Support for MFMA_16x16x128 for fp8/bf8 (#2125)

* Adding 16x16x128 support for gfx950

* Support for fp8 and bf8

* fix input arguments for MFMA scale instruction

* clang-formatted

* Fixes for lwpck-3145 (#2138)

* Fix lds tile & cmake dep & default epilogue

* Fallback BTypeToUse to ADataType in WOQ cases

* reverting instance json file

* reverting instance json file

---------

Co-authored-by: Yi DING <yi.ding@amd.com>
This commit is contained in:
Khushbu Agarwal
2025-04-28 18:19:50 -07:00
committed by GitHub
parent 768c99eca9
commit d107f3c3a5
8 changed files with 143 additions and 10 deletions

View File

@@ -298,7 +298,7 @@ struct BlockUniversalGemmAsBsCr
using BLdsTile = decltype(make_static_distributed_tensor<ComputeDataType>(BLdsTileDistr));
ALdsTile a_warp_tile_;
ALdsTile b_warp_tile_;
BLdsTile b_warp_tile_;
template <typename ASmemBlockWindow, typename BSmemBlockWindow>
CK_TILE_DEVICE void LocalPrefetch(const ASmemBlockWindow& a_block_window,