mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-04 13:41:24 +00:00
ffb52783d0
* Implement tile loop persistent gemm kernel * Enable timing * Add tests for persistent gemm * Fix formatting * Fix gemm_basic * Rename True/False to Persistent/NonPersistent * Use only one set of layouts for persistent tests * Fix gemm example persistent template parameter * Fix formatting
31 lines
1.4 KiB
CMake
31 lines
1.4 KiB
CMake
# Currently ck_tile_gemm is only built on gfx94/gfx95
|
|
set(EXAMPLE_GEMM_COMPILE_OPTIONS "")
|
|
if(CK_USE_OCP_FP8)
|
|
list(APPEND EXAMPLE_GEMM_COMPILE_OPTIONS -DCK_TILE_USE_OCP_FP8)
|
|
endif()
|
|
set(EXAMPLE_GEMM_COMPILE_COMPUTE_V4_OPTIONS "")
|
|
if(CK_USE_OCP_FP8)
|
|
list(APPEND EXAMPLE_GEMM_COMPILE_COMPUTE_V4_OPTIONS -DCK_TILE_USE_OCP_FP8)
|
|
endif()
|
|
list(APPEND EXAMPLE_GEMM_COMPILE_COMPUTE_V4_OPTIONS
|
|
-mllvm
|
|
-enable-noalias-to-md-conversion=0
|
|
)
|
|
|
|
if(GPU_TARGETS MATCHES "gfx94" OR GPU_TARGETS MATCHES "gfx95")
|
|
add_gtest_executable(test_ck_tile_gemm_pipeline_mem test_gemm_pipeline_mem.cpp)
|
|
add_gtest_executable(test_ck_tile_gemm_pipeline_compv3 test_gemm_pipeline_compv3.cpp)
|
|
add_gtest_executable(test_ck_tile_gemm_pipeline_compv4 test_gemm_pipeline_compv4.cpp)
|
|
|
|
target_compile_options(test_ck_tile_gemm_pipeline_mem PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
target_compile_options(test_ck_tile_gemm_pipeline_compv3 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
target_compile_options(test_ck_tile_gemm_pipeline_compv4 PRIVATE ${EXAMPLE_GEMM_COMPILE_COMPUTE_V4_OPTIONS})
|
|
else()
|
|
message("Skipping ck_tile_gemm tests for current target")
|
|
endif()
|
|
|
|
if(GPU_TARGETS MATCHES "gfx94" OR GPU_TARGETS MATCHES "gfx95" OR GPU_TARGETS MATCHES "gfx90a")
|
|
add_gtest_executable(test_ck_tile_gemm_pipeline_persistent test_gemm_pipeline_persistent.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_persistent PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
endif()
|