AVX512 : VPERMILPS/PD implementation

[igorb]

Summary

AVX512: Implemented encoding and intrinsics for VPERMILPS/PD instructions.
Added tests for intrinsics and encoding.

[delena]

LGTM

lib/Target/X86/X86InstrAVX512.td

3901

Please add a comment here.

include/llvm/IR/IntrinsicsX86.td

def int_x86_avx512_maskz_vpermt2var_q_256 :
           [IntrNoMem]>;
 
   def int_x86_avx512_maskz_vpermt2var_q_512 : 
         GCCBuiltin<"__builtin_ia32_vpermt2varq512_maskz">,
           Intrinsic<[llvm_v8i64_ty],
           [llvm_v8i64_ty, llvm_v8i64_ty, llvm_v8i64_ty, llvm_i8_ty],
           [IntrNoMem]>;
 
+  def int_x86_avx512_mask_vpermil_pd_128 :
+        GCCBuiltin<"__builtin_ia32_vpermilpd_mask">,
+          Intrinsic<[llvm_v2f64_ty],
+          [llvm_v2f64_ty, llvm_i32_ty, llvm_v2f64_ty, llvm_i8_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_mask_vpermil_pd_256 :
+        GCCBuiltin<"__builtin_ia32_vpermilpd256_mask">,
+          Intrinsic<[llvm_v4f64_ty],
+          [llvm_v4f64_ty, llvm_i32_ty, llvm_v4f64_ty, llvm_i8_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_mask_vpermil_pd_512 :
+        GCCBuiltin<"__builtin_ia32_vpermilpd512_mask">,
+          Intrinsic<[llvm_v8f64_ty],
+          [llvm_v8f64_ty, llvm_i32_ty, llvm_v8f64_ty, llvm_i8_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_mask_vpermil_ps_128 :
+        GCCBuiltin<"__builtin_ia32_vpermilps_mask">,
+          Intrinsic<[llvm_v4f32_ty],
+          [llvm_v4f32_ty, llvm_i32_ty, llvm_v4f32_ty, llvm_i8_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_mask_vpermil_ps_256 :
+        GCCBuiltin<"__builtin_ia32_vpermilps256_mask">,
+          Intrinsic<[llvm_v8f32_ty],
+          [llvm_v8f32_ty, llvm_i32_ty, llvm_v8f32_ty, llvm_i8_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_mask_vpermil_ps_512 :
+        GCCBuiltin<"__builtin_ia32_vpermilps512_mask">,
+          Intrinsic<[llvm_v16f32_ty],
+          [llvm_v16f32_ty, llvm_i32_ty, llvm_v16f32_ty, llvm_i16_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_mask_vpermilvar_pd_256 :
+        GCCBuiltin<"__builtin_ia32_vpermilvarpd256_mask">,
+          Intrinsic<[llvm_v4f64_ty],
+          [llvm_v4f64_ty, llvm_v4i64_ty, llvm_v4f64_ty, llvm_i8_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_mask_vpermilvar_pd_512 :
+        GCCBuiltin<"__builtin_ia32_vpermilvarpd512_mask">,
+          Intrinsic<[llvm_v8f64_ty],
+          [llvm_v8f64_ty, llvm_v8i64_ty, llvm_v8f64_ty, llvm_i8_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_mask_vpermilvar_pd_128 :
+        GCCBuiltin<"__builtin_ia32_vpermilvarpd_mask">,
+          Intrinsic<[llvm_v2f64_ty],
+          [llvm_v2f64_ty, llvm_v2i64_ty, llvm_v2f64_ty, llvm_i8_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_mask_vpermilvar_ps_256 :
+        GCCBuiltin<"__builtin_ia32_vpermilvarps256_mask">,
+          Intrinsic<[llvm_v8f32_ty],
+          [llvm_v8f32_ty, llvm_v8i32_ty, llvm_v8f32_ty, llvm_i8_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_mask_vpermilvar_ps_512 :
+        GCCBuiltin<"__builtin_ia32_vpermilvarps512_mask">,
+          Intrinsic<[llvm_v16f32_ty],
+          [llvm_v16f32_ty, llvm_v16i32_ty, llvm_v16f32_ty, llvm_i16_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_mask_vpermilvar_ps_128 :
+        GCCBuiltin<"__builtin_ia32_vpermilvarps_mask">,
+          Intrinsic<[llvm_v4f32_ty],
+          [llvm_v4f32_ty, llvm_v4i32_ty, llvm_v4f32_ty, llvm_i8_ty],
+          [IntrNoMem]>;
+
   def int_x86_avx512_mask_pshuf_b_128 : 
         GCCBuiltin<"__builtin_ia32_pshufb128_mask">,
           Intrinsic<[llvm_v16i8_ty],
           [llvm_v16i8_ty, llvm_v16i8_ty, llvm_v16i8_ty,  llvm_i16_ty],
           [IntrNoMem]>;
 
   def int_x86_avx512_mask_pshuf_b_256 : 
         GCCBuiltin<"__builtin_ia32_pshufb256_mask">,

lib/Target/X86/X86ISelLowering.cpp

case INTR_TYPE_SCALAR_MASK_RM: {
       }
       assert(Op.getNumOperands() == 7 && "Unexpected intrinsic form");
       SDValue RoundingMode  = Op.getOperand(5);
       SDValue Sae  = Op.getOperand(6);
       return getScalarMaskingNode(DAG.getNode(IntrData->Opc0, dl, VT, Src1, Src2,
                                               RoundingMode, Sae),
                                   Mask, Src0, Subtarget, DAG);
     }
-    case INTR_TYPE_2OP_MASK: {
+    case INTR_TYPE_2OP_MASK:
+    case INTR_TYPE_2OP_IMM8_MASK: {
       SDValue Src1 = Op.getOperand(1);
       SDValue Src2 = Op.getOperand(2);
       SDValue PassThru = Op.getOperand(3);
       SDValue Mask = Op.getOperand(4);
+
+      if (IntrData->Type == INTR_TYPE_2OP_IMM8_MASK)
+        Src2 = DAG.getNode(ISD::TRUNCATE, dl, MVT::i8, Src2);
+
       // We specify 2 possible opcodes for intrinsics with rounding modes.
       // First, we check if the intrinsic may have non-default rounding mode,
       // (IntrData->Opc1 != 0), then we check the rounding mode operand.
       unsigned IntrWithRoundingModeOpcode = IntrData->Opc1;
       if (IntrWithRoundingModeOpcode != 0) {
         SDValue Rnd = Op.getOperand(5);
         unsigned Round = cast<ConstantSDNode>(Rnd)->getZExtValue();
         if (Round != X86::STATIC_ROUNDING::CUR_DIRECTION) {

lib/Target/X86/X86InstrAVX512.td

 let Predicates = [HasCDI] in {
 defm VPBROADCASTMW2D : avx512_mask_broadcast<0x3A, "vpbroadcastmw2d",
                                              VK16>;
 defm VPBROADCASTMB2Q : avx512_mask_broadcast<0x2A, "vpbroadcastmb2q",
                                              VK8>, VEX_W;
 }
 
 //===----------------------------------------------------------------------===//
-// AVX-512 - VPERM
-//
-// -- immediate form --
-multiclass avx512_perm_imm<bits<8> opc, string OpcodeStr, SDNode OpNode,
-                           X86VectorVTInfo _> {
-  let ExeDomain = _.ExeDomain in {
-  def ri : AVX512AIi8<opc, MRMSrcReg, (outs _.RC:$dst),
-                     (ins _.RC:$src1, u8imm:$src2),
-                     !strconcat(OpcodeStr,
-                         "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-                     [(set _.RC:$dst,
-                       (_.VT (OpNode _.RC:$src1, (i8 imm:$src2))))]>,
-                     EVEX;
-  def mi : AVX512AIi8<opc, MRMSrcMem, (outs _.RC:$dst),
-                     (ins _.MemOp:$src1, u8imm:$src2),
-                     !strconcat(OpcodeStr,
-                         "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-                     [(set _.RC:$dst,
-                       (_.VT (OpNode (_.LdFrag addr:$src1),
-                              (i8 imm:$src2))))]>,
-           EVEX, EVEX_CD8<_.EltSize, CD8VF>;
-}
-}
-
-multiclass avx512_permil<bits<8> OpcImm, bits<8> OpcVar, X86VectorVTInfo _,
-                         X86VectorVTInfo Ctrl> :
-     avx512_perm_imm<OpcImm, "vpermil" # _.Suffix, X86VPermilpi, _> {
-  let ExeDomain = _.ExeDomain in {
-    def rr : AVX5128I<OpcVar, MRMSrcReg, (outs _.RC:$dst),
-                     (ins _.RC:$src1, _.RC:$src2),
-                     !strconcat("vpermil" # _.Suffix,
-                         "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-                     [(set _.RC:$dst,
-                         (_.VT (X86VPermilpv _.RC:$src1,
-                                  (Ctrl.VT Ctrl.RC:$src2))))]>,
-             EVEX_4V;
-    def rm : AVX5128I<OpcVar, MRMSrcMem, (outs _.RC:$dst),
-                     (ins _.RC:$src1, Ctrl.MemOp:$src2),
-                     !strconcat("vpermil" # _.Suffix,
-                         "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-                     [(set _.RC:$dst,
-                         (_.VT (X86VPermilpv _.RC:$src1,
-                                  (Ctrl.VT (Ctrl.LdFrag addr:$src2)))))]>,
-             EVEX_4V;
-  }
-}
-defm VPERMILPSZ : avx512_permil<0x04, 0x0C, v16f32_info, v16i32_info>,
-                  EVEX_V512;
-defm VPERMILPDZ : avx512_permil<0x05, 0x0D, v8f64_info, v8i64_info>,
-                  EVEX_V512, VEX_W;
-
-def : Pat<(v16i32 (X86VPermilpi VR512:$src1, (i8 imm:$imm))),
-          (VPERMILPSZri VR512:$src1, imm:$imm)>;
-def : Pat<(v8i64 (X86VPermilpi VR512:$src1, (i8 imm:$imm))),
-          (VPERMILPDZri VR512:$src1, imm:$imm)>;
-
 // -- VPERM2I - 3 source operands form --
 multiclass avx512_perm_3src<bits<8> opc, string OpcodeStr,
                             SDNode OpNode, X86VectorVTInfo _> {
 let Constraints = "$src1 = $dst" in {
   defm rr: AVX512_maskable_3src<opc, MRMSrcReg, _, (outs _.RC:$dst),
           (ins _.RC:$src2, _.RC:$src3),
           OpcodeStr, "$src3, $src2", "$src2, $src3",
           (_.VT (OpNode _.RC:$src1, _.RC:$src2, _.RC:$src3))>, EVEX_4V,
defm VPERMPD : avx512_vperm_dq_sizes<0x16, "vpermpd", X86VPermv,
                                     avx512vl_f64_info>, VEX_W;
 
 defm VPERMQ : avx512_vpermi_dq_sizes<0x00, MRMSrcReg, MRMSrcMem, "vpermq",
                              X86VPermi, avx512vl_i64_info>,
                              EVEX, AVX512AIi8Base, EVEX_CD8<64, CD8VF>, VEX_W;
 defm VPERMPD : avx512_vpermi_dq_sizes<0x01, MRMSrcReg, MRMSrcMem, "vpermpd",
                              X86VPermi, avx512vl_f64_info>,
                              EVEX, AVX512AIi8Base, EVEX_CD8<64, CD8VF>, VEX_W;
+//===----------------------------------------------------------------------===//
+// AVX-512 - VPERMIL 
+//===----------------------------------------------------------------------===//
+
+multiclass avx512_permil_vec<bits<8> OpcVar, string OpcodeStr,  SDNode OpNode,
+                             X86VectorVTInfo _, X86VectorVTInfo Ctrl> {
+  defm rr: AVX512_maskable<OpcVar, MRMSrcReg, _, (outs _.RC:$dst),
+                  (ins _.RC:$src1, Ctrl.RC:$src2), OpcodeStr,
+                  "$src2, $src1", "$src1, $src2",
+                  (_.VT (OpNode _.RC:$src1,
+                               (Ctrl.VT Ctrl.RC:$src2)))>,
+                  T8PD, EVEX_4V;
+  let mayLoad = 1 in {
+    defm rm: AVX512_maskable<OpcVar, MRMSrcMem, _, (outs _.RC:$dst),
+                    (ins _.RC:$src1, Ctrl.MemOp:$src2), OpcodeStr,
+                    "$src2, $src1", "$src1, $src2",
+                    (_.VT (OpNode
+                             _.RC:$src1,
+                             (Ctrl.VT (bitconvert(Ctrl.LdFrag addr:$src2)))))>,
+                    T8PD, EVEX_4V, EVEX_CD8<_.EltSize, CD8VF>;
+    defm rmb: AVX512_maskable<OpcVar, MRMSrcMem, _, (outs _.RC:$dst),
+                     (ins _.RC:$src1, _.ScalarMemOp:$src2), OpcodeStr,
+                     "${src2}"##_.BroadcastStr##", $src1",
+                     "$src1, ${src2}"##_.BroadcastStr,
+                     (_.VT (OpNode
+                              _.RC:$src1,
+                              (Ctrl.VT (X86VBroadcast
+                                         (Ctrl.ScalarLdFrag addr:$src2)))))>,
+                     T8PD, EVEX_4V, EVEX_B, EVEX_CD8<_.EltSize, CD8VF>;
+  }//let mayLoad = 1
+}
+
+multiclass avx512_permil_vec_common<string OpcodeStr, bits<8> OpcVar,
+                             AVX512VLVectorVTInfo _, AVX512VLVectorVTInfo Ctrl>{
+  let Predicates = [HasAVX512] in {
+    defm Z    : avx512_permil_vec<OpcVar, OpcodeStr, X86VPermilpv, _.info512,
+                                  Ctrl.info512>, EVEX_V512;
+  }
+  let Predicates = [HasAVX512, HasVLX] in {
+    defm Z128 : avx512_permil_vec<OpcVar, OpcodeStr, X86VPermilpv, _.info128,
+                                  Ctrl.info128>, EVEX_V128;
+    defm Z256 : avx512_permil_vec<OpcVar, OpcodeStr, X86VPermilpv, _.info256,
+                                  Ctrl.info256>, EVEX_V256;
+  }
+}
+
+multiclass avx512_permil<string OpcodeStr, bits<8> OpcImm, bits<8> OpcVar,
+                         AVX512VLVectorVTInfo _, AVX512VLVectorVTInfo Ctrl>{
+
+  defm NAME: avx512_permil_vec_common<OpcodeStr, OpcVar, _, Ctrl>;
+  defm NAME: avx512_shift_rmi_sizes<OpcImm, MRMSrcReg, MRMSrcMem, OpcodeStr,
+                                    X86VPermilpi, _>,
+                    EVEX, AVX512AIi8Base, EVEX_CD8<_.info128.EltSize, CD8VF>;
 
+  let isCodeGenOnly = 1 in {
+    defm NAME#_I: avx512_permil_vec_common<OpcodeStr, OpcVar, Ctrl, Ctrl>;
+    defm NAME#_I: avx512_shift_rmi_sizes<OpcImm, MRMSrcReg, MRMSrcMem,
+                                         OpcodeStr, X86VPermilpi, Ctrl>,
+                    EVEX, AVX512AIi8Base, EVEX_CD8<_.info128.EltSize, CD8VF>;
+  }
+}
+
+defm VPERMILPS : avx512_permil<"vpermilps", 0x04, 0x0C, avx512vl_f32_info,
+                               avx512vl_i32_info>;
+defm VPERMILPD : avx512_permil<"vpermilpd", 0x05, 0x0D, avx512vl_f64_info,
+                               avx512vl_i64_info>, VEX_W;
 //===----------------------------------------------------------------------===//
 // AVX-512 - VPSHUFD, VPSHUFLW, VPSHUFHW
 //===----------------------------------------------------------------------===//
 
 defm VPSHUFD : avx512_shift_rmi_sizes<0x70, MRMSrcReg, MRMSrcMem, "vpshufd",
                              X86PShufd, avx512vl_i32_info>,
                              EVEX, AVX512BIi8Base, EVEX_CD8<32, CD8VF>;
 defm VPSHUFH : avx512_shift_rmi_w<0x70, MRMSrcReg, MRMSrcMem, "vpshufhw",

lib/Target/X86/X86InstrSSE.td

def rr  : AVX8I<opc_rm, MRMSrcReg, (outs RC:$dst),
              Sched<[WriteFShuffle]>;
   def rm  : AVX8I<opc_rm, MRMSrcMem, (outs RC:$dst),
              (ins RC:$src1, x86memop_i:$src2),
              !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
              [(set RC:$dst, (IntVar RC:$src1,
                              (bitconvert (i_frag addr:$src2))))]>, VEX_4V,
              Sched<[WriteFShuffleLd, ReadAfterLd]>;
 
+  let Predicates = [HasAVX, NoVLX] in {
     def ri  : AVXAIi8<opc_rmi, MRMSrcReg, (outs RC:$dst),
              (ins RC:$src1, u8imm:$src2),
              !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
              [(set RC:$dst, (vt (X86VPermilpi RC:$src1, (i8 imm:$src2))))]>, VEX,
              Sched<[WriteFShuffle]>;
     def mi  : AVXAIi8<opc_rmi, MRMSrcMem, (outs RC:$dst),
              (ins x86memop_f:$src1, u8imm:$src2),
              !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
              [(set RC:$dst,
                (vt (X86VPermilpi (load addr:$src1), (i8 imm:$src2))))]>, VEX,
              Sched<[WriteFShuffleLd]>;
+  }// Predicates = [HasAVX, NoVLX]
 }
 
 let ExeDomain = SSEPackedSingle in {
   defm VPERMILPS  : avx_permil<0x0C, 0x04, "vpermilps", VR128, f128mem, i128mem,
                                loadv2i64, int_x86_avx_vpermilvar_ps, v4f32>;
   defm VPERMILPSY : avx_permil<0x0C, 0x04, "vpermilps", VR256, f256mem, i256mem,
                        loadv4i64, int_x86_avx_vpermilvar_ps_256, v8f32>, VEX_L;
 }
 let ExeDomain = SSEPackedDouble in {
   defm VPERMILPD  : avx_permil<0x0D, 0x05, "vpermilpd", VR128, f128mem, i128mem,
                                loadv2i64, int_x86_avx_vpermilvar_pd, v2f64>;
   defm VPERMILPDY : avx_permil<0x0D, 0x05, "vpermilpd", VR256, f256mem, i256mem,
                        loadv4i64, int_x86_avx_vpermilvar_pd_256, v4f64>, VEX_L;
 }
 
-let Predicates = [HasAVX] in {
+let Predicates = [HasAVX, NoVLX] in {
 def : Pat<(v8f32 (X86VPermilpv VR256:$src1, (v8i32 VR256:$src2))),
           (VPERMILPSYrr VR256:$src1, VR256:$src2)>;
 def : Pat<(v8f32 (X86VPermilpv VR256:$src1, (bc_v8i32 (loadv4i64 addr:$src2)))),
           (VPERMILPSYrm VR256:$src1, addr:$src2)>;
 def : Pat<(v4f64 (X86VPermilpv VR256:$src1, (v4i64 VR256:$src2))),
           (VPERMILPDYrr VR256:$src1, VR256:$src2)>;
 def : Pat<(v4f64 (X86VPermilpv VR256:$src1, (loadv4i64 addr:$src2))),
           (VPERMILPDYrm VR256:$src1, addr:$src2)>;

lib/Target/X86/X86IntrinsicsInfo.h

 namespace llvm {
 
 enum IntrinsicType {
   INTR_NO_TYPE,
   GATHER, SCATTER, PREFETCH, RDSEED, RDRAND, RDPMC, RDTSC, XTEST, ADX,
   INTR_TYPE_1OP, INTR_TYPE_2OP, INTR_TYPE_2OP_IMM8, INTR_TYPE_3OP, INTR_TYPE_4OP,
   CMP_MASK, CMP_MASK_CC, VSHIFT, VSHIFT_MASK, COMI,
   INTR_TYPE_1OP_MASK, INTR_TYPE_1OP_MASK_RM,
-  INTR_TYPE_2OP_MASK, INTR_TYPE_2OP_MASK_RM,
+  INTR_TYPE_2OP_MASK, INTR_TYPE_2OP_MASK_RM, INTR_TYPE_2OP_IMM8_MASK,
   INTR_TYPE_3OP_MASK, INTR_TYPE_3OP_MASK_RM, INTR_TYPE_3OP_IMM8_MASK,
   FMA_OP_MASK, FMA_OP_MASKZ, FMA_OP_MASK3, VPERM_3OP_MASK,
   VPERM_3OP_MASKZ,
   INTR_TYPE_SCALAR_MASK_RM, INTR_TYPE_3OP_SCALAR_MASK_RM,
   COMPRESS_EXPAND_IN_REG, COMPRESS_TO_MEM,
   TRUNCATE_TO_MEM_VI8, TRUNCATE_TO_MEM_VI16, TRUNCATE_TO_MEM_VI32,
   EXPAND_FROM_MEM, BLEND
 };

   X86_INTRINSIC_DATA(avx512_mask_vpermi2var_ps_512, VPERM_3OP_MASK,
                     X86ISD::VPERMIV3, 0),
   X86_INTRINSIC_DATA(avx512_mask_vpermi2var_q_128, VPERM_3OP_MASK,
                     X86ISD::VPERMIV3, 0),
   X86_INTRINSIC_DATA(avx512_mask_vpermi2var_q_256, VPERM_3OP_MASK,
                     X86ISD::VPERMIV3, 0),
   X86_INTRINSIC_DATA(avx512_mask_vpermi2var_q_512, VPERM_3OP_MASK,
                     X86ISD::VPERMIV3, 0),
+  X86_INTRINSIC_DATA(avx512_mask_vpermil_pd_128, INTR_TYPE_2OP_IMM8_MASK,
+                     X86ISD::VPERMILPI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_vpermil_pd_256, INTR_TYPE_2OP_IMM8_MASK,
+                     X86ISD::VPERMILPI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_vpermil_pd_512, INTR_TYPE_2OP_IMM8_MASK,
+                     X86ISD::VPERMILPI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_vpermil_ps_128, INTR_TYPE_2OP_IMM8_MASK,
+                     X86ISD::VPERMILPI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_vpermil_ps_256, INTR_TYPE_2OP_IMM8_MASK,
+                     X86ISD::VPERMILPI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_vpermil_ps_512, INTR_TYPE_2OP_IMM8_MASK,
+                     X86ISD::VPERMILPI, 0),
+  X86_INTRINSIC_DATA(avx512_mask_vpermilvar_pd_128, INTR_TYPE_2OP_MASK,
+                     X86ISD::VPERMILPV, 0),
+  X86_INTRINSIC_DATA(avx512_mask_vpermilvar_pd_256, INTR_TYPE_2OP_MASK,
+                     X86ISD::VPERMILPV, 0),
+  X86_INTRINSIC_DATA(avx512_mask_vpermilvar_pd_512, INTR_TYPE_2OP_MASK,
+                     X86ISD::VPERMILPV, 0),
+  X86_INTRINSIC_DATA(avx512_mask_vpermilvar_ps_128, INTR_TYPE_2OP_MASK,
+                     X86ISD::VPERMILPV, 0),
+  X86_INTRINSIC_DATA(avx512_mask_vpermilvar_ps_256, INTR_TYPE_2OP_MASK,
+                     X86ISD::VPERMILPV, 0),
+  X86_INTRINSIC_DATA(avx512_mask_vpermilvar_ps_512, INTR_TYPE_2OP_MASK,
+                     X86ISD::VPERMILPV, 0),
   X86_INTRINSIC_DATA(avx512_mask_vpermt2var_d_128, VPERM_3OP_MASK,
                     X86ISD::VPERMV3, 0),
   X86_INTRINSIC_DATA(avx512_mask_vpermt2var_d_128, VPERM_3OP_MASK,
                     X86ISD::VPERMV3, 0),
   X86_INTRINSIC_DATA(avx512_mask_vpermt2var_d_256, VPERM_3OP_MASK,
                     X86ISD::VPERMV3, 0),
   X86_INTRINSIC_DATA(avx512_mask_vpermt2var_d_256, VPERM_3OP_MASK,
                     X86ISD::VPERMV3, 0),

test/CodeGen/X86/avx512-intrinsics.ll

 ; CHECK-NEXT:    vaddps %zmm0, %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %res = call <16 x float> @llvm.x86.avx512.mask.shuf.ps.512(<16 x float> %x0, <16 x float> %x1, i32 22, <16 x float> %x3, i16 %x4)
   %res1 = call <16 x float> @llvm.x86.avx512.mask.shuf.ps.512(<16 x float> %x0, <16 x float> %x1, i32 22, <16 x float> %x3, i16 -1)  
   %res2 = fadd <16 x float> %res, %res1
   ret <16 x float> %res2
 }
 
+declare <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double>, i32, <8 x double>, i8)
+
+define <8 x double>@test_int_x86_avx512_mask_vpermil_pd_512(<8 x double> %x0, <8 x double> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_pd_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %dil, %eax
+; CHECK-NEXT:    kmovw %eax, %k1 
+; CHECK-NEXT:    vpermilpd $22, %zmm0, %zmm1 {%k1} 
+; CHECK-NEXT:    vpermilpd $22, %zmm0, %zmm2 {%k1} {z} 
+; CHECK-NEXT:    vpermilpd $22, %zmm0, %zmm0 
+; CHECK-NEXT:    vaddpd %zmm2, %zmm1, %zmm1 
+; CHECK-NEXT:    vaddpd %zmm0, %zmm1, %zmm0 
+; CHECK-NEXT:    retq 
+  %res = call <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double> %x0, i32 22, <8 x double> %x2, i8 %x3)
+  %res1 = call <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double> %x0, i32 22, <8 x double> zeroinitializer, i8 %x3)
+  %res2 = call <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double> %x0, i32 22, <8 x double> %x2, i8 -1)
+  %res3 = fadd <8 x double> %res, %res1
+  %res4 = fadd <8 x double> %res3, %res2
+  ret <8 x double> %res4
+}
+
+declare <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float>, i32, <16 x float>, i16)
+
+define <16 x float>@test_int_x86_avx512_mask_vpermil_ps_512(<16 x float> %x0, <16 x float> %x2, i16 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_ps_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 
+; CHECK-NEXT:    vpermilps $22, %zmm0, %zmm1 {%k1} 
+; CHECK-NEXT:    vpermilps $22, %zmm0, %zmm2 {%k1} {z} 
+; CHECK-NEXT:    vpermilps $22, %zmm0, %zmm0 
+; CHECK-NEXT:    vaddps %zmm2, %zmm1, %zmm1 
+; CHECK-NEXT:    vaddps %zmm0, %zmm1, %zmm0 
+; CHECK-NEXT:    retq 
+  %res = call <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float> %x0, i32 22, <16 x float> %x2, i16 %x3)
+  %res1 = call <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float> %x0, i32 22, <16 x float> zeroinitializer, i16 %x3)
+  %res2 = call <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float> %x0, i32 22, <16 x float> %x2, i16 -1)
+  %res3 = fadd <16 x float> %res, %res1
+  %res4 = fadd <16 x float> %res3, %res2
+  ret <16 x float> %res4
+}
+
+declare <8 x double> @llvm.x86.avx512.mask.vpermilvar.pd.512(<8 x double>, <8 x i64>, <8 x double>, i8)
+
+define <8 x double>@test_int_x86_avx512_mask_vpermilvar_pd_512(<8 x double> %x0, <8 x i64> %x1, <8 x double> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_vpermilvar_pd_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %dil, %eax
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpermilpd %zmm1, %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vpermilpd %zmm1, %zmm0, %zmm3 {%k1} {z}
+; CHECK-NEXT:    vpermilpd %zmm1, %zmm0, %zmm0
+; CHECK-NEXT:    vaddpd %zmm3, %zmm2, %zmm1
+; CHECK-NEXT:    vaddpd %zmm1, %zmm0, %zmm0
+; CHECK-NEXT:    retq
+  %res = call <8 x double> @llvm.x86.avx512.mask.vpermilvar.pd.512(<8 x double> %x0, <8 x i64> %x1, <8 x double> %x2, i8 %x3)
+  %res1 = call <8 x double> @llvm.x86.avx512.mask.vpermilvar.pd.512(<8 x double> %x0, <8 x i64> %x1, <8 x double> zeroinitializer, i8 %x3)
+  %res2 = call <8 x double> @llvm.x86.avx512.mask.vpermilvar.pd.512(<8 x double> %x0, <8 x i64> %x1, <8 x double> %x2, i8 -1)
+  %res3 = fadd <8 x double> %res, %res1
+  %res4 = fadd <8 x double> %res2, %res3
+  ret <8 x double> %res4
+}
+
+declare <16 x float> @llvm.x86.avx512.mask.vpermilvar.ps.512(<16 x float>, <16 x i32>, <16 x float>, i16)
+
+define <16 x float>@test_int_x86_avx512_mask_vpermilvar_ps_512(<16 x float> %x0, <16 x i32> %x1, <16 x float> %x2, i16 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_vpermilvar_ps_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1
+; CHECK-NEXT:    vpermilps %zmm1, %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vpermilps %zmm1, %zmm0, %zmm3 {%k1} {z}
+; CHECK-NEXT:    vpermilps %zmm1, %zmm0, %zmm0
+; CHECK-NEXT:    vaddps %zmm3, %zmm2, %zmm1
+; CHECK-NEXT:    vaddps %zmm1, %zmm0, %zmm0
+; CHECK-NEXT:    retq
+  %res = call <16 x float> @llvm.x86.avx512.mask.vpermilvar.ps.512(<16 x float> %x0, <16 x i32> %x1, <16 x float> %x2, i16 %x3)
+  %res1 = call <16 x float> @llvm.x86.avx512.mask.vpermilvar.ps.512(<16 x float> %x0, <16 x i32> %x1, <16 x float> zeroinitializer, i16 %x3)
+  %res2 = call <16 x float> @llvm.x86.avx512.mask.vpermilvar.ps.512(<16 x float> %x0, <16 x i32> %x1, <16 x float> %x2, i16 -1)
+  %res3 = fadd <16 x float> %res, %res1
+  %res4 = fadd <16 x float> %res2, %res3
+  ret <16 x float> %res4
+}

test/CodeGen/X86/avx512vl-intrinsics.ll

 ; CHECK-NEXT:    vpaddq %ymm0, %ymm2, %ymm0
 ; CHECK-NEXT:    retq
   %res = call <4 x i64> @llvm.x86.avx512.mask.valign.q.256(<4 x i64> %x0, <4 x i64> %x1, i32 22, <4 x i64> %x3, i8 %x4)
   %res1 = call <4 x i64> @llvm.x86.avx512.mask.valign.q.256(<4 x i64> %x0, <4 x i64> %x1, i32 22, <4 x i64> %x3, i8 -1)
   %res2 = add <4 x i64> %res, %res1
   ret <4 x i64> %res2
 }
 
+declare <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double>, i32, <4 x double>, i8)
+
+define <4 x double>@test_int_x86_avx512_mask_vpermil_pd_256(<4 x double> %x0, <4 x double> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_pd_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %dil, %eax
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpermilpd $22, %ymm0, %ymm1 {%k1}
+; CHECK-NEXT:    vpermilpd $22, %ymm0, %ymm2 {%k1} {z}
+; CHECK-NEXT:    vpermilpd $22, %ymm0, %ymm0
+; CHECK-NEXT:    vaddpd %ymm2, %ymm1, %ymm1
+; CHECK-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retq
+  %res = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> %x2, i8 %x3)
+  %res1 = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> zeroinitializer, i8 %x3)
+  %res2 = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> %x2, i8 -1)
+  %res3 = fadd <4 x double> %res, %res1
+  %res4 = fadd <4 x double> %res2, %res3
+  ret <4 x double> %res4
+}
+
+declare <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double>, i32, <2 x double>, i8)
+
+define <2 x double>@test_int_x86_avx512_mask_vpermil_pd_128(<2 x double> %x0, <2 x double> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_pd_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %dil, %eax
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpermilpd $1, %xmm0, %xmm1 {%k1}
+; CHECK-NEXT:    vpermilpd $1, %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT:    vpermilpd $1, %xmm0, %xmm0
+; CHECK-NEXT:    vaddpd %xmm2, %xmm1, %xmm1
+; CHECK-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; CHECK-NEXT:    retq
+  %res = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> %x2, i8 %x3)
+  %res1 = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> zeroinitializer, i8 %x3)
+  %res2 = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> %x2, i8 -1)
+  %res3 = fadd <2 x double> %res, %res1
+  %res4 = fadd <2 x double> %res3, %res2
+  ret <2 x double> %res4
+}
+
+declare <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float>, i32, <8 x float>, i8)
+
+define <8 x float>@test_int_x86_avx512_mask_vpermil_ps_256(<8 x float> %x0, <8 x float> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_ps_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %dil, %eax
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpermilps $22, %ymm0, %ymm1 {%k1}
+; CHECK-NEXT:    vpermilps $22, %ymm0, %ymm2 {%k1} {z}
+; CHECK-NEXT:    vpermilps $22, %ymm0, %ymm0
+; CHECK-NEXT:    vaddps %ymm2, %ymm1, %ymm1
+; CHECK-NEXT:    vaddps %ymm0, %ymm1, %ymm0
+; CHECK-NEXT:    retq
+  %res = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> %x2, i8 %x3)
+  %res1 = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> zeroinitializer, i8 %x3)
+  %res2 = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> %x2, i8 -1)
+  %res3 = fadd <8 x float> %res, %res1
+  %res4 = fadd <8 x float> %res3, %res2
+  ret <8 x float> %res4
+}
+
+declare <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float>, i32, <4 x float>, i8)
+
+define <4 x float>@test_int_x86_avx512_mask_vpermil_ps_128(<4 x float> %x0, <4 x float> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_vpermil_ps_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %dil, %eax
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpermilps $22, %xmm0, %xmm1 {%k1}
+; CHECK-NEXT:    vpermilps $22, %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT:    vpermilps $22, %xmm0, %xmm0
+; CHECK-NEXT:    vaddps %xmm2, %xmm1, %xmm1
+; CHECK-NEXT:    vaddps %xmm1, %xmm0, %xmm0
+; CHECK-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> %x2, i8 %x3)
+  %res1 = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> zeroinitializer, i8 %x3)
+  %res2 = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> %x2, i8 -1)
+  %res3 = fadd <4 x float> %res, %res1
+  %res4 = fadd <4 x float> %res2, %res3
+  ret <4 x float> %res4
+}
+
+declare <4 x double> @llvm.x86.avx512.mask.vpermilvar.pd.256(<4 x double>, <4 x i64>, <4 x double>, i8)
+
+define <4 x double>@test_int_x86_avx512_mask_vpermilvar_pd_256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_vpermilvar_pd_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %dil, %eax
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpermilpd %ymm1, %ymm0, %ymm2 {%k1}
+; CHECK-NEXT:    vpermilpd %ymm1, %ymm0, %ymm3 {%k1} {z}
+; CHECK-NEXT:    vpermilpd %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    vaddpd %ymm3, %ymm2, %ymm1
+; CHECK-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retq
+  %res = call <4 x double> @llvm.x86.avx512.mask.vpermilvar.pd.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 %x3)
+  %res1 = call <4 x double> @llvm.x86.avx512.mask.vpermilvar.pd.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> zeroinitializer, i8 %x3)
+  %res2 = call <4 x double> @llvm.x86.avx512.mask.vpermilvar.pd.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 -1)
+  %res3 = fadd <4 x double> %res, %res1
+  %res4 = fadd <4 x double> %res2, %res3
+  ret <4 x double> %res4
+}
+
+declare <2 x double> @llvm.x86.avx512.mask.vpermilvar.pd.128(<2 x double>, <2 x i64>, <2 x double>, i8)
+
+define <2 x double>@test_int_x86_avx512_mask_vpermilvar_pd_128(<2 x double> %x0, <2 x i64> %x1, <2 x double> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_vpermilvar_pd_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %dil, %eax
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpermilpd %xmm1, %xmm0, %xmm2 {%k1}
+; CHECK-NEXT:    vpermilpd %xmm1, %xmm0, %xmm3 {%k1} {z}
+; CHECK-NEXT:    vpermilpd %xmm1, %xmm0, %xmm0
+; CHECK-NEXT:    vaddpd %xmm3, %xmm2, %xmm1
+; CHECK-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; CHECK-NEXT:    retq
+  %res = call <2 x double> @llvm.x86.avx512.mask.vpermilvar.pd.128(<2 x double> %x0, <2 x i64> %x1, <2 x double> %x2, i8 %x3)
+  %res1 = call <2 x double> @llvm.x86.avx512.mask.vpermilvar.pd.128(<2 x double> %x0, <2 x i64> %x1, <2 x double> zeroinitializer, i8 %x3)
+  %res2 = call <2 x double> @llvm.x86.avx512.mask.vpermilvar.pd.128(<2 x double> %x0, <2 x i64> %x1, <2 x double> %x2, i8 -1)
+  %res3 = fadd <2 x double> %res, %res1
+  %res4 = fadd <2 x double> %res3, %res2
+  ret <2 x double> %res4
+}
+
+declare <8 x float> @llvm.x86.avx512.mask.vpermilvar.ps.256(<8 x float>, <8 x i32>, <8 x float>, i8)
+
+define <8 x float>@test_int_x86_avx512_mask_vpermilvar_ps_256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_vpermilvar_ps_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %dil, %eax
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpermilps %ymm1, %ymm0, %ymm2 {%k1}
+; CHECK-NEXT:    vpermilps %ymm1, %ymm0, %ymm3 {%k1} {z}
+; CHECK-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    vaddps %ymm3, %ymm2, %ymm1
+; CHECK-NEXT:    vaddps %ymm0, %ymm1, %ymm0
+; CHECK-NEXT:    retq
+  %res = call <8 x float> @llvm.x86.avx512.mask.vpermilvar.ps.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2, i8 %x3)
+  %res1 = call <8 x float> @llvm.x86.avx512.mask.vpermilvar.ps.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> zeroinitializer, i8 %x3)
+  %res2 = call <8 x float> @llvm.x86.avx512.mask.vpermilvar.ps.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2, i8 -1)
+  %res3 = fadd <8 x float> %res, %res1
+  %res4 = fadd <8 x float> %res3, %res2
+  ret <8 x float> %res4
+}
+
+declare <4 x float> @llvm.x86.avx512.mask.vpermilvar.ps.128(<4 x float>, <4 x i32>, <4 x float>, i8)
+
+define <4 x float>@test_int_x86_avx512_mask_vpermilvar_ps_128(<4 x float> %x0, <4 x i32> %x1, <4 x float> %x2, i8 %x3) {
+; CHECK-LABEL: test_int_x86_avx512_mask_vpermilvar_ps_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    movzbl %dil, %eax
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpermilps %xmm1, %xmm0, %xmm2 {%k1}
+; CHECK-NEXT:    vpermilps %xmm1, %xmm0, %xmm3 {%k1} {z}
+; CHECK-NEXT:    vpermilps %xmm1, %xmm0, %xmm0
+; CHECK-NEXT:    vaddps %xmm3, %xmm2, %xmm1
+; CHECK-NEXT:    vaddps %xmm1, %xmm0, %xmm0
+; CHECK-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.avx512.mask.vpermilvar.ps.128(<4 x float> %x0, <4 x i32> %x1, <4 x float> %x2, i8 %x3)
+  %res1 = call <4 x float> @llvm.x86.avx512.mask.vpermilvar.ps.128(<4 x float> %x0, <4 x i32> %x1, <4 x float> zeroinitializer, i8 %x3)
+  %res2 = call <4 x float> @llvm.x86.avx512.mask.vpermilvar.ps.128(<4 x float> %x0, <4 x i32> %x1, <4 x float> %x2, i8 -1)
+  %res3 = fadd <4 x float> %res, %res1
+  %res4 = fadd <4 x float> %res2, %res3
+  ret <4 x float> %res4
+}
+
+

test/CodeGen/X86/vector-shuffle-128-v2.ll

 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+sse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE3
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+ssse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSSE3
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+sse4.1 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE41
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=knl -mattr=+avx512vl | FileCheck %s --check-prefix=AVX512VL
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-unknown-unknown"
 
 define <2 x i64> @shuffle_v2i64_00(<2 x i64> %a, <2 x i64> %b) {
 ; SSE-LABEL: shuffle_v2i64_00:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]

 ; SSE:       # BB#0:
 ; SSE-NEXT:    shufpd {{.*#+}} xmm0 = xmm0[1,0]
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: shuffle_v2f64_10:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
 ; AVX-NEXT:    retq
+
+; AVX512VL-LABEL: shuffle_v2f64_10:
+; AVX512VL:       # BB#0:
+; AVX512VL-NEXT:    vpermilpd $1, %xmm0, %xmm0
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <2 x double> %a, <2 x double> %b, <2 x i32> <i32 1, i32 0>
   ret <2 x double> %shuffle
 }
 define <2 x double> @shuffle_v2f64_11(<2 x double> %a, <2 x double> %b) {
 ; SSE-LABEL: shuffle_v2f64_11:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
 ; SSE-NEXT:    retq

 ; SSE-NEXT:    shufpd {{.*#+}} xmm1 = xmm1[1,0]
 ; SSE-NEXT:    movapd %xmm1, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: shuffle_v2f64_32:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm1[1,0]
 ; AVX-NEXT:    retq
+
+; AVX512VL-LABEL: shuffle_v2f64_32:
+; AVX512VL:       # BB#0:
+; AVX512VL-NEXT:    vpermilpd $1, %xmm1, %xmm0
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <2 x double> %a, <2 x double> %b, <2 x i32> <i32 3, i32 2>
   ret <2 x double> %shuffle
 }
 define <2 x double> @shuffle_v2f64_33(<2 x double> %a, <2 x double> %b) {
 ; SSE-LABEL: shuffle_v2f64_33:
 ; SSE:       # BB#0:
 ; SSE-NEXT:    movhlps {{.*#+}} xmm1 = xmm1[1,1]
 ; SSE-NEXT:    movaps %xmm1, %xmm0

 ; SSE-NEXT:    movapd (%rdi), %xmm0
 ; SSE-NEXT:    shufpd {{.*#+}} xmm0 = xmm0[1,0]
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: shuffle_mem_v2f64_10:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm0 = mem[1,0]
 ; AVX-NEXT:    retq
+
+; AVX512VL-LABEL: shuffle_mem_v2f64_10:
+; AVX512VL:       # BB#0:
+; AVX512VL-NEXT:    vpermilpd $1, (%rdi), %xmm0
+; AVX512VL-NEXT:    retq
   %a = load <2 x double>, <2 x double>* %ptr
   %shuffle = shufflevector <2 x double> %a, <2 x double> undef, <2 x i32> <i32 1, i32 0>
   ret <2 x double> %shuffle
 }

test/CodeGen/X86/vector-shuffle-256-v4.ll

 ; RUN: llc < %s -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
 ; RUN: llc < %s -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2
+; RUN: llc < %s -mcpu=knl -mattr=+avx512vl | FileCheck %s --check-prefix=AVX512VL
 
 target triple = "x86_64-unknown-unknown"
 
 define <4 x double> @shuffle_v4f64_0000(<4 x double> %a, <4 x double> %b) {
 ; AVX1-LABEL: shuffle_v4f64_0000:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-NEXT:    retq
   ret <4 x double> %shuffle
 }
 
 define <4 x double> @shuffle_v4f64_0023(<4 x double> %a, <4 x double> %b) {
 ; ALL-LABEL: shuffle_v4f64_0023:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
 ; ALL-NEXT:    retq
+
+; AVX512VL-LABEL: shuffle_v4f64_0023:
+; AVX512VL:       # BB#0:
+; AVX512VL-NEXT:    vpermilpd $8, %ymm0, %ymm0
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 0, i32 2, i32 3>
   ret <4 x double> %shuffle
 }
 
 define <4 x double> @shuffle_v4f64_0022(<4 x double> %a, <4 x double> %b) {
 ; ALL-LABEL: shuffle_v4f64_0022:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]

test/MC/X86/avx512-encodings.s

 // CHECK: vscatterdpd %zmm18, 256(%r9,%ymm24) {%k1}
 // CHECK:  encoding: [0x62,0x82,0xfd,0x41,0xa2,0x54,0x01,0x20]
           vscatterdpd %zmm18, 256(%r9, %ymm24) {%k1}
 
 // CHECK: vscatterdpd %zmm18, 1024(%rcx,%ymm24,4) {%k1}
 // CHECK:  encoding: [0x62,0xa2,0xfd,0x41,0xa2,0x94,0x81,0x00,0x04,0x00,0x00]
           vscatterdpd %zmm18, 1024(%rcx, %ymm24,4) {%k1}
 
+// CHECK: vpermilps $171, %zmm22, %zmm2
+// CHECK:  encoding: [0x62,0xb3,0x7d,0x48,0x04,0xd6,0xab]
+          vpermilps $0xab, %zmm22, %zmm2
+
+// CHECK: vpermilps $171, %zmm22, %zmm2 {%k2}
+// CHECK:  encoding: [0x62,0xb3,0x7d,0x4a,0x04,0xd6,0xab]
+          vpermilps $0xab, %zmm22, %zmm2 {%k2}
+
+// CHECK: vpermilps $171, %zmm22, %zmm2 {%k2} {z}
+// CHECK:  encoding: [0x62,0xb3,0x7d,0xca,0x04,0xd6,0xab]
+          vpermilps $0xab, %zmm22, %zmm2 {%k2} {z}
+
+// CHECK: vpermilps $123, %zmm22, %zmm2
+// CHECK:  encoding: [0x62,0xb3,0x7d,0x48,0x04,0xd6,0x7b]
+          vpermilps $0x7b, %zmm22, %zmm2
+
+// CHECK: vpermilps $123, (%rcx), %zmm2
+// CHECK:  encoding: [0x62,0xf3,0x7d,0x48,0x04,0x11,0x7b]
+          vpermilps $0x7b, (%rcx), %zmm2
+
+// CHECK: vpermilps $123, 291(%rax,%r14,8), %zmm2
+// CHECK:  encoding: [0x62,0xb3,0x7d,0x48,0x04,0x94,0xf0,0x23,0x01,0x00,0x00,0x7b]
+          vpermilps $0x7b, 291(%rax,%r14,8), %zmm2
+
+// CHECK: vpermilps $123, (%rcx){1to16}, %zmm2
+// CHECK:  encoding: [0x62,0xf3,0x7d,0x58,0x04,0x11,0x7b]
+          vpermilps $0x7b, (%rcx){1to16}, %zmm2
+
+// CHECK: vpermilps $123, 8128(%rdx), %zmm2
+// CHECK:  encoding: [0x62,0xf3,0x7d,0x48,0x04,0x52,0x7f,0x7b]
+          vpermilps $0x7b, 8128(%rdx), %zmm2
+
+// CHECK: vpermilps $123, 8192(%rdx), %zmm2
+// CHECK:  encoding: [0x62,0xf3,0x7d,0x48,0x04,0x92,0x00,0x20,0x00,0x00,0x7b]
+          vpermilps $0x7b, 8192(%rdx), %zmm2
+
+// CHECK: vpermilps $123, -8192(%rdx), %zmm2
+// CHECK:  encoding: [0x62,0xf3,0x7d,0x48,0x04,0x52,0x80,0x7b]
+          vpermilps $0x7b, -8192(%rdx), %zmm2
+
+// CHECK: vpermilps $123, -8256(%rdx), %zmm2
+// CHECK:  encoding: [0x62,0xf3,0x7d,0x48,0x04,0x92,0xc0,0xdf,0xff,0xff,0x7b]
+          vpermilps $0x7b, -8256(%rdx), %zmm2
+
+// CHECK: vpermilps $123, 508(%rdx){1to16}, %zmm2
+// CHECK:  encoding: [0x62,0xf3,0x7d,0x58,0x04,0x52,0x7f,0x7b]
+          vpermilps $0x7b, 508(%rdx){1to16}, %zmm2
+
+// CHECK: vpermilps $123, 512(%rdx){1to16}, %zmm2
+// CHECK:  encoding: [0x62,0xf3,0x7d,0x58,0x04,0x92,0x00,0x02,0x00,0x00,0x7b]
+          vpermilps $0x7b, 512(%rdx){1to16}, %zmm2
+
+// CHECK: vpermilps $123, -512(%rdx){1to16}, %zmm2
+// CHECK:  encoding: [0x62,0xf3,0x7d,0x58,0x04,0x52,0x80,0x7b]
+          vpermilps $0x7b, -512(%rdx){1to16}, %zmm2
+
+// CHECK: vpermilps $123, -516(%rdx){1to16}, %zmm2
+// CHECK:  encoding: [0x62,0xf3,0x7d,0x58,0x04,0x92,0xfc,0xfd,0xff,0xff,0x7b]
+          vpermilps $0x7b, -516(%rdx){1to16}, %zmm2
+
+// CHECK: vpermilps %zmm2, %zmm20, %zmm13
+// CHECK:  encoding: [0x62,0x72,0x5d,0x40,0x0c,0xea]
+          vpermilps %zmm2, %zmm20, %zmm13
+
+// CHECK: vpermilps %zmm2, %zmm20, %zmm13 {%k1}
+// CHECK:  encoding: [0x62,0x72,0x5d,0x41,0x0c,0xea]
+          vpermilps %zmm2, %zmm20, %zmm13 {%k1}
+
+// CHECK: vpermilps %zmm2, %zmm20, %zmm13 {%k1} {z}
+// CHECK:  encoding: [0x62,0x72,0x5d,0xc1,0x0c,0xea]
+          vpermilps %zmm2, %zmm20, %zmm13 {%k1} {z}
+
+// CHECK: vpermilps (%rcx), %zmm20, %zmm13
+// CHECK:  encoding: [0x62,0x72,0x5d,0x40,0x0c,0x29]
+          vpermilps (%rcx), %zmm20, %zmm13
+
+// CHECK: vpermilps 291(%rax,%r14,8), %zmm20, %zmm13
+// CHECK:  encoding: [0x62,0x32,0x5d,0x40,0x0c,0xac,0xf0,0x23,0x01,0x00,0x00]
+          vpermilps 291(%rax,%r14,8), %zmm20, %zmm13
+
+// CHECK: vpermilps (%rcx){1to16}, %zmm20, %zmm13
+// CHECK:  encoding: [0x62,0x72,0x5d,0x50,0x0c,0x29]
+          vpermilps (%rcx){1to16}, %zmm20, %zmm13
+
+// CHECK: vpermilps 8128(%rdx), %zmm20, %zmm13
+// CHECK:  encoding: [0x62,0x72,0x5d,0x40,0x0c,0x6a,0x7f]
+          vpermilps 8128(%rdx), %zmm20, %zmm13
+
+// CHECK: vpermilps 8192(%rdx), %zmm20, %zmm13
+// CHECK:  encoding: [0x62,0x72,0x5d,0x40,0x0c,0xaa,0x00,0x20,0x00,0x00]
+          vpermilps 8192(%rdx), %zmm20, %zmm13
+
+// CHECK: vpermilps -8192(%rdx), %zmm20, %zmm13
+// CHECK:  encoding: [0x62,0x72,0x5d,0x40,0x0c,0x6a,0x80]
+          vpermilps -8192(%rdx), %zmm20, %zmm13
+
+// CHECK: vpermilps -8256(%rdx), %zmm20, %zmm13
+// CHECK:  encoding: [0x62,0x72,0x5d,0x40,0x0c,0xaa,0xc0,0xdf,0xff,0xff]
+          vpermilps -8256(%rdx), %zmm20, %zmm13
+
+// CHECK: vpermilps 508(%rdx){1to16}, %zmm20, %zmm13
+// CHECK:  encoding: [0x62,0x72,0x5d,0x50,0x0c,0x6a,0x7f]
+          vpermilps 508(%rdx){1to16}, %zmm20, %zmm13
+
+// CHECK: vpermilps 512(%rdx){1to16}, %zmm20, %zmm13
+// CHECK:  encoding: [0x62,0x72,0x5d,0x50,0x0c,0xaa,0x00,0x02,0x00,0x00]
+          vpermilps 512(%rdx){1to16}, %zmm20, %zmm13
+
+// CHECK: vpermilps -512(%rdx){1to16}, %zmm20, %zmm13
+// CHECK:  encoding: [0x62,0x72,0x5d,0x50,0x0c,0x6a,0x80]
+          vpermilps -512(%rdx){1to16}, %zmm20, %zmm13
+
+// CHECK: vpermilps -516(%rdx){1to16}, %zmm20, %zmm13
+// CHECK:  encoding: [0x62,0x72,0x5d,0x50,0x0c,0xaa,0xfc,0xfd,0xff,0xff]
+          vpermilps -516(%rdx){1to16}, %zmm20, %zmm13
+
+// CHECK: vpermilpd $171, %zmm4, %zmm19
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x48,0x05,0xdc,0xab]
+          vpermilpd $0xab, %zmm4, %zmm19
+
+// CHECK: vpermilpd $171, %zmm4, %zmm19 {%k1}
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x49,0x05,0xdc,0xab]
+          vpermilpd $0xab, %zmm4, %zmm19 {%k1}
+
+// CHECK: vpermilpd $171, %zmm4, %zmm19 {%k1} {z}
+// CHECK:  encoding: [0x62,0xe3,0xfd,0xc9,0x05,0xdc,0xab]
+          vpermilpd $0xab, %zmm4, %zmm19 {%k1} {z}
+
+// CHECK: vpermilpd $123, %zmm4, %zmm19
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x48,0x05,0xdc,0x7b]
+          vpermilpd $0x7b, %zmm4, %zmm19
+
+// CHECK: vpermilpd $123, (%rcx), %zmm19
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x48,0x05,0x19,0x7b]
+          vpermilpd $0x7b, (%rcx), %zmm19
+
+// CHECK: vpermilpd $123, 291(%rax,%r14,8), %zmm19
+// CHECK:  encoding: [0x62,0xa3,0xfd,0x48,0x05,0x9c,0xf0,0x23,0x01,0x00,0x00,0x7b]
+          vpermilpd $0x7b, 291(%rax,%r14,8), %zmm19
+
+// CHECK: vpermilpd $123, (%rcx){1to8}, %zmm19
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x58,0x05,0x19,0x7b]
+          vpermilpd $0x7b, (%rcx){1to8}, %zmm19
+
+// CHECK: vpermilpd $123, 8128(%rdx), %zmm19
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x48,0x05,0x5a,0x7f,0x7b]
+          vpermilpd $0x7b, 8128(%rdx), %zmm19
+
+// CHECK: vpermilpd $123, 8192(%rdx), %zmm19
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x48,0x05,0x9a,0x00,0x20,0x00,0x00,0x7b]
+          vpermilpd $0x7b, 8192(%rdx), %zmm19
+
+// CHECK: vpermilpd $123, -8192(%rdx), %zmm19
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x48,0x05,0x5a,0x80,0x7b]
+          vpermilpd $0x7b, -8192(%rdx), %zmm19
+
+// CHECK: vpermilpd $123, -8256(%rdx), %zmm19
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x48,0x05,0x9a,0xc0,0xdf,0xff,0xff,0x7b]
+          vpermilpd $0x7b, -8256(%rdx), %zmm19
+
+// CHECK: vpermilpd $123, 1016(%rdx){1to8}, %zmm19
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x58,0x05,0x5a,0x7f,0x7b]
+          vpermilpd $0x7b, 1016(%rdx){1to8}, %zmm19
+
+// CHECK: vpermilpd $123, 1024(%rdx){1to8}, %zmm19
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x58,0x05,0x9a,0x00,0x04,0x00,0x00,0x7b]
+          vpermilpd $0x7b, 1024(%rdx){1to8}, %zmm19
+
+// CHECK: vpermilpd $123, -1024(%rdx){1to8}, %zmm19
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x58,0x05,0x5a,0x80,0x7b]
+          vpermilpd $0x7b, -1024(%rdx){1to8}, %zmm19
+
+// CHECK: vpermilpd $123, -1032(%rdx){1to8}, %zmm19
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x58,0x05,0x9a,0xf8,0xfb,0xff,0xff,0x7b]
+          vpermilpd $0x7b, -1032(%rdx){1to8}, %zmm19
+
+// CHECK: vpermilpd %zmm21, %zmm26, %zmm1
+// CHECK:  encoding: [0x62,0xb2,0xad,0x40,0x0d,0xcd]
+          vpermilpd %zmm21, %zmm26, %zmm1
+
+// CHECK: vpermilpd %zmm21, %zmm26, %zmm1 {%k2}
+// CHECK:  encoding: [0x62,0xb2,0xad,0x42,0x0d,0xcd]
+          vpermilpd %zmm21, %zmm26, %zmm1 {%k2}
+
+// CHECK: vpermilpd %zmm21, %zmm26, %zmm1 {%k2} {z}
+// CHECK:  encoding: [0x62,0xb2,0xad,0xc2,0x0d,0xcd]
+          vpermilpd %zmm21, %zmm26, %zmm1 {%k2} {z}
+
+// CHECK: vpermilpd (%rcx), %zmm26, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0xad,0x40,0x0d,0x09]
+          vpermilpd (%rcx), %zmm26, %zmm1
+
+// CHECK: vpermilpd 291(%rax,%r14,8), %zmm26, %zmm1
+// CHECK:  encoding: [0x62,0xb2,0xad,0x40,0x0d,0x8c,0xf0,0x23,0x01,0x00,0x00]
+          vpermilpd 291(%rax,%r14,8), %zmm26, %zmm1
+
+// CHECK: vpermilpd (%rcx){1to8}, %zmm26, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0xad,0x50,0x0d,0x09]
+          vpermilpd (%rcx){1to8}, %zmm26, %zmm1
+
+// CHECK: vpermilpd 8128(%rdx), %zmm26, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0xad,0x40,0x0d,0x4a,0x7f]
+          vpermilpd 8128(%rdx), %zmm26, %zmm1
+
+// CHECK: vpermilpd 8192(%rdx), %zmm26, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0xad,0x40,0x0d,0x8a,0x00,0x20,0x00,0x00]
+          vpermilpd 8192(%rdx), %zmm26, %zmm1
+
+// CHECK: vpermilpd -8192(%rdx), %zmm26, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0xad,0x40,0x0d,0x4a,0x80]
+          vpermilpd -8192(%rdx), %zmm26, %zmm1
+
+// CHECK: vpermilpd -8256(%rdx), %zmm26, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0xad,0x40,0x0d,0x8a,0xc0,0xdf,0xff,0xff]
+          vpermilpd -8256(%rdx), %zmm26, %zmm1
+
+// CHECK: vpermilpd 1016(%rdx){1to8}, %zmm26, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0xad,0x50,0x0d,0x4a,0x7f]
+          vpermilpd 1016(%rdx){1to8}, %zmm26, %zmm1
+
+// CHECK: vpermilpd 1024(%rdx){1to8}, %zmm26, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0xad,0x50,0x0d,0x8a,0x00,0x04,0x00,0x00]
+          vpermilpd 1024(%rdx){1to8}, %zmm26, %zmm1
+
+// CHECK: vpermilpd -1024(%rdx){1to8}, %zmm26, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0xad,0x50,0x0d,0x4a,0x80]
+          vpermilpd -1024(%rdx){1to8}, %zmm26, %zmm1
+
+// CHECK: vpermilpd -1032(%rdx){1to8}, %zmm26, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0xad,0x50,0x0d,0x8a,0xf8,0xfb,0xff,0xff]
+          vpermilpd -1032(%rdx){1to8}, %zmm26, %zmm1
+

test/MC/X86/x86-64-avx512f_vl.s

 // CHECK: vscatterdpd %ymm30, 256(%r9,%xmm31) {%k1}
 // CHECK:  encoding: [0x62,0x02,0xfd,0x21,0xa2,0x74,0x39,0x20]
           vscatterdpd %ymm30, 256(%r9, %xmm31) {%k1}
 
 // CHECK: vscatterdpd %ymm30, 1024(%rcx,%xmm31,4) {%k1}
 // CHECK:  encoding: [0x62,0x22,0xfd,0x21,0xa2,0xb4,0xb9,0x00,0x04,0x00,0x00]
           vscatterdpd %ymm30, 1024(%rcx, %xmm31,4) {%k1}
 
+// CHECK: vpermilps $171, %xmm28, %xmm20
+// CHECK:  encoding: [0x62,0x83,0x7d,0x08,0x04,0xe4,0xab]
+          vpermilps $0xab, %xmm28, %xmm20
+
+// CHECK: vpermilps $171, %xmm28, %xmm20 {%k4}
+// CHECK:  encoding: [0x62,0x83,0x7d,0x0c,0x04,0xe4,0xab]
+          vpermilps $0xab, %xmm28, %xmm20 {%k4}
+
+// CHECK: vpermilps $171, %xmm28, %xmm20 {%k4} {z}
+// CHECK:  encoding: [0x62,0x83,0x7d,0x8c,0x04,0xe4,0xab]
+          vpermilps $0xab, %xmm28, %xmm20 {%k4} {z}
+
+// CHECK: vpermilps $123, %xmm28, %xmm20
+// CHECK:  encoding: [0x62,0x83,0x7d,0x08,0x04,0xe4,0x7b]
+          vpermilps $0x7b, %xmm28, %xmm20
+
+// CHECK: vpermilps $123, (%rcx), %xmm20
+// CHECK:  encoding: [0x62,0xe3,0x7d,0x08,0x04,0x21,0x7b]
+          vpermilps $0x7b, (%rcx), %xmm20
+
+// CHECK: vpermilps $123, 291(%rax,%r14,8), %xmm20
+// CHECK:  encoding: [0x62,0xa3,0x7d,0x08,0x04,0xa4,0xf0,0x23,0x01,0x00,0x00,0x7b]
+          vpermilps $0x7b, 291(%rax,%r14,8), %xmm20
+
+// CHECK: vpermilps $123, (%rcx){1to4}, %xmm20
+// CHECK:  encoding: [0x62,0xe3,0x7d,0x18,0x04,0x21,0x7b]
+          vpermilps $0x7b, (%rcx){1to4}, %xmm20
+
+// CHECK: vpermilps $123, 2032(%rdx), %xmm20
+// CHECK:  encoding: [0x62,0xe3,0x7d,0x08,0x04,0x62,0x7f,0x7b]
+          vpermilps $0x7b, 2032(%rdx), %xmm20
+
+// CHECK: vpermilps $123, 2048(%rdx), %xmm20
+// CHECK:  encoding: [0x62,0xe3,0x7d,0x08,0x04,0xa2,0x00,0x08,0x00,0x00,0x7b]
+          vpermilps $0x7b, 2048(%rdx), %xmm20
+
+// CHECK: vpermilps $123, -2048(%rdx), %xmm20
+// CHECK:  encoding: [0x62,0xe3,0x7d,0x08,0x04,0x62,0x80,0x7b]
+          vpermilps $0x7b, -2048(%rdx), %xmm20
+
+// CHECK: vpermilps $123, -2064(%rdx), %xmm20
+// CHECK:  encoding: [0x62,0xe3,0x7d,0x08,0x04,0xa2,0xf0,0xf7,0xff,0xff,0x7b]
+          vpermilps $0x7b, -2064(%rdx), %xmm20
+
+// CHECK: vpermilps $123, 508(%rdx){1to4}, %xmm20
+// CHECK:  encoding: [0x62,0xe3,0x7d,0x18,0x04,0x62,0x7f,0x7b]
+          vpermilps $0x7b, 508(%rdx){1to4}, %xmm20
+
+// CHECK: vpermilps $123, 512(%rdx){1to4}, %xmm20
+// CHECK:  encoding: [0x62,0xe3,0x7d,0x18,0x04,0xa2,0x00,0x02,0x00,0x00,0x7b]
+          vpermilps $0x7b, 512(%rdx){1to4}, %xmm20
+
+// CHECK: vpermilps $123, -512(%rdx){1to4}, %xmm20
+// CHECK:  encoding: [0x62,0xe3,0x7d,0x18,0x04,0x62,0x80,0x7b]
+          vpermilps $0x7b, -512(%rdx){1to4}, %xmm20
+
+// CHECK: vpermilps $123, -516(%rdx){1to4}, %xmm20
+// CHECK:  encoding: [0x62,0xe3,0x7d,0x18,0x04,0xa2,0xfc,0xfd,0xff,0xff,0x7b]
+          vpermilps $0x7b, -516(%rdx){1to4}, %xmm20
+
+// CHECK: vpermilps $171, %ymm17, %ymm30
+// CHECK:  encoding: [0x62,0x23,0x7d,0x28,0x04,0xf1,0xab]
+          vpermilps $0xab, %ymm17, %ymm30
+
+// CHECK: vpermilps $171, %ymm17, %ymm30 {%k5}
+// CHECK:  encoding: [0x62,0x23,0x7d,0x2d,0x04,0xf1,0xab]
+          vpermilps $0xab, %ymm17, %ymm30 {%k5}
+
+// CHECK: vpermilps $171, %ymm17, %ymm30 {%k5} {z}
+// CHECK:  encoding: [0x62,0x23,0x7d,0xad,0x04,0xf1,0xab]
+          vpermilps $0xab, %ymm17, %ymm30 {%k5} {z}
+
+// CHECK: vpermilps $123, %ymm17, %ymm30
+// CHECK:  encoding: [0x62,0x23,0x7d,0x28,0x04,0xf1,0x7b]
+          vpermilps $0x7b, %ymm17, %ymm30
+
+// CHECK: vpermilps $123, (%rcx), %ymm30
+// CHECK:  encoding: [0x62,0x63,0x7d,0x28,0x04,0x31,0x7b]
+          vpermilps $0x7b, (%rcx), %ymm30
+
+// CHECK: vpermilps $123, 291(%rax,%r14,8), %ymm30
+// CHECK:  encoding: [0x62,0x23,0x7d,0x28,0x04,0xb4,0xf0,0x23,0x01,0x00,0x00,0x7b]
+          vpermilps $0x7b, 291(%rax,%r14,8), %ymm30
+
+// CHECK: vpermilps $123, (%rcx){1to8}, %ymm30
+// CHECK:  encoding: [0x62,0x63,0x7d,0x38,0x04,0x31,0x7b]
+          vpermilps $0x7b, (%rcx){1to8}, %ymm30
+
+// CHECK: vpermilps $123, 4064(%rdx), %ymm30
+// CHECK:  encoding: [0x62,0x63,0x7d,0x28,0x04,0x72,0x7f,0x7b]
+          vpermilps $0x7b, 4064(%rdx), %ymm30
+
+// CHECK: vpermilps $123, 4096(%rdx), %ymm30
+// CHECK:  encoding: [0x62,0x63,0x7d,0x28,0x04,0xb2,0x00,0x10,0x00,0x00,0x7b]
+          vpermilps $0x7b, 4096(%rdx), %ymm30
+
+// CHECK: vpermilps $123, -4096(%rdx), %ymm30
+// CHECK:  encoding: [0x62,0x63,0x7d,0x28,0x04,0x72,0x80,0x7b]
+          vpermilps $0x7b, -4096(%rdx), %ymm30
+
+// CHECK: vpermilps $123, -4128(%rdx), %ymm30
+// CHECK:  encoding: [0x62,0x63,0x7d,0x28,0x04,0xb2,0xe0,0xef,0xff,0xff,0x7b]
+          vpermilps $0x7b, -4128(%rdx), %ymm30
+
+// CHECK: vpermilps $123, 508(%rdx){1to8}, %ymm30
+// CHECK:  encoding: [0x62,0x63,0x7d,0x38,0x04,0x72,0x7f,0x7b]
+          vpermilps $0x7b, 508(%rdx){1to8}, %ymm30
+
+// CHECK: vpermilps $123, 512(%rdx){1to8}, %ymm30
+// CHECK:  encoding: [0x62,0x63,0x7d,0x38,0x04,0xb2,0x00,0x02,0x00,0x00,0x7b]
+          vpermilps $0x7b, 512(%rdx){1to8}, %ymm30
+
+// CHECK: vpermilps $123, -512(%rdx){1to8}, %ymm30
+// CHECK:  encoding: [0x62,0x63,0x7d,0x38,0x04,0x72,0x80,0x7b]
+          vpermilps $0x7b, -512(%rdx){1to8}, %ymm30
+
+// CHECK: vpermilps $123, -516(%rdx){1to8}, %ymm30
+// CHECK:  encoding: [0x62,0x63,0x7d,0x38,0x04,0xb2,0xfc,0xfd,0xff,0xff,0x7b]
+          vpermilps $0x7b, -516(%rdx){1to8}, %ymm30
+
+// CHECK: vpermilps %xmm22, %xmm28, %xmm28
+// CHECK:  encoding: [0x62,0x22,0x1d,0x00,0x0c,0xe6]
+          vpermilps %xmm22, %xmm28, %xmm28
+
+// CHECK: vpermilps %xmm22, %xmm28, %xmm28 {%k6}
+// CHECK:  encoding: [0x62,0x22,0x1d,0x06,0x0c,0xe6]
+          vpermilps %xmm22, %xmm28, %xmm28 {%k6}
+
+// CHECK: vpermilps %xmm22, %xmm28, %xmm28 {%k6} {z}
+// CHECK:  encoding: [0x62,0x22,0x1d,0x86,0x0c,0xe6]
+          vpermilps %xmm22, %xmm28, %xmm28 {%k6} {z}
+
+// CHECK: vpermilps (%rcx), %xmm28, %xmm28
+// CHECK:  encoding: [0x62,0x62,0x1d,0x00,0x0c,0x21]
+          vpermilps (%rcx), %xmm28, %xmm28
+
+// CHECK: vpermilps 291(%rax,%r14,8), %xmm28, %xmm28
+// CHECK:  encoding: [0x62,0x22,0x1d,0x00,0x0c,0xa4,0xf0,0x23,0x01,0x00,0x00]
+          vpermilps 291(%rax,%r14,8), %xmm28, %xmm28
+
+// CHECK: vpermilps (%rcx){1to4}, %xmm28, %xmm28
+// CHECK:  encoding: [0x62,0x62,0x1d,0x10,0x0c,0x21]
+          vpermilps (%rcx){1to4}, %xmm28, %xmm28
+
+// CHECK: vpermilps 2032(%rdx), %xmm28, %xmm28
+// CHECK:  encoding: [0x62,0x62,0x1d,0x00,0x0c,0x62,0x7f]
+          vpermilps 2032(%rdx), %xmm28, %xmm28
+
+// CHECK: vpermilps 2048(%rdx), %xmm28, %xmm28
+// CHECK:  encoding: [0x62,0x62,0x1d,0x00,0x0c,0xa2,0x00,0x08,0x00,0x00]
+          vpermilps 2048(%rdx), %xmm28, %xmm28
+
+// CHECK: vpermilps -2048(%rdx), %xmm28, %xmm28
+// CHECK:  encoding: [0x62,0x62,0x1d,0x00,0x0c,0x62,0x80]
+          vpermilps -2048(%rdx), %xmm28, %xmm28
+
+// CHECK: vpermilps -2064(%rdx), %xmm28, %xmm28
+// CHECK:  encoding: [0x62,0x62,0x1d,0x00,0x0c,0xa2,0xf0,0xf7,0xff,0xff]
+          vpermilps -2064(%rdx), %xmm28, %xmm28
+
+// CHECK: vpermilps 508(%rdx){1to4}, %xmm28, %xmm28
+// CHECK:  encoding: [0x62,0x62,0x1d,0x10,0x0c,0x62,0x7f]
+          vpermilps 508(%rdx){1to4}, %xmm28, %xmm28
+
+// CHECK: vpermilps 512(%rdx){1to4}, %xmm28, %xmm28
+// CHECK:  encoding: [0x62,0x62,0x1d,0x10,0x0c,0xa2,0x00,0x02,0x00,0x00]
+          vpermilps 512(%rdx){1to4}, %xmm28, %xmm28
+
+// CHECK: vpermilps -512(%rdx){1to4}, %xmm28, %xmm28
+// CHECK:  encoding: [0x62,0x62,0x1d,0x10,0x0c,0x62,0x80]
+          vpermilps -512(%rdx){1to4}, %xmm28, %xmm28
+
+// CHECK: vpermilps -516(%rdx){1to4}, %xmm28, %xmm28
+// CHECK:  encoding: [0x62,0x62,0x1d,0x10,0x0c,0xa2,0xfc,0xfd,0xff,0xff]
+          vpermilps -516(%rdx){1to4}, %xmm28, %xmm28
+
+// CHECK: vpermilps %ymm21, %ymm28, %ymm29
+// CHECK:  encoding: [0x62,0x22,0x1d,0x20,0x0c,0xed]
+          vpermilps %ymm21, %ymm28, %ymm29
+
+// CHECK: vpermilps %ymm21, %ymm28, %ymm29 {%k2}
+// CHECK:  encoding: [0x62,0x22,0x1d,0x22,0x0c,0xed]
+          vpermilps %ymm21, %ymm28, %ymm29 {%k2}
+
+// CHECK: vpermilps %ymm21, %ymm28, %ymm29 {%k2} {z}
+// CHECK:  encoding: [0x62,0x22,0x1d,0xa2,0x0c,0xed]
+          vpermilps %ymm21, %ymm28, %ymm29 {%k2} {z}
+
+// CHECK: vpermilps (%rcx), %ymm28, %ymm29
+// CHECK:  encoding: [0x62,0x62,0x1d,0x20,0x0c,0x29]
+          vpermilps (%rcx), %ymm28, %ymm29
+
+// CHECK: vpermilps 291(%rax,%r14,8), %ymm28, %ymm29
+// CHECK:  encoding: [0x62,0x22,0x1d,0x20,0x0c,0xac,0xf0,0x23,0x01,0x00,0x00]
+          vpermilps 291(%rax,%r14,8), %ymm28, %ymm29
+
+// CHECK: vpermilps (%rcx){1to8}, %ymm28, %ymm29
+// CHECK:  encoding: [0x62,0x62,0x1d,0x30,0x0c,0x29]
+          vpermilps (%rcx){1to8}, %ymm28, %ymm29
+
+// CHECK: vpermilps 4064(%rdx), %ymm28, %ymm29
+// CHECK:  encoding: [0x62,0x62,0x1d,0x20,0x0c,0x6a,0x7f]
+          vpermilps 4064(%rdx), %ymm28, %ymm29
+
+// CHECK: vpermilps 4096(%rdx), %ymm28, %ymm29
+// CHECK:  encoding: [0x62,0x62,0x1d,0x20,0x0c,0xaa,0x00,0x10,0x00,0x00]
+          vpermilps 4096(%rdx), %ymm28, %ymm29
+
+// CHECK: vpermilps -4096(%rdx), %ymm28, %ymm29
+// CHECK:  encoding: [0x62,0x62,0x1d,0x20,0x0c,0x6a,0x80]
+          vpermilps -4096(%rdx), %ymm28, %ymm29
+
+// CHECK: vpermilps -4128(%rdx), %ymm28, %ymm29
+// CHECK:  encoding: [0x62,0x62,0x1d,0x20,0x0c,0xaa,0xe0,0xef,0xff,0xff]
+          vpermilps -4128(%rdx), %ymm28, %ymm29
+
+// CHECK: vpermilps 508(%rdx){1to8}, %ymm28, %ymm29
+// CHECK:  encoding: [0x62,0x62,0x1d,0x30,0x0c,0x6a,0x7f]
+          vpermilps 508(%rdx){1to8}, %ymm28, %ymm29
+
+// CHECK: vpermilps 512(%rdx){1to8}, %ymm28, %ymm29
+// CHECK:  encoding: [0x62,0x62,0x1d,0x30,0x0c,0xaa,0x00,0x02,0x00,0x00]
+          vpermilps 512(%rdx){1to8}, %ymm28, %ymm29
+
+// CHECK: vpermilps -512(%rdx){1to8}, %ymm28, %ymm29
+// CHECK:  encoding: [0x62,0x62,0x1d,0x30,0x0c,0x6a,0x80]
+          vpermilps -512(%rdx){1to8}, %ymm28, %ymm29
+
+// CHECK: vpermilps -516(%rdx){1to8}, %ymm28, %ymm29
+// CHECK:  encoding: [0x62,0x62,0x1d,0x30,0x0c,0xaa,0xfc,0xfd,0xff,0xff]
+          vpermilps -516(%rdx){1to8}, %ymm28, %ymm29
+
+// CHECK: vpermilpd $171, %xmm19, %xmm29
+// CHECK:  encoding: [0x62,0x23,0xfd,0x08,0x05,0xeb,0xab]
+          vpermilpd $0xab, %xmm19, %xmm29
+
+// CHECK: vpermilpd $171, %xmm19, %xmm29 {%k7}
+// CHECK:  encoding: [0x62,0x23,0xfd,0x0f,0x05,0xeb,0xab]
+          vpermilpd $0xab, %xmm19, %xmm29 {%k7}
+
+// CHECK: vpermilpd $171, %xmm19, %xmm29 {%k7} {z}
+// CHECK:  encoding: [0x62,0x23,0xfd,0x8f,0x05,0xeb,0xab]
+          vpermilpd $0xab, %xmm19, %xmm29 {%k7} {z}
+
+// CHECK: vpermilpd $123, %xmm19, %xmm29
+// CHECK:  encoding: [0x62,0x23,0xfd,0x08,0x05,0xeb,0x7b]
+          vpermilpd $0x7b, %xmm19, %xmm29
+
+// CHECK: vpermilpd $123, (%rcx), %xmm29
+// CHECK:  encoding: [0x62,0x63,0xfd,0x08,0x05,0x29,0x7b]
+          vpermilpd $0x7b, (%rcx), %xmm29
+
+// CHECK: vpermilpd $123, 291(%rax,%r14,8), %xmm29
+// CHECK:  encoding: [0x62,0x23,0xfd,0x08,0x05,0xac,0xf0,0x23,0x01,0x00,0x00,0x7b]
+          vpermilpd $0x7b, 291(%rax,%r14,8), %xmm29
+
+// CHECK: vpermilpd $123, (%rcx){1to2}, %xmm29
+// CHECK:  encoding: [0x62,0x63,0xfd,0x18,0x05,0x29,0x7b]
+          vpermilpd $0x7b, (%rcx){1to2}, %xmm29
+
+// CHECK: vpermilpd $123, 2032(%rdx), %xmm29
+// CHECK:  encoding: [0x62,0x63,0xfd,0x08,0x05,0x6a,0x7f,0x7b]
+          vpermilpd $0x7b, 2032(%rdx), %xmm29
+
+// CHECK: vpermilpd $123, 2048(%rdx), %xmm29
+// CHECK:  encoding: [0x62,0x63,0xfd,0x08,0x05,0xaa,0x00,0x08,0x00,0x00,0x7b]
+          vpermilpd $0x7b, 2048(%rdx), %xmm29
+
+// CHECK: vpermilpd $123, -2048(%rdx), %xmm29
+// CHECK:  encoding: [0x62,0x63,0xfd,0x08,0x05,0x6a,0x80,0x7b]
+          vpermilpd $0x7b, -2048(%rdx), %xmm29
+
+// CHECK: vpermilpd $123, -2064(%rdx), %xmm29
+// CHECK:  encoding: [0x62,0x63,0xfd,0x08,0x05,0xaa,0xf0,0xf7,0xff,0xff,0x7b]
+          vpermilpd $0x7b, -2064(%rdx), %xmm29
+
+// CHECK: vpermilpd $123, 1016(%rdx){1to2}, %xmm29
+// CHECK:  encoding: [0x62,0x63,0xfd,0x18,0x05,0x6a,0x7f,0x7b]
+          vpermilpd $0x7b, 1016(%rdx){1to2}, %xmm29
+
+// CHECK: vpermilpd $123, 1024(%rdx){1to2}, %xmm29
+// CHECK:  encoding: [0x62,0x63,0xfd,0x18,0x05,0xaa,0x00,0x04,0x00,0x00,0x7b]
+          vpermilpd $0x7b, 1024(%rdx){1to2}, %xmm29
+
+// CHECK: vpermilpd $123, -1024(%rdx){1to2}, %xmm29
+// CHECK:  encoding: [0x62,0x63,0xfd,0x18,0x05,0x6a,0x80,0x7b]
+          vpermilpd $0x7b, -1024(%rdx){1to2}, %xmm29
+
+// CHECK: vpermilpd $123, -1032(%rdx){1to2}, %xmm29
+// CHECK:  encoding: [0x62,0x63,0xfd,0x18,0x05,0xaa,0xf8,0xfb,0xff,0xff,0x7b]
+          vpermilpd $0x7b, -1032(%rdx){1to2}, %xmm29
+
+// CHECK: vpermilpd $171, %ymm24, %ymm17
+// CHECK:  encoding: [0x62,0x83,0xfd,0x28,0x05,0xc8,0xab]
+          vpermilpd $0xab, %ymm24, %ymm17
+
+// CHECK: vpermilpd $171, %ymm24, %ymm17 {%k6}
+// CHECK:  encoding: [0x62,0x83,0xfd,0x2e,0x05,0xc8,0xab]
+          vpermilpd $0xab, %ymm24, %ymm17 {%k6}
+
+// CHECK: vpermilpd $171, %ymm24, %ymm17 {%k6} {z}
+// CHECK:  encoding: [0x62,0x83,0xfd,0xae,0x05,0xc8,0xab]
+          vpermilpd $0xab, %ymm24, %ymm17 {%k6} {z}
+
+// CHECK: vpermilpd $123, %ymm24, %ymm17
+// CHECK:  encoding: [0x62,0x83,0xfd,0x28,0x05,0xc8,0x7b]
+          vpermilpd $0x7b, %ymm24, %ymm17
+
+// CHECK: vpermilpd $123, (%rcx), %ymm17
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x28,0x05,0x09,0x7b]
+          vpermilpd $0x7b, (%rcx), %ymm17
+
+// CHECK: vpermilpd $123, 291(%rax,%r14,8), %ymm17
+// CHECK:  encoding: [0x62,0xa3,0xfd,0x28,0x05,0x8c,0xf0,0x23,0x01,0x00,0x00,0x7b]
+          vpermilpd $0x7b, 291(%rax,%r14,8), %ymm17
+
+// CHECK: vpermilpd $123, (%rcx){1to4}, %ymm17
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x38,0x05,0x09,0x7b]
+          vpermilpd $0x7b, (%rcx){1to4}, %ymm17
+
+// CHECK: vpermilpd $123, 4064(%rdx), %ymm17
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x28,0x05,0x4a,0x7f,0x7b]
+          vpermilpd $0x7b, 4064(%rdx), %ymm17
+
+// CHECK: vpermilpd $123, 4096(%rdx), %ymm17
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x28,0x05,0x8a,0x00,0x10,0x00,0x00,0x7b]
+          vpermilpd $0x7b, 4096(%rdx), %ymm17
+
+// CHECK: vpermilpd $123, -4096(%rdx), %ymm17
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x28,0x05,0x4a,0x80,0x7b]
+          vpermilpd $0x7b, -4096(%rdx), %ymm17
+
+// CHECK: vpermilpd $123, -4128(%rdx), %ymm17
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x28,0x05,0x8a,0xe0,0xef,0xff,0xff,0x7b]
+          vpermilpd $0x7b, -4128(%rdx), %ymm17
+
+// CHECK: vpermilpd $123, 1016(%rdx){1to4}, %ymm17
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x38,0x05,0x4a,0x7f,0x7b]
+          vpermilpd $0x7b, 1016(%rdx){1to4}, %ymm17
+
+// CHECK: vpermilpd $123, 1024(%rdx){1to4}, %ymm17
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x38,0x05,0x8a,0x00,0x04,0x00,0x00,0x7b]
+          vpermilpd $0x7b, 1024(%rdx){1to4}, %ymm17
+
+// CHECK: vpermilpd $123, -1024(%rdx){1to4}, %ymm17
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x38,0x05,0x4a,0x80,0x7b]
+          vpermilpd $0x7b, -1024(%rdx){1to4}, %ymm17
+
+// CHECK: vpermilpd $123, -1032(%rdx){1to4}, %ymm17
+// CHECK:  encoding: [0x62,0xe3,0xfd,0x38,0x05,0x8a,0xf8,0xfb,0xff,0xff,0x7b]
+          vpermilpd $0x7b, -1032(%rdx){1to4}, %ymm17
+
+// CHECK: vpermilpd %xmm17, %xmm27, %xmm26
+// CHECK:  encoding: [0x62,0x22,0xa5,0x00,0x0d,0xd1]
+          vpermilpd %xmm17, %xmm27, %xmm26
+
+// CHECK: vpermilpd %xmm17, %xmm27, %xmm26 {%k2}
+// CHECK:  encoding: [0x62,0x22,0xa5,0x02,0x0d,0xd1]
+          vpermilpd %xmm17, %xmm27, %xmm26 {%k2}
+
+// CHECK: vpermilpd %xmm17, %xmm27, %xmm26 {%k2} {z}
+// CHECK:  encoding: [0x62,0x22,0xa5,0x82,0x0d,0xd1]
+          vpermilpd %xmm17, %xmm27, %xmm26 {%k2} {z}
+
+// CHECK: vpermilpd (%rcx), %xmm27, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xa5,0x00,0x0d,0x11]
+          vpermilpd (%rcx), %xmm27, %xmm26
+
+// CHECK: vpermilpd 291(%rax,%r14,8), %xmm27, %xmm26
+// CHECK:  encoding: [0x62,0x22,0xa5,0x00,0x0d,0x94,0xf0,0x23,0x01,0x00,0x00]
+          vpermilpd 291(%rax,%r14,8), %xmm27, %xmm26
+
+// CHECK: vpermilpd (%rcx){1to2}, %xmm27, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xa5,0x10,0x0d,0x11]
+          vpermilpd (%rcx){1to2}, %xmm27, %xmm26
+
+// CHECK: vpermilpd 2032(%rdx), %xmm27, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xa5,0x00,0x0d,0x52,0x7f]
+          vpermilpd 2032(%rdx), %xmm27, %xmm26
+
+// CHECK: vpermilpd 2048(%rdx), %xmm27, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xa5,0x00,0x0d,0x92,0x00,0x08,0x00,0x00]
+          vpermilpd 2048(%rdx), %xmm27, %xmm26
+
+// CHECK: vpermilpd -2048(%rdx), %xmm27, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xa5,0x00,0x0d,0x52,0x80]
+          vpermilpd -2048(%rdx), %xmm27, %xmm26
+
+// CHECK: vpermilpd -2064(%rdx), %xmm27, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xa5,0x00,0x0d,0x92,0xf0,0xf7,0xff,0xff]
+          vpermilpd -2064(%rdx), %xmm27, %xmm26
+
+// CHECK: vpermilpd 1016(%rdx){1to2}, %xmm27, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xa5,0x10,0x0d,0x52,0x7f]
+          vpermilpd 1016(%rdx){1to2}, %xmm27, %xmm26
+
+// CHECK: vpermilpd 1024(%rdx){1to2}, %xmm27, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xa5,0x10,0x0d,0x92,0x00,0x04,0x00,0x00]
+          vpermilpd 1024(%rdx){1to2}, %xmm27, %xmm26
+
+// CHECK: vpermilpd -1024(%rdx){1to2}, %xmm27, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xa5,0x10,0x0d,0x52,0x80]
+          vpermilpd -1024(%rdx){1to2}, %xmm27, %xmm26
+
+// CHECK: vpermilpd -1032(%rdx){1to2}, %xmm27, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xa5,0x10,0x0d,0x92,0xf8,0xfb,0xff,0xff]
+          vpermilpd -1032(%rdx){1to2}, %xmm27, %xmm26
+
+// CHECK: vpermilpd %ymm24, %ymm26, %ymm26
+// CHECK:  encoding: [0x62,0x02,0xad,0x20,0x0d,0xd0]
+          vpermilpd %ymm24, %ymm26, %ymm26
+
+// CHECK: vpermilpd %ymm24, %ymm26, %ymm26 {%k5}
+// CHECK:  encoding: [0x62,0x02,0xad,0x25,0x0d,0xd0]
+          vpermilpd %ymm24, %ymm26, %ymm26 {%k5}
+
+// CHECK: vpermilpd %ymm24, %ymm26, %ymm26 {%k5} {z}
+// CHECK:  encoding: [0x62,0x02,0xad,0xa5,0x0d,0xd0]
+          vpermilpd %ymm24, %ymm26, %ymm26 {%k5} {z}
+
+// CHECK: vpermilpd (%rcx), %ymm26, %ymm26
+// CHECK:  encoding: [0x62,0x62,0xad,0x20,0x0d,0x11]
+          vpermilpd (%rcx), %ymm26, %ymm26
+
+// CHECK: vpermilpd 291(%rax,%r14,8), %ymm26, %ymm26
+// CHECK:  encoding: [0x62,0x22,0xad,0x20,0x0d,0x94,0xf0,0x23,0x01,0x00,0x00]
+          vpermilpd 291(%rax,%r14,8), %ymm26, %ymm26
+
+// CHECK: vpermilpd (%rcx){1to4}, %ymm26, %ymm26
+// CHECK:  encoding: [0x62,0x62,0xad,0x30,0x0d,0x11]
+          vpermilpd (%rcx){1to4}, %ymm26, %ymm26
+
+// CHECK: vpermilpd 4064(%rdx), %ymm26, %ymm26
+// CHECK:  encoding: [0x62,0x62,0xad,0x20,0x0d,0x52,0x7f]
+          vpermilpd 4064(%rdx), %ymm26, %ymm26
+
+// CHECK: vpermilpd 4096(%rdx), %ymm26, %ymm26
+// CHECK:  encoding: [0x62,0x62,0xad,0x20,0x0d,0x92,0x00,0x10,0x00,0x00]
+          vpermilpd 4096(%rdx), %ymm26, %ymm26
+
+// CHECK: vpermilpd -4096(%rdx), %ymm26, %ymm26
+// CHECK:  encoding: [0x62,0x62,0xad,0x20,0x0d,0x52,0x80]
+          vpermilpd -4096(%rdx), %ymm26, %ymm26
+
+// CHECK: vpermilpd -4128(%rdx), %ymm26, %ymm26
+// CHECK:  encoding: [0x62,0x62,0xad,0x20,0x0d,0x92,0xe0,0xef,0xff,0xff]
+          vpermilpd -4128(%rdx), %ymm26, %ymm26
+
+// CHECK: vpermilpd 1016(%rdx){1to4}, %ymm26, %ymm26
+// CHECK:  encoding: [0x62,0x62,0xad,0x30,0x0d,0x52,0x7f]
+          vpermilpd 1016(%rdx){1to4}, %ymm26, %ymm26
+
+// CHECK: vpermilpd 1024(%rdx){1to4}, %ymm26, %ymm26
+// CHECK:  encoding: [0x62,0x62,0xad,0x30,0x0d,0x92,0x00,0x04,0x00,0x00]
+          vpermilpd 1024(%rdx){1to4}, %ymm26, %ymm26
+
+// CHECK: vpermilpd -1024(%rdx){1to4}, %ymm26, %ymm26
+// CHECK:  encoding: [0x62,0x62,0xad,0x30,0x0d,0x52,0x80]
+          vpermilpd -1024(%rdx){1to4}, %ymm26, %ymm26
+
+// CHECK: vpermilpd -1032(%rdx){1to4}, %ymm26, %ymm26
+// CHECK:  encoding: [0x62,0x62,0xad,0x30,0x0d,0x92,0xf8,0xfb,0xff,0xff]
+          vpermilpd -1032(%rdx){1to4}, %ymm26, %ymm26
+