[Mlir-commits] [mlir] [mlir][SPIR-V] Lower AND/OR/XOR vector reductions (PR #192293)

Igor Wodiany llvmlistbot at llvm.org
Fri Apr 17 03:31:34 PDT 2026


================
@@ -442,10 +442,17 @@ struct VectorReductionPattern final : OpConversionPattern<vector::ReductionOp> {
         INT_OR_FLOAT_CASE(MAXUI, SPIRVUMaxOp);
         INT_OR_FLOAT_CASE(MAXSI, SPIRVSMaxOp);
 
-      case vector::CombiningKind::AND:
-      case vector::CombiningKind::OR:
-      case vector::CombiningKind::XOR:
-        return rewriter.notifyMatchFailure(reduceOp, "unimplemented");
+#define INT_CASE(kind, iop)                                                    \
----------------
IgWod wrote:

Maybe move the whole `#define` above the first `INT_AND_FLOAT_CASE` for consistency? So all defines are on the top of the switch.

https://github.com/llvm/llvm-project/pull/192293


More information about the Mlir-commits mailing list