InstCombine: fsub nsz 0, X ==> fsub nsz -0.0, X

[spatel]

Summary

Some day the backend may handle instruction-level fast math flags and make this transform unnecessary, but it's still better practice to use the canonical representation of fneg when possible (use a -0.0).

This is a partial fix for PR20870 ( http://llvm.org/bugs/show_bug.cgi?id=20870 ). See also http://reviews.llvm.org/D6723.

[hfinkel]

LGTM.

llvm/trunk/lib/Transforms/InstCombine/InstCombineAddSub.cpp

Instruction *InstCombiner::visitFSub(BinaryOperator &I) {
 
   if (Value *V = SimplifyVectorOp(I))
     return ReplaceInstUsesWith(I, V);
 
   if (Value *V = SimplifyFSubInst(Op0, Op1, I.getFastMathFlags(), DL,
                                   TLI, DT, AT))
     return ReplaceInstUsesWith(I, V);
 
+  // fsub nsz 0, X ==> fsub nsz -0.0, X
+  if (I.getFastMathFlags().noSignedZeros() && match(Op0, m_Zero())) {
+    // Subtraction from -0.0 is the canonical form of fneg.
+    Instruction *NewI = BinaryOperator::CreateFNeg(Op1);
+    NewI->copyFastMathFlags(&I);
+    return NewI;
+  }
+
   if (isa<Constant>(Op0))
     if (SelectInst *SI = dyn_cast<SelectInst>(Op1))
       if (Instruction *NV = FoldOpIntoSelect(I, SI))
         return NV;
 
   // If this is a 'B = x-(-A)', change to B = x+A, potentially looking
   // through FP extensions/truncations along the way.
   if (Value *V = dyn_castFNegVal(Op1)) {

llvm/trunk/test/Transforms/InstCombine/fast-math.ll

define float @fneg1(float %f1, float %f2) {
   %sub = fsub float -0.000000e+00, %f1
   %sub1 = fsub nsz float 0.000000e+00, %f2
   %mul = fmul float %sub, %sub1
   ret float %mul
 ; CHECK-LABEL: @fneg1(
 ; CHECK: fmul float %f1, %f2
 }
 
+define float @fneg2(float %x) {
+  %sub = fsub nsz float 0.0, %x
+  ret float %sub
+; CHECK-LABEL: @fneg2(
+; CHECK-NEXT: fsub nsz float -0.000000e+00, %x
+; CHECK-NEXT: ret float 
+}
+
 ; =========================================================================
 ;
 ;   Testing-cases about div
 ;
 ; =========================================================================
 
 ; X/C1 / C2 => X * (1/(C2*C1))
 define float @fdiv1(float %x) {