ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-20 06:49:15 +00:00
Code Issues Packages Projects Releases Wiki Activity

Simulate TF32 with BF16x3 (#3142)

Browse Source 
* tf32:bf16x3:use bf16x3 emulate tf32 gemm

* change blockwiseGemm to demo bf16x3

* temp push

* self review

* self review

* fix multi-device compile error

* bug fix

* code refactor

* limit to gfx950

* enhance gemm gfx942 threshold

* lower change from blockwise to warpwise

* refact codes

* refact codes

* error fix

* change threshold

* bug fix

* fix threshold error

* change host reference implement to same as device

* bug fix

* bug fix

* code refact

* fix clang-format fail

* code refine
This commit is contained in:
yinglu
2025-11-14 08:21:09 +08:00
committed by GitHub
parent f2cfc6b94e
commit 2a73eb3bc0
16 changed files with 419 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
example/09_convnd_fwd/CMakeLists.txt
View File

@@ -21,7 +21,7 @@ foreach(gpu IN LISTS GPU_TARGETS)
endif()
endforeach()
list(APPEND gpu_list_tf32 gfx942)
list(APPEND gpu_list_tf32 gfx942 gfx950)
set(target 0)
foreach(gpu IN LISTS GPU_TARGETS)
if(gpu IN_LIST gpu_list_tf32 AND target EQUAL 0)

2
example/09_convnd_fwd/convnd_fwd_common.hpp
View File

@@ -77,7 +77,7 @@ inline __host__ __device__ constexpr double get_atol()
{
if constexpr(std::is_same_v<DataType, float> && std::is_same_v<GemmType, ck::tf32_t>)
{
return 1e-2;
return 1e-3;
}
else if constexpr(std::is_same_v<DataType, float>)
{