| commit | 85cba3b6fb4b797d26a16fefb6611c796a71cd1c | [log] [tgz] |
|---|---|---|
| author | Sanjay Patel <spatel@rotateright.com> | Wed Oct 31 14:57:23 2018 +0000 |
| committer | Sanjay Patel <spatel@rotateright.com> | Wed Oct 31 14:57:23 2018 +0000 |
| tree | fad9bfae0d72067f2eef9aba810b9364188dedab | |
| parent | 7d6d9eb6889c3f78a4466488d03744e408f39f1c [diff] [blame] |
[InstSimplify] fold 'fcmp nnan oge X, 0.0' when X is not negative This re-raises some of the open questions about how to apply and use fast-math-flags in IR from PR38086: https://bugs.llvm.org/show_bug.cgi?id=38086 ...but given the current implementation (no FMF on casts), this is likely the only way to predicate the transform. This is part of solving PR39475: https://bugs.llvm.org/show_bug.cgi?id=39475 Differential Revision: https://reviews.llvm.org/D53874 llvm-svn: 345725
diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp index b138193..efe8850 100644 --- a/llvm/lib/Analysis/InstructionSimplify.cpp +++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -3612,6 +3612,10 @@ } if (C->isZero()) { switch (Pred) { + case FCmpInst::FCMP_OGE: + if (FMF.noNaNs() && CannotBeOrderedLessThanZero(LHS, Q.TLI)) + return getTrue(RetTy); + break; case FCmpInst::FCMP_UGE: if (CannotBeOrderedLessThanZero(LHS, Q.TLI)) return getTrue(RetTy);