diff --git a/lib/Target/CellSPU/SPUISelLowering.cpp b/lib/Target/CellSPU/SPUISelLowering.cpp
index 65eecbaf562..148cb46c97f 100644
--- a/lib/Target/CellSPU/SPUISelLowering.cpp
+++ b/lib/Target/CellSPU/SPUISelLowering.cpp
@@ -1746,15 +1746,20 @@ static SDValue LowerVECTOR_SHUFFLE(SDValue Op, SelectionDAG &DAG) {
   unsigned V0Elt = 0;
   bool monotonic = true;
   bool rotate = true;
+  EVT maskVT;             // which of the c?d instructions to use
 
   if (EltVT == MVT::i8) {
     V2EltIdx0 = 16;
+    maskVT = MVT::v16i8; 
   } else if (EltVT == MVT::i16) {
     V2EltIdx0 = 8;
+    maskVT = MVT::v8i16;
   } else if (EltVT == MVT::i32 || EltVT == MVT::f32) {
     V2EltIdx0 = 4;
+    maskVT = MVT::v4i32;
   } else if (EltVT == MVT::i64 || EltVT == MVT::f64) {
     V2EltIdx0 = 2;
+    maskVT = MVT::v2i64;
   } else
     llvm_unreachable("Unhandled vector type in LowerVECTOR_SHUFFLE");
 
@@ -1800,16 +1805,16 @@ static SDValue LowerVECTOR_SHUFFLE(SDValue Op, SelectionDAG &DAG) {
     // Compute mask and shuffle
     MachineFunction &MF = DAG.getMachineFunction();
     MachineRegisterInfo &RegInfo = MF.getRegInfo();
-    unsigned VReg = RegInfo.createVirtualRegister(&SPU::R32CRegClass);
     EVT PtrVT = DAG.getTargetLoweringInfo().getPointerTy();
-    // Initialize temporary register to 0
-    SDValue InitTempReg =
-      DAG.getCopyToReg(DAG.getEntryNode(), dl, VReg, DAG.getConstant(0, PtrVT));
-    // Copy register's contents as index in SHUFFLE_MASK:
-    SDValue ShufMaskOp =
-      DAG.getNode(SPUISD::SHUFFLE_MASK, dl, MVT::v4i32,
-                  DAG.getTargetConstant(V2Elt, MVT::i32),
-                  DAG.getCopyFromReg(InitTempReg, dl, VReg, PtrVT));
+
+    // As SHUFFLE_MASK becomes a c?d instruction, feed it an address
+    // R1 ($sp) is used here only as it is guaranteed to have last bits zero
+    SDValue Pointer = DAG.getNode(SPUISD::IndirectAddr, dl, PtrVT,
+                                DAG.getRegister(SPU::R1, PtrVT),
+                                DAG.getConstant(V2Elt, MVT::i32));
+    SDValue ShufMaskOp = DAG.getNode(SPUISD::SHUFFLE_MASK, dl, 
+                                     maskVT, Pointer);
+
     // Use shuffle mask in SHUFB synthetic instruction:
     return DAG.getNode(SPUISD::SHUFB, dl, V1.getValueType(), V2, V1,
                        ShufMaskOp);
diff --git a/test/CodeGen/CellSPU/shuffles.ll b/test/CodeGen/CellSPU/shuffles.ll
new file mode 100644
index 00000000000..20ac5705586
--- /dev/null
+++ b/test/CodeGen/CellSPU/shuffles.ll
@@ -0,0 +1,10 @@
+; RUN: llc --march=cellspu < %s | FileCheck %s
+
+define <4 x float> @shuffle(<4 x float> %param1, <4 x float> %param2) {
+  ; CHECK: cwd {{\$.}}, 0($sp)
+  ; CHECK: shufb {{\$., \$4, \$3, \$.}}
+  %val= shufflevector <4 x float> %param1, <4 x float> %param2, <4 x i32> <i32 4,i32 1,i32 2,i32 3>
+  ret <4 x float> %val
+  
+}
+