mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-04-20 14:59:17 +00:00
Enable gfx941 and gfx942 architectures. (#752)
* enable gfx941/942 targets * fix clang format * fix the cmake logic for multiple targets * fix cmake syntax for looping over targets * add gfx941/942 support for gemm_xdl instances
This commit is contained in:
@@ -1,12 +1,17 @@
|
||||
if(GPU_TARGETS MATCHES "gfx908" OR GPU_TARGETS MATCHES "gfx90a" OR GPU_TARGETS MATCHES "gfx940")
|
||||
list(APPEND gpu_list gfx908 gfx90a gfx940 gfx941 gfx942)
|
||||
set(target 0)
|
||||
foreach(gpu IN LISTS GPU_TARGETS)
|
||||
if(gpu IN_LIST gpu_list AND target EQUAL 0)
|
||||
add_example_executable(example_convnd_fwd_xdl_fp32 convnd_fwd_xdl_fp32.cpp)
|
||||
add_example_executable(example_convnd_fwd_xdl_fp16 convnd_fwd_xdl_fp16.cpp)
|
||||
add_example_executable(example_convnd_fwd_xdl_bf16 convnd_fwd_xdl_bf16.cpp)
|
||||
add_example_executable(example_convnd_fwd_xdl_int8 convnd_fwd_xdl_int8.cpp)
|
||||
# FIXME: re-enable this exampe as test when SWDEV-335738 is fixed
|
||||
add_example_executable_no_testing(example_convnd_fwd_xdl_fp64 convnd_fwd_xdl_fp64.cpp)
|
||||
endif()
|
||||
set(target 1)
|
||||
endif()
|
||||
endforeach()
|
||||
|
||||
add_example_executable(example_convnd_fwd_dl_fp16 convnd_fwd_dl_fp16.cpp)
|
||||
add_example_executable(example_convnd_fwd_dl_fp32 convnd_fwd_dl_fp32.cpp)
|
||||
add_example_executable(example_convnd_fwd_dl_int8 convnd_fwd_dl_int8.cpp)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user