portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2026-06-18 01:43:43 +01:00
Code Releases Activity

Merge pull request #492 from giltirn/develop

Browse Source 
Fixes to support CUDA > 13
This commit is contained in:
Peter Boyle
2026-05-19 15:26:58 -04:00
committed by GitHub
parent f8b2eacf99 77b8657fcc
commit a5a04929fb
2 changed files with 11 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Grid/lattice/Lattice_slicesum_core.h
+9 -3
View File
@@ -1,7 +1,6 @@
#pragma once
#if defined(GRID_CUDA)
#include <cub/cub.cuh>
#define gpucub cub
#define gpuError_t cudaError_t
@@ -57,8 +56,13 @@ inline void sliceSumReduction_cub_small(const vobj *Data,
//copy offsets to device
acceleratorCopyToDeviceAsynch(&offsets[0],d_offsets,sizeof(int)*(rd+1),computeStream);
#if defined(__CUDACC__) && (__CUDACC_VER_MAJOR__ >= 13)
#define GRID_CUB_SUM_OP ::cuda::std::plus<>{}
#else
#define GRID_CUB_SUM_OP ::cub::Sum()
#endif
gpuError_t gpuErr = gpucub::DeviceSegmentedReduce::Reduce(temp_storage_array, temp_storage_bytes, rb_p,d_out, rd, d_offsets, d_offsets+1, ::gpucub::Sum(), zero_init, computeStream);
gpuError_t gpuErr = gpucub::DeviceSegmentedReduce::Reduce(temp_storage_array, temp_storage_bytes, rb_p,d_out, rd, d_offsets, d_offsets+1, GRID_CUB_SUM_OP, zero_init, computeStream);
if (gpuErr!=gpuSuccess) {
std::cout << GridLogError << "Lattice_slicesum_gpu.h: Encountered error during gpucub::DeviceSegmentedReduce::Reduce (setup)! Error: " << gpuErr <<std::endl;
exit(EXIT_FAILURE);
@@ -82,11 +86,13 @@ inline void sliceSumReduction_cub_small(const vobj *Data,
});
//issue segmented reductions in computeStream
gpuErr = gpucub::DeviceSegmentedReduce::Reduce(temp_storage_array, temp_storage_bytes, rb_p, d_out, rd, d_offsets, d_offsets+1,::gpucub::Sum(), zero_init, computeStream);
gpuErr = gpucub::DeviceSegmentedReduce::Reduce(temp_storage_array, temp_storage_bytes, rb_p, d_out, rd, d_offsets, d_offsets+1, GRID_CUB_SUM_OP, zero_init, computeStream);
if (gpuErr!=gpuSuccess) {
std::cout << GridLogError << "Lattice_slicesum_gpu.h: Encountered error during gpucub::DeviceSegmentedReduce::Reduce! Error: " << gpuErr <<std::endl;
exit(EXIT_FAILURE);
}
#undef GRID_CUB_SUM_OP
acceleratorCopyFromDeviceAsynch(d_out,&lvSum[0],rd*sizeof(vobj),computeStream);
Grid/threads/Accelerator.h
+2
View File
@@ -96,7 +96,9 @@ void acceleratorInit(void);
#ifdef GRID_CUDA
NAMESPACE_END(Grid);
#include <cuda.h>
NAMESPACE_BEGIN(Grid);
#ifdef __CUDA_ARCH__
#define GRID_SIMT