Make the sqrt intrinsic return undef for a negative input.

As discussed here:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20140609/220598.html

And again here:
http://lists.cs.uiuc.edu/pipermail/llvmdev/2014-September/077168.html

The sqrt of a negative number when using the llvm intrinsic is undefined.
We should return undef rather than 0.0 to match the definition in the LLVM IR lang ref.

This change should not affect any code that isn't using "no-nans-fp-math";
ie, no-nans is a requirement for generating the llvm intrinsic in place of a sqrt function call.

Unfortunately, the behavior introduced by this patch will not match current gcc, xlc, icc, and
possibly other compilers. The current clang/llvm behavior of returning 0.0 doesn't either.
We knowingly approve of this difference with the other compilers in an attempt to flag code
that is invoking undefined behavior.

A front-end warning should also try to convince the user that the program will fail:
http://llvm.org/bugs/show_bug.cgi?id=21093

Differential Revision: http://reviews.llvm.org/D5527

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218803 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Analysis/ConstantFolding.cpp b/lib/Analysis/ConstantFolding.cpp
index 88e255832f1d2329fba7179cca2d18c75cea9e81..c703491ace22267058010ec067318e0feac9ff9c 100644 (file)
--- a/lib/Analysis/ConstantFolding.cpp
+++ b/lib/Analysis/ConstantFolding.cpp
@@ -1520,8 +1520,14 @@ static Constant *ConstantFoldScalarCall(StringRef Name, unsigned IntrinsicID,
                  (Ty->isHalfTy() || Ty->isFloatTy() || Ty->isDoubleTy())) {
           if (V >= -0.0)
             return ConstantFoldFP(sqrt, V, Ty);
-          else // Undefined
-            return Constant::getNullValue(Ty);
+          else {
+            // Unlike the sqrt definitions in C/C++, POSIX, and IEEE-754 - which
+            // all guarantee or favor returning NaN - the square root of a
+            // negative number is not defined for the LLVM sqrt intrinsic.
+            // This is because the intrinsic should only be emitted in place of
+            // libm's sqrt function when using "no-nans-fp-math".
+            return UndefValue::get(Ty);
+          }
         }
         break;
       case 's':

diff --git a/test/Transforms/InstCombine/constant-fold-math.ll b/test/Transforms/InstCombine/constant-fold-math.ll
index 14377df37299b9d536766a9095fdcd18744e93ec..ce8d337c08bfc97eeef001b5466c88d8125798d2 100644 (file)
--- a/test/Transforms/InstCombine/constant-fold-math.ll
+++ b/test/Transforms/InstCombine/constant-fold-math.ll
@@ -7,6 +7,7 @@ declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>) #0
 declare double @llvm.fma.f64(double, double, double) #0
 declare double @llvm.fmuladd.f64(double, double, double) #0
 
+declare double @llvm.sqrt.f64(double) #0
 
 
 ; CHECK-LABEL: @constant_fold_fma_f32
@@ -44,4 +45,12 @@ define double @constant_fold_fmuladd_f64() #0 {
   ret double %x
 }
 
+; The sqrt intrinsic is undefined for negative inputs besides -0.0.
+; CHECK-LABEL: @bad_sqrt
+; CHECK-NEXT: ret double undef
+define double @bad_sqrt() {
+  %x = call double @llvm.sqrt.f64(double -2.000000e+00)
+  ret double %x
+}
+
 attributes #0 = { nounwind readnone }