[llvm] [SROA] Rewrite invariant group intrinsics after splitting alloca (PR #107557)

[Antonio Frighetto via llvm-commits]

================
@@ -3531,8 +3531,19 @@ class AllocaSliceRewriter : public InstVisitor<AllocaSliceRewriter, bool> {
       return true;
     }
 
-    if (II.isLaunderOrStripInvariantGroup())
+    if (II.isLaunderOrStripInvariantGroup()) {
+      Value *AdjustedPtr = getNewAllocaSlicePtr(IRB, OldPtr->getType());
+      Value *New = nullptr;
+      if (II.getIntrinsicID() == Intrinsic::launder_invariant_group)
+        New = IRB.CreateLaunderInvariantGroup(AdjustedPtr);
+      else if (II.getIntrinsicID() == Intrinsic::strip_invariant_group)
+        New = IRB.CreateStripInvariantGroup(AdjustedPtr);
+
+      New->takeName(&II);
+      II.replaceAllUsesWith(New);
----------------
antoniofrighetto wrote:

Isn't this already being done as part of visiting each slice in:

https://github.com/llvm/llvm-project/blob/e03f427196ec67a8a5cfbdd658f9eabe9bce83ce/llvm/lib/Transforms/Scalar/SROA.cpp#L4921-L4924

Or am I mistaking it? It looks like the intrinsic has been rewritten for each part of the alloca (this is before mem2reg for the test case):
```
define void @partial_promotion_of_alloca() {
  %struct_ptr.sroa.0 = alloca i32, align 4
  %struct_ptr.sroa.2 = alloca i32, align 4
  store i32 0, ptr %struct_ptr.sroa.0, align 4
  store volatile i32 0, ptr %struct_ptr.sroa.2, align 4
  %barr = call ptr @llvm.launder.invariant.group.p0(ptr %struct_ptr.sroa.0)
  %1 = call ptr @llvm.launder.invariant.group.p0(ptr %struct_ptr.sroa.2)
  %struct_ptr.sroa.2.0.struct_ptr.sroa.2.4.load_val = load volatile i32, ptr %struct_ptr.sroa.2, align 4
  ret void
}
```

https://github.com/llvm/llvm-project/pull/107557