+   * If the precision is higher, force optimization on so that usable
+   * tables will be computed.  This test can be removed if default tables
+   * are supplied that are valid for the desired precision.
+   */
+  if (cinfo->data_precision > 8)
+    cinfo->optimize_coding = TRUE;
+
+  /* By default, use the simpler non-cosited sampling alignment */
+  cinfo->CCIR601_sampling = FALSE;
+
+  /* No input smoothing */
+  cinfo->smoothing_factor = 0;
+
+  /* DCT algorithm preference */
+  cinfo->dct_method = JDCT_DEFAULT;
+
+  /* No restart markers */
+  cinfo->restart_interval = 0;
+  cinfo->restart_in_rows = 0;
+
+  /* Fill in default JFIF marker parameters.  Note that whether the marker
+   * will actually be written is determined by jpeg_set_colorspace.
+   *
+   * By default, the library emits JFIF version code 1.01.
+   * An application that wants to emit JFIF 1.02 extension markers should set
+   * JFIF_minor_version to 2.  We could probably get away with just defaulting
+   * to 1.02, but there may still be some decoders in use that will complain
+   * about that; saying 1.01 should minimize compatibility problems.
+   */
+  cinfo->JFIF_major_version = 1; /* Default JFIF version = 1.01 */
+  cinfo->JFIF_minor_version = 1;
+  cinfo->density_unit = 0;	/* Pixel size is unknown by default */
+  cinfo->X_density = 1;		/* Pixel aspect ratio is square by default */
+  cinfo->Y_density = 1;
+
+  /* Choose JPEG colorspace based on input space, set defaults accordingly */
+
+  jpeg_default_colorspace(cinfo);
+}
+
+
+/*
+ * Select an appropriate JPEG colorspace for in_color_space.
+ */
+
+GLOBAL(void)
+jpeg_default_colorspace (j_compress_ptr cinfo)
+{
+  switch (cinfo->in_color_space) {
+  case JCS_GRAYSCALE:
+    jpeg_set_colorspace(cinfo, JCS_GRAYSCALE);
+    break;
+  case JCS_RGB:
+    jpeg_set_colorspace(cinfo, JCS_YCbCr);
+    break;
+  case JCS_YCbCr:
+    jpeg_set_colorspace(cinfo, JCS_YCbCr);
+    break;
+  case JCS_CMYK:
+    jpeg_set_colorspace(cinfo, JCS_CMYK); /* By default, no translation */
+    break;
+  case JCS_YCCK:
+    jpeg_set_colorspace(cinfo, JCS_YCCK);
+    break;
+  case JCS_UNKNOWN:
+    jpeg_set_colorspace(cinfo, JCS_UNKNOWN);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+  }
+}
+
+
+/*
+ * Set the JPEG colorspace, and choose colorspace-dependent default values.
+ */
+
+GLOBAL(void)
+jpeg_set_colorspace (j_compress_ptr cinfo, J_COLOR_SPACE colorspace)
+{
+  jpeg_component_info * compptr;
+  int ci;
+
+#define SET_COMP(index,id,hsamp,vsamp,quant,dctbl,actbl)  \
+  (compptr = &cinfo->comp_info[index], \
+   compptr->component_id = (id), \
+   compptr->h_samp_factor = (hsamp), \
+   compptr->v_samp_factor = (vsamp), \
+   compptr->quant_tbl_no = (quant), \
+   compptr->dc_tbl_no = (dctbl), \
+   compptr->ac_tbl_no = (actbl) )
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* For all colorspaces, we use Q and Huff tables 0 for luminance components,
+   * tables 1 for chrominance components.
+   */
+
+  cinfo->jpeg_color_space = colorspace;
+
+  cinfo->write_JFIF_header = FALSE; /* No marker for non-JFIF colorspaces */
+  cinfo->write_Adobe_marker = FALSE; /* write no Adobe marker by default */
+
+  switch (colorspace) {
+  case JCS_GRAYSCALE:
+    cinfo->write_JFIF_header = TRUE; /* Write a JFIF marker */
+    cinfo->num_components = 1;
+    /* JFIF specifies component ID 1 */
+    SET_COMP(0, 1, 1,1, 0, 0,0);
+    break;
+  case JCS_RGB:
+    cinfo->write_Adobe_marker = TRUE; /* write Adobe marker to flag RGB */
+    cinfo->num_components = 3;
+    SET_COMP(0, 0x52 /* 'R' */, 1,1, 0, 0,0);
+    SET_COMP(1, 0x47 /* 'G' */, 1,1, 0, 0,0);
+    SET_COMP(2, 0x42 /* 'B' */, 1,1, 0, 0,0);
+    break;
+  case JCS_YCbCr:
+    cinfo->write_JFIF_header = TRUE; /* Write a JFIF marker */
+    cinfo->num_components = 3;
+    /* JFIF specifies component IDs 1,2,3 */
+    /* We default to 2x2 subsamples of chrominance */
+    SET_COMP(0, 1, 2,2, 0, 0,0);
+    SET_COMP(1, 2, 1,1, 1, 1,1);
+    SET_COMP(2, 3, 1,1, 1, 1,1);
+    break;
+  case JCS_CMYK:
+    cinfo->write_Adobe_marker = TRUE; /* write Adobe marker to flag CMYK */
+    cinfo->num_components = 4;
+    SET_COMP(0, 0x43 /* 'C' */, 1,1, 0, 0,0);
+    SET_COMP(1, 0x4D /* 'M' */, 1,1, 0, 0,0);
+    SET_COMP(2, 0x59 /* 'Y' */, 1,1, 0, 0,0);
+    SET_COMP(3, 0x4B /* 'K' */, 1,1, 0, 0,0);
+    break;
+  case JCS_YCCK:
+    cinfo->write_Adobe_marker = TRUE; /* write Adobe marker to flag YCCK */
+    cinfo->num_components = 4;
+    SET_COMP(0, 1, 2,2, 0, 0,0);
+    SET_COMP(1, 2, 1,1, 1, 1,1);
+    SET_COMP(2, 3, 1,1, 1, 1,1);
+    SET_COMP(3, 4, 2,2, 0, 0,0);
+    break;
+  case JCS_UNKNOWN:
+    cinfo->num_components = cinfo->input_components;
+    if (cinfo->num_components < 1 || cinfo->num_components > MAX_COMPONENTS)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
+	       MAX_COMPONENTS);
+    for (ci = 0; ci < cinfo->num_components; ci++) {
+      SET_COMP(ci, ci, 1,1, 0, 0,0);
+    }
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+  }
+}
+
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+
+LOCAL(jpeg_scan_info *)
+fill_a_scan (jpeg_scan_info * scanptr, int ci,
+	     int Ss, int Se, int Ah, int Al)
+/* Support routine: generate one scan for specified component */
+{
+  scanptr->comps_in_scan = 1;
+  scanptr->component_index[0] = ci;
+  scanptr->Ss = Ss;
+  scanptr->Se = Se;
+  scanptr->Ah = Ah;
+  scanptr->Al = Al;
+  scanptr++;
+  return scanptr;
+}
+
+LOCAL(jpeg_scan_info *)
+fill_scans (jpeg_scan_info * scanptr, int ncomps,
+	    int Ss, int Se, int Ah, int Al)
+/* Support routine: generate one scan for each component */
+{
+  int ci;
+
+  for (ci = 0; ci < ncomps; ci++) {
+    scanptr->comps_in_scan = 1;
+    scanptr->component_index[0] = ci;
+    scanptr->Ss = Ss;
+    scanptr->Se = Se;
+    scanptr->Ah = Ah;
+    scanptr->Al = Al;
+    scanptr++;
+  }
+  return scanptr;
+}
+
+LOCAL(jpeg_scan_info *)
+fill_dc_scans (jpeg_scan_info * scanptr, int ncomps, int Ah, int Al)
+/* Support routine: generate interleaved DC scan if possible, else N scans */
+{
+  int ci;
+
+  if (ncomps <= MAX_COMPS_IN_SCAN) {
+    /* Single interleaved DC scan */
+    scanptr->comps_in_scan = ncomps;
+    for (ci = 0; ci < ncomps; ci++)
+      scanptr->component_index[ci] = ci;
+    scanptr->Ss = scanptr->Se = 0;
+    scanptr->Ah = Ah;
+    scanptr->Al = Al;
+    scanptr++;
+  } else {
+    /* Noninterleaved DC scan for each component */
+    scanptr = fill_scans(scanptr, ncomps, 0, 0, Ah, Al);
+  }
+  return scanptr;
+}
+
+
+/*
+ * Create a recommended progressive-JPEG script.
+ * cinfo->num_components and cinfo->jpeg_color_space must be correct.
+ */
+
+GLOBAL(void)
+jpeg_simple_progression (j_compress_ptr cinfo)
+{
+  int ncomps = cinfo->num_components;
+  int nscans;
+  jpeg_scan_info * scanptr;
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* Figure space needed for script.  Calculation must match code below! */
+  if (ncomps == 3 && cinfo->jpeg_color_space == JCS_YCbCr) {
+    /* Custom script for YCbCr color images. */
+    nscans = 10;
+  } else {
+    /* All-purpose script for other color spaces. */
+    if (ncomps > MAX_COMPS_IN_SCAN)
+      nscans = 6 * ncomps;	/* 2 DC + 4 AC scans per component */
+    else
+      nscans = 2 + 4 * ncomps;	/* 2 DC scans; 4 AC scans per component */
+  }
+
+  /* Allocate space for script.
+   * We need to put it in the permanent pool in case the application performs
+   * multiple compressions without changing the settings.  To avoid a memory
+   * leak if jpeg_simple_progression is called repeatedly for the same JPEG
+   * object, we try to re-use previously allocated space, and we allocate
+   * enough space to handle YCbCr even if initially asked for grayscale.
+   */
+  if (cinfo->script_space == NULL || cinfo->script_space_size < nscans) {
+    cinfo->script_space_size = MAX(nscans, 10);
+    cinfo->script_space = (jpeg_scan_info *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+			cinfo->script_space_size * SIZEOF(jpeg_scan_info));
+  }
+  scanptr = cinfo->script_space;
+  cinfo->scan_info = scanptr;
+  cinfo->num_scans = nscans;
+
+  if (ncomps == 3 && cinfo->jpeg_color_space == JCS_YCbCr) {
+    /* Custom script for YCbCr color images. */
+    /* Initial DC scan */
+    scanptr = fill_dc_scans(scanptr, ncomps, 0, 1);
+    /* Initial AC scan: get some luma data out in a hurry */
+    scanptr = fill_a_scan(scanptr, 0, 1, 5, 0, 2);
+    /* Chroma data is too small to be worth expending many scans on */
+    scanptr = fill_a_scan(scanptr, 2, 1, 63, 0, 1);
+    scanptr = fill_a_scan(scanptr, 1, 1, 63, 0, 1);
+    /* Complete spectral selection for luma AC */
+    scanptr = fill_a_scan(scanptr, 0, 6, 63, 0, 2);
+    /* Refine next bit of luma AC */
+    scanptr = fill_a_scan(scanptr, 0, 1, 63, 2, 1);
+    /* Finish DC successive approximation */
+    scanptr = fill_dc_scans(scanptr, ncomps, 1, 0);
+    /* Finish AC successive approximation */
+    scanptr = fill_a_scan(scanptr, 2, 1, 63, 1, 0);
+    scanptr = fill_a_scan(scanptr, 1, 1, 63, 1, 0);
+    /* Luma bottom bit comes last since it's usually largest scan */
+    scanptr = fill_a_scan(scanptr, 0, 1, 63, 1, 0);
+  } else {
+    /* All-purpose script for other color spaces. */
+    /* Successive approximation first pass */
+    scanptr = fill_dc_scans(scanptr, ncomps, 0, 1);
+    scanptr = fill_scans(scanptr, ncomps, 1, 5, 0, 2);
+    scanptr = fill_scans(scanptr, ncomps, 6, 63, 0, 2);
+    /* Successive approximation second pass */
+    scanptr = fill_scans(scanptr, ncomps, 1, 63, 2, 1);
+    /* Successive approximation final pass */
+    scanptr = fill_dc_scans(scanptr, ncomps, 1, 0);
+    scanptr = fill_scans(scanptr, ncomps, 1, 63, 1, 0);
+  }
+#ifdef NEEDS_ARM_ERRATA_754319_754320
+  asm volatile ( "vmov s0,s0\n" );
+#endif
+
+}
+
+#endif /* C_PROGRESSIVE_SUPPORTED */
diff --git a/libjpegtwrp/jcphuff.c b/libjpegtwrp/jcphuff.c
new file mode 100644
index 000000000..07f9178b0
--- /dev/null
+++ b/libjpegtwrp/jcphuff.c
@@ -0,0 +1,833 @@
+/*
+ * jcphuff.c
+ *
+ * Copyright (C) 1995-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains Huffman entropy encoding routines for progressive JPEG.
+ *
+ * We do not support output suspension in this module, since the library
+ * currently does not allow multiple-scan files to be written with output
+ * suspension.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jchuff.h"		/* Declarations shared with jchuff.c */
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+
+/* Expanded entropy encoder object for progressive Huffman encoding. */
+
+typedef struct {
+  struct jpeg_entropy_encoder pub; /* public fields */
+
+  /* Mode flag: TRUE for optimization, FALSE for actual data output */
+  boolean gather_statistics;
+
+  /* Bit-level coding status.
+   * next_output_byte/free_in_buffer are local copies of cinfo->dest fields.
+   */
+  JOCTET * next_output_byte;	/* => next byte to write in buffer */
+  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
+  INT32 put_buffer;		/* current bit-accumulation buffer */
+  int put_bits;			/* # of bits now in it */
+  j_compress_ptr cinfo;		/* link to cinfo (needed for dump_buffer) */
+
+  /* Coding status for DC components */
+  int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
+
+  /* Coding status for AC components */
+  int ac_tbl_no;		/* the table number of the single component */
+  unsigned int EOBRUN;		/* run length of EOBs */
+  unsigned int BE;		/* # of buffered correction bits before MCU */
+  char * bit_buffer;		/* buffer for correction bits (1 per char) */
+  /* packing correction bits tightly would save some space but cost time... */
+
+  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+  int next_restart_num;		/* next restart number to write (0-7) */
+
+  /* Pointers to derived tables (these workspaces have image lifespan).
+   * Since any one scan codes only DC or only AC, we only need one set
+   * of tables, not one for DC and one for AC.
+   */
+  c_derived_tbl * derived_tbls[NUM_HUFF_TBLS];
+
+  /* Statistics tables for optimization; again, one set is enough */
+  long * count_ptrs[NUM_HUFF_TBLS];
+} phuff_entropy_encoder;
+
+typedef phuff_entropy_encoder * phuff_entropy_ptr;
+
+/* MAX_CORR_BITS is the number of bits the AC refinement correction-bit
+ * buffer can hold.  Larger sizes may slightly improve compression, but
+ * 1000 is already well into the realm of overkill.
+ * The minimum safe size is 64 bits.
+ */
+
+#define MAX_CORR_BITS  1000	/* Max # of correction bits I can buffer */
+
+/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than INT32.
+ * We assume that int right shift is unsigned if INT32 right shift is,
+ * which should be safe.
+ */
+
+#ifdef RIGHT_SHIFT_IS_UNSIGNED
+#define ISHIFT_TEMPS	int ishift_temp;
+#define IRIGHT_SHIFT(x,shft)  \
+	((ishift_temp = (x)) < 0 ? \
+	 (ishift_temp >> (shft)) | ((~0) << (16-(shft))) : \
+	 (ishift_temp >> (shft)))
+#else
+#define ISHIFT_TEMPS
+#define IRIGHT_SHIFT(x,shft)	((x) >> (shft))
+#endif
+
+/* Forward declarations */
+METHODDEF(boolean) encode_mcu_DC_first JPP((j_compress_ptr cinfo,
+					    JBLOCKROW *MCU_data));
+METHODDEF(boolean) encode_mcu_AC_first JPP((j_compress_ptr cinfo,
+					    JBLOCKROW *MCU_data));
+METHODDEF(boolean) encode_mcu_DC_refine JPP((j_compress_ptr cinfo,
+					     JBLOCKROW *MCU_data));
+METHODDEF(boolean) encode_mcu_AC_refine JPP((j_compress_ptr cinfo,
+					     JBLOCKROW *MCU_data));
+METHODDEF(void) finish_pass_phuff JPP((j_compress_ptr cinfo));
+METHODDEF(void) finish_pass_gather_phuff JPP((j_compress_ptr cinfo));
+
+
+/*
+ * Initialize for a Huffman-compressed scan using progressive JPEG.
+ */
+
+METHODDEF(void)
+start_pass_phuff (j_compress_ptr cinfo, boolean gather_statistics)
+{  
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  boolean is_DC_band;
+  int ci, tbl;
+  jpeg_component_info * compptr;
+
+  entropy->cinfo = cinfo;
+  entropy->gather_statistics = gather_statistics;
+
+  is_DC_band = (cinfo->Ss == 0);
+
+  /* We assume jcmaster.c already validated the scan parameters. */
+
+  /* Select execution routines */
+  if (cinfo->Ah == 0) {
+    if (is_DC_band)
+      entropy->pub.encode_mcu = encode_mcu_DC_first;
+    else
+      entropy->pub.encode_mcu = encode_mcu_AC_first;
+  } else {
+    if (is_DC_band)
+      entropy->pub.encode_mcu = encode_mcu_DC_refine;
+    else {
+      entropy->pub.encode_mcu = encode_mcu_AC_refine;
+      /* AC refinement needs a correction bit buffer */
+      if (entropy->bit_buffer == NULL)
+	entropy->bit_buffer = (char *)
+	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				      MAX_CORR_BITS * SIZEOF(char));
+    }
+  }
+  if (gather_statistics)
+    entropy->pub.finish_pass = finish_pass_gather_phuff;
+  else
+    entropy->pub.finish_pass = finish_pass_phuff;
+
+  /* Only DC coefficients may be interleaved, so cinfo->comps_in_scan = 1
+   * for AC coefficients.
+   */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    /* Initialize DC predictions to 0 */
+    entropy->last_dc_val[ci] = 0;
+    /* Get table index */
+    if (is_DC_band) {
+      if (cinfo->Ah != 0)	/* DC refinement needs no table */
+	continue;
+      tbl = compptr->dc_tbl_no;
+    } else {
+      entropy->ac_tbl_no = tbl = compptr->ac_tbl_no;
+    }
+    if (gather_statistics) {
+      /* Check for invalid table index */
+      /* (make_c_derived_tbl does this in the other path) */
+      if (tbl < 0 || tbl >= NUM_HUFF_TBLS)
+        ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tbl);
+      /* Allocate and zero the statistics tables */
+      /* Note that jpeg_gen_optimal_table expects 257 entries in each table! */
+      if (entropy->count_ptrs[tbl] == NULL)
+	entropy->count_ptrs[tbl] = (long *)
+	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				      257 * SIZEOF(long));
+      MEMZERO(entropy->count_ptrs[tbl], 257 * SIZEOF(long));
+    } else {
+      /* Compute derived values for Huffman table */
+      /* We may do this more than once for a table, but it's not expensive */
+      jpeg_make_c_derived_tbl(cinfo, is_DC_band, tbl,
+			      & entropy->derived_tbls[tbl]);
+    }
+  }
+
+  /* Initialize AC stuff */
+  entropy->EOBRUN = 0;
+  entropy->BE = 0;
+
+  /* Initialize bit buffer to empty */
+  entropy->put_buffer = 0;
+  entropy->put_bits = 0;
+
+  /* Initialize restart stuff */
+  entropy->restarts_to_go = cinfo->restart_interval;
+  entropy->next_restart_num = 0;
+}
+
+
+/* Outputting bytes to the file.
+ * NB: these must be called only when actually outputting,
+ * that is, entropy->gather_statistics == FALSE.
+ */
+
+/* Emit a byte */
+#define emit_byte(entropy,val)  \
+	{ *(entropy)->next_output_byte++ = (JOCTET) (val);  \
+	  if (--(entropy)->free_in_buffer == 0)  \
+	    dump_buffer(entropy); }
+
+
+LOCAL(void)
+dump_buffer (phuff_entropy_ptr entropy)
+/* Empty the output buffer; we do not support suspension in this module. */
+{
+  struct jpeg_destination_mgr * dest = entropy->cinfo->dest;
+
+  if (! (*dest->empty_output_buffer) (entropy->cinfo))
+    ERREXIT(entropy->cinfo, JERR_CANT_SUSPEND);
+  /* After a successful buffer dump, must reset buffer pointers */
+  entropy->next_output_byte = dest->next_output_byte;
+  entropy->free_in_buffer = dest->free_in_buffer;
+}
+
+
+/* Outputting bits to the file */
+
+/* Only the right 24 bits of put_buffer are used; the valid bits are
+ * left-justified in this part.  At most 16 bits can be passed to emit_bits
+ * in one call, and we never retain more than 7 bits in put_buffer
+ * between calls, so 24 bits are sufficient.
+ */
+
+INLINE
+LOCAL(void)
+emit_bits (phuff_entropy_ptr entropy, unsigned int code, int size)
+/* Emit some bits, unless we are in gather mode */
+{
+  /* This routine is heavily used, so it's worth coding tightly. */
+  register INT32 put_buffer = (INT32) code;
+  register int put_bits = entropy->put_bits;
+
+  /* if size is 0, caller used an invalid Huffman table entry */
+  if (size == 0)
+    ERREXIT(entropy->cinfo, JERR_HUFF_MISSING_CODE);
+
+  if (entropy->gather_statistics)
+    return;			/* do nothing if we're only getting stats */
+
+  put_buffer &= (((INT32) 1)<<size) - 1; /* mask off any extra bits in code */
+  
+  put_bits += size;		/* new number of bits in buffer */
+  
+  put_buffer <<= 24 - put_bits; /* align incoming bits */
+
+  put_buffer |= entropy->put_buffer; /* and merge with old buffer contents */
+
+  while (put_bits >= 8) {
+    int c = (int) ((put_buffer >> 16) & 0xFF);
+    
+    emit_byte(entropy, c);
+    if (c == 0xFF) {		/* need to stuff a zero byte? */
+      emit_byte(entropy, 0);
+    }
+    put_buffer <<= 8;
+    put_bits -= 8;
+  }
+
+  entropy->put_buffer = put_buffer; /* update variables */
+  entropy->put_bits = put_bits;
+}
+
+
+LOCAL(void)
+flush_bits (phuff_entropy_ptr entropy)
+{
+  emit_bits(entropy, 0x7F, 7); /* fill any partial byte with ones */
+  entropy->put_buffer = 0;     /* and reset bit-buffer to empty */
+  entropy->put_bits = 0;
+}
+
+
+/*
+ * Emit (or just count) a Huffman symbol.
+ */
+
+INLINE
+LOCAL(void)
+emit_symbol (phuff_entropy_ptr entropy, int tbl_no, int symbol)
+{
+  if (entropy->gather_statistics)
+    entropy->count_ptrs[tbl_no][symbol]++;
+  else {
+    c_derived_tbl * tbl = entropy->derived_tbls[tbl_no];
+    emit_bits(entropy, tbl->ehufco[symbol], tbl->ehufsi[symbol]);
+  }
+}
+
+
+/*
+ * Emit bits from a correction bit buffer.
+ */
+
+LOCAL(void)
+emit_buffered_bits (phuff_entropy_ptr entropy, char * bufstart,
+		    unsigned int nbits)
+{
+  if (entropy->gather_statistics)
+    return;			/* no real work */
+
+  while (nbits > 0) {
+    emit_bits(entropy, (unsigned int) (*bufstart), 1);
+    bufstart++;
+    nbits--;
+  }
+}
+
+
+/*
+ * Emit any pending EOBRUN symbol.
+ */
+
+LOCAL(void)
+emit_eobrun (phuff_entropy_ptr entropy)
+{
+  register int temp, nbits;
+
+  if (entropy->EOBRUN > 0) {	/* if there is any pending EOBRUN */
+    temp = entropy->EOBRUN;
+    nbits = 0;
+    while ((temp >>= 1))
+      nbits++;
+    /* safety check: shouldn't happen given limited correction-bit buffer */
+    if (nbits > 14)
+      ERREXIT(entropy->cinfo, JERR_HUFF_MISSING_CODE);
+
+    emit_symbol(entropy, entropy->ac_tbl_no, nbits << 4);
+    if (nbits)
+      emit_bits(entropy, entropy->EOBRUN, nbits);
+
+    entropy->EOBRUN = 0;
+
+    /* Emit any buffered correction bits */
+    emit_buffered_bits(entropy, entropy->bit_buffer, entropy->BE);
+    entropy->BE = 0;
+  }
+}
+
+
+/*
+ * Emit a restart marker & resynchronize predictions.
+ */
+
+LOCAL(void)
+emit_restart (phuff_entropy_ptr entropy, int restart_num)
+{
+  int ci;
+
+  emit_eobrun(entropy);
+
+  if (! entropy->gather_statistics) {
+    flush_bits(entropy);
+    emit_byte(entropy, 0xFF);
+    emit_byte(entropy, JPEG_RST0 + restart_num);
+  }
+
+  if (entropy->cinfo->Ss == 0) {
+    /* Re-initialize DC predictions to 0 */
+    for (ci = 0; ci < entropy->cinfo->comps_in_scan; ci++)
+      entropy->last_dc_val[ci] = 0;
+  } else {
+    /* Re-initialize all AC-related fields to 0 */
+    entropy->EOBRUN = 0;
+    entropy->BE = 0;
+  }
+}
+
+
+/*
+ * MCU encoding for DC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+encode_mcu_DC_first (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  register int temp, temp2;
+  register int nbits;
+  int blkn, ci;
+  int Al = cinfo->Al;
+  JBLOCKROW block;
+  jpeg_component_info * compptr;
+  ISHIFT_TEMPS
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval)
+    if (entropy->restarts_to_go == 0)
+      emit_restart(entropy, entropy->next_restart_num);
+
+  /* Encode the MCU data blocks */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_data[blkn];
+    ci = cinfo->MCU_membership[blkn];
+    compptr = cinfo->cur_comp_info[ci];
+
+    /* Compute the DC value after the required point transform by Al.
+     * This is simply an arithmetic right shift.
+     */
+    temp2 = IRIGHT_SHIFT((int) ((*block)[0]), Al);
+
+    /* DC differences are figured on the point-transformed values. */
+    temp = temp2 - entropy->last_dc_val[ci];
+    entropy->last_dc_val[ci] = temp2;
+
+    /* Encode the DC coefficient difference per section G.1.2.1 */
+    temp2 = temp;
+    if (temp < 0) {
+      temp = -temp;		/* temp is abs value of input */
+      /* For a negative input, want temp2 = bitwise complement of abs(input) */
+      /* This code assumes we are on a two's complement machine */
+      temp2--;
+    }
+    
+    /* Find the number of bits needed for the magnitude of the coefficient */
+    nbits = 0;
+    while (temp) {
+      nbits++;
+      temp >>= 1;
+    }
+    /* Check for out-of-range coefficient values.
+     * Since we're encoding a difference, the range limit is twice as much.
+     */
+    if (nbits > MAX_COEF_BITS+1)
+      ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+    
+    /* Count/emit the Huffman-coded symbol for the number of bits */
+    emit_symbol(entropy, compptr->dc_tbl_no, nbits);
+    
+    /* Emit that number of bits of the value, if positive, */
+    /* or the complement of its magnitude, if negative. */
+    if (nbits)			/* emit_bits rejects calls with size 0 */
+      emit_bits(entropy, (unsigned int) temp2, nbits);
+  }
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU encoding for AC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+encode_mcu_AC_first (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  register int temp, temp2;
+  register int nbits;
+  register int r, k;
+  int Se = cinfo->Se;
+  int Al = cinfo->Al;
+  JBLOCKROW block;
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval)
+    if (entropy->restarts_to_go == 0)
+      emit_restart(entropy, entropy->next_restart_num);
+
+  /* Encode the MCU data block */
+  block = MCU_data[0];
+
+  /* Encode the AC coefficients per section G.1.2.2, fig. G.3 */
+  
+  r = 0;			/* r = run length of zeros */
+   
+  for (k = cinfo->Ss; k <= Se; k++) {
+    if ((temp = (*block)[jpeg_natural_order[k]]) == 0) {
+      r++;
+      continue;
+    }
+    /* We must apply the point transform by Al.  For AC coefficients this
+     * is an integer division with rounding towards 0.  To do this portably
+     * in C, we shift after obtaining the absolute value; so the code is
+     * interwoven with finding the abs value (temp) and output bits (temp2).
+     */
+    if (temp < 0) {
+      temp = -temp;		/* temp is abs value of input */
+      temp >>= Al;		/* apply the point transform */
+      /* For a negative coef, want temp2 = bitwise complement of abs(coef) */
+      temp2 = ~temp;
+    } else {
+      temp >>= Al;		/* apply the point transform */
+      temp2 = temp;
+    }
+    /* Watch out for case that nonzero coef is zero after point transform */
+    if (temp == 0) {
+      r++;
+      continue;
+    }
+
+    /* Emit any pending EOBRUN */
+    if (entropy->EOBRUN > 0)
+      emit_eobrun(entropy);
+    /* if run length > 15, must emit special run-length-16 codes (0xF0) */
+    while (r > 15) {
+      emit_symbol(entropy, entropy->ac_tbl_no, 0xF0);
+      r -= 16;
+    }
+
+    /* Find the number of bits needed for the magnitude of the coefficient */
+    nbits = 1;			/* there must be at least one 1 bit */
+    while ((temp >>= 1))
+      nbits++;
+    /* Check for out-of-range coefficient values */
+    if (nbits > MAX_COEF_BITS)
+      ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+
+    /* Count/emit Huffman symbol for run length / number of bits */
+    emit_symbol(entropy, entropy->ac_tbl_no, (r << 4) + nbits);
+
+    /* Emit that number of bits of the value, if positive, */
+    /* or the complement of its magnitude, if negative. */
+    emit_bits(entropy, (unsigned int) temp2, nbits);
+
+    r = 0;			/* reset zero run length */
+  }
+
+  if (r > 0) {			/* If there are trailing zeroes, */
+    entropy->EOBRUN++;		/* count an EOB */
+    if (entropy->EOBRUN == 0x7FFF)
+      emit_eobrun(entropy);	/* force it out to avoid overflow */
+  }
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU encoding for DC successive approximation refinement scan.
+ * Note: we assume such scans can be multi-component, although the spec
+ * is not very clear on the point.
+ */
+
+METHODDEF(boolean)
+encode_mcu_DC_refine (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  register int temp;
+  int blkn;
+  int Al = cinfo->Al;
+  JBLOCKROW block;
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval)
+    if (entropy->restarts_to_go == 0)
+      emit_restart(entropy, entropy->next_restart_num);
+
+  /* Encode the MCU data blocks */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_data[blkn];
+
+    /* We simply emit the Al'th bit of the DC coefficient value. */
+    temp = (*block)[0];
+    emit_bits(entropy, (unsigned int) (temp >> Al), 1);
+  }
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU encoding for AC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+encode_mcu_AC_refine (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  register int temp;
+  register int r, k;
+  int EOB;
+  char *BR_buffer;
+  unsigned int BR;
+  int Se = cinfo->Se;
+  int Al = cinfo->Al;
+  JBLOCKROW block;
+  int absvalues[DCTSIZE2];
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval)
+    if (entropy->restarts_to_go == 0)
+      emit_restart(entropy, entropy->next_restart_num);
+
+  /* Encode the MCU data block */
+  block = MCU_data[0];
+
+  /* It is convenient to make a pre-pass to determine the transformed
+   * coefficients' absolute values and the EOB position.
+   */
+  EOB = 0;
+  for (k = cinfo->Ss; k <= Se; k++) {
+    temp = (*block)[jpeg_natural_order[k]];
+    /* We must apply the point transform by Al.  For AC coefficients this
+     * is an integer division with rounding towards 0.  To do this portably
+     * in C, we shift after obtaining the absolute value.
+     */
+    if (temp < 0)
+      temp = -temp;		/* temp is abs value of input */
+    temp >>= Al;		/* apply the point transform */
+    absvalues[k] = temp;	/* save abs value for main pass */
+    if (temp == 1)
+      EOB = k;			/* EOB = index of last newly-nonzero coef */
+  }
+
+  /* Encode the AC coefficients per section G.1.2.3, fig. G.7 */
+  
+  r = 0;			/* r = run length of zeros */
+  BR = 0;			/* BR = count of buffered bits added now */
+  BR_buffer = entropy->bit_buffer + entropy->BE; /* Append bits to buffer */
+
+  for (k = cinfo->Ss; k <= Se; k++) {
+    if ((temp = absvalues[k]) == 0) {
+      r++;
+      continue;
+    }
+
+    /* Emit any required ZRLs, but not if they can be folded into EOB */
+    while (r > 15 && k <= EOB) {
+      /* emit any pending EOBRUN and the BE correction bits */
+      emit_eobrun(entropy);
+      /* Emit ZRL */
+      emit_symbol(entropy, entropy->ac_tbl_no, 0xF0);
+      r -= 16;
+      /* Emit buffered correction bits that must be associated with ZRL */
+      emit_buffered_bits(entropy, BR_buffer, BR);
+      BR_buffer = entropy->bit_buffer; /* BE bits are gone now */
+      BR = 0;
+    }
+
+    /* If the coef was previously nonzero, it only needs a correction bit.
+     * NOTE: a straight translation of the spec's figure G.7 would suggest
+     * that we also need to test r > 15.  But if r > 15, we can only get here
+     * if k > EOB, which implies that this coefficient is not 1.
+     */
+    if (temp > 1) {
+      /* The correction bit is the next bit of the absolute value. */
+      BR_buffer[BR++] = (char) (temp & 1);
+      continue;
+    }
+
+    /* Emit any pending EOBRUN and the BE correction bits */
+    emit_eobrun(entropy);
+
+    /* Count/emit Huffman symbol for run length / number of bits */
+    emit_symbol(entropy, entropy->ac_tbl_no, (r << 4) + 1);
+
+    /* Emit output bit for newly-nonzero coef */
+    temp = ((*block)[jpeg_natural_order[k]] < 0) ? 0 : 1;
+    emit_bits(entropy, (unsigned int) temp, 1);
+
+    /* Emit buffered correction bits that must be associated with this code */
+    emit_buffered_bits(entropy, BR_buffer, BR);
+    BR_buffer = entropy->bit_buffer; /* BE bits are gone now */
+    BR = 0;
+    r = 0;			/* reset zero run length */
+  }
+
+  if (r > 0 || BR > 0) {	/* If there are trailing zeroes, */
+    entropy->EOBRUN++;		/* count an EOB */
+    entropy->BE += BR;		/* concat my correction bits to older ones */
+    /* We force out the EOB if we risk either:
+     * 1. overflow of the EOB counter;
+     * 2. overflow of the correction bit buffer during the next MCU.
+     */
+    if (entropy->EOBRUN == 0x7FFF || entropy->BE > (MAX_CORR_BITS-DCTSIZE2+1))
+      emit_eobrun(entropy);
+  }
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Finish up at the end of a Huffman-compressed progressive scan.
+ */
+
+METHODDEF(void)
+finish_pass_phuff (j_compress_ptr cinfo)
+{   
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Flush out any buffered data */
+  emit_eobrun(entropy);
+  flush_bits(entropy);
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+}
+
+
+/*
+ * Finish up a statistics-gathering pass and create the new Huffman tables.
+ */
+
+METHODDEF(void)
+finish_pass_gather_phuff (j_compress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  boolean is_DC_band;
+  int ci, tbl;
+  jpeg_component_info * compptr;
+  JHUFF_TBL **htblptr;
+  boolean did[NUM_HUFF_TBLS];
+
+  /* Flush out buffered data (all we care about is counting the EOB symbol) */
+  emit_eobrun(entropy);
+
+  is_DC_band = (cinfo->Ss == 0);
+
+  /* It's important not to apply jpeg_gen_optimal_table more than once
+   * per table, because it clobbers the input frequency counts!
+   */
+  MEMZERO(did, SIZEOF(did));
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    if (is_DC_band) {
+      if (cinfo->Ah != 0)	/* DC refinement needs no table */
+	continue;
+      tbl = compptr->dc_tbl_no;
+    } else {
+      tbl = compptr->ac_tbl_no;
+    }
+    if (! did[tbl]) {
+      if (is_DC_band)
+        htblptr = & cinfo->dc_huff_tbl_ptrs[tbl];
+      else
+        htblptr = & cinfo->ac_huff_tbl_ptrs[tbl];
+      if (*htblptr == NULL)
+        *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+      jpeg_gen_optimal_table(cinfo, *htblptr, entropy->count_ptrs[tbl]);
+      did[tbl] = TRUE;
+    }
+  }
+}
+
+
+/*
+ * Module initialization routine for progressive Huffman entropy encoding.
+ */
+
+GLOBAL(void)
+jinit_phuff_encoder (j_compress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy;
+  int i;
+
+  entropy = (phuff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(phuff_entropy_encoder));
+  cinfo->entropy = (struct jpeg_entropy_encoder *) entropy;
+  entropy->pub.start_pass = start_pass_phuff;
+
+  /* Mark tables unallocated */
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    entropy->derived_tbls[i] = NULL;
+    entropy->count_ptrs[i] = NULL;
+  }
+  entropy->bit_buffer = NULL;	/* needed only in AC refinement scan */
+}
+
+#endif /* C_PROGRESSIVE_SUPPORTED */
diff --git a/libjpegtwrp/jcprepct.c b/libjpegtwrp/jcprepct.c
new file mode 100644
index 000000000..fa93333db
--- /dev/null
+++ b/libjpegtwrp/jcprepct.c
@@ -0,0 +1,354 @@
+/*
+ * jcprepct.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the compression preprocessing controller.
+ * This controller manages the color conversion, downsampling,
+ * and edge expansion steps.
+ *
+ * Most of the complexity here is associated with buffering input rows
+ * as required by the downsampler.  See the comments at the head of
+ * jcsample.c for the downsampler's needs.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* At present, jcsample.c can request context rows only for smoothing.
+ * In the future, we might also need context rows for CCIR601 sampling
+ * or other more-complex downsampling procedures.  The code to support
+ * context rows should be compiled only if needed.
+ */
+#ifdef INPUT_SMOOTHING_SUPPORTED
+#define CONTEXT_ROWS_SUPPORTED
+#endif
+
+
+/*
+ * For the simple (no-context-row) case, we just need to buffer one
+ * row group's worth of pixels for the downsampling step.  At the bottom of
+ * the image, we pad to a full row group by replicating the last pixel row.
+ * The downsampler's last output row is then replicated if needed to pad
+ * out to a full iMCU row.
+ *
+ * When providing context rows, we must buffer three row groups' worth of
+ * pixels.  Three row groups are physically allocated, but the row pointer
+ * arrays are made five row groups high, with the extra pointers above and
+ * below "wrapping around" to point to the last and first real row groups.
+ * This allows the downsampler to access the proper context rows.
+ * At the top and bottom of the image, we create dummy context rows by
+ * copying the first or last real pixel row.  This copying could be avoided
+ * by pointer hacking as is done in jdmainct.c, but it doesn't seem worth the
+ * trouble on the compression side.
+ */
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_c_prep_controller pub; /* public fields */
+
+  /* Downsampling input buffer.  This buffer holds color-converted data
+   * until we have enough to do a downsample step.
+   */
+  JSAMPARRAY color_buf[MAX_COMPONENTS];
+
+  JDIMENSION rows_to_go;	/* counts rows remaining in source image */
+  int next_buf_row;		/* index of next row to store in color_buf */
+
+#ifdef CONTEXT_ROWS_SUPPORTED	/* only needed for context case */
+  int this_row_group;		/* starting row index of group to process */
+  int next_buf_stop;		/* downsample when we reach this index */
+#endif
+} my_prep_controller;
+
+typedef my_prep_controller * my_prep_ptr;
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_prep (j_compress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
+
+  if (pass_mode != JBUF_PASS_THRU)
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  /* Initialize total-height counter for detecting bottom of image */
+  prep->rows_to_go = cinfo->image_height;
+  /* Mark the conversion buffer empty */
+  prep->next_buf_row = 0;
+#ifdef CONTEXT_ROWS_SUPPORTED
+  /* Preset additional state variables for context mode.
+   * These aren't used in non-context mode, so we needn't test which mode.
+   */
+  prep->this_row_group = 0;
+  /* Set next_buf_stop to stop after two row groups have been read in. */
+  prep->next_buf_stop = 2 * cinfo->max_v_samp_factor;
+#endif
+}
+
+
+/*
+ * Expand an image vertically from height input_rows to height output_rows,
+ * by duplicating the bottom row.
+ */
+
+LOCAL(void)
+expand_bottom_edge (JSAMPARRAY image_data, JDIMENSION num_cols,
+		    int input_rows, int output_rows)
+{
+  register int row;
+
+  for (row = input_rows; row < output_rows; row++) {
+    jcopy_sample_rows(image_data, input_rows-1, image_data, row,
+		      1, num_cols);
+  }
+}
+
+
+/*
+ * Process some data in the simple no-context case.
+ *
+ * Preprocessor output data is counted in "row groups".  A row group
+ * is defined to be v_samp_factor sample rows of each component.
+ * Downsampling will produce this much data from each max_v_samp_factor
+ * input rows.
+ */
+
+METHODDEF(void)
+pre_process_data (j_compress_ptr cinfo,
+		  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+		  JDIMENSION in_rows_avail,
+		  JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
+		  JDIMENSION out_row_groups_avail)
+{
+  my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
+  int numrows, ci;
+  JDIMENSION inrows;
+  jpeg_component_info * compptr;
+
+  while (*in_row_ctr < in_rows_avail &&
+	 *out_row_group_ctr < out_row_groups_avail) {
+    /* Do color conversion to fill the conversion buffer. */
+    inrows = in_rows_avail - *in_row_ctr;
+    numrows = cinfo->max_v_samp_factor - prep->next_buf_row;
+    numrows = (int) MIN((JDIMENSION) numrows, inrows);
+    (*cinfo->cconvert->color_convert) (cinfo, input_buf + *in_row_ctr,
+				       prep->color_buf,
+				       (JDIMENSION) prep->next_buf_row,
+				       numrows);
+    *in_row_ctr += numrows;
+    prep->next_buf_row += numrows;
+    prep->rows_to_go -= numrows;
+    /* If at bottom of image, pad to fill the conversion buffer. */
+    if (prep->rows_to_go == 0 &&
+	prep->next_buf_row < cinfo->max_v_samp_factor) {
+      for (ci = 0; ci < cinfo->num_components; ci++) {
+	expand_bottom_edge(prep->color_buf[ci], cinfo->image_width,
+			   prep->next_buf_row, cinfo->max_v_samp_factor);
+      }
+      prep->next_buf_row = cinfo->max_v_samp_factor;
+    }
+    /* If we've filled the conversion buffer, empty it. */
+    if (prep->next_buf_row == cinfo->max_v_samp_factor) {
+      (*cinfo->downsample->downsample) (cinfo,
+					prep->color_buf, (JDIMENSION) 0,
+					output_buf, *out_row_group_ctr);
+      prep->next_buf_row = 0;
+      (*out_row_group_ctr)++;
+    }
+    /* If at bottom of image, pad the output to a full iMCU height.
+     * Note we assume the caller is providing a one-iMCU-height output buffer!
+     */
+    if (prep->rows_to_go == 0 &&
+	*out_row_group_ctr < out_row_groups_avail) {
+      for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	   ci++, compptr++) {
+	expand_bottom_edge(output_buf[ci],
+			   compptr->width_in_blocks * DCTSIZE,
+			   (int) (*out_row_group_ctr * compptr->v_samp_factor),
+			   (int) (out_row_groups_avail * compptr->v_samp_factor));
+      }
+      *out_row_group_ctr = out_row_groups_avail;
+      break;			/* can exit outer loop without test */
+    }
+  }
+}
+
+
+#ifdef CONTEXT_ROWS_SUPPORTED
+
+/*
+ * Process some data in the context case.
+ */
+
+METHODDEF(void)
+pre_process_context (j_compress_ptr cinfo,
+		     JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+		     JDIMENSION in_rows_avail,
+		     JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
+		     JDIMENSION out_row_groups_avail)
+{
+  my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
+  int numrows, ci;
+  int buf_height = cinfo->max_v_samp_factor * 3;
+  JDIMENSION inrows;
+
+  while (*out_row_group_ctr < out_row_groups_avail) {
+    if (*in_row_ctr < in_rows_avail) {
+      /* Do color conversion to fill the conversion buffer. */
+      inrows = in_rows_avail - *in_row_ctr;
+      numrows = prep->next_buf_stop - prep->next_buf_row;
+      numrows = (int) MIN((JDIMENSION) numrows, inrows);
+      (*cinfo->cconvert->color_convert) (cinfo, input_buf + *in_row_ctr,
+					 prep->color_buf,
+					 (JDIMENSION) prep->next_buf_row,
+					 numrows);
+      /* Pad at top of image, if first time through */
+      if (prep->rows_to_go == cinfo->image_height) {
+	for (ci = 0; ci < cinfo->num_components; ci++) {
+	  int row;
+	  for (row = 1; row <= cinfo->max_v_samp_factor; row++) {
+	    jcopy_sample_rows(prep->color_buf[ci], 0,
+			      prep->color_buf[ci], -row,
+			      1, cinfo->image_width);
+	  }
+	}
+      }
+      *in_row_ctr += numrows;
+      prep->next_buf_row += numrows;
+      prep->rows_to_go -= numrows;
+    } else {
+      /* Return for more data, unless we are at the bottom of the image. */
+      if (prep->rows_to_go != 0)
+	break;
+      /* When at bottom of image, pad to fill the conversion buffer. */
+      if (prep->next_buf_row < prep->next_buf_stop) {
+	for (ci = 0; ci < cinfo->num_components; ci++) {
+	  expand_bottom_edge(prep->color_buf[ci], cinfo->image_width,
+			     prep->next_buf_row, prep->next_buf_stop);
+	}
+	prep->next_buf_row = prep->next_buf_stop;
+      }
+    }
+    /* If we've gotten enough data, downsample a row group. */
+    if (prep->next_buf_row == prep->next_buf_stop) {
+      (*cinfo->downsample->downsample) (cinfo,
+					prep->color_buf,
+					(JDIMENSION) prep->this_row_group,
+					output_buf, *out_row_group_ctr);
+      (*out_row_group_ctr)++;
+      /* Advance pointers with wraparound as necessary. */
+      prep->this_row_group += cinfo->max_v_samp_factor;
+      if (prep->this_row_group >= buf_height)
+	prep->this_row_group = 0;
+      if (prep->next_buf_row >= buf_height)
+	prep->next_buf_row = 0;
+      prep->next_buf_stop = prep->next_buf_row + cinfo->max_v_samp_factor;
+    }
+  }
+}
+
+
+/*
+ * Create the wrapped-around downsampling input buffer needed for context mode.
+ */
+
+LOCAL(void)
+create_context_buffer (j_compress_ptr cinfo)
+{
+  my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
+  int rgroup_height = cinfo->max_v_samp_factor;
+  int ci, i;
+  jpeg_component_info * compptr;
+  JSAMPARRAY true_buffer, fake_buffer;
+
+  /* Grab enough space for fake row pointers for all the components;
+   * we need five row groups' worth of pointers for each component.
+   */
+  fake_buffer = (JSAMPARRAY)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(cinfo->num_components * 5 * rgroup_height) *
+				SIZEOF(JSAMPROW));
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Allocate the actual buffer space (3 row groups) for this component.
+     * We make the buffer wide enough to allow the downsampler to edge-expand
+     * horizontally within the buffer, if it so chooses.
+     */
+    true_buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) (((long) compptr->width_in_blocks * DCTSIZE *
+		      cinfo->max_h_samp_factor) / compptr->h_samp_factor),
+       (JDIMENSION) (3 * rgroup_height));
+    /* Copy true buffer row pointers into the middle of the fake row array */
+    MEMCOPY(fake_buffer + rgroup_height, true_buffer,
+	    3 * rgroup_height * SIZEOF(JSAMPROW));
+    /* Fill in the above and below wraparound pointers */
+    for (i = 0; i < rgroup_height; i++) {
+      fake_buffer[i] = true_buffer[2 * rgroup_height + i];
+      fake_buffer[4 * rgroup_height + i] = true_buffer[i];
+    }
+    prep->color_buf[ci] = fake_buffer + rgroup_height;
+    fake_buffer += 5 * rgroup_height; /* point to space for next component */
+  }
+}
+
+#endif /* CONTEXT_ROWS_SUPPORTED */
+
+
+/*
+ * Initialize preprocessing controller.
+ */
+
+GLOBAL(void)
+jinit_c_prep_controller (j_compress_ptr cinfo, boolean need_full_buffer)
+{
+  my_prep_ptr prep;
+  int ci;
+  jpeg_component_info * compptr;
+
+  if (need_full_buffer)		/* safety check */
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  prep = (my_prep_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_prep_controller));
+  cinfo->prep = (struct jpeg_c_prep_controller *) prep;
+  prep->pub.start_pass = start_pass_prep;
+
+  /* Allocate the color conversion buffer.
+   * We make the buffer wide enough to allow the downsampler to edge-expand
+   * horizontally within the buffer, if it so chooses.
+   */
+  if (cinfo->downsample->need_context_rows) {
+    /* Set up to provide context rows */
+#ifdef CONTEXT_ROWS_SUPPORTED
+    prep->pub.pre_process_data = pre_process_context;
+    create_context_buffer(cinfo);
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    /* No context, just make it tall enough for one row group */
+    prep->pub.pre_process_data = pre_process_data;
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      prep->color_buf[ci] = (*cinfo->mem->alloc_sarray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE,
+	 (JDIMENSION) (((long) compptr->width_in_blocks * DCTSIZE *
+			cinfo->max_h_samp_factor) / compptr->h_samp_factor),
+	 (JDIMENSION) cinfo->max_v_samp_factor);
+    }
+  }
+}
diff --git a/libjpegtwrp/jcsample.c b/libjpegtwrp/jcsample.c
new file mode 100644
index 000000000..212ec8757
--- /dev/null
+++ b/libjpegtwrp/jcsample.c
@@ -0,0 +1,519 @@
+/*
+ * jcsample.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains downsampling routines.
+ *
+ * Downsampling input data is counted in "row groups".  A row group
+ * is defined to be max_v_samp_factor pixel rows of each component,
+ * from which the downsampler produces v_samp_factor sample rows.
+ * A single row group is processed in each call to the downsampler module.
+ *
+ * The downsampler is responsible for edge-expansion of its output data
+ * to fill an integral number of DCT blocks horizontally.  The source buffer
+ * may be modified if it is helpful for this purpose (the source buffer is
+ * allocated wide enough to correspond to the desired output width).
+ * The caller (the prep controller) is responsible for vertical padding.
+ *
+ * The downsampler may request "context rows" by setting need_context_rows
+ * during startup.  In this case, the input arrays will contain at least
+ * one row group's worth of pixels above and below the passed-in data;
+ * the caller will create dummy rows at image top and bottom by replicating
+ * the first or last real pixel row.
+ *
+ * An excellent reference for image resampling is
+ *   Digital Image Warping, George Wolberg, 1990.
+ *   Pub. by IEEE Computer Society Press, Los Alamitos, CA. ISBN 0-8186-8944-7.
+ *
+ * The downsampling algorithm used here is a simple average of the source
+ * pixels covered by the output pixel.  The hi-falutin sampling literature
+ * refers to this as a "box filter".  In general the characteristics of a box
+ * filter are not very good, but for the specific cases we normally use (1:1
+ * and 2:1 ratios) the box is equivalent to a "triangle filter" which is not
+ * nearly so bad.  If you intend to use other sampling ratios, you'd be well
+ * advised to improve this code.
+ *
+ * A simple input-smoothing capability is provided.  This is mainly intended
+ * for cleaning up color-dithered GIF input files (if you find it inadequate,
+ * we suggest using an external filtering program such as pnmconvol).  When
+ * enabled, each input pixel P is replaced by a weighted sum of itself and its
+ * eight neighbors.  P's weight is 1-8*SF and each neighbor's weight is SF,
+ * where SF = (smoothing_factor / 1024).
+ * Currently, smoothing is only supported for 2h2v sampling factors.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Pointer to routine to downsample a single component */
+typedef JMETHOD(void, downsample1_ptr,
+		(j_compress_ptr cinfo, jpeg_component_info * compptr,
+		 JSAMPARRAY input_data, JSAMPARRAY output_data));
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_downsampler pub;	/* public fields */
+
+  /* Downsampling method pointers, one per component */
+  downsample1_ptr methods[MAX_COMPONENTS];
+} my_downsampler;
+
+typedef my_downsampler * my_downsample_ptr;
+
+
+/*
+ * Initialize for a downsampling pass.
+ */
+
+METHODDEF(void)
+start_pass_downsample (j_compress_ptr cinfo)
+{
+  /* no work for now */
+}
+
+
+/*
+ * Expand a component horizontally from width input_cols to width output_cols,
+ * by duplicating the rightmost samples.
+ */
+
+LOCAL(void)
+expand_right_edge (JSAMPARRAY image_data, int num_rows,
+		   JDIMENSION input_cols, JDIMENSION output_cols)
+{
+  register JSAMPROW ptr;
+  register JSAMPLE pixval;
+  register int count;
+  int row;
+  int numcols = (int) (output_cols - input_cols);
+
+  if (numcols > 0) {
+    for (row = 0; row < num_rows; row++) {
+      ptr = image_data[row] + input_cols;
+      pixval = ptr[-1];		/* don't need GETJSAMPLE() here */
+      for (count = numcols; count > 0; count--)
+	*ptr++ = pixval;
+    }
+  }
+}
+
+
+/*
+ * Do downsampling for a whole row group (all components).
+ *
+ * In this version we simply downsample each component independently.
+ */
+
+METHODDEF(void)
+sep_downsample (j_compress_ptr cinfo,
+		JSAMPIMAGE input_buf, JDIMENSION in_row_index,
+		JSAMPIMAGE output_buf, JDIMENSION out_row_group_index)
+{
+  my_downsample_ptr downsample = (my_downsample_ptr) cinfo->downsample;
+  int ci;
+  jpeg_component_info * compptr;
+  JSAMPARRAY in_ptr, out_ptr;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    in_ptr = input_buf[ci] + in_row_index;
+    out_ptr = output_buf[ci] + (out_row_group_index * compptr->v_samp_factor);
+    (*downsample->methods[ci]) (cinfo, compptr, in_ptr, out_ptr);
+  }
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * One row group is processed per call.
+ * This version handles arbitrary integral sampling ratios, without smoothing.
+ * Note that this version is not actually used for customary sampling ratios.
+ */
+
+METHODDEF(void)
+int_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+		JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow, h_expand, v_expand, numpix, numpix2, h, v;
+  JDIMENSION outcol, outcol_h;	/* outcol_h == outcol*h_expand */
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  JSAMPROW inptr, outptr;
+  INT32 outvalue;
+
+  h_expand = cinfo->max_h_samp_factor / compptr->h_samp_factor;
+  v_expand = cinfo->max_v_samp_factor / compptr->v_samp_factor;
+  numpix = h_expand * v_expand;
+  numpix2 = numpix/2;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data, cinfo->max_v_samp_factor,
+		    cinfo->image_width, output_cols * h_expand);
+
+  inrow = 0;
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_data[outrow];
+    for (outcol = 0, outcol_h = 0; outcol < output_cols;
+	 outcol++, outcol_h += h_expand) {
+      outvalue = 0;
+      for (v = 0; v < v_expand; v++) {
+	inptr = input_data[inrow+v] + outcol_h;
+	for (h = 0; h < h_expand; h++) {
+	  outvalue += (INT32) GETJSAMPLE(*inptr++);
+	}
+      }
+      *outptr++ = (JSAMPLE) ((outvalue + numpix2) / numpix);
+    }
+    inrow += v_expand;
+  }
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the special case of a full-size component,
+ * without smoothing.
+ */
+
+METHODDEF(void)
+fullsize_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+		     JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  /* Copy the data */
+  jcopy_sample_rows(input_data, 0, output_data, 0,
+		    cinfo->max_v_samp_factor, cinfo->image_width);
+  /* Edge-expand */
+  expand_right_edge(output_data, cinfo->max_v_samp_factor,
+		    cinfo->image_width, compptr->width_in_blocks * DCTSIZE);
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the common case of 2:1 horizontal and 1:1 vertical,
+ * without smoothing.
+ *
+ * A note about the "bias" calculations: when rounding fractional values to
+ * integer, we do not want to always round 0.5 up to the next integer.
+ * If we did that, we'd introduce a noticeable bias towards larger values.
+ * Instead, this code is arranged so that 0.5 will be rounded up or down at
+ * alternate pixel locations (a simple ordered dither pattern).
+ */
+
+METHODDEF(void)
+h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+		 JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int outrow;
+  JDIMENSION outcol;
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  register JSAMPROW inptr, outptr;
+  register int bias;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data, cinfo->max_v_samp_factor,
+		    cinfo->image_width, output_cols * 2);
+
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_data[outrow];
+    inptr = input_data[outrow];
+    bias = 0;			/* bias = 0,1,0,1,... for successive samples */
+    for (outcol = 0; outcol < output_cols; outcol++) {
+      *outptr++ = (JSAMPLE) ((GETJSAMPLE(*inptr) + GETJSAMPLE(inptr[1])
+			      + bias) >> 1);
+      bias ^= 1;		/* 0=>1, 1=>0 */
+      inptr += 2;
+    }
+  }
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+ * without smoothing.
+ */
+
+METHODDEF(void)
+h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+		 JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow;
+  JDIMENSION outcol;
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  register JSAMPROW inptr0, inptr1, outptr;
+  register int bias;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data, cinfo->max_v_samp_factor,
+		    cinfo->image_width, output_cols * 2);
+
+  inrow = 0;
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_data[outrow];
+    inptr0 = input_data[inrow];
+    inptr1 = input_data[inrow+1];
+    bias = 1;			/* bias = 1,2,1,2,... for successive samples */
+    for (outcol = 0; outcol < output_cols; outcol++) {
+      *outptr++ = (JSAMPLE) ((GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
+			      GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1])
+			      + bias) >> 2);
+      bias ^= 3;		/* 1=>2, 2=>1 */
+      inptr0 += 2; inptr1 += 2;
+    }
+    inrow += 2;
+  }
+}
+
+
+#ifdef INPUT_SMOOTHING_SUPPORTED
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+ * with smoothing.  One row of context is required.
+ */
+
+METHODDEF(void)
+h2v2_smooth_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+			JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow;
+  JDIMENSION colctr;
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  register JSAMPROW inptr0, inptr1, above_ptr, below_ptr, outptr;
+  INT32 membersum, neighsum, memberscale, neighscale;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data - 1, cinfo->max_v_samp_factor + 2,
+		    cinfo->image_width, output_cols * 2);
+
+  /* We don't bother to form the individual "smoothed" input pixel values;
+   * we can directly compute the output which is the average of the four
+   * smoothed values.  Each of the four member pixels contributes a fraction
+   * (1-8*SF) to its own smoothed image and a fraction SF to each of the three
+   * other smoothed pixels, therefore a total fraction (1-5*SF)/4 to the final
+   * output.  The four corner-adjacent neighbor pixels contribute a fraction
+   * SF to just one smoothed pixel, or SF/4 to the final output; while the
+   * eight edge-adjacent neighbors contribute SF to each of two smoothed
+   * pixels, or SF/2 overall.  In order to use integer arithmetic, these
+   * factors are scaled by 2^16 = 65536.
+   * Also recall that SF = smoothing_factor / 1024.
+   */
+
+  memberscale = 16384 - cinfo->smoothing_factor * 80; /* scaled (1-5*SF)/4 */
+  neighscale = cinfo->smoothing_factor * 16; /* scaled SF/4 */
+
+  inrow = 0;
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_data[outrow];
+    inptr0 = input_data[inrow];
+    inptr1 = input_data[inrow+1];
+    above_ptr = input_data[inrow-1];
+    below_ptr = input_data[inrow+2];
+
+    /* Special case for first column: pretend column -1 is same as column 0 */
+    membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
+		GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+    neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
+	       GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+	       GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[2]) +
+	       GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[2]);
+    neighsum += neighsum;
+    neighsum += GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[2]) +
+		GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[2]);
+    membersum = membersum * memberscale + neighsum * neighscale;
+    *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
+    inptr0 += 2; inptr1 += 2; above_ptr += 2; below_ptr += 2;
+
+    for (colctr = output_cols - 2; colctr > 0; colctr--) {
+      /* sum of pixels directly mapped to this output element */
+      membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
+		  GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+      /* sum of edge-neighbor pixels */
+      neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
+		 GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+		 GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[2]) +
+		 GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[2]);
+      /* The edge-neighbors count twice as much as corner-neighbors */
+      neighsum += neighsum;
+      /* Add in the corner-neighbors */
+      neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[2]) +
+		  GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[2]);
+      /* form final output scaled up by 2^16 */
+      membersum = membersum * memberscale + neighsum * neighscale;
+      /* round, descale and output it */
+      *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
+      inptr0 += 2; inptr1 += 2; above_ptr += 2; below_ptr += 2;
+    }
+
+    /* Special case for last column */
+    membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
+		GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+    neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
+	       GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+	       GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[1]) +
+	       GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[1]);
+    neighsum += neighsum;
+    neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[1]) +
+		GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[1]);
+    membersum = membersum * memberscale + neighsum * neighscale;
+    *outptr = (JSAMPLE) ((membersum + 32768) >> 16);
+
+    inrow += 2;
+  }
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the special case of a full-size component,
+ * with smoothing.  One row of context is required.
+ */
+
+METHODDEF(void)
+fullsize_smooth_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+			    JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int outrow;
+  JDIMENSION colctr;
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  register JSAMPROW inptr, above_ptr, below_ptr, outptr;
+  INT32 membersum, neighsum, memberscale, neighscale;
+  int colsum, lastcolsum, nextcolsum;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data - 1, cinfo->max_v_samp_factor + 2,
+		    cinfo->image_width, output_cols);
+
+  /* Each of the eight neighbor pixels contributes a fraction SF to the
+   * smoothed pixel, while the main pixel contributes (1-8*SF).  In order
+   * to use integer arithmetic, these factors are multiplied by 2^16 = 65536.
+   * Also recall that SF = smoothing_factor / 1024.
+   */
+
+  memberscale = 65536L - cinfo->smoothing_factor * 512L; /* scaled 1-8*SF */
+  neighscale = cinfo->smoothing_factor * 64; /* scaled SF */
+
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_data[outrow];
+    inptr = input_data[outrow];
+    above_ptr = input_data[outrow-1];
+    below_ptr = input_data[outrow+1];
+
+    /* Special case for first column */
+    colsum = GETJSAMPLE(*above_ptr++) + GETJSAMPLE(*below_ptr++) +
+	     GETJSAMPLE(*inptr);
+    membersum = GETJSAMPLE(*inptr++);
+    nextcolsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(*below_ptr) +
+		 GETJSAMPLE(*inptr);
+    neighsum = colsum + (colsum - membersum) + nextcolsum;
+    membersum = membersum * memberscale + neighsum * neighscale;
+    *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
+    lastcolsum = colsum; colsum = nextcolsum;
+
+    for (colctr = output_cols - 2; colctr > 0; colctr--) {
+      membersum = GETJSAMPLE(*inptr++);
+      above_ptr++; below_ptr++;
+      nextcolsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(*below_ptr) +
+		   GETJSAMPLE(*inptr);
+      neighsum = lastcolsum + (colsum - membersum) + nextcolsum;
+      membersum = membersum * memberscale + neighsum * neighscale;
+      *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
+      lastcolsum = colsum; colsum = nextcolsum;
+    }
+
+    /* Special case for last column */
+    membersum = GETJSAMPLE(*inptr);
+    neighsum = lastcolsum + (colsum - membersum) + colsum;
+    membersum = membersum * memberscale + neighsum * neighscale;
+    *outptr = (JSAMPLE) ((membersum + 32768) >> 16);
+
+  }
+}
+
+#endif /* INPUT_SMOOTHING_SUPPORTED */
+
+
+/*
+ * Module initialization routine for downsampling.
+ * Note that we must select a routine for each component.
+ */
+
+GLOBAL(void)
+jinit_downsampler (j_compress_ptr cinfo)
+{
+  my_downsample_ptr downsample;
+  int ci;
+  jpeg_component_info * compptr;
+  boolean smoothok = TRUE;
+
+  downsample = (my_downsample_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_downsampler));
+  cinfo->downsample = (struct jpeg_downsampler *) downsample;
+  downsample->pub.start_pass = start_pass_downsample;
+  downsample->pub.downsample = sep_downsample;
+  downsample->pub.need_context_rows = FALSE;
+
+  if (cinfo->CCIR601_sampling)
+    ERREXIT(cinfo, JERR_CCIR601_NOTIMPL);
+
+  /* Verify we can handle the sampling factors, and set up method pointers */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    if (compptr->h_samp_factor == cinfo->max_h_samp_factor &&
+	compptr->v_samp_factor == cinfo->max_v_samp_factor) {
+#ifdef INPUT_SMOOTHING_SUPPORTED
+      if (cinfo->smoothing_factor) {
+	downsample->methods[ci] = fullsize_smooth_downsample;
+	downsample->pub.need_context_rows = TRUE;
+      } else
+#endif
+	downsample->methods[ci] = fullsize_downsample;
+    } else if (compptr->h_samp_factor * 2 == cinfo->max_h_samp_factor &&
+	       compptr->v_samp_factor == cinfo->max_v_samp_factor) {
+      smoothok = FALSE;
+      downsample->methods[ci] = h2v1_downsample;
+    } else if (compptr->h_samp_factor * 2 == cinfo->max_h_samp_factor &&
+	       compptr->v_samp_factor * 2 == cinfo->max_v_samp_factor) {
+#ifdef INPUT_SMOOTHING_SUPPORTED
+      if (cinfo->smoothing_factor) {
+	downsample->methods[ci] = h2v2_smooth_downsample;
+	downsample->pub.need_context_rows = TRUE;
+      } else
+#endif
+	downsample->methods[ci] = h2v2_downsample;
+    } else if ((cinfo->max_h_samp_factor % compptr->h_samp_factor) == 0 &&
+	       (cinfo->max_v_samp_factor % compptr->v_samp_factor) == 0) {
+      smoothok = FALSE;
+      downsample->methods[ci] = int_downsample;
+    } else
+      ERREXIT(cinfo, JERR_FRACT_SAMPLE_NOTIMPL);
+  }
+
+#ifdef INPUT_SMOOTHING_SUPPORTED
+  if (cinfo->smoothing_factor && !smoothok)
+    TRACEMS(cinfo, 0, JTRC_SMOOTH_NOTIMPL);
+#endif
+}
diff --git a/libjpegtwrp/jctrans.c b/libjpegtwrp/jctrans.c
new file mode 100644
index 000000000..f574ed14f
--- /dev/null
+++ b/libjpegtwrp/jctrans.c
@@ -0,0 +1,395 @@
+/*
+ * jctrans.c
+ *
+ * Copyright (C) 1995-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains library routines for transcoding compression,
+ * that is, writing raw DCT coefficient arrays to an output JPEG file.
+ * The routines in jcapimin.c will also be needed by a transcoder.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Forward declarations */
+LOCAL(void) transencode_master_selection
+	JPP((j_compress_ptr cinfo, jvirt_barray_ptr * coef_arrays));
+LOCAL(void) transencode_coef_controller
+	JPP((j_compress_ptr cinfo, jvirt_barray_ptr * coef_arrays));
+
+
+/*
+ * Compression initialization for writing raw-coefficient data.
+ * Before calling this, all parameters and a data destination must be set up.
+ * Call jpeg_finish_compress() to actually write the data.
+ *
+ * The number of passed virtual arrays must match cinfo->num_components.
+ * Note that the virtual arrays need not be filled or even realized at
+ * the time write_coefficients is called; indeed, if the virtual arrays
+ * were requested from this compression object's memory manager, they
+ * typically will be realized during this routine and filled afterwards.
+ */
+
+GLOBAL(void)
+jpeg_write_coefficients (j_compress_ptr cinfo, jvirt_barray_ptr * coef_arrays)
+{
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Mark all tables to be written */
+  jpeg_suppress_tables(cinfo, FALSE);
+  /* (Re)initialize error mgr and destination modules */
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr) cinfo);
+  (*cinfo->dest->init_destination) (cinfo);
+  /* Perform master selection of active modules */
+  transencode_master_selection(cinfo, coef_arrays);
+  /* Wait for jpeg_finish_compress() call */
+  cinfo->next_scanline = 0;	/* so jpeg_write_marker works */
+  cinfo->global_state = CSTATE_WRCOEFS;
+}
+
+
+/*
+ * Initialize the compression object with default parameters,
+ * then copy from the source object all parameters needed for lossless
+ * transcoding.  Parameters that can be varied without loss (such as
+ * scan script and Huffman optimization) are left in their default states.
+ */
+
+GLOBAL(void)
+jpeg_copy_critical_parameters (j_decompress_ptr srcinfo,
+			       j_compress_ptr dstinfo)
+{
+  JQUANT_TBL ** qtblptr;
+  jpeg_component_info *incomp, *outcomp;
+  JQUANT_TBL *c_quant, *slot_quant;
+  int tblno, ci, coefi;
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (dstinfo->global_state != CSTATE_START)
+    ERREXIT1(dstinfo, JERR_BAD_STATE, dstinfo->global_state);
+  /* Copy fundamental image dimensions */
+  dstinfo->image_width = srcinfo->image_width;
+  dstinfo->image_height = srcinfo->image_height;
+  dstinfo->input_components = srcinfo->num_components;
+  dstinfo->in_color_space = srcinfo->jpeg_color_space;
+  /* Initialize all parameters to default values */
+  jpeg_set_defaults(dstinfo);
+  /* jpeg_set_defaults may choose wrong colorspace, eg YCbCr if input is RGB.
+   * Fix it to get the right header markers for the image colorspace.
+   */
+  jpeg_set_colorspace(dstinfo, srcinfo->jpeg_color_space);
+  dstinfo->data_precision = srcinfo->data_precision;
+  dstinfo->CCIR601_sampling = srcinfo->CCIR601_sampling;
+  /* Copy the source's quantization tables. */
+  for (tblno = 0; tblno < NUM_QUANT_TBLS; tblno++) {
+    if (srcinfo->quant_tbl_ptrs[tblno] != NULL) {
+      qtblptr = & dstinfo->quant_tbl_ptrs[tblno];
+      if (*qtblptr == NULL)
+	*qtblptr = jpeg_alloc_quant_table((j_common_ptr) dstinfo);
+      MEMCOPY((*qtblptr)->quantval,
+	      srcinfo->quant_tbl_ptrs[tblno]->quantval,
+	      SIZEOF((*qtblptr)->quantval));
+      (*qtblptr)->sent_table = FALSE;
+    }
+  }
+  /* Copy the source's per-component info.
+   * Note we assume jpeg_set_defaults has allocated the dest comp_info array.
+   */
+  dstinfo->num_components = srcinfo->num_components;
+  if (dstinfo->num_components < 1 || dstinfo->num_components > MAX_COMPONENTS)
+    ERREXIT2(dstinfo, JERR_COMPONENT_COUNT, dstinfo->num_components,
+	     MAX_COMPONENTS);
+  for (ci = 0, incomp = srcinfo->comp_info, outcomp = dstinfo->comp_info;
+       ci < dstinfo->num_components; ci++, incomp++, outcomp++) {
+    outcomp->component_id = incomp->component_id;
+    outcomp->h_samp_factor = incomp->h_samp_factor;
+    outcomp->v_samp_factor = incomp->v_samp_factor;
+    outcomp->quant_tbl_no = incomp->quant_tbl_no;
+    /* Make sure saved quantization table for component matches the qtable
+     * slot.  If not, the input file re-used this qtable slot.
+     * IJG encoder currently cannot duplicate this.
+     */
+    tblno = outcomp->quant_tbl_no;
+    if (tblno < 0 || tblno >= NUM_QUANT_TBLS ||
+	srcinfo->quant_tbl_ptrs[tblno] == NULL)
+      ERREXIT1(dstinfo, JERR_NO_QUANT_TABLE, tblno);
+    slot_quant = srcinfo->quant_tbl_ptrs[tblno];
+    c_quant = incomp->quant_table;
+    if (c_quant != NULL) {
+      for (coefi = 0; coefi < DCTSIZE2; coefi++) {
+	if (c_quant->quantval[coefi] != slot_quant->quantval[coefi])
+	  ERREXIT1(dstinfo, JERR_MISMATCHED_QUANT_TABLE, tblno);
+      }
+    }
+    /* Note: we do not copy the source's Huffman table assignments;
+     * instead we rely on jpeg_set_colorspace to have made a suitable choice.
+     */
+  }
+  /* Also copy JFIF version and resolution information, if available.
+   * Strictly speaking this isn't "critical" info, but it's nearly
+   * always appropriate to copy it if available.  In particular,
+   * if the application chooses to copy JFIF 1.02 extension markers from
+   * the source file, we need to copy the version to make sure we don't
+   * emit a file that has 1.02 extensions but a claimed version of 1.01.
+   * We will *not*, however, copy version info from mislabeled "2.01" files.
+   */
+  if (srcinfo->saw_JFIF_marker) {
+    if (srcinfo->JFIF_major_version == 1) {
+      dstinfo->JFIF_major_version = srcinfo->JFIF_major_version;
+      dstinfo->JFIF_minor_version = srcinfo->JFIF_minor_version;
+    }
+    dstinfo->density_unit = srcinfo->density_unit;
+    dstinfo->X_density = srcinfo->X_density;
+    dstinfo->Y_density = srcinfo->Y_density;
+  }
+}
+
+
+/*
+ * Master selection of compression modules for transcoding.
+ * This substitutes for jcinit.c's initialization of the full compressor.
+ */
+
+LOCAL(void)
+transencode_master_selection (j_compress_ptr cinfo,
+			      jvirt_barray_ptr * coef_arrays)
+{
+  /* Although we don't actually use input_components for transcoding,
+   * jcmaster.c's initial_setup will complain if input_components is 0.
+   */
+  cinfo->input_components = 1;
+  /* Initialize master control (includes parameter checking/processing) */
+  jinit_c_master_control(cinfo, TRUE /* transcode only */);
+
+  /* Entropy encoding: either Huffman or arithmetic coding. */
+  if (cinfo->arith_code) {
+    ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+  } else {
+    if (cinfo->progressive_mode) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+      jinit_phuff_encoder(cinfo);
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else
+      jinit_huff_encoder(cinfo);
+  }
+
+  /* We need a special coefficient buffer controller. */
+  transencode_coef_controller(cinfo, coef_arrays);
+
+  jinit_marker_writer(cinfo);
+
+  /* We can now tell the memory manager to allocate virtual arrays. */
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr) cinfo);
+
+  /* Write the datastream header (SOI, JFIF) immediately.
+   * Frame and scan headers are postponed till later.
+   * This lets application insert special markers after the SOI.
+   */
+  (*cinfo->marker->write_file_header) (cinfo);
+}
+
+
+/*
+ * The rest of this file is a special implementation of the coefficient
+ * buffer controller.  This is similar to jccoefct.c, but it handles only
+ * output from presupplied virtual arrays.  Furthermore, we generate any
+ * dummy padding blocks on-the-fly rather than expecting them to be present
+ * in the arrays.
+ */
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_c_coef_controller pub; /* public fields */
+
+  JDIMENSION iMCU_row_num;	/* iMCU row # within image */
+  JDIMENSION mcu_ctr;		/* counts MCUs processed in current row */
+  int MCU_vert_offset;		/* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;	/* number of such rows needed */
+
+  /* Virtual block array for each component. */
+  jvirt_barray_ptr * whole_image;
+
+  /* Workspace for constructing dummy blocks at right/bottom edges. */
+  JBLOCKROW dummy_buffer[C_MAX_BLOCKS_IN_MCU];
+} my_coef_controller;
+
+typedef my_coef_controller * my_coef_ptr;
+
+
+LOCAL(void)
+start_iMCU_row (j_compress_ptr cinfo)
+/* Reset within-iMCU-row counters for a new row */
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  /* In an interleaved scan, an MCU row is the same as an iMCU row.
+   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
+   * But at the bottom of the image, process only what's left.
+   */
+  if (cinfo->comps_in_scan > 1) {
+    coef->MCU_rows_per_iMCU_row = 1;
+  } else {
+    if (coef->iMCU_row_num < (cinfo->total_iMCU_rows-1))
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->v_samp_factor;
+    else
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->last_row_height;
+  }
+
+  coef->mcu_ctr = 0;
+  coef->MCU_vert_offset = 0;
+}
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_coef (j_compress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  if (pass_mode != JBUF_CRANK_DEST)
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  coef->iMCU_row_num = 0;
+  start_iMCU_row(cinfo);
+}
+
+
+/*
+ * Process some data.
+ * We process the equivalent of one fully interleaved MCU row ("iMCU" row)
+ * per call, ie, v_samp_factor block rows for each component in the scan.
+ * The data is obtained from the virtual arrays and fed to the entropy coder.
+ * Returns TRUE if the iMCU row is completed, FALSE if suspended.
+ *
+ * NB: input_buf is ignored; it is likely to be a NULL pointer.
+ */
+
+METHODDEF(boolean)
+compress_output (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  int blkn, ci, xindex, yindex, yoffset, blockcnt;
+  JDIMENSION start_col;
+  JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN];
+  JBLOCKROW MCU_buffer[C_MAX_BLOCKS_IN_MCU];
+  JBLOCKROW buffer_ptr;
+  jpeg_component_info *compptr;
+
+  /* Align the virtual buffers for the components used in this scan. */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    buffer[ci] = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr) cinfo, coef->whole_image[compptr->component_index],
+       coef->iMCU_row_num * compptr->v_samp_factor,
+       (JDIMENSION) compptr->v_samp_factor, FALSE);
+  }
+
+  /* Loop to process one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->mcu_ctr; MCU_col_num < cinfo->MCUs_per_row;
+	 MCU_col_num++) {
+      /* Construct list of pointers to DCT blocks belonging to this MCU */
+      blkn = 0;			/* index of current DCT block within MCU */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+	compptr = cinfo->cur_comp_info[ci];
+	start_col = MCU_col_num * compptr->MCU_width;
+	blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
+						: compptr->last_col_width;
+	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+	  if (coef->iMCU_row_num < last_iMCU_row ||
+	      yindex+yoffset < compptr->last_row_height) {
+	    /* Fill in pointers to real blocks in this row */
+	    buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+	    for (xindex = 0; xindex < blockcnt; xindex++)
+	      MCU_buffer[blkn++] = buffer_ptr++;
+	  } else {
+	    /* At bottom of image, need a whole row of dummy blocks */
+	    xindex = 0;
+	  }
+	  /* Fill in any dummy blocks needed in this row.
+	   * Dummy blocks are filled in the same way as in jccoefct.c:
+	   * all zeroes in the AC entries, DC entries equal to previous
+	   * block's DC value.  The init routine has already zeroed the
+	   * AC entries, so we need only set the DC entries correctly.
+	   */
+	  for (; xindex < compptr->MCU_width; xindex++) {
+	    MCU_buffer[blkn] = coef->dummy_buffer[blkn];
+	    MCU_buffer[blkn][0][0] = MCU_buffer[blkn-1][0][0];
+	    blkn++;
+	  }
+	}
+      }
+      /* Try to write the MCU. */
+      if (! (*cinfo->entropy->encode_mcu) (cinfo, MCU_buffer)) {
+	/* Suspension forced; update state counters and exit */
+	coef->MCU_vert_offset = yoffset;
+	coef->mcu_ctr = MCU_col_num;
+#ifdef NEEDS_ARM_ERRATA_754319_754320
+    asm volatile ( "vmov s0,s0\n" );
+#endif
+	return FALSE;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->mcu_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  coef->iMCU_row_num++;
+  start_iMCU_row(cinfo);
+#ifdef NEEDS_ARM_ERRATA_754319_754320
+  asm volatile ( "vmov s0,s0\n" );
+#endif
+
+  return TRUE;
+}
+
+
+/*
+ * Initialize coefficient buffer controller.
+ *
+ * Each passed coefficient array must be the right size for that
+ * coefficient: width_in_blocks wide and height_in_blocks high,
+ * with unitheight at least v_samp_factor.
+ */
+
+LOCAL(void)
+transencode_coef_controller (j_compress_ptr cinfo,
+			     jvirt_barray_ptr * coef_arrays)
+{
+  my_coef_ptr coef;
+  JBLOCKROW buffer;
+  int i;
+
+  coef = (my_coef_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_coef_controller));
+  cinfo->coef = (struct jpeg_c_coef_controller *) coef;
+  coef->pub.start_pass = start_pass_coef;
+  coef->pub.compress_data = compress_output;
+
+  /* Save pointer to virtual arrays */
+  coef->whole_image = coef_arrays;
+
+  /* Allocate and pre-zero space for dummy DCT blocks. */
+  buffer = (JBLOCKROW)
+    (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				C_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+  jzero_far((void FAR *) buffer, C_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+  for (i = 0; i < C_MAX_BLOCKS_IN_MCU; i++) {
+    coef->dummy_buffer[i] = buffer + i;
+  }
+}
diff --git a/libjpegtwrp/jdapimin.c b/libjpegtwrp/jdapimin.c
new file mode 100644
index 000000000..5c9607ed5
--- /dev/null
+++ b/libjpegtwrp/jdapimin.c
@@ -0,0 +1,401 @@
+/*
+ * jdapimin.c
+ *
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains application interface code for the decompression half
+ * of the JPEG library.  These are the "minimum" API routines that may be
+ * needed in either the normal full-decompression case or the
+ * transcoding-only case.
+ *
+ * Most of the routines intended to be called directly by an application
+ * are in this file or in jdapistd.c.  But also see jcomapi.c for routines
+ * shared by compression and decompression, and jdtrans.c for the transcoding
+ * case.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Initialization of a JPEG decompression object.
+ * The error manager must already be set up (in case memory manager fails).
+ */
+
+GLOBAL(void)
+jpeg_CreateDecompress (j_decompress_ptr cinfo, int version, size_t structsize)
+{
+  int i;
+
+  /* Guard against version mismatches between library and caller. */
+  cinfo->mem = NULL;		/* so jpeg_destroy knows mem mgr not called */
+  if (version != JPEG_LIB_VERSION)
+    ERREXIT2(cinfo, JERR_BAD_LIB_VERSION, JPEG_LIB_VERSION, version);
+  if (structsize != SIZEOF(struct jpeg_decompress_struct))
+    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE, 
+	     (int) SIZEOF(struct jpeg_decompress_struct), (int) structsize);
+
+  /* For debugging purposes, we zero the whole master structure.
+   * But the application has already set the err pointer, and may have set
+   * client_data, so we have to save and restore those fields.
+   * Note: if application hasn't set client_data, tools like Purify may
+   * complain here.
+   */
+  {
+    struct jpeg_error_mgr * err = cinfo->err;
+    void * client_data = cinfo->client_data; /* ignore Purify complaint here */
+    MEMZERO(cinfo, SIZEOF(struct jpeg_decompress_struct));
+    cinfo->err = err;
+    cinfo->client_data = client_data;
+  }
+  cinfo->is_decompressor = TRUE;
+  cinfo->tile_decode = FALSE;
+
+  /* Initialize a memory manager instance for this object */
+  jinit_memory_mgr((j_common_ptr) cinfo);
+
+  /* Zero out pointers to permanent structures. */
+  cinfo->progress = NULL;
+  cinfo->src = NULL;
+
+  for (i = 0; i < NUM_QUANT_TBLS; i++)
+    cinfo->quant_tbl_ptrs[i] = NULL;
+
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    cinfo->dc_huff_tbl_ptrs[i] = NULL;
+    cinfo->ac_huff_tbl_ptrs[i] = NULL;
+  }
+
+  /* Initialize marker processor so application can override methods
+   * for COM, APPn markers before calling jpeg_read_header.
+   */
+  cinfo->marker_list = NULL;
+  jinit_marker_reader(cinfo);
+
+  /* And initialize the overall input controller. */
+  jinit_input_controller(cinfo);
+
+  /* OK, I'm ready */
+  cinfo->global_state = DSTATE_START;
+}
+
+
+/*
+ * Destruction of a JPEG decompression object
+ */
+
+GLOBAL(void)
+jpeg_destroy_decompress (j_decompress_ptr cinfo)
+{
+  jpeg_destroy((j_common_ptr) cinfo); /* use common routine */
+}
+
+
+/*
+ * Abort processing of a JPEG decompression operation,
+ * but don't destroy the object itself.
+ */
+
+GLOBAL(void)
+jpeg_abort_decompress (j_decompress_ptr cinfo)
+{
+  jpeg_abort((j_common_ptr) cinfo); /* use common routine */
+}
+
+
+/*
+ * Set default decompression parameters.
+ */
+
+LOCAL(void)
+default_decompress_parms (j_decompress_ptr cinfo)
+{
+  /* Guess the input colorspace, and set output colorspace accordingly. */
+  /* (Wish JPEG committee had provided a real way to specify this...) */
+  /* Note application may override our guesses. */
+  switch (cinfo->num_components) {
+  case 1:
+    cinfo->jpeg_color_space = JCS_GRAYSCALE;
+    cinfo->out_color_space = JCS_GRAYSCALE;
+    break;
+    
+  case 3:
+    if (cinfo->saw_JFIF_marker) {
+      cinfo->jpeg_color_space = JCS_YCbCr; /* JFIF implies YCbCr */
+    } else if (cinfo->saw_Adobe_marker) {
+      switch (cinfo->Adobe_transform) {
+      case 0:
+	cinfo->jpeg_color_space = JCS_RGB;
+	break;
+      case 1:
+	cinfo->jpeg_color_space = JCS_YCbCr;
+	break;
+      default:
+	WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
+	cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
+	break;
+      }
+    } else {
+      /* Saw no special markers, try to guess from the component IDs */
+      int cid0 = cinfo->comp_info[0].component_id;
+      int cid1 = cinfo->comp_info[1].component_id;
+      int cid2 = cinfo->comp_info[2].component_id;
+
+      if (cid0 == 1 && cid1 == 2 && cid2 == 3)
+	cinfo->jpeg_color_space = JCS_YCbCr; /* assume JFIF w/out marker */
+      else if (cid0 == 82 && cid1 == 71 && cid2 == 66)
+	cinfo->jpeg_color_space = JCS_RGB; /* ASCII 'R', 'G', 'B' */
+      else {
+	TRACEMS3(cinfo, 1, JTRC_UNKNOWN_IDS, cid0, cid1, cid2);
+	cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
+      }
+    }
+    /* Always guess RGB is proper output colorspace. */
+    cinfo->out_color_space = JCS_RGB;
+    break;
+    
+  case 4:
+    if (cinfo->saw_Adobe_marker) {
+      switch (cinfo->Adobe_transform) {
+      case 0:
+	cinfo->jpeg_color_space = JCS_CMYK;
+	break;
+      case 2:
+	cinfo->jpeg_color_space = JCS_YCCK;
+	break;
+      default:
+	WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
+	cinfo->jpeg_color_space = JCS_YCCK; /* assume it's YCCK */
+	break;
+      }
+    } else {
+      /* No special markers, assume straight CMYK. */
+      cinfo->jpeg_color_space = JCS_CMYK;
+    }
+    cinfo->out_color_space = JCS_CMYK;
+    break;
+    
+  default:
+    cinfo->jpeg_color_space = JCS_UNKNOWN;
+    cinfo->out_color_space = JCS_UNKNOWN;
+    break;
+  }
+
+  /* Set defaults for other decompression parameters. */
+  cinfo->scale_num = 1;		/* 1:1 scaling */
+  cinfo->scale_denom = 1;
+  cinfo->output_gamma = 1.0;
+  cinfo->buffered_image = FALSE;
+  cinfo->raw_data_out = FALSE;
+  cinfo->dct_method = JDCT_DEFAULT;
+  cinfo->do_fancy_upsampling = TRUE;
+  cinfo->do_block_smoothing = TRUE;
+  cinfo->quantize_colors = FALSE;
+  /* We set these in case application only sets quantize_colors. */
+  cinfo->dither_mode = JDITHER_FS;
+#ifdef QUANT_2PASS_SUPPORTED
+  cinfo->two_pass_quantize = TRUE;
+#else
+  cinfo->two_pass_quantize = FALSE;
+#endif
+  cinfo->desired_number_of_colors = 256;
+  cinfo->colormap = NULL;
+  /* Initialize for no mode change in buffered-image mode. */
+  cinfo->enable_1pass_quant = FALSE;
+  cinfo->enable_external_quant = FALSE;
+  cinfo->enable_2pass_quant = FALSE;
+}
+
+
+/*
+ * Decompression startup: read start of JPEG datastream to see what's there.
+ * Need only initialize JPEG object and supply a data source before calling.
+ *
+ * This routine will read as far as the first SOS marker (ie, actual start of
+ * compressed data), and will save all tables and parameters in the JPEG
+ * object.  It will also initialize the decompression parameters to default
+ * values, and finally return JPEG_HEADER_OK.  On return, the application may
+ * adjust the decompression parameters and then call jpeg_start_decompress.
+ * (Or, if the application only wanted to determine the image parameters,
+ * the data need not be decompressed.  In that case, call jpeg_abort or
+ * jpeg_destroy to release any temporary space.)
+ * If an abbreviated (tables only) datastream is presented, the routine will
+ * return JPEG_HEADER_TABLES_ONLY upon reaching EOI.  The application may then
+ * re-use the JPEG object to read the abbreviated image datastream(s).
+ * It is unnecessary (but OK) to call jpeg_abort in this case.
+ * The JPEG_SUSPENDED return code only occurs if the data source module
+ * requests suspension of the decompressor.  In this case the application
+ * should load more source data and then re-call jpeg_read_header to resume
+ * processing.
+ * If a non-suspending data source is used and require_image is TRUE, then the
+ * return code need not be inspected since only JPEG_HEADER_OK is possible.
+ *
+ * This routine is now just a front end to jpeg_consume_input, with some
+ * extra error checking.
+ */
+
+GLOBAL(int)
+jpeg_read_header (j_decompress_ptr cinfo, boolean require_image)
+{
+  int retcode;
+
+  if (cinfo->global_state != DSTATE_START &&
+      cinfo->global_state != DSTATE_INHEADER)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  retcode = jpeg_consume_input(cinfo);
+
+  switch (retcode) {
+  case JPEG_REACHED_SOS:
+    retcode = JPEG_HEADER_OK;
+    break;
+  case JPEG_REACHED_EOI:
+    if (require_image)		/* Complain if application wanted an image */
+      ERREXIT(cinfo, JERR_NO_IMAGE);
+    /* Reset to start state; it would be safer to require the application to
+     * call jpeg_abort, but we can't change it now for compatibility reasons.
+     * A side effect is to free any temporary memory (there shouldn't be any).
+     */
+    jpeg_abort((j_common_ptr) cinfo); /* sets state = DSTATE_START */
+    retcode = JPEG_HEADER_TABLES_ONLY;
+    break;
+  case JPEG_SUSPENDED:
+    /* no work */
+    break;
+  }
+
+  return retcode;
+}
+
+
+/*
+ * Consume data in advance of what the decompressor requires.
+ * This can be called at any time once the decompressor object has
+ * been created and a data source has been set up.
+ *
+ * This routine is essentially a state machine that handles a couple
+ * of critical state-transition actions, namely initial setup and
+ * transition from header scanning to ready-for-start_decompress.
+ * All the actual input is done via the input controller's consume_input
+ * method.
+ */
+
+GLOBAL(int)
+jpeg_consume_input (j_decompress_ptr cinfo)
+{
+  int retcode = JPEG_SUSPENDED;
+
+  /* NB: every possible DSTATE value should be listed in this switch */
+  switch (cinfo->global_state) {
+  case DSTATE_START:
+    /* Start-of-datastream actions: reset appropriate modules */
+    (*cinfo->inputctl->reset_input_controller) (cinfo);
+    /* Initialize application's data source module */
+    (*cinfo->src->init_source) (cinfo);
+    cinfo->global_state = DSTATE_INHEADER;
+    /*FALLTHROUGH*/
+  case DSTATE_INHEADER:
+    retcode = (*cinfo->inputctl->consume_input) (cinfo);
+    if (retcode == JPEG_REACHED_SOS) { /* Found SOS, prepare to decompress */
+      /* Set up default parameters based on header data */
+      default_decompress_parms(cinfo);
+      /* Set global state: ready for start_decompress */
+      cinfo->global_state = DSTATE_READY;
+    }
+    break;
+  case DSTATE_READY:
+    /* Can't advance past first SOS until start_decompress is called */
+    retcode = JPEG_REACHED_SOS;
+    break;
+  case DSTATE_PRELOAD:
+  case DSTATE_PRESCAN:
+  case DSTATE_SCANNING:
+  case DSTATE_RAW_OK:
+  case DSTATE_BUFIMAGE:
+  case DSTATE_BUFPOST:
+  case DSTATE_STOPPING:
+    retcode = (*cinfo->inputctl->consume_input) (cinfo);
+    break;
+  default:
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  }
+  return retcode;
+}
+
+
+/*
+ * Have we finished reading the input file?
+ */
+
+GLOBAL(boolean)
+jpeg_input_complete (j_decompress_ptr cinfo)
+{
+  /* Check for valid jpeg object */
+  if (cinfo->global_state < DSTATE_START ||
+      cinfo->global_state > DSTATE_STOPPING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  return cinfo->inputctl->eoi_reached;
+}
+
+
+/*
+ * Is there more than one scan?
+ */
+
+GLOBAL(boolean)
+jpeg_has_multiple_scans (j_decompress_ptr cinfo)
+{
+  /* Only valid after jpeg_read_header completes */
+  if (cinfo->global_state < DSTATE_READY ||
+      cinfo->global_state > DSTATE_STOPPING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  return cinfo->inputctl->has_multiple_scans;
+}
+
+
+/*
+ * Finish JPEG decompression.
+ *
+ * This will normally just verify the file trailer and release temp storage.
+ *
+ * Returns FALSE if suspended.  The return value need be inspected only if
+ * a suspending data source is used.
+ */
+
+GLOBAL(boolean)
+jpeg_finish_decompress (j_decompress_ptr cinfo)
+{
+  if ((cinfo->global_state == DSTATE_SCANNING ||
+       cinfo->global_state == DSTATE_RAW_OK) && ! cinfo->buffered_image) {
+    /* Terminate final pass of non-buffered mode */
+#ifdef ANDROID_TILE_BASED_DECODE
+    cinfo->output_scanline = cinfo->output_height;
+#endif
+    if (cinfo->output_scanline < cinfo->output_height)
+      ERREXIT(cinfo, JERR_TOO_LITTLE_DATA);
+    (*cinfo->master->finish_output_pass) (cinfo);
+    cinfo->global_state = DSTATE_STOPPING;
+  } else if (cinfo->global_state == DSTATE_BUFIMAGE) {
+    /* Finishing after a buffered-image operation */
+    cinfo->global_state = DSTATE_STOPPING;
+  } else if (cinfo->global_state != DSTATE_STOPPING) {
+    /* STOPPING = repeat call after a suspension, anything else is error */
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  }
+  /* Read until EOI */
+#ifndef ANDROID_TILE_BASED_DECODE
+  while (! cinfo->inputctl->eoi_reached) {
+    if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED)
+      return FALSE;		/* Suspend, come back later */
+  }
+#endif
+  /* Do final cleanup */
+  (*cinfo->src->term_source) (cinfo);
+  /* We can use jpeg_abort to release memory and reset global_state */
+  jpeg_abort((j_common_ptr) cinfo);
+  return TRUE;
+}
diff --git a/libjpegtwrp/jdapistd.c b/libjpegtwrp/jdapistd.c
new file mode 100644
index 000000000..e1233df36
--- /dev/null
+++ b/libjpegtwrp/jdapistd.c
@@ -0,0 +1,397 @@
+/*
+ * jdapistd.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains application interface code for the decompression half
+ * of the JPEG library.  These are the "standard" API routines that are
+ * used in the normal full-decompression case.  They are not used by a
+ * transcoding-only application.  Note that if an application links in
+ * jpeg_start_decompress, it will end up linking in the entire decompressor.
+ * We thus must separate this file from jdapimin.c to avoid linking the
+ * whole decompression library into a transcoder.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Forward declarations */
+LOCAL(boolean) output_pass_setup JPP((j_decompress_ptr cinfo));
+
+
+/*
+ * Decompression initialization.
+ * jpeg_read_header must be completed before calling this.
+ *
+ * If a multipass operating mode was selected, this will do all but the
+ * last pass, and thus may take a great deal of time.
+ *
+ * Returns FALSE if suspended.  The return value need be inspected only if
+ * a suspending data source is used.
+ */
+
+GLOBAL(boolean)
+jpeg_start_decompress (j_decompress_ptr cinfo)
+{
+  if (cinfo->global_state == DSTATE_READY) {
+    /* First call: initialize master control, select active modules */
+    jinit_master_decompress(cinfo);
+    if (cinfo->buffered_image) {
+      /* No more work here; expecting jpeg_start_output next */
+      cinfo->global_state = DSTATE_BUFIMAGE;
+      return TRUE;
+    }
+    cinfo->global_state = DSTATE_PRELOAD;
+  }
+  if (cinfo->global_state == DSTATE_PRELOAD) {
+    /* If file has multiple scans, absorb them all into the coef buffer */
+    if (cinfo->inputctl->has_multiple_scans) {
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+      for (;;) {
+	int retcode;
+	/* Call progress monitor hook if present */
+	if (cinfo->progress != NULL)
+	  (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+	/* Absorb some more input */
+	retcode = (*cinfo->inputctl->consume_input) (cinfo);
+	if (retcode == JPEG_SUSPENDED)
+	  return FALSE;
+	if (retcode == JPEG_REACHED_EOI)
+	  break;
+	/* Advance progress counter if appropriate */
+	if (cinfo->progress != NULL &&
+	    (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
+	  if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
+	    /* jdmaster underestimated number of scans; ratchet up one scan */
+	    cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
+	  }
+	}
+      }
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
+    }
+    cinfo->output_scan_number = cinfo->input_scan_number;
+  } else if (cinfo->global_state != DSTATE_PRESCAN)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Perform any dummy output passes, and set up for the final pass */
+  return output_pass_setup(cinfo);
+}
+
+/*
+ * Tile decompression initialization.
+ * jpeg_read_header must be completed before calling this.
+ */
+
+GLOBAL(boolean)
+jpeg_start_tile_decompress (j_decompress_ptr cinfo)
+{
+  if (cinfo->global_state == DSTATE_READY) {
+    /* First call: initialize master control, select active modules */
+    cinfo->tile_decode = TRUE;
+    jinit_master_decompress(cinfo);
+    if (cinfo->buffered_image) {
+      cinfo->global_state = DSTATE_BUFIMAGE;
+      return TRUE;
+    }
+    cinfo->global_state = DSTATE_PRELOAD;
+  }
+  if (cinfo->global_state == DSTATE_PRELOAD) {
+    cinfo->output_scan_number = cinfo->input_scan_number;
+  } else if (cinfo->global_state != DSTATE_PRESCAN)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Perform any dummy output passes, and set up for the final pass */
+  return output_pass_setup(cinfo);
+}
+
+
+/*
+ * Set up for an output pass, and perform any dummy pass(es) needed.
+ * Common subroutine for jpeg_start_decompress and jpeg_start_output.
+ * Entry: global_state = DSTATE_PRESCAN only if previously suspended.
+ * Exit: If done, returns TRUE and sets global_state for proper output mode.
+ *       If suspended, returns FALSE and sets global_state = DSTATE_PRESCAN.
+ */
+
+LOCAL(boolean)
+output_pass_setup (j_decompress_ptr cinfo)
+{
+  if (cinfo->global_state != DSTATE_PRESCAN) {
+    /* First call: do pass setup */
+    (*cinfo->master->prepare_for_output_pass) (cinfo);
+    cinfo->output_scanline = 0;
+    cinfo->global_state = DSTATE_PRESCAN;
+  }
+  /* Loop over any required dummy passes */
+  while (cinfo->master->is_dummy_pass) {
+#ifdef QUANT_2PASS_SUPPORTED
+    /* Crank through the dummy pass */
+    while (cinfo->output_scanline < cinfo->output_height) {
+      JDIMENSION last_scanline;
+      /* Call progress monitor hook if present */
+      if (cinfo->progress != NULL) {
+	cinfo->progress->pass_counter = (long) cinfo->output_scanline;
+	cinfo->progress->pass_limit = (long) cinfo->output_height;
+	(*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+      }
+      /* Process some data */
+      last_scanline = cinfo->output_scanline;
+      (*cinfo->main->process_data) (cinfo, (JSAMPARRAY) NULL,
+				    &cinfo->output_scanline, (JDIMENSION) 0);
+      if (cinfo->output_scanline == last_scanline)
+	return FALSE;		/* No progress made, must suspend */
+    }
+    /* Finish up dummy pass, and set up for another one */
+    (*cinfo->master->finish_output_pass) (cinfo);
+    (*cinfo->master->prepare_for_output_pass) (cinfo);
+    cinfo->output_scanline = 0;
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif /* QUANT_2PASS_SUPPORTED */
+  }
+  /* Ready for application to drive output pass through
+   * jpeg_read_scanlines or jpeg_read_raw_data.
+   */
+  cinfo->global_state = cinfo->raw_data_out ? DSTATE_RAW_OK : DSTATE_SCANNING;
+  return TRUE;
+}
+
+
+/*
+ * Read some scanlines of data from the JPEG decompressor.
+ *
+ * The return value will be the number of lines actually read.
+ * This may be less than the number requested in several cases,
+ * including bottom of image, data source suspension, and operating
+ * modes that emit multiple scanlines at a time.
+ *
+ * Note: we warn about excess calls to jpeg_read_scanlines() since
+ * this likely signals an application programmer error.  However,
+ * an oversize buffer (max_lines > scanlines remaining) is not an error.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_read_scanlines (j_decompress_ptr cinfo, JSAMPARRAY scanlines,
+		     JDIMENSION max_lines)
+{
+  JDIMENSION row_ctr;
+
+  if (cinfo->global_state != DSTATE_SCANNING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  if (cinfo->output_scanline >= cinfo->output_height) {
+    WARNMS(cinfo, JWRN_TOO_MUCH_DATA);
+    return 0;
+  }
+
+  /* Call progress monitor hook if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->pass_counter = (long) cinfo->output_scanline;
+    cinfo->progress->pass_limit = (long) cinfo->output_height;
+    (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+  }
+
+  /* Process some data */
+  row_ctr = 0;
+  (*cinfo->main->process_data) (cinfo, scanlines, &row_ctr, max_lines);
+  cinfo->output_scanline += row_ctr;
+  return row_ctr;
+}
+/*
+ * Initialize the jpeg decoder to decompressing a rectangle with size of (width, height)
+ * and its upper-left corner located at (start_x, start_y).
+ * Align start_x and start_y to multiplies of iMCU width and height, respectively.
+ * Also, the new reader position and sampled image size will be returned in
+ * (start_x, start_y) and (width, height), respectively.
+ */
+
+GLOBAL(void)
+jpeg_init_read_tile_scanline(j_decompress_ptr cinfo, huffman_index *index,
+		     int *start_x, int *start_y, int *width, int *height)
+{
+  // Calculates the boundary of iMCU
+  int lines_per_iMCU_row = cinfo->max_v_samp_factor * DCTSIZE;
+  int lines_per_iMCU_col = cinfo->max_h_samp_factor * DCTSIZE;
+  int row_offset = *start_y / lines_per_iMCU_row;
+  int col_left_boundary = ((*start_x / lines_per_iMCU_col)
+            / index->MCU_sample_size) * index->MCU_sample_size;
+  int col_right_boundary =
+                  jdiv_round_up(*start_x + *width, lines_per_iMCU_col);
+
+  cinfo->coef->MCU_columns_to_skip =
+      *start_x / lines_per_iMCU_col - col_left_boundary;
+
+  *height = (*start_y - row_offset * lines_per_iMCU_row) + *height;
+  *start_x = col_left_boundary * lines_per_iMCU_col;
+  *start_y = row_offset * lines_per_iMCU_row;
+  cinfo->image_width = jmin(cinfo->original_image_width,
+          col_right_boundary * lines_per_iMCU_col) -
+          col_left_boundary * lines_per_iMCU_col;
+  cinfo->input_iMCU_row = row_offset;
+  cinfo->output_iMCU_row = row_offset;
+
+  // Updates JPEG decoder parameter
+  jinit_color_deconverter(cinfo);
+  jpeg_calc_output_dimensions(cinfo);
+  jinit_upsampler(cinfo);
+  (*cinfo->master->prepare_for_output_pass) (cinfo);
+  if (cinfo->progressive_mode)
+    (*cinfo->entropy->start_pass) (cinfo);
+  else
+    jpeg_decompress_per_scan_setup(cinfo);
+
+  int sample_size = DCTSIZE / cinfo->min_DCT_scaled_size;
+
+  *height = jdiv_round_up(*height, sample_size);
+  *width = cinfo->output_width;
+  cinfo->output_scanline = lines_per_iMCU_row * row_offset / sample_size;
+  cinfo->inputctl->consume_input = cinfo->coef->consume_data;
+  cinfo->inputctl->consume_input_build_huffman_index =
+      cinfo->coef->consume_data_build_huffman_index;
+  cinfo->entropy->index = index;
+  cinfo->input_iMCU_row = row_offset;
+  cinfo->output_iMCU_row = row_offset;
+  cinfo->coef->MCU_column_left_boundary = col_left_boundary;
+  cinfo->coef->MCU_column_right_boundary = col_right_boundary;
+  cinfo->coef->column_left_boundary =
+      col_left_boundary / index->MCU_sample_size;
+  cinfo->coef->column_right_boundary =
+      jdiv_round_up(col_right_boundary, index->MCU_sample_size);
+}
+
+/*
+ * Read a scanline from the current position.
+ *
+ * Return the number of lines actually read.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_read_tile_scanline (j_decompress_ptr cinfo, huffman_index *index,
+        JSAMPARRAY scanlines)
+{
+  // Calculates the boundary of iMCU
+  int lines_per_iMCU_row = cinfo->max_v_samp_factor * DCTSIZE;
+  int lines_per_iMCU_col = cinfo->max_h_samp_factor * DCTSIZE;
+  int sample_size = DCTSIZE / cinfo->min_DCT_scaled_size;
+  JDIMENSION row_ctr = 0;
+
+  if (cinfo->progressive_mode) {
+    (*cinfo->main->process_data) (cinfo, scanlines, &row_ctr, 1);
+  } else {
+    if (cinfo->output_scanline % (lines_per_iMCU_row / sample_size) == 0) {
+      // Set the read head to the next iMCU row
+      int iMCU_row_offset = cinfo->output_scanline /
+            (lines_per_iMCU_row / sample_size);
+      int offset_data_col_position = cinfo->coef->MCU_column_left_boundary /
+            index->MCU_sample_size;
+      huffman_offset_data offset_data =
+          index->scan[0].offset[iMCU_row_offset][offset_data_col_position];
+      (*cinfo->entropy->configure_huffman_decoder) (cinfo, offset_data);
+    }
+    (*cinfo->main->process_data) (cinfo, scanlines, &row_ctr, 1);
+  }
+
+  cinfo->output_scanline += row_ctr;
+  return row_ctr;
+}
+
+/*
+ * Alternate entry point to read raw data.
+ * Processes exactly one iMCU row per call, unless suspended.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_read_raw_data (j_decompress_ptr cinfo, JSAMPIMAGE data,
+		    JDIMENSION max_lines)
+{
+  JDIMENSION lines_per_iMCU_row;
+
+  if (cinfo->global_state != DSTATE_RAW_OK)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  if (cinfo->output_scanline >= cinfo->output_height) {
+    WARNMS(cinfo, JWRN_TOO_MUCH_DATA);
+    return 0;
+  }
+
+  /* Call progress monitor hook if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->pass_counter = (long) cinfo->output_scanline;
+    cinfo->progress->pass_limit = (long) cinfo->output_height;
+    (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+  }
+
+  /* Verify that at least one iMCU row can be returned. */
+  lines_per_iMCU_row = cinfo->max_v_samp_factor * cinfo->min_DCT_scaled_size;
+  if (max_lines < lines_per_iMCU_row)
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+
+  /* Decompress directly into user's buffer. */
+  if (! (*cinfo->coef->decompress_data) (cinfo, data))
+    return 0;			/* suspension forced, can do nothing more */
+
+  /* OK, we processed one iMCU row. */
+  cinfo->output_scanline += lines_per_iMCU_row;
+  return lines_per_iMCU_row;
+}
+
+
+/* Additional entry points for buffered-image mode. */
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+
+/*
+ * Initialize for an output pass in buffered-image mode.
+ */
+
+GLOBAL(boolean)
+jpeg_start_output (j_decompress_ptr cinfo, int scan_number)
+{
+  if (cinfo->global_state != DSTATE_BUFIMAGE &&
+      cinfo->global_state != DSTATE_PRESCAN)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Limit scan number to valid range */
+  if (scan_number <= 0)
+    scan_number = 1;
+  if (cinfo->inputctl->eoi_reached &&
+      scan_number > cinfo->input_scan_number)
+    scan_number = cinfo->input_scan_number;
+  cinfo->output_scan_number = scan_number;
+  /* Perform any dummy output passes, and set up for the real pass */
+  return output_pass_setup(cinfo);
+}
+
+
+/*
+ * Finish up after an output pass in buffered-image mode.
+ *
+ * Returns FALSE if suspended.  The return value need be inspected only if
+ * a suspending data source is used.
+ */
+
+GLOBAL(boolean)
+jpeg_finish_output (j_decompress_ptr cinfo)
+{
+  if ((cinfo->global_state == DSTATE_SCANNING ||
+       cinfo->global_state == DSTATE_RAW_OK) && cinfo->buffered_image) {
+    /* Terminate this pass. */
+    /* We do not require the whole pass to have been completed. */
+    (*cinfo->master->finish_output_pass) (cinfo);
+    cinfo->global_state = DSTATE_BUFPOST;
+  } else if (cinfo->global_state != DSTATE_BUFPOST) {
+    /* BUFPOST = repeat call after a suspension, anything else is error */
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  }
+  /* Read markers looking for SOS or EOI */
+  while (cinfo->input_scan_number <= cinfo->output_scan_number &&
+	 ! cinfo->inputctl->eoi_reached) {
+    if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED)
+      return FALSE;		/* Suspend, come back later */
+  }
+  cinfo->global_state = DSTATE_BUFIMAGE;
+  return TRUE;
+}
+
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
diff --git a/libjpegtwrp/jdatadst.c b/libjpegtwrp/jdatadst.c
new file mode 100644
index 000000000..a8f6fb0e0
--- /dev/null
+++ b/libjpegtwrp/jdatadst.c
@@ -0,0 +1,151 @@
+/*
+ * jdatadst.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains compression data destination routines for the case of
+ * emitting JPEG data to a file (or any stdio stream).  While these routines
+ * are sufficient for most applications, some will want to use a different
+ * destination manager.
+ * IMPORTANT: we assume that fwrite() will correctly transcribe an array of
+ * JOCTETs into 8-bit-wide elements on external storage.  If char is wider
+ * than 8 bits on your machine, you may need to do some tweaking.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+
+/* Expanded data destination object for stdio output */
+
+typedef struct {
+  struct jpeg_destination_mgr pub; /* public fields */
+
+  FILE * outfile;		/* target stream */
+  JOCTET * buffer;		/* start of buffer */
+} my_destination_mgr;
+
+typedef my_destination_mgr * my_dest_ptr;
+
+#define OUTPUT_BUF_SIZE  4096	/* choose an efficiently fwrite'able size */
+
+
+/*
+ * Initialize destination --- called by jpeg_start_compress
+ * before any data is actually written.
+ */
+
+METHODDEF(void)
+init_destination (j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;
+
+  /* Allocate the output buffer --- it will be released when done with image */
+  dest->buffer = (JOCTET *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  OUTPUT_BUF_SIZE * SIZEOF(JOCTET));
+
+  dest->pub.next_output_byte = dest->buffer;
+  dest->pub.free_in_buffer = OUTPUT_BUF_SIZE;
+}
+
+
+/*
+ * Empty the output buffer --- called whenever buffer fills up.
+ *
+ * In typical applications, this should write the entire output buffer
+ * (ignoring the current state of next_output_byte & free_in_buffer),
+ * reset the pointer & count to the start of the buffer, and return TRUE
+ * indicating that the buffer has been dumped.
+ *
+ * In applications that need to be able to suspend compression due to output
+ * overrun, a FALSE return indicates that the buffer cannot be emptied now.
+ * In this situation, the compressor will return to its caller (possibly with
+ * an indication that it has not accepted all the supplied scanlines).  The
+ * application should resume compression after it has made more room in the
+ * output buffer.  Note that there are substantial restrictions on the use of
+ * suspension --- see the documentation.
+ *
+ * When suspending, the compressor will back up to a convenient restart point
+ * (typically the start of the current MCU). next_output_byte & free_in_buffer
+ * indicate where the restart point will be if the current call returns FALSE.
+ * Data beyond this point will be regenerated after resumption, so do not
+ * write it out when emptying the buffer externally.
+ */
+
+METHODDEF(boolean)
+empty_output_buffer (j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;
+
+  if (JFWRITE(dest->outfile, dest->buffer, OUTPUT_BUF_SIZE) !=
+      (size_t) OUTPUT_BUF_SIZE)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  dest->pub.next_output_byte = dest->buffer;
+  dest->pub.free_in_buffer = OUTPUT_BUF_SIZE;
+
+  return TRUE;
+}
+
+
+/*
+ * Terminate destination --- called by jpeg_finish_compress
+ * after all data has been written.  Usually needs to flush buffer.
+ *
+ * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding
+ * application must deal with any cleanup that should happen even
+ * for error exit.
+ */
+
+METHODDEF(void)
+term_destination (j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;
+  size_t datacount = OUTPUT_BUF_SIZE - dest->pub.free_in_buffer;
+
+  /* Write any data remaining in the buffer */
+  if (datacount > 0) {
+    if (JFWRITE(dest->outfile, dest->buffer, datacount) != datacount)
+      ERREXIT(cinfo, JERR_FILE_WRITE);
+  }
+  fflush(dest->outfile);
+  /* Make sure we wrote the output file OK */
+  if (ferror(dest->outfile))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * Prepare for output to a stdio stream.
+ * The caller must have already opened the stream, and is responsible
+ * for closing it after finishing compression.
+ */
+
+GLOBAL(void)
+jpeg_stdio_dest (j_compress_ptr cinfo, FILE * outfile)
+{
+  my_dest_ptr dest;
+
+  /* The destination object is made permanent so that multiple JPEG images
+   * can be written to the same file without re-executing jpeg_stdio_dest.
+   * This makes it dangerous to use this manager and a different destination
+   * manager serially with the same JPEG object, because their private object
+   * sizes may be different.  Caveat programmer.
+   */
+  if (cinfo->dest == NULL) {	/* first time for this JPEG object? */
+    cinfo->dest = (struct jpeg_destination_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  SIZEOF(my_destination_mgr));
+  }
+
+  dest = (my_dest_ptr) cinfo->dest;
+  dest->pub.init_destination = init_destination;
+  dest->pub.empty_output_buffer = empty_output_buffer;
+  dest->pub.term_destination = term_destination;
+  dest->outfile = outfile;
+}
diff --git a/libjpegtwrp/jdatasrc.c b/libjpegtwrp/jdatasrc.c
new file mode 100644
index 000000000..edc752bf5
--- /dev/null
+++ b/libjpegtwrp/jdatasrc.c
@@ -0,0 +1,212 @@
+/*
+ * jdatasrc.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains decompression data source routines for the case of
+ * reading JPEG data from a file (or any stdio stream).  While these routines
+ * are sufficient for most applications, some will want to use a different
+ * source manager.
+ * IMPORTANT: we assume that fread() will correctly transcribe an array of
+ * JOCTETs from 8-bit-wide elements on external storage.  If char is wider
+ * than 8 bits on your machine, you may need to do some tweaking.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+
+/* Expanded data source object for stdio input */
+
+typedef struct {
+  struct jpeg_source_mgr pub;	/* public fields */
+
+  FILE * infile;		/* source stream */
+  JOCTET * buffer;		/* start of buffer */
+  boolean start_of_file;	/* have we gotten any data yet? */
+} my_source_mgr;
+
+typedef my_source_mgr * my_src_ptr;
+
+#define INPUT_BUF_SIZE  4096	/* choose an efficiently fread'able size */
+
+
+/*
+ * Initialize source --- called by jpeg_read_header
+ * before any data is actually read.
+ */
+
+METHODDEF(void)
+init_source (j_decompress_ptr cinfo)
+{
+  my_src_ptr src = (my_src_ptr) cinfo->src;
+
+  /* We reset the empty-input-file flag for each image,
+   * but we don't clear the input buffer.
+   * This is correct behavior for reading a series of images from one source.
+   */
+  src->start_of_file = TRUE;
+}
+
+
+/*
+ * Fill the input buffer --- called whenever buffer is emptied.
+ *
+ * In typical applications, this should read fresh data into the buffer
+ * (ignoring the current state of next_input_byte & bytes_in_buffer),
+ * reset the pointer & count to the start of the buffer, and return TRUE
+ * indicating that the buffer has been reloaded.  It is not necessary to
+ * fill the buffer entirely, only to obtain at least one more byte.
+ *
+ * There is no such thing as an EOF return.  If the end of the file has been
+ * reached, the routine has a choice of ERREXIT() or inserting fake data into
+ * the buffer.  In most cases, generating a warning message and inserting a
+ * fake EOI marker is the best course of action --- this will allow the
+ * decompressor to output however much of the image is there.  However,
+ * the resulting error message is misleading if the real problem is an empty
+ * input file, so we handle that case specially.
+ *
+ * In applications that need to be able to suspend compression due to input
+ * not being available yet, a FALSE return indicates that no more data can be
+ * obtained right now, but more may be forthcoming later.  In this situation,
+ * the decompressor will return to its caller (with an indication of the
+ * number of scanlines it has read, if any).  The application should resume
+ * decompression after it has loaded more data into the input buffer.  Note
+ * that there are substantial restrictions on the use of suspension --- see
+ * the documentation.
+ *
+ * When suspending, the decompressor will back up to a convenient restart point
+ * (typically the start of the current MCU). next_input_byte & bytes_in_buffer
+ * indicate where the restart point will be if the current call returns FALSE.
+ * Data beyond this point must be rescanned after resumption, so move it to
+ * the front of the buffer rather than discarding it.
+ */
+
+METHODDEF(boolean)
+fill_input_buffer (j_decompress_ptr cinfo)
+{
+  my_src_ptr src = (my_src_ptr) cinfo->src;
+  size_t nbytes;
+
+  nbytes = JFREAD(src->infile, src->buffer, INPUT_BUF_SIZE);
+
+  if (nbytes <= 0) {
+    if (src->start_of_file)	/* Treat empty input file as fatal error */
+      ERREXIT(cinfo, JERR_INPUT_EMPTY);
+    WARNMS(cinfo, JWRN_JPEG_EOF);
+    /* Insert a fake EOI marker */
+    src->buffer[0] = (JOCTET) 0xFF;
+    src->buffer[1] = (JOCTET) JPEG_EOI;
+    nbytes = 2;
+  }
+
+  src->pub.next_input_byte = src->buffer;
+  src->pub.bytes_in_buffer = nbytes;
+  src->start_of_file = FALSE;
+
+  return TRUE;
+}
+
+
+/*
+ * Skip data --- used to skip over a potentially large amount of
+ * uninteresting data (such as an APPn marker).
+ *
+ * Writers of suspendable-input applications must note that skip_input_data
+ * is not granted the right to give a suspension return.  If the skip extends
+ * beyond the data currently in the buffer, the buffer can be marked empty so
+ * that the next read will cause a fill_input_buffer call that can suspend.
+ * Arranging for additional bytes to be discarded before reloading the input
+ * buffer is the application writer's problem.
+ */
+
+METHODDEF(void)
+skip_input_data (j_decompress_ptr cinfo, long num_bytes)
+{
+  my_src_ptr src = (my_src_ptr) cinfo->src;
+
+  /* Just a dumb implementation for now.  Could use fseek() except
+   * it doesn't work on pipes.  Not clear that being smart is worth
+   * any trouble anyway --- large skips are infrequent.
+   */
+  if (num_bytes > 0) {
+    while (num_bytes > (long) src->pub.bytes_in_buffer) {
+      num_bytes -= (long) src->pub.bytes_in_buffer;
+      (void) fill_input_buffer(cinfo);
+      /* note we assume that fill_input_buffer will never return FALSE,
+       * so suspension need not be handled.
+       */
+    }
+    src->pub.next_input_byte += (size_t) num_bytes;
+    src->pub.bytes_in_buffer -= (size_t) num_bytes;
+  }
+}
+
+
+/*
+ * An additional method that can be provided by data source modules is the
+ * resync_to_restart method for error recovery in the presence of RST markers.
+ * For the moment, this source module just uses the default resync method
+ * provided by the JPEG library.  That method assumes that no backtracking
+ * is possible.
+ */
+
+
+/*
+ * Terminate source --- called by jpeg_finish_decompress
+ * after all data has been read.  Often a no-op.
+ *
+ * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding
+ * application must deal with any cleanup that should happen even
+ * for error exit.
+ */
+
+METHODDEF(void)
+term_source (j_decompress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+
+
+/*
+ * Prepare for input from a stdio stream.
+ * The caller must have already opened the stream, and is responsible
+ * for closing it after finishing decompression.
+ */
+
+GLOBAL(void)
+jpeg_stdio_src (j_decompress_ptr cinfo, FILE * infile)
+{
+  my_src_ptr src;
+
+  /* The source object and input buffer are made permanent so that a series
+   * of JPEG images can be read from the same file by calling jpeg_stdio_src
+   * only before the first one.  (If we discarded the buffer at the end of
+   * one image, we'd likely lose the start of the next one.)
+   * This makes it unsafe to use this manager and a different source
+   * manager serially with the same JPEG object.  Caveat programmer.
+   */
+  if (cinfo->src == NULL) {	/* first time for this JPEG object? */
+    cinfo->src = (struct jpeg_source_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  SIZEOF(my_source_mgr));
+    src = (my_src_ptr) cinfo->src;
+    src->buffer = (JOCTET *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  INPUT_BUF_SIZE * SIZEOF(JOCTET));
+  }
+
+  src = (my_src_ptr) cinfo->src;
+  src->pub.init_source = init_source;
+  src->pub.fill_input_buffer = fill_input_buffer;
+  src->pub.skip_input_data = skip_input_data;
+  src->pub.resync_to_restart = jpeg_resync_to_restart; /* use default method */
+  src->pub.term_source = term_source;
+  src->infile = infile;
+  src->pub.bytes_in_buffer = 0; /* forces fill_input_buffer on first read */
+  src->pub.next_input_byte = NULL; /* until buffer loaded */
+}
diff --git a/libjpegtwrp/jdcoefct.c b/libjpegtwrp/jdcoefct.c
new file mode 100644
index 000000000..84ce56fb7
--- /dev/null
+++ b/libjpegtwrp/jdcoefct.c
@@ -0,0 +1,1038 @@
+/*
+ * jdcoefct.c
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the coefficient buffer controller for decompression.
+ * This controller is the top level of the JPEG decompressor proper.
+ * The coefficient buffer lies between entropy decoding and inverse-DCT steps.
+ *
+ * In buffered-image mode, this controller is the interface between
+ * input-oriented processing and output-oriented processing.
+ * Also, the input side (only) is used when reading a file for transcoding.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+/* Block smoothing is only applicable for progressive JPEG, so: */
+#ifndef D_PROGRESSIVE_SUPPORTED
+#undef BLOCK_SMOOTHING_SUPPORTED
+#endif
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_coef_controller pub; /* public fields */
+
+  /* These variables keep track of the current location of the input side. */
+  /* cinfo->input_iMCU_row is also used for this. */
+  JDIMENSION MCU_ctr;		/* counts MCUs processed in current row */
+  int MCU_vert_offset;		/* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;	/* number of such rows needed */
+
+  /* The output side's location is represented by cinfo->output_iMCU_row. */
+
+  /* In single-pass modes, it's sufficient to buffer just one MCU.
+   * We allocate a workspace of D_MAX_BLOCKS_IN_MCU coefficient blocks,
+   * and let the entropy decoder write into that workspace each time.
+   * (On 80x86, the workspace is FAR even though it's not really very big;
+   * this is to keep the module interfaces unchanged when a large coefficient
+   * buffer is necessary.)
+   * In multi-pass modes, this array points to the current MCU's blocks
+   * within the virtual arrays; it is used only by the input side.
+   */
+  JBLOCKROW MCU_buffer[D_MAX_BLOCKS_IN_MCU];
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+  /* In multi-pass modes, we need a virtual block array for each component. */
+  jvirt_barray_ptr whole_image[MAX_COMPONENTS];
+#endif
+
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+  /* When doing block smoothing, we latch coefficient Al values here */
+  int * coef_bits_latch;
+#define SAVED_COEFS  6		/* we save coef_bits[0..5] */
+#endif
+} my_coef_controller;
+
+typedef my_coef_controller * my_coef_ptr;
+
+/* Forward declarations */
+METHODDEF(int) decompress_onepass
+	JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf));
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+METHODDEF(int) decompress_data
+	JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf));
+#endif
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+LOCAL(boolean) smoothing_ok JPP((j_decompress_ptr cinfo));
+METHODDEF(int) decompress_smooth_data
+	JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf));
+#endif
+
+
+LOCAL(void)
+start_iMCU_row (j_decompress_ptr cinfo)
+/* Reset within-iMCU-row counters for a new row (input side) */
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  /* In an interleaved scan, an MCU row is the same as an iMCU row.
+   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
+   * But at the bottom of the image, process only what's left.
+   */
+  if (cinfo->comps_in_scan > 1) {
+    coef->MCU_rows_per_iMCU_row = 1;
+  } else {
+    if (cinfo->input_iMCU_row < (cinfo->total_iMCU_rows-1))
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->v_samp_factor;
+    else
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->last_row_height;
+  }
+
+  coef->MCU_ctr = 0;
+  coef->MCU_vert_offset = 0;
+}
+
+
+/*
+ * Initialize for an input processing pass.
+ */
+
+METHODDEF(void)
+start_input_pass (j_decompress_ptr cinfo)
+{
+  cinfo->input_iMCU_row = 0;
+  start_iMCU_row(cinfo);
+}
+
+
+/*
+ * Initialize for an output processing pass.
+ */
+
+METHODDEF(void)
+start_output_pass (j_decompress_ptr cinfo)
+{
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  /* If multipass, check to see whether to use block smoothing on this pass */
+  if (coef->pub.coef_arrays != NULL) {
+    if (cinfo->do_block_smoothing && smoothing_ok(cinfo))
+      coef->pub.decompress_data = decompress_smooth_data;
+    else
+      coef->pub.decompress_data = decompress_data;
+  }
+#endif
+  cinfo->output_iMCU_row = 0;
+}
+
+
+/*
+ * Decompress and return some data in the single-pass case.
+ * Always attempts to emit one fully interleaved MCU row ("iMCU" row).
+ * Input and output must run in lockstep since we have only a one-MCU buffer.
+ * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED.
+ *
+ * NB: output_buf contains a plane for each component in image,
+ * which we index according to the component's SOF position.
+ */
+
+METHODDEF(int)
+decompress_onepass (j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  int blkn, ci, xindex, yindex, yoffset, useful_width;
+  JSAMPARRAY output_ptr;
+  JDIMENSION start_col, output_col;
+  jpeg_component_info *compptr;
+  inverse_DCT_method_ptr inverse_DCT;
+
+#ifdef ANDROID_TILE_BASED_DECODE
+  if (cinfo->tile_decode) {
+    last_MCU_col =
+        (cinfo->coef->MCU_column_right_boundary -
+         cinfo->coef->MCU_column_left_boundary) - 1;
+  }
+#endif
+
+  /* Loop to process as much as one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->MCU_ctr; MCU_col_num <= last_MCU_col;
+	 MCU_col_num++) {
+      /* Try to fetch an MCU.  Entropy decoder expects buffer to be zeroed. */
+      if (MCU_col_num < (unsigned int)coef->pub.MCU_columns_to_skip) {
+        (*cinfo->entropy->decode_mcu_discard_coef) (cinfo);
+        continue;
+      } else {
+        jzero_far((void FAR *) coef->MCU_buffer[0],
+		(size_t) (cinfo->blocks_in_MCU * SIZEOF(JBLOCK)));
+        if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
+	  /* Suspension forced; update state counters and exit */
+	  coef->MCU_vert_offset = yoffset;
+	  coef->MCU_ctr = MCU_col_num;
+	  return JPEG_SUSPENDED;
+        }
+      }
+      /* Determine where data should go in output_buf and do the IDCT thing.
+       * We skip dummy blocks at the right and bottom edges (but blkn gets
+       * incremented past them!).  Note the inner loop relies on having
+       * allocated the MCU_buffer[] blocks sequentially.
+       */
+      blkn = 0;			/* index of current DCT block within MCU */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+	compptr = cinfo->cur_comp_info[ci];
+	/* Don't bother to IDCT an uninteresting component. */
+	if (! compptr->component_needed) {
+	  blkn += compptr->MCU_blocks;
+	  continue;
+	}
+	inverse_DCT = cinfo->idct->inverse_DCT[compptr->component_index];
+	useful_width = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
+						    : compptr->last_col_width;
+	output_ptr = output_buf[compptr->component_index] +
+	  yoffset * compptr->DCT_scaled_size;
+	start_col = MCU_col_num * compptr->MCU_sample_width;
+	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+	  if (cinfo->input_iMCU_row < last_iMCU_row ||
+	      yoffset+yindex < compptr->last_row_height) {
+	    output_col = start_col;
+	    for (xindex = 0; xindex < useful_width; xindex++) {
+	      (*inverse_DCT) (cinfo, compptr,
+		        (JCOEFPTR) coef->MCU_buffer[blkn+xindex],
+		        output_ptr, output_col);
+	      output_col += compptr->DCT_scaled_size;
+	    }
+	  }
+	  blkn += compptr->MCU_width;
+	  output_ptr += compptr->DCT_scaled_size;
+	}
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->MCU_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  cinfo->output_iMCU_row++;
+  if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) {
+    start_iMCU_row(cinfo);
+    return JPEG_ROW_COMPLETED;
+  }
+  /* Completed the scan */
+  (*cinfo->inputctl->finish_input_pass) (cinfo);
+  return JPEG_SCAN_COMPLETED;
+}
+
+
+/*
+ * Dummy consume-input routine for single-pass operation.
+ */
+
+METHODDEF(int)
+dummy_consume_data (j_decompress_ptr cinfo)
+{
+  return JPEG_SUSPENDED;	/* Always indicate nothing was done */
+}
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+/*
+ * Consume input data and store it in the full-image coefficient buffer.
+ * We read as much as one fully interleaved MCU row ("iMCU" row) per call,
+ * ie, v_samp_factor block rows for each component in the scan.
+ * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED.
+ */
+
+METHODDEF(int)
+consume_data (j_decompress_ptr cinfo)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  int blkn, ci, xindex, yindex, yoffset;
+  JDIMENSION start_col;
+  JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN];
+  JBLOCKROW buffer_ptr;
+  jpeg_component_info *compptr;
+
+  /* Align the virtual buffers for the components used in this scan. */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    buffer[ci] = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr) cinfo, coef->whole_image[compptr->component_index],
+       cinfo->tile_decode ? 0 : cinfo->input_iMCU_row * compptr->v_samp_factor,
+       (JDIMENSION) compptr->v_samp_factor, TRUE);
+    /* Note: entropy decoder expects buffer to be zeroed,
+     * but this is handled automatically by the memory manager
+     * because we requested a pre-zeroed array.
+     */
+  }
+  unsigned int MCUs_per_row = cinfo->MCUs_per_row;
+#ifdef ANDROID_TILE_BASED_DECODE
+  if (cinfo->tile_decode) {
+    int iMCU_width_To_MCU_width;
+    if (cinfo->comps_in_scan > 1) {
+      // Interleaved
+      iMCU_width_To_MCU_width = 1;
+    } else {
+      // Non-intervleaved
+      iMCU_width_To_MCU_width = cinfo->cur_comp_info[0]->h_samp_factor;
+    }
+    MCUs_per_row = jmin(MCUs_per_row,
+        (cinfo->coef->column_right_boundary - cinfo->coef->column_left_boundary)
+        * cinfo->entropy->index->MCU_sample_size * iMCU_width_To_MCU_width);
+  }
+#endif
+
+  /* Loop to process one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+   // configure huffman decoder
+#ifdef ANDROID_TILE_BASED_DECODE
+    if (cinfo->tile_decode) {
+      huffman_scan_header scan_header =
+            cinfo->entropy->index->scan[cinfo->input_scan_number];
+      int col_offset = cinfo->coef->column_left_boundary;
+      (*cinfo->entropy->configure_huffman_decoder) (cinfo,
+              scan_header.offset[cinfo->input_iMCU_row]
+              [col_offset + yoffset * scan_header.MCUs_per_row]);
+    }
+#endif
+
+    // zero all blocks
+    for (MCU_col_num = coef->MCU_ctr; MCU_col_num < MCUs_per_row;
+          MCU_col_num++) {
+      /* Construct list of pointers to DCT blocks belonging to this MCU */
+      blkn = 0;			/* index of current DCT block within MCU */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+        compptr = cinfo->cur_comp_info[ci];
+        start_col = MCU_col_num * compptr->MCU_width;
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+          for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
+            coef->MCU_buffer[blkn++] = buffer_ptr++;
+#ifdef ANDROID_TILE_BASED_DECODE
+            if (cinfo->tile_decode && cinfo->input_scan_number == 0) {
+              // need to do pre-zero ourselves.
+              jzero_far((void FAR *) coef->MCU_buffer[blkn-1],
+                        (size_t) (SIZEOF(JBLOCK)));
+            }
+#endif
+          }
+        }
+      }
+
+
+      /* Try to fetch the MCU. */
+      if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->MCU_ctr = MCU_col_num;
+        return JPEG_SUSPENDED;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->MCU_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) {
+    start_iMCU_row(cinfo);
+    return JPEG_ROW_COMPLETED;
+  }
+  /* Completed the scan */
+  (*cinfo->inputctl->finish_input_pass) (cinfo);
+  return JPEG_SCAN_COMPLETED;
+}
+
+/*
+ * Consume input data and store it in the coefficient buffer.
+ * Read one fully interleaved MCU row ("iMCU" row) per call.
+ */
+
+METHODDEF(int)
+consume_data_multi_scan (j_decompress_ptr cinfo)
+{
+  huffman_index *index = cinfo->entropy->index;
+  int i, retcode, ci;
+  int mcu = cinfo->input_iMCU_row;
+  jinit_phuff_decoder(cinfo);
+  for (i = 0; i < index->scan_count; i++) {
+    (*cinfo->inputctl->finish_input_pass) (cinfo);
+    jset_input_stream_position(cinfo, index->scan[i].bitstream_offset);
+    cinfo->output_iMCU_row = mcu;
+    cinfo->unread_marker = 0;
+    // Consume SOS and DHT headers
+    retcode = (*cinfo->inputctl->consume_markers) (cinfo, index, i);
+    cinfo->input_iMCU_row = mcu;
+    cinfo->input_scan_number = i;
+    cinfo->entropy->index = index;
+    // Consume scan block data
+    consume_data(cinfo);
+  }
+  cinfo->input_iMCU_row = mcu + 1;
+  cinfo->input_scan_number = 0;
+  cinfo->output_scan_number = 0;
+  return JPEG_ROW_COMPLETED;
+}
+
+/*
+ * Same as consume_data, expect for saving the Huffman decode information
+ * - bitstream offset and DC coefficient to index.
+ */
+
+METHODDEF(int)
+consume_data_build_huffman_index_baseline (j_decompress_ptr cinfo,
+        huffman_index *index, int current_scan)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  int ci, xindex, yindex, yoffset;
+  JDIMENSION start_col;
+  JBLOCKROW buffer_ptr;
+
+  huffman_scan_header *scan_header = index->scan + current_scan;
+  scan_header->MCU_rows_per_iMCU_row = coef->MCU_rows_per_iMCU_row;
+
+  size_t allocate_size = coef->MCU_rows_per_iMCU_row
+      * jdiv_round_up(cinfo->MCUs_per_row, index->MCU_sample_size)
+      * sizeof(huffman_offset_data);
+  scan_header->offset[cinfo->input_iMCU_row] =
+        (huffman_offset_data*)malloc(allocate_size);
+  index->mem_used += allocate_size;
+
+  huffman_offset_data *offset_data = scan_header->offset[cinfo->input_iMCU_row];
+
+  /* Loop to process one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->MCU_ctr; MCU_col_num < cinfo->MCUs_per_row;
+	 MCU_col_num++) {
+      // Record huffman bit offset
+      if (MCU_col_num % index->MCU_sample_size == 0) {
+        (*cinfo->entropy->get_huffman_decoder_configuration)
+                (cinfo, offset_data);
+        ++offset_data;
+      }
+
+      /* Try to fetch the MCU. */
+      if (! (*cinfo->entropy->decode_mcu_discard_coef) (cinfo)) {
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->MCU_ctr = MCU_col_num;
+        return JPEG_SUSPENDED;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->MCU_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) {
+    start_iMCU_row(cinfo);
+    return JPEG_ROW_COMPLETED;
+  }
+  /* Completed the scan */
+  (*cinfo->inputctl->finish_input_pass) (cinfo);
+  return JPEG_SCAN_COMPLETED;
+}
+
+/*
+ * Same as consume_data, expect for saving the Huffman decode information
+ * - bitstream offset and DC coefficient to index.
+ */
+
+METHODDEF(int)
+consume_data_build_huffman_index_progressive (j_decompress_ptr cinfo,
+        huffman_index *index, int current_scan)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  int blkn, ci, xindex, yindex, yoffset;
+  JDIMENSION start_col;
+  JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN];
+  JBLOCKROW buffer_ptr;
+  jpeg_component_info *compptr;
+
+  int factor = 4; // maximum factor is 4.
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++)
+    factor = jmin(factor, cinfo->cur_comp_info[ci]->h_samp_factor);
+
+  int sample_size = index->MCU_sample_size * factor;
+  huffman_scan_header *scan_header = index->scan + current_scan;
+  scan_header->MCU_rows_per_iMCU_row = coef->MCU_rows_per_iMCU_row;
+  scan_header->MCUs_per_row = jdiv_round_up(cinfo->MCUs_per_row, sample_size);
+  scan_header->comps_in_scan = cinfo->comps_in_scan;
+
+  size_t allocate_size = coef->MCU_rows_per_iMCU_row
+      * scan_header->MCUs_per_row * sizeof(huffman_offset_data);
+  scan_header->offset[cinfo->input_iMCU_row] =
+        (huffman_offset_data*)malloc(allocate_size);
+  index->mem_used += allocate_size;
+
+  huffman_offset_data *offset_data = scan_header->offset[cinfo->input_iMCU_row];
+
+  /* Align the virtual buffers for the components used in this scan. */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    buffer[ci] = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr) cinfo, coef->whole_image[compptr->component_index],
+       0, // Only need one row buffer
+       (JDIMENSION) compptr->v_samp_factor, TRUE);
+  }
+  /* Loop to process one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->MCU_ctr; MCU_col_num < cinfo->MCUs_per_row;
+	 MCU_col_num++) {
+      /* For each MCU, we loop through different color components.
+       * Then, for each color component we will get a list of pointers to DCT
+       * blocks in the virtual buffer.
+       */
+      blkn = 0; /* index of current DCT block within MCU */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+        compptr = cinfo->cur_comp_info[ci];
+        start_col = MCU_col_num * compptr->MCU_width;
+        /* Get the list of pointers to DCT blocks in
+         * the virtual buffer in a color component of the MCU.
+         */
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+          for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
+            coef->MCU_buffer[blkn++] = buffer_ptr++;
+            if (cinfo->input_scan_number == 0) {
+              // need to do pre-zero by ourself.
+              jzero_far((void FAR *) coef->MCU_buffer[blkn-1],
+                        (size_t) (SIZEOF(JBLOCK)));
+            }
+          }
+        }
+      }
+      // Record huffman bit offset
+      if (MCU_col_num % sample_size == 0) {
+        (*cinfo->entropy->get_huffman_decoder_configuration)
+                (cinfo, offset_data);
+        ++offset_data;
+      }
+      /* Try to fetch the MCU. */
+      if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
+	/* Suspension forced; update state counters and exit */
+	coef->MCU_vert_offset = yoffset;
+	coef->MCU_ctr = MCU_col_num;
+	return JPEG_SUSPENDED;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->MCU_ctr = 0;
+  }
+  (*cinfo->entropy->get_huffman_decoder_configuration)
+        (cinfo, &scan_header->prev_MCU_offset);
+  /* Completed the iMCU row, advance counters for next one */
+  if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) {
+    start_iMCU_row(cinfo);
+    return JPEG_ROW_COMPLETED;
+  }
+  /* Completed the scan */
+  (*cinfo->inputctl->finish_input_pass) (cinfo);
+  return JPEG_SCAN_COMPLETED;
+}
+
+/*
+ * Decompress and return some data in the multi-pass case.
+ * Always attempts to emit one fully interleaved MCU row ("iMCU" row).
+ * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED.
+ *
+ * NB: output_buf contains a plane for each component in image.
+ */
+
+METHODDEF(int)
+decompress_data (j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  JDIMENSION block_num;
+  int ci, block_row, block_rows;
+  JBLOCKARRAY buffer;
+  JBLOCKROW buffer_ptr;
+  JSAMPARRAY output_ptr;
+  JDIMENSION output_col;
+  jpeg_component_info *compptr;
+  inverse_DCT_method_ptr inverse_DCT;
+
+  /* Force some input to be done if we are getting ahead of the input. */
+  while (cinfo->input_scan_number < cinfo->output_scan_number ||
+	 (cinfo->input_scan_number == cinfo->output_scan_number &&
+	  cinfo->input_iMCU_row <= cinfo->output_iMCU_row)) {
+    if ((*cinfo->inputctl->consume_input)(cinfo) == JPEG_SUSPENDED)
+      return JPEG_SUSPENDED;
+  }
+
+  /* OK, output from the virtual arrays. */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Don't bother to IDCT an uninteresting component. */
+    if (! compptr->component_needed)
+      continue;
+    /* Align the virtual buffer for this component. */
+    buffer = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr) cinfo, coef->whole_image[ci],
+       cinfo->tile_decode ? 0 : cinfo->output_iMCU_row * compptr->v_samp_factor,
+       (JDIMENSION) compptr->v_samp_factor, FALSE);
+    /* Count non-dummy DCT block rows in this iMCU row. */
+    if (cinfo->output_iMCU_row < last_iMCU_row)
+      block_rows = compptr->v_samp_factor;
+    else {
+      /* NB: can't use last_row_height here; it is input-side-dependent! */
+      block_rows = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
+      if (block_rows == 0) block_rows = compptr->v_samp_factor;
+    }
+    inverse_DCT = cinfo->idct->inverse_DCT[ci];
+    output_ptr = output_buf[ci];
+    int width_in_blocks = compptr->width_in_blocks;
+    int start_block = 0;
+#if ANDROID_TILE_BASED_DECODE
+    if (cinfo->tile_decode) {
+      // width_in_blocks for a component depends on its h_samp_factor.
+      width_in_blocks = jmin(width_in_blocks,
+        (cinfo->coef->MCU_column_right_boundary -
+         cinfo->coef->MCU_column_left_boundary) *
+         compptr->h_samp_factor);
+      start_block = coef->pub.MCU_columns_to_skip *
+         compptr->h_samp_factor;
+   }
+#endif
+    /* Loop over all DCT blocks to be processed. */
+    for (block_row = 0; block_row < block_rows; block_row++) {
+      buffer_ptr = buffer[block_row];
+      output_col = start_block * compptr->DCT_scaled_size;
+      buffer_ptr += start_block;
+      for (block_num = start_block; block_num < (unsigned int)width_in_blocks; block_num++) {
+	(*inverse_DCT) (cinfo, compptr, (JCOEFPTR) buffer_ptr,
+			output_ptr, output_col);
+	buffer_ptr++;
+	output_col += compptr->DCT_scaled_size;
+      }
+      output_ptr += compptr->DCT_scaled_size;
+    }
+  }
+
+  if (++(cinfo->output_iMCU_row) < cinfo->total_iMCU_rows)
+    return JPEG_ROW_COMPLETED;
+  return JPEG_SCAN_COMPLETED;
+}
+
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
+
+
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+
+/*
+ * This code applies interblock smoothing as described by section K.8
+ * of the JPEG standard: the first 5 AC coefficients are estimated from
+ * the DC values of a DCT block and its 8 neighboring blocks.
+ * We apply smoothing only for progressive JPEG decoding, and only if
+ * the coefficients it can estimate are not yet known to full precision.
+ */
+
+/* Natural-order array positions of the first 5 zigzag-order coefficients */
+#define Q01_POS  1
+#define Q10_POS  8
+#define Q20_POS  16
+#define Q11_POS  9
+#define Q02_POS  2
+
+/*
+ * Determine whether block smoothing is applicable and safe.
+ * We also latch the current states of the coef_bits[] entries for the
+ * AC coefficients; otherwise, if the input side of the decompressor
+ * advances into a new scan, we might think the coefficients are known
+ * more accurately than they really are.
+ */
+
+LOCAL(boolean)
+smoothing_ok (j_decompress_ptr cinfo)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  boolean smoothing_useful = FALSE;
+  int ci, coefi;
+  jpeg_component_info *compptr;
+  JQUANT_TBL * qtable;
+  int * coef_bits;
+  int * coef_bits_latch;
+
+  if (! cinfo->progressive_mode || cinfo->coef_bits == NULL)
+    return FALSE;
+
+  /* Allocate latch area if not already done */
+  if (coef->coef_bits_latch == NULL)
+    coef->coef_bits_latch = (int *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  cinfo->num_components *
+				  (SAVED_COEFS * SIZEOF(int)));
+  coef_bits_latch = coef->coef_bits_latch;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* All components' quantization values must already be latched. */
+    if ((qtable = compptr->quant_table) == NULL)
+      return FALSE;
+    /* Verify DC & first 5 AC quantizers are nonzero to avoid zero-divide. */
+    if (qtable->quantval[0] == 0 ||
+	qtable->quantval[Q01_POS] == 0 ||
+	qtable->quantval[Q10_POS] == 0 ||
+	qtable->quantval[Q20_POS] == 0 ||
+	qtable->quantval[Q11_POS] == 0 ||
+	qtable->quantval[Q02_POS] == 0)
+      return FALSE;
+    /* DC values must be at least partly known for all components. */
+    coef_bits = cinfo->coef_bits[ci];
+    if (coef_bits[0] < 0)
+      return FALSE;
+    /* Block smoothing is helpful if some AC coefficients remain inaccurate. */
+    for (coefi = 1; coefi <= 5; coefi++) {
+      coef_bits_latch[coefi] = coef_bits[coefi];
+      if (coef_bits[coefi] != 0)
+	smoothing_useful = TRUE;
+    }
+    coef_bits_latch += SAVED_COEFS;
+  }
+
+  return smoothing_useful;
+}
+
+
+/*
+ * Variant of decompress_data for use when doing block smoothing.
+ */
+
+METHODDEF(int)
+decompress_smooth_data (j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  JDIMENSION block_num, last_block_column;
+  int ci, block_row, block_rows, access_rows;
+  JBLOCKARRAY buffer;
+  JBLOCKROW buffer_ptr, prev_block_row, next_block_row;
+  JSAMPARRAY output_ptr;
+  JDIMENSION output_col;
+  jpeg_component_info *compptr;
+  inverse_DCT_method_ptr inverse_DCT;
+  boolean first_row, last_row;
+  JBLOCK workspace;
+  int *coef_bits;
+  JQUANT_TBL *quanttbl;
+  INT32 Q00,Q01,Q02,Q10,Q11,Q20, num;
+  int DC1,DC2,DC3,DC4,DC5,DC6,DC7,DC8,DC9;
+  int Al, pred;
+
+  /* Force some input to be done if we are getting ahead of the input. */
+  while (cinfo->input_scan_number <= cinfo->output_scan_number &&
+	 ! cinfo->inputctl->eoi_reached) {
+    if (cinfo->input_scan_number == cinfo->output_scan_number) {
+      /* If input is working on current scan, we ordinarily want it to
+       * have completed the current row.  But if input scan is DC,
+       * we want it to keep one row ahead so that next block row's DC
+       * values are up to date.
+       */
+      JDIMENSION delta = (cinfo->Ss == 0) ? 1 : 0;
+      if (cinfo->input_iMCU_row > cinfo->output_iMCU_row+delta)
+	break;
+    }
+    if ((*cinfo->inputctl->consume_input)(cinfo) == JPEG_SUSPENDED)
+      return JPEG_SUSPENDED;
+  }
+
+  /* OK, output from the virtual arrays. */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Don't bother to IDCT an uninteresting component. */
+    if (! compptr->component_needed)
+      continue;
+    /* Count non-dummy DCT block rows in this iMCU row. */
+    if (cinfo->output_iMCU_row < last_iMCU_row) {
+      block_rows = compptr->v_samp_factor;
+      access_rows = block_rows * 2; /* this and next iMCU row */
+      last_row = FALSE;
+    } else {
+      /* NB: can't use last_row_height here; it is input-side-dependent! */
+      block_rows = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
+      if (block_rows == 0) block_rows = compptr->v_samp_factor;
+      access_rows = block_rows; /* this iMCU row only */
+      last_row = TRUE;
+    }
+    /* Align the virtual buffer for this component. */
+    if (cinfo->output_iMCU_row > 0) {
+      access_rows += compptr->v_samp_factor; /* prior iMCU row too */
+      buffer = (*cinfo->mem->access_virt_barray)
+	((j_common_ptr) cinfo, coef->whole_image[ci],
+	 (cinfo->output_iMCU_row - 1) * compptr->v_samp_factor,
+	 (JDIMENSION) access_rows, FALSE);
+      buffer += compptr->v_samp_factor;	/* point to current iMCU row */
+      first_row = FALSE;
+    } else {
+      buffer = (*cinfo->mem->access_virt_barray)
+	((j_common_ptr) cinfo, coef->whole_image[ci],
+	 (JDIMENSION) 0, (JDIMENSION) access_rows, FALSE);
+      first_row = TRUE;
+    }
+    /* Fetch component-dependent info */
+    coef_bits = coef->coef_bits_latch + (ci * SAVED_COEFS);
+    quanttbl = compptr->quant_table;
+    Q00 = quanttbl->quantval[0];
+    Q01 = quanttbl->quantval[Q01_POS];
+    Q10 = quanttbl->quantval[Q10_POS];
+    Q20 = quanttbl->quantval[Q20_POS];
+    Q11 = quanttbl->quantval[Q11_POS];
+    Q02 = quanttbl->quantval[Q02_POS];
+    inverse_DCT = cinfo->idct->inverse_DCT[ci];
+    output_ptr = output_buf[ci];
+    /* Loop over all DCT blocks to be processed. */
+    for (block_row = 0; block_row < block_rows; block_row++) {
+      buffer_ptr = buffer[block_row];
+      if (first_row && block_row == 0)
+	prev_block_row = buffer_ptr;
+      else
+	prev_block_row = buffer[block_row-1];
+      if (last_row && block_row == block_rows-1)
+	next_block_row = buffer_ptr;
+      else
+	next_block_row = buffer[block_row+1];
+      /* We fetch the surrounding DC values using a sliding-register approach.
+       * Initialize all nine here so as to do the right thing on narrow pics.
+       */
+      DC1 = DC2 = DC3 = (int) prev_block_row[0][0];
+      DC4 = DC5 = DC6 = (int) buffer_ptr[0][0];
+      DC7 = DC8 = DC9 = (int) next_block_row[0][0];
+      output_col = 0;
+      last_block_column = compptr->width_in_blocks - 1;
+      for (block_num = 0; block_num <= last_block_column; block_num++) {
+	/* Fetch current DCT block into workspace so we can modify it. */
+	jcopy_block_row(buffer_ptr, (JBLOCKROW) workspace, (JDIMENSION) 1);
+	/* Update DC values */
+	if (block_num < last_block_column) {
+	  DC3 = (int) prev_block_row[1][0];
+	  DC6 = (int) buffer_ptr[1][0];
+	  DC9 = (int) next_block_row[1][0];
+	}
+	/* Compute coefficient estimates per K.8.
+	 * An estimate is applied only if coefficient is still zero,
+	 * and is not known to be fully accurate.
+	 */
+	/* AC01 */
+	if ((Al=coef_bits[1]) != 0 && workspace[1] == 0) {
+	  num = 36 * Q00 * (DC4 - DC6);
+	  if (num >= 0) {
+	    pred = (int) (((Q01<<7) + num) / (Q01<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	  } else {
+	    pred = (int) (((Q01<<7) - num) / (Q01<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	    pred = -pred;
+	  }
+	  workspace[1] = (JCOEF) pred;
+	}
+	/* AC10 */
+	if ((Al=coef_bits[2]) != 0 && workspace[8] == 0) {
+	  num = 36 * Q00 * (DC2 - DC8);
+	  if (num >= 0) {
+	    pred = (int) (((Q10<<7) + num) / (Q10<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	  } else {
+	    pred = (int) (((Q10<<7) - num) / (Q10<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	    pred = -pred;
+	  }
+	  workspace[8] = (JCOEF) pred;
+	}
+	/* AC20 */
+	if ((Al=coef_bits[3]) != 0 && workspace[16] == 0) {
+	  num = 9 * Q00 * (DC2 + DC8 - 2*DC5);
+	  if (num >= 0) {
+	    pred = (int) (((Q20<<7) + num) / (Q20<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	  } else {
+	    pred = (int) (((Q20<<7) - num) / (Q20<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	    pred = -pred;
+	  }
+	  workspace[16] = (JCOEF) pred;
+	}
+	/* AC11 */
+	if ((Al=coef_bits[4]) != 0 && workspace[9] == 0) {
+	  num = 5 * Q00 * (DC1 - DC3 - DC7 + DC9);
+	  if (num >= 0) {
+	    pred = (int) (((Q11<<7) + num) / (Q11<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	  } else {
+	    pred = (int) (((Q11<<7) - num) / (Q11<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	    pred = -pred;
+	  }
+	  workspace[9] = (JCOEF) pred;
+	}
+	/* AC02 */
+	if ((Al=coef_bits[5]) != 0 && workspace[2] == 0) {
+	  num = 9 * Q00 * (DC4 + DC6 - 2*DC5);
+	  if (num >= 0) {
+	    pred = (int) (((Q02<<7) + num) / (Q02<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	  } else {
+	    pred = (int) (((Q02<<7) - num) / (Q02<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	    pred = -pred;
+	  }
+	  workspace[2] = (JCOEF) pred;
+	}
+	/* OK, do the IDCT */
+	(*inverse_DCT) (cinfo, compptr, (JCOEFPTR) workspace,
+			output_ptr, output_col);
+	/* Advance for next column */
+	DC1 = DC2; DC2 = DC3;
+	DC4 = DC5; DC5 = DC6;
+	DC7 = DC8; DC8 = DC9;
+	buffer_ptr++, prev_block_row++, next_block_row++;
+	output_col += compptr->DCT_scaled_size;
+      }
+      output_ptr += compptr->DCT_scaled_size;
+    }
+  }
+
+  if (++(cinfo->output_iMCU_row) < cinfo->total_iMCU_rows)
+    return JPEG_ROW_COMPLETED;
+  return JPEG_SCAN_COMPLETED;
+}
+
+#endif /* BLOCK_SMOOTHING_SUPPORTED */
+
+
+/*
+ * Initialize coefficient buffer controller.
+ */
+
+GLOBAL(void)
+jinit_d_coef_controller (j_decompress_ptr cinfo, boolean need_full_buffer)
+{
+  my_coef_ptr coef;
+
+  coef = (my_coef_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_coef_controller));
+  cinfo->coef = (struct jpeg_d_coef_controller *) coef;
+  coef->pub.start_input_pass = start_input_pass;
+  coef->pub.start_output_pass = start_output_pass;
+  coef->pub.column_left_boundary = 0;
+  coef->pub.column_right_boundary = 0;
+  coef->pub.MCU_columns_to_skip = 0;
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+  coef->coef_bits_latch = NULL;
+#endif
+
+#ifdef ANDROID_TILE_BASED_DECODE
+  if (cinfo->tile_decode) {
+    if (cinfo->progressive_mode) {
+      /* Allocate one iMCU row virtual array, coef->whole_image[ci],
+       * for each color component, padded to a multiple of h_samp_factor
+       * DCT blocks in the horizontal direction.
+       */
+      int ci, access_rows;
+      jpeg_component_info *compptr;
+
+      for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	   ci++, compptr++) {
+        access_rows = compptr->v_samp_factor;
+        coef->whole_image[ci] = (*cinfo->mem->request_virt_barray)
+	  ((j_common_ptr) cinfo, JPOOL_IMAGE, TRUE,
+	   (JDIMENSION) jround_up((long) compptr->width_in_blocks,
+				(long) compptr->h_samp_factor),
+	   (JDIMENSION) compptr->v_samp_factor, // one iMCU row
+	   (JDIMENSION) access_rows);
+      }
+      coef->pub.consume_data_build_huffman_index =
+            consume_data_build_huffman_index_progressive;
+      coef->pub.consume_data = consume_data_multi_scan;
+      coef->pub.coef_arrays = coef->whole_image; /* link to virtual arrays */
+      coef->pub.decompress_data = decompress_onepass;
+    } else {
+      /* We only need a single-MCU buffer. */
+      JBLOCKROW buffer;
+      int i;
+
+      buffer = (JBLOCKROW)
+      (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  D_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+      for (i = 0; i < D_MAX_BLOCKS_IN_MCU; i++) {
+        coef->MCU_buffer[i] = buffer + i;
+      }
+      coef->pub.consume_data_build_huffman_index =
+            consume_data_build_huffman_index_baseline;
+      coef->pub.consume_data = dummy_consume_data;
+      coef->pub.coef_arrays = NULL; /* flag for no virtual arrays */
+      coef->pub.decompress_data = decompress_onepass;
+    }
+    return;
+  }
+#endif
+
+  /* Create the coefficient buffer. */
+  if (need_full_buffer) {
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+    /* Allocate a full-image virtual array for each component, */
+    /* padded to a multiple of samp_factor DCT blocks in each direction. */
+    /* Note we ask for a pre-zeroed array. */
+    int ci, access_rows;
+    jpeg_component_info *compptr;
+
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      access_rows = compptr->v_samp_factor;
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+      /* If block smoothing could be used, need a bigger window */
+      if (cinfo->progressive_mode)
+	access_rows *= 3;
+#endif
+      coef->whole_image[ci] = (*cinfo->mem->request_virt_barray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE, TRUE,
+	 (JDIMENSION) jround_up((long) compptr->width_in_blocks,
+				(long) compptr->h_samp_factor),
+	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
+				(long) compptr->v_samp_factor),
+	 (JDIMENSION) access_rows);
+    }
+    coef->pub.consume_data = consume_data;
+    coef->pub.decompress_data = decompress_data;
+    coef->pub.coef_arrays = coef->whole_image; /* link to virtual arrays */
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    /* We only need a single-MCU buffer. */
+    JBLOCKROW buffer;
+    int i;
+
+    buffer = (JBLOCKROW)
+      (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+		  D_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+    for (i = 0; i < D_MAX_BLOCKS_IN_MCU; i++) {
+      coef->MCU_buffer[i] = buffer + i;
+    }
+    coef->pub.consume_data = dummy_consume_data;
+    coef->pub.decompress_data = decompress_onepass;
+    coef->pub.coef_arrays = NULL; /* flag for no virtual arrays */
+  }
+}
diff --git a/libjpegtwrp/jdcolor.c b/libjpegtwrp/jdcolor.c
new file mode 100644
index 000000000..9a6519e3e
--- /dev/null
+++ b/libjpegtwrp/jdcolor.c
@@ -0,0 +1,976 @@
+/*
+ * jdcolor.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Copyright (c) 2010, Code Aurora Forum. All rights reserved.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains output colorspace conversion routines.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_color_deconverter pub; /* public fields */
+
+  /* Private state for YCC->RGB conversion */
+  int * Cr_r_tab;		/* => table for Cr to R conversion */
+  int * Cb_b_tab;		/* => table for Cb to B conversion */
+  INT32 * Cr_g_tab;		/* => table for Cr to G conversion */
+  INT32 * Cb_g_tab;		/* => table for Cb to G conversion */
+} my_color_deconverter;
+
+typedef my_color_deconverter * my_cconvert_ptr;
+
+
+#ifdef ANDROID_RGB
+
+/* Declarations for ordered dithering.
+ * 
+ * We use 4x4 ordered dither array packed into 32 bits. This array is
+ * sufficent for dithering RGB_888 to RGB_565.
+ */
+
+#define DITHER_MASK         0x3
+#define DITHER_ROTATE(x)    (((x)<<24) | (((x)>>8)&0x00FFFFFF))
+static const INT32 dither_matrix[4] = {
+  0x0008020A,
+  0x0C040E06,
+  0x030B0109,
+  0x0F070D05
+};
+
+#endif
+
+
+/**************** YCbCr -> RGB conversion: most common case **************/
+
+/*
+ * YCbCr is defined per CCIR 601-1, except that Cb and Cr are
+ * normalized to the range 0..MAXJSAMPLE rather than -0.5 .. 0.5.
+ * The conversion equations to be implemented are therefore
+ *	R = Y                + 1.40200 * Cr
+ *	G = Y - 0.34414 * Cb - 0.71414 * Cr
+ *	B = Y + 1.77200 * Cb
+ * where Cb and Cr represent the incoming values less CENTERJSAMPLE.
+ * (These numbers are derived from TIFF 6.0 section 21, dated 3-June-92.)
+ *
+ * To avoid floating-point arithmetic, we represent the fractional constants
+ * as integers scaled up by 2^16 (about 4 digits precision); we have to divide
+ * the products by 2^16, with appropriate rounding, to get the correct answer.
+ * Notice that Y, being an integral input, does not contribute any fraction
+ * so it need not participate in the rounding.
+ *
+ * For even more speed, we avoid doing any multiplications in the inner loop
+ * by precalculating the constants times Cb and Cr for all possible values.
+ * For 8-bit JSAMPLEs this is very reasonable (only 256 entries per table);
+ * for 12-bit samples it is still acceptable.  It's not very reasonable for
+ * 16-bit samples, but if you want lossless storage you shouldn't be changing
+ * colorspace anyway.
+ * The Cr=>R and Cb=>B values can be rounded to integers in advance; the
+ * values for the G calculation are left scaled up, since we must add them
+ * together before rounding.
+ */
+
+#define SCALEBITS	16	/* speediest right-shift on some machines */
+#define ONE_HALF	((INT32) 1 << (SCALEBITS-1))
+#define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
+
+
+/*
+ * Initialize tables for YCC->RGB colorspace conversion.
+ */
+
+LOCAL(void)
+build_ycc_rgb_table (j_decompress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  int i;
+  INT32 x;
+  SHIFT_TEMPS
+
+  cconvert->Cr_r_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                (MAXJSAMPLE+1) * SIZEOF(int));
+  cconvert->Cb_b_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                (MAXJSAMPLE+1) * SIZEOF(int));
+  cconvert->Cr_g_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                (MAXJSAMPLE+1) * SIZEOF(INT32));
+  cconvert->Cb_g_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                (MAXJSAMPLE+1) * SIZEOF(INT32));
+
+  for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) {
+    /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */
+    /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */
+    /* Cr=>R value is nearest int to 1.40200 * x */
+    cconvert->Cr_r_tab[i] = (int)
+                    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
+    /* Cb=>B value is nearest int to 1.77200 * x */
+    cconvert->Cb_b_tab[i] = (int)
+                    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
+    /* Cr=>G value is scaled-up -0.71414 * x */
+    cconvert->Cr_g_tab[i] = (- FIX(0.71414)) * x;
+    /* Cb=>G value is scaled-up -0.34414 * x */
+    /* We also add in ONE_HALF so that need not do it in inner loop */
+    cconvert->Cb_g_tab[i] = (- FIX(0.34414)) * x + ONE_HALF;
+  }
+#ifdef NEEDS_ARM_ERRATA_754319_754320
+  asm volatile ( "vmov s0,s0\n" );
+#endif
+}
+
+/*
+ * Convert some rows of samples to the output colorspace.
+ *
+ * Note that we change from noninterleaved, one-plane-per-component format
+ * to interleaved-pixel format.  The output buffer is therefore three times
+ * as wide as the input buffer.
+ * A starting row offset is provided only for the input buffer.  The caller
+ * can easily adjust the passed output_buf value to accommodate any row
+ * offset required on that side.
+ */
+
+METHODDEF(void)
+ycc_rgb_convert (j_decompress_ptr cinfo,
+		 JSAMPIMAGE input_buf, JDIMENSION input_row,
+		 JSAMPARRAY output_buf, int num_rows)
+#ifdef ANDROID_JPEG_USE_VENUM
+/*
+ * Converts YCC->RGB888 using VeNum instructions.
+ */
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JSAMPROW outptr;
+  JDIMENSION row;
+
+  for (row = 0; row < (JDIMENSION)num_rows; row++)
+  {
+    inptr0     = input_buf[0][input_row];
+    inptr1     = input_buf[1][input_row];
+    inptr2     = input_buf[2][input_row];
+
+    input_row++;
+    outptr = *output_buf++;
+
+    yvup2bgr888_venum((UINT8*) inptr0,
+                      (UINT8*) inptr2,
+                      (UINT8*) inptr1,
+                      (UINT8*) outptr,
+                      cinfo->output_width);
+  }
+}
+#else
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register INT32 * Crgtab = cconvert->Cr_g_tab;
+  register INT32 * Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      y  = GETJSAMPLE(inptr0[col]);
+      cb = GETJSAMPLE(inptr1[col]);
+      cr = GETJSAMPLE(inptr2[col]);
+      /* Range-limiting is essential due to noise introduced by DCT losses. */
+      outptr[RGB_RED] =   range_limit[y + Crrtab[cr]];
+      outptr[RGB_GREEN] = range_limit[y +
+                              ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                 SCALEBITS))];
+      outptr[RGB_BLUE] =  range_limit[y + Cbbtab[cb]];
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
+#endif /* ANDROID_JPEG_USE_VENUM */
+
+#ifdef ANDROID_RGB
+METHODDEF(void)
+ycc_rgba_8888_convert (j_decompress_ptr cinfo,
+         JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows)
+#ifdef ANDROID_JPEG_USE_VENUM
+/*
+ * Converts YCC->RGBA8888 using VeNum instructions.
+ */
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JSAMPROW outptr;
+  JDIMENSION row;
+
+  for (row = 0; row < (JDIMENSION)num_rows; row++)
+  {
+    inptr0     = input_buf[0][input_row];
+    inptr1     = input_buf[1][input_row];
+    inptr2     = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+
+    yvup2abgr8888_venum((UINT8*) inptr0,
+                        (UINT8*) inptr2,
+                        (UINT8*) inptr1,
+                        (UINT8*) outptr,
+                        cinfo->output_width);
+  }
+}
+
+#else
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register INT32 * Crgtab = cconvert->Cr_g_tab;
+  register INT32 * Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      y  = GETJSAMPLE(inptr0[col]);
+      cb = GETJSAMPLE(inptr1[col]);
+      cr = GETJSAMPLE(inptr2[col]);
+      /* Range-limiting is essential due to noise introduced by DCT losses. */
+      outptr[RGB_RED] =   range_limit[y + Crrtab[cr]];
+      outptr[RGB_GREEN] = range_limit[y +
+                              ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                 SCALEBITS))];
+      outptr[RGB_BLUE] =  range_limit[y + Cbbtab[cb]];
+      outptr[RGB_ALPHA] =  0xFF;
+      outptr += 4;
+    }
+  }
+}
+#endif /* ANDROID_JPEG_USE_VENUM */
+
+METHODDEF(void)
+ycc_rgb_565_convert (j_decompress_ptr cinfo,
+         JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows)
+#if defined(ANDROID_JPEG_USE_VENUM)
+/*
+ * Converts YCC->RGB565 using VeNum instructions.
+ */
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JSAMPROW outptr;
+  JDIMENSION row;
+
+  for (row = 0; row < (JDIMENSION)num_rows; row++)
+  {
+    inptr0     = input_buf[0][input_row];
+    inptr1     = input_buf[1][input_row];
+    inptr2     = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+
+    yvup2rgb565_venum((UINT8*) inptr0,
+                      (UINT8*) inptr2,
+                      (UINT8*) inptr1,
+                      (UINT8*) outptr,
+                      cinfo->output_width);
+  }
+}
+#else
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register INT32 * Crgtab = cconvert->Cr_g_tab;
+  register INT32 * Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    INT32 rgb;
+    unsigned int r, g, b;
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+        y  = GETJSAMPLE(*inptr0++);
+        cb = GETJSAMPLE(*inptr1++);
+        cr = GETJSAMPLE(*inptr2++);
+        r = range_limit[y + Crrtab[cr]];
+        g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb]+Crgtab[cr], SCALEBITS))];
+        b = range_limit[y + Cbbtab[cb]];
+        rgb = PACK_SHORT_565(r,g,b);
+        *(INT16*)outptr = rgb;
+        outptr += 2;
+        num_cols--;
+    }
+    for (col = 0; col < (num_cols>>1); col++) {
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[y + Crrtab[cr]];
+      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb]+Crgtab[cr], SCALEBITS))];
+      b = range_limit[y + Cbbtab[cb]];
+      rgb = PACK_SHORT_565(r,g,b);
+
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[y + Crrtab[cr]];
+      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb]+Crgtab[cr], SCALEBITS))];
+      b = range_limit[y + Cbbtab[cb]];
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r,g,b));
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols&1) {
+      y  = GETJSAMPLE(*inptr0);
+      cb = GETJSAMPLE(*inptr1);
+      cr = GETJSAMPLE(*inptr2);
+      r = range_limit[y + Crrtab[cr]];
+      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb]+Crgtab[cr], SCALEBITS))];
+      b = range_limit[y + Cbbtab[cb]];
+      rgb = PACK_SHORT_565(r,g,b);
+      *(INT16*)outptr = rgb;
+    }
+  }
+}
+#endif /* ANDROID_JPEG_USE_VENUM */
+
+METHODDEF(void)
+ycc_rgb_565D_convert (j_decompress_ptr cinfo,
+         JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register INT32 * Crgtab = cconvert->Cr_g_tab;
+  register INT32 * Cbgtab = cconvert->Cb_g_tab;
+  INT32 d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    INT32 rgb;
+    unsigned int r, g, b;
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+        y  = GETJSAMPLE(*inptr0++);
+        cb = GETJSAMPLE(*inptr1++);
+        cr = GETJSAMPLE(*inptr2++);
+        r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
+        g = range_limit[DITHER_565_G(y + ((int)RIGHT_SHIFT(Cbgtab[cb]+Crgtab[cr], SCALEBITS)), d0)];
+        b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)];
+        rgb = PACK_SHORT_565(r,g,b);
+        *(INT16*)outptr = rgb;
+        outptr += 2;
+        num_cols--;
+    }
+    for (col = 0; col < (num_cols>>1); col++) {
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
+      g = range_limit[DITHER_565_G(y + ((int)RIGHT_SHIFT(Cbgtab[cb]+Crgtab[cr], SCALEBITS)), d0)];
+      b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)];
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_SHORT_565(r,g,b);
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
+      g = range_limit[DITHER_565_G(y + ((int)RIGHT_SHIFT(Cbgtab[cb]+Crgtab[cr], SCALEBITS)), d0)];
+      b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)];
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r,g,b));
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols&1) {
+      y  = GETJSAMPLE(*inptr0);
+      cb = GETJSAMPLE(*inptr1);
+      cr = GETJSAMPLE(*inptr2);
+      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
+      g = range_limit[DITHER_565_G(y + ((int)RIGHT_SHIFT(Cbgtab[cb]+Crgtab[cr], SCALEBITS)), d0)];
+      b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)];
+      rgb = PACK_SHORT_565(r,g,b);
+      *(INT16*)outptr = rgb;
+    }
+  }
+}
+
+#endif
+
+/**************** Cases other than YCbCr -> RGB(A) **************/
+
+#ifdef ANDROID_RGB
+METHODDEF(void)
+rgb_rgba_8888_convert (j_decompress_ptr cinfo,
+         JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      *outptr++ = *inptr0++;
+      *outptr++ = *inptr1++;
+      *outptr++ = *inptr2++;
+      *outptr++ = 0xFF;
+    }
+  }
+}
+
+METHODDEF(void)
+rgb_rgb_565_convert (j_decompress_ptr cinfo,
+         JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    INT32 rgb;
+    unsigned int r, g, b;
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+        r = GETJSAMPLE(*inptr0++);
+        g = GETJSAMPLE(*inptr1++);
+        b = GETJSAMPLE(*inptr2++);
+        rgb = PACK_SHORT_565(r,g,b);
+        *(INT16*)outptr = rgb;
+        outptr += 2;
+        num_cols--;
+    }
+    for (col = 0; col < (num_cols>>1); col++) {
+      r = GETJSAMPLE(*inptr0++);
+      g = GETJSAMPLE(*inptr1++);
+      b = GETJSAMPLE(*inptr2++);
+      rgb = PACK_SHORT_565(r,g,b);
+      r = GETJSAMPLE(*inptr0++);
+      g = GETJSAMPLE(*inptr1++);
+      b = GETJSAMPLE(*inptr2++);
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r,g,b));
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols&1) {
+      r = GETJSAMPLE(*inptr0);
+      g = GETJSAMPLE(*inptr1);
+      b = GETJSAMPLE(*inptr2);
+      rgb = PACK_SHORT_565(r,g,b);
+      *(INT16*)outptr = rgb;
+    }
+  }
+}
+
+
+METHODDEF(void)
+rgb_rgb_565D_convert (j_decompress_ptr cinfo,
+         JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  JDIMENSION num_cols = cinfo->output_width;
+  INT32 d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    INT32 rgb;
+    unsigned int r, g, b;
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+        r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
+        g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
+        b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
+        rgb = PACK_SHORT_565(r,g,b);
+        *(INT16*)outptr = rgb;
+        outptr += 2;
+        num_cols--;
+    }
+    for (col = 0; col < (num_cols>>1); col++) {
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_SHORT_565(r,g,b);
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r,g,b));
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols&1) {
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2), d0)];
+      rgb = PACK_SHORT_565(r,g,b);
+      *(INT16*)outptr = rgb;
+    }
+  }
+}
+
+#endif
+
+/*
+ * Color conversion for no colorspace change: just copy the data,
+ * converting from separate-planes to interleaved representation.
+ */
+
+METHODDEF(void)
+null_convert (j_decompress_ptr cinfo,
+	      JSAMPIMAGE input_buf, JDIMENSION input_row,
+	      JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION count;
+  register int num_components = cinfo->num_components;
+  JDIMENSION num_cols = cinfo->output_width;
+  int ci;
+
+  while (--num_rows >= 0) {
+    for (ci = 0; ci < num_components; ci++) {
+      inptr = input_buf[ci][input_row];
+      outptr = output_buf[0] + ci;
+      for (count = num_cols; count > 0; count--) {
+	*outptr = *inptr++;	/* needn't bother with GETJSAMPLE() here */
+	outptr += num_components;
+      }
+    }
+    input_row++;
+    output_buf++;
+  }
+}
+
+
+/*
+ * Color conversion for grayscale: just copy the data.
+ * This also works for YCbCr -> grayscale conversion, in which
+ * we just copy the Y (luminance) component and ignore chrominance.
+ */
+
+METHODDEF(void)
+grayscale_convert (j_decompress_ptr cinfo,
+		   JSAMPIMAGE input_buf, JDIMENSION input_row,
+		   JSAMPARRAY output_buf, int num_rows)
+{
+  jcopy_sample_rows(input_buf[0], (int) input_row, output_buf, 0,
+		    num_rows, cinfo->output_width);
+}
+
+
+/*
+ * Convert grayscale to RGB: just duplicate the graylevel three times.
+ * This is provided to support applications that don't want to cope
+ * with grayscale as a separate case.
+ */
+
+METHODDEF(void)
+gray_rgb_convert (j_decompress_ptr cinfo,
+		  JSAMPIMAGE input_buf, JDIMENSION input_row,
+		  JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr = input_buf[0][input_row++];
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      /* We can dispense with GETJSAMPLE() here */
+      outptr[RGB_RED] = outptr[RGB_GREEN] = outptr[RGB_BLUE] = inptr[col];
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
+
+#ifdef ANDROID_RGB
+METHODDEF(void)
+gray_rgba_8888_convert (j_decompress_ptr cinfo,
+          JSAMPIMAGE input_buf, JDIMENSION input_row,
+          JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr = input_buf[0][input_row++];
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      /* We can dispense with GETJSAMPLE() here */
+      outptr[RGB_RED] = outptr[RGB_GREEN] = outptr[RGB_BLUE] = inptr[col];
+      outptr[RGB_ALPHA] = 0xff;
+      outptr += 4;
+    }
+  }
+}
+
+METHODDEF(void)
+gray_rgb_565_convert (j_decompress_ptr cinfo,
+          JSAMPIMAGE input_buf, JDIMENSION input_row,
+          JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    INT32 rgb;
+    unsigned int g;
+    inptr = input_buf[0][input_row++];
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+        g = *inptr++;
+        rgb = PACK_SHORT_565(g, g, g);
+        *(INT16*)outptr = rgb;
+        outptr += 2;
+        num_cols--;
+    }
+    for (col = 0; col < (num_cols>>1); col++) {
+      g = *inptr++;
+      rgb = PACK_SHORT_565(g, g, g);
+      g = *inptr++;
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(g, g, g));
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols&1) {
+      g = *inptr;
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16*)outptr = rgb;
+    }
+  }
+}
+
+METHODDEF(void)
+gray_rgb_565D_convert (j_decompress_ptr cinfo,
+          JSAMPIMAGE input_buf, JDIMENSION input_row,
+          JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  JDIMENSION num_cols = cinfo->output_width;
+  INT32 d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+
+  while (--num_rows >= 0) {
+    INT32 rgb;
+    unsigned int g;
+    inptr = input_buf[0][input_row++];
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+        g = *inptr++;
+        g = range_limit[DITHER_565_R(g, d0)];
+        rgb = PACK_SHORT_565(g, g, g);
+        *(INT16*)outptr = rgb;
+        outptr += 2;
+        num_cols--;
+    }
+    for (col = 0; col < (num_cols>>1); col++) {
+      g = *inptr++;
+      g = range_limit[DITHER_565_R(g, d0)];
+      rgb = PACK_SHORT_565(g, g, g);
+      d0 = DITHER_ROTATE(d0);
+      g = *inptr++;
+      g = range_limit[DITHER_565_R(g, d0)];
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(g, g, g));
+      d0 = DITHER_ROTATE(d0);
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols&1) {
+      g = *inptr;
+      g = range_limit[DITHER_565_R(g, d0)];
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16*)outptr = rgb;
+    }
+  }
+}
+#endif
+
+/*
+ * Adobe-style YCCK->CMYK conversion.
+ * We convert YCbCr to R=1-C, G=1-M, and B=1-Y using the same
+ * conversion as above, while passing K (black) unchanged.
+ * We assume build_ycc_rgb_table has been called.
+ */
+
+METHODDEF(void)
+ycck_cmyk_convert (j_decompress_ptr cinfo,
+		   JSAMPIMAGE input_buf, JDIMENSION input_row,
+		   JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2, inptr3;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register INT32 * Crgtab = cconvert->Cr_g_tab;
+  register INT32 * Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    inptr3 = input_buf[3][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      y  = GETJSAMPLE(inptr0[col]);
+      cb = GETJSAMPLE(inptr1[col]);
+      cr = GETJSAMPLE(inptr2[col]);
+      /* Range-limiting is essential due to noise introduced by DCT losses. */
+      outptr[0] = range_limit[MAXJSAMPLE - (y + Crrtab[cr])];   /* red */
+      outptr[1] = range_limit[MAXJSAMPLE - (y +                 /* green */
+                              ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                 SCALEBITS)))];
+      outptr[2] = range_limit[MAXJSAMPLE - (y + Cbbtab[cb])];   /* blue */
+      /* K passes through unchanged */
+      outptr[3] = inptr3[col];	/* don't need GETJSAMPLE here */
+      outptr += 4;
+    }
+  }
+}
+
+
+/*
+ * Empty method for start_pass.
+ */
+
+METHODDEF(void)
+start_pass_dcolor (j_decompress_ptr cinfo)
+{
+  /* no work needed */
+}
+
+
+/*
+ * Module initialization routine for output colorspace conversion.
+ */
+
+GLOBAL(void)
+jinit_color_deconverter (j_decompress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert;
+  int ci;
+
+  cconvert = (my_cconvert_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_color_deconverter));
+  cinfo->cconvert = (struct jpeg_color_deconverter *) cconvert;
+  cconvert->pub.start_pass = start_pass_dcolor;
+
+  /* Make sure num_components agrees with jpeg_color_space */
+  switch (cinfo->jpeg_color_space) {
+  case JCS_GRAYSCALE:
+    if (cinfo->num_components != 1)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
+
+  case JCS_RGB:
+  case JCS_YCbCr:
+    if (cinfo->num_components != 3)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
+
+  case JCS_CMYK:
+  case JCS_YCCK:
+    if (cinfo->num_components != 4)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
+
+  default:			/* JCS_UNKNOWN can be anything */
+    if (cinfo->num_components < 1)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
+  }
+
+  /* Set out_color_components and conversion method based on requested space.
+   * Also clear the component_needed flags for any unused components,
+   * so that earlier pipeline stages can avoid useless computation.
+   */
+
+  switch (cinfo->out_color_space) {
+  case JCS_GRAYSCALE:
+    cinfo->out_color_components = 1;
+    if (cinfo->jpeg_color_space == JCS_GRAYSCALE ||
+	cinfo->jpeg_color_space == JCS_YCbCr) {
+      cconvert->pub.color_convert = grayscale_convert;
+      /* For color->grayscale conversion, only the Y (0) component is needed */
+      for (ci = 1; ci < cinfo->num_components; ci++)
+	cinfo->comp_info[ci].component_needed = FALSE;
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_RGB:
+    cinfo->out_color_components = RGB_PIXELSIZE;
+    if (cinfo->jpeg_color_space == JCS_YCbCr) {
+      cconvert->pub.color_convert = ycc_rgb_convert;
+#ifndef ANDROID_JPEG_USE_VENUM
+      build_ycc_rgb_table(cinfo);
+#endif
+    } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
+      cconvert->pub.color_convert = gray_rgb_convert;
+    } else if (cinfo->jpeg_color_space == JCS_RGB && RGB_PIXELSIZE == 3) {
+      cconvert->pub.color_convert = null_convert;
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+#ifdef ANDROID_RGB
+  case JCS_RGBA_8888:
+    cinfo->out_color_components = 4;
+    if (cinfo->jpeg_color_space == JCS_YCbCr) {
+      cconvert->pub.color_convert = ycc_rgba_8888_convert;
+      build_ycc_rgb_table(cinfo);
+    } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
+      cconvert->pub.color_convert = gray_rgba_8888_convert;
+    } else if (cinfo->jpeg_color_space == JCS_RGB) {
+      cconvert->pub.color_convert = rgb_rgba_8888_convert;
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_RGB_565:
+    cinfo->out_color_components = RGB_PIXELSIZE;
+    if (cinfo->dither_mode == JDITHER_NONE) {
+      if (cinfo->jpeg_color_space == JCS_YCbCr) {
+        cconvert->pub.color_convert = ycc_rgb_565_convert;
+#if !defined(ANDROID_JPEG_USE_VENUM)
+        build_ycc_rgb_table(cinfo);
+#endif /* ANDROID_JPEG_USE_VENUM */
+      } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
+        cconvert->pub.color_convert = gray_rgb_565_convert;
+      } else if (cinfo->jpeg_color_space == JCS_RGB) {
+        cconvert->pub.color_convert = rgb_rgb_565_convert;
+      } else
+        ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    } else {
+      /* only ordered dither is supported */
+      if (cinfo->jpeg_color_space == JCS_YCbCr) {
+#if defined(ANDROID_JPEG_USE_VENUM)
+        /* Use VeNum routine even if dithering option is selected. */
+        cconvert->pub.color_convert = ycc_rgb_565_convert;
+#else
+        cconvert->pub.color_convert = ycc_rgb_565D_convert;
+        build_ycc_rgb_table(cinfo);
+#endif /* ANDROID_JPEG_USE_VENUM */
+      } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
+        cconvert->pub.color_convert = gray_rgb_565D_convert;
+      } else if (cinfo->jpeg_color_space == JCS_RGB) {
+        cconvert->pub.color_convert = rgb_rgb_565D_convert;
+      } else
+        ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    }
+    break;
+#endif
+    
+  case JCS_CMYK:
+    cinfo->out_color_components = 4;
+    if (cinfo->jpeg_color_space == JCS_YCCK) {
+      cconvert->pub.color_convert = ycck_cmyk_convert;
+      build_ycc_rgb_table(cinfo);
+    } else if (cinfo->jpeg_color_space == JCS_CMYK) {
+      cconvert->pub.color_convert = null_convert;
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  default:
+    /* Permit null conversion to same output space */
+    if (cinfo->out_color_space == cinfo->jpeg_color_space) {
+      cinfo->out_color_components = cinfo->num_components;
+      cconvert->pub.color_convert = null_convert;
+    } else			/* unsupported non-null conversion */
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+  }
+
+  if (cinfo->quantize_colors)
+    cinfo->output_components = 1; /* single colormapped output component */
+  else
+    cinfo->output_components = cinfo->out_color_components;
+}
diff --git a/libjpegtwrp/jdct.h b/libjpegtwrp/jdct.h
new file mode 100644
index 000000000..04192a266
--- /dev/null
+++ b/libjpegtwrp/jdct.h
@@ -0,0 +1,176 @@
+/*
+ * jdct.h
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This include file contains common declarations for the forward and
+ * inverse DCT modules.  These declarations are private to the DCT managers
+ * (jcdctmgr.c, jddctmgr.c) and the individual DCT algorithms.
+ * The individual DCT algorithms are kept in separate files to ease 
+ * machine-dependent tuning (e.g., assembly coding).
+ */
+
+
+/*
+ * A forward DCT routine is given a pointer to a work area of type DCTELEM[];
+ * the DCT is to be performed in-place in that buffer.  Type DCTELEM is int
+ * for 8-bit samples, INT32 for 12-bit samples.  (NOTE: Floating-point DCT
+ * implementations use an array of type FAST_FLOAT, instead.)
+ * The DCT inputs are expected to be signed (range +-CENTERJSAMPLE).
+ * The DCT outputs are returned scaled up by a factor of 8; they therefore
+ * have a range of +-8K for 8-bit data, +-128K for 12-bit data.  This
+ * convention improves accuracy in integer implementations and saves some
+ * work in floating-point ones.
+ * Quantization of the output coefficients is done by jcdctmgr.c.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+typedef int DCTELEM;		/* 16 or 32 bits is fine */
+#else
+typedef INT32 DCTELEM;		/* must have 32 bits */
+#endif
+
+typedef JMETHOD(void, forward_DCT_method_ptr, (DCTELEM * data));
+typedef JMETHOD(void, float_DCT_method_ptr, (FAST_FLOAT * data));
+
+
+/*
+ * An inverse DCT routine is given a pointer to the input JBLOCK and a pointer
+ * to an output sample array.  The routine must dequantize the input data as
+ * well as perform the IDCT; for dequantization, it uses the multiplier table
+ * pointed to by compptr->dct_table.  The output data is to be placed into the
+ * sample array starting at a specified column.  (Any row offset needed will
+ * be applied to the array pointer before it is passed to the IDCT code.)
+ * Note that the number of samples emitted by the IDCT routine is
+ * DCT_scaled_size * DCT_scaled_size.
+ */
+
+/* typedef inverse_DCT_method_ptr is declared in jpegint.h */
+
+/*
+ * Each IDCT routine has its own ideas about the best dct_table element type.
+ */
+
+typedef MULTIPLIER ISLOW_MULT_TYPE; /* short or int, whichever is faster */
+#if BITS_IN_JSAMPLE == 8
+typedef MULTIPLIER IFAST_MULT_TYPE; /* 16 bits is OK, use short if faster */
+#define IFAST_SCALE_BITS  2	/* fractional bits in scale factors */
+#else
+typedef INT32 IFAST_MULT_TYPE;	/* need 32 bits for scaled quantizers */
+#define IFAST_SCALE_BITS  13	/* fractional bits in scale factors */
+#endif
+typedef FAST_FLOAT FLOAT_MULT_TYPE; /* preferred floating type */
+
+
+/*
+ * Each IDCT routine is responsible for range-limiting its results and
+ * converting them to unsigned form (0..MAXJSAMPLE).  The raw outputs could
+ * be quite far out of range if the input data is corrupt, so a bulletproof
+ * range-limiting step is required.  We use a mask-and-table-lookup method
+ * to do the combined operations quickly.  See the comments with
+ * prepare_range_limit_table (in jdmaster.c) for more info.
+ */
+
+#define IDCT_range_limit(cinfo)  ((cinfo)->sample_range_limit + CENTERJSAMPLE)
+
+#define RANGE_MASK  (MAXJSAMPLE * 4 + 3) /* 2 bits wider than legal samples */
+
+
+/* Short forms of external names for systems with brain-damaged linkers. */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jpeg_fdct_islow		jFDislow
+#define jpeg_fdct_ifast		jFDifast
+#define jpeg_fdct_float		jFDfloat
+#define jpeg_idct_islow		jRDislow
+#define jpeg_idct_ifast		jRDifast
+#define jpeg_idct_float		jRDfloat
+#define jpeg_idct_4x4		jRD4x4
+#define jpeg_idct_2x2		jRD2x2
+#define jpeg_idct_1x1		jRD1x1
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+/* Extern declarations for the forward and inverse DCT routines. */
+
+EXTERN(void) jpeg_fdct_islow JPP((DCTELEM * data));
+EXTERN(void) jpeg_fdct_ifast JPP((DCTELEM * data));
+EXTERN(void) jpeg_fdct_float JPP((FAST_FLOAT * data));
+
+EXTERN(void) jpeg_idct_islow
+    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+EXTERN(void) jpeg_idct_ifast
+    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+EXTERN(void) jpeg_idct_float
+    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+EXTERN(void) jpeg_idct_4x4
+    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+EXTERN(void) jpeg_idct_2x2
+    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+EXTERN(void) jpeg_idct_1x1
+    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+
+
+/*
+ * Macros for handling fixed-point arithmetic; these are used by many
+ * but not all of the DCT/IDCT modules.
+ *
+ * All values are expected to be of type INT32.
+ * Fractional constants are scaled left by CONST_BITS bits.
+ * CONST_BITS is defined within each module using these macros,
+ * and may differ from one module to the next.
+ */
+
+#define ONE	((INT32) 1)
+#define CONST_SCALE (ONE << CONST_BITS)
+
+/* Convert a positive real constant to an integer scaled by CONST_SCALE.
+ * Caution: some C compilers fail to reduce "FIX(constant)" at compile time,
+ * thus causing a lot of useless floating-point operations at run time.
+ */
+
+#define FIX(x)	((INT32) ((x) * CONST_SCALE + 0.5))
+
+/* Descale and correctly round an INT32 value that's scaled by N bits.
+ * We assume RIGHT_SHIFT rounds towards minus infinity, so adding
+ * the fudge factor is correct for either sign of X.
+ */
+
+#define DESCALE(x,n)  RIGHT_SHIFT((x) + (ONE << ((n)-1)), n)
+
+/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+ * This macro is used only when the two inputs will actually be no more than
+ * 16 bits wide, so that a 16x16->32 bit multiply can be used instead of a
+ * full 32x32 multiply.  This provides a useful speedup on many machines.
+ * Unfortunately there is no way to specify a 16x16->32 multiply portably
+ * in C, but some C compilers will do the right thing if you provide the
+ * correct combination of casts.
+ */
+
+#ifdef SHORTxSHORT_32		/* may work if 'int' is 32 bits */
+#define MULTIPLY16C16(var,const)  (((INT16) (var)) * ((INT16) (const)))
+#endif
+#ifdef SHORTxLCONST_32		/* known to work with Microsoft C 6.0 */
+#define MULTIPLY16C16(var,const)  (((INT16) (var)) * ((INT32) (const)))
+#endif
+
+#ifndef MULTIPLY16C16		/* default definition */
+#define MULTIPLY16C16(var,const)  ((var) * (const))
+#endif
+
+/* Same except both inputs are variables. */
+
+#ifdef SHORTxSHORT_32		/* may work if 'int' is 32 bits */
+#define MULTIPLY16V16(var1,var2)  (((INT16) (var1)) * ((INT16) (var2)))
+#endif
+
+#ifndef MULTIPLY16V16		/* default definition */
+#define MULTIPLY16V16(var1,var2)  ((var1) * (var2))
+#endif
diff --git a/libjpegtwrp/jddctmgr.c b/libjpegtwrp/jddctmgr.c
new file mode 100644
index 000000000..674bfa75f
--- /dev/null
+++ b/libjpegtwrp/jddctmgr.c
@@ -0,0 +1,335 @@
+/*
+ * jddctmgr.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Copyright (c) 2010, Code Aurora Forum. All rights reserved.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the inverse-DCT management logic.
+ * This code selects a particular IDCT implementation to be used,
+ * and it performs related housekeeping chores.  No code in this file
+ * is executed per IDCT step, only during output pass setup.
+ *
+ * Note that the IDCT routines are responsible for performing coefficient
+ * dequantization as well as the IDCT proper.  This module sets up the
+ * dequantization multiplier table needed by the IDCT routine.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef ANDROID_ARMV6_IDCT
+  #undef ANDROID_ARMV6_IDCT
+  #ifdef __arm__
+    #include <machine/cpu-features.h>
+    #if __ARM_ARCH__ >= 6
+      #define ANDROID_ARMV6_IDCT
+    #else
+      #warning "ANDROID_ARMV6_IDCT is disabled"
+    #endif
+  #endif
+#endif
+
+#ifdef ANDROID_ARMV6_IDCT
+
+/* Intentionally declare the prototype with arguments of primitive types instead
+ * of type-defined ones. This will at least generate some warnings if jmorecfg.h
+ * is changed and becomes incompatible with the assembly code.
+ */
+extern void armv6_idct(short *coefs, int *quans, unsigned char **rows, int col);
+
+void jpeg_idct_armv6 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		 JCOEFPTR coef_block,
+		 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  IFAST_MULT_TYPE *dct_table = (IFAST_MULT_TYPE *)compptr->dct_table;
+  armv6_idct(coef_block, dct_table, output_buf, output_col);
+}
+
+#endif
+
+/*
+ * The decompressor input side (jdinput.c) saves away the appropriate
+ * quantization table for each component at the start of the first scan
+ * involving that component.  (This is necessary in order to correctly
+ * decode files that reuse Q-table slots.)
+ * When we are ready to make an output pass, the saved Q-table is converted
+ * to a multiplier table that will actually be used by the IDCT routine.
+ * The multiplier table contents are IDCT-method-dependent.  To support
+ * application changes in IDCT method between scans, we can remake the
+ * multiplier tables if necessary.
+ * In buffered-image mode, the first output pass may occur before any data
+ * has been seen for some components, and thus before their Q-tables have
+ * been saved away.  To handle this case, multiplier tables are preset
+ * to zeroes; the result of the IDCT will be a neutral gray level.
+ */
+
+
+/* Private subobject for this module */
+
+typedef struct {
+  struct jpeg_inverse_dct pub;	/* public fields */
+
+  /* This array contains the IDCT method code that each multiplier table
+   * is currently set up for, or -1 if it's not yet set up.
+   * The actual multiplier tables are pointed to by dct_table in the
+   * per-component comp_info structures.
+   */
+  int cur_method[MAX_COMPONENTS];
+} my_idct_controller;
+
+typedef my_idct_controller * my_idct_ptr;
+
+
+/* Allocated multiplier tables: big enough for any supported variant */
+
+typedef union {
+  ISLOW_MULT_TYPE islow_array[DCTSIZE2];
+#ifdef DCT_IFAST_SUPPORTED
+  IFAST_MULT_TYPE ifast_array[DCTSIZE2];
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  FLOAT_MULT_TYPE float_array[DCTSIZE2];
+#endif
+} multiplier_table;
+
+
+/* The current scaled-IDCT routines require ISLOW-style multiplier tables,
+ * so be sure to compile that code if either ISLOW or SCALING is requested.
+ */
+#ifdef DCT_ISLOW_SUPPORTED
+#define PROVIDE_ISLOW_TABLES
+#else
+#ifdef IDCT_SCALING_SUPPORTED
+#define PROVIDE_ISLOW_TABLES
+#endif
+#endif
+
+
+/*
+ * Prepare for an output pass.
+ * Here we select the proper IDCT routine for each component and build
+ * a matching multiplier table.
+ */
+
+METHODDEF(void)
+start_pass (j_decompress_ptr cinfo)
+{
+  my_idct_ptr idct = (my_idct_ptr) cinfo->idct;
+  int ci, i;
+  jpeg_component_info *compptr;
+  int method = 0;
+  inverse_DCT_method_ptr method_ptr = NULL;
+  JQUANT_TBL * qtbl;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Select the proper IDCT routine for this component's scaling */
+    switch (compptr->DCT_scaled_size) {
+#ifdef IDCT_SCALING_SUPPORTED
+    case 1:
+      method_ptr = jpeg_idct_1x1;
+      method = JDCT_ISLOW;	/* jidctred uses islow-style table */
+      break;
+    case 2:
+      method_ptr = jpeg_idct_2x2;
+      method = JDCT_ISLOW;	/* jidctred uses islow-style table */
+      break;
+    case 4:
+      method_ptr = jpeg_idct_4x4;
+      method = JDCT_ISLOW;	/* jidctred uses islow-style table */
+      break;
+#endif
+    case DCTSIZE:
+      switch (cinfo->dct_method) {
+#ifdef ANDROID_ARMV6_IDCT
+      case JDCT_ISLOW:
+      case JDCT_IFAST:
+	method_ptr = jpeg_idct_armv6;
+	method = JDCT_IFAST;
+	break;
+#else /* ANDROID_ARMV6_IDCT */
+#ifdef DCT_ISLOW_SUPPORTED
+      case JDCT_ISLOW:
+	method_ptr = jpeg_idct_islow;
+	method = JDCT_ISLOW;
+	break;
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+      case JDCT_IFAST:
+#ifdef ANDROID_JPEG_USE_VENUM
+        /* Use VeNum implementation of jpeg_idct_islow even if fast DCT option is selected */
+	method_ptr = jpeg_idct_islow;
+	method = JDCT_ISLOW;
+#else
+	method_ptr = jpeg_idct_ifast;
+	method = JDCT_IFAST;
+#endif /* ANDROID_JPEG_USE_VENUM */
+	break;
+#endif
+#endif /* ANDROID_ARMV6_IDCT */
+#ifdef DCT_FLOAT_SUPPORTED
+      case JDCT_FLOAT:
+	method_ptr = jpeg_idct_float;
+	method = JDCT_FLOAT;
+	break;
+#endif
+      default:
+	ERREXIT(cinfo, JERR_NOT_COMPILED);
+	break;
+      }
+      break;
+    default:
+      ERREXIT1(cinfo, JERR_BAD_DCTSIZE, compptr->DCT_scaled_size);
+      break;
+    }
+    idct->pub.inverse_DCT[ci] = method_ptr;
+    /* Create multiplier table from quant table.
+     * However, we can skip this if the component is uninteresting
+     * or if we already built the table.  Also, if no quant table
+     * has yet been saved for the component, we leave the
+     * multiplier table all-zero; we'll be reading zeroes from the
+     * coefficient controller's buffer anyway.
+     */
+    if (! compptr->component_needed || idct->cur_method[ci] == method)
+      continue;
+    qtbl = compptr->quant_table;
+    if (qtbl == NULL)		/* happens if no data yet for component */
+      continue;
+    idct->cur_method[ci] = method;
+    switch (method) {
+#ifdef PROVIDE_ISLOW_TABLES
+    case JDCT_ISLOW:
+      {
+	/* For LL&M IDCT method, multipliers are equal to raw quantization
+	 * coefficients, but are stored as ints to ensure access efficiency.
+	 */
+	ISLOW_MULT_TYPE * ismtbl = (ISLOW_MULT_TYPE *) compptr->dct_table;
+	for (i = 0; i < DCTSIZE2; i++) {
+	  ismtbl[i] = (ISLOW_MULT_TYPE) qtbl->quantval[i];
+	}
+      }
+      break;
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+    case JDCT_IFAST:
+      {
+	/* For AA&N IDCT method, multipliers are equal to quantization
+	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
+	 *   scalefactor[0] = 1
+	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+	 * For integer operation, the multiplier table is to be scaled by
+	 * IFAST_SCALE_BITS.
+	 */
+	IFAST_MULT_TYPE * ifmtbl = (IFAST_MULT_TYPE *) compptr->dct_table;
+#ifdef ANDROID_ARMV6_IDCT
+	/* Precomputed values scaled up by 15 bits. */
+	static const unsigned short scales[DCTSIZE2] = {
+	  32768, 45451, 42813, 38531, 32768, 25746, 17734,  9041,
+	  45451, 63042, 59384, 53444, 45451, 35710, 24598, 12540,
+	  42813, 59384, 55938, 50343, 42813, 33638, 23170, 11812,
+	  38531, 53444, 50343, 45308, 38531, 30274, 20853, 10631,
+	  32768, 45451, 42813, 38531, 32768, 25746, 17734,  9041,
+	  25746, 35710, 33638, 30274, 25746, 20228, 13933,  7103,
+	  17734, 24598, 23170, 20853, 17734, 13933,  9598,  4893,
+	   9041, 12540, 11812, 10631,  9041,  7103,  4893,  2494,
+	};
+	/* Inverse map of [7, 5, 1, 3, 0, 2, 4, 6]. */
+	static const char orders[DCTSIZE] = {4, 2, 5, 3, 6, 1, 7, 0};
+	/* Reorder the columns after transposing. */
+	for (i = 0; i < DCTSIZE2; ++i) {
+	  int j = ((i & 7) << 3) + orders[i >> 3];
+	  ifmtbl[j] = (qtbl->quantval[i] * scales[i] + 2) >> 2;
+	}
+#else /* ANDROID_ARMV6_IDCT */
+
+#define CONST_BITS 14
+	static const INT16 aanscales[DCTSIZE2] = {
+	  /* precomputed values scaled up by 14 bits */
+	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+	  22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
+	  21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
+	  19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
+	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+	  12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
+	   8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
+	   4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
+	};
+	SHIFT_TEMPS
+
+	for (i = 0; i < DCTSIZE2; i++) {
+	  ifmtbl[i] = (IFAST_MULT_TYPE)
+	    DESCALE(MULTIPLY16V16((INT32) qtbl->quantval[i],
+				  (INT32) aanscales[i]),
+		    CONST_BITS-IFAST_SCALE_BITS);
+	}
+#endif /* ANDROID_ARMV6_IDCT */
+      }
+      break;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+    case JDCT_FLOAT:
+      {
+	/* For float AA&N IDCT method, multipliers are equal to quantization
+	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
+	 *   scalefactor[0] = 1
+	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+	 */
+	FLOAT_MULT_TYPE * fmtbl = (FLOAT_MULT_TYPE *) compptr->dct_table;
+	int row, col;
+	static const double aanscalefactor[DCTSIZE] = {
+	  1.0, 1.387039845, 1.306562965, 1.175875602,
+	  1.0, 0.785694958, 0.541196100, 0.275899379
+	};
+
+	i = 0;
+	for (row = 0; row < DCTSIZE; row++) {
+	  for (col = 0; col < DCTSIZE; col++) {
+	    fmtbl[i] = (FLOAT_MULT_TYPE)
+	      ((double) qtbl->quantval[i] *
+	       aanscalefactor[row] * aanscalefactor[col]);
+	    i++;
+	  }
+	}
+      }
+      break;
+#endif
+    default:
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+      break;
+    }
+  }
+}
+
+
+/*
+ * Initialize IDCT manager.
+ */
+
+GLOBAL(void)
+jinit_inverse_dct (j_decompress_ptr cinfo)
+{
+  my_idct_ptr idct;
+  int ci;
+  jpeg_component_info *compptr;
+
+  idct = (my_idct_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_idct_controller));
+  cinfo->idct = (struct jpeg_inverse_dct *) idct;
+  idct->pub.start_pass = start_pass;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Allocate and pre-zero a multiplier table for each component */
+    compptr->dct_table =
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(multiplier_table));
+    MEMZERO(compptr->dct_table, SIZEOF(multiplier_table));
+    /* Mark multiplier table not yet set up for any method */
+    idct->cur_method[ci] = -1;
+  }
+}
diff --git a/libjpegtwrp/jdhuff.c b/libjpegtwrp/jdhuff.c
new file mode 100644
index 000000000..bc5d4fdd2
--- /dev/null
+++ b/libjpegtwrp/jdhuff.c
@@ -0,0 +1,894 @@
+/*
+ * jdhuff.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains Huffman entropy decoding routines.
+ *
+ * Much of the complexity here has to do with supporting input suspension.
+ * If the data source module demands suspension, we want to be able to back
+ * up to the start of the current MCU.  To do this, we copy state variables
+ * into local working storage, and update them back to the permanent
+ * storage only upon successful completion of an MCU.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdhuff.h"		/* Declarations shared with jdphuff.c */
+
+LOCAL(boolean) process_restart (j_decompress_ptr cinfo);
+
+
+/*
+ * Expanded entropy decoder object for Huffman decoding.
+ *
+ * The savable_state subrecord contains fields that change within an MCU,
+ * but must not be updated permanently until we complete the MCU.
+ */
+
+typedef struct {
+  int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
+} savable_state;
+
+/* This macro is to work around compilers with missing or broken
+ * structure assignment.  You'll need to fix this code if you have
+ * such a compiler and you change MAX_COMPS_IN_SCAN.
+ */
+
+#ifndef NO_STRUCT_ASSIGN
+#define ASSIGN_STATE(dest,src)  ((dest) = (src))
+#else
+#if MAX_COMPS_IN_SCAN == 4
+#define ASSIGN_STATE(dest,src)  \
+	((dest).last_dc_val[0] = (src).last_dc_val[0], \
+	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
+	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
+	 (dest).last_dc_val[3] = (src).last_dc_val[3])
+#endif
+#endif
+
+
+typedef struct {
+  struct jpeg_entropy_decoder pub; /* public fields */
+
+  /* These fields are loaded into local variables at start of each MCU.
+   * In case of suspension, we exit WITHOUT updating them.
+   */
+  bitread_perm_state bitstate;	/* Bit buffer at start of MCU */
+  savable_state saved;		/* Other state at start of MCU */
+
+  /* These fields are NOT loaded into local working state. */
+  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+
+  /* Pointers to derived tables (these workspaces have image lifespan) */
+  d_derived_tbl * dc_derived_tbls[NUM_HUFF_TBLS];
+  d_derived_tbl * ac_derived_tbls[NUM_HUFF_TBLS];
+
+  /* Precalculated info set up by start_pass for use in decode_mcu: */
+
+  /* Pointers to derived tables to be used for each block within an MCU */
+  d_derived_tbl * dc_cur_tbls[D_MAX_BLOCKS_IN_MCU];
+  d_derived_tbl * ac_cur_tbls[D_MAX_BLOCKS_IN_MCU];
+  /* Whether we care about the DC and AC coefficient values for each block */
+  boolean dc_needed[D_MAX_BLOCKS_IN_MCU];
+  boolean ac_needed[D_MAX_BLOCKS_IN_MCU];
+} huff_entropy_decoder;
+
+typedef huff_entropy_decoder * huff_entropy_ptr;
+
+/*
+ * Initialize for a Huffman-compressed scan.
+ */
+
+METHODDEF(void)
+start_pass_huff_decoder (j_decompress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int ci, blkn, dctbl, actbl;
+  jpeg_component_info * compptr;
+
+  /* Check that the scan parameters Ss, Se, Ah/Al are OK for sequential JPEG.
+   * This ought to be an error condition, but we make it a warning because
+   * there are some baseline files out there with all zeroes in these bytes.
+   */
+  if (cinfo->Ss != 0 || cinfo->Se != DCTSIZE2-1 ||
+      cinfo->Ah != 0 || cinfo->Al != 0)
+    WARNMS(cinfo, JWRN_NOT_SEQUENTIAL);
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    dctbl = compptr->dc_tbl_no;
+    actbl = compptr->ac_tbl_no;
+    /* Compute derived values for Huffman tables */
+    /* We may do this more than once for a table, but it's not expensive */
+    jpeg_make_d_derived_tbl(cinfo, TRUE, dctbl,
+			    & entropy->dc_derived_tbls[dctbl]);
+    jpeg_make_d_derived_tbl(cinfo, FALSE, actbl,
+			    & entropy->ac_derived_tbls[actbl]);
+    /* Initialize DC predictions to 0 */
+    entropy->saved.last_dc_val[ci] = 0;
+  }
+
+  /* Precalculate decoding info for each block in an MCU of this scan */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    ci = cinfo->MCU_membership[blkn];
+    compptr = cinfo->cur_comp_info[ci];
+    /* Precalculate which table to use for each block */
+    entropy->dc_cur_tbls[blkn] = entropy->dc_derived_tbls[compptr->dc_tbl_no];
+    entropy->ac_cur_tbls[blkn] = entropy->ac_derived_tbls[compptr->ac_tbl_no];
+    /* Decide whether we really care about the coefficient values */
+    if (compptr->component_needed) {
+      entropy->dc_needed[blkn] = TRUE;
+      /* we don't need the ACs if producing a 1/8th-size image */
+      entropy->ac_needed[blkn] = (compptr->DCT_scaled_size > 1);
+    } else {
+      entropy->dc_needed[blkn] = entropy->ac_needed[blkn] = FALSE;
+    }
+  }
+
+  /* Initialize bitread state variables */
+  entropy->bitstate.bits_left = 0;
+  entropy->bitstate.get_buffer = 0; /* unnecessary, but keeps Purify quiet */
+  entropy->pub.insufficient_data = FALSE;
+
+  /* Initialize restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+}
+
+
+/*
+ * Compute the derived values for a Huffman table.
+ * This routine also performs some validation checks on the table.
+ *
+ * Note this is also used by jdphuff.c.
+ */
+
+GLOBAL(void)
+jpeg_make_d_derived_tbl (j_decompress_ptr cinfo, boolean isDC, int tblno,
+			 d_derived_tbl ** pdtbl)
+{
+  JHUFF_TBL *htbl;
+  d_derived_tbl *dtbl;
+  int p, i, l, si, numsymbols;
+  int lookbits, ctr;
+  char huffsize[257];
+  unsigned int huffcode[257];
+  unsigned int code;
+
+  /* Note that huffsize[] and huffcode[] are filled in code-length order,
+   * paralleling the order of the symbols themselves in htbl->huffval[].
+   */
+
+  /* Find the input Huffman table */
+  if (tblno < 0 || tblno >= NUM_HUFF_TBLS)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
+  htbl =
+    isDC ? cinfo->dc_huff_tbl_ptrs[tblno] : cinfo->ac_huff_tbl_ptrs[tblno];
+  if (htbl == NULL)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
+
+  /* Allocate a workspace if we haven't already done so. */
+  if (*pdtbl == NULL)
+    *pdtbl = (d_derived_tbl *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(d_derived_tbl));
+  dtbl = *pdtbl;
+  dtbl->pub = htbl;		/* fill in back link */
+  
+  /* Figure C.1: make table of Huffman code length for each symbol */
+
+  p = 0;
+  for (l = 1; l <= 16; l++) {
+    i = (int) htbl->bits[l];
+    if (i < 0 || p + i > 256)	/* protect against table overrun */
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    while (i--)
+      huffsize[p++] = (char) l;
+  }
+  huffsize[p] = 0;
+  numsymbols = p;
+  
+  /* Figure C.2: generate the codes themselves */
+  /* We also validate that the counts represent a legal Huffman code tree. */
+  
+  code = 0;
+  si = huffsize[0];
+  p = 0;
+  while (huffsize[p]) {
+    while (((int) huffsize[p]) == si) {
+      huffcode[p++] = code;
+      code++;
+    }
+    /* code is now 1 more than the last code used for codelength si; but
+     * it must still fit in si bits, since no code is allowed to be all ones.
+     */
+    if (((INT32) code) >= (((INT32) 1) << si))
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    code <<= 1;
+    si++;
+  }
+
+  /* Figure F.15: generate decoding tables for bit-sequential decoding */
+
+  p = 0;
+  for (l = 1; l <= 16; l++) {
+    if (htbl->bits[l]) {
+      /* valoffset[l] = huffval[] index of 1st symbol of code length l,
+       * minus the minimum code of length l
+       */
+      dtbl->valoffset[l] = (INT32) p - (INT32) huffcode[p];
+      p += htbl->bits[l];
+      dtbl->maxcode[l] = huffcode[p-1]; /* maximum code of length l */
+    } else {
+      dtbl->maxcode[l] = -1;	/* -1 if no codes of this length */
+    }
+  }
+  dtbl->maxcode[17] = 0xFFFFFL; /* ensures jpeg_huff_decode terminates */
+
+  /* Compute lookahead tables to speed up decoding.
+   * First we set all the table entries to 0, indicating "too long";
+   * then we iterate through the Huffman codes that are short enough and
+   * fill in all the entries that correspond to bit sequences starting
+   * with that code.
+   */
+
+  MEMZERO(dtbl->look_nbits, SIZEOF(dtbl->look_nbits));
+
+  p = 0;
+  for (l = 1; l <= HUFF_LOOKAHEAD; l++) {
+    for (i = 1; i <= (int) htbl->bits[l]; i++, p++) {
+      /* l = current code's length, p = its index in huffcode[] & huffval[]. */
+      /* Generate left-justified code followed by all possible bit sequences */
+      lookbits = huffcode[p] << (HUFF_LOOKAHEAD-l);
+      for (ctr = 1 << (HUFF_LOOKAHEAD-l); ctr > 0; ctr--) {
+	dtbl->look_nbits[lookbits] = l;
+	dtbl->look_sym[lookbits] = htbl->huffval[p];
+	lookbits++;
+      }
+    }
+  }
+
+  /* Validate symbols as being reasonable.
+   * For AC tables, we make no check, but accept all byte values 0..255.
+   * For DC tables, we require the symbols to be in range 0..15.
+   * (Tighter bounds could be applied depending on the data depth and mode,
+   * but this is sufficient to ensure safe decoding.)
+   */
+  if (isDC) {
+    for (i = 0; i < numsymbols; i++) {
+      int sym = htbl->huffval[i];
+      if (sym < 0 || sym > 15)
+	ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    }
+  }
+}
+
+
+/*
+ * Out-of-line code for bit fetching (shared with jdphuff.c).
+ * See jdhuff.h for info about usage.
+ * Note: current values of get_buffer and bits_left are passed as parameters,
+ * but are returned in the corresponding fields of the state struct.
+ *
+ * On most machines MIN_GET_BITS should be 25 to allow the full 32-bit width
+ * of get_buffer to be used.  (On machines with wider words, an even larger
+ * buffer could be used.)  However, on some machines 32-bit shifts are
+ * quite slow and take time proportional to the number of places shifted.
+ * (This is true with most PC compilers, for instance.)  In this case it may
+ * be a win to set MIN_GET_BITS to the minimum value of 15.  This reduces the
+ * average shift distance at the cost of more calls to jpeg_fill_bit_buffer.
+ */
+
+#ifdef SLOW_SHIFT_32
+#define MIN_GET_BITS  15	/* minimum allowable value */
+#else
+#define MIN_GET_BITS  (BIT_BUF_SIZE-7)
+#endif
+
+
+GLOBAL(boolean)
+jpeg_fill_bit_buffer (bitread_working_state * state,
+		      register bit_buf_type get_buffer, register int bits_left,
+		      int nbits)
+/* Load up the bit buffer to a depth of at least nbits */
+{
+  /* Copy heavily used state fields into locals (hopefully registers) */
+  register const JOCTET * next_input_byte = state->next_input_byte;
+  register size_t bytes_in_buffer = state->bytes_in_buffer;
+  j_decompress_ptr cinfo = state->cinfo;
+
+  /* Attempt to load at least MIN_GET_BITS bits into get_buffer. */
+  /* (It is assumed that no request will be for more than that many bits.) */
+  /* We fail to do so only if we hit a marker or are forced to suspend. */
+
+  if (cinfo->unread_marker == 0) {	/* cannot advance past a marker */
+    while (bits_left < MIN_GET_BITS) {
+      register int c;
+
+      /* Attempt to read a byte */
+      if (bytes_in_buffer == 0) {
+	if (! (*cinfo->src->fill_input_buffer) (cinfo))
+	  return FALSE;
+	next_input_byte = cinfo->src->next_input_byte;
+	bytes_in_buffer = cinfo->src->bytes_in_buffer;
+      }
+      bytes_in_buffer--;
+      c = GETJOCTET(*next_input_byte++);
+
+      /* If it's 0xFF, check and discard stuffed zero byte */
+      if (c == 0xFF) {
+	/* Loop here to discard any padding FF's on terminating marker,
+	 * so that we can save a valid unread_marker value.  NOTE: we will
+	 * accept multiple FF's followed by a 0 as meaning a single FF data
+	 * byte.  This data pattern is not valid according to the standard.
+	 */
+	do {
+	  if (bytes_in_buffer == 0) {
+	    if (! (*cinfo->src->fill_input_buffer) (cinfo))
+	      return FALSE;
+	    next_input_byte = cinfo->src->next_input_byte;
+	    bytes_in_buffer = cinfo->src->bytes_in_buffer;
+	  }
+	  bytes_in_buffer--;
+	  c = GETJOCTET(*next_input_byte++);
+	} while (c == 0xFF);
+
+	if (c == 0) {
+	  /* Found FF/00, which represents an FF data byte */
+	  c = 0xFF;
+	} else {
+	  /* Oops, it's actually a marker indicating end of compressed data.
+	   * Save the marker code for later use.
+	   * Fine point: it might appear that we should save the marker into
+	   * bitread working state, not straight into permanent state.  But
+	   * once we have hit a marker, we cannot need to suspend within the
+	   * current MCU, because we will read no more bytes from the data
+	   * source.  So it is OK to update permanent state right away.
+	   */
+	  cinfo->unread_marker = c;
+	  /* See if we need to insert some fake zero bits. */
+	  goto no_more_bytes;
+	}
+      }
+
+      /* OK, load c into get_buffer */
+      get_buffer = (get_buffer << 8) | c;
+      bits_left += 8;
+    } /* end while */
+  } else {
+  no_more_bytes:
+    /* We get here if we've read the marker that terminates the compressed
+     * data segment.  There should be enough bits in the buffer register
+     * to satisfy the request; if so, no problem.
+     */
+    if (nbits > bits_left) {
+      /* Uh-oh.  Report corrupted data to user and stuff zeroes into
+       * the data stream, so that we can produce some kind of image.
+       * We use a nonvolatile flag to ensure that only one warning message
+       * appears per data segment.
+       */
+      if (! cinfo->entropy->insufficient_data) {
+	WARNMS(cinfo, JWRN_HIT_MARKER);
+	cinfo->entropy->insufficient_data = TRUE;
+      }
+      /* Fill the buffer with zero bits */
+      get_buffer <<= MIN_GET_BITS - bits_left;
+      bits_left = MIN_GET_BITS;
+    }
+  }
+
+  /* Unload the local registers */
+  state->next_input_byte = next_input_byte;
+  state->bytes_in_buffer = bytes_in_buffer;
+  state->get_buffer = get_buffer;
+  state->bits_left = bits_left;
+
+  return TRUE;
+}
+
+
+/*
+ * Out-of-line code for Huffman code decoding.
+ * See jdhuff.h for info about usage.
+ */
+
+GLOBAL(int)
+jpeg_huff_decode (bitread_working_state * state,
+		  register bit_buf_type get_buffer, register int bits_left,
+		  d_derived_tbl * htbl, int min_bits)
+{
+  register int l = min_bits;
+  register INT32 code;
+
+  /* HUFF_DECODE has determined that the code is at least min_bits */
+  /* bits long, so fetch that many bits in one swoop. */
+
+  CHECK_BIT_BUFFER(*state, l, return -1);
+  code = GET_BITS(l);
+
+  /* Collect the rest of the Huffman code one bit at a time. */
+  /* This is per Figure F.16 in the JPEG spec. */
+
+  while (code > htbl->maxcode[l]) {
+    code <<= 1;
+    CHECK_BIT_BUFFER(*state, 1, return -1);
+    code |= GET_BITS(1);
+    l++;
+  }
+
+  /* Unload the local registers */
+  state->get_buffer = get_buffer;
+  state->bits_left = bits_left;
+
+  /* With garbage input we may reach the sentinel value l = 17. */
+
+  if (l > 16) {
+    WARNMS(state->cinfo, JWRN_HUFF_BAD_CODE);
+    return 0;			/* fake a zero as the safest result */
+  }
+
+  return htbl->pub->huffval[ (int) (code + htbl->valoffset[l]) ];
+}
+
+
+/*
+ * Figure F.12: extend sign bit.
+ * On some machines, a shift and add will be faster than a table lookup.
+ */
+
+#ifdef AVOID_TABLES
+
+#define HUFF_EXTEND(x,s)  ((x) < (1<<((s)-1)) ? (x) + (((-1)<<(s)) + 1) : (x))
+
+#else
+
+#define HUFF_EXTEND(x,s)  ((x) < extend_test[s] ? (x) + extend_offset[s] : (x))
+
+static const int extend_test[16] =   /* entry n is 2**(n-1) */
+  { 0, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080,
+    0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000 };
+
+static const int extend_offset[16] = /* entry n is (-1 << n) + 1 */
+  { 0, ((-1)<<1) + 1, ((-1)<<2) + 1, ((-1)<<3) + 1, ((-1)<<4) + 1,
+    ((-1)<<5) + 1, ((-1)<<6) + 1, ((-1)<<7) + 1, ((-1)<<8) + 1,
+    ((-1)<<9) + 1, ((-1)<<10) + 1, ((-1)<<11) + 1, ((-1)<<12) + 1,
+    ((-1)<<13) + 1, ((-1)<<14) + 1, ((-1)<<15) + 1 };
+
+#endif /* AVOID_TABLES */
+
+
+/*
+ * Check for a restart marker & resynchronize decoder.
+ * Returns FALSE if must suspend.
+ */
+
+LOCAL(boolean)
+process_restart (j_decompress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int ci;
+
+  /* Throw away any unused bits remaining in bit buffer; */
+  /* include any full bytes in next_marker's count of discarded bytes */
+  cinfo->marker->discarded_bytes += entropy->bitstate.bits_left / 8;
+  entropy->bitstate.bits_left = 0;
+
+  /* Advance past the RSTn marker */
+  if (! (*cinfo->marker->read_restart_marker) (cinfo))
+    return FALSE;
+
+  /* Re-initialize DC predictions to 0 */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++)
+    entropy->saved.last_dc_val[ci] = 0;
+
+  /* Reset restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+
+  /* Reset out-of-data flag, unless read_restart_marker left us smack up
+   * against a marker.  In that case we will end up treating the next data
+   * segment as empty, and we can avoid producing bogus output pixels by
+   * leaving the flag set.
+   */
+  if (cinfo->unread_marker == 0)
+    entropy->pub.insufficient_data = FALSE;
+
+  return TRUE;
+}
+
+/*
+ * Save the current Huffman deocde position and the DC coefficients
+ * for each component into bitstream_offset and dc_info[], respectively.
+ */
+METHODDEF(void)
+get_huffman_decoder_configuration(j_decompress_ptr cinfo,
+        huffman_offset_data *offset)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  short int *dc_info = offset->prev_dc;
+  int i;
+  jpeg_get_huffman_decoder_configuration(cinfo, offset);
+  for (i = 0; i < cinfo->comps_in_scan; i++) {
+    dc_info[i] = entropy->saved.last_dc_val[i];
+  }
+}
+
+/*
+ * Save the current Huffman decoder position and the bit buffer
+ * into bitstream_offset and get_buffer, respectively.
+ */
+GLOBAL(void)
+jpeg_get_huffman_decoder_configuration(j_decompress_ptr cinfo,
+        huffman_offset_data *offset)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+
+  if (cinfo->restart_interval) {
+    // We are at the end of a data segment
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return;
+  }
+
+  // Save restarts_to_go and next_restart_num
+  offset->restarts_to_go = (unsigned short) entropy->restarts_to_go;
+  offset->next_restart_num = cinfo->marker->next_restart_num;
+
+  offset->bitstream_offset =
+      (jget_input_stream_position(cinfo) << LOG_TWO_BIT_BUF_SIZE)
+      + entropy->bitstate.bits_left;
+
+  offset->get_buffer = entropy->bitstate.get_buffer;
+}
+
+/*
+ * Configure the Huffman decoder to decode the image
+ * starting from the bitstream position recorded in offset.
+ */
+METHODDEF(void)
+configure_huffman_decoder(j_decompress_ptr cinfo, huffman_offset_data offset)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  short int *dc_info = offset.prev_dc;
+  int i;
+  jpeg_configure_huffman_decoder(cinfo, offset);
+  for (i = 0; i < cinfo->comps_in_scan; i++) {
+    entropy->saved.last_dc_val[i] = dc_info[i];
+  }
+}
+
+/*
+ * Configure the Huffman decoder reader position and bit buffer.
+ */
+GLOBAL(void)
+jpeg_configure_huffman_decoder(j_decompress_ptr cinfo,
+        huffman_offset_data offset)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+
+  // Restore restarts_to_go and next_restart_num
+  cinfo->unread_marker = 0;
+  entropy->restarts_to_go = offset.restarts_to_go;
+  cinfo->marker->next_restart_num = offset.next_restart_num;
+
+  unsigned int bitstream_offset = offset.bitstream_offset;
+  int blkn, i;
+
+  unsigned int byte_offset = bitstream_offset >> LOG_TWO_BIT_BUF_SIZE;
+  unsigned int bit_in_bit_buffer =
+      bitstream_offset & ((1 << LOG_TWO_BIT_BUF_SIZE) - 1);
+
+  jset_input_stream_position_bit(cinfo, byte_offset,
+          bit_in_bit_buffer, offset.get_buffer);
+}
+
+/*
+ * Decode and return one MCU's worth of Huffman-compressed coefficients.
+ * The coefficients are reordered from zigzag order into natural array order,
+ * but are not dequantized.
+ *
+ * The i'th block of the MCU is stored into the block pointed to by
+ * MCU_data[i].  WE ASSUME THIS AREA HAS BEEN ZEROED BY THE CALLER.
+ * (Wholesale zeroing is usually a little faster than retail...)
+ *
+ * Returns FALSE if data source requested suspension.  In that case no
+ * changes have been made to permanent state.  (Exception: some output
+ * coefficients may already have been assigned.  This is harmless for
+ * this module, since we'll just re-assign them on the next call.)
+ */
+
+METHODDEF(boolean)
+decode_mcu (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int blkn;
+  BITREAD_STATE_VARS;
+  savable_state state;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return FALSE;
+  }
+
+  /* If we've run out of data, just leave the MCU set to zeroes.
+   * This way, we return uniform gray for the remainder of the segment.
+   */
+  if (! entropy->pub.insufficient_data) {
+    /* Load up working state */
+    BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
+    ASSIGN_STATE(state, entropy->saved);
+
+    /* Outer loop handles each block in the MCU */
+
+    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+      JBLOCKROW block = MCU_data[blkn];
+      d_derived_tbl * dctbl = entropy->dc_cur_tbls[blkn];
+      d_derived_tbl * actbl = entropy->ac_cur_tbls[blkn];
+      register int s, k, r;
+
+      /* Decode a single block's worth of coefficients */
+
+      /* Section F.2.2.1: decode the DC coefficient difference */
+      HUFF_DECODE(s, br_state, dctbl, return FALSE, label1);
+      if (s) {
+	CHECK_BIT_BUFFER(br_state, s, return FALSE);
+	r = GET_BITS(s);
+	s = HUFF_EXTEND(r, s);
+      }
+
+      if (entropy->dc_needed[blkn]) {
+	/* Convert DC difference to actual value, update last_dc_val */
+	int ci = cinfo->MCU_membership[blkn];
+	s += state.last_dc_val[ci];
+	state.last_dc_val[ci] = s;
+	/* Output the DC coefficient (assumes jpeg_natural_order[0] = 0) */
+	(*block)[0] = (JCOEF) s;
+      }
+
+      if (entropy->ac_needed[blkn]) {
+
+	/* Section F.2.2.2: decode the AC coefficients */
+	/* Since zeroes are skipped, output area must be cleared beforehand */
+	for (k = 1; k < DCTSIZE2; k++) {
+	  HUFF_DECODE(s, br_state, actbl, return FALSE, label2);
+      
+	  r = s >> 4;
+	  s &= 15;
+      
+	  if (s) {
+	    k += r;
+	    CHECK_BIT_BUFFER(br_state, s, return FALSE);
+	    r = GET_BITS(s);
+	    s = HUFF_EXTEND(r, s);
+	    /* Output coefficient in natural (dezigzagged) order.
+	     * Note: the extra entries in jpeg_natural_order[] will save us
+	     * if k >= DCTSIZE2, which could happen if the data is corrupted.
+	     */
+	    (*block)[jpeg_natural_order[k]] = (JCOEF) s;
+	  } else {
+	    if (r != 15)
+	      break;
+	    k += 15;
+	  }
+	}
+
+      } else {
+
+	/* Section F.2.2.2: decode the AC coefficients */
+	/* In this path we just discard the values */
+	for (k = 1; k < DCTSIZE2; k++) {
+	  HUFF_DECODE(s, br_state, actbl, return FALSE, label3);
+      
+	  r = s >> 4;
+	  s &= 15;
+      
+	  if (s) {
+	    k += r;
+	    CHECK_BIT_BUFFER(br_state, s, return FALSE);
+	    DROP_BITS(s);
+	  } else {
+	    if (r != 15)
+	      break;
+	    k += 15;
+	  }
+	}
+
+      }
+    }
+
+    /* Completed MCU, so update state */
+    BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
+    ASSIGN_STATE(entropy->saved, state);
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+/*
+ * Decode one MCU's worth of Huffman-compressed coefficients.
+ * The propose of this method is to calculate the
+ * data length of one MCU in Huffman-coded format.
+ * Therefore, all coefficients are discarded.
+ */
+
+METHODDEF(boolean)
+decode_mcu_discard_coef (j_decompress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int blkn;
+  BITREAD_STATE_VARS;
+  savable_state state;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return FALSE;
+  }
+
+  if (! entropy->pub.insufficient_data) {
+
+    /* Load up working state */
+    BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
+    ASSIGN_STATE(state, entropy->saved);
+
+    /* Outer loop handles each block in the MCU */
+
+    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+      d_derived_tbl * dctbl = entropy->dc_cur_tbls[blkn];
+      d_derived_tbl * actbl = entropy->ac_cur_tbls[blkn];
+      register int s, k, r;
+
+      /* Decode a single block's worth of coefficients */
+
+      /* Section F.2.2.1: decode the DC coefficient difference */
+      HUFF_DECODE(s, br_state, dctbl, return FALSE, label1);
+      if (s) {
+	CHECK_BIT_BUFFER(br_state, s, return FALSE);
+	r = GET_BITS(s);
+	s = HUFF_EXTEND(r, s);
+      }
+
+      /* discard all coefficients */
+      if (entropy->dc_needed[blkn]) {
+	/* Convert DC difference to actual value, update last_dc_val */
+	int ci = cinfo->MCU_membership[blkn];
+	s += state.last_dc_val[ci];
+	state.last_dc_val[ci] = s;
+      }
+      for (k = 1; k < DCTSIZE2; k++) {
+        HUFF_DECODE(s, br_state, actbl, return FALSE, label3);
+
+        r = s >> 4;
+        s &= 15;
+
+        if (s) {
+          k += r;
+          CHECK_BIT_BUFFER(br_state, s, return FALSE);
+          DROP_BITS(s);
+        } else {
+          if (r != 15)
+            break;
+          k += 15;
+        }
+      }
+    }
+
+    /* Completed MCU, so update state */
+    BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
+    ASSIGN_STATE(entropy->saved, state);
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+
+/*
+ * Module initialization routine for Huffman entropy decoding.
+ */
+
+GLOBAL(void)
+jinit_huff_decoder (j_decompress_ptr cinfo)
+{
+  huff_entropy_ptr entropy;
+  int i;
+
+  entropy = (huff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(huff_entropy_decoder));
+  cinfo->entropy = (struct jpeg_entropy_decoder *) entropy;
+  entropy->pub.start_pass = start_pass_huff_decoder;
+  entropy->pub.decode_mcu = decode_mcu;
+  entropy->pub.decode_mcu_discard_coef = decode_mcu_discard_coef;
+  entropy->pub.configure_huffman_decoder = configure_huffman_decoder;
+  entropy->pub.get_huffman_decoder_configuration =
+        get_huffman_decoder_configuration;
+  entropy->pub.index = NULL;
+
+  /* Mark tables unallocated */
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    entropy->dc_derived_tbls[i] = entropy->ac_derived_tbls[i] = NULL;
+  }
+}
+
+/*
+ * Call after jpeg_read_header
+ */
+GLOBAL(void)
+jpeg_create_huffman_index(j_decompress_ptr cinfo, huffman_index *index)
+{
+  int i, s;
+  index->scan_count = 1;
+  index->total_iMCU_rows = cinfo->total_iMCU_rows;
+  index->scan = (huffman_scan_header*)malloc(index->scan_count
+          * sizeof(huffman_scan_header));
+  index->scan[0].offset = (huffman_offset_data**)malloc(cinfo->total_iMCU_rows
+          * sizeof(huffman_offset_data*));
+  index->scan[0].prev_MCU_offset.bitstream_offset = 0;
+  index->MCU_sample_size = DEFAULT_MCU_SAMPLE_SIZE;
+
+  index->mem_used = sizeof(huffman_scan_header)
+      + cinfo->total_iMCU_rows * sizeof(huffman_offset_data*);
+}
+
+GLOBAL(void)
+jpeg_destroy_huffman_index(huffman_index *index)
+{
+    int i, j;
+    for (i = 0; i < index->scan_count; i++) {
+        for(j = 0; j < index->total_iMCU_rows; j++) {
+            free(index->scan[i].offset[j]);
+        }
+        free(index->scan[i].offset);
+    }
+    free(index->scan);
+}
+
+/*
+ * Set the reader byte position to offset
+ */
+GLOBAL(void)
+jset_input_stream_position(j_decompress_ptr cinfo, int offset)
+{
+  if (cinfo->src->seek_input_data) {
+    cinfo->src->seek_input_data(cinfo, offset);
+  } else {
+    cinfo->src->bytes_in_buffer = cinfo->src->current_offset - offset;
+    cinfo->src->next_input_byte = cinfo->src->start_input_byte + offset;
+  }
+}
+
+/*
+ * Set the reader byte position to offset and bit position to bit_left
+ * with bit buffer set to buf.
+ */
+GLOBAL(void)
+jset_input_stream_position_bit(j_decompress_ptr cinfo,
+        int byte_offset, int bit_left, INT32 buf)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+
+  entropy->bitstate.bits_left = bit_left;
+  entropy->bitstate.get_buffer = buf;
+
+  jset_input_stream_position(cinfo, byte_offset);
+}
+
+/*
+ * Get the current reader byte position.
+ */
+GLOBAL(int)
+jget_input_stream_position(j_decompress_ptr cinfo)
+{
+  return cinfo->src->current_offset - cinfo->src->bytes_in_buffer;
+}
diff --git a/libjpegtwrp/jdhuff.h b/libjpegtwrp/jdhuff.h
new file mode 100644
index 000000000..5760a134e
--- /dev/null
+++ b/libjpegtwrp/jdhuff.h
@@ -0,0 +1,202 @@
+/*
+ * jdhuff.h
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains declarations for Huffman entropy decoding routines
+ * that are shared between the sequential decoder (jdhuff.c) and the
+ * progressive decoder (jdphuff.c).  No other modules need to see these.
+ */
+
+/* Short forms of external names for systems with brain-damaged linkers. */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jpeg_make_d_derived_tbl	jMkDDerived
+#define jpeg_fill_bit_buffer	jFilBitBuf
+#define jpeg_huff_decode	jHufDecode
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+
+/* Derived data constructed for each Huffman table */
+
+#define HUFF_LOOKAHEAD	8	/* # of bits of lookahead */
+
+typedef struct {
+  /* Basic tables: (element [0] of each array is unused) */
+  INT32 maxcode[18];		/* largest code of length k (-1 if none) */
+  /* (maxcode[17] is a sentinel to ensure jpeg_huff_decode terminates) */
+  INT32 valoffset[17];		/* huffval[] offset for codes of length k */
+  /* valoffset[k] = huffval[] index of 1st symbol of code length k, less
+   * the smallest code of length k; so given a code of length k, the
+   * corresponding symbol is huffval[code + valoffset[k]]
+   */
+
+  /* Link to public Huffman table (needed only in jpeg_huff_decode) */
+  JHUFF_TBL *pub;
+
+  /* Lookahead tables: indexed by the next HUFF_LOOKAHEAD bits of
+   * the input data stream.  If the next Huffman code is no more
+   * than HUFF_LOOKAHEAD bits long, we can obtain its length and
+   * the corresponding symbol directly from these tables.
+   */
+  int look_nbits[1<<HUFF_LOOKAHEAD]; /* # bits, or 0 if too long */
+  UINT8 look_sym[1<<HUFF_LOOKAHEAD]; /* symbol, or unused */
+} d_derived_tbl;
+
+/* Expand a Huffman table definition into the derived format */
+EXTERN(void) jpeg_make_d_derived_tbl
+	JPP((j_decompress_ptr cinfo, boolean isDC, int tblno,
+	     d_derived_tbl ** pdtbl));
+
+
+/*
+ * Fetching the next N bits from the input stream is a time-critical operation
+ * for the Huffman decoders.  We implement it with a combination of inline
+ * macros and out-of-line subroutines.  Note that N (the number of bits
+ * demanded at one time) never exceeds 15 for JPEG use.
+ *
+ * We read source bytes into get_buffer and dole out bits as needed.
+ * If get_buffer already contains enough bits, they are fetched in-line
+ * by the macros CHECK_BIT_BUFFER and GET_BITS.  When there aren't enough
+ * bits, jpeg_fill_bit_buffer is called; it will attempt to fill get_buffer
+ * as full as possible (not just to the number of bits needed; this
+ * prefetching reduces the overhead cost of calling jpeg_fill_bit_buffer).
+ * Note that jpeg_fill_bit_buffer may return FALSE to indicate suspension.
+ * On TRUE return, jpeg_fill_bit_buffer guarantees that get_buffer contains
+ * at least the requested number of bits --- dummy zeroes are inserted if
+ * necessary.
+ */
+
+typedef INT32 bit_buf_type;	/* type of bit-extraction buffer */
+#define BIT_BUF_SIZE  32	/* size of buffer in bits */
+#define LOG_TWO_BIT_BUF_SIZE  5	/* log_2(BIT_BUF_SIZE) */
+
+/* If long is > 32 bits on your machine, and shifting/masking longs is
+ * reasonably fast, making bit_buf_type be long and setting BIT_BUF_SIZE
+ * appropriately should be a win.  Unfortunately we can't define the size
+ * with something like  #define BIT_BUF_SIZE (sizeof(bit_buf_type)*8)
+ * because not all machines measure sizeof in 8-bit bytes.
+ */
+
+typedef struct {		/* Bitreading state saved across MCUs */
+  bit_buf_type get_buffer;	/* current bit-extraction buffer */
+  int bits_left;		/* # of unused bits in it */
+} bitread_perm_state;
+
+typedef struct {		/* Bitreading working state within an MCU */
+  /* Current data source location */
+  /* We need a copy, rather than munging the original, in case of suspension */
+  const JOCTET * next_input_byte; /* => next byte to read from source */
+  size_t bytes_in_buffer;	/* # of bytes remaining in source buffer */
+  /* Bit input buffer --- note these values are kept in register variables,
+   * not in this struct, inside the inner loops.
+   */
+  bit_buf_type get_buffer;	/* current bit-extraction buffer */
+  int bits_left;		/* # of unused bits in it */
+  /* Pointer needed by jpeg_fill_bit_buffer. */
+  j_decompress_ptr cinfo;	/* back link to decompress master record */
+} bitread_working_state;
+
+/* Macros to declare and load/save bitread local variables. */
+#define BITREAD_STATE_VARS  \
+	register bit_buf_type get_buffer;  \
+	register int bits_left;  \
+	bitread_working_state br_state
+
+#define BITREAD_LOAD_STATE(cinfop,permstate)  \
+	br_state.cinfo = cinfop; \
+	br_state.next_input_byte = cinfop->src->next_input_byte; \
+	br_state.bytes_in_buffer = cinfop->src->bytes_in_buffer; \
+	get_buffer = permstate.get_buffer; \
+	bits_left = permstate.bits_left;
+
+#define BITREAD_SAVE_STATE(cinfop,permstate)  \
+	cinfop->src->next_input_byte = br_state.next_input_byte; \
+	cinfop->src->bytes_in_buffer = br_state.bytes_in_buffer; \
+	permstate.get_buffer = get_buffer; \
+	permstate.bits_left = bits_left
+
+/*
+ * These macros provide the in-line portion of bit fetching.
+ * Use CHECK_BIT_BUFFER to ensure there are N bits in get_buffer
+ * before using GET_BITS, PEEK_BITS, or DROP_BITS.
+ * The variables get_buffer and bits_left are assumed to be locals,
+ * but the state struct might not be (jpeg_huff_decode needs this).
+ *	CHECK_BIT_BUFFER(state,n,action);
+ *		Ensure there are N bits in get_buffer; if suspend, take action.
+ *      val = GET_BITS(n);
+ *		Fetch next N bits.
+ *      val = PEEK_BITS(n);
+ *		Fetch next N bits without removing them from the buffer.
+ *	DROP_BITS(n);
+ *		Discard next N bits.
+ * The value N should be a simple variable, not an expression, because it
+ * is evaluated multiple times.
+ */
+
+#define CHECK_BIT_BUFFER(state,nbits,action) \
+	{ if (bits_left < (nbits)) {  \
+	    if (! jpeg_fill_bit_buffer(&(state),get_buffer,bits_left,nbits))  \
+	      { action; }  \
+	    get_buffer = (state).get_buffer; bits_left = (state).bits_left; } }
+
+#define GET_BITS(nbits) \
+	(((int) (get_buffer >> (bits_left -= (nbits)))) & ((1<<(nbits))-1))
+
+#define PEEK_BITS(nbits) \
+	(((int) (get_buffer >> (bits_left -  (nbits)))) & ((1<<(nbits))-1))
+
+#define DROP_BITS(nbits) \
+	(bits_left -= (nbits))
+
+/* Load up the bit buffer to a depth of at least nbits */
+EXTERN(boolean) jpeg_fill_bit_buffer
+	JPP((bitread_working_state * state, register bit_buf_type get_buffer,
+	     register int bits_left, int nbits));
+
+
+/*
+ * Code for extracting next Huffman-coded symbol from input bit stream.
+ * Again, this is time-critical and we make the main paths be macros.
+ *
+ * We use a lookahead table to process codes of up to HUFF_LOOKAHEAD bits
+ * without looping.  Usually, more than 95% of the Huffman codes will be 8
+ * or fewer bits long.  The few overlength codes are handled with a loop,
+ * which need not be inline code.
+ *
+ * Notes about the HUFF_DECODE macro:
+ * 1. Near the end of the data segment, we may fail to get enough bits
+ *    for a lookahead.  In that case, we do it the hard way.
+ * 2. If the lookahead table contains no entry, the next code must be
+ *    more than HUFF_LOOKAHEAD bits long.
+ * 3. jpeg_huff_decode returns -1 if forced to suspend.
+ */
+
+#define HUFF_DECODE(result,state,htbl,failaction,slowlabel) \
+{ register int nb, look; \
+  if (bits_left < HUFF_LOOKAHEAD) { \
+    if (! jpeg_fill_bit_buffer(&state,get_buffer,bits_left, 0)) {failaction;} \
+    get_buffer = state.get_buffer; bits_left = state.bits_left; \
+    if (bits_left < HUFF_LOOKAHEAD) { \
+      nb = 1; goto slowlabel; \
+    } \
+  } \
+  look = PEEK_BITS(HUFF_LOOKAHEAD); \
+  if ((nb = htbl->look_nbits[look]) != 0) { \
+    DROP_BITS(nb); \
+    result = htbl->look_sym[look]; \
+  } else { \
+    nb = HUFF_LOOKAHEAD+1; \
+slowlabel: \
+    if ((result=jpeg_huff_decode(&state,get_buffer,bits_left,htbl,nb)) < 0) \
+	{ failaction; } \
+    get_buffer = state.get_buffer; bits_left = state.bits_left; \
+  } \
+}
+
+/* Out-of-line case for Huffman code fetching */
+EXTERN(int) jpeg_huff_decode
+	JPP((bitread_working_state * state, register bit_buf_type get_buffer,
+	     register int bits_left, d_derived_tbl * htbl, int min_bits));
diff --git a/libjpegtwrp/jdinput.c b/libjpegtwrp/jdinput.c
new file mode 100644
index 000000000..4261c1a1c
--- /dev/null
+++ b/libjpegtwrp/jdinput.c
@@ -0,0 +1,415 @@
+/*
+ * jdinput.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains input control logic for the JPEG decompressor.
+ * These routines are concerned with controlling the decompressor's input
+ * processing (marker reading and coefficient decoding).  The actual input
+ * reading is done in jdmarker.c, jdhuff.c, and jdphuff.c.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_input_controller pub; /* public fields */
+
+  boolean inheaders;		/* TRUE until first SOS is reached */
+} my_input_controller;
+
+typedef my_input_controller * my_inputctl_ptr;
+
+
+/* Forward declarations */
+METHODDEF(int) consume_markers JPP((j_decompress_ptr cinfo));
+METHODDEF(int) consume_markers_with_huffman_index JPP((j_decompress_ptr cinfo,
+                    huffman_index *index, int current_scan));
+
+
+/*
+ * Routines to calculate various quantities related to the size of the image.
+ */
+
+LOCAL(void)
+initial_setup (j_decompress_ptr cinfo)
+/* Called once, when first SOS marker is reached */
+{
+  int ci;
+  jpeg_component_info *compptr;
+
+  /* Make sure image isn't bigger than I can handle */
+  if ((long) cinfo->image_height > (long) JPEG_MAX_DIMENSION ||
+      (long) cinfo->image_width > (long) JPEG_MAX_DIMENSION)
+    ERREXIT1(cinfo, JERR_IMAGE_TOO_BIG, (unsigned int) JPEG_MAX_DIMENSION);
+
+  /* For now, precision must match compiled-in value... */
+  if (cinfo->data_precision != BITS_IN_JSAMPLE)
+    ERREXIT1(cinfo, JERR_BAD_PRECISION, cinfo->data_precision);
+
+  /* Check that number of components won't exceed internal array sizes */
+  if (cinfo->num_components > MAX_COMPONENTS)
+    ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
+	     MAX_COMPONENTS);
+
+  /* Compute maximum sampling factors; check factor validity */
+  cinfo->max_h_samp_factor = 1;
+  cinfo->max_v_samp_factor = 1;
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    if (compptr->h_samp_factor<=0 || compptr->h_samp_factor>MAX_SAMP_FACTOR ||
+	compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR)
+      ERREXIT(cinfo, JERR_BAD_SAMPLING);
+    cinfo->max_h_samp_factor = MAX(cinfo->max_h_samp_factor,
+				   compptr->h_samp_factor);
+    cinfo->max_v_samp_factor = MAX(cinfo->max_v_samp_factor,
+				   compptr->v_samp_factor);
+  }
+
+  /* We initialize DCT_scaled_size and min_DCT_scaled_size to DCTSIZE.
+   * In the full decompressor, this will be overridden by jdmaster.c;
+   * but in the transcoder, jdmaster.c is not used, so we must do it here.
+   */
+  cinfo->min_DCT_scaled_size = DCTSIZE;
+
+  /* Compute dimensions of components */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    compptr->DCT_scaled_size = DCTSIZE;
+    /* Size in DCT blocks */
+    compptr->width_in_blocks = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor,
+		    (long) (cinfo->max_h_samp_factor * DCTSIZE));
+    compptr->height_in_blocks = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor,
+		    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+    /* downsampled_width and downsampled_height will also be overridden by
+     * jdmaster.c if we are doing full decompression.  The transcoder library
+     * doesn't use these values, but the calling application might.
+     */
+    /* Size in samples */
+    compptr->downsampled_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor,
+		    (long) cinfo->max_h_samp_factor);
+    compptr->downsampled_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor,
+		    (long) cinfo->max_v_samp_factor);
+    /* Mark component needed, until color conversion says otherwise */
+    compptr->component_needed = TRUE;
+    /* Mark no quantization table yet saved for component */
+    compptr->quant_table = NULL;
+  }
+
+  /* Compute number of fully interleaved MCU rows. */
+  cinfo->total_iMCU_rows = (JDIMENSION)
+    jdiv_round_up((long) cinfo->image_height,
+		  (long) (cinfo->max_v_samp_factor*DCTSIZE));
+
+  /* Decide whether file contains multiple scans */
+  if (cinfo->comps_in_scan < cinfo->num_components || cinfo->progressive_mode)
+    cinfo->inputctl->has_multiple_scans = TRUE;
+  else
+    cinfo->inputctl->has_multiple_scans = FALSE;
+  cinfo->original_image_width = cinfo->image_width;
+}
+
+LOCAL(void)
+per_scan_setup (j_decompress_ptr cinfo)
+/* Do computations that are needed before processing a JPEG scan */
+/* cinfo->comps_in_scan and cinfo->cur_comp_info[] were set from SOS marker */
+{
+  int ci, mcublks, tmp;
+  jpeg_component_info *compptr;
+  
+  if (cinfo->comps_in_scan == 1) {
+    
+    /* Noninterleaved (single-component) scan */
+    compptr = cinfo->cur_comp_info[0];
+    
+    /* Overall image size in MCUs */
+    cinfo->MCUs_per_row = compptr->width_in_blocks;
+    cinfo->MCU_rows_in_scan = compptr->height_in_blocks;
+    
+    /* For noninterleaved scan, always one block per MCU */
+    compptr->MCU_width = 1;
+    compptr->MCU_height = 1;
+    compptr->MCU_blocks = 1;
+    compptr->MCU_sample_width = compptr->DCT_scaled_size;
+    compptr->last_col_width = 1;
+    /* For noninterleaved scans, it is convenient to define last_row_height
+     * as the number of block rows present in the last iMCU row.
+     */
+    tmp = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
+    if (tmp == 0) tmp = compptr->v_samp_factor;
+    compptr->last_row_height = tmp;
+    
+    /* Prepare array describing MCU composition */
+    cinfo->blocks_in_MCU = 1;
+    cinfo->MCU_membership[0] = 0;
+    
+  } else {
+    
+    /* Interleaved (multi-component) scan */
+    if (cinfo->comps_in_scan <= 0 || cinfo->comps_in_scan > MAX_COMPS_IN_SCAN)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->comps_in_scan,
+	       MAX_COMPS_IN_SCAN);
+    
+    /* Overall image size in MCUs */
+    cinfo->MCUs_per_row = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width,
+		    (long) (cinfo->max_h_samp_factor*DCTSIZE));
+    cinfo->MCU_rows_in_scan = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height,
+		    (long) (cinfo->max_v_samp_factor*DCTSIZE));
+    
+    cinfo->blocks_in_MCU = 0;
+    
+    for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+      compptr = cinfo->cur_comp_info[ci];
+      /* Sampling factors give # of blocks of component in each MCU */
+      compptr->MCU_width = compptr->h_samp_factor;
+      compptr->MCU_height = compptr->v_samp_factor;
+      compptr->MCU_blocks = compptr->MCU_width * compptr->MCU_height;
+      compptr->MCU_sample_width = compptr->MCU_width * compptr->DCT_scaled_size;
+      /* Figure number of non-dummy blocks in last MCU column & row */
+      tmp = (int) (compptr->width_in_blocks % compptr->MCU_width);
+      if (tmp == 0) tmp = compptr->MCU_width;
+      compptr->last_col_width = tmp;
+#ifdef ANDROID_TILE_BASED_DECODE
+      if (cinfo->tile_decode) {
+        tmp = (int) (jdiv_round_up(cinfo->image_width, 8)
+                % compptr->MCU_width);
+        if (tmp == 0) tmp = compptr->MCU_width;
+        compptr->last_col_width = tmp;
+      }
+#endif
+
+      tmp = (int) (compptr->height_in_blocks % compptr->MCU_height);
+      if (tmp == 0) tmp = compptr->MCU_height;
+      compptr->last_row_height = tmp;
+      /* Prepare array describing MCU composition */
+      mcublks = compptr->MCU_blocks;
+      if (cinfo->blocks_in_MCU + mcublks > D_MAX_BLOCKS_IN_MCU)
+	ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
+      while (mcublks-- > 0) {
+	cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci;
+      }
+    }
+    
+  }
+}
+
+GLOBAL(void)
+jpeg_decompress_per_scan_setup(j_decompress_ptr cinfo)
+{
+    per_scan_setup(cinfo);
+}
+
+
+
+/*
+ * Save away a copy of the Q-table referenced by each component present
+ * in the current scan, unless already saved during a prior scan.
+ *
+ * In a multiple-scan JPEG file, the encoder could assign different components
+ * the same Q-table slot number, but change table definitions between scans
+ * so that each component uses a different Q-table.  (The IJG encoder is not
+ * currently capable of doing this, but other encoders might.)  Since we want
+ * to be able to dequantize all the components at the end of the file, this
+ * means that we have to save away the table actually used for each component.
+ * We do this by copying the table at the start of the first scan containing
+ * the component.
+ * The JPEG spec prohibits the encoder from changing the contents of a Q-table
+ * slot between scans of a component using that slot.  If the encoder does so
+ * anyway, this decoder will simply use the Q-table values that were current
+ * at the start of the first scan for the component.
+ *
+ * The decompressor output side looks only at the saved quant tables,
+ * not at the current Q-table slots.
+ */
+
+LOCAL(void)
+latch_quant_tables (j_decompress_ptr cinfo)
+{
+  int ci, qtblno;
+  jpeg_component_info *compptr;
+  JQUANT_TBL * qtbl;
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    /* No work if we already saved Q-table for this component */
+    if (compptr->quant_table != NULL)
+      continue;
+    /* Make sure specified quantization table is present */
+    qtblno = compptr->quant_tbl_no;
+    if (qtblno < 0 || qtblno >= NUM_QUANT_TBLS ||
+	cinfo->quant_tbl_ptrs[qtblno] == NULL)
+      ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, qtblno);
+    /* OK, save away the quantization table */
+    qtbl = (JQUANT_TBL *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(JQUANT_TBL));
+    MEMCOPY(qtbl, cinfo->quant_tbl_ptrs[qtblno], SIZEOF(JQUANT_TBL));
+    compptr->quant_table = qtbl;
+  }
+}
+
+
+/*
+ * Initialize the input modules to read a scan of compressed data.
+ * The first call to this is done by jdmaster.c after initializing
+ * the entire decompressor (during jpeg_start_decompress).
+ * Subsequent calls come from consume_markers, below.
+ */
+
+METHODDEF(void)
+start_input_pass (j_decompress_ptr cinfo)
+{
+  per_scan_setup(cinfo);
+  latch_quant_tables(cinfo);
+  (*cinfo->entropy->start_pass) (cinfo);
+  (*cinfo->coef->start_input_pass) (cinfo);
+  cinfo->inputctl->consume_input = cinfo->coef->consume_data;
+  cinfo->inputctl->consume_input_build_huffman_index =
+        cinfo->coef->consume_data_build_huffman_index;
+}
+
+
+/*
+ * Finish up after inputting a compressed-data scan.
+ * This is called by the coefficient controller after it's read all
+ * the expected data of the scan.
+ */
+
+METHODDEF(void)
+finish_input_pass (j_decompress_ptr cinfo)
+{
+  cinfo->inputctl->consume_input = consume_markers;
+  cinfo->inputctl->consume_input_build_huffman_index =
+        consume_markers_with_huffman_index;
+}
+
+
+METHODDEF(int)
+consume_markers_with_huffman_index (j_decompress_ptr cinfo,
+        huffman_index *index, int current_scan)
+{
+    return consume_markers(cinfo);
+}
+/*
+ * Read JPEG markers before, between, or after compressed-data scans.
+ * Change state as necessary when a new scan is reached.
+ * Return value is JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI.
+ *
+ * The consume_input method pointer points either here or to the
+ * coefficient controller's consume_data routine, depending on whether
+ * we are reading a compressed data segment or inter-segment markers.
+ */
+
+METHODDEF(int)
+consume_markers (j_decompress_ptr cinfo)
+{
+  my_inputctl_ptr inputctl = (my_inputctl_ptr) cinfo->inputctl;
+  int val;
+
+  if (inputctl->pub.eoi_reached) /* After hitting EOI, read no further */
+    return JPEG_REACHED_EOI;
+
+  val = (*cinfo->marker->read_markers) (cinfo);
+
+  switch (val) {
+  case JPEG_REACHED_SOS:	/* Found SOS */
+    if (inputctl->inheaders) {	/* 1st SOS */
+      initial_setup(cinfo);
+      inputctl->inheaders = FALSE;
+      /* Note: start_input_pass must be called by jdmaster.c
+       * before any more input can be consumed.  jdapimin.c is
+       * responsible for enforcing this sequencing.
+       */
+    } else {			/* 2nd or later SOS marker */
+      if (! inputctl->pub.has_multiple_scans)
+	ERREXIT(cinfo, JERR_EOI_EXPECTED); /* Oops, I wasn't expecting this! */
+      start_input_pass(cinfo);
+    }
+    break;
+  case JPEG_REACHED_EOI:	/* Found EOI */
+    inputctl->pub.eoi_reached = TRUE;
+    if (inputctl->inheaders) {	/* Tables-only datastream, apparently */
+      if (cinfo->marker->saw_SOF)
+	ERREXIT(cinfo, JERR_SOF_NO_SOS);
+    } else {
+      /* Prevent infinite loop in coef ctlr's decompress_data routine
+       * if user set output_scan_number larger than number of scans.
+       */
+      if (cinfo->output_scan_number > cinfo->input_scan_number)
+	cinfo->output_scan_number = cinfo->input_scan_number;
+    }
+    break;
+  case JPEG_SUSPENDED:
+    break;
+  }
+
+  return val;
+}
+
+
+/*
+ * Reset state to begin a fresh datastream.
+ */
+
+METHODDEF(void)
+reset_input_controller (j_decompress_ptr cinfo)
+{
+  my_inputctl_ptr inputctl = (my_inputctl_ptr) cinfo->inputctl;
+
+  inputctl->pub.consume_input = consume_markers;
+  inputctl->pub.consume_input_build_huffman_index =
+        consume_markers_with_huffman_index;
+  inputctl->pub.has_multiple_scans = FALSE; /* "unknown" would be better */
+  inputctl->pub.eoi_reached = FALSE;
+  inputctl->inheaders = TRUE;
+  /* Reset other modules */
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr) cinfo);
+  (*cinfo->marker->reset_marker_reader) (cinfo);
+  /* Reset progression state -- would be cleaner if entropy decoder did this */
+  cinfo->coef_bits = NULL;
+}
+
+
+/*
+ * Initialize the input controller module.
+ * This is called only once, when the decompression object is created.
+ */
+
+GLOBAL(void)
+jinit_input_controller (j_decompress_ptr cinfo)
+{
+  my_inputctl_ptr inputctl;
+
+  /* Create subobject in permanent pool */
+  inputctl = (my_inputctl_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				SIZEOF(my_input_controller));
+  cinfo->inputctl = (struct jpeg_input_controller *) inputctl;
+  /* Initialize method pointers */
+  inputctl->pub.consume_input = consume_markers;
+  inputctl->pub.reset_input_controller = reset_input_controller;
+  inputctl->pub.start_input_pass = start_input_pass;
+  inputctl->pub.finish_input_pass = finish_input_pass;
+
+  inputctl->pub.consume_markers = consume_markers_with_huffman_index;
+  inputctl->pub.consume_input_build_huffman_index =
+        consume_markers_with_huffman_index;
+  /* Initialize state: can't use reset_input_controller since we don't
+   * want to try to reset other modules yet.
+   */
+  inputctl->pub.has_multiple_scans = FALSE; /* "unknown" would be better */
+  inputctl->pub.eoi_reached = FALSE;
+  inputctl->inheaders = TRUE;
+}
diff --git a/libjpegtwrp/jdmainct.c b/libjpegtwrp/jdmainct.c
new file mode 100644
index 000000000..c27217e55
--- /dev/null
+++ b/libjpegtwrp/jdmainct.c
@@ -0,0 +1,512 @@
+/*
+ * jdmainct.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the main buffer controller for decompression.
+ * The main buffer lies between the JPEG decompressor proper and the
+ * post-processor; it holds downsampled data in the JPEG colorspace.
+ *
+ * Note that this code is bypassed in raw-data mode, since the application
+ * supplies the equivalent of the main buffer in that case.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * In the current system design, the main buffer need never be a full-image
+ * buffer; any full-height buffers will be found inside the coefficient or
+ * postprocessing controllers.  Nonetheless, the main controller is not
+ * trivial.  Its responsibility is to provide context rows for upsampling/
+ * rescaling, and doing this in an efficient fashion is a bit tricky.
+ *
+ * Postprocessor input data is counted in "row groups".  A row group
+ * is defined to be (v_samp_factor * DCT_scaled_size / min_DCT_scaled_size)
+ * sample rows of each component.  (We require DCT_scaled_size values to be
+ * chosen such that these numbers are integers.  In practice DCT_scaled_size
+ * values will likely be powers of two, so we actually have the stronger
+ * condition that DCT_scaled_size / min_DCT_scaled_size is an integer.)
+ * Upsampling will typically produce max_v_samp_factor pixel rows from each
+ * row group (times any additional scale factor that the upsampler is
+ * applying).
+ *
+ * The coefficient controller will deliver data to us one iMCU row at a time;
+ * each iMCU row contains v_samp_factor * DCT_scaled_size sample rows, or
+ * exactly min_DCT_scaled_size row groups.  (This amount of data corresponds
+ * to one row of MCUs when the image is fully interleaved.)  Note that the
+ * number of sample rows varies across components, but the number of row
+ * groups does not.  Some garbage sample rows may be included in the last iMCU
+ * row at the bottom of the image.
+ *
+ * Depending on the vertical scaling algorithm used, the upsampler may need
+ * access to the sample row(s) above and below its current input row group.
+ * The upsampler is required to set need_context_rows TRUE at global selection
+ * time if so.  When need_context_rows is FALSE, this controller can simply
+ * obtain one iMCU row at a time from the coefficient controller and dole it
+ * out as row groups to the postprocessor.
+ *
+ * When need_context_rows is TRUE, this controller guarantees that the buffer
+ * passed to postprocessing contains at least one row group's worth of samples
+ * above and below the row group(s) being processed.  Note that the context
+ * rows "above" the first passed row group appear at negative row offsets in
+ * the passed buffer.  At the top and bottom of the image, the required
+ * context rows are manufactured by duplicating the first or last real sample
+ * row; this avoids having special cases in the upsampling inner loops.
+ *
+ * The amount of context is fixed at one row group just because that's a
+ * convenient number for this controller to work with.  The existing
+ * upsamplers really only need one sample row of context.  An upsampler
+ * supporting arbitrary output rescaling might wish for more than one row
+ * group of context when shrinking the image; tough, we don't handle that.
+ * (This is justified by the assumption that downsizing will be handled mostly
+ * by adjusting the DCT_scaled_size values, so that the actual scale factor at
+ * the upsample step needn't be much less than one.)
+ *
+ * To provide the desired context, we have to retain the last two row groups
+ * of one iMCU row while reading in the next iMCU row.  (The last row group
+ * can't be processed until we have another row group for its below-context,
+ * and so we have to save the next-to-last group too for its above-context.)
+ * We could do this most simply by copying data around in our buffer, but
+ * that'd be very slow.  We can avoid copying any data by creating a rather
+ * strange pointer structure.  Here's how it works.  We allocate a workspace
+ * consisting of M+2 row groups (where M = min_DCT_scaled_size is the number
+ * of row groups per iMCU row).  We create two sets of redundant pointers to
+ * the workspace.  Labeling the physical row groups 0 to M+1, the synthesized
+ * pointer lists look like this:
+ *                   M+1                          M-1
+ * master pointer --> 0         master pointer --> 0
+ *                    1                            1
+ *                   ...                          ...
+ *                   M-3                          M-3
+ *                   M-2                           M
+ *                   M-1                          M+1
+ *                    M                           M-2
+ *                   M+1                          M-1
+ *                    0                            0
+ * We read alternate iMCU rows using each master pointer; thus the last two
+ * row groups of the previous iMCU row remain un-overwritten in the workspace.
+ * The pointer lists are set up so that the required context rows appear to
+ * be adjacent to the proper places when we pass the pointer lists to the
+ * upsampler.
+ *
+ * The above pictures describe the normal state of the pointer lists.
+ * At top and bottom of the image, we diddle the pointer lists to duplicate
+ * the first or last sample row as necessary (this is cheaper than copying
+ * sample rows around).
+ *
+ * This scheme breaks down if M < 2, ie, min_DCT_scaled_size is 1.  In that
+ * situation each iMCU row provides only one row group so the buffering logic
+ * must be different (eg, we must read two iMCU rows before we can emit the
+ * first row group).  For now, we simply do not support providing context
+ * rows when min_DCT_scaled_size is 1.  That combination seems unlikely to
+ * be worth providing --- if someone wants a 1/8th-size preview, they probably
+ * want it quick and dirty, so a context-free upsampler is sufficient.
+ */
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_main_controller pub; /* public fields */
+
+  /* Pointer to allocated workspace (M or M+2 row groups). */
+  JSAMPARRAY buffer[MAX_COMPONENTS];
+
+  boolean buffer_full;		/* Have we gotten an iMCU row from decoder? */
+  JDIMENSION rowgroup_ctr;	/* counts row groups output to postprocessor */
+
+  /* Remaining fields are only used in the context case. */
+
+  /* These are the master pointers to the funny-order pointer lists. */
+  JSAMPIMAGE xbuffer[2];	/* pointers to weird pointer lists */
+
+  int whichptr;			/* indicates which pointer set is now in use */
+  int context_state;		/* process_data state machine status */
+  JDIMENSION rowgroups_avail;	/* row groups available to postprocessor */
+  JDIMENSION iMCU_row_ctr;	/* counts iMCU rows to detect image top/bot */
+} my_main_controller;
+
+typedef my_main_controller * my_main_ptr;
+
+/* context_state values: */
+#define CTX_PREPARE_FOR_IMCU	0	/* need to prepare for MCU row */
+#define CTX_PROCESS_IMCU	1	/* feeding iMCU to postprocessor */
+#define CTX_POSTPONED_ROW	2	/* feeding postponed row group */
+
+
+/* Forward declarations */
+METHODDEF(void) process_data_simple_main
+	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+	     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
+METHODDEF(void) process_data_context_main
+	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+	     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
+#ifdef QUANT_2PASS_SUPPORTED
+METHODDEF(void) process_data_crank_post
+	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+	     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
+#endif
+
+
+LOCAL(void)
+alloc_funny_pointers (j_decompress_ptr cinfo)
+/* Allocate space for the funny pointer lists.
+ * This is done only once, not once per pass.
+ */
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+  int ci, rgroup;
+  int M = cinfo->min_DCT_scaled_size;
+  jpeg_component_info *compptr;
+  JSAMPARRAY xbuf;
+
+  /* Get top-level space for component array pointers.
+   * We alloc both arrays with one call to save a few cycles.
+   */
+  main_ptr->xbuffer[0] = (JSAMPIMAGE)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				cinfo->num_components * 2 * SIZEOF(JSAMPARRAY));
+  main_ptr->xbuffer[1] = main_ptr->xbuffer[0] + cinfo->num_components;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->DCT_scaled_size) /
+      cinfo->min_DCT_scaled_size; /* height of a row group of component */
+    /* Get space for pointer lists --- M+4 row groups in each list.
+     * We alloc both pointer lists with one call to save a few cycles.
+     */
+    xbuf = (JSAMPARRAY)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  2 * (rgroup * (M + 4)) * SIZEOF(JSAMPROW));
+    xbuf += rgroup;		/* want one row group at negative offsets */
+    main_ptr->xbuffer[0][ci] = xbuf;
+    xbuf += rgroup * (M + 4);
+    main_ptr->xbuffer[1][ci] = xbuf;
+  }
+}
+
+
+LOCAL(void)
+make_funny_pointers (j_decompress_ptr cinfo)
+/* Create the funny pointer lists discussed in the comments above.
+ * The actual workspace is already allocated (in main->buffer),
+ * and the space for the pointer lists is allocated too.
+ * This routine just fills in the curiously ordered lists.
+ * This will be repeated at the beginning of each pass.
+ */
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+  int ci, i, rgroup;
+  int M = cinfo->min_DCT_scaled_size;
+  jpeg_component_info *compptr;
+  JSAMPARRAY buf, xbuf0, xbuf1;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->DCT_scaled_size) /
+      cinfo->min_DCT_scaled_size; /* height of a row group of component */
+    xbuf0 = main_ptr->xbuffer[0][ci];
+    xbuf1 = main_ptr->xbuffer[1][ci];
+    /* First copy the workspace pointers as-is */
+    buf = main_ptr->buffer[ci];
+    for (i = 0; i < rgroup * (M + 2); i++) {
+      xbuf0[i] = xbuf1[i] = buf[i];
+    }
+    /* In the second list, put the last four row groups in swapped order */
+    for (i = 0; i < rgroup * 2; i++) {
+      xbuf1[rgroup*(M-2) + i] = buf[rgroup*M + i];
+      xbuf1[rgroup*M + i] = buf[rgroup*(M-2) + i];
+    }
+    /* The wraparound pointers at top and bottom will be filled later
+     * (see set_wraparound_pointers, below).  Initially we want the "above"
+     * pointers to duplicate the first actual data line.  This only needs
+     * to happen in xbuffer[0].
+     */
+    for (i = 0; i < rgroup; i++) {
+      xbuf0[i - rgroup] = xbuf0[0];
+    }
+  }
+}
+
+
+LOCAL(void)
+set_wraparound_pointers (j_decompress_ptr cinfo)
+/* Set up the "wraparound" pointers at top and bottom of the pointer lists.
+ * This changes the pointer list state from top-of-image to the normal state.
+ */
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+  int ci, i, rgroup;
+  int M = cinfo->min_DCT_scaled_size;
+  jpeg_component_info *compptr;
+  JSAMPARRAY xbuf0, xbuf1;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->DCT_scaled_size) /
+      cinfo->min_DCT_scaled_size; /* height of a row group of component */
+    xbuf0 = main_ptr->xbuffer[0][ci];
+    xbuf1 = main_ptr->xbuffer[1][ci];
+    for (i = 0; i < rgroup; i++) {
+      xbuf0[i - rgroup] = xbuf0[rgroup*(M+1) + i];
+      xbuf1[i - rgroup] = xbuf1[rgroup*(M+1) + i];
+      xbuf0[rgroup*(M+2) + i] = xbuf0[i];
+      xbuf1[rgroup*(M+2) + i] = xbuf1[i];
+    }
+  }
+}
+
+
+LOCAL(void)
+set_bottom_pointers (j_decompress_ptr cinfo)
+/* Change the pointer lists to duplicate the last sample row at the bottom
+ * of the image.  whichptr indicates which xbuffer holds the final iMCU row.
+ * Also sets rowgroups_avail to indicate number of nondummy row groups in row.
+ */
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+  int ci, i, rgroup, iMCUheight, rows_left;
+  jpeg_component_info *compptr;
+  JSAMPARRAY xbuf;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Count sample rows in one iMCU row and in one row group */
+    iMCUheight = compptr->v_samp_factor * compptr->DCT_scaled_size;
+    rgroup = iMCUheight / cinfo->min_DCT_scaled_size;
+    /* Count nondummy sample rows remaining for this component */
+    rows_left = (int) (compptr->downsampled_height % (JDIMENSION) iMCUheight);
+    if (rows_left == 0) rows_left = iMCUheight;
+    /* Count nondummy row groups.  Should get same answer for each component,
+     * so we need only do it once.
+     */
+    if (ci == 0) {
+      main_ptr->rowgroups_avail = (JDIMENSION) ((rows_left-1) / rgroup + 1);
+    }
+    /* Duplicate the last real sample row rgroup*2 times; this pads out the
+     * last partial rowgroup and ensures at least one full rowgroup of context.
+     */
+    xbuf = main_ptr->xbuffer[main_ptr->whichptr][ci];
+    for (i = 0; i < rgroup * 2; i++) {
+      xbuf[rows_left + i] = xbuf[rows_left-1];
+    }
+  }
+}
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_main (j_decompress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+
+  switch (pass_mode) {
+  case JBUF_PASS_THRU:
+    if (cinfo->upsample->need_context_rows) {
+      main_ptr->pub.process_data = process_data_context_main;
+      make_funny_pointers(cinfo); /* Create the xbuffer[] lists */
+      main_ptr->whichptr = 0;	/* Read first iMCU row into xbuffer[0] */
+      main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
+      main_ptr->iMCU_row_ctr = 0;
+    } else {
+      /* Simple case with no context needed */
+      main_ptr->pub.process_data = process_data_simple_main;
+    }
+    main_ptr->buffer_full = FALSE;	/* Mark buffer empty */
+    main_ptr->rowgroup_ctr = 0;
+    break;
+#ifdef QUANT_2PASS_SUPPORTED
+  case JBUF_CRANK_DEST:
+    /* For last pass of 2-pass quantization, just crank the postprocessor */
+    main_ptr->pub.process_data = process_data_crank_post;
+    break;
+#endif
+  default:
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    break;
+  }
+}
+
+
+/*
+ * Process some data.
+ * This handles the simple case where no context is required.
+ */
+
+METHODDEF(void)
+process_data_simple_main (j_decompress_ptr cinfo,
+			  JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+			  JDIMENSION out_rows_avail)
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+  JDIMENSION rowgroups_avail;
+
+  /* Read input data if we haven't filled the main buffer yet */
+  if (! main_ptr->buffer_full) {
+    if (! (*cinfo->coef->decompress_data) (cinfo, main_ptr->buffer))
+      return;			/* suspension forced, can do nothing more */
+    main_ptr->buffer_full = TRUE;	/* OK, we have an iMCU row to work with */
+  }
+
+  /* There are always min_DCT_scaled_size row groups in an iMCU row. */
+  rowgroups_avail = (JDIMENSION) cinfo->min_DCT_scaled_size;
+  /* Note: at the bottom of the image, we may pass extra garbage row groups
+   * to the postprocessor.  The postprocessor has to check for bottom
+   * of image anyway (at row resolution), so no point in us doing it too.
+   */
+
+  /* Feed the postprocessor */
+  (*cinfo->post->post_process_data) (cinfo, main_ptr->buffer,
+				     &main_ptr->rowgroup_ctr, rowgroups_avail,
+				     output_buf, out_row_ctr, out_rows_avail);
+
+  /* Has postprocessor consumed all the data yet? If so, mark buffer empty */
+  if (main_ptr->rowgroup_ctr >= rowgroups_avail) {
+    main_ptr->buffer_full = FALSE;
+    main_ptr->rowgroup_ctr = 0;
+  }
+}
+
+
+/*
+ * Process some data.
+ * This handles the case where context rows must be provided.
+ */
+
+METHODDEF(void)
+process_data_context_main (j_decompress_ptr cinfo,
+			   JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+			   JDIMENSION out_rows_avail)
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+
+  /* Read input data if we haven't filled the main buffer yet */
+  if (! main_ptr->buffer_full) {
+    if (! (*cinfo->coef->decompress_data) (cinfo,
+					   main_ptr->xbuffer[main_ptr->whichptr]))
+      return;			/* suspension forced, can do nothing more */
+    main_ptr->buffer_full = TRUE;	/* OK, we have an iMCU row to work with */
+    main_ptr->iMCU_row_ctr++;	/* count rows received */
+  }
+
+  /* Postprocessor typically will not swallow all the input data it is handed
+   * in one call (due to filling the output buffer first).  Must be prepared
+   * to exit and restart.  This switch lets us keep track of how far we got.
+   * Note that each case falls through to the next on successful completion.
+   */
+  switch (main_ptr->context_state) {
+  case CTX_POSTPONED_ROW:
+    /* Call postprocessor using previously set pointers for postponed row */
+    (*cinfo->post->post_process_data) (cinfo, main_ptr->xbuffer[main_ptr->whichptr],
+			&main_ptr->rowgroup_ctr, main_ptr->rowgroups_avail,
+			output_buf, out_row_ctr, out_rows_avail);
+    if (main_ptr->rowgroup_ctr < main_ptr->rowgroups_avail)
+      return;			/* Need to suspend */
+    main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
+    if (*out_row_ctr >= out_rows_avail)
+      return;			/* Postprocessor exactly filled output buf */
+    /*FALLTHROUGH*/
+  case CTX_PREPARE_FOR_IMCU:
+    /* Prepare to process first M-1 row groups of this iMCU row */
+    main_ptr->rowgroup_ctr = 0;
+    main_ptr->rowgroups_avail = (JDIMENSION) (cinfo->min_DCT_scaled_size - 1);
+    /* Check for bottom of image: if so, tweak pointers to "duplicate"
+     * the last sample row, and adjust rowgroups_avail to ignore padding rows.
+     */
+    if (main_ptr->iMCU_row_ctr == cinfo->total_iMCU_rows)
+      set_bottom_pointers(cinfo);
+    main_ptr->context_state = CTX_PROCESS_IMCU;
+    /*FALLTHROUGH*/
+  case CTX_PROCESS_IMCU:
+    /* Call postprocessor using previously set pointers */
+    (*cinfo->post->post_process_data) (cinfo, main_ptr->xbuffer[main_ptr->whichptr],
+			&main_ptr->rowgroup_ctr, main_ptr->rowgroups_avail,
+			output_buf, out_row_ctr, out_rows_avail);
+    if (main_ptr->rowgroup_ctr < main_ptr->rowgroups_avail)
+      return;			/* Need to suspend */
+    /* After the first iMCU, change wraparound pointers to normal state */
+    if (main_ptr->iMCU_row_ctr == 1)
+      set_wraparound_pointers(cinfo);
+    /* Prepare to load new iMCU row using other xbuffer list */
+    main_ptr->whichptr ^= 1;	/* 0=>1 or 1=>0 */
+    main_ptr->buffer_full = FALSE;
+    /* Still need to process last row group of this iMCU row, */
+    /* which is saved at index M+1 of the other xbuffer */
+    main_ptr->rowgroup_ctr = (JDIMENSION) (cinfo->min_DCT_scaled_size + 1);
+    main_ptr->rowgroups_avail = (JDIMENSION) (cinfo->min_DCT_scaled_size + 2);
+    main_ptr->context_state = CTX_POSTPONED_ROW;
+  }
+}
+
+
+/*
+ * Process some data.
+ * Final pass of two-pass quantization: just call the postprocessor.
+ * Source data will be the postprocessor controller's internal buffer.
+ */
+
+#ifdef QUANT_2PASS_SUPPORTED
+
+METHODDEF(void)
+process_data_crank_post (j_decompress_ptr cinfo,
+			 JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+			 JDIMENSION out_rows_avail)
+{
+  (*cinfo->post->post_process_data) (cinfo, (JSAMPIMAGE) NULL,
+				     (JDIMENSION *) NULL, (JDIMENSION) 0,
+				     output_buf, out_row_ctr, out_rows_avail);
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */
+
+
+/*
+ * Initialize main buffer controller.
+ */
+
+GLOBAL(void)
+jinit_d_main_controller (j_decompress_ptr cinfo, boolean need_full_buffer)
+{
+  my_main_ptr main_ptr;
+  int ci, rgroup, ngroups;
+  jpeg_component_info *compptr;
+
+  main_ptr = (my_main_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_main_controller));
+  cinfo->main = (struct jpeg_d_main_controller *) main_ptr;
+  main_ptr->pub.start_pass = start_pass_main;
+
+  if (need_full_buffer)		/* shouldn't happen */
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  /* Allocate the workspace.
+   * ngroups is the number of row groups we need.
+   */
+  if (cinfo->upsample->need_context_rows) {
+    if (cinfo->min_DCT_scaled_size < 2) /* unsupported, see comments above */
+      ERREXIT(cinfo, JERR_NOTIMPL);
+    alloc_funny_pointers(cinfo); /* Alloc space for xbuffer[] lists */
+    ngroups = cinfo->min_DCT_scaled_size + 2;
+  } else {
+    ngroups = cinfo->min_DCT_scaled_size;
+  }
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->DCT_scaled_size) /
+      cinfo->min_DCT_scaled_size; /* height of a row group of component */
+    main_ptr->buffer[ci] = (*cinfo->mem->alloc_sarray)
+			((j_common_ptr) cinfo, JPOOL_IMAGE,
+			 compptr->width_in_blocks * compptr->DCT_scaled_size,
+			 (JDIMENSION) (rgroup * ngroups));
+  }
+}
diff --git a/libjpegtwrp/jdmarker.c b/libjpegtwrp/jdmarker.c
new file mode 100644
index 000000000..733294050
--- /dev/null
+++ b/libjpegtwrp/jdmarker.c
@@ -0,0 +1,1397 @@
+/*
+ * jdmarker.c
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to decode JPEG datastream markers.
+ * Most of the complexity arises from our desire to support input
+ * suspension: if not all of the data for a marker is available,
+ * we must exit back to the application.  On resumption, we reprocess
+ * the marker.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+typedef enum {			/* JPEG marker codes */
+  M_SOF0  = 0xc0,
+  M_SOF1  = 0xc1,
+  M_SOF2  = 0xc2,
+  M_SOF3  = 0xc3,
+  
+  M_SOF5  = 0xc5,
+  M_SOF6  = 0xc6,
+  M_SOF7  = 0xc7,
+  
+  M_JPG   = 0xc8,
+  M_SOF9  = 0xc9,
+  M_SOF10 = 0xca,
+  M_SOF11 = 0xcb,
+  
+  M_SOF13 = 0xcd,
+  M_SOF14 = 0xce,
+  M_SOF15 = 0xcf,
+  
+  M_DHT   = 0xc4,
+  
+  M_DAC   = 0xcc,
+  
+  M_RST0  = 0xd0,
+  M_RST1  = 0xd1,
+  M_RST2  = 0xd2,
+  M_RST3  = 0xd3,
+  M_RST4  = 0xd4,
+  M_RST5  = 0xd5,
+  M_RST6  = 0xd6,
+  M_RST7  = 0xd7,
+  
+  M_SOI   = 0xd8,
+  M_EOI   = 0xd9,
+  M_SOS   = 0xda,
+  M_DQT   = 0xdb,
+  M_DNL   = 0xdc,
+  M_DRI   = 0xdd,
+  M_DHP   = 0xde,
+  M_EXP   = 0xdf,
+  
+  M_APP0  = 0xe0,
+  M_APP1  = 0xe1,
+  M_APP2  = 0xe2,
+  M_APP3  = 0xe3,
+  M_APP4  = 0xe4,
+  M_APP5  = 0xe5,
+  M_APP6  = 0xe6,
+  M_APP7  = 0xe7,
+  M_APP8  = 0xe8,
+  M_APP9  = 0xe9,
+  M_APP10 = 0xea,
+  M_APP11 = 0xeb,
+  M_APP12 = 0xec,
+  M_APP13 = 0xed,
+  M_APP14 = 0xee,
+  M_APP15 = 0xef,
+  
+  M_JPG0  = 0xf0,
+  M_JPG13 = 0xfd,
+  M_COM   = 0xfe,
+  
+  M_TEM   = 0x01,
+  
+  M_ERROR = 0x100
+} JPEG_MARKER;
+
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_marker_reader pub; /* public fields */
+
+  /* Application-overridable marker processing methods */
+  jpeg_marker_parser_method process_COM;
+  jpeg_marker_parser_method process_APPn[16];
+
+  /* Limit on marker data length to save for each marker type */
+  unsigned int length_limit_COM;
+  unsigned int length_limit_APPn[16];
+
+  /* Status of COM/APPn marker saving */
+  jpeg_saved_marker_ptr cur_marker;	/* NULL if not processing a marker */
+  unsigned int bytes_read;		/* data bytes read so far in marker */
+  /* Note: cur_marker is not linked into marker_list until it's all read. */
+} my_marker_reader;
+
+typedef my_marker_reader * my_marker_ptr;
+
+
+/*
+ * Macros for fetching data from the data source module.
+ *
+ * At all times, cinfo->src->next_input_byte and ->bytes_in_buffer reflect
+ * the current restart point; we update them only when we have reached a
+ * suitable place to restart if a suspension occurs.
+ */
+
+/* Declare and initialize local copies of input pointer/count */
+#define INPUT_VARS(cinfo)  \
+	struct jpeg_source_mgr * datasrc = (cinfo)->src;  \
+	const JOCTET * next_input_byte = datasrc->next_input_byte;  \
+	size_t bytes_in_buffer = datasrc->bytes_in_buffer
+
+/* Unload the local copies --- do this only at a restart boundary */
+#define INPUT_SYNC(cinfo)  \
+	( datasrc->next_input_byte = next_input_byte,  \
+	  datasrc->bytes_in_buffer = bytes_in_buffer )
+
+/* Reload the local copies --- used only in MAKE_BYTE_AVAIL */
+#define INPUT_RELOAD(cinfo)  \
+	( next_input_byte = datasrc->next_input_byte,  \
+	  bytes_in_buffer = datasrc->bytes_in_buffer )
+
+/* Internal macro for INPUT_BYTE and INPUT_2BYTES: make a byte available.
+ * Note we do *not* do INPUT_SYNC before calling fill_input_buffer,
+ * but we must reload the local copies after a successful fill.
+ */
+#define MAKE_BYTE_AVAIL(cinfo,action)  \
+	if (bytes_in_buffer == 0) {  \
+	  if (! (*datasrc->fill_input_buffer) (cinfo))  \
+	    { action; }  \
+	  INPUT_RELOAD(cinfo);  \
+	}
+
+/* Read a byte into variable V.
+ * If must suspend, take the specified action (typically "return FALSE").
+ */
+#define INPUT_BYTE(cinfo,V,action)  \
+	MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \
+		  bytes_in_buffer--; \
+		  V = GETJOCTET(*next_input_byte++); )
+
+/* As above, but read two bytes interpreted as an unsigned 16-bit integer.
+ * V should be declared unsigned int or perhaps INT32.
+ */
+#define INPUT_2BYTES(cinfo,V,action)  \
+	MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \
+		  bytes_in_buffer--; \
+		  V = ((unsigned int) GETJOCTET(*next_input_byte++)) << 8; \
+		  MAKE_BYTE_AVAIL(cinfo,action); \
+		  bytes_in_buffer--; \
+		  V += GETJOCTET(*next_input_byte++); )
+
+
+/*
+ * Routines to process JPEG markers.
+ *
+ * Entry condition: JPEG marker itself has been read and its code saved
+ *   in cinfo->unread_marker; input restart point is just after the marker.
+ *
+ * Exit: if return TRUE, have read and processed any parameters, and have
+ *   updated the restart point to point after the parameters.
+ *   If return FALSE, was forced to suspend before reaching end of
+ *   marker parameters; restart point has not been moved.  Same routine
+ *   will be called again after application supplies more input data.
+ *
+ * This approach to suspension assumes that all of a marker's parameters
+ * can fit into a single input bufferload.  This should hold for "normal"
+ * markers.  Some COM/APPn markers might have large parameter segments
+ * that might not fit.  If we are simply dropping such a marker, we use
+ * skip_input_data to get past it, and thereby put the problem on the
+ * source manager's shoulders.  If we are saving the marker's contents
+ * into memory, we use a slightly different convention: when forced to
+ * suspend, the marker processor updates the restart point to the end of
+ * what it's consumed (ie, the end of the buffer) before returning FALSE.
+ * On resumption, cinfo->unread_marker still contains the marker code,
+ * but the data source will point to the next chunk of marker data.
+ * The marker processor must retain internal state to deal with this.
+ *
+ * Note that we don't bother to avoid duplicate trace messages if a
+ * suspension occurs within marker parameters.  Other side effects
+ * require more care.
+ */
+
+
+LOCAL(boolean)
+get_soi (j_decompress_ptr cinfo)
+/* Process an SOI marker */
+{
+  int i;
+  
+  TRACEMS(cinfo, 1, JTRC_SOI);
+
+  if (cinfo->marker->saw_SOI)
+    ERREXIT(cinfo, JERR_SOI_DUPLICATE);
+
+  /* Reset all parameters that are defined to be reset by SOI */
+
+  for (i = 0; i < NUM_ARITH_TBLS; i++) {
+    cinfo->arith_dc_L[i] = 0;
+    cinfo->arith_dc_U[i] = 1;
+    cinfo->arith_ac_K[i] = 5;
+  }
+  cinfo->restart_interval = 0;
+
+  /* Set initial assumptions for colorspace etc */
+
+  cinfo->jpeg_color_space = JCS_UNKNOWN;
+  cinfo->CCIR601_sampling = FALSE; /* Assume non-CCIR sampling??? */
+
+  cinfo->saw_JFIF_marker = FALSE;
+  cinfo->JFIF_major_version = 1; /* set default JFIF APP0 values */
+  cinfo->JFIF_minor_version = 1;
+  cinfo->density_unit = 0;
+  cinfo->X_density = 1;
+  cinfo->Y_density = 1;
+  cinfo->saw_Adobe_marker = FALSE;
+  cinfo->Adobe_transform = 0;
+
+  cinfo->marker->saw_SOI = TRUE;
+
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+get_sof (j_decompress_ptr cinfo, boolean is_prog, boolean is_arith)
+/* Process a SOFn marker */
+{
+  INT32 length;
+  int c, ci;
+  jpeg_component_info * compptr;
+  INPUT_VARS(cinfo);
+
+  cinfo->progressive_mode = is_prog;
+  cinfo->arith_code = is_arith;
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+
+  INPUT_BYTE(cinfo, cinfo->data_precision, return FALSE);
+  INPUT_2BYTES(cinfo, cinfo->image_height, return FALSE);
+  INPUT_2BYTES(cinfo, cinfo->image_width, return FALSE);
+  INPUT_BYTE(cinfo, cinfo->num_components, return FALSE);
+
+  length -= 8;
+
+  TRACEMS4(cinfo, 1, JTRC_SOF, cinfo->unread_marker,
+	   (int) cinfo->image_width, (int) cinfo->image_height,
+	   cinfo->num_components);
+
+  if (cinfo->marker->saw_SOF)
+    ERREXIT(cinfo, JERR_SOF_DUPLICATE);
+
+  /* We don't support files in which the image height is initially specified */
+  /* as 0 and is later redefined by DNL.  As long as we have to check that,  */
+  /* might as well have a general sanity check. */
+  if (cinfo->image_height <= 0 || cinfo->image_width <= 0
+      || cinfo->num_components <= 0)
+    ERREXIT(cinfo, JERR_EMPTY_IMAGE);
+
+  if (length != (cinfo->num_components * 3))
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  if (cinfo->comp_info == NULL)	/* do only once, even if suspend */
+    cinfo->comp_info = (jpeg_component_info *) (*cinfo->mem->alloc_small)
+			((j_common_ptr) cinfo, JPOOL_IMAGE,
+			 cinfo->num_components * SIZEOF(jpeg_component_info));
+  
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    compptr->component_index = ci;
+    INPUT_BYTE(cinfo, compptr->component_id, return FALSE);
+    INPUT_BYTE(cinfo, c, return FALSE);
+    compptr->h_samp_factor = (c >> 4) & 15;
+    compptr->v_samp_factor = (c     ) & 15;
+    INPUT_BYTE(cinfo, compptr->quant_tbl_no, return FALSE);
+
+    TRACEMS4(cinfo, 1, JTRC_SOF_COMPONENT,
+	     compptr->component_id, compptr->h_samp_factor,
+	     compptr->v_samp_factor, compptr->quant_tbl_no);
+  }
+
+  cinfo->marker->saw_SOF = TRUE;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+get_sos (j_decompress_ptr cinfo)
+/* Process a SOS marker */
+{
+  INT32 length;
+  int i, ci, n, c, cc;
+  jpeg_component_info * compptr;
+  INPUT_VARS(cinfo);
+
+  if (! cinfo->marker->saw_SOF)
+    ERREXIT(cinfo, JERR_SOS_NO_SOF);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+
+  INPUT_BYTE(cinfo, n, return FALSE); /* Number of components */
+
+  TRACEMS1(cinfo, 1, JTRC_SOS, n);
+
+  if (length != (n * 2 + 6) || n < 1 || n > MAX_COMPS_IN_SCAN)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  cinfo->comps_in_scan = n;
+
+  /* Collect the component-spec parameters */
+
+  for (i = 0; i < n; i++) {
+    INPUT_BYTE(cinfo, cc, return FALSE);
+    INPUT_BYTE(cinfo, c, return FALSE);
+    
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      if (cc == compptr->component_id)
+	goto id_found;
+    }
+
+    ERREXIT1(cinfo, JERR_BAD_COMPONENT_ID, cc);
+
+  id_found:
+
+    cinfo->cur_comp_info[i] = compptr;
+    compptr->dc_tbl_no = (c >> 4) & 15;
+    compptr->ac_tbl_no = (c     ) & 15;
+    
+    TRACEMS3(cinfo, 1, JTRC_SOS_COMPONENT, cc,
+	     compptr->dc_tbl_no, compptr->ac_tbl_no);
+  }
+
+  /* Collect the additional scan parameters Ss, Se, Ah/Al. */
+  INPUT_BYTE(cinfo, c, return FALSE);
+  cinfo->Ss = c;
+  INPUT_BYTE(cinfo, c, return FALSE);
+  cinfo->Se = c;
+  INPUT_BYTE(cinfo, c, return FALSE);
+  cinfo->Ah = (c >> 4) & 15;
+  cinfo->Al = (c     ) & 15;
+
+  TRACEMS4(cinfo, 1, JTRC_SOS_PARAMS, cinfo->Ss, cinfo->Se,
+	   cinfo->Ah, cinfo->Al);
+
+  /* Prepare to scan data & restart markers */
+  cinfo->marker->next_restart_num = 0;
+
+  /* Count another SOS marker */
+  cinfo->input_scan_number++;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+#ifdef D_ARITH_CODING_SUPPORTED
+
+LOCAL(boolean)
+get_dac (j_decompress_ptr cinfo)
+/* Process a DAC marker */
+{
+  INT32 length;
+  int index, val;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+  
+  while (length > 0) {
+    INPUT_BYTE(cinfo, index, return FALSE);
+    INPUT_BYTE(cinfo, val, return FALSE);
+
+    length -= 2;
+
+    TRACEMS2(cinfo, 1, JTRC_DAC, index, val);
+
+    if (index < 0 || index >= (2*NUM_ARITH_TBLS))
+      ERREXIT1(cinfo, JERR_DAC_INDEX, index);
+
+    if (index >= NUM_ARITH_TBLS) { /* define AC table */
+      cinfo->arith_ac_K[index-NUM_ARITH_TBLS] = (UINT8) val;
+    } else {			/* define DC table */
+      cinfo->arith_dc_L[index] = (UINT8) (val & 0x0F);
+      cinfo->arith_dc_U[index] = (UINT8) (val >> 4);
+      if (cinfo->arith_dc_L[index] > cinfo->arith_dc_U[index])
+	ERREXIT1(cinfo, JERR_DAC_VALUE, val);
+    }
+  }
+
+  if (length != 0)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+#else /* ! D_ARITH_CODING_SUPPORTED */
+
+#define get_dac(cinfo)  skip_variable(cinfo)
+
+#endif /* D_ARITH_CODING_SUPPORTED */
+
+
+LOCAL(boolean)
+get_dht (j_decompress_ptr cinfo)
+/* Process a DHT marker */
+{
+  INT32 length;
+  UINT8 bits[17];
+  UINT8 huffval[256];
+  int i, index, count;
+  JHUFF_TBL **htblptr;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+  
+  while (length > 16) {
+    INPUT_BYTE(cinfo, index, return FALSE);
+
+    TRACEMS1(cinfo, 1, JTRC_DHT, index);
+      
+    bits[0] = 0;
+    count = 0;
+    for (i = 1; i <= 16; i++) {
+      INPUT_BYTE(cinfo, bits[i], return FALSE);
+      count += bits[i];
+    }
+
+    length -= 1 + 16;
+
+    TRACEMS8(cinfo, 2, JTRC_HUFFBITS,
+	     bits[1], bits[2], bits[3], bits[4],
+	     bits[5], bits[6], bits[7], bits[8]);
+    TRACEMS8(cinfo, 2, JTRC_HUFFBITS,
+	     bits[9], bits[10], bits[11], bits[12],
+	     bits[13], bits[14], bits[15], bits[16]);
+
+    /* Here we just do minimal validation of the counts to avoid walking
+     * off the end of our table space.  jdhuff.c will check more carefully.
+     */
+    if (count > 256 || ((INT32) count) > length)
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+
+    for (i = 0; i < count; i++)
+      INPUT_BYTE(cinfo, huffval[i], return FALSE);
+
+    length -= count;
+
+    if (index & 0x10) {		/* AC table definition */
+      index -= 0x10;
+      htblptr = &cinfo->ac_huff_tbl_ptrs[index];
+    } else {			/* DC table definition */
+      htblptr = &cinfo->dc_huff_tbl_ptrs[index];
+    }
+
+    if (index < 0 || index >= NUM_HUFF_TBLS)
+      ERREXIT1(cinfo, JERR_DHT_INDEX, index);
+
+    if (*htblptr == NULL)
+      *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+  
+    MEMCOPY((*htblptr)->bits, bits, SIZEOF((*htblptr)->bits));
+    MEMCOPY((*htblptr)->huffval, huffval, SIZEOF((*htblptr)->huffval));
+  }
+
+  if (length != 0)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+get_dqt (j_decompress_ptr cinfo)
+/* Process a DQT marker */
+{
+  INT32 length;
+  int n, i, prec;
+  unsigned int tmp;
+  JQUANT_TBL *quant_ptr;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+
+  while (length > 0) {
+    INPUT_BYTE(cinfo, n, return FALSE);
+    prec = n >> 4;
+    n &= 0x0F;
+
+    TRACEMS2(cinfo, 1, JTRC_DQT, n, prec);
+
+    if (n >= NUM_QUANT_TBLS)
+      ERREXIT1(cinfo, JERR_DQT_INDEX, n);
+      
+    if (cinfo->quant_tbl_ptrs[n] == NULL)
+      cinfo->quant_tbl_ptrs[n] = jpeg_alloc_quant_table((j_common_ptr) cinfo);
+    quant_ptr = cinfo->quant_tbl_ptrs[n];
+
+    for (i = 0; i < DCTSIZE2; i++) {
+      if (prec)
+	INPUT_2BYTES(cinfo, tmp, return FALSE);
+      else
+	INPUT_BYTE(cinfo, tmp, return FALSE);
+      /* We convert the zigzag-order table to natural array order. */
+      quant_ptr->quantval[jpeg_natural_order[i]] = (UINT16) tmp;
+    }
+
+    if (cinfo->err->trace_level >= 2) {
+      for (i = 0; i < DCTSIZE2; i += 8) {
+	TRACEMS8(cinfo, 2, JTRC_QUANTVALS,
+		 quant_ptr->quantval[i],   quant_ptr->quantval[i+1],
+		 quant_ptr->quantval[i+2], quant_ptr->quantval[i+3],
+		 quant_ptr->quantval[i+4], quant_ptr->quantval[i+5],
+		 quant_ptr->quantval[i+6], quant_ptr->quantval[i+7]);
+      }
+    }
+
+    length -= DCTSIZE2+1;
+    if (prec) length -= DCTSIZE2;
+  }
+
+  if (length != 0)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+get_dri (j_decompress_ptr cinfo)
+/* Process a DRI marker */
+{
+  INT32 length;
+  unsigned int tmp;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  
+  if (length != 4)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  INPUT_2BYTES(cinfo, tmp, return FALSE);
+
+  TRACEMS1(cinfo, 1, JTRC_DRI, tmp);
+
+  cinfo->restart_interval = tmp;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+/*
+ * Routines for processing APPn and COM markers.
+ * These are either saved in memory or discarded, per application request.
+ * APP0 and APP14 are specially checked to see if they are
+ * JFIF and Adobe markers, respectively.
+ */
+
+#define APP0_DATA_LEN	14	/* Length of interesting data in APP0 */
+#define APP14_DATA_LEN	12	/* Length of interesting data in APP14 */
+#define APPN_DATA_LEN	14	/* Must be the largest of the above!! */
+
+
+LOCAL(void)
+examine_app0 (j_decompress_ptr cinfo, JOCTET FAR * data,
+	      unsigned int datalen, INT32 remaining)
+/* Examine first few bytes from an APP0.
+ * Take appropriate action if it is a JFIF marker.
+ * datalen is # of bytes at data[], remaining is length of rest of marker data.
+ */
+{
+  INT32 totallen = (INT32) datalen + remaining;
+
+  if (datalen >= APP0_DATA_LEN &&
+      GETJOCTET(data[0]) == 0x4A &&
+      GETJOCTET(data[1]) == 0x46 &&
+      GETJOCTET(data[2]) == 0x49 &&
+      GETJOCTET(data[3]) == 0x46 &&
+      GETJOCTET(data[4]) == 0) {
+    /* Found JFIF APP0 marker: save info */
+    cinfo->saw_JFIF_marker = TRUE;
+    cinfo->JFIF_major_version = GETJOCTET(data[5]);
+    cinfo->JFIF_minor_version = GETJOCTET(data[6]);
+    cinfo->density_unit = GETJOCTET(data[7]);
+    cinfo->X_density = (GETJOCTET(data[8]) << 8) + GETJOCTET(data[9]);
+    cinfo->Y_density = (GETJOCTET(data[10]) << 8) + GETJOCTET(data[11]);
+    /* Check version.
+     * Major version must be 1, anything else signals an incompatible change.
+     * (We used to treat this as an error, but now it's a nonfatal warning,
+     * because some bozo at Hijaak couldn't read the spec.)
+     * Minor version should be 0..2, but process anyway if newer.
+     */
+    if (cinfo->JFIF_major_version != 1)
+      WARNMS2(cinfo, JWRN_JFIF_MAJOR,
+	      cinfo->JFIF_major_version, cinfo->JFIF_minor_version);
+    /* Generate trace messages */
+    TRACEMS5(cinfo, 1, JTRC_JFIF,
+	     cinfo->JFIF_major_version, cinfo->JFIF_minor_version,
+	     cinfo->X_density, cinfo->Y_density, cinfo->density_unit);
+    /* Validate thumbnail dimensions and issue appropriate messages */
+    if (GETJOCTET(data[12]) | GETJOCTET(data[13]))
+      TRACEMS2(cinfo, 1, JTRC_JFIF_THUMBNAIL,
+	       GETJOCTET(data[12]), GETJOCTET(data[13]));
+    totallen -= APP0_DATA_LEN;
+    if (totallen !=
+	((INT32)GETJOCTET(data[12]) * (INT32)GETJOCTET(data[13]) * (INT32) 3))
+      TRACEMS1(cinfo, 1, JTRC_JFIF_BADTHUMBNAILSIZE, (int) totallen);
+  } else if (datalen >= 6 &&
+      GETJOCTET(data[0]) == 0x4A &&
+      GETJOCTET(data[1]) == 0x46 &&
+      GETJOCTET(data[2]) == 0x58 &&
+      GETJOCTET(data[3]) == 0x58 &&
+      GETJOCTET(data[4]) == 0) {
+    /* Found JFIF "JFXX" extension APP0 marker */
+    /* The library doesn't actually do anything with these,
+     * but we try to produce a helpful trace message.
+     */
+    switch (GETJOCTET(data[5])) {
+    case 0x10:
+      TRACEMS1(cinfo, 1, JTRC_THUMB_JPEG, (int) totallen);
+      break;
+    case 0x11:
+      TRACEMS1(cinfo, 1, JTRC_THUMB_PALETTE, (int) totallen);
+      break;
+    case 0x13:
+      TRACEMS1(cinfo, 1, JTRC_THUMB_RGB, (int) totallen);
+      break;
+    default:
+      TRACEMS2(cinfo, 1, JTRC_JFIF_EXTENSION,
+	       GETJOCTET(data[5]), (int) totallen);
+      break;
+    }
+  } else {
+    /* Start of APP0 does not match "JFIF" or "JFXX", or too short */
+    TRACEMS1(cinfo, 1, JTRC_APP0, (int) totallen);
+  }
+}
+
+
+LOCAL(void)
+examine_app14 (j_decompress_ptr cinfo, JOCTET FAR * data,
+	       unsigned int datalen, INT32 remaining)
+/* Examine first few bytes from an APP14.
+ * Take appropriate action if it is an Adobe marker.
+ * datalen is # of bytes at data[], remaining is length of rest of marker data.
+ */
+{
+  unsigned int version, flags0, flags1, transform;
+
+  if (datalen >= APP14_DATA_LEN &&
+      GETJOCTET(data[0]) == 0x41 &&
+      GETJOCTET(data[1]) == 0x64 &&
+      GETJOCTET(data[2]) == 0x6F &&
+      GETJOCTET(data[3]) == 0x62 &&
+      GETJOCTET(data[4]) == 0x65) {
+    /* Found Adobe APP14 marker */
+    version = (GETJOCTET(data[5]) << 8) + GETJOCTET(data[6]);
+    flags0 = (GETJOCTET(data[7]) << 8) + GETJOCTET(data[8]);
+    flags1 = (GETJOCTET(data[9]) << 8) + GETJOCTET(data[10]);
+    transform = GETJOCTET(data[11]);
+    TRACEMS4(cinfo, 1, JTRC_ADOBE, version, flags0, flags1, transform);
+    cinfo->saw_Adobe_marker = TRUE;
+    cinfo->Adobe_transform = (UINT8) transform;
+  } else {
+    /* Start of APP14 does not match "Adobe", or too short */
+    TRACEMS1(cinfo, 1, JTRC_APP14, (int) (datalen + remaining));
+  }
+}
+
+
+METHODDEF(boolean)
+get_interesting_appn (j_decompress_ptr cinfo)
+/* Process an APP0 or APP14 marker without saving it */
+{
+  INT32 length;
+  JOCTET b[APPN_DATA_LEN];
+  unsigned int i, numtoread;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+
+  /* get the interesting part of the marker data */
+  if (length >= APPN_DATA_LEN)
+    numtoread = APPN_DATA_LEN;
+  else if (length > 0)
+    numtoread = (unsigned int) length;
+  else
+    numtoread = 0;
+  for (i = 0; i < numtoread; i++)
+    INPUT_BYTE(cinfo, b[i], return FALSE);
+  length -= numtoread;
+
+  /* process it */
+  switch (cinfo->unread_marker) {
+  case M_APP0:
+    examine_app0(cinfo, (JOCTET FAR *) b, numtoread, length);
+    break;
+  case M_APP14:
+    examine_app14(cinfo, (JOCTET FAR *) b, numtoread, length);
+    break;
+  default:
+    /* can't get here unless jpeg_save_markers chooses wrong processor */
+    ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, cinfo->unread_marker);
+    break;
+  }
+
+  /* skip any remaining data -- could be lots */
+  INPUT_SYNC(cinfo);
+  if (length > 0)
+    (*cinfo->src->skip_input_data) (cinfo, (long) length);
+
+  return TRUE;
+}
+
+
+#ifdef SAVE_MARKERS_SUPPORTED
+
+METHODDEF(boolean)
+save_marker (j_decompress_ptr cinfo)
+/* Save an APPn or COM marker into the marker list */
+{
+  my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
+  jpeg_saved_marker_ptr cur_marker = marker->cur_marker;
+  unsigned int bytes_read, data_length;
+  JOCTET FAR * data;
+  INT32 length = 0;
+  INPUT_VARS(cinfo);
+
+  if (cur_marker == NULL) {
+    /* begin reading a marker */
+    INPUT_2BYTES(cinfo, length, return FALSE);
+    length -= 2;
+    if (length >= 0) {		/* watch out for bogus length word */
+      /* figure out how much we want to save */
+      unsigned int limit;
+      if (cinfo->unread_marker == (int) M_COM)
+	limit = marker->length_limit_COM;
+      else
+	limit = marker->length_limit_APPn[cinfo->unread_marker - (int) M_APP0];
+      if ((unsigned int) length < limit)
+	limit = (unsigned int) length;
+      /* allocate and initialize the marker item */
+      cur_marker = (jpeg_saved_marker_ptr)
+	(*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				    SIZEOF(struct jpeg_marker_struct) + limit);
+      cur_marker->next = NULL;
+      cur_marker->marker = (UINT8) cinfo->unread_marker;
+      cur_marker->original_length = (unsigned int) length;
+      cur_marker->data_length = limit;
+      /* data area is just beyond the jpeg_marker_struct */
+      data = cur_marker->data = (JOCTET FAR *) (cur_marker + 1);
+      marker->cur_marker = cur_marker;
+      marker->bytes_read = 0;
+      bytes_read = 0;
+      data_length = limit;
+    } else {
+      /* deal with bogus length word */
+      bytes_read = data_length = 0;
+      data = NULL;
+    }
+  } else {
+    /* resume reading a marker */
+    bytes_read = marker->bytes_read;
+    data_length = cur_marker->data_length;
+    data = cur_marker->data + bytes_read;
+  }
+
+  while (bytes_read < data_length) {
+    INPUT_SYNC(cinfo);		/* move the restart point to here */
+    marker->bytes_read = bytes_read;
+    /* If there's not at least one byte in buffer, suspend */
+    MAKE_BYTE_AVAIL(cinfo, return FALSE);
+    /* Copy bytes with reasonable rapidity */
+    while (bytes_read < data_length && bytes_in_buffer > 0) {
+      *data++ = *next_input_byte++;
+      bytes_in_buffer--;
+      bytes_read++;
+    }
+  }
+
+  /* Done reading what we want to read */
+  if (cur_marker != NULL) {	/* will be NULL if bogus length word */
+    /* Add new marker to end of list */
+    if (cinfo->marker_list == NULL) {
+      cinfo->marker_list = cur_marker;
+    } else {
+      jpeg_saved_marker_ptr prev = cinfo->marker_list;
+      while (prev->next != NULL)
+	prev = prev->next;
+      prev->next = cur_marker;
+    }
+    /* Reset pointer & calc remaining data length */
+    data = cur_marker->data;
+    length = cur_marker->original_length - data_length;
+  }
+  /* Reset to initial state for next marker */
+  marker->cur_marker = NULL;
+
+  /* Process the marker if interesting; else just make a generic trace msg */
+  switch (cinfo->unread_marker) {
+  case M_APP0:
+    examine_app0(cinfo, data, data_length, length);
+    break;
+  case M_APP14:
+    examine_app14(cinfo, data, data_length, length);
+    break;
+  default:
+    TRACEMS2(cinfo, 1, JTRC_MISC_MARKER, cinfo->unread_marker,
+	     (int) (data_length + length));
+    break;
+  }
+
+  /* skip any remaining data -- could be lots */
+  INPUT_SYNC(cinfo);		/* do before skip_input_data */
+  if (length > 0)
+    (*cinfo->src->skip_input_data) (cinfo, (long) length);
+
+  return TRUE;
+}
+
+#endif /* SAVE_MARKERS_SUPPORTED */
+
+
+METHODDEF(boolean)
+skip_variable (j_decompress_ptr cinfo)
+/* Skip over an unknown or uninteresting variable-length marker */
+{
+  INT32 length;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+  
+  TRACEMS2(cinfo, 1, JTRC_MISC_MARKER, cinfo->unread_marker, (int) length);
+
+  INPUT_SYNC(cinfo);		/* do before skip_input_data */
+  if (length > 0)
+    (*cinfo->src->skip_input_data) (cinfo, (long) length);
+
+  return TRUE;
+}
+
+
+/*
+ * Find the next JPEG marker, save it in cinfo->unread_marker.
+ * Returns FALSE if had to suspend before reaching a marker;
+ * in that case cinfo->unread_marker is unchanged.
+ *
+ * Note that the result might not be a valid marker code,
+ * but it will never be 0 or FF.
+ */
+
+LOCAL(boolean)
+next_marker (j_decompress_ptr cinfo)
+{
+  int c;
+  INPUT_VARS(cinfo);
+
+  for (;;) {
+    INPUT_BYTE(cinfo, c, return FALSE);
+    /* Skip any non-FF bytes.
+     * This may look a bit inefficient, but it will not occur in a valid file.
+     * We sync after each discarded byte so that a suspending data source
+     * can discard the byte from its buffer.
+     */
+    while (c != 0xFF) {
+      cinfo->marker->discarded_bytes++;
+      INPUT_SYNC(cinfo);
+      INPUT_BYTE(cinfo, c, return FALSE);
+    }
+    /* This loop swallows any duplicate FF bytes.  Extra FFs are legal as
+     * pad bytes, so don't count them in discarded_bytes.  We assume there
+     * will not be so many consecutive FF bytes as to overflow a suspending
+     * data source's input buffer.
+     */
+    do {
+      INPUT_BYTE(cinfo, c, return FALSE);
+    } while (c == 0xFF);
+    if (c != 0)
+      break;			/* found a valid marker, exit loop */
+    /* Reach here if we found a stuffed-zero data sequence (FF/00).
+     * Discard it and loop back to try again.
+     */
+    cinfo->marker->discarded_bytes += 2;
+    INPUT_SYNC(cinfo);
+  }
+
+  if (cinfo->marker->discarded_bytes != 0) {
+    WARNMS2(cinfo, JWRN_EXTRANEOUS_DATA, cinfo->marker->discarded_bytes, c);
+    cinfo->marker->discarded_bytes = 0;
+  }
+
+  cinfo->unread_marker = c;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+first_marker (j_decompress_ptr cinfo)
+/* Like next_marker, but used to obtain the initial SOI marker. */
+/* For this marker, we do not allow preceding garbage or fill; otherwise,
+ * we might well scan an entire input file before realizing it ain't JPEG.
+ * If an application wants to process non-JFIF files, it must seek to the
+ * SOI before calling the JPEG library.
+ */
+{
+  int c, c2;
+  INPUT_VARS(cinfo);
+
+  INPUT_BYTE(cinfo, c, return FALSE);
+  INPUT_BYTE(cinfo, c2, return FALSE);
+  if (c != 0xFF || c2 != (int) M_SOI)
+    ERREXIT2(cinfo, JERR_NO_SOI, c, c2);
+
+  cinfo->unread_marker = c2;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+/*
+ * Read markers until SOS or EOI.
+ *
+ * Returns same codes as are defined for jpeg_consume_input:
+ * JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI.
+ */
+
+METHODDEF(int)
+read_markers (j_decompress_ptr cinfo)
+{
+  /* Outer loop repeats once for each marker. */
+  for (;;) {
+    /* Collect the marker proper, unless we already did. */
+    /* NB: first_marker() enforces the requirement that SOI appear first. */
+    if (cinfo->unread_marker == 0) {
+      if (! cinfo->marker->saw_SOI) {
+	if (! first_marker(cinfo))
+	  return JPEG_SUSPENDED;
+      } else {
+	if (! next_marker(cinfo))
+	  return JPEG_SUSPENDED;
+      }
+    }
+
+    /*
+     * Save the position of the fist marker after SOF.
+     */
+    if (cinfo->marker->current_sos_marker_position == -1)
+      cinfo->marker->current_sos_marker_position =
+          jget_input_stream_position(cinfo) - 2;
+
+    /* At this point cinfo->unread_marker contains the marker code and the
+     * input point is just past the marker proper, but before any parameters.
+     * A suspension will cause us to return with this state still true.
+     */
+    switch (cinfo->unread_marker) {
+    case M_SOI:
+      if (! get_soi(cinfo))
+	return JPEG_SUSPENDED;
+      break;
+
+    case M_SOF0:		/* Baseline */
+    case M_SOF1:		/* Extended sequential, Huffman */
+      if (! get_sof(cinfo, FALSE, FALSE))
+	return JPEG_SUSPENDED;
+      break;
+
+    case M_SOF2:		/* Progressive, Huffman */
+      cinfo->marker->current_sos_marker_position = -1;
+      if (! get_sof(cinfo, TRUE, FALSE))
+	return JPEG_SUSPENDED;
+      break;
+
+    case M_SOF9:		/* Extended sequential, arithmetic */
+      if (! get_sof(cinfo, FALSE, TRUE))
+	return JPEG_SUSPENDED;
+      break;
+
+    case M_SOF10:		/* Progressive, arithmetic */
+      if (! get_sof(cinfo, TRUE, TRUE))
+	return JPEG_SUSPENDED;
+      break;
+
+    /* Currently unsupported SOFn types */
+    case M_SOF3:		/* Lossless, Huffman */
+    case M_SOF5:		/* Differential sequential, Huffman */
+    case M_SOF6:		/* Differential progressive, Huffman */
+    case M_SOF7:		/* Differential lossless, Huffman */
+    case M_JPG:			/* Reserved for JPEG extensions */
+    case M_SOF11:		/* Lossless, arithmetic */
+    case M_SOF13:		/* Differential sequential, arithmetic */
+    case M_SOF14:		/* Differential progressive, arithmetic */
+    case M_SOF15:		/* Differential lossless, arithmetic */
+      ERREXIT1(cinfo, JERR_SOF_UNSUPPORTED, cinfo->unread_marker);
+      break;
+
+    case M_SOS:
+      if (! get_sos(cinfo))
+	return JPEG_SUSPENDED;
+      cinfo->unread_marker = 0;	/* processed the marker */
+      return JPEG_REACHED_SOS;
+    
+    case M_EOI:
+      TRACEMS(cinfo, 1, JTRC_EOI);
+      cinfo->unread_marker = 0;	/* processed the marker */
+      return JPEG_REACHED_EOI;
+      
+    case M_DAC:
+      if (! get_dac(cinfo))
+	return JPEG_SUSPENDED;
+      break;
+      
+    case M_DHT:
+      if (! get_dht(cinfo))
+	return JPEG_SUSPENDED;
+      break;
+      
+    case M_DQT:
+      if (! get_dqt(cinfo))
+	return JPEG_SUSPENDED;
+      break;
+      
+    case M_DRI:
+      if (! get_dri(cinfo))
+	return JPEG_SUSPENDED;
+      break;
+      
+    case M_APP0:
+    case M_APP1:
+    case M_APP2:
+    case M_APP3:
+    case M_APP4:
+    case M_APP5:
+    case M_APP6:
+    case M_APP7:
+    case M_APP8:
+    case M_APP9:
+    case M_APP10:
+    case M_APP11:
+    case M_APP12:
+    case M_APP13:
+    case M_APP14:
+    case M_APP15:
+      if (! (*((my_marker_ptr) cinfo->marker)->process_APPn[
+		cinfo->unread_marker - (int) M_APP0]) (cinfo))
+	return JPEG_SUSPENDED;
+      break;
+      
+    case M_COM:
+      if (! (*((my_marker_ptr) cinfo->marker)->process_COM) (cinfo))
+	return JPEG_SUSPENDED;
+      break;
+
+    case M_RST0:		/* these are all parameterless */
+    case M_RST1:
+    case M_RST2:
+    case M_RST3:
+    case M_RST4:
+    case M_RST5:
+    case M_RST6:
+    case M_RST7:
+    case M_TEM:
+      TRACEMS1(cinfo, 1, JTRC_PARMLESS_MARKER, cinfo->unread_marker);
+      break;
+
+    case M_DNL:			/* Ignore DNL ... perhaps the wrong thing */
+      if (! skip_variable(cinfo))
+	return JPEG_SUSPENDED;
+      break;
+
+    default:			/* must be DHP, EXP, JPGn, or RESn */
+      /* For now, we treat the reserved markers as fatal errors since they are
+       * likely to be used to signal incompatible JPEG Part 3 extensions.
+       * Once the JPEG 3 version-number marker is well defined, this code
+       * ought to change!
+       */
+      ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, cinfo->unread_marker);
+      break;
+    }
+    /* Successfully processed marker, so reset state variable */
+    cinfo->unread_marker = 0;
+  } /* end loop */
+}
+
+
+/*
+ * Read a restart marker, which is expected to appear next in the datastream;
+ * if the marker is not there, take appropriate recovery action.
+ * Returns FALSE if suspension is required.
+ *
+ * This is called by the entropy decoder after it has read an appropriate
+ * number of MCUs.  cinfo->unread_marker may be nonzero if the entropy decoder
+ * has already read a marker from the data source.  Under normal conditions
+ * cinfo->unread_marker will be reset to 0 before returning; if not reset,
+ * it holds a marker which the decoder will be unable to read past.
+ */
+
+METHODDEF(boolean)
+read_restart_marker (j_decompress_ptr cinfo)
+{
+  /* Obtain a marker unless we already did. */
+  /* Note that next_marker will complain if it skips any data. */
+  if (cinfo->unread_marker == 0) {
+    if (! next_marker(cinfo))
+      return FALSE;
+  }
+
+  if (cinfo->unread_marker ==
+      ((int) M_RST0 + cinfo->marker->next_restart_num)) {
+    /* Normal case --- swallow the marker and let entropy decoder continue */
+    TRACEMS1(cinfo, 3, JTRC_RST, cinfo->marker->next_restart_num);
+    cinfo->unread_marker = 0;
+  } else {
+    /* Uh-oh, the restart markers have been messed up. */
+    /* Let the data source manager determine how to resync. */
+    if (! (*cinfo->src->resync_to_restart) (cinfo,
+					    cinfo->marker->next_restart_num))
+      return FALSE;
+  }
+
+  /* Update next-restart state */
+  cinfo->marker->next_restart_num = (cinfo->marker->next_restart_num + 1) & 7;
+
+  return TRUE;
+}
+
+
+/*
+ * This is the default resync_to_restart method for data source managers
+ * to use if they don't have any better approach.  Some data source managers
+ * may be able to back up, or may have additional knowledge about the data
+ * which permits a more intelligent recovery strategy; such managers would
+ * presumably supply their own resync method.
+ *
+ * read_restart_marker calls resync_to_restart if it finds a marker other than
+ * the restart marker it was expecting.  (This code is *not* used unless
+ * a nonzero restart interval has been declared.)  cinfo->unread_marker is
+ * the marker code actually found (might be anything, except 0 or FF).
+ * The desired restart marker number (0..7) is passed as a parameter.
+ * This routine is supposed to apply whatever error recovery strategy seems
+ * appropriate in order to position the input stream to the next data segment.
+ * Note that cinfo->unread_marker is treated as a marker appearing before
+ * the current data-source input point; usually it should be reset to zero
+ * before returning.
+ * Returns FALSE if suspension is required.
+ *
+ * This implementation is substantially constrained by wanting to treat the
+ * input as a data stream; this means we can't back up.  Therefore, we have
+ * only the following actions to work with:
+ *   1. Simply discard the marker and let the entropy decoder resume at next
+ *      byte of file.
+ *   2. Read forward until we find another marker, discarding intervening
+ *      data.  (In theory we could look ahead within the current bufferload,
+ *      without having to discard data if we don't find the desired marker.
+ *      This idea is not implemented here, in part because it makes behavior
+ *      dependent on buffer size and chance buffer-boundary positions.)
+ *   3. Leave the marker unread (by failing to zero cinfo->unread_marker).
+ *      This will cause the entropy decoder to process an empty data segment,
+ *      inserting dummy zeroes, and then we will reprocess the marker.
+ *
+ * #2 is appropriate if we think the desired marker lies ahead, while #3 is
+ * appropriate if the found marker is a future restart marker (indicating
+ * that we have missed the desired restart marker, probably because it got
+ * corrupted).
+ * We apply #2 or #3 if the found marker is a restart marker no more than
+ * two counts behind or ahead of the expected one.  We also apply #2 if the
+ * found marker is not a legal JPEG marker code (it's certainly bogus data).
+ * If the found marker is a restart marker more than 2 counts away, we do #1
+ * (too much risk that the marker is erroneous; with luck we will be able to
+ * resync at some future point).
+ * For any valid non-restart JPEG marker, we apply #3.  This keeps us from
+ * overrunning the end of a scan.  An implementation limited to single-scan
+ * files might find it better to apply #2 for markers other than EOI, since
+ * any other marker would have to be bogus data in that case.
+ */
+
+GLOBAL(boolean)
+jpeg_resync_to_restart (j_decompress_ptr cinfo, int desired)
+{
+  int marker = cinfo->unread_marker;
+  int action = 1;
+  
+  /* Always put up a warning. */
+  WARNMS2(cinfo, JWRN_MUST_RESYNC, marker, desired);
+  
+  /* Outer loop handles repeated decision after scanning forward. */
+  for (;;) {
+    if (marker < (int) M_SOF0)
+      action = 2;		/* invalid marker */
+    else if (marker < (int) M_RST0 || marker > (int) M_RST7)
+      action = 3;		/* valid non-restart marker */
+    else {
+      if (marker == ((int) M_RST0 + ((desired+1) & 7)) ||
+	  marker == ((int) M_RST0 + ((desired+2) & 7)))
+	action = 3;		/* one of the next two expected restarts */
+      else if (marker == ((int) M_RST0 + ((desired-1) & 7)) ||
+	       marker == ((int) M_RST0 + ((desired-2) & 7)))
+	action = 2;		/* a prior restart, so advance */
+      else
+	action = 1;		/* desired restart or too far away */
+    }
+    TRACEMS2(cinfo, 4, JTRC_RECOVERY_ACTION, marker, action);
+    switch (action) {
+    case 1:
+      /* Discard marker and let entropy decoder resume processing. */
+      cinfo->unread_marker = 0;
+      return TRUE;
+    case 2:
+      /* Scan to the next marker, and repeat the decision loop. */
+      if (! next_marker(cinfo))
+	return FALSE;
+      marker = cinfo->unread_marker;
+      break;
+    case 3:
+      /* Return without advancing past this marker. */
+      /* Entropy decoder will be forced to process an empty segment. */
+      return TRUE;
+    }
+  } /* end loop */
+}
+
+/*
+ * Get the position for all SOS markers in the image.
+ */
+
+METHODDEF(void)
+get_sos_marker_position(j_decompress_ptr cinfo, huffman_index *index)
+{
+  unsigned char *head;
+  int count = 0;
+  int retcode = JPEG_REACHED_SOS;
+
+  while (cinfo->src->bytes_in_buffer > 0) {
+    if (retcode == JPEG_REACHED_SOS) {
+      jpeg_configure_huffman_index_scan(cinfo, index, count++,
+              cinfo->marker->current_sos_marker_position);
+      // Skips scan content to the next non-RST JPEG marker.
+      while(next_marker(cinfo) &&
+              cinfo->unread_marker >= M_RST0 && cinfo->unread_marker <= M_RST7)
+          ;
+      cinfo->marker->current_sos_marker_position =
+        jget_input_stream_position(cinfo) - 2;
+      retcode = read_markers(cinfo);
+    } else {
+      break;
+    }
+  }
+}
+
+/*
+ * Reset marker processing state to begin a fresh datastream.
+ */
+
+METHODDEF(void)
+reset_marker_reader (j_decompress_ptr cinfo)
+{
+  my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
+
+  cinfo->comp_info = NULL;		/* until allocated by get_sof */
+  cinfo->input_scan_number = 0;		/* no SOS seen yet */
+  cinfo->unread_marker = 0;		/* no pending marker */
+  marker->pub.saw_SOI = FALSE;		/* set internal state too */
+  marker->pub.saw_SOF = FALSE;
+  marker->pub.discarded_bytes = 0;
+  marker->cur_marker = NULL;
+}
+
+
+/*
+ * Initialize the marker reader module.
+ * This is called only once, when the decompression object is created.
+ */
+
+GLOBAL(void)
+jinit_marker_reader (j_decompress_ptr cinfo)
+{
+  my_marker_ptr marker;
+  int i;
+
+  /* Create subobject in permanent pool */
+  marker = (my_marker_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				SIZEOF(my_marker_reader));
+  cinfo->marker = (struct jpeg_marker_reader *) marker;
+  /* Initialize public method pointers */
+  marker->pub.reset_marker_reader = reset_marker_reader;
+  marker->pub.read_markers = read_markers;
+  marker->pub.read_restart_marker = read_restart_marker;
+  marker->pub.get_sos_marker_position = get_sos_marker_position;
+  /* Initialize COM/APPn processing.
+   * By default, we examine and then discard APP0 and APP14,
+   * but simply discard COM and all other APPn.
+   */
+  marker->process_COM = skip_variable;
+  marker->length_limit_COM = 0;
+  for (i = 0; i < 16; i++) {
+    marker->process_APPn[i] = skip_variable;
+    marker->length_limit_APPn[i] = 0;
+  }
+  marker->process_APPn[0] = get_interesting_appn;
+  marker->process_APPn[14] = get_interesting_appn;
+  /* Reset marker processing state */
+  reset_marker_reader(cinfo);
+}
+
+
+/*
+ * Control saving of COM and APPn markers into marker_list.
+ */
+
+#ifdef SAVE_MARKERS_SUPPORTED
+
+GLOBAL(void)
+jpeg_save_markers (j_decompress_ptr cinfo, int marker_code,
+		   unsigned int length_limit)
+{
+  my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
+  long maxlength;
+  jpeg_marker_parser_method processor;
+
+  /* Length limit mustn't be larger than what we can allocate
+   * (should only be a concern in a 16-bit environment).
+   */
+  maxlength = cinfo->mem->max_alloc_chunk - SIZEOF(struct jpeg_marker_struct);
+  if (((long) length_limit) > maxlength)
+    length_limit = (unsigned int) maxlength;
+
+  /* Choose processor routine to use.
+   * APP0/APP14 have special requirements.
+   */
+  if (length_limit) {
+    processor = save_marker;
+    /* If saving APP0/APP14, save at least enough for our internal use. */
+    if (marker_code == (int) M_APP0 && length_limit < APP0_DATA_LEN)
+      length_limit = APP0_DATA_LEN;
+    else if (marker_code == (int) M_APP14 && length_limit < APP14_DATA_LEN)
+      length_limit = APP14_DATA_LEN;
+  } else {
+    processor = skip_variable;
+    /* If discarding APP0/APP14, use our regular on-the-fly processor. */
+    if (marker_code == (int) M_APP0 || marker_code == (int) M_APP14)
+      processor = get_interesting_appn;
+  }
+
+  if (marker_code == (int) M_COM) {
+    marker->process_COM = processor;
+    marker->length_limit_COM = length_limit;
+  } else if (marker_code >= (int) M_APP0 && marker_code <= (int) M_APP15) {
+    marker->process_APPn[marker_code - (int) M_APP0] = processor;
+    marker->length_limit_APPn[marker_code - (int) M_APP0] = length_limit;
+  } else
+    ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, marker_code);
+}
+
+#endif /* SAVE_MARKERS_SUPPORTED */
+
+
+/*
+ * Install a special processing method for COM or APPn markers.
+ */
+
+GLOBAL(void)
+jpeg_set_marker_processor (j_decompress_ptr cinfo, int marker_code,
+			   jpeg_marker_parser_method routine)
+{
+  my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
+
+  if (marker_code == (int) M_COM)
+    marker->process_COM = routine;
+  else if (marker_code >= (int) M_APP0 && marker_code <= (int) M_APP15)
+    marker->process_APPn[marker_code - (int) M_APP0] = routine;
+  else
+    ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, marker_code);
+}
diff --git a/libjpegtwrp/jdmaster.c b/libjpegtwrp/jdmaster.c
new file mode 100644
index 000000000..e3da758b1
--- /dev/null
+++ b/libjpegtwrp/jdmaster.c
@@ -0,0 +1,580 @@
+/*
+ * jdmaster.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains master control logic for the JPEG decompressor.
+ * These routines are concerned with selecting the modules to be executed
+ * and with determining the number of passes and the work to be done in each
+ * pass.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_decomp_master pub; /* public fields */
+
+  int pass_number;		/* # of passes completed */
+
+  boolean using_merged_upsample; /* TRUE if using merged upsample/cconvert */
+
+  /* Saved references to initialized quantizer modules,
+   * in case we need to switch modes.
+   */
+  struct jpeg_color_quantizer * quantizer_1pass;
+  struct jpeg_color_quantizer * quantizer_2pass;
+} my_decomp_master;
+
+typedef my_decomp_master * my_master_ptr;
+
+
+/*
+ * Determine whether merged upsample/color conversion should be used.
+ * CRUCIAL: this must match the actual capabilities of jdmerge.c!
+ */
+
+LOCAL(boolean)
+use_merged_upsample (j_decompress_ptr cinfo)
+{
+#ifdef UPSAMPLE_MERGING_SUPPORTED
+  /* Merging is the equivalent of plain box-filter upsampling */
+  if (cinfo->do_fancy_upsampling || cinfo->CCIR601_sampling)
+    return FALSE;
+
+#ifdef ANDROID_RGB
+  /* jdmerge.c only supports YCC=>RGB565 and YCC=>RGB color conversion */
+  if (cinfo->jpeg_color_space != JCS_YCbCr || 
+      cinfo->num_components != 3 ||
+      cinfo->out_color_components != 3 ||
+      (cinfo->out_color_space != JCS_RGB_565 && 
+         cinfo->out_color_space != JCS_RGB)) {
+    return FALSE;
+  }
+#else
+  /* jdmerge.c only supports YCC=>RGB color conversion */
+  if (cinfo->jpeg_color_space != JCS_YCbCr || cinfo->num_components != 3 ||
+      cinfo->out_color_space != JCS_RGB ||
+      cinfo->out_color_components != RGB_PIXELSIZE)
+    return FALSE;
+#endif
+
+  /* and it only handles 2h1v or 2h2v sampling ratios */
+  if (cinfo->comp_info[0].h_samp_factor != 2 ||
+      cinfo->comp_info[1].h_samp_factor != 1 ||
+      cinfo->comp_info[2].h_samp_factor != 1 ||
+      cinfo->comp_info[0].v_samp_factor >  2 ||
+      cinfo->comp_info[1].v_samp_factor != 1 ||
+      cinfo->comp_info[2].v_samp_factor != 1)
+    return FALSE;
+  /* furthermore, it doesn't work if we've scaled the IDCTs differently */
+  if (cinfo->comp_info[0].DCT_scaled_size != cinfo->min_DCT_scaled_size ||
+      cinfo->comp_info[1].DCT_scaled_size != cinfo->min_DCT_scaled_size ||
+      cinfo->comp_info[2].DCT_scaled_size != cinfo->min_DCT_scaled_size)
+    return FALSE;
+  /* ??? also need to test for upsample-time rescaling, when & if supported */
+  return TRUE;			/* by golly, it'll work... */
+#else
+  return FALSE;
+#endif
+}
+
+
+/*
+ * Compute output image dimensions and related values.
+ * NOTE: this is exported for possible use by application.
+ * Hence it mustn't do anything that can't be done twice.
+ * Also note that it may be called before the master module is initialized!
+ */
+
+GLOBAL(void)
+jpeg_calc_output_dimensions (j_decompress_ptr cinfo)
+/* Do computations that are needed before master selection phase */
+{
+#ifdef IDCT_SCALING_SUPPORTED
+  int ci;
+  jpeg_component_info *compptr;
+#endif
+
+  /* Prevent application from calling me at wrong times */
+#if ANDROID_TILE_BASED_DECODE
+  // Tile based decoding may call this function several times.
+  if (!cinfo->tile_decode)
+#endif
+    if (cinfo->global_state != DSTATE_READY)
+      ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+#ifdef IDCT_SCALING_SUPPORTED
+
+  /* Compute actual output image dimensions and DCT scaling choices. */
+  if (cinfo->scale_num * 8 <= cinfo->scale_denom) {
+    /* Provide 1/8 scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width, 8L);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height, 8L);
+    cinfo->min_DCT_scaled_size = 1;
+  } else if (cinfo->scale_num * 4 <= cinfo->scale_denom) {
+    /* Provide 1/4 scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width, 4L);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height, 4L);
+    cinfo->min_DCT_scaled_size = 2;
+  } else if (cinfo->scale_num * 2 <= cinfo->scale_denom) {
+    /* Provide 1/2 scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width, 2L);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height, 2L);
+    cinfo->min_DCT_scaled_size = 4;
+  } else {
+    /* Provide 1/1 scaling */
+    cinfo->output_width = cinfo->image_width;
+    cinfo->output_height = cinfo->image_height;
+    cinfo->min_DCT_scaled_size = DCTSIZE;
+  }
+  /* In selecting the actual DCT scaling for each component, we try to
+   * scale up the chroma components via IDCT scaling rather than upsampling.
+   * This saves time if the upsampler gets to use 1:1 scaling.
+   * Note this code assumes that the supported DCT scalings are powers of 2.
+   */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    int ssize = cinfo->min_DCT_scaled_size;
+    while (ssize < DCTSIZE &&
+	   (compptr->h_samp_factor * ssize * 2 <=
+	    cinfo->max_h_samp_factor * cinfo->min_DCT_scaled_size) &&
+	   (compptr->v_samp_factor * ssize * 2 <=
+	    cinfo->max_v_samp_factor * cinfo->min_DCT_scaled_size)) {
+      ssize = ssize * 2;
+    }
+    compptr->DCT_scaled_size = ssize;
+  }
+
+  /* Recompute downsampled dimensions of components;
+   * application needs to know these if using raw downsampled data.
+   */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Size in samples, after IDCT scaling */
+    compptr->downsampled_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width *
+		    (long) (compptr->h_samp_factor * compptr->DCT_scaled_size),
+		    (long) (cinfo->max_h_samp_factor * DCTSIZE));
+    compptr->downsampled_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height *
+		    (long) (compptr->v_samp_factor * compptr->DCT_scaled_size),
+		    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+  }
+
+#else /* !IDCT_SCALING_SUPPORTED */
+
+  /* Hardwire it to "no scaling" */
+  cinfo->output_width = cinfo->image_width;
+  cinfo->output_height = cinfo->image_height;
+  /* jdinput.c has already initialized DCT_scaled_size to DCTSIZE,
+   * and has computed unscaled downsampled_width and downsampled_height.
+   */
+
+#endif /* IDCT_SCALING_SUPPORTED */
+
+  /* Report number of components in selected colorspace. */
+  /* Probably this should be in the color conversion module... */
+  switch (cinfo->out_color_space) {
+  case JCS_GRAYSCALE:
+    cinfo->out_color_components = 1;
+    break;
+  case JCS_RGB:
+#if RGB_PIXELSIZE != 3
+    cinfo->out_color_components = RGB_PIXELSIZE;
+    break;
+#endif /* else share code with YCbCr */
+#ifdef ANDROID_RGB
+  case JCS_RGB_565:
+#endif
+  case JCS_YCbCr:
+    cinfo->out_color_components = 3;
+    break;
+  case JCS_CMYK:
+  case JCS_YCCK:
+#ifdef ANDROID_RGB
+  case JCS_RGBA_8888:
+#endif
+    cinfo->out_color_components = 4;
+    break;
+  default:			/* else must be same colorspace as in file */
+    cinfo->out_color_components = cinfo->num_components;
+    break;
+  }
+  cinfo->output_components = (cinfo->quantize_colors ? 1 :
+			      cinfo->out_color_components);
+
+  /* See if upsampler will want to emit more than one row at a time */
+  if (use_merged_upsample(cinfo))
+    cinfo->rec_outbuf_height = cinfo->max_v_samp_factor;
+  else
+    cinfo->rec_outbuf_height = 1;
+}
+
+
+/*
+ * Several decompression processes need to range-limit values to the range
+ * 0..MAXJSAMPLE; the input value may fall somewhat outside this range
+ * due to noise introduced by quantization, roundoff error, etc.  These
+ * processes are inner loops and need to be as fast as possible.  On most
+ * machines, particularly CPUs with pipelines or instruction prefetch,
+ * a (subscript-check-less) C table lookup
+ *		x = sample_range_limit[x];
+ * is faster than explicit tests
+ *		if (x < 0)  x = 0;
+ *		else if (x > MAXJSAMPLE)  x = MAXJSAMPLE;
+ * These processes all use a common table prepared by the routine below.
+ *
+ * For most steps we can mathematically guarantee that the initial value
+ * of x is within MAXJSAMPLE+1 of the legal range, so a table running from
+ * -(MAXJSAMPLE+1) to 2*MAXJSAMPLE+1 is sufficient.  But for the initial
+ * limiting step (just after the IDCT), a wildly out-of-range value is
+ * possible if the input data is corrupt.  To avoid any chance of indexing
+ * off the end of memory and getting a bad-pointer trap, we perform the
+ * post-IDCT limiting thus:
+ *		x = range_limit[x & MASK];
+ * where MASK is 2 bits wider than legal sample data, ie 10 bits for 8-bit
+ * samples.  Under normal circumstances this is more than enough range and
+ * a correct output will be generated; with bogus input data the mask will
+ * cause wraparound, and we will safely generate a bogus-but-in-range output.
+ * For the post-IDCT step, we want to convert the data from signed to unsigned
+ * representation by adding CENTERJSAMPLE at the same time that we limit it.
+ * So the post-IDCT limiting table ends up looking like this:
+ *   CENTERJSAMPLE,CENTERJSAMPLE+1,...,MAXJSAMPLE,
+ *   MAXJSAMPLE (repeat 2*(MAXJSAMPLE+1)-CENTERJSAMPLE times),
+ *   0          (repeat 2*(MAXJSAMPLE+1)-CENTERJSAMPLE times),
+ *   0,1,...,CENTERJSAMPLE-1
+ * Negative inputs select values from the upper half of the table after
+ * masking.
+ *
+ * We can save some space by overlapping the start of the post-IDCT table
+ * with the simpler range limiting table.  The post-IDCT table begins at
+ * sample_range_limit + CENTERJSAMPLE.
+ *
+ * Note that the table is allocated in near data space on PCs; it's small
+ * enough and used often enough to justify this.
+ */
+
+LOCAL(void)
+prepare_range_limit_table (j_decompress_ptr cinfo)
+/* Allocate and fill in the sample_range_limit table */
+{
+  JSAMPLE * table;
+  int i;
+
+  table = (JSAMPLE *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+		(5 * (MAXJSAMPLE+1) + CENTERJSAMPLE) * SIZEOF(JSAMPLE));
+  table += (MAXJSAMPLE+1);	/* allow negative subscripts of simple table */
+  cinfo->sample_range_limit = table;
+  /* First segment of "simple" table: limit[x] = 0 for x < 0 */
+  MEMZERO(table - (MAXJSAMPLE+1), (MAXJSAMPLE+1) * SIZEOF(JSAMPLE));
+  /* Main part of "simple" table: limit[x] = x */
+  for (i = 0; i <= MAXJSAMPLE; i++)
+    table[i] = (JSAMPLE) i;
+  table += CENTERJSAMPLE;	/* Point to where post-IDCT table starts */
+  /* End of simple table, rest of first half of post-IDCT table */
+  for (i = CENTERJSAMPLE; i < 2*(MAXJSAMPLE+1); i++)
+    table[i] = MAXJSAMPLE;
+  /* Second half of post-IDCT table */
+  MEMZERO(table + (2 * (MAXJSAMPLE+1)),
+	  (2 * (MAXJSAMPLE+1) - CENTERJSAMPLE) * SIZEOF(JSAMPLE));
+  MEMCOPY(table + (4 * (MAXJSAMPLE+1) - CENTERJSAMPLE),
+	  cinfo->sample_range_limit, CENTERJSAMPLE * SIZEOF(JSAMPLE));
+}
+
+
+/*
+ * Master selection of decompression modules.
+ * This is done once at jpeg_start_decompress time.  We determine
+ * which modules will be used and give them appropriate initialization calls.
+ * We also initialize the decompressor input side to begin consuming data.
+ *
+ * Since jpeg_read_header has finished, we know what is in the SOF
+ * and (first) SOS markers.  We also have all the application parameter
+ * settings.
+ */
+
+LOCAL(void)
+master_selection (j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr) cinfo->master;
+  boolean use_c_buffer;
+  long samplesperrow;
+  JDIMENSION jd_samplesperrow;
+
+  /* Initialize dimensions and other stuff */
+  jpeg_calc_output_dimensions(cinfo);
+  prepare_range_limit_table(cinfo);
+
+  /* Width of an output scanline must be representable as JDIMENSION. */
+  samplesperrow = (long) cinfo->output_width * (long) cinfo->out_color_components;
+  jd_samplesperrow = (JDIMENSION) samplesperrow;
+  if ((long) jd_samplesperrow != samplesperrow)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+
+  /* Initialize my private state */
+  master->pass_number = 0;
+  master->using_merged_upsample = use_merged_upsample(cinfo);
+
+  /* Color quantizer selection */
+  master->quantizer_1pass = NULL;
+  master->quantizer_2pass = NULL;
+  /* No mode changes if not using buffered-image mode. */
+  if (! cinfo->quantize_colors || ! cinfo->buffered_image) {
+    cinfo->enable_1pass_quant = FALSE;
+    cinfo->enable_external_quant = FALSE;
+    cinfo->enable_2pass_quant = FALSE;
+  }
+  if (cinfo->quantize_colors) {
+    if (cinfo->raw_data_out)
+      ERREXIT(cinfo, JERR_NOTIMPL);
+    /* 2-pass quantizer only works in 3-component color space. */
+    if (cinfo->out_color_components != 3) {
+      cinfo->enable_1pass_quant = TRUE;
+      cinfo->enable_external_quant = FALSE;
+      cinfo->enable_2pass_quant = FALSE;
+      cinfo->colormap = NULL;
+    } else if (cinfo->colormap != NULL) {
+      cinfo->enable_external_quant = TRUE;
+    } else if (cinfo->two_pass_quantize) {
+      cinfo->enable_2pass_quant = TRUE;
+    } else {
+      cinfo->enable_1pass_quant = TRUE;
+    }
+
+    if (cinfo->enable_1pass_quant) {
+#ifdef QUANT_1PASS_SUPPORTED
+      jinit_1pass_quantizer(cinfo);
+      master->quantizer_1pass = cinfo->cquantize;
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    }
+
+    /* We use the 2-pass code to map to external colormaps. */
+    if (cinfo->enable_2pass_quant || cinfo->enable_external_quant) {
+#ifdef QUANT_2PASS_SUPPORTED
+      jinit_2pass_quantizer(cinfo);
+      master->quantizer_2pass = cinfo->cquantize;
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    }
+    /* If both quantizers are initialized, the 2-pass one is left active;
+     * this is necessary for starting with quantization to an external map.
+     */
+  }
+
+  /* Post-processing: in particular, color conversion first */
+  if (! cinfo->raw_data_out) {
+    if (master->using_merged_upsample) {
+#ifdef UPSAMPLE_MERGING_SUPPORTED
+      jinit_merged_upsampler(cinfo); /* does color conversion too */
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else {
+      jinit_color_deconverter(cinfo);
+      jinit_upsampler(cinfo);
+    }
+    jinit_d_post_controller(cinfo, cinfo->enable_2pass_quant);
+  }
+  /* Inverse DCT */
+  jinit_inverse_dct(cinfo);
+  /* Entropy decoding: either Huffman or arithmetic coding. */
+  if (cinfo->arith_code) {
+    ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+  } else {
+    if (cinfo->progressive_mode) {
+#ifdef D_PROGRESSIVE_SUPPORTED
+      jinit_phuff_decoder(cinfo);
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else
+      jinit_huff_decoder(cinfo);
+  }
+
+  /* Initialize principal buffer controllers. */
+  use_c_buffer = cinfo->inputctl->has_multiple_scans || cinfo->buffered_image;
+  jinit_d_coef_controller(cinfo, use_c_buffer);
+
+  if (! cinfo->raw_data_out)
+    jinit_d_main_controller(cinfo, FALSE /* never need full buffer here */);
+
+  /* We can now tell the memory manager to allocate virtual arrays. */
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr) cinfo);
+
+  /* Initialize input side of decompressor to consume first scan. */
+  (*cinfo->inputctl->start_input_pass) (cinfo);
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+  /* If jpeg_start_decompress will read the whole file, initialize
+   * progress monitoring appropriately.  The input step is counted
+   * as one pass.
+   */
+  if (cinfo->progress != NULL && ! cinfo->buffered_image &&
+      cinfo->inputctl->has_multiple_scans) {
+    int nscans;
+    /* Estimate number of scans to set pass_limit. */
+    if (cinfo->progressive_mode) {
+      /* Arbitrarily estimate 2 interleaved DC scans + 3 AC scans/component. */
+      nscans = 2 + 3 * cinfo->num_components;
+    } else {
+      /* For a nonprogressive multiscan file, estimate 1 scan per component. */
+      nscans = cinfo->num_components;
+    }
+    cinfo->progress->pass_counter = 0L;
+    cinfo->progress->pass_limit = (long) cinfo->total_iMCU_rows * nscans;
+    cinfo->progress->completed_passes = 0;
+    cinfo->progress->total_passes = (cinfo->enable_2pass_quant ? 3 : 2);
+    /* Count the input pass as done */
+    master->pass_number++;
+  }
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
+}
+
+
+/*
+ * Per-pass setup.
+ * This is called at the beginning of each output pass.  We determine which
+ * modules will be active during this pass and give them appropriate
+ * start_pass calls.  We also set is_dummy_pass to indicate whether this
+ * is a "real" output pass or a dummy pass for color quantization.
+ * (In the latter case, jdapistd.c will crank the pass to completion.)
+ */
+
+METHODDEF(void)
+prepare_for_output_pass (j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr) cinfo->master;
+
+  if (master->pub.is_dummy_pass) {
+#ifdef QUANT_2PASS_SUPPORTED
+    /* Final pass of 2-pass quantization */
+    master->pub.is_dummy_pass = FALSE;
+    (*cinfo->cquantize->start_pass) (cinfo, FALSE);
+    (*cinfo->post->start_pass) (cinfo, JBUF_CRANK_DEST);
+    (*cinfo->main->start_pass) (cinfo, JBUF_CRANK_DEST);
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif /* QUANT_2PASS_SUPPORTED */
+  } else {
+    if (cinfo->quantize_colors && cinfo->colormap == NULL) {
+      /* Select new quantization method */
+      if (cinfo->two_pass_quantize && cinfo->enable_2pass_quant) {
+	cinfo->cquantize = master->quantizer_2pass;
+	master->pub.is_dummy_pass = TRUE;
+      } else if (cinfo->enable_1pass_quant) {
+	cinfo->cquantize = master->quantizer_1pass;
+      } else {
+	ERREXIT(cinfo, JERR_MODE_CHANGE);
+      }
+    }
+    (*cinfo->idct->start_pass) (cinfo);
+    (*cinfo->coef->start_output_pass) (cinfo);
+    if (! cinfo->raw_data_out) {
+      if (! master->using_merged_upsample)
+	(*cinfo->cconvert->start_pass) (cinfo);
+      (*cinfo->upsample->start_pass) (cinfo);
+      if (cinfo->quantize_colors)
+	(*cinfo->cquantize->start_pass) (cinfo, master->pub.is_dummy_pass);
+      (*cinfo->post->start_pass) (cinfo,
+	    (master->pub.is_dummy_pass ? JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
+      (*cinfo->main->start_pass) (cinfo, JBUF_PASS_THRU);
+    }
+  }
+
+  /* Set up progress monitor's pass info if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->completed_passes = master->pass_number;
+    cinfo->progress->total_passes = master->pass_number +
+				    (master->pub.is_dummy_pass ? 2 : 1);
+    /* In buffered-image mode, we assume one more output pass if EOI not
+     * yet reached, but no more passes if EOI has been reached.
+     */
+    if (cinfo->buffered_image && ! cinfo->inputctl->eoi_reached) {
+      cinfo->progress->total_passes += (cinfo->enable_2pass_quant ? 2 : 1);
+    }
+  }
+}
+
+
+/*
+ * Finish up at end of an output pass.
+ */
+
+METHODDEF(void)
+finish_output_pass (j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr) cinfo->master;
+
+  if (cinfo->quantize_colors)
+    (*cinfo->cquantize->finish_pass) (cinfo);
+  master->pass_number++;
+}
+
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+
+/*
+ * Switch to a new external colormap between output passes.
+ */
+
+GLOBAL(void)
+jpeg_new_colormap (j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr) cinfo->master;
+
+  /* Prevent application from calling me at wrong times */
+  if (cinfo->global_state != DSTATE_BUFIMAGE)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  if (cinfo->quantize_colors && cinfo->enable_external_quant &&
+      cinfo->colormap != NULL) {
+    /* Select 2-pass quantizer for external colormap use */
+    cinfo->cquantize = master->quantizer_2pass;
+    /* Notify quantizer of colormap change */
+    (*cinfo->cquantize->new_color_map) (cinfo);
+    master->pub.is_dummy_pass = FALSE; /* just in case */
+  } else
+    ERREXIT(cinfo, JERR_MODE_CHANGE);
+}
+
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
+
+
+/*
+ * Initialize master decompression control and select active modules.
+ * This is performed at the start of jpeg_start_decompress.
+ */
+
+GLOBAL(void)
+jinit_master_decompress (j_decompress_ptr cinfo)
+{
+  my_master_ptr master;
+
+  master = (my_master_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(my_decomp_master));
+  cinfo->master = (struct jpeg_decomp_master *) master;
+  master->pub.prepare_for_output_pass = prepare_for_output_pass;
+  master->pub.finish_output_pass = finish_output_pass;
+
+  master->pub.is_dummy_pass = FALSE;
+
+  master_selection(cinfo);
+}
diff --git a/libjpegtwrp/jdmerge.c b/libjpegtwrp/jdmerge.c
new file mode 100644
index 000000000..171fbd3ab
--- /dev/null
+++ b/libjpegtwrp/jdmerge.c
@@ -0,0 +1,902 @@
+/*
+ * jdmerge.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Copyright (c) 2010, Code Aurora Forum. All rights reserved.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains code for merged upsampling/color conversion.
+ *
+ * This file combines functions from jdsample.c and jdcolor.c;
+ * read those files first to understand what's going on.
+ *
+ * When the chroma components are to be upsampled by simple replication
+ * (ie, box filtering), we can save some work in color conversion by
+ * calculating all the output pixels corresponding to a pair of chroma
+ * samples at one time.  In the conversion equations
+ *	R = Y           + K1 * Cr
+ *	G = Y + K2 * Cb + K3 * Cr
+ *	B = Y + K4 * Cb
+ * only the Y term varies among the group of pixels corresponding to a pair
+ * of chroma samples, so the rest of the terms can be calculated just once.
+ * At typical sampling ratios, this eliminates half or three-quarters of the
+ * multiplications needed for color conversion.
+ *
+ * This file currently provides implementations for the following cases:
+ *	YCbCr => RGB color conversion only.
+ *	Sampling ratios of 2h1v or 2h2v.
+ *	No scaling needed at upsample time.
+ *	Corner-aligned (non-CCIR601) sampling alignment.
+ * Other special cases could be added, but in most applications these are
+ * the only common cases.  (For uncommon cases we fall back on the more
+ * general code in jdsample.c and jdcolor.c.)
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+#ifdef UPSAMPLE_MERGING_SUPPORTED
+
+#ifdef ANDROID_RGB
+
+/* Declarations for ordered dithering.
+ * 
+ * We use 4x4 ordered dither array packed into 32 bits. This array is
+ * sufficent for dithering RGB_888 to RGB_565.
+ */
+
+#define DITHER_MASK         0x3
+#define DITHER_ROTATE(x)    (((x)<<24) | (((x)>>8)&0x00FFFFFF))
+static const INT32 dither_matrix[4] = {
+  0x0008020A,
+  0x0C040E06,
+  0x030B0109,
+  0x0F070D05
+};
+
+#endif
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_upsampler pub;	/* public fields */
+
+  /* Pointer to routine to do actual upsampling/conversion of one row group */
+  JMETHOD(void, upmethod, (j_decompress_ptr cinfo,
+			   JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+			   JSAMPARRAY output_buf));
+
+  /* Private state for YCC->RGB conversion */
+  int * Cr_r_tab;		/* => table for Cr to R conversion */
+  int * Cb_b_tab;		/* => table for Cb to B conversion */
+  INT32 * Cr_g_tab;		/* => table for Cr to G conversion */
+  INT32 * Cb_g_tab;		/* => table for Cb to G conversion */
+
+  /* For 2:1 vertical sampling, we produce two output rows at a time.
+   * We need a "spare" row buffer to hold the second output row if the
+   * application provides just a one-row buffer; we also use the spare
+   * to discard the dummy last row if the image height is odd.
+   */
+  JSAMPROW spare_row;
+  boolean spare_full;		/* T if spare buffer is occupied */
+
+  JDIMENSION out_row_width;	/* samples per output row */
+  JDIMENSION rows_to_go;	/* counts rows remaining in image */
+} my_upsampler;
+
+typedef my_upsampler * my_upsample_ptr;
+
+#define SCALEBITS	16	/* speediest right-shift on some machines */
+#define ONE_HALF	((INT32) 1 << (SCALEBITS-1))
+#define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
+
+
+/*
+ * Initialize tables for YCC->RGB colorspace conversion.
+ * This is taken directly from jdcolor.c; see that file for more info.
+ */
+
+LOCAL(void)
+build_ycc_rgb_table (j_decompress_ptr cinfo)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  int i;
+  INT32 x;
+  SHIFT_TEMPS
+
+  upsample->Cr_r_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(int));
+  upsample->Cb_b_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(int));
+  upsample->Cr_g_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(INT32));
+  upsample->Cb_g_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(INT32));
+
+  for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) {
+    /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */
+    /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */
+    /* Cr=>R value is nearest int to 1.40200 * x */
+    upsample->Cr_r_tab[i] = (int)
+		    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
+    /* Cb=>B value is nearest int to 1.77200 * x */
+    upsample->Cb_b_tab[i] = (int)
+		    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
+    /* Cr=>G value is scaled-up -0.71414 * x */
+    upsample->Cr_g_tab[i] = (- FIX(0.71414)) * x;
+    /* Cb=>G value is scaled-up -0.34414 * x */
+    /* We also add in ONE_HALF so that need not do it in inner loop */
+    upsample->Cb_g_tab[i] = (- FIX(0.34414)) * x + ONE_HALF;
+  }
+}
+
+
+/*
+ * Initialize for an upsampling pass.
+ */
+
+METHODDEF(void)
+start_pass_merged_upsample (j_decompress_ptr cinfo)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+
+  /* Mark the spare buffer empty */
+  upsample->spare_full = FALSE;
+  /* Initialize total-height counter for detecting bottom of image */
+  upsample->rows_to_go = cinfo->output_height;
+}
+
+
+/*
+ * Control routine to do upsampling (and color conversion).
+ *
+ * The control routine just handles the row buffering considerations.
+ */
+
+METHODDEF(void)
+merged_2v_upsample (j_decompress_ptr cinfo,
+		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+		    JDIMENSION in_row_groups_avail,
+		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+		    JDIMENSION out_rows_avail)
+/* 2:1 vertical sampling case: may need a spare row. */
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  JSAMPROW work_ptrs[2];
+  JDIMENSION num_rows;		/* number of rows returned to caller */
+
+  if (upsample->spare_full) {
+    /* If we have a spare row saved from a previous cycle, just return it. */
+      JDIMENSION size = upsample->out_row_width;
+#ifdef ANDROID_RGB
+    if (cinfo->out_color_space == JCS_RGB_565)
+      size = cinfo->output_width*2;
+#endif
+    jcopy_sample_rows(& upsample->spare_row, 0, output_buf + *out_row_ctr, 0,
+		      1, size);
+
+    num_rows = 1;
+    upsample->spare_full = FALSE;
+  } else {
+    /* Figure number of rows to return to caller. */
+    num_rows = 2;
+    /* Not more than the distance to the end of the image. */
+    if (num_rows > upsample->rows_to_go)
+      num_rows = upsample->rows_to_go;
+    /* And not more than what the client can accept: */
+    out_rows_avail -= *out_row_ctr;
+    if (num_rows > out_rows_avail)
+      num_rows = out_rows_avail;
+    /* Create output pointer array for upsampler. */
+    work_ptrs[0] = output_buf[*out_row_ctr];
+    if (num_rows > 1) {
+      work_ptrs[1] = output_buf[*out_row_ctr + 1];
+    } else {
+      work_ptrs[1] = upsample->spare_row;
+      upsample->spare_full = TRUE;
+    }
+    /* Now do the upsampling. */
+    (*upsample->upmethod) (cinfo, input_buf, *in_row_group_ctr, work_ptrs);
+  }
+
+  /* Adjust counts */
+  *out_row_ctr += num_rows;
+  upsample->rows_to_go -= num_rows;
+  /* When the buffer is emptied, declare this input row group consumed */
+  if (! upsample->spare_full)
+    (*in_row_group_ctr)++;
+}
+
+
+METHODDEF(void)
+merged_1v_upsample (j_decompress_ptr cinfo,
+		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+		    JDIMENSION in_row_groups_avail,
+		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+		    JDIMENSION out_rows_avail)
+/* 1:1 vertical sampling case: much easier, never need a spare row. */
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+
+  /* Just do the upsampling. */
+  (*upsample->upmethod) (cinfo, input_buf, *in_row_group_ctr,
+			 output_buf + *out_row_ctr);
+  /* Adjust counts */
+  (*out_row_ctr)++;
+  (*in_row_group_ctr)++;
+}
+
+
+/*
+ * These are the routines invoked by the control routines to do
+ * the actual upsampling/conversion.  One row group is processed per call.
+ *
+ * Note: since we may be writing directly into application-supplied buffers,
+ * we have to be honest about the output width; we can't assume the buffer
+ * has been rounded up to an even width.
+ */
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+ */
+
+METHODDEF(void)
+h2v1_merged_upsample (j_decompress_ptr cinfo,
+		      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+		      JSAMPARRAY output_buf)
+#ifdef ANDROID_JPEG_USE_VENUM
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JSAMPROW outptr;
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+
+#ifdef ANDROID_RGB
+  if (cinfo->out_color_space == JCS_RGBA_8888) {
+    yyvup2abgr8888_venum((UINT8*) inptr0,
+                         (UINT8*) inptr2,
+                         (UINT8*) inptr1,
+                         (UINT8*) outptr,
+                         cinfo->output_width);
+  } else
+#endif
+  {
+    yyvup2bgr888_venum((UINT8*) inptr0,
+                       (UINT8*) inptr2,
+                       (UINT8*) inptr1,
+                       (UINT8*) outptr,
+                       cinfo->output_width);
+  }
+}
+#else
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  INT32 * Crgtab = upsample->Cr_g_tab;
+  INT32 * Cbgtab = upsample->Cb_g_tab;
+  JDIMENSION pixelWidth = RGB_PIXELSIZE;
+  SHIFT_TEMPS
+
+#ifdef ANDROID_RGB
+  if ((cinfo->out_color_components == 4) &&
+      (cinfo->out_color_space == JCS_RGBA_8888))
+     pixelWidth = 4;  // RGBA pixel size
+#endif /* ANDROID_RGB */
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = GETJSAMPLE(*inptr0++);
+    outptr[RGB_RED] = range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] = range_limit[y + cblue];
+    outptr += pixelWidth;
+    y  = GETJSAMPLE(*inptr0++);
+    outptr[RGB_RED] = range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] = range_limit[y + cblue];
+    outptr += pixelWidth;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr0);
+    outptr[RGB_RED] = range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] = range_limit[y + cblue];
+  }
+}
+#endif /* ANDROID_JPEG_USE_VENUM */
+
+
+#ifdef ANDROID_RGB
+METHODDEF(void)
+h2v1_merged_upsample_565 (j_decompress_ptr cinfo,
+              JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+              JSAMPARRAY output_buf)
+#if defined(ANDROID_JPEG_USE_VENUM)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JSAMPROW outptr;
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+
+  yyvup2rgb565_venum((UINT8*) inptr0,
+                     (UINT8*) inptr2,
+                     (UINT8*) inptr1,
+                     (UINT8*) outptr,
+                     cinfo->output_width);
+}
+#else
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  INT32 * Crgtab = upsample->Cr_g_tab;
+  INT32 * Cbgtab = upsample->Cb_g_tab;
+  unsigned int r, g, b;
+  INT32 rgb;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = GETJSAMPLE(*inptr0++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r,g,b);
+    y  = GETJSAMPLE(*inptr0++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r,g,b));
+    WRITE_TWO_PIXELS(outptr, rgb);
+    outptr += 4;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr0);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r,g,b);
+    *(INT16*)outptr = rgb;
+  }
+}
+#endif /* ANDROID_JPEG_USE_VENUM */
+
+
+METHODDEF(void)
+h2v1_merged_upsample_565D (j_decompress_ptr cinfo,
+              JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+              JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  INT32 * Crgtab = upsample->Cr_g_tab;
+  INT32 * Cbgtab = upsample->Cb_g_tab;
+  JDIMENSION col_index = 0;
+  INT32 d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+  unsigned int r, g, b;
+  INT32 rgb;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = GETJSAMPLE(*inptr0++);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_SHORT_565(r,g,b);
+    y  = GETJSAMPLE(*inptr0++);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r,g,b));
+    WRITE_TWO_PIXELS(outptr, rgb);
+    outptr += 4;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr0);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    rgb = PACK_SHORT_565(r,g,b);
+    *(INT16*)outptr = rgb;
+  }
+}
+
+
+#endif
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+ */
+
+METHODDEF(void)
+h2v2_merged_upsample (j_decompress_ptr cinfo,
+		      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+		      JSAMPARRAY output_buf)
+#ifdef ANDROID_JPEG_USE_VENUM
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  inptr00 = input_buf[0][in_row_group_ctr*2];
+  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
+  inptr1  = input_buf[1][in_row_group_ctr];
+  inptr2  = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+#ifdef ANDROID_RGB
+  if (cinfo->out_color_space == JCS_RGBA_8888) {
+    yyvup2abgr8888_venum((UINT8*) inptr00,
+                         (UINT8*) inptr2,
+                         (UINT8*) inptr1,
+                         (UINT8*) outptr0,
+                         cinfo->output_width);
+    yyvup2abgr8888_venum((UINT8*) inptr01,
+                         (UINT8*) inptr2,
+                         (UINT8*) inptr1,
+                         (UINT8*) outptr1,
+                         cinfo->output_width);
+  } else
+#endif
+  {
+    yyvup2bgr888_venum((UINT8*) inptr00,
+                       (UINT8*) inptr2,
+                       (UINT8*) inptr1,
+                       (UINT8*) outptr0,
+                       cinfo->output_width);
+
+    yyvup2bgr888_venum((UINT8*) inptr01,
+                       (UINT8*) inptr2,
+                       (UINT8*) inptr1,
+                       (UINT8*) outptr1,
+                       cinfo->output_width);
+  }
+}
+#else
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  INT32 * Crgtab = upsample->Cr_g_tab;
+  INT32 * Cbgtab = upsample->Cb_g_tab;
+  JDIMENSION pixelWidth = RGB_PIXELSIZE;
+  SHIFT_TEMPS
+
+#ifdef ANDROID_RGB
+  if ((cinfo->out_color_components == 4) &&
+      (cinfo->out_color_space == JCS_RGBA_8888))
+     pixelWidth = 4;  // RGBA pixel size
+#endif /* ANDROID_RGB */
+
+  inptr00 = input_buf[0][in_row_group_ctr*2];
+  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = GETJSAMPLE(*inptr00++);
+    outptr0[RGB_RED] = range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] = range_limit[y + cblue];
+    outptr0 += pixelWidth;
+    y  = GETJSAMPLE(*inptr00++);
+    outptr0[RGB_RED] = range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] = range_limit[y + cblue];
+    outptr0 += pixelWidth;
+    y  = GETJSAMPLE(*inptr01++);
+    outptr1[RGB_RED] = range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] = range_limit[y + cblue];
+    outptr1 += pixelWidth;
+    y  = GETJSAMPLE(*inptr01++);
+    outptr1[RGB_RED] = range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] = range_limit[y + cblue];
+    outptr1 += pixelWidth;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr00);
+    outptr0[RGB_RED] = range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] = range_limit[y + cblue];
+    y  = GETJSAMPLE(*inptr01);
+    outptr1[RGB_RED] = range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] = range_limit[y + cblue];
+  }
+}
+#endif /* ANDROID_JPEG_USE_VENUM */
+
+
+#ifdef ANDROID_RGB
+
+METHODDEF(void)
+h2v2_merged_upsample_565 (j_decompress_ptr cinfo,
+              JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+              JSAMPARRAY output_buf)
+#if defined(ANDROID_JPEG_USE_VENUM)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  inptr00 = input_buf[0][in_row_group_ctr*2];
+  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
+  inptr1  = input_buf[1][in_row_group_ctr];
+  inptr2  = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+
+  yyvup2rgb565_venum((UINT8*) inptr00,
+                     (UINT8*) inptr2,
+                     (UINT8*) inptr1,
+                     (UINT8*) outptr0,
+                     cinfo->output_width);
+
+  yyvup2rgb565_venum((UINT8*) inptr01,
+                     (UINT8*) inptr2,
+                     (UINT8*) inptr1,
+                     (UINT8*) outptr1,
+                     cinfo->output_width);
+}
+#else
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  INT32 * Crgtab = upsample->Cr_g_tab;
+  INT32 * Cbgtab = upsample->Cb_g_tab;
+  unsigned int r, g, b;
+  INT32 rgb;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf[0][in_row_group_ctr*2];
+  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = GETJSAMPLE(*inptr00++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r,g,b);
+    y  = GETJSAMPLE(*inptr00++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r,g,b));
+    WRITE_TWO_PIXELS(outptr0, rgb);
+    outptr0 += 4;
+    y  = GETJSAMPLE(*inptr01++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r,g,b);
+    y  = GETJSAMPLE(*inptr01++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r,g,b));
+    WRITE_TWO_PIXELS(outptr1, rgb);
+    outptr1 += 4;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr00);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r,g,b);
+    *(INT16*)outptr0 = rgb;
+   y  = GETJSAMPLE(*inptr01);
+   r = range_limit[y + cred];
+   g = range_limit[y + cgreen];
+   b = range_limit[y + cblue];
+   rgb = PACK_SHORT_565(r,g,b);
+   *(INT16*)outptr1 = rgb;
+  }
+}
+#endif /* ANDROID_JPEG_USE_VENUM */
+
+
+METHODDEF(void)
+h2v2_merged_upsample_565D (j_decompress_ptr cinfo,
+              JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+              JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  INT32 * Crgtab = upsample->Cr_g_tab;
+  INT32 * Cbgtab = upsample->Cb_g_tab;
+  JDIMENSION col_index = 0;
+  INT32 d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+  INT32 d1 = dither_matrix[(cinfo->output_scanline+1) & DITHER_MASK];
+  unsigned int r, g, b;
+  INT32 rgb;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf[0][in_row_group_ctr*2];
+  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 4 Y values and emit 4 pixels */    
+    y  = GETJSAMPLE(*inptr00++);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_SHORT_565(r,g,b);
+    y  = GETJSAMPLE(*inptr00++);
+    r = range_limit[DITHER_565_R(y + cred, d1)];
+    g = range_limit[DITHER_565_G(y + cgreen, d1)];
+    b = range_limit[DITHER_565_B(y + cblue, d1)];
+    d1 = DITHER_ROTATE(d1);
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r,g,b));
+    WRITE_TWO_PIXELS(outptr0, rgb);
+    outptr0 += 4;
+    y  = GETJSAMPLE(*inptr01++);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_SHORT_565(r,g,b);
+    y  = GETJSAMPLE(*inptr01++);
+    r = range_limit[DITHER_565_R(y + cred, d1)];
+    g = range_limit[DITHER_565_G(y + cgreen, d1)];
+    b = range_limit[DITHER_565_B(y + cblue, d1)];
+    d1 = DITHER_ROTATE(d1);
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r,g,b));
+    WRITE_TWO_PIXELS(outptr1, rgb);
+    outptr1 += 4;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr00);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    rgb = PACK_SHORT_565(r,g,b);
+    *(INT16*)outptr0 = rgb;
+   y  = GETJSAMPLE(*inptr01);
+   r = range_limit[DITHER_565_R(y + cred, d1)];
+   g = range_limit[DITHER_565_G(y + cgreen, d1)];
+   b = range_limit[DITHER_565_B(y + cblue, d1)];
+   rgb = PACK_SHORT_565(r,g,b);
+   *(INT16*)outptr1 = rgb;
+  }
+}
+
+#endif
+
+/*
+ * Module initialization routine for merged upsampling/color conversion.
+ *
+ * NB: this is called under the conditions determined by use_merged_upsample()
+ * in jdmaster.c.  That routine MUST correspond to the actual capabilities
+ * of this module; no safety checks are made here.
+ */
+
+GLOBAL(void)
+jinit_merged_upsampler (j_decompress_ptr cinfo)
+{
+  my_upsample_ptr upsample;
+
+  upsample = (my_upsample_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_upsampler));
+  cinfo->upsample = (struct jpeg_upsampler *) upsample;
+  upsample->pub.start_pass = start_pass_merged_upsample;
+  upsample->pub.need_context_rows = FALSE;
+
+  upsample->out_row_width = cinfo->output_width * cinfo->out_color_components;
+  
+  if (cinfo->max_v_samp_factor == 2) {
+    upsample->pub.upsample = merged_2v_upsample;
+    upsample->upmethod = h2v2_merged_upsample;
+#ifdef ANDROID_RGB
+    if (cinfo->out_color_space == JCS_RGB_565) {
+#if !defined(ANDROID_JPEG_USE_VENUM)
+      if (cinfo->dither_mode != JDITHER_NONE) {
+        upsample->upmethod = h2v2_merged_upsample_565D;
+      } else
+#endif
+      {
+        /* If VeNum routines are enabled, use h2v2_merged_upsample_565
+         * function regardless of dither mode. */
+        upsample->upmethod = h2v2_merged_upsample_565;
+      }
+    }
+#endif
+    /* Allocate a spare row buffer */
+    upsample->spare_row = (JSAMPROW)
+      (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+		(size_t) (upsample->out_row_width * SIZEOF(JSAMPLE)));
+  } else {
+    upsample->pub.upsample = merged_1v_upsample;
+    upsample->upmethod = h2v1_merged_upsample;
+#ifdef ANDROID_RGB
+    if (cinfo->out_color_space == JCS_RGB_565) {
+#ifndef ANDROID_JPEG_USE_VENUM
+      if (cinfo->dither_mode != JDITHER_NONE) {
+        upsample->upmethod = h2v1_merged_upsample_565D;
+      } else
+#endif
+      {
+        /* If VeNum routines are enabled, use h2v1_merged_upsample_565
+         * function regardless of dither mode. */
+        upsample->upmethod = h2v1_merged_upsample_565;
+      }
+    }
+#endif
+    /* No spare row needed */
+    upsample->spare_row = NULL;
+  }
+
+#if !defined(ANDROID_JPEG_USE_VENUM)
+    build_ycc_rgb_table(cinfo);
+#endif
+#ifdef NEEDS_ARM_ERRATA_754319_754320
+  asm volatile ( "vmov s0,s0\n" );
+#endif
+}
+
+#endif /* UPSAMPLE_MERGING_SUPPORTED */
diff --git a/libjpegtwrp/jdphuff.c b/libjpegtwrp/jdphuff.c
new file mode 100644
index 000000000..2f856e07b
--- /dev/null
+++ b/libjpegtwrp/jdphuff.c
@@ -0,0 +1,770 @@
+/*
+ * jdphuff.c
+ *
+ * Copyright (C) 1995-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains Huffman entropy decoding routines for progressive JPEG.
+ *
+ * Much of the complexity here has to do with supporting input suspension.
+ * If the data source module demands suspension, we want to be able to back
+ * up to the start of the current MCU.  To do this, we copy state variables
+ * into local working storage, and update them back to the permanent
+ * storage only upon successful completion of an MCU.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdhuff.h"		/* Declarations shared with jdhuff.c */
+
+
+#ifdef D_PROGRESSIVE_SUPPORTED
+
+/*
+ * Expanded entropy decoder object for progressive Huffman decoding.
+ *
+ * The savable_state subrecord contains fields that change within an MCU,
+ * but must not be updated permanently until we complete the MCU.
+ */
+
+typedef struct {
+  unsigned int EOBRUN;			/* remaining EOBs in EOBRUN */
+  int last_dc_val[MAX_COMPS_IN_SCAN];	/* last DC coef for each component */
+} savable_state;
+
+/* This macro is to work around compilers with missing or broken
+ * structure assignment.  You'll need to fix this code if you have
+ * such a compiler and you change MAX_COMPS_IN_SCAN.
+ */
+
+#ifndef NO_STRUCT_ASSIGN
+#define ASSIGN_STATE(dest,src)  ((dest) = (src))
+#else
+#if MAX_COMPS_IN_SCAN == 4
+#define ASSIGN_STATE(dest,src)  \
+	((dest).EOBRUN = (src).EOBRUN, \
+	 (dest).last_dc_val[0] = (src).last_dc_val[0], \
+	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
+	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
+	 (dest).last_dc_val[3] = (src).last_dc_val[3])
+#endif
+#endif
+
+
+typedef struct {
+  struct jpeg_entropy_decoder pub; /* public fields */
+
+  /* These fields are loaded into local variables at start of each MCU.
+   * In case of suspension, we exit WITHOUT updating them.
+   */
+  bitread_perm_state bitstate;	/* Bit buffer at start of MCU */
+  savable_state saved;		/* Other state at start of MCU */
+
+  /* These fields are NOT loaded into local working state. */
+  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+
+  /* Pointers to derived tables (these workspaces have image lifespan) */
+  d_derived_tbl * derived_tbls[NUM_HUFF_TBLS];
+
+  d_derived_tbl * ac_derived_tbl; /* active table during an AC scan */
+} phuff_entropy_decoder;
+
+typedef phuff_entropy_decoder * phuff_entropy_ptr;
+
+/* Forward declarations */
+METHODDEF(boolean) decode_mcu_DC_first JPP((j_decompress_ptr cinfo,
+					    JBLOCKROW *MCU_data));
+METHODDEF(boolean) decode_mcu_AC_first JPP((j_decompress_ptr cinfo,
+					    JBLOCKROW *MCU_data));
+METHODDEF(boolean) decode_mcu_DC_refine JPP((j_decompress_ptr cinfo,
+					     JBLOCKROW *MCU_data));
+METHODDEF(boolean) decode_mcu_AC_refine JPP((j_decompress_ptr cinfo,
+					     JBLOCKROW *MCU_data));
+
+/*
+ * Initialize for a Huffman-compressed scan.
+ */
+
+METHODDEF(void)
+start_pass_phuff_decoder (j_decompress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  boolean is_DC_band, bad;
+  int ci, coefi, tbl;
+  int *coef_bit_ptr;
+  jpeg_component_info * compptr;
+
+  is_DC_band = (cinfo->Ss == 0);
+
+  /* Validate scan parameters */
+  bad = FALSE;
+  if (is_DC_band) {
+    if (cinfo->Se != 0)
+      bad = TRUE;
+  } else {
+    /* need not check Ss/Se < 0 since they came from unsigned bytes */
+    if (cinfo->Ss > cinfo->Se || cinfo->Se >= DCTSIZE2)
+      bad = TRUE;
+    /* AC scans may have only one component */
+    if (cinfo->comps_in_scan != 1)
+      bad = TRUE;
+  }
+  if (cinfo->Ah != 0) {
+    /* Successive approximation refinement scan: must have Al = Ah-1. */
+    if (cinfo->Al != cinfo->Ah-1)
+      bad = TRUE;
+  }
+  if (cinfo->Al > 13)		/* need not check for < 0 */
+    bad = TRUE;
+  /* Arguably the maximum Al value should be less than 13 for 8-bit precision,
+   * but the spec doesn't say so, and we try to be liberal about what we
+   * accept.  Note: large Al values could result in out-of-range DC
+   * coefficients during early scans, leading to bizarre displays due to
+   * overflows in the IDCT math.  But we won't crash.
+   */
+  if (bad)
+    ERREXIT4(cinfo, JERR_BAD_PROGRESSION,
+	     cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
+  /* Update progression status, and verify that scan order is legal.
+   * Note that inter-scan inconsistencies are treated as warnings
+   * not fatal errors ... not clear if this is right way to behave.
+   */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    int cindex = cinfo->cur_comp_info[ci]->component_index;
+    coef_bit_ptr = & cinfo->coef_bits[cindex][0];
+    if (!is_DC_band && coef_bit_ptr[0] < 0) /* AC without prior DC scan */
+      WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
+    for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
+      int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
+      if (cinfo->Ah != expected)
+	WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
+      coef_bit_ptr[coefi] = cinfo->Al;
+    }
+  }
+
+  /* Select MCU decoding routine */
+  if (cinfo->Ah == 0) {
+    if (is_DC_band)
+      entropy->pub.decode_mcu = decode_mcu_DC_first;
+    else
+      entropy->pub.decode_mcu = decode_mcu_AC_first;
+  } else {
+    if (is_DC_band)
+      entropy->pub.decode_mcu = decode_mcu_DC_refine;
+    else
+      entropy->pub.decode_mcu = decode_mcu_AC_refine;
+  }
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    /* Make sure requested tables are present, and compute derived tables.
+     * We may build same derived table more than once, but it's not expensive.
+     */
+    if (is_DC_band) {
+      if (cinfo->Ah == 0) {	/* DC refinement needs no table */
+	tbl = compptr->dc_tbl_no;
+	jpeg_make_d_derived_tbl(cinfo, TRUE, tbl,
+				& entropy->derived_tbls[tbl]);
+      }
+    } else {
+      tbl = compptr->ac_tbl_no;
+      jpeg_make_d_derived_tbl(cinfo, FALSE, tbl,
+			      & entropy->derived_tbls[tbl]);
+      /* remember the single active table */
+      entropy->ac_derived_tbl = entropy->derived_tbls[tbl];
+    }
+    /* Initialize DC predictions to 0 */
+    entropy->saved.last_dc_val[ci] = 0;
+  }
+
+  /* Initialize bitread state variables */
+  entropy->bitstate.bits_left = 0;
+  entropy->bitstate.get_buffer = 0; /* unnecessary, but keeps Purify quiet */
+  entropy->pub.insufficient_data = FALSE;
+
+  /* Initialize private state variables */
+  entropy->saved.EOBRUN = 0;
+
+  /* Initialize restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+}
+
+
+/*
+ * Figure F.12: extend sign bit.
+ * On some machines, a shift and add will be faster than a table lookup.
+ */
+
+#ifdef AVOID_TABLES
+
+#define HUFF_EXTEND(x,s)  ((x) < (1<<((s)-1)) ? (x) + (((-1)<<(s)) + 1) : (x))
+
+#else
+
+#define HUFF_EXTEND(x,s)  ((x) < extend_test[s] ? (x) + extend_offset[s] : (x))
+
+static const int extend_test[16] =   /* entry n is 2**(n-1) */
+  { 0, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080,
+    0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000 };
+
+static const int extend_offset[16] = /* entry n is (-1 << n) + 1 */
+  { 0, ((-1)<<1) + 1, ((-1)<<2) + 1, ((-1)<<3) + 1, ((-1)<<4) + 1,
+    ((-1)<<5) + 1, ((-1)<<6) + 1, ((-1)<<7) + 1, ((-1)<<8) + 1,
+    ((-1)<<9) + 1, ((-1)<<10) + 1, ((-1)<<11) + 1, ((-1)<<12) + 1,
+    ((-1)<<13) + 1, ((-1)<<14) + 1, ((-1)<<15) + 1 };
+
+#endif /* AVOID_TABLES */
+
+
+/*
+ * Check for a restart marker & resynchronize decoder.
+ * Returns FALSE if must suspend.
+ */
+
+LOCAL(boolean)
+process_restart (j_decompress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  int ci;
+
+  /* Throw away any unused bits remaining in bit buffer; */
+  /* include any full bytes in next_marker's count of discarded bytes */
+  cinfo->marker->discarded_bytes += entropy->bitstate.bits_left / 8;
+  entropy->bitstate.bits_left = 0;
+
+  /* Advance past the RSTn marker */
+  if (! (*cinfo->marker->read_restart_marker) (cinfo))
+    return FALSE;
+
+  /* Re-initialize DC predictions to 0 */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++)
+    entropy->saved.last_dc_val[ci] = 0;
+  /* Re-init EOB run count, too */
+  entropy->saved.EOBRUN = 0;
+
+  /* Reset restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+
+  /* Reset out-of-data flag, unless read_restart_marker left us smack up
+   * against a marker.  In that case we will end up treating the next data
+   * segment as empty, and we can avoid producing bogus output pixels by
+   * leaving the flag set.
+   */
+  if (cinfo->unread_marker == 0)
+    entropy->pub.insufficient_data = FALSE;
+
+  return TRUE;
+}
+
+
+/*
+ * Huffman MCU decoding.
+ * Each of these routines decodes and returns one MCU's worth of
+ * Huffman-compressed coefficients. 
+ * The coefficients are reordered from zigzag order into natural array order,
+ * but are not dequantized.
+ *
+ * The i'th block of the MCU is stored into the block pointed to by
+ * MCU_data[i].  WE ASSUME THIS AREA IS INITIALLY ZEROED BY THE CALLER.
+ *
+ * We return FALSE if data source requested suspension.  In that case no
+ * changes have been made to permanent state.  (Exception: some output
+ * coefficients may already have been assigned.  This is harmless for
+ * spectral selection, since we'll just re-assign them on the next call.
+ * Successive approximation AC refinement has to be more careful, however.)
+ */
+
+/*
+ * MCU decoding for DC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+decode_mcu_DC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{   
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  int Al = cinfo->Al;
+  register int s, r;
+  int blkn, ci;
+  JBLOCKROW block;
+  BITREAD_STATE_VARS;
+  savable_state state;
+  d_derived_tbl * tbl;
+  jpeg_component_info * compptr;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return FALSE;
+  }
+
+  /* If we've run out of data, just leave the MCU set to zeroes.
+   * This way, we return uniform gray for the remainder of the segment.
+   */
+  if (! entropy->pub.insufficient_data) {
+
+    /* Load up working state */
+    BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
+    ASSIGN_STATE(state, entropy->saved);
+
+    /* Outer loop handles each block in the MCU */
+
+    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+      block = MCU_data[blkn];
+      ci = cinfo->MCU_membership[blkn];
+      compptr = cinfo->cur_comp_info[ci];
+      tbl = entropy->derived_tbls[compptr->dc_tbl_no];
+
+      /* Decode a single block's worth of coefficients */
+
+      /* Section F.2.2.1: decode the DC coefficient difference */
+      HUFF_DECODE(s, br_state, tbl, return FALSE, label1);
+      if (s) {
+	CHECK_BIT_BUFFER(br_state, s, return FALSE);
+	r = GET_BITS(s);
+	s = HUFF_EXTEND(r, s);
+      }
+
+      /* Convert DC difference to actual value, update last_dc_val */
+      s += state.last_dc_val[ci];
+      state.last_dc_val[ci] = s;
+      /* Scale and output the coefficient (assumes jpeg_natural_order[0]=0) */
+      (*block)[0] = (JCOEF) (s << Al);
+    }
+
+    /* Completed MCU, so update state */
+    BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
+    ASSIGN_STATE(entropy->saved, state);
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+
+/*
+ * MCU decoding for AC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+decode_mcu_AC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{   
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  int Se = cinfo->Se;
+  int Al = cinfo->Al;
+  register int s, k, r;
+  unsigned int EOBRUN;
+  JBLOCKROW block;
+  BITREAD_STATE_VARS;
+  d_derived_tbl * tbl;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return FALSE;
+  }
+
+  /* If we've run out of data, just leave the MCU set to zeroes.
+   * This way, we return uniform gray for the remainder of the segment.
+   */
+  if (! entropy->pub.insufficient_data) {
+
+    /* Load up working state.
+     * We can avoid loading/saving bitread state if in an EOB run.
+     */
+    EOBRUN = entropy->saved.EOBRUN;	/* only part of saved state we need */
+
+    /* There is always only one block per MCU */
+
+    if (EOBRUN > 0)		/* if it's a band of zeroes... */
+      EOBRUN--;			/* ...process it now (we do nothing) */
+    else {
+      BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
+      block = MCU_data[0];
+      tbl = entropy->ac_derived_tbl;
+
+      for (k = cinfo->Ss; k <= Se; k++) {
+	HUFF_DECODE(s, br_state, tbl, return FALSE, label2);
+	r = s >> 4;
+	s &= 15;
+	if (s) {
+	  k += r;
+	  CHECK_BIT_BUFFER(br_state, s, return FALSE);
+	  r = GET_BITS(s);
+	  s = HUFF_EXTEND(r, s);
+	  /* Scale and output coefficient in natural (dezigzagged) order */
+	  (*block)[jpeg_natural_order[k]] = (JCOEF) (s << Al);
+	} else {
+	  if (r == 15) {	/* ZRL */
+	    k += 15;		/* skip 15 zeroes in band */
+	  } else {		/* EOBr, run length is 2^r + appended bits */
+	    EOBRUN = 1 << r;
+	    if (r) {		/* EOBr, r > 0 */
+	      CHECK_BIT_BUFFER(br_state, r, return FALSE);
+	      r = GET_BITS(r);
+	      EOBRUN += r;
+	    }
+	    EOBRUN--;		/* this band is processed at this moment */
+	    break;		/* force end-of-band */
+	  }
+	}
+      }
+
+      BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
+    }
+
+    /* Completed MCU, so update state */
+    entropy->saved.EOBRUN = EOBRUN;	/* only part of saved state we need */
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+
+/*
+ * MCU decoding for DC successive approximation refinement scan.
+ * Note: we assume such scans can be multi-component, although the spec
+ * is not very clear on the point.
+ */
+
+METHODDEF(boolean)
+decode_mcu_DC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{   
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  int p1 = 1 << cinfo->Al;	/* 1 in the bit position being coded */
+  int blkn;
+  JBLOCKROW block;
+  BITREAD_STATE_VARS;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return FALSE;
+  }
+
+  /* Not worth the cycles to check insufficient_data here,
+   * since we will not change the data anyway if we read zeroes.
+   */
+
+  /* Load up working state */
+  BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
+
+  /* Outer loop handles each block in the MCU */
+
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_data[blkn];
+
+    /* Encoded data is simply the next bit of the two's-complement DC value */
+    CHECK_BIT_BUFFER(br_state, 1, return FALSE);
+    if (GET_BITS(1))
+      (*block)[0] |= p1;
+    /* Note: since we use |=, repeating the assignment later is safe */
+  }
+
+  /* Completed MCU, so update state */
+  BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
+
+  /* Account for restart interval (no-op if not using restarts) */
+  entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+
+/*
+ * MCU decoding for AC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+decode_mcu_AC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{   
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  int Se = cinfo->Se;
+  int p1 = 1 << cinfo->Al;	/* 1 in the bit position being coded */
+  int m1 = (-1) << cinfo->Al;	/* -1 in the bit position being coded */
+  register int s, k, r;
+  unsigned int EOBRUN;
+  JBLOCKROW block;
+  JCOEFPTR thiscoef;
+  BITREAD_STATE_VARS;
+  d_derived_tbl * tbl;
+  int num_newnz;
+  int newnz_pos[DCTSIZE2];
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return FALSE;
+  }
+
+  /* If we've run out of data, don't modify the MCU.
+   */
+  if (! entropy->pub.insufficient_data) {
+
+    /* Load up working state */
+    BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
+    EOBRUN = entropy->saved.EOBRUN; /* only part of saved state we need */
+
+    /* There is always only one block per MCU */
+    block = MCU_data[0];
+    tbl = entropy->ac_derived_tbl;
+
+    /* If we are forced to suspend, we must undo the assignments to any newly
+     * nonzero coefficients in the block, because otherwise we'd get confused
+     * next time about which coefficients were already nonzero.
+     * But we need not undo addition of bits to already-nonzero coefficients;
+     * instead, we can test the current bit to see if we already did it.
+     */
+    num_newnz = 0;
+
+    /* initialize coefficient loop counter to start of band */
+    k = cinfo->Ss;
+
+    if (EOBRUN == 0) {
+      for (; k <= Se; k++) {
+	HUFF_DECODE(s, br_state, tbl, goto undoit, label3);
+	r = s >> 4;
+	s &= 15;
+	if (s) {
+	  if (s != 1)		/* size of new coef should always be 1 */
+	    WARNMS(cinfo, JWRN_HUFF_BAD_CODE);
+	  CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+	  if (GET_BITS(1))
+	    s = p1;		/* newly nonzero coef is positive */
+	  else
+	    s = m1;		/* newly nonzero coef is negative */
+	} else {
+	  if (r != 15) {
+	    EOBRUN = 1 << r;	/* EOBr, run length is 2^r + appended bits */
+	    if (r) {
+	      CHECK_BIT_BUFFER(br_state, r, goto undoit);
+	      r = GET_BITS(r);
+	      EOBRUN += r;
+	    }
+	    break;		/* rest of block is handled by EOB logic */
+	  }
+	  /* note s = 0 for processing ZRL */
+	}
+	/* Advance over already-nonzero coefs and r still-zero coefs,
+	 * appending correction bits to the nonzeroes.  A correction bit is 1
+	 * if the absolute value of the coefficient must be increased.
+	 */
+	do {
+	  thiscoef = *block + jpeg_natural_order[k];
+	  if (*thiscoef != 0) {
+	    CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+	    if (GET_BITS(1)) {
+	      if ((*thiscoef & p1) == 0) { /* do nothing if already set it */
+		if (*thiscoef >= 0)
+		  *thiscoef += p1;
+		else
+		  *thiscoef += m1;
+	      }
+	    }
+	  } else {
+	    if (--r < 0)
+	      break;		/* reached target zero coefficient */
+	  }
+	  k++;
+	} while (k <= Se);
+	if (s) {
+	  int pos = jpeg_natural_order[k];
+	  /* Output newly nonzero coefficient */
+	  (*block)[pos] = (JCOEF) s;
+	  /* Remember its position in case we have to suspend */
+	  newnz_pos[num_newnz++] = pos;
+	}
+      }
+    }
+
+    if (EOBRUN > 0) {
+      /* Scan any remaining coefficient positions after the end-of-band
+       * (the last newly nonzero coefficient, if any).  Append a correction
+       * bit to each already-nonzero coefficient.  A correction bit is 1
+       * if the absolute value of the coefficient must be increased.
+       */
+      for (; k <= Se; k++) {
+	thiscoef = *block + jpeg_natural_order[k];
+	if (*thiscoef != 0) {
+	  CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+	  if (GET_BITS(1)) {
+	    if ((*thiscoef & p1) == 0) { /* do nothing if already changed it */
+	      if (*thiscoef >= 0)
+		*thiscoef += p1;
+	      else
+		*thiscoef += m1;
+	    }
+	  }
+	}
+      }
+      /* Count one block completed in EOB run */
+      EOBRUN--;
+    }
+
+    /* Completed MCU, so update state */
+    BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
+    entropy->saved.EOBRUN = EOBRUN; /* only part of saved state we need */
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  entropy->restarts_to_go--;
+
+  return TRUE;
+
+undoit:
+  /* Re-zero any output coefficients that we made newly nonzero */
+  while (num_newnz > 0)
+    (*block)[newnz_pos[--num_newnz]] = 0;
+
+  return FALSE;
+}
+
+/*
+ * Save the current Huffman decoder position and the bit buffer
+ * into bitstream_offset and get_buffer, respectively.
+ */
+GLOBAL(void)
+jpeg_get_huffman_decoder_configuration_progressive(j_decompress_ptr cinfo,
+        huffman_offset_data *offset)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+
+  if (cinfo->restart_interval) {
+    // We are at the end of a data segment
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return;
+  }
+
+  // Save restarts_to_go and next_restart_num.
+  offset->restarts_to_go = (unsigned short) entropy->restarts_to_go;
+  offset->next_restart_num = cinfo->marker->next_restart_num;
+
+  offset->bitstream_offset =
+      (jget_input_stream_position(cinfo) << LOG_TWO_BIT_BUF_SIZE)
+      + entropy->bitstate.bits_left;
+
+  offset->get_buffer = entropy->bitstate.get_buffer;
+}
+
+/*
+ * Save the current Huffman deocde position and the DC coefficients
+ * for each component into bitstream_offset and dc_info[], respectively.
+ */
+METHODDEF(void)
+get_huffman_decoder_configuration(j_decompress_ptr cinfo,
+        huffman_offset_data *offset)
+{
+  int i;
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  jpeg_get_huffman_decoder_configuration_progressive(cinfo, offset);
+  offset->EOBRUN = entropy->saved.EOBRUN;
+  for (i = 0; i < cinfo->comps_in_scan; i++)
+    offset->prev_dc[i] = entropy->saved.last_dc_val[i];
+}
+
+/*
+ * Configure the Huffman decoder reader position and bit buffer.
+ */
+GLOBAL(void)
+jpeg_configure_huffman_decoder_progressive(j_decompress_ptr cinfo,
+        huffman_offset_data offset)
+{
+	phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+
+  // Restore restarts_to_go and next_restart_num
+  cinfo->unread_marker = 0;
+  entropy->restarts_to_go = offset.restarts_to_go;
+  cinfo->marker->next_restart_num = offset.next_restart_num;
+
+  unsigned int bitstream_offset = offset.bitstream_offset;
+  int blkn, i;
+
+  unsigned int byte_offset = bitstream_offset >> LOG_TWO_BIT_BUF_SIZE;
+  unsigned int bit_in_bit_buffer =
+      bitstream_offset & ((1 << LOG_TWO_BIT_BUF_SIZE) - 1);
+
+  jset_input_stream_position_bit(cinfo, byte_offset,
+          bit_in_bit_buffer, offset.get_buffer);
+}
+
+/*
+ * Configure the Huffman decoder to decode the image
+ * starting from (iMCU_row_offset, iMCU_col_offset).
+ */
+METHODDEF(void)
+configure_huffman_decoder(j_decompress_ptr cinfo, huffman_offset_data offset)
+{
+  int i;
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  jpeg_configure_huffman_decoder_progressive(cinfo, offset);
+  entropy->saved.EOBRUN = offset.EOBRUN;
+  for (i = 0; i < cinfo->comps_in_scan; i++)
+    entropy->saved.last_dc_val[i] = offset.prev_dc[i];
+}
+
+GLOBAL(void)
+jpeg_configure_huffman_index_scan(j_decompress_ptr cinfo,
+        huffman_index *index, int scan_no, int offset)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  if (scan_no >= index->scan_count) {
+    index->scan = realloc(index->scan,
+                    (scan_no + 1) * sizeof(huffman_scan_header));
+    index->mem_used += (scan_no - index->scan_count + 1)
+      * (sizeof(huffman_scan_header) + cinfo->total_iMCU_rows
+      * sizeof(huffman_offset_data*));
+    index->scan_count = scan_no + 1;
+  }
+  index->scan[scan_no].offset = (huffman_offset_data**)malloc(
+          cinfo->total_iMCU_rows * sizeof(huffman_offset_data*));
+  index->scan[scan_no].bitstream_offset = offset;
+}
+
+/*
+ * Module initialization routine for progressive Huffman entropy decoding.
+ */
+GLOBAL(void)
+jinit_phuff_decoder (j_decompress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy;
+  int *coef_bit_ptr;
+  int ci, i;
+
+  entropy = (phuff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(phuff_entropy_decoder));
+  cinfo->entropy = (struct jpeg_entropy_decoder *) entropy;
+  entropy->pub.start_pass = start_pass_phuff_decoder;
+  entropy->pub.configure_huffman_decoder = configure_huffman_decoder;
+  entropy->pub.get_huffman_decoder_configuration =
+        get_huffman_decoder_configuration;
+
+  /* Mark derived tables unallocated */
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    entropy->derived_tbls[i] = NULL;
+  }
+
+  /* Create progression status table */
+  cinfo->coef_bits = (int (*)[DCTSIZE2])
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				cinfo->num_components*DCTSIZE2*SIZEOF(int));
+  coef_bit_ptr = & cinfo->coef_bits[0][0];
+  for (ci = 0; ci < cinfo->num_components; ci++) 
+    for (i = 0; i < DCTSIZE2; i++)
+      *coef_bit_ptr++ = -1;
+}
+
+#endif /* D_PROGRESSIVE_SUPPORTED */
diff --git a/libjpegtwrp/jdpostct.c b/libjpegtwrp/jdpostct.c
new file mode 100644
index 000000000..571563d72
--- /dev/null
+++ b/libjpegtwrp/jdpostct.c
@@ -0,0 +1,290 @@
+/*
+ * jdpostct.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the decompression postprocessing controller.
+ * This controller manages the upsampling, color conversion, and color
+ * quantization/reduction steps; specifically, it controls the buffering
+ * between upsample/color conversion and color quantization/reduction.
+ *
+ * If no color quantization/reduction is required, then this module has no
+ * work to do, and it just hands off to the upsample/color conversion code.
+ * An integrated upsample/convert/quantize process would replace this module
+ * entirely.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_post_controller pub; /* public fields */
+
+  /* Color quantization source buffer: this holds output data from
+   * the upsample/color conversion step to be passed to the quantizer.
+   * For two-pass color quantization, we need a full-image buffer;
+   * for one-pass operation, a strip buffer is sufficient.
+   */
+  jvirt_sarray_ptr whole_image;	/* virtual array, or NULL if one-pass */
+  JSAMPARRAY buffer;		/* strip buffer, or current strip of virtual */
+  JDIMENSION strip_height;	/* buffer size in rows */
+  /* for two-pass mode only: */
+  JDIMENSION starting_row;	/* row # of first row in current strip */
+  JDIMENSION next_row;		/* index of next row to fill/empty in strip */
+} my_post_controller;
+
+typedef my_post_controller * my_post_ptr;
+
+
+/* Forward declarations */
+METHODDEF(void) post_process_1pass
+	JPP((j_decompress_ptr cinfo,
+	     JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+	     JDIMENSION in_row_groups_avail,
+	     JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+	     JDIMENSION out_rows_avail));
+#ifdef QUANT_2PASS_SUPPORTED
+METHODDEF(void) post_process_prepass
+	JPP((j_decompress_ptr cinfo,
+	     JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+	     JDIMENSION in_row_groups_avail,
+	     JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+	     JDIMENSION out_rows_avail));
+METHODDEF(void) post_process_2pass
+	JPP((j_decompress_ptr cinfo,
+	     JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+	     JDIMENSION in_row_groups_avail,
+	     JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+	     JDIMENSION out_rows_avail));
+#endif
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_dpost (j_decompress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_post_ptr post = (my_post_ptr) cinfo->post;
+
+  switch (pass_mode) {
+  case JBUF_PASS_THRU:
+    if (cinfo->quantize_colors) {
+      /* Single-pass processing with color quantization. */
+      post->pub.post_process_data = post_process_1pass;
+      /* We could be doing buffered-image output before starting a 2-pass
+       * color quantization; in that case, jinit_d_post_controller did not
+       * allocate a strip buffer.  Use the virtual-array buffer as workspace.
+       */
+      if (post->buffer == NULL) {
+	post->buffer = (*cinfo->mem->access_virt_sarray)
+	  ((j_common_ptr) cinfo, post->whole_image,
+	   (JDIMENSION) 0, post->strip_height, TRUE);
+      }
+    } else {
+      /* For single-pass processing without color quantization,
+       * I have no work to do; just call the upsampler directly.
+       */
+      post->pub.post_process_data = cinfo->upsample->upsample;
+    }
+    break;
+#ifdef QUANT_2PASS_SUPPORTED
+  case JBUF_SAVE_AND_PASS:
+    /* First pass of 2-pass quantization */
+    if (post->whole_image == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    post->pub.post_process_data = post_process_prepass;
+    break;
+  case JBUF_CRANK_DEST:
+    /* Second pass of 2-pass quantization */
+    if (post->whole_image == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    post->pub.post_process_data = post_process_2pass;
+    break;
+#endif /* QUANT_2PASS_SUPPORTED */
+  default:
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    break;
+  }
+  post->starting_row = post->next_row = 0;
+}
+
+
+/*
+ * Process some data in the one-pass (strip buffer) case.
+ * This is used for color precision reduction as well as one-pass quantization.
+ */
+
+METHODDEF(void)
+post_process_1pass (j_decompress_ptr cinfo,
+		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+		    JDIMENSION in_row_groups_avail,
+		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+		    JDIMENSION out_rows_avail)
+{
+  my_post_ptr post = (my_post_ptr) cinfo->post;
+  JDIMENSION num_rows, max_rows;
+
+  /* Fill the buffer, but not more than what we can dump out in one go. */
+  /* Note we rely on the upsampler to detect bottom of image. */
+  max_rows = out_rows_avail - *out_row_ctr;
+  if (max_rows > post->strip_height)
+    max_rows = post->strip_height;
+  num_rows = 0;
+  (*cinfo->upsample->upsample) (cinfo,
+		input_buf, in_row_group_ctr, in_row_groups_avail,
+		post->buffer, &num_rows, max_rows);
+  /* Quantize and emit data. */
+  (*cinfo->cquantize->color_quantize) (cinfo,
+		post->buffer, output_buf + *out_row_ctr, (int) num_rows);
+  *out_row_ctr += num_rows;
+}
+
+
+#ifdef QUANT_2PASS_SUPPORTED
+
+/*
+ * Process some data in the first pass of 2-pass quantization.
+ */
+
+METHODDEF(void)
+post_process_prepass (j_decompress_ptr cinfo,
+		      JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+		      JDIMENSION in_row_groups_avail,
+		      JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+		      JDIMENSION out_rows_avail)
+{
+  my_post_ptr post = (my_post_ptr) cinfo->post;
+  JDIMENSION old_next_row, num_rows;
+
+  /* Reposition virtual buffer if at start of strip. */
+  if (post->next_row == 0) {
+    post->buffer = (*cinfo->mem->access_virt_sarray)
+	((j_common_ptr) cinfo, post->whole_image,
+	 post->starting_row, post->strip_height, TRUE);
+  }
+
+  /* Upsample some data (up to a strip height's worth). */
+  old_next_row = post->next_row;
+  (*cinfo->upsample->upsample) (cinfo,
+		input_buf, in_row_group_ctr, in_row_groups_avail,
+		post->buffer, &post->next_row, post->strip_height);
+
+  /* Allow quantizer to scan new data.  No data is emitted, */
+  /* but we advance out_row_ctr so outer loop can tell when we're done. */
+  if (post->next_row > old_next_row) {
+    num_rows = post->next_row - old_next_row;
+    (*cinfo->cquantize->color_quantize) (cinfo, post->buffer + old_next_row,
+					 (JSAMPARRAY) NULL, (int) num_rows);
+    *out_row_ctr += num_rows;
+  }
+
+  /* Advance if we filled the strip. */
+  if (post->next_row >= post->strip_height) {
+    post->starting_row += post->strip_height;
+    post->next_row = 0;
+  }
+}
+
+
+/*
+ * Process some data in the second pass of 2-pass quantization.
+ */
+
+METHODDEF(void)
+post_process_2pass (j_decompress_ptr cinfo,
+		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+		    JDIMENSION in_row_groups_avail,
+		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+		    JDIMENSION out_rows_avail)
+{
+  my_post_ptr post = (my_post_ptr) cinfo->post;
+  JDIMENSION num_rows, max_rows;
+
+  /* Reposition virtual buffer if at start of strip. */
+  if (post->next_row == 0) {
+    post->buffer = (*cinfo->mem->access_virt_sarray)
+	((j_common_ptr) cinfo, post->whole_image,
+	 post->starting_row, post->strip_height, FALSE);
+  }
+
+  /* Determine number of rows to emit. */
+  num_rows = post->strip_height - post->next_row; /* available in strip */
+  max_rows = out_rows_avail - *out_row_ctr; /* available in output area */
+  if (num_rows > max_rows)
+    num_rows = max_rows;
+  /* We have to check bottom of image here, can't depend on upsampler. */
+  max_rows = cinfo->output_height - post->starting_row;
+  if (num_rows > max_rows)
+    num_rows = max_rows;
+
+  /* Quantize and emit data. */
+  (*cinfo->cquantize->color_quantize) (cinfo,
+		post->buffer + post->next_row, output_buf + *out_row_ctr,
+		(int) num_rows);
+  *out_row_ctr += num_rows;
+
+  /* Advance if we filled the strip. */
+  post->next_row += num_rows;
+  if (post->next_row >= post->strip_height) {
+    post->starting_row += post->strip_height;
+    post->next_row = 0;
+  }
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */
+
+
+/*
+ * Initialize postprocessing controller.
+ */
+
+GLOBAL(void)
+jinit_d_post_controller (j_decompress_ptr cinfo, boolean need_full_buffer)
+{
+  my_post_ptr post;
+
+  post = (my_post_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_post_controller));
+  cinfo->post = (struct jpeg_d_post_controller *) post;
+  post->pub.start_pass = start_pass_dpost;
+  post->whole_image = NULL;	/* flag for no virtual arrays */
+  post->buffer = NULL;		/* flag for no strip buffer */
+
+  /* Create the quantization buffer, if needed */
+  if (cinfo->quantize_colors) {
+    /* The buffer strip height is max_v_samp_factor, which is typically
+     * an efficient number of rows for upsampling to return.
+     * (In the presence of output rescaling, we might want to be smarter?)
+     */
+    post->strip_height = (JDIMENSION) cinfo->max_v_samp_factor;
+    if (need_full_buffer) {
+      /* Two-pass color quantization: need full-image storage. */
+      /* We round up the number of rows to a multiple of the strip height. */
+#ifdef QUANT_2PASS_SUPPORTED
+      post->whole_image = (*cinfo->mem->request_virt_sarray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+	 cinfo->output_width * cinfo->out_color_components,
+	 (JDIMENSION) jround_up((long) cinfo->output_height,
+				(long) post->strip_height),
+	 post->strip_height);
+#else
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+#endif /* QUANT_2PASS_SUPPORTED */
+    } else {
+      /* One-pass color quantization: just make a strip buffer. */
+      post->buffer = (*cinfo->mem->alloc_sarray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE,
+	 cinfo->output_width * cinfo->out_color_components,
+	 post->strip_height);
+    }
+  }
+}
diff --git a/libjpegtwrp/jdsample.c b/libjpegtwrp/jdsample.c
new file mode 100644
index 000000000..80ffefb2a
--- /dev/null
+++ b/libjpegtwrp/jdsample.c
@@ -0,0 +1,478 @@
+/*
+ * jdsample.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains upsampling routines.
+ *
+ * Upsampling input data is counted in "row groups".  A row group
+ * is defined to be (v_samp_factor * DCT_scaled_size / min_DCT_scaled_size)
+ * sample rows of each component.  Upsampling will normally produce
+ * max_v_samp_factor pixel rows from each row group (but this could vary
+ * if the upsampler is applying a scale factor of its own).
+ *
+ * An excellent reference for image resampling is
+ *   Digital Image Warping, George Wolberg, 1990.
+ *   Pub. by IEEE Computer Society Press, Los Alamitos, CA. ISBN 0-8186-8944-7.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Pointer to routine to upsample a single component */
+typedef JMETHOD(void, upsample1_ptr,
+		(j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		 JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_upsampler pub;	/* public fields */
+
+  /* Color conversion buffer.  When using separate upsampling and color
+   * conversion steps, this buffer holds one upsampled row group until it
+   * has been color converted and output.
+   * Note: we do not allocate any storage for component(s) which are full-size,
+   * ie do not need rescaling.  The corresponding entry of color_buf[] is
+   * simply set to point to the input data array, thereby avoiding copying.
+   */
+  JSAMPARRAY color_buf[MAX_COMPONENTS];
+
+  /* Per-component upsampling method pointers */
+  upsample1_ptr methods[MAX_COMPONENTS];
+
+  int next_row_out;		/* counts rows emitted from color_buf */
+  JDIMENSION rows_to_go;	/* counts rows remaining in image */
+
+  /* Height of an input row group for each component. */
+  int rowgroup_height[MAX_COMPONENTS];
+
+  /* These arrays save pixel expansion factors so that int_expand need not
+   * recompute them each time.  They are unused for other upsampling methods.
+   */
+  UINT8 h_expand[MAX_COMPONENTS];
+  UINT8 v_expand[MAX_COMPONENTS];
+} my_upsampler;
+
+typedef my_upsampler * my_upsample_ptr;
+
+
+/*
+ * Initialize for an upsampling pass.
+ */
+
+METHODDEF(void)
+start_pass_upsample (j_decompress_ptr cinfo)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+
+  /* Mark the conversion buffer empty */
+  upsample->next_row_out = cinfo->max_v_samp_factor;
+  /* Initialize total-height counter for detecting bottom of image */
+  upsample->rows_to_go = cinfo->output_height;
+}
+
+
+/*
+ * Control routine to do upsampling (and color conversion).
+ *
+ * In this version we upsample each component independently.
+ * We upsample one row group into the conversion buffer, then apply
+ * color conversion a row at a time.
+ */
+
+METHODDEF(void)
+sep_upsample (j_decompress_ptr cinfo,
+	      JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+	      JDIMENSION in_row_groups_avail,
+	      JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+	      JDIMENSION out_rows_avail)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  int ci;
+  jpeg_component_info * compptr;
+  JDIMENSION num_rows;
+
+  /* Fill the conversion buffer, if it's empty */
+  if (upsample->next_row_out >= cinfo->max_v_samp_factor) {
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      /* Invoke per-component upsample method.  Notice we pass a POINTER
+       * to color_buf[ci], so that fullsize_upsample can change it.
+       */
+      (*upsample->methods[ci]) (cinfo, compptr,
+	input_buf[ci] + (*in_row_group_ctr * upsample->rowgroup_height[ci]),
+	upsample->color_buf + ci);
+    }
+    upsample->next_row_out = 0;
+  }
+
+  /* Color-convert and emit rows */
+
+  /* How many we have in the buffer: */
+  num_rows = (JDIMENSION) (cinfo->max_v_samp_factor - upsample->next_row_out);
+  /* Not more than the distance to the end of the image.  Need this test
+   * in case the image height is not a multiple of max_v_samp_factor:
+   */
+  if (num_rows > upsample->rows_to_go) 
+    num_rows = upsample->rows_to_go;
+  /* And not more than what the client can accept: */
+  out_rows_avail -= *out_row_ctr;
+  if (num_rows > out_rows_avail)
+    num_rows = out_rows_avail;
+
+  (*cinfo->cconvert->color_convert) (cinfo, upsample->color_buf,
+				     (JDIMENSION) upsample->next_row_out,
+				     output_buf + *out_row_ctr,
+				     (int) num_rows);
+
+  /* Adjust counts */
+  *out_row_ctr += num_rows;
+  upsample->rows_to_go -= num_rows;
+  upsample->next_row_out += num_rows;
+  /* When the buffer is emptied, declare this input row group consumed */
+  if (upsample->next_row_out >= cinfo->max_v_samp_factor)
+    (*in_row_group_ctr)++;
+}
+
+
+/*
+ * These are the routines invoked by sep_upsample to upsample pixel values
+ * of a single component.  One row group is processed per call.
+ */
+
+
+/*
+ * For full-size components, we just make color_buf[ci] point at the
+ * input buffer, and thus avoid copying any data.  Note that this is
+ * safe only because sep_upsample doesn't declare the input row group
+ * "consumed" until we are done color converting and emitting it.
+ */
+
+METHODDEF(void)
+fullsize_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		   JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  *output_data_ptr = input_data;
+}
+
+
+/*
+ * This is a no-op version used for "uninteresting" components.
+ * These components will not be referenced by color conversion.
+ */
+
+METHODDEF(void)
+noop_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  *output_data_ptr = NULL;	/* safety check */
+}
+
+
+/*
+ * This version handles any integral sampling ratios.
+ * This is not used for typical JPEG files, so it need not be fast.
+ * Nor, for that matter, is it particularly accurate: the algorithm is
+ * simple replication of the input pixel onto the corresponding output
+ * pixels.  The hi-falutin sampling literature refers to this as a
+ * "box filter".  A box filter tends to introduce visible artifacts,
+ * so if you are actually going to use 3:1 or 4:1 sampling ratios
+ * you would be well advised to improve this code.
+ */
+
+METHODDEF(void)
+int_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	      JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr, outptr;
+  register JSAMPLE invalue;
+  register int h;
+  JSAMPROW outend;
+  int h_expand, v_expand;
+  int inrow, outrow;
+
+  h_expand = upsample->h_expand[compptr->component_index];
+  v_expand = upsample->v_expand[compptr->component_index];
+
+  inrow = outrow = 0;
+  while (outrow < cinfo->max_v_samp_factor) {
+    /* Generate one output row with proper horizontal expansion */
+    inptr = input_data[inrow];
+    outptr = output_data[outrow];
+    outend = outptr + cinfo->output_width;
+    while (outptr < outend) {
+      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
+      for (h = h_expand; h > 0; h--) {
+	*outptr++ = invalue;
+      }
+    }
+    /* Generate any additional output rows by duplicating the first one */
+    if (v_expand > 1) {
+      jcopy_sample_rows(output_data, outrow, output_data, outrow+1,
+			v_expand-1, cinfo->output_width);
+    }
+    inrow++;
+    outrow += v_expand;
+  }
+}
+
+
+/*
+ * Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+ * It's still a box filter.
+ */
+
+METHODDEF(void)
+h2v1_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr, outptr;
+  register JSAMPLE invalue;
+  JSAMPROW outend;
+  int inrow;
+
+  for (inrow = 0; inrow < cinfo->max_v_samp_factor; inrow++) {
+    inptr = input_data[inrow];
+    outptr = output_data[inrow];
+    outend = outptr + cinfo->output_width;
+    while (outptr < outend) {
+      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
+      *outptr++ = invalue;
+      *outptr++ = invalue;
+    }
+  }
+}
+
+
+/*
+ * Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+ * It's still a box filter.
+ */
+
+METHODDEF(void)
+h2v2_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr, outptr;
+  register JSAMPLE invalue;
+  JSAMPROW outend;
+  int inrow, outrow;
+
+  inrow = outrow = 0;
+  while (outrow < cinfo->max_v_samp_factor) {
+    inptr = input_data[inrow];
+    outptr = output_data[outrow];
+    outend = outptr + cinfo->output_width;
+    while (outptr < outend) {
+      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
+      *outptr++ = invalue;
+      *outptr++ = invalue;
+    }
+    jcopy_sample_rows(output_data, outrow, output_data, outrow+1,
+		      1, cinfo->output_width);
+    inrow++;
+    outrow += 2;
+  }
+}
+
+
+/*
+ * Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+ *
+ * The upsampling algorithm is linear interpolation between pixel centers,
+ * also known as a "triangle filter".  This is a good compromise between
+ * speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+ * of the way between input pixel centers.
+ *
+ * A note about the "bias" calculations: when rounding fractional values to
+ * integer, we do not want to always round 0.5 up to the next integer.
+ * If we did that, we'd introduce a noticeable bias towards larger values.
+ * Instead, this code is arranged so that 0.5 will be rounded up or down at
+ * alternate pixel locations (a simple ordered dither pattern).
+ */
+
+METHODDEF(void)
+h2v1_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr, outptr;
+  register int invalue;
+  register JDIMENSION colctr;
+  int inrow;
+
+  for (inrow = 0; inrow < cinfo->max_v_samp_factor; inrow++) {
+    inptr = input_data[inrow];
+    outptr = output_data[inrow];
+    /* Special case for first column */
+    invalue = GETJSAMPLE(*inptr++);
+    *outptr++ = (JSAMPLE) invalue;
+    *outptr++ = (JSAMPLE) ((invalue * 3 + GETJSAMPLE(*inptr) + 2) >> 2);
+
+    for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
+      /* General case: 3/4 * nearer pixel + 1/4 * further pixel */
+      invalue = GETJSAMPLE(*inptr++) * 3;
+      *outptr++ = (JSAMPLE) ((invalue + GETJSAMPLE(inptr[-2]) + 1) >> 2);
+      *outptr++ = (JSAMPLE) ((invalue + GETJSAMPLE(*inptr) + 2) >> 2);
+    }
+
+    /* Special case for last column */
+    invalue = GETJSAMPLE(*inptr);
+    *outptr++ = (JSAMPLE) ((invalue * 3 + GETJSAMPLE(inptr[-1]) + 1) >> 2);
+    *outptr++ = (JSAMPLE) invalue;
+  }
+}
+
+
+/*
+ * Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+ * Again a triangle filter; see comments for h2v1 case, above.
+ *
+ * It is OK for us to reference the adjacent input rows because we demanded
+ * context from the main buffer controller (see initialization code).
+ */
+
+METHODDEF(void)
+h2v2_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr0, inptr1, outptr;
+#if BITS_IN_JSAMPLE == 8
+  register int thiscolsum, lastcolsum, nextcolsum;
+#else
+  register INT32 thiscolsum, lastcolsum, nextcolsum;
+#endif
+  register JDIMENSION colctr;
+  int inrow, outrow, v;
+
+  inrow = outrow = 0;
+  while (outrow < cinfo->max_v_samp_factor) {
+    for (v = 0; v < 2; v++) {
+      /* inptr0 points to nearest input row, inptr1 points to next nearest */
+      inptr0 = input_data[inrow];
+      if (v == 0)		/* next nearest is row above */
+	inptr1 = input_data[inrow-1];
+      else			/* next nearest is row below */
+	inptr1 = input_data[inrow+1];
+      outptr = output_data[outrow++];
+
+      /* Special case for first column */
+      thiscolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+      nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+      *outptr++ = (JSAMPLE) ((thiscolsum * 4 + 8) >> 4);
+      *outptr++ = (JSAMPLE) ((thiscolsum * 3 + nextcolsum + 7) >> 4);
+      lastcolsum = thiscolsum; thiscolsum = nextcolsum;
+
+      for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
+	/* General case: 3/4 * nearer pixel + 1/4 * further pixel in each */
+	/* dimension, thus 9/16, 3/16, 3/16, 1/16 overall */
+	nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+	*outptr++ = (JSAMPLE) ((thiscolsum * 3 + lastcolsum + 8) >> 4);
+	*outptr++ = (JSAMPLE) ((thiscolsum * 3 + nextcolsum + 7) >> 4);
+	lastcolsum = thiscolsum; thiscolsum = nextcolsum;
+      }
+
+      /* Special case for last column */
+      *outptr++ = (JSAMPLE) ((thiscolsum * 3 + lastcolsum + 8) >> 4);
+      *outptr++ = (JSAMPLE) ((thiscolsum * 4 + 7) >> 4);
+    }
+    inrow++;
+  }
+}
+
+
+/*
+ * Module initialization routine for upsampling.
+ */
+
+GLOBAL(void)
+jinit_upsampler (j_decompress_ptr cinfo)
+{
+  my_upsample_ptr upsample;
+  int ci;
+  jpeg_component_info * compptr;
+  boolean need_buffer, do_fancy;
+  int h_in_group, v_in_group, h_out_group, v_out_group;
+
+  upsample = (my_upsample_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_upsampler));
+  cinfo->upsample = (struct jpeg_upsampler *) upsample;
+  upsample->pub.start_pass = start_pass_upsample;
+  upsample->pub.upsample = sep_upsample;
+  upsample->pub.need_context_rows = FALSE; /* until we find out differently */
+
+  if (cinfo->CCIR601_sampling)	/* this isn't supported */
+    ERREXIT(cinfo, JERR_CCIR601_NOTIMPL);
+
+  /* jdmainct.c doesn't support context rows when min_DCT_scaled_size = 1,
+   * so don't ask for it.
+   */
+  do_fancy = cinfo->do_fancy_upsampling && cinfo->min_DCT_scaled_size > 1;
+
+  /* Verify we can handle the sampling factors, select per-component methods,
+   * and create storage as needed.
+   */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Compute size of an "input group" after IDCT scaling.  This many samples
+     * are to be converted to max_h_samp_factor * max_v_samp_factor pixels.
+     */
+    h_in_group = (compptr->h_samp_factor * compptr->DCT_scaled_size) /
+		 cinfo->min_DCT_scaled_size;
+    v_in_group = (compptr->v_samp_factor * compptr->DCT_scaled_size) /
+		 cinfo->min_DCT_scaled_size;
+    h_out_group = cinfo->max_h_samp_factor;
+    v_out_group = cinfo->max_v_samp_factor;
+    upsample->rowgroup_height[ci] = v_in_group; /* save for use later */
+    need_buffer = TRUE;
+    if (! compptr->component_needed) {
+      /* Don't bother to upsample an uninteresting component. */
+      upsample->methods[ci] = noop_upsample;
+      need_buffer = FALSE;
+    } else if (h_in_group == h_out_group && v_in_group == v_out_group) {
+      /* Fullsize components can be processed without any work. */
+      upsample->methods[ci] = fullsize_upsample;
+      need_buffer = FALSE;
+    } else if (h_in_group * 2 == h_out_group &&
+	       v_in_group == v_out_group) {
+      /* Special cases for 2h1v upsampling */
+      if (do_fancy && compptr->downsampled_width > 2)
+	upsample->methods[ci] = h2v1_fancy_upsample;
+      else
+	upsample->methods[ci] = h2v1_upsample;
+    } else if (h_in_group * 2 == h_out_group &&
+	       v_in_group * 2 == v_out_group) {
+      /* Special cases for 2h2v upsampling */
+      if (do_fancy && compptr->downsampled_width > 2) {
+	upsample->methods[ci] = h2v2_fancy_upsample;
+	upsample->pub.need_context_rows = TRUE;
+      } else
+	upsample->methods[ci] = h2v2_upsample;
+    } else if ((h_out_group % h_in_group) == 0 &&
+	       (v_out_group % v_in_group) == 0) {
+      /* Generic integral-factors upsampling method */
+      upsample->methods[ci] = int_upsample;
+      upsample->h_expand[ci] = (UINT8) (h_out_group / h_in_group);
+      upsample->v_expand[ci] = (UINT8) (v_out_group / v_in_group);
+    } else
+      ERREXIT(cinfo, JERR_FRACT_SAMPLE_NOTIMPL);
+    if (need_buffer) {
+      upsample->color_buf[ci] = (*cinfo->mem->alloc_sarray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE,
+	 (JDIMENSION) jround_up((long) cinfo->output_width,
+				(long) cinfo->max_h_samp_factor),
+	 (JDIMENSION) cinfo->max_v_samp_factor);
+    }
+  }
+}
diff --git a/libjpegtwrp/jdtrans.c b/libjpegtwrp/jdtrans.c
new file mode 100644
index 000000000..e06e170f4
--- /dev/null
+++ b/libjpegtwrp/jdtrans.c
@@ -0,0 +1,271 @@
+/*
+ * jdtrans.c
+ *
+ * Copyright (C) 1995-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains library routines for transcoding decompression,
+ * that is, reading raw DCT coefficient arrays from an input JPEG file.
+ * The routines in jdapimin.c will also be needed by a transcoder.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Forward declarations */
+LOCAL(void) transdecode_master_selection JPP((j_decompress_ptr cinfo));
+
+
+/*
+ * Read the coefficient arrays from a JPEG file.
+ * jpeg_read_header must be completed before calling this.
+ *
+ * The entire image is read into a set of virtual coefficient-block arrays,
+ * one per component.  The return value is a pointer to the array of
+ * virtual-array descriptors.  These can be manipulated directly via the
+ * JPEG memory manager, or handed off to jpeg_write_coefficients().
+ * To release the memory occupied by the virtual arrays, call
+ * jpeg_finish_decompress() when done with the data.
+ *
+ * An alternative usage is to simply obtain access to the coefficient arrays
+ * during a buffered-image-mode decompression operation.  This is allowed
+ * after any jpeg_finish_output() call.  The arrays can be accessed until
+ * jpeg_finish_decompress() is called.  (Note that any call to the library
+ * may reposition the arrays, so don't rely on access_virt_barray() results
+ * to stay valid across library calls.)
+ *
+ * Returns NULL if suspended.  This case need be checked only if
+ * a suspending data source is used.
+ */
+
+GLOBAL(jvirt_barray_ptr *)
+jpeg_read_coefficients (j_decompress_ptr cinfo)
+{
+  if (cinfo->global_state == DSTATE_READY) {
+    /* First call: initialize active modules */
+    transdecode_master_selection(cinfo);
+    cinfo->global_state = DSTATE_RDCOEFS;
+  }
+  if (cinfo->global_state == DSTATE_RDCOEFS) {
+    /* Absorb whole file into the coef buffer */
+    for (;;) {
+      int retcode;
+      /* Call progress monitor hook if present */
+      if (cinfo->progress != NULL)
+        (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+      /* Absorb some more input */
+      retcode = (*cinfo->inputctl->consume_input) (cinfo);
+      if (retcode == JPEG_SUSPENDED)
+        return NULL;
+      if (retcode == JPEG_REACHED_EOI)
+        break;
+      /* Advance progress counter if appropriate */
+      if (cinfo->progress != NULL &&
+	  (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
+        if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
+	  /* startup underestimated number of scans; ratchet up one scan */
+        cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
+        }
+      }
+    }
+    /* Set state so that jpeg_finish_decompress does the right thing */
+    cinfo->global_state = DSTATE_STOPPING;
+  }
+  /* At this point we should be in state DSTATE_STOPPING if being used
+   * standalone, or in state DSTATE_BUFIMAGE if being invoked to get access
+   * to the coefficients during a full buffered-image-mode decompression.
+   */
+  if ((cinfo->global_state == DSTATE_STOPPING ||
+       cinfo->global_state == DSTATE_BUFIMAGE) && cinfo->buffered_image) {
+    return cinfo->coef->coef_arrays;
+  }
+  /* Oops, improper usage */
+  ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  return NULL;			/* keep compiler happy */
+}
+
+LOCAL(boolean)
+jpeg_build_huffman_index_progressive(j_decompress_ptr cinfo,
+        huffman_index *index)
+{
+  if (cinfo->global_state == DSTATE_READY) {
+    printf("Progressive Mode\n");
+    /* First call: initialize active modules */
+    transdecode_master_selection(cinfo);
+    cinfo->global_state = DSTATE_RDCOEFS;
+  }
+  if (cinfo->global_state == DSTATE_RDCOEFS) {
+    unsigned int mcu;
+    int i;
+    cinfo->marker->get_sos_marker_position(cinfo, index);
+
+    /* Absorb whole file into the coef buffer */
+    for (mcu = 0; mcu < cinfo->total_iMCU_rows; mcu++) {
+      int retcode = 0;
+      /* Call progress monitor hook if present */
+      if (cinfo->progress != NULL)
+        (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+      /* Absorb some more input */
+      jinit_phuff_decoder(cinfo);
+      for (i = 0; i < index->scan_count; i++) {
+        (*cinfo->inputctl->finish_input_pass) (cinfo);
+        jset_input_stream_position(cinfo, index->scan[i].bitstream_offset);
+        cinfo->unread_marker = 0;
+        retcode = (*cinfo->inputctl->consume_input_build_huffman_index)
+                    (cinfo, index, i);
+        if (retcode == JPEG_REACHED_EOI)
+          break;
+        cinfo->input_iMCU_row = mcu;
+        if (mcu != 0)
+          (*cinfo->entropy->configure_huffman_decoder)
+                (cinfo, index->scan[i].prev_MCU_offset);
+        cinfo->input_scan_number = i;
+        retcode = (*cinfo->inputctl->consume_input_build_huffman_index)
+                    (cinfo, index, i);
+      }
+      if (retcode == JPEG_SUSPENDED)
+        return FALSE;
+      if (retcode == JPEG_REACHED_EOI)
+        break;
+      /* Advance progress counter if appropriate */
+      if (cinfo->progress != NULL &&
+	  (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
+        if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
+	  /* startup underestimated number of scans; ratchet up one scan */
+          cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
+        }
+      }
+    }
+    cinfo->global_state = DSTATE_STOPPING;
+  }
+  /* At this point we should be in state DSTATE_STOPPING if being used
+   * standalone, or in state DSTATE_BUFIMAGE if being invoked to get access
+   * to the coefficients during a full buffered-image-mode decompression.
+   */
+  if ((cinfo->global_state == DSTATE_STOPPING ||
+       cinfo->global_state == DSTATE_BUFIMAGE) && cinfo->buffered_image) {
+    return TRUE;
+  }
+  /* Oops, improper usage */
+  ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  return FALSE;			/* keep compiler happy */
+}
+
+LOCAL(boolean)
+jpeg_build_huffman_index_baseline(j_decompress_ptr cinfo, huffman_index *index)
+{
+  if (cinfo->global_state == DSTATE_READY) {
+    printf("Baseline Mode\n");
+    /* First call: initialize active modules */
+    transdecode_master_selection(cinfo);
+    cinfo->global_state = DSTATE_RDCOEFS;
+  }
+  if (cinfo->global_state == DSTATE_RDCOEFS) {
+    /* Absorb whole file into the coef buffer */
+    for (;;) {
+      int retcode;
+      /* Call progress monitor hook if present */
+      if (cinfo->progress != NULL)
+        (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+      /* Absorb some more input */
+      retcode = (*cinfo->inputctl->consume_input_build_huffman_index)
+                    (cinfo, index, 0);
+      if (retcode == JPEG_SUSPENDED)
+        return FALSE;
+      if (retcode == JPEG_REACHED_EOI)
+        break;
+      if (retcode == JPEG_SCAN_COMPLETED)
+        break;
+
+      /* Advance progress counter if appropriate */
+      if (cinfo->progress != NULL &&
+	  (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
+        if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
+	  /* startup underestimated number of scans; ratchet up one scan */
+        cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
+        }
+      }
+    }
+    /* Set state so that jpeg_finish_decompress does the right thing */
+    cinfo->global_state = DSTATE_STOPPING;
+  }
+  /* At this point we should be in state DSTATE_STOPPING if being used
+   * standalone, or in state DSTATE_BUFIMAGE if being invoked to get access
+   * to the coefficients during a full buffered-image-mode decompression.
+   */
+  if ((cinfo->global_state == DSTATE_STOPPING ||
+       cinfo->global_state == DSTATE_BUFIMAGE) && cinfo->buffered_image) {
+    return TRUE;
+  }
+  /* Oops, improper usage */
+  ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  return FALSE;			/* keep compiler happy */
+}
+
+GLOBAL(boolean)
+jpeg_build_huffman_index(j_decompress_ptr cinfo, huffman_index *index)
+{
+    cinfo->tile_decode = TRUE;
+    if (cinfo->progressive_mode)
+      return jpeg_build_huffman_index_progressive(cinfo, index);
+    else
+      return jpeg_build_huffman_index_baseline(cinfo, index);
+}
+
+/*
+ * Master selection of decompression modules for transcoding.
+ * This substitutes for jdmaster.c's initialization of the full decompressor.
+ */
+
+LOCAL(void)
+transdecode_master_selection (j_decompress_ptr cinfo)
+{
+  /* This is effectively a buffered-image operation. */
+  cinfo->buffered_image = TRUE;
+
+  /* Entropy decoding: either Huffman or arithmetic coding. */
+  if (cinfo->arith_code) {
+    ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+  } else {
+    if (cinfo->progressive_mode) {
+#ifdef D_PROGRESSIVE_SUPPORTED
+      jinit_phuff_decoder(cinfo);
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else {
+      jinit_huff_decoder(cinfo);
+    }
+  }
+
+  /* Always get a full-image coefficient buffer. */
+  jinit_d_coef_controller(cinfo, TRUE);
+
+  /* We can now tell the memory manager to allocate virtual arrays. */
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr) cinfo);
+
+  /* Initialize input side of decompressor to consume first scan. */
+  (*cinfo->inputctl->start_input_pass) (cinfo);
+
+  /* Initialize progress monitoring. */
+  if (cinfo->progress != NULL) {
+    int nscans;
+    /* Estimate number of scans to set pass_limit. */
+    if (cinfo->progressive_mode) {
+      /* Arbitrarily estimate 2 interleaved DC scans + 3 AC scans/component. */
+      nscans = 2 + 3 * cinfo->num_components;
+    } else if (cinfo->inputctl->has_multiple_scans) {
+      /* For a nonprogressive multiscan file, estimate 1 scan per component. */
+      nscans = cinfo->num_components;
+    } else {
+      nscans = 1;
+    }
+    cinfo->progress->pass_counter = 0L;
+    cinfo->progress->pass_limit = (long) cinfo->total_iMCU_rows * nscans;
+    cinfo->progress->completed_passes = 0;
+    cinfo->progress->total_passes = 1;
+  }
+}
diff --git a/libjpegtwrp/jerror.c b/libjpegtwrp/jerror.c
new file mode 100644
index 000000000..3da7be86a
--- /dev/null
+++ b/libjpegtwrp/jerror.c
@@ -0,0 +1,252 @@
+/*
+ * jerror.c
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains simple error-reporting and trace-message routines.
+ * These are suitable for Unix-like systems and others where writing to
+ * stderr is the right thing to do.  Many applications will want to replace
+ * some or all of these routines.
+ *
+ * If you define USE_WINDOWS_MESSAGEBOX in jconfig.h or in the makefile,
+ * you get a Windows-specific hack to display error messages in a dialog box.
+ * It ain't much, but it beats dropping error messages into the bit bucket,
+ * which is what happens to output to stderr under most Windows C compilers.
+ *
+ * These routines are used by both the compression and decompression code.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jversion.h"
+#include "jerror.h"
+
+#ifdef USE_WINDOWS_MESSAGEBOX
+#include <windows.h>
+#endif
+
+#ifndef EXIT_FAILURE		/* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+
+
+/*
+ * Create the message string table.
+ * We do this from the master message list in jerror.h by re-reading
+ * jerror.h with a suitable definition for macro JMESSAGE.
+ * The message table is made an external symbol just in case any applications
+ * want to refer to it directly.
+ */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jpeg_std_message_table	jMsgTable
+#endif
+
+#define JMESSAGE(code,string)	string ,
+
+const char * const jpeg_std_message_table[] = {
+#include "jerror.h"
+  NULL
+};
+
+
+/*
+ * Error exit handler: must not return to caller.
+ *
+ * Applications may override this if they want to get control back after
+ * an error.  Typically one would longjmp somewhere instead of exiting.
+ * The setjmp buffer can be made a private field within an expanded error
+ * handler object.  Note that the info needed to generate an error message
+ * is stored in the error object, so you can generate the message now or
+ * later, at your convenience.
+ * You should make sure that the JPEG object is cleaned up (with jpeg_abort
+ * or jpeg_destroy) at some point.
+ */
+
+METHODDEF(void)
+error_exit (j_common_ptr cinfo)
+{
+  /* Always display the message */
+  (*cinfo->err->output_message) (cinfo);
+
+  /* Let the memory manager delete any temp files before we die */
+  jpeg_destroy(cinfo);
+
+  exit(EXIT_FAILURE);
+}
+
+
+/*
+ * Actual output of an error or trace message.
+ * Applications may override this method to send JPEG messages somewhere
+ * other than stderr.
+ *
+ * On Windows, printing to stderr is generally completely useless,
+ * so we provide optional code to produce an error-dialog popup.
+ * Most Windows applications will still prefer to override this routine,
+ * but if they don't, it'll do something at least marginally useful.
+ *
+ * NOTE: to use the library in an environment that doesn't support the
+ * C stdio library, you may have to delete the call to fprintf() entirely,
+ * not just not use this routine.
+ */
+
+METHODDEF(void)
+output_message (j_common_ptr cinfo)
+{
+  char buffer[JMSG_LENGTH_MAX];
+
+  /* Create the message */
+  (*cinfo->err->format_message) (cinfo, buffer);
+
+#ifdef USE_WINDOWS_MESSAGEBOX
+  /* Display it in a message dialog box */
+  MessageBox(GetActiveWindow(), buffer, "JPEG Library Error",
+	     MB_OK | MB_ICONERROR);
+#else
+  /* Send it to stderr, adding a newline */
+  fprintf(stderr, "%s\n", buffer);
+#endif
+}
+
+
+/*
+ * Decide whether to emit a trace or warning message.
+ * msg_level is one of:
+ *   -1: recoverable corrupt-data warning, may want to abort.
+ *    0: important advisory messages (always display to user).
+ *    1: first level of tracing detail.
+ *    2,3,...: successively more detailed tracing messages.
+ * An application might override this method if it wanted to abort on warnings
+ * or change the policy about which messages to display.
+ */
+
+METHODDEF(void)
+emit_message (j_common_ptr cinfo, int msg_level)
+{
+  struct jpeg_error_mgr * err = cinfo->err;
+
+  if (msg_level < 0) {
+    /* It's a warning message.  Since corrupt files may generate many warnings,
+     * the policy implemented here is to show only the first warning,
+     * unless trace_level >= 3.
+     */
+    if (err->num_warnings == 0 || err->trace_level >= 3)
+      (*err->output_message) (cinfo);
+    /* Always count warnings in num_warnings. */
+    err->num_warnings++;
+  } else {
+    /* It's a trace message.  Show it if trace_level >= msg_level. */
+    if (err->trace_level >= msg_level)
+      (*err->output_message) (cinfo);
+  }
+}
+
+
+/*
+ * Format a message string for the most recent JPEG error or message.
+ * The message is stored into buffer, which should be at least JMSG_LENGTH_MAX
+ * characters.  Note that no '\n' character is added to the string.
+ * Few applications should need to override this method.
+ */
+
+METHODDEF(void)
+format_message (j_common_ptr cinfo, char * buffer)
+{
+  struct jpeg_error_mgr * err = cinfo->err;
+  int msg_code = err->msg_code;
+  const char * msgtext = NULL;
+  const char * msgptr;
+  char ch;
+  boolean isstring;
+
+  /* Look up message string in proper table */
+  if (msg_code > 0 && msg_code <= err->last_jpeg_message) {
+    msgtext = err->jpeg_message_table[msg_code];
+  } else if (err->addon_message_table != NULL &&
+	     msg_code >= err->first_addon_message &&
+	     msg_code <= err->last_addon_message) {
+    msgtext = err->addon_message_table[msg_code - err->first_addon_message];
+  }
+
+  /* Defend against bogus message number */
+  if (msgtext == NULL) {
+    err->msg_parm.i[0] = msg_code;
+    msgtext = err->jpeg_message_table[0];
+  }
+
+  /* Check for string parameter, as indicated by %s in the message text */
+  isstring = FALSE;
+  msgptr = msgtext;
+  while ((ch = *msgptr++) != '\0') {
+    if (ch == '%') {
+      if (*msgptr == 's') isstring = TRUE;
+      break;
+    }
+  }
+
+  /* Format the message into the passed buffer */
+  if (isstring)
+    sprintf(buffer, msgtext, err->msg_parm.s);
+  else
+    sprintf(buffer, msgtext,
+	    err->msg_parm.i[0], err->msg_parm.i[1],
+	    err->msg_parm.i[2], err->msg_parm.i[3],
+	    err->msg_parm.i[4], err->msg_parm.i[5],
+	    err->msg_parm.i[6], err->msg_parm.i[7]);
+}
+
+
+/*
+ * Reset error state variables at start of a new image.
+ * This is called during compression startup to reset trace/error
+ * processing to default state, without losing any application-specific
+ * method pointers.  An application might possibly want to override
+ * this method if it has additional error processing state.
+ */
+
+METHODDEF(void)
+reset_error_mgr (j_common_ptr cinfo)
+{
+  cinfo->err->num_warnings = 0;
+  /* trace_level is not reset since it is an application-supplied parameter */
+  cinfo->err->msg_code = 0;	/* may be useful as a flag for "no error" */
+}
+
+
+/*
+ * Fill in the standard error-handling methods in a jpeg_error_mgr object.
+ * Typical call is:
+ *	struct jpeg_compress_struct cinfo;
+ *	struct jpeg_error_mgr err;
+ *
+ *	cinfo.err = jpeg_std_error(&err);
+ * after which the application may override some of the methods.
+ */
+
+GLOBAL(struct jpeg_error_mgr *)
+jpeg_std_error (struct jpeg_error_mgr * err)
+{
+  err->error_exit = error_exit;
+  err->emit_message = emit_message;
+  err->output_message = output_message;
+  err->format_message = format_message;
+  err->reset_error_mgr = reset_error_mgr;
+
+  err->trace_level = 0;		/* default = no tracing */
+  err->num_warnings = 0;	/* no warnings emitted yet */
+  err->msg_code = 0;		/* may be useful as a flag for "no error" */
+
+  /* Initialize message table pointers */
+  err->jpeg_message_table = jpeg_std_message_table;
+  err->last_jpeg_message = (int) JMSG_LASTMSGCODE - 1;
+
+  err->addon_message_table = NULL;
+  err->first_addon_message = 0;	/* for safety */
+  err->last_addon_message = 0;
+
+  return err;
+}
diff --git a/libjpegtwrp/jerror.h b/libjpegtwrp/jerror.h
new file mode 100644
index 000000000..fc2fffeac
--- /dev/null
+++ b/libjpegtwrp/jerror.h
@@ -0,0 +1,291 @@
+/*
+ * jerror.h
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file defines the error and message codes for the JPEG library.
+ * Edit this file to add new codes, or to translate the message strings to
+ * some other language.
+ * A set of error-reporting macros are defined too.  Some applications using
+ * the JPEG library may wish to include this file to get the error codes
+ * and/or the macros.
+ */
+
+/*
+ * To define the enum list of message codes, include this file without
+ * defining macro JMESSAGE.  To create a message string table, include it
+ * again with a suitable JMESSAGE definition (see jerror.c for an example).
+ */
+#ifndef JMESSAGE
+#ifndef JERROR_H
+/* First time through, define the enum list */
+#define JMAKE_ENUM_LIST
+#else
+/* Repeated inclusions of this file are no-ops unless JMESSAGE is defined */
+#define JMESSAGE(code,string)
+#endif /* JERROR_H */
+#endif /* JMESSAGE */
+
+#ifdef JMAKE_ENUM_LIST
+
+typedef enum {
+
+#define JMESSAGE(code,string)	code ,
+
+#endif /* JMAKE_ENUM_LIST */
+
+JMESSAGE(JMSG_NOMESSAGE, "Bogus message code %d") /* Must be first entry! */
+
+/* For maintenance convenience, list is alphabetical by message code name */
+JMESSAGE(JERR_ARITH_NOTIMPL,
+	 "Sorry, there are legal restrictions on arithmetic coding")
+JMESSAGE(JERR_BAD_ALIGN_TYPE, "ALIGN_TYPE is wrong, please fix")
+JMESSAGE(JERR_BAD_ALLOC_CHUNK, "MAX_ALLOC_CHUNK is wrong, please fix")
+JMESSAGE(JERR_BAD_BUFFER_MODE, "Bogus buffer control mode")
+JMESSAGE(JERR_BAD_COMPONENT_ID, "Invalid component ID %d in SOS")
+JMESSAGE(JERR_BAD_DCT_COEF, "DCT coefficient out of range")
+JMESSAGE(JERR_BAD_DCTSIZE, "IDCT output block size %d not supported")
+JMESSAGE(JERR_BAD_HUFF_TABLE, "Bogus Huffman table definition")
+JMESSAGE(JERR_BAD_IN_COLORSPACE, "Bogus input colorspace")
+JMESSAGE(JERR_BAD_J_COLORSPACE, "Bogus JPEG colorspace")
+JMESSAGE(JERR_BAD_LENGTH, "Bogus marker length")
+JMESSAGE(JERR_BAD_LIB_VERSION,
+	 "Wrong JPEG library version: library is %d, caller expects %d")
+JMESSAGE(JERR_BAD_MCU_SIZE, "Sampling factors too large for interleaved scan")
+JMESSAGE(JERR_BAD_POOL_ID, "Invalid memory pool code %d")
+JMESSAGE(JERR_BAD_PRECISION, "Unsupported JPEG data precision %d")
+JMESSAGE(JERR_BAD_PROGRESSION,
+	 "Invalid progressive parameters Ss=%d Se=%d Ah=%d Al=%d")
+JMESSAGE(JERR_BAD_PROG_SCRIPT,
+	 "Invalid progressive parameters at scan script entry %d")
+JMESSAGE(JERR_BAD_SAMPLING, "Bogus sampling factors")
+JMESSAGE(JERR_BAD_SCAN_SCRIPT, "Invalid scan script at entry %d")
+JMESSAGE(JERR_BAD_STATE, "Improper call to JPEG library in state %d")
+JMESSAGE(JERR_BAD_STRUCT_SIZE,
+	 "JPEG parameter struct mismatch: library thinks size is %u, caller expects %u")
+JMESSAGE(JERR_BAD_VIRTUAL_ACCESS, "Bogus virtual array access")
+JMESSAGE(JERR_BUFFER_SIZE, "Buffer passed to JPEG library is too small")
+JMESSAGE(JERR_CANT_SUSPEND, "Suspension not allowed here")
+JMESSAGE(JERR_CCIR601_NOTIMPL, "CCIR601 sampling not implemented yet")
+JMESSAGE(JERR_COMPONENT_COUNT, "Too many color components: %d, max %d")
+JMESSAGE(JERR_CONVERSION_NOTIMPL, "Unsupported color conversion request")
+JMESSAGE(JERR_DAC_INDEX, "Bogus DAC index %d")
+JMESSAGE(JERR_DAC_VALUE, "Bogus DAC value 0x%x")
+JMESSAGE(JERR_DHT_INDEX, "Bogus DHT index %d")
+JMESSAGE(JERR_DQT_INDEX, "Bogus DQT index %d")
+JMESSAGE(JERR_EMPTY_IMAGE, "Empty JPEG image (DNL not supported)")
+JMESSAGE(JERR_EMS_READ, "Read from EMS failed")
+JMESSAGE(JERR_EMS_WRITE, "Write to EMS failed")
+JMESSAGE(JERR_EOI_EXPECTED, "Didn't expect more than one scan")
+JMESSAGE(JERR_FILE_READ, "Input file read error")
+JMESSAGE(JERR_FILE_WRITE, "Output file write error --- out of disk space?")
+JMESSAGE(JERR_FRACT_SAMPLE_NOTIMPL, "Fractional sampling not implemented yet")
+JMESSAGE(JERR_HUFF_CLEN_OVERFLOW, "Huffman code size table overflow")
+JMESSAGE(JERR_HUFF_MISSING_CODE, "Missing Huffman code table entry")
+JMESSAGE(JERR_IMAGE_TOO_BIG, "Maximum supported image dimension is %u pixels")
+JMESSAGE(JERR_INPUT_EMPTY, "Empty input file")
+JMESSAGE(JERR_INPUT_EOF, "Premature end of input file")
+JMESSAGE(JERR_MISMATCHED_QUANT_TABLE,
+	 "Cannot transcode due to multiple use of quantization table %d")
+JMESSAGE(JERR_MISSING_DATA, "Scan script does not transmit all data")
+JMESSAGE(JERR_MODE_CHANGE, "Invalid color quantization mode change")
+JMESSAGE(JERR_NOTIMPL, "Not implemented yet")
+JMESSAGE(JERR_NOT_COMPILED, "Requested feature was omitted at compile time")
+JMESSAGE(JERR_NO_BACKING_STORE, "Backing store not supported")
+JMESSAGE(JERR_NO_HUFF_TABLE, "Huffman table 0x%02x was not defined")
+JMESSAGE(JERR_NO_IMAGE, "JPEG datastream contains no image")
+JMESSAGE(JERR_NO_QUANT_TABLE, "Quantization table 0x%02x was not defined")
+JMESSAGE(JERR_NO_SOI, "Not a JPEG file: starts with 0x%02x 0x%02x")
+JMESSAGE(JERR_OUT_OF_MEMORY, "Insufficient memory (case %d)")
+JMESSAGE(JERR_QUANT_COMPONENTS,
+	 "Cannot quantize more than %d color components")
+JMESSAGE(JERR_QUANT_FEW_COLORS, "Cannot quantize to fewer than %d colors")
+JMESSAGE(JERR_QUANT_MANY_COLORS, "Cannot quantize to more than %d colors")
+JMESSAGE(JERR_SOF_DUPLICATE, "Invalid JPEG file structure: two SOF markers")
+JMESSAGE(JERR_SOF_NO_SOS, "Invalid JPEG file structure: missing SOS marker")
+JMESSAGE(JERR_SOF_UNSUPPORTED, "Unsupported JPEG process: SOF type 0x%02x")
+JMESSAGE(JERR_SOI_DUPLICATE, "Invalid JPEG file structure: two SOI markers")
+JMESSAGE(JERR_SOS_NO_SOF, "Invalid JPEG file structure: SOS before SOF")
+JMESSAGE(JERR_TFILE_CREATE, "Failed to create temporary file %s")
+JMESSAGE(JERR_TFILE_READ, "Read failed on temporary file")
+JMESSAGE(JERR_TFILE_SEEK, "Seek failed on temporary file")
+JMESSAGE(JERR_TFILE_WRITE,
+	 "Write failed on temporary file --- out of disk space?")
+JMESSAGE(JERR_TOO_LITTLE_DATA, "Application transferred too few scanlines")
+JMESSAGE(JERR_UNKNOWN_MARKER, "Unsupported marker type 0x%02x")
+JMESSAGE(JERR_VIRTUAL_BUG, "Virtual array controller messed up")
+JMESSAGE(JERR_WIDTH_OVERFLOW, "Image too wide for this implementation")
+JMESSAGE(JERR_XMS_READ, "Read from XMS failed")
+JMESSAGE(JERR_XMS_WRITE, "Write to XMS failed")
+JMESSAGE(JMSG_COPYRIGHT, JCOPYRIGHT)
+JMESSAGE(JMSG_VERSION, JVERSION)
+JMESSAGE(JTRC_16BIT_TABLES,
+	 "Caution: quantization tables are too coarse for baseline JPEG")
+JMESSAGE(JTRC_ADOBE,
+	 "Adobe APP14 marker: version %d, flags 0x%04x 0x%04x, transform %d")
+JMESSAGE(JTRC_APP0, "Unknown APP0 marker (not JFIF), length %u")
+JMESSAGE(JTRC_APP14, "Unknown APP14 marker (not Adobe), length %u")
+JMESSAGE(JTRC_DAC, "Define Arithmetic Table 0x%02x: 0x%02x")
+JMESSAGE(JTRC_DHT, "Define Huffman Table 0x%02x")
+JMESSAGE(JTRC_DQT, "Define Quantization Table %d  precision %d")
+JMESSAGE(JTRC_DRI, "Define Restart Interval %u")
+JMESSAGE(JTRC_EMS_CLOSE, "Freed EMS handle %u")
+JMESSAGE(JTRC_EMS_OPEN, "Obtained EMS handle %u")
+JMESSAGE(JTRC_EOI, "End Of Image")
+JMESSAGE(JTRC_HUFFBITS, "        %3d %3d %3d %3d %3d %3d %3d %3d")
+JMESSAGE(JTRC_JFIF, "JFIF APP0 marker: version %d.%02d, density %dx%d  %d")
+JMESSAGE(JTRC_JFIF_BADTHUMBNAILSIZE,
+	 "Warning: thumbnail image size does not match data length %u")
+JMESSAGE(JTRC_JFIF_EXTENSION,
+	 "JFIF extension marker: type 0x%02x, length %u")
+JMESSAGE(JTRC_JFIF_THUMBNAIL, "    with %d x %d thumbnail image")
+JMESSAGE(JTRC_MISC_MARKER, "Miscellaneous marker 0x%02x, length %u")
+JMESSAGE(JTRC_PARMLESS_MARKER, "Unexpected marker 0x%02x")
+JMESSAGE(JTRC_QUANTVALS, "        %4u %4u %4u %4u %4u %4u %4u %4u")
+JMESSAGE(JTRC_QUANT_3_NCOLORS, "Quantizing to %d = %d*%d*%d colors")
+JMESSAGE(JTRC_QUANT_NCOLORS, "Quantizing to %d colors")
+JMESSAGE(JTRC_QUANT_SELECTED, "Selected %d colors for quantization")
+JMESSAGE(JTRC_RECOVERY_ACTION, "At marker 0x%02x, recovery action %d")
+JMESSAGE(JTRC_RST, "RST%d")
+JMESSAGE(JTRC_SMOOTH_NOTIMPL,
+	 "Smoothing not supported with nonstandard sampling ratios")
+JMESSAGE(JTRC_SOF, "Start Of Frame 0x%02x: width=%u, height=%u, components=%d")
+JMESSAGE(JTRC_SOF_COMPONENT, "    Component %d: %dhx%dv q=%d")
+JMESSAGE(JTRC_SOI, "Start of Image")
+JMESSAGE(JTRC_SOS, "Start Of Scan: %d components")
+JMESSAGE(JTRC_SOS_COMPONENT, "    Component %d: dc=%d ac=%d")
+JMESSAGE(JTRC_SOS_PARAMS, "  Ss=%d, Se=%d, Ah=%d, Al=%d")
+JMESSAGE(JTRC_TFILE_CLOSE, "Closed temporary file %s")
+JMESSAGE(JTRC_TFILE_OPEN, "Opened temporary file %s")
+JMESSAGE(JTRC_THUMB_JPEG,
+	 "JFIF extension marker: JPEG-compressed thumbnail image, length %u")
+JMESSAGE(JTRC_THUMB_PALETTE,
+	 "JFIF extension marker: palette thumbnail image, length %u")
+JMESSAGE(JTRC_THUMB_RGB,
+	 "JFIF extension marker: RGB thumbnail image, length %u")
+JMESSAGE(JTRC_UNKNOWN_IDS,
+	 "Unrecognized component IDs %d %d %d, assuming YCbCr")
+JMESSAGE(JTRC_XMS_CLOSE, "Freed XMS handle %u")
+JMESSAGE(JTRC_XMS_OPEN, "Obtained XMS handle %u")
+JMESSAGE(JWRN_ADOBE_XFORM, "Unknown Adobe color transform code %d")
+JMESSAGE(JWRN_BOGUS_PROGRESSION,
+	 "Inconsistent progression sequence for component %d coefficient %d")
+JMESSAGE(JWRN_EXTRANEOUS_DATA,
+	 "Corrupt JPEG data: %u extraneous bytes before marker 0x%02x")
+JMESSAGE(JWRN_HIT_MARKER, "Corrupt JPEG data: premature end of data segment")
+JMESSAGE(JWRN_HUFF_BAD_CODE, "Corrupt JPEG data: bad Huffman code")
+JMESSAGE(JWRN_JFIF_MAJOR, "Warning: unknown JFIF revision number %d.%02d")
+JMESSAGE(JWRN_JPEG_EOF, "Premature end of JPEG file")
+JMESSAGE(JWRN_MUST_RESYNC,
+	 "Corrupt JPEG data: found marker 0x%02x instead of RST%d")
+JMESSAGE(JWRN_NOT_SEQUENTIAL, "Invalid SOS parameters for sequential JPEG")
+JMESSAGE(JWRN_TOO_MUCH_DATA, "Application transferred too many scanlines")
+
+#ifdef JMAKE_ENUM_LIST
+
+  JMSG_LASTMSGCODE
+} J_MESSAGE_CODE;
+
+#undef JMAKE_ENUM_LIST
+#endif /* JMAKE_ENUM_LIST */
+
+/* Zap JMESSAGE macro so that future re-inclusions do nothing by default */
+#undef JMESSAGE
+
+
+#ifndef JERROR_H
+#define JERROR_H
+
+/* Macros to simplify using the error and trace message stuff */
+/* The first parameter is either type of cinfo pointer */
+
+/* Fatal errors (print message and exit) */
+#define ERREXIT(cinfo,code)  \
+  ((cinfo)->err->msg_code = (code), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
+#define ERREXIT1(cinfo,code,p1)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
+#define ERREXIT2(cinfo,code,p1,p2)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (cinfo)->err->msg_parm.i[1] = (p2), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
+#define ERREXIT3(cinfo,code,p1,p2,p3)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (cinfo)->err->msg_parm.i[1] = (p2), \
+   (cinfo)->err->msg_parm.i[2] = (p3), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
+#define ERREXIT4(cinfo,code,p1,p2,p3,p4)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (cinfo)->err->msg_parm.i[1] = (p2), \
+   (cinfo)->err->msg_parm.i[2] = (p3), \
+   (cinfo)->err->msg_parm.i[3] = (p4), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
+#define ERREXITS(cinfo,code,str)  \
+  ((cinfo)->err->msg_code = (code), \
+   strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
+
+#define MAKESTMT(stuff)		do { stuff } while (0)
+
+/* Nonfatal errors (we can keep going, but the data is probably corrupt) */
+#define WARNMS(cinfo,code)  \
+  ((cinfo)->err->msg_code = (code), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), -1))
+#define WARNMS1(cinfo,code,p1)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), -1))
+#define WARNMS2(cinfo,code,p1,p2)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (cinfo)->err->msg_parm.i[1] = (p2), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), -1))
+
+/* Informational/debugging messages */
+#define TRACEMS(cinfo,lvl,code)  \
+  ((cinfo)->err->msg_code = (code), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)))
+#define TRACEMS1(cinfo,lvl,code,p1)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)))
+#define TRACEMS2(cinfo,lvl,code,p1,p2)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (cinfo)->err->msg_parm.i[1] = (p2), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)))
+#define TRACEMS3(cinfo,lvl,code,p1,p2,p3)  \
+  MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
+	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); \
+	   (cinfo)->err->msg_code = (code); \
+	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+#define TRACEMS4(cinfo,lvl,code,p1,p2,p3,p4)  \
+  MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
+	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
+	   (cinfo)->err->msg_code = (code); \
+	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+#define TRACEMS5(cinfo,lvl,code,p1,p2,p3,p4,p5)  \
+  MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
+	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
+	   _mp[4] = (p5); \
+	   (cinfo)->err->msg_code = (code); \
+	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+#define TRACEMS8(cinfo,lvl,code,p1,p2,p3,p4,p5,p6,p7,p8)  \
+  MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
+	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
+	   _mp[4] = (p5); _mp[5] = (p6); _mp[6] = (p7); _mp[7] = (p8); \
+	   (cinfo)->err->msg_code = (code); \
+	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+#define TRACEMSS(cinfo,lvl,code,str)  \
+  ((cinfo)->err->msg_code = (code), \
+   strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)))
+
+#endif /* JERROR_H */
diff --git a/libjpegtwrp/jfdctflt.c b/libjpegtwrp/jfdctflt.c
new file mode 100644
index 000000000..79d7a0078
--- /dev/null
+++ b/libjpegtwrp/jfdctflt.c
@@ -0,0 +1,168 @@
+/*
+ * jfdctflt.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a floating-point implementation of the
+ * forward DCT (Discrete Cosine Transform).
+ *
+ * This implementation should be more accurate than either of the integer
+ * DCT implementations.  However, it may not give the same results on all
+ * machines because of differences in roundoff behavior.  Speed will depend
+ * on the hardware's floating point capacity.
+ *
+ * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
+ * on each column.  Direct algorithms are also available, but they are
+ * much more complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README).  The following code
+ * is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with a fixed-point
+ * implementation, accuracy is lost due to imprecise representation of the
+ * scaled quantization values.  However, that problem does not arise if
+ * we use floating point arithmetic.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef DCT_FLOAT_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/*
+ * Perform the forward DCT on one block of samples.
+ */
+
+GLOBAL(void)
+jpeg_fdct_float (FAST_FLOAT * data)
+{
+  FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  FAST_FLOAT tmp10, tmp11, tmp12, tmp13;
+  FAST_FLOAT z1, z2, z3, z4, z5, z11, z13;
+  FAST_FLOAT *dataptr;
+  int ctr;
+
+  /* Pass 1: process rows. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[0] + dataptr[7];
+    tmp7 = dataptr[0] - dataptr[7];
+    tmp1 = dataptr[1] + dataptr[6];
+    tmp6 = dataptr[1] - dataptr[6];
+    tmp2 = dataptr[2] + dataptr[5];
+    tmp5 = dataptr[2] - dataptr[5];
+    tmp3 = dataptr[3] + dataptr[4];
+    tmp4 = dataptr[3] - dataptr[4];
+    
+    /* Even part */
+    
+    tmp10 = tmp0 + tmp3;	/* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[0] = tmp10 + tmp11; /* phase 3 */
+    dataptr[4] = tmp10 - tmp11;
+    
+    z1 = (tmp12 + tmp13) * ((FAST_FLOAT) 0.707106781); /* c4 */
+    dataptr[2] = tmp13 + z1;	/* phase 5 */
+    dataptr[6] = tmp13 - z1;
+    
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = (tmp10 - tmp12) * ((FAST_FLOAT) 0.382683433); /* c6 */
+    z2 = ((FAST_FLOAT) 0.541196100) * tmp10 + z5; /* c2-c6 */
+    z4 = ((FAST_FLOAT) 1.306562965) * tmp12 + z5; /* c2+c6 */
+    z3 = tmp11 * ((FAST_FLOAT) 0.707106781); /* c4 */
+
+    z11 = tmp7 + z3;		/* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptr[5] = z13 + z2;	/* phase 6 */
+    dataptr[3] = z13 - z2;
+    dataptr[1] = z11 + z4;
+    dataptr[7] = z11 - z4;
+
+    dataptr += DCTSIZE;		/* advance pointer to next row */
+  }
+
+  /* Pass 2: process columns. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7];
+    tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7];
+    tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6];
+    tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6];
+    tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5];
+    tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
+    tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
+    tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
+    
+    /* Even part */
+    
+    tmp10 = tmp0 + tmp3;	/* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[DCTSIZE*0] = tmp10 + tmp11; /* phase 3 */
+    dataptr[DCTSIZE*4] = tmp10 - tmp11;
+    
+    z1 = (tmp12 + tmp13) * ((FAST_FLOAT) 0.707106781); /* c4 */
+    dataptr[DCTSIZE*2] = tmp13 + z1; /* phase 5 */
+    dataptr[DCTSIZE*6] = tmp13 - z1;
+    
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = (tmp10 - tmp12) * ((FAST_FLOAT) 0.382683433); /* c6 */
+    z2 = ((FAST_FLOAT) 0.541196100) * tmp10 + z5; /* c2-c6 */
+    z4 = ((FAST_FLOAT) 1.306562965) * tmp12 + z5; /* c2+c6 */
+    z3 = tmp11 * ((FAST_FLOAT) 0.707106781); /* c4 */
+
+    z11 = tmp7 + z3;		/* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptr[DCTSIZE*5] = z13 + z2; /* phase 6 */
+    dataptr[DCTSIZE*3] = z13 - z2;
+    dataptr[DCTSIZE*1] = z11 + z4;
+    dataptr[DCTSIZE*7] = z11 - z4;
+
+    dataptr++;			/* advance pointer to next column */
+  }
+}
+
+#endif /* DCT_FLOAT_SUPPORTED */
diff --git a/libjpegtwrp/jfdctfst.c b/libjpegtwrp/jfdctfst.c
new file mode 100644
index 000000000..ccb378a3b
--- /dev/null
+++ b/libjpegtwrp/jfdctfst.c
@@ -0,0 +1,224 @@
+/*
+ * jfdctfst.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a fast, not so accurate integer implementation of the
+ * forward DCT (Discrete Cosine Transform).
+ *
+ * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
+ * on each column.  Direct algorithms are also available, but they are
+ * much more complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README).  The following code
+ * is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with fixed-point math,
+ * accuracy is lost due to imprecise representation of the scaled
+ * quantization values.  The smaller the quantization table entry, the less
+ * precise the scaled value, so this implementation does worse with high-
+ * quality-setting files than with low-quality ones.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef DCT_IFAST_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Scaling decisions are generally the same as in the LL&M algorithm;
+ * see jfdctint.c for more details.  However, we choose to descale
+ * (right shift) multiplication products as soon as they are formed,
+ * rather than carrying additional fractional bits into subsequent additions.
+ * This compromises accuracy slightly, but it lets us save a few shifts.
+ * More importantly, 16-bit arithmetic is then adequate (for 8-bit samples)
+ * everywhere except in the multiplications proper; this saves a good deal
+ * of work on 16-bit-int machines.
+ *
+ * Again to save a few shifts, the intermediate results between pass 1 and
+ * pass 2 are not upscaled, but are represented only to integral precision.
+ *
+ * A final compromise is to represent the multiplicative constants to only
+ * 8 fractional bits, rather than 13.  This saves some shifting work on some
+ * machines, and may also reduce the cost of multiplication (since there
+ * are fewer one-bits in the constants).
+ */
+
+#define CONST_BITS  8
+
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 8
+#define FIX_0_382683433  ((INT32)   98)		/* FIX(0.382683433) */
+#define FIX_0_541196100  ((INT32)  139)		/* FIX(0.541196100) */
+#define FIX_0_707106781  ((INT32)  181)		/* FIX(0.707106781) */
+#define FIX_1_306562965  ((INT32)  334)		/* FIX(1.306562965) */
+#else
+#define FIX_0_382683433  FIX(0.382683433)
+#define FIX_0_541196100  FIX(0.541196100)
+#define FIX_0_707106781  FIX(0.707106781)
+#define FIX_1_306562965  FIX(1.306562965)
+#endif
+
+
+/* We can gain a little more speed, with a further compromise in accuracy,
+ * by omitting the addition in a descaling shift.  This yields an incorrectly
+ * rounded result half the time...
+ */
+
+#ifndef USE_ACCURATE_ROUNDING
+#undef DESCALE
+#define DESCALE(x,n)  RIGHT_SHIFT(x, n)
+#endif
+
+
+/* Multiply a DCTELEM variable by an INT32 constant, and immediately
+ * descale to yield a DCTELEM result.
+ */
+
+#define MULTIPLY(var,const)  ((DCTELEM) DESCALE((var) * (const), CONST_BITS))
+
+
+/*
+ * Perform the forward DCT on one block of samples.
+ */
+
+GLOBAL(void)
+jpeg_fdct_ifast (DCTELEM * data)
+{
+  DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  DCTELEM tmp10, tmp11, tmp12, tmp13;
+  DCTELEM z1, z2, z3, z4, z5, z11, z13;
+  DCTELEM *dataptr;
+  int ctr;
+  SHIFT_TEMPS
+
+  /* Pass 1: process rows. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[0] + dataptr[7];
+    tmp7 = dataptr[0] - dataptr[7];
+    tmp1 = dataptr[1] + dataptr[6];
+    tmp6 = dataptr[1] - dataptr[6];
+    tmp2 = dataptr[2] + dataptr[5];
+    tmp5 = dataptr[2] - dataptr[5];
+    tmp3 = dataptr[3] + dataptr[4];
+    tmp4 = dataptr[3] - dataptr[4];
+    
+    /* Even part */
+    
+    tmp10 = tmp0 + tmp3;	/* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[0] = tmp10 + tmp11; /* phase 3 */
+    dataptr[4] = tmp10 - tmp11;
+    
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
+    dataptr[2] = tmp13 + z1;	/* phase 5 */
+    dataptr[6] = tmp13 - z1;
+    
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */
+    z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */
+    z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
+    z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
+
+    z11 = tmp7 + z3;		/* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptr[5] = z13 + z2;	/* phase 6 */
+    dataptr[3] = z13 - z2;
+    dataptr[1] = z11 + z4;
+    dataptr[7] = z11 - z4;
+
+    dataptr += DCTSIZE;		/* advance pointer to next row */
+  }
+
+  /* Pass 2: process columns. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7];
+    tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7];
+    tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6];
+    tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6];
+    tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5];
+    tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
+    tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
+    tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
+    
+    /* Even part */
+    
+    tmp10 = tmp0 + tmp3;	/* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[DCTSIZE*0] = tmp10 + tmp11; /* phase 3 */
+    dataptr[DCTSIZE*4] = tmp10 - tmp11;
+    
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
+    dataptr[DCTSIZE*2] = tmp13 + z1; /* phase 5 */
+    dataptr[DCTSIZE*6] = tmp13 - z1;
+    
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */
+    z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */
+    z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
+    z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
+
+    z11 = tmp7 + z3;		/* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptr[DCTSIZE*5] = z13 + z2; /* phase 6 */
+    dataptr[DCTSIZE*3] = z13 - z2;
+    dataptr[DCTSIZE*1] = z11 + z4;
+    dataptr[DCTSIZE*7] = z11 - z4;
+
+    dataptr++;			/* advance pointer to next column */
+  }
+}
+
+#endif /* DCT_IFAST_SUPPORTED */
diff --git a/libjpegtwrp/jfdctint.c b/libjpegtwrp/jfdctint.c
new file mode 100644
index 000000000..0a78b64ae
--- /dev/null
+++ b/libjpegtwrp/jfdctint.c
@@ -0,0 +1,283 @@
+/*
+ * jfdctint.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a slow-but-accurate integer implementation of the
+ * forward DCT (Discrete Cosine Transform).
+ *
+ * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
+ * on each column.  Direct algorithms are also available, but they are
+ * much more complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on an algorithm described in
+ *   C. Loeffler, A. Ligtenberg and G. Moschytz, "Practical Fast 1-D DCT
+ *   Algorithms with 11 Multiplications", Proc. Int'l. Conf. on Acoustics,
+ *   Speech, and Signal Processing 1989 (ICASSP '89), pp. 988-991.
+ * The primary algorithm described there uses 11 multiplies and 29 adds.
+ * We use their alternate method with 12 multiplies and 32 adds.
+ * The advantage of this method is that no data path contains more than one
+ * multiplication; this allows a very simple and accurate implementation in
+ * scaled fixed-point arithmetic, with a minimal number of shifts.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef DCT_ISLOW_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/*
+ * The poop on this scaling stuff is as follows:
+ *
+ * Each 1-D DCT step produces outputs which are a factor of sqrt(N)
+ * larger than the true DCT outputs.  The final outputs are therefore
+ * a factor of N larger than desired; since N=8 this can be cured by
+ * a simple right shift at the end of the algorithm.  The advantage of
+ * this arrangement is that we save two multiplications per 1-D DCT,
+ * because the y0 and y4 outputs need not be divided by sqrt(N).
+ * In the IJG code, this factor of 8 is removed by the quantization step
+ * (in jcdctmgr.c), NOT in this module.
+ *
+ * We have to do addition and subtraction of the integer inputs, which
+ * is no problem, and multiplication by fractional constants, which is
+ * a problem to do in integer arithmetic.  We multiply all the constants
+ * by CONST_SCALE and convert them to integer constants (thus retaining
+ * CONST_BITS bits of precision in the constants).  After doing a
+ * multiplication we have to divide the product by CONST_SCALE, with proper
+ * rounding, to produce the correct output.  This division can be done
+ * cheaply as a right shift of CONST_BITS bits.  We postpone shifting
+ * as long as possible so that partial sums can be added together with
+ * full fractional precision.
+ *
+ * The outputs of the first pass are scaled up by PASS1_BITS bits so that
+ * they are represented to better-than-integral precision.  These outputs
+ * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
+ * with the recommended scaling.  (For 12-bit sample data, the intermediate
+ * array is INT32 anyway.)
+ *
+ * To avoid overflow of the 32-bit intermediate results in pass 2, we must
+ * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
+ * shows that the values given below are the most effective.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#else
+#define CONST_BITS  13
+#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#endif
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 13
+#define FIX_0_298631336  ((INT32)  2446)	/* FIX(0.298631336) */
+#define FIX_0_390180644  ((INT32)  3196)	/* FIX(0.390180644) */
+#define FIX_0_541196100  ((INT32)  4433)	/* FIX(0.541196100) */
+#define FIX_0_765366865  ((INT32)  6270)	/* FIX(0.765366865) */
+#define FIX_0_899976223  ((INT32)  7373)	/* FIX(0.899976223) */
+#define FIX_1_175875602  ((INT32)  9633)	/* FIX(1.175875602) */
+#define FIX_1_501321110  ((INT32)  12299)	/* FIX(1.501321110) */
+#define FIX_1_847759065  ((INT32)  15137)	/* FIX(1.847759065) */
+#define FIX_1_961570560  ((INT32)  16069)	/* FIX(1.961570560) */
+#define FIX_2_053119869  ((INT32)  16819)	/* FIX(2.053119869) */
+#define FIX_2_562915447  ((INT32)  20995)	/* FIX(2.562915447) */
+#define FIX_3_072711026  ((INT32)  25172)	/* FIX(3.072711026) */
+#else
+#define FIX_0_298631336  FIX(0.298631336)
+#define FIX_0_390180644  FIX(0.390180644)
+#define FIX_0_541196100  FIX(0.541196100)
+#define FIX_0_765366865  FIX(0.765366865)
+#define FIX_0_899976223  FIX(0.899976223)
+#define FIX_1_175875602  FIX(1.175875602)
+#define FIX_1_501321110  FIX(1.501321110)
+#define FIX_1_847759065  FIX(1.847759065)
+#define FIX_1_961570560  FIX(1.961570560)
+#define FIX_2_053119869  FIX(2.053119869)
+#define FIX_2_562915447  FIX(2.562915447)
+#define FIX_3_072711026  FIX(3.072711026)
+#endif
+
+
+/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+ * For 8-bit samples with the recommended scaling, all the variable
+ * and constant values involved are no more than 16 bits wide, so a
+ * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
+ * For 12-bit samples, a full 32-bit multiplication will be needed.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define MULTIPLY(var,const)  MULTIPLY16C16(var,const)
+#else
+#define MULTIPLY(var,const)  ((var) * (const))
+#endif
+
+
+/*
+ * Perform the forward DCT on one block of samples.
+ */
+
+GLOBAL(void)
+jpeg_fdct_islow (DCTELEM * data)
+{
+  INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  INT32 tmp10, tmp11, tmp12, tmp13;
+  INT32 z1, z2, z3, z4, z5;
+  DCTELEM *dataptr;
+  int ctr;
+  SHIFT_TEMPS
+
+  /* Pass 1: process rows. */
+  /* Note results are scaled up by sqrt(8) compared to a true DCT; */
+  /* furthermore, we scale the results by 2**PASS1_BITS. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[0] + dataptr[7];
+    tmp7 = dataptr[0] - dataptr[7];
+    tmp1 = dataptr[1] + dataptr[6];
+    tmp6 = dataptr[1] - dataptr[6];
+    tmp2 = dataptr[2] + dataptr[5];
+    tmp5 = dataptr[2] - dataptr[5];
+    tmp3 = dataptr[3] + dataptr[4];
+    tmp4 = dataptr[3] - dataptr[4];
+    
+    /* Even part per LL&M figure 1 --- note that published figure is faulty;
+     * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
+     */
+    
+    tmp10 = tmp0 + tmp3;
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[0] = (DCTELEM) ((tmp10 + tmp11) << PASS1_BITS);
+    dataptr[4] = (DCTELEM) ((tmp10 - tmp11) << PASS1_BITS);
+    
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
+    dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
+				   CONST_BITS-PASS1_BITS);
+    dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
+				   CONST_BITS-PASS1_BITS);
+    
+    /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
+     * cK represents cos(K*pi/16).
+     * i0..i3 in the paper are tmp4..tmp7 here.
+     */
+    
+    z1 = tmp4 + tmp7;
+    z2 = tmp5 + tmp6;
+    z3 = tmp4 + tmp6;
+    z4 = tmp5 + tmp7;
+    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
+    
+    tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
+    tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
+    tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
+    tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
+    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
+    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
+    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
+    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
+    
+    z3 += z5;
+    z4 += z5;
+    
+    dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS);
+    dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS);
+    dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS);
+    dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS);
+    
+    dataptr += DCTSIZE;		/* advance pointer to next row */
+  }
+
+  /* Pass 2: process columns.
+   * We remove the PASS1_BITS scaling, but leave the results scaled up
+   * by an overall factor of 8.
+   */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7];
+    tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7];
+    tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6];
+    tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6];
+    tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5];
+    tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
+    tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
+    tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
+    
+    /* Even part per LL&M figure 1 --- note that published figure is faulty;
+     * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
+     */
+    
+    tmp10 = tmp0 + tmp3;
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
+    dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);
+    
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
+    dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
+					   CONST_BITS+PASS1_BITS);
+    dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
+					   CONST_BITS+PASS1_BITS);
+    
+    /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
+     * cK represents cos(K*pi/16).
+     * i0..i3 in the paper are tmp4..tmp7 here.
+     */
+    
+    z1 = tmp4 + tmp7;
+    z2 = tmp5 + tmp6;
+    z3 = tmp4 + tmp6;
+    z4 = tmp5 + tmp7;
+    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
+    
+    tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
+    tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
+    tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
+    tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
+    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
+    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
+    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
+    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
+    
+    z3 += z5;
+    z4 += z5;
+    
+    dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp4 + z1 + z3,
+					   CONST_BITS+PASS1_BITS);
+    dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp5 + z2 + z4,
+					   CONST_BITS+PASS1_BITS);
+    dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp6 + z2 + z3,
+					   CONST_BITS+PASS1_BITS);
+    dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp7 + z1 + z4,
+					   CONST_BITS+PASS1_BITS);
+    
+    dataptr++;			/* advance pointer to next column */
+  }
+}
+
+#endif /* DCT_ISLOW_SUPPORTED */
diff --git a/libjpegtwrp/jidctflt.c b/libjpegtwrp/jidctflt.c
new file mode 100644
index 000000000..0188ce3df
--- /dev/null
+++ b/libjpegtwrp/jidctflt.c
@@ -0,0 +1,242 @@
+/*
+ * jidctflt.c
+ *
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a floating-point implementation of the
+ * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
+ * must also perform dequantization of the input coefficients.
+ *
+ * This implementation should be more accurate than either of the integer
+ * IDCT implementations.  However, it may not give the same results on all
+ * machines because of differences in roundoff behavior.  Speed will depend
+ * on the hardware's floating point capacity.
+ *
+ * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT
+ * on each row (or vice versa, but it's more convenient to emit a row at
+ * a time).  Direct algorithms are also available, but they are much more
+ * complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README).  The following code
+ * is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with a fixed-point
+ * implementation, accuracy is lost due to imprecise representation of the
+ * scaled quantization values.  However, that problem does not arise if
+ * we use floating point arithmetic.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef DCT_FLOAT_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce a float result.
+ */
+
+#define DEQUANTIZE(coef,quantval)  (((FAST_FLOAT) (coef)) * (quantval))
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ */
+
+GLOBAL(void)
+jpeg_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		 JCOEFPTR coef_block,
+		 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  FAST_FLOAT tmp10, tmp11, tmp12, tmp13;
+  FAST_FLOAT z5, z10, z11, z12, z13;
+  JCOEFPTR inptr;
+  FLOAT_MULT_TYPE * quantptr;
+  FAST_FLOAT * wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  FAST_FLOAT workspace[DCTSIZE2]; /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (FLOAT_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; ctr--) {
+    /* Due to quantization, we will usually find that many of the input
+     * coefficients are zero, especially the AC terms.  We can exploit this
+     * by short-circuiting the IDCT calculation for any column in which all
+     * the AC terms are zero.  In that case each output is equal to the
+     * DC coefficient (with scale factor as needed).
+     * With typical images and quantization tables, half or more of the
+     * column DCT calculations can be simplified this way.
+     */
+    
+    if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
+	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
+	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
+	inptr[DCTSIZE*7] == 0) {
+      /* AC terms all zero */
+      FAST_FLOAT dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+      
+      wsptr[DCTSIZE*0] = dcval;
+      wsptr[DCTSIZE*1] = dcval;
+      wsptr[DCTSIZE*2] = dcval;
+      wsptr[DCTSIZE*3] = dcval;
+      wsptr[DCTSIZE*4] = dcval;
+      wsptr[DCTSIZE*5] = dcval;
+      wsptr[DCTSIZE*6] = dcval;
+      wsptr[DCTSIZE*7] = dcval;
+      
+      inptr++;			/* advance pointers to next column */
+      quantptr++;
+      wsptr++;
+      continue;
+    }
+    
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp10 = tmp0 + tmp2;	/* phase 3 */
+    tmp11 = tmp0 - tmp2;
+
+    tmp13 = tmp1 + tmp3;	/* phases 5-3 */
+    tmp12 = (tmp1 - tmp3) * ((FAST_FLOAT) 1.414213562) - tmp13; /* 2*c4 */
+
+    tmp0 = tmp10 + tmp13;	/* phase 2 */
+    tmp3 = tmp10 - tmp13;
+    tmp1 = tmp11 + tmp12;
+    tmp2 = tmp11 - tmp12;
+    
+    /* Odd part */
+
+    tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    tmp5 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    z13 = tmp6 + tmp5;		/* phase 6 */
+    z10 = tmp6 - tmp5;
+    z11 = tmp4 + tmp7;
+    z12 = tmp4 - tmp7;
+
+    tmp7 = z11 + z13;		/* phase 5 */
+    tmp11 = (z11 - z13) * ((FAST_FLOAT) 1.414213562); /* 2*c4 */
+
+    z5 = (z10 + z12) * ((FAST_FLOAT) 1.847759065); /* 2*c2 */
+    tmp10 = ((FAST_FLOAT) 1.082392200) * z12 - z5; /* 2*(c2-c6) */
+    tmp12 = ((FAST_FLOAT) -2.613125930) * z10 + z5; /* -2*(c2+c6) */
+
+    tmp6 = tmp12 - tmp7;	/* phase 2 */
+    tmp5 = tmp11 - tmp6;
+    tmp4 = tmp10 + tmp5;
+
+    wsptr[DCTSIZE*0] = tmp0 + tmp7;
+    wsptr[DCTSIZE*7] = tmp0 - tmp7;
+    wsptr[DCTSIZE*1] = tmp1 + tmp6;
+    wsptr[DCTSIZE*6] = tmp1 - tmp6;
+    wsptr[DCTSIZE*2] = tmp2 + tmp5;
+    wsptr[DCTSIZE*5] = tmp2 - tmp5;
+    wsptr[DCTSIZE*4] = tmp3 + tmp4;
+    wsptr[DCTSIZE*3] = tmp3 - tmp4;
+
+    inptr++;			/* advance pointers to next column */
+    quantptr++;
+    wsptr++;
+  }
+  
+  /* Pass 2: process rows from work array, store into output array. */
+  /* Note that we must descale the results by a factor of 8 == 2**3. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < DCTSIZE; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+    /* Rows of zeroes can be exploited in the same way as we did with columns.
+     * However, the column calculation has created many nonzero AC terms, so
+     * the simplification applies less often (typically 5% to 10% of the time).
+     * And testing floats for zero is relatively expensive, so we don't bother.
+     */
+    
+    /* Even part */
+
+    tmp10 = wsptr[0] + wsptr[4];
+    tmp11 = wsptr[0] - wsptr[4];
+
+    tmp13 = wsptr[2] + wsptr[6];
+    tmp12 = (wsptr[2] - wsptr[6]) * ((FAST_FLOAT) 1.414213562) - tmp13;
+
+    tmp0 = tmp10 + tmp13;
+    tmp3 = tmp10 - tmp13;
+    tmp1 = tmp11 + tmp12;
+    tmp2 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    z13 = wsptr[5] + wsptr[3];
+    z10 = wsptr[5] - wsptr[3];
+    z11 = wsptr[1] + wsptr[7];
+    z12 = wsptr[1] - wsptr[7];
+
+    tmp7 = z11 + z13;
+    tmp11 = (z11 - z13) * ((FAST_FLOAT) 1.414213562);
+
+    z5 = (z10 + z12) * ((FAST_FLOAT) 1.847759065); /* 2*c2 */
+    tmp10 = ((FAST_FLOAT) 1.082392200) * z12 - z5; /* 2*(c2-c6) */
+    tmp12 = ((FAST_FLOAT) -2.613125930) * z10 + z5; /* -2*(c2+c6) */
+
+    tmp6 = tmp12 - tmp7;
+    tmp5 = tmp11 - tmp6;
+    tmp4 = tmp10 + tmp5;
+
+    /* Final output stage: scale down by a factor of 8 and range-limit */
+
+    outptr[0] = range_limit[(int) DESCALE((INT32) (tmp0 + tmp7), 3)
+			    & RANGE_MASK];
+    outptr[7] = range_limit[(int) DESCALE((INT32) (tmp0 - tmp7), 3)
+			    & RANGE_MASK];
+    outptr[1] = range_limit[(int) DESCALE((INT32) (tmp1 + tmp6), 3)
+			    & RANGE_MASK];
+    outptr[6] = range_limit[(int) DESCALE((INT32) (tmp1 - tmp6), 3)
+			    & RANGE_MASK];
+    outptr[2] = range_limit[(int) DESCALE((INT32) (tmp2 + tmp5), 3)
+			    & RANGE_MASK];
+    outptr[5] = range_limit[(int) DESCALE((INT32) (tmp2 - tmp5), 3)
+			    & RANGE_MASK];
+    outptr[4] = range_limit[(int) DESCALE((INT32) (tmp3 + tmp4), 3)
+			    & RANGE_MASK];
+    outptr[3] = range_limit[(int) DESCALE((INT32) (tmp3 - tmp4), 3)
+			    & RANGE_MASK];
+    
+    wsptr += DCTSIZE;		/* advance pointer to next row */
+  }
+}
+
+#endif /* DCT_FLOAT_SUPPORTED */
diff --git a/libjpegtwrp/jidctfst.c b/libjpegtwrp/jidctfst.c
new file mode 100644
index 000000000..dba4216fb
--- /dev/null
+++ b/libjpegtwrp/jidctfst.c
@@ -0,0 +1,368 @@
+/*
+ * jidctfst.c
+ *
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a fast, not so accurate integer implementation of the
+ * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
+ * must also perform dequantization of the input coefficients.
+ *
+ * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT
+ * on each row (or vice versa, but it's more convenient to emit a row at
+ * a time).  Direct algorithms are also available, but they are much more
+ * complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README).  The following code
+ * is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with fixed-point math,
+ * accuracy is lost due to imprecise representation of the scaled
+ * quantization values.  The smaller the quantization table entry, the less
+ * precise the scaled value, so this implementation does worse with high-
+ * quality-setting files than with low-quality ones.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef DCT_IFAST_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Scaling decisions are generally the same as in the LL&M algorithm;
+ * see jidctint.c for more details.  However, we choose to descale
+ * (right shift) multiplication products as soon as they are formed,
+ * rather than carrying additional fractional bits into subsequent additions.
+ * This compromises accuracy slightly, but it lets us save a few shifts.
+ * More importantly, 16-bit arithmetic is then adequate (for 8-bit samples)
+ * everywhere except in the multiplications proper; this saves a good deal
+ * of work on 16-bit-int machines.
+ *
+ * The dequantized coefficients are not integers because the AA&N scaling
+ * factors have been incorporated.  We represent them scaled up by PASS1_BITS,
+ * so that the first and second IDCT rounds have the same input scaling.
+ * For 8-bit JSAMPLEs, we choose IFAST_SCALE_BITS = PASS1_BITS so as to
+ * avoid a descaling shift; this compromises accuracy rather drastically
+ * for small quantization table entries, but it saves a lot of shifts.
+ * For 12-bit JSAMPLEs, there's no hope of using 16x16 multiplies anyway,
+ * so we use a much larger scaling factor to preserve accuracy.
+ *
+ * A final compromise is to represent the multiplicative constants to only
+ * 8 fractional bits, rather than 13.  This saves some shifting work on some
+ * machines, and may also reduce the cost of multiplication (since there
+ * are fewer one-bits in the constants).
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define CONST_BITS  8
+#define PASS1_BITS  2
+#else
+#define CONST_BITS  8
+#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#endif
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 8
+#define FIX_1_082392200  ((INT32)  277)		/* FIX(1.082392200) */
+#define FIX_1_414213562  ((INT32)  362)		/* FIX(1.414213562) */
+#define FIX_1_847759065  ((INT32)  473)		/* FIX(1.847759065) */
+#define FIX_2_613125930  ((INT32)  669)		/* FIX(2.613125930) */
+#else
+#define FIX_1_082392200  FIX(1.082392200)
+#define FIX_1_414213562  FIX(1.414213562)
+#define FIX_1_847759065  FIX(1.847759065)
+#define FIX_2_613125930  FIX(2.613125930)
+#endif
+
+
+/* We can gain a little more speed, with a further compromise in accuracy,
+ * by omitting the addition in a descaling shift.  This yields an incorrectly
+ * rounded result half the time...
+ */
+
+#ifndef USE_ACCURATE_ROUNDING
+#undef DESCALE
+#define DESCALE(x,n)  RIGHT_SHIFT(x, n)
+#endif
+
+
+/* Multiply a DCTELEM variable by an INT32 constant, and immediately
+ * descale to yield a DCTELEM result.
+ */
+
+#define MULTIPLY(var,const)  ((DCTELEM) DESCALE((var) * (const), CONST_BITS))
+
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce a DCTELEM result.  For 8-bit data a 16x16->16
+ * multiplication will do.  For 12-bit data, the multiplier table is
+ * declared INT32, so a 32-bit multiply will be used.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define DEQUANTIZE(coef,quantval)  (((IFAST_MULT_TYPE) (coef)) * (quantval))
+#else
+#define DEQUANTIZE(coef,quantval)  \
+	DESCALE((coef)*(quantval), IFAST_SCALE_BITS-PASS1_BITS)
+#endif
+
+
+/* Like DESCALE, but applies to a DCTELEM and produces an int.
+ * We assume that int right shift is unsigned if INT32 right shift is.
+ */
+
+#ifdef RIGHT_SHIFT_IS_UNSIGNED
+#define ISHIFT_TEMPS	DCTELEM ishift_temp;
+#if BITS_IN_JSAMPLE == 8
+#define DCTELEMBITS  16		/* DCTELEM may be 16 or 32 bits */
+#else
+#define DCTELEMBITS  32		/* DCTELEM must be 32 bits */
+#endif
+#define IRIGHT_SHIFT(x,shft)  \
+    ((ishift_temp = (x)) < 0 ? \
+     (ishift_temp >> (shft)) | ((~((DCTELEM) 0)) << (DCTELEMBITS-(shft))) : \
+     (ishift_temp >> (shft)))
+#else
+#define ISHIFT_TEMPS
+#define IRIGHT_SHIFT(x,shft)	((x) >> (shft))
+#endif
+
+#ifdef USE_ACCURATE_ROUNDING
+#define IDESCALE(x,n)  ((int) IRIGHT_SHIFT((x) + (1 << ((n)-1)), n))
+#else
+#define IDESCALE(x,n)  ((int) IRIGHT_SHIFT(x, n))
+#endif
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ */
+
+GLOBAL(void)
+jpeg_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		 JCOEFPTR coef_block,
+		 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  DCTELEM tmp10, tmp11, tmp12, tmp13;
+  DCTELEM z5, z10, z11, z12, z13;
+  JCOEFPTR inptr;
+  IFAST_MULT_TYPE * quantptr;
+  int * wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[DCTSIZE2];	/* buffers data between passes */
+  SHIFT_TEMPS			/* for DESCALE */
+  ISHIFT_TEMPS			/* for IDESCALE */
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (IFAST_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; ctr--) {
+    /* Due to quantization, we will usually find that many of the input
+     * coefficients are zero, especially the AC terms.  We can exploit this
+     * by short-circuiting the IDCT calculation for any column in which all
+     * the AC terms are zero.  In that case each output is equal to the
+     * DC coefficient (with scale factor as needed).
+     * With typical images and quantization tables, half or more of the
+     * column DCT calculations can be simplified this way.
+     */
+    
+    if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
+	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
+	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
+	inptr[DCTSIZE*7] == 0) {
+      /* AC terms all zero */
+      int dcval = (int) DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+
+      wsptr[DCTSIZE*0] = dcval;
+      wsptr[DCTSIZE*1] = dcval;
+      wsptr[DCTSIZE*2] = dcval;
+      wsptr[DCTSIZE*3] = dcval;
+      wsptr[DCTSIZE*4] = dcval;
+      wsptr[DCTSIZE*5] = dcval;
+      wsptr[DCTSIZE*6] = dcval;
+      wsptr[DCTSIZE*7] = dcval;
+      
+      inptr++;			/* advance pointers to next column */
+      quantptr++;
+      wsptr++;
+      continue;
+    }
+    
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp10 = tmp0 + tmp2;	/* phase 3 */
+    tmp11 = tmp0 - tmp2;
+
+    tmp13 = tmp1 + tmp3;	/* phases 5-3 */
+    tmp12 = MULTIPLY(tmp1 - tmp3, FIX_1_414213562) - tmp13; /* 2*c4 */
+
+    tmp0 = tmp10 + tmp13;	/* phase 2 */
+    tmp3 = tmp10 - tmp13;
+    tmp1 = tmp11 + tmp12;
+    tmp2 = tmp11 - tmp12;
+    
+    /* Odd part */
+
+    tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    tmp5 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    z13 = tmp6 + tmp5;		/* phase 6 */
+    z10 = tmp6 - tmp5;
+    z11 = tmp4 + tmp7;
+    z12 = tmp4 - tmp7;
+
+    tmp7 = z11 + z13;		/* phase 5 */
+    tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562); /* 2*c4 */
+
+    z5 = MULTIPLY(z10 + z12, FIX_1_847759065); /* 2*c2 */
+    tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5; /* 2*(c2-c6) */
+    tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5; /* -2*(c2+c6) */
+
+    tmp6 = tmp12 - tmp7;	/* phase 2 */
+    tmp5 = tmp11 - tmp6;
+    tmp4 = tmp10 + tmp5;
+
+    wsptr[DCTSIZE*0] = (int) (tmp0 + tmp7);
+    wsptr[DCTSIZE*7] = (int) (tmp0 - tmp7);
+    wsptr[DCTSIZE*1] = (int) (tmp1 + tmp6);
+    wsptr[DCTSIZE*6] = (int) (tmp1 - tmp6);
+    wsptr[DCTSIZE*2] = (int) (tmp2 + tmp5);
+    wsptr[DCTSIZE*5] = (int) (tmp2 - tmp5);
+    wsptr[DCTSIZE*4] = (int) (tmp3 + tmp4);
+    wsptr[DCTSIZE*3] = (int) (tmp3 - tmp4);
+
+    inptr++;			/* advance pointers to next column */
+    quantptr++;
+    wsptr++;
+  }
+  
+  /* Pass 2: process rows from work array, store into output array. */
+  /* Note that we must descale the results by a factor of 8 == 2**3, */
+  /* and also undo the PASS1_BITS scaling. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < DCTSIZE; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+    /* Rows of zeroes can be exploited in the same way as we did with columns.
+     * However, the column calculation has created many nonzero AC terms, so
+     * the simplification applies less often (typically 5% to 10% of the time).
+     * On machines with very fast multiplication, it's possible that the
+     * test takes more time than it's worth.  In that case this section
+     * may be commented out.
+     */
+    
+#ifndef NO_ZERO_ROW_TEST
+    if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 && wsptr[4] == 0 &&
+	wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
+      /* AC terms all zero */
+      JSAMPLE dcval = range_limit[IDESCALE(wsptr[0], PASS1_BITS+3)
+				  & RANGE_MASK];
+      
+      outptr[0] = dcval;
+      outptr[1] = dcval;
+      outptr[2] = dcval;
+      outptr[3] = dcval;
+      outptr[4] = dcval;
+      outptr[5] = dcval;
+      outptr[6] = dcval;
+      outptr[7] = dcval;
+
+      wsptr += DCTSIZE;		/* advance pointer to next row */
+      continue;
+    }
+#endif
+    
+    /* Even part */
+
+    tmp10 = ((DCTELEM) wsptr[0] + (DCTELEM) wsptr[4]);
+    tmp11 = ((DCTELEM) wsptr[0] - (DCTELEM) wsptr[4]);
+
+    tmp13 = ((DCTELEM) wsptr[2] + (DCTELEM) wsptr[6]);
+    tmp12 = MULTIPLY((DCTELEM) wsptr[2] - (DCTELEM) wsptr[6], FIX_1_414213562)
+	    - tmp13;
+
+    tmp0 = tmp10 + tmp13;
+    tmp3 = tmp10 - tmp13;
+    tmp1 = tmp11 + tmp12;
+    tmp2 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    z13 = (DCTELEM) wsptr[5] + (DCTELEM) wsptr[3];
+    z10 = (DCTELEM) wsptr[5] - (DCTELEM) wsptr[3];
+    z11 = (DCTELEM) wsptr[1] + (DCTELEM) wsptr[7];
+    z12 = (DCTELEM) wsptr[1] - (DCTELEM) wsptr[7];
+
+    tmp7 = z11 + z13;		/* phase 5 */
+    tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562); /* 2*c4 */
+
+    z5 = MULTIPLY(z10 + z12, FIX_1_847759065); /* 2*c2 */
+    tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5; /* 2*(c2-c6) */
+    tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5; /* -2*(c2+c6) */
+
+    tmp6 = tmp12 - tmp7;	/* phase 2 */
+    tmp5 = tmp11 - tmp6;
+    tmp4 = tmp10 + tmp5;
+
+    /* Final output stage: scale down by a factor of 8 and range-limit */
+
+    outptr[0] = range_limit[IDESCALE(tmp0 + tmp7, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[7] = range_limit[IDESCALE(tmp0 - tmp7, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[1] = range_limit[IDESCALE(tmp1 + tmp6, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[6] = range_limit[IDESCALE(tmp1 - tmp6, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[2] = range_limit[IDESCALE(tmp2 + tmp5, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[5] = range_limit[IDESCALE(tmp2 - tmp5, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[4] = range_limit[IDESCALE(tmp3 + tmp4, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[3] = range_limit[IDESCALE(tmp3 - tmp4, PASS1_BITS+3)
+			    & RANGE_MASK];
+
+    wsptr += DCTSIZE;		/* advance pointer to next row */
+  }
+}
+
+#endif /* DCT_IFAST_SUPPORTED */
diff --git a/libjpegtwrp/jidctint.c b/libjpegtwrp/jidctint.c
new file mode 100644
index 000000000..a72b3207c
--- /dev/null
+++ b/libjpegtwrp/jidctint.c
@@ -0,0 +1,389 @@
+/*
+ * jidctint.c
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a slow-but-accurate integer implementation of the
+ * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
+ * must also perform dequantization of the input coefficients.
+ *
+ * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT
+ * on each row (or vice versa, but it's more convenient to emit a row at
+ * a time).  Direct algorithms are also available, but they are much more
+ * complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on an algorithm described in
+ *   C. Loeffler, A. Ligtenberg and G. Moschytz, "Practical Fast 1-D DCT
+ *   Algorithms with 11 Multiplications", Proc. Int'l. Conf. on Acoustics,
+ *   Speech, and Signal Processing 1989 (ICASSP '89), pp. 988-991.
+ * The primary algorithm described there uses 11 multiplies and 29 adds.
+ * We use their alternate method with 12 multiplies and 32 adds.
+ * The advantage of this method is that no data path contains more than one
+ * multiplication; this allows a very simple and accurate implementation in
+ * scaled fixed-point arithmetic, with a minimal number of shifts.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef DCT_ISLOW_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/*
+ * The poop on this scaling stuff is as follows:
+ *
+ * Each 1-D IDCT step produces outputs which are a factor of sqrt(N)
+ * larger than the true IDCT outputs.  The final outputs are therefore
+ * a factor of N larger than desired; since N=8 this can be cured by
+ * a simple right shift at the end of the algorithm.  The advantage of
+ * this arrangement is that we save two multiplications per 1-D IDCT,
+ * because the y0 and y4 inputs need not be divided by sqrt(N).
+ *
+ * We have to do addition and subtraction of the integer inputs, which
+ * is no problem, and multiplication by fractional constants, which is
+ * a problem to do in integer arithmetic.  We multiply all the constants
+ * by CONST_SCALE and convert them to integer constants (thus retaining
+ * CONST_BITS bits of precision in the constants).  After doing a
+ * multiplication we have to divide the product by CONST_SCALE, with proper
+ * rounding, to produce the correct output.  This division can be done
+ * cheaply as a right shift of CONST_BITS bits.  We postpone shifting
+ * as long as possible so that partial sums can be added together with
+ * full fractional precision.
+ *
+ * The outputs of the first pass are scaled up by PASS1_BITS bits so that
+ * they are represented to better-than-integral precision.  These outputs
+ * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
+ * with the recommended scaling.  (To scale up 12-bit sample data further, an
+ * intermediate INT32 array would be needed.)
+ *
+ * To avoid overflow of the 32-bit intermediate results in pass 2, we must
+ * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
+ * shows that the values given below are the most effective.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#else
+#define CONST_BITS  13
+#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#endif
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 13
+#define FIX_0_298631336  ((INT32)  2446)	/* FIX(0.298631336) */
+#define FIX_0_390180644  ((INT32)  3196)	/* FIX(0.390180644) */
+#define FIX_0_541196100  ((INT32)  4433)	/* FIX(0.541196100) */
+#define FIX_0_765366865  ((INT32)  6270)	/* FIX(0.765366865) */
+#define FIX_0_899976223  ((INT32)  7373)	/* FIX(0.899976223) */
+#define FIX_1_175875602  ((INT32)  9633)	/* FIX(1.175875602) */
+#define FIX_1_501321110  ((INT32)  12299)	/* FIX(1.501321110) */
+#define FIX_1_847759065  ((INT32)  15137)	/* FIX(1.847759065) */
+#define FIX_1_961570560  ((INT32)  16069)	/* FIX(1.961570560) */
+#define FIX_2_053119869  ((INT32)  16819)	/* FIX(2.053119869) */
+#define FIX_2_562915447  ((INT32)  20995)	/* FIX(2.562915447) */
+#define FIX_3_072711026  ((INT32)  25172)	/* FIX(3.072711026) */
+#else
+#define FIX_0_298631336  FIX(0.298631336)
+#define FIX_0_390180644  FIX(0.390180644)
+#define FIX_0_541196100  FIX(0.541196100)
+#define FIX_0_765366865  FIX(0.765366865)
+#define FIX_0_899976223  FIX(0.899976223)
+#define FIX_1_175875602  FIX(1.175875602)
+#define FIX_1_501321110  FIX(1.501321110)
+#define FIX_1_847759065  FIX(1.847759065)
+#define FIX_1_961570560  FIX(1.961570560)
+#define FIX_2_053119869  FIX(2.053119869)
+#define FIX_2_562915447  FIX(2.562915447)
+#define FIX_3_072711026  FIX(3.072711026)
+#endif
+
+
+/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+ * For 8-bit samples with the recommended scaling, all the variable
+ * and constant values involved are no more than 16 bits wide, so a
+ * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
+ * For 12-bit samples, a full 32-bit multiplication will be needed.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define MULTIPLY(var,const)  MULTIPLY16C16(var,const)
+#else
+#define MULTIPLY(var,const)  ((var) * (const))
+#endif
+
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce an int result.  In this module, both inputs and result
+ * are 16 bits or less, so either int or short multiply will work.
+ */
+
+#define DEQUANTIZE(coef,quantval)  (((ISLOW_MULT_TYPE) (coef)) * (quantval))
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ */
+
+GLOBAL(void)
+jpeg_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		 JCOEFPTR coef_block,
+		 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  INT32 tmp0, tmp1, tmp2, tmp3;
+  INT32 tmp10, tmp11, tmp12, tmp13;
+  INT32 z1, z2, z3, z4, z5;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE * quantptr;
+  int * wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[DCTSIZE2];	/* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+  /* Note results are scaled up by sqrt(8) compared to a true IDCT; */
+  /* furthermore, we scale the results by 2**PASS1_BITS. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; ctr--) {
+    /* Due to quantization, we will usually find that many of the input
+     * coefficients are zero, especially the AC terms.  We can exploit this
+     * by short-circuiting the IDCT calculation for any column in which all
+     * the AC terms are zero.  In that case each output is equal to the
+     * DC coefficient (with scale factor as needed).
+     * With typical images and quantization tables, half or more of the
+     * column DCT calculations can be simplified this way.
+     */
+    
+    if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
+	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
+	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
+	inptr[DCTSIZE*7] == 0) {
+      /* AC terms all zero */
+      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
+      
+      wsptr[DCTSIZE*0] = dcval;
+      wsptr[DCTSIZE*1] = dcval;
+      wsptr[DCTSIZE*2] = dcval;
+      wsptr[DCTSIZE*3] = dcval;
+      wsptr[DCTSIZE*4] = dcval;
+      wsptr[DCTSIZE*5] = dcval;
+      wsptr[DCTSIZE*6] = dcval;
+      wsptr[DCTSIZE*7] = dcval;
+      
+      inptr++;			/* advance pointers to next column */
+      quantptr++;
+      wsptr++;
+      continue;
+    }
+    
+    /* Even part: reverse the even part of the forward DCT. */
+    /* The rotator is sqrt(2)*c(-6). */
+    
+    z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+    
+    z1 = MULTIPLY(z2 + z3, FIX_0_541196100);
+    tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065);
+    tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865);
+    
+    z2 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+
+    tmp0 = (z2 + z3) << CONST_BITS;
+    tmp1 = (z2 - z3) << CONST_BITS;
+    
+    tmp10 = tmp0 + tmp3;
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+    
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+    tmp1 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    tmp2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    tmp3 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    
+    z1 = tmp0 + tmp3;
+    z2 = tmp1 + tmp2;
+    z3 = tmp0 + tmp2;
+    z4 = tmp1 + tmp3;
+    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
+    
+    tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
+    tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
+    tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
+    tmp3 = MULTIPLY(tmp3, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
+    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
+    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
+    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
+    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
+    
+    z3 += z5;
+    z4 += z5;
+    
+    tmp0 += z1 + z3;
+    tmp1 += z2 + z4;
+    tmp2 += z2 + z3;
+    tmp3 += z1 + z4;
+    
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+    
+    wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS);
+    
+    inptr++;			/* advance pointers to next column */
+    quantptr++;
+    wsptr++;
+  }
+  
+  /* Pass 2: process rows from work array, store into output array. */
+  /* Note that we must descale the results by a factor of 8 == 2**3, */
+  /* and also undo the PASS1_BITS scaling. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < DCTSIZE; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+    /* Rows of zeroes can be exploited in the same way as we did with columns.
+     * However, the column calculation has created many nonzero AC terms, so
+     * the simplification applies less often (typically 5% to 10% of the time).
+     * On machines with very fast multiplication, it's possible that the
+     * test takes more time than it's worth.  In that case this section
+     * may be commented out.
+     */
+    
+#ifndef NO_ZERO_ROW_TEST
+    if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 && wsptr[4] == 0 &&
+	wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
+      /* AC terms all zero */
+      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
+				  & RANGE_MASK];
+      
+      outptr[0] = dcval;
+      outptr[1] = dcval;
+      outptr[2] = dcval;
+      outptr[3] = dcval;
+      outptr[4] = dcval;
+      outptr[5] = dcval;
+      outptr[6] = dcval;
+      outptr[7] = dcval;
+
+      wsptr += DCTSIZE;		/* advance pointer to next row */
+      continue;
+    }
+#endif
+    
+    /* Even part: reverse the even part of the forward DCT. */
+    /* The rotator is sqrt(2)*c(-6). */
+    
+    z2 = (INT32) wsptr[2];
+    z3 = (INT32) wsptr[6];
+    
+    z1 = MULTIPLY(z2 + z3, FIX_0_541196100);
+    tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065);
+    tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865);
+    
+    tmp0 = ((INT32) wsptr[0] + (INT32) wsptr[4]) << CONST_BITS;
+    tmp1 = ((INT32) wsptr[0] - (INT32) wsptr[4]) << CONST_BITS;
+    
+    tmp10 = tmp0 + tmp3;
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+    
+    tmp0 = (INT32) wsptr[7];
+    tmp1 = (INT32) wsptr[5];
+    tmp2 = (INT32) wsptr[3];
+    tmp3 = (INT32) wsptr[1];
+    
+    z1 = tmp0 + tmp3;
+    z2 = tmp1 + tmp2;
+    z3 = tmp0 + tmp2;
+    z4 = tmp1 + tmp3;
+    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
+    
+    tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
+    tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
+    tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
+    tmp3 = MULTIPLY(tmp3, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
+    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
+    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
+    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
+    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
+    
+    z3 += z5;
+    z4 += z5;
+    
+    tmp0 += z1 + z3;
+    tmp1 += z2 + z4;
+    tmp2 += z2 + z3;
+    tmp3 += z1 + z4;
+    
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+    
+    outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp3,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[7] = range_limit[(int) DESCALE(tmp10 - tmp3,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[1] = range_limit[(int) DESCALE(tmp11 + tmp2,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[6] = range_limit[(int) DESCALE(tmp11 - tmp2,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[2] = range_limit[(int) DESCALE(tmp12 + tmp1,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[5] = range_limit[(int) DESCALE(tmp12 - tmp1,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[3] = range_limit[(int) DESCALE(tmp13 + tmp0,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[4] = range_limit[(int) DESCALE(tmp13 - tmp0,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    
+    wsptr += DCTSIZE;		/* advance pointer to next row */
+  }
+}
+
+#endif /* DCT_ISLOW_SUPPORTED */
diff --git a/libjpegtwrp/jidctred.c b/libjpegtwrp/jidctred.c
new file mode 100644
index 000000000..421f3c7ca
--- /dev/null
+++ b/libjpegtwrp/jidctred.c
@@ -0,0 +1,398 @@
+/*
+ * jidctred.c
+ *
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains inverse-DCT routines that produce reduced-size output:
+ * either 4x4, 2x2, or 1x1 pixels from an 8x8 DCT block.
+ *
+ * The implementation is based on the Loeffler, Ligtenberg and Moschytz (LL&M)
+ * algorithm used in jidctint.c.  We simply replace each 8-to-8 1-D IDCT step
+ * with an 8-to-4 step that produces the four averages of two adjacent outputs
+ * (or an 8-to-2 step producing two averages of four outputs, for 2x2 output).
+ * These steps were derived by computing the corresponding values at the end
+ * of the normal LL&M code, then simplifying as much as possible.
+ *
+ * 1x1 is trivial: just take the DC coefficient divided by 8.
+ *
+ * See jidctint.c for additional comments.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef IDCT_SCALING_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Scaling is the same as in jidctint.c. */
+
+#if BITS_IN_JSAMPLE == 8
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#else
+#define CONST_BITS  13
+#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#endif
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 13
+#define FIX_0_211164243  ((INT32)  1730)	/* FIX(0.211164243) */
+#define FIX_0_509795579  ((INT32)  4176)	/* FIX(0.509795579) */
+#define FIX_0_601344887  ((INT32)  4926)	/* FIX(0.601344887) */
+#define FIX_0_720959822  ((INT32)  5906)	/* FIX(0.720959822) */
+#define FIX_0_765366865  ((INT32)  6270)	/* FIX(0.765366865) */
+#define FIX_0_850430095  ((INT32)  6967)	/* FIX(0.850430095) */
+#define FIX_0_899976223  ((INT32)  7373)	/* FIX(0.899976223) */
+#define FIX_1_061594337  ((INT32)  8697)	/* FIX(1.061594337) */
+#define FIX_1_272758580  ((INT32)  10426)	/* FIX(1.272758580) */
+#define FIX_1_451774981  ((INT32)  11893)	/* FIX(1.451774981) */
+#define FIX_1_847759065  ((INT32)  15137)	/* FIX(1.847759065) */
+#define FIX_2_172734803  ((INT32)  17799)	/* FIX(2.172734803) */
+#define FIX_2_562915447  ((INT32)  20995)	/* FIX(2.562915447) */
+#define FIX_3_624509785  ((INT32)  29692)	/* FIX(3.624509785) */
+#else
+#define FIX_0_211164243  FIX(0.211164243)
+#define FIX_0_509795579  FIX(0.509795579)
+#define FIX_0_601344887  FIX(0.601344887)
+#define FIX_0_720959822  FIX(0.720959822)
+#define FIX_0_765366865  FIX(0.765366865)
+#define FIX_0_850430095  FIX(0.850430095)
+#define FIX_0_899976223  FIX(0.899976223)
+#define FIX_1_061594337  FIX(1.061594337)
+#define FIX_1_272758580  FIX(1.272758580)
+#define FIX_1_451774981  FIX(1.451774981)
+#define FIX_1_847759065  FIX(1.847759065)
+#define FIX_2_172734803  FIX(2.172734803)
+#define FIX_2_562915447  FIX(2.562915447)
+#define FIX_3_624509785  FIX(3.624509785)
+#endif
+
+
+/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+ * For 8-bit samples with the recommended scaling, all the variable
+ * and constant values involved are no more than 16 bits wide, so a
+ * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
+ * For 12-bit samples, a full 32-bit multiplication will be needed.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define MULTIPLY(var,const)  MULTIPLY16C16(var,const)
+#else
+#define MULTIPLY(var,const)  ((var) * (const))
+#endif
+
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce an int result.  In this module, both inputs and result
+ * are 16 bits or less, so either int or short multiply will work.
+ */
+
+#define DEQUANTIZE(coef,quantval)  (((ISLOW_MULT_TYPE) (coef)) * (quantval))
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 4x4 output block.
+ */
+
+GLOBAL(void)
+jpeg_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	       JCOEFPTR coef_block,
+	       JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  INT32 tmp0, tmp2, tmp10, tmp12;
+  INT32 z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE * quantptr;
+  int * wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[DCTSIZE*4];	/* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; inptr++, quantptr++, wsptr++, ctr--) {
+    /* Don't bother to process column 4, because second pass won't use it */
+    if (ctr == DCTSIZE-4)
+      continue;
+    if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
+	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*5] == 0 &&
+	inptr[DCTSIZE*6] == 0 && inptr[DCTSIZE*7] == 0) {
+      /* AC terms all zero; we need not examine term 4 for 4x4 output */
+      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
+      
+      wsptr[DCTSIZE*0] = dcval;
+      wsptr[DCTSIZE*1] = dcval;
+      wsptr[DCTSIZE*2] = dcval;
+      wsptr[DCTSIZE*3] = dcval;
+      
+      continue;
+    }
+    
+    /* Even part */
+    
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp0 <<= (CONST_BITS+1);
+    
+    z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp2 = MULTIPLY(z2, FIX_1_847759065) + MULTIPLY(z3, - FIX_0_765366865);
+    
+    tmp10 = tmp0 + tmp2;
+    tmp12 = tmp0 - tmp2;
+    
+    /* Odd part */
+    
+    z1 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    
+    tmp0 = MULTIPLY(z1, - FIX_0_211164243) /* sqrt(2) * (c3-c1) */
+	 + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */
+	 + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */
+	 + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */
+    
+    tmp2 = MULTIPLY(z1, - FIX_0_509795579) /* sqrt(2) * (c7-c5) */
+	 + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */
+	 + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */
+	 + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */
+
+    /* Final output stage */
+    
+    wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp2, CONST_BITS-PASS1_BITS+1);
+    wsptr[DCTSIZE*3] = (int) DESCALE(tmp10 - tmp2, CONST_BITS-PASS1_BITS+1);
+    wsptr[DCTSIZE*1] = (int) DESCALE(tmp12 + tmp0, CONST_BITS-PASS1_BITS+1);
+    wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 - tmp0, CONST_BITS-PASS1_BITS+1);
+  }
+  
+  /* Pass 2: process 4 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 4; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+    /* It's not clear whether a zero row test is worthwhile here ... */
+
+#ifndef NO_ZERO_ROW_TEST
+    if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 &&
+	wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
+      /* AC terms all zero */
+      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
+				  & RANGE_MASK];
+      
+      outptr[0] = dcval;
+      outptr[1] = dcval;
+      outptr[2] = dcval;
+      outptr[3] = dcval;
+      
+      wsptr += DCTSIZE;		/* advance pointer to next row */
+      continue;
+    }
+#endif
+    
+    /* Even part */
+    
+    tmp0 = ((INT32) wsptr[0]) << (CONST_BITS+1);
+    
+    tmp2 = MULTIPLY((INT32) wsptr[2], FIX_1_847759065)
+	 + MULTIPLY((INT32) wsptr[6], - FIX_0_765366865);
+    
+    tmp10 = tmp0 + tmp2;
+    tmp12 = tmp0 - tmp2;
+    
+    /* Odd part */
+    
+    z1 = (INT32) wsptr[7];
+    z2 = (INT32) wsptr[5];
+    z3 = (INT32) wsptr[3];
+    z4 = (INT32) wsptr[1];
+    
+    tmp0 = MULTIPLY(z1, - FIX_0_211164243) /* sqrt(2) * (c3-c1) */
+	 + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */
+	 + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */
+	 + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */
+    
+    tmp2 = MULTIPLY(z1, - FIX_0_509795579) /* sqrt(2) * (c7-c5) */
+	 + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */
+	 + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */
+	 + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */
+
+    /* Final output stage */
+    
+    outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp2,
+					  CONST_BITS+PASS1_BITS+3+1)
+			    & RANGE_MASK];
+    outptr[3] = range_limit[(int) DESCALE(tmp10 - tmp2,
+					  CONST_BITS+PASS1_BITS+3+1)
+			    & RANGE_MASK];
+    outptr[1] = range_limit[(int) DESCALE(tmp12 + tmp0,
+					  CONST_BITS+PASS1_BITS+3+1)
+			    & RANGE_MASK];
+    outptr[2] = range_limit[(int) DESCALE(tmp12 - tmp0,
+					  CONST_BITS+PASS1_BITS+3+1)
+			    & RANGE_MASK];
+    
+    wsptr += DCTSIZE;		/* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 2x2 output block.
+ */
+
+GLOBAL(void)
+jpeg_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	       JCOEFPTR coef_block,
+	       JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  INT32 tmp0, tmp10, z1;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE * quantptr;
+  int * wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[DCTSIZE*2];	/* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; inptr++, quantptr++, wsptr++, ctr--) {
+    /* Don't bother to process columns 2,4,6 */
+    if (ctr == DCTSIZE-2 || ctr == DCTSIZE-4 || ctr == DCTSIZE-6)
+      continue;
+    if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*3] == 0 &&
+	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*7] == 0) {
+      /* AC terms all zero; we need not examine terms 2,4,6 for 2x2 output */
+      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
+      
+      wsptr[DCTSIZE*0] = dcval;
+      wsptr[DCTSIZE*1] = dcval;
+      
+      continue;
+    }
+    
+    /* Even part */
+    
+    z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp10 = z1 << (CONST_BITS+2);
+    
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+    tmp0 = MULTIPLY(z1, - FIX_0_720959822); /* sqrt(2) * (c7-c5+c3-c1) */
+    z1 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    tmp0 += MULTIPLY(z1, FIX_0_850430095); /* sqrt(2) * (-c1+c3+c5+c7) */
+    z1 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    tmp0 += MULTIPLY(z1, - FIX_1_272758580); /* sqrt(2) * (-c1+c3-c5-c7) */
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    tmp0 += MULTIPLY(z1, FIX_3_624509785); /* sqrt(2) * (c1+c3+c5+c7) */
+
+    /* Final output stage */
+    
+    wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp0, CONST_BITS-PASS1_BITS+2);
+    wsptr[DCTSIZE*1] = (int) DESCALE(tmp10 - tmp0, CONST_BITS-PASS1_BITS+2);
+  }
+  
+  /* Pass 2: process 2 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 2; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+    /* It's not clear whether a zero row test is worthwhile here ... */
+
+#ifndef NO_ZERO_ROW_TEST
+    if (wsptr[1] == 0 && wsptr[3] == 0 && wsptr[5] == 0 && wsptr[7] == 0) {
+      /* AC terms all zero */
+      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
+				  & RANGE_MASK];
+      
+      outptr[0] = dcval;
+      outptr[1] = dcval;
+      
+      wsptr += DCTSIZE;		/* advance pointer to next row */
+      continue;
+    }
+#endif
+    
+    /* Even part */
+    
+    tmp10 = ((INT32) wsptr[0]) << (CONST_BITS+2);
+    
+    /* Odd part */
+
+    tmp0 = MULTIPLY((INT32) wsptr[7], - FIX_0_720959822) /* sqrt(2) * (c7-c5+c3-c1) */
+	 + MULTIPLY((INT32) wsptr[5], FIX_0_850430095) /* sqrt(2) * (-c1+c3+c5+c7) */
+	 + MULTIPLY((INT32) wsptr[3], - FIX_1_272758580) /* sqrt(2) * (-c1+c3-c5-c7) */
+	 + MULTIPLY((INT32) wsptr[1], FIX_3_624509785); /* sqrt(2) * (c1+c3+c5+c7) */
+
+    /* Final output stage */
+    
+    outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp0,
+					  CONST_BITS+PASS1_BITS+3+2)
+			    & RANGE_MASK];
+    outptr[1] = range_limit[(int) DESCALE(tmp10 - tmp0,
+					  CONST_BITS+PASS1_BITS+3+2)
+			    & RANGE_MASK];
+    
+    wsptr += DCTSIZE;		/* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 1x1 output block.
+ */
+
+GLOBAL(void)
+jpeg_idct_1x1 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	       JCOEFPTR coef_block,
+	       JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  int dcval;
+  ISLOW_MULT_TYPE * quantptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  SHIFT_TEMPS
+
+  /* We hardly need an inverse DCT routine for this: just take the
+   * average pixel value, which is one-eighth of the DC coefficient.
+   */
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  dcval = DEQUANTIZE(coef_block[0], quantptr[0]);
+  dcval = (int) DESCALE((INT32) dcval, 3);
+
+  output_buf[0][output_col] = range_limit[dcval & RANGE_MASK];
+}
+
+#endif /* IDCT_SCALING_SUPPORTED */
diff --git a/libjpegtwrp/jidctvenum.c b/libjpegtwrp/jidctvenum.c
new file mode 100644
index 000000000..f458bb565
--- /dev/null
+++ b/libjpegtwrp/jidctvenum.c
@@ -0,0 +1,229 @@
+/*
+ * jidctvenum.c
+ *
+ * Copyright (c) 2010, Code Aurora Forum. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above
+ *       copyright notice, this list of conditions and the following
+ *       disclaimer in the documentation and/or other materials provided
+ *       with the distribution.
+ *     * Neither the name of Code Aurora Forum, Inc. nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED "AS IS" AND ANY EXPRESS OR IMPLIED
+ * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
+ * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
+ * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
+ * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN
+ * IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"    /* Private declarations for DCT subsystem */
+
+#ifdef ANDROID_JPEG_USE_VENUM
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce an int result.  In this module, both inputs and result
+ * are 16 bits or less, so either int or short multiply will work.
+ */
+
+#define DEQUANTIZE(coef,quantval)  ((coef) * ((INT16)quantval))
+
+GLOBAL(void)
+jpeg_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+         JCOEFPTR coef_block,
+         JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  ISLOW_MULT_TYPE * quantptr;
+  JCOEFPTR coefptr;
+  int ctr;
+
+  /* idct_out temp buffer is needed because output_buf sample allocation is 8 bits,
+   * while IDCT output expects 16 bits.
+   */
+  INT16 idct_out[DCTSIZE2];  /* buffers data between passes */
+  JSAMPROW outptr;
+  INT16*  idctptr;
+
+  coefptr  = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+
+  /* Dequantize the coeff buffer and write it back to the same location */
+  for (ctr = DCTSIZE; ctr > 0; ctr--) {
+    coefptr[0]         = DEQUANTIZE(coefptr[0]        , quantptr[0]        );
+    coefptr[DCTSIZE*1] = DEQUANTIZE(coefptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    coefptr[DCTSIZE*2] = DEQUANTIZE(coefptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    coefptr[DCTSIZE*3] = DEQUANTIZE(coefptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    coefptr[DCTSIZE*4] = DEQUANTIZE(coefptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    coefptr[DCTSIZE*5] = DEQUANTIZE(coefptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    coefptr[DCTSIZE*6] = DEQUANTIZE(coefptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+    coefptr[DCTSIZE*7] = DEQUANTIZE(coefptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    /* advance pointers to next column */
+    quantptr++;
+    coefptr++;
+  }
+
+  idct_8x8_venum((INT16*)coef_block,
+                 (INT16*)idct_out,
+                 DCTSIZE * sizeof(INT16));
+
+  idctptr = idct_out;
+  for (ctr = 0; ctr < DCTSIZE; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+    // outptr sample size is 1 byte while idctptr sample size is 2 bytes
+    outptr[0] = idctptr[0];
+    outptr[1] = idctptr[1];
+    outptr[2] = idctptr[2];
+    outptr[3] = idctptr[3];
+    outptr[4] = idctptr[4];
+    outptr[5] = idctptr[5];
+    outptr[6] = idctptr[6];
+    outptr[7] = idctptr[7];
+    idctptr  += DCTSIZE;      /* advance pointers to next row */
+  }
+}
+
+GLOBAL(void)
+jpeg_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+         JCOEFPTR coef_block,
+         JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  ISLOW_MULT_TYPE * quantptr;
+  JSAMPROW outptr;
+
+  /* Note: Must allocate 8x4 even though only 4x4 is used because
+   * IDCT function expects stride of 8. Stride input to function is ignored.
+   */
+  INT16    idct_out[DCTSIZE * (DCTSIZE>>1)];  /* buffers data between passes */
+  INT16*   idctptr;
+  JCOEFPTR coefptr;
+  int ctr;
+
+  coefptr  = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+
+  /* Dequantize the coeff buffer and write it back to the same location */
+  for (ctr = (DCTSIZE>>1); ctr > 0; ctr--) {
+    coefptr[0]         = DEQUANTIZE(coefptr[0]        , quantptr[0]        );
+    coefptr[DCTSIZE*1] = DEQUANTIZE(coefptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    coefptr[DCTSIZE*2] = DEQUANTIZE(coefptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    coefptr[DCTSIZE*3] = DEQUANTIZE(coefptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+
+    /* advance pointers to next column */
+    quantptr++;
+    coefptr++;
+  }
+
+  idct_4x4_venum((INT16*)coef_block,
+                 (INT16*)idct_out,
+                  DCTSIZE * sizeof(INT16));
+
+  idctptr = idct_out;
+  for (ctr = 0; ctr < (DCTSIZE>>1); ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* outptr sample size is 1 byte while idctptr sample size is 2 bytes */
+    outptr[0] = idctptr[0];
+    outptr[1] = idctptr[1];
+    outptr[2] = idctptr[2];
+    outptr[3] = idctptr[3];
+    /* IDCT function assumes stride of 8 units */
+    idctptr += (DCTSIZE);    /* advance pointers to next row */
+  }
+}
+
+GLOBAL(void)
+jpeg_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+         JCOEFPTR coef_block,
+         JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  ISLOW_MULT_TYPE * quantptr;
+  JSAMPROW outptr;
+
+  /* Note: Must allocate 8x2 even though only 2x2 is used because
+   * IDCT function expects stride of 8. Stride input to function is ignored.
+   * There is also a hw limitation requiring input size to be 8x2.
+   */
+  INT16    idct_out[DCTSIZE * (DCTSIZE>>2)];  /* buffers data between passes */
+  INT16*   idctptr;
+  JCOEFPTR coefptr;
+  int ctr;
+
+  coefptr  = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+
+  /* Dequantize the coeff buffer and write it back to the same location */
+  for (ctr = (DCTSIZE>>2); ctr > 0; ctr--) {
+    coefptr[0]         = DEQUANTIZE(coefptr[0]        , quantptr[0]        );
+    coefptr[DCTSIZE*1] = DEQUANTIZE(coefptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+
+    /* advance pointers to next column */
+    quantptr++;
+    coefptr++;
+  }
+
+  idct_2x2_venum((INT16*)coef_block,
+                 (INT16*)idct_out,
+                  DCTSIZE * sizeof(INT16));
+
+  idctptr = idct_out;
+  for (ctr = 0; ctr < (DCTSIZE>>2); ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* outptr sample size is 1 bytes, idctptr sample size is 2 bytes */
+    outptr[0] = idctptr[0];
+    outptr[1] = idctptr[1];
+
+    /* IDCT function assumes stride of 8 units */
+    idctptr += (DCTSIZE);    /* advance pointers to next row */
+  }
+}
+
+
+GLOBAL(void)
+jpeg_idct_1x1 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+         JCOEFPTR coef_block,
+         JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  ISLOW_MULT_TYPE * quantptr;
+  JSAMPROW outptr; // 8-bit type
+  INT16    idct_out[DCTSIZE]; /* Required to allocate 8 samples, even though we only use one. */
+  JCOEFPTR coefptr;
+  int ctr;
+
+  coefptr  = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  outptr   = output_buf[0] + output_col;
+
+  /* Dequantize the coeff buffer and write it back to the same location */
+  coefptr[0] = DEQUANTIZE(coefptr[0], quantptr[0]);
+
+  idct_1x1_venum((INT16*)coef_block,
+                 (INT16*)idct_out,
+                 DCTSIZE * sizeof(INT16));
+  outptr[0] = idct_out[0];
+}
+
+
+#endif /* ANDROID_JPEG_USE_VENUM */
diff --git a/libjpegtwrp/jinclude.h b/libjpegtwrp/jinclude.h
new file mode 100644
index 000000000..0a4f15146
--- /dev/null
+++ b/libjpegtwrp/jinclude.h
@@ -0,0 +1,91 @@
+/*
+ * jinclude.h
+ *
+ * Copyright (C) 1991-1994, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file exists to provide a single place to fix any problems with
+ * including the wrong system include files.  (Common problems are taken
+ * care of by the standard jconfig symbols, but on really weird systems
+ * you may have to edit this file.)
+ *
+ * NOTE: this file is NOT intended to be included by applications using the
+ * JPEG library.  Most applications need only include jpeglib.h.
+ */
+
+
+/* Include auto-config file to find out which system include files we need. */
+
+#include "jconfig.h"		/* auto configuration options */
+#define JCONFIG_INCLUDED	/* so that jpeglib.h doesn't do it again */
+
+/*
+ * We need the NULL macro and size_t typedef.
+ * On an ANSI-conforming system it is sufficient to include <stddef.h>.
+ * Otherwise, we get them from <stdlib.h> or <stdio.h>; we may have to
+ * pull in <sys/types.h> as well.
+ * Note that the core JPEG library does not require <stdio.h>;
+ * only the default error handler and data source/destination modules do.
+ * But we must pull it in because of the references to FILE in jpeglib.h.
+ * You can remove those references if you want to compile without <stdio.h>.
+ */
+
+#ifdef HAVE_STDDEF_H
+#include <stddef.h>
+#endif
+
+#ifdef HAVE_STDLIB_H
+#include <stdlib.h>
+#endif
+
+#ifdef NEED_SYS_TYPES_H
+#include <sys/types.h>
+#endif
+
+#include <stdio.h>
+
+/*
+ * We need memory copying and zeroing functions, plus strncpy().
+ * ANSI and System V implementations declare these in <string.h>.
+ * BSD doesn't have the mem() functions, but it does have bcopy()/bzero().
+ * Some systems may declare memset and memcpy in <memory.h>.
+ *
+ * NOTE: we assume the size parameters to these functions are of type size_t.
+ * Change the casts in these macros if not!
+ */
+
+#ifdef NEED_BSD_STRINGS
+
+#include <strings.h>
+#define MEMZERO(target,size)	bzero((void *)(target), (size_t)(size))
+#define MEMCOPY(dest,src,size)	bcopy((const void *)(src), (void *)(dest), (size_t)(size))
+
+#else /* not BSD, assume ANSI/SysV string lib */
+
+#include <string.h>
+#define MEMZERO(target,size)	memset((void *)(target), 0, (size_t)(size))
+#define MEMCOPY(dest,src,size)	memcpy((void *)(dest), (const void *)(src), (size_t)(size))
+
+#endif
+
+/*
+ * In ANSI C, and indeed any rational implementation, size_t is also the
+ * type returned by sizeof().  However, it seems there are some irrational
+ * implementations out there, in which sizeof() returns an int even though
+ * size_t is defined as long or unsigned long.  To ensure consistent results
+ * we always use this SIZEOF() macro in place of using sizeof() directly.
+ */
+
+#define SIZEOF(object)	((size_t) sizeof(object))
+
+/*
+ * The modules that use fread() and fwrite() always invoke them through
+ * these macros.  On some systems you may need to twiddle the argument casts.
+ * CAUTION: argument order is different from underlying functions!
+ */
+
+#define JFREAD(file,buf,sizeofbuf)  \
+  ((size_t) fread((void *) (buf), (size_t) 1, (size_t) (sizeofbuf), (file)))
+#define JFWRITE(file,buf,sizeofbuf)  \
+  ((size_t) fwrite((const void *) (buf), (size_t) 1, (size_t) (sizeofbuf), (file)))
diff --git a/libjpegtwrp/jmem-android.c b/libjpegtwrp/jmem-android.c
new file mode 100644
index 000000000..5da4d0a6d
--- /dev/null
+++ b/libjpegtwrp/jmem-android.c
@@ -0,0 +1,186 @@
+/*
+ * Copyright (C) 2007-2008 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jmemsys.h"		/* import the system-dependent declarations */
+
+#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare malloc(),free() */
+extern void * malloc JPP((size_t size));
+extern void free JPP((void *ptr));
+#endif
+
+#ifndef SEEK_SET		/* pre-ANSI systems may not define this; */
+#define SEEK_SET  0		/* if not, assume 0 is correct */
+#endif
+
+
+/*
+ * Memory allocation and freeing are controlled by the regular library
+ * routines malloc() and free().
+ */
+
+GLOBAL(void *)
+jpeg_get_small (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_small (j_common_ptr cinfo, void * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * "Large" objects are treated the same as "small" ones.
+ * NB: although we include FAR keywords in the routine declarations,
+ * this file won't actually work in 80x86 small/medium model; at least,
+ * you probably won't be able to process useful-size images in only 64KB.
+ */
+
+GLOBAL(void FAR *)
+jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void FAR *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_large (j_common_ptr cinfo, void FAR * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * This routine computes the total memory space available for allocation.
+ * It's impossible to do this in a portable way; our current solution is
+ * to make the user tell us (with a default value set at compile time).
+ * If you can actually get the available space, it's a good idea to subtract
+ * a slop factor of 5% or so.
+ */
+
+#ifndef DEFAULT_MAX_MEM		/* so can override from makefile */
+#define DEFAULT_MAX_MEM		10000000L /* default: ten megabyte */
+#endif
+
+GLOBAL(long)
+jpeg_mem_available (j_common_ptr cinfo, long min_bytes_needed,
+		    long max_bytes_needed, long already_allocated)
+{
+  return cinfo->mem->max_memory_to_use - already_allocated;
+}
+
+
+/*
+ * Backing store (temporary file) management.
+ * Backing store objects are only used when the value returned by
+ * jpeg_mem_available is less than the total space needed.  You can dispense
+ * with these routines if you have plenty of virtual memory; see jmemnobs.c.
+ */
+
+
+METHODDEF(void)
+read_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+		    void FAR * buffer_address,
+		    long file_offset, long byte_count)
+{
+  if (fseek(info->temp_file, file_offset, SEEK_SET))
+    ERREXIT(cinfo, JERR_TFILE_SEEK);
+  if (JFREAD(info->temp_file, buffer_address, byte_count)
+      != (size_t) byte_count)
+    ERREXIT(cinfo, JERR_TFILE_READ);
+}
+
+
+METHODDEF(void)
+write_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+		     void FAR * buffer_address,
+		     long file_offset, long byte_count)
+{
+  if (fseek(info->temp_file, file_offset, SEEK_SET))
+    ERREXIT(cinfo, JERR_TFILE_SEEK);
+  if (JFWRITE(info->temp_file, buffer_address, byte_count)
+      != (size_t) byte_count)
+    ERREXIT(cinfo, JERR_TFILE_WRITE);
+}
+
+
+METHODDEF(void)
+close_backing_store (j_common_ptr cinfo, backing_store_ptr info)
+{
+  fclose(info->temp_file);
+  /* Since this implementation uses tmpfile() to create the file,
+   * no explicit file deletion is needed.
+   */
+}
+
+static FILE* getTempFileFromPath(const char * path) {
+    FILE * fd = fopen(path, "w+");
+    unlink(path);
+    return fd;
+}
+
+static FILE* getTempFile() {
+    char path[1024];
+    snprintf(path, 1023, "/sdcard/.%d.tmp", getpid());
+    FILE * fd = getTempFileFromPath(path);
+    if (fd == NULL) {
+        // anywhere else we can create a temp file?
+		//	    snprintf(path, 1023, "/data/data/.%d.tmp", getpid());
+		//      fd = getTempFileFromPath(path);
+    }
+    return fd;
+}
+
+/*
+ * Initial opening of a backing-store object.
+ *
+ * This version uses tmpfile(), which constructs a suitable file name
+ * behind the scenes.  We don't have to use info->temp_name[] at all;
+ * indeed, we can't even find out the actual name of the temp file.
+ */
+
+GLOBAL(void)
+jpeg_open_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+			 long total_bytes_needed)
+{
+  if ((info->temp_file = getTempFile()) == NULL)
+    ERREXITS(cinfo, JERR_TFILE_CREATE, "");
+  info->read_backing_store = read_backing_store;
+  info->write_backing_store = write_backing_store;
+  info->close_backing_store = close_backing_store;
+}
+
+
+/*
+ * These routines take care of any system-dependent initialization and
+ * cleanup required.
+ */
+
+GLOBAL(long)
+jpeg_mem_init (j_common_ptr cinfo)
+{
+  return DEFAULT_MAX_MEM;	/* default for max_memory_to_use */
+}
+
+GLOBAL(void)
+jpeg_mem_term (j_common_ptr cinfo)
+{
+  /* no work */
+}
diff --git a/libjpegtwrp/jmem-ashmem.c b/libjpegtwrp/jmem-ashmem.c
new file mode 100644
index 000000000..3a17b02c3
--- /dev/null
+++ b/libjpegtwrp/jmem-ashmem.c
@@ -0,0 +1,170 @@
+/*
+ * Copyright (C) 2007-2008 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jmemsys.h"		/* import the system-dependent declarations */
+
+#include <cutils/ashmem.h>
+#include <unistd.h>
+#include <sys/mman.h>
+
+#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare malloc(),free() */
+extern void * malloc JPP((size_t size));
+extern void free JPP((void *ptr));
+#endif
+
+/*
+ * Memory allocation and freeing are controlled by the regular library
+ * routines malloc() and free().
+ */
+
+GLOBAL(void *)
+jpeg_get_small (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_small (j_common_ptr cinfo, void * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * "Large" objects are treated the same as "small" ones.
+ * NB: although we include FAR keywords in the routine declarations,
+ * this file won't actually work in 80x86 small/medium model; at least,
+ * you probably won't be able to process useful-size images in only 64KB.
+ */
+
+GLOBAL(void FAR *)
+jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void FAR *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_large (j_common_ptr cinfo, void FAR * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+/*
+ * This routine computes the total memory space available for allocation.
+ * It's impossible to do this in a portable way; our current solution is
+ * to make the user tell us (with a default value set at compile time).
+ * If you can actually get the available space, it's a good idea to subtract
+ * a slop factor of 5% or so.
+ */
+
+#ifndef DEFAULT_MAX_MEM		/* so can override from makefile */
+#define DEFAULT_MAX_MEM		10000000L /* default: ten megabyte */
+#endif
+
+GLOBAL(long)
+jpeg_mem_available (j_common_ptr cinfo, long min_bytes_needed,
+		    long max_bytes_needed, long already_allocated)
+{
+  return cinfo->mem->max_memory_to_use - already_allocated;
+}
+
+
+/*
+ * Backing store (temporary file) management.
+ * Backing store objects are only used when the value returned by
+ * jpeg_mem_available is less than the total space needed.  You can dispense
+ * with these routines if you have plenty of virtual memory; see jmemnobs.c.
+ */
+
+METHODDEF(void)
+read_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+		    void FAR * buffer_address,
+		    long file_offset, long byte_count)
+{
+  memmove(buffer_address, info->addr + file_offset, byte_count);
+}
+
+
+METHODDEF(void)
+write_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+		     void FAR * buffer_address,
+		     long file_offset, long byte_count)
+{
+  memmove(info->addr + file_offset, buffer_address, byte_count);
+}
+
+
+METHODDEF(void)
+close_backing_store (j_common_ptr cinfo, backing_store_ptr info)
+{
+  munmap(info->addr, info->size);
+  close(info->temp_file);
+}
+
+LOCAL(int)
+get_ashmem(backing_store_ptr info, long total_bytes_needed)
+{
+  char path[1024];
+  snprintf(path, 1023, "%d.tmp.ashmem", getpid());
+  int fd = ashmem_create_region(path, total_bytes_needed);
+  if (fd == -1) {
+      return -1;
+  }
+  int err = ashmem_set_prot_region(fd, PROT_READ | PROT_WRITE);
+  if (err) {
+      return -1;
+  }
+  info->addr = mmap(NULL, total_bytes_needed, PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0);
+  info->size = total_bytes_needed;
+  info->temp_file = fd;
+  return fd;
+}
+
+/*
+ * Initial opening of a backing-store object.
+ * This version uses ashmem to get a shared memory of total-bytes_needed.
+ */
+
+GLOBAL(void)
+jpeg_open_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+			 long total_bytes_needed)
+{
+  if (get_ashmem(info, total_bytes_needed) == -1)
+    ERREXITS(cinfo, JERR_TFILE_CREATE, "");
+  info->read_backing_store = read_backing_store;
+  info->write_backing_store = write_backing_store;
+  info->close_backing_store = close_backing_store;
+}
+
+/*
+ * These routines take care of any system-dependent initialization and
+ * cleanup required.
+ */
+
+GLOBAL(long)
+jpeg_mem_init (j_common_ptr cinfo)
+{
+  return DEFAULT_MAX_MEM;	/* default for max_memory_to_use */
+}
+
+GLOBAL(void)
+jpeg_mem_term (j_common_ptr cinfo)
+{
+  /* no work */
+}
diff --git a/libjpegtwrp/jmemansi.c b/libjpegtwrp/jmemansi.c
new file mode 100644
index 000000000..2d93e4962
--- /dev/null
+++ b/libjpegtwrp/jmemansi.c
@@ -0,0 +1,167 @@
+/*
+ * jmemansi.c
+ *
+ * Copyright (C) 1992-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file provides a simple generic implementation of the system-
+ * dependent portion of the JPEG memory manager.  This implementation
+ * assumes that you have the ANSI-standard library routine tmpfile().
+ * Also, the problem of determining the amount of memory available
+ * is shoved onto the user.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jmemsys.h"		/* import the system-dependent declarations */
+
+#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare malloc(),free() */
+extern void * malloc JPP((size_t size));
+extern void free JPP((void *ptr));
+#endif
+
+#ifndef SEEK_SET		/* pre-ANSI systems may not define this; */
+#define SEEK_SET  0		/* if not, assume 0 is correct */
+#endif
+
+
+/*
+ * Memory allocation and freeing are controlled by the regular library
+ * routines malloc() and free().
+ */
+
+GLOBAL(void *)
+jpeg_get_small (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_small (j_common_ptr cinfo, void * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * "Large" objects are treated the same as "small" ones.
+ * NB: although we include FAR keywords in the routine declarations,
+ * this file won't actually work in 80x86 small/medium model; at least,
+ * you probably won't be able to process useful-size images in only 64KB.
+ */
+
+GLOBAL(void FAR *)
+jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void FAR *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_large (j_common_ptr cinfo, void FAR * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * This routine computes the total memory space available for allocation.
+ * It's impossible to do this in a portable way; our current solution is
+ * to make the user tell us (with a default value set at compile time).
+ * If you can actually get the available space, it's a good idea to subtract
+ * a slop factor of 5% or so.
+ */
+
+#ifndef DEFAULT_MAX_MEM		/* so can override from makefile */
+#define DEFAULT_MAX_MEM		1000000L /* default: one megabyte */
+#endif
+
+GLOBAL(long)
+jpeg_mem_available (j_common_ptr cinfo, long min_bytes_needed,
+		    long max_bytes_needed, long already_allocated)
+{
+  return cinfo->mem->max_memory_to_use - already_allocated;
+}
+
+
+/*
+ * Backing store (temporary file) management.
+ * Backing store objects are only used when the value returned by
+ * jpeg_mem_available is less than the total space needed.  You can dispense
+ * with these routines if you have plenty of virtual memory; see jmemnobs.c.
+ */
+
+
+METHODDEF(void)
+read_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+		    void FAR * buffer_address,
+		    long file_offset, long byte_count)
+{
+  if (fseek(info->temp_file, file_offset, SEEK_SET))
+    ERREXIT(cinfo, JERR_TFILE_SEEK);
+  if (JFREAD(info->temp_file, buffer_address, byte_count)
+      != (size_t) byte_count)
+    ERREXIT(cinfo, JERR_TFILE_READ);
+}
+
+
+METHODDEF(void)
+write_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+		     void FAR * buffer_address,
+		     long file_offset, long byte_count)
+{
+  if (fseek(info->temp_file, file_offset, SEEK_SET))
+    ERREXIT(cinfo, JERR_TFILE_SEEK);
+  if (JFWRITE(info->temp_file, buffer_address, byte_count)
+      != (size_t) byte_count)
+    ERREXIT(cinfo, JERR_TFILE_WRITE);
+}
+
+
+METHODDEF(void)
+close_backing_store (j_common_ptr cinfo, backing_store_ptr info)
+{
+  fclose(info->temp_file);
+  /* Since this implementation uses tmpfile() to create the file,
+   * no explicit file deletion is needed.
+   */
+}
+
+
+/*
+ * Initial opening of a backing-store object.
+ *
+ * This version uses tmpfile(), which constructs a suitable file name
+ * behind the scenes.  We don't have to use info->temp_name[] at all;
+ * indeed, we can't even find out the actual name of the temp file.
+ */
+
+GLOBAL(void)
+jpeg_open_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+			 long total_bytes_needed)
+{
+  if ((info->temp_file = tmpfile()) == NULL)
+    ERREXITS(cinfo, JERR_TFILE_CREATE, "");
+  info->read_backing_store = read_backing_store;
+  info->write_backing_store = write_backing_store;
+  info->close_backing_store = close_backing_store;
+}
+
+
+/*
+ * These routines take care of any system-dependent initialization and
+ * cleanup required.
+ */
+
+GLOBAL(long)
+jpeg_mem_init (j_common_ptr cinfo)
+{
+  return DEFAULT_MAX_MEM;	/* default for max_memory_to_use */
+}
+
+GLOBAL(void)
+jpeg_mem_term (j_common_ptr cinfo)
+{
+  /* no work */
+}
diff --git a/libjpegtwrp/jmemdos.c b/libjpegtwrp/jmemdos.c
new file mode 100644
index 000000000..60b45c693
--- /dev/null
+++ b/libjpegtwrp/jmemdos.c
@@ -0,0 +1,638 @@
+/*
+ * jmemdos.c
+ *
+ * Copyright (C) 1992-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file provides an MS-DOS-compatible implementation of the system-
+ * dependent portion of the JPEG memory manager.  Temporary data can be
+ * stored in extended or expanded memory as well as in regular DOS files.
+ *
+ * If you use this file, you must be sure that NEED_FAR_POINTERS is defined
+ * if you compile in a small-data memory model; it should NOT be defined if
+ * you use a large-data memory model.  This file is not recommended if you
+ * are using a flat-memory-space 386 environment such as DJGCC or Watcom C.
+ * Also, this code will NOT work if struct fields are aligned on greater than
+ * 2-byte boundaries.
+ *
+ * Based on code contributed by Ge' Weijers.
+ */
+
+/*
+ * If you have both extended and expanded memory, you may want to change the
+ * order in which they are tried in jopen_backing_store.  On a 286 machine
+ * expanded memory is usually faster, since extended memory access involves
+ * an expensive protected-mode-and-back switch.  On 386 and better, extended
+ * memory is usually faster.  As distributed, the code tries extended memory
+ * first (what? not everyone has a 386? :-).
+ *
+ * You can disable use of extended/expanded memory entirely by altering these
+ * definitions or overriding them from the Makefile (eg, -DEMS_SUPPORTED=0).
+ */
+
+#ifndef XMS_SUPPORTED
+#define XMS_SUPPORTED  1
+#endif
+#ifndef EMS_SUPPORTED
+#define EMS_SUPPORTED  1
+#endif
+
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jmemsys.h"		/* import the system-dependent declarations */
+
+#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare these */
+extern void * malloc JPP((size_t size));
+extern void free JPP((void *ptr));
+extern char * getenv JPP((const char * name));
+#endif
+
+#ifdef NEED_FAR_POINTERS
+
+#ifdef __TURBOC__
+/* These definitions work for Borland C (Turbo C) */
+#include <alloc.h>		/* need farmalloc(), farfree() */
+#define far_malloc(x)	farmalloc(x)
+#define far_free(x)	farfree(x)
+#else
+/* These definitions work for Microsoft C and compatible compilers */
+#include <malloc.h>		/* need _fmalloc(), _ffree() */
+#define far_malloc(x)	_fmalloc(x)
+#define far_free(x)	_ffree(x)
+#endif
+
+#else /* not NEED_FAR_POINTERS */
+
+#define far_malloc(x)	malloc(x)
+#define far_free(x)	free(x)
+
+#endif /* NEED_FAR_POINTERS */
+
+#ifdef DONT_USE_B_MODE		/* define mode parameters for fopen() */
+#define READ_BINARY	"r"
+#else
+#define READ_BINARY	"rb"
+#endif
+
+#ifndef USE_MSDOS_MEMMGR	/* make sure user got configuration right */
+  You forgot to define USE_MSDOS_MEMMGR in jconfig.h. /* deliberate syntax error */
+#endif
+
+#if MAX_ALLOC_CHUNK >= 65535L	/* make sure jconfig.h got this right */
+  MAX_ALLOC_CHUNK should be less than 64K. /* deliberate syntax error */
+#endif
+
+
+/*
+ * Declarations for assembly-language support routines (see jmemdosa.asm).
+ *
+ * The functions are declared "far" as are all their pointer arguments;
+ * this ensures the assembly source code will work regardless of the
+ * compiler memory model.  We assume "short" is 16 bits, "long" is 32.
+ */
+
+typedef void far * XMSDRIVER;	/* actually a pointer to code */
+typedef struct {		/* registers for calling XMS driver */
+	unsigned short ax, dx, bx;
+	void far * ds_si;
+      } XMScontext;
+typedef struct {		/* registers for calling EMS driver */
+	unsigned short ax, dx, bx;
+	void far * ds_si;
+      } EMScontext;
+
+extern short far jdos_open JPP((short far * handle, char far * filename));
+extern short far jdos_close JPP((short handle));
+extern short far jdos_seek JPP((short handle, long offset));
+extern short far jdos_read JPP((short handle, void far * buffer,
+				unsigned short count));
+extern short far jdos_write JPP((short handle, void far * buffer,
+				 unsigned short count));
+extern void far jxms_getdriver JPP((XMSDRIVER far *));
+extern void far jxms_calldriver JPP((XMSDRIVER, XMScontext far *));
+extern short far jems_available JPP((void));
+extern void far jems_calldriver JPP((EMScontext far *));
+
+
+/*
+ * Selection of a file name for a temporary file.
+ * This is highly system-dependent, and you may want to customize it.
+ */
+
+static int next_file_num;	/* to distinguish among several temp files */
+
+LOCAL(void)
+select_file_name (char * fname)
+{
+  const char * env;
+  char * ptr;
+  FILE * tfile;
+
+  /* Keep generating file names till we find one that's not in use */
+  for (;;) {
+    /* Get temp directory name from environment TMP or TEMP variable;
+     * if none, use "."
+     */
+    if ((env = (const char *) getenv("TMP")) == NULL)
+      if ((env = (const char *) getenv("TEMP")) == NULL)
+	env = ".";
+    if (*env == '\0')		/* null string means "." */
+      env = ".";
+    ptr = fname;		/* copy name to fname */
+    while (*env != '\0')
+      *ptr++ = *env++;
+    if (ptr[-1] != '\\' && ptr[-1] != '/')
+      *ptr++ = '\\';		/* append backslash if not in env variable */
+    /* Append a suitable file name */
+    next_file_num++;		/* advance counter */
+    sprintf(ptr, "JPG%03d.TMP", next_file_num);
+    /* Probe to see if file name is already in use */
+    if ((tfile = fopen(fname, READ_BINARY)) == NULL)
+      break;
+    fclose(tfile);		/* oops, it's there; close tfile & try again */
+  }
+}
+
+
+/*
+ * Near-memory allocation and freeing are controlled by the regular library
+ * routines malloc() and free().
+ */
+
+GLOBAL(void *)
+jpeg_get_small (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_small (j_common_ptr cinfo, void * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * "Large" objects are allocated in far memory, if possible
+ */
+
+GLOBAL(void FAR *)
+jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void FAR *) far_malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_large (j_common_ptr cinfo, void FAR * object, size_t sizeofobject)
+{
+  far_free(object);
+}
+
+
+/*
+ * This routine computes the total memory space available for allocation.
+ * It's impossible to do this in a portable way; our current solution is
+ * to make the user tell us (with a default value set at compile time).
+ * If you can actually get the available space, it's a good idea to subtract
+ * a slop factor of 5% or so.
+ */
+
+#ifndef DEFAULT_MAX_MEM		/* so can override from makefile */
+#define DEFAULT_MAX_MEM		300000L /* for total usage about 450K */
+#endif
+
+GLOBAL(long)
+jpeg_mem_available (j_common_ptr cinfo, long min_bytes_needed,
+		    long max_bytes_needed, long already_allocated)
+{
+  return cinfo->mem->max_memory_to_use - already_allocated;
+}
+
+
+/*
+ * Backing store (temporary file) management.
+ * Backing store objects are only used when the value returned by
+ * jpeg_mem_available is less than the total space needed.  You can dispense
+ * with these routines if you have plenty of virtual memory; see jmemnobs.c.
+ */
+
+/*
+ * For MS-DOS we support three types of backing storage:
+ *   1. Conventional DOS files.  We access these by direct DOS calls rather
+ *      than via the stdio package.  This provides a bit better performance,
+ *      but the real reason is that the buffers to be read or written are FAR.
+ *      The stdio library for small-data memory models can't cope with that.
+ *   2. Extended memory, accessed per the XMS V2.0 specification.
+ *   3. Expanded memory, accessed per the LIM/EMS 4.0 specification.
+ * You'll need copies of those specs to make sense of the related code.
+ * The specs are available by Internet FTP from the SIMTEL archives 
+ * (oak.oakland.edu and its various mirror sites).  See files
+ * pub/msdos/microsoft/xms20.arc and pub/msdos/info/limems41.zip.
+ */
+
+
+/*
+ * Access methods for a DOS file.
+ */
+
+
+METHODDEF(void)
+read_file_store (j_common_ptr cinfo, backing_store_ptr info,
+		 void FAR * buffer_address,
+		 long file_offset, long byte_count)
+{
+  if (jdos_seek(info->handle.file_handle, file_offset))
+    ERREXIT(cinfo, JERR_TFILE_SEEK);
+  /* Since MAX_ALLOC_CHUNK is less than 64K, byte_count will be too. */
+  if (byte_count > 65535L)	/* safety check */
+    ERREXIT(cinfo, JERR_BAD_ALLOC_CHUNK);
+  if (jdos_read(info->handle.file_handle, buffer_address,
+		(unsigned short) byte_count))
+    ERREXIT(cinfo, JERR_TFILE_READ);
+}
+
+
+METHODDEF(void)
+write_file_store (j_common_ptr cinfo, backing_store_ptr info,
+		  void FAR * buffer_address,
+		  long file_offset, long byte_count)
+{
+  if (jdos_seek(info->handle.file_handle, file_offset))
+    ERREXIT(cinfo, JERR_TFILE_SEEK);
+  /* Since MAX_ALLOC_CHUNK is less than 64K, byte_count will be too. */
+  if (byte_count > 65535L)	/* safety check */
+    ERREXIT(cinfo, JERR_BAD_ALLOC_CHUNK);
+  if (jdos_write(info->handle.file_handle, buffer_address,
+		 (unsigned short) byte_count))
+    ERREXIT(cinfo, JERR_TFILE_WRITE);
+}
+
+
+METHODDEF(void)
+close_file_store (j_common_ptr cinfo, backing_store_ptr info)
+{
+  jdos_close(info->handle.file_handle);	/* close the file */
+  remove(info->temp_name);	/* delete the file */
+/* If your system doesn't have remove(), try unlink() instead.
+ * remove() is the ANSI-standard name for this function, but
+ * unlink() was more common in pre-ANSI systems.
+ */
+  TRACEMSS(cinfo, 1, JTRC_TFILE_CLOSE, info->temp_name);
+}
+
+
+LOCAL(boolean)
+open_file_store (j_common_ptr cinfo, backing_store_ptr info,
+		 long total_bytes_needed)
+{
+  short handle;
+
+  select_file_name(info->temp_name);
+  if (jdos_open((short far *) & handle, (char far *) info->temp_name)) {
+    /* might as well exit since jpeg_open_backing_store will fail anyway */
+    ERREXITS(cinfo, JERR_TFILE_CREATE, info->temp_name);
+    return FALSE;
+  }
+  info->handle.file_handle = handle;
+  info->read_backing_store = read_file_store;
+  info->write_backing_store = write_file_store;
+  info->close_backing_store = close_file_store;
+  TRACEMSS(cinfo, 1, JTRC_TFILE_OPEN, info->temp_name);
+  return TRUE;			/* succeeded */
+}
+
+
+/*
+ * Access methods for extended memory.
+ */
+
+#if XMS_SUPPORTED
+
+static XMSDRIVER xms_driver;	/* saved address of XMS driver */
+
+typedef union {			/* either long offset or real-mode pointer */
+	long offset;
+	void far * ptr;
+      } XMSPTR;
+
+typedef struct {		/* XMS move specification structure */
+	long length;
+	XMSH src_handle;
+	XMSPTR src;
+	XMSH dst_handle;
+	XMSPTR dst;
+      } XMSspec;
+
+#define ODD(X)	(((X) & 1L) != 0)
+
+
+METHODDEF(void)
+read_xms_store (j_common_ptr cinfo, backing_store_ptr info,
+		void FAR * buffer_address,
+		long file_offset, long byte_count)
+{
+  XMScontext ctx;
+  XMSspec spec;
+  char endbuffer[2];
+
+  /* The XMS driver can't cope with an odd length, so handle the last byte
+   * specially if byte_count is odd.  We don't expect this to be common.
+   */
+
+  spec.length = byte_count & (~ 1L);
+  spec.src_handle = info->handle.xms_handle;
+  spec.src.offset = file_offset;
+  spec.dst_handle = 0;
+  spec.dst.ptr = buffer_address;
+  
+  ctx.ds_si = (void far *) & spec;
+  ctx.ax = 0x0b00;		/* EMB move */
+  jxms_calldriver(xms_driver, (XMScontext far *) & ctx);
+  if (ctx.ax != 1)
+    ERREXIT(cinfo, JERR_XMS_READ);
+
+  if (ODD(byte_count)) {
+    read_xms_store(cinfo, info, (void FAR *) endbuffer,
+		   file_offset + byte_count - 1L, 2L);
+    ((char FAR *) buffer_address)[byte_count - 1L] = endbuffer[0];
+  }
+}
+
+
+METHODDEF(void)
+write_xms_store (j_common_ptr cinfo, backing_store_ptr info,
+		 void FAR * buffer_address,
+		 long file_offset, long byte_count)
+{
+  XMScontext ctx;
+  XMSspec spec;
+  char endbuffer[2];
+
+  /* The XMS driver can't cope with an odd length, so handle the last byte
+   * specially if byte_count is odd.  We don't expect this to be common.
+   */
+
+  spec.length = byte_count & (~ 1L);
+  spec.src_handle = 0;
+  spec.src.ptr = buffer_address;
+  spec.dst_handle = info->handle.xms_handle;
+  spec.dst.offset = file_offset;
+
+  ctx.ds_si = (void far *) & spec;
+  ctx.ax = 0x0b00;		/* EMB move */
+  jxms_calldriver(xms_driver, (XMScontext far *) & ctx);
+  if (ctx.ax != 1)
+    ERREXIT(cinfo, JERR_XMS_WRITE);
+
+  if (ODD(byte_count)) {
+    read_xms_store(cinfo, info, (void FAR *) endbuffer,
+		   file_offset + byte_count - 1L, 2L);
+    endbuffer[0] = ((char FAR *) buffer_address)[byte_count - 1L];
+    write_xms_store(cinfo, info, (void FAR *) endbuffer,
+		    file_offset + byte_count - 1L, 2L);
+  }
+}
+
+
+METHODDEF(void)
+close_xms_store (j_common_ptr cinfo, backing_store_ptr info)
+{
+  XMScontext ctx;
+
+  ctx.dx = info->handle.xms_handle;
+  ctx.ax = 0x0a00;
+  jxms_calldriver(xms_driver, (XMScontext far *) & ctx);
+  TRACEMS1(cinfo, 1, JTRC_XMS_CLOSE, info->handle.xms_handle);
+  /* we ignore any error return from the driver */
+}
+
+
+LOCAL(boolean)
+open_xms_store (j_common_ptr cinfo, backing_store_ptr info,
+		long total_bytes_needed)
+{
+  XMScontext ctx;
+
+  /* Get address of XMS driver */
+  jxms_getdriver((XMSDRIVER far *) & xms_driver);
+  if (xms_driver == NULL)
+    return FALSE;		/* no driver to be had */
+
+  /* Get version number, must be >= 2.00 */
+  ctx.ax = 0x0000;
+  jxms_calldriver(xms_driver, (XMScontext far *) & ctx);
+  if (ctx.ax < (unsigned short) 0x0200)
+    return FALSE;
+
+  /* Try to get space (expressed in kilobytes) */
+  ctx.dx = (unsigned short) ((total_bytes_needed + 1023L) >> 10);
+  ctx.ax = 0x0900;
+  jxms_calldriver(xms_driver, (XMScontext far *) & ctx);
+  if (ctx.ax != 1)
+    return FALSE;
+
+  /* Succeeded, save the handle and away we go */
+  info->handle.xms_handle = ctx.dx;
+  info->read_backing_store = read_xms_store;
+  info->write_backing_store = write_xms_store;
+  info->close_backing_store = close_xms_store;
+  TRACEMS1(cinfo, 1, JTRC_XMS_OPEN, ctx.dx);
+  return TRUE;			/* succeeded */
+}
+
+#endif /* XMS_SUPPORTED */
+
+
+/*
+ * Access methods for expanded memory.
+ */
+
+#if EMS_SUPPORTED
+
+/* The EMS move specification structure requires word and long fields aligned
+ * at odd byte boundaries.  Some compilers will align struct fields at even
+ * byte boundaries.  While it's usually possible to force byte alignment,
+ * that causes an overall performance penalty and may pose problems in merging
+ * JPEG into a larger application.  Instead we accept some rather dirty code
+ * here.  Note this code would fail if the hardware did not allow odd-byte
+ * word & long accesses, but all 80x86 CPUs do.
+ */
+
+typedef void far * EMSPTR;
+
+typedef union {			/* EMS move specification structure */
+	long length;		/* It's easy to access first 4 bytes */
+	char bytes[18];		/* Misaligned fields in here! */
+      } EMSspec;
+
+/* Macros for accessing misaligned fields */
+#define FIELD_AT(spec,offset,type)  (*((type *) &(spec.bytes[offset])))
+#define SRC_TYPE(spec)		FIELD_AT(spec,4,char)
+#define SRC_HANDLE(spec)	FIELD_AT(spec,5,EMSH)
+#define SRC_OFFSET(spec)	FIELD_AT(spec,7,unsigned short)
+#define SRC_PAGE(spec)		FIELD_AT(spec,9,unsigned short)
+#define SRC_PTR(spec)		FIELD_AT(spec,7,EMSPTR)
+#define DST_TYPE(spec)		FIELD_AT(spec,11,char)
+#define DST_HANDLE(spec)	FIELD_AT(spec,12,EMSH)
+#define DST_OFFSET(spec)	FIELD_AT(spec,14,unsigned short)
+#define DST_PAGE(spec)		FIELD_AT(spec,16,unsigned short)
+#define DST_PTR(spec)		FIELD_AT(spec,14,EMSPTR)
+
+#define EMSPAGESIZE	16384L	/* gospel, see the EMS specs */
+
+#define HIBYTE(W)  (((W) >> 8) & 0xFF)
+#define LOBYTE(W)  ((W) & 0xFF)
+
+
+METHODDEF(void)
+read_ems_store (j_common_ptr cinfo, backing_store_ptr info,
+		void FAR * buffer_address,
+		long file_offset, long byte_count)
+{
+  EMScontext ctx;
+  EMSspec spec;
+
+  spec.length = byte_count;
+  SRC_TYPE(spec) = 1;
+  SRC_HANDLE(spec) = info->handle.ems_handle;
+  SRC_PAGE(spec)   = (unsigned short) (file_offset / EMSPAGESIZE);
+  SRC_OFFSET(spec) = (unsigned short) (file_offset % EMSPAGESIZE);
+  DST_TYPE(spec) = 0;
+  DST_HANDLE(spec) = 0;
+  DST_PTR(spec)    = buffer_address;
+  
+  ctx.ds_si = (void far *) & spec;
+  ctx.ax = 0x5700;		/* move memory region */
+  jems_calldriver((EMScontext far *) & ctx);
+  if (HIBYTE(ctx.ax) != 0)
+    ERREXIT(cinfo, JERR_EMS_READ);
+}
+
+
+METHODDEF(void)
+write_ems_store (j_common_ptr cinfo, backing_store_ptr info,
+		 void FAR * buffer_address,
+		 long file_offset, long byte_count)
+{
+  EMScontext ctx;
+  EMSspec spec;
+
+  spec.length = byte_count;
+  SRC_TYPE(spec) = 0;
+  SRC_HANDLE(spec) = 0;
+  SRC_PTR(spec)    = buffer_address;
+  DST_TYPE(spec) = 1;
+  DST_HANDLE(spec) = info->handle.ems_handle;
+  DST_PAGE(spec)   = (unsigned short) (file_offset / EMSPAGESIZE);
+  DST_OFFSET(spec) = (unsigned short) (file_offset % EMSPAGESIZE);
+  
+  ctx.ds_si = (void far *) & spec;
+  ctx.ax = 0x5700;		/* move memory region */
+  jems_calldriver((EMScontext far *) & ctx);
+  if (HIBYTE(ctx.ax) != 0)
+    ERREXIT(cinfo, JERR_EMS_WRITE);
+}
+
+
+METHODDEF(void)
+close_ems_store (j_common_ptr cinfo, backing_store_ptr info)
+{
+  EMScontext ctx;
+
+  ctx.ax = 0x4500;
+  ctx.dx = info->handle.ems_handle;
+  jems_calldriver((EMScontext far *) & ctx);
+  TRACEMS1(cinfo, 1, JTRC_EMS_CLOSE, info->handle.ems_handle);
+  /* we ignore any error return from the driver */
+}
+
+
+LOCAL(boolean)
+open_ems_store (j_common_ptr cinfo, backing_store_ptr info,
+		long total_bytes_needed)
+{
+  EMScontext ctx;
+
+  /* Is EMS driver there? */
+  if (! jems_available())
+    return FALSE;
+
+  /* Get status, make sure EMS is OK */
+  ctx.ax = 0x4000;
+  jems_calldriver((EMScontext far *) & ctx);
+  if (HIBYTE(ctx.ax) != 0)
+    return FALSE;
+
+  /* Get version, must be >= 4.0 */
+  ctx.ax = 0x4600;
+  jems_calldriver((EMScontext far *) & ctx);
+  if (HIBYTE(ctx.ax) != 0 || LOBYTE(ctx.ax) < 0x40)
+    return FALSE;
+
+  /* Try to allocate requested space */
+  ctx.ax = 0x4300;
+  ctx.bx = (unsigned short) ((total_bytes_needed + EMSPAGESIZE-1L) / EMSPAGESIZE);
+  jems_calldriver((EMScontext far *) & ctx);
+  if (HIBYTE(ctx.ax) != 0)
+    return FALSE;
+
+  /* Succeeded, save the handle and away we go */
+  info->handle.ems_handle = ctx.dx;
+  info->read_backing_store = read_ems_store;
+  info->write_backing_store = write_ems_store;
+  info->close_backing_store = close_ems_store;
+  TRACEMS1(cinfo, 1, JTRC_EMS_OPEN, ctx.dx);
+  return TRUE;			/* succeeded */
+}
+
+#endif /* EMS_SUPPORTED */
+
+
+/*
+ * Initial opening of a backing-store object.
+ */
+
+GLOBAL(void)
+jpeg_open_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+			 long total_bytes_needed)
+{
+  /* Try extended memory, then expanded memory, then regular file. */
+#if XMS_SUPPORTED
+  if (open_xms_store(cinfo, info, total_bytes_needed))
+    return;
+#endif
+#if EMS_SUPPORTED
+  if (open_ems_store(cinfo, info, total_bytes_needed))
+    return;
+#endif
+  if (open_file_store(cinfo, info, total_bytes_needed))
+    return;
+  ERREXITS(cinfo, JERR_TFILE_CREATE, "");
+}
+
+
+/*
+ * These routines take care of any system-dependent initialization and
+ * cleanup required.
+ */
+
+GLOBAL(long)
+jpeg_mem_init (j_common_ptr cinfo)
+{
+  next_file_num = 0;		/* initialize temp file name generator */
+  return DEFAULT_MAX_MEM;	/* default for max_memory_to_use */
+}
+
+GLOBAL(void)
+jpeg_mem_term (j_common_ptr cinfo)
+{
+  /* Microsoft C, at least in v6.00A, will not successfully reclaim freed
+   * blocks of size > 32Kbytes unless we give it a kick in the rear, like so:
+   */
+#ifdef NEED_FHEAPMIN
+  _fheapmin();
+#endif
+}
diff --git a/libjpegtwrp/jmemdosa.asm b/libjpegtwrp/jmemdosa.asm
new file mode 100644
index 000000000..ecd43729f
--- /dev/null
+++ b/libjpegtwrp/jmemdosa.asm
@@ -0,0 +1,379 @@
+;
+; jmemdosa.asm
+;
+; Copyright (C) 1992, Thomas G. Lane.
+; This file is part of the Independent JPEG Group's software.
+; For conditions of distribution and use, see the accompanying README file.
+;
+; This file contains low-level interface routines to support the MS-DOS
+; backing store manager (jmemdos.c).  Routines are provided to access disk
+; files through direct DOS calls, and to access XMS and EMS drivers.
+;
+; This file should assemble with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).  If you haven't got
+; a compatible assembler, better fall back to jmemansi.c or jmemname.c.
+;
+; To minimize dependence on the C compiler's register usage conventions,
+; we save and restore all 8086 registers, even though most compilers only
+; require SI,DI,DS to be preserved.  Also, we use only 16-bit-wide return
+; values, which everybody returns in AX.
+;
+; Based on code contributed by Ge' Weijers.
+;
+
+JMEMDOSA_TXT	segment byte public 'CODE'
+
+		assume	cs:JMEMDOSA_TXT
+
+		public	_jdos_open
+		public	_jdos_close
+		public	_jdos_seek
+		public	_jdos_read
+		public	_jdos_write
+		public	_jxms_getdriver
+		public	_jxms_calldriver
+		public	_jems_available
+		public	_jems_calldriver
+
+;
+; short far jdos_open (short far * handle, char far * filename)
+;
+; Create and open a temporary file
+;
+_jdos_open	proc	far
+		push	bp			; linkage
+		mov 	bp,sp
+		push	si			; save all registers for safety
+		push	di
+		push	bx
+		push	cx
+		push	dx
+		push	es
+		push	ds
+		mov	cx,0			; normal file attributes
+		lds	dx,dword ptr [bp+10]	; get filename pointer
+		mov	ah,3ch			; create file
+		int	21h
+		jc	open_err		; if failed, return error code
+		lds	bx,dword ptr [bp+6]	; get handle pointer
+		mov	word ptr [bx],ax	; save the handle
+		xor	ax,ax			; return zero for OK
+open_err:	pop	ds			; restore registers and exit
+		pop	es
+		pop	dx
+		pop	cx
+		pop	bx
+		pop	di
+		pop	si
+		pop 	bp
+		ret
+_jdos_open	endp
+
+
+;
+; short far jdos_close (short handle)
+;
+; Close the file handle
+;
+_jdos_close	proc	far
+		push	bp			; linkage
+		mov 	bp,sp
+		push	si			; save all registers for safety
+		push	di
+		push	bx
+		push	cx
+		push	dx
+		push	es
+		push	ds
+		mov	bx,word ptr [bp+6]	; file handle
+		mov	ah,3eh			; close file
+		int	21h
+		jc	close_err		; if failed, return error code
+		xor	ax,ax			; return zero for OK
+close_err:	pop	ds			; restore registers and exit
+		pop	es
+		pop	dx
+		pop	cx
+		pop	bx
+		pop	di
+		pop	si
+		pop 	bp
+		ret
+_jdos_close	endp
+
+
+;
+; short far jdos_seek (short handle, long offset)
+;
+; Set file position
+;
+_jdos_seek	proc	far
+		push	bp			; linkage
+		mov 	bp,sp
+		push	si			; save all registers for safety
+		push	di
+		push	bx
+		push	cx
+		push	dx
+		push	es
+		push	ds
+		mov	bx,word ptr [bp+6]	; file handle
+		mov	dx,word ptr [bp+8]	; LS offset
+		mov	cx,word ptr [bp+10]	; MS offset
+		mov	ax,4200h		; absolute seek
+		int	21h
+		jc	seek_err		; if failed, return error code
+		xor	ax,ax			; return zero for OK
+seek_err:	pop	ds			; restore registers and exit
+		pop	es
+		pop	dx
+		pop	cx
+		pop	bx
+		pop	di
+		pop	si
+		pop 	bp
+		ret
+_jdos_seek	endp
+
+
+;
+; short far jdos_read (short handle, void far * buffer, unsigned short count)
+;
+; Read from file
+;
+_jdos_read	proc	far
+		push	bp			; linkage
+		mov 	bp,sp
+		push	si			; save all registers for safety
+		push	di
+		push	bx
+		push	cx
+		push	dx
+		push	es
+		push	ds
+		mov	bx,word ptr [bp+6]	; file handle
+		lds	dx,dword ptr [bp+8]	; buffer address
+		mov	cx,word ptr [bp+12]	; number of bytes
+		mov	ah,3fh			; read file
+		int	21h
+		jc	read_err		; if failed, return error code
+		cmp	ax,word ptr [bp+12]	; make sure all bytes were read
+		je	read_ok
+		mov	ax,1			; else return 1 for not OK
+		jmp	short read_err
+read_ok:	xor	ax,ax			; return zero for OK
+read_err:	pop	ds			; restore registers and exit
+		pop	es
+		pop	dx
+		pop	cx
+		pop	bx
+		pop	di
+		pop	si
+		pop 	bp
+		ret
+_jdos_read	endp
+
+
+;
+; short far jdos_write (short handle, void far * buffer, unsigned short count)
+;
+; Write to file
+;
+_jdos_write	proc	far
+		push	bp			; linkage
+		mov 	bp,sp
+		push	si			; save all registers for safety
+		push	di
+		push	bx
+		push	cx
+		push	dx
+		push	es
+		push	ds
+		mov	bx,word ptr [bp+6]	; file handle
+		lds	dx,dword ptr [bp+8]	; buffer address
+		mov	cx,word ptr [bp+12]	; number of bytes
+		mov	ah,40h			; write file
+		int	21h
+		jc	write_err		; if failed, return error code
+		cmp	ax,word ptr [bp+12]	; make sure all bytes written
+		je	write_ok
+		mov	ax,1			; else return 1 for not OK
+		jmp	short write_err
+write_ok:	xor	ax,ax			; return zero for OK
+write_err:	pop	ds			; restore registers and exit
+		pop	es
+		pop	dx
+		pop	cx
+		pop	bx
+		pop	di
+		pop	si
+		pop 	bp
+		ret
+_jdos_write	endp
+
+
+;
+; void far jxms_getdriver (XMSDRIVER far *)
+;
+; Get the address of the XMS driver, or NULL if not available
+;
+_jxms_getdriver	proc	far
+		push	bp			; linkage
+		mov 	bp,sp
+		push	si			; save all registers for safety
+		push	di
+		push	bx
+		push	cx
+		push	dx
+		push	es
+		push	ds
+		mov 	ax,4300h		; call multiplex interrupt with
+		int	2fh			; a magic cookie, hex 4300
+		cmp 	al,80h			; AL should contain hex 80
+		je	xmsavail
+		xor 	dx,dx			; no XMS driver available
+		xor 	ax,ax			; return a nil pointer
+		jmp	short xmsavail_done
+xmsavail:	mov 	ax,4310h		; fetch driver address with
+		int	2fh			; another magic cookie
+		mov 	dx,es			; copy address to dx:ax
+		mov 	ax,bx
+xmsavail_done:	les 	bx,dword ptr [bp+6]	; get pointer to return value
+		mov	word ptr es:[bx],ax
+		mov	word ptr es:[bx+2],dx
+		pop	ds			; restore registers and exit
+		pop	es
+		pop	dx
+		pop	cx
+		pop	bx
+		pop	di
+		pop	si
+		pop	bp
+		ret
+_jxms_getdriver	endp
+
+
+;
+; void far jxms_calldriver (XMSDRIVER, XMScontext far *)
+;
+; The XMScontext structure contains values for the AX,DX,BX,SI,DS registers.
+; These are loaded, the XMS call is performed, and the new values of the
+; AX,DX,BX registers are written back to the context structure.
+;
+_jxms_calldriver 	proc	far
+		push	bp			; linkage
+		mov 	bp,sp
+		push	si			; save all registers for safety
+		push	di
+		push	bx
+		push	cx
+		push	dx
+		push	es
+		push	ds
+		les 	bx,dword ptr [bp+10]	; get XMScontext pointer
+		mov 	ax,word ptr es:[bx]	; load registers
+		mov 	dx,word ptr es:[bx+2]
+		mov 	si,word ptr es:[bx+6]
+		mov 	ds,word ptr es:[bx+8]
+		mov 	bx,word ptr es:[bx+4]
+		call	dword ptr [bp+6]	; call the driver
+		mov	cx,bx			; save returned BX for a sec
+		les 	bx,dword ptr [bp+10]	; get XMScontext pointer
+		mov 	word ptr es:[bx],ax	; put back ax,dx,bx
+		mov 	word ptr es:[bx+2],dx
+		mov 	word ptr es:[bx+4],cx
+		pop	ds			; restore registers and exit
+		pop	es
+		pop	dx
+		pop	cx
+		pop	bx
+		pop	di
+		pop	si
+		pop 	bp
+		ret
+_jxms_calldriver 	endp
+
+
+;
+; short far jems_available (void)
+;
+; Have we got an EMS driver? (this comes straight from the EMS 4.0 specs)
+;
+_jems_available	proc	far
+		push	si			; save all registers for safety
+		push	di
+		push	bx
+		push	cx
+		push	dx
+		push	es
+		push	ds
+		mov	ax,3567h		; get interrupt vector 67h
+		int	21h
+		push	cs
+		pop	ds
+		mov	di,000ah		; check offs 10 in returned seg
+		lea	si,ASCII_device_name	; against literal string
+		mov	cx,8
+		cld
+		repe cmpsb
+		jne	no_ems
+		mov	ax,1			; match, it's there
+		jmp	short avail_done
+no_ems:		xor	ax,ax			; it's not there
+avail_done:	pop	ds			; restore registers and exit
+		pop	es
+		pop	dx
+		pop	cx
+		pop	bx
+		pop	di
+		pop	si
+		ret
+
+ASCII_device_name	db	"EMMXXXX0"
+
+_jems_available	endp
+
+
+;
+; void far jems_calldriver (EMScontext far *)
+;
+; The EMScontext structure contains values for the AX,DX,BX,SI,DS registers.
+; These are loaded, the EMS trap is performed, and the new values of the
+; AX,DX,BX registers are written back to the context structure.
+;
+_jems_calldriver	proc far
+		push	bp			; linkage
+		mov 	bp,sp
+		push	si			; save all registers for safety
+		push	di
+		push	bx
+		push	cx
+		push	dx
+		push	es
+		push	ds
+		les 	bx,dword ptr [bp+6]	; get EMScontext pointer
+		mov 	ax,word ptr es:[bx]	; load registers
+		mov 	dx,word ptr es:[bx+2]
+		mov 	si,word ptr es:[bx+6]
+		mov 	ds,word ptr es:[bx+8]
+		mov 	bx,word ptr es:[bx+4]
+		int	67h			; call the EMS driver
+		mov	cx,bx			; save returned BX for a sec
+		les 	bx,dword ptr [bp+6]	; get EMScontext pointer
+		mov 	word ptr es:[bx],ax	; put back ax,dx,bx
+		mov 	word ptr es:[bx+2],dx
+		mov 	word ptr es:[bx+4],cx
+		pop	ds			; restore registers and exit
+		pop	es
+		pop	dx
+		pop	cx
+		pop	bx
+		pop	di
+		pop	si
+		pop 	bp
+		ret
+_jems_calldriver	endp
+
+JMEMDOSA_TXT	ends
+
+		end
diff --git a/libjpegtwrp/jmemmac.c b/libjpegtwrp/jmemmac.c
new file mode 100644
index 000000000..106f9bea0
--- /dev/null
+++ b/libjpegtwrp/jmemmac.c
@@ -0,0 +1,289 @@
+/*
+ * jmemmac.c
+ *
+ * Copyright (C) 1992-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * jmemmac.c provides an Apple Macintosh implementation of the system-
+ * dependent portion of the JPEG memory manager.
+ *
+ * If you use jmemmac.c, then you must define USE_MAC_MEMMGR in the
+ * JPEG_INTERNALS part of jconfig.h.
+ *
+ * jmemmac.c uses the Macintosh toolbox routines NewPtr and DisposePtr
+ * instead of malloc and free.  It accurately determines the amount of
+ * memory available by using CompactMem.  Notice that if left to its
+ * own devices, this code can chew up all available space in the
+ * application's zone, with the exception of the rather small "slop"
+ * factor computed in jpeg_mem_available().  The application can ensure
+ * that more space is left over by reducing max_memory_to_use.
+ *
+ * Large images are swapped to disk using temporary files and System 7.0+'s
+ * temporary folder functionality.
+ *
+ * Note that jmemmac.c depends on two features of MacOS that were first
+ * introduced in System 7: FindFolder and the FSSpec-based calls.
+ * If your application uses jmemmac.c and is run under System 6 or earlier,
+ * and the jpeg library decides it needs a temporary file, it will abort,
+ * printing error messages about requiring System 7.  (If no temporary files
+ * are created, it will run fine.)
+ *
+ * If you want to use jmemmac.c in an application that might be used with
+ * System 6 or earlier, then you should remove dependencies on FindFolder
+ * and the FSSpec calls.  You will need to replace FindFolder with some
+ * other mechanism for finding a place to put temporary files, and you
+ * should replace the FSSpec calls with their HFS equivalents:
+ *
+ *     FSpDelete     ->  HDelete
+ *     FSpGetFInfo   ->  HGetFInfo
+ *     FSpCreate     ->  HCreate
+ *     FSpOpenDF     ->  HOpen      *** Note: not HOpenDF ***
+ *     FSMakeFSSpec  ->  (fill in spec by hand.)
+ *
+ * (Use HOpen instead of HOpenDF.  HOpen is just a glue-interface to PBHOpen,
+ * which is on all HFS macs.  HOpenDF is a System 7 addition which avoids the
+ * ages-old problem of names starting with a period.)
+ *
+ * Contributed by Sam Bushell (jsam@iagu.on.net) and
+ * Dan Gildor (gyld@in-touch.com).
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jmemsys.h"    /* import the system-dependent declarations */
+
+#ifndef USE_MAC_MEMMGR	/* make sure user got configuration right */
+  You forgot to define USE_MAC_MEMMGR in jconfig.h. /* deliberate syntax error */
+#endif
+
+#include <Memory.h>     /* we use the MacOS memory manager */
+#include <Files.h>      /* we use the MacOS File stuff */
+#include <Folders.h>    /* we use the MacOS HFS stuff */
+#include <Script.h>     /* for smSystemScript */
+#include <Gestalt.h>    /* we use Gestalt to test for specific functionality */
+
+#ifndef TEMP_FILE_NAME		/* can override from jconfig.h or Makefile */
+#define TEMP_FILE_NAME  "JPG%03d.TMP"
+#endif
+
+static int next_file_num;	/* to distinguish among several temp files */
+
+
+/*
+ * Memory allocation and freeing are controlled by the MacOS library
+ * routines NewPtr() and DisposePtr(), which allocate fixed-address
+ * storage.  Unfortunately, the IJG library isn't smart enough to cope
+ * with relocatable storage.
+ */
+
+GLOBAL(void *)
+jpeg_get_small (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void *) NewPtr(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_small (j_common_ptr cinfo, void * object, size_t sizeofobject)
+{
+  DisposePtr((Ptr) object);
+}
+
+
+/*
+ * "Large" objects are treated the same as "small" ones.
+ * NB: we include FAR keywords in the routine declarations simply for
+ * consistency with the rest of the IJG code; FAR should expand to empty
+ * on rational architectures like the Mac.
+ */
+
+GLOBAL(void FAR *)
+jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void FAR *) NewPtr(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_large (j_common_ptr cinfo, void FAR * object, size_t sizeofobject)
+{
+  DisposePtr((Ptr) object);
+}
+
+
+/*
+ * This routine computes the total memory space available for allocation.
+ */
+
+GLOBAL(long)
+jpeg_mem_available (j_common_ptr cinfo, long min_bytes_needed,
+		    long max_bytes_needed, long already_allocated)
+{
+  long limit = cinfo->mem->max_memory_to_use - already_allocated;
+  long slop, mem;
+
+  /* Don't ask for more than what application has told us we may use */
+  if (max_bytes_needed > limit && limit > 0)
+    max_bytes_needed = limit;
+  /* Find whether there's a big enough free block in the heap.
+   * CompactMem tries to create a contiguous block of the requested size,
+   * and then returns the size of the largest free block (which could be
+   * much more or much less than we asked for).
+   * We add some slop to ensure we don't use up all available memory.
+   */
+  slop = max_bytes_needed / 16 + 32768L;
+  mem = CompactMem(max_bytes_needed + slop) - slop;
+  if (mem < 0)
+    mem = 0;			/* sigh, couldn't even get the slop */
+  /* Don't take more than the application says we can have */
+  if (mem > limit && limit > 0)
+    mem = limit;
+  return mem;
+}
+
+
+/*
+ * Backing store (temporary file) management.
+ * Backing store objects are only used when the value returned by
+ * jpeg_mem_available is less than the total space needed.  You can dispense
+ * with these routines if you have plenty of virtual memory; see jmemnobs.c.
+ */
+
+
+METHODDEF(void)
+read_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+		    void FAR * buffer_address,
+		    long file_offset, long byte_count)
+{
+  long bytes = byte_count;
+  long retVal;
+
+  if ( SetFPos ( info->temp_file, fsFromStart, file_offset ) != noErr )
+    ERREXIT(cinfo, JERR_TFILE_SEEK);
+
+  retVal = FSRead ( info->temp_file, &bytes,
+		    (unsigned char *) buffer_address );
+  if ( retVal != noErr || bytes != byte_count )
+    ERREXIT(cinfo, JERR_TFILE_READ);
+}
+
+
+METHODDEF(void)
+write_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+		     void FAR * buffer_address,
+		     long file_offset, long byte_count)
+{
+  long bytes = byte_count;
+  long retVal;
+
+  if ( SetFPos ( info->temp_file, fsFromStart, file_offset ) != noErr )
+    ERREXIT(cinfo, JERR_TFILE_SEEK);
+
+  retVal = FSWrite ( info->temp_file, &bytes,
+		     (unsigned char *) buffer_address );
+  if ( retVal != noErr || bytes != byte_count )
+    ERREXIT(cinfo, JERR_TFILE_WRITE);
+}
+
+
+METHODDEF(void)
+close_backing_store (j_common_ptr cinfo, backing_store_ptr info)
+{
+  FSClose ( info->temp_file );
+  FSpDelete ( &(info->tempSpec) );
+}
+
+
+/*
+ * Initial opening of a backing-store object.
+ *
+ * This version uses FindFolder to find the Temporary Items folder,
+ * and puts the temporary file in there.
+ */
+
+GLOBAL(void)
+jpeg_open_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+			 long total_bytes_needed)
+{
+  short         tmpRef, vRefNum;
+  long          dirID;
+  FInfo         finderInfo;
+  FSSpec        theSpec;
+  Str255        fName;
+  OSErr         osErr;
+  long          gestaltResponse = 0;
+
+  /* Check that FSSpec calls are available. */
+  osErr = Gestalt( gestaltFSAttr, &gestaltResponse );
+  if ( ( osErr != noErr )
+       || !( gestaltResponse & (1<<gestaltHasFSSpecCalls) ) )
+    ERREXITS(cinfo, JERR_TFILE_CREATE, "- System 7.0 or later required");
+  /* TO DO: add a proper error message to jerror.h. */
+
+  /* Check that FindFolder is available. */
+  osErr = Gestalt( gestaltFindFolderAttr, &gestaltResponse );
+  if ( ( osErr != noErr )
+       || !( gestaltResponse & (1<<gestaltFindFolderPresent) ) )
+    ERREXITS(cinfo, JERR_TFILE_CREATE, "- System 7.0 or later required.");
+  /* TO DO: add a proper error message to jerror.h. */
+
+  osErr = FindFolder ( kOnSystemDisk, kTemporaryFolderType, kCreateFolder,
+                       &vRefNum, &dirID );
+  if ( osErr != noErr )
+    ERREXITS(cinfo, JERR_TFILE_CREATE, "- temporary items folder unavailable");
+  /* TO DO: Try putting the temp files somewhere else. */
+
+  /* Keep generating file names till we find one that's not in use */
+  for (;;) {
+    next_file_num++;		/* advance counter */
+
+    sprintf(info->temp_name, TEMP_FILE_NAME, next_file_num);
+    strcpy ( (Ptr)fName+1, info->temp_name );
+    *fName = strlen (info->temp_name);
+    osErr = FSMakeFSSpec ( vRefNum, dirID, fName, &theSpec );
+
+    if ( (osErr = FSpGetFInfo ( &theSpec, &finderInfo ) ) != noErr )
+      break;
+  }
+
+  osErr = FSpCreate ( &theSpec, '????', '????', smSystemScript );
+  if ( osErr != noErr )
+    ERREXITS(cinfo, JERR_TFILE_CREATE, info->temp_name);
+
+  osErr = FSpOpenDF ( &theSpec, fsRdWrPerm, &(info->temp_file) );
+  if ( osErr != noErr )
+    ERREXITS(cinfo, JERR_TFILE_CREATE, info->temp_name);
+
+  info->tempSpec = theSpec;
+
+  info->read_backing_store = read_backing_store;
+  info->write_backing_store = write_backing_store;
+  info->close_backing_store = close_backing_store;
+  TRACEMSS(cinfo, 1, JTRC_TFILE_OPEN, info->temp_name);
+}
+
+
+/*
+ * These routines take care of any system-dependent initialization and
+ * cleanup required.
+ */
+
+GLOBAL(long)
+jpeg_mem_init (j_common_ptr cinfo)
+{
+  next_file_num = 0;
+
+  /* max_memory_to_use will be initialized to FreeMem()'s result;
+   * the calling application might later reduce it, for example
+   * to leave room to invoke multiple JPEG objects.
+   * Note that FreeMem returns the total number of free bytes;
+   * it may not be possible to allocate a single block of this size.
+   */
+  return FreeMem();
+}
+
+GLOBAL(void)
+jpeg_mem_term (j_common_ptr cinfo)
+{
+  /* no work */
+}
diff --git a/libjpegtwrp/jmemmgr.c b/libjpegtwrp/jmemmgr.c
new file mode 100644
index 000000000..d801b322d
--- /dev/null
+++ b/libjpegtwrp/jmemmgr.c
@@ -0,0 +1,1118 @@
+/*
+ * jmemmgr.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the JPEG system-independent memory management
+ * routines.  This code is usable across a wide variety of machines; most
+ * of the system dependencies have been isolated in a separate file.
+ * The major functions provided here are:
+ *   * pool-based allocation and freeing of memory;
+ *   * policy decisions about how to divide available memory among the
+ *     virtual arrays;
+ *   * control logic for swapping virtual arrays between main memory and
+ *     backing storage.
+ * The separate system-dependent file provides the actual backing-storage
+ * access code, and it contains the policy decision about how much total
+ * main memory to use.
+ * This file is system-dependent in the sense that some of its functions
+ * are unnecessary in some systems.  For example, if there is enough virtual
+ * memory so that backing storage will never be used, much of the virtual
+ * array control logic could be removed.  (Of course, if you have that much
+ * memory then you shouldn't care about a little bit of unused code...)
+ */
+
+#define JPEG_INTERNALS
+#define AM_MEMORY_MANAGER	/* we define jvirt_Xarray_control structs */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jmemsys.h"		/* import the system-dependent declarations */
+
+#ifndef NO_GETENV
+#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare getenv() */
+extern char * getenv JPP((const char * name));
+#endif
+#endif
+
+
+/*
+ * Some important notes:
+ *   The allocation routines provided here must never return NULL.
+ *   They should exit to error_exit if unsuccessful.
+ *
+ *   It's not a good idea to try to merge the sarray and barray routines,
+ *   even though they are textually almost the same, because samples are
+ *   usually stored as bytes while coefficients are shorts or ints.  Thus,
+ *   in machines where byte pointers have a different representation from
+ *   word pointers, the resulting machine code could not be the same.
+ */
+
+
+/*
+ * Many machines require storage alignment: longs must start on 4-byte
+ * boundaries, doubles on 8-byte boundaries, etc.  On such machines, malloc()
+ * always returns pointers that are multiples of the worst-case alignment
+ * requirement, and we had better do so too.
+ * There isn't any really portable way to determine the worst-case alignment
+ * requirement.  This module assumes that the alignment requirement is
+ * multiples of sizeof(ALIGN_TYPE).
+ * By default, we define ALIGN_TYPE as double.  This is necessary on some
+ * workstations (where doubles really do need 8-byte alignment) and will work
+ * fine on nearly everything.  If your machine has lesser alignment needs,
+ * you can save a few bytes by making ALIGN_TYPE smaller.
+ * The only place I know of where this will NOT work is certain Macintosh
+ * 680x0 compilers that define double as a 10-byte IEEE extended float.
+ * Doing 10-byte alignment is counterproductive because longwords won't be
+ * aligned well.  Put "#define ALIGN_TYPE long" in jconfig.h if you have
+ * such a compiler.
+ */
+
+#ifndef ALIGN_TYPE		/* so can override from jconfig.h */
+#define ALIGN_TYPE  double
+#endif
+
+
+/*
+ * We allocate objects from "pools", where each pool is gotten with a single
+ * request to jpeg_get_small() or jpeg_get_large().  There is no per-object
+ * overhead within a pool, except for alignment padding.  Each pool has a
+ * header with a link to the next pool of the same class.
+ * Small and large pool headers are identical except that the latter's
+ * link pointer must be FAR on 80x86 machines.
+ * Notice that the "real" header fields are union'ed with a dummy ALIGN_TYPE
+ * field.  This forces the compiler to make SIZEOF(small_pool_hdr) a multiple
+ * of the alignment requirement of ALIGN_TYPE.
+ */
+
+typedef union small_pool_struct * small_pool_ptr;
+
+typedef union small_pool_struct {
+  struct {
+    small_pool_ptr next;	/* next in list of pools */
+    size_t bytes_used;		/* how many bytes already used within pool */
+    size_t bytes_left;		/* bytes still available in this pool */
+  } hdr;
+  ALIGN_TYPE dummy;		/* included in union to ensure alignment */
+} small_pool_hdr;
+
+typedef union large_pool_struct FAR * large_pool_ptr;
+
+typedef union large_pool_struct {
+  struct {
+    large_pool_ptr next;	/* next in list of pools */
+    size_t bytes_used;		/* how many bytes already used within pool */
+    size_t bytes_left;		/* bytes still available in this pool */
+  } hdr;
+  ALIGN_TYPE dummy;		/* included in union to ensure alignment */
+} large_pool_hdr;
+
+
+/*
+ * Here is the full definition of a memory manager object.
+ */
+
+typedef struct {
+  struct jpeg_memory_mgr pub;	/* public fields */
+
+  /* Each pool identifier (lifetime class) names a linked list of pools. */
+  small_pool_ptr small_list[JPOOL_NUMPOOLS];
+  large_pool_ptr large_list[JPOOL_NUMPOOLS];
+
+  /* Since we only have one lifetime class of virtual arrays, only one
+   * linked list is necessary (for each datatype).  Note that the virtual
+   * array control blocks being linked together are actually stored somewhere
+   * in the small-pool list.
+   */
+  jvirt_sarray_ptr virt_sarray_list;
+  jvirt_barray_ptr virt_barray_list;
+
+  /* This counts total space obtained from jpeg_get_small/large */
+  long total_space_allocated;
+
+  /* alloc_sarray and alloc_barray set this value for use by virtual
+   * array routines.
+   */
+  JDIMENSION last_rowsperchunk;	/* from most recent alloc_sarray/barray */
+} my_memory_mgr;
+
+typedef my_memory_mgr * my_mem_ptr;
+
+
+/*
+ * The control blocks for virtual arrays.
+ * Note that these blocks are allocated in the "small" pool area.
+ * System-dependent info for the associated backing store (if any) is hidden
+ * inside the backing_store_info struct.
+ */
+
+struct jvirt_sarray_control {
+  JSAMPARRAY mem_buffer;	/* => the in-memory buffer */
+  JDIMENSION rows_in_array;	/* total virtual array height */
+  JDIMENSION samplesperrow;	/* width of array (and of memory buffer) */
+  JDIMENSION maxaccess;		/* max rows accessed by access_virt_sarray */
+  JDIMENSION rows_in_mem;	/* height of memory buffer */
+  JDIMENSION rowsperchunk;	/* allocation chunk size in mem_buffer */
+  JDIMENSION cur_start_row;	/* first logical row # in the buffer */
+  JDIMENSION first_undef_row;	/* row # of first uninitialized row */
+  boolean pre_zero;		/* pre-zero mode requested? */
+  boolean dirty;		/* do current buffer contents need written? */
+  boolean b_s_open;		/* is backing-store data valid? */
+  jvirt_sarray_ptr next;	/* link to next virtual sarray control block */
+  backing_store_info b_s_info;	/* System-dependent control info */
+};
+
+struct jvirt_barray_control {
+  JBLOCKARRAY mem_buffer;	/* => the in-memory buffer */
+  JDIMENSION rows_in_array;	/* total virtual array height */
+  JDIMENSION blocksperrow;	/* width of array (and of memory buffer) */
+  JDIMENSION maxaccess;		/* max rows accessed by access_virt_barray */
+  JDIMENSION rows_in_mem;	/* height of memory buffer */
+  JDIMENSION rowsperchunk;	/* allocation chunk size in mem_buffer */
+  JDIMENSION cur_start_row;	/* first logical row # in the buffer */
+  JDIMENSION first_undef_row;	/* row # of first uninitialized row */
+  boolean pre_zero;		/* pre-zero mode requested? */
+  boolean dirty;		/* do current buffer contents need written? */
+  boolean b_s_open;		/* is backing-store data valid? */
+  jvirt_barray_ptr next;	/* link to next virtual barray control block */
+  backing_store_info b_s_info;	/* System-dependent control info */
+};
+
+
+#ifdef MEM_STATS		/* optional extra stuff for statistics */
+
+LOCAL(void)
+print_mem_stats (j_common_ptr cinfo, int pool_id)
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  small_pool_ptr shdr_ptr;
+  large_pool_ptr lhdr_ptr;
+
+  /* Since this is only a debugging stub, we can cheat a little by using
+   * fprintf directly rather than going through the trace message code.
+   * This is helpful because message parm array can't handle longs.
+   */
+  fprintf(stderr, "Freeing pool %d, total space = %ld\n",
+	  pool_id, mem->total_space_allocated);
+
+  for (lhdr_ptr = mem->large_list[pool_id]; lhdr_ptr != NULL;
+       lhdr_ptr = lhdr_ptr->hdr.next) {
+    fprintf(stderr, "  Large chunk used %ld\n",
+	    (long) lhdr_ptr->hdr.bytes_used);
+  }
+
+  for (shdr_ptr = mem->small_list[pool_id]; shdr_ptr != NULL;
+       shdr_ptr = shdr_ptr->hdr.next) {
+    fprintf(stderr, "  Small chunk used %ld free %ld\n",
+	    (long) shdr_ptr->hdr.bytes_used,
+	    (long) shdr_ptr->hdr.bytes_left);
+  }
+}
+
+#endif /* MEM_STATS */
+
+
+LOCAL(void)
+out_of_memory (j_common_ptr cinfo, int which)
+/* Report an out-of-memory error and stop execution */
+/* If we compiled MEM_STATS support, report alloc requests before dying */
+{
+#ifdef MEM_STATS
+  cinfo->err->trace_level = 2;	/* force self_destruct to report stats */
+#endif
+  ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, which);
+}
+
+
+/*
+ * Allocation of "small" objects.
+ *
+ * For these, we use pooled storage.  When a new pool must be created,
+ * we try to get enough space for the current request plus a "slop" factor,
+ * where the slop will be the amount of leftover space in the new pool.
+ * The speed vs. space tradeoff is largely determined by the slop values.
+ * A different slop value is provided for each pool class (lifetime),
+ * and we also distinguish the first pool of a class from later ones.
+ * NOTE: the values given work fairly well on both 16- and 32-bit-int
+ * machines, but may be too small if longs are 64 bits or more.
+ */
+
+static const size_t first_pool_slop[JPOOL_NUMPOOLS] = 
+{
+	1600,			/* first PERMANENT pool */
+	16000			/* first IMAGE pool */
+};
+
+static const size_t extra_pool_slop[JPOOL_NUMPOOLS] = 
+{
+	0,			/* additional PERMANENT pools */
+	5000			/* additional IMAGE pools */
+};
+
+#define MIN_SLOP  50		/* greater than 0 to avoid futile looping */
+
+
+METHODDEF(void *)
+alloc_small (j_common_ptr cinfo, int pool_id, size_t sizeofobject)
+/* Allocate a "small" object */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  small_pool_ptr hdr_ptr, prev_hdr_ptr;
+  char * data_ptr;
+  size_t odd_bytes, min_request, slop;
+
+  /* Check for unsatisfiable request (do now to ensure no overflow below) */
+  if (sizeofobject > (size_t) (MAX_ALLOC_CHUNK-SIZEOF(small_pool_hdr)))
+    out_of_memory(cinfo, 1);	/* request exceeds malloc's ability */
+
+  /* Round up the requested size to a multiple of SIZEOF(ALIGN_TYPE) */
+  odd_bytes = sizeofobject % SIZEOF(ALIGN_TYPE);
+  if (odd_bytes > 0)
+    sizeofobject += SIZEOF(ALIGN_TYPE) - odd_bytes;
+
+  /* See if space is available in any existing pool */
+  if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+  prev_hdr_ptr = NULL;
+  hdr_ptr = mem->small_list[pool_id];
+  while (hdr_ptr != NULL) {
+    if (hdr_ptr->hdr.bytes_left >= sizeofobject)
+      break;			/* found pool with enough space */
+    prev_hdr_ptr = hdr_ptr;
+    hdr_ptr = hdr_ptr->hdr.next;
+  }
+
+  /* Time to make a new pool? */
+  if (hdr_ptr == NULL) {
+    /* min_request is what we need now, slop is what will be leftover */
+    min_request = sizeofobject + SIZEOF(small_pool_hdr);
+    if (prev_hdr_ptr == NULL)	/* first pool in class? */
+      slop = first_pool_slop[pool_id];
+    else
+      slop = extra_pool_slop[pool_id];
+    /* Don't ask for more than MAX_ALLOC_CHUNK */
+    if (slop > (size_t) (MAX_ALLOC_CHUNK-min_request))
+      slop = (size_t) (MAX_ALLOC_CHUNK-min_request);
+    /* Try to get space, if fail reduce slop and try again */
+    for (;;) {
+      hdr_ptr = (small_pool_ptr) jpeg_get_small(cinfo, min_request + slop);
+      if (hdr_ptr != NULL)
+	break;
+      slop /= 2;
+      if (slop < MIN_SLOP)	/* give up when it gets real small */
+	out_of_memory(cinfo, 2); /* jpeg_get_small failed */
+    }
+    mem->total_space_allocated += min_request + slop;
+    /* Success, initialize the new pool header and add to end of list */
+    hdr_ptr->hdr.next = NULL;
+    hdr_ptr->hdr.bytes_used = 0;
+    hdr_ptr->hdr.bytes_left = sizeofobject + slop;
+    if (prev_hdr_ptr == NULL)	/* first pool in class? */
+      mem->small_list[pool_id] = hdr_ptr;
+    else
+      prev_hdr_ptr->hdr.next = hdr_ptr;
+  }
+
+  /* OK, allocate the object from the current pool */
+  data_ptr = (char *) (hdr_ptr + 1); /* point to first data byte in pool */
+  data_ptr += hdr_ptr->hdr.bytes_used; /* point to place for object */
+  hdr_ptr->hdr.bytes_used += sizeofobject;
+  hdr_ptr->hdr.bytes_left -= sizeofobject;
+
+  return (void *) data_ptr;
+}
+
+
+/*
+ * Allocation of "large" objects.
+ *
+ * The external semantics of these are the same as "small" objects,
+ * except that FAR pointers are used on 80x86.  However the pool
+ * management heuristics are quite different.  We assume that each
+ * request is large enough that it may as well be passed directly to
+ * jpeg_get_large; the pool management just links everything together
+ * so that we can free it all on demand.
+ * Note: the major use of "large" objects is in JSAMPARRAY and JBLOCKARRAY
+ * structures.  The routines that create these structures (see below)
+ * deliberately bunch rows together to ensure a large request size.
+ */
+
+METHODDEF(void FAR *)
+alloc_large (j_common_ptr cinfo, int pool_id, size_t sizeofobject)
+/* Allocate a "large" object */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  large_pool_ptr hdr_ptr;
+  size_t odd_bytes;
+
+  /* Check for unsatisfiable request (do now to ensure no overflow below) */
+  if (sizeofobject > (size_t) (MAX_ALLOC_CHUNK-SIZEOF(large_pool_hdr)))
+    out_of_memory(cinfo, 3);	/* request exceeds malloc's ability */
+
+  /* Round up the requested size to a multiple of SIZEOF(ALIGN_TYPE) */
+  odd_bytes = sizeofobject % SIZEOF(ALIGN_TYPE);
+  if (odd_bytes > 0)
+    sizeofobject += SIZEOF(ALIGN_TYPE) - odd_bytes;
+
+  /* Always make a new pool */
+  if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+
+  hdr_ptr = (large_pool_ptr) jpeg_get_large(cinfo, sizeofobject +
+					    SIZEOF(large_pool_hdr));
+  if (hdr_ptr == NULL)
+    out_of_memory(cinfo, 4);	/* jpeg_get_large failed */
+  mem->total_space_allocated += sizeofobject + SIZEOF(large_pool_hdr);
+
+  /* Success, initialize the new pool header and add to list */
+  hdr_ptr->hdr.next = mem->large_list[pool_id];
+  /* We maintain space counts in each pool header for statistical purposes,
+   * even though they are not needed for allocation.
+   */
+  hdr_ptr->hdr.bytes_used = sizeofobject;
+  hdr_ptr->hdr.bytes_left = 0;
+  mem->large_list[pool_id] = hdr_ptr;
+
+  return (void FAR *) (hdr_ptr + 1); /* point to first data byte in pool */
+}
+
+
+/*
+ * Creation of 2-D sample arrays.
+ * The pointers are in near heap, the samples themselves in FAR heap.
+ *
+ * To minimize allocation overhead and to allow I/O of large contiguous
+ * blocks, we allocate the sample rows in groups of as many rows as possible
+ * without exceeding MAX_ALLOC_CHUNK total bytes per allocation request.
+ * NB: the virtual array control routines, later in this file, know about
+ * this chunking of rows.  The rowsperchunk value is left in the mem manager
+ * object so that it can be saved away if this sarray is the workspace for
+ * a virtual array.
+ */
+
+METHODDEF(JSAMPARRAY)
+alloc_sarray (j_common_ptr cinfo, int pool_id,
+	      JDIMENSION samplesperrow, JDIMENSION numrows)
+/* Allocate a 2-D sample array */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  JSAMPARRAY result;
+  JSAMPROW workspace;
+  JDIMENSION rowsperchunk, currow, i;
+  long ltemp;
+
+  /* Calculate max # of rows allowed in one allocation chunk */
+  ltemp = (MAX_ALLOC_CHUNK-SIZEOF(large_pool_hdr)) /
+	  ((long) samplesperrow * SIZEOF(JSAMPLE));
+  if (ltemp <= 0)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+  if (ltemp < (long) numrows)
+    rowsperchunk = (JDIMENSION) ltemp;
+  else
+    rowsperchunk = numrows;
+  mem->last_rowsperchunk = rowsperchunk;
+
+  /* Get space for row pointers (small object) */
+  result = (JSAMPARRAY) alloc_small(cinfo, pool_id,
+				    (size_t) (numrows * SIZEOF(JSAMPROW)));
+
+  /* Get the rows themselves (large objects) */
+  currow = 0;
+  while (currow < numrows) {
+    rowsperchunk = MIN(rowsperchunk, numrows - currow);
+    workspace = (JSAMPROW) alloc_large(cinfo, pool_id,
+	(size_t) ((size_t) rowsperchunk * (size_t) samplesperrow
+		  * SIZEOF(JSAMPLE)));
+    for (i = rowsperchunk; i > 0; i--) {
+      result[currow++] = workspace;
+      workspace += samplesperrow;
+    }
+  }
+
+  return result;
+}
+
+
+/*
+ * Creation of 2-D coefficient-block arrays.
+ * This is essentially the same as the code for sample arrays, above.
+ */
+
+METHODDEF(JBLOCKARRAY)
+alloc_barray (j_common_ptr cinfo, int pool_id,
+	      JDIMENSION blocksperrow, JDIMENSION numrows)
+/* Allocate a 2-D coefficient-block array */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  JBLOCKARRAY result;
+  JBLOCKROW workspace;
+  JDIMENSION rowsperchunk, currow, i;
+  long ltemp;
+
+  /* Calculate max # of rows allowed in one allocation chunk */
+  ltemp = (MAX_ALLOC_CHUNK-SIZEOF(large_pool_hdr)) /
+	  ((long) blocksperrow * SIZEOF(JBLOCK));
+  if (ltemp <= 0)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+  if (ltemp < (long) numrows)
+    rowsperchunk = (JDIMENSION) ltemp;
+  else
+    rowsperchunk = numrows;
+  mem->last_rowsperchunk = rowsperchunk;
+
+  /* Get space for row pointers (small object) */
+  result = (JBLOCKARRAY) alloc_small(cinfo, pool_id,
+				     (size_t) (numrows * SIZEOF(JBLOCKROW)));
+
+  /* Get the rows themselves (large objects) */
+  currow = 0;
+  while (currow < numrows) {
+    rowsperchunk = MIN(rowsperchunk, numrows - currow);
+    workspace = (JBLOCKROW) alloc_large(cinfo, pool_id,
+	(size_t) ((size_t) rowsperchunk * (size_t) blocksperrow
+		  * SIZEOF(JBLOCK)));
+    for (i = rowsperchunk; i > 0; i--) {
+      result[currow++] = workspace;
+      workspace += blocksperrow;
+    }
+  }
+
+  return result;
+}
+
+
+/*
+ * About virtual array management:
+ *
+ * The above "normal" array routines are only used to allocate strip buffers
+ * (as wide as the image, but just a few rows high).  Full-image-sized buffers
+ * are handled as "virtual" arrays.  The array is still accessed a strip at a
+ * time, but the memory manager must save the whole array for repeated
+ * accesses.  The intended implementation is that there is a strip buffer in
+ * memory (as high as is possible given the desired memory limit), plus a
+ * backing file that holds the rest of the array.
+ *
+ * The request_virt_array routines are told the total size of the image and
+ * the maximum number of rows that will be accessed at once.  The in-memory
+ * buffer must be at least as large as the maxaccess value.
+ *
+ * The request routines create control blocks but not the in-memory buffers.
+ * That is postponed until realize_virt_arrays is called.  At that time the
+ * total amount of space needed is known (approximately, anyway), so free
+ * memory can be divided up fairly.
+ *
+ * The access_virt_array routines are responsible for making a specific strip
+ * area accessible (after reading or writing the backing file, if necessary).
+ * Note that the access routines are told whether the caller intends to modify
+ * the accessed strip; during a read-only pass this saves having to rewrite
+ * data to disk.  The access routines are also responsible for pre-zeroing
+ * any newly accessed rows, if pre-zeroing was requested.
+ *
+ * In current usage, the access requests are usually for nonoverlapping
+ * strips; that is, successive access start_row numbers differ by exactly
+ * num_rows = maxaccess.  This means we can get good performance with simple
+ * buffer dump/reload logic, by making the in-memory buffer be a multiple
+ * of the access height; then there will never be accesses across bufferload
+ * boundaries.  The code will still work with overlapping access requests,
+ * but it doesn't handle bufferload overlaps very efficiently.
+ */
+
+
+METHODDEF(jvirt_sarray_ptr)
+request_virt_sarray (j_common_ptr cinfo, int pool_id, boolean pre_zero,
+		     JDIMENSION samplesperrow, JDIMENSION numrows,
+		     JDIMENSION maxaccess)
+/* Request a virtual 2-D sample array */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  jvirt_sarray_ptr result;
+
+  /* Only IMAGE-lifetime virtual arrays are currently supported */
+  if (pool_id != JPOOL_IMAGE)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+
+  /* get control block */
+  result = (jvirt_sarray_ptr) alloc_small(cinfo, pool_id,
+					  SIZEOF(struct jvirt_sarray_control));
+
+  result->mem_buffer = NULL;	/* marks array not yet realized */
+  result->rows_in_array = numrows;
+  result->samplesperrow = samplesperrow;
+  result->maxaccess = maxaccess;
+  result->pre_zero = pre_zero;
+  result->b_s_open = FALSE;	/* no associated backing-store object */
+  result->next = mem->virt_sarray_list; /* add to list of virtual arrays */
+  mem->virt_sarray_list = result;
+
+  return result;
+}
+
+
+METHODDEF(jvirt_barray_ptr)
+request_virt_barray (j_common_ptr cinfo, int pool_id, boolean pre_zero,
+		     JDIMENSION blocksperrow, JDIMENSION numrows,
+		     JDIMENSION maxaccess)
+/* Request a virtual 2-D coefficient-block array */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  jvirt_barray_ptr result;
+
+  /* Only IMAGE-lifetime virtual arrays are currently supported */
+  if (pool_id != JPOOL_IMAGE)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+
+  /* get control block */
+  result = (jvirt_barray_ptr) alloc_small(cinfo, pool_id,
+					  SIZEOF(struct jvirt_barray_control));
+
+  result->mem_buffer = NULL;	/* marks array not yet realized */
+  result->rows_in_array = numrows;
+  result->blocksperrow = blocksperrow;
+  result->maxaccess = maxaccess;
+  result->pre_zero = pre_zero;
+  result->b_s_open = FALSE;	/* no associated backing-store object */
+  result->next = mem->virt_barray_list; /* add to list of virtual arrays */
+  mem->virt_barray_list = result;
+
+  return result;
+}
+
+
+METHODDEF(void)
+realize_virt_arrays (j_common_ptr cinfo)
+/* Allocate the in-memory buffers for any unrealized virtual arrays */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  long space_per_minheight, maximum_space, avail_mem;
+  long minheights, max_minheights;
+  jvirt_sarray_ptr sptr;
+  jvirt_barray_ptr bptr;
+
+  /* Compute the minimum space needed (maxaccess rows in each buffer)
+   * and the maximum space needed (full image height in each buffer).
+   * These may be of use to the system-dependent jpeg_mem_available routine.
+   */
+  space_per_minheight = 0;
+  maximum_space = 0;
+  for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
+    if (sptr->mem_buffer == NULL) { /* if not realized yet */
+      space_per_minheight += (long) sptr->maxaccess *
+			     (long) sptr->samplesperrow * SIZEOF(JSAMPLE);
+      maximum_space += (long) sptr->rows_in_array *
+		       (long) sptr->samplesperrow * SIZEOF(JSAMPLE);
+    }
+  }
+  for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
+    if (bptr->mem_buffer == NULL) { /* if not realized yet */
+      space_per_minheight += (long) bptr->maxaccess *
+			     (long) bptr->blocksperrow * SIZEOF(JBLOCK);
+      maximum_space += (long) bptr->rows_in_array *
+		       (long) bptr->blocksperrow * SIZEOF(JBLOCK);
+    }
+  }
+
+  if (space_per_minheight <= 0)
+    return;			/* no unrealized arrays, no work */
+
+  /* Determine amount of memory to actually use; this is system-dependent. */
+  avail_mem = jpeg_mem_available(cinfo, space_per_minheight, maximum_space,
+				 mem->total_space_allocated);
+
+  /* If the maximum space needed is available, make all the buffers full
+   * height; otherwise parcel it out with the same number of minheights
+   * in each buffer.
+   */
+  if (avail_mem >= maximum_space)
+    max_minheights = 1000000000L;
+  else {
+    max_minheights = avail_mem / space_per_minheight;
+    /* If there doesn't seem to be enough space, try to get the minimum
+     * anyway.  This allows a "stub" implementation of jpeg_mem_available().
+     */
+    if (max_minheights <= 0)
+      max_minheights = 1;
+  }
+
+  /* Allocate the in-memory buffers and initialize backing store as needed. */
+
+  for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
+    if (sptr->mem_buffer == NULL) { /* if not realized yet */
+      minheights = ((long) sptr->rows_in_array - 1L) / sptr->maxaccess + 1L;
+      if (minheights <= max_minheights) {
+	/* This buffer fits in memory */
+	sptr->rows_in_mem = sptr->rows_in_array;
+      } else {
+	/* It doesn't fit in memory, create backing store. */
+	sptr->rows_in_mem = (JDIMENSION) (max_minheights * sptr->maxaccess);
+	jpeg_open_backing_store(cinfo, & sptr->b_s_info,
+				(long) sptr->rows_in_array *
+				(long) sptr->samplesperrow *
+				(long) SIZEOF(JSAMPLE));
+	sptr->b_s_open = TRUE;
+      }
+      sptr->mem_buffer = alloc_sarray(cinfo, JPOOL_IMAGE,
+				      sptr->samplesperrow, sptr->rows_in_mem);
+      sptr->rowsperchunk = mem->last_rowsperchunk;
+      sptr->cur_start_row = 0;
+      sptr->first_undef_row = 0;
+      sptr->dirty = FALSE;
+    }
+  }
+
+  for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
+    if (bptr->mem_buffer == NULL) { /* if not realized yet */
+      minheights = ((long) bptr->rows_in_array - 1L) / bptr->maxaccess + 1L;
+      if (minheights <= max_minheights) {
+	/* This buffer fits in memory */
+	bptr->rows_in_mem = bptr->rows_in_array;
+      } else {
+	/* It doesn't fit in memory, create backing store. */
+	bptr->rows_in_mem = (JDIMENSION) (max_minheights * bptr->maxaccess);
+	jpeg_open_backing_store(cinfo, & bptr->b_s_info,
+				(long) bptr->rows_in_array *
+				(long) bptr->blocksperrow *
+				(long) SIZEOF(JBLOCK));
+	bptr->b_s_open = TRUE;
+      }
+      bptr->mem_buffer = alloc_barray(cinfo, JPOOL_IMAGE,
+				      bptr->blocksperrow, bptr->rows_in_mem);
+      bptr->rowsperchunk = mem->last_rowsperchunk;
+      bptr->cur_start_row = 0;
+      bptr->first_undef_row = 0;
+      bptr->dirty = FALSE;
+    }
+  }
+}
+
+
+LOCAL(void)
+do_sarray_io (j_common_ptr cinfo, jvirt_sarray_ptr ptr, boolean writing)
+/* Do backing store read or write of a virtual sample array */
+{
+  long bytesperrow, file_offset, byte_count, rows, thisrow, i;
+
+  bytesperrow = (long) ptr->samplesperrow * SIZEOF(JSAMPLE);
+  file_offset = ptr->cur_start_row * bytesperrow;
+  /* Loop to read or write each allocation chunk in mem_buffer */
+  for (i = 0; i < (long) ptr->rows_in_mem; i += ptr->rowsperchunk) {
+    /* One chunk, but check for short chunk at end of buffer */
+    rows = MIN((long) ptr->rowsperchunk, (long) ptr->rows_in_mem - i);
+    /* Transfer no more than is currently defined */
+    thisrow = (long) ptr->cur_start_row + i;
+    rows = MIN(rows, (long) ptr->first_undef_row - thisrow);
+    /* Transfer no more than fits in file */
+    rows = MIN(rows, (long) ptr->rows_in_array - thisrow);
+    if (rows <= 0)		/* this chunk might be past end of file! */
+      break;
+    byte_count = rows * bytesperrow;
+    if (writing)
+      (*ptr->b_s_info.write_backing_store) (cinfo, & ptr->b_s_info,
+					    (void FAR *) ptr->mem_buffer[i],
+					    file_offset, byte_count);
+    else
+      (*ptr->b_s_info.read_backing_store) (cinfo, & ptr->b_s_info,
+					   (void FAR *) ptr->mem_buffer[i],
+					   file_offset, byte_count);
+    file_offset += byte_count;
+  }
+}
+
+
+LOCAL(void)
+do_barray_io (j_common_ptr cinfo, jvirt_barray_ptr ptr, boolean writing)
+/* Do backing store read or write of a virtual coefficient-block array */
+{
+  long bytesperrow, file_offset, byte_count, rows, thisrow, i;
+
+  bytesperrow = (long) ptr->blocksperrow * SIZEOF(JBLOCK);
+  file_offset = ptr->cur_start_row * bytesperrow;
+  /* Loop to read or write each allocation chunk in mem_buffer */
+  for (i = 0; i < (long) ptr->rows_in_mem; i += ptr->rowsperchunk) {
+    /* One chunk, but check for short chunk at end of buffer */
+    rows = MIN((long) ptr->rowsperchunk, (long) ptr->rows_in_mem - i);
+    /* Transfer no more than is currently defined */
+    thisrow = (long) ptr->cur_start_row + i;
+    rows = MIN(rows, (long) ptr->first_undef_row - thisrow);
+    /* Transfer no more than fits in file */
+    rows = MIN(rows, (long) ptr->rows_in_array - thisrow);
+    if (rows <= 0)		/* this chunk might be past end of file! */
+      break;
+    byte_count = rows * bytesperrow;
+    if (writing)
+      (*ptr->b_s_info.write_backing_store) (cinfo, & ptr->b_s_info,
+					    (void FAR *) ptr->mem_buffer[i],
+					    file_offset, byte_count);
+    else
+      (*ptr->b_s_info.read_backing_store) (cinfo, & ptr->b_s_info,
+					   (void FAR *) ptr->mem_buffer[i],
+					   file_offset, byte_count);
+    file_offset += byte_count;
+  }
+}
+
+
+METHODDEF(JSAMPARRAY)
+access_virt_sarray (j_common_ptr cinfo, jvirt_sarray_ptr ptr,
+		    JDIMENSION start_row, JDIMENSION num_rows,
+		    boolean writable)
+/* Access the part of a virtual sample array starting at start_row */
+/* and extending for num_rows rows.  writable is true if  */
+/* caller intends to modify the accessed area. */
+{
+  JDIMENSION end_row = start_row + num_rows;
+  JDIMENSION undef_row;
+
+  /* debugging check */
+  if (end_row > ptr->rows_in_array || num_rows > ptr->maxaccess ||
+      ptr->mem_buffer == NULL)
+    ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+
+  /* Make the desired part of the virtual array accessible */
+  if (start_row < ptr->cur_start_row ||
+      end_row > ptr->cur_start_row+ptr->rows_in_mem) {
+    if (! ptr->b_s_open)
+      ERREXIT(cinfo, JERR_VIRTUAL_BUG);
+    /* Flush old buffer contents if necessary */
+    if (ptr->dirty) {
+      do_sarray_io(cinfo, ptr, TRUE);
+      ptr->dirty = FALSE;
+    }
+    /* Decide what part of virtual array to access.
+     * Algorithm: if target address > current window, assume forward scan,
+     * load starting at target address.  If target address < current window,
+     * assume backward scan, load so that target area is top of window.
+     * Note that when switching from forward write to forward read, will have
+     * start_row = 0, so the limiting case applies and we load from 0 anyway.
+     */
+    if (start_row > ptr->cur_start_row) {
+      ptr->cur_start_row = start_row;
+    } else {
+      /* use long arithmetic here to avoid overflow & unsigned problems */
+      long ltemp;
+
+      ltemp = (long) end_row - (long) ptr->rows_in_mem;
+      if (ltemp < 0)
+	ltemp = 0;		/* don't fall off front end of file */
+      ptr->cur_start_row = (JDIMENSION) ltemp;
+    }
+    /* Read in the selected part of the array.
+     * During the initial write pass, we will do no actual read
+     * because the selected part is all undefined.
+     */
+    do_sarray_io(cinfo, ptr, FALSE);
+  }
+  /* Ensure the accessed part of the array is defined; prezero if needed.
+   * To improve locality of access, we only prezero the part of the array
+   * that the caller is about to access, not the entire in-memory array.
+   */
+  if (ptr->first_undef_row < end_row) {
+    if (ptr->first_undef_row < start_row) {
+      if (writable)		/* writer skipped over a section of array */
+	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      undef_row = start_row;	/* but reader is allowed to read ahead */
+    } else {
+      undef_row = ptr->first_undef_row;
+    }
+    if (writable)
+      ptr->first_undef_row = end_row;
+    if (ptr->pre_zero) {
+      size_t bytesperrow = (size_t) ptr->samplesperrow * SIZEOF(JSAMPLE);
+      undef_row -= ptr->cur_start_row; /* make indexes relative to buffer */
+      end_row -= ptr->cur_start_row;
+      while (undef_row < end_row) {
+	jzero_far((void FAR *) ptr->mem_buffer[undef_row], bytesperrow);
+	undef_row++;
+      }
+    } else {
+      if (! writable)		/* reader looking at undefined data */
+	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+    }
+  }
+  /* Flag the buffer dirty if caller will write in it */
+  if (writable)
+    ptr->dirty = TRUE;
+  /* Return address of proper part of the buffer */
+  return ptr->mem_buffer + (start_row - ptr->cur_start_row);
+}
+
+
+METHODDEF(JBLOCKARRAY)
+access_virt_barray (j_common_ptr cinfo, jvirt_barray_ptr ptr,
+		    JDIMENSION start_row, JDIMENSION num_rows,
+		    boolean writable)
+/* Access the part of a virtual block array starting at start_row */
+/* and extending for num_rows rows.  writable is true if  */
+/* caller intends to modify the accessed area. */
+{
+  JDIMENSION end_row = start_row + num_rows;
+  JDIMENSION undef_row;
+
+  /* debugging check */
+  if (end_row > ptr->rows_in_array || num_rows > ptr->maxaccess ||
+      ptr->mem_buffer == NULL)
+    ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+
+  /* Make the desired part of the virtual array accessible */
+  if (start_row < ptr->cur_start_row ||
+      end_row > ptr->cur_start_row+ptr->rows_in_mem) {
+    if (! ptr->b_s_open)
+      ERREXIT(cinfo, JERR_VIRTUAL_BUG);
+    /* Flush old buffer contents if necessary */
+    if (ptr->dirty) {
+      do_barray_io(cinfo, ptr, TRUE);
+      ptr->dirty = FALSE;
+    }
+    /* Decide what part of virtual array to access.
+     * Algorithm: if target address > current window, assume forward scan,
+     * load starting at target address.  If target address < current window,
+     * assume backward scan, load so that target area is top of window.
+     * Note that when switching from forward write to forward read, will have
+     * start_row = 0, so the limiting case applies and we load from 0 anyway.
+     */
+    if (start_row > ptr->cur_start_row) {
+      ptr->cur_start_row = start_row;
+    } else {
+      /* use long arithmetic here to avoid overflow & unsigned problems */
+      long ltemp;
+
+      ltemp = (long) end_row - (long) ptr->rows_in_mem;
+      if (ltemp < 0)
+	ltemp = 0;		/* don't fall off front end of file */
+      ptr->cur_start_row = (JDIMENSION) ltemp;
+    }
+    /* Read in the selected part of the array.
+     * During the initial write pass, we will do no actual read
+     * because the selected part is all undefined.
+     */
+    do_barray_io(cinfo, ptr, FALSE);
+  }
+  /* Ensure the accessed part of the array is defined; prezero if needed.
+   * To improve locality of access, we only prezero the part of the array
+   * that the caller is about to access, not the entire in-memory array.
+   */
+  if (ptr->first_undef_row < end_row) {
+    if (ptr->first_undef_row < start_row) {
+      if (writable)		/* writer skipped over a section of array */
+	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      undef_row = start_row;	/* but reader is allowed to read ahead */
+    } else {
+      undef_row = ptr->first_undef_row;
+    }
+    if (writable)
+      ptr->first_undef_row = end_row;
+    if (ptr->pre_zero) {
+      size_t bytesperrow = (size_t) ptr->blocksperrow * SIZEOF(JBLOCK);
+      undef_row -= ptr->cur_start_row; /* make indexes relative to buffer */
+      end_row -= ptr->cur_start_row;
+      while (undef_row < end_row) {
+	jzero_far((void FAR *) ptr->mem_buffer[undef_row], bytesperrow);
+	undef_row++;
+      }
+    } else {
+      if (! writable)		/* reader looking at undefined data */
+	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+    }
+  }
+  /* Flag the buffer dirty if caller will write in it */
+  if (writable)
+    ptr->dirty = TRUE;
+  /* Return address of proper part of the buffer */
+  return ptr->mem_buffer + (start_row - ptr->cur_start_row);
+}
+
+
+/*
+ * Release all objects belonging to a specified pool.
+ */
+
+METHODDEF(void)
+free_pool (j_common_ptr cinfo, int pool_id)
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  small_pool_ptr shdr_ptr;
+  large_pool_ptr lhdr_ptr;
+  size_t space_freed;
+
+  if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+
+#ifdef MEM_STATS
+  if (cinfo->err->trace_level > 1)
+    print_mem_stats(cinfo, pool_id); /* print pool's memory usage statistics */
+#endif
+
+  /* If freeing IMAGE pool, close any virtual arrays first */
+  if (pool_id == JPOOL_IMAGE) {
+    jvirt_sarray_ptr sptr;
+    jvirt_barray_ptr bptr;
+
+    for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
+      if (sptr->b_s_open) {	/* there may be no backing store */
+	sptr->b_s_open = FALSE;	/* prevent recursive close if error */
+	(*sptr->b_s_info.close_backing_store) (cinfo, & sptr->b_s_info);
+      }
+    }
+    mem->virt_sarray_list = NULL;
+    for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
+      if (bptr->b_s_open) {	/* there may be no backing store */
+	bptr->b_s_open = FALSE;	/* prevent recursive close if error */
+	(*bptr->b_s_info.close_backing_store) (cinfo, & bptr->b_s_info);
+      }
+    }
+    mem->virt_barray_list = NULL;
+  }
+
+  /* Release large objects */
+  lhdr_ptr = mem->large_list[pool_id];
+  mem->large_list[pool_id] = NULL;
+
+  while (lhdr_ptr != NULL) {
+    large_pool_ptr next_lhdr_ptr = lhdr_ptr->hdr.next;
+    space_freed = lhdr_ptr->hdr.bytes_used +
+		  lhdr_ptr->hdr.bytes_left +
+		  SIZEOF(large_pool_hdr);
+    jpeg_free_large(cinfo, (void FAR *) lhdr_ptr, space_freed);
+    mem->total_space_allocated -= space_freed;
+    lhdr_ptr = next_lhdr_ptr;
+  }
+
+  /* Release small objects */
+  shdr_ptr = mem->small_list[pool_id];
+  mem->small_list[pool_id] = NULL;
+
+  while (shdr_ptr != NULL) {
+    small_pool_ptr next_shdr_ptr = shdr_ptr->hdr.next;
+    space_freed = shdr_ptr->hdr.bytes_used +
+		  shdr_ptr->hdr.bytes_left +
+		  SIZEOF(small_pool_hdr);
+    jpeg_free_small(cinfo, (void *) shdr_ptr, space_freed);
+    mem->total_space_allocated -= space_freed;
+    shdr_ptr = next_shdr_ptr;
+  }
+}
+
+
+/*
+ * Close up shop entirely.
+ * Note that this cannot be called unless cinfo->mem is non-NULL.
+ */
+
+METHODDEF(void)
+self_destruct (j_common_ptr cinfo)
+{
+  int pool;
+
+  /* Close all backing store, release all memory.
+   * Releasing pools in reverse order might help avoid fragmentation
+   * with some (brain-damaged) malloc libraries.
+   */
+  for (pool = JPOOL_NUMPOOLS-1; pool >= JPOOL_PERMANENT; pool--) {
+    free_pool(cinfo, pool);
+  }
+
+  /* Release the memory manager control block too. */
+  jpeg_free_small(cinfo, (void *) cinfo->mem, SIZEOF(my_memory_mgr));
+  cinfo->mem = NULL;		/* ensures I will be called only once */
+
+  jpeg_mem_term(cinfo);		/* system-dependent cleanup */
+}
+
+
+/*
+ * Memory manager initialization.
+ * When this is called, only the error manager pointer is valid in cinfo!
+ */
+
+GLOBAL(void)
+jinit_memory_mgr (j_common_ptr cinfo)
+{
+  my_mem_ptr mem;
+  long max_to_use;
+  int pool;
+  size_t test_mac;
+
+  cinfo->mem = NULL;		/* for safety if init fails */
+
+  /* Check for configuration errors.
+   * SIZEOF(ALIGN_TYPE) should be a power of 2; otherwise, it probably
+   * doesn't reflect any real hardware alignment requirement.
+   * The test is a little tricky: for X>0, X and X-1 have no one-bits
+   * in common if and only if X is a power of 2, ie has only one one-bit.
+   * Some compilers may give an "unreachable code" warning here; ignore it.
+   */
+  if ((SIZEOF(ALIGN_TYPE) & (SIZEOF(ALIGN_TYPE)-1)) != 0)
+    ERREXIT(cinfo, JERR_BAD_ALIGN_TYPE);
+  /* MAX_ALLOC_CHUNK must be representable as type size_t, and must be
+   * a multiple of SIZEOF(ALIGN_TYPE).
+   * Again, an "unreachable code" warning may be ignored here.
+   * But a "constant too large" warning means you need to fix MAX_ALLOC_CHUNK.
+   */
+  test_mac = (size_t) MAX_ALLOC_CHUNK;
+  if ((long) test_mac != MAX_ALLOC_CHUNK ||
+      (MAX_ALLOC_CHUNK % SIZEOF(ALIGN_TYPE)) != 0)
+    ERREXIT(cinfo, JERR_BAD_ALLOC_CHUNK);
+
+  max_to_use = jpeg_mem_init(cinfo); /* system-dependent initialization */
+
+  /* Attempt to allocate memory manager's control block */
+  mem = (my_mem_ptr) jpeg_get_small(cinfo, SIZEOF(my_memory_mgr));
+
+  if (mem == NULL) {
+    jpeg_mem_term(cinfo);	/* system-dependent cleanup */
+    ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 0);
+  }
+
+  /* OK, fill in the method pointers */
+  mem->pub.alloc_small = alloc_small;
+  mem->pub.alloc_large = alloc_large;
+  mem->pub.alloc_sarray = alloc_sarray;
+  mem->pub.alloc_barray = alloc_barray;
+  mem->pub.request_virt_sarray = request_virt_sarray;
+  mem->pub.request_virt_barray = request_virt_barray;
+  mem->pub.realize_virt_arrays = realize_virt_arrays;
+  mem->pub.access_virt_sarray = access_virt_sarray;
+  mem->pub.access_virt_barray = access_virt_barray;
+  mem->pub.free_pool = free_pool;
+  mem->pub.self_destruct = self_destruct;
+
+  /* Make MAX_ALLOC_CHUNK accessible to other modules */
+  mem->pub.max_alloc_chunk = MAX_ALLOC_CHUNK;
+
+  /* Initialize working state */
+  mem->pub.max_memory_to_use = max_to_use;
+
+  for (pool = JPOOL_NUMPOOLS-1; pool >= JPOOL_PERMANENT; pool--) {
+    mem->small_list[pool] = NULL;
+    mem->large_list[pool] = NULL;
+  }
+  mem->virt_sarray_list = NULL;
+  mem->virt_barray_list = NULL;
+
+  mem->total_space_allocated = SIZEOF(my_memory_mgr);
+
+  /* Declare ourselves open for business */
+  cinfo->mem = & mem->pub;
+
+  /* Check for an environment variable JPEGMEM; if found, override the
+   * default max_memory setting from jpeg_mem_init.  Note that the
+   * surrounding application may again override this value.
+   * If your system doesn't support getenv(), define NO_GETENV to disable
+   * this feature.
+   */
+#ifndef NO_GETENV
+  { char * memenv;
+
+    if ((memenv = getenv("JPEGMEM")) != NULL) {
+      char ch = 'x';
+
+      if (sscanf(memenv, "%ld%c", &max_to_use, &ch) > 0) {
+	if (ch == 'm' || ch == 'M')
+	  max_to_use *= 1000L;
+	mem->pub.max_memory_to_use = max_to_use * 1000L;
+      }
+    }
+  }
+#endif
+
+}
diff --git a/libjpegtwrp/jmemname.c b/libjpegtwrp/jmemname.c
new file mode 100644
index 000000000..ed96dee1b
--- /dev/null
+++ b/libjpegtwrp/jmemname.c
@@ -0,0 +1,276 @@
+/*
+ * jmemname.c
+ *
+ * Copyright (C) 1992-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file provides a generic implementation of the system-dependent
+ * portion of the JPEG memory manager.  This implementation assumes that
+ * you must explicitly construct a name for each temp file.
+ * Also, the problem of determining the amount of memory available
+ * is shoved onto the user.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jmemsys.h"		/* import the system-dependent declarations */
+
+#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare malloc(),free() */
+extern void * malloc JPP((size_t size));
+extern void free JPP((void *ptr));
+#endif
+
+#ifndef SEEK_SET		/* pre-ANSI systems may not define this; */
+#define SEEK_SET  0		/* if not, assume 0 is correct */
+#endif
+
+#ifdef DONT_USE_B_MODE		/* define mode parameters for fopen() */
+#define READ_BINARY	"r"
+#define RW_BINARY	"w+"
+#else
+#ifdef VMS			/* VMS is very nonstandard */
+#define READ_BINARY	"rb", "ctx=stm"
+#define RW_BINARY	"w+b", "ctx=stm"
+#else				/* standard ANSI-compliant case */
+#define READ_BINARY	"rb"
+#define RW_BINARY	"w+b"
+#endif
+#endif
+
+
+/*
+ * Selection of a file name for a temporary file.
+ * This is system-dependent!
+ *
+ * The code as given is suitable for most Unix systems, and it is easily
+ * modified for most non-Unix systems.  Some notes:
+ *  1.  The temp file is created in the directory named by TEMP_DIRECTORY.
+ *      The default value is /usr/tmp, which is the conventional place for
+ *      creating large temp files on Unix.  On other systems you'll probably
+ *      want to change the file location.  You can do this by editing the
+ *      #define, or (preferred) by defining TEMP_DIRECTORY in jconfig.h.
+ *
+ *  2.  If you need to change the file name as well as its location,
+ *      you can override the TEMP_FILE_NAME macro.  (Note that this is
+ *      actually a printf format string; it must contain %s and %d.)
+ *      Few people should need to do this.
+ *
+ *  3.  mktemp() is used to ensure that multiple processes running
+ *      simultaneously won't select the same file names.  If your system
+ *      doesn't have mktemp(), define NO_MKTEMP to do it the hard way.
+ *      (If you don't have <errno.h>, also define NO_ERRNO_H.)
+ *
+ *  4.  You probably want to define NEED_SIGNAL_CATCHER so that cjpeg.c/djpeg.c
+ *      will cause the temp files to be removed if you stop the program early.
+ */
+
+#ifndef TEMP_DIRECTORY		/* can override from jconfig.h or Makefile */
+#define TEMP_DIRECTORY  "/usr/tmp/" /* recommended setting for Unix */
+#endif
+
+static int next_file_num;	/* to distinguish among several temp files */
+
+#ifdef NO_MKTEMP
+
+#ifndef TEMP_FILE_NAME		/* can override from jconfig.h or Makefile */
+#define TEMP_FILE_NAME  "%sJPG%03d.TMP"
+#endif
+
+#ifndef NO_ERRNO_H
+#include <errno.h>		/* to define ENOENT */
+#endif
+
+/* ANSI C specifies that errno is a macro, but on older systems it's more
+ * likely to be a plain int variable.  And not all versions of errno.h
+ * bother to declare it, so we have to in order to be most portable.  Thus:
+ */
+#ifndef errno
+extern int errno;
+#endif
+
+
+LOCAL(void)
+select_file_name (char * fname)
+{
+  FILE * tfile;
+
+  /* Keep generating file names till we find one that's not in use */
+  for (;;) {
+    next_file_num++;		/* advance counter */
+    sprintf(fname, TEMP_FILE_NAME, TEMP_DIRECTORY, next_file_num);
+    if ((tfile = fopen(fname, READ_BINARY)) == NULL) {
+      /* fopen could have failed for a reason other than the file not
+       * being there; for example, file there but unreadable.
+       * If <errno.h> isn't available, then we cannot test the cause.
+       */
+#ifdef ENOENT
+      if (errno != ENOENT)
+	continue;
+#endif
+      break;
+    }
+    fclose(tfile);		/* oops, it's there; close tfile & try again */
+  }
+}
+
+#else /* ! NO_MKTEMP */
+
+/* Note that mktemp() requires the initial filename to end in six X's */
+#ifndef TEMP_FILE_NAME		/* can override from jconfig.h or Makefile */
+#define TEMP_FILE_NAME  "%sJPG%dXXXXXX"
+#endif
+
+LOCAL(void)
+select_file_name (char * fname)
+{
+  next_file_num++;		/* advance counter */
+  sprintf(fname, TEMP_FILE_NAME, TEMP_DIRECTORY, next_file_num);
+  mktemp(fname);		/* make sure file name is unique */
+  /* mktemp replaces the trailing XXXXXX with a unique string of characters */
+}
+
+#endif /* NO_MKTEMP */
+
+
+/*
+ * Memory allocation and freeing are controlled by the regular library
+ * routines malloc() and free().
+ */
+
+GLOBAL(void *)
+jpeg_get_small (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_small (j_common_ptr cinfo, void * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * "Large" objects are treated the same as "small" ones.
+ * NB: although we include FAR keywords in the routine declarations,
+ * this file won't actually work in 80x86 small/medium model; at least,
+ * you probably won't be able to process useful-size images in only 64KB.
+ */
+
+GLOBAL(void FAR *)
+jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void FAR *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_large (j_common_ptr cinfo, void FAR * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * This routine computes the total memory space available for allocation.
+ * It's impossible to do this in a portable way; our current solution is
+ * to make the user tell us (with a default value set at compile time).
+ * If you can actually get the available space, it's a good idea to subtract
+ * a slop factor of 5% or so.
+ */
+
+#ifndef DEFAULT_MAX_MEM		/* so can override from makefile */
+#define DEFAULT_MAX_MEM		1000000L /* default: one megabyte */
+#endif
+
+GLOBAL(long)
+jpeg_mem_available (j_common_ptr cinfo, long min_bytes_needed,
+		    long max_bytes_needed, long already_allocated)
+{
+  return cinfo->mem->max_memory_to_use - already_allocated;
+}
+
+
+/*
+ * Backing store (temporary file) management.
+ * Backing store objects are only used when the value returned by
+ * jpeg_mem_available is less than the total space needed.  You can dispense
+ * with these routines if you have plenty of virtual memory; see jmemnobs.c.
+ */
+
+
+METHODDEF(void)
+read_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+		    void FAR * buffer_address,
+		    long file_offset, long byte_count)
+{
+  if (fseek(info->temp_file, file_offset, SEEK_SET))
+    ERREXIT(cinfo, JERR_TFILE_SEEK);
+  if (JFREAD(info->temp_file, buffer_address, byte_count)
+      != (size_t) byte_count)
+    ERREXIT(cinfo, JERR_TFILE_READ);
+}
+
+
+METHODDEF(void)
+write_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+		     void FAR * buffer_address,
+		     long file_offset, long byte_count)
+{
+  if (fseek(info->temp_file, file_offset, SEEK_SET))
+    ERREXIT(cinfo, JERR_TFILE_SEEK);
+  if (JFWRITE(info->temp_file, buffer_address, byte_count)
+      != (size_t) byte_count)
+    ERREXIT(cinfo, JERR_TFILE_WRITE);
+}
+
+
+METHODDEF(void)
+close_backing_store (j_common_ptr cinfo, backing_store_ptr info)
+{
+  fclose(info->temp_file);	/* close the file */
+  unlink(info->temp_name);	/* delete the file */
+/* If your system doesn't have unlink(), use remove() instead.
+ * remove() is the ANSI-standard name for this function, but if
+ * your system was ANSI you'd be using jmemansi.c, right?
+ */
+  TRACEMSS(cinfo, 1, JTRC_TFILE_CLOSE, info->temp_name);
+}
+
+
+/*
+ * Initial opening of a backing-store object.
+ */
+
+GLOBAL(void)
+jpeg_open_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+			 long total_bytes_needed)
+{
+  select_file_name(info->temp_name);
+  if ((info->temp_file = fopen(info->temp_name, RW_BINARY)) == NULL)
+    ERREXITS(cinfo, JERR_TFILE_CREATE, info->temp_name);
+  info->read_backing_store = read_backing_store;
+  info->write_backing_store = write_backing_store;
+  info->close_backing_store = close_backing_store;
+  TRACEMSS(cinfo, 1, JTRC_TFILE_OPEN, info->temp_name);
+}
+
+
+/*
+ * These routines take care of any system-dependent initialization and
+ * cleanup required.
+ */
+
+GLOBAL(long)
+jpeg_mem_init (j_common_ptr cinfo)
+{
+  next_file_num = 0;		/* initialize temp file name generator */
+  return DEFAULT_MAX_MEM;	/* default for max_memory_to_use */
+}
+
+GLOBAL(void)
+jpeg_mem_term (j_common_ptr cinfo)
+{
+  /* no work */
+}
diff --git a/libjpegtwrp/jmemnobs.c b/libjpegtwrp/jmemnobs.c
new file mode 100644
index 000000000..eb8c33772
--- /dev/null
+++ b/libjpegtwrp/jmemnobs.c
@@ -0,0 +1,109 @@
+/*
+ * jmemnobs.c
+ *
+ * Copyright (C) 1992-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file provides a really simple implementation of the system-
+ * dependent portion of the JPEG memory manager.  This implementation
+ * assumes that no backing-store files are needed: all required space
+ * can be obtained from malloc().
+ * This is very portable in the sense that it'll compile on almost anything,
+ * but you'd better have lots of main memory (or virtual memory) if you want
+ * to process big images.
+ * Note that the max_memory_to_use option is ignored by this implementation.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jmemsys.h"		/* import the system-dependent declarations */
+
+#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare malloc(),free() */
+extern void * malloc JPP((size_t size));
+extern void free JPP((void *ptr));
+#endif
+
+
+/*
+ * Memory allocation and freeing are controlled by the regular library
+ * routines malloc() and free().
+ */
+
+GLOBAL(void *)
+jpeg_get_small (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_small (j_common_ptr cinfo, void * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * "Large" objects are treated the same as "small" ones.
+ * NB: although we include FAR keywords in the routine declarations,
+ * this file won't actually work in 80x86 small/medium model; at least,
+ * you probably won't be able to process useful-size images in only 64KB.
+ */
+
+GLOBAL(void FAR *)
+jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void FAR *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_large (j_common_ptr cinfo, void FAR * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * This routine computes the total memory space available for allocation.
+ * Here we always say, "we got all you want bud!"
+ */
+
+GLOBAL(long)
+jpeg_mem_available (j_common_ptr cinfo, long min_bytes_needed,
+		    long max_bytes_needed, long already_allocated)
+{
+  return max_bytes_needed;
+}
+
+
+/*
+ * Backing store (temporary file) management.
+ * Since jpeg_mem_available always promised the moon,
+ * this should never be called and we can just error out.
+ */
+
+GLOBAL(void)
+jpeg_open_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+			 long total_bytes_needed)
+{
+  ERREXIT(cinfo, JERR_NO_BACKING_STORE);
+}
+
+
+/*
+ * These routines take care of any system-dependent initialization and
+ * cleanup required.  Here, there isn't any.
+ */
+
+GLOBAL(long)
+jpeg_mem_init (j_common_ptr cinfo)
+{
+  return 0;			/* just set max_memory_to_use to 0 */
+}
+
+GLOBAL(void)
+jpeg_mem_term (j_common_ptr cinfo)
+{
+  /* no work */
+}
diff --git a/libjpegtwrp/jmemsys.h b/libjpegtwrp/jmemsys.h
new file mode 100644
index 000000000..2ed1c63a1
--- /dev/null
+++ b/libjpegtwrp/jmemsys.h
@@ -0,0 +1,204 @@
+/*
+ * jmemsys.h
+ *
+ * Copyright (C) 1992-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This include file defines the interface between the system-independent
+ * and system-dependent portions of the JPEG memory manager.  No other
+ * modules need include it.  (The system-independent portion is jmemmgr.c;
+ * there are several different versions of the system-dependent portion.)
+ *
+ * This file works as-is for the system-dependent memory managers supplied
+ * in the IJG distribution.  You may need to modify it if you write a
+ * custom memory manager.  If system-dependent changes are needed in
+ * this file, the best method is to #ifdef them based on a configuration
+ * symbol supplied in jconfig.h, as we have done with USE_MSDOS_MEMMGR
+ * and USE_MAC_MEMMGR.
+ */
+
+
+/* Short forms of external names for systems with brain-damaged linkers. */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jpeg_get_small		jGetSmall
+#define jpeg_free_small		jFreeSmall
+#define jpeg_get_large		jGetLarge
+#define jpeg_free_large		jFreeLarge
+#define jpeg_mem_available	jMemAvail
+#define jpeg_open_backing_store	jOpenBackStore
+#define jpeg_mem_init		jMemInit
+#define jpeg_mem_term		jMemTerm
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+
+/*
+ * These two functions are used to allocate and release small chunks of
+ * memory.  (Typically the total amount requested through jpeg_get_small is
+ * no more than 20K or so; this will be requested in chunks of a few K each.)
+ * Behavior should be the same as for the standard library functions malloc
+ * and free; in particular, jpeg_get_small must return NULL on failure.
+ * On most systems, these ARE malloc and free.  jpeg_free_small is passed the
+ * size of the object being freed, just in case it's needed.
+ * On an 80x86 machine using small-data memory model, these manage near heap.
+ */
+
+EXTERN(void *) jpeg_get_small JPP((j_common_ptr cinfo, size_t sizeofobject));
+EXTERN(void) jpeg_free_small JPP((j_common_ptr cinfo, void * object,
+				  size_t sizeofobject));
+
+/*
+ * These two functions are used to allocate and release large chunks of
+ * memory (up to the total free space designated by jpeg_mem_available).
+ * The interface is the same as above, except that on an 80x86 machine,
+ * far pointers are used.  On most other machines these are identical to
+ * the jpeg_get/free_small routines; but we keep them separate anyway,
+ * in case a different allocation strategy is desirable for large chunks.
+ */
+
+EXTERN(void FAR *) jpeg_get_large JPP((j_common_ptr cinfo,
+				       size_t sizeofobject));
+EXTERN(void) jpeg_free_large JPP((j_common_ptr cinfo, void FAR * object,
+				  size_t sizeofobject));
+
+/*
+ * The macro MAX_ALLOC_CHUNK designates the maximum number of bytes that may
+ * be requested in a single call to jpeg_get_large (and jpeg_get_small for that
+ * matter, but that case should never come into play).  This macro is needed
+ * to model the 64Kb-segment-size limit of far addressing on 80x86 machines.
+ * On those machines, we expect that jconfig.h will provide a proper value.
+ * On machines with 32-bit flat address spaces, any large constant may be used.
+ *
+ * NB: jmemmgr.c expects that MAX_ALLOC_CHUNK will be representable as type
+ * size_t and will be a multiple of sizeof(align_type).
+ */
+
+#ifndef MAX_ALLOC_CHUNK		/* may be overridden in jconfig.h */
+#define MAX_ALLOC_CHUNK  1000000000L
+#endif
+
+/*
+ * This routine computes the total space still available for allocation by
+ * jpeg_get_large.  If more space than this is needed, backing store will be
+ * used.  NOTE: any memory already allocated must not be counted.
+ *
+ * There is a minimum space requirement, corresponding to the minimum
+ * feasible buffer sizes; jmemmgr.c will request that much space even if
+ * jpeg_mem_available returns zero.  The maximum space needed, enough to hold
+ * all working storage in memory, is also passed in case it is useful.
+ * Finally, the total space already allocated is passed.  If no better
+ * method is available, cinfo->mem->max_memory_to_use - already_allocated
+ * is often a suitable calculation.
+ *
+ * It is OK for jpeg_mem_available to underestimate the space available
+ * (that'll just lead to more backing-store access than is really necessary).
+ * However, an overestimate will lead to failure.  Hence it's wise to subtract
+ * a slop factor from the true available space.  5% should be enough.
+ *
+ * On machines with lots of virtual memory, any large constant may be returned.
+ * Conversely, zero may be returned to always use the minimum amount of memory.
+ */
+
+EXTERN(long) jpeg_mem_available JPP((j_common_ptr cinfo,
+				     long min_bytes_needed,
+				     long max_bytes_needed,
+				     long already_allocated));
+
+
+/*
+ * This structure holds whatever state is needed to access a single
+ * backing-store object.  The read/write/close method pointers are called
+ * by jmemmgr.c to manipulate the backing-store object; all other fields
+ * are private to the system-dependent backing store routines.
+ */
+
+#define TEMP_NAME_LENGTH   64	/* max length of a temporary file's name */
+
+
+#ifdef USE_MSDOS_MEMMGR		/* DOS-specific junk */
+
+typedef unsigned short XMSH;	/* type of extended-memory handles */
+typedef unsigned short EMSH;	/* type of expanded-memory handles */
+
+typedef union {
+  short file_handle;		/* DOS file handle if it's a temp file */
+  XMSH xms_handle;		/* handle if it's a chunk of XMS */
+  EMSH ems_handle;		/* handle if it's a chunk of EMS */
+} handle_union;
+
+#endif /* USE_MSDOS_MEMMGR */
+
+#ifdef USE_MAC_MEMMGR		/* Mac-specific junk */
+#include <Files.h>
+#endif /* USE_MAC_MEMMGR */
+
+
+typedef struct backing_store_struct * backing_store_ptr;
+
+typedef struct backing_store_struct {
+  /* Methods for reading/writing/closing this backing-store object */
+  JMETHOD(void, read_backing_store, (j_common_ptr cinfo,
+				     backing_store_ptr info,
+				     void FAR * buffer_address,
+				     long file_offset, long byte_count));
+  JMETHOD(void, write_backing_store, (j_common_ptr cinfo,
+				      backing_store_ptr info,
+				      void FAR * buffer_address,
+				      long file_offset, long byte_count));
+  JMETHOD(void, close_backing_store, (j_common_ptr cinfo,
+				      backing_store_ptr info));
+
+  /* Private fields for system-dependent backing-store management */
+#ifdef USE_MSDOS_MEMMGR
+  /* For the MS-DOS manager (jmemdos.c), we need: */
+  handle_union handle;		/* reference to backing-store storage object */
+  char temp_name[TEMP_NAME_LENGTH]; /* name if it's a file */
+#else
+#ifdef USE_MAC_MEMMGR
+  /* For the Mac manager (jmemmac.c), we need: */
+  short temp_file;		/* file reference number to temp file */
+  FSSpec tempSpec;		/* the FSSpec for the temp file */
+  char temp_name[TEMP_NAME_LENGTH]; /* name if it's a file */
+#else
+#ifdef USE_ANDROID_ASHMEM
+  short temp_file;		/* file reference number to temp file */
+  unsigned char* addr;  /* the memory address mapped to ashmem */
+  long size;            /* the requested ashmem size */
+#else
+  /* For a typical implementation with temp files, we need: */
+  FILE * temp_file;		/* stdio reference to temp file */
+  char temp_name[TEMP_NAME_LENGTH]; /* name of temp file */
+#endif
+#endif
+#endif
+} backing_store_info;
+
+
+/*
+ * Initial opening of a backing-store object.  This must fill in the
+ * read/write/close pointers in the object.  The read/write routines
+ * may take an error exit if the specified maximum file size is exceeded.
+ * (If jpeg_mem_available always returns a large value, this routine can
+ * just take an error exit.)
+ */
+
+EXTERN(void) jpeg_open_backing_store JPP((j_common_ptr cinfo,
+					  backing_store_ptr info,
+					  long total_bytes_needed));
+
+
+/*
+ * These routines take care of any system-dependent initialization and
+ * cleanup required.  jpeg_mem_init will be called before anything is
+ * allocated (and, therefore, nothing in cinfo is of use except the error
+ * manager pointer).  It should return a suitable default value for
+ * max_memory_to_use; this may subsequently be overridden by the surrounding
+ * application.  (Note that max_memory_to_use is only important if
+ * jpeg_mem_available chooses to consult it ... no one else will.)
+ * jpeg_mem_term may assume that all requested memory has been freed and that
+ * all opened backing-store objects have been closed.
+ */
+
+EXTERN(long) jpeg_mem_init JPP((j_common_ptr cinfo));
+EXTERN(void) jpeg_mem_term JPP((j_common_ptr cinfo));
diff --git a/libjpegtwrp/jmorecfg.h b/libjpegtwrp/jmorecfg.h
new file mode 100644
index 000000000..236bbcbc9
--- /dev/null
+++ b/libjpegtwrp/jmorecfg.h
@@ -0,0 +1,387 @@
+/*
+ * jmorecfg.h
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains additional configuration options that customize the
+ * JPEG software for special applications or support machine-dependent
+ * optimizations.  Most users will not need to touch this file.
+ */
+
+/*
+ * Define ANDROID_RGB to enable specific optimizations for Android
+ *   JCS_RGBA_8888 support
+ *   JCS_RGB_565 support
+ * 
+ */
+
+#define ANDROID_RGB
+
+#ifdef ANDROID_RGB
+#define PACK_SHORT_565(r,g,b)  ((((r)<<8)&0xf800)|(((g)<<3)&0x7E0)|((b)>>3))
+#define PACK_TWO_PIXELS(l,r)   ((r<<16) | l)
+#define PACK_NEED_ALIGNMENT(ptr) (((int)(ptr))&3)
+#define WRITE_TWO_PIXELS(addr, pixels) do {     \
+         ((INT16*)(addr))[0] = (pixels);        \
+         ((INT16*)(addr))[1] = (pixels)>>16;    \
+    } while(0)
+#define WRITE_TWO_ALIGNED_PIXELS(addr, pixels)  ((*(INT32*)(addr)) = pixels)
+#define DITHER_565_R(r, dither) ((r) + ((dither)&0xFF))
+#define DITHER_565_G(g, dither) ((g) + (((dither)&0xFF)>>1))
+#define DITHER_565_B(b, dither) ((b) + ((dither)&0xFF))
+#endif
+
+/*
+ * Define BITS_IN_JSAMPLE as either
+ *   8   for 8-bit sample values (the usual setting)
+ *   12  for 12-bit sample values
+ * Only 8 and 12 are legal data precisions for lossy JPEG according to the
+ * JPEG standard, and the IJG code does not support anything else!
+ * We do not support run-time selection of data precision, sorry.
+ */
+
+#define BITS_IN_JSAMPLE  8	/* use 8 or 12 */
+
+
+/*
+ * Maximum number of components (color channels) allowed in JPEG image.
+ * To meet the letter of the JPEG spec, set this to 255.  However, darn
+ * few applications need more than 4 channels (maybe 5 for CMYK + alpha
+ * mask).  We recommend 10 as a reasonable compromise; use 4 if you are
+ * really short on memory.  (Each allowed component costs a hundred or so
+ * bytes of storage, whether actually used in an image or not.)
+ */
+
+#define MAX_COMPONENTS  10	/* maximum number of image components */
+
+
+/*
+ * Basic data types.
+ * You may need to change these if you have a machine with unusual data
+ * type sizes; for example, "char" not 8 bits, "short" not 16 bits,
+ * or "long" not 32 bits.  We don't care whether "int" is 16 or 32 bits,
+ * but it had better be at least 16.
+ */
+
+/* Representation of a single sample (pixel element value).
+ * We frequently allocate large arrays of these, so it's important to keep
+ * them small.  But if you have memory to burn and access to char or short
+ * arrays is very slow on your hardware, you might want to change these.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+/* JSAMPLE should be the smallest type that will hold the values 0..255.
+ * You can use a signed char by having GETJSAMPLE mask it with 0xFF.
+ */
+
+#ifdef HAVE_UNSIGNED_CHAR
+
+typedef unsigned char JSAMPLE;
+#define GETJSAMPLE(value)  ((int) (value))
+
+#else /* not HAVE_UNSIGNED_CHAR */
+
+typedef char JSAMPLE;
+#ifdef CHAR_IS_UNSIGNED
+#define GETJSAMPLE(value)  ((int) (value))
+#else
+#define GETJSAMPLE(value)  ((int) (value) & 0xFF)
+#endif /* CHAR_IS_UNSIGNED */
+
+#endif /* HAVE_UNSIGNED_CHAR */
+
+#define MAXJSAMPLE	255
+#define CENTERJSAMPLE	128
+
+#endif /* BITS_IN_JSAMPLE == 8 */
+
+
+#if BITS_IN_JSAMPLE == 12
+/* JSAMPLE should be the smallest type that will hold the values 0..4095.
+ * On nearly all machines "short" will do nicely.
+ */
+
+typedef short JSAMPLE;
+#define GETJSAMPLE(value)  ((int) (value))
+
+#define MAXJSAMPLE	4095
+#define CENTERJSAMPLE	2048
+
+#endif /* BITS_IN_JSAMPLE == 12 */
+
+
+/* Representation of a DCT frequency coefficient.
+ * This should be a signed value of at least 16 bits; "short" is usually OK.
+ * Again, we allocate large arrays of these, but you can change to int
+ * if you have memory to burn and "short" is really slow.
+ */
+
+typedef short JCOEF;
+
+
+/* Compressed datastreams are represented as arrays of JOCTET.
+ * These must be EXACTLY 8 bits wide, at least once they are written to
+ * external storage.  Note that when using the stdio data source/destination
+ * managers, this is also the data type passed to fread/fwrite.
+ */
+
+#ifdef HAVE_UNSIGNED_CHAR
+
+typedef unsigned char JOCTET;
+#define GETJOCTET(value)  (value)
+
+#else /* not HAVE_UNSIGNED_CHAR */
+
+typedef char JOCTET;
+#ifdef CHAR_IS_UNSIGNED
+#define GETJOCTET(value)  (value)
+#else
+#define GETJOCTET(value)  ((value) & 0xFF)
+#endif /* CHAR_IS_UNSIGNED */
+
+#endif /* HAVE_UNSIGNED_CHAR */
+
+
+/* These typedefs are used for various table entries and so forth.
+ * They must be at least as wide as specified; but making them too big
+ * won't cost a huge amount of memory, so we don't provide special
+ * extraction code like we did for JSAMPLE.  (In other words, these
+ * typedefs live at a different point on the speed/space tradeoff curve.)
+ */
+
+/* UINT8 must hold at least the values 0..255. */
+
+#ifdef HAVE_UNSIGNED_CHAR
+typedef unsigned char UINT8;
+#else /* not HAVE_UNSIGNED_CHAR */
+#ifdef CHAR_IS_UNSIGNED
+typedef char UINT8;
+#else /* not CHAR_IS_UNSIGNED */
+typedef short UINT8;
+#endif /* CHAR_IS_UNSIGNED */
+#endif /* HAVE_UNSIGNED_CHAR */
+
+/* UINT16 must hold at least the values 0..65535. */
+
+#ifdef HAVE_UNSIGNED_SHORT
+typedef unsigned short UINT16;
+#else /* not HAVE_UNSIGNED_SHORT */
+typedef unsigned int UINT16;
+#endif /* HAVE_UNSIGNED_SHORT */
+
+/* INT16 must hold at least the values -32768..32767. */
+
+#ifndef XMD_H			/* X11/xmd.h correctly defines INT16 */
+typedef short INT16;
+#endif
+
+/* INT32 must hold at least signed 32-bit values. */
+
+#ifndef XMD_H			/* X11/xmd.h correctly defines INT32 */
+typedef long INT32;
+#endif
+
+/* Datatype used for image dimensions.  The JPEG standard only supports
+ * images up to 64K*64K due to 16-bit fields in SOF markers.  Therefore
+ * "unsigned int" is sufficient on all machines.  However, if you need to
+ * handle larger images and you don't mind deviating from the spec, you
+ * can change this datatype.
+ */
+
+typedef unsigned int JDIMENSION;
+
+#define JPEG_MAX_DIMENSION  65500L  /* a tad under 64K to prevent overflows */
+
+
+/* These macros are used in all function definitions and extern declarations.
+ * You could modify them if you need to change function linkage conventions;
+ * in particular, you'll need to do that to make the library a Windows DLL.
+ * Another application is to make all functions global for use with debuggers
+ * or code profilers that require it.
+ */
+
+/* a function called through method pointers: */
+#define METHODDEF(type)		static type
+/* a function used only in its module: */
+#define LOCAL(type)		static type
+/* a function referenced thru EXTERNs: */
+#define GLOBAL(type)		type
+/* a reference to a GLOBAL function: */
+#define EXTERN(type)		extern type
+
+
+/* This macro is used to declare a "method", that is, a function pointer.
+ * We want to supply prototype parameters if the compiler can cope.
+ * Note that the arglist parameter must be parenthesized!
+ * Again, you can customize this if you need special linkage keywords.
+ */
+
+#ifdef HAVE_PROTOTYPES
+#define JMETHOD(type,methodname,arglist)  type (*methodname) arglist
+#else
+#define JMETHOD(type,methodname,arglist)  type (*methodname) ()
+#endif
+
+
+/* Here is the pseudo-keyword for declaring pointers that must be "far"
+ * on 80x86 machines.  Most of the specialized coding for 80x86 is handled
+ * by just saying "FAR *" where such a pointer is needed.  In a few places
+ * explicit coding is needed; see uses of the NEED_FAR_POINTERS symbol.
+ */
+
+#ifdef NEED_FAR_POINTERS
+#define FAR  far
+#else
+#define FAR
+#endif
+
+
+/*
+ * On a few systems, type boolean and/or its values FALSE, TRUE may appear
+ * in standard header files.  Or you may have conflicts with application-
+ * specific header files that you want to include together with these files.
+ * Defining HAVE_BOOLEAN before including jpeglib.h should make it work.
+ */
+
+#ifndef HAVE_BOOLEAN
+typedef int boolean;
+#endif
+#ifndef FALSE			/* in case these macros already exist */
+#define FALSE	0		/* values of boolean */
+#endif
+#ifndef TRUE
+#define TRUE	1
+#endif
+
+
+/*
+ * The remaining options affect code selection within the JPEG library,
+ * but they don't need to be visible to most applications using the library.
+ * To minimize application namespace pollution, the symbols won't be
+ * defined unless JPEG_INTERNALS or JPEG_INTERNAL_OPTIONS has been defined.
+ */
+
+#ifdef JPEG_INTERNALS
+#define JPEG_INTERNAL_OPTIONS
+#endif
+
+#ifdef JPEG_INTERNAL_OPTIONS
+
+
+/*
+ * These defines indicate whether to include various optional functions.
+ * Undefining some of these symbols will produce a smaller but less capable
+ * library.  Note that you can leave certain source files out of the
+ * compilation/linking process if you've #undef'd the corresponding symbols.
+ * (You may HAVE to do that if your compiler doesn't like null source files.)
+ */
+
+/* Arithmetic coding is unsupported for legal reasons.  Complaints to IBM. */
+
+/* Capability options common to encoder and decoder: */
+
+#define DCT_ISLOW_SUPPORTED	/* slow but accurate integer algorithm */
+#define DCT_IFAST_SUPPORTED	/* faster, less accurate integer method */
+#define DCT_FLOAT_SUPPORTED	/* floating-point: accurate, fast on fast HW */
+
+/* Encoder capability options: */
+
+#undef  C_ARITH_CODING_SUPPORTED    /* Arithmetic coding back end? */
+#define C_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */
+#define C_PROGRESSIVE_SUPPORTED	    /* Progressive JPEG? (Requires MULTISCAN)*/
+#define ENTROPY_OPT_SUPPORTED	    /* Optimization of entropy coding parms? */
+/* Note: if you selected 12-bit data precision, it is dangerous to turn off
+ * ENTROPY_OPT_SUPPORTED.  The standard Huffman tables are only good for 8-bit
+ * precision, so jchuff.c normally uses entropy optimization to compute
+ * usable tables for higher precision.  If you don't want to do optimization,
+ * you'll have to supply different default Huffman tables.
+ * The exact same statements apply for progressive JPEG: the default tables
+ * don't work for progressive mode.  (This may get fixed, however.)
+ */
+#define INPUT_SMOOTHING_SUPPORTED   /* Input image smoothing option? */
+
+/* Decoder capability options: */
+
+#undef  D_ARITH_CODING_SUPPORTED    /* Arithmetic coding back end? */
+#define D_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */
+#define D_PROGRESSIVE_SUPPORTED	    /* Progressive JPEG? (Requires MULTISCAN)*/
+#define SAVE_MARKERS_SUPPORTED	    /* jpeg_save_markers() needed? */
+#define BLOCK_SMOOTHING_SUPPORTED   /* Block smoothing? (Progressive only) */
+#define IDCT_SCALING_SUPPORTED	    /* Output rescaling via IDCT? */
+#undef  UPSAMPLE_SCALING_SUPPORTED  /* Output rescaling at upsample stage? */
+#define UPSAMPLE_MERGING_SUPPORTED  /* Fast path for sloppy upsampling? */
+#define QUANT_1PASS_SUPPORTED	    /* 1-pass color quantization? */
+#define QUANT_2PASS_SUPPORTED	    /* 2-pass color quantization? */
+
+/* more capability options later, no doubt */
+
+
+/*
+ * Ordering of RGB data in scanlines passed to or from the application.
+ * If your application wants to deal with data in the order B,G,R, just
+ * change these macros.  You can also deal with formats such as R,G,B,X
+ * (one extra byte per pixel) by changing RGB_PIXELSIZE.  Note that changing
+ * the offsets will also change the order in which colormap data is organized.
+ * RESTRICTIONS:
+ * 1. The sample applications cjpeg,djpeg do NOT support modified RGB formats.
+ * 2. These macros only affect RGB<=>YCbCr color conversion, so they are not
+ *    useful if you are using JPEG color spaces other than YCbCr or grayscale.
+ * 3. The color quantizer modules will not behave desirably if RGB_PIXELSIZE
+ *    is not 3 (they don't understand about dummy color components!).  So you
+ *    can't use color quantization if you change that value.
+ */
+
+#define RGB_RED		0	/* Offset of Red in an RGB scanline element */
+#define RGB_GREEN	1	/* Offset of Green */
+#define RGB_BLUE	2	/* Offset of Blue */
+#ifdef ANDROID_RGB
+#define RGB_ALPHA   3   /* Offset of Alpha */
+#endif
+#define RGB_PIXELSIZE   3   /* JSAMPLEs per RGB scanline element */
+
+/* Definitions for speed-related optimizations. */
+
+
+/* If your compiler supports inline functions, define INLINE
+ * as the inline keyword; otherwise define it as empty.
+ */
+
+#ifndef INLINE
+#ifdef __GNUC__			/* for instance, GNU C knows about inline */
+#define INLINE __inline__
+#endif
+#ifndef INLINE
+#define INLINE			/* default is to define it as empty */
+#endif
+#endif
+
+
+/* On some machines (notably 68000 series) "int" is 32 bits, but multiplying
+ * two 16-bit shorts is faster than multiplying two ints.  Define MULTIPLIER
+ * as short on such a machine.  MULTIPLIER must be at least 16 bits wide.
+ */
+
+#ifndef MULTIPLIER
+#define MULTIPLIER  int		/* type for fastest integer multiply */
+#endif
+
+
+/* FAST_FLOAT should be either float or double, whichever is done faster
+ * by your compiler.  (Note that this type is only used in the floating point
+ * DCT routines, so it only matters if you've defined DCT_FLOAT_SUPPORTED.)
+ * Typically, float is faster in ANSI C compilers, while double is faster in
+ * pre-ANSI compilers (because they insist on converting to double anyway).
+ * The code below therefore chooses float if we have ANSI-style prototypes.
+ */
+
+#ifndef FAST_FLOAT
+#ifdef HAVE_PROTOTYPES
+#define FAST_FLOAT  float
+#else
+#define FAST_FLOAT  double
+#endif
+#endif
+
+#endif /* JPEG_INTERNAL_OPTIONS */
diff --git a/libjpegtwrp/jpegint.h b/libjpegtwrp/jpegint.h
new file mode 100644
index 000000000..7bf9ffe95
--- /dev/null
+++ b/libjpegtwrp/jpegint.h
@@ -0,0 +1,475 @@
+/*
+ * jpegint.h
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Copyright (c) 2010, Code Aurora Forum. All rights reserved.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file provides common declarations for the various JPEG modules.
+ * These declarations are considered internal to the JPEG library; most
+ * applications using the library shouldn't need to include this file.
+ */
+
+
+/* Declarations for both compression & decompression */
+
+typedef enum {			/* Operating modes for buffer controllers */
+	JBUF_PASS_THRU,		/* Plain stripwise operation */
+	/* Remaining modes require a full-image buffer to have been created */
+	JBUF_SAVE_SOURCE,	/* Run source subobject only, save output */
+	JBUF_CRANK_DEST,	/* Run dest subobject only, using saved data */
+	JBUF_SAVE_AND_PASS	/* Run both subobjects, save output */
+} J_BUF_MODE;
+
+/* Values of global_state field (jdapi.c has some dependencies on ordering!) */
+#define CSTATE_START	100	/* after create_compress */
+#define CSTATE_SCANNING	101	/* start_compress done, write_scanlines OK */
+#define CSTATE_RAW_OK	102	/* start_compress done, write_raw_data OK */
+#define CSTATE_WRCOEFS	103	/* jpeg_write_coefficients done */
+#define DSTATE_START	200	/* after create_decompress */
+#define DSTATE_INHEADER	201	/* reading header markers, no SOS yet */
+#define DSTATE_READY	202	/* found SOS, ready for start_decompress */
+#define DSTATE_PRELOAD	203	/* reading multiscan file in start_decompress*/
+#define DSTATE_PRESCAN	204	/* performing dummy pass for 2-pass quant */
+#define DSTATE_SCANNING	205	/* start_decompress done, read_scanlines OK */
+#define DSTATE_RAW_OK	206	/* start_decompress done, read_raw_data OK */
+#define DSTATE_BUFIMAGE	207	/* expecting jpeg_start_output */
+#define DSTATE_BUFPOST	208	/* looking for SOS/EOI in jpeg_finish_output */
+#define DSTATE_RDCOEFS	209	/* reading file in jpeg_read_coefficients */
+#define DSTATE_STOPPING	210	/* looking for EOI in jpeg_finish_decompress */
+
+
+/* Declarations for compression modules */
+
+/* Master control module */
+struct jpeg_comp_master {
+  JMETHOD(void, prepare_for_pass, (j_compress_ptr cinfo));
+  JMETHOD(void, pass_startup, (j_compress_ptr cinfo));
+  JMETHOD(void, finish_pass, (j_compress_ptr cinfo));
+
+  /* State variables made visible to other modules */
+  boolean call_pass_startup;	/* True if pass_startup must be called */
+  boolean is_last_pass;		/* True during last pass */
+};
+
+/* Main buffer control (downsampled-data buffer) */
+struct jpeg_c_main_controller {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo, J_BUF_MODE pass_mode));
+  JMETHOD(void, process_data, (j_compress_ptr cinfo,
+			       JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+			       JDIMENSION in_rows_avail));
+};
+
+/* Compression preprocessing (downsampling input buffer control) */
+struct jpeg_c_prep_controller {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo, J_BUF_MODE pass_mode));
+  JMETHOD(void, pre_process_data, (j_compress_ptr cinfo,
+				   JSAMPARRAY input_buf,
+				   JDIMENSION *in_row_ctr,
+				   JDIMENSION in_rows_avail,
+				   JSAMPIMAGE output_buf,
+				   JDIMENSION *out_row_group_ctr,
+				   JDIMENSION out_row_groups_avail));
+};
+
+/* Coefficient buffer control */
+struct jpeg_c_coef_controller {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo, J_BUF_MODE pass_mode));
+  JMETHOD(boolean, compress_data, (j_compress_ptr cinfo,
+				   JSAMPIMAGE input_buf));
+};
+
+/* Colorspace conversion */
+struct jpeg_color_converter {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo));
+  JMETHOD(void, color_convert, (j_compress_ptr cinfo,
+				JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+				JDIMENSION output_row, int num_rows));
+};
+
+/* Downsampling */
+struct jpeg_downsampler {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo));
+  JMETHOD(void, downsample, (j_compress_ptr cinfo,
+			     JSAMPIMAGE input_buf, JDIMENSION in_row_index,
+			     JSAMPIMAGE output_buf,
+			     JDIMENSION out_row_group_index));
+
+  boolean need_context_rows;	/* TRUE if need rows above & below */
+};
+
+/* Forward DCT (also controls coefficient quantization) */
+struct jpeg_forward_dct {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo));
+  /* perhaps this should be an array??? */
+  JMETHOD(void, forward_DCT, (j_compress_ptr cinfo,
+			      jpeg_component_info * compptr,
+			      JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+			      JDIMENSION start_row, JDIMENSION start_col,
+			      JDIMENSION num_blocks));
+};
+
+/* Entropy encoding */
+struct jpeg_entropy_encoder {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo, boolean gather_statistics));
+  JMETHOD(boolean, encode_mcu, (j_compress_ptr cinfo, JBLOCKROW *MCU_data));
+  JMETHOD(void, finish_pass, (j_compress_ptr cinfo));
+};
+
+/* Marker writing */
+struct jpeg_marker_writer {
+  JMETHOD(void, write_file_header, (j_compress_ptr cinfo));
+  JMETHOD(void, write_frame_header, (j_compress_ptr cinfo));
+  JMETHOD(void, write_scan_header, (j_compress_ptr cinfo));
+  JMETHOD(void, write_file_trailer, (j_compress_ptr cinfo));
+  JMETHOD(void, write_tables_only, (j_compress_ptr cinfo));
+  /* These routines are exported to allow insertion of extra markers */
+  /* Probably only COM and APPn markers should be written this way */
+  JMETHOD(void, write_marker_header, (j_compress_ptr cinfo, int marker,
+				      unsigned int datalen));
+  JMETHOD(void, write_marker_byte, (j_compress_ptr cinfo, int val));
+};
+
+
+/* Declarations for decompression modules */
+
+/* Master control module */
+struct jpeg_decomp_master {
+  JMETHOD(void, prepare_for_output_pass, (j_decompress_ptr cinfo));
+  JMETHOD(void, finish_output_pass, (j_decompress_ptr cinfo));
+
+  /* State variables made visible to other modules */
+  boolean is_dummy_pass;	/* True during 1st pass for 2-pass quant */
+};
+
+/* Input control module */
+struct jpeg_input_controller {
+  JMETHOD(int, consume_input, (j_decompress_ptr cinfo));
+  JMETHOD(int, consume_input_build_huffman_index, (j_decompress_ptr cinfo,
+                    huffman_index *index, int scan_count));
+  JMETHOD(int, consume_markers, (j_decompress_ptr cinfo,
+                    huffman_index *index, int scan_count));
+  JMETHOD(void, reset_input_controller, (j_decompress_ptr cinfo));
+  JMETHOD(void, start_input_pass, (j_decompress_ptr cinfo));
+  JMETHOD(void, finish_input_pass, (j_decompress_ptr cinfo));
+
+  /* State variables made visible to other modules */
+  boolean has_multiple_scans;	/* True if file has multiple scans */
+  boolean eoi_reached;		/* True when EOI has been consumed */
+};
+
+/* Main buffer control (downsampled-data buffer) */
+struct jpeg_d_main_controller {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo, J_BUF_MODE pass_mode));
+  JMETHOD(void, process_data, (j_decompress_ptr cinfo,
+			       JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+			       JDIMENSION out_rows_avail));
+};
+
+/* Coefficient buffer control */
+struct jpeg_d_coef_controller {
+  JMETHOD(void, start_input_pass, (j_decompress_ptr cinfo));
+  JMETHOD(int, consume_data, (j_decompress_ptr cinfo));
+  JMETHOD(int, consume_data_build_huffman_index, (j_decompress_ptr cinfo,
+                    huffman_index* index, int scan_count));
+  JMETHOD(void, start_output_pass, (j_decompress_ptr cinfo));
+  JMETHOD(int, decompress_data, (j_decompress_ptr cinfo,
+				 JSAMPIMAGE output_buf));
+  /* Pointer to array of coefficient virtual arrays, or NULL if none */
+  jvirt_barray_ptr *coef_arrays;
+
+  /* column number of the first and last tile, respectively */
+  int column_left_boundary;
+  int column_right_boundary;
+
+  /* column number of the first and last MCU, respectively */
+  int MCU_column_left_boundary;
+  int MCU_column_right_boundary;
+
+  /* the number of MCU columns to skip from the indexed MCU, iM,
+   * to the requested MCU boundary, rM, where iM is the MCU that we sample
+   * into our index and is the nearest one to the left of rM.
+   */
+  int MCU_columns_to_skip;
+};
+
+/* Decompression postprocessing (color quantization buffer control) */
+struct jpeg_d_post_controller {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo, J_BUF_MODE pass_mode));
+  JMETHOD(void, post_process_data, (j_decompress_ptr cinfo,
+				    JSAMPIMAGE input_buf,
+				    JDIMENSION *in_row_group_ctr,
+				    JDIMENSION in_row_groups_avail,
+				    JSAMPARRAY output_buf,
+				    JDIMENSION *out_row_ctr,
+				    JDIMENSION out_rows_avail));
+};
+
+/* Marker reading & parsing */
+struct jpeg_marker_reader {
+  JMETHOD(void, reset_marker_reader, (j_decompress_ptr cinfo));
+  /* Read markers until SOS or EOI.
+   * Returns same codes as are defined for jpeg_consume_input:
+   * JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI.
+   */
+  JMETHOD(int, read_markers, (j_decompress_ptr cinfo));
+  JMETHOD(void, get_sos_marker_position, (j_decompress_ptr cinfo,
+                    huffman_index *index));
+  /* Read a restart marker --- exported for use by entropy decoder only */
+  jpeg_marker_parser_method read_restart_marker;
+
+  /* State of marker reader --- nominally internal, but applications
+   * supplying COM or APPn handlers might like to know the state.
+   */
+  boolean saw_SOI;		/* found SOI? */
+  boolean saw_SOF;		/* found SOF? */
+  int next_restart_num;		/* next restart number expected (0-7) */
+  int current_sos_marker_position;
+  unsigned int discarded_bytes;	/* # of bytes skipped looking for a marker */
+};
+
+/* Entropy decoding */
+struct jpeg_entropy_decoder {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
+  JMETHOD(boolean, decode_mcu, (j_decompress_ptr cinfo,
+				JBLOCKROW *MCU_data));
+  JMETHOD(boolean, decode_mcu_discard_coef, (j_decompress_ptr cinfo));
+  JMETHOD(void, configure_huffman_decoder, (j_decompress_ptr cinfo,
+                    huffman_offset_data offset));
+  JMETHOD(void, get_huffman_decoder_configuration, (j_decompress_ptr cinfo,
+                    huffman_offset_data *offset));
+
+  /* This is here to share code between baseline and progressive decoders; */
+  /* other modules probably should not use it */
+  boolean insufficient_data;	/* set TRUE after emitting warning */
+
+  huffman_index *index;
+};
+
+/* Inverse DCT (also performs dequantization) */
+typedef JMETHOD(void, inverse_DCT_method_ptr,
+		(j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		 JCOEFPTR coef_block,
+		 JSAMPARRAY output_buf, JDIMENSION output_col));
+
+struct jpeg_inverse_dct {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
+  /* It is useful to allow each component to have a separate IDCT method. */
+  inverse_DCT_method_ptr inverse_DCT[MAX_COMPONENTS];
+};
+
+/* Upsampling (note that upsampler must also call color converter) */
+struct jpeg_upsampler {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
+  JMETHOD(void, upsample, (j_decompress_ptr cinfo,
+			   JSAMPIMAGE input_buf,
+			   JDIMENSION *in_row_group_ctr,
+			   JDIMENSION in_row_groups_avail,
+			   JSAMPARRAY output_buf,
+			   JDIMENSION *out_row_ctr,
+			   JDIMENSION out_rows_avail));
+
+  boolean need_context_rows;	/* TRUE if need rows above & below */
+};
+
+/* Colorspace conversion */
+struct jpeg_color_deconverter {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
+  JMETHOD(void, color_convert, (j_decompress_ptr cinfo,
+				JSAMPIMAGE input_buf, JDIMENSION input_row,
+				JSAMPARRAY output_buf, int num_rows));
+};
+
+/* Color quantization or color precision reduction */
+struct jpeg_color_quantizer {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo, boolean is_pre_scan));
+  JMETHOD(void, color_quantize, (j_decompress_ptr cinfo,
+				 JSAMPARRAY input_buf, JSAMPARRAY output_buf,
+				 int num_rows));
+  JMETHOD(void, finish_pass, (j_decompress_ptr cinfo));
+  JMETHOD(void, new_color_map, (j_decompress_ptr cinfo));
+};
+
+#ifdef ANDROID_JPEG_USE_VENUM
+
+/* IDCT routines */
+EXTERN (void) idct_1x1_venum (INT16 * coeffPtr, INT16 * samplePtr, INT32 stride);
+EXTERN (void) idct_2x2_venum (INT16 * coeffPtr, INT16 * samplePtr, INT32 stride);
+EXTERN (void) idct_4x4_venum (INT16 * coeffPtr, INT16 * samplePtr, INT32 stride);
+EXTERN (void) idct_8x8_venum (INT16 * coeffPtr, INT16 * samplePtr, INT32 stride);
+
+#ifndef ANDROID_JPEG_DISABLE_VENUM_YCC_RGB_565
+/* Color conversion routines */
+EXTERN (void) yvup2rgb565_venum (UINT8 *pLumaLine,
+                UINT8 *pCrLine,
+                UINT8 *pCbLine,
+                UINT8 *pRGB565Line,
+                JDIMENSION nLineWidth);
+EXTERN (void) yyvup2rgb565_venum (UINT8 * pLumaLine,
+                UINT8 *pCrLine,
+                UINT8 *pCbLine,
+                UINT8 * pRGB565Line,
+                JDIMENSION nLineWidth);
+#endif
+EXTERN (void) yvup2bgr888_venum (UINT8 * pLumaLine,
+                UINT8 *pCrLine,
+                UINT8 *pCbLine,
+                UINT8 * pBGR888Line,
+                JDIMENSION nLineWidth);
+EXTERN (void) yyvup2bgr888_venum (UINT8 * pLumaLine,
+                UINT8 *pCrLine,
+                UINT8 *pCbLine,
+                UINT8 * pBGR888Line,
+                JDIMENSION nLineWidth);
+EXTERN (void) yvup2abgr8888_venum (UINT8 * pLumaLine,
+                UINT8 *pCrLine,
+                UINT8 *pCbLine,
+                UINT8 * pABGR888Line,
+                JDIMENSION nLineWidth);
+EXTERN (void) yyvup2abgr8888_venum (UINT8 * pLumaLine,
+                UINT8 *pCrLine,
+                UINT8 *pCbLine,
+                UINT8 * pABGR888Line,
+                JDIMENSION nLineWidth);
+#endif
+
+/* Miscellaneous useful macros */
+
+#undef MAX
+#define MAX(a,b)	((a) > (b) ? (a) : (b))
+#undef MIN
+#define MIN(a,b)	((a) < (b) ? (a) : (b))
+
+
+/* We assume that right shift corresponds to signed division by 2 with
+ * rounding towards minus infinity.  This is correct for typical "arithmetic
+ * shift" instructions that shift in copies of the sign bit.  But some
+ * C compilers implement >> with an unsigned shift.  For these machines you
+ * must define RIGHT_SHIFT_IS_UNSIGNED.
+ * RIGHT_SHIFT provides a proper signed right shift of an INT32 quantity.
+ * It is only applied with constant shift counts.  SHIFT_TEMPS must be
+ * included in the variables of any routine using RIGHT_SHIFT.
+ */
+
+#ifdef RIGHT_SHIFT_IS_UNSIGNED
+#define SHIFT_TEMPS	INT32 shift_temp;
+#define RIGHT_SHIFT(x,shft)  \
+	((shift_temp = (x)) < 0 ? \
+	 (shift_temp >> (shft)) | ((~((INT32) 0)) << (32-(shft))) : \
+	 (shift_temp >> (shft)))
+#else
+#define SHIFT_TEMPS
+#define RIGHT_SHIFT(x,shft)	((x) >> (shft))
+#endif
+
+
+/* Short forms of external names for systems with brain-damaged linkers. */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jinit_compress_master	jICompress
+#define jinit_c_master_control	jICMaster
+#define jinit_c_main_controller	jICMainC
+#define jinit_c_prep_controller	jICPrepC
+#define jinit_c_coef_controller	jICCoefC
+#define jinit_color_converter	jICColor
+#define jinit_downsampler	jIDownsampler
+#define jinit_forward_dct	jIFDCT
+#define jinit_huff_encoder	jIHEncoder
+#define jinit_phuff_encoder	jIPHEncoder
+#define jinit_marker_writer	jIMWriter
+#define jinit_master_decompress	jIDMaster
+#define jinit_d_main_controller	jIDMainC
+#define jinit_d_coef_controller	jIDCoefC
+#define jinit_d_post_controller	jIDPostC
+#define jinit_input_controller	jIInCtlr
+#define jinit_marker_reader	jIMReader
+#define jinit_huff_decoder	jIHDecoder
+#define jinit_phuff_decoder	jIPHDecoder
+#define jinit_inverse_dct	jIIDCT
+#define jinit_upsampler		jIUpsampler
+#define jinit_color_deconverter	jIDColor
+#define jinit_1pass_quantizer	jI1Quant
+#define jinit_2pass_quantizer	jI2Quant
+#define jinit_merged_upsampler	jIMUpsampler
+#define jinit_memory_mgr	jIMemMgr
+#define jdiv_round_up		jDivRound
+#define jround_up		jRound
+#define jcopy_sample_rows	jCopySamples
+#define jcopy_block_row		jCopyBlocks
+#define jzero_far		jZeroFar
+#define jpeg_zigzag_order	jZIGTable
+#define jpeg_natural_order	jZAGTable
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+
+/* Compression module initialization routines */
+EXTERN(void) jinit_compress_master JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_c_master_control JPP((j_compress_ptr cinfo,
+					 boolean transcode_only));
+EXTERN(void) jinit_c_main_controller JPP((j_compress_ptr cinfo,
+					  boolean need_full_buffer));
+EXTERN(void) jinit_c_prep_controller JPP((j_compress_ptr cinfo,
+					  boolean need_full_buffer));
+EXTERN(void) jinit_c_coef_controller JPP((j_compress_ptr cinfo,
+					  boolean need_full_buffer));
+EXTERN(void) jinit_color_converter JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_downsampler JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_forward_dct JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_huff_encoder JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_phuff_encoder JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_marker_writer JPP((j_compress_ptr cinfo));
+/* Decompression module initialization routines */
+EXTERN(void) jinit_master_decompress JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_d_main_controller JPP((j_decompress_ptr cinfo,
+					  boolean need_full_buffer));
+EXTERN(void) jinit_d_coef_controller JPP((j_decompress_ptr cinfo,
+					  boolean need_full_buffer));
+EXTERN(void) jinit_d_post_controller JPP((j_decompress_ptr cinfo,
+					  boolean need_full_buffer));
+EXTERN(void) jinit_input_controller JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_marker_reader JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_huff_decoder JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_huff_decoder_no_data JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_phuff_decoder JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_inverse_dct JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_upsampler JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_color_deconverter JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_1pass_quantizer JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_2pass_quantizer JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_merged_upsampler JPP((j_decompress_ptr cinfo));
+EXTERN(void) jpeg_decompress_per_scan_setup (j_decompress_ptr cinfo);
+/* Memory manager initialization */
+EXTERN(void) jinit_memory_mgr JPP((j_common_ptr cinfo));
+
+/* Utility routines in jutils.c */
+EXTERN(long) jdiv_round_up JPP((long a, long b));
+EXTERN(long) jround_up JPP((long a, long b));
+EXTERN(long) jmin JPP((long a, long b));
+EXTERN(void) jcopy_sample_rows JPP((JSAMPARRAY input_array, int source_row,
+				    JSAMPARRAY output_array, int dest_row,
+				    int num_rows, JDIMENSION num_cols));
+EXTERN(void) jcopy_block_row JPP((JBLOCKROW input_row, JBLOCKROW output_row,
+				  JDIMENSION num_blocks));
+EXTERN(void) jzero_far JPP((void FAR * target, size_t bytestozero));
+
+EXTERN(void) jset_input_stream_position JPP((j_decompress_ptr cinfo,
+                    int offset));
+EXTERN(void) jset_input_stream_position_bit JPP((j_decompress_ptr cinfo,
+                    int byte_offset, int bit_left, INT32 buf));
+
+EXTERN(int) jget_input_stream_position JPP((j_decompress_ptr cinfo));
+/* Constant tables in jutils.c */
+#if 0				/* This table is not actually needed in v6a */
+extern const int jpeg_zigzag_order[]; /* natural coef order to zigzag order */
+#endif
+extern const int jpeg_natural_order[]; /* zigzag coef order to natural order */
+
+/* Suppress undefined-structure complaints if necessary. */
+
+#ifdef INCOMPLETE_TYPES_BROKEN
+#ifndef AM_MEMORY_MANAGER	/* only jmemmgr.c defines these */
+struct jvirt_sarray_control { long dummy; };
+struct jvirt_barray_control { long dummy; };
+#endif
+#endif /* INCOMPLETE_TYPES_BROKEN */
diff --git a/libjpegtwrp/jpeglib.h b/libjpegtwrp/jpeglib.h
new file mode 100644
index 000000000..07e687294
--- /dev/null
+++ b/libjpegtwrp/jpeglib.h
@@ -0,0 +1,1184 @@
+/*
+ * jpeglib.h
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file defines the application interface for the JPEG library.
+ * Most applications using the library need only include this file,
+ * and perhaps jerror.h if they want to know the exact error codes.
+ */
+
+#ifndef JPEGLIB_H
+#define JPEGLIB_H
+
+/*
+ * First we include the configuration files that record how this
+ * installation of the JPEG library is set up.  jconfig.h can be
+ * generated automatically for many systems.  jmorecfg.h contains
+ * manual configuration options that most people need not worry about.
+ */
+
+#ifndef JCONFIG_INCLUDED	/* in case jinclude.h already did */
+#include "jconfig.h"		/* widely used configuration options */
+#endif
+#include "jmorecfg.h"		/* seldom changed options */
+
+
+/* Version ID for the JPEG library.
+ * Might be useful for tests like "#if JPEG_LIB_VERSION >= 60".
+ */
+
+#define JPEG_LIB_VERSION  62	/* Version 6b */
+
+
+/* Various constants determining the sizes of things.
+ * All of these are specified by the JPEG standard, so don't change them
+ * if you want to be compatible.
+ */
+
+#define DCTSIZE		    8	/* The basic DCT block is 8x8 samples */
+#define DCTSIZE2	    64	/* DCTSIZE squared; # of elements in a block */
+#define NUM_QUANT_TBLS      4	/* Quantization tables are numbered 0..3 */
+#define NUM_HUFF_TBLS       4	/* Huffman tables are numbered 0..3 */
+#define NUM_ARITH_TBLS      16	/* Arith-coding tables are numbered 0..15 */
+#define MAX_COMPS_IN_SCAN   4	/* JPEG limit on # of components in one scan */
+#define MAX_SAMP_FACTOR     4	/* JPEG limit on sampling factors */
+/* Unfortunately, some bozo at Adobe saw no reason to be bound by the standard;
+ * the PostScript DCT filter can emit files with many more than 10 blocks/MCU.
+ * If you happen to run across such a file, you can up D_MAX_BLOCKS_IN_MCU
+ * to handle it.  We even let you do this from the jconfig.h file.  However,
+ * we strongly discourage changing C_MAX_BLOCKS_IN_MCU; just because Adobe
+ * sometimes emits noncompliant files doesn't mean you should too.
+ */
+#define C_MAX_BLOCKS_IN_MCU   10 /* compressor's limit on blocks per MCU */
+#ifndef D_MAX_BLOCKS_IN_MCU
+#define D_MAX_BLOCKS_IN_MCU   10 /* decompressor's limit on blocks per MCU */
+#endif
+
+
+/* Data structures for images (arrays of samples and of DCT coefficients).
+ * On 80x86 machines, the image arrays are too big for near pointers,
+ * but the pointer arrays can fit in near memory.
+ */
+
+typedef JSAMPLE FAR *JSAMPROW;	/* ptr to one image row of pixel samples. */
+typedef JSAMPROW *JSAMPARRAY;	/* ptr to some rows (a 2-D sample array) */
+typedef JSAMPARRAY *JSAMPIMAGE;	/* a 3-D sample array: top index is color */
+
+typedef JCOEF JBLOCK[DCTSIZE2];	/* one block of coefficients */
+typedef JBLOCK FAR *JBLOCKROW;	/* pointer to one row of coefficient blocks */
+typedef JBLOCKROW *JBLOCKARRAY;		/* a 2-D array of coefficient blocks */
+typedef JBLOCKARRAY *JBLOCKIMAGE;	/* a 3-D array of coefficient blocks */
+
+typedef JCOEF FAR *JCOEFPTR;	/* useful in a couple of places */
+
+
+/* Types for JPEG compression parameters and working tables. */
+
+
+/* DCT coefficient quantization tables. */
+
+typedef struct {
+  /* This array gives the coefficient quantizers in natural array order
+   * (not the zigzag order in which they are stored in a JPEG DQT marker).
+   * CAUTION: IJG versions prior to v6a kept this array in zigzag order.
+   */
+  UINT16 quantval[DCTSIZE2];	/* quantization step for each coefficient */
+  /* This field is used only during compression.  It's initialized FALSE when
+   * the table is created, and set TRUE when it's been output to the file.
+   * You could suppress output of a table by setting this to TRUE.
+   * (See jpeg_suppress_tables for an example.)
+   */
+  boolean sent_table;		/* TRUE when table has been output */
+} JQUANT_TBL;
+
+
+/* Huffman coding tables. */
+
+typedef struct {
+  /* These two fields directly represent the contents of a JPEG DHT marker */
+  UINT8 bits[17];		/* bits[k] = # of symbols with codes of */
+				/* length k bits; bits[0] is unused */
+  UINT8 huffval[256];		/* The symbols, in order of incr code length */
+  /* This field is used only during compression.  It's initialized FALSE when
+   * the table is created, and set TRUE when it's been output to the file.
+   * You could suppress output of a table by setting this to TRUE.
+   * (See jpeg_suppress_tables for an example.)
+   */
+  boolean sent_table;		/* TRUE when table has been output */
+} JHUFF_TBL;
+
+
+/* Basic info about one component (color channel). */
+
+typedef struct {
+  /* These values are fixed over the whole image. */
+  /* For compression, they must be supplied by parameter setup; */
+  /* for decompression, they are read from the SOF marker. */
+  int component_id;		/* identifier for this component (0..255) */
+  int component_index;		/* its index in SOF or cinfo->comp_info[] */
+  int h_samp_factor;		/* horizontal sampling factor (1..4) */
+  int v_samp_factor;		/* vertical sampling factor (1..4) */
+  int quant_tbl_no;		/* quantization table selector (0..3) */
+  /* These values may vary between scans. */
+  /* For compression, they must be supplied by parameter setup; */
+  /* for decompression, they are read from the SOS marker. */
+  /* The decompressor output side may not use these variables. */
+  int dc_tbl_no;		/* DC entropy table selector (0..3) */
+  int ac_tbl_no;		/* AC entropy table selector (0..3) */
+
+  /* Remaining fields should be treated as private by applications. */
+
+  /* These values are computed during compression or decompression startup: */
+  /* Component's size in DCT blocks.
+   * Any dummy blocks added to complete an MCU are not counted; therefore
+   * these values do not depend on whether a scan is interleaved or not.
+   */
+  JDIMENSION width_in_blocks;
+  JDIMENSION height_in_blocks;
+  /* Size of a DCT block in samples.  Always DCTSIZE for compression.
+   * For decompression this is the size of the output from one DCT block,
+   * reflecting any scaling we choose to apply during the IDCT step.
+   * Values of 1,2,4,8 are likely to be supported.  Note that different
+   * components may receive different IDCT scalings.
+   */
+  int DCT_scaled_size;
+  /* The downsampled dimensions are the component's actual, unpadded number
+   * of samples at the main buffer (preprocessing/compression interface), thus
+   * downsampled_width = ceil(image_width * Hi/Hmax)
+   * and similarly for height.  For decompression, IDCT scaling is included, so
+   * downsampled_width = ceil(image_width * Hi/Hmax * DCT_scaled_size/DCTSIZE)
+   */
+  JDIMENSION downsampled_width;	 /* actual width in samples */
+  JDIMENSION downsampled_height; /* actual height in samples */
+  /* This flag is used only for decompression.  In cases where some of the
+   * components will be ignored (eg grayscale output from YCbCr image),
+   * we can skip most computations for the unused components.
+   */
+  boolean component_needed;	/* do we need the value of this component? */
+
+  /* These values are computed before starting a scan of the component. */
+  /* The decompressor output side may not use these variables. */
+  int MCU_width;		/* number of blocks per MCU, horizontally */
+  int MCU_height;		/* number of blocks per MCU, vertically */
+  int MCU_blocks;		/* MCU_width * MCU_height */
+  int MCU_sample_width;		/* MCU width in samples, MCU_width*DCT_scaled_size */
+  int last_col_width;		/* # of non-dummy blocks across in last MCU */
+  int last_row_height;		/* # of non-dummy blocks down in last MCU */
+
+  /* Saved quantization table for component; NULL if none yet saved.
+   * See jdinput.c comments about the need for this information.
+   * This field is currently used only for decompression.
+   */
+  JQUANT_TBL * quant_table;
+
+  /* Private per-component storage for DCT or IDCT subsystem. */
+  void * dct_table;
+} jpeg_component_info;
+
+
+/* The script for encoding a multiple-scan file is an array of these: */
+
+typedef struct {
+  int comps_in_scan;		/* number of components encoded in this scan */
+  int component_index[MAX_COMPS_IN_SCAN]; /* their SOF/comp_info[] indexes */
+  int Ss, Se;			/* progressive JPEG spectral selection parms */
+  int Ah, Al;			/* progressive JPEG successive approx. parms */
+} jpeg_scan_info;
+
+/* The decompressor can save APPn and COM markers in a list of these: */
+
+typedef struct jpeg_marker_struct FAR * jpeg_saved_marker_ptr;
+
+struct jpeg_marker_struct {
+  jpeg_saved_marker_ptr next;	/* next in list, or NULL */
+  UINT8 marker;			/* marker code: JPEG_COM, or JPEG_APP0+n */
+  unsigned int original_length;	/* # bytes of data in the file */
+  unsigned int data_length;	/* # bytes of data saved at data[] */
+  JOCTET FAR * data;		/* the data contained in the marker */
+  /* the marker length word is not counted in data_length or original_length */
+};
+
+/* Known color spaces. */
+
+typedef enum {
+	JCS_UNKNOWN,		/* error/unspecified */
+	JCS_GRAYSCALE,		/* monochrome */
+	JCS_RGB,		/* red/green/blue */
+	JCS_YCbCr,		/* Y/Cb/Cr (also known as YUV) */
+	JCS_CMYK,		/* C/M/Y/K */
+	JCS_YCCK,		/* Y/Cb/Cr/K */
+#ifdef ANDROID_RGB
+    JCS_RGBA_8888,  /* red/green/blue/alpha */
+    JCS_RGB_565     /* red/green/blue in 565 format */
+#endif
+} J_COLOR_SPACE;
+
+/* DCT/IDCT algorithm options. */
+
+typedef enum {
+	JDCT_ISLOW,		/* slow but accurate integer algorithm */
+	JDCT_IFAST,		/* faster, less accurate integer method */
+	JDCT_FLOAT		/* floating-point: accurate, fast on fast HW */
+} J_DCT_METHOD;
+
+#ifndef JDCT_DEFAULT		/* may be overridden in jconfig.h */
+#define JDCT_DEFAULT  JDCT_ISLOW
+#endif
+#ifndef JDCT_FASTEST		/* may be overridden in jconfig.h */
+#define JDCT_FASTEST  JDCT_IFAST
+#endif
+
+/* Dithering options for decompression. */
+
+typedef enum {
+	JDITHER_NONE,		/* no dithering */
+	JDITHER_ORDERED,	/* simple ordered dither */
+	JDITHER_FS		/* Floyd-Steinberg error diffusion dither */
+} J_DITHER_MODE;
+
+
+/* Common fields between JPEG compression and decompression master structs. */
+
+#define jpeg_common_fields \
+  struct jpeg_error_mgr * err;	/* Error handler module */\
+  struct jpeg_memory_mgr * mem;	/* Memory manager module */\
+  struct jpeg_progress_mgr * progress; /* Progress monitor, or NULL if none */\
+  void * client_data;		/* Available for use by application */\
+  boolean is_decompressor;	/* So common code can tell which is which */\
+  int global_state		/* For checking call sequence validity */
+
+/* Routines that are to be used by both halves of the library are declared
+ * to receive a pointer to this structure.  There are no actual instances of
+ * jpeg_common_struct, only of jpeg_compress_struct and jpeg_decompress_struct.
+ */
+struct jpeg_common_struct {
+  jpeg_common_fields;		/* Fields common to both master struct types */
+  /* Additional fields follow in an actual jpeg_compress_struct or
+   * jpeg_decompress_struct.  All three structs must agree on these
+   * initial fields!  (This would be a lot cleaner in C++.)
+   */
+};
+
+typedef struct jpeg_common_struct * j_common_ptr;
+typedef struct jpeg_compress_struct * j_compress_ptr;
+typedef struct jpeg_decompress_struct * j_decompress_ptr;
+
+
+/* Master record for a compression instance */
+
+struct jpeg_compress_struct {
+  jpeg_common_fields;		/* Fields shared with jpeg_decompress_struct */
+
+  /* Destination for compressed data */
+  struct jpeg_destination_mgr * dest;
+
+  /* Description of source image --- these fields must be filled in by
+   * outer application before starting compression.  in_color_space must
+   * be correct before you can even call jpeg_set_defaults().
+   */
+
+  JDIMENSION image_width;	/* input image width */
+  JDIMENSION image_height;	/* input image height */
+  int input_components;		/* # of color components in input image */
+  J_COLOR_SPACE in_color_space;	/* colorspace of input image */
+
+  double input_gamma;		/* image gamma of input image */
+
+  /* Compression parameters --- these fields must be set before calling
+   * jpeg_start_compress().  We recommend calling jpeg_set_defaults() to
+   * initialize everything to reasonable defaults, then changing anything
+   * the application specifically wants to change.  That way you won't get
+   * burnt when new parameters are added.  Also note that there are several
+   * helper routines to simplify changing parameters.
+   */
+
+  int data_precision;		/* bits of precision in image data */
+
+  int num_components;		/* # of color components in JPEG image */
+  J_COLOR_SPACE jpeg_color_space; /* colorspace of JPEG image */
+
+  jpeg_component_info * comp_info;
+  /* comp_info[i] describes component that appears i'th in SOF */
+
+  JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS];
+  /* ptrs to coefficient quantization tables, or NULL if not defined */
+
+  JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  /* ptrs to Huffman coding tables, or NULL if not defined */
+
+  UINT8 arith_dc_L[NUM_ARITH_TBLS]; /* L values for DC arith-coding tables */
+  UINT8 arith_dc_U[NUM_ARITH_TBLS]; /* U values for DC arith-coding tables */
+  UINT8 arith_ac_K[NUM_ARITH_TBLS]; /* Kx values for AC arith-coding tables */
+
+  int num_scans;		/* # of entries in scan_info array */
+  const jpeg_scan_info * scan_info; /* script for multi-scan file, or NULL */
+  /* The default value of scan_info is NULL, which causes a single-scan
+   * sequential JPEG file to be emitted.  To create a multi-scan file,
+   * set num_scans and scan_info to point to an array of scan definitions.
+   */
+
+  boolean raw_data_in;		/* TRUE=caller supplies downsampled data */
+  boolean arith_code;		/* TRUE=arithmetic coding, FALSE=Huffman */
+  boolean optimize_coding;	/* TRUE=optimize entropy encoding parms */
+  boolean CCIR601_sampling;	/* TRUE=first samples are cosited */
+  int smoothing_factor;		/* 1..100, or 0 for no input smoothing */
+  J_DCT_METHOD dct_method;	/* DCT algorithm selector */
+
+  /* The restart interval can be specified in absolute MCUs by setting
+   * restart_interval, or in MCU rows by setting restart_in_rows
+   * (in which case the correct restart_interval will be figured
+   * for each scan).
+   */
+  unsigned int restart_interval; /* MCUs per restart, or 0 for no restart */
+  int restart_in_rows;		/* if > 0, MCU rows per restart interval */
+
+  /* Parameters controlling emission of special markers. */
+
+  boolean write_JFIF_header;	/* should a JFIF marker be written? */
+  UINT8 JFIF_major_version;	/* What to write for the JFIF version number */
+  UINT8 JFIF_minor_version;
+  /* These three values are not used by the JPEG code, merely copied */
+  /* into the JFIF APP0 marker.  density_unit can be 0 for unknown, */
+  /* 1 for dots/inch, or 2 for dots/cm.  Note that the pixel aspect */
+  /* ratio is defined by X_density/Y_density even when density_unit=0. */
+  UINT8 density_unit;		/* JFIF code for pixel size units */
+  UINT16 X_density;		/* Horizontal pixel density */
+  UINT16 Y_density;		/* Vertical pixel density */
+  boolean write_Adobe_marker;	/* should an Adobe marker be written? */
+
+  /* State variable: index of next scanline to be written to
+   * jpeg_write_scanlines().  Application may use this to control its
+   * processing loop, e.g., "while (next_scanline < image_height)".
+   */
+
+  JDIMENSION next_scanline;	/* 0 .. image_height-1  */
+
+  /* Remaining fields are known throughout compressor, but generally
+   * should not be touched by a surrounding application.
+   */
+
+  /*
+   * These fields are computed during compression startup
+   */
+  boolean progressive_mode;	/* TRUE if scan script uses progressive mode */
+  int max_h_samp_factor;	/* largest h_samp_factor */
+  int max_v_samp_factor;	/* largest v_samp_factor */
+
+  JDIMENSION total_iMCU_rows;	/* # of iMCU rows to be input to coef ctlr */
+  /* The coefficient controller receives data in units of MCU rows as defined
+   * for fully interleaved scans (whether the JPEG file is interleaved or not).
+   * There are v_samp_factor * DCTSIZE sample rows of each component in an
+   * "iMCU" (interleaved MCU) row.
+   */
+
+  /*
+   * These fields are valid during any one scan.
+   * They describe the components and MCUs actually appearing in the scan.
+   */
+  int comps_in_scan;		/* # of JPEG components in this scan */
+  jpeg_component_info * cur_comp_info[MAX_COMPS_IN_SCAN];
+  /* *cur_comp_info[i] describes component that appears i'th in SOS */
+
+  JDIMENSION MCUs_per_row;	/* # of MCUs across the image */
+  JDIMENSION MCU_rows_in_scan;	/* # of MCU rows in the image */
+
+  int blocks_in_MCU;		/* # of DCT blocks per MCU */
+  int MCU_membership[C_MAX_BLOCKS_IN_MCU];
+  /* MCU_membership[i] is index in cur_comp_info of component owning */
+  /* i'th block in an MCU */
+
+  int Ss, Se, Ah, Al;		/* progressive JPEG parameters for scan */
+
+  /*
+   * Links to compression subobjects (methods and private variables of modules)
+   */
+  struct jpeg_comp_master * master;
+  struct jpeg_c_main_controller * main;
+  struct jpeg_c_prep_controller * prep;
+  struct jpeg_c_coef_controller * coef;
+  struct jpeg_marker_writer * marker;
+  struct jpeg_color_converter * cconvert;
+  struct jpeg_downsampler * downsample;
+  struct jpeg_forward_dct * fdct;
+  struct jpeg_entropy_encoder * entropy;
+  jpeg_scan_info * script_space; /* workspace for jpeg_simple_progression */
+  int script_space_size;
+};
+
+
+/* Master record for a decompression instance */
+
+struct jpeg_decompress_struct {
+  jpeg_common_fields;		/* Fields shared with jpeg_compress_struct */
+
+  /* Source of compressed data */
+  struct jpeg_source_mgr * src;
+
+  /* Basic description of image --- filled in by jpeg_read_header(). */
+  /* Application may inspect these values to decide how to process image. */
+
+  JDIMENSION original_image_width;	/* nominal image width (from SOF marker) */
+
+  JDIMENSION image_width;	/* nominal image width (from SOF marker)
+                               may be changed by tile decode */
+  JDIMENSION image_height;	/* nominal image height */
+  int num_components;		/* # of color components in JPEG image */
+  J_COLOR_SPACE jpeg_color_space; /* colorspace of JPEG image */
+
+  /* Decompression processing parameters --- these fields must be set before
+   * calling jpeg_start_decompress().  Note that jpeg_read_header() initializes
+   * them to default values.
+   */
+
+  J_COLOR_SPACE out_color_space; /* colorspace for output */
+
+  unsigned int scale_num, scale_denom; /* fraction by which to scale image */
+
+  double output_gamma;		/* image gamma wanted in output */
+
+  boolean buffered_image;	/* TRUE=multiple output passes */
+  boolean raw_data_out;		/* TRUE=downsampled data wanted */
+
+  J_DCT_METHOD dct_method;	/* IDCT algorithm selector */
+  boolean do_fancy_upsampling;	/* TRUE=apply fancy upsampling */
+  boolean do_block_smoothing;	/* TRUE=apply interblock smoothing */
+
+  boolean quantize_colors;	/* TRUE=colormapped output wanted */
+  /* the following are ignored if not quantize_colors: */
+  J_DITHER_MODE dither_mode;	/* type of color dithering to use */
+  boolean two_pass_quantize;	/* TRUE=use two-pass color quantization */
+  int desired_number_of_colors;	/* max # colors to use in created colormap */
+  /* these are significant only in buffered-image mode: */
+  boolean enable_1pass_quant;	/* enable future use of 1-pass quantizer */
+  boolean enable_external_quant;/* enable future use of external colormap */
+  boolean enable_2pass_quant;	/* enable future use of 2-pass quantizer */
+
+  /* Description of actual output image that will be returned to application.
+   * These fields are computed by jpeg_start_decompress().
+   * You can also use jpeg_calc_output_dimensions() to determine these values
+   * in advance of calling jpeg_start_decompress().
+   */
+
+  JDIMENSION output_width;	/* scaled image width */
+  JDIMENSION output_height;	/* scaled image height */
+  int out_color_components;	/* # of color components in out_color_space */
+  int output_components;	/* # of color components returned */
+  /* output_components is 1 (a colormap index) when quantizing colors;
+   * otherwise it equals out_color_components.
+   */
+  int rec_outbuf_height;	/* min recommended height of scanline buffer */
+  /* If the buffer passed to jpeg_read_scanlines() is less than this many rows
+   * high, space and time will be wasted due to unnecessary data copying.
+   * Usually rec_outbuf_height will be 1 or 2, at most 4.
+   */
+
+  /* When quantizing colors, the output colormap is described by these fields.
+   * The application can supply a colormap by setting colormap non-NULL before
+   * calling jpeg_start_decompress; otherwise a colormap is created during
+   * jpeg_start_decompress or jpeg_start_output.
+   * The map has out_color_components rows and actual_number_of_colors columns.
+   */
+  int actual_number_of_colors;	/* number of entries in use */
+  JSAMPARRAY colormap;		/* The color map as a 2-D pixel array */
+
+  /* State variables: these variables indicate the progress of decompression.
+   * The application may examine these but must not modify them.
+   */
+
+  /* Row index of next scanline to be read from jpeg_read_scanlines().
+   * Application may use this to control its processing loop, e.g.,
+   * "while (output_scanline < output_height)".
+   */
+  JDIMENSION output_scanline;	/* 0 .. output_height-1  */
+
+  /* Current input scan number and number of iMCU rows completed in scan.
+   * These indicate the progress of the decompressor input side.
+   */
+  int input_scan_number;	/* Number of SOS markers seen so far */
+  JDIMENSION input_iMCU_row;	/* Number of iMCU rows completed */
+
+  /* The "output scan number" is the notional scan being displayed by the
+   * output side.  The decompressor will not allow output scan/row number
+   * to get ahead of input scan/row, but it can fall arbitrarily far behind.
+   */
+  int output_scan_number;	/* Nominal scan number being displayed */
+  JDIMENSION output_iMCU_row;	/* Number of iMCU rows read */
+
+  /* Current progression status.  coef_bits[c][i] indicates the precision
+   * with which component c's DCT coefficient i (in zigzag order) is known.
+   * It is -1 when no data has yet been received, otherwise it is the point
+   * transform (shift) value for the most recent scan of the coefficient
+   * (thus, 0 at completion of the progression).
+   * This pointer is NULL when reading a non-progressive file.
+   */
+  int (*coef_bits)[DCTSIZE2];	/* -1 or current Al value for each coef */
+
+  /* Internal JPEG parameters --- the application usually need not look at
+   * these fields.  Note that the decompressor output side may not use
+   * any parameters that can change between scans.
+   */
+
+  /* Quantization and Huffman tables are carried forward across input
+   * datastreams when processing abbreviated JPEG datastreams.
+   */
+
+  JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS];
+  /* ptrs to coefficient quantization tables, or NULL if not defined */
+
+  JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  /* ptrs to Huffman coding tables, or NULL if not defined */
+
+  /* These parameters are never carried across datastreams, since they
+   * are given in SOF/SOS markers or defined to be reset by SOI.
+   */
+
+  int data_precision;		/* bits of precision in image data */
+
+  jpeg_component_info * comp_info;
+  /* comp_info[i] describes component that appears i'th in SOF */
+
+  boolean tile_decode;         /* TRUE if using tile based decoding */
+  boolean progressive_mode;	/* TRUE if SOFn specifies progressive mode */
+  boolean arith_code;		/* TRUE=arithmetic coding, FALSE=Huffman */
+
+  UINT8 arith_dc_L[NUM_ARITH_TBLS]; /* L values for DC arith-coding tables */
+  UINT8 arith_dc_U[NUM_ARITH_TBLS]; /* U values for DC arith-coding tables */
+  UINT8 arith_ac_K[NUM_ARITH_TBLS]; /* Kx values for AC arith-coding tables */
+
+  unsigned int restart_interval; /* MCUs per restart interval, or 0 for no restart */
+
+  /* These fields record data obtained from optional markers recognized by
+   * the JPEG library.
+   */
+  boolean saw_JFIF_marker;	/* TRUE iff a JFIF APP0 marker was found */
+  /* Data copied from JFIF marker; only valid if saw_JFIF_marker is TRUE: */
+  UINT8 JFIF_major_version;	/* JFIF version number */
+  UINT8 JFIF_minor_version;
+  UINT8 density_unit;		/* JFIF code for pixel size units */
+  UINT16 X_density;		/* Horizontal pixel density */
+  UINT16 Y_density;		/* Vertical pixel density */
+  boolean saw_Adobe_marker;	/* TRUE iff an Adobe APP14 marker was found */
+  UINT8 Adobe_transform;	/* Color transform code from Adobe marker */
+
+  boolean CCIR601_sampling;	/* TRUE=first samples are cosited */
+
+  /* Aside from the specific data retained from APPn markers known to the
+   * library, the uninterpreted contents of any or all APPn and COM markers
+   * can be saved in a list for examination by the application.
+   */
+  jpeg_saved_marker_ptr marker_list; /* Head of list of saved markers */
+
+  /* Remaining fields are known throughout decompressor, but generally
+   * should not be touched by a surrounding application.
+   */
+
+  /*
+   * These fields are computed during decompression startup
+   */
+  int max_h_samp_factor;	/* largest h_samp_factor */
+  int max_v_samp_factor;	/* largest v_samp_factor */
+
+  int min_DCT_scaled_size;	/* smallest DCT_scaled_size of any component */
+
+  JDIMENSION total_iMCU_rows;	/* # of iMCU rows in image */
+  /* The coefficient controller's input and output progress is measured in
+   * units of "iMCU" (interleaved MCU) rows.  These are the same as MCU rows
+   * in fully interleaved JPEG scans, but are used whether the scan is
+   * interleaved or not.  We define an iMCU row as v_samp_factor DCT block
+   * rows of each component.  Therefore, the IDCT output contains
+   * v_samp_factor*DCT_scaled_size sample rows of a component per iMCU row.
+   */
+
+  JSAMPLE * sample_range_limit; /* table for fast range-limiting */
+
+  /*
+   * These fields are valid during any one scan.
+   * They describe the components and MCUs actually appearing in the scan.
+   * Note that the decompressor output side must not use these fields.
+   */
+  int comps_in_scan;		/* # of JPEG components in this scan */
+  jpeg_component_info * cur_comp_info[MAX_COMPS_IN_SCAN];
+  /* *cur_comp_info[i] describes component that appears i'th in SOS */
+
+  JDIMENSION MCUs_per_row;	/* # of MCUs across the image */
+  JDIMENSION MCU_rows_in_scan;	/* # of MCU rows in the image */
+
+  int blocks_in_MCU;		/* # of DCT blocks per MCU */
+  int MCU_membership[D_MAX_BLOCKS_IN_MCU];
+  /* MCU_membership[i] is index in cur_comp_info of component owning */
+  /* i'th block in an MCU */
+
+  int Ss, Se, Ah, Al;		/* progressive JPEG parameters for scan */
+
+  /* This field is shared between entropy decoder and marker parser.
+   * It is either zero or the code of a JPEG marker that has been
+   * read from the data source, but has not yet been processed.
+   */
+  int unread_marker;
+
+  /*
+   * Links to decompression subobjects (methods, private variables of modules)
+   */
+  struct jpeg_decomp_master * master;
+  struct jpeg_d_main_controller * main;
+  struct jpeg_d_coef_controller * coef;
+  struct jpeg_d_post_controller * post;
+  struct jpeg_input_controller * inputctl;
+  struct jpeg_marker_reader * marker;
+  struct jpeg_entropy_decoder * entropy;
+  struct jpeg_inverse_dct * idct;
+  struct jpeg_upsampler * upsample;
+  struct jpeg_color_deconverter * cconvert;
+  struct jpeg_color_quantizer * cquantize;
+};
+
+typedef struct {
+
+  // |--- byte_offset ---|- bit_left -|
+  //  \------ 27 -------/ \---- 5 ----/
+  unsigned int bitstream_offset;
+  short prev_dc[3];
+
+  // remaining EOBs in EOBRUN
+  unsigned short EOBRUN;
+
+  // save the decoder current bit buffer, entropy->bitstate.get_buffer.
+  INT32 get_buffer;
+
+  // save the restart info.
+  unsigned short restarts_to_go;
+  unsigned char next_restart_num;
+} huffman_offset_data;
+
+typedef struct {
+
+  // The header starting position of this scan
+  unsigned int bitstream_offset;
+
+  // Number of components in this scan
+  int comps_in_scan;
+
+  // Number of MCUs in each row
+  int MCUs_per_row;
+  int MCU_rows_per_iMCU_row;
+
+  // The last MCU position and its dc value in this scan
+  huffman_offset_data prev_MCU_offset;
+
+  huffman_offset_data **offset;
+} huffman_scan_header;
+
+#define DEFAULT_MCU_SAMPLE_SIZE 16
+
+typedef struct {
+
+  // The number of MCUs that we sample each time as an index point
+  int MCU_sample_size;
+
+  // Number of scan in this image
+  int scan_count;
+
+  // Number of iMCUs rows in this image
+  int total_iMCU_rows;
+
+  // Memory used by scan struct
+  size_t mem_used;
+  huffman_scan_header *scan;
+} huffman_index;
+
+/* "Object" declarations for JPEG modules that may be supplied or called
+ * directly by the surrounding application.
+ * As with all objects in the JPEG library, these structs only define the
+ * publicly visible methods and state variables of a module.  Additional
+ * private fields may exist after the public ones.
+ */
+
+
+/* Error handler object */
+
+struct jpeg_error_mgr {
+  /* Error exit handler: does not return to caller */
+  JMETHOD(void, error_exit, (j_common_ptr cinfo));
+  /* Conditionally emit a trace or warning message */
+  JMETHOD(void, emit_message, (j_common_ptr cinfo, int msg_level));
+  /* Routine that actually outputs a trace or error message */
+  JMETHOD(void, output_message, (j_common_ptr cinfo));
+  /* Format a message string for the most recent JPEG error or message */
+  JMETHOD(void, format_message, (j_common_ptr cinfo, char * buffer));
+#define JMSG_LENGTH_MAX  200	/* recommended size of format_message buffer */
+  /* Reset error state variables at start of a new image */
+  JMETHOD(void, reset_error_mgr, (j_common_ptr cinfo));
+
+  /* The message ID code and any parameters are saved here.
+   * A message can have one string parameter or up to 8 int parameters.
+   */
+  int msg_code;
+#define JMSG_STR_PARM_MAX  80
+  union {
+    int i[8];
+    char s[JMSG_STR_PARM_MAX];
+  } msg_parm;
+
+  /* Standard state variables for error facility */
+
+  int trace_level;		/* max msg_level that will be displayed */
+
+  /* For recoverable corrupt-data errors, we emit a warning message,
+   * but keep going unless emit_message chooses to abort.  emit_message
+   * should count warnings in num_warnings.  The surrounding application
+   * can check for bad data by seeing if num_warnings is nonzero at the
+   * end of processing.
+   */
+  long num_warnings;		/* number of corrupt-data warnings */
+
+  /* These fields point to the table(s) of error message strings.
+   * An application can change the table pointer to switch to a different
+   * message list (typically, to change the language in which errors are
+   * reported).  Some applications may wish to add additional error codes
+   * that will be handled by the JPEG library error mechanism; the second
+   * table pointer is used for this purpose.
+   *
+   * First table includes all errors generated by JPEG library itself.
+   * Error code 0 is reserved for a "no such error string" message.
+   */
+  const char * const * jpeg_message_table; /* Library errors */
+  int last_jpeg_message;    /* Table contains strings 0..last_jpeg_message */
+  /* Second table can be added by application (see cjpeg/djpeg for example).
+   * It contains strings numbered first_addon_message..last_addon_message.
+   */
+  const char * const * addon_message_table; /* Non-library errors */
+  int first_addon_message;	/* code for first string in addon table */
+  int last_addon_message;	/* code for last string in addon table */
+};
+
+
+/* Progress monitor object */
+
+struct jpeg_progress_mgr {
+  JMETHOD(void, progress_monitor, (j_common_ptr cinfo));
+
+  long pass_counter;		/* work units completed in this pass */
+  long pass_limit;		/* total number of work units in this pass */
+  int completed_passes;		/* passes completed so far */
+  int total_passes;		/* total number of passes expected */
+};
+
+
+/* Data destination object for compression */
+
+struct jpeg_destination_mgr {
+  JOCTET * next_output_byte;	/* => next byte to write in buffer */
+  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
+
+  JMETHOD(void, init_destination, (j_compress_ptr cinfo));
+  JMETHOD(boolean, empty_output_buffer, (j_compress_ptr cinfo));
+  JMETHOD(void, term_destination, (j_compress_ptr cinfo));
+};
+
+
+/* Data source object for decompression */
+
+struct jpeg_source_mgr {
+  const JOCTET * next_input_byte; /* => next byte to read from buffer */
+  const JOCTET * start_input_byte; /* => first byte to read from input */
+  size_t bytes_in_buffer;	/* # of bytes remaining in buffer */
+  size_t current_offset; /* current readed input offset */
+
+  JMETHOD(void, init_source, (j_decompress_ptr cinfo));
+  JMETHOD(boolean, fill_input_buffer, (j_decompress_ptr cinfo));
+  JMETHOD(void, skip_input_data, (j_decompress_ptr cinfo, long num_bytes));
+  JMETHOD(boolean, resync_to_restart, (j_decompress_ptr cinfo, int desired));
+  JMETHOD(void, term_source, (j_decompress_ptr cinfo));
+  JMETHOD(boolean, seek_input_data, (j_decompress_ptr cinfo, long byte_offset));
+};
+
+
+/* Memory manager object.
+ * Allocates "small" objects (a few K total), "large" objects (tens of K),
+ * and "really big" objects (virtual arrays with backing store if needed).
+ * The memory manager does not allow individual objects to be freed; rather,
+ * each created object is assigned to a pool, and whole pools can be freed
+ * at once.  This is faster and more convenient than remembering exactly what
+ * to free, especially where malloc()/free() are not too speedy.
+ * NB: alloc routines never return NULL.  They exit to error_exit if not
+ * successful.
+ */
+
+#define JPOOL_PERMANENT	0	/* lasts until master record is destroyed */
+#define JPOOL_IMAGE	1	/* lasts until done with image/datastream */
+#define JPOOL_NUMPOOLS	2
+
+typedef struct jvirt_sarray_control * jvirt_sarray_ptr;
+typedef struct jvirt_barray_control * jvirt_barray_ptr;
+
+
+struct jpeg_memory_mgr {
+  /* Method pointers */
+  JMETHOD(void *, alloc_small, (j_common_ptr cinfo, int pool_id,
+				size_t sizeofobject));
+  JMETHOD(void FAR *, alloc_large, (j_common_ptr cinfo, int pool_id,
+				     size_t sizeofobject));
+  JMETHOD(JSAMPARRAY, alloc_sarray, (j_common_ptr cinfo, int pool_id,
+				     JDIMENSION samplesperrow,
+				     JDIMENSION numrows));
+  JMETHOD(JBLOCKARRAY, alloc_barray, (j_common_ptr cinfo, int pool_id,
+				      JDIMENSION blocksperrow,
+				      JDIMENSION numrows));
+  JMETHOD(jvirt_sarray_ptr, request_virt_sarray, (j_common_ptr cinfo,
+						  int pool_id,
+						  boolean pre_zero,
+						  JDIMENSION samplesperrow,
+						  JDIMENSION numrows,
+						  JDIMENSION maxaccess));
+  JMETHOD(jvirt_barray_ptr, request_virt_barray, (j_common_ptr cinfo,
+						  int pool_id,
+						  boolean pre_zero,
+						  JDIMENSION blocksperrow,
+						  JDIMENSION numrows,
+						  JDIMENSION maxaccess));
+  JMETHOD(void, realize_virt_arrays, (j_common_ptr cinfo));
+  JMETHOD(JSAMPARRAY, access_virt_sarray, (j_common_ptr cinfo,
+					   jvirt_sarray_ptr ptr,
+					   JDIMENSION start_row,
+					   JDIMENSION num_rows,
+					   boolean writable));
+  JMETHOD(JBLOCKARRAY, access_virt_barray, (j_common_ptr cinfo,
+					    jvirt_barray_ptr ptr,
+					    JDIMENSION start_row,
+					    JDIMENSION num_rows,
+					    boolean writable));
+  JMETHOD(void, free_pool, (j_common_ptr cinfo, int pool_id));
+  JMETHOD(void, self_destruct, (j_common_ptr cinfo));
+
+  /* Limit on memory allocation for this JPEG object.  (Note that this is
+   * merely advisory, not a guaranteed maximum; it only affects the space
+   * used for virtual-array buffers.)  May be changed by outer application
+   * after creating the JPEG object.
+   */
+  long max_memory_to_use;
+
+  /* Maximum allocation request accepted by alloc_large. */
+  long max_alloc_chunk;
+};
+
+
+/* Routine signature for application-supplied marker processing methods.
+ * Need not pass marker code since it is stored in cinfo->unread_marker.
+ */
+typedef JMETHOD(boolean, jpeg_marker_parser_method, (j_decompress_ptr cinfo));
+
+
+/* Declarations for routines called by application.
+ * The JPP macro hides prototype parameters from compilers that can't cope.
+ * Note JPP requires double parentheses.
+ */
+
+#ifdef HAVE_PROTOTYPES
+#define JPP(arglist)	arglist
+#else
+#define JPP(arglist)	()
+#endif
+
+
+/* Short forms of external names for systems with brain-damaged linkers.
+ * We shorten external names to be unique in the first six letters, which
+ * is good enough for all known systems.
+ * (If your compiler itself needs names to be unique in less than 15
+ * characters, you are out of luck.  Get a better compiler.)
+ */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jpeg_std_error		jStdError
+#define jpeg_CreateCompress	jCreaCompress
+#define jpeg_CreateDecompress	jCreaDecompress
+#define jpeg_destroy_compress	jDestCompress
+#define jpeg_destroy_decompress	jDestDecompress
+#define jpeg_stdio_dest		jStdDest
+#define jpeg_stdio_src		jStdSrc
+#define jpeg_set_defaults	jSetDefaults
+#define jpeg_set_colorspace	jSetColorspace
+#define jpeg_default_colorspace	jDefColorspace
+#define jpeg_set_quality	jSetQuality
+#define jpeg_set_linear_quality	jSetLQuality
+#define jpeg_add_quant_table	jAddQuantTable
+#define jpeg_quality_scaling	jQualityScaling
+#define jpeg_simple_progression	jSimProgress
+#define jpeg_suppress_tables	jSuppressTables
+#define jpeg_alloc_quant_table	jAlcQTable
+#define jpeg_alloc_huff_table	jAlcHTable
+#define jpeg_start_compress	jStrtCompress
+#define jpeg_write_scanlines	jWrtScanlines
+#define jpeg_finish_compress	jFinCompress
+#define jpeg_write_raw_data	jWrtRawData
+#define jpeg_write_marker	jWrtMarker
+#define jpeg_write_m_header	jWrtMHeader
+#define jpeg_write_m_byte	jWrtMByte
+#define jpeg_write_tables	jWrtTables
+#define jpeg_read_header	jReadHeader
+#define jpeg_start_decompress	jStrtDecompress
+#define jpeg_read_scanlines	jReadScanlines
+#define jpeg_finish_decompress	jFinDecompress
+#define jpeg_read_raw_data	jReadRawData
+#define jpeg_has_multiple_scans	jHasMultScn
+#define jpeg_start_output	jStrtOutput
+#define jpeg_finish_output	jFinOutput
+#define jpeg_input_complete	jInComplete
+#define jpeg_new_colormap	jNewCMap
+#define jpeg_consume_input	jConsumeInput
+#define jpeg_calc_output_dimensions	jCalcDimensions
+#define jpeg_save_markers	jSaveMarkers
+#define jpeg_set_marker_processor	jSetMarker
+#define jpeg_read_coefficients	jReadCoefs
+#define jpeg_write_coefficients	jWrtCoefs
+#define jpeg_copy_critical_parameters	jCopyCrit
+#define jpeg_abort_compress	jAbrtCompress
+#define jpeg_abort_decompress	jAbrtDecompress
+#define jpeg_abort		jAbort
+#define jpeg_destroy		jDestroy
+#define jpeg_resync_to_restart	jResyncRestart
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+
+/* Default error-management setup */
+EXTERN(struct jpeg_error_mgr *) jpeg_std_error
+	JPP((struct jpeg_error_mgr * err));
+
+/* Initialization of JPEG compression objects.
+ * jpeg_create_compress() and jpeg_create_decompress() are the exported
+ * names that applications should call.  These expand to calls on
+ * jpeg_CreateCompress and jpeg_CreateDecompress with additional information
+ * passed for version mismatch checking.
+ * NB: you must set up the error-manager BEFORE calling jpeg_create_xxx.
+ */
+#define jpeg_create_compress(cinfo) \
+    jpeg_CreateCompress((cinfo), JPEG_LIB_VERSION, \
+			(size_t) sizeof(struct jpeg_compress_struct))
+#define jpeg_create_decompress(cinfo) \
+    jpeg_CreateDecompress((cinfo), JPEG_LIB_VERSION, \
+			  (size_t) sizeof(struct jpeg_decompress_struct))
+EXTERN(void) jpeg_CreateCompress JPP((j_compress_ptr cinfo,
+				      int version, size_t structsize));
+EXTERN(void) jpeg_CreateDecompress JPP((j_decompress_ptr cinfo,
+					int version, size_t structsize));
+/* Destruction of JPEG compression objects */
+EXTERN(void) jpeg_destroy_compress JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_destroy_decompress JPP((j_decompress_ptr cinfo));
+
+/* Standard data source and destination managers: stdio streams. */
+/* Caller is responsible for opening the file before and closing after. */
+EXTERN(void) jpeg_stdio_dest JPP((j_compress_ptr cinfo, FILE * outfile));
+EXTERN(void) jpeg_stdio_src JPP((j_decompress_ptr cinfo, FILE * infile));
+
+/* Default parameter setup for compression */
+EXTERN(void) jpeg_set_defaults JPP((j_compress_ptr cinfo));
+/* Compression parameter setup aids */
+EXTERN(void) jpeg_set_colorspace JPP((j_compress_ptr cinfo,
+				      J_COLOR_SPACE colorspace));
+EXTERN(void) jpeg_default_colorspace JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_set_quality JPP((j_compress_ptr cinfo, int quality,
+				   boolean force_baseline));
+EXTERN(void) jpeg_set_linear_quality JPP((j_compress_ptr cinfo,
+					  int scale_factor,
+					  boolean force_baseline));
+EXTERN(void) jpeg_add_quant_table JPP((j_compress_ptr cinfo, int which_tbl,
+				       const unsigned int *basic_table,
+				       int scale_factor,
+				       boolean force_baseline));
+EXTERN(int) jpeg_quality_scaling JPP((int quality));
+EXTERN(void) jpeg_simple_progression JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_suppress_tables JPP((j_compress_ptr cinfo,
+				       boolean suppress));
+EXTERN(JQUANT_TBL *) jpeg_alloc_quant_table JPP((j_common_ptr cinfo));
+EXTERN(JHUFF_TBL *) jpeg_alloc_huff_table JPP((j_common_ptr cinfo));
+
+/* Main entry points for compression */
+EXTERN(void) jpeg_start_compress JPP((j_compress_ptr cinfo,
+				      boolean write_all_tables));
+EXTERN(JDIMENSION) jpeg_write_scanlines JPP((j_compress_ptr cinfo,
+					     JSAMPARRAY scanlines,
+					     JDIMENSION num_lines));
+EXTERN(void) jpeg_finish_compress JPP((j_compress_ptr cinfo));
+
+/* Replaces jpeg_write_scanlines when writing raw downsampled data. */
+EXTERN(JDIMENSION) jpeg_write_raw_data JPP((j_compress_ptr cinfo,
+					    JSAMPIMAGE data,
+					    JDIMENSION num_lines));
+
+/* Write a special marker.  See libjpeg.doc concerning safe usage. */
+EXTERN(void) jpeg_write_marker
+	JPP((j_compress_ptr cinfo, int marker,
+	     const JOCTET * dataptr, unsigned int datalen));
+/* Same, but piecemeal. */
+EXTERN(void) jpeg_write_m_header
+	JPP((j_compress_ptr cinfo, int marker, unsigned int datalen));
+EXTERN(void) jpeg_write_m_byte
+	JPP((j_compress_ptr cinfo, int val));
+
+/* Alternate compression function: just write an abbreviated table file */
+EXTERN(void) jpeg_write_tables JPP((j_compress_ptr cinfo));
+
+/* Decompression startup: read start of JPEG datastream to see what's there */
+EXTERN(int) jpeg_read_header JPP((j_decompress_ptr cinfo,
+				  boolean require_image));
+/* Return value is one of: */
+#define JPEG_SUSPENDED		0 /* Suspended due to lack of input data */
+#define JPEG_HEADER_OK		1 /* Found valid image datastream */
+#define JPEG_HEADER_TABLES_ONLY	2 /* Found valid table-specs-only datastream */
+/* If you pass require_image = TRUE (normal case), you need not check for
+ * a TABLES_ONLY return code; an abbreviated file will cause an error exit.
+ * JPEG_SUSPENDED is only possible if you use a data source module that can
+ * give a suspension return (the stdio source module doesn't).
+ */
+
+/* Main entry points for decompression */
+EXTERN(boolean) jpeg_start_decompress JPP((j_decompress_ptr cinfo));
+EXTERN(boolean) jpeg_start_tile_decompress JPP((j_decompress_ptr cinfo));
+EXTERN(JDIMENSION) jpeg_read_scanlines JPP((j_decompress_ptr cinfo,
+					    JSAMPARRAY scanlines,
+					    JDIMENSION max_lines));
+EXTERN(JDIMENSION) jpeg_read_scanlines_from JPP((j_decompress_ptr cinfo,
+					    JSAMPARRAY scanlines,
+					    int line_offset,
+					    JDIMENSION max_lines));
+EXTERN(JDIMENSION) jpeg_read_tile_scanline JPP((j_decompress_ptr cinfo,
+                        huffman_index *index,
+                        JSAMPARRAY scanlines));
+EXTERN(void) jpeg_init_read_tile_scanline JPP((j_decompress_ptr cinfo,
+                        huffman_index *index,
+		                int *start_x, int *start_y,
+                        int *width, int *height));
+EXTERN(boolean) jpeg_finish_decompress JPP((j_decompress_ptr cinfo));
+
+/* Replaces jpeg_read_scanlines when reading raw downsampled data. */
+EXTERN(JDIMENSION) jpeg_read_raw_data JPP((j_decompress_ptr cinfo,
+					   JSAMPIMAGE data,
+					   JDIMENSION max_lines));
+
+/* Additional entry points for buffered-image mode. */
+EXTERN(boolean) jpeg_has_multiple_scans JPP((j_decompress_ptr cinfo));
+EXTERN(boolean) jpeg_start_output JPP((j_decompress_ptr cinfo,
+				       int scan_number));
+EXTERN(boolean) jpeg_finish_output JPP((j_decompress_ptr cinfo));
+EXTERN(boolean) jpeg_input_complete JPP((j_decompress_ptr cinfo));
+EXTERN(void) jpeg_new_colormap JPP((j_decompress_ptr cinfo));
+EXTERN(int) jpeg_consume_input JPP((j_decompress_ptr cinfo));
+/* Return value is one of: */
+/* #define JPEG_SUSPENDED	0    Suspended due to lack of input data */
+#define JPEG_REACHED_SOS	1 /* Reached start of new scan */
+#define JPEG_REACHED_EOI	2 /* Reached end of image */
+#define JPEG_ROW_COMPLETED	3 /* Completed one iMCU row */
+#define JPEG_SCAN_COMPLETED	4 /* Completed last iMCU row of a scan */
+
+/* Precalculate output dimensions for current decompression parameters. */
+EXTERN(void) jpeg_calc_output_dimensions JPP((j_decompress_ptr cinfo));
+
+/* Control saving of COM and APPn markers into marker_list. */
+EXTERN(void) jpeg_save_markers
+	JPP((j_decompress_ptr cinfo, int marker_code,
+	     unsigned int length_limit));
+
+/* Install a special processing method for COM or APPn markers. */
+EXTERN(void) jpeg_set_marker_processor
+	JPP((j_decompress_ptr cinfo, int marker_code,
+	     jpeg_marker_parser_method routine));
+
+/* Read or write raw DCT coefficients --- useful for lossless transcoding. */
+EXTERN(jvirt_barray_ptr *) jpeg_read_coefficients JPP((j_decompress_ptr cinfo));
+EXTERN(boolean) jpeg_build_huffman_index
+    JPP((j_decompress_ptr cinfo, huffman_index *index));
+EXTERN(void) jpeg_write_coefficients JPP((j_compress_ptr cinfo,
+					  jvirt_barray_ptr * coef_arrays));
+EXTERN(void) jpeg_copy_critical_parameters JPP((j_decompress_ptr srcinfo,
+						j_compress_ptr dstinfo));
+
+/* If you choose to abort compression or decompression before completing
+ * jpeg_finish_(de)compress, then you need to clean up to release memory,
+ * temporary files, etc.  You can just call jpeg_destroy_(de)compress
+ * if you're done with the JPEG object, but if you want to clean it up and
+ * reuse it, call this:
+ */
+EXTERN(void) jpeg_abort_compress JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_abort_decompress JPP((j_decompress_ptr cinfo));
+
+/* Generic versions of jpeg_abort and jpeg_destroy that work on either
+ * flavor of JPEG object.  These may be more convenient in some places.
+ */
+EXTERN(void) jpeg_abort JPP((j_common_ptr cinfo));
+EXTERN(void) jpeg_destroy JPP((j_common_ptr cinfo));
+
+/* Default restart-marker-resync procedure for use by data source modules */
+EXTERN(boolean) jpeg_resync_to_restart JPP((j_decompress_ptr cinfo,
+					    int desired));
+
+EXTERN(void) jpeg_configure_huffman_decoder(j_decompress_ptr cinfo,
+                        huffman_offset_data offset);
+EXTERN(void) jpeg_get_huffman_decoder_configuration(j_decompress_ptr cinfo,
+                        huffman_offset_data *offset);
+EXTERN(void) jpeg_create_huffman_index(j_decompress_ptr cinfo,
+                        huffman_index *index);
+EXTERN(void) jpeg_configure_huffman_index_scan(j_decompress_ptr cinfo,
+                        huffman_index *index, int scan_no, int offset);
+EXTERN(void) jpeg_destroy_huffman_index(huffman_index *index);
+
+
+/* These marker codes are exported since applications and data source modules
+ * are likely to want to use them.
+ */
+
+#define JPEG_RST0	0xD0	/* RST0 marker code */
+#define JPEG_EOI	0xD9	/* EOI marker code */
+#define JPEG_APP0	0xE0	/* APP0 marker code */
+#define JPEG_COM	0xFE	/* COM marker code */
+
+
+/* If we have a brain-damaged compiler that emits warnings (or worse, errors)
+ * for structure definitions that are never filled in, keep it quiet by
+ * supplying dummy definitions for the various substructures.
+ */
+
+#ifdef INCOMPLETE_TYPES_BROKEN
+#ifndef JPEG_INTERNALS		/* will be defined in jpegint.h */
+struct jvirt_sarray_control { long dummy; };
+struct jvirt_barray_control { long dummy; };
+struct jpeg_comp_master { long dummy; };
+struct jpeg_c_main_controller { long dummy; };
+struct jpeg_c_prep_controller { long dummy; };
+struct jpeg_c_coef_controller { long dummy; };
+struct jpeg_marker_writer { long dummy; };
+struct jpeg_color_converter { long dummy; };
+struct jpeg_downsampler { long dummy; };
+struct jpeg_forward_dct { long dummy; };
+struct jpeg_entropy_encoder { long dummy; };
+struct jpeg_decomp_master { long dummy; };
+struct jpeg_d_main_controller { long dummy; };
+struct jpeg_d_coef_controller { long dummy; };
+struct jpeg_d_post_controller { long dummy; };
+struct jpeg_input_controller { long dummy; };
+struct jpeg_marker_reader { long dummy; };
+struct jpeg_entropy_decoder { long dummy; };
+struct jpeg_inverse_dct { long dummy; };
+struct jpeg_upsampler { long dummy; };
+struct jpeg_color_deconverter { long dummy; };
+struct jpeg_color_quantizer { long dummy; };
+#endif /* JPEG_INTERNALS */
+#endif /* INCOMPLETE_TYPES_BROKEN */
+
+
+/*
+ * The JPEG library modules define JPEG_INTERNALS before including this file.
+ * The internal structure declarations are read only when that is true.
+ * Applications using the library should not include jpegint.h, but may wish
+ * to include jerror.h.
+ */
+
+#ifdef JPEG_INTERNALS
+#include "jpegint.h"		/* fetch private declarations */
+#include "jerror.h"		/* fetch error codes too */
+#endif
+
+#endif /* JPEGLIB_H */
diff --git a/libjpegtwrp/jpegtran.1 b/libjpegtwrp/jpegtran.1
new file mode 100644
index 000000000..6de18e2af
--- /dev/null
+++ b/libjpegtwrp/jpegtran.1
@@ -0,0 +1,238 @@
+.TH JPEGTRAN 1 "3 August 1997"
+.SH NAME
+jpegtran \- lossless transformation of JPEG files
+.SH SYNOPSIS
+.B jpegtran
+[
+.I options
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B jpegtran
+performs various useful transformations of JPEG files.
+It can translate the coded representation from one variant of JPEG to another,
+for example from baseline JPEG to progressive JPEG or vice versa.  It can also
+perform some rearrangements of the image data, for example turning an image
+from landscape to portrait format by rotation.
+.PP
+.B jpegtran
+works by rearranging the compressed data (DCT coefficients), without
+ever fully decoding the image.  Therefore, its transformations are lossless:
+there is no image degradation at all, which would not be true if you used
+.B djpeg
+followed by
+.B cjpeg
+to accomplish the same conversion.  But by the same token,
+.B jpegtran
+cannot perform lossy operations such as changing the image quality.
+.PP
+.B jpegtran
+reads the named JPEG/JFIF file, or the standard input if no file is
+named, and produces a JPEG/JFIF file on the standard output.
+.SH OPTIONS
+All switch names may be abbreviated; for example,
+.B \-optimize
+may be written
+.B \-opt
+or
+.BR \-o .
+Upper and lower case are equivalent.
+British spellings are also accepted (e.g.,
+.BR \-optimise ),
+though for brevity these are not mentioned below.
+.PP
+To specify the coded JPEG representation used in the output file,
+.B jpegtran
+accepts a subset of the switches recognized by
+.BR cjpeg :
+.TP
+.B \-optimize
+Perform optimization of entropy encoding parameters.
+.TP
+.B \-progressive
+Create progressive JPEG file.
+.TP
+.BI \-restart " N"
+Emit a JPEG restart marker every N MCU rows, or every N MCU blocks if "B" is
+attached to the number.
+.TP
+.BI \-scans " file"
+Use the scan script given in the specified text file.
+.PP
+See
+.BR cjpeg (1)
+for more details about these switches.
+If you specify none of these switches, you get a plain baseline-JPEG output
+file.  The quality setting and so forth are determined by the input file.
+.PP
+The image can be losslessly transformed by giving one of these switches:
+.TP
+.B \-flip horizontal
+Mirror image horizontally (left-right).
+.TP
+.B \-flip vertical
+Mirror image vertically (top-bottom).
+.TP
+.B \-rotate 90
+Rotate image 90 degrees clockwise.
+.TP
+.B \-rotate 180
+Rotate image 180 degrees.
+.TP
+.B \-rotate 270
+Rotate image 270 degrees clockwise (or 90 ccw).
+.TP
+.B \-transpose
+Transpose image (across UL-to-LR axis).
+.TP
+.B \-transverse
+Transverse transpose (across UR-to-LL axis).
+.PP
+The transpose transformation has no restrictions regarding image dimensions.
+The other transformations operate rather oddly if the image dimensions are not
+a multiple of the iMCU size (usually 8 or 16 pixels), because they can only
+transform complete blocks of DCT coefficient data in the desired way.
+.PP
+.BR jpegtran 's
+default behavior when transforming an odd-size image is designed
+to preserve exact reversibility and mathematical consistency of the
+transformation set.  As stated, transpose is able to flip the entire image
+area.  Horizontal mirroring leaves any partial iMCU column at the right edge
+untouched, but is able to flip all rows of the image.  Similarly, vertical
+mirroring leaves any partial iMCU row at the bottom edge untouched, but is
+able to flip all columns.  The other transforms can be built up as sequences
+of transpose and flip operations; for consistency, their actions on edge
+pixels are defined to be the same as the end result of the corresponding
+transpose-and-flip sequence.
+.PP
+For practical use, you may prefer to discard any untransformable edge pixels
+rather than having a strange-looking strip along the right and/or bottom edges
+of a transformed image.  To do this, add the
+.B \-trim
+switch:
+.TP
+.B \-trim
+Drop non-transformable edge blocks.
+.PP
+Obviously, a transformation with
+.B \-trim
+is not reversible, so strictly speaking
+.B jpegtran
+with this switch is not lossless.  Also, the expected mathematical
+equivalences between the transformations no longer hold.  For example,
+.B \-rot 270 -trim
+trims only the bottom edge, but
+.B \-rot 90 -trim
+followed by
+.B \-rot 180 -trim
+trims both edges.
+.PP
+Another not-strictly-lossless transformation switch is:
+.TP
+.B \-grayscale
+Force grayscale output.
+.PP
+This option discards the chrominance channels if the input image is YCbCr
+(ie, a standard color JPEG), resulting in a grayscale JPEG file.  The
+luminance channel is preserved exactly, so this is a better method of reducing
+to grayscale than decompression, conversion, and recompression.  This switch
+is particularly handy for fixing a monochrome picture that was mistakenly
+encoded as a color JPEG.  (In such a case, the space savings from getting rid
+of the near-empty chroma channels won't be large; but the decoding time for
+a grayscale JPEG is substantially less than that for a color JPEG.)
+.PP
+.B jpegtran
+also recognizes these switches that control what to do with "extra" markers,
+such as comment blocks:
+.TP
+.B \-copy none
+Copy no extra markers from source file.  This setting suppresses all
+comments and other excess baggage present in the source file.
+.TP
+.B \-copy comments
+Copy only comment markers.  This setting copies comments from the source file,
+but discards any other inessential data.
+.TP
+.B \-copy all
+Copy all extra markers.  This setting preserves miscellaneous markers
+found in the source file, such as JFIF thumbnails and Photoshop settings.
+In some files these extra markers can be sizable.
+.PP
+The default behavior is
+.BR "\-copy comments" .
+(Note: in IJG releases v6 and v6a,
+.B jpegtran
+always did the equivalent of
+.BR "\-copy none" .)
+.PP
+Additional switches recognized by jpegtran are:
+.TP
+.BI \-maxmemory " N"
+Set limit for amount of memory to use in processing large images.  Value is
+in thousands of bytes, or millions of bytes if "M" is attached to the
+number.  For example,
+.B \-max 4m
+selects 4000000 bytes.  If more space is needed, temporary files will be used.
+.TP
+.BI \-outfile " name"
+Send output image to the named file, not to standard output.
+.TP
+.B \-verbose
+Enable debug printout.  More
+.BR \-v 's
+give more output.  Also, version information is printed at startup.
+.TP
+.B \-debug
+Same as
+.BR \-verbose .
+.SH EXAMPLES
+.LP
+This example converts a baseline JPEG file to progressive form:
+.IP
+.B jpegtran \-progressive
+.I foo.jpg
+.B >
+.I fooprog.jpg
+.PP
+This example rotates an image 90 degrees clockwise, discarding any
+unrotatable edge pixels:
+.IP
+.B jpegtran \-rot 90 -trim
+.I foo.jpg
+.B >
+.I foo90.jpg
+.SH ENVIRONMENT
+.TP
+.B JPEGMEM
+If this environment variable is set, its value is the default memory limit.
+The value is specified as described for the
+.B \-maxmemory
+switch.
+.B JPEGMEM
+overrides the default value specified when the program was compiled, and
+itself is overridden by an explicit
+.BR \-maxmemory .
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR djpeg (1),
+.BR rdjpgcom (1),
+.BR wrjpgcom (1)
+.br
+Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
+.SH AUTHOR
+Independent JPEG Group
+.SH BUGS
+Arithmetic coding is not supported for legal reasons.
+.PP
+The transform options can't transform odd-size images perfectly.  Use
+.B \-trim
+if you don't like the results without it.
+.PP
+The entire image is read into memory and then written out again, even in
+cases where this isn't really necessary.  Expect swapping on large images,
+especially when using the more complex transform options.
diff --git a/libjpegtwrp/jpegtran.c b/libjpegtwrp/jpegtran.c
new file mode 100644
index 000000000..20ef111b3
--- /dev/null
+++ b/libjpegtwrp/jpegtran.c
@@ -0,0 +1,504 @@
+/*
+ * jpegtran.c
+ *
+ * Copyright (C) 1995-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a command-line user interface for JPEG transcoding.
+ * It is very similar to cjpeg.c, but provides lossless transcoding between
+ * different JPEG file formats.  It also provides some lossless and sort-of-
+ * lossless transformations of JPEG data.
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+#include "transupp.h"		/* Support routines for jpegtran */
+#include "jversion.h"		/* for version message */
+
+#ifdef USE_CCOMMAND		/* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>		/* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>		/* Think declares it here */
+#endif
+#endif
+
+
+/*
+ * Argument-parsing code.
+ * The switch parser is designed to be useful with DOS-style command line
+ * syntax, ie, intermixed switches and file names, where only the switches
+ * to the left of a given file name affect processing of that file.
+ * The main program in this file doesn't actually use this capability...
+ */
+
+
+static const char * progname;	/* program name for error messages */
+static char * outfilename;	/* for -outfile switch */
+static JCOPY_OPTION copyoption;	/* -copy switch */
+static jpeg_transform_info transformoption; /* image transformation options */
+
+
+LOCAL(void)
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "usage: %s [switches] ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "[inputfile]\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -copy none     Copy no extra markers from source file\n");
+  fprintf(stderr, "  -copy comments Copy only comment markers (default)\n");
+  fprintf(stderr, "  -copy all      Copy all extra markers\n");
+#ifdef ENTROPY_OPT_SUPPORTED
+  fprintf(stderr, "  -optimize      Optimize Huffman table (smaller file, but slow compression)\n");
+#endif
+#ifdef C_PROGRESSIVE_SUPPORTED
+  fprintf(stderr, "  -progressive   Create progressive JPEG file\n");
+#endif
+#if TRANSFORMS_SUPPORTED
+  fprintf(stderr, "Switches for modifying the image:\n");
+  fprintf(stderr, "  -grayscale     Reduce to grayscale (omit color data)\n");
+  fprintf(stderr, "  -flip [horizontal|vertical]  Mirror image (left-right or top-bottom)\n");
+  fprintf(stderr, "  -rotate [90|180|270]         Rotate image (degrees clockwise)\n");
+  fprintf(stderr, "  -transpose     Transpose image\n");
+  fprintf(stderr, "  -transverse    Transverse transpose image\n");
+  fprintf(stderr, "  -trim          Drop non-transformable edge blocks\n");
+#endif /* TRANSFORMS_SUPPORTED */
+  fprintf(stderr, "Switches for advanced users:\n");
+  fprintf(stderr, "  -restart N     Set restart interval in rows, or in blocks with B\n");
+  fprintf(stderr, "  -maxmemory N   Maximum memory to use (in kbytes)\n");
+  fprintf(stderr, "  -outfile name  Specify name for output file\n");
+  fprintf(stderr, "  -verbose  or  -debug   Emit debug output\n");
+  fprintf(stderr, "Switches for wizards:\n");
+#ifdef C_ARITH_CODING_SUPPORTED
+  fprintf(stderr, "  -arithmetic    Use arithmetic coding\n");
+#endif
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+  fprintf(stderr, "  -scans file    Create multi-scan JPEG per script file\n");
+#endif
+  exit(EXIT_FAILURE);
+}
+
+
+LOCAL(void)
+select_transform (JXFORM_CODE transform)
+/* Silly little routine to detect multiple transform options,
+ * which we can't handle.
+ */
+{
+#if TRANSFORMS_SUPPORTED
+  if (transformoption.transform == JXFORM_NONE ||
+      transformoption.transform == transform) {
+    transformoption.transform = transform;
+  } else {
+    fprintf(stderr, "%s: can only do one image transformation at a time\n",
+	    progname);
+    usage();
+  }
+#else
+  fprintf(stderr, "%s: sorry, image transformation was not compiled\n",
+	  progname);
+  exit(EXIT_FAILURE);
+#endif
+}
+
+
+LOCAL(int)
+parse_switches (j_compress_ptr cinfo, int argc, char **argv,
+		int last_file_arg_seen, boolean for_real)
+/* Parse optional switches.
+ * Returns argv[] index of first file-name argument (== argc if none).
+ * Any file names with indexes <= last_file_arg_seen are ignored;
+ * they have presumably been processed in a previous iteration.
+ * (Pass 0 for last_file_arg_seen on the first or only iteration.)
+ * for_real is FALSE on the first (dummy) pass; we may skip any expensive
+ * processing.
+ */
+{
+  int argn;
+  char * arg;
+  boolean simple_progressive;
+  char * scansarg = NULL;	/* saves -scans parm if any */
+
+  /* Set up default JPEG parameters. */
+  simple_progressive = FALSE;
+  outfilename = NULL;
+  copyoption = JCOPYOPT_DEFAULT;
+  transformoption.transform = JXFORM_NONE;
+  transformoption.trim = FALSE;
+  transformoption.force_grayscale = FALSE;
+  cinfo->err->trace_level = 0;
+
+  /* Scan command line options, adjust parameters */
+
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (*arg != '-') {
+      /* Not a switch, must be a file name argument */
+      if (argn <= last_file_arg_seen) {
+	outfilename = NULL;	/* -outfile applies to just one input file */
+	continue;		/* ignore this name if previously processed */
+      }
+      break;			/* else done parsing switches */
+    }
+    arg++;			/* advance past switch marker character */
+
+    if (keymatch(arg, "arithmetic", 1)) {
+      /* Use arithmetic coding. */
+#ifdef C_ARITH_CODING_SUPPORTED
+      cinfo->arith_code = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, arithmetic coding not supported\n",
+	      progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "copy", 1)) {
+      /* Select which extra markers to copy. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (keymatch(argv[argn], "none", 1)) {
+	copyoption = JCOPYOPT_NONE;
+      } else if (keymatch(argv[argn], "comments", 1)) {
+	copyoption = JCOPYOPT_COMMENTS;
+      } else if (keymatch(argv[argn], "all", 1)) {
+	copyoption = JCOPYOPT_ALL;
+      } else
+	usage();
+
+    } else if (keymatch(arg, "debug", 1) || keymatch(arg, "verbose", 1)) {
+      /* Enable debug printouts. */
+      /* On first -d, print version identification */
+      static boolean printed_version = FALSE;
+
+      if (! printed_version) {
+	fprintf(stderr, "Independent JPEG Group's JPEGTRAN, version %s\n%s\n",
+		JVERSION, JCOPYRIGHT);
+	printed_version = TRUE;
+      }
+      cinfo->err->trace_level++;
+
+    } else if (keymatch(arg, "flip", 1)) {
+      /* Mirror left-right or top-bottom. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (keymatch(argv[argn], "horizontal", 1))
+	select_transform(JXFORM_FLIP_H);
+      else if (keymatch(argv[argn], "vertical", 1))
+	select_transform(JXFORM_FLIP_V);
+      else
+	usage();
+
+    } else if (keymatch(arg, "grayscale", 1) || keymatch(arg, "greyscale",1)) {
+      /* Force to grayscale. */
+#if TRANSFORMS_SUPPORTED
+      transformoption.force_grayscale = TRUE;
+#else
+      select_transform(JXFORM_NONE);	/* force an error */
+#endif
+
+    } else if (keymatch(arg, "maxmemory", 3)) {
+      /* Maximum memory in Kb (or Mb with 'm'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+	usage();
+      if (ch == 'm' || ch == 'M')
+	lval *= 1000L;
+      cinfo->mem->max_memory_to_use = lval * 1000L;
+
+    } else if (keymatch(arg, "optimize", 1) || keymatch(arg, "optimise", 1)) {
+      /* Enable entropy parm optimization. */
+#ifdef ENTROPY_OPT_SUPPORTED
+      cinfo->optimize_coding = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, entropy optimization was not compiled\n",
+	      progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "outfile", 4)) {
+      /* Set output file name. */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      outfilename = argv[argn];	/* save it away for later use */
+
+    } else if (keymatch(arg, "progressive", 1)) {
+      /* Select simple progressive mode. */
+#ifdef C_PROGRESSIVE_SUPPORTED
+      simple_progressive = TRUE;
+      /* We must postpone execution until num_components is known. */
+#else
+      fprintf(stderr, "%s: sorry, progressive output was not compiled\n",
+	      progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "restart", 1)) {
+      /* Restart interval in MCU rows (or in MCUs with 'b'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+	usage();
+      if (lval < 0 || lval > 65535L)
+	usage();
+      if (ch == 'b' || ch == 'B') {
+	cinfo->restart_interval = (unsigned int) lval;
+	cinfo->restart_in_rows = 0; /* else prior '-restart n' overrides me */
+      } else {
+	cinfo->restart_in_rows = (int) lval;
+	/* restart_interval will be computed during startup */
+      }
+
+    } else if (keymatch(arg, "rotate", 2)) {
+      /* Rotate 90, 180, or 270 degrees (measured clockwise). */
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      if (keymatch(argv[argn], "90", 2))
+	select_transform(JXFORM_ROT_90);
+      else if (keymatch(argv[argn], "180", 3))
+	select_transform(JXFORM_ROT_180);
+      else if (keymatch(argv[argn], "270", 3))
+	select_transform(JXFORM_ROT_270);
+      else
+	usage();
+
+    } else if (keymatch(arg, "scans", 1)) {
+      /* Set scan script. */
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+      if (++argn >= argc)	/* advance to next argument */
+	usage();
+      scansarg = argv[argn];
+      /* We must postpone reading the file in case -progressive appears. */
+#else
+      fprintf(stderr, "%s: sorry, multi-scan output was not compiled\n",
+	      progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "transpose", 1)) {
+      /* Transpose (across UL-to-LR axis). */
+      select_transform(JXFORM_TRANSPOSE);
+
+    } else if (keymatch(arg, "transverse", 6)) {
+      /* Transverse transpose (across UR-to-LL axis). */
+      select_transform(JXFORM_TRANSVERSE);
+
+    } else if (keymatch(arg, "trim", 3)) {
+      /* Trim off any partial edge MCUs that the transform can't handle. */
+      transformoption.trim = TRUE;
+
+    } else {
+      usage();			/* bogus switch */
+    }
+  }
+
+  /* Post-switch-scanning cleanup */
+
+  if (for_real) {
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+    if (simple_progressive)	/* process -progressive; -scans can override */
+      jpeg_simple_progression(cinfo);
+#endif
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+    if (scansarg != NULL)	/* process -scans if it was present */
+      if (! read_scan_script(cinfo, scansarg))
+	usage();
+#endif
+  }
+
+  return argn;			/* return index of next arg (file name) */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  struct jpeg_decompress_struct srcinfo;
+  struct jpeg_compress_struct dstinfo;
+  struct jpeg_error_mgr jsrcerr, jdsterr;
+#ifdef PROGRESS_REPORT
+  struct cdjpeg_progress_mgr progress;
+#endif
+  jvirt_barray_ptr * src_coef_arrays;
+  jvirt_barray_ptr * dst_coef_arrays;
+  int file_index;
+  FILE * input_file;
+  FILE * output_file;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "jpegtran";	/* in case C library doesn't provide it */
+
+  /* Initialize the JPEG decompression object with default error handling. */
+  srcinfo.err = jpeg_std_error(&jsrcerr);
+  jpeg_create_decompress(&srcinfo);
+  /* Initialize the JPEG compression object with default error handling. */
+  dstinfo.err = jpeg_std_error(&jdsterr);
+  jpeg_create_compress(&dstinfo);
+
+  /* Now safe to enable signal catcher.
+   * Note: we assume only the decompression object will have virtual arrays.
+   */
+#ifdef NEED_SIGNAL_CATCHER
+  enable_signal_catcher((j_common_ptr) &srcinfo);
+#endif
+
+  /* Scan command line to find file names.
+   * It is convenient to use just one switch-parsing routine, but the switch
+   * values read here are mostly ignored; we will rescan the switches after
+   * opening the input file.  Also note that most of the switches affect the
+   * destination JPEG object, so we parse into that and then copy over what
+   * needs to affects the source too.
+   */
+
+  file_index = parse_switches(&dstinfo, argc, argv, 0, FALSE);
+  jsrcerr.trace_level = jdsterr.trace_level;
+  srcinfo.mem->max_memory_to_use = dstinfo.mem->max_memory_to_use;
+
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have either -outfile switch or explicit output file name */
+  if (outfilename == NULL) {
+    if (file_index != argc-2) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+	      progname);
+      usage();
+    }
+    outfilename = argv[file_index+1];
+  } else {
+    if (file_index != argc-1) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+	      progname);
+      usage();
+    }
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (file_index < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Open the input file. */
+  if (file_index < argc) {
+    if ((input_file = fopen(argv[file_index], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[file_index]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+    input_file = read_stdin();
+  }
+
+  /* Open the output file. */
+  if (outfilename != NULL) {
+    if ((output_file = fopen(outfilename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, outfilename);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default output file is stdout */
+    output_file = write_stdout();
+  }
+
+#ifdef PROGRESS_REPORT
+  start_progress_monitor((j_common_ptr) &dstinfo, &progress);
+#endif
+
+  /* Specify data source for decompression */
+  jpeg_stdio_src(&srcinfo, input_file);
+
+  /* Enable saving of extra markers that we want to copy */
+  jcopy_markers_setup(&srcinfo, copyoption);
+
+  /* Read file header */
+  (void) jpeg_read_header(&srcinfo, TRUE);
+
+  /* Any space needed by a transform option must be requested before
+   * jpeg_read_coefficients so that memory allocation will be done right.
+   */
+#if TRANSFORMS_SUPPORTED
+  jtransform_request_workspace(&srcinfo, &transformoption);
+#endif
+
+  /* Read source file as DCT coefficients */
+  src_coef_arrays = jpeg_read_coefficients(&srcinfo);
+
+  /* Initialize destination compression parameters from source values */
+  jpeg_copy_critical_parameters(&srcinfo, &dstinfo);
+
+  /* Adjust destination parameters if required by transform options;
+   * also find out which set of coefficient arrays will hold the output.
+   */
+#if TRANSFORMS_SUPPORTED
+  dst_coef_arrays = jtransform_adjust_parameters(&srcinfo, &dstinfo,
+						 src_coef_arrays,
+						 &transformoption);
+#else
+  dst_coef_arrays = src_coef_arrays;
+#endif
+
+  /* Adjust default compression parameters by re-parsing the options */
+  file_index = parse_switches(&dstinfo, argc, argv, 0, TRUE);
+
+  /* Specify data destination for compression */
+  jpeg_stdio_dest(&dstinfo, output_file);
+
+  /* Start compressor (note no image data is actually written here) */
+  jpeg_write_coefficients(&dstinfo, dst_coef_arrays);
+
+  /* Copy to the output file any extra markers that we want to preserve */
+  jcopy_markers_execute(&srcinfo, &dstinfo, copyoption);
+
+  /* Execute image transformation, if any */
+#if TRANSFORMS_SUPPORTED
+  jtransform_execute_transformation(&srcinfo, &dstinfo,
+				    src_coef_arrays,
+				    &transformoption);
+#endif
+
+  /* Finish compression and release memory */
+  jpeg_finish_compress(&dstinfo);
+  jpeg_destroy_compress(&dstinfo);
+  (void) jpeg_finish_decompress(&srcinfo);
+  jpeg_destroy_decompress(&srcinfo);
+
+  /* Close files, if we opened them */
+  if (input_file != stdin)
+    fclose(input_file);
+  if (output_file != stdout)
+    fclose(output_file);
+
+#ifdef PROGRESS_REPORT
+  end_progress_monitor((j_common_ptr) &dstinfo);
+#endif
+
+  /* All done. */
+  exit(jsrcerr.num_warnings + jdsterr.num_warnings ?EXIT_WARNING:EXIT_SUCCESS);
+  return 0;			/* suppress no-return-value warnings */
+}
diff --git a/libjpegtwrp/jquant1.c b/libjpegtwrp/jquant1.c
new file mode 100644
index 000000000..b2f96aa15
--- /dev/null
+++ b/libjpegtwrp/jquant1.c
@@ -0,0 +1,856 @@
+/*
+ * jquant1.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains 1-pass color quantization (color mapping) routines.
+ * These routines provide mapping to a fixed color map using equally spaced
+ * color values.  Optional Floyd-Steinberg or ordered dithering is available.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+#ifdef QUANT_1PASS_SUPPORTED
+
+
+/*
+ * The main purpose of 1-pass quantization is to provide a fast, if not very
+ * high quality, colormapped output capability.  A 2-pass quantizer usually
+ * gives better visual quality; however, for quantized grayscale output this
+ * quantizer is perfectly adequate.  Dithering is highly recommended with this
+ * quantizer, though you can turn it off if you really want to.
+ *
+ * In 1-pass quantization the colormap must be chosen in advance of seeing the
+ * image.  We use a map consisting of all combinations of Ncolors[i] color
+ * values for the i'th component.  The Ncolors[] values are chosen so that
+ * their product, the total number of colors, is no more than that requested.
+ * (In most cases, the product will be somewhat less.)
+ *
+ * Since the colormap is orthogonal, the representative value for each color
+ * component can be determined without considering the other components;
+ * then these indexes can be combined into a colormap index by a standard
+ * N-dimensional-array-subscript calculation.  Most of the arithmetic involved
+ * can be precalculated and stored in the lookup table colorindex[].
+ * colorindex[i][j] maps pixel value j in component i to the nearest
+ * representative value (grid plane) for that component; this index is
+ * multiplied by the array stride for component i, so that the
+ * index of the colormap entry closest to a given pixel value is just
+ *    sum( colorindex[component-number][pixel-component-value] )
+ * Aside from being fast, this scheme allows for variable spacing between
+ * representative values with no additional lookup cost.
+ *
+ * If gamma correction has been applied in color conversion, it might be wise
+ * to adjust the color grid spacing so that the representative colors are
+ * equidistant in linear space.  At this writing, gamma correction is not
+ * implemented by jdcolor, so nothing is done here.
+ */
+
+
+/* Declarations for ordered dithering.
+ *
+ * We use a standard 16x16 ordered dither array.  The basic concept of ordered
+ * dithering is described in many references, for instance Dale Schumacher's
+ * chapter II.2 of Graphics Gems II (James Arvo, ed. Academic Press, 1991).
+ * In place of Schumacher's comparisons against a "threshold" value, we add a
+ * "dither" value to the input pixel and then round the result to the nearest
+ * output value.  The dither value is equivalent to (0.5 - threshold) times
+ * the distance between output values.  For ordered dithering, we assume that
+ * the output colors are equally spaced; if not, results will probably be
+ * worse, since the dither may be too much or too little at a given point.
+ *
+ * The normal calculation would be to form pixel value + dither, range-limit
+ * this to 0..MAXJSAMPLE, and then index into the colorindex table as usual.
+ * We can skip the separate range-limiting step by extending the colorindex
+ * table in both directions.
+ */
+
+#define ODITHER_SIZE  16	/* dimension of dither matrix */
+/* NB: if ODITHER_SIZE is not a power of 2, ODITHER_MASK uses will break */
+#define ODITHER_CELLS (ODITHER_SIZE*ODITHER_SIZE)	/* # cells in matrix */
+#define ODITHER_MASK  (ODITHER_SIZE-1) /* mask for wrapping around counters */
+
+typedef int ODITHER_MATRIX[ODITHER_SIZE][ODITHER_SIZE];
+typedef int (*ODITHER_MATRIX_PTR)[ODITHER_SIZE];
+
+static const UINT8 base_dither_matrix[ODITHER_SIZE][ODITHER_SIZE] = {
+  /* Bayer's order-4 dither array.  Generated by the code given in
+   * Stephen Hawley's article "Ordered Dithering" in Graphics Gems I.
+   * The values in this array must range from 0 to ODITHER_CELLS-1.
+   */
+  {   0,192, 48,240, 12,204, 60,252,  3,195, 51,243, 15,207, 63,255 },
+  { 128, 64,176,112,140, 76,188,124,131, 67,179,115,143, 79,191,127 },
+  {  32,224, 16,208, 44,236, 28,220, 35,227, 19,211, 47,239, 31,223 },
+  { 160, 96,144, 80,172,108,156, 92,163, 99,147, 83,175,111,159, 95 },
+  {   8,200, 56,248,  4,196, 52,244, 11,203, 59,251,  7,199, 55,247 },
+  { 136, 72,184,120,132, 68,180,116,139, 75,187,123,135, 71,183,119 },
+  {  40,232, 24,216, 36,228, 20,212, 43,235, 27,219, 39,231, 23,215 },
+  { 168,104,152, 88,164,100,148, 84,171,107,155, 91,167,103,151, 87 },
+  {   2,194, 50,242, 14,206, 62,254,  1,193, 49,241, 13,205, 61,253 },
+  { 130, 66,178,114,142, 78,190,126,129, 65,177,113,141, 77,189,125 },
+  {  34,226, 18,210, 46,238, 30,222, 33,225, 17,209, 45,237, 29,221 },
+  { 162, 98,146, 82,174,110,158, 94,161, 97,145, 81,173,109,157, 93 },
+  {  10,202, 58,250,  6,198, 54,246,  9,201, 57,249,  5,197, 53,245 },
+  { 138, 74,186,122,134, 70,182,118,137, 73,185,121,133, 69,181,117 },
+  {  42,234, 26,218, 38,230, 22,214, 41,233, 25,217, 37,229, 21,213 },
+  { 170,106,154, 90,166,102,150, 86,169,105,153, 89,165,101,149, 85 }
+};
+
+
+/* Declarations for Floyd-Steinberg dithering.
+ *
+ * Errors are accumulated into the array fserrors[], at a resolution of
+ * 1/16th of a pixel count.  The error at a given pixel is propagated
+ * to its not-yet-processed neighbors using the standard F-S fractions,
+ *		...	(here)	7/16
+ *		3/16	5/16	1/16
+ * We work left-to-right on even rows, right-to-left on odd rows.
+ *
+ * We can get away with a single array (holding one row's worth of errors)
+ * by using it to store the current row's errors at pixel columns not yet
+ * processed, but the next row's errors at columns already processed.  We
+ * need only a few extra variables to hold the errors immediately around the
+ * current column.  (If we are lucky, those variables are in registers, but
+ * even if not, they're probably cheaper to access than array elements are.)
+ *
+ * The fserrors[] array is indexed [component#][position].
+ * We provide (#columns + 2) entries per component; the extra entry at each
+ * end saves us from special-casing the first and last pixels.
+ *
+ * Note: on a wide image, we might not have enough room in a PC's near data
+ * segment to hold the error array; so it is allocated with alloc_large.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+typedef INT16 FSERROR;		/* 16 bits should be enough */
+typedef int LOCFSERROR;		/* use 'int' for calculation temps */
+#else
+typedef INT32 FSERROR;		/* may need more than 16 bits */
+typedef INT32 LOCFSERROR;	/* be sure calculation temps are big enough */
+#endif
+
+typedef FSERROR FAR *FSERRPTR;	/* pointer to error array (in FAR storage!) */
+
+
+/* Private subobject */
+
+#define MAX_Q_COMPS 4		/* max components I can handle */
+
+typedef struct {
+  struct jpeg_color_quantizer pub; /* public fields */
+
+  /* Initially allocated colormap is saved here */
+  JSAMPARRAY sv_colormap;	/* The color map as a 2-D pixel array */
+  int sv_actual;		/* number of entries in use */
+
+  JSAMPARRAY colorindex;	/* Precomputed mapping for speed */
+  /* colorindex[i][j] = index of color closest to pixel value j in component i,
+   * premultiplied as described above.  Since colormap indexes must fit into
+   * JSAMPLEs, the entries of this array will too.
+   */
+  boolean is_padded;		/* is the colorindex padded for odither? */
+
+  int Ncolors[MAX_Q_COMPS];	/* # of values alloced to each component */
+
+  /* Variables for ordered dithering */
+  int row_index;		/* cur row's vertical index in dither matrix */
+  ODITHER_MATRIX_PTR odither[MAX_Q_COMPS]; /* one dither array per component */
+
+  /* Variables for Floyd-Steinberg dithering */
+  FSERRPTR fserrors[MAX_Q_COMPS]; /* accumulated errors */
+  boolean on_odd_row;		/* flag to remember which row we are on */
+} my_cquantizer;
+
+typedef my_cquantizer * my_cquantize_ptr;
+
+
+/*
+ * Policy-making subroutines for create_colormap and create_colorindex.
+ * These routines determine the colormap to be used.  The rest of the module
+ * only assumes that the colormap is orthogonal.
+ *
+ *  * select_ncolors decides how to divvy up the available colors
+ *    among the components.
+ *  * output_value defines the set of representative values for a component.
+ *  * largest_input_value defines the mapping from input values to
+ *    representative values for a component.
+ * Note that the latter two routines may impose different policies for
+ * different components, though this is not currently done.
+ */
+
+
+LOCAL(int)
+select_ncolors (j_decompress_ptr cinfo, int Ncolors[])
+/* Determine allocation of desired colors to components, */
+/* and fill in Ncolors[] array to indicate choice. */
+/* Return value is total number of colors (product of Ncolors[] values). */
+{
+  int nc = cinfo->out_color_components; /* number of color components */
+  int max_colors = cinfo->desired_number_of_colors;
+  int total_colors, iroot, i, j;
+  boolean changed;
+  long temp;
+  static const int RGB_order[3] = { RGB_GREEN, RGB_RED, RGB_BLUE };
+
+  /* We can allocate at least the nc'th root of max_colors per component. */
+  /* Compute floor(nc'th root of max_colors). */
+  iroot = 1;
+  do {
+    iroot++;
+    temp = iroot;		/* set temp = iroot ** nc */
+    for (i = 1; i < nc; i++)
+      temp *= iroot;
+  } while (temp <= (long) max_colors); /* repeat till iroot exceeds root */
+  iroot--;			/* now iroot = floor(root) */
+
+  /* Must have at least 2 color values per component */
+  if (iroot < 2)
+    ERREXIT1(cinfo, JERR_QUANT_FEW_COLORS, (int) temp);
+
+  /* Initialize to iroot color values for each component */
+  total_colors = 1;
+  for (i = 0; i < nc; i++) {
+    Ncolors[i] = iroot;
+    total_colors *= iroot;
+  }
+  /* We may be able to increment the count for one or more components without
+   * exceeding max_colors, though we know not all can be incremented.
+   * Sometimes, the first component can be incremented more than once!
+   * (Example: for 16 colors, we start at 2*2*2, go to 3*2*2, then 4*2*2.)
+   * In RGB colorspace, try to increment G first, then R, then B.
+   */
+  do {
+    changed = FALSE;
+    for (i = 0; i < nc; i++) {
+      j = (cinfo->out_color_space == JCS_RGB ? RGB_order[i] : i);
+      /* calculate new total_colors if Ncolors[j] is incremented */
+      temp = total_colors / Ncolors[j];
+      temp *= Ncolors[j]+1;	/* done in long arith to avoid oflo */
+      if (temp > (long) max_colors)
+	break;			/* won't fit, done with this pass */
+      Ncolors[j]++;		/* OK, apply the increment */
+      total_colors = (int) temp;
+      changed = TRUE;
+    }
+  } while (changed);
+
+  return total_colors;
+}
+
+
+LOCAL(int)
+output_value (j_decompress_ptr cinfo, int ci, int j, int maxj)
+/* Return j'th output value, where j will range from 0 to maxj */
+/* The output values must fall in 0..MAXJSAMPLE in increasing order */
+{
+  /* We always provide values 0 and MAXJSAMPLE for each component;
+   * any additional values are equally spaced between these limits.
+   * (Forcing the upper and lower values to the limits ensures that
+   * dithering can't produce a color outside the selected gamut.)
+   */
+  return (int) (((INT32) j * MAXJSAMPLE + maxj/2) / maxj);
+}
+
+
+LOCAL(int)
+largest_input_value (j_decompress_ptr cinfo, int ci, int j, int maxj)
+/* Return largest input value that should map to j'th output value */
+/* Must have largest(j=0) >= 0, and largest(j=maxj) >= MAXJSAMPLE */
+{
+  /* Breakpoints are halfway between values returned by output_value */
+  return (int) (((INT32) (2*j + 1) * MAXJSAMPLE + maxj) / (2*maxj));
+}
+
+
+/*
+ * Create the colormap.
+ */
+
+LOCAL(void)
+create_colormap (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  JSAMPARRAY colormap;		/* Created colormap */
+  int total_colors;		/* Number of distinct output colors */
+  int i,j,k, nci, blksize, blkdist, ptr, val;
+
+  /* Select number of colors for each component */
+  total_colors = select_ncolors(cinfo, cquantize->Ncolors);
+
+  /* Report selected color counts */
+  if (cinfo->out_color_components == 3)
+    TRACEMS4(cinfo, 1, JTRC_QUANT_3_NCOLORS,
+	     total_colors, cquantize->Ncolors[0],
+	     cquantize->Ncolors[1], cquantize->Ncolors[2]);
+  else
+    TRACEMS1(cinfo, 1, JTRC_QUANT_NCOLORS, total_colors);
+
+  /* Allocate and fill in the colormap. */
+  /* The colors are ordered in the map in standard row-major order, */
+  /* i.e. rightmost (highest-indexed) color changes most rapidly. */
+
+  colormap = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     (JDIMENSION) total_colors, (JDIMENSION) cinfo->out_color_components);
+
+  /* blksize is number of adjacent repeated entries for a component */
+  /* blkdist is distance between groups of identical entries for a component */
+  blkdist = total_colors;
+
+  for (i = 0; i < cinfo->out_color_components; i++) {
+    /* fill in colormap entries for i'th color component */
+    nci = cquantize->Ncolors[i]; /* # of distinct values for this color */
+    blksize = blkdist / nci;
+    for (j = 0; j < nci; j++) {
+      /* Compute j'th output value (out of nci) for component */
+      val = output_value(cinfo, i, j, nci-1);
+      /* Fill in all colormap entries that have this value of this component */
+      for (ptr = j * blksize; ptr < total_colors; ptr += blkdist) {
+	/* fill in blksize entries beginning at ptr */
+	for (k = 0; k < blksize; k++)
+	  colormap[i][ptr+k] = (JSAMPLE) val;
+      }
+    }
+    blkdist = blksize;		/* blksize of this color is blkdist of next */
+  }
+
+  /* Save the colormap in private storage,
+   * where it will survive color quantization mode changes.
+   */
+  cquantize->sv_colormap = colormap;
+  cquantize->sv_actual = total_colors;
+}
+
+
+/*
+ * Create the color index table.
+ */
+
+LOCAL(void)
+create_colorindex (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  JSAMPROW indexptr;
+  int i,j,k, nci, blksize, val, pad;
+
+  /* For ordered dither, we pad the color index tables by MAXJSAMPLE in
+   * each direction (input index values can be -MAXJSAMPLE .. 2*MAXJSAMPLE).
+   * This is not necessary in the other dithering modes.  However, we
+   * flag whether it was done in case user changes dithering mode.
+   */
+  if (cinfo->dither_mode == JDITHER_ORDERED) {
+    pad = MAXJSAMPLE*2;
+    cquantize->is_padded = TRUE;
+  } else {
+    pad = 0;
+    cquantize->is_padded = FALSE;
+  }
+
+  cquantize->colorindex = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     (JDIMENSION) (MAXJSAMPLE+1 + pad),
+     (JDIMENSION) cinfo->out_color_components);
+
+  /* blksize is number of adjacent repeated entries for a component */
+  blksize = cquantize->sv_actual;
+
+  for (i = 0; i < cinfo->out_color_components; i++) {
+    /* fill in colorindex entries for i'th color component */
+    nci = cquantize->Ncolors[i]; /* # of distinct values for this color */
+    blksize = blksize / nci;
+
+    /* adjust colorindex pointers to provide padding at negative indexes. */
+    if (pad)
+      cquantize->colorindex[i] += MAXJSAMPLE;
+
+    /* in loop, val = index of current output value, */
+    /* and k = largest j that maps to current val */
+    indexptr = cquantize->colorindex[i];
+    val = 0;
+    k = largest_input_value(cinfo, i, 0, nci-1);
+    for (j = 0; j <= MAXJSAMPLE; j++) {
+      while (j > k)		/* advance val if past boundary */
+	k = largest_input_value(cinfo, i, ++val, nci-1);
+      /* premultiply so that no multiplication needed in main processing */
+      indexptr[j] = (JSAMPLE) (val * blksize);
+    }
+    /* Pad at both ends if necessary */
+    if (pad)
+      for (j = 1; j <= MAXJSAMPLE; j++) {
+	indexptr[-j] = indexptr[0];
+	indexptr[MAXJSAMPLE+j] = indexptr[MAXJSAMPLE];
+      }
+  }
+}
+
+
+/*
+ * Create an ordered-dither array for a component having ncolors
+ * distinct output values.
+ */
+
+LOCAL(ODITHER_MATRIX_PTR)
+make_odither_array (j_decompress_ptr cinfo, int ncolors)
+{
+  ODITHER_MATRIX_PTR odither;
+  int j,k;
+  INT32 num,den;
+
+  odither = (ODITHER_MATRIX_PTR)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(ODITHER_MATRIX));
+  /* The inter-value distance for this color is MAXJSAMPLE/(ncolors-1).
+   * Hence the dither value for the matrix cell with fill order f
+   * (f=0..N-1) should be (N-1-2*f)/(2*N) * MAXJSAMPLE/(ncolors-1).
+   * On 16-bit-int machine, be careful to avoid overflow.
+   */
+  den = 2 * ODITHER_CELLS * ((INT32) (ncolors - 1));
+  for (j = 0; j < ODITHER_SIZE; j++) {
+    for (k = 0; k < ODITHER_SIZE; k++) {
+      num = ((INT32) (ODITHER_CELLS-1 - 2*((int)base_dither_matrix[j][k])))
+	    * MAXJSAMPLE;
+      /* Ensure round towards zero despite C's lack of consistency
+       * about rounding negative values in integer division...
+       */
+      odither[j][k] = (int) (num<0 ? -((-num)/den) : num/den);
+    }
+  }
+  return odither;
+}
+
+
+/*
+ * Create the ordered-dither tables.
+ * Components having the same number of representative colors may 
+ * share a dither table.
+ */
+
+LOCAL(void)
+create_odither_tables (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  ODITHER_MATRIX_PTR odither;
+  int i, j, nci;
+
+  for (i = 0; i < cinfo->out_color_components; i++) {
+    nci = cquantize->Ncolors[i]; /* # of distinct values for this color */
+    odither = NULL;		/* search for matching prior component */
+    for (j = 0; j < i; j++) {
+      if (nci == cquantize->Ncolors[j]) {
+	odither = cquantize->odither[j];
+	break;
+      }
+    }
+    if (odither == NULL)	/* need a new table? */
+      odither = make_odither_array(cinfo, nci);
+    cquantize->odither[i] = odither;
+  }
+}
+
+
+/*
+ * Map some rows of pixels to the output colormapped representation.
+ */
+
+METHODDEF(void)
+color_quantize (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+		JSAMPARRAY output_buf, int num_rows)
+/* General case, no dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  JSAMPARRAY colorindex = cquantize->colorindex;
+  register int pixcode, ci;
+  register JSAMPROW ptrin, ptrout;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+  register int nc = cinfo->out_color_components;
+
+  for (row = 0; row < num_rows; row++) {
+    ptrin = input_buf[row];
+    ptrout = output_buf[row];
+    for (col = width; col > 0; col--) {
+      pixcode = 0;
+      for (ci = 0; ci < nc; ci++) {
+	pixcode += GETJSAMPLE(colorindex[ci][GETJSAMPLE(*ptrin++)]);
+      }
+      *ptrout++ = (JSAMPLE) pixcode;
+    }
+  }
+}
+
+
+METHODDEF(void)
+color_quantize3 (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+		 JSAMPARRAY output_buf, int num_rows)
+/* Fast path for out_color_components==3, no dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  register int pixcode;
+  register JSAMPROW ptrin, ptrout;
+  JSAMPROW colorindex0 = cquantize->colorindex[0];
+  JSAMPROW colorindex1 = cquantize->colorindex[1];
+  JSAMPROW colorindex2 = cquantize->colorindex[2];
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    ptrin = input_buf[row];
+    ptrout = output_buf[row];
+    for (col = width; col > 0; col--) {
+      pixcode  = GETJSAMPLE(colorindex0[GETJSAMPLE(*ptrin++)]);
+      pixcode += GETJSAMPLE(colorindex1[GETJSAMPLE(*ptrin++)]);
+      pixcode += GETJSAMPLE(colorindex2[GETJSAMPLE(*ptrin++)]);
+      *ptrout++ = (JSAMPLE) pixcode;
+    }
+  }
+}
+
+
+METHODDEF(void)
+quantize_ord_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+		     JSAMPARRAY output_buf, int num_rows)
+/* General case, with ordered dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  register JSAMPROW input_ptr;
+  register JSAMPROW output_ptr;
+  JSAMPROW colorindex_ci;
+  int * dither;			/* points to active row of dither matrix */
+  int row_index, col_index;	/* current indexes into dither matrix */
+  int nc = cinfo->out_color_components;
+  int ci;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    /* Initialize output values to 0 so can process components separately */
+    jzero_far((void FAR *) output_buf[row],
+	      (size_t) (width * SIZEOF(JSAMPLE)));
+    row_index = cquantize->row_index;
+    for (ci = 0; ci < nc; ci++) {
+      input_ptr = input_buf[row] + ci;
+      output_ptr = output_buf[row];
+      colorindex_ci = cquantize->colorindex[ci];
+      dither = cquantize->odither[ci][row_index];
+      col_index = 0;
+
+      for (col = width; col > 0; col--) {
+	/* Form pixel value + dither, range-limit to 0..MAXJSAMPLE,
+	 * select output value, accumulate into output code for this pixel.
+	 * Range-limiting need not be done explicitly, as we have extended
+	 * the colorindex table to produce the right answers for out-of-range
+	 * inputs.  The maximum dither is +- MAXJSAMPLE; this sets the
+	 * required amount of padding.
+	 */
+	*output_ptr += colorindex_ci[GETJSAMPLE(*input_ptr)+dither[col_index]];
+	input_ptr += nc;
+	output_ptr++;
+	col_index = (col_index + 1) & ODITHER_MASK;
+      }
+    }
+    /* Advance row index for next row */
+    row_index = (row_index + 1) & ODITHER_MASK;
+    cquantize->row_index = row_index;
+  }
+}
+
+
+METHODDEF(void)
+quantize3_ord_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+		      JSAMPARRAY output_buf, int num_rows)
+/* Fast path for out_color_components==3, with ordered dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  register int pixcode;
+  register JSAMPROW input_ptr;
+  register JSAMPROW output_ptr;
+  JSAMPROW colorindex0 = cquantize->colorindex[0];
+  JSAMPROW colorindex1 = cquantize->colorindex[1];
+  JSAMPROW colorindex2 = cquantize->colorindex[2];
+  int * dither0;		/* points to active row of dither matrix */
+  int * dither1;
+  int * dither2;
+  int row_index, col_index;	/* current indexes into dither matrix */
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    row_index = cquantize->row_index;
+    input_ptr = input_buf[row];
+    output_ptr = output_buf[row];
+    dither0 = cquantize->odither[0][row_index];
+    dither1 = cquantize->odither[1][row_index];
+    dither2 = cquantize->odither[2][row_index];
+    col_index = 0;
+
+    for (col = width; col > 0; col--) {
+      pixcode  = GETJSAMPLE(colorindex0[GETJSAMPLE(*input_ptr++) +
+					dither0[col_index]]);
+      pixcode += GETJSAMPLE(colorindex1[GETJSAMPLE(*input_ptr++) +
+					dither1[col_index]]);
+      pixcode += GETJSAMPLE(colorindex2[GETJSAMPLE(*input_ptr++) +
+					dither2[col_index]]);
+      *output_ptr++ = (JSAMPLE) pixcode;
+      col_index = (col_index + 1) & ODITHER_MASK;
+    }
+    row_index = (row_index + 1) & ODITHER_MASK;
+    cquantize->row_index = row_index;
+  }
+}
+
+
+METHODDEF(void)
+quantize_fs_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+		    JSAMPARRAY output_buf, int num_rows)
+/* General case, with Floyd-Steinberg dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  register LOCFSERROR cur;	/* current error or pixel value */
+  LOCFSERROR belowerr;		/* error for pixel below cur */
+  LOCFSERROR bpreverr;		/* error for below/prev col */
+  LOCFSERROR bnexterr;		/* error for below/next col */
+  LOCFSERROR delta;
+  register FSERRPTR errorptr;	/* => fserrors[] at column before current */
+  register JSAMPROW input_ptr;
+  register JSAMPROW output_ptr;
+  JSAMPROW colorindex_ci;
+  JSAMPROW colormap_ci;
+  int pixcode;
+  int nc = cinfo->out_color_components;
+  int dir;			/* 1 for left-to-right, -1 for right-to-left */
+  int dirnc;			/* dir * nc */
+  int ci;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+  JSAMPLE *range_limit = cinfo->sample_range_limit;
+  SHIFT_TEMPS
+
+  for (row = 0; row < num_rows; row++) {
+    /* Initialize output values to 0 so can process components separately */
+    jzero_far((void FAR *) output_buf[row],
+	      (size_t) (width * SIZEOF(JSAMPLE)));
+    for (ci = 0; ci < nc; ci++) {
+      input_ptr = input_buf[row] + ci;
+      output_ptr = output_buf[row];
+      if (cquantize->on_odd_row) {
+	/* work right to left in this row */
+	input_ptr += (width-1) * nc; /* so point to rightmost pixel */
+	output_ptr += width-1;
+	dir = -1;
+	dirnc = -nc;
+	errorptr = cquantize->fserrors[ci] + (width+1); /* => entry after last column */
+      } else {
+	/* work left to right in this row */
+	dir = 1;
+	dirnc = nc;
+	errorptr = cquantize->fserrors[ci]; /* => entry before first column */
+      }
+      colorindex_ci = cquantize->colorindex[ci];
+      colormap_ci = cquantize->sv_colormap[ci];
+      /* Preset error values: no error propagated to first pixel from left */
+      cur = 0;
+      /* and no error propagated to row below yet */
+      belowerr = bpreverr = 0;
+
+      for (col = width; col > 0; col--) {
+	/* cur holds the error propagated from the previous pixel on the
+	 * current line.  Add the error propagated from the previous line
+	 * to form the complete error correction term for this pixel, and
+	 * round the error term (which is expressed * 16) to an integer.
+	 * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct
+	 * for either sign of the error value.
+	 * Note: errorptr points to *previous* column's array entry.
+	 */
+	cur = RIGHT_SHIFT(cur + errorptr[dir] + 8, 4);
+	/* Form pixel value + error, and range-limit to 0..MAXJSAMPLE.
+	 * The maximum error is +- MAXJSAMPLE; this sets the required size
+	 * of the range_limit array.
+	 */
+	cur += GETJSAMPLE(*input_ptr);
+	cur = GETJSAMPLE(range_limit[cur]);
+	/* Select output value, accumulate into output code for this pixel */
+	pixcode = GETJSAMPLE(colorindex_ci[cur]);
+	*output_ptr += (JSAMPLE) pixcode;
+	/* Compute actual representation error at this pixel */
+	/* Note: we can do this even though we don't have the final */
+	/* pixel code, because the colormap is orthogonal. */
+	cur -= GETJSAMPLE(colormap_ci[pixcode]);
+	/* Compute error fractions to be propagated to adjacent pixels.
+	 * Add these into the running sums, and simultaneously shift the
+	 * next-line error sums left by 1 column.
+	 */
+	bnexterr = cur;
+	delta = cur * 2;
+	cur += delta;		/* form error * 3 */
+	errorptr[0] = (FSERROR) (bpreverr + cur);
+	cur += delta;		/* form error * 5 */
+	bpreverr = belowerr + cur;
+	belowerr = bnexterr;
+	cur += delta;		/* form error * 7 */
+	/* At this point cur contains the 7/16 error value to be propagated
+	 * to the next pixel on the current line, and all the errors for the
+	 * next line have been shifted over. We are therefore ready to move on.
+	 */
+	input_ptr += dirnc;	/* advance input ptr to next column */
+	output_ptr += dir;	/* advance output ptr to next column */
+	errorptr += dir;	/* advance errorptr to current column */
+      }
+      /* Post-loop cleanup: we must unload the final error value into the
+       * final fserrors[] entry.  Note we need not unload belowerr because
+       * it is for the dummy column before or after the actual array.
+       */
+      errorptr[0] = (FSERROR) bpreverr; /* unload prev err into array */
+    }
+    cquantize->on_odd_row = (cquantize->on_odd_row ? FALSE : TRUE);
+  }
+}
+
+
+/*
+ * Allocate workspace for Floyd-Steinberg errors.
+ */
+
+LOCAL(void)
+alloc_fs_workspace (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  size_t arraysize;
+  int i;
+
+  arraysize = (size_t) ((cinfo->output_width + 2) * SIZEOF(FSERROR));
+  for (i = 0; i < cinfo->out_color_components; i++) {
+    cquantize->fserrors[i] = (FSERRPTR)
+      (*cinfo->mem->alloc_large)((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize);
+  }
+}
+
+
+/*
+ * Initialize for one-pass color quantization.
+ */
+
+METHODDEF(void)
+start_pass_1_quant (j_decompress_ptr cinfo, boolean is_pre_scan)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  size_t arraysize;
+  int i;
+
+  /* Install my colormap. */
+  cinfo->colormap = cquantize->sv_colormap;
+  cinfo->actual_number_of_colors = cquantize->sv_actual;
+
+  /* Initialize for desired dithering mode. */
+  switch (cinfo->dither_mode) {
+  case JDITHER_NONE:
+    if (cinfo->out_color_components == 3)
+      cquantize->pub.color_quantize = color_quantize3;
+    else
+      cquantize->pub.color_quantize = color_quantize;
+    break;
+  case JDITHER_ORDERED:
+    if (cinfo->out_color_components == 3)
+      cquantize->pub.color_quantize = quantize3_ord_dither;
+    else
+      cquantize->pub.color_quantize = quantize_ord_dither;
+    cquantize->row_index = 0;	/* initialize state for ordered dither */
+    /* If user changed to ordered dither from another mode,
+     * we must recreate the color index table with padding.
+     * This will cost extra space, but probably isn't very likely.
+     */
+    if (! cquantize->is_padded)
+      create_colorindex(cinfo);
+    /* Create ordered-dither tables if we didn't already. */
+    if (cquantize->odither[0] == NULL)
+      create_odither_tables(cinfo);
+    break;
+  case JDITHER_FS:
+    cquantize->pub.color_quantize = quantize_fs_dither;
+    cquantize->on_odd_row = FALSE; /* initialize state for F-S dither */
+    /* Allocate Floyd-Steinberg workspace if didn't already. */
+    if (cquantize->fserrors[0] == NULL)
+      alloc_fs_workspace(cinfo);
+    /* Initialize the propagated errors to zero. */
+    arraysize = (size_t) ((cinfo->output_width + 2) * SIZEOF(FSERROR));
+    for (i = 0; i < cinfo->out_color_components; i++)
+      jzero_far((void FAR *) cquantize->fserrors[i], arraysize);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+    break;
+  }
+}
+
+
+/*
+ * Finish up at the end of the pass.
+ */
+
+METHODDEF(void)
+finish_pass_1_quant (j_decompress_ptr cinfo)
+{
+  /* no work in 1-pass case */
+}
+
+
+/*
+ * Switch to a new external colormap between output passes.
+ * Shouldn't get to this module!
+ */
+
+METHODDEF(void)
+new_color_map_1_quant (j_decompress_ptr cinfo)
+{
+  ERREXIT(cinfo, JERR_MODE_CHANGE);
+}
+
+
+/*
+ * Module initialization routine for 1-pass color quantization.
+ */
+
+GLOBAL(void)
+jinit_1pass_quantizer (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize;
+
+  cquantize = (my_cquantize_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_cquantizer));
+  cinfo->cquantize = (struct jpeg_color_quantizer *) cquantize;
+  cquantize->pub.start_pass = start_pass_1_quant;
+  cquantize->pub.finish_pass = finish_pass_1_quant;
+  cquantize->pub.new_color_map = new_color_map_1_quant;
+  cquantize->fserrors[0] = NULL; /* Flag FS workspace not allocated */
+  cquantize->odither[0] = NULL;	/* Also flag odither arrays not allocated */
+
+  /* Make sure my internal arrays won't overflow */
+  if (cinfo->out_color_components > MAX_Q_COMPS)
+    ERREXIT1(cinfo, JERR_QUANT_COMPONENTS, MAX_Q_COMPS);
+  /* Make sure colormap indexes can be represented by JSAMPLEs */
+  if (cinfo->desired_number_of_colors > (MAXJSAMPLE+1))
+    ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXJSAMPLE+1);
+
+  /* Create the colormap and color index table. */
+  create_colormap(cinfo);
+  create_colorindex(cinfo);
+
+  /* Allocate Floyd-Steinberg workspace now if requested.
+   * We do this now since it is FAR storage and may affect the memory
+   * manager's space calculations.  If the user changes to FS dither
+   * mode in a later pass, we will allocate the space then, and will
+   * possibly overrun the max_memory_to_use setting.
+   */
+  if (cinfo->dither_mode == JDITHER_FS)
+    alloc_fs_workspace(cinfo);
+}
+
+#endif /* QUANT_1PASS_SUPPORTED */
diff --git a/libjpegtwrp/jquant2.c b/libjpegtwrp/jquant2.c
new file mode 100644
index 000000000..af601e334
--- /dev/null
+++ b/libjpegtwrp/jquant2.c
@@ -0,0 +1,1310 @@
+/*
+ * jquant2.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains 2-pass color quantization (color mapping) routines.
+ * These routines provide selection of a custom color map for an image,
+ * followed by mapping of the image to that color map, with optional
+ * Floyd-Steinberg dithering.
+ * It is also possible to use just the second pass to map to an arbitrary
+ * externally-given color map.
+ *
+ * Note: ordered dithering is not supported, since there isn't any fast
+ * way to compute intercolor distances; it's unclear that ordered dither's
+ * fundamental assumptions even hold with an irregularly spaced color map.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+#ifdef QUANT_2PASS_SUPPORTED
+
+
+/*
+ * This module implements the well-known Heckbert paradigm for color
+ * quantization.  Most of the ideas used here can be traced back to
+ * Heckbert's seminal paper
+ *   Heckbert, Paul.  "Color Image Quantization for Frame Buffer Display",
+ *   Proc. SIGGRAPH '82, Computer Graphics v.16 #3 (July 1982), pp 297-304.
+ *
+ * In the first pass over the image, we accumulate a histogram showing the
+ * usage count of each possible color.  To keep the histogram to a reasonable
+ * size, we reduce the precision of the input; typical practice is to retain
+ * 5 or 6 bits per color, so that 8 or 4 different input values are counted
+ * in the same histogram cell.
+ *
+ * Next, the color-selection step begins with a box representing the whole
+ * color space, and repeatedly splits the "largest" remaining box until we
+ * have as many boxes as desired colors.  Then the mean color in each
+ * remaining box becomes one of the possible output colors.
+ * 
+ * The second pass over the image maps each input pixel to the closest output
+ * color (optionally after applying a Floyd-Steinberg dithering correction).
+ * This mapping is logically trivial, but making it go fast enough requires
+ * considerable care.
+ *
+ * Heckbert-style quantizers vary a good deal in their policies for choosing
+ * the "largest" box and deciding where to cut it.  The particular policies
+ * used here have proved out well in experimental comparisons, but better ones
+ * may yet be found.
+ *
+ * In earlier versions of the IJG code, this module quantized in YCbCr color
+ * space, processing the raw upsampled data without a color conversion step.
+ * This allowed the color conversion math to be done only once per colormap
+ * entry, not once per pixel.  However, that optimization precluded other
+ * useful optimizations (such as merging color conversion with upsampling)
+ * and it also interfered with desired capabilities such as quantizing to an
+ * externally-supplied colormap.  We have therefore abandoned that approach.
+ * The present code works in the post-conversion color space, typically RGB.
+ *
+ * To improve the visual quality of the results, we actually work in scaled
+ * RGB space, giving G distances more weight than R, and R in turn more than
+ * B.  To do everything in integer math, we must use integer scale factors.
+ * The 2/3/1 scale factors used here correspond loosely to the relative
+ * weights of the colors in the NTSC grayscale equation.
+ * If you want to use this code to quantize a non-RGB color space, you'll
+ * probably need to change these scale factors.
+ */
+
+#define R_SCALE 2		/* scale R distances by this much */
+#define G_SCALE 3		/* scale G distances by this much */
+#define B_SCALE 1		/* and B by this much */
+
+/* Relabel R/G/B as components 0/1/2, respecting the RGB ordering defined
+ * in jmorecfg.h.  As the code stands, it will do the right thing for R,G,B
+ * and B,G,R orders.  If you define some other weird order in jmorecfg.h,
+ * you'll get compile errors until you extend this logic.  In that case
+ * you'll probably want to tweak the histogram sizes too.
+ */
+
+#if RGB_RED == 0
+#define C0_SCALE R_SCALE
+#endif
+#if RGB_BLUE == 0
+#define C0_SCALE B_SCALE
+#endif
+#if RGB_GREEN == 1
+#define C1_SCALE G_SCALE
+#endif
+#if RGB_RED == 2
+#define C2_SCALE R_SCALE
+#endif
+#if RGB_BLUE == 2
+#define C2_SCALE B_SCALE
+#endif
+
+
+/*
+ * First we have the histogram data structure and routines for creating it.
+ *
+ * The number of bits of precision can be adjusted by changing these symbols.
+ * We recommend keeping 6 bits for G and 5 each for R and B.
+ * If you have plenty of memory and cycles, 6 bits all around gives marginally
+ * better results; if you are short of memory, 5 bits all around will save
+ * some space but degrade the results.
+ * To maintain a fully accurate histogram, we'd need to allocate a "long"
+ * (preferably unsigned long) for each cell.  In practice this is overkill;
+ * we can get by with 16 bits per cell.  Few of the cell counts will overflow,
+ * and clamping those that do overflow to the maximum value will give close-
+ * enough results.  This reduces the recommended histogram size from 256Kb
+ * to 128Kb, which is a useful savings on PC-class machines.
+ * (In the second pass the histogram space is re-used for pixel mapping data;
+ * in that capacity, each cell must be able to store zero to the number of
+ * desired colors.  16 bits/cell is plenty for that too.)
+ * Since the JPEG code is intended to run in small memory model on 80x86
+ * machines, we can't just allocate the histogram in one chunk.  Instead
+ * of a true 3-D array, we use a row of pointers to 2-D arrays.  Each
+ * pointer corresponds to a C0 value (typically 2^5 = 32 pointers) and
+ * each 2-D array has 2^6*2^5 = 2048 or 2^6*2^6 = 4096 entries.  Note that
+ * on 80x86 machines, the pointer row is in near memory but the actual
+ * arrays are in far memory (same arrangement as we use for image arrays).
+ */
+
+#define MAXNUMCOLORS  (MAXJSAMPLE+1) /* maximum size of colormap */
+
+/* These will do the right thing for either R,G,B or B,G,R color order,
+ * but you may not like the results for other color orders.
+ */
+#define HIST_C0_BITS  5		/* bits of precision in R/B histogram */
+#define HIST_C1_BITS  6		/* bits of precision in G histogram */
+#define HIST_C2_BITS  5		/* bits of precision in B/R histogram */
+
+/* Number of elements along histogram axes. */
+#define HIST_C0_ELEMS  (1<<HIST_C0_BITS)
+#define HIST_C1_ELEMS  (1<<HIST_C1_BITS)
+#define HIST_C2_ELEMS  (1<<HIST_C2_BITS)
+
+/* These are the amounts to shift an input value to get a histogram index. */
+#define C0_SHIFT  (BITS_IN_JSAMPLE-HIST_C0_BITS)
+#define C1_SHIFT  (BITS_IN_JSAMPLE-HIST_C1_BITS)
+#define C2_SHIFT  (BITS_IN_JSAMPLE-HIST_C2_BITS)
+
+
+typedef UINT16 histcell;	/* histogram cell; prefer an unsigned type */
+
+typedef histcell FAR * histptr;	/* for pointers to histogram cells */
+
+typedef histcell hist1d[HIST_C2_ELEMS]; /* typedefs for the array */
+typedef hist1d FAR * hist2d;	/* type for the 2nd-level pointers */
+typedef hist2d * hist3d;	/* type for top-level pointer */
+
+
+/* Declarations for Floyd-Steinberg dithering.
+ *
+ * Errors are accumulated into the array fserrors[], at a resolution of
+ * 1/16th of a pixel count.  The error at a given pixel is propagated
+ * to its not-yet-processed neighbors using the standard F-S fractions,
+ *		...	(here)	7/16
+ *		3/16	5/16	1/16
+ * We work left-to-right on even rows, right-to-left on odd rows.
+ *
+ * We can get away with a single array (holding one row's worth of errors)
+ * by using it to store the current row's errors at pixel columns not yet
+ * processed, but the next row's errors at columns already processed.  We
+ * need only a few extra variables to hold the errors immediately around the
+ * current column.  (If we are lucky, those variables are in registers, but
+ * even if not, they're probably cheaper to access than array elements are.)
+ *
+ * The fserrors[] array has (#columns + 2) entries; the extra entry at
+ * each end saves us from special-casing the first and last pixels.
+ * Each entry is three values long, one value for each color component.
+ *
+ * Note: on a wide image, we might not have enough room in a PC's near data
+ * segment to hold the error array; so it is allocated with alloc_large.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+typedef INT16 FSERROR;		/* 16 bits should be enough */
+typedef int LOCFSERROR;		/* use 'int' for calculation temps */
+#else
+typedef INT32 FSERROR;		/* may need more than 16 bits */
+typedef INT32 LOCFSERROR;	/* be sure calculation temps are big enough */
+#endif
+
+typedef FSERROR FAR *FSERRPTR;	/* pointer to error array (in FAR storage!) */
+
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_color_quantizer pub; /* public fields */
+
+  /* Space for the eventually created colormap is stashed here */
+  JSAMPARRAY sv_colormap;	/* colormap allocated at init time */
+  int desired;			/* desired # of colors = size of colormap */
+
+  /* Variables for accumulating image statistics */
+  hist3d histogram;		/* pointer to the histogram */
+
+  boolean needs_zeroed;		/* TRUE if next pass must zero histogram */
+
+  /* Variables for Floyd-Steinberg dithering */
+  FSERRPTR fserrors;		/* accumulated errors */
+  boolean on_odd_row;		/* flag to remember which row we are on */
+  int * error_limiter;		/* table for clamping the applied error */
+} my_cquantizer;
+
+typedef my_cquantizer * my_cquantize_ptr;
+
+
+/*
+ * Prescan some rows of pixels.
+ * In this module the prescan simply updates the histogram, which has been
+ * initialized to zeroes by start_pass.
+ * An output_buf parameter is required by the method signature, but no data
+ * is actually output (in fact the buffer controller is probably passing a
+ * NULL pointer).
+ */
+
+METHODDEF(void)
+prescan_quantize (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+		  JSAMPARRAY output_buf, int num_rows)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  register JSAMPROW ptr;
+  register histptr histp;
+  register hist3d histogram = cquantize->histogram;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    ptr = input_buf[row];
+    for (col = width; col > 0; col--) {
+      /* get pixel value and index into the histogram */
+      histp = & histogram[GETJSAMPLE(ptr[0]) >> C0_SHIFT]
+			 [GETJSAMPLE(ptr[1]) >> C1_SHIFT]
+			 [GETJSAMPLE(ptr[2]) >> C2_SHIFT];
+      /* increment, check for overflow and undo increment if so. */
+      if (++(*histp) <= 0)
+	(*histp)--;
+      ptr += 3;
+    }
+  }
+}
+
+
+/*
+ * Next we have the really interesting routines: selection of a colormap
+ * given the completed histogram.
+ * These routines work with a list of "boxes", each representing a rectangular
+ * subset of the input color space (to histogram precision).
+ */
+
+typedef struct {
+  /* The bounds of the box (inclusive); expressed as histogram indexes */
+  int c0min, c0max;
+  int c1min, c1max;
+  int c2min, c2max;
+  /* The volume (actually 2-norm) of the box */
+  INT32 volume;
+  /* The number of nonzero histogram cells within this box */
+  long colorcount;
+} box;
+
+typedef box * boxptr;
+
+
+LOCAL(boxptr)
+find_biggest_color_pop (boxptr boxlist, int numboxes)
+/* Find the splittable box with the largest color population */
+/* Returns NULL if no splittable boxes remain */
+{
+  register boxptr boxp;
+  register int i;
+  register long maxc = 0;
+  boxptr which = NULL;
+  
+  for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) {
+    if (boxp->colorcount > maxc && boxp->volume > 0) {
+      which = boxp;
+      maxc = boxp->colorcount;
+    }
+  }
+  return which;
+}
+
+
+LOCAL(boxptr)
+find_biggest_volume (boxptr boxlist, int numboxes)
+/* Find the splittable box with the largest (scaled) volume */
+/* Returns NULL if no splittable boxes remain */
+{
+  register boxptr boxp;
+  register int i;
+  register INT32 maxv = 0;
+  boxptr which = NULL;
+  
+  for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) {
+    if (boxp->volume > maxv) {
+      which = boxp;
+      maxv = boxp->volume;
+    }
+  }
+  return which;
+}
+
+
+LOCAL(void)
+update_box (j_decompress_ptr cinfo, boxptr boxp)
+/* Shrink the min/max bounds of a box to enclose only nonzero elements, */
+/* and recompute its volume and population */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  histptr histp;
+  int c0,c1,c2;
+  int c0min,c0max,c1min,c1max,c2min,c2max;
+  INT32 dist0,dist1,dist2;
+  long ccount;
+  
+  c0min = boxp->c0min;  c0max = boxp->c0max;
+  c1min = boxp->c1min;  c1max = boxp->c1max;
+  c2min = boxp->c2min;  c2max = boxp->c2max;
+  
+  if (c0max > c0min)
+    for (c0 = c0min; c0 <= c0max; c0++)
+      for (c1 = c1min; c1 <= c1max; c1++) {
+	histp = & histogram[c0][c1][c2min];
+	for (c2 = c2min; c2 <= c2max; c2++)
+	  if (*histp++ != 0) {
+	    boxp->c0min = c0min = c0;
+	    goto have_c0min;
+	  }
+      }
+ have_c0min:
+  if (c0max > c0min)
+    for (c0 = c0max; c0 >= c0min; c0--)
+      for (c1 = c1min; c1 <= c1max; c1++) {
+	histp = & histogram[c0][c1][c2min];
+	for (c2 = c2min; c2 <= c2max; c2++)
+	  if (*histp++ != 0) {
+	    boxp->c0max = c0max = c0;
+	    goto have_c0max;
+	  }
+      }
+ have_c0max:
+  if (c1max > c1min)
+    for (c1 = c1min; c1 <= c1max; c1++)
+      for (c0 = c0min; c0 <= c0max; c0++) {
+	histp = & histogram[c0][c1][c2min];
+	for (c2 = c2min; c2 <= c2max; c2++)
+	  if (*histp++ != 0) {
+	    boxp->c1min = c1min = c1;
+	    goto have_c1min;
+	  }
+      }
+ have_c1min:
+  if (c1max > c1min)
+    for (c1 = c1max; c1 >= c1min; c1--)
+      for (c0 = c0min; c0 <= c0max; c0++) {
+	histp = & histogram[c0][c1][c2min];
+	for (c2 = c2min; c2 <= c2max; c2++)
+	  if (*histp++ != 0) {
+	    boxp->c1max = c1max = c1;
+	    goto have_c1max;
+	  }
+      }
+ have_c1max:
+  if (c2max > c2min)
+    for (c2 = c2min; c2 <= c2max; c2++)
+      for (c0 = c0min; c0 <= c0max; c0++) {
+	histp = & histogram[c0][c1min][c2];
+	for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
+	  if (*histp != 0) {
+	    boxp->c2min = c2min = c2;
+	    goto have_c2min;
+	  }
+      }
+ have_c2min:
+  if (c2max > c2min)
+    for (c2 = c2max; c2 >= c2min; c2--)
+      for (c0 = c0min; c0 <= c0max; c0++) {
+	histp = & histogram[c0][c1min][c2];
+	for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
+	  if (*histp != 0) {
+	    boxp->c2max = c2max = c2;
+	    goto have_c2max;
+	  }
+      }
+ have_c2max:
+
+  /* Update box volume.
+   * We use 2-norm rather than real volume here; this biases the method
+   * against making long narrow boxes, and it has the side benefit that
+   * a box is splittable iff norm > 0.
+   * Since the differences are expressed in histogram-cell units,
+   * we have to shift back to JSAMPLE units to get consistent distances;
+   * after which, we scale according to the selected distance scale factors.
+   */
+  dist0 = ((c0max - c0min) << C0_SHIFT) * C0_SCALE;
+  dist1 = ((c1max - c1min) << C1_SHIFT) * C1_SCALE;
+  dist2 = ((c2max - c2min) << C2_SHIFT) * C2_SCALE;
+  boxp->volume = dist0*dist0 + dist1*dist1 + dist2*dist2;
+  
+  /* Now scan remaining volume of box and compute population */
+  ccount = 0;
+  for (c0 = c0min; c0 <= c0max; c0++)
+    for (c1 = c1min; c1 <= c1max; c1++) {
+      histp = & histogram[c0][c1][c2min];
+      for (c2 = c2min; c2 <= c2max; c2++, histp++)
+	if (*histp != 0) {
+	  ccount++;
+	}
+    }
+  boxp->colorcount = ccount;
+}
+
+
+LOCAL(int)
+median_cut (j_decompress_ptr cinfo, boxptr boxlist, int numboxes,
+	    int desired_colors)
+/* Repeatedly select and split the largest box until we have enough boxes */
+{
+  int n,lb;
+  int c0,c1,c2,cmax;
+  register boxptr b1,b2;
+
+  while (numboxes < desired_colors) {
+    /* Select box to split.
+     * Current algorithm: by population for first half, then by volume.
+     */
+    if (numboxes*2 <= desired_colors) {
+      b1 = find_biggest_color_pop(boxlist, numboxes);
+    } else {
+      b1 = find_biggest_volume(boxlist, numboxes);
+    }
+    if (b1 == NULL)		/* no splittable boxes left! */
+      break;
+    b2 = &boxlist[numboxes];	/* where new box will go */
+    /* Copy the color bounds to the new box. */
+    b2->c0max = b1->c0max; b2->c1max = b1->c1max; b2->c2max = b1->c2max;
+    b2->c0min = b1->c0min; b2->c1min = b1->c1min; b2->c2min = b1->c2min;
+    /* Choose which axis to split the box on.
+     * Current algorithm: longest scaled axis.
+     * See notes in update_box about scaling distances.
+     */
+    c0 = ((b1->c0max - b1->c0min) << C0_SHIFT) * C0_SCALE;
+    c1 = ((b1->c1max - b1->c1min) << C1_SHIFT) * C1_SCALE;
+    c2 = ((b1->c2max - b1->c2min) << C2_SHIFT) * C2_SCALE;
+    /* We want to break any ties in favor of green, then red, blue last.
+     * This code does the right thing for R,G,B or B,G,R color orders only.
+     */
+#if RGB_RED == 0
+    cmax = c1; n = 1;
+    if (c0 > cmax) { cmax = c0; n = 0; }
+    if (c2 > cmax) { n = 2; }
+#else
+    cmax = c1; n = 1;
+    if (c2 > cmax) { cmax = c2; n = 2; }
+    if (c0 > cmax) { n = 0; }
+#endif
+    /* Choose split point along selected axis, and update box bounds.
+     * Current algorithm: split at halfway point.
+     * (Since the box has been shrunk to minimum volume,
+     * any split will produce two nonempty subboxes.)
+     * Note that lb value is max for lower box, so must be < old max.
+     */
+    switch (n) {
+    case 0:
+      lb = (b1->c0max + b1->c0min) / 2;
+      b1->c0max = lb;
+      b2->c0min = lb+1;
+      break;
+    case 1:
+      lb = (b1->c1max + b1->c1min) / 2;
+      b1->c1max = lb;
+      b2->c1min = lb+1;
+      break;
+    case 2:
+      lb = (b1->c2max + b1->c2min) / 2;
+      b1->c2max = lb;
+      b2->c2min = lb+1;
+      break;
+    }
+    /* Update stats for boxes */
+    update_box(cinfo, b1);
+    update_box(cinfo, b2);
+    numboxes++;
+  }
+  return numboxes;
+}
+
+
+LOCAL(void)
+compute_color (j_decompress_ptr cinfo, boxptr boxp, int icolor)
+/* Compute representative color for a box, put it in colormap[icolor] */
+{
+  /* Current algorithm: mean weighted by pixels (not colors) */
+  /* Note it is important to get the rounding correct! */
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  histptr histp;
+  int c0,c1,c2;
+  int c0min,c0max,c1min,c1max,c2min,c2max;
+  long count;
+  long total = 0;
+  long c0total = 0;
+  long c1total = 0;
+  long c2total = 0;
+  
+  c0min = boxp->c0min;  c0max = boxp->c0max;
+  c1min = boxp->c1min;  c1max = boxp->c1max;
+  c2min = boxp->c2min;  c2max = boxp->c2max;
+  
+  for (c0 = c0min; c0 <= c0max; c0++)
+    for (c1 = c1min; c1 <= c1max; c1++) {
+      histp = & histogram[c0][c1][c2min];
+      for (c2 = c2min; c2 <= c2max; c2++) {
+	if ((count = *histp++) != 0) {
+	  total += count;
+	  c0total += ((c0 << C0_SHIFT) + ((1<<C0_SHIFT)>>1)) * count;
+	  c1total += ((c1 << C1_SHIFT) + ((1<<C1_SHIFT)>>1)) * count;
+	  c2total += ((c2 << C2_SHIFT) + ((1<<C2_SHIFT)>>1)) * count;
+	}
+      }
+    }
+  
+  cinfo->colormap[0][icolor] = (JSAMPLE) ((c0total + (total>>1)) / total);
+  cinfo->colormap[1][icolor] = (JSAMPLE) ((c1total + (total>>1)) / total);
+  cinfo->colormap[2][icolor] = (JSAMPLE) ((c2total + (total>>1)) / total);
+}
+
+
+LOCAL(void)
+select_colors (j_decompress_ptr cinfo, int desired_colors)
+/* Master routine for color selection */
+{
+  boxptr boxlist;
+  int numboxes;
+  int i;
+
+  /* Allocate workspace for box list */
+  boxlist = (boxptr) (*cinfo->mem->alloc_small)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, desired_colors * SIZEOF(box));
+  /* Initialize one box containing whole space */
+  numboxes = 1;
+  boxlist[0].c0min = 0;
+  boxlist[0].c0max = MAXJSAMPLE >> C0_SHIFT;
+  boxlist[0].c1min = 0;
+  boxlist[0].c1max = MAXJSAMPLE >> C1_SHIFT;
+  boxlist[0].c2min = 0;
+  boxlist[0].c2max = MAXJSAMPLE >> C2_SHIFT;
+  /* Shrink it to actually-used volume and set its statistics */
+  update_box(cinfo, & boxlist[0]);
+  /* Perform median-cut to produce final box list */
+  numboxes = median_cut(cinfo, boxlist, numboxes, desired_colors);
+  /* Compute the representative color for each box, fill colormap */
+  for (i = 0; i < numboxes; i++)
+    compute_color(cinfo, & boxlist[i], i);
+  cinfo->actual_number_of_colors = numboxes;
+  TRACEMS1(cinfo, 1, JTRC_QUANT_SELECTED, numboxes);
+}
+
+
+/*
+ * These routines are concerned with the time-critical task of mapping input
+ * colors to the nearest color in the selected colormap.
+ *
+ * We re-use the histogram space as an "inverse color map", essentially a
+ * cache for the results of nearest-color searches.  All colors within a
+ * histogram cell will be mapped to the same colormap entry, namely the one
+ * closest to the cell's center.  This may not be quite the closest entry to
+ * the actual input color, but it's almost as good.  A zero in the cache
+ * indicates we haven't found the nearest color for that cell yet; the array
+ * is cleared to zeroes before starting the mapping pass.  When we find the
+ * nearest color for a cell, its colormap index plus one is recorded in the
+ * cache for future use.  The pass2 scanning routines call fill_inverse_cmap
+ * when they need to use an unfilled entry in the cache.
+ *
+ * Our method of efficiently finding nearest colors is based on the "locally
+ * sorted search" idea described by Heckbert and on the incremental distance
+ * calculation described by Spencer W. Thomas in chapter III.1 of Graphics
+ * Gems II (James Arvo, ed.  Academic Press, 1991).  Thomas points out that
+ * the distances from a given colormap entry to each cell of the histogram can
+ * be computed quickly using an incremental method: the differences between
+ * distances to adjacent cells themselves differ by a constant.  This allows a
+ * fairly fast implementation of the "brute force" approach of computing the
+ * distance from every colormap entry to every histogram cell.  Unfortunately,
+ * it needs a work array to hold the best-distance-so-far for each histogram
+ * cell (because the inner loop has to be over cells, not colormap entries).
+ * The work array elements have to be INT32s, so the work array would need
+ * 256Kb at our recommended precision.  This is not feasible in DOS machines.
+ *
+ * To get around these problems, we apply Thomas' method to compute the
+ * nearest colors for only the cells within a small subbox of the histogram.
+ * The work array need be only as big as the subbox, so the memory usage
+ * problem is solved.  Furthermore, we need not fill subboxes that are never
+ * referenced in pass2; many images use only part of the color gamut, so a
+ * fair amount of work is saved.  An additional advantage of this
+ * approach is that we can apply Heckbert's locality criterion to quickly
+ * eliminate colormap entries that are far away from the subbox; typically
+ * three-fourths of the colormap entries are rejected by Heckbert's criterion,
+ * and we need not compute their distances to individual cells in the subbox.
+ * The speed of this approach is heavily influenced by the subbox size: too
+ * small means too much overhead, too big loses because Heckbert's criterion
+ * can't eliminate as many colormap entries.  Empirically the best subbox
+ * size seems to be about 1/512th of the histogram (1/8th in each direction).
+ *
+ * Thomas' article also describes a refined method which is asymptotically
+ * faster than the brute-force method, but it is also far more complex and
+ * cannot efficiently be applied to small subboxes.  It is therefore not
+ * useful for programs intended to be portable to DOS machines.  On machines
+ * with plenty of memory, filling the whole histogram in one shot with Thomas'
+ * refined method might be faster than the present code --- but then again,
+ * it might not be any faster, and it's certainly more complicated.
+ */
+
+
+/* log2(histogram cells in update box) for each axis; this can be adjusted */
+#define BOX_C0_LOG  (HIST_C0_BITS-3)
+#define BOX_C1_LOG  (HIST_C1_BITS-3)
+#define BOX_C2_LOG  (HIST_C2_BITS-3)
+
+#define BOX_C0_ELEMS  (1<<BOX_C0_LOG) /* # of hist cells in update box */
+#define BOX_C1_ELEMS  (1<<BOX_C1_LOG)
+#define BOX_C2_ELEMS  (1<<BOX_C2_LOG)
+
+#define BOX_C0_SHIFT  (C0_SHIFT + BOX_C0_LOG)
+#define BOX_C1_SHIFT  (C1_SHIFT + BOX_C1_LOG)
+#define BOX_C2_SHIFT  (C2_SHIFT + BOX_C2_LOG)
+
+
+/*
+ * The next three routines implement inverse colormap filling.  They could
+ * all be folded into one big routine, but splitting them up this way saves
+ * some stack space (the mindist[] and bestdist[] arrays need not coexist)
+ * and may allow some compilers to produce better code by registerizing more
+ * inner-loop variables.
+ */
+
+LOCAL(int)
+find_nearby_colors (j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
+		    JSAMPLE colorlist[])
+/* Locate the colormap entries close enough to an update box to be candidates
+ * for the nearest entry to some cell(s) in the update box.  The update box
+ * is specified by the center coordinates of its first cell.  The number of
+ * candidate colormap entries is returned, and their colormap indexes are
+ * placed in colorlist[].
+ * This routine uses Heckbert's "locally sorted search" criterion to select
+ * the colors that need further consideration.
+ */
+{
+  int numcolors = cinfo->actual_number_of_colors;
+  int maxc0, maxc1, maxc2;
+  int centerc0, centerc1, centerc2;
+  int i, x, ncolors;
+  INT32 minmaxdist, min_dist, max_dist, tdist;
+  INT32 mindist[MAXNUMCOLORS];	/* min distance to colormap entry i */
+
+  /* Compute true coordinates of update box's upper corner and center.
+   * Actually we compute the coordinates of the center of the upper-corner
+   * histogram cell, which are the upper bounds of the volume we care about.
+   * Note that since ">>" rounds down, the "center" values may be closer to
+   * min than to max; hence comparisons to them must be "<=", not "<".
+   */
+  maxc0 = minc0 + ((1 << BOX_C0_SHIFT) - (1 << C0_SHIFT));
+  centerc0 = (minc0 + maxc0) >> 1;
+  maxc1 = minc1 + ((1 << BOX_C1_SHIFT) - (1 << C1_SHIFT));
+  centerc1 = (minc1 + maxc1) >> 1;
+  maxc2 = minc2 + ((1 << BOX_C2_SHIFT) - (1 << C2_SHIFT));
+  centerc2 = (minc2 + maxc2) >> 1;
+
+  /* For each color in colormap, find:
+   *  1. its minimum squared-distance to any point in the update box
+   *     (zero if color is within update box);
+   *  2. its maximum squared-distance to any point in the update box.
+   * Both of these can be found by considering only the corners of the box.
+   * We save the minimum distance for each color in mindist[];
+   * only the smallest maximum distance is of interest.
+   */
+  minmaxdist = 0x7FFFFFFFL;
+
+  for (i = 0; i < numcolors; i++) {
+    /* We compute the squared-c0-distance term, then add in the other two. */
+    x = GETJSAMPLE(cinfo->colormap[0][i]);
+    if (x < minc0) {
+      tdist = (x - minc0) * C0_SCALE;
+      min_dist = tdist*tdist;
+      tdist = (x - maxc0) * C0_SCALE;
+      max_dist = tdist*tdist;
+    } else if (x > maxc0) {
+      tdist = (x - maxc0) * C0_SCALE;
+      min_dist = tdist*tdist;
+      tdist = (x - minc0) * C0_SCALE;
+      max_dist = tdist*tdist;
+    } else {
+      /* within cell range so no contribution to min_dist */
+      min_dist = 0;
+      if (x <= centerc0) {
+	tdist = (x - maxc0) * C0_SCALE;
+	max_dist = tdist*tdist;
+      } else {
+	tdist = (x - minc0) * C0_SCALE;
+	max_dist = tdist*tdist;
+      }
+    }
+
+    x = GETJSAMPLE(cinfo->colormap[1][i]);
+    if (x < minc1) {
+      tdist = (x - minc1) * C1_SCALE;
+      min_dist += tdist*tdist;
+      tdist = (x - maxc1) * C1_SCALE;
+      max_dist += tdist*tdist;
+    } else if (x > maxc1) {
+      tdist = (x - maxc1) * C1_SCALE;
+      min_dist += tdist*tdist;
+      tdist = (x - minc1) * C1_SCALE;
+      max_dist += tdist*tdist;
+    } else {
+      /* within cell range so no contribution to min_dist */
+      if (x <= centerc1) {
+	tdist = (x - maxc1) * C1_SCALE;
+	max_dist += tdist*tdist;
+      } else {
+	tdist = (x - minc1) * C1_SCALE;
+	max_dist += tdist*tdist;
+      }
+    }
+
+    x = GETJSAMPLE(cinfo->colormap[2][i]);
+    if (x < minc2) {
+      tdist = (x - minc2) * C2_SCALE;
+      min_dist += tdist*tdist;
+      tdist = (x - maxc2) * C2_SCALE;
+      max_dist += tdist*tdist;
+    } else if (x > maxc2) {
+      tdist = (x - maxc2) * C2_SCALE;
+      min_dist += tdist*tdist;
+      tdist = (x - minc2) * C2_SCALE;
+      max_dist += tdist*tdist;
+    } else {
+      /* within cell range so no contribution to min_dist */
+      if (x <= centerc2) {
+	tdist = (x - maxc2) * C2_SCALE;
+	max_dist += tdist*tdist;
+      } else {
+	tdist = (x - minc2) * C2_SCALE;
+	max_dist += tdist*tdist;
+      }
+    }
+
+    mindist[i] = min_dist;	/* save away the results */
+    if (max_dist < minmaxdist)
+      minmaxdist = max_dist;
+  }
+
+  /* Now we know that no cell in the update box is more than minmaxdist
+   * away from some colormap entry.  Therefore, only colors that are
+   * within minmaxdist of some part of the box need be considered.
+   */
+  ncolors = 0;
+  for (i = 0; i < numcolors; i++) {
+    if (mindist[i] <= minmaxdist)
+      colorlist[ncolors++] = (JSAMPLE) i;
+  }
+  return ncolors;
+}
+
+
+LOCAL(void)
+find_best_colors (j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
+		  int numcolors, JSAMPLE colorlist[], JSAMPLE bestcolor[])
+/* Find the closest colormap entry for each cell in the update box,
+ * given the list of candidate colors prepared by find_nearby_colors.
+ * Return the indexes of the closest entries in the bestcolor[] array.
+ * This routine uses Thomas' incremental distance calculation method to
+ * find the distance from a colormap entry to successive cells in the box.
+ */
+{
+  int ic0, ic1, ic2;
+  int i, icolor;
+  register INT32 * bptr;	/* pointer into bestdist[] array */
+  JSAMPLE * cptr;		/* pointer into bestcolor[] array */
+  INT32 dist0, dist1;		/* initial distance values */
+  register INT32 dist2;		/* current distance in inner loop */
+  INT32 xx0, xx1;		/* distance increments */
+  register INT32 xx2;
+  INT32 inc0, inc1, inc2;	/* initial values for increments */
+  /* This array holds the distance to the nearest-so-far color for each cell */
+  INT32 bestdist[BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS];
+
+  /* Initialize best-distance for each cell of the update box */
+  bptr = bestdist;
+  for (i = BOX_C0_ELEMS*BOX_C1_ELEMS*BOX_C2_ELEMS-1; i >= 0; i--)
+    *bptr++ = 0x7FFFFFFFL;
+  
+  /* For each color selected by find_nearby_colors,
+   * compute its distance to the center of each cell in the box.
+   * If that's less than best-so-far, update best distance and color number.
+   */
+  
+  /* Nominal steps between cell centers ("x" in Thomas article) */
+#define STEP_C0  ((1 << C0_SHIFT) * C0_SCALE)
+#define STEP_C1  ((1 << C1_SHIFT) * C1_SCALE)
+#define STEP_C2  ((1 << C2_SHIFT) * C2_SCALE)
+  
+  for (i = 0; i < numcolors; i++) {
+    icolor = GETJSAMPLE(colorlist[i]);
+    /* Compute (square of) distance from minc0/c1/c2 to this color */
+    inc0 = (minc0 - GETJSAMPLE(cinfo->colormap[0][icolor])) * C0_SCALE;
+    dist0 = inc0*inc0;
+    inc1 = (minc1 - GETJSAMPLE(cinfo->colormap[1][icolor])) * C1_SCALE;
+    dist0 += inc1*inc1;
+    inc2 = (minc2 - GETJSAMPLE(cinfo->colormap[2][icolor])) * C2_SCALE;
+    dist0 += inc2*inc2;
+    /* Form the initial difference increments */
+    inc0 = inc0 * (2 * STEP_C0) + STEP_C0 * STEP_C0;
+    inc1 = inc1 * (2 * STEP_C1) + STEP_C1 * STEP_C1;
+    inc2 = inc2 * (2 * STEP_C2) + STEP_C2 * STEP_C2;
+    /* Now loop over all cells in box, updating distance per Thomas method */
+    bptr = bestdist;
+    cptr = bestcolor;
+    xx0 = inc0;
+    for (ic0 = BOX_C0_ELEMS-1; ic0 >= 0; ic0--) {
+      dist1 = dist0;
+      xx1 = inc1;
+      for (ic1 = BOX_C1_ELEMS-1; ic1 >= 0; ic1--) {
+	dist2 = dist1;
+	xx2 = inc2;
+	for (ic2 = BOX_C2_ELEMS-1; ic2 >= 0; ic2--) {
+	  if (dist2 < *bptr) {
+	    *bptr = dist2;
+	    *cptr = (JSAMPLE) icolor;
+	  }
+	  dist2 += xx2;
+	  xx2 += 2 * STEP_C2 * STEP_C2;
+	  bptr++;
+	  cptr++;
+	}
+	dist1 += xx1;
+	xx1 += 2 * STEP_C1 * STEP_C1;
+      }
+      dist0 += xx0;
+      xx0 += 2 * STEP_C0 * STEP_C0;
+    }
+  }
+}
+
+
+LOCAL(void)
+fill_inverse_cmap (j_decompress_ptr cinfo, int c0, int c1, int c2)
+/* Fill the inverse-colormap entries in the update box that contains */
+/* histogram cell c0/c1/c2.  (Only that one cell MUST be filled, but */
+/* we can fill as many others as we wish.) */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  int minc0, minc1, minc2;	/* lower left corner of update box */
+  int ic0, ic1, ic2;
+  register JSAMPLE * cptr;	/* pointer into bestcolor[] array */
+  register histptr cachep;	/* pointer into main cache array */
+  /* This array lists the candidate colormap indexes. */
+  JSAMPLE colorlist[MAXNUMCOLORS];
+  int numcolors;		/* number of candidate colors */
+  /* This array holds the actually closest colormap index for each cell. */
+  JSAMPLE bestcolor[BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS];
+
+  /* Convert cell coordinates to update box ID */
+  c0 >>= BOX_C0_LOG;
+  c1 >>= BOX_C1_LOG;
+  c2 >>= BOX_C2_LOG;
+
+  /* Compute true coordinates of update box's origin corner.
+   * Actually we compute the coordinates of the center of the corner
+   * histogram cell, which are the lower bounds of the volume we care about.
+   */
+  minc0 = (c0 << BOX_C0_SHIFT) + ((1 << C0_SHIFT) >> 1);
+  minc1 = (c1 << BOX_C1_SHIFT) + ((1 << C1_SHIFT) >> 1);
+  minc2 = (c2 << BOX_C2_SHIFT) + ((1 << C2_SHIFT) >> 1);
+  
+  /* Determine which colormap entries are close enough to be candidates
+   * for the nearest entry to some cell in the update box.
+   */
+  numcolors = find_nearby_colors(cinfo, minc0, minc1, minc2, colorlist);
+
+  /* Determine the actually nearest colors. */
+  find_best_colors(cinfo, minc0, minc1, minc2, numcolors, colorlist,
+		   bestcolor);
+
+  /* Save the best color numbers (plus 1) in the main cache array */
+  c0 <<= BOX_C0_LOG;		/* convert ID back to base cell indexes */
+  c1 <<= BOX_C1_LOG;
+  c2 <<= BOX_C2_LOG;
+  cptr = bestcolor;
+  for (ic0 = 0; ic0 < BOX_C0_ELEMS; ic0++) {
+    for (ic1 = 0; ic1 < BOX_C1_ELEMS; ic1++) {
+      cachep = & histogram[c0+ic0][c1+ic1][c2];
+      for (ic2 = 0; ic2 < BOX_C2_ELEMS; ic2++) {
+	*cachep++ = (histcell) (GETJSAMPLE(*cptr++) + 1);
+      }
+    }
+  }
+}
+
+
+/*
+ * Map some rows of pixels to the output colormapped representation.
+ */
+
+METHODDEF(void)
+pass2_no_dither (j_decompress_ptr cinfo,
+		 JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows)
+/* This version performs no dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  register JSAMPROW inptr, outptr;
+  register histptr cachep;
+  register int c0, c1, c2;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    inptr = input_buf[row];
+    outptr = output_buf[row];
+    for (col = width; col > 0; col--) {
+      /* get pixel value and index into the cache */
+      c0 = GETJSAMPLE(*inptr++) >> C0_SHIFT;
+      c1 = GETJSAMPLE(*inptr++) >> C1_SHIFT;
+      c2 = GETJSAMPLE(*inptr++) >> C2_SHIFT;
+      cachep = & histogram[c0][c1][c2];
+      /* If we have not seen this color before, find nearest colormap entry */
+      /* and update the cache */
+      if (*cachep == 0)
+	fill_inverse_cmap(cinfo, c0,c1,c2);
+      /* Now emit the colormap index for this cell */
+      *outptr++ = (JSAMPLE) (*cachep - 1);
+    }
+  }
+}
+
+
+METHODDEF(void)
+pass2_fs_dither (j_decompress_ptr cinfo,
+		 JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows)
+/* This version performs Floyd-Steinberg dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  register LOCFSERROR cur0, cur1, cur2;	/* current error or pixel value */
+  LOCFSERROR belowerr0, belowerr1, belowerr2; /* error for pixel below cur */
+  LOCFSERROR bpreverr0, bpreverr1, bpreverr2; /* error for below/prev col */
+  register FSERRPTR errorptr;	/* => fserrors[] at column before current */
+  JSAMPROW inptr;		/* => current input pixel */
+  JSAMPROW outptr;		/* => current output pixel */
+  histptr cachep;
+  int dir;			/* +1 or -1 depending on direction */
+  int dir3;			/* 3*dir, for advancing inptr & errorptr */
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+  JSAMPLE *range_limit = cinfo->sample_range_limit;
+  int *error_limit = cquantize->error_limiter;
+  JSAMPROW colormap0 = cinfo->colormap[0];
+  JSAMPROW colormap1 = cinfo->colormap[1];
+  JSAMPROW colormap2 = cinfo->colormap[2];
+  SHIFT_TEMPS
+
+  for (row = 0; row < num_rows; row++) {
+    inptr = input_buf[row];
+    outptr = output_buf[row];
+    if (cquantize->on_odd_row) {
+      /* work right to left in this row */
+      inptr += (width-1) * 3;	/* so point to rightmost pixel */
+      outptr += width-1;
+      dir = -1;
+      dir3 = -3;
+      errorptr = cquantize->fserrors + (width+1)*3; /* => entry after last column */
+      cquantize->on_odd_row = FALSE; /* flip for next time */
+    } else {
+      /* work left to right in this row */
+      dir = 1;
+      dir3 = 3;
+      errorptr = cquantize->fserrors; /* => entry before first real column */
+      cquantize->on_odd_row = TRUE; /* flip for next time */
+    }
+    /* Preset error values: no error propagated to first pixel from left */
+    cur0 = cur1 = cur2 = 0;
+    /* and no error propagated to row below yet */
+    belowerr0 = belowerr1 = belowerr2 = 0;
+    bpreverr0 = bpreverr1 = bpreverr2 = 0;
+
+    for (col = width; col > 0; col--) {
+      /* curN holds the error propagated from the previous pixel on the
+       * current line.  Add the error propagated from the previous line
+       * to form the complete error correction term for this pixel, and
+       * round the error term (which is expressed * 16) to an integer.
+       * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct
+       * for either sign of the error value.
+       * Note: errorptr points to *previous* column's array entry.
+       */
+      cur0 = RIGHT_SHIFT(cur0 + errorptr[dir3+0] + 8, 4);
+      cur1 = RIGHT_SHIFT(cur1 + errorptr[dir3+1] + 8, 4);
+      cur2 = RIGHT_SHIFT(cur2 + errorptr[dir3+2] + 8, 4);
+      /* Limit the error using transfer function set by init_error_limit.
+       * See comments with init_error_limit for rationale.
+       */
+      cur0 = error_limit[cur0];
+      cur1 = error_limit[cur1];
+      cur2 = error_limit[cur2];
+      /* Form pixel value + error, and range-limit to 0..MAXJSAMPLE.
+       * The maximum error is +- MAXJSAMPLE (or less with error limiting);
+       * this sets the required size of the range_limit array.
+       */
+      cur0 += GETJSAMPLE(inptr[0]);
+      cur1 += GETJSAMPLE(inptr[1]);
+      cur2 += GETJSAMPLE(inptr[2]);
+      cur0 = GETJSAMPLE(range_limit[cur0]);
+      cur1 = GETJSAMPLE(range_limit[cur1]);
+      cur2 = GETJSAMPLE(range_limit[cur2]);
+      /* Index into the cache with adjusted pixel value */
+      cachep = & histogram[cur0>>C0_SHIFT][cur1>>C1_SHIFT][cur2>>C2_SHIFT];
+      /* If we have not seen this color before, find nearest colormap */
+      /* entry and update the cache */
+      if (*cachep == 0)
+	fill_inverse_cmap(cinfo, cur0>>C0_SHIFT,cur1>>C1_SHIFT,cur2>>C2_SHIFT);
+      /* Now emit the colormap index for this cell */
+      { register int pixcode = *cachep - 1;
+	*outptr = (JSAMPLE) pixcode;
+	/* Compute representation error for this pixel */
+	cur0 -= GETJSAMPLE(colormap0[pixcode]);
+	cur1 -= GETJSAMPLE(colormap1[pixcode]);
+	cur2 -= GETJSAMPLE(colormap2[pixcode]);
+      }
+      /* Compute error fractions to be propagated to adjacent pixels.
+       * Add these into the running sums, and simultaneously shift the
+       * next-line error sums left by 1 column.
+       */
+      { register LOCFSERROR bnexterr, delta;
+
+	bnexterr = cur0;	/* Process component 0 */
+	delta = cur0 * 2;
+	cur0 += delta;		/* form error * 3 */
+	errorptr[0] = (FSERROR) (bpreverr0 + cur0);
+	cur0 += delta;		/* form error * 5 */
+	bpreverr0 = belowerr0 + cur0;
+	belowerr0 = bnexterr;
+	cur0 += delta;		/* form error * 7 */
+	bnexterr = cur1;	/* Process component 1 */
+	delta = cur1 * 2;
+	cur1 += delta;		/* form error * 3 */
+	errorptr[1] = (FSERROR) (bpreverr1 + cur1);
+	cur1 += delta;		/* form error * 5 */
+	bpreverr1 = belowerr1 + cur1;
+	belowerr1 = bnexterr;
+	cur1 += delta;		/* form error * 7 */
+	bnexterr = cur2;	/* Process component 2 */
+	delta = cur2 * 2;
+	cur2 += delta;		/* form error * 3 */
+	errorptr[2] = (FSERROR) (bpreverr2 + cur2);
+	cur2 += delta;		/* form error * 5 */
+	bpreverr2 = belowerr2 + cur2;
+	belowerr2 = bnexterr;
+	cur2 += delta;		/* form error * 7 */
+      }
+      /* At this point curN contains the 7/16 error value to be propagated
+       * to the next pixel on the current line, and all the errors for the
+       * next line have been shifted over.  We are therefore ready to move on.
+       */
+      inptr += dir3;		/* Advance pixel pointers to next column */
+      outptr += dir;
+      errorptr += dir3;		/* advance errorptr to current column */
+    }
+    /* Post-loop cleanup: we must unload the final error values into the
+     * final fserrors[] entry.  Note we need not unload belowerrN because
+     * it is for the dummy column before or after the actual array.
+     */
+    errorptr[0] = (FSERROR) bpreverr0; /* unload prev errs into array */
+    errorptr[1] = (FSERROR) bpreverr1;
+    errorptr[2] = (FSERROR) bpreverr2;
+  }
+}
+
+
+/*
+ * Initialize the error-limiting transfer function (lookup table).
+ * The raw F-S error computation can potentially compute error values of up to
+ * +- MAXJSAMPLE.  But we want the maximum correction applied to a pixel to be
+ * much less, otherwise obviously wrong pixels will be created.  (Typical
+ * effects include weird fringes at color-area boundaries, isolated bright
+ * pixels in a dark area, etc.)  The standard advice for avoiding this problem
+ * is to ensure that the "corners" of the color cube are allocated as output
+ * colors; then repeated errors in the same direction cannot cause cascading
+ * error buildup.  However, that only prevents the error from getting
+ * completely out of hand; Aaron Giles reports that error limiting improves
+ * the results even with corner colors allocated.
+ * A simple clamping of the error values to about +- MAXJSAMPLE/8 works pretty
+ * well, but the smoother transfer function used below is even better.  Thanks
+ * to Aaron Giles for this idea.
+ */
+
+LOCAL(void)
+init_error_limit (j_decompress_ptr cinfo)
+/* Allocate and fill in the error_limiter table */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  int * table;
+  int in, out;
+
+  table = (int *) (*cinfo->mem->alloc_small)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE*2+1) * SIZEOF(int));
+  table += MAXJSAMPLE;		/* so can index -MAXJSAMPLE .. +MAXJSAMPLE */
+  cquantize->error_limiter = table;
+
+#define STEPSIZE ((MAXJSAMPLE+1)/16)
+  /* Map errors 1:1 up to +- MAXJSAMPLE/16 */
+  out = 0;
+  for (in = 0; in < STEPSIZE; in++, out++) {
+    table[in] = out; table[-in] = -out;
+  }
+  /* Map errors 1:2 up to +- 3*MAXJSAMPLE/16 */
+  for (; in < STEPSIZE*3; in++, out += (in&1) ? 0 : 1) {
+    table[in] = out; table[-in] = -out;
+  }
+  /* Clamp the rest to final out value (which is (MAXJSAMPLE+1)/8) */
+  for (; in <= MAXJSAMPLE; in++) {
+    table[in] = out; table[-in] = -out;
+  }
+#undef STEPSIZE
+}
+
+
+/*
+ * Finish up at the end of each pass.
+ */
+
+METHODDEF(void)
+finish_pass1 (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+
+  /* Select the representative colors and fill in cinfo->colormap */
+  cinfo->colormap = cquantize->sv_colormap;
+  select_colors(cinfo, cquantize->desired);
+  /* Force next pass to zero the color index table */
+  cquantize->needs_zeroed = TRUE;
+}
+
+
+METHODDEF(void)
+finish_pass2 (j_decompress_ptr cinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * Initialize for each processing pass.
+ */
+
+METHODDEF(void)
+start_pass_2_quant (j_decompress_ptr cinfo, boolean is_pre_scan)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  int i;
+
+  /* Only F-S dithering or no dithering is supported. */
+  /* If user asks for ordered dither, give him F-S. */
+  if (cinfo->dither_mode != JDITHER_NONE)
+    cinfo->dither_mode = JDITHER_FS;
+
+  if (is_pre_scan) {
+    /* Set up method pointers */
+    cquantize->pub.color_quantize = prescan_quantize;
+    cquantize->pub.finish_pass = finish_pass1;
+    cquantize->needs_zeroed = TRUE; /* Always zero histogram */
+  } else {
+    /* Set up method pointers */
+    if (cinfo->dither_mode == JDITHER_FS)
+      cquantize->pub.color_quantize = pass2_fs_dither;
+    else
+      cquantize->pub.color_quantize = pass2_no_dither;
+    cquantize->pub.finish_pass = finish_pass2;
+
+    /* Make sure color count is acceptable */
+    i = cinfo->actual_number_of_colors;
+    if (i < 1)
+      ERREXIT1(cinfo, JERR_QUANT_FEW_COLORS, 1);
+    if (i > MAXNUMCOLORS)
+      ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXNUMCOLORS);
+
+    if (cinfo->dither_mode == JDITHER_FS) {
+      size_t arraysize = (size_t) ((cinfo->output_width + 2) *
+				   (3 * SIZEOF(FSERROR)));
+      /* Allocate Floyd-Steinberg workspace if we didn't already. */
+      if (cquantize->fserrors == NULL)
+	cquantize->fserrors = (FSERRPTR) (*cinfo->mem->alloc_large)
+	  ((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize);
+      /* Initialize the propagated errors to zero. */
+      jzero_far((void FAR *) cquantize->fserrors, arraysize);
+      /* Make the error-limit table if we didn't already. */
+      if (cquantize->error_limiter == NULL)
+	init_error_limit(cinfo);
+      cquantize->on_odd_row = FALSE;
+    }
+
+  }
+  /* Zero the histogram or inverse color map, if necessary */
+  if (cquantize->needs_zeroed) {
+    for (i = 0; i < HIST_C0_ELEMS; i++) {
+      jzero_far((void FAR *) histogram[i],
+		HIST_C1_ELEMS*HIST_C2_ELEMS * SIZEOF(histcell));
+    }
+    cquantize->needs_zeroed = FALSE;
+  }
+}
+
+
+/*
+ * Switch to a new external colormap between output passes.
+ */
+
+METHODDEF(void)
+new_color_map_2_quant (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+
+  /* Reset the inverse color map */
+  cquantize->needs_zeroed = TRUE;
+}
+
+
+/*
+ * Module initialization routine for 2-pass color quantization.
+ */
+
+GLOBAL(void)
+jinit_2pass_quantizer (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize;
+  int i;
+
+  cquantize = (my_cquantize_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_cquantizer));
+  cinfo->cquantize = (struct jpeg_color_quantizer *) cquantize;
+  cquantize->pub.start_pass = start_pass_2_quant;
+  cquantize->pub.new_color_map = new_color_map_2_quant;
+  cquantize->fserrors = NULL;	/* flag optional arrays not allocated */
+  cquantize->error_limiter = NULL;
+
+  /* Make sure jdmaster didn't give me a case I can't handle */
+  if (cinfo->out_color_components != 3)
+    ERREXIT(cinfo, JERR_NOTIMPL);
+
+  /* Allocate the histogram/inverse colormap storage */
+  cquantize->histogram = (hist3d) (*cinfo->mem->alloc_small)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, HIST_C0_ELEMS * SIZEOF(hist2d));
+  for (i = 0; i < HIST_C0_ELEMS; i++) {
+    cquantize->histogram[i] = (hist2d) (*cinfo->mem->alloc_large)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       HIST_C1_ELEMS*HIST_C2_ELEMS * SIZEOF(histcell));
+  }
+  cquantize->needs_zeroed = TRUE; /* histogram is garbage now */
+
+  /* Allocate storage for the completed colormap, if required.
+   * We do this now since it is FAR storage and may affect
+   * the memory manager's space calculations.
+   */
+  if (cinfo->enable_2pass_quant) {
+    /* Make sure color count is acceptable */
+    int desired = cinfo->desired_number_of_colors;
+    /* Lower bound on # of colors ... somewhat arbitrary as long as > 0 */
+    if (desired < 8)
+      ERREXIT1(cinfo, JERR_QUANT_FEW_COLORS, 8);
+    /* Make sure colormap indexes can be represented by JSAMPLEs */
+    if (desired > MAXNUMCOLORS)
+      ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXNUMCOLORS);
+    cquantize->sv_colormap = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo,JPOOL_IMAGE, (JDIMENSION) desired, (JDIMENSION) 3);
+    cquantize->desired = desired;
+  } else
+    cquantize->sv_colormap = NULL;
+
+  /* Only F-S dithering or no dithering is supported. */
+  /* If user asks for ordered dither, give him F-S. */
+  if (cinfo->dither_mode != JDITHER_NONE)
+    cinfo->dither_mode = JDITHER_FS;
+
+  /* Allocate Floyd-Steinberg workspace if necessary.
+   * This isn't really needed until pass 2, but again it is FAR storage.
+   * Although we will cope with a later change in dither_mode,
+   * we do not promise to honor max_memory_to_use if dither_mode changes.
+   */
+  if (cinfo->dither_mode == JDITHER_FS) {
+    cquantize->fserrors = (FSERRPTR) (*cinfo->mem->alloc_large)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (size_t) ((cinfo->output_width + 2) * (3 * SIZEOF(FSERROR))));
+    /* Might as well create the error-limiting table too. */
+    init_error_limit(cinfo);
+  }
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */
diff --git a/libjpegtwrp/jutils.c b/libjpegtwrp/jutils.c
new file mode 100644
index 000000000..616ad0511
--- /dev/null
+++ b/libjpegtwrp/jutils.c
@@ -0,0 +1,185 @@
+/*
+ * jutils.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains tables and miscellaneous utility routines needed
+ * for both compression and decompression.
+ * Note we prefix all global names with "j" to minimize conflicts with
+ * a surrounding application.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * jpeg_zigzag_order[i] is the zigzag-order position of the i'th element
+ * of a DCT block read in natural order (left to right, top to bottom).
+ */
+
+#if 0				/* This table is not actually needed in v6a */
+
+const int jpeg_zigzag_order[DCTSIZE2] = {
+   0,  1,  5,  6, 14, 15, 27, 28,
+   2,  4,  7, 13, 16, 26, 29, 42,
+   3,  8, 12, 17, 25, 30, 41, 43,
+   9, 11, 18, 24, 31, 40, 44, 53,
+  10, 19, 23, 32, 39, 45, 52, 54,
+  20, 22, 33, 38, 46, 51, 55, 60,
+  21, 34, 37, 47, 50, 56, 59, 61,
+  35, 36, 48, 49, 57, 58, 62, 63
+};
+
+#endif
+
+/*
+ * jpeg_natural_order[i] is the natural-order position of the i'th element
+ * of zigzag order.
+ *
+ * When reading corrupted data, the Huffman decoders could attempt
+ * to reference an entry beyond the end of this array (if the decoded
+ * zero run length reaches past the end of the block).  To prevent
+ * wild stores without adding an inner-loop test, we put some extra
+ * "63"s after the real entries.  This will cause the extra coefficient
+ * to be stored in location 63 of the block, not somewhere random.
+ * The worst case would be a run-length of 15, which means we need 16
+ * fake entries.
+ */
+
+const int jpeg_natural_order[DCTSIZE2+16] = {
+  0,  1,  8, 16,  9,  2,  3, 10,
+ 17, 24, 32, 25, 18, 11,  4,  5,
+ 12, 19, 26, 33, 40, 48, 41, 34,
+ 27, 20, 13,  6,  7, 14, 21, 28,
+ 35, 42, 49, 56, 57, 50, 43, 36,
+ 29, 22, 15, 23, 30, 37, 44, 51,
+ 58, 59, 52, 45, 38, 31, 39, 46,
+ 53, 60, 61, 54, 47, 55, 62, 63,
+ 63, 63, 63, 63, 63, 63, 63, 63, /* extra entries for safety in decoder */
+ 63, 63, 63, 63, 63, 63, 63, 63
+};
+
+
+/*
+ * Arithmetic utilities
+ */
+
+GLOBAL(long)
+jdiv_round_up (long a, long b)
+/* Compute a/b rounded up to next integer, ie, ceil(a/b) */
+/* Assumes a >= 0, b > 0 */
+{
+  return (a + b - 1L) / b;
+}
+
+
+GLOBAL(long)
+jround_up (long a, long b)
+/* Compute a rounded up to next multiple of b, ie, ceil(a/b)*b */
+/* Assumes a >= 0, b > 0 */
+{
+  a += b - 1L;
+  return a - (a % b);
+}
+
+GLOBAL(long)
+jmin (long a, long b)
+{
+  return a < b ? a : b;
+}
+
+
+/* On normal machines we can apply MEMCOPY() and MEMZERO() to sample arrays
+ * and coefficient-block arrays.  This won't work on 80x86 because the arrays
+ * are FAR and we're assuming a small-pointer memory model.  However, some
+ * DOS compilers provide far-pointer versions of memcpy() and memset() even
+ * in the small-model libraries.  These will be used if USE_FMEM is defined.
+ * Otherwise, the routines below do it the hard way.  (The performance cost
+ * is not all that great, because these routines aren't very heavily used.)
+ */
+
+#ifndef NEED_FAR_POINTERS	/* normal case, same as regular macros */
+#define FMEMCOPY(dest,src,size)	MEMCOPY(dest,src,size)
+#define FMEMZERO(target,size)	MEMZERO(target,size)
+#else				/* 80x86 case, define if we can */
+#ifdef USE_FMEM
+#define FMEMCOPY(dest,src,size)	_fmemcpy((void FAR *)(dest), (const void FAR *)(src), (size_t)(size))
+#define FMEMZERO(target,size)	_fmemset((void FAR *)(target), 0, (size_t)(size))
+#endif
+#endif
+
+
+GLOBAL(void)
+jcopy_sample_rows (JSAMPARRAY input_array, int source_row,
+		   JSAMPARRAY output_array, int dest_row,
+		   int num_rows, JDIMENSION num_cols)
+/* Copy some rows of samples from one place to another.
+ * num_rows rows are copied from input_array[source_row++]
+ * to output_array[dest_row++]; these areas may overlap for duplication.
+ * The source and destination arrays must be at least as wide as num_cols.
+ */
+{
+  register JSAMPROW inptr, outptr;
+#ifdef FMEMCOPY
+  register size_t count = (size_t) (num_cols * SIZEOF(JSAMPLE));
+#else
+  register JDIMENSION count;
+#endif
+  register int row;
+
+  input_array += source_row;
+  output_array += dest_row;
+
+  for (row = num_rows; row > 0; row--) {
+    inptr = *input_array++;
+    outptr = *output_array++;
+#ifdef FMEMCOPY
+    FMEMCOPY(outptr, inptr, count);
+#else
+    for (count = num_cols; count > 0; count--)
+      *outptr++ = *inptr++;	/* needn't bother with GETJSAMPLE() here */
+#endif
+  }
+}
+
+
+GLOBAL(void)
+jcopy_block_row (JBLOCKROW input_row, JBLOCKROW output_row,
+		 JDIMENSION num_blocks)
+/* Copy a row of coefficient blocks from one place to another. */
+{
+#ifdef FMEMCOPY
+  FMEMCOPY(output_row, input_row, num_blocks * (DCTSIZE2 * SIZEOF(JCOEF)));
+#else
+  register JCOEFPTR inptr, outptr;
+  register long count;
+
+  inptr = (JCOEFPTR) input_row;
+  outptr = (JCOEFPTR) output_row;
+  for (count = (long) num_blocks * DCTSIZE2; count > 0; count--) {
+    *outptr++ = *inptr++;
+  }
+#endif
+}
+
+
+GLOBAL(void)
+jzero_far (void FAR * target, size_t bytestozero)
+/* Zero out a chunk of FAR memory. */
+/* This might be sample-array data, block-array data, or alloc_large data. */
+{
+#ifdef FMEMZERO
+  FMEMZERO(target, bytestozero);
+#else
+  register char FAR * ptr = (char FAR *) target;
+  register size_t count;
+
+  for (count = bytestozero; count > 0; count--) {
+    *ptr++ = 0;
+  }
+#endif
+}
diff --git a/libjpegtwrp/jversion.h b/libjpegtwrp/jversion.h
new file mode 100644
index 000000000..6472c58d3
--- /dev/null
+++ b/libjpegtwrp/jversion.h
@@ -0,0 +1,14 @@
+/*
+ * jversion.h
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains software version identification.
+ */
+
+
+#define JVERSION	"6b  27-Mar-1998"
+
+#define JCOPYRIGHT	"Copyright (C) 1998, Thomas G. Lane"
diff --git a/libjpegtwrp/libjpeg.doc b/libjpegtwrp/libjpeg.doc
new file mode 100644
index 000000000..689b206c0
--- /dev/null
+++ b/libjpegtwrp/libjpeg.doc
@@ -0,0 +1,3006 @@
+USING THE IJG JPEG LIBRARY
+
+Copyright (C) 1994-1998, Thomas G. Lane.
+This file is part of the Independent JPEG Group's software.
+For conditions of distribution and use, see the accompanying README file.
+
+
+This file describes how to use the IJG JPEG library within an application
+program.  Read it if you want to write a program that uses the library.
+
+The file example.c provides heavily commented skeleton code for calling the
+JPEG library.  Also see jpeglib.h (the include file to be used by application
+programs) for full details about data structures and function parameter lists.
+The library source code, of course, is the ultimate reference.
+
+Note that there have been *major* changes from the application interface
+presented by IJG version 4 and earlier versions.  The old design had several
+inherent limitations, and it had accumulated a lot of cruft as we added
+features while trying to minimize application-interface changes.  We have
+sacrificed backward compatibility in the version 5 rewrite, but we think the
+improvements justify this.
+
+
+TABLE OF CONTENTS
+-----------------
+
+Overview:
+	Functions provided by the library
+	Outline of typical usage
+Basic library usage:
+	Data formats
+	Compression details
+	Decompression details
+	Mechanics of usage: include files, linking, etc
+Advanced features:
+	Compression parameter selection
+	Decompression parameter selection
+	Special color spaces
+	Error handling
+	Compressed data handling (source and destination managers)
+	I/O suspension
+	Progressive JPEG support
+	Buffered-image mode
+	Abbreviated datastreams and multiple images
+	Special markers
+	Raw (downsampled) image data
+	Really raw data: DCT coefficients
+	Progress monitoring
+	Memory management
+	Memory usage
+	Library compile-time options
+	Portability considerations
+	Notes for MS-DOS implementors
+
+You should read at least the overview and basic usage sections before trying
+to program with the library.  The sections on advanced features can be read
+if and when you need them.
+
+
+OVERVIEW
+========
+
+Functions provided by the library
+---------------------------------
+
+The IJG JPEG library provides C code to read and write JPEG-compressed image
+files.  The surrounding application program receives or supplies image data a
+scanline at a time, using a straightforward uncompressed image format.  All
+details of color conversion and other preprocessing/postprocessing can be
+handled by the library.
+
+The library includes a substantial amount of code that is not covered by the
+JPEG standard but is necessary for typical applications of JPEG.  These
+functions preprocess the image before JPEG compression or postprocess it after
+decompression.  They include colorspace conversion, downsampling/upsampling,
+and color quantization.  The application indirectly selects use of this code
+by specifying the format in which it wishes to supply or receive image data.
+For example, if colormapped output is requested, then the decompression
+library automatically invokes color quantization.
+
+A wide range of quality vs. speed tradeoffs are possible in JPEG processing,
+and even more so in decompression postprocessing.  The decompression library
+provides multiple implementations that cover most of the useful tradeoffs,
+ranging from very-high-quality down to fast-preview operation.  On the
+compression side we have generally not provided low-quality choices, since
+compression is normally less time-critical.  It should be understood that the
+low-quality modes may not meet the JPEG standard's accuracy requirements;
+nonetheless, they are useful for viewers.
+
+A word about functions *not* provided by the library.  We handle a subset of
+the ISO JPEG standard; most baseline, extended-sequential, and progressive
+JPEG processes are supported.  (Our subset includes all features now in common
+use.)  Unsupported ISO options include:
+	* Hierarchical storage
+	* Lossless JPEG
+	* Arithmetic entropy coding (unsupported for legal reasons)
+	* DNL marker
+	* Nonintegral subsampling ratios
+We support both 8- and 12-bit data precision, but this is a compile-time
+choice rather than a run-time choice; hence it is difficult to use both
+precisions in a single application.
+
+By itself, the library handles only interchange JPEG datastreams --- in
+particular the widely used JFIF file format.  The library can be used by
+surrounding code to process interchange or abbreviated JPEG datastreams that
+are embedded in more complex file formats.  (For example, this library is
+used by the free LIBTIFF library to support JPEG compression in TIFF.)
+
+
+Outline of typical usage
+------------------------
+
+The rough outline of a JPEG compression operation is:
+
+	Allocate and initialize a JPEG compression object
+	Specify the destination for the compressed data (eg, a file)
+	Set parameters for compression, including image size & colorspace
+	jpeg_start_compress(...);
+	while (scan lines remain to be written)
+		jpeg_write_scanlines(...);
+	jpeg_finish_compress(...);
+	Release the JPEG compression object
+
+A JPEG compression object holds parameters and working state for the JPEG
+library.  We make creation/destruction of the object separate from starting
+or finishing compression of an image; the same object can be re-used for a
+series of image compression operations.  This makes it easy to re-use the
+same parameter settings for a sequence of images.  Re-use of a JPEG object
+also has important implications for processing abbreviated JPEG datastreams,
+as discussed later.
+
+The image data to be compressed is supplied to jpeg_write_scanlines() from
+in-memory buffers.  If the application is doing file-to-file compression,
+reading image data from the source file is the application's responsibility.
+The library emits compressed data by calling a "data destination manager",
+which typically will write the data into a file; but the application can
+provide its own destination manager to do something else.
+
+Similarly, the rough outline of a JPEG decompression operation is:
+
+	Allocate and initialize a JPEG decompression object
+	Specify the source of the compressed data (eg, a file)
+	Call jpeg_read_header() to obtain image info
+	Set parameters for decompression
+	jpeg_start_decompress(...);
+	while (scan lines remain to be read)
+		jpeg_read_scanlines(...);
+	jpeg_finish_decompress(...);
+	Release the JPEG decompression object
+
+This is comparable to the compression outline except that reading the
+datastream header is a separate step.  This is helpful because information
+about the image's size, colorspace, etc is available when the application
+selects decompression parameters.  For example, the application can choose an
+output scaling ratio that will fit the image into the available screen size.
+
+The decompression library obtains compressed data by calling a data source
+manager, which typically will read the data from a file; but other behaviors
+can be obtained with a custom source manager.  Decompressed data is delivered
+into in-memory buffers passed to jpeg_read_scanlines().
+
+It is possible to abort an incomplete compression or decompression operation
+by calling jpeg_abort(); or, if you do not need to retain the JPEG object,
+simply release it by calling jpeg_destroy().
+
+JPEG compression and decompression objects are two separate struct types.
+However, they share some common fields, and certain routines such as
+jpeg_destroy() can work on either type of object.
+
+The JPEG library has no static variables: all state is in the compression
+or decompression object.  Therefore it is possible to process multiple
+compression and decompression operations concurrently, using multiple JPEG
+objects.
+
+Both compression and decompression can be done in an incremental memory-to-
+memory fashion, if suitable source/destination managers are used.  See the
+section on "I/O suspension" for more details.
+
+
+BASIC LIBRARY USAGE
+===================
+
+Data formats
+------------
+
+Before diving into procedural details, it is helpful to understand the
+image data format that the JPEG library expects or returns.
+
+The standard input image format is a rectangular array of pixels, with each
+pixel having the same number of "component" or "sample" values (color
+channels).  You must specify how many components there are and the colorspace
+interpretation of the components.  Most applications will use RGB data
+(three components per pixel) or grayscale data (one component per pixel).
+PLEASE NOTE THAT RGB DATA IS THREE SAMPLES PER PIXEL, GRAYSCALE ONLY ONE.
+A remarkable number of people manage to miss this, only to find that their
+programs don't work with grayscale JPEG files.
+
+There is no provision for colormapped input.  JPEG files are always full-color
+or full grayscale (or sometimes another colorspace such as CMYK).  You can
+feed in a colormapped image by expanding it to full-color format.  However
+JPEG often doesn't work very well with source data that has been colormapped,
+because of dithering noise.  This is discussed in more detail in the JPEG FAQ
+and the other references mentioned in the README file.
+
+Pixels are stored by scanlines, with each scanline running from left to
+right.  The component values for each pixel are adjacent in the row; for
+example, R,G,B,R,G,B,R,G,B,... for 24-bit RGB color.  Each scanline is an
+array of data type JSAMPLE --- which is typically "unsigned char", unless
+you've changed jmorecfg.h.  (You can also change the RGB pixel layout, say
+to B,G,R order, by modifying jmorecfg.h.  But see the restrictions listed in
+that file before doing so.)
+
+A 2-D array of pixels is formed by making a list of pointers to the starts of
+scanlines; so the scanlines need not be physically adjacent in memory.  Even
+if you process just one scanline at a time, you must make a one-element
+pointer array to conform to this structure.  Pointers to JSAMPLE rows are of
+type JSAMPROW, and the pointer to the pointer array is of type JSAMPARRAY.
+
+The library accepts or supplies one or more complete scanlines per call.
+It is not possible to process part of a row at a time.  Scanlines are always
+processed top-to-bottom.  You can process an entire image in one call if you
+have it all in memory, but usually it's simplest to process one scanline at
+a time.
+
+For best results, source data values should have the precision specified by
+BITS_IN_JSAMPLE (normally 8 bits).  For instance, if you choose to compress
+data that's only 6 bits/channel, you should left-justify each value in a
+byte before passing it to the compressor.  If you need to compress data
+that has more than 8 bits/channel, compile with BITS_IN_JSAMPLE = 12.
+(See "Library compile-time options", later.)
+
+
+The data format returned by the decompressor is the same in all details,
+except that colormapped output is supported.  (Again, a JPEG file is never
+colormapped.  But you can ask the decompressor to perform on-the-fly color
+quantization to deliver colormapped output.)  If you request colormapped
+output then the returned data array contains a single JSAMPLE per pixel;
+its value is an index into a color map.  The color map is represented as
+a 2-D JSAMPARRAY in which each row holds the values of one color component,
+that is, colormap[i][j] is the value of the i'th color component for pixel
+value (map index) j.  Note that since the colormap indexes are stored in
+JSAMPLEs, the maximum number of colors is limited by the size of JSAMPLE
+(ie, at most 256 colors for an 8-bit JPEG library).
+
+
+Compression details
+-------------------
+
+Here we revisit the JPEG compression outline given in the overview.
+
+1. Allocate and initialize a JPEG compression object.
+
+A JPEG compression object is a "struct jpeg_compress_struct".  (It also has
+a bunch of subsidiary structures which are allocated via malloc(), but the
+application doesn't control those directly.)  This struct can be just a local
+variable in the calling routine, if a single routine is going to execute the
+whole JPEG compression sequence.  Otherwise it can be static or allocated
+from malloc().
+
+You will also need a structure representing a JPEG error handler.  The part
+of this that the library cares about is a "struct jpeg_error_mgr".  If you
+are providing your own error handler, you'll typically want to embed the
+jpeg_error_mgr struct in a larger structure; this is discussed later under
+"Error handling".  For now we'll assume you are just using the default error
+handler.  The default error handler will print JPEG error/warning messages
+on stderr, and it will call exit() if a fatal error occurs.
+
+You must initialize the error handler structure, store a pointer to it into
+the JPEG object's "err" field, and then call jpeg_create_compress() to
+initialize the rest of the JPEG object.
+
+Typical code for this step, if you are using the default error handler, is
+
+	struct jpeg_compress_struct cinfo;
+	struct jpeg_error_mgr jerr;
+	...
+	cinfo.err = jpeg_std_error(&jerr);
+	jpeg_create_compress(&cinfo);
+
+jpeg_create_compress allocates a small amount of memory, so it could fail
+if you are out of memory.  In that case it will exit via the error handler;
+that's why the error handler must be initialized first.
+
+
+2. Specify the destination for the compressed data (eg, a file).
+
+As previously mentioned, the JPEG library delivers compressed data to a
+"data destination" module.  The library includes one data destination
+module which knows how to write to a stdio stream.  You can use your own
+destination module if you want to do something else, as discussed later.
+
+If you use the standard destination module, you must open the target stdio
+stream beforehand.  Typical code for this step looks like:
+
+	FILE * outfile;
+	...
+	if ((outfile = fopen(filename, "wb")) == NULL) {
+	    fprintf(stderr, "can't open %s\n", filename);
+	    exit(1);
+	}
+	jpeg_stdio_dest(&cinfo, outfile);
+
+where the last line invokes the standard destination module.
+
+WARNING: it is critical that the binary compressed data be delivered to the
+output file unchanged.  On non-Unix systems the stdio library may perform
+newline translation or otherwise corrupt binary data.  To suppress this
+behavior, you may need to use a "b" option to fopen (as shown above), or use
+setmode() or another routine to put the stdio stream in binary mode.  See
+cjpeg.c and djpeg.c for code that has been found to work on many systems.
+
+You can select the data destination after setting other parameters (step 3),
+if that's more convenient.  You may not change the destination between
+calling jpeg_start_compress() and jpeg_finish_compress().
+
+
+3. Set parameters for compression, including image size & colorspace.
+
+You must supply information about the source image by setting the following
+fields in the JPEG object (cinfo structure):
+
+	image_width		Width of image, in pixels
+	image_height		Height of image, in pixels
+	input_components	Number of color channels (samples per pixel)
+	in_color_space		Color space of source image
+
+The image dimensions are, hopefully, obvious.  JPEG supports image dimensions
+of 1 to 64K pixels in either direction.  The input color space is typically
+RGB or grayscale, and input_components is 3 or 1 accordingly.  (See "Special
+color spaces", later, for more info.)  The in_color_space field must be
+assigned one of the J_COLOR_SPACE enum constants, typically JCS_RGB or
+JCS_GRAYSCALE.
+
+JPEG has a large number of compression parameters that determine how the
+image is encoded.  Most applications don't need or want to know about all
+these parameters.  You can set all the parameters to reasonable defaults by
+calling jpeg_set_defaults(); then, if there are particular values you want
+to change, you can do so after that.  The "Compression parameter selection"
+section tells about all the parameters.
+
+You must set in_color_space correctly before calling jpeg_set_defaults(),
+because the defaults depend on the source image colorspace.  However the
+other three source image parameters need not be valid until you call
+jpeg_start_compress().  There's no harm in calling jpeg_set_defaults() more
+than once, if that happens to be convenient.
+
+Typical code for a 24-bit RGB source image is
+
+	cinfo.image_width = Width; 	/* image width and height, in pixels */
+	cinfo.image_height = Height;
+	cinfo.input_components = 3;	/* # of color components per pixel */
+	cinfo.in_color_space = JCS_RGB; /* colorspace of input image */
+
+	jpeg_set_defaults(&cinfo);
+	/* Make optional parameter settings here */
+
+
+4. jpeg_start_compress(...);
+
+After you have established the data destination and set all the necessary
+source image info and other parameters, call jpeg_start_compress() to begin
+a compression cycle.  This will initialize internal state, allocate working
+storage, and emit the first few bytes of the JPEG datastream header.
+
+Typical code:
+
+	jpeg_start_compress(&cinfo, TRUE);
+
+The "TRUE" parameter ensures that a complete JPEG interchange datastream
+will be written.  This is appropriate in most cases.  If you think you might
+want to use an abbreviated datastream, read the section on abbreviated
+datastreams, below.
+
+Once you have called jpeg_start_compress(), you may not alter any JPEG
+parameters or other fields of the JPEG object until you have completed
+the compression cycle.
+
+
+5. while (scan lines remain to be written)
+	jpeg_write_scanlines(...);
+
+Now write all the required image data by calling jpeg_write_scanlines()
+one or more times.  You can pass one or more scanlines in each call, up
+to the total image height.  In most applications it is convenient to pass
+just one or a few scanlines at a time.  The expected format for the passed
+data is discussed under "Data formats", above.
+
+Image data should be written in top-to-bottom scanline order.  The JPEG spec
+contains some weasel wording about how top and bottom are application-defined
+terms (a curious interpretation of the English language...) but if you want
+your files to be compatible with everyone else's, you WILL use top-to-bottom
+order.  If the source data must be read in bottom-to-top order, you can use
+the JPEG library's virtual array mechanism to invert the data efficiently.
+Examples of this can be found in the sample application cjpeg.
+
+The library maintains a count of the number of scanlines written so far
+in the next_scanline field of the JPEG object.  Usually you can just use
+this variable as the loop counter, so that the loop test looks like
+"while (cinfo.next_scanline < cinfo.image_height)".
+
+Code for this step depends heavily on the way that you store the source data.
+example.c shows the following code for the case of a full-size 2-D source
+array containing 3-byte RGB pixels:
+
+	JSAMPROW row_pointer[1];	/* pointer to a single row */
+	int row_stride;			/* physical row width in buffer */
+
+	row_stride = image_width * 3;	/* JSAMPLEs per row in image_buffer */
+
+	while (cinfo.next_scanline < cinfo.image_height) {
+	    row_pointer[0] = & image_buffer[cinfo.next_scanline * row_stride];
+	    jpeg_write_scanlines(&cinfo, row_pointer, 1);
+	}
+
+jpeg_write_scanlines() returns the number of scanlines actually written.
+This will normally be equal to the number passed in, so you can usually
+ignore the return value.  It is different in just two cases:
+  * If you try to write more scanlines than the declared image height,
+    the additional scanlines are ignored.
+  * If you use a suspending data destination manager, output buffer overrun
+    will cause the compressor to return before accepting all the passed lines.
+    This feature is discussed under "I/O suspension", below.  The normal
+    stdio destination manager will NOT cause this to happen.
+In any case, the return value is the same as the change in the value of
+next_scanline.
+
+
+6. jpeg_finish_compress(...);
+
+After all the image data has been written, call jpeg_finish_compress() to
+complete the compression cycle.  This step is ESSENTIAL to ensure that the
+last bufferload of data is written to the data destination.
+jpeg_finish_compress() also releases working memory associated with the JPEG
+object.
+
+Typical code:
+
+	jpeg_finish_compress(&cinfo);
+
+If using the stdio destination manager, don't forget to close the output
+stdio stream (if necessary) afterwards.
+
+If you have requested a multi-pass operating mode, such as Huffman code
+optimization, jpeg_finish_compress() will perform the additional passes using
+data buffered by the first pass.  In this case jpeg_finish_compress() may take
+quite a while to complete.  With the default compression parameters, this will
+not happen.
+
+It is an error to call jpeg_finish_compress() before writing the necessary
+total number of scanlines.  If you wish to abort compression, call
+jpeg_abort() as discussed below.
+
+After completing a compression cycle, you may dispose of the JPEG object
+as discussed next, or you may use it to compress another image.  In that case
+return to step 2, 3, or 4 as appropriate.  If you do not change the
+destination manager, the new datastream will be written to the same target.
+If you do not change any JPEG parameters, the new datastream will be written
+with the same parameters as before.  Note that you can change the input image
+dimensions freely between cycles, but if you change the input colorspace, you
+should call jpeg_set_defaults() to adjust for the new colorspace; and then
+you'll need to repeat all of step 3.
+
+
+7. Release the JPEG compression object.
+
+When you are done with a JPEG compression object, destroy it by calling
+jpeg_destroy_compress().  This will free all subsidiary memory (regardless of
+the previous state of the object).  Or you can call jpeg_destroy(), which
+works for either compression or decompression objects --- this may be more
+convenient if you are sharing code between compression and decompression
+cases.  (Actually, these routines are equivalent except for the declared type
+of the passed pointer.  To avoid gripes from ANSI C compilers, jpeg_destroy()
+should be passed a j_common_ptr.)
+
+If you allocated the jpeg_compress_struct structure from malloc(), freeing
+it is your responsibility --- jpeg_destroy() won't.  Ditto for the error
+handler structure.
+
+Typical code:
+
+	jpeg_destroy_compress(&cinfo);
+
+
+8. Aborting.
+
+If you decide to abort a compression cycle before finishing, you can clean up
+in either of two ways:
+
+* If you don't need the JPEG object any more, just call
+  jpeg_destroy_compress() or jpeg_destroy() to release memory.  This is
+  legitimate at any point after calling jpeg_create_compress() --- in fact,
+  it's safe even if jpeg_create_compress() fails.
+
+* If you want to re-use the JPEG object, call jpeg_abort_compress(), or call
+  jpeg_abort() which works on both compression and decompression objects.
+  This will return the object to an idle state, releasing any working memory.
+  jpeg_abort() is allowed at any time after successful object creation.
+
+Note that cleaning up the data destination, if required, is your
+responsibility; neither of these routines will call term_destination().
+(See "Compressed data handling", below, for more about that.)
+
+jpeg_destroy() and jpeg_abort() are the only safe calls to make on a JPEG
+object that has reported an error by calling error_exit (see "Error handling"
+for more info).  The internal state of such an object is likely to be out of
+whack.  Either of these two routines will return the object to a known state.
+
+
+Decompression details
+---------------------
+
+Here we revisit the JPEG decompression outline given in the overview.
+
+1. Allocate and initialize a JPEG decompression object.
+
+This is just like initialization for compression, as discussed above,
+except that the object is a "struct jpeg_decompress_struct" and you
+call jpeg_create_decompress().  Error handling is exactly the same.
+
+Typical code:
+
+	struct jpeg_decompress_struct cinfo;
+	struct jpeg_error_mgr jerr;
+	...
+	cinfo.err = jpeg_std_error(&jerr);
+	jpeg_create_decompress(&cinfo);
+
+(Both here and in the IJG code, we usually use variable name "cinfo" for
+both compression and decompression objects.)
+
+
+2. Specify the source of the compressed data (eg, a file).
+
+As previously mentioned, the JPEG library reads compressed data from a "data
+source" module.  The library includes one data source module which knows how
+to read from a stdio stream.  You can use your own source module if you want
+to do something else, as discussed later.
+
+If you use the standard source module, you must open the source stdio stream
+beforehand.  Typical code for this step looks like:
+
+	FILE * infile;
+	...
+	if ((infile = fopen(filename, "rb")) == NULL) {
+	    fprintf(stderr, "can't open %s\n", filename);
+	    exit(1);
+	}
+	jpeg_stdio_src(&cinfo, infile);
+
+where the last line invokes the standard source module.
+
+WARNING: it is critical that the binary compressed data be read unchanged.
+On non-Unix systems the stdio library may perform newline translation or
+otherwise corrupt binary data.  To suppress this behavior, you may need to use
+a "b" option to fopen (as shown above), or use setmode() or another routine to
+put the stdio stream in binary mode.  See cjpeg.c and djpeg.c for code that
+has been found to work on many systems.
+
+You may not change the data source between calling jpeg_read_header() and
+jpeg_finish_decompress().  If you wish to read a series of JPEG images from
+a single source file, you should repeat the jpeg_read_header() to
+jpeg_finish_decompress() sequence without reinitializing either the JPEG
+object or the data source module; this prevents buffered input data from
+being discarded.
+
+
+3. Call jpeg_read_header() to obtain image info.
+
+Typical code for this step is just
+
+	jpeg_read_header(&cinfo, TRUE);
+
+This will read the source datastream header markers, up to the beginning
+of the compressed data proper.  On return, the image dimensions and other
+info have been stored in the JPEG object.  The application may wish to
+consult this information before selecting decompression parameters.
+
+More complex code is necessary if
+  * A suspending data source is used --- in that case jpeg_read_header()
+    may return before it has read all the header data.  See "I/O suspension",
+    below.  The normal stdio source manager will NOT cause this to happen.
+  * Abbreviated JPEG files are to be processed --- see the section on
+    abbreviated datastreams.  Standard applications that deal only in
+    interchange JPEG files need not be concerned with this case either.
+
+It is permissible to stop at this point if you just wanted to find out the
+image dimensions and other header info for a JPEG file.  In that case,
+call jpeg_destroy() when you are done with the JPEG object, or call
+jpeg_abort() to return it to an idle state before selecting a new data
+source and reading another header.
+
+
+4. Set parameters for decompression.
+
+jpeg_read_header() sets appropriate default decompression parameters based on
+the properties of the image (in particular, its colorspace).  However, you
+may well want to alter these defaults before beginning the decompression.
+For example, the default is to produce full color output from a color file.
+If you want colormapped output you must ask for it.  Other options allow the
+returned image to be scaled and allow various speed/quality tradeoffs to be
+selected.  "Decompression parameter selection", below, gives details.
+
+If the defaults are appropriate, nothing need be done at this step.
+
+Note that all default values are set by each call to jpeg_read_header().
+If you reuse a decompression object, you cannot expect your parameter
+settings to be preserved across cycles, as you can for compression.
+You must set desired parameter values each time.
+
+
+5. jpeg_start_decompress(...);
+
+Once the parameter values are satisfactory, call jpeg_start_decompress() to
+begin decompression.  This will initialize internal state, allocate working
+memory, and prepare for returning data.
+
+Typical code is just
+
+	jpeg_start_decompress(&cinfo);
+
+If you have requested a multi-pass operating mode, such as 2-pass color
+quantization, jpeg_start_decompress() will do everything needed before data
+output can begin.  In this case jpeg_start_decompress() may take quite a while
+to complete.  With a single-scan (non progressive) JPEG file and default
+decompression parameters, this will not happen; jpeg_start_decompress() will
+return quickly.
+
+After this call, the final output image dimensions, including any requested
+scaling, are available in the JPEG object; so is the selected colormap, if
+colormapped output has been requested.  Useful fields include
+
+	output_width		image width and height, as scaled
+	output_height
+	out_color_components	# of color components in out_color_space
+	output_components	# of color components returned per pixel
+	colormap		the selected colormap, if any
+	actual_number_of_colors		number of entries in colormap
+
+output_components is 1 (a colormap index) when quantizing colors; otherwise it
+equals out_color_components.  It is the number of JSAMPLE values that will be
+emitted per pixel in the output arrays.
+
+Typically you will need to allocate data buffers to hold the incoming image.
+You will need output_width * output_components JSAMPLEs per scanline in your
+output buffer, and a total of output_height scanlines will be returned.
+
+Note: if you are using the JPEG library's internal memory manager to allocate
+data buffers (as djpeg does), then the manager's protocol requires that you
+request large buffers *before* calling jpeg_start_decompress().  This is a
+little tricky since the output_XXX fields are not normally valid then.  You
+can make them valid by calling jpeg_calc_output_dimensions() after setting the
+relevant parameters (scaling, output color space, and quantization flag).
+
+
+6. while (scan lines remain to be read)
+	jpeg_read_scanlines(...);
+
+Now you can read the decompressed image data by calling jpeg_read_scanlines()
+one or more times.  At each call, you pass in the maximum number of scanlines
+to be read (ie, the height of your working buffer); jpeg_read_scanlines()
+will return up to that many lines.  The return value is the number of lines
+actually read.  The format of the returned data is discussed under "Data
+formats", above.  Don't forget that grayscale and color JPEGs will return
+different data formats!
+
+Image data is returned in top-to-bottom scanline order.  If you must write
+out the image in bottom-to-top order, you can use the JPEG library's virtual
+array mechanism to invert the data efficiently.  Examples of this can be
+found in the sample application djpeg.
+
+The library maintains a count of the number of scanlines returned so far
+in the output_scanline field of the JPEG object.  Usually you can just use
+this variable as the loop counter, so that the loop test looks like
+"while (cinfo.output_scanline < cinfo.output_height)".  (Note that the test
+should NOT be against image_height, unless you never use scaling.  The
+image_height field is the height of the original unscaled image.)
+The return value always equals the change in the value of output_scanline.
+
+If you don't use a suspending data source, it is safe to assume that
+jpeg_read_scanlines() reads at least one scanline per call, until the
+bottom of the image has been reached.
+
+If you use a buffer larger than one scanline, it is NOT safe to assume that
+jpeg_read_scanlines() fills it.  (The current implementation returns only a
+few scanlines per call, no matter how large a buffer you pass.)  So you must
+always provide a loop that calls jpeg_read_scanlines() repeatedly until the
+whole image has been read.
+
+
+7. jpeg_finish_decompress(...);
+
+After all the image data has been read, call jpeg_finish_decompress() to
+complete the decompression cycle.  This causes working memory associated
+with the JPEG object to be released.
+
+Typical code:
+
+	jpeg_finish_decompress(&cinfo);
+
+If using the stdio source manager, don't forget to close the source stdio
+stream if necessary.
+
+It is an error to call jpeg_finish_decompress() before reading the correct
+total number of scanlines.  If you wish to abort decompression, call
+jpeg_abort() as discussed below.
+
+After completing a decompression cycle, you may dispose of the JPEG object as
+discussed next, or you may use it to decompress another image.  In that case
+return to step 2 or 3 as appropriate.  If you do not change the source
+manager, the next image will be read from the same source.
+
+
+8. Release the JPEG decompression object.
+
+When you are done with a JPEG decompression object, destroy it by calling
+jpeg_destroy_decompress() or jpeg_destroy().  The previous discussion of
+destroying compression objects applies here too.
+
+Typical code:
+
+	jpeg_destroy_decompress(&cinfo);
+
+
+9. Aborting.
+
+You can abort a decompression cycle by calling jpeg_destroy_decompress() or
+jpeg_destroy() if you don't need the JPEG object any more, or
+jpeg_abort_decompress() or jpeg_abort() if you want to reuse the object.
+The previous discussion of aborting compression cycles applies here too.
+
+
+Mechanics of usage: include files, linking, etc
+-----------------------------------------------
+
+Applications using the JPEG library should include the header file jpeglib.h
+to obtain declarations of data types and routines.  Before including
+jpeglib.h, include system headers that define at least the typedefs FILE and
+size_t.  On ANSI-conforming systems, including <stdio.h> is sufficient; on
+older Unix systems, you may need <sys/types.h> to define size_t.
+
+If the application needs to refer to individual JPEG library error codes, also
+include jerror.h to define those symbols.
+
+jpeglib.h indirectly includes the files jconfig.h and jmorecfg.h.  If you are
+installing the JPEG header files in a system directory, you will want to
+install all four files: jpeglib.h, jerror.h, jconfig.h, jmorecfg.h.
+
+The most convenient way to include the JPEG code into your executable program
+is to prepare a library file ("libjpeg.a", or a corresponding name on non-Unix
+machines) and reference it at your link step.  If you use only half of the
+library (only compression or only decompression), only that much code will be
+included from the library, unless your linker is hopelessly brain-damaged.
+The supplied makefiles build libjpeg.a automatically (see install.doc).
+
+While you can build the JPEG library as a shared library if the whim strikes
+you, we don't really recommend it.  The trouble with shared libraries is that
+at some point you'll probably try to substitute a new version of the library
+without recompiling the calling applications.  That generally doesn't work
+because the parameter struct declarations usually change with each new
+version.  In other words, the library's API is *not* guaranteed binary
+compatible across versions; we only try to ensure source-code compatibility.
+(In hindsight, it might have been smarter to hide the parameter structs from
+applications and introduce a ton of access functions instead.  Too late now,
+however.)
+
+On some systems your application may need to set up a signal handler to ensure
+that temporary files are deleted if the program is interrupted.  This is most
+critical if you are on MS-DOS and use the jmemdos.c memory manager back end;
+it will try to grab extended memory for temp files, and that space will NOT be
+freed automatically.  See cjpeg.c or djpeg.c for an example signal handler.
+
+It may be worth pointing out that the core JPEG library does not actually
+require the stdio library: only the default source/destination managers and
+error handler need it.  You can use the library in a stdio-less environment
+if you replace those modules and use jmemnobs.c (or another memory manager of
+your own devising).  More info about the minimum system library requirements
+may be found in jinclude.h.
+
+
+ADVANCED FEATURES
+=================
+
+Compression parameter selection
+-------------------------------
+
+This section describes all the optional parameters you can set for JPEG
+compression, as well as the "helper" routines provided to assist in this
+task.  Proper setting of some parameters requires detailed understanding
+of the JPEG standard; if you don't know what a parameter is for, it's best
+not to mess with it!  See REFERENCES in the README file for pointers to
+more info about JPEG.
+
+It's a good idea to call jpeg_set_defaults() first, even if you plan to set
+all the parameters; that way your code is more likely to work with future JPEG
+libraries that have additional parameters.  For the same reason, we recommend
+you use a helper routine where one is provided, in preference to twiddling
+cinfo fields directly.
+
+The helper routines are:
+
+jpeg_set_defaults (j_compress_ptr cinfo)
+	This routine sets all JPEG parameters to reasonable defaults, using
+	only the input image's color space (field in_color_space, which must
+	already be set in cinfo).  Many applications will only need to use
+	this routine and perhaps jpeg_set_quality().
+
+jpeg_set_colorspace (j_compress_ptr cinfo, J_COLOR_SPACE colorspace)
+	Sets the JPEG file's colorspace (field jpeg_color_space) as specified,
+	and sets other color-space-dependent parameters appropriately.  See
+	"Special color spaces", below, before using this.  A large number of
+	parameters, including all per-component parameters, are set by this
+	routine; if you want to twiddle individual parameters you should call
+	jpeg_set_colorspace() before rather than after.
+
+jpeg_default_colorspace (j_compress_ptr cinfo)
+	Selects an appropriate JPEG colorspace based on cinfo->in_color_space,
+	and calls jpeg_set_colorspace().  This is actually a subroutine of
+	jpeg_set_defaults().  It's broken out in case you want to change
+	just the colorspace-dependent JPEG parameters.
+
+jpeg_set_quality (j_compress_ptr cinfo, int quality, boolean force_baseline)
+	Constructs JPEG quantization tables appropriate for the indicated
+	quality setting.  The quality value is expressed on the 0..100 scale
+	recommended by IJG (cjpeg's "-quality" switch uses this routine).
+	Note that the exact mapping from quality values to tables may change
+	in future IJG releases as more is learned about DCT quantization.
+	If the force_baseline parameter is TRUE, then the quantization table
+	entries are constrained to the range 1..255 for full JPEG baseline
+	compatibility.  In the current implementation, this only makes a
+	difference for quality settings below 25, and it effectively prevents
+	very small/low quality files from being generated.  The IJG decoder
+	is capable of reading the non-baseline files generated at low quality
+	settings when force_baseline is FALSE, but other decoders may not be.
+
+jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
+			 boolean force_baseline)
+	Same as jpeg_set_quality() except that the generated tables are the
+	sample tables given in the JPEC spec section K.1, multiplied by the
+	specified scale factor (which is expressed as a percentage; thus
+	scale_factor = 100 reproduces the spec's tables).  Note that larger
+	scale factors give lower quality.  This entry point is useful for
+	conforming to the Adobe PostScript DCT conventions, but we do not
+	recommend linear scaling as a user-visible quality scale otherwise.
+	force_baseline again constrains the computed table entries to 1..255.
+
+int jpeg_quality_scaling (int quality)
+	Converts a value on the IJG-recommended quality scale to a linear
+	scaling percentage.  Note that this routine may change or go away
+	in future releases --- IJG may choose to adopt a scaling method that
+	can't be expressed as a simple scalar multiplier, in which case the
+	premise of this routine collapses.  Caveat user.
+
+jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
+		      const unsigned int *basic_table,
+		      int scale_factor, boolean force_baseline)
+	Allows an arbitrary quantization table to be created.  which_tbl
+	indicates which table slot to fill.  basic_table points to an array
+	of 64 unsigned ints given in normal array order.  These values are
+	multiplied by scale_factor/100 and then clamped to the range 1..65535
+	(or to 1..255 if force_baseline is TRUE).
+	CAUTION: prior to library version 6a, jpeg_add_quant_table expected
+	the basic table to be given in JPEG zigzag order.  If you need to
+	write code that works with either older or newer versions of this
+	routine, you must check the library version number.  Something like
+	"#if JPEG_LIB_VERSION >= 61" is the right test.
+
+jpeg_simple_progression (j_compress_ptr cinfo)
+	Generates a default scan script for writing a progressive-JPEG file.
+	This is the recommended method of creating a progressive file,
+	unless you want to make a custom scan sequence.  You must ensure that
+	the JPEG color space is set correctly before calling this routine.
+
+
+Compression parameters (cinfo fields) include:
+
+J_DCT_METHOD dct_method
+	Selects the algorithm used for the DCT step.  Choices are:
+		JDCT_ISLOW: slow but accurate integer algorithm
+		JDCT_IFAST: faster, less accurate integer method
+		JDCT_FLOAT: floating-point method
+		JDCT_DEFAULT: default method (normally JDCT_ISLOW)
+		JDCT_FASTEST: fastest method (normally JDCT_IFAST)
+	The FLOAT method is very slightly more accurate than the ISLOW method,
+	but may give different results on different machines due to varying
+	roundoff behavior.  The integer methods should give the same results
+	on all machines.  On machines with sufficiently fast FP hardware, the
+	floating-point method may also be the fastest.  The IFAST method is
+	considerably less accurate than the other two; its use is not
+	recommended if high quality is a concern.  JDCT_DEFAULT and
+	JDCT_FASTEST are macros configurable by each installation.
+
+J_COLOR_SPACE jpeg_color_space
+int num_components
+	The JPEG color space and corresponding number of components; see
+	"Special color spaces", below, for more info.  We recommend using
+	jpeg_set_color_space() if you want to change these.
+
+boolean optimize_coding
+	TRUE causes the compressor to compute optimal Huffman coding tables
+	for the image.  This requires an extra pass over the data and
+	therefore costs a good deal of space and time.  The default is
+	FALSE, which tells the compressor to use the supplied or default
+	Huffman tables.  In most cases optimal tables save only a few percent
+	of file size compared to the default tables.  Note that when this is
+	TRUE, you need not supply Huffman tables at all, and any you do
+	supply will be overwritten.
+
+unsigned int restart_interval
+int restart_in_rows
+	To emit restart markers in the JPEG file, set one of these nonzero.
+	Set restart_interval to specify the exact interval in MCU blocks.
+	Set restart_in_rows to specify the interval in MCU rows.  (If
+	restart_in_rows is not 0, then restart_interval is set after the
+	image width in MCUs is computed.)  Defaults are zero (no restarts).
+	One restart marker per MCU row is often a good choice.
+	NOTE: the overhead of restart markers is higher in grayscale JPEG
+	files than in color files, and MUCH higher in progressive JPEGs.
+	If you use restarts, you may want to use larger intervals in those
+	cases.
+
+const jpeg_scan_info * scan_info
+int num_scans
+	By default, scan_info is NULL; this causes the compressor to write a
+	single-scan sequential JPEG file.  If not NULL, scan_info points to
+	an array of scan definition records of length num_scans.  The
+	compressor will then write a JPEG file having one scan for each scan
+	definition record.  This is used to generate noninterleaved or
+	progressive JPEG files.  The library checks that the scan array
+	defines a valid JPEG scan sequence.  (jpeg_simple_progression creates
+	a suitable scan definition array for progressive JPEG.)  This is
+	discussed further under "Progressive JPEG support".
+
+int smoothing_factor
+	If non-zero, the input image is smoothed; the value should be 1 for
+	minimal smoothing to 100 for maximum smoothing.  Consult jcsample.c
+	for details of the smoothing algorithm.  The default is zero.
+
+boolean write_JFIF_header
+	If TRUE, a JFIF APP0 marker is emitted.  jpeg_set_defaults() and
+	jpeg_set_colorspace() set this TRUE if a JFIF-legal JPEG color space
+	(ie, YCbCr or grayscale) is selected, otherwise FALSE.
+
+UINT8 JFIF_major_version
+UINT8 JFIF_minor_version
+	The version number to be written into the JFIF marker.
+	jpeg_set_defaults() initializes the version to 1.01 (major=minor=1).
+	You should set it to 1.02 (major=1, minor=2) if you plan to write
+	any JFIF 1.02 extension markers.
+
+UINT8 density_unit
+UINT16 X_density
+UINT16 Y_density
+	The resolution information to be written into the JFIF marker;
+	not used otherwise.  density_unit may be 0 for unknown,
+	1 for dots/inch, or 2 for dots/cm.  The default values are 0,1,1
+	indicating square pixels of unknown size.
+
+boolean write_Adobe_marker
+	If TRUE, an Adobe APP14 marker is emitted.  jpeg_set_defaults() and
+	jpeg_set_colorspace() set this TRUE if JPEG color space RGB, CMYK,
+	or YCCK is selected, otherwise FALSE.  It is generally a bad idea
+	to set both write_JFIF_header and write_Adobe_marker.  In fact,
+	you probably shouldn't change the default settings at all --- the
+	default behavior ensures that the JPEG file's color space can be
+	recognized by the decoder.
+
+JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS]
+	Pointers to coefficient quantization tables, one per table slot,
+	or NULL if no table is defined for a slot.  Usually these should
+	be set via one of the above helper routines; jpeg_add_quant_table()
+	is general enough to define any quantization table.  The other
+	routines will set up table slot 0 for luminance quality and table
+	slot 1 for chrominance.
+
+JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS]
+JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS]
+	Pointers to Huffman coding tables, one per table slot, or NULL if
+	no table is defined for a slot.  Slots 0 and 1 are filled with the
+	JPEG sample tables by jpeg_set_defaults().  If you need to allocate
+	more table structures, jpeg_alloc_huff_table() may be used.
+	Note that optimal Huffman tables can be computed for an image
+	by setting optimize_coding, as discussed above; there's seldom
+	any need to mess with providing your own Huffman tables.
+
+There are some additional cinfo fields which are not documented here
+because you currently can't change them; for example, you can't set
+arith_code TRUE because arithmetic coding is unsupported.
+
+
+Per-component parameters are stored in the struct cinfo.comp_info[i] for
+component number i.  Note that components here refer to components of the
+JPEG color space, *not* the source image color space.  A suitably large
+comp_info[] array is allocated by jpeg_set_defaults(); if you choose not
+to use that routine, it's up to you to allocate the array.
+
+int component_id
+	The one-byte identifier code to be recorded in the JPEG file for
+	this component.  For the standard color spaces, we recommend you
+	leave the default values alone.
+
+int h_samp_factor
+int v_samp_factor
+	Horizontal and vertical sampling factors for the component; must
+	be 1..4 according to the JPEG standard.  Note that larger sampling
+	factors indicate a higher-resolution component; many people find
+	this behavior quite unintuitive.  The default values are 2,2 for
+	luminance components and 1,1 for chrominance components, except
+	for grayscale where 1,1 is used.
+
+int quant_tbl_no
+	Quantization table number for component.  The default value is
+	0 for luminance components and 1 for chrominance components.
+
+int dc_tbl_no
+int ac_tbl_no
+	DC and AC entropy coding table numbers.  The default values are
+	0 for luminance components and 1 for chrominance components.
+
+int component_index
+	Must equal the component's index in comp_info[].  (Beginning in
+	release v6, the compressor library will fill this in automatically;
+	you don't have to.)
+
+
+Decompression parameter selection
+---------------------------------
+
+Decompression parameter selection is somewhat simpler than compression
+parameter selection, since all of the JPEG internal parameters are
+recorded in the source file and need not be supplied by the application.
+(Unless you are working with abbreviated files, in which case see
+"Abbreviated datastreams", below.)  Decompression parameters control
+the postprocessing done on the image to deliver it in a format suitable
+for the application's use.  Many of the parameters control speed/quality
+tradeoffs, in which faster decompression may be obtained at the price of
+a poorer-quality image.  The defaults select the highest quality (slowest)
+processing.
+
+The following fields in the JPEG object are set by jpeg_read_header() and
+may be useful to the application in choosing decompression parameters:
+
+JDIMENSION image_width			Width and height of image
+JDIMENSION image_height
+int num_components			Number of color components
+J_COLOR_SPACE jpeg_color_space		Colorspace of image
+boolean saw_JFIF_marker			TRUE if a JFIF APP0 marker was seen
+  UINT8 JFIF_major_version		Version information from JFIF marker
+  UINT8 JFIF_minor_version
+  UINT8 density_unit			Resolution data from JFIF marker
+  UINT16 X_density
+  UINT16 Y_density
+boolean saw_Adobe_marker		TRUE if an Adobe APP14 marker was seen
+  UINT8 Adobe_transform			Color transform code from Adobe marker
+
+The JPEG color space, unfortunately, is something of a guess since the JPEG
+standard proper does not provide a way to record it.  In practice most files
+adhere to the JFIF or Adobe conventions, and the decoder will recognize these
+correctly.  See "Special color spaces", below, for more info.
+
+
+The decompression parameters that determine the basic properties of the
+returned image are:
+
+J_COLOR_SPACE out_color_space
+	Output color space.  jpeg_read_header() sets an appropriate default
+	based on jpeg_color_space; typically it will be RGB or grayscale.
+	The application can change this field to request output in a different
+	colorspace.  For example, set it to JCS_GRAYSCALE to get grayscale
+	output from a color file.  (This is useful for previewing: grayscale
+	output is faster than full color since the color components need not
+	be processed.)  Note that not all possible color space transforms are
+	currently implemented; you may need to extend jdcolor.c if you want an
+	unusual conversion.
+
+unsigned int scale_num, scale_denom
+	Scale the image by the fraction scale_num/scale_denom.  Default is
+	1/1, or no scaling.  Currently, the only supported scaling ratios
+	are 1/1, 1/2, 1/4, and 1/8.  (The library design allows for arbitrary
+	scaling ratios but this is not likely to be implemented any time soon.)
+	Smaller scaling ratios permit significantly faster decoding since
+	fewer pixels need be processed and a simpler IDCT method can be used.
+
+boolean quantize_colors
+	If set TRUE, colormapped output will be delivered.  Default is FALSE,
+	meaning that full-color output will be delivered.
+
+The next three parameters are relevant only if quantize_colors is TRUE.
+
+int desired_number_of_colors
+	Maximum number of colors to use in generating a library-supplied color
+	map (the actual number of colors is returned in a different field).
+	Default 256.  Ignored when the application supplies its own color map.
+
+boolean two_pass_quantize
+	If TRUE, an extra pass over the image is made to select a custom color
+	map for the image.  This usually looks a lot better than the one-size-
+	fits-all colormap that is used otherwise.  Default is TRUE.  Ignored
+	when the application supplies its own color map.
+
+J_DITHER_MODE dither_mode
+	Selects color dithering method.  Supported values are:
+		JDITHER_NONE	no dithering: fast, very low quality
+		JDITHER_ORDERED	ordered dither: moderate speed and quality
+		JDITHER_FS	Floyd-Steinberg dither: slow, high quality
+	Default is JDITHER_FS.  (At present, ordered dither is implemented
+	only in the single-pass, standard-colormap case.  If you ask for
+	ordered dither when two_pass_quantize is TRUE or when you supply
+	an external color map, you'll get F-S dithering.)
+
+When quantize_colors is TRUE, the target color map is described by the next
+two fields.  colormap is set to NULL by jpeg_read_header().  The application
+can supply a color map by setting colormap non-NULL and setting
+actual_number_of_colors to the map size.  Otherwise, jpeg_start_decompress()
+selects a suitable color map and sets these two fields itself.
+[Implementation restriction: at present, an externally supplied colormap is
+only accepted for 3-component output color spaces.]
+
+JSAMPARRAY colormap
+	The color map, represented as a 2-D pixel array of out_color_components
+	rows and actual_number_of_colors columns.  Ignored if not quantizing.
+	CAUTION: if the JPEG library creates its own colormap, the storage
+	pointed to by this field is released by jpeg_finish_decompress().
+	Copy the colormap somewhere else first, if you want to save it.
+
+int actual_number_of_colors
+	The number of colors in the color map.
+
+Additional decompression parameters that the application may set include:
+
+J_DCT_METHOD dct_method
+	Selects the algorithm used for the DCT step.  Choices are the same
+	as described above for compression.
+
+boolean do_fancy_upsampling
+	If TRUE, do careful upsampling of chroma components.  If FALSE,
+	a faster but sloppier method is used.  Default is TRUE.  The visual
+	impact of the sloppier method is often very small.
+
+boolean do_block_smoothing
+	If TRUE, interblock smoothing is applied in early stages of decoding
+	progressive JPEG files; if FALSE, not.  Default is TRUE.  Early
+	progression stages look "fuzzy" with smoothing, "blocky" without.
+	In any case, block smoothing ceases to be applied after the first few
+	AC coefficients are known to full accuracy, so it is relevant only
+	when using buffered-image mode for progressive images.
+
+boolean enable_1pass_quant
+boolean enable_external_quant
+boolean enable_2pass_quant
+	These are significant only in buffered-image mode, which is
+	described in its own section below.
+
+
+The output image dimensions are given by the following fields.  These are
+computed from the source image dimensions and the decompression parameters
+by jpeg_start_decompress().  You can also call jpeg_calc_output_dimensions()
+to obtain the values that will result from the current parameter settings.
+This can be useful if you are trying to pick a scaling ratio that will get
+close to a desired target size.  It's also important if you are using the
+JPEG library's memory manager to allocate output buffer space, because you
+are supposed to request such buffers *before* jpeg_start_decompress().
+
+JDIMENSION output_width		Actual dimensions of output image.
+JDIMENSION output_height
+int out_color_components	Number of color components in out_color_space.
+int output_components		Number of color components returned.
+int rec_outbuf_height		Recommended height of scanline buffer.
+
+When quantizing colors, output_components is 1, indicating a single color map
+index per pixel.  Otherwise it equals out_color_components.  The output arrays
+are required to be output_width * output_components JSAMPLEs wide.
+
+rec_outbuf_height is the recommended minimum height (in scanlines) of the
+buffer passed to jpeg_read_scanlines().  If the buffer is smaller, the
+library will still work, but time will be wasted due to unnecessary data
+copying.  In high-quality modes, rec_outbuf_height is always 1, but some
+faster, lower-quality modes set it to larger values (typically 2 to 4).
+If you are going to ask for a high-speed processing mode, you may as well
+go to the trouble of honoring rec_outbuf_height so as to avoid data copying.
+(An output buffer larger than rec_outbuf_height lines is OK, but won't
+provide any material speed improvement over that height.)
+
+
+Special color spaces
+--------------------
+
+The JPEG standard itself is "color blind" and doesn't specify any particular
+color space.  It is customary to convert color data to a luminance/chrominance
+color space before compressing, since this permits greater compression.  The
+existing de-facto JPEG file format standards specify YCbCr or grayscale data
+(JFIF), or grayscale, RGB, YCbCr, CMYK, or YCCK (Adobe).  For special
+applications such as multispectral images, other color spaces can be used,
+but it must be understood that such files will be unportable.
+
+The JPEG library can handle the most common colorspace conversions (namely
+RGB <=> YCbCr and CMYK <=> YCCK).  It can also deal with data of an unknown
+color space, passing it through without conversion.  If you deal extensively
+with an unusual color space, you can easily extend the library to understand
+additional color spaces and perform appropriate conversions.
+
+For compression, the source data's color space is specified by field
+in_color_space.  This is transformed to the JPEG file's color space given
+by jpeg_color_space.  jpeg_set_defaults() chooses a reasonable JPEG color
+space depending on in_color_space, but you can override this by calling
+jpeg_set_colorspace().  Of course you must select a supported transformation.
+jccolor.c currently supports the following transformations:
+	RGB => YCbCr
+	RGB => GRAYSCALE
+	YCbCr => GRAYSCALE
+	CMYK => YCCK
+plus the null transforms: GRAYSCALE => GRAYSCALE, RGB => RGB,
+YCbCr => YCbCr, CMYK => CMYK, YCCK => YCCK, and UNKNOWN => UNKNOWN.
+
+The de-facto file format standards (JFIF and Adobe) specify APPn markers that
+indicate the color space of the JPEG file.  It is important to ensure that
+these are written correctly, or omitted if the JPEG file's color space is not
+one of the ones supported by the de-facto standards.  jpeg_set_colorspace()
+will set the compression parameters to include or omit the APPn markers
+properly, so long as it is told the truth about the JPEG color space.
+For example, if you are writing some random 3-component color space without
+conversion, don't try to fake out the library by setting in_color_space and
+jpeg_color_space to JCS_YCbCr; use JCS_UNKNOWN.  You may want to write an
+APPn marker of your own devising to identify the colorspace --- see "Special
+markers", below.
+
+When told that the color space is UNKNOWN, the library will default to using
+luminance-quality compression parameters for all color components.  You may
+well want to change these parameters.  See the source code for
+jpeg_set_colorspace(), in jcparam.c, for details.
+
+For decompression, the JPEG file's color space is given in jpeg_color_space,
+and this is transformed to the output color space out_color_space.
+jpeg_read_header's setting of jpeg_color_space can be relied on if the file
+conforms to JFIF or Adobe conventions, but otherwise it is no better than a
+guess.  If you know the JPEG file's color space for certain, you can override
+jpeg_read_header's guess by setting jpeg_color_space.  jpeg_read_header also
+selects a default output color space based on (its guess of) jpeg_color_space;
+set out_color_space to override this.  Again, you must select a supported
+transformation.  jdcolor.c currently supports
+	YCbCr => GRAYSCALE
+	YCbCr => RGB
+	GRAYSCALE => RGB
+	YCCK => CMYK
+as well as the null transforms.  (Since GRAYSCALE=>RGB is provided, an
+application can force grayscale JPEGs to look like color JPEGs if it only
+wants to handle one case.)
+
+The two-pass color quantizer, jquant2.c, is specialized to handle RGB data
+(it weights distances appropriately for RGB colors).  You'll need to modify
+the code if you want to use it for non-RGB output color spaces.  Note that
+jquant2.c is used to map to an application-supplied colormap as well as for
+the normal two-pass colormap selection process.
+
+CAUTION: it appears that Adobe Photoshop writes inverted data in CMYK JPEG
+files: 0 represents 100% ink coverage, rather than 0% ink as you'd expect.
+This is arguably a bug in Photoshop, but if you need to work with Photoshop
+CMYK files, you will have to deal with it in your application.  We cannot
+"fix" this in the library by inverting the data during the CMYK<=>YCCK
+transform, because that would break other applications, notably Ghostscript.
+Photoshop versions prior to 3.0 write EPS files containing JPEG-encoded CMYK
+data in the same inverted-YCCK representation used in bare JPEG files, but
+the surrounding PostScript code performs an inversion using the PS image
+operator.  I am told that Photoshop 3.0 will write uninverted YCCK in
+EPS/JPEG files, and will omit the PS-level inversion.  (But the data
+polarity used in bare JPEG files will not change in 3.0.)  In either case,
+the JPEG library must not invert the data itself, or else Ghostscript would
+read these EPS files incorrectly.
+
+
+Error handling
+--------------
+
+When the default error handler is used, any error detected inside the JPEG
+routines will cause a message to be printed on stderr, followed by exit().
+You can supply your own error handling routines to override this behavior
+and to control the treatment of nonfatal warnings and trace/debug messages.
+The file example.c illustrates the most common case, which is to have the
+application regain control after an error rather than exiting.
+
+The JPEG library never writes any message directly; it always goes through
+the error handling routines.  Three classes of messages are recognized:
+  * Fatal errors: the library cannot continue.
+  * Warnings: the library can continue, but the data is corrupt, and a
+    damaged output image is likely to result.
+  * Trace/informational messages.  These come with a trace level indicating
+    the importance of the message; you can control the verbosity of the
+    program by adjusting the maximum trace level that will be displayed.
+
+You may, if you wish, simply replace the entire JPEG error handling module
+(jerror.c) with your own code.  However, you can avoid code duplication by
+only replacing some of the routines depending on the behavior you need.
+This is accomplished by calling jpeg_std_error() as usual, but then overriding
+some of the method pointers in the jpeg_error_mgr struct, as illustrated by
+example.c.
+
+All of the error handling routines will receive a pointer to the JPEG object
+(a j_common_ptr which points to either a jpeg_compress_struct or a
+jpeg_decompress_struct; if you need to tell which, test the is_decompressor
+field).  This struct includes a pointer to the error manager struct in its
+"err" field.  Frequently, custom error handler routines will need to access
+additional data which is not known to the JPEG library or the standard error
+handler.  The most convenient way to do this is to embed either the JPEG
+object or the jpeg_error_mgr struct in a larger structure that contains
+additional fields; then casting the passed pointer provides access to the
+additional fields.  Again, see example.c for one way to do it.  (Beginning
+with IJG version 6b, there is also a void pointer "client_data" in each
+JPEG object, which the application can also use to find related data.
+The library does not touch client_data at all.)
+
+The individual methods that you might wish to override are:
+
+error_exit (j_common_ptr cinfo)
+	Receives control for a fatal error.  Information sufficient to
+	generate the error message has been stored in cinfo->err; call
+	output_message to display it.  Control must NOT return to the caller;
+	generally this routine will exit() or longjmp() somewhere.
+	Typically you would override this routine to get rid of the exit()
+	default behavior.  Note that if you continue processing, you should
+	clean up the JPEG object with jpeg_abort() or jpeg_destroy().
+
+output_message (j_common_ptr cinfo)
+	Actual output of any JPEG message.  Override this to send messages
+	somewhere other than stderr.  Note that this method does not know
+	how to generate a message, only where to send it.
+
+format_message (j_common_ptr cinfo, char * buffer)
+	Constructs a readable error message string based on the error info
+	stored in cinfo->err.  This method is called by output_message.  Few
+	applications should need to override this method.  One possible
+	reason for doing so is to implement dynamic switching of error message
+	language.
+
+emit_message (j_common_ptr cinfo, int msg_level)
+	Decide whether or not to emit a warning or trace message; if so,
+	calls output_message.  The main reason for overriding this method
+	would be to abort on warnings.  msg_level is -1 for warnings,
+	0 and up for trace messages.
+
+Only error_exit() and emit_message() are called from the rest of the JPEG
+library; the other two are internal to the error handler.
+
+The actual message texts are stored in an array of strings which is pointed to
+by the field err->jpeg_message_table.  The messages are numbered from 0 to
+err->last_jpeg_message, and it is these code numbers that are used in the
+JPEG library code.  You could replace the message texts (for instance, with
+messages in French or German) by changing the message table pointer.  See
+jerror.h for the default texts.  CAUTION: this table will almost certainly
+change or grow from one library version to the next.
+
+It may be useful for an application to add its own message texts that are
+handled by the same mechanism.  The error handler supports a second "add-on"
+message table for this purpose.  To define an addon table, set the pointer
+err->addon_message_table and the message numbers err->first_addon_message and
+err->last_addon_message.  If you number the addon messages beginning at 1000
+or so, you won't have to worry about conflicts with the library's built-in
+messages.  See the sample applications cjpeg/djpeg for an example of using
+addon messages (the addon messages are defined in cderror.h).
+
+Actual invocation of the error handler is done via macros defined in jerror.h:
+	ERREXITn(...)	for fatal errors
+	WARNMSn(...)	for corrupt-data warnings
+	TRACEMSn(...)	for trace and informational messages.
+These macros store the message code and any additional parameters into the
+error handler struct, then invoke the error_exit() or emit_message() method.
+The variants of each macro are for varying numbers of additional parameters.
+The additional parameters are inserted into the generated message using
+standard printf() format codes.
+
+See jerror.h and jerror.c for further details.
+
+
+Compressed data handling (source and destination managers)
+----------------------------------------------------------
+
+The JPEG compression library sends its compressed data to a "destination
+manager" module.  The default destination manager just writes the data to a
+stdio stream, but you can provide your own manager to do something else.
+Similarly, the decompression library calls a "source manager" to obtain the
+compressed data; you can provide your own source manager if you want the data
+to come from somewhere other than a stdio stream.
+
+In both cases, compressed data is processed a bufferload at a time: the
+destination or source manager provides a work buffer, and the library invokes
+the manager only when the buffer is filled or emptied.  (You could define a
+one-character buffer to force the manager to be invoked for each byte, but
+that would be rather inefficient.)  The buffer's size and location are
+controlled by the manager, not by the library.  For example, if you desired to
+decompress a JPEG datastream that was all in memory, you could just make the
+buffer pointer and length point to the original data in memory.  Then the
+buffer-reload procedure would be invoked only if the decompressor ran off the
+end of the datastream, which would indicate an erroneous datastream.
+
+The work buffer is defined as an array of datatype JOCTET, which is generally
+"char" or "unsigned char".  On a machine where char is not exactly 8 bits
+wide, you must define JOCTET as a wider data type and then modify the data
+source and destination modules to transcribe the work arrays into 8-bit units
+on external storage.
+
+A data destination manager struct contains a pointer and count defining the
+next byte to write in the work buffer and the remaining free space:
+
+	JOCTET * next_output_byte;  /* => next byte to write in buffer */
+	size_t free_in_buffer;      /* # of byte spaces remaining in buffer */
+
+The library increments the pointer and decrements the count until the buffer
+is filled.  The manager's empty_output_buffer method must reset the pointer
+and count.  The manager is expected to remember the buffer's starting address
+and total size in private fields not visible to the library.
+
+A data destination manager provides three methods:
+
+init_destination (j_compress_ptr cinfo)
+	Initialize destination.  This is called by jpeg_start_compress()
+	before any data is actually written.  It must initialize
+	next_output_byte and free_in_buffer.  free_in_buffer must be
+	initialized to a positive value.
+
+empty_output_buffer (j_compress_ptr cinfo)
+	This is called whenever the buffer has filled (free_in_buffer
+	reaches zero).  In typical applications, it should write out the
+	*entire* buffer (use the saved start address and buffer length;
+	ignore the current state of next_output_byte and free_in_buffer).
+	Then reset the pointer & count to the start of the buffer, and
+	return TRUE indicating that the buffer has been dumped.
+	free_in_buffer must be set to a positive value when TRUE is
+	returned.  A FALSE return should only be used when I/O suspension is
+	desired (this operating mode is discussed in the next section).
+
+term_destination (j_compress_ptr cinfo)
+	Terminate destination --- called by jpeg_finish_compress() after all
+	data has been written.  In most applications, this must flush any
+	data remaining in the buffer.  Use either next_output_byte or
+	free_in_buffer to determine how much data is in the buffer.
+
+term_destination() is NOT called by jpeg_abort() or jpeg_destroy().  If you
+want the destination manager to be cleaned up during an abort, you must do it
+yourself.
+
+You will also need code to create a jpeg_destination_mgr struct, fill in its
+method pointers, and insert a pointer to the struct into the "dest" field of
+the JPEG compression object.  This can be done in-line in your setup code if
+you like, but it's probably cleaner to provide a separate routine similar to
+the jpeg_stdio_dest() routine of the supplied destination manager.
+
+Decompression source managers follow a parallel design, but with some
+additional frammishes.  The source manager struct contains a pointer and count
+defining the next byte to read from the work buffer and the number of bytes
+remaining:
+
+	const JOCTET * next_input_byte; /* => next byte to read from buffer */
+	size_t bytes_in_buffer;         /* # of bytes remaining in buffer */
+
+The library increments the pointer and decrements the count until the buffer
+is emptied.  The manager's fill_input_buffer method must reset the pointer and
+count.  In most applications, the manager must remember the buffer's starting
+address and total size in private fields not visible to the library.
+
+A data source manager provides five methods:
+
+init_source (j_decompress_ptr cinfo)
+	Initialize source.  This is called by jpeg_read_header() before any
+	data is actually read.  Unlike init_destination(), it may leave
+	bytes_in_buffer set to 0 (in which case a fill_input_buffer() call
+	will occur immediately).
+
+fill_input_buffer (j_decompress_ptr cinfo)
+	This is called whenever bytes_in_buffer has reached zero and more
+	data is wanted.  In typical applications, it should read fresh data
+	into the buffer (ignoring the current state of next_input_byte and
+	bytes_in_buffer), reset the pointer & count to the start of the
+	buffer, and return TRUE indicating that the buffer has been reloaded.
+	It is not necessary to fill the buffer entirely, only to obtain at
+	least one more byte.  bytes_in_buffer MUST be set to a positive value
+	if TRUE is returned.  A FALSE return should only be used when I/O
+	suspension is desired (this mode is discussed in the next section).
+
+skip_input_data (j_decompress_ptr cinfo, long num_bytes)
+	Skip num_bytes worth of data.  The buffer pointer and count should
+	be advanced over num_bytes input bytes, refilling the buffer as
+	needed.  This is used to skip over a potentially large amount of
+	uninteresting data (such as an APPn marker).  In some applications
+	it may be possible to optimize away the reading of the skipped data,
+	but it's not clear that being smart is worth much trouble; large
+	skips are uncommon.  bytes_in_buffer may be zero on return.
+	A zero or negative skip count should be treated as a no-op.
+
+resync_to_restart (j_decompress_ptr cinfo, int desired)
+	This routine is called only when the decompressor has failed to find
+	a restart (RSTn) marker where one is expected.  Its mission is to
+	find a suitable point for resuming decompression.  For most
+	applications, we recommend that you just use the default resync
+	procedure, jpeg_resync_to_restart().  However, if you are able to back
+	up in the input data stream, or if you have a-priori knowledge about
+	the likely location of restart markers, you may be able to do better.
+	Read the read_restart_marker() and jpeg_resync_to_restart() routines
+	in jdmarker.c if you think you'd like to implement your own resync
+	procedure.
+
+term_source (j_decompress_ptr cinfo)
+	Terminate source --- called by jpeg_finish_decompress() after all
+	data has been read.  Often a no-op.
+
+For both fill_input_buffer() and skip_input_data(), there is no such thing
+as an EOF return.  If the end of the file has been reached, the routine has
+a choice of exiting via ERREXIT() or inserting fake data into the buffer.
+In most cases, generating a warning message and inserting a fake EOI marker
+is the best course of action --- this will allow the decompressor to output
+however much of the image is there.  In pathological cases, the decompressor
+may swallow the EOI and again demand data ... just keep feeding it fake EOIs.
+jdatasrc.c illustrates the recommended error recovery behavior.
+
+term_source() is NOT called by jpeg_abort() or jpeg_destroy().  If you want
+the source manager to be cleaned up during an abort, you must do it yourself.
+
+You will also need code to create a jpeg_source_mgr struct, fill in its method
+pointers, and insert a pointer to the struct into the "src" field of the JPEG
+decompression object.  This can be done in-line in your setup code if you
+like, but it's probably cleaner to provide a separate routine similar to the
+jpeg_stdio_src() routine of the supplied source manager.
+
+For more information, consult the stdio source and destination managers
+in jdatasrc.c and jdatadst.c.
+
+
+I/O suspension
+--------------
+
+Some applications need to use the JPEG library as an incremental memory-to-
+memory filter: when the compressed data buffer is filled or emptied, they want
+control to return to the outer loop, rather than expecting that the buffer can
+be emptied or reloaded within the data source/destination manager subroutine.
+The library supports this need by providing an "I/O suspension" mode, which we
+describe in this section.
+
+The I/O suspension mode is not a panacea: nothing is guaranteed about the
+maximum amount of time spent in any one call to the library, so it will not
+eliminate response-time problems in single-threaded applications.  If you
+need guaranteed response time, we suggest you "bite the bullet" and implement
+a real multi-tasking capability.
+
+To use I/O suspension, cooperation is needed between the calling application
+and the data source or destination manager; you will always need a custom
+source/destination manager.  (Please read the previous section if you haven't
+already.)  The basic idea is that the empty_output_buffer() or
+fill_input_buffer() routine is a no-op, merely returning FALSE to indicate
+that it has done nothing.  Upon seeing this, the JPEG library suspends
+operation and returns to its caller.  The surrounding application is
+responsible for emptying or refilling the work buffer before calling the
+JPEG library again.
+
+Compression suspension:
+
+For compression suspension, use an empty_output_buffer() routine that returns
+FALSE; typically it will not do anything else.  This will cause the
+compressor to return to the caller of jpeg_write_scanlines(), with the return
+value indicating that not all the supplied scanlines have been accepted.
+The application must make more room in the output buffer, adjust the output
+buffer pointer/count appropriately, and then call jpeg_write_scanlines()
+again, pointing to the first unconsumed scanline.
+
+When forced to suspend, the compressor will backtrack to a convenient stopping
+point (usually the start of the current MCU); it will regenerate some output
+data when restarted.  Therefore, although empty_output_buffer() is only
+called when the buffer is filled, you should NOT write out the entire buffer
+after a suspension.  Write only the data up to the current position of
+next_output_byte/free_in_buffer.  The data beyond that point will be
+regenerated after resumption.
+
+Because of the backtracking behavior, a good-size output buffer is essential
+for efficiency; you don't want the compressor to suspend often.  (In fact, an
+overly small buffer could lead to infinite looping, if a single MCU required
+more data than would fit in the buffer.)  We recommend a buffer of at least
+several Kbytes.  You may want to insert explicit code to ensure that you don't
+call jpeg_write_scanlines() unless there is a reasonable amount of space in
+the output buffer; in other words, flush the buffer before trying to compress
+more data.
+
+The compressor does not allow suspension while it is trying to write JPEG
+markers at the beginning and end of the file.  This means that:
+  * At the beginning of a compression operation, there must be enough free
+    space in the output buffer to hold the header markers (typically 600 or
+    so bytes).  The recommended buffer size is bigger than this anyway, so
+    this is not a problem as long as you start with an empty buffer.  However,
+    this restriction might catch you if you insert large special markers, such
+    as a JFIF thumbnail image, without flushing the buffer afterwards.
+  * When you call jpeg_finish_compress(), there must be enough space in the
+    output buffer to emit any buffered data and the final EOI marker.  In the
+    current implementation, half a dozen bytes should suffice for this, but
+    for safety's sake we recommend ensuring that at least 100 bytes are free
+    before calling jpeg_finish_compress().
+
+A more significant restriction is that jpeg_finish_compress() cannot suspend.
+This means you cannot use suspension with multi-pass operating modes, namely
+Huffman code optimization and multiple-scan output.  Those modes write the
+whole file during jpeg_finish_compress(), which will certainly result in
+buffer overrun.  (Note that this restriction applies only to compression,
+not decompression.  The decompressor supports input suspension in all of its
+operating modes.)
+
+Decompression suspension:
+
+For decompression suspension, use a fill_input_buffer() routine that simply
+returns FALSE (except perhaps during error recovery, as discussed below).
+This will cause the decompressor to return to its caller with an indication
+that suspension has occurred.  This can happen at four places:
+  * jpeg_read_header(): will return JPEG_SUSPENDED.
+  * jpeg_start_decompress(): will return FALSE, rather than its usual TRUE.
+  * jpeg_read_scanlines(): will return the number of scanlines already
+	completed (possibly 0).
+  * jpeg_finish_decompress(): will return FALSE, rather than its usual TRUE.
+The surrounding application must recognize these cases, load more data into
+the input buffer, and repeat the call.  In the case of jpeg_read_scanlines(),
+increment the passed pointers past any scanlines successfully read.
+
+Just as with compression, the decompressor will typically backtrack to a
+convenient restart point before suspending.  When fill_input_buffer() is
+called, next_input_byte/bytes_in_buffer point to the current restart point,
+which is where the decompressor will backtrack to if FALSE is returned.
+The data beyond that position must NOT be discarded if you suspend; it needs
+to be re-read upon resumption.  In most implementations, you'll need to shift
+this data down to the start of your work buffer and then load more data after
+it.  Again, this behavior means that a several-Kbyte work buffer is essential
+for decent performance; furthermore, you should load a reasonable amount of
+new data before resuming decompression.  (If you loaded, say, only one new
+byte each time around, you could waste a LOT of cycles.)
+
+The skip_input_data() source manager routine requires special care in a
+suspension scenario.  This routine is NOT granted the ability to suspend the
+decompressor; it can decrement bytes_in_buffer to zero, but no more.  If the
+requested skip distance exceeds the amount of data currently in the input
+buffer, then skip_input_data() must set bytes_in_buffer to zero and record the
+additional skip distance somewhere else.  The decompressor will immediately
+call fill_input_buffer(), which should return FALSE, which will cause a
+suspension return.  The surrounding application must then arrange to discard
+the recorded number of bytes before it resumes loading the input buffer.
+(Yes, this design is rather baroque, but it avoids complexity in the far more
+common case where a non-suspending source manager is used.)
+
+If the input data has been exhausted, we recommend that you emit a warning
+and insert dummy EOI markers just as a non-suspending data source manager
+would do.  This can be handled either in the surrounding application logic or
+within fill_input_buffer(); the latter is probably more efficient.  If
+fill_input_buffer() knows that no more data is available, it can set the
+pointer/count to point to a dummy EOI marker and then return TRUE just as
+though it had read more data in a non-suspending situation.
+
+The decompressor does not attempt to suspend within standard JPEG markers;
+instead it will backtrack to the start of the marker and reprocess the whole
+marker next time.  Hence the input buffer must be large enough to hold the
+longest standard marker in the file.  Standard JPEG markers should normally
+not exceed a few hundred bytes each (DHT tables are typically the longest).
+We recommend at least a 2K buffer for performance reasons, which is much
+larger than any correct marker is likely to be.  For robustness against
+damaged marker length counts, you may wish to insert a test in your
+application for the case that the input buffer is completely full and yet
+the decoder has suspended without consuming any data --- otherwise, if this
+situation did occur, it would lead to an endless loop.  (The library can't
+provide this test since it has no idea whether "the buffer is full", or
+even whether there is a fixed-size input buffer.)
+
+The input buffer would need to be 64K to allow for arbitrary COM or APPn
+markers, but these are handled specially: they are either saved into allocated
+memory, or skipped over by calling skip_input_data().  In the former case,
+suspension is handled correctly, and in the latter case, the problem of
+buffer overrun is placed on skip_input_data's shoulders, as explained above.
+Note that if you provide your own marker handling routine for large markers,
+you should consider how to deal with buffer overflow.
+
+Multiple-buffer management:
+
+In some applications it is desirable to store the compressed data in a linked
+list of buffer areas, so as to avoid data copying.  This can be handled by
+having empty_output_buffer() or fill_input_buffer() set the pointer and count
+to reference the next available buffer; FALSE is returned only if no more
+buffers are available.  Although seemingly straightforward, there is a
+pitfall in this approach: the backtrack that occurs when FALSE is returned
+could back up into an earlier buffer.  For example, when fill_input_buffer()
+is called, the current pointer & count indicate the backtrack restart point.
+Since fill_input_buffer() will set the pointer and count to refer to a new
+buffer, the restart position must be saved somewhere else.  Suppose a second
+call to fill_input_buffer() occurs in the same library call, and no
+additional input data is available, so fill_input_buffer must return FALSE.
+If the JPEG library has not moved the pointer/count forward in the current
+buffer, then *the correct restart point is the saved position in the prior
+buffer*.  Prior buffers may be discarded only after the library establishes
+a restart point within a later buffer.  Similar remarks apply for output into
+a chain of buffers.
+
+The library will never attempt to backtrack over a skip_input_data() call,
+so any skipped data can be permanently discarded.  You still have to deal
+with the case of skipping not-yet-received data, however.
+
+It's much simpler to use only a single buffer; when fill_input_buffer() is
+called, move any unconsumed data (beyond the current pointer/count) down to
+the beginning of this buffer and then load new data into the remaining buffer
+space.  This approach requires a little more data copying but is far easier
+to get right.
+
+
+Progressive JPEG support
+------------------------
+
+Progressive JPEG rearranges the stored data into a series of scans of
+increasing quality.  In situations where a JPEG file is transmitted across a
+slow communications link, a decoder can generate a low-quality image very
+quickly from the first scan, then gradually improve the displayed quality as
+more scans are received.  The final image after all scans are complete is
+identical to that of a regular (sequential) JPEG file of the same quality
+setting.  Progressive JPEG files are often slightly smaller than equivalent
+sequential JPEG files, but the possibility of incremental display is the main
+reason for using progressive JPEG.
+
+The IJG encoder library generates progressive JPEG files when given a
+suitable "scan script" defining how to divide the data into scans.
+Creation of progressive JPEG files is otherwise transparent to the encoder.
+Progressive JPEG files can also be read transparently by the decoder library.
+If the decoding application simply uses the library as defined above, it
+will receive a final decoded image without any indication that the file was
+progressive.  Of course, this approach does not allow incremental display.
+To perform incremental display, an application needs to use the decoder
+library's "buffered-image" mode, in which it receives a decoded image
+multiple times.
+
+Each displayed scan requires about as much work to decode as a full JPEG
+image of the same size, so the decoder must be fairly fast in relation to the
+data transmission rate in order to make incremental display useful.  However,
+it is possible to skip displaying the image and simply add the incoming bits
+to the decoder's coefficient buffer.  This is fast because only Huffman
+decoding need be done, not IDCT, upsampling, colorspace conversion, etc.
+The IJG decoder library allows the application to switch dynamically between
+displaying the image and simply absorbing the incoming bits.  A properly
+coded application can automatically adapt the number of display passes to
+suit the time available as the image is received.  Also, a final
+higher-quality display cycle can be performed from the buffered data after
+the end of the file is reached.
+
+Progressive compression:
+
+To create a progressive JPEG file (or a multiple-scan sequential JPEG file),
+set the scan_info cinfo field to point to an array of scan descriptors, and
+perform compression as usual.  Instead of constructing your own scan list,
+you can call the jpeg_simple_progression() helper routine to create a
+recommended progression sequence; this method should be used by all
+applications that don't want to get involved in the nitty-gritty of
+progressive scan sequence design.  (If you want to provide user control of
+scan sequences, you may wish to borrow the scan script reading code found
+in rdswitch.c, so that you can read scan script files just like cjpeg's.)
+When scan_info is not NULL, the compression library will store DCT'd data
+into a buffer array as jpeg_write_scanlines() is called, and will emit all
+the requested scans during jpeg_finish_compress().  This implies that
+multiple-scan output cannot be created with a suspending data destination
+manager, since jpeg_finish_compress() does not support suspension.  We
+should also note that the compressor currently forces Huffman optimization
+mode when creating a progressive JPEG file, because the default Huffman
+tables are unsuitable for progressive files.
+
+Progressive decompression:
+
+When buffered-image mode is not used, the decoder library will read all of
+a multi-scan file during jpeg_start_decompress(), so that it can provide a
+final decoded image.  (Here "multi-scan" means either progressive or
+multi-scan sequential.)  This makes multi-scan files transparent to the
+decoding application.  However, existing applications that used suspending
+input with version 5 of the IJG library will need to be modified to check
+for a suspension return from jpeg_start_decompress().
+
+To perform incremental display, an application must use the library's
+buffered-image mode.  This is described in the next section.
+
+
+Buffered-image mode
+-------------------
+
+In buffered-image mode, the library stores the partially decoded image in a
+coefficient buffer, from which it can be read out as many times as desired.
+This mode is typically used for incremental display of progressive JPEG files,
+but it can be used with any JPEG file.  Each scan of a progressive JPEG file
+adds more data (more detail) to the buffered image.  The application can
+display in lockstep with the source file (one display pass per input scan),
+or it can allow input processing to outrun display processing.  By making
+input and display processing run independently, it is possible for the
+application to adapt progressive display to a wide range of data transmission
+rates.
+
+The basic control flow for buffered-image decoding is
+
+	jpeg_create_decompress()
+	set data source
+	jpeg_read_header()
+	set overall decompression parameters
+	cinfo.buffered_image = TRUE;	/* select buffered-image mode */
+	jpeg_start_decompress()
+	for (each output pass) {
+	    adjust output decompression parameters if required
+	    jpeg_start_output()		/* start a new output pass */
+	    for (all scanlines in image) {
+	        jpeg_read_scanlines()
+	        display scanlines
+	    }
+	    jpeg_finish_output()	/* terminate output pass */
+	}
+	jpeg_finish_decompress()
+	jpeg_destroy_decompress()
+
+This differs from ordinary unbuffered decoding in that there is an additional
+level of looping.  The application can choose how many output passes to make
+and how to display each pass.
+
+The simplest approach to displaying progressive images is to do one display
+pass for each scan appearing in the input file.  In this case the outer loop
+condition is typically
+	while (! jpeg_input_complete(&cinfo))
+and the start-output call should read
+	jpeg_start_output(&cinfo, cinfo.input_scan_number);
+The second parameter to jpeg_start_output() indicates which scan of the input
+file is to be displayed; the scans are numbered starting at 1 for this
+purpose.  (You can use a loop counter starting at 1 if you like, but using
+the library's input scan counter is easier.)  The library automatically reads
+data as necessary to complete each requested scan, and jpeg_finish_output()
+advances to the next scan or end-of-image marker (hence input_scan_number
+will be incremented by the time control arrives back at jpeg_start_output()).
+With this technique, data is read from the input file only as needed, and
+input and output processing run in lockstep.
+
+After reading the final scan and reaching the end of the input file, the
+buffered image remains available; it can be read additional times by
+repeating the jpeg_start_output()/jpeg_read_scanlines()/jpeg_finish_output()
+sequence.  For example, a useful technique is to use fast one-pass color
+quantization for display passes made while the image is arriving, followed by
+a final display pass using two-pass quantization for highest quality.  This
+is done by changing the library parameters before the final output pass.
+Changing parameters between passes is discussed in detail below.
+
+In general the last scan of a progressive file cannot be recognized as such
+until after it is read, so a post-input display pass is the best approach if
+you want special processing in the final pass.
+
+When done with the image, be sure to call jpeg_finish_decompress() to release
+the buffered image (or just use jpeg_destroy_decompress()).
+
+If input data arrives faster than it can be displayed, the application can
+cause the library to decode input data in advance of what's needed to produce
+output.  This is done by calling the routine jpeg_consume_input().
+The return value is one of the following:
+	JPEG_REACHED_SOS:    reached an SOS marker (the start of a new scan)
+	JPEG_REACHED_EOI:    reached the EOI marker (end of image)
+	JPEG_ROW_COMPLETED:  completed reading one MCU row of compressed data
+	JPEG_SCAN_COMPLETED: completed reading last MCU row of current scan
+	JPEG_SUSPENDED:      suspended before completing any of the above
+(JPEG_SUSPENDED can occur only if a suspending data source is used.)  This
+routine can be called at any time after initializing the JPEG object.  It
+reads some additional data and returns when one of the indicated significant
+events occurs.  (If called after the EOI marker is reached, it will
+immediately return JPEG_REACHED_EOI without attempting to read more data.)
+
+The library's output processing will automatically call jpeg_consume_input()
+whenever the output processing overtakes the input; thus, simple lockstep
+display requires no direct calls to jpeg_consume_input().  But by adding
+calls to jpeg_consume_input(), you can absorb data in advance of what is
+being displayed.  This has two benefits:
+  * You can limit buildup of unprocessed data in your input buffer.
+  * You can eliminate extra display passes by paying attention to the
+    state of the library's input processing.
+
+The first of these benefits only requires interspersing calls to
+jpeg_consume_input() with your display operations and any other processing
+you may be doing.  To avoid wasting cycles due to backtracking, it's best to
+call jpeg_consume_input() only after a hundred or so new bytes have arrived.
+This is discussed further under "I/O suspension", above.  (Note: the JPEG
+library currently is not thread-safe.  You must not call jpeg_consume_input()
+from one thread of control if a different library routine is working on the
+same JPEG object in another thread.)
+
+When input arrives fast enough that more than one new scan is available
+before you start a new output pass, you may as well skip the output pass
+corresponding to the completed scan.  This occurs for free if you pass
+cinfo.input_scan_number as the target scan number to jpeg_start_output().
+The input_scan_number field is simply the index of the scan currently being
+consumed by the input processor.  You can ensure that this is up-to-date by
+emptying the input buffer just before calling jpeg_start_output(): call
+jpeg_consume_input() repeatedly until it returns JPEG_SUSPENDED or
+JPEG_REACHED_EOI.
+
+The target scan number passed to jpeg_start_output() is saved in the
+cinfo.output_scan_number field.  The library's output processing calls
+jpeg_consume_input() whenever the current input scan number and row within
+that scan is less than or equal to the current output scan number and row.
+Thus, input processing can "get ahead" of the output processing but is not
+allowed to "fall behind".  You can achieve several different effects by
+manipulating this interlock rule.  For example, if you pass a target scan
+number greater than the current input scan number, the output processor will
+wait until that scan starts to arrive before producing any output.  (To avoid
+an infinite loop, the target scan number is automatically reset to the last
+scan number when the end of image is reached.  Thus, if you specify a large
+target scan number, the library will just absorb the entire input file and
+then perform an output pass.  This is effectively the same as what
+jpeg_start_decompress() does when you don't select buffered-image mode.)
+When you pass a target scan number equal to the current input scan number,
+the image is displayed no faster than the current input scan arrives.  The
+final possibility is to pass a target scan number less than the current input
+scan number; this disables the input/output interlock and causes the output
+processor to simply display whatever it finds in the image buffer, without
+waiting for input.  (However, the library will not accept a target scan
+number less than one, so you can't avoid waiting for the first scan.)
+
+When data is arriving faster than the output display processing can advance
+through the image, jpeg_consume_input() will store data into the buffered
+image beyond the point at which the output processing is reading data out
+again.  If the input arrives fast enough, it may "wrap around" the buffer to
+the point where the input is more than one whole scan ahead of the output.
+If the output processing simply proceeds through its display pass without
+paying attention to the input, the effect seen on-screen is that the lower
+part of the image is one or more scans better in quality than the upper part.
+Then, when the next output scan is started, you have a choice of what target
+scan number to use.  The recommended choice is to use the current input scan
+number at that time, which implies that you've skipped the output scans
+corresponding to the input scans that were completed while you processed the
+previous output scan.  In this way, the decoder automatically adapts its
+speed to the arriving data, by skipping output scans as necessary to keep up
+with the arriving data.
+
+When using this strategy, you'll want to be sure that you perform a final
+output pass after receiving all the data; otherwise your last display may not
+be full quality across the whole screen.  So the right outer loop logic is
+something like this:
+	do {
+	    absorb any waiting input by calling jpeg_consume_input()
+	    final_pass = jpeg_input_complete(&cinfo);
+	    adjust output decompression parameters if required
+	    jpeg_start_output(&cinfo, cinfo.input_scan_number);
+	    ...
+	    jpeg_finish_output()
+	} while (! final_pass);
+rather than quitting as soon as jpeg_input_complete() returns TRUE.  This
+arrangement makes it simple to use higher-quality decoding parameters
+for the final pass.  But if you don't want to use special parameters for
+the final pass, the right loop logic is like this:
+	for (;;) {
+	    absorb any waiting input by calling jpeg_consume_input()
+	    jpeg_start_output(&cinfo, cinfo.input_scan_number);
+	    ...
+	    jpeg_finish_output()
+	    if (jpeg_input_complete(&cinfo) &&
+	        cinfo.input_scan_number == cinfo.output_scan_number)
+	      break;
+	}
+In this case you don't need to know in advance whether an output pass is to
+be the last one, so it's not necessary to have reached EOF before starting
+the final output pass; rather, what you want to test is whether the output
+pass was performed in sync with the final input scan.  This form of the loop
+will avoid an extra output pass whenever the decoder is able (or nearly able)
+to keep up with the incoming data.
+
+When the data transmission speed is high, you might begin a display pass,
+then find that much or all of the file has arrived before you can complete
+the pass.  (You can detect this by noting the JPEG_REACHED_EOI return code
+from jpeg_consume_input(), or equivalently by testing jpeg_input_complete().)
+In this situation you may wish to abort the current display pass and start a
+new one using the newly arrived information.  To do so, just call
+jpeg_finish_output() and then start a new pass with jpeg_start_output().
+
+A variant strategy is to abort and restart display if more than one complete
+scan arrives during an output pass; this can be detected by noting
+JPEG_REACHED_SOS returns and/or examining cinfo.input_scan_number.  This
+idea should be employed with caution, however, since the display process
+might never get to the bottom of the image before being aborted, resulting
+in the lower part of the screen being several passes worse than the upper.
+In most cases it's probably best to abort an output pass only if the whole
+file has arrived and you want to begin the final output pass immediately.
+
+When receiving data across a communication link, we recommend always using
+the current input scan number for the output target scan number; if a
+higher-quality final pass is to be done, it should be started (aborting any
+incomplete output pass) as soon as the end of file is received.  However,
+many other strategies are possible.  For example, the application can examine
+the parameters of the current input scan and decide whether to display it or
+not.  If the scan contains only chroma data, one might choose not to use it
+as the target scan, expecting that the scan will be small and will arrive
+quickly.  To skip to the next scan, call jpeg_consume_input() until it
+returns JPEG_REACHED_SOS or JPEG_REACHED_EOI.  Or just use the next higher
+number as the target scan for jpeg_start_output(); but that method doesn't
+let you inspect the next scan's parameters before deciding to display it.
+
+
+In buffered-image mode, jpeg_start_decompress() never performs input and
+thus never suspends.  An application that uses input suspension with
+buffered-image mode must be prepared for suspension returns from these
+routines:
+* jpeg_start_output() performs input only if you request 2-pass quantization
+  and the target scan isn't fully read yet.  (This is discussed below.)
+* jpeg_read_scanlines(), as always, returns the number of scanlines that it
+  was able to produce before suspending.
+* jpeg_finish_output() will read any markers following the target scan,
+  up to the end of the file or the SOS marker that begins another scan.
+  (But it reads no input if jpeg_consume_input() has already reached the
+  end of the file or a SOS marker beyond the target output scan.)
+* jpeg_finish_decompress() will read until the end of file, and thus can
+  suspend if the end hasn't already been reached (as can be tested by
+  calling jpeg_input_complete()).
+jpeg_start_output(), jpeg_finish_output(), and jpeg_finish_decompress()
+all return TRUE if they completed their tasks, FALSE if they had to suspend.
+In the event of a FALSE return, the application must load more input data
+and repeat the call.  Applications that use non-suspending data sources need
+not check the return values of these three routines.
+
+
+It is possible to change decoding parameters between output passes in the
+buffered-image mode.  The decoder library currently supports only very
+limited changes of parameters.  ONLY THE FOLLOWING parameter changes are
+allowed after jpeg_start_decompress() is called:
+* dct_method can be changed before each call to jpeg_start_output().
+  For example, one could use a fast DCT method for early scans, changing
+  to a higher quality method for the final scan.
+* dither_mode can be changed before each call to jpeg_start_output();
+  of course this has no impact if not using color quantization.  Typically
+  one would use ordered dither for initial passes, then switch to
+  Floyd-Steinberg dither for the final pass.  Caution: changing dither mode
+  can cause more memory to be allocated by the library.  Although the amount
+  of memory involved is not large (a scanline or so), it may cause the
+  initial max_memory_to_use specification to be exceeded, which in the worst
+  case would result in an out-of-memory failure.
+* do_block_smoothing can be changed before each call to jpeg_start_output().
+  This setting is relevant only when decoding a progressive JPEG image.
+  During the first DC-only scan, block smoothing provides a very "fuzzy" look
+  instead of the very "blocky" look seen without it; which is better seems a
+  matter of personal taste.  But block smoothing is nearly always a win
+  during later stages, especially when decoding a successive-approximation
+  image: smoothing helps to hide the slight blockiness that otherwise shows
+  up on smooth gradients until the lowest coefficient bits are sent.
+* Color quantization mode can be changed under the rules described below.
+  You *cannot* change between full-color and quantized output (because that
+  would alter the required I/O buffer sizes), but you can change which
+  quantization method is used.
+
+When generating color-quantized output, changing quantization method is a
+very useful way of switching between high-speed and high-quality display.
+The library allows you to change among its three quantization methods:
+1. Single-pass quantization to a fixed color cube.
+   Selected by cinfo.two_pass_quantize = FALSE and cinfo.colormap = NULL.
+2. Single-pass quantization to an application-supplied colormap.
+   Selected by setting cinfo.colormap to point to the colormap (the value of
+   two_pass_quantize is ignored); also set cinfo.actual_number_of_colors.
+3. Two-pass quantization to a colormap chosen specifically for the image.
+   Selected by cinfo.two_pass_quantize = TRUE and cinfo.colormap = NULL.
+   (This is the default setting selected by jpeg_read_header, but it is
+   probably NOT what you want for the first pass of progressive display!)
+These methods offer successively better quality and lesser speed.  However,
+only the first method is available for quantizing in non-RGB color spaces.
+
+IMPORTANT: because the different quantizer methods have very different
+working-storage requirements, the library requires you to indicate which
+one(s) you intend to use before you call jpeg_start_decompress().  (If we did
+not require this, the max_memory_to_use setting would be a complete fiction.)
+You do this by setting one or more of these three cinfo fields to TRUE:
+	enable_1pass_quant		Fixed color cube colormap
+	enable_external_quant		Externally-supplied colormap
+	enable_2pass_quant		Two-pass custom colormap
+All three are initialized FALSE by jpeg_read_header().  But
+jpeg_start_decompress() automatically sets TRUE the one selected by the
+current two_pass_quantize and colormap settings, so you only need to set the
+enable flags for any other quantization methods you plan to change to later.
+
+After setting the enable flags correctly at jpeg_start_decompress() time, you
+can change to any enabled quantization method by setting two_pass_quantize
+and colormap properly just before calling jpeg_start_output().  The following
+special rules apply:
+1. You must explicitly set cinfo.colormap to NULL when switching to 1-pass
+   or 2-pass mode from a different mode, or when you want the 2-pass
+   quantizer to be re-run to generate a new colormap.
+2. To switch to an external colormap, or to change to a different external
+   colormap than was used on the prior pass, you must call
+   jpeg_new_colormap() after setting cinfo.colormap.
+NOTE: if you want to use the same colormap as was used in the prior pass,
+you should not do either of these things.  This will save some nontrivial
+switchover costs.
+(These requirements exist because cinfo.colormap will always be non-NULL
+after completing a prior output pass, since both the 1-pass and 2-pass
+quantizers set it to point to their output colormaps.  Thus you have to
+do one of these two things to notify the library that something has changed.
+Yup, it's a bit klugy, but it's necessary to do it this way for backwards
+compatibility.)
+
+Note that in buffered-image mode, the library generates any requested colormap
+during jpeg_start_output(), not during jpeg_start_decompress().
+
+When using two-pass quantization, jpeg_start_output() makes a pass over the
+buffered image to determine the optimum color map; it therefore may take a
+significant amount of time, whereas ordinarily it does little work.  The
+progress monitor hook is called during this pass, if defined.  It is also
+important to realize that if the specified target scan number is greater than
+or equal to the current input scan number, jpeg_start_output() will attempt
+to consume input as it makes this pass.  If you use a suspending data source,
+you need to check for a FALSE return from jpeg_start_output() under these
+conditions.  The combination of 2-pass quantization and a not-yet-fully-read
+target scan is the only case in which jpeg_start_output() will consume input.
+
+
+Application authors who support buffered-image mode may be tempted to use it
+for all JPEG images, even single-scan ones.  This will work, but it is
+inefficient: there is no need to create an image-sized coefficient buffer for
+single-scan images.  Requesting buffered-image mode for such an image wastes
+memory.  Worse, it can cost time on large images, since the buffered data has
+to be swapped out or written to a temporary file.  If you are concerned about
+maximum performance on baseline JPEG files, you should use buffered-image
+mode only when the incoming file actually has multiple scans.  This can be
+tested by calling jpeg_has_multiple_scans(), which will return a correct
+result at any time after jpeg_read_header() completes.
+
+It is also worth noting that when you use jpeg_consume_input() to let input
+processing get ahead of output processing, the resulting pattern of access to
+the coefficient buffer is quite nonsequential.  It's best to use the memory
+manager jmemnobs.c if you can (ie, if you have enough real or virtual main
+memory).  If not, at least make sure that max_memory_to_use is set as high as
+possible.  If the JPEG memory manager has to use a temporary file, you will
+probably see a lot of disk traffic and poor performance.  (This could be
+improved with additional work on the memory manager, but we haven't gotten
+around to it yet.)
+
+In some applications it may be convenient to use jpeg_consume_input() for all
+input processing, including reading the initial markers; that is, you may
+wish to call jpeg_consume_input() instead of jpeg_read_header() during
+startup.  This works, but note that you must check for JPEG_REACHED_SOS and
+JPEG_REACHED_EOI return codes as the equivalent of jpeg_read_header's codes.
+Once the first SOS marker has been reached, you must call
+jpeg_start_decompress() before jpeg_consume_input() will consume more input;
+it'll just keep returning JPEG_REACHED_SOS until you do.  If you read a
+tables-only file this way, jpeg_consume_input() will return JPEG_REACHED_EOI
+without ever returning JPEG_REACHED_SOS; be sure to check for this case.
+If this happens, the decompressor will not read any more input until you call
+jpeg_abort() to reset it.  It is OK to call jpeg_consume_input() even when not
+using buffered-image mode, but in that case it's basically a no-op after the
+initial markers have been read: it will just return JPEG_SUSPENDED.
+
+
+Abbreviated datastreams and multiple images
+-------------------------------------------
+
+A JPEG compression or decompression object can be reused to process multiple
+images.  This saves a small amount of time per image by eliminating the
+"create" and "destroy" operations, but that isn't the real purpose of the
+feature.  Rather, reuse of an object provides support for abbreviated JPEG
+datastreams.  Object reuse can also simplify processing a series of images in
+a single input or output file.  This section explains these features.
+
+A JPEG file normally contains several hundred bytes worth of quantization
+and Huffman tables.  In a situation where many images will be stored or
+transmitted with identical tables, this may represent an annoying overhead.
+The JPEG standard therefore permits tables to be omitted.  The standard
+defines three classes of JPEG datastreams:
+  * "Interchange" datastreams contain an image and all tables needed to decode
+     the image.  These are the usual kind of JPEG file.
+  * "Abbreviated image" datastreams contain an image, but are missing some or
+    all of the tables needed to decode that image.
+  * "Abbreviated table specification" (henceforth "tables-only") datastreams
+    contain only table specifications.
+To decode an abbreviated image, it is necessary to load the missing table(s)
+into the decoder beforehand.  This can be accomplished by reading a separate
+tables-only file.  A variant scheme uses a series of images in which the first
+image is an interchange (complete) datastream, while subsequent ones are
+abbreviated and rely on the tables loaded by the first image.  It is assumed
+that once the decoder has read a table, it will remember that table until a
+new definition for the same table number is encountered.
+
+It is the application designer's responsibility to figure out how to associate
+the correct tables with an abbreviated image.  While abbreviated datastreams
+can be useful in a closed environment, their use is strongly discouraged in
+any situation where data exchange with other applications might be needed.
+Caveat designer.
+
+The JPEG library provides support for reading and writing any combination of
+tables-only datastreams and abbreviated images.  In both compression and
+decompression objects, a quantization or Huffman table will be retained for
+the lifetime of the object, unless it is overwritten by a new table definition.
+
+
+To create abbreviated image datastreams, it is only necessary to tell the
+compressor not to emit some or all of the tables it is using.  Each
+quantization and Huffman table struct contains a boolean field "sent_table",
+which normally is initialized to FALSE.  For each table used by the image, the
+header-writing process emits the table and sets sent_table = TRUE unless it is
+already TRUE.  (In normal usage, this prevents outputting the same table
+definition multiple times, as would otherwise occur because the chroma
+components typically share tables.)  Thus, setting this field to TRUE before
+calling jpeg_start_compress() will prevent the table from being written at
+all.
+
+If you want to create a "pure" abbreviated image file containing no tables,
+just call "jpeg_suppress_tables(&cinfo, TRUE)" after constructing all the
+tables.  If you want to emit some but not all tables, you'll need to set the
+individual sent_table fields directly.
+
+To create an abbreviated image, you must also call jpeg_start_compress()
+with a second parameter of FALSE, not TRUE.  Otherwise jpeg_start_compress()
+will force all the sent_table fields to FALSE.  (This is a safety feature to
+prevent abbreviated images from being created accidentally.)
+
+To create a tables-only file, perform the same parameter setup that you
+normally would, but instead of calling jpeg_start_compress() and so on, call
+jpeg_write_tables(&cinfo).  This will write an abbreviated datastream
+containing only SOI, DQT and/or DHT markers, and EOI.  All the quantization
+and Huffman tables that are currently defined in the compression object will
+be emitted unless their sent_tables flag is already TRUE, and then all the
+sent_tables flags will be set TRUE.
+
+A sure-fire way to create matching tables-only and abbreviated image files
+is to proceed as follows:
+
+	create JPEG compression object
+	set JPEG parameters
+	set destination to tables-only file
+	jpeg_write_tables(&cinfo);
+	set destination to image file
+	jpeg_start_compress(&cinfo, FALSE);
+	write data...
+	jpeg_finish_compress(&cinfo);
+
+Since the JPEG parameters are not altered between writing the table file and
+the abbreviated image file, the same tables are sure to be used.  Of course,
+you can repeat the jpeg_start_compress() ... jpeg_finish_compress() sequence
+many times to produce many abbreviated image files matching the table file.
+
+You cannot suppress output of the computed Huffman tables when Huffman
+optimization is selected.  (If you could, there'd be no way to decode the
+image...)  Generally, you don't want to set optimize_coding = TRUE when
+you are trying to produce abbreviated files.
+
+In some cases you might want to compress an image using tables which are
+not stored in the application, but are defined in an interchange or
+tables-only file readable by the application.  This can be done by setting up
+a JPEG decompression object to read the specification file, then copying the
+tables into your compression object.  See jpeg_copy_critical_parameters()
+for an example of copying quantization tables.
+
+
+To read abbreviated image files, you simply need to load the proper tables
+into the decompression object before trying to read the abbreviated image.
+If the proper tables are stored in the application program, you can just
+allocate the table structs and fill in their contents directly.  For example,
+to load a fixed quantization table into table slot "n":
+
+    if (cinfo.quant_tbl_ptrs[n] == NULL)
+      cinfo.quant_tbl_ptrs[n] = jpeg_alloc_quant_table((j_common_ptr) &cinfo);
+    quant_ptr = cinfo.quant_tbl_ptrs[n];	/* quant_ptr is JQUANT_TBL* */
+    for (i = 0; i < 64; i++) {
+      /* Qtable[] is desired quantization table, in natural array order */
+      quant_ptr->quantval[i] = Qtable[i];
+    }
+
+Code to load a fixed Huffman table is typically (for AC table "n"):
+
+    if (cinfo.ac_huff_tbl_ptrs[n] == NULL)
+      cinfo.ac_huff_tbl_ptrs[n] = jpeg_alloc_huff_table((j_common_ptr) &cinfo);
+    huff_ptr = cinfo.ac_huff_tbl_ptrs[n];	/* huff_ptr is JHUFF_TBL* */
+    for (i = 1; i <= 16; i++) {
+      /* counts[i] is number of Huffman codes of length i bits, i=1..16 */
+      huff_ptr->bits[i] = counts[i];
+    }
+    for (i = 0; i < 256; i++) {
+      /* symbols[] is the list of Huffman symbols, in code-length order */
+      huff_ptr->huffval[i] = symbols[i];
+    }
+
+(Note that trying to set cinfo.quant_tbl_ptrs[n] to point directly at a
+constant JQUANT_TBL object is not safe.  If the incoming file happened to
+contain a quantization table definition, your master table would get
+overwritten!  Instead allocate a working table copy and copy the master table
+into it, as illustrated above.  Ditto for Huffman tables, of course.)
+
+You might want to read the tables from a tables-only file, rather than
+hard-wiring them into your application.  The jpeg_read_header() call is
+sufficient to read a tables-only file.  You must pass a second parameter of
+FALSE to indicate that you do not require an image to be present.  Thus, the
+typical scenario is
+
+	create JPEG decompression object
+	set source to tables-only file
+	jpeg_read_header(&cinfo, FALSE);
+	set source to abbreviated image file
+	jpeg_read_header(&cinfo, TRUE);
+	set decompression parameters
+	jpeg_start_decompress(&cinfo);
+	read data...
+	jpeg_finish_decompress(&cinfo);
+
+In some cases, you may want to read a file without knowing whether it contains
+an image or just tables.  In that case, pass FALSE and check the return value
+from jpeg_read_header(): it will be JPEG_HEADER_OK if an image was found,
+JPEG_HEADER_TABLES_ONLY if only tables were found.  (A third return value,
+JPEG_SUSPENDED, is possible when using a suspending data source manager.)
+Note that jpeg_read_header() will not complain if you read an abbreviated
+image for which you haven't loaded the missing tables; the missing-table check
+occurs later, in jpeg_start_decompress().
+
+
+It is possible to read a series of images from a single source file by
+repeating the jpeg_read_header() ... jpeg_finish_decompress() sequence,
+without releasing/recreating the JPEG object or the data source module.
+(If you did reinitialize, any partial bufferload left in the data source
+buffer at the end of one image would be discarded, causing you to lose the
+start of the next image.)  When you use this method, stored tables are
+automatically carried forward, so some of the images can be abbreviated images
+that depend on tables from earlier images.
+
+If you intend to write a series of images into a single destination file,
+you might want to make a specialized data destination module that doesn't
+flush the output buffer at term_destination() time.  This would speed things
+up by some trifling amount.  Of course, you'd need to remember to flush the
+buffer after the last image.  You can make the later images be abbreviated
+ones by passing FALSE to jpeg_start_compress().
+
+
+Special markers
+---------------
+
+Some applications may need to insert or extract special data in the JPEG
+datastream.  The JPEG standard provides marker types "COM" (comment) and
+"APP0" through "APP15" (application) to hold application-specific data.
+Unfortunately, the use of these markers is not specified by the standard.
+COM markers are fairly widely used to hold user-supplied text.  The JFIF file
+format spec uses APP0 markers with specified initial strings to hold certain
+data.  Adobe applications use APP14 markers beginning with the string "Adobe"
+for miscellaneous data.  Other APPn markers are rarely seen, but might
+contain almost anything.
+
+If you wish to store user-supplied text, we recommend you use COM markers
+and place readable 7-bit ASCII text in them.  Newline conventions are not
+standardized --- expect to find LF (Unix style), CR/LF (DOS style), or CR
+(Mac style).  A robust COM reader should be able to cope with random binary
+garbage, including nulls, since some applications generate COM markers
+containing non-ASCII junk.  (But yours should not be one of them.)
+
+For program-supplied data, use an APPn marker, and be sure to begin it with an
+identifying string so that you can tell whether the marker is actually yours.
+It's probably best to avoid using APP0 or APP14 for any private markers.
+(NOTE: the upcoming SPIFF standard will use APP8 markers; we recommend you
+not use APP8 markers for any private purposes, either.)
+
+Keep in mind that at most 65533 bytes can be put into one marker, but you
+can have as many markers as you like.
+
+By default, the IJG compression library will write a JFIF APP0 marker if the
+selected JPEG colorspace is grayscale or YCbCr, or an Adobe APP14 marker if
+the selected colorspace is RGB, CMYK, or YCCK.  You can disable this, but
+we don't recommend it.  The decompression library will recognize JFIF and
+Adobe markers and will set the JPEG colorspace properly when one is found.
+
+
+You can write special markers immediately following the datastream header by
+calling jpeg_write_marker() after jpeg_start_compress() and before the first
+call to jpeg_write_scanlines().  When you do this, the markers appear after
+the SOI and the JFIF APP0 and Adobe APP14 markers (if written), but before
+all else.  Specify the marker type parameter as "JPEG_COM" for COM or
+"JPEG_APP0 + n" for APPn.  (Actually, jpeg_write_marker will let you write
+any marker type, but we don't recommend writing any other kinds of marker.)
+For example, to write a user comment string pointed to by comment_text:
+	jpeg_write_marker(cinfo, JPEG_COM, comment_text, strlen(comment_text));
+
+If it's not convenient to store all the marker data in memory at once,
+you can instead call jpeg_write_m_header() followed by multiple calls to
+jpeg_write_m_byte().  If you do it this way, it's your responsibility to
+call jpeg_write_m_byte() exactly the number of times given in the length
+parameter to jpeg_write_m_header().  (This method lets you empty the
+output buffer partway through a marker, which might be important when
+using a suspending data destination module.  In any case, if you are using
+a suspending destination, you should flush its buffer after inserting
+any special markers.  See "I/O suspension".)
+
+Or, if you prefer to synthesize the marker byte sequence yourself,
+you can just cram it straight into the data destination module.
+
+If you are writing JFIF 1.02 extension markers (thumbnail images), don't
+forget to set cinfo.JFIF_minor_version = 2 so that the encoder will write the
+correct JFIF version number in the JFIF header marker.  The library's default
+is to write version 1.01, but that's wrong if you insert any 1.02 extension
+markers.  (We could probably get away with just defaulting to 1.02, but there
+used to be broken decoders that would complain about unknown minor version
+numbers.  To reduce compatibility risks it's safest not to write 1.02 unless
+you are actually using 1.02 extensions.)
+
+
+When reading, two methods of handling special markers are available:
+1. You can ask the library to save the contents of COM and/or APPn markers
+into memory, and then examine them at your leisure afterwards.
+2. You can supply your own routine to process COM and/or APPn markers
+on-the-fly as they are read.
+The first method is simpler to use, especially if you are using a suspending
+data source; writing a marker processor that copes with input suspension is
+not easy (consider what happens if the marker is longer than your available
+input buffer).  However, the second method conserves memory since the marker
+data need not be kept around after it's been processed.
+
+For either method, you'd normally set up marker handling after creating a
+decompression object and before calling jpeg_read_header(), because the
+markers of interest will typically be near the head of the file and so will
+be scanned by jpeg_read_header.  Once you've established a marker handling
+method, it will be used for the life of that decompression object
+(potentially many datastreams), unless you change it.  Marker handling is
+determined separately for COM markers and for each APPn marker code.
+
+
+To save the contents of special markers in memory, call
+	jpeg_save_markers(cinfo, marker_code, length_limit)
+where marker_code is the marker type to save, JPEG_COM or JPEG_APP0+n.
+(To arrange to save all the special marker types, you need to call this
+routine 17 times, for COM and APP0-APP15.)  If the incoming marker is longer
+than length_limit data bytes, only length_limit bytes will be saved; this
+parameter allows you to avoid chewing up memory when you only need to see the
+first few bytes of a potentially large marker.  If you want to save all the
+data, set length_limit to 0xFFFF; that is enough since marker lengths are only
+16 bits.  As a special case, setting length_limit to 0 prevents that marker
+type from being saved at all.  (That is the default behavior, in fact.)
+
+After jpeg_read_header() completes, you can examine the special markers by
+following the cinfo->marker_list pointer chain.  All the special markers in
+the file appear in this list, in order of their occurrence in the file (but
+omitting any markers of types you didn't ask for).  Both the original data
+length and the saved data length are recorded for each list entry; the latter
+will not exceed length_limit for the particular marker type.  Note that these
+lengths exclude the marker length word, whereas the stored representation
+within the JPEG file includes it.  (Hence the maximum data length is really
+only 65533.)
+
+It is possible that additional special markers appear in the file beyond the
+SOS marker at which jpeg_read_header stops; if so, the marker list will be
+extended during reading of the rest of the file.  This is not expected to be
+common, however.  If you are short on memory you may want to reset the length
+limit to zero for all marker types after finishing jpeg_read_header, to
+ensure that the max_memory_to_use setting cannot be exceeded due to addition
+of later markers.
+
+The marker list remains stored until you call jpeg_finish_decompress or
+jpeg_abort, at which point the memory is freed and the list is set to empty.
+(jpeg_destroy also releases the storage, of course.)
+
+Note that the library is internally interested in APP0 and APP14 markers;
+if you try to set a small nonzero length limit on these types, the library
+will silently force the length up to the minimum it wants.  (But you can set
+a zero length limit to prevent them from being saved at all.)  Also, in a
+16-bit environment, the maximum length limit may be constrained to less than
+65533 by malloc() limitations.  It is therefore best not to assume that the
+effective length limit is exactly what you set it to be.
+
+
+If you want to supply your own marker-reading routine, you do it by calling
+jpeg_set_marker_processor().  A marker processor routine must have the
+signature
+	boolean jpeg_marker_parser_method (j_decompress_ptr cinfo)
+Although the marker code is not explicitly passed, the routine can find it
+in cinfo->unread_marker.  At the time of call, the marker proper has been
+read from the data source module.  The processor routine is responsible for
+reading the marker length word and the remaining parameter bytes, if any.
+Return TRUE to indicate success.  (FALSE should be returned only if you are
+using a suspending data source and it tells you to suspend.  See the standard
+marker processors in jdmarker.c for appropriate coding methods if you need to
+use a suspending data source.)
+
+If you override the default APP0 or APP14 processors, it is up to you to
+recognize JFIF and Adobe markers if you want colorspace recognition to occur
+properly.  We recommend copying and extending the default processors if you
+want to do that.  (A better idea is to save these marker types for later
+examination by calling jpeg_save_markers(); that method doesn't interfere
+with the library's own processing of these markers.)
+
+jpeg_set_marker_processor() and jpeg_save_markers() are mutually exclusive
+--- if you call one it overrides any previous call to the other, for the
+particular marker type specified.
+
+A simple example of an external COM processor can be found in djpeg.c.
+Also, see jpegtran.c for an example of using jpeg_save_markers.
+
+
+Raw (downsampled) image data
+----------------------------
+
+Some applications need to supply already-downsampled image data to the JPEG
+compressor, or to receive raw downsampled data from the decompressor.  The
+library supports this requirement by allowing the application to write or
+read raw data, bypassing the normal preprocessing or postprocessing steps.
+The interface is different from the standard one and is somewhat harder to
+use.  If your interest is merely in bypassing color conversion, we recommend
+that you use the standard interface and simply set jpeg_color_space =
+in_color_space (or jpeg_color_space = out_color_space for decompression).
+The mechanism described in this section is necessary only to supply or
+receive downsampled image data, in which not all components have the same
+dimensions.
+
+
+To compress raw data, you must supply the data in the colorspace to be used
+in the JPEG file (please read the earlier section on Special color spaces)
+and downsampled to the sampling factors specified in the JPEG parameters.
+You must supply the data in the format used internally by the JPEG library,
+namely a JSAMPIMAGE array.  This is an array of pointers to two-dimensional
+arrays, each of type JSAMPARRAY.  Each 2-D array holds the values for one
+color component.  This structure is necessary since the components are of
+different sizes.  If the image dimensions are not a multiple of the MCU size,
+you must also pad the data correctly (usually, this is done by replicating
+the last column and/or row).  The data must be padded to a multiple of a DCT
+block in each component: that is, each downsampled row must contain a
+multiple of 8 valid samples, and there must be a multiple of 8 sample rows
+for each component.  (For applications such as conversion of digital TV
+images, the standard image size is usually a multiple of the DCT block size,
+so that no padding need actually be done.)
+
+The procedure for compression of raw data is basically the same as normal
+compression, except that you call jpeg_write_raw_data() in place of
+jpeg_write_scanlines().  Before calling jpeg_start_compress(), you must do
+the following:
+  * Set cinfo->raw_data_in to TRUE.  (It is set FALSE by jpeg_set_defaults().)
+    This notifies the library that you will be supplying raw data.
+  * Ensure jpeg_color_space is correct --- an explicit jpeg_set_colorspace()
+    call is a good idea.  Note that since color conversion is bypassed,
+    in_color_space is ignored, except that jpeg_set_defaults() uses it to
+    choose the default jpeg_color_space setting.
+  * Ensure the sampling factors, cinfo->comp_info[i].h_samp_factor and
+    cinfo->comp_info[i].v_samp_factor, are correct.  Since these indicate the
+    dimensions of the data you are supplying, it's wise to set them
+    explicitly, rather than assuming the library's defaults are what you want.
+
+To pass raw data to the library, call jpeg_write_raw_data() in place of
+jpeg_write_scanlines().  The two routines work similarly except that
+jpeg_write_raw_data takes a JSAMPIMAGE data array rather than JSAMPARRAY.
+The scanlines count passed to and returned from jpeg_write_raw_data is
+measured in terms of the component with the largest v_samp_factor.
+
+jpeg_write_raw_data() processes one MCU row per call, which is to say
+v_samp_factor*DCTSIZE sample rows of each component.  The passed num_lines
+value must be at least max_v_samp_factor*DCTSIZE, and the return value will
+be exactly that amount (or possibly some multiple of that amount, in future
+library versions).  This is true even on the last call at the bottom of the
+image; don't forget to pad your data as necessary.
+
+The required dimensions of the supplied data can be computed for each
+component as
+	cinfo->comp_info[i].width_in_blocks*DCTSIZE  samples per row
+	cinfo->comp_info[i].height_in_blocks*DCTSIZE rows in image
+after jpeg_start_compress() has initialized those fields.  If the valid data
+is smaller than this, it must be padded appropriately.  For some sampling
+factors and image sizes, additional dummy DCT blocks are inserted to make
+the image a multiple of the MCU dimensions.  The library creates such dummy
+blocks itself; it does not read them from your supplied data.  Therefore you
+need never pad by more than DCTSIZE samples.  An example may help here.
+Assume 2h2v downsampling of YCbCr data, that is
+	cinfo->comp_info[0].h_samp_factor = 2		for Y
+	cinfo->comp_info[0].v_samp_factor = 2
+	cinfo->comp_info[1].h_samp_factor = 1		for Cb
+	cinfo->comp_info[1].v_samp_factor = 1
+	cinfo->comp_info[2].h_samp_factor = 1		for Cr
+	cinfo->comp_info[2].v_samp_factor = 1
+and suppose that the nominal image dimensions (cinfo->image_width and
+cinfo->image_height) are 101x101 pixels.  Then jpeg_start_compress() will
+compute downsampled_width = 101 and width_in_blocks = 13 for Y,
+downsampled_width = 51 and width_in_blocks = 7 for Cb and Cr (and the same
+for the height fields).  You must pad the Y data to at least 13*8 = 104
+columns and rows, the Cb/Cr data to at least 7*8 = 56 columns and rows.  The
+MCU height is max_v_samp_factor = 2 DCT rows so you must pass at least 16
+scanlines on each call to jpeg_write_raw_data(), which is to say 16 actual
+sample rows of Y and 8 each of Cb and Cr.  A total of 7 MCU rows are needed,
+so you must pass a total of 7*16 = 112 "scanlines".  The last DCT block row
+of Y data is dummy, so it doesn't matter what you pass for it in the data
+arrays, but the scanlines count must total up to 112 so that all of the Cb
+and Cr data gets passed.
+
+Output suspension is supported with raw-data compression: if the data
+destination module suspends, jpeg_write_raw_data() will return 0.
+In this case the same data rows must be passed again on the next call.
+
+
+Decompression with raw data output implies bypassing all postprocessing:
+you cannot ask for rescaling or color quantization, for instance.  More
+seriously, you must deal with the color space and sampling factors present in
+the incoming file.  If your application only handles, say, 2h1v YCbCr data,
+you must check for and fail on other color spaces or other sampling factors.
+The library will not convert to a different color space for you.
+
+To obtain raw data output, set cinfo->raw_data_out = TRUE before
+jpeg_start_decompress() (it is set FALSE by jpeg_read_header()).  Be sure to
+verify that the color space and sampling factors are ones you can handle.
+Then call jpeg_read_raw_data() in place of jpeg_read_scanlines().  The
+decompression process is otherwise the same as usual.
+
+jpeg_read_raw_data() returns one MCU row per call, and thus you must pass a
+buffer of at least max_v_samp_factor*DCTSIZE scanlines (scanline counting is
+the same as for raw-data compression).  The buffer you pass must be large
+enough to hold the actual data plus padding to DCT-block boundaries.  As with
+compression, any entirely dummy DCT blocks are not processed so you need not
+allocate space for them, but the total scanline count includes them.  The
+above example of computing buffer dimensions for raw-data compression is
+equally valid for decompression.
+
+Input suspension is supported with raw-data decompression: if the data source
+module suspends, jpeg_read_raw_data() will return 0.  You can also use
+buffered-image mode to read raw data in multiple passes.
+
+
+Really raw data: DCT coefficients
+---------------------------------
+
+It is possible to read or write the contents of a JPEG file as raw DCT
+coefficients.  This facility is mainly intended for use in lossless
+transcoding between different JPEG file formats.  Other possible applications
+include lossless cropping of a JPEG image, lossless reassembly of a
+multi-strip or multi-tile TIFF/JPEG file into a single JPEG datastream, etc.
+
+To read the contents of a JPEG file as DCT coefficients, open the file and do
+jpeg_read_header() as usual.  But instead of calling jpeg_start_decompress()
+and jpeg_read_scanlines(), call jpeg_read_coefficients().  This will read the
+entire image into a set of virtual coefficient-block arrays, one array per
+component.  The return value is a pointer to an array of virtual-array
+descriptors.  Each virtual array can be accessed directly using the JPEG
+memory manager's access_virt_barray method (see Memory management, below,
+and also read structure.doc's discussion of virtual array handling).  Or,
+for simple transcoding to a different JPEG file format, the array list can
+just be handed directly to jpeg_write_coefficients().
+
+Each block in the block arrays contains quantized coefficient values in
+normal array order (not JPEG zigzag order).  The block arrays contain only
+DCT blocks containing real data; any entirely-dummy blocks added to fill out
+interleaved MCUs at the right or bottom edges of the image are discarded
+during reading and are not stored in the block arrays.  (The size of each
+block array can be determined from the width_in_blocks and height_in_blocks
+fields of the component's comp_info entry.)  This is also the data format
+expected by jpeg_write_coefficients().
+
+When you are done using the virtual arrays, call jpeg_finish_decompress()
+to release the array storage and return the decompression object to an idle
+state; or just call jpeg_destroy() if you don't need to reuse the object.
+
+If you use a suspending data source, jpeg_read_coefficients() will return
+NULL if it is forced to suspend; a non-NULL return value indicates successful
+completion.  You need not test for a NULL return value when using a
+non-suspending data source.
+
+It is also possible to call jpeg_read_coefficients() to obtain access to the
+decoder's coefficient arrays during a normal decode cycle in buffered-image
+mode.  This frammish might be useful for progressively displaying an incoming
+image and then re-encoding it without loss.  To do this, decode in buffered-
+image mode as discussed previously, then call jpeg_read_coefficients() after
+the last jpeg_finish_output() call.  The arrays will be available for your use
+until you call jpeg_finish_decompress().
+
+
+To write the contents of a JPEG file as DCT coefficients, you must provide
+the DCT coefficients stored in virtual block arrays.  You can either pass
+block arrays read from an input JPEG file by jpeg_read_coefficients(), or
+allocate virtual arrays from the JPEG compression object and fill them
+yourself.  In either case, jpeg_write_coefficients() is substituted for
+jpeg_start_compress() and jpeg_write_scanlines().  Thus the sequence is
+  * Create compression object
+  * Set all compression parameters as necessary
+  * Request virtual arrays if needed
+  * jpeg_write_coefficients()
+  * jpeg_finish_compress()
+  * Destroy or re-use compression object
+jpeg_write_coefficients() is passed a pointer to an array of virtual block
+array descriptors; the number of arrays is equal to cinfo.num_components.
+
+The virtual arrays need only have been requested, not realized, before
+jpeg_write_coefficients() is called.  A side-effect of
+jpeg_write_coefficients() is to realize any virtual arrays that have been
+requested from the compression object's memory manager.  Thus, when obtaining
+the virtual arrays from the compression object, you should fill the arrays
+after calling jpeg_write_coefficients().  The data is actually written out
+when you call jpeg_finish_compress(); jpeg_write_coefficients() only writes
+the file header.
+
+When writing raw DCT coefficients, it is crucial that the JPEG quantization
+tables and sampling factors match the way the data was encoded, or the
+resulting file will be invalid.  For transcoding from an existing JPEG file,
+we recommend using jpeg_copy_critical_parameters().  This routine initializes
+all the compression parameters to default values (like jpeg_set_defaults()),
+then copies the critical information from a source decompression object.
+The decompression object should have just been used to read the entire
+JPEG input file --- that is, it should be awaiting jpeg_finish_decompress().
+
+jpeg_write_coefficients() marks all tables stored in the compression object
+as needing to be written to the output file (thus, it acts like
+jpeg_start_compress(cinfo, TRUE)).  This is for safety's sake, to avoid
+emitting abbreviated JPEG files by accident.  If you really want to emit an
+abbreviated JPEG file, call jpeg_suppress_tables(), or set the tables'
+individual sent_table flags, between calling jpeg_write_coefficients() and
+jpeg_finish_compress().
+
+
+Progress monitoring
+-------------------
+
+Some applications may need to regain control from the JPEG library every so
+often.  The typical use of this feature is to produce a percent-done bar or
+other progress display.  (For a simple example, see cjpeg.c or djpeg.c.)
+Although you do get control back frequently during the data-transferring pass
+(the jpeg_read_scanlines or jpeg_write_scanlines loop), any additional passes
+will occur inside jpeg_finish_compress or jpeg_start_decompress; those
+routines may take a long time to execute, and you don't get control back
+until they are done.
+
+You can define a progress-monitor routine which will be called periodically
+by the library.  No guarantees are made about how often this call will occur,
+so we don't recommend you use it for mouse tracking or anything like that.
+At present, a call will occur once per MCU row, scanline, or sample row
+group, whichever unit is convenient for the current processing mode; so the
+wider the image, the longer the time between calls.  During the data
+transferring pass, only one call occurs per call of jpeg_read_scanlines or
+jpeg_write_scanlines, so don't pass a large number of scanlines at once if
+you want fine resolution in the progress count.  (If you really need to use
+the callback mechanism for time-critical tasks like mouse tracking, you could
+insert additional calls inside some of the library's inner loops.)
+
+To establish a progress-monitor callback, create a struct jpeg_progress_mgr,
+fill in its progress_monitor field with a pointer to your callback routine,
+and set cinfo->progress to point to the struct.  The callback will be called
+whenever cinfo->progress is non-NULL.  (This pointer is set to NULL by
+jpeg_create_compress or jpeg_create_decompress; the library will not change
+it thereafter.  So if you allocate dynamic storage for the progress struct,
+make sure it will live as long as the JPEG object does.  Allocating from the
+JPEG memory manager with lifetime JPOOL_PERMANENT will work nicely.)  You
+can use the same callback routine for both compression and decompression.
+
+The jpeg_progress_mgr struct contains four fields which are set by the library:
+	long pass_counter;	/* work units completed in this pass */
+	long pass_limit;	/* total number of work units in this pass */
+	int completed_passes;	/* passes completed so far */
+	int total_passes;	/* total number of passes expected */
+During any one pass, pass_counter increases from 0 up to (not including)
+pass_limit; the step size is usually but not necessarily 1.  The pass_limit
+value may change from one pass to another.  The expected total number of
+passes is in total_passes, and the number of passes already completed is in
+completed_passes.  Thus the fraction of work completed may be estimated as
+		completed_passes + (pass_counter/pass_limit)
+		--------------------------------------------
+				total_passes
+ignoring the fact that the passes may not be equal amounts of work.
+
+When decompressing, pass_limit can even change within a pass, because it
+depends on the number of scans in the JPEG file, which isn't always known in
+advance.  The computed fraction-of-work-done may jump suddenly (if the library
+discovers it has overestimated the number of scans) or even decrease (in the
+opposite case).  It is not wise to put great faith in the work estimate.
+
+When using the decompressor's buffered-image mode, the progress monitor work
+estimate is likely to be completely unhelpful, because the library has no way
+to know how many output passes will be demanded of it.  Currently, the library
+sets total_passes based on the assumption that there will be one more output
+pass if the input file end hasn't yet been read (jpeg_input_complete() isn't
+TRUE), but no more output passes if the file end has been reached when the
+output pass is started.  This means that total_passes will rise as additional
+output passes are requested.  If you have a way of determining the input file
+size, estimating progress based on the fraction of the file that's been read
+will probably be more useful than using the library's value.
+
+
+Memory management
+-----------------
+
+This section covers some key facts about the JPEG library's built-in memory
+manager.  For more info, please read structure.doc's section about the memory
+manager, and consult the source code if necessary.
+
+All memory and temporary file allocation within the library is done via the
+memory manager.  If necessary, you can replace the "back end" of the memory
+manager to control allocation yourself (for example, if you don't want the
+library to use malloc() and free() for some reason).
+
+Some data is allocated "permanently" and will not be freed until the JPEG
+object is destroyed.  Most data is allocated "per image" and is freed by
+jpeg_finish_compress, jpeg_finish_decompress, or jpeg_abort.  You can call the
+memory manager yourself to allocate structures that will automatically be
+freed at these times.  Typical code for this is
+  ptr = (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, size);
+Use JPOOL_PERMANENT to get storage that lasts as long as the JPEG object.
+Use alloc_large instead of alloc_small for anything bigger than a few Kbytes.
+There are also alloc_sarray and alloc_barray routines that automatically
+build 2-D sample or block arrays.
+
+The library's minimum space requirements to process an image depend on the
+image's width, but not on its height, because the library ordinarily works
+with "strip" buffers that are as wide as the image but just a few rows high.
+Some operating modes (eg, two-pass color quantization) require full-image
+buffers.  Such buffers are treated as "virtual arrays": only the current strip
+need be in memory, and the rest can be swapped out to a temporary file.
+
+If you use the simplest memory manager back end (jmemnobs.c), then no
+temporary files are used; virtual arrays are simply malloc()'d.  Images bigger
+than memory can be processed only if your system supports virtual memory.
+The other memory manager back ends support temporary files of various flavors
+and thus work in machines without virtual memory.  They may also be useful on
+Unix machines if you need to process images that exceed available swap space.
+
+When using temporary files, the library will make the in-memory buffers for
+its virtual arrays just big enough to stay within a "maximum memory" setting.
+Your application can set this limit by setting cinfo->mem->max_memory_to_use
+after creating the JPEG object.  (Of course, there is still a minimum size for
+the buffers, so the max-memory setting is effective only if it is bigger than
+the minimum space needed.)  If you allocate any large structures yourself, you
+must allocate them before jpeg_start_compress() or jpeg_start_decompress() in
+order to have them counted against the max memory limit.  Also keep in mind
+that space allocated with alloc_small() is ignored, on the assumption that
+it's too small to be worth worrying about; so a reasonable safety margin
+should be left when setting max_memory_to_use.
+
+If you use the jmemname.c or jmemdos.c memory manager back end, it is
+important to clean up the JPEG object properly to ensure that the temporary
+files get deleted.  (This is especially crucial with jmemdos.c, where the
+"temporary files" may be extended-memory segments; if they are not freed,
+DOS will require a reboot to recover the memory.)  Thus, with these memory
+managers, it's a good idea to provide a signal handler that will trap any
+early exit from your program.  The handler should call either jpeg_abort()
+or jpeg_destroy() for any active JPEG objects.  A handler is not needed with
+jmemnobs.c, and shouldn't be necessary with jmemansi.c or jmemmac.c either,
+since the C library is supposed to take care of deleting files made with
+tmpfile().
+
+
+Memory usage
+------------
+
+Working memory requirements while performing compression or decompression
+depend on image dimensions, image characteristics (such as colorspace and
+JPEG process), and operating mode (application-selected options).
+
+As of v6b, the decompressor requires:
+ 1. About 24K in more-or-less-fixed-size data.  This varies a bit depending
+    on operating mode and image characteristics (particularly color vs.
+    grayscale), but it doesn't depend on image dimensions.
+ 2. Strip buffers (of size proportional to the image width) for IDCT and
+    upsampling results.  The worst case for commonly used sampling factors
+    is about 34 bytes * width in pixels for a color image.  A grayscale image
+    only needs about 8 bytes per pixel column.
+ 3. A full-image DCT coefficient buffer is needed to decode a multi-scan JPEG
+    file (including progressive JPEGs), or whenever you select buffered-image
+    mode.  This takes 2 bytes/coefficient.  At typical 2x2 sampling, that's
+    3 bytes per pixel for a color image.  Worst case (1x1 sampling) requires
+    6 bytes/pixel.  For grayscale, figure 2 bytes/pixel.
+ 4. To perform 2-pass color quantization, the decompressor also needs a
+    128K color lookup table and a full-image pixel buffer (3 bytes/pixel).
+This does not count any memory allocated by the application, such as a
+buffer to hold the final output image.
+
+The above figures are valid for 8-bit JPEG data precision and a machine with
+32-bit ints.  For 12-bit JPEG data, double the size of the strip buffers and
+quantization pixel buffer.  The "fixed-size" data will be somewhat smaller
+with 16-bit ints, larger with 64-bit ints.  Also, CMYK or other unusual
+color spaces will require different amounts of space.
+
+The full-image coefficient and pixel buffers, if needed at all, do not
+have to be fully RAM resident; you can have the library use temporary
+files instead when the total memory usage would exceed a limit you set.
+(But if your OS supports virtual memory, it's probably better to just use
+jmemnobs and let the OS do the swapping.)
+
+The compressor's memory requirements are similar, except that it has no need
+for color quantization.  Also, it needs a full-image DCT coefficient buffer
+if Huffman-table optimization is asked for, even if progressive mode is not
+requested.
+
+If you need more detailed information about memory usage in a particular
+situation, you can enable the MEM_STATS code in jmemmgr.c.
+
+
+Library compile-time options
+----------------------------
+
+A number of compile-time options are available by modifying jmorecfg.h.
+
+The JPEG standard provides for both the baseline 8-bit DCT process and
+a 12-bit DCT process.  The IJG code supports 12-bit lossy JPEG if you define
+BITS_IN_JSAMPLE as 12 rather than 8.  Note that this causes JSAMPLE to be
+larger than a char, so it affects the surrounding application's image data.
+The sample applications cjpeg and djpeg can support 12-bit mode only for PPM
+and GIF file formats; you must disable the other file formats to compile a
+12-bit cjpeg or djpeg.  (install.doc has more information about that.)
+At present, a 12-bit library can handle *only* 12-bit images, not both
+precisions.  (If you need to include both 8- and 12-bit libraries in a single
+application, you could probably do it by defining NEED_SHORT_EXTERNAL_NAMES
+for just one of the copies.  You'd have to access the 8-bit and 12-bit copies
+from separate application source files.  This is untested ... if you try it,
+we'd like to hear whether it works!)
+
+Note that a 12-bit library always compresses in Huffman optimization mode,
+in order to generate valid Huffman tables.  This is necessary because our
+default Huffman tables only cover 8-bit data.  If you need to output 12-bit
+files in one pass, you'll have to supply suitable default Huffman tables.
+You may also want to supply your own DCT quantization tables; the existing
+quality-scaling code has been developed for 8-bit use, and probably doesn't
+generate especially good tables for 12-bit.
+
+The maximum number of components (color channels) in the image is determined
+by MAX_COMPONENTS.  The JPEG standard allows up to 255 components, but we
+expect that few applications will need more than four or so.
+
+On machines with unusual data type sizes, you may be able to improve
+performance or reduce memory space by tweaking the various typedefs in
+jmorecfg.h.  In particular, on some RISC CPUs, access to arrays of "short"s
+is quite slow; consider trading memory for speed by making JCOEF, INT16, and
+UINT16 be "int" or "unsigned int".  UINT8 is also a candidate to become int.
+You probably don't want to make JSAMPLE be int unless you have lots of memory
+to burn.
+
+You can reduce the size of the library by compiling out various optional
+functions.  To do this, undefine xxx_SUPPORTED symbols as necessary.
+
+You can also save a few K by not having text error messages in the library;
+the standard error message table occupies about 5Kb.  This is particularly
+reasonable for embedded applications where there's no good way to display 
+a message anyway.  To do this, remove the creation of the message table
+(jpeg_std_message_table[]) from jerror.c, and alter format_message to do
+something reasonable without it.  You could output the numeric value of the
+message code number, for example.  If you do this, you can also save a couple
+more K by modifying the TRACEMSn() macros in jerror.h to expand to nothing;
+you don't need trace capability anyway, right?
+
+
+Portability considerations
+--------------------------
+
+The JPEG library has been written to be extremely portable; the sample
+applications cjpeg and djpeg are slightly less so.  This section summarizes
+the design goals in this area.  (If you encounter any bugs that cause the
+library to be less portable than is claimed here, we'd appreciate hearing
+about them.)
+
+The code works fine on ANSI C, C++, and pre-ANSI C compilers, using any of
+the popular system include file setups, and some not-so-popular ones too.
+See install.doc for configuration procedures.
+
+The code is not dependent on the exact sizes of the C data types.  As
+distributed, we make the assumptions that
+	char	is at least 8 bits wide
+	short	is at least 16 bits wide
+	int	is at least 16 bits wide
+	long	is at least 32 bits wide
+(These are the minimum requirements of the ANSI C standard.)  Wider types will
+work fine, although memory may be used inefficiently if char is much larger
+than 8 bits or short is much bigger than 16 bits.  The code should work
+equally well with 16- or 32-bit ints.
+
+In a system where these assumptions are not met, you may be able to make the
+code work by modifying the typedefs in jmorecfg.h.  However, you will probably
+have difficulty if int is less than 16 bits wide, since references to plain
+int abound in the code.
+
+char can be either signed or unsigned, although the code runs faster if an
+unsigned char type is available.  If char is wider than 8 bits, you will need
+to redefine JOCTET and/or provide custom data source/destination managers so
+that JOCTET represents exactly 8 bits of data on external storage.
+
+The JPEG library proper does not assume ASCII representation of characters.
+But some of the image file I/O modules in cjpeg/djpeg do have ASCII
+dependencies in file-header manipulation; so does cjpeg's select_file_type()
+routine.
+
+The JPEG library does not rely heavily on the C library.  In particular, C
+stdio is used only by the data source/destination modules and the error
+handler, all of which are application-replaceable.  (cjpeg/djpeg are more
+heavily dependent on stdio.)  malloc and free are called only from the memory
+manager "back end" module, so you can use a different memory allocator by
+replacing that one file.
+
+The code generally assumes that C names must be unique in the first 15
+characters.  However, global function names can be made unique in the
+first 6 characters by defining NEED_SHORT_EXTERNAL_NAMES.
+
+More info about porting the code may be gleaned by reading jconfig.doc,
+jmorecfg.h, and jinclude.h.
+
+
+Notes for MS-DOS implementors
+-----------------------------
+
+The IJG code is designed to work efficiently in 80x86 "small" or "medium"
+memory models (i.e., data pointers are 16 bits unless explicitly declared
+"far"; code pointers can be either size).  You may be able to use small
+model to compile cjpeg or djpeg by itself, but you will probably have to use
+medium model for any larger application.  This won't make much difference in
+performance.  You *will* take a noticeable performance hit if you use a
+large-data memory model (perhaps 10%-25%), and you should avoid "huge" model
+if at all possible.
+
+The JPEG library typically needs 2Kb-3Kb of stack space.  It will also
+malloc about 20K-30K of near heap space while executing (and lots of far
+heap, but that doesn't count in this calculation).  This figure will vary
+depending on selected operating mode, and to a lesser extent on image size.
+There is also about 5Kb-6Kb of constant data which will be allocated in the
+near data segment (about 4Kb of this is the error message table).
+Thus you have perhaps 20K available for other modules' static data and near
+heap space before you need to go to a larger memory model.  The C library's
+static data will account for several K of this, but that still leaves a good
+deal for your needs.  (If you are tight on space, you could reduce the sizes
+of the I/O buffers allocated by jdatasrc.c and jdatadst.c, say from 4K to
+1K.  Another possibility is to move the error message table to far memory;
+this should be doable with only localized hacking on jerror.c.)
+
+About 2K of the near heap space is "permanent" memory that will not be
+released until you destroy the JPEG object.  This is only an issue if you
+save a JPEG object between compression or decompression operations.
+
+Far data space may also be a tight resource when you are dealing with large
+images.  The most memory-intensive case is decompression with two-pass color
+quantization, or single-pass quantization to an externally supplied color
+map.  This requires a 128Kb color lookup table plus strip buffers amounting
+to about 40 bytes per column for typical sampling ratios (eg, about 25600
+bytes for a 640-pixel-wide image).  You may not be able to process wide
+images if you have large data structures of your own.
+
+Of course, all of these concerns vanish if you use a 32-bit flat-memory-model
+compiler, such as DJGPP or Watcom C.  We highly recommend flat model if you
+can use it; the JPEG library is significantly faster in flat model.
diff --git a/libjpegtwrp/ltconfig b/libjpegtwrp/ltconfig
new file mode 100755
index 000000000..2347e6943
--- /dev/null
+++ b/libjpegtwrp/ltconfig
@@ -0,0 +1,1512 @@
+#! /bin/sh
+
+# ltconfig - Create a system-specific libtool.
+# Copyright (C) 1996-1998 Free Software Foundation, Inc.
+# Gordon Matzigkeit <gord@gnu.ai.mit.edu>, 1996
+#
+# This file is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+# General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+#
+# As a special exception to the GNU General Public License, if you
+# distribute this file as part of a program that contains a
+# configuration script generated by Autoconf, you may include it under
+# the same distribution terms that you use for the rest of that program.
+
+# A lot of this script is taken from autoconf-2.10.
+
+# The HP-UX ksh and POSIX shell print the target directory to stdout
+# if CDPATH is set.
+if test "${CDPATH+set}" = set; then CDPATH=; export CDPATH; fi
+
+echo=echo
+if test "X`($echo '\t') 2>/dev/null`" = 'X\t'; then :
+else
+  # The Solaris and AIX default echo program unquotes backslashes.
+  # This makes it impossible to quote backslashes using
+  #   echo "$something" | sed 's/\\/\\\\/g'
+  # So, we emulate echo with printf '%s\n'
+  echo="printf %s\\n"
+  if test "X`($echo '\t') 2>/dev/null`" = 'X\t'; then :
+  else
+    # Oops.  We have no working printf.  Try to find a not-so-buggy echo.
+    echo=echo
+    IFS="${IFS= 	}"; save_ifs="$IFS"; IFS="${IFS}:"
+    for dir in $PATH /usr/ucb; do
+      if test -f $dir/echo && test "X`$dir/echo '\t'`" = 'X\t'; then
+        echo="$dir/echo"
+        break
+      fi
+    done
+    IFS="$save_ifs"
+  fi
+fi
+
+# Sed substitution that helps us do robust quoting.  It backslashifies
+# metacharacters that are still active within double-quoted strings.
+Xsed='sed -e s/^X//'
+sed_quote_subst='s/\([\\"\\`$\\\\]\)/\\\1/g'
+
+# Same as above, but do not quote variable references.
+double_quote_subst='s/\([\\"\\`\\\\]\)/\\\1/g'
+
+# The name of this program.
+progname=`$echo "X$0" | $Xsed -e 's%^.*/%%'`
+
+# Constants:
+PROGRAM=ltconfig
+PACKAGE=libtool
+VERSION=1.2
+ac_compile='${CC-cc} -c $CFLAGS $CPPFLAGS conftest.c 1>&5'
+ac_link='${CC-cc} -o conftest $CFLAGS $CPPFLAGS $LDFLAGS conftest.c $LIBS 1>&5'
+rm="rm -f"
+
+help="Try \`$progname --help' for more information."
+
+# Global variables:
+can_build_shared=yes
+enable_shared=yes
+# All known linkers require a `.a' archive for static linking.
+enable_static=yes
+ltmain=
+silent=
+srcdir=
+ac_config_guess=
+ac_config_sub=
+host=
+nonopt=
+verify_host=yes
+with_gcc=no
+with_gnu_ld=no
+
+old_AR="$AR"
+old_CC="$CC"
+old_CFLAGS="$CFLAGS"
+old_CPPFLAGS="$CPPFLAGS"
+old_LD="$LD"
+old_LN_S="$LN_S"
+old_NM="$NM"
+old_RANLIB="$RANLIB"
+
+# Parse the command line options.
+args=
+prev=
+for option
+do
+  case "$option" in
+  -*=*) optarg=`echo "$option" | sed 's/[-_a-zA-Z0-9]*=//'` ;;
+  *) optarg= ;;
+  esac
+
+  # If the previous option needs an argument, assign it.
+  if test -n "$prev"; then
+    eval "$prev=\$option"
+    prev=
+    continue
+  fi
+
+  case "$option" in
+  --help) cat <<EOM
+Usage: $progname [OPTION]... LTMAIN [HOST]
+
+Generate a system-specific libtool script.
+
+    --disable-shared       do not build shared libraries
+    --disable-static       do not build static libraries
+    --help                 display this help and exit
+    --no-verify            do not verify that HOST is a valid host type
+    --quiet                same as \`--silent'
+    --silent               do not print informational messages
+    --srcdir=DIR           find \`config.guess' in DIR
+    --version              output version information and exit
+    --with-gcc             assume that the GNU C compiler will be used
+    --with-gnu-ld          assume that the C compiler uses the GNU linker
+
+LTMAIN is the \`ltmain.sh' shell script fragment that provides basic libtool
+functionality.
+
+HOST is the canonical host system name [default=guessed].
+EOM
+  exit 0
+  ;;
+
+  --disable-shared) enable_shared=no ;;
+
+  --disable-static) enable_static=no ;;
+
+  --quiet | --silent) silent=yes ;;
+
+  --srcdir) prev=srcdir ;;
+  --srcdir=*) srcdir="$optarg" ;;
+
+  --no-verify) verify_host=no ;;
+
+  --version) echo "$PROGRAM (GNU $PACKAGE) $VERSION"; exit 0 ;;
+
+  --with-gcc) with_gcc=yes ;;
+  --with-gnu-ld) with_gnu_ld=yes ;;
+
+  -*)
+    echo "$progname: unrecognized option \`$option'" 1>&2
+    echo "$help" 1>&2
+    exit 1
+    ;;
+
+  *)
+    if test -z "$ltmain"; then
+      ltmain="$option"
+    elif test -z "$host"; then
+# This generates an unnecessary warning for sparc-sun-solaris4.1.3_U1
+#      if test -n "`echo $option| sed 's/[-a-z0-9.]//g'`"; then
+#        echo "$progname: warning \`$option' is not a valid host type" 1>&2
+#      fi
+      host="$option"
+    else
+      echo "$progname: too many arguments" 1>&2
+      echo "$help" 1>&2
+      exit 1
+    fi ;;
+  esac
+done
+
+if test -z "$ltmain"; then
+  echo "$progname: you must specify a LTMAIN file" 1>&2
+  echo "$help" 1>&2
+  exit 1
+fi
+
+if test -f "$ltmain"; then :
+else
+  echo "$progname: \`$ltmain' does not exist" 1>&2
+  echo "$help" 1>&2
+  exit 1
+fi
+
+# Quote any args containing shell metacharacters.
+ltconfig_args=
+for arg
+do
+  case "$arg" in
+  *" "*|*"	"*|*[\[\]\~\#\$\^\&\*\(\)\{\}\\\|\;\<\>\?]*)
+  ltconfig_args="$ltconfig_args '$arg'" ;;
+  *) ltconfig_args="$ltconfig_args $arg" ;;
+  esac
+done
+
+# A relevant subset of AC_INIT.
+
+# File descriptor usage:
+# 0 standard input
+# 1 file creation
+# 2 errors and warnings
+# 3 some systems may open it to /dev/tty
+# 4 used on the Kubota Titan
+# 5 compiler messages saved in config.log
+# 6 checking for... messages and results
+if test "$silent" = yes; then
+  exec 6>/dev/null
+else
+  exec 6>&1
+fi
+exec 5>>./config.log
+
+# NLS nuisances.
+# Only set LANG and LC_ALL to C if already set.
+# These must not be set unconditionally because not all systems understand
+# e.g. LANG=C (notably SCO).
+if test "${LC_ALL+set}" = set; then LC_ALL=C; export LC_ALL; fi
+if test "${LANG+set}"   = set; then LANG=C;   export LANG;   fi
+
+if (echo "testing\c"; echo 1,2,3) | grep c >/dev/null; then
+  # Stardent Vistra SVR4 grep lacks -e, says ghazi@caip.rutgers.edu.
+  if (echo -n testing; echo 1,2,3) | sed s/-n/xn/ | grep xn >/dev/null; then
+    ac_n= ac_c='
+' ac_t='	'
+  else
+    ac_n=-n ac_c= ac_t=
+  fi
+else
+  ac_n= ac_c='\c' ac_t=
+fi
+
+if test -z "$srcdir"; then
+  # Assume the source directory is the same one as the path to ltmain.sh.
+  srcdir=`$echo "$ltmain" | $Xsed -e 's%/[^/]*$%%'`
+  test "$srcdir" = "$ltmain" && srcdir=.
+fi
+
+trap "$rm conftest*; exit 1" 1 2 15
+if test "$verify_host" = yes; then
+  # Check for config.guess and config.sub.
+  ac_aux_dir=
+  for ac_dir in $srcdir $srcdir/.. $srcdir/../..; do
+    if test -f $ac_dir/config.guess; then
+      ac_aux_dir=$ac_dir
+      break
+    fi
+  done
+  if test -z "$ac_aux_dir"; then
+    echo "$progname: cannot find config.guess in $srcdir $srcdir/.. $srcdir/../.." 1>&2
+    echo "$help" 1>&2
+    exit 1
+  fi
+  ac_config_guess=$ac_aux_dir/config.guess
+  ac_config_sub=$ac_aux_dir/config.sub
+
+  # Make sure we can run config.sub.
+  if $ac_config_sub sun4 >/dev/null 2>&1; then :
+  else
+    echo "$progname: cannot run $ac_config_sub" 1>&2
+    echo "$help" 1>&2
+    exit 1
+  fi
+
+  echo $ac_n "checking host system type""... $ac_c" 1>&6
+
+  host_alias=$host
+  case "$host_alias" in
+  "")
+    if host_alias=`$ac_config_guess`; then :
+    else
+      echo "$progname: cannot guess host type; you must specify one" 1>&2
+      echo "$help" 1>&2
+      exit 1
+    fi ;;
+  esac
+  host=`$ac_config_sub $host_alias`
+  echo "$ac_t$host" 1>&6
+
+  # Make sure the host verified.
+  test -z "$host" && exit 1
+
+elif test -z "$host"; then
+  echo "$progname: you must specify a host type if you use \`--no-verify'" 1>&2
+  echo "$help" 1>&2
+  exit 1
+else
+  host_alias=$host
+fi
+
+# Transform linux* to *-*-linux-gnu*, to support old configure scripts.
+case "$host_os" in
+linux-gnu*) ;;
+linux*) host=`echo $host | sed 's/^\(.*-.*-linux\)\(.*\)$/\1-gnu\2/'`
+esac
+
+host_cpu=`echo $host | sed 's/^\([^-]*\)-\([^-]*\)-\(.*\)$/\1/'`
+host_vendor=`echo $host | sed 's/^\([^-]*\)-\([^-]*\)-\(.*\)$/\2/'`
+host_os=`echo $host | sed 's/^\([^-]*\)-\([^-]*\)-\(.*\)$/\3/'`
+
+case "$host_os" in
+aix3*)
+  # AIX sometimes has problems with the GCC collect2 program.  For some
+  # reason, if we set the COLLECT_NAMES environment variable, the problems
+  # vanish in a puff of smoke.
+  if test "${COLLECT_NAMES+set}" != set; then
+    COLLECT_NAMES=
+    export COLLECT_NAMES
+  fi
+  ;;
+esac
+
+# Determine commands to create old-style static archives.
+old_archive_cmds='$AR cru $oldlib$oldobjs'
+old_postinstall_cmds='chmod 644 $oldlib'
+old_postuninstall_cmds=
+
+# Set a sane default for `AR'.
+test -z "$AR" && AR=ar
+
+# If RANLIB is not set, then run the test.
+if test "${RANLIB+set}" != "set"; then
+  result=no
+
+  echo $ac_n "checking for ranlib... $ac_c" 1>&6
+  IFS="${IFS= 	}"; save_ifs="$IFS"; IFS="${IFS}:"
+  for dir in $PATH; do
+    test -z "$dir" && dir=.
+    if test -f $dir/ranlib; then
+      RANLIB="ranlib"
+      result="ranlib"
+      break
+    fi
+  done
+  IFS="$save_ifs"
+
+  echo "$ac_t$result" 1>&6
+fi
+
+if test -n "$RANLIB"; then
+  old_archive_cmds="$old_archive_cmds;\$RANLIB \$oldlib"
+  old_postinstall_cmds="\$RANLIB \$oldlib;$old_postinstall_cmds"
+fi
+
+# Check to see if we are using GCC.
+if test "$with_gcc" != yes || test -z "$CC"; then
+  # If CC is not set, then try to find GCC or a usable CC.
+  if test -z "$CC"; then
+    echo $ac_n "checking for gcc... $ac_c" 1>&6
+    IFS="${IFS= 	}"; save_ifs="$IFS"; IFS="${IFS}:"
+    for dir in $PATH; do
+      IFS="$save_ifs"
+      test -z "$dir" && dir=.
+      if test -f $dir/gcc; then
+	CC="gcc"
+	break
+      fi
+    done
+    IFS="$save_ifs"
+
+    if test -n "$CC"; then
+      echo "$ac_t$CC" 1>&6
+    else
+      echo "$ac_t"no 1>&6
+    fi
+  fi
+
+  # Not "gcc", so try "cc", rejecting "/usr/ucb/cc".
+  if test -z "$CC"; then
+    echo $ac_n "checking for cc... $ac_c" 1>&6
+    IFS="${IFS= 	}"; save_ifs="$IFS"; IFS="${IFS}:"
+    cc_rejected=no
+    for dir in $PATH; do
+      test -z "$dir" && dir=.
+      if test -f $dir/cc; then
+	if test "$dir/cc" = "/usr/ucb/cc"; then
+	  cc_rejected=yes
+	  continue
+	fi
+	CC="cc"
+	break
+      fi
+    done
+    IFS="$save_ifs"
+    if test $cc_rejected = yes; then
+      # We found a bogon in the path, so make sure we never use it.
+      set dummy $CC
+      shift
+      if test $# -gt 0; then
+	# We chose a different compiler from the bogus one.
+	# However, it has the same name, so the bogon will be chosen
+	# first if we set CC to just the name; use the full file name.
+	shift
+	set dummy "$dir/cc" "$@"
+	shift
+	CC="$@"
+      fi
+    fi
+
+    if test -n "$CC"; then
+      echo "$ac_t$CC" 1>&6
+    else
+      echo "$ac_t"no 1>&6
+    fi
+
+    if test -z "$CC"; then
+      echo "$progname: error: no acceptable cc found in \$PATH" 1>&2
+      exit 1
+    fi
+  fi
+
+  # Now see if the compiler is really GCC.
+  with_gcc=no
+  echo $ac_n "checking whether we are using GNU C... $ac_c" 1>&6
+  echo "$progname:424: checking whether we are using GNU C" >&5
+
+  $rm conftest.c
+  cat > conftest.c <<EOF
+#ifdef __GNUC__
+  yes;
+#endif
+EOF
+  if { ac_try='${CC-cc} -E conftest.c'; { (eval echo $progname:432: \"$ac_try\") 1>&5; (eval $ac_try) 2>&5; }; } | egrep yes >/dev/null 2>&1; then
+    with_gcc=yes
+  fi
+  $rm conftest.c
+  echo "$ac_t$with_gcc" 1>&6
+fi
+
+# Allow CC to be a program name with arguments.
+set dummy $CC
+compiler="$2"
+
+echo $ac_n "checking for $compiler option to produce PIC... $ac_c" 1>&6
+pic_flag=
+special_shlib_compile_flags=
+wl=
+link_static_flag=
+no_builtin_flag=
+
+if test "$with_gcc" = yes; then
+  wl='-Wl,'
+  link_static_flag='-static'
+  no_builtin_flag=' -fno-builtin'
+
+  case "$host_os" in
+  aix3* | aix4* | irix5* | irix6* | osf3* | osf4*)
+    # PIC is the default for these OSes.
+    ;;
+  os2*)
+    # We can build DLLs from non-PIC.
+    ;;
+  amigaos*)
+    # FIXME: we need at least 68020 code to build shared libraries, but
+    # adding the `-m68020' flag to GCC prevents building anything better,
+    # like `-m68040'.
+    pic_flag='-m68020 -resident32 -malways-restore-a4'
+    ;;
+  *)
+    pic_flag='-fPIC'
+    ;;
+  esac
+else
+  # PORTME Check for PIC flags for the system compiler.
+  case "$host_os" in
+  aix3* | aix4*)
+    # All AIX code is PIC.
+    link_static_flag='-bnso -bI:/lib/syscalls.exp'
+    ;;
+
+  hpux9* | hpux10*)
+    # Is there a better link_static_flag that works with the bundled CC?
+    wl='-Wl,'
+    link_static_flag="${wl}-a ${wl}archive"
+    pic_flag='+Z'
+    ;;
+
+  irix5* | irix6*)
+    wl='-Wl,'
+    link_static_flag='-non_shared'
+    # PIC (with -KPIC) is the default.
+    ;;
+
+  os2*)
+    # We can build DLLs from non-PIC.
+    ;;
+
+  osf3* | osf4*)
+    # All OSF/1 code is PIC.
+    wl='-Wl,'
+    link_static_flag='-non_shared'
+    ;;
+
+  sco3.2v5*)
+    pic_flag='-Kpic'
+    link_static_flag='-dn'
+    special_shlib_compile_flags='-belf'
+    ;;
+
+  solaris2*)
+    pic_flag='-KPIC'
+    link_static_flag='-Bstatic'
+    wl='-Wl,'
+    ;;
+
+  sunos4*)
+    pic_flag='-PIC'
+    link_static_flag='-Bstatic'
+    wl='-Qoption ld '
+    ;;
+
+  sysv4.2uw2*)
+    pic_flag='-KPIC'
+    link_static_flag='-Bstatic'
+    wl='-Wl,'
+    ;;
+
+  uts4*)
+    pic_flag='-pic'
+    link_static_flag='-Bstatic'
+    ;;
+
+  *)
+    can_build_shared=no
+    ;;
+  esac
+fi
+
+if test -n "$pic_flag"; then
+  echo "$ac_t$pic_flag" 1>&6
+
+  # Check to make sure the pic_flag actually works.
+  echo $ac_n "checking if $compiler PIC flag $pic_flag works... $ac_c" 1>&6
+  $rm conftest*
+  echo > conftest.c
+  save_CFLAGS="$CFLAGS"
+  CFLAGS="$CFLAGS $pic_flag -DPIC"
+  echo "$progname:547: checking if $compiler PIC flag $pic_flag works" >&5
+  if { (eval echo $progname:548: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>conftest.err; } && test -s conftest.o; then
+    # Append any warnings to the config.log.
+    cat conftest.err 1>&5
+
+    # On HP-UX, both CC and GCC only warn that PIC is supported... then they
+    # create non-PIC objects.  So, if there were any warnings, we assume that
+    # PIC is not supported.
+    if test -s conftest.err; then
+      echo "$ac_t"no 1>&6
+      can_build_shared=no
+      pic_flag=
+    else
+      echo "$ac_t"yes 1>&6
+      pic_flag=" $pic_flag"
+    fi
+  else
+    # Append any errors to the config.log.
+    cat conftest.err 1>&5
+    can_build_shared=no
+    pic_flag=
+    echo "$ac_t"no 1>&6
+  fi
+  CFLAGS="$save_CFLAGS"
+  $rm conftest*
+else
+  echo "$ac_t"none 1>&6
+fi
+
+# Check for any special shared library compilation flags.
+if test -n "$special_shlib_compile_flags"; then
+  echo "$progname: warning: \`$CC' requires \`$special_shlib_compile_flags' to build shared libraries" 1>&2
+  if echo "$old_CC $old_CFLAGS " | egrep -e "[ 	]$special_shlib_compile_flags[ 	]" >/dev/null; then :
+  else
+    echo "$progname: add \`$special_shlib_compile_flags' to the CC or CFLAGS env variable and reconfigure" 1>&2
+    can_build_shared=no
+  fi
+fi
+
+echo $ac_n "checking if $compiler static flag $link_static_flag works... $ac_c" 1>&6
+$rm conftest*
+echo 'main(){return(0);}' > conftest.c
+save_LDFLAGS="$LDFLAGS"
+LDFLAGS="$LDFLAGS $link_static_flag"
+echo "$progname:591: checking if $compiler static flag $link_static_flag works" >&5
+if { (eval echo $progname:592: \"$ac_link\") 1>&5; (eval $ac_link) 2>&5; } && test -s conftest; then
+  echo "$ac_t$link_static_flag" 1>&6
+else
+  echo "$ac_t"none 1>&6
+  link_static_flag=
+fi
+LDFLAGS="$save_LDFLAGS"
+$rm conftest*
+
+if test -z "$LN_S"; then
+  # Check to see if we can use ln -s, or we need hard links.
+  echo $ac_n "checking whether ln -s works... $ac_c" 1>&6
+  $rm conftestdata
+  if ln -s X conftestdata 2>/dev/null; then
+    $rm conftestdata
+    LN_S="ln -s"
+  else
+    LN_S=ln
+  fi
+  if test "$LN_S" = "ln -s"; then
+    echo "$ac_t"yes 1>&6
+  else
+    echo "$ac_t"no 1>&6
+  fi
+fi
+
+# Make sure LD is an absolute path.
+if test -z "$LD"; then
+  ac_prog=ld
+  if test "$with_gcc" = yes; then
+    # Check if gcc -print-prog-name=ld gives a path.
+    echo $ac_n "checking for ld used by GCC... $ac_c" 1>&6
+    echo "$progname:624: checking for ld used by GCC" >&5
+    ac_prog=`($CC -print-prog-name=ld) 2>&5`
+    case "$ac_prog" in
+    # Accept absolute paths.
+    /* | [A-Za-z]:\\*)
+      test -z "$LD" && LD="$ac_prog"
+      ;;
+    "")
+      # If it fails, then pretend we are not using GCC.
+      ac_prog=ld
+      ;;
+    *)
+      # If it is relative, then search for the first ld in PATH.
+      with_gnu_ld=unknown
+      ;;
+    esac
+  elif test "$with_gnu_ld" = yes; then
+    echo $ac_n "checking for GNU ld... $ac_c" 1>&6
+    echo "$progname:642: checking for GNU ld" >&5
+  else
+    echo $ac_n "checking for non-GNU ld""... $ac_c" 1>&6
+    echo "$progname:645: checking for non-GNU ld" >&5
+  fi
+
+  if test -z "$LD"; then
+    IFS="${IFS= 	}"; ac_save_ifs="$IFS"; IFS="${IFS}:"
+    for ac_dir in $PATH; do
+      test -z "$ac_dir" && ac_dir=.
+      if test -f "$ac_dir/$ac_prog"; then
+	LD="$ac_dir/$ac_prog"
+	# Check to see if the program is GNU ld.  I'd rather use --version,
+	# but apparently some GNU ld's only accept -v.
+	# Break only if it was the GNU/non-GNU ld that we prefer.
+	if "$LD" -v 2>&1 < /dev/null | egrep '(GNU|with BFD)' > /dev/null; then
+	  test "$with_gnu_ld" != no && break
+	else
+	  test "$with_gnu_ld" != yes && break
+	fi
+      fi
+    done
+    IFS="$ac_save_ifs"
+  fi
+
+  if test -n "$LD"; then
+    echo "$ac_t$LD" 1>&6
+  else
+    echo "$ac_t"no 1>&6
+  fi
+
+  if test -z "$LD"; then
+    echo "$progname: error: no acceptable ld found in \$PATH" 1>&2
+    exit 1
+  fi
+fi
+
+# Check to see if it really is or is not GNU ld.
+echo $ac_n "checking if the linker ($LD) is GNU ld... $ac_c" 1>&6
+# I'd rather use --version here, but apparently some GNU ld's only accept -v.
+if $LD -v 2>&1 </dev/null | egrep '(GNU|with BFD)' 1>&5; then
+  with_gnu_ld=yes
+else
+  with_gnu_ld=no
+fi
+echo "$ac_t$with_gnu_ld" 1>&6
+
+# See if the linker supports building shared libraries.
+echo $ac_n "checking whether the linker ($LD) supports shared libraries... $ac_c" 1>&6
+
+allow_undefined_flag=
+no_undefined_flag=
+archive_cmds=
+old_archive_from_new_cmds=
+export_dynamic_flag_spec=
+hardcode_libdir_flag_spec=
+hardcode_libdir_separator=
+hardcode_direct=no
+hardcode_minus_L=no
+hardcode_shlibpath_var=unsupported
+runpath_var=
+
+case "$host_os" in
+amigaos* | sunos4*)
+  # On these operating systems, we should treat GNU ld like the system ld.
+  gnu_ld_acts_native=yes
+  ;;
+*)
+  gnu_ld_acts_native=no
+  ;;
+esac
+
+ld_shlibs=yes
+if test "$with_gnu_ld" = yes && test "$gnu_ld_acts_native" != yes; then
+
+  # See if GNU ld supports shared libraries.
+  if $LD --help 2>&1 | egrep ': supported targets:.* elf' > /dev/null; then
+    archive_cmds='$CC -shared ${wl}-soname $wl$soname -o $lib$libobjs'
+    runpath_var=LD_RUN_PATH
+    ld_shlibs=yes
+  else
+    ld_shlibs=no
+  fi
+
+  if test "$ld_shlibs" = yes; then
+    hardcode_libdir_flag_spec='${wl}--rpath ${wl}$libdir'
+    export_dynamic_flag_spec='${wl}--export-dynamic'
+  fi
+else
+  # PORTME fill in a description of your system's linker (not GNU ld)
+  case "$host_os" in
+  aix3*)
+    allow_undefined_flag=unsupported
+    archive_cmds='$NM$libobjs | $global_symbol_pipe | sed '\''s/.* //'\'' > $lib.exp;$LD -o $objdir/$soname$libobjs -bE:$lib.exp -T512 -H512 -bM:SRE;$AR cru $lib $objdir/$soname'
+    # Note: this linker hardcodes the directories in LIBPATH if there
+    # are no directories specified by -L.
+    hardcode_minus_L=yes
+    if test "$with_gcc" = yes && test -z "$link_static_flag"; then
+      # Neither direct hardcoding nor static linking is supported with a
+      # broken collect2.
+      hardcode_direct=unsupported
+    fi
+    ;;
+
+  aix4*)
+    allow_undefined_flag=unsupported
+    archive_cmds='$NM$libobjs | $global_symbol_pipe | sed '\''s/.* //'\'' > $lib.exp;$CC -o $objdir/$soname$libobjs ${wl}-bE:$lib.exp ${wl}-bM:SRE ${wl}-bnoentry;$AR cru $lib $objdir/$soname'
+    hardcode_direct=yes
+    hardcode_minus_L=yes
+    ;;
+
+  amigaos*)
+    archive_cmds='$rm $objdir/a2ixlibrary.data;$echo "#define NAME $libname" > $objdir/a2ixlibrary.data;$echo "#define LIBRARY_ID 1" >> $objdir/a2ixlibrary.data;$echo "#define VERSION $major" >> $objdir/a2ixlibrary.data;$echo "#define REVISION $revision" >> $objdir/a2ixlibrary.data;$AR cru $lib$libobjs;$RANLIB $lib;(cd $objdir && a2ixlibrary -32)'
+    hardcode_libdir_flag_spec='-L$libdir'
+    hardcode_minus_L=yes
+    ;;
+
+  # FreeBSD 2.2.[012] allows us to include c++rt0.o to get C++ constructor
+  # support.  Future versions do this automatically, but an explicit c++rt0.o
+  # does not break anything, and helps significantly (at the cost of a little
+  # extra space).
+  freebsd2.2*)
+    archive_cmds='$LD -Bshareable -o $lib$libobjs /usr/lib/c++rt0.o'
+    hardcode_libdir_flag_spec='-R$libdir'
+    hardcode_direct=yes
+    hardcode_minus_L=yes
+    hardcode_shlibpath_var=no
+    ;;
+
+  # Unfortunately, older versions of FreeBSD 2 do not have this feature.
+  freebsd2*)
+    archive_cmds='$LD -Bshareable -o $lib$libobjs'
+    hardcode_direct=yes
+    hardcode_minus_L=yes
+    hardcode_shlibpath_var=no
+    ;;
+
+  # FreeBSD 3, at last, uses gcc -shared to do shared libraries.
+  freebsd3*)
+    archive_cmds='$CC -shared -o $lib$libobjs'
+    hardcode_libdir_flag_spec='-R$libdir'
+    hardcode_direct=yes
+    hardcode_minus_L=yes
+    hardcode_shlibpath_var=no
+    ;;
+
+  hpux9*)
+    archive_cmds='$rm $objdir/$soname;$LD -b +s +b $install_libdir -o $objdir/$soname$libobjs;mv $objdir/$soname $lib'
+    hardcode_libdir_flag_spec='${wl}+b ${wl}$libdir'
+    hardcode_direct=yes
+    hardcode_minus_L=yes
+    export_dynamic_flag_spec='${wl}-E'
+    ;;
+
+  hpux10*)
+    archive_cmds='$LD -b +h $soname +s +b $install_libdir -o $lib$libobjs'
+    hardcode_libdir_flag_spec='${wl}+b ${wl}$libdir'
+    hardcode_direct=yes
+    hardcode_minus_L=yes
+    export_dynamic_flag_spec='${wl}-E'
+    ;;
+
+  irix5* | irix6*)
+    archive_cmds='$LD -shared -o $lib -soname $soname -set_version $verstring$libobjs'
+    hardcode_libdir_flag_spec='${wl}-rpath ${wl}$libdir'
+    ;;
+
+  netbsd*)
+    # Tested with NetBSD 1.2 ld
+    archive_cmds='$LD -Bshareable -o $lib$libobjs'
+    hardcode_libdir_flag_spec='-R$libdir'
+    hardcode_direct=yes
+    hardcode_shlibpath_var=no
+    ;;
+
+  openbsd*)
+    archive_cmds='$LD -Bshareable -o $lib$libobjs'
+    hardcode_libdir_flag_spec='-R$libdir'
+    hardcode_direct=yes
+    hardcode_shlibpath_var=no
+    ;;
+
+  os2*)
+    hardcode_libdir_flag_spec='-L$libdir'
+    hardcode_minus_L=yes
+    allow_undefined_flag=unsupported
+    archive_cmds='$echo "LIBRARY $libname INITINSTANCE" > $objdir/$libname.def;$echo "DESCRIPTION \"$libname\"" >> $objdir/$libname.def;$echo DATA >> $objdir/$libname.def;$echo " SINGLE NONSHARED" >> $objdir/$libname.def;$echo EXPORTS >> $objdir/$libname.def;emxexp$libobjs >> $objdir/$libname.def;$CC -Zdll -Zcrtdll -o $lib$libobjs $objdir/$libname.def'
+    old_archive_from_new_cmds='emximp -o $objdir/$libname.a $objdir/$libname.def'
+    ;;
+
+  osf3* | osf4*)
+    allow_undefined_flag=' -expect_unresolved \*'
+    archive_cmds='$LD -shared${allow_undefined_flag} -o $lib -soname $soname -set_version $verstring$libobjs$deplibs'
+    hardcode_libdir_flag_spec='${wl}-rpath ${wl}$libdir'
+    hardcode_libdir_separator=:
+    ;;
+
+  sco3.2v5*)
+    archive_cmds='$LD -G -o $lib$libobjs'
+    hardcode_direct=yes
+    ;;
+
+  solaris2*)
+    no_undefined_flag=' -z text'
+    archive_cmds='$LD -G${allow_undefined_flag} -h $soname -o $lib$libobjs'
+    hardcode_libdir_flag_spec='-R$libdir'
+    hardcode_shlibpath_var=no
+
+    # Solaris 2 before 2.5 hardcodes -L paths.
+    case "$host_os" in
+    solaris2.[0-4]*)
+      hardcode_minus_L=yes
+      ;;
+    esac
+    ;;
+
+  sunos4*)
+    if test "$with_gcc" = yes; then
+      archive_cmds='$CC -shared -o $lib$libobjs'
+    else
+      archive_cmds='$LD -assert pure-text -Bstatic -o $lib$libobjs'
+    fi
+
+    if test "$with_gnu_ld" = yes; then
+      export_dynamic_flag_spec='${wl}-export-dynamic'
+    fi
+    hardcode_libdir_flag_spec='-L$libdir'
+    hardcode_direct=yes
+    hardcode_minus_L=yes
+    hardcode_shlibpath_var=no
+    ;;
+
+  uts4*)
+    archive_cmds='$LD -G -h $soname -o $lib$libobjs'
+    hardcode_libdir_flag_spec='-L$libdir'
+    hardcode_direct=no
+    hardcode_minus_L=no
+    hardcode_shlibpath_var=no
+    ;;
+
+  *)
+    ld_shlibs=no
+    can_build_shared=no
+    ;;
+  esac
+fi
+echo "$ac_t$ld_shlibs" 1>&6
+
+if test -z "$NM"; then
+  echo $ac_n "checking for BSD-compatible nm... $ac_c" 1>&6
+  case "$NM" in
+  /* | [A-Za-z]:\\*) ;; # Let the user override the test with a path.
+  *)
+    IFS="${IFS= 	}"; ac_save_ifs="$IFS"; IFS="${IFS}:"
+    for ac_dir in /usr/ucb /usr/ccs/bin $PATH /bin; do
+      test -z "$ac_dir" && ac_dir=.
+      if test -f $ac_dir/nm; then
+        # Check to see if the nm accepts a BSD-compat flag.
+        # Adding the `sed 1q' prevents false positives on HP-UX, which says:
+        #   nm: unknown option "B" ignored
+        if ($ac_dir/nm -B /dev/null 2>&1 | sed '1q'; exit 0) | egrep /dev/null >/dev/null; then
+          NM="$ac_dir/nm -B"
+        elif ($ac_dir/nm -p /dev/null 2>&1 | sed '1q'; exit 0) | egrep /dev/null >/dev/null; then
+          NM="$ac_dir/nm -p"
+	else
+          NM="$ac_dir/nm"
+	fi
+        break
+      fi
+    done
+    IFS="$ac_save_ifs"
+    test -z "$NM" && NM=nm
+    ;;
+  esac
+  echo "$ac_t$NM" 1>&6
+fi
+
+# Check for command to grab the raw symbol name followed by C symbol from nm.
+echo $ac_n "checking command to parse $NM output... $ac_c" 1>&6
+
+# These are sane defaults that work on at least a few old systems.
+# [They come from Ultrix.  What could be older than Ultrix?!! ;)]
+
+# Character class describing NM global symbol codes.
+symcode='[BCDEGRSTU]'
+
+# Regexp to match symbols that can be accessed directly from C.
+sympat='\([_A-Za-z][_A-Za-z0-9]*\)'
+
+# Transform the above into a raw symbol and a C symbol.
+symxfrm='\1 \1'
+
+# Define system-specific variables.
+case "$host_os" in
+aix*)
+  symcode='[BCDTU]'
+  ;;
+irix*)
+  # Cannot use undefined symbols on IRIX because inlined functions mess us up.
+  symcode='[BCDEGRST]'
+  ;;
+solaris2*)
+  symcode='[BDTU]'
+  ;;
+esac
+
+# If we're using GNU nm, then use its standard symbol codes.
+if $NM -V 2>&1 | egrep '(GNU|with BFD)' > /dev/null; then
+  symcode='[ABCDGISTUW]'
+fi
+
+# Write the raw and C identifiers.
+global_symbol_pipe="sed -n -e 's/^.* $symcode $sympat$/$symxfrm/p'"
+
+# Check to see that the pipe works correctly.
+pipe_works=no
+$rm conftest*
+cat > conftest.c <<EOF
+#ifdef __cplusplus
+extern "C" {
+#endif
+char nm_test_var;
+void nm_test_func(){}
+#ifdef __cplusplus
+}
+#endif
+main(){nm_test_var='a';nm_test_func();return(0);}
+EOF
+
+echo "$progname:971: checking if global_symbol_pipe works" >&5
+if { (eval echo $progname:972: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; } && test -s conftest.o; then
+  # Now try to grab the symbols.
+  nlist=conftest.nm
+  if { echo "$progname:975: eval \"$NM conftest.o | $global_symbol_pipe > $nlist\"" >&5; eval "$NM conftest.o | $global_symbol_pipe > $nlist 2>&5"; } && test -s "$nlist"; then
+
+    # Try sorting and uniquifying the output.
+    if sort "$nlist" | uniq > "$nlist"T; then
+      mv -f "$nlist"T "$nlist"
+      wcout=`wc "$nlist" 2>/dev/null`
+      count=`$echo "X$wcout" | $Xsed -e 's/^[ 	]*\([0-9][0-9]*\).*$/\1/'`
+      (test "$count" -ge 0) 2>/dev/null || count=-1
+    else
+      rm -f "$nlist"T
+      count=-1
+    fi
+
+    # Make sure that we snagged all the symbols we need.
+    if egrep ' nm_test_var$' "$nlist" >/dev/null; then
+      if egrep ' nm_test_func$' "$nlist" >/dev/null; then
+	cat <<EOF > conftest.c
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+EOF
+        # Now generate the symbol file.
+        sed 's/^.* \(.*\)$/extern char \1;/' < "$nlist" >> conftest.c
+
+	cat <<EOF >> conftest.c
+#if defined (__STDC__) && __STDC__
+# define __ptr_t void *
+#else
+# define __ptr_t char *
+#endif
+
+/* The number of symbols in dld_preloaded_symbols, -1 if unsorted. */
+int dld_preloaded_symbol_count = $count;
+
+/* The mapping between symbol names and symbols. */
+struct {
+  char *name;
+  __ptr_t address;
+}
+dld_preloaded_symbols[] =
+{
+EOF
+        sed 's/^\(.*\) \(.*\)$/  {"\1", (__ptr_t) \&\2},/' < "$nlist" >> conftest.c
+        cat <<\EOF >> conftest.c
+  {0, (__ptr_t) 0}
+};
+
+#ifdef __cplusplus
+}
+#endif
+EOF
+        # Now try linking the two files.
+        mv conftest.o conftestm.o
+	save_LIBS="$LIBS"
+	save_CFLAGS="$CFLAGS"
+        LIBS='conftestm.o'
+	CFLAGS="$CFLAGS$no_builtin_flag"
+        if { (eval echo $progname:1033: \"$ac_link\") 1>&5; (eval $ac_link) 2>&5; } && test -s conftest; then
+          pipe_works=yes
+        else
+          echo "$progname: failed program was:" >&5
+          cat conftest.c >&5
+        fi
+        LIBS="$save_LIBS"
+      else
+        echo "cannot find nm_test_func in $nlist" >&5
+      fi
+    else
+      echo "cannot find nm_test_var in $nlist" >&5
+    fi
+  else
+    echo "cannot run $global_symbol_pipe" >&5
+  fi
+else
+  echo "$progname: failed program was:" >&5
+  cat conftest.c >&5
+fi
+$rm conftest*
+
+# Do not use the global_symbol_pipe unless it works.
+echo "$ac_t$pipe_works" 1>&6
+test "$pipe_works" = yes || global_symbol_pipe=
+
+# Check hardcoding attributes.
+echo $ac_n "checking how to hardcode library paths into programs... $ac_c" 1>&6
+hardcode_action=
+if test -n "$hardcode_libdir_flag_spec" || \
+   test -n "$runpath_var"; then
+
+  # We can hardcode non-existant directories.
+  if test "$hardcode_direct" != no && \
+     test "$hardcode_minus_L" != no && \
+     test "$hardcode_shlibpath_var" != no; then
+
+    # Linking always hardcodes the temporary library directory.
+    hardcode_action=relink
+  else
+    # We can link without hardcoding, and we can hardcode nonexisting dirs.
+    hardcode_action=immediate
+  fi
+elif test "$hardcode_direct" != yes && \
+     test "$hardcode_minus_L" != yes && \
+     test "$hardcode_shlibpath_var" != yes; then
+  # We cannot hardcode anything.
+  hardcode_action=unsupported
+else
+  # We can only hardcode existing directories.
+  hardcode_action=relink
+fi
+echo "$ac_t$hardcode_action" 1>&6
+test "$hardcode_action" = unsupported && can_build_shared=no
+
+
+reload_flag=
+reload_cmds='$LD$reload_flag -o $output$reload_objs'
+echo $ac_n "checking for $LD option to reload object files... $ac_c" 1>&6
+# PORTME Some linker may need a different reload flag.
+reload_flag='-r'
+echo "$ac_t$reload_flag"
+test -n "$reload_flag" && reload_flag=" $reload_flag"
+
+# PORTME Fill in your ld.so characteristics
+library_names_spec=
+libname_spec='lib$name'
+soname_spec=
+postinstall_cmds=
+postuninstall_cmds=
+finish_cmds=
+finish_eval=
+shlibpath_var=
+version_type=none
+dynamic_linker="$host_os ld.so"
+
+echo $ac_n "checking dynamic linker characteristics... $ac_c" 1>&6
+case "$host_os" in
+aix3* | aix4*)
+  version_type=linux
+  library_names_spec='${libname}${release}.so.$versuffix $libname.a'
+  shlibpath_var=LIBPATH
+
+  # AIX has no versioning support, so we append a major version to the name.
+  soname_spec='${libname}${release}.so.$major'
+  ;;
+
+amigaos*)
+  library_names_spec='$libname.ixlibrary $libname.a'
+  # Create ${libname}_ixlibrary.a entries in /sys/libs.
+  finish_eval='for lib in `ls $libdir/*.ixlibrary 2>/dev/null`; do libname=`$echo "X$lib" | $Xsed -e '\''s%^.*/\([^/]*\)\.ixlibrary$%\1%'\''`; test $rm /sys/libs/${libname}_ixlibrary.a; $show "(cd /sys/libs && $LN_S $lib ${libname}_ixlibrary.a)"; (cd /sys/libs && $LN_S $lib ${libname}_ixlibrary.a) || exit 1; done'
+  ;;
+
+freebsd2* | freebsd3*)
+  version_type=sunos
+  library_names_spec='${libname}${release}.so.$versuffix $libname.so'
+  finish_cmds='PATH="$PATH:/sbin" ldconfig -m $libdir'
+  shlibpath_var=LD_LIBRARY_PATH
+  ;;
+
+gnu*)
+  version_type=sunos
+  library_names_spec='${libname}${release}.so.$versuffix'
+  shlibpath_var=LD_LIBRARY_PATH
+  ;;
+
+hpux9* | hpux10*)
+  # Give a soname corresponding to the major version so that dld.sl refuses to
+  # link against other versions.
+  dynamic_linker="$host_os dld.sl"
+  version_type=sunos
+  shlibpath_var=SHLIB_PATH
+  library_names_spec='${libname}${release}.sl.$versuffix ${libname}${release}.sl.$major $libname.sl'
+  soname_spec='${libname}${release}.sl.$major'
+  # HP-UX runs *really* slowly unless shared libraries are mode 555.
+  postinstall_cmds='chmod 555 $lib'
+  ;;
+
+irix5* | irix6*)
+  version_type=osf
+  soname_spec='${libname}${release}.so'
+  library_names_spec='${libname}${release}.so.$versuffix $libname.so'
+  shlibpath_var=LD_LIBRARY_PATH
+  ;;
+
+# No shared lib support for Linux oldld, aout, or coff.
+linux-gnuoldld* | linux-gnuaout* | linux-gnucoff*)
+  dynamic_linker=no
+  ;;
+
+# This must be Linux ELF.
+linux-gnu*)
+  version_type=linux
+  library_names_spec='${libname}${release}.so.$versuffix ${libname}${release}.so.$major $libname.so'
+  soname_spec='${libname}${release}.so.$major'
+  finish_cmds='PATH="$PATH:/sbin" ldconfig -n $libdir'
+  shlibpath_var=LD_LIBRARY_PATH
+
+  if test -f /lib/ld.so.1; then
+    dynamic_linker='GNU ld.so'
+  else
+    # Only the GNU ld.so supports shared libraries on MkLinux.
+    case "$host_cpu" in
+    powerpc*) dynamic_linker=no ;;
+    *) dynamic_linker='Linux ld.so' ;;
+    esac
+  fi
+  ;;
+
+netbsd* | openbsd*)
+  version_type=sunos
+  library_names_spec='${libname}${release}.so.$versuffix'
+  finish_cmds='PATH="$PATH:/sbin" ldconfig -m $libdir'
+  shlibpath_var=LD_LIBRARY_PATH
+  ;;
+
+os2*)
+  libname_spec='$name'
+  library_names_spec='$libname.dll $libname.a'
+  dynamic_linker='OS/2 ld.exe'
+  shlibpath_var=LIBPATH
+  ;;
+
+osf3* | osf4*)
+  version_type=osf
+  soname_spec='${libname}${release}.so'
+  library_names_spec='${libname}${release}.so.$versuffix $libname.so'
+  shlibpath_var=LD_LIBRARY_PATH
+  ;;
+
+sco3.2v5*)
+  version_type=osf
+  soname_spec='${libname}${release}.so.$major'
+  library_names_spec='${libname}${release}.so.$versuffix ${libname}${release}.so.$major $libname.so'
+  shlibpath_var=LD_LIBRARY_PATH
+  ;;
+
+solaris2*)
+  version_type=linux
+  library_names_spec='${libname}${release}.so.$versuffix ${libname}${release}.so.$major $libname.so'
+  soname_spec='${libname}${release}.so.$major'
+  shlibpath_var=LD_LIBRARY_PATH
+  ;;
+
+sunos4*)
+  version_type=sunos
+  library_names_spec='${libname}${release}.so.$versuffix'
+  finish_cmds='PATH="$PATH:/usr/etc" ldconfig $libdir'
+  shlibpath_var=LD_LIBRARY_PATH
+  ;;
+
+sysv4.2uw2*)
+  version_type=linux
+  library_names_spec='${libname}${release}.so.$versuffix ${libname}${release}.so.$major $libname.so'
+  soname_spec='${libname}${release}.so.$major'
+  shlibpath_var=LD_LIBRARY_PATH
+  ;;
+
+uts4*)
+  version_type=linux
+  library_names_spec='${libname}${release}.so.$versuffix ${libname}${release}.so.$major $libname.so'
+  soname_spec='${libname}${release}.so.$major'
+  shlibpath_var=LD_LIBRARY_PATH
+  ;;
+
+*)
+  dynamic_linker=no
+  ;;
+esac
+echo "$ac_t$dynamic_linker"
+test "$dynamic_linker" = no && can_build_shared=no
+
+# Report the final consequences.
+echo "checking if libtool supports shared libraries... $can_build_shared" 1>&6
+
+echo $ac_n "checking whether to build shared libraries... $ac_c" 1>&6
+test "$can_build_shared" = "no" && enable_shared=no
+
+# On AIX, shared libraries and static libraries use the same namespace, and
+# are all built from PIC.
+case "$host_os" in
+aix*)
+  test "$enable_shared" = yes && enable_static=no
+  if test -n "$RANLIB"; then
+    archive_cmds="$archive_cmds;\$RANLIB \$lib"
+    postinstall_cmds='$RANLIB $lib'
+  fi
+  ;;
+esac
+
+echo "$ac_t$enable_shared" 1>&6
+
+# Make sure either enable_shared or enable_static is yes.
+test "$enable_shared" = yes || enable_static=yes
+
+echo "checking whether to build static libraries... $enable_static" 1>&6
+
+echo $ac_n "checking for objdir... $ac_c" 1>&6
+rm -f .libs 2>/dev/null
+mkdir .libs 2>/dev/null
+if test -d .libs; then
+  objdir=.libs
+else
+  # MS-DOS does not allow filenames that begin with a dot.
+  objdir=_libs
+fi
+rmdir .libs 2>/dev/null
+echo "$ac_t$objdir" 1>&6
+
+# Copy echo and quote the copy, instead of the original, because it is
+# used later.
+ltecho="$echo"
+
+# Now quote all the things that may contain metacharacters.
+for var in ltecho old_CC old_CFLAGS old_CPPFLAGS old_LD old_NM old_RANLIB \
+  old_LN_S AR CC LD LN_S NM reload_flag reload_cmds wl pic_flag \
+  link_static_flag no_builtin_flag export_dynamic_flag_spec \
+  libname_spec library_names_spec soname_spec RANLIB \
+  old_archive_cmds old_archive_from_new_cmds old_postinstall_cmds \
+  old_postuninstall_cmds archive_cmds postinstall_cmds postuninstall_cmds \
+  allow_undefined_flag no_undefined_flag \
+  finish_cmds finish_eval global_symbol_pipe \
+  hardcode_libdir_flag_spec hardcode_libdir_separator; do
+
+  case "$var" in
+  reload_cmds | old_archive_cmds | old_archive_from_new_cmds | \
+  old_postinstall_cmds | old_postuninstall_cmds | archive_cmds | \
+  postinstall_cmds | postuninstall_cmds | finish_cmds)
+    # Double-quote double-evaled strings.
+    eval "$var=\`\$echo \"X\$$var\" | \$Xsed -e \"\$double_quote_subst\" -e \"\$sed_quote_subst\"\`"
+    ;;
+  *)
+    eval "$var=\`\$echo \"X\$$var\" | \$Xsed -e \"\$sed_quote_subst\"\`"
+    ;;
+  esac
+done
+
+ofile=libtool
+trap "$rm $ofile; exit 1" 1 2 15
+echo creating $ofile
+$rm $ofile
+cat <<EOF > $ofile
+#! /bin/sh
+
+# libtool - Provide generalized library-building support services.
+# Generated automatically by $PROGRAM - GNU $PACKAGE $VERSION
+# NOTE: Changes made to this file will be lost: look at ltconfig or ltmain.sh.
+#
+# Copyright (C) 1996-1998 Free Software Foundation, Inc.
+# Gordon Matzigkeit <gord@gnu.ai.mit.edu>, 1996
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+# General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+#
+# As a special exception to the GNU General Public License, if you
+# distribute this file as part of a program that contains a
+# configuration script generated by Autoconf, you may include it under
+# the same distribution terms that you use for the rest of that program.
+
+# This program was configured as follows,
+# on host `(hostname || uname -n) 2>/dev/null | sed 1q`:
+#
+# CC="$old_CC" CFLAGS="$old_CFLAGS" CPPFLAGS="$old_CPPFLAGS" \\
+# LD="$old_LD" NM="$old_NM" RANLIB="$old_RANLIB" LN_S="$old_LN_S" \\
+#   $0$ltconfig_args
+#
+# Compiler and other test output produced by $progname, useful for
+# debugging $progname, is in ./config.log if it exists.
+
+# Sed that helps us avoid accidentally triggering echo(1) options like -n.
+Xsed="sed -e s/^X//"
+
+# The HP-UX ksh and POSIX shell print the target directory to stdout
+# if CDPATH is set.
+if test "\${CDPATH+set}" = set; then CDPATH=; export CDPATH; fi
+
+# An echo program that does not interpret backslashes.
+echo="$ltecho"
+
+# The version of $progname that generated this script.
+LTCONFIG_VERSION="$VERSION"
+
+# Shell to use when invoking shell scripts.
+SHELL=${CONFIG_SHELL-/bin/sh}
+
+# Whether or not to build libtool libraries.
+build_libtool_libs=$enable_shared
+
+# Whether or not to build old-style libraries.
+build_old_libs=$enable_static
+
+# The host system.
+host_alias="$host_alias"
+host="$host"
+
+# The archiver.
+AR="$AR"
+
+# The default C compiler.
+CC="$CC"
+
+# The linker used to build libraries.
+LD="$LD"
+
+# Whether we need hard or soft links.
+LN_S="$LN_S"
+
+# A BSD-compatible nm program.
+NM="$NM"
+
+# The name of the directory that contains temporary libtool files.
+objdir="$objdir"
+
+# How to create reloadable object files.
+reload_flag="$reload_flag"
+reload_cmds="$reload_cmds"
+
+# How to pass a linker flag through the compiler.
+wl="$wl"
+
+# Additional compiler flags for building library objects.
+pic_flag="$pic_flag"
+
+# Compiler flag to prevent dynamic linking.
+link_static_flag="$link_static_flag"
+
+# Compiler flag to turn off builtin functions.
+no_builtin_flag="$no_builtin_flag"
+
+# Compiler flag to allow reflexive dlopens.
+export_dynamic_flag_spec="$export_dynamic_flag_spec"
+
+# Library versioning type.
+version_type=$version_type
+
+# Format of library name prefix.
+libname_spec="$libname_spec"
+
+# List of archive names.  First name is the real one, the rest are links.
+# The last name is the one that the linker finds with -lNAME.
+library_names_spec="$library_names_spec"
+
+# The coded name of the library, if different from the real name.
+soname_spec="$soname_spec"
+
+# Commands used to build and install an old-style archive.
+RANLIB="$RANLIB"
+old_archive_cmds="$old_archive_cmds"
+old_postinstall_cmds="$old_postinstall_cmds"
+old_postuninstall_cmds="$old_postuninstall_cmds"
+
+# Create an old-style archive from a shared archive.
+old_archive_from_new_cmds="$old_archive_from_new_cmds"
+
+# Commands used to build and install a shared archive.
+archive_cmds="$archive_cmds"
+postinstall_cmds="$postinstall_cmds"
+postuninstall_cmds="$postuninstall_cmds"
+
+# Flag that allows shared libraries with undefined symbols to be built.
+allow_undefined_flag="$allow_undefined_flag"
+
+# Flag that forces no undefined symbols.
+no_undefined_flag="$no_undefined_flag"
+
+# Commands used to finish a libtool library installation in a directory.
+finish_cmds="$finish_cmds"
+
+# Same as above, but a single script fragment to be evaled but not shown.
+finish_eval="$finish_eval"
+
+# Take the output of nm and produce a listing of raw symbols and C names.
+global_symbol_pipe="$global_symbol_pipe"
+
+# This is the shared library runtime path variable.
+runpath_var=$runpath_var
+
+# This is the shared library path variable.
+shlibpath_var=$shlibpath_var
+
+# How to hardcode a shared library path into an executable.
+hardcode_action=$hardcode_action
+
+# Flag to hardcode \$libdir into a binary during linking.
+# This must work even if \$libdir does not exist.
+hardcode_libdir_flag_spec="$hardcode_libdir_flag_spec"
+
+# Whether we need a single -rpath flag with a separated argument.
+hardcode_libdir_separator="$hardcode_libdir_separator"
+
+# Set to yes if using DIR/libNAME.so during linking hardcodes DIR into the
+# resulting binary.
+hardcode_direct=$hardcode_direct
+
+# Set to yes if using the -LDIR flag during linking hardcodes DIR into the
+# resulting binary.
+hardcode_minus_L=$hardcode_minus_L
+
+# Set to yes if using SHLIBPATH_VAR=DIR during linking hardcodes DIR into
+# the resulting binary.
+hardcode_shlibpath_var=$hardcode_shlibpath_var
+
+EOF
+
+case "$host_os" in
+aix3*)
+  cat <<\EOF >> $ofile
+# AIX sometimes has problems with the GCC collect2 program.  For some
+# reason, if we set the COLLECT_NAMES environment variable, the problems
+# vanish in a puff of smoke.
+if test "${COLLECT_NAMES+set}" != set; then
+  COLLECT_NAMES=
+  export COLLECT_NAMES
+fi
+
+EOF
+  ;;
+esac
+
+# Append the ltmain.sh script.
+cat "$ltmain" >> $ofile || (rm -f $ofile; exit 1)
+
+chmod +x $ofile
+exit 0
+
+# Local Variables:
+# mode:shell-script
+# sh-indentation:2
+# End:
diff --git a/libjpegtwrp/ltmain.sh b/libjpegtwrp/ltmain.sh
new file mode 100644
index 000000000..e9350b3fa
--- /dev/null
+++ b/libjpegtwrp/ltmain.sh
@@ -0,0 +1,2453 @@
+# ltmain.sh - Provide generalized library-building support services.
+# NOTE: Changing this file will not affect anything until you rerun ltconfig.
+#
+# Copyright (C) 1996-1998 Free Software Foundation, Inc.
+# Gordon Matzigkeit <gord@gnu.ai.mit.edu>, 1996
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+# General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+#
+# As a special exception to the GNU General Public License, if you
+# distribute this file as part of a program that contains a
+# configuration script generated by Autoconf, you may include it under
+# the same distribution terms that you use for the rest of that program.
+
+# The name of this program.
+progname=`$echo "$0" | sed 's%^.*/%%'`
+modename="$progname"
+
+# Constants.
+PROGRAM=ltmain.sh
+PACKAGE=libtool
+VERSION=1.2
+
+default_mode=
+help="Try \`$progname --help' for more information."
+magic="%%%MAGIC variable%%%"
+mkdir="mkdir"
+mv="mv -f"
+rm="rm -f"
+
+# Sed substitution that helps us do robust quoting.  It backslashifies
+# metacharacters that are still active within double-quoted strings.
+Xsed='sed -e s/^X//'
+sed_quote_subst='s/\([\\`\\"$\\\\]\)/\\\1/g'
+
+# NLS nuisances.
+# Only set LANG and LC_ALL to C if already set.
+# These must not be set unconditionally because not all systems understand
+# e.g. LANG=C (notably SCO).
+if test "${LC_ALL+set}" = set; then LC_ALL=C; export LC_ALL; fi
+if test "${LANG+set}"   = set; then LANG=C;   export LANG;   fi
+
+if test "$LTCONFIG_VERSION" != "$VERSION"; then
+  echo "$modename: ltconfig version \`$LTCONFIG_VERSION' does not match $PROGRAM version \`$VERSION'" 1>&2
+  echo "Fatal configuration error.  See the $PACKAGE docs for more information." 1>&2
+  exit 1
+fi
+
+if test "$build_libtool_libs" != yes && test "$build_old_libs" != yes; then
+  echo "$modename: not configured to build any kind of library" 1>&2
+  echo "Fatal configuration error.  See the $PACKAGE docs for more information." 1>&2
+  exit 1
+fi
+
+# Global variables.
+mode=$default_mode
+nonopt=
+prev=
+prevopt=
+run=
+show="$echo"
+show_help=
+execute_dlfiles=
+
+# Parse our command line options once, thoroughly.
+while test $# -gt 0
+do
+  arg="$1"
+  shift
+
+  case "$arg" in
+  -*=*) optarg=`$echo "X$arg" | $Xsed -e 's/[-_a-zA-Z0-9]*=//'` ;;
+  *) optarg= ;;
+  esac
+
+  # If the previous option needs an argument, assign it.
+  if test -n "$prev"; then
+    case "$prev" in
+    execute_dlfiles)
+      eval "$prev=\"\$$prev \$arg\""
+      ;;
+    *)
+      eval "$prev=\$arg"
+      ;;
+    esac
+
+    prev=
+    prevopt=
+    continue
+  fi
+
+  # Have we seen a non-optional argument yet?
+  case "$arg" in
+  --help)
+    show_help=yes
+    ;;
+
+  --version)
+    echo "$PROGRAM (GNU $PACKAGE) $VERSION"
+    exit 0
+    ;;
+
+  --dry-run | -n)
+    run=:
+    ;;
+
+  --features)
+    echo "host: $host"
+    if test "$build_libtool_libs" = yes; then
+      echo "enable shared libraries"
+    else
+      echo "disable shared libraries"
+    fi
+    if test "$build_old_libs" = yes; then
+      echo "enable static libraries"
+    else
+      echo "disable static libraries"
+    fi
+    exit 0
+    ;;
+
+  --finish) mode="finish" ;;
+
+  --mode) prevopt="--mode" prev=mode ;;
+  --mode=*) mode="$optarg" ;;
+
+  --quiet | --silent)
+    show=:
+    ;;
+
+  -dlopen)
+    prevopt="-dlopen"
+    prev=execute_dlfiles
+    ;;
+
+  -*)
+    $echo "$modename: unrecognized option \`$arg'" 1>&2
+    $echo "$help" 1>&2
+    exit 1
+    ;;
+
+  *)
+    nonopt="$arg"
+    break
+    ;;
+  esac
+done
+
+if test -n "$prevopt"; then
+  $echo "$modename: option \`$prevopt' requires an argument" 1>&2
+  $echo "$help" 1>&2
+  exit 1
+fi
+
+if test -z "$show_help"; then
+
+  # Infer the operation mode.
+  if test -z "$mode"; then
+    case "$nonopt" in
+    *cc | *++ | gcc* | *-gcc*)
+      mode=link
+      for arg
+      do
+        case "$arg" in
+        -c)
+           mode=compile
+           break
+           ;;
+        esac
+      done
+      ;;
+    *db | *dbx)
+      mode=execute
+      ;;
+    *install*|cp|mv)
+      mode=install
+      ;;
+    *rm)
+      mode=uninstall
+      ;;
+    *)
+      # If we have no mode, but dlfiles were specified, then do execute mode.
+      test -n "$execute_dlfiles" && mode=execute
+
+      # Just use the default operation mode.
+      if test -z "$mode"; then
+        if test -n "$nonopt"; then
+          $echo "$modename: warning: cannot infer operation mode from \`$nonopt'" 1>&2
+        else
+          $echo "$modename: warning: cannot infer operation mode without MODE-ARGS" 1>&2
+        fi
+      fi
+      ;;
+    esac
+  fi
+
+  # Only execute mode is allowed to have -dlopen flags.
+  if test -n "$execute_dlfiles" && test "$mode" != execute; then
+    $echo "$modename: unrecognized option \`-dlopen'" 1>&2
+    $echo "$help" 1>&2
+    exit 1
+  fi
+
+  # Change the help message to a mode-specific one.
+  generic_help="$help"
+  help="Try \`$modename --help --mode=$mode' for more information."
+
+  # These modes are in order of execution frequency so that they run quickly.
+  case "$mode" in
+  # libtool compile mode
+  compile)
+    modename="$modename: compile"
+    # Get the compilation command and the source file.
+    base_compile=
+    lastarg=
+    srcfile="$nonopt"
+    suppress_output=
+
+    for arg
+    do
+      # Accept any command-line options.
+      case "$arg" in
+      -o)
+	$echo "$modename: you cannot specify the output filename with \`-o'" 1>&2
+	$echo "$help" 1>&2
+	exit 1
+	;;
+
+      -static)
+	build_libtool_libs=no
+	build_old_libs=yes
+	continue
+	;;
+      esac
+
+      # Accept the current argument as the source file.
+      lastarg="$srcfile"
+      srcfile="$arg"
+
+      # Aesthetically quote the previous argument.
+
+      # Backslashify any backslashes, double quotes, and dollar signs.
+      # These are the only characters that are still specially
+      # interpreted inside of double-quoted scrings.
+      lastarg=`$echo "X$lastarg" | $Xsed -e "$sed_quote_subst"`
+
+      # Double-quote args containing other shell metacharacters.
+      # Many Bourne shells cannot handle close brackets correctly in scan
+      # sets, so we specify it separately.
+      case "$lastarg" in
+      *[\[\~\#\^\&\*\(\)\{\}\|\;\<\>\?\'\ \	]*|*]*)
+	lastarg="\"$lastarg\""
+	;;
+      esac
+
+      # Add the previous argument to base_compile.
+      if test -z "$base_compile"; then
+	base_compile="$lastarg"
+      else
+	base_compile="$base_compile $lastarg"
+      fi
+    done
+
+    # Get the name of the library object.
+    libobj=`$echo "X$srcfile" | $Xsed -e 's%^.*/%%'`
+
+    # Recognize several different file suffixes.
+    xform='[cCFSfms]'
+    case "$libobj" in
+    *.ada) xform=ada ;;
+    *.adb) xform=adb ;;
+    *.ads) xform=ads ;;
+    *.asm) xform=asm ;;
+    *.c++) xform=c++ ;;
+    *.cc) xform=cc ;;
+    *.cpp) xform=cpp ;;
+    *.cxx) xform=cxx ;;
+    *.f90) xform=f90 ;;
+    *.for) xform=for ;;
+    esac
+
+    libobj=`$echo "X$libobj" | $Xsed -e "s/\.$xform$/.lo/"`
+
+    case "$libobj" in
+    *.lo) obj=`$echo "X$libobj" | $Xsed -e 's/\.lo$/.o/'` ;;
+    *)
+      $echo "$modename: cannot determine name of library object from \`$srcfile'" 1>&2
+      exit 1
+      ;;
+    esac
+
+    if test -z "$base_compile"; then
+      $echo "$modename: you must specify a compilation command" 1>&2
+      $echo "$help" 1>&2
+      exit 1
+    fi
+
+    # Delete any leftover library objects.
+    if test "$build_old_libs" = yes; then
+      $run $rm $obj $libobj
+      trap "$run $rm $obj $libobj; exit 1" 1 2 15
+    else
+      $run $rm $libobj
+      trap "$run $rm $libobj; exit 1" 1 2 15
+    fi
+
+    # Only build a PIC object if we are building libtool libraries.
+    if test "$build_libtool_libs" = yes; then
+      # Without this assignment, base_compile gets emptied.
+      fbsd_hideous_sh_bug=$base_compile
+
+      # All platforms use -DPIC, to notify preprocessed assembler code.
+      $show "$base_compile$pic_flag -DPIC $srcfile"
+      if $run eval "$base_compile\$pic_flag -DPIC \$srcfile"; then :
+      else
+        test -n "$obj" && $run $rm $obj
+        exit 1
+      fi
+
+      # If we have no pic_flag, then copy the object into place and finish.
+      if test -z "$pic_flag"; then
+        $show "$LN_S $obj $libobj"
+        $run $LN_S $obj $libobj
+        exit $?
+      fi
+
+      # Just move the object, then go on to compile the next one
+      $show "$mv $obj $libobj"
+      $run $mv $obj $libobj || exit 1
+
+      # Allow error messages only from the first compilation.
+      suppress_output=' >/dev/null 2>&1'
+    fi
+
+    # Only build a position-dependent object if we build old libraries.
+    if test "$build_old_libs" = yes; then
+      # Suppress compiler output if we already did a PIC compilation.
+      $show "$base_compile $srcfile$suppress_output"
+      if $run eval "$base_compile \$srcfile$suppress_output"; then :
+      else
+        $run $rm $obj $libobj
+        exit 1
+      fi
+    fi
+
+    # Create an invalid libtool object if no PIC, so that we do not
+    # accidentally link it into a program.
+    if test "$build_libtool_libs" != yes; then
+      $show "echo timestamp > $libobj"
+      $run eval "echo timestamp > \$libobj" || exit $?
+    fi
+
+    exit 0
+    ;;
+
+  # libtool link mode
+  link)
+    modename="$modename: link"
+    CC="$nonopt"
+    allow_undefined=yes
+    compile_command="$CC"
+    finalize_command="$CC"
+
+    compile_shlibpath=
+    finalize_shlibpath=
+    deplibs=
+    dlfiles=
+    dlprefiles=
+    export_dynamic=no
+    hardcode_libdirs=
+    libobjs=
+    link_against_libtool_libs=
+    ltlibs=
+    objs=
+    prev=
+    prevarg=
+    release=
+    rpath=
+    perm_rpath=
+    temp_rpath=
+    vinfo=
+
+    # We need to know -static, to get the right output filenames.
+    for arg
+    do
+      case "$arg" in
+      -all-static | -static)
+        if test "X$arg" = "X-all-static" && test "$build_libtool_libs" = yes && test -z "$link_static_flag"; then
+	    $echo "$modename: warning: complete static linking is impossible in this configuration" 1>&2
+        fi
+        build_libtool_libs=no
+	build_old_libs=yes
+        break
+        ;;
+      esac
+    done
+
+    # See if our shared archives depend on static archives.
+    test -n "$old_archive_from_new_cmds" && build_old_libs=yes
+
+    # Go through the arguments, transforming them on the way.
+    for arg
+    do
+      # If the previous option needs an argument, assign it.
+      if test -n "$prev"; then
+        case "$prev" in
+        output)
+          compile_command="$compile_command @OUTPUT@"
+          finalize_command="$finalize_command @OUTPUT@"
+          ;;
+        esac
+
+        case "$prev" in
+        dlfiles|dlprefiles)
+          case "$arg" in
+          *.la | *.lo) ;;  # We handle these cases below.
+          *)
+            dlprefiles="$dlprefiles $arg"
+            test "$prev" = dlfiles && dlfiles="$dlfiles $arg"
+            prev=
+            ;;
+          esac
+          ;;
+	release)
+	  release="-$arg"
+	  prev=
+	  continue
+	  ;;
+        rpath)
+          rpath="$rpath $arg"
+	  prev=
+	  continue
+	  ;;
+        *)
+          eval "$prev=\"\$arg\""
+          prev=
+          continue
+          ;;
+        esac
+      fi
+
+      prevarg="$arg"
+
+      case "$arg" in
+      -all-static)
+	if test -n "$link_static_flag"; then
+          compile_command="$compile_command $link_static_flag"
+	  finalize_command="$finalize_command $link_static_flag"
+        fi
+        continue
+	;;
+
+      -allow-undefined)
+	# FIXME: remove this flag sometime in the future.
+	$echo "$modename: \`-allow-undefined' is deprecated because it is the default" 1>&2
+	continue
+	;;
+
+      -dlopen)
+        prev=dlfiles
+        continue
+        ;;
+
+      -dlpreopen)
+        prev=dlprefiles
+        continue
+        ;;
+
+      -export-dynamic)
+        if test "$export_dynamic" != yes; then
+          export_dynamic=yes
+	  if test -n "$export_dynamic_flag_spec"; then
+	    eval arg=\"$export_dynamic_flag_spec\"
+	  else
+	    arg=
+	  fi
+
+          # Add the symbol object into the linking commands.
+	  compile_command="$compile_command @SYMFILE@"
+	  finalize_command="$finalize_command @SYMFILE@"
+        fi
+        ;;
+
+      -L*)
+        dir=`$echo "X$arg" | $Xsed -e 's%^-L\(.*\)$%\1%'`
+        case "$dir" in
+        /* | [A-Za-z]:\\*)
+	  # Add the corresponding hardcode_libdir_flag, if it is not identical.
+          ;;
+        *)
+          $echo "$modename: \`-L$dir' cannot specify a relative directory" 1>&2
+          exit 1
+          ;;
+        esac
+        deplibs="$deplibs $arg"
+        ;;
+
+      -l*) deplibs="$deplibs $arg" ;;
+
+      -no-undefined)
+	allow_undefined=no
+	continue
+	;;
+
+      -o) prev=output ;;
+
+      -release)
+	prev=release
+	continue
+	;;
+
+      -rpath)
+        prev=rpath
+        continue
+        ;;
+
+      -static)
+	# If we have no pic_flag, then this is the same as -all-static.
+	if test -z "$pic_flag" && test -n "$link_static_flag"; then
+          compile_command="$compile_command $link_static_flag"
+	  finalize_command="$finalize_command $link_static_flag"
+        fi
+	continue
+	;;
+
+      -version-info)
+        prev=vinfo
+        continue
+        ;;
+
+      # Some other compiler flag.
+      -* | +*)
+	# Unknown arguments in both finalize_command and compile_command need
+	# to be aesthetically quoted because they are evaled later.
+	arg=`$echo "X$arg" | $Xsed -e "$sed_quote_subst"`
+	case "$arg" in
+	*[\[\~\#\^\&\*\(\)\{\}\|\;\<\>\?\'\ \	]*|*]*)
+	  arg="\"$arg\""
+	  ;;
+	esac
+        ;;
+
+      *.o | *.a)
+        # A standard object.
+        objs="$objs $arg"
+        ;;
+
+      *.lo)
+        # A library object.
+	if test "$prev" = dlfiles; then
+	  dlfiles="$dlfiles $arg"
+	  if test "$build_libtool_libs" = yes; then
+	    prev=
+	    continue
+	  else
+	    # If libtool objects are unsupported, then we need to preload.
+	    prev=dlprefiles
+	  fi
+	fi
+
+	if test "$prev" = dlprefiles; then
+	  # Preload the old-style object.
+	  dlprefiles="$dlprefiles "`$echo "X$arg" | $Xsed -e 's/\.lo$/\.o/'`
+	  prev=
+	fi
+	libobjs="$libobjs $arg"
+        ;;
+
+      *.la)
+        # A libtool-controlled library.
+
+        dlname=
+        libdir=
+        library_names=
+        old_library=
+
+        # Check to see that this really is a libtool archive.
+        if (sed -e '2q' $arg | egrep '^# Generated by ltmain\.sh') >/dev/null 2>&1; then :
+        else
+          $echo "$modename: \`$arg' is not a valid libtool archive" 1>&2
+          exit 1
+        fi
+
+        # If there is no directory component, then add one.
+        case "$arg" in
+        */* | *\\*) . $arg ;;
+        *) . ./$arg ;;
+        esac
+
+        if test -z "$libdir"; then
+          $echo "$modename: \`$arg' contains no -rpath information" 1>&2
+          exit 1
+        fi
+
+        # Get the name of the library we link against.
+        linklib=
+        for l in $old_library $library_names; do
+          linklib="$l"
+        done
+
+        if test -z "$linklib"; then
+          $echo "$modename: cannot find name of link library for \`$arg'" 1>&2
+          exit 1
+        fi
+
+        # Find the relevant object directory and library name.
+        name=`$echo "X$arg" | $Xsed -e 's%^.*/%%' -e 's/\.la$//' -e 's/^lib//'`
+        dir=`$echo "X$arg" | $Xsed -e 's%/[^/]*$%%'`
+        if test "X$dir" = "X$arg"; then
+          dir="$objdir"
+        else
+          dir="$dir/$objdir"
+        fi
+
+        # This library was specified with -dlopen.
+        if test "$prev" = dlfiles; then
+          dlfiles="$dlfiles $arg"
+          if test -z "$dlname"; then
+            # If there is no dlname, we need to preload.
+            prev=dlprefiles
+          else
+            # We should not create a dependency on this library, but we
+	    # may need any libraries it requires.
+	    compile_command="$compile_command$dependency_libs"
+	    finalize_command="$finalize_command$dependency_libs"
+            prev=
+            continue
+          fi
+        fi
+
+        # The library was specified with -dlpreopen.
+        if test "$prev" = dlprefiles; then
+          # Prefer using a static library (so that no silly _DYNAMIC symbols
+          # are required to link).
+          if test -n "$old_library"; then
+            dlprefiles="$dlprefiles $dir/$old_library"
+          else
+            dlprefiles="$dlprefiles $dir/$linklib"
+          fi
+          prev=
+        fi
+
+        if test "$build_libtool_libs" = yes && test -n "$library_names"; then
+          link_against_libtool_libs="$link_against_libtool_libs $arg"
+          if test -n "$shlibpath_var"; then
+            # Make sure the rpath contains only unique directories.
+            case "$temp_rpath " in
+            *" $dir "*) ;;
+            *) temp_rpath="$temp_rpath $dir" ;;
+            esac
+          fi
+
+	  # This is the magic to use -rpath.
+          if test -n "$hardcode_libdir_flag_spec"; then
+            if test -n "$hardcode_libdir_separator"; then
+              if test -z "$hardcode_libdirs"; then
+                # Put the magic libdir with the hardcode flag.
+                hardcode_libdirs="$libdir"
+                libdir="@HARDCODE_LIBDIRS@"
+              else
+                # Just accumulate the unique libdirs.
+		case "$hardcode_libdir_separator$hardcode_libdirs$hardcode_libdir_separator" in
+		*"$hardcode_libdir_separator$libdir$hardcode_libdir_separator"*)
+		  ;;
+		*)
+		  hardcode_libdirs="$hardcode_libdirs$hardcode_libdir_separator$libdir"
+		  ;;
+		esac
+                libdir=
+              fi
+            fi
+
+            if test -n "$libdir"; then
+              eval flag=\"$hardcode_libdir_flag_spec\"
+
+              compile_command="$compile_command $flag"
+              finalize_command="$finalize_command $flag"
+            fi
+          elif test -n "$runpath_var"; then
+            # Do the same for the permanent run path.
+            case "$perm_rpath " in
+            *" $libdir "*) ;;
+            *) perm_rpath="$perm_rpath $libdir" ;;
+            esac
+          fi
+
+
+          case "$hardcode_action" in
+          immediate)
+            if test "$hardcode_direct" = no; then
+              compile_command="$compile_command $dir/$linklib"
+            elif test "$hardcode_minus_L" = no; then
+              compile_command="$compile_command -L$dir -l$name"
+            elif test "$hardcode_shlibpath_var" = no; then
+              compile_shlibpath="$compile_shlibpath$dir:"
+              compile_command="$compile_command -l$name"
+            fi
+            ;;
+
+          relink)
+            # We need an absolute path.
+            case "$dir" in
+            /* | [A-Za-z]:\\*) ;;
+            *)
+              absdir=`cd "$dir" && pwd`
+              if test -z "$absdir"; then
+                $echo "$modename: cannot determine absolute directory name of \`$dir'" 1>&2
+                exit 1
+              fi
+              dir="$absdir"
+              ;;
+            esac
+
+            if test "$hardcode_direct" = yes; then
+              compile_command="$compile_command $dir/$linklib"
+            elif test "$hardcode_minus_L" = yes; then
+              compile_command="$compile_command -L$dir -l$name"
+            elif test "$hardcode_shlibpath_var" = yes; then
+              compile_shlibpath="$compile_shlibpath$dir:"
+              compile_command="$compile_command -l$name"
+            fi
+            ;;
+
+          *)
+            $echo "$modename: \`$hardcode_action' is an unknown hardcode action" 1>&2
+            exit 1
+            ;;
+          esac
+
+          # Finalize command for both is simple: just hardcode it.
+          if test "$hardcode_direct" = yes; then
+            finalize_command="$finalize_command $libdir/$linklib"
+          elif test "$hardcode_minus_L" = yes; then
+            finalize_command="$finalize_command -L$libdir -l$name"
+          elif test "$hardcode_shlibpath_var" = yes; then
+            finalize_shlibpath="$finalize_shlibpath$libdir:"
+            finalize_command="$finalize_command -l$name"
+          else
+            # We cannot seem to hardcode it, guess we'll fake it.
+            finalize_command="$finalize_command -L$libdir -l$name"
+          fi
+        else
+          # Transform directly to old archives if we don't build new libraries.
+          if test -n "$pic_flag" && test -z "$old_library"; then
+            $echo "$modename: cannot find static library for \`$arg'" 1>&2
+            exit 1
+          fi
+
+	  # Here we assume that one of hardcode_direct or hardcode_minus_L
+	  # is not unsupported.  This is valid on all known static and
+	  # shared platforms.
+	  if test "$hardcode_direct" != unsupported; then
+	    test -n "$old_library" && linklib="$old_library"
+	    compile_command="$compile_command $dir/$linklib"
+	    finalize_command="$finalize_command $dir/$linklib"
+	  else
+	    compile_command="$compile_command -L$dir -l$name"
+	    finalize_command="$finalize_command -L$dir -l$name"
+	  fi
+        fi
+
+	# Add in any libraries that this one depends upon.
+	compile_command="$compile_command$dependency_libs"
+	finalize_command="$finalize_command$dependency_libs"
+	continue
+        ;;
+
+      # Some other compiler argument.
+      *)
+	# Unknown arguments in both finalize_command and compile_command need
+	# to be aesthetically quoted because they are evaled later.
+	arg=`$echo "X$arg" | $Xsed -e "$sed_quote_subst"`
+	case "$arg" in
+	*[\[\~\#\^\&\*\(\)\{\}\|\;\<\>\?\'\ \	]*|*]*)
+	  arg="\"$arg\""
+	  ;;
+	esac
+        ;;
+      esac
+
+      # Now actually substitute the argument into the commands.
+      if test -n "$arg"; then
+	compile_command="$compile_command $arg"
+	finalize_command="$finalize_command $arg"
+      fi
+    done
+
+    if test -n "$prev"; then
+      $echo "$modename: the \`$prevarg' option requires an argument" 1>&2
+      $echo "$help" 1>&2
+      exit 1
+    fi
+
+    if test -n "$vinfo" && test -n "$release"; then
+      $echo "$modename: you cannot specify both \`-version-info' and \`-release'" 1>&2
+      $echo "$help" 1>&2
+      exit 1
+    fi
+
+    oldlib=
+    oldobjs=
+    case "$output" in
+    "")
+      $echo "$modename: you must specify an output file" 1>&2
+      $echo "$help" 1>&2
+      exit 1
+      ;;
+
+    */* | *\\*)
+      $echo "$modename: output file \`$output' must have no directory components" 1>&2
+      exit 1
+      ;;
+
+    *.a)
+      # Now set the variables for building old libraries.
+      build_libtool_libs=no
+      build_old_libs=yes
+      oldlib="$output"
+      $show "$rm $oldlib"
+      $run $rm $oldlib
+      ;;
+
+    *.la)
+      # Make sure we only generate libraries of the form `libNAME.la'.
+      case "$output" in
+      lib*) ;;
+      *)
+	$echo "$modename: libtool library \`$arg' must begin with \`lib'" 1>&2
+	$echo "$help" 1>&2
+	exit 1
+	;;
+      esac
+
+      name=`$echo "X$output" | $Xsed -e 's/\.la$//' -e 's/^lib//'`
+      eval libname=\"$libname_spec\"
+
+      # All the library-specific variables (install_libdir is set above).
+      library_names=
+      old_library=
+      dlname=
+      current=0
+      revision=0
+      age=0
+
+      if test -n "$objs"; then
+        $echo "$modename: cannot build libtool library \`$output' from non-libtool objects:$objs" 2>&1
+        exit 1
+      fi
+
+      # How the heck are we supposed to write a wrapper for a shared library?
+      if test -n "$link_against_libtool_libs"; then
+        $echo "$modename: libtool library \`$output' may not depend on uninstalled libraries:$link_against_libtool_libs" 1>&2
+        exit 1
+      fi
+
+      if test -n "$dlfiles$dlprefiles"; then
+        $echo "$modename: warning: \`-dlopen' is ignored while creating libtool libraries" 1>&2
+        # Nullify the symbol file.
+        compile_command=`$echo "X$compile_command" | $Xsed -e "s% @SYMFILE@%%"`
+        finalize_command=`$echo "X$finalize_command" | $Xsed -e "s% @SYMFILE@%%"`
+      fi
+
+      if test -z "$rpath"; then
+        $echo "$modename: you must specify an installation directory with \`-rpath'" 1>&2
+	$echo "$help" 1>&2
+        exit 1
+      fi
+
+      set dummy $rpath
+      if test $# -gt 2; then
+	$echo "$modename: warning: ignoring multiple \`-rpath's for a libtool library" 1>&2
+      fi
+      install_libdir="$2"
+
+      # Parse the version information argument.
+      IFS="${IFS= 	}"; save_ifs="$IFS"; IFS=':'
+      set dummy $vinfo
+      IFS="$save_ifs"
+
+      if test -n "$5"; then
+        $echo "$modename: too many parameters to \`-version-info'" 1>&2
+        $echo "$help" 1>&2
+        exit 1
+      fi
+
+      test -n "$2" && current="$2"
+      test -n "$3" && revision="$3"
+      test -n "$4" && age="$4"
+
+      # Check that each of the things are valid numbers.
+      case "$current" in
+      0 | [1-9] | [1-9][0-9]*) ;;
+      *)
+        $echo "$modename: CURRENT \`$current' is not a nonnegative integer" 1>&2
+        $echo "$modename: \`$vinfo' is not valid version information" 1>&2
+        exit 1
+        ;;
+      esac
+
+      case "$revision" in
+      0 | [1-9] | [1-9][0-9]*) ;;
+      *)
+        $echo "$modename: REVISION \`$revision' is not a nonnegative integer" 1>&2
+        $echo "$modename: \`$vinfo' is not valid version information" 1>&2
+        exit 1
+        ;;
+      esac
+
+      case "$age" in
+      0 | [1-9] | [1-9][0-9]*) ;;
+      *)
+        $echo "$modename: AGE \`$age' is not a nonnegative integer" 1>&2
+        $echo "$modename: \`$vinfo' is not valid version information" 1>&2
+        exit 1
+        ;;
+      esac
+
+      if test $age -gt $current; then
+        $echo "$modename: AGE \`$age' is greater than the current interface number \`$current'" 1>&2
+        $echo "$modename: \`$vinfo' is not valid version information" 1>&2
+        exit 1
+      fi
+
+      # Calculate the version variables.
+      version_vars="version_type current age revision"
+      case "$version_type" in
+      none) ;;
+
+      linux)
+        version_vars="$version_vars major versuffix"
+        major=`expr $current - $age`
+        versuffix="$major.$age.$revision"
+        ;;
+
+      osf)
+        version_vars="$version_vars versuffix verstring"
+        major=`expr $current - $age`
+        versuffix="$current.$age.$revision"
+        verstring="$versuffix"
+
+        # Add in all the interfaces that we are compatible with.
+        loop=$age
+        while test $loop != 0; do
+          iface=`expr $current - $loop`
+          loop=`expr $loop - 1`
+          verstring="$verstring:${iface}.0"
+        done
+
+        # Make executables depend on our current version.
+        verstring="$verstring:${current}.0"
+        ;;
+
+      sunos)
+        version_vars="$version_vars major versuffix"
+        major="$current"
+        versuffix="$current.$revision"
+        ;;
+
+      *)
+        $echo "$modename: unknown library version type \`$version_type'" 1>&2
+        echo "Fatal configuration error.  See the $PACKAGE docs for more information." 1>&2
+        exit 1
+        ;;
+      esac
+
+      # Create the output directory, or remove our outputs if we need to.
+      if test -d $objdir; then
+        $show "$rm $objdir/$output $objdir/$libname.* $objdir/${libname}${release}.*"
+        $run $rm $objdir/$output $objdir/$libname.* $objdir/${libname}${release}.*
+      else
+        $show "$mkdir $objdir"
+        $run $mkdir $objdir
+	status=$?
+	if test $status -eq 0 || test -d $objdir; then :
+	else
+	  exit $status
+	fi
+      fi
+
+      # Check to see if the archive will have undefined symbols.
+      if test "$allow_undefined" = yes; then
+        if test "$allow_undefined_flag" = unsupported; then
+          $echo "$modename: warning: undefined symbols not allowed in $host shared libraries" 1>&2
+          build_libtool_libs=no
+	  build_old_libs=yes
+        fi
+      else
+        # Don't allow undefined symbols.
+        allow_undefined_flag="$no_undefined_flag"
+      fi
+
+      # Add libc to deplibs on all systems.
+      dependency_libs="$deplibs"
+      deplibs="$deplibs -lc"
+
+      if test "$build_libtool_libs" = yes; then
+        # Get the real and link names of the library.
+        eval library_names=\"$library_names_spec\"
+        set dummy $library_names
+        realname="$2"
+        shift; shift
+
+        if test -n "$soname_spec"; then
+          eval soname=\"$soname_spec\"
+        else
+          soname="$realname"
+        fi
+
+        lib="$objdir/$realname"
+	for link
+	do
+	  linknames="$linknames $link"
+	done
+
+        # Use standard objects if they are PIC.
+        test -z "$pic_flag" && libobjs=`$echo "X$libobjs " | $Xsed -e 's/\.lo /.o /g' -e 's/ $//g'`
+
+        # Do each of the archive commands.
+        eval cmds=\"$archive_cmds\"
+        IFS="${IFS= 	}"; save_ifs="$IFS"; IFS=';'
+        for cmd in $cmds; do
+          IFS="$save_ifs"
+          $show "$cmd"
+          $run eval "$cmd" || exit $?
+        done
+        IFS="$save_ifs"
+
+        # Create links to the real library.
+        for linkname in $linknames; do
+          $show "(cd $objdir && $LN_S $realname $linkname)"
+          $run eval '(cd $objdir && $LN_S $realname $linkname)' || exit $?
+        done
+
+        # If -export-dynamic was specified, set the dlname.
+        if test "$export_dynamic" = yes; then
+          # On all known operating systems, these are identical.
+          dlname="$soname"
+        fi
+      fi
+
+      # Now set the variables for building old libraries.
+      oldlib="$objdir/$libname.a"
+      ;;
+
+    *.lo | *.o)
+      if test -n "$link_against_libtool_libs"; then
+        $echo "$modename: error: cannot link libtool libraries into reloadable objects" 1>&2
+        exit 1
+      fi
+
+      if test -n "$deplibs"; then
+        $echo "$modename: warning: \`-l' and \`-L' are ignored while creating objects" 1>&2
+      fi
+
+      if test -n "$dlfiles$dlprefiles"; then
+        $echo "$modename: warning: \`-dlopen' is ignored while creating objects" 1>&2
+        # Nullify the symbol file.
+        compile_command=`$echo "X$compile_command" | $Xsed -e "s% @SYMFILE@%%"`
+        finalize_command=`$echo "X$finalize_command" | $Xsed -e "s% @SYMFILE@%%"`
+      fi
+
+      if test -n "$rpath"; then
+        $echo "$modename: warning: \`-rpath' is ignored while creating objects" 1>&2
+      fi
+
+      if test -n "$vinfo"; then
+        $echo "$modename: warning: \`-version-info' is ignored while creating objects" 1>&2
+      fi
+
+      if test -n "$release"; then
+        $echo "$modename: warning: \`-release' is ignored while creating objects" 1>&2
+      fi
+
+      case "$output" in
+      *.lo)
+        if test -n "$objs"; then
+          $echo "$modename: cannot build library object \`$output' from non-libtool objects" 1>&2
+          exit 1
+        fi
+        libobj="$output"
+        obj=`$echo "X$output" | $Xsed -e 's/\.lo$/.o/'`
+        ;;
+      *)
+        libobj=
+        obj="$output"
+        ;;
+      esac
+
+      # Delete the old objects.
+      $run $rm $obj $libobj
+
+      # Create the old-style object.
+      reload_objs="$objs"`$echo "X$libobjs " | $Xsed -e 's/[^       ]*\.a //g' -e 's/\.lo /.o /g' -e 's/ $//g'`
+
+      output="$obj"
+      eval cmds=\"$reload_cmds\"
+      IFS="${IFS= 	}"; save_ifs="$IFS"; IFS=';'
+      for cmd in $cmds; do
+        IFS="$save_ifs"
+        $show "$cmd"
+        $run eval "$cmd" || exit $?
+      done
+      IFS="$save_ifs"
+
+      # Exit if we aren't doing a library object file.
+      test -z "$libobj" && exit 0
+
+      if test "$build_libtool_libs" != yes; then
+        # Create an invalid libtool object if no PIC, so that we don't
+        # accidentally link it into a program.
+        $show "echo timestamp > $libobj"
+        $run eval "echo timestamp > $libobj" || exit $?
+        exit 0
+      fi
+
+      if test -n "$pic_flag"; then
+        # Only do commands if we really have different PIC objects.
+        reload_objs="$libobjs"
+        output="$libobj"
+        eval cmds=\"$reload_cmds\"
+        IFS="${IFS= 	}"; save_ifs="$IFS"; IFS=';'
+        for cmd in $cmds; do
+          IFS="$save_ifs"
+          $show "$cmd"
+          $run eval "$cmd" || exit $?
+        done
+        IFS="$save_ifs"
+      else
+        # Just create a symlink.
+        $show "$LN_S $obj $libobj"
+        $run $LN_S $obj $libobj || exit 1
+      fi
+
+      exit 0
+      ;;
+
+    *)
+      if test -n "$vinfo"; then
+        $echo "$modename: warning: \`-version-info' is ignored while linking programs" 1>&2
+      fi
+
+      if test -n "$release"; then
+        $echo "$modename: warning: \`-release' is ignored while creating objects" 1>&2
+      fi
+
+      if test -n "$rpath"; then
+	# If the user specified any rpath flags, then add them.
+	for libdir in $rpath; do
+          if test -n "$hardcode_libdir_flag_spec"; then
+            if test -n "$hardcode_libdir_separator"; then
+              if test -z "$hardcode_libdirs"; then
+                # Put the magic libdir with the hardcode flag.
+                hardcode_libdirs="$libdir"
+                libdir="@HARDCODE_LIBDIRS@"
+              else
+                # Just accumulate the unique libdirs.
+		case "$hardcode_libdir_separator$hardcode_libdirs$hardcode_libdir_separator" in
+		*"$hardcode_libdir_separator$libdir$hardcode_libdir_separator"*)
+		  ;;
+		*)
+		  hardcode_libdirs="$hardcode_libdirs$hardcode_libdir_separator$libdir"
+		  ;;
+		esac
+                libdir=
+              fi
+            fi
+
+            if test -n "$libdir"; then
+              eval flag=\"$hardcode_libdir_flag_spec\"
+
+              compile_command="$compile_command $flag"
+              finalize_command="$finalize_command $flag"
+            fi
+          elif test -n "$runpath_var"; then
+            case "$perm_rpath " in
+            *" $libdir "*) ;;
+            *) perm_rpath="$perm_rpath $libdir" ;;
+            esac
+          fi
+	done
+      fi
+
+      # Substitute the hardcoded libdirs into the compile commands.
+      if test -n "$hardcode_libdir_separator"; then
+	compile_command=`$echo "X$compile_command" | $Xsed -e "s%@HARDCODE_LIBDIRS@%$hardcode_libdirs%g"`
+	finalize_command=`$echo "X$finalize_command" | $Xsed -e "s%@HARDCODE_LIBDIRS@%$hardcode_libdirs%g"`
+      fi
+
+      if test -n "$libobjs" && test "$build_old_libs" = yes; then
+        # Transform all the library objects into standard objects.
+        compile_command=`$echo "X$compile_command " | $Xsed -e 's/\.lo /.o /g' -e 's/ $//'`
+        finalize_command=`$echo "X$finalize_command " | $Xsed -e 's/\.lo /.o /g' -e 's/ $//'`
+      fi
+
+      if test "$export_dynamic" = yes && test -n "$NM" && test -n "$global_symbol_pipe"; then
+        dlsyms="${output}S.c"
+      else
+        dlsyms=
+      fi
+
+      if test -n "$dlsyms"; then
+        # Add our own program objects to the preloaded list.
+        dlprefiles=`$echo "X$objs$dlprefiles " | $Xsed -e 's/\.lo /.o /g' -e 's/ $//'`
+
+	# Discover the nlist of each of the dlfiles.
+        nlist="$objdir/${output}.nm"
+
+	if test -d $objdir; then
+	  $show "$rm $nlist ${nlist}T"
+	  $run $rm "$nlist" "${nlist}T"
+	else
+	  $show "$mkdir $objdir"
+	  $run $mkdir $objdir
+	  status=$?
+	  if test $status -eq 0 || test -d $objdir; then :
+	  else
+	    exit $status
+	  fi
+	fi
+
+        for arg in $dlprefiles; do
+	  $show "extracting global C symbols from \`$arg'"
+	  $run eval "$NM $arg | $global_symbol_pipe >> '$nlist'"
+        done
+
+        # Parse the name list into a source file.
+        $show "creating $objdir/$dlsyms"
+        if test -z "$run"; then
+	  # Make sure we at least have an empty file.
+	  test -f "$nlist" || : > "$nlist"
+
+	  # Try sorting and uniquifying the output.
+	  if sort "$nlist" | uniq > "$nlist"T; then
+	    mv -f "$nlist"T "$nlist"
+	    wcout=`wc "$nlist" 2>/dev/null`
+	    count=`echo "X$wcout" | $Xsed -e 's/^[ 	]*\([0-9][0-9]*\).*$/\1/'`
+	    (test "$count" -ge 0) 2>/dev/null || count=-1
+	  else
+	    $rm "$nlist"T
+	    count=-1
+	  fi
+
+	  case "$dlsyms" in
+	  "") ;;
+	  *.c)
+	    $echo > "$objdir/$dlsyms" "\
+/* $dlsyms - symbol resolution table for \`$output' dlsym emulation. */
+/* Generated by $PROGRAM - GNU $PACKAGE $VERSION */
+
+#ifdef __cplusplus
+extern \"C\" {
+#endif
+
+/* Prevent the only kind of declaration conflicts we can make. */
+#define dld_preloaded_symbol_count some_other_symbol
+#define dld_preloaded_symbols some_other_symbol
+
+/* External symbol declarations for the compiler. */\
+"
+
+	    if test -f "$nlist"; then
+	      sed -e 's/^.* \(.*\)$/extern char \1;/' < "$nlist" >> "$objdir/$dlsyms"
+	    else
+	      echo '/* NONE */' >> "$objdir/$dlsyms"
+	    fi
+
+	    $echo >> "$objdir/$dlsyms" "\
+
+#undef dld_preloaded_symbol_count
+#undef dld_preloaded_symbols
+
+#if defined (__STDC__) && __STDC__
+# define __ptr_t void *
+#else
+# define __ptr_t char *
+#endif
+
+/* The number of symbols in dld_preloaded_symbols, -1 if unsorted. */
+int dld_preloaded_symbol_count = $count;
+
+/* The mapping between symbol names and symbols. */
+struct {
+  char *name;
+  __ptr_t address;
+}
+dld_preloaded_symbols[] =
+{\
+"
+
+	    if test -f "$nlist"; then
+	      sed 's/^\(.*\) \(.*\)$/  {"\1", (__ptr_t) \&\2},/' < "$nlist" >> "$objdir/$dlsyms"
+	    fi
+
+	    $echo >> "$objdir/$dlsyms" "\
+  {0, (__ptr_t) 0}
+};
+
+#ifdef __cplusplus
+}
+#endif\
+"
+	    ;;
+
+	  *)
+	    $echo "$modename: unknown suffix for \`$dlsyms'" 1>&2
+	    exit 1
+	    ;;
+	  esac
+        fi
+
+        # Now compile the dynamic symbol file.
+        $show "(cd $objdir && $CC -c$no_builtin_flag \"$dlsyms\")"
+        $run eval '(cd $objdir && $CC -c$no_builtin_flag "$dlsyms")' || exit $?
+
+        # Transform the symbol file into the correct name.
+        compile_command=`$echo "X$compile_command" | $Xsed -e "s%@SYMFILE@%$objdir/${output}S.o%"`
+        finalize_command=`$echo "X$finalize_command" | $Xsed -e "s%@SYMFILE@%$objdir/${output}S.o%"`
+      elif test "$export_dynamic" != yes; then
+        test -n "$dlfiles$dlprefiles" && $echo "$modename: warning: \`-dlopen' and \`-dlpreopen' are ignored without \`-export-dynamic'" 1>&2
+      else
+        # We keep going just in case the user didn't refer to
+        # dld_preloaded_symbols.  The linker will fail if global_symbol_pipe
+        # really was required.
+        $echo "$modename: not configured to extract global symbols from dlpreopened files" 1>&2
+
+        # Nullify the symbol file.
+        compile_command=`$echo "X$compile_command" | $Xsed -e "s% @SYMFILE@%%"`
+        finalize_command=`$echo "X$finalize_command" | $Xsed -e "s% @SYMFILE@%%"`
+      fi
+
+      if test -z "$link_against_libtool_libs" || test "$build_libtool_libs" != yes; then
+        # Replace the output file specification.
+        compile_command=`$echo "X$compile_command" | $Xsed -e 's%@OUTPUT@%'"$output"'%g'`
+        finalize_command=`$echo "X$finalize_command" | $Xsed -e 's%@OUTPUT@%'"$output"'%g'`
+
+        # We have no uninstalled library dependencies, so finalize right now.
+        $show "$compile_command"
+        $run eval "$compile_command"
+        exit $?
+      fi
+
+      # Replace the output file specification.
+      compile_command=`$echo "X$compile_command" | $Xsed -e 's%@OUTPUT@%'"$objdir/$output"'%g'`
+      finalize_command=`$echo "X$finalize_command" | $Xsed -e 's%@OUTPUT@%'"$objdir/$output"'T%g'`
+
+      # Create the binary in the object directory, then wrap it.
+      if test -d $objdir; then :
+      else
+        $show "$mkdir $objdir"
+	$run $mkdir $objdir
+	status=$?
+	if test $status -eq 0 || test -d $objdir; then :
+	else
+	  exit $status
+	fi
+      fi
+
+      if test -n "$shlibpath_var"; then
+        # We should set the shlibpath_var
+        rpath=
+        for dir in $temp_rpath; do
+          case "$dir" in
+          /* | [A-Za-z]:\\*)
+            # Absolute path.
+            rpath="$rpath$dir:"
+            ;;
+          *)
+            # Relative path: add a thisdir entry.
+            rpath="$rpath\$thisdir/$dir:"
+            ;;
+          esac
+        done
+        temp_rpath="$rpath"
+      fi
+
+      # Delete the old output file.
+      $run $rm $output
+
+      if test -n "$compile_shlibpath"; then
+        compile_command="$shlibpath_var=\"$compile_shlibpath\$$shlibpath_var\" $compile_command"
+      fi
+      if test -n "$finalize_shlibpath"; then
+        finalize_command="$shlibpath_var=\"$finalize_shlibpath\$$shlibpath_var\" $finalize_command"
+      fi
+
+      if test -n "$runpath_var" && test -n "$perm_rpath"; then
+        # We should set the runpath_var.
+        rpath=
+        for dir in $perm_rpath; do
+          rpath="$rpath$dir:"
+        done
+        compile_command="$runpath_var=\"$rpath\$$runpath_var\" $compile_command"
+        finalize_command="$runpath_var=\"$rpath\$$runpath_var\" $finalize_command"
+      fi
+
+      case "$hardcode_action" in
+      relink)
+        # AGH! Flame the AIX and HP-UX people for me, will ya?
+        $echo "$modename: warning: using a buggy system linker" 1>&2
+        $echo "$modename: relinking will be required before \`$output' can be installed" 1>&2
+        ;;
+      esac
+
+      $show "$compile_command"
+      $run eval "$compile_command" || exit $?
+
+      # Now create the wrapper script.
+      $show "creating $output"
+
+      # Quote the finalize command for shipping.
+      finalize_command=`$echo "X$finalize_command" | $Xsed -e "$sed_quote_subst"`
+
+      # Quote $echo for shipping.
+      qecho=`$echo "X$echo" | $Xsed -e "$sed_quote_subst"`
+
+      # Only actually do things if our run command is non-null.
+      if test -z "$run"; then
+        $rm $output
+        trap "$rm $output; exit 1" 1 2 15
+
+        $echo > $output "\
+#! /bin/sh
+
+# $output - temporary wrapper script for $objdir/$output
+# Generated by ltmain.sh - GNU $PACKAGE $VERSION
+#
+# The $output program cannot be directly executed until all the libtool
+# libraries that it depends on are installed.
+#
+# This wrapper script should never be moved out of \``pwd`'.
+# If it is, it will not operate correctly.
+
+# Sed substitution that helps us do robust quoting.  It backslashifies
+# metacharacters that are still active within double-quoted strings.
+Xsed='sed -e s/^X//'
+sed_quote_subst='$sed_quote_subst'
+
+# The HP-UX ksh and POSIX shell print the target directory to stdout
+# if CDPATH is set.
+if test \"\${CDPATH+set}\" = set; then CDPATH=; export CDPATH; fi
+
+# This environment variable determines our operation mode.
+if test \"\$libtool_install_magic\" = \"$magic\"; then
+  # install mode needs the following variables:
+  link_against_libtool_libs='$link_against_libtool_libs'
+  finalize_command=\"$finalize_command\"
+else
+  # When we are sourced in execute mode, \$file and \$echo are already set.
+  if test \"\$libtool_execute_magic\" = \"$magic\"; then :
+  else
+    echo=\"$qecho\"
+    file=\"\$0\"
+  fi\
+"
+        $echo >> $output "\
+
+  # Find the directory that this script lives in.
+  thisdir=\`\$echo \"X\$file\" | \$Xsed -e 's%/[^/]*$%%'\`
+  test \"x\$thisdir\" = \"x\$file\" && thisdir=.
+
+  # Follow symbolic links until we get to the real thisdir.
+  file=\`ls -ld \"\$file\" | sed -n 's/.*-> //p'\`
+  while test -n \"\$file\"; do
+    destdir=\`\$echo \"X\$file\" | \$Xsed -e 's%/[^/]*\$%%'\`
+
+    # If there was a directory component, then change thisdir.
+    if test \"x\$destdir\" != \"x\$file\"; then
+      case \"\$destdir\" in
+      /* | [A-Za-z]:\\*) thisdir=\"\$destdir\" ;;
+      *) thisdir=\"\$thisdir/\$destdir\" ;;
+      esac
+    fi
+
+    file=\`\$echo \"X\$file\" | \$Xsed -e 's%^.*/%%'\`
+    file=\`ls -ld \"\$thisdir/\$file\" | sed -n 's/.*-> //p'\`
+  done
+
+  # Try to get the absolute directory name.
+  absdir=\`cd \"\$thisdir\" && pwd\`
+  test -n \"\$absdir\" && thisdir=\"\$absdir\"
+
+  progdir=\"\$thisdir/$objdir\"
+  program='$output'
+
+  if test -f \"\$progdir/\$program\"; then"
+
+        # Export our shlibpath_var if we have one.
+        if test -n "$shlibpath_var" && test -n "$temp_rpath"; then
+          $echo >> $output "\
+    # Add our own library path to $shlibpath_var
+    $shlibpath_var=\"$temp_rpath\$$shlibpath_var\"
+
+    # Some systems cannot cope with colon-terminated $shlibpath_var
+    $shlibpath_var=\`\$echo \"X\$$shlibpath_var\" | \$Xsed -e 's/:*\$//'\`
+
+    export $shlibpath_var
+"
+        fi
+
+        $echo >> $output "\
+    if test \"\$libtool_execute_magic\" != \"$magic\"; then
+      # Run the actual program with our arguments.
+
+      # Export the path to the program.
+      PATH=\"\$progdir:\$PATH\"
+      export PATH
+
+      exec \$program \${1+\"\$@\"}
+
+      \$echo \"\$0: cannot exec \$program \${1+\"\$@\"}\"
+      exit 1
+    fi
+  else
+    # The program doesn't exist.
+    \$echo \"\$0: error: \$progdir/\$program does not exist\" 1>&2
+    \$echo \"This script is just a wrapper for \$program.\" 1>&2
+    echo \"See the $PACKAGE documentation for more information.\" 1>&2
+    exit 1
+  fi
+fi\
+"
+        chmod +x $output
+      fi
+      exit 0
+      ;;
+    esac
+
+    # See if we need to build an old-fashioned archive.
+    if test "$build_old_libs" = "yes"; then
+      # Transform .lo files to .o files.
+      oldobjs="$objs"`$echo "X$libobjs " | $Xsed -e 's/[^   ]*\.a //g' -e 's/\.lo /.o /g' -e 's/ $//g'`
+
+      # Do each command in the archive commands.
+      if test -n "$old_archive_from_new_cmds" && test "$build_libtool_libs" = yes; then
+	eval cmds=\"$old_archive_from_new_cmds\"
+      else
+	eval cmds=\"$old_archive_cmds\"
+      fi
+      IFS="${IFS= 	}"; save_ifs="$IFS"; IFS=';'
+      for cmd in $cmds; do
+        IFS="$save_ifs"
+        $show "$cmd"
+        $run eval "$cmd" || exit $?
+      done
+      IFS="$save_ifs"
+    fi
+
+    # Now create the libtool archive.
+    case "$output" in
+    *.la)
+      old_library=
+      test "$build_old_libs" = yes && old_library="$libname.a"
+
+      $show "creating $output"
+
+      # Only create the output if not a dry run.
+      if test -z "$run"; then
+        $echo > $output "\
+# $output - a libtool library file
+# Generated by ltmain.sh - GNU $PACKAGE $VERSION
+
+# The name that we can dlopen(3).
+dlname='$dlname'
+
+# Names of this library.
+library_names='$library_names'
+
+# The name of the static archive.
+old_library='$old_library'
+
+# Libraries that this one depends upon.
+dependency_libs='$dependency_libs'
+
+# Version information for $libname.
+current=$current
+age=$age
+revision=$revision
+
+# Directory that this library needs to be installed in:
+libdir='$install_libdir'\
+"
+      fi
+
+      # Do a symbolic link so that the libtool archive can be found in
+      # LD_LIBRARY_PATH before the program is installed.
+      $show "(cd $objdir && $LN_S ../$output $output)"
+      $run eval "(cd $objdir && $LN_S ../$output $output)" || exit 1
+      ;;
+    esac
+    exit 0
+    ;;
+
+  # libtool install mode
+  install)
+    modename="$modename: install"
+
+    # There may be an optional /bin/sh argument at the beginning of
+    # install_prog (especially on Windows NT).
+    if test "$nonopt" = "$SHELL"; then
+      # Aesthetically quote it.
+      arg=`$echo "X$nonopt" | $Xsed -e "$sed_quote_subst"`
+      case "$arg" in
+      *[\[\~\#\^\&\*\(\)\{\}\|\;\<\>\?\'\ \	]*|*]*)
+	arg="\"$arg\""
+	;;
+      esac
+      install_prog="$arg "
+      arg="$1"
+      shift
+    else
+      install_prog=
+      arg="$nonopt"
+    fi
+
+    # The real first argument should be the name of the installation program.
+    # Aesthetically quote it.
+    arg=`$echo "X$arg" | $Xsed -e "$sed_quote_subst"`
+    case "$arg" in
+    *[\[\~\#\^\&\*\(\)\{\}\|\;\<\>\?\'\ \	]*|*]*)
+      arg="\"$arg\""
+      ;;
+    esac
+    install_prog="$install_prog$arg"
+
+    # We need to accept at least all the BSD install flags.
+    dest=
+    files=
+    opts=
+    prev=
+    install_type=
+    isdir=
+    stripme=
+    for arg
+    do
+      if test -n "$dest"; then
+        files="$files $dest"
+        dest="$arg"
+        continue
+      fi
+
+      case "$arg" in
+      -d) isdir=yes ;;
+      -f) prev="-f" ;;
+      -g) prev="-g" ;;
+      -m) prev="-m" ;;
+      -o) prev="-o" ;;
+      -s)
+        stripme=" -s"
+        continue
+        ;;
+      -*) ;;
+
+      *)
+        # If the previous option needed an argument, then skip it.
+        if test -n "$prev"; then
+          prev=
+        else
+          dest="$arg"
+          continue
+        fi
+        ;;
+      esac
+
+      # Aesthetically quote the argument.
+      arg=`$echo "X$arg" | $Xsed -e "$sed_quote_subst"`
+      case "$arg" in
+      *[\[\~\#\^\&\*\(\)\{\}\|\;\<\>\?\'\ \	]*|*]*)
+	arg="\"$arg\""
+	;;
+      esac
+      install_prog="$install_prog $arg"
+    done
+
+    if test -z "$install_prog"; then
+      $echo "$modename: you must specify an install program" 1>&2
+      $echo "$help" 1>&2
+      exit 1
+    fi
+
+    if test -n "$prev"; then
+      $echo "$modename: the \`$prev' option requires an argument" 1>&2
+      $echo "$help" 1>&2
+      exit 1
+    fi
+
+    if test -z "$files"; then
+      if test -z "$dest"; then
+        $echo "$modename: no file or destination specified" 1>&2
+      else
+        $echo "$modename: you must specify a destination" 1>&2
+      fi
+      $echo "$help" 1>&2
+      exit 1
+    fi
+
+    # Strip any trailing slash from the destination.
+    dest=`$echo "X$dest" | $Xsed -e 's%/$%%'`
+
+    # Check to see that the destination is a directory.
+    test -d "$dest" && isdir=yes
+    if test -n "$isdir"; then
+      destdir="$dest"
+      destname=
+    else
+      destdir=`$echo "X$dest" | $Xsed -e 's%/[^/]*$%%'`
+      test "X$destdir" = "X$dest" && destdir=.
+      destname=`$echo "X$dest" | $Xsed -e 's%^.*/%%'`
+
+      # Not a directory, so check to see that there is only one file specified.
+      set dummy $files
+      if test $# -gt 2; then
+        $echo "$modename: \`$dest' is not a directory" 1>&2
+        $echo "$help" 1>&2
+        exit 1
+      fi
+    fi
+    case "$destdir" in
+    /* | [A-Za-z]:\\*) ;;
+    *)
+      for file in $files; do
+        case "$file" in
+        *.lo) ;;
+        *)
+          $echo "$modename: \`$destdir' must be an absolute directory name" 1>&2
+          $echo "$help" 1>&2
+          exit 1
+          ;;
+        esac
+      done
+      ;;
+    esac
+
+    # This variable tells wrapper scripts just to set variables rather
+    # than running their programs.
+    libtool_install_magic="$magic"
+
+    staticlibs=
+    future_libdirs=
+    current_libdirs=
+    for file in $files; do
+
+      # Do each installation.
+      case "$file" in
+      *.a)
+        # Do the static libraries later.
+        staticlibs="$staticlibs $file"
+        ;;
+
+      *.la)
+        # Check to see that this really is a libtool archive.
+        if (sed -e '2q' $file | egrep '^# Generated by ltmain\.sh') >/dev/null 2>&1; then :
+        else
+          $echo "$modename: \`$file' is not a valid libtool archive" 1>&2
+          $echo "$help" 1>&2
+          exit 1
+        fi
+
+        library_names=
+        old_library=
+        # If there is no directory component, then add one.
+        case "$file" in
+        */* | *\\*) . $file ;;
+        *) . ./$file ;;
+        esac
+
+        # Add the libdir to current_libdirs if it is the destination.
+        if test "X$destdir" = "X$libdir"; then
+          case "$current_libdirs " in
+          *" $libdir "*) ;;
+          *) current_libdirs="$current_libdirs $libdir" ;;
+          esac
+        else
+          # Note the libdir as a future libdir.
+          case "$future_libdirs " in
+          *" $libdir "*) ;;
+          *) future_libdirs="$future_libdirs $libdir" ;;
+          esac
+        fi
+
+        dir="`$echo "X$file" | $Xsed -e 's%/[^/]*$%%'`/"
+        test "X$dir" = "X$file/" && dir=
+        dir="$dir$objdir"
+
+        # See the names of the shared library.
+        set dummy $library_names
+        if test -n "$2"; then
+          realname="$2"
+          shift
+          shift
+
+          # Install the shared library and build the symlinks.
+          $show "$install_prog $dir/$realname $destdir/$realname"
+          $run eval "$install_prog $dir/$realname $destdir/$realname" || exit $?
+          test "X$dlname" = "X$realname" && dlname=
+
+          if test $# -gt 0; then
+            # Delete the old symlinks.
+            rmcmd="$rm"
+            for linkname
+            do
+              rmcmd="$rmcmd $destdir/$linkname"
+            done
+            $show "$rmcmd"
+            $run $rmcmd
+
+            # ... and create new ones.
+            for linkname
+            do
+              test "X$dlname" = "X$linkname" && dlname=
+              $show "(cd $destdir && $LN_S $realname $linkname)"
+              $run eval "(cd $destdir && $LN_S $realname $linkname)"
+            done
+          fi
+
+          if test -n "$dlname"; then
+            # Install the dynamically-loadable library.
+            $show "$install_prog $dir/$dlname $destdir/$dlname"
+            $run eval "$install_prog $dir/$dlname $destdir/$dlname" || exit $?
+          fi
+
+          # Do each command in the postinstall commands.
+          lib="$destdir/$realname"
+          eval cmds=\"$postinstall_cmds\"
+          IFS="${IFS= 	}"; save_ifs="$IFS"; IFS=';'
+          for cmd in $cmds; do
+            IFS="$save_ifs"
+            $show "$cmd"
+            $run eval "$cmd" || exit $?
+          done
+          IFS="$save_ifs"
+        fi
+
+        # Install the pseudo-library for information purposes.
+        name=`$echo "X$file" | $Xsed -e 's%^.*/%%'`
+        $show "$install_prog $file $destdir/$name"
+        $run eval "$install_prog $file $destdir/$name" || exit $?
+
+        # Maybe install the static library, too.
+        test -n "$old_library" && staticlibs="$staticlibs $dir/$old_library"
+        ;;
+
+      *.lo)
+        # Install (i.e. copy) a libtool object.
+
+        # Figure out destination file name, if it wasn't already specified.
+        if test -n "$destname"; then
+          destfile="$destdir/$destname"
+        else
+          destfile=`$echo "X$file" | $Xsed -e 's%^.*/%%'`
+          destfile="$destdir/$destfile"
+        fi
+
+        # Deduce the name of the destination old-style object file.
+        case "$destfile" in
+        *.lo)
+          staticdest=`$echo "X$destfile" | $Xsed -e 's/\.lo$/\.o/'`
+          ;;
+        *.o)
+          staticdest="$destfile"
+          destfile=
+          ;;
+        *)
+          $echo "$modename: cannot copy a libtool object to \`$destfile'" 1>&2
+          $echo "$help" 1>&2
+          exit 1
+          ;;
+        esac
+
+        # Install the libtool object if requested.
+        if test -n "$destfile"; then
+          $show "$install_prog $file $destfile"
+          $run eval "$install_prog $file $destfile" || exit $?
+        fi
+
+        # Install the old object if enabled.
+        if test "$build_old_libs" = yes; then
+          # Deduce the name of the old-style object file.
+          staticobj=`$echo "X$file" | $Xsed -e 's/\.lo$/\.o/'`
+
+          $show "$install_prog $staticobj $staticdest"
+          $run eval "$install_prog \$staticobj \$staticdest" || exit $?
+        fi
+        exit 0
+        ;;
+
+      *)
+        # Do a test to see if this is really a libtool program.
+        if (sed -e '4q' $file | egrep '^# Generated by ltmain\.sh') >/dev/null 2>&1; then
+          link_against_libtool_libs=
+          finalize_command=
+
+          # If there is no directory component, then add one.
+          case "$file" in
+          */* | *\\*) . $file ;;
+          *) . ./$file ;;
+          esac
+
+          # Check the variables that should have been set.
+          if test -z "$link_against_libtool_libs" || test -z "$finalize_command"; then
+            $echo "$modename: invalid libtool wrapper script \`$file'" 1>&2
+            exit 1
+          fi
+
+          finalize=yes
+          for lib in $link_against_libtool_libs; do
+            # Check to see that each library is installed.
+            libdir=
+            if test -f "$lib"; then
+              # If there is no directory component, then add one.
+              case "$lib" in
+              */* | *\\*) . $lib ;;
+              *) . ./$lib ;;
+              esac
+            fi
+            libfile="$libdir/`$echo "X$lib" | $Xsed -e 's%^.*/%%g'`"
+            if test -z "$libdir"; then
+              $echo "$modename: warning: \`$lib' contains no -rpath information" 1>&2
+            elif test -f "$libfile"; then :
+            else
+              $echo "$modename: warning: \`$lib' has not been installed in \`$libdir'" 1>&2
+              finalize=no
+            fi
+          done
+
+          if test "$hardcode_action" = relink; then
+            if test "$finalize" = yes; then
+              $echo "$modename: warning: relinking \`$file' on behalf of your buggy system linker" 1>&2
+              $show "$finalize_command"
+              if $run eval "$finalize_command"; then :
+              else
+                $echo "$modename: error: relink \`$file' with the above command before installing it" 1>&2
+                continue
+              fi
+              file="$objdir/$file"T
+            else
+              $echo "$modename: warning: cannot relink \`$file' on behalf of your buggy system linker" 1>&2
+            fi
+          else
+            # Install the binary that we compiled earlier.
+	    file=`$echo "X$file" | $Xsed -e "s%\([^/]*\)$%$objdir/\1%"`
+          fi
+        fi
+
+        $show "$install_prog$stripme $file $dest"
+        $run eval "$install_prog\$stripme \$file \$dest" || exit $?
+        ;;
+      esac
+    done
+
+    for file in $staticlibs; do
+      name=`$echo "X$file" | $Xsed -e 's%^.*/%%'`
+
+      # Set up the ranlib parameters.
+      oldlib="$destdir/$name"
+
+      $show "$install_prog $file $oldlib"
+      $run eval "$install_prog \$file \$oldlib" || exit $?
+
+      # Do each command in the postinstall commands.
+      eval cmds=\"$old_postinstall_cmds\"
+      IFS="${IFS= 	}"; save_ifs="$IFS"; IFS=';'
+      for cmd in $cmds; do
+        IFS="$save_ifs"
+        $show "$cmd"
+        $run eval "$cmd" || exit $?
+      done
+      IFS="$save_ifs"
+    done
+
+    if test -n "$future_libdirs"; then
+      $echo "$modename: warning: remember to run \`$progname --finish$future_libdirs'" 1>&2
+    fi
+
+    if test -n "$current_libdirs"; then
+      # Maybe just do a dry run.
+      test -n "$run" && current_libdirs=" -n$current_libdirs"
+      exec $SHELL $0 --finish$current_libdirs
+      exit 1
+    fi
+
+    exit 0
+    ;;
+
+  # libtool finish mode
+  finish)
+    modename="$modename: finish"
+    libdirs="$nonopt"
+
+    if test -n "$finish_cmds$finish_eval" && test -n "$libdirs"; then
+      for dir
+      do
+        libdirs="$libdirs $dir"
+      done
+
+      for libdir in $libdirs; do
+	if test -n "$finish_cmds"; then
+	  # Do each command in the finish commands.
+	  eval cmds=\"$finish_cmds\"
+          IFS="${IFS= 	}"; save_ifs="$IFS"; IFS=';'
+          for cmd in $cmds; do
+            IFS="$save_ifs"
+            $show "$cmd"
+            $run eval "$cmd"
+          done
+          IFS="$save_ifs"
+	fi
+	if test -n "$finish_eval"; then
+	  # Do the single finish_eval.
+	  eval cmds=\"$finish_eval\"
+	  $run eval "$cmds"
+	fi
+      done
+    fi
+
+    echo "------------------------------------------------------------------------------"
+    echo "Libraries have been installed in:"
+    for libdir in $libdirs; do
+      echo "   $libdir"
+    done
+    echo
+    echo "To link against installed libraries in a given directory, LIBDIR,"
+    echo "you must use the \`-LLIBDIR' flag during linking."
+    echo
+    echo " You will also need to do one of the following:"
+    if test -n "$shlibpath_var"; then
+      echo "   - add LIBDIR to the \`$shlibpath_var' environment variable"
+      echo "     during execution"
+    fi
+    if test -n "$runpath_var"; then
+      echo "   - add LIBDIR to the \`$runpath_var' environment variable"
+      echo "     during linking"
+    fi
+    if test -n "$hardcode_libdir_flag_spec"; then
+      libdir=LIBDIR
+      eval flag=\"$hardcode_libdir_flag_spec\"
+
+      echo "   - use the \`$flag' linker flag"
+    fi
+    if test -f /etc/ld.so.conf; then
+      echo "   - have your system administrator add LIBDIR to \`/etc/ld.so.conf'"
+    fi
+    echo
+    echo "See any operating system documentation about shared libraries for"
+    echo "more information, such as the ld(1) and ld.so(8) manual pages."
+    echo "------------------------------------------------------------------------------"
+    exit 0
+    ;;
+
+  # libtool execute mode
+  execute)
+    modename="$modename: execute"
+
+    # The first argument is the command name.
+    cmd="$nonopt"
+    if test -z "$cmd"; then
+      $echo "$modename: you must specify a COMMAND" 1>&2
+      $echo "$help"
+      exit 1
+    fi
+
+    # Handle -dlopen flags immediately.
+    for file in $execute_dlfiles; do
+      if test -f "$file"; then :
+      else
+	$echo "$modename: \`$file' is not a file" 1>&2
+	$echo "$help" 1>&2
+	exit 1
+      fi
+
+      dir=
+      case "$file" in
+      *.la)
+        # Check to see that this really is a libtool archive.
+        if (sed -e '2q' $file | egrep '^# Generated by ltmain\.sh') >/dev/null 2>&1; then :
+        else
+          $echo "$modename: \`$lib' is not a valid libtool archive" 1>&2
+          $echo "$help" 1>&2
+          exit 1
+        fi
+
+	# Read the libtool library.
+	dlname=
+	library_names=
+
+        # If there is no directory component, then add one.
+	case "$file" in
+	*/* | *\\*) . $file ;;
+        *) . ./$file ;;
+	esac
+
+	# Skip this library if it cannot be dlopened.
+	if test -z "$dlname"; then
+	  # Warn if it was a shared library.
+	  test -n "$library_names" && $echo "$modename: warning: \`$file' was not linked with \`-export-dynamic'"
+	  continue
+	fi
+
+	dir=`$echo "X$file" | $Xsed -e 's%/[^/]*$%%'`
+	test "X$dir" = "X$file" && dir=.
+
+	if test -f "$dir/$objdir/$dlname"; then
+	  dir="$dir/$objdir"
+	else
+	  $echo "$modename: cannot find \`$dlname' in \`$dir' or \`$dir/$objdir'" 1>&2
+	  exit 1
+	fi
+	;;
+
+      *.lo)
+	# Just add the directory containing the .lo file.
+	dir=`$echo "X$file" | $Xsed -e 's%/[^/]*$%%'`
+	test "X$dir" = "X$file" && dir=.
+	;;
+
+      *)
+	$echo "$modename: warning \`-dlopen' is ignored for non-libtool libraries and objects" 1>&2
+        continue
+	;;
+      esac
+
+      # Get the absolute pathname.
+      absdir=`cd "$dir" && pwd`
+      test -n "$absdir" && dir="$absdir"
+
+      # Now add the directory to shlibpath_var.
+      if eval "test -z \"\$$shlibpath_var\""; then
+	eval "$shlibpath_var=\"\$dir\""
+      else
+	eval "$shlibpath_var=\"\$dir:\$$shlibpath_var\""
+      fi
+    done
+
+    # This variable tells wrapper scripts just to set shlibpath_var
+    # rather than running their programs.
+    libtool_execute_magic="$magic"
+
+    # Check if any of the arguments is a wrapper script.
+    args=
+    for file
+    do
+      case "$file" in
+      -*) ;;
+      *)
+        # Do a test to see if this is really a libtool program.
+        if (sed -e '4q' $file | egrep '^# Generated by ltmain\.sh') >/dev/null 2>&1; then
+	  # If there is no directory component, then add one.
+	  case "$file" in
+	  */* | *\\*) . $file ;;
+	  *) . ./$file ;;
+	  esac
+
+	  # Transform arg to wrapped name.
+	  file="$progdir/$program"
+	fi
+        ;;
+      esac
+      # Quote arguments (to preserve shell metacharacters).
+      file=`$echo "X$file" | $Xsed -e "$sed_quote_subst"`
+      args="$args \"$file\""
+    done
+
+    if test -z "$run"; then
+      # Export the shlibpath_var.
+      eval "export $shlibpath_var"
+
+      # Now actually exec the command.
+      eval "exec \$cmd$args"
+
+      $echo "$modename: cannot exec \$cmd$args"
+      exit 1
+    else
+      # Display what would be done.
+      eval "\$echo \"\$shlibpath_var=\$$shlibpath_var\""
+      $echo "export $shlibpath_var"
+      $echo "$cmd$args"
+      exit 0
+    fi
+    ;;
+
+  # libtool uninstall mode
+  uninstall)
+    modename="$modename: uninstall"
+    rm="$nonopt"
+    files=
+
+    for arg
+    do
+      case "$arg" in
+      -*) rm="$rm $arg" ;;
+      *) files="$files $arg" ;;
+      esac
+    done
+
+    if test -z "$rm"; then
+      $echo "$modename: you must specify an RM program" 1>&2
+      $echo "$help" 1>&2
+      exit 1
+    fi
+
+    for file in $files; do
+      dir=`$echo "X$file" | $Xsed -e 's%/[^/]*$%%'`
+      test "X$dir" = "X$file" && dir=.
+      name=`$echo "X$file" | $Xsed -e 's%^.*/%%'`
+
+      rmfiles="$file"
+
+      case "$name" in
+      *.la)
+        # Possibly a libtool archive, so verify it.
+        if (sed -e '2q' $file | egrep '^# Generated by ltmain\.sh') >/dev/null 2>&1; then
+          . $dir/$name
+
+          # Delete the libtool libraries and symlinks.
+          for n in $library_names; do
+            rmfiles="$rmfiles $dir/$n"
+            test "X$n" = "X$dlname" && dlname=
+          done
+          test -n "$dlname" && rmfiles="$rmfiles $dir/$dlname"
+          test -n "$old_library" && rmfiles="$rmfiles $dir/$old_library"
+
+	  $show "$rm $rmfiles"
+	  $run $rm $rmfiles
+
+	  if test -n "$library_names"; then
+	    # Do each command in the postuninstall commands.
+	    eval cmds=\"$postuninstall_cmds\"
+	    IFS="${IFS= 	}"; save_ifs="$IFS"; IFS=';'
+	    for cmd in $cmds; do
+	      IFS="$save_ifs"
+	      $show "$cmd"
+	      $run eval "$cmd"
+	    done
+	    IFS="$save_ifs"
+	  fi
+
+          if test -n "$old_library"; then
+	    # Do each command in the old_postuninstall commands.
+	    eval cmds=\"$old_postuninstall_cmds\"
+	    IFS="${IFS= 	}"; save_ifs="$IFS"; IFS=';'
+	    for cmd in $cmds; do
+	      IFS="$save_ifs"
+	      $show "$cmd"
+	      $run eval "$cmd"
+	    done
+	    IFS="$save_ifs"
+	  fi
+
+          # FIXME: should reinstall the best remaining shared library.
+        fi
+        ;;
+
+      *.lo)
+        if test "$build_old_libs" = yes; then
+          oldobj=`$echo "X$name" | $Xsed -e 's/\.lo$/\.o/'`
+          rmfiles="$rmfiles $dir/$oldobj"
+        fi
+	$show "$rm $rmfiles"
+	$run $rm $rmfiles
+        ;;
+
+      *)
+      	$show "$rm $rmfiles"
+	$run $rm $rmfiles
+	;;
+      esac
+    done
+    exit 0
+    ;;
+
+  "")
+    $echo "$modename: you must specify a MODE" 1>&2
+    $echo "$generic_help" 1>&2
+    exit 1
+    ;;
+  esac
+
+  $echo "$modename: invalid operation mode \`$mode'" 1>&2
+  $echo "$generic_help" 1>&2
+  exit 1
+fi # test -z "$show_help"
+
+# We need to display help for each of the modes.
+case "$mode" in
+"") $echo \
+"Usage: $modename [OPTION]... [MODE-ARG]...
+
+Provide generalized library-building support services.
+
+-n, --dry-run         display commands without modifying any files
+    --features        display configuration information and exit
+    --finish          same as \`--mode=finish'
+    --help            display this help message and exit
+    --mode=MODE       use operation mode MODE [default=inferred from MODE-ARGS]
+    --quiet           same as \`--silent'
+    --silent          don't print informational messages
+    --version         print version information
+
+MODE must be one of the following:
+
+      compile         compile a source file into a libtool object
+      execute         automatically set library path, then run a program
+      finish          complete the installation of libtool libraries
+      install         install libraries or executables
+      link            create a library or an executable
+      uninstall       remove libraries from an installed directory
+
+MODE-ARGS vary depending on the MODE.  Try \`$modename --help --mode=MODE' for
+a more detailed description of MODE."
+  exit 0
+  ;;
+
+compile)
+  $echo \
+"Usage: $modename [OPTION]... --mode=compile COMPILE-COMMAND... SOURCEFILE
+
+Compile a source file into a libtool library object.
+
+COMPILE-COMMAND is a command to be used in creating a \`standard' object file
+from the given SOURCEFILE.
+
+The output file name is determined by removing the directory component from
+SOURCEFILE, then substituting the C source code suffix \`.c' with the
+library object suffix, \`.lo'."
+  ;;
+
+execute)
+  $echo \
+"Usage: $modename [OPTION]... --mode=execute COMMAND [ARGS]...
+
+Automatically set library path, then run a program.
+
+This mode accepts the following additional options:
+
+  -dlopen FILE      add the directory containing FILE to the library path
+
+This mode sets the library path environment variable according to \`-dlopen'
+flags.
+
+If any of the ARGS are libtool executable wrappers, then they are translated
+into their corresponding uninstalled binary, and any of their required library
+directories are added to the library path.
+
+Then, COMMAND is executed, with ARGS as arguments."
+  ;;
+
+finish)
+  $echo \
+"Usage: $modename [OPTION]... --mode=finish [LIBDIR]...
+
+Complete the installation of libtool libraries.
+
+Each LIBDIR is a directory that contains libtool libraries.
+
+The commands that this mode executes may require superuser privileges.  Use
+the \`--dry-run' option if you just want to see what would be executed."
+  ;;
+
+install)
+  $echo \
+"Usage: $modename [OPTION]... --mode=install INSTALL-COMMAND...
+
+Install executables or libraries.
+
+INSTALL-COMMAND is the installation command.  The first component should be
+either the \`install' or \`cp' program.
+
+The rest of the components are interpreted as arguments to that command (only
+BSD-compatible install options are recognized)."
+  ;;
+
+link)
+  $echo \
+"Usage: $modename [OPTION]... --mode=link LINK-COMMAND...
+
+Link object files or libraries together to form another library, or to
+create an executable program.
+
+LINK-COMMAND is a command using the C compiler that you would use to create
+a program from several object files.
+
+The following components of LINK-COMMAND are treated specially:
+
+  -all-static       do not do any dynamic linking at all
+  -dlopen FILE      \`-dlpreopen' FILE if it cannot be dlopened at runtime
+  -dlpreopen FILE   link in FILE and add its symbols to dld_preloaded_symbols
+  -export-dynamic   allow symbols from OUTPUT-FILE to be resolved with dlsym(3)
+  -LLIBDIR          search LIBDIR for required installed libraries
+  -lNAME            OUTPUT-FILE requires the installed library libNAME
+  -no-undefined     declare that a library does not refer to external symbols
+  -o OUTPUT-FILE    create OUTPUT-FILE from the specified objects
+  -release RELEASE  specify package release information
+  -rpath LIBDIR     the created library will eventually be installed in LIBDIR
+  -static           do not do any dynamic linking of libtool libraries
+  -version-info CURRENT[:REVISION[:AGE]]
+                    specify library version info [each variable defaults to 0]
+
+All other options (arguments beginning with \`-') are ignored.
+
+Every other argument is treated as a filename.  Files ending in \`.la' are
+treated as uninstalled libtool libraries, other files are standard or library
+object files.
+
+If the OUTPUT-FILE ends in \`.la', then a libtool library is created, only
+library objects (\`.lo' files) may be specified, and \`-rpath' is required.
+
+If OUTPUT-FILE ends in \`.a', then a standard library is created using \`ar'
+and \`ranlib'.
+
+If OUTPUT-FILE ends in \`.lo' or \`.o', then a reloadable object file is
+created, otherwise an executable program is created."
+  ;;
+
+uninstall)
+  $echo
+"Usage: $modename [OPTION]... --mode=uninstall RM [RM-OPTION]... FILE...
+
+Remove libraries from an installation directory.
+
+RM is the name of the program to use to delete files associated with each FILE
+(typically \`/bin/rm').  RM-OPTIONS are options (such as \`-f') to be passed
+to RM.
+
+If FILE is a libtool library, all the files associated with it are deleted.
+Otherwise, only FILE itself is deleted using RM."
+  ;;
+
+*)
+  $echo "$modename: invalid operation mode \`$mode'" 1>&2
+  $echo "$help" 1>&2
+  exit 1
+  ;;
+esac
+
+echo
+$echo "Try \`$modename --help' for more information about other modes."
+
+exit 0
+
+# Local Variables:
+# mode:shell-script
+# sh-indentation:2
+# End:
diff --git a/libjpegtwrp/makcjpeg.st b/libjpegtwrp/makcjpeg.st
new file mode 100644
index 000000000..fc72c8982
--- /dev/null
+++ b/libjpegtwrp/makcjpeg.st
@@ -0,0 +1,38 @@
+; Project file for Independent JPEG Group's software
+;
+; This project file is for Atari ST/STE/TT systems using Pure C or Turbo C.
+; Thanks to Frank Moehle (Frank.Moehle@arbi.informatik.uni-oldenburg.de),
+; Dr. B. Setzepfandt (bernd@gina.uni-muenster.de),
+; and Guido Vollbeding (guivol@esc.de).
+;
+; To use this file, rename it to cjpeg.prj.
+; If you are using Turbo C, change filenames beginning with "pc..." to "tc..."
+; Read installation instructions before trying to make the program!
+;
+;
+;      * * * Output file * * *
+cjpeg.ttp
+;
+; * * * COMPILER OPTIONS * * *  
+.C[-P]        ; absolute calls
+.C[-M]        ; and no string merging, folks
+.C[-w-cln]    ; no "constant is long" warnings
+.C[-w-par]    ; no "parameter xxxx unused"
+.C[-w-rch]    ; no "unreachable code"
+.C[-wsig]     ; warn if significant digits may be lost
+=
+; * * * * List of modules * * * * 
+pcstart.o
+cjpeg.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h,jversion.h)
+cdjpeg.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+rdswitch.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+rdppm.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+rdgif.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+rdtarga.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+rdbmp.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+rdrle.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+libjpeg.lib        ; built by libjpeg.prj
+pcfltlib.lib       ; floating point library
+; the float library can be omitted if you've turned off DCT_FLOAT_SUPPORTED
+pcstdlib.lib       ; standard library
+pcextlib.lib       ; extended library
diff --git a/libjpegtwrp/makdjpeg.st b/libjpegtwrp/makdjpeg.st
new file mode 100644
index 000000000..322672631
--- /dev/null
+++ b/libjpegtwrp/makdjpeg.st
@@ -0,0 +1,38 @@
+; Project file for Independent JPEG Group's software
+;
+; This project file is for Atari ST/STE/TT systems using Pure C or Turbo C.
+; Thanks to Frank Moehle (Frank.Moehle@arbi.informatik.uni-oldenburg.de),
+; Dr. B. Setzepfandt (bernd@gina.uni-muenster.de),
+; and Guido Vollbeding (guivol@esc.de).
+;
+; To use this file, rename it to djpeg.prj.
+; If you are using Turbo C, change filenames beginning with "pc..." to "tc..."
+; Read installation instructions before trying to make the program!
+;
+;
+;      * * * Output file * * *
+djpeg.ttp
+;
+; * * * COMPILER OPTIONS * * *  
+.C[-P]        ; absolute calls
+.C[-M]        ; and no string merging, folks
+.C[-w-cln]    ; no "constant is long" warnings
+.C[-w-par]    ; no "parameter xxxx unused"
+.C[-w-rch]    ; no "unreachable code"
+.C[-wsig]     ; warn if significant digits may be lost
+=
+; * * * * List of modules * * * * 
+pcstart.o
+djpeg.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h,jversion.h)
+cdjpeg.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+rdcolmap.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+wrppm.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+wrgif.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+wrtarga.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+wrbmp.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+wrrle.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+libjpeg.lib        ; built by libjpeg.prj
+pcfltlib.lib       ; floating point library
+; the float library can be omitted if you've turned off DCT_FLOAT_SUPPORTED
+pcstdlib.lib       ; standard library
+pcextlib.lib       ; extended library
diff --git a/libjpegtwrp/makeapps.ds b/libjpegtwrp/makeapps.ds
new file mode 100644
index 000000000..bedd038a3
--- /dev/null
+++ b/libjpegtwrp/makeapps.ds
@@ -0,0 +1,828 @@
+# Microsoft Developer Studio Generated NMAKE File, Format Version 4.20
+# ** DO NOT EDIT **
+
+# TARGTYPE "Win32 (x86) Console Application" 0x0103
+
+!IF "$(CFG)" == ""
+CFG=cjpeg - Win32
+!MESSAGE No configuration specified.  Defaulting to cjpeg - Win32.
+!ENDIF 
+
+!IF "$(CFG)" != "cjpeg - Win32" && "$(CFG)" != "djpeg - Win32" &&\
+ "$(CFG)" != "jpegtran - Win32" && "$(CFG)" != "rdjpgcom - Win32" &&\
+ "$(CFG)" != "wrjpgcom - Win32"
+!MESSAGE Invalid configuration "$(CFG)" specified.
+!MESSAGE You can specify a configuration when running NMAKE on this makefile
+!MESSAGE by defining the macro CFG on the command line.  For example:
+!MESSAGE 
+!MESSAGE NMAKE /f "apps.mak" CFG="cjpeg - Win32"
+!MESSAGE 
+!MESSAGE Possible choices for configuration are:
+!MESSAGE 
+!MESSAGE "cjpeg - Win32" (based on "Win32 (x86) Console Application")
+!MESSAGE "djpeg - Win32" (based on "Win32 (x86) Console Application")
+!MESSAGE "jpegtran - Win32" (based on "Win32 (x86) Console Application")
+!MESSAGE "rdjpgcom - Win32" (based on "Win32 (x86) Console Application")
+!MESSAGE "wrjpgcom - Win32" (based on "Win32 (x86) Console Application")
+!MESSAGE 
+!ERROR An invalid configuration is specified.
+!ENDIF 
+
+!IF "$(OS)" == "Windows_NT"
+NULL=
+!ELSE 
+NULL=nul
+!ENDIF 
+################################################################################
+# Begin Project
+# PROP Target_Last_Scanned "cjpeg - Win32"
+CPP=cl.exe
+RSC=rc.exe
+
+!IF  "$(CFG)" == "cjpeg - Win32"
+
+# PROP BASE Use_MFC 0
+# PROP BASE Use_Debug_Libraries 0
+# PROP BASE Output_Dir "cjpeg\Release"
+# PROP BASE Intermediate_Dir "cjpeg\Release"
+# PROP BASE Target_Dir "cjpeg"
+# PROP Use_MFC 0
+# PROP Use_Debug_Libraries 0
+# PROP Output_Dir "cjpeg\Release"
+# PROP Intermediate_Dir "cjpeg\Release"
+# PROP Target_Dir "cjpeg"
+OUTDIR=.\cjpeg\Release
+INTDIR=.\cjpeg\Release
+
+ALL : "$(OUTDIR)\cjpeg.exe"
+
+CLEAN : 
+	-@erase "$(INTDIR)\cjpeg.obj"
+	-@erase "$(INTDIR)\rdppm.obj"
+	-@erase "$(INTDIR)\rdgif.obj"
+	-@erase "$(INTDIR)\rdtarga.obj"
+	-@erase "$(INTDIR)\rdrle.obj"
+	-@erase "$(INTDIR)\rdbmp.obj"
+	-@erase "$(INTDIR)\rdswitch.obj"
+	-@erase "$(INTDIR)\cdjpeg.obj"
+	-@erase "$(OUTDIR)\cjpeg.exe"
+
+"$(OUTDIR)" :
+    if not exist "$(OUTDIR)/$(NULL)" mkdir "$(OUTDIR)"
+
+# ADD BASE CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /YX /c
+# ADD CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /YX /c
+CPP_PROJ=/nologo /ML /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE"\
+ /Fp"$(INTDIR)/cjpeg.pch" /YX /Fo"$(INTDIR)/" /c 
+CPP_OBJS=.\cjpeg\Release/
+CPP_SBRS=.\.
+# ADD BASE RSC /l 0x409 /d "NDEBUG"
+# ADD RSC /l 0x409 /d "NDEBUG"
+BSC32=bscmake.exe
+# ADD BASE BSC32 /nologo
+# ADD BSC32 /nologo
+BSC32_FLAGS=/nologo /o"$(OUTDIR)/cjpeg.bsc" 
+BSC32_SBRS= \
+	
+LINK32=link.exe
+# ADD BASE LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386
+# ADD LINK32 Release\jpeg.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386
+LINK32_FLAGS=Release\jpeg.lib kernel32.lib user32.lib gdi32.lib winspool.lib\
+ comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib\
+ odbc32.lib odbccp32.lib /nologo /subsystem:console /incremental:no\
+ /pdb:"$(OUTDIR)/cjpeg.pdb" /machine:I386 /out:"$(OUTDIR)/cjpeg.exe" 
+LINK32_OBJS= \
+	"$(INTDIR)\cjpeg.obj" \
+	"$(INTDIR)\rdppm.obj" \
+	"$(INTDIR)\rdgif.obj" \
+	"$(INTDIR)\rdtarga.obj" \
+	"$(INTDIR)\rdrle.obj" \
+	"$(INTDIR)\rdbmp.obj" \
+	"$(INTDIR)\rdswitch.obj" \
+	"$(INTDIR)\cdjpeg.obj" \
+
+
+"$(OUTDIR)\cjpeg.exe" : "$(OUTDIR)" $(DEF_FILE) $(LINK32_OBJS)
+    $(LINK32) @<<
+  $(LINK32_FLAGS) $(LINK32_OBJS)
+<<
+
+!ELSEIF  "$(CFG)" == "djpeg - Win32"
+
+# PROP BASE Use_MFC 0
+# PROP BASE Use_Debug_Libraries 0
+# PROP BASE Output_Dir "djpeg\Release"
+# PROP BASE Intermediate_Dir "djpeg\Release"
+# PROP BASE Target_Dir "djpeg"
+# PROP Use_MFC 0
+# PROP Use_Debug_Libraries 0
+# PROP Output_Dir "djpeg\Release"
+# PROP Intermediate_Dir "djpeg\Release"
+# PROP Target_Dir "djpeg"
+OUTDIR=.\djpeg\Release
+INTDIR=.\djpeg\Release
+
+ALL : "$(OUTDIR)\djpeg.exe"
+
+CLEAN : 
+	-@erase "$(INTDIR)\djpeg.obj"
+	-@erase "$(INTDIR)\wrppm.obj"
+	-@erase "$(INTDIR)\wrgif.obj"
+	-@erase "$(INTDIR)\wrtarga.obj"
+	-@erase "$(INTDIR)\wrrle.obj"
+	-@erase "$(INTDIR)\wrbmp.obj"
+	-@erase "$(INTDIR)\rdcolmap.obj"
+	-@erase "$(INTDIR)\cdjpeg.obj"
+	-@erase "$(OUTDIR)\djpeg.exe"
+
+"$(OUTDIR)" :
+    if not exist "$(OUTDIR)/$(NULL)" mkdir "$(OUTDIR)"
+
+# ADD BASE CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /YX /c
+# ADD CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /YX /c
+CPP_PROJ=/nologo /ML /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE"\
+ /Fp"$(INTDIR)/djpeg.pch" /YX /Fo"$(INTDIR)/" /c 
+CPP_OBJS=.\djpeg\Release/
+CPP_SBRS=.\.
+# ADD BASE RSC /l 0x409 /d "NDEBUG"
+# ADD RSC /l 0x409 /d "NDEBUG"
+BSC32=bscmake.exe
+# ADD BASE BSC32 /nologo
+# ADD BSC32 /nologo
+BSC32_FLAGS=/nologo /o"$(OUTDIR)/djpeg.bsc" 
+BSC32_SBRS= \
+	
+LINK32=link.exe
+# ADD BASE LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386
+# ADD LINK32 Release\jpeg.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386
+LINK32_FLAGS=Release\jpeg.lib kernel32.lib user32.lib gdi32.lib winspool.lib\
+ comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib\
+ odbc32.lib odbccp32.lib /nologo /subsystem:console /incremental:no\
+ /pdb:"$(OUTDIR)/djpeg.pdb" /machine:I386 /out:"$(OUTDIR)/djpeg.exe" 
+LINK32_OBJS= \
+	"$(INTDIR)\djpeg.obj" \
+	"$(INTDIR)\wrppm.obj" \
+	"$(INTDIR)\wrgif.obj" \
+	"$(INTDIR)\wrtarga.obj" \
+	"$(INTDIR)\wrrle.obj" \
+	"$(INTDIR)\wrbmp.obj" \
+	"$(INTDIR)\rdcolmap.obj" \
+	"$(INTDIR)\cdjpeg.obj" \
+
+
+"$(OUTDIR)\djpeg.exe" : "$(OUTDIR)" $(DEF_FILE) $(LINK32_OBJS)
+    $(LINK32) @<<
+  $(LINK32_FLAGS) $(LINK32_OBJS)
+<<
+
+!ELSEIF  "$(CFG)" == "jpegtran - Win32"
+
+# PROP BASE Use_MFC 0
+# PROP BASE Use_Debug_Libraries 0
+# PROP BASE Output_Dir "jpegtran\Release"
+# PROP BASE Intermediate_Dir "jpegtran\Release"
+# PROP BASE Target_Dir "jpegtran"
+# PROP Use_MFC 0
+# PROP Use_Debug_Libraries 0
+# PROP Output_Dir "jpegtran\Release"
+# PROP Intermediate_Dir "jpegtran\Release"
+# PROP Target_Dir "jpegtran"
+OUTDIR=.\jpegtran\Release
+INTDIR=.\jpegtran\Release
+
+ALL : "$(OUTDIR)\jpegtran.exe"
+
+CLEAN : 
+	-@erase "$(INTDIR)\jpegtran.obj"
+	-@erase "$(INTDIR)\rdswitch.obj"
+	-@erase "$(INTDIR)\cdjpeg.obj"
+	-@erase "$(INTDIR)\transupp.obj"
+	-@erase "$(OUTDIR)\jpegtran.exe"
+
+"$(OUTDIR)" :
+    if not exist "$(OUTDIR)/$(NULL)" mkdir "$(OUTDIR)"
+
+# ADD BASE CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /YX /c
+# ADD CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /YX /c
+CPP_PROJ=/nologo /ML /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE"\
+ /Fp"$(INTDIR)/jpegtran.pch" /YX /Fo"$(INTDIR)/" /c 
+CPP_OBJS=.\jpegtran\Release/
+CPP_SBRS=.\.
+# ADD BASE RSC /l 0x409 /d "NDEBUG"
+# ADD RSC /l 0x409 /d "NDEBUG"
+BSC32=bscmake.exe
+# ADD BASE BSC32 /nologo
+# ADD BSC32 /nologo
+BSC32_FLAGS=/nologo /o"$(OUTDIR)/jpegtran.bsc" 
+BSC32_SBRS= \
+	
+LINK32=link.exe
+# ADD BASE LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386
+# ADD LINK32 Release\jpeg.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386
+LINK32_FLAGS=Release\jpeg.lib kernel32.lib user32.lib gdi32.lib winspool.lib\
+ comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib\
+ odbc32.lib odbccp32.lib /nologo /subsystem:console /incremental:no\
+ /pdb:"$(OUTDIR)/jpegtran.pdb" /machine:I386 /out:"$(OUTDIR)/jpegtran.exe" 
+LINK32_OBJS= \
+	"$(INTDIR)\jpegtran.obj" \
+	"$(INTDIR)\rdswitch.obj" \
+	"$(INTDIR)\cdjpeg.obj" \
+	"$(INTDIR)\transupp.obj" \
+
+
+"$(OUTDIR)\jpegtran.exe" : "$(OUTDIR)" $(DEF_FILE) $(LINK32_OBJS)
+    $(LINK32) @<<
+  $(LINK32_FLAGS) $(LINK32_OBJS)
+<<
+
+!ELSEIF  "$(CFG)" == "rdjpgcom - Win32"
+
+# PROP BASE Use_MFC 0
+# PROP BASE Use_Debug_Libraries 0
+# PROP BASE Output_Dir "rdjpgcom\Release"
+# PROP BASE Intermediate_Dir "rdjpgcom\Release"
+# PROP BASE Target_Dir "rdjpgcom"
+# PROP Use_MFC 0
+# PROP Use_Debug_Libraries 0
+# PROP Output_Dir "rdjpgcom\Release"
+# PROP Intermediate_Dir "rdjpgcom\Release"
+# PROP Target_Dir "rdjpgcom"
+OUTDIR=.\rdjpgcom\Release
+INTDIR=.\rdjpgcom\Release
+
+ALL : "$(OUTDIR)\rdjpgcom.exe"
+
+CLEAN : 
+	-@erase "$(INTDIR)\rdjpgcom.obj"
+	-@erase "$(OUTDIR)\rdjpgcom.exe"
+
+"$(OUTDIR)" :
+    if not exist "$(OUTDIR)/$(NULL)" mkdir "$(OUTDIR)"
+
+# ADD BASE CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /YX /c
+# ADD CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /YX /c
+CPP_PROJ=/nologo /ML /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE"\
+ /Fp"$(INTDIR)/rdjpgcom.pch" /YX /Fo"$(INTDIR)/" /c 
+CPP_OBJS=.\rdjpgcom\Release/
+CPP_SBRS=.\.
+# ADD BASE RSC /l 0x409 /d "NDEBUG"
+# ADD RSC /l 0x409 /d "NDEBUG"
+BSC32=bscmake.exe
+# ADD BASE BSC32 /nologo
+# ADD BSC32 /nologo
+BSC32_FLAGS=/nologo /o"$(OUTDIR)/rdjpgcom.bsc" 
+BSC32_SBRS= \
+	
+LINK32=link.exe
+# ADD BASE LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386
+# ADD LINK32 Release\jpeg.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386
+LINK32_FLAGS=Release\jpeg.lib kernel32.lib user32.lib gdi32.lib winspool.lib\
+ comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib\
+ odbc32.lib odbccp32.lib /nologo /subsystem:console /incremental:no\
+ /pdb:"$(OUTDIR)/rdjpgcom.pdb" /machine:I386 /out:"$(OUTDIR)/rdjpgcom.exe" 
+LINK32_OBJS= \
+	"$(INTDIR)\rdjpgcom.obj"
+
+"$(OUTDIR)\rdjpgcom.exe" : "$(OUTDIR)" $(DEF_FILE) $(LINK32_OBJS)
+    $(LINK32) @<<
+  $(LINK32_FLAGS) $(LINK32_OBJS)
+<<
+
+!ELSEIF  "$(CFG)" == "wrjpgcom - Win32"
+
+# PROP BASE Use_MFC 0
+# PROP BASE Use_Debug_Libraries 0
+# PROP BASE Output_Dir "wrjpgcom\Release"
+# PROP BASE Intermediate_Dir "wrjpgcom\Release"
+# PROP BASE Target_Dir "wrjpgcom"
+# PROP Use_MFC 0
+# PROP Use_Debug_Libraries 0
+# PROP Output_Dir "wrjpgcom\Release"
+# PROP Intermediate_Dir "wrjpgcom\Release"
+# PROP Target_Dir "wrjpgcom"
+OUTDIR=.\wrjpgcom\Release
+INTDIR=.\wrjpgcom\Release
+
+ALL : "$(OUTDIR)\wrjpgcom.exe"
+
+CLEAN : 
+	-@erase "$(INTDIR)\wrjpgcom.obj"
+	-@erase "$(OUTDIR)\wrjpgcom.exe"
+
+"$(OUTDIR)" :
+    if not exist "$(OUTDIR)/$(NULL)" mkdir "$(OUTDIR)"
+
+# ADD BASE CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /YX /c
+# ADD CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /YX /c
+CPP_PROJ=/nologo /ML /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE"\
+ /Fp"$(INTDIR)/wrjpgcom.pch" /YX /Fo"$(INTDIR)/" /c 
+CPP_OBJS=.\wrjpgcom\Release/
+CPP_SBRS=.\.
+# ADD BASE RSC /l 0x409 /d "NDEBUG"
+# ADD RSC /l 0x409 /d "NDEBUG"
+BSC32=bscmake.exe
+# ADD BASE BSC32 /nologo
+# ADD BSC32 /nologo
+BSC32_FLAGS=/nologo /o"$(OUTDIR)/wrjpgcom.bsc" 
+BSC32_SBRS= \
+	
+LINK32=link.exe
+# ADD BASE LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386
+# ADD LINK32 Release\jpeg.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386
+LINK32_FLAGS=Release\jpeg.lib kernel32.lib user32.lib gdi32.lib winspool.lib\
+ comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib\
+ odbc32.lib odbccp32.lib /nologo /subsystem:console /incremental:no\
+ /pdb:"$(OUTDIR)/wrjpgcom.pdb" /machine:I386 /out:"$(OUTDIR)/wrjpgcom.exe" 
+LINK32_OBJS= \
+	"$(INTDIR)\wrjpgcom.obj"
+
+"$(OUTDIR)\wrjpgcom.exe" : "$(OUTDIR)" $(DEF_FILE) $(LINK32_OBJS)
+    $(LINK32) @<<
+  $(LINK32_FLAGS) $(LINK32_OBJS)
+<<
+
+!ENDIF 
+
+.c{$(CPP_OBJS)}.obj:
+   $(CPP) $(CPP_PROJ) $<  
+
+.cpp{$(CPP_OBJS)}.obj:
+   $(CPP) $(CPP_PROJ) $<  
+
+.cxx{$(CPP_OBJS)}.obj:
+   $(CPP) $(CPP_PROJ) $<  
+
+.c{$(CPP_SBRS)}.sbr:
+   $(CPP) $(CPP_PROJ) $<  
+
+.cpp{$(CPP_SBRS)}.sbr:
+   $(CPP) $(CPP_PROJ) $<  
+
+.cxx{$(CPP_SBRS)}.sbr:
+   $(CPP) $(CPP_PROJ) $<  
+
+################################################################################
+# Begin Target
+
+# Name "cjpeg - Win32"
+
+!IF  "$(CFG)" == "cjpeg - Win32"
+
+!ENDIF 
+
+################################################################################
+# Begin Source File
+
+SOURCE="cjpeg.c"
+DEP_CPP_CJPEG=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	"jversion.h"\
+	
+
+"$(INTDIR)\cjpeg.obj" : $(SOURCE) $(DEP_CPP_CJPEG) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="cdjpeg.c"
+DEP_CPP_CDJPE=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\cdjpeg.obj" : $(SOURCE) $(DEP_CPP_CDJPE) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="rdswitch.c"
+DEP_CPP_RDSWI=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\rdswitch.obj" : $(SOURCE) $(DEP_CPP_RDSWI) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="rdppm.c"
+DEP_CPP_RDPPM=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\rdppm.obj" : $(SOURCE) $(DEP_CPP_RDPPM) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="rdgif.c"
+DEP_CPP_RDGIF=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\rdgif.obj" : $(SOURCE) $(DEP_CPP_RDGIF) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="rdtarga.c"
+DEP_CPP_RDTAR=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\rdtarga.obj" : $(SOURCE) $(DEP_CPP_RDTAR) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="rdbmp.c"
+DEP_CPP_RDBMP=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\rdbmp.obj" : $(SOURCE) $(DEP_CPP_RDBMP) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="rdrle.c"
+DEP_CPP_RDRLE=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\rdrle.obj" : $(SOURCE) $(DEP_CPP_RDRLE) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+# End Target
+################################################################################
+# Begin Target
+
+# Name "djpeg - Win32"
+
+!IF  "$(CFG)" == "djpeg - Win32"
+
+!ENDIF 
+
+################################################################################
+# Begin Source File
+
+SOURCE="djpeg.c"
+DEP_CPP_DJPEG=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	"jversion.h"\
+	
+
+"$(INTDIR)\djpeg.obj" : $(SOURCE) $(DEP_CPP_DJPEG) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="cdjpeg.c"
+DEP_CPP_CDJPE=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\cdjpeg.obj" : $(SOURCE) $(DEP_CPP_CDJPE) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="rdcolmap.c"
+DEP_CPP_RDCOL=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\rdcolmap.obj" : $(SOURCE) $(DEP_CPP_RDCOL) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="wrppm.c"
+DEP_CPP_WRPPM=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\wrppm.obj" : $(SOURCE) $(DEP_CPP_WRPPM) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="wrgif.c"
+DEP_CPP_WRGIF=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\wrgif.obj" : $(SOURCE) $(DEP_CPP_WRGIF) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="wrtarga.c"
+DEP_CPP_WRTAR=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\wrtarga.obj" : $(SOURCE) $(DEP_CPP_WRTAR) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="wrbmp.c"
+DEP_CPP_WRBMP=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\wrbmp.obj" : $(SOURCE) $(DEP_CPP_WRBMP) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="wrrle.c"
+DEP_CPP_WRRLE=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\wrrle.obj" : $(SOURCE) $(DEP_CPP_WRRLE) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+# End Target
+################################################################################
+# Begin Target
+
+# Name "jpegtran - Win32"
+
+!IF  "$(CFG)" == "jpegtran - Win32"
+
+!ENDIF 
+
+################################################################################
+# Begin Source File
+
+SOURCE="jpegtran.c"
+DEP_CPP_JPEGT=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	"transupp.h"\
+	"jversion.h"\
+	
+
+"$(INTDIR)\jpegtran.obj" : $(SOURCE) $(DEP_CPP_JPEGT) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="cdjpeg.c"
+DEP_CPP_CDJPE=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\cdjpeg.obj" : $(SOURCE) $(DEP_CPP_CDJPE) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="rdswitch.c"
+DEP_CPP_RDSWI=\
+	"cdjpeg.h"\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	"cderror.h"\
+	
+
+"$(INTDIR)\rdswitch.obj" : $(SOURCE) $(DEP_CPP_RDSWI) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="transupp.c"
+DEP_CPP_TRANS=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"transupp.h"\
+	
+
+"$(INTDIR)\transupp.obj" : $(SOURCE) $(DEP_CPP_TRANS) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+# End Target
+################################################################################
+# Begin Target
+
+# Name "rdjpgcom - Win32"
+
+!IF  "$(CFG)" == "rdjpgcom - Win32"
+
+!ENDIF 
+
+################################################################################
+# Begin Source File
+
+SOURCE="rdjpgcom.c"
+DEP_CPP_RDJPG=\
+	"jinclude.h"\
+	"jconfig.h"\
+	
+
+"$(INTDIR)\rdjpgcom.obj" : $(SOURCE) $(DEP_CPP_RDJPG) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+# End Target
+################################################################################
+# Begin Target
+
+# Name "wrjpgcom - Win32"
+
+!IF  "$(CFG)" == "wrjpgcom - Win32"
+
+!ENDIF 
+
+################################################################################
+# Begin Source File
+
+SOURCE="wrjpgcom.c"
+DEP_CPP_WRJPG=\
+	"jinclude.h"\
+	"jconfig.h"\
+	
+
+"$(INTDIR)\wrjpgcom.obj" : $(SOURCE) $(DEP_CPP_WRJPG) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+# End Target
+# End Project
+################################################################################
+
diff --git a/libjpegtwrp/makefile.ansi b/libjpegtwrp/makefile.ansi
new file mode 100644
index 000000000..829191359
--- /dev/null
+++ b/libjpegtwrp/makefile.ansi
@@ -0,0 +1,214 @@
+# Makefile for Independent JPEG Group's software
+
+# This makefile is suitable for Unix-like systems with ANSI-capable compilers.
+# If you have a non-ANSI compiler, makefile.unix is a better starting point.
+
+# Read installation instructions before saying "make" !!
+
+# The name of your C compiler:
+CC= cc
+
+# You may need to adjust these cc options:
+CFLAGS= -O
+# Generally, we recommend defining any configuration symbols in jconfig.h,
+# NOT via -D switches here.
+
+# Link-time cc options:
+LDFLAGS= 
+
+# To link any special libraries, add the necessary -l commands here.
+LDLIBS= 
+
+# Put here the object file name for the correct system-dependent memory
+# manager file.  For Unix this is usually jmemnobs.o, but you may want
+# to use jmemansi.o or jmemname.o if you have limited swap space.
+SYSDEPMEM= jmemnobs.o
+
+# miscellaneous OS-dependent stuff
+# linker
+LN= $(CC)
+# file deletion command
+RM= rm -f
+# library (.a) file creation command
+AR= ar rc
+# second step in .a creation (use "touch" if not needed)
+AR2= ranlib
+
+# End of configurable options.
+
+
+# source files: JPEG library proper
+LIBSOURCES= jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c \
+        jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c \
+        jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c \
+        jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c \
+        jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c \
+        jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c \
+        jfdctint.c jidctflt.c jidctfst.c jidctint.c jidctred.c jquant1.c \
+        jquant2.c jutils.c jmemmgr.c
+# memmgr back ends: compile only one of these into a working library
+SYSDEPSOURCES= jmemansi.c jmemname.c jmemnobs.c jmemdos.c jmemmac.c
+# source files: cjpeg/djpeg/jpegtran applications, also rdjpgcom/wrjpgcom
+APPSOURCES= cjpeg.c djpeg.c jpegtran.c rdjpgcom.c wrjpgcom.c cdjpeg.c \
+        rdcolmap.c rdswitch.c transupp.c rdppm.c wrppm.c rdgif.c wrgif.c \
+        rdtarga.c wrtarga.c rdbmp.c wrbmp.c rdrle.c wrrle.c
+SOURCES= $(LIBSOURCES) $(SYSDEPSOURCES) $(APPSOURCES)
+# files included by source files
+INCLUDES= jchuff.h jdhuff.h jdct.h jerror.h jinclude.h jmemsys.h jmorecfg.h \
+        jpegint.h jpeglib.h jversion.h cdjpeg.h cderror.h transupp.h
+# documentation, test, and support files
+DOCS= README install.doc usage.doc cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 \
+        wrjpgcom.1 wizard.doc example.c libjpeg.doc structure.doc \
+        coderules.doc filelist.doc change.log
+MKFILES= configure makefile.cfg makefile.ansi makefile.unix makefile.bcc \
+        makefile.mc6 makefile.dj makefile.wat makefile.vc makelib.ds \
+        makeapps.ds makeproj.mac makcjpeg.st makdjpeg.st makljpeg.st \
+        maktjpeg.st makefile.manx makefile.sas makefile.mms makefile.vms \
+        makvms.opt
+CONFIGFILES= jconfig.cfg jconfig.bcc jconfig.mc6 jconfig.dj jconfig.wat \
+        jconfig.vc jconfig.mac jconfig.st jconfig.manx jconfig.sas \
+        jconfig.vms
+CONFIGUREFILES= config.guess config.sub install-sh ltconfig ltmain.sh
+OTHERFILES= jconfig.doc ckconfig.c ansi2knr.c ansi2knr.1 jmemdosa.asm
+TESTFILES= testorig.jpg testimg.ppm testimg.bmp testimg.jpg testprog.jpg \
+        testimgp.jpg
+DISTFILES= $(DOCS) $(MKFILES) $(CONFIGFILES) $(SOURCES) $(INCLUDES) \
+        $(CONFIGUREFILES) $(OTHERFILES) $(TESTFILES)
+# library object files common to compression and decompression
+COMOBJECTS= jcomapi.o jutils.o jerror.o jmemmgr.o $(SYSDEPMEM)
+# compression library object files
+CLIBOBJECTS= jcapimin.o jcapistd.o jctrans.o jcparam.o jdatadst.o jcinit.o \
+        jcmaster.o jcmarker.o jcmainct.o jcprepct.o jccoefct.o jccolor.o \
+        jcsample.o jchuff.o jcphuff.o jcdctmgr.o jfdctfst.o jfdctflt.o \
+        jfdctint.o
+# decompression library object files
+DLIBOBJECTS= jdapimin.o jdapistd.o jdtrans.o jdatasrc.o jdmaster.o \
+        jdinput.o jdmarker.o jdhuff.o jdphuff.o jdmainct.o jdcoefct.o \
+        jdpostct.o jddctmgr.o jidctfst.o jidctflt.o jidctint.o jidctred.o \
+        jdsample.o jdcolor.o jquant1.o jquant2.o jdmerge.o
+# These objectfiles are included in libjpeg.a
+LIBOBJECTS= $(CLIBOBJECTS) $(DLIBOBJECTS) $(COMOBJECTS)
+# object files for sample applications (excluding library files)
+COBJECTS= cjpeg.o rdppm.o rdgif.o rdtarga.o rdrle.o rdbmp.o rdswitch.o \
+        cdjpeg.o
+DOBJECTS= djpeg.o wrppm.o wrgif.o wrtarga.o wrrle.o wrbmp.o rdcolmap.o \
+        cdjpeg.o
+TROBJECTS= jpegtran.o rdswitch.o cdjpeg.o transupp.o
+
+
+all: libjpeg.a cjpeg djpeg jpegtran rdjpgcom wrjpgcom
+
+libjpeg.a: $(LIBOBJECTS)
+	$(RM) libjpeg.a
+	$(AR) libjpeg.a  $(LIBOBJECTS)
+	$(AR2) libjpeg.a
+
+cjpeg: $(COBJECTS) libjpeg.a
+	$(LN) $(LDFLAGS) -o cjpeg $(COBJECTS) libjpeg.a $(LDLIBS)
+
+djpeg: $(DOBJECTS) libjpeg.a
+	$(LN) $(LDFLAGS) -o djpeg $(DOBJECTS) libjpeg.a $(LDLIBS)
+
+jpegtran: $(TROBJECTS) libjpeg.a
+	$(LN) $(LDFLAGS) -o jpegtran $(TROBJECTS) libjpeg.a $(LDLIBS)
+
+rdjpgcom: rdjpgcom.o
+	$(LN) $(LDFLAGS) -o rdjpgcom rdjpgcom.o $(LDLIBS)
+
+wrjpgcom: wrjpgcom.o
+	$(LN) $(LDFLAGS) -o wrjpgcom wrjpgcom.o $(LDLIBS)
+
+jconfig.h: jconfig.doc
+	echo You must prepare a system-dependent jconfig.h file.
+	echo Please read the installation directions in install.doc.
+	exit 1
+
+clean:
+	$(RM) *.o cjpeg djpeg jpegtran libjpeg.a rdjpgcom wrjpgcom
+	$(RM) core testout*
+
+test: cjpeg djpeg jpegtran
+	$(RM) testout*
+	./djpeg -dct int -ppm -outfile testout.ppm  testorig.jpg
+	./djpeg -dct int -bmp -colors 256 -outfile testout.bmp  testorig.jpg
+	./cjpeg -dct int -outfile testout.jpg  testimg.ppm
+	./djpeg -dct int -ppm -outfile testoutp.ppm testprog.jpg
+	./cjpeg -dct int -progressive -opt -outfile testoutp.jpg testimg.ppm
+	./jpegtran -outfile testoutt.jpg testprog.jpg
+	cmp testimg.ppm testout.ppm
+	cmp testimg.bmp testout.bmp
+	cmp testimg.jpg testout.jpg
+	cmp testimg.ppm testoutp.ppm
+	cmp testimgp.jpg testoutp.jpg
+	cmp testorig.jpg testoutt.jpg
+
+
+jcapimin.o: jcapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcapistd.o: jcapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccoefct.o: jccoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccolor.o: jccolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcdctmgr.o: jcdctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jchuff.o: jchuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcinit.o: jcinit.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmainct.o: jcmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmarker.o: jcmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmaster.o: jcmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcomapi.o: jcomapi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcparam.o: jcparam.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcphuff.o: jcphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcprepct.o: jcprepct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcsample.o: jcsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jctrans.o: jctrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapimin.o: jdapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapistd.o: jdapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdatadst.o: jdatadst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdatasrc.o: jdatasrc.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdcoefct.o: jdcoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdcolor.o: jdcolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jddctmgr.o: jddctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jdhuff.o: jdhuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdinput.o: jdinput.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmainct.o: jdmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmarker.o: jdmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmaster.o: jdmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmerge.o: jdmerge.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdphuff.o: jdphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdpostct.o: jdpostct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdsample.o: jdsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdtrans.o: jdtrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jerror.o: jerror.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jversion.h jerror.h
+jfdctflt.o: jfdctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctfst.o: jfdctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctint.o: jfdctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctflt.o: jidctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctfst.o: jidctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctint.o: jidctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctred.o: jidctred.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jquant1.o: jquant1.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jquant2.o: jquant2.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jutils.o: jutils.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jmemmgr.o: jmemmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemansi.o: jmemansi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemname.o: jmemname.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemnobs.o: jmemnobs.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemdos.o: jmemdos.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemmac.o: jmemmac.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+cjpeg.o: cjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+djpeg.o: djpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+jpegtran.o: jpegtran.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h transupp.h jversion.h
+rdjpgcom.o: rdjpgcom.c jinclude.h jconfig.h
+wrjpgcom.o: wrjpgcom.c jinclude.h jconfig.h
+cdjpeg.o: cdjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdcolmap.o: rdcolmap.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdswitch.o: rdswitch.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+transupp.o: transupp.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h transupp.h
+rdppm.o: rdppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrppm.o: wrppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdgif.o: rdgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrgif.o: wrgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdtarga.o: rdtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrtarga.o: wrtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdbmp.o: rdbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrbmp.o: wrbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdrle.o: rdrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrrle.o: wrrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
diff --git a/libjpegtwrp/makefile.bcc b/libjpegtwrp/makefile.bcc
new file mode 100644
index 000000000..a1cfcde66
--- /dev/null
+++ b/libjpegtwrp/makefile.bcc
@@ -0,0 +1,285 @@
+# Makefile for Independent JPEG Group's software
+
+# This makefile is suitable for Borland C on MS-DOS or OS/2.
+# It works with Borland C++ for DOS, revision 3.0 or later,
+# and has been tested with Borland C++ for OS/2.
+# Watch out for optimization bugs in the OS/2 compilers --- see notes below!
+# Thanks to Tom Wright and Ge' Weijers (original DOS) and
+# Ken Porter (OS/2) for this file.
+
+# Read installation instructions before saying "make" !!
+
+# Are we under DOS or OS/2?
+!if !$d(DOS) && !$d(OS2)
+!if $d(__OS2__)
+OS2=1
+!else
+DOS=1
+!endif
+!endif
+
+# The name of your C compiler:
+CC= bcc
+
+# You may need to adjust these cc options:
+!if $d(DOS)
+CFLAGS= -O2 -mm -w-par -w-stu -w-ccc -w-rch
+!else
+CFLAGS= -O1 -w-par -w-stu -w-ccc -w-rch
+!endif
+# -O2 enables full code optimization (for pre-3.0 Borland C++, use -O -G -Z).
+# -O2 is buggy in Borland OS/2 C++ revision 2.0, so use -O1 there for now.
+# If you have Borland OS/2 C++ revision 1.0, use -O or no optimization at all.
+# -mm selects medium memory model (near data, far code pointers; DOS only!)
+# -w-par suppresses warnings about unused function parameters
+# -w-stu suppresses warnings about incomplete structures
+# -w-ccc suppresses warnings about compile-time-constant conditions
+# -w-rch suppresses warnings about unreachable code
+# Generally, we recommend defining any configuration symbols in jconfig.h,
+# NOT via -D switches here.
+
+# Link-time cc options:
+!if $d(DOS)
+LDFLAGS= -mm
+# memory model option here must match CFLAGS!
+!else
+LDFLAGS=
+# -lai full-screen app
+# -lc case-significant link
+!endif
+
+# Put here the object file name for the correct system-dependent memory
+# manager file.
+# For DOS, we recommend jmemdos.c and jmemdosa.asm.
+# For OS/2, we recommend jmemnobs.c (flat memory!)
+# SYSDEPMEMLIB must list the same files with "+" signs for the librarian.
+!if $d(DOS)
+SYSDEPMEM= jmemdos.obj jmemdosa.obj
+SYSDEPMEMLIB= +jmemdos.obj +jmemdosa.obj
+!else
+SYSDEPMEM= jmemnobs.obj
+SYSDEPMEMLIB= +jmemnobs.obj
+!endif
+
+# End of configurable options.
+
+
+# source files: JPEG library proper
+LIBSOURCES= jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c \
+        jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c \
+        jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c \
+        jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c \
+        jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c \
+        jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c \
+        jfdctint.c jidctflt.c jidctfst.c jidctint.c jidctred.c jquant1.c \
+        jquant2.c jutils.c jmemmgr.c
+# memmgr back ends: compile only one of these into a working library
+SYSDEPSOURCES= jmemansi.c jmemname.c jmemnobs.c jmemdos.c jmemmac.c
+# source files: cjpeg/djpeg/jpegtran applications, also rdjpgcom/wrjpgcom
+APPSOURCES= cjpeg.c djpeg.c jpegtran.c rdjpgcom.c wrjpgcom.c cdjpeg.c \
+        rdcolmap.c rdswitch.c transupp.c rdppm.c wrppm.c rdgif.c wrgif.c \
+        rdtarga.c wrtarga.c rdbmp.c wrbmp.c rdrle.c wrrle.c
+SOURCES= $(LIBSOURCES) $(SYSDEPSOURCES) $(APPSOURCES)
+# files included by source files
+INCLUDES= jchuff.h jdhuff.h jdct.h jerror.h jinclude.h jmemsys.h jmorecfg.h \
+        jpegint.h jpeglib.h jversion.h cdjpeg.h cderror.h transupp.h
+# documentation, test, and support files
+DOCS= README install.doc usage.doc cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 \
+        wrjpgcom.1 wizard.doc example.c libjpeg.doc structure.doc \
+        coderules.doc filelist.doc change.log
+MKFILES= configure makefile.cfg makefile.ansi makefile.unix makefile.bcc \
+        makefile.mc6 makefile.dj makefile.wat makefile.vc makelib.ds \
+        makeapps.ds makeproj.mac makcjpeg.st makdjpeg.st makljpeg.st \
+        maktjpeg.st makefile.manx makefile.sas makefile.mms makefile.vms \
+        makvms.opt
+CONFIGFILES= jconfig.cfg jconfig.bcc jconfig.mc6 jconfig.dj jconfig.wat \
+        jconfig.vc jconfig.mac jconfig.st jconfig.manx jconfig.sas \
+        jconfig.vms
+CONFIGUREFILES= config.guess config.sub install-sh ltconfig ltmain.sh
+OTHERFILES= jconfig.doc ckconfig.c ansi2knr.c ansi2knr.1 jmemdosa.asm
+TESTFILES= testorig.jpg testimg.ppm testimg.bmp testimg.jpg testprog.jpg \
+        testimgp.jpg
+DISTFILES= $(DOCS) $(MKFILES) $(CONFIGFILES) $(SOURCES) $(INCLUDES) \
+        $(CONFIGUREFILES) $(OTHERFILES) $(TESTFILES)
+# library object files common to compression and decompression
+COMOBJECTS= jcomapi.obj jutils.obj jerror.obj jmemmgr.obj $(SYSDEPMEM)
+# compression library object files
+CLIBOBJECTS= jcapimin.obj jcapistd.obj jctrans.obj jcparam.obj jdatadst.obj \
+        jcinit.obj jcmaster.obj jcmarker.obj jcmainct.obj jcprepct.obj \
+        jccoefct.obj jccolor.obj jcsample.obj jchuff.obj jcphuff.obj \
+        jcdctmgr.obj jfdctfst.obj jfdctflt.obj jfdctint.obj
+# decompression library object files
+DLIBOBJECTS= jdapimin.obj jdapistd.obj jdtrans.obj jdatasrc.obj \
+        jdmaster.obj jdinput.obj jdmarker.obj jdhuff.obj jdphuff.obj \
+        jdmainct.obj jdcoefct.obj jdpostct.obj jddctmgr.obj jidctfst.obj \
+        jidctflt.obj jidctint.obj jidctred.obj jdsample.obj jdcolor.obj \
+        jquant1.obj jquant2.obj jdmerge.obj
+# These objectfiles are included in libjpeg.lib
+LIBOBJECTS= $(CLIBOBJECTS) $(DLIBOBJECTS) $(COMOBJECTS)
+# object files for sample applications (excluding library files)
+COBJECTS= cjpeg.obj rdppm.obj rdgif.obj rdtarga.obj rdrle.obj rdbmp.obj \
+        rdswitch.obj cdjpeg.obj
+DOBJECTS= djpeg.obj wrppm.obj wrgif.obj wrtarga.obj wrrle.obj wrbmp.obj \
+        rdcolmap.obj cdjpeg.obj
+TROBJECTS= jpegtran.obj rdswitch.obj cdjpeg.obj transupp.obj
+
+
+all: libjpeg.lib cjpeg.exe djpeg.exe jpegtran.exe rdjpgcom.exe wrjpgcom.exe
+
+libjpeg.lib: $(LIBOBJECTS)
+	- del libjpeg.lib
+	tlib libjpeg.lib /E /C @&&|
++jcapimin.obj +jcapistd.obj +jctrans.obj +jcparam.obj +jdatadst.obj &
++jcinit.obj +jcmaster.obj +jcmarker.obj +jcmainct.obj +jcprepct.obj &
++jccoefct.obj +jccolor.obj +jcsample.obj +jchuff.obj +jcphuff.obj &
++jcdctmgr.obj +jfdctfst.obj +jfdctflt.obj +jfdctint.obj +jdapimin.obj &
++jdapistd.obj +jdtrans.obj +jdatasrc.obj +jdmaster.obj +jdinput.obj &
++jdmarker.obj +jdhuff.obj +jdphuff.obj +jdmainct.obj +jdcoefct.obj &
++jdpostct.obj +jddctmgr.obj +jidctfst.obj +jidctflt.obj +jidctint.obj &
++jidctred.obj +jdsample.obj +jdcolor.obj +jquant1.obj +jquant2.obj &
++jdmerge.obj +jcomapi.obj +jutils.obj +jerror.obj +jmemmgr.obj &
+$(SYSDEPMEMLIB)
+|
+
+cjpeg.exe: $(COBJECTS) libjpeg.lib
+	$(CC) $(LDFLAGS) -ecjpeg.exe $(COBJECTS) libjpeg.lib
+
+djpeg.exe: $(DOBJECTS) libjpeg.lib
+	$(CC) $(LDFLAGS) -edjpeg.exe $(DOBJECTS) libjpeg.lib
+
+jpegtran.exe: $(TROBJECTS) libjpeg.lib
+	$(CC) $(LDFLAGS) -ejpegtran.exe $(TROBJECTS) libjpeg.lib
+
+rdjpgcom.exe: rdjpgcom.c
+!if $d(DOS)
+	$(CC) -ms -O rdjpgcom.c
+!else
+	$(CC) $(CFLAGS) rdjpgcom.c
+!endif
+
+# On DOS, wrjpgcom needs large model so it can malloc a 64K chunk
+wrjpgcom.exe: wrjpgcom.c
+!if $d(DOS)
+	$(CC) -ml -O wrjpgcom.c
+!else
+	$(CC) $(CFLAGS) wrjpgcom.c
+!endif
+
+# This "{}" syntax allows Borland Make to "batch" source files.
+# In this way, each run of the compiler can build many modules.
+.c.obj:
+	$(CC) $(CFLAGS) -c{ $<}
+
+jconfig.h: jconfig.doc
+	echo You must prepare a system-dependent jconfig.h file.
+	echo Please read the installation directions in install.doc.
+	exit 1
+
+clean:
+	- del *.obj
+	- del libjpeg.lib
+	- del cjpeg.exe
+	- del djpeg.exe
+	- del jpegtran.exe
+	- del rdjpgcom.exe
+	- del wrjpgcom.exe
+	- del testout*.*
+
+test: cjpeg.exe djpeg.exe jpegtran.exe
+	- del testout*.*
+	djpeg -dct int -ppm -outfile testout.ppm  testorig.jpg
+	djpeg -dct int -bmp -colors 256 -outfile testout.bmp  testorig.jpg
+	cjpeg -dct int -outfile testout.jpg  testimg.ppm
+	djpeg -dct int -ppm -outfile testoutp.ppm testprog.jpg
+	cjpeg -dct int -progressive -opt -outfile testoutp.jpg testimg.ppm
+	jpegtran -outfile testoutt.jpg testprog.jpg
+!if $d(DOS)
+	fc /b testimg.ppm testout.ppm
+	fc /b testimg.bmp testout.bmp
+	fc /b testimg.jpg testout.jpg
+	fc /b testimg.ppm testoutp.ppm
+	fc /b testimgp.jpg testoutp.jpg
+	fc /b testorig.jpg testoutt.jpg
+!else
+	echo n > n.tmp
+	comp testimg.ppm testout.ppm < n.tmp
+	comp testimg.bmp testout.bmp < n.tmp
+	comp testimg.jpg testout.jpg < n.tmp
+	comp testimg.ppm testoutp.ppm < n.tmp
+	comp testimgp.jpg testoutp.jpg < n.tmp
+	comp testorig.jpg testoutt.jpg < n.tmp
+	del n.tmp
+!endif
+
+
+jcapimin.obj: jcapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcapistd.obj: jcapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccoefct.obj: jccoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccolor.obj: jccolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcdctmgr.obj: jcdctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jchuff.obj: jchuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcinit.obj: jcinit.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmainct.obj: jcmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmarker.obj: jcmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmaster.obj: jcmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcomapi.obj: jcomapi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcparam.obj: jcparam.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcphuff.obj: jcphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcprepct.obj: jcprepct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcsample.obj: jcsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jctrans.obj: jctrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapimin.obj: jdapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapistd.obj: jdapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdatadst.obj: jdatadst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdatasrc.obj: jdatasrc.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdcoefct.obj: jdcoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdcolor.obj: jdcolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jddctmgr.obj: jddctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jdhuff.obj: jdhuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdinput.obj: jdinput.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmainct.obj: jdmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmarker.obj: jdmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmaster.obj: jdmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmerge.obj: jdmerge.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdphuff.obj: jdphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdpostct.obj: jdpostct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdsample.obj: jdsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdtrans.obj: jdtrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jerror.obj: jerror.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jversion.h jerror.h
+jfdctflt.obj: jfdctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctfst.obj: jfdctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctint.obj: jfdctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctflt.obj: jidctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctfst.obj: jidctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctint.obj: jidctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctred.obj: jidctred.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jquant1.obj: jquant1.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jquant2.obj: jquant2.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jutils.obj: jutils.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jmemmgr.obj: jmemmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemansi.obj: jmemansi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemname.obj: jmemname.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemnobs.obj: jmemnobs.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemdos.obj: jmemdos.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemmac.obj: jmemmac.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+cjpeg.obj: cjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+djpeg.obj: djpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+jpegtran.obj: jpegtran.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h transupp.h jversion.h
+rdjpgcom.obj: rdjpgcom.c jinclude.h jconfig.h
+wrjpgcom.obj: wrjpgcom.c jinclude.h jconfig.h
+cdjpeg.obj: cdjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdcolmap.obj: rdcolmap.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdswitch.obj: rdswitch.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+transupp.obj: transupp.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h transupp.h
+rdppm.obj: rdppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrppm.obj: wrppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdgif.obj: rdgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrgif.obj: wrgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdtarga.obj: rdtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrtarga.obj: wrtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdbmp.obj: rdbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrbmp.obj: wrbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdrle.obj: rdrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrrle.obj: wrrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+jmemdosa.obj: jmemdosa.asm
+	tasm /mx jmemdosa.asm
diff --git a/libjpegtwrp/makefile.cfg b/libjpegtwrp/makefile.cfg
new file mode 100644
index 000000000..f25e42e3e
--- /dev/null
+++ b/libjpegtwrp/makefile.cfg
@@ -0,0 +1,319 @@
+# Makefile for Independent JPEG Group's software
+
+# makefile.cfg is edited by configure to produce a custom Makefile.
+
+# Read installation instructions before saying "make" !!
+
+# For compiling with source and object files in different directories.
+srcdir = @srcdir@
+VPATH = @srcdir@
+
+# Where to install the programs and man pages.
+prefix = @prefix@
+exec_prefix = @exec_prefix@
+bindir = $(exec_prefix)/bin
+libdir = $(exec_prefix)/lib
+includedir = $(prefix)/include
+binprefix =
+manprefix =
+manext = 1
+mandir = $(prefix)/man/man$(manext)
+
+# The name of your C compiler:
+CC= @CC@
+
+# You may need to adjust these cc options:
+CFLAGS= @CFLAGS@ @CPPFLAGS@ @INCLUDEFLAGS@
+# Generally, we recommend defining any configuration symbols in jconfig.h,
+# NOT via -D switches here.
+# However, any special defines for ansi2knr.c may be included here:
+ANSI2KNRFLAGS= @ANSI2KNRFLAGS@
+
+# Link-time cc options:
+LDFLAGS= @LDFLAGS@
+
+# To link any special libraries, add the necessary -l commands here.
+LDLIBS= @LIBS@
+
+# If using GNU libtool, LIBTOOL references it; if not, LIBTOOL is empty.
+LIBTOOL = @LIBTOOL@
+# $(O) expands to "lo" if using libtool, plain "o" if not.
+# Similarly, $(A) expands to "la" or "a".
+O = @O@
+A = @A@
+
+# Library version ID; libtool uses this for the shared library version number.
+# Note: we suggest this match the macro of the same name in jpeglib.h.
+JPEG_LIB_VERSION = @JPEG_LIB_VERSION@
+
+# Put here the object file name for the correct system-dependent memory
+# manager file.  For Unix this is usually jmemnobs.o, but you may want
+# to use jmemansi.o or jmemname.o if you have limited swap space.
+SYSDEPMEM= @MEMORYMGR@
+
+# miscellaneous OS-dependent stuff
+SHELL= /bin/sh
+# linker
+LN= @LN@
+# file deletion command
+RM= rm -f
+# directory creation command
+MKDIR= mkdir
+# library (.a) file creation command
+AR= ar rc
+# second step in .a creation (use "touch" if not needed)
+AR2= @RANLIB@
+# installation program
+INSTALL= @INSTALL@
+INSTALL_PROGRAM= @INSTALL_PROGRAM@
+INSTALL_LIB= @INSTALL_LIB@
+INSTALL_DATA= @INSTALL_DATA@
+
+# End of configurable options.
+
+
+# source files: JPEG library proper
+LIBSOURCES= jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c \
+        jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c \
+        jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c \
+        jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c \
+        jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c \
+        jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c \
+        jfdctint.c jidctflt.c jidctfst.c jidctint.c jidctred.c jquant1.c \
+        jquant2.c jutils.c jmemmgr.c
+# memmgr back ends: compile only one of these into a working library
+SYSDEPSOURCES= jmemansi.c jmemname.c jmemnobs.c jmemdos.c jmemmac.c
+# source files: cjpeg/djpeg/jpegtran applications, also rdjpgcom/wrjpgcom
+APPSOURCES= cjpeg.c djpeg.c jpegtran.c rdjpgcom.c wrjpgcom.c cdjpeg.c \
+        rdcolmap.c rdswitch.c transupp.c rdppm.c wrppm.c rdgif.c wrgif.c \
+        rdtarga.c wrtarga.c rdbmp.c wrbmp.c rdrle.c wrrle.c
+SOURCES= $(LIBSOURCES) $(SYSDEPSOURCES) $(APPSOURCES)
+# files included by source files
+INCLUDES= jchuff.h jdhuff.h jdct.h jerror.h jinclude.h jmemsys.h jmorecfg.h \
+        jpegint.h jpeglib.h jversion.h cdjpeg.h cderror.h transupp.h
+# documentation, test, and support files
+DOCS= README install.doc usage.doc cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 \
+        wrjpgcom.1 wizard.doc example.c libjpeg.doc structure.doc \
+        coderules.doc filelist.doc change.log
+MKFILES= configure makefile.cfg makefile.ansi makefile.unix makefile.bcc \
+        makefile.mc6 makefile.dj makefile.wat makefile.vc makelib.ds \
+        makeapps.ds makeproj.mac makcjpeg.st makdjpeg.st makljpeg.st \
+        maktjpeg.st makefile.manx makefile.sas makefile.mms makefile.vms \
+        makvms.opt
+CONFIGFILES= jconfig.cfg jconfig.bcc jconfig.mc6 jconfig.dj jconfig.wat \
+        jconfig.vc jconfig.mac jconfig.st jconfig.manx jconfig.sas \
+        jconfig.vms
+CONFIGUREFILES= config.guess config.sub install-sh ltconfig ltmain.sh
+OTHERFILES= jconfig.doc ckconfig.c ansi2knr.c ansi2knr.1 jmemdosa.asm
+TESTFILES= testorig.jpg testimg.ppm testimg.bmp testimg.jpg testprog.jpg \
+        testimgp.jpg
+DISTFILES= $(DOCS) $(MKFILES) $(CONFIGFILES) $(SOURCES) $(INCLUDES) \
+        $(CONFIGUREFILES) $(OTHERFILES) $(TESTFILES)
+# library object files common to compression and decompression
+COMOBJECTS= jcomapi.$(O) jutils.$(O) jerror.$(O) jmemmgr.$(O) $(SYSDEPMEM)
+# compression library object files
+CLIBOBJECTS= jcapimin.$(O) jcapistd.$(O) jctrans.$(O) jcparam.$(O) \
+        jdatadst.$(O) jcinit.$(O) jcmaster.$(O) jcmarker.$(O) jcmainct.$(O) \
+        jcprepct.$(O) jccoefct.$(O) jccolor.$(O) jcsample.$(O) jchuff.$(O) \
+        jcphuff.$(O) jcdctmgr.$(O) jfdctfst.$(O) jfdctflt.$(O) \
+        jfdctint.$(O)
+# decompression library object files
+DLIBOBJECTS= jdapimin.$(O) jdapistd.$(O) jdtrans.$(O) jdatasrc.$(O) \
+        jdmaster.$(O) jdinput.$(O) jdmarker.$(O) jdhuff.$(O) jdphuff.$(O) \
+        jdmainct.$(O) jdcoefct.$(O) jdpostct.$(O) jddctmgr.$(O) \
+        jidctfst.$(O) jidctflt.$(O) jidctint.$(O) jidctred.$(O) \
+        jdsample.$(O) jdcolor.$(O) jquant1.$(O) jquant2.$(O) jdmerge.$(O)
+# These objectfiles are included in libjpeg.a
+LIBOBJECTS= $(CLIBOBJECTS) $(DLIBOBJECTS) $(COMOBJECTS)
+# object files for sample applications (excluding library files)
+COBJECTS= cjpeg.$(O) rdppm.$(O) rdgif.$(O) rdtarga.$(O) rdrle.$(O) \
+        rdbmp.$(O) rdswitch.$(O) cdjpeg.$(O)
+DOBJECTS= djpeg.$(O) wrppm.$(O) wrgif.$(O) wrtarga.$(O) wrrle.$(O) \
+        wrbmp.$(O) rdcolmap.$(O) cdjpeg.$(O)
+TROBJECTS= jpegtran.$(O) rdswitch.$(O) cdjpeg.$(O) transupp.$(O)
+
+
+all: @A2K_DEPS@ libjpeg.$(A) cjpeg djpeg jpegtran rdjpgcom wrjpgcom
+
+# Special compilation rules to support ansi2knr and libtool.
+.SUFFIXES: .lo .la
+
+# How to compile with libtool.
+@COM_LT@.c.lo:
+@COM_LT@	$(LIBTOOL) --mode=compile $(CC) $(CFLAGS) -c $(srcdir)/$*.c
+
+# How to use ansi2knr, when not using libtool.
+@COM_A2K@.c.o:
+@COM_A2K@	./ansi2knr $(srcdir)/$*.c knr/$*.c
+@COM_A2K@	$(CC) $(CFLAGS) -c knr/$*.c
+@COM_A2K@	$(RM) knr/$*.c
+
+# How to use ansi2knr AND libtool.
+@COM_A2K@.c.lo:
+@COM_A2K@	./ansi2knr $(srcdir)/$*.c knr/$*.c
+@COM_A2K@	$(LIBTOOL) --mode=compile $(CC) $(CFLAGS) -c knr/$*.c
+@COM_A2K@	$(RM) knr/$*.c
+
+ansi2knr: ansi2knr.c
+	$(CC) $(CFLAGS) $(ANSI2KNRFLAGS) -o ansi2knr $(srcdir)/ansi2knr.c
+	$(MKDIR) knr
+
+# the library:
+
+# without libtool:
+libjpeg.a: @A2K_DEPS@ $(LIBOBJECTS)
+	$(RM) libjpeg.a
+	$(AR) libjpeg.a  $(LIBOBJECTS)
+	$(AR2) libjpeg.a
+
+# with libtool:
+libjpeg.la: @A2K_DEPS@ $(LIBOBJECTS)
+	$(LIBTOOL) --mode=link $(CC) -o libjpeg.la $(LIBOBJECTS) \
+		-rpath $(libdir) -version-info $(JPEG_LIB_VERSION)
+
+# sample programs:
+
+cjpeg: $(COBJECTS) libjpeg.$(A)
+	$(LN) $(LDFLAGS) -o cjpeg $(COBJECTS) libjpeg.$(A) $(LDLIBS)
+
+djpeg: $(DOBJECTS) libjpeg.$(A)
+	$(LN) $(LDFLAGS) -o djpeg $(DOBJECTS) libjpeg.$(A) $(LDLIBS)
+
+jpegtran: $(TROBJECTS) libjpeg.$(A)
+	$(LN) $(LDFLAGS) -o jpegtran $(TROBJECTS) libjpeg.$(A) $(LDLIBS)
+
+rdjpgcom: rdjpgcom.$(O)
+	$(LN) $(LDFLAGS) -o rdjpgcom rdjpgcom.$(O) $(LDLIBS)
+
+wrjpgcom: wrjpgcom.$(O)
+	$(LN) $(LDFLAGS) -o wrjpgcom wrjpgcom.$(O) $(LDLIBS)
+
+# Installation rules:
+
+install: cjpeg djpeg jpegtran rdjpgcom wrjpgcom @FORCE_INSTALL_LIB@
+	$(INSTALL_PROGRAM) cjpeg $(bindir)/$(binprefix)cjpeg
+	$(INSTALL_PROGRAM) djpeg $(bindir)/$(binprefix)djpeg
+	$(INSTALL_PROGRAM) jpegtran $(bindir)/$(binprefix)jpegtran
+	$(INSTALL_PROGRAM) rdjpgcom $(bindir)/$(binprefix)rdjpgcom
+	$(INSTALL_PROGRAM) wrjpgcom $(bindir)/$(binprefix)wrjpgcom
+	$(INSTALL_DATA) $(srcdir)/cjpeg.1 $(mandir)/$(manprefix)cjpeg.$(manext)
+	$(INSTALL_DATA) $(srcdir)/djpeg.1 $(mandir)/$(manprefix)djpeg.$(manext)
+	$(INSTALL_DATA) $(srcdir)/jpegtran.1 $(mandir)/$(manprefix)jpegtran.$(manext)
+	$(INSTALL_DATA) $(srcdir)/rdjpgcom.1 $(mandir)/$(manprefix)rdjpgcom.$(manext)
+	$(INSTALL_DATA) $(srcdir)/wrjpgcom.1 $(mandir)/$(manprefix)wrjpgcom.$(manext)
+
+install-lib: libjpeg.$(A) install-headers
+	$(INSTALL_LIB) libjpeg.$(A) $(libdir)/$(binprefix)libjpeg.$(A)
+
+install-headers: jconfig.h
+	$(INSTALL_DATA) jconfig.h $(includedir)/jconfig.h
+	$(INSTALL_DATA) $(srcdir)/jpeglib.h $(includedir)/jpeglib.h
+	$(INSTALL_DATA) $(srcdir)/jmorecfg.h $(includedir)/jmorecfg.h
+	$(INSTALL_DATA) $(srcdir)/jerror.h $(includedir)/jerror.h
+
+clean:
+	$(RM) *.o *.lo libjpeg.a libjpeg.la
+	$(RM) cjpeg djpeg jpegtran rdjpgcom wrjpgcom
+	$(RM) ansi2knr core testout* config.log config.status
+	$(RM) -r knr .libs _libs
+
+distclean: clean
+	$(RM) Makefile jconfig.h libtool config.cache
+
+test: cjpeg djpeg jpegtran
+	$(RM) testout*
+	./djpeg -dct int -ppm -outfile testout.ppm  $(srcdir)/testorig.jpg
+	./djpeg -dct int -bmp -colors 256 -outfile testout.bmp  $(srcdir)/testorig.jpg
+	./cjpeg -dct int -outfile testout.jpg  $(srcdir)/testimg.ppm
+	./djpeg -dct int -ppm -outfile testoutp.ppm $(srcdir)/testprog.jpg
+	./cjpeg -dct int -progressive -opt -outfile testoutp.jpg $(srcdir)/testimg.ppm
+	./jpegtran -outfile testoutt.jpg $(srcdir)/testprog.jpg
+	cmp $(srcdir)/testimg.ppm testout.ppm
+	cmp $(srcdir)/testimg.bmp testout.bmp
+	cmp $(srcdir)/testimg.jpg testout.jpg
+	cmp $(srcdir)/testimg.ppm testoutp.ppm
+	cmp $(srcdir)/testimgp.jpg testoutp.jpg
+	cmp $(srcdir)/testorig.jpg testoutt.jpg
+
+check: test
+
+# Mistake catcher:
+
+jconfig.h: jconfig.doc
+	echo You must prepare a system-dependent jconfig.h file.
+	echo Please read the installation directions in install.doc.
+	exit 1
+
+# GNU Make likes to know which target names are not really files to be made:
+.PHONY: all install install-lib install-headers clean distclean test check
+
+
+jcapimin.$(O): jcapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcapistd.$(O): jcapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccoefct.$(O): jccoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccolor.$(O): jccolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcdctmgr.$(O): jcdctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jchuff.$(O): jchuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcinit.$(O): jcinit.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmainct.$(O): jcmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmarker.$(O): jcmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmaster.$(O): jcmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcomapi.$(O): jcomapi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcparam.$(O): jcparam.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcphuff.$(O): jcphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcprepct.$(O): jcprepct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcsample.$(O): jcsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jctrans.$(O): jctrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapimin.$(O): jdapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapistd.$(O): jdapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdatadst.$(O): jdatadst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdatasrc.$(O): jdatasrc.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdcoefct.$(O): jdcoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdcolor.$(O): jdcolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jddctmgr.$(O): jddctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jdhuff.$(O): jdhuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdinput.$(O): jdinput.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmainct.$(O): jdmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmarker.$(O): jdmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmaster.$(O): jdmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmerge.$(O): jdmerge.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdphuff.$(O): jdphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdpostct.$(O): jdpostct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdsample.$(O): jdsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdtrans.$(O): jdtrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jerror.$(O): jerror.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jversion.h jerror.h
+jfdctflt.$(O): jfdctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctfst.$(O): jfdctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctint.$(O): jfdctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctflt.$(O): jidctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctfst.$(O): jidctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctint.$(O): jidctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctred.$(O): jidctred.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jquant1.$(O): jquant1.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jquant2.$(O): jquant2.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jutils.$(O): jutils.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jmemmgr.$(O): jmemmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemansi.$(O): jmemansi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemname.$(O): jmemname.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemnobs.$(O): jmemnobs.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemdos.$(O): jmemdos.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemmac.$(O): jmemmac.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+cjpeg.$(O): cjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+djpeg.$(O): djpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+jpegtran.$(O): jpegtran.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h transupp.h jversion.h
+rdjpgcom.$(O): rdjpgcom.c jinclude.h jconfig.h
+wrjpgcom.$(O): wrjpgcom.c jinclude.h jconfig.h
+cdjpeg.$(O): cdjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdcolmap.$(O): rdcolmap.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdswitch.$(O): rdswitch.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+transupp.$(O): transupp.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h transupp.h
+rdppm.$(O): rdppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrppm.$(O): wrppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdgif.$(O): rdgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrgif.$(O): wrgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdtarga.$(O): rdtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrtarga.$(O): wrtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdbmp.$(O): rdbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrbmp.$(O): wrbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdrle.$(O): rdrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrrle.$(O): wrrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
diff --git a/libjpegtwrp/makefile.dj b/libjpegtwrp/makefile.dj
new file mode 100644
index 000000000..f766d25e3
--- /dev/null
+++ b/libjpegtwrp/makefile.dj
@@ -0,0 +1,220 @@
+# Makefile for Independent JPEG Group's software
+
+# This makefile is for DJGPP (Delorie's GNU C port on MS-DOS), v2.0 or later.
+# Thanks to Frank J. Donahoe for this version.
+
+# Read installation instructions before saying "make" !!
+
+# The name of your C compiler:
+CC= gcc
+
+# You may need to adjust these cc options:
+CFLAGS= -O2 -Wall -I.
+# Generally, we recommend defining any configuration symbols in jconfig.h,
+# NOT via -D switches here.
+
+# Link-time cc options:
+LDFLAGS= -s
+
+# To link any special libraries, add the necessary -l commands here.
+LDLIBS= 
+
+# Put here the object file name for the correct system-dependent memory
+# manager file.  For DJGPP this is usually jmemnobs.o, but you could
+# use jmemname.o if you want to use named temp files instead of swap space.
+SYSDEPMEM= jmemnobs.o
+
+# miscellaneous OS-dependent stuff
+# linker
+LN= $(CC)
+# file deletion command
+RM= del
+# library (.a) file creation command
+AR= ar rc
+# second step in .a creation (use "touch" if not needed)
+AR2= ranlib
+
+# End of configurable options.
+
+
+# source files: JPEG library proper
+LIBSOURCES= jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c \
+        jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c \
+        jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c \
+        jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c \
+        jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c \
+        jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c \
+        jfdctint.c jidctflt.c jidctfst.c jidctint.c jidctred.c jquant1.c \
+        jquant2.c jutils.c jmemmgr.c
+# memmgr back ends: compile only one of these into a working library
+SYSDEPSOURCES= jmemansi.c jmemname.c jmemnobs.c jmemdos.c jmemmac.c
+# source files: cjpeg/djpeg/jpegtran applications, also rdjpgcom/wrjpgcom
+APPSOURCES= cjpeg.c djpeg.c jpegtran.c rdjpgcom.c wrjpgcom.c cdjpeg.c \
+        rdcolmap.c rdswitch.c transupp.c rdppm.c wrppm.c rdgif.c wrgif.c \
+        rdtarga.c wrtarga.c rdbmp.c wrbmp.c rdrle.c wrrle.c
+SOURCES= $(LIBSOURCES) $(SYSDEPSOURCES) $(APPSOURCES)
+# files included by source files
+INCLUDES= jchuff.h jdhuff.h jdct.h jerror.h jinclude.h jmemsys.h jmorecfg.h \
+        jpegint.h jpeglib.h jversion.h cdjpeg.h cderror.h transupp.h
+# documentation, test, and support files
+DOCS= README install.doc usage.doc cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 \
+        wrjpgcom.1 wizard.doc example.c libjpeg.doc structure.doc \
+        coderules.doc filelist.doc change.log
+MKFILES= configure makefile.cfg makefile.ansi makefile.unix makefile.bcc \
+        makefile.mc6 makefile.dj makefile.wat makefile.vc makelib.ds \
+        makeapps.ds makeproj.mac makcjpeg.st makdjpeg.st makljpeg.st \
+        maktjpeg.st makefile.manx makefile.sas makefile.mms makefile.vms \
+        makvms.opt
+CONFIGFILES= jconfig.cfg jconfig.bcc jconfig.mc6 jconfig.dj jconfig.wat \
+        jconfig.vc jconfig.mac jconfig.st jconfig.manx jconfig.sas \
+        jconfig.vms
+CONFIGUREFILES= config.guess config.sub install-sh ltconfig ltmain.sh
+OTHERFILES= jconfig.doc ckconfig.c ansi2knr.c ansi2knr.1 jmemdosa.asm
+TESTFILES= testorig.jpg testimg.ppm testimg.bmp testimg.jpg testprog.jpg \
+        testimgp.jpg
+DISTFILES= $(DOCS) $(MKFILES) $(CONFIGFILES) $(SOURCES) $(INCLUDES) \
+        $(CONFIGUREFILES) $(OTHERFILES) $(TESTFILES)
+# library object files common to compression and decompression
+COMOBJECTS= jcomapi.o jutils.o jerror.o jmemmgr.o $(SYSDEPMEM)
+# compression library object files
+CLIBOBJECTS= jcapimin.o jcapistd.o jctrans.o jcparam.o jdatadst.o jcinit.o \
+        jcmaster.o jcmarker.o jcmainct.o jcprepct.o jccoefct.o jccolor.o \
+        jcsample.o jchuff.o jcphuff.o jcdctmgr.o jfdctfst.o jfdctflt.o \
+        jfdctint.o
+# decompression library object files
+DLIBOBJECTS= jdapimin.o jdapistd.o jdtrans.o jdatasrc.o jdmaster.o \
+        jdinput.o jdmarker.o jdhuff.o jdphuff.o jdmainct.o jdcoefct.o \
+        jdpostct.o jddctmgr.o jidctfst.o jidctflt.o jidctint.o jidctred.o \
+        jdsample.o jdcolor.o jquant1.o jquant2.o jdmerge.o
+# These objectfiles are included in libjpeg.a
+LIBOBJECTS= $(CLIBOBJECTS) $(DLIBOBJECTS) $(COMOBJECTS)
+# object files for sample applications (excluding library files)
+COBJECTS= cjpeg.o rdppm.o rdgif.o rdtarga.o rdrle.o rdbmp.o rdswitch.o \
+        cdjpeg.o
+DOBJECTS= djpeg.o wrppm.o wrgif.o wrtarga.o wrrle.o wrbmp.o rdcolmap.o \
+        cdjpeg.o
+TROBJECTS= jpegtran.o rdswitch.o cdjpeg.o transupp.o
+
+
+all: libjpeg.a cjpeg.exe djpeg.exe jpegtran.exe rdjpgcom.exe wrjpgcom.exe
+
+libjpeg.a: $(LIBOBJECTS)
+	$(RM) libjpeg.a
+	$(AR) libjpeg.a  $(LIBOBJECTS)
+	$(AR2) libjpeg.a
+
+cjpeg.exe: $(COBJECTS) libjpeg.a
+	$(LN) $(LDFLAGS) -o cjpeg.exe $(COBJECTS) libjpeg.a $(LDLIBS)
+
+djpeg.exe: $(DOBJECTS) libjpeg.a
+	$(LN) $(LDFLAGS) -o djpeg.exe $(DOBJECTS) libjpeg.a $(LDLIBS)
+
+jpegtran.exe: $(TROBJECTS) libjpeg.a
+	$(LN) $(LDFLAGS) -o jpegtran.exe $(TROBJECTS) libjpeg.a $(LDLIBS)
+
+rdjpgcom.exe: rdjpgcom.o
+	$(LN) $(LDFLAGS) -o rdjpgcom.exe rdjpgcom.o $(LDLIBS)
+
+wrjpgcom.exe: wrjpgcom.o
+	$(LN) $(LDFLAGS) -o wrjpgcom.exe wrjpgcom.o $(LDLIBS)
+
+jconfig.h: jconfig.doc
+	echo You must prepare a system-dependent jconfig.h file.
+	echo Please read the installation directions in install.doc.
+	exit 1
+
+clean:
+	$(RM) *.o
+	$(RM) cjpeg.exe
+	$(RM) djpeg.exe
+	$(RM) jpegtran.exe
+	$(RM) rdjpgcom.exe
+	$(RM) wrjpgcom.exe
+	$(RM) libjpeg.a
+	$(RM) testout*.*
+
+test: cjpeg.exe djpeg.exe jpegtran.exe
+	$(RM) testout*.*
+	./djpeg -dct int -ppm -outfile testout.ppm  testorig.jpg
+	./djpeg -dct int -bmp -colors 256 -outfile testout.bmp  testorig.jpg
+	./cjpeg -dct int -outfile testout.jpg  testimg.ppm
+	./djpeg -dct int -ppm -outfile testoutp.ppm testprog.jpg
+	./cjpeg -dct int -progressive -opt -outfile testoutp.jpg testimg.ppm
+	./jpegtran -outfile testoutt.jpg testprog.jpg
+	fc /b testimg.ppm testout.ppm
+	fc /b testimg.bmp testout.bmp
+	fc /b testimg.jpg testout.jpg
+	fc /b testimg.ppm testoutp.ppm
+	fc /b testimgp.jpg testoutp.jpg
+	fc /b testorig.jpg testoutt.jpg
+
+
+jcapimin.o: jcapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcapistd.o: jcapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccoefct.o: jccoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccolor.o: jccolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcdctmgr.o: jcdctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jchuff.o: jchuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcinit.o: jcinit.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmainct.o: jcmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmarker.o: jcmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmaster.o: jcmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcomapi.o: jcomapi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcparam.o: jcparam.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcphuff.o: jcphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcprepct.o: jcprepct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcsample.o: jcsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jctrans.o: jctrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapimin.o: jdapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapistd.o: jdapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdatadst.o: jdatadst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdatasrc.o: jdatasrc.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdcoefct.o: jdcoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdcolor.o: jdcolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jddctmgr.o: jddctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jdhuff.o: jdhuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdinput.o: jdinput.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmainct.o: jdmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmarker.o: jdmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmaster.o: jdmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmerge.o: jdmerge.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdphuff.o: jdphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdpostct.o: jdpostct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdsample.o: jdsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdtrans.o: jdtrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jerror.o: jerror.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jversion.h jerror.h
+jfdctflt.o: jfdctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctfst.o: jfdctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctint.o: jfdctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctflt.o: jidctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctfst.o: jidctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctint.o: jidctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctred.o: jidctred.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jquant1.o: jquant1.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jquant2.o: jquant2.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jutils.o: jutils.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jmemmgr.o: jmemmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemansi.o: jmemansi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemname.o: jmemname.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemnobs.o: jmemnobs.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemdos.o: jmemdos.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemmac.o: jmemmac.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+cjpeg.o: cjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+djpeg.o: djpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+jpegtran.o: jpegtran.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h transupp.h jversion.h
+rdjpgcom.o: rdjpgcom.c jinclude.h jconfig.h
+wrjpgcom.o: wrjpgcom.c jinclude.h jconfig.h
+cdjpeg.o: cdjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdcolmap.o: rdcolmap.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdswitch.o: rdswitch.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+transupp.o: transupp.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h transupp.h
+rdppm.o: rdppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrppm.o: wrppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdgif.o: rdgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrgif.o: wrgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdtarga.o: rdtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrtarga.o: wrtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdbmp.o: rdbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrbmp.o: wrbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdrle.o: rdrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrrle.o: wrrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
diff --git a/libjpegtwrp/makefile.manx b/libjpegtwrp/makefile.manx
new file mode 100644
index 000000000..4cb42d17c
--- /dev/null
+++ b/libjpegtwrp/makefile.manx
@@ -0,0 +1,214 @@
+# Makefile for Independent JPEG Group's software
+
+# This makefile is for Amiga systems using Manx Aztec C ver 5.x.
+# Thanks to D.J. James (djjames@cup.portal.com) for this version.
+
+# Read installation instructions before saying "make" !!
+
+# The name of your C compiler:
+CC= cc
+
+# You may need to adjust these cc options:
+# Uncomment for generic 68000 code (will work on any Amiga)
+ARCHFLAGS= -sn
+
+# Uncomment for 68020/68030 code (faster, but won't run on 68000 CPU)
+#ARCHFLAGS= -c2
+
+CFLAGS= -MC -MD $(ARCHFLAGS) -spfam -r4
+
+# Link-time cc options:
+LDFLAGS= -g
+
+# To link any special libraries, add the necessary -l commands here.
+LDLIBS= -lml -lcl
+
+# Put here the object file name for the correct system-dependent memory
+# manager file.  For Amiga we recommend jmemname.o.
+SYSDEPMEM= jmemname.o
+
+# miscellaneous OS-dependent stuff
+# linker
+LN= ln
+# file deletion command
+RM= delete quiet
+# library (.lib) file creation command
+AR= lb
+
+# End of configurable options.
+
+
+# source files: JPEG library proper
+LIBSOURCES= jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c \
+        jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c \
+        jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c \
+        jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c \
+        jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c \
+        jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c \
+        jfdctint.c jidctflt.c jidctfst.c jidctint.c jidctred.c jquant1.c \
+        jquant2.c jutils.c jmemmgr.c
+# memmgr back ends: compile only one of these into a working library
+SYSDEPSOURCES= jmemansi.c jmemname.c jmemnobs.c jmemdos.c jmemmac.c
+# source files: cjpeg/djpeg/jpegtran applications, also rdjpgcom/wrjpgcom
+APPSOURCES= cjpeg.c djpeg.c jpegtran.c rdjpgcom.c wrjpgcom.c cdjpeg.c \
+        rdcolmap.c rdswitch.c transupp.c rdppm.c wrppm.c rdgif.c wrgif.c \
+        rdtarga.c wrtarga.c rdbmp.c wrbmp.c rdrle.c wrrle.c
+SOURCES= $(LIBSOURCES) $(SYSDEPSOURCES) $(APPSOURCES)
+# files included by source files
+INCLUDES= jchuff.h jdhuff.h jdct.h jerror.h jinclude.h jmemsys.h jmorecfg.h \
+        jpegint.h jpeglib.h jversion.h cdjpeg.h cderror.h transupp.h
+# documentation, test, and support files
+DOCS= README install.doc usage.doc cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 \
+        wrjpgcom.1 wizard.doc example.c libjpeg.doc structure.doc \
+        coderules.doc filelist.doc change.log
+MKFILES= configure makefile.cfg makefile.ansi makefile.unix makefile.bcc \
+        makefile.mc6 makefile.dj makefile.wat makefile.vc makelib.ds \
+        makeapps.ds makeproj.mac makcjpeg.st makdjpeg.st makljpeg.st \
+        maktjpeg.st makefile.manx makefile.sas makefile.mms makefile.vms \
+        makvms.opt
+CONFIGFILES= jconfig.cfg jconfig.bcc jconfig.mc6 jconfig.dj jconfig.wat \
+        jconfig.vc jconfig.mac jconfig.st jconfig.manx jconfig.sas \
+        jconfig.vms
+CONFIGUREFILES= config.guess config.sub install-sh ltconfig ltmain.sh
+OTHERFILES= jconfig.doc ckconfig.c ansi2knr.c ansi2knr.1 jmemdosa.asm
+TESTFILES= testorig.jpg testimg.ppm testimg.bmp testimg.jpg testprog.jpg \
+        testimgp.jpg
+DISTFILES= $(DOCS) $(MKFILES) $(CONFIGFILES) $(SOURCES) $(INCLUDES) \
+        $(CONFIGUREFILES) $(OTHERFILES) $(TESTFILES)
+# library object files common to compression and decompression
+COMOBJECTS= jcomapi.o jutils.o jerror.o jmemmgr.o $(SYSDEPMEM)
+# compression library object files
+CLIBOBJECTS= jcapimin.o jcapistd.o jctrans.o jcparam.o jdatadst.o jcinit.o \
+        jcmaster.o jcmarker.o jcmainct.o jcprepct.o jccoefct.o jccolor.o \
+        jcsample.o jchuff.o jcphuff.o jcdctmgr.o jfdctfst.o jfdctflt.o \
+        jfdctint.o
+# decompression library object files
+DLIBOBJECTS= jdapimin.o jdapistd.o jdtrans.o jdatasrc.o jdmaster.o \
+        jdinput.o jdmarker.o jdhuff.o jdphuff.o jdmainct.o jdcoefct.o \
+        jdpostct.o jddctmgr.o jidctfst.o jidctflt.o jidctint.o jidctred.o \
+        jdsample.o jdcolor.o jquant1.o jquant2.o jdmerge.o
+# These objectfiles are included in libjpeg.lib
+LIBOBJECTS= $(CLIBOBJECTS) $(DLIBOBJECTS) $(COMOBJECTS)
+# object files for sample applications (excluding library files)
+COBJECTS= cjpeg.o rdppm.o rdgif.o rdtarga.o rdrle.o rdbmp.o rdswitch.o \
+        cdjpeg.o
+DOBJECTS= djpeg.o wrppm.o wrgif.o wrtarga.o wrrle.o wrbmp.o rdcolmap.o \
+        cdjpeg.o
+TROBJECTS= jpegtran.o rdswitch.o cdjpeg.o transupp.o
+
+
+all: libjpeg.lib cjpeg djpeg jpegtran rdjpgcom wrjpgcom
+
+libjpeg.lib: $(LIBOBJECTS)
+	-$(RM) libjpeg.lib
+	$(AR) libjpeg.lib  $(LIBOBJECTS)
+
+cjpeg: $(COBJECTS) libjpeg.lib
+	$(LN) $(LDFLAGS) -o cjpeg $(COBJECTS) libjpeg.lib $(LDLIBS)
+
+djpeg: $(DOBJECTS) libjpeg.lib
+	$(LN) $(LDFLAGS) -o djpeg $(DOBJECTS) libjpeg.lib $(LDLIBS)
+
+jpegtran: $(TROBJECTS) libjpeg.lib
+	$(LN) $(LDFLAGS) -o jpegtran $(TROBJECTS) libjpeg.lib $(LDLIBS)
+
+rdjpgcom: rdjpgcom.o
+	$(LN) $(LDFLAGS) -o rdjpgcom rdjpgcom.o $(LDLIBS)
+
+wrjpgcom: wrjpgcom.o
+	$(LN) $(LDFLAGS) -o wrjpgcom wrjpgcom.o $(LDLIBS)
+
+jconfig.h: jconfig.doc
+	echo You must prepare a system-dependent jconfig.h file.
+	echo Please read the installation directions in install.doc.
+	exit 1
+
+clean:
+	-$(RM) *.o cjpeg djpeg jpegtran libjpeg.lib rdjpgcom wrjpgcom
+	-$(RM) core testout*.*
+
+test: cjpeg djpeg jpegtran
+	-$(RM) testout*.*
+	djpeg -dct int -ppm -outfile testout.ppm  testorig.jpg
+	djpeg -dct int -bmp -colors 256 -outfile testout.bmp  testorig.jpg
+	cjpeg -dct int -outfile testout.jpg  testimg.ppm
+	djpeg -dct int -ppm -outfile testoutp.ppm testprog.jpg
+	cjpeg -dct int -progressive -opt -outfile testoutp.jpg testimg.ppm
+	jpegtran -outfile testoutt.jpg testprog.jpg
+	cmp testimg.ppm testout.ppm
+	cmp testimg.bmp testout.bmp
+	cmp testimg.jpg testout.jpg
+	cmp testimg.ppm testoutp.ppm
+	cmp testimgp.jpg testoutp.jpg
+	cmp testorig.jpg testoutt.jpg
+
+
+jcapimin.o: jcapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcapistd.o: jcapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccoefct.o: jccoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccolor.o: jccolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcdctmgr.o: jcdctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jchuff.o: jchuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcinit.o: jcinit.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmainct.o: jcmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmarker.o: jcmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmaster.o: jcmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcomapi.o: jcomapi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcparam.o: jcparam.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcphuff.o: jcphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcprepct.o: jcprepct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcsample.o: jcsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jctrans.o: jctrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapimin.o: jdapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapistd.o: jdapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdatadst.o: jdatadst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdatasrc.o: jdatasrc.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdcoefct.o: jdcoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdcolor.o: jdcolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jddctmgr.o: jddctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jdhuff.o: jdhuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdinput.o: jdinput.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmainct.o: jdmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmarker.o: jdmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmaster.o: jdmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmerge.o: jdmerge.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdphuff.o: jdphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdpostct.o: jdpostct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdsample.o: jdsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdtrans.o: jdtrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jerror.o: jerror.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jversion.h jerror.h
+jfdctflt.o: jfdctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctfst.o: jfdctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctint.o: jfdctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctflt.o: jidctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctfst.o: jidctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctint.o: jidctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctred.o: jidctred.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jquant1.o: jquant1.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jquant2.o: jquant2.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jutils.o: jutils.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jmemmgr.o: jmemmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemansi.o: jmemansi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemname.o: jmemname.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemnobs.o: jmemnobs.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemdos.o: jmemdos.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemmac.o: jmemmac.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+cjpeg.o: cjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+djpeg.o: djpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+jpegtran.o: jpegtran.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h transupp.h jversion.h
+rdjpgcom.o: rdjpgcom.c jinclude.h jconfig.h
+wrjpgcom.o: wrjpgcom.c jinclude.h jconfig.h
+cdjpeg.o: cdjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdcolmap.o: rdcolmap.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdswitch.o: rdswitch.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+transupp.o: transupp.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h transupp.h
+rdppm.o: rdppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrppm.o: wrppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdgif.o: rdgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrgif.o: wrgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdtarga.o: rdtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrtarga.o: wrtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdbmp.o: rdbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrbmp.o: wrbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdrle.o: rdrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrrle.o: wrrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
diff --git a/libjpegtwrp/makefile.mc6 b/libjpegtwrp/makefile.mc6
new file mode 100644
index 000000000..6aff05464
--- /dev/null
+++ b/libjpegtwrp/makefile.mc6
@@ -0,0 +1,249 @@
+# Makefile for Independent JPEG Group's software
+
+# This makefile is for Microsoft C for MS-DOS, version 6.00A and up.
+# Use NMAKE, not Microsoft's brain-damaged MAKE.
+# Thanks to Alan Wright and Chris Turner of Olivetti Research Ltd.
+
+# Read installation instructions before saying "nmake" !!
+
+# You may need to adjust these compiler options:
+CFLAGS = -AM -Oecigt -Gs -W3
+# -AM medium memory model (or use -AS for small model, if you remove features)
+# -Oecigt -Gs  maximum safe optimisation (-Ol has bugs in MSC 6.00A)
+# -W3 warning level 3
+# You might also want to add -G2 if you have an 80286, etc.
+# Generally, we recommend defining any configuration symbols in jconfig.h,
+# NOT via -D switches here.
+
+# Jan-Herman Buining suggests the following switches for MS C 8.0 and a 486:
+# CFLAGS = /AM /f- /FPi87 /G3 /Gs /Gy /Ob1 /Oc /Oe /Og /Oi /Ol /On /Oo /Ot \
+#          /OV4 /W3
+# except for jquant1.c, which must be compiled with /Oo- to avoid a compiler
+# crash.
+
+# Ingar Steinsland suggests the following switches when building
+# a 16-bit Windows DLL:
+# CFLAGS = -ALw -Gsw -Zpe -W3 -O2 -Zi -Zd
+
+# Put here the object file name for the correct system-dependent memory
+# manager file.  For DOS, we recommend jmemdos.c and jmemdosa.asm.
+# (But not for Windows; see install.doc if you use this makefile for Windows.)
+SYSDEPMEM= jmemdos.obj jmemdosa.obj
+# SYSDEPMEMLIB must list the same files with "+" signs for the librarian.
+SYSDEPMEMLIB= +jmemdos.obj +jmemdosa.obj
+
+# End of configurable options.
+
+
+# source files: JPEG library proper
+LIBSOURCES= jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c \
+        jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c \
+        jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c \
+        jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c \
+        jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c \
+        jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c \
+        jfdctint.c jidctflt.c jidctfst.c jidctint.c jidctred.c jquant1.c \
+        jquant2.c jutils.c jmemmgr.c
+# memmgr back ends: compile only one of these into a working library
+SYSDEPSOURCES= jmemansi.c jmemname.c jmemnobs.c jmemdos.c jmemmac.c
+# source files: cjpeg/djpeg/jpegtran applications, also rdjpgcom/wrjpgcom
+APPSOURCES= cjpeg.c djpeg.c jpegtran.c rdjpgcom.c wrjpgcom.c cdjpeg.c \
+        rdcolmap.c rdswitch.c transupp.c rdppm.c wrppm.c rdgif.c wrgif.c \
+        rdtarga.c wrtarga.c rdbmp.c wrbmp.c rdrle.c wrrle.c
+SOURCES= $(LIBSOURCES) $(SYSDEPSOURCES) $(APPSOURCES)
+# files included by source files
+INCLUDES= jchuff.h jdhuff.h jdct.h jerror.h jinclude.h jmemsys.h jmorecfg.h \
+        jpegint.h jpeglib.h jversion.h cdjpeg.h cderror.h transupp.h
+# documentation, test, and support files
+DOCS= README install.doc usage.doc cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 \
+        wrjpgcom.1 wizard.doc example.c libjpeg.doc structure.doc \
+        coderules.doc filelist.doc change.log
+MKFILES= configure makefile.cfg makefile.ansi makefile.unix makefile.bcc \
+        makefile.mc6 makefile.dj makefile.wat makefile.vc makelib.ds \
+        makeapps.ds makeproj.mac makcjpeg.st makdjpeg.st makljpeg.st \
+        maktjpeg.st makefile.manx makefile.sas makefile.mms makefile.vms \
+        makvms.opt
+CONFIGFILES= jconfig.cfg jconfig.bcc jconfig.mc6 jconfig.dj jconfig.wat \
+        jconfig.vc jconfig.mac jconfig.st jconfig.manx jconfig.sas \
+        jconfig.vms
+CONFIGUREFILES= config.guess config.sub install-sh ltconfig ltmain.sh
+OTHERFILES= jconfig.doc ckconfig.c ansi2knr.c ansi2knr.1 jmemdosa.asm
+TESTFILES= testorig.jpg testimg.ppm testimg.bmp testimg.jpg testprog.jpg \
+        testimgp.jpg
+DISTFILES= $(DOCS) $(MKFILES) $(CONFIGFILES) $(SOURCES) $(INCLUDES) \
+        $(CONFIGUREFILES) $(OTHERFILES) $(TESTFILES)
+# library object files common to compression and decompression
+COMOBJECTS= jcomapi.obj jutils.obj jerror.obj jmemmgr.obj $(SYSDEPMEM)
+# compression library object files
+CLIBOBJECTS= jcapimin.obj jcapistd.obj jctrans.obj jcparam.obj jdatadst.obj \
+        jcinit.obj jcmaster.obj jcmarker.obj jcmainct.obj jcprepct.obj \
+        jccoefct.obj jccolor.obj jcsample.obj jchuff.obj jcphuff.obj \
+        jcdctmgr.obj jfdctfst.obj jfdctflt.obj jfdctint.obj
+# decompression library object files
+DLIBOBJECTS= jdapimin.obj jdapistd.obj jdtrans.obj jdatasrc.obj \
+        jdmaster.obj jdinput.obj jdmarker.obj jdhuff.obj jdphuff.obj \
+        jdmainct.obj jdcoefct.obj jdpostct.obj jddctmgr.obj jidctfst.obj \
+        jidctflt.obj jidctint.obj jidctred.obj jdsample.obj jdcolor.obj \
+        jquant1.obj jquant2.obj jdmerge.obj
+# These objectfiles are included in libjpeg.lib
+LIBOBJECTS= $(CLIBOBJECTS) $(DLIBOBJECTS) $(COMOBJECTS)
+# object files for sample applications (excluding library files)
+COBJECTS= cjpeg.obj rdppm.obj rdgif.obj rdtarga.obj rdrle.obj rdbmp.obj \
+        rdswitch.obj cdjpeg.obj
+DOBJECTS= djpeg.obj wrppm.obj wrgif.obj wrtarga.obj wrrle.obj wrbmp.obj \
+        rdcolmap.obj cdjpeg.obj
+TROBJECTS= jpegtran.obj rdswitch.obj cdjpeg.obj transupp.obj
+
+# need linker response file because file list > 128 chars
+RFILE = libjpeg.ans
+
+
+all: libjpeg.lib cjpeg.exe djpeg.exe jpegtran.exe rdjpgcom.exe wrjpgcom.exe
+
+libjpeg.lib: $(LIBOBJECTS) $(RFILE)
+	del libjpeg.lib
+	lib @$(RFILE)
+
+# linker response file for building libjpeg.lib
+$(RFILE) : makefile
+	del $(RFILE)
+	echo libjpeg.lib >$(RFILE)
+# silly want-to-create-it prompt:
+	echo y >>$(RFILE)
+	echo +jcapimin.obj +jcapistd.obj +jctrans.obj +jcparam.obj & >>$(RFILE)
+	echo +jdatadst.obj +jcinit.obj +jcmaster.obj +jcmarker.obj & >>$(RFILE)
+	echo +jcmainct.obj +jcprepct.obj +jccoefct.obj & >>$(RFILE)
+	echo +jccolor.obj +jcsample.obj +jchuff.obj +jcphuff.obj & >>$(RFILE)
+	echo +jcdctmgr.obj +jfdctfst.obj +jfdctflt.obj & >>$(RFILE)
+	echo +jfdctint.obj +jdapimin.obj +jdapistd.obj & >>$(RFILE)
+	echo +jdtrans.obj +jdatasrc.obj +jdmaster.obj +jdinput.obj & >>$(RFILE)
+	echo +jdmarker.obj +jdhuff.obj +jdphuff.obj +jdmainct.obj & >>$(RFILE)
+	echo +jdcoefct.obj +jdpostct.obj +jddctmgr.obj & >>$(RFILE)
+	echo +jidctfst.obj +jidctflt.obj +jidctint.obj & >>$(RFILE)
+	echo +jidctred.obj +jdsample.obj +jdcolor.obj +jquant1.obj & >>$(RFILE)
+	echo +jquant2.obj +jdmerge.obj +jcomapi.obj +jutils.obj & >>$(RFILE)
+	echo +jerror.obj +jmemmgr.obj & >>$(RFILE)
+	echo $(SYSDEPMEMLIB) ; >>$(RFILE)
+
+cjpeg.exe: $(COBJECTS) libjpeg.lib
+	echo $(COBJECTS) >cjpeg.lst
+	link /STACK:4096 /EXEPACK @cjpeg.lst, cjpeg.exe, , libjpeg.lib, ;
+	del cjpeg.lst
+
+djpeg.exe: $(DOBJECTS) libjpeg.lib
+	echo $(DOBJECTS) >djpeg.lst
+	link /STACK:4096 /EXEPACK @djpeg.lst, djpeg.exe, , libjpeg.lib, ;
+	del djpeg.lst
+
+jpegtran.exe: $(TROBJECTS) libjpeg.lib
+	link /STACK:4096 /EXEPACK $(TROBJECTS), jpegtran.exe, , libjpeg.lib, ;
+
+rdjpgcom.exe: rdjpgcom.c
+	$(CC) -AS -O -W3 rdjpgcom.c
+
+# wrjpgcom needs large model so it can malloc a 64K chunk
+wrjpgcom.exe: wrjpgcom.c
+	$(CC) -AL -O -W3 wrjpgcom.c
+
+jconfig.h: jconfig.doc
+	echo You must prepare a system-dependent jconfig.h file.
+	echo Please read the installation directions in install.doc.
+	exit 1
+
+clean:
+	del *.obj
+	del libjpeg.lib
+	del cjpeg.exe
+	del djpeg.exe
+	del jpegtran.exe
+	del rdjpgcom.exe
+	del wrjpgcom.exe
+	del testout*.*
+
+test: cjpeg.exe djpeg.exe jpegtran.exe
+	del testout*.*
+	djpeg -dct int -ppm -outfile testout.ppm  testorig.jpg
+	djpeg -dct int -bmp -colors 256 -outfile testout.bmp  testorig.jpg
+	cjpeg -dct int -outfile testout.jpg  testimg.ppm
+	djpeg -dct int -ppm -outfile testoutp.ppm testprog.jpg
+	cjpeg -dct int -progressive -opt -outfile testoutp.jpg testimg.ppm
+	jpegtran -outfile testoutt.jpg testprog.jpg
+	fc /b testimg.ppm testout.ppm
+	fc /b testimg.bmp testout.bmp
+	fc /b testimg.jpg testout.jpg
+	fc /b testimg.ppm testoutp.ppm
+	fc /b testimgp.jpg testoutp.jpg
+	fc /b testorig.jpg testoutt.jpg
+
+
+jcapimin.obj: jcapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcapistd.obj: jcapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccoefct.obj: jccoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccolor.obj: jccolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcdctmgr.obj: jcdctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jchuff.obj: jchuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcinit.obj: jcinit.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmainct.obj: jcmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmarker.obj: jcmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmaster.obj: jcmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcomapi.obj: jcomapi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcparam.obj: jcparam.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcphuff.obj: jcphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcprepct.obj: jcprepct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcsample.obj: jcsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jctrans.obj: jctrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapimin.obj: jdapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapistd.obj: jdapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdatadst.obj: jdatadst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdatasrc.obj: jdatasrc.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdcoefct.obj: jdcoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdcolor.obj: jdcolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jddctmgr.obj: jddctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jdhuff.obj: jdhuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdinput.obj: jdinput.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmainct.obj: jdmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmarker.obj: jdmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmaster.obj: jdmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmerge.obj: jdmerge.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdphuff.obj: jdphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdpostct.obj: jdpostct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdsample.obj: jdsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdtrans.obj: jdtrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jerror.obj: jerror.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jversion.h jerror.h
+jfdctflt.obj: jfdctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctfst.obj: jfdctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctint.obj: jfdctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctflt.obj: jidctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctfst.obj: jidctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctint.obj: jidctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctred.obj: jidctred.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jquant1.obj: jquant1.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jquant2.obj: jquant2.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jutils.obj: jutils.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jmemmgr.obj: jmemmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemansi.obj: jmemansi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemname.obj: jmemname.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemnobs.obj: jmemnobs.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemdos.obj: jmemdos.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemmac.obj: jmemmac.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+cjpeg.obj: cjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+djpeg.obj: djpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+jpegtran.obj: jpegtran.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h transupp.h jversion.h
+rdjpgcom.obj: rdjpgcom.c jinclude.h jconfig.h
+wrjpgcom.obj: wrjpgcom.c jinclude.h jconfig.h
+cdjpeg.obj: cdjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdcolmap.obj: rdcolmap.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdswitch.obj: rdswitch.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+transupp.obj: transupp.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h transupp.h
+rdppm.obj: rdppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrppm.obj: wrppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdgif.obj: rdgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrgif.obj: wrgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdtarga.obj: rdtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrtarga.obj: wrtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdbmp.obj: rdbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrbmp.obj: wrbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdrle.obj: rdrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrrle.obj: wrrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+jmemdosa.obj : jmemdosa.asm
+	masm /mx $*;
diff --git a/libjpegtwrp/makefile.mms b/libjpegtwrp/makefile.mms
new file mode 100644
index 000000000..cf130e5b9
--- /dev/null
+++ b/libjpegtwrp/makefile.mms
@@ -0,0 +1,218 @@
+# Makefile for Independent JPEG Group's software
+
+# This makefile is for use with MMS on Digital VMS systems.
+# Thanks to Rick Dyson (dyson@iowasp.physics.uiowa.edu)
+# and Tim Bell (tbell@netcom.com) for their help.
+
+# Read installation instructions before saying "MMS" !!
+
+# You may need to adjust these cc options:
+CFLAGS= $(CFLAGS) /NoDebug /Optimize
+# Generally, we recommend defining any configuration symbols in jconfig.h,
+# NOT via /Define switches here.
+.ifdef ALPHA
+OPT=
+.else
+OPT= ,Sys$Disk:[]MAKVMS.OPT/Option
+.endif
+
+# Put here the object file name for the correct system-dependent memory
+# manager file.  For Unix this is usually jmemnobs.o, but you may want
+# to use jmemansi.o or jmemname.o if you have limited swap space.
+SYSDEPMEM= jmemnobs.obj
+
+# End of configurable options.
+
+
+# source files: JPEG library proper
+LIBSOURCES= jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c \
+        jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c \
+        jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c \
+        jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c \
+        jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c \
+        jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c \
+        jfdctint.c jidctflt.c jidctfst.c jidctint.c jidctred.c jquant1.c \
+        jquant2.c jutils.c jmemmgr.c
+# memmgr back ends: compile only one of these into a working library
+SYSDEPSOURCES= jmemansi.c jmemname.c jmemnobs.c jmemdos.c jmemmac.c
+# source files: cjpeg/djpeg/jpegtran applications, also rdjpgcom/wrjpgcom
+APPSOURCES= cjpeg.c djpeg.c jpegtran.c rdjpgcom.c wrjpgcom.c cdjpeg.c \
+        rdcolmap.c rdswitch.c transupp.c rdppm.c wrppm.c rdgif.c wrgif.c \
+        rdtarga.c wrtarga.c rdbmp.c wrbmp.c rdrle.c wrrle.c
+SOURCES= $(LIBSOURCES) $(SYSDEPSOURCES) $(APPSOURCES)
+# files included by source files
+INCLUDES= jchuff.h jdhuff.h jdct.h jerror.h jinclude.h jmemsys.h jmorecfg.h \
+        jpegint.h jpeglib.h jversion.h cdjpeg.h cderror.h transupp.h
+# documentation, test, and support files
+DOCS= README install.doc usage.doc cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 \
+        wrjpgcom.1 wizard.doc example.c libjpeg.doc structure.doc \
+        coderules.doc filelist.doc change.log
+MKFILES= configure makefile.cfg makefile.ansi makefile.unix makefile.bcc \
+        makefile.mc6 makefile.dj makefile.wat makefile.vc makelib.ds \
+        makeapps.ds makeproj.mac makcjpeg.st makdjpeg.st makljpeg.st \
+        maktjpeg.st makefile.manx makefile.sas makefile.mms makefile.vms \
+        makvms.opt
+CONFIGFILES= jconfig.cfg jconfig.bcc jconfig.mc6 jconfig.dj jconfig.wat \
+        jconfig.vc jconfig.mac jconfig.st jconfig.manx jconfig.sas \
+        jconfig.vms
+CONFIGUREFILES= config.guess config.sub install-sh ltconfig ltmain.sh
+OTHERFILES= jconfig.doc ckconfig.c ansi2knr.c ansi2knr.1 jmemdosa.asm
+TESTFILES= testorig.jpg testimg.ppm testimg.bmp testimg.jpg testprog.jpg \
+        testimgp.jpg
+DISTFILES= $(DOCS) $(MKFILES) $(CONFIGFILES) $(SOURCES) $(INCLUDES) \
+        $(CONFIGUREFILES) $(OTHERFILES) $(TESTFILES)
+# library object files common to compression and decompression
+COMOBJECTS= jcomapi.obj jutils.obj jerror.obj jmemmgr.obj $(SYSDEPMEM)
+# compression library object files
+CLIBOBJECTS= jcapimin.obj jcapistd.obj jctrans.obj jcparam.obj jdatadst.obj \
+        jcinit.obj jcmaster.obj jcmarker.obj jcmainct.obj jcprepct.obj \
+        jccoefct.obj jccolor.obj jcsample.obj jchuff.obj jcphuff.obj \
+        jcdctmgr.obj jfdctfst.obj jfdctflt.obj jfdctint.obj
+# decompression library object files
+DLIBOBJECTS= jdapimin.obj jdapistd.obj jdtrans.obj jdatasrc.obj \
+        jdmaster.obj jdinput.obj jdmarker.obj jdhuff.obj jdphuff.obj \
+        jdmainct.obj jdcoefct.obj jdpostct.obj jddctmgr.obj jidctfst.obj \
+        jidctflt.obj jidctint.obj jidctred.obj jdsample.obj jdcolor.obj \
+        jquant1.obj jquant2.obj jdmerge.obj
+# These objectfiles are included in libjpeg.olb
+LIBOBJECTS= $(CLIBOBJECTS) $(DLIBOBJECTS) $(COMOBJECTS)
+# object files for sample applications (excluding library files)
+COBJECTS= cjpeg.obj rdppm.obj rdgif.obj rdtarga.obj rdrle.obj rdbmp.obj \
+        rdswitch.obj cdjpeg.obj
+DOBJECTS= djpeg.obj wrppm.obj wrgif.obj wrtarga.obj wrrle.obj wrbmp.obj \
+        rdcolmap.obj cdjpeg.obj
+TROBJECTS= jpegtran.obj rdswitch.obj cdjpeg.obj transupp.obj
+# objectfile lists with commas --- what a crock
+COBJLIST= cjpeg.obj,rdppm.obj,rdgif.obj,rdtarga.obj,rdrle.obj,rdbmp.obj,\
+          rdswitch.obj,cdjpeg.obj
+DOBJLIST= djpeg.obj,wrppm.obj,wrgif.obj,wrtarga.obj,wrrle.obj,wrbmp.obj,\
+          rdcolmap.obj,cdjpeg.obj
+TROBJLIST= jpegtran.obj,rdswitch.obj,cdjpeg.obj,transupp.obj
+LIBOBJLIST= jcapimin.obj,jcapistd.obj,jctrans.obj,jcparam.obj,jdatadst.obj,\
+          jcinit.obj,jcmaster.obj,jcmarker.obj,jcmainct.obj,jcprepct.obj,\
+          jccoefct.obj,jccolor.obj,jcsample.obj,jchuff.obj,jcphuff.obj,\
+          jcdctmgr.obj,jfdctfst.obj,jfdctflt.obj,jfdctint.obj,jdapimin.obj,\
+          jdapistd.obj,jdtrans.obj,jdatasrc.obj,jdmaster.obj,jdinput.obj,\
+          jdmarker.obj,jdhuff.obj,jdphuff.obj,jdmainct.obj,jdcoefct.obj,\
+          jdpostct.obj,jddctmgr.obj,jidctfst.obj,jidctflt.obj,jidctint.obj,\
+          jidctred.obj,jdsample.obj,jdcolor.obj,jquant1.obj,jquant2.obj,\
+          jdmerge.obj,jcomapi.obj,jutils.obj,jerror.obj,jmemmgr.obj,$(SYSDEPMEM)
+
+
+.first
+	@- Define /NoLog Sys Sys$Library
+
+ALL : libjpeg.olb cjpeg.exe djpeg.exe jpegtran.exe rdjpgcom.exe wrjpgcom.exe
+	@ Continue
+
+libjpeg.olb : $(LIBOBJECTS)
+	Library /Create libjpeg.olb $(LIBOBJLIST)
+
+cjpeg.exe : $(COBJECTS) libjpeg.olb
+	$(LINK) $(LFLAGS) /Executable = cjpeg.exe $(COBJLIST),libjpeg.olb/Library$(OPT)
+
+djpeg.exe : $(DOBJECTS) libjpeg.olb
+	$(LINK) $(LFLAGS) /Executable = djpeg.exe $(DOBJLIST),libjpeg.olb/Library$(OPT)
+
+jpegtran.exe : $(TROBJECTS) libjpeg.olb
+	$(LINK) $(LFLAGS) /Executable = jpegtran.exe $(TROBJLIST),libjpeg.olb/Library$(OPT)
+
+rdjpgcom.exe : rdjpgcom.obj
+	$(LINK) $(LFLAGS) /Executable = rdjpgcom.exe rdjpgcom.obj$(OPT)
+
+wrjpgcom.exe : wrjpgcom.obj
+	$(LINK) $(LFLAGS) /Executable = wrjpgcom.exe wrjpgcom.obj$(OPT)
+
+jconfig.h : jconfig.vms
+	@- Copy jconfig.vms jconfig.h
+
+clean :
+	@- Set Protection = Owner:RWED *.*;-1
+	@- Set Protection = Owner:RWED *.OBJ
+	- Purge /NoLog /NoConfirm *.*
+	- Delete /NoLog /NoConfirm *.OBJ;
+
+test : cjpeg.exe djpeg.exe jpegtran.exe
+	mcr sys$disk:[]djpeg -dct int -ppm -outfile testout.ppm testorig.jpg
+	mcr sys$disk:[]djpeg -dct int -bmp -colors 256 -outfile testout.bmp testorig.jpg
+	mcr sys$disk:[]cjpeg -dct int      -outfile testout.jpg testimg.ppm
+	mcr sys$disk:[]djpeg -dct int -ppm -outfile testoutp.ppm testprog.jpg
+	mcr sys$disk:[]cjpeg -dct int -progressive -opt -outfile testoutp.jpg testimg.ppm
+	mcr sys$disk:[]jpegtran -outfile testoutt.jpg testprog.jpg
+	- Backup /Compare/Log	  testimg.ppm testout.ppm
+	- Backup /Compare/Log	  testimg.bmp testout.bmp
+	- Backup /Compare/Log	  testimg.jpg testout.jpg
+	- Backup /Compare/Log	  testimg.ppm testoutp.ppm
+	- Backup /Compare/Log	  testimgp.jpg testoutp.jpg
+	- Backup /Compare/Log	  testorig.jpg testoutt.jpg
+
+
+jcapimin.obj : jcapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcapistd.obj : jcapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccoefct.obj : jccoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccolor.obj : jccolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcdctmgr.obj : jcdctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jchuff.obj : jchuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcinit.obj : jcinit.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmainct.obj : jcmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmarker.obj : jcmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmaster.obj : jcmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcomapi.obj : jcomapi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcparam.obj : jcparam.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcphuff.obj : jcphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcprepct.obj : jcprepct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcsample.obj : jcsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jctrans.obj : jctrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapimin.obj : jdapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapistd.obj : jdapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdatadst.obj : jdatadst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdatasrc.obj : jdatasrc.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdcoefct.obj : jdcoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdcolor.obj : jdcolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jddctmgr.obj : jddctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jdhuff.obj : jdhuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdinput.obj : jdinput.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmainct.obj : jdmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmarker.obj : jdmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmaster.obj : jdmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmerge.obj : jdmerge.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdphuff.obj : jdphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdpostct.obj : jdpostct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdsample.obj : jdsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdtrans.obj : jdtrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jerror.obj : jerror.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jversion.h jerror.h
+jfdctflt.obj : jfdctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctfst.obj : jfdctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctint.obj : jfdctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctflt.obj : jidctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctfst.obj : jidctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctint.obj : jidctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctred.obj : jidctred.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jquant1.obj : jquant1.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jquant2.obj : jquant2.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jutils.obj : jutils.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jmemmgr.obj : jmemmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemansi.obj : jmemansi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemname.obj : jmemname.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemnobs.obj : jmemnobs.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemdos.obj : jmemdos.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemmac.obj : jmemmac.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+cjpeg.obj : cjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+djpeg.obj : djpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+jpegtran.obj : jpegtran.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h transupp.h jversion.h
+rdjpgcom.obj : rdjpgcom.c jinclude.h jconfig.h
+wrjpgcom.obj : wrjpgcom.c jinclude.h jconfig.h
+cdjpeg.obj : cdjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdcolmap.obj : rdcolmap.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdswitch.obj : rdswitch.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+transupp.obj : transupp.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h transupp.h
+rdppm.obj : rdppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrppm.obj : wrppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdgif.obj : rdgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrgif.obj : wrgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdtarga.obj : rdtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrtarga.obj : wrtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdbmp.obj : rdbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrbmp.obj : wrbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdrle.obj : rdrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrrle.obj : wrrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
diff --git a/libjpegtwrp/makefile.sas b/libjpegtwrp/makefile.sas
new file mode 100644
index 000000000..f296faf06
--- /dev/null
+++ b/libjpegtwrp/makefile.sas
@@ -0,0 +1,252 @@
+# Makefile for Independent JPEG Group's software
+
+# This makefile is for Amiga systems using SAS C 6.0 and up.
+# Thanks to Ed Hanway, Mark Rinfret, and Jim Zepeda.
+
+# Read installation instructions before saying "make" !!
+
+# The name of your C compiler:
+CC= sc
+
+# You may need to adjust these cc options:
+# Uncomment the following lines for generic 680x0 version
+ARCHFLAGS= cpu=any
+SUFFIX=
+
+# Uncomment the following lines for 68030-only version
+#ARCHFLAGS= cpu=68030
+#SUFFIX=.030
+
+CFLAGS= nostackcheck data=near parms=register optimize $(ARCHFLAGS) \
+	ignore=104 ignore=304 ignore=306
+# ignore=104 disables warnings for mismatched const qualifiers
+# ignore=304 disables warnings for variables being optimized out
+# ignore=306 disables warnings for the inlining of functions
+# Generally, we recommend defining any configuration symbols in jconfig.h,
+# NOT via define switches here.
+
+# Link-time cc options:
+LDFLAGS= SC SD ND BATCH
+
+# To link any special libraries, add the necessary commands here.
+LDLIBS= LIB:scm.lib LIB:sc.lib
+
+# Put here the object file name for the correct system-dependent memory
+# manager file.  For Amiga we recommend jmemname.o.
+SYSDEPMEM= jmemname.o
+
+# miscellaneous OS-dependent stuff
+# linker
+LN= slink
+# file deletion command
+RM= delete quiet
+# library (.lib) file creation command
+AR= oml
+
+# End of configurable options.
+
+
+# source files: JPEG library proper
+LIBSOURCES= jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c \
+        jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c \
+        jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c \
+        jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c \
+        jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c \
+        jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c \
+        jfdctint.c jidctflt.c jidctfst.c jidctint.c jidctred.c jquant1.c \
+        jquant2.c jutils.c jmemmgr.c
+# memmgr back ends: compile only one of these into a working library
+SYSDEPSOURCES= jmemansi.c jmemname.c jmemnobs.c jmemdos.c jmemmac.c
+# source files: cjpeg/djpeg/jpegtran applications, also rdjpgcom/wrjpgcom
+APPSOURCES= cjpeg.c djpeg.c jpegtran.c rdjpgcom.c wrjpgcom.c cdjpeg.c \
+        rdcolmap.c rdswitch.c transupp.c rdppm.c wrppm.c rdgif.c wrgif.c \
+        rdtarga.c wrtarga.c rdbmp.c wrbmp.c rdrle.c wrrle.c
+SOURCES= $(LIBSOURCES) $(SYSDEPSOURCES) $(APPSOURCES)
+# files included by source files
+INCLUDES= jchuff.h jdhuff.h jdct.h jerror.h jinclude.h jmemsys.h jmorecfg.h \
+        jpegint.h jpeglib.h jversion.h cdjpeg.h cderror.h transupp.h
+# documentation, test, and support files
+DOCS= README install.doc usage.doc cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 \
+        wrjpgcom.1 wizard.doc example.c libjpeg.doc structure.doc \
+        coderules.doc filelist.doc change.log
+MKFILES= configure makefile.cfg makefile.ansi makefile.unix makefile.bcc \
+        makefile.mc6 makefile.dj makefile.wat makefile.vc makelib.ds \
+        makeapps.ds makeproj.mac makcjpeg.st makdjpeg.st makljpeg.st \
+        maktjpeg.st makefile.manx makefile.sas makefile.mms makefile.vms \
+        makvms.opt
+CONFIGFILES= jconfig.cfg jconfig.bcc jconfig.mc6 jconfig.dj jconfig.wat \
+        jconfig.vc jconfig.mac jconfig.st jconfig.manx jconfig.sas \
+        jconfig.vms
+CONFIGUREFILES= config.guess config.sub install-sh ltconfig ltmain.sh
+OTHERFILES= jconfig.doc ckconfig.c ansi2knr.c ansi2knr.1 jmemdosa.asm
+TESTFILES= testorig.jpg testimg.ppm testimg.bmp testimg.jpg testprog.jpg \
+        testimgp.jpg
+DISTFILES= $(DOCS) $(MKFILES) $(CONFIGFILES) $(SOURCES) $(INCLUDES) \
+        $(CONFIGUREFILES) $(OTHERFILES) $(TESTFILES)
+# library object files common to compression and decompression
+COMOBJECTS= jcomapi.o jutils.o jerror.o jmemmgr.o $(SYSDEPMEM)
+# compression library object files
+CLIBOBJECTS= jcapimin.o jcapistd.o jctrans.o jcparam.o jdatadst.o jcinit.o \
+        jcmaster.o jcmarker.o jcmainct.o jcprepct.o jccoefct.o jccolor.o \
+        jcsample.o jchuff.o jcphuff.o jcdctmgr.o jfdctfst.o jfdctflt.o \
+        jfdctint.o
+# decompression library object files
+DLIBOBJECTS= jdapimin.o jdapistd.o jdtrans.o jdatasrc.o jdmaster.o \
+        jdinput.o jdmarker.o jdhuff.o jdphuff.o jdmainct.o jdcoefct.o \
+        jdpostct.o jddctmgr.o jidctfst.o jidctflt.o jidctint.o jidctred.o \
+        jdsample.o jdcolor.o jquant1.o jquant2.o jdmerge.o
+# These objectfiles are included in libjpeg.lib
+LIBOBJECTS= $(CLIBOBJECTS) $(DLIBOBJECTS) $(COMOBJECTS)
+# object files for sample applications (excluding library files)
+COBJECTS= cjpeg.o rdppm.o rdgif.o rdtarga.o rdrle.o rdbmp.o rdswitch.o \
+        cdjpeg.o
+DOBJECTS= djpeg.o wrppm.o wrgif.o wrtarga.o wrrle.o wrbmp.o rdcolmap.o \
+        cdjpeg.o
+TROBJECTS= jpegtran.o rdswitch.o cdjpeg.o transupp.o
+
+
+all: libjpeg.lib cjpeg$(SUFFIX) djpeg$(SUFFIX) jpegtran$(SUFFIX) rdjpgcom$(SUFFIX) wrjpgcom$(SUFFIX)
+
+# note: do several AR steps to avoid command line length limitations
+
+libjpeg.lib: $(LIBOBJECTS)
+	-$(RM) libjpeg.lib
+	$(AR) libjpeg.lib r $(CLIBOBJECTS)
+	$(AR) libjpeg.lib r $(DLIBOBJECTS)
+	$(AR) libjpeg.lib r $(COMOBJECTS)
+
+cjpeg$(SUFFIX): $(COBJECTS) libjpeg.lib
+	$(LN) <WITH <
+$(LDFLAGS)
+TO cjpeg$(SUFFIX)
+FROM LIB:c.o $(COBJECTS)
+LIB libjpeg.lib $(LDLIBS)
+<
+
+djpeg$(SUFFIX): $(DOBJECTS) libjpeg.lib
+	$(LN) <WITH <
+$(LDFLAGS)
+TO djpeg$(SUFFIX)
+FROM LIB:c.o $(DOBJECTS)
+LIB libjpeg.lib $(LDLIBS)
+<
+
+jpegtran$(SUFFIX): $(TROBJECTS) libjpeg.lib
+	$(LN) <WITH <
+$(LDFLAGS)
+TO jpegtran$(SUFFIX)
+FROM LIB:c.o $(TROBJECTS)
+LIB libjpeg.lib $(LDLIBS)
+<
+
+rdjpgcom$(SUFFIX): rdjpgcom.o
+	$(LN) <WITH <
+$(LDFLAGS)
+TO rdjpgcom$(SUFFIX)
+FROM LIB:c.o rdjpgcom.o
+LIB $(LDLIBS)
+<
+
+wrjpgcom$(SUFFIX): wrjpgcom.o
+	$(LN) <WITH <
+$(LDFLAGS)
+TO wrjpgcom$(SUFFIX)
+FROM LIB:c.o wrjpgcom.o
+LIB $(LDLIBS)
+<
+
+jconfig.h: jconfig.doc
+	echo You must prepare a system-dependent jconfig.h file.
+	echo Please read the installation directions in install.doc.
+	exit 1
+
+clean:
+	-$(RM) *.o cjpeg djpeg jpegtran cjpeg.030 djpeg.030 jpegtran.030
+	-$(RM) rdjpgcom wrjpgcom rdjpgcom.030 wrjpgcom.030
+	-$(RM) libjpeg.lib core testout*.*
+
+test: cjpeg djpeg jpegtran
+	-$(RM) testout*.*
+	djpeg -dct int -ppm -outfile testout.ppm  testorig.jpg
+	djpeg -dct int -bmp -colors 256 -outfile testout.bmp  testorig.jpg
+	cjpeg -dct int -outfile testout.jpg  testimg.ppm
+	djpeg -dct int -ppm -outfile testoutp.ppm testprog.jpg
+	cjpeg -dct int -progressive -opt -outfile testoutp.jpg testimg.ppm
+	jpegtran -outfile testoutt.jpg testprog.jpg
+	cmp testimg.ppm testout.ppm
+	cmp testimg.bmp testout.bmp
+	cmp testimg.jpg testout.jpg
+	cmp testimg.ppm testoutp.ppm
+	cmp testimgp.jpg testoutp.jpg
+	cmp testorig.jpg testoutt.jpg
+
+
+jcapimin.o: jcapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcapistd.o: jcapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccoefct.o: jccoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccolor.o: jccolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcdctmgr.o: jcdctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jchuff.o: jchuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcinit.o: jcinit.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmainct.o: jcmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmarker.o: jcmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmaster.o: jcmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcomapi.o: jcomapi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcparam.o: jcparam.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcphuff.o: jcphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcprepct.o: jcprepct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcsample.o: jcsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jctrans.o: jctrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapimin.o: jdapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapistd.o: jdapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdatadst.o: jdatadst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdatasrc.o: jdatasrc.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdcoefct.o: jdcoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdcolor.o: jdcolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jddctmgr.o: jddctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jdhuff.o: jdhuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdinput.o: jdinput.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmainct.o: jdmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmarker.o: jdmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmaster.o: jdmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmerge.o: jdmerge.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdphuff.o: jdphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdpostct.o: jdpostct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdsample.o: jdsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdtrans.o: jdtrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jerror.o: jerror.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jversion.h jerror.h
+jfdctflt.o: jfdctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctfst.o: jfdctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctint.o: jfdctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctflt.o: jidctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctfst.o: jidctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctint.o: jidctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctred.o: jidctred.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jquant1.o: jquant1.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jquant2.o: jquant2.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jutils.o: jutils.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jmemmgr.o: jmemmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemansi.o: jmemansi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemname.o: jmemname.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemnobs.o: jmemnobs.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemdos.o: jmemdos.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemmac.o: jmemmac.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+cjpeg.o: cjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+djpeg.o: djpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+jpegtran.o: jpegtran.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h transupp.h jversion.h
+rdjpgcom.o: rdjpgcom.c jinclude.h jconfig.h
+wrjpgcom.o: wrjpgcom.c jinclude.h jconfig.h
+cdjpeg.o: cdjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdcolmap.o: rdcolmap.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdswitch.o: rdswitch.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+transupp.o: transupp.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h transupp.h
+rdppm.o: rdppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrppm.o: wrppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdgif.o: rdgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrgif.o: wrgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdtarga.o: rdtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrtarga.o: wrtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdbmp.o: rdbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrbmp.o: wrbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdrle.o: rdrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrrle.o: wrrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
diff --git a/libjpegtwrp/makefile.unix b/libjpegtwrp/makefile.unix
new file mode 100644
index 000000000..00455ab6a
--- /dev/null
+++ b/libjpegtwrp/makefile.unix
@@ -0,0 +1,228 @@
+# Makefile for Independent JPEG Group's software
+
+# This makefile is suitable for Unix-like systems with non-ANSI compilers.
+# If you have an ANSI compiler, makefile.ansi is a better starting point.
+
+# Read installation instructions before saying "make" !!
+
+# The name of your C compiler:
+CC= cc
+
+# You may need to adjust these cc options:
+CFLAGS= -O
+# Generally, we recommend defining any configuration symbols in jconfig.h,
+# NOT via -D switches here.
+# However, any special defines for ansi2knr.c may be included here:
+ANSI2KNRFLAGS= 
+
+# Link-time cc options:
+LDFLAGS= 
+
+# To link any special libraries, add the necessary -l commands here.
+LDLIBS= 
+
+# Put here the object file name for the correct system-dependent memory
+# manager file.  For Unix this is usually jmemnobs.o, but you may want
+# to use jmemansi.o or jmemname.o if you have limited swap space.
+SYSDEPMEM= jmemnobs.o
+
+# miscellaneous OS-dependent stuff
+# linker
+LN= $(CC)
+# file deletion command
+RM= rm -f
+# file rename command
+MV= mv
+# library (.a) file creation command
+AR= ar rc
+# second step in .a creation (use "touch" if not needed)
+AR2= ranlib
+
+# End of configurable options.
+
+
+# source files: JPEG library proper
+LIBSOURCES= jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c \
+        jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c \
+        jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c \
+        jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c \
+        jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c \
+        jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c \
+        jfdctint.c jidctflt.c jidctfst.c jidctint.c jidctred.c jquant1.c \
+        jquant2.c jutils.c jmemmgr.c
+# memmgr back ends: compile only one of these into a working library
+SYSDEPSOURCES= jmemansi.c jmemname.c jmemnobs.c jmemdos.c jmemmac.c
+# source files: cjpeg/djpeg/jpegtran applications, also rdjpgcom/wrjpgcom
+APPSOURCES= cjpeg.c djpeg.c jpegtran.c rdjpgcom.c wrjpgcom.c cdjpeg.c \
+        rdcolmap.c rdswitch.c transupp.c rdppm.c wrppm.c rdgif.c wrgif.c \
+        rdtarga.c wrtarga.c rdbmp.c wrbmp.c rdrle.c wrrle.c
+SOURCES= $(LIBSOURCES) $(SYSDEPSOURCES) $(APPSOURCES)
+# files included by source files
+INCLUDES= jchuff.h jdhuff.h jdct.h jerror.h jinclude.h jmemsys.h jmorecfg.h \
+        jpegint.h jpeglib.h jversion.h cdjpeg.h cderror.h transupp.h
+# documentation, test, and support files
+DOCS= README install.doc usage.doc cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 \
+        wrjpgcom.1 wizard.doc example.c libjpeg.doc structure.doc \
+        coderules.doc filelist.doc change.log
+MKFILES= configure makefile.cfg makefile.ansi makefile.unix makefile.bcc \
+        makefile.mc6 makefile.dj makefile.wat makefile.vc makelib.ds \
+        makeapps.ds makeproj.mac makcjpeg.st makdjpeg.st makljpeg.st \
+        maktjpeg.st makefile.manx makefile.sas makefile.mms makefile.vms \
+        makvms.opt
+CONFIGFILES= jconfig.cfg jconfig.bcc jconfig.mc6 jconfig.dj jconfig.wat \
+        jconfig.vc jconfig.mac jconfig.st jconfig.manx jconfig.sas \
+        jconfig.vms
+CONFIGUREFILES= config.guess config.sub install-sh ltconfig ltmain.sh
+OTHERFILES= jconfig.doc ckconfig.c ansi2knr.c ansi2knr.1 jmemdosa.asm
+TESTFILES= testorig.jpg testimg.ppm testimg.bmp testimg.jpg testprog.jpg \
+        testimgp.jpg
+DISTFILES= $(DOCS) $(MKFILES) $(CONFIGFILES) $(SOURCES) $(INCLUDES) \
+        $(CONFIGUREFILES) $(OTHERFILES) $(TESTFILES)
+# library object files common to compression and decompression
+COMOBJECTS= jcomapi.o jutils.o jerror.o jmemmgr.o $(SYSDEPMEM)
+# compression library object files
+CLIBOBJECTS= jcapimin.o jcapistd.o jctrans.o jcparam.o jdatadst.o jcinit.o \
+        jcmaster.o jcmarker.o jcmainct.o jcprepct.o jccoefct.o jccolor.o \
+        jcsample.o jchuff.o jcphuff.o jcdctmgr.o jfdctfst.o jfdctflt.o \
+        jfdctint.o
+# decompression library object files
+DLIBOBJECTS= jdapimin.o jdapistd.o jdtrans.o jdatasrc.o jdmaster.o \
+        jdinput.o jdmarker.o jdhuff.o jdphuff.o jdmainct.o jdcoefct.o \
+        jdpostct.o jddctmgr.o jidctfst.o jidctflt.o jidctint.o jidctred.o \
+        jdsample.o jdcolor.o jquant1.o jquant2.o jdmerge.o
+# These objectfiles are included in libjpeg.a
+LIBOBJECTS= $(CLIBOBJECTS) $(DLIBOBJECTS) $(COMOBJECTS)
+# object files for sample applications (excluding library files)
+COBJECTS= cjpeg.o rdppm.o rdgif.o rdtarga.o rdrle.o rdbmp.o rdswitch.o \
+        cdjpeg.o
+DOBJECTS= djpeg.o wrppm.o wrgif.o wrtarga.o wrrle.o wrbmp.o rdcolmap.o \
+        cdjpeg.o
+TROBJECTS= jpegtran.o rdswitch.o cdjpeg.o transupp.o
+
+
+all: ansi2knr libjpeg.a cjpeg djpeg jpegtran rdjpgcom wrjpgcom
+
+# This rule causes ansi2knr to be invoked.
+.c.o:
+	./ansi2knr $*.c T$*.c
+	$(CC) $(CFLAGS) -c T$*.c
+	$(RM) T$*.c $*.o
+	$(MV) T$*.o $*.o
+
+ansi2knr: ansi2knr.c
+	$(CC) $(CFLAGS) $(ANSI2KNRFLAGS) -o ansi2knr ansi2knr.c
+
+libjpeg.a: ansi2knr $(LIBOBJECTS)
+	$(RM) libjpeg.a
+	$(AR) libjpeg.a  $(LIBOBJECTS)
+	$(AR2) libjpeg.a
+
+cjpeg: ansi2knr $(COBJECTS) libjpeg.a
+	$(LN) $(LDFLAGS) -o cjpeg $(COBJECTS) libjpeg.a $(LDLIBS)
+
+djpeg: ansi2knr $(DOBJECTS) libjpeg.a
+	$(LN) $(LDFLAGS) -o djpeg $(DOBJECTS) libjpeg.a $(LDLIBS)
+
+jpegtran: ansi2knr $(TROBJECTS) libjpeg.a
+	$(LN) $(LDFLAGS) -o jpegtran $(TROBJECTS) libjpeg.a $(LDLIBS)
+
+rdjpgcom: rdjpgcom.o
+	$(LN) $(LDFLAGS) -o rdjpgcom rdjpgcom.o $(LDLIBS)
+
+wrjpgcom: wrjpgcom.o
+	$(LN) $(LDFLAGS) -o wrjpgcom wrjpgcom.o $(LDLIBS)
+
+jconfig.h: jconfig.doc
+	echo You must prepare a system-dependent jconfig.h file.
+	echo Please read the installation directions in install.doc.
+	exit 1
+
+clean:
+	$(RM) *.o cjpeg djpeg jpegtran libjpeg.a rdjpgcom wrjpgcom
+	$(RM) ansi2knr core testout*
+
+test: cjpeg djpeg jpegtran
+	$(RM) testout*
+	./djpeg -dct int -ppm -outfile testout.ppm  testorig.jpg
+	./djpeg -dct int -bmp -colors 256 -outfile testout.bmp  testorig.jpg
+	./cjpeg -dct int -outfile testout.jpg  testimg.ppm
+	./djpeg -dct int -ppm -outfile testoutp.ppm testprog.jpg
+	./cjpeg -dct int -progressive -opt -outfile testoutp.jpg testimg.ppm
+	./jpegtran -outfile testoutt.jpg testprog.jpg
+	cmp testimg.ppm testout.ppm
+	cmp testimg.bmp testout.bmp
+	cmp testimg.jpg testout.jpg
+	cmp testimg.ppm testoutp.ppm
+	cmp testimgp.jpg testoutp.jpg
+	cmp testorig.jpg testoutt.jpg
+
+
+jcapimin.o: jcapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcapistd.o: jcapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccoefct.o: jccoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccolor.o: jccolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcdctmgr.o: jcdctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jchuff.o: jchuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcinit.o: jcinit.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmainct.o: jcmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmarker.o: jcmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmaster.o: jcmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcomapi.o: jcomapi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcparam.o: jcparam.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcphuff.o: jcphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcprepct.o: jcprepct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcsample.o: jcsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jctrans.o: jctrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapimin.o: jdapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapistd.o: jdapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdatadst.o: jdatadst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdatasrc.o: jdatasrc.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdcoefct.o: jdcoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdcolor.o: jdcolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jddctmgr.o: jddctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jdhuff.o: jdhuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdinput.o: jdinput.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmainct.o: jdmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmarker.o: jdmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmaster.o: jdmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmerge.o: jdmerge.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdphuff.o: jdphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdpostct.o: jdpostct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdsample.o: jdsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdtrans.o: jdtrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jerror.o: jerror.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jversion.h jerror.h
+jfdctflt.o: jfdctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctfst.o: jfdctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctint.o: jfdctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctflt.o: jidctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctfst.o: jidctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctint.o: jidctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctred.o: jidctred.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jquant1.o: jquant1.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jquant2.o: jquant2.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jutils.o: jutils.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jmemmgr.o: jmemmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemansi.o: jmemansi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemname.o: jmemname.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemnobs.o: jmemnobs.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemdos.o: jmemdos.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemmac.o: jmemmac.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+cjpeg.o: cjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+djpeg.o: djpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+jpegtran.o: jpegtran.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h transupp.h jversion.h
+rdjpgcom.o: rdjpgcom.c jinclude.h jconfig.h
+wrjpgcom.o: wrjpgcom.c jinclude.h jconfig.h
+cdjpeg.o: cdjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdcolmap.o: rdcolmap.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdswitch.o: rdswitch.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+transupp.o: transupp.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h transupp.h
+rdppm.o: rdppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrppm.o: wrppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdgif.o: rdgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrgif.o: wrgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdtarga.o: rdtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrtarga.o: wrtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdbmp.o: rdbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrbmp.o: wrbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdrle.o: rdrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrrle.o: wrrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
diff --git a/libjpegtwrp/makefile.vc b/libjpegtwrp/makefile.vc
new file mode 100644
index 000000000..2acf06940
--- /dev/null
+++ b/libjpegtwrp/makefile.vc
@@ -0,0 +1,211 @@
+# Makefile for Independent JPEG Group's software
+
+# This makefile is for Microsoft Visual C++ on Windows NT (and 95?).
+# It builds the IJG library as a statically linkable library (.LIB),
+# and builds the sample applications as console-mode apps.
+# Thanks to Xingong Chang, Raymond Everly and others.
+
+# Read installation instructions before saying "nmake" !!
+# To build an optimized library without debug info, say "nmake nodebug=1".
+
+# Pull in standard variable definitions
+!include <win32.mak>
+
+# You may want to adjust these compiler options:
+CFLAGS= $(cflags) $(cdebug) $(cvars) -I.
+# Generally, we recommend defining any configuration symbols in jconfig.h,
+# NOT via -D switches here.
+
+# Link-time options:
+LDFLAGS= $(ldebug) $(conlflags)
+
+# To link any special libraries, add the necessary commands here.
+LDLIBS= $(conlibs)
+
+# Put here the object file name for the correct system-dependent memory
+# manager file.  For NT we suggest jmemnobs.obj, which expects the OS to
+# provide adequate virtual memory.
+SYSDEPMEM= jmemnobs.obj
+
+# miscellaneous OS-dependent stuff
+# file deletion command
+RM= del
+
+# End of configurable options.
+
+
+# source files: JPEG library proper
+LIBSOURCES= jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c \
+        jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c \
+        jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c \
+        jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c \
+        jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c \
+        jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c \
+        jfdctint.c jidctflt.c jidctfst.c jidctint.c jidctred.c jquant1.c \
+        jquant2.c jutils.c jmemmgr.c
+# memmgr back ends: compile only one of these into a working library
+SYSDEPSOURCES= jmemansi.c jmemname.c jmemnobs.c jmemdos.c jmemmac.c
+# source files: cjpeg/djpeg/jpegtran applications, also rdjpgcom/wrjpgcom
+APPSOURCES= cjpeg.c djpeg.c jpegtran.c rdjpgcom.c wrjpgcom.c cdjpeg.c \
+        rdcolmap.c rdswitch.c transupp.c rdppm.c wrppm.c rdgif.c wrgif.c \
+        rdtarga.c wrtarga.c rdbmp.c wrbmp.c rdrle.c wrrle.c
+SOURCES= $(LIBSOURCES) $(SYSDEPSOURCES) $(APPSOURCES)
+# files included by source files
+INCLUDES= jchuff.h jdhuff.h jdct.h jerror.h jinclude.h jmemsys.h jmorecfg.h \
+        jpegint.h jpeglib.h jversion.h cdjpeg.h cderror.h transupp.h
+# documentation, test, and support files
+DOCS= README install.doc usage.doc cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 \
+        wrjpgcom.1 wizard.doc example.c libjpeg.doc structure.doc \
+        coderules.doc filelist.doc change.log
+MKFILES= configure makefile.cfg makefile.ansi makefile.unix makefile.bcc \
+        makefile.mc6 makefile.dj makefile.wat makefile.vc makelib.ds \
+        makeapps.ds makeproj.mac makcjpeg.st makdjpeg.st makljpeg.st \
+        maktjpeg.st makefile.manx makefile.sas makefile.mms makefile.vms \
+        makvms.opt
+CONFIGFILES= jconfig.cfg jconfig.bcc jconfig.mc6 jconfig.dj jconfig.wat \
+        jconfig.vc jconfig.mac jconfig.st jconfig.manx jconfig.sas \
+        jconfig.vms
+CONFIGUREFILES= config.guess config.sub install-sh ltconfig ltmain.sh
+OTHERFILES= jconfig.doc ckconfig.c ansi2knr.c ansi2knr.1 jmemdosa.asm
+TESTFILES= testorig.jpg testimg.ppm testimg.bmp testimg.jpg testprog.jpg \
+        testimgp.jpg
+DISTFILES= $(DOCS) $(MKFILES) $(CONFIGFILES) $(SOURCES) $(INCLUDES) \
+        $(CONFIGUREFILES) $(OTHERFILES) $(TESTFILES)
+# library object files common to compression and decompression
+COMOBJECTS= jcomapi.obj jutils.obj jerror.obj jmemmgr.obj $(SYSDEPMEM)
+# compression library object files
+CLIBOBJECTS= jcapimin.obj jcapistd.obj jctrans.obj jcparam.obj jdatadst.obj \
+        jcinit.obj jcmaster.obj jcmarker.obj jcmainct.obj jcprepct.obj \
+        jccoefct.obj jccolor.obj jcsample.obj jchuff.obj jcphuff.obj \
+        jcdctmgr.obj jfdctfst.obj jfdctflt.obj jfdctint.obj
+# decompression library object files
+DLIBOBJECTS= jdapimin.obj jdapistd.obj jdtrans.obj jdatasrc.obj \
+        jdmaster.obj jdinput.obj jdmarker.obj jdhuff.obj jdphuff.obj \
+        jdmainct.obj jdcoefct.obj jdpostct.obj jddctmgr.obj jidctfst.obj \
+        jidctflt.obj jidctint.obj jidctred.obj jdsample.obj jdcolor.obj \
+        jquant1.obj jquant2.obj jdmerge.obj
+# These objectfiles are included in libjpeg.lib
+LIBOBJECTS= $(CLIBOBJECTS) $(DLIBOBJECTS) $(COMOBJECTS)
+# object files for sample applications (excluding library files)
+COBJECTS= cjpeg.obj rdppm.obj rdgif.obj rdtarga.obj rdrle.obj rdbmp.obj \
+        rdswitch.obj cdjpeg.obj
+DOBJECTS= djpeg.obj wrppm.obj wrgif.obj wrtarga.obj wrrle.obj wrbmp.obj \
+        rdcolmap.obj cdjpeg.obj
+TROBJECTS= jpegtran.obj rdswitch.obj cdjpeg.obj transupp.obj
+
+# Template command for compiling .c to .obj
+.c.obj:
+	$(cc) $(CFLAGS) $*.c
+
+
+all: libjpeg.lib cjpeg.exe djpeg.exe jpegtran.exe rdjpgcom.exe wrjpgcom.exe
+
+libjpeg.lib: $(LIBOBJECTS)
+	$(RM) libjpeg.lib
+	lib -out:libjpeg.lib  $(LIBOBJECTS)
+
+cjpeg.exe: $(COBJECTS) libjpeg.lib
+	$(link) $(LDFLAGS) -out:cjpeg.exe $(COBJECTS) libjpeg.lib $(LDLIBS)
+
+djpeg.exe: $(DOBJECTS) libjpeg.lib
+	$(link) $(LDFLAGS) -out:djpeg.exe $(DOBJECTS) libjpeg.lib $(LDLIBS)
+
+jpegtran.exe: $(TROBJECTS) libjpeg.lib
+	$(link) $(LDFLAGS) -out:jpegtran.exe $(TROBJECTS) libjpeg.lib $(LDLIBS)
+
+rdjpgcom.exe: rdjpgcom.obj
+	$(link) $(LDFLAGS) -out:rdjpgcom.exe rdjpgcom.obj $(LDLIBS)
+
+wrjpgcom.exe: wrjpgcom.obj
+	$(link) $(LDFLAGS) -out:wrjpgcom.exe wrjpgcom.obj $(LDLIBS)
+
+
+clean:
+	$(RM) *.obj *.exe libjpeg.lib
+	$(RM) testout*
+
+test: cjpeg.exe djpeg.exe jpegtran.exe
+	$(RM) testout*
+	.\djpeg -dct int -ppm -outfile testout.ppm  testorig.jpg
+	.\djpeg -dct int -bmp -colors 256 -outfile testout.bmp  testorig.jpg
+	.\cjpeg -dct int -outfile testout.jpg  testimg.ppm
+	.\djpeg -dct int -ppm -outfile testoutp.ppm testprog.jpg
+	.\cjpeg -dct int -progressive -opt -outfile testoutp.jpg testimg.ppm
+	.\jpegtran -outfile testoutt.jpg testprog.jpg
+	fc /b testimg.ppm testout.ppm
+	fc /b testimg.bmp testout.bmp
+	fc /b testimg.jpg testout.jpg
+	fc /b testimg.ppm testoutp.ppm
+	fc /b testimgp.jpg testoutp.jpg
+	fc /b testorig.jpg testoutt.jpg
+
+
+jcapimin.obj: jcapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcapistd.obj: jcapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccoefct.obj: jccoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccolor.obj: jccolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcdctmgr.obj: jcdctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jchuff.obj: jchuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcinit.obj: jcinit.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmainct.obj: jcmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmarker.obj: jcmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmaster.obj: jcmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcomapi.obj: jcomapi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcparam.obj: jcparam.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcphuff.obj: jcphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcprepct.obj: jcprepct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcsample.obj: jcsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jctrans.obj: jctrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapimin.obj: jdapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapistd.obj: jdapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdatadst.obj: jdatadst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdatasrc.obj: jdatasrc.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdcoefct.obj: jdcoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdcolor.obj: jdcolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jddctmgr.obj: jddctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jdhuff.obj: jdhuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdinput.obj: jdinput.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmainct.obj: jdmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmarker.obj: jdmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmaster.obj: jdmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmerge.obj: jdmerge.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdphuff.obj: jdphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdpostct.obj: jdpostct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdsample.obj: jdsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdtrans.obj: jdtrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jerror.obj: jerror.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jversion.h jerror.h
+jfdctflt.obj: jfdctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctfst.obj: jfdctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctint.obj: jfdctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctflt.obj: jidctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctfst.obj: jidctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctint.obj: jidctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctred.obj: jidctred.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jquant1.obj: jquant1.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jquant2.obj: jquant2.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jutils.obj: jutils.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jmemmgr.obj: jmemmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemansi.obj: jmemansi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemname.obj: jmemname.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemnobs.obj: jmemnobs.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemdos.obj: jmemdos.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemmac.obj: jmemmac.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+cjpeg.obj: cjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+djpeg.obj: djpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+jpegtran.obj: jpegtran.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h transupp.h jversion.h
+rdjpgcom.obj: rdjpgcom.c jinclude.h jconfig.h
+wrjpgcom.obj: wrjpgcom.c jinclude.h jconfig.h
+cdjpeg.obj: cdjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdcolmap.obj: rdcolmap.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdswitch.obj: rdswitch.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+transupp.obj: transupp.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h transupp.h
+rdppm.obj: rdppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrppm.obj: wrppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdgif.obj: rdgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrgif.obj: wrgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdtarga.obj: rdtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrtarga.obj: wrtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdbmp.obj: rdbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrbmp.obj: wrbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdrle.obj: rdrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrrle.obj: wrrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
diff --git a/libjpegtwrp/makefile.vms b/libjpegtwrp/makefile.vms
new file mode 100644
index 000000000..a42358d05
--- /dev/null
+++ b/libjpegtwrp/makefile.vms
@@ -0,0 +1,142 @@
+$! Makefile for Independent JPEG Group's software
+$!
+$! This is a command procedure for Digital VMS systems that do not have MMS.
+$! It builds the JPEG software by brute force, recompiling everything whether
+$! or not it is necessary.  It then runs the basic self-test.
+$! Thanks to Rick Dyson (dyson@iowasp.physics.uiowa.edu)
+$! and Tim Bell (tbell@netcom.com) for their help.
+$!
+$! Read installation instructions before running this!!
+$!
+$ If F$Mode () .eqs. "INTERACTIVE"
+$   Then
+$       VERIFY = F$Verify (0)
+$   Else
+$       VERIFY = F$Verify (1)
+$ EndIf
+$ On Control_Y Then GoTo End
+$ On Error     Then GoTo End
+$
+$ If F$GetSyi ("HW_MODEL") .gt. 1023 
+$   Then
+$       OPT = ""
+$   Else
+$       OPT = ",Sys$Disk:[]makvms.opt/Option"
+$ EndIf
+$ 
+$ DoCompile := CC /NoDebug /Optimize /NoList
+$!
+$ DoCompile jcapimin.c
+$ DoCompile jcapistd.c
+$ DoCompile jctrans.c
+$ DoCompile jcparam.c
+$ DoCompile jdatadst.c
+$ DoCompile jcinit.c
+$ DoCompile jcmaster.c
+$ DoCompile jcmarker.c
+$ DoCompile jcmainct.c
+$ DoCompile jcprepct.c
+$ DoCompile jccoefct.c
+$ DoCompile jccolor.c
+$ DoCompile jcsample.c
+$ DoCompile jchuff.c
+$ DoCompile jcphuff.c
+$ DoCompile jcdctmgr.c
+$ DoCompile jfdctfst.c
+$ DoCompile jfdctflt.c
+$ DoCompile jfdctint.c
+$ DoCompile jdapimin.c
+$ DoCompile jdapistd.c
+$ DoCompile jdtrans.c
+$ DoCompile jdatasrc.c
+$ DoCompile jdmaster.c
+$ DoCompile jdinput.c
+$ DoCompile jdmarker.c
+$ DoCompile jdhuff.c
+$ DoCompile jdphuff.c
+$ DoCompile jdmainct.c
+$ DoCompile jdcoefct.c
+$ DoCompile jdpostct.c
+$ DoCompile jddctmgr.c
+$ DoCompile jidctfst.c
+$ DoCompile jidctflt.c
+$ DoCompile jidctint.c
+$ DoCompile jidctred.c
+$ DoCompile jdsample.c
+$ DoCompile jdcolor.c
+$ DoCompile jquant1.c
+$ DoCompile jquant2.c
+$ DoCompile jdmerge.c
+$ DoCompile jcomapi.c
+$ DoCompile jutils.c
+$ DoCompile jerror.c
+$ DoCompile jmemmgr.c
+$ DoCompile jmemnobs.c
+$!
+$ Library /Create libjpeg.olb  jcapimin.obj,jcapistd.obj,jctrans.obj, -
+          jcparam.obj,jdatadst.obj,jcinit.obj,jcmaster.obj,jcmarker.obj, -
+          jcmainct.obj,jcprepct.obj,jccoefct.obj,jccolor.obj,jcsample.obj, -
+          jchuff.obj,jcphuff.obj,jcdctmgr.obj,jfdctfst.obj,jfdctflt.obj, -
+          jfdctint.obj,jdapimin.obj,jdapistd.obj,jdtrans.obj,jdatasrc.obj, -
+          jdmaster.obj,jdinput.obj,jdmarker.obj,jdhuff.obj,jdphuff.obj, -
+          jdmainct.obj,jdcoefct.obj,jdpostct.obj,jddctmgr.obj,jidctfst.obj, -
+          jidctflt.obj,jidctint.obj,jidctred.obj,jdsample.obj,jdcolor.obj, -
+          jquant1.obj,jquant2.obj,jdmerge.obj,jcomapi.obj,jutils.obj, -
+          jerror.obj,jmemmgr.obj,jmemnobs.obj
+$!
+$ DoCompile cjpeg.c
+$ DoCompile rdppm.c
+$ DoCompile rdgif.c
+$ DoCompile rdtarga.c
+$ DoCompile rdrle.c
+$ DoCompile rdbmp.c
+$ DoCompile rdswitch.c
+$ DoCompile cdjpeg.c
+$!
+$ Link /NoMap /Executable = cjpeg.exe  cjpeg.obj,rdppm.obj,rdgif.obj, -
+          rdtarga.obj,rdrle.obj,rdbmp.obj,rdswitch.obj,cdjpeg.obj,libjpeg.olb/Library'OPT'
+$!
+$ DoCompile djpeg.c
+$ DoCompile wrppm.c
+$ DoCompile wrgif.c
+$ DoCompile wrtarga.c
+$ DoCompile wrrle.c
+$ DoCompile wrbmp.c
+$ DoCompile rdcolmap.c
+$ DoCompile cdjpeg.c
+$!
+$ Link /NoMap /Executable = djpeg.exe  djpeg.obj,wrppm.obj,wrgif.obj, -
+          wrtarga.obj,wrrle.obj,wrbmp.obj,rdcolmap.obj,cdjpeg.obj,libjpeg.olb/Library'OPT'
+$!
+$ DoCompile jpegtran.c
+$ DoCompile rdswitch.c
+$ DoCompile cdjpeg.c
+$ DoCompile transupp.c
+$!
+$ Link /NoMap /Executable = jpegtran.exe  jpegtran.obj,rdswitch.obj, -
+          cdjpeg.obj,transupp.obj,libjpeg.olb/Library'OPT'
+$!
+$ DoCompile rdjpgcom.c
+$ Link /NoMap /Executable = rdjpgcom.exe  rdjpgcom.obj'OPT'
+$!
+$ DoCompile wrjpgcom.c
+$ Link /NoMap /Executable = wrjpgcom.exe  wrjpgcom.obj'OPT'
+$!
+$! Run the self-test
+$!
+$ mcr sys$disk:[]djpeg -dct int -ppm -outfile testout.ppm testorig.jpg
+$ mcr sys$disk:[]djpeg -dct int -bmp -colors 256 -outfile testout.bmp testorig.jpg
+$ mcr sys$disk:[]cjpeg -dct int      -outfile testout.jpg testimg.ppm
+$ mcr sys$disk:[]djpeg -dct int -ppm -outfile testoutp.ppm testprog.jpg
+$ mcr sys$disk:[]cjpeg -dct int -progressive -opt -outfile testoutp.jpg testimg.ppm
+$ mcr sys$disk:[]jpegtran -outfile testoutt.jpg testprog.jpg
+$ Backup /Compare/Log testimg.ppm testout.ppm
+$ Backup /Compare/Log testimg.bmp testout.bmp
+$ Backup /Compare/Log testimg.jpg testout.jpg
+$ Backup /Compare/Log testimg.ppm testoutp.ppm
+$ Backup /Compare/Log testimgp.jpg testoutp.jpg
+$ Backup /Compare/Log testorig.jpg testoutt.jpg
+$!
+$End:
+$   If Verify Then Set Verify
+$ Exit
diff --git a/libjpegtwrp/makefile.wat b/libjpegtwrp/makefile.wat
new file mode 100644
index 000000000..d953e466f
--- /dev/null
+++ b/libjpegtwrp/makefile.wat
@@ -0,0 +1,233 @@
+# Makefile for Independent JPEG Group's software
+
+# This makefile is suitable for Watcom C/C++ 10.0 on MS-DOS (using
+# dos4g extender), OS/2, and Windows NT console mode.
+# Thanks to Janos Haide, jhaide@btrvtech.com.
+
+# Read installation instructions before saying "wmake" !!
+
+# Uncomment line for desired system
+SYSTEM=DOS
+#SYSTEM=OS2
+#SYSTEM=NT
+
+# The name of your C compiler:
+CC= wcl386
+
+# You may need to adjust these cc options:
+CFLAGS= -4r -ort -wx -zq -bt=$(SYSTEM)
+# Caution: avoid -ol or -ox; these generate bad code with 10.0 or 10.0a.
+# Generally, we recommend defining any configuration symbols in jconfig.h,
+# NOT via -D switches here.
+
+# Link-time cc options:
+!ifeq SYSTEM DOS
+LDFLAGS= -zq -l=dos4g
+!else ifeq SYSTEM OS2
+LDFLAGS= -zq -l=os2v2
+!else ifeq SYSTEM NT
+LDFLAGS= -zq -l=nt
+!endif
+
+# Put here the object file name for the correct system-dependent memory
+# manager file.  jmemnobs should work fine for dos4g or OS/2 environment.
+SYSDEPMEM= jmemnobs.obj
+
+# End of configurable options.
+
+
+# source files: JPEG library proper
+LIBSOURCES= jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c &
+        jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c &
+        jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c &
+        jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c &
+        jdinput.c jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c &
+        jdpostct.c jdsample.c jdtrans.c jerror.c jfdctflt.c jfdctfst.c &
+        jfdctint.c jidctflt.c jidctfst.c jidctint.c jidctred.c jquant1.c &
+        jquant2.c jutils.c jmemmgr.c
+# memmgr back ends: compile only one of these into a working library
+SYSDEPSOURCES= jmemansi.c jmemname.c jmemnobs.c jmemdos.c jmemmac.c
+# source files: cjpeg/djpeg/jpegtran applications, also rdjpgcom/wrjpgcom
+APPSOURCES= cjpeg.c djpeg.c jpegtran.c rdjpgcom.c wrjpgcom.c cdjpeg.c &
+        rdcolmap.c rdswitch.c transupp.c rdppm.c wrppm.c rdgif.c wrgif.c &
+        rdtarga.c wrtarga.c rdbmp.c wrbmp.c rdrle.c wrrle.c
+SOURCES= $(LIBSOURCES) $(SYSDEPSOURCES) $(APPSOURCES)
+# files included by source files
+INCLUDES= jchuff.h jdhuff.h jdct.h jerror.h jinclude.h jmemsys.h jmorecfg.h &
+        jpegint.h jpeglib.h jversion.h cdjpeg.h cderror.h transupp.h
+# documentation, test, and support files
+DOCS= README install.doc usage.doc cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 &
+        wrjpgcom.1 wizard.doc example.c libjpeg.doc structure.doc &
+        coderules.doc filelist.doc change.log
+MKFILES= configure makefile.cfg makefile.ansi makefile.unix makefile.bcc &
+        makefile.mc6 makefile.dj makefile.wat makefile.vc makelib.ds &
+        makeapps.ds makeproj.mac makcjpeg.st makdjpeg.st makljpeg.st &
+        maktjpeg.st makefile.manx makefile.sas makefile.mms makefile.vms &
+        makvms.opt
+CONFIGFILES= jconfig.cfg jconfig.bcc jconfig.mc6 jconfig.dj jconfig.wat &
+        jconfig.vc jconfig.mac jconfig.st jconfig.manx jconfig.sas &
+        jconfig.vms
+CONFIGUREFILES= config.guess config.sub install-sh ltconfig ltmain.sh
+OTHERFILES= jconfig.doc ckconfig.c ansi2knr.c ansi2knr.1 jmemdosa.asm
+TESTFILES= testorig.jpg testimg.ppm testimg.bmp testimg.jpg testprog.jpg &
+        testimgp.jpg
+DISTFILES= $(DOCS) $(MKFILES) $(CONFIGFILES) $(SOURCES) $(INCLUDES) &
+        $(CONFIGUREFILES) $(OTHERFILES) $(TESTFILES)
+# library object files common to compression and decompression
+COMOBJECTS= jcomapi.obj jutils.obj jerror.obj jmemmgr.obj $(SYSDEPMEM)
+# compression library object files
+CLIBOBJECTS= jcapimin.obj jcapistd.obj jctrans.obj jcparam.obj jdatadst.obj &
+        jcinit.obj jcmaster.obj jcmarker.obj jcmainct.obj jcprepct.obj &
+        jccoefct.obj jccolor.obj jcsample.obj jchuff.obj jcphuff.obj &
+        jcdctmgr.obj jfdctfst.obj jfdctflt.obj jfdctint.obj
+# decompression library object files
+DLIBOBJECTS= jdapimin.obj jdapistd.obj jdtrans.obj jdatasrc.obj &
+        jdmaster.obj jdinput.obj jdmarker.obj jdhuff.obj jdphuff.obj &
+        jdmainct.obj jdcoefct.obj jdpostct.obj jddctmgr.obj jidctfst.obj &
+        jidctflt.obj jidctint.obj jidctred.obj jdsample.obj jdcolor.obj &
+        jquant1.obj jquant2.obj jdmerge.obj
+# These objectfiles are included in libjpeg.lib
+LIBOBJECTS= $(CLIBOBJECTS) $(DLIBOBJECTS) $(COMOBJECTS)
+# object files for sample applications (excluding library files)
+COBJECTS= cjpeg.obj rdppm.obj rdgif.obj rdtarga.obj rdrle.obj rdbmp.obj &
+        rdswitch.obj cdjpeg.obj
+DOBJECTS= djpeg.obj wrppm.obj wrgif.obj wrtarga.obj wrrle.obj wrbmp.obj &
+        rdcolmap.obj cdjpeg.obj
+TROBJECTS= jpegtran.obj rdswitch.obj cdjpeg.obj transupp.obj
+
+
+all: libjpeg.lib cjpeg.exe djpeg.exe jpegtran.exe rdjpgcom.exe wrjpgcom.exe
+
+libjpeg.lib: $(LIBOBJECTS)
+	- del libjpeg.lib
+	* wlib -n libjpeg.lib $(LIBOBJECTS)
+
+cjpeg.exe: $(COBJECTS) libjpeg.lib
+	$(CC) $(LDFLAGS) $(COBJECTS) libjpeg.lib
+
+djpeg.exe: $(DOBJECTS) libjpeg.lib
+	$(CC) $(LDFLAGS) $(DOBJECTS) libjpeg.lib
+
+jpegtran.exe: $(TROBJECTS) libjpeg.lib
+	$(CC) $(LDFLAGS) $(TROBJECTS) libjpeg.lib
+
+rdjpgcom.exe: rdjpgcom.c
+	$(CC) $(CFLAGS) $(LDFLAGS) rdjpgcom.c
+
+wrjpgcom.exe: wrjpgcom.c
+	$(CC) $(CFLAGS) $(LDFLAGS) wrjpgcom.c
+
+.c.obj:
+	$(CC) $(CFLAGS) -c $<
+
+jconfig.h: jconfig.doc
+	echo You must prepare a system-dependent jconfig.h file.
+	echo Please read the installation directions in install.doc.
+	exit 1
+
+clean: .SYMBOLIC
+	- del *.obj
+	- del libjpeg.lib
+	- del cjpeg.exe
+	- del djpeg.exe
+	- del jpegtran.exe
+	- del rdjpgcom.exe
+	- del wrjpgcom.exe
+	- del testout*.*
+
+test: cjpeg.exe djpeg.exe jpegtran.exe  .SYMBOLIC
+	- del testout*.*
+	djpeg -dct int -ppm -outfile testout.ppm  testorig.jpg
+	djpeg -dct int -bmp -colors 256 -outfile testout.bmp  testorig.jpg
+	cjpeg -dct int -outfile testout.jpg  testimg.ppm
+	djpeg -dct int -ppm -outfile testoutp.ppm testprog.jpg
+	cjpeg -dct int -progressive -opt -outfile testoutp.jpg testimg.ppm
+	jpegtran -outfile testoutt.jpg testprog.jpg
+!ifeq SYSTEM DOS
+	fc /b testimg.ppm testout.ppm
+	fc /b testimg.bmp testout.bmp
+	fc /b testimg.jpg testout.jpg
+	fc /b testimg.ppm testoutp.ppm
+	fc /b testimgp.jpg testoutp.jpg
+	fc /b testorig.jpg testoutt.jpg
+!else
+	echo n > n.tmp
+	comp testimg.ppm testout.ppm < n.tmp
+	comp testimg.bmp testout.bmp < n.tmp
+	comp testimg.jpg testout.jpg < n.tmp
+	comp testimg.ppm testoutp.ppm < n.tmp
+	comp testimgp.jpg testoutp.jpg < n.tmp
+	comp testorig.jpg testoutt.jpg < n.tmp
+	del n.tmp
+!endif
+
+
+jcapimin.obj: jcapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcapistd.obj: jcapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccoefct.obj: jccoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jccolor.obj: jccolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcdctmgr.obj: jcdctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jchuff.obj: jchuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcinit.obj: jcinit.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmainct.obj: jcmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmarker.obj: jcmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcmaster.obj: jcmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcomapi.obj: jcomapi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcparam.obj: jcparam.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcphuff.obj: jcphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jchuff.h
+jcprepct.obj: jcprepct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jcsample.obj: jcsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jctrans.obj: jctrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapimin.obj: jdapimin.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdapistd.obj: jdapistd.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdatadst.obj: jdatadst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdatasrc.obj: jdatasrc.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h
+jdcoefct.obj: jdcoefct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdcolor.obj: jdcolor.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jddctmgr.obj: jddctmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jdhuff.obj: jdhuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdinput.obj: jdinput.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmainct.obj: jdmainct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmarker.obj: jdmarker.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmaster.obj: jdmaster.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdmerge.obj: jdmerge.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdphuff.obj: jdphuff.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdhuff.h
+jdpostct.obj: jdpostct.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdsample.obj: jdsample.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jdtrans.obj: jdtrans.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jerror.obj: jerror.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jversion.h jerror.h
+jfdctflt.obj: jfdctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctfst.obj: jfdctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jfdctint.obj: jfdctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctflt.obj: jidctflt.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctfst.obj: jidctfst.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctint.obj: jidctint.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jidctred.obj: jidctred.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jdct.h
+jquant1.obj: jquant1.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jquant2.obj: jquant2.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jutils.obj: jutils.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h
+jmemmgr.obj: jmemmgr.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemansi.obj: jmemansi.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemname.obj: jmemname.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemnobs.obj: jmemnobs.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemdos.obj: jmemdos.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+jmemmac.obj: jmemmac.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h jmemsys.h
+cjpeg.obj: cjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+djpeg.obj: djpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h jversion.h
+jpegtran.obj: jpegtran.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h transupp.h jversion.h
+rdjpgcom.obj: rdjpgcom.c jinclude.h jconfig.h
+wrjpgcom.obj: wrjpgcom.c jinclude.h jconfig.h
+cdjpeg.obj: cdjpeg.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdcolmap.obj: rdcolmap.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdswitch.obj: rdswitch.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+transupp.obj: transupp.c jinclude.h jconfig.h jpeglib.h jmorecfg.h jpegint.h jerror.h transupp.h
+rdppm.obj: rdppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrppm.obj: wrppm.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdgif.obj: rdgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrgif.obj: wrgif.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdtarga.obj: rdtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrtarga.obj: wrtarga.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdbmp.obj: rdbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrbmp.obj: wrbmp.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+rdrle.obj: rdrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
+wrrle.obj: wrrle.c cdjpeg.h jinclude.h jconfig.h jpeglib.h jmorecfg.h jerror.h cderror.h
diff --git a/libjpegtwrp/makelib.ds b/libjpegtwrp/makelib.ds
new file mode 100644
index 000000000..c7ad36d09
--- /dev/null
+++ b/libjpegtwrp/makelib.ds
@@ -0,0 +1,1046 @@
+# Microsoft Developer Studio Generated NMAKE File, Format Version 4.20
+# ** DO NOT EDIT **
+
+# TARGTYPE "Win32 (x86) Static Library" 0x0104
+
+!IF "$(CFG)" == ""
+CFG=jpeg - Win32
+!MESSAGE No configuration specified.  Defaulting to jpeg - Win32.
+!ENDIF 
+
+!IF "$(CFG)" != "jpeg - Win32"
+!MESSAGE Invalid configuration "$(CFG)" specified.
+!MESSAGE You can specify a configuration when running NMAKE on this makefile
+!MESSAGE by defining the macro CFG on the command line.  For example:
+!MESSAGE 
+!MESSAGE NMAKE /f "jpeg.mak" CFG="jpeg - Win32"
+!MESSAGE 
+!MESSAGE Possible choices for configuration are:
+!MESSAGE 
+!MESSAGE "jpeg - Win32" (based on "Win32 (x86) Static Library")
+!MESSAGE 
+!ERROR An invalid configuration is specified.
+!ENDIF 
+
+!IF "$(OS)" == "Windows_NT"
+NULL=
+!ELSE 
+NULL=nul
+!ENDIF 
+################################################################################
+# Begin Project
+# PROP Target_Last_Scanned "jpeg - Win32"
+CPP=cl.exe
+
+!IF  "$(CFG)" == "jpeg - Win32"
+
+# PROP BASE Use_MFC 0
+# PROP BASE Use_Debug_Libraries 0
+# PROP BASE Output_Dir "Release"
+# PROP BASE Intermediate_Dir "Release"
+# PROP BASE Target_Dir ""
+# PROP Use_MFC 0
+# PROP Use_Debug_Libraries 0
+# PROP Output_Dir "Release"
+# PROP Intermediate_Dir "Release"
+# PROP Target_Dir ""
+OUTDIR=.\Release
+INTDIR=.\Release
+
+ALL : "$(OUTDIR)\jpeg.lib"
+
+CLEAN : 
+	-@erase "$(INTDIR)\jcapimin.obj"
+	-@erase "$(INTDIR)\jcapistd.obj"
+	-@erase "$(INTDIR)\jctrans.obj"
+	-@erase "$(INTDIR)\jcparam.obj"
+	-@erase "$(INTDIR)\jdatadst.obj"
+	-@erase "$(INTDIR)\jcinit.obj"
+	-@erase "$(INTDIR)\jcmaster.obj"
+	-@erase "$(INTDIR)\jcmarker.obj"
+	-@erase "$(INTDIR)\jcmainct.obj"
+	-@erase "$(INTDIR)\jcprepct.obj"
+	-@erase "$(INTDIR)\jccoefct.obj"
+	-@erase "$(INTDIR)\jccolor.obj"
+	-@erase "$(INTDIR)\jcsample.obj"
+	-@erase "$(INTDIR)\jchuff.obj"
+	-@erase "$(INTDIR)\jcphuff.obj"
+	-@erase "$(INTDIR)\jcdctmgr.obj"
+	-@erase "$(INTDIR)\jfdctfst.obj"
+	-@erase "$(INTDIR)\jfdctflt.obj"
+	-@erase "$(INTDIR)\jfdctint.obj"
+	-@erase "$(INTDIR)\jdapimin.obj"
+	-@erase "$(INTDIR)\jdapistd.obj"
+	-@erase "$(INTDIR)\jdtrans.obj"
+	-@erase "$(INTDIR)\jdatasrc.obj"
+	-@erase "$(INTDIR)\jdmaster.obj"
+	-@erase "$(INTDIR)\jdinput.obj"
+	-@erase "$(INTDIR)\jdmarker.obj"
+	-@erase "$(INTDIR)\jdhuff.obj"
+	-@erase "$(INTDIR)\jdphuff.obj"
+	-@erase "$(INTDIR)\jdmainct.obj"
+	-@erase "$(INTDIR)\jdcoefct.obj"
+	-@erase "$(INTDIR)\jdpostct.obj"
+	-@erase "$(INTDIR)\jddctmgr.obj"
+	-@erase "$(INTDIR)\jidctfst.obj"
+	-@erase "$(INTDIR)\jidctflt.obj"
+	-@erase "$(INTDIR)\jidctint.obj"
+	-@erase "$(INTDIR)\jidctred.obj"
+	-@erase "$(INTDIR)\jdsample.obj"
+	-@erase "$(INTDIR)\jdcolor.obj"
+	-@erase "$(INTDIR)\jquant1.obj"
+	-@erase "$(INTDIR)\jquant2.obj"
+	-@erase "$(INTDIR)\jdmerge.obj"
+	-@erase "$(INTDIR)\jcomapi.obj"
+	-@erase "$(INTDIR)\jutils.obj"
+	-@erase "$(INTDIR)\jerror.obj"
+	-@erase "$(INTDIR)\jmemmgr.obj"
+	-@erase "$(INTDIR)\jmemnobs.obj"
+	-@erase "$(OUTDIR)\jpeg.lib"
+
+"$(OUTDIR)" :
+    if not exist "$(OUTDIR)/$(NULL)" mkdir "$(OUTDIR)"
+
+# ADD BASE CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_WINDOWS" /YX /c
+# ADD CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_WINDOWS" /YX /c
+CPP_PROJ=/nologo /ML /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_WINDOWS"\
+ /Fp"$(INTDIR)/jpeg.pch" /YX /Fo"$(INTDIR)/" /c 
+CPP_OBJS=.\Release/
+CPP_SBRS=.\.
+BSC32=bscmake.exe
+# ADD BASE BSC32 /nologo
+# ADD BSC32 /nologo
+BSC32_FLAGS=/nologo /o"$(OUTDIR)/jpeg.bsc" 
+BSC32_SBRS= \
+	
+LIB32=link.exe -lib
+# ADD BASE LIB32 /nologo
+# ADD LIB32 /nologo
+LIB32_FLAGS=/nologo /out:"$(OUTDIR)/jpeg.lib" 
+LIB32_OBJS= \
+	"$(INTDIR)\jcapimin.obj" \
+	"$(INTDIR)\jcapistd.obj" \
+	"$(INTDIR)\jctrans.obj" \
+	"$(INTDIR)\jcparam.obj" \
+	"$(INTDIR)\jdatadst.obj" \
+	"$(INTDIR)\jcinit.obj" \
+	"$(INTDIR)\jcmaster.obj" \
+	"$(INTDIR)\jcmarker.obj" \
+	"$(INTDIR)\jcmainct.obj" \
+	"$(INTDIR)\jcprepct.obj" \
+	"$(INTDIR)\jccoefct.obj" \
+	"$(INTDIR)\jccolor.obj" \
+	"$(INTDIR)\jcsample.obj" \
+	"$(INTDIR)\jchuff.obj" \
+	"$(INTDIR)\jcphuff.obj" \
+	"$(INTDIR)\jcdctmgr.obj" \
+	"$(INTDIR)\jfdctfst.obj" \
+	"$(INTDIR)\jfdctflt.obj" \
+	"$(INTDIR)\jfdctint.obj" \
+	"$(INTDIR)\jdapimin.obj" \
+	"$(INTDIR)\jdapistd.obj" \
+	"$(INTDIR)\jdtrans.obj" \
+	"$(INTDIR)\jdatasrc.obj" \
+	"$(INTDIR)\jdmaster.obj" \
+	"$(INTDIR)\jdinput.obj" \
+	"$(INTDIR)\jdmarker.obj" \
+	"$(INTDIR)\jdhuff.obj" \
+	"$(INTDIR)\jdphuff.obj" \
+	"$(INTDIR)\jdmainct.obj" \
+	"$(INTDIR)\jdcoefct.obj" \
+	"$(INTDIR)\jdpostct.obj" \
+	"$(INTDIR)\jddctmgr.obj" \
+	"$(INTDIR)\jidctfst.obj" \
+	"$(INTDIR)\jidctflt.obj" \
+	"$(INTDIR)\jidctint.obj" \
+	"$(INTDIR)\jidctred.obj" \
+	"$(INTDIR)\jdsample.obj" \
+	"$(INTDIR)\jdcolor.obj" \
+	"$(INTDIR)\jquant1.obj" \
+	"$(INTDIR)\jquant2.obj" \
+	"$(INTDIR)\jdmerge.obj" \
+	"$(INTDIR)\jcomapi.obj" \
+	"$(INTDIR)\jutils.obj" \
+	"$(INTDIR)\jerror.obj" \
+	"$(INTDIR)\jmemmgr.obj" \
+	"$(INTDIR)\jmemnobs.obj"
+
+"$(OUTDIR)\jpeg.lib" : "$(OUTDIR)" $(DEF_FILE) $(LIB32_OBJS)
+    $(LIB32) @<<
+  $(LIB32_FLAGS) $(DEF_FLAGS) $(LIB32_OBJS)
+<<
+
+!ENDIF 
+
+.c{$(CPP_OBJS)}.obj:
+   $(CPP) $(CPP_PROJ) $<  
+
+.cpp{$(CPP_OBJS)}.obj:
+   $(CPP) $(CPP_PROJ) $<  
+
+.cxx{$(CPP_OBJS)}.obj:
+   $(CPP) $(CPP_PROJ) $<  
+
+.c{$(CPP_SBRS)}.sbr:
+   $(CPP) $(CPP_PROJ) $<  
+
+.cpp{$(CPP_SBRS)}.sbr:
+   $(CPP) $(CPP_PROJ) $<  
+
+.cxx{$(CPP_SBRS)}.sbr:
+   $(CPP) $(CPP_PROJ) $<  
+
+################################################################################
+# Begin Target
+
+# Name "jpeg - Win32"
+
+!IF  "$(CFG)" == "jpeg - Win32"
+
+!ENDIF 
+
+################################################################################
+# Begin Source File
+
+SOURCE="jcapimin.c"
+DEP_CPP_JCAPI=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jcapimin.obj" : $(SOURCE) $(DEP_CPP_JCAPI) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jcapistd.c"
+DEP_CPP_JCAPIS=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jcapistd.obj" : $(SOURCE) $(DEP_CPP_JCAPIS) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jccoefct.c"
+DEP_CPP_JCCOE=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jccoefct.obj" : $(SOURCE) $(DEP_CPP_JCCOE) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jccolor.c"
+DEP_CPP_JCCOL=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jccolor.obj" : $(SOURCE) $(DEP_CPP_JCCOL) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jcdctmgr.c"
+DEP_CPP_JCDCT=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jdct.h"\
+	
+
+"$(INTDIR)\jcdctmgr.obj" : $(SOURCE) $(DEP_CPP_JCDCT) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jchuff.c"
+DEP_CPP_JCHUF=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jchuff.h"\
+	
+
+"$(INTDIR)\jchuff.obj" : $(SOURCE) $(DEP_CPP_JCHUF) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jcinit.c"
+DEP_CPP_JCINI=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jcinit.obj" : $(SOURCE) $(DEP_CPP_JCINI) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jcmainct.c"
+DEP_CPP_JCMAI=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jcmainct.obj" : $(SOURCE) $(DEP_CPP_JCMAI) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jcmarker.c"
+DEP_CPP_JCMAR=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jcmarker.obj" : $(SOURCE) $(DEP_CPP_JCMAR) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jcmaster.c"
+DEP_CPP_JCMAS=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jcmaster.obj" : $(SOURCE) $(DEP_CPP_JCMAS) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jcomapi.c"
+DEP_CPP_JCOMA=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jcomapi.obj" : $(SOURCE) $(DEP_CPP_JCOMA) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jcparam.c"
+DEP_CPP_JCPAR=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jcparam.obj" : $(SOURCE) $(DEP_CPP_JCPAR) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jcphuff.c"
+DEP_CPP_JCPHU=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jchuff.h"\
+	
+
+"$(INTDIR)\jcphuff.obj" : $(SOURCE) $(DEP_CPP_JCPHU) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jcprepct.c"
+DEP_CPP_JCPRE=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jcprepct.obj" : $(SOURCE) $(DEP_CPP_JCPRE) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jcsample.c"
+DEP_CPP_JCSAM=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jcsample.obj" : $(SOURCE) $(DEP_CPP_JCSAM) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jctrans.c"
+DEP_CPP_JCTRA=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jctrans.obj" : $(SOURCE) $(DEP_CPP_JCTRA) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdapimin.c"
+DEP_CPP_JDAPI=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdapimin.obj" : $(SOURCE) $(DEP_CPP_JDAPI) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdapistd.c"
+DEP_CPP_JDAPIS=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdapistd.obj" : $(SOURCE) $(DEP_CPP_JDAPIS) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdatadst.c"
+DEP_CPP_JDATA=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdatadst.obj" : $(SOURCE) $(DEP_CPP_JDATA) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdatasrc.c"
+DEP_CPP_JDATAS=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdatasrc.obj" : $(SOURCE) $(DEP_CPP_JDATAS) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdcoefct.c"
+DEP_CPP_JDCOE=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdcoefct.obj" : $(SOURCE) $(DEP_CPP_JDCOE) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdcolor.c"
+DEP_CPP_JDCOL=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdcolor.obj" : $(SOURCE) $(DEP_CPP_JDCOL) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jddctmgr.c"
+DEP_CPP_JDDCT=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jdct.h"\
+	
+
+"$(INTDIR)\jddctmgr.obj" : $(SOURCE) $(DEP_CPP_JDDCT) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdhuff.c"
+DEP_CPP_JDHUF=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jdhuff.h"\
+	
+
+"$(INTDIR)\jdhuff.obj" : $(SOURCE) $(DEP_CPP_JDHUF) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdinput.c"
+DEP_CPP_JDINP=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdinput.obj" : $(SOURCE) $(DEP_CPP_JDINP) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdmainct.c"
+DEP_CPP_JDMAI=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdmainct.obj" : $(SOURCE) $(DEP_CPP_JDMAI) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdmarker.c"
+DEP_CPP_JDMAR=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdmarker.obj" : $(SOURCE) $(DEP_CPP_JDMAR) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdmaster.c"
+DEP_CPP_JDMAS=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdmaster.obj" : $(SOURCE) $(DEP_CPP_JDMAS) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdmerge.c"
+DEP_CPP_JDMER=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdmerge.obj" : $(SOURCE) $(DEP_CPP_JDMER) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdphuff.c"
+DEP_CPP_JDPHU=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jdhuff.h"\
+	
+
+"$(INTDIR)\jdphuff.obj" : $(SOURCE) $(DEP_CPP_JDPHU) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdpostct.c"
+DEP_CPP_JDPOS=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdpostct.obj" : $(SOURCE) $(DEP_CPP_JDPOS) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdsample.c"
+DEP_CPP_JDSAM=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdsample.obj" : $(SOURCE) $(DEP_CPP_JDSAM) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jdtrans.c"
+DEP_CPP_JDTRA=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jdtrans.obj" : $(SOURCE) $(DEP_CPP_JDTRA) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jerror.c"
+DEP_CPP_JERRO=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jversion.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jerror.obj" : $(SOURCE) $(DEP_CPP_JERRO) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jfdctflt.c"
+DEP_CPP_JFDCT=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jdct.h"\
+	
+
+"$(INTDIR)\jfdctflt.obj" : $(SOURCE) $(DEP_CPP_JFDCT) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jfdctfst.c"
+DEP_CPP_JFDCTF=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jdct.h"\
+	
+
+"$(INTDIR)\jfdctfst.obj" : $(SOURCE) $(DEP_CPP_JFDCTF) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jfdctint.c"
+DEP_CPP_JFDCTI=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jdct.h"\
+	
+
+"$(INTDIR)\jfdctint.obj" : $(SOURCE) $(DEP_CPP_JFDCTI) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jidctflt.c"
+DEP_CPP_JIDCT=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jdct.h"\
+	
+
+"$(INTDIR)\jidctflt.obj" : $(SOURCE) $(DEP_CPP_JIDCT) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jidctfst.c"
+DEP_CPP_JIDCTF=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jdct.h"\
+	
+
+"$(INTDIR)\jidctfst.obj" : $(SOURCE) $(DEP_CPP_JIDCTF) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jidctint.c"
+DEP_CPP_JIDCTI=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jdct.h"\
+	
+
+"$(INTDIR)\jidctint.obj" : $(SOURCE) $(DEP_CPP_JIDCTI) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jidctred.c"
+DEP_CPP_JIDCTR=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jdct.h"\
+	
+
+"$(INTDIR)\jidctred.obj" : $(SOURCE) $(DEP_CPP_JIDCTR) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jquant1.c"
+DEP_CPP_JQUAN=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jquant1.obj" : $(SOURCE) $(DEP_CPP_JQUAN) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jquant2.c"
+DEP_CPP_JQUANT=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jquant2.obj" : $(SOURCE) $(DEP_CPP_JQUANT) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jutils.c"
+DEP_CPP_JUTIL=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	
+
+"$(INTDIR)\jutils.obj" : $(SOURCE) $(DEP_CPP_JUTIL) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jmemmgr.c"
+DEP_CPP_JMEMM=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jmemsys.h"\
+	
+
+"$(INTDIR)\jmemmgr.obj" : $(SOURCE) $(DEP_CPP_JMEMM) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+################################################################################
+# Begin Source File
+
+SOURCE="jmemnobs.c"
+DEP_CPP_JMEMN=\
+	"jinclude.h"\
+	"jconfig.h"\
+	"jpeglib.h"\
+	"jmorecfg.h"\
+	"jpegint.h"\
+	"jerror.h"\
+	"jmemsys.h"\
+	
+
+"$(INTDIR)\jmemnobs.obj" : $(SOURCE) $(DEP_CPP_JMEMN) "$(INTDIR)"
+   $(CPP) $(CPP_PROJ) $(SOURCE)
+
+
+# End Source File
+# End Target
+# End Project
+################################################################################
+
diff --git a/libjpegtwrp/makeproj.mac b/libjpegtwrp/makeproj.mac
new file mode 100644
index 000000000..ed277c83d
--- /dev/null
+++ b/libjpegtwrp/makeproj.mac
@@ -0,0 +1,213 @@
+--
+-- makeproj.mac
+--
+-- This AppleScript builds Code Warrior PRO Release 2 project files for the
+-- libjpeg library as well as the test programs 'cjpeg', 'djpeg', 'jpegtran'.
+-- (We'd distribute real project files, except they're not text
+-- and would create maintenance headaches.)
+--
+-- The script then compiles and links the library and the test programs.
+-- NOTE: if you haven't already created a 'jconfig.h' file, the script
+-- automatically copies 'jconfig.mac' to 'jconfig.h'.
+--
+-- To use this script, you must have AppleScript 1.1 or later installed
+-- and a suitable AppleScript editor like Script Editor or Script Debugger
+-- (http://www.latenightsw.com). Open this file with your AppleScript
+-- editor and execute the "run" command to build the projects.
+--
+-- Thanks to Dan Sears and Don Agro for this script.
+-- Questions about this script can be addressed to dogpark@interlog.com
+--
+
+on run
+
+	choose folder with prompt ">>> Select IJG source folder <<<"
+	set ijg_folder to result
+
+	choose folder with prompt ">>> Select MetroWerks folder <<<"
+	set cw_folder to result
+
+	-- if jconfig.h doesn't already exist, copy jconfig.mac
+
+	tell application "Finder"
+		if not (exists file "jconfig.h" of ijg_folder) then
+			duplicate {file "jconfig.mac" of folder ijg_folder}
+			select file "jconfig.mac copy" of folder ijg_folder
+			set name of selection to "jconfig.h"
+		end if
+	end tell
+
+	tell application "CodeWarrior IDE 2.1"
+	  with timeout of 10000 seconds
+
+		-- create libjpeg project
+
+		activate
+		Create Project (ijg_folder as string) & "libjpeg.proj"
+		Set Preferences of panel "Target Settings" to {Target Name:"libjpeg"}
+		Set Preferences of panel "PPC Project" to {File Name:"libjpeg"}
+		Set Preferences of panel "Target Settings" to {Linker:"MacOS PPC Linker"}
+		Set Preferences of panel "PPC Project" to {Project Type:library}
+		Set Preferences of panel "C/C++ Compiler" to {ANSI Strict:true}
+		Set Preferences of panel "C/C++ Compiler" to {Enums Always Ints:true}
+		Set Preferences of panel "PPC Codegen" to {Struct Alignment:PowerPC}
+		Set Preferences of panel "PPC Linker" to {Generate SYM File:false}
+
+		Add Files (ijg_folder as string) & "jcapimin.c" To Segment 1
+		Add Files (ijg_folder as string) & "jcapistd.c" To Segment 1
+		Add Files (ijg_folder as string) & "jctrans.c" To Segment 1
+		Add Files (ijg_folder as string) & "jcparam.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdatadst.c" To Segment 1
+		Add Files (ijg_folder as string) & "jcinit.c" To Segment 1
+		Add Files (ijg_folder as string) & "jcmaster.c" To Segment 1
+		Add Files (ijg_folder as string) & "jcmarker.c" To Segment 1
+		Add Files (ijg_folder as string) & "jcmainct.c" To Segment 1
+		Add Files (ijg_folder as string) & "jcprepct.c" To Segment 1
+		Add Files (ijg_folder as string) & "jccoefct.c" To Segment 1
+		Add Files (ijg_folder as string) & "jccolor.c" To Segment 1
+		Add Files (ijg_folder as string) & "jcsample.c" To Segment 1
+		Add Files (ijg_folder as string) & "jchuff.c" To Segment 1
+		Add Files (ijg_folder as string) & "jcphuff.c" To Segment 1
+		Add Files (ijg_folder as string) & "jcdctmgr.c" To Segment 1
+		Add Files (ijg_folder as string) & "jfdctfst.c" To Segment 1
+		Add Files (ijg_folder as string) & "jfdctflt.c" To Segment 1
+		Add Files (ijg_folder as string) & "jfdctint.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdapimin.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdapistd.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdtrans.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdatasrc.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdmaster.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdinput.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdmarker.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdhuff.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdphuff.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdmainct.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdcoefct.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdpostct.c" To Segment 1
+		Add Files (ijg_folder as string) & "jddctmgr.c" To Segment 1
+		Add Files (ijg_folder as string) & "jidctfst.c" To Segment 1
+		Add Files (ijg_folder as string) & "jidctflt.c" To Segment 1
+		Add Files (ijg_folder as string) & "jidctint.c" To Segment 1
+		Add Files (ijg_folder as string) & "jidctred.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdsample.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdcolor.c" To Segment 1
+		Add Files (ijg_folder as string) & "jquant1.c" To Segment 1
+		Add Files (ijg_folder as string) & "jquant2.c" To Segment 1
+		Add Files (ijg_folder as string) & "jdmerge.c" To Segment 1
+		Add Files (ijg_folder as string) & "jcomapi.c" To Segment 1
+		Add Files (ijg_folder as string) & "jutils.c" To Segment 1
+		Add Files (ijg_folder as string) & "jerror.c" To Segment 1
+		Add Files (ijg_folder as string) & "jmemmgr.c" To Segment 1
+		Add Files (ijg_folder as string) & "jmemmac.c" To Segment 1
+
+		-- compile and link the library
+
+		Make Project
+		Close Project
+
+		-- create cjpeg project
+
+		activate
+		Create Project (ijg_folder as string) & "cjpeg.proj"
+		Set Preferences of panel "Target Settings" to {Target Name:"cjpeg"}
+		Set Preferences of panel "PPC Project" to {File Name:"cjpeg"}
+		Set Preferences of panel "Target Settings" to {Linker:"MacOS PPC Linker"}
+		Set Preferences of panel "C/C++ Compiler" to {ANSI Strict:true}
+		Set Preferences of panel "C/C++ Compiler" to {Enums Always Ints:true}
+		Set Preferences of panel "PPC Codegen" to {Struct Alignment:PowerPC}
+		Set Preferences of panel "PPC Linker" to {Generate SYM File:false}
+
+		Add Files (ijg_folder as string) & "cjpeg.c" To Segment 1
+		Add Files (ijg_folder as string) & "rdppm.c" To Segment 1
+		Add Files (ijg_folder as string) & "rdgif.c" To Segment 1
+		Add Files (ijg_folder as string) & "rdtarga.c" To Segment 1
+		Add Files (ijg_folder as string) & "rdrle.c" To Segment 1
+		Add Files (ijg_folder as string) & "rdbmp.c" To Segment 1
+		Add Files (ijg_folder as string) & "rdswitch.c" To Segment 1
+		Add Files (ijg_folder as string) & "cdjpeg.c" To Segment 1
+
+		Add Files (ijg_folder as string) & "libjpeg" To Segment 2
+
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:Metrowerks Standard Library:MSL C:Bin:MSL C.PPC.Lib" To Segment 3
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:Metrowerks Standard Library:MSL C:Bin:MSL SIOUX.PPC.Lib" To Segment 3
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:MacOS Support:Libraries:Runtime:Runtime PPC:MSL RuntimePPC.Lib" To Segment 3
+
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:MacOS Support:Libraries:MacOS Common:InterfaceLib" To Segment 4
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:MacOS Support:Libraries:MacOS Common:MathLib" To Segment 4
+
+		-- compile and link cjpeg
+
+		Make Project
+		Close Project
+
+		-- create djpeg project
+
+		activate
+		Create Project (ijg_folder as string) & "djpeg.proj"
+		Set Preferences of panel "Target Settings" to {Target Name:"djpeg"}
+		Set Preferences of panel "PPC Project" to {File Name:"djpeg"}
+		Set Preferences of panel "Target Settings" to {Linker:"MacOS PPC Linker"}
+		Set Preferences of panel "C/C++ Compiler" to {ANSI Strict:true}
+		Set Preferences of panel "C/C++ Compiler" to {Enums Always Ints:true}
+		Set Preferences of panel "PPC Codegen" to {Struct Alignment:PowerPC}
+		Set Preferences of panel "PPC Linker" to {Generate SYM File:false}
+
+		Add Files (ijg_folder as string) & "djpeg.c" To Segment 1
+		Add Files (ijg_folder as string) & "wrppm.c" To Segment 1
+		Add Files (ijg_folder as string) & "wrgif.c" To Segment 1
+		Add Files (ijg_folder as string) & "wrtarga.c" To Segment 1
+		Add Files (ijg_folder as string) & "wrrle.c" To Segment 1
+		Add Files (ijg_folder as string) & "wrbmp.c" To Segment 1
+		Add Files (ijg_folder as string) & "rdcolmap.c" To Segment 1
+		Add Files (ijg_folder as string) & "cdjpeg.c" To Segment 1
+
+		Add Files (ijg_folder as string) & "libjpeg" To Segment 2
+
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:Metrowerks Standard Library:MSL C:Bin:MSL C.PPC.Lib" To Segment 3
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:Metrowerks Standard Library:MSL C:Bin:MSL SIOUX.PPC.Lib" To Segment 3
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:MacOS Support:Libraries:Runtime:Runtime PPC:MSL RuntimePPC.Lib" To Segment 3
+
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:MacOS Support:Libraries:MacOS Common:InterfaceLib" To Segment 4
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:MacOS Support:Libraries:MacOS Common:MathLib" To Segment 4
+
+		-- compile and link djpeg
+
+		Make Project
+		Close Project
+
+		-- create jpegtran project
+
+		activate
+		Create Project (ijg_folder as string) & "jpegtran.proj"
+		Set Preferences of panel "Target Settings" to {Target Name:"jpegtran"}
+		Set Preferences of panel "PPC Project" to {File Name:"jpegtran"}
+		Set Preferences of panel "Target Settings" to {Linker:"MacOS PPC Linker"}
+		Set Preferences of panel "C/C++ Compiler" to {ANSI Strict:true}
+		Set Preferences of panel "C/C++ Compiler" to {Enums Always Ints:true}
+		Set Preferences of panel "PPC Codegen" to {Struct Alignment:PowerPC}
+		Set Preferences of panel "PPC Linker" to {Generate SYM File:false}
+
+		Add Files (ijg_folder as string) & "jpegtran.c" To Segment 1
+		Add Files (ijg_folder as string) & "rdswitch.c" To Segment 1
+		Add Files (ijg_folder as string) & "cdjpeg.c" To Segment 1
+		Add Files (ijg_folder as string) & "transupp.c" To Segment 1
+
+		Add Files (ijg_folder as string) & "libjpeg" To Segment 2
+
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:Metrowerks Standard Library:MSL C:Bin:MSL C.PPC.Lib" To Segment 3
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:Metrowerks Standard Library:MSL C:Bin:MSL SIOUX.PPC.Lib" To Segment 3
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:MacOS Support:Libraries:Runtime:Runtime PPC:MSL RuntimePPC.Lib" To Segment 3
+
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:MacOS Support:Libraries:MacOS Common:InterfaceLib" To Segment 4
+		Add Files (cw_folder as string) & "Metrowerks CodeWarrior:MacOS Support:Libraries:MacOS Common:MathLib" To Segment 4
+
+		-- compile and link jpegtran
+
+		Make Project
+		Close Project
+
+		quit
+
+	  end timeout
+	end tell
+end run
diff --git a/libjpegtwrp/makljpeg.st b/libjpegtwrp/makljpeg.st
new file mode 100644
index 000000000..813493ea8
--- /dev/null
+++ b/libjpegtwrp/makljpeg.st
@@ -0,0 +1,70 @@
+; Project file for Independent JPEG Group's software
+;
+; This project file is for Atari ST/STE/TT systems using Pure C or Turbo C.
+; Thanks to Frank Moehle (Frank.Moehle@arbi.informatik.uni-oldenburg.de),
+; Dr. B. Setzepfandt (bernd@gina.uni-muenster.de),
+; and Guido Vollbeding (guivol@esc.de).
+;
+; To use this file, rename it to libjpeg.prj.
+; Read installation instructions before trying to make the program!
+;
+;
+;      * * * Output file * * *
+libjpeg.lib
+;
+; * * * COMPILER OPTIONS * * *  
+.C[-P]        ; absolute calls
+.C[-M]        ; and no string merging, folks
+.C[-w-cln]    ; no "constant is long" warnings
+.C[-w-par]    ; no "parameter xxxx unused"
+.C[-w-rch]    ; no "unreachable code"
+.C[-wsig]     ; warn if significant digits may be lost
+.L[-J]        ; link new Obj-format (so we get a library)
+=
+; * * * * List of modules * * * * 
+jcapimin.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jcapistd.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jccoefct.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jccolor.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jcdctmgr.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jdct.h)
+jchuff.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jchuff.h)
+jcinit.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jcmainct.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jcmarker.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jcmaster.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jcomapi.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jcparam.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jcphuff.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jchuff.h)
+jcprepct.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jcsample.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jctrans.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jdapimin.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jdapistd.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jdatadst.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h)
+jdatasrc.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h)
+jdcoefct.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jdcolor.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jddctmgr.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jdct.h)
+jdhuff.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jdhuff.h)
+jdinput.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jdmainct.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jdmarker.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jdmaster.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jdmerge.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jdphuff.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jdhuff.h)
+jdpostct.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jdsample.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jdtrans.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jerror.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jversion.h,jerror.h)
+jfdctflt.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jdct.h)
+jfdctfst.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jdct.h)
+jfdctint.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jdct.h)
+jidctflt.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jdct.h)
+jidctfst.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jdct.h)
+jidctint.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jdct.h)
+jidctred.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jdct.h)
+jquant1.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jquant2.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jutils.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h)
+jmemmgr.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jmemsys.h)
+jmemansi.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,jmemsys.h)
diff --git a/libjpegtwrp/maktjpeg.st b/libjpegtwrp/maktjpeg.st
new file mode 100644
index 000000000..31f4d169c
--- /dev/null
+++ b/libjpegtwrp/maktjpeg.st
@@ -0,0 +1,32 @@
+; Project file for Independent JPEG Group's software
+;
+; This project file is for Atari ST/STE/TT systems using Pure C or Turbo C.
+; Thanks to Frank Moehle (Frank.Moehle@arbi.informatik.uni-oldenburg.de),
+; Dr. B. Setzepfandt (bernd@gina.uni-muenster.de),
+; and Guido Vollbeding (guivol@esc.de).
+;
+; To use this file, rename it to jpegtran.prj.
+; If you are using Turbo C, change filenames beginning with "pc..." to "tc..."
+; Read installation instructions before trying to make the program!
+;
+;
+;      * * * Output file * * *
+jpegtran.ttp
+;
+; * * * COMPILER OPTIONS * * *  
+.C[-P]        ; absolute calls
+.C[-M]        ; and no string merging, folks
+.C[-w-cln]    ; no "constant is long" warnings
+.C[-w-par]    ; no "parameter xxxx unused"
+.C[-w-rch]    ; no "unreachable code"
+.C[-wsig]     ; warn if significant digits may be lost
+=
+; * * * * List of modules * * * * 
+pcstart.o
+jpegtran.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h,transupp.h,jversion.h)
+cdjpeg.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+rdswitch.c	(cdjpeg.h,jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jerror.h,cderror.h)
+transupp.c	(jinclude.h,jconfig.h,jpeglib.h,jmorecfg.h,jpegint.h,jerror.h,transupp.h)
+libjpeg.lib        ; built by libjpeg.prj
+pcstdlib.lib       ; standard library
+pcextlib.lib       ; extended library
diff --git a/libjpegtwrp/makvms.opt b/libjpegtwrp/makvms.opt
new file mode 100644
index 000000000..675e8fe98
--- /dev/null
+++ b/libjpegtwrp/makvms.opt
@@ -0,0 +1,4 @@
+! A pointer to the VAX/VMS C Run-Time Shareable Library.
+! This file is needed by makefile.mms and makefile.vms,
+! but only for the older VAX C compiler.  DEC C does not need it.
+Sys$Library:VAXCRTL.EXE /Share
diff --git a/libjpegtwrp/rdbmp.c b/libjpegtwrp/rdbmp.c
new file mode 100644
index 000000000..b05fe2ac4
--- /dev/null
+++ b/libjpegtwrp/rdbmp.c
@@ -0,0 +1,439 @@
+/*
+ * rdbmp.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to read input images in Microsoft "BMP"
+ * format (MS Windows 3.x, OS/2 1.x, and OS/2 2.x flavors).
+ * Currently, only 8-bit and 24-bit images are supported, not 1-bit or
+ * 4-bit (feeding such low-depth images into JPEG would be silly anyway).
+ * Also, we don't support RLE-compressed files.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed BMP format).
+ *
+ * This code contributed by James Arthur Boucher.
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+
+#ifdef BMP_SUPPORTED
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+#ifdef HAVE_UNSIGNED_CHAR
+typedef unsigned char U_CHAR;
+#define UCH(x)	((int) (x))
+#else /* !HAVE_UNSIGNED_CHAR */
+#ifdef CHAR_IS_UNSIGNED
+typedef char U_CHAR;
+#define UCH(x)	((int) (x))
+#else
+typedef char U_CHAR;
+#define UCH(x)	((int) (x) & 0xFF)
+#endif
+#endif /* HAVE_UNSIGNED_CHAR */
+
+
+#define	ReadOK(file,buffer,len)	(JFREAD(file,buffer,len) == ((size_t) (len)))
+
+
+/* Private version of data source object */
+
+typedef struct _bmp_source_struct * bmp_source_ptr;
+
+typedef struct _bmp_source_struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  j_compress_ptr cinfo;		/* back link saves passing separate parm */
+
+  JSAMPARRAY colormap;		/* BMP colormap (converted to my format) */
+
+  jvirt_sarray_ptr whole_image;	/* Needed to reverse row order */
+  JDIMENSION source_row;	/* Current source row number */
+  JDIMENSION row_width;		/* Physical width of scanlines in file */
+
+  int bits_per_pixel;		/* remembers 8- or 24-bit format */
+} bmp_source_struct;
+
+
+LOCAL(int)
+read_byte (bmp_source_ptr sinfo)
+/* Read next byte from BMP file */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int c;
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(sinfo->cinfo, JERR_INPUT_EOF);
+  return c;
+}
+
+
+LOCAL(void)
+read_colormap (bmp_source_ptr sinfo, int cmaplen, int mapentrysize)
+/* Read the colormap from a BMP file */
+{
+  int i;
+
+  switch (mapentrysize) {
+  case 3:
+    /* BGR format (occurs in OS/2 files) */
+    for (i = 0; i < cmaplen; i++) {
+      sinfo->colormap[2][i] = (JSAMPLE) read_byte(sinfo);
+      sinfo->colormap[1][i] = (JSAMPLE) read_byte(sinfo);
+      sinfo->colormap[0][i] = (JSAMPLE) read_byte(sinfo);
+    }
+    break;
+  case 4:
+    /* BGR0 format (occurs in MS Windows files) */
+    for (i = 0; i < cmaplen; i++) {
+      sinfo->colormap[2][i] = (JSAMPLE) read_byte(sinfo);
+      sinfo->colormap[1][i] = (JSAMPLE) read_byte(sinfo);
+      sinfo->colormap[0][i] = (JSAMPLE) read_byte(sinfo);
+      (void) read_byte(sinfo);
+    }
+    break;
+  default:
+    ERREXIT(sinfo->cinfo, JERR_BMP_BADCMAP);
+    break;
+  }
+}
+
+
+/*
+ * Read one row of pixels.
+ * The image has been read into the whole_image array, but is otherwise
+ * unprocessed.  We must read it out in top-to-bottom row order, and if
+ * it is an 8-bit image, we must expand colormapped pixels to 24bit format.
+ */
+
+METHODDEF(JDIMENSION)
+get_8bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 8-bit colormap indexes */
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  register JSAMPARRAY colormap = source->colormap;
+  JSAMPARRAY image_ptr;
+  register int t;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+
+  /* Fetch next row from virtual array */
+  source->source_row--;
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->whole_image,
+     source->source_row, (JDIMENSION) 1, FALSE);
+
+  /* Expand the colormap indexes to real data */
+  inptr = image_ptr[0];
+  outptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    t = GETJSAMPLE(*inptr++);
+    *outptr++ = colormap[0][t];	/* can omit GETJSAMPLE() safely */
+    *outptr++ = colormap[1][t];
+    *outptr++ = colormap[2][t];
+  }
+
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_24bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 24-bit pixels */
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+
+  /* Fetch next row from virtual array */
+  source->source_row--;
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->whole_image,
+     source->source_row, (JDIMENSION) 1, FALSE);
+
+  /* Transfer data.  Note source values are in BGR order
+   * (even though Microsoft's own documents say the opposite).
+   */
+  inptr = image_ptr[0];
+  outptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    outptr[2] = *inptr++;	/* can omit GETJSAMPLE() safely */
+    outptr[1] = *inptr++;
+    outptr[0] = *inptr++;
+    outptr += 3;
+  }
+
+  return 1;
+}
+
+
+/*
+ * This method loads the image into whole_image during the first call on
+ * get_pixel_rows.  The get_pixel_rows pointer is then adjusted to call
+ * get_8bit_row or get_24bit_row on subsequent calls.
+ */
+
+METHODDEF(JDIMENSION)
+preload_image (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  register FILE *infile = source->pub.input_file;
+  register int c;
+  register JSAMPROW out_ptr;
+  JSAMPARRAY image_ptr;
+  JDIMENSION row, col;
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+
+  /* Read the data into a virtual array in input-file row order. */
+  for (row = 0; row < cinfo->image_height; row++) {
+    if (progress != NULL) {
+      progress->pub.pass_counter = (long) row;
+      progress->pub.pass_limit = (long) cinfo->image_height;
+      (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+    }
+    image_ptr = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr) cinfo, source->whole_image,
+       row, (JDIMENSION) 1, TRUE);
+    out_ptr = image_ptr[0];
+    for (col = source->row_width; col > 0; col--) {
+      /* inline copy of read_byte() for speed */
+      if ((c = getc(infile)) == EOF)
+	ERREXIT(cinfo, JERR_INPUT_EOF);
+      *out_ptr++ = (JSAMPLE) c;
+    }
+  }
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+
+  /* Set up to read from the virtual array in top-to-bottom order */
+  switch (source->bits_per_pixel) {
+  case 8:
+    source->pub.get_pixel_rows = get_8bit_row;
+    break;
+  case 24:
+    source->pub.get_pixel_rows = get_24bit_row;
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+  }
+  source->source_row = cinfo->image_height;
+
+  /* And read the first row */
+  return (*source->pub.get_pixel_rows) (cinfo, sinfo);
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_bmp (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  U_CHAR bmpfileheader[14];
+  U_CHAR bmpinfoheader[64];
+#define GET_2B(array,offset)  ((unsigned int) UCH(array[offset]) + \
+			       (((unsigned int) UCH(array[offset+1])) << 8))
+#define GET_4B(array,offset)  ((INT32) UCH(array[offset]) + \
+			       (((INT32) UCH(array[offset+1])) << 8) + \
+			       (((INT32) UCH(array[offset+2])) << 16) + \
+			       (((INT32) UCH(array[offset+3])) << 24))
+  INT32 bfOffBits;
+  INT32 headerSize;
+  INT32 biWidth = 0;		/* initialize to avoid compiler warning */
+  INT32 biHeight = 0;
+  unsigned int biPlanes;
+  INT32 biCompression;
+  INT32 biXPelsPerMeter,biYPelsPerMeter;
+  INT32 biClrUsed = 0;
+  int mapentrysize = 0;		/* 0 indicates no colormap */
+  INT32 bPad;
+  JDIMENSION row_width;
+
+  /* Read and verify the bitmap file header */
+  if (! ReadOK(source->pub.input_file, bmpfileheader, 14))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  if (GET_2B(bmpfileheader,0) != 0x4D42) /* 'BM' */
+    ERREXIT(cinfo, JERR_BMP_NOT);
+  bfOffBits = (INT32) GET_4B(bmpfileheader,10);
+  /* We ignore the remaining fileheader fields */
+
+  /* The infoheader might be 12 bytes (OS/2 1.x), 40 bytes (Windows),
+   * or 64 bytes (OS/2 2.x).  Check the first 4 bytes to find out which.
+   */
+  if (! ReadOK(source->pub.input_file, bmpinfoheader, 4))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  headerSize = (INT32) GET_4B(bmpinfoheader,0);
+  if (headerSize < 12 || headerSize > 64)
+    ERREXIT(cinfo, JERR_BMP_BADHEADER);
+  if (! ReadOK(source->pub.input_file, bmpinfoheader+4, headerSize-4))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+
+  switch ((int) headerSize) {
+  case 12:
+    /* Decode OS/2 1.x header (Microsoft calls this a BITMAPCOREHEADER) */
+    biWidth = (INT32) GET_2B(bmpinfoheader,4);
+    biHeight = (INT32) GET_2B(bmpinfoheader,6);
+    biPlanes = GET_2B(bmpinfoheader,8);
+    source->bits_per_pixel = (int) GET_2B(bmpinfoheader,10);
+
+    switch (source->bits_per_pixel) {
+    case 8:			/* colormapped image */
+      mapentrysize = 3;		/* OS/2 uses RGBTRIPLE colormap */
+      TRACEMS2(cinfo, 1, JTRC_BMP_OS2_MAPPED, (int) biWidth, (int) biHeight);
+      break;
+    case 24:			/* RGB image */
+      TRACEMS2(cinfo, 1, JTRC_BMP_OS2, (int) biWidth, (int) biHeight);
+      break;
+    default:
+      ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+      break;
+    }
+    if (biPlanes != 1)
+      ERREXIT(cinfo, JERR_BMP_BADPLANES);
+    break;
+  case 40:
+  case 64:
+    /* Decode Windows 3.x header (Microsoft calls this a BITMAPINFOHEADER) */
+    /* or OS/2 2.x header, which has additional fields that we ignore */
+    biWidth = GET_4B(bmpinfoheader,4);
+    biHeight = GET_4B(bmpinfoheader,8);
+    biPlanes = GET_2B(bmpinfoheader,12);
+    source->bits_per_pixel = (int) GET_2B(bmpinfoheader,14);
+    biCompression = GET_4B(bmpinfoheader,16);
+    biXPelsPerMeter = GET_4B(bmpinfoheader,24);
+    biYPelsPerMeter = GET_4B(bmpinfoheader,28);
+    biClrUsed = GET_4B(bmpinfoheader,32);
+    /* biSizeImage, biClrImportant fields are ignored */
+
+    switch (source->bits_per_pixel) {
+    case 8:			/* colormapped image */
+      mapentrysize = 4;		/* Windows uses RGBQUAD colormap */
+      TRACEMS2(cinfo, 1, JTRC_BMP_MAPPED, (int) biWidth, (int) biHeight);
+      break;
+    case 24:			/* RGB image */
+      TRACEMS2(cinfo, 1, JTRC_BMP, (int) biWidth, (int) biHeight);
+      break;
+    default:
+      ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+      break;
+    }
+    if (biPlanes != 1)
+      ERREXIT(cinfo, JERR_BMP_BADPLANES);
+    if (biCompression != 0)
+      ERREXIT(cinfo, JERR_BMP_COMPRESSED);
+
+    if (biXPelsPerMeter > 0 && biYPelsPerMeter > 0) {
+      /* Set JFIF density parameters from the BMP data */
+      cinfo->X_density = (UINT16) (biXPelsPerMeter/100); /* 100 cm per meter */
+      cinfo->Y_density = (UINT16) (biYPelsPerMeter/100);
+      cinfo->density_unit = 2;	/* dots/cm */
+    }
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BMP_BADHEADER);
+    break;
+  }
+
+  /* Compute distance to bitmap data --- will adjust for colormap below */
+  bPad = bfOffBits - (headerSize + 14);
+
+  /* Read the colormap, if any */
+  if (mapentrysize > 0) {
+    if (biClrUsed <= 0)
+      biClrUsed = 256;		/* assume it's 256 */
+    else if (biClrUsed > 256)
+      ERREXIT(cinfo, JERR_BMP_BADCMAP);
+    /* Allocate space to store the colormap */
+    source->colormap = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) biClrUsed, (JDIMENSION) 3);
+    /* and read it from the file */
+    read_colormap(source, (int) biClrUsed, mapentrysize);
+    /* account for size of colormap */
+    bPad -= biClrUsed * mapentrysize;
+  }
+
+  /* Skip any remaining pad bytes */
+  if (bPad < 0)			/* incorrect bfOffBits value? */
+    ERREXIT(cinfo, JERR_BMP_BADHEADER);
+  while (--bPad >= 0) {
+    (void) read_byte(source);
+  }
+
+  /* Compute row width in file, including padding to 4-byte boundary */
+  if (source->bits_per_pixel == 24)
+    row_width = (JDIMENSION) (biWidth * 3);
+  else
+    row_width = (JDIMENSION) biWidth;
+  while ((row_width & 3) != 0) row_width++;
+  source->row_width = row_width;
+
+  /* Allocate space for inversion array, prepare for preload pass */
+  source->whole_image = (*cinfo->mem->request_virt_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+     row_width, (JDIMENSION) biHeight, (JDIMENSION) 1);
+  source->pub.get_pixel_rows = preload_image;
+  if (cinfo->progress != NULL) {
+    cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+    progress->total_extra_passes++; /* count file input as separate pass */
+  }
+
+  /* Allocate one-row buffer for returned data */
+  source->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     (JDIMENSION) (biWidth * 3), (JDIMENSION) 1);
+  source->pub.buffer_height = 1;
+
+  cinfo->in_color_space = JCS_RGB;
+  cinfo->input_components = 3;
+  cinfo->data_precision = 8;
+  cinfo->image_width = (JDIMENSION) biWidth;
+  cinfo->image_height = (JDIMENSION) biHeight;
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_bmp (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for BMP format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_bmp (j_compress_ptr cinfo)
+{
+  bmp_source_ptr source;
+
+  /* Create module interface object */
+  source = (bmp_source_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(bmp_source_struct));
+  source->cinfo = cinfo;	/* make back link for subroutines */
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_bmp;
+  source->pub.finish_input = finish_input_bmp;
+
+  return (cjpeg_source_ptr) source;
+}
+
+#endif /* BMP_SUPPORTED */
diff --git a/libjpegtwrp/rdcolmap.c b/libjpegtwrp/rdcolmap.c
new file mode 100644
index 000000000..42b343763
--- /dev/null
+++ b/libjpegtwrp/rdcolmap.c
@@ -0,0 +1,253 @@
+/*
+ * rdcolmap.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file implements djpeg's "-map file" switch.  It reads a source image
+ * and constructs a colormap to be supplied to the JPEG decompressor.
+ *
+ * Currently, these file formats are supported for the map file:
+ *   GIF: the contents of the GIF's global colormap are used.
+ *   PPM (either text or raw flavor): the entire file is read and
+ *      each unique pixel value is entered in the map.
+ * Note that reading a large PPM file will be horrendously slow.
+ * Typically, a PPM-format map file should contain just one pixel
+ * of each desired color.  Such a file can be extracted from an
+ * ordinary image PPM file with ppmtomap(1).
+ *
+ * Rescaling a PPM that has a maxval unequal to MAXJSAMPLE is not
+ * currently implemented.
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+
+#ifdef QUANT_2PASS_SUPPORTED	/* otherwise can't quantize to supplied map */
+
+/* Portions of this code are based on the PBMPLUS library, which is:
+**
+** Copyright (C) 1988 by Jef Poskanzer.
+**
+** Permission to use, copy, modify, and distribute this software and its
+** documentation for any purpose and without fee is hereby granted, provided
+** that the above copyright notice appear in all copies and that both that
+** copyright notice and this permission notice appear in supporting
+** documentation.  This software is provided "as is" without express or
+** implied warranty.
+*/
+
+
+/*
+ * Add a (potentially) new color to the color map.
+ */
+
+LOCAL(void)
+add_map_entry (j_decompress_ptr cinfo, int R, int G, int B)
+{
+  JSAMPROW colormap0 = cinfo->colormap[0];
+  JSAMPROW colormap1 = cinfo->colormap[1];
+  JSAMPROW colormap2 = cinfo->colormap[2];
+  int ncolors = cinfo->actual_number_of_colors;
+  int index;
+
+  /* Check for duplicate color. */
+  for (index = 0; index < ncolors; index++) {
+    if (GETJSAMPLE(colormap0[index]) == R &&
+	GETJSAMPLE(colormap1[index]) == G &&
+	GETJSAMPLE(colormap2[index]) == B)
+      return;			/* color is already in map */
+  }
+
+  /* Check for map overflow. */
+  if (ncolors >= (MAXJSAMPLE+1))
+    ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, (MAXJSAMPLE+1));
+
+  /* OK, add color to map. */
+  colormap0[ncolors] = (JSAMPLE) R;
+  colormap1[ncolors] = (JSAMPLE) G;
+  colormap2[ncolors] = (JSAMPLE) B;
+  cinfo->actual_number_of_colors++;
+}
+
+
+/*
+ * Extract color map from a GIF file.
+ */
+
+LOCAL(void)
+read_gif_map (j_decompress_ptr cinfo, FILE * infile)
+{
+  int header[13];
+  int i, colormaplen;
+  int R, G, B;
+
+  /* Initial 'G' has already been read by read_color_map */
+  /* Read the rest of the GIF header and logical screen descriptor */
+  for (i = 1; i < 13; i++) {
+    if ((header[i] = getc(infile)) == EOF)
+      ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+  }
+
+  /* Verify GIF Header */
+  if (header[1] != 'I' || header[2] != 'F')
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  /* There must be a global color map. */
+  if ((header[10] & 0x80) == 0)
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  /* OK, fetch it. */
+  colormaplen = 2 << (header[10] & 0x07);
+
+  for (i = 0; i < colormaplen; i++) {
+    R = getc(infile);
+    G = getc(infile);
+    B = getc(infile);
+    if (R == EOF || G == EOF || B == EOF)
+      ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+    add_map_entry(cinfo,
+		  R << (BITS_IN_JSAMPLE-8),
+		  G << (BITS_IN_JSAMPLE-8),
+		  B << (BITS_IN_JSAMPLE-8));
+  }
+}
+
+
+/* Support routines for reading PPM */
+
+
+LOCAL(int)
+pbm_getc (FILE * infile)
+/* Read next char, skipping over any comments */
+/* A comment/newline sequence is returned as a newline */
+{
+  register int ch;
+  
+  ch = getc(infile);
+  if (ch == '#') {
+    do {
+      ch = getc(infile);
+    } while (ch != '\n' && ch != EOF);
+  }
+  return ch;
+}
+
+
+LOCAL(unsigned int)
+read_pbm_integer (j_decompress_ptr cinfo, FILE * infile)
+/* Read an unsigned decimal integer from the PPM file */
+/* Swallows one trailing character after the integer */
+/* Note that on a 16-bit-int machine, only values up to 64k can be read. */
+/* This should not be a problem in practice. */
+{
+  register int ch;
+  register unsigned int val;
+  
+  /* Skip any leading whitespace */
+  do {
+    ch = pbm_getc(infile);
+    if (ch == EOF)
+      ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+  } while (ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r');
+  
+  if (ch < '0' || ch > '9')
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+  
+  val = ch - '0';
+  while ((ch = pbm_getc(infile)) >= '0' && ch <= '9') {
+    val *= 10;
+    val += ch - '0';
+  }
+  return val;
+}
+
+
+/*
+ * Extract color map from a PPM file.
+ */
+
+LOCAL(void)
+read_ppm_map (j_decompress_ptr cinfo, FILE * infile)
+{
+  int c;
+  unsigned int w, h, maxval, row, col;
+  int R, G, B;
+
+  /* Initial 'P' has already been read by read_color_map */
+  c = getc(infile);		/* save format discriminator for a sec */
+
+  /* while we fetch the remaining header info */
+  w = read_pbm_integer(cinfo, infile);
+  h = read_pbm_integer(cinfo, infile);
+  maxval = read_pbm_integer(cinfo, infile);
+
+  if (w <= 0 || h <= 0 || maxval <= 0) /* error check */
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  /* For now, we don't support rescaling from an unusual maxval. */
+  if (maxval != (unsigned int) MAXJSAMPLE)
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  switch (c) {
+  case '3':			/* it's a text-format PPM file */
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+	R = read_pbm_integer(cinfo, infile);
+	G = read_pbm_integer(cinfo, infile);
+	B = read_pbm_integer(cinfo, infile);
+	add_map_entry(cinfo, R, G, B);
+      }
+    }
+    break;
+
+  case '6':			/* it's a raw-format PPM file */
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+	R = getc(infile);
+	G = getc(infile);
+	B = getc(infile);
+	if (R == EOF || G == EOF || B == EOF)
+	  ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+	add_map_entry(cinfo, R, G, B);
+      }
+    }
+    break;
+
+  default:
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+    break;
+  }
+}
+
+
+/*
+ * Main entry point from djpeg.c.
+ *  Input: opened input file (from file name argument on command line).
+ *  Output: colormap and actual_number_of_colors fields are set in cinfo.
+ */
+
+GLOBAL(void)
+read_color_map (j_decompress_ptr cinfo, FILE * infile)
+{
+  /* Allocate space for a color map of maximum supported size. */
+  cinfo->colormap = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     (JDIMENSION) (MAXJSAMPLE+1), (JDIMENSION) 3);
+  cinfo->actual_number_of_colors = 0; /* initialize map to empty */
+
+  /* Read first byte to determine file format */
+  switch (getc(infile)) {
+  case 'G':
+    read_gif_map(cinfo, infile);
+    break;
+  case 'P':
+    read_ppm_map(cinfo, infile);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+    break;
+  }
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */
diff --git a/libjpegtwrp/rdgif.c b/libjpegtwrp/rdgif.c
new file mode 100644
index 000000000..b27c1675d
--- /dev/null
+++ b/libjpegtwrp/rdgif.c
@@ -0,0 +1,38 @@
+/*
+ * rdgif.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to read input images in GIF format.
+ *
+ *****************************************************************************
+ * NOTE: to avoid entanglements with Unisys' patent on LZW compression,      *
+ * the ability to read GIF files has been removed from the IJG distribution. *
+ * Sorry about that.                                                         *
+ *****************************************************************************
+ *
+ * We are required to state that
+ *    "The Graphics Interchange Format(c) is the Copyright property of
+ *    CompuServe Incorporated. GIF(sm) is a Service Mark property of
+ *    CompuServe Incorporated."
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+
+#ifdef GIF_SUPPORTED
+
+/*
+ * The module selection routine for GIF format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_gif (j_compress_ptr cinfo)
+{
+  fprintf(stderr, "GIF input is unsupported for legal reasons.  Sorry.\n");
+  exit(EXIT_FAILURE);
+  return NULL;			/* keep compiler happy */
+}
+
+#endif /* GIF_SUPPORTED */
diff --git a/libjpegtwrp/rdjpgcom.1 b/libjpegtwrp/rdjpgcom.1
new file mode 100644
index 000000000..2bba04e1a
--- /dev/null
+++ b/libjpegtwrp/rdjpgcom.1
@@ -0,0 +1,54 @@
+.TH RDJPGCOM 1 "11 October 1997"
+.SH NAME
+rdjpgcom \- display text comments from a JPEG file
+.SH SYNOPSIS
+.B rdjpgcom
+[
+.B \-verbose
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B rdjpgcom
+reads the named JPEG/JFIF file, or the standard input if no file is named,
+and prints any text comments found in the file on the standard output.
+.PP
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+.SH OPTIONS
+.TP
+.B \-verbose
+Causes
+.B rdjpgcom
+to also display the JPEG image dimensions.
+.PP
+Switch names may be abbreviated, and are not case sensitive.
+.SH HINTS
+.B rdjpgcom
+does not depend on the IJG JPEG library.  Its source code is intended as an
+illustration of the minimum amount of code required to parse a JPEG file
+header correctly.
+.PP
+In
+.B \-verbose
+mode,
+.B rdjpgcom
+will also attempt to print the contents of any "APP12" markers as text.
+Some digital cameras produce APP12 markers containing useful textual
+information.  If you like, you can modify the source code to print
+other APPn marker types as well.
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR djpeg (1),
+.BR jpegtran (1),
+.BR wrjpgcom (1)
+.SH AUTHOR
+Independent JPEG Group
diff --git a/libjpegtwrp/rdjpgcom.c b/libjpegtwrp/rdjpgcom.c
new file mode 100644
index 000000000..ffe6fc621
--- /dev/null
+++ b/libjpegtwrp/rdjpgcom.c
@@ -0,0 +1,496 @@
+/*
+ * rdjpgcom.c
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a very simple stand-alone application that displays
+ * the text in COM (comment) markers in a JFIF file.
+ * This may be useful as an example of the minimum logic needed to parse
+ * JPEG markers.
+ */
+
+#define JPEG_CJPEG_DJPEG	/* to get the command-line config symbols */
+#include "jinclude.h"		/* get auto-config symbols, <stdio.h> */
+
+#include <ctype.h>		/* to declare isupper(), tolower() */
+#ifdef USE_SETMODE
+#include <fcntl.h>		/* to declare setmode()'s parameter macros */
+/* If you have setmode() but not <io.h>, just delete this line: */
+#include <io.h>			/* to declare setmode() */
+#endif
+
+#ifdef USE_CCOMMAND		/* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>		/* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>		/* Think declares it here */
+#endif
+#endif
+
+#ifdef DONT_USE_B_MODE		/* define mode parameters for fopen() */
+#define READ_BINARY	"r"
+#else
+#ifdef VMS			/* VMS is very nonstandard */
+#define READ_BINARY	"rb", "ctx=stm"
+#else				/* standard ANSI-compliant case */
+#define READ_BINARY	"rb"
+#endif
+#endif
+
+#ifndef EXIT_FAILURE		/* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+#ifndef EXIT_SUCCESS
+#ifdef VMS
+#define EXIT_SUCCESS  1		/* VMS is very nonstandard */
+#else
+#define EXIT_SUCCESS  0
+#endif
+#endif
+
+
+/*
+ * These macros are used to read the input file.
+ * To reuse this code in another application, you might need to change these.
+ */
+
+static FILE * infile;		/* input JPEG file */
+
+/* Return next input byte, or EOF if no more */
+#define NEXTBYTE()  getc(infile)
+
+
+/* Error exit handler */
+#define ERREXIT(msg)  (fprintf(stderr, "%s\n", msg), exit(EXIT_FAILURE))
+
+
+/* Read one byte, testing for EOF */
+static int
+read_1_byte (void)
+{
+  int c;
+
+  c = NEXTBYTE();
+  if (c == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return c;
+}
+
+/* Read 2 bytes, convert to unsigned int */
+/* All 2-byte quantities in JPEG markers are MSB first */
+static unsigned int
+read_2_bytes (void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  if (c1 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  c2 = NEXTBYTE();
+  if (c2 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return (((unsigned int) c1) << 8) + ((unsigned int) c2);
+}
+
+
+/*
+ * JPEG markers consist of one or more 0xFF bytes, followed by a marker
+ * code byte (which is not an FF).  Here are the marker codes of interest
+ * in this program.  (See jdmarker.c for a more complete list.)
+ */
+
+#define M_SOF0  0xC0		/* Start Of Frame N */
+#define M_SOF1  0xC1		/* N indicates which compression process */
+#define M_SOF2  0xC2		/* Only SOF0-SOF2 are now in common use */
+#define M_SOF3  0xC3
+#define M_SOF5  0xC5		/* NB: codes C4 and CC are NOT SOF markers */
+#define M_SOF6  0xC6
+#define M_SOF7  0xC7
+#define M_SOF9  0xC9
+#define M_SOF10 0xCA
+#define M_SOF11 0xCB
+#define M_SOF13 0xCD
+#define M_SOF14 0xCE
+#define M_SOF15 0xCF
+#define M_SOI   0xD8		/* Start Of Image (beginning of datastream) */
+#define M_EOI   0xD9		/* End Of Image (end of datastream) */
+#define M_SOS   0xDA		/* Start Of Scan (begins compressed data) */
+#define M_APP0	0xE0		/* Application-specific marker, type N */
+#define M_APP12	0xEC		/* (we don't bother to list all 16 APPn's) */
+#define M_COM   0xFE		/* COMment */
+
+
+/*
+ * Find the next JPEG marker and return its marker code.
+ * We expect at least one FF byte, possibly more if the compressor used FFs
+ * to pad the file.
+ * There could also be non-FF garbage between markers.  The treatment of such
+ * garbage is unspecified; we choose to skip over it but emit a warning msg.
+ * NB: this routine must not be used after seeing SOS marker, since it will
+ * not deal correctly with FF/00 sequences in the compressed image data...
+ */
+
+static int
+next_marker (void)
+{
+  int c;
+  int discarded_bytes = 0;
+
+  /* Find 0xFF byte; count and skip any non-FFs. */
+  c = read_1_byte();
+  while (c != 0xFF) {
+    discarded_bytes++;
+    c = read_1_byte();
+  }
+  /* Get marker code byte, swallowing any duplicate FF bytes.  Extra FFs
+   * are legal as pad bytes, so don't count them in discarded_bytes.
+   */
+  do {
+    c = read_1_byte();
+  } while (c == 0xFF);
+
+  if (discarded_bytes != 0) {
+    fprintf(stderr, "Warning: garbage data found in JPEG file\n");
+  }
+
+  return c;
+}
+
+
+/*
+ * Read the initial marker, which should be SOI.
+ * For a JFIF file, the first two bytes of the file should be literally
+ * 0xFF M_SOI.  To be more general, we could use next_marker, but if the
+ * input file weren't actually JPEG at all, next_marker might read the whole
+ * file and then return a misleading error message...
+ */
+
+static int
+first_marker (void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  c2 = NEXTBYTE();
+  if (c1 != 0xFF || c2 != M_SOI)
+    ERREXIT("Not a JPEG file");
+  return c2;
+}
+
+
+/*
+ * Most types of marker are followed by a variable-length parameter segment.
+ * This routine skips over the parameters for any marker we don't otherwise
+ * want to process.
+ * Note that we MUST skip the parameter segment explicitly in order not to
+ * be fooled by 0xFF bytes that might appear within the parameter segment;
+ * such bytes do NOT introduce new markers.
+ */
+
+static void
+skip_variable (void)
+/* Skip over an unknown or uninteresting variable-length marker */
+{
+  unsigned int length;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+  /* Skip over the remaining bytes */
+  while (length > 0) {
+    (void) read_1_byte();
+    length--;
+  }
+}
+
+
+/*
+ * Process a COM marker.
+ * We want to print out the marker contents as legible text;
+ * we must guard against non-text junk and varying newline representations.
+ */
+
+static void
+process_COM (void)
+{
+  unsigned int length;
+  int ch;
+  int lastch = 0;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+
+  while (length > 0) {
+    ch = read_1_byte();
+    /* Emit the character in a readable form.
+     * Nonprintables are converted to \nnn form,
+     * while \ is converted to \\.
+     * Newlines in CR, CR/LF, or LF form will be printed as one newline.
+     */
+    if (ch == '\r') {
+      printf("\n");
+    } else if (ch == '\n') {
+      if (lastch != '\r')
+	printf("\n");
+    } else if (ch == '\\') {
+      printf("\\\\");
+    } else if (isprint(ch)) {
+      putc(ch, stdout);
+    } else {
+      printf("\\%03o", ch);
+    }
+    lastch = ch;
+    length--;
+  }
+  printf("\n");
+}
+
+
+/*
+ * Process a SOFn marker.
+ * This code is only needed if you want to know the image dimensions...
+ */
+
+static void
+process_SOFn (int marker)
+{
+  unsigned int length;
+  unsigned int image_height, image_width;
+  int data_precision, num_components;
+  const char * process;
+  int ci;
+
+  length = read_2_bytes();	/* usual parameter length count */
+
+  data_precision = read_1_byte();
+  image_height = read_2_bytes();
+  image_width = read_2_bytes();
+  num_components = read_1_byte();
+
+  switch (marker) {
+  case M_SOF0:	process = "Baseline";  break;
+  case M_SOF1:	process = "Extended sequential";  break;
+  case M_SOF2:	process = "Progressive";  break;
+  case M_SOF3:	process = "Lossless";  break;
+  case M_SOF5:	process = "Differential sequential";  break;
+  case M_SOF6:	process = "Differential progressive";  break;
+  case M_SOF7:	process = "Differential lossless";  break;
+  case M_SOF9:	process = "Extended sequential, arithmetic coding";  break;
+  case M_SOF10:	process = "Progressive, arithmetic coding";  break;
+  case M_SOF11:	process = "Lossless, arithmetic coding";  break;
+  case M_SOF13:	process = "Differential sequential, arithmetic coding";  break;
+  case M_SOF14:	process = "Differential progressive, arithmetic coding"; break;
+  case M_SOF15:	process = "Differential lossless, arithmetic coding";  break;
+  default:	process = "Unknown";  break;
+  }
+
+  printf("JPEG image is %uw * %uh, %d color components, %d bits per sample\n",
+	 image_width, image_height, num_components, data_precision);
+  printf("JPEG process: %s\n", process);
+
+  if (length != (unsigned int) (8 + num_components * 3))
+    ERREXIT("Bogus SOF marker length");
+
+  for (ci = 0; ci < num_components; ci++) {
+    (void) read_1_byte();	/* Component ID code */
+    (void) read_1_byte();	/* H, V sampling factors */
+    (void) read_1_byte();	/* Quantization table number */
+  }
+}
+
+
+/*
+ * Parse the marker stream until SOS or EOI is seen;
+ * display any COM markers.
+ * While the companion program wrjpgcom will always insert COM markers before
+ * SOFn, other implementations might not, so we scan to SOS before stopping.
+ * If we were only interested in the image dimensions, we would stop at SOFn.
+ * (Conversely, if we only cared about COM markers, there would be no need
+ * for special code to handle SOFn; we could treat it like other markers.)
+ */
+
+static int
+scan_JPEG_header (int verbose)
+{
+  int marker;
+
+  /* Expect SOI at start of file */
+  if (first_marker() != M_SOI)
+    ERREXIT("Expected SOI marker first");
+
+  /* Scan miscellaneous markers until we reach SOS. */
+  for (;;) {
+    marker = next_marker();
+    switch (marker) {
+      /* Note that marker codes 0xC4, 0xC8, 0xCC are not, and must not be,
+       * treated as SOFn.  C4 in particular is actually DHT.
+       */
+    case M_SOF0:		/* Baseline */
+    case M_SOF1:		/* Extended sequential, Huffman */
+    case M_SOF2:		/* Progressive, Huffman */
+    case M_SOF3:		/* Lossless, Huffman */
+    case M_SOF5:		/* Differential sequential, Huffman */
+    case M_SOF6:		/* Differential progressive, Huffman */
+    case M_SOF7:		/* Differential lossless, Huffman */
+    case M_SOF9:		/* Extended sequential, arithmetic */
+    case M_SOF10:		/* Progressive, arithmetic */
+    case M_SOF11:		/* Lossless, arithmetic */
+    case M_SOF13:		/* Differential sequential, arithmetic */
+    case M_SOF14:		/* Differential progressive, arithmetic */
+    case M_SOF15:		/* Differential lossless, arithmetic */
+      if (verbose)
+	process_SOFn(marker);
+      else
+	skip_variable();
+      break;
+
+    case M_SOS:			/* stop before hitting compressed data */
+      return marker;
+
+    case M_EOI:			/* in case it's a tables-only JPEG stream */
+      return marker;
+
+    case M_COM:
+      process_COM();
+      break;
+
+    case M_APP12:
+      /* Some digital camera makers put useful textual information into
+       * APP12 markers, so we print those out too when in -verbose mode.
+       */
+      if (verbose) {
+	printf("APP12 contains:\n");
+	process_COM();
+      } else
+	skip_variable();
+      break;
+
+    default:			/* Anything else just gets skipped */
+      skip_variable();		/* we assume it has a parameter count... */
+      break;
+    }
+  } /* end loop */
+}
+
+
+/* Command line parsing code */
+
+static const char * progname;	/* program name for error messages */
+
+
+static void
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "rdjpgcom displays any textual comments in a JPEG file.\n");
+
+  fprintf(stderr, "Usage: %s [switches] [inputfile]\n", progname);
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -verbose    Also display dimensions of JPEG image\n");
+
+  exit(EXIT_FAILURE);
+}
+
+
+static int
+keymatch (char * arg, const char * keyword, int minchars)
+/* Case-insensitive matching of (possibly abbreviated) keyword switches. */
+/* keyword is the constant keyword (must be lower case already), */
+/* minchars is length of minimum legal abbreviation. */
+{
+  register int ca, ck;
+  register int nmatched = 0;
+
+  while ((ca = *arg++) != '\0') {
+    if ((ck = *keyword++) == '\0')
+      return 0;			/* arg longer than keyword, no good */
+    if (isupper(ca))		/* force arg to lcase (assume ck is already) */
+      ca = tolower(ca);
+    if (ca != ck)
+      return 0;			/* no good */
+    nmatched++;			/* count matched characters */
+  }
+  /* reached end of argument; fail if it's too short for unique abbrev */
+  if (nmatched < minchars)
+    return 0;
+  return 1;			/* A-OK */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  int argn;
+  char * arg;
+  int verbose = 0;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "rdjpgcom";	/* in case C library doesn't provide it */
+
+  /* Parse switches, if any */
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (arg[0] != '-')
+      break;			/* not switch, must be file name */
+    arg++;			/* advance over '-' */
+    if (keymatch(arg, "verbose", 1)) {
+      verbose++;
+    } else
+      usage();
+  }
+
+  /* Open the input file. */
+  /* Unix style: expect zero or one file name */
+  if (argn < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+  if (argn < argc) {
+    if ((infile = fopen(argv[argn], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[argn]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+#ifdef USE_SETMODE		/* need to hack file mode? */
+    setmode(fileno(stdin), O_BINARY);
+#endif
+#ifdef USE_FDOPEN		/* need to re-open in binary mode? */
+    if ((infile = fdopen(fileno(stdin), READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open stdin\n", progname);
+      exit(EXIT_FAILURE);
+    }
+#else
+    infile = stdin;
+#endif
+  }
+
+  /* Scan the JPEG headers. */
+  (void) scan_JPEG_header(verbose);
+
+  /* All done. */
+  exit(EXIT_SUCCESS);
+  return 0;			/* suppress no-return-value warnings */
+}
diff --git a/libjpegtwrp/rdppm.c b/libjpegtwrp/rdppm.c
new file mode 100644
index 000000000..1df35c1b3
--- /dev/null
+++ b/libjpegtwrp/rdppm.c
@@ -0,0 +1,458 @@
+/*
+ * rdppm.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to read input images in PPM/PGM format.
+ * The extended 2-byte-per-sample raw PPM/PGM formats are supported.
+ * The PBMPLUS library is NOT required to compile this software
+ * (but it is highly useful as a set of PPM image manipulation programs).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed PPM format).
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+
+#ifdef PPM_SUPPORTED
+
+
+/* Portions of this code are based on the PBMPLUS library, which is:
+**
+** Copyright (C) 1988 by Jef Poskanzer.
+**
+** Permission to use, copy, modify, and distribute this software and its
+** documentation for any purpose and without fee is hereby granted, provided
+** that the above copyright notice appear in all copies and that both that
+** copyright notice and this permission notice appear in supporting
+** documentation.  This software is provided "as is" without express or
+** implied warranty.
+*/
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+#ifdef HAVE_UNSIGNED_CHAR
+typedef unsigned char U_CHAR;
+#define UCH(x)	((int) (x))
+#else /* !HAVE_UNSIGNED_CHAR */
+#ifdef CHAR_IS_UNSIGNED
+typedef char U_CHAR;
+#define UCH(x)	((int) (x))
+#else
+typedef char U_CHAR;
+#define UCH(x)	((int) (x) & 0xFF)
+#endif
+#endif /* HAVE_UNSIGNED_CHAR */
+
+
+#define	ReadOK(file,buffer,len)	(JFREAD(file,buffer,len) == ((size_t) (len)))
+
+
+/*
+ * On most systems, reading individual bytes with getc() is drastically less
+ * efficient than buffering a row at a time with fread().  On PCs, we must
+ * allocate the buffer in near data space, because we are assuming small-data
+ * memory model, wherein fread() can't reach far memory.  If you need to
+ * process very wide images on a PC, you might have to compile in large-memory
+ * model, or else replace fread() with a getc() loop --- which will be much
+ * slower.
+ */
+
+
+/* Private version of data source object */
+
+typedef struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  U_CHAR *iobuffer;		/* non-FAR pointer to I/O buffer */
+  JSAMPROW pixrow;		/* FAR pointer to same */
+  size_t buffer_width;		/* width of I/O buffer */
+  JSAMPLE *rescale;		/* => maxval-remapping array, or NULL */
+} ppm_source_struct;
+
+typedef ppm_source_struct * ppm_source_ptr;
+
+
+LOCAL(int)
+pbm_getc (FILE * infile)
+/* Read next char, skipping over any comments */
+/* A comment/newline sequence is returned as a newline */
+{
+  register int ch;
+
+  ch = getc(infile);
+  if (ch == '#') {
+    do {
+      ch = getc(infile);
+    } while (ch != '\n' && ch != EOF);
+  }
+  return ch;
+}
+
+
+LOCAL(unsigned int)
+read_pbm_integer (j_compress_ptr cinfo, FILE * infile)
+/* Read an unsigned decimal integer from the PPM file */
+/* Swallows one trailing character after the integer */
+/* Note that on a 16-bit-int machine, only values up to 64k can be read. */
+/* This should not be a problem in practice. */
+{
+  register int ch;
+  register unsigned int val;
+
+  /* Skip any leading whitespace */
+  do {
+    ch = pbm_getc(infile);
+    if (ch == EOF)
+      ERREXIT(cinfo, JERR_INPUT_EOF);
+  } while (ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r');
+
+  if (ch < '0' || ch > '9')
+    ERREXIT(cinfo, JERR_PPM_NONNUMERIC);
+
+  val = ch - '0';
+  while ((ch = pbm_getc(infile)) >= '0' && ch <= '9') {
+    val *= 10;
+    val += ch - '0';
+  }
+  return val;
+}
+
+
+/*
+ * Read one row of pixels.
+ *
+ * We provide several different versions depending on input file format.
+ * In all cases, input is scaled to the size of JSAMPLE.
+ *
+ * A really fast path is provided for reading byte/sample raw files with
+ * maxval = MAXJSAMPLE, which is the normal case for 8-bit data.
+ */
+
+
+METHODDEF(JDIMENSION)
+get_text_gray_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading text-format PGM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  FILE * infile = source->pub.input_file;
+  register JSAMPROW ptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescale[read_pbm_integer(cinfo, infile)];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_text_rgb_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading text-format PPM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  FILE * infile = source->pub.input_file;
+  register JSAMPROW ptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescale[read_pbm_integer(cinfo, infile)];
+    *ptr++ = rescale[read_pbm_integer(cinfo, infile)];
+    *ptr++ = rescale[read_pbm_integer(cinfo, infile)];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_scaled_gray_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format PGM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR * bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer[0];
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescale[UCH(*bufferptr++)];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_scaled_rgb_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format PPM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR * bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer[0];
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescale[UCH(*bufferptr++)];
+    *ptr++ = rescale[UCH(*bufferptr++)];
+    *ptr++ = rescale[UCH(*bufferptr++)];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_raw_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format files with maxval = MAXJSAMPLE.
+ * In this case we just read right into the JSAMPLE buffer!
+ * Note that same code works for PPM and PGM files.
+ */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_word_gray_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-word-format PGM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR * bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer[0];
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    register int temp;
+    temp  = UCH(*bufferptr++);
+    temp |= UCH(*bufferptr++) << 8;
+    *ptr++ = rescale[temp];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_word_rgb_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-word-format PPM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR * bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer[0];
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    register int temp;
+    temp  = UCH(*bufferptr++);
+    temp |= UCH(*bufferptr++) << 8;
+    *ptr++ = rescale[temp];
+    temp  = UCH(*bufferptr++);
+    temp |= UCH(*bufferptr++) << 8;
+    *ptr++ = rescale[temp];
+    temp  = UCH(*bufferptr++);
+    temp |= UCH(*bufferptr++) << 8;
+    *ptr++ = rescale[temp];
+  }
+  return 1;
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_ppm (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  int c;
+  unsigned int w, h, maxval;
+  boolean need_iobuffer, use_raw_buffer, need_rescale;
+
+  if (getc(source->pub.input_file) != 'P')
+    ERREXIT(cinfo, JERR_PPM_NOT);
+
+  c = getc(source->pub.input_file); /* subformat discriminator character */
+
+  /* detect unsupported variants (ie, PBM) before trying to read header */
+  switch (c) {
+  case '2':			/* it's a text-format PGM file */
+  case '3':			/* it's a text-format PPM file */
+  case '5':			/* it's a raw-format PGM file */
+  case '6':			/* it's a raw-format PPM file */
+    break;
+  default:
+    ERREXIT(cinfo, JERR_PPM_NOT);
+    break;
+  }
+
+  /* fetch the remaining header info */
+  w = read_pbm_integer(cinfo, source->pub.input_file);
+  h = read_pbm_integer(cinfo, source->pub.input_file);
+  maxval = read_pbm_integer(cinfo, source->pub.input_file);
+
+  if (w <= 0 || h <= 0 || maxval <= 0) /* error check */
+    ERREXIT(cinfo, JERR_PPM_NOT);
+
+  cinfo->data_precision = BITS_IN_JSAMPLE; /* we always rescale data to this */
+  cinfo->image_width = (JDIMENSION) w;
+  cinfo->image_height = (JDIMENSION) h;
+
+  /* initialize flags to most common settings */
+  need_iobuffer = TRUE;		/* do we need an I/O buffer? */
+  use_raw_buffer = FALSE;	/* do we map input buffer onto I/O buffer? */
+  need_rescale = TRUE;		/* do we need a rescale array? */
+
+  switch (c) {
+  case '2':			/* it's a text-format PGM file */
+    cinfo->input_components = 1;
+    cinfo->in_color_space = JCS_GRAYSCALE;
+    TRACEMS2(cinfo, 1, JTRC_PGM_TEXT, w, h);
+    source->pub.get_pixel_rows = get_text_gray_row;
+    need_iobuffer = FALSE;
+    break;
+
+  case '3':			/* it's a text-format PPM file */
+    cinfo->input_components = 3;
+    cinfo->in_color_space = JCS_RGB;
+    TRACEMS2(cinfo, 1, JTRC_PPM_TEXT, w, h);
+    source->pub.get_pixel_rows = get_text_rgb_row;
+    need_iobuffer = FALSE;
+    break;
+
+  case '5':			/* it's a raw-format PGM file */
+    cinfo->input_components = 1;
+    cinfo->in_color_space = JCS_GRAYSCALE;
+    TRACEMS2(cinfo, 1, JTRC_PGM, w, h);
+    if (maxval > 255) {
+      source->pub.get_pixel_rows = get_word_gray_row;
+    } else if (maxval == MAXJSAMPLE && SIZEOF(JSAMPLE) == SIZEOF(U_CHAR)) {
+      source->pub.get_pixel_rows = get_raw_row;
+      use_raw_buffer = TRUE;
+      need_rescale = FALSE;
+    } else {
+      source->pub.get_pixel_rows = get_scaled_gray_row;
+    }
+    break;
+
+  case '6':			/* it's a raw-format PPM file */
+    cinfo->input_components = 3;
+    cinfo->in_color_space = JCS_RGB;
+    TRACEMS2(cinfo, 1, JTRC_PPM, w, h);
+    if (maxval > 255) {
+      source->pub.get_pixel_rows = get_word_rgb_row;
+    } else if (maxval == MAXJSAMPLE && SIZEOF(JSAMPLE) == SIZEOF(U_CHAR)) {
+      source->pub.get_pixel_rows = get_raw_row;
+      use_raw_buffer = TRUE;
+      need_rescale = FALSE;
+    } else {
+      source->pub.get_pixel_rows = get_scaled_rgb_row;
+    }
+    break;
+  }
+
+  /* Allocate space for I/O buffer: 1 or 3 bytes or words/pixel. */
+  if (need_iobuffer) {
+    source->buffer_width = (size_t) w * cinfo->input_components *
+      ((maxval<=255) ? SIZEOF(U_CHAR) : (2*SIZEOF(U_CHAR)));
+    source->iobuffer = (U_CHAR *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  source->buffer_width);
+  }
+
+  /* Create compressor input buffer. */
+  if (use_raw_buffer) {
+    /* For unscaled raw-input case, we can just map it onto the I/O buffer. */
+    /* Synthesize a JSAMPARRAY pointer structure */
+    /* Cast here implies near->far pointer conversion on PCs */
+    source->pixrow = (JSAMPROW) source->iobuffer;
+    source->pub.buffer = & source->pixrow;
+    source->pub.buffer_height = 1;
+  } else {
+    /* Need to translate anyway, so make a separate sample buffer. */
+    source->pub.buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) w * cinfo->input_components, (JDIMENSION) 1);
+    source->pub.buffer_height = 1;
+  }
+
+  /* Compute the rescaling array if required. */
+  if (need_rescale) {
+    INT32 val, half_maxval;
+
+    /* On 16-bit-int machines we have to be careful of maxval = 65535 */
+    source->rescale = (JSAMPLE *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  (size_t) (((long) maxval + 1L) * SIZEOF(JSAMPLE)));
+    half_maxval = maxval / 2;
+    for (val = 0; val <= (INT32) maxval; val++) {
+      /* The multiplication here must be done in 32 bits to avoid overflow */
+      source->rescale[val] = (JSAMPLE) ((val*MAXJSAMPLE + half_maxval)/maxval);
+    }
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_ppm (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for PPM format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_ppm (j_compress_ptr cinfo)
+{
+  ppm_source_ptr source;
+
+  /* Create module interface object */
+  source = (ppm_source_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(ppm_source_struct));
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_ppm;
+  source->pub.finish_input = finish_input_ppm;
+
+  return (cjpeg_source_ptr) source;
+}
+
+#endif /* PPM_SUPPORTED */
diff --git a/libjpegtwrp/rdrle.c b/libjpegtwrp/rdrle.c
new file mode 100644
index 000000000..542bc3749
--- /dev/null
+++ b/libjpegtwrp/rdrle.c
@@ -0,0 +1,387 @@
+/*
+ * rdrle.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to read input images in Utah RLE format.
+ * The Utah Raster Toolkit library is required (version 3.1 or later).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed RLE format).
+ *
+ * Based on code contributed by Mike Lijewski,
+ * with updates from Robert Hutchinson.
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+
+#ifdef RLE_SUPPORTED
+
+/* rle.h is provided by the Utah Raster Toolkit. */
+
+#include <rle.h>
+
+/*
+ * We assume that JSAMPLE has the same representation as rle_pixel,
+ * to wit, "unsigned char".  Hence we can't cope with 12- or 16-bit samples.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+/*
+ * We support the following types of RLE files:
+ *   
+ *   GRAYSCALE   - 8 bits, no colormap
+ *   MAPPEDGRAY  - 8 bits, 1 channel colomap
+ *   PSEUDOCOLOR - 8 bits, 3 channel colormap
+ *   TRUECOLOR   - 24 bits, 3 channel colormap
+ *   DIRECTCOLOR - 24 bits, no colormap
+ *
+ * For now, we ignore any alpha channel in the image.
+ */
+
+typedef enum
+  { GRAYSCALE, MAPPEDGRAY, PSEUDOCOLOR, TRUECOLOR, DIRECTCOLOR } rle_kind;
+
+
+/*
+ * Since RLE stores scanlines bottom-to-top, we have to invert the image
+ * to conform to JPEG's top-to-bottom order.  To do this, we read the
+ * incoming image into a virtual array on the first get_pixel_rows call,
+ * then fetch the required row from the virtual array on subsequent calls.
+ */
+
+typedef struct _rle_source_struct * rle_source_ptr;
+
+typedef struct _rle_source_struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  rle_kind visual;              /* actual type of input file */
+  jvirt_sarray_ptr image;       /* virtual array to hold the image */
+  JDIMENSION row;		/* current row # in the virtual array */
+  rle_hdr header;               /* Input file information */
+  rle_pixel** rle_row;          /* holds a row returned by rle_getrow() */
+
+} rle_source_struct;
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_rle (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  rle_source_ptr source = (rle_source_ptr) sinfo;
+  JDIMENSION width, height;
+#ifdef PROGRESS_REPORT
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+#endif
+
+  /* Use RLE library routine to get the header info */
+  source->header = *rle_hdr_init(NULL);
+  source->header.rle_file = source->pub.input_file;
+  switch (rle_get_setup(&(source->header))) {
+  case RLE_SUCCESS:
+    /* A-OK */
+    break;
+  case RLE_NOT_RLE:
+    ERREXIT(cinfo, JERR_RLE_NOT);
+    break;
+  case RLE_NO_SPACE:
+    ERREXIT(cinfo, JERR_RLE_MEM);
+    break;
+  case RLE_EMPTY:
+    ERREXIT(cinfo, JERR_RLE_EMPTY);
+    break;
+  case RLE_EOF:
+    ERREXIT(cinfo, JERR_RLE_EOF);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_RLE_BADERROR);
+    break;
+  }
+
+  /* Figure out what we have, set private vars and return values accordingly */
+  
+  width  = source->header.xmax - source->header.xmin + 1;
+  height = source->header.ymax - source->header.ymin + 1;
+  source->header.xmin = 0;		/* realign horizontally */
+  source->header.xmax = width-1;
+
+  cinfo->image_width      = width;
+  cinfo->image_height     = height;
+  cinfo->data_precision   = 8;  /* we can only handle 8 bit data */
+
+  if (source->header.ncolors == 1 && source->header.ncmap == 0) {
+    source->visual     = GRAYSCALE;
+    TRACEMS2(cinfo, 1, JTRC_RLE_GRAY, width, height);
+  } else if (source->header.ncolors == 1 && source->header.ncmap == 1) {
+    source->visual     = MAPPEDGRAY;
+    TRACEMS3(cinfo, 1, JTRC_RLE_MAPGRAY, width, height,
+             1 << source->header.cmaplen);
+  } else if (source->header.ncolors == 1 && source->header.ncmap == 3) {
+    source->visual     = PSEUDOCOLOR;
+    TRACEMS3(cinfo, 1, JTRC_RLE_MAPPED, width, height,
+	     1 << source->header.cmaplen);
+  } else if (source->header.ncolors == 3 && source->header.ncmap == 3) {
+    source->visual     = TRUECOLOR;
+    TRACEMS3(cinfo, 1, JTRC_RLE_FULLMAP, width, height,
+	     1 << source->header.cmaplen);
+  } else if (source->header.ncolors == 3 && source->header.ncmap == 0) {
+    source->visual     = DIRECTCOLOR;
+    TRACEMS2(cinfo, 1, JTRC_RLE, width, height);
+  } else
+    ERREXIT(cinfo, JERR_RLE_UNSUPPORTED);
+  
+  if (source->visual == GRAYSCALE || source->visual == MAPPEDGRAY) {
+    cinfo->in_color_space   = JCS_GRAYSCALE;
+    cinfo->input_components = 1;
+  } else {
+    cinfo->in_color_space   = JCS_RGB;
+    cinfo->input_components = 3;
+  }
+
+  /*
+   * A place to hold each scanline while it's converted.
+   * (GRAYSCALE scanlines don't need converting)
+   */
+  if (source->visual != GRAYSCALE) {
+    source->rle_row = (rle_pixel**) (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) width, (JDIMENSION) cinfo->input_components);
+  }
+
+  /* request a virtual array to hold the image */
+  source->image = (*cinfo->mem->request_virt_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+     (JDIMENSION) (width * source->header.ncolors),
+     (JDIMENSION) height, (JDIMENSION) 1);
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL) {
+    /* count file input as separate pass */
+    progress->total_extra_passes++;
+  }
+#endif
+
+  source->pub.buffer_height = 1;
+}
+
+
+/*
+ * Read one row of pixels.
+ * Called only after load_image has read the image into the virtual array.
+ * Used for GRAYSCALE, MAPPEDGRAY, TRUECOLOR, and DIRECTCOLOR images.
+ */
+
+METHODDEF(JDIMENSION)
+get_rle_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  rle_source_ptr source = (rle_source_ptr) sinfo;
+
+  source->row--;
+  source->pub.buffer = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->image, source->row, (JDIMENSION) 1, FALSE);
+
+  return 1;
+}
+
+/*
+ * Read one row of pixels.
+ * Called only after load_image has read the image into the virtual array.
+ * Used for PSEUDOCOLOR images.
+ */
+
+METHODDEF(JDIMENSION)
+get_pseudocolor_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  rle_source_ptr source = (rle_source_ptr) sinfo;
+  JSAMPROW src_row, dest_row;
+  JDIMENSION col;
+  rle_map *colormap;
+  int val;
+
+  colormap = source->header.cmap;
+  dest_row = source->pub.buffer[0];
+  source->row--;
+  src_row = * (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->image, source->row, (JDIMENSION) 1, FALSE);
+
+  for (col = cinfo->image_width; col > 0; col--) {
+    val = GETJSAMPLE(*src_row++);
+    *dest_row++ = (JSAMPLE) (colormap[val      ] >> 8);
+    *dest_row++ = (JSAMPLE) (colormap[val + 256] >> 8);
+    *dest_row++ = (JSAMPLE) (colormap[val + 512] >> 8);
+  }
+
+  return 1;
+}
+
+
+/*
+ * Load the image into a virtual array.  We have to do this because RLE
+ * files start at the lower left while the JPEG standard has them starting
+ * in the upper left.  This is called the first time we want to get a row
+ * of input.  What we do is load the RLE data into the array and then call
+ * the appropriate routine to read one row from the array.  Before returning,
+ * we set source->pub.get_pixel_rows so that subsequent calls go straight to
+ * the appropriate row-reading routine.
+ */
+
+METHODDEF(JDIMENSION)
+load_image (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  rle_source_ptr source = (rle_source_ptr) sinfo;
+  JDIMENSION row, col;
+  JSAMPROW  scanline, red_ptr, green_ptr, blue_ptr;
+  rle_pixel **rle_row;
+  rle_map *colormap;
+  char channel;
+#ifdef PROGRESS_REPORT
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+#endif
+
+  colormap = source->header.cmap;
+  rle_row = source->rle_row;
+
+  /* Read the RLE data into our virtual array.
+   * We assume here that (a) rle_pixel is represented the same as JSAMPLE,
+   * and (b) we are not on a machine where FAR pointers differ from regular.
+   */
+  RLE_CLR_BIT(source->header, RLE_ALPHA); /* don't read the alpha channel */
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL) {
+    progress->pub.pass_limit = cinfo->image_height;
+    progress->pub.pass_counter = 0;
+    (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+  }
+#endif
+
+  switch (source->visual) {
+
+  case GRAYSCALE:
+  case PSEUDOCOLOR:
+    for (row = 0; row < cinfo->image_height; row++) {
+      rle_row = (rle_pixel **) (*cinfo->mem->access_virt_sarray)
+         ((j_common_ptr) cinfo, source->image, row, (JDIMENSION) 1, TRUE);
+      rle_getrow(&source->header, rle_row);
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+    break;
+
+  case MAPPEDGRAY:
+  case TRUECOLOR:
+    for (row = 0; row < cinfo->image_height; row++) {
+      scanline = * (*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr) cinfo, source->image, row, (JDIMENSION) 1, TRUE);
+      rle_row = source->rle_row;
+      rle_getrow(&source->header, rle_row);
+
+      for (col = 0; col < cinfo->image_width; col++) {
+        for (channel = 0; channel < source->header.ncolors; channel++) {
+          *scanline++ = (JSAMPLE)
+            (colormap[GETJSAMPLE(rle_row[channel][col]) + 256 * channel] >> 8);
+        }
+      }
+
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+    break;
+
+  case DIRECTCOLOR:
+    for (row = 0; row < cinfo->image_height; row++) {
+      scanline = * (*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr) cinfo, source->image, row, (JDIMENSION) 1, TRUE);
+      rle_getrow(&source->header, rle_row);
+
+      red_ptr   = rle_row[0];
+      green_ptr = rle_row[1];
+      blue_ptr  = rle_row[2];
+
+      for (col = cinfo->image_width; col > 0; col--) {
+        *scanline++ = *red_ptr++;
+        *scanline++ = *green_ptr++;
+        *scanline++ = *blue_ptr++;
+      }
+
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+  }
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+#endif
+
+  /* Set up to call proper row-extraction routine in future */
+  if (source->visual == PSEUDOCOLOR) {
+    source->pub.buffer = source->rle_row;
+    source->pub.get_pixel_rows = get_pseudocolor_row;
+  } else {
+    source->pub.get_pixel_rows = get_rle_row;
+  }
+  source->row = cinfo->image_height;
+
+  /* And fetch the topmost (bottommost) row */
+  return (*source->pub.get_pixel_rows) (cinfo, sinfo);   
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_rle (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for RLE format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_rle (j_compress_ptr cinfo)
+{
+  rle_source_ptr source;
+
+  /* Create module interface object */
+  source = (rle_source_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  SIZEOF(rle_source_struct));
+  /* Fill in method ptrs */
+  source->pub.start_input = start_input_rle;
+  source->pub.finish_input = finish_input_rle;
+  source->pub.get_pixel_rows = load_image;
+
+  return (cjpeg_source_ptr) source;
+}
+
+#endif /* RLE_SUPPORTED */
diff --git a/libjpegtwrp/rdswitch.c b/libjpegtwrp/rdswitch.c
new file mode 100644
index 000000000..4f4bb4f58
--- /dev/null
+++ b/libjpegtwrp/rdswitch.c
@@ -0,0 +1,332 @@
+/*
+ * rdswitch.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to process some of cjpeg's more complicated
+ * command-line switches.  Switches processed here are:
+ *	-qtables file		Read quantization tables from text file
+ *	-scans file		Read scan script from text file
+ *	-qslots N[,N,...]	Set component quantization table selectors
+ *	-sample HxV[,HxV,...]	Set component sampling factors
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+#include <ctype.h>		/* to declare isdigit(), isspace() */
+
+
+LOCAL(int)
+text_getc (FILE * file)
+/* Read next char, skipping over any comments (# to end of line) */
+/* A comment/newline sequence is returned as a newline */
+{
+  register int ch;
+  
+  ch = getc(file);
+  if (ch == '#') {
+    do {
+      ch = getc(file);
+    } while (ch != '\n' && ch != EOF);
+  }
+  return ch;
+}
+
+
+LOCAL(boolean)
+read_text_integer (FILE * file, long * result, int * termchar)
+/* Read an unsigned decimal integer from a file, store it in result */
+/* Reads one trailing character after the integer; returns it in termchar */
+{
+  register int ch;
+  register long val;
+  
+  /* Skip any leading whitespace, detect EOF */
+  do {
+    ch = text_getc(file);
+    if (ch == EOF) {
+      *termchar = ch;
+      return FALSE;
+    }
+  } while (isspace(ch));
+  
+  if (! isdigit(ch)) {
+    *termchar = ch;
+    return FALSE;
+  }
+
+  val = ch - '0';
+  while ((ch = text_getc(file)) != EOF) {
+    if (! isdigit(ch))
+      break;
+    val *= 10;
+    val += ch - '0';
+  }
+  *result = val;
+  *termchar = ch;
+  return TRUE;
+}
+
+
+GLOBAL(boolean)
+read_quant_tables (j_compress_ptr cinfo, char * filename,
+		   int scale_factor, boolean force_baseline)
+/* Read a set of quantization tables from the specified file.
+ * The file is plain ASCII text: decimal numbers with whitespace between.
+ * Comments preceded by '#' may be included in the file.
+ * There may be one to NUM_QUANT_TBLS tables in the file, each of 64 values.
+ * The tables are implicitly numbered 0,1,etc.
+ * NOTE: does not affect the qslots mapping, which will default to selecting
+ * table 0 for luminance (or primary) components, 1 for chrominance components.
+ * You must use -qslots if you want a different component->table mapping.
+ */
+{
+  FILE * fp;
+  int tblno, i, termchar;
+  long val;
+  unsigned int table[DCTSIZE2];
+
+  if ((fp = fopen(filename, "r")) == NULL) {
+    fprintf(stderr, "Can't open table file %s\n", filename);
+    return FALSE;
+  }
+  tblno = 0;
+
+  while (read_text_integer(fp, &val, &termchar)) { /* read 1st element of table */
+    if (tblno >= NUM_QUANT_TBLS) {
+      fprintf(stderr, "Too many tables in file %s\n", filename);
+      fclose(fp);
+      return FALSE;
+    }
+    table[0] = (unsigned int) val;
+    for (i = 1; i < DCTSIZE2; i++) {
+      if (! read_text_integer(fp, &val, &termchar)) {
+	fprintf(stderr, "Invalid table data in file %s\n", filename);
+	fclose(fp);
+	return FALSE;
+      }
+      table[i] = (unsigned int) val;
+    }
+    jpeg_add_quant_table(cinfo, tblno, table, scale_factor, force_baseline);
+    tblno++;
+  }
+
+  if (termchar != EOF) {
+    fprintf(stderr, "Non-numeric data in file %s\n", filename);
+    fclose(fp);
+    return FALSE;
+  }
+
+  fclose(fp);
+  return TRUE;
+}
+
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+
+LOCAL(boolean)
+read_scan_integer (FILE * file, long * result, int * termchar)
+/* Variant of read_text_integer that always looks for a non-space termchar;
+ * this simplifies parsing of punctuation in scan scripts.
+ */
+{
+  register int ch;
+
+  if (! read_text_integer(file, result, termchar))
+    return FALSE;
+  ch = *termchar;
+  while (ch != EOF && isspace(ch))
+    ch = text_getc(file);
+  if (isdigit(ch)) {		/* oops, put it back */
+    if (ungetc(ch, file) == EOF)
+      return FALSE;
+    ch = ' ';
+  } else {
+    /* Any separators other than ';' and ':' are ignored;
+     * this allows user to insert commas, etc, if desired.
+     */
+    if (ch != EOF && ch != ';' && ch != ':')
+      ch = ' ';
+  }
+  *termchar = ch;
+  return TRUE;
+}
+
+
+GLOBAL(boolean)
+read_scan_script (j_compress_ptr cinfo, char * filename)
+/* Read a scan script from the specified text file.
+ * Each entry in the file defines one scan to be emitted.
+ * Entries are separated by semicolons ';'.
+ * An entry contains one to four component indexes,
+ * optionally followed by a colon ':' and four progressive-JPEG parameters.
+ * The component indexes denote which component(s) are to be transmitted
+ * in the current scan.  The first component has index 0.
+ * Sequential JPEG is used if the progressive-JPEG parameters are omitted.
+ * The file is free format text: any whitespace may appear between numbers
+ * and the ':' and ';' punctuation marks.  Also, other punctuation (such
+ * as commas or dashes) can be placed between numbers if desired.
+ * Comments preceded by '#' may be included in the file.
+ * Note: we do very little validity checking here;
+ * jcmaster.c will validate the script parameters.
+ */
+{
+  FILE * fp;
+  int scanno, ncomps, termchar;
+  long val;
+  jpeg_scan_info * scanptr;
+#define MAX_SCANS  100		/* quite arbitrary limit */
+  jpeg_scan_info scans[MAX_SCANS];
+
+  if ((fp = fopen(filename, "r")) == NULL) {
+    fprintf(stderr, "Can't open scan definition file %s\n", filename);
+    return FALSE;
+  }
+  scanptr = scans;
+  scanno = 0;
+
+  while (read_scan_integer(fp, &val, &termchar)) {
+    if (scanno >= MAX_SCANS) {
+      fprintf(stderr, "Too many scans defined in file %s\n", filename);
+      fclose(fp);
+      return FALSE;
+    }
+    scanptr->component_index[0] = (int) val;
+    ncomps = 1;
+    while (termchar == ' ') {
+      if (ncomps >= MAX_COMPS_IN_SCAN) {
+	fprintf(stderr, "Too many components in one scan in file %s\n",
+		filename);
+	fclose(fp);
+	return FALSE;
+      }
+      if (! read_scan_integer(fp, &val, &termchar))
+	goto bogus;
+      scanptr->component_index[ncomps] = (int) val;
+      ncomps++;
+    }
+    scanptr->comps_in_scan = ncomps;
+    if (termchar == ':') {
+      if (! read_scan_integer(fp, &val, &termchar) || termchar != ' ')
+	goto bogus;
+      scanptr->Ss = (int) val;
+      if (! read_scan_integer(fp, &val, &termchar) || termchar != ' ')
+	goto bogus;
+      scanptr->Se = (int) val;
+      if (! read_scan_integer(fp, &val, &termchar) || termchar != ' ')
+	goto bogus;
+      scanptr->Ah = (int) val;
+      if (! read_scan_integer(fp, &val, &termchar))
+	goto bogus;
+      scanptr->Al = (int) val;
+    } else {
+      /* set non-progressive parameters */
+      scanptr->Ss = 0;
+      scanptr->Se = DCTSIZE2-1;
+      scanptr->Ah = 0;
+      scanptr->Al = 0;
+    }
+    if (termchar != ';' && termchar != EOF) {
+bogus:
+      fprintf(stderr, "Invalid scan entry format in file %s\n", filename);
+      fclose(fp);
+      return FALSE;
+    }
+    scanptr++, scanno++;
+  }
+
+  if (termchar != EOF) {
+    fprintf(stderr, "Non-numeric data in file %s\n", filename);
+    fclose(fp);
+    return FALSE;
+  }
+
+  if (scanno > 0) {
+    /* Stash completed scan list in cinfo structure.
+     * NOTE: for cjpeg's use, JPOOL_IMAGE is the right lifetime for this data,
+     * but if you want to compress multiple images you'd want JPOOL_PERMANENT.
+     */
+    scanptr = (jpeg_scan_info *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  scanno * SIZEOF(jpeg_scan_info));
+    MEMCOPY(scanptr, scans, scanno * SIZEOF(jpeg_scan_info));
+    cinfo->scan_info = scanptr;
+    cinfo->num_scans = scanno;
+  }
+
+  fclose(fp);
+  return TRUE;
+}
+
+#endif /* C_MULTISCAN_FILES_SUPPORTED */
+
+
+GLOBAL(boolean)
+set_quant_slots (j_compress_ptr cinfo, char *arg)
+/* Process a quantization-table-selectors parameter string, of the form
+ *     N[,N,...]
+ * If there are more components than parameters, the last value is replicated.
+ */
+{
+  int val = 0;			/* default table # */
+  int ci;
+  char ch;
+
+  for (ci = 0; ci < MAX_COMPONENTS; ci++) {
+    if (*arg) {
+      ch = ',';			/* if not set by sscanf, will be ',' */
+      if (sscanf(arg, "%d%c", &val, &ch) < 1)
+	return FALSE;
+      if (ch != ',')		/* syntax check */
+	return FALSE;
+      if (val < 0 || val >= NUM_QUANT_TBLS) {
+	fprintf(stderr, "JPEG quantization tables are numbered 0..%d\n",
+		NUM_QUANT_TBLS-1);
+	return FALSE;
+      }
+      cinfo->comp_info[ci].quant_tbl_no = val;
+      while (*arg && *arg++ != ',') /* advance to next segment of arg string */
+	;
+    } else {
+      /* reached end of parameter, set remaining components to last table */
+      cinfo->comp_info[ci].quant_tbl_no = val;
+    }
+  }
+  return TRUE;
+}
+
+
+GLOBAL(boolean)
+set_sample_factors (j_compress_ptr cinfo, char *arg)
+/* Process a sample-factors parameter string, of the form
+ *     HxV[,HxV,...]
+ * If there are more components than parameters, "1x1" is assumed for the rest.
+ */
+{
+  int ci, val1, val2;
+  char ch1, ch2;
+
+  for (ci = 0; ci < MAX_COMPONENTS; ci++) {
+    if (*arg) {
+      ch2 = ',';		/* if not set by sscanf, will be ',' */
+      if (sscanf(arg, "%d%c%d%c", &val1, &ch1, &val2, &ch2) < 3)
+	return FALSE;
+      if ((ch1 != 'x' && ch1 != 'X') || ch2 != ',') /* syntax check */
+	return FALSE;
+      if (val1 <= 0 || val1 > 4 || val2 <= 0 || val2 > 4) {
+	fprintf(stderr, "JPEG sampling factors must be 1..4\n");
+	return FALSE;
+      }
+      cinfo->comp_info[ci].h_samp_factor = val1;
+      cinfo->comp_info[ci].v_samp_factor = val2;
+      while (*arg && *arg++ != ',') /* advance to next segment of arg string */
+	;
+    } else {
+      /* reached end of parameter, set remaining components to 1x1 sampling */
+      cinfo->comp_info[ci].h_samp_factor = 1;
+      cinfo->comp_info[ci].v_samp_factor = 1;
+    }
+  }
+  return TRUE;
+}
diff --git a/libjpegtwrp/rdtarga.c b/libjpegtwrp/rdtarga.c
new file mode 100644
index 000000000..4c2cd2673
--- /dev/null
+++ b/libjpegtwrp/rdtarga.c
@@ -0,0 +1,500 @@
+/*
+ * rdtarga.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to read input images in Targa format.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed Targa format).
+ *
+ * Based on code contributed by Lee Daniel Crocker.
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+
+#ifdef TARGA_SUPPORTED
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+#ifdef HAVE_UNSIGNED_CHAR
+typedef unsigned char U_CHAR;
+#define UCH(x)	((int) (x))
+#else /* !HAVE_UNSIGNED_CHAR */
+#ifdef CHAR_IS_UNSIGNED
+typedef char U_CHAR;
+#define UCH(x)	((int) (x))
+#else
+typedef char U_CHAR;
+#define UCH(x)	((int) (x) & 0xFF)
+#endif
+#endif /* HAVE_UNSIGNED_CHAR */
+
+
+#define	ReadOK(file,buffer,len)	(JFREAD(file,buffer,len) == ((size_t) (len)))
+
+
+/* Private version of data source object */
+
+typedef struct _tga_source_struct * tga_source_ptr;
+
+typedef struct _tga_source_struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  j_compress_ptr cinfo;		/* back link saves passing separate parm */
+
+  JSAMPARRAY colormap;		/* Targa colormap (converted to my format) */
+
+  jvirt_sarray_ptr whole_image;	/* Needed if funny input row order */
+  JDIMENSION current_row;	/* Current logical row number to read */
+
+  /* Pointer to routine to extract next Targa pixel from input file */
+  JMETHOD(void, read_pixel, (tga_source_ptr sinfo));
+
+  /* Result of read_pixel is delivered here: */
+  U_CHAR tga_pixel[4];
+
+  int pixel_size;		/* Bytes per Targa pixel (1 to 4) */
+
+  /* State info for reading RLE-coded pixels; both counts must be init to 0 */
+  int block_count;		/* # of pixels remaining in RLE block */
+  int dup_pixel_count;		/* # of times to duplicate previous pixel */
+
+  /* This saves the correct pixel-row-expansion method for preload_image */
+  JMETHOD(JDIMENSION, get_pixel_rows, (j_compress_ptr cinfo,
+				       cjpeg_source_ptr sinfo));
+} tga_source_struct;
+
+
+/* For expanding 5-bit pixel values to 8-bit with best rounding */
+
+static const UINT8 c5to8bits[32] = {
+    0,   8,  16,  25,  33,  41,  49,  58,
+   66,  74,  82,  90,  99, 107, 115, 123,
+  132, 140, 148, 156, 165, 173, 181, 189,
+  197, 206, 214, 222, 230, 239, 247, 255
+};
+
+
+
+LOCAL(int)
+read_byte (tga_source_ptr sinfo)
+/* Read next byte from Targa file */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int c;
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(sinfo->cinfo, JERR_INPUT_EOF);
+  return c;
+}
+
+
+LOCAL(void)
+read_colormap (tga_source_ptr sinfo, int cmaplen, int mapentrysize)
+/* Read the colormap from a Targa file */
+{
+  int i;
+
+  /* Presently only handles 24-bit BGR format */
+  if (mapentrysize != 24)
+    ERREXIT(sinfo->cinfo, JERR_TGA_BADCMAP);
+
+  for (i = 0; i < cmaplen; i++) {
+    sinfo->colormap[2][i] = (JSAMPLE) read_byte(sinfo);
+    sinfo->colormap[1][i] = (JSAMPLE) read_byte(sinfo);
+    sinfo->colormap[0][i] = (JSAMPLE) read_byte(sinfo);
+  }
+}
+
+
+/*
+ * read_pixel methods: get a single pixel from Targa file into tga_pixel[]
+ */
+
+METHODDEF(void)
+read_non_rle_pixel (tga_source_ptr sinfo)
+/* Read one Targa pixel from the input file; no RLE expansion */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int i;
+
+  for (i = 0; i < sinfo->pixel_size; i++) {
+    sinfo->tga_pixel[i] = (U_CHAR) getc(infile);
+  }
+}
+
+
+METHODDEF(void)
+read_rle_pixel (tga_source_ptr sinfo)
+/* Read one Targa pixel from the input file, expanding RLE data as needed */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int i;
+
+  /* Duplicate previously read pixel? */
+  if (sinfo->dup_pixel_count > 0) {
+    sinfo->dup_pixel_count--;
+    return;
+  }
+
+  /* Time to read RLE block header? */
+  if (--sinfo->block_count < 0) { /* decrement pixels remaining in block */
+    i = read_byte(sinfo);
+    if (i & 0x80) {		/* Start of duplicate-pixel block? */
+      sinfo->dup_pixel_count = i & 0x7F; /* number of dups after this one */
+      sinfo->block_count = 0;	/* then read new block header */
+    } else {
+      sinfo->block_count = i & 0x7F; /* number of pixels after this one */
+    }
+  }
+
+  /* Read next pixel */
+  for (i = 0; i < sinfo->pixel_size; i++) {
+    sinfo->tga_pixel[i] = (U_CHAR) getc(infile);
+  }
+}
+
+
+/*
+ * Read one row of pixels.
+ *
+ * We provide several different versions depending on input file format.
+ */
+
+
+METHODDEF(JDIMENSION)
+get_8bit_gray_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 8-bit grayscale pixels */
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+  
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    *ptr++ = (JSAMPLE) UCH(source->tga_pixel[0]);
+  }
+  return 1;
+}
+
+METHODDEF(JDIMENSION)
+get_8bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 8-bit colormap indexes */
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  register int t;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+  register JSAMPARRAY colormap = source->colormap;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    t = UCH(source->tga_pixel[0]);
+    *ptr++ = colormap[0][t];
+    *ptr++ = colormap[1][t];
+    *ptr++ = colormap[2][t];
+  }
+  return 1;
+}
+
+METHODDEF(JDIMENSION)
+get_16bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 16-bit pixels */
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  register int t;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+  
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    t = UCH(source->tga_pixel[0]);
+    t += UCH(source->tga_pixel[1]) << 8;
+    /* We expand 5 bit data to 8 bit sample width.
+     * The format of the 16-bit (LSB first) input word is
+     *     xRRRRRGGGGGBBBBB
+     */
+    ptr[2] = (JSAMPLE) c5to8bits[t & 0x1F];
+    t >>= 5;
+    ptr[1] = (JSAMPLE) c5to8bits[t & 0x1F];
+    t >>= 5;
+    ptr[0] = (JSAMPLE) c5to8bits[t & 0x1F];
+    ptr += 3;
+  }
+  return 1;
+}
+
+METHODDEF(JDIMENSION)
+get_24bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 24-bit pixels */
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+  
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    *ptr++ = (JSAMPLE) UCH(source->tga_pixel[2]); /* change BGR to RGB order */
+    *ptr++ = (JSAMPLE) UCH(source->tga_pixel[1]);
+    *ptr++ = (JSAMPLE) UCH(source->tga_pixel[0]);
+  }
+  return 1;
+}
+
+/*
+ * Targa also defines a 32-bit pixel format with order B,G,R,A.
+ * We presently ignore the attribute byte, so the code for reading
+ * these pixels is identical to the 24-bit routine above.
+ * This works because the actual pixel length is only known to read_pixel.
+ */
+
+#define get_32bit_row  get_24bit_row
+
+
+/*
+ * This method is for re-reading the input data in standard top-down
+ * row order.  The entire image has already been read into whole_image
+ * with proper conversion of pixel format, but it's in a funny row order.
+ */
+
+METHODDEF(JDIMENSION)
+get_memory_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  JDIMENSION source_row;
+
+  /* Compute row of source that maps to current_row of normal order */
+  /* For now, assume image is bottom-up and not interlaced. */
+  /* NEEDS WORK to support interlaced images! */
+  source_row = cinfo->image_height - source->current_row - 1;
+
+  /* Fetch that row from virtual array */
+  source->pub.buffer = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->whole_image,
+     source_row, (JDIMENSION) 1, FALSE);
+
+  source->current_row++;
+  return 1;
+}
+
+
+/*
+ * This method loads the image into whole_image during the first call on
+ * get_pixel_rows.  The get_pixel_rows pointer is then adjusted to call
+ * get_memory_row on subsequent calls.
+ */
+
+METHODDEF(JDIMENSION)
+preload_image (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  JDIMENSION row;
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+
+  /* Read the data into a virtual array in input-file row order. */
+  for (row = 0; row < cinfo->image_height; row++) {
+    if (progress != NULL) {
+      progress->pub.pass_counter = (long) row;
+      progress->pub.pass_limit = (long) cinfo->image_height;
+      (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+    }
+    source->pub.buffer = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr) cinfo, source->whole_image, row, (JDIMENSION) 1, TRUE);
+    (*source->get_pixel_rows) (cinfo, sinfo);
+  }
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+
+  /* Set up to read from the virtual array in unscrambled order */
+  source->pub.get_pixel_rows = get_memory_row;
+  source->current_row = 0;
+  /* And read the first row */
+  return get_memory_row(cinfo, sinfo);
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_tga (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  U_CHAR targaheader[18];
+  int idlen, cmaptype, subtype, flags, interlace_type, components;
+  unsigned int width, height, maplen;
+  boolean is_bottom_up;
+
+#define GET_2B(offset)	((unsigned int) UCH(targaheader[offset]) + \
+			 (((unsigned int) UCH(targaheader[offset+1])) << 8))
+
+  if (! ReadOK(source->pub.input_file, targaheader, 18))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+
+  /* Pretend "15-bit" pixels are 16-bit --- we ignore attribute bit anyway */
+  if (targaheader[16] == 15)
+    targaheader[16] = 16;
+
+  idlen = UCH(targaheader[0]);
+  cmaptype = UCH(targaheader[1]);
+  subtype = UCH(targaheader[2]);
+  maplen = GET_2B(5);
+  width = GET_2B(12);
+  height = GET_2B(14);
+  source->pixel_size = UCH(targaheader[16]) >> 3;
+  flags = UCH(targaheader[17]);	/* Image Descriptor byte */
+
+  is_bottom_up = ((flags & 0x20) == 0);	/* bit 5 set => top-down */
+  interlace_type = flags >> 6;	/* bits 6/7 are interlace code */
+
+  if (cmaptype > 1 ||		/* cmaptype must be 0 or 1 */
+      source->pixel_size < 1 || source->pixel_size > 4 ||
+      (UCH(targaheader[16]) & 7) != 0 || /* bits/pixel must be multiple of 8 */
+      interlace_type != 0)	/* currently don't allow interlaced image */
+    ERREXIT(cinfo, JERR_TGA_BADPARMS);
+  
+  if (subtype > 8) {
+    /* It's an RLE-coded file */
+    source->read_pixel = read_rle_pixel;
+    source->block_count = source->dup_pixel_count = 0;
+    subtype -= 8;
+  } else {
+    /* Non-RLE file */
+    source->read_pixel = read_non_rle_pixel;
+  }
+
+  /* Now should have subtype 1, 2, or 3 */
+  components = 3;		/* until proven different */
+  cinfo->in_color_space = JCS_RGB;
+
+  switch (subtype) {
+  case 1:			/* Colormapped image */
+    if (source->pixel_size == 1 && cmaptype == 1)
+      source->get_pixel_rows = get_8bit_row;
+    else
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    TRACEMS2(cinfo, 1, JTRC_TGA_MAPPED, width, height);
+    break;
+  case 2:			/* RGB image */
+    switch (source->pixel_size) {
+    case 2:
+      source->get_pixel_rows = get_16bit_row;
+      break;
+    case 3:
+      source->get_pixel_rows = get_24bit_row;
+      break;
+    case 4:
+      source->get_pixel_rows = get_32bit_row;
+      break;
+    default:
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+      break;
+    }
+    TRACEMS2(cinfo, 1, JTRC_TGA, width, height);
+    break;
+  case 3:			/* Grayscale image */
+    components = 1;
+    cinfo->in_color_space = JCS_GRAYSCALE;
+    if (source->pixel_size == 1)
+      source->get_pixel_rows = get_8bit_gray_row;
+    else
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    TRACEMS2(cinfo, 1, JTRC_TGA_GRAY, width, height);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    break;
+  }
+
+  if (is_bottom_up) {
+    /* Create a virtual array to buffer the upside-down image. */
+    source->whole_image = (*cinfo->mem->request_virt_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+       (JDIMENSION) width * components, (JDIMENSION) height, (JDIMENSION) 1);
+    if (cinfo->progress != NULL) {
+      cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+      progress->total_extra_passes++; /* count file input as separate pass */
+    }
+    /* source->pub.buffer will point to the virtual array. */
+    source->pub.buffer_height = 1; /* in case anyone looks at it */
+    source->pub.get_pixel_rows = preload_image;
+  } else {
+    /* Don't need a virtual array, but do need a one-row input buffer. */
+    source->whole_image = NULL;
+    source->pub.buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) width * components, (JDIMENSION) 1);
+    source->pub.buffer_height = 1;
+    source->pub.get_pixel_rows = source->get_pixel_rows;
+  }
+  
+  while (idlen--)		/* Throw away ID field */
+    (void) read_byte(source);
+
+  if (maplen > 0) {
+    if (maplen > 256 || GET_2B(3) != 0)
+      ERREXIT(cinfo, JERR_TGA_BADCMAP);
+    /* Allocate space to store the colormap */
+    source->colormap = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE, (JDIMENSION) maplen, (JDIMENSION) 3);
+    /* and read it from the file */
+    read_colormap(source, (int) maplen, UCH(targaheader[7]));
+  } else {
+    if (cmaptype)		/* but you promised a cmap! */
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    source->colormap = NULL;
+  }
+
+  cinfo->input_components = components;
+  cinfo->data_precision = 8;
+  cinfo->image_width = width;
+  cinfo->image_height = height;
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_tga (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for Targa format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_targa (j_compress_ptr cinfo)
+{
+  tga_source_ptr source;
+
+  /* Create module interface object */
+  source = (tga_source_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(tga_source_struct));
+  source->cinfo = cinfo;	/* make back link for subroutines */
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_tga;
+  source->pub.finish_input = finish_input_tga;
+
+  return (cjpeg_source_ptr) source;
+}
+
+#endif /* TARGA_SUPPORTED */
diff --git a/libjpegtwrp/structure.doc b/libjpegtwrp/structure.doc
new file mode 100644
index 000000000..51c9def7e
--- /dev/null
+++ b/libjpegtwrp/structure.doc
@@ -0,0 +1,948 @@
+IJG JPEG LIBRARY:  SYSTEM ARCHITECTURE
+
+Copyright (C) 1991-1995, Thomas G. Lane.
+This file is part of the Independent JPEG Group's software.
+For conditions of distribution and use, see the accompanying README file.
+
+
+This file provides an overview of the architecture of the IJG JPEG software;
+that is, the functions of the various modules in the system and the interfaces
+between modules.  For more precise details about any data structure or calling
+convention, see the include files and comments in the source code.
+
+We assume that the reader is already somewhat familiar with the JPEG standard.
+The README file includes references for learning about JPEG.  The file
+libjpeg.doc describes the library from the viewpoint of an application
+programmer using the library; it's best to read that file before this one.
+Also, the file coderules.doc describes the coding style conventions we use.
+
+In this document, JPEG-specific terminology follows the JPEG standard:
+  A "component" means a color channel, e.g., Red or Luminance.
+  A "sample" is a single component value (i.e., one number in the image data).
+  A "coefficient" is a frequency coefficient (a DCT transform output number).
+  A "block" is an 8x8 group of samples or coefficients.
+  An "MCU" (minimum coded unit) is an interleaved set of blocks of size
+	determined by the sampling factors, or a single block in a
+	noninterleaved scan.
+We do not use the terms "pixel" and "sample" interchangeably.  When we say
+pixel, we mean an element of the full-size image, while a sample is an element
+of the downsampled image.  Thus the number of samples may vary across
+components while the number of pixels does not.  (This terminology is not used
+rigorously throughout the code, but it is used in places where confusion would
+otherwise result.)
+
+
+*** System features ***
+
+The IJG distribution contains two parts:
+  * A subroutine library for JPEG compression and decompression.
+  * cjpeg/djpeg, two sample applications that use the library to transform
+    JFIF JPEG files to and from several other image formats.
+cjpeg/djpeg are of no great intellectual complexity: they merely add a simple
+command-line user interface and I/O routines for several uncompressed image
+formats.  This document concentrates on the library itself.
+
+We desire the library to be capable of supporting all JPEG baseline, extended
+sequential, and progressive DCT processes.  Hierarchical processes are not
+supported.
+
+The library does not support the lossless (spatial) JPEG process.  Lossless
+JPEG shares little or no code with lossy JPEG, and would normally be used
+without the extensive pre- and post-processing provided by this library.
+We feel that lossless JPEG is better handled by a separate library.
+
+Within these limits, any set of compression parameters allowed by the JPEG
+spec should be readable for decompression.  (We can be more restrictive about
+what formats we can generate.)  Although the system design allows for all
+parameter values, some uncommon settings are not yet implemented and may
+never be; nonintegral sampling ratios are the prime example.  Furthermore,
+we treat 8-bit vs. 12-bit data precision as a compile-time switch, not a
+run-time option, because most machines can store 8-bit pixels much more
+compactly than 12-bit.
+
+For legal reasons, JPEG arithmetic coding is not currently supported, but
+extending the library to include it would be straightforward.
+
+By itself, the library handles only interchange JPEG datastreams --- in
+particular the widely used JFIF file format.  The library can be used by
+surrounding code to process interchange or abbreviated JPEG datastreams that
+are embedded in more complex file formats.  (For example, libtiff uses this
+library to implement JPEG compression within the TIFF file format.)
+
+The library includes a substantial amount of code that is not covered by the
+JPEG standard but is necessary for typical applications of JPEG.  These
+functions preprocess the image before JPEG compression or postprocess it after
+decompression.  They include colorspace conversion, downsampling/upsampling,
+and color quantization.  This code can be omitted if not needed.
+
+A wide range of quality vs. speed tradeoffs are possible in JPEG processing,
+and even more so in decompression postprocessing.  The decompression library
+provides multiple implementations that cover most of the useful tradeoffs,
+ranging from very-high-quality down to fast-preview operation.  On the
+compression side we have generally not provided low-quality choices, since
+compression is normally less time-critical.  It should be understood that the
+low-quality modes may not meet the JPEG standard's accuracy requirements;
+nonetheless, they are useful for viewers.
+
+
+*** Portability issues ***
+
+Portability is an essential requirement for the library.  The key portability
+issues that show up at the level of system architecture are:
+
+1.  Memory usage.  We want the code to be able to run on PC-class machines
+with limited memory.  Images should therefore be processed sequentially (in
+strips), to avoid holding the whole image in memory at once.  Where a
+full-image buffer is necessary, we should be able to use either virtual memory
+or temporary files.
+
+2.  Near/far pointer distinction.  To run efficiently on 80x86 machines, the
+code should distinguish "small" objects (kept in near data space) from
+"large" ones (kept in far data space).  This is an annoying restriction, but
+fortunately it does not impact code quality for less brain-damaged machines,
+and the source code clutter turns out to be minimal with sufficient use of
+pointer typedefs.
+
+3. Data precision.  We assume that "char" is at least 8 bits, "short" and
+"int" at least 16, "long" at least 32.  The code will work fine with larger
+data sizes, although memory may be used inefficiently in some cases.  However,
+the JPEG compressed datastream must ultimately appear on external storage as a
+sequence of 8-bit bytes if it is to conform to the standard.  This may pose a
+problem on machines where char is wider than 8 bits.  The library represents
+compressed data as an array of values of typedef JOCTET.  If no data type
+exactly 8 bits wide is available, custom data source and data destination
+modules must be written to unpack and pack the chosen JOCTET datatype into
+8-bit external representation.
+
+
+*** System overview ***
+
+The compressor and decompressor are each divided into two main sections:
+the JPEG compressor or decompressor proper, and the preprocessing or
+postprocessing functions.  The interface between these two sections is the
+image data that the official JPEG spec regards as its input or output: this
+data is in the colorspace to be used for compression, and it is downsampled
+to the sampling factors to be used.  The preprocessing and postprocessing
+steps are responsible for converting a normal image representation to or from
+this form.  (Those few applications that want to deal with YCbCr downsampled
+data can skip the preprocessing or postprocessing step.)
+
+Looking more closely, the compressor library contains the following main
+elements:
+
+  Preprocessing:
+    * Color space conversion (e.g., RGB to YCbCr).
+    * Edge expansion and downsampling.  Optionally, this step can do simple
+      smoothing --- this is often helpful for low-quality source data.
+  JPEG proper:
+    * MCU assembly, DCT, quantization.
+    * Entropy coding (sequential or progressive, Huffman or arithmetic).
+
+In addition to these modules we need overall control, marker generation,
+and support code (memory management & error handling).  There is also a
+module responsible for physically writing the output data --- typically
+this is just an interface to fwrite(), but some applications may need to
+do something else with the data.
+
+The decompressor library contains the following main elements:
+
+  JPEG proper:
+    * Entropy decoding (sequential or progressive, Huffman or arithmetic).
+    * Dequantization, inverse DCT, MCU disassembly.
+  Postprocessing:
+    * Upsampling.  Optionally, this step may be able to do more general
+      rescaling of the image.
+    * Color space conversion (e.g., YCbCr to RGB).  This step may also
+      provide gamma adjustment [ currently it does not ].
+    * Optional color quantization (e.g., reduction to 256 colors).
+    * Optional color precision reduction (e.g., 24-bit to 15-bit color).
+      [This feature is not currently implemented.]
+
+We also need overall control, marker parsing, and a data source module.
+The support code (memory management & error handling) can be shared with
+the compression half of the library.
+
+There may be several implementations of each of these elements, particularly
+in the decompressor, where a wide range of speed/quality tradeoffs is very
+useful.  It must be understood that some of the best speedups involve
+merging adjacent steps in the pipeline.  For example, upsampling, color space
+conversion, and color quantization might all be done at once when using a
+low-quality ordered-dither technique.  The system architecture is designed to
+allow such merging where appropriate.
+
+
+Note: it is convenient to regard edge expansion (padding to block boundaries)
+as a preprocessing/postprocessing function, even though the JPEG spec includes
+it in compression/decompression.  We do this because downsampling/upsampling
+can be simplified a little if they work on padded data: it's not necessary to
+have special cases at the right and bottom edges.  Therefore the interface
+buffer is always an integral number of blocks wide and high, and we expect
+compression preprocessing to pad the source data properly.  Padding will occur
+only to the next block (8-sample) boundary.  In an interleaved-scan situation,
+additional dummy blocks may be used to fill out MCUs, but the MCU assembly and
+disassembly logic will create or discard these blocks internally.  (This is
+advantageous for speed reasons, since we avoid DCTing the dummy blocks.
+It also permits a small reduction in file size, because the compressor can
+choose dummy block contents so as to minimize their size in compressed form.
+Finally, it makes the interface buffer specification independent of whether
+the file is actually interleaved or not.)  Applications that wish to deal
+directly with the downsampled data must provide similar buffering and padding
+for odd-sized images.
+
+
+*** Poor man's object-oriented programming ***
+
+It should be clear by now that we have a lot of quasi-independent processing
+steps, many of which have several possible behaviors.  To avoid cluttering the
+code with lots of switch statements, we use a simple form of object-style
+programming to separate out the different possibilities.
+
+For example, two different color quantization algorithms could be implemented
+as two separate modules that present the same external interface; at runtime,
+the calling code will access the proper module indirectly through an "object".
+
+We can get the limited features we need while staying within portable C.
+The basic tool is a function pointer.  An "object" is just a struct
+containing one or more function pointer fields, each of which corresponds to
+a method name in real object-oriented languages.  During initialization we
+fill in the function pointers with references to whichever module we have
+determined we need to use in this run.  Then invocation of the module is done
+by indirecting through a function pointer; on most machines this is no more
+expensive than a switch statement, which would be the only other way of
+making the required run-time choice.  The really significant benefit, of
+course, is keeping the source code clean and well structured.
+
+We can also arrange to have private storage that varies between different
+implementations of the same kind of object.  We do this by making all the
+module-specific object structs be separately allocated entities, which will
+be accessed via pointers in the master compression or decompression struct.
+The "public" fields or methods for a given kind of object are specified by
+a commonly known struct.  But a module's initialization code can allocate
+a larger struct that contains the common struct as its first member, plus
+additional private fields.  With appropriate pointer casting, the module's
+internal functions can access these private fields.  (For a simple example,
+see jdatadst.c, which implements the external interface specified by struct
+jpeg_destination_mgr, but adds extra fields.)
+
+(Of course this would all be a lot easier if we were using C++, but we are
+not yet prepared to assume that everyone has a C++ compiler.)
+
+An important benefit of this scheme is that it is easy to provide multiple
+versions of any method, each tuned to a particular case.  While a lot of
+precalculation might be done to select an optimal implementation of a method,
+the cost per invocation is constant.  For example, the upsampling step might
+have a "generic" method, plus one or more "hardwired" methods for the most
+popular sampling factors; the hardwired methods would be faster because they'd
+use straight-line code instead of for-loops.  The cost to determine which
+method to use is paid only once, at startup, and the selection criteria are
+hidden from the callers of the method.
+
+This plan differs a little bit from usual object-oriented structures, in that
+only one instance of each object class will exist during execution.  The
+reason for having the class structure is that on different runs we may create
+different instances (choose to execute different modules).  You can think of
+the term "method" as denoting the common interface presented by a particular
+set of interchangeable functions, and "object" as denoting a group of related
+methods, or the total shared interface behavior of a group of modules.
+
+
+*** Overall control structure ***
+
+We previously mentioned the need for overall control logic in the compression
+and decompression libraries.  In IJG implementations prior to v5, overall
+control was mostly provided by "pipeline control" modules, which proved to be
+large, unwieldy, and hard to understand.  To improve the situation, the
+control logic has been subdivided into multiple modules.  The control modules
+consist of:
+
+1. Master control for module selection and initialization.  This has two
+responsibilities:
+
+   1A.  Startup initialization at the beginning of image processing.
+        The individual processing modules to be used in this run are selected
+        and given initialization calls.
+
+   1B.  Per-pass control.  This determines how many passes will be performed
+        and calls each active processing module to configure itself
+        appropriately at the beginning of each pass.  End-of-pass processing,
+	where necessary, is also invoked from the master control module.
+
+   Method selection is partially distributed, in that a particular processing
+   module may contain several possible implementations of a particular method,
+   which it will select among when given its initialization call.  The master
+   control code need only be concerned with decisions that affect more than
+   one module.
+ 
+2. Data buffering control.  A separate control module exists for each
+   inter-processing-step data buffer.  This module is responsible for
+   invoking the processing steps that write or read that data buffer.
+
+Each buffer controller sees the world as follows:
+
+input data => processing step A => buffer => processing step B => output data
+                      |              |               |
+              ------------------ controller ------------------
+
+The controller knows the dataflow requirements of steps A and B: how much data
+they want to accept in one chunk and how much they output in one chunk.  Its
+function is to manage its buffer and call A and B at the proper times.
+
+A data buffer control module may itself be viewed as a processing step by a
+higher-level control module; thus the control modules form a binary tree with
+elementary processing steps at the leaves of the tree.
+
+The control modules are objects.  A considerable amount of flexibility can
+be had by replacing implementations of a control module.  For example:
+* Merging of adjacent steps in the pipeline is done by replacing a control
+  module and its pair of processing-step modules with a single processing-
+  step module.  (Hence the possible merges are determined by the tree of
+  control modules.)
+* In some processing modes, a given interstep buffer need only be a "strip"
+  buffer large enough to accommodate the desired data chunk sizes.  In other
+  modes, a full-image buffer is needed and several passes are required.
+  The control module determines which kind of buffer is used and manipulates
+  virtual array buffers as needed.  One or both processing steps may be
+  unaware of the multi-pass behavior.
+
+In theory, we might be able to make all of the data buffer controllers
+interchangeable and provide just one set of implementations for all.  In
+practice, each one contains considerable special-case processing for its
+particular job.  The buffer controller concept should be regarded as an
+overall system structuring principle, not as a complete description of the
+task performed by any one controller.
+
+
+*** Compression object structure ***
+
+Here is a sketch of the logical structure of the JPEG compression library:
+
+                                                 |-- Colorspace conversion
+                  |-- Preprocessing controller --|
+                  |                              |-- Downsampling
+Main controller --|
+                  |                            |-- Forward DCT, quantize
+                  |-- Coefficient controller --|
+                                               |-- Entropy encoding
+
+This sketch also describes the flow of control (subroutine calls) during
+typical image data processing.  Each of the components shown in the diagram is
+an "object" which may have several different implementations available.  One
+or more source code files contain the actual implementation(s) of each object.
+
+The objects shown above are:
+
+* Main controller: buffer controller for the subsampled-data buffer, which
+  holds the preprocessed input data.  This controller invokes preprocessing to
+  fill the subsampled-data buffer, and JPEG compression to empty it.  There is
+  usually no need for a full-image buffer here; a strip buffer is adequate.
+
+* Preprocessing controller: buffer controller for the downsampling input data
+  buffer, which lies between colorspace conversion and downsampling.  Note
+  that a unified conversion/downsampling module would probably replace this
+  controller entirely.
+
+* Colorspace conversion: converts application image data into the desired
+  JPEG color space; also changes the data from pixel-interleaved layout to
+  separate component planes.  Processes one pixel row at a time.
+
+* Downsampling: performs reduction of chroma components as required.
+  Optionally may perform pixel-level smoothing as well.  Processes a "row
+  group" at a time, where a row group is defined as Vmax pixel rows of each
+  component before downsampling, and Vk sample rows afterwards (remember Vk
+  differs across components).  Some downsampling or smoothing algorithms may
+  require context rows above and below the current row group; the
+  preprocessing controller is responsible for supplying these rows via proper
+  buffering.  The downsampler is responsible for edge expansion at the right
+  edge (i.e., extending each sample row to a multiple of 8 samples); but the
+  preprocessing controller is responsible for vertical edge expansion (i.e.,
+  duplicating the bottom sample row as needed to make a multiple of 8 rows).
+
+* Coefficient controller: buffer controller for the DCT-coefficient data.
+  This controller handles MCU assembly, including insertion of dummy DCT
+  blocks when needed at the right or bottom edge.  When performing
+  Huffman-code optimization or emitting a multiscan JPEG file, this
+  controller is responsible for buffering the full image.  The equivalent of
+  one fully interleaved MCU row of subsampled data is processed per call,
+  even when the JPEG file is noninterleaved.
+
+* Forward DCT and quantization: Perform DCT, quantize, and emit coefficients.
+  Works on one or more DCT blocks at a time.  (Note: the coefficients are now
+  emitted in normal array order, which the entropy encoder is expected to
+  convert to zigzag order as necessary.  Prior versions of the IJG code did
+  the conversion to zigzag order within the quantization step.)
+
+* Entropy encoding: Perform Huffman or arithmetic entropy coding and emit the
+  coded data to the data destination module.  Works on one MCU per call.
+  For progressive JPEG, the same DCT blocks are fed to the entropy coder
+  during each pass, and the coder must emit the appropriate subset of
+  coefficients.
+
+In addition to the above objects, the compression library includes these
+objects:
+
+* Master control: determines the number of passes required, controls overall
+  and per-pass initialization of the other modules.
+
+* Marker writing: generates JPEG markers (except for RSTn, which is emitted
+  by the entropy encoder when needed).
+
+* Data destination manager: writes the output JPEG datastream to its final
+  destination (e.g., a file).  The destination manager supplied with the
+  library knows how to write to a stdio stream; for other behaviors, the
+  surrounding application may provide its own destination manager.
+
+* Memory manager: allocates and releases memory, controls virtual arrays
+  (with backing store management, where required).
+
+* Error handler: performs formatting and output of error and trace messages;
+  determines handling of nonfatal errors.  The surrounding application may
+  override some or all of this object's methods to change error handling.
+
+* Progress monitor: supports output of "percent-done" progress reports.
+  This object represents an optional callback to the surrounding application:
+  if wanted, it must be supplied by the application.
+
+The error handler, destination manager, and progress monitor objects are
+defined as separate objects in order to simplify application-specific
+customization of the JPEG library.  A surrounding application may override
+individual methods or supply its own all-new implementation of one of these
+objects.  The object interfaces for these objects are therefore treated as
+part of the application interface of the library, whereas the other objects
+are internal to the library.
+
+The error handler and memory manager are shared by JPEG compression and
+decompression; the progress monitor, if used, may be shared as well.
+
+
+*** Decompression object structure ***
+
+Here is a sketch of the logical structure of the JPEG decompression library:
+
+                                               |-- Entropy decoding
+                  |-- Coefficient controller --|
+                  |                            |-- Dequantize, Inverse DCT
+Main controller --|
+                  |                               |-- Upsampling
+                  |-- Postprocessing controller --|   |-- Colorspace conversion
+                                                  |-- Color quantization
+                                                  |-- Color precision reduction
+
+As before, this diagram also represents typical control flow.  The objects
+shown are:
+
+* Main controller: buffer controller for the subsampled-data buffer, which
+  holds the output of JPEG decompression proper.  This controller's primary
+  task is to feed the postprocessing procedure.  Some upsampling algorithms
+  may require context rows above and below the current row group; when this
+  is true, the main controller is responsible for managing its buffer so as
+  to make context rows available.  In the current design, the main buffer is
+  always a strip buffer; a full-image buffer is never required.
+
+* Coefficient controller: buffer controller for the DCT-coefficient data.
+  This controller handles MCU disassembly, including deletion of any dummy
+  DCT blocks at the right or bottom edge.  When reading a multiscan JPEG
+  file, this controller is responsible for buffering the full image.
+  (Buffering DCT coefficients, rather than samples, is necessary to support
+  progressive JPEG.)  The equivalent of one fully interleaved MCU row of
+  subsampled data is processed per call, even when the source JPEG file is
+  noninterleaved.
+
+* Entropy decoding: Read coded data from the data source module and perform
+  Huffman or arithmetic entropy decoding.  Works on one MCU per call.
+  For progressive JPEG decoding, the coefficient controller supplies the prior
+  coefficients of each MCU (initially all zeroes), which the entropy decoder
+  modifies in each scan.
+
+* Dequantization and inverse DCT: like it says.  Note that the coefficients
+  buffered by the coefficient controller have NOT been dequantized; we
+  merge dequantization and inverse DCT into a single step for speed reasons.
+  When scaled-down output is asked for, simplified DCT algorithms may be used
+  that emit only 1x1, 2x2, or 4x4 samples per DCT block, not the full 8x8.
+  Works on one DCT block at a time.
+
+* Postprocessing controller: buffer controller for the color quantization
+  input buffer, when quantization is in use.  (Without quantization, this
+  controller just calls the upsampler.)  For two-pass quantization, this
+  controller is responsible for buffering the full-image data.
+
+* Upsampling: restores chroma components to full size.  (May support more
+  general output rescaling, too.  Note that if undersized DCT outputs have
+  been emitted by the DCT module, this module must adjust so that properly
+  sized outputs are created.)  Works on one row group at a time.  This module
+  also calls the color conversion module, so its top level is effectively a
+  buffer controller for the upsampling->color conversion buffer.  However, in
+  all but the highest-quality operating modes, upsampling and color
+  conversion are likely to be merged into a single step.
+
+* Colorspace conversion: convert from JPEG color space to output color space,
+  and change data layout from separate component planes to pixel-interleaved.
+  Works on one pixel row at a time.
+
+* Color quantization: reduce the data to colormapped form, using either an
+  externally specified colormap or an internally generated one.  This module
+  is not used for full-color output.  Works on one pixel row at a time; may
+  require two passes to generate a color map.  Note that the output will
+  always be a single component representing colormap indexes.  In the current
+  design, the output values are JSAMPLEs, so an 8-bit compilation cannot
+  quantize to more than 256 colors.  This is unlikely to be a problem in
+  practice.
+
+* Color reduction: this module handles color precision reduction, e.g.,
+  generating 15-bit color (5 bits/primary) from JPEG's 24-bit output.
+  Not quite clear yet how this should be handled... should we merge it with
+  colorspace conversion???
+
+Note that some high-speed operating modes might condense the entire
+postprocessing sequence to a single module (upsample, color convert, and
+quantize in one step).
+
+In addition to the above objects, the decompression library includes these
+objects:
+
+* Master control: determines the number of passes required, controls overall
+  and per-pass initialization of the other modules.  This is subdivided into
+  input and output control: jdinput.c controls only input-side processing,
+  while jdmaster.c handles overall initialization and output-side control.
+
+* Marker reading: decodes JPEG markers (except for RSTn).
+
+* Data source manager: supplies the input JPEG datastream.  The source
+  manager supplied with the library knows how to read from a stdio stream;
+  for other behaviors, the surrounding application may provide its own source
+  manager.
+
+* Memory manager: same as for compression library.
+
+* Error handler: same as for compression library.
+
+* Progress monitor: same as for compression library.
+
+As with compression, the data source manager, error handler, and progress
+monitor are candidates for replacement by a surrounding application.
+
+
+*** Decompression input and output separation ***
+
+To support efficient incremental display of progressive JPEG files, the
+decompressor is divided into two sections that can run independently:
+
+1. Data input includes marker parsing, entropy decoding, and input into the
+   coefficient controller's DCT coefficient buffer.  Note that this
+   processing is relatively cheap and fast.
+
+2. Data output reads from the DCT coefficient buffer and performs the IDCT
+   and all postprocessing steps.
+
+For a progressive JPEG file, the data input processing is allowed to get
+arbitrarily far ahead of the data output processing.  (This occurs only
+if the application calls jpeg_consume_input(); otherwise input and output
+run in lockstep, since the input section is called only when the output
+section needs more data.)  In this way the application can avoid making
+extra display passes when data is arriving faster than the display pass
+can run.  Furthermore, it is possible to abort an output pass without
+losing anything, since the coefficient buffer is read-only as far as the
+output section is concerned.  See libjpeg.doc for more detail.
+
+A full-image coefficient array is only created if the JPEG file has multiple
+scans (or if the application specifies buffered-image mode anyway).  When
+reading a single-scan file, the coefficient controller normally creates only
+a one-MCU buffer, so input and output processing must run in lockstep in this
+case.  jpeg_consume_input() is effectively a no-op in this situation.
+
+The main impact of dividing the decompressor in this fashion is that we must
+be very careful with shared variables in the cinfo data structure.  Each
+variable that can change during the course of decompression must be
+classified as belonging to data input or data output, and each section must
+look only at its own variables.  For example, the data output section may not
+depend on any of the variables that describe the current scan in the JPEG
+file, because these may change as the data input section advances into a new
+scan.
+
+The progress monitor is (somewhat arbitrarily) defined to treat input of the
+file as one pass when buffered-image mode is not used, and to ignore data
+input work completely when buffered-image mode is used.  Note that the
+library has no reliable way to predict the number of passes when dealing
+with a progressive JPEG file, nor can it predict the number of output passes
+in buffered-image mode.  So the work estimate is inherently bogus anyway.
+
+No comparable division is currently made in the compression library, because
+there isn't any real need for it.
+
+
+*** Data formats ***
+
+Arrays of pixel sample values use the following data structure:
+
+    typedef something JSAMPLE;		a pixel component value, 0..MAXJSAMPLE
+    typedef JSAMPLE *JSAMPROW;		ptr to a row of samples
+    typedef JSAMPROW *JSAMPARRAY;	ptr to a list of rows
+    typedef JSAMPARRAY *JSAMPIMAGE;	ptr to a list of color-component arrays
+
+The basic element type JSAMPLE will typically be one of unsigned char,
+(signed) char, or short.  Short will be used if samples wider than 8 bits are
+to be supported (this is a compile-time option).  Otherwise, unsigned char is
+used if possible.  If the compiler only supports signed chars, then it is
+necessary to mask off the value when reading.  Thus, all reads of JSAMPLE
+values must be coded as "GETJSAMPLE(value)", where the macro will be defined
+as "((value) & 0xFF)" on signed-char machines and "((int) (value))" elsewhere.
+
+With these conventions, JSAMPLE values can be assumed to be >= 0.  This helps
+simplify correct rounding during downsampling, etc.  The JPEG standard's
+specification that sample values run from -128..127 is accommodated by
+subtracting 128 just as the sample value is copied into the source array for
+the DCT step (this will be an array of signed ints).  Similarly, during
+decompression the output of the IDCT step will be immediately shifted back to
+0..255.  (NB: different values are required when 12-bit samples are in use.
+The code is written in terms of MAXJSAMPLE and CENTERJSAMPLE, which will be
+defined as 255 and 128 respectively in an 8-bit implementation, and as 4095
+and 2048 in a 12-bit implementation.)
+
+We use a pointer per row, rather than a two-dimensional JSAMPLE array.  This
+choice costs only a small amount of memory and has several benefits:
+* Code using the data structure doesn't need to know the allocated width of
+  the rows.  This simplifies edge expansion/compression, since we can work
+  in an array that's wider than the logical picture width.
+* Indexing doesn't require multiplication; this is a performance win on many
+  machines.
+* Arrays with more than 64K total elements can be supported even on machines
+  where malloc() cannot allocate chunks larger than 64K.
+* The rows forming a component array may be allocated at different times
+  without extra copying.  This trick allows some speedups in smoothing steps
+  that need access to the previous and next rows.
+
+Note that each color component is stored in a separate array; we don't use the
+traditional layout in which the components of a pixel are stored together.
+This simplifies coding of modules that work on each component independently,
+because they don't need to know how many components there are.  Furthermore,
+we can read or write each component to a temporary file independently, which
+is helpful when dealing with noninterleaved JPEG files.
+
+In general, a specific sample value is accessed by code such as
+	GETJSAMPLE(image[colorcomponent][row][col])
+where col is measured from the image left edge, but row is measured from the
+first sample row currently in memory.  Either of the first two indexings can
+be precomputed by copying the relevant pointer.
+
+
+Since most image-processing applications prefer to work on images in which
+the components of a pixel are stored together, the data passed to or from the
+surrounding application uses the traditional convention: a single pixel is
+represented by N consecutive JSAMPLE values, and an image row is an array of
+(# of color components)*(image width) JSAMPLEs.  One or more rows of data can
+be represented by a pointer of type JSAMPARRAY in this scheme.  This scheme is
+converted to component-wise storage inside the JPEG library.  (Applications
+that want to skip JPEG preprocessing or postprocessing will have to contend
+with component-wise storage.)
+
+
+Arrays of DCT-coefficient values use the following data structure:
+
+    typedef short JCOEF;		a 16-bit signed integer
+    typedef JCOEF JBLOCK[DCTSIZE2];	an 8x8 block of coefficients
+    typedef JBLOCK *JBLOCKROW;		ptr to one horizontal row of 8x8 blocks
+    typedef JBLOCKROW *JBLOCKARRAY;	ptr to a list of such rows
+    typedef JBLOCKARRAY *JBLOCKIMAGE;	ptr to a list of color component arrays
+
+The underlying type is at least a 16-bit signed integer; while "short" is big
+enough on all machines of interest, on some machines it is preferable to use
+"int" for speed reasons, despite the storage cost.  Coefficients are grouped
+into 8x8 blocks (but we always use #defines DCTSIZE and DCTSIZE2 rather than
+"8" and "64").
+
+The contents of a coefficient block may be in either "natural" or zigzagged
+order, and may be true values or divided by the quantization coefficients,
+depending on where the block is in the processing pipeline.  In the current
+library, coefficient blocks are kept in natural order everywhere; the entropy
+codecs zigzag or dezigzag the data as it is written or read.  The blocks
+contain quantized coefficients everywhere outside the DCT/IDCT subsystems.
+(This latter decision may need to be revisited to support variable
+quantization a la JPEG Part 3.)
+
+Notice that the allocation unit is now a row of 8x8 blocks, corresponding to
+eight rows of samples.  Otherwise the structure is much the same as for
+samples, and for the same reasons.
+
+On machines where malloc() can't handle a request bigger than 64Kb, this data
+structure limits us to rows of less than 512 JBLOCKs, or a picture width of
+4000+ pixels.  This seems an acceptable restriction.
+
+
+On 80x86 machines, the bottom-level pointer types (JSAMPROW and JBLOCKROW)
+must be declared as "far" pointers, but the upper levels can be "near"
+(implying that the pointer lists are allocated in the DS segment).
+We use a #define symbol FAR, which expands to the "far" keyword when
+compiling on 80x86 machines and to nothing elsewhere.
+
+
+*** Suspendable processing ***
+
+In some applications it is desirable to use the JPEG library as an
+incremental, memory-to-memory filter.  In this situation the data source or
+destination may be a limited-size buffer, and we can't rely on being able to
+empty or refill the buffer at arbitrary times.  Instead the application would
+like to have control return from the library at buffer overflow/underrun, and
+then resume compression or decompression at a later time.
+
+This scenario is supported for simple cases.  (For anything more complex, we
+recommend that the application "bite the bullet" and develop real multitasking
+capability.)  The libjpeg.doc file goes into more detail about the usage and
+limitations of this capability; here we address the implications for library
+structure.
+
+The essence of the problem is that the entropy codec (coder or decoder) must
+be prepared to stop at arbitrary times.  In turn, the controllers that call
+the entropy codec must be able to stop before having produced or consumed all
+the data that they normally would handle in one call.  That part is reasonably
+straightforward: we make the controller call interfaces include "progress
+counters" which indicate the number of data chunks successfully processed, and
+we require callers to test the counter rather than just assume all of the data
+was processed.
+
+Rather than trying to restart at an arbitrary point, the current Huffman
+codecs are designed to restart at the beginning of the current MCU after a
+suspension due to buffer overflow/underrun.  At the start of each call, the
+codec's internal state is loaded from permanent storage (in the JPEG object
+structures) into local variables.  On successful completion of the MCU, the
+permanent state is updated.  (This copying is not very expensive, and may even
+lead to *improved* performance if the local variables can be registerized.)
+If a suspension occurs, the codec simply returns without updating the state,
+thus effectively reverting to the start of the MCU.  Note that this implies
+leaving some data unprocessed in the source/destination buffer (ie, the
+compressed partial MCU).  The data source/destination module interfaces are
+specified so as to make this possible.  This also implies that the data buffer
+must be large enough to hold a worst-case compressed MCU; a couple thousand
+bytes should be enough.
+
+In a successive-approximation AC refinement scan, the progressive Huffman
+decoder has to be able to undo assignments of newly nonzero coefficients if it
+suspends before the MCU is complete, since decoding requires distinguishing
+previously-zero and previously-nonzero coefficients.  This is a bit tedious
+but probably won't have much effect on performance.  Other variants of Huffman
+decoding need not worry about this, since they will just store the same values
+again if forced to repeat the MCU.
+
+This approach would probably not work for an arithmetic codec, since its
+modifiable state is quite large and couldn't be copied cheaply.  Instead it
+would have to suspend and resume exactly at the point of the buffer end.
+
+The JPEG marker reader is designed to cope with suspension at an arbitrary
+point.  It does so by backing up to the start of the marker parameter segment,
+so the data buffer must be big enough to hold the largest marker of interest.
+Again, a couple KB should be adequate.  (A special "skip" convention is used
+to bypass COM and APPn markers, so these can be larger than the buffer size
+without causing problems; otherwise a 64K buffer would be needed in the worst
+case.)
+
+The JPEG marker writer currently does *not* cope with suspension.  I feel that
+this is not necessary; it is much easier simply to require the application to
+ensure there is enough buffer space before starting.  (An empty 2K buffer is
+more than sufficient for the header markers; and ensuring there are a dozen or
+two bytes available before calling jpeg_finish_compress() will suffice for the
+trailer.)  This would not work for writing multi-scan JPEG files, but
+we simply do not intend to support that capability with suspension.
+
+
+*** Memory manager services ***
+
+The JPEG library's memory manager controls allocation and deallocation of
+memory, and it manages large "virtual" data arrays on machines where the
+operating system does not provide virtual memory.  Note that the same
+memory manager serves both compression and decompression operations.
+
+In all cases, allocated objects are tied to a particular compression or
+decompression master record, and they will be released when that master
+record is destroyed.
+
+The memory manager does not provide explicit deallocation of objects.
+Instead, objects are created in "pools" of free storage, and a whole pool
+can be freed at once.  This approach helps prevent storage-leak bugs, and
+it speeds up operations whenever malloc/free are slow (as they often are).
+The pools can be regarded as lifetime identifiers for objects.  Two
+pools/lifetimes are defined:
+  * JPOOL_PERMANENT	lasts until master record is destroyed
+  * JPOOL_IMAGE		lasts until done with image (JPEG datastream)
+Permanent lifetime is used for parameters and tables that should be carried
+across from one datastream to another; this includes all application-visible
+parameters.  Image lifetime is used for everything else.  (A third lifetime,
+JPOOL_PASS = one processing pass, was originally planned.  However it was
+dropped as not being worthwhile.  The actual usage patterns are such that the
+peak memory usage would be about the same anyway; and having per-pass storage
+substantially complicates the virtual memory allocation rules --- see below.)
+
+The memory manager deals with three kinds of object:
+1. "Small" objects.  Typically these require no more than 10K-20K total.
+2. "Large" objects.  These may require tens to hundreds of K depending on
+   image size.  Semantically they behave the same as small objects, but we
+   distinguish them for two reasons:
+     * On MS-DOS machines, large objects are referenced by FAR pointers,
+       small objects by NEAR pointers.
+     * Pool allocation heuristics may differ for large and small objects.
+   Note that individual "large" objects cannot exceed the size allowed by
+   type size_t, which may be 64K or less on some machines.
+3. "Virtual" objects.  These are large 2-D arrays of JSAMPLEs or JBLOCKs
+   (typically large enough for the entire image being processed).  The
+   memory manager provides stripwise access to these arrays.  On machines
+   without virtual memory, the rest of the array may be swapped out to a
+   temporary file.
+
+(Note: JSAMPARRAY and JBLOCKARRAY data structures are a combination of large
+objects for the data proper and small objects for the row pointers.  For
+convenience and speed, the memory manager provides single routines to create
+these structures.  Similarly, virtual arrays include a small control block
+and a JSAMPARRAY or JBLOCKARRAY working buffer, all created with one call.)
+
+In the present implementation, virtual arrays are only permitted to have image
+lifespan.  (Permanent lifespan would not be reasonable, and pass lifespan is
+not very useful since a virtual array's raison d'etre is to store data for
+multiple passes through the image.)  We also expect that only "small" objects
+will be given permanent lifespan, though this restriction is not required by
+the memory manager.
+
+In a non-virtual-memory machine, some performance benefit can be gained by
+making the in-memory buffers for virtual arrays be as large as possible.
+(For small images, the buffers might fit entirely in memory, so blind
+swapping would be very wasteful.)  The memory manager will adjust the height
+of the buffers to fit within a prespecified maximum memory usage.  In order
+to do this in a reasonably optimal fashion, the manager needs to allocate all
+of the virtual arrays at once.  Therefore, there isn't a one-step allocation
+routine for virtual arrays; instead, there is a "request" routine that simply
+allocates the control block, and a "realize" routine (called just once) that
+determines space allocation and creates all of the actual buffers.  The
+realize routine must allow for space occupied by non-virtual large objects.
+(We don't bother to factor in the space needed for small objects, on the
+grounds that it isn't worth the trouble.)
+
+To support all this, we establish the following protocol for doing business
+with the memory manager:
+  1. Modules must request virtual arrays (which may have only image lifespan)
+     during the initial setup phase, i.e., in their jinit_xxx routines.
+  2. All "large" objects (including JSAMPARRAYs and JBLOCKARRAYs) must also be
+     allocated during initial setup.
+  3. realize_virt_arrays will be called at the completion of initial setup.
+     The above conventions ensure that sufficient information is available
+     for it to choose a good size for virtual array buffers.
+Small objects of any lifespan may be allocated at any time.  We expect that
+the total space used for small objects will be small enough to be negligible
+in the realize_virt_arrays computation.
+
+In a virtual-memory machine, we simply pretend that the available space is
+infinite, thus causing realize_virt_arrays to decide that it can allocate all
+the virtual arrays as full-size in-memory buffers.  The overhead of the
+virtual-array access protocol is very small when no swapping occurs.
+
+A virtual array can be specified to be "pre-zeroed"; when this flag is set,
+never-yet-written sections of the array are set to zero before being made
+available to the caller.  If this flag is not set, never-written sections
+of the array contain garbage.  (This feature exists primarily because the
+equivalent logic would otherwise be needed in jdcoefct.c for progressive
+JPEG mode; we may as well make it available for possible other uses.)
+
+The first write pass on a virtual array is required to occur in top-to-bottom
+order; read passes, as well as any write passes after the first one, may
+access the array in any order.  This restriction exists partly to simplify
+the virtual array control logic, and partly because some file systems may not
+support seeking beyond the current end-of-file in a temporary file.  The main
+implication of this restriction is that rearrangement of rows (such as
+converting top-to-bottom data order to bottom-to-top) must be handled while
+reading data out of the virtual array, not while putting it in.
+
+
+*** Memory manager internal structure ***
+
+To isolate system dependencies as much as possible, we have broken the
+memory manager into two parts.  There is a reasonably system-independent
+"front end" (jmemmgr.c) and a "back end" that contains only the code
+likely to change across systems.  All of the memory management methods
+outlined above are implemented by the front end.  The back end provides
+the following routines for use by the front end (none of these routines
+are known to the rest of the JPEG code):
+
+jpeg_mem_init, jpeg_mem_term	system-dependent initialization/shutdown
+
+jpeg_get_small, jpeg_free_small	interface to malloc and free library routines
+				(or their equivalents)
+
+jpeg_get_large, jpeg_free_large	interface to FAR malloc/free in MSDOS machines;
+				else usually the same as
+				jpeg_get_small/jpeg_free_small
+
+jpeg_mem_available		estimate available memory
+
+jpeg_open_backing_store		create a backing-store object
+
+read_backing_store,		manipulate a backing-store object
+write_backing_store,
+close_backing_store
+
+On some systems there will be more than one type of backing-store object
+(specifically, in MS-DOS a backing store file might be an area of extended
+memory as well as a disk file).  jpeg_open_backing_store is responsible for
+choosing how to implement a given object.  The read/write/close routines
+are method pointers in the structure that describes a given object; this
+lets them be different for different object types.
+
+It may be necessary to ensure that backing store objects are explicitly
+released upon abnormal program termination.  For example, MS-DOS won't free
+extended memory by itself.  To support this, we will expect the main program
+or surrounding application to arrange to call self_destruct (typically via
+jpeg_destroy) upon abnormal termination.  This may require a SIGINT signal
+handler or equivalent.  We don't want to have the back end module install its
+own signal handler, because that would pre-empt the surrounding application's
+ability to control signal handling.
+
+The IJG distribution includes several memory manager back end implementations.
+Usually the same back end should be suitable for all applications on a given
+system, but it is possible for an application to supply its own back end at
+need.
+
+
+*** Implications of DNL marker ***
+
+Some JPEG files may use a DNL marker to postpone definition of the image
+height (this would be useful for a fax-like scanner's output, for instance).
+In these files the SOF marker claims the image height is 0, and you only
+find out the true image height at the end of the first scan.
+
+We could read these files as follows:
+1. Upon seeing zero image height, replace it by 65535 (the maximum allowed).
+2. When the DNL is found, update the image height in the global image
+   descriptor.
+This implies that control modules must avoid making copies of the image
+height, and must re-test for termination after each MCU row.  This would
+be easy enough to do.
+
+In cases where image-size data structures are allocated, this approach will
+result in very inefficient use of virtual memory or much-larger-than-necessary
+temporary files.  This seems acceptable for something that probably won't be a
+mainstream usage.  People might have to forgo use of memory-hogging options
+(such as two-pass color quantization or noninterleaved JPEG files) if they
+want efficient conversion of such files.  (One could improve efficiency by
+demanding a user-supplied upper bound for the height, less than 65536; in most
+cases it could be much less.)
+
+The standard also permits the SOF marker to overestimate the image height,
+with a DNL to give the true, smaller height at the end of the first scan.
+This would solve the space problems if the overestimate wasn't too great.
+However, it implies that you don't even know whether DNL will be used.
+
+This leads to a couple of very serious objections:
+1. Testing for a DNL marker must occur in the inner loop of the decompressor's
+   Huffman decoder; this implies a speed penalty whether the feature is used
+   or not.
+2. There is no way to hide the last-minute change in image height from an
+   application using the decoder.  Thus *every* application using the IJG
+   library would suffer a complexity penalty whether it cared about DNL or
+   not.
+We currently do not support DNL because of these problems.
+
+A different approach is to insist that DNL-using files be preprocessed by a
+separate program that reads ahead to the DNL, then goes back and fixes the SOF
+marker.  This is a much simpler solution and is probably far more efficient.
+Even if one wants piped input, buffering the first scan of the JPEG file needs
+a lot smaller temp file than is implied by the maximum-height method.  For
+this approach we'd simply treat DNL as a no-op in the decompressor (at most,
+check that it matches the SOF image height).
+
+We will not worry about making the compressor capable of outputting DNL.
+Something similar to the first scheme above could be applied if anyone ever
+wants to make that work.
diff --git a/libjpegtwrp/testimg.bmp b/libjpegtwrp/testimg.bmp
new file mode 100644
index 000000000..8603d1548
--- /dev/null
+++ b/libjpegtwrp/testimg.bmp
diff --git a/libjpegtwrp/testimg.jpg b/libjpegtwrp/testimg.jpg
new file mode 100644
index 000000000..b34ca5d31
--- /dev/null
+++ b/libjpegtwrp/testimg.jpg
diff --git a/libjpegtwrp/testimg.ppm b/libjpegtwrp/testimg.ppm
new file mode 100644
index 000000000..9d81ce24a
--- /dev/null
+++ b/libjpegtwrp/testimg.ppm
@@ -0,0 +1,4 @@
+P6
+227 149
+255
+0/-0/-10.21/51.51.62/62/83/83/:3-:3-:3-:3-:3-:3-:2/91.91.80-80-91.91.:2/80-80-80-80-80-80-80-80-6.+6.+6.+5-*5-*4,)4,)4,)4,)4,)4,)4,)4,)4,)4,)2-)/*$/,%/,%0-&1.'2/(30)30)63,63,74-85.85.96/:70:7.A:0B<0D>2F@4IA4JB5KC6KC6MD5MD5OC3NB2OC3OC3PD4RE5R?1Y?2b@4nB5}E6‹H8™G9£F7¯G:¸G9¾E:ÅG;ÇG>ÊG?ËH@ÐE@çFLíCLëDKëEIîCIïBDñ>Bô=Aø;A÷:@ô:?ð<?é?@â@>×?<ËA7»=/µ@.µ@.´?-´?-³@-²?-¯@-­@,ªA,¦A-¢B,Ÿ@*›A)˜@*–A,”>-’?/’?/‘>.‘>,�=+’<+’<+”?+”?+”=*”=*”=*•>+–?,–@/–?6•>5—=2Ÿ?1©B3³D3¼D4¿D4¹?0µA2¬F8žH;‡H9oA2T8*C3&=5295495473271160050-50-72/72/72/61.61-50,50,41,//-.0-//-//-0/-0/-2.-2.-5,-4+,4*+3)*7(+=.1E69P:<jBC|IH�MM•OOŸW[ªdnªoƒŸt�”{£‡®€†º~ˆ½sy­`a‘TKvPDhSJgOG^MH^TQbfdo|}‚�‘�™ž˜£©Ÿ¢¨šž “�’�{|lfgUXWEQNEUR[UQbUQb0/-0/-10.10.40-51.62/62/83/83/:3-:3-:3-:3-:3-:3-91.91.80-80-80-80-91.91.80-80-80-80-80-80-80-80-6.+6.+5-*5-*5-*4,)4,)4,)5-*5-*5-*5-*5-*5-*5-*3.*0+%0-&0-&1.'2/(30)41*41*63,63,74-74-85.96/:70:7.@9/A;/C=1E?3H@3IA4JB5JB5LC4LC4MA1MA1MA1NB2OC3QD4P>0U?1^A3jC4xD6„E4’E5œC3§C4¯A4µA4¼B7ÀD:ÄE<ÅF=ÍC@áEIçBIèCIêDHíDGïBDó@Cö?Cø;A÷:@ô:?ð<?é?@àA>Õ@<Ê@6¹>/µ@.´?-´?-´?-²?,°?-¯@-­@,©@+¦A-¡A+Ÿ@*›A)˜@*–A,”>-’?/‘>.‘>.‘>,�=+’<+’<+”?+“>*”=*”=*”=*•>+–?,–@/”@5•>5˜>3 >1«A3µD4½C4¿D5»A2·C6¬F8œI;…G:l@3S9*B4)>63:6595484382271161.61.72/72/72/61.61-50,50,41,//-.0-//-//-0/-0/-2.-2.-3--5,-4*+3)*5)+<-0C47N8:d>=vEA†JI�NLšTV¤aj¥l}�rŽ‘{¢†€®…¹{„»ou©[[�QHuOCiOFeOG_PH_RN_[Yfnotƒ‡ˆ”™•™ž—š ”™���‘ƒ~ojkY][LVSJXSZVRaXQa/.,/.,0/-10.40-40-51.51.72.72.72.72.92,92,92,92,91.80.7/-7/-7/-7/-80.91/80.80.80.80.80.80.80.80.6.,5-+5-+5-+4,*4,*4,*4,*5-+5-+5-+5-+5-+5-+5-+3.*2-'1.'2/(30)30)41*41*52+63,63,63,74-85.96/96/:7.?8.@:.B<0D>2G?4H@5H@3H@3I@1I@1K?1K?1K?/L@0MA1NB2MA1QA1YB2dC2qC3|C2‡B2’A0˜<- :+§;.¯=2µ@6ºD:¿F=ÅD>ÙCEá@FãBGçBFêDFðCEôADø?Dú;@ù:?õ;@ð=@è@@ÜA=Ñ@;Æ@5·=.³@-³@-²?,²?-°?-¯>,­@,ª?-§@-¥@,¡A+�A,š?*˜@*•@+”>-‘>.‘>.‘>.�=+�=+�=+�=+‘>,‘>,’<+’<+“=,“=,”?+•?.•A6–?5š>3£>2¯A4¹C5¿D5ÁC5ÀD8¸F;®I=™J=�G;h@4Q:,B5,?74=77<66:4494183072/72/62/62/62/51.52-41,41,21,.0-,1-.0-.0-//-//-0/-2.-5//4..5,-4*+4*+9-/>24I56[97l?9|E@†ID�OM˜[`›fv”mˆŒwžƒ}­}‚¹u~·fm¤TV‰MEvLAkMAeOFcQHcMH^NK\[[eqty…‰ˆ‡Œ†Š�ˆ��…†Š|xzlfiXZ[MVSLZU[ZT`[S`.-+/.,/.,0/-3/,40-40-40-61-61-61-61-81+81+81+81+7/-7/-6.,6.,6.,6.,7/-7/-80.80.80.80.80.80.80.80.5-+5-+5-+4,*4,*4,*3+)3+)6.,6.,6.,6.,6.,6.,6.,4/,30+30)30)41*41*52+52+52+52+52+63,74-85.85.96/96->7-?9-@:.B<0E=2E=2F>1F>1G=1G>/H<.I=/I=/J>.L@0JA0KD2NE4UD4^D3iD2sB1~A/†?-Œ9)”9'�9*£<-¬@3³E8¸H<ÁF>ÒDCÚACÞBCâDCçCDìBCó@C÷?Aú;@ù:?õ;@î>@åA@ÚB=Í@9Â@3µ=-°@,°@,°@,¯>,®?,®?,¬?+©@-¦?,£@+ @*œ@+˜@*–@)”?*‘>,‘>.‘>.�=-�=+�=+�<*�<*�=+�=+�<*�<*’<+‘>,”>-’?-•A6–?5œ>2¦@4²B6¼C8ÁC7ÂB7ÂF<ºJ?¬L@—K>|F:b@4L:.A7-@85>96=77<74:5294183083062/62/62/32.52-21,21,12--2.-2./1./1.00.00.10.10.5106005//5,-4+,6,-:01D22T71c;3qA7{E;‚HD‰RU�_l‹i‚ƒs˜}y«x}µowµ`f¢QR‹LEyL@pL@hPEgQFfLC^GBVMLZ^^fjnquyxx}wz€vwzokoa`bTWYLTTL]WY]V]]V^------.-+/.,0/-10.3/,40-3/,3/,4/+4/+4/+4/+6/)6/)4/,4/,3.+3.+3.+3.+4/,4/,50-50-50-50-50-50-50-50-3.+3.+2-*2-*2-*1,)1,)1,)4/,4/,4/,4/,4/,4/,4/,4/,41,41,41,41,41,52-52-52-52-52-63.63.74/85096196/<5-=6,?8.@9/B:/C;0C;0C;.D:.D:.F:.G;-H<.I=/J>0I@1JG6MH5RG5YF5bE3jD1uB/}>,‚;)‹:)“:*š=,£B2¬F8²J=»I?ÌGBÔDCØDBÝEBâBBéAAð=@ô<>ù:?ø:<ô<>í?>áB>ÓC:ÅA5¹?0²?-®?,®?,®?,­>-¬>-¬>-ª?-¨>.¤?- ?,ž?+š?,—?+•>*“>*‘>,�?.�>-�>-�>-Ž=,Ž=,Ž=,Ž=,Ž=,Ž=,Ž=,�<,�>-‘>.�?.”B4—A4�@1¨@3¶A7¿C9ÅB8ÄA7¾C;·H?¦LC�J@tE;Z>2E9-<5+@93@85?75>63=52<4194083/62/43/43/23.32.12-12-02--2.,2.-2.-2./1./1.00.10.3205105104..2,,4+,7./=/.N5.Y9.e=1n@3sB;yKK€Zeƒg€p—zxªu{·ks´_d¦TT”OGƒLBwNAmNBhMAeJA`GBYGFXKKWMPU]cc`fbbia`f\Z`TW[MUXMXXP^YV`WX`WZ,,,,,,.-+.-+/.,0/-3/,3/,2.+2.+3.*3.*3.*3.*5.(5-*3.+3.+2-*1,)1,)2-*3.+3.+3.+3.+3.+3.+3.+3.+3.+3.+2-*2-*2-*2-*1,)1,)1,)0+(3.+3.+3.+3.+3.+3.+3.+3.+41,41,41,41,41,41,41,41,41,52-52-63.74/85085085.;4,<5+=6,>7-@7.A9.A9.A9.C9/C9-E9-F:.G;/H<.J>0HA1JG6IH6NG5VF6\E3dC2n@0v>-{<+‚;)Œ;*”=,œ@1£F5ªJ:´J=ÄH@ÌEAÑFAÖE@ÞCAä@>ì>?ñ:<÷;<ô:;ð<=é@=ÜC=ÍC8¾@2±>,®?,«@,ª?+ª?+ª?-©>,©>,¨?,¥>-¢?, ?,›>,—?+•>*“>)‘?*�>+�>-�>-�>-Ž=,Ž=,Ž=,�<+Ž=,‹<+�<+‹<+‹<-Œ=,�>/Ž?.”B4—A2 @2¬B5¹C9ÂC:ÅB:ÂA;ºA9±I@£NG�NEoG=S?4A;/96-@93A75?74>63<4194083/74/43/43/34/23.23.02-02-.3--3/-3/.3/.3/02/02/11/11/21/32032040/2.-1-,4..8.,G4-O4)X8+`<0e?6mGFyYd‚k…€uŸ||²w|¼nu»dh¯[[¡RM�LB~OArL@hI=cH>`HB^ECX@BO;@FBGJDMJJQJJQIIQFKQEOUIVWO^YS`YS`XU++++++,,,---/.,/.,0/-0/-1-*1-*1-*1-*2-)2-)2-)2-)2-*2,,1++1++1++1++2,,2,,1++1++1++1++1++1++1++1++2,,2,,2,,1++1++1++0**0**3--3--3--3--3--3--3--3.+41,41,41,30+30+30+30+30+41,41,52-63.74/74/85085.:3+;4,<5-=6.?6/?6-?6-?7,B8.B8.E8/E9-G;/H<0J>2H@3HE6GF4KE5QD4XC2_B2f?.n=,v=,|:*…9)Œ;*“=.›B2¢F7¬F8»F<ÂF>ÉF>ÐE>ÙD@âC?ê@@ð>>ò::ñ;:ì<<äA<ÖC;ÆD6µ@/ª=)ª?-©@-©@-¨?,¨>.¨>.§=-¥>-£=. ?.ž?-š?-–?,“>*‘?*�>)�>+�>-�>-Œ=,Œ=.Œ=.‹<-‹<-‹<-Š=-Š;,‰<,Š</‹>.‹=0Œ?/’C2˜B1¡A1®B6¼C:ÂC<ÄC=ÀC=¹FA²QJ¥XR�XQsRIWI>CC7<?6>93@72>71=60:5/94.83/63.43.43.34/13.13..3-.3-.3--3/-3/-3/-3/.3/.3/02/02/00.11/22021/0/-/.,2.-4/,?0+D0)K3)T8-Z<4dFFu]jƒs�‰€«…„¾~ƒÇtzÆmp½ce¯VSšLC‚K?qI<gG;cE>_FB]DBW?AN;?H:BE>HGDMHGQIGQHJRGNVKUXM^ZOaYNaXO++++++,,,,,,.,-/.,0/-0/-1-*1-*1-*1-*2-)2-)2-)2-*2,,1++1++0**0**1++1++2,,0**0**0**0**0**0**0**0**2,,2,,2,,1++1++0**0**0**2,,2,,2,,2,,2,,2,,2,,2,,3/,30+30+30+30+30+30+30+41,41,52-63.63.74/85085092,:3+;4,<5->5.>5.>5.>5,B8/B8.E8/E8/G:1I=1J>2I?3FC4FC4JB5OA4TA2\@2b>0j<-q<.w9*}8)…8(Œ:,•=/›B4¤B5²F:ºE;ÁF>ÊG?ÔG@ÞFAçCAîB@í;;ë;;ç>;ßB;ÑD:¿D4¯@-¤>(¦A-¦A-¥@,¥@.¥@.¤?-¤?-¤>/¢>.Ÿ@.œ?.˜?-•>+‘?*�>)�>+�>-�>-�>-Œ=,Œ=.‹<-‹<-Š=-Š=-ˆ<.ˆ<.ˆ<.ˆ</‡>/ˆ>1‹?1‘D2–C1¢B4­C6ºC;ÁD>ÁD>»EA¹PL²[T¥e\‘f]u_T[UIGMACI?<92?82>71;6094.74-63.43.43.34.24/13./4..3-.3-.3--3/-3/-3/-3/.3/.3/02/02///-00.22022010.0/-0/-3/,8,,;,)C0*K70S<6^IHtbn‡z”�Š¶Œ�Æ„ˆÏz€ÌrxÆik¶WWŸID„E=nG<dD<aC>^CAY@CV@DP>EKGQRKWUQ^WU`XS_UR^TT^SY_S^[LaZJaZJ,-/,-/,-/,.-------.-+.-+/.,/.,1-*0,)0,)0,)/+(/+(/+*/+*/+*/+*/+*/+*/+*0,+/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*0,+0,+1-,1-,2.-1-,1-,1-,1-,1-,1-,1-,1-,0,)1-*2.+3/,3/,3/,3/,3/,3/,3/,3/,3/,40-51.62/74/80-92,:3-;4.=4/>5.>5.>5.?4.?5,B5-C6.D7/F90G:1F<2G?4H@5J@6P?5T>3X<1^90b6+m9.t8-|8+ƒ9,�;/“=0˜?1ž>0§A3­A4µC8¾E:ÊG=ÔG>ÞE?åC@è@?êBAæDAÚE>ÈD8·B1ªA.¢B,¢A.¡@-¢?,¢>.¡=-¡=-¢>. ?/œ<,š=,˜<-•>-“=,�=+Ž=*Œ>*‹<+‹<+Š=-‰<,‰<,‰<,ˆ:-‡;-‰=/†</†</‡=0ˆ>3ˆ@2‰A5‹A4‘E5—D4£E9±I>ºG@»D>»EA¸MG´ZR¯f_£qf‘sh~rdjj^V^SIQFLLBJF=B>5<8/95,74+63,33+43.34.14-14-02-/1,,1+,1+-2.-2.-2.-2./1./1./1./1.02/02/11/11/11/11/11/40/4+0;/3A32C4/J;6]OOymy‹…Ÿ“�»”–Ï�•Ùƒ‰ÓtzÆjn·_b§Z[”LItHBdA>]>>X?BUAIVLU\U`bbqnn}xv†|rulyoguh_k_W_P]\Hb\Fc]G,-/,-/,-/,-/------.,-.-+/.,.-+0,)0,)/+(/+(/+(/+(.*).*).*).*)/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*0,+0,+0,+1-,1-,1-,1-,1-,1-,1-,1-,1-,1-,0,+0,)1-*2.+3/,3/,2.+2.+2.+2.+2.+2.+3/,40-51.62/80.91.:2/;4.=4/>50>50=4-?4.?4.A4.B5-C6.E80G:2H;3H>5H>5L=6O>6Q=4V;2Z90_7/h70p7.w7-�9-‰9.�<1–<1›=1¢@3§A3¬B5´D8¾E:ÉF<ÔE=ÛD=ßD@àE@ÞFAÔG>ÄF8²C2¥B-žB-ŸC.žB-Ÿ@.ž?-ž?-ž?-ž>.ž>.š=,™>,–=-”=,�=+�>+Œ>*Œ=,‹<+Š=+‰<,‰<,‰<,‡;+‡;-…<-†</…=/…=1ƒ=1„>2†@6‡A7ˆB6“G9—E7£G<¯J@¸IB¸GA·LF³TN±d\ªqfŸ~o�o|mmseZfZNXMLNAKI=EC7@=4=:188.44,11)23-23-03,/2+/1,.0+.0+.0+/1.-2./1./1./1./1./1./1.02/02/11/11/11/11/11/2015+49-7<23?53H?:^VTxr|Šˆž““¹”—Ê�–Ô„‹Ïy�Âqy¸kt­hnž\_€XZqSUjRWjT^hZgmfvvr‚tˆ~’‡ƒ•‰~Ž�v†yr€qfteZeT[ZE`Z@b\D,-/,-/,-/,-/,-/,.-------.-+.-+.-+-,*/+(.*'.*'.*',+),*+,*+,*++)*+)*+)*+)*-+,-+,-+,-+,-+,-+,-+,-+,,*+,*+-+,-+,-+,.,-.,-.,-.,-.,-.,-/-./-./-./-./.,0,+0,+1-,2.-2.-2.-2.-1-,1-,1-,1-,1-,2.-3/.40/51.80.91.:2/;30=31=4/=4/=4/?40?4.A4.A4.C60D71F93G:4H;5J;6K<7N=6P;6S:5V72[6.c60k6.t5,}7/‡9/�;0”<0–<1�?3 @4¢@3¨@3±C6ºD8ÅE:ÍD:ÕF@×H@ÔIBÌI?¾E:®C3¡B.œA.�B/œA.œ?.›>-›>-›>-›>/›?0˜<-–=-”<.“=.�>-Œ=,‹<+Š=+‰<*‰<*‰<,‡;+‡;-…<-„:-ƒ;-„<0‚<0‚<2‚>3ƒ?4…A8‡C:ˆD9”J=—H; H>¬KD²KF³LG²SM®`V­sg¦�q�Œz‘Ž{‚‰ws€ocqbXcUNRDMN@HI;DD8@@49;.46+/1&01)01)/0*/0*./*./*//-//-//-.0-//-//-//-//-//-//-00.00.00.00.00.00.00.1/26+97+98/4;63HE>_^Yzz|‹�š�”±’—¿�•Ç„ŽÃ}‰»{‰¶|‹²}Œ«}ˆšwƒ�q~‡n}‚n~~o‚~yŽ…ƒ™�‹¡”‘¨˜�¥”ˆ�Šƒ–ƒ|�{j{i\hTXX@]Y<_[@-.0-.0-.0-.0-.0-.0.......-+.-+-,*-,*.*'.*'.*'.*),*++)*+)*+)*+)**()*()*(),*+,*+,*+,*+,*+,*+,*+,*+,*+,*+,*+,*+,*+-+,-+,-+,-+,.,-.,-.,-.,-/-./-./-./+*0,+1-,1-,1-,1-,0,+0,+0,+0,+0,+0,+1-,2.-3/.40/91/:20;31<41=31=31=31=4/>3/>3/@2/@3-A4.C60D71E82F93H94I:5J;6L:6N94Q83T50^72e60o6/x8/‚90Š:/�<1”>1™?4›?4›?2Ÿ?1¥A2®B5¸D7¿E:ÇG>ËH@ÊJAÃI>¶F:ªB5žA0™@.™A-™A-˜?-—>,—>,™>,™=.—>.•<,”=,“=.�=-Ž=,Š=+Š=-‰<,ˆ;)ˆ;)‡;+‡;+„;,„;,ƒ;-€;,�;/€<1€<1�>5ƒ@7ƒC:‡D<ˆE<”L@–H<žG>§JC®LI®QL­]Vªj^§€o¡�yšš‚�›ƒ†•€z‹xm{jbm]SZJQUFKO@EI:@D6;=057,13(01)/0(./).-(.-).-)/.,0/-/.,/.,/.,/.,/.,..,/.,..,0/-//-0/-//-//-//-//-2-17,:6*83-1961HJ?bfX{€z‹�“�“£�–°Œ•¶ƒ�²��¯…™²�¤¶’¨³“¨«�¤£Œ ž‡ž˜‚™�}˜‰�œ‹ˆ£�”°š˜±›”­—‹£�…›„’|k|iXfOSV;ZV9^Z=+/2+/2+/2+/2-.0-.0......------.-+-,*-,*,+),+),+),*+,*+,*++)*+)**()*()*(),*+,*+,*+,*+,*+,*+,*+,*+,*+,*+,*+,*+,*++)*+)*+)*,*+-+,-+,.,-.,-/-./-./-./+*0,+0,+0,+0,+0,+/+*.*)/+*/+*/+*/+*0,+1-,3/.40/901:20;31<42=32=31<20<20=20=2.?1.?1.@2/A30B5/C60D63C84D95E:6G96H94K84N50X72_60i70r7/}:1†<1Œ>2�>2—@6—?5—?5—?3œ@3£C5«C6³E8ºE;½G=¾H>¹G=°D8¦A5›?0—@/—@-—@-–?.•>-”=,•<,–=/–=/•<.“;-’</Ž<.‹<-‰<,‰<,ˆ<,‡;+†:*†:*†:*ƒ:+ƒ:+‚:,€:.€:0€<1€=4?6‚B9ƒD=‡F@ˆH?‘KA’H=šG?£LE¨PL¨XQ©f]¦vh¢Œw›™€•¤‡Œ£‡†�ƒ~‘}t‚qjue\eRV_LNUCEL:?F6<@27:/68-01)00(.-(/,'/+(/+*1-,2.-1-,1-,1-,1-,1-,/.,1-,/.,2.-0/-2.-0/-0/-0/-0/-2-18,:5)51++66,GL8`hQzƒpˆ’‡�—–�š �™¥†–£‡›¢�ª©›¹± À³˜·¨š¶§›·¨˜·¥‘±œ‰«’‰«��±”’³–•´•�­�ˆ¢…„›�{’xgzdTbIQT9VU9XW;,03,03,03,03./1./1//////....../.,.-+.-+.-+-,*-,*-+,-+,-+,,*++)**()*()*()+)*+)*+)*+)*+)*+)*+)*+)*,*+,*++)*+)*+)**()*()*(),*+,*+,*+-+,.,-.,-/-./-./+*/+*0,+0,+/+*/+*.*)-)(0,+0,+/+*0,+1-,2.-3/.40/:12:12;31<42=32<21<20;1/=20<1/>0->0-?1.@2/A4.?4.@51@72@93A:4B94C84F74H5/Q51X5/a6/l8-u9.€</†>0‹=1“?5•>5“?5“?4•B4šB4 C4¥D4¬B5°D8´E:±E;ªB7¡?4š>1–>0•?.”?+“=,“=,“<+“<+”<.”<.“;-’<-�<.�;-Š;,‰<,‡;+‡;+†:*†:*…9)ƒ:)ƒ:+�9*9-9-€<1<3?6€A8‚C<…F?ˆIBŠICŽJA�G>—HCŸNJ§VR§`Z©pe¤ƒrœ”}–¡ƒ�¨Šˆ§ˆƒ „~–~yˆup{jdp\]iSR^JJS@BK:>E5:@29<134,22*1.)/,'/))0**1++2,,1++1++1++1++1++0,+0,+0,+1-,1-,1-,1-,/.,/.,/.,2,.8*75(13+(56&EK1\gGu‚d†“yŽ›Š‘ž”�Ÿš‹ž˜�¥——´ž¢Å¥©Î­¥É­¦Ê°§Ë±¤È¬—¾Ÿ‹³‘ˆ°��²�Ž²Ž�²��¬Š„¡‚€™{wŽrdx]Q_ENR7QQ5SR6,03,03,03,03./1./1./1./1/////////////.,/.,.-+.-+/-..,-.,--+,,*++)*+)**()+)*+)*+)*+)*+)*+)*+)*+)*+)*+)*+)**()*()*())'()'(+)*+)*,*+-+,.,-.,-/-./-./+*/+*/+*/+*/+*.*)-)(,('0,+0,+0,+0,+1-,2.-40/40/:12;23;23<34=32<21<21;10<1/<1/>0/=/.>0->0-?1.>3/=60;60;83<94=:5>93@72C60G4.O4+Y4+d5+n8,x:-;.…;.�<4‘<5�>3�@3�A2“B1—A2™@0 >1¦@4ªB7ªB9¥@6Ÿ>5—=2•?2’?-’?-‘>,�=+‘;*‘;*’<-’<-‘;.‘;.Ž;-�;-Š;.ˆ:-†:,„;,…9)…9)ƒ:)‚9(‚9*�9*~8,~8,€<1€=4€@7�B9„E>…IAˆKFŒMFŒIAŽH@˜JF¡SO¨]X©hb©{n¤Žy™œ�’§†‰¬‹‚ªˆ€¡„™€~‹wx�njxadr[ZhQQ]IITCCL;>D6<?467/44,30+0+'1()1()2)*4+,2)*2)*2)*2)*2)*0**0**0**1++1++0,+0,+0,+0,+0,+1+-6)25)-4-%46!AH&Wc;q€W‚’m�œ}� †�¢Œ‹¡Š�¨‰•¸Ž È”¤Ð�©Ó«¨Ò®§Ñ­¡Ë¥“½—…¯‡‚¬„‡°†ˆ®…Œ±ˆ‹«†�Ÿ{{”tqˆk_sWM]BIO3JM0KK/./1,01./1./1.0/.0/.0////00.00.00.0/-0/-0/+/.,1-,2.-1-,1-,0,+/+,,*++)*+)*+)*+)*+)*+)*+)*+)*+),+)*+),+)*-(,*(),())'()'((&'*()+)*,*+,+)-+,.-+/.,/.,-,*-,*/+*/+*.*)-)(,(',('1-,2-*2-*2-*3.+4/,50-61.;31;31<42<42=31<20;1/;1/<1/<1/=/.=/.=/,>0-?1.=2.=4-=6.;819919:29:2;81?6/C2+J1,T2)^3*g7+o9-u=.z</‡=2‹=3‹?2Œ@2Ž@3�@1’?1•=/˜</œ>2¢B6¡C7žB7™?4–>2”>1”>/“=.�=-�<,Ž=,Œ=,Œ=.�>/Ž<.Ž<.‹</Š;.ˆ:-ˆ:-†:,„;,‚:+‚:+:*~;*~;+|:*}9,|:.€>2�>5€@6‚C:ƒG=…JB‡NE‹OGŒICŽID“PJ˜\Tžg`�uiž‡už˜€”¡…�ª‰‰¬‹ƒ¨‡€¡„~›z�ytˆoi}bby]\pUUgOO\HGRAAH8=A388.85.7/,3+)2()2()3)*4*+0*,/+*0*,0**0*,0*,2),2),3*-1+-1+-1+-0,-0,-0,-1+/4)/4*+4-%46!?F%T`8m|Q��g‰™t�žz�¡~ˆ¡zŠ§{‘¶‚›ÄˆŸË�ŸÉ™žÇ��Æš˜Á•Œ·Š‚­€…®‚Œµ‰„«‰®…‰©‚~œxv�olƒfZnSJZ?GM3FJ1DF./0+.0+/0+01,01,01,12-21-32.43/43/62/51.41,3/,4/,50-50-4/,3.+2,,1-,0,+0,+.,-.,---/--/,,.++-*).))+.)/.)-/(/.)-/)-.)-.*+.*+/+*/+*-,*.-)--+-.)./*./*------.,-/-./.,0/-2.+2.+2-*4,)5-*6.+8/*:1,:1*;2+=4-=4->50>50>50=4/<3.;2-=2.<1-<1/;0.=/.>0/>0/@1.A0)@2);4*77-39-28.56.:3-?-)F*'L)'S*&Z/(`5,d<0k@0yA0@/†A1‹B3�@3“=0“;/’8-“;/“=.”A1•B2”C2’A0‘>.‘;,—;0—:2’;1Œ=0ˆ?0„?/‚?.ƒ>.ˆ@1ˆ?0Š>1Š<0‰;/ˆ:.†:-ƒ;-{9+~@3w<.q7)w>-w>-v;-}?2{;/‚@4‡E9‰I?ˆLA†MB„PC‡NE‘KI—SP”bY‘paŒ}j‰‰q‰–|Šž‚�¤‡�¦‰Ž§Š�§Œ‡¤ˆ~Ÿ‚uš{o—ui�ib„_[zXZsUZnSWeNPWEJK=C?6@93;0.6**3')3'+3'+1&*,*-)+***,*(),'+.(,1(-2'-3(.3(.3(02(00)10)1/*1/*0/)-1++0-(//#<?*V\@pzX‹g�œu‹�sŒ¡vŒ¦wŠ¨v‰ªuŽ´{™¿†ŸÁ��¾�–ºŠ�³ƒˆ®}…«|ƒª{…ª~ƒ¨}ƒ¦~€ž|y”ur‰mh|cYkUMZFIQ<CH4?B//0*01+01+01+12,12,21,32-43.43.74/74/63.52-50,50,7/,7/,6.,6.,3.+2-*0,+0,+0,-/-./-0--/-,1+*/)*.)(-.)//(//(//(//)-/)-.*+.*+/+*/+*.-).-)./*./*./)./*.....0....../.,/.,1-*3.+5-*5.(7.)8/(:/);0*<2)=3*>5,>5,>5.>5.>5.=4-<3.<3.=2.<1-;0.;0.=/.=/.>0/@1.C0*C0)A2+>4+:6-95,:3-<1-?-+D*)I*(N+'T/)Z5-_:1e>/pA/xA-€A0ˆ@2�?4‘=3“:2“;1�9.‹</‹>.Œ@0�A1�A1ŽA/‘?1•;2•;2�<2‹=0†>0ƒ@/�?/‚?.…@1†>/‰=0ˆ<.‡;.„;,‚:,€;,x8,x<1s9-n9+s>.s>.r8*u:,}=1ƒA5‰E:‹I=‰K@†NA„PC‡NE‘JH—SR’f]Œwf„†p€“w�ž„¥†ˆ§ˆ�§Š�§��§�Š¤‰�Ÿƒwœ}o˜ve’i`‹`Y‚ZY{X\wXZnSSaJNUCFH;C@7<737/-3*+2)*1(+.(*,*+**,+),+),-(,/)-2(02(02(02(02(10)1/*1-*3-*3-*12-12.-0,)--%8:,SXBox]€‹i‡–o†™l†žnˆ£p†¤p„¥pŠ­w“¶€™¹‡–¶„“³�Œ®{†«x‚¨w‚¨yƒ¨|ƒ¦|‚¤�{x’uqˆnh{eYjWMZHEL<@D5;=/12,12,12,12,23-23-43.43.54/54/85085085074/72.61-80-80-7/,6.+4/,3.+2,,1-,1-./-.0.1/-0--/,,.+*/+)./(/1'//)-/)-/)-/)-.*+.*+0,+0,+/.*/.*/.)0/*0/*0/+//-///0./0./0/-/.,1-*2-*6.+70*90+:1*<1+=3*>4+?5,?6-?6-@7.@7.?6/>5.=4-=4-=2.<1-;0.;0.;0.;0.<1/?1.C2+E0+H/+K--L,/K+.I*/E+.A-,@.*A.(E.(M.)X1*b3-g5,j:,o;-w;0�=4‡<6Œ<5�:4�:4Š<2‡=2†>2…?3†A2‰A3ŒB5�C5Ž?2�=/Œ>1Š>.‡>/…>,†=.ˆ<,‰;.‰<,ˆ:-†:*ƒ;,<)~=+{<+}=1z<1v:/u;/x@1x@1v<.v;-?5‚B8‰F=‹H?‰JA‡KA‡NCŠNF‘JH˜TQ–f\�ve‡…n‚‘t�{�¤ƒ…¨‡ˆª‰�ªŒ�©ŒŽ¥‰ˆŸƒ�š}y˜xi–ma’e\‹a\†`]‚`[|]UrVPhPFYEBP?:D93:2.3,*/)*,)**(0'(1&*1&*1&*0'*1(+0*.0+//*.-+.,+0+,0)-0(-1(-1)-01/23/..*'**"57*PUAmv[|Ši€�k~“j™jƒŸn€ n~ m€¥q‡¬x�¯|‹­zˆªx„©v�§v§u¦w¦y¤y€¢}}›yw‘tp‡mh|cYkUMZHDH9=?299-23-23-34.34.34.45/54/54/650761961:72:72961940940:2/91.91.80-50-4/,4..4..3/03/01/01/20.1..0--/-+.0)02).0*.0*.0*.0*,/+,/+*1-,1-,0/+0/+10+10+10+10,11/1111/010.10.2.+3.+3.*92,92,;2+<3,>4+@6-@6-A7.B8/A8/B90A8/A81@70>5.>5.=2.=2.<1/;0.;0.<1/=20=2.B3.E2.L..R+0V'/U&.P'/I).C/.<1+;2)?2)G0(R/)_.*d/)i9/k;/u<3}<6…;8‰;9Š;7ˆ:6‰>8†@8†B9…B9…C7†B7‡A7‰A5ˆ@1ˆA/‡@.‡@.‡>-ˆ<,ˆ;+‰:+Š8*ˆ9*‡:*ƒ:)<){=(x>(x>*‚>1{7.z7.z<1v;-w=/|A3{@2€B7ƒE:†H=ˆJ?†J@‡MBˆODŠQHŒSJŽ\QŽh[‹tb…g€‹m~–vž|ƒ§ƒ„©‡‡ª‰‹ªŠ‹¦‡‡ ‚ƒš~}˜yq•oi”if�gfŒeg‹gd†e_~_Zw[PhPK^JBP?8D62:/.4*,/(+*%2&&5%(4%(2&(1'(/)+/+,.,-+++*+-*+-(,-(,/',/',/*+-.*+1++0+'.+"88,PUAiqYv„c{Œhy�h|–iœl}Ÿmyžkz m¥r‚§t€¥q}£p}£p}¥s~¥v~¥x}¤x|¡x}Ÿzz™wu�qn‡je|bWkRN[GDF9?=1:8,45/45/45/560560560761761872872;83<94<94<94<73<73<41<41;30;3083072/61.61.5106216213122011/00./1-.2,04+.4+.4+.2,.2,,2,,2,,3/.3/,3/,3/,21,21,32-32.32032032032051051.61.61-;4.<5/=4-?6-A7.B8/E8/C9/D:1D:1E;2D:1C90B8/@5/@5/>3/>3/=2.=2.=20=20>31>31@51E31M02T,4X)3W(2R)1K,1B30:6-77+:6*B4)M2)X/)^/)f:1j;3s<7z=:‚<:†<;‡;;†::‚;7�>8�A8�B9ƒC:ƒC:…@9„@7‚C2‚C1ƒB0„?/†=.ˆ<.‰:-‹9-‰7+‡8+…9+‚:+~=+x>*v?*x>*�9-|/'‚8/„>4w4+s5*}A6}C7E:€G<‚I>ƒJ?„KB†MD‡QG…WJ|aNzjQ€pY�u]‚|bƒ†i†“uˆ�|ƒ£~�§‚‚ª…‚¬†�©„¥€{ž}z™wz’pz�lx�lwŒku�mr‰lm„gkd`rX[iRR[HHL=@@4;7,70&5*$6('5''3''1'&.)&+*(++)+-*',('+*)+*)+***,+),+),-)*,#$2*'50*86*BC5UZFfpWn}^tˆeqŠbu�ex˜iw›ks™htšiwŸkz¢nx lwŸmx ny£s{¥u{¥w|£wyžuzœww–tsŽom†hd{_WkPN[GCC7>:195,560560671671782782872872983:94=:5>;6>;6>;6>95>95?74?74>63=52;63:5294194184184395484373243151240/6-06-.6-06-.4..4..4..4/,40-40-40-51.32-32-43.43.43/431542540841850940:5/=60>7/@7.A8/C90D:0G:1H;2F<3F<3F<3F<3E;2C90B71A60@51@51?40>3/>31?42@53?53@72C52I35P16T/6S.5P05J22C52=90<:-=:+C7)I6(Q3)W2)]2+d3,l50v64}77‚87ƒ77‚66~75}:4}<6}>5€?9ƒ@8†?9…A8€B3€C1�B1ƒ@0…=/‡;.ˆ:.‡9-…9,ƒ9,‚:,<,|=,y>,x?,|=,‡5*‹2,¡LE§XQ‹A8|90‚F;€K=yH:zJ<{M>|N?}OB�QE‡UJ‚_LrkOosR|uX‡w]�yb˜iŸ‹s ”z––z��|‹£ˆ¨�…§‚€¤~{Ÿy~™vƒ‘p…ŽoƒŽp€�p�p{ŽpwŒms‡kj~bfv\_hSV[GOM>GA3@6*=0';,'9+(6+'3+(/,',-'+.').().(+-(-,*/+*3)*4(*7'*7'(3($<3,E>4IG8QR@]bKgqVjyZn‚]k„\l‰]p‘bq•eo•do—eršfuŸmrœjq›itžnx¢rz¤vy¢vyŸvvštw™vu”ro�kj…fc|^UlON\ECC7@91;4,671782782782893893983:94:94;:5>;6?<7?<7?<7@;7@;7B:7B:7A96@85=84=84<73<73<73<74<74<74;639529338308/09/.8/080.80.80.61.61-61-61-52-52-63.63.74/74/540540841952:63<94=84@93@70A8/C90D:0G:1H<0I=1I=1J=4J=4J=4I<3F<3D:1B8/A7.A81@70@72?61?61@72@72A83?74@85B86D97G96H96H96H94E80E8/E9-E9+G9,I9*K9+Q7*Z/&d/'n3-z63ƒ98‰;9‹;:‹=;‹A>‡@:‚=6<3:3‚<4ˆ=7‰@7ƒA5�B3‚@2ƒ?2ƒ=1„<0�;/€</~<.|=.{=.|>/|>/}>/=/†9/�1+¢<8ÍlfÙ~y­ZTŒC<ˆLAN@tJ<qM=pQ?qR@tS@zWD‚[J~eOmsOo{U„y[˜u_©oc¶mf¾qk½wo»‚w±Œz§—~žŸ€–¡��¡~…�y…™vˆ‘r‹�rˆ‘r„“r€–r|–sx“pt’pm‹ii„edx\]kQV^GMP;ED0B;+@3+?2,;0*70*30)00(./'./)/0*2/*6,+:*+>(+C&+E$+C&(F5-LC4VQ>[YD`bJgmQiwVj{Wl‚[g‚Wf†WlŽ\o”an”an–bršfs�ko›jo™irœnw ty¢xxžwu›vs—su—vs’rn‹li„cb{[TmMM]CGH:E<5@707827828938938939:4:94:94;:5<;6?<7@=8@=8@=8A<8A<8C;8C;8C;8B:7?:6>95>95=84>95>95>95>95=85<73:51;30:0.:0.91.91.91/91.91.72.61-61-63.63.63.74/74/74/540651952;83<94?:4B;5B;3A8/B:/C9/E;/H<0I=1J>2J>2K>5K>5J=4J=4F<3E;2C90B8/B92B92A83@72@72A83B94A:4?82@93B;5D=7F=6G<6K<5N;4M6.N6,Q6+Q6+Q7*P9+P9)V6'f6*r6,~;3‰@9•D@›HDŸJGŸLF QJ™LDŽD;…;0„7-…7-Œ91�=4ˆ>5‡>7†=6…<5…<5�=4}=3z>3x@3vA3x@3z>3<3ƒ:3ˆ73‘31˜(&³=;ì|zý•’Åhc–G@‰K@xH:nM<jQ=fT>hV>lX@t[E`L€hPysQ„wUžt\´l^É__ÓV\ÙQ[×T\äouÙzx̆|À�~µ•€«•}£’xŸ�vœŠrœŠt™�u•�u‘‘u‹’s†‘s‚‘r|‹ny†ju{amoWgbN_TBUE5R</O4-M1-I0,D/*>/(9/&7.'6/'81):/)=.+A,+F)+H(+K'+H)'TB4YQ<d^FgeLilOnuVm{Xl~Xk„Zg„Vg‡Vm�]p•bq–br™du�it›lo™ko–jr™mw�vyŸxwšyt–us”uu”us�qoŠki„ea|[TmMM]BMN@KB9F=69:49:49:49:49:4:;5=<7=<7=<7=<7@=8@=8A>9B?:D?;D?;E@<E@<E@:D?9D?9C>8E>8D=7B;5B;5B;5B;5B:7B;5A:4A:4A83A83A83@72@64>71>71>71<71<71;60:5/85.85.74/74/961961961:70<71=82A:2B;1C:1D<1F<2J>2K?3L@2N@3N@5N@7N@7L?6K>5I<3H;2E;2E;2B90A8/@91?80?80?80@93@93<5/MD=M@:K:3T@9R62O0+a<6\3-`5.`4+^/'^/%f6,m=1q=0‰I=�G;—I?ŸKAŸF>›>7š=6žD;¦ND¤PE¨VJ«YK¥OB™A5—;0›>6“98‘98Œ65†52†84„?8|@6r>1rB4oA2q=0v:0‚72Œ43’-1š',ÈHIèbaÅEDÍWUáyv«VO†F<„VFmR=i[AibFjeHj`En^D|dL�jT’bL²m]ÑnhÛX]à;Lç-Eò+Hõ0Lö@YìI\ä[eånrâ{|Û€}Ûƒ؉‚Ãv½�wº„x¶†x²ˆx­Šw¨‹y¥Œxš€o™~m˜xi“oa�bY„UM{IBxA>u:<i.2d02a11P&'G%#L0-M51G4.E2,D1-F1.F.,G+*K--Q6/YH4`W:f_BgdEkoNu|Zu„]o�YlƒWl‡Zp�]r’as•bu—dw™fv›hr–js˜os—qq•os•tw™xz˜|y–zx•ys�rm‡jk„fi‚dazZTmMO_DNO?SK@PG>9:49:49:49:4:;5;<6>=8?>9>=8>=8A>9A>9B?:C@;D?;D?;FA=E@<E@:E@:D?9D?9E>8E>8E>8E>8E>8D=5C<6C<4D;4C:1B92B92B92B92A83A83@93@93>71<71;60;6096/85.74/74/96196/96/:70<71>:1A:0B<0C;0E;/I=1J>0L@2L@2N@3N@3N@5N@7M?6J=4I<3H;2E;2E;2C:1B90@91@91@91@91A:4B92D93I81L/+V.,j76u99{;;…ECƒD?„H@…IA†G>ŽIB™NH¡PL¢OG“D7”B4šB6ŸC8 >3œ7-œ7-Ÿ=0£C5›?0™@0 E3¤G6§F6®H:¶NE´LM¯HL¥@D™9;‘98�>:ƒ@8x>3o>0zI;…LA„?8„,+’).³9DÒLWÜKNßMMÊ>=À@?Ð`\³YQ„@5‚TDyaIe[@^]?ihIslOvgJ~dK–cN¼dXÖ_[æUZëANò,Eþ%Dÿ&Hÿ'Iÿ.Nù3Nò<RðIZëP^åQ]ãS]àXbÛbiÕflÔjnÒmqÏqrÍutÉxwÇywÊ||ÉyzÈvxÇqtÆkrÅfnÅakÃ_kÅ`n³S_¥KUšGQ�DK†EIm69HT1-M0*H1+J6/K81K:2O>6UE8[O5cZ9gaAifEnrOz‚]z‰bt†^r‰_p‰_p‹`qŽ`u’dw”dw”dw”fs’ix–rz—xy–x~˜}ƒ�„…�‡ƒ›…�™ƒz’zr‰om„hh�d`y[TmMNaERSAVPBUOC8938938939:4;<6<=7?>9@?:@?:@?:C@;C@;C@;C@;D?;D?;FA;FA;FA;E@:E@:E@:G@8F?7JA:I@9H?8G>5F=6E<3F<3E;1D:1D:1D:1D:1D;4D;4D;4C<4?80?80<71;60:5/:5/94.94.96/96/:5/;7.>7/@9/B90C;0E;1F<0I=1K?1M?2M@0NA1M@0P?5M?4L>3K=2I<3H;2E;1D:0C:1C:1A:0A:0?;2@<3@<3D;4UD<R3.j23ŽAG­LWÃU`Ë]fÆ_b›B>—I?–LAœNB¬RJ¾URÌOSÄLK¥@4�B0 @0¥?1¥;.£7+¤8+§=/«E6¡>+˜7$�:'¥>-¬A/²C2¶D:§02¯:B¹DL½LRºMR­KL–A>‚71{8/v6,x2*„1-œ37¹=GÕGWèM[âFIÏ53Ð:9½31ÈNKÆc]”J?Œ[J}cLj_CgdEtoOpQ�hJ�hN°hYäbbõQZôDQö7Iý1Jÿ2Nÿ/Mÿ*Hÿ2Oü0K÷1Jõ5Lò4Mð2Kó3Ló7Pò@XïC[îF]íIaëNcêQeèUhèVkåSjäRiãOiäNiçNlëPoòSsóVuþh…ña|å[tÕUlÍYlÉdr¨R]r-2^&'Y0,W:4Q?5B:-:8)>B1LN9VO3aV6e_?heDquT†e~ŒiwˆdxŒiu‹er‹dr‹cvŒey�hz�iz�j€•v„›~Šž…� Š’£�—¨–˜©™–¦™‘¢’‡˜†z�ypƒmg~b^wYTmMPcGSXDXUDYVG7827828939:4;<6=>8@?:A@;BA<BA<DA<DA<DA<DA<E@<E@<GB<GB<GB<FA;FA;FB9HA9HA9LC<KB9JA8I@7H>5G=3F<2E;1D:0D:0E;1E;2D;2E<3E<5E<5@91?80=82<71;60:5/:5/:5/96/96-;7.;7,?8.@:.C;0D</F<0F=.J>0K?/M@0M@0M@/M@0O?2O>4L>3K=2J<3G:1E;1D:0D;2D;2B;1@<1@<3A=4B>5F<3S81[*&‰:?¾T`ÙTeæUfãUcËJOŸ30–:/‘>0™A3®F=ÅIGÕBHÍ>@±B7§F5©B3¬@3¯?3°>3³B4´D6²G5©B/£<)¤;(©<(«:(®6&­3(¶97º9=½7>¼7<¿<BÆJLÄTS¾XS™?7Œ4*‰,%�61ÃHKÜOWâAQÛ3@Þ:9Í.*Ó84È74ÆE@È]U©WK�WFwW@veI~pSpQ„fJ�dK©o[ÕuiîSWüDP÷;Jø9Iþ=Nÿ@Qÿ:Lù4Hÿ?Sþ9Mú4Ký3Ký1Jÿ/Iÿ0Nÿ6Sÿ3Tý4Vý6Wú7Wø9Zø=\÷>^õ@aøCfö@fõ?eõ?gø?hüCmÿErÿIsÿOuÿQuþUvòSqçTnåaxÖfv·Wb}15j0/X1*P8,G>/>A.;D/?G/PK._T8g`CokN||`ˆŽr‡‘v}‹qz�q|�s|�sz�ozŠm}Œoƒ’uˆ—z•£Šœ©•£®�¦±¡¬´§±¹®°·°«µ­¡«¢” ”ƒ‘‚s„qh|c]tXSlNPdHRZCWYDYZH671671782893:;5<=7?>9@?:BA<BA<DA<DA<DA<EB=FA=FA=FC<FC<GB<HC=HD;HD;JC;JC9LC:KB9KA8J@6J=4I=1H<0G;/E;/E;/E;/E;1E;1E;1D;2D;2@9/@9/@91?80<71;60;60:5/;7.;7.;7.<8-?8.A;/C;0D</G=1G>/K?1M@0M@0NA0O?/O?0O?2N=3N=3J<1I;0G;/D:0D:0D<1D<1B;1@<1A=2A?3B?6K=4\5.w32­LSÛ]iæM_å@Q×6E·&+¨1)�7(’:&—<)¬@3ÃD=Ô>?Í;;±?5¨B4ª@2­?2²@5¸C9¼H;»J<¬>/ª?/¨=+¦;)ª9)°:,·;/¼<1ÊFAÎDAÐ>?Ð79Ö7<ÞBEâLNÝROÄC>ÆKCÍTLÔSMÙKJÛ?Bà5>á27Û4.Ú7.Ð1+Ï:4¿84³@9µ[P–UC{N9‚bI†kP„`F’[F­fTËrdçmhêEKò:Dê9Cç<DçBHçFKèGLêFMôJSòBMò9Iö5Hù2Gý/Fÿ2Lÿ8Rÿ9Vÿ9Xÿ:Xý:Zþ;]ÿ=_ÿ@bÿAeÿCiþDkþDmýBmüCoüCoüBqýBoÿGpÿ@hÿGmÿMpòIhéOkå[rÙcs½YcŽ?D`&$Q+"V@3VO=IL7@D-RJ3eYCujV€{gŽ�y™œ‹•œŒˆ“ƒ‚Ž€‰—ˆ�›�Œ™ˆ‡’‚Š“‚—žŽ¤¨™²¶§¸¼®¿Á¶ÀÁ¹ÂýÅÆÁÃÃþ¾¾°µ±¢©¢�—Œy‡vi{e\sYTmORfJQ[BUZCW\F560560671782893:;5=<7>=8@?:@?:C@;DA<EB=EB=GB>GB>GD=GD=HC=ID>IE<IE<KD:LE;LC:KC8LB8KA5L@4K?3J>2I=1G=1G=1F<0F<0E;1E;1D;2C:1A:0A:0@91@91=82<71;60;60;7.;7.;7,<8-?9-A;/D</E=0H?0J>0L?/M@0NA0NA0O?/O?/O?2N=3M<2L;1I;0G;/F90C9/C;0B<0B;1@<1@>2A?3B?6N;4m84—FEÄY_ÛWbßCQÞ8FÔ3;½++­1'¡8%˜:!—9 ¤;&¶>.Ã:2¾71§7,¢:-¥9-§7,¯:1¸B8ºE;·E:®>2­?0§<,¤6'ª6)¶>0ÃD;ÊG=¾8-Î@6áD?ìBBô=Aó9>ë27Þ.0Û75âGBèSLåNGÜ=9Ù10ã/2ê67Ù1(Þ</Í,"Ð71¾3,°7/Ég\¸l\ŽWC‡ZCƒV?‰R=©\LÎlaágbäTTðEKñ=Fä;@Ù=>ÒA>ÒGBÝOMëWWíNRïDMð:Gö5F÷2Fø.Dû1Gþ7Nþ8Sý8Tý8Vþ9Wÿ;\ÿ>aÿAeÿCkþ;eý<hý>jý?mú>nø>oô=mö:jÿHrÿ;aÿ?dÿKnÿKn÷NmïUoãZnèr‚Âaj˜ILt:8\6-P9+PC2UM:^QAreUˆ|n•�‚Ÿ�‘¦§Ÿ¡¦Ÿ–œ˜•ž›ž§¤¦¯¬¦¬¨¢§£¦¨£µ´°ÂÁ¼ÌÉÀÑÎÅÕÐÌÕÐÍÔÎÎÕÏÑÑËÏËÆÊ¿½À°²¯˜Ÿ˜€Œ~m~k`t[WnRVjNS`FV`EX`H560560560560671893:94;:5=<7>=8A>9C@;DA<FC>HC?HC?GD=HE>ID>ID>JF=JF=MF<MF<ME:LD9MC9MC7NB6MA3N@3MA3JA2JA2I@1H?0G=1F<0D<1D<1C:1C:1A:2@91?82?82<71<71<8/<8-<8-<8-@:.A;/D</E=0H?0K?1M@0NA0P@0P@0O@-O?/O=1O=1M=0L</I;0F:.E8/C9/B:/A;/A:0?;0?=1@>2@@4P91r1/¨JKÁSV»>D½06Â03Á//º1)©1!£9#œ=!™;›< ¡=#¥;%¤6%ž6) 8-¢6,¥4,­81·@:¸C<²@6±B7­A4¨</¦8+¯:0¼C8ÃD=Å@7Ã>-Ï>-Ù5+á*&í"%ú%+ÿ*1ÿ.4í)*ç.+á3,ß4,à3,ã2,é0-ç2+Ú2%Ö6&Ð2&Ï7,Å6.½>7È`UÒq¦eS�VBŠM:£WIÈf[ßd_çRTèCIõFMí@Dß??Ô@<É@8Æ@7ÑFAãMLïJNô@Kù9Hý7Hü5Hø3E÷6Gú;Mú8Pü7Rû6Rý6Uÿ6Xÿ9]ÿ;aÿ<gÿ>kÿ?mÿ@qÿBtÿBuþBuüBtùBpÿFmúAaþEeÿKjþKkÿVtÿ_zõ]tòj~æp~Óow®^aƒCAg6/hB7sUJye\Œ}v£–�­£¡²®«¶¶¶²¶¹ª¯³°·½¶½Å½ÁÊÀÃÊÃÂÊÉÆÍÕÎÖÝ×ÛâÚØçÝÛéÝÝæÚÜåØßä×ààÔÞÙÏØÊÅ˺ºº¢§£ˆ’‡t‚qexb\sW[oS[jM\iK]gL561561561560561671872983<;6=<7@=8B?:DA<FC<GD=HE>HE>HE>HE<JF=JF;KG;KG;NH:MG;MG;ME:MD;MD;MD=NC?NC=MC7MD5KB3JA2H>2G=1E=2D;2C:1C:3A:2A:4@93?74<73<71>7/>7/<8/=90>:1A:2C<2D=3H@5I?3L@2MA1NA0P@0O@-O@-L?/L>1M=0L;1J91I81E80C90@9/>:/;;/<<0=>0>?1A?0S8-‡<7Ä\[ÊZY®86«.*²1+®.%­2#¬9&¬B*ªF,¦D'£A&£A&¢<#Ÿ9# 9*£;0¦:0¦7.­<4¶C<·E;°@5«=0¨</©;.¬<0»E9ÆLAÅE<¼8,¿6$Í:(Ý;.æ3,ñ+*ù(+ý&+û%'ø**õ.+ï2,ë4,è3*å3)ä2(à3%à:*Ñ2Ü?0Î6)È8-Â?5³A6Ñl`¿gY¥RB¨OAËcZçhbæRRêAFóBJë>Dá;=Ö<<Í@9Á>4¶8,¹6.Ä94âHHé@Eð;Dö:Hú:Iø8G÷;Iû?NþAUÿ?Vþ<Tþ9Tÿ8Uÿ9Zÿ:^ÿ;cÿ>iÿ>lÿ>oÿ>pý>qú@rø@t÷Aq÷CjÿMmÿSs÷Jhë@`ýTsÿg…ÿe‚÷]wï_xçh{Üp}Æqx§gg�^X{ZQ—~wª™’¾°­Ä¼ºÆÂÃÈÇÌÇÇÏÃÃÏËÊØÌÊØÑÊÚ×ÎßßÔäæÛéêàëíãëïåæòèæôèèðäæîáèíàéêÝçäØâÑÊÑÁ¿Â««©‘–�z†xl{hbu_`rXbpVboScmT21/320431651875984984984;:5<;6==5??7AA9CC9EE;EE;HH@HH>HH<JH;KJ8LK7ML8NK8MJ9JF:GD;JFCNIMNIPGDOD?ENE>LC4J@4KA5MC9LB8F=4B90E<5C:5@93?74@85?75<74<42C:5B94A96>95?:7>:7?;8@<9B?:D@7G@6J@4L@0O@-O@+L?,G@.F?/I;0K81M53K65J88F;9B?:6904</9D4:A/69$?=(bC1¼j^¹JA¦7,¥5)§7+¥5'¥7&¨=+£8$¥<&ª@*­@)¬=)¬9&®6%«6%¤6'¡7)¢8*¦<.ª@2¬B4¬B2¬A1©>,«=,­<,³=/½C4ÄF8À>1º2&Ç9+Ì8*Ð8+Ö8-Û7-â5.é3/î1-ñ.,ô.-ó0,ñ1,ê5*ã7)Ý:)Û:(Ü9&Ô3Ð3 Ï8'Æ7'¾6(ÂB5ÏSIÈNCÏSKá\Wî]ZðPRí>Cò9?÷?GêAFßCDÕAAÈ?9»<3±;/°</²<0ÊL@ÔKCÜGCá>?ë<A÷?Gý?Iù;GþDQüCQþDRÿDUý@Vü<Uÿ<ZÿBdÿFkÿAjú;gõ:gö<mùCsýIyÿKzÿMyÿMtøPsðOpçMiêMjüYxÿgˆÿ_ƒõGlöWwè_yÃ^n½{‡a^�xp§‘„»§œÑÄ»ÝÖÐÝÜÚÜÛàÞÜçâÛëæ×ìëØîñÙñöÝó÷âó÷çôõëôôîòùôñùôðùñïöíîóèìîãééÞäæÝâÖÐÔÍÉÊ»º¸£¥ ‹�‰yƒxr}op{k}†sx�lv}k0./1/0320542653762873872:94;:5==5??7AA7CC9DD8EE;HH>HH>HH<JH9KJ6LK6MM5NM8KJ8KI=KJEPPRWU`YXjVTjRO`OIKMD?H?:F=8G>9H?:F=8B;5D:8A96?74?74@85@85=85;62=4/=4/=52<74=96>:9=<:>=9B?:D?9G@6J@4M@0O@-P?+L@*B?,B?.F<0H:1J65I56F35@65@<9:=6;>5@@4E<-J9)^B4€L>®QB«;- 2#¡6&£:'ž7$ž9%£>*¡;%§<(®?+³@-¶=,·9*¸6(µ7)ª9)¥:*¤9)¥;+¨>.«A1«B/«@.®?,«:(±<+¿E6ÆH9Á?1¼8+À6)Ë7+Ï7,Ð9.Ñ;-Ô</Ø:.à8/ç4-î1-ô.+ô.+ñ0+ê4)á7(×:'Õ:$Þ<'Ù6#Õ8%Ô=*Ì;*Á5&Ä:/ÏF<ÑF?ÚIDéOMóPQôGIð;@õ:AøCHåBEÚDCÒBAÅ>8·;1®:-ª<-«=.¶E5ÆL?ÔNEÜGCã?>ï?Bú?Fþ@JøCJ÷CLúDPÿFTýBSù>Sü>XÿBaÿFhÿCiü?iù@jùBpüFvþJzÿK{ÿJzÿHtõKpñQsõ\zù_{ûXwüNqýEkúDlïEiÚKgÈ_p±nukLI�~t°–‰Ì²¥äÓÉìãÜëçæêéîêçòíâóöãùûáúÿâýÿçÿÿìÿÿòþÿöþÿùûþú÷ýúõü÷ôúõòøïðôëîñèíïæëæÝàÜÖØÊÆų²®ž¡šŽ“Œ‰�…ˆ�ƒ�•‡‰�€†Œ~/.,0/-10.21/43/540762761:94::2<<4>>6@@6BB8CC9DD:IF=IG;JH;LI8MJ7NL7NL7OL9KI:NKBTRS_^fihxmm…lkŠkhƒ`YiYQ\OGRH@KH>GG=EE<A@:<C:;?:7<74<73=82>93>:1=9083-94.;60<92=:3>;4?<5@<3E>4G?4I?3L@0O?0P?/P?-L?,<;&:=(?<+A;-B71A62>42;30=84B;5H94N2.Y,)l/.‡;= FE£;0¢2$Ÿ1 ¤9'¥<)ž7$ž8"¢<&¤;&ª=)±A-¶A/¹>.»<-¾:-¾</²<.¬<.©9+©9+«<+­>-­>-­<,²?-°8(¹>.ËL=ËG:¼6*º2&É9.Ð6,Ô6+Ô8,Ö:.Ø:/Ü8.ã6/è3,ð0-ô.+ô.+ð1)ç2'Ý6&Ô8"Ò7!Ý8$Ü5#Ú9'Ú>/Ó=.È6)Å9,ËA7É<5Í>8ÙEAåKIêJJéCCì?AëCCÞCAÕD?ÍB=Â=4¶:.®8*«:*¬;+¬8)ÀD8ÔNEÝIEà@@è>A÷BIÿHPôAGô@IùCOýGTþEUú@Uû?XþA_ÿDfÿCiÿCkÿFpÿIwÿK{ÿL|ÿJzÿFxÿHwÿKtúNtÿZ|ÿa€ýUvñCdûIkÿTvÜ?\ÍI`Ø{…¸||aC;�|n®Ž�Õ³§óÛÑûìåúòðù÷ú÷ôýôìûúëÿÿêÿÿìÿÿîÿÿóÿÿ÷ÿÿûÿÿýüÿþûÿýùÿúøÿøöÿõöýóôûðôùðóðçêåßáÓÏξ½¹­®¨¢¥ž�£™�£—£«žž¦—›£–/.,/.,0/-10,21-32.54/650880991;;3==5??5AA7BB8CC9HE<JF;KH9MJ9PK8OM8QL8OL;LJ>QPL^]bmlzzz’�‚¡€‚¨€€¦{u—rkŠe_{YSmTLdMEZG@PB<HD?F@;?<87;63;60<8/=9.;8/:70991;;3>;2?=1@<1?;/A;-F?/H?.K?/M@/O?0O?/P>0L?/@=*?>,@=.?;/?;2>93=:5:94<94D95M51V-+j)-„28ž8C«>C¦7.¥7(¥:(ª?-¨?, 9&¡8#¤;&¥:&«<)³>,¹@/½>/Á=0Å=1Ä>2½?3¶=2²9.®8*¯9+°:,²:,³9*¶;,¹;-ÄB4ËE9È>3¾1'À2(Í9/Ö5+Ø4*Ú6-Û7-Þ7.á6,æ3,ì1*ó0,ô.+ô/)ï0(ä2&Û4$Ò6 Ð5×2Ú0Ú4$Ü<.Ö=/Ð9.Í=2ÒD:Ä5-Å60Ê;5ØD@âMIåKIÞC?Ö=8Ó@9ÎA8Ç>4¾</µ:+°8*¯9+±;-¬4&½?3ÑH@ÙHCÝB@å@DóFLýNUð?Eð?EöBMþHUþHWüBWú@XüA^ÿAcÿCfÿFmÿIrÿMyÿO|ÿL{ÿIyÿDvÿN~ÿR}þKrøMoÿVuÿUu÷MjðFcïKfÒ>VádtþŸ§Ó�Žd92lJ>¥ynÐ¥œõÔËÿéãÿôòÿýÿÿüÿùóÿüïÿÿïÿÿðÿÿòÿÿôÿÿ÷ÿÿûýÿüûÿýúÿüøÿúøÿù÷ÿö÷ÿõöÿôøþó÷öëïìãäÚÕÒÉÆÁº¹´³´¬±´©°¶ª¶¾³°¸­¬´©10,10,0/+0/+10,21,43.54/77/880991;;3==3??5AA7DA8IE:LF:NH:PJ:RK9RM:SL:QK=OJDVTUfdqwx�††¨��¹�“Ã��ÆŽ‹À‡ƒ¶{w©pmšid�_[~UPnNJaKGXEBM?=B;7895296/85,85,671783891;;1></?<+B=*C<)G?,J?+K@,LA/M@0M?2L@4K>5L?6K>5F<3B92=82:946;47<59<5>:1H4-Y2-w78”?D¦>E§9<ª;0ª<+«@.¬A/©@-¥<)¦;'«>*§8$­:'µ<+¼=.Á<-Å;.Ê<0Ê=3Å?6¾>3º:/·7,·7,¸8+º8+º8+»7+ÇA5ÍC8Å8.Á1&Æ3)Ë7-Ï5+Û4+ß3)á4-â6,ã5,ç5+ì1*ð/*õ/,ö/*ò/)ì0'â2#Ù4!Ð5Ï4Ö1Ù/Ø2"Ø6)×9-Ô:0Ö?6ØE=ÖF>Í@7É<5Ë@9ÖKDÙNGÒE>Å;1Å=1Â<0À</º;,·9*´:+¶<-¹?0µ9-¿<2É@:ÒC?ÛCBãDHíJOöOVì?Eí>CòBLûIUÿJYýFXúCYûC]û@_ûBdÿElÿJsÿNzÿO|ÿLyþIvÿJwÿT�ÿT}ýJqùImÿStÿVtøNiëD^äF]êXkÿ–¢ÿµ¼ó�žŽHFk.)¦kcÍ–�óÈÁÿãÞÿñðÿüýÿýÿü÷þþöÿÿöÿÿ÷ÿÿøÿÿúÿÿüÿÿýýÿýúÿüøÿûøÿùøÿø÷ÿöøÿõ÷ÿôøÿôøýñóóéêäÜÚÖÑÍÍÊÃÉÉ¿ÊÊÀÊÍÂÉÐȾȿ·Á¸65143.32.10+10+21,32-43.66.77/880::2<<2>>4@@6B@4JD8ME8OH8RK;TK:TM;SL<RKAQLIZW^li|~�Œ�¸–™Ì™žØšžÞ™˜Ú”‘ÔŒŠÉ†…¿�~µxv§nk–fe‡]ZwVTiMKYDAJ><?;:8;74762555457664872<:.?<+C>*F@*G?*H@+IA,IB0IA4HB6HA9G@:K=<I;;C9:=77875384/83/917<574+C1']80~C=–GCž>?ž51§;/ª<+ª?-ª?-©>,©>*¬?+°?-«8%²9(¹:+¿9-Ã9,È:.Í:0Ï<4Ë>5Æ;4Â91Á8.Á8.Â9/Â8.Ã6,Å7-ÑC9Ð@7Ã0&Ã,#Ð7/Ö<4Ó2*à3,ä2(ç2+è3*ê3+í2+ð/*ó-*ö/*õ.)ñ0)ì1(á4$Ú5"Ñ6 Ð5Ø7#Ù3#Ö3$Õ3&Õ5)Ô8,Ô<1Ö?6ìYQáRJÓHAÌC;ÊE<ÌI?ÈE;À>1¼:*¼;(¼:*¹:)·:(¸:+¼>0¿A3ÁA6Á<3Å<6ÏA=×EEÝGHäIMëLPèBFê@CðCIùKTþNYüJZüF\üF_øA_ùBaýEiÿJpÿNwÿOzÿNxÿLvÿQ{ÿRyþOvýOtÿVyÿ]{ÿSoîD^úTlîQd÷dtÿ’žÿ‹•ôƒ‰Ç`d§MMµjgÎ�Šï¼¸ÿÞÙÿïîÿúùÿüþýûÿýúÿþûÿÿûÿÿûÿÿüÿÿûýÿûûÿûøÿûøÿüùÿûúÿûúÿøúÿöøÿó÷ýñóùíïòææçÝÛßØÒÛ×ÎÝÛÏàÞÒßáÖÜãÛÏØÓÅÎÉ<94;8185052+41*41*52+63,74-85.96/;81=:1?<3A>5C?4JB7MC7PG8SJ;WK;UL=UK?SJCSJK]Wcnl‚‚‚¦‘•Åš ÚŸ¦ê¢©ñ¡¥ïž ë™›å•˜Ý”•Ö�‘ʉ‰½‚ƒ±zy¡rq‘fc~XVkPN\IGRFCLCBJ??K==G;:@;9:<94?;/C=-E@,F@*FA+EB/EC4CC7BC;AC>@ACB<FD>LCANEEOCHNAIL>HI>GDCHAA=2L:.gF7ƒOA�I<Ž?2”8)¥>/©>,«>*¬?+­@,¯@-°?-±<*±8'¶8)½9*Ã9,È8-Ì8.Ò91Ô;5Ï:4Í:3Ë81Ë81Ì92Í:2Ì70Ë6/Ó<5Õ>5Ò91Ì2(Î4*Ø;2Ú<3×3*ä1*é1'ì1*î2)ð1)ò/)ô-(ö,(÷-)ô/)ð1)é3(â5'Ù6%Ò7!Ï7 Ô9%Õ8%Ó6%Ñ4%Ò4(Ó7+Ò8.Ð7/ãNGèXPçZSÚQIÌG>ÄD9¿@7¹>/¹<*¹<&¹<(¹:'¸9(º;*À>0ÃA4ÉD;Æ@7É@:ÑFCÕIHÖHGØGJßIKæFHèBDíDIõLSüQZûO]ûL]ûK`öD^÷DaùFfýIlÿNuÿOxÿOxÿPw÷VxóUvôTvüYxÿ]{ÿZwûQkòI`ýWköUgêM^ö^mâKZæTaåTa×XaÁefȃ~嫧ÿÔÑÿëèÿóòÿøùÿþÿûüÿûüÿûüÿüüþüúýýùúý÷÷ü÷ôÿú÷ÿûøÿüûÿûúÿøøÿóóýîñúëîðáäéÝÝäÙÕâÙÒæßÕëçÛñíáññåìóìÛæâÎÙÕ@=6=:3:7074-52+52+52+63,74-74-96/;81=:1?<3@=4B>3JB7MC7RF8VJ:WK;XL<WK?TICSJM^Xfpmˆ‚„«‘—Ëœ¤ã£­õ§°ýª±ÿ¦¬ú¡§ó ¤îŸ£ê�¡á™›Ö•–Ì’’ĉˆ´yyŸji‹`]|XUpRPhOMeNJcKG^FBS@=H?:>?:6@<1A>/C@/CB0BC3BD7?D=>D@<ED;BJ>CVCIaLRhU\o\br`dobbjd`afXWaJDlG>ƒSE•XF“J7‘@+™>+¨A.®@/®A-¯@-°A.´A/²=+°7&µ7(¼8+Â8+È8-Í6-Ñ7/Õ81Ø;4Ò72Ð72Ð72Ñ82Ô94Ô94Ó83Ó6/ÞA:Õ8/Ñ3*Ö8/Þ=5Þ=5Ú91Ü5,ç2)ë0'ï0(ñ0)ô/)÷-)÷,(÷,(÷-)ô/)î2)é4)á5'Ú7&Ô7$Ï8#Î7"Ï8%Î7$Ï6&Ð7)Ò9+Ñ7+Î4*Ë4+ãNGód\ë`YÕOFÄA7º;2µ9-¸>)¸>'¹<&¹;%º9&½9*À</Ä>2ÊD9ÊA9ÎE?ÕKHÓLIÎGDÍEEÖHGãIIæCDèEHòMSøSZùR\÷O^úOaôH^õG`øGdûIiÿNrÿQxÿRyÿTzðUuïZwù^}ý^|ùUpõMgùOiÿYnÿ[mÿctðM^÷TeôO`üUgõL_äR_À^_»vqל˜úÉÅÿåãÿíëÿôôÿÿýüÿÿûÿÿûÿÿûÿþýýýýüúþùöýøõÿøõÿùöÿúùÿúùÿööüððøéìôææëÝÝèÚÙäÙÓèÞÕðèÝùóåÿúìþþòôúöáëêÓÝÜB>5A=4@<3>:1<8/:6-84+73*62)62)73*84+;60>93A<6E>6I@7MC7RF8UI9WJ:XK;ZLA[NHTIMXR`gd�|~§�’Ê— ãžªö¥°ÿª³ÿ«²ÿ¬´ÿ¬²üª±ùª¯ó¨¬ì©«èŸ¡Ú™šÒ��Á�±tr¡he’]Z‡XR~XQzYQvVOnMH_C?M?:@?;:B?8>>4@B5BE:?D=<B>:CB>FH?LUDVnQfƒarŽisŽtv��x�„o€€^lŠ[c’WYžWU¦VM¦M?Ÿ@.œ9$ :$§<(¬=*®?,±@.³@-´?-¶=,¹;,º6'¾6(Ä6*Ë7-Ñ7-Ö8/Ú91Ú83×84Ø95Ù:6Ø93Ø61Ø61Ú83Ý:3æC<à=6Ú70ã@7þ[Rÿlcÿ]Tç?6å3)í2)ñ2*ô/)ô*&õ(%ø(&ø+(÷-+ó0,í2+ã1'Ú0#Ô1"Ô4$Ó:(Ì;&Ê<(Í<)Ï<*Ò:,Ò9+Ò8,Ñ7-Ú@8Ï81ÜGAôc^åXQËB<Å@;·4*·:&¶<$¹<&º<&½:(¿:+Ä<0Æ>2Ç>4ÌC;ÙPJßXRÔOJÃ?:À?:ÏHDÞHGâGEæJKïPTñRWïNVíKXñL\ùRføOf÷Ke÷JfüMlÿQsÿRvýRvôWvîXuõXuû[wÿ\xÿZsÿWnÿVkÿbuýVgÿ\mÿ[lñFXôDXÿQeîVeÆfgªieÊ�‹ç¶²øÒÏÿëçýïîüø÷ùýüùÿÿùÿÿúÿþüþýÿþüÿûøþöôþùõÿû÷ÿùöÿöôÿööÿõõöèèæØØãÕÔÞÐÍáÖÐñçÞüôçþøèÿüéÿÿóúÿùèñðØáàC?6B>3A=4?;2=90;7.:6-95,73*73*73*84+:5/=82@;5D=5H?6LB6QE7TH8VI8WJ9YK>ZMEUJNWQ_eby{¤Š�É”�⛦õ£®þ«´ÿ¬µÿ®¶ÿ¯·ÿ°·ÿ¯¶ü¯´ø¯²õ©ªë¢£ã˜˜ØŒŠÉ€»uq®ie c^˜`WŽ`Vˆ]TVPtMIbFCTBAIAAC<<:AB=FEACD?A?@DBEOMRWVdch†quš}¢‡}Ÿ’}œ {š¥pŠ¤`u¬Yi¶VaºQU¸HF°@4¬;)¬<&­='¨8$¨7%«8&®9(°8'²9(µ7(¸6(»3%Â4(Ê7-Ó:2Ù<3Þ=5ß<5ß<5á>9ß<7ß<7á>9åB=èC=çB<æA;å@:æB9ä@7Ü8/Õ1(Ø6+ëI>ÿ[Oñ@6î6,ê+#î)#ø-)ÿ/-þ,+ö((ú0.õ4/ï61ç6.Þ3)Ö2&Ô2%Ï6&Ë:'È;'Ê<(Í:(Ï9*Ð7)Ñ5)Ð4(Ó9/Ï5-Ô=6åPJåTQÛLHÍB?·1(¼;(»=&½<'¾;'¾9(À8(Â:,Ä:/È>4ËB8ÕOFÜWPÒRIÃD=¿C;ÌHCÛLHßJFåMLëSRíTWëPVéNVìOZ÷VføUhùSiúRküRmÿSqÿStÿRuòMmøUtÿ\zÿ_{ÿ\vÿUoûRiûReüUf÷RbÿZhÿZhóN^øScþYißS^¶`_—^W¶�}Ù©¥îÈÅýáÞùééýøõúüùùÿÿøÿÿ÷ÿýúþýÿÿýÿûûÿ÷õýøôþùõþùõþöóÿ÷öÿõõöêêêÜÛäÖÓÝÐÊßÕÌïåÛüõåÿùçÿýéÿÿïúÿøèñîÛáßEA6EA5C?4A=2?;0=9.<8-;7,84+84+73*84+:5/=82?:4C<4G>5JB7ND8RF6TG6WG7YI<YJCWKKXP]b^ysuœ…ŠÂ�™Ü˜¤ð «úª³þ­µþ¯·ÿ±¹ÿ³ºÿ´»ÿµ¹ÿµ¹ÿ¯²ù¬¬ô££ë™—àŽ�Óƒ€Çxu¼sm³k`¢i]›cZ‘]W…VRwNKhEDV@@L85<A<@KABL@BQ>BZEJpW]‚cs˜r—¥z§®«²{£¸vœÀp“ÂcƒÀTnÂG\ÐK\ÑHPÄ<<º5.¸9*µ<'°<%°;'°;)²:)´;*·<,»<-À<-Ä<.Ä6*Ë8.Ó<1Ù?5ßA8á>7à<3ß:4Ü71Ù4.Ø3-Þ93æA;ìE?éB<ä=5ã<4Û4,Ø4+Û7-Ô2'Ì, Ö6*éG:õK>ò@6ï4-ð-)ö,*ú,,ü,,ø*,ð*)ì/+è2.ã5.Ü4+Ø1(Õ1'Ñ3'Í7(Ë:)Í:(Ð:)Ò;*Ó:*Ô8+Ô8,Ñ4+Ö<4Õ:5Õ<7åONíYWÙEEÂ3/À;,¿<(À;*¿:)Á9)À8(À8*À8,Æ=3ÇA6ÏJAØUKÒRGÆG>ÀD:ÇG>ÖKDØICÝLIäSPåSTâPSâMSåNWòXdöZhý[pþZrüVnúRlûQlÿQpýGmÿPvÿZ}ÿ[{ÿVrýTkþWkÿ\mûYhøZhûamö`kí]gïfnìfmÉ\_ ]W{OF˜mfÁ˜”ݺ¶ðÔÑôàßÿõôùù÷ùÿýùÿÿ÷ÿýúþýÿÿÿÿýÿÿøùüùôüùòüùôýøôÿùöÿøöùîìïäâåÚÖÛÐÊÜÒÈîäØýöäÿûèÿýçÿÿíúýôêðìÞãßIE9HD8FB6D@4B>3@<1?;0>:/:6-95,95,95,:5/<71>93@;5E>6I@7LD7OF5RE4UE5WG8WI>XMKXNW^Zqpp–€†ºŒ–Ô•¡é�©õ¨°ù©±ù­³ý°·ÿ±·ÿ²¹ÿ´¸ÿ´¸ÿ¯±ü«­ø¥§òŸžê—–⌋ׂ�Í}yÅum¶pgªf`ž_Z�YX„QQuGGcA>Q=5DH8BT=C`@EnCJ€OU–]d«fx¿k�ÉnšËo˜Éi�È`ƒÉWxÆIgÄ;UÆ2HÙAPàEMÑ;<Æ71Ä?0½@,±:$­6"­5$®5$³5&¶7(¼8+Ã;-È</Ó@6×@7Ú@6Ü?6Þ=5Þ93Ý6.Û4,Ó.(Ò-'Ó.(Ø3-á:4ã<4á81Ü5-Ü5-Ð,"Í)Ù7,âB6Ü?0Ð4%Ë.éF7óI<ýH?ü<7ô-*ñ#%õ')û/2ñ-.ë/.ä2.Þ3,Ú1*×0*Ö/)Ó0'Ó7*Ò9+Ö:+Ø<-Ú>/Û?0Ü@3Ü@3Ò6*æLBåJEÎ50ßGFóZ\ßIKÕA?È</Ä;+Ã:*Ã:*Â:*Â:*Á9+¾9*Â<0Â>2ÈF9ÏOBÎPDÆH<¿C7ÁC7ÎH?ÑF?ÕJEÛPKÝROÙMLÚKMÜKPëX`ó[güaqÿauúXmõOgõMgÿNkÿOuÿRyÿUwÿTqüTmý[pÿcuÿjxÿguúboõamídlãflÙhjÎghµfa�^TeG<^U©ˆ�Í«©åÇÅïÚÙÿóóùøöùýüøÿÿ÷ÿÿûÿÿÿþÿÿüÿÿùúþûöûûóûúõþûöÿýùÿûøþôòøíéêßÙÞÔËÝÓÉíæÖÿøåÿþèÿÿæÿÿëøúïêïèãæßNH:MG9LF8JD6HB6F@4D=3C<2?80>7/=6.=6.=60>71@93?:4C>8EA8KC8NE6QD3RE2VF6VH;ZMGVMR[Whlk‹~‚±Š“Î’Ÿãš¦î¦®÷©¯ù­°ý¯³ý°³ÿ°´þ°³ÿ¯³ý«®û¨«ø¤§ô ¢ïšœé“•âŠŒÙ†…уÉ{u»mk¬dež_a’Y[„RQsNHbWFY_CQmBL�EM˜MT¬U]¼[bÊZhÙRpâRußTuÙPlÒJbÍBWÇ8JÅ.?Ô8EãCKæGKØ>>Í>6ÌF:ÃH6¶?+·>-¸=-º<-½>/Ã?0ÊB4ÐD7ÕE:ÙE9ÚB7Û>5Ù;2Ù6-Ù5,Ú3+×3*Ô1*×4-Ú7.Ý90ß80Þ7/Ý6.Ü5,Ö/&Ù5+×7+Ï2#Í1"Ò9)Õ>-Ô;)Ø<-æD7øJAýD?ù64õ*-ö*-÷/2÷67í55â30Ú1,×0*×/,Ù0-Ù0+ã81ä91â:1â;2à<0ß=0Ü?0Ú>1Ñ7+ïWLøaZÓ;6Ñ;:æPQÜFHæPQÏ<4Ê:/Æ8,Å9,Ä:-Ä<.Ã;-À;,¾:-¾<.ÃA3ÇH9ÇI;ÂD6¾@2¼>0ÉF<ÊE<ÎIBÕPI×RMÕNKÓKK×LOéZ`ð^hüfrÿhxü]qòPeóMeýNkÿVzÿTxÿRrøTmö]qûhxûjw÷erüetó_mì`kêkrßruÃjfªd\žla}fVXH9kXJ”|r½ œÞÂÁîÖÖþîïûõõûûûûÿÿùÿÿýþÿÿþÿÿûÿÿúýÿþùûþõûüöÿþùÿÿúÿþúÿûõÿ÷òñèáäÜÑáÙÌðé×ÿúäÿÿæÿþåÿÿê÷ùëíðçèéáQK=PJ<OI;MG9KE9IC7G@6G@6B;3A:2@91?80?82?82A:4@;5B?8EB9KE9MF6PE3RE2UF3UH7YLCUKLZScii…|�«‰“È“žÞš¤ë¦¬ø©¬û¬­ý­°ÿ®±ÿ®±þ­°ý­°ý©¬ù§ª÷¤¨ò¢¦ðŸ£íšžè•™ã’”Þ’’Ú‰ŠÍ|€Àsx²ou©mp�igŽj_}y_x‚WjŽP_£O\»S\ÍT]ÖPWÛHRå@Qè?RäCRÜDPÖCKÏ@DÊ9<É46äJLçIJáAAÕ74Î95ÎA8ÈD8¾>1ÃE6ÃE6ÅF7ÉE8ÎF8ÑG:×G<ÚG=Ö>3Õ;1Ö8/Õ4,Ö3*Ø4+Û4,Ú6-Û81Ý<4à=4ß<3Ý90Ü5-Ü5,Û7-Ò.$Ý;0Ú=.Í1"Æ-Î7&Ò=)Í:&Ì9'Ô8)â:/ó=9ÿ@@ÿ<>þ37ó,/ê,.á+*Ø*)Ô+(×/,Þ44ä88é99ï75ð74î73è71ã7-Ü6*×5(Ñ5&Í5(çPEÿmcÛHAÈ42×CCÓ>BêVVÙA<Ò>4Ë8.Æ6+Æ:-Æ</Ã>/Á=.»9+»<-¾?0ÀA2¿A2½@.¼=.»<-ÃC6ÄD9ÈH?ÐPGÓRLÐOIÒNLÕONé^cîagûitÿnzÿduõVjõQiÿUoÿWvÿVtúXpñ^pônyöw€íksÞXañ`mñ`mìboïs}숊Ì�|ªth™yjskXON:_VGƒqg¯–’ÜÀ¿ïÕØüéëþôõýûüýþÿúþÿþýÿÿýÿÿûÿÿùþÿÿûùÿõúýöÿÿúÿÿúÿÿøÿý÷ÿþöùðçíåÚéáÔôïÜÿûåÿþåÿÿãÿÿèùúêòôçîðåSM=RL<QK=OI;MG9KE7JD8IC7E>4D=3B;3A:2@93A:4A96@;7A@;CC;IE9MG7OG2RF0UF1UH5WK?RHFXR^jg‚}�§‰“Ä’žÚ˜¢ç¤§ö¦§ú«©ü«¬þ¬­ÿ­®þ¬®û¬®û¨¬ö¦ªô¤¨ñ£§ð¡¨î §í�¤è›¢æ™žâ’šÛ‹“Ò†�ʃ�„ˆ·ƒ�¨ˆyš�mˆšcz©Xi»Q_ÏMYàIRèAIê=Aé=;ã>8ÝC9ÖH:ÎI8ÅF3ÂC0Å@1äTIàG?Ú;7Ø64Ø88Ø:;Ñ98È74À:/¾<.¿;.À:.Â8+Ã7*Æ4'Æ2&Î4*Ð3*Ñ3(Ô3)Ö3*Ù5,Ý6.Ü8/Þ=3Û=2Ü;1Û8/Û7-Ú6,Ø4*×3)Õ3&Ó3%Ð4%Ï6&Ò=)Ñ>*Ê9$À2È:&Ê4%Õ1'ç51ú<<ÿ=@ÿ7=÷37ê-1á-.Ù--×/.Ü43ã9:é;=ð9=ö26ø03ô01í1/ä1*Þ2(Õ1%Ï3$É1$Ñ=1új_äTLÈ95Ð@?Ë:=äRSäJHÜC=Ñ:3É6.Æ8.Å;0Â<0¿=/º;*¼?-½@.¼?-º;*º;(½;+¾<,¼>0¼>2ÀD8ÈLBÌPHÊNFÌLIÐNLèaeìaføhrÿo{ÿhx÷Zk÷UjþZrÿYsû]tîaræjtë}€ñ‡‰èwyÚ_dñgtþm|óeuëlwö‘•ì¢ŸÄ”Š¢�|hjUHQ<WUFth\¤ŒˆÛ¿¾òÕÙùãæÿóöÿúüþþÿüýÿþûÿÿüÿÿúÿÿøýÿÿûùÿõ÷ýóýÿ÷ÿÿøÿýöÿý÷ÿÿöÿùïõïáòìÜúõáÿýçÿýäÿþâÿÿçýþìùúì÷÷ëVO?TN>SK>PJ<OG:LF8LD7JD8H@5F@4E<3B;1B92A:2C:3A<6C@9DD:JF:MG7OG2QF0TH2UH5UI=QGEWQ]jh€~‚¨Œ”Å“�Ø™¢çŸ ò£ ÷¦£ú§§ý¨ªÿ©«þ©¬ý§­û¥«õ£ªò£§î¢§ë¤©í¦©î¥¨í£§î˜ è”¡å”Ÿß”Ÿ×•›Íš•¾ Œ¯ªƒ ¥e´]pÂQcÐIWÜBNå>Fë<Aì<<ê?8ãA4ÜE4ÔI4ÊI3ÂF.¿B,Â?-âSEÞD<Ý97æ>>êDFå@DÙ8=Î65ÍB;ÈD8ÉC8ÇA5Æ>2Ã9,Â6)Â2'Æ2(Ë3(Î4*Ó5,Ø5.Ù4.Û4.Û4,äB7Ý=1Ù7,Ú6,ß9-ß9-Û5)Õ1%Ø8(Í2 Ê1!Ï9(Î8'Ç4"Ê7%ÔC0È7&Î8*×6,Þ5.æ3/î53ö:9û?>ÿLKúHFòBBì>=ê<=ç7:å26æ,1ô+1ø)/ô,.ï/.è1+à4*Ù5)Ó7(Æ0!À.ôdYë]SÏ@:ÔDCË;;ßMMêRQàHEÓ;6É6/Æ8.Æ9/Ã;-¾:+¾<,Á@-ÀA.»<)·8%¸9&½<)¾?.µ7)³7+¸>1ÁG:ÅK@ÅIAÈIBËJEåa_æ^`ñdjþnwþkuó]iðXeø]mú^sõbtèdoãlrì~ù�‹ö†…éqsøryÿy†õaqàXföˆ‘ÿ´´Ð®¢ ›‡]jPDR9QT?jbU›ˆ�ÚÀ¿òØÙöàãÿô÷ÿúýÿþÿýüÿþüÿÿýÿÿûÿþùýûÿúôÿòòþòùÿôüÿöûýòÿýôÿÿôÿÿóûõç÷ñáþùåÿýçþüãþýáÿÿêÿÿñýýóüüòXO@WP@WN?UN>TK<RK;RI:PH;MC7KC6KA5H@3H>2G?2H>2F@4GC:GE9JG8NI6RJ5UJ4UJ4VK9XNDULMZTbkhƒ{~§‰�љטŸç£¤ö¥¢ù¢¢ü¡¢ü ¤ÿ¢¨ÿ£«ÿ¤­ü¡ªõ¡©ñ¡¨ì¥¨íª©ï­©ó±©ö¬©ø ©ø—§ò’¡â–ŸÖ¦£Î¶ž¾¹…œ¸gz¿L]ÑERß>Mç;Gç<Dæ=Bà@@à@@â>?ß<=ÝEBÉ:4ÊD9ÊF:º1)ÛJEèJIéCEë>Bè;?æ<?á=>Ú<=Ô<;ÏB;È?7ÊD;ÑMAÍK>¿=0¸6)½9,¿7+Ã6,È5-Ë2*Ñ/*Ô/+Û2/Þ5.çA5ß9+Ù1$Ý3&å9+ç;-ß8&Õ2Ï2Ì4Ê5!Ç4"Ç4"Ê4%Í5'Ï5)Ê0&Ï2)Ñ5)Ò6*Ò4(Õ3&×3'Ü4)à2)æ3.ë52ë33é/2ç-2è-4í.5ö-3ö+/ï+,ì/-ê5.ã9,Ù7(Ð4%Ã-Ä3"çXHéYNË;3ßNIÅ41ÚHHáOPÜJJÔC@Ì;6È80Æ8,Ç9+È<+Å<*Â;(¾9&¼;(»<)º=)º=+·<,´8,µ;0¹=1»?3»?3¾@4À@5ÄA7ïjaõpiêc_ômjûqqìadôgmÿpyÿlzûhxðdoébiëdjîlnîqoîqoûy{ÿ}…ùZlücwßbpû¬¯¶ªš}‘u]rQJY:LR8snZ¥•ˆË·°íÕÓÿîðÿö÷ÿ÷ûýøüüüþþÿÿþÿÿÿþÿûÿþðÿôåþéæûêïÿîøÿöüÿôþÿóÿþñÿýñÿûîÿúêÿùæÿùãÿùáÿùáüúåÿÿõÿÿûÿÿûYPAXO@XO@VM>UL=TK<SJ;RI:NE6MD5LC4KB3JA2JA2JA2IA4GC8HD9LF8OH6RJ5SK4VK7TK:XNEULMYUckhƒ{}¦†ŒÀ�•Ó–šáŸžìŸ�ñžžôžŸù�¢ýŸ¦ÿ ªÿ¡¬û ¬ô «íŸ§è¢¥è¥¤ê©£í« î¤Ÿï §÷š¦ðœ¡á£œÐ±˜À¿�ªÂq„ÄTbÑCOâ<Fì8Aï7?ê:=å<?Ü@AÛACà<CÚ9?ÚADË;;ÌB?Ê@=Á31åOPæCFè?Dê;Bç8?â8;Û89Õ:8Ï;7ÔE?Æ=5Å<4ÌG>ÌLAÁC5¸:,·9+¹7*¿7+Æ6-Ë4-Ñ2.Ø3/á53ä84ä>2á;-Þ6)ß5&á5'á5'Ú5"Ó4Ñ9$Ë9"Ç9%Ç:&É<+Ë<,Ì8,Î4*Õ3.Ö5-Ó7+Ï7)É6&Ç4"Ë4!Ï4"Ø2$Þ2&ã0+æ.,æ,-æ,/æ-2è.3ì+.í,-ì0.è2.à4*Û5'Õ8'Ò9'Õ?.»*ÖD5Ð>1ÙF>ëVPÚB?ØBAçUVâRRÜKHÔC>Í=5É9.È9+Ç8(È;*Ä;(¿:'½<)½>+º?-¹@/¶>.²:,´;0·=0¸>1º?0½?1ÁB3ÅC5ëi\òmdçb[ðkfõolèbañklütxÿq~ÿo|öirí]fêY`ñbf÷qpþzxûvwÿx~ùRdþ[pä]nö¦©©¦“jŒkZtON_;SZ;us\©��ÖĸöáÜÿõñÿúúÿúûÿûüüüüúþýûÿþúþÿõÿúíÿóáÿéáýæéÿëóÿñøÿòýÿòÿÿñÿýïÿýíÿúéÿöåûóàúòÝüôßü÷äýúóüüúÿÿý\PB[OA[OAYM?XL>WK=VJ<VJ<SG9RF8QE7OC5OC5OC5OC5MC7JC9JF;MG9PI7SK6VK5VK7TK:WMCVMNZWbki�y{¡…‰¹‹‘Ë‘–Ö˜˜Þ™™ã˜™éš�î› ô�£÷Ÿ§ø ©ô ªï�¨èœ¤ãœ¢àžŸàž�á�™ã›™â–›ßššÚ¤•Ð±ŒÀ¿‚¬Çr“ÉYqÇBS×ALß>Dã=?å==å<?ã=?á>Aá>Cà<C×8=Õ?AÏA@ÊC?Ã<8Ä96éWWàBCä>Bè;?ã9<Ü68Ö66Ñ96Ì;6ÕHAÄ;1¾5+ÅA5ÍK>ÆH:»=/·8)º6)¿7)Ç7,Ï6.×50Þ71æ95é=9×3*×5*Ú6*Ü6*Ý5*Ý7)Ú:*×>,Å4!¿4¹4!¹6"¼8)½8)½3(À-%ã@;å>8Þ=3Õ<.Ì9)È7$È7"Ì7#Ò7%Õ3$×/&Ù.'Û.*Ü.-Û//Ü./Û,)ß0+à5-Ü5,Õ3&Ð4%Ð9&Ñ>*ÔA/Â1 çSEÒ;0ÚA;ÞC?èJIðTUíUTçSQãOMÝJCÖC;Ï<2Ê8+Å6&Ä7&À7%¼7&¹:'¹<*·>-µ=,²=,°:,±;/´<.µ=/·=.¹?0¾C4ÃE7Ü\Qêg]äaYðlgöpmça`ìfgójqàR^ï^kõhqòemñ`gôeiùqqþxwútuÿnuóJ]øRhä[mñœ¡£œŠa€`UoHOd=W`Aww]­¥’áÑÂÿîæÿøñÿüøÿþúþÿúûÿüûÿýûÿýùÿûóÿöèÿîÜþãÚøÞáùáèûåïüèöýëýþìýúéÿúêÿ÷æúñàôèØòçÕöëÙøñáþúñýüøÿþû\PB\PB[OAZN@YM?YM?XL>XL>UI;TH:SG9RF8RF8RF8RF8PF:LE;KG<OI;RK9TL7WL6WL8WK;VLBUMK\V`jh~yy›‚…²Š�Â�”Γ•Ò•”Ö••Û–˜ã˜›è˜žìš¢í›£ëœ¥è˜¢ß–ŸÚ•›Õ”˜Õ“—Ö’“Ö‘�Ò…„½‘ƒ¶¥}±·s¢Èf�ÒVzÕGaÔ<KÞAJàBCÞCAàB?áAAä?Cç>Eç=Fã?FÖ:>Ñ?@ÑEDÆA<º61Å>:êZYÝABã@Cæ=@ã:=Û89Ó97Ð<8ÎA8ÏE;Â:.»2(À</ÇE7ÄE6½>/¸9(½8)Â9)Ë8.Ò:/Ù80à91è;7ë>8Ø1+Ù2*Ü3,Ý5,à5-ß7.Û9.Ö=/É7(Æ:)Á<+¾<,¿;.Á;0Å<6Î95ëC@í@<ä@7Ù=1Ï9*Ç9%È:&É;%Ï<(Ð7'Ï1%Ñ1%Ô1(Ö3,Ô1,Ñ/*Ó2*Ó5,Ó7+Ð7)Í5'Ê7%Ì;(Ì>*Ì;(Ï<,ô^PàF<Ò3/Ò.,æ@@ûWXÿusÿroÿjgö^YçPGÕA7Ç5(½. Ä8'À8(»8&¸;)·<,µ=-²=,°<-¯;.­<.°</°<-±;-¶>.¼B3ÀF7ËMAß`Wâa[ôpløtræ`aä]aæ]dÖHTçYeógrôgoöelükpþsvþvvûpsûenñDXôKbåXkë’– •ƒ_zYQlCSh?\gEy|_°ª”çÜÊÿõèÿùðÿûñþþôýÿ÷ûÿúùÿúöÿøòüóêûëÚøÞÎóÒÊëÌÏìÎ×îÒàðÖéóÛò÷áöõáûöãüõãõìÛíáÑéÝÍíáÑñèÙüõíü÷ñþùó\PB\PB[OA[OAZN@YM?YM?YM?WK=VJ<UI;UI;TH:UI;UI;SI=OG<OI=QI<SL:UM8XM7YL9XL<UKAUKI\U]hexut“|~¥…‡·Œ�ÂŽ�Ã��È�’Ï�”Ô’•Ü“—à”˜ß•šÞ•�Ü’›ÖŽ”Ή�Ç…‹Å�‡Ã}ƒÁ~¸}tŸŽp”¨g�Á\„ÖNvâAcè:Uê9KæAGàECÜGAÛGCßEEæAGî<Jí=JæBI×>@Ð@?ÐIE¿>8³2,ÈC<àUPÜDCáACã?@á>?Ú?=Ô@<ÑD;ÏF<Ç=2À8*½5)½9*¿=-¿=-½;+½<)Â:*Ç;*Ï;/Õ;/Û8/à8/ç83é:5æ95ä52ä20æ21ç32æ40Ý2+Ô0'Ë1'Ë7-È:0Ä7.Á4-Ã40Ì:;Ý?@ë8;í76ä71Ø5,Ì4&Æ5"Ä7#Å:%Ë=)Ë8&Ì4&Î5'Ô8,Ô:0Ó9/Ð8-Ð>1Ê;-Å8'Ä7%Æ9'Ç<)Ç<)Ç:(Í>-Í:*Ø>2áC:Ô/-ä::à24ß56×53Ö;6ÞC>åKCêQIêSHèTHäUGË?0Ä?.¾<,¸=-·>-´?.°?/®>0¬<.¬<.¬<.¬=,­<,±=.·A3ºD6ÆLAÞbXâc]ðnlôrrå`cå`eç`g÷kvûoz÷kví`hð_fýlqÿx{ÿy|úmsø^jóAWôC]èUhæ‡�£“ƒg]YtI^uIhwP�ˆf³²–èâÌÿúéÿýíÿþïýÿòûÿôøÿôôÿñíûêäòáØî×Åèǻ伺޺¿Þ¼ÇàÀÍâÃØæÌãëÓìîØôñÞ÷òßóêÙìàÒèÚÍêÜÏïáÖòèßñèáòéâ[OA[OA[OAZN@ZN@YM?YM?YM?XL>XL>WK=VJ<VJ<WK=WK=XL>QI>QI>SK>UL;XM9XM7YL9XM;WK?ULGYSWe^nnjƒut–~}¥„…±†‰´‡‰¹Š‹Á‹�È�ŽÏŽŽÔ��Õ��Ó�”Ó‹�ʃˆÀ|�¸u|²ov­io«ljœ€lˆ“cy®YvËPoãBdð7Vö4L÷8Hï@EåFBÝJCÛJEáGGèCJò=Nð>LäCI×CCÌA>ÎIDº;4°4,ÉI@ÕLFØBAÚ?=Û;;Ú<;Ø@=ÔE?ÍG<ËG:À8*Á8(À8*¾9(¼9'»8&½:(À;(Ä;)Ê=,Ò<.×;.Ü8.á6.ä6/è50é32è./é,0î02õ47õ77î45æ21Ú.*×4/Õ62Ñ32Ï/1Ò/4Û6=ê:Dí06î02æ3/Ü3,Ñ5)Ê7'È9(É<(Æ8$É6$Î6(Ó:,Ú>2ÝA5ÝA5ØA6ÓG8ÉA1Â;(À9&Ã<)Æ=*Å<*Å8&Ì:+Í7)Í0'æC<Û2/ï?Aã/2Ú*,Õ1/Ò5.Õ81Ö90Ò8.Ì5*Æ2&À1#ÎE5ÆA2¼=.µ:*±9)¯9+«:,ª:,«;/«=0¬>/¬<.«<+­<,²>/µA2ÉSGàg^àc_êjiînoæchðjqõoxúq{ýt~ömuîbködnÿrzÿx~ÿquúipøWfùAYô@[êQfá~ƒ¬—†y�kmˆ]o‰ZyŠ`�—r¶·˜ßÞÂùôÞüúåÿÿïûÿïõÿïïÿëçúäÜïÙÐãÍÅÞÁ°Ö­¨Ó¦ªÐ§°Ó«¶Ô®¼Ö±ÇÚºÒßÁàæÌèêÔòíÚòéØîâÔìÛÑêÙÏìÛÓéÚÓæÙÑåØÐ[N>[N>[N>[N>ZM=ZM=ZM=ZM=YL<YL<XK;XK;XK;YL<YL<YM?SI?TJ>UL=XL<YN:ZM:ZM:YN<YM?ULEXOR_Wbd_sjf�so�yy�|~¤~€©�‚²ƒ„¼††Æˆ‡Ë‰ˆÌ‰ˆÊˆˆÈƒƒ¿|}µvw­pq§kl¤fg l`Ž‚`q•V^³N^ÒJ^ì@X÷6Kþ3Fþ9Có?BéD@áHBÞICâFIéBJò=Nï>NÝAEÔFDÈA=ÉHB´:/±7,ÎOFÉB<Î=8Ð64Ï10Ð51Ó>8ÏE;ÈF9ÃD5¾9(Ã:(Â:*¿:'½8%¾9&¿:'Á:&Ä9&Ë:)Ó:,Ù9+Ý7+á5+ä3+ê3-ë*+ñ*-ø-3ü/4ÿ17ÿ37ÿ38ý58ù8;õ8<ò9>ò9Aô9D÷7Fû6Hþ5Eô+5ô-2í12ã4/Ù6-Ð8*Î;+Î=,Ë5&Ñ8*Ù;/ß=2â>4ä=4å>5ßA6ÙJ<ÏG7ÉA1Ä<,Â9)Â9'Æ8*È9)É2'Ó9/á@8ëD>Þ0/à..ë46è66Ø3-Õ7.×90×:1Õ;1Ò;0Í:0É;/ÖL?ÌF:¿@1´9*°6)¬6(«7*©8*¬;-­=/®>0­=/«<+«<+­<.±=0ÀKAÚdZÞc^èkiðosêinõrzüvîfp÷oyùryöjsþlvÿwÿv}ùhoùcn÷RbþB[÷<YêKaÝt{µš‰�žz}˜k{—f‚•h‘�u¬²ŽËÌ­ààÄéìÑîöÞåóÙÙíÑÏæÉÅÞÀ¼Õ·´Ë®ªÈ¤šÁ”–Â�™Ã“¡È™§Ë�¬Ë¡µÎ§ÁÒ°ÎÙ»ÙÝÄåãÎìåÕïáÖëÚÒçÒÍãÐÊàÏÈÚÌÃÕǾ[N>ZM=ZM=ZM=ZM=ZM=YL<YL<ZM=YL<YL<XK;XK;YL<ZM=ZL?VJ>VJ>XL>YN<ZM<ZM:ZM:ZM<[OAWMDWML\RZ^Xfb]qjfsp�wv–xxœ|{§�~µ„¿…‚ň‚ȉ„Ç‚}¿y·zu­xq§vo¥tk¢pg za‹„QZ™HE³EHÓHOîBNø:Fþ6Aþ9@õ==îA=æD?ãEBäCHèAKî=Më?MØ?BÑGDÅA<ÃG?²9.³:/ÑUKÃ>5É83Ç/,Å*&È/*Í:3ÌC9ÃC6¼>/À;*Ä;)Ã<)Á:'Â;(Ã<)Â;'Ã8%Ä5$Ê7%Ô8)Ù7*Ý5*á3*å3)ê2*õ33ÿ58ÿ8<ÿ4:ÿ-2ÿ(-ÿ(-þ+1ü-3ö)0ò&1÷)6ÿ,?ÿ*Aü 9ó/ö"0ó(.í,1æ0/Ú1,Ô3+Ñ5)Ð6*Ô6+Ý90æ=6ë>7ë:4è71è50â92ßH=×K<ÓE7Ë=/Ä6(Â3%È6)Ï8-Î4*Ô6-æC<ß82à21Ò  ì89å63Õ1(Ò4)Ò4)Ï5)Î6+Ê6*È5+Å7+ãYNØRFÊH;¾@2·;/µ;.µ<1µ<1¯9-±;/±=0±=.¯;,­9*­9,®:-±<2ÐYQÜb]ìppôvyîmrõq|ör}ÿzƒÿ‰ÿy€ôhq÷epÿq|ÿuÿktø^jöOaÿA^÷8WéG^Ûntº�‹™©„‚žn~šg�—h�œq¡©‚´º–ÈÍ­ÕܽÌÙ»ÁÔ´²Ì©§ÂŸž¼˜š¶�–²Œ�±†‰³�†µ~�¹„•À‹œÂ�ŸÄ‘¨Å™²È¡¾ÎªÉÒµÚÙÄæßÍêÞÒéÖÏáÌÉÛÆÃÔÁ»Î½µÈ·°ZM=ZM=ZM=ZM=ZM=ZM=ZM=ZM=XK;XK;XK;YL<YL<ZM=ZM=ZM=ZL?ZL?ZM=[N=]M=^O<\O<\O>YM=WMCWLHXNOZPX]Ub`Yia\rjf�okŽsp›vq§{u³ƒ~ÁŠƒÉŠƒÇˆÀŠ~¼ˆy²ƒr¨†r§�t«‡l£†\„�HL¤B9¶A:ÍC@Þ@?è:;ô<>ÿDE÷><ó=<ì>=ê@AëBIêCMëANãALÐ>>Å@9¾?6·>3°</²<0ÀD8ÒMDÇ61Ô;6Í2.Ì3.ÖC;Ç?3µ6'»>,º7%Á:'Ã<)Â;(À9%Â9&Å:'Ç:(Ë:'Ð9(Ö6(Ø2$Ú."à/%ì7.õ=5õ82÷40÷0-ø**û&(û%'û%'û%'û&*÷!)ø".ÿ'7ÿ(>ÿ!;ÿ6ý1ÿ'8ï#,æ#+ã+-Ý--Ô+(Ò-)Ü41â62è64î66ò65ò12ï./î,,å/,Ü92Ñ:1Ò91ìSK»"Ä+#ßF>È.&Í.(Þ<7èE@â:7Ø/*Ú.*â51å<7ß>4Õ<.Ï5)Ë3&Ì5*Í;.Ë;0Å8.ìbXàZOÈC:¹6,º:/¸8-±3'´6*°0%°2&®2&¯4%°6'²:*´<.³=1°:0ºC;Ö\Wìppñsvöx|ûyƒõq|ÿ|„ýw€út}üs{ÿr~ÿr}ÿmxÿgsü`nôI\ÿ>\ÿ>[äAVÔek´•€’¡zƒ�m}™f~”c‡—j”Ÿwž¦� ¨ƒœ¨„’£Š£|€Ÿvzšqyšo|�r� vƒ¤u‚­wƒ³y‹¸‘¼„“¾†—¾‡œ¾Œ¢½�§¼“¯½š¾Â§Ï̹ßÓÇãÒÊÜÇÄÔ¿¼Ìº¶Á²«¹ª¥ZM=ZM=ZM=ZM=ZM=ZM=ZM=ZM=XK;XK;XK;YL<YL<ZM=ZM=ZM=ZL?ZM=\L=]M=]N;^O<\O<\O>]P@[OC[NFYNJZPQ\SX_U^_Wfd\tibƒng‘qj�xp¬€y¼‡€ÄŠ�Ä�€Á�~º�t«Œn¢—q¤¡u¨šjž™W{�=Aª5+±5+Â<3Ó@8Ý>8è?:óA=õ=;õ;<ô;@ñ>Bî@IèAIâ@KÚAFÇ<9½=4¶<1±;/«:,®:+½A5ÐJAÚG@ÜA=Ô63Ð51Õ@9ÍC8»9+²5#½8'Â;(Ä=*Ä=*Ä;(Æ;(Ç:(Ê9(Ï9(Ï6&Ô2%Ú2%ã5*ê8.ï80ð91ç1&ç/%ê.%ï,&ö+'û+)þ,+ÿ-.ü*-ú%+ú$.ÿ'7ÿ$;ÿ7ÿ3ü0õ0ö.9õ3<å*1Û(,Ü.0Þ02Û+.ñ:>ô7=÷4:ö26ö/4ø03ù25ô87Û2-Ú;5×82äE?äE?Î/+Ñ2.Á"Ô51á?:èE@à=8Ø3-Ù2,Ý60Ý:1Ô:.Ï9*É5'Ç5(È9+Ê<0Ê<2Æ9/ícYáXNÉ@8º4+Â91Ã:0¿6,Â9/¿5+¾4*½4*»5)¸6)µ5(´6(°6)´;0»B9ÓYTèljïqtöx|þ|„øv€þ|„üy�ùv~þuÿsÿp}ÿkwüdqý_nöI]ÿ>\ÿ>\éCYÓeh®Žw‰˜oz•bu’\xŽ]�cŠ•k�˜p‹•p…”mp‡]l‰]h‰\g�^k�_q•ey›izŸk�¬t„±vŠ·|�»}�¼“¼‚–½†›»‰›µ†¡³‹­¶—¾½¨ÏÆ·ØÇ¿ÖÁ¼Ï¼¸Æ·´º¬©²¤£\L<\L<\L<\L<\L<\L<\L<\L<ZJ:ZJ:ZJ:[K;[K;\L<\L<\L<\L=\L=]K=]M=]N;^O<^O<\O<_RA]QA\PD[NFZOK[PN]QS]RZ_SgcXvh]…kb‘sk¤vµ†}¾‰}½—…Á™€·›v©žpž®u¢ºv¥µi˜±Rt¸<D¾5-º4+À<0ËC7ÓE9ÛC8ä?9ñ=<÷:>ù:Aö=Eï@GæAHÙ@EÏ@BÀ;6¸90±9+­9*©8(­9*»?3ÏF<ÙD>×96Ù74Ú;7ÜC=ÝOEÎH<µ3#¾9(Á:'Ä;)Æ=+É=,È<+É:)Ì9)Ï7)Ï2#Ó/#ß4*î=3ô?6ñ91é4+à4&ß5&â2%å/$ë,$î+%ñ+(ò+(ó+-ð(+ò&/ø(6þ%:ý7ü3ø3ô%7ò0;ó5Aò8Cð;Dð<Eé5>ä,6ð2<ó/;ö.9ö-7õ.3õ.1ó/1í42Ü0,â=9Ú64ß;9ÿmjåCAÑ/-Î/,Ù:7ß@<àA;Û<6Õ7.Ö5-Ö5+Ò6*È6'Ã6%Â6'Ä8)Å;.Æ<1Æ<1Å;1ë`YáTMÉ<3¿0(Ë;3Ñ>6Î;3Ò=6Ø=8×<7Ô=6Ï<2É;1Ã9.¼6*¶6)¹;/¼@6ÏTMãgeíorøz~ÿ�ˆû|ƒû|ƒûy�ûx€ÿw�ÿs�ÿn}ÿgvû`pü[mõG^ÿ<Zÿ>\ìFZÎ`až~ev…Zi„Qh…OmƒRt…X{ˆ]}ˆ^x„\q‚X`zM_Pa…Ue�Yn•`uœg} j~£mƒ¬r„²tˆ¶x‹¹y‹¹y�¹z‘º~”º�“±•¬€›©†©­’ºµ¢Ç¹®Ë¸²È¶´¸¬¬ª¡¢ —š[K;[K;[K;[K;[K;[K;[K;[K;ZJ:ZJ:ZJ:[K;[K;\L<\L<\L<]K=]K=]L<^M=^M;^O<^O<^O<^Q@^Q@]OB[OC\NE[NF\OI]OO^P_bTmdY{i^‰pgœ|t­…{·ˆ{³™…º¥…´¯€ª¸z¡ÆxŸÑuœÌcŒÈMlÜFRßA@ÕA=ÏE;ÌH;ÍG;ÓE9ÞA:î@Aõ<Aú=Dø?GðAHâAFÒ@AÆ?<¿?6¶<1¯9+«:*ª9)®8*¼>0ÎD:Õ<7Õ31à;9åA?áD?èSLßUJ¿:+Á9)¾7$À7%Ä;)É=,Ê=,Ë9*Î8)Í4&Ò2&Ù2)å9/ñ>7õ>6ð5.ä/&à6)Þ6)à4*ã1'ç.)ê-)ì*(ì**ê),è',ë'1ò)9ö&<÷!;û=û%Aû6Hä(6ç.<ÿP\ÿ`lÿP\ò<Ió9Gè)8í*8ñ-9ô0:ñ27ê01á+*×(%Ø0-Ú72Ô2/åC@ÿspõVSÝ>;åGDÛ=:Ú<9Õ:5Ò80Ñ7-Ð6*Í4&Æ3#¾5#¹6$»8&¿;,À</¿9-À:/Å<2åXQÜMGÊ70Ç0)×<7ÞA:Ú;5ß<5ç>9é=9æ?9á@8Ù?5Ð<2È:.Á9-»9,»=1ËNHàc_ìnoú|ÿ„ˆý~…øy€øy€üy�ÿx‚ÿt‚ÿn}ýetù^n÷VhóE\ý8Vÿ<ZîH\ÇYZŠlRcrG\tB\yCdzIl}Pt�VvƒXr�Xo�Wm‡ZlŽ\o•bu�h{¤l�¨o„ªq…«p…­q…±tˆ´u‰·v‰·v‰·v‹¸w�¹z‘´|�«{’¤|›¤…­«–¾²¤È·¯Ç¸µ´«®¡ž¥—’™[K;[K;[K;[K;[K;[K;[K;[K;ZJ:ZJ:ZJ:[K;[K;\L<\L<\L<]L<^K<^K<^M=^M;_N<^O<^O<^O<^N>^N>\O?^NA]OB^PE^OJbNYbQdcTqdZ}le‘xr¢w¨ƒv¤”}§¬†«Â‰©Î‚žÚw”ál‰ÛXwÚD_óBTúDPïJPáIHÐE>ËE<ÐF<ÙD>èBBð@Cõ@GõBHíDGßCDÏA=Â?7¿@7µ=/®:+«:*ª9)®8*»;.Ì?6áFAâ=;îBBëAAÞ;6ãJBæXLÕK>È@2À8(¼3#Á8(È<-É:*Ë7)Ð8+Î0%×3)à8/é;4í:5í60ë0+å,'ã0,ã1-ç10ì31ó25ö37÷48÷6;ì-5ê,6í.=ñ0Có-Fö)Gý+Nÿ3Rô9Lå3?ð@Mÿ_lÿlzÿWfõCSõ>Pî3Dí3Aí3>é6<ã99Û83Î4*Ç/$Ë2*Ç.(Ð72òYTö\ZòXVÛA?âHFÜB@Ô<7Î70Í6-Ï8-Ï9+Ê7'Â5#¹8#³9$µ:(¸=-¹;,¶6)¼7.Æ=5ÛNGÙEAÎ50Ð1-á>9ç@:ã81ç51ë20î21ë52ç83á:2Ø:/Ð9.È:.¾6*º8+ÈIBÜ_Yënlû}€ÿ…‰ýƒõv}öw~üy�ÿx‚ÿsƒÿk}ýdvù^pôSeôF]ý8Vþ=ZïL]ÀTRz_BWh<Yq?[vCfyKn~Qt�Vx…Zyˆ_wŒay•exšhz m|§q�ªr„«r…«p†©o‡­rˆ°t‰³tˆµt‡µt…·r‰¸t‹¸w�¶{Ž¬xŽ£x™¤‚­­•À¹§ÎÀµÒÄÁ¼¹À¨ª¶œœ¨ZJ:ZJ:ZJ:ZJ:ZJ:ZJ:ZJ:ZJ:ZJ:ZJ:ZJ:[K;[K;\L<\L<]L<^K<^K<^K<_L=^M;_N<^O<^O<]N;]N;^O<_O?_O?`P@aQAbPFbNPcN]_Pe`Usga…to—{t�}q—�v–®ƒŸÌ‹¡Û‚–ær…ëduéPdé;Rù3Lÿ<QûERéDKÕ?@Î@>ÑC?ÖEBãEFèBDîAEíCFçDEÚEAËB:¿?4»?3±;-©8(¨9(ª9)°8*¼:-Ê;3èIEé??ë=>å78Ú2/ÞA:îZPôh[ÚPCÉA3¾5%Â6'Ç9+È9+Ë7+Ò8,Ñ0&Ú6-æ;4é;4ê40è/,ê-+ë--ó49õ3;ø3=ü3=ÿ1>þ0=ý1=ù1>õ0Aò1Bô4Kõ5Nõ0Nö-Oÿ1Xÿ>_æ3HùO\ÿanÿ_nÿWfûP`ôDXé8JóAQê:GÝ2:Ò/2Ê2-Â8-½;+¼:*Ä:/À2(ÙJBúkcÝLGáPKÏ;9Ï;9àLJÔ@<Ê70Ê7-Î</Í<+Ë:)Ä;(´;&­:%¯<)³>-±9)¯3'º:/ÊD;ÖGAØC=Ô72×2.ç;7ì:6è2.í2-ò,-ô,,ò./ï31è71â:1Ú<1Ò>2Â6)º4)ÅB:×XRçjhú|}ÿ†‰þ€„õv}÷xýz‚ÿx‚ÿq�ÿi{ûbtø]oòQc÷I`ÿ:Xý?[ðO_ºQNqX:Sf8]uEa|In�Sv†Y{ˆ]~‹`��g‚—l€œl}Ÿm|¤o}¨p€©o‚ªnƒ©l…©lˆ«qˆ®qˆ°r†³r„³oƒµp…·r‰¸t‹µv‰¬tŽ¥wš¨„°´™ÈïØ̾ÚÑÌÂÂÌ«±Á›¡±ZK8ZK8ZK8ZK8ZK8ZK8ZK8ZK8ZK8ZK8ZK8[L9[L9\M:\M:]L:^K<`J<^K<_L=^M;_N<^O<^O<^O:_P;`O;aP>aP>aP>aP<bOAeMMdMW`O_aUmgb€uq’|v˜€s‘�uŽ­€”͆”ß~‡îqyùclúP[ú;Mÿ*Fÿ2Qÿ<Sñ<Mà<E×@EÕCDÔBCÜCEàBCäADâBDÞDBÔC>É@8¾>3¸>1¯9+¨7'©:)­<*´<,Á=1Ï?7ß=:å99ã03â/2à42Û94æMEülaòh[ÙQCÆ</Æ:-È:.É7*Í6+Õ8/Õ4,Þ7/æ93è62é1/è/,î1/õ15ÿ2Cÿ0Eÿ-Fÿ*Bÿ%@ÿ#=ü 9ö 8÷'Aõ+Eõ.Mõ/Pó+Pô)Tÿ2`ÿAhóCXÿ`mÿhuøVeíKZñL\ëBUÚ2Cë@RÝ7CÉ-1º((±.$«8&¥?&¨@'¸?.¼:,ÝYMá]QÍG>ÍD>Á63É;7äUQÕF@É91È8-É:,È9(È9(Ä=)±?'¦<&¨;&¬=*«7(¬3(¼=4ÓLFÖGAÝE@Û96Ü30é73ì51ë0+ô1-ý,/ÿ+-ú./ô1/í4/æ81ß;1Ø>2É9.½3(À;2ÏNHáa^÷yzÿˆŠÿƒ‡ùz�û|ƒÿ{†ÿx„ÿn~þew÷^põZlëL`õIaü9Wû=YîM]²JGhQ1Qd6\tDf~Nuˆ[Žc„�h†’jˆ—n‰žsˆ¤t„¦t§r}¨p¨n�©mƒ©l…©l…¨n‡«n…­o…¯o�°l€²k‚¶n…·r‡³t‡­t�§wœ¬…²¹šÈƯ×νÙÒÌ»¿Ê¤¬¿’š­[J8ZK8[J8ZK8[J8ZK8[J8ZK8[J8ZK8[J8[L9\K9\M:]L:]L:^K<^K<^K<^M=^M=^O<^O<\O<`Q>aR=aR?bQ?bQ?aP>aO;bL>hOKgNRbP\cWkjfuv”}|œx—‘z–«�•Æ„’Ú}…ðuzþlmÿY[ýDLþ/Iÿ5Rÿ<Uó<Né?LãCKÝBHÖ=@ÛACÜ@AÞ@AÛA?ØC?ÐA9Ç>4½=0µ?1¬;+©8(­<,´?.»@1ÊA7ØC=Ù74á85Þ..â30æ:6Ó0'Ï5)ëUGÿwjë[PÑC7Ê<0Ë;0É7*Í6+Ô:0Ú70ß82å63ç32ê01í12ô36ý4<þ%6ÿ#9ÿ"9ÿ"9ÿ!8ÿ6ý5ø6ï2î9ð%Bï(Gî&Jñ'Mÿ3\ÿEhÿbuÿ`l÷WcñQ]òP]ïMZâ@MÔ2=Ú=FÌ7;»/.®/(¦5'¢;(›?&œ>%ª9'²8+ÒVJ»;2Ä?8¾41½2/ÎC@åWSÕHAÈ91Ä6*Å6(Ä5%Æ5$Ã:(±<(§:&¨7%­9*¬4&¯1%Â?5ÛRJÙIAáHBÞ;6Þ2.ê41í1/ï-+ü22þ,/ý+.ù+-ó++í-*æ1*ß4,Ö8,Ñ=1¿2(¼7.ÈIBÙ\Xówwÿ‰‹ÿ‡Šû€…þ†ÿ}ˆÿv„ÿj|þ_s÷XlöTiêD\õD^õ6Uñ:VàLZ¥HC^K-Mc5Uo?b}JtŒ\€–g‡˜lˆ›nŠ r‹¥vŒ¬z†«w�ªr~§m|¥i}¥g�¥g‚¦i‚¥k‚¨mƒ«m‚¬l®h~°i�³j„¶o‰¶s‰¯rŽ¬x™°„­»˜¿Å©ËʵËͶ½Åžªº‹—§]K7\K7]K7\K7]K7\K7]K7\K7^L8]L8^L8]L8^L8]L8^L8]L:]J<]K=^L>\L<\L<[N=]P?^SA`SB_R?`P@bQAcRBdQBdNAdLBjMIhKMdPYf[lnk†w{ž€…­‰ˆ²�ƒ©¥‡©¾ˆ Ð‚’ây~ðsqújbÿa^ùJWñ>Që:Lî?NïCQê@Kæ<Gå>Fá<Bß<?Ú<=Ö<:Ï>9Ê=4Á;0¸:,§6&¯A0­<,¬6(¿A5ÌF=Ì=5Ð72Ú85Ü71Þ5.Þ6-Þ8*Ý<*×:'Ó6%Ù;0ðSLòZOÙE9Ç8*Í>.ÔB3Ò:-Ô3+ðGBá//î5:ó6=î,7ÿBMñ'3ú&2ÿ)2ÿ(2ù'ý"*ÿ'/þ'/ð&ð)ù+7ê .ì&7ñ-Aç#;ÿXrç-Dÿ]lÿ_iþZcùU\õPWëHMÝ?@Í84Ë>7ÄA7·?1¦7&™2!™6#š;'Ÿ<)§8-®7/ÂE?¿;7½31Ä64Å54ÔDCÛLHÚKEÑD;È:0Ä4)Ç5(Ë4)Å5*¼</³9,­/#«(³-$Ã9/ÒC;ØH?äPFßE=Ú70Ý0*ç0,ò21ù13ý14÷/1ö01ö01ó0.ò0.í2-æ5-Ý9/Õ=0¾2%À>1¹=3Ö_Yésqÿ��û„†û‡Šÿ„Œÿxˆÿh}ÿ_wÿ[uÿUoøLføEcñ:Yô;[æ?YÛ]i‹@;VF-L^6YuBg†Mw–]}œc�g€¡jƒ¦n‚¨o€§n¦m~¥l}¥i|¤h}£f£e£f¥l¦m�«l¬k~­g®f�±gƒ³i‡µm‹´r�°vŽ®|™´‰§½™±Á¤²Á®¦²°�˜žz…‹]K7]K7]K7]K7]K7]K7]K7]K7^L8^L8^L8^L8^L8^L8^L8^K:\K;^L>_M?^N>[N=[N=\Q?^RB_SC]QA]OB_OB`NDbPFcOHeNHiMJgNQeS_g_tnn�y~¨€‡»‡ŠÁž•Ê£‹»«£¹x�Óyƒê|{ôvjöi`ø]aóOZìFRòHSøIVõEPï?Jë;Eä9?à:<Ú::Õ=8Î?7Ç>4À</·<,«<+ª?-©8(²:,ÁA6Å<4Ç61Ô<7Ò50Ö5-×5*Ø4(Ù8&Û:&Û;%Ú9'Ý6.æA;ëMDáI<Ð>/Ç8(Ê8)Ó;.Þ;4öJFä01ð6;ö6Añ.<ÿ@Nï'4ó$,÷%(û),ý+,ü*+ø((ö((÷+,ï$'í%(ê%,ý;DÙ'ð4CüARØ&6ÿjuÿbjþZaüW]ôOSåBC×<8Ñ>6¾8,¿D5¹H6©>,™/�)�(’(ž-%°:6ÑTPÎHGÃ54Æ45Ñ==ëWUØGBÖG?ÐA9Ç9/Æ3)Ê6,Ð7/Î70Ç=3Â91Â5.Ã2-É4.Ñ83Ø?9ÞE=äJ@ÞA8Ú6-Ý0*ç0,ò21û03û03ô02ó12ô01ó/0ó0.î1-ç4-ß9-×=1¿3$¾</·=2Ó^Wévsÿ��ù…ˆø…Šÿ‚Šÿt…ÿf|ÿ\vÿXtÿQoþIhü@aó:Zñ?_áI`ÁS\{?7RF.M_9ZvCe‡Ks•Yxœ_w�b{¡f}¥i}¦j|¥i|¥i|¤h{£e{¡d{¡d}¡c~¢e|£j}¦l}©l}ªi|«e}­c¯c‚±c…°h‡²kˆ¯pˆ«s‹«|�¬„�ª‡Œ¢‹x‡€dppR^^]L8]L8]L8]L8]L8]L8]L8]L8]L8]L8]L8]L8]L8]L8]L8]L8^K:_N>aP@`P@]P?\O>[O?[RC[QE[QG\OG\OI_PMcQOfTTgUUgUUgV\h]nli†sv£|ƒº‚ŠËŠ�Ô˜‘× ŠÊª‚´¹}ŸÏ€“䄆îvósjûeg÷S\íGQðFOöHR÷EQòAKî=Eä7=à7:Ù99Ô<7Í@7ÅA4¾?0¶>-­>-ª<+«:,¸B6ÇH?À;2»0)Ê:2Í6/Ð6.Õ7,Ø8,Ù7(Ù8&Ü9&à:*Ý4-Û4.âA9éOCÞH:È5%È2#Ù@2â>5õHDè13ò5;ù7@ø4@ÿESø2?ð)0ë#&ì$'ù13õ-/ï''ì&%õ12ë)*ö8:è+1â)1ë3=ÿR^ï=Kÿ[gþ^fñV\îOTðMRêEIÝ:;Õ74Ô?9À6,Á?2¼B5³=1¬7-¥3)¡,#œ' š%©/*ÇECÂ::¼,,È35×ABô^]Ð?:ÒC;ÐA9Ë=3Ì9/Ô=4Ú@8Û@;Û@<Ø=;Ý>;ãA?ä>>à::ã?=ëIDàB9Ý<4Ù5,Ý2+ç1-ð31õ12ø02ô02ô02ô01ó/0ó0.î1-ç4-ß9-Ø>2À4%º8+µ;0Ð[Tízwÿ“’û‡Šúƒ‰ý~‰ÿr‚ÿdyÿ[tÿTqÿNmÿFgû=_ô=\îEbßTg¢AHp;3SH2Qa<]xEf‡Nr“Zu˜^tš_xžc{£g{£gy¢fy¢fy¡cx byŸby a| bz cz¢f{¤h{§h{¨e{©a{©`}«`®`‡²jŠµn‹²s‰¬t†¦w‚žuy“no†lSeYAPM2A>^M9^M9^M9^M9^M9^M9^M9^M9^M9^M9^M9^M9^M9^M9^M9^M9^L8`N:aP<`Q>]P?ZN>YOCXPEXOH[QO_UTbXYeX_iYcl\fk^glaildqok‚rtšy~µ†Ê…ŒÚŒŽáŽ‡×Ÿ‹Ó­ŠÄ³€©½yŽÌ{�Þ�zò‚wÿsrÿcgùU\ôMTõGPôCMò>Gï<Bå6;à88Û97Ô<7Í@6ÅA4¼A1·?.«:*­>-±=0¼F:ÌRGÇH?º5.º1)È91Í81Ó;0Ù=1Û;-Ù6'Ü6&à8+à5-Û2+à=4ìNBãK=Ò:,Ñ8*àB6ß7.ð>:é/0ð16÷4<ü8BÿNZÿLUý>Eò38â$&í/1ê,,ð22è**ë/.ò::Ñæ37Ý,2è9@ðEMÿXcóMWØ=CÚADáBFá>AÞ9=Ü89Ø88Ó97É83Â91½:2ÀA:ÍNHÖVSÐNNÃC@²72¶95ÏKIÙMLãQRïYZêRQñYVÉ51Î>6ÒB9Ð@7Ó?5ÙB9àC<â@;é?@ç8=ì9=ò?Cï:?è58ì>=öMJÞ93Û60Ú3+Þ3,å4.ì30ò21ô01ô02ö/2ö01õ/0õ/.ð0-é4-à8-Ù?3Â6'·5(³9.ËVOï|yÿ“’üˆ‹ý�‰þz‡ýoücwÿXqÿOlÿFhÿ@cû9\ô?^éKdØZh…35g;0WO8WgCa{Kj‰Ps’Yt•\u˜^w�by¡ez¢fw dv awŸavž`xŸ`xŸ`{Ÿaz cz¢fz£gz§fz§bz¨`z¨]|«]~­_†²g‹´nŒ³tŠ­s†¦u€�qtŽii€dQcUDSL8G@^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^M9]K5_M5`P9_P;\O>ZN@XOFXQKYPQ`W\g`hnfsshysg{sg}qh}ol}pq†rw—v}«z‚Á‡Ò„ŠàŠ‹å“�ã�ŒÚ „¿žuŸ l‚¬nq¼vlÒviòsmÿllÿgiþ^`ûRWõFMï<Bê7=æ5;á78Ý98Ö=7ÎA7ÆB5½B0¸@/­7)´@3¯:0°;1ÈRHÕ\QÉMC»;0¾90Á7-Ì9/Ö>3Û=1Ú6*Ü4'â6*à3,â70ä=5åE9âF9ÞB3ÞA2â>2Û0&ê72ì0/ï.1ñ.4ú7?ÿPXÿ]dÿZ_ÿOSá.1ã03à,-÷EEè66Ü,,Û-.ë>@Ü25Ù37ÿbhà=BË*2Ê-4Á&*Õ;=á@EÛ6:Ù37ß9=Ü7;Í/0Ì43Í;;ÕGFÞRSì`cõilôekê^aÄ@>ÆE@å^[øhgÿopÿjjåKK×=;É40Ñ>7×D<ÖC;Õ>5Õ<4Ø:1Ü41î5;ñ-7ð,6ò.8ï,4ë,3ó9<þJKß3/Ü3.Ú3+Ü3,ã5.ê40ï4/ò21ô02ö/2÷/1ö.0õ/.ð0-ê3-â7-Ù?3Æ8*´2%²8-ÃNGï|yÿ’‘ÿˆŒÿ‰ÿw…úl|øatûTnýHgÿ>bÿ:_ú6\óAaàNeÂWap/+_?0[W>^lIe}Mk‰St’\u–_v™_xžcy¡cx buŸ`uŸ`u�_vž_wž_xŸ`{Ÿaz¡bz¢d{¥e|¦d|§`y§^y¨Z{ª\}¬^€¬a…¯g‡®m‡ªp…¥sƒ r{•nrŠjexe[k^RbU_N<_N<_N<_N<_N<_N<_N<_N<_N<_N<_N<_N<_N<_N<_N<_N:^L4_N4`N8^O:[N>YOEZSMZTT_Zahbprm�zuŒ{u‘{q“zp“vp’qt“rxšt}¨w�¶zƒÄ~†Ï„‰Ú‹ŒÞ–�Ü™ŠÍ›…¶¢ƒ£ª„‘«~yŸl[šQ>¼MBÔQIâWRëWUñSRñKMí>Cå4:é9<æ9;à;9Ø=8Ó@8ÊB6ÀA0º?/³9,ºA6«5+£.$¾I?Ùg\ÙdZËUI¸<2¸6)À3)Î:0Ù;0Ú6,Þ3)ä6-ß1*ä71ã;2Ý9/Þ>0ãF5äB3ß9+ß1(é4-ð31ï-.í*.ô17ÿEJÿV[ÿ]bÿ\aâ9<ã9<Õ+,ôJKâ89×//Û57Ì()Ì'+øXZÝ>BÕ9=Ñ6:º#Ä(+Ü<>æAEÝ49Ù,2Þ17Ù05Í(.Ò37äKNöaeûhnùcl÷`iú`løbkÖLLÕNJê\ZêVTçMMãHFÏ42Î3/ÙA<àKDäOHßJCÙ@8×:3×6.Ü.-õ2:û-:ø*7õ'4õ)4ø0:ÿ>CÿJJà1.Þ3,Ü3,Ý5,á4-ç4/í4/ð3/ô02÷/2ù.1ø-0ö..ñ/-ë2-â7-Ù=1É;-³1$±7,ºC=ìyvÿ��ÿˆ�ÿ|ˆÿq�öhxö_rùRlûDcÿ9^ÿ6]ú6\ðFcÑN`¢IMe5+VD0[Y@^lIe|NkˆRt‘[w•_x™`{ždz cvž`s›]rœ\s›\tœ]v�^wž_{Ÿaz¡by¡bz¤b|§b{¦_x¦[x§Yz©Y{ªZ~ª_ƒ­e…¬k‡©m‰§s‰¥u„�s|”ru‰pnlfwe_P=_P=_P=_P=_P=_P=_P=_P=^O<^O<^O<^O<^O<^O<^O<_N:aP6aP6aO9^O<[OA[RK\VV_Zakgvso†}z—‚~¡�|¥}w£{u¥xv§sy©s}°x�¸{„¿~†Ä„‰É�ŽÎ”“Íš“Ç •½«›µ¼¦²É­©Å¦”ªˆmœfJœE2¯A4¶@4ÂB9ÖHDèNNîJKì?Cë=?é;=ä<;ß=:Ù@8Ð@5È>1Á=0½:0¾?6¯4,§1'»I?ÑcVÖh[ÕeWÃOBº>2º4)Ç7,Ö90Ù5,ß2+ç60â0,ã4/â7/Þ7.à>1åC4ä>0Þ4%æ4*é0+ð3/ð..ï-.ò14ó78öBEöMRüY^éFKöQWØ37ëFJÝ8<æAEÚ7:Î.0Ñ05äHKÒ7;Ð7:¹#%Õ;=Û8;å;>ë<Aç6<â-6Ý)2á1;ç<DðLSý]eÿksÿjsú`jöZgùXh÷[fôddèZVèWTØ@=Ï10Ô44Ó33áC@ãJEéRKêSLáJAÙ?7Ø93Û81ã33ó.7ÿ0>ÿ0=ú*7ü.:ÿ;EÿCHûABã1/ß4-Ý4-Ý5,â4-æ3,í4/ñ40ô02÷/2ù.1ø-0ö..ñ/-ë2-ã7-Ø:/Í>0µ1%±7,²;5ésqÿŠŠÿ‡�ÿv†þi}óató[p÷PjúCbÿ7`ÿ7aù;_ìLf¿O[~86]=0NF1VX@ZeEcxMj…RrŽ[v”^yša{ždxžas›]p˜Yp™WqšXqšXt›\uœ]y�_xŸ`x¡_y¢^z¥^y¤\x¤Yx¤Wy¥V{§X©_‚¬d„«h†¨l‡¥o†¢q€šmx‘jq…ij~cbv]_P=_P=_P=_P=_P=_P=_P=_P=^O<^O<^O<^O<^O<^O<^O<_N:cQ9cQ9aP<]P@ZPG\TR_Zaa_mtrŠyx˜�€¨ƒ�°~~°yy¯yu°wwµw|¾x€Á}…Ä‚ˆÄ…ŠÁ��¿—–¾ �¼¬§»¶¬µ¾³±È¸«Ð¾¦ÖÁ¢Ñ»–Ьˆ¯oT®\G¤G5¨>0ÂG?àWQïWVñNOé@Cê=?æ<=â=;ß=:×=5Ñ:1É90Á80Á<5¹:3¸=5ÁOEÇXMÇ]OÑeXÓcUÂL>¼:-Å8.Ò91×4-ß2,ç60æ40á2-à5.ã<3çC7æC4â:-ß3%ê5,æ.&ð0-ï/,ð31ñ54ç-.â24çBF÷X]öW\ÿouåDJëHMà<Cþ[`ÿouÞ=Bÿ}‚ÿmqêQTÔ>@Æ02Ö:=òHKî9>ð7?ø=Fò7Bè-8ô>Jÿ]gÿmwÿjrÿcn÷]gú^kÿaoÿ]nùYeò\]ãPIâJGÕ74Ò/0Ù54Ò0.Ø95ßD?ãJDâIAÙ@8Ó6/Ö4/Ü71è66ê'/ü0<þ2=÷+6ù0:ÿ>Eý?Aì44â3.à5.Þ6-ß4,á3*è3,ï4/ô41ö01÷/2ù.1ø-0ø..ò/-ë2-ã7-×9.Ñ?2¶0%²6,°61èpoÿ‡ˆÿ†Žÿp‚ücxñ\pôXn÷OiûBbÿ9aÿ<döAbäRg°QWd/'WD3GG/RT<U^?btLi�Qq‹[v’_z˜b|�dxœ_r™Zn—Un—Un—Uo˜Vr™Zt›\w›]wž]wž[x¡[y£[w£Xv¢Wv¢Uw£Ty¥Xz¤Z~¦_�¦c¡d~œf{—frŒ]i‚X`vRZoPRgH\L<^N>`P@`P@_O?_O?_O?aQAbRB_O?]M=]M=aQAcSCbRB`Q>dS?aP>]P@^QI[QPZSZeapsq‰yxšz|¥~€±€�¹}ºxz·rs´lo¶twÈv{Ì€ƒÊ‹�Æ–—䤾²°»¼¹´ËƳÓÌ°ÙϬÛϧÜÏ£ÝУÙ˜٘Ѩˆ¸€g ^FËzgÖue½H?×QN÷ccéKLå?Aâ89ä::â:9Ü75Ù61Ö;6Ë81À5.ÇB;»<3¿I?ÆXK¹OAÎdVÑcTÌXIÅG9Ä:/Í6-Ø7/á6/ã4/ß0-ß2,à70ä=4à<0Ý7)ß5(å7,ä/$ê/&ì/)î1-î20ë40è51â64×79æJNõY]íNS×8=×6<ùX`ÿ~„ÿ~†ÿouÿgoÿipÿflõ\añX[ûZ_ÿUZÿDKü8Bû7Aö2>î+9ì0>ñ=HÿS_ÿ`iÿgrÿepþbmÿcnÿ`oý]gíSSßF@Ö85Ö41×11×/.Ø31Û94áD=ÜB:Ù?7Ö<4×90Û81à93æ66ð7<ó4;ó4;ö4<õ6;ò59ë33ã1/Þ3,Þ5.ß7.à5+ä3+é2,ï2,õ20÷12ø03ú/2ù.1ù//ó0.ì3.å7.à?5Ì8,¿6,¯/&¸;7ÒXWÿ˜šþ{ƒÿm€ÿe|ø\rðPhôGeüCeÿ;dù9`õNlÓRd‰=?P, C;&CG.KM5Y_CZjEcxMqˆZx’bx–`t•\r–Yq˜Yl•Sm–Rn—So˜Tr™Xt›Zuœ]uœ[t›Xt�Wv Xw¡Wv¢Wv¢Uu¡Tt S{¥[{£]zž^u—[m‹Ub~MXrCSj>AW0?U/<R,[K;^N>`P@`P@`P@_O?`P@aQAbRB`P@_O?`P@bRBcSCbRB`P@fVG_QF\OG^TSaZad`omkƒvw–wy¢{}®}¸x|¹quµkp´lo¶nq¾uvÐ|~Õ‰‰Ó˜˜Î¨§Ç¸¶ÁÉźÓαÛתâÛ¥çÞ¥æÝ¢äÛ¤âØ£ÛÑ ÔǛӼš»œ€©}dÅ‹wÊ}m´VJÒa[örpìZ[çKLã?@ç=>ë>@ç;;â66Û64Ö=8É83Ê?8¹6.¸>3½K@¶H;È]MÝo`Ô`QÆH:½5)Ä0&Ô6-Ü5-Û0)â51å95ä=7à<3Ü8.Û5)ß4*ã5*ì7.í5+í2+ë0+ç0,ã2,á4.Ý52Ö87Ø>@äJLóW[ú^bü\dþ]eÿaiôS[ïNVñQYÿ_gÿjqÿflõZ`óNTí4:ñ,5ô*6÷-9ñ'5å!-æ'6í7CþPZÿZdÿblÿdmÿenÿcmý[hõQZåEEÞ<7Û64Þ63ß55Ü30Ú2/Û62à>9Û>7Ø;4×:3Û81ß82å95ç:6é9;ê7:ë7:î79î79ì65å31ß2,Ü3,Ý6-ß7,â6,å2+ê1,ò1,õ20÷12÷03ú/2ù.1÷//ó0.ì3.å7.ß<3Ô=2Â8.¶1*·74ØZ[ÿ�“ûs}ÿfzÿ^vùWo÷OiùFfû?bü<cô@cåOh¹KVx;8L2#<:#<B(EG/OS8VdA`rJnƒXu�_v“]t“Xq“Vp”Tk’Ok”Pl•Qm–Rp—Vr™Xsš[sšYt›Xu�WuŸWuŸWtžTržQq�RpœQn˜Pl”NhŒLb�FXu?Ni6F]/@W+@W-@W-AX.YL<[N>^QA^QA^QA^QA_RB`SCaTD`SC_RB`SCaTDaTD`SC^PCbUM\QM[QRcYbkdtqm„vu•y{¤y{¬y~¶x|»nsµei°`g¯jn¸ruÆ}}ׇ„Ý—”Û©§Ø»¹ÏËÊÈÜØ¿åá´éäªíè¦ðé¥íå¤çá§âÛ§×Ï¡ËÃœ¸¯�©›�›ƒk¨ƒp§o`œRGº]UÝmi÷uuï__äJLå@DëADì>@è8:â66ß=:Ô;6ÔA:Â70·7,·>3®=/¶F8Ûj\ßi[Ù[MÊB6Å2(Ñ4+Ù6-Ù2,Þ52éA>ìGCà=6Õ2)Õ1'Ú2'Þ0%æ4*è1)ç.)æ/)ã2,â7/á=4àB9Æ-(Ê43ÜDCðVXú^b÷ZaóS[ñQ[èFQêHSïOYü\fÿmvÿntôY_ã>Dé.5ô-4û2<ü2>õ+9í'4ö6EÿKW÷ISõPWøW_þ^fÿckÿ`i÷RYéBIÞ97Þ71ã75é<8ë;;æ95á51Ý60Ý:3Ù;2×90×90Ü71á83ç:6é<8ã99ã99å97ç98è88æ74á40Û2+Ú6-Û7-Þ8,â6,ç2)ë0)ó0,÷10õ12õ13ø02÷/1÷//ò0.î3.ç6.Ý90ÜD9Ç80»4.µ/.ä_bÿƒˆølwþbxþWsúPmþLlÿCfö:]ó;_ìIfÔSg�DJl;4O>,=<':?(ED/FH0Q\<YkEg|Qr‰[t�Zq�Um�Rk�Oi�Mi‘Ki’Lk”Pm–Tn—Up—Xq˜WuœYu�WtœUrœTpšRm—Mk•Mj”Le�GaˆE[AUt;Li3E`-AX*>U'B\/E_2Ic6WJ:ZM=\O?^QA]P@^QA_RBaTD`SC`SC`SCaTDaTD`SC_RB\PDZOMZQVbXcjcsqm†vu•xz£z|­w|´sy·mr´bi¯]d¬ag±lr¾x{ʉ‰ß•‘夡䶵áÇÄÙÕÑÎáÝÄçä¹ëæ¯îê­îè¬éâ«ãÛªÚÓ©Ë¡»¶™¥¥�›™„“ˆv˜�q™rc›bW³i`Ïrmû‹‰õutê[]æLNèCGí@Dì<?é;<ß:8Û<8ÞE?ÓC;Ã:0º:/°6)§1#¿I;ÛaRêk\ÛSEÉ:,Ì4'Ô6+×4-Ù42ñMKþZXíKFÙ80Ö3*Û4+Þ2(ß.&á,%á,%Þ-%Ü1)Ú6,Ù;0×?4Â/'Ä4,Î:6ÙCBßFHàEIäGNêMVñQ]öVbøXbù\eÿgpÿmuôYaÞ;@ê17ñ.4ò-6ð+4ë%2ì(4û=IÿS_í?IèAIêGLôQVÿY`ÿX]ôGMä68á53ä73ê;8ò@>ô@?ï=;æ95à72Ý82Ú91Ù80Ù80Ü71ã73è96é:7â:7à;9ã:7æ:8ç:6å84à5.Ú3+Ú6,Û7-Þ8,â6,ç2)ë0)ò/+ö0/ô22ô23÷12ö01ö0/ò0.î3.ç6.Ü8/àF<É91¾3.¶,,îfjÿy€øgtÿbyýTqþKkÿGjÿ>cò6Yë=^ãQhÄXeˆBBgC7[N;GF1AC-JH3EE-LU6Sc>btLlƒUp‹Xn�Rk�PiŽKg�Ih�Jh‘Kj“Mk”Pm–To–Wp—Vt›Xs›Uq™So—Pj“Mg�JeŽHd�G`‡D]�AWy=Rq8Mi6Lf6Ne7Of8Lf7Oi9Sm>UH8WJ:[N>\O?]P@]P@_RBaTD_RB`SCaTDaTD`SC_RB_RB^QIVMR_Xhnf{tp‰xt—xw¡xz«y{´sy·jr³`g­[c«_g°jr»v|Æ€„Γ•àžžä­®ç»½æËÉßÔÓØÝÛÎáßÆåá¾æâ¼äß¹à׶ÚеÎƱ¾µ¦­§›˜šŒŒŽ€‰…yŠ|q‘vk›si¨meºoi扄î�~ósrñedñWYðMNîDEì@@ã:7à;7äB=ÞE?Ì<3Æ=3ÄB5¯1#®0"ÇI;ÜXKÖN@Ì=/Î8*Ð6*Ñ3*Ö42õUUÿkhú]XãD>Ü;3ß;2ã80â4-ã2,â1+Ý0)Ô0'Í/$Æ/$Â0#Å9,À4'Á3)Ë:5×CAÛEGÛBGÚ@HàDOëOZòVaòVaú^iÿgoóYaáBGê:<í57í38ï28ñ3=÷9CÿHSÿV_ä6?Ý4;ß6;ëADùJOüIMô;Aé13è51ë95ï<8ò<9ó;9ð:7é73â70ß;2Ü;1Û:0Ü90ß82ä73è64é75æ:8ä;8æ:8è:9è;7æ95á6/Ü3,Û7-Ü8.ß9-á7*å3)é1)ñ0)ô1-ó32ñ33ô22ô01ô1/ò0.î3.ç6.Þ7/àC:Ê70À1-À13÷jpÿqzúftû]túPmÿCfÿ=dÿ9cõ8^æEdÙZm«V[wE>dJ=aYFTO;KI4PI6GE.GP3L\8ZlFf|Nm…SmŠRlŒMh�JfŽHg�Hg‘Ii’Lj“Ol•Sn•Vo–Ut™Vs˜Un•Rj’Lf�KcŒHaŠF_ˆFZ�BX|?St;Pn8Ok:Rl=Wn@ZqCUo?Vq>WqARE5UH8XK;ZM=[N>\O?^QA`SC`SCaTDaTDaTD`SC`SCbUEaVP^Vcje|yu�}yšzy¡xy©vy°sw´jr³cj°[c«]e®hp¹v~Ç�‡ÑˆŒÓ™�Ý£¥à¯³ã»¿åÆÇãÏÎÞÔÓØ×ÖÑÙ×ËÙÖÇ×ÑÅÒÉÀÌþŹ¹´¨¬£š�Š‹†x{tyvqulele�ogŽa[�`[ÀsmØyuð{ü|yÿrpýccöSTðHGïC?ê>:á<6àA;Ñ:1ÓC:ÝTJÂ>2¸6)½9,Â=.È?/ÏB1×C5Ö@1Ñ7-Ð2/êLMú_]òWSßD?Ù:4Û81à70æ93ç92ç92â:1Ú9/Ñ9,Ê8+Ä8)°(¸3"È@2ÓI>ÒD@Ë;:Ì7;Ï:@Ñ7AâHRíV_ðYbõ^g÷`iêU[ÜAEå@>ì?;ñACúGJÿLTÿQYÿT]ÿV^á4:Ú/5Ü/3ç7:ô=Aø;?÷48ó12í42ï95ð86í42ë20ê20ç40á4.ã<4à=4ß<3Þ;2á83ä73è43è43è88ç98è88ë99ë97é75â5/Þ3,Û7-Ü8,Ý:+ß7*ã2(ç0(ï0(ò1,ñ42ð43ó32ó11ó1/ñ1.ì3.ç6.à91Ù<3Ì71Á0-Ñ@CÿnuýkuùbqðRiõHfÿ;aÿ4^ÿ6bû>däNiÊ\i‹GFjF:]L<]VCXQ>RJ7OG4ID.DJ.GU2Sc>^sHgMj‡OjŠKgŒGgŒFfŽGf�Hg�Ji’Nk”Rm”Um”Uq•Uo”Qj‘NgŽKcŒH`‰E_ˆF^†G\‚E[~DWxAVr?Uo@YpB]rG^uG[sCYtAXs@NB2QE5UI9WK;XL<ZN>\P@^RBaUEbVFbVFaUE_SCaUEfZJh]YngwupŽ|x™{y wx¦vw­pu¯jo¯ah¬^f®_g°em¶pxÀ}…͉�Ø�—Ø™ Ö¡¨Öª±Û´¹ßº¿ßÀÁÝÃÃÛÇÆØÆÃÔÅÁÐÁ»É¼³Ä»°Áµ¨¼¦˜¯–‹œˆ…Œqqqtpopfewhe†nj„_Y•d_›ZT¹f`Ùsnðzvÿ{wÿwsÿhfüZWþRNõHBæ=6çD=Ø;4ÛG=ê\RÅ;0Ä<0½7+¾6(Ç>.ÑD3ÔB3ÔA1×A3Ò86Ø>>ÝCCÜB@Ö=7Õ81Ø7/Þ7/å:3æ93å:3â;3Û>5ÕA5ÑE8ÌG6®-¼;%ÍJ8ÑL=Æ=5»0+À00Ê9>×AJáKTëU`ðZeð]eìYaßLRÔ>@Ï4/Ø7/à;9ê@@òCHõDJñBIî?Fâ5;ß26ã36ï8<õ79ø14û/2ü14ï20ð95ð74ì0.è,*ç0,æ3.à3,ä=5à=4à=4à<3ã:5æ95ê65ì65ì57ì57í57î68î66ë54ä2.Þ1+Þ7.Ý9-Ý:+ß7*á4&æ1&í1(ñ1,ð42î53ñ42ò21ó1/ñ1.ì3.ç6.â;3Ò4+Ò:5Æ20çUXÿpxùcnòZiçI`óEfÿ8bÿ/]ÿ5bÿBhÜLe±PYj5/^G7XK:UN;[P>VI8M@0JC0@F,AO.K[6Xj@bxGf�JhˆIf‹FeŠDc‹Bc�EeŽHg�Lh‘Oi‘Rk’Sn’Rl�Ph�Nd‹JaŠH_ˆF_‡H_‡H_…H_‚H]~G]yF]wH`wKcxOczN]uC[vAZu@JC3MF6QH9TK<WK=YM?\O?^QAdWFeXGdXHbVF_UIaWMg^Ulcfwr‰yušxvžtsŸrr¤tu­lp¯bg©Za©_e¯fl¸ms¿v|È�‡Ó�“Ý—œÞ›¡Õ §Ó©­Ú®³Ý±´Ý²´Û´´Ú¶´Ûµ±Ö³­Ñ®§É«¡ÄªžÂ¤˜¼™‹¯ˆ~™|x†dbgiefcYZj[X{c_yVP’c]ŠNF¢UM½`XÖkcñwrÿzÿzwÿnjþa\÷TOéGBîOIÝB=ÞG@êWOº,"¾1'¾4'Ç>.ÑE6Ë>-Á2"Ê7'ÛE7ÞE?Ó97Ì3.Í50Ô;5×=5Ý<4á=4ä;4ä92Þ5.×4+Ï5)Ê8+Å<,Â?-ÔS>ÈG2½<)¾:-Ã=2Ä;5Â74Â43ÜKNÞMRáPUåTYèY]åVXÜMOÓD@É5+Ñ7+Ø7/Ý52â24å26æ18æ.6ì4>ì3;ó5?û9Bý4>ú+5ÿ(3ÿ/9ø-3ù57ù59ð./ë*+é0.è51â5/ã<4à=4à=4á=4ä;6ç:6ë76î66ì46ë35ì25í34í55ê41ã1-Û0(Þ8,Ý:+Ý:+Þ8(à4&ã3$é3&ì4*î50î52ð42ð3/ò2/ï2,ì4,ç6.ä=4Í/&×>9Ì35ùbgÿoy÷[iðNcêC]ùEhÿ:eÿ0]ÿ6aöGhÉJ]“CFR,#VI8SL:OH6[O?UI9F?-HF1>B'@J(IU1Sd:_rEe}Ig„JeˆHcˆCa‰BbŠCc�EeŽHf�Mh�Pi�Ql�Pk�Mf�JcŠIaˆG`‡F`‡H`†I^�G^F_}G`|Ia{Kd{Mf|Nf~N]xC]z@^{CDB6FD8JF;PG>RH>WI>]M>_O?`Q>aR?`SB_VGbYRgb_ojnso~wq“zt |y¨yw©pp¦gg£`a¢\_¤`b­fhµopÀyzÌ„‚Ö�‹ß–’眙栞ݡ¡×££Ù¤¤Ú¦£Ø¥¢×¥¡Ô¤ Ó¥žÑ£�Í¢šÉ¢™ÈŸ–Ù�»Ž‚°ƒ|žrn}igld^`fZZkYUrWP}WNŒWO•RI¨ZPµXPÁZSØkdìyrù‚|ÿ‹†ùuqÿusí_]ßPLÙEAïZTÌ71Å2*Ä6*Ã7(À7'Ã7&Æ9(Ê;+Ð<.Õ?1Ò8.Ö<2×?4Ó;0Í5(Í3'Ø:/â@5à90á90Þ7.×7+Ï7)É:*Æ?,ÄA/Â=.Æ@4ËE:ÍG<ÌC;Å<4»2,¶+&Ç<7ÐEBÙOLàVSãYVáXRØOIÐC:Ì;*Ó:(Ü8,â5.è./ì*2ð(3ô'6ú->ú+?û*?ý)?ÿ(?ÿ&?ÿ$<ÿ$<ÿ$8ÿ'9ÿ+<þ,9ó+5í.5è45ç;9Ý84Ý:3Þ;4à;5ã:5æ95é73ë54ë35ì46ë54é54ç53â5/Ý5,Û5)Ü8,Û9*Ü9(Þ9&ß8&â7%ã6%æ4&ì5-í4/ð50ñ4.ñ5,í5+ê5*å7,à90Ø7/Õ:6Û?BÿmwÿbsûSjõIcÿKjö6[ÿ6_ÿ<gø>cçNj¯IV^$"D, ?:'HE4PM<PI9EB1>C/?G/BH&EK%KS.S]8[kDavKd~NdƒJb†F`ˆBa‰@c‹Be�Fh�Jl�Pm‘Ql‘LiŽHgŒGf‹HfŠJe‰Ic…HaƒGdƒJd�IbGc~GdJdHe~Ga}C[{<\=_‚B=?4?A6FC<JE?PE?TG>[K>]M=`O;`Q<^SA^VIc[Xhcinlyso†tp•vpžtp¢pm¢gfŸaaŸ]^¡\\¤ed°kjºtrÅ}{φ‚Ù�‰à”Žæ˜“å›–Ùœ˜Õ�™Öžš×ž˜Ö�—Õœ–Ô›•Ñž–Ñœ”Íœ“Ê›’Çš�Ä“‰»ˆ®‚x�tk|mdif]^f[YiZSoWM{UJŠVKšVK°\RºYRÁXRÓfaàqjãvoí~wý‡…ÿ‡…÷usîgdãUSëZUÇ2,Â/%Å7)Ã7&À7%Á8&Ä8'È;*Ï;-Ó=/Õ9,Õ9,Ó:,Ò:,Ñ9+Ñ9+Ô8+Ø8,ß;1à90Ü8.Ö8,Ï9*È9(Á:&¾9&¾6*Á80Æ=5É@8ÊA9È?7Å<4Â91¿4-Ç>6ÑH@ØOGÚQI×NFÌC;Ã7*Ê7%Ó8$Ý7)å4,í//ô+3ù)7ý(:ÿ+@ÿ*Aÿ)Aÿ&Aÿ#@ÿ!<ÿ=ÿ:ÿ:ÿ 9ÿ%;ÿ);÷-9ð19ê7:å;;Ü94Ü:5Þ;4à;5ã:5æ95é73ë52ë35ì46ë54è64ä71ß6/Ü5,Ù5)Û9*Û9*Ü9(Ü9&Þ9&ß8%á8%ã6%ç5+ê3+ë4,ë5*ë5*ê6(ä7)á7*ß;1Õ4,Ó54óTYÿesÿ[põGbÿIhÿAdÿ;`ý3[ù;aòMmÍLb‰8?S*$A5'69(;<,B@1DA0@?-=B.>G,=CKP'\b<eoJhvRezQd~Od‚LaƒFa†Aa‰@cŒ@g�DiŽHk�Ml‘Nl‘Kj�Gh�Hf‹FgŠJf‰If…JdƒHfƒKd�IdHd€Fe�Ge�GfEb~A]€>_„?b‡B69.:<1??7DA:KB;OE<VH=YL<^O<\O<\SB^WMc]]helom{sq‰so”qk™jf˜c`•\[”ZZ˜\] \_¦hjµno¿wxÊ€Ò†„ØŠˆÜ�‹à‘ŽÝ“�Ó“‘Г‘Г‘Ò”�Ñ“ŽÑ“ŒÐ’‹Î•ŒÍ“‹Ç”‹Ä”‹À“‹¾Œ„³ƒz¥€r•|ewu^fi[[d[Vd[RfXKrVH„TH£ZQ½`YÉ\YÏYWÛcbákißokãvqûŠ„ÿŠ†ÿ…�ÿ}y÷nhîaXÄ4+À2&Ä8)Á:'À9&À9%Â9&Æ;(Ì:+Ð:,Ö:-Ó5)Î5'Ð8*Ô>/Ô>/Î6(Ë/"Ú:.Ú8-Ù7,Ö8,Ð:+É:)Â9&¾7$½5)À6,Â8.Ä:0Ç=3Ê@6ËA7ÌB8Ç:1Ê@6ÏE;ÐF<ÒH>ÑG=Ê@6Å7+Ì6%Ö6&ß7,ç4/ñ03ö-5û+9ÿ*<ü';ü'=û%=ù#;û!:ú7ü7þ6ÿ8ÿ6ÿ$8ý*;÷2<ñ6=ë8;ä::Ü94Ü:5Þ;4à;5å95ç85é73ë52í36î47í55ê65å61á6/Ü5,Ø4(Ú8)Ù9)Ú9'Û8%Ý8%Þ7$à7$á6%á5'â4)ã5*ã6(ã6(á7(Þ6)Û7+Ý<2Ð1+Ø88ÿkrÿ]nþPiñ<[ÿHlÿ7]ÿ>eù5[î>bæVqª?Qa#(L/);7+07'37&9<)@=*A<(@>)BB&?FXa4s~T€Œdw‡`hSa}La€G`ƒC`…?a‰@cŒ@hŽEj�Jk�Ml‘Nn“Mk‘Hh�HgŒGh‹Kh‹KfˆLd†Je„KdƒJc€Fd�Gf‚Hf‚Ge�Fb€Bc†DeŠEh�H25*58-;<4A>7G@:KB;RE<VJ<XL<YM=YQD^WOc^bigron€roŒrm•jg”`^�WW‹TUŽWX˜\_¢ac­ln»rsÃ{|΂ƒÕˆ†ÙŠˆÛŒˆÜ�Š×�ŽÒ��Î�ŒÏŽ‹Ð�‰ÑŽˆÐŽ†Ï�…Î�…ÊŽ…Æ�…Á�‡¼�‡¶‰‚¬€x�oŒ€_p|Y`lXYd[T^_Q_\IjWF}SEžUL¾\YÐY[ÙX\äcgçkkáplåxsì}vízsþƒ~ÿ†�ÿ�{õkaÉ<2Ä8)Â;(¿:'¿;&¾:%Á:&Å:'È9)Î8*Ø:.Ô6*Î5'Î8)Ñ>.Ñ>.Ë8(Ç/!Ò6)Ó5)Ò4(Ò6)Ñ9+Í<+È=*Ä;)Ä:-Ä:/Æ90Æ9/Ç:1È;1É<3Ë=3ÓD<ÓE;ÐA9Ë=3Ë<4ÏA7Ï@8Î;1Î5'Ù5)â5.ë31ò/5÷+6ù)7ü):ù&9÷&;÷&;ø'<û&<ý%:ÿ$:ÿ#:ÿ"7þ#7ú%7ù,;÷5>ò9?è8:à87Ü94Ü:5Þ;4à;5å95ç85é73ë54î47î47î66ê65æ72á6/Ü5,×5*Ø8*Ø8(Ø8(Ù8&Ú7&Ü7$Ý6$Ý6$Û5%Ü6(Ý7'Ü9(Ü9(Û:(Ù9)×:+×:1Ð3.ëHKÿoyÿVjøD_ø<_ÿAgÿ3\ÿ8bø>cèMlÅOe€0=J C1-11'.6'4=,=B.B@+B<$E>$HE$U\0p~M�Ÿr—¨|ƒ˜mj„W_{J]~E_ƒCa†@bŠAf�Ck�Jl‘Lm’Om’Oo”Nl’IiŽIh�HiŒLiŒLhŠNg‰Mf…LdƒJd�Ge‚HgƒHh„If‚Gd‚DgŠHiŽIl‘L-3'36+891<;6B=9G@:MD=QG=SI=SK>UNDZUQa^eigumk�lkŠjh�b`�XVˆPP†QQ�XYš`cªfj´orÁuxÉ|Ђ…Ö†‡Ù‡ˆØŠ‰Ù‹ŠÖ‹ŒÏ‹ŠÌŠ‰Í‰‡Ð‰…ш„ÑŠƒÑ‰‚ÐŒ„Ï‹„È�…Á�†º�†°‡�¥~w–m…„Yj€SZnVTc\RZaOZ_HeZDxUB’LD¸TRÐSWÛRYå^dèejãklæuqäunâoh÷~vÿ‚{ÿ‡~ôl`É?2Á8(À;(¿='¿='À<'Â;'Å:'Ç8(Í7)Ø:/Ù8.Ò8,Ï9+Ì:+É:*È9)É7(Ï9+Ñ7+Ò8,Ñ7+Ï7*Ì8*Ë<,Ë>-Ê<0É=0Ê<2É;/È:0Æ8,Å7-Å5*Ð@7Ð@5Í=4É9.Ë80Î;1Í:2Ì5,Ð/%Ù0)ä20ì25ô/8ù-9û+;ü,<ü-?û.?ü/@ü/@þ/Aÿ/?ÿ-?ÿ.@ÿ)9ü)8÷+7õ0:ñ6=í8=ã77Ü43Ü94Ü:5Þ;6â:7å95ç85ê65ì44ñ48ñ48ï56í76æ72á6/Ú6,×5*Ø8*Ö9(Ö9(Ø9&Ø9&Ù8&Ú7&Ù8&Õ8%Õ:&Ö;'Õ=(Ô=(Ô=*Ô=,Ô<.Ñ7/Ú;8ÿ^eÿaqÿPi÷<[ÿAiÿ4_ÿ4`û1YôKlÛYq•>NZ%-C),:2/-0)08+7@/:B+<<"A<UH(`[5{…S�£lª¾‹¨¾�Ž¥wo‹[aK^F`„DcˆBe�Dj“Go”Np•Po“So“So”Nl’IiŽIgŒGiŒLiŒLi‹OhŠNf…Je„IdƒHe„Ig…Ih†Hg…Ge…DiŽIi‘Kl”N+1%.4*470893>:7B=:HA;KD<NH<NG=OJDWSR_^ffeugggf†ba‰[YŠQQ…NO‡RS“[^£ei³lo¾uwÊz|Ñ�Öƒ…؆‡Ù†‡×ˆ‡Õ‰‰Ó‡ˆË†‡È†…Ë…ƒÎ…€Ð…€Ò‡Ò‡€Ð�…ÒŒ…Ë�…ÁŒ…¸‹…«„�{v�|jz„Vc�PVnUQd^P[eMXbIb^EsXC�OE´WRÐUZÚSZç\cèagågjèpoìyræqh÷|tósjþxmæ\QÄ;+¾5#Â;'Á=(Á=(Ã<(Ä;(È;)Ë9*Ï7*Ö5+Ø7-Ô:.Î:,Ç:)Ä8'Æ:)Ê=,Ï=.Ö@2ÙA4Ö>1Î8*È4&Ê8)Ë<,É:,È:,É;/É;-Ê:/È9+È5+Ç5(Ê6,Ï;/Ò>4Ó?3Ö?6Õ>3Î7.É,#Ò+%Ü-*ç02ð39ö1;ú0<ý0?ý0?ù0@÷1@÷1@ö0=÷/<÷-9ú,9ú*7ý/<ù/;õ0:ð3:í6;æ69ß55Ù42Ü94Ý:5ß:6â:7å95ç85ê65ì44ò59ò59ð67í76æ74ß6/Ú6-Ö6*Õ7+Ô8)Ô8)Ô9'Õ8'Õ8%×7'Ô9%Ð9$Î<%Ï=&Ï?'Î@(Î@*Ð?.Ò;0Ñ61ëHIÿfrÿOdÿHfû7[ÿAmý/]ÿ7aò4XãPj¼Ufm19E(*E697325406906=-3;#38CCja:…�TŸ®w©Á‡µÎ—¯Ç“’®}v”be†Q`‚F`…BcˆBg�Fm–Jr˜Or—Rq•Up”To”Nl‘Kh�Hf‹Fh‹KiŒLi‹Ni‹Og†Kf…Je„If…IhˆIi‰Hi‡Gf†Cj�Jk“Mn–P(0#+1'/4-350764<85@=8C>8IE<HE<KHCRPQ[[eabtbb|``‚_^ˆZXŠTTŠTT�\] gi³psÄvxÍ{Ø~‚Ûƒ„Þ„†Ý„…؃„Ö…„Ô†…у†Ë‚…ȃƒË‚�̓€Ñ…€Ôˆ€Õˆ€Ó�…Ò‹„È‹ƒ¾‹ƒ²‰‚¤�|“wp€veo~U[|PQnUNf_O]eMZdIaaGq]E‰TF­]TÈZ[ÔVZâ[aæ]déaeîllõzsírj÷xoÞYPå[PÒD8Ã5'Ç:)Å:'Ä<&Å='È=(Ê;*Í:*Ð8*Ô6*Ó0'Ô3)Ò8,Í;,Ç;*Â;(À;(Ã<)È;*ÕC4áK=ÞH:Ó=/Ê4&É5'Í;,È9)È9)È9+È9)Ê8+Ê8)Ë7+Ë7)Í6+Ò<.×?4Ø@3Ú@6ÛA5×:1Ò1)Ü0.å14î49õ6>ù4>ù1>ú0@ø/?ò.<ï/<í/9í07ï-5ï,2ò+2ñ*1õ2:õ3;ò5;ì59å57á55Þ65Ü75Ü86Ý97ß:6ã:7æ87é77ê67ì46ô5:ô5:ñ7:î87æ74à70Ú6-Ô6*Ó7*Ñ8*Ð8*Ð9(Ð9(Ð9&Ñ8(Ð9&Ë9"È:"È>$Ç?%È@(È@*Ì=-Ï;1×96üU\ÿ^qÿGcÿ?aÿ:aÿ7gþ3bû8`íFdÅNa‰@IS12A64@:<:46?56<7179+4="9CSZ.‚‚P¢§q±ÅŠ¯ËŽ°Ì’¦Ã�’®}{™egˆQ_�E^ƒ@c‰@i�Do•Js™Ps˜Sq•Uo“Sn“Nk�JgŒGeŠEgŠJh‹Ki‹Ni‹NfˆKd†Id†IeˆHgŠJh‹Ii‰Hf‰Ej’Ll–No™Q&.!(0%-2+13.331764:95<;6GD=DD<GFBOOQZZd_ap^`y\\~`_‰\ZŒYX‘\\šgg­qsÀz|Ñ€Ü�„ႅ䅆ㄅჃÛ�‚Õƒ‚Òƒ‚΄‡Ìƒ†É„„̃‚Î…‚Ó‡‚ÖŠ‚׌ƒÔ�…ÐŒƒÄ‹‚·Š‚«…›xˆsnuqbevWUtSLjVKf_MbfM_eIcbFp^FzQ?�[M·ZSÅSR×VZâW\êY`òce÷rmñnf÷qhÑD;ÔA7Å1%È4&ÔC2Ç9%Ç:&É;'Í<)Ð:+Ó:,Ø8,Ù7,Ï,#Î0%Î6)Ë<,Ç@-Á>*¾;'¼7$À4#ÑB2ãOAäN@ØB4Ë7)Ë7)Ð>/Ê;+É:*Ç8(Æ7'È6'Ê6(Î8*Ð8+Ó;.Õ;/Õ9-Ò6*Õ7,Ú<1Ý<2Ü71è88ñ8=÷:Aû9Bú4Aõ/<ò,;ï,:í/;ê19ê38ë48í57ñ56ô36ó57ð37ï6;í9<é69â45Ý33ß76à;9Ý86Þ97á96ã:7æ87é77ê67ì46ô5:ô5:ñ7:ì89æ95ß82Ø7/Ó7+Ð8+Í:*Í9+Ì;*Í:*Í:(Í:*Ì;(Å9"Â:"Ã=$Â>%Ã?(Æ?+Ê<.Ï81ß<?ÿ\fÿQhÿIfÿ9]ÿ=hù.]ý<hé;\å\p¦LV^..G83BC><89?48E26C41=:)>E&FU*bs?�›a­½±ËŒ«É‹¦Ã‹�º„Ž¬z{™ef‡PZ|@^�?a‡>hŽCo•Jr˜Or—Rp”Tn’Rn“Nj�If‹Fd‰Df‰Ih‹Ki‹Ni‹NfˆKe‡Jd‡GeˆFgŠHh‹Gh‹GeŠDk“Lm—OpšR(0!'/"(-&).(.0-3317839:4==5?@8DE@JLKORYVXeZ\s]^}YX‚[Y‹^]–ee£nn¶wxÈ�Ø„…⌌ðŠŠì†‡ä‚‚Ü€€Ö‚€Óƒ‚΀ƒÈ�„ǃƒË„„ΆƒÐˆƒÓ‹„ÒŒ„ÏŠ€ÅŠ€¼‰€¯‡}Ÿ€x�wozkegf]Xm[Qm[Mi\Lf^Kd_IeaHhaGo_FoP;…UA¥]O¿aYÑYXÝQTîS[ø\`ùheæYPÔD<Î70Ñ7/×:1Ö=/Ô;+Ï9(Í:&Ï9(Ñ:)Ô8)Ø8*Ü5,Ü5,Ð/%Í3)Ì8,Ê>/Å@/Á@-¾=*¼9'Â9)¾0"éWJßK?Ã/#Ñ=1Î</Ç8(Ê;+É:)È9(Ç8'È7&É6&Ë5&Ì4&Ð7)×;.Ü>2Ü<0Ù7,Ú8-â>4ìC<õBEò9?ð2<ö4?ö2@ñ-;í-:í2=ì7>æ5;ä68å78ë99ð:9ñ78ï77ò;=î<<ê::æ87á85ß74Ü75Ü75Þ97á98ã99æ::é9;ì9<í9<ï8<ô7;ó6:ð8:í9:ç;9à;5×90Ï7,Ê8+È9)Ç9+Æ:)Ç:)Ç:(È;*Ç<)Á9#Ä='¾:#¶4¼9%ÉD3Ï?6Ê1,ÿ^eÿVgÿIbÿ<\ø1X÷0Yû6cõBiçYo¬FQo33N3,B?6>C<@78@-1L/3K2.A9&DJ(Zn=~š_ž¸w¨Å‚¬ÊŒŸ¾‚˜¶€”²~‚ nf„RXvBY{?]€>c‡Ai�Dn”Im•Ll”Nk’Ol‘NiŽIfŠDc†Bc†Bf‰Ih‹KfŠLdˆJg‹MhŒNg‹Kf‹HeŠEf‹Ej�Ik“Jq›QržSt U*0")/#).').(-/,130561782;<4>?7AD=HJGOPTSUaXYkZ\u\Z^]‰db”kj¤utº~·‡ß‹Œé��ð‰‰ë…„ ؀~Ҁς‚Ì…†É†‡È‡ˆË‰ˆÎ‹ˆÏŒ†Î�†ÌŽ…ÈŠ€¼‰±†|¡‚x‘{s€tknlb`e]Re`Me^Kf^Kh]Ki\Ij^Hk_Gn^Ew_G[C’YF­ZLÇVPÞRSõQXÿX_ðUSãLEÙ>9Õ60Ù6/Ü90Û9,Ö9(Ó8&Ñ8&Ñ8(Ô8)×7)Ü6*ß4,Ý5,Î0'É5)È9+Æ=-ÄA/Á@-¿<*¾9(Ä8)Ë<.äREÜH<È4(Ì:-È:,Â4&Ê=,Ë<+Ê;*Ê;*Ë:)Ì9)Î8)Ï6(Ó7*Ö8,Ù7,×3)Õ1'Ø1(Þ7.ç<5øDGñ8>í/9ï/:ð0=î.;ë1<ë6?ä5:Ü36Ø22ß:8îEBøIFøDCó?>é:7ç98å97â96á85à85á98á98à87â88å99ç9:é9;ì9<í9<î7;ó6:ò59î68ê88ä;8Ü:5Ó9/Ë7+Æ8*Ã:*Â:*À;*À;*Á<)Á<+Á<)½9$Á?)»:%³5¹:'Á;/Ê:2Ø<=ÿ^gÿPdÿA[ù8Wö6[õ;`ô>dãIe«?Lƒ?>_:2G;/:>08>2F<:R>=N0.S8/VK5ciEx�[�°p�Â}¡Æ�š¹}š¹€š¶ƒ�©xrŽ^YuETp=\{B_@b†@i�Cm“Hl”Kk“Li�Mi�MgŒGe‰Cc†Bc†DgŠJiŒLhŒNg‹Mi�Oi�MiŽKgŒGgŒFfŽEl’Im•LpœQo�Rt U-1"-1#,/&,.)//-11/34/45-9;0<>1@B7EHAKMLQRWUUaVWi\Yt^]fdŒpn zx·…ƒÎ�‹ß��ê�ŽïŽŒíŠ‰ç‡„ß�Ó|Ë}ȀƅLjˆÆŠŠÊŒŠÉŒŠÉŒˆÅŒ„¿‹‚·‰€­†|¡€v�{p€ujpndbi_Vc]M_aL_aKc^Jg\Jl[In\Ho]Gq]EwaIxX?‡S=£VDÇXOãUSøOTþMSâ><Þ<7Ý84Þ71á6/á6.Ý7+Ú7(Ô7&Ó8$Ó8&Õ8'Ø6)Ý5*à3,Ý5,Ë3(Ä6(Â9)Á<+Â?-Â?-Á<+À8(Ã5'ÜJ=ÚF:ÕA5Î</Ä6(Ä;+¾6&Æ:)Ç:)Ç:)É:*Ê8)Ë7)Ì6(Í3'Õ9-Õ7,Ö3*×3*Ú3+ß6/å:3é<8ð<=í49ë07ì18î3<í4<ë7@ê=CÝ7;áAAìNKøYUýZUúQLïB>å84â62à72à72à72á85â96ä;8ä;8á77ä88æ8:è8;ë8<ë8<ë8<ì8;ï58î47ì57ç77â:7Ù:4Ï8/Ç7,Â8+¿:+½;+»<+»<+»<)»<+»<)¸9&½@*³9$±6$¶;+µ0'Ã40ìPTÿ[iÿJ`õ:Uð7Vñ@]ïHfçHdÍNa‚89e?6R@2DA09=,57)>6+K9/E-#R=,g]B�ˆ_’«t—»{–Áz–¿{”³xœ¸…š´…€šm\uKHb5Nh8[xB_}?cƒ@hŒDl’Gj’Ih’Jg�Lf�Kd‰Dd‡Cc†DeˆFh‹KkŽNk�OjŽNi�Oi�Oi�Mh�Jg�Hh’Hl”Ko™Ms¡Vs£Wv¤Y24&04&01)01+12-23.34.46+9;-;>-?A3DF9IJDMONRQWSS]XVd\Zoda~nl”zw¬„‚ÃŒ‰Ö‘�㔑ê“�뎋≅قÌ|Ã|¿€~½„‚Á†…Á‰ˆÂ�ŠÃ‹ˆ¿‰…¸‡®…}¤…{�€vŽyn~rgmmaak^Vi[Pc\J]`K]aJc^Jh[JmZKqZJtZIv[FsWAxR;ŠQ=ªYFÍ[PãTPðFIñ>AÛ2/Ü5/á51ã60ã5.á5+à6)Û8)Ö7$Ó8$Ñ8&Ô9'Ø6)Ü6*à3,Ú6-È6)½8'¼7&½:(¿>+Â?-Ã;+Ä8)Æ4'éUIÓ<1Ì8,ÑB4¾5%À=+À=+À8(Ä8'Å9(Ç:)È9)Ë7)Ì6(Í3'Ò6*Ô3)Õ2)Û4,ã:3é>7ì?9ì=:è45ì59ï6;î5;ê2:æ39ä7=â<@óUVübbÿmjÿidóVQäA<ß63Ý4/ß61Þ71Þ71ß82á85â96å97æ:8å78æ89è8;é9<ë8<ë8<ë8<ì7<î5:î68ê67å97ß<7Ö<4Ì9/Å9,¾9*º;*¶;)µ<)µ<+´=)³;*³<(³:'·@,­8&­8'³;-«+"Ã54ý`gÿTdúG\ð<Uë=XçG_áOdÒO_´SZvGAXH9KD2FE1BE0=?*:8#;3?3PG*nkHŠ“f•¬t‘´tŽ·s‘ºxž»…¡¸Š�§}l‚[K_:AW0Lc7Wr?_{@dƒ@iŠCk‘Fi‘Hg‘Gd�HeŽJb‡Bc†Bd‡EgŠHj�Ml�Om‘Qn’Rj‘Pj‘Nj’Lh’Jg‘Gi“Gm—KpœOv¥Wv¦Zx¨\78(68*66,56.56056067/68+;>-=@-@C.DG4IK>MNFQPNRRRVUSYX]a_lki�us›€}²ˆ†ÇŒŠÓ•’ᔑâ��Ü‹‡Ó„�Ä�}º|µ|³�~·„�º‡„¹‰…·Š„²†¨�z›€x�s‡znzrfjm_^j[Ti[PjZKfZJb_Lb_Le^Li\Kn[Lr[Kw\K{ZG~UA†R=›RA¶VFÎRHÚG@â88å33Ý1-ß3/â5/ã5.á3,ß3)Ý5(Ü9(Ó8$Ð9$Ï:&Ñ:'Ô8)Ù7*Ý5,Ø7-Ä;+¹:)µ6%·8'½;+Â=.Ç;.Ç7,Ð90õ\TÖ=5É5+ÏA5¸3$·:(¼?+¿:)Ã:*Æ:+È:,Ê;-Ì:-Ï8-Ñ7-Ð3*Ò1)Ö1+Þ71å<7ê>:è;7æ74é75ï;<ð<?ê69ä15ä5:èBFíMOÿxwÿqnñ`[ÝJCÑ83Ñ2.Ü73ç@:à93à91à72à72â62ã73å76å76æ68ç79ê7;ë8<ë8>ë8>è7=ê7;î7;î79ê7:ä::ß=:Ö=7Ë;2Ã;/º8*µ:*³;*¯<)¯<*¬<(¬;)¬;)ª;(¬=*¦8'§9*ª8-¬/)Ë@CÿhrôM^óH[íDYåH[ÙO^ÉT]·TWœXUdM?OL9LG4JE1FD-?B'=D#>H#KV.\g<x�T‹™f‘¤m�©n�²r—¹}Ÿ¶ˆ’§€u†dRcC?P0BS1Oa9Vl=b{Be‚BiŠCj�Ei’Fe‘Fb�HcŽGb‡Bf†CgŠHiŒJkŽNm�Pn’Ro“Sj‘Nk’Oj“Mi“Ki“Ij–InšKržOv¥Wv§Xw¨Y?=.=>.==1==3==5;<4;=2;=/?B/@D-CG.FJ3KL:NOAQQIRRHYWJZZN_^\gerqn‰{y¡ƒ€·‡…Ä�ŒÑ�ŒÓ�ŠÏŠ…ǃº~y¯zw¦yv¥{v¬|y®€|®ƒ}©ƒ}£x—|t‹zr�vksrffm_\j\Sk[NlZLl[Kk[Ki]Mg^Mi]Mj]Mn]Ms^Mz]M�ZI“`O›VG§N@¸J=Å@7Ï6.Ø1+Þ1+á51á6/â5.â4-à2)Þ4'Ü6&Ú;(Ñ9$Í;$Ì;&Î;'Ñ:)Õ9*Ù7,Ó9-Á>,´;(°5#²5#»9)Ä<.É;/Ë7-Ù?7ø[TãIAÎ:0Ì@3¸6&¬3 µ<)¼9'¿7'À7'Ã7(Å7)É7*Î7,Ñ7-Ù<3Û:2Þ93å<7ê>:ê=9ç85ã41è96ë<9ì::è8:ì>@øPPÿbcÿppùheâTPÊ=6À1+É40Ö;7ß=:à;7â;5á:2â94á83ã73ä73å55æ66é69é69ê7;ë8<ë8>ê7=è7=é6:ï8<ì89è8:ã;:Ý>:Ó>7É<2À</¶8)²:)®;)ª;(©<(¨;'¥:&¥:(¥<) 9( ;) 9*¢6*µ>8ÛTXûepîK\ïI]éJ\ÛM[ÉQZ´VV ZRŠ`RPI7HM9NI5H@+;567>O%Lf6lŠTw”\„›e‰šd�›f�£k”­t˜°|ƒ•mn}\Q`C<J09F,AP1O_;Wj=czDe‚Bj‹Dk‘Fi’Fe‘Fa�FcŽGcˆCi‰Fj�Kl�Ml�Ol�Om‘Qn’Rm”Qm•Om–Pl–Lk—Jm™JrŸNs¢Q{¬]{¬]z«\DA0DB3DB6CC9CC;CC;BB6BC3DF0EH-GJ-JM2ON9RQ?TREUSD\ZC\\D\]Oaaaihxrq�{z¦�~µ†‚¿ˆƒÃˆ„Á„�º{®ys¡to—om•plžrn¡vpžwršwr’to…skzrinmb`l_Wj\Qk[Lm\Lo\Mo\Kq[Mq[No\Nm]Nl_Om`Os`Oz_N†ZMŸ_S¤PE®B8º;2Æ6-Î1(Ö2)Ý5,â70á6/á4-à3,à4*Ý5(Ø7%Ô9%Ï:"Ê="É=$Ê>'Í<)Ò;*Õ9,Ï;-¿@-±<(¬5!®3!º8(Ä<.Ê:/Î5-Ø93ðNIñTMÙB9ÌB5¾?.¦1³<(¾<,Á9+Á7*Â6)Å7+É9.Ï;1Ô;3áD=ãA<ä?;ç>;é=;é;:è88ç85ç85æ95æ74é;:ôJJÿ\[ÿdcøc_Ì=9Á82½4.Ã81ÕA=ÞE@ß=;Ù42â;5â;3â94â94ä84å84ç77é77ê69ë7:ë6;ë8<ë8>è7=è7=é6:ë7:é69æ89à;9Û>9Ñ>6Æ<1¾<.³8)®;)ª;(¦;'¤;(£:'£:' 9&¡>+–7%œ=+š9)š2'ÃPKíkmî\fìP^éJ\ßHWÍKS¹RS¤[R’cQ€jUJM8DL7JB/H9$B7FH#Up=g“V~³o‚³qƒ¥h}’Yƒ�YŽ–c‰—d|Œ_XfCGT89E-6B*:D,=J0JX7Wh>e|Fh‚Cl‹El�Fi“Gc‘Fa‘Ga�Gf‹FlŒIm�Nn‘Ol�OkŽNk�Om‘Qp—Tq™Sp™SpšPo›Nq�Nt£Rw¦U|­^z­^y¬]IC3JD6IE9IE:IG;IG;HF9HG5IH3JJ2LL2NN4RP;TQ>XRDYTA]Y>[Z>\YH^[Tdahnk~xs“}x �{§ƒ}«…­ƒ|¦}w�un�oi…kg‚ieˆjf‹mhˆmi‚mh|lfrkbgj`^g]Th\Nh[Kk\In]Kq_Kq_Kp]Lp]Np]Op]Ns]Ow]N~]N‡\L’XL¡UH¤F<«9/º7-Ê7/Ó6-Ú6-Þ6-ß6/ß4-ß4,ß4,ß7,Ý7)Ø6'Ñ6$Ï:$Ì<$Ê<&Ë='Î;)Ñ;*Ó:,Î<-¾A/¯<'¨3!¬3 ¶9'Á=.É;/Ì8.Ì2*àA;ø[TàI@ÐD7ÊF7«2¹>,ÊF7ÌB5Ê@3Ë?2Í?3ÔA7ÜE<ãF?â@;â=9ã:7â64ã54ä65å97å<9ä?9à=6Ü:5ß@:êMHêQLÓ@9¸)#¸-&¾5-É>7Ñ@;Ö>9Ú;8ã;;ç==â96â96â96ä86å95ç:6é99ê::ê88ê88ë8;ë8;é9<è8;ç7:ç7:é6:æ68â88Þ:8Ö=8Ì=5Â<1¹;-®9(ª;*¦;)¢;( ;'ž;&Ÿ:(›:'˜?-Œ5"™=.˜8*—/&Í\X÷y|ßX_æXdÚQ[ÊKR¹LO©VPž`U�gU€nVPW8?J*<;C=UQ+lvDƒ¢f”Á~‡¼v†¸s|žan‚Mu~Q‚‡^u}V[b@;D)4<%5=(=D2<E29C+CP2Ue>g{HiƒFl‹Gm‘Gj“Gf’Gb�GcŽGh�Jl�Mo’Pn“Pl�MiŽKj�Jj’Ls›Ts�Us�SržQržOs¢Rw¦Uz«Zv¦Zu¥[s£YRK;RK;RJ=RJ=RJ=RJ=RJ=RK;UN<VO=WP>XQ?YR@[TD\UE\UEb[HaZHaYLbYRe][kaiqftuj{zn„|pˆs‹s‰|p„uj{mbre]jc`k`_g_\c_Z^^ZY^YUaZRbZOe[Og[Ki\Kk\Ik\Gm]Fm]Fj^HicMqfTqZJuOB‹WJœ\P£YL¬ZL¥N=¦E4§<,±7(Â8-Ñ80Ü41Ý1-ß6/Þ7.Þ7.Þ7.Þ7.Û7+Ø6)×7)Õ8)Ô8)Ò9)Ò9)Ò9+Ò9+Ô8+Í;,½;+¶>-®9(©1 «2!·9*À</Ã9.È5+Î5-Ô:0ÛC8ÝK>ÔH9Â=,·2#ÝPFØJ@ÔF<ÔD;ÚG?ÞG@àC>à<:à74æ87ì::ì::ç77â64à93Ü=7ÙF>ÏB9Ä:/¼4(·1&¹5)¿;/Ã?3Ã:0É<3Ò?7Ú?:â=;ç;9ì89í9:ç7:å8:å8:å99å97å97å95å95à40á51ä65å76æ89ç9:ç9:ç9;ë;>æ9;ß99Ü=:ÖA=ÍB;ÁA6·?1©:)£<+ =*›<(™:&–:%—8&‘8&�A0‹<-‡/#‘/&¼LJämoçloÍVZ½LNµNOªQM¢UO›XO—]R•aTŠjSWZ/:M7FTa)~ŽP›°oŸ¼yœ½x’µs~�ah€L\mCYdBT]BHP9?D0:=,@C2<@1:>0<C3:C0AN2Rd>]sBgƒHp�Ko“Kl’Gj’Ii“Kg�Jl�Pm‘So“Sl“Pm’Mk“Jm–Jm˜IqœLqžMrŸNr¡Pt£Rv§UyªYzª^y¦cm™ZbŽOWN?WN?WN?WN?WN?WN?WN?WN?XO@YPAYPA[RC\SD]TE^UF_VGd[JcZIcYMcYOf[Uj_]nbdpdhreltgpvirvirsfmm`gg[_bXY^[VZZRYVOWTKXTIXTH\VH^WGcZKcZIf[Ig\Hi]Gj^Fk_Gh`IcbMngTv\M�SFšVM«VO°RJ·QE¯H9¬C0¬=,±9)¾8,Ë80Õ62Ù40Ý6.Ü8.Ü8.Ü8.Û7-Ú8-Ù7,×7+×7+Õ7+Õ7+Õ7+Ô8+Ô8+Ô8+Î:,¾<.¶>.®9(¦1 §2 ²:)¹>.¾<.Æ:-Ç4*Ë4)Ò;0ÜE:áOBßQCÝOCÛF?Ø@;Ó<5Ò;4Ö=7Ú=8Ü86Ý33ç77ë78î79ë78ç77å97â=9ÛB:ÊA7¿?4¸:,±6'¯5&²8)¸=-¿@1ÅA4Í@6×@7ß=8ã:7é77ï56ñ7:é6:ç6<ç7:æ89æ87å95å95ã:5â92ã:3ã:5ä;6ä;8ä;8å99å99ä:;ã;;ß<=Ù?=ÐA;Ä?6·=2¬;-¥;+ =*œ=+—<)–;(’:&“8%Œ:%€9'€8)‰5*�<5¿QP×eeÑ^a·MM¬NL¤TM�YP˜\Q—^S—^S˜_VŒfQ`_/Sg*^r3z�O™¯n¥¾|™²pˆ bj‚HauBTe;JX7FP7?H59?158-:</AB4=?4:=2=A3:C0@M3Rd>^tEgƒHq�Mp”Ln’Hk”Hk“Lh‘Km‘Qn’Tm”Sm”Qk“Lk”Hl—Hm˜Hp�Jp�Jp Ls¢Qu¦Tx©Xx©Xx¨^m—Xb‹QY‚H]TE]TE]TE]TE]TE]TE]TE]TE[RC\SD]TE^UF_VG`WHaXIaXGe]Je]Hd[Jd[Le[Qf[Uh]Yh][i]]j^^k__k__i^\f[YbWS_VQZVMWTKUQHRNCQMBSOCWQCXRB^WG^WEaYFc[Fg\Hi^Hk`JhbLbaMngUy_PˆXN¢[U±VQ²IE¸E@¹@7¸>1·;/¸8+¼8+Ã9,É;/Ñ:/Ù8.Û7-Û7-Û9.Ú8-Ù7,Ù7,×7+×7+Õ7+Õ7+Ô8+Ô8+Ô8+Ò9+Í;,À>0¶>.­:(£2 £3ª:&³>,¹>.Ä?0Ä8+Å2(È4*Ð<0ØF9ÜM?âNDÙ:6Ø43Ô20Ö42Û97á=<æ<<é;<í9<î5:ë27é26è58å99á<:Ù@:¾8-µ:+±6'®4%­5%°8(µ<+»=.ÈD5ÏC6ÖB8Ü=7â94è64í55ï79ê69é69é69ç77ç77æ95å95å:3ã:3ä;4ä;6ã<6ã;8â:9á99Þ88Ý9:Ü<<Ú@>ÓB=È?9¹:1­7+¥7(¡;,�<+™<+”;)“:(�9&�7%Š9&z6#y6%ˆ9, F=¹SOÀWT¸PO«IF¢MHœSL™YP—^S—aW—aW˜_V‹ePsrFz�V“¦n¤¹€«À‡ž³{|�[[o<?R$AS+AP/?L2:C06<.69058/9;.@A3=?49<1=A39B/@M3Qc=`vGi…Js’Or–Np”Jm–Jm•Nk”Np”Tq•Uo–Sn–Pl”Kl•Il—Hn™Ip�Lp�LpŸNu¤Sx©Xyª[u¦Wq X_‰JVEOx>aXIaXIaXIaXIaXIaXIaXIaXI_VG`WH`WHaXIbYJcZKd[Ld[Je]Hf^Gf^Ig_Lg^Of\Pf\Rf\Sh]Wh]Wh]Wh^Ug]Te[QdZPc[N]YMZVJVRFRNBPL@PM>TN@UO?WQAYTA[VC]XDaZGd]Jf_LfaMdcQleUv\OŠ[Q©b\¸ZX¶FD¹<:Ã<8Æ;6Ä:0Â8-½8)¼:*¾?.Ç?/Ó9-Ù7,Ú8-Ù9-Ú8-Ø8,Ø8,Õ7+Õ7+Ô8+Ô8+Ô8+Ô8+Ô8+Ò9+Í;,Á?1¶>.¬9'¢3 �3¢8"¬=)µ@.¼?-¿:+Ã7*Å5*Ë7+Î:.Î</Ó:2Þ65à24Û12Ý34à88å;<ê=?ì<>î7<ì38é06æ25å58á77Ú65Î61·5(¯7'®6&¬7&¬7&®9(²:)µ:*ÃA1ÊB4ÔA7Ü?8â;5æ95ë76î87í68ë78ë78ê86ê86è94ç:4æ;4â92á:2ß:4à;7Þ:8Ü:8Û97Ù99Ö<:Õ@<ÒC=ÊA;»;2¯6+¤6'ž7(œ;*˜;)”;)’;(�:&�8$Œ7#‡9%€=*v5#�5'™E:ªPH¬MG©LG¦QL�PJ˜UL”ZO”^R•aV•aV—aW�gT‰…_™©x­½Ž¦¸ˆ�¡ts†YN`8/A->3B%:F.=F3;A3:=2:=4<=599-?@2<>39<1<@29B/?L2Pb<awHj†Kt“Ps—Or–Lo˜Lo—Pm—Op—Ts˜Up˜Ro—Pm–Jm–Hn™Jp›KqžMrŸNt£Sy¨X{¬]w§[m�Se’M[…FS|BNw=cZKcZKcZKcZKcZKcZKcZKcZKcZKcZKd[Ld[Le\Mf]Nf]Nf]Lf^Gg_Hh`IiaLiaNi`Oh_Pg^Oj`Ti`Qi`Qh_Nh_Nh`Mh`MhaNd^N`ZL[UGVPBPM>NK:NK:NK:PM<RO>TQ@VS@XWC[ZF]\H^]Kb`Qf^QmWJ†[R¬ic¾c`ÁMMÈBCÐ;=Ô89Ó84Ì70Á9+º=)µC+¼B+Î;+×7)Ø8*×:+Ù9+Ö9*Ö9*Ô8)Ô8)Ô8)Ô8)Ò9)Ò9)Ò9)Ò9)Í;,Á?1¶>0«:(¡6"š4›7 £=&¯B-³;*º;,Â:,Ç9-Ë8.Î:0Ð<0Ù:4é9<ì4<æ39ã28â38á48â38â25é6<é49ç4:ç7:ã9<Û76Ð21À0(´6(­:(­:(«:(«:(«:(¬9&¯7&¸:+Á;/Î>3Ø?7à=8ä;8è96ì87î66î66î66í74ê84è:3ç:3ä<3á:2ß;2ß<5Ü=7Û=:Ù><×=;Ô>=Ï@<ËB<ÅB:¹=5­7-£5(ž7(š:*–;)“<)‘;*�:(Š9&‰8%ˆ7$ƒ8%ƒ@-u4"{3%“G:¤QI¡NFžNG¡WN˜UL’XL�ZN�\N�]R”`U—aW�hW�‰f˜¥z�©�€Žj]kHDS42@&$2-:&0;*5=.9?3<?4=?4:;3::099->?1;=28;0;?18A.>K1Pb<bxIk‡Lu”Qt˜Ps—MqšNq™Ro™QršSršSršQp™Mm—Km˜Ip›LsžOt Qv¢Sx§Y{©^z¨_qŸVd’JY†C]†JW~GRyBe\Me\Me\Me\Me\Me\Me\Me\Me\Mf]Nf]Nf]Ng^Og^Oh_Ph_Ng_Jh`IiaLjbMjbOjbOjaPjaPjaPiaNh`Mh`Kh`Kh`IiaJjbKf_Mb]J^XHXRBSM=MJ9KH7IH6KJ8LK9LM;NO=PQ?QTASVESVE[YL^VIcQEzXN¡ha¶eaÄVUÖQRÙ@Cß:>Ü87Ô70Ç;.¼?+±C*µA(Ê;*Ô7(Õ8)Õ9*Ö9*Õ9*Õ9*Ô8)Ô8)Ò9)Ò9)Ò9)Ò9)Ò9)Ò9)Í;,Á?1µ=/«<)¢9$™6•5›;"§A*²B.¹@/¿;,Ã7*Å2(Ë4+Ò;0ß=8ê5:î3<ë6=é8>ç:>ä;>ã:=â9<ß58á7:â9>á=>Û=>Ñ96Ã2/¶0'¯9+«>*«>*©>*ª=)©<(©:'«8%¯6%¹7)Å;0Ð=3Ù<5ß<7ä;8è;7î87ï75ï75î85ë:4é;4ç<4ä=4â>5à?7ÞA:ÚA;ÙA>ÕA?Ñ@=ÍB?Ã?:»@8³=3ª8-¡5(›5'˜8(•<*‘<(�<)Š<(ˆ:&†9'„7%ƒ6$6%}:)t3!~9*—OA£YNœRG•OE˜VJ•YN�[MŽZMŒ[MŽ\Q“_T™`W‘gW~xXyƒ^r|ZU`B;E,/;%0;*0<.3=24<16<2<?6@A9?A6<<077+89+>?1:<17:/:>07@-=J0Oa;bxIk‡Lu”Qu™Qt˜NqšNršSpšRršQsœPr›Op™Kn™Jp›LsžOv¡Ry¥X{§Zz¨]x¦]qŸWg”O]ŠEX‚B`‡NY€ISzCg^Og^Og^Og^Og^Og^Og^Og^Og^Og^Og^Oh_Ph_Ph_Ph_Pi`OjaPjbOjbOjbOi`Oi`OiaNiaNiaNh`Kh`Kh`Ih`IhaGhaGhaGf_Le^Kb[I\WDWR?PM:LI8IH6IH6GH6GJ9GK:HL;IM<HO?JN?PQCXRF[OCiRDƒYM–VL°SLÔZYÛIJâ?Bá99Ú72Ï;1Â?-´B*³<$É:)Ó7(Ô8)Ó:*Õ9*Ó:*Ó:*Ó:*Ò9)Ñ:)Ñ:)Ñ:)Ñ:)Ñ:)Ñ:)Í;,Á=0µ=/¬=,£<)˜8 �4•9 ¡A)­C-¶A/¿=/À6+Â/%Ç0'Ð6.Þ:8è59î6>ì;Aê@CæAEàBCÜ@AÙ??Í12Ð66Ó:<Ñ=;É;9¾71µ4.­4)ª<-¥>+¥>+¥@,¥>+¤>(¦;'©:'ª7%²7'¼8+Å8.Ì70Ô94Ü=9æ>;é:7ì95ì95ë:4é;4è;4å=4â>4àB7ÝC9ÚC<ÕB;ÏA=Ê?<Ç=;¿>9±<3©:/¢8+œ6(˜7'”8)‘:)�:(‹=)‰<*‡<)ƒ:'‚9(€7&6%}6$x3#x5%…B2˜UE�WK”PC�PD•ZL“YM�ZN�\Q�\Q’[T•\U˜[VŽbUkbEV`=JS4=G,6?*2=,4>35?66=68?8=B;CF=EG<CD6@>/;9*78*=>0:<17:/:>06?,=J0N`:awHk‡Lu”Qu™Qt˜Nr›OršSpšPq›Or�NqœMp›LošKr�Nu¡Ty¥X«`~©az¥]qœUf‘L_‰G]‡E^ˆIb‰PY€ISzCh_Ph_Ph_Ph_Ph_Ph_Ph_Ph_Pg^Oh_Ph_Ph_Ph_Ph_Pi`Qi`QmdUlcTjaRi`Qh_Pg^Og^Mh`MiaNiaLiaLiaLh`Ih`IhaGh`IgaKf_Le^Kb[H\WDUR?PM<ML:EF4DG4DH7DH7BI9BK:BK:CJ:IK>SQDSPAXL<eK:tE5•H>ÂYSÙROàDEà::Ü75Õ;3Ê>/º=)µ8"É:)Ï8'Ð9(Ò;*Ò;*Ò;*Ò;*Ò;*Ñ:)Ñ:)Ñ:)Ñ:)Ñ:)Ñ:)Ñ:)Í;,Á;0¶:.­<,§>+™9#�3“7Ÿ?'£9#°;)À>0Æ<1Ì8.Ð7/Ù<5ã?=ì=Bî?DéCEâDCØ@?Ê;7À50»1.¹/,¼51¾:6»<6³:2ª8.¤8,¡:+¡=-Ÿ>+ ?,¡@-¡@-¡A+£>*¦=*©;*­:(²7(¹5)¿5+É90Ô?9àC>ä=7è;5è;5ç<4å<5ä=4á>5Þ@5ÜD9ØD:ÒC;ÊA9Ä?:¼;6¸85°93¢7-™8(•5'“6'‘:)�<*‹:)…8&ˆ=*…<+ƒ<*€<)~:'|8%{6&{6&y1#€8*ŒG8–SC’PB‹L=�RE™_S‘WL’YN”[R–]T–]V—\V–YTŒ^QaX;EM(6@8A&>G2<E43=2-7.1817>6BE<GJ?HI;BC1?>*;:&78(=>09;069.:>06?,<I/N`:awHj†Ku”Qt˜Pt˜Nr›OršSpšPqœMqœLqœLp›Kp›LsžOw£V{§\‚­e€«dx£\j•P]‡EYƒA^ˆHe�PcŠQZ�JSzCi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi_Sj`Wj`Wj`Wj`Wj`Vj`Vj`TjaRjaRjaPjaPjbOjbOjbOjbMjbMg_Jg_Jf^Ie_Ib]IZXCSP=ML:HI7EH7AE6@D5?E7?G8?G8>F7GK=IK=AE4KJ6\Q;aE/ƒJ9Àj]Üg`ãUSäFEåA@àA=Õ>5É=0Æ=-Æ4%Ë5&Ì6'Í7(Í7(Î8)Ñ;,Ó=.Ò<-Ò<-Ò<-Ò<-Ñ;,Ñ;,Ñ;,Î<-»2(¶7.±;/¨:)š7"”5•7œ9"®A-³;+º6)Á4*Í6/×<7âC?êHEèDEèEHÝCCÉ:6»61´80®8.©4+«9/©9.¤8. 8-›9,–:+“<+’<+–=+˜=*™>+š?,›@-�A, A-£@-¢;*¥:(«7(°6)¶6+¾8-Ä;3Ì<4Ú=6à=6â?8âA7âA9ÞB6Ú@6×@5ØH=ÒE;ÇA8»;2²5/ª3-¦1*Ÿ2+™9-’;*‘9+�:*‹:)ˆ;)‡;+…<+€9'}8(|9(|9(|9({8'y6&x3$w- ‹A4—OC“MAŽL@‘QE“UJŽRG˜\R˜\R‘UM�TL—ZU•XS”WT“gZ]T5IQ*AK(BK,>H/7@+2=-4?14<1HPCZ^PW[JMO:DF.AA'==%:;)>?1;=2:=2;?14=*;H.Pb<cyJlˆMu”Qs—Or–Lp™MršSq›Qp�LqžKrŸNrŸNsŸPv¢U{§\ªb‚­fx£^j“O`‰G^‡Ea‰JfŽPh�Td‰S[€LTyEi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi_Sj_Yj_[j_[j_Yj_Yj`Wj`Vj`Tj`TjaRjaRjaRjaRjaRjaPjaPh`Mh`Kg_Jg_Jd]J^YEUR?ON:IJ:EH7BD6@D5?E7>F7>F7=E6@F8>E5>F1HL5MI0U?'Q:¼sbËcXà_ZëUTçGGâ@>àA=Ø?9Î:0Í9+Î=,Ï=.Ï>-Ï;-Î;+Í9+Í:*Î:,Ð:+Ð:,Ð:+Ð:,Ð:+Ð:,Í;.É?5ÅB:¼B7®=/Ÿ6#–1–1›2¨5#´6(Ã;/ÑA8ÞE?æIDéJFêKHÞCAÛEDÐB>¼<3¯9/©=0¥?1Ÿ=0Ÿ?1œ>2—>0”>/�=/Š>.‡@.ˆ?,Ž=*‘;*’<+“=,”?+•>+—>,™>,Ÿ@.¡=-¤:,¨8,¬8+±8-¶:0¼9/Ê<2Ð<2Ó?5ÕA7×C9ÖC9ÔD9ÑE8ÊB6ÃA4¹=3°:0©6/¤5.Ÿ4,™7,’:.Œ=,Š;,ˆ;+†:*„;*ƒ<*�<,~;*|;){9)y:)x9(w8'u6'w4$€4'�D7˜NC‘K?ŒJ>�QD’TI�RG“SJ˜XO˜UO—VPœ]X”WRŽSOŠcT]W5Xc9_jBZfBIT66C)4@*<H4OXGYbQ_fTW\FJN7BE*??#<<"9:(=>0:<19<1<@25>+;H.Pb<dzKlˆMu”Qt˜Pr–LqšNršSq›Qs OrŸLp�LqžMt Qx¤W|§_ªc{¦aošUcŒJ]†Da‰Je�Ne�Nc‹Od‰S[€LTyEi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi_Uj_[j^^j^^j_]j_Yj`Wj`Vj`Tj`Tj`Tj`Tj`Tj`Tj`Tj`Vj`TlaOk`LiaNiaNf_La\HYVCSR>MK<HI9DF8AE6AE6?F6>D6<D5=F57@-@H1IK3F?%W?'‡ZC·s`¶VJÙbZðebìTSæFHëGHæCDØ;6Ë7+É:*É:,É:*Ë9,Ê8)È6)È6'Ê8+Ë7)Ë7+Ë7)Ë7+Ì8*Ì8,Ë8.Â3-¾5/¶6-­3(¡0"ž/ 1 §3$¼A2ÇC7ÖG?áLFçLGéJFåFBßD@×CAÏDAÁ@:±;1¤</¡A1œE4˜E3“D3‘D4�B2ŠA0…@0‚A/~A.@.†:*‹9+‹:)Š;*‹<+�<+Ž=*�<*–?.–=-›;+�9* 8+¥9-©;.®9/¸:.¾90Á;0Ä>3Å?3ÅA4ÃA3ÁA4¶;,²:,ª8-£7+ž6+›6,˜7.”:/�;/ˆ<,‡;-„;,‚:+€;+€;+<,};+z;*z;,w9*v8)u7(t6)v4&ˆ;1”F<–NBŽJ=ˆH<�OB‘SH’RI’OI™TO™SQšVSš]Z“YU�[T�l[�}X~‰^€‹alxRLX46D#=J.LZ@eqYeoW]eMPU>EI0?C(=@#:<$89';<.8:/9<1=A36?,<I/Pb<dzKm‰Nv•Rt˜Ps—MqšNršSpšPu¢Qp Ln�LpŸOu¤Vy§\}¨`|§`p™Uf�M\„E\„EeŒMi�Qf�N_…HcˆRZLSxEi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`QjaRjaRjaRjaRjaRjaRjaRj`Vk`\k__k`^k`\k`ZkaXkaWkaUkaUkaUkaUkaUkaWkaWkaXkaWnbRmbPjbOjbOhaNe^K_ZGYVERO>MK<GH:DF8CE7BD6@D5<C3<H49B-FJ3LF0O:%qL9šcOª^N¡A5ÆSLçc_ð^^íSUëJOéDHáAAÌ91Ç9-Å7+Å7+È8-Ê:/Ë;0Í:0Ñ>4Ñ>4Ò>4Ò>4Ó?5Ó?5Ó?5Ó>7Ó?;ÑB>ËD>ÆF=ÂF<ÃI>ÇM@ÎPDÙSHÝPGâMGäIEäGBâC?àA=ÛB=ÕFBÉD?º?7«=0Ÿ>.šC0•G3”I6�F4ŒE3ˆC3…B1€A0}@.zA.|?-‚:,†8,…9+†:,†:,‡;-‡;+ˆ;+‰:+‹:)Ž8)�8*•9,™;/�=/¢<0ª:/­8.¯9/°:.°:.°:,¯9+®8*¨7)¥7(ž6)›6*•7+’8-‘9/Ž<0ˆ<.…<-ƒ;,�9*:*~9*}:*};+|:,z;,y;,w9*s8*q6(r7)u5)ŒD8’H=‘K?ŠH:‡I<‹OD“TK—TL–QL›TP˜QO”SO–\X�_X�h^—~j””p…’fu‚WYh?CR+?O*P`<crQetU\iKNX=CL1?E+>B)=A(;='78(:;-79.9<1=A37@-<I/Oa;bxIlˆMv•Ru™Qt˜Nr›OršSo›Pq OožMožMs¢Rz¨]|ªaz¥^u [cŒJ^‡E[ƒD_‡HgŽOj‘Pf�L`‡Hb‡QY~KRwDi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`QjaRjaRjaRjaRjaRjaRjaRj`Tk`Zk`\k`ZkaXkaWkaUkbSkbSkbSkbSkbSkaUkaWkaWkaXkaWocUnbRkbQkbQkbQhaOd]M`ZJXRBRO>NK<IG8GE6DE5BC3>B1=I1?G/LG1S>+g=-•ZL­_Sœ>4ž3+¶?9ØXUóigödeëRUæJNèNPÙIAÓF<Í@6Ë>4Î@6ÒD:ÕG=×G>ÚJAÚJAÝJBÝJBÞKCÞKCàKDàKEëSRèTTãSRÜROØSLØSLÛULàUNàLHáIFãEBâC@äB@âC@äEBàFDÙJFÊE@¹>7¬=2¡?2˜B1’E3�G4ŒE3ˆC3‡B3ƒ@0�?1~?0z?/|>/€:.ƒ9.ƒ9.ƒ9.ƒ;-ƒ;-ƒ;-ƒ;-�9*‚9*…9+‡:*‰:+�;-�<.“;/›;/ž9-Ÿ9- :,¡9,¢8+¢8*¢8*Ÿ8)œ8)˜8(”8)�9*Œ:,‰:-‡;-„:-‚:,�9+€8*}8)}7+|8+{9+z8,y9-x:-v8+q5*o5)q7+t9+ŒH;�I<‰I=„I;…K?ŒRF“WM˜XO˜QMœUQ˜RP“VQ’`YˆbWƒg[†zdnrQZi@JY0AR(IZ0Wi?dvNj|VXiEL[:?L.:D)<D,>C-<A-:=,89+:;-68-8;0>B47@-;H.L^8_uFi…Jt“Pt˜Pt˜Nr›Os›TpœQn�MmžLp¡Pw¨Y|ªaz¨`r�Xi“QZƒA\„EaˆIeŒMgŽMgŽKh�JfŠLc†PY{IRtBi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`QkbSkbSkbSkbSkbSkbSkbSkaUlbYlbYlbYlbVlbVlcRlcRldQldQldQlcRlcTlbVlbXlbXlbXocUnbRkbQlcRlcRkbQg`Pd]M^WGYSCSM?MJ;KH9GE6EB3?B/;H.?H-OC-a=-ƒE:«XR´NJž,+²:9¹=;ÓSRòpnþvvñehëX^ð^_ãUQÚPFÔG@ÐC:ÑD=ÔG>ÖGAÖG?ÙHCÚJBÜHDÝJCÞJFÞKDàKGàKGâHHáGIÜHHÙGGÖHD×HDØGBÛFBàBAäABæBAèBBçCBäB@ßA>ÛA?ÛJGÊC?¹<6®>3¢@3˜A0�A0�C3ŠA2ˆ@1‡?1…=1‚<0€<1=1=1�;1�;1ƒ:3�;1�;1€</€<1=/€</€=-�<-€;,‚:,�9*‚8+ƒ7)Œ:.�8.‘9-”:/—;.š=.ž</Ÿ>.™9)–9(“:*�:+Š;,‡;+„;,‚:+‚:,�9+7)~6({5){5){5)z6)x6*x8,w9,s8*p4)o5)r:-v<0‹M@‰K<„J<‚M=…QC‹VH’XM—WN“NI™TO—VR•\U�e\{^PfVF`^GEO->N)?O(J\2_rEk~QexKXj@M_7@Q-6E&7D*=F1=D4:>07;-8:,:;-68-8;0>B47@-9F,I[5[qBe�Fr‘Ns—Os—Mr›Os›TpœQmœLo Os¤Ux¨\z¨`s [i“Q`ŠHY�B^†GeŒMgŽMf�Je�GgŒGhŒLa„NXzHQsAi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`QkbSkbSkbSkbSkbSkbSkbSkbSlbXlbVlbVlcRlcRldOldOldMldMldMldOldQlcRlcTlbVlbVocUmaSkbSlcTmdUlcTi`Qf_Ob[K^WGXQARL<OI9JG6IC3AC.:G+@F*U@+xG9£PL´HH¶8<º39ÈAEÄ?@ÊJIßb`ñqpðnpçbeâZ\ÛQNÕLDÏD?ÍB;ÏD?ÐE>ÐB>Í@9ÒC?ÒC=ÔC@ÕD?×CAØD@ÙECÚDCÝADÝADÞDFÝGHßIHàJIãIIåGHçACêADìBEìBEçAAâ@>Û=:Õ=:ÚFFÉ=<¹83°;2¦@4˜?1‘>0�@1Š</Š<0‰;1ˆ:0‡81†93…:4„;4�;3€<3�:4€<3<3}=1}=3|<0|<0|=.|=.{<-|:,{9)z8*{7*‚:.ƒ9.‡9-‰9.�9.‘;.•<.–=-’9)‘:)�:*Š;*‡;+ƒ<*€;+~;*€8*€8,~6*}5)|4(z4(z4*y5*u2)v6,v8-r6+o3)o5*s:/x@3ŠPB†N?�N=‚Q@…VF‰XIŽUJ‘RI’OI—TN“TO‘\VŠfZn[JPK7EK1AN0KZ9WgC`rJgyOdzLVl>G]/AV-7H$1@!6B*=F5<D98=67:39;.;<.68-9<1?C57@-7D*FX2Vl=a}BoŽKq•Ms—Mr›OtœUq�RožPq¢Qu¦Wv¦Zs¡Yj—RbŒJ[…E^†H`ˆIcŠKeŒKf�JfŽHf‹Fe‰I`ƒMWyGPr@i`Qi`Qi`Qi`Qi`Qi`Qi`Qi`QkbSkbSkbSkbSkbSkbSkbSkbSlbVlcTlcTlcRldQldOldOldMldMldOldOldQlcRlcTlcTlcTnbTmaSmaSmaSmdUlcTjaRg`Pd]M`YIZSCUN>PK8MH5ID1DC.@F*DD([B,ˆTG¯\X«>A¬/3ÎINÆ?EÀ;>»;<ÄFGØZ[ãefÛY[ÌGHÑJGÐEBÍB?ÐB>ÔFBÔFBÐB>É>7ÏD=ÏD=ÒD@ÔEAÖEBÙECÚDEÝDFÞ?Cß@DÞBEßEEáGGàFFàBCá>?èBDé@Cè>Aç=>ä>>ãA?áC@ßECÖBBÅ98¶50°;4¦?6š>3�=/�?2ˆ</‡:0‡:0‡81‡83‡83ˆ94ˆ;5‚92€:2€:2;2~<0~<0|<0|<0x8,y9-z:.z:.{<-{<-z;,z;,};/9/€7.�7,„6*…5*†7*†7(Š8*ˆ9*ˆ;+…<-„<-€=,~<,};+}9,}7-|6,y5*z4*x4)x4)x5,r2(u5+v8-r6+n4)n5*t;0xB6ŠUG„PB€O@„SDˆWIŒVJŒRG�NG•TN”UN‹RK„XMdSeZDGK2=J,IX9ZkIgxTcxQYnESh=Nc8J_4:L&0?,95>)=C5<B8:<79<5;=2;=079.9<1?C58?-8B)EU1Uh;azCnŠMr“Nr–NršQs�Sq�Rr Uu£Xv¤[t¢ZošUf�N^ˆH[ƒDdŒNa‰K`‡HcŠIf�JgŽKeŠEa…E`ƒMWyGPr@i`Qi`Qi`QjaRjaRkbSkbSkbSjaRjaRjaRjaRjaRjaRjaRjaRkbSkbSkbSkbSkbQkbQkbQkbQlcRlcRlcRlcRlcTlcTlcTnbTk]Pp`QtdWrdWnbVj`Th`ShbTgaSgaQc]M[VCVQ=SN8NH2GA)KG,I?$bI3‘gW¶xm»le¸YW¹OO¶BE½BEÂDGÈFHÍHKÐJKÒLMÓKMÏEEÐDEÓEDÖFEÕFBÒC=ÎC<ÏF<ÈD8ÉE9ÊE<ÌE?ÒD@×CCÚADÛ?CáBFâBDáCDàDEàDEáFDãEDäDDãCCäB@ä@?ã?>ã?>â@=â@=ÞC?ÓB?Å>:¶93ª70Ÿ:0—=2Ž>3†>2ƒ?2�?3�=2�;3„93†93ˆ81ˆ92…;0„<.„<.ƒ;-�;/€:.€:.€:.~:-~:-~:-~:-};/};/};/};/}90}90~80~8.€7.€7.€7.~8,€8,~8,}9,|8-{9-z8,x8,x8,x8.x8.w7-v8-v6,t6+s5*q5+m1'r6,n4)m3(o6+m4)q8-I=…SH…SJ†RGˆRH‹PH�QJ”QK•RL“TM�TK‰[N‚cQo_HYU:MV7O_;j~YbxQYoHUjCRg@Mb;IY5CR17D&4>#2:#6;'<>0@B5@@6<>358-39-4:.7:/9=/;?.=B,<E(S_;arFn†Rs‘Ur•QršSs�Sr�UužZ~§e~§ep˜YbŠK^†G^†H^„G`†IbˆKdŠMf�NeŒMcŠK`‡H^„G_‚LTvDKm;i`Qi`Qi`QjaRjaRkbSkbSkbSjaRjaRjaRjaRjaRjaRjaRjaRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTnbTrbSqaRrbUqcVrfZogZmeXjdVgcWkhYoiYlgThcOc^H[U=RL4PJ0H>%Q?)kP=XG„NBˆE=“D?±WV¸TT»QQÀPOÀNMÁMMÂLJÃKJÐTTÈHGÂ=>Ä=:ÐB@ÓEAÏB;Ç>4ÊD9ÉE9ËF=ÍG>ÔFBØDDÜAEÝAEáBFáCDáCDâDEâDCáFDáFDáFDàEAßD@àB?ßA>àA=àA=àA=ÞC?ÔE?ÈC<º>6¬:0 90”:/Š<0‚<0€A2~@3~>4<3ƒ:3†91ˆ81‰90‡;.…<-„;,„;,ƒ;-‚:,‚:,‚:,9-9-9-9-~:/~:/~:/~:/}90}90}90}90}90}90|90|90|90z:0z:0y9/y9/y9/w9.w9.w8/w8/t8.s7-r6,r6,p6+p6+k1&q7,m4)l3(o6+l3(p7,~H>„RI…SLˆQJ‹PJ�QL”PM•QN”SOŠOG…SH€[K{ePsiPhkLfqQgxT^tMXnGPf?La:K_:K\:GV7CP4<F-8@)6;'7:)<=/?@2>>2;=04:03;05;16<0:</<?.>A,>D*Q[9^mDk�RpŒRp“Sq™Rs�SsžVxŸ^|¢exžak‘T`†I]ƒF]ƒF\‚E`†Ia‡Jc‰Le‹NdŠMbˆK`†I^„I]LRtBIk9i`Qi`Qi`QjaRjaRkbSkbSkbSjaRjaRjaRjaRjaRjaRjaRjaRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTnbTqdTo_Pm]Pm_RqeYog\ldYhbVfbWpl`{xi~{jyizubpiVe`JTT<FF.C>(MB.S@/U8(]7*l=3n70u91~;5‡@:’IB›PJ¥XP­ZR»^YºSN·GE¼B?ÈDBÑFCÓD@Ï@:ÐA;ÐC<ÒE>ÔG@ÖGCÚFFÛEFÞDFßCDßCDßCDßCDàDEßEEßECßECÞDBÞDBÜD?ÛC>ÜC=ÜC=ÜC=ÛC>ÖE@ÎE?ÃB<³>5¢:1”8-ˆ9,ƒ;-‚@2@1~>2=1ƒ:1†91‡81ˆ:0‡;.„;,„;,„;,‚:,‚:,€:.€:.9-9-~:/~:/~:/~:/~:/~:/}90}90}90}90|90{8/{8/{8/{;1{;1z:0y;0x:/x:/x:/x:/t8.t8.s7-s7-r6,q5+o5*o5*j1&o6+m4)k2'l6*i3'm7+}G=‡PI‡PI‹PJŽQL‘RM’SN”UP‘VPŒWO‡[P~^OtbNoiQorUlwYfwUQeBK_:EY6DU3EV6GV9GT8ER8?I0<D-9>*7;*:<.<>0;=/9;.6<25=26<26<0:</<>0=@-=C)MW5Zi@f|MlˆNn‘Qq™RuŸUtŸWz¡`w�`n”Wc‰L\‚E\‚E\‚EZ€C_…Ha‡JbˆKdŠMc‰La‡J_…H]ƒH[}JPr@Gi7i`Qi`Qi`QjaRjaRkbSkbSkbSjaRjaRjaRjaRjaRjaRjaRjaRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTnbTrdWoaTm_Rk_SkaWlcZle[jf]jf]xtiˆ„x‘��•��’Œ~‰‚r}zg`cNOT>DG2FE1HC0F=,L<,VC4P9+R6*V4*Y5)\6+a9/e=1m=1ŠKB�SJ®UO¶NK¼FDÆABÒDCØFFÔ@>ÕA?ÖB>ÖE@ÙECÚFDÛEDÝEDÞDDßCDßCDÞDDÞDDÝEDÝEBÞFCÛFBÚEAÙD>ØE>×D<×D<ÙD=ØE>ÔD<ÓF?ÉF>ºA8§;1˜8,Œ9+…<-ƒ@0@1~?0=/ƒ:1„:/‡9/ˆ:.„:-„;,„;,ƒ:+‚:,‚:,9-9-9-9-~:/~:/~:/~:/~:/~:/}90|90}90{8/{8/{8/{8/x8.{;1y;0y;0y;0x:/x:/x:/v:/s7-s7-s7-r6,q5+o5*o5*m4)j1&o6+l3(h2&k5)h2&l6*|F<ŠOGŒOJ�PK�RM�SN‘VP�WPŒZQ‰]R…aUy_PiZGd_IdhO\fKN]@EU8AQ4=M0<K.?M3CQ7FR:GS;BL4>G2:A/9=,9=.9=.8<-7;-7=36=56;46<2:<1;=/<?,<A*JS4Ve>bxIj†Ln‘Qs›Tw¡Wx¢Z{¢ar˜[e‹N\‚EZ€C\‚E\‚EZ€C_…H`†Ia‡JbˆKa‡J`†I^„G]ƒHY{HNp>Eg5haQhaQhaQibRibRjcSjcSjcSibRibRibRibRibRibRibRibRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTlcTrfXth\sg[mcYjaXle]snhzwp�~w��† �”¨¥œ«§œ¨¤™£�‘™–‡z{kcfSOR?GJ7EH5BE2BE2EH5KL<JH9GD5D>0A9,A7+@6*F4(S5*qB8ŒPHŸPK¯IG¿GIÌEIÓBEÖ@BÙ@BÚBAÛCBÚEAÚEAÜDAÜDAÞDDÞDFÞDFÞDFÞDDÝEDÝEDÚDCÚFB×FAÖF>ÔE=ÔE=ÓF=ÔE=ÕF>ÔA:ÓC;ÎE=ÁB9®>3ž9-‘;.Š</„?0�?/€>.€=-�;/„:-†:-‡;.„:-ƒ;-‚:,‚:,‚:,�9+9-9-~:/~:/~:/~:/}:1}:1}:1}:1|91z:1|91y90y90x8/x8/v7.y:1y:1x90v:0v:0u9/u9/t:/r7/q6.q6.p5-o4,n3+n3+m4+j1(m7-j4*h2(j4*f2'j6+}D;�NG�NH�QJ�TL�UOŽWPŠYRƒ[Q{YMv\OkZJ]UBYXDY^HOYA?K3:H/7E.4B+5A+8D.=I3CL7EN9BK6?H5<E2:A/8?/7>.5<,4:,5<45<56;56;49;0:</;>-;@)GP1Ra:`vGi…Ko’Ru�Vy£Yy£[xŸ`n‘W`ƒIZ}C[~D^�G^�G]€F`ƒIa„Jb…Kb…Ka„J`ƒI_‚H^�IW|ILp@Bf6haQhaQhaQibRibRjcSjcSjcSibRibRibRibRibRibRibRibRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTlcTkaUndZpf\lcZkd\rmg„�|‘�Œ§£ µ±®ÄÁ¼ÉÆÁÉľÅÀºÀ¹±¹²¨¤œ�ˆ‚rfcTPQ?FI8>E3;D1:F28E38E38C2:B3<B4@C8CE:GE9>4(P:/e?6€EA¡PO¿X[ÊQVÊCIÕFJ×EFÚFFÛEDÝEBÞDBÝD?ÞCAÞDDÞDFÞDFÞDFÜDCÜDCÜDCÚDCÙEA×FAÕF@ÓF=ÒE<ÐF<ÑG=ÔE=Ô?8Õ@9ÐC:ÆC9¶A7¥=0–=/Œ=.†>/‚?/�>.€=-�<-ƒ;-„:-…;.‚:,‚:,‚:,‚:,�9+:+9-~8,~:/~:/~:/~:/}:1}:1}:1}:1z:1z:1y90y90x8/v7.v7.t8.w8/u9/u9/s9.t8.s9.r8-r8-q6.q6.p5-o4,n3+m4+l3*j4*h2(l6,i3)f2'h4)e1&i5*{B9ŽMG’MH‘PJ�TL�VO‡XN�YO|\Qs\Nk[L`WFVSBPSBJQ?@I64@,3<+2;*09(09(2;*5>-9B1<E4?H7?H5>G4:F2:C25@/3>-1<,3:33954954928919;.;>-;@*CL/O^7]sDi„Mp“SwŸYy£Yw¡Yq˜Yf‰OZ}CX{A\E]€F]€F^�G`ƒI`ƒI`ƒI`ƒI`ƒI_‚H^�G]€JV{HJn@Ae7haQhaQhaQibRibRjcSjcSjcSibRibRibRibRibRibRibRibRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTlbVj`VjaXkdZkg^upjˆ…€Ÿžš°°®ÃÂÀÓÒÐâáßçæäçãàãÞÚÞÖÓÙÎÈμ²¸¥——‡zym]_YIKL:@E1;D/>J6=H7=F5;C4;>5:;3983880==3C<2N71a84ˆHH¯]_Å`dÊWZÍQSÓNO×MKÛKJÞHGáFDãDAâBBßCDÞDFÞDFÝCEÝCEÜBDÜBBÛCBÚDCØD@ÕD?ÓF?ÒE<ÐF<ÐG=ÔE=Ù@:Ù@:ÑB:ÈC:¼C:¬@4™;/Œ9+‰=/ƒ>.�>-€=,�<,ƒ;,ƒ;,„<-‚:,‚:,‚:,�9+9-9-~8.}9.~:/~:/}:1}:1}:1}:1}:1}:1z:1y:1x90x90w8/t8.s7-r8-t8.s9.r8-q8-r8-p7,p7,p7,o6-o6-n5,n5,m4+l3*k2)i3)g1'k7,h4)e1&h4)b0%f4)zA8�MG’MH‘PJŽULˆXN�YOxZOr\Nk^N^WGRPAKN=CJ:9B12:+08)17+17-06,/5+/5)08+2:+4<-9D4;F5=H7<I7;F56C13@,1>-.800621622717829;0:<.:?)?H+KY5[qChƒLq“Vwž[w¡YsœVg�P^�IUx@WzB[~F[~F[~F]€H_‚J_‚J_‚J_‚J^�I^�I]€H]€JUvIJk@@a6haQhaQhaQibRibRjcSjcSjcSibRibRibRibRibRibRibRibRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTlbVpg^ng_lgatqj‡„¡ œ¼¼ºÎÐÏÏÏÏàààòòòúøùû÷ö÷óðñìéðâßèÐÆÝÁµÁ«�¢‘�}ub^\GKM7CH2AI2AI4@G5BD7AA7B=9C97?:69<5>?7B71J.+d66�QR³giÆnmÀ\\ÇZWÐVS×QNÜLKâHHæCDåBCßCFÝDFÝDFÝCEÝCEÜBDÜBBÜBBÜDCÙCB×C?ÓD>ÒE>ÐF<ÐF<ÔE=ÜC=ÛA9ÒB:ÊD;¿E:°A6œ:-‹5&Š</…=.‚=-€=,�<,�<,ƒ<*„<-‚:,‚:,‚:,:+9-~8,}9.}9.~:/~:/}:1}:1}:1}:1}:1{;1y:1w;1x90u9/t8.s9.r8-q7,r8-q8-p7,n8,p7,m7+o6+o6+o6-o6-n5,m4+l3*j4*i3)h4)f2'k7,h4)c1&e3(b0%e3(xB8ŽMG’OIŽRJŠUMƒXOzZOq[Mh\LZTDIJ:?C4>D69A208+/7*6<247058157247025.06,08-19,6A3:E5=J9>K9<I77F32A,1>-,6..400511606718:/:<.:?+<E*IW4YoAg‚Mp’Vv�\užXp™U`†IX{CSv>WzB[~FY|DY|D\G_‚J_‚J^�I^�I^�I]€H]€H]LStIHhA?]7icSicSicSicSicSicSicSicSicSicSicSicSicSicSicSjcSkbSkbSkbSlcTlcTmdUmdUmdUmdUmdUmdUmdUmdUmdUmdUldWibZgd]vsn“’Ž¯¯­ÄÆÅÚÜÛêîïóôöö÷ùüüþÿÿÿÿÿýÿûúü÷ôýòì÷áÔòØÇàʵƶŸª ‡ˆ„ibbFEH-DH/CG0FE1GD5F<3C60F42H:9>=9>?:B:7I75cGF’jjÅ“’è­©Ò‹‡ÇtnÀ]XÆSPÕONßJLçFKéFKßEGßIKÚDFÔ;>Ø>@áGIàDGÖ:=Ú@@Ú@@ÙA@ÖB>ÔC>ÑB<ÐA;ÑA9ÞC>ÛA9Ó@9ÊA9¾B8°@5Ÿ:.�8*Œ=0‡>/ƒ>.�>-€=,<+�=*�<,:+9-9-~8,~8.|8-|8-|8-|90|90|90|90z:1z:1z:1z:1y=5w<4w;3u:2t91q8/p7.p7.o6-m7-m7-l8-m7-l8-m7-m7-i2+m6/p92o81k4-g2*g2*h3+e0(i7.d2)]-#`0&_/%b2(s>6ŒOJ�PKˆQJ�RJ}YMv_QfYITN>DE5>E5:C25@04</3;04:04:068378366446135016//6..6+.9+2=-6C2:G6:I68G44C.1?.*4+,2.-2..3-45/79.:</:?+6>&DR1WlCh‚Rs”]wž_r›Wk”R[~DXyDTu@Tu@WxCZ{F[|GZ{F^JbƒNbƒN^J\}H_€K`�L^~MTsJFd@<W4icSicSicSicSicSicSicSicSicSicSicSicSicSicSicSicSkbSkbSkbSlcTlcTmdUmdUmdUmdUmdUmdUmdUmdUmdUmdUldYgd]onj†…ƒ¤¤¢ÁÃÂÖÚÛêîï÷ûü÷ûþøüÿüýÿþþþÿÿýÿþúÿüùÿúóÿðàûëÔíßÅ×ͲÀ»�¤£„‚ƒcgjKLQ3GJ/EC.GB/H>2I:3L95K<9?:7BA?MHE]RPyjg£Ž‹Î²®èÅ¿ÿ×Ñ쳪͂|Àc^ÊVVØRSßHMÚADÖDEÕEEØHHÜJKÛEFÕ<>Ö<>ÛACÙ??Ù??ÙA@×A@ÖB>ÓB=ÒA<Ó@9ÜA<Û@;Ó@9ËB:¿C;±A6 ;1’8-Œ=0‡>/ƒ>.�>-€=,<+�<,�<,:+9-9-~8,}9.|8-|8-|8-|90|90|90|90z:1{;2{;2z;2w<4u<3u:2s:1r90n8.o6-m7-m7-l8-l8-l8-l8-j8-l8-l8-k6.n70o81n91l7/i4,g2*e3*c1(d4*_/%^.$b2(^.$`0&p>5�JE‡RL‡XR~XOrVKdRDQJ:>?/:A16A05@03>04<14<15;17:379477577557246116//6./7,.9+1<.4A08E49H58G44E24B1.8/.5..3--2+23+46+8:-9=,7?(DQ3WlEh‚Ut•`w�bp˜Yg�PY|DWxCTu@Tu@VwBYzEYzEYzE[|G_€K_€K\}H[|G^J^J^{MSnKF_A9R4icSicSicSicSicSicSicSicSicSicSicSicSicSicSicSicSkbSkbSkbSlcTlcTmdUmdUmdUmdUmdUmdUmdUmdUneVneVlfZed_y{x—™˜²¶·ÍÑÒäéìôùüúÿÿûÿÿüÿÿýþÿþþþÿþüÿÿûÿÿúÿÿôþüçúùÝïðÑáåÄÒØ´¼ÄŸ ¨ƒ‡‘l`gEQX7DG,DC.JD4LD7KA8I>:C;9SJKia_�wvœ’�¾³¯ÙÎÈêÜÓþèÝÿóéÿäÛð©£ÃecµBEÇDIÚRVÛUTÑMKÊDAÍCAÔFE×EFØDDÙCDÖ=?×>@Ø@?Ö@?ÖB@ÓB?ÓB=ÔA:Ù@:Ù@:ÒB:ËE<ÀD<²B7¡<2“9.Ž<0‰=0ƒ>/�>.=-~<,€=-€=-9-9-~:/}9.}9.|8-{8/{8/}:1}:1{;2{;2{;2{;2z;4x<4t;2q;1r90o9/n8.l8-l6,k7,l8-j8-j8-j8-j8-i9-j8-j8-n91m80m80m80m80i7.f4+c1(d2)a1']-#_0&c4*\-#_0&sD<}PJ�VP~YQpTI^J?OC7?<-46(2;*0;+1<.2=/4<14<15;17:379479668357257227007/08-.9+0;-2?.6C27E48G48G49F54>33:2/4-,2(/0(13(57*6:)6>'DQ5YmJk„Zw—fwœfl“Z_‡KWxCWuCTr@Tr@VtBXvDYwEYwEZxF]{I_}K]{I]{I_}K^|JZwKPgJAW@3I2icSicSicSicSicSicSicSicSicSicSicSicSicSicSicSicSjcSkbSkbSlcTlcTmdUmdUmdUmdUmdUmdUneVneVneVneVmg[jkfƒ‡ˆ£§¨¹¾ÁÐÕØåíïóûýõýÿûÿÿûÿÿüÿÿüþûýþùþýøÿþ÷ÿÿóøýæôýÞî÷ØæòÎÝéÃÎÚ´¶Á™¡«†€‡efmLMQ6FH0HJ5KI:HD9D@7OGDj`_‹�€¤š™·¯¬ËÆÂÞÛÔêæÝúñèÿûñÿ÷íÿÝÕû¶±å�ŽÊcd°@?ÃPKÉTMÉPHÃF@ÈGBÔMJÖKHÐ@?Ò>>Ó??Õ?>ÕA?ÕA?ÓB?ÓB=ÓB=Ö>9Õ@:ÒC=ÌE?ÀE>²A9¢=5•;2Ž<1‰=0ƒ>/�>.=/~<,€=-€=-9-9-~:/}9.}9.|8-{8/{8/~;2~;2|<3{;2{;2{;2z;4x<4r90o9/o9/m9.l8-k7,j6+h6+i7,i7,i7,h8,h8,h8,h8,h8,m;2l7/j5-i7.j8/j8/f4+a1'e5+a2(].$a2(b3)Y* a2(yNE�\VvWRiNGWC:I<3@:.:;-8<.08)/:*3;.4</5=25;169069079468368349238139/19.19..9+/:,0=,2?.5B17E4;H6<I89C87?428.-3'.0%/1$24&48'4<'ER8]pPr‹d{šnwœif�XU|ESsATr@Sq?Sq?Tr@UsAVtBVtBWuC[yG]{I\zH^|J_}K\zHWrIJ^E<M;.?-gdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSicSjcSkbSkbSlcTlcTmdUmdUmdUmdUmdUmdUneVneVofWofWng]qqo‡Œ�¤©¬·¿ÂÍÕØãíïòüþôþÿøÿÿúÿÿûÿüýÿúüýõúüñùùíöøêöÿèóÿäðýáîùÙéôÒÞèÅÊÒ­¶¾™�£�}‚bY]BGL5FK7FJ9CG9AC8c^X…}z«¡ŸÁ·µÌÂÁ×ÏÍæáÝñîéú÷ðüõëÿñæÿòéÿôìÿåß騤·hc¦LAµPDÁYNÇ[QÆQH¿D=ÂA<ËDAË@=Î@>Ð@?ÑA@ÒC?ÑC?ÑC?ÒC=Ó?;ÒA<ÐE@ÉE@½C>¯@9 >5•=3Ž<1‰<2ƒ=1�=0=1~<.=/=/~:/~:/~:/}9.|90{8/{8/{8/|<3|<3|<3{;2z;4y:3y:3w;3n70m80l7/l7/k6.h6-h6-h6-i7.h8.h8.g8.g8.g8.g8.h8.k92h6/f4-g5.i70h70e4-b1*c2+b3+_0(]1(]1(V*!b6-{ULtYRaNHN?8A7.=6,;9-9;.9<139-3;.5;/6<06<07:/58-47.69049238139/39/2:/19.19,.9+.9+/:,0;-3>.6C2:E5<G7=E8:B55;/06(02%/1$13%15$2:%GS;buWw�k|›rq•g\‚QJp=Oo>Qn>Qn>Qn>Qn>Qn>Sp@Sp@WtD[xH]zJ]zJ_|L_|LZwGRlECU?6D7(6)gdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSicSjcSkbSkbSlcTlcTmdUmdUmdUmdUmdUneVneVofWpgXpgXng]lnmz‚…“š «µ·ÄÎÐÜæèí÷ùôþÿõþýöÿüúÿúûÿùûþóøúíõöèñôãôýêôýèôüåóùßòöÛéìÍØ×¹ÅÄ¥±°’�ŽregOMP;GK:FM=GOBKQG{{sŸš”ž¸ØÎÌßÓÓçÛÛóëéü÷ôðïêüüôÿÿöÿýóÿùïÿóêÿçÞÿ×ËØ“ƒ¸eS£J:´TF½WIµF;·@8ÈKEÇC>ÊC?ÌB?ÎC@ÎC>ÎC>ÍD>ÍB=ÑB<ÒC?ÎDAÅD?¸A=ª=8ž;5•<4Ž;3‰<2ƒ=3�=2=1~<0=1=/~:/~:/~:/}9.|90{8/{8/{8/|<3{;2{;2z:1y:3x92w81u91n70l7/l7/j8/i7.i7.h6-g7-g7-g7-f7-f7-f7-f7-f7-f7-h70h6/g5.g5.g6/h70g6/f5.a2*e6.^2)\0'`4+a5,i=4uSJXG@E>6<5/95,;8/8:/57,36+69.5;/7:/69.69.58-57,36+28,19,28.19,19.19,19,19,.9+.9+.9+/:,0;+3>.6A17B2<D5:C28?/5;-46)24'13%04%07%HT>ex\u�mt’lf‰_RwKBh9Kk<Nk;Pm=Pm=Nk;Nk;Ol<Qn>VsCZwG]zJ^{K_|L_|LVsCNg@<J9/:2$.&gdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSicSjcSkbSkbSlcTlcTmdUmdUmdUmdUmdUneVneVofWpgXqhYoh^hikmtz…Œ”£¬±¿ÈÍÓÞâåðòðüüòüûõþùøÿ÷úÿöüÿòúýìøùçö÷çööê÷÷ëú÷èüöæýõâ÷ìÖçØÃÕƯÀ±šž“}un[ZWFPQCSVK[bZeld•˜�²²¨ÕÎÆèÝÙóããúêëÿñôÿ÷ùÿþûþýùùúôøúïüüðÿÿóÿÿñÿûêÿæÑñª”¼o[©VD®RCµPD»LA¿JAÂG?ÆE?ÉE@ÊE@ÊE>ÉD=ÈC<ÉB<ÑC?ÐE@ÌEAÂC=³=9¦;5œ;5•<6Œ<3‡=4…<5�=4<3~<0=1=1~:/~:/~:/}9.|90{8/{8/{8/{;2{;2y:3x92w81v70v70s7/n70j8/j8/j8/j8/h8.h8.h8.g7-f7-f7-f7-d8-d8-d8-f7-g6/h70i81i81h70i81j;3l=5g80k<4b6-a5,oC:xLCtI@nMDC9056.45/8918;247.14+25,58-58-57,46+46+46+46)37)/7(.9)08+.9).9+.9)08)08)19*08)08+/7*08+19,3;.4<-9B1:C0;B0:A1:<.68*24&/3$29'IU?cu[m„gf„bWyTInE?d8Hg;Li;Nk=Nk=Li;Li;Mj<Ol>UrDYvH[xJ]zL^{M\yKQn@G_;6B4+4/#)%gdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSicSjcSjcSjbUlbVlbVmcWmdUmdUmdUleUmfVngWogZnh\oh^mjekormtzƒŠ�¥®³ÃÌÑÕßáæîñôüþôýú÷ýùùþ÷úÿôüÿòüÿñüÿîýþðûøïþúñÿùïÿúíÿùéÿòßïàËÝηÁ²›¢—�|ubb_NXZL]cWmwn|†}¦¬¢¾ÀµÚÕÏîäâüîîÿôôÿô÷þôõÿùùûúøûýøüÿúøÿõôüñ÷ýñÿÿñÿùèÿôãÿÌ»½p›QD©SFµSH²C:ÀJ@ÄH@ÇG>ÉF>ÈE=ÇD<ÆC;ÆC;ÍG>ÌG>ÇG>¼C;­>5Ÿ:2—:3‘>6Š=5†=4ƒ=5�=4€<3;0�;1�;1~:/~:/~:1}90|90{8/{8/{8/z:1z:1x92w81v70u6/u6/q6.m80k90j8/j8/i9/i9/i9/h8.g7-f7-g7-f7-f7-f7-f7-f7-f5.j81m;4k:3j92j:0m>4oC8l@5oD;d<2c=2zVJ†dZyWMbLA>:13814927<54:0/5+25,9<336+25*24)03(13(25*46+47,.6),7)/7*-8*/7*/7*/7*/7*19,08+08-/7,/7,/7,08-08+7?09B1<E4=D4;A38<.26'-4$2;*JVBcr[h|a]vXOmKFfAAa:Hd;Kh<Nk?Nk?Kh<Jg9Li=Nk=TqEWtHZwK[xL]zNZvMNjAC[;2>4*00!''heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVjdVjdVjdXldYlcZmeZmeXmfVg`NjeRnkXol]he\feasqr�‚†Š�“–ž¡«³¶ÃÈËØÝàêîïö÷ùùûúüþýþþüÿÿúýþöúýòùüñúýòûþóùúòúúòùùíüúëþüçù÷ÞéåÊÕÔ¶º¹›˜˜|qtY_dMbkXr}l„�‚‘›�´¹²ËÌÇãâÞòîë÷óòû÷öýùøüø÷ÿþüþþüþþüýÿúýÿúýÿúýÿúýÿúúü÷ÿÿúÿûöÿêåا £`X›D=µNE¿MCÂF<ÄD9ÊE<ÎI@ÎKAËH@ÇG<ÆG8ÃH8½J8³H8¦E5šB4‘B5‹C5†B5„B6‚@4‚>3‚<2„:1†91†9191~:1|91{80z7/z7/w7.v6-w7.w7.w8/w8/v7.u6-q5+n3+l7/i81i81h70h70g6/g6/g6/f5.g6/i70i81h70g6/e4-e3,h3-j5/m80k90j:0i;.i=0kA3gB2jI:dH:v_O~k\ŠxjŒoRJ=79.4:04:039/39/39/28.28.17-17-36-06,25,/5+14+/5+/4-.5-/4-.5-/4-/4-/4-/4-05./4-/4-.3,.3,/4-/4-/6.1;23=26@58B79D67B45@03>-4?.DQ=WeN[kQPbHEX<@T8BW6Ic>Ke>Kf=Kf=Je<Je:Lg>Oj?UpGYtK\vO]wP]wRWqNHb?;O6&1+&+.$),heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVjcYkdZlc\md[meZmfVidQjhSmjWjk[gg_lll~„�“š¨¯µ¸ÂÄÑÖÙâçêòóõûüþÿþÿÿþÿþýûÿþúÿÿúþÿùûüôøûòöüò÷ýó÷úñøúïøúí÷úéùûåòõÚßâÃÌЯ´¸—‘–vkqUYaI_kUt�oŒ˜Š�§œ¿Á¼ÓÓÑççåóóñ÷÷õûûùüüúûûùþþüþþüþþüþþüþþüþþüþþüûÿþõÿÿõÿÿþÿÿÿûúÿíêð¿ºÂ~uŸKA¯M@¹K>ÂL@ÅK>ÃE9¼=4¾?6ÃE9ÄF7ÃJ7½L:³J7¦F6™C2‘C6�E7…C7‚B8�A7�>5‚<4ƒ:3†91„93~92}:2z:1z:1y90x8/v7.v7.u6-v7.t8.t8.t8.r6,o5*m4+l7/i81j81h70i70h70g6/g6/e3,e3,g5.h6/j81j81j81k92k60l71k90k;1k;/j>1j@0iD2gH6kP=gRAufS�ueŒ…s†ƒrKL<69.39/39/39/28.28.28.17-17-17-17-06,06,/5+/5+/5+/4./4./4./4./4./4./4./4.05//4..3-.3-.3-.3-/4./6/.80/:21=34@66B66B66B45B16C1CP>P]IR`IIW@AO6BP7EU:Jb@Jd?Ke@Ke>Ic<Hc:Id;Je<SnEXrK\vQ]wT[tTTmOC\>6I5&1-',0$)-heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVjcYkdZlc\md[meZmfVmhUkiTmjWjk[kkcwww‘’—§¬²¿ÆÌÒÜÞìñô÷üÿþÿÿþÿÿÿþÿÿþÿþýûÿþúÿÿúýþøúûóõøïñ÷ëñ÷ëô÷ìõùëöøêõøåõ÷áîîÔÛÛ¿ÉÊ«°±’“–yuy`kpZt{iˆ��Ÿ§œ®µ­ÍÏÊÞÞÜïïíøøöûûùþþüþþüýýûþþüþþüþþüþþüþþüþþüþþüýÿþûÿÿûÿÿÿÿÿÿûùÿúôÿ÷íýÑÆØž�”L=¥RB©L;§@1·G;ÏYMÏUJ¼@4ÄD7ÆH:ÃK=¹I;ªB7�?3—@7“E;‰@9„A9�A8?6~>5}=4<4}=4z;4x<4w;3w;3u:2t91t91t91r90r90r90r90o9/n8.l6,k5+l7/j81l71j81k60i70i70h6/g5.g5.f5.g6/h70i81k:3k:3l;4j;3i:0h<1k?2jC4iD2fE2fI7lVAjYGujV†m‘Ž{€mEF658-28.28.28.17-17-17-06,17-17-17-06,06,/5+/5+/5+/4./4./4./4./4./4./4./4./4./4..3--2,-2,.3-/4..5.+5-+6.-9//;12>24@46B47D38E3?L:ER>CQ:;I28F->L3DT9J_@Jb@LdBKc?Ia=G`9F_8G`9RkDXpL]uS^vVZqTPgK=T8.A.&1-',0%*.heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVjcYkdZlc\md[meZmfVniVljUlkWlm]pqi€‚�ž¡¦¹¾ÄÄËÑÛåçõúýûÿÿüýÿüýÿþüýüûùÿþüÿÿûÿþùûüôõöîîñæéïãçíßíñâðôåñôãñôßòñÜêêÐØؼÇÇ«²±•¡¡‰‘“}�“€œž‘ª­¢¶»´ÀÅ¿ÜÜÚééçööôûûùýýûÿÿýÿÿýýýûþþüþþüþþüþþüþþüþþüþþüþþþûûýÿþÿÿþÿ÷òïüóìÿüóÿûíÿôäìñ¹�p•P@¢O?¸XJ·M?µE9ÀG<ÁA6ÈE;ÅI?½G=¯@9¡<4š=6–A<�?;‡@<�@:}@;z?9y@9x?8x?8v=4v=4v=4u<3s=3r<2r<2q;1o;0o;0o;0n:/k9.j8-j8-i7.m82n72m61l71l71k60i70i70l:3j92h70f7/f7/f7/h91f:1j?6f>4e=1f>2hC3iE5fG3cG2cJ6hV@i\IskV…‚o�Žzpq_9=,47,28.17-17-17-06,06,06,17-17-17-06,06,/5+/5+/5+.3-.3-.3-.3-.3-.3-.3-.3-.3-.3--2,-2,-2,-2,.3--4-*4,)4,*6,+7-.:.1=14@25A36C2:G5<I78E12?+2@)8F/>N4DX<E]=K`AJb@K`?G_;H^:F^:SiEXpN`uV^uXZnSLcI9M4):(%.+%*.$)-heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVjcYkdZlc\md[meZmfVlgTkiTnmYop`tum…‡†¤§¬¿ÄÊÎÕÛæðòûÿÿûÿÿûüþýþÿÿþÿþýûÿÿýÿÿûþýø÷øðïðèçêßàæÚÞåÕçëÚêïÛíðÛìðÙîîÖèèÐÚØ¿Ìʱ½§·µ ±®�¶´§Á¾µÉÈÃÒÓÎÙÙ×èèæòòðúúøüüúýýûÿÿýÿÿýüüúþþüþþüþþüþþüþþüþþüþþüÿþüÿüÿÿüÿÿûúüû÷ýÿùþÿøþÿôÿÿíÿþìøßËÅœŠ�bP–J:¥L<´P@¼L>¿@7ÆC;ÅF?¿D=±>;¦:7ž:8™>;‘=;Š?<„@=}@=xA<tB;rC;rC;q?6s>6s>6r=5r=5r=5p>5o=4n>4m=3l<2k;1j;1i:0i:0j:0n72p62p62m61m61l71i70i70l;4k:3h91e90e90e90e:1d<2fB6cA5cC4dD5dG5dI6bI3_I2^L6eV?jbMrmW„ƒo„‡r[`L3:(28,28.17-17-17-06,06,06,17-17-17-06,06,/5+/5+/5+.3-.3-.3-.3-.3-.3-.3-.3--2,-2,,1+,1+,1+,1+-2,,3,,6.+5-*4+*4++5,.8-0;-2=/2=-5@/6A05A-1=)1=)5A-8F/=O5@T8FY=H\@J]?I^=K^>K`?TgG[pQcvZauZYkSJ]G6G4&4%$-*$),"'*heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVjcYkdZlc\md[meZlfVjgTkjUlnYorasvm�…„Ÿ¤¨¸¿ÇÖßäí÷ùûÿÿûÿÿûüÿþÿÿÿþÿþýûÿÿûÿÿúüüôóõêëíâãçÙÝãÕÜãÑäéÓçìÕéíÖêìÔëëÓèæÏßÚÄÓθÓκÌƶËøÐÉÁ×ÒÎàÛØêæåðïíóòðùù÷ýýûýýûýýûÿÿýþþüüüúþþüþþüþþüþþüþþüþþüþþüÿýüÿûüÿøùÿþýþÿýùÿýôÿøïÿõ÷ÿöð÷çÿÿíÿîÛͧ”¤kX¢[GªWE­L;¹D:ÀD<ÀE>¼E?´A>ª=:¢:9š;9•=<�?=…?=}@;wB<rC;oD;oD;p?8q?6q?6q?6q?6p>5o?5o?5o@6n?5m>4k<2h<1h<1h<1l<2o83q62p62p62m61l71j81i81h70g80e90e:1d<2e?4f@5dB6^B4aG8cL:dM;cM8`K6]K3]M4]Q9bX?mhRss[€ƒnv{eFO:4=*39-28.28.28.17-17-17-06,17-17-17-06,06,/5+/5+/5+-2,-2,-2,-2,-2,-2,-2,-2,-2,,1++0*+0*+0*+0*,1+,3,.5.,6.*4+*4+*4++5,-7,.9+-8*0;+4?/5@/4?.3?+3>-3@,7F/9K3@O8BT:GW<H[?M]BM`DUeJ\oSdtZ`rZXgRGXE4B1$2%$-*$),"'*heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVjdVjdVjcYkdZlc\md[lfZlfVliVlkVkmXkn]lrhy}•� ®·¾ÒÛàêóøûÿÿùþÿùúþþÿÿÿþÿúù÷ÿÿûþýøùùññóèéëÞãçØßæÖàèÓäéÒçíÓèìÓçéÑèèÐçåÎàÛÇ×оÚÓÃÙÑÆÞÔËæÝØîäãóéêúñôÿùûúù÷þþüÿÿýýýûýýûÿÿýÿÿýüüúþþüþþüþþüþþüþþüþþüþþüÿýüÿùøÿúùÿþýüÿý÷ÿþîÿùñÿÿòÿýõÿúóùëÿýíÿúèôλ«wb�Q:ª\H²N>¸J=¹I>¸I@´G@¯D>¥=:œ:7–=9�>:ˆ?9€A:xA:tD:pF:pE<p?8q?8q?8q?8q?8q?8p?8p?8qB:pA9n?7l=5i=4j>5j>5n=6o83r73p62p62n72l71j81i81d8/d90c;1d>3d@4cA5cA5_C5ZE4_N<gVBhWCcS<^O8[O5\R7]V<]Y>nlUtv^|�kcmU2>(6B.3;.39/39/39/28.28.28.17-17-17-17-06,06,/5+/5+/5+-2,-2,-2,-2,-2,-2,-2,-2,,1+,1++0**/)*/)+0*,1++2+-4-+5-*4,*4,+5,+5,,6+-7,,6+/:,2=/6A17B27B15@04?.2?-4C.:G3=L5AO8DT:JX?L\BTbI\kTcqZ^mXUbPDRA2>0#/%&/,&,,$**heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVjdVjdVjcYkdZlc\md[lfZlfVolYmlWjlWgkZflbr{xŽ˜š¨±¸ÊÓØãìñøýÿ÷üÿúûÿþÿÿÿþÿúùõÿÿúýýõøøîðòåéëÝåéØãêØåíÖæìÒéíÒêíÒèèÎèæÏèãÏáÚÈØÐÃØÎÄÝÒÌêßÛøíëÿôöÿõùÿõûÿùüþüýÿÿýÿÿýþþüþþüÿÿýÿÿýýýûþþüþþüþþüþþüþþüþþüþþüÿþúÿýûÿýûýüúùýüøÿÿõÿÿóÿÿíýúóÿûûÿúÿÿôÿöæÿóßÿãÍ×­•©oY¨WD®Q@¬O>¯OA±OB­K@¤C<™<5•>7�?6ˆ?8�A7{B7uD6qE8qE8p?8q?8s>8q?8q?8q?8q?8p?8sB;qB:o@8m>6l=5j>5m>6o>7o83q73o83m82m82i81h91f:1f;2d>3d@4bB5`C5]A3\@2WB1TG4[S>f^Gi^HbX?ZS9ZS7\W:ZV;XW;kmUsw^u}fUaI&28D03;.4:04:039/39/39/28.28.17-17-17-06,06,/5+/5+/5+-2,-2,-2,-2,-2,-2,-2,-2,,1++0*+0**/)*/)+0*+0*,1++2+)3+*4,+5-+5,,6--7.-7,,6+.8-2<16A39D69D47B46A10;+2?-5B09F2<I5@N7FT=JX?R`I[hTanZ\iWQ^MBN@/;/",#+1-(.,&,,heVheVheVheVheVheVheVheVifWifWifWifWifWifWifWifWkeWkeWkdZle[md]ne\mg[mgWqn[kmWknYkr`jrgpyv‡‘“Ÿª°»ÄÉÐÙÞêïóöûþüýÿþÿÿÿþÿþýùÿÿúþþöúúðöøëôöèðôãêñßçïØðöÚíòÔëîÑììÒèæÑáÜÉÝÕÈÞÔÊáÔÎèÚÙòääúëîþòôÿ÷ûÿùÿÿüÿÿýþþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþÿúÿþúÿþúþþüýÿþûÿÿúÿÿúÿÿûÿÿþÿÿÿÿýÿûõÿûñÿüìÿûåÿóÛÿêѺ‚i¦bK—O9ŸT?¨ZF£RA�J: M?Ž<0‹=0†>2=/x=/t?/sB3tD6r=5r=7t=8s>8u@:u@:t?9q?8sA:q@9q@9q@9p?8n?7o>7o>7k92m82l:3l;4l;4h<3g<3c=2c?3^>1dG9cG9T=-N9(M8'@1WQ;PN7KI0OK2]Y>ieJeaDVU7XX<]`CosZz€fdlU=I1-9#2>*4</5;14:04:04:039/39/39/39/39/39/28.17-17-17-06,16016005/05//4./4./4./4.,1+,1+,1+,1+,1+,1++0*+0*.5.-4--4-,3,-4--4,.5-.5-+2*-4,07/5=29A6=E8?G<?J<7B25@/3>-1=)2>*9E/BN8IU?O[GVbN]hWYdTLWI<F;/9.)3*)/+(.*'-+heVheVheVheVheVheVheVheVifWifWifWifWifWifWifWifWkeWkeWkdZle[md]ne\mg[khWonZjlVknYkr`jtiq|x‰”–£®´¾ÇÌÒÛàëðô÷üÿýþÿþÿÿÿþüþýùÿÿúþþôûûï÷ùëô÷æðõáêòÝèíÖêïÑæëËäçÊææÌãáÌßÙÉÞÔÊßÔÎçÙØíßßöçêûïóÿôøÿ÷ûÿúþÿüÿÿþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüýÿüûÿúûÿúýÿüýþÿþýÿÿüÿÿüÿÿûÿÿüÿÿüÿÿûùÿüõÿþñÿÿíÿûåÿòÙÿëÑà±—§u\‘[C–ZBšYCšVC™R@ŽG5‹F6ˆE4ƒD3|C2wB0s@/o>/u@8s>8u>9u@:t?9o=6p>7sA:q?8p?8p?8n?7o>7m>6n=6m>6k<4l=5m>6k?6j?6gA6eA5bB5dG9[A2^G7^I8N=+F7$G8%>6!LL4GK2DG,DD*MM1\[?baC`aBYY=dgJsw^sy_X`I:D,/;%7C/6>16<26<26<25;15;15;15;15;15;15;14:04:039/39/39/27127127116016016005/05/.3-.3--2,-2,,1+,1++0*+0**1**1*)0))0))0)*1*+2*,3+.5--4,.5-07/4;39A6?F>BJ?;F8:E57B14?.3?+6B.<H2@L6LXDS_KYdTWbRLVK=G<1;2+5,+2+*0,)/+heVheVheVheVheVheVheVheVifWifWifWifWifWifWifWifWkeWkeWlcZmd[md]ne\mg[khWmlXimVjoYktaiuiq}yŠ˜™¥²¸¿ÊÐÔÝâíòö÷üÿüýÿýÿþÿþüþýøÿÿ÷ýýóúûí÷úéô÷äïôÞçïØåëÑâçÉÞáÂÝÝÁÞÞÆÞÛÈÝ×ËàÕÏäÙ×îââóçëúîòÿôøÿ÷ûÿùüÿúýÿýýÿþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüýÿüøÿüøÿüûÿþþþþÿüÿÿúÿÿùÿÿøÿÿùÿÿúÿÿúúÿýøÿÿöÿÿóýÿïÿúæÿôÜÿøàÿìÓ㿧¬‚j‡YB…Q;’[F—]I�S?F2u>)r>)s@+tD0vE4q?4n<3q<4tB9sA8o?5qA7xH>n>4m>4m>4k?4l=3j>3l=3j>3g?5gA6gA6gC7eC7cC6`C5^D5bK;UB1ZI7`S@RG3C;&E=(FA+AE,BH.AE*<@%?B%NN2_`AijKaaEnqTvzagmSJO94<%4=(=F38>28>48>48>47=37=37=36<28>48>47=37=37=36<26<26<25:449349349338238238227105/05//4..3--2,,1++0*+0*).().().().().(*/)+0*,1+160/4.,1+,1+/4.6;4=B<AH@@K=>K:<I78E34A-2?+3A*5C,DQ=KXDQ^MQ^MHTH<H<2>4.8/-4--4-,3,gdUgdUheVheVheVheVifWifWifWifWifWifWifWifWifWifWkeWkeWlcZmd[md]mf\mg[jiWkmXimVjoYjs`hthm|wˆ˜˜¦³¹¿ÊÐÓÜãëðööûÿüýÿýÿþÿÿýÿþùþþöýýñúûí÷úçô÷âîóÜåíÕâèÎÛàÀ×Ú»ÖÖ¼Ù×ÂÝ×ÇÞ×ÍåÛÙìààøìðûðöÿõûÿùþÿúýÿûûÿûûÿýüÿþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüýÿüøÿþøÿþûÿþþþþÿüþÿúÿÿùÿÿøþÿúýÿùüÿùùÿýúÿÿøÿÿøûÿõüþðÿÿíÿüéÿñßÿðÝÿïÚñÒ½´�z{R>tJ4uH3vI4zM8}P;yL7nA.e7'{M@sD:qB8sD:oC8g;0d8-g=1i?3i?3i?3g?3f>2f>2f>2e?2b@4bB5`C5_C5_C5[D4ZE4WD3XI6OB/\T?oiSc^HMK4IG0IK3>D*?H-@F*<B&<A#FI,XY:deFmmQxx\tv^[_F@C.6;%8@+<C19?39?59?58>48>47=37=37=39?58>48>48>48>48>48>48>47<67<67<66;56;56;55:45:438238216005//4.-2,,1+,1++0*+0**/)*/)+0*,1+-2,-2,05/.3-+0*+0*-2.2718=9;B:@K=@M;@M;=J68E13@,1?(0>'<I5BO;HUCIVEDPB;G;2>4.:0,6..5.-4-gdUgdUgdUheVheVifWifWifWifWifWifWifWifWifWifWifWkeWldWlcZmd[md]mf\mg[jiWkmXinWiqZgs_drejyt…••£²·½ÈÎÒÛâêïõôùýûüÿýÿþÿÿýÿÿúÿÿøÿÿóþÿñûþë÷úåðõÞæîÖâèÌÙÜ¿ÕÕ¹ÒйÔѾÚÔÈàÙÑêàßôéíþóùÿ÷üÿûÿÿýÿÿþÿÿþüÿþúÿþúþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüýÿþúÿÿúÿÿûÿÿþþþÿýþÿüüÿûüÿûúÿúøÿù÷ÿú÷ÿýøÿÿûÿÿûýÿúýþùùôîÿþöÿýôÿüóÿûîÿ÷çûæÕéÒÀÏ´¡¶š…”uawVCjI6lI6rM;uP>mF7iB3gB2jE5nI9nJ:oK;oK;eA1cB1cB1cB1cB1cB1bC1`C3]F6\G6\G6ZG6ZG6WH5TG4RG3PH3NH2fdM~~frrZVX@JL4GM3<E*<F+?H+AG+AF(DG*MN/TU6xw[{z^nnVUU=CB.?A,>A.:>-;>39?59?58>48>47=37=37=38>48>48>48>48>48>48>48>49>89>89>88=78=78=77<67<66;55:449338227105//4./4.-2.-2.,1-+0,+0,,1--2.-2.,1-,1-,1-,1-.210513764;4;G9=L9@O<@O:<K67F/4C,2A*6E.:I4@O<BP??M>8F71?2-9-,6--4,,3+fcTfcTgdUheVheVifWjgXjgXifWifWifWifWifWifWifWifWldWldWlcZmd[md]mf\mg[jiWkmXinWiqZgs_bpcgxr‚””£²·ÀËÑÔÝäëðöõúþûüÿýÿþÿÿýÿÿúÿÿøÿÿôÿÿóÿÿïúýèò÷àçðÕâèÌ×Ú½ÑѵÍË´Î˺ÖÏÅÞ×Ñìâã÷îóÿöþÿúÿÿýÿÿþÿÿÿýþÿúþÿùþÿúþÿúþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþþýþÿýþÿýþÿýÿþýÿüþÿúþÿùÿÿ÷ýýóýýõþýøþýùÿþüþþþÿýÿÿþÿþùÿÿüÿÿûüûöóÿøòÿýôÿþñÿýíÿúéÿòà÷äÓçÒ¿Ò»©¹ŸŽŸ„s�rb|]K}^LtUCaD2W:(Y<*[>,X=*`E2`E2`E2_F2`G3`G3`G3]H3\K7\M:\M:YL9XM9UM8RL6QL6NL5QQ9ikS|€gmqXSY?FL2>G,<F+<F+?H+DJ.EJ,DG*HI*LM.yx\tsWfdMTR;LI6KJ6FE3<=-<?49?59?59?58>48>48>48>47=37=38>48>48>48>49?59?5:?9:?9:?9:?99>89>89>88=78=78=77<66;55:4493382382/40.3/-2.,1-+0,+0,+0,+0,+0,,1--10.21/32/32/32-4-3?17F3=L7@O:?N7<K49H18G05D-8G0<K6>M:<J97E61?2-9-+5,-4,,3+fcTfcTgdUheVheVifWjgXjgXifWifWifWifWifWifWifWifWldWldWlcZmd[md]mf\mg[jiWjlWhmVhrZfs_cqdhys…——¥·»ÇÒØÚãêðõûøýÿüýÿýÿþÿÿýÿþùÿÿõÿÿòÿÿðüÿìøûæïôÝâëÐÝãÇÕÕ¹Î̳ÉÄ°ÊÄ´ÐÉ¿ÚÒÏéàãöíòÿ÷ÿÿúÿÿýÿÿþÿÿÿýýÿúüÿøüÿøýÿúþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþþÿüÿÿüÿþýÿýÿþúÿúøÿùøÿõøÿóøÿóøÿõüÿøýÿúÿýþÿüÿÿüÿÿûÿý÷ÿÿúÿÿüÿÿýÿÿþÿþýûûûóúøìÿÿñÿÿïÿÿíÿþíÿýíÿöçÿîàýêÛúãÑÿæÐãÊ´¥�xt^GeO8bL5\H0^J2\J2\J2]K3]K3^L4^L4\M6YN8WO:WO:VP:UO9SN8PN7NN6MO7PT;`fLfoTU^CEO4AK0<F+?I.?I.CL/HN2HM/FI,OP1YX:qmRjfKa\F[V@YTATQ>MJ9BC3=@5;A7;A7;A7:@6:@6:@69?58>48>48>49?59?5:@6:@6:@6;@:;@:;@::?9:?99>89>89>8:?9:?99>89>88=78=77<67<6495273162/40-2.,1-+/.+/.,0/,0/-10.21.23.23-12+1-+9*/@-6H2;M7=O7<N4:L2:L25G/7I1:L6<M::K96G70@3-;.,6--4,,3+cdRcdRdeSefTgeVhfWigXigXhfWifWifWifWifWifWifWkeWkeWkeWldYle[md]mf\kg[jiWikVhmWhrZgt`dqgj{u‰›�ª¼ÀÍØÞàçíôùÿúÿÿûÿÿýÿüþÿúþþöúúîúûíúûéøúåóöáêî×ßãÊÙÜÁÓѸÌDZž¬Æ¾±ÌüÖÎËæÝàôêòþ÷ÿÿúÿÿþÿÿÿÿþþþýÿúüÿúüÿúýÿüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüÿþüÿþüÿþüÿýþÿüÿÿüÿÿýÿýÿþûÿüøÿù÷ÿ÷÷ÿõøÿöùÿöüÿøÿÿûÿþýÿüÿÿûÿÿúÿÿúÿÿûÿÿüÿÿþÿþýÿûýüüþûþÿúùúòúüñýûïþúîÿùìÿûìÿýïÿýìÿðÛÿûâìÜÞŽu`P7RB)XH/XH.[M3[M3ZN4ZN4[O5[O5\P6[Q8UO7TO9TO9QO8PN7NN6KM5IM4JN5LR8S\APY<@I,;F(AL.?J,DM0EN1JQ2MR4JM0LL0YY=kjNhdIb^E^ZA_ZD`[GZUAQO:KJ8AB4@B5@B5>A6>A6>A6=@7=@79?59?59?5:@6:@6:B7:B7:B7;@9;@9;@::?8:?99>79>89>7;@:;@9;@:;@9:?9:?8:?9:?98?87=94;4382160/4..3/.3/.21.21,2.,2.+2++2*,4))6%$6)=!2F+8L1;O6;M5;M5;M58J29K3;M5<N6<N87H51B0.<--7,.5--4,]eN^fO_gP`hQdiUfiVgjWijXfgUhgUheVheVjdVkeWkeYlfXlfXkhYmg[kg\mf^jf[ieYgeVghVfkUenYbo]dqhsƒ€–¨¬¹ÈÏÒÛââéïò÷û÷üÿùýüüþùüÿöüþñ÷øêùúêø÷åòñÝîíÙëêÖàßÊ×Ò¾×ϼËÀ®½´¥½³©Â·±ÌÂÁÛÒ×êàèúóûüöÿþûÿÿþÿþýÿüýÿüþýüþýþþþþþþþþþþþþþþþþþþÿþüÿþüÿþüÿþüþýùþýùÿüùÿýúÿþûÿþûÿþÿÿýþÿüýÿýþÿÿÿþÿýýÿüûýøÿÿûüý÷ÿýøÿþúÿýøÿúöÿ÷ôÿùùÿúÿÿüÿÿýÿÿÿÿÿÿÿýÿüýÿþüþýþÿÿÿÿÿÿþüÿþùÿþóÿþîÿýêÿýæúôÜÿúàçàÆ�ˆlUN2WN1TK.SJ+WN1WN1UN1UN1TO2TO2TN4SO4QM2RN5QO6PP6MO7KO6HM6GL5JP6FO4AJ-=F'>H&DK*FM+EL*PV4PT3KN/HK,TT8bbFccIZZ@[Y@\ZA][B][B[Y@XV=US:SQ8HF1GD1DA0A?0@>1>>2??5>@5;>59?59?59A69A48C58C58C5:B79A69@88@58?79A69@8:B7:A9:B79@8:B7;B:<D9=D<>E=8B:8B:7A88@56>14</39/271.40/51-7/-9-.;)0@&6G'7M';V+:W+;U0;T4;R8=P:>O<>P::L6@R8EW=FY=DX=@T;5H2+<),6+,3++2*ZgM[hN\hP]iQbjSckTglXhkXghVghVifWifWkeWlfXmeZmgYmgYkhYkg[kg\jf]jf[hfZefVhiWglVfo\erajwnzŠ‰œ«²ºÈÑÐÙàáéìò÷ûøþþùþúúÿøûþóùýïøùë÷øæõôâíìØéæÓåâÏÜ×ÄÒ˹ʿ­Á´¤¹­¡¼¯¦Á¶²ËÁÀÛÐÖæßæûóþý÷ÿþûÿÿþÿþýÿüýÿüýÿüýÿþþþþþþþþþþþþþþþþþþÿþüÿþüÿþüÿþüÿýúÿüùÿüùÿýúÿýùÿþúÿþýÿýüÿüýÿýþÿþÿÿþÿÿýþýûüÿþýÿþûÿüúÿøóøíçòåÝöèßûðêÿúýÿüÿÿýþÿÿýÿÿûýÿüýÿþüýÿýþÿýþÿÿýÿÿþüÿþõÿÿîÿþèÿÿäúöÝýûâäàÇ‹‡lRM0SN0QK+QK+UO/UO/TO1TO1SO2SO2QP4QP4QM2PO3PN5NQ6MO7JP6HM6EM5HQ6EN1BK.?I'BI(FM+JP,KQ-NR/PT1VZ7^bA_bCY\?TW<UU=VV>WU>XV=YW>XV=XW;WV:XT9QK3OI3LG4ID1DA2B@3A?3>@5<=59?59?59A67B48C57D37D3:B79A69A68@58@59A69A6:B7:B7:B7:B7:B7;C8<D9=E:<F;=G?<H><F;;F6<D5:B39?38=67>74>63=40>-3C)9M*BY/Ga1Op;Ln;Li=Fa>BY?<R=;L::K8=O9DV<K^BM`BL`DH\@:Q71D.+5*+2*)0(ZgM[hN\hP^jRbjSdlUglXilYijXijXkhYkhYlfXmgYnf[nhZmgYkhYkg[jf[jf]ieZhfZfgWhiWejTdmZgtco|s�Ž�¬³¶ÄÍÍÖÝÞæéñöúøþþúþýùþøøûòôøêóôæðñßêé×áàÌÜÙÆ×ÔÁÐ˸ÇÀ®¿³£½° ¼°¤Ã¶®ËÀ¼ÕËÊãØÞëäëüôÿý÷ÿþûÿÿþÿþýÿüýÿüýÿýÿþþþþþþþþþþþþþþþþþþþÿþüÿþüÿþüÿþüÿýúÿýúÿüùÿüùÿüøÿüøÿýüÿüûÿûüÿüýÿþÿÿþÿÿýþþüýÿþýÿþûÿýûüóîêßÙâÕÍëÝÔ÷ìæÿúýÿüÿÿýþÿÿýÿÿûýÿüýÿþüýÿýþÿýþÿÿþÿÿÿýÿþõÿþíÿýçÿþãû÷ÞþüããßƉ…jPK.QL.QK+QK+TN.UO/TO1UP2TP3SO2QP4PO3QM2ON2PN5MP5MO7JP6IN7FN6HQ6FO2CL/CM+FM,JQ/OU1SY5QY2W_8jpJy[qvVY^@KO4MP5QQ9QQ9RP9SQ8TR9UT8YU:ZV;XR:VP:RM:NI6HE6DB5CA5?A6<=59?5:@69A67B47B47D37D3:B7:B79A69A69A69A6:B7:B7:B7:B7:B7:B7;C8<D9>F;<F;?IA>J@?I>>I9?G8>F7>D8=B;>E>9C;6@74B19I/BV3Pg=XrB^J[}JXuIMhECZ@9O:6G54E2<N8EW=NaESfHSgKNbF@W=6I3/9./6.-4,ZgM[hN\hP^jRblTemVhmYinZklZklZlj[liZnhZnhZog\oi[liZkhYjfZjf[ie\ieZhfZfgWfgUbgQajWerao|s~Ž�–¥¬«¹ÂÇÐ×Úáçîó÷öüüøüûõúôóöíîòäéêÜãäÒÜÛÉÓÒ¾Î˸ÉƳþ«½¶¤Ã·§Ç¹¬ÍÁµÖÉÁßÔÐéßÞóèîøñøýõÿþøÿÿüÿþýÿýüÿüýÿüþýýÿþþþþþþþþþþþþþþþþþþþÿþüÿþüþýûÿþüÿýúÿýúÿýúÿüùÿû÷ÿû÷ÿýüÿüûþúûÿûüÿýþÿþÿÿýþþüýÿüûÿþûÿýûýôïíâÜæÙÑðâÙüñëÿúýÿüÿÿýþÿÿýÿÿûýÿúýÿþüýÿýþÿýþÿÿþÿÿÿýÿþõÿþíÿüæÿüáþùãÿþçäßÉŠ†mPJ0RM0SL/SM-TN.UO/UP2UP2TP3SO2PO3ON2PL1NM1OM4LO4LN6JP6JO8HP8JS8GP3GP3IS1MT3OV4V\8\d=grHt�Uˆ“i�›s…�lnuTZaBRV;LN6NN6MM5NL5OM6RP7WS:YU<ZT>XR<TO<QL9KH9GE8EC7@B7=>6:@6:@69A67B47B46C26C2;C8:B79A69A69A69A6:B7;C8;C8:B7:B7;C8;C8=E:>F;=G<=G?<H>>H=>I9?G8>F7>D8=B;=D=9C;6@74B19I/DX5Ri?[uEZ{FWyFTqEIdA=T:2H3/@.->+6H2@R8L_CReGSgKOcG@W=6I30:/07/-4,YgMZhN\hP^jRblTemVinZjo[mn\lm[mk\mj[nhZnhZoi]oi[liZkhYjfZieZie\ieZig[ghXghVchRajWerao|s{‹Š�Ÿ¦¢°¹ÁËÔÔÝäéñôòúüôúøñöðíðçèìÞâãÕÚÛÉÒÑ¿ÌɶÉıþ«¿¸¦¼³¢Ê¾®ÓŸÜÐÄæÙÑíâàöìíÿôúÿùÿÿ÷ÿÿúÿÿüÿþýÿýüÿûüþüþýþÿÿþþþþþþþþþþþþþþþþþþÿþüÿþüþýûþýûÿýúÿýúÿýúÿûøÿúöþùõÿýüÿûúýùúþúûþüýÿýþÿþÿÿýþÿüûÿýúÿüúÿøó÷ìæòåÝøêáÿôîÿúýÿüÿÿýþÿÿýÿÿûýÿúýÿþüýÿüýÿýþÿÿþÿÿÿýÿþõÿýìÿûãþúßÿüæÿÿéçâÌŒˆoRL2TO2TM0UN1TN.UO/VQ3VQ3UQ4SO2ON2NM1PL1NM1NL3KN3LN6KQ7KP9JR:LU:IR5JS6OY7SZ9T[9]c?fnG‚’c’¦sŸ±�›ª’¡zˆ”pr{\\bFLP7KM5LK6KI4LJ5PK5TN8VP:WQ;VP:SN;QL9LI:IG:GE9CE:=>6:@6:@69A67B47B46C26C2;C8;C8:B79A69A6:B7;C8;C8;C8;C8;C8;C8<D9=E:>F;=G<=G?<H>=G<=H8>F7=E6=C7=B;:A:7A95?64B19I/BV3Ne;Uo?Tu@RtAPmAE`=9P6-C.+<*+<)1C-<N4H[?PcERfJNbF@W=5H2,6++2*'.&
\ No newline at end of file
diff --git a/libjpegtwrp/testimgp.jpg b/libjpegtwrp/testimgp.jpg
new file mode 100644
index 000000000..8cbb658bb
--- /dev/null
+++ b/libjpegtwrp/testimgp.jpg
diff --git a/libjpegtwrp/testorig.jpg b/libjpegtwrp/testorig.jpg
new file mode 100644
index 000000000..9816a0c62
--- /dev/null
+++ b/libjpegtwrp/testorig.jpg
diff --git a/libjpegtwrp/testprog.jpg b/libjpegtwrp/testprog.jpg
new file mode 100644
index 000000000..920fee2e3
--- /dev/null
+++ b/libjpegtwrp/testprog.jpg
diff --git a/libjpegtwrp/transupp.c b/libjpegtwrp/transupp.c
new file mode 100644
index 000000000..e5ec5642f
--- /dev/null
+++ b/libjpegtwrp/transupp.c
@@ -0,0 +1,928 @@
+/*
+ * transupp.c
+ *
+ * Copyright (C) 1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains image transformation routines and other utility code
+ * used by the jpegtran sample application.  These are NOT part of the core
+ * JPEG library.  But we keep these routines separate from jpegtran.c to
+ * ease the task of maintaining jpegtran-like programs that have other user
+ * interfaces.
+ */
+
+/* Although this file really shouldn't have access to the library internals,
+ * it's helpful to let it call jround_up() and jcopy_block_row().
+ */
+#define JPEG_INTERNALS
+
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "transupp.h"		/* My own external interface */
+
+
+#if TRANSFORMS_SUPPORTED
+
+/*
+ * Lossless image transformation routines.  These routines work on DCT
+ * coefficient arrays and thus do not require any lossy decompression
+ * or recompression of the image.
+ * Thanks to Guido Vollbeding for the initial design and code of this feature.
+ *
+ * Horizontal flipping is done in-place, using a single top-to-bottom
+ * pass through the virtual source array.  It will thus be much the
+ * fastest option for images larger than main memory.
+ *
+ * The other routines require a set of destination virtual arrays, so they
+ * need twice as much memory as jpegtran normally does.  The destination
+ * arrays are always written in normal scan order (top to bottom) because
+ * the virtual array manager expects this.  The source arrays will be scanned
+ * in the corresponding order, which means multiple passes through the source
+ * arrays for most of the transforms.  That could result in much thrashing
+ * if the image is larger than main memory.
+ *
+ * Some notes about the operating environment of the individual transform
+ * routines:
+ * 1. Both the source and destination virtual arrays are allocated from the
+ *    source JPEG object, and therefore should be manipulated by calling the
+ *    source's memory manager.
+ * 2. The destination's component count should be used.  It may be smaller
+ *    than the source's when forcing to grayscale.
+ * 3. Likewise the destination's sampling factors should be used.  When
+ *    forcing to grayscale the destination's sampling factors will be all 1,
+ *    and we may as well take that as the effective iMCU size.
+ * 4. When "trim" is in effect, the destination's dimensions will be the
+ *    trimmed values but the source's will be untrimmed.
+ * 5. All the routines assume that the source and destination buffers are
+ *    padded out to a full iMCU boundary.  This is true, although for the
+ *    source buffer it is an undocumented property of jdcoefct.c.
+ * Notes 2,3,4 boil down to this: generally we should use the destination's
+ * dimensions and ignore the source's.
+ */
+
+
+LOCAL(void)
+do_flip_h (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+	   jvirt_barray_ptr *src_coef_arrays)
+/* Horizontal flip; done in-place, so no separate dest array is required */
+{
+  JDIMENSION MCU_cols, comp_width, blk_x, blk_y;
+  int ci, k, offset_y;
+  JBLOCKARRAY buffer;
+  JCOEFPTR ptr1, ptr2;
+  JCOEF temp1, temp2;
+  jpeg_component_info *compptr;
+
+  /* Horizontal mirroring of DCT blocks is accomplished by swapping
+   * pairs of blocks in-place.  Within a DCT block, we perform horizontal
+   * mirroring by changing the signs of odd-numbered columns.
+   * Partial iMCUs at the right edge are left untouched.
+   */
+  MCU_cols = dstinfo->image_width / (dstinfo->max_h_samp_factor * DCTSIZE);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    for (blk_y = 0; blk_y < compptr->height_in_blocks;
+	 blk_y += compptr->v_samp_factor) {
+      buffer = (*srcinfo->mem->access_virt_barray)
+	((j_common_ptr) srcinfo, src_coef_arrays[ci], blk_y,
+	 (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+	for (blk_x = 0; blk_x * 2 < comp_width; blk_x++) {
+	  ptr1 = buffer[offset_y][blk_x];
+	  ptr2 = buffer[offset_y][comp_width - blk_x - 1];
+	  /* this unrolled loop doesn't need to know which row it's on... */
+	  for (k = 0; k < DCTSIZE2; k += 2) {
+	    temp1 = *ptr1;	/* swap even column */
+	    temp2 = *ptr2;
+	    *ptr1++ = temp2;
+	    *ptr2++ = temp1;
+	    temp1 = *ptr1;	/* swap odd column with sign change */
+	    temp2 = *ptr2;
+	    *ptr1++ = -temp2;
+	    *ptr2++ = -temp1;
+	  }
+	}
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_flip_v (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+	   jvirt_barray_ptr *src_coef_arrays,
+	   jvirt_barray_ptr *dst_coef_arrays)
+/* Vertical flip */
+{
+  JDIMENSION MCU_rows, comp_height, dst_blk_x, dst_blk_y;
+  int ci, i, j, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* We output into a separate array because we can't touch different
+   * rows of the source virtual array simultaneously.  Otherwise, this
+   * is a pretty straightforward analog of horizontal flip.
+   * Within a DCT block, vertical mirroring is done by changing the signs
+   * of odd-numbered rows.
+   * Partial iMCUs at the bottom edge are copied verbatim.
+   */
+  MCU_rows = dstinfo->image_height / (dstinfo->max_v_samp_factor * DCTSIZE);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+	 dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+	((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+	 (JDIMENSION) compptr->v_samp_factor, TRUE);
+      if (dst_blk_y < comp_height) {
+	/* Row is within the mirrorable area. */
+	src_buffer = (*srcinfo->mem->access_virt_barray)
+	  ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+	   comp_height - dst_blk_y - (JDIMENSION) compptr->v_samp_factor,
+	   (JDIMENSION) compptr->v_samp_factor, FALSE);
+      } else {
+	/* Bottom-edge blocks will be copied verbatim. */
+	src_buffer = (*srcinfo->mem->access_virt_barray)
+	  ((j_common_ptr) srcinfo, src_coef_arrays[ci], dst_blk_y,
+	   (JDIMENSION) compptr->v_samp_factor, FALSE);
+      }
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+	if (dst_blk_y < comp_height) {
+	  /* Row is within the mirrorable area. */
+	  dst_row_ptr = dst_buffer[offset_y];
+	  src_row_ptr = src_buffer[compptr->v_samp_factor - offset_y - 1];
+	  for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+	       dst_blk_x++) {
+	    dst_ptr = dst_row_ptr[dst_blk_x];
+	    src_ptr = src_row_ptr[dst_blk_x];
+	    for (i = 0; i < DCTSIZE; i += 2) {
+	      /* copy even row */
+	      for (j = 0; j < DCTSIZE; j++)
+		*dst_ptr++ = *src_ptr++;
+	      /* copy odd row with sign change */
+	      for (j = 0; j < DCTSIZE; j++)
+		*dst_ptr++ = - *src_ptr++;
+	    }
+	  }
+	} else {
+	  /* Just copy row verbatim. */
+	  jcopy_block_row(src_buffer[offset_y], dst_buffer[offset_y],
+			  compptr->width_in_blocks);
+	}
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_transpose (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+	      jvirt_barray_ptr *src_coef_arrays,
+	      jvirt_barray_ptr *dst_coef_arrays)
+/* Transpose source into destination */
+{
+  JDIMENSION dst_blk_x, dst_blk_y;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Transposing pixels within a block just requires transposing the
+   * DCT coefficients.
+   * Partial iMCUs at the edges require no special treatment; we simply
+   * process all the available DCT blocks for every component.
+   */
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+	 dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+	((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+	 (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+	for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+	     dst_blk_x += compptr->h_samp_factor) {
+	  src_buffer = (*srcinfo->mem->access_virt_barray)
+	    ((j_common_ptr) srcinfo, src_coef_arrays[ci], dst_blk_x,
+	     (JDIMENSION) compptr->h_samp_factor, FALSE);
+	  for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+	    src_ptr = src_buffer[offset_x][dst_blk_y + offset_y];
+	    dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+	    for (i = 0; i < DCTSIZE; i++)
+	      for (j = 0; j < DCTSIZE; j++)
+		dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+	  }
+	}
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_rot_90 (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+	   jvirt_barray_ptr *src_coef_arrays,
+	   jvirt_barray_ptr *dst_coef_arrays)
+/* 90 degree rotation is equivalent to
+ *   1. Transposing the image;
+ *   2. Horizontal mirroring.
+ * These two steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_cols, comp_width, dst_blk_x, dst_blk_y;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Because of the horizontal mirror step, we can't process partial iMCUs
+   * at the (output) right edge properly.  They just get transposed and
+   * not mirrored.
+   */
+  MCU_cols = dstinfo->image_width / (dstinfo->max_h_samp_factor * DCTSIZE);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+	 dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+	((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+	 (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+	for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+	     dst_blk_x += compptr->h_samp_factor) {
+	  src_buffer = (*srcinfo->mem->access_virt_barray)
+	    ((j_common_ptr) srcinfo, src_coef_arrays[ci], dst_blk_x,
+	     (JDIMENSION) compptr->h_samp_factor, FALSE);
+	  for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+	    src_ptr = src_buffer[offset_x][dst_blk_y + offset_y];
+	    if (dst_blk_x < comp_width) {
+	      /* Block is within the mirrorable area. */
+	      dst_ptr = dst_buffer[offset_y]
+		[comp_width - dst_blk_x - offset_x - 1];
+	      for (i = 0; i < DCTSIZE; i++) {
+		for (j = 0; j < DCTSIZE; j++)
+		  dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+		i++;
+		for (j = 0; j < DCTSIZE; j++)
+		  dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+	      }
+	    } else {
+	      /* Edge blocks are transposed but not mirrored. */
+	      dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+	      for (i = 0; i < DCTSIZE; i++)
+		for (j = 0; j < DCTSIZE; j++)
+		  dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+	    }
+	  }
+	}
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_rot_270 (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+	    jvirt_barray_ptr *src_coef_arrays,
+	    jvirt_barray_ptr *dst_coef_arrays)
+/* 270 degree rotation is equivalent to
+ *   1. Horizontal mirroring;
+ *   2. Transposing the image.
+ * These two steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_rows, comp_height, dst_blk_x, dst_blk_y;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Because of the horizontal mirror step, we can't process partial iMCUs
+   * at the (output) bottom edge properly.  They just get transposed and
+   * not mirrored.
+   */
+  MCU_rows = dstinfo->image_height / (dstinfo->max_v_samp_factor * DCTSIZE);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+	 dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+	((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+	 (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+	for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+	     dst_blk_x += compptr->h_samp_factor) {
+	  src_buffer = (*srcinfo->mem->access_virt_barray)
+	    ((j_common_ptr) srcinfo, src_coef_arrays[ci], dst_blk_x,
+	     (JDIMENSION) compptr->h_samp_factor, FALSE);
+	  for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+	    dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+	    if (dst_blk_y < comp_height) {
+	      /* Block is within the mirrorable area. */
+	      src_ptr = src_buffer[offset_x]
+		[comp_height - dst_blk_y - offset_y - 1];
+	      for (i = 0; i < DCTSIZE; i++) {
+		for (j = 0; j < DCTSIZE; j++) {
+		  dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+		  j++;
+		  dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+		}
+	      }
+	    } else {
+	      /* Edge blocks are transposed but not mirrored. */
+	      src_ptr = src_buffer[offset_x][dst_blk_y + offset_y];
+	      for (i = 0; i < DCTSIZE; i++)
+		for (j = 0; j < DCTSIZE; j++)
+		  dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+	    }
+	  }
+	}
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_rot_180 (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+	    jvirt_barray_ptr *src_coef_arrays,
+	    jvirt_barray_ptr *dst_coef_arrays)
+/* 180 degree rotation is equivalent to
+ *   1. Vertical mirroring;
+ *   2. Horizontal mirroring.
+ * These two steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_cols, MCU_rows, comp_width, comp_height, dst_blk_x, dst_blk_y;
+  int ci, i, j, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  MCU_cols = dstinfo->image_width / (dstinfo->max_h_samp_factor * DCTSIZE);
+  MCU_rows = dstinfo->image_height / (dstinfo->max_v_samp_factor * DCTSIZE);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+	 dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+	((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+	 (JDIMENSION) compptr->v_samp_factor, TRUE);
+      if (dst_blk_y < comp_height) {
+	/* Row is within the vertically mirrorable area. */
+	src_buffer = (*srcinfo->mem->access_virt_barray)
+	  ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+	   comp_height - dst_blk_y - (JDIMENSION) compptr->v_samp_factor,
+	   (JDIMENSION) compptr->v_samp_factor, FALSE);
+      } else {
+	/* Bottom-edge rows are only mirrored horizontally. */
+	src_buffer = (*srcinfo->mem->access_virt_barray)
+	  ((j_common_ptr) srcinfo, src_coef_arrays[ci], dst_blk_y,
+	   (JDIMENSION) compptr->v_samp_factor, FALSE);
+      }
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+	if (dst_blk_y < comp_height) {
+	  /* Row is within the mirrorable area. */
+	  dst_row_ptr = dst_buffer[offset_y];
+	  src_row_ptr = src_buffer[compptr->v_samp_factor - offset_y - 1];
+	  /* Process the blocks that can be mirrored both ways. */
+	  for (dst_blk_x = 0; dst_blk_x < comp_width; dst_blk_x++) {
+	    dst_ptr = dst_row_ptr[dst_blk_x];
+	    src_ptr = src_row_ptr[comp_width - dst_blk_x - 1];
+	    for (i = 0; i < DCTSIZE; i += 2) {
+	      /* For even row, negate every odd column. */
+	      for (j = 0; j < DCTSIZE; j += 2) {
+		*dst_ptr++ = *src_ptr++;
+		*dst_ptr++ = - *src_ptr++;
+	      }
+	      /* For odd row, negate every even column. */
+	      for (j = 0; j < DCTSIZE; j += 2) {
+		*dst_ptr++ = - *src_ptr++;
+		*dst_ptr++ = *src_ptr++;
+	      }
+	    }
+	  }
+	  /* Any remaining right-edge blocks are only mirrored vertically. */
+	  for (; dst_blk_x < compptr->width_in_blocks; dst_blk_x++) {
+	    dst_ptr = dst_row_ptr[dst_blk_x];
+	    src_ptr = src_row_ptr[dst_blk_x];
+	    for (i = 0; i < DCTSIZE; i += 2) {
+	      for (j = 0; j < DCTSIZE; j++)
+		*dst_ptr++ = *src_ptr++;
+	      for (j = 0; j < DCTSIZE; j++)
+		*dst_ptr++ = - *src_ptr++;
+	    }
+	  }
+	} else {
+	  /* Remaining rows are just mirrored horizontally. */
+	  dst_row_ptr = dst_buffer[offset_y];
+	  src_row_ptr = src_buffer[offset_y];
+	  /* Process the blocks that can be mirrored. */
+	  for (dst_blk_x = 0; dst_blk_x < comp_width; dst_blk_x++) {
+	    dst_ptr = dst_row_ptr[dst_blk_x];
+	    src_ptr = src_row_ptr[comp_width - dst_blk_x - 1];
+	    for (i = 0; i < DCTSIZE2; i += 2) {
+	      *dst_ptr++ = *src_ptr++;
+	      *dst_ptr++ = - *src_ptr++;
+	    }
+	  }
+	  /* Any remaining right-edge blocks are only copied. */
+	  for (; dst_blk_x < compptr->width_in_blocks; dst_blk_x++) {
+	    dst_ptr = dst_row_ptr[dst_blk_x];
+	    src_ptr = src_row_ptr[dst_blk_x];
+	    for (i = 0; i < DCTSIZE2; i++)
+	      *dst_ptr++ = *src_ptr++;
+	  }
+	}
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_transverse (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+	       jvirt_barray_ptr *src_coef_arrays,
+	       jvirt_barray_ptr *dst_coef_arrays)
+/* Transverse transpose is equivalent to
+ *   1. 180 degree rotation;
+ *   2. Transposition;
+ * or
+ *   1. Horizontal mirroring;
+ *   2. Transposition;
+ *   3. Horizontal mirroring.
+ * These steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_cols, MCU_rows, comp_width, comp_height, dst_blk_x, dst_blk_y;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  MCU_cols = dstinfo->image_width / (dstinfo->max_h_samp_factor * DCTSIZE);
+  MCU_rows = dstinfo->image_height / (dstinfo->max_v_samp_factor * DCTSIZE);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+	 dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+	((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+	 (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+	for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+	     dst_blk_x += compptr->h_samp_factor) {
+	  src_buffer = (*srcinfo->mem->access_virt_barray)
+	    ((j_common_ptr) srcinfo, src_coef_arrays[ci], dst_blk_x,
+	     (JDIMENSION) compptr->h_samp_factor, FALSE);
+	  for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+	    if (dst_blk_y < comp_height) {
+	      src_ptr = src_buffer[offset_x]
+		[comp_height - dst_blk_y - offset_y - 1];
+	      if (dst_blk_x < comp_width) {
+		/* Block is within the mirrorable area. */
+		dst_ptr = dst_buffer[offset_y]
+		  [comp_width - dst_blk_x - offset_x - 1];
+		for (i = 0; i < DCTSIZE; i++) {
+		  for (j = 0; j < DCTSIZE; j++) {
+		    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+		    j++;
+		    dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+		  }
+		  i++;
+		  for (j = 0; j < DCTSIZE; j++) {
+		    dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+		    j++;
+		    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+		  }
+		}
+	      } else {
+		/* Right-edge blocks are mirrored in y only */
+		dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+		for (i = 0; i < DCTSIZE; i++) {
+		  for (j = 0; j < DCTSIZE; j++) {
+		    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+		    j++;
+		    dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+		  }
+		}
+	      }
+	    } else {
+	      src_ptr = src_buffer[offset_x][dst_blk_y + offset_y];
+	      if (dst_blk_x < comp_width) {
+		/* Bottom-edge blocks are mirrored in x only */
+		dst_ptr = dst_buffer[offset_y]
+		  [comp_width - dst_blk_x - offset_x - 1];
+		for (i = 0; i < DCTSIZE; i++) {
+		  for (j = 0; j < DCTSIZE; j++)
+		    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+		  i++;
+		  for (j = 0; j < DCTSIZE; j++)
+		    dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+		}
+	      } else {
+		/* At lower right corner, just transpose, no mirroring */
+		dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+		for (i = 0; i < DCTSIZE; i++)
+		  for (j = 0; j < DCTSIZE; j++)
+		    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+	      }
+	    }
+	  }
+	}
+      }
+    }
+  }
+}
+
+
+/* Request any required workspace.
+ *
+ * We allocate the workspace virtual arrays from the source decompression
+ * object, so that all the arrays (both the original data and the workspace)
+ * will be taken into account while making memory management decisions.
+ * Hence, this routine must be called after jpeg_read_header (which reads
+ * the image dimensions) and before jpeg_read_coefficients (which realizes
+ * the source's virtual arrays).
+ */
+
+GLOBAL(void)
+jtransform_request_workspace (j_decompress_ptr srcinfo,
+			      jpeg_transform_info *info)
+{
+  jvirt_barray_ptr *coef_arrays = NULL;
+  jpeg_component_info *compptr;
+  int ci;
+
+  if (info->force_grayscale &&
+      srcinfo->jpeg_color_space == JCS_YCbCr &&
+      srcinfo->num_components == 3) {
+    /* We'll only process the first component */
+    info->num_components = 1;
+  } else {
+    /* Process all the components */
+    info->num_components = srcinfo->num_components;
+  }
+
+  switch (info->transform) {
+  case JXFORM_NONE:
+  case JXFORM_FLIP_H:
+    /* Don't need a workspace array */
+    break;
+  case JXFORM_FLIP_V:
+  case JXFORM_ROT_180:
+    /* Need workspace arrays having same dimensions as source image.
+     * Note that we allocate arrays padded out to the next iMCU boundary,
+     * so that transform routines need not worry about missing edge blocks.
+     */
+    coef_arrays = (jvirt_barray_ptr *)
+      (*srcinfo->mem->alloc_small) ((j_common_ptr) srcinfo, JPOOL_IMAGE,
+	SIZEOF(jvirt_barray_ptr) * info->num_components);
+    for (ci = 0; ci < info->num_components; ci++) {
+      compptr = srcinfo->comp_info + ci;
+      coef_arrays[ci] = (*srcinfo->mem->request_virt_barray)
+	((j_common_ptr) srcinfo, JPOOL_IMAGE, FALSE,
+	 (JDIMENSION) jround_up((long) compptr->width_in_blocks,
+				(long) compptr->h_samp_factor),
+	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
+				(long) compptr->v_samp_factor),
+	 (JDIMENSION) compptr->v_samp_factor);
+    }
+    break;
+  case JXFORM_TRANSPOSE:
+  case JXFORM_TRANSVERSE:
+  case JXFORM_ROT_90:
+  case JXFORM_ROT_270:
+    /* Need workspace arrays having transposed dimensions.
+     * Note that we allocate arrays padded out to the next iMCU boundary,
+     * so that transform routines need not worry about missing edge blocks.
+     */
+    coef_arrays = (jvirt_barray_ptr *)
+      (*srcinfo->mem->alloc_small) ((j_common_ptr) srcinfo, JPOOL_IMAGE,
+	SIZEOF(jvirt_barray_ptr) * info->num_components);
+    for (ci = 0; ci < info->num_components; ci++) {
+      compptr = srcinfo->comp_info + ci;
+      coef_arrays[ci] = (*srcinfo->mem->request_virt_barray)
+	((j_common_ptr) srcinfo, JPOOL_IMAGE, FALSE,
+	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
+				(long) compptr->v_samp_factor),
+	 (JDIMENSION) jround_up((long) compptr->width_in_blocks,
+				(long) compptr->h_samp_factor),
+	 (JDIMENSION) compptr->h_samp_factor);
+    }
+    break;
+  }
+  info->workspace_coef_arrays = coef_arrays;
+}
+
+
+/* Transpose destination image parameters */
+
+LOCAL(void)
+transpose_critical_parameters (j_compress_ptr dstinfo)
+{
+  int tblno, i, j, ci, itemp;
+  jpeg_component_info *compptr;
+  JQUANT_TBL *qtblptr;
+  JDIMENSION dtemp;
+  UINT16 qtemp;
+
+  /* Transpose basic image dimensions */
+  dtemp = dstinfo->image_width;
+  dstinfo->image_width = dstinfo->image_height;
+  dstinfo->image_height = dtemp;
+
+  /* Transpose sampling factors */
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    itemp = compptr->h_samp_factor;
+    compptr->h_samp_factor = compptr->v_samp_factor;
+    compptr->v_samp_factor = itemp;
+  }
+
+  /* Transpose quantization tables */
+  for (tblno = 0; tblno < NUM_QUANT_TBLS; tblno++) {
+    qtblptr = dstinfo->quant_tbl_ptrs[tblno];
+    if (qtblptr != NULL) {
+      for (i = 0; i < DCTSIZE; i++) {
+	for (j = 0; j < i; j++) {
+	  qtemp = qtblptr->quantval[i*DCTSIZE+j];
+	  qtblptr->quantval[i*DCTSIZE+j] = qtblptr->quantval[j*DCTSIZE+i];
+	  qtblptr->quantval[j*DCTSIZE+i] = qtemp;
+	}
+      }
+    }
+  }
+}
+
+
+/* Trim off any partial iMCUs on the indicated destination edge */
+
+LOCAL(void)
+trim_right_edge (j_compress_ptr dstinfo)
+{
+  int ci, max_h_samp_factor;
+  JDIMENSION MCU_cols;
+
+  /* We have to compute max_h_samp_factor ourselves,
+   * because it hasn't been set yet in the destination
+   * (and we don't want to use the source's value).
+   */
+  max_h_samp_factor = 1;
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    int h_samp_factor = dstinfo->comp_info[ci].h_samp_factor;
+    max_h_samp_factor = MAX(max_h_samp_factor, h_samp_factor);
+  }
+  MCU_cols = dstinfo->image_width / (max_h_samp_factor * DCTSIZE);
+  if (MCU_cols > 0)		/* can't trim to 0 pixels */
+    dstinfo->image_width = MCU_cols * (max_h_samp_factor * DCTSIZE);
+}
+
+LOCAL(void)
+trim_bottom_edge (j_compress_ptr dstinfo)
+{
+  int ci, max_v_samp_factor;
+  JDIMENSION MCU_rows;
+
+  /* We have to compute max_v_samp_factor ourselves,
+   * because it hasn't been set yet in the destination
+   * (and we don't want to use the source's value).
+   */
+  max_v_samp_factor = 1;
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    int v_samp_factor = dstinfo->comp_info[ci].v_samp_factor;
+    max_v_samp_factor = MAX(max_v_samp_factor, v_samp_factor);
+  }
+  MCU_rows = dstinfo->image_height / (max_v_samp_factor * DCTSIZE);
+  if (MCU_rows > 0)		/* can't trim to 0 pixels */
+    dstinfo->image_height = MCU_rows * (max_v_samp_factor * DCTSIZE);
+}
+
+
+/* Adjust output image parameters as needed.
+ *
+ * This must be called after jpeg_copy_critical_parameters()
+ * and before jpeg_write_coefficients().
+ *
+ * The return value is the set of virtual coefficient arrays to be written
+ * (either the ones allocated by jtransform_request_workspace, or the
+ * original source data arrays).  The caller will need to pass this value
+ * to jpeg_write_coefficients().
+ */
+
+GLOBAL(jvirt_barray_ptr *)
+jtransform_adjust_parameters (j_decompress_ptr srcinfo,
+			      j_compress_ptr dstinfo,
+			      jvirt_barray_ptr *src_coef_arrays,
+			      jpeg_transform_info *info)
+{
+  /* If force-to-grayscale is requested, adjust destination parameters */
+  if (info->force_grayscale) {
+    /* We use jpeg_set_colorspace to make sure subsidiary settings get fixed
+     * properly.  Among other things, the target h_samp_factor & v_samp_factor
+     * will get set to 1, which typically won't match the source.
+     * In fact we do this even if the source is already grayscale; that
+     * provides an easy way of coercing a grayscale JPEG with funny sampling
+     * factors to the customary 1,1.  (Some decoders fail on other factors.)
+     */
+    if ((dstinfo->jpeg_color_space == JCS_YCbCr &&
+	 dstinfo->num_components == 3) ||
+	(dstinfo->jpeg_color_space == JCS_GRAYSCALE &&
+	 dstinfo->num_components == 1)) {
+      /* We have to preserve the source's quantization table number. */
+      int sv_quant_tbl_no = dstinfo->comp_info[0].quant_tbl_no;
+      jpeg_set_colorspace(dstinfo, JCS_GRAYSCALE);
+      dstinfo->comp_info[0].quant_tbl_no = sv_quant_tbl_no;
+    } else {
+      /* Sorry, can't do it */
+      ERREXIT(dstinfo, JERR_CONVERSION_NOTIMPL);
+    }
+  }
+
+  /* Correct the destination's image dimensions etc if necessary */
+  switch (info->transform) {
+  case JXFORM_NONE:
+    /* Nothing to do */
+    break;
+  case JXFORM_FLIP_H:
+    if (info->trim)
+      trim_right_edge(dstinfo);
+    break;
+  case JXFORM_FLIP_V:
+    if (info->trim)
+      trim_bottom_edge(dstinfo);
+    break;
+  case JXFORM_TRANSPOSE:
+    transpose_critical_parameters(dstinfo);
+    /* transpose does NOT have to trim anything */
+    break;
+  case JXFORM_TRANSVERSE:
+    transpose_critical_parameters(dstinfo);
+    if (info->trim) {
+      trim_right_edge(dstinfo);
+      trim_bottom_edge(dstinfo);
+    }
+    break;
+  case JXFORM_ROT_90:
+    transpose_critical_parameters(dstinfo);
+    if (info->trim)
+      trim_right_edge(dstinfo);
+    break;
+  case JXFORM_ROT_180:
+    if (info->trim) {
+      trim_right_edge(dstinfo);
+      trim_bottom_edge(dstinfo);
+    }
+    break;
+  case JXFORM_ROT_270:
+    transpose_critical_parameters(dstinfo);
+    if (info->trim)
+      trim_bottom_edge(dstinfo);
+    break;
+  }
+
+  /* Return the appropriate output data set */
+  if (info->workspace_coef_arrays != NULL)
+    return info->workspace_coef_arrays;
+  return src_coef_arrays;
+}
+
+
+/* Execute the actual transformation, if any.
+ *
+ * This must be called *after* jpeg_write_coefficients, because it depends
+ * on jpeg_write_coefficients to have computed subsidiary values such as
+ * the per-component width and height fields in the destination object.
+ *
+ * Note that some transformations will modify the source data arrays!
+ */
+
+GLOBAL(void)
+jtransform_execute_transformation (j_decompress_ptr srcinfo,
+				   j_compress_ptr dstinfo,
+				   jvirt_barray_ptr *src_coef_arrays,
+				   jpeg_transform_info *info)
+{
+  jvirt_barray_ptr *dst_coef_arrays = info->workspace_coef_arrays;
+
+  switch (info->transform) {
+  case JXFORM_NONE:
+    break;
+  case JXFORM_FLIP_H:
+    do_flip_h(srcinfo, dstinfo, src_coef_arrays);
+    break;
+  case JXFORM_FLIP_V:
+    do_flip_v(srcinfo, dstinfo, src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_TRANSPOSE:
+    do_transpose(srcinfo, dstinfo, src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_TRANSVERSE:
+    do_transverse(srcinfo, dstinfo, src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_ROT_90:
+    do_rot_90(srcinfo, dstinfo, src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_ROT_180:
+    do_rot_180(srcinfo, dstinfo, src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_ROT_270:
+    do_rot_270(srcinfo, dstinfo, src_coef_arrays, dst_coef_arrays);
+    break;
+  }
+}
+
+#endif /* TRANSFORMS_SUPPORTED */
+
+
+/* Setup decompression object to save desired markers in memory.
+ * This must be called before jpeg_read_header() to have the desired effect.
+ */
+
+GLOBAL(void)
+jcopy_markers_setup (j_decompress_ptr srcinfo, JCOPY_OPTION option)
+{
+#ifdef SAVE_MARKERS_SUPPORTED
+  int m;
+
+  /* Save comments except under NONE option */
+  if (option != JCOPYOPT_NONE) {
+    jpeg_save_markers(srcinfo, JPEG_COM, 0xFFFF);
+  }
+  /* Save all types of APPn markers iff ALL option */
+  if (option == JCOPYOPT_ALL) {
+    for (m = 0; m < 16; m++)
+      jpeg_save_markers(srcinfo, JPEG_APP0 + m, 0xFFFF);
+  }
+#endif /* SAVE_MARKERS_SUPPORTED */
+}
+
+/* Copy markers saved in the given source object to the destination object.
+ * This should be called just after jpeg_start_compress() or
+ * jpeg_write_coefficients().
+ * Note that those routines will have written the SOI, and also the
+ * JFIF APP0 or Adobe APP14 markers if selected.
+ */
+
+GLOBAL(void)
+jcopy_markers_execute (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+		       JCOPY_OPTION option)
+{
+  jpeg_saved_marker_ptr marker;
+
+  /* In the current implementation, we don't actually need to examine the
+   * option flag here; we just copy everything that got saved.
+   * But to avoid confusion, we do not output JFIF and Adobe APP14 markers
+   * if the encoder library already wrote one.
+   */
+  for (marker = srcinfo->marker_list; marker != NULL; marker = marker->next) {
+    if (dstinfo->write_JFIF_header &&
+	marker->marker == JPEG_APP0 &&
+	marker->data_length >= 5 &&
+	GETJOCTET(marker->data[0]) == 0x4A &&
+	GETJOCTET(marker->data[1]) == 0x46 &&
+	GETJOCTET(marker->data[2]) == 0x49 &&
+	GETJOCTET(marker->data[3]) == 0x46 &&
+	GETJOCTET(marker->data[4]) == 0)
+      continue;			/* reject duplicate JFIF */
+    if (dstinfo->write_Adobe_marker &&
+	marker->marker == JPEG_APP0+14 &&
+	marker->data_length >= 5 &&
+	GETJOCTET(marker->data[0]) == 0x41 &&
+	GETJOCTET(marker->data[1]) == 0x64 &&
+	GETJOCTET(marker->data[2]) == 0x6F &&
+	GETJOCTET(marker->data[3]) == 0x62 &&
+	GETJOCTET(marker->data[4]) == 0x65)
+      continue;			/* reject duplicate Adobe */
+#ifdef NEED_FAR_POINTERS
+    /* We could use jpeg_write_marker if the data weren't FAR... */
+    {
+      unsigned int i;
+      jpeg_write_m_header(dstinfo, marker->marker, marker->data_length);
+      for (i = 0; i < marker->data_length; i++)
+	jpeg_write_m_byte(dstinfo, marker->data[i]);
+    }
+#else
+    jpeg_write_marker(dstinfo, marker->marker,
+		      marker->data, marker->data_length);
+#endif
+  }
+}
diff --git a/libjpegtwrp/transupp.h b/libjpegtwrp/transupp.h
new file mode 100644
index 000000000..5c2d32aff
--- /dev/null
+++ b/libjpegtwrp/transupp.h
@@ -0,0 +1,135 @@
+/*
+ * transupp.h
+ *
+ * Copyright (C) 1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains declarations for image transformation routines and
+ * other utility code used by the jpegtran sample application.  These are
+ * NOT part of the core JPEG library.  But we keep these routines separate
+ * from jpegtran.c to ease the task of maintaining jpegtran-like programs
+ * that have other user interfaces.
+ *
+ * NOTE: all the routines declared here have very specific requirements
+ * about when they are to be executed during the reading and writing of the
+ * source and destination files.  See the comments in transupp.c, or see
+ * jpegtran.c for an example of correct usage.
+ */
+
+/* If you happen not to want the image transform support, disable it here */
+#ifndef TRANSFORMS_SUPPORTED
+#define TRANSFORMS_SUPPORTED 1		/* 0 disables transform code */
+#endif
+
+/* Short forms of external names for systems with brain-damaged linkers. */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jtransform_request_workspace		jTrRequest
+#define jtransform_adjust_parameters		jTrAdjust
+#define jtransform_execute_transformation	jTrExec
+#define jcopy_markers_setup			jCMrkSetup
+#define jcopy_markers_execute			jCMrkExec
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+
+/*
+ * Codes for supported types of image transformations.
+ */
+
+typedef enum {
+	JXFORM_NONE,		/* no transformation */
+	JXFORM_FLIP_H,		/* horizontal flip */
+	JXFORM_FLIP_V,		/* vertical flip */
+	JXFORM_TRANSPOSE,	/* transpose across UL-to-LR axis */
+	JXFORM_TRANSVERSE,	/* transpose across UR-to-LL axis */
+	JXFORM_ROT_90,		/* 90-degree clockwise rotation */
+	JXFORM_ROT_180,		/* 180-degree rotation */
+	JXFORM_ROT_270		/* 270-degree clockwise (or 90 ccw) */
+} JXFORM_CODE;
+
+/*
+ * Although rotating and flipping data expressed as DCT coefficients is not
+ * hard, there is an asymmetry in the JPEG format specification for images
+ * whose dimensions aren't multiples of the iMCU size.  The right and bottom
+ * image edges are padded out to the next iMCU boundary with junk data; but
+ * no padding is possible at the top and left edges.  If we were to flip
+ * the whole image including the pad data, then pad garbage would become
+ * visible at the top and/or left, and real pixels would disappear into the
+ * pad margins --- perhaps permanently, since encoders & decoders may not
+ * bother to preserve DCT blocks that appear to be completely outside the
+ * nominal image area.  So, we have to exclude any partial iMCUs from the
+ * basic transformation.
+ *
+ * Transpose is the only transformation that can handle partial iMCUs at the
+ * right and bottom edges completely cleanly.  flip_h can flip partial iMCUs
+ * at the bottom, but leaves any partial iMCUs at the right edge untouched.
+ * Similarly flip_v leaves any partial iMCUs at the bottom edge untouched.
+ * The other transforms are defined as combinations of these basic transforms
+ * and process edge blocks in a way that preserves the equivalence.
+ *
+ * The "trim" option causes untransformable partial iMCUs to be dropped;
+ * this is not strictly lossless, but it usually gives the best-looking
+ * result for odd-size images.  Note that when this option is active,
+ * the expected mathematical equivalences between the transforms may not hold.
+ * (For example, -rot 270 -trim trims only the bottom edge, but -rot 90 -trim
+ * followed by -rot 180 -trim trims both edges.)
+ *
+ * We also offer a "force to grayscale" option, which simply discards the
+ * chrominance channels of a YCbCr image.  This is lossless in the sense that
+ * the luminance channel is preserved exactly.  It's not the same kind of
+ * thing as the rotate/flip transformations, but it's convenient to handle it
+ * as part of this package, mainly because the transformation routines have to
+ * be aware of the option to know how many components to work on.
+ */
+
+typedef struct {
+  /* Options: set by caller */
+  JXFORM_CODE transform;	/* image transform operator */
+  boolean trim;			/* if TRUE, trim partial MCUs as needed */
+  boolean force_grayscale;	/* if TRUE, convert color image to grayscale */
+
+  /* Internal workspace: caller should not touch these */
+  int num_components;		/* # of components in workspace */
+  jvirt_barray_ptr * workspace_coef_arrays; /* workspace for transformations */
+} jpeg_transform_info;
+
+
+#if TRANSFORMS_SUPPORTED
+
+/* Request any required workspace */
+EXTERN(void) jtransform_request_workspace
+	JPP((j_decompress_ptr srcinfo, jpeg_transform_info *info));
+/* Adjust output image parameters */
+EXTERN(jvirt_barray_ptr *) jtransform_adjust_parameters
+	JPP((j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+	     jvirt_barray_ptr *src_coef_arrays,
+	     jpeg_transform_info *info));
+/* Execute the actual transformation, if any */
+EXTERN(void) jtransform_execute_transformation
+	JPP((j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+	     jvirt_barray_ptr *src_coef_arrays,
+	     jpeg_transform_info *info));
+
+#endif /* TRANSFORMS_SUPPORTED */
+
+
+/*
+ * Support for copying optional markers from source to destination file.
+ */
+
+typedef enum {
+	JCOPYOPT_NONE,		/* copy no optional markers */
+	JCOPYOPT_COMMENTS,	/* copy only comment (COM) markers */
+	JCOPYOPT_ALL		/* copy all optional markers */
+} JCOPY_OPTION;
+
+#define JCOPYOPT_DEFAULT  JCOPYOPT_COMMENTS	/* recommended default */
+
+/* Setup decompression object to save desired markers in memory */
+EXTERN(void) jcopy_markers_setup
+	JPP((j_decompress_ptr srcinfo, JCOPY_OPTION option));
+/* Copy markers saved in the given source object to the destination object */
+EXTERN(void) jcopy_markers_execute
+	JPP((j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+	     JCOPY_OPTION option));
diff --git a/libjpegtwrp/usage.doc b/libjpegtwrp/usage.doc
new file mode 100644
index 000000000..8c4970af0
--- /dev/null
+++ b/libjpegtwrp/usage.doc
@@ -0,0 +1,562 @@
+USAGE instructions for the Independent JPEG Group's JPEG software
+=================================================================
+
+This file describes usage of the JPEG conversion programs cjpeg and djpeg,
+as well as the utility programs jpegtran, rdjpgcom and wrjpgcom.  (See
+the other documentation files if you wish to use the JPEG library within
+your own programs.)
+
+If you are on a Unix machine you may prefer to read the Unix-style manual
+pages in files cjpeg.1, djpeg.1, jpegtran.1, rdjpgcom.1, wrjpgcom.1.
+
+
+INTRODUCTION
+
+These programs implement JPEG image compression and decompression.  JPEG
+(pronounced "jay-peg") is a standardized compression method for full-color
+and gray-scale images.  JPEG is designed to handle "real-world" scenes,
+for example scanned photographs.  Cartoons, line drawings, and other
+non-realistic images are not JPEG's strong suit; on that sort of material
+you may get poor image quality and/or little compression.
+
+JPEG is lossy, meaning that the output image is not necessarily identical to
+the input image.  Hence you should not use JPEG if you have to have identical
+output bits.  However, on typical real-world images, very good compression
+levels can be obtained with no visible change, and amazingly high compression
+is possible if you can tolerate a low-quality image.  You can trade off image
+quality against file size by adjusting the compressor's "quality" setting.
+
+
+GENERAL USAGE
+
+We provide two programs, cjpeg to compress an image file into JPEG format,
+and djpeg to decompress a JPEG file back into a conventional image format.
+
+On Unix-like systems, you say:
+	cjpeg [switches] [imagefile] >jpegfile
+or
+	djpeg [switches] [jpegfile]  >imagefile
+The programs read the specified input file, or standard input if none is
+named.  They always write to standard output (with trace/error messages to
+standard error).  These conventions are handy for piping images between
+programs.
+
+On most non-Unix systems, you say:
+	cjpeg [switches] imagefile jpegfile
+or
+	djpeg [switches] jpegfile  imagefile
+i.e., both the input and output files are named on the command line.  This
+style is a little more foolproof, and it loses no functionality if you don't
+have pipes.  (You can get this style on Unix too, if you prefer, by defining
+TWO_FILE_COMMANDLINE when you compile the programs; see install.doc.)
+
+You can also say:
+	cjpeg [switches] -outfile jpegfile  imagefile
+or
+	djpeg [switches] -outfile imagefile  jpegfile
+This syntax works on all systems, so it is useful for scripts.
+
+The currently supported image file formats are: PPM (PBMPLUS color format),
+PGM (PBMPLUS gray-scale format), BMP, Targa, and RLE (Utah Raster Toolkit
+format).  (RLE is supported only if the URT library is available.)
+cjpeg recognizes the input image format automatically, with the exception
+of some Targa-format files.  You have to tell djpeg which format to generate.
+
+JPEG files are in the defacto standard JFIF file format.  There are other,
+less widely used JPEG-based file formats, but we don't support them.
+
+All switch names may be abbreviated; for example, -grayscale may be written
+-gray or -gr.  Most of the "basic" switches can be abbreviated to as little as
+one letter.  Upper and lower case are equivalent (-BMP is the same as -bmp).
+British spellings are also accepted (e.g., -greyscale), though for brevity
+these are not mentioned below.
+
+
+CJPEG DETAILS
+
+The basic command line switches for cjpeg are:
+
+	-quality N	Scale quantization tables to adjust image quality.
+			Quality is 0 (worst) to 100 (best); default is 75.
+			(See below for more info.)
+
+	-grayscale	Create monochrome JPEG file from color input.
+			Be sure to use this switch when compressing a grayscale
+			BMP file, because cjpeg isn't bright enough to notice
+			whether a BMP file uses only shades of gray.  By
+			saying -grayscale, you'll get a smaller JPEG file that
+			takes less time to process.
+
+	-optimize	Perform optimization of entropy encoding parameters.
+			Without this, default encoding parameters are used.
+			-optimize usually makes the JPEG file a little smaller,
+			but cjpeg runs somewhat slower and needs much more
+			memory.  Image quality and speed of decompression are
+			unaffected by -optimize.
+
+	-progressive	Create progressive JPEG file (see below).
+
+	-targa		Input file is Targa format.  Targa files that contain
+			an "identification" field will not be automatically
+			recognized by cjpeg; for such files you must specify
+			-targa to make cjpeg treat the input as Targa format.
+			For most Targa files, you won't need this switch.
+
+The -quality switch lets you trade off compressed file size against quality of
+the reconstructed image: the higher the quality setting, the larger the JPEG
+file, and the closer the output image will be to the original input.  Normally
+you want to use the lowest quality setting (smallest file) that decompresses
+into something visually indistinguishable from the original image.  For this
+purpose the quality setting should be between 50 and 95; the default of 75 is
+often about right.  If you see defects at -quality 75, then go up 5 or 10
+counts at a time until you are happy with the output image.  (The optimal
+setting will vary from one image to another.)
+
+-quality 100 will generate a quantization table of all 1's, minimizing loss
+in the quantization step (but there is still information loss in subsampling,
+as well as roundoff error).  This setting is mainly of interest for
+experimental purposes.  Quality values above about 95 are NOT recommended for
+normal use; the compressed file size goes up dramatically for hardly any gain
+in output image quality.
+
+In the other direction, quality values below 50 will produce very small files
+of low image quality.  Settings around 5 to 10 might be useful in preparing an
+index of a large image library, for example.  Try -quality 2 (or so) for some
+amusing Cubist effects.  (Note: quality values below about 25 generate 2-byte
+quantization tables, which are considered optional in the JPEG standard.
+cjpeg emits a warning message when you give such a quality value, because some
+other JPEG programs may be unable to decode the resulting file.  Use -baseline
+if you need to ensure compatibility at low quality values.)
+
+The -progressive switch creates a "progressive JPEG" file.  In this type of
+JPEG file, the data is stored in multiple scans of increasing quality.  If the
+file is being transmitted over a slow communications link, the decoder can use
+the first scan to display a low-quality image very quickly, and can then
+improve the display with each subsequent scan.  The final image is exactly
+equivalent to a standard JPEG file of the same quality setting, and the total
+file size is about the same --- often a little smaller.  CAUTION: progressive
+JPEG is not yet widely implemented, so many decoders will be unable to view a
+progressive JPEG file at all.
+
+Switches for advanced users:
+
+	-dct int	Use integer DCT method (default).
+	-dct fast	Use fast integer DCT (less accurate).
+	-dct float	Use floating-point DCT method.
+			The float method is very slightly more accurate than
+			the int method, but is much slower unless your machine
+			has very fast floating-point hardware.  Also note that
+			results of the floating-point method may vary slightly
+			across machines, while the integer methods should give
+			the same results everywhere.  The fast integer method
+			is much less accurate than the other two.
+
+	-restart N	Emit a JPEG restart marker every N MCU rows, or every
+			N MCU blocks if "B" is attached to the number.
+			-restart 0 (the default) means no restart markers.
+
+	-smooth N	Smooth the input image to eliminate dithering noise.
+			N, ranging from 1 to 100, indicates the strength of
+			smoothing.  0 (the default) means no smoothing.
+
+	-maxmemory N	Set limit for amount of memory to use in processing
+			large images.  Value is in thousands of bytes, or
+			millions of bytes if "M" is attached to the number.
+			For example, -max 4m selects 4000000 bytes.  If more
+			space is needed, temporary files will be used.
+
+	-verbose	Enable debug printout.  More -v's give more printout.
+	or  -debug	Also, version information is printed at startup.
+
+The -restart option inserts extra markers that allow a JPEG decoder to
+resynchronize after a transmission error.  Without restart markers, any damage
+to a compressed file will usually ruin the image from the point of the error
+to the end of the image; with restart markers, the damage is usually confined
+to the portion of the image up to the next restart marker.  Of course, the
+restart markers occupy extra space.  We recommend -restart 1 for images that
+will be transmitted across unreliable networks such as Usenet.
+
+The -smooth option filters the input to eliminate fine-scale noise.  This is
+often useful when converting dithered images to JPEG: a moderate smoothing
+factor of 10 to 50 gets rid of dithering patterns in the input file, resulting
+in a smaller JPEG file and a better-looking image.  Too large a smoothing
+factor will visibly blur the image, however.
+
+Switches for wizards:
+
+	-baseline	Force baseline-compatible quantization tables to be
+			generated.  This clamps quantization values to 8 bits
+			even at low quality settings.  (This switch is poorly
+			named, since it does not ensure that the output is
+			actually baseline JPEG.  For example, you can use
+			-baseline and -progressive together.)
+
+	-qtables file	Use the quantization tables given in the specified
+			text file.
+
+	-qslots N[,...] Select which quantization table to use for each color
+			component.
+
+	-sample HxV[,...]  Set JPEG sampling factors for each color component.
+
+	-scans file	Use the scan script given in the specified text file.
+
+The "wizard" switches are intended for experimentation with JPEG.  If you
+don't know what you are doing, DON'T USE THEM.  These switches are documented
+further in the file wizard.doc.
+
+
+DJPEG DETAILS
+
+The basic command line switches for djpeg are:
+
+	-colors N	Reduce image to at most N colors.  This reduces the
+	or -quantize N	number of colors used in the output image, so that it
+			can be displayed on a colormapped display or stored in
+			a colormapped file format.  For example, if you have
+			an 8-bit display, you'd need to reduce to 256 or fewer
+			colors.  (-colors is the recommended name, -quantize
+			is provided only for backwards compatibility.)
+
+	-fast		Select recommended processing options for fast, low
+			quality output.  (The default options are chosen for
+			highest quality output.)  Currently, this is equivalent
+			to "-dct fast -nosmooth -onepass -dither ordered".
+
+	-grayscale	Force gray-scale output even if JPEG file is color.
+			Useful for viewing on monochrome displays; also,
+			djpeg runs noticeably faster in this mode.
+
+	-scale M/N	Scale the output image by a factor M/N.  Currently
+			the scale factor must be 1/1, 1/2, 1/4, or 1/8.
+			Scaling is handy if the image is larger than your
+			screen; also, djpeg runs much faster when scaling
+			down the output.
+
+	-bmp		Select BMP output format (Windows flavor).  8-bit
+			colormapped format is emitted if -colors or -grayscale
+			is specified, or if the JPEG file is gray-scale;
+			otherwise, 24-bit full-color format is emitted.
+
+	-gif		Select GIF output format.  Since GIF does not support
+			more than 256 colors, -colors 256 is assumed (unless
+			you specify a smaller number of colors).  If you
+			specify -fast, the default number of colors is 216.
+
+	-os2		Select BMP output format (OS/2 1.x flavor).  8-bit
+			colormapped format is emitted if -colors or -grayscale
+			is specified, or if the JPEG file is gray-scale;
+			otherwise, 24-bit full-color format is emitted.
+
+	-pnm		Select PBMPLUS (PPM/PGM) output format (this is the
+			default format).  PGM is emitted if the JPEG file is
+			gray-scale or if -grayscale is specified; otherwise
+			PPM is emitted.
+
+	-rle		Select RLE output format.  (Requires URT library.)
+
+	-targa		Select Targa output format.  Gray-scale format is
+			emitted if the JPEG file is gray-scale or if
+			-grayscale is specified; otherwise, colormapped format
+			is emitted if -colors is specified; otherwise, 24-bit
+			full-color format is emitted.
+
+Switches for advanced users:
+
+	-dct int	Use integer DCT method (default).
+	-dct fast	Use fast integer DCT (less accurate).
+	-dct float	Use floating-point DCT method.
+			The float method is very slightly more accurate than
+			the int method, but is much slower unless your machine
+			has very fast floating-point hardware.  Also note that
+			results of the floating-point method may vary slightly
+			across machines, while the integer methods should give
+			the same results everywhere.  The fast integer method
+			is much less accurate than the other two.
+
+	-dither fs	Use Floyd-Steinberg dithering in color quantization.
+	-dither ordered	Use ordered dithering in color quantization.
+	-dither none	Do not use dithering in color quantization.
+			By default, Floyd-Steinberg dithering is applied when
+			quantizing colors; this is slow but usually produces
+			the best results.  Ordered dither is a compromise
+			between speed and quality; no dithering is fast but
+			usually looks awful.  Note that these switches have
+			no effect unless color quantization is being done.
+			Ordered dither is only available in -onepass mode.
+
+	-map FILE	Quantize to the colors used in the specified image
+			file.  This is useful for producing multiple files
+			with identical color maps, or for forcing a predefined
+			set of colors to be used.  The FILE must be a GIF
+			or PPM file.  This option overrides -colors and
+			-onepass.
+
+	-nosmooth	Use a faster, lower-quality upsampling routine.
+
+	-onepass	Use one-pass instead of two-pass color quantization.
+			The one-pass method is faster and needs less memory,
+			but it produces a lower-quality image.  -onepass is
+			ignored unless you also say -colors N.  Also,
+			the one-pass method is always used for gray-scale
+			output (the two-pass method is no improvement then).
+
+	-maxmemory N	Set limit for amount of memory to use in processing
+			large images.  Value is in thousands of bytes, or
+			millions of bytes if "M" is attached to the number.
+			For example, -max 4m selects 4000000 bytes.  If more
+			space is needed, temporary files will be used.
+
+	-verbose	Enable debug printout.  More -v's give more printout.
+	or  -debug	Also, version information is printed at startup.
+
+
+HINTS FOR CJPEG
+
+Color GIF files are not the ideal input for JPEG; JPEG is really intended for
+compressing full-color (24-bit) images.  In particular, don't try to convert
+cartoons, line drawings, and other images that have only a few distinct
+colors.  GIF works great on these, JPEG does not.  If you want to convert a
+GIF to JPEG, you should experiment with cjpeg's -quality and -smooth options
+to get a satisfactory conversion.  -smooth 10 or so is often helpful.
+
+Avoid running an image through a series of JPEG compression/decompression
+cycles.  Image quality loss will accumulate; after ten or so cycles the image
+may be noticeably worse than it was after one cycle.  It's best to use a
+lossless format while manipulating an image, then convert to JPEG format when
+you are ready to file the image away.
+
+The -optimize option to cjpeg is worth using when you are making a "final"
+version for posting or archiving.  It's also a win when you are using low
+quality settings to make very small JPEG files; the percentage improvement
+is often a lot more than it is on larger files.  (At present, -optimize
+mode is always selected when generating progressive JPEG files.)
+
+GIF input files are no longer supported, to avoid the Unisys LZW patent.
+Use a Unisys-licensed program if you need to read a GIF file.  (Conversion
+of GIF files to JPEG is usually a bad idea anyway.)
+
+
+HINTS FOR DJPEG
+
+To get a quick preview of an image, use the -grayscale and/or -scale switches.
+"-grayscale -scale 1/8" is the fastest case.
+
+Several options are available that trade off image quality to gain speed.
+"-fast" turns on the recommended settings.
+
+"-dct fast" and/or "-nosmooth" gain speed at a small sacrifice in quality.
+When producing a color-quantized image, "-onepass -dither ordered" is fast but
+much lower quality than the default behavior.  "-dither none" may give
+acceptable results in two-pass mode, but is seldom tolerable in one-pass mode.
+
+If you are fortunate enough to have very fast floating point hardware,
+"-dct float" may be even faster than "-dct fast".  But on most machines
+"-dct float" is slower than "-dct int"; in this case it is not worth using,
+because its theoretical accuracy advantage is too small to be significant
+in practice.
+
+Two-pass color quantization requires a good deal of memory; on MS-DOS machines
+it may run out of memory even with -maxmemory 0.  In that case you can still
+decompress, with some loss of image quality, by specifying -onepass for
+one-pass quantization.
+
+To avoid the Unisys LZW patent, djpeg produces uncompressed GIF files.  These
+are larger than they should be, but are readable by standard GIF decoders.
+
+
+HINTS FOR BOTH PROGRAMS
+
+If more space is needed than will fit in the available main memory (as
+determined by -maxmemory), temporary files will be used.  (MS-DOS versions
+will try to get extended or expanded memory first.)  The temporary files are
+often rather large: in typical cases they occupy three bytes per pixel, for
+example 3*800*600 = 1.44Mb for an 800x600 image.  If you don't have enough
+free disk space, leave out -progressive and -optimize (for cjpeg) or specify
+-onepass (for djpeg).
+
+On MS-DOS, the temporary files are created in the directory named by the TMP
+or TEMP environment variable, or in the current directory if neither of those
+exist.  Amiga implementations put the temp files in the directory named by
+JPEGTMP:, so be sure to assign JPEGTMP: to a disk partition with adequate free
+space.
+
+The default memory usage limit (-maxmemory) is set when the software is
+compiled.  If you get an "insufficient memory" error, try specifying a smaller
+-maxmemory value, even -maxmemory 0 to use the absolute minimum space.  You
+may want to recompile with a smaller default value if this happens often.
+
+On machines that have "environment" variables, you can define the environment
+variable JPEGMEM to set the default memory limit.  The value is specified as
+described for the -maxmemory switch.  JPEGMEM overrides the default value
+specified when the program was compiled, and itself is overridden by an
+explicit -maxmemory switch.
+
+On MS-DOS machines, -maxmemory is the amount of main (conventional) memory to
+use.  (Extended or expanded memory is also used if available.)  Most
+DOS-specific versions of this software do their own memory space estimation
+and do not need you to specify -maxmemory.
+
+
+JPEGTRAN
+
+jpegtran performs various useful transformations of JPEG files.
+It can translate the coded representation from one variant of JPEG to another,
+for example from baseline JPEG to progressive JPEG or vice versa.  It can also
+perform some rearrangements of the image data, for example turning an image
+from landscape to portrait format by rotation.
+
+jpegtran works by rearranging the compressed data (DCT coefficients), without
+ever fully decoding the image.  Therefore, its transformations are lossless:
+there is no image degradation at all, which would not be true if you used
+djpeg followed by cjpeg to accomplish the same conversion.  But by the same
+token, jpegtran cannot perform lossy operations such as changing the image
+quality.
+
+jpegtran uses a command line syntax similar to cjpeg or djpeg.
+On Unix-like systems, you say:
+	jpegtran [switches] [inputfile] >outputfile
+On most non-Unix systems, you say:
+	jpegtran [switches] inputfile outputfile
+where both the input and output files are JPEG files.
+
+To specify the coded JPEG representation used in the output file,
+jpegtran accepts a subset of the switches recognized by cjpeg:
+	-optimize	Perform optimization of entropy encoding parameters.
+	-progressive	Create progressive JPEG file.
+	-restart N	Emit a JPEG restart marker every N MCU rows, or every
+			N MCU blocks if "B" is attached to the number.
+	-scans file	Use the scan script given in the specified text file.
+See the previous discussion of cjpeg for more details about these switches.
+If you specify none of these switches, you get a plain baseline-JPEG output
+file.  The quality setting and so forth are determined by the input file.
+
+The image can be losslessly transformed by giving one of these switches:
+	-flip horizontal	Mirror image horizontally (left-right).
+	-flip vertical		Mirror image vertically (top-bottom).
+	-rotate 90		Rotate image 90 degrees clockwise.
+	-rotate 180		Rotate image 180 degrees.
+	-rotate 270		Rotate image 270 degrees clockwise (or 90 ccw).
+	-transpose		Transpose image (across UL-to-LR axis).
+	-transverse		Transverse transpose (across UR-to-LL axis).
+
+The transpose transformation has no restrictions regarding image dimensions.
+The other transformations operate rather oddly if the image dimensions are not
+a multiple of the iMCU size (usually 8 or 16 pixels), because they can only
+transform complete blocks of DCT coefficient data in the desired way.
+
+jpegtran's default behavior when transforming an odd-size image is designed
+to preserve exact reversibility and mathematical consistency of the
+transformation set.  As stated, transpose is able to flip the entire image
+area.  Horizontal mirroring leaves any partial iMCU column at the right edge
+untouched, but is able to flip all rows of the image.  Similarly, vertical
+mirroring leaves any partial iMCU row at the bottom edge untouched, but is
+able to flip all columns.  The other transforms can be built up as sequences
+of transpose and flip operations; for consistency, their actions on edge
+pixels are defined to be the same as the end result of the corresponding
+transpose-and-flip sequence.
+
+For practical use, you may prefer to discard any untransformable edge pixels
+rather than having a strange-looking strip along the right and/or bottom edges
+of a transformed image.  To do this, add the -trim switch:
+	-trim		Drop non-transformable edge blocks.
+Obviously, a transformation with -trim is not reversible, so strictly speaking
+jpegtran with this switch is not lossless.  Also, the expected mathematical
+equivalences between the transformations no longer hold.  For example,
+"-rot 270 -trim" trims only the bottom edge, but "-rot 90 -trim" followed by
+"-rot 180 -trim" trims both edges.
+
+Another not-strictly-lossless transformation switch is:
+	-grayscale	Force grayscale output.
+This option discards the chrominance channels if the input image is YCbCr
+(ie, a standard color JPEG), resulting in a grayscale JPEG file.  The
+luminance channel is preserved exactly, so this is a better method of reducing
+to grayscale than decompression, conversion, and recompression.  This switch
+is particularly handy for fixing a monochrome picture that was mistakenly
+encoded as a color JPEG.  (In such a case, the space savings from getting rid
+of the near-empty chroma channels won't be large; but the decoding time for
+a grayscale JPEG is substantially less than that for a color JPEG.)
+
+jpegtran also recognizes these switches that control what to do with "extra"
+markers, such as comment blocks:
+	-copy none	Copy no extra markers from source file.  This setting
+			suppresses all comments and other excess baggage
+			present in the source file.
+	-copy comments	Copy only comment markers.  This setting copies
+			comments from the source file, but discards
+			any other inessential data. 
+	-copy all	Copy all extra markers.  This setting preserves
+			miscellaneous markers found in the source file, such
+			as JFIF thumbnails and Photoshop settings.  In some
+			files these extra markers can be sizable.
+The default behavior is -copy comments.  (Note: in IJG releases v6 and v6a,
+jpegtran always did the equivalent of -copy none.)
+
+Additional switches recognized by jpegtran are:
+	-outfile filename
+	-maxmemory N
+	-verbose
+	-debug
+These work the same as in cjpeg or djpeg.
+
+
+THE COMMENT UTILITIES
+
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+
+We provide two utility programs to display COM block contents and add COM
+blocks to a JPEG file.
+
+rdjpgcom searches a JPEG file and prints the contents of any COM blocks on
+standard output.  The command line syntax is
+	rdjpgcom [-verbose] [inputfilename]
+The switch "-verbose" (or just "-v") causes rdjpgcom to also display the JPEG
+image dimensions.  If you omit the input file name from the command line,
+the JPEG file is read from standard input.  (This may not work on some
+operating systems, if binary data can't be read from stdin.)
+
+wrjpgcom adds a COM block, containing text you provide, to a JPEG file.
+Ordinarily, the COM block is added after any existing COM blocks, but you
+can delete the old COM blocks if you wish.  wrjpgcom produces a new JPEG
+file; it does not modify the input file.  DO NOT try to overwrite the input
+file by directing wrjpgcom's output back into it; on most systems this will
+just destroy your file.
+
+The command line syntax for wrjpgcom is similar to cjpeg's.  On Unix-like
+systems, it is
+	wrjpgcom [switches] [inputfilename]
+The output file is written to standard output.  The input file comes from
+the named file, or from standard input if no input file is named.
+
+On most non-Unix systems, the syntax is
+	wrjpgcom [switches] inputfilename outputfilename
+where both input and output file names must be given explicitly.
+
+wrjpgcom understands three switches:
+	-replace		 Delete any existing COM blocks from the file.
+	-comment "Comment text"	 Supply new COM text on command line.
+	-cfile name		 Read text for new COM block from named file.
+(Switch names can be abbreviated.)  If you have only one line of comment text
+to add, you can provide it on the command line with -comment.  The comment
+text must be surrounded with quotes so that it is treated as a single
+argument.  Longer comments can be read from a text file.
+
+If you give neither -comment nor -cfile, then wrjpgcom will read the comment
+text from standard input.  (In this case an input image file name MUST be
+supplied, so that the source JPEG file comes from somewhere else.)  You can
+enter multiple lines, up to 64KB worth.  Type an end-of-file indicator
+(usually control-D or control-Z) to terminate the comment text entry.
+
+wrjpgcom will not add a COM block if the provided comment string is empty.
+Therefore -replace -comment "" can be used to delete all COM blocks from a
+file.
+
+These utility programs do not depend on the IJG JPEG library.  In
+particular, the source code for rdjpgcom is intended as an illustration of
+the minimum amount of code required to parse a JPEG file header correctly.
diff --git a/libjpegtwrp/wizard.doc b/libjpegtwrp/wizard.doc
new file mode 100644
index 000000000..54170b227
--- /dev/null
+++ b/libjpegtwrp/wizard.doc
@@ -0,0 +1,211 @@
+Advanced usage instructions for the Independent JPEG Group's JPEG software
+==========================================================================
+
+This file describes cjpeg's "switches for wizards".
+
+The "wizard" switches are intended for experimentation with JPEG by persons
+who are reasonably knowledgeable about the JPEG standard.  If you don't know
+what you are doing, DON'T USE THESE SWITCHES.  You'll likely produce files
+with worse image quality and/or poorer compression than you'd get from the
+default settings.  Furthermore, these switches must be used with caution
+when making files intended for general use, because not all JPEG decoders
+will support unusual JPEG parameter settings.
+
+
+Quantization Table Adjustment
+-----------------------------
+
+Ordinarily, cjpeg starts with a default set of tables (the same ones given
+as examples in the JPEG standard) and scales them up or down according to
+the -quality setting.  The details of the scaling algorithm can be found in
+jcparam.c.  At very low quality settings, some quantization table entries
+can get scaled up to values exceeding 255.  Although 2-byte quantization
+values are supported by the IJG software, this feature is not in baseline
+JPEG and is not supported by all implementations.  If you need to ensure
+wide compatibility of low-quality files, you can constrain the scaled
+quantization values to no more than 255 by giving the -baseline switch.
+Note that use of -baseline will result in poorer quality for the same file
+size, since more bits than necessary are expended on higher AC coefficients.
+
+You can substitute a different set of quantization values by using the
+-qtables switch:
+
+	-qtables file	Use the quantization tables given in the named file.
+
+The specified file should be a text file containing decimal quantization
+values.  The file should contain one to four tables, each of 64 elements.
+The tables are implicitly numbered 0,1,etc. in order of appearance.  Table
+entries appear in normal array order (NOT in the zigzag order in which they
+will be stored in the JPEG file).
+
+Quantization table files are free format, in that arbitrary whitespace can
+appear between numbers.  Also, comments can be included: a comment starts
+with '#' and extends to the end of the line.  Here is an example file that
+duplicates the default quantization tables:
+
+	# Quantization tables given in JPEG spec, section K.1
+
+	# This is table 0 (the luminance table):
+	  16  11  10  16  24  40  51  61
+	  12  12  14  19  26  58  60  55
+	  14  13  16  24  40  57  69  56
+	  14  17  22  29  51  87  80  62
+	  18  22  37  56  68 109 103  77
+	  24  35  55  64  81 104 113  92
+	  49  64  78  87 103 121 120 101
+	  72  92  95  98 112 100 103  99
+
+	# This is table 1 (the chrominance table):
+	  17  18  24  47  99  99  99  99
+	  18  21  26  66  99  99  99  99
+	  24  26  56  99  99  99  99  99
+	  47  66  99  99  99  99  99  99
+	  99  99  99  99  99  99  99  99
+	  99  99  99  99  99  99  99  99
+	  99  99  99  99  99  99  99  99
+	  99  99  99  99  99  99  99  99
+
+If the -qtables switch is used without -quality, then the specified tables
+are used exactly as-is.  If both -qtables and -quality are used, then the
+tables taken from the file are scaled in the same fashion that the default
+tables would be scaled for that quality setting.  If -baseline appears, then
+the quantization values are constrained to the range 1-255.
+
+By default, cjpeg will use quantization table 0 for luminance components and
+table 1 for chrominance components.  To override this choice, use the -qslots
+switch:
+
+	-qslots N[,...]		Select which quantization table to use for
+				each color component.
+
+The -qslots switch specifies a quantization table number for each color
+component, in the order in which the components appear in the JPEG SOF marker.
+For example, to create a separate table for each of Y,Cb,Cr, you could
+provide a -qtables file that defines three quantization tables and say
+"-qslots 0,1,2".  If -qslots gives fewer table numbers than there are color
+components, then the last table number is repeated as necessary.
+
+
+Sampling Factor Adjustment
+--------------------------
+
+By default, cjpeg uses 2:1 horizontal and vertical downsampling when
+compressing YCbCr data, and no downsampling for all other color spaces.
+You can override this default with the -sample switch:
+
+	-sample HxV[,...]	Set JPEG sampling factors for each color
+				component.
+
+The -sample switch specifies the JPEG sampling factors for each color
+component, in the order in which they appear in the JPEG SOF marker.
+If you specify fewer HxV pairs than there are components, the remaining
+components are set to 1x1 sampling.  For example, the default YCbCr setting
+is equivalent to "-sample 2x2,1x1,1x1", which can be abbreviated to
+"-sample 2x2".
+
+There are still some JPEG decoders in existence that support only 2x1
+sampling (also called 4:2:2 sampling).  Compatibility with such decoders can
+be achieved by specifying "-sample 2x1".  This is not recommended unless
+really necessary, since it increases file size and encoding/decoding time
+with very little quality gain.
+
+
+Multiple Scan / Progression Control
+-----------------------------------
+
+By default, cjpeg emits a single-scan sequential JPEG file.  The
+-progressive switch generates a progressive JPEG file using a default series
+of progression parameters.  You can create multiple-scan sequential JPEG
+files or progressive JPEG files with custom progression parameters by using
+the -scans switch:
+
+	-scans file	Use the scan sequence given in the named file.
+
+The specified file should be a text file containing a "scan script".
+The script specifies the contents and ordering of the scans to be emitted.
+Each entry in the script defines one scan.  A scan definition specifies
+the components to be included in the scan, and for progressive JPEG it also
+specifies the progression parameters Ss,Se,Ah,Al for the scan.  Scan
+definitions are separated by semicolons (';').  A semicolon after the last
+scan definition is optional.
+
+Each scan definition contains one to four component indexes, optionally
+followed by a colon (':') and the four progressive-JPEG parameters.  The
+component indexes denote which color component(s) are to be transmitted in
+the scan.  Components are numbered in the order in which they appear in the
+JPEG SOF marker, with the first component being numbered 0.  (Note that these
+indexes are not the "component ID" codes assigned to the components, just
+positional indexes.)
+
+The progression parameters for each scan are:
+	Ss	Zigzag index of first coefficient included in scan
+	Se	Zigzag index of last coefficient included in scan
+	Ah	Zero for first scan of a coefficient, else Al of prior scan
+	Al	Successive approximation low bit position for scan
+If the progression parameters are omitted, the values 0,63,0,0 are used,
+producing a sequential JPEG file.  cjpeg automatically determines whether
+the script represents a progressive or sequential file, by observing whether
+Ss and Se values other than 0 and 63 appear.  (The -progressive switch is
+not needed to specify this; in fact, it is ignored when -scans appears.)
+The scan script must meet the JPEG restrictions on progression sequences.
+(cjpeg checks that the spec's requirements are obeyed.)
+
+Scan script files are free format, in that arbitrary whitespace can appear
+between numbers and around punctuation.  Also, comments can be included: a
+comment starts with '#' and extends to the end of the line.  For additional
+legibility, commas or dashes can be placed between values.  (Actually, any
+single punctuation character other than ':' or ';' can be inserted.)  For
+example, the following two scan definitions are equivalent:
+	0 1 2: 0 63 0 0;
+	0,1,2 : 0-63, 0,0 ;
+
+Here is an example of a scan script that generates a partially interleaved
+sequential JPEG file:
+
+	0;			# Y only in first scan
+	1 2;			# Cb and Cr in second scan
+
+Here is an example of a progressive scan script using only spectral selection
+(no successive approximation):
+
+	# Interleaved DC scan for Y,Cb,Cr:
+	0,1,2: 0-0,   0, 0 ;
+	# AC scans:
+	0:     1-2,   0, 0 ;	# First two Y AC coefficients
+	0:     3-5,   0, 0 ;	# Three more
+	1:     1-63,  0, 0 ;	# All AC coefficients for Cb
+	2:     1-63,  0, 0 ;	# All AC coefficients for Cr
+	0:     6-9,   0, 0 ;	# More Y coefficients
+	0:     10-63, 0, 0 ;	# Remaining Y coefficients
+
+Here is an example of a successive-approximation script.  This is equivalent
+to the default script used by "cjpeg -progressive" for YCbCr images:
+
+	# Initial DC scan for Y,Cb,Cr (lowest bit not sent)
+	0,1,2: 0-0,   0, 1 ;
+	# First AC scan: send first 5 Y AC coefficients, minus 2 lowest bits:
+	0:     1-5,   0, 2 ;
+	# Send all Cr,Cb AC coefficients, minus lowest bit:
+	# (chroma data is usually too small to be worth subdividing further;
+	#  but note we send Cr first since eye is least sensitive to Cb)
+	2:     1-63,  0, 1 ;
+	1:     1-63,  0, 1 ;
+	# Send remaining Y AC coefficients, minus 2 lowest bits:
+	0:     6-63,  0, 2 ;
+	# Send next-to-lowest bit of all Y AC coefficients:
+	0:     1-63,  2, 1 ;
+	# At this point we've sent all but the lowest bit of all coefficients.
+	# Send lowest bit of DC coefficients
+	0,1,2: 0-0,   1, 0 ;
+	# Send lowest bit of AC coefficients
+	2:     1-63,  1, 0 ;
+	1:     1-63,  1, 0 ;
+	# Y AC lowest bit scan is last; it's usually the largest scan
+	0:     1-63,  1, 0 ;
+
+It may be worth pointing out that this script is tuned for quality settings
+of around 50 to 75.  For lower quality settings, you'd probably want to use
+a script with fewer stages of successive approximation (otherwise the
+initial scans will be really bad).  For higher quality settings, you might
+want to use more stages of successive approximation (so that the initial
+scans are not too large).
diff --git a/libjpegtwrp/wrbmp.c b/libjpegtwrp/wrbmp.c
new file mode 100644
index 000000000..3283b0f15
--- /dev/null
+++ b/libjpegtwrp/wrbmp.c
@@ -0,0 +1,442 @@
+/*
+ * wrbmp.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to write output images in Microsoft "BMP"
+ * format (MS Windows 3.x and OS/2 1.x flavors).
+ * Either 8-bit colormapped or 24-bit full-color format can be written.
+ * No compression is supported.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ *
+ * This code contributed by James Arthur Boucher.
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+
+#ifdef BMP_SUPPORTED
+
+
+/*
+ * To support 12-bit JPEG data, we'd have to scale output down to 8 bits.
+ * This is not yet implemented.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+/*
+ * Since BMP stores scanlines bottom-to-top, we have to invert the image
+ * from JPEG's top-to-bottom order.  To do this, we save the outgoing data
+ * in a virtual array during put_pixel_row calls, then actually emit the
+ * BMP file during finish_output.  The virtual array contains one JSAMPLE per
+ * pixel if the output is grayscale or colormapped, three if it is full color.
+ */
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub;	/* public fields */
+
+  boolean is_os2;		/* saves the OS2 format request flag */
+
+  jvirt_sarray_ptr whole_image;	/* needed to reverse row order */
+  JDIMENSION data_width;	/* JSAMPLEs per row */
+  JDIMENSION row_width;		/* physical width of one row in the BMP file */
+  int pad_bytes;		/* number of padding bytes needed per row */
+  JDIMENSION cur_output_row;	/* next row# to write to virtual array */
+} bmp_dest_struct;
+
+typedef bmp_dest_struct * bmp_dest_ptr;
+
+
+/* Forward declarations */
+LOCAL(void) write_colormap
+	JPP((j_decompress_ptr cinfo, bmp_dest_ptr dest,
+	     int map_colors, int map_entry_size));
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ */
+
+METHODDEF(void)
+put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+		JDIMENSION rows_supplied)
+/* This version is for writing 24-bit pixels */
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr) dinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  int pad;
+
+  /* Access next row in virtual array */
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, dest->whole_image,
+     dest->cur_output_row, (JDIMENSION) 1, TRUE);
+  dest->cur_output_row++;
+
+  /* Transfer data.  Note destination values must be in BGR order
+   * (even though Microsoft's own documents say the opposite).
+   */
+  inptr = dest->pub.buffer[0];
+  outptr = image_ptr[0];
+  for (col = cinfo->output_width; col > 0; col--) {
+    outptr[2] = *inptr++;	/* can omit GETJSAMPLE() safely */
+    outptr[1] = *inptr++;
+    outptr[0] = *inptr++;
+    outptr += 3;
+  }
+
+  /* Zero out the pad bytes. */
+  pad = dest->pad_bytes;
+  while (--pad >= 0)
+    *outptr++ = 0;
+}
+
+METHODDEF(void)
+put_gray_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+	       JDIMENSION rows_supplied)
+/* This version is for grayscale OR quantized color output */
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr) dinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  int pad;
+
+  /* Access next row in virtual array */
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, dest->whole_image,
+     dest->cur_output_row, (JDIMENSION) 1, TRUE);
+  dest->cur_output_row++;
+
+  /* Transfer data. */
+  inptr = dest->pub.buffer[0];
+  outptr = image_ptr[0];
+  for (col = cinfo->output_width; col > 0; col--) {
+    *outptr++ = *inptr++;	/* can omit GETJSAMPLE() safely */
+  }
+
+  /* Zero out the pad bytes. */
+  pad = dest->pad_bytes;
+  while (--pad >= 0)
+    *outptr++ = 0;
+}
+
+
+/*
+ * Startup: normally writes the file header.
+ * In this module we may as well postpone everything until finish_output.
+ */
+
+METHODDEF(void)
+start_output_bmp (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  /* no work here */
+}
+
+
+/*
+ * Finish up at the end of the file.
+ *
+ * Here is where we really output the BMP file.
+ *
+ * First, routines to write the Windows and OS/2 variants of the file header.
+ */
+
+LOCAL(void)
+write_bmp_header (j_decompress_ptr cinfo, bmp_dest_ptr dest)
+/* Write a Windows-style BMP file header, including colormap if needed */
+{
+  char bmpfileheader[14];
+  char bmpinfoheader[40];
+#define PUT_2B(array,offset,value)  \
+	(array[offset] = (char) ((value) & 0xFF), \
+	 array[offset+1] = (char) (((value) >> 8) & 0xFF))
+#define PUT_4B(array,offset,value)  \
+	(array[offset] = (char) ((value) & 0xFF), \
+	 array[offset+1] = (char) (((value) >> 8) & 0xFF), \
+	 array[offset+2] = (char) (((value) >> 16) & 0xFF), \
+	 array[offset+3] = (char) (((value) >> 24) & 0xFF))
+  INT32 headersize, bfSize;
+  int bits_per_pixel, cmap_entries;
+
+  /* Compute colormap size and total file size */
+  if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors) {
+      /* Colormapped RGB */
+      bits_per_pixel = 8;
+      cmap_entries = 256;
+    } else {
+      /* Unquantized, full color RGB */
+      bits_per_pixel = 24;
+      cmap_entries = 0;
+    }
+  } else {
+    /* Grayscale output.  We need to fake a 256-entry colormap. */
+    bits_per_pixel = 8;
+    cmap_entries = 256;
+  }
+  /* File size */
+  headersize = 14 + 40 + cmap_entries * 4; /* Header and colormap */
+  bfSize = headersize + (INT32) dest->row_width * (INT32) cinfo->output_height;
+  
+  /* Set unused fields of header to 0 */
+  MEMZERO(bmpfileheader, SIZEOF(bmpfileheader));
+  MEMZERO(bmpinfoheader, SIZEOF(bmpinfoheader));
+
+  /* Fill the file header */
+  bmpfileheader[0] = 0x42;	/* first 2 bytes are ASCII 'B', 'M' */
+  bmpfileheader[1] = 0x4D;
+  PUT_4B(bmpfileheader, 2, bfSize); /* bfSize */
+  /* we leave bfReserved1 & bfReserved2 = 0 */
+  PUT_4B(bmpfileheader, 10, headersize); /* bfOffBits */
+
+  /* Fill the info header (Microsoft calls this a BITMAPINFOHEADER) */
+  PUT_2B(bmpinfoheader, 0, 40);	/* biSize */
+  PUT_4B(bmpinfoheader, 4, cinfo->output_width); /* biWidth */
+  PUT_4B(bmpinfoheader, 8, cinfo->output_height); /* biHeight */
+  PUT_2B(bmpinfoheader, 12, 1);	/* biPlanes - must be 1 */
+  PUT_2B(bmpinfoheader, 14, bits_per_pixel); /* biBitCount */
+  /* we leave biCompression = 0, for none */
+  /* we leave biSizeImage = 0; this is correct for uncompressed data */
+  if (cinfo->density_unit == 2) { /* if have density in dots/cm, then */
+    PUT_4B(bmpinfoheader, 24, (INT32) (cinfo->X_density*100)); /* XPels/M */
+    PUT_4B(bmpinfoheader, 28, (INT32) (cinfo->Y_density*100)); /* XPels/M */
+  }
+  PUT_2B(bmpinfoheader, 32, cmap_entries); /* biClrUsed */
+  /* we leave biClrImportant = 0 */
+
+  if (JFWRITE(dest->pub.output_file, bmpfileheader, 14) != (size_t) 14)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+  if (JFWRITE(dest->pub.output_file, bmpinfoheader, 40) != (size_t) 40)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  if (cmap_entries > 0)
+    write_colormap(cinfo, dest, cmap_entries, 4);
+}
+
+
+LOCAL(void)
+write_os2_header (j_decompress_ptr cinfo, bmp_dest_ptr dest)
+/* Write an OS2-style BMP file header, including colormap if needed */
+{
+  char bmpfileheader[14];
+  char bmpcoreheader[12];
+  INT32 headersize, bfSize;
+  int bits_per_pixel, cmap_entries;
+
+  /* Compute colormap size and total file size */
+  if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors) {
+      /* Colormapped RGB */
+      bits_per_pixel = 8;
+      cmap_entries = 256;
+    } else {
+      /* Unquantized, full color RGB */
+      bits_per_pixel = 24;
+      cmap_entries = 0;
+    }
+  } else {
+    /* Grayscale output.  We need to fake a 256-entry colormap. */
+    bits_per_pixel = 8;
+    cmap_entries = 256;
+  }
+  /* File size */
+  headersize = 14 + 12 + cmap_entries * 3; /* Header and colormap */
+  bfSize = headersize + (INT32) dest->row_width * (INT32) cinfo->output_height;
+  
+  /* Set unused fields of header to 0 */
+  MEMZERO(bmpfileheader, SIZEOF(bmpfileheader));
+  MEMZERO(bmpcoreheader, SIZEOF(bmpcoreheader));
+
+  /* Fill the file header */
+  bmpfileheader[0] = 0x42;	/* first 2 bytes are ASCII 'B', 'M' */
+  bmpfileheader[1] = 0x4D;
+  PUT_4B(bmpfileheader, 2, bfSize); /* bfSize */
+  /* we leave bfReserved1 & bfReserved2 = 0 */
+  PUT_4B(bmpfileheader, 10, headersize); /* bfOffBits */
+
+  /* Fill the info header (Microsoft calls this a BITMAPCOREHEADER) */
+  PUT_2B(bmpcoreheader, 0, 12);	/* bcSize */
+  PUT_2B(bmpcoreheader, 4, cinfo->output_width); /* bcWidth */
+  PUT_2B(bmpcoreheader, 6, cinfo->output_height); /* bcHeight */
+  PUT_2B(bmpcoreheader, 8, 1);	/* bcPlanes - must be 1 */
+  PUT_2B(bmpcoreheader, 10, bits_per_pixel); /* bcBitCount */
+
+  if (JFWRITE(dest->pub.output_file, bmpfileheader, 14) != (size_t) 14)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+  if (JFWRITE(dest->pub.output_file, bmpcoreheader, 12) != (size_t) 12)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  if (cmap_entries > 0)
+    write_colormap(cinfo, dest, cmap_entries, 3);
+}
+
+
+/*
+ * Write the colormap.
+ * Windows uses BGR0 map entries; OS/2 uses BGR entries.
+ */
+
+LOCAL(void)
+write_colormap (j_decompress_ptr cinfo, bmp_dest_ptr dest,
+		int map_colors, int map_entry_size)
+{
+  JSAMPARRAY colormap = cinfo->colormap;
+  int num_colors = cinfo->actual_number_of_colors;
+  FILE * outfile = dest->pub.output_file;
+  int i;
+
+  if (colormap != NULL) {
+    if (cinfo->out_color_components == 3) {
+      /* Normal case with RGB colormap */
+      for (i = 0; i < num_colors; i++) {
+	putc(GETJSAMPLE(colormap[2][i]), outfile);
+	putc(GETJSAMPLE(colormap[1][i]), outfile);
+	putc(GETJSAMPLE(colormap[0][i]), outfile);
+	if (map_entry_size == 4)
+	  putc(0, outfile);
+      }
+    } else {
+      /* Grayscale colormap (only happens with grayscale quantization) */
+      for (i = 0; i < num_colors; i++) {
+	putc(GETJSAMPLE(colormap[0][i]), outfile);
+	putc(GETJSAMPLE(colormap[0][i]), outfile);
+	putc(GETJSAMPLE(colormap[0][i]), outfile);
+	if (map_entry_size == 4)
+	  putc(0, outfile);
+      }
+    }
+  } else {
+    /* If no colormap, must be grayscale data.  Generate a linear "map". */
+    for (i = 0; i < 256; i++) {
+      putc(i, outfile);
+      putc(i, outfile);
+      putc(i, outfile);
+      if (map_entry_size == 4)
+	putc(0, outfile);
+    }
+  }
+  /* Pad colormap with zeros to ensure specified number of colormap entries */ 
+  if (i > map_colors)
+    ERREXIT1(cinfo, JERR_TOO_MANY_COLORS, i);
+  for (; i < map_colors; i++) {
+    putc(0, outfile);
+    putc(0, outfile);
+    putc(0, outfile);
+    if (map_entry_size == 4)
+      putc(0, outfile);
+  }
+}
+
+
+METHODDEF(void)
+finish_output_bmp (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr) dinfo;
+  register FILE * outfile = dest->pub.output_file;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW data_ptr;
+  JDIMENSION row;
+  register JDIMENSION col;
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+
+  /* Write the header and colormap */
+  if (dest->is_os2)
+    write_os2_header(cinfo, dest);
+  else
+    write_bmp_header(cinfo, dest);
+
+  /* Write the file body from our virtual array */
+  for (row = cinfo->output_height; row > 0; row--) {
+    if (progress != NULL) {
+      progress->pub.pass_counter = (long) (cinfo->output_height - row);
+      progress->pub.pass_limit = (long) cinfo->output_height;
+      (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+    }
+    image_ptr = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr) cinfo, dest->whole_image, row-1, (JDIMENSION) 1, FALSE);
+    data_ptr = image_ptr[0];
+    for (col = dest->row_width; col > 0; col--) {
+      putc(GETJSAMPLE(*data_ptr), outfile);
+      data_ptr++;
+    }
+  }
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+
+  /* Make sure we wrote the output file OK */
+  fflush(outfile);
+  if (ferror(outfile))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for BMP format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_bmp (j_decompress_ptr cinfo, boolean is_os2)
+{
+  bmp_dest_ptr dest;
+  JDIMENSION row_width;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (bmp_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(bmp_dest_struct));
+  dest->pub.start_output = start_output_bmp;
+  dest->pub.finish_output = finish_output_bmp;
+  dest->is_os2 = is_os2;
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE) {
+    dest->pub.put_pixel_rows = put_gray_rows;
+  } else if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors)
+      dest->pub.put_pixel_rows = put_gray_rows;
+    else
+      dest->pub.put_pixel_rows = put_pixel_rows;
+  } else {
+    ERREXIT(cinfo, JERR_BMP_COLORSPACE);
+  }
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Determine width of rows in the BMP file (padded to 4-byte boundary). */
+  row_width = cinfo->output_width * cinfo->output_components;
+  dest->data_width = row_width;
+  while ((row_width & 3) != 0) row_width++;
+  dest->row_width = row_width;
+  dest->pad_bytes = (int) (row_width - dest->data_width);
+
+  /* Allocate space for inversion array, prepare for write pass */
+  dest->whole_image = (*cinfo->mem->request_virt_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+     row_width, cinfo->output_height, (JDIMENSION) 1);
+  dest->cur_output_row = 0;
+  if (cinfo->progress != NULL) {
+    cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+    progress->total_extra_passes++; /* count file input as separate pass */
+  }
+
+  /* Create decompressor output buffer. */
+  dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, row_width, (JDIMENSION) 1);
+  dest->pub.buffer_height = 1;
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* BMP_SUPPORTED */
diff --git a/libjpegtwrp/wrgif.c b/libjpegtwrp/wrgif.c
new file mode 100644
index 000000000..5fe832839
--- /dev/null
+++ b/libjpegtwrp/wrgif.c
@@ -0,0 +1,399 @@
+/*
+ * wrgif.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to write output images in GIF format.
+ *
+ **************************************************************************
+ * NOTE: to avoid entanglements with Unisys' patent on LZW compression,   *
+ * this code has been modified to output "uncompressed GIF" files.        *
+ * There is no trace of the LZW algorithm in this file.                   *
+ **************************************************************************
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ */
+
+/*
+ * This code is loosely based on ppmtogif from the PBMPLUS distribution
+ * of Feb. 1991.  That file contains the following copyright notice:
+ *    Based on GIFENCODE by David Rowley <mgardi@watdscu.waterloo.edu>.
+ *    Lempel-Ziv compression based on "compress" by Spencer W. Thomas et al.
+ *    Copyright (C) 1989 by Jef Poskanzer.
+ *    Permission to use, copy, modify, and distribute this software and its
+ *    documentation for any purpose and without fee is hereby granted, provided
+ *    that the above copyright notice appear in all copies and that both that
+ *    copyright notice and this permission notice appear in supporting
+ *    documentation.  This software is provided "as is" without express or
+ *    implied warranty.
+ *
+ * We are also required to state that
+ *    "The Graphics Interchange Format(c) is the Copyright property of
+ *    CompuServe Incorporated. GIF(sm) is a Service Mark property of
+ *    CompuServe Incorporated."
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+
+#ifdef GIF_SUPPORTED
+
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub;	/* public fields */
+
+  j_decompress_ptr cinfo;	/* back link saves passing separate parm */
+
+  /* State for packing variable-width codes into a bitstream */
+  int n_bits;			/* current number of bits/code */
+  int maxcode;			/* maximum code, given n_bits */
+  INT32 cur_accum;		/* holds bits not yet output */
+  int cur_bits;			/* # of bits in cur_accum */
+
+  /* State for GIF code assignment */
+  int ClearCode;		/* clear code (doesn't change) */
+  int EOFCode;			/* EOF code (ditto) */
+  int code_counter;		/* counts output symbols */
+
+  /* GIF data packet construction buffer */
+  int bytesinpkt;		/* # of bytes in current packet */
+  char packetbuf[256];		/* workspace for accumulating packet */
+
+} gif_dest_struct;
+
+typedef gif_dest_struct * gif_dest_ptr;
+
+/* Largest value that will fit in N bits */
+#define MAXCODE(n_bits)	((1 << (n_bits)) - 1)
+
+
+/*
+ * Routines to package finished data bytes into GIF data blocks.
+ * A data block consists of a count byte (1..255) and that many data bytes.
+ */
+
+LOCAL(void)
+flush_packet (gif_dest_ptr dinfo)
+/* flush any accumulated data */
+{
+  if (dinfo->bytesinpkt > 0) {	/* never write zero-length packet */
+    dinfo->packetbuf[0] = (char) dinfo->bytesinpkt++;
+    if (JFWRITE(dinfo->pub.output_file, dinfo->packetbuf, dinfo->bytesinpkt)
+	!= (size_t) dinfo->bytesinpkt)
+      ERREXIT(dinfo->cinfo, JERR_FILE_WRITE);
+    dinfo->bytesinpkt = 0;
+  }
+}
+
+
+/* Add a character to current packet; flush to disk if necessary */
+#define CHAR_OUT(dinfo,c)  \
+	{ (dinfo)->packetbuf[++(dinfo)->bytesinpkt] = (char) (c);  \
+	    if ((dinfo)->bytesinpkt >= 255)  \
+	      flush_packet(dinfo);  \
+	}
+
+
+/* Routine to convert variable-width codes into a byte stream */
+
+LOCAL(void)
+output (gif_dest_ptr dinfo, int code)
+/* Emit a code of n_bits bits */
+/* Uses cur_accum and cur_bits to reblock into 8-bit bytes */
+{
+  dinfo->cur_accum |= ((INT32) code) << dinfo->cur_bits;
+  dinfo->cur_bits += dinfo->n_bits;
+
+  while (dinfo->cur_bits >= 8) {
+    CHAR_OUT(dinfo, dinfo->cur_accum & 0xFF);
+    dinfo->cur_accum >>= 8;
+    dinfo->cur_bits -= 8;
+  }
+}
+
+
+/* The pseudo-compression algorithm.
+ *
+ * In this module we simply output each pixel value as a separate symbol;
+ * thus, no compression occurs.  In fact, there is expansion of one bit per
+ * pixel, because we use a symbol width one bit wider than the pixel width.
+ *
+ * GIF ordinarily uses variable-width symbols, and the decoder will expect
+ * to ratchet up the symbol width after a fixed number of symbols.
+ * To simplify the logic and keep the expansion penalty down, we emit a
+ * GIF Clear code to reset the decoder just before the width would ratchet up.
+ * Thus, all the symbols in the output file will have the same bit width.
+ * Note that emitting the Clear codes at the right times is a mere matter of
+ * counting output symbols and is in no way dependent on the LZW patent.
+ *
+ * With a small basic pixel width (low color count), Clear codes will be
+ * needed very frequently, causing the file to expand even more.  So this
+ * simplistic approach wouldn't work too well on bilevel images, for example.
+ * But for output of JPEG conversions the pixel width will usually be 8 bits
+ * (129 to 256 colors), so the overhead added by Clear symbols is only about
+ * one symbol in every 256.
+ */
+
+LOCAL(void)
+compress_init (gif_dest_ptr dinfo, int i_bits)
+/* Initialize pseudo-compressor */
+{
+  /* init all the state variables */
+  dinfo->n_bits = i_bits;
+  dinfo->maxcode = MAXCODE(dinfo->n_bits);
+  dinfo->ClearCode = (1 << (i_bits - 1));
+  dinfo->EOFCode = dinfo->ClearCode + 1;
+  dinfo->code_counter = dinfo->ClearCode + 2;
+  /* init output buffering vars */
+  dinfo->bytesinpkt = 0;
+  dinfo->cur_accum = 0;
+  dinfo->cur_bits = 0;
+  /* GIF specifies an initial Clear code */
+  output(dinfo, dinfo->ClearCode);
+}
+
+
+LOCAL(void)
+compress_pixel (gif_dest_ptr dinfo, int c)
+/* Accept and "compress" one pixel value.
+ * The given value must be less than n_bits wide.
+ */
+{
+  /* Output the given pixel value as a symbol. */
+  output(dinfo, c);
+  /* Issue Clear codes often enough to keep the reader from ratcheting up
+   * its symbol size.
+   */
+  if (dinfo->code_counter < dinfo->maxcode) {
+    dinfo->code_counter++;
+  } else {
+    output(dinfo, dinfo->ClearCode);
+    dinfo->code_counter = dinfo->ClearCode + 2;	/* reset the counter */
+  }
+}
+
+
+LOCAL(void)
+compress_term (gif_dest_ptr dinfo)
+/* Clean up at end */
+{
+  /* Send an EOF code */
+  output(dinfo, dinfo->EOFCode);
+  /* Flush the bit-packing buffer */
+  if (dinfo->cur_bits > 0) {
+    CHAR_OUT(dinfo, dinfo->cur_accum & 0xFF);
+  }
+  /* Flush the packet buffer */
+  flush_packet(dinfo);
+}
+
+
+/* GIF header construction */
+
+
+LOCAL(void)
+put_word (gif_dest_ptr dinfo, unsigned int w)
+/* Emit a 16-bit word, LSB first */
+{
+  putc(w & 0xFF, dinfo->pub.output_file);
+  putc((w >> 8) & 0xFF, dinfo->pub.output_file);
+}
+
+
+LOCAL(void)
+put_3bytes (gif_dest_ptr dinfo, int val)
+/* Emit 3 copies of same byte value --- handy subr for colormap construction */
+{
+  putc(val, dinfo->pub.output_file);
+  putc(val, dinfo->pub.output_file);
+  putc(val, dinfo->pub.output_file);
+}
+
+
+LOCAL(void)
+emit_header (gif_dest_ptr dinfo, int num_colors, JSAMPARRAY colormap)
+/* Output the GIF file header, including color map */
+/* If colormap==NULL, synthesize a gray-scale colormap */
+{
+  int BitsPerPixel, ColorMapSize, InitCodeSize, FlagByte;
+  int cshift = dinfo->cinfo->data_precision - 8;
+  int i;
+
+  if (num_colors > 256)
+    ERREXIT1(dinfo->cinfo, JERR_TOO_MANY_COLORS, num_colors);
+  /* Compute bits/pixel and related values */
+  BitsPerPixel = 1;
+  while (num_colors > (1 << BitsPerPixel))
+    BitsPerPixel++;
+  ColorMapSize = 1 << BitsPerPixel;
+  if (BitsPerPixel <= 1)
+    InitCodeSize = 2;
+  else
+    InitCodeSize = BitsPerPixel;
+  /*
+   * Write the GIF header.
+   * Note that we generate a plain GIF87 header for maximum compatibility.
+   */
+  putc('G', dinfo->pub.output_file);
+  putc('I', dinfo->pub.output_file);
+  putc('F', dinfo->pub.output_file);
+  putc('8', dinfo->pub.output_file);
+  putc('7', dinfo->pub.output_file);
+  putc('a', dinfo->pub.output_file);
+  /* Write the Logical Screen Descriptor */
+  put_word(dinfo, (unsigned int) dinfo->cinfo->output_width);
+  put_word(dinfo, (unsigned int) dinfo->cinfo->output_height);
+  FlagByte = 0x80;		/* Yes, there is a global color table */
+  FlagByte |= (BitsPerPixel-1) << 4; /* color resolution */
+  FlagByte |= (BitsPerPixel-1);	/* size of global color table */
+  putc(FlagByte, dinfo->pub.output_file);
+  putc(0, dinfo->pub.output_file); /* Background color index */
+  putc(0, dinfo->pub.output_file); /* Reserved (aspect ratio in GIF89) */
+  /* Write the Global Color Map */
+  /* If the color map is more than 8 bits precision, */
+  /* we reduce it to 8 bits by shifting */
+  for (i=0; i < ColorMapSize; i++) {
+    if (i < num_colors) {
+      if (colormap != NULL) {
+	if (dinfo->cinfo->out_color_space == JCS_RGB) {
+	  /* Normal case: RGB color map */
+	  putc(GETJSAMPLE(colormap[0][i]) >> cshift, dinfo->pub.output_file);
+	  putc(GETJSAMPLE(colormap[1][i]) >> cshift, dinfo->pub.output_file);
+	  putc(GETJSAMPLE(colormap[2][i]) >> cshift, dinfo->pub.output_file);
+	} else {
+	  /* Grayscale "color map": possible if quantizing grayscale image */
+	  put_3bytes(dinfo, GETJSAMPLE(colormap[0][i]) >> cshift);
+	}
+      } else {
+	/* Create a gray-scale map of num_colors values, range 0..255 */
+	put_3bytes(dinfo, (i * 255 + (num_colors-1)/2) / (num_colors-1));
+      }
+    } else {
+      /* fill out the map to a power of 2 */
+      put_3bytes(dinfo, 0);
+    }
+  }
+  /* Write image separator and Image Descriptor */
+  putc(',', dinfo->pub.output_file); /* separator */
+  put_word(dinfo, 0);		/* left/top offset */
+  put_word(dinfo, 0);
+  put_word(dinfo, (unsigned int) dinfo->cinfo->output_width); /* image size */
+  put_word(dinfo, (unsigned int) dinfo->cinfo->output_height);
+  /* flag byte: not interlaced, no local color map */
+  putc(0x00, dinfo->pub.output_file);
+  /* Write Initial Code Size byte */
+  putc(InitCodeSize, dinfo->pub.output_file);
+
+  /* Initialize for "compression" of image data */
+  compress_init(dinfo, InitCodeSize+1);
+}
+
+
+/*
+ * Startup: write the file header.
+ */
+
+METHODDEF(void)
+start_output_gif (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  gif_dest_ptr dest = (gif_dest_ptr) dinfo;
+
+  if (cinfo->quantize_colors)
+    emit_header(dest, cinfo->actual_number_of_colors, cinfo->colormap);
+  else
+    emit_header(dest, 256, (JSAMPARRAY) NULL);
+}
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ */
+
+METHODDEF(void)
+put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+		JDIMENSION rows_supplied)
+{
+  gif_dest_ptr dest = (gif_dest_ptr) dinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer[0];
+  for (col = cinfo->output_width; col > 0; col--) {
+    compress_pixel(dest, GETJSAMPLE(*ptr++));
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_output_gif (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  gif_dest_ptr dest = (gif_dest_ptr) dinfo;
+
+  /* Flush "compression" mechanism */
+  compress_term(dest);
+  /* Write a zero-length data block to end the series */
+  putc(0, dest->pub.output_file);
+  /* Write the GIF terminator mark */
+  putc(';', dest->pub.output_file);
+  /* Make sure we wrote the output file OK */
+  fflush(dest->pub.output_file);
+  if (ferror(dest->pub.output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for GIF format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_gif (j_decompress_ptr cinfo)
+{
+  gif_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (gif_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(gif_dest_struct));
+  dest->cinfo = cinfo;		/* make back link for subroutines */
+  dest->pub.start_output = start_output_gif;
+  dest->pub.put_pixel_rows = put_pixel_rows;
+  dest->pub.finish_output = finish_output_gif;
+
+  if (cinfo->out_color_space != JCS_GRAYSCALE &&
+      cinfo->out_color_space != JCS_RGB)
+    ERREXIT(cinfo, JERR_GIF_COLORSPACE);
+
+  /* Force quantization if color or if > 8 bits input */
+  if (cinfo->out_color_space != JCS_GRAYSCALE || cinfo->data_precision > 8) {
+    /* Force quantization to at most 256 colors */
+    cinfo->quantize_colors = TRUE;
+    if (cinfo->desired_number_of_colors > 256)
+      cinfo->desired_number_of_colors = 256;
+  }
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  if (cinfo->output_components != 1) /* safety check: just one component? */
+    ERREXIT(cinfo, JERR_GIF_BUG);
+
+  /* Create decompressor output buffer. */
+  dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, cinfo->output_width, (JDIMENSION) 1);
+  dest->pub.buffer_height = 1;
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* GIF_SUPPORTED */
diff --git a/libjpegtwrp/wrjpgcom.1 b/libjpegtwrp/wrjpgcom.1
new file mode 100644
index 000000000..d419a9999
--- /dev/null
+++ b/libjpegtwrp/wrjpgcom.1
@@ -0,0 +1,103 @@
+.TH WRJPGCOM 1 "15 June 1995"
+.SH NAME
+wrjpgcom \- insert text comments into a JPEG file
+.SH SYNOPSIS
+.B wrjpgcom
+[
+.B \-replace
+]
+[
+.BI \-comment " text"
+]
+[
+.BI \-cfile " name"
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B wrjpgcom
+reads the named JPEG/JFIF file, or the standard input if no file is named,
+and generates a new JPEG/JFIF file on standard output.  A comment block is
+added to the file.
+.PP
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+.PP
+.B wrjpgcom
+adds a COM block, containing text you provide, to a JPEG file.
+Ordinarily, the COM block is added after any existing COM blocks; but you
+can delete the old COM blocks if you wish.
+.SH OPTIONS
+Switch names may be abbreviated, and are not case sensitive.
+.TP
+.B \-replace
+Delete any existing COM blocks from the file.
+.TP
+.BI \-comment " text"
+Supply text for new COM block on command line.
+.TP
+.BI \-cfile " name"
+Read text for new COM block from named file.
+.PP
+If you have only one line of comment text to add, you can provide it on the
+command line with
+.BR \-comment .
+The comment text must be surrounded with quotes so that it is treated as a
+single argument.  Longer comments can be read from a text file.
+.PP
+If you give neither
+.B \-comment
+nor
+.BR \-cfile ,
+then
+.B wrjpgcom
+will read the comment text from standard input.  (In this case an input image
+file name MUST be supplied, so that the source JPEG file comes from somewhere
+else.)  You can enter multiple lines, up to 64KB worth.  Type an end-of-file
+indicator (usually control-D) to terminate the comment text entry.
+.PP
+.B wrjpgcom
+will not add a COM block if the provided comment string is empty.  Therefore
+\fB\-replace \-comment ""\fR can be used to delete all COM blocks from a file.
+.SH EXAMPLES
+.LP
+Add a short comment to in.jpg, producing out.jpg:
+.IP
+.B wrjpgcom \-c
+\fI"View of my back yard" in.jpg
+.B >
+.I out.jpg
+.PP
+Attach a long comment previously stored in comment.txt:
+.IP
+.B wrjpgcom
+.I in.jpg
+.B <
+.I comment.txt
+.B >
+.I out.jpg
+.PP
+or equivalently
+.IP
+.B wrjpgcom
+.B -cfile
+.I comment.txt
+.B <
+.I in.jpg
+.B >
+.I out.jpg
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR djpeg (1),
+.BR jpegtran (1),
+.BR rdjpgcom (1)
+.SH AUTHOR
+Independent JPEG Group
diff --git a/libjpegtwrp/wrjpgcom.c b/libjpegtwrp/wrjpgcom.c
new file mode 100644
index 000000000..8c04b0551
--- /dev/null
+++ b/libjpegtwrp/wrjpgcom.c
@@ -0,0 +1,583 @@
+/*
+ * wrjpgcom.c
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a very simple stand-alone application that inserts
+ * user-supplied text as a COM (comment) marker in a JFIF file.
+ * This may be useful as an example of the minimum logic needed to parse
+ * JPEG markers.
+ */
+
+#define JPEG_CJPEG_DJPEG	/* to get the command-line config symbols */
+#include "jinclude.h"		/* get auto-config symbols, <stdio.h> */
+
+#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare malloc() */
+extern void * malloc ();
+#endif
+#include <ctype.h>		/* to declare isupper(), tolower() */
+#ifdef USE_SETMODE
+#include <fcntl.h>		/* to declare setmode()'s parameter macros */
+/* If you have setmode() but not <io.h>, just delete this line: */
+#include <io.h>			/* to declare setmode() */
+#endif
+
+#ifdef USE_CCOMMAND		/* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>		/* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>		/* Think declares it here */
+#endif
+#endif
+
+#ifdef DONT_USE_B_MODE		/* define mode parameters for fopen() */
+#define READ_BINARY	"r"
+#define WRITE_BINARY	"w"
+#else
+#ifdef VMS			/* VMS is very nonstandard */
+#define READ_BINARY	"rb", "ctx=stm"
+#define WRITE_BINARY	"wb", "ctx=stm"
+#else				/* standard ANSI-compliant case */
+#define READ_BINARY	"rb"
+#define WRITE_BINARY	"wb"
+#endif
+#endif
+
+#ifndef EXIT_FAILURE		/* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+#ifndef EXIT_SUCCESS
+#ifdef VMS
+#define EXIT_SUCCESS  1		/* VMS is very nonstandard */
+#else
+#define EXIT_SUCCESS  0
+#endif
+#endif
+
+/* Reduce this value if your malloc() can't allocate blocks up to 64K.
+ * On DOS, compiling in large model is usually a better solution.
+ */
+
+#ifndef MAX_COM_LENGTH
+#define MAX_COM_LENGTH 65000L	/* must be <= 65533 in any case */
+#endif
+
+
+/*
+ * These macros are used to read the input file and write the output file.
+ * To reuse this code in another application, you might need to change these.
+ */
+
+static FILE * infile;		/* input JPEG file */
+
+/* Return next input byte, or EOF if no more */
+#define NEXTBYTE()  getc(infile)
+
+static FILE * outfile;		/* output JPEG file */
+
+/* Emit an output byte */
+#define PUTBYTE(x)  putc((x), outfile)
+
+
+/* Error exit handler */
+#define ERREXIT(msg)  (fprintf(stderr, "%s\n", msg), exit(EXIT_FAILURE))
+
+
+/* Read one byte, testing for EOF */
+static int
+read_1_byte (void)
+{
+  int c;
+
+  c = NEXTBYTE();
+  if (c == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return c;
+}
+
+/* Read 2 bytes, convert to unsigned int */
+/* All 2-byte quantities in JPEG markers are MSB first */
+static unsigned int
+read_2_bytes (void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  if (c1 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  c2 = NEXTBYTE();
+  if (c2 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return (((unsigned int) c1) << 8) + ((unsigned int) c2);
+}
+
+
+/* Routines to write data to output file */
+
+static void
+write_1_byte (int c)
+{
+  PUTBYTE(c);
+}
+
+static void
+write_2_bytes (unsigned int val)
+{
+  PUTBYTE((val >> 8) & 0xFF);
+  PUTBYTE(val & 0xFF);
+}
+
+static void
+write_marker (int marker)
+{
+  PUTBYTE(0xFF);
+  PUTBYTE(marker);
+}
+
+static void
+copy_rest_of_file (void)
+{
+  int c;
+
+  while ((c = NEXTBYTE()) != EOF)
+    PUTBYTE(c);
+}
+
+
+/*
+ * JPEG markers consist of one or more 0xFF bytes, followed by a marker
+ * code byte (which is not an FF).  Here are the marker codes of interest
+ * in this program.  (See jdmarker.c for a more complete list.)
+ */
+
+#define M_SOF0  0xC0		/* Start Of Frame N */
+#define M_SOF1  0xC1		/* N indicates which compression process */
+#define M_SOF2  0xC2		/* Only SOF0-SOF2 are now in common use */
+#define M_SOF3  0xC3
+#define M_SOF5  0xC5		/* NB: codes C4 and CC are NOT SOF markers */
+#define M_SOF6  0xC6
+#define M_SOF7  0xC7
+#define M_SOF9  0xC9
+#define M_SOF10 0xCA
+#define M_SOF11 0xCB
+#define M_SOF13 0xCD
+#define M_SOF14 0xCE
+#define M_SOF15 0xCF
+#define M_SOI   0xD8		/* Start Of Image (beginning of datastream) */
+#define M_EOI   0xD9		/* End Of Image (end of datastream) */
+#define M_SOS   0xDA		/* Start Of Scan (begins compressed data) */
+#define M_COM   0xFE		/* COMment */
+
+
+/*
+ * Find the next JPEG marker and return its marker code.
+ * We expect at least one FF byte, possibly more if the compressor used FFs
+ * to pad the file.  (Padding FFs will NOT be replicated in the output file.)
+ * There could also be non-FF garbage between markers.  The treatment of such
+ * garbage is unspecified; we choose to skip over it but emit a warning msg.
+ * NB: this routine must not be used after seeing SOS marker, since it will
+ * not deal correctly with FF/00 sequences in the compressed image data...
+ */
+
+static int
+next_marker (void)
+{
+  int c;
+  int discarded_bytes = 0;
+
+  /* Find 0xFF byte; count and skip any non-FFs. */
+  c = read_1_byte();
+  while (c != 0xFF) {
+    discarded_bytes++;
+    c = read_1_byte();
+  }
+  /* Get marker code byte, swallowing any duplicate FF bytes.  Extra FFs
+   * are legal as pad bytes, so don't count them in discarded_bytes.
+   */
+  do {
+    c = read_1_byte();
+  } while (c == 0xFF);
+
+  if (discarded_bytes != 0) {
+    fprintf(stderr, "Warning: garbage data found in JPEG file\n");
+  }
+
+  return c;
+}
+
+
+/*
+ * Read the initial marker, which should be SOI.
+ * For a JFIF file, the first two bytes of the file should be literally
+ * 0xFF M_SOI.  To be more general, we could use next_marker, but if the
+ * input file weren't actually JPEG at all, next_marker might read the whole
+ * file and then return a misleading error message...
+ */
+
+static int
+first_marker (void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  c2 = NEXTBYTE();
+  if (c1 != 0xFF || c2 != M_SOI)
+    ERREXIT("Not a JPEG file");
+  return c2;
+}
+
+
+/*
+ * Most types of marker are followed by a variable-length parameter segment.
+ * This routine skips over the parameters for any marker we don't otherwise
+ * want to process.
+ * Note that we MUST skip the parameter segment explicitly in order not to
+ * be fooled by 0xFF bytes that might appear within the parameter segment;
+ * such bytes do NOT introduce new markers.
+ */
+
+static void
+copy_variable (void)
+/* Copy an unknown or uninteresting variable-length marker */
+{
+  unsigned int length;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  write_2_bytes(length);
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+  /* Skip over the remaining bytes */
+  while (length > 0) {
+    write_1_byte(read_1_byte());
+    length--;
+  }
+}
+
+static void
+skip_variable (void)
+/* Skip over an unknown or uninteresting variable-length marker */
+{
+  unsigned int length;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+  /* Skip over the remaining bytes */
+  while (length > 0) {
+    (void) read_1_byte();
+    length--;
+  }
+}
+
+
+/*
+ * Parse the marker stream until SOFn or EOI is seen;
+ * copy data to output, but discard COM markers unless keep_COM is true.
+ */
+
+static int
+scan_JPEG_header (int keep_COM)
+{
+  int marker;
+
+  /* Expect SOI at start of file */
+  if (first_marker() != M_SOI)
+    ERREXIT("Expected SOI marker first");
+  write_marker(M_SOI);
+
+  /* Scan miscellaneous markers until we reach SOFn. */
+  for (;;) {
+    marker = next_marker();
+    switch (marker) {
+      /* Note that marker codes 0xC4, 0xC8, 0xCC are not, and must not be,
+       * treated as SOFn.  C4 in particular is actually DHT.
+       */
+    case M_SOF0:		/* Baseline */
+    case M_SOF1:		/* Extended sequential, Huffman */
+    case M_SOF2:		/* Progressive, Huffman */
+    case M_SOF3:		/* Lossless, Huffman */
+    case M_SOF5:		/* Differential sequential, Huffman */
+    case M_SOF6:		/* Differential progressive, Huffman */
+    case M_SOF7:		/* Differential lossless, Huffman */
+    case M_SOF9:		/* Extended sequential, arithmetic */
+    case M_SOF10:		/* Progressive, arithmetic */
+    case M_SOF11:		/* Lossless, arithmetic */
+    case M_SOF13:		/* Differential sequential, arithmetic */
+    case M_SOF14:		/* Differential progressive, arithmetic */
+    case M_SOF15:		/* Differential lossless, arithmetic */
+      return marker;
+
+    case M_SOS:			/* should not see compressed data before SOF */
+      ERREXIT("SOS without prior SOFn");
+      break;
+
+    case M_EOI:			/* in case it's a tables-only JPEG stream */
+      return marker;
+
+    case M_COM:			/* Existing COM: conditionally discard */
+      if (keep_COM) {
+	write_marker(marker);
+	copy_variable();
+      } else {
+	skip_variable();
+      }
+      break;
+
+    default:			/* Anything else just gets copied */
+      write_marker(marker);
+      copy_variable();		/* we assume it has a parameter count... */
+      break;
+    }
+  } /* end loop */
+}
+
+
+/* Command line parsing code */
+
+static const char * progname;	/* program name for error messages */
+
+
+static void
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "wrjpgcom inserts a textual comment in a JPEG file.\n");
+  fprintf(stderr, "You can add to or replace any existing comment(s).\n");
+
+  fprintf(stderr, "Usage: %s [switches] ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "[inputfile]\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -replace         Delete any existing comments\n");
+  fprintf(stderr, "  -comment \"text\"  Insert comment with given text\n");
+  fprintf(stderr, "  -cfile name      Read comment from named file\n");
+  fprintf(stderr, "Notice that you must put quotes around the comment text\n");
+  fprintf(stderr, "when you use -comment.\n");
+  fprintf(stderr, "If you do not give either -comment or -cfile on the command line,\n");
+  fprintf(stderr, "then the comment text is read from standard input.\n");
+  fprintf(stderr, "It can be multiple lines, up to %u characters total.\n",
+	  (unsigned int) MAX_COM_LENGTH);
+#ifndef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "You must specify an input JPEG file name when supplying\n");
+  fprintf(stderr, "comment text from standard input.\n");
+#endif
+
+  exit(EXIT_FAILURE);
+}
+
+
+static int
+keymatch (char * arg, const char * keyword, int minchars)
+/* Case-insensitive matching of (possibly abbreviated) keyword switches. */
+/* keyword is the constant keyword (must be lower case already), */
+/* minchars is length of minimum legal abbreviation. */
+{
+  register int ca, ck;
+  register int nmatched = 0;
+
+  while ((ca = *arg++) != '\0') {
+    if ((ck = *keyword++) == '\0')
+      return 0;			/* arg longer than keyword, no good */
+    if (isupper(ca))		/* force arg to lcase (assume ck is already) */
+      ca = tolower(ca);
+    if (ca != ck)
+      return 0;			/* no good */
+    nmatched++;			/* count matched characters */
+  }
+  /* reached end of argument; fail if it's too short for unique abbrev */
+  if (nmatched < minchars)
+    return 0;
+  return 1;			/* A-OK */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  int argn;
+  char * arg;
+  int keep_COM = 1;
+  char * comment_arg = NULL;
+  FILE * comment_file = NULL;
+  unsigned int comment_length = 0;
+  int marker;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "wrjpgcom";	/* in case C library doesn't provide it */
+
+  /* Parse switches, if any */
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (arg[0] != '-')
+      break;			/* not switch, must be file name */
+    arg++;			/* advance over '-' */
+    if (keymatch(arg, "replace", 1)) {
+      keep_COM = 0;
+    } else if (keymatch(arg, "cfile", 2)) {
+      if (++argn >= argc) usage();
+      if ((comment_file = fopen(argv[argn], "r")) == NULL) {
+	fprintf(stderr, "%s: can't open %s\n", progname, argv[argn]);
+	exit(EXIT_FAILURE);
+      }
+    } else if (keymatch(arg, "comment", 1)) {
+      if (++argn >= argc) usage();
+      comment_arg = argv[argn];
+      /* If the comment text starts with '"', then we are probably running
+       * under MS-DOG and must parse out the quoted string ourselves.  Sigh.
+       */
+      if (comment_arg[0] == '"') {
+	comment_arg = (char *) malloc((size_t) MAX_COM_LENGTH);
+	if (comment_arg == NULL)
+	  ERREXIT("Insufficient memory");
+	strcpy(comment_arg, argv[argn]+1);
+	for (;;) {
+	  comment_length = (unsigned int) strlen(comment_arg);
+	  if (comment_length > 0 && comment_arg[comment_length-1] == '"') {
+	    comment_arg[comment_length-1] = '\0'; /* zap terminating quote */
+	    break;
+	  }
+	  if (++argn >= argc)
+	    ERREXIT("Missing ending quote mark");
+	  strcat(comment_arg, " ");
+	  strcat(comment_arg, argv[argn]);
+	}
+      }
+      comment_length = (unsigned int) strlen(comment_arg);
+    } else
+      usage();
+  }
+
+  /* Cannot use both -comment and -cfile. */
+  if (comment_arg != NULL && comment_file != NULL)
+    usage();
+  /* If there is neither -comment nor -cfile, we will read the comment text
+   * from stdin; in this case there MUST be an input JPEG file name.
+   */
+  if (comment_arg == NULL && comment_file == NULL && argn >= argc)
+    usage();
+
+  /* Open the input file. */
+  if (argn < argc) {
+    if ((infile = fopen(argv[argn], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[argn]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+#ifdef USE_SETMODE		/* need to hack file mode? */
+    setmode(fileno(stdin), O_BINARY);
+#endif
+#ifdef USE_FDOPEN		/* need to re-open in binary mode? */
+    if ((infile = fdopen(fileno(stdin), READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open stdin\n", progname);
+      exit(EXIT_FAILURE);
+    }
+#else
+    infile = stdin;
+#endif
+  }
+
+  /* Open the output file. */
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have explicit output file name */
+  if (argn != argc-2) {
+    fprintf(stderr, "%s: must name one input and one output file\n",
+	    progname);
+    usage();
+  }
+  if ((outfile = fopen(argv[argn+1], WRITE_BINARY)) == NULL) {
+    fprintf(stderr, "%s: can't open %s\n", progname, argv[argn+1]);
+    exit(EXIT_FAILURE);
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (argn < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+  /* default output file is stdout */
+#ifdef USE_SETMODE		/* need to hack file mode? */
+  setmode(fileno(stdout), O_BINARY);
+#endif
+#ifdef USE_FDOPEN		/* need to re-open in binary mode? */
+  if ((outfile = fdopen(fileno(stdout), WRITE_BINARY)) == NULL) {
+    fprintf(stderr, "%s: can't open stdout\n", progname);
+    exit(EXIT_FAILURE);
+  }
+#else
+  outfile = stdout;
+#endif
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Collect comment text from comment_file or stdin, if necessary */
+  if (comment_arg == NULL) {
+    FILE * src_file;
+    int c;
+
+    comment_arg = (char *) malloc((size_t) MAX_COM_LENGTH);
+    if (comment_arg == NULL)
+      ERREXIT("Insufficient memory");
+    comment_length = 0;
+    src_file = (comment_file != NULL ? comment_file : stdin);
+    while ((c = getc(src_file)) != EOF) {
+      if (comment_length >= (unsigned int) MAX_COM_LENGTH) {
+	fprintf(stderr, "Comment text may not exceed %u bytes\n",
+		(unsigned int) MAX_COM_LENGTH);
+	exit(EXIT_FAILURE);
+      }
+      comment_arg[comment_length++] = (char) c;
+    }
+    if (comment_file != NULL)
+      fclose(comment_file);
+  }
+
+  /* Copy JPEG headers until SOFn marker;
+   * we will insert the new comment marker just before SOFn.
+   * This (a) causes the new comment to appear after, rather than before,
+   * existing comments; and (b) ensures that comments come after any JFIF
+   * or JFXX markers, as required by the JFIF specification.
+   */
+  marker = scan_JPEG_header(keep_COM);
+  /* Insert the new COM marker, but only if nonempty text has been supplied */
+  if (comment_length > 0) {
+    write_marker(M_COM);
+    write_2_bytes(comment_length + 2);
+    while (comment_length > 0) {
+      write_1_byte(*comment_arg++);
+      comment_length--;
+    }
+  }
+  /* Duplicate the remainder of the source file.
+   * Note that any COM markers occuring after SOF will not be touched.
+   */
+  write_marker(marker);
+  copy_rest_of_file();
+
+  /* All done. */
+  exit(EXIT_SUCCESS);
+  return 0;			/* suppress no-return-value warnings */
+}
diff --git a/libjpegtwrp/wrppm.c b/libjpegtwrp/wrppm.c
new file mode 100644
index 000000000..6c6d90881
--- /dev/null
+++ b/libjpegtwrp/wrppm.c
@@ -0,0 +1,268 @@
+/*
+ * wrppm.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to write output images in PPM/PGM format.
+ * The extended 2-byte-per-sample raw PPM/PGM formats are supported.
+ * The PBMPLUS library is NOT required to compile this software
+ * (but it is highly useful as a set of PPM image manipulation programs).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+
+#ifdef PPM_SUPPORTED
+
+
+/*
+ * For 12-bit JPEG data, we either downscale the values to 8 bits
+ * (to write standard byte-per-sample PPM/PGM files), or output
+ * nonstandard word-per-sample PPM/PGM files.  Downscaling is done
+ * if PPM_NORAWWORD is defined (this can be done in the Makefile
+ * or in jconfig.h).
+ * (When the core library supports data precision reduction, a cleaner
+ * implementation will be to ask for that instead.)
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define PUTPPMSAMPLE(ptr,v)  *ptr++ = (char) (v)
+#define BYTESPERSAMPLE 1
+#define PPM_MAXVAL 255
+#else
+#ifdef PPM_NORAWWORD
+#define PUTPPMSAMPLE(ptr,v)  *ptr++ = (char) ((v) >> (BITS_IN_JSAMPLE-8))
+#define BYTESPERSAMPLE 1
+#define PPM_MAXVAL 255
+#else
+/* The word-per-sample format always puts the LSB first. */
+#define PUTPPMSAMPLE(ptr,v)			\
+	{ register int val_ = v;		\
+	  *ptr++ = (char) (val_ & 0xFF);	\
+	  *ptr++ = (char) ((val_ >> 8) & 0xFF);	\
+	}
+#define BYTESPERSAMPLE 2
+#define PPM_MAXVAL ((1<<BITS_IN_JSAMPLE)-1)
+#endif
+#endif
+
+
+/*
+ * When JSAMPLE is the same size as char, we can just fwrite() the
+ * decompressed data to the PPM or PGM file.  On PCs, in order to make this
+ * work the output buffer must be allocated in near data space, because we are
+ * assuming small-data memory model wherein fwrite() can't reach far memory.
+ * If you need to process very wide images on a PC, you might have to compile
+ * in large-memory model, or else replace fwrite() with a putc() loop ---
+ * which will be much slower.
+ */
+
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub;	/* public fields */
+
+  /* Usually these two pointers point to the same place: */
+  char *iobuffer;		/* fwrite's I/O buffer */
+  JSAMPROW pixrow;		/* decompressor output buffer */
+  size_t buffer_width;		/* width of I/O buffer */
+  JDIMENSION samples_per_row;	/* JSAMPLEs per output row */
+} ppm_dest_struct;
+
+typedef ppm_dest_struct * ppm_dest_ptr;
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ *
+ * put_pixel_rows handles the "normal" 8-bit case where the decompressor
+ * output buffer is physically the same as the fwrite buffer.
+ */
+
+METHODDEF(void)
+put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+		JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * This code is used when we have to copy the data and apply a pixel
+ * format translation.  Typically this only happens in 12-bit mode.
+ */
+
+METHODDEF(void)
+copy_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+		 JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+  register char * bufferptr;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer[0];
+  bufferptr = dest->iobuffer;
+  for (col = dest->samples_per_row; col > 0; col--) {
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(*ptr++));
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Write some pixel data when color quantization is in effect.
+ * We have to demap the color index values to straight data.
+ */
+
+METHODDEF(void)
+put_demapped_rgb (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+		  JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+  register char * bufferptr;
+  register int pixval;
+  register JSAMPROW ptr;
+  register JSAMPROW color_map0 = cinfo->colormap[0];
+  register JSAMPROW color_map1 = cinfo->colormap[1];
+  register JSAMPROW color_map2 = cinfo->colormap[2];
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer[0];
+  bufferptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    pixval = GETJSAMPLE(*ptr++);
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(color_map0[pixval]));
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(color_map1[pixval]));
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(color_map2[pixval]));
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+METHODDEF(void)
+put_demapped_gray (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+		   JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+  register char * bufferptr;
+  register JSAMPROW ptr;
+  register JSAMPROW color_map = cinfo->colormap[0];
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer[0];
+  bufferptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(color_map[GETJSAMPLE(*ptr++)]));
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Startup: write the file header.
+ */
+
+METHODDEF(void)
+start_output_ppm (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+
+  /* Emit file header */
+  switch (cinfo->out_color_space) {
+  case JCS_GRAYSCALE:
+    /* emit header for raw PGM format */
+    fprintf(dest->pub.output_file, "P5\n%ld %ld\n%d\n",
+	    (long) cinfo->output_width, (long) cinfo->output_height,
+	    PPM_MAXVAL);
+    break;
+  case JCS_RGB:
+    /* emit header for raw PPM format */
+    fprintf(dest->pub.output_file, "P6\n%ld %ld\n%d\n",
+	    (long) cinfo->output_width, (long) cinfo->output_height,
+	    PPM_MAXVAL);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_PPM_COLORSPACE);
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_output_ppm (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  /* Make sure we wrote the output file OK */
+  fflush(dinfo->output_file);
+  if (ferror(dinfo->output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for PPM format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_ppm (j_decompress_ptr cinfo)
+{
+  ppm_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (ppm_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(ppm_dest_struct));
+  dest->pub.start_output = start_output_ppm;
+  dest->pub.finish_output = finish_output_ppm;
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Create physical I/O buffer.  Note we make this near on a PC. */
+  dest->samples_per_row = cinfo->output_width * cinfo->out_color_components;
+  dest->buffer_width = dest->samples_per_row * (BYTESPERSAMPLE * SIZEOF(char));
+  dest->iobuffer = (char *) (*cinfo->mem->alloc_small)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, dest->buffer_width);
+
+  if (cinfo->quantize_colors || BITS_IN_JSAMPLE != 8 ||
+      SIZEOF(JSAMPLE) != SIZEOF(char)) {
+    /* When quantizing, we need an output buffer for colormap indexes
+     * that's separate from the physical I/O buffer.  We also need a
+     * separate buffer if pixel format translation must take place.
+     */
+    dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       cinfo->output_width * cinfo->output_components, (JDIMENSION) 1);
+    dest->pub.buffer_height = 1;
+    if (! cinfo->quantize_colors)
+      dest->pub.put_pixel_rows = copy_pixel_rows;
+    else if (cinfo->out_color_space == JCS_GRAYSCALE)
+      dest->pub.put_pixel_rows = put_demapped_gray;
+    else
+      dest->pub.put_pixel_rows = put_demapped_rgb;
+  } else {
+    /* We will fwrite() directly from decompressor output buffer. */
+    /* Synthesize a JSAMPARRAY pointer structure */
+    /* Cast here implies near->far pointer conversion on PCs */
+    dest->pixrow = (JSAMPROW) dest->iobuffer;
+    dest->pub.buffer = & dest->pixrow;
+    dest->pub.buffer_height = 1;
+    dest->pub.put_pixel_rows = put_pixel_rows;
+  }
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* PPM_SUPPORTED */
diff --git a/libjpegtwrp/wrrle.c b/libjpegtwrp/wrrle.c
new file mode 100644
index 000000000..a4e73372d
--- /dev/null
+++ b/libjpegtwrp/wrrle.c
@@ -0,0 +1,305 @@
+/*
+ * wrrle.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to write output images in RLE format.
+ * The Utah Raster Toolkit library is required (version 3.1 or later).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ *
+ * Based on code contributed by Mike Lijewski,
+ * with updates from Robert Hutchinson.
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+
+#ifdef RLE_SUPPORTED
+
+/* rle.h is provided by the Utah Raster Toolkit. */
+
+#include <rle.h>
+
+/*
+ * We assume that JSAMPLE has the same representation as rle_pixel,
+ * to wit, "unsigned char".  Hence we can't cope with 12- or 16-bit samples.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+
+/*
+ * Since RLE stores scanlines bottom-to-top, we have to invert the image
+ * from JPEG's top-to-bottom order.  To do this, we save the outgoing data
+ * in a virtual array during put_pixel_row calls, then actually emit the
+ * RLE file during finish_output.
+ */
+
+
+/*
+ * For now, if we emit an RLE color map then it is always 256 entries long,
+ * though not all of the entries need be used.
+ */
+
+#define CMAPBITS	8
+#define CMAPLENGTH	(1<<(CMAPBITS))
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  jvirt_sarray_ptr image;	/* virtual array to store the output image */
+  rle_map *colormap;	 	/* RLE-style color map, or NULL if none */
+  rle_pixel **rle_row;		/* To pass rows to rle_putrow() */
+
+} rle_dest_struct;
+
+typedef rle_dest_struct * rle_dest_ptr;
+
+/* Forward declarations */
+METHODDEF(void) rle_put_pixel_rows
+    JPP((j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+	 JDIMENSION rows_supplied));
+
+
+/*
+ * Write the file header.
+ *
+ * In this module it's easier to wait till finish_output to write anything.
+ */
+
+METHODDEF(void)
+start_output_rle (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  rle_dest_ptr dest = (rle_dest_ptr) dinfo;
+  size_t cmapsize;
+  int i, ci;
+#ifdef PROGRESS_REPORT
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+#endif
+
+  /*
+   * Make sure the image can be stored in RLE format.
+   *
+   * - RLE stores image dimensions as *signed* 16 bit integers.  JPEG
+   *   uses unsigned, so we have to check the width.
+   *
+   * - Colorspace is expected to be grayscale or RGB.
+   *
+   * - The number of channels (components) is expected to be 1 (grayscale/
+   *   pseudocolor) or 3 (truecolor/directcolor).
+   *   (could be 2 or 4 if using an alpha channel, but we aren't)
+   */
+
+  if (cinfo->output_width > 32767 || cinfo->output_height > 32767)
+    ERREXIT2(cinfo, JERR_RLE_DIMENSIONS, cinfo->output_width, 
+	     cinfo->output_height);
+
+  if (cinfo->out_color_space != JCS_GRAYSCALE &&
+      cinfo->out_color_space != JCS_RGB)
+    ERREXIT(cinfo, JERR_RLE_COLORSPACE);
+
+  if (cinfo->output_components != 1 && cinfo->output_components != 3)
+    ERREXIT1(cinfo, JERR_RLE_TOOMANYCHANNELS, cinfo->num_components);
+
+  /* Convert colormap, if any, to RLE format. */
+
+  dest->colormap = NULL;
+
+  if (cinfo->quantize_colors) {
+    /* Allocate storage for RLE-style cmap, zero any extra entries */
+    cmapsize = cinfo->out_color_components * CMAPLENGTH * SIZEOF(rle_map);
+    dest->colormap = (rle_map *) (*cinfo->mem->alloc_small)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE, cmapsize);
+    MEMZERO(dest->colormap, cmapsize);
+
+    /* Save away data in RLE format --- note 8-bit left shift! */
+    /* Shifting would need adjustment for JSAMPLEs wider than 8 bits. */
+    for (ci = 0; ci < cinfo->out_color_components; ci++) {
+      for (i = 0; i < cinfo->actual_number_of_colors; i++) {
+        dest->colormap[ci * CMAPLENGTH + i] =
+          GETJSAMPLE(cinfo->colormap[ci][i]) << 8;
+      }
+    }
+  }
+
+  /* Set the output buffer to the first row */
+  dest->pub.buffer = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, dest->image, (JDIMENSION) 0, (JDIMENSION) 1, TRUE);
+  dest->pub.buffer_height = 1;
+
+  dest->pub.put_pixel_rows = rle_put_pixel_rows;
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL) {
+    progress->total_extra_passes++;  /* count file writing as separate pass */
+  }
+#endif
+}
+
+
+/*
+ * Write some pixel data.
+ *
+ * This routine just saves the data away in a virtual array.
+ */
+
+METHODDEF(void)
+rle_put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+		    JDIMENSION rows_supplied)
+{
+  rle_dest_ptr dest = (rle_dest_ptr) dinfo;
+
+  if (cinfo->output_scanline < cinfo->output_height) {
+    dest->pub.buffer = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr) cinfo, dest->image,
+       cinfo->output_scanline, (JDIMENSION) 1, TRUE);
+  }
+}
+
+/*
+ * Finish up at the end of the file.
+ *
+ * Here is where we really output the RLE file.
+ */
+
+METHODDEF(void)
+finish_output_rle (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  rle_dest_ptr dest = (rle_dest_ptr) dinfo;
+  rle_hdr header;		/* Output file information */
+  rle_pixel **rle_row, *red, *green, *blue;
+  JSAMPROW output_row;
+  char cmapcomment[80];
+  int row, col;
+  int ci;
+#ifdef PROGRESS_REPORT
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+#endif
+
+  /* Initialize the header info */
+  header = *rle_hdr_init(NULL);
+  header.rle_file = dest->pub.output_file;
+  header.xmin     = 0;
+  header.xmax     = cinfo->output_width  - 1;
+  header.ymin     = 0;
+  header.ymax     = cinfo->output_height - 1;
+  header.alpha    = 0;
+  header.ncolors  = cinfo->output_components;
+  for (ci = 0; ci < cinfo->output_components; ci++) {
+    RLE_SET_BIT(header, ci);
+  }
+  if (cinfo->quantize_colors) {
+    header.ncmap   = cinfo->out_color_components;
+    header.cmaplen = CMAPBITS;
+    header.cmap    = dest->colormap;
+    /* Add a comment to the output image with the true colormap length. */
+    sprintf(cmapcomment, "color_map_length=%d", cinfo->actual_number_of_colors);
+    rle_putcom(cmapcomment, &header);
+  }
+
+  /* Emit the RLE header and color map (if any) */
+  rle_put_setup(&header);
+
+  /* Now output the RLE data from our virtual array.
+   * We assume here that (a) rle_pixel is represented the same as JSAMPLE,
+   * and (b) we are not on a machine where FAR pointers differ from regular.
+   */
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL) {
+    progress->pub.pass_limit = cinfo->output_height;
+    progress->pub.pass_counter = 0;
+    (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+  }
+#endif
+
+  if (cinfo->output_components == 1) {
+    for (row = cinfo->output_height-1; row >= 0; row--) {
+      rle_row = (rle_pixel **) (*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr) cinfo, dest->image,
+	 (JDIMENSION) row, (JDIMENSION) 1, FALSE);
+      rle_putrow(rle_row, (int) cinfo->output_width, &header);
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+  } else {
+    for (row = cinfo->output_height-1; row >= 0; row--) {
+      rle_row = (rle_pixel **) dest->rle_row;
+      output_row = * (*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr) cinfo, dest->image,
+	 (JDIMENSION) row, (JDIMENSION) 1, FALSE);
+      red = rle_row[0];
+      green = rle_row[1];
+      blue = rle_row[2];
+      for (col = cinfo->output_width; col > 0; col--) {
+        *red++ = GETJSAMPLE(*output_row++);
+        *green++ = GETJSAMPLE(*output_row++);
+        *blue++ = GETJSAMPLE(*output_row++);
+      }
+      rle_putrow(rle_row, (int) cinfo->output_width, &header);
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+  }
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+#endif
+
+  /* Emit file trailer */
+  rle_puteof(&header);
+  fflush(dest->pub.output_file);
+  if (ferror(dest->pub.output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for RLE format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_rle (j_decompress_ptr cinfo)
+{
+  rle_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (rle_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  SIZEOF(rle_dest_struct));
+  dest->pub.start_output = start_output_rle;
+  dest->pub.finish_output = finish_output_rle;
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Allocate a work array for output to the RLE library. */
+  dest->rle_row = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     cinfo->output_width, (JDIMENSION) cinfo->output_components);
+
+  /* Allocate a virtual array to hold the image. */
+  dest->image = (*cinfo->mem->request_virt_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+     (JDIMENSION) (cinfo->output_width * cinfo->output_components),
+     cinfo->output_height, (JDIMENSION) 1);
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* RLE_SUPPORTED */
diff --git a/libjpegtwrp/wrtarga.c b/libjpegtwrp/wrtarga.c
new file mode 100644
index 000000000..cf104d2de
--- /dev/null
+++ b/libjpegtwrp/wrtarga.c
@@ -0,0 +1,253 @@
+/*
+ * wrtarga.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to write output images in Targa format.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ *
+ * Based on code contributed by Lee Daniel Crocker.
+ */
+
+#include "cdjpeg.h"		/* Common decls for cjpeg/djpeg applications */
+
+#ifdef TARGA_SUPPORTED
+
+
+/*
+ * To support 12-bit JPEG data, we'd have to scale output down to 8 bits.
+ * This is not yet implemented.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+/*
+ * The output buffer needs to be writable by fwrite().  On PCs, we must
+ * allocate the buffer in near data space, because we are assuming small-data
+ * memory model, wherein fwrite() can't reach far memory.  If you need to
+ * process very wide images on a PC, you might have to compile in large-memory
+ * model, or else replace fwrite() with a putc() loop --- which will be much
+ * slower.
+ */
+
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub;	/* public fields */
+
+  char *iobuffer;		/* physical I/O buffer */
+  JDIMENSION buffer_width;	/* width of one row */
+} tga_dest_struct;
+
+typedef tga_dest_struct * tga_dest_ptr;
+
+
+LOCAL(void)
+write_header (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo, int num_colors)
+/* Create and write a Targa header */
+{
+  char targaheader[18];
+
+  /* Set unused fields of header to 0 */
+  MEMZERO(targaheader, SIZEOF(targaheader));
+
+  if (num_colors > 0) {
+    targaheader[1] = 1;		/* color map type 1 */
+    targaheader[5] = (char) (num_colors & 0xFF);
+    targaheader[6] = (char) (num_colors >> 8);
+    targaheader[7] = 24;	/* 24 bits per cmap entry */
+  }
+
+  targaheader[12] = (char) (cinfo->output_width & 0xFF);
+  targaheader[13] = (char) (cinfo->output_width >> 8);
+  targaheader[14] = (char) (cinfo->output_height & 0xFF);
+  targaheader[15] = (char) (cinfo->output_height >> 8);
+  targaheader[17] = 0x20;	/* Top-down, non-interlaced */
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE) {
+    targaheader[2] = 3;		/* image type = uncompressed gray-scale */
+    targaheader[16] = 8;	/* bits per pixel */
+  } else {			/* must be RGB */
+    if (num_colors > 0) {
+      targaheader[2] = 1;	/* image type = colormapped RGB */
+      targaheader[16] = 8;
+    } else {
+      targaheader[2] = 2;	/* image type = uncompressed RGB */
+      targaheader[16] = 24;
+    }
+  }
+
+  if (JFWRITE(dinfo->output_file, targaheader, 18) != (size_t) 18)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ */
+
+METHODDEF(void)
+put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+		JDIMENSION rows_supplied)
+/* used for unquantized full-color output */
+{
+  tga_dest_ptr dest = (tga_dest_ptr) dinfo;
+  register JSAMPROW inptr;
+  register char * outptr;
+  register JDIMENSION col;
+
+  inptr = dest->pub.buffer[0];
+  outptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    outptr[0] = (char) GETJSAMPLE(inptr[2]); /* RGB to BGR order */
+    outptr[1] = (char) GETJSAMPLE(inptr[1]);
+    outptr[2] = (char) GETJSAMPLE(inptr[0]);
+    inptr += 3, outptr += 3;
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+METHODDEF(void)
+put_gray_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+	       JDIMENSION rows_supplied)
+/* used for grayscale OR quantized color output */
+{
+  tga_dest_ptr dest = (tga_dest_ptr) dinfo;
+  register JSAMPROW inptr;
+  register char * outptr;
+  register JDIMENSION col;
+
+  inptr = dest->pub.buffer[0];
+  outptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    *outptr++ = (char) GETJSAMPLE(*inptr++);
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Write some demapped pixel data when color quantization is in effect.
+ * For Targa, this is only applied to grayscale data.
+ */
+
+METHODDEF(void)
+put_demapped_gray (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+		   JDIMENSION rows_supplied)
+{
+  tga_dest_ptr dest = (tga_dest_ptr) dinfo;
+  register JSAMPROW inptr;
+  register char * outptr;
+  register JSAMPROW color_map0 = cinfo->colormap[0];
+  register JDIMENSION col;
+
+  inptr = dest->pub.buffer[0];
+  outptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    *outptr++ = (char) GETJSAMPLE(color_map0[GETJSAMPLE(*inptr++)]);
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Startup: write the file header.
+ */
+
+METHODDEF(void)
+start_output_tga (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  tga_dest_ptr dest = (tga_dest_ptr) dinfo;
+  int num_colors, i;
+  FILE *outfile;
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE) {
+    /* Targa doesn't have a mapped grayscale format, so we will */
+    /* demap quantized gray output.  Never emit a colormap. */
+    write_header(cinfo, dinfo, 0);
+    if (cinfo->quantize_colors)
+      dest->pub.put_pixel_rows = put_demapped_gray;
+    else
+      dest->pub.put_pixel_rows = put_gray_rows;
+  } else if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors) {
+      /* We only support 8-bit colormap indexes, so only 256 colors */
+      num_colors = cinfo->actual_number_of_colors;
+      if (num_colors > 256)
+	ERREXIT1(cinfo, JERR_TOO_MANY_COLORS, num_colors);
+      write_header(cinfo, dinfo, num_colors);
+      /* Write the colormap.  Note Targa uses BGR byte order */
+      outfile = dest->pub.output_file;
+      for (i = 0; i < num_colors; i++) {
+	putc(GETJSAMPLE(cinfo->colormap[2][i]), outfile);
+	putc(GETJSAMPLE(cinfo->colormap[1][i]), outfile);
+	putc(GETJSAMPLE(cinfo->colormap[0][i]), outfile);
+      }
+      dest->pub.put_pixel_rows = put_gray_rows;
+    } else {
+      write_header(cinfo, dinfo, 0);
+      dest->pub.put_pixel_rows = put_pixel_rows;
+    }
+  } else {
+    ERREXIT(cinfo, JERR_TGA_COLORSPACE);
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_output_tga (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  /* Make sure we wrote the output file OK */
+  fflush(dinfo->output_file);
+  if (ferror(dinfo->output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for Targa format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_targa (j_decompress_ptr cinfo)
+{
+  tga_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (tga_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(tga_dest_struct));
+  dest->pub.start_output = start_output_tga;
+  dest->pub.finish_output = finish_output_tga;
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Create I/O buffer.  Note we make this near on a PC. */
+  dest->buffer_width = cinfo->output_width * cinfo->output_components;
+  dest->iobuffer = (char *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(size_t) (dest->buffer_width * SIZEOF(char)));
+
+  /* Create decompressor output buffer. */
+  dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, dest->buffer_width, (JDIMENSION) 1);
+  dest->pub.buffer_height = 1;
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* TARGA_SUPPORTED */