ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-04 13:41:24 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK Tile] enable building examples by default (#3259)

Browse Source 
* remove EXCLUDE_FROM_ALL from ck-tile examples
-> +15 min build time w/ 64 threads for a single arch

* fix cpp17 compile error in the ck-tile examples

---------

Co-authored-by: khuagarw <khuagarw@amd.com>
Co-authored-by: Ding, Yi <yi.ding@amd.com>
This commit is contained in:
Max Podkorytov
2025-11-26 16:24:44 -08:00
committed by GitHub
parent 40d7217ac7
commit 79aae7c7f7
39 changed files with 175 additions and 174 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
example/ck_tile/15_fused_moe/CMakeLists.txt
View File

@@ -1,19 +1,19 @@
set(TILE_EXAPMLE_FUSED_MOE "tile_example_fused_moe")
# not using add_example_executable() to add this target, since we don't want this to have
# to be included in "make all/install/check"
message(DEBUG "adding ${TILE_EXAPMLE_FUSED_MOE}")
file(GLOB INSTANCE_SRCS instances/*.cpp)
add_executable(${TILE_EXAPMLE_FUSED_MOE} EXCLUDE_FROM_ALL main.cpp)
target_include_directories(${TILE_EXAPMLE_FUSED_MOE} PRIVATE ${CMAKE_CURRENT_LIST_DIR})
target_sources(${TILE_EXAPMLE_FUSED_MOE} PRIVATE ${INSTANCE_SRCS})
if(GPU_TARGETS MATCHES "gfx94|gfx95")
set(TILE_EXAMPLE_FUSED_MOE "tile_example_fused_moe")
message(DEBUG "adding ${TILE_EXAMPLE_FUSED_MOE}")
file(GLOB INSTANCE_SRCS instances/*.cpp)
add_executable(${TILE_EXAMPLE_FUSED_MOE} main.cpp)
target_include_directories(${TILE_EXAMPLE_FUSED_MOE} PRIVATE ${CMAKE_CURRENT_LIST_DIR})
target_sources(${TILE_EXAMPLE_FUSED_MOE} PRIVATE ${INSTANCE_SRCS})
set(TILE_EXAPMLE_FUSED_MOE_COMPILE_OPTIONS)
set(TILE_EXAMPLE_FUSED_MOE_COMPILE_OPTIONS)
# NOTE: we turn off undefined-func-template to let source compile without explicit declare function specializations
list(APPEND TILE_EXAPMLE_FUSED_MOE_COMPILE_OPTIONS -Wno-undefined-func-template -Wno-float-equal)
list(APPEND TILE_EXAPMLE_FUSED_MOE_COMPILE_OPTIONS -DCK_TILE_BUFFER_LOAD_AGPR=1) # TODO: enable load to a
list(APPEND TILE_EXAPMLE_FUSED_MOE_COMPILE_OPTIONS -DCK_TILE_FLOAT_TO_BFLOAT16_DEFAULT=4) # rta
# list(APPEND TILE_EXAPMLE_FUSED_MOE_COMPILE_OPTIONS -mllvm -greedy-reverse-local-assignment=1)
# list(APPEND TILE_EXAPMLE_FUSED_MOE_COMPILE_OPTIONS -v --save-temps -Wno-gnu-line-marker)
# NOTE: we turn off undefined-func-template to let source compile without explicit declare function specializations
list(APPEND TILE_EXAMPLE_FUSED_MOE_COMPILE_OPTIONS -Wno-undefined-func-template -Wno-float-equal)
list(APPEND TILE_EXAMPLE_FUSED_MOE_COMPILE_OPTIONS -DCK_TILE_BUFFER_LOAD_AGPR=1) # TODO: enable load to a
list(APPEND TILE_EXAMPLE_FUSED_MOE_COMPILE_OPTIONS -DCK_TILE_FLOAT_TO_BFLOAT16_DEFAULT=4) # rta
# list(APPEND TILE_EXAMPLE_FUSED_MOE_COMPILE_OPTIONS -mllvm -greedy-reverse-local-assignment=1)
# list(APPEND TILE_EXAMPLE_FUSED_MOE_COMPILE_OPTIONS -v --save-temps -Wno-gnu-line-marker)
target_compile_options(${TILE_EXAPMLE_FUSED_MOE} PRIVATE ${TILE_EXAPMLE_FUSED_MOE_COMPILE_OPTIONS})
target_compile_options(${TILE_EXAMPLE_FUSED_MOE} PRIVATE ${TILE_EXAMPLE_FUSED_MOE_COMPILE_OPTIONS})
endif()

16
example/ck_tile/15_fused_moe/instances/fused_moesorting_api.cpp
View File

@@ -402,22 +402,6 @@ float fused_moesorting_mp(fused_moesorting_trait t,
using ms_index_t = ck_tile::index_t;
using ms_weight_type = float;
auto maybe_clear_workspace = [=](const ck_tile::stream_config& s_) {
if(t.clear_workspace_inside_api)
{
if(is_local_token)
{
auto k = MOR_SORTING_CLEAR_WS_DISPATCH_(true, 1024, 1);
k(s_);
}
else
{
auto k = MOR_SORTING_CLEAR_WS_DISPATCH_(false, 1024, 1);
k(s_);
}
}
};
if(a.tokens < 2048)
{
if(ck_tile::impl::moe_sorting_get_smem_size_p23(a.num_experts) >