[Mlir-commits] [mlir] [mlir][gpu] Add gpu.ballot operation to GPU dialect (PR #188647)

[Krzysztof Drewniak]

================
@@ -2619,6 +2619,17 @@ OpFoldResult gpu::SubgroupBroadcastOp::fold(FoldAdaptor /*adaptor*/) {
   return nullptr;
 }
 
+//===----------------------------------------------------------------------===//
+// GPU_BallotOp
+//===----------------------------------------------------------------------===//
+
+LogicalResult gpu::BallotOp::verify() {
+  auto intType = dyn_cast<IntegerType>(getResult().getType());
----------------
krzysz00 wrote:

I'm pretty sure this is redundant with the tablegen constraints and can be removed entirely.

https://github.com/llvm/llvm-project/pull/188647