[AArch64][SVE] Add SVE2 intrinsics for uniform DSP operations

[kmclaughlin]

Summary

Implements the following intrinsics:

• sqrdmlah, sqrdmlsh, sqrdmulh & sqdmulh
• [s|u]hadd, [s|u]hsub, [s|u]rhadd & [s|u]hsubr
• urecpe, ursqrte, sqabs & sqneg

[efriedma]

LGTM

llvm/include/llvm/IR/IntrinsicsAArch64.td

 // Integer arithmetic
 //
 
 def int_aarch64_sve_add   : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_sub   : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_subr  : AdvSIMD_Pred2VectorArg_Intrinsic;
 
 def int_aarch64_sve_pmul       : AdvSIMD_2VectorArg_Intrinsic;
-def int_aarch64_sve_sqdmulh    : AdvSIMD_2VectorArg_Intrinsic;
-def int_aarch64_sve_sqrdmulh   : AdvSIMD_2VectorArg_Intrinsic;
 
 def int_aarch64_sve_mul        : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_smulh      : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_umulh      : AdvSIMD_Pred2VectorArg_Intrinsic;
 
 def int_aarch64_sve_sdiv       : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_udiv       : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_sdivr      : AdvSIMD_Pred2VectorArg_Intrinsic;

 
 //
 // Scatter stores: vector base + scalar offset
 //
 
 def int_aarch64_sve_st1_scatter_scalar_offset : AdvSIMD_ScatterStore_VectorBase_Intrinsic;
 
 //
+// SVE2 - Uniform DSP operations
+//
+
+def int_aarch64_sve_shadd         : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_shsub         : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_shsubr        : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_sqabs         : AdvSIMD_Merged1VectorArg_Intrinsic;
+def int_aarch64_sve_sqdmulh       : AdvSIMD_2VectorArg_Intrinsic;
+def int_aarch64_sve_sqdmulh_lane  : AdvSIMD_2VectorArgIndexed_Intrinsic;
+def int_aarch64_sve_sqneg         : AdvSIMD_Merged1VectorArg_Intrinsic;
+def int_aarch64_sve_sqrdmlah      : AdvSIMD_3VectorArg_Intrinsic;
+def int_aarch64_sve_sqrdmlah_lane : AdvSIMD_3VectorArgIndexed_Intrinsic;
+def int_aarch64_sve_sqrdmlsh      : AdvSIMD_3VectorArg_Intrinsic;
+def int_aarch64_sve_sqrdmlsh_lane : AdvSIMD_3VectorArgIndexed_Intrinsic;
+def int_aarch64_sve_sqrdmulh      : AdvSIMD_2VectorArg_Intrinsic;
+def int_aarch64_sve_sqrdmulh_lane : AdvSIMD_2VectorArgIndexed_Intrinsic;
+def int_aarch64_sve_srhadd        : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_uhadd         : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_uhsub         : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_uhsubr        : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_urecpe        : AdvSIMD_Merged1VectorArg_Intrinsic;
+def int_aarch64_sve_urhadd        : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_ursqrte       : AdvSIMD_Merged1VectorArg_Intrinsic;
+
+//
 // SVE2 - Non-widening pairwise arithmetic
 //
 
 def int_aarch64_sve_addp    : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_faddp   : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fmaxp   : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fmaxnmp : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fminp   : AdvSIMD_Pred2VectorArg_Intrinsic;

llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td

   defm : ldff1<LDFF1H,    nxv8f16, AArch64ldff1,  nxv8i1, nxv8f16>;
 
   // 16-element contiguous first faulting loads
   defm : ldff1<LDFF1B, nxv16i8, AArch64ldff1, nxv16i1, nxv16i8>;
 }
 
 let Predicates = [HasSVE2] in {
   // SVE2 integer multiply-add (indexed)
-  defm MLA_ZZZI : sve2_int_mla_by_indexed_elem<0b01, 0b0, "mla">;
-  defm MLS_ZZZI : sve2_int_mla_by_indexed_elem<0b01, 0b1, "mls">;
+  defm MLA_ZZZI : sve2_int_mla_by_indexed_elem<0b01, 0b0, "mla", null_frag>;
+  defm MLS_ZZZI : sve2_int_mla_by_indexed_elem<0b01, 0b1, "mls", null_frag>;
 
   // SVE2 saturating multiply-add high (indexed)
-  defm SQRDMLAH_ZZZI : sve2_int_mla_by_indexed_elem<0b10, 0b0, "sqrdmlah">;
-  defm SQRDMLSH_ZZZI : sve2_int_mla_by_indexed_elem<0b10, 0b1, "sqrdmlsh">;
+  defm SQRDMLAH_ZZZI : sve2_int_mla_by_indexed_elem<0b10, 0b0, "sqrdmlah", int_aarch64_sve_sqrdmlah_lane>;
+  defm SQRDMLSH_ZZZI : sve2_int_mla_by_indexed_elem<0b10, 0b1, "sqrdmlsh", int_aarch64_sve_sqrdmlsh_lane>;
 
   // SVE2 saturating multiply-add high (vectors, unpredicated)
-  defm SQRDMLAH_ZZZ : sve2_int_mla<0b0, "sqrdmlah">;
-  defm SQRDMLSH_ZZZ : sve2_int_mla<0b1, "sqrdmlsh">;
+  defm SQRDMLAH_ZZZ : sve2_int_mla<0b0, "sqrdmlah", int_aarch64_sve_sqrdmlah>;
+  defm SQRDMLSH_ZZZ : sve2_int_mla<0b1, "sqrdmlsh", int_aarch64_sve_sqrdmlsh>;
 
   // SVE2 integer multiply (indexed)
-  defm MUL_ZZZI : sve2_int_mul_by_indexed_elem<0b1110, "mul">;
+  defm MUL_ZZZI : sve2_int_mul_by_indexed_elem<0b1110, "mul", null_frag>;
 
   // SVE2 saturating multiply high (indexed)
-  defm SQDMULH_ZZZI  : sve2_int_mul_by_indexed_elem<0b1100, "sqdmulh">;
-  defm SQRDMULH_ZZZI : sve2_int_mul_by_indexed_elem<0b1101, "sqrdmulh">;
+  defm SQDMULH_ZZZI  : sve2_int_mul_by_indexed_elem<0b1100, "sqdmulh",  int_aarch64_sve_sqdmulh_lane>;
+  defm SQRDMULH_ZZZI : sve2_int_mul_by_indexed_elem<0b1101, "sqrdmulh", int_aarch64_sve_sqrdmulh_lane>;
 
   // SVE2 signed saturating doubling multiply high (unpredicated)
   defm SQDMULH_ZZZ  : sve2_int_mul<0b100, "sqdmulh",  int_aarch64_sve_sqdmulh>;
   defm SQRDMULH_ZZZ : sve2_int_mul<0b101, "sqrdmulh", int_aarch64_sve_sqrdmulh>;
 
   // SVE2 integer multiply vectors (unpredicated)
   defm MUL_ZZZ    : sve2_int_mul<0b000,  "mul",   mul>;
   defm SMULH_ZZZ  : sve2_int_mul<0b010,  "smulh", null_frag>;
let Predicates = [HasSVE2] in {
   defm SQDMLSLB_ZZZ : sve2_int_mla_long<0b11010, "sqdmlslb">;
   defm SQDMLSLT_ZZZ : sve2_int_mla_long<0b11011, "sqdmlslt">;
 
   // SVE2 saturating multiply-add interleaved long
   defm SQDMLALBT_ZZZ : sve2_int_mla_long<0b00010, "sqdmlalbt">;
   defm SQDMLSLBT_ZZZ : sve2_int_mla_long<0b00011, "sqdmlslbt">;
 
   // SVE2 integer halving add/subtract (predicated)
-  defm SHADD_ZPmZ  : sve2_int_arith_pred<0b100000, "shadd",  null_frag>;
-  defm UHADD_ZPmZ  : sve2_int_arith_pred<0b100010, "uhadd",  null_frag>;
-  defm SHSUB_ZPmZ  : sve2_int_arith_pred<0b100100, "shsub",  null_frag>;
-  defm UHSUB_ZPmZ  : sve2_int_arith_pred<0b100110, "uhsub",  null_frag>;
-  defm SRHADD_ZPmZ : sve2_int_arith_pred<0b101000, "srhadd", null_frag>;
-  defm URHADD_ZPmZ : sve2_int_arith_pred<0b101010, "urhadd", null_frag>;
-  defm SHSUBR_ZPmZ : sve2_int_arith_pred<0b101100, "shsubr", null_frag>;
-  defm UHSUBR_ZPmZ : sve2_int_arith_pred<0b101110, "uhsubr", null_frag>;
+  defm SHADD_ZPmZ  : sve2_int_arith_pred<0b100000, "shadd",  int_aarch64_sve_shadd>;
+  defm UHADD_ZPmZ  : sve2_int_arith_pred<0b100010, "uhadd",  int_aarch64_sve_uhadd>;
+  defm SHSUB_ZPmZ  : sve2_int_arith_pred<0b100100, "shsub",  int_aarch64_sve_shsub>;
+  defm UHSUB_ZPmZ  : sve2_int_arith_pred<0b100110, "uhsub",  int_aarch64_sve_uhsub>;
+  defm SRHADD_ZPmZ : sve2_int_arith_pred<0b101000, "srhadd", int_aarch64_sve_srhadd>;
+  defm URHADD_ZPmZ : sve2_int_arith_pred<0b101010, "urhadd", int_aarch64_sve_urhadd>;
+  defm SHSUBR_ZPmZ : sve2_int_arith_pred<0b101100, "shsubr", int_aarch64_sve_shsubr>;
+  defm UHSUBR_ZPmZ : sve2_int_arith_pred<0b101110, "uhsubr", int_aarch64_sve_uhsubr>;
 
   // SVE2 integer pairwise add and accumulate long
   defm SADALP_ZPmZ : sve2_int_sadd_long_accum_pairwise<0, "sadalp", int_aarch64_sve_sadalp>;
   defm UADALP_ZPmZ : sve2_int_sadd_long_accum_pairwise<1, "uadalp", int_aarch64_sve_uadalp>;
 
   // SVE2 integer pairwise arithmetic
   defm ADDP_ZPmZ  : sve2_int_arith_pred<0b100011, "addp",  int_aarch64_sve_addp>;
   defm SMAXP_ZPmZ : sve2_int_arith_pred<0b101001, "smaxp", int_aarch64_sve_smaxp>;
   defm UMAXP_ZPmZ : sve2_int_arith_pred<0b101011, "umaxp", int_aarch64_sve_umaxp>;
   defm SMINP_ZPmZ : sve2_int_arith_pred<0b101101, "sminp", int_aarch64_sve_sminp>;
   defm UMINP_ZPmZ : sve2_int_arith_pred<0b101111, "uminp", int_aarch64_sve_uminp>;
 
   // SVE2 integer unary operations (predicated)
-  defm URECPE_ZPmZ  : sve2_int_un_pred_arit_s<0b000, "urecpe">;
-  defm URSQRTE_ZPmZ : sve2_int_un_pred_arit_s<0b001, "ursqrte">;
-  defm SQABS_ZPmZ   : sve2_int_un_pred_arit<0b100, "sqabs">;
-  defm SQNEG_ZPmZ   : sve2_int_un_pred_arit<0b101, "sqneg">;
+  defm URECPE_ZPmZ  : sve2_int_un_pred_arit_s<0b000, "urecpe",  int_aarch64_sve_urecpe>;
+  defm URSQRTE_ZPmZ : sve2_int_un_pred_arit_s<0b001, "ursqrte", int_aarch64_sve_ursqrte>;
+  defm SQABS_ZPmZ   : sve2_int_un_pred_arit<0b100,   "sqabs",   int_aarch64_sve_sqabs>;
+  defm SQNEG_ZPmZ   : sve2_int_un_pred_arit<0b101,   "sqneg",   int_aarch64_sve_sqneg>;
 
   // SVE2 saturating add/subtract
   defm SQADD_ZPmZ  : sve2_int_arith_pred<0b110000, "sqadd",  null_frag>;
   defm UQADD_ZPmZ  : sve2_int_arith_pred<0b110010, "uqadd",  null_frag>;
   defm SQSUB_ZPmZ  : sve2_int_arith_pred<0b110100, "sqsub",  null_frag>;
   defm UQSUB_ZPmZ  : sve2_int_arith_pred<0b110110, "uqsub",  null_frag>;
   defm SUQADD_ZPmZ : sve2_int_arith_pred<0b111000, "suqadd", null_frag>;
   defm USQADD_ZPmZ : sve2_int_arith_pred<0b111010, "usqadd", null_frag>;

llvm/lib/Target/AArch64/SVEInstrFormats.td

: I<(outs zprty1:$Zda), (ins zprty1:$_Zda, zprty2:$Zn, zprty2:$Zm),
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zda;
 
   let Constraints = "$Zda = $_Zda";
   let DestructiveInstType = Destructive;
   let ElementSize = ElementSizeNone;
 }
 
-multiclass sve2_int_mla<bit S, string asm> {
+multiclass sve2_int_mla<bit S, string asm, SDPatternOperator op> {
   def _B : sve2_int_mla<0b00, { 0b1110, S }, asm, ZPR8, ZPR8>;
   def _H : sve2_int_mla<0b01, { 0b1110, S }, asm, ZPR16, ZPR16>;
   def _S : sve2_int_mla<0b10, { 0b1110, S }, asm, ZPR32, ZPR32>;
   def _D : sve2_int_mla<0b11, { 0b1110, S }, asm, ZPR64, ZPR64>;
+
+  def : SVE_3_Op_Pat<nxv16i8, op, nxv16i8, nxv16i8, nxv16i8, !cast<Instruction>(NAME # _B)>;
+  def : SVE_3_Op_Pat<nxv8i16, op, nxv8i16, nxv8i16, nxv8i16, !cast<Instruction>(NAME # _H)>;
+  def : SVE_3_Op_Pat<nxv4i32, op, nxv4i32, nxv4i32, nxv4i32, !cast<Instruction>(NAME # _S)>;
+  def : SVE_3_Op_Pat<nxv2i64, op, nxv2i64, nxv2i64, nxv2i64, !cast<Instruction>(NAME # _D)>;
 }
 
 multiclass sve2_int_mla_long<bits<5> opc, string asm> {
   def _H : sve2_int_mla<0b01, opc, asm, ZPR16, ZPR8>;
   def _S : sve2_int_mla<0b10, opc, asm, ZPR32, ZPR16>;
   def _D : sve2_int_mla<0b11, opc, asm, ZPR64, ZPR32>;
 }
 
: I<(outs zprty1:$Zda), (ins zprty1:$_Zda, zprty2:$Zn, zprty3:$Zm, itype:$iop),
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zda;
 
   let Constraints = "$Zda = $_Zda";
   let DestructiveInstType = Destructive;
   let ElementSize = ElementSizeNone;
 }
 
-multiclass sve2_int_mla_by_indexed_elem<bits<2> opc, bit S, string asm> {
-  def _H : sve2_int_mla_by_indexed_elem<{0, ?}, { 0b000, opc, S }, asm, ZPR16, ZPR16, ZPR3b16, VectorIndexH> {
+multiclass sve2_int_mla_by_indexed_elem<bits<2> opc, bit S, string asm,
+                                        SDPatternOperator op> {
+  def _H : sve2_int_mla_by_indexed_elem<{0, ?}, { 0b000, opc, S }, asm, ZPR16, ZPR16, ZPR3b16, VectorIndexH32b> {
     bits<3> Zm;
     bits<3> iop;
     let Inst{22} = iop{2};
     let Inst{20-19} = iop{1-0};
     let Inst{18-16} = Zm;
   }
-  def _S : sve2_int_mla_by_indexed_elem<0b10, { 0b000, opc, S }, asm, ZPR32, ZPR32, ZPR3b32, VectorIndexS> {
+  def _S : sve2_int_mla_by_indexed_elem<0b10, { 0b000, opc, S }, asm, ZPR32, ZPR32, ZPR3b32, VectorIndexS32b> {
     bits<3> Zm;
     bits<2> iop;
     let Inst{20-19} = iop;
     let Inst{18-16} = Zm;
   }
-  def _D : sve2_int_mla_by_indexed_elem<0b11, { 0b000, opc, S }, asm, ZPR64, ZPR64, ZPR4b64, VectorIndexD> {
+  def _D : sve2_int_mla_by_indexed_elem<0b11, { 0b000, opc, S }, asm, ZPR64, ZPR64, ZPR4b64, VectorIndexD32b> {
     bits<4> Zm;
     bit iop;
     let Inst{20} = iop;
     let Inst{19-16} = Zm;
   }
+
+  def : SVE_4_Op_Imm_Pat<nxv8i16, op, nxv8i16, nxv8i16, nxv8i16, i32, VectorIndexH32b_timm, !cast<Instruction>(NAME # _H)>;
+  def : SVE_4_Op_Imm_Pat<nxv4i32, op, nxv4i32, nxv4i32, nxv4i32, i32, VectorIndexS32b_timm, !cast<Instruction>(NAME # _S)>;
+  def : SVE_4_Op_Imm_Pat<nxv2i64, op, nxv2i64, nxv2i64, nxv2i64, i32, VectorIndexD32b_timm, !cast<Instruction>(NAME # _D)>;
 }
 
 //===----------------------------------------------------------------------===//
 // SVE2 Integer Multiply-Add Long - Indexed Group
 //===----------------------------------------------------------------------===//
 
 multiclass sve2_int_mla_long_by_indexed_elem<bits<4> opc, string asm, SDPatternOperator op> {
   def _S : sve2_int_mla_by_indexed_elem<0b10, { opc{3}, 0b0, opc{2-1}, ?, opc{0} },
: I<(outs zprty1:$Zd), (ins zprty2:$Zn, zprty3:$Zm, itype:$iop),
   let Inst{23-22} = sz;
   let Inst{21}    = 0b1;
   let Inst{15-14} = 0b11;
   let Inst{13-10} = opc;
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zd;
 }
 
-multiclass sve2_int_mul_by_indexed_elem<bits<4> opc, string asm> {
-  def _H : sve2_int_mul_by_indexed_elem<{0, ?}, opc, asm, ZPR16, ZPR16, ZPR3b16, VectorIndexH> {
+multiclass sve2_int_mul_by_indexed_elem<bits<4> opc, string asm,
+                                        SDPatternOperator op> {
+  def _H : sve2_int_mul_by_indexed_elem<{0, ?}, opc, asm, ZPR16, ZPR16, ZPR3b16, VectorIndexH32b> {
     bits<3> Zm;
     bits<3> iop;
     let Inst{22} = iop{2};
     let Inst{20-19} = iop{1-0};
     let Inst{18-16} = Zm;
   }
-  def _S : sve2_int_mul_by_indexed_elem<0b10, opc, asm, ZPR32, ZPR32, ZPR3b32, VectorIndexS> {
+  def _S : sve2_int_mul_by_indexed_elem<0b10, opc, asm, ZPR32, ZPR32, ZPR3b32, VectorIndexS32b> {
     bits<3> Zm;
     bits<2> iop;
     let Inst{20-19} = iop;
     let Inst{18-16} = Zm;
   }
-  def _D : sve2_int_mul_by_indexed_elem<0b11, opc, asm, ZPR64, ZPR64, ZPR4b64, VectorIndexD> {
+  def _D : sve2_int_mul_by_indexed_elem<0b11, opc, asm, ZPR64, ZPR64, ZPR4b64, VectorIndexD32b> {
     bits<4> Zm;
     bit iop;
     let Inst{20} = iop;
     let Inst{19-16} = Zm;
   }
+
+  def : SVE_3_Op_Imm_Pat<nxv8i16, op, nxv8i16, nxv8i16, i32, VectorIndexH32b_timm, !cast<Instruction>(NAME # _H)>;
+  def : SVE_3_Op_Imm_Pat<nxv4i32, op, nxv4i32, nxv4i32, i32, VectorIndexS32b_timm, !cast<Instruction>(NAME # _S)>;
+  def : SVE_3_Op_Imm_Pat<nxv2i64, op, nxv2i64, nxv2i64, i32, VectorIndexD32b_timm, !cast<Instruction>(NAME # _D)>;
 }
 
 multiclass sve2_int_mul_long_by_indexed_elem<bits<3> opc, string asm> {
   def _S : sve2_int_mul_by_indexed_elem<0b10, { opc{2-1}, ?, opc{0} }, asm,
                                         ZPR32, ZPR16, ZPR3b16, VectorIndexH> {
     bits<3> Zm;
     bits<3> iop;
     let Inst{20-19} = iop{2-1};
: I<(outs zprty:$Zd), (ins zprty:$_Zd, PPR3bAny:$Pg, zprty:$Zn),
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zd;
 
   let Constraints = "$Zd = $_Zd";
   let DestructiveInstType = Destructive;
   let ElementSize = zprty.ElementSize;
 }
 
-multiclass sve2_int_un_pred_arit_s<bits<3> opc, string asm> {
+multiclass sve2_int_un_pred_arit_s<bits<3> opc, string asm,
+                                   SDPatternOperator op> {
   def _S : sve2_int_un_pred_arit<0b10, opc{2}, opc{1-0}, asm, ZPR32>;
+  def : SVE_3_Op_Pat<nxv4i32, op, nxv4i32, nxv4i1, nxv4i32, !cast<Instruction>(NAME # _S)>;
 }
 
-multiclass sve2_int_un_pred_arit<bits<3> opc, string asm> {
+multiclass sve2_int_un_pred_arit<bits<3> opc, string asm, SDPatternOperator op> {
   def _B : sve2_int_un_pred_arit<0b00, opc{2}, opc{1-0}, asm, ZPR8>;
   def _H : sve2_int_un_pred_arit<0b01, opc{2}, opc{1-0}, asm, ZPR16>;
   def _S : sve2_int_un_pred_arit<0b10, opc{2}, opc{1-0}, asm, ZPR32>;
   def _D : sve2_int_un_pred_arit<0b11, opc{2}, opc{1-0}, asm, ZPR64>;
+
+  def : SVE_3_Op_Pat<nxv16i8, op, nxv16i8, nxv16i1, nxv16i8, !cast<Instruction>(NAME # _B)>;
+  def : SVE_3_Op_Pat<nxv8i16, op, nxv8i16, nxv8i1,  nxv8i16, !cast<Instruction>(NAME # _H)>;
+  def : SVE_3_Op_Pat<nxv4i32, op, nxv4i32, nxv4i1,  nxv4i32, !cast<Instruction>(NAME # _S)>;
+  def : SVE_3_Op_Pat<nxv2i64, op, nxv2i64, nxv2i1,  nxv2i64, !cast<Instruction>(NAME # _D)>;
 }
 
 //===----------------------------------------------------------------------===//
 // SVE2 Widening Integer Arithmetic Group
 //===----------------------------------------------------------------------===//
 
 class sve2_wide_int_arith<bits<2> sz, bits<5> opc, string asm,
                           ZPRRegOp zprty1, ZPRRegOp zprty2, ZPRRegOp zprty3>

llvm/test/CodeGen/AArch64/sve2-intrinsics-uniform-dsp.ll

+; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve2 -asm-verbose=0 < %s | FileCheck %s
+
+;
+; SHADD
+;
+
+define <vscale x 16 x i8> @shadd_i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
+; CHECK-LABEL: shadd_i8:
+; CHECK: shadd z0.b, p0/m, z0.b, z1.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.shadd.nxv16i8(<vscale x 16 x i1> %pg,
+                                                                 <vscale x 16 x i8> %a,
+                                                                 <vscale x 16 x i8> %b)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @shadd_i16(<vscale x 8 x i1> %pg, <vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: shadd_i16:
+; CHECK: shadd z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.shadd.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i16> %b)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @shadd_i32(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: shadd_i32:
+; CHECK: shadd z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.shadd.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @shadd_i64(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: shadd_i64:
+; CHECK: shadd z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.shadd.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i64> %b)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SHSUB
+;
+
+define <vscale x 16 x i8> @shsub_i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
+; CHECK-LABEL: shsub_i8:
+; CHECK: shsub z0.b, p0/m, z0.b, z1.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.shsub.nxv16i8(<vscale x 16 x i1> %pg,
+                                                                 <vscale x 16 x i8> %a,
+                                                                 <vscale x 16 x i8> %b)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @shsub_i16(<vscale x 8 x i1> %pg, <vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: shsub_i16:
+; CHECK: shsub z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.shsub.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i16> %b)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @shsub_i32(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: shsub_i32:
+; CHECK: shsub z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.shsub.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @shsub_i64(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: shsub_i64:
+; CHECK: shsub z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.shsub.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i64> %b)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SHSUBR
+;
+
+define <vscale x 16 x i8> @shsubr_i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
+; CHECK-LABEL: shsubr_i8:
+; CHECK: shsubr z0.b, p0/m, z0.b, z1.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.shsubr.nxv16i8(<vscale x 16 x i1> %pg,
+                                                                  <vscale x 16 x i8> %a,
+                                                                  <vscale x 16 x i8> %b)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @shsubr_i16(<vscale x 8 x i1> %pg, <vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: shsubr_i16:
+; CHECK: shsubr z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.shsubr.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                  <vscale x 8 x i16> %a,
+                                                                  <vscale x 8 x i16> %b)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @shsubr_i32(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: shsubr_i32:
+; CHECK: shsubr z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.shsubr.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                  <vscale x 4 x i32> %a,
+                                                                  <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @shsubr_i64(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: shsubr_i64:
+; CHECK: shsubr z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.shsubr.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                  <vscale x 2 x i64> %a,
+                                                                  <vscale x 2 x i64> %b)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SQABS
+;
+
+define <vscale x 16 x i8> @sqabs_i8(<vscale x 16 x i8> %a, <vscale x 16 x i1> %pg, <vscale x 16 x i8> %b) {
+; CHECK-LABEL: sqabs_i8:
+; CHECK: sqabs z0.b, p0/m, z1.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sqabs.nxv16i8(<vscale x 16 x i8> %a,
+                                                                 <vscale x 16 x i1> %pg,
+                                                                 <vscale x 16 x i8> %b)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @sqabs_i16(<vscale x 8 x i16> %a, <vscale x 8 x i1> %pg, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: sqabs_i16:
+; CHECK: sqabs z0.h, p0/m, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqabs.nxv8i16(<vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %b)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sqabs_i32(<vscale x 4 x i32> %a, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: sqabs_i32:
+; CHECK: sqabs z0.s, p0/m, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqabs.nxv4i32(<vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sqabs_i64(<vscale x 2 x i64> %a, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: sqabs_i64:
+; CHECK: sqabs z0.d, p0/m, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqabs.nxv2i64(<vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %b)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SQDMULH (Vector)
+;
+
+define <vscale x 16 x i8> @sqdmulh_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
+; CHECK-LABEL: sqdmulh_i8:
+; CHECK: sqdmulh z0.b, z0.b, z1.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sqdmulh.nxv16i8(<vscale x 16 x i8> %a,
+                                                                   <vscale x 16 x i8> %b)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @sqdmulh_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: sqdmulh_i16:
+; CHECK: sqdmulh z0.h, z0.h, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqdmulh.nxv8i16(<vscale x 8 x i16> %a,
+                                                                   <vscale x 8 x i16> %b)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sqdmulh_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: sqdmulh_i32:
+; CHECK: sqdmulh z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqdmulh.nxv4i32(<vscale x 4 x i32> %a,
+                                                                   <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sqdmulh_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: sqdmulh_i64:
+; CHECK: sqdmulh z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqdmulh.nxv2i64(<vscale x 2 x i64> %a,
+                                                                   <vscale x 2 x i64> %b)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SQDMULH (Indexed)
+;
+
+define <vscale x 8 x i16> @sqdmulh_lane_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: sqdmulh_lane_i16:
+; CHECK: sqdmulh z0.h, z0.h, z1.h[7]
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqdmulh.lane.nxv8i16(<vscale x 8 x i16> %a,
+                                                                        <vscale x 8 x i16> %b,
+                                                                        i32 7)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sqdmulh_lane_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: sqdmulh_lane_i32:
+; CHECK: sqdmulh z0.s, z0.s, z1.s[3]
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqdmulh.lane.nxv4i32(<vscale x 4 x i32> %a,
+                                                                        <vscale x 4 x i32> %b,
+                                                                        i32 3);
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sqdmulh_lane_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: sqdmulh_lane_i64:
+; CHECK: sqdmulh z0.d, z0.d, z1.d[1]
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqdmulh.lane.nxv2i64(<vscale x 2 x i64> %a,
+                                                                        <vscale x 2 x i64> %b,
+                                                                        i32 1)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SQNEG
+;
+
+define <vscale x 16 x i8> @sqneg_i8(<vscale x 16 x i8> %a, <vscale x 16 x i1> %pg, <vscale x 16 x i8> %b) {
+; CHECK-LABEL: sqneg_i8:
+; CHECK: sqneg z0.b, p0/m, z1.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sqneg.nxv16i8(<vscale x 16 x i8> %a,
+                                                                 <vscale x 16 x i1> %pg,
+                                                                 <vscale x 16 x i8> %b)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @sqneg_i16(<vscale x 8 x i16> %a, <vscale x 8 x i1> %pg, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: sqneg_i16:
+; CHECK: sqneg z0.h, p0/m, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqneg.nxv8i16(<vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %b)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sqneg_i32(<vscale x 4 x i32> %a, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: sqneg_i32:
+; CHECK: sqneg z0.s, p0/m, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqneg.nxv4i32(<vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sqneg_i64(<vscale x 2 x i64> %a, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: sqneg_i64:
+; CHECK: sqneg z0.d, p0/m, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqneg.nxv2i64(<vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %b)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SQRDMALH (Vectors)
+;
+
+define <vscale x 16 x i8> @sqrdmlah_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b, <vscale x 16 x i8> %c) {
+; CHECK-LABEL: sqrdmlah_i8:
+; CHECK: sqrdmlah z0.b, z1.b, z2.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sqrdmlah.nxv16i8(<vscale x 16 x i8> %a,
+                                                                    <vscale x 16 x i8> %b,
+                                                                    <vscale x 16 x i8> %c)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @sqrdmlah_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c) {
+; CHECK-LABEL: sqrdmlah_i16:
+; CHECK: sqrdmlah z0.h, z1.h, z2.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqrdmlah.nxv8i16(<vscale x 8 x i16> %a,
+                                                                    <vscale x 8 x i16> %b,
+                                                                    <vscale x 8 x i16> %c)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sqrdmlah_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c) {
+; CHECK-LABEL: sqrdmlah_i32:
+; CHECK: sqrdmlah z0.s, z1.s, z2.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqrdmlah.nxv4i32(<vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b,
+                                                                    <vscale x 4 x i32> %c)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sqrdmlah_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c) {
+; CHECK-LABEL: sqrdmlah_i64:
+; CHECK: sqrdmlah z0.d, z1.d, z2.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqrdmlah.nxv2i64(<vscale x 2 x i64> %a,
+                                                                    <vscale x 2 x i64> %b,
+                                                                    <vscale x 2 x i64> %c)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SQRDMALH (Indexed)
+;
+
+define <vscale x 8 x i16> @sqrdmlah_lane_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c) {
+; CHECK-LABEL: sqrdmlah_lane_i16:
+; CHECK: sqrdmlah z0.h, z1.h, z2.h[5]
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqrdmlah.lane.nxv8i16(<vscale x 8 x i16> %a,
+                                                                         <vscale x 8 x i16> %b,
+                                                                         <vscale x 8 x i16> %c,
+                                                                         i32 5)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sqrdmlah_lane_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c) {
+; CHECK-LABEL: sqrdmlah_lane_i32:
+; CHECK: sqrdmlah z0.s, z1.s, z2.s[1]
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqrdmlah.lane.nxv4i32(<vscale x 4 x i32> %a,
+                                                                         <vscale x 4 x i32> %b,
+                                                                         <vscale x 4 x i32> %c,
+                                                                         i32 1);
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sqrdmlah_lane_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c) {
+; CHECK-LABEL: sqrdmlah_lane_i64:
+; CHECK: sqrdmlah z0.d, z1.d, z2.d[1]
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqrdmlah.lane.nxv2i64(<vscale x 2 x i64> %a,
+                                                                         <vscale x 2 x i64> %b,
+                                                                         <vscale x 2 x i64> %c,
+                                                                         i32 1)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SQRDMSLH (Vectors)
+;
+
+define <vscale x 16 x i8> @sqrdmlsh_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b, <vscale x 16 x i8> %c) {
+; CHECK-LABEL: sqrdmlsh_i8:
+; CHECK: sqrdmlsh z0.b, z1.b, z2.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sqrdmlsh.nxv16i8(<vscale x 16 x i8> %a,
+                                                                    <vscale x 16 x i8> %b,
+                                                                    <vscale x 16 x i8> %c)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @sqrdmlsh_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c) {
+; CHECK-LABEL: sqrdmlsh_i16:
+; CHECK: sqrdmlsh z0.h, z1.h, z2.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqrdmlsh.nxv8i16(<vscale x 8 x i16> %a,
+                                                                    <vscale x 8 x i16> %b,
+                                                                    <vscale x 8 x i16> %c)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sqrdmlsh_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c) {
+; CHECK-LABEL: sqrdmlsh_i32:
+; CHECK: sqrdmlsh z0.s, z1.s, z2.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqrdmlsh.nxv4i32(<vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b,
+                                                                    <vscale x 4 x i32> %c)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sqrdmlsh_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c) {
+; CHECK-LABEL: sqrdmlsh_i64:
+; CHECK: sqrdmlsh z0.d, z1.d, z2.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqrdmlsh.nxv2i64(<vscale x 2 x i64> %a,
+                                                                    <vscale x 2 x i64> %b,
+                                                                    <vscale x 2 x i64> %c)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SQRDMSLH (Indexed)
+;
+
+define <vscale x 8 x i16> @sqrdmlsh_lane_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c) {
+; CHECK-LABEL: sqrdmlsh_lane_i16:
+; CHECK: sqrdmlsh z0.h, z1.h, z2.h[4]
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqrdmlsh.lane.nxv8i16(<vscale x 8 x i16> %a,
+                                                                         <vscale x 8 x i16> %b,
+                                                                         <vscale x 8 x i16> %c,
+                                                                         i32 4)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sqrdmlsh_lane_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c) {
+; CHECK-LABEL: sqrdmlsh_lane_i32:
+; CHECK: sqrdmlsh z0.s, z1.s, z2.s[0]
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqrdmlsh.lane.nxv4i32(<vscale x 4 x i32> %a,
+                                                                         <vscale x 4 x i32> %b,
+                                                                         <vscale x 4 x i32> %c,
+                                                                         i32 0);
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sqrdmlsh_lane_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c) {
+; CHECK-LABEL: sqrdmlsh_lane_i64:
+; CHECK: sqrdmlsh z0.d, z1.d, z2.d[1]
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqrdmlsh.lane.nxv2i64(<vscale x 2 x i64> %a,
+                                                                         <vscale x 2 x i64> %b,
+                                                                         <vscale x 2 x i64> %c,
+                                                                         i32 1)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SQRDMULH (Vectors)
+;
+
+define <vscale x 16 x i8> @sqrdmulh_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
+; CHECK-LABEL: sqrdmulh_i8:
+; CHECK: sqrdmulh z0.b, z0.b, z1.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sqrdmulh.nxv16i8(<vscale x 16 x i8> %a,
+                                                                    <vscale x 16 x i8> %b)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @sqrdmulh_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: sqrdmulh_i16:
+; CHECK: sqrdmulh z0.h, z0.h, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqrdmulh.nxv8i16(<vscale x 8 x i16> %a,
+                                                                    <vscale x 8 x i16> %b)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sqrdmulh_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: sqrdmulh_i32:
+; CHECK: sqrdmulh z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqrdmulh.nxv4i32(<vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sqrdmulh_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: sqrdmulh_i64:
+; CHECK: sqrdmulh z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqrdmulh.nxv2i64(<vscale x 2 x i64> %a,
+                                                                    <vscale x 2 x i64> %b)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SQRDMULH (Indexed)
+;
+
+define <vscale x 8 x i16> @sqrdmulh_lane_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: sqrdmulh_lane_i16:
+; CHECK: sqrdmulh z0.h, z0.h, z1.h[6]
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqrdmulh.lane.nxv8i16(<vscale x 8 x i16> %a,
+                                                                         <vscale x 8 x i16> %b,
+                                                                         i32 6)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sqrdmulh_lane_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: sqrdmulh_lane_i32:
+; CHECK: sqrdmulh z0.s, z0.s, z1.s[2]
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqrdmulh.lane.nxv4i32(<vscale x 4 x i32> %a,
+                                                                         <vscale x 4 x i32> %b,
+                                                                         i32 2);
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sqrdmulh_lane_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: sqrdmulh_lane_i64:
+; CHECK: sqrdmulh z0.d, z0.d, z1.d[1]
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqrdmulh.lane.nxv2i64(<vscale x 2 x i64> %a,
+                                                                         <vscale x 2 x i64> %b,
+                                                                         i32 1)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; SRHADD
+;
+
+define <vscale x 16 x i8> @srhadd_i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
+; CHECK-LABEL: srhadd_i8:
+; CHECK: srhadd z0.b, p0/m, z0.b, z1.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.srhadd.nxv16i8(<vscale x 16 x i1> %pg,
+                                                                  <vscale x 16 x i8> %a,
+                                                                  <vscale x 16 x i8> %b)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @srhadd_i16(<vscale x 8 x i1> %pg, <vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: srhadd_i16:
+; CHECK: srhadd z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.srhadd.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                  <vscale x 8 x i16> %a,
+                                                                  <vscale x 8 x i16> %b)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @srhadd_i32(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: srhadd_i32:
+; CHECK: srhadd z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.srhadd.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                  <vscale x 4 x i32> %a,
+                                                                  <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @srhadd_i64(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: srhadd_i64:
+; CHECK: srhadd z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.srhadd.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                  <vscale x 2 x i64> %a,
+                                                                  <vscale x 2 x i64> %b)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; UHADD
+;
+
+define <vscale x 16 x i8> @uhadd_i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
+; CHECK-LABEL: uhadd_i8:
+; CHECK: uhadd z0.b, p0/m, z0.b, z1.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.uhadd.nxv16i8(<vscale x 16 x i1> %pg,
+                                                                 <vscale x 16 x i8> %a,
+                                                                 <vscale x 16 x i8> %b)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @uhadd_i16(<vscale x 8 x i1> %pg, <vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: uhadd_i16:
+; CHECK: uhadd z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.uhadd.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i16> %b)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @uhadd_i32(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: uhadd_i32:
+; CHECK: uhadd z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.uhadd.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @uhadd_i64(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: uhadd_i64:
+; CHECK: uhadd z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.uhadd.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i64> %b)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; UHSUB
+;
+
+define <vscale x 16 x i8> @uhsub_i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
+; CHECK-LABEL: uhsub_i8:
+; CHECK: uhsub z0.b, p0/m, z0.b, z1.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.uhsub.nxv16i8(<vscale x 16 x i1> %pg,
+                                                                 <vscale x 16 x i8> %a,
+                                                                 <vscale x 16 x i8> %b)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @uhsub_i16(<vscale x 8 x i1> %pg, <vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: uhsub_i16:
+; CHECK: uhsub z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.uhsub.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i16> %b)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @uhsub_i32(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: uhsub_i32:
+; CHECK: uhsub z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.uhsub.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @uhsub_i64(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: uhsub_i64:
+; CHECK: uhsub z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.uhsub.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i64> %b)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; UHSUBR
+;
+
+define <vscale x 16 x i8> @uhsubr_i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
+; CHECK-LABEL: uhsubr_i8:
+; CHECK: uhsubr z0.b, p0/m, z0.b, z1.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.uhsubr.nxv16i8(<vscale x 16 x i1> %pg,
+                                                                  <vscale x 16 x i8> %a,
+                                                                  <vscale x 16 x i8> %b)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @uhsubr_i16(<vscale x 8 x i1> %pg, <vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: uhsubr_i16:
+; CHECK: uhsubr z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.uhsubr.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                  <vscale x 8 x i16> %a,
+                                                                  <vscale x 8 x i16> %b)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @uhsubr_i32(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: uhsubr_i32:
+; CHECK: uhsubr z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.uhsubr.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                  <vscale x 4 x i32> %a,
+                                                                  <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @uhsubr_i64(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: uhsubr_i64:
+; CHECK: uhsubr z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.uhsubr.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                  <vscale x 2 x i64> %a,
+                                                                  <vscale x 2 x i64> %b)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; URECPE
+;
+
+define <vscale x 4 x i32> @urecpe_i32(<vscale x 4 x i32> %a, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: urecpe_i32:
+; CHECK: urecpe z0.s, p0/m, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.urecpe.nxv4i32(<vscale x 4 x i32> %a,
+                                                                  <vscale x 4 x i1> %pg,
+                                                                  <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+;
+; URHADD
+;
+
+define <vscale x 16 x i8> @urhadd_i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
+; CHECK-LABEL: urhadd_i8:
+; CHECK: urhadd z0.b, p0/m, z0.b, z1.b
+; CHECK-NEXT: ret
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.urhadd.nxv16i8(<vscale x 16 x i1> %pg,
+                                                                  <vscale x 16 x i8> %a,
+                                                                  <vscale x 16 x i8> %b)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @urhadd_i16(<vscale x 8 x i1> %pg, <vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: urhadd_i16:
+; CHECK: urhadd z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.urhadd.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                  <vscale x 8 x i16> %a,
+                                                                  <vscale x 8 x i16> %b)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @urhadd_i32(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: urhadd_i32:
+; CHECK: urhadd z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.urhadd.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                  <vscale x 4 x i32> %a,
+                                                                  <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @urhadd_i64(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: urhadd_i64:
+; CHECK: urhadd z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.urhadd.nxv2i64(<vscale x 2 x i1> %pg,
+                                                             <vscale x 2 x i64> %a,
+                                                             <vscale x 2 x i64> %b)
+  ret <vscale x 2 x i64> %out
+}
+
+;
+; URSQRTE
+;
+
+define <vscale x 4 x i32> @ursqrte_i32(<vscale x 4 x i32> %a, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: ursqrte_i32:
+; CHECK: ursqrte z0.s, p0/m, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.ursqrte.nxv4i32(<vscale x 4 x i32> %a,
+                                                                   <vscale x 4 x i1> %pg,
+                                                                   <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.shadd.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x i8>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.shadd.nxv8i16(<vscale x 8 x i1>, <vscale x 8 x i16>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.shadd.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.shadd.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.shsub.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x i8>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.shsub.nxv8i16(<vscale x 8 x i1>, <vscale x 8 x i16>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.shsub.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.shsub.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.shsubr.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x i8>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.shsubr.nxv8i16(<vscale x 8 x i1>, <vscale x 8 x i16>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.shsubr.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.shsubr.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.sqabs.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i1>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.sqabs.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i1>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.sqabs.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i1>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.sqabs.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i1>, <vscale x 2 x i64>)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.sqdmulh.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.sqdmulh.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.sqdmulh.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.sqdmulh.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 8 x i16> @llvm.aarch64.sve.sqdmulh.lane.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, i32)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.sqdmulh.lane.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, i32)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.sqdmulh.lane.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, i32)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.sqneg.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i1>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.sqneg.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i1>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.sqneg.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i1>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.sqneg.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i1>, <vscale x 2 x i64>)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.sqrdmlah.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.sqrdmlah.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.sqrdmlah.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.sqrdmlah.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 8 x i16> @llvm.aarch64.sve.sqrdmlah.lane.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.sqrdmlah.lane.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, i32)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.sqrdmlah.lane.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, i32)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.sqrdmlsh.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.sqrdmlsh.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.sqrdmlsh.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.sqrdmlsh.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 8 x i16> @llvm.aarch64.sve.sqrdmlsh.lane.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.sqrdmlsh.lane.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, i32)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.sqrdmlsh.lane.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, i32)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.sqrdmulh.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.sqrdmulh.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.sqrdmulh.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.sqrdmulh.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 8 x i16> @llvm.aarch64.sve.sqrdmulh.lane.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, i32)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.sqrdmulh.lane.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, i32)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.sqrdmulh.lane.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, i32)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.srhadd.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x i8>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.srhadd.nxv8i16(<vscale x 8 x i1>, <vscale x 8 x i16>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.srhadd.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.srhadd.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.uhadd.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x i8>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.uhadd.nxv8i16(<vscale x 8 x i1>, <vscale x 8 x i16>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.uhadd.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.uhadd.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.uhsub.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x i8>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.uhsub.nxv8i16(<vscale x 8 x i1>, <vscale x 8 x i16>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.uhsub.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.uhsub.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.uhsubr.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x i8>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.uhsubr.nxv8i16(<vscale x 8 x i1>, <vscale x 8 x i16>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.uhsubr.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.uhsubr.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 4 x i32> @llvm.aarch64.sve.urecpe.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i1>, <vscale x 4 x i32>)
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.urhadd.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x i8>, <vscale x 16 x i8>)
+declare <vscale x 8 x i16> @llvm.aarch64.sve.urhadd.nxv8i16(<vscale x 8 x i1>, <vscale x 8 x i16>, <vscale x 8 x i16>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.urhadd.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.urhadd.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 4 x i32> @llvm.aarch64.sve.ursqrte.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i1>, <vscale x 4 x i32>)