mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-03 21:21:22 +00:00
a7da3c68b9
* check in pipeline and policy for async load in mi350, need to make sure TileAccessPattern is warp_raked or block_raked solve merge conflicts * fix cmakelists * make it build * fix? buffer async fence * relax fences; it appears it only is needed between pairs of ping-pongs * remove fences * remove fences * cleanup and reformat * add steps annotations * comment all pipeline steps / remove unexplainable syncs * clang-format * add comment * cleanup kernel types for test * fix comment * fix hardcoded warp size * faithfully copy block gemm from compute v4 policy to async policy * make async test gfx950 only * fix cmake logic * set separate compile options for async * refine comment in policy * try update hotloop scheduler * cleanup comments * test more K block sizes * unhardcode Ks, sort of * add large odd test case * fix build for quant * add comment to hot loop scheduler and rename enum * reformat * reword the pipeline description * reformat * address review / add static asserts / typo fix * update changelog
83 lines
5.3 KiB
CMake
83 lines
5.3 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
|
|
)
|
|
set(EXAMPLE_GEMM_COMPILE_COMPUTE_ASYNC_OPTIONS ${EXAMPLE_GEMM_COMPILE_COMPUTE_V4_OPTIONS})
|
|
|
|
if(GPU_TARGETS MATCHES "gfx94|gfx95|gfx11|gfx12")
|
|
add_test_executable(test_ck_tile_gemm_pipeline_universal_int8 test_gemm_pipeline_universal_int8.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_universal_int8 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
add_test_executable(test_ck_tile_gemm_pipeline_universal_pk_int4 test_gemm_pipeline_universal_pk_int4.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_universal_pk_int4 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
else()
|
|
message(DEBUG "Skipping ck_tile_gemm tests for current target")
|
|
endif()
|
|
|
|
if(GPU_TARGETS MATCHES "gfx94|gfx95|gfx12")
|
|
add_test_executable(test_ck_tile_gemm_pipeline_universal_fp8 test_gemm_pipeline_universal_fp8.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_universal_fp8 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
add_test_executable(test_ck_tile_gemm_pipeline_universal_bf8 test_gemm_pipeline_universal_bf8.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_universal_bf8 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
add_test_executable(test_ck_tile_gemm_pipeline_basic_fp8 test_gemm_pipeline_basic_fp8.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_basic_fp8 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
add_test_executable(test_ck_tile_gemm_pipeline_basic_bf8 test_gemm_pipeline_basic_bf8.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_basic_bf8 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
else()
|
|
message(DEBUG "Skipping ck_tile_gemm tests for current target")
|
|
endif()
|
|
|
|
if(GPU_TARGETS MATCHES "gfx94|gfx95|gfx90a|gfx11|gfx12")
|
|
add_test_executable(test_ck_tile_gemm_pipeline_universal_fp16 test_gemm_pipeline_universal_fp16.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_universal_fp16 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
target_compile_options(test_ck_tile_gemm_pipeline_universal_fp16 PRIVATE --save-temps -Wno-gnu-line-marker)
|
|
add_test_executable(test_ck_tile_gemm_pipeline_universal_bf16 test_gemm_pipeline_universal_bf16.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_universal_bf16 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
add_test_executable(test_ck_tile_gemm_pipeline_basic_fp16 test_gemm_pipeline_basic_fp16.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_basic_fp16 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
add_test_executable(test_ck_tile_gemm_pipeline_basic_bf16 test_gemm_pipeline_basic_bf16.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_basic_bf16 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
else()
|
|
message(DEBUG "Skipping ck_tile_gemm tests for current target ")
|
|
endif()
|
|
|
|
if(GPU_TARGETS MATCHES "gfx94|gfx95|gfx90a|gfx11|gfx12")
|
|
if(GPU_TARGETS MATCHES "gfx94|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)
|
|
add_gtest_executable(test_ck_tile_gemm_pipeline_persistent test_gemm_pipeline_persistent.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})
|
|
target_compile_options(test_ck_tile_gemm_pipeline_persistent PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
endif()
|
|
|
|
if(GPU_TARGETS MATCHES "gfx95")
|
|
add_gtest_executable(test_ck_tile_gemm_pipeline_comp_async test_gemm_pipeline_comp_async.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_comp_async PRIVATE ${EXAMPLE_GEMM_COMPILE_COMPUTE_ASYNC_OPTIONS})
|
|
endif()
|
|
|
|
if(GPU_TARGETS MATCHES "gfx11|gfx12")
|
|
# On Radeon devices, build the WMMA version instead
|
|
add_gtest_executable(test_ck_tile_gemm_pipeline_mem_wmma test_gemm_pipeline_mem_wmma.cpp)
|
|
add_gtest_executable(test_ck_tile_gemm_pipeline_compv3_wmma test_gemm_pipeline_compv3_wmma.cpp)
|
|
add_gtest_executable(test_ck_tile_gemm_pipeline_compv4_wmma test_gemm_pipeline_compv4_wmma.cpp)
|
|
add_gtest_executable(test_ck_tile_gemm_pipeline_persistent_wmma test_gemm_pipeline_persistent_wmma.cpp)
|
|
target_compile_options(test_ck_tile_gemm_pipeline_mem_wmma PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
target_compile_options(test_ck_tile_gemm_pipeline_compv3_wmma PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
target_compile_options(test_ck_tile_gemm_pipeline_compv4_wmma PRIVATE ${EXAMPLE_GEMM_COMPILE_COMPUTE_V4_OPTIONS})
|
|
target_compile_options(test_ck_tile_gemm_pipeline_persistent_wmma PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
endif()
|
|
else()
|
|
message(DEBUG "Skipping ck_tile_gemm tests for current target test_ck_tile_gemm_pipeline")
|
|
endif()
|