[Mlir-commits] [mlir] [mlir][gpu] Pattern to promote `gpu.shuffle` to specialized AMDGPU ops (PR #137109)

[Krzysztof Drewniak]

krzysz00 wrote:

I think the pattern is fine, though part of me wants to just add this as a case to the shuffle lowering in LowerGPUOpsToROCDL

https://github.com/llvm/llvm-project/pull/137109