22 files changed, 476 insertions, 119 deletions

diff --git a/src/video_core/command_classes/codecs/codec.cpp b/src/video_core/host1x/codecs/codec.cpp
index a5eb97b7f..42e7d6e4f 100644
--- a/src/video_core/command_classes/codecs/codec.cpp
+++ b/src/video_core/host1x/codecs/codec.cpp
@@ -6,11 +6,11 @@
 #include <vector>
 #include "common/assert.h"
 #include "common/settings.h"
-#include "video_core/command_classes/codecs/codec.h"
-#include "video_core/command_classes/codecs/h264.h"
-#include "video_core/command_classes/codecs/vp8.h"
-#include "video_core/command_classes/codecs/vp9.h"
-#include "video_core/gpu.h"
+#include "video_core/host1x/codecs/codec.h"
+#include "video_core/host1x/codecs/h264.h"
+#include "video_core/host1x/codecs/vp8.h"
+#include "video_core/host1x/codecs/vp9.h"
+#include "video_core/host1x/host1x.h"
 #include "video_core/memory_manager.h"
 
 extern "C" {
@@ -73,10 +73,10 @@ void AVFrameDeleter(AVFrame* ptr) {
     av_frame_free(&ptr);
 }
 
-Codec::Codec(GPU& gpu_, const NvdecCommon::NvdecRegisters& regs)
-    : gpu(gpu_), state{regs}, h264_decoder(std::make_unique<Decoder::H264>(gpu)),
-      vp8_decoder(std::make_unique<Decoder::VP8>(gpu)),
-      vp9_decoder(std::make_unique<Decoder::VP9>(gpu)) {}
+Codec::Codec(Host1x::Host1x& host1x_, const Host1x::NvdecCommon::NvdecRegisters& regs)
+    : host1x(host1x_), state{regs}, h264_decoder(std::make_unique<Decoder::H264>(host1x)),
+      vp8_decoder(std::make_unique<Decoder::VP8>(host1x)),
+      vp9_decoder(std::make_unique<Decoder::VP9>(host1x)) {}
 
 Codec::~Codec() {
     if (!initialized) {
@@ -168,11 +168,11 @@ void Codec::InitializeGpuDecoder() {
 void Codec::Initialize() {
     const AVCodecID codec = [&] {
         switch (current_codec) {
-        case NvdecCommon::VideoCodec::H264:
+        case Host1x::NvdecCommon::VideoCodec::H264:
             return AV_CODEC_ID_H264;
-        case NvdecCommon::VideoCodec::VP8:
+        case Host1x::NvdecCommon::VideoCodec::VP8:
             return AV_CODEC_ID_VP8;
-        case NvdecCommon::VideoCodec::VP9:
+        case Host1x::NvdecCommon::VideoCodec::VP9:
             return AV_CODEC_ID_VP9;
         default:
             UNIMPLEMENTED_MSG("Unknown codec {}", current_codec);
@@ -197,7 +197,7 @@ void Codec::Initialize() {
     initialized = true;
 }
 
-void Codec::SetTargetCodec(NvdecCommon::VideoCodec codec) {
+void Codec::SetTargetCodec(Host1x::NvdecCommon::VideoCodec codec) {
     if (current_codec != codec) {
         current_codec = codec;
         LOG_INFO(Service_NVDRV, "NVDEC video codec initialized to {}", GetCurrentCodecName());
@@ -215,11 +215,11 @@ void Codec::Decode() {
     bool vp9_hidden_frame = false;
     const auto& frame_data = [&]() {
         switch (current_codec) {
-        case Tegra::NvdecCommon::VideoCodec::H264:
+        case Tegra::Host1x::NvdecCommon::VideoCodec::H264:
             return h264_decoder->ComposeFrame(state, is_first_frame);
-        case Tegra::NvdecCommon::VideoCodec::VP8:
+        case Tegra::Host1x::NvdecCommon::VideoCodec::VP8:
             return vp8_decoder->ComposeFrame(state);
-        case Tegra::NvdecCommon::VideoCodec::VP9:
+        case Tegra::Host1x::NvdecCommon::VideoCodec::VP9:
             vp9_decoder->ComposeFrame(state);
             vp9_hidden_frame = vp9_decoder->WasFrameHidden();
             return vp9_decoder->GetFrameBytes();
@@ -287,21 +287,21 @@ AVFramePtr Codec::GetCurrentFrame() {
     return frame;
 }
 
-NvdecCommon::VideoCodec Codec::GetCurrentCodec() const {
+Host1x::NvdecCommon::VideoCodec Codec::GetCurrentCodec() const {
     return current_codec;
 }
 
 std::string_view Codec::GetCurrentCodecName() const {
     switch (current_codec) {
-    case NvdecCommon::VideoCodec::None:
+    case Host1x::NvdecCommon::VideoCodec::None:
         return "None";
-    case NvdecCommon::VideoCodec::H264:
+    case Host1x::NvdecCommon::VideoCodec::H264:
         return "H264";
-    case NvdecCommon::VideoCodec::VP8:
+    case Host1x::NvdecCommon::VideoCodec::VP8:
         return "VP8";
-    case NvdecCommon::VideoCodec::H265:
+    case Host1x::NvdecCommon::VideoCodec::H265:
         return "H265";
-    case NvdecCommon::VideoCodec::VP9:
+    case Host1x::NvdecCommon::VideoCodec::VP9:
         return "VP9";
     default:
         return "Unknown";
diff --git a/src/video_core/command_classes/codecs/codec.h b/src/video_core/host1x/codecs/codec.h
index 0c2405465..0d45fb7fe 100644
--- a/src/video_core/command_classes/codecs/codec.h
+++ b/src/video_core/host1x/codecs/codec.h
@@ -6,8 +6,8 @@
 #include <memory>
 #include <string_view>
 #include <queue>
-
-#include "video_core/command_classes/nvdec_common.h"
+#include "common/common_types.h"
+#include "video_core/host1x/nvdec_common.h"
 
 extern "C" {
 #if defined(__GNUC__) || defined(__clang__)
@@ -21,7 +21,6 @@ extern "C" {
 }
 
 namespace Tegra {
-class GPU;
 
 void AVFrameDeleter(AVFrame* ptr);
 using AVFramePtr = std::unique_ptr<AVFrame, decltype(&AVFrameDeleter)>;
@@ -32,16 +31,20 @@ class VP8;
 class VP9;
 } // namespace Decoder
 
+namespace Host1x {
+class Host1x;
+} // namespace Host1x
+
 class Codec {
 public:
-    explicit Codec(GPU& gpu, const NvdecCommon::NvdecRegisters& regs);
+    explicit Codec(Host1x::Host1x& host1x, const Host1x::NvdecCommon::NvdecRegisters& regs);
     ~Codec();
 
     /// Initialize the codec, returning success or failure
     void Initialize();
 
     /// Sets NVDEC video stream codec
-    void SetTargetCodec(NvdecCommon::VideoCodec codec);
+    void SetTargetCodec(Host1x::NvdecCommon::VideoCodec codec);
 
     /// Call decoders to construct headers, decode AVFrame with ffmpeg
     void Decode();
@@ -50,7 +53,7 @@ public:
     [[nodiscard]] AVFramePtr GetCurrentFrame();
 
     /// Returns the value of current_codec
-    [[nodiscard]] NvdecCommon::VideoCodec GetCurrentCodec() const;
+    [[nodiscard]] Host1x::NvdecCommon::VideoCodec GetCurrentCodec() const;
 
     /// Return name of the current codec
     [[nodiscard]] std::string_view GetCurrentCodecName() const;
@@ -63,14 +66,14 @@ private:
     bool CreateGpuAvDevice();
 
     bool initialized{};
-    NvdecCommon::VideoCodec current_codec{NvdecCommon::VideoCodec::None};
+    Host1x::NvdecCommon::VideoCodec current_codec{Host1x::NvdecCommon::VideoCodec::None};
 
     const AVCodec* av_codec{nullptr};
     AVCodecContext* av_codec_ctx{nullptr};
     AVBufferRef* av_gpu_decoder{nullptr};
 
-    GPU& gpu;
-    const NvdecCommon::NvdecRegisters& state;
+    Host1x::Host1x& host1x;
+    const Host1x::NvdecCommon::NvdecRegisters& state;
     std::unique_ptr<Decoder::H264> h264_decoder;
     std::unique_ptr<Decoder::VP8> vp8_decoder;
     std::unique_ptr<Decoder::VP9> vp9_decoder;
diff --git a/src/video_core/command_classes/codecs/h264.cpp b/src/video_core/host1x/codecs/h264.cpp
index e2acd54d4..e87bd65fa 100644
--- a/src/video_core/command_classes/codecs/h264.cpp
+++ b/src/video_core/host1x/codecs/h264.cpp
@@ -5,8 +5,8 @@
 #include <bit>
 
 #include "common/settings.h"
-#include "video_core/command_classes/codecs/h264.h"
-#include "video_core/gpu.h"
+#include "video_core/host1x/codecs/h264.h"
+#include "video_core/host1x/host1x.h"
 #include "video_core/memory_manager.h"
 
 namespace Tegra::Decoder {
@@ -24,19 +24,20 @@ constexpr std::array<u8, 16> zig_zag_scan{
 };
 } // Anonymous namespace
 
-H264::H264(GPU& gpu_) : gpu(gpu_) {}
+H264::H264(Host1x::Host1x& host1x_) : host1x{host1x_} {}
 
 H264::~H264() = default;
 
-const std::vector<u8>& H264::ComposeFrame(const NvdecCommon::NvdecRegisters& state,
+const std::vector<u8>& H264::ComposeFrame(const Host1x::NvdecCommon::NvdecRegisters& state,
                                           bool is_first_frame) {
     H264DecoderContext context;
-    gpu.MemoryManager().ReadBlock(state.picture_info_offset, &context, sizeof(H264DecoderContext));
+    host1x.MemoryManager().ReadBlock(state.picture_info_offset, &context,
+                                     sizeof(H264DecoderContext));
 
     const s64 frame_number = context.h264_parameter_set.frame_number.Value();
     if (!is_first_frame && frame_number != 0) {
         frame.resize(context.stream_len);
-        gpu.MemoryManager().ReadBlock(state.frame_bitstream_offset, frame.data(), frame.size());
+        host1x.MemoryManager().ReadBlock(state.frame_bitstream_offset, frame.data(), frame.size());
         return frame;
     }
 
@@ -155,8 +156,8 @@ const std::vector<u8>& H264::ComposeFrame(const NvdecCommon::NvdecRegisters& sta
     frame.resize(encoded_header.size() + context.stream_len);
     std::memcpy(frame.data(), encoded_header.data(), encoded_header.size());
 
-    gpu.MemoryManager().ReadBlock(state.frame_bitstream_offset,
-                                  frame.data() + encoded_header.size(), context.stream_len);
+    host1x.MemoryManager().ReadBlock(state.frame_bitstream_offset,
+                                     frame.data() + encoded_header.size(), context.stream_len);
 
     return frame;
 }
diff --git a/src/video_core/command_classes/codecs/h264.h b/src/video_core/host1x/codecs/h264.h
index 261574364..5cc86454e 100644
--- a/src/video_core/command_classes/codecs/h264.h
+++ b/src/video_core/host1x/codecs/h264.h
@@ -8,10 +8,14 @@
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
-#include "video_core/command_classes/nvdec_common.h"
+#include "video_core/host1x/nvdec_common.h"
 
 namespace Tegra {
-class GPU;
+
+namespace Host1x {
+class Host1x;
+} // namespace Host1x
+
 namespace Decoder {
 
 class H264BitWriter {
@@ -55,16 +59,16 @@ private:
 
 class H264 {
 public:
-    explicit H264(GPU& gpu);
+    explicit H264(Host1x::Host1x& host1x);
     ~H264();
 
     /// Compose the H264 frame for FFmpeg decoding
-    [[nodiscard]] const std::vector<u8>& ComposeFrame(const NvdecCommon::NvdecRegisters& state,
-                                                      bool is_first_frame = false);
+    [[nodiscard]] const std::vector<u8>& ComposeFrame(
+        const Host1x::NvdecCommon::NvdecRegisters& state, bool is_first_frame = false);
 
 private:
     std::vector<u8> frame;
-    GPU& gpu;
+    Host1x::Host1x& host1x;
 
     struct H264ParameterSet {
         s32 log2_max_pic_order_cnt_lsb_minus4; ///< 0x00
diff --git a/src/video_core/command_classes/codecs/vp8.cpp b/src/video_core/host1x/codecs/vp8.cpp
index c83b9bbc2..28fb12cb8 100644
--- a/src/video_core/command_classes/codecs/vp8.cpp
+++ b/src/video_core/host1x/codecs/vp8.cpp
@@ -3,18 +3,18 @@
 
 #include <vector>
 
-#include "video_core/command_classes/codecs/vp8.h"
-#include "video_core/gpu.h"
+#include "video_core/host1x/codecs/vp8.h"
+#include "video_core/host1x/host1x.h"
 #include "video_core/memory_manager.h"
 
 namespace Tegra::Decoder {
-VP8::VP8(GPU& gpu_) : gpu(gpu_) {}
+VP8::VP8(Host1x::Host1x& host1x_) : host1x{host1x_} {}
 
 VP8::~VP8() = default;
 
-const std::vector<u8>& VP8::ComposeFrame(const NvdecCommon::NvdecRegisters& state) {
+const std::vector<u8>& VP8::ComposeFrame(const Host1x::NvdecCommon::NvdecRegisters& state) {
     VP8PictureInfo info;
-    gpu.MemoryManager().ReadBlock(state.picture_info_offset, &info, sizeof(VP8PictureInfo));
+    host1x.MemoryManager().ReadBlock(state.picture_info_offset, &info, sizeof(VP8PictureInfo));
 
     const bool is_key_frame = info.key_frame == 1u;
     const auto bitstream_size = static_cast<size_t>(info.vld_buffer_size);
@@ -45,7 +45,7 @@ const std::vector<u8>& VP8::ComposeFrame(const NvdecCommon::NvdecRegisters& stat
         frame[9] = static_cast<u8>(((info.frame_height >> 8) & 0x3f));
     }
     const u64 bitstream_offset = state.frame_bitstream_offset;
-    gpu.MemoryManager().ReadBlock(bitstream_offset, frame.data() + header_size, bitstream_size);
+    host1x.MemoryManager().ReadBlock(bitstream_offset, frame.data() + header_size, bitstream_size);
 
     return frame;
 }
diff --git a/src/video_core/command_classes/codecs/vp8.h b/src/video_core/host1x/codecs/vp8.h
index 3357667b0..5bf07ecab 100644
--- a/src/video_core/command_classes/codecs/vp8.h
+++ b/src/video_core/host1x/codecs/vp8.h
@@ -8,23 +8,28 @@
 
 #include "common/common_funcs.h"
 #include "common/common_types.h"
-#include "video_core/command_classes/nvdec_common.h"
+#include "video_core/host1x/nvdec_common.h"
 
 namespace Tegra {
-class GPU;
+
+namespace Host1x {
+class Host1x;
+} // namespace Host1x
+
 namespace Decoder {
 
 class VP8 {
 public:
-    explicit VP8(GPU& gpu);
+    explicit VP8(Host1x::Host1x& host1x);
     ~VP8();
 
     /// Compose the VP8 frame for FFmpeg decoding
-    [[nodiscard]] const std::vector<u8>& ComposeFrame(const NvdecCommon::NvdecRegisters& state);
+    [[nodiscard]] const std::vector<u8>& ComposeFrame(
+        const Host1x::NvdecCommon::NvdecRegisters& state);
 
 private:
     std::vector<u8> frame;
-    GPU& gpu;
+    Host1x::Host1x& host1x;
 
     struct VP8PictureInfo {
         INSERT_PADDING_WORDS_NOINIT(14);
diff --git a/src/video_core/command_classes/codecs/vp9.cpp b/src/video_core/host1x/codecs/vp9.cpp
index c01431441..cf40c9012 100644
--- a/src/video_core/command_classes/codecs/vp9.cpp
+++ b/src/video_core/host1x/codecs/vp9.cpp
@@ -4,8 +4,8 @@
 #include <algorithm> // for std::copy
 #include <numeric>
 #include "common/assert.h"
-#include "video_core/command_classes/codecs/vp9.h"
-#include "video_core/gpu.h"
+#include "video_core/host1x/codecs/vp9.h"
+#include "video_core/host1x/host1x.h"
 #include "video_core/memory_manager.h"
 
 namespace Tegra::Decoder {
@@ -236,7 +236,7 @@ constexpr std::array<u8, 254> map_lut{
 }
 } // Anonymous namespace
 
-VP9::VP9(GPU& gpu_) : gpu{gpu_} {}
+VP9::VP9(Host1x::Host1x& host1x_) : host1x{host1x_} {}
 
 VP9::~VP9() = default;
 
@@ -355,9 +355,9 @@ void VP9::WriteMvProbabilityUpdate(VpxRangeEncoder& writer, u8 new_prob, u8 old_
     }
 }
 
-Vp9PictureInfo VP9::GetVp9PictureInfo(const NvdecCommon::NvdecRegisters& state) {
+Vp9PictureInfo VP9::GetVp9PictureInfo(const Host1x::NvdecCommon::NvdecRegisters& state) {
     PictureInfo picture_info;
-    gpu.MemoryManager().ReadBlock(state.picture_info_offset, &picture_info, sizeof(PictureInfo));
+    host1x.MemoryManager().ReadBlock(state.picture_info_offset, &picture_info, sizeof(PictureInfo));
     Vp9PictureInfo vp9_info = picture_info.Convert();
 
     InsertEntropy(state.vp9_entropy_probs_offset, vp9_info.entropy);
@@ -372,18 +372,19 @@ Vp9PictureInfo VP9::GetVp9PictureInfo(const NvdecCommon::NvdecRegisters& state)
 
 void VP9::InsertEntropy(u64 offset, Vp9EntropyProbs& dst) {
     EntropyProbs entropy;
-    gpu.MemoryManager().ReadBlock(offset, &entropy, sizeof(EntropyProbs));
+    host1x.MemoryManager().ReadBlock(offset, &entropy, sizeof(EntropyProbs));
     entropy.Convert(dst);
 }
 
-Vp9FrameContainer VP9::GetCurrentFrame(const NvdecCommon::NvdecRegisters& state) {
+Vp9FrameContainer VP9::GetCurrentFrame(const Host1x::NvdecCommon::NvdecRegisters& state) {
     Vp9FrameContainer current_frame{};
     {
-        gpu.SyncGuestHost();
+        // gpu.SyncGuestHost(); epic, why?
         current_frame.info = GetVp9PictureInfo(state);
         current_frame.bit_stream.resize(current_frame.info.bitstream_size);
-        gpu.MemoryManager().ReadBlock(state.frame_bitstream_offset, current_frame.bit_stream.data(),
-                                      current_frame.info.bitstream_size);
+        host1x.MemoryManager().ReadBlock(state.frame_bitstream_offset,
+                                         current_frame.bit_stream.data(),
+                                         current_frame.info.bitstream_size);
     }
     if (!next_frame.bit_stream.empty()) {
         Vp9FrameContainer temp{
@@ -769,7 +770,7 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
     return uncomp_writer;
 }
 
-void VP9::ComposeFrame(const NvdecCommon::NvdecRegisters& state) {
+void VP9::ComposeFrame(const Host1x::NvdecCommon::NvdecRegisters& state) {
     std::vector<u8> bitstream;
     {
         Vp9FrameContainer curr_frame = GetCurrentFrame(state);
diff --git a/src/video_core/command_classes/codecs/vp9.h b/src/video_core/host1x/codecs/vp9.h
index ecc40e8b1..d4083e8d3 100644
--- a/src/video_core/command_classes/codecs/vp9.h
+++ b/src/video_core/host1x/codecs/vp9.h
@@ -8,11 +8,15 @@
 
 #include "common/common_types.h"
 #include "common/stream.h"
-#include "video_core/command_classes/codecs/vp9_types.h"
-#include "video_core/command_classes/nvdec_common.h"
+#include "video_core/host1x/codecs/vp9_types.h"
+#include "video_core/host1x/nvdec_common.h"
 
 namespace Tegra {
-class GPU;
+
+namespace Host1x {
+class Host1x;
+} // namespace Host1x
+
 namespace Decoder {
 
 /// The VpxRangeEncoder, and VpxBitStreamWriter classes are used to compose the
@@ -106,7 +110,7 @@ private:
 
 class VP9 {
 public:
-    explicit VP9(GPU& gpu_);
+    explicit VP9(Host1x::Host1x& host1x);
     ~VP9();
 
     VP9(const VP9&) = delete;
@@ -117,7 +121,7 @@ public:
 
     /// Composes the VP9 frame from the GPU state information.
     /// Based on the official VP9 spec documentation
-    void ComposeFrame(const NvdecCommon::NvdecRegisters& state);
+    void ComposeFrame(const Host1x::NvdecCommon::NvdecRegisters& state);
 
     /// Returns true if the most recent frame was a hidden frame.
     [[nodiscard]] bool WasFrameHidden() const {
@@ -162,19 +166,21 @@ private:
     void WriteMvProbabilityUpdate(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob);
 
     /// Returns VP9 information from NVDEC provided offset and size
-    [[nodiscard]] Vp9PictureInfo GetVp9PictureInfo(const NvdecCommon::NvdecRegisters& state);
+    [[nodiscard]] Vp9PictureInfo GetVp9PictureInfo(
+        const Host1x::NvdecCommon::NvdecRegisters& state);
 
     /// Read and convert NVDEC provided entropy probs to Vp9EntropyProbs struct
     void InsertEntropy(u64 offset, Vp9EntropyProbs& dst);
 
     /// Returns frame to be decoded after buffering
-    [[nodiscard]] Vp9FrameContainer GetCurrentFrame(const NvdecCommon::NvdecRegisters& state);
+    [[nodiscard]] Vp9FrameContainer GetCurrentFrame(
+        const Host1x::NvdecCommon::NvdecRegisters& state);
 
     /// Use NVDEC providied information to compose the headers for the current frame
     [[nodiscard]] std::vector<u8> ComposeCompressedHeader();
     [[nodiscard]] VpxBitStreamWriter ComposeUncompressedHeader();
 
-    GPU& gpu;
+    Host1x::Host1x& host1x;
     std::vector<u8> frame;
 
     std::array<s8, 4> loop_filter_ref_deltas{};
diff --git a/src/video_core/command_classes/codecs/vp9_types.h b/src/video_core/host1x/codecs/vp9_types.h
index bb3d8df6e..adad8ed7e 100644
--- a/src/video_core/command_classes/codecs/vp9_types.h
+++ b/src/video_core/host1x/codecs/vp9_types.h
@@ -9,7 +9,6 @@
 #include "common/common_types.h"
 
 namespace Tegra {
-class GPU;
 
 namespace Decoder {
 struct Vp9FrameDimensions {
diff --git a/src/video_core/host1x/control.cpp b/src/video_core/host1x/control.cpp
new file mode 100644
index 000000000..dceefdb7f
--- /dev/null
+++ b/src/video_core/host1x/control.cpp
@@ -0,0 +1,33 @@
+// SPDX-FileCopyrightText: 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#include "common/assert.h"
+#include "video_core/host1x/control.h"
+#include "video_core/host1x/host1x.h"
+
+namespace Tegra::Host1x {
+
+Control::Control(Host1x& host1x_) : host1x(host1x_) {}
+
+Control::~Control() = default;
+
+void Control::ProcessMethod(Method method, u32 argument) {
+    switch (method) {
+    case Method::LoadSyncptPayload32:
+        syncpoint_value = argument;
+        break;
+    case Method::WaitSyncpt:
+    case Method::WaitSyncpt32:
+        Execute(argument);
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Control method 0x{:X}", static_cast<u32>(method));
+        break;
+    }
+}
+
+void Control::Execute(u32 data) {
+    host1x.GetSyncpointManager().WaitHost(data, syncpoint_value);
+}
+
+} // namespace Tegra::Host1x
diff --git a/src/video_core/command_classes/host1x.h b/src/video_core/host1x/control.h
index bb48a4381..e117888a3 100644
--- a/src/video_core/command_classes/host1x.h
+++ b/src/video_core/host1x/control.h
@@ -1,15 +1,19 @@
-// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
+// SPDX-FileCopyrightText: 2021 yuzu Emulator Project
+// SPDX-FileCopyrightText: 2021 Skyline Team and Contributors
+// SPDX-License-Identifier: GPL-3.0-or-later
 
 #pragma once
 
 #include "common/common_types.h"
 
 namespace Tegra {
-class GPU;
+
+namespace Host1x {
+
+class Host1x;
 class Nvdec;
 
-class Host1x {
+class Control {
 public:
     enum class Method : u32 {
         WaitSyncpt = 0x8,
@@ -17,8 +21,8 @@ public:
         WaitSyncpt32 = 0x50,
     };
 
-    explicit Host1x(GPU& gpu);
-    ~Host1x();
+    explicit Control(Host1x& host1x);
+    ~Control();
 
     /// Writes the method into the state, Invoke Execute() if encountered
     void ProcessMethod(Method method, u32 argument);
@@ -28,7 +32,9 @@ private:
     void Execute(u32 data);
 
     u32 syncpoint_value{};
-    GPU& gpu;
+    Host1x& host1x;
 };
 
+} // namespace Host1x
+
 } // namespace Tegra
diff --git a/src/video_core/host1x/host1x.cpp b/src/video_core/host1x/host1x.cpp
new file mode 100644
index 000000000..7c317a85d
--- /dev/null
+++ b/src/video_core/host1x/host1x.cpp
@@ -0,0 +1,17 @@
+// SPDX-FileCopyrightText: 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#include "core/core.h"
+#include "video_core/host1x/host1x.h"
+
+namespace Tegra {
+
+namespace Host1x {
+
+Host1x::Host1x(Core::System& system_)
+    : system{system_}, syncpoint_manager{}, memory_manager{system, 32, 12},
+      allocator{std::make_unique<Common::FlatAllocator<u32, 0, 32>>(1 << 12)} {}
+
+} // namespace Host1x
+
+} // namespace Tegra
diff --git a/src/video_core/host1x/host1x.h b/src/video_core/host1x/host1x.h
new file mode 100644
index 000000000..57082ae54
--- /dev/null
+++ b/src/video_core/host1x/host1x.h
@@ -0,0 +1,57 @@
+// SPDX-FileCopyrightText: 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+#include "common/common_types.h"
+
+#include "common/address_space.h"
+#include "video_core/host1x/syncpoint_manager.h"
+#include "video_core/memory_manager.h"
+
+namespace Core {
+class System;
+} // namespace Core
+
+namespace Tegra {
+
+namespace Host1x {
+
+class Host1x {
+public:
+    explicit Host1x(Core::System& system);
+
+    SyncpointManager& GetSyncpointManager() {
+        return syncpoint_manager;
+    }
+
+    const SyncpointManager& GetSyncpointManager() const {
+        return syncpoint_manager;
+    }
+
+    Tegra::MemoryManager& MemoryManager() {
+        return memory_manager;
+    }
+
+    const Tegra::MemoryManager& MemoryManager() const {
+        return memory_manager;
+    }
+
+    Common::FlatAllocator<u32, 0, 32>& Allocator() {
+        return *allocator;
+    }
+
+    const Common::FlatAllocator<u32, 0, 32>& Allocator() const {
+        return *allocator;
+    }
+
+private:
+    Core::System& system;
+    SyncpointManager syncpoint_manager;
+    Tegra::MemoryManager memory_manager;
+    std::unique_ptr<Common::FlatAllocator<u32, 0, 32>> allocator;
+};
+
+} // namespace Host1x
+
+} // namespace Tegra
diff --git a/src/video_core/command_classes/nvdec.cpp b/src/video_core/host1x/nvdec.cpp
index 4fbbe3da6..a4bd5b79f 100644
--- a/src/video_core/command_classes/nvdec.cpp
+++ b/src/video_core/host1x/nvdec.cpp
@@ -2,15 +2,16 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include "common/assert.h"
-#include "video_core/command_classes/nvdec.h"
-#include "video_core/gpu.h"
+#include "video_core/host1x/host1x.h"
+#include "video_core/host1x/nvdec.h"
 
-namespace Tegra {
+namespace Tegra::Host1x {
 
 #define NVDEC_REG_INDEX(field_name)                                                                \
     (offsetof(NvdecCommon::NvdecRegisters, field_name) / sizeof(u64))
 
-Nvdec::Nvdec(GPU& gpu_) : gpu(gpu_), state{}, codec(std::make_unique<Codec>(gpu, state)) {}
+Nvdec::Nvdec(Host1x& host1x_)
+    : host1x(host1x_), state{}, codec(std::make_unique<Codec>(host1x, state)) {}
 
 Nvdec::~Nvdec() = default;
 
@@ -44,4 +45,4 @@ void Nvdec::Execute() {
     }
 }
 
-} // namespace Tegra
+} // namespace Tegra::Host1x
diff --git a/src/video_core/command_classes/nvdec.h b/src/video_core/host1x/nvdec.h
index 488531fc6..3949d5181 100644
--- a/src/video_core/command_classes/nvdec.h
+++ b/src/video_core/host1x/nvdec.h
@@ -6,14 +6,17 @@
 #include <memory>
 #include <vector>
 #include "common/common_types.h"
-#include "video_core/command_classes/codecs/codec.h"
+#include "video_core/host1x/codecs/codec.h"
 
 namespace Tegra {
-class GPU;
+
+namespace Host1x {
+
+class Host1x;
 
 class Nvdec {
 public:
-    explicit Nvdec(GPU& gpu);
+    explicit Nvdec(Host1x& host1x);
     ~Nvdec();
 
     /// Writes the method into the state, Invoke Execute() if encountered
@@ -26,8 +29,11 @@ private:
     /// Invoke codec to decode a frame
     void Execute();
 
-    GPU& gpu;
+    Host1x& host1x;
     NvdecCommon::NvdecRegisters state;
     std::unique_ptr<Codec> codec;
 };
+
+} // namespace Host1x
+
 } // namespace Tegra
diff --git a/src/video_core/command_classes/nvdec_common.h b/src/video_core/host1x/nvdec_common.h
index 521e5b52b..49d67ebbe 100644
--- a/src/video_core/command_classes/nvdec_common.h
+++ b/src/video_core/host1x/nvdec_common.h
@@ -7,7 +7,7 @@
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 
-namespace Tegra::NvdecCommon {
+namespace Tegra::Host1x::NvdecCommon {
 
 enum class VideoCodec : u64 {
     None = 0x0,
@@ -94,4 +94,4 @@ ASSERT_REG_POSITION(vp9_curr_frame_mvs_offset, 0x176);
 
 #undef ASSERT_REG_POSITION
 
-} // namespace Tegra::NvdecCommon
+} // namespace Tegra::Host1x::NvdecCommon
diff --git a/src/video_core/command_classes/sync_manager.cpp b/src/video_core/host1x/sync_manager.cpp
index 67e58046f..5ef9ea217 100644
--- a/src/video_core/command_classes/sync_manager.cpp
+++ b/src/video_core/host1x/sync_manager.cpp
@@ -3,10 +3,13 @@
 
 #include <algorithm>
 #include "sync_manager.h"
-#include "video_core/gpu.h"
+#include "video_core/host1x/host1x.h"
+#include "video_core/host1x/syncpoint_manager.h"
 
 namespace Tegra {
-SyncptIncrManager::SyncptIncrManager(GPU& gpu_) : gpu(gpu_) {}
+namespace Host1x {
+
+SyncptIncrManager::SyncptIncrManager(Host1x& host1x_) : host1x(host1x_) {}
 SyncptIncrManager::~SyncptIncrManager() = default;
 
 void SyncptIncrManager::Increment(u32 id) {
@@ -36,8 +39,12 @@ void SyncptIncrManager::IncrementAllDone() {
         if (!increments[done_count].complete) {
             break;
         }
-        gpu.IncrementSyncPoint(increments[done_count].syncpt_id);
+        auto& syncpoint_manager = host1x.GetSyncpointManager();
+        syncpoint_manager.IncrementGuest(increments[done_count].syncpt_id);
+        syncpoint_manager.IncrementHost(increments[done_count].syncpt_id);
     }
     increments.erase(increments.begin(), increments.begin() + done_count);
 }
+
+} // namespace Host1x
 } // namespace Tegra
diff --git a/src/video_core/command_classes/sync_manager.h b/src/video_core/host1x/sync_manager.h
index 6dfaae080..7bb77fa27 100644
--- a/src/video_core/command_classes/sync_manager.h
+++ b/src/video_core/host1x/sync_manager.h
@@ -8,7 +8,11 @@
 #include "common/common_types.h"
 
 namespace Tegra {
-class GPU;
+
+namespace Host1x {
+
+class Host1x;
+
 struct SyncptIncr {
     u32 id;
     u32 class_id;
@@ -21,7 +25,7 @@ struct SyncptIncr {
 
 class SyncptIncrManager {
 public:
-    explicit SyncptIncrManager(GPU& gpu);
+    explicit SyncptIncrManager(Host1x& host1x);
     ~SyncptIncrManager();
 
     /// Add syncpoint id and increment all
@@ -41,7 +45,9 @@ private:
     std::mutex increment_lock;
     u32 current_id{};
 
-    GPU& gpu;
+    Host1x& host1x;
 };
 
+} // namespace Host1x
+
 } // namespace Tegra
diff --git a/src/video_core/host1x/syncpoint_manager.cpp b/src/video_core/host1x/syncpoint_manager.cpp
new file mode 100644
index 000000000..326e8355a
--- /dev/null
+++ b/src/video_core/host1x/syncpoint_manager.cpp
@@ -0,0 +1,96 @@
+// SPDX-FileCopyrightText: 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#include "common/microprofile.h"
+#include "video_core/host1x/syncpoint_manager.h"
+
+namespace Tegra {
+
+namespace Host1x {
+
+MICROPROFILE_DEFINE(GPU_wait, "GPU", "Wait for the GPU", MP_RGB(128, 128, 192));
+
+SyncpointManager::ActionHandle SyncpointManager::RegisterAction(
+    std::atomic<u32>& syncpoint, std::list<RegisteredAction>& action_storage, u32 expected_value,
+    std::function<void()>&& action) {
+    if (syncpoint.load(std::memory_order_acquire) >= expected_value) {
+        action();
+        return {};
+    }
+
+    std::unique_lock lk(guard);
+    if (syncpoint.load(std::memory_order_relaxed) >= expected_value) {
+        action();
+        return {};
+    }
+    auto it = action_storage.begin();
+    while (it != action_storage.end()) {
+        if (it->expected_value >= expected_value) {
+            break;
+        }
+        ++it;
+    }
+    return action_storage.emplace(it, expected_value, std::move(action));
+}
+
+void SyncpointManager::DeregisterAction(std::list<RegisteredAction>& action_storage,
+                                        ActionHandle& handle) {
+    std::unique_lock lk(guard);
+    action_storage.erase(handle);
+}
+
+void SyncpointManager::DeregisterGuestAction(u32 syncpoint_id, ActionHandle& handle) {
+    DeregisterAction(guest_action_storage[syncpoint_id], handle);
+}
+
+void SyncpointManager::DeregisterHostAction(u32 syncpoint_id, ActionHandle& handle) {
+    DeregisterAction(host_action_storage[syncpoint_id], handle);
+}
+
+void SyncpointManager::IncrementGuest(u32 syncpoint_id) {
+    Increment(syncpoints_guest[syncpoint_id], wait_guest_cv, guest_action_storage[syncpoint_id]);
+}
+
+void SyncpointManager::IncrementHost(u32 syncpoint_id) {
+    Increment(syncpoints_host[syncpoint_id], wait_host_cv, host_action_storage[syncpoint_id]);
+}
+
+void SyncpointManager::WaitGuest(u32 syncpoint_id, u32 expected_value) {
+    Wait(syncpoints_guest[syncpoint_id], wait_guest_cv, expected_value);
+}
+
+void SyncpointManager::WaitHost(u32 syncpoint_id, u32 expected_value) {
+    MICROPROFILE_SCOPE(GPU_wait);
+    Wait(syncpoints_host[syncpoint_id], wait_host_cv, expected_value);
+}
+
+void SyncpointManager::Increment(std::atomic<u32>& syncpoint, std::condition_variable& wait_cv,
+                                 std::list<RegisteredAction>& action_storage) {
+    auto new_value{syncpoint.fetch_add(1, std::memory_order_acq_rel) + 1};
+
+    std::unique_lock lk(guard);
+    auto it = action_storage.begin();
+    while (it != action_storage.end()) {
+        if (it->expected_value > new_value) {
+            break;
+        }
+        it->action();
+        it = action_storage.erase(it);
+    }
+    wait_cv.notify_all();
+}
+
+void SyncpointManager::Wait(std::atomic<u32>& syncpoint, std::condition_variable& wait_cv,
+                            u32 expected_value) {
+    const auto pred = [&]() { return syncpoint.load(std::memory_order_acquire) >= expected_value; };
+    if (pred()) {
+        return;
+    }
+
+    std::unique_lock lk(guard);
+    wait_cv.wait(lk, pred);
+}
+
+} // namespace Host1x
+
+} // namespace Tegra
diff --git a/src/video_core/host1x/syncpoint_manager.h b/src/video_core/host1x/syncpoint_manager.h
new file mode 100644
index 000000000..50a264e23
--- /dev/null
+++ b/src/video_core/host1x/syncpoint_manager.h
@@ -0,0 +1,98 @@
+// SPDX-FileCopyrightText: 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+#include <array>
+#include <atomic>
+#include <condition_variable>
+#include <functional>
+#include <list>
+#include <mutex>
+
+#include "common/common_types.h"
+
+namespace Tegra {
+
+namespace Host1x {
+
+class SyncpointManager {
+public:
+    u32 GetGuestSyncpointValue(u32 id) const {
+        return syncpoints_guest[id].load(std::memory_order_acquire);
+    }
+
+    u32 GetHostSyncpointValue(u32 id) const {
+        return syncpoints_host[id].load(std::memory_order_acquire);
+    }
+
+    struct RegisteredAction {
+        explicit RegisteredAction(u32 expected_value_, std::function<void()>&& action_)
+            : expected_value{expected_value_}, action{std::move(action_)} {}
+        u32 expected_value;
+        std::function<void()> action;
+    };
+    using ActionHandle = std::list<RegisteredAction>::iterator;
+
+    template <typename Func>
+    ActionHandle RegisterGuestAction(u32 syncpoint_id, u32 expected_value, Func&& action) {
+        std::function<void()> func(action);
+        return RegisterAction(syncpoints_guest[syncpoint_id], guest_action_storage[syncpoint_id],
+                              expected_value, std::move(func));
+    }
+
+    template <typename Func>
+    ActionHandle RegisterHostAction(u32 syncpoint_id, u32 expected_value, Func&& action) {
+        std::function<void()> func(action);
+        return RegisterAction(syncpoints_host[syncpoint_id], host_action_storage[syncpoint_id],
+                              expected_value, std::move(func));
+    }
+
+    void DeregisterGuestAction(u32 syncpoint_id, ActionHandle& handle);
+
+    void DeregisterHostAction(u32 syncpoint_id, ActionHandle& handle);
+
+    void IncrementGuest(u32 syncpoint_id);
+
+    void IncrementHost(u32 syncpoint_id);
+
+    void WaitGuest(u32 syncpoint_id, u32 expected_value);
+
+    void WaitHost(u32 syncpoint_id, u32 expected_value);
+
+    bool IsReadyGuest(u32 syncpoint_id, u32 expected_value) const {
+        return syncpoints_guest[syncpoint_id].load(std::memory_order_acquire) >= expected_value;
+    }
+
+    bool IsReadyHost(u32 syncpoint_id, u32 expected_value) const {
+        return syncpoints_host[syncpoint_id].load(std::memory_order_acquire) >= expected_value;
+    }
+
+private:
+    void Increment(std::atomic<u32>& syncpoint, std::condition_variable& wait_cv,
+                   std::list<RegisteredAction>& action_storage);
+
+    ActionHandle RegisterAction(std::atomic<u32>& syncpoint,
+                                std::list<RegisteredAction>& action_storage, u32 expected_value,
+                                std::function<void()>&& action);
+
+    void DeregisterAction(std::list<RegisteredAction>& action_storage, ActionHandle& handle);
+
+    void Wait(std::atomic<u32>& syncpoint, std::condition_variable& wait_cv, u32 expected_value);
+
+    static constexpr size_t NUM_MAX_SYNCPOINTS = 192;
+
+    std::array<std::atomic<u32>, NUM_MAX_SYNCPOINTS> syncpoints_guest{};
+    std::array<std::atomic<u32>, NUM_MAX_SYNCPOINTS> syncpoints_host{};
+
+    std::array<std::list<RegisteredAction>, NUM_MAX_SYNCPOINTS> guest_action_storage;
+    std::array<std::list<RegisteredAction>, NUM_MAX_SYNCPOINTS> host_action_storage;
+
+    std::mutex guard;
+    std::condition_variable wait_guest_cv;
+    std::condition_variable wait_host_cv;
+};
+
+} // namespace Host1x
+
+} // namespace Tegra
diff --git a/src/video_core/command_classes/vic.cpp b/src/video_core/host1x/vic.cpp
index 7c17df353..ac0b7d20e 100644
--- a/src/video_core/command_classes/vic.cpp
+++ b/src/video_core/host1x/vic.cpp
@@ -18,14 +18,17 @@ extern "C" {
 #include "common/bit_field.h"
 #include "common/logging/log.h"
 
-#include "video_core/command_classes/nvdec.h"
-#include "video_core/command_classes/vic.h"
 #include "video_core/engines/maxwell_3d.h"
-#include "video_core/gpu.h"
+#include "video_core/host1x/host1x.h"
+#include "video_core/host1x/nvdec.h"
+#include "video_core/host1x/vic.h"
 #include "video_core/memory_manager.h"
 #include "video_core/textures/decoders.h"
 
 namespace Tegra {
+
+namespace Host1x {
+
 namespace {
 enum class VideoPixelFormat : u64_le {
     RGBA8 = 0x1f,
@@ -46,8 +49,8 @@ union VicConfig {
     BitField<46, 14, u64_le> surface_height_minus1;
 };
 
-Vic::Vic(GPU& gpu_, std::shared_ptr<Nvdec> nvdec_processor_)
-    : gpu(gpu_),
+Vic::Vic(Host1x& host1x_, std::shared_ptr<Nvdec> nvdec_processor_)
+    : host1x(host1x_),
       nvdec_processor(std::move(nvdec_processor_)), converted_frame_buffer{nullptr, av_free} {}
 
 Vic::~Vic() = default;
@@ -78,7 +81,7 @@ void Vic::Execute() {
         LOG_ERROR(Service_NVDRV, "VIC Luma address not set.");
         return;
     }
-    const VicConfig config{gpu.MemoryManager().Read<u64>(config_struct_address + 0x20)};
+    const VicConfig config{host1x.MemoryManager().Read<u64>(config_struct_address + 0x20)};
     const AVFramePtr frame_ptr = nvdec_processor->GetFrame();
     const auto* frame = frame_ptr.get();
     if (!frame) {
@@ -153,15 +156,16 @@ void Vic::WriteRGBFrame(const AVFrame* frame, const VicConfig& config) {
         const u32 block_height = static_cast<u32>(config.block_linear_height_log2);
         const auto size = Texture::CalculateSize(true, 4, width, height, 1, block_height, 0);
         luma_buffer.resize(size);
-        Texture::SwizzleSubrect(width, height, width * 4, width, 4, luma_buffer.data(),
-                                converted_frame_buf_addr, block_height, 0, 0);
+        std::span<const u8> frame_buff(converted_frame_buf_addr, 4 * width * height);
+        Texture::SwizzleSubrect(luma_buffer, frame_buff, 4, width, height, 1, 0, 0, width, height,
+                                block_height, 0, width * 4);
 
-        gpu.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(), size);
+        host1x.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(), size);
     } else {
         // send pitch linear frame
         const size_t linear_size = width * height * 4;
-        gpu.MemoryManager().WriteBlock(output_surface_luma_address, converted_frame_buf_addr,
-                                       linear_size);
+        host1x.MemoryManager().WriteBlock(output_surface_luma_address, converted_frame_buf_addr,
+                                          linear_size);
     }
 }
 
@@ -189,8 +193,8 @@ void Vic::WriteYUVFrame(const AVFrame* frame, const VicConfig& config) {
             luma_buffer[dst + x] = luma_src[src + x];
         }
     }
-    gpu.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(),
-                                   luma_buffer.size());
+    host1x.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(),
+                                      luma_buffer.size());
 
     // Chroma
     const std::size_t half_height = frame_height / 2;
@@ -231,8 +235,10 @@ void Vic::WriteYUVFrame(const AVFrame* frame, const VicConfig& config) {
         ASSERT(false);
         break;
     }
-    gpu.MemoryManager().WriteBlock(output_surface_chroma_address, chroma_buffer.data(),
-                                   chroma_buffer.size());
+    host1x.MemoryManager().WriteBlock(output_surface_chroma_address, chroma_buffer.data(),
+                                      chroma_buffer.size());
 }
 
+} // namespace Host1x
+
 } // namespace Tegra
diff --git a/src/video_core/command_classes/vic.h b/src/video_core/host1x/vic.h
index 010daa6b6..2b78786e8 100644
--- a/src/video_core/command_classes/vic.h
+++ b/src/video_core/host1x/vic.h
@@ -10,7 +10,10 @@
 struct SwsContext;
 
 namespace Tegra {
-class GPU;
+
+namespace Host1x {
+
+class Host1x;
 class Nvdec;
 union VicConfig;
 
@@ -25,7 +28,7 @@ public:
         SetOutputSurfaceChromaUnusedOffset = 0x1ca
     };
 
-    explicit Vic(GPU& gpu, std::shared_ptr<Nvdec> nvdec_processor);
+    explicit Vic(Host1x& host1x, std::shared_ptr<Nvdec> nvdec_processor);
 
     ~Vic();
 
@@ -39,8 +42,8 @@ private:
 
     void WriteYUVFrame(const AVFrame* frame, const VicConfig& config);
 
-    GPU& gpu;
-    std::shared_ptr<Tegra::Nvdec> nvdec_processor;
+    Host1x& host1x;
+    std::shared_ptr<Tegra::Host1x::Nvdec> nvdec_processor;
 
     /// Avoid reallocation of the following buffers every frame, as their
     /// size does not change during a stream
@@ -58,4 +61,6 @@ private:
     s32 scaler_height{};
 };
 
+} // namespace Host1x
+
 } // namespace Tegra