ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-06-30 03:37:38 +00:00
Code Issues Packages Projects Releases Wiki Activity

Cmakelist diable compiler-bound flags

Browse Source
This commit is contained in:
OscarXu
2025-05-17 00:51:44 -05:00
parent e4516b71c9
commit 19e43cb175
1 changed files with 4 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
example/65_gemm_multiply_multiply/CMakeLists.txt
View File

@@ -17,6 +17,7 @@ target_compile_options(example_gemm_multiply_multiply_xdl_fp8_bpreshuffle PRIVAT
add_example_executable(example_moe_gemm1_xdl_fp8 moe_gemm1_xdl_fp8.cpp)
add_example_executable(example_moe_gemm2_xdl_fp8 moe_gemm2_xdl_fp8.cpp)
add_example_executable(example_moe_gemm2_xdl_fp8_blockscale moe_gemm2_xdl_fp8_blockscale.cpp)
add_example_executable(example_moe_gemm1_xdl_fp8_blockscale moe_gemm1_xdl_fp8_blockscale.cpp)
list(APPEND gpu_list gfx942 gfx950)
set(target 0)
@@ -41,16 +42,8 @@ set(GEMM_OPTIONS)
list(APPEND GEMM_OPTIONS "SHELL: -mllvm -greedy-reverse-local-assignment=1 -mllvm --slp-threshold=-32")
list(APPEND GEMM_OPTIONS -v --save-temps -Wno-gnu-line-marker)
set(BLOCKSCALE_GEMM_OPTIONS)
list(APPEND BLOCKSCALE_GEMM_OPTIONS "SHELL: -mllvm -greedy-reverse-local-assignment=1 -mllvm --slp-threshold=-32 -mllvm --schedmodel=0 -mllvm --amdgpu-sched-strategy=gcn-iterative-max-occupancy-experimental")
list(APPEND BLOCKSCALE_GEMM_OPTIONS -v --save-temps -Wno-gnu-line-marker)
target_compile_options(example_gemm_multiply_multiply_xdl_fp8_bpreshuffle PRIVATE ${GEMM_OPTIONS})
target_compile_options(example_moe_gemm1_xdl_fp8 PRIVATE ${GEMM_OPTIONS})
target_compile_options(example_moe_gemm2_xdl_fp8 PRIVATE ${GEMM_OPTIONS})
target_compile_options(example_gemm_multiply_multiply_xdl_fp8_ab_scale PRIVATE ${BLOCKSCALE_GEMM_OPTIONS})
target_compile_options(example_gemm_multiply_multiply_xdl_fp8_blockscale_bpreshuffle PRIVATE ${BLOCKSCALE_GEMM_OPTIONS})
# list(APPEND BLOCKSCALE_GEMM_OPTIONS "SHELL: -mllvm -greedy-reverse-local-assignment=1 -mllvm --slp-threshold=-32 -mllvm --disable-schedmodel-in-sched-mi=1 -mllvm --amdgpu-sched-strategy=gcn-iterative-max-occupancy-experimental -mllvm --misched-bottomup=1")
list(APPEND BLOCKSCALE_GEMM_OPTIONS "SHELL: -mllvm -greedy-reverse-local-assignment=1 -mllvm --slp-threshold=-32 -mllvm --schedmodel=0 -mllvm --amdgpu-sched-strategy=gcn-iterative-max-occupancy-experimental")
list(APPEND BLOCKSCALE_GEMM_OPTIONS "SHELL: -mllvm -greedy-reverse-local-assignment=1 -mllvm --slp-threshold=-32")
list(APPEND BLOCKSCALE_GEMM_OPTIONS -v --save-temps -Wno-gnu-line-marker)
target_compile_options(example_gemm_multiply_multiply_xdl_fp8_bpreshuffle PRIVATE ${GEMM_OPTIONS})
target_compile_options(example_moe_gemm1_xdl_fp8 PRIVATE ${GEMM_OPTIONS})
@@ -59,6 +52,7 @@ target_compile_options(example_gemm_multiply_multiply_xdl_fp8_ab_scale PRIVATE $
target_compile_options(example_gemm_multiply_multiply_xdl_fp8_blockscale_bpreshuffle PRIVATE ${BLOCKSCALE_GEMM_OPTIONS})
target_compile_options(example_moe_gemm2_xdl_fp8_blockscale PRIVATE ${BLOCKSCALE_GEMM_OPTIONS})
target_compile_options(example_moe_gemm1_xdl_fp8_blockscale PRIVATE ${BLOCKSCALE_GEMM_OPTIONS})
#hacky fix for bs_moe_stage2 with rocm < 6.4
target_compile_definitions(example_moe_gemm2_xdl_fp8_blockscale PRIVATE MOE_STAGE2_ASM_DIR="${CMAKE_CURRENT_SOURCE_DIR}/hsa/")
target_compile_definitions(example_moe_gemm2_xdl_fp8_blockscale PRIVATE MOE_STAGE2_ASM_DIR="${CMAKE_CURRENT_SOURCE_DIR}/hsa/")