[llvm] [AMDGPU] Promote nestedGEP allocas to vectors (PR #141199)

[Harrison Hao via llvm-commits]

================
@@ -70,7 +70,7 @@ entry:
   store i32 %extra, ptr addrspace(5) %v
   store <2 x i32> %in, ptr addrspace(5) %v1
   %e = getelementptr [2 x i32], ptr addrspace(5) %v1, i32 0, i32 %idx
-  %x = load i32, ptr addrspace(5) %e
+  %x = load volatile i32, ptr addrspace(5) %e
----------------
harrisonGPU wrote:

You can see here: https://github.com/llvm/llvm-project/pull/141199#discussion_r2111295138

https://github.com/llvm/llvm-project/pull/141199