[AVX512] add getexp instruction

[AsafBadouh]

Summary

add getexp instruction with sae
add rounding mode for sqrt

[delena]

lib/Target/X86/X86ISelLowering.h

205

Don't need 2 nodes for getexp.

[AsafBadouh]

remove the extra SDNode

[delena]

Please fix the {} and commit.
Thanks

lib/Target/X86/X86ISelLowering.cpp

15009

remove {} from one line "if"

include/llvm/IR/IntrinsicsX86.td

def int_x86_avx512_mask_rndscale_sd : GCCBuiltin<"__builtin_ia32_rndscalesd_mask">,
                                      [IntrNoMem]>;
   def int_x86_avx512_sqrt_ss        : GCCBuiltin<"__builtin_ia32_sqrtrndss">,
               Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_v4f32_ty],
                         [IntrNoMem]>;
   def int_x86_avx512_sqrt_sd        : GCCBuiltin<"__builtin_ia32_sqrtrndsd">,
               Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty, llvm_v2f64_ty],
                         [IntrNoMem]>;
 
-  def int_x86_avx512_sqrt_pd_512 : GCCBuiltin<"__builtin_ia32_sqrtpd512_mask">,
+  def int_x86_avx512_mask_sqrt_pd_128 : GCCBuiltin<"__builtin_ia32_sqrtpd128_mask">,
+        Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty, llvm_v2f64_ty,
+                                    llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_sqrt_pd_256 : GCCBuiltin<"__builtin_ia32_sqrtpd256_mask">,
+        Intrinsic<[llvm_v4f64_ty], [llvm_v4f64_ty, llvm_v4f64_ty,
+                                    llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_sqrt_pd_512 : GCCBuiltin<"__builtin_ia32_sqrtpd512_mask">,
+        Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty, llvm_v8f64_ty,
+                                    llvm_i8_ty, llvm_i32_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_sqrt_ps_128 : GCCBuiltin<"__builtin_ia32_sqrtps128_mask">,
+        Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_v4f32_ty,
+                                     llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_sqrt_ps_256 : GCCBuiltin<"__builtin_ia32_sqrtps256_mask">,
+        Intrinsic<[llvm_v8f32_ty], [llvm_v8f32_ty, llvm_v8f32_ty,
+                                     llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_sqrt_ps_512 : GCCBuiltin<"__builtin_ia32_sqrtps512_mask">,
+        Intrinsic<[llvm_v16f32_ty], [llvm_v16f32_ty, llvm_v16f32_ty,
+                                     llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_getexp_pd_128 : GCCBuiltin<"__builtin_ia32_getexppd128_mask">,
+        Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty, llvm_v2f64_ty,
+                                    llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_getexp_pd_256 : GCCBuiltin<"__builtin_ia32_getexppd256_mask">,
+        Intrinsic<[llvm_v4f64_ty], [llvm_v4f64_ty, llvm_v4f64_ty,
+                                    llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_getexp_pd_512 : GCCBuiltin<"__builtin_ia32_getexppd512_mask">,
         Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty, llvm_v8f64_ty,
                                     llvm_i8_ty, llvm_i32_ty], [IntrNoMem]>;
-  def int_x86_avx512_sqrt_ps_512 : GCCBuiltin<"__builtin_ia32_sqrtps512_mask">,
+  def int_x86_avx512_mask_getexp_ps_128 : GCCBuiltin<"__builtin_ia32_getexpps128_mask">,
+        Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_v4f32_ty,
+                                     llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_getexp_ps_256 : GCCBuiltin<"__builtin_ia32_getexpps256_mask">,
+        Intrinsic<[llvm_v8f32_ty], [llvm_v8f32_ty, llvm_v8f32_ty,
+                                     llvm_i8_ty], [IntrNoMem]>;
+  def int_x86_avx512_mask_getexp_ps_512 : GCCBuiltin<"__builtin_ia32_getexpps512_mask">,
         Intrinsic<[llvm_v16f32_ty], [llvm_v16f32_ty, llvm_v16f32_ty,
                                      llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>;
 
   def int_x86_avx512_rsqrt14_ss : GCCBuiltin<"__builtin_ia32_rsqrt14ss_mask">,
         Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty, llvm_v4f32_ty, llvm_v4f32_ty,
                                     llvm_i8_ty], [IntrNoMem]>;
   def int_x86_avx512_rsqrt14_sd : GCCBuiltin<"__builtin_ia32_rsqrt14sd_mask">,
         Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty, llvm_v2f64_ty, llvm_v2f64_ty,

lib/Target/X86/X86ISelLowering.h

enum NodeType : unsigned {
 
       /// Blend where the condition has been shrunk.
       /// This is used to emphasize that the condition mask is
       /// no more valid for generic VSELECT optimizations.
       SHRUNKBLEND,
 
       /// Combined add and sub on an FP vector.
       ADDSUB,
+
+      // FP vector get exponent 
+      FGETEXP,
+
       //  FP vector ops with rounding mode.
       FADD_RND,
       FSUB_RND,
       FMUL_RND,
       FDIV_RND,
       FMAX_RND,
       FMIN_RND,
+      FSQRT_RND,
+      FGETEXP_RND,
       
       // Integer add/sub with unsigned saturation.
       ADDUS,
       SUBUS,
       // Integer add/sub with signed saturation.
       ADDS,
       SUBS,
 

lib/Target/X86/X86ISelLowering.cpp

if (IntrData) {
     case INTR_TYPE_2OP:
       return DAG.getNode(IntrData->Opc0, dl, Op.getValueType(), Op.getOperand(1),
         Op.getOperand(2));
     case INTR_TYPE_3OP:
       return DAG.getNode(IntrData->Opc0, dl, Op.getValueType(), Op.getOperand(1),
         Op.getOperand(2), Op.getOperand(3));
     case INTR_TYPE_1OP_MASK_RM: {
       SDValue Src = Op.getOperand(1);
-      SDValue Src0 = Op.getOperand(2);
+      SDValue PassThru = Op.getOperand(2);
       SDValue Mask = Op.getOperand(3);
       SDValue RoundingMode = Op.getOperand(4);
+      unsigned IntrWithRoundingModeOpcode = IntrData->Opc1;
+      if (IntrWithRoundingModeOpcode != 0) {
+        unsigned Round = cast<ConstantSDNode>(RoundingMode)->getZExtValue();
+        if (Round != X86::STATIC_ROUNDING::CUR_DIRECTION) {
+          return getVectorMaskingNode(DAG.getNode(IntrWithRoundingModeOpcode,
+                                      dl, Op.getValueType(), Src, RoundingMode),
+                                      Mask, PassThru, Subtarget, DAG);
+        }
+      }
       return getVectorMaskingNode(DAG.getNode(IntrData->Opc0, dl, VT, Src,
                                               RoundingMode),
-                                  Mask, Src0, Subtarget, DAG);
+                                  Mask, PassThru, Subtarget, DAG);
+    }
+    case INTR_TYPE_1OP_MASK: {
+      SDValue Src = Op.getOperand(1);
+      SDValue Passthru = Op.getOperand(2);
+      SDValue Mask = Op.getOperand(3);
+      return getVectorMaskingNode(DAG.getNode(IntrData->Opc0, dl, VT, Src),
+                                  Mask, Passthru, Subtarget, DAG);
     }
     case INTR_TYPE_SCALAR_MASK_RM: {
       SDValue Src1 = Op.getOperand(1);
       SDValue Src2 = Op.getOperand(2);
       SDValue Src0 = Op.getOperand(3);
       SDValue Mask = Op.getOperand(4);
       // There are 2 kinds of intrinsics in this group:
       // (1) With supress-all-exceptions (sae) or rounding mode- 6 operands

lib/Target/X86/X86InstrAVX512.td

 
 let Predicates = [HasERI], hasSideEffects = 0 in {
 
  defm VRSQRT28 : avx512_eri<0xCC, "vrsqrt28", X86rsqrt28>, EVEX, EVEX_V512, T8PD;
  defm VRCP28   : avx512_eri<0xCA, "vrcp28",   X86rcp28>,   EVEX, EVEX_V512, T8PD;
  defm VEXP2    : avx512_eri<0xC8, "vexp2",    X86exp2>,    EVEX, EVEX_V512, T8PD;
 }
 
-multiclass avx512_sqrt_packed<bits<8> opc, string OpcodeStr,
+multiclass avx512_fp_unaryop_packed_round<bits<8> opc, string OpcodeStr,
+                              SDNode OpNodeRnd, X86VectorVTInfo _>{
+  defm rb: AVX512_maskable<opc, MRMSrcReg, _, (outs _.RC:$dst),
+                         (ins _.RC:$src, AVX512RC:$rc), OpcodeStr, "$rc, $src", "$src, $rc",
+                         (_.VT (OpNodeRnd _.RC:$src, (i32 imm:$rc)))>,
+                         EVEX, EVEX_B, EVEX_RC;
+}
+
+multiclass avx512_fp_unaryop_packed_sae<bits<8> opc, string OpcodeStr,
+                              SDNode OpNodeRnd, X86VectorVTInfo _>{
+  defm rb: AVX512_maskable<opc, MRMSrcReg, _, (outs _.RC:$dst),
+                         (ins _.RC:$src), OpcodeStr, "{sae}, $src", "$src, {sae}",
+                         (_.VT (OpNodeRnd _.RC:$src, (i32 FROUND_NO_EXC)))>,
+                         EVEX, EVEX_B;
+}
+
+multiclass avx512_fp_unaryop_packed<bits<8> opc, string OpcodeStr,
                               SDNode OpNode, X86VectorVTInfo _>{
   defm r: AVX512_maskable<opc, MRMSrcReg, _, (outs _.RC:$dst),
                          (ins _.RC:$src), OpcodeStr, "$src", "$src",
                          (_.FloatVT (OpNode _.RC:$src))>, EVEX;
   let mayLoad = 1 in {
     defm m: AVX512_maskable<opc, MRMSrcMem, _, (outs _.RC:$dst),
                            (ins _.MemOp:$src), OpcodeStr, "$src", "$src",
                            (OpNode (_.FloatVT
def SDZm_Int : SIi8<opc, MRMSrcMem, (outs VR128X:$dst),
                    !strconcat(OpcodeStr,
                   "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
                   [(set VR128X:$dst,
                     (F64Int VR128X:$src1, sse_load_f64:$src2))]>,
                   XD, EVEX_4V, VEX_W, EVEX_CD8<64, CD8VT1>;
   }
 }
 
-multiclass avx512_sqrt_packed_all<bits<8> opc, string OpcodeStr,
+multiclass avx512_fp_unaryop_packed_all<bits<8> opc, string OpcodeStr,
                                   SDNode OpNode> {
-  defm PSZ : avx512_sqrt_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode,
+  defm PSZ : avx512_fp_unaryop_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode,
                                 v16f32_info>,
                                 EVEX_V512, PS, EVEX_CD8<32, CD8VF>;
-  defm PDZ : avx512_sqrt_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode,
+  defm PDZ : avx512_fp_unaryop_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode,
                                 v8f64_info>,
                                 EVEX_V512, VEX_W, PD, EVEX_CD8<64, CD8VF>;
   // Define only if AVX512VL feature is present.
   let Predicates = [HasVLX] in {
-    defm PSZ128 : avx512_sqrt_packed<opc, !strconcat(OpcodeStr, "ps"),
+    defm PSZ128 : avx512_fp_unaryop_packed<opc, !strconcat(OpcodeStr, "ps"),
                                      OpNode, v4f32x_info>,
                                      EVEX_V128, PS, EVEX_CD8<32, CD8VF>;
-    defm PSZ256 : avx512_sqrt_packed<opc, !strconcat(OpcodeStr, "ps"),
+    defm PSZ256 : avx512_fp_unaryop_packed<opc, !strconcat(OpcodeStr, "ps"),
                                      OpNode, v8f32x_info>,
                                      EVEX_V256, PS, EVEX_CD8<32, CD8VF>;
-    defm PDZ128 : avx512_sqrt_packed<opc, !strconcat(OpcodeStr, "pd"),
+    defm PDZ128 : avx512_fp_unaryop_packed<opc, !strconcat(OpcodeStr, "pd"),
                                      OpNode, v2f64x_info>,
                                      EVEX_V128, VEX_W, PD, EVEX_CD8<64, CD8VF>;
-    defm PDZ256 : avx512_sqrt_packed<opc, !strconcat(OpcodeStr, "pd"),
+    defm PDZ256 : avx512_fp_unaryop_packed<opc, !strconcat(OpcodeStr, "pd"),
                                      OpNode, v4f64x_info>,
                                      EVEX_V256, VEX_W, PD, EVEX_CD8<64, CD8VF>;
   }
 }
 
-defm VSQRT : avx512_sqrt_packed_all<0x51, "vsqrt", fsqrt>;
+multiclass avx512_fp_unaryop_packed_all_round<bits<8> opc, string OpcodeStr,
+                                          SDNode OpNodeRnd> {
+  defm PSZ : avx512_fp_unaryop_packed_round<opc, !strconcat(OpcodeStr, "ps"), OpNodeRnd,
+                                v16f32_info>, EVEX_V512, PS, EVEX_CD8<32, CD8VF>;
+  defm PDZ : avx512_fp_unaryop_packed_round<opc, !strconcat(OpcodeStr, "pd"), OpNodeRnd,
+                                v8f64_info>, EVEX_V512, VEX_W, PD, EVEX_CD8<64, CD8VF>;
+}
+
+multiclass avx512_fp_unaryop_packed_all_sae<bits<8> opc, string OpcodeStr,
+                                        SDNode OpNodeRnd> {
+  defm PSZ : avx512_fp_unaryop_packed_sae<opc, !strconcat(OpcodeStr, "ps"), OpNodeRnd,
+                                v16f32_info>, EVEX_V512, EVEX_CD8<32, CD8VF>;
+  defm PDZ : avx512_fp_unaryop_packed_sae<opc, !strconcat(OpcodeStr, "pd"), OpNodeRnd,
+                                v8f64_info>, EVEX_V512, VEX_W, EVEX_CD8<64, CD8VF>;
+}
+
+defm VSQRT   : avx512_fp_unaryop_packed_all<0x51, "vsqrt", fsqrt>,
+               avx512_fp_unaryop_packed_all_round<0x51, "vsqrt", X86fsqrtRnd>;
+defm VGETEXP : avx512_fp_unaryop_packed_all<0x42, "vgetexp", X86fgetexp>,
+               avx512_fp_unaryop_packed_all_sae<0x42, "vgetexp", X86fgetexpRnd>, T8PD;
 
 defm VSQRT  : avx512_sqrt_scalar<0x51, "sqrt",
                 int_x86_avx512_sqrt_ss, int_x86_avx512_sqrt_sd,
                 SSE_SQRTSS, SSE_SQRTSD>;
 
 let Predicates = [HasAVX512] in {
-  def : Pat<(v16f32 (int_x86_avx512_sqrt_ps_512 (v16f32 VR512:$src1),
-                    (bc_v16f32 (v16i32 immAllZerosV)), (i16 -1), FROUND_CURRENT)),
-                   (VSQRTPSZr VR512:$src1)>;
-  def : Pat<(v8f64 (int_x86_avx512_sqrt_pd_512 (v8f64 VR512:$src1),
-                    (bc_v8f64 (v16i32 immAllZerosV)), (i8 -1), FROUND_CURRENT)),
-                   (VSQRTPDZr VR512:$src1)>;
-
   def : Pat<(f32 (fsqrt FR32X:$src)),
             (VSQRTSSZr (f32 (IMPLICIT_DEF)), FR32X:$src)>;
   def : Pat<(f32 (fsqrt (load addr:$src))),
             (VSQRTSSZm (f32 (IMPLICIT_DEF)), addr:$src)>,
             Requires<[OptForSize]>;
   def : Pat<(f64 (fsqrt FR64X:$src)),
             (VSQRTSDZr (f64 (IMPLICIT_DEF)), FR64X:$src)>;
   def : Pat<(f64 (fsqrt (load addr:$src))),

lib/Target/X86/X86InstrFragmentsSIMD.td

 def X86for     : SDNode<"X86ISD::FOR",       SDTFPBinOp,
                         [SDNPCommutative, SDNPAssociative]>;
 def X86fxor    : SDNode<"X86ISD::FXOR",      SDTFPBinOp,
                         [SDNPCommutative, SDNPAssociative]>;
 def X86fandn   : SDNode<"X86ISD::FANDN",     SDTFPBinOp,
                         [SDNPCommutative, SDNPAssociative]>;
 def X86frsqrt  : SDNode<"X86ISD::FRSQRT",    SDTFPUnaryOp>;
 def X86frcp    : SDNode<"X86ISD::FRCP",      SDTFPUnaryOp>;
+def X86fgetexp : SDNode<"X86ISD::FGETEXP",  SDTFPUnaryOp>;
 def X86fsrl    : SDNode<"X86ISD::FSRL",      SDTX86FPShiftOp>;
 def X86fgetsign: SDNode<"X86ISD::FGETSIGNx86",SDTFPToIntOp>;
 def X86fhadd   : SDNode<"X86ISD::FHADD",     SDTFPBinOp>;
 def X86fhsub   : SDNode<"X86ISD::FHSUB",     SDTFPBinOp>;
 def X86hadd    : SDNode<"X86ISD::HADD",      SDTIntBinOp>;
 def X86hsub    : SDNode<"X86ISD::HSUB",      SDTIntBinOp>;
 def X86comi    : SDNode<"X86ISD::COMI",      SDTX86CmpTest>;
 def X86ucomi   : SDNode<"X86ISD::UCOMI",     SDTX86CmpTest>;

 def SDTVBroadcastm : SDTypeProfile<1, 2, [SDTCisVec<0>, SDTCisVec<1>]>;
 
 def SDTBlend : SDTypeProfile<1, 3, [SDTCisVec<0>, SDTCisSameAs<0,1>,
                              SDTCisSameAs<1,2>, SDTCisVT<3, i8>]>;
 
 def SDTFPBinOpRound : SDTypeProfile<1, 3, [      // fadd_round, fmul_round, etc.
   SDTCisSameAs<0, 1>, SDTCisSameAs<0, 2>, SDTCisFP<0>, SDTCisInt<3>]>;
 
+def SDTFPUnaryOpRound : SDTypeProfile<1, 2, [      // fsqrt_round, fgetexp_round, etc.
+  SDTCisSameAs<0, 1>, SDTCisFP<0>, SDTCisInt<2>]>;
+
 def SDTFma : SDTypeProfile<1, 3, [SDTCisSameAs<0,1>,
                            SDTCisSameAs<1,2>, SDTCisSameAs<1,3>]>;
 def SDTFmaRound : SDTypeProfile<1, 4, [SDTCisSameAs<0,1>,
                            SDTCisSameAs<1,2>, SDTCisSameAs<1,3>, SDTCisInt<4>]>;
 def STDFp1SrcRm : SDTypeProfile<1, 2, [SDTCisSameAs<0,1>,
                            SDTCisVec<0>, SDTCisInt<2>]>;
 def STDFp2SrcRm : SDTypeProfile<1, 3, [SDTCisSameAs<0,1>,
                            SDTCisVec<0>, SDTCisInt<3>]>;

 def X86Addsub    : SDNode<"X86ISD::ADDSUB", SDTFPBinOp>;
 
 def X86faddRnd   : SDNode<"X86ISD::FADD_RND",  SDTFPBinOpRound>;
 def X86fsubRnd   : SDNode<"X86ISD::FSUB_RND",  SDTFPBinOpRound>;
 def X86fmulRnd   : SDNode<"X86ISD::FMUL_RND",  SDTFPBinOpRound>;
 def X86fdivRnd   : SDNode<"X86ISD::FDIV_RND",  SDTFPBinOpRound>;
 def X86fmaxRnd   : SDNode<"X86ISD::FMAX_RND",      SDTFPBinOpRound>;
 def X86fminRnd   : SDNode<"X86ISD::FMIN_RND",      SDTFPBinOpRound>;
+def X86fsqrtRnd     : SDNode<"X86ISD::FSQRT_RND",  SDTFPUnaryOpRound>;
+def X86fgetexpRnd   : SDNode<"X86ISD::FGETEXP_RND",  SDTFPUnaryOpRound>;
 
 def X86Fmadd     : SDNode<"X86ISD::FMADD",     SDTFma>;
 def X86Fnmadd    : SDNode<"X86ISD::FNMADD",    SDTFma>;
 def X86Fmsub     : SDNode<"X86ISD::FMSUB",     SDTFma>;
 def X86Fnmsub    : SDNode<"X86ISD::FNMSUB",    SDTFma>;
 def X86Fmaddsub  : SDNode<"X86ISD::FMADDSUB",  SDTFma>;
 def X86Fmsubadd  : SDNode<"X86ISD::FMSUBADD",  SDTFma>;
 

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_3OP,
   CMP_MASK, CMP_MASK_CC, VSHIFT, VSHIFT_MASK, COMI,
-  INTR_TYPE_1OP_MASK_RM, INTR_TYPE_2OP_MASK, FMA_OP_MASK,
+  INTR_TYPE_1OP_MASK, INTR_TYPE_1OP_MASK_RM, INTR_TYPE_2OP_MASK, FMA_OP_MASK,
   INTR_TYPE_SCALAR_MASK_RM, COMPRESS_EXPAND_IN_REG, COMPRESS_TO_MEM,
   EXPAND_FROM_MEM, BLEND
 };
 
 struct IntrinsicData {
 
   unsigned      Id;
   IntrinsicType Type;
static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_div_pd_256, INTR_TYPE_2OP_MASK, ISD::FDIV, 0),
   X86_INTRINSIC_DATA(avx512_mask_div_pd_512, INTR_TYPE_2OP_MASK, ISD::FDIV,
                      X86ISD::FDIV_RND),
   X86_INTRINSIC_DATA(avx512_mask_div_ps_128, INTR_TYPE_2OP_MASK, ISD::FDIV, 0),
   X86_INTRINSIC_DATA(avx512_mask_div_ps_256, INTR_TYPE_2OP_MASK, ISD::FDIV, 0),
   X86_INTRINSIC_DATA(avx512_mask_div_ps_512, INTR_TYPE_2OP_MASK, ISD::FDIV,
                      X86ISD::FDIV_RND),
   X86_INTRINSIC_DATA(avx512_mask_div_sd_round, INTR_TYPE_SCALAR_MASK_RM, ISD::FDIV,
                      X86ISD::FDIV_RND),
   X86_INTRINSIC_DATA(avx512_mask_div_ss_round, INTR_TYPE_SCALAR_MASK_RM, ISD::FDIV,
                      X86ISD::FDIV_RND),
   X86_INTRINSIC_DATA(avx512_mask_expand_d_128,  COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_d_256,  COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_d_512,  COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_pd_128, COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_pd_256, COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_pd_512, COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_ps_128, COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_ps_256, COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_ps_512, COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_q_128,  COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_q_256,  COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_q_512,  COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
+  X86_INTRINSIC_DATA(avx512_mask_getexp_pd_128, INTR_TYPE_1OP_MASK,
+                     X86ISD::FGETEXP, 0),
+  X86_INTRINSIC_DATA(avx512_mask_getexp_pd_256, INTR_TYPE_1OP_MASK,
+                     X86ISD::FGETEXP, 0),
+  X86_INTRINSIC_DATA(avx512_mask_getexp_pd_512, INTR_TYPE_1OP_MASK_RM,
+                     X86ISD::FGETEXP, X86ISD::FGETEXP_RND),
+  X86_INTRINSIC_DATA(avx512_mask_getexp_ps_128, INTR_TYPE_1OP_MASK,
+                     X86ISD::FGETEXP, 0),
+  X86_INTRINSIC_DATA(avx512_mask_getexp_ps_256, INTR_TYPE_1OP_MASK, 
+                     X86ISD::FGETEXP, 0),
+  X86_INTRINSIC_DATA(avx512_mask_getexp_ps_512, INTR_TYPE_1OP_MASK_RM, 
+                     X86ISD::FGETEXP, X86ISD::FGETEXP_RND),
   X86_INTRINSIC_DATA(avx512_mask_max_pd_128, INTR_TYPE_2OP_MASK, X86ISD::FMAX, 0),
   X86_INTRINSIC_DATA(avx512_mask_max_pd_256, INTR_TYPE_2OP_MASK, X86ISD::FMAX, 0),
   X86_INTRINSIC_DATA(avx512_mask_max_pd_512, INTR_TYPE_2OP_MASK, X86ISD::FMAX,
                      X86ISD::FMAX_RND),
   X86_INTRINSIC_DATA(avx512_mask_max_ps_128, INTR_TYPE_2OP_MASK, X86ISD::FMAX, 0),
   X86_INTRINSIC_DATA(avx512_mask_max_ps_256, INTR_TYPE_2OP_MASK, X86ISD::FMAX, 0),
   X86_INTRINSIC_DATA(avx512_mask_max_ps_512, INTR_TYPE_2OP_MASK, X86ISD::FMAX,
                      X86ISD::FMAX_RND),  
static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_pxor_d_512, INTR_TYPE_2OP_MASK, ISD::XOR, 0),
   X86_INTRINSIC_DATA(avx512_mask_pxor_q_128, INTR_TYPE_2OP_MASK, ISD::XOR, 0),
   X86_INTRINSIC_DATA(avx512_mask_pxor_q_256, INTR_TYPE_2OP_MASK, ISD::XOR, 0),
   X86_INTRINSIC_DATA(avx512_mask_pxor_q_512, INTR_TYPE_2OP_MASK, ISD::XOR, 0),
   X86_INTRINSIC_DATA(avx512_mask_rndscale_sd,   INTR_TYPE_SCALAR_MASK_RM,
                      X86ISD::RNDSCALE, 0),
   X86_INTRINSIC_DATA(avx512_mask_rndscale_ss,   INTR_TYPE_SCALAR_MASK_RM,
                      X86ISD::RNDSCALE, 0),
+  X86_INTRINSIC_DATA(avx512_mask_sqrt_pd_128, INTR_TYPE_1OP_MASK, ISD::FSQRT, 0),
+  X86_INTRINSIC_DATA(avx512_mask_sqrt_pd_256, INTR_TYPE_1OP_MASK, ISD::FSQRT, 0),
+  X86_INTRINSIC_DATA(avx512_mask_sqrt_pd_512, INTR_TYPE_1OP_MASK_RM, ISD::FSQRT,
+                     X86ISD::FSQRT_RND),
+  X86_INTRINSIC_DATA(avx512_mask_sqrt_ps_128, INTR_TYPE_1OP_MASK, ISD::FSQRT, 0),
+  X86_INTRINSIC_DATA(avx512_mask_sqrt_ps_256, INTR_TYPE_1OP_MASK, ISD::FSQRT, 0),
+  X86_INTRINSIC_DATA(avx512_mask_sqrt_ps_512, INTR_TYPE_1OP_MASK_RM, ISD::FSQRT,
+                     X86ISD::FSQRT_RND),
   X86_INTRINSIC_DATA(avx512_mask_sub_pd_128, INTR_TYPE_2OP_MASK, ISD::FSUB, 0),
   X86_INTRINSIC_DATA(avx512_mask_sub_pd_256, INTR_TYPE_2OP_MASK, ISD::FSUB, 0),
   X86_INTRINSIC_DATA(avx512_mask_sub_pd_512, INTR_TYPE_2OP_MASK, ISD::FSUB,
                      X86ISD::FSUB_RND),
   X86_INTRINSIC_DATA(avx512_mask_sub_ps_128, INTR_TYPE_2OP_MASK, ISD::FSUB, 0),
   X86_INTRINSIC_DATA(avx512_mask_sub_ps_256, INTR_TYPE_2OP_MASK, ISD::FSUB, 0),
   X86_INTRINSIC_DATA(avx512_mask_sub_ps_512, INTR_TYPE_2OP_MASK, ISD::FSUB,
                      X86ISD::FSUB_RND),

test/CodeGen/X86/avx512-intrinsics.ll

 define <4 x float> @test_rcp14_ss(<4 x float> %a0) {
   ; CHECK: vrcp14ss {{.*}}encoding: [0x62,0xf2,0x7d,0x08,0x4d,0xc0]
   %res = call <4 x float> @llvm.x86.avx512.rcp14.ss(<4 x float> %a0, <4 x float> %a0, <4 x float> zeroinitializer, i8 -1) ; <<4 x float>> [#uses=1]
   ret <4 x float> %res
 }
 declare <4 x float> @llvm.x86.avx512.rcp14.ss(<4 x float>, <4 x float>, <4 x float>, i8) nounwind readnone
 
 define <8 x double> @test_sqrt_pd_512(<8 x double> %a0) {
+  ; CHECK-LABEL: test_sqrt_pd_512
   ; CHECK: vsqrtpd
-  %res = call <8 x double> @llvm.x86.avx512.sqrt.pd.512(<8 x double> %a0,  <8 x double> zeroinitializer, i8 -1, i32 4) ; <<8 x double>> [#uses=1]
+  %res = call <8 x double> @llvm.x86.avx512.mask.sqrt.pd.512(<8 x double> %a0,  <8 x double> zeroinitializer, i8 -1, i32 4) 
   ret <8 x double> %res
 }
-declare <8 x double> @llvm.x86.avx512.sqrt.pd.512(<8 x double>, <8 x double>, i8, i32) nounwind readnone
+declare <8 x double> @llvm.x86.avx512.mask.sqrt.pd.512(<8 x double>, <8 x double>, i8, i32) nounwind readnone
 
 define <16 x float> @test_sqrt_ps_512(<16 x float> %a0) {
+  ; CHECK-LABEL: test_sqrt_ps_512
   ; CHECK: vsqrtps
-  %res = call <16 x float> @llvm.x86.avx512.sqrt.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 4) ; <<16 x float>> [#uses=1]
+  %res = call <16 x float> @llvm.x86.avx512.mask.sqrt.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 4) 
   ret <16 x float> %res
 }
-declare <16 x float> @llvm.x86.avx512.sqrt.ps.512(<16 x float>, <16 x float>, i16, i32) nounwind readnone
+define <16 x float> @test_sqrt_round_ps_512(<16 x float> %a0) {
+  ; CHECK-LABEL: test_sqrt_round_ps_512
+  ; CHECK: vsqrtps {rz-sae}
+  %res = call <16 x float> @llvm.x86.avx512.mask.sqrt.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 3) 
+  ret <16 x float> %res
+}
+declare <16 x float> @llvm.x86.avx512.mask.sqrt.ps.512(<16 x float>, <16 x float>, i16, i32) nounwind readnone
+
+define <8 x double> @test_getexp_pd_512(<8 x double> %a0) {
+  ; CHECK-LABEL: test_getexp_pd_512
+  ; CHECK: vgetexppd
+  %res = call <8 x double> @llvm.x86.avx512.mask.getexp.pd.512(<8 x double> %a0,  <8 x double> zeroinitializer, i8 -1, i32 4) 
+  ret <8 x double> %res
+}
+define <8 x double> @test_getexp_round_pd_512(<8 x double> %a0) {
+  ; CHECK-LABEL: test_getexp_round_pd_512
+  ; CHECK: vgetexppd {sae}
+  %res = call <8 x double> @llvm.x86.avx512.mask.getexp.pd.512(<8 x double> %a0,  <8 x double> zeroinitializer, i8 -1, i32 8) 
+  ret <8 x double> %res
+}
+declare <8 x double> @llvm.x86.avx512.mask.getexp.pd.512(<8 x double>, <8 x double>, i8, i32) nounwind readnone
+
+define <16 x float> @test_getexp_ps_512(<16 x float> %a0) {
+  ; CHECK-LABEL: test_getexp_ps_512
+  ; CHECK: vgetexpps
+  %res = call <16 x float> @llvm.x86.avx512.mask.getexp.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 4) 
+  ret <16 x float> %res
+}
+
+define <16 x float> @test_getexp_round_ps_512(<16 x float> %a0) {
+  ; CHECK-LABEL: test_getexp_round_ps_512
+  ; CHECK: vgetexpps {sae}
+  %res = call <16 x float> @llvm.x86.avx512.mask.getexp.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 8) 
+  ret <16 x float> %res
+}
+declare <16 x float> @llvm.x86.avx512.mask.getexp.ps.512(<16 x float>, <16 x float>, i16, i32) nounwind readnone
 
 define <4 x float> @test_sqrt_ss(<4 x float> %a0, <4 x float> %a1) {
   ; CHECK: vsqrtss {{.*}}encoding: [0x62
   %res = call <4 x float> @llvm.x86.avx512.sqrt.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
   ret <4 x float> %res
 }
 declare <4 x float> @llvm.x86.avx512.sqrt.ss(<4 x float>, <4 x float>) nounwind readnone
 

test/CodeGen/X86/avx512vl-intrinsics.ll

 }
 
 define <4 x float> @test_mm512_min_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
   ;CHECK-LABEL: test_mm512_min_ps_128
   ;CHECK: vminps %xmm1, %xmm0, %xmm0
   %res = call <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 -1)
   ret <4 x float> %res
 }
-declare <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)
+declare <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)
+
+define <4 x double> @test_sqrt_pd_256(<4 x double> %a0, i8 %mask) {
+  ; CHECK-LABEL: test_sqrt_pd_256
+  ; CHECK: vsqrtpd
+  %res = call <4 x double> @llvm.x86.avx512.mask.sqrt.pd.256(<4 x double> %a0,  <4 x double> zeroinitializer, i8 %mask)
+  ret <4 x double> %res
+}
+declare <4 x double> @llvm.x86.avx512.mask.sqrt.pd.256(<4 x double>, <4 x double>, i8) nounwind readnone
+
+define <8 x float> @test_sqrt_ps_256(<8 x float> %a0, i8 %mask) {
+  ; CHECK-LABEL: test_sqrt_ps_256
+  ; CHECK: vsqrtps
+  %res = call <8 x float> @llvm.x86.avx512.mask.sqrt.ps.256(<8 x float> %a0, <8 x float> zeroinitializer, i8 %mask)
+  ret <8 x float> %res
+}
+
+declare <8 x float> @llvm.x86.avx512.mask.sqrt.ps.256(<8 x float>, <8 x float>, i8) nounwind readnone
+
+define <4 x double> @test_getexp_pd_256(<4 x double> %a0) {
+  ; CHECK-LABEL: test_getexp_pd_256
+  ; CHECK: vgetexppd
+  %res = call <4 x double> @llvm.x86.avx512.mask.getexp.pd.256(<4 x double> %a0,  <4 x double> zeroinitializer, i8 -1)
+  ret <4 x double> %res
+}
+
+declare <4 x double> @llvm.x86.avx512.mask.getexp.pd.256(<4 x double>, <4 x double>, i8) nounwind readnone
+
+define <8 x float> @test_getexp_ps_256(<8 x float> %a0) {
+  ; CHECK-LABEL: test_getexp_ps_256
+  ; CHECK: vgetexpps
+  %res = call <8 x float> @llvm.x86.avx512.mask.getexp.ps.256(<8 x float> %a0, <8 x float> zeroinitializer, i8 -1)
+  ret <8 x float> %res
+}
+declare <8 x float> @llvm.x86.avx512.mask.getexp.ps.256(<8 x float>, <8 x float>, i8) nounwind readnone
 No newline at end of file

test/MC/X86/avx512-encodings.s

 // CHECK: vsqrtps -512(%rdx){1to16}, %zmm28
 // CHECK:  encoding: [0x62,0x61,0x7c,0x58,0x51,0x62,0x80]
           vsqrtps -512(%rdx){1to16}, %zmm28
 
 // CHECK: vsqrtps -516(%rdx){1to16}, %zmm28
 // CHECK:  encoding: [0x62,0x61,0x7c,0x58,0x51,0xa2,0xfc,0xfd,0xff,0xff]
           vsqrtps -516(%rdx){1to16}, %zmm28
 
+// CHECK: vsqrtpd {rn-sae}, %zmm19, %zmm19
+// CHECK:  encoding: [0x62,0xa1,0xfd,0x18,0x51,0xdb]
+          vsqrtpd {rn-sae}, %zmm19, %zmm19
+
+// CHECK: vsqrtpd {ru-sae}, %zmm19, %zmm19
+// CHECK:  encoding: [0x62,0xa1,0xfd,0x58,0x51,0xdb]
+          vsqrtpd {ru-sae}, %zmm19, %zmm19
+
+// CHECK: vsqrtpd {rd-sae}, %zmm19, %zmm19
+// CHECK:  encoding: [0x62,0xa1,0xfd,0x38,0x51,0xdb]
+          vsqrtpd {rd-sae}, %zmm19, %zmm19
+
+// CHECK: vsqrtpd {rz-sae}, %zmm19, %zmm19
+// CHECK:  encoding: [0x62,0xa1,0xfd,0x78,0x51,0xdb]
+          vsqrtpd {rz-sae}, %zmm19, %zmm19
+
+// CHECK: vsqrtps {rn-sae}, %zmm29, %zmm28
+// CHECK:  encoding: [0x62,0x01,0x7c,0x18,0x51,0xe5]
+          vsqrtps {rn-sae}, %zmm29, %zmm28
+
+// CHECK: vsqrtps {ru-sae}, %zmm29, %zmm28
+// CHECK:  encoding: [0x62,0x01,0x7c,0x58,0x51,0xe5]
+          vsqrtps {ru-sae}, %zmm29, %zmm28
+
+// CHECK: vsqrtps {rd-sae}, %zmm29, %zmm28
+// CHECK:  encoding: [0x62,0x01,0x7c,0x38,0x51,0xe5]
+          vsqrtps {rd-sae}, %zmm29, %zmm28
+
+// CHECK: vsqrtps {rz-sae}, %zmm29, %zmm28
+// CHECK:  encoding: [0x62,0x01,0x7c,0x78,0x51,0xe5]
+          vsqrtps {rz-sae}, %zmm29, %zmm28
+
 // CHECK: vsubpd %zmm9, %zmm12, %zmm9
 // CHECK:  encoding: [0x62,0x51,0x9d,0x48,0x5c,0xc9]
           vsubpd %zmm9, %zmm12, %zmm9
 
 // CHECK: vsubpd %zmm9, %zmm12, %zmm9 {%k7}
 // CHECK:  encoding: [0x62,0x51,0x9d,0x4f,0x5c,0xc9]
           vsubpd %zmm9, %zmm12, %zmm9 {%k7}
 

 // CHECK: vpsraq $123, -1024(%rdx){1to8}, %zmm29
 // CHECK:  encoding: [0x62,0xf1,0x95,0x50,0x72,0x62,0x80,0x7b]
           vpsraq $123, -1024(%rdx){1to8}, %zmm29
 
 // CHECK: vpsraq $123, -1032(%rdx){1to8}, %zmm29
 // CHECK:  encoding: [0x62,0xf1,0x95,0x50,0x72,0xa2,0xf8,0xfb,0xff,0xff,0x7b]
           vpsraq $123, -1032(%rdx){1to8}, %zmm29
 
+// CHECK: vgetexppd %zmm25, %zmm14
+// CHECK:  encoding: [0x62,0x12,0xfd,0x48,0x42,0xf1]
+          vgetexppd %zmm25, %zmm14
+
+// CHECK: vgetexppd %zmm25, %zmm14 {%k5}
+// CHECK:  encoding: [0x62,0x12,0xfd,0x4d,0x42,0xf1]
+          vgetexppd %zmm25, %zmm14 {%k5}
+
+// CHECK: vgetexppd %zmm25, %zmm14 {%k5} {z}
+// CHECK:  encoding: [0x62,0x12,0xfd,0xcd,0x42,0xf1]
+          vgetexppd %zmm25, %zmm14 {%k5} {z}
+
+// CHECK: vgetexppd {sae}, %zmm25, %zmm14
+// CHECK:  encoding: [0x62,0x12,0xfd,0x18,0x42,0xf1]
+          vgetexppd {sae}, %zmm25, %zmm14
+
+// CHECK: vgetexppd (%rcx), %zmm14
+// CHECK:  encoding: [0x62,0x72,0xfd,0x48,0x42,0x31]
+          vgetexppd (%rcx), %zmm14
+
+// CHECK: vgetexppd 291(%rax,%r14,8), %zmm14
+// CHECK:  encoding: [0x62,0x32,0xfd,0x48,0x42,0xb4,0xf0,0x23,0x01,0x00,0x00]
+          vgetexppd 291(%rax,%r14,8), %zmm14
+
+// CHECK: vgetexppd (%rcx){1to8}, %zmm14
+// CHECK:  encoding: [0x62,0x72,0xfd,0x58,0x42,0x31]
+          vgetexppd (%rcx){1to8}, %zmm14
+
+// CHECK: vgetexppd 8128(%rdx), %zmm14
+// CHECK:  encoding: [0x62,0x72,0xfd,0x48,0x42,0x72,0x7f]
+          vgetexppd 8128(%rdx), %zmm14
+
+// CHECK: vgetexppd 8192(%rdx), %zmm14
+// CHECK:  encoding: [0x62,0x72,0xfd,0x48,0x42,0xb2,0x00,0x20,0x00,0x00]
+          vgetexppd 8192(%rdx), %zmm14
+
+// CHECK: vgetexppd -8192(%rdx), %zmm14
+// CHECK:  encoding: [0x62,0x72,0xfd,0x48,0x42,0x72,0x80]
+          vgetexppd -8192(%rdx), %zmm14
+
+// CHECK: vgetexppd -8256(%rdx), %zmm14
+// CHECK:  encoding: [0x62,0x72,0xfd,0x48,0x42,0xb2,0xc0,0xdf,0xff,0xff]
+          vgetexppd -8256(%rdx), %zmm14
+
+// CHECK: vgetexppd 1016(%rdx){1to8}, %zmm14
+// CHECK:  encoding: [0x62,0x72,0xfd,0x58,0x42,0x72,0x7f]
+          vgetexppd 1016(%rdx){1to8}, %zmm14
+
+// CHECK: vgetexppd 1024(%rdx){1to8}, %zmm14
+// CHECK:  encoding: [0x62,0x72,0xfd,0x58,0x42,0xb2,0x00,0x04,0x00,0x00]
+          vgetexppd 1024(%rdx){1to8}, %zmm14
+
+// CHECK: vgetexppd -1024(%rdx){1to8}, %zmm14
+// CHECK:  encoding: [0x62,0x72,0xfd,0x58,0x42,0x72,0x80]
+          vgetexppd -1024(%rdx){1to8}, %zmm14
+
+// CHECK: vgetexppd -1032(%rdx){1to8}, %zmm14
+// CHECK:  encoding: [0x62,0x72,0xfd,0x58,0x42,0xb2,0xf8,0xfb,0xff,0xff]
+          vgetexppd -1032(%rdx){1to8}, %zmm14
+
+// CHECK: vgetexpps %zmm6, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x48,0x42,0xce]
+          vgetexpps %zmm6, %zmm1
+
+// CHECK: vgetexpps %zmm6, %zmm1 {%k3}
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x4b,0x42,0xce]
+          vgetexpps %zmm6, %zmm1 {%k3}
+
+// CHECK: vgetexpps %zmm6, %zmm1 {%k3} {z}
+// CHECK:  encoding: [0x62,0xf2,0x7d,0xcb,0x42,0xce]
+          vgetexpps %zmm6, %zmm1 {%k3} {z}
+
+// CHECK: vgetexpps {sae}, %zmm6, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x18,0x42,0xce]
+          vgetexpps {sae}, %zmm6, %zmm1
+
+// CHECK: vgetexpps (%rcx), %zmm1
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x48,0x42,0x09]
+          vgetexpps (%rcx), %zmm1
+
+// CHECK: vgetexpps 291(%rax,%r14,8), %zmm1
+// CHECK:  encoding: [0x62,0xb2,0x7d,0x48,0x42,0x8c,0xf0,0x23,0x01,0x00,0x00]
+          vgetexpps 291(%rax,%r14,8), %zmm1
+
+// CHECK: vgetexpps (%rcx){1to16}, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x58,0x42,0x09]
+          vgetexpps (%rcx){1to16}, %zmm1
+
+// CHECK: vgetexpps 8128(%rdx), %zmm1
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x48,0x42,0x4a,0x7f]
+          vgetexpps 8128(%rdx), %zmm1
+
+// CHECK: vgetexpps 8192(%rdx), %zmm1
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x48,0x42,0x8a,0x00,0x20,0x00,0x00]
+          vgetexpps 8192(%rdx), %zmm1
+
+// CHECK: vgetexpps -8192(%rdx), %zmm1
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x48,0x42,0x4a,0x80]
+          vgetexpps -8192(%rdx), %zmm1
+
+// CHECK: vgetexpps -8256(%rdx), %zmm1
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x48,0x42,0x8a,0xc0,0xdf,0xff,0xff]
+          vgetexpps -8256(%rdx), %zmm1
+
+// CHECK: vgetexpps 508(%rdx){1to16}, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x58,0x42,0x4a,0x7f]
+          vgetexpps 508(%rdx){1to16}, %zmm1
+
+// CHECK: vgetexpps 512(%rdx){1to16}, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x58,0x42,0x8a,0x00,0x02,0x00,0x00]
+          vgetexpps 512(%rdx){1to16}, %zmm1
+
+// CHECK: vgetexpps -512(%rdx){1to16}, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x58,0x42,0x4a,0x80]
+          vgetexpps -512(%rdx){1to16}, %zmm1
+
+// CHECK: vgetexpps -516(%rdx){1to16}, %zmm1
+// CHECK:  encoding: [0x62,0xf2,0x7d,0x58,0x42,0x8a,0xfc,0xfd,0xff,0xff]
+          vgetexpps -516(%rdx){1to16}, %zmm1
+

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

 // CHECK: vbroadcasti32x4 -2048(%rdx), %ymm19
 // CHECK:  encoding: [0x62,0xe2,0x7d,0x28,0x5a,0x5a,0x80]
           vbroadcasti32x4 -2048(%rdx), %ymm19
 
 // CHECK: vbroadcasti32x4 -2064(%rdx), %ymm19
 // CHECK:  encoding: [0x62,0xe2,0x7d,0x28,0x5a,0x9a,0xf0,0xf7,0xff,0xff]
           vbroadcasti32x4 -2064(%rdx), %ymm19
 
-
-
-
-
-
-
-
-
-
+// CHECK: vgetexppd %xmm18, %xmm17
+// CHECK:  encoding: [0x62,0xa2,0xfd,0x08,0x42,0xca]
+          vgetexppd %xmm18, %xmm17
+
+// CHECK: vgetexppd %xmm18, %xmm17 {%k1}
+// CHECK:  encoding: [0x62,0xa2,0xfd,0x09,0x42,0xca]
+          vgetexppd %xmm18, %xmm17 {%k1}
+
+// CHECK: vgetexppd %xmm18, %xmm17 {%k1} {z}
+// CHECK:  encoding: [0x62,0xa2,0xfd,0x89,0x42,0xca]
+          vgetexppd %xmm18, %xmm17 {%k1} {z}
+
+// CHECK: vgetexppd (%rcx), %xmm17
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x08,0x42,0x09]
+          vgetexppd (%rcx), %xmm17
+
+// CHECK: vgetexppd 291(%rax,%r14,8), %xmm17
+// CHECK:  encoding: [0x62,0xa2,0xfd,0x08,0x42,0x8c,0xf0,0x23,0x01,0x00,0x00]
+          vgetexppd 291(%rax,%r14,8), %xmm17
+
+// CHECK: vgetexppd (%rcx){1to2}, %xmm17
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x18,0x42,0x09]
+          vgetexppd (%rcx){1to2}, %xmm17
+
+// CHECK: vgetexppd 2032(%rdx), %xmm17
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x08,0x42,0x4a,0x7f]
+          vgetexppd 2032(%rdx), %xmm17
+
+// CHECK: vgetexppd 2048(%rdx), %xmm17
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x08,0x42,0x8a,0x00,0x08,0x00,0x00]
+          vgetexppd 2048(%rdx), %xmm17
+
+// CHECK: vgetexppd -2048(%rdx), %xmm17
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x08,0x42,0x4a,0x80]
+          vgetexppd -2048(%rdx), %xmm17
+
+// CHECK: vgetexppd -2064(%rdx), %xmm17
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x08,0x42,0x8a,0xf0,0xf7,0xff,0xff]
+          vgetexppd -2064(%rdx), %xmm17
+
+// CHECK: vgetexppd 1016(%rdx){1to2}, %xmm17
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x18,0x42,0x4a,0x7f]
+          vgetexppd 1016(%rdx){1to2}, %xmm17
+
+// CHECK: vgetexppd 1024(%rdx){1to2}, %xmm17
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x18,0x42,0x8a,0x00,0x04,0x00,0x00]
+          vgetexppd 1024(%rdx){1to2}, %xmm17
+
+// CHECK: vgetexppd -1024(%rdx){1to2}, %xmm17
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x18,0x42,0x4a,0x80]
+          vgetexppd -1024(%rdx){1to2}, %xmm17
+
+// CHECK: vgetexppd -1032(%rdx){1to2}, %xmm17
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x18,0x42,0x8a,0xf8,0xfb,0xff,0xff]
+          vgetexppd -1032(%rdx){1to2}, %xmm17
+
+// CHECK: vgetexppd %ymm17, %ymm20
+// CHECK:  encoding: [0x62,0xa2,0xfd,0x28,0x42,0xe1]
+          vgetexppd %ymm17, %ymm20
+
+// CHECK: vgetexppd %ymm17, %ymm20 {%k3}
+// CHECK:  encoding: [0x62,0xa2,0xfd,0x2b,0x42,0xe1]
+          vgetexppd %ymm17, %ymm20 {%k3}
+
+// CHECK: vgetexppd %ymm17, %ymm20 {%k3} {z}
+// CHECK:  encoding: [0x62,0xa2,0xfd,0xab,0x42,0xe1]
+          vgetexppd %ymm17, %ymm20 {%k3} {z}
+
+// CHECK: vgetexppd (%rcx), %ymm20
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x28,0x42,0x21]
+          vgetexppd (%rcx), %ymm20
+
+// CHECK: vgetexppd 291(%rax,%r14,8), %ymm20
+// CHECK:  encoding: [0x62,0xa2,0xfd,0x28,0x42,0xa4,0xf0,0x23,0x01,0x00,0x00]
+          vgetexppd 291(%rax,%r14,8), %ymm20
+
+// CHECK: vgetexppd (%rcx){1to4}, %ymm20
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x38,0x42,0x21]
+          vgetexppd (%rcx){1to4}, %ymm20
+
+// CHECK: vgetexppd 4064(%rdx), %ymm20
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x28,0x42,0x62,0x7f]
+          vgetexppd 4064(%rdx), %ymm20
+
+// CHECK: vgetexppd 4096(%rdx), %ymm20
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x28,0x42,0xa2,0x00,0x10,0x00,0x00]
+          vgetexppd 4096(%rdx), %ymm20
+
+// CHECK: vgetexppd -4096(%rdx), %ymm20
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x28,0x42,0x62,0x80]
+          vgetexppd -4096(%rdx), %ymm20
+
+// CHECK: vgetexppd -4128(%rdx), %ymm20
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x28,0x42,0xa2,0xe0,0xef,0xff,0xff]
+          vgetexppd -4128(%rdx), %ymm20
+
+// CHECK: vgetexppd 1016(%rdx){1to4}, %ymm20
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x38,0x42,0x62,0x7f]
+          vgetexppd 1016(%rdx){1to4}, %ymm20
+
+// CHECK: vgetexppd 1024(%rdx){1to4}, %ymm20
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x38,0x42,0xa2,0x00,0x04,0x00,0x00]
+          vgetexppd 1024(%rdx){1to4}, %ymm20
+
+// CHECK: vgetexppd -1024(%rdx){1to4}, %ymm20
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x38,0x42,0x62,0x80]
+          vgetexppd -1024(%rdx){1to4}, %ymm20
+
+// CHECK: vgetexppd -1032(%rdx){1to4}, %ymm20
+// CHECK:  encoding: [0x62,0xe2,0xfd,0x38,0x42,0xa2,0xf8,0xfb,0xff,0xff]
+          vgetexppd -1032(%rdx){1to4}, %ymm20
+
+// CHECK: vgetexpps %xmm27, %xmm17
+// CHECK:  encoding: [0x62,0x82,0x7d,0x08,0x42,0xcb]
+          vgetexpps %xmm27, %xmm17
+
+// CHECK: vgetexpps %xmm27, %xmm17 {%k2}
+// CHECK:  encoding: [0x62,0x82,0x7d,0x0a,0x42,0xcb]
+          vgetexpps %xmm27, %xmm17 {%k2}
+
+// CHECK: vgetexpps %xmm27, %xmm17 {%k2} {z}
+// CHECK:  encoding: [0x62,0x82,0x7d,0x8a,0x42,0xcb]
+          vgetexpps %xmm27, %xmm17 {%k2} {z}
+
+// CHECK: vgetexpps (%rcx), %xmm17
+// CHECK:  encoding: [0x62,0xe2,0x7d,0x08,0x42,0x09]
+          vgetexpps (%rcx), %xmm17
+
+// CHECK: vgetexpps 291(%rax,%r14,8), %xmm17
+// CHECK:  encoding: [0x62,0xa2,0x7d,0x08,0x42,0x8c,0xf0,0x23,0x01,0x00,0x00]
+          vgetexpps 291(%rax,%r14,8), %xmm17
+
+// CHECK: vgetexpps (%rcx){1to4}, %xmm17
+// CHECK:  encoding: [0x62,0xe2,0x7d,0x18,0x42,0x09]
+          vgetexpps (%rcx){1to4}, %xmm17
+
+// CHECK: vgetexpps 2032(%rdx), %xmm17
+// CHECK:  encoding: [0x62,0xe2,0x7d,0x08,0x42,0x4a,0x7f]
+          vgetexpps 2032(%rdx), %xmm17
+
+// CHECK: vgetexpps 2048(%rdx), %xmm17
+// CHECK:  encoding: [0x62,0xe2,0x7d,0x08,0x42,0x8a,0x00,0x08,0x00,0x00]
+          vgetexpps 2048(%rdx), %xmm17
+
+// CHECK: vgetexpps -2048(%rdx), %xmm17
+// CHECK:  encoding: [0x62,0xe2,0x7d,0x08,0x42,0x4a,0x80]
+          vgetexpps -2048(%rdx), %xmm17
+
+// CHECK: vgetexpps -2064(%rdx), %xmm17
+// CHECK:  encoding: [0x62,0xe2,0x7d,0x08,0x42,0x8a,0xf0,0xf7,0xff,0xff]
+          vgetexpps -2064(%rdx), %xmm17
+
+// CHECK: vgetexpps 508(%rdx){1to4}, %xmm17
+// CHECK:  encoding: [0x62,0xe2,0x7d,0x18,0x42,0x4a,0x7f]
+          vgetexpps 508(%rdx){1to4}, %xmm17
+
+// CHECK: vgetexpps 512(%rdx){1to4}, %xmm17
+// CHECK:  encoding: [0x62,0xe2,0x7d,0x18,0x42,0x8a,0x00,0x02,0x00,0x00]
+          vgetexpps 512(%rdx){1to4}, %xmm17
+
+// CHECK: vgetexpps -512(%rdx){1to4}, %xmm17
+// CHECK:  encoding: [0x62,0xe2,0x7d,0x18,0x42,0x4a,0x80]
+          vgetexpps -512(%rdx){1to4}, %xmm17
+
+// CHECK: vgetexpps -516(%rdx){1to4}, %xmm17
+// CHECK:  encoding: [0x62,0xe2,0x7d,0x18,0x42,0x8a,0xfc,0xfd,0xff,0xff]
+          vgetexpps -516(%rdx){1to4}, %xmm17
+
+// CHECK: vgetexpps %ymm29, %ymm30
+// CHECK:  encoding: [0x62,0x02,0x7d,0x28,0x42,0xf5]
+          vgetexpps %ymm29, %ymm30
+
+// CHECK: vgetexpps %ymm29, %ymm30 {%k6}
+// CHECK:  encoding: [0x62,0x02,0x7d,0x2e,0x42,0xf5]
+          vgetexpps %ymm29, %ymm30 {%k6}
+
+// CHECK: vgetexpps %ymm29, %ymm30 {%k6} {z}
+// CHECK:  encoding: [0x62,0x02,0x7d,0xae,0x42,0xf5]
+          vgetexpps %ymm29, %ymm30 {%k6} {z}
+
+// CHECK: vgetexpps (%rcx), %ymm30
+// CHECK:  encoding: [0x62,0x62,0x7d,0x28,0x42,0x31]
+          vgetexpps (%rcx), %ymm30
+
+// CHECK: vgetexpps 291(%rax,%r14,8), %ymm30
+// CHECK:  encoding: [0x62,0x22,0x7d,0x28,0x42,0xb4,0xf0,0x23,0x01,0x00,0x00]
+          vgetexpps 291(%rax,%r14,8), %ymm30
+
+// CHECK: vgetexpps (%rcx){1to8}, %ymm30
+// CHECK:  encoding: [0x62,0x62,0x7d,0x38,0x42,0x31]
+          vgetexpps (%rcx){1to8}, %ymm30
+
+// CHECK: vgetexpps 4064(%rdx), %ymm30
+// CHECK:  encoding: [0x62,0x62,0x7d,0x28,0x42,0x72,0x7f]
+          vgetexpps 4064(%rdx), %ymm30
+
+// CHECK: vgetexpps 4096(%rdx), %ymm30
+// CHECK:  encoding: [0x62,0x62,0x7d,0x28,0x42,0xb2,0x00,0x10,0x00,0x00]
+          vgetexpps 4096(%rdx), %ymm30
+
+// CHECK: vgetexpps -4096(%rdx), %ymm30
+// CHECK:  encoding: [0x62,0x62,0x7d,0x28,0x42,0x72,0x80]
+          vgetexpps -4096(%rdx), %ymm30
+
+// CHECK: vgetexpps -4128(%rdx), %ymm30
+// CHECK:  encoding: [0x62,0x62,0x7d,0x28,0x42,0xb2,0xe0,0xef,0xff,0xff]
+          vgetexpps -4128(%rdx), %ymm30
+
+// CHECK: vgetexpps 508(%rdx){1to8}, %ymm30
+// CHECK:  encoding: [0x62,0x62,0x7d,0x38,0x42,0x72,0x7f]
+          vgetexpps 508(%rdx){1to8}, %ymm30
+
+// CHECK: vgetexpps 512(%rdx){1to8}, %ymm30
+// CHECK:  encoding: [0x62,0x62,0x7d,0x38,0x42,0xb2,0x00,0x02,0x00,0x00]
+          vgetexpps 512(%rdx){1to8}, %ymm30
+
+// CHECK: vgetexpps -512(%rdx){1to8}, %ymm30
+// CHECK:  encoding: [0x62,0x62,0x7d,0x38,0x42,0x72,0x80]
+          vgetexpps -512(%rdx){1to8}, %ymm30
+
+// CHECK: vgetexpps -516(%rdx){1to8}, %ymm30
+// CHECK:  encoding: [0x62,0x62,0x7d,0x38,0x42,0xb2,0xfc,0xfd,0xff,0xff]
+          vgetexpps -516(%rdx){1to8}, %ymm30