diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 26b1e6d51ea..16a21a96616 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -21776,7 +21776,7 @@ static SDValue PerformTruncateCombine(SDNode *N, SelectionDAG &DAG,
 /// XFormVExtractWithShuffleIntoLoad - Check if a vector extract from a target
 /// specific shuffle of a load can be folded into a single element load.
 /// Similar handling for VECTOR_SHUFFLE is performed by DAGCombiner, but
-/// shuffles have been customed lowered so we need to handle those here.
+/// shuffles have been custom lowered so we need to handle those here.
 static SDValue XFormVExtractWithShuffleIntoLoad(SDNode *N, SelectionDAG &DAG,
                                          TargetLowering::DAGCombinerInfo &DCI) {
   if (DCI.isBeforeLegalizeOps())
@@ -21788,18 +21788,20 @@ static SDValue XFormVExtractWithShuffleIntoLoad(SDNode *N, SelectionDAG &DAG,
   if (!isa<ConstantSDNode>(EltNo))
     return SDValue();
 
-  EVT VT = InVec.getValueType();
+  EVT OriginalVT = InVec.getValueType();
 
   if (InVec.getOpcode() == ISD::BITCAST) {
     // Don't duplicate a load with other uses.
     if (!InVec.hasOneUse())
       return SDValue();
     EVT BCVT = InVec.getOperand(0).getValueType();
-    if (BCVT.getVectorNumElements() != VT.getVectorNumElements())
+    if (BCVT.getVectorNumElements() != OriginalVT.getVectorNumElements())
       return SDValue();
     InVec = InVec.getOperand(0);
   }
 
+  EVT CurrentVT = InVec.getValueType();
+
   if (!isTargetShuffle(InVec.getOpcode()))
     return SDValue();
 
@@ -21809,12 +21811,12 @@ static SDValue XFormVExtractWithShuffleIntoLoad(SDNode *N, SelectionDAG &DAG,
 
   SmallVector<int, 16> ShuffleMask;
   bool UnaryShuffle;
-  if (!getTargetShuffleMask(InVec.getNode(), VT.getSimpleVT(), ShuffleMask,
-                            UnaryShuffle))
+  if (!getTargetShuffleMask(InVec.getNode(), CurrentVT.getSimpleVT(),
+                            ShuffleMask, UnaryShuffle))
     return SDValue();
 
   // Select the input vector, guarding against out of range extract vector.
-  unsigned NumElems = VT.getVectorNumElements();
+  unsigned NumElems = CurrentVT.getVectorNumElements();
   int Elt = cast<ConstantSDNode>(EltNo)->getZExtValue();
   int Idx = (Elt > (int)NumElems) ? -1 : ShuffleMask[Elt];
   SDValue LdNode = (Idx < (int)NumElems) ? InVec.getOperand(0)
@@ -21856,11 +21858,12 @@ static SDValue XFormVExtractWithShuffleIntoLoad(SDNode *N, SelectionDAG &DAG,
   SDLoc dl(N);
 
   // Create shuffle node taking into account the case that its a unary shuffle
-  SDValue Shuffle = (UnaryShuffle) ? DAG.getUNDEF(VT) : InVec.getOperand(1);
-  Shuffle = DAG.getVectorShuffle(InVec.getValueType(), dl,
+  SDValue Shuffle = (UnaryShuffle) ? DAG.getUNDEF(CurrentVT)
+                                   : InVec.getOperand(1);
+  Shuffle = DAG.getVectorShuffle(CurrentVT, dl,
                                  InVec.getOperand(0), Shuffle,
                                  &ShuffleMask[0]);
-  Shuffle = DAG.getNode(ISD::BITCAST, dl, VT, Shuffle);
+  Shuffle = DAG.getNode(ISD::BITCAST, dl, OriginalVT, Shuffle);
   return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, N->getValueType(0), Shuffle,
                      EltNo);
 }
diff --git a/test/CodeGen/X86/extractelement-load.ll b/test/CodeGen/X86/extractelement-load.ll
index b5b320d4d4b..86475997693 100644
--- a/test/CodeGen/X86/extractelement-load.ll
+++ b/test/CodeGen/X86/extractelement-load.ll
@@ -46,3 +46,19 @@ bb:
   store double %.sroa.3.24.vec.extract, double* undef, align 8
   unreachable
 }
+
+; Case where a load is unary shuffled, then bitcast (to a type with the same
+; number of elements) before extractelement.
+; This is testing for an assertion - the extraction was assuming that the undef
+; second shuffle operand was a post-bitcast type instead of a pre-bitcast type.
+define i64 @t4(<2 x double>* %a) {
+; CHECK-LABEL: t4:
+; CHECK: mov
+; CHECK: ret
+  %b = load <2 x double>* %a, align 16
+  %c = shufflevector <2 x double> %b, <2 x double> %b, <2 x i32> <i32 1, i32 0>
+  %d = bitcast <2 x double> %c to <2 x i64>
+  %e = extractelement <2 x i64> %d, i32 1
+  ret i64 %e
+}
+