ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-14 10:09:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

disable tensor contraction f64 on MI100 (#602)

Browse Source 
[ROCm/composable_kernel commit: 209baee299]
This commit is contained in:
zjing14
2023-02-23 18:59:37 -06:00
committed by GitHub
parent 76c23de0a8
commit 84a4731c15
2 changed files with 7 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
include/ck/tensor_operation/gpu/device/impl/device_contraction_multiple_d_xdl_cshuffle.hpp
View File

@@ -586,6 +586,11 @@ struct DeviceContractionMultipleD_Xdl_CShuffle
return false;
}
if(ck::get_device_name() != "gfx90a" && std::is_same<ADataType, double>::value)
{
return false;
}
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,

4
script/cmake-ck-dev.sh
View File

@@ -10,8 +10,8 @@ cmake
-D CMAKE_CXX_COMPILER=/opt/rocm/bin/hipcc \
-D CMAKE_CXX_FLAGS="-O3 -ftemplate-backtrace-limit=0 -gline-tables-only -save-temps=$PWD" \
-D CMAKE_BUILD_TYPE=Release \
-D BUILD_DEV=OFF \
-D GPU_TARGETS="gfx90a" \
-D BUILD_DEV=ON \
-D GPU_TARGETS="gfx908;gfx90a" \
-D CMAKE_VERBOSE_MAKEFILE:BOOL=ON \
-D USE_BITINT_EXTENSION_INT4=OFF \
${MY_PROJECT_SOURCE}