[Mlir-commits] [mlir] [mlir][amdgpu] Define an amdgpu.scaling_mfma wrapper (PR #137498)

Krzysztof Drewniak llvmlistbot at llvm.org
Thu May 1 14:43:39 PDT 2025


================
@@ -830,4 +834,52 @@ def AMDGPU_GatherToLDSOp :
   let hasVerifier = 1;
 }
 
+def AMDGPU_ScaledMFMAOp :
+    AMDGPU_Op<"scaled_mfma", [AllTypesMatch<["destC", "destD"]>,
+                        Pure]>,
+    Arguments<(ins
+                   I32Attr:$m,
+                   I32Attr:$n,
+                   I32Attr:$k,
+                   ScaledMFMAInTypes:$sourceA,
+                   ScaledMFMAInTypes:$sourceB,
+                   ScaledMFMAOutTypes:$destC,
+                   AnyTypeOf<[I8, FixedVectorOfLengthAndType<[4], [I8]>]>:$scalesA,
----------------
krzysz00 wrote:

Quick note: could you make this `Float8E8M0FNU` instead of bytes? I didn't realize we had the type, but since we do, we should use it

https://github.com/llvm/llvm-project/pull/137498


More information about the Mlir-commits mailing list