[Mlir-commits] [mlir] [mlir][AMDGPU] Enable emulating vector buffer_atomic_fadd on gfx11 (PR #108312)

[Krzysztof Drewniak]

krzysz00 wrote:

@arsenm Agreed, but for reasons of ... inertia and wanting to get around to, say `load <16 x i8>, ptr addrspace(7) %M` (or even `<32 x i4>`, those pointers haven't been plumbed through MLIR yet 

(And I *think* this pass came into existence while fat pointers were still under review)

https://github.com/llvm/llvm-project/pull/108312