[Mlir-commits] [mlir] [MLIR][AMDGPU] Add support for fp8 ops on gfx12 (PR #106388)

Jakub Kuderski llvmlistbot at llvm.org
Thu Aug 29 18:36:45 PDT 2024


================
@@ -235,7 +235,9 @@ LogicalResult WMMAOp::verify() {
 
   bool isDestFloat =
       (destElemType.isF32() || destElemType.isF16() || destElemType.isBF16());
-  bool isSrcFloat = (sourceAElemType.isF16() || sourceAElemType.isBF16());
+  bool isSrcFloat =
+      (sourceAElemType.isF16() || sourceAElemType.isBF16() ||
+       sourceAElemType.isFloat8E4M3FN() || sourceAElemType.isFloat8E5M2());
----------------
kuhar wrote:

You can use `isa<Type1, Type2, Type3>(sourceAElemType)`. Also above.

https://github.com/llvm/llvm-project/pull/106388


More information about the Mlir-commits mailing list