diff options
| author | Cedric Nugteren <web@cedricnugteren.nl> | 2023-01-21 20:28:32 +0100 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2023-01-21 20:28:32 +0100 |
| commit | e72f87ae5eca5e2ea8aea4f2ce49408c1faa0521 (patch) | |
| tree | 588a426b6350a5c982d89d98749ae78667fd23b4 /src/kernels/level3/copy_fast.opencl | |
| parent | 03cffa83c5f7742f8ec0c5e762bb7048e38952f3 (diff) | |
| parent | 73f49e9b3d4abc4214122e4b8c07a736e01626ee (diff) | |
Merge pull request #451 from CodeLinaro/master
CLBlast modifications to address Qualcomm Adreno performance
Diffstat (limited to 'src/kernels/level3/copy_fast.opencl')
| -rw-r--r-- | src/kernels/level3/copy_fast.opencl | 6 |
1 files changed, 5 insertions, 1 deletions
diff --git a/src/kernels/level3/copy_fast.opencl b/src/kernels/level3/copy_fast.opencl index ef8a9017..04e4b2ab 100644 --- a/src/kernels/level3/copy_fast.opencl +++ b/src/kernels/level3/copy_fast.opencl @@ -35,7 +35,11 @@ R"( // Fast copy kernel. Requires 'ld' and the number of threads in dimension 0 to be a multiple of // COPY_VW. Also requires both matrices to be of the same dimensions and without offset. -__kernel __attribute__((reqd_work_group_size(COPY_DIMX, COPY_DIMY, 1))) +#if RELAX_WORKGROUP_SIZE == 1 + __kernel +#else + __kernel __attribute__((reqd_work_group_size(COPY_DIMX, COPY_DIMY, 1))) +#endif void CopyMatrixFast(const int ld, __global const realC* restrict src, __global realC* dest, |