[CodeGen] Teach LLVM how to lower @llvm.{min,max}num to {MIN,MAX}NAN

The behavior of {MIN,MAX}NAN differs from that of {MIN,MAX}NUM when only
one of the inputs is NaN: -NUM will return the non-NaN argument while
-NAN would return NaN.

It is desirable to lower to @llvm.{min,max}num to -NAN if they don't
have a native instruction for -NUM.  Notably, ARMv7 NEON's vmin has the
-NAN semantics.

N.B.  Of course, it is only safe to do this if the intrinsic call is
marked nnan.

llvm-svn: 266279

1 files changed, 5 insertions, 1 deletions

diff --git a/llvm/lib/Analysis/CostModel.cpp b/llvm/lib/Analysis/CostModel.cpp
index 0383cbfbbe4..36a1db664e1 100644
--- a/llvm/lib/Analysis/CostModel.cpp
+++ b/llvm/lib/Analysis/CostModel.cpp
@@ -504,8 +504,12 @@ unsigned CostModelAnalysis::getInstructionCost(const Instruction *I) const {
       for (unsigned J = 0, JE = II->getNumArgOperands(); J != JE; ++J)
         Args.push_back(II->getArgOperand(J));
 
+      FastMathFlags FMF;
+      if (auto *FPMO = dyn_cast<FPMathOperator>(II))
+        FMF = FPMO->getFastMathFlags();
+
       return TTI->getIntrinsicInstrCost(II->getIntrinsicID(), II->getType(),
-                                        Args);
+                                        Args, FMF);
     }
     return -1;
   default: