diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 0ea46af..d575e8a 100644
--- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -6413,6 +6413,13 @@
           break;
       }
 
+      // If the vector element type is not legal, the BUILD_VECTOR operands
+      // are promoted and implicitly truncated.  Make that explicit here.
+      if (LHSOp.getValueType() != EltType)
+        LHSOp = DAG.getNode(ISD::TRUNCATE, LHS.getDebugLoc(), EltType, LHSOp);
+      if (RHSOp.getValueType() != EltType)
+        RHSOp = DAG.getNode(ISD::TRUNCATE, RHS.getDebugLoc(), EltType, RHSOp);
+
       SDValue FoldOp = DAG.getNode(N->getOpcode(), LHS.getDebugLoc(), EltType,
                                    LHSOp, RHSOp);
       if (FoldOp.getOpcode() != ISD::UNDEF &&
diff --git a/test/CodeGen/ARM/2009-11-02-NegativeLane.ll b/test/CodeGen/ARM/2009-11-02-NegativeLane.ll
index f2288c3..89c9037 100644
--- a/test/CodeGen/ARM/2009-11-02-NegativeLane.ll
+++ b/test/CodeGen/ARM/2009-11-02-NegativeLane.ll
@@ -1,4 +1,4 @@
-; RUN: llc -mcpu=cortex-a8 < %s | grep vdup.32
+; RUN: llc -mcpu=cortex-a8 < %s | grep vdup.16
 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64"
 target triple = "armv7-eabi"
 
@@ -7,12 +7,12 @@
   br i1 undef, label %return, label %bb
 
 bb:                                               ; preds = %bb, %entry
-  %0 = load float* undef, align 4                 ; <float> [#uses=1]
-  %1 = insertelement <4 x float> undef, float %0, i32 2 ; <<4 x float>> [#uses=1]
-  %2 = insertelement <4 x float> %1, float undef, i32 3 ; <<4 x float>> [#uses=1]
-  %3 = fmul <4 x float> undef, %2                 ; <<4 x float>> [#uses=1]
-  %4 = extractelement <4 x float> %3, i32 1       ; <float> [#uses=1]
-  store float %4, float* undef, align 4
+  %0 = load i16* undef, align 2
+  %1 = insertelement <8 x i16> undef, i16 %0, i32 2
+  %2 = insertelement <8 x i16> %1, i16 undef, i32 3
+  %3 = mul <8 x i16> %2, %2
+  %4 = extractelement <8 x i16> %3, i32 2
+  store i16 %4, i16* undef, align 2
   br i1 undef, label %return, label %bb
 
 return:                                           ; preds = %bb, %entry