[PATCH] D111846: [LV] Drop integer poison-generating flags from instructions that need predication

[Nuno Lopes via Phabricator via llvm-commits]

nlopes accepted this revision.
nlopes added a comment.

looks correct to me. Left just one perf suggestion.



================
Comment at: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:1199
+    SmallVector<VPDef *, 16> Worklist;
+    SmallPtrSet<VPDef *, 16> Visited;
+    Worklist.push_back(Root);
----------------
this visited set can be hoisted outside the lambda. You should only traverse each instruction at most once, as you need to drop poison flags if the instructions contributes to any address (doesn't matter which)


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D111846/new/

https://reviews.llvm.org/D111846