[Mlir-commits] [mlir] [mlir][gpu] Preserve explicit async objects in gpu.launch_func lowering (PR #195436)

[Fabian Mora]

fabianmcg wrote:

Welcome, thank you for submitting a PR to MLIR!

This seems like a missing verification condition, as one is not supposed to have both async objects and async tokens in the same op.
You can transform this PR into the above, just amend the LanchFuncOp verifier with the check that only one is available.

https://github.com/llvm/llvm-project/pull/195436