[llvm] AMDGPU: Improve vector of pointer handling in amdgpu-promote-alloca (PR #114144)

[Pierre van Houtryve via llvm-commits]

================
@@ -1115,9 +1115,10 @@ bool AMDGPUPromoteAllocaImpl::binaryOpIsDerivedFromSameAlloca(
   if (Val == OtherOp)
     OtherOp = Inst->getOperand(OpIdx1);
 
-  if (isa<ConstantPointerNull>(OtherOp))
+  if (isa<ConstantPointerNull>(OtherOp) || isa<ConstantAggregateZero>(OtherOp))
----------------
Pierre-vh wrote:

```suggestion
  if (isa<ConstantPointerNull, ConstantAggregateZero>(OtherOp))
```


https://github.com/llvm/llvm-project/pull/114144