[PATCH] D62997: [LV] Share the LV illegality reporting with LoopVectorize. NFC.
Hideki Saito via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Fri Jun 14 12:16:54 PDT 2019
hsaito added a comment.
Sorry, I missed the original review request in the e-mail pile up. Yes, this is the direction I was suggesting. Thank you.
My preference is to have the utility declared in LoopVectorize.h as a more neutral place. If we had vectorizer utility .h/.cpp, that would have been ideal, What bothers me is Hints.vectorizeAnalysisPassName(). Anybody know why we use AlwaysPrint()?
Not a big fan of creating a new entry just for two parameters --- especially when those two parameters are probably the same for all cases. If we can always use LV_NAME, one of the two parameters can be eliminated.
If theLoop is the only thing left, passing it from all calls seems better than having two functions with the same name with same functionality (on in class and one at top level).
================
Comment at: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7339
if (F->hasFnAttribute(Attribute::NoImplicitFloat)) {
- LLVM_DEBUG(dbgs() << "LV: Can't vectorize when the NoImplicitFloat"
- "attribute is used.\n");
- ORE->emit(createLVMissedAnalysis(Hints.vectorizeAnalysisPassName(),
- "NoImplicitFloat", L)
- << "loop not vectorized due to NoImplicitFloat attribute");
+ ::reportVectorizationFailure(
+ "Can't vectorize when the NoImplicitFloat attribute is used",
----------------
Do we need :: here?
================
Comment at: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:7353
TTI->isFPVectorizationPotentiallyUnsafe()) {
- LLVM_DEBUG(
- dbgs() << "LV: Potentially unsafe FP op prevents vectorization.\n");
- ORE->emit(
- createLVMissedAnalysis(Hints.vectorizeAnalysisPassName(), "UnsafeFP", L)
- << "loop not vectorized due to unsafe FP support.");
+ ::reportVectorizationFailure(
+ "Potentially unsafe FP op prevents vectorization",
----------------
Same as above
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D62997/new/
https://reviews.llvm.org/D62997
More information about the llvm-commits
mailing list