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

fix:tf32:fix build fail for all supported targets (#2942)

Browse Source 
* fix:tf32:fix build fail for all supported targets

* new fix code
This commit is contained in:
yinglu
2025-09-29 23:04:11 +08:00
committed by GitHub
parent 769c58f133
commit 0f04f020d9
2 changed files with 19 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
include/ck/tensor_operation/gpu/warp/xdlops_gemm.hpp
View File

@@ -1277,13 +1277,29 @@ struct MfmaSelector
template <>
constexpr auto GetMfma<tf32_t, 32, 32>()
{
#if defined(__gfx12__)
return MfmaInstr::wmma_unsupport_16x16_gfx12;
#elif defined(__gfx11__)
return MfmaInstr::wmma_unsupport_16x16_gfx11;
#elif defined(__gfx942__)
return MfmaInstr::mfma_f32_32x32x4xf32;
#else
return MfmaInstr::mfma_f32_32x32x2f32;
#endif
}
template <>
constexpr auto GetMfma<tf32_t, 16, 16>()
{
#if defined(__gfx12__)
return MfmaInstr::wmma_unsupport_16x16_gfx12;
#elif defined(__gfx11__)
return MfmaInstr::wmma_unsupport_16x16_gfx11;
#elif defined(__gfx942__)
return MfmaInstr::mfma_f32_16x16x8xf32;
#else
return MfmaInstr::mfma_f32_16x16x4f32;
#endif
}
template <>