ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-19 22:39:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

ck: add tf32 in DTYPES to control instances build(#3317)

Browse Source
This commit is contained in:
yinglu
2025-12-08 16:24:20 +08:00
committed by GitHub
parent 86a84ae611
commit 8fec8054b2
24 changed files with 177 additions and 140 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
client_example/CMakeLists.txt
View File

@@ -27,6 +27,9 @@ if (DTYPES)
add_definitions(-DCK_ENABLE_FP32)
set(CK_ENABLE_FP32 "ON")
endif()
if (DTYPES MATCHES "tf32")
set(CK_ENABLE_TF32 "ON")
endif()
if (DTYPES MATCHES "fp64")
add_definitions(-DCK_ENABLE_FP64)
set(CK_ENABLE_FP64 "ON")
@@ -41,6 +44,7 @@ else()
set(CK_ENABLE_INT8 "ON")
set(CK_ENABLE_FP16 "ON")
set(CK_ENABLE_FP32 "ON")
set(CK_ENABLE_TF32 "ON")
set(CK_ENABLE_FP64 "ON")
set(CK_ENABLE_BF16 "ON")
if (GPU_TARGETS MATCHES "gfx94")
@@ -67,6 +71,14 @@ if (GPU_TARGETS)
add_definitions(-DCK_USE_FNUZ_FP8)
set(CK_USE_FNUZ_FP8 "ON")
endif()
if ((GPU_TARGETS MATCHES "gfx942" OR GPU_TARGETS MATCHES "gfx95") AND CK_ENABLE_TF32)
add_definitions(-DCK_ENABLE_TF32)
set(CK_ENABLE_TF32 "ON")
else()
message(STATUS "Disabling TF32 instances for this target")
remove_definitions(-DCK_ENABLE_TF32)
set(CK_ENABLE_TF32 "OFF")
endif()
else()
add_definitions(-DCK_USE_WMMA -DCK_USE_XDL)
set(CK_USE_XDL "ON")