diff options
| author | Cedric Nugteren <web@cedricnugteren.nl> | 2018-07-28 14:36:33 +0200 |
|---|---|---|
| committer | Cedric Nugteren <web@cedricnugteren.nl> | 2018-07-28 14:36:33 +0200 |
| commit | 0f0baa561b6c215a1052b5c70d72215e2ab38745 (patch) | |
| tree | 944ee44ac071bd8ca6da9f4a16b9b8c9bba1889f /src/kernels/common.opencl | |
| parent | 03bed8633eade7b22e72389b36e2f63ad8f3809d (diff) | |
Disabled the use of staggered indices on AMD GPUs for the new GEMMK == 1 kernels to improve performance
Diffstat (limited to 'src/kernels/common.opencl')
| -rw-r--r-- | src/kernels/common.opencl | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/src/kernels/common.opencl b/src/kernels/common.opencl index 4a476a8b..0ad38919 100644 --- a/src/kernels/common.opencl +++ b/src/kernels/common.opencl @@ -260,7 +260,7 @@ R"( // Staggered/shuffled group indices to avoid partition camping (AMD GPUs). Formula's are taken from: // http://docs.nvidia.com/cuda/samples/6_Advanced/transpose/doc/MatrixTranspose.pdf // More details: https://github.com/CNugteren/CLBlast/issues/53 -#if USE_STAGGERED_INDICES == 1 +#if USE_STAGGERED_INDICES == 1 && GEMMK == 0 INLINE_FUNC int GetGroupIDFlat() { return get_group_id(0) + get_num_groups(0) * get_group_id(1); } |