diff --git a/lib/Target/X86/InstPrinter/X86InstComments.cpp b/lib/Target/X86/InstPrinter/X86InstComments.cpp
index 6e87efa..a6eec37 100644
--- a/lib/Target/X86/InstPrinter/X86InstComments.cpp
+++ b/lib/Target/X86/InstPrinter/X86InstComments.cpp
@@ -163,14 +163,22 @@
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::SHUFPDrmi:
-    DecodeSHUFPSMask(2, MI->getOperand(3).getImm(), ShuffleMask);
+    DecodeSHUFPMask(MVT::v2f64, MI->getOperand(3).getImm(), ShuffleMask);
     Src1Name = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VSHUFPDrri:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VSHUFPDrmi:
-    DecodeSHUFPSMask(2, MI->getOperand(3).getImm(), ShuffleMask);
+    DecodeSHUFPMask(MVT::v2f64, MI->getOperand(3).getImm(), ShuffleMask);
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    break;
+  case X86::VSHUFPDYrri:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VSHUFPDYrmi:
+    DecodeSHUFPMask(MVT::v4f64, MI->getOperand(3).getImm(), ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
@@ -179,14 +187,22 @@
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::SHUFPSrmi:
-    DecodeSHUFPSMask(4, MI->getOperand(3).getImm(), ShuffleMask);
+    DecodeSHUFPMask(MVT::v4f32, MI->getOperand(3).getImm(), ShuffleMask);
     Src1Name = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VSHUFPSrri:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     // FALL THROUGH.
   case X86::VSHUFPSrmi:
-    DecodeSHUFPSMask(4, MI->getOperand(3).getImm(), ShuffleMask);
+    DecodeSHUFPMask(MVT::v4f32, MI->getOperand(3).getImm(), ShuffleMask);
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
+    break;
+  case X86::VSHUFPSYrri:
+    Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VSHUFPSYrmi:
+    DecodeSHUFPMask(MVT::v8f32, MI->getOperand(3).getImm(), ShuffleMask);
     Src1Name = getRegName(MI->getOperand(1).getReg());
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
@@ -284,29 +300,47 @@
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VPERMILPSri:
-    DecodeVPERMILPSMask(4, MI->getOperand(2).getImm(),
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    // FALL THROUGH.
+  case X86::VPERMILPSmi:
+    DecodeVPERMILPSMask(4, MI->getOperand(MI->getNumOperands()-1).getImm(),
                         ShuffleMask);
-    Src1Name = getRegName(MI->getOperand(0).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VPERMILPSYri:
-    DecodeVPERMILPSMask(8, MI->getOperand(2).getImm(),
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    // FALL THROUGH.
+  case X86::VPERMILPSYmi:
+    DecodeVPERMILPSMask(8, MI->getOperand(MI->getNumOperands()-1).getImm(),
                         ShuffleMask);
-    Src1Name = getRegName(MI->getOperand(0).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VPERMILPDri:
-    DecodeVPERMILPDMask(2, MI->getOperand(2).getImm(),
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    // FALL THROUGH.
+  case X86::VPERMILPDmi:
+    DecodeVPERMILPDMask(2, MI->getOperand(MI->getNumOperands()-1).getImm(),
                         ShuffleMask);
-    Src1Name = getRegName(MI->getOperand(0).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VPERMILPDYri:
-    DecodeVPERMILPDMask(4, MI->getOperand(2).getImm(),
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    // FALL THROUGH.
+  case X86::VPERMILPDYmi:
+    DecodeVPERMILPDMask(4, MI->getOperand(MI->getNumOperands()-1).getImm(),
                         ShuffleMask);
-    Src1Name = getRegName(MI->getOperand(0).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
     break;
   case X86::VPERM2F128rr:
-    DecodeVPERM2F128Mask(MI->getOperand(3).getImm(), ShuffleMask);
-    Src1Name = getRegName(MI->getOperand(1).getReg());
+  case X86::VPERM2I128rr:
     Src2Name = getRegName(MI->getOperand(2).getReg());
+    // FALL THROUGH.
+  case X86::VPERM2F128rm:
+  case X86::VPERM2I128rm:
+    DecodeVPERM2F128Mask(MI->getOperand(MI->getNumOperands()-1).getImm(),
+                         ShuffleMask);
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DestName = getRegName(MI->getOperand(0).getReg());
     break;
   }
 
diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
index f6c9d7b..cf1c55e 100644
--- a/lib/Target/X86/Utils/X86ShuffleDecode.cpp
+++ b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
@@ -128,17 +128,27 @@
   }
 }
 
-void DecodeSHUFPSMask(unsigned NElts, unsigned Imm,
-                      SmallVectorImpl<unsigned> &ShuffleMask) {
-  // Part that reads from dest.
-  for (unsigned i = 0; i != NElts/2; ++i) {
-    ShuffleMask.push_back(Imm % NElts);
-    Imm /= NElts;
-  }
-  // Part that reads from src.
-  for (unsigned i = 0; i != NElts/2; ++i) {
-    ShuffleMask.push_back(Imm % NElts + NElts);
-    Imm /= NElts;
+void DecodeSHUFPMask(EVT VT, unsigned Imm,
+                     SmallVectorImpl<unsigned> &ShuffleMask) {
+  unsigned NumElts = VT.getVectorNumElements();
+
+  unsigned NumLanes = VT.getSizeInBits() / 128;
+  unsigned NumLaneElts = NumElts / NumLanes;
+
+  int NewImm = Imm;
+  for (unsigned l = 0; l < NumLanes; ++l) {
+    unsigned LaneStart = l * NumLaneElts;
+    // Part that reads from dest.
+    for (unsigned i = 0; i != NumLaneElts/2; ++i) {
+      ShuffleMask.push_back(NewImm % NumLaneElts + LaneStart);
+      NewImm /= NumLaneElts;
+    }
+    // Part that reads from src.
+    for (unsigned i = 0; i != NumLaneElts/2; ++i) {
+      ShuffleMask.push_back(NewImm % NumLaneElts + NumElts + LaneStart);
+      NewImm /= NumLaneElts;
+    }
+    if (NumLaneElts == 4) NewImm = Imm; // reload imm
   }
 }
 
diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.h b/lib/Target/X86/Utils/X86ShuffleDecode.h
index 35f6530..48b9ef2 100644
--- a/lib/Target/X86/Utils/X86ShuffleDecode.h
+++ b/lib/Target/X86/Utils/X86ShuffleDecode.h
@@ -64,8 +64,8 @@
 void DecodePUNPCKHMask(unsigned NElts,
                        SmallVectorImpl<unsigned> &ShuffleMask);
 
-void DecodeSHUFPSMask(unsigned NElts, unsigned Imm,
-                      SmallVectorImpl<unsigned> &ShuffleMask);
+void DecodeSHUFPMask(EVT VT, unsigned Imm,
+                     SmallVectorImpl<unsigned> &ShuffleMask);
 
 /// DecodeUNPCKHPMask - This decodes the shuffle masks for unpckhps/unpckhpd
 /// etc.  VT indicates the type of the vector allowing it to handle different
diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index cde20c3..7626db0 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -4567,9 +4567,8 @@
     case X86ISD::SHUFPS:
     case X86ISD::SHUFPD:
       ImmN = N->getOperand(N->getNumOperands()-1);
-      DecodeSHUFPSMask(NumElems,
-                       cast<ConstantSDNode>(ImmN)->getZExtValue(),
-                       ShuffleMask);
+      DecodeSHUFPMask(VT, cast<ConstantSDNode>(ImmN)->getZExtValue(),
+                      ShuffleMask);
       break;
     case X86ISD::PUNPCKH:
       DecodePUNPCKHMask(NumElems, ShuffleMask);
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index 1d8e3ce..97737d1 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -7332,6 +7332,15 @@
           (VPERMILPSYri VR256:$src1, imm:$imm)>;
 def : Pat<(v4i64 (X86VPermilpd VR256:$src1, (i8 imm:$imm))),
           (VPERMILPDYri VR256:$src1, imm:$imm)>;
+def : Pat<(v8f32 (X86VPermilps (memopv8f32 addr:$src1), (i8 imm:$imm))),
+          (VPERMILPSYmi addr:$src1, imm:$imm)>;
+def : Pat<(v4f64 (X86VPermilpd (memopv4f64 addr:$src1), (i8 imm:$imm))),
+          (VPERMILPDYmi addr:$src1, imm:$imm)>;
+def : Pat<(v8i32 (X86VPermilps (bc_v8i32 (memopv4i64 addr:$src1)),
+                               (i8 imm:$imm))),
+          (VPERMILPSYmi addr:$src1, imm:$imm)>;
+def : Pat<(v4i64 (X86VPermilpd (memopv4i64 addr:$src1), (i8 imm:$imm))),
+          (VPERMILPDYmi addr:$src1, imm:$imm)>;
 
 //===----------------------------------------------------------------------===//
 // VPERM2F128 - Permute Floating-Point Values in 128-bit chunks
diff --git a/test/CodeGen/X86/avx-vpermil.ll b/test/CodeGen/X86/avx-vpermil.ll
index 49b2f54..3d521e7 100644
--- a/test/CodeGen/X86/avx-vpermil.ll
+++ b/test/CodeGen/X86/avx-vpermil.ll
@@ -28,6 +28,14 @@
   ret <4 x i64> %shuffle
 }
 
+; CHECK: vpermilpd
+define <4 x i64> @funcQ(<4 x i64>* %a) nounwind uwtable readnone ssp {
+entry:
+  %a2 = load <4 x i64>* %a
+  %shuffle = shufflevector <4 x i64> %a2, <4 x i64> undef, <4 x i32> <i32 1, i32 0, i32 3, i32 3>
+  ret <4 x i64> %shuffle
+}
+
 ; vpermil should match masks like this: <u,3,1,2,4,u,5,6>. Check that the
 ; target specific mask was correctly generated.
 ; CHECK: vpermilps $-100