[PATCH] D153392: [PhaseOrdering] Don't speculate blocks in simplifycfg before jump-threading

[Arthur Eubanks via Phabricator via llvm-commits]

aeubanks added a comment.

I've also found that

  diff --git a/llvm/lib/Transforms/Scalar/IndVarSimplify.cpp b/llvm/lib/Transforms/Scalar/IndVarSimplify.cpp
  index 40475d9563b2..dee3a1308405 100644
  --- a/llvm/lib/Transforms/Scalar/IndVarSimplify.cpp
  +++ b/llvm/lib/Transforms/Scalar/IndVarSimplify.cpp
  @@ -2064,6 +2064,8 @@ PreservedAnalyses IndVarSimplifyPass::run(Loop &L, LoopAnalysisManager &AM,
     if (!IVS.run(&L))
       return PreservedAnalyses::all();
   
  +  AR.SE.forgetLoop(&L);
  +
     auto PA = getLoopPassPreservedAnalyses();
     PA.preserveSet<CFGAnalyses>();
     if (AR.MSSA)

fixes the issue and allows loop-unroll-full to see the proper trip count. I'm not familiar with SCEV caching, is this a stale analysis issue?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D153392/new/

https://reviews.llvm.org/D153392