[Mlir-commits] [flang] [mlir] [Flang] [OpenMP] atomic compare (PR #184761)

[Tom Eccles]

https://github.com/tblah commented:

I wonder if you could avoid some of the pattern matching by encoding the information into the operation (instead of trying to handle a nested region with somewhat arbitrary code)? For example, the lowering for atomic write just reads everything it needs straight out of accessors on the operation.

https://github.com/llvm/llvm-project/pull/184761