[clang] [llvm] [NFC][AMDGPU] Pre-commit clang and llvm tests for dynamic allocas (PR #120063)

[Matt Arsenault via llvm-commits]

================
@@ -418,3 +418,88 @@ define void @func_dynamic_stackalloc_sgpr_align32(ptr addrspace(1) %out) {
   store i32 0, ptr addrspace(5) %alloca
   ret void
 }
+
+define amdgpu_kernel void @kernel_non_entry_block_static_alloca(ptr addrspace(1) %out, i32 %arg.cond, i32 %in) {
----------------
arsenm wrote:

This case is tested already? 

https://github.com/llvm/llvm-project/pull/120063