[PATCH] D68128: [InstCombine] Fold PHIs with equal incoming pointers

[Roman Lebedev via Phabricator via llvm-commits]

lebedev.ri marked an inline comment as done.
lebedev.ri added inline comments.


================
Comment at: llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp:1157-1168
+  Instruction *InsertPt = nullptr;
+  if (auto *BaseInst = dyn_cast<Instruction>(Base)) {
+    if (isa<PHINode>(BaseInst)) {
+      BasicBlock *InsertBB = BaseInst->getParent();
+      BasicBlock::iterator InsertPtIter = InsertBB->getFirstInsertionPt();
+      // Make sure the insertion point exists.
+      if (InsertPtIter != InsertBB->end())
----------------
DaniilSuchkov wrote:
> lebedev.ri wrote:
> > I'm not sure if this is correct.
> > I'd expect that we actually need to check to where we can move these geps.
> > 
> Why do you think it's incorrect? Given the offset is constant, this code just looks for an insertion point right after base pointer definition and bails out in case when such point doesn't exist.
> 
> What restrictions did I miss?
Actually, i guess this might be ok,
we don't do anything with the users,
and `gep inbounds` at worst produces `poison`.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D68128/new/

https://reviews.llvm.org/D68128