5 files changed, 62 insertions, 4 deletions
diff --git a/include/llvm/ADT/APFloat.h b/include/llvm/ADT/APFloat.h
index dfe4e0f49a..ca4138b825 100644
--- a/include/llvm/ADT/APFloat.h
+++ b/include/llvm/ADT/APFloat.h
@@ -246,6 +246,13 @@ namespace llvm {
     static APFloat getSmallestNormalized(const fltSemantics &Sem,
                                          bool Negative = false);
 
+    /// getAllOnesValue - Returns a float which is bitcasted from
+    /// an all one value int.
+    ///
+    /// \param BitWidth - Select float type
+    /// \param isIEEE   - If 128 bit number, select between PPC and IEEE
+    static APFloat getAllOnesValue(unsigned BitWidth, bool isIEEE = false);
+
     /// Profile - Used to insert APFloat objects, or objects that contain
     ///  APFloat objects, into FoldingSets.
     void Profile(FoldingSetNodeID& NID) const;
diff --git a/lib/Support/APFloat.cpp b/lib/Support/APFloat.cpp
index 6d018afa3f..e765ba0a27 100644
--- a/lib/Support/APFloat.cpp
+++ b/lib/Support/APFloat.cpp
@@ -3197,6 +3197,12 @@ APFloat::initFromAPInt(const APInt& api, bool isIEEE)
     llvm_unreachable(0);
 }
 
+APFloat
+APFloat::getAllOnesValue(unsigned BitWidth, bool isIEEE)
+{
+  return APFloat(APInt::getAllOnesValue(BitWidth), isIEEE);
+}
+
 APFloat APFloat::getLargest(const fltSemantics &Sem, bool Negative) {
   APFloat Val(Sem, fcNormal, Negative);
 
diff --git a/lib/VMCore/ConstantFold.cpp b/lib/VMCore/ConstantFold.cpp
index c8292c6d61..573efb7e57 100644
--- a/lib/VMCore/ConstantFold.cpp
+++ b/lib/VMCore/ConstantFold.cpp
@@ -43,8 +43,7 @@ using namespace llvm;
 static Constant *BitCastConstantVector(ConstantVector *CV,
                                        const VectorType *DstTy) {
 
-  if (CV->isAllOnesValue() && DstTy->getElementType()->isIntegerTy())
-      return Constant::getAllOnesValue(DstTy);
+  if (CV->isAllOnesValue()) return Constant::getAllOnesValue(DstTy);
   if (CV->isNullValue()) return Constant::getNullValue(DstTy);
 
   // If this cast changes element count then we can't handle it here:
diff --git a/lib/VMCore/Constants.cpp b/lib/VMCore/Constants.cpp
index cac37cbcc3..a2ac768d22 100644
--- a/lib/VMCore/Constants.cpp
+++ b/lib/VMCore/Constants.cpp
@@ -93,7 +93,13 @@ Constant *Constant::getAllOnesValue(const Type *Ty) {
   if (const IntegerType *ITy = dyn_cast<IntegerType>(Ty))
     return ConstantInt::get(Ty->getContext(),
                             APInt::getAllOnesValue(ITy->getBitWidth()));
-  
+
+  if (Ty->isFloatingPointTy()) {
+    APFloat FL = APFloat::getAllOnesValue(Ty->getPrimitiveSizeInBits(),
+                                          !Ty->isPPC_FP128Ty());
+    return ConstantFP::get(Ty->getContext(), FL);
+  }
+
   SmallVector<Constant*, 16> Elts;
   const VectorType *VTy = cast<VectorType>(Ty);
   Elts.resize(VTy->getNumElements(), getAllOnesValue(VTy->getElementType()));
diff --git a/test/Transforms/InstCombine/bitcast-vec-uniform.ll b/test/Transforms/InstCombine/bitcast-vec-uniform.ll
index a0d4678b0a..5975f1ec39 100644
--- a/test/Transforms/InstCombine/bitcast-vec-uniform.ll
+++ b/test/Transforms/InstCombine/bitcast-vec-uniform.ll
@@ -18,7 +18,8 @@ define <4 x i32> @b(<1 x i64> %y) {
 }
 
 ; CHECK: @foo
-; CHECK: bitcast
+; CHECK-NOT: bitcast
+; CHECK: ret
 
 ; from MultiSource/Benchmarks/Bullet
 define <2 x float> @foo() {
@@ -27,4 +28,43 @@ define <2 x float> @foo() {
 }
 
 
+; CHECK: @foo2
+; CHECK-NOT: bitcast
+; CHECK: ret
+define <2 x double> @foo2() {
+  %cast = bitcast i128 -1 to <2 x double>
+  ret <2 x double> %cast
+}
 
+; CHECK: @foo3
+; CHECK-NOT: bitcast
+; CHECK: ret
+define <1 x float> @foo3() {
+  %cast = bitcast i32 -1 to <1 x float>
+  ret <1 x float> %cast
+}
+
+; CHECK: @foo4
+; CHECK-NOT: bitcast
+; CHECK: ret
+define float @foo4() {
+  %cast = bitcast <1 x i32 ><i32 -1> to float
+  ret float %cast
+}
+
+; CHECK: @foo5
+; CHECK-NOT: bitcast
+; CHECK: ret
+define double @foo5() {
+  %cast = bitcast <2 x i32 ><i32 -1, i32 -1> to double
+  ret double %cast
+}
+
+
+; CHECK: @foo6
+; CHECK-NOT: bitcast
+; CHECK: ret
+define <2 x double> @foo6() {
+  %cast = bitcast <4 x i32><i32 -1, i32 -1, i32 -1, i32 -1> to <2 x double>
+  ret <2 x double> %cast
+}