[AArch64][SVE] Implement additional floating-point arithmetic intrinsics

[kmclaughlin]

Summary

Adds intrinsics for the following:

• ftssel
• fcadd, fcmla
• fmla, fmls, fnmla, fnmls
• fmad, fmsb, fnmad, fnmsb

[sdesmalen]

llvm/lib/Target/AArch64/AArch64InstrFormats.td

10068

should the target constant not be MVT::i32 here?

10075

Same here

[kmclaughlin]

• Changed target constant to MVT::i32 in complexrotateop & complexrotateopodd definitions

[kmclaughlin]

• Rebased & removed unused llvm.aarch64.sve.fcmla.lane.nxv2f64 from sve-intrinsics-fp-arith.ll

[sdesmalen]

Thanks, LGTM!

llvm/include/llvm/IR/IntrinsicsAArch64.td

 let TargetPrefix = "aarch64" in {  // All intrinsics start with "llvm.aarch64.".
   class AdvSIMD_Merged1VectorArg_Intrinsic
     : Intrinsic<[llvm_anyvector_ty],
                 [LLVMMatchType<0>,
                  LLVMScalarOrSameVectorWidth<0, llvm_i1_ty>,
                  LLVMMatchType<0>],
                 [IntrNoMem]>;
 
+  class AdvSIMD_2VectorArgIndexed_Intrinsic
+    : Intrinsic<[llvm_anyvector_ty],
+                [LLVMMatchType<0>,
+                 LLVMMatchType<0>,
+                 llvm_i32_ty],
+                [IntrNoMem]>;
+
+  class AdvSIMD_3VectorArgIndexed_Intrinsic
+    : Intrinsic<[llvm_anyvector_ty],
+                [LLVMMatchType<0>,
+                 LLVMMatchType<0>,
+                 LLVMMatchType<0>,
+                 llvm_i32_ty],
+                [IntrNoMem]>;
+
   class AdvSIMD_SVE_CNT_Intrinsic
     : Intrinsic<[LLVMVectorOfBitcastsToInt<0>],
                 [LLVMVectorOfBitcastsToInt<0>,
                  LLVMScalarOrSameVectorWidth<0, llvm_i1_ty>,
                  llvm_anyvector_ty],
                 [IntrNoMem]>;
 
   class AdvSIMD_SVE_Unpack_Intrinsic
    : Intrinsic<[llvm_anyvector_ty],
                [LLVMSubdivide2VectorType<0>],
                [IntrNoMem]>;
 
+  class AdvSIMD_SVE_CADD_Intrinsic
+    : Intrinsic<[llvm_anyvector_ty],
+                [LLVMScalarOrSameVectorWidth<0, llvm_i1_ty>,
+                 LLVMMatchType<0>,
+                 LLVMMatchType<0>,
+                 llvm_i32_ty],
+                [IntrNoMem]>;
+
+  class AdvSIMD_SVE_CMLA_Intrinsic
+    : Intrinsic<[llvm_anyvector_ty],
+                [LLVMScalarOrSameVectorWidth<0, llvm_i1_ty>,
+                 LLVMMatchType<0>,
+                 LLVMMatchType<0>,
+                 LLVMMatchType<0>,
+                 llvm_i32_ty],
+                [IntrNoMem]>;
+
+  class AdvSIMD_SVE_CMLA_LANE_Intrinsic
+    : Intrinsic<[llvm_anyvector_ty],
+                [LLVMMatchType<0>,
+                 LLVMMatchType<0>,
+                 LLVMMatchType<0>,
+                 llvm_i32_ty,
+                 llvm_i32_ty],
+                [IntrNoMem]>;
+
   class AdvSIMD_SVE_PUNPKHI_Intrinsic
     : Intrinsic<[LLVMHalfElementsVectorType<0>],
                 [llvm_anyvector_ty],
                 [IntrNoMem]>;
 
   class AdvSIMD_SVE_SCALE_Intrinsic
     : Intrinsic<[llvm_anyvector_ty],
                 [LLVMScalarOrSameVectorWidth<0, llvm_i1_ty>,

 def int_aarch64_sve_uxtw : AdvSIMD_Merged1VectorArg_Intrinsic;
 
 //
 // Floating-point arithmetic
 //
 
 def int_aarch64_sve_fabd       : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fadd       : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fcadd      : AdvSIMD_SVE_CADD_Intrinsic;
+def int_aarch64_sve_fcmla      : AdvSIMD_SVE_CMLA_Intrinsic;
+def int_aarch64_sve_fcmla_lane : AdvSIMD_SVE_CMLA_LANE_Intrinsic;
 def int_aarch64_sve_fdiv       : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fdivr      : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fmad       : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_fmax       : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fmaxnm     : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fmin       : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fminnm     : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fmla       : AdvSIMD_Pred3VectorArg_Intrinsic;
+def int_aarch64_sve_fmla_lane  : AdvSIMD_3VectorArgIndexed_Intrinsic;
+def int_aarch64_sve_fmls       : AdvSIMD_Pred3VectorArg_Intrinsic;
+def int_aarch64_sve_fmls_lane  : AdvSIMD_3VectorArgIndexed_Intrinsic;
+def int_aarch64_sve_fmsb       : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_fmul       : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fmulx      : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fmul_lane  : AdvSIMD_2VectorArgIndexed_Intrinsic;
+def int_aarch64_sve_fnmad      : AdvSIMD_Pred3VectorArg_Intrinsic;
+def int_aarch64_sve_fnmla      : AdvSIMD_Pred3VectorArg_Intrinsic;
+def int_aarch64_sve_fnmls      : AdvSIMD_Pred3VectorArg_Intrinsic;
+def int_aarch64_sve_fnmsb      : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_fscale     : AdvSIMD_SVE_SCALE_Intrinsic;
 def int_aarch64_sve_fsub       : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fsubr      : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_ftmad_x    : AdvSIMD_2VectorArgIndexed_Intrinsic;
 def int_aarch64_sve_ftsmul_x   : AdvSIMD_SVE_TSMUL_Intrinsic;
+def int_aarch64_sve_ftssel_x   : AdvSIMD_SVE_TSMUL_Intrinsic;
 
 //
 // Floating-point comparisons
 //
 
 def int_aarch64_sve_fcvtzs_i32f16 : Builtin_SVCVT<"svcvt_s32_f16_m", llvm_nxv4i32_ty, llvm_nxv8f16_ty>;
 
 //

llvm/lib/Target/AArch64/AArch64InstrFormats.td

 
 // imm0_7 predicate - True if the immediate is in the range [0,7]
 def imm0_7 : Operand<i64>, ImmLeaf<i64, [{
   return ((uint64_t)Imm) < 8;
 }]> {
   let ParserMatchClass = Imm0_7Operand;
 }
 
+// imm32_0_7 predicate - True if the 32-bit immediate is in the range [0,7]
+def imm32_0_7 : Operand<i32>, ImmLeaf<i32, [{
+  return ((uint32_t)Imm) < 8;
+}]> {
+  let ParserMatchClass = Imm0_7Operand;
+}
+
 // imm32_0_15 predicate - True if the 32-bit immediate is in the range [0,15]
 def imm32_0_15 : Operand<i32>, ImmLeaf<i32, [{
   return ((uint32_t)Imm) < 16;
 }]> {
   let ParserMatchClass = Imm0_15Operand;
 }
 
 // An arithmetic shifter operand:

 //----------------------------------------------------------------------------
 
 class ComplexRotationOperand<int Angle, int Remainder, string Type>
   : AsmOperandClass {
   let PredicateMethod = "isComplexRotation<" # Angle # ", " # Remainder # ">";
   let DiagnosticType = "InvalidComplexRotation" # Type;
   let Name = "ComplexRotation" # Type;
 }
-def complexrotateop : Operand<i32> {
+def complexrotateop : Operand<i32>, ImmLeaf<i32, [{ return Imm >= 0 && Imm <= 270;  }],
+                                                 SDNodeXForm<imm, [{
+  return CurDAG->getTargetConstant((N->getSExtValue() / 90), SDLoc(N), MVT::i32);
+}]>> {
   let ParserMatchClass = ComplexRotationOperand<90, 0, "Even">;
   let PrintMethod = "printComplexRotationOp<90, 0>";
 }
-def complexrotateopodd : Operand<i32> {
+def complexrotateopodd : Operand<i32>, ImmLeaf<i32, [{ return Imm >= 0 && Imm <= 270;  }],
+                                                 SDNodeXForm<imm, [{
+  return CurDAG->getTargetConstant(((N->getSExtValue() - 90) / 180), SDLoc(N), MVT::i32);
+}]>> {
   let ParserMatchClass = ComplexRotationOperand<180, 90, "Odd">;
   let PrintMethod = "printComplexRotationOp<180, 90>";
 }
-
 let mayLoad = 0, mayStore = 0, hasSideEffects = 0 in
 class BaseSIMDThreeSameVectorComplex<bit Q, bit U, bits<2> size, bits<3> opcode,
                                      RegisterOperand regtype, Operand rottype,
                                      string asm, string kind, list<dag> pattern>
   : I<(outs regtype:$Rd), (ins regtype:$Rn, regtype:$Rm, rottype:$rot), asm,
       "{\t$Rd" # kind # ", $Rn" # kind # ", $Rm" # kind # ", $rot"
       "|" # kind # "\t$Rd, $Rn, $Rm, $rot}", "", pattern>,
     Sched<[WriteV]> {

llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td

let Predicates = [HasSVE] in {
 
   defm FADD_ZZZ    : sve_fp_3op_u_zd<0b000, "fadd",    fadd>;
   defm FSUB_ZZZ    : sve_fp_3op_u_zd<0b001, "fsub",    fsub>;
   defm FMUL_ZZZ    : sve_fp_3op_u_zd<0b010, "fmul",    fmul>;
   defm FTSMUL_ZZZ  : sve_fp_3op_u_zd_ftsmul<0b011, "ftsmul",  int_aarch64_sve_ftsmul_x>;
   defm FRECPS_ZZZ  : sve_fp_3op_u_zd<0b110, "frecps",  int_aarch64_sve_frecps_x>;
   defm FRSQRTS_ZZZ : sve_fp_3op_u_zd<0b111, "frsqrts", int_aarch64_sve_frsqrts_x>;
 
-  defm FTSSEL_ZZZ : sve_int_bin_cons_misc_0_b<"ftssel">;
+  defm FTSSEL_ZZZ : sve_int_bin_cons_misc_0_b<"ftssel", int_aarch64_sve_ftssel_x>;
 
-  defm FCADD_ZPmZ : sve_fp_fcadd<"fcadd">;
-  defm FCMLA_ZPmZZ : sve_fp_fcmla<"fcmla">;
+  defm FCADD_ZPmZ : sve_fp_fcadd<"fcadd", int_aarch64_sve_fcadd>;
+  defm FCMLA_ZPmZZ : sve_fp_fcmla<"fcmla", int_aarch64_sve_fcmla>;
 
-  defm FMLA_ZPmZZ  : sve_fp_3op_p_zds_a<0b00, "fmla">;
-  defm FMLS_ZPmZZ  : sve_fp_3op_p_zds_a<0b01, "fmls">;
-  defm FNMLA_ZPmZZ : sve_fp_3op_p_zds_a<0b10, "fnmla">;
-  defm FNMLS_ZPmZZ : sve_fp_3op_p_zds_a<0b11, "fnmls">;
+  defm FMLA_ZPmZZ  : sve_fp_3op_p_zds_a<0b00, "fmla",  int_aarch64_sve_fmla>;
+  defm FMLS_ZPmZZ  : sve_fp_3op_p_zds_a<0b01, "fmls",  int_aarch64_sve_fmls>;
+  defm FNMLA_ZPmZZ : sve_fp_3op_p_zds_a<0b10, "fnmla", int_aarch64_sve_fnmla>;
+  defm FNMLS_ZPmZZ : sve_fp_3op_p_zds_a<0b11, "fnmls", int_aarch64_sve_fnmls>;
 
-  defm FMAD_ZPmZZ  : sve_fp_3op_p_zds_b<0b00, "fmad">;
-  defm FMSB_ZPmZZ  : sve_fp_3op_p_zds_b<0b01, "fmsb">;
-  defm FNMAD_ZPmZZ : sve_fp_3op_p_zds_b<0b10, "fnmad">;
-  defm FNMSB_ZPmZZ : sve_fp_3op_p_zds_b<0b11, "fnmsb">;
+  defm FMAD_ZPmZZ  : sve_fp_3op_p_zds_b<0b00, "fmad",  int_aarch64_sve_fmad>;
+  defm FMSB_ZPmZZ  : sve_fp_3op_p_zds_b<0b01, "fmsb",  int_aarch64_sve_fmsb>;
+  defm FNMAD_ZPmZZ : sve_fp_3op_p_zds_b<0b10, "fnmad", int_aarch64_sve_fnmad>;
+  defm FNMSB_ZPmZZ : sve_fp_3op_p_zds_b<0b11, "fnmsb", int_aarch64_sve_fnmsb>;
 
-  defm FTMAD_ZZI : sve_fp_ftmad<"ftmad">;
+  defm FTMAD_ZZI : sve_fp_ftmad<"ftmad", int_aarch64_sve_ftmad_x>;
 
-  defm FMLA_ZZZI : sve_fp_fma_by_indexed_elem<0b0, "fmla">;
-  defm FMLS_ZZZI : sve_fp_fma_by_indexed_elem<0b1, "fmls">;
+  defm FMLA_ZZZI : sve_fp_fma_by_indexed_elem<0b0, "fmla", int_aarch64_sve_fmla_lane>;
+  defm FMLS_ZZZI : sve_fp_fma_by_indexed_elem<0b1, "fmls", int_aarch64_sve_fmls_lane>;
 
-  defm FCMLA_ZZZI : sve_fp_fcmla_by_indexed_elem<"fcmla">;
-  defm FMUL_ZZZI   : sve_fp_fmul_by_indexed_elem<"fmul">;
+  defm FCMLA_ZZZI : sve_fp_fcmla_by_indexed_elem<"fcmla", int_aarch64_sve_fcmla_lane>;
+  defm FMUL_ZZZI   : sve_fp_fmul_by_indexed_elem<"fmul", int_aarch64_sve_fmul_lane>;
 
   // SVE floating point reductions.
   defm FADDA_VPZ   : sve_fp_2op_p_vd<0b000, "fadda">;
   defm FADDV_VPZ   : sve_fp_fast_red<0b000, "faddv">;
   defm FMAXNMV_VPZ : sve_fp_fast_red<0b100, "fmaxnmv">;
   defm FMINNMV_VPZ : sve_fp_fast_red<0b101, "fminnmv">;
   defm FMAXV_VPZ   : sve_fp_fast_red<0b110, "fmaxv">;
   defm FMINV_VPZ   : sve_fp_fast_red<0b111, "fminv">;

llvm/lib/Target/AArch64/SVEInstrFormats.td

: Pat<(vtd (op vt1:$Op1, vt2:$Op2)),
       (inst $Op1, $Op2)>;
 
 class SVE_3_Op_Pat<ValueType vtd, SDPatternOperator op, ValueType vt1,
                    ValueType vt2, ValueType vt3, Instruction inst>
 : Pat<(vtd (op vt1:$Op1, vt2:$Op2, vt3:$Op3)),
       (inst $Op1, $Op2, $Op3)>;
 
 class SVE_4_Op_Pat<ValueType vtd, SDPatternOperator op, ValueType vt1,
-                   ValueType vt2, ValueType vt3, ValueType vt4, Instruction inst>
+                   ValueType vt2, ValueType vt3, ValueType vt4,
+                   Instruction inst>
 : Pat<(vtd (op vt1:$Op1, vt2:$Op2, vt3:$Op3, vt4:$Op4)),
       (inst $Op1, $Op2, $Op3, $Op4)>;
 
 def SVEDup0Undef : ComplexPattern<i64, 0, "SelectDupZeroOrUndef", []>;
 
 //===----------------------------------------------------------------------===//
 // SVE Predicate Misc Group
 //===----------------------------------------------------------------------===//
multiclass sve_fp_2op_p_zds_fscale<bits<4> opc, string asm,
   def _D : sve_fp_2op_p_zds<0b11, opc, asm, ZPR64>;
 
   def : SVE_3_Op_Pat<nxv8f16, op, nxv8i1, nxv8f16, nxv8i16, !cast<Instruction>(NAME # _H)>;
   def : SVE_3_Op_Pat<nxv4f32, op, nxv4i1, nxv4f32, nxv4i32, !cast<Instruction>(NAME # _S)>;
   def : SVE_3_Op_Pat<nxv2f64, op, nxv2i1, nxv2f64, nxv2i64, !cast<Instruction>(NAME # _D)>;
 }
 
 class sve_fp_ftmad<bits<2> sz, string asm, ZPRRegOp zprty>
-: I<(outs zprty:$Zdn), (ins zprty:$_Zdn, zprty:$Zm, imm0_7:$imm3),
+: I<(outs zprty:$Zdn), (ins zprty:$_Zdn, zprty:$Zm, imm32_0_7:$imm3),
   asm, "\t$Zdn, $_Zdn, $Zm, $imm3",
   "",
   []>, Sched<[]> {
   bits<5> Zdn;
   bits<5> Zm;
   bits<3> imm3;
   let Inst{31-24} = 0b01100101;
   let Inst{23-22} = sz;
   let Inst{21-19} = 0b010;
   let Inst{18-16} = imm3;
   let Inst{15-10} = 0b100000;
   let Inst{9-5}   = Zm;
   let Inst{4-0}   = Zdn;
 
   let Constraints = "$Zdn = $_Zdn";
   let DestructiveInstType = Destructive;
   let ElementSize = ElementSizeNone;
 }
 
-multiclass sve_fp_ftmad<string asm> {
+multiclass sve_fp_ftmad<string asm, SDPatternOperator op> {
   def _H : sve_fp_ftmad<0b01, asm, ZPR16>;
   def _S : sve_fp_ftmad<0b10, asm, ZPR32>;
   def _D : sve_fp_ftmad<0b11, asm, ZPR64>;
+
+  def : Pat<(nxv8f16 (op (nxv8f16 ZPR16:$Zn), (nxv8f16 ZPR16:$Zm), (i32 imm32_0_7:$imm))),
+            (!cast<Instruction>(NAME # _H) ZPR16:$Zn, ZPR16:$Zm, imm32_0_7:$imm)>;
+  def : Pat<(nxv4f32 (op (nxv4f32 ZPR32:$Zn), (nxv4f32 ZPR32:$Zm), (i32 imm32_0_7:$imm))),
+            (!cast<Instruction>(NAME # _S) ZPR32:$Zn, ZPR32:$Zm, imm32_0_7:$imm)>;
+  def : Pat<(nxv2f64 (op (nxv2f64 ZPR64:$Zn), (nxv2f64 ZPR64:$Zm), (i32 imm32_0_7:$imm))),
+            (!cast<Instruction>(NAME # _D) ZPR64:$Zn, ZPR64:$Zm, imm32_0_7:$imm)>;
 }
 
 
 //===----------------------------------------------------------------------===//
 // SVE Floating Point Arithmetic - Unpredicated Group
 //===----------------------------------------------------------------------===//
 
 class sve_fp_3op_u_zd<bits<2> sz, bits<3> opc, string asm, ZPRRegOp zprty>
: I<(outs zprty:$Zda), (ins PPR3bAny:$Pg, zprty:$_Zda, zprty:$Zn, zprty:$Zm),
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zda;
 
   let Constraints = "$Zda = $_Zda";
   let DestructiveInstType = Destructive;
   let ElementSize = zprty.ElementSize;
 }
 
-multiclass sve_fp_3op_p_zds_a<bits<2> opc, string asm> {
+multiclass sve_fp_3op_p_zds_a<bits<2> opc, string asm, SDPatternOperator op> {
   def _H : sve_fp_3op_p_zds_a<0b01, opc, asm, ZPR16>;
   def _S : sve_fp_3op_p_zds_a<0b10, opc, asm, ZPR32>;
   def _D : sve_fp_3op_p_zds_a<0b11, opc, asm, ZPR64>;
+
+  def : SVE_4_Op_Pat<nxv8f16, op, nxv8i1, nxv8f16, nxv8f16, nxv8f16, !cast<Instruction>(NAME # _H)>;
+  def : SVE_4_Op_Pat<nxv4f32, op, nxv4i1, nxv4f32, nxv4f32, nxv4f32, !cast<Instruction>(NAME # _S)>;
+  def : SVE_4_Op_Pat<nxv2f64, op, nxv2i1, nxv2f64, nxv2f64, nxv2f64, !cast<Instruction>(NAME # _D)>;
 }
 
 class sve_fp_3op_p_zds_b<bits<2> sz, bits<2> opc, string asm,
                          ZPRRegOp zprty>
 : I<(outs zprty:$Zdn), (ins PPR3bAny:$Pg, zprty:$_Zdn, zprty:$Zm, zprty:$Za),
   asm, "\t$Zdn, $Pg/m, $Zm, $Za",
   "",
   []>, Sched<[]> {
: I<(outs zprty:$Zdn), (ins PPR3bAny:$Pg, zprty:$_Zdn, zprty:$Zm, zprty:$Za),
   let Inst{9-5}   = Zm;
   let Inst{4-0}   = Zdn;
 
   let Constraints = "$Zdn = $_Zdn";
   let DestructiveInstType = Destructive;
   let ElementSize = zprty.ElementSize;
 }
 
-multiclass sve_fp_3op_p_zds_b<bits<2> opc, string asm> {
+multiclass sve_fp_3op_p_zds_b<bits<2> opc, string asm, SDPatternOperator op> {
   def _H : sve_fp_3op_p_zds_b<0b01, opc, asm, ZPR16>;
   def _S : sve_fp_3op_p_zds_b<0b10, opc, asm, ZPR32>;
   def _D : sve_fp_3op_p_zds_b<0b11, opc, asm, ZPR64>;
+
+  def : SVE_4_Op_Pat<nxv8f16, op, nxv8i1, nxv8f16, nxv8f16, nxv8f16, !cast<Instruction>(NAME # _H)>;
+  def : SVE_4_Op_Pat<nxv4f32, op, nxv4i1, nxv4f32, nxv4f32, nxv4f32, !cast<Instruction>(NAME # _S)>;
+  def : SVE_4_Op_Pat<nxv2f64, op, nxv2i1, nxv2f64, nxv2f64, nxv2f64, !cast<Instruction>(NAME # _D)>;
 }
 
 //===----------------------------------------------------------------------===//
 // SVE Floating Point Multiply-Add - Indexed Group
 //===----------------------------------------------------------------------===//
 
 class sve_fp_fma_by_indexed_elem<bits<2> sz, bit opc, string asm,
                                  ZPRRegOp zprty1,
: I<(outs zprty1:$Zda), (ins zprty1:$_Zda, zprty1:$Zn, zprty2:$Zm, itype:$iop),
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zda;
 
   let Constraints = "$Zda = $_Zda";
   let DestructiveInstType = Destructive;
   let ElementSize = ElementSizeNone;
 }
 
-multiclass sve_fp_fma_by_indexed_elem<bit opc, string asm> {
-  def _H : sve_fp_fma_by_indexed_elem<{0, ?}, opc, asm, ZPR16, ZPR3b16, VectorIndexH> {
+multiclass sve_fp_fma_by_indexed_elem<bit opc, string asm,
+                                      SDPatternOperator op> {
+  def _H : sve_fp_fma_by_indexed_elem<{0, ?}, opc, asm, 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 : sve_fp_fma_by_indexed_elem<0b10, opc, asm, ZPR32, ZPR3b32, VectorIndexS> {
+  def _S : sve_fp_fma_by_indexed_elem<0b10, opc, asm, ZPR32, ZPR3b32, VectorIndexS32b> {
     bits<3> Zm;
     bits<2> iop;
     let Inst{20-19} = iop;
     let Inst{18-16} = Zm;
   }
-  def _D : sve_fp_fma_by_indexed_elem<0b11, opc, asm, ZPR64, ZPR4b64, VectorIndexD> {
+  def _D : sve_fp_fma_by_indexed_elem<0b11, opc, asm, ZPR64, ZPR4b64, VectorIndexD32b> {
     bits<4> Zm;
     bit iop;
     let Inst{20} = iop;
     let Inst{19-16} = Zm;
   }
+
+  def : Pat<(nxv8f16 (op nxv8f16:$Op1, nxv8f16:$Op2, nxv8f16:$Op3, (i32 VectorIndexH32b:$idx))),
+            (!cast<Instruction>(NAME # _H) $Op1, $Op2, $Op3, VectorIndexH32b:$idx)>;
+  def : Pat<(nxv4f32 (op nxv4f32:$Op1, nxv4f32:$Op2, nxv4f32:$Op3, (i32 VectorIndexS32b:$idx))),
+            (!cast<Instruction>(NAME # _S) $Op1, $Op2, $Op3, VectorIndexS32b:$idx)>;
+  def : Pat<(nxv2f64 (op nxv2f64:$Op1, nxv2f64:$Op2, nxv2f64:$Op3, (i32 VectorIndexD32b:$idx))),
+            (!cast<Instruction>(NAME # _D) $Op1, $Op2, $Op3, VectorIndexD32b:$idx)>;
 }
 
 
 //===----------------------------------------------------------------------===//
 // SVE Floating Point Multiply - Indexed Group
 //===----------------------------------------------------------------------===//
 
 class sve_fp_fmul_by_indexed_elem<bits<2> sz, string asm, ZPRRegOp zprty,
                                       ZPRRegOp zprty2, Operand itype>
 : I<(outs zprty:$Zd), (ins zprty:$Zn, zprty2:$Zm, itype:$iop),
   asm, "\t$Zd, $Zn, $Zm$iop", "", []>, Sched<[]> {
   bits<5> Zd;
   bits<5> Zn;
   let Inst{31-24} = 0b01100100;
   let Inst{23-22} = sz;
   let Inst{21}    = 0b1;
   let Inst{15-10} = 0b001000;
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zd;
 }
 
-multiclass sve_fp_fmul_by_indexed_elem<string asm> {
-  def _H : sve_fp_fmul_by_indexed_elem<{0, ?}, asm, ZPR16, ZPR3b16, VectorIndexH> {
+multiclass sve_fp_fmul_by_indexed_elem<string asm, SDPatternOperator op> {
+  def _H : sve_fp_fmul_by_indexed_elem<{0, ?}, asm, 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 : sve_fp_fmul_by_indexed_elem<0b10, asm, ZPR32, ZPR3b32, VectorIndexS> {
+  def _S : sve_fp_fmul_by_indexed_elem<0b10, asm, ZPR32, ZPR3b32, VectorIndexS32b> {
     bits<3> Zm;
     bits<2> iop;
     let Inst{20-19} = iop;
     let Inst{18-16} = Zm;
   }
-  def _D : sve_fp_fmul_by_indexed_elem<0b11, asm, ZPR64, ZPR4b64, VectorIndexD> {
+  def _D : sve_fp_fmul_by_indexed_elem<0b11, asm, ZPR64, ZPR4b64, VectorIndexD32b> {
     bits<4> Zm;
     bit iop;
     let Inst{20} = iop;
     let Inst{19-16} = Zm;
   }
+
+  def : Pat<(nxv8f16 (op nxv8f16:$Op1, nxv8f16:$Op2, (i32 VectorIndexH32b:$idx))),
+            (!cast<Instruction>(NAME # _H) $Op1, $Op2, VectorIndexH32b:$idx)>;
+  def : Pat<(nxv4f32 (op nxv4f32:$Op1, nxv4f32:$Op2, (i32 VectorIndexS32b:$idx))),
+            (!cast<Instruction>(NAME # _S) $Op1, $Op2, VectorIndexS32b:$idx)>;
+  def : Pat<(nxv2f64 (op nxv2f64:$Op1, nxv2f64:$Op2, (i32 VectorIndexD32b:$idx))),
+            (!cast<Instruction>(NAME # _D) $Op1, $Op2, VectorIndexD32b:$idx)>;
 }
 
 //===----------------------------------------------------------------------===//
 // SVE Floating Point Complex Multiply-Add Group
 //===----------------------------------------------------------------------===//
 
 class sve_fp_fcmla<bits<2> sz, string asm, ZPRRegOp zprty>
 : I<(outs zprty:$Zda), (ins PPR3bAny:$Pg, zprty:$_Zda, zprty:$Zn, zprty:$Zm,
: I<(outs zprty:$Zda), (ins PPR3bAny:$Pg, zprty:$_Zda, zprty:$Zn, zprty:$Zm,
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zda;
 
   let Constraints = "$Zda = $_Zda";
   let DestructiveInstType = Destructive;
   let ElementSize = zprty.ElementSize;
 }
 
-multiclass sve_fp_fcmla<string asm> {
+multiclass sve_fp_fcmla<string asm, SDPatternOperator op> {
   def _H : sve_fp_fcmla<0b01, asm, ZPR16>;
   def _S : sve_fp_fcmla<0b10, asm, ZPR32>;
   def _D : sve_fp_fcmla<0b11, asm, ZPR64>;
+
+  def : Pat<(nxv8f16 (op nxv8i1:$Op1, nxv8f16:$Op2, nxv8f16:$Op3, nxv8f16:$Op4, (i32 complexrotateop:$imm))),
+            (!cast<Instruction>(NAME # _H) $Op1, $Op2, $Op3, $Op4, complexrotateop:$imm)>;
+  def : Pat<(nxv4f32 (op nxv4i1:$Op1, nxv4f32:$Op2, nxv4f32:$Op3, nxv4f32:$Op4, (i32 complexrotateop:$imm))),
+            (!cast<Instruction>(NAME # _S) $Op1, $Op2, $Op3, $Op4, complexrotateop:$imm)>;
+  def : Pat<(nxv2f64 (op nxv2i1:$Op1, nxv2f64:$Op2, nxv2f64:$Op3, nxv2f64:$Op4, (i32 complexrotateop:$imm))),
+            (!cast<Instruction>(NAME # _D) $Op1, $Op2, $Op3, $Op4, complexrotateop:$imm)>;
 }
 
 //===----------------------------------------------------------------------===//
 // SVE Floating Point Complex Multiply-Add - Indexed Group
 //===----------------------------------------------------------------------===//
 
 class sve_fp_fcmla_by_indexed_elem<bits<2> sz, string asm,
                                    ZPRRegOp zprty,
: I<(outs zprty:$Zda), (ins zprty:$_Zda, zprty:$Zn, zprty2:$Zm, itype:$iop,
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zda;
 
   let Constraints = "$Zda = $_Zda";
   let DestructiveInstType = Destructive;
   let ElementSize = ElementSizeNone;
 }
 
-multiclass sve_fp_fcmla_by_indexed_elem<string asm> {
-  def _H : sve_fp_fcmla_by_indexed_elem<0b10, asm, ZPR16, ZPR3b16, VectorIndexS> {
+multiclass sve_fp_fcmla_by_indexed_elem<string asm, SDPatternOperator op> {
+  def _H : sve_fp_fcmla_by_indexed_elem<0b10, asm, ZPR16, ZPR3b16, VectorIndexS32b> {
     bits<3> Zm;
     bits<2> iop;
     let Inst{20-19} = iop;
     let Inst{18-16} = Zm;
   }
-  def _S : sve_fp_fcmla_by_indexed_elem<0b11, asm, ZPR32, ZPR4b32, VectorIndexD> {
+  def _S : sve_fp_fcmla_by_indexed_elem<0b11, asm, ZPR32, ZPR4b32, VectorIndexD32b> {
     bits<4> Zm;
     bits<1> iop;
     let Inst{20} = iop;
     let Inst{19-16} = Zm;
   }
+
+  def : Pat<(nxv8f16 (op nxv8f16:$Op1, nxv8f16:$Op2, nxv8f16:$Op3, (i32 VectorIndexS32b:$idx), (i32 complexrotateop:$imm))),
+            (!cast<Instruction>(NAME # _H) $Op1, $Op2, $Op3, VectorIndexS32b:$idx, complexrotateop:$imm)>;
+  def : Pat<(nxv4f32 (op nxv4f32:$Op1, nxv4f32:$Op2, nxv4f32:$Op3, (i32 VectorIndexD32b:$idx), (i32 complexrotateop:$imm))),
+            (!cast<Instruction>(NAME # _S) $Op1, $Op2, $Op3, VectorIndexD32b:$idx, complexrotateop:$imm)>;
 }
 
 //===----------------------------------------------------------------------===//
 // SVE Floating Point Complex Addition Group
 //===----------------------------------------------------------------------===//
 
 class sve_fp_fcadd<bits<2> sz, string asm, ZPRRegOp zprty>
 : I<(outs zprty:$Zdn), (ins PPR3bAny:$Pg, zprty:$_Zdn, zprty:$Zm,
: I<(outs zprty:$Zdn), (ins PPR3bAny:$Pg, zprty:$_Zdn, zprty:$Zm,
   let Inst{9-5}   = Zm;
   let Inst{4-0}   = Zdn;
 
   let Constraints = "$Zdn = $_Zdn";
   let DestructiveInstType = Destructive;
   let ElementSize = zprty.ElementSize;
 }
 
-multiclass sve_fp_fcadd<string asm> {
+multiclass sve_fp_fcadd<string asm, SDPatternOperator op> {
   def _H : sve_fp_fcadd<0b01, asm, ZPR16>;
   def _S : sve_fp_fcadd<0b10, asm, ZPR32>;
   def _D : sve_fp_fcadd<0b11, asm, ZPR64>;
+
+  def : Pat<(nxv8f16 (op nxv8i1:$Op1, nxv8f16:$Op2, nxv8f16:$Op3, (i32 complexrotateopodd:$imm))),
+            (!cast<Instruction>(NAME # _H) $Op1, $Op2, $Op3, complexrotateopodd:$imm)>;
+  def : Pat<(nxv4f32 (op nxv4i1:$Op1, nxv4f32:$Op2, nxv4f32:$Op3, (i32 complexrotateopodd:$imm))),
+            (!cast<Instruction>(NAME # _S) $Op1, $Op2, $Op3, complexrotateopodd:$imm)>;
+  def : Pat<(nxv2f64 (op nxv2i1:$Op1, nxv2f64:$Op2, nxv2f64:$Op3, (i32 complexrotateopodd:$imm))),
+            (!cast<Instruction>(NAME # _D) $Op1, $Op2, $Op3, complexrotateopodd:$imm)>;
 }
 
 //===----------------------------------------------------------------------===//
 // SVE2 Floating Point Convert Group
 //===----------------------------------------------------------------------===//
 
 class sve2_fp_convert_precision<bits<4> opc, string asm,
                                 ZPRRegOp zprty1, ZPRRegOp zprty2>
: I<(outs zprty:$Zd), (ins zprty:$Zn, zprty:$Zm),
   let Inst{23-22} = sz;
   let Inst{21}    = 0b1;
   let Inst{20-16} = Zm;
   let Inst{15-10} = 0b101100;
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zd;
 }
 
-multiclass sve_int_bin_cons_misc_0_b<string asm> {
+multiclass sve_int_bin_cons_misc_0_b<string asm, SDPatternOperator op> {
   def _H : sve_int_bin_cons_misc_0_b<0b01, asm, ZPR16>;
   def _S : sve_int_bin_cons_misc_0_b<0b10, asm, ZPR32>;
   def _D : sve_int_bin_cons_misc_0_b<0b11, asm, ZPR64>;
+
+  def : SVE_2_Op_Pat<nxv8f16, op, nxv8f16, nxv8i16, !cast<Instruction>(NAME # _H)>;
+  def : SVE_2_Op_Pat<nxv4f32, op, nxv4f32, nxv4i32, !cast<Instruction>(NAME # _S)>;
+  def : SVE_2_Op_Pat<nxv2f64, op, nxv2f64, nxv2i64, !cast<Instruction>(NAME # _D)>;
 }
 
 class sve_int_bin_cons_misc_0_c<bits<8> opc, string asm, ZPRRegOp zprty>
 : I<(outs zprty:$Zd), (ins zprty:$Zn),
   asm, "\t$Zd, $Zn",
   "",
   []>, Sched<[]> {
   bits<5> Zd;

llvm/test/CodeGen/AArch64/sve-intrinsics-fp-arith.ll

 ; CHECK-NEXT: ret
   %out = call <vscale x 2 x double> @llvm.aarch64.sve.fadd.nxv2f64(<vscale x 2 x i1> %pg,
                                                                    <vscale x 2 x double> %a,
                                                                    <vscale x 2 x double> %b)
   ret <vscale x 2 x double> %out
 }
 
 ;
+; FCADD
+;
+
+define <vscale x 8 x half> @fcadd_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fcadd_h:
+; CHECK: fcadd z0.h, p0/m, z0.h, z1.h, #90
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fcadd.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                  <vscale x 8 x half> %a,
+                                                                  <vscale x 8 x half> %b,
+                                                                  i32 90)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fcadd_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fcadd_s:
+; CHECK: fcadd z0.s, p0/m, z0.s, z1.s, #270
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fcadd.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                   <vscale x 4 x float> %a,
+                                                                   <vscale x 4 x float> %b,
+                                                                   i32 270)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fcadd_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fcadd_d:
+; CHECK: fcadd z0.d, p0/m, z0.d, z1.d, #90
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fcadd.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                    <vscale x 2 x double> %a,
+                                                                    <vscale x 2 x double> %b,
+                                                                    i32 90)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FCMLA
+;
+
+define <vscale x 8 x half> @fcmla_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fcmla_h:
+; CHECK: fcmla z0.h, p0/m, z1.h, z2.h, #90
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fcmla.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                  <vscale x 8 x half> %a,
+                                                                  <vscale x 8 x half> %b,
+                                                                  <vscale x 8 x half> %c,
+                                                                  i32 90)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fcmla_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fcmla_s:
+; CHECK: fcmla z0.s, p0/m, z1.s, z2.s, #180
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fcmla.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                   <vscale x 4 x float> %a,
+                                                                   <vscale x 4 x float> %b,
+                                                                   <vscale x 4 x float> %c,
+                                                                   i32 180)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fcmla_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fcmla_d:
+; CHECK: fcmla z0.d, p0/m, z1.d, z2.d, #270
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fcmla.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                    <vscale x 2 x double> %a,
+                                                                    <vscale x 2 x double> %b,
+                                                                    <vscale x 2 x double> %c,
+                                                                    i32 270)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FCMLA (Indexed)
+;
+
+define <vscale x 8 x half> @fcmla_lane_h(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fcmla_lane_h:
+; CHECK: fcmla z0.h, z1.h, z2.h[3], #0
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fcmla.lane.nxv8f16(<vscale x 8 x half> %a,
+                                                                       <vscale x 8 x half> %b,
+                                                                       <vscale x 8 x half> %c,
+                                                                       i32 3,
+                                                                       i32 0)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fcmla_lane_s(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fcmla_lane_s:
+; CHECK: fcmla z0.s, z1.s, z2.s[1], #90
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fcmla.lane.nxv4f32(<vscale x 4 x float> %a,
+                                                                        <vscale x 4 x float> %b,
+                                                                        <vscale x 4 x float> %c,
+                                                                        i32 1,
+                                                                        i32 90)
+  ret <vscale x 4 x float> %out
+}
+
+;
 ; FDIV
 ;
 
 define <vscale x 8 x half> @fdiv_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
 ; CHECK-LABEL: fdiv_h:
 ; CHECK: fdiv z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT: ret
   %out = call <vscale x 8 x half> @llvm.aarch64.sve.fdiv.nxv8f16(<vscale x 8 x i1> %pg,

 ; CHECK-NEXT: ret
   %out = call <vscale x 2 x double> @llvm.aarch64.sve.fdivr.nxv2f64(<vscale x 2 x i1> %pg,
                                                                     <vscale x 2 x double> %a,
                                                                     <vscale x 2 x double> %b)
   ret <vscale x 2 x double> %out
 }
 
 ;
+; FMAD
+;
+
+define <vscale x 8 x half> @fmad_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fmad_h:
+; CHECK: fmad z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmad.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x half> %a,
+                                                                 <vscale x 8 x half> %b,
+                                                                 <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmad_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fmad_s:
+; CHECK: fmad z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmad.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                  <vscale x 4 x float> %a,
+                                                                  <vscale x 4 x float> %b,
+                                                                  <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmad_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fmad_d:
+; CHECK: fmad z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmad.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                   <vscale x 2 x double> %a,
+                                                                   <vscale x 2 x double> %b,
+                                                                   <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %out
+}
+
+;
 ; FMAX
 ;
 
 define <vscale x 8 x half> @fmax_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
 ; CHECK-LABEL: fmax_h:
 ; CHECK: fmax z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT: ret
   %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmax.nxv8f16(<vscale x 8 x i1> %pg,

 ; CHECK-NEXT: ret
   %out = call <vscale x 2 x double> @llvm.aarch64.sve.fminnm.nxv2f64(<vscale x 2 x i1> %pg,
                                                                      <vscale x 2 x double> %a,
                                                                      <vscale x 2 x double> %b)
   ret <vscale x 2 x double> %out
 }
 
 ;
+; FMLA
+;
+
+define <vscale x 8 x half> @fmla_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fmla_h:
+; CHECK: fmla z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmla.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x half> %a,
+                                                                 <vscale x 8 x half> %b,
+                                                                 <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmla_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fmla_s:
+; CHECK: fmla z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmla.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                  <vscale x 4 x float> %a,
+                                                                  <vscale x 4 x float> %b,
+                                                                  <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmla_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fmla_d:
+; CHECK: fmla z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmla.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                   <vscale x 2 x double> %a,
+                                                                   <vscale x 2 x double> %b,
+                                                                   <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMLA (Indexed)
+;
+
+define <vscale x 8 x half> @fmla_lane_h(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fmla_lane_h:
+; CHECK: fmla z0.h, z1.h, z2.h[3]
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmla.lane.nxv8f16(<vscale x 8 x half> %a,
+                                                                      <vscale x 8 x half> %b,
+                                                                      <vscale x 8 x half> %c,
+                                                                      i32 3)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmla_lane_s(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fmla_lane_s:
+; CHECK: fmla z0.s, z1.s, z2.s[2]
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmla.lane.nxv4f32(<vscale x 4 x float> %a,
+                                                                       <vscale x 4 x float> %b,
+                                                                       <vscale x 4 x float> %c,
+                                                                       i32 2)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmla_lane_d(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fmla_lane_d:
+; CHECK: fmla z0.d, z1.d, z2.d[1]
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmla.lane.nxv2f64(<vscale x 2 x double> %a,
+                                                                        <vscale x 2 x double> %b,
+                                                                        <vscale x 2 x double> %c,
+                                                                        i32 1)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMLS
+;
+
+define <vscale x 8 x half> @fmls_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fmls_h:
+; CHECK: fmls z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmls.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x half> %a,
+                                                                 <vscale x 8 x half> %b,
+                                                                 <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmls_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fmls_s:
+; CHECK: fmls z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmls.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                  <vscale x 4 x float> %a,
+                                                                  <vscale x 4 x float> %b,
+                                                                  <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmls_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fmls_d:
+; CHECK: fmls z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmls.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                   <vscale x 2 x double> %a,
+                                                                   <vscale x 2 x double> %b,
+                                                                   <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMLS (Indexed)
+;
+
+define <vscale x 8 x half> @fmls_lane_h(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fmls_lane_h:
+; CHECK: fmls z0.h, z1.h, z2.h[3]
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmls.lane.nxv8f16(<vscale x 8 x half> %a,
+                                                                 <vscale x 8 x half> %b,
+                                                                 <vscale x 8 x half> %c,
+                                                                 i32 3)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmls_lane_s(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fmls_lane_s:
+; CHECK: fmls z0.s, z1.s, z2.s[2]
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmls.lane.nxv4f32(<vscale x 4 x float> %a,
+                                                                  <vscale x 4 x float> %b,
+                                                                  <vscale x 4 x float> %c,
+                                                                  i32 2)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmls_lane_d(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fmls_lane_d:
+; CHECK: fmls z0.d, z1.d, z2.d[1]
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmls.lane.nxv2f64(<vscale x 2 x double> %a,
+                                                                   <vscale x 2 x double> %b,
+                                                                   <vscale x 2 x double> %c,
+                                                                   i32 1)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMSB
+;
+
+define <vscale x 8 x half> @fmsb_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fmsb_h:
+; CHECK: fmsb z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmsb.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x half> %a,
+                                                                 <vscale x 8 x half> %b,
+                                                                 <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmsb_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fmsb_s:
+; CHECK: fmsb z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmsb.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                  <vscale x 4 x float> %a,
+                                                                  <vscale x 4 x float> %b,
+                                                                  <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmsb_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fmsb_d:
+; CHECK: fmsb z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmsb.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                   <vscale x 2 x double> %a,
+                                                                   <vscale x 2 x double> %b,
+                                                                   <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %out
+}
+
+;
 ; FMUL
 ;
 
 define <vscale x 8 x half> @fmul_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
 ; CHECK-LABEL: fmul_h:
 ; CHECK: fmul z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT: ret
   %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmul.nxv8f16(<vscale x 8 x i1> %pg,

 ; CHECK-NEXT: ret
   %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmul.nxv2f64(<vscale x 2 x i1> %pg,
                                                                    <vscale x 2 x double> %a,
                                                                    <vscale x 2 x double> %b)
   ret <vscale x 2 x double> %out
 }
 
 ;
+; FMUL (Indexed)
+;
+
+define <vscale x 8 x half> @fmul_lane_h(<vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fmul_lane_h:
+; CHECK: fmul z0.h, z0.h, z1.h[3]
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmul.lane.nxv8f16(<vscale x 8 x half> %a,
+                                                                      <vscale x 8 x half> %b,
+                                                                      i32 3)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmul_lane_s(<vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fmul_lane_s:
+; CHECK: fmul z0.s, z0.s, z1.s[2]
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmul.lane.nxv4f32(<vscale x 4 x float> %a,
+                                                                       <vscale x 4 x float> %b,
+                                                                       i32 2)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmul_lane_d(<vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fmul_lane_d:
+; CHECK: fmul z0.d, z0.d, z1.d[1]
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmul.lane.nxv2f64(<vscale x 2 x double> %a,
+                                                                        <vscale x 2 x double> %b,
+                                                                        i32 1)
+  ret <vscale x 2 x double> %out
+}
+
+;
 ; FMULX
 ;
 
 define <vscale x 8 x half> @fmulx_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
 ; CHECK-LABEL: fmulx_h:
 ; CHECK: fmulx z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT: ret
   %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmulx.nxv8f16(<vscale x 8 x i1> %pg,

 ; CHECK-NEXT: ret
   %out = call <vscale x 2 x double> @llvm.aarch64.sve.fscale.nxv2f64(<vscale x 2 x i1> %pg,
                                                                      <vscale x 2 x double> %a,
                                                                      <vscale x 2 x i64> %b)
   ret <vscale x 2 x double> %out
 }
 
 ;
+; FNMAD
+;
+
+define <vscale x 8 x half> @fnmad_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fnmad_h:
+; CHECK: fnmad z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fnmad.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                  <vscale x 8 x half> %a,
+                                                                  <vscale x 8 x half> %b,
+                                                                  <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fnmad_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fnmad_s:
+; CHECK: fnmad z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fnmad.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                   <vscale x 4 x float> %a,
+                                                                   <vscale x 4 x float> %b,
+                                                                   <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fnmad_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fnmad_d:
+; CHECK: fnmad z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fnmad.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                    <vscale x 2 x double> %a,
+                                                                    <vscale x 2 x double> %b,
+                                                                    <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FNMLA
+;
+
+define <vscale x 8 x half> @fnmla_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fnmla_h:
+; CHECK: fnmla z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fnmla.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                  <vscale x 8 x half> %a,
+                                                                  <vscale x 8 x half> %b,
+                                                                  <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fnmla_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fnmla_s:
+; CHECK: fnmla z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fnmla.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                   <vscale x 4 x float> %a,
+                                                                   <vscale x 4 x float> %b,
+                                                                   <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fnmla_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fnmla_d:
+; CHECK: fnmla z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fnmla.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                    <vscale x 2 x double> %a,
+                                                                    <vscale x 2 x double> %b,
+                                                                    <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FNMLS
+;
+
+define <vscale x 8 x half> @fnmls_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fnmls_h:
+; CHECK: fnmls z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fnmls.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                  <vscale x 8 x half> %a,
+                                                                  <vscale x 8 x half> %b,
+                                                                  <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fnmls_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fnmls_s:
+; CHECK: fnmls z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fnmls.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                   <vscale x 4 x float> %a,
+                                                                   <vscale x 4 x float> %b,
+                                                                   <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fnmls_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fnmls_d:
+; CHECK: fnmls z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fnmls.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                    <vscale x 2 x double> %a,
+                                                                    <vscale x 2 x double> %b,
+                                                                    <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FNMSB
+;
+
+define <vscale x 8 x half> @fnmsb_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fnmsb_h:
+; CHECK: fnmsb z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fnmsb.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                  <vscale x 8 x half> %a,
+                                                                  <vscale x 8 x half> %b,
+                                                                  <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fnmsb_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fnmsb_s:
+; CHECK: fnmsb z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fnmsb.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                   <vscale x 4 x float> %a,
+                                                                   <vscale x 4 x float> %b,
+                                                                   <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fnmsb_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fnmsb_d:
+; CHECK: fnmsb z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fnmsb.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                    <vscale x 2 x double> %a,
+                                                                    <vscale x 2 x double> %b,
+                                                                    <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %out
+}
+
+;
 ; FSUB
 ;
 
 define <vscale x 8 x half> @fsub_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
 ; CHECK-LABEL: fsub_h:
 ; CHECK: fsub z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT: ret
   %out = call <vscale x 8 x half> @llvm.aarch64.sve.fsub.nxv8f16(<vscale x 8 x i1> %pg,

 ; CHECK-NEXT: ret
   %out = call <vscale x 2 x double> @llvm.aarch64.sve.fsubr.nxv2f64(<vscale x 2 x i1> %pg,
                                                                     <vscale x 2 x double> %a,
                                                                     <vscale x 2 x double> %b)
   ret <vscale x 2 x double> %out
 }
 
 ;
+; FTMAD
+;
+
+define <vscale x 8 x half> @ftmad_h(<vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: ftmad_h:
+; CHECK: ftmad z0.h, z0.h, z1.h, #0
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.ftmad.x.nxv8f16(<vscale x 8 x half> %a,
+                                                                    <vscale x 8 x half> %b,
+                                                                    i32 0)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @ftmad_s(<vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: ftmad_s:
+; CHECK: ftmad z0.s, z0.s, z1.s, #0
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.ftmad.x.nxv4f32(<vscale x 4 x float> %a,
+                                                                     <vscale x 4 x float> %b,
+                                                                     i32 0)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @ftmad_d(<vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: ftmad_d:
+; CHECK: ftmad z0.d, z0.d, z1.d, #7
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.ftmad.x.nxv2f64(<vscale x 2 x double> %a,
+                                                                      <vscale x 2 x double> %b,
+                                                                      i32 7)
+  ret <vscale x 2 x double> %out
+}
+
+;
 ; FTSMUL
 ;
 
 define <vscale x 8 x half> @ftsmul_h(<vscale x 8 x half> %a, <vscale x 8 x i16> %b) {
 ; CHECK-LABEL: ftsmul_h:
 ; CHECK: ftsmul z0.h, z0.h, z1.h
 ; CHECK-NEXT: ret
   %out = call <vscale x 8 x half> @llvm.aarch64.sve.ftsmul.x.nxv8f16(<vscale x 8 x half> %a,

 ; CHECK-LABEL: ftsmul_d:
 ; CHECK: ftsmul z0.d, z0.d, z1.d
 ; CHECK-NEXT: ret
   %out = call <vscale x 2 x double> @llvm.aarch64.sve.ftsmul.x.nxv2f64(<vscale x 2 x double> %a,
                                                                        <vscale x 2 x i64> %b)
   ret <vscale x 2 x double> %out
 }
 
+;
+; FTSSEL
+;
+
+define <vscale x 8 x half> @ftssel_h(<vscale x 8 x half> %a, <vscale x 8 x i16> %b) {
+; CHECK-LABEL: ftssel_h:
+; CHECK: ftssel z0.h, z0.h, z1.h
+; CHECK-NEXT: ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.ftssel.x.nxv8f16(<vscale x 8 x half> %a,
+                                                                     <vscale x 8 x i16> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @ftssel_s(<vscale x 4 x float> %a, <vscale x 4 x i32> %b) {
+; CHECK-LABEL: ftssel_s:
+; CHECK: ftssel z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.ftssel.x.nxv4f32(<vscale x 4 x float> %a,
+                                                                      <vscale x 4 x i32> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @ftssel_d(<vscale x 2 x double> %a, <vscale x 2 x i64> %b) {
+; CHECK-LABEL: ftssel_d:
+; CHECK: ftssel z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.ftssel.x.nxv2f64(<vscale x 2 x double> %a,
+                                                                       <vscale x 2 x i64> %b)
+  ret <vscale x 2 x double> %out
+}
+
 declare <vscale x 8 x half> @llvm.aarch64.sve.fabd.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fabd.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fabd.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
 
 declare <vscale x 8 x half> @llvm.aarch64.sve.fadd.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fadd.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fadd.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
 
+declare <vscale x 8 x half> @llvm.aarch64.sve.fcadd.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fcadd.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fcadd.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fcmla.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fcmla.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fcmla.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fcmla.lane.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, i32, i32)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fcmla.lane.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, i32, i32)
+
 declare <vscale x 8 x half> @llvm.aarch64.sve.fdiv.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fdiv.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fdiv.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
 
 declare <vscale x 8 x half> @llvm.aarch64.sve.fdivr.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fdivr.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fdivr.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
 
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmad.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmad.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmad.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
+
 declare <vscale x 8 x half> @llvm.aarch64.sve.fmax.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fmax.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fmax.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
 
 declare <vscale x 8 x half> @llvm.aarch64.sve.fmaxnm.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fmaxnm.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fmaxnm.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
 
 declare <vscale x 8 x half> @llvm.aarch64.sve.fmin.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fmin.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fmin.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
 
 declare <vscale x 8 x half> @llvm.aarch64.sve.fminnm.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fminnm.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fminnm.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
 
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmla.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmla.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmla.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmla.lane.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmla.lane.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmla.lane.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmls.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmls.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmls.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmls.lane.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmls.lane.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmls.lane.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmsb.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmsb.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmsb.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
+
 declare <vscale x 8 x half> @llvm.aarch64.sve.fmul.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fmul.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fmul.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
 
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmul.lane.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, i32)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmul.lane.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, i32)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmul.lane.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, i32)
+
 declare <vscale x 8 x half> @llvm.aarch64.sve.fmulx.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fmulx.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fmulx.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
 
+declare <vscale x 8 x half> @llvm.aarch64.sve.fnmad.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fnmad.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fnmad.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fnmla.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fnmla.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fnmla.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fnmls.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fnmls.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fnmls.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fnmsb.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fnmsb.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fnmsb.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
+
 declare <vscale x 8 x half> @llvm.aarch64.sve.fscale.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x i16>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fscale.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x i32>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fscale.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x i64>)
 
 declare <vscale x 8 x half> @llvm.aarch64.sve.fsub.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fsub.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fsub.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
 
 declare <vscale x 8 x half> @llvm.aarch64.sve.fsubr.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.fsubr.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.fsubr.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
 
+declare <vscale x 8 x half> @llvm.aarch64.sve.ftmad.x.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, i32)
+declare <vscale x 4 x float> @llvm.aarch64.sve.ftmad.x.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, i32)
+declare <vscale x 2 x double> @llvm.aarch64.sve.ftmad.x.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, i32)
+
 declare <vscale x 8 x half> @llvm.aarch64.sve.ftsmul.x.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i16>)
 declare <vscale x 4 x float> @llvm.aarch64.sve.ftsmul.x.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i32>)
 declare <vscale x 2 x double> @llvm.aarch64.sve.ftsmul.x.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i64>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.ftssel.x.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i16>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.ftssel.x.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i32>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.ftssel.x.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i64>)