Skip to content
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

Fix race condition in CUDA, ROCm, and TensorRT EP GetKernelRegistry() implementations. #10200

Merged
merged 7 commits into from
Mar 2, 2022
Merged

Fix race condition in CUDA, ROCm, and TensorRT EP GetKernelRegistry() implementations. #10200

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
c73f198
Fix race condition in CUDA, ROCm, and TensorRT EP GetKernelRegistry()…
edgchen1 Jan 5, 2022
0a68c6a
Use function-local static variable instead.
edgchen1 Jan 7, 2022
794f3d3
Merge remote-tracking branch 'origin/master' into edgchen1/ep_get_ker…
edgchen1 Jan 7, 2022
cc12013
Revert "Use function-local static variable instead."
edgchen1 Jan 8, 2022
5b8db3f
Merge branch 'master' into edgchen1/ep_get_kernel_registry_thread_safety
edgchen1 Feb 24, 2022
03c63af
Merge remote-tracking branch 'origin/master' into edgchen1/ep_get_ker…
edgchen1 Mar 1, 2022
97e4c85
Revert "Revert "Use function-local static variable instead.""
edgchen1 Mar 1, 2022
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
23 changes: 12 additions & 11 deletions onnxruntime/core/providers/cuda/cuda_execution_provider.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2073,22 +2073,23 @@ static Status RegisterCudaKernels(KernelRegistry& kernel_registry) {

} // namespace cuda

static std::shared_ptr<onnxruntime::KernelRegistry> s_kernel_registry;
static std::shared_ptr<KernelRegistry>& CudaKernelRegistry() {
// static local variable ensures thread-safe initialization
static std::shared_ptr<KernelRegistry> cuda_kernel_registry = []() {
std::shared_ptr<KernelRegistry> registry = KernelRegistry::Create();
ORT_THROW_IF_ERROR(cuda::RegisterCudaKernels(*registry));
return registry;
}();

return cuda_kernel_registry;
}

void Shutdown_DeleteRegistry() {
s_kernel_registry.reset();
CudaKernelRegistry().reset();
}

std::shared_ptr<KernelRegistry> CUDAExecutionProvider::GetKernelRegistry() const {
if (!s_kernel_registry) {
s_kernel_registry = KernelRegistry::Create();
auto status = cuda::RegisterCudaKernels(*s_kernel_registry);
if (!status.IsOK())
s_kernel_registry.reset();
ORT_THROW_IF_ERROR(status);
}

return s_kernel_registry;
return CudaKernelRegistry();
pranavsharma marked this conversation as resolved.
Show resolved Hide resolved
}

static bool RNNNeedFallbackToCPU(const onnxruntime::Node& node,
Expand Down
Loading