mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-04 05:31:24 +00:00
634634f5c0
* First checkpoint * Second checkpoint - hot loop scheduler * Third checkpoint - init main operator * Fourth checkpoint - main loop ready * Fifth checkpoint - main loop fix * Sixth checkpoint - ReadWritecompFunc * Seventh checkpoint - Tail finished * [CK_TILE] Blockwise gemm pipeline v5 complete * Working * Working fixes 2 * Rename v5 to v77 temporarily * Data type adjustment * Data type adjustment 2 * [CK_TILE] Blockwise Gemm pipeline v5 add tests * [CK_TILE] Fix calculation error * TEMP: check pipeline * Fix name to V6 * naming and documentation changes * WIP dump * Try fixing v1 * Failing tests v5 * Debugging * Changes v2 * F16 tests working great * Working BlockwiseGemmPipelineV5 as V6 * Cleanup and format * Merging changes part1 * [CK_TILE] Blockwise Gemm Pipeline Comp V5/V6 * Remove commented code * Fix gfx950 build issues * Fix file formatting * Review changes, more concat info, add bf16 bf8 tests * Fix formatting * Add bf16 and bf8 tests --------- Co-authored-by: Adam Osewski <Adam.Osewski@amd.com>
87 lines
5.5 KiB
CMake
87 lines
5.5 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)
|
|
add_test_executable(test_ck_tile_gemm_pipeline_universal_bf8 test_gemm_pipeline_universal_bf8.cpp)
|
|
add_test_executable(test_ck_tile_gemm_pipeline_basic_fp8 test_gemm_pipeline_basic_fp8.cpp)
|
|
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_fp8 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
target_compile_options(test_ck_tile_gemm_pipeline_universal_bf8 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
target_compile_options(test_ck_tile_gemm_pipeline_universal_fp8 PRIVATE ${EXAMPLE_GEMM_COMPILE_OPTIONS})
|
|
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)
|
|
add_gtest_executable(test_ck_tile_gemm_pipeline_compv6 test_gemm_pipeline_compv6.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})
|
|
target_compile_options(test_ck_tile_gemm_pipeline_compv6 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()
|