mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-03 13:11:25 +00:00
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:
@@ -216,6 +216,18 @@ using WarpGemmMfma_f32_16x16x64_bf8_bf8 = WarpGemmImpl<WarpGemmAtrributeMfmaIter
|
||||
using WarpGemmMfma_f32_16x16x32_bf8_bf8 = WarpGemmImpl<
|
||||
WarpGemmAtrributeMfma<WarpGemmAttributeMfmaImpl_f32_16x16x32_bf8_bf8<WGAttrCtlEnum::Default_>>>;
|
||||
|
||||
using WarpGemmMfma_f32_16x16x128_fp8_fp8 = WarpGemmImpl<WarpGemmAtrributeMfma<
|
||||
WarpGemmAttributeMfmaImpl_f32_16x16x128_fp8_fp8<WGAttrCtlEnum::Default_>>>;
|
||||
|
||||
using WarpGemmMfma_f32_16x16x128_fp8_bf8 = WarpGemmImpl<WarpGemmAtrributeMfma<
|
||||
WarpGemmAttributeMfmaImpl_f32_16x16x128_fp8_bf8<WGAttrCtlEnum::Default_>>>;
|
||||
|
||||
using WarpGemmMfma_f32_16x16x128_bf8_fp8 = WarpGemmImpl<WarpGemmAtrributeMfma<
|
||||
WarpGemmAttributeMfmaImpl_f32_16x16x128_bf8_fp8<WGAttrCtlEnum::Default_>>>;
|
||||
|
||||
using WarpGemmMfma_f32_16x16x128_bf8_bf8 = WarpGemmImpl<WarpGemmAtrributeMfma<
|
||||
WarpGemmAttributeMfmaImpl_f32_16x16x128_bf8_bf8<WGAttrCtlEnum::Default_>>>;
|
||||
|
||||
using WarpGemmMfma_f32_32x32x16_fp8_fp8_CTransposed =
|
||||
WarpGemmImpl<WarpGemmAtrributeMfmaTransposedCDistribution<
|
||||
WarpGemmAttributeMfmaImpl_f32_32x32x16_fp8_fp8<WGAttrCtlEnum::Default_>>>;
|
||||
|
||||
Reference in New Issue
Block a user