[Mlir-commits] [mlir] [MLIR][AMDGPU] Add OCP FP8 support for new hardware (PR #106160)

[Jakub Kuderski]

================
@@ -68,6 +72,15 @@ struct TruncfToFloat16RewritePattern final
 
 } // end namespace
 
+static LogicalResult isSupportedFp8(Type elementType, Chipset chipset) {
+  if (chipset.isGfx940())
+    return success(elementType.isFloat8E5M2FNUZ() ||
+                   elementType.isFloat8E4M3FNUZ());
+  if (chipset.hasOcpFp8())
+    return success(elementType.isFloat8E5M2() || elementType.isFloat8E4M3FN());
----------------
kuhar wrote:

use `isa<T1, T2>(elementType)`

https://github.com/llvm/llvm-project/pull/106160