[llvm] AMDGPU: Use freeze poison instead of undef in alloca promotion (PR #131285)

[Matt Arsenault via llvm-commits]

================
@@ -170,7 +173,9 @@ entry:
 ; OPT-LABEL: define i64 @ptr_alloca_bitcast
 
 ; GCN-NOT: buffer_
-; GCN: v_mov_b32_e32 v1, 0
+; GCN: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GCN-NEXT: s_setpc_b64
+
----------------
arsenm wrote:

Yes, this now optimizes out fully and avoids the previous avoidable zero initialize 

https://github.com/llvm/llvm-project/pull/131285