mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-11 17:00:18 +00:00
eb041079a3
* feat: grouped gemm tile loop support for RDNA4 * fix: removed extra parameter from grouped gemm example instance * fix: FP8 check incorrectly enabling FP8 on RDNA3
59 lines
3.1 KiB
CMake
59 lines
3.1 KiB
CMake
# Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
|
|
# SPDX-License-Identifier: MIT
|
|
|
|
add_custom_target(example_grouped_gemm_xdl)
|
|
add_example_executable(example_grouped_gemm_xdl_fp32 grouped_gemm_xdl_fp32.cpp)
|
|
add_example_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_fp32)
|
|
|
|
add_example_executable(example_grouped_gemm_xdl_fp16 grouped_gemm_xdl_fp16.cpp)
|
|
add_example_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_fp16)
|
|
|
|
add_example_executable(example_grouped_gemm_multiple_d_dl_fp16 grouped_gemm_multiple_d_dl_fp16.cpp)
|
|
add_example_dependencies(example_grouped_gemm_xdl example_grouped_gemm_multiple_d_dl_fp16)
|
|
|
|
add_example_executable(example_grouped_gemm_xdl_splitk_fp16 grouped_gemm_xdl_splitk_fp16.cpp)
|
|
add_example_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_splitk_fp16)
|
|
|
|
add_example_executable(example_grouped_gemm_xdl_fixed_nk_fp16 grouped_gemm_xdl_fixed_nk_fp16.cpp)
|
|
add_example_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_fixed_nk_fp16)
|
|
|
|
add_example_executable(example_grouped_gemm_xdl_fixed_nk_bias_fp16 grouped_gemm_xdl_fixed_nk_bias_fp16.cpp)
|
|
add_example_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_fixed_nk_bias_fp16)
|
|
|
|
add_example_executable(example_grouped_gemm_xdl_bf16 grouped_gemm_xdl_bf16.cpp)
|
|
add_example_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_bf16)
|
|
|
|
add_example_executable(example_grouped_gemm_xdl_int8 grouped_gemm_xdl_int8.cpp)
|
|
add_example_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_int8)
|
|
|
|
add_example_executable(example_grouped_gemm_xdl_fixed_nk_fp16_fp8 grouped_gemm_xdl_fixed_nk_fp16_fp8.cpp)
|
|
add_example_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_fixed_nk_fp16_fp8)
|
|
|
|
add_example_executable(example_grouped_gemm_multiple_d_xdl_fp16 grouped_gemm_multiple_d_xdl_fp16.cpp)
|
|
add_example_dependencies(example_grouped_gemm_xdl example_grouped_gemm_multiple_d_xdl_fp16)
|
|
|
|
if(USE_BITINT_EXTENSION_INT4)
|
|
add_example_executable(example_grouped_gemm_xdl_int4 grouped_gemm_xdl_int4.cpp)
|
|
add_example_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_int4)
|
|
endif()
|
|
|
|
add_custom_target(example_grouped_gemm_wmma)
|
|
add_example_executable(example_grouped_gemm_wmma_splitk_fp16 grouped_gemm_wmma_splitk_fp16.cpp)
|
|
add_example_dependencies(example_grouped_gemm_wmma example_grouped_gemm_wmma_splitk_fp16)
|
|
|
|
add_example_executable(example_grouped_gemm_wmma_splitk_bf16 grouped_gemm_wmma_splitk_bf16.cpp)
|
|
add_example_dependencies(example_grouped_gemm_wmma example_grouped_gemm_wmma_splitk_bf16)
|
|
|
|
add_example_executable(example_grouped_gemm_multiple_d_wmma_fp16 grouped_gemm_multiple_d_wmma_fp16.cpp)
|
|
add_example_dependencies(example_grouped_gemm_wmma example_grouped_gemm_multiple_d_wmma_fp16)
|
|
|
|
list(APPEND gpu_list_tf32 gfx942 gfx950)
|
|
set(target 0)
|
|
foreach(gpu IN LISTS GPU_TARGETS)
|
|
if(gpu IN_LIST gpu_list_tf32 AND target EQUAL 0)
|
|
add_example_executable(example_grouped_gemm_xdl_fp32_tf32 grouped_gemm_xdl_fp32_tf32.cpp)
|
|
add_example_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_fp32_tf32)
|
|
set(target 1)
|
|
endif()
|
|
endforeach()
|