MFMA 16x16x32fp8 (#2103)

* add mfma_16x16x32_fp8

* clang format code

* Finished the fix for gemm basic

* clang foramt

* rebuild CI

* recover gemm.hpp

* add MFMA 16*16*32bf8

---------

Co-authored-by: solin <bingzhou@amd.com>
This commit is contained in:
Thomas Ning
2025-04-21 10:21:35 -07:00
committed by GitHub
parent ce61759538
commit a738e43445
5 changed files with 188 additions and 2 deletions

View File

@@ -145,6 +145,20 @@ using WarpGemmMfma_f32_32x32x16_bf8_fp8 = WarpGemmImpl<
using WarpGemmMfma_f32_32x32x16_bf8_bf8 = WarpGemmImpl<
WarpGemmAtrributeMfma<WarpGemmAttributeMfmaImpl_f32_32x32x16_bf8_bf8<WGAttrCtlEnum::Default_>>>;
using WarpGemmMfma_f32_16x16x64_fp8_fp8 = WarpGemmImpl<WarpGemmAtrributeMfmaIterateK<
WarpGemmAttributeMfmaImpl_f32_16x16x32_fp8_fp8<WGAttrCtlEnum::Default_>,
2>>;
using WarpGemmMfma_f32_16x16x32_fp8_fp8 = WarpGemmImpl<
WarpGemmAtrributeMfma<WarpGemmAttributeMfmaImpl_f32_16x16x32_fp8_fp8<WGAttrCtlEnum::Default_>>>;
using WarpGemmMfma_f32_16x16x64_bf8_bf8 = WarpGemmImpl<WarpGemmAtrributeMfmaIterateK<
WarpGemmAttributeMfmaImpl_f32_16x16x32_bf8_bf8<WGAttrCtlEnum::Default_>,
2>>;
using WarpGemmMfma_f32_16x16x32_bf8_bf8 = WarpGemmImpl<
WarpGemmAtrributeMfma<WarpGemmAttributeMfmaImpl_f32_16x16x32_bf8_bf8<WGAttrCtlEnum::Default_>>>;
using WarpGemmMfma_f32_32x32x16_fp8_fp8_CTransposed =
WarpGemmImpl<WarpGemmAtrributeMfmaTransposedCDistribution<
WarpGemmAttributeMfmaImpl_f32_32x32x16_fp8_fp8<WGAttrCtlEnum::Default_>>>;