Index: llvm/trunk/include/llvm/Transforms/Utils/SimplifyLibCalls.h
===================================================================
--- llvm/trunk/include/llvm/Transforms/Utils/SimplifyLibCalls.h
+++ llvm/trunk/include/llvm/Transforms/Utils/SimplifyLibCalls.h
@@ -92,6 +92,7 @@
   Value *optimizeCos(CallInst *CI, IRBuilder<> &B);
   Value *optimizePow(CallInst *CI, IRBuilder<> &B);
   Value *optimizeExp2(CallInst *CI, IRBuilder<> &B);
+  Value *optimizeFabs(CallInst *CI, IRBuilder<> &B);
   Value *optimizeSinCosPi(CallInst *CI, IRBuilder<> &B);
 
   // Integer Library Call Optimizations
Index: llvm/trunk/lib/Transforms/Utils/SimplifyLibCalls.cpp
===================================================================
--- llvm/trunk/lib/Transforms/Utils/SimplifyLibCalls.cpp
+++ llvm/trunk/lib/Transforms/Utils/SimplifyLibCalls.cpp
@@ -1230,6 +1230,30 @@
   return Ret;
 }
 
+Value *LibCallSimplifier::optimizeFabs(CallInst *CI, IRBuilder<> &B) {
+  Function *Callee = CI->getCalledFunction();
+
+  Value *Ret = nullptr;
+  if (Callee->getName() == "fabs" && TLI->has(LibFunc::fabsf)) {
+    Ret = optimizeUnaryDoubleFP(CI, B, false);
+  }
+
+  FunctionType *FT = Callee->getFunctionType();
+  // Make sure this has 1 argument of FP type which matches the result type.
+  if (FT->getNumParams() != 1 || FT->getReturnType() != FT->getParamType(0) ||
+      !FT->getParamType(0)->isFloatingPointTy())
+    return Ret;
+
+  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;
+}
+
 static bool isTrigLibCall(CallInst *CI);
 static void insertSinCosCall(IRBuilder<> &B, Function *OrigCallee, Value *Arg,
                              bool UseFloat, Value *&Sin, Value *&Cos,
@@ -1893,6 +1917,8 @@
       return optimizePow(CI, Builder);
     case Intrinsic::exp2:
       return optimizeExp2(CI, Builder);
+    case Intrinsic::fabs:
+      return optimizeFabs(CI, Builder);
     default:
       return nullptr;
     }
@@ -1965,6 +1991,10 @@
     case LibFunc::exp2:
     case LibFunc::exp2f:
       return optimizeExp2(CI, Builder);
+    case LibFunc::fabsf:
+    case LibFunc::fabs:
+    case LibFunc::fabsl:
+      return optimizeFabs(CI, Builder);
     case LibFunc::ffs:
     case LibFunc::ffsl:
     case LibFunc::ffsll:
@@ -1999,7 +2029,6 @@
     case LibFunc::fputc:
       return optimizeErrorReporting(CI, Builder, 1);
     case LibFunc::ceil:
-    case LibFunc::fabs:
     case LibFunc::floor:
     case LibFunc::rint:
     case LibFunc::round:
Index: llvm/trunk/test/Transforms/InstCombine/fabs.ll
===================================================================
--- llvm/trunk/test/Transforms/InstCombine/fabs.ll
+++ llvm/trunk/test/Transforms/InstCombine/fabs.ll
@@ -0,0 +1,100 @@
+; RUN: opt < %s -instcombine -S | FileCheck %s
+
+; Make sure all library calls are eliminated when the input is known positive.
+
+declare float @fabsf(float)
+declare double @fabs(double)
+declare fp128 @fabsl(fp128)
+
+define float @square_fabs_call_f32(float %x) {
+  %mul = fmul float %x, %x
+  %fabsf = tail call float @fabsf(float %mul)
+  ret float %fabsf
+
+; CHECK-LABEL: square_fabs_call_f32(
+; CHECK-NEXT: %mul = fmul float %x, %x
+; CHECK-NEXT: ret float %mul
+}
+
+define double @square_fabs_call_f64(double %x) {
+  %mul = fmul double %x, %x
+  %fabs = tail call double @fabs(double %mul)
+  ret double %fabs
+
+; CHECK-LABEL: square_fabs_call_f64(
+; CHECK-NEXT: %mul = fmul double %x, %x
+; CHECK-NEXT: ret double %mul
+}
+
+define fp128 @square_fabs_call_f128(fp128 %x) {
+  %mul = fmul fp128 %x, %x
+  %fabsl = tail call fp128 @fabsl(fp128 %mul)
+  ret fp128 %fabsl
+
+; CHECK-LABEL: square_fabs_call_f128(
+; CHECK-NEXT: %mul = fmul fp128 %x, %x
+; CHECK-NEXT: ret fp128 %mul
+}
+
+; Make sure all intrinsic calls are eliminated when the input is known positive.
+
+declare float @llvm.fabs.f32(float)
+declare double @llvm.fabs.f64(double)
+declare fp128 @llvm.fabs.f128(fp128)
+
+define float @square_fabs_intrinsic_f32(float %x) {
+  %mul = fmul float %x, %x
+  %fabsf = tail call float @llvm.fabs.f32(float %mul)
+  ret float %fabsf
+
+; CHECK-LABEL: square_fabs_intrinsic_f32(
+; CHECK-NEXT: %mul = fmul float %x, %x
+; CHECK-NEXT: ret float %mul
+}
+
+define double @square_fabs_intrinsic_f64(double %x) {
+  %mul = fmul double %x, %x
+  %fabs = tail call double @llvm.fabs.f64(double %mul)
+  ret double %fabs
+
+; CHECK-LABEL: square_fabs_intrinsic_f64(
+; CHECK-NEXT: %mul = fmul double %x, %x
+; CHECK-NEXT: ret double %mul
+}
+
+define fp128 @square_fabs_intrinsic_f128(fp128 %x) {
+  %mul = fmul fp128 %x, %x
+  %fabsl = tail call fp128 @llvm.fabs.f128(fp128 %mul)
+  ret fp128 %fabsl
+
+; CHECK-LABEL: square_fabs_intrinsic_f128(
+; CHECK-NEXT: %mul = fmul fp128 %x, %x
+; CHECK-NEXT: ret fp128 %mul
+}
+
+; Shrinking a library call to a smaller type should not be inhibited by nor inhibit the square optimization.
+
+define float @square_fabs_shrink_call1(float %x) {
+  %ext = fpext float %x to double
+  %sq = fmul double %ext, %ext
+  %fabs = call double @fabs(double %sq)
+  %trunc = fptrunc double %fabs to float
+  ret float %trunc
+
+; CHECK-LABEL: square_fabs_shrink_call1(
+; CHECK-NEXT: %trunc = fmul float %x, %x
+; CHECK-NEXT: ret float %trunc
+}
+
+define float @square_fabs_shrink_call2(float %x) {
+  %sq = fmul float %x, %x
+  %ext = fpext float %sq to double
+  %fabs = call double @fabs(double %ext)
+  %trunc = fptrunc double %fabs to float
+  ret float %trunc
+
+; CHECK-LABEL: square_fabs_shrink_call2(
+; CHECK-NEXT: %sq = fmul float %x, %x
+; CHECK-NEXT: ret float %sq
+}
+