diff --git a/lib/Target/X86/InstPrinter/X86InstComments.cpp b/lib/Target/X86/InstPrinter/X86InstComments.cpp
index a6eec371e2b..55a05f46259 100644
--- a/lib/Target/X86/InstPrinter/X86InstComments.cpp
+++ b/lib/Target/X86/InstPrinter/X86InstComments.cpp
@@ -163,14 +163,16 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::SHUFPDrmi:
-    DecodeSHUFPMask(MVT::v2f64, MI->getOperand(3).getImm(), ShuffleMask);
+    DecodeSHUFPMask(MVT::v2f64, MI->getOperand(MI->getNumOperands()-1).getImm(),
+                    ShuffleMask);
     Src1Name = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VSHUFPDrri:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VSHUFPDrmi:
-    DecodeSHUFPMask(MVT::v2f64, MI->getOperand(3).getImm(), ShuffleMask);
+    DecodeSHUFPMask(MVT::v2f64, MI->getOperand(MI->getNumOperands()-1).getImm(),
+                    ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
@@ -178,7 +180,8 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VSHUFPDYrmi:
-    DecodeSHUFPMask(MVT::v4f64, MI->getOperand(3).getImm(), ShuffleMask);
+    DecodeSHUFPMask(MVT::v4f64, MI->getOperand(MI->getNumOperands()-1).getImm(),
+                    ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
@@ -187,14 +190,16 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::SHUFPSrmi:
-    DecodeSHUFPMask(MVT::v4f32, MI->getOperand(3).getImm(), ShuffleMask);
+    DecodeSHUFPMask(MVT::v4f32, MI->getOperand(MI->getNumOperands()-1).getImm(),
+                    ShuffleMask);
     Src1Name = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VSHUFPSrri:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VSHUFPSrmi:
-    DecodeSHUFPMask(MVT::v4f32, MI->getOperand(3).getImm(), ShuffleMask);
+    DecodeSHUFPMask(MVT::v4f32, MI->getOperand(MI->getNumOperands()-1).getImm(),
+                    ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
@@ -202,7 +207,8 @@ void llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VSHUFPSYrmi:
-    DecodeSHUFPMask(MVT::v8f32, MI->getOperand(3).getImm(), ShuffleMask);
+    DecodeSHUFPMask(MVT::v8f32, MI->getOperand(MI->getNumOperands()-1).getImm(),
+                    ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
diff --git a/test/CodeGen/X86/avx-vshufp.ll b/test/CodeGen/X86/avx-vshufp.ll
index 7ec3a44e799..0ccbc594a0d 100644
--- a/test/CodeGen/X86/avx-vshufp.ll
+++ b/test/CodeGen/X86/avx-vshufp.ll
@@ -7,6 +7,15 @@ entry:
   ret <8 x float> %shuffle
 }
 
+; CHECK: vshufps  $-53, (%
+define <8 x float> @A2(<8 x float>* %a, <8 x float>* %b) nounwind uwtable readnone ssp {
+entry:
+  %a2 = load <8 x float>* %a
+  %b2 = load <8 x float>* %b
+  %shuffle = shufflevector <8 x float> %a2, <8 x float> %b2, <8 x i32> <i32 3, i32 2, i32 8, i32 11, i32 7, i32 6, i32 12, i32 15>
+  ret <8 x float> %shuffle
+}
+
 ; CHECK: vshufpd  $10, %ymm
 define <4 x double> @B(<4 x double> %a, <4 x double> %b) nounwind uwtable readnone ssp {
 entry:
@@ -14,6 +23,15 @@ entry:
   ret <4 x double> %shuffle
 }
 
+; CHECK: vshufpd  $10, (%
+define <4 x double> @B2(<4 x double>* %a, <4 x double>* %b) nounwind uwtable readnone ssp {
+entry:
+  %a2 = load <4 x double>* %a
+  %b2 = load <4 x double>* %b
+  %shuffle = shufflevector <4 x double> %a2, <4 x double> %b2, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
+  ret <4 x double> %shuffle
+}
+
 ; CHECK: vshufps  $-53, %ymm
 define <8 x float> @C(<8 x float> %a, <8 x float> %b) nounwind uwtable readnone ssp {
 entry: