[PATCH] D37163: [LICM] sink through non-trivially replicable PHI

[Eli Friedman via Phabricator via llvm-commits]

efriedma added inline comments.


================
Comment at: lib/Transforms/Scalar/LICM.cpp:803
+  Instruction *New;
+  auto It = SunkCopies.find(ExitBlock);
+  if (It != SunkCopies.end())
----------------
Do you need to do something to keep SunkCopies up to date?


================
Comment at: lib/Transforms/Scalar/LICM.cpp:858
+  while (i != e) {
+    if (CurLoop->contains(PN->getIncomingBlock(i))) {
+      SplitBlockPredecessors(PN->getParent(), PN->getIncomingBlock(i),
----------------
The indexing here is a little weird; can you first compute all the blocks to split, then split them?


https://reviews.llvm.org/D37163