[PATCH] D86156: [BFI] Preserve BFI information through loop passes via VH callbacks inside LoopStandardAnalysisResults

Thu Aug 27 12:41:22 PDT 2020

nikic added a comment.

Current compile-time numbers show a 1% geomean regression: https://llvm-compile-time-tracker.com/compare.php?from=d7c119d89c5f6d0789cfd0a139c80e23912c0bb0&to=127615d90c7b4424ec83f5a8ab4256d08f7a8362&stat=instructions

I've left a comment inline with a possible cause.

================
Comment at: llvm/lib/Transforms/Scalar/LICM.cpp:220
                           &getAnalysis<DominatorTreeWrapperPass>().getDomTree(),
+                          &getAnalysis<LazyBlockFrequencyInfoPass>().getBFI(),
                           &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI(
----------------
I believe that to make this actually lazy the getBFI() call needs to be guarded by some other check for presence of profiling data, otherwise it will be computed unconditionally at this point. Typically something like F.hasProfileData() or PSI.hasProfileSummary().

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D86156/new/

https://reviews.llvm.org/D86156