[InstSimplify] Add unary fneg to `fsub 0.0, (fneg X) ==> X` transform

[cameron.mcinally]

Summary

Also, fix a test mistake I made when adding this new test. That was r360808.

Note that we could probably do a little better with the pattern matching here, but functionality first...

[cameron.mcinally]

Fix another test I botched in r360808.

[arsenm]

llvm/lib/Analysis/InstructionSimplify.cpp

4372

Should keep the original comment, and add another line for the fneg case since both are handled

[spatel]

llvm/test/Transforms/InstSimplify/fast-math.ll

224–226

Better to commit the test fixes ahead of this patch as an NFC step, so we don't lose those diffs in case the code change has to be reverted.

[cameron.mcinally]

llvm/lib/Analysis/InstructionSimplify.cpp

4372

That's a fair change.

There is some clunkiness here though, since an FSub(+-0.0. X) should match to an FNeg under NSZ. However, the PatternMatcher doesn't currently have a good way to match that pattern when the FMFs are inferred from a different instruction. Just FYI.

[cameron.mcinally]

Address reviews...

[spatel]

LGTM

llvm/lib/Analysis/InstructionSimplify.cpp

if (match(Op1, m_NegZeroFP()) &&
     return Op0;
 
   // fsub -0.0, (fsub -0.0, X) ==> X
   Value *X;
   if (match(Op0, m_NegZeroFP()) &&
       match(Op1, m_FSub(m_NegZeroFP(), m_Value(X))))
     return X;
 
-  // fsub 0.0, (fsub 0.0, X) ==> X if signed zeros are ignored.
+  // fsub 0.0, (fneg X) ==> X if signed zeros are ignored.
   if (FMF.noSignedZeros() && match(Op0, m_AnyZeroFP()) &&
-      match(Op1, m_FSub(m_AnyZeroFP(), m_Value(X))))
+      (match(Op1, m_FSub(m_AnyZeroFP(), m_Value(X))) ||
+       match(Op1, m_FNeg(m_Value(X)))))
     return X;
 
   // fsub nnan x, x ==> 0.0
   if (FMF.noNaNs() && Op0 == Op1)
     return Constant::getNullValue(Op0->getType());
 
   // Y - (Y - X) --> X
   // (X + Y) - Y --> X

llvm/test/Transforms/InstSimplify/fast-math.ll

;
   %t1 = fsub float 0.0, %a
   %ret = fsub nsz float 0.0, %t1
   ret float %ret
 }
 
 ; fsub nsz 0.0, (fneg X) ==> X
 define float @fneg_x(float %a) {
 ; CHECK-LABEL: @fneg_x(
-; CHECK-NEXT:    [[T1:%.*]] = fsub float 0.000000e+00, [[A:%.*]]
-; CHECK-NEXT:    [[RET:%.*]] = fneg nsz float [[T1]]
-; CHECK-NEXT:    ret float [[RET]]
+; CHECK-NEXT:    ret float [[A:%.*]]
 ;
-  %t1 = fsub float 0.0, %a
-  %ret = fneg nsz float %t1
+  %t1 = fneg float %a
+  %ret = fsub nsz float 0.0, %t1
   ret float %ret
 }
 
 define <2 x float> @fsub_0_0_x_vec_undef1(<2 x float> %a) {
 ; CHECK-LABEL: @fsub_0_0_x_vec_undef1(
 ; CHECK-NEXT:    ret <2 x float> [[A:%.*]]
 ;
   %t1 = fsub <2 x float> <float 0.0, float undef>, %a
   %ret = fsub nsz <2 x float> zeroinitializer, %t1