Skip to content

[flang][cuda] Use the provided stream in kernel launch #135267

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

Merged
merged 2 commits into from
Apr 11, 2025
Merged

[flang][cuda] Use the provided stream in kernel launch #135267

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b2752c9
[flang][cuda] Use the provided stream in kernel launch
clementval Apr 10, 2025
23bd4e7
Fix typo
clementval Apr 10, 2025
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
20 changes: 13 additions & 7 deletions flang-rt/lib/cuda/kernel.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,7 @@
//===----------------------------------------------------------------------===//

#include "flang/Runtime/CUDA/kernel.h"
#include "flang-rt/runtime/descriptor.h"
#include "flang-rt/runtime/terminator.h"
#include "flang/Runtime/CUDA/common.h"

Expand Down Expand Up @@ -74,9 +75,9 @@ void RTDEF(CUFLaunchKernel)(const void *kernel, intptr_t gridX, intptr_t gridY,
Fortran::runtime::Terminator terminator{__FILE__, __LINE__};
terminator.Crash("Too many invalid grid dimensions");
}
cudaStream_t cuStream = 0; // TODO stream managment
CUDA_REPORT_IF_ERROR(
cudaLaunchKernel(kernel, gridDim, blockDim, params, smem, cuStream));
cudaStream_t defaultStream = 0;
CUDA_REPORT_IF_ERROR(cudaLaunchKernel(kernel, gridDim, blockDim, params, smem,
stream != kNoAsyncId ? (cudaStream_t)stream : defaultStream));
}

void RTDEF(CUFLaunchClusterKernel)(const void *kernel, intptr_t clusterX,
Expand Down Expand Up @@ -140,7 +141,11 @@ void RTDEF(CUFLaunchClusterKernel)(const void *kernel, intptr_t clusterX,
terminator.Crash("Too many invalid grid dimensions");
}
config.dynamicSmemBytes = smem;
config.stream = 0; // TODO stream managment
if (stream != kNoAsyncId) {
config.stream = (cudaStream_t)stream;
} else {
config.stream = 0;
}
cudaLaunchAttribute launchAttr[1];
launchAttr[0].id = cudaLaunchAttributeClusterDimension;
launchAttr[0].val.clusterDim.x = clusterX;
Expand Down Expand Up @@ -212,9 +217,10 @@ void RTDEF(CUFLaunchCooperativeKernel)(const void *kernel, intptr_t gridX,
Fortran::runtime::Terminator terminator{__FILE__, __LINE__};
terminator.Crash("Too many invalid grid dimensions");
}
cudaStream_t cuStream = 0; // TODO stream managment
CUDA_REPORT_IF_ERROR(cudaLaunchCooperativeKernel(
kernel, gridDim, blockDim, params, smem, cuStream));
cudaStream_t defaultStream = 0;
CUDA_REPORT_IF_ERROR(
cudaLaunchCooperativeKernel(kernel, gridDim, blockDim, params, smem,
stream != kNoAsyncId ? (cudaStream_t)stream : defaultStream));
}

} // extern "C"