Skip to content

Enable Intel GPU support in torchcodec on Linux (xpu device) #558

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Enable Intel GPU support in torchcodec on Linux (xpu device) #558

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions setup.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -112,6 +112,7 @@ def _build_all_extensions_with_cmake(self):
torch_dir = Path(torch.utils.cmake_prefix_path) / "Torch"
cmake_build_type = os.environ.get("CMAKE_BUILD_TYPE", "Release")
enable_cuda = os.environ.get("ENABLE_CUDA", "")
enable_xpu = os.environ.get("ENABLE_XPU", "")
python_version = sys.version_info
cmake_args = [
f"-DCMAKE_INSTALL_PREFIX={self._install_prefix}",
Expand All @@ -120,6 +121,7 @@ def _build_all_extensions_with_cmake(self):
f"-DCMAKE_BUILD_TYPE={cmake_build_type}",
f"-DPYTHON_VERSION={python_version.major}.{python_version.minor}",
f"-DENABLE_CUDA={enable_cuda}",
f"-DENABLE_XPU={enable_xpu}",
]

Path(self.build_temp).mkdir(parents=True, exist_ok=True)
Expand Down
21 changes: 19 additions & 2 deletions src/torchcodec/decoders/_core/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,15 @@ find_package(Torch REQUIRED)
find_package(Python3 ${PYTHON_VERSION} EXACT COMPONENTS Development)

set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wextra -pedantic -Werror ${TORCH_CXX_FLAGS}")
if(ENABLE_CUDA)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DENABLE_CUDA")
endif()
if(ENABLE_XPU)
find_package(PkgConfig REQUIRED)
pkg_check_modules(L0 REQUIRED IMPORTED_TARGET level-zero)
pkg_check_modules(LIBVA REQUIRED IMPORTED_TARGET libva)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DENABLE_XPU")
endif()

function(make_torchcodec_sublibrary
library_name
Expand Down Expand Up @@ -61,12 +70,15 @@ function(make_torchcodec_libraries
AVIOContextHolder.cpp
FFMPEGCommon.cpp
VideoDecoder.cpp
CPUOnlyDevice.cpp
)

if(ENABLE_CUDA)
list(APPEND decoder_sources CudaDevice.cpp)
else()
list(APPEND decoder_sources CPUOnlyDevice.cpp)
endif()

if(ENABLE_XPU)
list(APPEND decoder_sources XpuDevice.cpp)
endif()

set(decoder_library_dependencies
Expand All @@ -81,6 +93,11 @@ function(make_torchcodec_libraries
)
endif()

if(ENABLE_XPU)
list(APPEND decoder_library_dependencies
PkgConfig::L0 PkgConfig::LIBVA)
endif()

make_torchcodec_sublibrary(
"${decoder_library_name}"
SHARED
Expand Down
31 changes: 31 additions & 0 deletions src/torchcodec/decoders/_core/CPUOnlyDevice.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@ namespace facebook::torchcodec {
TORCH_CHECK(false, "Unsupported device: " + device.str());
}

#ifndef ENABLE_CUDA
void convertAVFrameToFrameOutputOnCuda(
const torch::Device& device,
[[maybe_unused]] const VideoDecoder::VideoStreamOptions& videoStreamOptions,
Expand All @@ -40,5 +41,35 @@ std::optional<const AVCodec*> findCudaCodec(
[[maybe_unused]] const AVCodecID& codecId) {
throwUnsupportedDeviceError(device);
}
#endif // ENABLE_CUDA

#ifndef ENABLE_XPU
void convertAVFrameToFrameOutputOnXpu(
const torch::Device& device,
[[maybe_unused]] const VideoDecoder::VideoStreamOptions& videoStreamOptions,
[[maybe_unused]] UniqueAVFrame& avFrame,
[[maybe_unused]] VideoDecoder::FrameOutput& frameOutput,
[[maybe_unused]] std::optional<torch::Tensor> preAllocatedOutputTensor) {
throwUnsupportedDeviceError(device);
}

void initializeContextOnXpu(
const torch::Device& device,
[[maybe_unused]] AVCodecContext* codecContext) {
throwUnsupportedDeviceError(device);
}

void releaseContextOnXpu(
const torch::Device& device,
[[maybe_unused]] AVCodecContext* codecContext) {
throwUnsupportedDeviceError(device);
}

std::optional<const AVCodec*> findXpuCodec(
const torch::Device& device,
[[maybe_unused]] const AVCodecID& codecId) {
throwUnsupportedDeviceError(device);
}
#endif // ENABLE_XPU

} // namespace facebook::torchcodec
19 changes: 19 additions & 0 deletions src/torchcodec/decoders/_core/DeviceInterface.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,19 +29,38 @@ void initializeContextOnCuda(
const torch::Device& device,
AVCodecContext* codecContext);

void initializeContextOnXpu(
const torch::Device& device,
AVCodecContext* codecContext);

void convertAVFrameToFrameOutputOnCuda(
const torch::Device& device,
const VideoDecoder::VideoStreamOptions& videoStreamOptions,
UniqueAVFrame& avFrame,
VideoDecoder::FrameOutput& frameOutput,
std::optional<torch::Tensor> preAllocatedOutputTensor = std::nullopt);

void convertAVFrameToFrameOutputOnXpu(
const torch::Device& device,
const VideoDecoder::VideoStreamOptions& videoStreamOptions,
UniqueAVFrame& avFrame,
VideoDecoder::FrameOutput& frameOutput,
std::optional<torch::Tensor> preAllocatedOutputTensor = std::nullopt);

void releaseContextOnCuda(
const torch::Device& device,
AVCodecContext* codecContext);

void releaseContextOnXpu(
const torch::Device& device,
AVCodecContext* codecContext);

std::optional<const AVCodec*> findCudaCodec(
const torch::Device& device,
const AVCodecID& codecId);

std::optional<const AVCodec*> findXpuCodec(
const torch::Device& device,
const AVCodecID& codecId);

} // namespace facebook::torchcodec
34 changes: 27 additions & 7 deletions src/torchcodec/decoders/_core/VideoDecoder.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -101,6 +101,8 @@ VideoDecoder::~VideoDecoder() {
if (device.type() == torch::kCPU) {
} else if (device.type() == torch::kCUDA) {
releaseContextOnCuda(device, streamInfo.codecContext.get());
} else if (device.type() == torch::kXPU) {
releaseContextOnXpu(device, streamInfo.codecContext.get());
} else {
TORCH_CHECK(false, "Invalid device type: " + device.str());
}
Expand Down Expand Up @@ -429,10 +431,16 @@ void VideoDecoder::addStream(

// TODO_CODE_QUALITY it's pretty meh to have a video-specific logic within
// addStream() which is supposed to be generic
if (mediaType == AVMEDIA_TYPE_VIDEO && device.type() == torch::kCUDA) {
avCodec = makeAVCodecOnlyUseForCallingAVFindBestStream(
findCudaCodec(device, streamInfo.stream->codecpar->codec_id)
.value_or(avCodec));
if (mediaType == AVMEDIA_TYPE_VIDEO) {
if (device.type() == torch::kCUDA) {
avCodec = makeAVCodecOnlyUseForCallingAVFindBestStream(
findCudaCodec(device, streamInfo.stream->codecpar->codec_id)
.value_or(avCodec));
} else if (device.type() == torch::kXPU) {
avCodec = makeAVCodecOnlyUseForCallingAVFindBestStream(
findXpuCodec(device, streamInfo.stream->codecpar->codec_id)
.value_or(avCodec));
}
}

AVCodecContext* codecContext = avcodec_alloc_context3(avCodec);
Expand All @@ -447,8 +455,12 @@ void VideoDecoder::addStream(
streamInfo.codecContext->pkt_timebase = streamInfo.stream->time_base;

// TODO_CODE_QUALITY same as above.
if (mediaType == AVMEDIA_TYPE_VIDEO && device.type() == torch::kCUDA) {
initializeContextOnCuda(device, codecContext);
if (mediaType == AVMEDIA_TYPE_VIDEO) {
if (device.type() == torch::kCUDA) {
initializeContextOnCuda(device, codecContext);
} else if (device.type() == torch::kXPU) {
initializeContextOnXpu(device, codecContext);
}
}

retVal = avcodec_open2(streamInfo.codecContext.get(), avCodec, nullptr);
Expand Down Expand Up @@ -476,7 +488,8 @@ void VideoDecoder::addVideoStream(
const VideoStreamOptions& videoStreamOptions) {
TORCH_CHECK(
videoStreamOptions.device.type() == torch::kCPU ||
videoStreamOptions.device.type() == torch::kCUDA,
videoStreamOptions.device.type() == torch::kCUDA ||
videoStreamOptions.device.type() == torch::kXPU,
"Invalid device type: " + videoStreamOptions.device.str());

addStream(
Expand Down Expand Up @@ -1226,6 +1239,13 @@ VideoDecoder::FrameOutput VideoDecoder::convertAVFrameToFrameOutput(
avFrame,
frameOutput,
preAllocatedOutputTensor);
} else if (streamInfo.videoStreamOptions.device.type() == torch::kXPU) {
convertAVFrameToFrameOutputOnXpu(
streamInfo.videoStreamOptions.device,
streamInfo.videoStreamOptions,
avFrame,
frameOutput,
preAllocatedOutputTensor);
} else {
TORCH_CHECK(
false,
Expand Down
5 changes: 4 additions & 1 deletion src/torchcodec/decoders/_core/VideoDecoderOps.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -210,10 +210,13 @@ void _add_video_stream(
} else if (device.value().rfind("cuda", 0) == 0) { // starts with "cuda"
std::string deviceStr(device.value());
videoStreamOptions.device = torch::Device(deviceStr);
} else if (device.value().rfind("xpu", 0) == 0) { // starts with "xpu"
std::string deviceStr(device.value());
videoStreamOptions.device = torch::Device(deviceStr);
} else {
throw std::runtime_error(
"Invalid device=" + std::string(device.value()) +
". device must be either cpu or cuda.");
". device must be either cpu, cuda or xpu.");
}
}

Expand Down
Loading