Add support for fast-math flags to the FCmp instruction.
FCmp behaves a lot like a floating-point binary operator in many ways,
and can benefit from fast-math information. Flags such as nsz and nnan
can affect if this fcmp (in combination with a select) can be treated
as a fminnum/fmaxnum operation.
This adds backwards-compatible bitcode support, IR parsing and writing,
LangRef changes and IRBuilder changes. I'll need to audit InstSimplify
and InstCombine in a followup to find places where flags should be
copied.
llvm-svn: 241901
diff --git a/llvm/lib/AsmParser/LLParser.cpp b/llvm/lib/AsmParser/LLParser.cpp
index 91a88bc..aa96d60 100644
--- a/llvm/lib/AsmParser/LLParser.cpp
+++ b/llvm/lib/AsmParser/LLParser.cpp
@@ -4534,8 +4534,17 @@
case lltok::kw_and:
case lltok::kw_or:
case lltok::kw_xor: return ParseLogical(Inst, PFS, KeywordVal);
- case lltok::kw_icmp:
- case lltok::kw_fcmp: return ParseCompare(Inst, PFS, KeywordVal);
+ case lltok::kw_icmp: return ParseCompare(Inst, PFS, KeywordVal);
+ case lltok::kw_fcmp: {
+ FastMathFlags FMF = EatFastMathFlagsIfPresent();
+ int Res = ParseCompare(Inst, PFS, KeywordVal);
+ if (Res != 0)
+ return Res;
+ if (FMF.any())
+ Inst->setFastMathFlags(FMF);
+ return 0;
+ }
+
// Casts.
case lltok::kw_trunc:
case lltok::kw_zext: