[llvm] [AMDGPU][PromoteAlloca] Don't stop when an alloca is too big to promote (PR #93466)

[Pierre van Houtryve via llvm-commits]

================
@@ -0,0 +1,41 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -mcpu=kaveri -amdgpu-promote-alloca-to-vector-limit=128 -passes=amdgpu-promote-alloca-to-vector %s -o - | FileCheck %s
+
+; Check that when we see an alloca that's too big to vectorize given the remaining budget,
+; we don't give up and we keep looking for other allocas to vectorize.
+
+define amdgpu_kernel void @simple_users_scores() {
----------------
Pierre-vh wrote:

Before this patch, `%simpleuser` wouldn't be promoted because `%manyusers` is seen first, but it's too big and we'd simply stop.

https://github.com/llvm/llvm-project/pull/93466