[llvm] [AMDGPU][PromoteAlloca] Whole-function alloca promotion to vector (PR #84735)

[Pierre van Houtryve via llvm-commits]

================
@@ -225,6 +249,55 @@ FunctionPass *llvm::createAMDGPUPromoteAllocaToVector() {
   return new AMDGPUPromoteAllocaToVector();
 }
 
+static void collectAllocaUses(AllocaInst &Alloca,
+                              SmallVectorImpl<Use *> &Uses) {
+  SmallVector<Instruction *, 4> WorkList({&Alloca});
+  while (!WorkList.empty()) {
+    auto *Cur = WorkList.pop_back_val();
+    for (auto &U : Cur->uses()) {
+      Uses.push_back(&U);
+
+      if (isa<GetElementPtrInst>(U.getUser()) || isa<BitCastInst>(U.getUser()))
----------------
Pierre-vh wrote:

I still see a lot of hits for bitcast ptr in the codebase, I think it doesn't hurt to leave it in

https://github.com/llvm/llvm-project/pull/84735