[InstSimplify] fix folds for (0.0 - X) + X --> 0 (PR27151)

As shown in:
https://bugs.llvm.org/show_bug.cgi?id=27151
...the existing fold could miscompile when X is NaN.

The fold was also dependent on 'ninf' but that's not necessary.

From IEEE-754 (with default rounding which we can assume for these opcodes):
"When the sum of two operands with opposite signs (or the difference of two 
operands with like signs) is exactly zero, the sign of that sum (or difference) 
shall be +0...However, x + x = x − (−x) retains the same sign as x even when 
x is zero."

llvm-svn: 327575

2 files changed, 22 insertions, 32 deletions

diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index a95a775ead6..2b8e23b9569 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -4168,20 +4168,16 @@ static Value *SimplifyFAddInst(Value *Op0, Value *Op1, FastMathFlags FMF,
       (FMF.noSignedZeros() || CannotBeNegativeZero(Op0, Q.TLI)))
     return Op0;
 
-  // fadd [nnan ninf] X, (fsub [nnan ninf] 0, X) ==> 0
-  //   where nnan and ninf have to occur at least once somewhere in this
-  //   expression
-  Value *SubOp = nullptr;
-  if (match(Op1, m_FSub(m_AnyZero(), m_Specific(Op0))))
-    SubOp = Op1;
-  else if (match(Op0, m_FSub(m_AnyZero(), m_Specific(Op1))))
-    SubOp = Op0;
-  if (SubOp) {
-    Instruction *FSub = cast<Instruction>(SubOp);
-    if ((FMF.noNaNs() || FSub->hasNoNaNs()) &&
-        (FMF.noInfs() || FSub->hasNoInfs()))
-      return Constant::getNullValue(Op0->getType());
-  }
+  // With nnan: (+/-0.0 - X) + X --> 0.0 (and commuted variant)
+  // We don't have to explicitly exclude infinities (ninf): INF + -INF == NaN.
+  // Negative zeros are allowed because we always end up with positive zero:
+  // X = -0.0: (-0.0 - (-0.0)) + (-0.0) == ( 0.0) + (-0.0) == 0.0
+  // X = -0.0: ( 0.0 - (-0.0)) + (-0.0) == ( 0.0) + (-0.0) == 0.0
+  // X =  0.0: (-0.0 - ( 0.0)) + ( 0.0) == (-0.0) + ( 0.0) == 0.0
+  // X =  0.0: ( 0.0 - ( 0.0)) + ( 0.0) == ( 0.0) + ( 0.0) == 0.0
+  if (FMF.noNaNs() && (match(Op0, m_FSub(m_AnyZero(), m_Specific(Op1))) ||
+                       match(Op1, m_FSub(m_AnyZero(), m_Specific(Op0)))))
+    return ConstantFP::getNullValue(Op0->getType());
 
   return nullptr;
 }
diff --git a/llvm/test/Transforms/InstSimplify/fast-math.ll b/llvm/test/Transforms/InstSimplify/fast-math.ll
index bf862933a9c..2a3086481f2 100644
--- a/llvm/test/Transforms/InstSimplify/fast-math.ll
+++ b/llvm/test/Transforms/InstSimplify/fast-math.ll
@@ -46,50 +46,47 @@ define float @no_mul_zero_3(float %a) {
   ret float %b
 }
 
-; FIXME: -X + X --> 0.0 (with nnan on the fadd)
+; -X + X --> 0.0 (with nnan on the fadd)
 
 define float @fadd_fnegx(float %x) {
 ; CHECK-LABEL: @fadd_fnegx(
-; CHECK-NEXT:    [[NEGX:%.*]] = fsub float -0.000000e+00, [[X:%.*]]
-; CHECK-NEXT:    [[R:%.*]] = fadd nnan float [[NEGX]], [[X]]
-; CHECK-NEXT:    ret float [[R]]
+; CHECK-NEXT:    ret float 0.000000e+00
 ;
   %negx = fsub float -0.0, %x
   %r = fadd nnan float %negx, %x
   ret float %r
 }
 
-; FIXME: X + -X --> 0.0 (with nnan on the fadd)
+; X + -X --> 0.0 (with nnan on the fadd)
 
 define <2 x float> @fadd_fnegx_commute_vec(<2 x float> %x) {
 ; CHECK-LABEL: @fadd_fnegx_commute_vec(
-; CHECK-NEXT:    [[NEGX:%.*]] = fsub <2 x float> <float -0.000000e+00, float -0.000000e+00>, [[X:%.*]]
-; CHECK-NEXT:    [[R:%.*]] = fadd nnan <2 x float> [[X]], [[NEGX]]
-; CHECK-NEXT:    ret <2 x float> [[R]]
+; CHECK-NEXT:    ret <2 x float> zeroinitializer
 ;
   %negx = fsub <2 x float> <float -0.0, float -0.0>, %x
   %r = fadd nnan <2 x float> %x, %negx
   ret <2 x float> %r
 }
 
-; FIXME: Could be NaN.
 ; https://bugs.llvm.org/show_bug.cgi?id=26958
 ; https://bugs.llvm.org/show_bug.cgi?id=27151
 
 define float @fadd_fneg_nan(float %x) {
 ; CHECK-LABEL: @fadd_fneg_nan(
-; CHECK-NEXT:    ret float 0.000000e+00
+; CHECK-NEXT:    [[T:%.*]] = fsub nnan float -0.000000e+00, [[X:%.*]]
+; CHECK-NEXT:    [[COULD_BE_NAN:%.*]] = fadd ninf float [[T]], [[X]]
+; CHECK-NEXT:    ret float [[COULD_BE_NAN]]
 ;
   %t = fsub nnan float -0.0, %x
   %could_be_nan = fadd ninf float %t, %x
   ret float %could_be_nan
 }
 
-; FIXME: Could be NaN.
-
 define float @fadd_fneg_nan_commute(float %x) {
 ; CHECK-LABEL: @fadd_fneg_nan_commute(
-; CHECK-NEXT:    ret float 0.000000e+00
+; CHECK-NEXT:    [[T:%.*]] = fsub nnan ninf float -0.000000e+00, [[X:%.*]]
+; CHECK-NEXT:    [[COULD_BE_NAN:%.*]] = fadd float [[X]], [[T]]
+; CHECK-NEXT:    ret float [[COULD_BE_NAN]]
 ;
   %t = fsub nnan ninf float -0.0, %x
   %could_be_nan = fadd float %x, %t
@@ -118,15 +115,12 @@ define <2 x float> @fadd_fsub_nnan_ninf_commute_vec(<2 x float> %x) {
   ret <2 x float> %zero
 }
 
-; FIXME: Do fold this.
 ; 'ninf' is not required because 'nnan' allows us to assume
-; that X is not INF/-INF (adding opposite INFs would be NaN).
+; that X is not INF or -INF (adding opposite INFs would be NaN).
 
 define float @fadd_fsub_nnan(float %x) {
 ; CHECK-LABEL: @fadd_fsub_nnan(
-; CHECK-NEXT:    [[SUB:%.*]] = fsub float 0.000000e+00, [[X:%.*]]
-; CHECK-NEXT:    [[ZERO:%.*]] = fadd nnan float [[SUB]], [[X]]
-; CHECK-NEXT:    ret float [[ZERO]]
+; CHECK-NEXT:    ret float 0.000000e+00
 ;
   %sub = fsub float 0.0, %x
   %zero = fadd nnan float %sub, %x