blob: ee51449c7a25b93c44528fc1506a248594c12d14 (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
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.
|
