[llvm] [AMDGPU] Use vector memory types for 64/128-bit cooperative atomics (PR #229501)

[Gheorghe-Teodor Bercea via llvm-commits]

================
@@ -1443,24 +1443,18 @@ static unsigned getIntrMemWidth(unsigned IntrID) {
   case Intrinsic::amdgcn_global_load_async_to_lds_b32:
   case Intrinsic::amdgcn_cluster_load_async_to_lds_b32:
   case Intrinsic::amdgcn_global_store_async_from_lds_b32:
-  case Intrinsic::amdgcn_cooperative_atomic_load_32x4B:
-  case Intrinsic::amdgcn_cooperative_atomic_store_32x4B:
----------------
doru1004 wrote:

Agreed, those cases are only dead after this change, not part of the fix

https://github.com/llvm/llvm-project/pull/229501