[llvm] [GVN] Look through select/phi when determining underlying object (PR #99509)

[Nikita Popov via llvm-commits]

================
@@ -6609,6 +6609,48 @@ void llvm::getUnderlyingObjects(const Value *V,
   } while (!Worklist.empty());
 }
 
+const Value *
+llvm::getUnderlyingObjectThroughPhisAndSelects(const Value *V,
+                                               unsigned MaxLookup) {
+  const unsigned MaxVisited = 8;
+
+  SmallPtrSet<const Value *, 8> Visited;
+  SmallVector<const Value *, 8> Worklist;
+  Worklist.push_back(V);
+  const Value *Object = nullptr, *FirstObject = nullptr;
+  do {
+    const Value *P = Worklist.pop_back_val();
+    P = getUnderlyingObject(P, MaxLookup);
+
+    if (!FirstObject)
+      FirstObject = P;
+
+    if (!Visited.insert(P).second)
+      continue;
+
+    if (Visited.size() == MaxVisited)
+      return FirstObject;
+
+    if (auto *SI = dyn_cast<SelectInst>(P)) {
+      Worklist.push_back(SI->getTrueValue());
+      Worklist.push_back(SI->getFalseValue());
+      continue;
+    }
+
+    if (auto *PN = dyn_cast<PHINode>(P)) {
----------------
nikic wrote:

I believe that we don't need that isSameUnderlyingObjectInLoop() logic as long as we're reducing down to a single underlying object. What this handles is the case where you have underlying objects from different loop iterations that have the same SSA value -- this is the same basic issue as "value equal in potential cycles" in BasicAA.

However, if we're reducing to a single object, then this can't happen, because we have both a contribution from the pre-header and the loop. The loop contribution has to reduce back to the loop phi to end up with a single object. If it reduces to a loop variant value, then we'd have at least two underlying objects.

https://github.com/llvm/llvm-project/pull/99509