mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-07 12:21:27 +01:00
8f0db9bbbb
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/12043 Re-trying D9979976, this time with all call sites fixed. D9979976 got reverted because there was a call site that wasn't covered by sandcastle it seems. I fixed it and used 'grep' to ensure there aren't any more call sites in fbsource. Reviewed By: ezyang Differential Revision: D10026392 fbshipit-source-id: cd341514a8e53a40147ea0ee3e52f63bb6444157
28 lines
910 B
C++
28 lines
910 B
C++
#include "caffe2/core/context_gpu.h"
|
|
#include "caffe2/operators/conv_op_shared.h"
|
|
|
|
namespace caffe2 {
|
|
|
|
template <>
|
|
void createSharedBuffer<CUDAContext>(Workspace* ws) {
|
|
auto* mutexPtr = ws->CreateBlob("__CAFFE2_SHARED_CONV_BUFFER_CUDA_MUTEX__")
|
|
->GetMutable<std::unique_ptr<std::mutex>>();
|
|
mutexPtr->reset(new std::mutex());
|
|
ws->CreateBlob("__CAFFE2_SHARED_CONV_BUFFER_CUDA__");
|
|
}
|
|
|
|
template <>
|
|
void runWithSharedBuffer<CUDAContext>(
|
|
Workspace* ws,
|
|
std::function<void(Tensor* buffer)> f) {
|
|
auto* mutexBlob = ws->GetBlob("__CAFFE2_SHARED_CONV_BUFFER_CUDA_MUTEX__");
|
|
CAFFE_ENFORCE(mutexBlob, "Must call createSharedBuffer() first");
|
|
|
|
auto* mutexPtr = mutexBlob->GetMutable<std::unique_ptr<std::mutex>>();
|
|
std::lock_guard<std::mutex> g(**mutexPtr);
|
|
auto* buffer = BlobGetMutableTensor(
|
|
ws->GetBlob("__CAFFE2_SHARED_CONV_BUFFER_CUDA__"), CUDA);
|
|
f(buffer);
|
|
}
|
|
}
|