float comparison to double 'zero' constant can just be a float 'zero.'

InstCombine was incorrectly considering the conversion of the constant
zero to be unsafe.

We want to transform:
define float @bar(float %x) nounwind readnone optsize ssp {
  %conv = fpext float %x to double
  %cmp = fcmp olt double %conv, 0.000000e+00
  %conv1 = zext i1 %cmp to i32
  %conv2 = sitofp i32 %conv1 to float
  ret float %conv2
}

Into:
define float @bar(float %x) nounwind readnone optsize ssp {
  %cmp = fcmp olt float %x, 0.000000e+00   ; <---- This
  %conv1 = zext i1 %cmp to i32
  %conv2 = sitofp i32 %conv1 to float
  ret float %conv2
}


rdar://10215914


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140869 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Transforms/InstCombine/InstCombineCompares.cpp

@@ -2837,10 +2837,13 @@ Instruction *InstCombiner::visitFCmpInst(FCmpInst &I) {
         APFloat F = RHSF->getValueAPF();
         F.convert(*Sem, APFloat::rmNearestTiesToEven, &Lossy);
 
-        // Avoid lossy conversions and denormals.
+        // Avoid lossy conversions and denormals. Zero is a special case
+        // that's OK to convert.
+        F.clearSign();
         if (!Lossy &&
-            F.compare(APFloat::getSmallestNormalized(*Sem)) !=
-                                                           APFloat::cmpLessThan)
+            ((F.compare(APFloat::getSmallestNormalized(*Sem)) !=
+                 APFloat::cmpLessThan) || F.isZero()))
+
           return new FCmpInst(I.getPredicate(), LHSExt->getOperand(0),
                               ConstantFP::get(RHSC->getContext(), F));
         break;