[llvm] RFC: [AMDGPU] Stop using attribute groups in CodeGen tests (PR #86157)

[Matt Arsenault via llvm-commits]

https://github.com/arsenm requested changes to this pull request.

You could argue we could just remove all the attributes used by intrinsics, but I think this is mostly a bad idea for any of the cases actually testing attribute handling. It deviates from the canonically printed IR. I also semi-regularly encounter test updating situations where I want to add another attribute to nearly every function, which is easy to update by modifying the one group 

https://github.com/llvm/llvm-project/pull/86157