ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-06-05 20:55:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

add flags to avoid vectorizer problem

Browse Source
This commit is contained in:
aska-0096
2025-04-23 02:08:44 +00:00
parent 5366d3415b
commit 25bb0d2fee
2 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
example/01_gemm/CMakeLists.txt
View File

@@ -39,7 +39,7 @@ add_example_dependencies(example_gemm_xdl example_gemm_xdl_fp16_fp8_streamk_v3)
add_example_executable(example_gemm_xdl_bf16_v3 gemm_xdl_bf16_v3.cpp)
add_example_dependencies(example_gemm_xdl example_gemm_xdl_bf16_v3)
set(GEMM_OPTIONS)
# list(APPEND GEMM_OPTIONS -mllvm -greedy-reverse-local-assignment=1)
list(APPEND GEMM_OPTIONS "SHELL: -mllvm -greedy-reverse-local-assignment=1 -mllvm --slp-threshold=-16")
list(APPEND GEMM_OPTIONS -v --save-temps -Wno-gnu-line-marker)
target_compile_options(example_gemm_xdl_bf16_v3 PRIVATE ${GEMM_OPTIONS})
target_compile_options(example_gemm_xdl_fp8_v3 PRIVATE ${GEMM_OPTIONS})

4
example/01_gemm/gemm_xdl_fp8_v3.cpp
View File

@@ -28,10 +28,10 @@ using DeviceGemmV2Instance =
ADataType, BDataType, CDataType, AccDataType, CShuffleDataType,
PassThrough, PassThrough, PassThrough, GemmDefault,
256,
128, 128,
256, 256,
128, 16, 16,
16, 16,
4, 4,
8, 8,
S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>,
2, 16, 16, 0,
S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>,