[llvm-branch-commits] [llvm] [LowerMemIntrinsics][AMDGPU] Optimize memset.pattern lowering (PR #185901)

[Krzysztof Drewniak via llvm-branch-commits]

krzysz00 wrote:

@ritter-x2a What you're seeing there is the fact that the "resource" part of the AS7 needs to be wave-uniform.

The easiest thing to do here is to put an `inreg` on the argument, and maybe add a non-`inreg` i32 offset if you want to test the full generality.

https://github.com/llvm/llvm-project/pull/185901