mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-03 05:01:25 +00:00
Added gemm + reduce instance library for RDNA4. This includes: - New device implementation running GEMM and reduction kernel - instances for wmma (xdl parity) - examples for wmma (xdl parity) - tests for existing xdl and wmma
43 lines
2.4 KiB
CMake
43 lines
2.4 KiB
CMake
add_custom_target(example_splitK_gemm_xdl)
|
|
add_example_executable(example_splitK_gemm_xdl_fp32 splitK_gemm_xdl_fp32.cpp)
|
|
add_example_dependencies(example_splitK_gemm_xdl example_splitK_gemm_xdl_fp32)
|
|
|
|
add_example_executable(example_splitK_gemm_xdl_fp16 splitK_gemm_xdl_fp16.cpp)
|
|
add_example_dependencies(example_splitK_gemm_xdl example_splitK_gemm_xdl_fp16)
|
|
|
|
add_example_executable(example_splitK_gemm_xdl_fp16_fp8 splitK_gemm_xdl_fp16_fp8.cpp)
|
|
add_example_dependencies(example_splitK_gemm_xdl example_splitK_gemm_xdl_fp16_fp8)
|
|
|
|
add_example_executable(example_splitK_gemm_xdl_lds_direct_load_fp16 splitK_gemm_xdl_lds_direct_load_fp16.cpp)
|
|
add_example_dependencies(example_splitK_gemm_xdl example_splitK_gemm_xdl_lds_direct_load_fp16)
|
|
|
|
add_example_executable(example_splitK_gemm_xdl_bf16 splitK_gemm_xdl_bf16.cpp)
|
|
add_example_dependencies(example_splitK_gemm_xdl example_splitK_gemm_xdl_bf16)
|
|
|
|
add_example_executable(example_splitK_gemm_xdl_int8 splitK_gemm_xdl_int8.cpp)
|
|
add_example_dependencies(example_splitK_gemm_xdl example_splitK_gemm_xdl_int8)
|
|
|
|
if(USE_BITINT_EXTENSION_INT4)
|
|
add_example_executable(example_splitK_gemm_xdl_int4 splitK_gemm_xdl_int4.cpp)
|
|
add_example_dependencies(example_splitK_gemm_xdl example_splitK_gemm_xdl_int4)
|
|
endif()
|
|
|
|
add_example_executable(example_gemm_xdl_splitk_reduce_multi_d_fp16 gemm_xdl_splitk_reduce_multi_d_fp16.cpp)
|
|
add_example_executable(example_gemm_xdl_splitk_reduce_multi_d_bf16 gemm_xdl_splitk_reduce_multi_d_bf16.cpp)
|
|
add_example_executable(example_gemm_xdl_splitk_reduce_bf16A_i8B gemm_xdl_splitk_reduce_bf16A_i8B.cpp)
|
|
|
|
add_example_executable(example_gemm_xdl_splitk_reduce_bfp16 gemm_xdl_splitk_reduce_bf16.cpp)
|
|
|
|
add_custom_target(example_splitK_gemm_wmma)
|
|
add_example_executable(example_gemm_wmma_splitk_reduce_bf16 gemm_wmma_splitk_reduce_bf16.cpp)
|
|
add_example_dependencies(example_splitK_gemm_wmma example_gemm_wmma_splitk_reduce_bf16)
|
|
|
|
add_example_executable(example_gemm_wmma_splitk_reduce_bf16A_i8B gemm_wmma_splitk_reduce_bf16A_i8B.cpp)
|
|
add_example_dependencies(example_splitK_gemm_wmma example_gemm_wmma_splitk_reduce_bf16A_i8B)
|
|
|
|
add_example_executable(example_gemm_wmma_splitk_reduce_multi_d_bf16 gemm_wmma_splitk_reduce_multi_d_bf16.cpp)
|
|
add_example_dependencies(example_splitK_gemm_wmma example_gemm_wmma_splitk_reduce_multi_d_bf16)
|
|
|
|
add_example_executable(example_gemm_wmma_splitk_reduce_multi_d_fp16 gemm_wmma_splitk_reduce_multi_d_fp16.cpp)
|
|
add_example_dependencies(example_splitK_gemm_wmma example_gemm_wmma_splitk_reduce_multi_d_fp16)
|