diff options
| author | Yong He <yonghe@outlook.com> | 2023-04-13 09:49:22 -0700 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2023-04-13 09:49:22 -0700 |
| commit | 813892cd023e216f6f6560eb47566522d3a82609 (patch) | |
| tree | 07cbf8851e0c178cbc895be73e17e6340cc22685 /tools/gfx/cuda/cuda-command-queue.cpp | |
| parent | 352a460fc866998da5f45a8c117d891c51ab5a47 (diff) | |
Set sharedMem argument to 0 when launching cuda kernel. (#2799)
Co-authored-by: Yong He <yhe@nvidia.com>
Diffstat (limited to 'tools/gfx/cuda/cuda-command-queue.cpp')
| -rw-r--r-- | tools/gfx/cuda/cuda-command-queue.cpp | 8 |
1 files changed, 1 insertions, 7 deletions
diff --git a/tools/gfx/cuda/cuda-command-queue.cpp b/tools/gfx/cuda/cuda-command-queue.cpp index 0c17a418e..4b0ab7d94 100644 --- a/tools/gfx/cuda/cuda-command-queue.cpp +++ b/tools/gfx/cuda/cuda-command-queue.cpp @@ -93,12 +93,6 @@ void CommandQueueImpl::dispatchCompute(int x, int y, int z) UInt threadGroupSize[3]; programLayout->getKernelThreadGroupSize(kernelId, threadGroupSize); - int sharedSizeInBytes; - cuFuncGetAttribute( - &sharedSizeInBytes, - CU_FUNC_ATTRIBUTE_SHARED_SIZE_BYTES, - currentPipeline->shaderProgram->cudaKernel); - // Copy global parameter data to the `SLANG_globalParams` symbol. { CUdeviceptr globalParamsSymbol = 0; @@ -144,7 +138,7 @@ void CommandQueueImpl::dispatchCompute(int x, int y, int z) int(threadGroupSize[0]), int(threadGroupSize[1]), int(threadGroupSize[2]), - sharedSizeInBytes, + 0, stream, nullptr, extraOptions); |