mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-03 05:01:25 +00:00
c5fff071e5
* add batched_transpose implement --------- Co-authored-by: root <root@ctr-ubbsmc16.amd.com> Co-authored-by: ThruptiRajLakshmanaGowda <tlakshma@amd.com> Co-authored-by: ThomasNing <thomas.ning@amd.com>
21 lines
577 B
CMake
21 lines
577 B
CMake
include_directories(AFTER
|
|
${CMAKE_CURRENT_LIST_DIR}
|
|
)
|
|
|
|
add_subdirectory(01_fmha)
|
|
add_subdirectory(02_layernorm2d)
|
|
add_subdirectory(03_gemm)
|
|
add_subdirectory(04_img2col)
|
|
add_subdirectory(05_reduce)
|
|
add_subdirectory(06_permute)
|
|
add_subdirectory(09_topk_softmax)
|
|
add_subdirectory(10_rmsnorm2d)
|
|
add_subdirectory(11_add_rmsnorm2d_rdquant)
|
|
add_subdirectory(12_smoothquant)
|
|
add_subdirectory(13_moe_sorting)
|
|
add_subdirectory(14_moe_smoothquant)
|
|
add_subdirectory(15_fused_moe)
|
|
add_subdirectory(16_batched_gemm)
|
|
add_subdirectory(17_grouped_gemm)
|
|
add_subdirectory(35_batched_transpose)
|