[X86] add RSQRT14 & RCP14 Scalar FP

[AsafBadouh]

Summary

add mask version include intrinsics

[AsafBadouh]

add encoding tests

[delena]

../llvmOrg/lib/Target/X86/X86InstrFragmentsSIMD.td

61 ↗

(On Diff #33681)

Can we use the same FRSQRT/FRCP for the scalar node?

llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

case INTR_TYPE_1OP_MASK: {
                                       dl, Op.getValueType(),
                                       Src, Rnd),
                                       Mask, PassThru, Subtarget, DAG);
         }
       }
       return getVectorMaskingNode(DAG.getNode(IntrData->Opc0, dl, VT, Src),
                                   Mask, PassThru, Subtarget, DAG);
     }
+    case INTR_TYPE_SCALAR_MASK: {
+      SDValue Src1 = Op.getOperand(1);
+      SDValue Src2 = Op.getOperand(2);
+      SDValue passThru = Op.getOperand(3);
+      SDValue Mask = Op.getOperand(4);
+      return getScalarMaskingNode(DAG.getNode(IntrData->Opc0, dl, VT, Src1, Src2),
+                                  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 suppress-all-exceptions (sae) or rounding mode- 6 operands
       // (2) With rounding mode and sae - 7 operands.
case FMA_OP_MASK: {
       }
       return getVectorMaskingNode(DAG.getNode(IntrData->Opc0,
                                               dl, Op.getValueType(),
                                               Src1, Src2, Src3),
                                   Mask, PassThru, Subtarget, DAG);
     }
     case FPCLASS: {
       // FPclass intrinsics with mask
-      //
        SDValue Src1 = Op.getOperand(1);
        EVT VT = Src1.getValueType();
        EVT MaskVT = EVT::getVectorVT(*DAG.getContext(), MVT::i1,
                                       VT.getVectorNumElements());
        SDValue Imm = Op.getOperand(2);
        SDValue Mask = Op.getOperand(3);
        EVT BitcastVT = EVT::getVectorVT(*DAG.getContext(), MVT::i1,
                                         Mask.getValueType().getSizeInBits());

llvm/trunk/lib/Target/X86/X86InstrAVX512.td

defm Int_VCOMISSZ  : sse12_ord_cmp<0x2F, VR128X, X86comi, v4f32, f128mem,
                               EVEX_CD8<32, CD8VT1>;
     defm Int_VCOMISDZ  : sse12_ord_cmp<0x2F, VR128X, X86comi, v2f64, f128mem,
                               load, "comisd">, PD, EVEX,
                               VEX_LIG, VEX_W, EVEX_CD8<64, CD8VT1>;
   }
 }
 
 /// avx512_fp14_s rcp14ss, rcp14sd, rsqrt14ss, rsqrt14sd
-multiclass avx512_fp14_s<bits<8> opc, string OpcodeStr, RegisterClass RC,
-                            X86MemOperand x86memop> {
-  let hasSideEffects = 0 in {
-  def rr : AVX5128I<opc, MRMSrcReg, (outs RC:$dst),
-               (ins RC:$src1, RC:$src2),
-               !strconcat(OpcodeStr,
-               "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>, EVEX_4V;
+multiclass avx512_fp14_s<bits<8> opc, string OpcodeStr, SDNode OpNode,
+                            X86VectorVTInfo _> {
+  let hasSideEffects = 0, AddedComplexity = 20 , Predicates = [HasAVX512] in {
+  defm rr : AVX512_maskable_scalar<opc, MRMSrcReg, _, (outs _.RC:$dst),
+                           (ins _.RC:$src1, _.RC:$src2), OpcodeStr,
+                           "$src2, $src1", "$src1, $src2",
+                           (OpNode (_.VT _.RC:$src1), (_.VT _.RC:$src2))>, EVEX_4V;
   let mayLoad = 1 in {
-  def rm : AVX5128I<opc, MRMSrcMem, (outs RC:$dst),
-               (ins RC:$src1, x86memop:$src2),
-               !strconcat(OpcodeStr,
-               "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>, EVEX_4V;
+  defm rm : AVX512_maskable_scalar<opc, MRMSrcMem, _, (outs _.RC:$dst),
+                         (ins _.RC:$src1, _.MemOp:$src2), OpcodeStr,
+                         "$src2, $src1", "$src1, $src2",
+                         (OpNode (_.VT _.RC:$src1),
+                          (_.VT (scalar_to_vector (_.ScalarLdFrag addr:$src2))))>, EVEX_4V;
   }
 }
 }
 
-defm VRCP14SS   : avx512_fp14_s<0x4D, "vrcp14ss", FR32X, f32mem>,
-                  EVEX_CD8<32, CD8VT1>;
-defm VRCP14SD   : avx512_fp14_s<0x4D, "vrcp14sd", FR64X, f64mem>,
-                  VEX_W, EVEX_CD8<64, CD8VT1>;
-defm VRSQRT14SS   : avx512_fp14_s<0x4F, "vrsqrt14ss", FR32X, f32mem>,
-                  EVEX_CD8<32, CD8VT1>;
-defm VRSQRT14SD   : avx512_fp14_s<0x4F, "vrsqrt14sd", FR64X, f64mem>,
-                  VEX_W, EVEX_CD8<64, CD8VT1>;
-
-def : Pat <(v4f32 (int_x86_avx512_rcp14_ss (v4f32 VR128X:$src1),
-              (v4f32 VR128X:$src2), (bc_v4f32 (v4i32 immAllZerosV)), (i8 -1))),
-           (COPY_TO_REGCLASS (VRCP14SSrr (COPY_TO_REGCLASS VR128X:$src1, FR32X),
-                       (COPY_TO_REGCLASS VR128X:$src2, FR32X)), VR128X)>;
-
-def : Pat <(v2f64 (int_x86_avx512_rcp14_sd (v2f64 VR128X:$src1),
-              (v2f64 VR128X:$src2), (bc_v2f64 (v4i32 immAllZerosV)), (i8 -1))),
-           (COPY_TO_REGCLASS (VRCP14SDrr (COPY_TO_REGCLASS VR128X:$src1, FR64X),
-                       (COPY_TO_REGCLASS VR128X:$src2, FR64X)), VR128X)>;
-
-def : Pat <(v4f32 (int_x86_avx512_rsqrt14_ss (v4f32 VR128X:$src1),
-              (v4f32 VR128X:$src2), (bc_v4f32 (v4i32 immAllZerosV)), (i8 -1))),
-           (COPY_TO_REGCLASS (VRSQRT14SSrr (COPY_TO_REGCLASS VR128X:$src1, FR32X),
-                       (COPY_TO_REGCLASS VR128X:$src2, FR32X)), VR128X)>;
-
-def : Pat <(v2f64 (int_x86_avx512_rsqrt14_sd (v2f64 VR128X:$src1),
-              (v2f64 VR128X:$src2), (bc_v2f64 (v4i32 immAllZerosV)), (i8 -1))),
-           (COPY_TO_REGCLASS (VRSQRT14SDrr (COPY_TO_REGCLASS VR128X:$src1, FR64X),
-                       (COPY_TO_REGCLASS VR128X:$src2, FR64X)), VR128X)>;
+defm VRCP14SS   : avx512_fp14_s<0x4D, "vrcp14ss", X86frcp14s, f32x_info>,
+                  EVEX_CD8<32, CD8VT1>, T8PD;
+defm VRCP14SD   : avx512_fp14_s<0x4D, "vrcp14sd", X86frcp14s, f64x_info>,
+                  VEX_W, EVEX_CD8<64, CD8VT1>, T8PD;
+defm VRSQRT14SS   : avx512_fp14_s<0x4F, "vrsqrt14ss", X86frsqrt14s, f32x_info>,
+                  EVEX_CD8<32, CD8VT1>, T8PD;
+defm VRSQRT14SD   : avx512_fp14_s<0x4F, "vrsqrt14sd", X86frsqrt14s, f64x_info>,
+                  VEX_W, EVEX_CD8<64, CD8VT1>, T8PD;
 
 /// avx512_fp14_p rcp14ps, rcp14pd, rsqrt14ps, rsqrt14pd
 multiclass avx512_fp14_p<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, T8PD;
   let mayLoad = 1 in {

 
 defm VSQRT   : avx512_sqrt_packed_all<0x51, "vsqrt", fsqrt>,
                avx512_sqrt_packed_all_round<0x51, "vsqrt", X86fsqrtRnd>;
 
 defm VSQRT   : avx512_sqrt_scalar_all<0x51, "vsqrt">, VEX_LIG;
 
 let Predicates = [HasAVX512] in {
   def : Pat<(f32 (X86frsqrt FR32X:$src)),
-            (VRSQRT14SSrr (f32 (IMPLICIT_DEF)), FR32X:$src)>;
+            (COPY_TO_REGCLASS (VRSQRT14SSrr (v4f32 (IMPLICIT_DEF)), (COPY_TO_REGCLASS FR32X:$src, VR128X)), VR128X)>;
   def : Pat<(f32 (X86frsqrt (load addr:$src))),
-            (VRSQRT14SSrm (f32 (IMPLICIT_DEF)), addr:$src)>,
+            (COPY_TO_REGCLASS (VRSQRT14SSrm (v4f32 (IMPLICIT_DEF)), addr:$src), VR128X)>,
             Requires<[OptForSize]>;
-
   def : Pat<(f32 (X86frcp FR32X:$src)),
-            (VRCP14SSrr (f32 (IMPLICIT_DEF)), FR32X:$src)>;
+            (COPY_TO_REGCLASS (VRCP14SSrr (v4f32 (IMPLICIT_DEF)), (COPY_TO_REGCLASS FR32X:$src, VR128X)), VR128X )>;
   def : Pat<(f32 (X86frcp (load addr:$src))),
-            (VRCP14SSrm (f32 (IMPLICIT_DEF)), addr:$src)>,
+            (COPY_TO_REGCLASS (VRCP14SSrm (v4f32 (IMPLICIT_DEF)), addr:$src), VR128X)>,
             Requires<[OptForSize]>;
 }
 
 multiclass
 avx512_rndscale_scalar<bits<8> opc, string OpcodeStr, X86VectorVTInfo _> {
 
   let ExeDomain = _.ExeDomain in {
   defm r : AVX512_maskable_scalar<opc, MRMSrcReg, _, (outs _.RC:$dst),

llvm/trunk/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 X86frsqrt14s: SDNode<"X86ISD::FRSQRT",  SDTFPBinOp>;
+def X86frcp14s : SDNode<"X86ISD::FRCP",    SDTFPBinOp>;
 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 X86cmps    : SDNode<"X86ISD::FSETCC",     SDTX86Cmps>;

llvm/trunk/lib/Target/X86/X86IntrinsicsInfo.h

enum IntrinsicType {
   INTR_NO_TYPE,
   GATHER, SCATTER, PREFETCH, RDSEED, RDRAND, RDPMC, RDTSC, XTEST, ADX, FPCLASS,
   INTR_TYPE_1OP, INTR_TYPE_2OP, INTR_TYPE_2OP_IMM8, INTR_TYPE_3OP, INTR_TYPE_4OP,
   CMP_MASK, CMP_MASK_CC,CMP_MASK_SCALAR_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_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,
+  VPERM_3OP_MASKZ, INTR_TYPE_SCALAR_MASK,
   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, INSERT_SUBVEC
 };
 
 struct IntrinsicData {
 
X86_INTRINSIC_DATA(avx512_maskz_vpermt2var_q_128, VPERM_3OP_MASKZ,
                      X86ISD::VPERMV3, 0),
   X86_INTRINSIC_DATA(avx512_maskz_vpermt2var_q_256, VPERM_3OP_MASKZ,
                      X86ISD::VPERMV3, 0),
   X86_INTRINSIC_DATA(avx512_maskz_vpermt2var_q_512, VPERM_3OP_MASKZ,
                      X86ISD::VPERMV3, 0),
   X86_INTRINSIC_DATA(avx512_psad_bw_512, INTR_TYPE_2OP, X86ISD::PSADBW, 0),
   X86_INTRINSIC_DATA(avx512_psll_dq_512, INTR_TYPE_2OP_IMM8, X86ISD::VSHLDQ, 0),
   X86_INTRINSIC_DATA(avx512_psrl_dq_512, INTR_TYPE_2OP_IMM8, X86ISD::VSRLDQ, 0),
+  X86_INTRINSIC_DATA(avx512_rcp14_sd, INTR_TYPE_SCALAR_MASK, X86ISD::FRCP, 0),
+  X86_INTRINSIC_DATA(avx512_rcp14_ss, INTR_TYPE_SCALAR_MASK, X86ISD::FRCP, 0),
   X86_INTRINSIC_DATA(avx512_rcp28_pd,   INTR_TYPE_1OP_MASK_RM,X86ISD::RCP28, 0),
   X86_INTRINSIC_DATA(avx512_rcp28_ps,   INTR_TYPE_1OP_MASK_RM,X86ISD::RCP28, 0),
   X86_INTRINSIC_DATA(avx512_rcp28_sd,   INTR_TYPE_SCALAR_MASK_RM, X86ISD::RCP28, 0),
   X86_INTRINSIC_DATA(avx512_rcp28_ss,   INTR_TYPE_SCALAR_MASK_RM, X86ISD::RCP28, 0),
+  X86_INTRINSIC_DATA(avx512_rsqrt14_sd, INTR_TYPE_SCALAR_MASK, X86ISD::FRSQRT, 0),
+  X86_INTRINSIC_DATA(avx512_rsqrt14_ss, INTR_TYPE_SCALAR_MASK, X86ISD::FRSQRT, 0),
   X86_INTRINSIC_DATA(avx512_rsqrt28_pd, INTR_TYPE_1OP_MASK_RM,X86ISD::RSQRT28, 0),
   X86_INTRINSIC_DATA(avx512_rsqrt28_ps, INTR_TYPE_1OP_MASK_RM,X86ISD::RSQRT28, 0),
   X86_INTRINSIC_DATA(avx512_rsqrt28_sd, INTR_TYPE_SCALAR_MASK_RM,X86ISD::RSQRT28, 0),
   X86_INTRINSIC_DATA(avx512_rsqrt28_ss, INTR_TYPE_SCALAR_MASK_RM,X86ISD::RSQRT28, 0),
   X86_INTRINSIC_DATA(avx_hadd_pd_256,   INTR_TYPE_2OP, X86ISD::FHADD, 0),
   X86_INTRINSIC_DATA(avx_hadd_ps_256,   INTR_TYPE_2OP, X86ISD::FHADD, 0),
   X86_INTRINSIC_DATA(avx_hsub_pd_256,   INTR_TYPE_2OP, X86ISD::FHSUB, 0),
   X86_INTRINSIC_DATA(avx_hsub_ps_256,   INTR_TYPE_2OP, X86ISD::FHSUB, 0),

llvm/trunk/test/MC/X86/avx512-encodings.s

 
 // CHECK: vcvttss2usi -512(%rdx), %r8
 // CHECK:  encoding: [0x62,0x71,0xfe,0x08,0x78,0x42,0x80]
           vcvttss2usi -512(%rdx), %r8
 
 // CHECK: vcvttss2usi -516(%rdx), %r8
 // CHECK:  encoding: [0x62,0x71,0xfe,0x08,0x78,0x82,0xfc,0xfd,0xff,0xff]
           vcvttss2usi -516(%rdx), %r8
+// CHECK: vrsqrt14sd %xmm10, %xmm6, %xmm26
+// CHECK:  encoding: [0x62,0x42,0xcd,0x08,0x4f,0xd2]
+          vrsqrt14sd %xmm10, %xmm6, %xmm26
+
+// CHECK: vrsqrt14sd %xmm10, %xmm6, %xmm26 {%k5}
+// CHECK:  encoding: [0x62,0x42,0xcd,0x0d,0x4f,0xd2]
+          vrsqrt14sd %xmm10, %xmm6, %xmm26 {%k5}
+
+// CHECK: vrsqrt14sd %xmm10, %xmm6, %xmm26 {%k5} {z}
+// CHECK:  encoding: [0x62,0x42,0xcd,0x8d,0x4f,0xd2]
+          vrsqrt14sd %xmm10, %xmm6, %xmm26 {%k5} {z}
+
+// CHECK: vrsqrt14sd (%rcx), %xmm6, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xcd,0x08,0x4f,0x11]
+          vrsqrt14sd (%rcx), %xmm6, %xmm26
+
+// CHECK: vrsqrt14sd 291(%rax,%r14,8), %xmm6, %xmm26
+// CHECK:  encoding: [0x62,0x22,0xcd,0x08,0x4f,0x94,0xf0,0x23,0x01,0x00,0x00]
+          vrsqrt14sd 291(%rax,%r14,8), %xmm6, %xmm26
+
+// CHECK: vrsqrt14sd 1016(%rdx), %xmm6, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xcd,0x08,0x4f,0x52,0x7f]
+          vrsqrt14sd 1016(%rdx), %xmm6, %xmm26
+
+// CHECK: vrsqrt14sd 1024(%rdx), %xmm6, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xcd,0x08,0x4f,0x92,0x00,0x04,0x00,0x00]
+          vrsqrt14sd 1024(%rdx), %xmm6, %xmm26
+
+// CHECK: vrsqrt14sd -1024(%rdx), %xmm6, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xcd,0x08,0x4f,0x52,0x80]
+          vrsqrt14sd -1024(%rdx), %xmm6, %xmm26
+
+// CHECK: vrsqrt14sd -1032(%rdx), %xmm6, %xmm26
+// CHECK:  encoding: [0x62,0x62,0xcd,0x08,0x4f,0x92,0xf8,0xfb,0xff,0xff]
+          vrsqrt14sd -1032(%rdx), %xmm6, %xmm26
+
+// CHECK: vrsqrt14ss %xmm9, %xmm14, %xmm14
+// CHECK:  encoding: [0x62,0x52,0x0d,0x08,0x4f,0xf1]
+          vrsqrt14ss %xmm9, %xmm14, %xmm14
+
+// CHECK: vrsqrt14ss %xmm9, %xmm14, %xmm14 {%k1}
+// CHECK:  encoding: [0x62,0x52,0x0d,0x09,0x4f,0xf1]
+          vrsqrt14ss %xmm9, %xmm14, %xmm14 {%k1}
+
+// CHECK: vrsqrt14ss %xmm9, %xmm14, %xmm14 {%k1} {z}
+// CHECK:  encoding: [0x62,0x52,0x0d,0x89,0x4f,0xf1]
+          vrsqrt14ss %xmm9, %xmm14, %xmm14 {%k1} {z}
+
+// CHECK: vrsqrt14ss (%rcx), %xmm14, %xmm14
+// CHECK:  encoding: [0x62,0x72,0x0d,0x08,0x4f,0x31]
+          vrsqrt14ss (%rcx), %xmm14, %xmm14
+
+// CHECK: vrsqrt14ss 291(%rax,%r14,8), %xmm14, %xmm14
+// CHECK:  encoding: [0x62,0x32,0x0d,0x08,0x4f,0xb4,0xf0,0x23,0x01,0x00,0x00]
+          vrsqrt14ss 291(%rax,%r14,8), %xmm14, %xmm14
+
+// CHECK: vrsqrt14ss 508(%rdx), %xmm14, %xmm14
+// CHECK:  encoding: [0x62,0x72,0x0d,0x08,0x4f,0x72,0x7f]
+          vrsqrt14ss 508(%rdx), %xmm14, %xmm14
+
+// CHECK: vrsqrt14ss 512(%rdx), %xmm14, %xmm14
+// CHECK:  encoding: [0x62,0x72,0x0d,0x08,0x4f,0xb2,0x00,0x02,0x00,0x00]
+          vrsqrt14ss 512(%rdx), %xmm14, %xmm14
+
+// CHECK: vrsqrt14ss -512(%rdx), %xmm14, %xmm14
+// CHECK:  encoding: [0x62,0x72,0x0d,0x08,0x4f,0x72,0x80]
+          vrsqrt14ss -512(%rdx), %xmm14, %xmm14
+
+// CHECK: vrsqrt14ss -516(%rdx), %xmm14, %xmm14
+// CHECK:  encoding: [0x62,0x72,0x0d,0x08,0x4f,0xb2,0xfc,0xfd,0xff,0xff]
+          vrsqrt14ss -516(%rdx), %xmm14, %xmm14
+
+// CHECK: vrcp14sd %xmm14, %xmm22, %xmm12
+// CHECK:  encoding: [0x62,0x52,0xcd,0x00,0x4d,0xe6]
+          vrcp14sd %xmm14, %xmm22, %xmm12
+
+// CHECK: vrcp14sd %xmm14, %xmm22, %xmm12 {%k2}
+// CHECK:  encoding: [0x62,0x52,0xcd,0x02,0x4d,0xe6]
+          vrcp14sd %xmm14, %xmm22, %xmm12 {%k2}
+
+// CHECK: vrcp14sd %xmm14, %xmm22, %xmm12 {%k2} {z}
+// CHECK:  encoding: [0x62,0x52,0xcd,0x82,0x4d,0xe6]
+          vrcp14sd %xmm14, %xmm22, %xmm12 {%k2} {z}
+
+// CHECK: vrcp14sd (%rcx), %xmm22, %xmm12
+// CHECK:  encoding: [0x62,0x72,0xcd,0x00,0x4d,0x21]
+          vrcp14sd (%rcx), %xmm22, %xmm12
+
+// CHECK: vrcp14sd 291(%rax,%r14,8), %xmm22, %xmm12
+// CHECK:  encoding: [0x62,0x32,0xcd,0x00,0x4d,0xa4,0xf0,0x23,0x01,0x00,0x00]
+          vrcp14sd 291(%rax,%r14,8), %xmm22, %xmm12
+
+// CHECK: vrcp14sd 1016(%rdx), %xmm22, %xmm12
+// CHECK:  encoding: [0x62,0x72,0xcd,0x00,0x4d,0x62,0x7f]
+          vrcp14sd 1016(%rdx), %xmm22, %xmm12
+
+// CHECK: vrcp14sd 1024(%rdx), %xmm22, %xmm12
+// CHECK:  encoding: [0x62,0x72,0xcd,0x00,0x4d,0xa2,0x00,0x04,0x00,0x00]
+          vrcp14sd 1024(%rdx), %xmm22, %xmm12
+
+// CHECK: vrcp14sd -1024(%rdx), %xmm22, %xmm12
+// CHECK:  encoding: [0x62,0x72,0xcd,0x00,0x4d,0x62,0x80]
+          vrcp14sd -1024(%rdx), %xmm22, %xmm12
+
+// CHECK: vrcp14sd -1032(%rdx), %xmm22, %xmm12
+// CHECK:  encoding: [0x62,0x72,0xcd,0x00,0x4d,0xa2,0xf8,0xfb,0xff,0xff]
+          vrcp14sd -1032(%rdx), %xmm22, %xmm12
+
+// CHECK: vrcp14ss %xmm3, %xmm8, %xmm8
+// CHECK:  encoding: [0x62,0x72,0x3d,0x08,0x4d,0xc3]
+          vrcp14ss %xmm3, %xmm8, %xmm8
+
+// CHECK: vrcp14ss %xmm3, %xmm8, %xmm8 {%k7}
+// CHECK:  encoding: [0x62,0x72,0x3d,0x0f,0x4d,0xc3]
+          vrcp14ss %xmm3, %xmm8, %xmm8 {%k7}
+
+// CHECK: vrcp14ss %xmm3, %xmm8, %xmm8 {%k7} {z}
+// CHECK:  encoding: [0x62,0x72,0x3d,0x8f,0x4d,0xc3]
+          vrcp14ss %xmm3, %xmm8, %xmm8 {%k7} {z}
+
+// CHECK: vrcp14ss (%rcx), %xmm8, %xmm8
+// CHECK:  encoding: [0x62,0x72,0x3d,0x08,0x4d,0x01]
+          vrcp14ss (%rcx), %xmm8, %xmm8
+
+// CHECK: vrcp14ss 291(%rax,%r14,8), %xmm8, %xmm8
+// CHECK:  encoding: [0x62,0x32,0x3d,0x08,0x4d,0x84,0xf0,0x23,0x01,0x00,0x00]
+          vrcp14ss 291(%rax,%r14,8), %xmm8, %xmm8
+
+// CHECK: vrcp14ss 508(%rdx), %xmm8, %xmm8
+// CHECK:  encoding: [0x62,0x72,0x3d,0x08,0x4d,0x42,0x7f]
+          vrcp14ss 508(%rdx), %xmm8, %xmm8
+
+// CHECK: vrcp14ss 512(%rdx), %xmm8, %xmm8
+// CHECK:  encoding: [0x62,0x72,0x3d,0x08,0x4d,0x82,0x00,0x02,0x00,0x00]
+          vrcp14ss 512(%rdx), %xmm8, %xmm8
+
+// CHECK: vrcp14ss -512(%rdx), %xmm8, %xmm8
+// CHECK:  encoding: [0x62,0x72,0x3d,0x08,0x4d,0x42,0x80]
+          vrcp14ss -512(%rdx), %xmm8, %xmm8
+
+// CHECK: vrcp14ss -516(%rdx), %xmm8, %xmm8
+// CHECK:  encoding: [0x62,0x72,0x3d,0x08,0x4d,0x82,0xfc,0xfd,0xff,0xff]
+          vrcp14ss -516(%rdx), %xmm8, %xmm8