[X86] Enable v16i8/v32i8/v64i8 rotation on AVX512 targets

[RKSimon]

Summary

We currently rely on generic promotion to vXi16/vXi32 types for rotation lowering on various AVX512 targets.

We can more efficiently perform this by making use of the shl(unpack(x,x),amt) style pattern that we already use for vXi8 rotation by splat amounts, either by widening to a larger vector type or unpacking lo/hi halves of the subvectors so we can access whatever vXi16/vXi32 per-element shifts are supported.

This uncovered an issue in the supportedVectorShiftWithImm/supportedVectorVarShift legality checkers which was using hasAVX512() instead of useAVX512Regs() to detect support for 512-bit vector shifts.

NOTE: I'm actually hoping to eventually reuse this code for shl(unpack(y,x),amt) funnel shift lowering (vXi8 and wider), but initially I just want to ensure we have efficient ISD::ROTL lowering for all targets.

[RKSimon]

Added v64i8 rotation handling as well

[RKSimon]

Delay 512-bit vector splitting until after constant-splat early-out, which avoids the need for a lot of additional VPTERNLOG combines on non-AVX512VL targets.

[LuoYuanke]

llvm/lib/Target/X86/X86ISelLowering.cpp

29896

I notice in line 29981, it is aext(x). Is it SIGN_EXTEND?

[RKSimon]

llvm/lib/Target/X86/X86ISelLowering.cpp

29896

I did have separate ZERO/ANY_EXTEND nodes at one point but (for pure tidyness) to reuse the ZERO_EXTEND since for the rotation expansion its what the ANY_EXTEND will end up as. I can split it though if you prefer - we'll need to do this if/when this code gets reused for funnel shifts.

[LuoYuanke]

llvm/lib/Target/X86/X86ISelLowering.cpp

29896

No, not necessary to split. I'm just curious on this node. :)

[RKSimon]

Any more comments?

[pengfei]

llvm/lib/Target/X86/X86ISelLowering.cpp

29873–29874

The NumElts may be 64 or 32 to the maximum, so the WideVT can be illegal one?

29880–29882

If the comments are usable for all if and else conditions, the second half comment in 29884 seems redundant. Or the comments should be moved to the if block?

29893

There's extra space here.

llvm/test/CodeGen/X86/vector-fshr-rot-128.ll

626–628

Why do we still widen here rather that use avx instruction like other cases?

llvm/test/CodeGen/X86/vector-fshr-rot-512.ll

859 ↗

(On Diff #392735)

Didn't dig much, why the immediate value keep unchanged while zmm1 and zmm0 swapped?

[RKSimon]

llvm/test/CodeGen/X86/vector-fshr-rot-128.ll

626–628

I'm sorry I don't quite understand - do you mean you'd expect the (much longer) vpblendvb sequence?

[pengfei]

llvm/test/CodeGen/X86/vector-fshr-rot-128.ll

626–628

Sorry, my mistake. I thought the use of zmm register results from the lack of avx512vl. Now I understand we are using the full 512 bit register. Sorry for the noise.

[RKSimon]

Add explicit 128/256/512-bit checks to the supportedVectorShift* helper checkers

[RKSimon]

ping?

[pengfei]

LGTM.

llvm/lib/Target/X86/X86ISelLowering.cpp

     for (auto VT : { MVT::v16i8, MVT::v8i16, MVT::v4i32, MVT::v2i64 }) {
       setOperationAction(ISD::SRL,              VT, Custom);
       setOperationAction(ISD::SHL,              VT, Custom);
       setOperationAction(ISD::SRA,              VT, Custom);
     }
 
     setOperationAction(ISD::ROTL,               MVT::v4i32, Custom);
     setOperationAction(ISD::ROTL,               MVT::v8i16, Custom);
-
-    // With 512-bit registers or AVX512VL+BW, expanding (and promoting the
-    // shifts) is better.
-    if (!Subtarget.useAVX512Regs() &&
-        !(Subtarget.hasBWI() && Subtarget.hasVLX()))
-      setOperationAction(ISD::ROTL,             MVT::v16i8, Custom);
+    setOperationAction(ISD::ROTL,               MVT::v16i8, Custom);
 
     setOperationAction(ISD::STRICT_FSQRT,       MVT::v2f64, Legal);
     setOperationAction(ISD::STRICT_FADD,        MVT::v2f64, Legal);
     setOperationAction(ISD::STRICT_FSUB,        MVT::v2f64, Legal);
     setOperationAction(ISD::STRICT_FMUL,        MVT::v2f64, Legal);
     setOperationAction(ISD::STRICT_FDIV,        MVT::v2f64, Legal);
   }
 

     // These types need custom splitting if their input is a 128-bit vector.
     setOperationAction(ISD::SIGN_EXTEND,       MVT::v8i64,  Custom);
     setOperationAction(ISD::SIGN_EXTEND,       MVT::v16i32, Custom);
     setOperationAction(ISD::ZERO_EXTEND,       MVT::v8i64,  Custom);
     setOperationAction(ISD::ZERO_EXTEND,       MVT::v16i32, Custom);
 
     setOperationAction(ISD::ROTL,              MVT::v8i32,  Custom);
     setOperationAction(ISD::ROTL,              MVT::v16i16, Custom);
-
-    // With BWI, expanding (and promoting the shifts) is the better.
-    if (!Subtarget.useBWIRegs())
-      setOperationAction(ISD::ROTL,            MVT::v32i8,  Custom);
+    setOperationAction(ISD::ROTL,              MVT::v32i8,  Custom);
 
     setOperationAction(ISD::SELECT,            MVT::v4f64, Custom);
     setOperationAction(ISD::SELECT,            MVT::v4i64, Custom);
     setOperationAction(ISD::SELECT,            MVT::v8i32, Custom);
     setOperationAction(ISD::SELECT,            MVT::v16i16, Custom);
     setOperationAction(ISD::SELECT,            MVT::v32i8, Custom);
     setOperationAction(ISD::SELECT,            MVT::v8f32, Custom);
 

 
 // Return true if the required (according to Opcode) shift-imm form is natively
 // supported by the Subtarget
 static bool supportedVectorShiftWithImm(MVT VT, const X86Subtarget &Subtarget,
                                         unsigned Opcode) {
   if (VT.getScalarSizeInBits() < 16)
     return false;
 
-  if (VT.is512BitVector() && Subtarget.hasAVX512() &&
+  if (VT.is512BitVector() && Subtarget.useAVX512Regs() &&
       (VT.getScalarSizeInBits() > 16 || Subtarget.hasBWI()))
     return true;
 
   bool LShift = (VT.is128BitVector() && Subtarget.hasSSE2()) ||
                 (VT.is256BitVector() && Subtarget.hasInt256());
 
   bool AShift = LShift && (Subtarget.hasAVX512() ||
                            (VT != MVT::v2i64 && VT != MVT::v4i64));

 
   if (!Subtarget.hasInt256() || VT.getScalarSizeInBits() < 16)
     return false;
 
   // vXi16 supported only on AVX-512, BWI
   if (VT.getScalarSizeInBits() == 16 && !Subtarget.hasBWI())
     return false;
 
-  if (Subtarget.hasAVX512())
+  if (Subtarget.hasAVX512() &&
+      (Subtarget.useAVX512Regs() || !VT.is512BitVector()))
     return true;
 
   bool LShift = VT.is128BitVector() || VT.is256BitVector();
   bool AShift = LShift &&  VT != MVT::v2i64 && VT != MVT::v4i64;
   return (Opcode == ISD::SRA) ? AShift : LShift;
 }
 
 static SDValue LowerScalarImmediateShift(SDValue Op, SelectionDAG &DAG,

   if (IsCstSplat)
     return SDValue();
 
   bool IsSplatAmt = DAG.isSplatValue(Amt);
   SDValue AmtMask = DAG.getConstant(EltSizeInBits - 1, DL, VT);
 
   // v16i8/v32i8: Split rotation into rot4/rot2/rot1 stages and select by
   // the amount bit.
+  // TODO: We're doing nothing here that we couldn't do for funnel shifts.
   if (EltSizeInBits == 8) {
     if (ISD::isBuildVectorOfConstantSDNodes(Amt.getNode()))
       return SDValue();
 
     // Check for a hidden ISD::ROTR, vXi8 lowering can handle both, but we
     // currently hit infinite loops in legalization if we allow ISD::ROTR.
     // FIXME: Infinite ROTL<->ROTR legalization in TargetLowering::expandROT.
     SDValue HiddenROTRAmt;
     if (Amt.getOpcode() == ISD::SUB &&
         ISD::isBuildVectorAllZeros(Amt.getOperand(0).getNode()))
       HiddenROTRAmt = Amt.getOperand(1);
 
     MVT ExtVT = MVT::getVectorVT(MVT::i16, NumElts / 2);
-
-    // If the amount is a splat, attempt to fold as unpack(x,x) << zext(y):
+    MVT WideVT =
+        MVT::getVectorVT(Subtarget.hasBWI() ? MVT::i16 : MVT::i32, NumElts);
+    unsigned ShiftOpc = HiddenROTRAmt ? ISD::SRL : ISD::SHL;
+    unsigned ShiftX86Opc = HiddenROTRAmt ? X86ISD::VSRLI : X86ISD::VSHLI;
+    SDValue AmtMod = DAG.getNode(ISD::AND, DL, VT,
+                                 HiddenROTRAmt ? HiddenROTRAmt : Amt, AmtMask);
+
+    // Attempt to fold as unpack(x,x) << zext(y):
     // rotl(x,y) -> (((aext(x) << bw) | zext(x)) << (y & (bw-1))) >> bw.
     // rotr(x,y) -> (((aext(x) << bw) | zext(x)) >> (y & (bw-1))).
-    if (SDValue BaseRotAmt = DAG.getSplatValue(DAG.getNode(
-            ISD::AND, DL, VT, HiddenROTRAmt ? HiddenROTRAmt : Amt, AmtMask))) {
-      unsigned ShiftX86Opc = HiddenROTRAmt ? X86ISD::VSRLI : X86ISD::VSHLI;
+    if (SDValue BaseRotAmt = DAG.getSplatValue(AmtMod)) {
+      // If the amount is a splat, attempt to fold as unpack(x,x) << zext(y).
       BaseRotAmt = DAG.getNode(ISD::ZERO_EXTEND, DL, MVT::i32, BaseRotAmt);
       SDValue Lo = DAG.getBitcast(ExtVT, getUnpackl(DAG, DL, VT, R, R));
       SDValue Hi = DAG.getBitcast(ExtVT, getUnpackh(DAG, DL, VT, R, R));
       Lo = getTargetVShiftNode(ShiftX86Opc, DL, ExtVT, Lo, BaseRotAmt,
                                Subtarget, DAG);
       Hi = getTargetVShiftNode(ShiftX86Opc, DL, ExtVT, Hi, BaseRotAmt,
                                Subtarget, DAG);
       return getPack(DAG, Subtarget, DL, VT, Lo, Hi, !HiddenROTRAmt);
+    } else if (supportedVectorVarShift(WideVT, Subtarget, ShiftOpc) && 
+               supportedVectorShiftWithImm(WideVT, Subtarget, ShiftOpc)) {
+      // See if we can perform this by widening to vXi16.
+      R = DAG.getNode(ISD::ZERO_EXTEND, DL, WideVT, R);
+      R = DAG.getNode(
+          ISD::OR, DL, WideVT, R,
+          getTargetVShiftByConstNode(X86ISD::VSHLI, DL, WideVT, R, 8, DAG));
+      Amt = DAG.getNode(ISD::ZERO_EXTEND, DL, WideVT, AmtMod);
+      R = DAG.getNode(ShiftOpc, DL, WideVT, R, Amt);
+      if (!HiddenROTRAmt)
+        R = getTargetVShiftByConstNode(X86ISD::VSRLI, DL, WideVT, R, 8, DAG);
+      return DAG.getNode(ISD::TRUNCATE, DL, VT, R);
+    } else if (supportedVectorVarShift(ExtVT, Subtarget, ShiftOpc)) {
+      // See if we can perform this by unpacking to lo/hi vXi16.
+      SDValue Z = getZeroVector(VT, Subtarget, DAG, DL);
+      SDValue RLo = DAG.getBitcast(ExtVT, getUnpackl(DAG, DL, VT, R, R));
+      SDValue RHi = DAG.getBitcast(ExtVT, getUnpackh(DAG, DL, VT, R, R));
+      SDValue ALo = DAG.getBitcast(ExtVT, getUnpackl(DAG, DL, VT, AmtMod, Z));
+      SDValue AHi = DAG.getBitcast(ExtVT, getUnpackh(DAG, DL, VT, AmtMod, Z));
+      SDValue Lo = DAG.getNode(ShiftOpc, DL, ExtVT, RLo, ALo);
+      SDValue Hi = DAG.getNode(ShiftOpc, DL, ExtVT, RHi, AHi);
+      return getPack(DAG, Subtarget, DL, VT, Lo, Hi, !HiddenROTRAmt);
     }
 
     // We don't need ModuloAmt here as we just peek at individual bits.
     auto SignBitSelect = [&](MVT SelVT, SDValue Sel, SDValue V0, SDValue V1) {
       if (Subtarget.hasSSE41()) {
         // On SSE41 targets we can use PBLENDVB which selects bytes based just
         // on the sign bit.
         V0 = DAG.getBitcast(VT, V0);

llvm/test/CodeGen/X86/min-legal-vector-width.ll

; CHECK-NEXT:    retq
   %y = load <8 x i64>, <8 x i64>* %yptr
   %cmp = icmp slt <8 x i64> %x, %y
   %ext = zext <8 x i1> %cmp to <8 x i64>
   store <8 x i64> %ext, <8 x i64>* %zptr
   ret void
 }
 
 define <16 x i8> @var_rotate_v16i8(<16 x i8> %a, <16 x i8> %b) nounwind "min-legal-vector-width"="256" {
-; CHECK-LABEL: var_rotate_v16i8:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; CHECK-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
-; CHECK-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; CHECK-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; CHECK-NEXT:    vpsllvw %ymm1, %ymm0, %ymm1
-; CHECK-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero
-; CHECK-NEXT:    vpsrlvw %ymm2, %ymm0, %ymm0
-; CHECK-NEXT:    vpor %ymm0, %ymm1, %ymm0
-; CHECK-NEXT:    vpmovwb %ymm0, %xmm0
-; CHECK-NEXT:    vzeroupper
-; CHECK-NEXT:    retq
+; CHECK-AVX512-LABEL: var_rotate_v16i8:
+; CHECK-AVX512:       # %bb.0:
+; CHECK-AVX512-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; CHECK-AVX512-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
+; CHECK-AVX512-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; CHECK-AVX512-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; CHECK-AVX512-NEXT:    vpsllvw %ymm1, %ymm0, %ymm0
+; CHECK-AVX512-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; CHECK-AVX512-NEXT:    vpmovwb %ymm0, %xmm0
+; CHECK-AVX512-NEXT:    vzeroupper
+; CHECK-AVX512-NEXT:    retq
+;
+; CHECK-VBMI-LABEL: var_rotate_v16i8:
+; CHECK-VBMI:       # %bb.0:
+; CHECK-VBMI-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
+; CHECK-VBMI-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; CHECK-VBMI-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; CHECK-VBMI-NEXT:    vpermb %ymm0, %ymm2, %ymm0
+; CHECK-VBMI-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
+; CHECK-VBMI-NEXT:    vpsllvw %ymm1, %ymm0, %ymm0
+; CHECK-VBMI-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; CHECK-VBMI-NEXT:    vpmovwb %ymm0, %xmm0
+; CHECK-VBMI-NEXT:    vzeroupper
+; CHECK-VBMI-NEXT:    retq
   %b8 = sub <16 x i8> <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>, %b
   %shl = shl <16 x i8> %a, %b
   %lshr = lshr <16 x i8> %a, %b8
   %or = or <16 x i8> %shl, %lshr
   ret <16 x i8> %or
 }
 
 define <32 x i8> @var_rotate_v32i8(<32 x i8> %a, <32 x i8> %b) nounwind "min-legal-vector-width"="256" {
 ; CHECK-LABEL: var_rotate_v32i8:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpsllw $4, %ymm0, %ymm2
-; CHECK-NEXT:    vpsrlw $4, %ymm0, %ymm3
-; CHECK-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm2, %ymm3
-; CHECK-NEXT:    vpsllw $5, %ymm1, %ymm1
-; CHECK-NEXT:    vpblendvb %ymm1, %ymm3, %ymm0, %ymm0
-; CHECK-NEXT:    vpsllw $2, %ymm0, %ymm2
-; CHECK-NEXT:    vpsrlw $6, %ymm0, %ymm3
-; CHECK-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm2, %ymm3
-; CHECK-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
-; CHECK-NEXT:    vpblendvb %ymm1, %ymm3, %ymm0, %ymm0
-; CHECK-NEXT:    vpsrlw $7, %ymm0, %ymm2
-; CHECK-NEXT:    vpaddb %ymm0, %ymm0, %ymm3
-; CHECK-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm3
-; CHECK-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
-; CHECK-NEXT:    vpblendvb %ymm1, %ymm3, %ymm0, %ymm0
+; CHECK-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
+; CHECK-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; CHECK-NEXT:    vpunpckhbw {{.*#+}} ymm3 = ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15],ymm1[24],ymm2[24],ymm1[25],ymm2[25],ymm1[26],ymm2[26],ymm1[27],ymm2[27],ymm1[28],ymm2[28],ymm1[29],ymm2[29],ymm1[30],ymm2[30],ymm1[31],ymm2[31]
+; CHECK-NEXT:    vpunpckhbw {{.*#+}} ymm4 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; CHECK-NEXT:    vpsllvw %ymm3, %ymm4, %ymm3
+; CHECK-NEXT:    vpsrlw $8, %ymm3, %ymm3
+; CHECK-NEXT:    vpunpcklbw {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[16],ymm2[16],ymm1[17],ymm2[17],ymm1[18],ymm2[18],ymm1[19],ymm2[19],ymm1[20],ymm2[20],ymm1[21],ymm2[21],ymm1[22],ymm2[22],ymm1[23],ymm2[23]
+; CHECK-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
+; CHECK-NEXT:    vpsllvw %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; CHECK-NEXT:    vpackuswb %ymm3, %ymm0, %ymm0
 ; CHECK-NEXT:    retq
   %b8 = sub <32 x i8> <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>, %b
   %shl = shl <32 x i8> %a, %b
   %lshr = lshr <32 x i8> %a, %b8
   %or = or <32 x i8> %shl, %lshr
   ret <32 x i8> %or
 }
 

llvm/test/CodeGen/X86/vector-fshl-rot-128.ll

 ; AVX-NEXT:    vpaddb %xmm0, %xmm0, %xmm3
 ; AVX-NEXT:    vpor %xmm2, %xmm3, %xmm2
 ; AVX-NEXT:    vpaddb %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    retq
 ;
 ; AVX512F-LABEL: var_funnnel_v16i8:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512F-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero,xmm3[4],zero,zero,zero,xmm3[5],zero,zero,zero,xmm3[6],zero,zero,zero,xmm3[7],zero,zero,zero,xmm3[8],zero,zero,zero,xmm3[9],zero,zero,zero,xmm3[10],zero,zero,zero,xmm3[11],zero,zero,zero,xmm3[12],zero,zero,zero,xmm3[13],zero,zero,zero,xmm3[14],zero,zero,zero,xmm3[15],zero,zero,zero
 ; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512F-NEXT:    vpsllvd %zmm3, %zmm0, %zmm3
-; AVX512F-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512F-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512F-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; AVX512F-NEXT:    vpslld $8, %zmm0, %zmm2
+; AVX512F-NEXT:    vpord %zmm2, %zmm0, %zmm0
+; AVX512F-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
-; AVX512F-NEXT:    vpsrlvd %zmm1, %zmm0, %zmm0
-; AVX512F-NEXT:    vpord %zmm0, %zmm3, %zmm0
+; AVX512F-NEXT:    vpsllvd %zmm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpsrld $8, %zmm0, %zmm0
 ; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
 ; AVX512F-NEXT:    vzeroupper
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: var_funnnel_v16i8:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VL-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero,xmm3[4],zero,zero,zero,xmm3[5],zero,zero,zero,xmm3[6],zero,zero,zero,xmm3[7],zero,zero,zero,xmm3[8],zero,zero,zero,xmm3[9],zero,zero,zero,xmm3[10],zero,zero,zero,xmm3[11],zero,zero,zero,xmm3[12],zero,zero,zero,xmm3[13],zero,zero,zero,xmm3[14],zero,zero,zero,xmm3[15],zero,zero,zero
 ; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512VL-NEXT:    vpsllvd %zmm3, %zmm0, %zmm3
-; AVX512VL-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VL-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VL-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; AVX512VL-NEXT:    vpslld $8, %zmm0, %zmm2
+; AVX512VL-NEXT:    vpord %zmm2, %zmm0, %zmm0
+; AVX512VL-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
-; AVX512VL-NEXT:    vpsrlvd %zmm1, %zmm0, %zmm0
-; AVX512VL-NEXT:    vpord %zmm0, %zmm3, %zmm0
+; AVX512VL-NEXT:    vpsllvd %zmm1, %zmm0, %zmm0
+; AVX512VL-NEXT:    vpsrld $8, %zmm0, %zmm0
 ; AVX512VL-NEXT:    vpmovdb %zmm0, %xmm0
 ; AVX512VL-NEXT:    vzeroupper
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: var_funnnel_v16i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512BW-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
-; AVX512BW-NEXT:    vpsllvw %zmm3, %zmm0, %zmm3
-; AVX512BW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512BW-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512BW-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX512BW-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm0
-; AVX512BW-NEXT:    vpor %ymm0, %ymm3, %ymm0
+; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX512BW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; AVX512BW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512BW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: var_funnnel_v16i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VLBW-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLBW-NEXT:    vpsllvw %ymm3, %ymm0, %ymm3
-; AVX512VLBW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLBW-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VLBW-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; AVX512VLBW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX512VLBW-NEXT:    vpsrlvw %ymm1, %ymm0, %ymm0
-; AVX512VLBW-NEXT:    vpor %ymm0, %ymm3, %ymm0
+; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; AVX512VLBW-NEXT:    vpsllvw %ymm1, %ymm0, %ymm0
+; AVX512VLBW-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512VLBW-NEXT:    vpmovwb %ymm0, %xmm0
 ; AVX512VLBW-NEXT:    vzeroupper
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: var_funnnel_v16i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
-; AVX512VBMI2-NEXT:    vpsllvw %zmm3, %zmm0, %zmm3
-; AVX512VBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VBMI2-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; AVX512VBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX512VBMI2-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm0
-; AVX512VBMI2-NEXT:    vpor %ymm0, %ymm3, %ymm0
+; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX512VBMI2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; AVX512VBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
+; AVX512VBMI2-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VBMI2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX512VBMI2-NEXT:    vzeroupper
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: var_funnnel_v16i8:
 ; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VLVBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLVBMI2-NEXT:    vpsllvw %ymm3, %ymm0, %ymm3
-; AVX512VLVBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLVBMI2-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VLVBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; AVX512VLVBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX512VLVBMI2-NEXT:    vpsrlvw %ymm1, %ymm0, %ymm0
-; AVX512VLVBMI2-NEXT:    vpor %ymm0, %ymm3, %ymm0
+; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX512VLVBMI2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; AVX512VLVBMI2-NEXT:    vpsllvw %ymm1, %ymm0, %ymm0
+; AVX512VLVBMI2-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    vpmovwb %ymm0, %xmm0
 ; AVX512VLVBMI2-NEXT:    vzeroupper
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOP-LABEL: var_funnnel_v16i8:
 ; XOP:       # %bb.0:
 ; XOP-NEXT:    vprotb %xmm1, %xmm0, %xmm0
 ; XOP-NEXT:    retq

 ; AVX-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; AVX-NEXT:    vpsllw %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vpsrlw $8, %xmm0, %xmm0
 ; AVX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    retq
 ;
 ; AVX512F-LABEL: splatvar_funnnel_v16i8:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512F-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512F-NEXT:    vpslld %xmm3, %zmm0, %zmm3
-; AVX512F-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512F-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512F-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512F-NEXT:    vpsrld %xmm1, %zmm0, %zmm0
-; AVX512F-NEXT:    vpord %zmm0, %zmm3, %zmm0
-; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
-; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512F-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512F-NEXT:    vpsllw %xmm1, %xmm2, %xmm2
+; AVX512F-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX512F-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512F-NEXT:    vpsllw %xmm1, %xmm0, %xmm0
+; AVX512F-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX512F-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: splatvar_funnnel_v16i8:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VL-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512VL-NEXT:    vpslld %xmm3, %zmm0, %zmm3
-; AVX512VL-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VL-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VL-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VL-NEXT:    vpsrld %xmm1, %zmm0, %zmm0
-; AVX512VL-NEXT:    vpord %zmm0, %zmm3, %zmm0
-; AVX512VL-NEXT:    vpmovdb %zmm0, %xmm0
-; AVX512VL-NEXT:    vzeroupper
+; AVX512VL-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VL-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512VL-NEXT:    vpsllw %xmm1, %xmm2, %xmm2
+; AVX512VL-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX512VL-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512VL-NEXT:    vpsllw %xmm1, %xmm0, %xmm0
+; AVX512VL-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX512VL-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: splatvar_funnnel_v16i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512BW-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512BW-NEXT:    vpsllw %xmm3, %ymm0, %ymm3
-; AVX512BW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512BW-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512BW-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512BW-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0
-; AVX512BW-NEXT:    vpor %ymm0, %ymm3, %ymm0
-; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
-; AVX512BW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
-; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512BW-NEXT:    vpsllw %xmm1, %xmm2, %xmm2
+; AVX512BW-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512BW-NEXT:    vpsllw %xmm1, %xmm0, %xmm0
+; AVX512BW-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX512BW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: splatvar_funnnel_v16i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VLBW-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLBW-NEXT:    vpsllw %xmm3, %ymm0, %ymm3
-; AVX512VLBW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLBW-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VLBW-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VLBW-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0
-; AVX512VLBW-NEXT:    vpor %ymm0, %ymm3, %ymm0
-; AVX512VLBW-NEXT:    vpmovwb %ymm0, %xmm0
-; AVX512VLBW-NEXT:    vzeroupper
+; AVX512VLBW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VLBW-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512VLBW-NEXT:    vpsllw %xmm1, %xmm2, %xmm2
+; AVX512VLBW-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX512VLBW-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512VLBW-NEXT:    vpsllw %xmm1, %xmm0, %xmm0
+; AVX512VLBW-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX512VLBW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: splatvar_funnnel_v16i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VBMI2-NEXT:    vpsllw %xmm3, %ymm0, %ymm3
-; AVX512VBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VBMI2-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VBMI2-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0
-; AVX512VBMI2-NEXT:    vpor %ymm0, %ymm3, %ymm0
-; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
-; AVX512VBMI2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
-; AVX512VBMI2-NEXT:    vzeroupper
+; AVX512VBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VBMI2-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512VBMI2-NEXT:    vpsllw %xmm1, %xmm2, %xmm2
+; AVX512VBMI2-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX512VBMI2-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512VBMI2-NEXT:    vpsllw %xmm1, %xmm0, %xmm0
+; AVX512VBMI2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX512VBMI2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: splatvar_funnnel_v16i8:
 ; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VLVBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLVBMI2-NEXT:    vpsllw %xmm3, %ymm0, %ymm3
-; AVX512VLVBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLVBMI2-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VLVBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VLVBMI2-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0
-; AVX512VLVBMI2-NEXT:    vpor %ymm0, %ymm3, %ymm0
-; AVX512VLVBMI2-NEXT:    vpmovwb %ymm0, %xmm0
-; AVX512VLVBMI2-NEXT:    vzeroupper
+; AVX512VLVBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VLVBMI2-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512VLVBMI2-NEXT:    vpsllw %xmm1, %xmm2, %xmm2
+; AVX512VLVBMI2-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX512VLVBMI2-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512VLVBMI2-NEXT:    vpsllw %xmm1, %xmm0, %xmm0
+; AVX512VLVBMI2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX512VLVBMI2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: splatvar_funnnel_v16i8:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vprotb %xmm1, %xmm0, %xmm0
 ; XOPAVX1-NEXT:    retq

llvm/test/CodeGen/X86/vector-fshl-rot-256.ll

 ; AVX512VL-NEXT:    vpaddb %ymm0, %ymm0, %ymm3
 ; AVX512VL-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm3
 ; AVX512VL-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
 ; AVX512VL-NEXT:    vpblendvb %ymm1, %ymm3, %ymm0, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: var_funnnel_v32i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm3
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero,ymm3[16],zero,ymm3[17],zero,ymm3[18],zero,ymm3[19],zero,ymm3[20],zero,ymm3[21],zero,ymm3[22],zero,ymm3[23],zero,ymm3[24],zero,ymm3[25],zero,ymm3[26],zero,ymm3[27],zero,ymm3[28],zero,ymm3[29],zero,ymm3[30],zero,ymm3[31],zero
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512BW-NEXT:    vpsllvw %zmm3, %zmm0, %zmm3
-; AVX512BW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512BW-NEXT:    vpsubb %ymm1, %ymm4, %ymm1
-; AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512BW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512BW-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm0
-; AVX512BW-NEXT:    vporq %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
+; AVX512BW-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30,32,32,34,34,36,36,38,38,40,40,42,42,44,44,46,46,48,48,50,50,52,52,54,54,56,56,58,58,60,60,62,62]
+; AVX512BW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: var_funnnel_v32i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VLBW-NEXT:    vpand %ymm2, %ymm1, %ymm3
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero,ymm3[16],zero,ymm3[17],zero,ymm3[18],zero,ymm3[19],zero,ymm3[20],zero,ymm3[21],zero,ymm3[22],zero,ymm3[23],zero,ymm3[24],zero,ymm3[25],zero,ymm3[26],zero,ymm3[27],zero,ymm3[28],zero,ymm3[29],zero,ymm3[30],zero,ymm3[31],zero
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLBW-NEXT:    vpsllvw %zmm3, %zmm0, %zmm3
-; AVX512VLBW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLBW-NEXT:    vpsubb %ymm1, %ymm4, %ymm1
-; AVX512VLBW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512VLBW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
 ; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512VLBW-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm0
-; AVX512VLBW-NEXT:    vporq %zmm0, %zmm3, %zmm0
+; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30,32,32,34,34,36,36,38,38,40,40,42,42,44,44,46,46,48,48,50,50,52,52,54,54,56,56,58,58,60,60,62,62]
+; AVX512VLBW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
+; AVX512VLBW-NEXT:    vpsrlw $8, %zmm0, %zmm0
 ; AVX512VLBW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: var_funnnel_v32i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} ymm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VBMI2-NEXT:    vpand %ymm2, %ymm1, %ymm3
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero,ymm3[16],zero,ymm3[17],zero,ymm3[18],zero,ymm3[19],zero,ymm3[20],zero,ymm3[21],zero,ymm3[22],zero,ymm3[23],zero,ymm3[24],zero,ymm3[25],zero,ymm3[26],zero,ymm3[27],zero,ymm3[28],zero,ymm3[29],zero,ymm3[30],zero,ymm3[31],zero
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VBMI2-NEXT:    vpsllvw %zmm3, %zmm0, %zmm3
-; AVX512VBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VBMI2-NEXT:    vpsubb %ymm1, %ymm4, %ymm1
-; AVX512VBMI2-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512VBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
 ; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512VBMI2-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm0
-; AVX512VBMI2-NEXT:    vporq %zmm0, %zmm3, %zmm0
+; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
+; AVX512VBMI2-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30,32,32,34,34,36,36,38,38,40,40,42,42,44,44,46,46,48,48,50,50,52,52,54,54,56,56,58,58,60,60,62,62]
+; AVX512VBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
+; AVX512VBMI2-NEXT:    vpsrlw $8, %zmm0, %zmm0
 ; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: var_funnnel_v32i8:
 ; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vmovdqa {{.*#+}} ymm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VLVBMI2-NEXT:    vpand %ymm2, %ymm1, %ymm3
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero,ymm3[16],zero,ymm3[17],zero,ymm3[18],zero,ymm3[19],zero,ymm3[20],zero,ymm3[21],zero,ymm3[22],zero,ymm3[23],zero,ymm3[24],zero,ymm3[25],zero,ymm3[26],zero,ymm3[27],zero,ymm3[28],zero,ymm3[29],zero,ymm3[30],zero,ymm3[31],zero
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLVBMI2-NEXT:    vpsllvw %zmm3, %zmm0, %zmm3
-; AVX512VLVBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLVBMI2-NEXT:    vpsubb %ymm1, %ymm4, %ymm1
-; AVX512VLVBMI2-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512VLVBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
 ; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512VLVBMI2-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm0
-; AVX512VLVBMI2-NEXT:    vporq %zmm0, %zmm3, %zmm0
+; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
+; AVX512VLVBMI2-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30,32,32,34,34,36,36,38,38,40,40,42,42,44,44,46,46,48,48,50,50,52,52,54,54,56,56,58,58,60,60,62,62]
+; AVX512VLVBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
+; AVX512VLVBMI2-NEXT:    vpsrlw $8, %zmm0, %zmm0
 ; AVX512VLVBMI2-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: var_funnnel_v32i8:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; XOPAVX1-NEXT:    vprotb %xmm2, %xmm3, %xmm2

 ; AVX512VL-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
 ; AVX512VL-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
 ; AVX512VL-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512VL-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: splatvar_funnnel_v32i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512BW-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512BW-NEXT:    vpsllw %xmm3, %zmm0, %zmm3
-; AVX512BW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512BW-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512BW-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512BW-NEXT:    vpsrlw %xmm1, %zmm0, %zmm0
-; AVX512BW-NEXT:    vporq %zmm0, %zmm3, %zmm0
-; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
+; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX512BW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512BW-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-NEXT:    vpsllw %xmm1, %ymm2, %ymm2
+; AVX512BW-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
+; AVX512BW-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
+; AVX512BW-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512BW-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: splatvar_funnnel_v32i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VLBW-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLBW-NEXT:    vpsllw %xmm3, %zmm0, %zmm3
-; AVX512VLBW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLBW-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VLBW-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VLBW-NEXT:    vpsrlw %xmm1, %zmm0, %zmm0
-; AVX512VLBW-NEXT:    vporq %zmm0, %zmm3, %zmm0
-; AVX512VLBW-NEXT:    vpmovwb %zmm0, %ymm0
+; AVX512VLBW-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX512VLBW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VLBW-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VLBW-NEXT:    vpsllw %xmm1, %ymm2, %ymm2
+; AVX512VLBW-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512VLBW-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
+; AVX512VLBW-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
+; AVX512VLBW-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512VLBW-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: splatvar_funnnel_v32i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VBMI2-NEXT:    vpsllw %xmm3, %zmm0, %zmm3
-; AVX512VBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VBMI2-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VBMI2-NEXT:    vpsrlw %xmm1, %zmm0, %zmm0
-; AVX512VBMI2-NEXT:    vporq %zmm0, %zmm3, %zmm0
-; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
+; AVX512VBMI2-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX512VBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VBMI2-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VBMI2-NEXT:    vpsllw %xmm1, %ymm2, %ymm2
+; AVX512VBMI2-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512VBMI2-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
+; AVX512VBMI2-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
+; AVX512VBMI2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512VBMI2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: splatvar_funnnel_v32i8:
 ; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VLVBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLVBMI2-NEXT:    vpsllw %xmm3, %zmm0, %zmm3
-; AVX512VLVBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLVBMI2-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VLVBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VLVBMI2-NEXT:    vpsrlw %xmm1, %zmm0, %zmm0
-; AVX512VLVBMI2-NEXT:    vporq %zmm0, %zmm3, %zmm0
-; AVX512VLVBMI2-NEXT:    vpmovwb %zmm0, %ymm0
+; AVX512VLVBMI2-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX512VLVBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VLVBMI2-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VLVBMI2-NEXT:    vpsllw %xmm1, %ymm2, %ymm2
+; AVX512VLVBMI2-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512VLVBMI2-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
+; AVX512VLVBMI2-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
+; AVX512VLVBMI2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512VLVBMI2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: splatvar_funnnel_v32i8:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; XOPAVX1-NEXT:    vprotb %xmm1, %xmm2, %xmm2

llvm/test/CodeGen/X86/vector-fshr-rot-128.ll

 ; AVX-NEXT:    vpaddb %xmm0, %xmm0, %xmm3
 ; AVX-NEXT:    vpor %xmm2, %xmm3, %xmm2
 ; AVX-NEXT:    vpaddb %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    retq
 ;
 ; AVX512F-LABEL: var_funnnel_v16i8:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512F-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero,xmm3[4],zero,zero,zero,xmm3[5],zero,zero,zero,xmm3[6],zero,zero,zero,xmm3[7],zero,zero,zero,xmm3[8],zero,zero,zero,xmm3[9],zero,zero,zero,xmm3[10],zero,zero,zero,xmm3[11],zero,zero,zero,xmm3[12],zero,zero,zero,xmm3[13],zero,zero,zero,xmm3[14],zero,zero,zero,xmm3[15],zero,zero,zero
 ; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512F-NEXT:    vpsrlvd %zmm3, %zmm0, %zmm3
-; AVX512F-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512F-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512F-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; AVX512F-NEXT:    vpslld $8, %zmm0, %zmm2
+; AVX512F-NEXT:    vpord %zmm2, %zmm0, %zmm0
+; AVX512F-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
-; AVX512F-NEXT:    vpsllvd %zmm1, %zmm0, %zmm0
-; AVX512F-NEXT:    vpord %zmm0, %zmm3, %zmm0
+; AVX512F-NEXT:    vpsrlvd %zmm1, %zmm0, %zmm0
 ; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
 ; AVX512F-NEXT:    vzeroupper
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: var_funnnel_v16i8:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VL-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero,xmm3[4],zero,zero,zero,xmm3[5],zero,zero,zero,xmm3[6],zero,zero,zero,xmm3[7],zero,zero,zero,xmm3[8],zero,zero,zero,xmm3[9],zero,zero,zero,xmm3[10],zero,zero,zero,xmm3[11],zero,zero,zero,xmm3[12],zero,zero,zero,xmm3[13],zero,zero,zero,xmm3[14],zero,zero,zero,xmm3[15],zero,zero,zero
 ; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512VL-NEXT:    vpsrlvd %zmm3, %zmm0, %zmm3
-; AVX512VL-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VL-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VL-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; AVX512VL-NEXT:    vpslld $8, %zmm0, %zmm2
+; AVX512VL-NEXT:    vpord %zmm2, %zmm0, %zmm0
+; AVX512VL-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
-; AVX512VL-NEXT:    vpsllvd %zmm1, %zmm0, %zmm0
-; AVX512VL-NEXT:    vpord %zmm0, %zmm3, %zmm0
+; AVX512VL-NEXT:    vpsrlvd %zmm1, %zmm0, %zmm0
 ; AVX512VL-NEXT:    vpmovdb %zmm0, %xmm0
 ; AVX512VL-NEXT:    vzeroupper
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: var_funnnel_v16i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512BW-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
-; AVX512BW-NEXT:    vpsrlvw %zmm3, %zmm0, %zmm3
-; AVX512BW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512BW-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512BW-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX512BW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
-; AVX512BW-NEXT:    vpor %ymm0, %ymm3, %ymm0
+; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX512BW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; AVX512BW-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512BW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: var_funnnel_v16i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VLBW-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLBW-NEXT:    vpsrlvw %ymm3, %ymm0, %ymm3
-; AVX512VLBW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLBW-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VLBW-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; AVX512VLBW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX512VLBW-NEXT:    vpsllvw %ymm1, %ymm0, %ymm0
-; AVX512VLBW-NEXT:    vpor %ymm0, %ymm3, %ymm0
+; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; AVX512VLBW-NEXT:    vpsrlvw %ymm1, %ymm0, %ymm0
 ; AVX512VLBW-NEXT:    vpmovwb %ymm0, %xmm0
 ; AVX512VLBW-NEXT:    vzeroupper
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: var_funnnel_v16i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
-; AVX512VBMI2-NEXT:    vpsrlvw %zmm3, %zmm0, %zmm3
-; AVX512VBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VBMI2-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; AVX512VBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX512VBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
-; AVX512VBMI2-NEXT:    vpor %ymm0, %ymm3, %ymm0
+; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX512VBMI2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; AVX512VBMI2-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm0
 ; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VBMI2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX512VBMI2-NEXT:    vzeroupper
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: var_funnnel_v16i8:
 ; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VLVBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLVBMI2-NEXT:    vpsrlvw %ymm3, %ymm0, %ymm3
-; AVX512VLVBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLVBMI2-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VLVBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; AVX512VLVBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX512VLVBMI2-NEXT:    vpsllvw %ymm1, %ymm0, %ymm0
-; AVX512VLVBMI2-NEXT:    vpor %ymm0, %ymm3, %ymm0
+; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX512VLVBMI2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; AVX512VLVBMI2-NEXT:    vpsrlvw %ymm1, %ymm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    vpmovwb %ymm0, %xmm0
 ; AVX512VLVBMI2-NEXT:    vzeroupper
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOP-LABEL: var_funnnel_v16i8:
 ; XOP:       # %bb.0:
 ; XOP-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOP-NEXT:    vpsubb %xmm1, %xmm2, %xmm1

 ; AVX-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; AVX-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vpand %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    retq
 ;
 ; AVX512F-LABEL: splatvar_funnnel_v16i8:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512F-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512F-NEXT:    vpsrld %xmm3, %zmm0, %zmm3
-; AVX512F-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512F-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512F-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512F-NEXT:    vpslld %xmm1, %zmm0, %zmm0
-; AVX512F-NEXT:    vpord %zmm0, %zmm3, %zmm0
-; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
-; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512F-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512F-NEXT:    vpsrlw %xmm1, %xmm2, %xmm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; AVX512F-NEXT:    vpand %xmm3, %xmm2, %xmm2
+; AVX512F-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512F-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0
+; AVX512F-NEXT:    vpand %xmm3, %xmm0, %xmm0
+; AVX512F-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: splatvar_funnnel_v16i8:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VL-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512VL-NEXT:    vpsrld %xmm3, %zmm0, %zmm3
-; AVX512VL-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VL-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VL-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VL-NEXT:    vpslld %xmm1, %zmm0, %zmm0
-; AVX512VL-NEXT:    vpord %zmm0, %zmm3, %zmm0
-; AVX512VL-NEXT:    vpmovdb %zmm0, %xmm0
-; AVX512VL-NEXT:    vzeroupper
+; AVX512VL-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VL-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512VL-NEXT:    vpsrlw %xmm1, %xmm2, %xmm2
+; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; AVX512VL-NEXT:    vpand %xmm3, %xmm2, %xmm2
+; AVX512VL-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512VL-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0
+; AVX512VL-NEXT:    vpand %xmm3, %xmm0, %xmm0
+; AVX512VL-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: splatvar_funnnel_v16i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512BW-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512BW-NEXT:    vpsrlw %xmm3, %ymm0, %ymm3
-; AVX512BW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512BW-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512BW-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512BW-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
-; AVX512BW-NEXT:    vpor %ymm0, %ymm3, %ymm0
-; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
-; AVX512BW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
-; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512BW-NEXT:    vpsrlw %xmm1, %xmm2, %xmm2
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; AVX512BW-NEXT:    vpand %xmm3, %xmm2, %xmm2
+; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512BW-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0
+; AVX512BW-NEXT:    vpand %xmm3, %xmm0, %xmm0
+; AVX512BW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: splatvar_funnnel_v16i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VLBW-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLBW-NEXT:    vpsrlw %xmm3, %ymm0, %ymm3
-; AVX512VLBW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLBW-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VLBW-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VLBW-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
-; AVX512VLBW-NEXT:    vpor %ymm0, %ymm3, %ymm0
+; AVX512VLBW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VLBW-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512VLBW-NEXT:    vpsrlw %xmm1, %xmm2, %xmm2
+; AVX512VLBW-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512VLBW-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0
+; AVX512VLBW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
 ; AVX512VLBW-NEXT:    vpmovwb %ymm0, %xmm0
 ; AVX512VLBW-NEXT:    vzeroupper
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: splatvar_funnnel_v16i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VBMI2-NEXT:    vpsrlw %xmm3, %ymm0, %ymm3
-; AVX512VBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VBMI2-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VBMI2-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
-; AVX512VBMI2-NEXT:    vpor %ymm0, %ymm3, %ymm0
-; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
-; AVX512VBMI2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
-; AVX512VBMI2-NEXT:    vzeroupper
+; AVX512VBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VBMI2-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512VBMI2-NEXT:    vpsrlw %xmm1, %xmm2, %xmm2
+; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; AVX512VBMI2-NEXT:    vpand %xmm3, %xmm2, %xmm2
+; AVX512VBMI2-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512VBMI2-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0
+; AVX512VBMI2-NEXT:    vpand %xmm3, %xmm0, %xmm0
+; AVX512VBMI2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: splatvar_funnnel_v16i8:
 ; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VLVBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLVBMI2-NEXT:    vpsrlw %xmm3, %ymm0, %ymm3
-; AVX512VLVBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLVBMI2-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VLVBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VLVBMI2-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
-; AVX512VLVBMI2-NEXT:    vpor %ymm0, %ymm3, %ymm0
+; AVX512VLVBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VLVBMI2-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512VLVBMI2-NEXT:    vpsrlw %xmm1, %xmm2, %xmm2
+; AVX512VLVBMI2-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512VLVBMI2-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0
+; AVX512VLVBMI2-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    vpmovwb %ymm0, %xmm0
 ; AVX512VLVBMI2-NEXT:    vzeroupper
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: splatvar_funnnel_v16i8:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1

 ; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: constant_funnnel_v16i8:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512F-NEXT:    vpsllvd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
-; AVX512F-NEXT:    vpsrlvd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
+; AVX512F-NEXT:    vpsrlvd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
+; AVX512F-NEXT:    vpsllvd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
 ; AVX512F-NEXT:    vpord %zmm1, %zmm0, %zmm0
 ; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
 ; AVX512F-NEXT:    vzeroupper
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: constant_funnnel_v16i8:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512VL-NEXT:    vpsllvd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
-; AVX512VL-NEXT:    vpsrlvd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
+; AVX512VL-NEXT:    vpsrlvd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
+; AVX512VL-NEXT:    vpsllvd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
 ; AVX512VL-NEXT:    vpord %zmm1, %zmm0, %zmm0
 ; AVX512VL-NEXT:    vpmovdb %zmm0, %xmm0
 ; AVX512VL-NEXT:    vzeroupper
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: constant_funnnel_v16i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,7,6,5,4,3,2,1,0,1,2,3,4,5,6,7]
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,1,2,3,4,5,6,7,0,7,6,5,4,3,2,1]
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512BW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm1
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,0,7,6,5,4,3,2,1]
-; AVX512BW-NEXT:    vpsrlvw %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm1
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,7,6,5,4,3,2,1,0,1,2,3,4,5,6,7]
+; AVX512BW-NEXT:    vpsllvw %zmm2, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpor %ymm1, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512BW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: constant_funnnel_v16i8:
 ; AVX512VLBW:       # %bb.0:
 ; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLBW-NEXT:    vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm1
-; AVX512VLBW-NEXT:    vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
+; AVX512VLBW-NEXT:    vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm1
+; AVX512VLBW-NEXT:    vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
 ; AVX512VLBW-NEXT:    vpor %ymm1, %ymm0, %ymm0
 ; AVX512VLBW-NEXT:    vpmovwb %ymm0, %xmm0
 ; AVX512VLBW-NEXT:    vzeroupper
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: constant_funnnel_v16i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,7,6,5,4,3,2,1,0,1,2,3,4,5,6,7]
+; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,1,2,3,4,5,6,7,0,7,6,5,4,3,2,1]
 ; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm1
-; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,2,3,4,5,6,7,0,7,6,5,4,3,2,1]
-; AVX512VBMI2-NEXT:    vpsrlvw %zmm2, %zmm0, %zmm0
+; AVX512VBMI2-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm1
+; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,7,6,5,4,3,2,1,0,1,2,3,4,5,6,7]
+; AVX512VBMI2-NEXT:    vpsllvw %zmm2, %zmm0, %zmm0
 ; AVX512VBMI2-NEXT:    vpor %ymm1, %ymm0, %ymm0
 ; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VBMI2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX512VBMI2-NEXT:    vzeroupper
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: constant_funnnel_v16i8:
 ; AVX512VLVBMI2:       # %bb.0:
 ; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLVBMI2-NEXT:    vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm1
-; AVX512VLVBMI2-NEXT:    vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
+; AVX512VLVBMI2-NEXT:    vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm1
+; AVX512VLVBMI2-NEXT:    vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    vpor %ymm1, %ymm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    vpmovwb %ymm0, %xmm0
 ; AVX512VLVBMI2-NEXT:    vzeroupper
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOP-LABEL: constant_funnnel_v16i8:
 ; XOP:       # %bb.0:
 ; XOP-NEXT:    vprotb {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0

 ; AVX-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX-NEXT:    vpsllw $4, %xmm0, %xmm0
 ; AVX-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    retq
 ;
 ; AVX512F-LABEL: splatconstant_funnnel_v16i8:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpsrlw $4, %xmm0, %xmm1
-; AVX512F-NEXT:    vpsllw $4, %xmm0, %xmm0
+; AVX512F-NEXT:    vpsllw $4, %xmm0, %xmm1
+; AVX512F-NEXT:    vpsrlw $4, %xmm0, %xmm0
 ; AVX512F-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm1, %zmm0
 ; AVX512F-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
 ; AVX512F-NEXT:    vzeroupper
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: splatconstant_funnnel_v16i8:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpsrlw $4, %xmm0, %xmm1
-; AVX512VL-NEXT:    vpsllw $4, %xmm0, %xmm0
+; AVX512VL-NEXT:    vpsllw $4, %xmm0, %xmm1
+; AVX512VL-NEXT:    vpsrlw $4, %xmm0, %xmm0
 ; AVX512VL-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: splatconstant_funnnel_v16i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpsrlw $4, %xmm0, %xmm1
-; AVX512BW-NEXT:    vpsllw $4, %xmm0, %xmm0
+; AVX512BW-NEXT:    vpsllw $4, %xmm0, %xmm1
+; AVX512BW-NEXT:    vpsrlw $4, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm1, %zmm0
 ; AVX512BW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: splatconstant_funnnel_v16i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vpsrlw $4, %xmm0, %xmm1
-; AVX512VLBW-NEXT:    vpsllw $4, %xmm0, %xmm0
+; AVX512VLBW-NEXT:    vpsllw $4, %xmm0, %xmm1
+; AVX512VLBW-NEXT:    vpsrlw $4, %xmm0, %xmm0
 ; AVX512VLBW-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm0
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: splatconstant_funnnel_v16i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vpsrlw $4, %xmm0, %xmm1
-; AVX512VBMI2-NEXT:    vpsllw $4, %xmm0, %xmm0
+; AVX512VBMI2-NEXT:    vpsllw $4, %xmm0, %xmm1
+; AVX512VBMI2-NEXT:    vpsrlw $4, %xmm0, %xmm0
 ; AVX512VBMI2-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm1, %zmm0
 ; AVX512VBMI2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
 ; AVX512VBMI2-NEXT:    vzeroupper
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: splatconstant_funnnel_v16i8:
 ; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vpsrlw $4, %xmm0, %xmm1
-; AVX512VLVBMI2-NEXT:    vpsllw $4, %xmm0, %xmm0
+; AVX512VLVBMI2-NEXT:    vpsllw $4, %xmm0, %xmm1
+; AVX512VLVBMI2-NEXT:    vpsrlw $4, %xmm0, %xmm0
 ; AVX512VLVBMI2-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm0
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOP-LABEL: splatconstant_funnnel_v16i8:
 ; XOP:       # %bb.0:
 ; XOP-NEXT:    vprotb $4, %xmm0, %xmm0
 ; XOP-NEXT:    retq
 ;

llvm/test/CodeGen/X86/vector-fshr-rot-256.ll

 ; AVX512VL-NEXT:    vpsllw $7, %ymm0, %ymm3
 ; AVX512VL-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm2, %ymm3
 ; AVX512VL-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
 ; AVX512VL-NEXT:    vpblendvb %ymm1, %ymm3, %ymm0, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: var_funnnel_v32i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm3
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero,ymm3[16],zero,ymm3[17],zero,ymm3[18],zero,ymm3[19],zero,ymm3[20],zero,ymm3[21],zero,ymm3[22],zero,ymm3[23],zero,ymm3[24],zero,ymm3[25],zero,ymm3[26],zero,ymm3[27],zero,ymm3[28],zero,ymm3[29],zero,ymm3[30],zero,ymm3[31],zero
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512BW-NEXT:    vpsrlvw %zmm3, %zmm0, %zmm3
-; AVX512BW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512BW-NEXT:    vpsubb %ymm1, %ymm4, %ymm1
-; AVX512BW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512BW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512BW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
-; AVX512BW-NEXT:    vporq %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
+; AVX512BW-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30,32,32,34,34,36,36,38,38,40,40,42,42,44,44,46,46,48,48,50,50,52,52,54,54,56,56,58,58,60,60,62,62]
+; AVX512BW-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: var_funnnel_v32i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VLBW-NEXT:    vpand %ymm2, %ymm1, %ymm3
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero,ymm3[16],zero,ymm3[17],zero,ymm3[18],zero,ymm3[19],zero,ymm3[20],zero,ymm3[21],zero,ymm3[22],zero,ymm3[23],zero,ymm3[24],zero,ymm3[25],zero,ymm3[26],zero,ymm3[27],zero,ymm3[28],zero,ymm3[29],zero,ymm3[30],zero,ymm3[31],zero
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLBW-NEXT:    vpsrlvw %zmm3, %zmm0, %zmm3
-; AVX512VLBW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLBW-NEXT:    vpsubb %ymm1, %ymm4, %ymm1
-; AVX512VLBW-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512VLBW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
 ; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512VLBW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
-; AVX512VLBW-NEXT:    vporq %zmm0, %zmm3, %zmm0
+; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30,32,32,34,34,36,36,38,38,40,40,42,42,44,44,46,46,48,48,50,50,52,52,54,54,56,56,58,58,60,60,62,62]
+; AVX512VLBW-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm0
 ; AVX512VLBW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: var_funnnel_v32i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} ymm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VBMI2-NEXT:    vpand %ymm2, %ymm1, %ymm3
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero,ymm3[16],zero,ymm3[17],zero,ymm3[18],zero,ymm3[19],zero,ymm3[20],zero,ymm3[21],zero,ymm3[22],zero,ymm3[23],zero,ymm3[24],zero,ymm3[25],zero,ymm3[26],zero,ymm3[27],zero,ymm3[28],zero,ymm3[29],zero,ymm3[30],zero,ymm3[31],zero
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VBMI2-NEXT:    vpsrlvw %zmm3, %zmm0, %zmm3
-; AVX512VBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VBMI2-NEXT:    vpsubb %ymm1, %ymm4, %ymm1
-; AVX512VBMI2-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512VBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
 ; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512VBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
-; AVX512VBMI2-NEXT:    vporq %zmm0, %zmm3, %zmm0
+; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
+; AVX512VBMI2-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30,32,32,34,34,36,36,38,38,40,40,42,42,44,44,46,46,48,48,50,50,52,52,54,54,56,56,58,58,60,60,62,62]
+; AVX512VBMI2-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm0
 ; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: var_funnnel_v32i8:
 ; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vmovdqa {{.*#+}} ymm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
-; AVX512VLVBMI2-NEXT:    vpand %ymm2, %ymm1, %ymm3
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero,ymm3[16],zero,ymm3[17],zero,ymm3[18],zero,ymm3[19],zero,ymm3[20],zero,ymm3[21],zero,ymm3[22],zero,ymm3[23],zero,ymm3[24],zero,ymm3[25],zero,ymm3[26],zero,ymm3[27],zero,ymm3[28],zero,ymm3[29],zero,ymm3[30],zero,ymm3[31],zero
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLVBMI2-NEXT:    vpsrlvw %zmm3, %zmm0, %zmm3
-; AVX512VLVBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLVBMI2-NEXT:    vpsubb %ymm1, %ymm4, %ymm1
-; AVX512VLVBMI2-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512VLVBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
 ; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512VLVBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
-; AVX512VLVBMI2-NEXT:    vporq %zmm0, %zmm3, %zmm0
+; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
+; AVX512VLVBMI2-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30,32,32,34,34,36,36,38,38,40,40,42,42,44,44,46,46,48,48,50,50,52,52,54,54,56,56,58,58,60,60,62,62]
+; AVX512VLVBMI2-NEXT:    vpsrlvw %zmm1, %zmm0, %zmm0
 ; AVX512VLVBMI2-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: var_funnnel_v32i8:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; XOPAVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
 ; XOPAVX1-NEXT:    vpsubb %xmm2, %xmm3, %xmm2

 ; AVX512VL-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
 ; AVX512VL-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0
 ; AVX512VL-NEXT:    vpand %ymm3, %ymm0, %ymm0
 ; AVX512VL-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: splatvar_funnnel_v32i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512BW-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512BW-NEXT:    vpsrlw %xmm3, %zmm0, %zmm3
-; AVX512BW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512BW-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512BW-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512BW-NEXT:    vpsllw %xmm1, %zmm0, %zmm0
-; AVX512BW-NEXT:    vporq %zmm0, %zmm3, %zmm0
-; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
+; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX512BW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512BW-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-NEXT:    vpsrlw %xmm1, %ymm2, %ymm2
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512BW-NEXT:    vpand %ymm3, %ymm2, %ymm2
+; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
+; AVX512BW-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0
+; AVX512BW-NEXT:    vpand %ymm3, %ymm0, %ymm0
+; AVX512BW-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: splatvar_funnnel_v32i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VLBW-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLBW-NEXT:    vpsrlw %xmm3, %zmm0, %zmm3
-; AVX512VLBW-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLBW-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VLBW-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VLBW-NEXT:    vpsllw %xmm1, %zmm0, %zmm0
-; AVX512VLBW-NEXT:    vporq %zmm0, %zmm3, %zmm0
-; AVX512VLBW-NEXT:    vpmovwb %zmm0, %ymm0
+; AVX512VLBW-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX512VLBW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VLBW-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VLBW-NEXT:    vpsrlw %xmm1, %ymm2, %ymm2
+; AVX512VLBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512VLBW-NEXT:    vpand %ymm3, %ymm2, %ymm2
+; AVX512VLBW-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
+; AVX512VLBW-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0
+; AVX512VLBW-NEXT:    vpand %ymm3, %ymm0, %ymm0
+; AVX512VLBW-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: splatvar_funnnel_v32i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VBMI2-NEXT:    vpsrlw %xmm3, %zmm0, %zmm3
-; AVX512VBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VBMI2-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VBMI2-NEXT:    vpsllw %xmm1, %zmm0, %zmm0
-; AVX512VBMI2-NEXT:    vporq %zmm0, %zmm3, %zmm0
-; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
+; AVX512VBMI2-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX512VBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VBMI2-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VBMI2-NEXT:    vpsrlw %xmm1, %ymm2, %ymm2
+; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512VBMI2-NEXT:    vpand %ymm3, %ymm2, %ymm2
+; AVX512VBMI2-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
+; AVX512VBMI2-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0
+; AVX512VBMI2-NEXT:    vpand %ymm3, %ymm0, %ymm0
+; AVX512VBMI2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: splatvar_funnnel_v32i8:
 ; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [7,7]
-; AVX512VLVBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm3
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLVBMI2-NEXT:    vpsrlw %xmm3, %zmm0, %zmm3
-; AVX512VLVBMI2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; AVX512VLVBMI2-NEXT:    vpsubb %xmm1, %xmm4, %xmm1
-; AVX512VLVBMI2-NEXT:    vpand %xmm2, %xmm1, %xmm1
-; AVX512VLVBMI2-NEXT:    vpsllw %xmm1, %zmm0, %zmm0
-; AVX512VLVBMI2-NEXT:    vporq %zmm0, %zmm3, %zmm0
-; AVX512VLVBMI2-NEXT:    vpmovwb %zmm0, %ymm0
+; AVX512VLVBMI2-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX512VLVBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VLVBMI2-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512VLVBMI2-NEXT:    vpsrlw %xmm1, %ymm2, %ymm2
+; AVX512VLVBMI2-NEXT:    vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512VLVBMI2-NEXT:    vpand %ymm3, %ymm2, %ymm2
+; AVX512VLVBMI2-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
+; AVX512VLVBMI2-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0
+; AVX512VLVBMI2-NEXT:    vpand %ymm3, %ymm0, %ymm0
+; AVX512VLVBMI2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: splatvar_funnnel_v32i8:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpsubb %xmm1, %xmm2, %xmm1
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2

 ; AVX512VL-NEXT:    vpand %ymm3, %ymm0, %ymm0
 ; AVX512VL-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; AVX512VL-NEXT:    vpor %ymm1, %ymm0, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: constant_funnnel_v32i8:
 ; AVX512BW:       # %bb.0:
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512BW-NEXT:    vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
-; AVX512BW-NEXT:    vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
+; AVX512BW-NEXT:    vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
 ; AVX512BW-NEXT:    vporq %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: constant_funnnel_v32i8:
 ; AVX512VLBW:       # %bb.0:
 ; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLBW-NEXT:    vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
-; AVX512VLBW-NEXT:    vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
+; AVX512VLBW-NEXT:    vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
+; AVX512VLBW-NEXT:    vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
 ; AVX512VLBW-NEXT:    vporq %zmm1, %zmm0, %zmm0
 ; AVX512VLBW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: constant_funnnel_v32i8:
 ; AVX512VBMI2:       # %bb.0:
 ; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VBMI2-NEXT:    vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
-; AVX512VBMI2-NEXT:    vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
+; AVX512VBMI2-NEXT:    vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
+; AVX512VBMI2-NEXT:    vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
 ; AVX512VBMI2-NEXT:    vporq %zmm1, %zmm0, %zmm0
 ; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: constant_funnnel_v32i8:
 ; AVX512VLVBMI2:       # %bb.0:
 ; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLVBMI2-NEXT:    vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
-; AVX512VLVBMI2-NEXT:    vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
+; AVX512VLVBMI2-NEXT:    vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
+; AVX512VLVBMI2-NEXT:    vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
 ; AVX512VLVBMI2-NEXT:    vporq %zmm1, %zmm0, %zmm0
 ; AVX512VLVBMI2-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: constant_funnnel_v32i8:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; XOPAVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,7,6,5,4,3,2,1,0,1,2,3,4,5,6,7]

 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vpsllw $4, %ymm0, %ymm1
 ; AVX512VL-NEXT:    vpsrlw $4, %ymm0, %ymm0
 ; AVX512VL-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm1, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: splatconstant_funnnel_v32i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpsrlw $4, %ymm0, %ymm1
-; AVX512BW-NEXT:    vpsllw $4, %ymm0, %ymm0
+; AVX512BW-NEXT:    vpsllw $4, %ymm0, %ymm1
+; AVX512BW-NEXT:    vpsrlw $4, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm1, %zmm0
 ; AVX512BW-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: splatconstant_funnnel_v32i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vpsrlw $4, %ymm0, %ymm1
-; AVX512VLBW-NEXT:    vpsllw $4, %ymm0, %ymm0
+; AVX512VLBW-NEXT:    vpsllw $4, %ymm0, %ymm1
+; AVX512VLBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
 ; AVX512VLBW-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm1, %ymm0
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: splatconstant_funnnel_v32i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vpsrlw $4, %ymm0, %ymm1
-; AVX512VBMI2-NEXT:    vpsllw $4, %ymm0, %ymm0
+; AVX512VBMI2-NEXT:    vpsllw $4, %ymm0, %ymm1
+; AVX512VBMI2-NEXT:    vpsrlw $4, %ymm0, %ymm0
 ; AVX512VBMI2-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm1, %zmm0
 ; AVX512VBMI2-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: splatconstant_funnnel_v32i8:
 ; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vpsrlw $4, %ymm0, %ymm1
-; AVX512VLVBMI2-NEXT:    vpsllw $4, %ymm0, %ymm0
+; AVX512VLVBMI2-NEXT:    vpsllw $4, %ymm0, %ymm1
+; AVX512VLVBMI2-NEXT:    vpsrlw $4, %ymm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm1, %ymm0
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: splatconstant_funnnel_v32i8:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vprotb $4, %xmm0, %xmm1
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; XOPAVX1-NEXT:    vprotb $4, %xmm0, %xmm0

llvm/test/CodeGen/X86/vector-rotate-128.ll

 ; AVX-NEXT:    vpaddb %xmm0, %xmm0, %xmm3
 ; AVX-NEXT:    vpor %xmm2, %xmm3, %xmm2
 ; AVX-NEXT:    vpaddb %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vpblendvb %xmm1, %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    retq
 ;
 ; AVX512F-LABEL: var_rotate_v16i8:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512F-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
-; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
 ; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512F-NEXT:    vpsllvd %zmm1, %zmm0, %zmm1
-; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero,xmm2[4],zero,zero,zero,xmm2[5],zero,zero,zero,xmm2[6],zero,zero,zero,xmm2[7],zero,zero,zero,xmm2[8],zero,zero,zero,xmm2[9],zero,zero,zero,xmm2[10],zero,zero,zero,xmm2[11],zero,zero,zero,xmm2[12],zero,zero,zero,xmm2[13],zero,zero,zero,xmm2[14],zero,zero,zero,xmm2[15],zero,zero,zero
-; AVX512F-NEXT:    vpsrlvd %zmm2, %zmm0, %zmm0
-; AVX512F-NEXT:    vpord %zmm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vpslld $8, %zmm0, %zmm2
+; AVX512F-NEXT:    vpord %zmm2, %zmm0, %zmm0
+; AVX512F-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
+; AVX512F-NEXT:    vpsllvd %zmm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpsrld $8, %zmm0, %zmm0
 ; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
 ; AVX512F-NEXT:    vzeroupper
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: var_rotate_v16i8:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VL-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
-; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
 ; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512VL-NEXT:    vpsllvd %zmm1, %zmm0, %zmm1
-; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero,xmm2[4],zero,zero,zero,xmm2[5],zero,zero,zero,xmm2[6],zero,zero,zero,xmm2[7],zero,zero,zero,xmm2[8],zero,zero,zero,xmm2[9],zero,zero,zero,xmm2[10],zero,zero,zero,xmm2[11],zero,zero,zero,xmm2[12],zero,zero,zero,xmm2[13],zero,zero,zero,xmm2[14],zero,zero,zero,xmm2[15],zero,zero,zero
-; AVX512VL-NEXT:    vpsrlvd %zmm2, %zmm0, %zmm0
-; AVX512VL-NEXT:    vpord %zmm0, %zmm1, %zmm0
+; AVX512VL-NEXT:    vpslld $8, %zmm0, %zmm2
+; AVX512VL-NEXT:    vpord %zmm2, %zmm0, %zmm0
+; AVX512VL-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero,xmm1[4],zero,zero,zero,xmm1[5],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[7],zero,zero,zero,xmm1[8],zero,zero,zero,xmm1[9],zero,zero,zero,xmm1[10],zero,zero,zero,xmm1[11],zero,zero,zero,xmm1[12],zero,zero,zero,xmm1[13],zero,zero,zero,xmm1[14],zero,zero,zero,xmm1[15],zero,zero,zero
+; AVX512VL-NEXT:    vpsllvd %zmm1, %zmm0, %zmm0
+; AVX512VL-NEXT:    vpsrld $8, %zmm0, %zmm0
 ; AVX512VL-NEXT:    vpmovdb %zmm0, %xmm0
 ; AVX512VL-NEXT:    vzeroupper
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: var_rotate_v16i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512BW-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
+; AVX512BW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512BW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm1
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero
-; AVX512BW-NEXT:    vpsrlvw %zmm2, %zmm0, %zmm0
-; AVX512BW-NEXT:    vpor %ymm0, %ymm1, %ymm0
+; AVX512BW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; AVX512BW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512BW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: var_rotate_v16i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VLBW-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
+; AVX512VLBW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
 ; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLBW-NEXT:    vpsllvw %ymm1, %ymm0, %ymm1
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero
-; AVX512VLBW-NEXT:    vpsrlvw %ymm2, %ymm0, %ymm0
-; AVX512VLBW-NEXT:    vpor %ymm0, %ymm1, %ymm0
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; AVX512VLBW-NEXT:    vpsllvw %ymm1, %ymm0, %ymm0
+; AVX512VLBW-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512VLBW-NEXT:    vpmovwb %ymm0, %xmm0
 ; AVX512VLBW-NEXT:    vzeroupper
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: var_rotate_v16i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VBMI2-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
+; AVX512VBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
 ; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm1
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero
-; AVX512VBMI2-NEXT:    vpsrlvw %zmm2, %zmm0, %zmm0
-; AVX512VBMI2-NEXT:    vpor %ymm0, %ymm1, %ymm0
+; AVX512VBMI2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; AVX512VBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
+; AVX512VBMI2-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VBMI2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX512VBMI2-NEXT:    vzeroupper
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: var_rotate_v16i8:
 ; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VLVBMI2-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
+; AVX512VLVBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
 ; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLVBMI2-NEXT:    vpsllvw %ymm1, %ymm0, %ymm1
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero
-; AVX512VLVBMI2-NEXT:    vpsrlvw %ymm2, %ymm0, %ymm0
-; AVX512VLVBMI2-NEXT:    vpor %ymm0, %ymm1, %ymm0
+; AVX512VLVBMI2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30]
+; AVX512VLVBMI2-NEXT:    vpsllvw %ymm1, %ymm0, %ymm0
+; AVX512VLVBMI2-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    vpmovwb %ymm0, %xmm0
 ; AVX512VLVBMI2-NEXT:    vzeroupper
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOP-LABEL: var_rotate_v16i8:
 ; XOP:       # %bb.0:
 ; XOP-NEXT:    vprotb %xmm1, %xmm0, %xmm0
 ; XOP-NEXT:    retq

 ; AVX-NEXT:    vpsllw %xmm1, %xmm2, %xmm2
 ; AVX-NEXT:    vpsrlw $8, %xmm2, %xmm2
 ; AVX-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; AVX-NEXT:    vpsllw %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vpsrlw $8, %xmm0, %xmm0
 ; AVX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    retq
 ;
-; AVX512F-LABEL: splatvar_rotate_v16i8:
-; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512F-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
-; AVX512F-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512F-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512F-NEXT:    vpslld %xmm1, %zmm0, %zmm1
-; AVX512F-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512F-NEXT:    vpsrld %xmm2, %zmm0, %zmm0
-; AVX512F-NEXT:    vpord %zmm0, %zmm1, %zmm0
-; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
-; AVX512F-NEXT:    vzeroupper
-; AVX512F-NEXT:    retq
-;
-; AVX512VL-LABEL: splatvar_rotate_v16i8:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VL-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
-; AVX512VL-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
-; AVX512VL-NEXT:    vpslld %xmm1, %zmm0, %zmm1
-; AVX512VL-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VL-NEXT:    vpsrld %xmm2, %zmm0, %zmm0
-; AVX512VL-NEXT:    vpord %zmm0, %zmm1, %zmm0
-; AVX512VL-NEXT:    vpmovdb %zmm0, %xmm0
-; AVX512VL-NEXT:    vzeroupper
-; AVX512VL-NEXT:    retq
-;
-; AVX512BW-LABEL: splatvar_rotate_v16i8:
-; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512BW-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
-; AVX512BW-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512BW-NEXT:    vpsllw %xmm1, %ymm0, %ymm1
-; AVX512BW-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-NEXT:    vpsrlw %xmm2, %ymm0, %ymm0
-; AVX512BW-NEXT:    vpor %ymm0, %ymm1, %ymm0
-; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
-; AVX512BW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
-; AVX512BW-NEXT:    vzeroupper
-; AVX512BW-NEXT:    retq
-;
-; AVX512VLBW-LABEL: splatvar_rotate_v16i8:
-; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VLBW-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
-; AVX512VLBW-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLBW-NEXT:    vpsllw %xmm1, %ymm0, %ymm1
-; AVX512VLBW-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLBW-NEXT:    vpsrlw %xmm2, %ymm0, %ymm0
-; AVX512VLBW-NEXT:    vpor %ymm0, %ymm1, %ymm0
-; AVX512VLBW-NEXT:    vpmovwb %ymm0, %xmm0
-; AVX512VLBW-NEXT:    vzeroupper
-; AVX512VLBW-NEXT:    retq
-;
-; AVX512VBMI2-LABEL: splatvar_rotate_v16i8:
-; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VBMI2-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
-; AVX512VBMI2-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VBMI2-NEXT:    vpsllw %xmm1, %ymm0, %ymm1
-; AVX512VBMI2-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VBMI2-NEXT:    vpsrlw %xmm2, %ymm0, %ymm0
-; AVX512VBMI2-NEXT:    vpor %ymm0, %ymm1, %ymm0
-; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
-; AVX512VBMI2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
-; AVX512VBMI2-NEXT:    vzeroupper
-; AVX512VBMI2-NEXT:    retq
-;
-; AVX512VLVBMI2-LABEL: splatvar_rotate_v16i8:
-; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vmovdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VLVBMI2-NEXT:    vpsubb %xmm1, %xmm2, %xmm2
-; AVX512VLVBMI2-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512VLVBMI2-NEXT:    vpsllw %xmm1, %ymm0, %ymm1
-; AVX512VLVBMI2-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLVBMI2-NEXT:    vpsrlw %xmm2, %ymm0, %ymm0
-; AVX512VLVBMI2-NEXT:    vpor %ymm0, %ymm1, %ymm0
-; AVX512VLVBMI2-NEXT:    vpmovwb %ymm0, %xmm0
-; AVX512VLVBMI2-NEXT:    vzeroupper
-; AVX512VLVBMI2-NEXT:    retq
+; AVX512-LABEL: splatvar_rotate_v16i8:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512-NEXT:    vpsllw %xmm1, %xmm2, %xmm2
+; AVX512-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX512-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512-NEXT:    vpsllw %xmm1, %xmm0, %xmm0
+; AVX512-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX512-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX512-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: splatvar_rotate_v16i8:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vprotb %xmm1, %xmm0, %xmm0
 ; XOPAVX1-NEXT:    retq
 ;

llvm/test/CodeGen/X86/vector-rotate-256.ll

 ; AVX512VL-NEXT:    vpaddb %ymm0, %ymm0, %ymm3
 ; AVX512VL-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm3
 ; AVX512VL-NEXT:    vpaddb %ymm1, %ymm1, %ymm1
 ; AVX512VL-NEXT:    vpblendvb %ymm1, %ymm3, %ymm0, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: var_rotate_v32i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512BW-NEXT:    vpsubb %ymm1, %ymm2, %ymm2
+; AVX512BW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512BW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm1
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero,ymm2[16],zero,ymm2[17],zero,ymm2[18],zero,ymm2[19],zero,ymm2[20],zero,ymm2[21],zero,ymm2[22],zero,ymm2[23],zero,ymm2[24],zero,ymm2[25],zero,ymm2[26],zero,ymm2[27],zero,ymm2[28],zero,ymm2[29],zero,ymm2[30],zero,ymm2[31],zero
-; AVX512BW-NEXT:    vpsrlvw %zmm2, %zmm0, %zmm0
-; AVX512BW-NEXT:    vporq %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30,32,32,34,34,36,36,38,38,40,40,42,42,44,44,46,46,48,48,50,50,52,52,54,54,56,56,58,58,60,60,62,62]
+; AVX512BW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: var_rotate_v32i8:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VLBW-NEXT:    vpsubb %ymm1, %ymm2, %ymm2
+; AVX512VLBW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
 ; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
 ; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLBW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm1
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero,ymm2[16],zero,ymm2[17],zero,ymm2[18],zero,ymm2[19],zero,ymm2[20],zero,ymm2[21],zero,ymm2[22],zero,ymm2[23],zero,ymm2[24],zero,ymm2[25],zero,ymm2[26],zero,ymm2[27],zero,ymm2[28],zero,ymm2[29],zero,ymm2[30],zero,ymm2[31],zero
-; AVX512VLBW-NEXT:    vpsrlvw %zmm2, %zmm0, %zmm0
-; AVX512VLBW-NEXT:    vporq %zmm0, %zmm1, %zmm0
+; AVX512VLBW-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30,32,32,34,34,36,36,38,38,40,40,42,42,44,44,46,46,48,48,50,50,52,52,54,54,56,56,58,58,60,60,62,62]
+; AVX512VLBW-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
+; AVX512VLBW-NEXT:    vpsrlw $8, %zmm0, %zmm0
 ; AVX512VLBW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VBMI2-LABEL: var_rotate_v32i8:
 ; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} ymm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VBMI2-NEXT:    vpsubb %ymm1, %ymm2, %ymm2
+; AVX512VBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
 ; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
 ; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm1
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero,ymm2[16],zero,ymm2[17],zero,ymm2[18],zero,ymm2[19],zero,ymm2[20],zero,ymm2[21],zero,ymm2[22],zero,ymm2[23],zero,ymm2[24],zero,ymm2[25],zero,ymm2[26],zero,ymm2[27],zero,ymm2[28],zero,ymm2[29],zero,ymm2[30],zero,ymm2[31],zero
-; AVX512VBMI2-NEXT:    vpsrlvw %zmm2, %zmm0, %zmm0
-; AVX512VBMI2-NEXT:    vporq %zmm0, %zmm1, %zmm0
+; AVX512VBMI2-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30,32,32,34,34,36,36,38,38,40,40,42,42,44,44,46,46,48,48,50,50,52,52,54,54,56,56,58,58,60,60,62,62]
+; AVX512VBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
+; AVX512VBMI2-NEXT:    vpsrlw $8, %zmm0, %zmm0
 ; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VBMI2-NEXT:    retq
 ;
 ; AVX512VLVBMI2-LABEL: var_rotate_v32i8:
 ; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vmovdqa {{.*#+}} ymm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VLVBMI2-NEXT:    vpsubb %ymm1, %ymm2, %ymm2
+; AVX512VLVBMI2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
 ; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
 ; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLVBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm1
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero,ymm2[16],zero,ymm2[17],zero,ymm2[18],zero,ymm2[19],zero,ymm2[20],zero,ymm2[21],zero,ymm2[22],zero,ymm2[23],zero,ymm2[24],zero,ymm2[25],zero,ymm2[26],zero,ymm2[27],zero,ymm2[28],zero,ymm2[29],zero,ymm2[30],zero,ymm2[31],zero
-; AVX512VLVBMI2-NEXT:    vpsrlvw %zmm2, %zmm0, %zmm0
-; AVX512VLVBMI2-NEXT:    vporq %zmm0, %zmm1, %zmm0
+; AVX512VLVBMI2-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14,16,16,18,18,20,20,22,22,24,24,26,26,28,28,30,30,32,32,34,34,36,36,38,38,40,40,42,42,44,44,46,46,48,48,50,50,52,52,54,54,56,56,58,58,60,60,62,62]
+; AVX512VLVBMI2-NEXT:    vpsllvw %zmm1, %zmm0, %zmm0
+; AVX512VLVBMI2-NEXT:    vpsrlw $8, %zmm0, %zmm0
 ; AVX512VLVBMI2-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512VLVBMI2-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: var_rotate_v32i8:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; XOPAVX1-NEXT:    vprotb %xmm2, %xmm3, %xmm2

 ; AVX2-NEXT:    vpsllw %xmm1, %ymm2, %ymm2
 ; AVX2-NEXT:    vpsrlw $8, %ymm2, %ymm2
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
 ; AVX2-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
 ;
-; AVX512F-LABEL: splatvar_rotate_v32i8:
-; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX512F-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX512F-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512F-NEXT:    vpsllw %xmm1, %ymm2, %ymm2
-; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
-; AVX512F-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX512F-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
-; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
-; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT:    retq
-;
-; AVX512VL-LABEL: splatvar_rotate_v32i8:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX512VL-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX512VL-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VL-NEXT:    vpsllw %xmm1, %ymm2, %ymm2
-; AVX512VL-NEXT:    vpsrlw $8, %ymm2, %ymm2
-; AVX512VL-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX512VL-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpsrlw $8, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
-; AVX512VL-NEXT:    retq
-;
-; AVX512BW-LABEL: splatvar_rotate_v32i8:
-; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512BW-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-NEXT:    vpsllw %xmm2, %zmm0, %zmm2
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512BW-NEXT:    vpsubb %xmm1, %xmm3, %xmm1
-; AVX512BW-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-NEXT:    vpsrlw %xmm1, %zmm0, %zmm0
-; AVX512BW-NEXT:    vporq %zmm0, %zmm2, %zmm0
-; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
-; AVX512BW-NEXT:    retq
-;
-; AVX512VLBW-LABEL: splatvar_rotate_v32i8:
-; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLBW-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLBW-NEXT:    vpsllw %xmm2, %zmm0, %zmm2
-; AVX512VLBW-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VLBW-NEXT:    vpsubb %xmm1, %xmm3, %xmm1
-; AVX512VLBW-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLBW-NEXT:    vpsrlw %xmm1, %zmm0, %zmm0
-; AVX512VLBW-NEXT:    vporq %zmm0, %zmm2, %zmm0
-; AVX512VLBW-NEXT:    vpmovwb %zmm0, %ymm0
-; AVX512VLBW-NEXT:    retq
-;
-; AVX512VBMI2-LABEL: splatvar_rotate_v32i8:
-; AVX512VBMI2:       # %bb.0:
-; AVX512VBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VBMI2-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VBMI2-NEXT:    vpsllw %xmm2, %zmm0, %zmm2
-; AVX512VBMI2-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VBMI2-NEXT:    vpsubb %xmm1, %xmm3, %xmm1
-; AVX512VBMI2-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VBMI2-NEXT:    vpsrlw %xmm1, %zmm0, %zmm0
-; AVX512VBMI2-NEXT:    vporq %zmm0, %zmm2, %zmm0
-; AVX512VBMI2-NEXT:    vpmovwb %zmm0, %ymm0
-; AVX512VBMI2-NEXT:    retq
-;
-; AVX512VLVBMI2-LABEL: splatvar_rotate_v32i8:
-; AVX512VLVBMI2:       # %bb.0:
-; AVX512VLVBMI2-NEXT:    vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512VLVBMI2-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLVBMI2-NEXT:    vpsllw %xmm2, %zmm0, %zmm2
-; AVX512VLVBMI2-NEXT:    vmovdqa {{.*#+}} xmm3 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
-; AVX512VLVBMI2-NEXT:    vpsubb %xmm1, %xmm3, %xmm1
-; AVX512VLVBMI2-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX512VLVBMI2-NEXT:    vpsrlw %xmm1, %zmm0, %zmm0
-; AVX512VLVBMI2-NEXT:    vporq %zmm0, %zmm2, %zmm0
-; AVX512VLVBMI2-NEXT:    vpmovwb %zmm0, %ymm0
-; AVX512VLVBMI2-NEXT:    retq
+; AVX512-LABEL: splatvar_rotate_v32i8:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX512-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; AVX512-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX512-NEXT:    vpsllw %xmm1, %ymm2, %ymm2
+; AVX512-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
+; AVX512-NEXT:    vpsllw %xmm1, %ymm0, %ymm0
+; AVX512-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: splatvar_rotate_v32i8:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; XOPAVX1-NEXT:    vprotb %xmm1, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vprotb %xmm1, %xmm0, %xmm0