Index: lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- lib/Target/X86/X86ISelLowering.cpp
+++ lib/Target/X86/X86ISelLowering.cpp
@@ -15053,6 +15053,42 @@
   }
 }
 
+// Determine if this shuffle can be implemented with a KSHIFT instruction.
+// Returns the shift amount if possible or -1 if not. This is a simplified
+// version of matchVectorShuffleAsShift.
+static int match1BitShuffleAsKSHIFT(unsigned &Opcode, ArrayRef<int> Mask,
+                                    int MaskOffset, const APInt &Zeroable) {
+  int Size = Mask.size();
+
+  auto CheckZeros = [&](int Shift, bool Left) {
+    for (int j = 0; j < Shift; ++j)
+      if (!Zeroable[j + (Left ? 0 : (Size - Shift))])
+        return false;
+
+    return true;
+  };
+
+  auto MatchShift = [&](int Shift, bool Left) {
+    unsigned Pos = Left ? Shift : 0;
+    unsigned Low = Left ? 0 : Shift;
+    unsigned Len = Size - Shift;
+    if (!isSequentialOrUndefInRange(Mask, Pos, Len, Low + MaskOffset))
+      return false;
+
+    return true;
+  };
+
+  for (int Shift = 1; Shift != Size; ++Shift)
+    for (bool Left : {true, false})
+      if (CheckZeros(Shift, Left) && MatchShift(Shift, Left)) {
+        Opcode = Left ? X86ISD::KSHIFTL : X86ISD::KSHIFTR;
+        return Shift;
+      }
+
+  return -1;
+}
+
+
 // Lower vXi1 vector shuffles.
 // There is no a dedicated instruction on AVX-512 that shuffles the masks.
 // The only way to shuffle bits is to sign-extend the mask vector to SIMD
@@ -15062,6 +15098,9 @@
                                       const APInt &Zeroable,
                                       const X86Subtarget &Subtarget,
                                       SelectionDAG &DAG) {
+  assert(Subtarget.hasAVX512() &&
+         "Cannot lower 512-bit vectors w/o basic ISA!");
+
   unsigned NumElts = Mask.size();
 
   // Try to recognize shuffles that are just padding a subvector with zeros.
@@ -15088,9 +15127,23 @@
                        Extract, DAG.getIntPtrConstant(0, DL));
   }
 
+  // Try to match KSHIFTs.
+  // TODO: Support narrower than legal shifts by widening and extracting.
+  if (NumElts >= 16 || (Subtarget.hasDQI() && NumElts == 8)) {
+    unsigned Opcode;
+    int ShiftAmt = match1BitShuffleAsKSHIFT(Opcode, Mask, 0, Zeroable);
+    if (ShiftAmt >= 0)
+      return DAG.getNode(Opcode, DL, VT, V1,
+                         DAG.getConstant(ShiftAmt, DL, MVT::i8));
+
+    // V1 didn't match, try V2.
+    ShiftAmt = match1BitShuffleAsKSHIFT(Opcode, Mask, NumElts, Zeroable);
+    if (ShiftAmt >= 0)
+      return DAG.getNode(Opcode, DL, VT, V2,
+                         DAG.getConstant(ShiftAmt, DL, MVT::i8));
+  }
+
 
-  assert(Subtarget.hasAVX512() &&
-         "Cannot lower 512-bit vectors w/o basic ISA!");
   MVT ExtVT;
   switch (VT.SimpleTy) {
   default:
Index: test/CodeGen/X86/avx512-skx-insert-subvec.ll
===================================================================
--- test/CodeGen/X86/avx512-skx-insert-subvec.ll
+++ test/CodeGen/X86/avx512-skx-insert-subvec.ll
@@ -30,11 +30,8 @@
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpsllq $63, %xmm0, %xmm0
 ; CHECK-NEXT:    vpmovq2m %xmm0, %k0
-; CHECK-NEXT:    vpmovm2d %k0, %ymm0
-; CHECK-NEXT:    vperm2i128 {{.*#+}} ymm0 = zero,zero,ymm0[0,1]
-; CHECK-NEXT:    vpmovd2m %ymm0, %k0
+; CHECK-NEXT:    kshiftlb $4, %k0, %k0
 ; CHECK-NEXT:    vpmovm2w %k0, %xmm0
-; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %res = shufflevector <2 x i1> %a, <2 x i1> zeroinitializer, <8 x i32> <i32 3, i32 3, i32 undef, i32 undef, i32 0, i32 1, i32 undef, i32 undef>
   ret <8 x i1> %res
Index: test/CodeGen/X86/kshift.ll
===================================================================
--- /dev/null
+++ test/CodeGen/X86/kshift.ll
@@ -0,0 +1,725 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=KNL
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx512vl,avx512dq,avx512bw | FileCheck %s --check-prefix=CHECK --check-prefix=SKX
+
+define i8 @kshiftl_v8i1_1(<8 x i64> %x, <8 x i64> %y) {
+; KNL-LABEL: kshiftl_v8i1_1:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vptestnmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; KNL-NEXT:    movb $-2, %al
+; KNL-NEXT:    kmovw %eax, %k1
+; KNL-NEXT:    vpexpandq %zmm0, %zmm0 {%k1} {z}
+; KNL-NEXT:    vptestmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    # kill: def $al killed $al killed $eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftl_v8i1_1:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmq %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftlb $1, %k0, %k1
+; SKX-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    # kill: def $al killed $al killed $eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <8 x i64> %x, zeroinitializer
+  %b = shufflevector <8 x i1> %a, <8 x i1> zeroinitializer, <8 x i32> <i32 8, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6>
+  %c = icmp eq <8 x i64> %y, zeroinitializer
+  %d = and <8 x i1> %b, %c
+  %e = bitcast <8 x i1> %d to i8
+  ret i8 %e
+}
+
+define i16 @kshiftl_v16i1_1(<16 x i32> %x, <16 x i32> %y) {
+; KNL-LABEL: kshiftl_v16i1_1:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vptestnmd %zmm0, %zmm0, %k0
+; KNL-NEXT:    kshiftlw $1, %k0, %k1
+; KNL-NEXT:    vptestnmd %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    # kill: def $ax killed $ax killed $eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftl_v16i1_1:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmd %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftlw $1, %k0, %k1
+; SKX-NEXT:    vptestnmd %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    # kill: def $ax killed $ax killed $eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <16 x i32> %x, zeroinitializer
+  %b = shufflevector <16 x i1> %a, <16 x i1> zeroinitializer, <16 x i32> <i32 16, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14>
+  %c = icmp eq <16 x i32> %y, zeroinitializer
+  %d = and <16 x i1> %b, %c
+  %e = bitcast <16 x i1> %d to i16
+  ret i16 %e
+}
+
+define i32 @kshiftl_v32i1_1(<32 x i16> %x, <32 x i16> %y) {
+; KNL-LABEL: kshiftl_v32i1_1:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; KNL-NEXT:    vpcmpeqw %ymm4, %ymm1, %ymm1
+; KNL-NEXT:    vpmovsxwd %ymm1, %zmm1
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k1
+; KNL-NEXT:    vpcmpeqw %ymm4, %ymm0, %ymm0
+; KNL-NEXT:    vpmovsxwd %ymm0, %zmm0
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k2
+; KNL-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k2} {z}
+; KNL-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
+; KNL-NEXT:    valignd {{.*#+}} zmm0 = zmm0[15],zmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14]
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k1
+; KNL-NEXT:    kshiftlw $1, %k2, %k2
+; KNL-NEXT:    vpcmpeqw %ymm4, %ymm3, %ymm0
+; KNL-NEXT:    vpmovsxwd %ymm0, %zmm0
+; KNL-NEXT:    vpcmpeqw %ymm4, %ymm2, %ymm1
+; KNL-NEXT:    vpmovsxwd %ymm1, %zmm1
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k0 {%k2}
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    shll $16, %eax
+; KNL-NEXT:    orl %ecx, %eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftl_v32i1_1:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmw %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftld $1, %k0, %k1
+; SKX-NEXT:    vptestnmw %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <32 x i16> %x, zeroinitializer
+  %b = shufflevector <32 x i1> %a, <32 x i1> zeroinitializer, <32 x i32> <i32 32, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
+  %c = icmp eq <32 x i16> %y, zeroinitializer
+  %d = and <32 x i1> %b, %c
+  %e = bitcast <32 x i1> %d to i32
+  ret i32 %e
+}
+
+define i64 @kshiftl_v64i1_1(<64 x i8> %x, <64 x i8> %y) {
+; KNL-LABEL: kshiftl_v64i1_1:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; KNL-NEXT:    vpcmpeqb %ymm4, %ymm0, %ymm0
+; KNL-NEXT:    vpmovsxbd %xmm0, %zmm5
+; KNL-NEXT:    vptestmd %zmm5, %zmm5, %k3
+; KNL-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k2
+; KNL-NEXT:    vpcmpeqb %ymm4, %ymm1, %ymm0
+; KNL-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; KNL-NEXT:    vpmovsxbd %xmm1, %zmm1
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k1
+; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k4
+; KNL-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k4} {z}
+; KNL-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
+; KNL-NEXT:    valignd {{.*#+}} zmm1 = zmm0[15],zmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14]
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k1
+; KNL-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k2} {z}
+; KNL-NEXT:    valignd {{.*#+}} zmm0 = zmm1[15],zmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14]
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k2
+; KNL-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k3} {z}
+; KNL-NEXT:    valignd {{.*#+}} zmm0 = zmm0[15],zmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14]
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k4
+; KNL-NEXT:    kshiftlw $1, %k3, %k3
+; KNL-NEXT:    vpcmpeqb %ymm4, %ymm3, %ymm0
+; KNL-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; KNL-NEXT:    vpmovsxbd %xmm1, %zmm1
+; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
+; KNL-NEXT:    vpcmpeqb %ymm4, %ymm2, %ymm2
+; KNL-NEXT:    vextracti128 $1, %ymm2, %xmm3
+; KNL-NEXT:    vpmovsxbd %xmm3, %zmm3
+; KNL-NEXT:    vpmovsxbd %xmm2, %zmm2
+; KNL-NEXT:    vptestmd %zmm2, %zmm2, %k0 {%k3}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vptestmd %zmm3, %zmm3, %k0 {%k4}
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    shll $16, %ecx
+; KNL-NEXT:    orl %eax, %ecx
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0 {%k2}
+; KNL-NEXT:    kmovw %k0, %edx
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    shll $16, %eax
+; KNL-NEXT:    orl %edx, %eax
+; KNL-NEXT:    shlq $32, %rax
+; KNL-NEXT:    orq %rcx, %rax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftl_v64i1_1:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmb %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftlq $1, %k0, %k1
+; SKX-NEXT:    vptestnmb %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovq %k0, %rax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <64 x i8> %x, zeroinitializer
+  %b = shufflevector <64 x i1> %a, <64 x i1> zeroinitializer, <64 x i32> <i32 64, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62>
+  %c = icmp eq <64 x i8> %y, zeroinitializer
+  %d = and <64 x i1> %b, %c
+  %e = bitcast <64 x i1> %d to i64
+  ret i64 %e
+}
+
+define i8 @kshiftl_v8i1_7(<8 x i64> %x, <8 x i64> %y) {
+; KNL-LABEL: kshiftl_v8i1_7:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vptestnmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; KNL-NEXT:    movb $-128, %al
+; KNL-NEXT:    kmovw %eax, %k1
+; KNL-NEXT:    vpexpandq %zmm0, %zmm0 {%k1} {z}
+; KNL-NEXT:    vptestmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    # kill: def $al killed $al killed $eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftl_v8i1_7:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmq %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftlb $7, %k0, %k1
+; SKX-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    # kill: def $al killed $al killed $eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <8 x i64> %x, zeroinitializer
+  %b = shufflevector <8 x i1> zeroinitializer, <8 x i1> %a, <8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
+  %c = icmp eq <8 x i64> %y, zeroinitializer
+  %d = and <8 x i1> %b, %c
+  %e = bitcast <8 x i1> %d to i8
+  ret i8 %e
+}
+
+define i16 @kshiftl_v16i1_15(<16 x i32> %x, <16 x i32> %y) {
+; KNL-LABEL: kshiftl_v16i1_15:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vptestnmd %zmm0, %zmm0, %k0
+; KNL-NEXT:    kshiftlw $15, %k0, %k1
+; KNL-NEXT:    vptestnmd %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    # kill: def $ax killed $ax killed $eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftl_v16i1_15:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmd %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftlw $15, %k0, %k1
+; SKX-NEXT:    vptestnmd %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    # kill: def $ax killed $ax killed $eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <16 x i32> %x, zeroinitializer
+  %b = shufflevector <16 x i1> zeroinitializer, <16 x i1> %a, <16 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16>
+  %c = icmp eq <16 x i32> %y, zeroinitializer
+  %d = and <16 x i1> %b, %c
+  %e = bitcast <16 x i1> %d to i16
+  ret i16 %e
+}
+
+define i32 @kshiftl_v32i1_31(<32 x i16> %x, <32 x i16> %y) {
+; KNL-LABEL: kshiftl_v32i1_31:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; KNL-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; KNL-NEXT:    vpmovsxwd %ymm0, %zmm0
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0
+; KNL-NEXT:    kshiftlw $15, %k0, %k1
+; KNL-NEXT:    vpcmpeqw %ymm1, %ymm3, %ymm0
+; KNL-NEXT:    vpmovsxwd %ymm0, %zmm0
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    shll $16, %eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftl_v32i1_31:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmw %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftld $31, %k0, %k1
+; SKX-NEXT:    vptestnmw %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <32 x i16> %x, zeroinitializer
+  %b = shufflevector <32 x i1> zeroinitializer, <32 x i1> %a, <32 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32>
+  %c = icmp eq <32 x i16> %y, zeroinitializer
+  %d = and <32 x i1> %b, %c
+  %e = bitcast <32 x i1> %d to i32
+  ret i32 %e
+}
+
+define i64 @kshiftl_v64i1_63(<64 x i8> %x, <64 x i8> %y) {
+; KNL-LABEL: kshiftl_v64i1_63:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; KNL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0
+; KNL-NEXT:    kshiftlw $15, %k0, %k1
+; KNL-NEXT:    vpcmpeqb %ymm1, %ymm3, %ymm0
+; KNL-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    movzwl %ax, %eax
+; KNL-NEXT:    shlq $48, %rax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftl_v64i1_63:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmb %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftlq $63, %k0, %k1
+; SKX-NEXT:    vptestnmb %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovq %k0, %rax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <64 x i8> %x, zeroinitializer
+  %b = shufflevector <64 x i1> zeroinitializer, <64 x i1> %a, <64 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64>
+  %c = icmp eq <64 x i8> %y, zeroinitializer
+  %d = and <64 x i1> %b, %c
+  %e = bitcast <64 x i1> %d to i64
+  ret i64 %e
+}
+
+define i8 @kshiftr_v8i1_1(<8 x i64> %x, <8 x i64> %y) {
+; KNL-LABEL: kshiftr_v8i1_1:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vptestnmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vpternlogq $255, %zmm2, %zmm2, %zmm2 {%k1} {z}
+; KNL-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [1,2,3,4,5,6,7,15]
+; KNL-NEXT:    vpermi2q %zmm0, %zmm2, %zmm3
+; KNL-NEXT:    vptestmq %zmm3, %zmm3, %k1
+; KNL-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    # kill: def $al killed $al killed $eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftr_v8i1_1:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmq %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftrb $1, %k0, %k1
+; SKX-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    # kill: def $al killed $al killed $eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <8 x i64> %x, zeroinitializer
+  %b = shufflevector <8 x i1> %a, <8 x i1> zeroinitializer, <8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
+  %c = icmp eq <8 x i64> %y, zeroinitializer
+  %d = and <8 x i1> %b, %c
+  %e = bitcast <8 x i1> %d to i8
+  ret i8 %e
+}
+
+define i16 @kshiftr_v16i1_1(<16 x i32> %x, <16 x i32> %y) {
+; KNL-LABEL: kshiftr_v16i1_1:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vptestnmd %zmm0, %zmm0, %k0
+; KNL-NEXT:    kshiftrw $1, %k0, %k1
+; KNL-NEXT:    vptestnmd %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    # kill: def $ax killed $ax killed $eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftr_v16i1_1:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmd %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftrw $1, %k0, %k1
+; SKX-NEXT:    vptestnmd %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    # kill: def $ax killed $ax killed $eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <16 x i32> %x, zeroinitializer
+  %b = shufflevector <16 x i1> %a, <16 x i1> zeroinitializer, <16 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16>
+  %c = icmp eq <16 x i32> %y, zeroinitializer
+  %d = and <16 x i1> %b, %c
+  %e = bitcast <16 x i1> %d to i16
+  ret i16 %e
+}
+
+define i32 @kshiftr_v32i1_1(<32 x i16> %x, <32 x i16> %y) {
+; KNL-LABEL: kshiftr_v32i1_1:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; KNL-NEXT:    vpcmpeqw %ymm4, %ymm1, %ymm1
+; KNL-NEXT:    vpmovsxwd %ymm1, %zmm1
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k1
+; KNL-NEXT:    vpcmpeqw %ymm4, %ymm0, %ymm0
+; KNL-NEXT:    vpmovsxwd %ymm0, %zmm0
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k2
+; KNL-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k2} {z}
+; KNL-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
+; KNL-NEXT:    valignd {{.*#+}} zmm0 = zmm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zmm1[0]
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k2
+; KNL-NEXT:    kshiftrw $1, %k1, %k1
+; KNL-NEXT:    vpcmpeqw %ymm4, %ymm2, %ymm0
+; KNL-NEXT:    vpmovsxwd %ymm0, %zmm0
+; KNL-NEXT:    vpcmpeqw %ymm4, %ymm3, %ymm1
+; KNL-NEXT:    vpmovsxwd %ymm1, %zmm1
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    shll $16, %ecx
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0 {%k2}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    orl %ecx, %eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftr_v32i1_1:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmw %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftrd $1, %k0, %k1
+; SKX-NEXT:    vptestnmw %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <32 x i16> %x, zeroinitializer
+  %b = shufflevector <32 x i1> %a, <32 x i1> zeroinitializer, <32 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32>
+  %c = icmp eq <32 x i16> %y, zeroinitializer
+  %d = and <32 x i1> %b, %c
+  %e = bitcast <32 x i1> %d to i32
+  ret i32 %e
+}
+
+define i64 @kshiftr_v64i1_1(<64 x i8> %x, <64 x i8> %y) {
+; KNL-LABEL: kshiftr_v64i1_1:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; KNL-NEXT:    vpcmpeqb %ymm4, %ymm1, %ymm1
+; KNL-NEXT:    vextracti128 $1, %ymm1, %xmm5
+; KNL-NEXT:    vpmovsxbd %xmm5, %zmm5
+; KNL-NEXT:    vptestmd %zmm5, %zmm5, %k3
+; KNL-NEXT:    vpcmpeqb %ymm4, %ymm0, %ymm0
+; KNL-NEXT:    vpmovsxbd %xmm0, %zmm5
+; KNL-NEXT:    vptestmd %zmm5, %zmm5, %k2
+; KNL-NEXT:    vpmovsxbd %xmm1, %zmm1
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k1
+; KNL-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k4
+; KNL-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k4} {z}
+; KNL-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
+; KNL-NEXT:    valignd {{.*#+}} zmm5 = zmm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zmm1[0]
+; KNL-NEXT:    vptestmd %zmm5, %zmm5, %k1
+; KNL-NEXT:    vpternlogd $255, %zmm5, %zmm5, %zmm5 {%k2} {z}
+; KNL-NEXT:    valignd {{.*#+}} zmm0 = zmm5[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zmm0[0]
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k2
+; KNL-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k3} {z}
+; KNL-NEXT:    valignd {{.*#+}} zmm0 = zmm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zmm0[0]
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k4
+; KNL-NEXT:    kshiftrw $1, %k3, %k3
+; KNL-NEXT:    vpcmpeqb %ymm4, %ymm2, %ymm0
+; KNL-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; KNL-NEXT:    vpmovsxbd %xmm1, %zmm1
+; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
+; KNL-NEXT:    vpcmpeqb %ymm4, %ymm3, %ymm2
+; KNL-NEXT:    vpmovsxbd %xmm2, %zmm3
+; KNL-NEXT:    vextracti128 $1, %ymm2, %xmm2
+; KNL-NEXT:    vpmovsxbd %xmm2, %zmm2
+; KNL-NEXT:    vptestmd %zmm2, %zmm2, %k0 {%k3}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    shll $16, %eax
+; KNL-NEXT:    vptestmd %zmm3, %zmm3, %k0 {%k4}
+; KNL-NEXT:    kmovw %k0, %ecx
+; KNL-NEXT:    orl %eax, %ecx
+; KNL-NEXT:    shlq $32, %rcx
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0 {%k2}
+; KNL-NEXT:    kmovw %k0, %edx
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    shll $16, %eax
+; KNL-NEXT:    orl %edx, %eax
+; KNL-NEXT:    orq %rcx, %rax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftr_v64i1_1:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmb %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftrq $1, %k0, %k1
+; SKX-NEXT:    vptestnmb %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovq %k0, %rax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <64 x i8> %x, zeroinitializer
+  %b = shufflevector <64 x i1> %a, <64 x i1> zeroinitializer, <64 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64>
+  %c = icmp eq <64 x i8> %y, zeroinitializer
+  %d = and <64 x i1> %b, %c
+  %e = bitcast <64 x i1> %d to i64
+  ret i64 %e
+}
+
+define i8 @kshiftr_v8i1_7(<8 x i64> %x, <8 x i64> %y) {
+; KNL-LABEL: kshiftr_v8i1_7:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vptestnmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; KNL-NEXT:    movb $-2, %al
+; KNL-NEXT:    kmovw %eax, %k1
+; KNL-NEXT:    vpexpandq %zmm0, %zmm0 {%k1} {z}
+; KNL-NEXT:    vptestmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    # kill: def $al killed $al killed $eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftr_v8i1_7:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmq %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftlb $1, %k0, %k1
+; SKX-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    # kill: def $al killed $al killed $eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <8 x i64> %x, zeroinitializer
+  %b = shufflevector <8 x i1> %a, <8 x i1> zeroinitializer, <8 x i32> <i32 15, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6>
+  %c = icmp eq <8 x i64> %y, zeroinitializer
+  %d = and <8 x i1> %b, %c
+  %e = bitcast <8 x i1> %d to i8
+  ret i8 %e
+}
+
+define i16 @kshiftr_v16i1_15(<16 x i32> %x, <16 x i32> %y) {
+; KNL-LABEL: kshiftr_v16i1_15:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vptestnmd %zmm0, %zmm0, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k1
+; KNL-NEXT:    vptestnmd %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    # kill: def $ax killed $ax killed $eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftr_v16i1_15:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmd %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftrw $15, %k0, %k1
+; SKX-NEXT:    vptestnmd %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    # kill: def $ax killed $ax killed $eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <16 x i32> %x, zeroinitializer
+  %b = shufflevector <16 x i1> zeroinitializer, <16 x i1> %a, <16 x i32> <i32 31, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14>
+  %c = icmp eq <16 x i32> %y, zeroinitializer
+  %d = and <16 x i1> %b, %c
+  %e = bitcast <16 x i1> %d to i16
+  ret i16 %e
+}
+
+define i32 @kshiftr_v32i1_31(<32 x i16> %x, <32 x i16> %y) {
+; KNL-LABEL: kshiftr_v32i1_31:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vpcmpeqw %ymm0, %ymm1, %ymm1
+; KNL-NEXT:    vpmovsxwd %ymm1, %zmm1
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k1
+; KNL-NEXT:    vpcmpeqw %ymm0, %ymm2, %ymm0
+; KNL-NEXT:    vpmovsxwd %ymm0, %zmm0
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftr_v32i1_31:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmw %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftrd $31, %k0, %k1
+; SKX-NEXT:    vptestnmw %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <32 x i16> %x, zeroinitializer
+  %b = shufflevector <32 x i1> zeroinitializer, <32 x i1> %a, <32 x i32> <i32 63, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
+  %c = icmp eq <32 x i16> %y, zeroinitializer
+  %d = and <32 x i1> %b, %c
+  %e = bitcast <32 x i1> %d to i32
+  ret i32 %e
+}
+
+define i64 @kshiftr_v64i1_63(<64 x i8> %x, <64 x i8> %y) {
+; KNL-LABEL: kshiftr_v64i1_63:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vpcmpeqb %ymm0, %ymm1, %ymm1
+; KNL-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; KNL-NEXT:    vpmovsxbd %xmm1, %zmm1
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k0
+; KNL-NEXT:    kshiftrw $15, %k0, %k1
+; KNL-NEXT:    vpcmpeqb %ymm0, %ymm2, %ymm0
+; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
+; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    movzwl %ax, %eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftr_v64i1_63:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmb %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftrq $63, %k0, %k1
+; SKX-NEXT:    vptestnmb %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovq %k0, %rax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <64 x i8> %x, zeroinitializer
+  %b = shufflevector <64 x i1> zeroinitializer, <64 x i1> %a, <64 x i32> <i32 127, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62>
+  %c = icmp eq <64 x i8> %y, zeroinitializer
+  %d = and <64 x i1> %b, %c
+  %e = bitcast <64 x i1> %d to i64
+  ret i64 %e
+}
+
+define i8 @kshiftl_v8i1_zu123u56(<8 x i64> %x, <8 x i64> %y) {
+; KNL-LABEL: kshiftl_v8i1_zu123u56:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vptestnmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vpternlogq $255, %zmm2, %zmm2, %zmm2 {%k1} {z}
+; KNL-NEXT:    vmovdqa64 {{.*#+}} zmm3 = <8,u,1,2,3,u,5,6>
+; KNL-NEXT:    vpermi2q %zmm0, %zmm2, %zmm3
+; KNL-NEXT:    vpsllq $63, %zmm3, %zmm0
+; KNL-NEXT:    vptestmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    # kill: def $al killed $al killed $eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftl_v8i1_zu123u56:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmq %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftlb $1, %k0, %k1
+; SKX-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    # kill: def $al killed $al killed $eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <8 x i64> %x, zeroinitializer
+  %b = shufflevector <8 x i1> %a, <8 x i1> zeroinitializer, <8 x i32> <i32 8, i32 undef, i32 1, i32 2, i32 3, i32 undef, i32 5, i32 6>
+  %c = icmp eq <8 x i64> %y, zeroinitializer
+  %d = and <8 x i1> %b, %c
+  %e = bitcast <8 x i1> %d to i8
+  ret i8 %e
+}
+
+define i8 @kshiftl_v8i1_u0123456(<8 x i64> %x, <8 x i64> %y) {
+; KNL-LABEL: kshiftl_v8i1_u0123456:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vptestnmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; KNL-NEXT:    valignq {{.*#+}} zmm0 = zmm0[7,0,1,2,3,4,5,6]
+; KNL-NEXT:    vpsllq $63, %zmm0, %zmm0
+; KNL-NEXT:    vptestmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    # kill: def $al killed $al killed $eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftl_v8i1_u0123456:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmq %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftlb $1, %k0, %k1
+; SKX-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    # kill: def $al killed $al killed $eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <8 x i64> %x, zeroinitializer
+  %b = shufflevector <8 x i1> %a, <8 x i1> undef, <8 x i32> <i32 undef, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6>
+  %c = icmp eq <8 x i64> %y, zeroinitializer
+  %d = and <8 x i1> %b, %c
+  %e = bitcast <8 x i1> %d to i8
+  ret i8 %e
+}
+
+define i8 @kshiftr_v8i1_1u3u567z(<8 x i64> %x, <8 x i64> %y) {
+; KNL-LABEL: kshiftr_v8i1_1u3u567z:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vptestnmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vpternlogq $255, %zmm2, %zmm2, %zmm2 {%k1} {z}
+; KNL-NEXT:    vmovdqa64 {{.*#+}} zmm3 = <1,u,3,u,5,6,7,15>
+; KNL-NEXT:    vpermi2q %zmm0, %zmm2, %zmm3
+; KNL-NEXT:    vpsllq $63, %zmm3, %zmm0
+; KNL-NEXT:    vptestmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    # kill: def $al killed $al killed $eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftr_v8i1_1u3u567z:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmq %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftrb $1, %k0, %k1
+; SKX-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    # kill: def $al killed $al killed $eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <8 x i64> %x, zeroinitializer
+  %b = shufflevector <8 x i1> %a, <8 x i1> zeroinitializer, <8 x i32> <i32 1, i32 undef, i32 3, i32 undef, i32 5, i32 6, i32 7, i32 8>
+  %c = icmp eq <8 x i64> %y, zeroinitializer
+  %d = and <8 x i1> %b, %c
+  %e = bitcast <8 x i1> %d to i8
+  ret i8 %e
+}
+
+define i8 @kshiftr_v8i1_234567uu(<8 x i64> %x, <8 x i64> %y) {
+; KNL-LABEL: kshiftr_v8i1_234567uu:
+; KNL:       # %bb.0:
+; KNL-NEXT:    vptestnmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; KNL-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[2,3,4,5,6,7,0,1]
+; KNL-NEXT:    vpsllq $63, %zmm0, %zmm0
+; KNL-NEXT:    vptestmq %zmm0, %zmm0, %k1
+; KNL-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    # kill: def $al killed $al killed $eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: kshiftr_v8i1_234567uu:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vptestnmq %zmm0, %zmm0, %k0
+; SKX-NEXT:    kshiftrb $2, %k0, %k1
+; SKX-NEXT:    vptestnmq %zmm1, %zmm1, %k0 {%k1}
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    # kill: def $al killed $al killed $eax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+  %a = icmp eq <8 x i64> %x, zeroinitializer
+  %b = shufflevector <8 x i1> %a, <8 x i1> undef, <8 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 10>
+  %c = icmp eq <8 x i64> %y, zeroinitializer
+  %d = and <8 x i1> %b, %c
+  %e = bitcast <8 x i1> %d to i8
+  ret i8 %e
+}