[PATCH] D130579: AMDGPU: Use tablegen patterns for buffer global and flat atomic fadd

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added a comment.

In D130579#3795273 <https://reviews.llvm.org/D130579#3795273>, @Petar.Avramovic wrote:

> then do I need to add additional sets of patterns for flat_intrinsic with global addr assuming we add checks for address space in each intrinsic pattern
>  global_intrinsic + addr_space1 -> global_atomic
>  flat_intrinsic + addr_space0 -> flat_atomic
>  flat_intrinsic + addr_space1 -> global_atomic

Yes, this should be the current behavior. Arguably flat intrinsic + addrspace1 should select to flat_atomic but I don't think it matters much. We should probably be swapping the intrinsic instead in the address space optimization


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D130579/new/

https://reviews.llvm.org/D130579