[AArch64][SVE] Better utilisation of unpredicated forms of remaining intrinsics

[bsmith]

Summary

When using predicated intrinsics, if the predicate used is all lanes active,
use an unpredicated form of the instruction, additionally this allows for
better use of immediate forms.

This only includes instructions where the unpredicated/predicated forms
matched in such a way that instruction selection would not introduce extra
ptrue instructions. This allows us to convert the intrinsics directly to
architecture independent ISD nodes.

Depends on D101062

[joechrisellis]

One nit, otherwise LGTM.

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

13784

nit: double negative -- I think the code would read better if this was PredicatedOp.

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

if (N.getOpcode() == AArch64ISD::PTRUE &&
     return N.getValueType().getVectorMinNumElements() >= NumElts;
 
   return false;
 }
 
 // If a merged operation has no inactive lanes we can relax it to a predicated
 // or unpredicated operation, which potentially allows better isel (perhaps
 // using immediate forms) or relaxing register reuse requirements.
-static SDValue convertMergedOpToPredOp(SDNode *N, unsigned PredOpc,
-                                       SelectionDAG &DAG) {
+static SDValue convertMergedOpToPredOp(SDNode *N, unsigned Opc,
+                                       SelectionDAG &DAG,
+                                       bool UnpredOp = false) {
   assert(N->getOpcode() == ISD::INTRINSIC_WO_CHAIN && "Expected intrinsic!");
   assert(N->getNumOperands() == 4 && "Expected 3 operand intrinsic!");
   SDValue Pg = N->getOperand(1);
 
   // ISD way to specify an all active predicate.
-  if (isAllActivePredicate(Pg))
-    return DAG.getNode(PredOpc, SDLoc(N), N->getValueType(0), Pg,
+  if (isAllActivePredicate(Pg)) {
+    if (UnpredOp)
+      return DAG.getNode(Opc, SDLoc(N), N->getValueType(0), N->getOperand(2),
+                         N->getOperand(3));
+    else
+      return DAG.getNode(Opc, SDLoc(N), N->getValueType(0), Pg,
                          N->getOperand(2), N->getOperand(3));
+  }
 
   // FUTURE: SplatVector(true)
   return SDValue();
 }
 
 static SDValue performIntrinsicCombine(SDNode *N,
                                        TargetLowering::DAGCombinerInfo &DCI,
                                        const AArch64Subtarget *Subtarget) {
static SDValue performIntrinsicCombine(SDNode *N,
   case Intrinsic::aarch64_sve_asr:
     return convertMergedOpToPredOp(N, AArch64ISD::SRA_PRED, DAG);
   case Intrinsic::aarch64_sve_fadd:
     return convertMergedOpToPredOp(N, AArch64ISD::FADD_PRED, DAG);
   case Intrinsic::aarch64_sve_fsub:
     return convertMergedOpToPredOp(N, AArch64ISD::FSUB_PRED, DAG);
   case Intrinsic::aarch64_sve_fmul:
     return convertMergedOpToPredOp(N, AArch64ISD::FMUL_PRED, DAG);
+  case Intrinsic::aarch64_sve_add:
+    return convertMergedOpToPredOp(N, ISD::ADD, DAG, true);
+  case Intrinsic::aarch64_sve_sub:
+    return convertMergedOpToPredOp(N, ISD::SUB, DAG, true);
+  case Intrinsic::aarch64_sve_and:
+    return convertMergedOpToPredOp(N, ISD::AND, DAG, true);
+  case Intrinsic::aarch64_sve_eor:
+    return convertMergedOpToPredOp(N, ISD::XOR, DAG, true);
+  case Intrinsic::aarch64_sve_orr:
+    return convertMergedOpToPredOp(N, ISD::OR, DAG, true);
+  case Intrinsic::aarch64_sve_sqadd:
+    return convertMergedOpToPredOp(N, ISD::SADDSAT, DAG, true);
+  case Intrinsic::aarch64_sve_sqsub:
+    return convertMergedOpToPredOp(N, ISD::SSUBSAT, DAG, true);
+  case Intrinsic::aarch64_sve_uqadd:
+    return convertMergedOpToPredOp(N, ISD::UADDSAT, DAG, true);
+  case Intrinsic::aarch64_sve_uqsub:
+    return convertMergedOpToPredOp(N, ISD::USUBSAT, DAG, true);
+  case Intrinsic::aarch64_sve_sqadd_x:
+    return DAG.getNode(ISD::SADDSAT, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(2));
+  case Intrinsic::aarch64_sve_sqsub_x:
+    return DAG.getNode(ISD::SSUBSAT, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(2));
+  case Intrinsic::aarch64_sve_uqadd_x:
+    return DAG.getNode(ISD::UADDSAT, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(2));
+  case Intrinsic::aarch64_sve_uqsub_x:
+    return DAG.getNode(ISD::USUBSAT, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(2));
   case Intrinsic::aarch64_sve_cmphs:
     if (!N->getOperand(2).getValueType().isFloatingPoint())
       return DAG.getNode(AArch64ISD::SETCC_MERGE_ZERO, SDLoc(N),
                          N->getValueType(0), N->getOperand(1), N->getOperand(2),
                          N->getOperand(3), DAG.getCondCode(ISD::SETUGE));
     break;
   case Intrinsic::aarch64_sve_cmphi:
     if (!N->getOperand(2).getValueType().isFloatingPoint())

llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td

 
 let Predicates = [HasSVE] in {
   defm RDFFR_PPz  : sve_int_rdffr_pred<0b0, "rdffr", int_aarch64_sve_rdffr_z>;
   def  RDFFRS_PPz : sve_int_rdffr_pred<0b1, "rdffrs">;
   defm RDFFR_P    : sve_int_rdffr_unpred<"rdffr", int_aarch64_sve_rdffr>;
   def  SETFFR     : sve_int_setffr<"setffr", int_aarch64_sve_setffr>;
   def  WRFFR      : sve_int_wrffr<"wrffr", int_aarch64_sve_wrffr>;
 
-  defm ADD_ZZZ   : sve_int_bin_cons_arit_0<0b000, "add", add, null_frag>;
-  defm SUB_ZZZ   : sve_int_bin_cons_arit_0<0b001, "sub", sub, null_frag>;
-  defm SQADD_ZZZ : sve_int_bin_cons_arit_0<0b100, "sqadd", saddsat, int_aarch64_sve_sqadd_x>;
-  defm UQADD_ZZZ : sve_int_bin_cons_arit_0<0b101, "uqadd", uaddsat, int_aarch64_sve_uqadd_x>;
-  defm SQSUB_ZZZ : sve_int_bin_cons_arit_0<0b110, "sqsub", ssubsat, int_aarch64_sve_sqsub_x>;
-  defm UQSUB_ZZZ : sve_int_bin_cons_arit_0<0b111, "uqsub", usubsat, int_aarch64_sve_uqsub_x>;
+  defm ADD_ZZZ   : sve_int_bin_cons_arit_0<0b000, "add", add>;
+  defm SUB_ZZZ   : sve_int_bin_cons_arit_0<0b001, "sub", sub>;
+  defm SQADD_ZZZ : sve_int_bin_cons_arit_0<0b100, "sqadd", saddsat>;
+  defm UQADD_ZZZ : sve_int_bin_cons_arit_0<0b101, "uqadd", uaddsat>;
+  defm SQSUB_ZZZ : sve_int_bin_cons_arit_0<0b110, "sqsub", ssubsat>;
+  defm UQSUB_ZZZ : sve_int_bin_cons_arit_0<0b111, "uqsub", usubsat>;
 
   defm AND_ZZZ : sve_int_bin_cons_log<0b00, "and", and>;
   defm ORR_ZZZ : sve_int_bin_cons_log<0b01, "orr", or>;
   defm EOR_ZZZ : sve_int_bin_cons_log<0b10, "eor", xor>;
   defm BIC_ZZZ : sve_int_bin_cons_log<0b11, "bic", null_frag>;
 
   defm ADD_ZPmZ  : sve_int_bin_pred_arit_0<0b000, "add",  "ADD_ZPZZ", int_aarch64_sve_add, DestructiveBinaryComm>;
   defm SUB_ZPmZ  : sve_int_bin_pred_arit_0<0b001, "sub",  "SUB_ZPZZ", int_aarch64_sve_sub, DestructiveBinaryCommWithRev, "SUBR_ZPmZ">;
   defm SUBR_ZPmZ : sve_int_bin_pred_arit_0<0b011, "subr", "SUBR_ZPZZ", int_aarch64_sve_subr, DestructiveBinaryCommWithRev, "SUB_ZPmZ", /*isReverseInstr*/ 1>;
 
   defm ADD_ZPZZ  : sve_int_bin_pred_bhsd<AArch64add_p>;
   defm SUB_ZPZZ  : sve_int_bin_pred_bhsd<AArch64sub_p>;
 
   let Predicates = [HasSVE, UseExperimentalZeroingPseudos] in {
     defm ADD_ZPZZ  : sve_int_bin_pred_zeroing_bhsd<int_aarch64_sve_add>;
     defm SUB_ZPZZ  : sve_int_bin_pred_zeroing_bhsd<int_aarch64_sve_sub>;
     defm SUBR_ZPZZ : sve_int_bin_pred_zeroing_bhsd<int_aarch64_sve_subr>;
   }
 
   defm ORR_ZPmZ : sve_int_bin_pred_log<0b000, "orr", int_aarch64_sve_orr>;
   defm EOR_ZPmZ : sve_int_bin_pred_log<0b001, "eor", int_aarch64_sve_eor>;
   defm AND_ZPmZ : sve_int_bin_pred_log<0b010, "and", int_aarch64_sve_and>;
   defm BIC_ZPmZ : sve_int_bin_pred_log<0b011, "bic", int_aarch64_sve_bic>;
 
-  defm ADD_ZI   : sve_int_arith_imm0<0b000, "add", add, null_frag>;
-  defm SUB_ZI   : sve_int_arith_imm0<0b001, "sub", sub, null_frag>;
+  defm ADD_ZI   : sve_int_arith_imm0<0b000, "add", add>;
+  defm SUB_ZI   : sve_int_arith_imm0<0b001, "sub", sub>;
   defm SUBR_ZI  : sve_int_arith_imm0_subr<0b011, "subr", sub>;
-  defm SQADD_ZI : sve_int_arith_imm0<0b100, "sqadd", saddsat, int_aarch64_sve_sqadd_x>;
-  defm UQADD_ZI : sve_int_arith_imm0<0b101, "uqadd", uaddsat, int_aarch64_sve_uqadd_x>;
-  defm SQSUB_ZI : sve_int_arith_imm0<0b110, "sqsub", ssubsat, int_aarch64_sve_sqsub_x>;
-  defm UQSUB_ZI : sve_int_arith_imm0<0b111, "uqsub", usubsat, int_aarch64_sve_uqsub_x>;
+  defm SQADD_ZI : sve_int_arith_imm0<0b100, "sqadd", saddsat>;
+  defm UQADD_ZI : sve_int_arith_imm0<0b101, "uqadd", uaddsat>;
+  defm SQSUB_ZI : sve_int_arith_imm0<0b110, "sqsub", ssubsat>;
+  defm UQSUB_ZI : sve_int_arith_imm0<0b111, "uqsub", usubsat>;
 
   defm MAD_ZPmZZ : sve_int_mladdsub_vvv_pred<0b0, "mad", int_aarch64_sve_mad>;
   defm MSB_ZPmZZ : sve_int_mladdsub_vvv_pred<0b1, "msb", int_aarch64_sve_msb>;
   defm MLA_ZPmZZ : sve_int_mlas_vvv_pred<0b0, "mla", int_aarch64_sve_mla, add, AArch64mul_p_oneuse>;
   defm MLS_ZPmZZ : sve_int_mlas_vvv_pred<0b1, "mls", int_aarch64_sve_mls, sub, AArch64mul_p_oneuse>;
 
   // SVE predicated integer reductions.
   defm SADDV_VPZ : sve_int_reduce_0_saddv<0b000, "saddv", AArch64saddv_p>;

llvm/lib/Target/AArch64/SVEInstrFormats.td

: I<(outs zprty:$Zd), (ins zprty:$Zn, zprty:$Zm),
   let Inst{21}    = 0b1;
   let Inst{20-16} = Zm;
   let Inst{15-13} = 0b000;
   let Inst{12-10} = opc;
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zd;
 }
 
-multiclass sve_int_bin_cons_arit_0<bits<3> opc, string asm,
-                                   SDPatternOperator op, SDPatternOperator int_op> {
+multiclass sve_int_bin_cons_arit_0<bits<3> opc, string asm, SDPatternOperator op> {
   def _B : sve_int_bin_cons_arit_0<0b00, opc, asm, ZPR8>;
   def _H : sve_int_bin_cons_arit_0<0b01, opc, asm, ZPR16>;
   def _S : sve_int_bin_cons_arit_0<0b10, opc, asm, ZPR32>;
   def _D : sve_int_bin_cons_arit_0<0b11, opc, asm, ZPR64>;
 
   def : SVE_2_Op_Pat<nxv16i8, op, nxv16i8, nxv16i8, !cast<Instruction>(NAME # _B)>;
   def : SVE_2_Op_Pat<nxv8i16, op, nxv8i16, nxv8i16, !cast<Instruction>(NAME # _H)>;
   def : SVE_2_Op_Pat<nxv4i32, op, nxv4i32, nxv4i32, !cast<Instruction>(NAME # _S)>;
   def : SVE_2_Op_Pat<nxv2i64, op, nxv2i64, nxv2i64, !cast<Instruction>(NAME # _D)>;
-
-  // Intrinsic version
-  def : SVE_2_Op_Pat<nxv16i8, int_op, nxv16i8, nxv16i8, !cast<Instruction>(NAME # _B)>;
-  def : SVE_2_Op_Pat<nxv8i16, int_op, nxv8i16, nxv8i16, !cast<Instruction>(NAME # _H)>;
-  def : SVE_2_Op_Pat<nxv4i32, int_op, nxv4i32, nxv4i32, !cast<Instruction>(NAME # _S)>;
-  def : SVE_2_Op_Pat<nxv2i64, int_op, nxv2i64, nxv2i64, !cast<Instruction>(NAME # _D)>;
 }
 
 //===----------------------------------------------------------------------===//
 // SVE Floating Point Arithmetic - Predicated Group
 //===----------------------------------------------------------------------===//
 
 class sve_fp_2op_i_p_zds<bits<2> sz, bits<3> opc, string asm,
                          ZPRRegOp zprty,
: I<(outs zprty:$Zdn), (ins zprty:$_Zdn, immtype:$imm),
   let Inst{12-5}  = imm{7-0}; // imm8
   let Inst{4-0}   = Zdn;
 
   let Constraints = "$Zdn = $_Zdn";
   let DestructiveInstType = DestructiveOther;
   let ElementSize = ElementSizeNone;
 }
 
-multiclass sve_int_arith_imm0<bits<3> opc, string asm,
-                              SDPatternOperator op, SDPatternOperator int_op> {
+multiclass sve_int_arith_imm0<bits<3> opc, string asm, SDPatternOperator op> {
   def _B : sve_int_arith_imm0<0b00, opc, asm, ZPR8,  addsub_imm8_opt_lsl_i8>;
   def _H : sve_int_arith_imm0<0b01, opc, asm, ZPR16, addsub_imm8_opt_lsl_i16>;
   def _S : sve_int_arith_imm0<0b10, opc, asm, ZPR32, addsub_imm8_opt_lsl_i32>;
   def _D : sve_int_arith_imm0<0b11, opc, asm, ZPR64, addsub_imm8_opt_lsl_i64>;
 
   def : SVE_1_Op_Imm_OptLsl_Pat<nxv16i8, op, ZPR8,  i32, SVEAddSubImm8Pat,  !cast<Instruction>(NAME # _B)>;
   def : SVE_1_Op_Imm_OptLsl_Pat<nxv8i16, op, ZPR16, i32, SVEAddSubImm16Pat, !cast<Instruction>(NAME # _H)>;
   def : SVE_1_Op_Imm_OptLsl_Pat<nxv4i32, op, ZPR32, i32, SVEAddSubImm32Pat, !cast<Instruction>(NAME # _S)>;
   def : SVE_1_Op_Imm_OptLsl_Pat<nxv2i64, op, ZPR64, i64, SVEAddSubImm64Pat, !cast<Instruction>(NAME # _D)>;
-
-  // Intrinsic version
-  def : SVE_1_Op_Imm_OptLsl_Pat<nxv16i8, int_op, ZPR8,  i32, SVEAddSubImm8Pat,  !cast<Instruction>(NAME # _B)>;
-  def : SVE_1_Op_Imm_OptLsl_Pat<nxv8i16, int_op, ZPR16, i32, SVEAddSubImm16Pat, !cast<Instruction>(NAME # _H)>;
-  def : SVE_1_Op_Imm_OptLsl_Pat<nxv4i32, int_op, ZPR32, i32, SVEAddSubImm32Pat, !cast<Instruction>(NAME # _S)>;
-  def : SVE_1_Op_Imm_OptLsl_Pat<nxv2i64, int_op, ZPR64, i64, SVEAddSubImm64Pat, !cast<Instruction>(NAME # _D)>;
 }
 
 multiclass sve_int_arith_imm0_subr<bits<3> opc, string asm, SDPatternOperator op> {
   def _B : sve_int_arith_imm0<0b00, opc, asm, ZPR8,  addsub_imm8_opt_lsl_i8>;
   def _H : sve_int_arith_imm0<0b01, opc, asm, ZPR16, addsub_imm8_opt_lsl_i16>;
   def _S : sve_int_arith_imm0<0b10, opc, asm, ZPR32, addsub_imm8_opt_lsl_i32>;
   def _D : sve_int_arith_imm0<0b11, opc, asm, ZPR64, addsub_imm8_opt_lsl_i64>;
 

llvm/test/CodeGen/AArch64/sve-intrinsics-int-arith-imm.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
 
+; ADD
+
+define <vscale x 16 x i8> @add_i8(<vscale x 16 x i8> %a) {
+; CHECK-LABEL: add_i8:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    add z0.b, z0.b, #127
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %elt = insertelement <vscale x 16 x i8> undef, i8 127, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %elt, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.add.nxv16i8(<vscale x 16 x i1> %pg,
+                                                               <vscale x 16 x i8> %a,
+                                                               <vscale x 16 x i8> %splat)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @add_i16(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: add_i16:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    add z0.h, z0.h, #127
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 127, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.add.nxv8i16(<vscale x 8 x i1> %pg,
+                                                               <vscale x 8 x i16> %a,
+                                                               <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 8 x i16> @add_i16_out_of_range(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: add_i16_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #257
+; CHECK-NEXT:    mov z1.h, w8
+; CHECK-NEXT:    add z0.h, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 257, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.add.nxv8i16(<vscale x 8 x i1> %pg,
+                                                               <vscale x 8 x i16> %a,
+                                                               <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @add_i32(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: add_i32:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    add z0.s, z0.s, #127
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 127, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.add.nxv4i32(<vscale x 4 x i1> %pg,
+                                                               <vscale x 4 x i32> %a,
+                                                               <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 4 x i32> @add_i32_out_of_range(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: add_i32_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #257
+; CHECK-NEXT:    mov z1.s, w8
+; CHECK-NEXT:    add z0.s, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 257, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.add.nxv4i32(<vscale x 4 x i1> %pg,
+                                                               <vscale x 4 x i32> %a,
+                                                               <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @add_i64(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: add_i64:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    add z0.d, z0.d, #127
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 127, i64 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.add.nxv2i64(<vscale x 2 x i1> %pg,
+                                                               <vscale x 2 x i64> %a,
+                                                               <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+define <vscale x 2 x i64> @add_i64_out_of_range(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: add_i64_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #257
+; CHECK-NEXT:    mov z1.d, x8
+; CHECK-NEXT:    add z0.d, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 257, i64 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.add.nxv2i64(<vscale x 2 x i1> %pg,
+                                                               <vscale x 2 x i64> %a,
+                                                               <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+; SUB
+
+define <vscale x 16 x i8> @sub_i8(<vscale x 16 x i8> %a) {
+; CHECK-LABEL: sub_i8:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sub z0.b, z0.b, #127
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %elt = insertelement <vscale x 16 x i8> undef, i8 127, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %elt, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sub.nxv16i8(<vscale x 16 x i1> %pg,
+                                                               <vscale x 16 x i8> %a,
+                                                               <vscale x 16 x i8> %splat)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @sub_i16(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: sub_i16:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sub z0.h, z0.h, #127
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 127, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sub.nxv8i16(<vscale x 8 x i1> %pg,
+                                                               <vscale x 8 x i16> %a,
+                                                               <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 8 x i16> @sub_i16_out_of_range(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: sub_i16_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #257
+; CHECK-NEXT:    mov z1.h, w8
+; CHECK-NEXT:    sub z0.h, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 257, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sub.nxv8i16(<vscale x 8 x i1> %pg,
+                                                               <vscale x 8 x i16> %a,
+                                                               <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sub_i32(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: sub_i32:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sub z0.s, z0.s, #127
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 127, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sub.nxv4i32(<vscale x 4 x i1> %pg,
+                                                               <vscale x 4 x i32> %a,
+                                                               <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 4 x i32> @sub_i32_out_of_range(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: sub_i32_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #257
+; CHECK-NEXT:    mov z1.s, w8
+; CHECK-NEXT:    sub z0.s, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 257, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sub.nxv4i32(<vscale x 4 x i1> %pg,
+                                                               <vscale x 4 x i32> %a,
+                                                               <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sub_i64(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: sub_i64:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sub z0.d, z0.d, #127
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 127, i64 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sub.nxv2i64(<vscale x 2 x i1> %pg,
+                                                               <vscale x 2 x i64> %a,
+                                                               <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+define <vscale x 2 x i64> @sub_i64_out_of_range(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: sub_i64_out_of_range:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #257
+; CHECK-NEXT:    mov z1.d, x8
+; CHECK-NEXT:    sub z0.d, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 257, i64 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sub.nxv2i64(<vscale x 2 x i1> %pg,
+                                                               <vscale x 2 x i64> %a,
+                                                               <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+; As sub_i32 but where pg is i8 based and thus compatible for i32.
+define <vscale x 4 x i32> @sub_i32_ptrue_all_b(<vscale x 4 x i32> %a) #0 {
+; CHECK-LABEL: sub_i32_ptrue_all_b:
+; CHECK: sub z0.s, z0.s, #1
+; CHECK-NEXT: ret
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %b = tail call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 1)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.sub.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+; As sub_i32 but where pg is i16 based and thus compatible for i32.
+define <vscale x 4 x i32> @sub_i32_ptrue_all_h(<vscale x 4 x i32> %a) #0 {
+; CHECK-LABEL: sub_i32_ptrue_all_h:
+; CHECK: sub z0.s, z0.s, #1
+; CHECK-NEXT: ret
+  %pg.h = tail call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> %pg.h)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %b = tail call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 1)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.sub.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+; As sub_i32 but where pg is i64 based, which is not compatibile for i32 and
+; thus inactive lanes are important and the immediate form cannot be used.
+define <vscale x 4 x i32> @sub_i32_ptrue_all_d(<vscale x 4 x i32> %a) #0 {
+; CHECK-LABEL: sub_i32_ptrue_all_d:
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].d
+; CHECK-DAG: mov [[DUP:z[0-9]+]].s, #1
+; CHECK-DAG: sub z0.s, [[PG]]/m, z0.s, [[DUP]].s
+; CHECK-NEXT: ret
+  %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> %pg.d)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %b = tail call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 1)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.sub.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
 ; SMAX
 
 define <vscale x 16 x i8> @smax_i8(<vscale x 16 x i8> %a) {
 ; CHECK-LABEL: smax_i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    smax z0.b, z0.b, #-128
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
; CHECK-NEXT: ret
   %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
   %b = tail call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 1)
   %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.mul.nxv4i32(<vscale x 4 x i1> %pg.s,
                                                                      <vscale x 4 x i32> %a,
                                                                      <vscale x 4 x i32> %b)
   ret <vscale x 4 x i32> %out
 }
 
+declare <vscale x 16 x i8> @llvm.aarch64.sve.add.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.add.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.add.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.add.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.sub.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.sub.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.sub.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.sub.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.sqadd.x.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
 declare <vscale x 8 x i16> @llvm.aarch64.sve.sqadd.x.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>)
 declare <vscale x 4 x i32> @llvm.aarch64.sve.sqadd.x.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
 declare <vscale x 2 x i64> @llvm.aarch64.sve.sqadd.x.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>)
 
 declare <vscale x 16 x i8> @llvm.aarch64.sve.sqsub.x.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
 declare <vscale x 8 x i16> @llvm.aarch64.sve.sqsub.x.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>)
 declare <vscale x 4 x i32> @llvm.aarch64.sve.sqsub.x.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)

llvm/test/CodeGen/AArch64/sve-intrinsics-logical-imm.ll

+; RUN: llc -mtriple=aarch64-linux-gnu < %s | FileCheck %s
+
+;
+; AND
+;
+
+define <vscale x 16 x i8> @and_i8(<vscale x 16 x i8> %a) #0 {
+; CHECK-LABEL: and_i8:
+; CHECK: and z0.b, z0.b, #0x7
+; CHECK-NEXT: ret
+  %pg = shufflevector <vscale x 16 x i1> insertelement (<vscale x 16 x i1> undef, i1 true, i32 0), <vscale x 16 x i1> undef, <vscale x 16 x i32> zeroinitializer
+  %b = shufflevector <vscale x 16 x i8> insertelement (<vscale x 16 x i8> undef, i8 7, i32 0), <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.and.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> @and_i16(<vscale x 8 x i16> %a) #0 {
+; CHECK-LABEL: and_i16:
+; CHECK: and z0.h, z0.h, #0xf0
+; CHECK-NEXT: ret
+  %pg = shufflevector <vscale x 8 x i1> insertelement (<vscale x 8 x i1> undef, i1 true, i32 0), <vscale x 8 x i1> undef, <vscale x 8 x i32> zeroinitializer
+  %b = shufflevector <vscale x 8 x i16> insertelement (<vscale x 8 x i16> undef, i16 240, i32 0), <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.and.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> @and_i32(<vscale x 4 x i32> %a) #0 {
+; CHECK-LABEL: and_i32:
+; CHECK: and z0.s, z0.s, #0xffff00
+; CHECK-NEXT: ret
+  %pg = shufflevector <vscale x 4 x i1> insertelement (<vscale x 4 x i1> undef, i1 true, i32 0), <vscale x 4 x i1> undef, <vscale x 4 x i32> zeroinitializer
+  %b = shufflevector <vscale x 4 x i32> insertelement (<vscale x 4 x i32> undef, i32 16776960, i32 0), <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.and.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> @and_i64(<vscale x 2 x i64> %a) #0 {
+; CHECK-LABEL: and_i64:
+; CHECK: and z0.d, z0.d, #0xfffc000000000000
+; CHECK-NEXT: ret
+  %pg = shufflevector <vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer
+  %b = shufflevector <vscale x 2 x i64> insertelement (<vscale x 2 x i64> undef, i64 18445618173802708992, i32 0), <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.and.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
+}
+
+;
+; EOR
+;
+
+define <vscale x 16 x i8> @eor_i8(<vscale x 16 x i8> %a) #0 {
+; CHECK-LABEL: eor_i8:
+; CHECK: eor z0.b, z0.b, #0xf
+; CHECK-NEXT: ret
+  %pg = shufflevector <vscale x 16 x i1> insertelement (<vscale x 16 x i1> undef, i1 true, i32 0), <vscale x 16 x i1> undef, <vscale x 16 x i32> zeroinitializer
+  %b = shufflevector <vscale x 16 x i8> insertelement (<vscale x 16 x i8> undef, i8 15, i32 0), <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.eor.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> @eor_i16(<vscale x 8 x i16> %a) #0 {
+; CHECK-LABEL: eor_i16:
+; CHECK: eor z0.h, z0.h, #0xfc07
+; CHECK-NEXT: ret
+  %pg = shufflevector <vscale x 8 x i1> insertelement (<vscale x 8 x i1> undef, i1 true, i32 0), <vscale x 8 x i1> undef, <vscale x 8 x i32> zeroinitializer
+  %b = shufflevector <vscale x 8 x i16> insertelement (<vscale x 8 x i16> undef, i16 64519, i32 0), <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.eor.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> @eor_i32(<vscale x 4 x i32> %a) #0 {
+; CHECK-LABEL: eor_i32:
+; CHECK: eor z0.s, z0.s, #0xffff00
+; CHECK-NEXT: ret
+  %pg = shufflevector <vscale x 4 x i1> insertelement (<vscale x 4 x i1> undef, i1 true, i32 0), <vscale x 4 x i1> undef, <vscale x 4 x i32> zeroinitializer
+  %b = shufflevector <vscale x 4 x i32> insertelement (<vscale x 4 x i32> undef, i32 16776960, i32 0), <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.eor.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> @eor_i64(<vscale x 2 x i64> %a) #0 {
+; CHECK-LABEL: eor_i64:
+; CHECK: eor z0.d, z0.d, #0x1000000000000
+; CHECK-NEXT: ret
+  %pg = shufflevector <vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer
+  %b = shufflevector <vscale x 2 x i64> insertelement (<vscale x 2 x i64> undef, i64 281474976710656, i32 0), <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.eor.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
+}
+
+;
+; ORR
+;
+
+define <vscale x 16 x i8> @orr_i8(<vscale x 16 x i8> %a) #0 {
+; CHECK-LABEL: orr_i8:
+; CHECK: orr z0.b, z0.b, #0x6
+; CHECK-NEXT: ret
+  %pg = shufflevector <vscale x 16 x i1> insertelement (<vscale x 16 x i1> undef, i1 true, i32 0), <vscale x 16 x i1> undef, <vscale x 16 x i32> zeroinitializer
+  %b = shufflevector <vscale x 16 x i8> insertelement (<vscale x 16 x i8> undef, i8 6, i32 0), <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.orr.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> @orr_i16(<vscale x 8 x i16> %a) #0 {
+; CHECK-LABEL: orr_i16:
+; CHECK: orr z0.h, z0.h, #0x8001
+; CHECK-NEXT: ret
+  %pg = shufflevector <vscale x 8 x i1> insertelement (<vscale x 8 x i1> undef, i1 true, i32 0), <vscale x 8 x i1> undef, <vscale x 8 x i32> zeroinitializer
+  %b = shufflevector <vscale x 8 x i16> insertelement (<vscale x 8 x i16> undef, i16 32769, i32 0), <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.orr.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> @orr_i32(<vscale x 4 x i32> %a) #0 {
+; CHECK-LABEL: orr_i32:
+; CHECK: orr z0.s, z0.s, #0xffff
+; CHECK-NEXT: ret
+  %pg = shufflevector <vscale x 4 x i1> insertelement (<vscale x 4 x i1> undef, i1 true, i32 0), <vscale x 4 x i1> undef, <vscale x 4 x i32> zeroinitializer
+  %b = shufflevector <vscale x 4 x i32> insertelement (<vscale x 4 x i32> undef, i32 65535, i32 0), <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.orr.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> @orr_i64(<vscale x 2 x i64> %a) #0 {
+; CHECK-LABEL: orr_i64:
+; CHECK: orr z0.d, z0.d, #0x7ffc000000000000
+; CHECK-NEXT: ret
+  %pg = shufflevector <vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer
+  %b = shufflevector <vscale x 2 x i64> insertelement (<vscale x 2 x i64> undef, i64 9222246136947933184, i32 0), <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.orr.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
+}
+
+; As orr_i32 but where pg is i8 based and thus compatible for i32.
+define <vscale x 4 x i32> @orr_i32_ptrue_all_b(<vscale x 4 x i32> %a) #0 {
+; CHECK-LABEL: orr_i32_ptrue_all_b:
+; CHECK: orr z0.s, z0.s, #0xffff
+; CHECK-NEXT: ret
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %b = tail call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 65535)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.orr.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+; As orr_i32 but where pg is i16 based and thus compatible for i32.
+define <vscale x 4 x i32> @orr_i32_ptrue_all_h(<vscale x 4 x i32> %a) #0 {
+; CHECK-LABEL: orr_i32_ptrue_all_h:
+; CHECK: orr z0.s, z0.s, #0xffff
+; CHECK-NEXT: ret
+  %pg.h = tail call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> %pg.h)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %b = tail call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 65535)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.orr.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+; As orr_i32 but where pg is i64 based, which is not compatibile for i32 and
+; thus inactive lanes are important and the immediate form cannot be used.
+define <vscale x 4 x i32> @orr_i32_ptrue_all_d(<vscale x 4 x i32> %a) #0 {
+; CHECK-LABEL: orr_i32_ptrue_all_d:
+; CHECK-DAG: mov [[IMM:w[0-9]+]], #65535
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].d
+; CHECK-DAG: mov [[DUP:z[0-9]+]].s, [[IMM]]
+; CHECK-DAG: orr z0.s, [[PG]]/m, z0.s, [[DUP]].s
+; CHECK-NEXT: ret
+  %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> %pg.d)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %b = tail call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 65535)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.orr.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.and.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.and.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.and.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.and.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.eor.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.eor.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.eor.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.eor.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.orr.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.orr.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.orr.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.orr.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1>)
+declare <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1>)
+declare <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1>)
+
+declare <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 16 x i1>)
+declare <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1>)
+declare <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1>)
+
+declare <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32)
+
+declare <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32)
+declare <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32)
+declare <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32)
+declare <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32)
+
+attributes #0 = { "target-features"="+sve" }

llvm/test/CodeGen/AArch64/sve-intrinsics-unpred-form.ll

 ; RUN: llc < %s | FileCheck %s
 
 target triple = "aarch64-unknown-linux-gnu"
 
 ;
+; ADD
+;
+
+define <vscale x 16 x i8> @add_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) #0 {
+; CHECK-LABEL: add_i8:
+; CHECK: add z0.b, z0.b, z1.b
+; CHECK-NEXT: ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.add.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> @add_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) #0 {
+; CHECK-LABEL: add_i16:
+; CHECK: add z0.h, z0.h, z1.h
+; CHECK-NEXT: ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.add.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> @add_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: add_i32:
+; CHECK: add z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.add.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> @add_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) #0 {
+; CHECK-LABEL: add_i64:
+; CHECK: add z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.add.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
+}
+
+;
+; SUB
+;
+
+define <vscale x 16 x i8> @sub_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) #0 {
+; CHECK-LABEL: sub_i8:
+; CHECK: sub z0.b, z0.b, z1.b
+; CHECK-NEXT: ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sub.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> @sub_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) #0 {
+; CHECK-LABEL: sub_i16:
+; CHECK: sub z0.h, z0.h, z1.h
+; CHECK-NEXT: ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sub.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> @sub_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: sub_i32:
+; CHECK: sub z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sub.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> @sub_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) #0 {
+; CHECK-LABEL: sub_i64:
+; CHECK: sub z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sub.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
+}
+
+; As sub_i32 but where pg is i8 based and thus compatible for i32.
+define <vscale x 4 x i32> @sub_i32_ptrue_all_b(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: sub_i32_ptrue_all_b:
+; CHECK: sub z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.sub.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+; As sub_i32 but where pg is i16 based and thus compatible for i32.
+define <vscale x 4 x i32> @sub_i32_ptrue_all_h(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: sub_i32_ptrue_all_h:
+; CHECK: sub z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg.h = tail call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> %pg.h)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.sub.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+; As sub_i32 but where pg is i64 based, which is not compatibile for i32 and
+; thus inactive lanes are important and the immediate form cannot be used.
+define <vscale x 4 x i32> @sub_i32_ptrue_all_d(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: sub_i32_ptrue_all_d:
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].d
+; CHECK-DAG: sub z0.s, [[PG]]/m, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> %pg.d)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.sub.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+;
 ; MUL
 ;
 
 define <vscale x 16 x i8> @mul_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) #0 {
 ; CHECK-LABEL: mul_i8:
 ; CHECK: mul z0.b, z0.b, z1.b
 ; CHECK-NEXT: ret
   %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
; CHECK-NEXT: ret
   %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
   %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.umulh.nxv4i32(<vscale x 4 x i1> %pg.s,
                                                                       <vscale x 4 x i32> %a,
                                                                       <vscale x 4 x i32> %b)
   ret <vscale x 4 x i32> %out
 }
 
 ;
+; AND
+;
+
+define <vscale x 16 x i8> @and_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) #0 {
+; CHECK-LABEL: and_i8:
+; CHECK: and z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.and.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> @and_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) #0 {
+; CHECK-LABEL: and_i16:
+; CHECK: and z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.and.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> @and_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: and_i32:
+; CHECK: and z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.and.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> @and_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) #0 {
+; CHECK-LABEL: and_i64:
+; CHECK: and z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.and.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
+}
+
+;
+; EOR
+;
+
+define <vscale x 16 x i8> @eor_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) #0 {
+; CHECK-LABEL: eor_i8:
+; CHECK: eor z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.eor.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> @eor_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) #0 {
+; CHECK-LABEL: eor_i16:
+; CHECK: eor z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.eor.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> @eor_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: eor_i32:
+; CHECK: eor z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.eor.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> @eor_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) #0 {
+; CHECK-LABEL: eor_i64:
+; CHECK: eor z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.eor.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
+}
+
+;
+; ORR
+;
+
+define <vscale x 16 x i8> @orr_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) #0 {
+; CHECK-LABEL: orr_i8:
+; CHECK: orr z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.orr.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> @orr_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) #0 {
+; CHECK-LABEL: orr_i16:
+; CHECK: orr z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.orr.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> @orr_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: orr_i32:
+; CHECK: orr z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.orr.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> @orr_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) #0 {
+; CHECK-LABEL: orr_i64:
+; CHECK: orr z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.orr.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
+}
+
+; As orr_i32 but where pg is i8 based and thus compatible for i32.
+define <vscale x 4 x i32> @orr_i32_ptrue_all_b(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: orr_i32_ptrue_all_b:
+; CHECK: orr z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.orr.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+; As orr_i32 but where pg is i16 based and thus compatible for i32.
+define <vscale x 4 x i32> @orr_i32_ptrue_all_h(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: orr_i32_ptrue_all_h:
+; CHECK: orr z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg.h = tail call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> %pg.h)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.orr.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+; As orr_i32 but where pg is i64 based, which is not compatibile for i32 and
+; thus inactive lanes are important and the immediate form cannot be used.
+define <vscale x 4 x i32> @orr_i32_ptrue_all_d(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: orr_i32_ptrue_all_d:
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].d
+; CHECK-DAG: orr z0.s, [[PG]]/m, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> %pg.d)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.orr.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+;
+; SQADD
+;
+
+define <vscale x 16 x i8> @sqadd_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) #0 {
+; CHECK-LABEL: sqadd_i8:
+; CHECK: sqadd z0.b, z0.b, z1.b
+; CHECK-NEXT: ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sqadd.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> @sqadd_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) #0 {
+; CHECK-LABEL: sqadd_i16:
+; CHECK: sqadd z0.h, z0.h, z1.h
+; CHECK-NEXT: ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqadd.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> @sqadd_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: sqadd_i32:
+; CHECK: sqadd z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqadd.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> @sqadd_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) #0 {
+; CHECK-LABEL: sqadd_i64:
+; CHECK: sqadd z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqadd.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
+}
+
+;
+; SQSUB
+;
+
+define <vscale x 16 x i8> @sqsub_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) #0 {
+; CHECK-LABEL: sqsub_i8:
+; CHECK: sqsub z0.b, z0.b, z1.b
+; CHECK-NEXT: ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sqsub.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> @sqsub_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) #0 {
+; CHECK-LABEL: sqsub_i16:
+; CHECK: sqsub z0.h, z0.h, z1.h
+; CHECK-NEXT: ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqsub.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> @sqsub_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: sqsub_i32:
+; CHECK: sqsub z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqsub.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> @sqsub_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) #0 {
+; CHECK-LABEL: sqsub_i64:
+; CHECK: sqsub z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqsub.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
+}
+
+;
+; UQADD
+;
+
+define <vscale x 16 x i8> @uqadd_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) #0 {
+; CHECK-LABEL: uqadd_i8:
+; CHECK: uqadd z0.b, z0.b, z1.b
+; CHECK-NEXT: ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.uqadd.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> @uqadd_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) #0 {
+; CHECK-LABEL: uqadd_i16:
+; CHECK: uqadd z0.h, z0.h, z1.h
+; CHECK-NEXT: ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.uqadd.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> @uqadd_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: uqadd_i32:
+; CHECK: uqadd z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.uqadd.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> @uqadd_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) #0 {
+; CHECK-LABEL: uqadd_i64:
+; CHECK: uqadd z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.uqadd.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
+}
+
+;
+; UQSUB
+;
+
+define <vscale x 16 x i8> @uqsub_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) #0 {
+; CHECK-LABEL: uqsub_i8:
+; CHECK: uqsub z0.b, z0.b, z1.b
+; CHECK-NEXT: ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.uqsub.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> @uqsub_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) #0 {
+; CHECK-LABEL: uqsub_i16:
+; CHECK: uqsub z0.h, z0.h, z1.h
+; CHECK-NEXT: ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.uqsub.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> @uqsub_i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: uqsub_i32:
+; CHECK: uqsub z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.uqsub.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> @uqsub_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b) #0 {
+; CHECK-LABEL: uqsub_i64:
+; CHECK: uqsub z0.d, z0.d, z1.d
+; CHECK-NEXT: ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.uqsub.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
+}
+
+; As uqsub_i32 but where pg is i8 based and thus compatible for i32.
+define <vscale x 4 x i32> @uqsub_i32_ptrue_all_b(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: uqsub_i32_ptrue_all_b:
+; CHECK: uqsub z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.uqsub.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                      <vscale x 4 x i32> %a,
+                                                                      <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+; As uqsub_i32 but where pg is i16 based and thus compatible for i32.
+define <vscale x 4 x i32> @uqsub_i32_ptrue_all_h(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: uqsub_i32_ptrue_all_h:
+; CHECK: uqsub z0.s, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg.h = tail call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> %pg.h)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.uqsub.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                      <vscale x 4 x i32> %a,
+                                                                      <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+; As uqsub_i32 but where pg is i64 based, which is not compatibile for i32 and
+; thus inactive lanes are important and the immediate form cannot be used.
+define <vscale x 4 x i32> @uqsub_i32_ptrue_all_d(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) #0 {
+; CHECK-LABEL: uqsub_i32_ptrue_all_d:
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].d
+; CHECK-DAG: uqsub z0.s, [[PG]]/m, z0.s, z1.s
+; CHECK-NEXT: ret
+  %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> %pg.d)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.uqsub.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                      <vscale x 4 x i32> %a,
+                                                                      <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+;
 ; ASR (wide)
 ;
 
 define <vscale x 16 x i8> @asr_i8(<vscale x 16 x i8> %a, <vscale x 2 x i64> %b) #0 {
 ; CHECK-LABEL: asr_i8:
 ; CHECK: asr z0.b, z0.b, z1.d
 ; CHECK-NEXT: ret
   %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)

 ; CHECK-NEXT: ret
   %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
   %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
 }
 
+declare <vscale x 16 x  i8> @llvm.aarch64.sve.add.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.add.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.add.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.add.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.sub.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.sub.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.sub.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.sub.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.mul.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.mul.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.mul.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.mul.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.smulh.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.smulh.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.smulh.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.smulh.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.umulh.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.umulh.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.umulh.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.umulh.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.and.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.and.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.and.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.and.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.eor.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.eor.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.eor.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.eor.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.orr.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.orr.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.orr.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.orr.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.sqadd.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.sqadd.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.sqadd.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.sqadd.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.sqsub.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.sqsub.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.sqsub.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.sqsub.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.uqadd.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.uqadd.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.uqadd.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.uqadd.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.uqsub.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.uqsub.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.uqsub.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.uqsub.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.asr.wide.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x  i8>, <vscale x 2 x i64>)
 declare <vscale x  8 x i16> @llvm.aarch64.sve.asr.wide.nxv8i16(<vscale x  8 x i1>, <vscale x  8 x i16>, <vscale x 2 x i64>)
 declare <vscale x  4 x i32> @llvm.aarch64.sve.asr.wide.nxv4i32(<vscale x  4 x i1>, <vscale x  4 x i32>, <vscale x 2 x i64>)
 
 declare <vscale x 16 x  i8> @llvm.aarch64.sve.lsl.wide.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x  i8>, <vscale x 2 x i64>)
 declare <vscale x  8 x i16> @llvm.aarch64.sve.lsl.wide.nxv8i16(<vscale x  8 x i1>, <vscale x  8 x i16>, <vscale x 2 x i64>)
 declare <vscale x  4 x i32> @llvm.aarch64.sve.lsl.wide.nxv4i32(<vscale x  4 x i1>, <vscale x  4 x i32>, <vscale x 2 x i64>)
 

llvm/test/CodeGen/AArch64/sve2-intrinsics-int-arith-imm.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve2 < %s | FileCheck %s
+
+; SQADD
+
+define <vscale x 16 x i8> @sqadd_b_lowimm(<vscale x 16 x i8> %a) {
+; CHECK-LABEL: sqadd_b_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqadd z0.b, z0.b, #27 // =0x1b
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %elt = insertelement <vscale x 16 x i8> undef, i8 27, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %elt, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sqadd.nxv16i8(<vscale x 16 x i1> %pg,
+                                                                 <vscale x 16 x i8> %a,
+                                                                 <vscale x 16 x i8> %splat)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @sqadd_h_lowimm(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: sqadd_h_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqadd z0.h, z0.h, #43 // =0x2b
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 43, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqadd.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 8 x i16> @sqadd_h_highimm(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: sqadd_h_highimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqadd z0.h, z0.h, #2048 // =0x800
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 2048, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqadd.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sqadd_s_lowimm(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: sqadd_s_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqadd z0.s, z0.s, #1 // =0x1
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 1, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqadd.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 4 x i32> @sqadd_s_highimm(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: sqadd_s_highimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqadd z0.s, z0.s, #8192 // =0x2000
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 8192, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqadd.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sqadd_d_lowimm(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: sqadd_d_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqadd z0.d, z0.d, #255 // =0xff
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 255, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqadd.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+define <vscale x 2 x i64> @sqadd_d_highimm(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: sqadd_d_highimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqadd z0.d, z0.d, #65280 // =0xff00
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 65280, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqadd.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+; SQSUB
+
+define <vscale x 16 x i8> @sqsub_b_lowimm(<vscale x 16 x i8> %a) {
+; CHECK-LABEL: sqsub_b_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqsub z0.b, z0.b, #27 // =0x1b
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %elt = insertelement <vscale x 16 x i8> undef, i8 27, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %elt, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sqsub.nxv16i8(<vscale x 16 x i1> %pg,
+                                                                 <vscale x 16 x i8> %a,
+                                                                 <vscale x 16 x i8> %splat)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @sqsub_h_lowimm(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: sqsub_h_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqsub z0.h, z0.h, #43 // =0x2b
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 43, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqsub.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 8 x i16> @sqsub_h_highimm(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: sqsub_h_highimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqsub z0.h, z0.h, #2048 // =0x800
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 2048, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqsub.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @sqsub_s_lowimm(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: sqsub_s_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqsub z0.s, z0.s, #1 // =0x1
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 1, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqsub.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 4 x i32> @sqsub_s_highimm(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: sqsub_s_highimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqsub z0.s, z0.s, #8192 // =0x2000
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 8192, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqsub.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @sqsub_d_lowimm(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: sqsub_d_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqsub z0.d, z0.d, #255 // =0xff
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 255, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqsub.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+define <vscale x 2 x i64> @sqsub_d_highimm(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: sqsub_d_highimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqsub z0.d, z0.d, #65280 // =0xff00
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 65280, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqsub.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+; UQADD
+
+define <vscale x 16 x i8> @uqadd_b_lowimm(<vscale x 16 x i8> %a) {
+; CHECK-LABEL: uqadd_b_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqadd z0.b, z0.b, #27 // =0x1b
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %elt = insertelement <vscale x 16 x i8> undef, i8 27, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %elt, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.uqadd.nxv16i8(<vscale x 16 x i1> %pg,
+                                                                 <vscale x 16 x i8> %a,
+                                                                 <vscale x 16 x i8> %splat)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @uqadd_h_lowimm(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: uqadd_h_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqadd z0.h, z0.h, #43 // =0x2b
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 43, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.uqadd.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 8 x i16> @uqadd_h_highimm(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: uqadd_h_highimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqadd z0.h, z0.h, #2048 // =0x800
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 2048, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.uqadd.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @uqadd_s_lowimm(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: uqadd_s_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqadd z0.s, z0.s, #1 // =0x1
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 1, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.uqadd.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 4 x i32> @uqadd_s_highimm(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: uqadd_s_highimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqadd z0.s, z0.s, #8192 // =0x2000
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 8192, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.uqadd.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @uqadd_d_lowimm(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: uqadd_d_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqadd z0.d, z0.d, #255 // =0xff
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 255, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.uqadd.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+define <vscale x 2 x i64> @uqadd_d_highimm(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: uqadd_d_highimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqadd z0.d, z0.d, #65280 // =0xff00
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 65280, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.uqadd.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+; UQSUB
+
+define <vscale x 16 x i8> @uqsub_b_lowimm(<vscale x 16 x i8> %a) {
+; CHECK-LABEL: uqsub_b_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqsub z0.b, z0.b, #27 // =0x1b
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %elt = insertelement <vscale x 16 x i8> undef, i8 27, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %elt, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.uqsub.nxv16i8(<vscale x 16 x i1> %pg,
+                                                                 <vscale x 16 x i8> %a,
+                                                                 <vscale x 16 x i8> %splat)
+  ret <vscale x 16 x i8> %out
+}
+
+define <vscale x 8 x i16> @uqsub_h_lowimm(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: uqsub_h_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqsub z0.h, z0.h, #43 // =0x2b
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 43, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.uqsub.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 8 x i16> @uqsub_h_highimm(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: uqsub_h_highimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqsub z0.h, z0.h, #2048 // =0x800
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %elt = insertelement <vscale x 8 x i16> undef, i16 2048, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.uqsub.nxv8i16(<vscale x 8 x i1> %pg,
+                                                                 <vscale x 8 x i16> %a,
+                                                                 <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
+define <vscale x 4 x i32> @uqsub_s_lowimm(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: uqsub_s_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqsub z0.s, z0.s, #1 // =0x1
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 1, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.uqsub.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 4 x i32> @uqsub_s_highimm(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: uqsub_s_highimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqsub z0.s, z0.s, #8192 // =0x2000
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
+  %elt = insertelement <vscale x 4 x i32> undef, i32 8192, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.uqsub.nxv4i32(<vscale x 4 x i1> %pg,
+                                                                 <vscale x 4 x i32> %a,
+                                                                 <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
+define <vscale x 2 x i64> @uqsub_d_lowimm(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: uqsub_d_lowimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqsub z0.d, z0.d, #255 // =0xff
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 255, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.uqsub.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+define <vscale x 2 x i64> @uqsub_d_highimm(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: uqsub_d_highimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    uqsub z0.d, z0.d, #65280 // =0xff00
+; CHECK-NEXT:    ret
+  %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %elt = insertelement <vscale x 2 x i64> undef, i64 65280, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.uqsub.nxv2i64(<vscale x 2 x i1> %pg,
+                                                                 <vscale x 2 x i64> %a,
+                                                                 <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
+; As uqsub_i32 but where pg is i8 based and thus compatible for i32.
+define <vscale x 4 x i32> @uqsub_i32_ptrue_all_b(<vscale x 4 x i32> %a) #0 {
+; CHECK-LABEL: uqsub_i32_ptrue_all_b:
+; CHECK: uqsub z0.s, z0.s, #1
+; CHECK-NEXT: ret
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %b = tail call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 1)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.uqsub.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                      <vscale x 4 x i32> %a,
+                                                                      <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+; As uqsub_i32 but where pg is i16 based and thus compatible for i32.
+define <vscale x 4 x i32> @uqsub_i32_ptrue_all_h(<vscale x 4 x i32> %a) #0 {
+; CHECK-LABEL: uqsub_i32_ptrue_all_h:
+; CHECK: uqsub z0.s, z0.s, #1
+; CHECK-NEXT: ret
+  %pg.h = tail call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> %pg.h)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %b = tail call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 1)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.uqsub.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+; As uqsub_i32 but where pg is i64 based, which is not compatibile for i32 and
+; thus inactive lanes are important and the immediate form cannot be used.
+define <vscale x 4 x i32> @uqsub_i32_ptrue_all_d(<vscale x 4 x i32> %a) #0 {
+; CHECK-LABEL: uqsub_i32_ptrue_all_d:
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].d
+; CHECK-DAG: mov [[DUP:z[0-9]+]].s, #1
+; CHECK-DAG: uqsub z0.s, [[PG]]/m, z0.s, [[DUP]].s
+; CHECK-NEXT: ret
+  %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
+  %pg.b = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> %pg.d)
+  %pg.s = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg.b)
+  %b = tail call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 1)
+  %out = tail call <vscale x 4 x i32> @llvm.aarch64.sve.uqsub.nxv4i32(<vscale x 4 x i1> %pg.s,
+                                                                    <vscale x 4 x i32> %a,
+                                                                    <vscale x 4 x i32> %b)
+  ret <vscale x 4 x i32> %out
+}
+
+declare <vscale x 16 x i8> @llvm.aarch64.sve.sqadd.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.sqadd.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.sqadd.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.sqadd.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.sqsub.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.sqsub.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.sqsub.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.sqsub.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.uqadd.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.uqadd.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.uqadd.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.uqadd.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.uqsub.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.uqsub.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.uqsub.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.uqsub.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, <vscale x 2 x i64>)
+
+declare <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1>)
+declare <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1>)
+declare <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1>)
+
+declare <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 16 x i1>)
+declare <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1>)
+declare <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1>)
+
+declare <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32)
+
+declare <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 %pattern)
+declare <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 %pattern)
+declare <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 %pattern)
+declare <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 %pattern)