hdkv/yuzu
1
Fork 0
Code Issues Pull Requests Releases 1 Activity

early-access version 1255

Browse Source
This commit is contained in:
pineappleEA
2020-12-28 15:15:37 +00:00
parent 84b39492d1
commit 78b48028e1
6254 changed files with 1868140 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

129
src/video_core/command_classes/codecs/codec.cpp Executable file
View File

@@ -0,0 +1,129 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <cstring>
#include <fstream>
#include <vector>
#include "common/assert.h"
#include "video_core/command_classes/codecs/codec.h"
#include "video_core/command_classes/codecs/h264.h"
#include "video_core/command_classes/codecs/vp9.h"
#include "video_core/gpu.h"
#include "video_core/memory_manager.h"
extern "C" {
#include <libavutil/opt.h>
}
namespace Tegra {
void AVFrameDeleter(AVFrame* ptr) {
av_frame_unref(ptr);
av_free(ptr);
}
Codec::Codec(GPU& gpu_)
: gpu(gpu_), h264_decoder(std::make_unique<Decoder::H264>(gpu)),
vp9_decoder(std::make_unique<Decoder::VP9>(gpu)) {}
Codec::~Codec() {
if (!initialized) {
return;
}
// Free libav memory
AVFrame* av_frame{nullptr};
avcodec_send_packet(av_codec_ctx, nullptr);
av_frame = av_frame_alloc();
avcodec_receive_frame(av_codec_ctx, av_frame);
avcodec_flush_buffers(av_codec_ctx);
av_frame_unref(av_frame);
av_free(av_frame);
avcodec_close(av_codec_ctx);
}
void Codec::SetTargetCodec(NvdecCommon::VideoCodec codec) {
LOG_INFO(Service_NVDRV, "NVDEC video codec initialized to {}", codec);
current_codec = codec;
}
void Codec::StateWrite(u32 offset, u64 arguments) {
u8* const state_offset = reinterpret_cast<u8*>(&state) + offset * sizeof(u64);
std::memcpy(state_offset, &arguments, sizeof(u64));
}
void Codec::Decode() {
bool is_first_frame = false;
if (!initialized) {
if (current_codec == NvdecCommon::VideoCodec::H264) {
av_codec = avcodec_find_decoder(AV_CODEC_ID_H264);
} else if (current_codec == NvdecCommon::VideoCodec::Vp9) {
av_codec = avcodec_find_decoder(AV_CODEC_ID_VP9);
} else {
LOG_ERROR(Service_NVDRV, "Unknown video codec {}", current_codec);
return;
}
av_codec_ctx = avcodec_alloc_context3(av_codec);
av_opt_set(av_codec_ctx->priv_data, "tune", "zerolatency", 0);
// TODO(ameerj): libavcodec gpu hw acceleration
const auto av_error = avcodec_open2(av_codec_ctx, av_codec, nullptr);
if (av_error < 0) {
LOG_ERROR(Service_NVDRV, "avcodec_open2() Failed.");
avcodec_close(av_codec_ctx);
return;
}
initialized = true;
is_first_frame = true;
}
bool vp9_hidden_frame = false;
AVPacket packet{};
av_init_packet(&packet);
std::vector<u8> frame_data;
if (current_codec == NvdecCommon::VideoCodec::H264) {
frame_data = h264_decoder->ComposeFrameHeader(state, is_first_frame);
} else if (current_codec == NvdecCommon::VideoCodec::Vp9) {
frame_data = vp9_decoder->ComposeFrameHeader(state);
vp9_hidden_frame = vp9_decoder->WasFrameHidden();
}
packet.data = frame_data.data();
packet.size = static_cast<int>(frame_data.size());
avcodec_send_packet(av_codec_ctx, &packet);
if (!vp9_hidden_frame) {
// Only receive/store visible frames
AVFramePtr frame = AVFramePtr{av_frame_alloc(), AVFrameDeleter};
avcodec_receive_frame(av_codec_ctx, frame.get());
av_frames.push(std::move(frame));
// Limit queue to 10 frames. Workaround for ZLA decode and queue spam
if (av_frames.size() > 10) {
av_frames.pop();
}
}
}
AVFramePtr Codec::GetCurrentFrame() {
// Sometimes VIC will request more frames than have been decoded.
// in this case, return a nullptr and don't overwrite previous frame data
if (av_frames.empty()) {
return AVFramePtr{nullptr, AVFrameDeleter};
}
AVFramePtr frame = std::move(av_frames.front());
av_frames.pop();
return frame;
}
NvdecCommon::VideoCodec Codec::GetCurrentCodec() const {
return current_codec;
}
} // namespace Tegra

70
src/video_core/command_classes/codecs/codec.h Executable file
View File

@@ -0,0 +1,70 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <memory>
#include <queue>
#include "common/common_types.h"
#include "video_core/command_classes/nvdec_common.h"
extern "C" {
#if defined(__GNUC__) || defined(__clang__)
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wconversion"
#endif
#include <libavcodec/avcodec.h>
#if defined(__GNUC__) || defined(__clang__)
#pragma GCC diagnostic pop
#endif
}
namespace Tegra {
class GPU;
struct VicRegisters;
void AVFrameDeleter(AVFrame* ptr);
using AVFramePtr = std::unique_ptr<AVFrame, decltype(&AVFrameDeleter)>;
namespace Decoder {
class H264;
class VP9;
} // namespace Decoder
class Codec {
public:
explicit Codec(GPU& gpu);
~Codec();
/// Sets NVDEC video stream codec
void SetTargetCodec(NvdecCommon::VideoCodec codec);
/// Populate NvdecRegisters state with argument value at the provided offset
void StateWrite(u32 offset, u64 arguments);
/// Call decoders to construct headers, decode AVFrame with ffmpeg
void Decode();
/// Returns next decoded frame
[[nodiscard]] AVFramePtr GetCurrentFrame();
/// Returns the value of current_codec
[[nodiscard]] NvdecCommon::VideoCodec GetCurrentCodec() const;
private:
bool initialized{};
NvdecCommon::VideoCodec current_codec{NvdecCommon::VideoCodec::None};
AVCodec* av_codec{nullptr};
AVCodecContext* av_codec_ctx{nullptr};
GPU& gpu;
std::unique_ptr<Decoder::H264> h264_decoder;
std::unique_ptr<Decoder::VP9> vp9_decoder;
NvdecCommon::NvdecRegisters state{};
std::queue<AVFramePtr> av_frames{};
};
} // namespace Tegra

293
src/video_core/command_classes/codecs/h264.cpp Executable file
View File

@@ -0,0 +1,293 @@
// MIT License
//
// Copyright (c) Ryujinx Team and Contributors
//
// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and
// associated documentation files (the "Software"), to deal in the Software without restriction,
// including without limitation the rights to use, copy, modify, merge, publish, distribute,
// sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in all copies or
// substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT
// NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
// DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
//
#include <array>
#include "common/bit_util.h"
#include "video_core/command_classes/codecs/h264.h"
#include "video_core/gpu.h"
#include "video_core/memory_manager.h"
namespace Tegra::Decoder {
namespace {
// ZigZag LUTs from libavcodec.
constexpr std::array<u8, 64> zig_zag_direct{
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,
};
constexpr std::array<u8, 16> zig_zag_scan{
0 + 0 * 4, 1 + 0 * 4, 0 + 1 * 4, 0 + 2 * 4, 1 + 1 * 4, 2 + 0 * 4, 3 + 0 * 4, 2 + 1 * 4,
1 + 2 * 4, 0 + 3 * 4, 1 + 3 * 4, 2 + 2 * 4, 3 + 1 * 4, 3 + 2 * 4, 2 + 3 * 4, 3 + 3 * 4,
};
} // Anonymous namespace
H264::H264(GPU& gpu_) : gpu(gpu_) {}
H264::~H264() = default;
const std::vector<u8>& H264::ComposeFrameHeader(const NvdecCommon::NvdecRegisters& state,
bool is_first_frame) {
H264DecoderContext context{};
gpu.MemoryManager().ReadBlock(state.picture_info_offset, &context, sizeof(H264DecoderContext));
const s32 frame_number = static_cast<s32>((context.h264_parameter_set.flags >> 46) & 0x1ffff);
if (!is_first_frame && frame_number != 0) {
frame.resize(context.frame_data_size);
gpu.MemoryManager().ReadBlock(state.frame_bitstream_offset, frame.data(), frame.size());
} else {
/// Encode header
H264BitWriter writer{};
writer.WriteU(1, 24);
writer.WriteU(0, 1);
writer.WriteU(3, 2);
writer.WriteU(7, 5);
writer.WriteU(100, 8);
writer.WriteU(0, 8);
writer.WriteU(31, 8);
writer.WriteUe(0);
const auto chroma_format_idc =
static_cast<u32>((context.h264_parameter_set.flags >> 12) & 3);
writer.WriteUe(chroma_format_idc);
if (chroma_format_idc == 3) {
writer.WriteBit(false);
}
writer.WriteUe(0);
writer.WriteUe(0);
writer.WriteBit(false); // QpprimeYZeroTransformBypassFlag
writer.WriteBit(false); // Scaling matrix present flag
const auto order_cnt_type = static_cast<u32>((context.h264_parameter_set.flags >> 14) & 3);
writer.WriteUe(static_cast<u32>((context.h264_parameter_set.flags >> 8) & 0xf));
writer.WriteUe(order_cnt_type);
if (order_cnt_type == 0) {
writer.WriteUe(context.h264_parameter_set.log2_max_pic_order_cnt);
} else if (order_cnt_type == 1) {
writer.WriteBit(context.h264_parameter_set.delta_pic_order_always_zero_flag != 0);
writer.WriteSe(0);
writer.WriteSe(0);
writer.WriteUe(0);
}
const s32 pic_height = context.h264_parameter_set.pic_height_in_map_units /
(context.h264_parameter_set.frame_mbs_only_flag ? 1 : 2);
writer.WriteUe(16);
writer.WriteBit(false);
writer.WriteUe(context.h264_parameter_set.pic_width_in_mbs - 1);
writer.WriteUe(pic_height - 1);
writer.WriteBit(context.h264_parameter_set.frame_mbs_only_flag != 0);
if (!context.h264_parameter_set.frame_mbs_only_flag) {
writer.WriteBit(((context.h264_parameter_set.flags >> 0) & 1) != 0);
}
writer.WriteBit(((context.h264_parameter_set.flags >> 1) & 1) != 0);
writer.WriteBit(false); // Frame cropping flag
writer.WriteBit(false); // VUI parameter present flag
writer.End();
// H264 PPS
writer.WriteU(1, 24);
writer.WriteU(0, 1);
writer.WriteU(3, 2);
writer.WriteU(8, 5);
writer.WriteUe(0);
writer.WriteUe(0);
writer.WriteBit(context.h264_parameter_set.entropy_coding_mode_flag != 0);
writer.WriteBit(false);
writer.WriteUe(0);
writer.WriteUe(context.h264_parameter_set.num_refidx_l0_default_active);
writer.WriteUe(context.h264_parameter_set.num_refidx_l1_default_active);
writer.WriteBit(((context.h264_parameter_set.flags >> 2) & 1) != 0);
writer.WriteU(static_cast<s32>((context.h264_parameter_set.flags >> 32) & 0x3), 2);
s32 pic_init_qp = static_cast<s32>((context.h264_parameter_set.flags >> 16) & 0x3f);
pic_init_qp = (pic_init_qp << 26) >> 26;
writer.WriteSe(pic_init_qp);
writer.WriteSe(0);
s32 chroma_qp_index_offset =
static_cast<s32>((context.h264_parameter_set.flags >> 22) & 0x1f);
chroma_qp_index_offset = (chroma_qp_index_offset << 27) >> 27;
writer.WriteSe(chroma_qp_index_offset);
writer.WriteBit(context.h264_parameter_set.deblocking_filter_control_flag != 0);
writer.WriteBit(((context.h264_parameter_set.flags >> 3) & 1) != 0);
writer.WriteBit(context.h264_parameter_set.redundant_pic_count_flag != 0);
writer.WriteBit(context.h264_parameter_set.transform_8x8_mode_flag != 0);
writer.WriteBit(true);
for (s32 index = 0; index < 6; index++) {
writer.WriteBit(true);
const auto matrix_x4 =
std::vector<u8>(context.scaling_matrix_4.begin(), context.scaling_matrix_4.end());
writer.WriteScalingList(matrix_x4, index * 16, 16);
}
if (context.h264_parameter_set.transform_8x8_mode_flag) {
for (s32 index = 0; index < 2; index++) {
writer.WriteBit(true);
const auto matrix_x8 = std::vector<u8>(context.scaling_matrix_8.begin(),
context.scaling_matrix_8.end());
writer.WriteScalingList(matrix_x8, index * 64, 64);
}
}
s32 chroma_qp_index_offset2 =
static_cast<s32>((context.h264_parameter_set.flags >> 27) & 0x1f);
chroma_qp_index_offset2 = (chroma_qp_index_offset2 << 27) >> 27;
writer.WriteSe(chroma_qp_index_offset2);
writer.End();
const auto& encoded_header = writer.GetByteArray();
frame.resize(encoded_header.size() + context.frame_data_size);
std::memcpy(frame.data(), encoded_header.data(), encoded_header.size());
gpu.MemoryManager().ReadBlock(state.frame_bitstream_offset,
frame.data() + encoded_header.size(),
context.frame_data_size);
}
return frame;
}
H264BitWriter::H264BitWriter() = default;
H264BitWriter::~H264BitWriter() = default;
void H264BitWriter::WriteU(s32 value, s32 value_sz) {
WriteBits(value, value_sz);
}
void H264BitWriter::WriteSe(s32 value) {
WriteExpGolombCodedInt(value);
}
void H264BitWriter::WriteUe(u32 value) {
WriteExpGolombCodedUInt(value);
}
void H264BitWriter::End() {
WriteBit(true);
Flush();
}
void H264BitWriter::WriteBit(bool state) {
WriteBits(state ? 1 : 0, 1);
}
void H264BitWriter::WriteScalingList(const std::vector<u8>& list, s32 start, s32 count) {
std::vector<u8> scan(count);
if (count == 16) {
std::memcpy(scan.data(), zig_zag_scan.data(), scan.size());
} else {
std::memcpy(scan.data(), zig_zag_direct.data(), scan.size());
}
u8 last_scale = 8;
for (s32 index = 0; index < count; index++) {
const u8 value = list[start + scan[index]];
const s32 delta_scale = static_cast<s32>(value - last_scale);
WriteSe(delta_scale);
last_scale = value;
}
}
std::vector<u8>& H264BitWriter::GetByteArray() {
return byte_array;
}
const std::vector<u8>& H264BitWriter::GetByteArray() const {
return byte_array;
}
void H264BitWriter::WriteBits(s32 value, s32 bit_count) {
s32 value_pos = 0;
s32 remaining = bit_count;
while (remaining > 0) {
s32 copy_size = remaining;
const s32 free_bits = GetFreeBufferBits();
if (copy_size > free_bits) {
copy_size = free_bits;
}
const s32 mask = (1 << copy_size) - 1;
const s32 src_shift = (bit_count - value_pos) - copy_size;
const s32 dst_shift = (buffer_size - buffer_pos) - copy_size;
buffer |= ((value >> src_shift) & mask) << dst_shift;
value_pos += copy_size;
buffer_pos += copy_size;
remaining -= copy_size;
}
}
void H264BitWriter::WriteExpGolombCodedInt(s32 value) {
const s32 sign = value <= 0 ? 0 : 1;
if (value < 0) {
value = -value;
}
value = (value << 1) - sign;
WriteExpGolombCodedUInt(value);
}
void H264BitWriter::WriteExpGolombCodedUInt(u32 value) {
const s32 size = 32 - Common::CountLeadingZeroes32(static_cast<s32>(value + 1));
WriteBits(1, size);
value -= (1U << (size - 1)) - 1;
WriteBits(static_cast<s32>(value), size - 1);
}
s32 H264BitWriter::GetFreeBufferBits() {
if (buffer_pos == buffer_size) {
Flush();
}
return buffer_size - buffer_pos;
}
void H264BitWriter::Flush() {
if (buffer_pos == 0) {
return;
}
byte_array.push_back(static_cast<u8>(buffer));
buffer = 0;
buffer_pos = 0;
}
} // namespace Tegra::Decoder

118
src/video_core/command_classes/codecs/h264.h Executable file
View File

@@ -0,0 +1,118 @@
// MIT License
//
// Copyright (c) Ryujinx Team and Contributors
//
// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and
// associated documentation files (the "Software"), to deal in the Software without restriction,
// including without limitation the rights to use, copy, modify, merge, publish, distribute,
// sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in all copies or
// substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT
// NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
// DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
//
#pragma once
#include <vector>
#include "common/common_funcs.h"
#include "common/common_types.h"
#include "video_core/command_classes/nvdec_common.h"
namespace Tegra {
class GPU;
namespace Decoder {
class H264BitWriter {
public:
H264BitWriter();
~H264BitWriter();
/// The following Write methods are based on clause 9.1 in the H.264 specification.
/// WriteSe and WriteUe write in the Exp-Golomb-coded syntax
void WriteU(s32 value, s32 value_sz);
void WriteSe(s32 value);
void WriteUe(u32 value);
/// Finalize the bitstream
void End();
/// append a bit to the stream, equivalent value to the state parameter
void WriteBit(bool state);
/// Based on section 7.3.2.1.1.1 and Table 7-4 in the H.264 specification
/// Writes the scaling matrices of the sream
void WriteScalingList(const std::vector<u8>& list, s32 start, s32 count);
/// Return the bitstream as a vector.
[[nodiscard]] std::vector<u8>& GetByteArray();
[[nodiscard]] const std::vector<u8>& GetByteArray() const;
private:
void WriteBits(s32 value, s32 bit_count);
void WriteExpGolombCodedInt(s32 value);
void WriteExpGolombCodedUInt(u32 value);
[[nodiscard]] s32 GetFreeBufferBits();
void Flush();
s32 buffer_size{8};
s32 buffer{};
s32 buffer_pos{};
std::vector<u8> byte_array;
};
class H264 {
public:
explicit H264(GPU& gpu);
~H264();
/// Compose the H264 header of the frame for FFmpeg decoding
[[nodiscard]] const std::vector<u8>& ComposeFrameHeader(
const NvdecCommon::NvdecRegisters& state, bool is_first_frame = false);
private:
struct H264ParameterSet {
u32 log2_max_pic_order_cnt{};
u32 delta_pic_order_always_zero_flag{};
u32 frame_mbs_only_flag{};
u32 pic_width_in_mbs{};
u32 pic_height_in_map_units{};
INSERT_PADDING_WORDS(1);
u32 entropy_coding_mode_flag{};
u32 bottom_field_pic_order_flag{};
u32 num_refidx_l0_default_active{};
u32 num_refidx_l1_default_active{};
u32 deblocking_filter_control_flag{};
u32 redundant_pic_count_flag{};
u32 transform_8x8_mode_flag{};
INSERT_PADDING_WORDS(9);
u64 flags{};
u32 frame_number{};
u32 frame_number2{};
};
static_assert(sizeof(H264ParameterSet) == 0x68, "H264ParameterSet is an invalid size");
struct H264DecoderContext {
INSERT_PADDING_BYTES(0x48);
u32 frame_data_size{};
INSERT_PADDING_BYTES(0xc);
H264ParameterSet h264_parameter_set{};
INSERT_PADDING_BYTES(0x100);
std::array<u8, 0x60> scaling_matrix_4;
std::array<u8, 0x80> scaling_matrix_8;
};
static_assert(sizeof(H264DecoderContext) == 0x2a0, "H264DecoderContext is an invalid size");
std::vector<u8> frame;
GPU& gpu;
};
} // namespace Decoder
} // namespace Tegra

989
src/video_core/command_classes/codecs/vp9.cpp Executable file
View File

@@ -0,0 +1,989 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <cstring> // for std::memcpy
#include <numeric>
#include "video_core/command_classes/codecs/vp9.h"
#include "video_core/gpu.h"
#include "video_core/memory_manager.h"
namespace Tegra::Decoder {
namespace {
// Default compressed header probabilities once frame context resets
constexpr Vp9EntropyProbs default_probs{
.y_mode_prob{
65, 32, 18, 144, 162, 194, 41, 51, 98, 132, 68, 18, 165, 217, 196, 45, 40, 78,
173, 80, 19, 176, 240, 193, 64, 35, 46, 221, 135, 38, 194, 248, 121, 96, 85, 29,
},
.partition_prob{
199, 122, 141, 0, 147, 63, 159, 0, 148, 133, 118, 0, 121, 104, 114, 0,
174, 73, 87, 0, 92, 41, 83, 0, 82, 99, 50, 0, 53, 39, 39, 0,
177, 58, 59, 0, 68, 26, 63, 0, 52, 79, 25, 0, 17, 14, 12, 0,
222, 34, 30, 0, 72, 16, 44, 0, 58, 32, 12, 0, 10, 7, 6, 0,
},
.coef_probs{
195, 29, 183, 84, 49, 136, 8, 42, 71, 0, 0, 0, 0, 0, 0, 0, 0, 0,
31, 107, 169, 35, 99, 159, 17, 82, 140, 8, 66, 114, 2, 44, 76, 1, 19, 32,
40, 132, 201, 29, 114, 187, 13, 91, 157, 7, 75, 127, 3, 58, 95, 1, 28, 47,
69, 142, 221, 42, 122, 201, 15, 91, 159, 6, 67, 121, 1, 42, 77, 1, 17, 31,
102, 148, 228, 67, 117, 204, 17, 82, 154, 6, 59, 114, 2, 39, 75, 1, 15, 29,
156, 57, 233, 119, 57, 212, 58, 48, 163, 29, 40, 124, 12, 30, 81, 3, 12, 31,
191, 107, 226, 124, 117, 204, 25, 99, 155, 0, 0, 0, 0, 0, 0, 0, 0, 0,
29, 148, 210, 37, 126, 194, 8, 93, 157, 2, 68, 118, 1, 39, 69, 1, 17, 33,
41, 151, 213, 27, 123, 193, 3, 82, 144, 1, 58, 105, 1, 32, 60, 1, 13, 26,
59, 159, 220, 23, 126, 198, 4, 88, 151, 1, 66, 114, 1, 38, 71, 1, 18, 34,
114, 136, 232, 51, 114, 207, 11, 83, 155, 3, 56, 105, 1, 33, 65, 1, 17, 34,
149, 65, 234, 121, 57, 215, 61, 49, 166, 28, 36, 114, 12, 25, 76, 3, 16, 42,
214, 49, 220, 132, 63, 188, 42, 65, 137, 0, 0, 0, 0, 0, 0, 0, 0, 0,
85, 137, 221, 104, 131, 216, 49, 111, 192, 21, 87, 155, 2, 49, 87, 1, 16, 28,
89, 163, 230, 90, 137, 220, 29, 100, 183, 10, 70, 135, 2, 42, 81, 1, 17, 33,
108, 167, 237, 55, 133, 222, 15, 97, 179, 4, 72, 135, 1, 45, 85, 1, 19, 38,
124, 146, 240, 66, 124, 224, 17, 88, 175, 4, 58, 122, 1, 36, 75, 1, 18, 37,
141, 79, 241, 126, 70, 227, 66, 58, 182, 30, 44, 136, 12, 34, 96, 2, 20, 47,
229, 99, 249, 143, 111, 235, 46, 109, 192, 0, 0, 0, 0, 0, 0, 0, 0, 0,
82, 158, 236, 94, 146, 224, 25, 117, 191, 9, 87, 149, 3, 56, 99, 1, 33, 57,
83, 167, 237, 68, 145, 222, 10, 103, 177, 2, 72, 131, 1, 41, 79, 1, 20, 39,
99, 167, 239, 47, 141, 224, 10, 104, 178, 2, 73, 133, 1, 44, 85, 1, 22, 47,
127, 145, 243, 71, 129, 228, 17, 93, 177, 3, 61, 124, 1, 41, 84, 1, 21, 52,
157, 78, 244, 140, 72, 231, 69, 58, 184, 31, 44, 137, 14, 38, 105, 8, 23, 61,
125, 34, 187, 52, 41, 133, 6, 31, 56, 0, 0, 0, 0, 0, 0, 0, 0, 0,
37, 109, 153, 51, 102, 147, 23, 87, 128, 8, 67, 101, 1, 41, 63, 1, 19, 29,
31, 154, 185, 17, 127, 175, 6, 96, 145, 2, 73, 114, 1, 51, 82, 1, 28, 45,
23, 163, 200, 10, 131, 185, 2, 93, 148, 1, 67, 111, 1, 41, 69, 1, 14, 24,
29, 176, 217, 12, 145, 201, 3, 101, 156, 1, 69, 111, 1, 39, 63, 1, 14, 23,
57, 192, 233, 25, 154, 215, 6, 109, 167, 3, 78, 118, 1, 48, 69, 1, 21, 29,
202, 105, 245, 108, 106, 216, 18, 90, 144, 0, 0, 0, 0, 0, 0, 0, 0, 0,
33, 172, 219, 64, 149, 206, 14, 117, 177, 5, 90, 141, 2, 61, 95, 1, 37, 57,
33, 179, 220, 11, 140, 198, 1, 89, 148, 1, 60, 104, 1, 33, 57, 1, 12, 21,
30, 181, 221, 8, 141, 198, 1, 87, 145, 1, 58, 100, 1, 31, 55, 1, 12, 20,
32, 186, 224, 7, 142, 198, 1, 86, 143, 1, 58, 100, 1, 31, 55, 1, 12, 22,
57, 192, 227, 20, 143, 204, 3, 96, 154, 1, 68, 112, 1, 42, 69, 1, 19, 32,
212, 35, 215, 113, 47, 169, 29, 48, 105, 0, 0, 0, 0, 0, 0, 0, 0, 0,
74, 129, 203, 106, 120, 203, 49, 107, 178, 19, 84, 144, 4, 50, 84, 1, 15, 25,
71, 172, 217, 44, 141, 209, 15, 102, 173, 6, 76, 133, 2, 51, 89, 1, 24, 42,
64, 185, 231, 31, 148, 216, 8, 103, 175, 3, 74, 131, 1, 46, 81, 1, 18, 30,
65, 196, 235, 25, 157, 221, 5, 105, 174, 1, 67, 120, 1, 38, 69, 1, 15, 30,
65, 204, 238, 30, 156, 224, 7, 107, 177, 2, 70, 124, 1, 42, 73, 1, 18, 34,
225, 86, 251, 144, 104, 235, 42, 99, 181, 0, 0, 0, 0, 0, 0, 0, 0, 0,
85, 175, 239, 112, 165, 229, 29, 136, 200, 12, 103, 162, 6, 77, 123, 2, 53, 84,
75, 183, 239, 30, 155, 221, 3, 106, 171, 1, 74, 128, 1, 44, 76, 1, 17, 28,
73, 185, 240, 27, 159, 222, 2, 107, 172, 1, 75, 127, 1, 42, 73, 1, 17, 29,
62, 190, 238, 21, 159, 222, 2, 107, 172, 1, 72, 122, 1, 40, 71, 1, 18, 32,
61, 199, 240, 27, 161, 226, 4, 113, 180, 1, 76, 129, 1, 46, 80, 1, 23, 41,
7, 27, 153, 5, 30, 95, 1, 16, 30, 0, 0, 0, 0, 0, 0, 0, 0, 0,
50, 75, 127, 57, 75, 124, 27, 67, 108, 10, 54, 86, 1, 33, 52, 1, 12, 18,
43, 125, 151, 26, 108, 148, 7, 83, 122, 2, 59, 89, 1, 38, 60, 1, 17, 27,
23, 144, 163, 13, 112, 154, 2, 75, 117, 1, 50, 81, 1, 31, 51, 1, 14, 23,
18, 162, 185, 6, 123, 171, 1, 78, 125, 1, 51, 86, 1, 31, 54, 1, 14, 23,
15, 199, 227, 3, 150, 204, 1, 91, 146, 1, 55, 95, 1, 30, 53, 1, 11, 20,
19, 55, 240, 19, 59, 196, 3, 52, 105, 0, 0, 0, 0, 0, 0, 0, 0, 0,
41, 166, 207, 104, 153, 199, 31, 123, 181, 14, 101, 152, 5, 72, 106, 1, 36, 52,
35, 176, 211, 12, 131, 190, 2, 88, 144, 1, 60, 101, 1, 36, 60, 1, 16, 28,
28, 183, 213, 8, 134, 191, 1, 86, 142, 1, 56, 96, 1, 30, 53, 1, 12, 20,
20, 190, 215, 4, 135, 192, 1, 84, 139, 1, 53, 91, 1, 28, 49, 1, 11, 20,
13, 196, 216, 2, 137, 192, 1, 86, 143, 1, 57, 99, 1, 32, 56, 1, 13, 24,
211, 29, 217, 96, 47, 156, 22, 43, 87, 0, 0, 0, 0, 0, 0, 0, 0, 0,
78, 120, 193, 111, 116, 186, 46, 102, 164, 15, 80, 128, 2, 49, 76, 1, 18, 28,
71, 161, 203, 42, 132, 192, 10, 98, 150, 3, 69, 109, 1, 44, 70, 1, 18, 29,
57, 186, 211, 30, 140, 196, 4, 93, 146, 1, 62, 102, 1, 38, 65, 1, 16, 27,
47, 199, 217, 14, 145, 196, 1, 88, 142, 1, 57, 98, 1, 36, 62, 1, 15, 26,
26, 219, 229, 5, 155, 207, 1, 94, 151, 1, 60, 104, 1, 36, 62, 1, 16, 28,
233, 29, 248, 146, 47, 220, 43, 52, 140, 0, 0, 0, 0, 0, 0, 0, 0, 0,
100, 163, 232, 179, 161, 222, 63, 142, 204, 37, 113, 174, 26, 89, 137, 18, 68, 97,
85, 181, 230, 32, 146, 209, 7, 100, 164, 3, 71, 121, 1, 45, 77, 1, 18, 30,
65, 187, 230, 20, 148, 207, 2, 97, 159, 1, 68, 116, 1, 40, 70, 1, 14, 29,
40, 194, 227, 8, 147, 204, 1, 94, 155, 1, 65, 112, 1, 39, 66, 1, 14, 26,
16, 208, 228, 3, 151, 207, 1, 98, 160, 1, 67, 117, 1, 41, 74, 1, 17, 31,
17, 38, 140, 7, 34, 80, 1, 17, 29, 0, 0, 0, 0, 0, 0, 0, 0, 0,
37, 75, 128, 41, 76, 128, 26, 66, 116, 12, 52, 94, 2, 32, 55, 1, 10, 16,
50, 127, 154, 37, 109, 152, 16, 82, 121, 5, 59, 85, 1, 35, 54, 1, 13, 20,
40, 142, 167, 17, 110, 157, 2, 71, 112, 1, 44, 72, 1, 27, 45, 1, 11, 17,
30, 175, 188, 9, 124, 169, 1, 74, 116, 1, 48, 78, 1, 30, 49, 1, 11, 18,
10, 222, 223, 2, 150, 194, 1, 83, 128, 1, 48, 79, 1, 27, 45, 1, 11, 17,
36, 41, 235, 29, 36, 193, 10, 27, 111, 0, 0, 0, 0, 0, 0, 0, 0, 0,
85, 165, 222, 177, 162, 215, 110, 135, 195, 57, 113, 168, 23, 83, 120, 10, 49, 61,
85, 190, 223, 36, 139, 200, 5, 90, 146, 1, 60, 103, 1, 38, 65, 1, 18, 30,
72, 202, 223, 23, 141, 199, 2, 86, 140, 1, 56, 97, 1, 36, 61, 1, 16, 27,
55, 218, 225, 13, 145, 200, 1, 86, 141, 1, 57, 99, 1, 35, 61, 1, 13, 22,
15, 235, 212, 1, 132, 184, 1, 84, 139, 1, 57, 97, 1, 34, 56, 1, 14, 23,
181, 21, 201, 61, 37, 123, 10, 38, 71, 0, 0, 0, 0, 0, 0, 0, 0, 0,
47, 106, 172, 95, 104, 173, 42, 93, 159, 18, 77, 131, 4, 50, 81, 1, 17, 23,
62, 147, 199, 44, 130, 189, 28, 102, 154, 18, 75, 115, 2, 44, 65, 1, 12, 19,
55, 153, 210, 24, 130, 194, 3, 93, 146, 1, 61, 97, 1, 31, 50, 1, 10, 16,
49, 186, 223, 17, 148, 204, 1, 96, 142, 1, 53, 83, 1, 26, 44, 1, 11, 17,
13, 217, 212, 2, 136, 180, 1, 78, 124, 1, 50, 83, 1, 29, 49, 1, 14, 23,
197, 13, 247, 82, 17, 222, 25, 17, 162, 0, 0, 0, 0, 0, 0, 0, 0, 0,
126, 186, 247, 234, 191, 243, 176, 177, 234, 104, 158, 220, 66, 128, 186, 55, 90, 137,
111, 197, 242, 46, 158, 219, 9, 104, 171, 2, 65, 125, 1, 44, 80, 1, 17, 91,
104, 208, 245, 39, 168, 224, 3, 109, 162, 1, 79, 124, 1, 50, 102, 1, 43, 102,
84, 220, 246, 31, 177, 231, 2, 115, 180, 1, 79, 134, 1, 55, 77, 1, 60, 79,
43, 243, 240, 8, 180, 217, 1, 115, 166, 1, 84, 121, 1, 51, 67, 1, 16, 6,
},
.switchable_interp_prob{235, 162, 36, 255, 34, 3, 149, 144},
.inter_mode_prob{
2, 173, 34, 0, 7, 145, 85, 0, 7, 166, 63, 0, 7, 94,
66, 0, 8, 64, 46, 0, 17, 81, 31, 0, 25, 29, 30, 0,
},
.intra_inter_prob{9, 102, 187, 225},
.comp_inter_prob{9, 102, 187, 225, 0},
.single_ref_prob{33, 16, 77, 74, 142, 142, 172, 170, 238, 247},
.comp_ref_prob{50, 126, 123, 221, 226},
.tx_32x32_prob{3, 136, 37, 5, 52, 13},
.tx_16x16_prob{20, 152, 15, 101},
.tx_8x8_prob{100, 66},
.skip_probs{192, 128, 64},
.joints{32, 64, 96},
.sign{128, 128},
.classes{
224, 144, 192, 168, 192, 176, 192, 198, 198, 245,
216, 128, 176, 160, 176, 176, 192, 198, 198, 208,
},
.class_0{216, 208},
.prob_bits{
136, 140, 148, 160, 176, 192, 224, 234, 234, 240,
136, 140, 148, 160, 176, 192, 224, 234, 234, 240,
},
.class_0_fr{128, 128, 64, 96, 112, 64, 128, 128, 64, 96, 112, 64},
.fr{64, 96, 64, 64, 96, 64},
.class_0_hp{160, 160},
.high_precision{128, 128},
};
constexpr std::array<s32, 256> norm_lut{
0, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
};
constexpr std::array<s32, 254> map_lut{
20, 21, 22, 23, 24, 25, 0, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37,
1, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 2, 50, 51, 52, 53, 54,
55, 56, 57, 58, 59, 60, 61, 3, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72,
73, 4, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 5, 86, 87, 88, 89,
90, 91, 92, 93, 94, 95, 96, 97, 6, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107,
108, 109, 7, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 8, 122, 123, 124,
125, 126, 127, 128, 129, 130, 131, 132, 133, 9, 134, 135, 136, 137, 138, 139, 140, 141, 142,
143, 144, 145, 10, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 11, 158, 159,
160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 12, 170, 171, 172, 173, 174, 175, 176, 177,
178, 179, 180, 181, 13, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 14, 194,
195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 15, 206, 207, 208, 209, 210, 211, 212,
213, 214, 215, 216, 217, 16, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 17,
230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 18, 242, 243, 244, 245, 246, 247,
248, 249, 250, 251, 252, 253, 19,
};
// 6.2.14 Tile size calculation
[[nodiscard]] s32 CalcMinLog2TileCols(s32 frame_width) {
const s32 sb64_cols = (frame_width + 63) / 64;
s32 min_log2 = 0;
while ((64 << min_log2) < sb64_cols) {
min_log2++;
}
return min_log2;
}
[[nodiscard]] s32 CalcMaxLog2TileCols(s32 frame_width) {
const s32 sb64_cols = (frame_width + 63) / 64;
s32 max_log2 = 1;
while ((sb64_cols >> max_log2) >= 4) {
max_log2++;
}
return max_log2 - 1;
}
// Recenters probability. Based on section 6.3.6 of VP9 Specification
[[nodiscard]] s32 RecenterNonNeg(s32 new_prob, s32 old_prob) {
if (new_prob > old_prob * 2) {
return new_prob;
}
if (new_prob >= old_prob) {
return (new_prob - old_prob) * 2;
}
return (old_prob - new_prob) * 2 - 1;
}
// Adjusts old_prob depending on new_prob. Based on section 6.3.5 of VP9 Specification
[[nodiscard]] s32 RemapProbability(s32 new_prob, s32 old_prob) {
new_prob--;
old_prob--;
std::size_t index{};
if (old_prob * 2 <= 0xff) {
index = static_cast<std::size_t>(std::max(0, RecenterNonNeg(new_prob, old_prob) - 1));
} else {
index = static_cast<std::size_t>(
std::max(0, RecenterNonNeg(0xff - 1 - new_prob, 0xff - 1 - old_prob) - 1));
}
return map_lut[index];
}
} // Anonymous namespace
VP9::VP9(GPU& gpu_) : gpu{gpu_} {}
VP9::~VP9() = default;
void VP9::WriteProbabilityUpdate(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob) {
const bool update = new_prob != old_prob;
writer.Write(update, diff_update_probability);
if (update) {
WriteProbabilityDelta(writer, new_prob, old_prob);
}
}
template <typename T, std::size_t N>
void VP9::WriteProbabilityUpdate(VpxRangeEncoder& writer, const std::array<T, N>& new_prob,
const std::array<T, N>& old_prob) {
for (std::size_t offset = 0; offset < new_prob.size(); ++offset) {
WriteProbabilityUpdate(writer, new_prob[offset], old_prob[offset]);
}
}
template <typename T, std::size_t N>
void VP9::WriteProbabilityUpdateAligned4(VpxRangeEncoder& writer, const std::array<T, N>& new_prob,
const std::array<T, N>& old_prob) {
for (std::size_t offset = 0; offset < new_prob.size(); offset += 4) {
WriteProbabilityUpdate(writer, new_prob[offset + 0], old_prob[offset + 0]);
WriteProbabilityUpdate(writer, new_prob[offset + 1], old_prob[offset + 1]);
WriteProbabilityUpdate(writer, new_prob[offset + 2], old_prob[offset + 2]);
}
}
void VP9::WriteProbabilityDelta(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob) {
const int delta = RemapProbability(new_prob, old_prob);
EncodeTermSubExp(writer, delta);
}
void VP9::EncodeTermSubExp(VpxRangeEncoder& writer, s32 value) {
if (WriteLessThan(writer, value, 16)) {
writer.Write(value, 4);
} else if (WriteLessThan(writer, value, 32)) {
writer.Write(value - 16, 4);
} else if (WriteLessThan(writer, value, 64)) {
writer.Write(value - 32, 5);
} else {
value -= 64;
constexpr s32 size = 8;
const s32 mask = (1 << size) - 191;
const s32 delta = value - mask;
if (delta < 0) {
writer.Write(value, size - 1);
} else {
writer.Write(delta / 2 + mask, size - 1);
writer.Write(delta & 1, 1);
}
}
}
bool VP9::WriteLessThan(VpxRangeEncoder& writer, s32 value, s32 test) {
const bool is_lt = value < test;
writer.Write(!is_lt);
return is_lt;
}
void VP9::WriteCoefProbabilityUpdate(VpxRangeEncoder& writer, s32 tx_mode,
const std::array<u8, 1728>& new_prob,
const std::array<u8, 1728>& old_prob) {
constexpr u32 block_bytes = 2 * 2 * 6 * 6 * 3;
const auto needs_update = [&](u32 base_index) {
return !std::equal(new_prob.begin() + base_index,
new_prob.begin() + base_index + block_bytes,
old_prob.begin() + base_index);
};
for (u32 block_index = 0; block_index < 4; block_index++) {
const u32 base_index = block_index * block_bytes;
const bool update = needs_update(base_index);
writer.Write(update);
if (update) {
u32 index = base_index;
for (s32 i = 0; i < 2; i++) {
for (s32 j = 0; j < 2; j++) {
for (s32 k = 0; k < 6; k++) {
for (s32 l = 0; l < 6; l++) {
if (k != 0 || l < 3) {
WriteProbabilityUpdate(writer, new_prob[index + 0],
old_prob[index + 0]);
WriteProbabilityUpdate(writer, new_prob[index + 1],
old_prob[index + 1]);
WriteProbabilityUpdate(writer, new_prob[index + 2],
old_prob[index + 2]);
}
index += 3;
}
}
}
}
}
if (block_index == static_cast<u32>(tx_mode)) {
break;
}
}
}
void VP9::WriteMvProbabilityUpdate(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob) {
const bool update = new_prob != old_prob;
writer.Write(update, diff_update_probability);
if (update) {
writer.Write(new_prob >> 1, 7);
}
}
Vp9PictureInfo VP9::GetVp9PictureInfo(const NvdecCommon::NvdecRegisters& state) {
PictureInfo picture_info{};
gpu.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);
// surface_luma_offset[0:3] contains the address of the reference frame offsets in the following
// order: last, golden, altref, current. It may be worthwhile to track the updates done here
// to avoid buffering frame data needed for reference frame updating in the header composition.
std::memcpy(vp9_info.frame_offsets.data(), state.surface_luma_offset.data(), 4 * sizeof(u64));
return vp9_info;
}
void VP9::InsertEntropy(u64 offset, Vp9EntropyProbs& dst) {
EntropyProbs entropy{};
gpu.MemoryManager().ReadBlock(offset, &entropy, sizeof(EntropyProbs));
entropy.Convert(dst);
}
Vp9FrameContainer VP9::GetCurrentFrame(const NvdecCommon::NvdecRegisters& state) {
Vp9FrameContainer current_frame{};
{
gpu.SyncGuestHost();
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);
}
// Buffer two frames, saving the last show frame info
if (!next_next_frame.bit_stream.empty()) {
Vp9FrameContainer temp{
.info = current_frame.info,
.bit_stream = std::move(current_frame.bit_stream),
};
next_next_frame.info.show_frame = current_frame.info.last_frame_shown;
current_frame.info = next_next_frame.info;
current_frame.bit_stream = std::move(next_next_frame.bit_stream);
next_next_frame = std::move(temp);
if (!next_frame.bit_stream.empty()) {
Vp9FrameContainer temp2{
.info = current_frame.info,
.bit_stream = std::move(current_frame.bit_stream),
};
next_frame.info.show_frame = current_frame.info.last_frame_shown;
current_frame.info = next_frame.info;
current_frame.bit_stream = std::move(next_frame.bit_stream);
next_frame = std::move(temp2);
} else {
next_frame.info = current_frame.info;
next_frame.bit_stream = std::move(current_frame.bit_stream);
}
} else {
next_next_frame.info = current_frame.info;
next_next_frame.bit_stream = std::move(current_frame.bit_stream);
}
return current_frame;
}
std::vector<u8> VP9::ComposeCompressedHeader() {
VpxRangeEncoder writer{};
const bool update_probs = current_frame_info.show_frame && !current_frame_info.is_key_frame;
if (!current_frame_info.lossless) {
if (static_cast<u32>(current_frame_info.transform_mode) >= 3) {
writer.Write(3, 2);
writer.Write(current_frame_info.transform_mode == 4);
} else {
writer.Write(current_frame_info.transform_mode, 2);
}
}
if (current_frame_info.transform_mode == 4) {
// tx_mode_probs() in the spec
WriteProbabilityUpdate(writer, current_frame_info.entropy.tx_8x8_prob,
prev_frame_probs.tx_8x8_prob);
WriteProbabilityUpdate(writer, current_frame_info.entropy.tx_16x16_prob,
prev_frame_probs.tx_16x16_prob);
WriteProbabilityUpdate(writer, current_frame_info.entropy.tx_32x32_prob,
prev_frame_probs.tx_32x32_prob);
if (update_probs) {
prev_frame_probs.tx_8x8_prob = current_frame_info.entropy.tx_8x8_prob;
prev_frame_probs.tx_16x16_prob = current_frame_info.entropy.tx_16x16_prob;
prev_frame_probs.tx_32x32_prob = current_frame_info.entropy.tx_32x32_prob;
}
}
// read_coef_probs() in the spec
WriteCoefProbabilityUpdate(writer, current_frame_info.transform_mode,
current_frame_info.entropy.coef_probs, prev_frame_probs.coef_probs);
// read_skip_probs() in the spec
WriteProbabilityUpdate(writer, current_frame_info.entropy.skip_probs,
prev_frame_probs.skip_probs);
if (update_probs) {
prev_frame_probs.coef_probs = current_frame_info.entropy.coef_probs;
prev_frame_probs.skip_probs = current_frame_info.entropy.skip_probs;
}
if (!current_frame_info.intra_only) {
// read_inter_probs() in the spec
WriteProbabilityUpdateAligned4(writer, current_frame_info.entropy.inter_mode_prob,
prev_frame_probs.inter_mode_prob);
if (current_frame_info.interp_filter == 4) {
// read_interp_filter_probs() in the spec
WriteProbabilityUpdate(writer, current_frame_info.entropy.switchable_interp_prob,
prev_frame_probs.switchable_interp_prob);
if (update_probs) {
prev_frame_probs.switchable_interp_prob =
current_frame_info.entropy.switchable_interp_prob;
}
}
// read_is_inter_probs() in the spec
WriteProbabilityUpdate(writer, current_frame_info.entropy.intra_inter_prob,
prev_frame_probs.intra_inter_prob);
// frame_reference_mode() in the spec
if ((current_frame_info.ref_frame_sign_bias[1] & 1) !=
(current_frame_info.ref_frame_sign_bias[2] & 1) ||
(current_frame_info.ref_frame_sign_bias[1] & 1) !=
(current_frame_info.ref_frame_sign_bias[3] & 1)) {
if (current_frame_info.reference_mode >= 1) {
writer.Write(1, 1);
writer.Write(current_frame_info.reference_mode == 2);
} else {
writer.Write(0, 1);
}
}
// frame_reference_mode_probs() in the spec
if (current_frame_info.reference_mode == 2) {
WriteProbabilityUpdate(writer, current_frame_info.entropy.comp_inter_prob,
prev_frame_probs.comp_inter_prob);
if (update_probs) {
prev_frame_probs.comp_inter_prob = current_frame_info.entropy.comp_inter_prob;
}
}
if (current_frame_info.reference_mode != 1) {
WriteProbabilityUpdate(writer, current_frame_info.entropy.single_ref_prob,
prev_frame_probs.single_ref_prob);
if (update_probs) {
prev_frame_probs.single_ref_prob = current_frame_info.entropy.single_ref_prob;
}
}
if (current_frame_info.reference_mode != 0) {
WriteProbabilityUpdate(writer, current_frame_info.entropy.comp_ref_prob,
prev_frame_probs.comp_ref_prob);
if (update_probs) {
prev_frame_probs.comp_ref_prob = current_frame_info.entropy.comp_ref_prob;
}
}
// read_y_mode_probs
for (std::size_t index = 0; index < current_frame_info.entropy.y_mode_prob.size();
++index) {
WriteProbabilityUpdate(writer, current_frame_info.entropy.y_mode_prob[index],
prev_frame_probs.y_mode_prob[index]);
}
// read_partition_probs
WriteProbabilityUpdateAligned4(writer, current_frame_info.entropy.partition_prob,
prev_frame_probs.partition_prob);
// mv_probs
for (s32 i = 0; i < 3; i++) {
WriteMvProbabilityUpdate(writer, current_frame_info.entropy.joints[i],
prev_frame_probs.joints[i]);
}
if (update_probs) {
prev_frame_probs.inter_mode_prob = current_frame_info.entropy.inter_mode_prob;
prev_frame_probs.intra_inter_prob = current_frame_info.entropy.intra_inter_prob;
prev_frame_probs.y_mode_prob = current_frame_info.entropy.y_mode_prob;
prev_frame_probs.partition_prob = current_frame_info.entropy.partition_prob;
prev_frame_probs.joints = current_frame_info.entropy.joints;
}
for (s32 i = 0; i < 2; i++) {
WriteMvProbabilityUpdate(writer, current_frame_info.entropy.sign[i],
prev_frame_probs.sign[i]);
for (s32 j = 0; j < 10; j++) {
const int index = i * 10 + j;
WriteMvProbabilityUpdate(writer, current_frame_info.entropy.classes[index],
prev_frame_probs.classes[index]);
}
WriteMvProbabilityUpdate(writer, current_frame_info.entropy.class_0[i],
prev_frame_probs.class_0[i]);
for (s32 j = 0; j < 10; j++) {
const int index = i * 10 + j;
WriteMvProbabilityUpdate(writer, current_frame_info.entropy.prob_bits[index],
prev_frame_probs.prob_bits[index]);
}
}
for (s32 i = 0; i < 2; i++) {
for (s32 j = 0; j < 2; j++) {
for (s32 k = 0; k < 3; k++) {
const int index = i * 2 * 3 + j * 3 + k;
WriteMvProbabilityUpdate(writer, current_frame_info.entropy.class_0_fr[index],
prev_frame_probs.class_0_fr[index]);
}
}
for (s32 j = 0; j < 3; j++) {
const int index = i * 3 + j;
WriteMvProbabilityUpdate(writer, current_frame_info.entropy.fr[index],
prev_frame_probs.fr[index]);
}
}
if (current_frame_info.allow_high_precision_mv) {
for (s32 index = 0; index < 2; index++) {
WriteMvProbabilityUpdate(writer, current_frame_info.entropy.class_0_hp[index],
prev_frame_probs.class_0_hp[index]);
WriteMvProbabilityUpdate(writer, current_frame_info.entropy.high_precision[index],
prev_frame_probs.high_precision[index]);
}
}
// save previous probs
if (update_probs) {
prev_frame_probs.sign = current_frame_info.entropy.sign;
prev_frame_probs.classes = current_frame_info.entropy.classes;
prev_frame_probs.class_0 = current_frame_info.entropy.class_0;
prev_frame_probs.prob_bits = current_frame_info.entropy.prob_bits;
prev_frame_probs.class_0_fr = current_frame_info.entropy.class_0_fr;
prev_frame_probs.fr = current_frame_info.entropy.fr;
prev_frame_probs.class_0_hp = current_frame_info.entropy.class_0_hp;
prev_frame_probs.high_precision = current_frame_info.entropy.high_precision;
}
}
writer.End();
return writer.GetBuffer();
}
VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
VpxBitStreamWriter uncomp_writer{};
uncomp_writer.WriteU(2, 2); // Frame marker.
uncomp_writer.WriteU(0, 2); // Profile.
uncomp_writer.WriteBit(false); // Show existing frame.
uncomp_writer.WriteBit(!current_frame_info.is_key_frame); // is key frame?
uncomp_writer.WriteBit(current_frame_info.show_frame); // show frame?
uncomp_writer.WriteBit(current_frame_info.error_resilient_mode); // error reslience
if (current_frame_info.is_key_frame) {
uncomp_writer.WriteU(frame_sync_code, 24);
uncomp_writer.WriteU(0, 3); // Color space.
uncomp_writer.WriteU(0, 1); // Color range.
uncomp_writer.WriteU(current_frame_info.frame_size.width - 1, 16);
uncomp_writer.WriteU(current_frame_info.frame_size.height - 1, 16);
uncomp_writer.WriteBit(false); // Render and frame size different.
// Reset context
prev_frame_probs = default_probs;
swap_next_golden = false;
loop_filter_ref_deltas.fill(0);
loop_filter_mode_deltas.fill(0);
// allow frames offsets to stabilize before checking for golden frames
grace_period = 4;
// On key frames, all frame slots are set to the current frame,
// so the value of the selected slot doesn't really matter.
frame_ctxs.fill({current_frame_number, false, default_probs});
// intra only, meaning the frame can be recreated with no other references
current_frame_info.intra_only = true;
} else {
if (!current_frame_info.show_frame) {
uncomp_writer.WriteBit(current_frame_info.intra_only);
if (!current_frame_info.last_frame_was_key) {
swap_next_golden = !swap_next_golden;
}
} else {
current_frame_info.intra_only = false;
}
if (!current_frame_info.error_resilient_mode) {
uncomp_writer.WriteU(0, 2); // Reset frame context.
}
// Last, Golden, Altref frames
std::array<s32, 3> ref_frame_index{0, 1, 2};
// Set when next frame is hidden
// altref and golden references are swapped
if (swap_next_golden) {
ref_frame_index = std::array<s32, 3>{0, 2, 1};
}
// update Last Frame
u64 refresh_frame_flags = 1;
// golden frame may refresh, determined if the next golden frame offset is changed
bool golden_refresh = false;
if (grace_period <= 0) {
for (s32 index = 1; index < 3; ++index) {
if (current_frame_info.frame_offsets[index] !=
next_frame.info.frame_offsets[index]) {
current_frame_info.refresh_frame[index] = true;
golden_refresh = true;
grace_period = 3;
}
}
}
if (current_frame_info.show_frame &&
(!next_frame.info.show_frame || next_frame.info.is_key_frame)) {
// Update golden frame
refresh_frame_flags = swap_next_golden ? 2 : 4;
}
if (!current_frame_info.show_frame) {
// Update altref
refresh_frame_flags = swap_next_golden ? 2 : 4;
} else if (golden_refresh) {
refresh_frame_flags = 3;
}
if (current_frame_info.intra_only) {
uncomp_writer.WriteU(frame_sync_code, 24);
uncomp_writer.WriteU(static_cast<s32>(refresh_frame_flags), 8);
uncomp_writer.WriteU(current_frame_info.frame_size.width - 1, 16);
uncomp_writer.WriteU(current_frame_info.frame_size.height - 1, 16);
uncomp_writer.WriteBit(false); // Render and frame size different.
} else {
uncomp_writer.WriteU(static_cast<s32>(refresh_frame_flags), 8);
for (s32 index = 1; index < 4; index++) {
uncomp_writer.WriteU(ref_frame_index[index - 1], 3);
uncomp_writer.WriteU(current_frame_info.ref_frame_sign_bias[index], 1);
}
uncomp_writer.WriteBit(true); // Frame size with refs.
uncomp_writer.WriteBit(false); // Render and frame size different.
uncomp_writer.WriteBit(current_frame_info.allow_high_precision_mv);
uncomp_writer.WriteBit(current_frame_info.interp_filter == 4);
if (current_frame_info.interp_filter != 4) {
uncomp_writer.WriteU(current_frame_info.interp_filter, 2);
}
}
}
if (!current_frame_info.error_resilient_mode) {
uncomp_writer.WriteBit(true); // Refresh frame context. where do i get this info from?
uncomp_writer.WriteBit(true); // Frame parallel decoding mode.
}
int frame_ctx_idx = 0;
if (!current_frame_info.show_frame) {
frame_ctx_idx = 1;
}
uncomp_writer.WriteU(frame_ctx_idx, 2); // Frame context index.
prev_frame_probs =
frame_ctxs[frame_ctx_idx].probs; // reference probabilities for compressed header
frame_ctxs[frame_ctx_idx] = {current_frame_number, false, current_frame_info.entropy};
uncomp_writer.WriteU(current_frame_info.first_level, 6);
uncomp_writer.WriteU(current_frame_info.sharpness_level, 3);
uncomp_writer.WriteBit(current_frame_info.mode_ref_delta_enabled);
if (current_frame_info.mode_ref_delta_enabled) {
// check if ref deltas are different, update accordingly
std::array<bool, 4> update_loop_filter_ref_deltas;
std::array<bool, 2> update_loop_filter_mode_deltas;
bool loop_filter_delta_update = false;
for (std::size_t index = 0; index < current_frame_info.ref_deltas.size(); index++) {
const s8 old_deltas = loop_filter_ref_deltas[index];
const s8 new_deltas = current_frame_info.ref_deltas[index];
const bool differing_delta = old_deltas != new_deltas;
update_loop_filter_ref_deltas[index] = differing_delta;
loop_filter_delta_update |= differing_delta;
}
for (std::size_t index = 0; index < current_frame_info.mode_deltas.size(); index++) {
const s8 old_deltas = loop_filter_mode_deltas[index];
const s8 new_deltas = current_frame_info.mode_deltas[index];
const bool differing_delta = old_deltas != new_deltas;
update_loop_filter_mode_deltas[index] = differing_delta;
loop_filter_delta_update |= differing_delta;
}
uncomp_writer.WriteBit(loop_filter_delta_update);
if (loop_filter_delta_update) {
for (std::size_t index = 0; index < current_frame_info.ref_deltas.size(); index++) {
uncomp_writer.WriteBit(update_loop_filter_ref_deltas[index]);
if (update_loop_filter_ref_deltas[index]) {
uncomp_writer.WriteS(current_frame_info.ref_deltas[index], 6);
}
}
for (std::size_t index = 0; index < current_frame_info.mode_deltas.size(); index++) {
uncomp_writer.WriteBit(update_loop_filter_mode_deltas[index]);
if (update_loop_filter_mode_deltas[index]) {
uncomp_writer.WriteS(current_frame_info.mode_deltas[index], 6);
}
}
// save new deltas
loop_filter_ref_deltas = current_frame_info.ref_deltas;
loop_filter_mode_deltas = current_frame_info.mode_deltas;
}
}
uncomp_writer.WriteU(current_frame_info.base_q_index, 8);
uncomp_writer.WriteDeltaQ(current_frame_info.y_dc_delta_q);
uncomp_writer.WriteDeltaQ(current_frame_info.uv_dc_delta_q);
uncomp_writer.WriteDeltaQ(current_frame_info.uv_ac_delta_q);
uncomp_writer.WriteBit(false); // Segmentation enabled (TODO).
const s32 min_tile_cols_log2 = CalcMinLog2TileCols(current_frame_info.frame_size.width);
const s32 max_tile_cols_log2 = CalcMaxLog2TileCols(current_frame_info.frame_size.width);
const s32 tile_cols_log2_diff = current_frame_info.log2_tile_cols - min_tile_cols_log2;
const s32 tile_cols_log2_inc_mask = (1 << tile_cols_log2_diff) - 1;
// If it's less than the maximum, we need to add an extra 0 on the bitstream
// to indicate that it should stop reading.
if (current_frame_info.log2_tile_cols < max_tile_cols_log2) {
uncomp_writer.WriteU(tile_cols_log2_inc_mask << 1, tile_cols_log2_diff + 1);
} else {
uncomp_writer.WriteU(tile_cols_log2_inc_mask, tile_cols_log2_diff);
}
const bool tile_rows_log2_is_nonzero = current_frame_info.log2_tile_rows != 0;
uncomp_writer.WriteBit(tile_rows_log2_is_nonzero);
if (tile_rows_log2_is_nonzero) {
uncomp_writer.WriteBit(current_frame_info.log2_tile_rows > 1);
}
return uncomp_writer;
}
const std::vector<u8>& VP9::ComposeFrameHeader(const NvdecCommon::NvdecRegisters& state) {
std::vector<u8> bitstream;
{
Vp9FrameContainer curr_frame = GetCurrentFrame(state);
current_frame_info = curr_frame.info;
bitstream = std::move(curr_frame.bit_stream);
}
// The uncompressed header routine sets PrevProb parameters needed for the compressed header
auto uncomp_writer = ComposeUncompressedHeader();
std::vector<u8> compressed_header = ComposeCompressedHeader();
uncomp_writer.WriteU(static_cast<s32>(compressed_header.size()), 16);
uncomp_writer.Flush();
std::vector<u8> uncompressed_header = uncomp_writer.GetByteArray();
// Write headers and frame to buffer
frame.resize(uncompressed_header.size() + compressed_header.size() + bitstream.size());
std::memcpy(frame.data(), uncompressed_header.data(), uncompressed_header.size());
std::memcpy(frame.data() + uncompressed_header.size(), compressed_header.data(),
compressed_header.size());
std::memcpy(frame.data() + uncompressed_header.size() + compressed_header.size(),
bitstream.data(), bitstream.size());
// keep track of frame number
current_frame_number++;
grace_period--;
// don't display hidden frames
hidden = !current_frame_info.show_frame;
return frame;
}
VpxRangeEncoder::VpxRangeEncoder() {
Write(false);
}
VpxRangeEncoder::~VpxRangeEncoder() = default;
void VpxRangeEncoder::Write(s32 value, s32 value_size) {
for (s32 bit = value_size - 1; bit >= 0; bit--) {
Write(((value >> bit) & 1) != 0);
}
}
void VpxRangeEncoder::Write(bool bit) {
Write(bit, half_probability);
}
void VpxRangeEncoder::Write(bool bit, s32 probability) {
u32 local_range = range;
const u32 split = 1 + (((local_range - 1) * static_cast<u32>(probability)) >> 8);
local_range = split;
if (bit) {
low_value += split;
local_range = range - split;
}
s32 shift = norm_lut[local_range];
local_range <<= shift;
count += shift;
if (count >= 0) {
const s32 offset = shift - count;
if (((low_value << (offset - 1)) >> 31) != 0) {
const s32 current_pos = static_cast<s32>(base_stream.GetPosition());
base_stream.Seek(-1, Common::SeekOrigin::FromCurrentPos);
while (PeekByte() == 0xff) {
base_stream.WriteByte(0);
base_stream.Seek(-2, Common::SeekOrigin::FromCurrentPos);
}
base_stream.WriteByte(static_cast<u8>((PeekByte() + 1)));
base_stream.Seek(current_pos, Common::SeekOrigin::SetOrigin);
}
base_stream.WriteByte(static_cast<u8>((low_value >> (24 - offset))));
low_value <<= offset;
shift = count;
low_value &= 0xffffff;
count -= 8;
}
low_value <<= shift;
range = local_range;
}
void VpxRangeEncoder::End() {
for (std::size_t index = 0; index < 32; ++index) {
Write(false);
}
}
u8 VpxRangeEncoder::PeekByte() {
const u8 value = base_stream.ReadByte();
base_stream.Seek(-1, Common::SeekOrigin::FromCurrentPos);
return value;
}
VpxBitStreamWriter::VpxBitStreamWriter() = default;
VpxBitStreamWriter::~VpxBitStreamWriter() = default;
void VpxBitStreamWriter::WriteU(u32 value, u32 value_size) {
WriteBits(value, value_size);
}
void VpxBitStreamWriter::WriteS(s32 value, u32 value_size) {
const bool sign = value < 0;
if (sign) {
value = -value;
}
WriteBits(static_cast<u32>(value << 1) | (sign ? 1 : 0), value_size + 1);
}
void VpxBitStreamWriter::WriteDeltaQ(u32 value) {
const bool delta_coded = value != 0;
WriteBit(delta_coded);
if (delta_coded) {
WriteBits(value, 4);
}
}
void VpxBitStreamWriter::WriteBits(u32 value, u32 bit_count) {
s32 value_pos = 0;
s32 remaining = bit_count;
while (remaining > 0) {
s32 copy_size = remaining;
const s32 free = GetFreeBufferBits();
if (copy_size > free) {
copy_size = free;
}
const s32 mask = (1 << copy_size) - 1;
const s32 src_shift = (bit_count - value_pos) - copy_size;
const s32 dst_shift = (buffer_size - buffer_pos) - copy_size;
buffer |= ((value >> src_shift) & mask) << dst_shift;
value_pos += copy_size;
buffer_pos += copy_size;
remaining -= copy_size;
}
}
void VpxBitStreamWriter::WriteBit(bool state) {
WriteBits(state ? 1 : 0, 1);
}
s32 VpxBitStreamWriter::GetFreeBufferBits() {
if (buffer_pos == buffer_size) {
Flush();
}
return buffer_size - buffer_pos;
}
void VpxBitStreamWriter::Flush() {
if (buffer_pos == 0) {
return;
}
byte_array.push_back(static_cast<u8>(buffer));
buffer = 0;
buffer_pos = 0;
}
std::vector<u8>& VpxBitStreamWriter::GetByteArray() {
return byte_array;
}
const std::vector<u8>& VpxBitStreamWriter::GetByteArray() const {
return byte_array;
}
} // namespace Tegra::Decoder

197
src/video_core/command_classes/codecs/vp9.h Executable file
View File

@@ -0,0 +1,197 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <array>
#include <vector>
#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"
namespace Tegra {
class GPU;
enum class FrameType { KeyFrame = 0, InterFrame = 1 };
namespace Decoder {
/// The VpxRangeEncoder, and VpxBitStreamWriter classes are used to compose the
/// VP9 header bitstreams.
class VpxRangeEncoder {
public:
VpxRangeEncoder();
~VpxRangeEncoder();
VpxRangeEncoder(const VpxRangeEncoder&) = delete;
VpxRangeEncoder& operator=(const VpxRangeEncoder&) = delete;
VpxRangeEncoder(VpxRangeEncoder&&) = default;
VpxRangeEncoder& operator=(VpxRangeEncoder&&) = default;
/// Writes the rightmost value_size bits from value into the stream
void Write(s32 value, s32 value_size);
/// Writes a single bit with half probability
void Write(bool bit);
/// Writes a bit to the base_stream encoded with probability
void Write(bool bit, s32 probability);
/// Signal the end of the bitstream
void End();
[[nodiscard]] std::vector<u8>& GetBuffer() {
return base_stream.GetBuffer();
}
[[nodiscard]] const std::vector<u8>& GetBuffer() const {
return base_stream.GetBuffer();
}
private:
u8 PeekByte();
Common::Stream base_stream{};
u32 low_value{};
u32 range{0xff};
s32 count{-24};
s32 half_probability{128};
};
class VpxBitStreamWriter {
public:
VpxBitStreamWriter();
~VpxBitStreamWriter();
VpxBitStreamWriter(const VpxBitStreamWriter&) = delete;
VpxBitStreamWriter& operator=(const VpxBitStreamWriter&) = delete;
VpxBitStreamWriter(VpxBitStreamWriter&&) = default;
VpxBitStreamWriter& operator=(VpxBitStreamWriter&&) = default;
/// Write an unsigned integer value
void WriteU(u32 value, u32 value_size);
/// Write a signed integer value
void WriteS(s32 value, u32 value_size);
/// Based on 6.2.10 of VP9 Spec, writes a delta coded value
void WriteDeltaQ(u32 value);
/// Write a single bit.
void WriteBit(bool state);
/// Pushes current buffer into buffer_array, resets buffer
void Flush();
/// Returns byte_array
[[nodiscard]] std::vector<u8>& GetByteArray();
/// Returns const byte_array
[[nodiscard]] const std::vector<u8>& GetByteArray() const;
private:
/// Write bit_count bits from value into buffer
void WriteBits(u32 value, u32 bit_count);
/// Gets next available position in buffer, invokes Flush() if buffer is full
s32 GetFreeBufferBits();
s32 buffer_size{8};
s32 buffer{};
s32 buffer_pos{};
std::vector<u8> byte_array;
};
class VP9 {
public:
explicit VP9(GPU& gpu_);
~VP9();
VP9(const VP9&) = delete;
VP9& operator=(const VP9&) = delete;
VP9(VP9&&) = default;
VP9& operator=(VP9&&) = delete;
/// Composes the VP9 frame from the GPU state information. Based on the official VP9 spec
/// documentation
[[nodiscard]] const std::vector<u8>& ComposeFrameHeader(
const NvdecCommon::NvdecRegisters& state);
/// Returns true if the most recent frame was a hidden frame.
[[nodiscard]] bool WasFrameHidden() const {
return hidden;
}
private:
/// Generates compressed header probability updates in the bitstream writer
template <typename T, std::size_t N>
void WriteProbabilityUpdate(VpxRangeEncoder& writer, const std::array<T, N>& new_prob,
const std::array<T, N>& old_prob);
/// Generates compressed header probability updates in the bitstream writer
/// If probs are not equal, WriteProbabilityDelta is invoked
void WriteProbabilityUpdate(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob);
/// Generates compressed header probability deltas in the bitstream writer
void WriteProbabilityDelta(VpxRangeEncoder& writer, u8 new_prob, u8 old_prob);
/// Inverse of 6.3.4 Decode term subexp
void EncodeTermSubExp(VpxRangeEncoder& writer, s32 value);
/// Writes if the value is less than the test value
bool WriteLessThan(VpxRangeEncoder& writer, s32 value, s32 test);
/// Writes probability updates for the Coef probabilities
void WriteCoefProbabilityUpdate(VpxRangeEncoder& writer, s32 tx_mode,
const std::array<u8, 1728>& new_prob,
const std::array<u8, 1728>& old_prob);
/// Write probabilities for 4-byte aligned structures
template <typename T, std::size_t N>
void WriteProbabilityUpdateAligned4(VpxRangeEncoder& writer, const std::array<T, N>& new_prob,
const std::array<T, N>& old_prob);
/// Write motion vector probability updates. 6.3.17 in the spec
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);
/// 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);
/// Use NVDEC providied information to compose the headers for the current frame
[[nodiscard]] std::vector<u8> ComposeCompressedHeader();
[[nodiscard]] VpxBitStreamWriter ComposeUncompressedHeader();
GPU& gpu;
std::vector<u8> frame;
std::array<s8, 4> loop_filter_ref_deltas{};
std::array<s8, 2> loop_filter_mode_deltas{};
bool hidden = false;
s64 current_frame_number = -2; // since we buffer 2 frames
s32 grace_period = 6; // frame offsets need to stabilize
std::array<FrameContexts, 4> frame_ctxs{};
Vp9FrameContainer next_frame{};
Vp9FrameContainer next_next_frame{};
bool swap_next_golden{};
Vp9PictureInfo current_frame_info{};
Vp9EntropyProbs prev_frame_probs{};
s32 diff_update_probability = 252;
s32 frame_sync_code = 0x498342;
};
} // namespace Decoder
} // namespace Tegra

302
src/video_core/command_classes/codecs/vp9_types.h Executable file
View File

@@ -0,0 +1,302 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <array>
#include <cstring>
#include <vector>
#include "common/common_funcs.h"
#include "common/common_types.h"
namespace Tegra {
class GPU;
namespace Decoder {
struct Vp9FrameDimensions {
s16 width{};
s16 height{};
s16 luma_pitch{};
s16 chroma_pitch{};
};
static_assert(sizeof(Vp9FrameDimensions) == 0x8, "Vp9 Vp9FrameDimensions is an invalid size");
enum FrameFlags : u32 {
IsKeyFrame = 1 << 0,
LastFrameIsKeyFrame = 1 << 1,
FrameSizeChanged = 1 << 2,
ErrorResilientMode = 1 << 3,
LastShowFrame = 1 << 4,
IntraOnly = 1 << 5,
};
enum class TxSize {
Tx4x4 = 0, // 4x4 transform
Tx8x8 = 1, // 8x8 transform
Tx16x16 = 2, // 16x16 transform
Tx32x32 = 3, // 32x32 transform
TxSizes = 4
};
enum class TxMode {
Only4X4 = 0, // Only 4x4 transform used
Allow8X8 = 1, // Allow block transform size up to 8x8
Allow16X16 = 2, // Allow block transform size up to 16x16
Allow32X32 = 3, // Allow block transform size up to 32x32
TxModeSelect = 4, // Transform specified for each block
TxModes = 5
};
struct Segmentation {
u8 enabled{};
u8 update_map{};
u8 temporal_update{};
u8 abs_delta{};
std::array<u32, 8> feature_mask{};
std::array<std::array<s16, 4>, 8> feature_data{};
};
static_assert(sizeof(Segmentation) == 0x64, "Segmentation is an invalid size");
struct LoopFilter {
u8 mode_ref_delta_enabled{};
std::array<s8, 4> ref_deltas{};
std::array<s8, 2> mode_deltas{};
};
static_assert(sizeof(LoopFilter) == 0x7, "LoopFilter is an invalid size");
struct Vp9EntropyProbs {
std::array<u8, 36> y_mode_prob{};
std::array<u8, 64> partition_prob{};
std::array<u8, 1728> coef_probs{};
std::array<u8, 8> switchable_interp_prob{};
std::array<u8, 28> inter_mode_prob{};
std::array<u8, 4> intra_inter_prob{};
std::array<u8, 5> comp_inter_prob{};
std::array<u8, 10> single_ref_prob{};
std::array<u8, 5> comp_ref_prob{};
std::array<u8, 6> tx_32x32_prob{};
std::array<u8, 4> tx_16x16_prob{};
std::array<u8, 2> tx_8x8_prob{};
std::array<u8, 3> skip_probs{};
std::array<u8, 3> joints{};
std::array<u8, 2> sign{};
std::array<u8, 20> classes{};
std::array<u8, 2> class_0{};
std::array<u8, 20> prob_bits{};
std::array<u8, 12> class_0_fr{};
std::array<u8, 6> fr{};
std::array<u8, 2> class_0_hp{};
std::array<u8, 2> high_precision{};
};
static_assert(sizeof(Vp9EntropyProbs) == 0x7B4, "Vp9EntropyProbs is an invalid size");
struct Vp9PictureInfo {
bool is_key_frame{};
bool intra_only{};
bool last_frame_was_key{};
bool frame_size_changed{};
bool error_resilient_mode{};
bool last_frame_shown{};
bool show_frame{};
std::array<s8, 4> ref_frame_sign_bias{};
s32 base_q_index{};
s32 y_dc_delta_q{};
s32 uv_dc_delta_q{};
s32 uv_ac_delta_q{};
bool lossless{};
s32 transform_mode{};
bool allow_high_precision_mv{};
s32 interp_filter{};
s32 reference_mode{};
s8 comp_fixed_ref{};
std::array<s8, 2> comp_var_ref{};
s32 log2_tile_cols{};
s32 log2_tile_rows{};
bool segment_enabled{};
bool segment_map_update{};
bool segment_map_temporal_update{};
s32 segment_abs_delta{};
std::array<u32, 8> segment_feature_enable{};
std::array<std::array<s16, 4>, 8> segment_feature_data{};
bool mode_ref_delta_enabled{};
bool use_prev_in_find_mv_refs{};
std::array<s8, 4> ref_deltas{};
std::array<s8, 2> mode_deltas{};
Vp9EntropyProbs entropy{};
Vp9FrameDimensions frame_size{};
u8 first_level{};
u8 sharpness_level{};
u32 bitstream_size{};
std::array<u64, 4> frame_offsets{};
std::array<bool, 4> refresh_frame{};
};
struct Vp9FrameContainer {
Vp9PictureInfo info{};
std::vector<u8> bit_stream;
};
struct PictureInfo {
INSERT_PADDING_WORDS(12);
u32 bitstream_size{};
INSERT_PADDING_WORDS(5);
Vp9FrameDimensions last_frame_size{};
Vp9FrameDimensions golden_frame_size{};
Vp9FrameDimensions alt_frame_size{};
Vp9FrameDimensions current_frame_size{};
u32 vp9_flags{};
std::array<s8, 4> ref_frame_sign_bias{};
u8 first_level{};
u8 sharpness_level{};
u8 base_q_index{};
u8 y_dc_delta_q{};
u8 uv_ac_delta_q{};
u8 uv_dc_delta_q{};
u8 lossless{};
u8 tx_mode{};
u8 allow_high_precision_mv{};
u8 interp_filter{};
u8 reference_mode{};
s8 comp_fixed_ref{};
std::array<s8, 2> comp_var_ref{};
u8 log2_tile_cols{};
u8 log2_tile_rows{};
Segmentation segmentation{};
LoopFilter loop_filter{};
INSERT_PADDING_BYTES(5);
u32 surface_params{};
INSERT_PADDING_WORDS(3);
[[nodiscard]] Vp9PictureInfo Convert() const {
return {
.is_key_frame = (vp9_flags & FrameFlags::IsKeyFrame) != 0,
.intra_only = (vp9_flags & FrameFlags::IntraOnly) != 0,
.last_frame_was_key = (vp9_flags & FrameFlags::LastFrameIsKeyFrame) != 0,
.frame_size_changed = (vp9_flags & FrameFlags::FrameSizeChanged) != 0,
.error_resilient_mode = (vp9_flags & FrameFlags::ErrorResilientMode) != 0,
.last_frame_shown = (vp9_flags & FrameFlags::LastShowFrame) != 0,
.ref_frame_sign_bias = ref_frame_sign_bias,
.base_q_index = base_q_index,
.y_dc_delta_q = y_dc_delta_q,
.uv_dc_delta_q = uv_dc_delta_q,
.uv_ac_delta_q = uv_ac_delta_q,
.lossless = lossless != 0,
.transform_mode = tx_mode,
.allow_high_precision_mv = allow_high_precision_mv != 0,
.interp_filter = interp_filter,
.reference_mode = reference_mode,
.comp_fixed_ref = comp_fixed_ref,
.comp_var_ref = comp_var_ref,
.log2_tile_cols = log2_tile_cols,
.log2_tile_rows = log2_tile_rows,
.segment_enabled = segmentation.enabled != 0,
.segment_map_update = segmentation.update_map != 0,
.segment_map_temporal_update = segmentation.temporal_update != 0,
.segment_abs_delta = segmentation.abs_delta,
.segment_feature_enable = segmentation.feature_mask,
.segment_feature_data = segmentation.feature_data,
.mode_ref_delta_enabled = loop_filter.mode_ref_delta_enabled != 0,
.use_prev_in_find_mv_refs = !(vp9_flags == (FrameFlags::ErrorResilientMode)) &&
!(vp9_flags == (FrameFlags::FrameSizeChanged)) &&
!(vp9_flags == (FrameFlags::IntraOnly)) &&
(vp9_flags == (FrameFlags::LastShowFrame)) &&
!(vp9_flags == (FrameFlags::LastFrameIsKeyFrame)),
.ref_deltas = loop_filter.ref_deltas,
.mode_deltas = loop_filter.mode_deltas,
.frame_size = current_frame_size,
.first_level = first_level,
.sharpness_level = sharpness_level,
.bitstream_size = bitstream_size,
};
}
};
static_assert(sizeof(PictureInfo) == 0x100, "PictureInfo is an invalid size");
struct EntropyProbs {
INSERT_PADDING_BYTES(1024);
std::array<u8, 28> inter_mode_prob{};
std::array<u8, 4> intra_inter_prob{};
INSERT_PADDING_BYTES(80);
std::array<u8, 2> tx_8x8_prob{};
std::array<u8, 4> tx_16x16_prob{};
std::array<u8, 6> tx_32x32_prob{};
std::array<u8, 4> y_mode_prob_e8{};
std::array<std::array<u8, 8>, 4> y_mode_prob_e0e7{};
INSERT_PADDING_BYTES(64);
std::array<u8, 64> partition_prob{};
INSERT_PADDING_BYTES(10);
std::array<u8, 8> switchable_interp_prob{};
std::array<u8, 5> comp_inter_prob{};
std::array<u8, 3> skip_probs{};
INSERT_PADDING_BYTES(1);
std::array<u8, 3> joints{};
std::array<u8, 2> sign{};
std::array<u8, 2> class_0{};
std::array<u8, 6> fr{};
std::array<u8, 2> class_0_hp{};
std::array<u8, 2> high_precision{};
std::array<u8, 20> classes{};
std::array<u8, 12> class_0_fr{};
std::array<u8, 20> pred_bits{};
std::array<u8, 10> single_ref_prob{};
std::array<u8, 5> comp_ref_prob{};
INSERT_PADDING_BYTES(17);
std::array<u8, 2304> coef_probs{};
void Convert(Vp9EntropyProbs& fc) {
fc.inter_mode_prob = inter_mode_prob;
fc.intra_inter_prob = intra_inter_prob;
fc.tx_8x8_prob = tx_8x8_prob;
fc.tx_16x16_prob = tx_16x16_prob;
fc.tx_32x32_prob = tx_32x32_prob;
for (std::size_t i = 0; i < 4; i++) {
for (std::size_t j = 0; j < 9; j++) {
fc.y_mode_prob[j + 9 * i] = j < 8 ? y_mode_prob_e0e7[i][j] : y_mode_prob_e8[i];
}
}
fc.partition_prob = partition_prob;
fc.switchable_interp_prob = switchable_interp_prob;
fc.comp_inter_prob = comp_inter_prob;
fc.skip_probs = skip_probs;
fc.joints = joints;
fc.sign = sign;
fc.class_0 = class_0;
fc.fr = fr;
fc.class_0_hp = class_0_hp;
fc.high_precision = high_precision;
fc.classes = classes;
fc.class_0_fr = class_0_fr;
fc.prob_bits = pred_bits;
fc.single_ref_prob = single_ref_prob;
fc.comp_ref_prob = comp_ref_prob;
// Skip the 4th element as it goes unused
for (std::size_t i = 0; i < coef_probs.size(); i += 4) {
const std::size_t j = i - i / 4;
fc.coef_probs[j] = coef_probs[i];
fc.coef_probs[j + 1] = coef_probs[i + 1];
fc.coef_probs[j + 2] = coef_probs[i + 2];
}
}
};
static_assert(sizeof(EntropyProbs) == 0xEA0, "EntropyProbs is an invalid size");
enum class Ref { Last, Golden, AltRef };
struct RefPoolElement {
s64 frame{};
Ref ref{};
bool refresh{};
};
struct FrameContexts {
s64 from{};
bool adapted{};
Vp9EntropyProbs probs{};
};
}; // namespace Decoder
}; // namespace Tegra

39
src/video_core/command_classes/host1x.cpp Executable file
View File

@@ -0,0 +1,39 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "common/assert.h"
#include "video_core/command_classes/host1x.h"
#include "video_core/gpu.h"
Tegra::Host1x::Host1x(GPU& gpu_) : gpu(gpu_) {}
Tegra::Host1x::~Host1x() = default;
void Tegra::Host1x::StateWrite(u32 offset, u32 arguments) {
u8* const state_offset = reinterpret_cast<u8*>(&state) + offset * sizeof(u32);
std::memcpy(state_offset, &arguments, sizeof(u32));
}
void Tegra::Host1x::ProcessMethod(Method method, const std::vector<u32>& arguments) {
StateWrite(static_cast<u32>(method), arguments[0]);
switch (method) {
case Method::WaitSyncpt:
Execute(arguments[0]);
break;
case Method::LoadSyncptPayload32:
syncpoint_value = arguments[0];
break;
case Method::WaitSyncpt32:
Execute(arguments[0]);
break;
default:
UNIMPLEMENTED_MSG("Host1x method 0x{:X}", static_cast<u32>(method));
break;
}
}
void Tegra::Host1x::Execute(u32 data) {
// This method waits on a valid syncpoint.
// TODO: Implement when proper Async is in place
}

78
src/video_core/command_classes/host1x.h Executable file
View File

@@ -0,0 +1,78 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <vector>
#include "common/common_funcs.h"
#include "common/common_types.h"
namespace Tegra {
class GPU;
class Nvdec;
class Host1x {
public:
struct Host1xClassRegisters {
u32 incr_syncpt{};
u32 incr_syncpt_ctrl{};
u32 incr_syncpt_error{};
INSERT_PADDING_WORDS(5);
u32 wait_syncpt{};
u32 wait_syncpt_base{};
u32 wait_syncpt_incr{};
u32 load_syncpt_base{};
u32 incr_syncpt_base{};
u32 clear{};
u32 wait{};
u32 wait_with_interrupt{};
u32 delay_use{};
u32 tick_count_high{};
u32 tick_count_low{};
u32 tick_ctrl{};
INSERT_PADDING_WORDS(23);
u32 ind_ctrl{};
u32 ind_off2{};
u32 ind_off{};
std::array<u32, 31> ind_data{};
INSERT_PADDING_WORDS(1);
u32 load_syncpoint_payload32{};
u32 stall_ctrl{};
u32 wait_syncpt32{};
u32 wait_syncpt_base32{};
u32 load_syncpt_base32{};
u32 incr_syncpt_base32{};
u32 stall_count_high{};
u32 stall_count_low{};
u32 xref_ctrl{};
u32 channel_xref_high{};
u32 channel_xref_low{};
};
static_assert(sizeof(Host1xClassRegisters) == 0x164, "Host1xClassRegisters is an invalid size");
enum class Method : u32 {
WaitSyncpt = offsetof(Host1xClassRegisters, wait_syncpt) / 4,
LoadSyncptPayload32 = offsetof(Host1xClassRegisters, load_syncpoint_payload32) / 4,
WaitSyncpt32 = offsetof(Host1xClassRegisters, wait_syncpt32) / 4,
};
explicit Host1x(GPU& gpu);
~Host1x();
/// Writes the method into the state, Invoke Execute() if encountered
void ProcessMethod(Method method, const std::vector<u32>& arguments);
private:
/// For Host1x, execute is waiting on a syncpoint previously written into the state
void Execute(u32 data);
/// Write argument into the provided offset
void StateWrite(u32 offset, u32 arguments);
u32 syncpoint_value{};
Host1xClassRegisters state{};
GPU& gpu;
};
} // namespace Tegra

48
src/video_core/command_classes/nvdec.cpp Executable file
View File

@@ -0,0 +1,48 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "common/assert.h"
#include "video_core/command_classes/nvdec.h"
#include "video_core/gpu.h"
namespace Tegra {
Nvdec::Nvdec(GPU& gpu_) : gpu(gpu_), codec(std::make_unique<Codec>(gpu)) {}
Nvdec::~Nvdec() = default;
void Nvdec::ProcessMethod(Method method, const std::vector<u32>& arguments) {
if (method == Method::SetVideoCodec) {
codec->StateWrite(static_cast<u32>(method), arguments[0]);
} else {
codec->StateWrite(static_cast<u32>(method), static_cast<u64>(arguments[0]) << 8);
}
switch (method) {
case Method::SetVideoCodec:
codec->SetTargetCodec(static_cast<NvdecCommon::VideoCodec>(arguments[0]));
break;
case Method::Execute:
Execute();
break;
}
}
AVFramePtr Nvdec::GetFrame() {
return codec->GetCurrentFrame();
}
void Nvdec::Execute() {
switch (codec->GetCurrentCodec()) {
case NvdecCommon::VideoCodec::H264:
case NvdecCommon::VideoCodec::Vp9:
codec->Decode();
break;
default:
UNIMPLEMENTED_MSG("Unknown codec {}", static_cast<u32>(codec->GetCurrentCodec()));
break;
}
}
} // namespace Tegra

38
src/video_core/command_classes/nvdec.h Executable file
View File

@@ -0,0 +1,38 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <memory>
#include <vector>
#include "common/common_types.h"
#include "video_core/command_classes/codecs/codec.h"
namespace Tegra {
class GPU;
class Nvdec {
public:
enum class Method : u32 {
SetVideoCodec = 0x80,
Execute = 0xc0,
};
explicit Nvdec(GPU& gpu);
~Nvdec();
/// Writes the method into the state, Invoke Execute() if encountered
void ProcessMethod(Method method, const std::vector<u32>& arguments);
/// Return most recently decoded frame
[[nodiscard]] AVFramePtr GetFrame();
private:
/// Invoke codec to decode a frame
void Execute();
GPU& gpu;
std::unique_ptr<Codec> codec;
};
} // namespace Tegra

48
src/video_core/command_classes/nvdec_common.h Executable file
View File

@@ -0,0 +1,48 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include "common/common_funcs.h"
#include "common/common_types.h"
namespace Tegra::NvdecCommon {
struct NvdecRegisters {
INSERT_PADDING_WORDS(256);
u64 set_codec_id{};
INSERT_PADDING_WORDS(254);
u64 set_platform_id{};
u64 picture_info_offset{};
u64 frame_bitstream_offset{};
u64 frame_number{};
u64 h264_slice_data_offsets{};
u64 h264_mv_dump_offset{};
INSERT_PADDING_WORDS(6);
u64 frame_stats_offset{};
u64 h264_last_surface_luma_offset{};
u64 h264_last_surface_chroma_offset{};
std::array<u64, 17> surface_luma_offset{};
std::array<u64, 17> surface_chroma_offset{};
INSERT_PADDING_WORDS(132);
u64 vp9_entropy_probs_offset{};
u64 vp9_backward_updates_offset{};
u64 vp9_last_frame_segmap_offset{};
u64 vp9_curr_frame_segmap_offset{};
INSERT_PADDING_WORDS(2);
u64 vp9_last_frame_mvs_offset{};
u64 vp9_curr_frame_mvs_offset{};
INSERT_PADDING_WORDS(2);
};
static_assert(sizeof(NvdecRegisters) == (0xBC0), "NvdecRegisters is incorrect size");
enum class VideoCodec : u32 {
None = 0x0,
H264 = 0x3,
Vp8 = 0x5,
H265 = 0x7,
Vp9 = 0x9,
};
} // namespace Tegra::NvdecCommon

60
src/video_core/command_classes/sync_manager.cpp Executable file
View File

@@ -0,0 +1,60 @@
// MIT License
//
// Copyright (c) Ryujinx Team and Contributors
//
// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and
// associated documentation files (the "Software"), to deal in the Software without restriction,
// including without limitation the rights to use, copy, modify, merge, publish, distribute,
// sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in all copies or
// substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT
// NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
// DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
//
#include <algorithm>
#include "sync_manager.h"
#include "video_core/gpu.h"
namespace Tegra {
SyncptIncrManager::SyncptIncrManager(GPU& gpu_) : gpu(gpu_) {}
SyncptIncrManager::~SyncptIncrManager() = default;
void SyncptIncrManager::Increment(u32 id) {
increments.emplace_back(0, 0, id, true);
IncrementAllDone();
}
u32 SyncptIncrManager::IncrementWhenDone(u32 class_id, u32 id) {
const u32 handle = current_id++;
increments.emplace_back(handle, class_id, id);
return handle;
}
void SyncptIncrManager::SignalDone(u32 handle) {
const auto done_incr =
std::find_if(increments.begin(), increments.end(),
[handle](const SyncptIncr& incr) { return incr.id == handle; });
if (done_incr != increments.cend()) {
done_incr->complete = true;
}
IncrementAllDone();
}
void SyncptIncrManager::IncrementAllDone() {
std::size_t done_count = 0;
for (; done_count < increments.size(); ++done_count) {
if (!increments[done_count].complete) {
break;
}
gpu.IncrementSyncPoint(increments[done_count].syncpt_id);
}
increments.erase(increments.begin(), increments.begin() + done_count);
}
} // namespace Tegra

64
src/video_core/command_classes/sync_manager.h Executable file
View File

@@ -0,0 +1,64 @@
// MIT License
//
// Copyright (c) Ryujinx Team and Contributors
//
// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and
// associated documentation files (the "Software"), to deal in the Software without restriction,
// including without limitation the rights to use, copy, modify, merge, publish, distribute,
// sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in all copies or
// substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT
// NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
// DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
//
#pragma once
#include <mutex>
#include <vector>
#include "common/common_types.h"
namespace Tegra {
class GPU;
struct SyncptIncr {
u32 id;
u32 class_id;
u32 syncpt_id;
bool complete;
SyncptIncr(u32 id_, u32 class_id_, u32 syncpt_id_, bool done = false)
: id(id_), class_id(class_id_), syncpt_id(syncpt_id_), complete(done) {}
};
class SyncptIncrManager {
public:
explicit SyncptIncrManager(GPU& gpu);
~SyncptIncrManager();
/// Add syncpoint id and increment all
void Increment(u32 id);
/// Returns a handle to increment later
u32 IncrementWhenDone(u32 class_id, u32 id);
/// IncrememntAllDone, including handle
void SignalDone(u32 handle);
/// Increment all sequential pending increments that are already done.
void IncrementAllDone();
private:
std::vector<SyncptIncr> increments;
std::mutex increment_lock;
u32 current_id{};
GPU& gpu;
};
} // namespace Tegra

175
src/video_core/command_classes/vic.cpp Executable file
View File

@@ -0,0 +1,175 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <array>
#include "common/assert.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/memory_manager.h"
#include "video_core/textures/decoders.h"
extern "C" {
#include <libswscale/swscale.h>
}
namespace Tegra {
Vic::Vic(GPU& gpu_, std::shared_ptr<Nvdec> nvdec_processor_)
: gpu(gpu_), nvdec_processor(std::move(nvdec_processor_)) {}
Vic::~Vic() = default;
void Vic::VicStateWrite(u32 offset, u32 arguments) {
u8* const state_offset = reinterpret_cast<u8*>(&vic_state) + offset * sizeof(u32);
std::memcpy(state_offset, &arguments, sizeof(u32));
}
void Vic::ProcessMethod(Method method, const std::vector<u32>& arguments) {
LOG_DEBUG(HW_GPU, "Vic method 0x{:X}", method);
VicStateWrite(static_cast<u32>(method), arguments[0]);
const u64 arg = static_cast<u64>(arguments[0]) << 8;
switch (method) {
case Method::Execute:
Execute();
break;
case Method::SetConfigStructOffset:
config_struct_address = arg;
break;
case Method::SetOutputSurfaceLumaOffset:
output_surface_luma_address = arg;
break;
case Method::SetOutputSurfaceChromaUOffset:
output_surface_chroma_u_address = arg;
break;
case Method::SetOutputSurfaceChromaVOffset:
output_surface_chroma_v_address = arg;
break;
default:
break;
}
}
void Vic::Execute() {
if (output_surface_luma_address == 0) {
LOG_ERROR(Service_NVDRV, "VIC Luma address not set. Recieved 0x{:X}",
vic_state.output_surface.luma_offset);
return;
}
const VicConfig config{gpu.MemoryManager().Read<u64>(config_struct_address + 0x20)};
const AVFramePtr frame_ptr = nvdec_processor->GetFrame();
const auto* frame = frame_ptr.get();
if (!frame || frame->width == 0 || frame->height == 0) {
return;
}
const VideoPixelFormat pixel_format =
static_cast<VideoPixelFormat>(config.pixel_format.Value());
switch (pixel_format) {
case VideoPixelFormat::BGRA8:
case VideoPixelFormat::RGBA8: {
LOG_TRACE(Service_NVDRV, "Writing RGB Frame");
if (scaler_ctx == nullptr || frame->width != scaler_width ||
frame->height != scaler_height) {
const AVPixelFormat target_format =
(pixel_format == VideoPixelFormat::RGBA8) ? AV_PIX_FMT_RGBA : AV_PIX_FMT_BGRA;
sws_freeContext(scaler_ctx);
scaler_ctx = nullptr;
// FFmpeg returns all frames in YUV420, convert it into expected format
scaler_ctx =
sws_getContext(frame->width, frame->height, AV_PIX_FMT_YUV420P, frame->width,
frame->height, target_format, 0, nullptr, nullptr, nullptr);
scaler_width = frame->width;
scaler_height = frame->height;
}
// Get Converted frame
const std::size_t linear_size = frame->width * frame->height * 4;
using AVMallocPtr = std::unique_ptr<u8, decltype(&av_free)>;
AVMallocPtr converted_frame_buffer{static_cast<u8*>(av_malloc(linear_size)), av_free};
const int converted_stride{frame->width * 4};
u8* const converted_frame_buf_addr{converted_frame_buffer.get()};
sws_scale(scaler_ctx, frame->data, frame->linesize, 0, frame->height,
&converted_frame_buf_addr, &converted_stride);
const u32 blk_kind = static_cast<u32>(config.block_linear_kind);
if (blk_kind != 0) {
// swizzle pitch linear to block linear
const u32 block_height = static_cast<u32>(config.block_linear_height_log2);
const auto size = Tegra::Texture::CalculateSize(true, 4, frame->width, frame->height, 1,
block_height, 0);
std::vector<u8> swizzled_data(size);
Tegra::Texture::SwizzleSubrect(frame->width, frame->height, frame->width * 4,
frame->width, 4, swizzled_data.data(),
converted_frame_buffer.get(), block_height, 0, 0);
gpu.MemoryManager().WriteBlock(output_surface_luma_address, swizzled_data.data(), size);
gpu.Maxwell3D().OnMemoryWrite();
} else {
// send pitch linear frame
gpu.MemoryManager().WriteBlock(output_surface_luma_address, converted_frame_buf_addr,
linear_size);
gpu.Maxwell3D().OnMemoryWrite();
}
break;
}
case VideoPixelFormat::Yuv420: {
LOG_TRACE(Service_NVDRV, "Writing YUV420 Frame");
const std::size_t surface_width = config.surface_width_minus1 + 1;
const std::size_t surface_height = config.surface_height_minus1 + 1;
const std::size_t half_width = surface_width / 2;
const std::size_t half_height = config.surface_height_minus1 / 2;
const std::size_t aligned_width = (surface_width + 0xff) & ~0xff;
const auto* luma_ptr = frame->data[0];
const auto* chroma_b_ptr = frame->data[1];
const auto* chroma_r_ptr = frame->data[2];
const auto stride = frame->linesize[0];
const auto half_stride = frame->linesize[1];
std::vector<u8> luma_buffer(aligned_width * surface_height);
std::vector<u8> chroma_buffer(aligned_width * half_height);
// Populate luma buffer
for (std::size_t y = 0; y < surface_height - 1; ++y) {
std::size_t src = y * stride;
std::size_t dst = y * aligned_width;
std::size_t size = surface_width;
for (std::size_t offset = 0; offset < size; ++offset) {
luma_buffer[dst + offset] = luma_ptr[src + offset];
}
}
gpu.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(),
luma_buffer.size());
// Populate chroma buffer from both channels with interleaving.
for (std::size_t y = 0; y < half_height; ++y) {
std::size_t src = y * half_stride;
std::size_t dst = y * aligned_width;
for (std::size_t x = 0; x < half_width; ++x) {
chroma_buffer[dst + x * 2] = chroma_b_ptr[src + x];
chroma_buffer[dst + x * 2 + 1] = chroma_r_ptr[src + x];
}
}
gpu.MemoryManager().WriteBlock(output_surface_chroma_u_address, chroma_buffer.data(),
chroma_buffer.size());
gpu.Maxwell3D().OnMemoryWrite();
break;
}
default:
UNIMPLEMENTED_MSG("Unknown video pixel format {}", config.pixel_format.Value());
break;
}
}
} // namespace Tegra

110
src/video_core/command_classes/vic.h Executable file
View File

@@ -0,0 +1,110 @@
// Copyright 2020 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <memory>
#include <vector>
#include "common/bit_field.h"
#include "common/common_types.h"
struct SwsContext;
namespace Tegra {
class GPU;
class Nvdec;
struct PlaneOffsets {
u32 luma_offset{};
u32 chroma_u_offset{};
u32 chroma_v_offset{};
};
struct VicRegisters {
INSERT_PADDING_WORDS(64);
u32 nop{};
INSERT_PADDING_WORDS(15);
u32 pm_trigger{};
INSERT_PADDING_WORDS(47);
u32 set_application_id{};
u32 set_watchdog_timer{};
INSERT_PADDING_WORDS(17);
u32 context_save_area{};
u32 context_switch{};
INSERT_PADDING_WORDS(43);
u32 execute{};
INSERT_PADDING_WORDS(63);
std::array<std::array<PlaneOffsets, 8>, 8> surfacex_slots{};
u32 picture_index{};
u32 control_params{};
u32 config_struct_offset{};
u32 filter_struct_offset{};
u32 palette_offset{};
u32 hist_offset{};
u32 context_id{};
u32 fce_ucode_size{};
PlaneOffsets output_surface{};
u32 fce_ucode_offset{};
INSERT_PADDING_WORDS(4);
std::array<u32, 8> slot_context_id{};
INSERT_PADDING_WORDS(16);
};
static_assert(sizeof(VicRegisters) == 0x7A0, "VicRegisters is an invalid size");
class Vic {
public:
enum class Method : u32 {
Execute = 0xc0,
SetControlParams = 0x1c1,
SetConfigStructOffset = 0x1c2,
SetOutputSurfaceLumaOffset = 0x1c8,
SetOutputSurfaceChromaUOffset = 0x1c9,
SetOutputSurfaceChromaVOffset = 0x1ca
};
explicit Vic(GPU& gpu, std::shared_ptr<Nvdec> nvdec_processor);
~Vic();
/// Write to the device state.
void ProcessMethod(Method method, const std::vector<u32>& arguments);
private:
void Execute();
void VicStateWrite(u32 offset, u32 arguments);
VicRegisters vic_state{};
enum class VideoPixelFormat : u64_le {
RGBA8 = 0x1f,
BGRA8 = 0x20,
Yuv420 = 0x44,
};
union VicConfig {
u64_le raw{};
BitField<0, 7, u64_le> pixel_format;
BitField<7, 2, u64_le> chroma_loc_horiz;
BitField<9, 2, u64_le> chroma_loc_vert;
BitField<11, 4, u64_le> block_linear_kind;
BitField<15, 4, u64_le> block_linear_height_log2;
BitField<19, 3, u64_le> reserved0;
BitField<22, 10, u64_le> reserved1;
BitField<32, 14, u64_le> surface_width_minus1;
BitField<46, 14, u64_le> surface_height_minus1;
};
GPU& gpu;
std::shared_ptr<Tegra::Nvdec> nvdec_processor;
GPUVAddr config_struct_address{};
GPUVAddr output_surface_luma_address{};
GPUVAddr output_surface_chroma_u_address{};
GPUVAddr output_surface_chroma_v_address{};
SwsContext* scaler_ctx{};
s32 scaler_width{};
s32 scaler_height{};
};
} // namespace Tegra