Index: lib/Transforms/Utils/SimplifyLibCalls.cpp
===================================================================
--- lib/Transforms/Utils/SimplifyLibCalls.cpp
+++ lib/Transforms/Utils/SimplifyLibCalls.cpp
@@ -1189,19 +1189,11 @@
 
 Value *LibCallSimplifier::optimizeFabs(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
-  Value *Ret = nullptr;
   StringRef Name = Callee->getName();
   if (Name == "fabs" && hasFloatVersion(Name))
-    Ret = optimizeUnaryDoubleFP(CI, B, false);
+    return optimizeUnaryDoubleFP(CI, B, false);
 
-  Value *Op = CI->getArgOperand(0);
-  if (Instruction *I = dyn_cast<Instruction>(Op)) {
-    // Fold fabs(x * x) -> x * x; any squared FP value must already be positive.
-    if (I->getOpcode() == Instruction::FMul)
-      if (I->getOperand(0) == I->getOperand(1))
-        return Op;
-  }
-  return Ret;
+  return nullptr;
 }
 
 Value *LibCallSimplifier::optimizeFMinFMax(CallInst *CI, IRBuilder<> &B) {
Index: test/Transforms/InstCombine/fabs.ll
===================================================================
--- test/Transforms/InstCombine/fabs.ll
+++ test/Transforms/InstCombine/fabs.ll
@@ -13,7 +13,8 @@
 
 ; CHECK-LABEL: square_fabs_call_f32(
 ; CHECK-NEXT: %mul = fmul float %x, %x
-; CHECK-NEXT: ret float %mul
+; CHECK-NEXT: %fabsf = tail call float @fabsf(float %mul)
+; CHECK-NEXT: ret float %fabsf
 }
 
 define double @square_fabs_call_f64(double %x) {
@@ -23,7 +24,8 @@
 
 ; CHECK-LABEL: square_fabs_call_f64(
 ; CHECK-NEXT: %mul = fmul double %x, %x
-; CHECK-NEXT: ret double %mul
+; CHECK-NEXT: %fabs = tail call double @fabs(double %mul)
+; CHECK-NEXT: ret double %fabs
 }
 
 define fp128 @square_fabs_call_f128(fp128 %x) {
@@ -33,7 +35,8 @@
 
 ; CHECK-LABEL: square_fabs_call_f128(
 ; CHECK-NEXT: %mul = fmul fp128 %x, %x
-; CHECK-NEXT: ret fp128 %mul
+; CHECK-NEXT: %fabsl = tail call fp128 @fabsl(fp128 %mul)
+; CHECK-NEXT: ret fp128 %fabsl
 }
 
 ; Make sure all intrinsic calls are eliminated when the input is known positive.
@@ -49,7 +52,8 @@
 
 ; CHECK-LABEL: square_fabs_intrinsic_f32(
 ; CHECK-NEXT: %mul = fmul float %x, %x
-; CHECK-NEXT: ret float %mul
+; CHECK-NEXT: %fabsf = tail call float @llvm.fabs.f32(float %mul)
+; CHECK-NEXT: ret float %fabsf
 }
 
 define double @square_fabs_intrinsic_f64(double %x) {
@@ -59,7 +63,8 @@
 
 ; CHECK-LABEL: square_fabs_intrinsic_f64(
 ; CHECK-NEXT: %mul = fmul double %x, %x
-; CHECK-NEXT: ret double %mul
+; CHECK-NEXT: %fabs = tail call double @llvm.fabs.f64(double %mul)
+; CHECK-NEXT: ret double %fabs
 }
 
 define fp128 @square_fabs_intrinsic_f128(fp128 %x) {
@@ -69,7 +74,8 @@
 
 ; CHECK-LABEL: square_fabs_intrinsic_f128(
 ; CHECK-NEXT: %mul = fmul fp128 %x, %x
-; CHECK-NEXT: ret fp128 %mul
+; CHECK-NEXT: %fabsl = tail call fp128 @llvm.fabs.f128(fp128 %mul)
+; CHECK-NEXT: ret fp128 %fabsl
 }
 
 ; Shrinking a library call to a smaller type should not be inhibited by nor inhibit the square optimization.
@@ -82,7 +88,10 @@
   ret float %trunc
 
 ; CHECK-LABEL: square_fabs_shrink_call1(
-; CHECK-NEXT: %trunc = fmul float %x, %x
+; CHECK-NEXT: %ext = fpext float %x to double
+; CHECK-NEXT: %sq = fmul double %ext, %ext
+; CHECK-NEXT: call double @fabs(double %sq)
+; CHECK-NEXT: %trunc = fptrunc double %fabs to float
 ; CHECK-NEXT: ret float %trunc
 }
 
@@ -95,7 +104,8 @@
 
 ; CHECK-LABEL: square_fabs_shrink_call2(
 ; CHECK-NEXT: %sq = fmul float %x, %x
-; CHECK-NEXT: ret float %sq
+; CHECK-NEXT: %fabsf = call float @fabsf(float %sq)
+; CHECK-NEXT: ret float %fabsf
 }
 
 ; CHECK-LABEL: @fabs_select_constant_negative_positive(
Index: test/Transforms/InstCombine/fast-math.ll
===================================================================
--- test/Transforms/InstCombine/fast-math.ll
+++ test/Transforms/InstCombine/fast-math.ll
@@ -672,7 +672,8 @@
 
 ; CHECK-LABEL: sqrt_intrinsic_arg_4th(
 ; CHECK-NEXT: %mul = fmul fast double %x, %x
-; CHECK-NEXT: ret double %mul
+; CHECK-NEXT: %fabs = call fast double @llvm.fabs.f64(double %mul)
+; CHECK-NEXT: ret double %fabs
 }
 
 define double @sqrt_intrinsic_arg_5th(double %x) {
@@ -684,8 +685,9 @@
 
 ; CHECK-LABEL: sqrt_intrinsic_arg_5th(
 ; CHECK-NEXT: %mul = fmul fast double %x, %x
+; CHECK-NEXT: %fabs = call fast double @llvm.fabs.f64(double %mul)
 ; CHECK-NEXT: %sqrt1 = call fast double @llvm.sqrt.f64(double %x)
-; CHECK-NEXT: %1 = fmul fast double %mul, %sqrt1
+; CHECK-NEXT: %1 = fmul fast double %fabs, %sqrt1
 ; CHECK-NEXT: ret double %1
 }