[clang] [llvm] [AMDGPU][gfx1250] Add 128B cooperative atomics (PR #156418)

[Matt Arsenault via llvm-commits]

================
@@ -6776,6 +6776,28 @@ void Verifier::visitIntrinsicCall(Intrinsic::ID ID, CallBase &Call) {
           "invalid vector type for format", &Call, Src1, Call.getArgOperand(2));
     break;
   }
+  case Intrinsic::amdgcn_cooperative_atomic_load_32x4B:
+  case Intrinsic::amdgcn_cooperative_atomic_load_16x8B:
+  case Intrinsic::amdgcn_cooperative_atomic_load_8x16B:
+  case Intrinsic::amdgcn_cooperative_atomic_store_32x4B:
+  case Intrinsic::amdgcn_cooperative_atomic_store_16x8B:
+  case Intrinsic::amdgcn_cooperative_atomic_store_8x16B: {
+    // Check we only use this intrinsic on the FLAT or GLOBAL address spaces.
----------------
arsenm wrote:

Missing test for this 

https://github.com/llvm/llvm-project/pull/156418