Implement AArch64 vector load/store multiple N-element structure class SIMD(lselem).

Including following 14 instructions:
4 ld1 insts: load multiple 1-element structure to sequential 1/2/3/4 registers.
ld2/ld3/ld4: load multiple N-element structure to sequential N registers (N=2,3,4).
4 st1 insts: store multiple 1-element structure from sequential 1/2/3/4 registers.
st2/st3/st4: store multiple N-element structure from sequential N registers (N = 2,3,4).


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@192361 91177308-0d34-0410-b5e6-96231b3b80d8

15 files changed, 2827 insertions, 3 deletions

diff --git a/include/llvm/CodeGen/ValueTypes.h b/include/llvm/CodeGen/ValueTypes.h
index 2e8f637f522..79f323341fd 100644
--- a/include/llvm/CodeGen/ValueTypes.h
+++ b/include/llvm/CodeGen/ValueTypes.h
@@ -208,7 +208,7 @@ namespace llvm {
     bool is64BitVector() const {
       return (SimpleTy == MVT::v8i8  || SimpleTy == MVT::v4i16 ||
               SimpleTy == MVT::v2i32 || SimpleTy == MVT::v1i64 ||
-              SimpleTy == MVT::v2f32);
+              SimpleTy == MVT::v1f64 || SimpleTy == MVT::v2f32);
     }
 
     /// is128BitVector - Return true if this is a 128-bit vector type.
diff --git a/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp b/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp
index a8655649f19..3b0dd64ee44 100644
--- a/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp
+++ b/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp
@@ -109,6 +109,23 @@ public:
 
   SDNode* Select(SDNode*);
 private:
+  /// Select NEON load intrinsics.  NumVecs should be 1, 2, 3 or 4.
+  SDNode *SelectVLD(SDNode *N, unsigned NumVecs, const uint16_t *Opcode);
+
+  /// Select NEON store intrinsics.  NumVecs should be 1, 2, 3 or 4.
+  SDNode *SelectVST(SDNode *N, unsigned NumVecs, const uint16_t *Opcodes);
+
+  // Form pairs of consecutive 64-bit/128-bit registers.
+  SDNode *createDPairNode(SDValue V0, SDValue V1);
+  SDNode *createQPairNode(SDValue V0, SDValue V1);
+
+  // Form sequences of 3 consecutive 64-bit/128-bit registers.
+  SDNode *createDTripleNode(SDValue V0, SDValue V1, SDValue V2);
+  SDNode *createQTripleNode(SDValue V0, SDValue V1, SDValue V2);
+
+  // Form sequences of 4 consecutive 64-bit/128-bit registers.
+  SDNode *createDQuadNode(SDValue V0, SDValue V1, SDValue V2, SDValue V3);
+  SDNode *createQQuadNode(SDValue V0, SDValue V1, SDValue V2, SDValue V3);
 };
 }
 
@@ -390,6 +407,221 @@ SDNode *AArch64DAGToDAGISel::SelectAtomic(SDNode *Node, unsigned Op8,
                               &Ops[0], Ops.size());
 }
 
+SDNode *AArch64DAGToDAGISel::createDPairNode(SDValue V0, SDValue V1) {
+  SDLoc dl(V0.getNode());
+  SDValue RegClass =
+      CurDAG->getTargetConstant(AArch64::DPairRegClassID, MVT::i32);
+  SDValue SubReg0 = CurDAG->getTargetConstant(AArch64::dsub_0, MVT::i32);
+  SDValue SubReg1 = CurDAG->getTargetConstant(AArch64::dsub_1, MVT::i32);
+  const SDValue Ops[] = { RegClass, V0, SubReg0, V1, SubReg1 };
+  return CurDAG->getMachineNode(TargetOpcode::REG_SEQUENCE, dl, MVT::v2i64,
+                                Ops);
+}
+
+SDNode *AArch64DAGToDAGISel::createQPairNode(SDValue V0, SDValue V1) {
+  SDLoc dl(V0.getNode());
+  SDValue RegClass =
+      CurDAG->getTargetConstant(AArch64::QPairRegClassID, MVT::i32);
+  SDValue SubReg0 = CurDAG->getTargetConstant(AArch64::qsub_0, MVT::i32);
+  SDValue SubReg1 = CurDAG->getTargetConstant(AArch64::qsub_1, MVT::i32);
+  const SDValue Ops[] = { RegClass, V0, SubReg0, V1, SubReg1 };
+  return CurDAG->getMachineNode(TargetOpcode::REG_SEQUENCE, dl, MVT::v4i64,
+                                Ops);
+}
+
+SDNode *AArch64DAGToDAGISel::createDTripleNode(SDValue V0, SDValue V1,
+                                               SDValue V2) {
+  SDLoc dl(V0.getNode());
+  SDValue RegClass =
+      CurDAG->getTargetConstant(AArch64::DTripleRegClassID, MVT::i32);
+  SDValue SubReg0 = CurDAG->getTargetConstant(AArch64::dsub_0, MVT::i32);
+  SDValue SubReg1 = CurDAG->getTargetConstant(AArch64::dsub_1, MVT::i32);
+  SDValue SubReg2 = CurDAG->getTargetConstant(AArch64::dsub_2, MVT::i32);
+  const SDValue Ops[] = { RegClass, V0, SubReg0, V1, SubReg1, V2, SubReg2 };
+  return CurDAG->getMachineNode(TargetOpcode::REG_SEQUENCE, dl, MVT::Untyped,
+                                Ops);
+}
+
+SDNode *AArch64DAGToDAGISel::createQTripleNode(SDValue V0, SDValue V1,
+                                               SDValue V2) {
+  SDLoc dl(V0.getNode());
+  SDValue RegClass =
+      CurDAG->getTargetConstant(AArch64::QTripleRegClassID, MVT::i32);
+  SDValue SubReg0 = CurDAG->getTargetConstant(AArch64::qsub_0, MVT::i32);
+  SDValue SubReg1 = CurDAG->getTargetConstant(AArch64::qsub_1, MVT::i32);
+  SDValue SubReg2 = CurDAG->getTargetConstant(AArch64::qsub_2, MVT::i32);
+  const SDValue Ops[] = { RegClass, V0, SubReg0, V1, SubReg1, V2, SubReg2 };
+  return CurDAG->getMachineNode(TargetOpcode::REG_SEQUENCE, dl, MVT::Untyped,
+                                Ops);
+}
+
+SDNode *AArch64DAGToDAGISel::createDQuadNode(SDValue V0, SDValue V1, SDValue V2,
+                                             SDValue V3) {
+  SDLoc dl(V0.getNode());
+  SDValue RegClass =
+      CurDAG->getTargetConstant(AArch64::DQuadRegClassID, MVT::i32);
+  SDValue SubReg0 = CurDAG->getTargetConstant(AArch64::dsub_0, MVT::i32);
+  SDValue SubReg1 = CurDAG->getTargetConstant(AArch64::dsub_1, MVT::i32);
+  SDValue SubReg2 = CurDAG->getTargetConstant(AArch64::dsub_2, MVT::i32);
+  SDValue SubReg3 = CurDAG->getTargetConstant(AArch64::dsub_3, MVT::i32);
+  const SDValue Ops[] = { RegClass, V0, SubReg0, V1, SubReg1, V2, SubReg2,  V3,
+                          SubReg3 };
+  return CurDAG->getMachineNode(TargetOpcode::REG_SEQUENCE, dl, MVT::v4i64,
+                                Ops);
+}
+
+SDNode *AArch64DAGToDAGISel::createQQuadNode(SDValue V0, SDValue V1, SDValue V2,
+                                             SDValue V3) {
+  SDLoc dl(V0.getNode());
+  SDValue RegClass =
+      CurDAG->getTargetConstant(AArch64::QQuadRegClassID, MVT::i32);
+  SDValue SubReg0 = CurDAG->getTargetConstant(AArch64::qsub_0, MVT::i32);
+  SDValue SubReg1 = CurDAG->getTargetConstant(AArch64::qsub_1, MVT::i32);
+  SDValue SubReg2 = CurDAG->getTargetConstant(AArch64::qsub_2, MVT::i32);
+  SDValue SubReg3 = CurDAG->getTargetConstant(AArch64::qsub_3, MVT::i32);
+  const SDValue Ops[] = { RegClass, V0, SubReg0, V1, SubReg1, V2, SubReg2,  V3,
+                          SubReg3 };
+  return CurDAG->getMachineNode(TargetOpcode::REG_SEQUENCE, dl, MVT::v8i64,
+                                Ops);
+}
+
+SDNode *AArch64DAGToDAGISel::SelectVLD(SDNode *N, unsigned NumVecs,
+                                       const uint16_t *Opcodes) {
+  assert(NumVecs >= 1 && NumVecs <= 4 && "VLD NumVecs out-of-range");
+
+  EVT VT = N->getValueType(0);
+  unsigned OpcodeIndex;
+  switch (VT.getSimpleVT().SimpleTy) {
+  default: llvm_unreachable("unhandled vector load type");
+  case MVT::v8i8:  OpcodeIndex = 0; break;
+  case MVT::v4i16: OpcodeIndex = 1; break;
+  case MVT::v2f32:
+  case MVT::v2i32: OpcodeIndex = 2; break;
+  case MVT::v1f64:
+  case MVT::v1i64: OpcodeIndex = 3; break;
+  case MVT::v16i8: OpcodeIndex = 4; break;
+  case MVT::v8f16:
+  case MVT::v8i16: OpcodeIndex = 5; break;
+  case MVT::v4f32:
+  case MVT::v4i32: OpcodeIndex = 6; break;
+  case MVT::v2f64:
+  case MVT::v2i64: OpcodeIndex = 7; break;
+  }
+  unsigned Opc = Opcodes[OpcodeIndex];
+
+  SmallVector<SDValue, 2> Ops;
+  Ops.push_back(N->getOperand(2)); // Push back the Memory Address
+  Ops.push_back(N->getOperand(0)); // Push back the Chain
+
+  std::vector<EVT> ResTys;
+  bool is64BitVector = VT.is64BitVector();
+
+  if (NumVecs == 1)
+    ResTys.push_back(VT);
+  else if (NumVecs == 3)
+    ResTys.push_back(MVT::Untyped);
+  else {
+    EVT ResTy = EVT::getVectorVT(*CurDAG->getContext(), MVT::i64,
+                                 is64BitVector ? NumVecs : NumVecs * 2);
+    ResTys.push_back(ResTy);
+  }
+
+  ResTys.push_back(MVT::Other); // Type of the Chain
+  SDLoc dl(N);
+  SDNode *VLd = CurDAG->getMachineNode(Opc, dl, ResTys, Ops);
+
+  // Transfer memoperands.
+  MachineSDNode::mmo_iterator MemOp = MF->allocateMemRefsArray(1);
+  MemOp[0] = cast<MemIntrinsicSDNode>(N)->getMemOperand();
+  cast<MachineSDNode>(VLd)->setMemRefs(MemOp, MemOp + 1);
+
+  if (NumVecs == 1)
+    return VLd;
+
+  // If NumVecs > 1, the return result is a super register containing 2-4
+  // consecutive vector registers.
+  SDValue SuperReg = SDValue(VLd, 0);
+
+  unsigned Sub0 = is64BitVector ? AArch64::dsub_0 : AArch64::qsub_0;
+  for (unsigned Vec = 0; Vec < NumVecs; ++Vec)
+    ReplaceUses(SDValue(N, Vec),
+                CurDAG->getTargetExtractSubreg(Sub0 + Vec, dl, VT, SuperReg));
+  // Update users of the Chain
+  ReplaceUses(SDValue(N, NumVecs), SDValue(VLd, 1));
+
+  return NULL;
+}
+
+SDNode *AArch64DAGToDAGISel::SelectVST(SDNode *N, unsigned NumVecs,
+                                       const uint16_t *Opcodes) {
+  assert(NumVecs >= 1 && NumVecs <= 4 && "VST NumVecs out-of-range");
+  SDLoc dl(N);
+
+  MachineSDNode::mmo_iterator MemOp = MF->allocateMemRefsArray(1);
+  MemOp[0] = cast<MemIntrinsicSDNode>(N)->getMemOperand();
+
+  unsigned Vec0Idx = 3;
+  EVT VT = N->getOperand(Vec0Idx).getValueType();
+  unsigned OpcodeIndex;
+  switch (VT.getSimpleVT().SimpleTy) {
+  default: llvm_unreachable("unhandled vector store type");
+  case MVT::v8i8:  OpcodeIndex = 0; break;
+  case MVT::v4i16: OpcodeIndex = 1; break;
+  case MVT::v2f32:
+  case MVT::v2i32: OpcodeIndex = 2; break;
+  case MVT::v1f64:
+  case MVT::v1i64: OpcodeIndex = 3; break;
+  case MVT::v16i8: OpcodeIndex = 4; break;
+  case MVT::v8f16:
+  case MVT::v8i16: OpcodeIndex = 5; break;
+  case MVT::v4f32:
+  case MVT::v4i32: OpcodeIndex = 6; break;
+  case MVT::v2f64:
+  case MVT::v2i64: OpcodeIndex = 7; break;
+  }
+  unsigned Opc = Opcodes[OpcodeIndex];
+
+  std::vector<EVT> ResTys;
+  ResTys.push_back(MVT::Other); // Type for the Chain
+
+  SmallVector<SDValue, 6> Ops;
+  Ops.push_back(N->getOperand(2)); // Push back the Memory Address
+
+  bool is64BitVector = VT.is64BitVector();
+
+  SDValue V0 = N->getOperand(Vec0Idx + 0);
+  SDValue SrcReg;
+  if (NumVecs == 1)
+    SrcReg = V0;
+  else {
+    SDValue V1 = N->getOperand(Vec0Idx + 1);
+    if (NumVecs == 2)
+      SrcReg = is64BitVector ? SDValue(createDPairNode(V0, V1), 0)
+                             : SDValue(createQPairNode(V0, V1), 0);
+    else {
+      SDValue V2 = N->getOperand(Vec0Idx + 2);
+      if (NumVecs == 3)
+        SrcReg = is64BitVector ? SDValue(createDTripleNode(V0, V1, V2), 0)
+                               : SDValue(createQTripleNode(V0, V1, V2), 0);
+      else {
+        SDValue V3 = N->getOperand(Vec0Idx + 3);
+        SrcReg = is64BitVector ? SDValue(createDQuadNode(V0, V1, V2, V3), 0)
+                               : SDValue(createQQuadNode(V0, V1, V2, V3), 0);
+      }
+    }
+  }
+  Ops.push_back(SrcReg);
+
+  // Push back the Chain
+  Ops.push_back(N->getOperand(0));
+
+  // Transfer memoperands.
+  SDNode *VSt = CurDAG->getMachineNode(Opc, dl, ResTys, Ops);
+  cast<MachineSDNode>(VSt)->setMemRefs(MemOp, MemOp + 1);
+
+  return VSt;
+}
+
 SDNode *AArch64DAGToDAGISel::Select(SDNode *Node) {
   // Dump information about the Node being selected
   DEBUG(dbgs() << "Selecting: "; Node->dump(CurDAG); dbgs() << "\n");
@@ -536,6 +768,72 @@ SDNode *AArch64DAGToDAGISel::Select(SDNode *Node) {
     Node = ResNode;
     break;
   }
+  case ISD::INTRINSIC_VOID:
+  case ISD::INTRINSIC_W_CHAIN: {
+    unsigned IntNo = cast<ConstantSDNode>(Node->getOperand(1))->getZExtValue();
+    switch (IntNo) {
+    default:
+      break;
+
+    case Intrinsic::arm_neon_vld1: {
+      static const uint16_t Opcodes[] = { AArch64::LD1_8B,  AArch64::LD1_4H,
+                                          AArch64::LD1_2S,  AArch64::LD1_1D,
+                                          AArch64::LD1_16B, AArch64::LD1_8H,
+                                          AArch64::LD1_4S,  AArch64::LD1_2D };
+      return SelectVLD(Node, 1, Opcodes);
+    }
+    case Intrinsic::arm_neon_vld2: {
+      static const uint16_t Opcodes[] = { AArch64::LD2_8B,  AArch64::LD2_4H,
+                                          AArch64::LD2_2S,  AArch64::LD1_2V_1D,
+                                          AArch64::LD2_16B, AArch64::LD2_8H,
+                                          AArch64::LD2_4S,  AArch64::LD2_2D };
+      return SelectVLD(Node, 2, Opcodes);
+    }
+    case Intrinsic::arm_neon_vld3: {
+      static const uint16_t Opcodes[] = { AArch64::LD3_8B,  AArch64::LD3_4H,
+                                          AArch64::LD3_2S,  AArch64::LD1_3V_1D,
+                                          AArch64::LD3_16B, AArch64::LD3_8H,
+                                          AArch64::LD3_4S,  AArch64::LD3_2D };
+      return SelectVLD(Node, 3, Opcodes);
+    }
+    case Intrinsic::arm_neon_vld4: {
+      static const uint16_t Opcodes[] = { AArch64::LD4_8B,  AArch64::LD4_4H,
+                                          AArch64::LD4_2S,  AArch64::LD1_4V_1D,
+                                          AArch64::LD4_16B, AArch64::LD4_8H,
+                                          AArch64::LD4_4S,  AArch64::LD4_2D };
+      return SelectVLD(Node, 4, Opcodes);
+    }
+    case Intrinsic::arm_neon_vst1: {
+      static const uint16_t Opcodes[] = { AArch64::ST1_8B,  AArch64::ST1_4H,
+                                          AArch64::ST1_2S,  AArch64::ST1_1D,
+                                          AArch64::ST1_16B, AArch64::ST1_8H,
+                                          AArch64::ST1_4S,  AArch64::ST1_2D };
+      return SelectVST(Node, 1, Opcodes);
+    }
+    case Intrinsic::arm_neon_vst2: {
+      static const uint16_t Opcodes[] = { AArch64::ST2_8B,  AArch64::ST2_4H,
+                                          AArch64::ST2_2S,  AArch64::ST1_2V_1D,
+                                          AArch64::ST2_16B, AArch64::ST2_8H,
+                                          AArch64::ST2_4S,  AArch64::ST2_2D };
+      return SelectVST(Node, 2, Opcodes);
+    }
+    case Intrinsic::arm_neon_vst3: {
+      static const uint16_t Opcodes[] = { AArch64::ST3_8B,  AArch64::ST3_4H,
+                                          AArch64::ST3_2S,  AArch64::ST1_3V_1D,
+                                          AArch64::ST3_16B, AArch64::ST3_8H,
+                                          AArch64::ST3_4S,  AArch64::ST3_2D };
+      return SelectVST(Node, 3, Opcodes);
+    }
+    case Intrinsic::arm_neon_vst4: {
+      static const uint16_t Opcodes[] = { AArch64::ST4_8B,  AArch64::ST4_4H,
+                                          AArch64::ST4_2S,  AArch64::ST1_4V_1D,
+                                          AArch64::ST4_16B, AArch64::ST4_8H,
+                                          AArch64::ST4_4S,  AArch64::ST4_2D };
+      return SelectVST(Node, 4, Opcodes);
+    }
+    }
+    break;
+  }
   default:
     break; // Let generic code handle it
   }
diff --git a/lib/Target/AArch64/AArch64ISelLowering.cpp b/lib/Target/AArch64/AArch64ISelLowering.cpp
index d70548aa357..d89213c80d7 100644
--- a/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -3681,3 +3681,57 @@ AArch64TargetLowering::getRegForInlineAsmConstraint(
   // constraint into a member of a register class.
   return TargetLowering::getRegForInlineAsmConstraint(Constraint, VT);
 }
+
+/// Represent NEON load and store intrinsics as MemIntrinsicNodes.
+/// The associated MachineMemOperands record the alignment specified
+/// in the intrinsic calls.
+bool AArch64TargetLowering::getTgtMemIntrinsic(IntrinsicInfo &Info,
+                                               const CallInst &I,
+                                               unsigned Intrinsic) const {
+  switch (Intrinsic) {
+  case Intrinsic::arm_neon_vld1:
+  case Intrinsic::arm_neon_vld2:
+  case Intrinsic::arm_neon_vld3:
+  case Intrinsic::arm_neon_vld4: {
+    Info.opc = ISD::INTRINSIC_W_CHAIN;
+    // Conservatively set memVT to the entire set of vectors loaded.
+    uint64_t NumElts = getDataLayout()->getTypeAllocSize(I.getType()) / 8;
+    Info.memVT = EVT::getVectorVT(I.getType()->getContext(), MVT::i64, NumElts);
+    Info.ptrVal = I.getArgOperand(0);
+    Info.offset = 0;
+    Value *AlignArg = I.getArgOperand(I.getNumArgOperands() - 1);
+    Info.align = cast<ConstantInt>(AlignArg)->getZExtValue();
+    Info.vol = false; // volatile loads with NEON intrinsics not supported
+    Info.readMem = true;
+    Info.writeMem = false;
+    return true;
+  }
+  case Intrinsic::arm_neon_vst1:
+  case Intrinsic::arm_neon_vst2:
+  case Intrinsic::arm_neon_vst3:
+  case Intrinsic::arm_neon_vst4: {
+    Info.opc = ISD::INTRINSIC_VOID;
+    // Conservatively set memVT to the entire set of vectors stored.
+    unsigned NumElts = 0;
+    for (unsigned ArgI = 1, ArgE = I.getNumArgOperands(); ArgI < ArgE; ++ArgI) {
+      Type *ArgTy = I.getArgOperand(ArgI)->getType();
+      if (!ArgTy->isVectorTy())
+        break;
+      NumElts += getDataLayout()->getTypeAllocSize(ArgTy) / 8;
+    }
+    Info.memVT = EVT::getVectorVT(I.getType()->getContext(), MVT::i64, NumElts);
+    Info.ptrVal = I.getArgOperand(0);
+    Info.offset = 0;
+    Value *AlignArg = I.getArgOperand(I.getNumArgOperands() - 1);
+    Info.align = cast<ConstantInt>(AlignArg)->getZExtValue();
+    Info.vol = false; // volatile stores with NEON intrinsics not supported
+    Info.readMem = false;
+    Info.writeMem = true;
+    return true;
+  }
+  default:
+    break;
+  }
+
+  return false;
+}
diff --git a/lib/Target/AArch64/AArch64ISelLowering.h b/lib/Target/AArch64/AArch64ISelLowering.h
index 3e309a95564..da7f62361ba 100644
--- a/lib/Target/AArch64/AArch64ISelLowering.h
+++ b/lib/Target/AArch64/AArch64ISelLowering.h
@@ -281,6 +281,10 @@ public:
 
   std::pair<unsigned, const TargetRegisterClass*>
   getRegForInlineAsmConstraint(const std::string &Constraint, MVT VT) const;
+
+  virtual bool getTgtMemIntrinsic(IntrinsicInfo &Info, const CallInst &I,
+                                  unsigned Intrinsic) const LLVM_OVERRIDE;
+
 private:
   const InstrItineraryData *Itins;
 
diff --git a/lib/Target/AArch64/AArch64InstrFormats.td b/lib/Target/AArch64/AArch64InstrFormats.td
index 5781578a552..ab4d083e0eb 100644
--- a/lib/Target/AArch64/AArch64InstrFormats.td
+++ b/lib/Target/AArch64/AArch64InstrFormats.td
@@ -1194,5 +1194,23 @@ class NeonI_Scalar2SameMisc<bit u, bits<2> size, bits<5> opcode, dag outs, dag i
   // Inherit Rd in 4-0
 }
 
+// Format AdvSIMD vector load/store multiple N-element structure
+class NeonI_LdStMult<bit q, bit l, bits<4> opcode, bits<2> size,
+                    dag outs, dag ins, string asmstr,
+                    list<dag> patterns, InstrItinClass itin>
+  : A64InstRtn<outs, ins, asmstr, patterns, itin>
+{
+  let Inst{31} = 0b0;
+  let Inst{30} = q;
+  let Inst{29-23} = 0b0011000;
+  let Inst{22} = l;
+  let Inst{21-16} = 0b000000;
+  let Inst{15-12} = opcode;
+  let Inst{11-10} = size;
+  
+  // Inherit Rn in 9-5
+  // Inherit Rt in 4-0
+}
+
 }
 
diff --git a/lib/Target/AArch64/AArch64InstrNEON.td b/lib/Target/AArch64/AArch64InstrNEON.td
index a9f60619a21..355de537686 100644
--- a/lib/Target/AArch64/AArch64InstrNEON.td
+++ b/lib/Target/AArch64/AArch64InstrNEON.td
@@ -2982,6 +2982,132 @@ defm PMULL2vvv : NeonI_3VDL2_2Op_mull_v3<0b0, 0b1110, "pmull2",
 
 // End of implementation for instruction class (3V Diff)
 
+// The followings are vector load/store multiple N-element structure
+// (class SIMD lselem).
+
+// ld1:         load multiple 1-element structure to 1/2/3/4 registers.
+// ld2/ld3/ld4: load multiple N-element structure to N registers (N = 2, 3, 4).
+//              The structure consists of a sequence of sets of N values.
+//              The first element of the structure is placed in the first lane
+//              of the first first vector, the second element in the first lane
+//              of the second vector, and so on. 
+// E.g. LD1_3V_2S will load 32-bit elements {A, B, C, D, E, F} sequentially into
+// the three 64-bit vectors list {BA, DC, FE}.
+// E.g. LD3_2S will load 32-bit elements {A, B, C, D, E, F} into the three
+// 64-bit vectors list {DA, EB, FC}.
+// Store instructions store multiple structure to N registers like load.
+
+
+class NeonI_LDVList<bit q, bits<4> opcode, bits<2> size,
+                    RegisterOperand VecList, string asmop>
+  : NeonI_LdStMult<q, 1, opcode, size,
+                 (outs VecList:$Rt), (ins GPR64xsp:$Rn),
+                 asmop # "\t$Rt, [$Rn]",
+                 [],
+                 NoItinerary> {
+  let mayLoad = 1;
+  let neverHasSideEffects = 1;
+}
+
+multiclass LDVList_BHSD<bits<4> opcode, string List, string asmop> {
+  def _8B : NeonI_LDVList<0, opcode, 0b00,
+                          !cast<RegisterOperand>(List # "8B_operand"), asmop>;
+
+  def _4H : NeonI_LDVList<0, opcode, 0b01,
+                          !cast<RegisterOperand>(List # "4H_operand"), asmop>;
+
+  def _2S : NeonI_LDVList<0, opcode, 0b10,
+                          !cast<RegisterOperand>(List # "2S_operand"), asmop>;
+
+  def _16B : NeonI_LDVList<1, opcode, 0b00,
+                           !cast<RegisterOperand>(List # "16B_operand"), asmop>;
+
+  def _8H : NeonI_LDVList<1, opcode, 0b01,
+                          !cast<RegisterOperand>(List # "8H_operand"), asmop>;
+
+  def _4S : NeonI_LDVList<1, opcode, 0b10,
+                          !cast<RegisterOperand>(List # "4S_operand"), asmop>;
+
+  def _2D : NeonI_LDVList<1, opcode, 0b11,
+                          !cast<RegisterOperand>(List # "2D_operand"), asmop>;
+}
+
+// Load multiple N-element structure to N consecutive registers (N = 1,2,3,4)
+defm LD1 : LDVList_BHSD<0b0111, "VOne", "ld1">;
+def LD1_1D : NeonI_LDVList<0, 0b0111, 0b11, VOne1D_operand, "ld1">;
+
+defm LD2 : LDVList_BHSD<0b1000, "VPair", "ld2">;
+
+defm LD3 : LDVList_BHSD<0b0100, "VTriple", "ld3">;
+
+defm LD4 : LDVList_BHSD<0b0000, "VQuad", "ld4">;
+
+// Load multiple 1-element structure to N consecutive registers (N = 2,3,4)
+defm LD1_2V : LDVList_BHSD<0b1010, "VPair", "ld1">;
+def LD1_2V_1D : NeonI_LDVList<0, 0b1010, 0b11, VPair1D_operand, "ld1">;
+
+defm LD1_3V : LDVList_BHSD<0b0110, "VTriple", "ld1">;
+def LD1_3V_1D : NeonI_LDVList<0, 0b0110, 0b11, VTriple1D_operand, "ld1">;
+
+defm LD1_4V : LDVList_BHSD<0b0010, "VQuad", "ld1">;
+def LD1_4V_1D : NeonI_LDVList<0, 0b0010, 0b11, VQuad1D_operand, "ld1">;
+
+class NeonI_STVList<bit q, bits<4> opcode, bits<2> size,
+                    RegisterOperand VecList, string asmop>
+  : NeonI_LdStMult<q, 0, opcode, size,
+                 (outs), (ins GPR64xsp:$Rn, VecList:$Rt), 
+                 asmop # "\t$Rt, [$Rn]",
+                 [], 
+                 NoItinerary> {
+  let mayStore = 1;
+  let neverHasSideEffects = 1;
+}
+
+multiclass STVList_BHSD<bits<4> opcode, string List, string asmop> {
+  def _8B : NeonI_STVList<0, opcode, 0b00,
+                          !cast<RegisterOperand>(List # "8B_operand"), asmop>;
+
+  def _4H : NeonI_STVList<0, opcode, 0b01,
+                          !cast<RegisterOperand>(List # "4H_operand"), asmop>;
+
+  def _2S : NeonI_STVList<0, opcode, 0b10,
+                          !cast<RegisterOperand>(List # "2S_operand"), asmop>;
+
+  def _16B : NeonI_STVList<1, opcode, 0b00,
+                           !cast<RegisterOperand>(List # "16B_operand"), asmop>;
+
+  def _8H : NeonI_STVList<1, opcode, 0b01,
+                          !cast<RegisterOperand>(List # "8H_operand"), asmop>;
+
+  def _4S : NeonI_STVList<1, opcode, 0b10,
+                          !cast<RegisterOperand>(List # "4S_operand"), asmop>;
+
+  def _2D : NeonI_STVList<1, opcode, 0b11,
+                          !cast<RegisterOperand>(List # "2D_operand"), asmop>;
+}
+
+// Store multiple N-element structures from N registers (N = 1,2,3,4)
+defm ST1 : STVList_BHSD<0b0111, "VOne", "st1">;
+def ST1_1D : NeonI_STVList<0, 0b0111, 0b11, VOne1D_operand, "st1">;
+
+defm ST2 : STVList_BHSD<0b1000, "VPair", "st2">;
+
+defm ST3 : STVList_BHSD<0b0100, "VTriple", "st3">;
+
+defm ST4 : STVList_BHSD<0b0000, "VQuad", "st4">;
+
+// Store multiple 1-element structures from N consecutive registers (N = 2,3,4)
+defm ST1_2V : STVList_BHSD<0b1010, "VPair", "st1">;
+def ST1_2V_1D : NeonI_STVList<0, 0b1010, 0b11, VPair1D_operand, "st1">;
+
+defm ST1_3V : STVList_BHSD<0b0110, "VTriple", "st1">;
+def ST1_3V_1D : NeonI_STVList<0, 0b0110, 0b11, VTriple1D_operand, "st1">;
+
+defm ST1_4V : STVList_BHSD<0b0010, "VQuad", "st1">;
+def ST1_4V_1D : NeonI_STVList<0, 0b0010, 0b11, VQuad1D_operand, "st1">;
+
+// End of vector load/store multiple N-element structure(class SIMD lselem)
+
 // Scalar Arithmetic
 
 class NeonI_Scalar3Same_D_size<bit u, bits<5> opcode, string asmop>
diff --git a/lib/Target/AArch64/AArch64RegisterInfo.td b/lib/Target/AArch64/AArch64RegisterInfo.td
index b7a6acb348e..5e2b1963977 100644
--- a/lib/Target/AArch64/AArch64RegisterInfo.td
+++ b/lib/Target/AArch64/AArch64RegisterInfo.td
@@ -17,6 +17,20 @@ def sub_64 : SubRegIndex<64>;
 def sub_32 : SubRegIndex<32>;
 def sub_16 : SubRegIndex<16>;
 def sub_8  : SubRegIndex<8>;
+
+// Note: Code depends on these having consecutive numbers.
+def qqsub : SubRegIndex<256, 256>;
+
+def qsub_0 : SubRegIndex<128>;
+def qsub_1 : SubRegIndex<128, 128>;
+def qsub_2 : ComposedSubRegIndex<qqsub, qsub_0>;
+def qsub_3 : ComposedSubRegIndex<qqsub, qsub_1>;
+
+def dsub_0 : SubRegIndex<64>;
+def dsub_1 : SubRegIndex<64, 64>;
+def dsub_2 : ComposedSubRegIndex<qsub_1, dsub_0>;
+def dsub_3 : ComposedSubRegIndex<qsub_1, dsub_1>;
+def dsub_4 : ComposedSubRegIndex<qsub_2, dsub_0>;
 }
 
 // Registers are identified with 5-bit ID numbers.
@@ -188,3 +202,90 @@ def FlagClass : RegisterClass<"AArch64", [i32], 32, (add NZCV)> {
   let CopyCost = -1;
   let isAllocatable = 0;
 }
+
+//===----------------------------------------------------------------------===//
+//  Consecutive vector registers
+//===----------------------------------------------------------------------===//
+// 2 Consecutive 64-bit registers: D0_D1, D1_D2, ..., D30_D31
+def Tuples2D : RegisterTuples<[dsub_0, dsub_1],
+                              [(rotl FPR64, 0), (rotl FPR64, 1)]>;
+                              
+// 3 Consecutive 64-bit registers: D0_D1_D2, ..., D31_D0_D1
+def Tuples3D : RegisterTuples<[dsub_0, dsub_1, dsub_2],
+                              [(rotl FPR64, 0), (rotl FPR64, 1),
+                               (rotl FPR64, 2)]>;
+                               
+// 4 Consecutive 64-bit registers: D0_D1_D2_D3, ..., D31_D0_D1_D2
+def Tuples4D : RegisterTuples<[dsub_0, dsub_1, dsub_2, dsub_3],
+                              [(rotl FPR64, 0), (rotl FPR64, 1),
+                               (rotl FPR64, 2), (rotl FPR64, 3)]>;
+
+// 2 Consecutive 128-bit registers: Q0_Q1, Q1_Q2, ..., Q30_Q31
+def Tuples2Q : RegisterTuples<[qsub_0, qsub_1],
+                              [(rotl FPR128, 0), (rotl FPR128, 1)]>;
+
+// 3 Consecutive 128-bit registers: Q0_Q1_Q2, ..., Q31_Q0_Q1
+def Tuples3Q : RegisterTuples<[qsub_0, qsub_1, qsub_2],
+                              [(rotl FPR128, 0), (rotl FPR128, 1),
+                               (rotl FPR128, 2)]>;
+                               
+// 4 Consecutive 128-bit registers: Q0_Q1_Q2_Q3, ..., Q31_Q0_Q1_Q2
+def Tuples4Q : RegisterTuples<[qsub_0, qsub_1, qsub_2, qsub_3],
+                              [(rotl FPR128, 0), (rotl FPR128, 1),
+                               (rotl FPR128, 2), (rotl FPR128, 3)]>;
+
+// The followings are super register classes to model 2/3/4 consecutive
+// 64-bit/128-bit registers.
+
+def DPair : RegisterClass<"AArch64", [v2i64], 64, (add Tuples2D)>;
+
+def DTriple : RegisterClass<"AArch64", [untyped], 64, (add Tuples3D)> {
+  let Size = 192; // 3 x 64 bits, we have no predefined type of that size.
+}
+
+def DQuad : RegisterClass<"AArch64", [v4i64], 64, (add Tuples4D)>;
+
+def QPair : RegisterClass<"AArch64", [v4i64], 128, (add Tuples2Q)>;
+
+def QTriple : RegisterClass<"AArch64", [untyped], 128, (add Tuples3Q)> {
+  let Size = 384; // 3 x 128 bits, we have no predefined type of that size.
+}
+
+def QQuad : RegisterClass<"AArch64", [v8i64], 128, (add Tuples4Q)>;
+
+
+// The followings are vector list operands
+multiclass VectorList_operands<string PREFIX, string LAYOUT, int Count,
+                               RegisterClass RegList> {
+  def _asmoperand : AsmOperandClass {
+    let Name = PREFIX # LAYOUT # Count;
+    let RenderMethod = "addVectorListOperands";
+    let PredicateMethod = 
+        "isVectorList<A64Layout::_" # LAYOUT # ", " # Count # ">";
+    let ParserMethod = "ParseVectorList";
+  }
+
+  def _operand : RegisterOperand<RegList,
+        "printVectorList<A64Layout::_" # LAYOUT # ", " # Count # ">"> {
+    let ParserMatchClass =
+      !cast<AsmOperandClass>(PREFIX # LAYOUT # "_asmoperand");
+  }
+}
+
+multiclass VectorList_BHSD<string PREFIX, int Count, RegisterClass DRegList,
+                           RegisterClass QRegList> {
+  defm 8B : VectorList_operands<PREFIX, "8B", Count, DRegList>;
+  defm 4H : VectorList_operands<PREFIX, "4H", Count, DRegList>;
+  defm 2S : VectorList_operands<PREFIX, "2S", Count, DRegList>;
+  defm 1D : VectorList_operands<PREFIX, "1D", Count, DRegList>;
+  defm 16B : VectorList_operands<PREFIX, "16B", Count, QRegList>;
+  defm 8H : VectorList_operands<PREFIX, "8H", Count, QRegList>;
+  defm 4S : VectorList_operands<PREFIX, "4S", Count, QRegList>;
+  defm 2D : VectorList_operands<PREFIX, "2D", Count, QRegList>;
+}
+
+// Vector list operand with 1/2/3/4 registers: VOne8B_operand,..., VQuad2D_operand
+defm VOne : VectorList_BHSD<"VOne", 1, FPR64, FPR128>;
+defm VPair : VectorList_BHSD<"VPair", 2, DPair, QPair>;
+defm VTriple : VectorList_BHSD<"VTriple", 3, DTriple, QTriple>;
+defm VQuad : VectorList_BHSD<"VQuad", 4, DQuad, QQuad>;
\ No newline at end of file
diff --git a/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp b/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp
index 51638d90e3c..127c7eca0ee 100644
--- a/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp
+++ b/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp
@@ -127,6 +127,11 @@ public:
   OperandMatchResultTy
   ParseSysRegOperand(SmallVectorImpl<MCParsedAsmOperand*> &Operands);
 
+  bool TryParseVector(uint32_t &RegNum, SMLoc &RegEndLoc, StringRef &Layout,
+                      SMLoc &LayoutLoc);
+
+  OperandMatchResultTy ParseVectorList(SmallVectorImpl<MCParsedAsmOperand *> &);
+
   bool validateInstruction(MCInst &Inst,
                           const SmallVectorImpl<MCParsedAsmOperand*> &Operands);
 
@@ -154,6 +159,7 @@ private:
     k_Immediate,      // Including expressions referencing symbols
     k_Register,
     k_ShiftExtend,
+    k_VectorList,     // A sequential list of 1 to 4 registers.
     k_SysReg,         // The register operand of MRS and MSR instructions
     k_Token,          // The mnemonic; other raw tokens the auto-generated
     k_WrappedRegister // Load/store exclusive permit a wrapped register.
@@ -189,6 +195,13 @@ private:
     bool ImplicitAmount;
   };
 
+  // A vector register list is a sequential list of 1 to 4 registers.
+  struct VectorListOp {
+    unsigned RegNum;
+    unsigned Count;
+    A64Layout::VectorLayout Layout;
+  };
+
   struct SysRegOp {
     const char *Data;
     unsigned Length;
@@ -206,6 +219,7 @@ private:
     struct ImmOp Imm;
     struct RegOp Reg;
     struct ShiftExtendOp ShiftExtend;
+    struct VectorListOp VectorList;
     struct SysRegOp SysReg;
     struct TokOp Tok;
   };
@@ -717,6 +731,12 @@ public:
     return ShiftExtend.Amount == 8 || ShiftExtend.Amount == 16;
   }
 
+  template <A64Layout::VectorLayout Layout, unsigned Count>
+  bool isVectorList() const {
+    return Kind == k_VectorList && VectorList.Layout == Layout &&
+           VectorList.Count == Count;
+  }
+
   template <int MemSize> bool isSImm7Scaled() const {
     if (!isImm())
       return false;
@@ -837,6 +857,18 @@ public:
     return Op;
   }
 
+  static AArch64Operand *CreateVectorList(unsigned RegNum, unsigned Count,
+                                          A64Layout::VectorLayout Layout,
+                                          SMLoc S, SMLoc E) {
+    AArch64Operand *Op = new AArch64Operand(k_VectorList, S, E);
+    Op->VectorList.RegNum = RegNum;
+    Op->VectorList.Count = Count;
+    Op->VectorList.Layout = Layout;
+    Op->StartLoc = S;
+    Op->EndLoc = E;
+    return Op;
+  }
+
   static AArch64Operand *CreateToken(StringRef Str, SMLoc S) {
     AArch64Operand *Op = new AArch64Operand(k_Token, S, S);
     Op->Tok.Data = Str.data();
@@ -1184,6 +1216,11 @@ public:
     }
     Inst.addOperand(MCOperand::CreateImm(Imm));
   }
+
+  void addVectorListOperands(MCInst &Inst, unsigned N) const {
+    assert(N == 1 && "Invalid number of operands!");
+    Inst.addOperand(MCOperand::CreateReg(VectorList.RegNum));
+  }
 };
 
 } // end anonymous namespace.
@@ -1223,7 +1260,6 @@ AArch64AsmParser::ParseOperand(SmallVectorImpl<MCParsedAsmOperand*> &Operands,
       else
         return MatchOperand_Success;
     }
-
     // ... or it might be a symbolish thing
   }
     // Fall through
@@ -1267,7 +1303,7 @@ AArch64AsmParser::ParseOperand(SmallVectorImpl<MCParsedAsmOperand*> &Operands,
     return ParseOperand(Operands, Mnemonic);
   }
   // The following will likely be useful later, but not in very early cases
-  case AsmToken::LCurly:  // Weird SIMD lists
+  case AsmToken::LCurly: // SIMD vector list is not parsed here
     llvm_unreachable("Don't know how to deal with '{' in operand");
     return MatchOperand_ParseFail;
   }
@@ -1890,6 +1926,132 @@ AArch64AsmParser::ParseShiftExtend(
   return MatchOperand_Success;
 }
 
+/// Try to parse a vector register token, If it is a vector register,
+/// the token is eaten and return true. Otherwise return false.
+bool AArch64AsmParser::TryParseVector(uint32_t &RegNum, SMLoc &RegEndLoc,
+                                      StringRef &Layout, SMLoc &LayoutLoc) {
+  bool IsVector = true;
+
+  if (!IdentifyRegister(RegNum, RegEndLoc, Layout, LayoutLoc))
+    IsVector = false;
+
+  if (!AArch64MCRegisterClasses[AArch64::FPR64RegClassID].contains(RegNum) &&
+      !AArch64MCRegisterClasses[AArch64::FPR128RegClassID].contains(RegNum))
+    IsVector = false;
+
+  if (Layout.size() == 0)
+    IsVector = false;
+
+  if (!IsVector)
+    Error(Parser.getTok().getLoc(), "expected vector type register");
+
+  Parser.Lex(); // Eat this token.
+  return IsVector;
+}
+
+
+// A vector list contains 1-4 consecutive registers.
+// Now there are two kinds of vector list when number of vector > 1:
+//   (1) {Vn.layout, Vn+1.layout, ... , Vm.layout}
+//   (2) {Vn.layout - Vm.layout}
+AArch64AsmParser::OperandMatchResultTy AArch64AsmParser::ParseVectorList(
+    SmallVectorImpl<MCParsedAsmOperand *> &Operands) {
+  if (Parser.getTok().isNot(AsmToken::LCurly)) {
+    Error(Parser.getTok().getLoc(), "'{' expected");
+    return MatchOperand_ParseFail;
+  }
+  SMLoc SLoc = Parser.getTok().getLoc();
+  Parser.Lex(); // Eat '{' token.
+
+  unsigned Reg, Count = 1;
+  StringRef LayoutStr;
+  SMLoc RegEndLoc, LayoutLoc;
+  if (!TryParseVector(Reg, RegEndLoc, LayoutStr, LayoutLoc))
+    return MatchOperand_ParseFail;
+
+  if (Parser.getTok().is(AsmToken::Minus)) {
+    Parser.Lex(); // Eat the minus.
+
+    unsigned Reg2;
+    StringRef LayoutStr2;
+    SMLoc RegEndLoc2, LayoutLoc2;
+    SMLoc RegLoc2 = Parser.getTok().getLoc();
+
+    if (!TryParseVector(Reg2, RegEndLoc2, LayoutStr2, LayoutLoc2))
+      return MatchOperand_ParseFail;
+    unsigned Space = (Reg < Reg2) ? (Reg2 - Reg) : (Reg2 + 32 - Reg);
+
+    if (LayoutStr != LayoutStr2) {
+      Error(LayoutLoc2, "expected the same vector layout");
+      return MatchOperand_ParseFail;
+    }
+    if (Space == 0 || Space > 3) {
+      Error(RegLoc2, "invalid number of vectors");
+      return MatchOperand_ParseFail;
+    }
+
+    Count += Space;
+  } else {
+    unsigned LastReg = Reg;
+    while (Parser.getTok().is(AsmToken::Comma)) {
+      Parser.Lex(); // Eat the comma.
+      unsigned Reg2;
+      StringRef LayoutStr2;
+      SMLoc RegEndLoc2, LayoutLoc2;
+      SMLoc RegLoc2 = Parser.getTok().getLoc();
+
+      if (!TryParseVector(Reg2, RegEndLoc2, LayoutStr2, LayoutLoc2))
+        return MatchOperand_ParseFail;
+      unsigned Space = (LastReg < Reg2) ? (Reg2 - LastReg)
+                                        : (Reg2 + 32 - LastReg);
+      Count++;
+
+      // The space between two vectors should be 1. And they should have the same layout.
+      // Total count shouldn't be great than 4
+      if (Space != 1) {
+        Error(RegLoc2, "invalid space between two vectors");
+        return MatchOperand_ParseFail;
+      }
+      if (LayoutStr != LayoutStr2) {
+        Error(LayoutLoc2, "expected the same vector layout");
+        return MatchOperand_ParseFail;
+      }
+      if (Count > 4) {
+        Error(RegLoc2, "invalid number of vectors");
+        return MatchOperand_ParseFail;
+      }
+
+      LastReg = Reg2;
+    }
+  }
+
+  if (Parser.getTok().isNot(AsmToken::RCurly)) {
+    Error(Parser.getTok().getLoc(), "'}' expected");
+    return MatchOperand_ParseFail;
+  }
+  SMLoc ELoc = Parser.getTok().getLoc();
+  Parser.Lex(); // Eat '}' token.
+
+  A64Layout::VectorLayout Layout = A64StringToVectorLayout(LayoutStr);
+  if (Count > 1) { // If count > 1, create vector list using super register.
+    bool IsVec64 = (Layout < A64Layout::_16B) ? true : false;
+    static unsigned SupRegIDs[3][2] = {
+      { AArch64::QPairRegClassID, AArch64::DPairRegClassID },
+      { AArch64::QTripleRegClassID, AArch64::DTripleRegClassID },
+      { AArch64::QQuadRegClassID, AArch64::DQuadRegClassID }
+    };
+    unsigned SupRegID = SupRegIDs[Count - 2][static_cast<int>(IsVec64)];
+    unsigned Sub0 = IsVec64 ? AArch64::dsub_0 : AArch64::qsub_0;
+    const MCRegisterInfo *MRI = getContext().getRegisterInfo();
+    Reg = MRI->getMatchingSuperReg(Reg, Sub0,
+                                   &AArch64MCRegisterClasses[SupRegID]);
+  }
+  Operands.push_back(
+      AArch64Operand::CreateVectorList(Reg, Count, Layout, SLoc, ELoc));
+
+  return MatchOperand_Success;
+}
+
 // FIXME: We would really like to be able to tablegen'erate this.
 bool AArch64AsmParser::
 validateInstruction(MCInst &Inst,
diff --git a/lib/Target/AArch64/Disassembler/AArch64Disassembler.cpp b/lib/Target/AArch64/Disassembler/AArch64Disassembler.cpp
index b9d7c1684d5..9a97fbe0178 100644
--- a/lib/Target/AArch64/Disassembler/AArch64Disassembler.cpp
+++ b/lib/Target/AArch64/Disassembler/AArch64Disassembler.cpp
@@ -89,6 +89,25 @@ static DecodeStatus DecodeFPR128LoRegisterClass(llvm::MCInst &Inst,
                                                 unsigned RegNo, uint64_t Address,
                                                 const void *Decoder);
 
+static DecodeStatus DecodeDPairRegisterClass(llvm::MCInst &Inst, unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder);
+static DecodeStatus DecodeQPairRegisterClass(llvm::MCInst &Inst, unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder);
+static DecodeStatus DecodeDTripleRegisterClass(llvm::MCInst &Inst,
+                                               unsigned RegNo, uint64_t Address,
+                                               const void *Decoder);
+static DecodeStatus DecodeQTripleRegisterClass(llvm::MCInst &Inst,
+                                               unsigned RegNo, uint64_t Address,
+                                               const void *Decoder);
+static DecodeStatus DecodeDQuadRegisterClass(llvm::MCInst &Inst, unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder);
+static DecodeStatus DecodeQQuadRegisterClass(llvm::MCInst &Inst, unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder);
+
 static DecodeStatus DecodeAddrRegExtendOperand(llvm::MCInst &Inst,
                                                unsigned OptionHiS,
                                                uint64_t Address,
@@ -361,6 +380,59 @@ DecodeFPR128LoRegisterClass(llvm::MCInst &Inst, unsigned RegNo,
   return DecodeFPR128RegisterClass(Inst, RegNo, Address, Decoder);
 }
 
+static DecodeStatus DecodeRegisterClassByID(llvm::MCInst &Inst, unsigned RegNo,
+                                            unsigned RegID,
+                                            const void *Decoder) {
+  if (RegNo > 31)
+    return MCDisassembler::Fail;
+
+  uint16_t Register = getReg(Decoder, RegID, RegNo);
+  Inst.addOperand(MCOperand::CreateReg(Register));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeDPairRegisterClass(llvm::MCInst &Inst, unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder) {
+  return DecodeRegisterClassByID(Inst, RegNo, AArch64::DPairRegClassID,
+                                 Decoder);
+}
+
+static DecodeStatus DecodeQPairRegisterClass(llvm::MCInst &Inst, unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder) {
+  return DecodeRegisterClassByID(Inst, RegNo, AArch64::QPairRegClassID,
+                                 Decoder);
+}
+
+static DecodeStatus DecodeDTripleRegisterClass(llvm::MCInst &Inst,
+                                               unsigned RegNo, uint64_t Address,
+                                               const void *Decoder) {
+  return DecodeRegisterClassByID(Inst, RegNo, AArch64::DTripleRegClassID,
+                                 Decoder);
+}
+
+static DecodeStatus DecodeQTripleRegisterClass(llvm::MCInst &Inst,
+                                               unsigned RegNo, uint64_t Address,
+                                               const void *Decoder) {
+  return DecodeRegisterClassByID(Inst, RegNo, AArch64::QTripleRegClassID,
+                                 Decoder);
+}
+
+static DecodeStatus DecodeDQuadRegisterClass(llvm::MCInst &Inst, unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder) {
+  return DecodeRegisterClassByID(Inst, RegNo, AArch64::DQuadRegClassID,
+                                 Decoder);
+}
+
+static DecodeStatus DecodeQQuadRegisterClass(llvm::MCInst &Inst, unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder) {
+  return DecodeRegisterClassByID(Inst, RegNo, AArch64::QQuadRegClassID,
+                                 Decoder);
+}
+
 static DecodeStatus DecodeAddrRegExtendOperand(llvm::MCInst &Inst,
                                                unsigned OptionHiS,
                                                uint64_t Address,
diff --git a/lib/Target/AArch64/InstPrinter/AArch64InstPrinter.cpp b/lib/Target/AArch64/InstPrinter/AArch64InstPrinter.cpp
index 26bd797e3b3..51335e145b7 100644
--- a/lib/Target/AArch64/InstPrinter/AArch64InstPrinter.cpp
+++ b/lib/Target/AArch64/InstPrinter/AArch64InstPrinter.cpp
@@ -507,3 +507,33 @@ void AArch64InstPrinter::printNeonUImm64MaskOperand(const MCInst *MI,
   O << "#0x";
   O.write_hex(Mask);
 }
+
+// If Count > 1, there are two valid kinds of vector list:
+//   (1) {Vn.layout, Vn+1.layout, ... , Vm.layout}
+//   (2) {Vn.layout - Vm.layout}
+// We choose the first kind as output.
+template <A64Layout::VectorLayout Layout, unsigned Count>
+void AArch64InstPrinter::printVectorList(const MCInst *MI, unsigned OpNum,
+                                         raw_ostream &O) {
+  assert(Count >= 1 && Count <= 4 && "Invalid Number of Vectors");
+
+  unsigned Reg = MI->getOperand(OpNum).getReg();
+  std::string LayoutStr = A64VectorLayoutToString(Layout);
+  O << "{";
+  if (Count > 1) { // Print sub registers separately
+    bool IsVec64 = (Layout < A64Layout::_16B) ? true : false;
+    unsigned SubRegIdx = IsVec64 ? AArch64::dsub_0 : AArch64::qsub_0;
+    for (unsigned I = 0; I < Count; I++) {
+      std::string Name = getRegisterName(MRI.getSubReg(Reg, SubRegIdx++));
+      Name[0] = 'v';
+      O << Name << LayoutStr;
+      if (I != Count - 1)
+        O << ", ";
+    }
+  } else { // Print the register directly when NumVecs is 1.
+    std::string Name = getRegisterName(Reg);
+    Name[0] = 'v';
+    O << Name << LayoutStr;
+  }
+  O << "}";
+}
diff --git a/lib/Target/AArch64/InstPrinter/AArch64InstPrinter.h b/lib/Target/AArch64/InstPrinter/AArch64InstPrinter.h
index 71c9f4a0291..28ebfc45f1f 100644
--- a/lib/Target/AArch64/InstPrinter/AArch64InstPrinter.h
+++ b/lib/Target/AArch64/InstPrinter/AArch64InstPrinter.h
@@ -174,6 +174,9 @@ public:
                                  raw_ostream &O);
   void printNeonUImm64MaskOperand(const MCInst *MI, unsigned OpNum,
                                   raw_ostream &O);
+
+  template <A64Layout::VectorLayout Layout, unsigned Count>
+  void printVectorList(const MCInst *MI, unsigned OpNum, raw_ostream &O);
 };
 }
 
diff --git a/lib/Target/AArch64/Utils/AArch64BaseInfo.h b/lib/Target/AArch64/Utils/AArch64BaseInfo.h
index e675efc9d9a..7db52381813 100644
--- a/lib/Target/AArch64/Utils/AArch64BaseInfo.h
+++ b/lib/Target/AArch64/Utils/AArch64BaseInfo.h
@@ -306,6 +306,50 @@ namespace A64SE {
     };
 }
 
+namespace A64Layout {
+    enum VectorLayout {
+        Invalid = -1,
+        _8B,
+        _4H,
+        _2S,
+        _1D,
+
+        _16B,
+        _8H,
+        _4S,
+        _2D
+    };
+}
+
+inline static const char *
+A64VectorLayoutToString(A64Layout::VectorLayout Layout) {
+  switch (Layout) {
+  case A64Layout::_8B:  return ".8b";
+  case A64Layout::_4H:  return ".4h";
+  case A64Layout::_2S:  return ".2s";
+  case A64Layout::_1D:  return ".1d";
+  case A64Layout::_16B:  return ".16b";
+  case A64Layout::_8H:  return ".8h";
+  case A64Layout::_4S:  return ".4s";
+  case A64Layout::_2D:  return ".2d";
+  default: llvm_unreachable("Unknown Vector Layout");
+  }
+}
+
+inline static A64Layout::VectorLayout
+A64StringToVectorLayout(StringRef LayoutStr) {
+  return StringSwitch<A64Layout::VectorLayout>(LayoutStr)
+             .Case(".8b", A64Layout::_8B)
+             .Case(".4h", A64Layout::_4H)
+             .Case(".2s", A64Layout::_2S)
+             .Case(".1d", A64Layout::_1D)
+             .Case(".16b", A64Layout::_16B)
+             .Case(".8h", A64Layout::_8H)
+             .Case(".4s", A64Layout::_4S)
+             .Case(".2d", A64Layout::_2D)
+             .Default(A64Layout::Invalid);
+}
+
 namespace A64SysReg {
   enum SysRegROValues {
     MDCCSR_EL0        = 0x9808, // 10  011  0000  0001  000
diff --git a/test/CodeGen/AArch64/neon-simd-ldst-multi-elem.ll b/test/CodeGen/AArch64/neon-simd-ldst-multi-elem.ll
new file mode 100644
index 00000000000..4cd76bcb00b
--- /dev/null
+++ b/test/CodeGen/AArch64/neon-simd-ldst-multi-elem.ll
@@ -0,0 +1,1228 @@
+; RUN: llc < %s -verify-machineinstrs -mtriple=aarch64-none-linux-gnu -mattr=+neon | FileCheck %s
+
+%struct.int8x16x2_t = type { [2 x <16 x i8>] }
+%struct.int16x8x2_t = type { [2 x <8 x i16>] }
+%struct.int32x4x2_t = type { [2 x <4 x i32>] }
+%struct.int64x2x2_t = type { [2 x <2 x i64>] }
+%struct.float32x4x2_t = type { [2 x <4 x float>] }
+%struct.float64x2x2_t = type { [2 x <2 x double>] }
+%struct.int8x8x2_t = type { [2 x <8 x i8>] }
+%struct.int16x4x2_t = type { [2 x <4 x i16>] }
+%struct.int32x2x2_t = type { [2 x <2 x i32>] }
+%struct.int64x1x2_t = type { [2 x <1 x i64>] }
+%struct.float32x2x2_t = type { [2 x <2 x float>] }
+%struct.float64x1x2_t = type { [2 x <1 x double>] }
+%struct.int8x16x3_t = type { [3 x <16 x i8>] }
+%struct.int16x8x3_t = type { [3 x <8 x i16>] }
+%struct.int32x4x3_t = type { [3 x <4 x i32>] }
+%struct.int64x2x3_t = type { [3 x <2 x i64>] }
+%struct.float32x4x3_t = type { [3 x <4 x float>] }
+%struct.float64x2x3_t = type { [3 x <2 x double>] }
+%struct.int8x8x3_t = type { [3 x <8 x i8>] }
+%struct.int16x4x3_t = type { [3 x <4 x i16>] }
+%struct.int32x2x3_t = type { [3 x <2 x i32>] }
+%struct.int64x1x3_t = type { [3 x <1 x i64>] }
+%struct.float32x2x3_t = type { [3 x <2 x float>] }
+%struct.float64x1x3_t = type { [3 x <1 x double>] }
+%struct.int8x16x4_t = type { [4 x <16 x i8>] }
+%struct.int16x8x4_t = type { [4 x <8 x i16>] }
+%struct.int32x4x4_t = type { [4 x <4 x i32>] }
+%struct.int64x2x4_t = type { [4 x <2 x i64>] }
+%struct.float32x4x4_t = type { [4 x <4 x float>] }
+%struct.float64x2x4_t = type { [4 x <2 x double>] }
+%struct.int8x8x4_t = type { [4 x <8 x i8>] }
+%struct.int16x4x4_t = type { [4 x <4 x i16>] }
+%struct.int32x2x4_t = type { [4 x <2 x i32>] }
+%struct.int64x1x4_t = type { [4 x <1 x i64>] }
+%struct.float32x2x4_t = type { [4 x <2 x float>] }
+%struct.float64x1x4_t = type { [4 x <1 x double>] }
+
+
+define <16 x i8> @test_vld1q_s8(i8* readonly %a) {
+; CHECK: test_vld1q_s8
+; CHECK: ld1 {v{{[0-9]+}}.16b}, [x{{[0-9]+|sp}}]
+  %vld1 = tail call <16 x i8> @llvm.arm.neon.vld1.v16i8(i8* %a, i32 1)
+  ret <16 x i8> %vld1
+}
+
+define <8 x i16> @test_vld1q_s16(i16* readonly %a) {
+; CHECK: test_vld1q_s16
+; CHECK: ld1 {v{{[0-9]+}}.8h}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i16* %a to i8*
+  %vld1 = tail call <8 x i16> @llvm.arm.neon.vld1.v8i16(i8* %1, i32 2)
+  ret <8 x i16> %vld1
+}
+
+define <4 x i32> @test_vld1q_s32(i32* readonly %a) {
+; CHECK: test_vld1q_s32
+; CHECK: ld1 {v{{[0-9]+}}.4s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i32* %a to i8*
+  %vld1 = tail call <4 x i32> @llvm.arm.neon.vld1.v4i32(i8* %1, i32 4)
+  ret <4 x i32> %vld1
+}
+
+define <2 x i64> @test_vld1q_s64(i64* readonly %a) {
+; CHECK: test_vld1q_s64
+; CHECK: ld1 {v{{[0-9]+}}.2d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i64* %a to i8*
+  %vld1 = tail call <2 x i64> @llvm.arm.neon.vld1.v2i64(i8* %1, i32 8)
+  ret <2 x i64> %vld1
+}
+
+define <4 x float> @test_vld1q_f32(float* readonly %a) {
+; CHECK: test_vld1q_f32
+; CHECK: ld1 {v{{[0-9]+}}.4s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast float* %a to i8*
+  %vld1 = tail call <4 x float> @llvm.arm.neon.vld1.v4f32(i8* %1, i32 4)
+  ret <4 x float> %vld1
+}
+
+define <2 x double> @test_vld1q_f64(double* readonly %a) {
+; CHECK: test_vld1q_f64
+; CHECK: ld1 {v{{[0-9]+}}.2d}, [{{x[0-9]+|sp}}]
+  %1 = bitcast double* %a to i8*
+  %vld1 = tail call <2 x double> @llvm.arm.neon.vld1.v2f64(i8* %1, i32 8)
+  ret <2 x double> %vld1
+}
+
+define <8 x i8> @test_vld1_s8(i8* readonly %a) {
+; CHECK: test_vld1_s8
+; CHECK: ld1 {v{{[0-9]+}}.8b}, [x{{[0-9]+|sp}}]
+  %vld1 = tail call <8 x i8> @llvm.arm.neon.vld1.v8i8(i8* %a, i32 1)
+  ret <8 x i8> %vld1
+}
+
+define <4 x i16> @test_vld1_s16(i16* readonly %a) {
+; CHECK: test_vld1_s16
+; CHECK: ld1 {v{{[0-9]+}}.4h}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i16* %a to i8*
+  %vld1 = tail call <4 x i16> @llvm.arm.neon.vld1.v4i16(i8* %1, i32 2)
+  ret <4 x i16> %vld1
+}
+
+define <2 x i32> @test_vld1_s32(i32* readonly %a) {
+; CHECK: test_vld1_s32
+; CHECK: ld1 {v{{[0-9]+}}.2s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i32* %a to i8*
+  %vld1 = tail call <2 x i32> @llvm.arm.neon.vld1.v2i32(i8* %1, i32 4)
+  ret <2 x i32> %vld1
+}
+
+define <1 x i64> @test_vld1_s64(i64* readonly %a) {
+; CHECK: test_vld1_s64
+; CHECK: ld1 {v{{[0-9]+}}.1d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i64* %a to i8*
+  %vld1 = tail call <1 x i64> @llvm.arm.neon.vld1.v1i64(i8* %1, i32 8)
+  ret <1 x i64> %vld1
+}
+
+define <2 x float> @test_vld1_f32(float* readonly %a) {
+; CHECK: test_vld1_f32
+; CHECK: ld1 {v{{[0-9]+}}.2s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast float* %a to i8*
+  %vld1 = tail call <2 x float> @llvm.arm.neon.vld1.v2f32(i8* %1, i32 4)
+  ret <2 x float> %vld1
+}
+
+define <1 x double> @test_vld1_f64(double* readonly %a) {
+; CHECK: test_vld1_f64
+; CHECK: ld1 {v{{[0-9]+}}.1d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast double* %a to i8*
+  %vld1 = tail call <1 x double> @llvm.arm.neon.vld1.v1f64(i8* %1, i32 8)
+  ret <1 x double> %vld1
+}
+
+define <8 x i8> @test_vld1_p8(i8* readonly %a) {
+; CHECK: test_vld1_p8
+; CHECK: ld1 {v{{[0-9]+}}.8b}, [x{{[0-9]+|sp}}]
+  %vld1 = tail call <8 x i8> @llvm.arm.neon.vld1.v8i8(i8* %a, i32 1)
+  ret <8 x i8> %vld1
+}
+
+define <4 x i16> @test_vld1_p16(i16* readonly %a) {
+; CHECK: test_vld1_p16
+; CHECK: ld1 {v{{[0-9]+}}.4h}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i16* %a to i8*
+  %vld1 = tail call <4 x i16> @llvm.arm.neon.vld1.v4i16(i8* %1, i32 2)
+  ret <4 x i16> %vld1
+}
+
+define %struct.int8x16x2_t @test_vld2q_s8(i8* readonly %a) {
+; CHECK: test_vld2q_s8
+; CHECK: ld2 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [x{{[0-9]+|sp}}]
+  %vld2 = tail call { <16 x i8>, <16 x i8> } @llvm.arm.neon.vld2.v16i8(i8* %a, i32 1)
+  %vld2.fca.0.extract = extractvalue { <16 x i8>, <16 x i8> } %vld2, 0
+  %vld2.fca.1.extract = extractvalue { <16 x i8>, <16 x i8> } %vld2, 1
+  %.fca.0.0.insert = insertvalue %struct.int8x16x2_t undef, <16 x i8> %vld2.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int8x16x2_t %.fca.0.0.insert, <16 x i8> %vld2.fca.1.extract, 0, 1
+  ret %struct.int8x16x2_t %.fca.0.1.insert
+}
+
+define %struct.int16x8x2_t @test_vld2q_s16(i16* readonly %a) {
+; CHECK: test_vld2q_s16
+; CHECK: ld2 {v{{[0-9]+}}.8h, v{{[0-9]+}}.8h}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i16* %a to i8*
+  %vld2 = tail call { <8 x i16>, <8 x i16> } @llvm.arm.neon.vld2.v8i16(i8* %1, i32 2)
+  %vld2.fca.0.extract = extractvalue { <8 x i16>, <8 x i16> } %vld2, 0
+  %vld2.fca.1.extract = extractvalue { <8 x i16>, <8 x i16> } %vld2, 1
+  %.fca.0.0.insert = insertvalue %struct.int16x8x2_t undef, <8 x i16> %vld2.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int16x8x2_t %.fca.0.0.insert, <8 x i16> %vld2.fca.1.extract, 0, 1
+  ret %struct.int16x8x2_t %.fca.0.1.insert
+}
+
+define %struct.int32x4x2_t @test_vld2q_s32(i32* readonly %a) {
+; CHECK: test_vld2q_s32
+; CHECK: ld2 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i32* %a to i8*
+  %vld2 = tail call { <4 x i32>, <4 x i32> } @llvm.arm.neon.vld2.v4i32(i8* %1, i32 4)
+  %vld2.fca.0.extract = extractvalue { <4 x i32>, <4 x i32> } %vld2, 0
+  %vld2.fca.1.extract = extractvalue { <4 x i32>, <4 x i32> } %vld2, 1
+  %.fca.0.0.insert = insertvalue %struct.int32x4x2_t undef, <4 x i32> %vld2.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int32x4x2_t %.fca.0.0.insert, <4 x i32> %vld2.fca.1.extract, 0, 1
+  ret %struct.int32x4x2_t %.fca.0.1.insert
+}
+
+define %struct.int64x2x2_t @test_vld2q_s64(i64* readonly %a) {
+; CHECK: test_vld2q_s64
+; CHECK: ld2 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i64* %a to i8*
+  %vld2 = tail call { <2 x i64>, <2 x i64> } @llvm.arm.neon.vld2.v2i64(i8* %1, i32 8)
+  %vld2.fca.0.extract = extractvalue { <2 x i64>, <2 x i64> } %vld2, 0
+  %vld2.fca.1.extract = extractvalue { <2 x i64>, <2 x i64> } %vld2, 1
+  %.fca.0.0.insert = insertvalue %struct.int64x2x2_t undef, <2 x i64> %vld2.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int64x2x2_t %.fca.0.0.insert, <2 x i64> %vld2.fca.1.extract, 0, 1
+  ret %struct.int64x2x2_t %.fca.0.1.insert
+}
+
+define %struct.float32x4x2_t @test_vld2q_f32(float* readonly %a) {
+; CHECK: test_vld2q_f32
+; CHECK: ld2 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast float* %a to i8*
+  %vld2 = tail call { <4 x float>, <4 x float> } @llvm.arm.neon.vld2.v4f32(i8* %1, i32 4)
+  %vld2.fca.0.extract = extractvalue { <4 x float>, <4 x float> } %vld2, 0
+  %vld2.fca.1.extract = extractvalue { <4 x float>, <4 x float> } %vld2, 1
+  %.fca.0.0.insert = insertvalue %struct.float32x4x2_t undef, <4 x float> %vld2.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.float32x4x2_t %.fca.0.0.insert, <4 x float> %vld2.fca.1.extract, 0, 1
+  ret %struct.float32x4x2_t %.fca.0.1.insert
+}
+
+define %struct.float64x2x2_t @test_vld2q_f64(double* readonly %a) {
+; CHECK: test_vld2q_f64
+; CHECK: ld2 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast double* %a to i8*
+  %vld2 = tail call { <2 x double>, <2 x double> } @llvm.arm.neon.vld2.v2f64(i8* %1, i32 8)
+  %vld2.fca.0.extract = extractvalue { <2 x double>, <2 x double> } %vld2, 0
+  %vld2.fca.1.extract = extractvalue { <2 x double>, <2 x double> } %vld2, 1
+  %.fca.0.0.insert = insertvalue %struct.float64x2x2_t undef, <2 x double> %vld2.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.float64x2x2_t %.fca.0.0.insert, <2 x double> %vld2.fca.1.extract, 0, 1
+  ret %struct.float64x2x2_t %.fca.0.1.insert
+}
+
+define %struct.int8x8x2_t @test_vld2_s8(i8* readonly %a) {
+; CHECK: test_vld2_s8
+; CHECK: ld2 {v{{[0-9]+}}.8b, v{{[0-9]+}}.8b}, [x{{[0-9]+|sp}}]
+  %vld2 = tail call { <8 x i8>, <8 x i8> } @llvm.arm.neon.vld2.v8i8(i8* %a, i32 1)
+  %vld2.fca.0.extract = extractvalue { <8 x i8>, <8 x i8> } %vld2, 0
+  %vld2.fca.1.extract = extractvalue { <8 x i8>, <8 x i8> } %vld2, 1
+  %.fca.0.0.insert = insertvalue %struct.int8x8x2_t undef, <8 x i8> %vld2.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int8x8x2_t %.fca.0.0.insert, <8 x i8> %vld2.fca.1.extract, 0, 1
+  ret %struct.int8x8x2_t %.fca.0.1.insert
+}
+
+define %struct.int16x4x2_t @test_vld2_s16(i16* readonly %a) {
+; CHECK: test_vld2_s16
+; CHECK: ld2 {v{{[0-9]+}}.4h, v{{[0-9]+}}.4h}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i16* %a to i8*
+  %vld2 = tail call { <4 x i16>, <4 x i16> } @llvm.arm.neon.vld2.v4i16(i8* %1, i32 2)
+  %vld2.fca.0.extract = extractvalue { <4 x i16>, <4 x i16> } %vld2, 0
+  %vld2.fca.1.extract = extractvalue { <4 x i16>, <4 x i16> } %vld2, 1
+  %.fca.0.0.insert = insertvalue %struct.int16x4x2_t undef, <4 x i16> %vld2.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int16x4x2_t %.fca.0.0.insert, <4 x i16> %vld2.fca.1.extract, 0, 1
+  ret %struct.int16x4x2_t %.fca.0.1.insert
+}
+
+define %struct.int32x2x2_t @test_vld2_s32(i32* readonly %a) {
+; CHECK: test_vld2_s32
+; CHECK: ld2 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i32* %a to i8*
+  %vld2 = tail call { <2 x i32>, <2 x i32> } @llvm.arm.neon.vld2.v2i32(i8* %1, i32 4)
+  %vld2.fca.0.extract = extractvalue { <2 x i32>, <2 x i32> } %vld2, 0
+  %vld2.fca.1.extract = extractvalue { <2 x i32>, <2 x i32> } %vld2, 1
+  %.fca.0.0.insert = insertvalue %struct.int32x2x2_t undef, <2 x i32> %vld2.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int32x2x2_t %.fca.0.0.insert, <2 x i32> %vld2.fca.1.extract, 0, 1
+  ret %struct.int32x2x2_t %.fca.0.1.insert
+}
+
+define %struct.int64x1x2_t @test_vld2_s64(i64* readonly %a) {
+; CHECK: test_vld2_s64
+; CHECK: ld1 {v{{[0-9]+}}.1d, v{{[0-9]+}}.1d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i64* %a to i8*
+  %vld2 = tail call { <1 x i64>, <1 x i64> } @llvm.arm.neon.vld2.v1i64(i8* %1, i32 8)
+  %vld2.fca.0.extract = extractvalue { <1 x i64>, <1 x i64> } %vld2, 0
+  %vld2.fca.1.extract = extractvalue { <1 x i64>, <1 x i64> } %vld2, 1
+  %.fca.0.0.insert = insertvalue %struct.int64x1x2_t undef, <1 x i64> %vld2.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int64x1x2_t %.fca.0.0.insert, <1 x i64> %vld2.fca.1.extract, 0, 1
+  ret %struct.int64x1x2_t %.fca.0.1.insert
+}
+
+define %struct.float32x2x2_t @test_vld2_f32(float* readonly %a) {
+; CHECK: test_vld2_f32
+; CHECK: ld2 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast float* %a to i8*
+  %vld2 = tail call { <2 x float>, <2 x float> } @llvm.arm.neon.vld2.v2f32(i8* %1, i32 4)
+  %vld2.fca.0.extract = extractvalue { <2 x float>, <2 x float> } %vld2, 0
+  %vld2.fca.1.extract = extractvalue { <2 x float>, <2 x float> } %vld2, 1
+  %.fca.0.0.insert = insertvalue %struct.float32x2x2_t undef, <2 x float> %vld2.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.float32x2x2_t %.fca.0.0.insert, <2 x float> %vld2.fca.1.extract, 0, 1
+  ret %struct.float32x2x2_t %.fca.0.1.insert
+}
+
+define %struct.float64x1x2_t @test_vld2_f64(double* readonly %a) {
+; CHECK: test_vld2_f64
+; CHECK: ld1 {v{{[0-9]+}}.1d, v{{[0-9]+}}.1d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast double* %a to i8*
+  %vld2 = tail call { <1 x double>, <1 x double> } @llvm.arm.neon.vld2.v1f64(i8* %1, i32 8)
+  %vld2.fca.0.extract = extractvalue { <1 x double>, <1 x double> } %vld2, 0
+  %vld2.fca.1.extract = extractvalue { <1 x double>, <1 x double> } %vld2, 1
+  %.fca.0.0.insert = insertvalue %struct.float64x1x2_t undef, <1 x double> %vld2.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.float64x1x2_t %.fca.0.0.insert, <1 x double> %vld2.fca.1.extract, 0, 1
+  ret %struct.float64x1x2_t %.fca.0.1.insert
+}
+
+define %struct.int8x16x3_t @test_vld3q_s8(i8* readonly %a) {
+; CHECK: test_vld3q_s8
+; CHECK: ld3 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [x{{[0-9]+|sp}}]
+  %vld3 = tail call { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.arm.neon.vld3.v16i8(i8* %a, i32 1)
+  %vld3.fca.0.extract = extractvalue { <16 x i8>, <16 x i8>, <16 x i8> } %vld3, 0
+  %vld3.fca.1.extract = extractvalue { <16 x i8>, <16 x i8>, <16 x i8> } %vld3, 1
+  %vld3.fca.2.extract = extractvalue { <16 x i8>, <16 x i8>, <16 x i8> } %vld3, 2
+  %.fca.0.0.insert = insertvalue %struct.int8x16x3_t undef, <16 x i8> %vld3.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int8x16x3_t %.fca.0.0.insert, <16 x i8> %vld3.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int8x16x3_t %.fca.0.1.insert, <16 x i8> %vld3.fca.2.extract, 0, 2
+  ret %struct.int8x16x3_t %.fca.0.2.insert
+}
+
+define %struct.int16x8x3_t @test_vld3q_s16(i16* readonly %a) {
+; CHECK: test_vld3q_s16
+; CHECK: ld3 {v{{[0-9]+}}.8h, v{{[0-9]+}}.8h, v{{[0-9]+}}.8h}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i16* %a to i8*
+  %vld3 = tail call { <8 x i16>, <8 x i16>, <8 x i16> } @llvm.arm.neon.vld3.v8i16(i8* %1, i32 2)
+  %vld3.fca.0.extract = extractvalue { <8 x i16>, <8 x i16>, <8 x i16> } %vld3, 0
+  %vld3.fca.1.extract = extractvalue { <8 x i16>, <8 x i16>, <8 x i16> } %vld3, 1
+  %vld3.fca.2.extract = extractvalue { <8 x i16>, <8 x i16>, <8 x i16> } %vld3, 2
+  %.fca.0.0.insert = insertvalue %struct.int16x8x3_t undef, <8 x i16> %vld3.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int16x8x3_t %.fca.0.0.insert, <8 x i16> %vld3.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int16x8x3_t %.fca.0.1.insert, <8 x i16> %vld3.fca.2.extract, 0, 2
+  ret %struct.int16x8x3_t %.fca.0.2.insert
+}
+
+define %struct.int32x4x3_t @test_vld3q_s32(i32* readonly %a) {
+; CHECK: test_vld3q_s32
+; CHECK: ld3 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i32* %a to i8*
+  %vld3 = tail call { <4 x i32>, <4 x i32>, <4 x i32> } @llvm.arm.neon.vld3.v4i32(i8* %1, i32 4)
+  %vld3.fca.0.extract = extractvalue { <4 x i32>, <4 x i32>, <4 x i32> } %vld3, 0
+  %vld3.fca.1.extract = extractvalue { <4 x i32>, <4 x i32>, <4 x i32> } %vld3, 1
+  %vld3.fca.2.extract = extractvalue { <4 x i32>, <4 x i32>, <4 x i32> } %vld3, 2
+  %.fca.0.0.insert = insertvalue %struct.int32x4x3_t undef, <4 x i32> %vld3.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int32x4x3_t %.fca.0.0.insert, <4 x i32> %vld3.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int32x4x3_t %.fca.0.1.insert, <4 x i32> %vld3.fca.2.extract, 0, 2
+  ret %struct.int32x4x3_t %.fca.0.2.insert
+}
+
+define %struct.int64x2x3_t @test_vld3q_s64(i64* readonly %a) {
+; CHECK: test_vld3q_s64
+; CHECK: ld3 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i64* %a to i8*
+  %vld3 = tail call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.arm.neon.vld3.v2i64(i8* %1, i32 8)
+  %vld3.fca.0.extract = extractvalue { <2 x i64>, <2 x i64>, <2 x i64> } %vld3, 0
+  %vld3.fca.1.extract = extractvalue { <2 x i64>, <2 x i64>, <2 x i64> } %vld3, 1
+  %vld3.fca.2.extract = extractvalue { <2 x i64>, <2 x i64>, <2 x i64> } %vld3, 2
+  %.fca.0.0.insert = insertvalue %struct.int64x2x3_t undef, <2 x i64> %vld3.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int64x2x3_t %.fca.0.0.insert, <2 x i64> %vld3.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int64x2x3_t %.fca.0.1.insert, <2 x i64> %vld3.fca.2.extract, 0, 2
+  ret %struct.int64x2x3_t %.fca.0.2.insert
+}
+
+define %struct.float32x4x3_t @test_vld3q_f32(float* readonly %a) {
+; CHECK: test_vld3q_f32
+; CHECK: ld3 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast float* %a to i8*
+  %vld3 = tail call { <4 x float>, <4 x float>, <4 x float> } @llvm.arm.neon.vld3.v4f32(i8* %1, i32 4)
+  %vld3.fca.0.extract = extractvalue { <4 x float>, <4 x float>, <4 x float> } %vld3, 0
+  %vld3.fca.1.extract = extractvalue { <4 x float>, <4 x float>, <4 x float> } %vld3, 1
+  %vld3.fca.2.extract = extractvalue { <4 x float>, <4 x float>, <4 x float> } %vld3, 2
+  %.fca.0.0.insert = insertvalue %struct.float32x4x3_t undef, <4 x float> %vld3.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.float32x4x3_t %.fca.0.0.insert, <4 x float> %vld3.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.float32x4x3_t %.fca.0.1.insert, <4 x float> %vld3.fca.2.extract, 0, 2
+  ret %struct.float32x4x3_t %.fca.0.2.insert
+}
+
+define %struct.float64x2x3_t @test_vld3q_f64(double* readonly %a) {
+; CHECK: test_vld3q_f64
+; CHECK: ld3 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast double* %a to i8*
+  %vld3 = tail call { <2 x double>, <2 x double>, <2 x double> } @llvm.arm.neon.vld3.v2f64(i8* %1, i32 8)
+  %vld3.fca.0.extract = extractvalue { <2 x double>, <2 x double>, <2 x double> } %vld3, 0
+  %vld3.fca.1.extract = extractvalue { <2 x double>, <2 x double>, <2 x double> } %vld3, 1
+  %vld3.fca.2.extract = extractvalue { <2 x double>, <2 x double>, <2 x double> } %vld3, 2
+  %.fca.0.0.insert = insertvalue %struct.float64x2x3_t undef, <2 x double> %vld3.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.float64x2x3_t %.fca.0.0.insert, <2 x double> %vld3.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.float64x2x3_t %.fca.0.1.insert, <2 x double> %vld3.fca.2.extract, 0, 2
+  ret %struct.float64x2x3_t %.fca.0.2.insert
+}
+
+define %struct.int8x8x3_t @test_vld3_s8(i8* readonly %a) {
+; CHECK: test_vld3_s8
+; CHECK: ld3 {v{{[0-9]+}}.8b, v{{[0-9]+}}.8b, v{{[0-9]+}}.8b}, [x{{[0-9]+|sp}}]
+  %vld3 = tail call { <8 x i8>, <8 x i8>, <8 x i8> } @llvm.arm.neon.vld3.v8i8(i8* %a, i32 1)
+  %vld3.fca.0.extract = extractvalue { <8 x i8>, <8 x i8>, <8 x i8> } %vld3, 0
+  %vld3.fca.1.extract = extractvalue { <8 x i8>, <8 x i8>, <8 x i8> } %vld3, 1
+  %vld3.fca.2.extract = extractvalue { <8 x i8>, <8 x i8>, <8 x i8> } %vld3, 2
+  %.fca.0.0.insert = insertvalue %struct.int8x8x3_t undef, <8 x i8> %vld3.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int8x8x3_t %.fca.0.0.insert, <8 x i8> %vld3.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int8x8x3_t %.fca.0.1.insert, <8 x i8> %vld3.fca.2.extract, 0, 2
+  ret %struct.int8x8x3_t %.fca.0.2.insert
+}
+
+define %struct.int16x4x3_t @test_vld3_s16(i16* readonly %a) {
+; CHECK: test_vld3_s16
+; CHECK: ld3 {v{{[0-9]+}}.4h, v{{[0-9]+}}.4h, v{{[0-9]+}}.4h}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i16* %a to i8*
+  %vld3 = tail call { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.arm.neon.vld3.v4i16(i8* %1, i32 2)
+  %vld3.fca.0.extract = extractvalue { <4 x i16>, <4 x i16>, <4 x i16> } %vld3, 0
+  %vld3.fca.1.extract = extractvalue { <4 x i16>, <4 x i16>, <4 x i16> } %vld3, 1
+  %vld3.fca.2.extract = extractvalue { <4 x i16>, <4 x i16>, <4 x i16> } %vld3, 2
+  %.fca.0.0.insert = insertvalue %struct.int16x4x3_t undef, <4 x i16> %vld3.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int16x4x3_t %.fca.0.0.insert, <4 x i16> %vld3.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int16x4x3_t %.fca.0.1.insert, <4 x i16> %vld3.fca.2.extract, 0, 2
+  ret %struct.int16x4x3_t %.fca.0.2.insert
+}
+
+define %struct.int32x2x3_t @test_vld3_s32(i32* readonly %a) {
+; CHECK: test_vld3_s32
+; CHECK: ld3 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i32* %a to i8*
+  %vld3 = tail call { <2 x i32>, <2 x i32>, <2 x i32> } @llvm.arm.neon.vld3.v2i32(i8* %1, i32 4)
+  %vld3.fca.0.extract = extractvalue { <2 x i32>, <2 x i32>, <2 x i32> } %vld3, 0
+  %vld3.fca.1.extract = extractvalue { <2 x i32>, <2 x i32>, <2 x i32> } %vld3, 1
+  %vld3.fca.2.extract = extractvalue { <2 x i32>, <2 x i32>, <2 x i32> } %vld3, 2
+  %.fca.0.0.insert = insertvalue %struct.int32x2x3_t undef, <2 x i32> %vld3.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int32x2x3_t %.fca.0.0.insert, <2 x i32> %vld3.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int32x2x3_t %.fca.0.1.insert, <2 x i32> %vld3.fca.2.extract, 0, 2
+  ret %struct.int32x2x3_t %.fca.0.2.insert
+}
+
+define %struct.int64x1x3_t @test_vld3_s64(i64* readonly %a) {
+; CHECK: test_vld3_s64
+; CHECK: ld1 {v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i64* %a to i8*
+  %vld3 = tail call { <1 x i64>, <1 x i64>, <1 x i64> } @llvm.arm.neon.vld3.v1i64(i8* %1, i32 8)
+  %vld3.fca.0.extract = extractvalue { <1 x i64>, <1 x i64>, <1 x i64> } %vld3, 0
+  %vld3.fca.1.extract = extractvalue { <1 x i64>, <1 x i64>, <1 x i64> } %vld3, 1
+  %vld3.fca.2.extract = extractvalue { <1 x i64>, <1 x i64>, <1 x i64> } %vld3, 2
+  %.fca.0.0.insert = insertvalue %struct.int64x1x3_t undef, <1 x i64> %vld3.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int64x1x3_t %.fca.0.0.insert, <1 x i64> %vld3.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int64x1x3_t %.fca.0.1.insert, <1 x i64> %vld3.fca.2.extract, 0, 2
+  ret %struct.int64x1x3_t %.fca.0.2.insert
+}
+
+define %struct.float32x2x3_t @test_vld3_f32(float* readonly %a) {
+; CHECK: test_vld3_f32
+; CHECK: ld3 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast float* %a to i8*
+  %vld3 = tail call { <2 x float>, <2 x float>, <2 x float> } @llvm.arm.neon.vld3.v2f32(i8* %1, i32 4)
+  %vld3.fca.0.extract = extractvalue { <2 x float>, <2 x float>, <2 x float> } %vld3, 0
+  %vld3.fca.1.extract = extractvalue { <2 x float>, <2 x float>, <2 x float> } %vld3, 1
+  %vld3.fca.2.extract = extractvalue { <2 x float>, <2 x float>, <2 x float> } %vld3, 2
+  %.fca.0.0.insert = insertvalue %struct.float32x2x3_t undef, <2 x float> %vld3.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.float32x2x3_t %.fca.0.0.insert, <2 x float> %vld3.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.float32x2x3_t %.fca.0.1.insert, <2 x float> %vld3.fca.2.extract, 0, 2
+  ret %struct.float32x2x3_t %.fca.0.2.insert
+}
+
+define %struct.float64x1x3_t @test_vld3_f64(double* readonly %a) {
+; CHECK: test_vld3_f64
+; CHECK: ld1 {v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast double* %a to i8*
+  %vld3 = tail call { <1 x double>, <1 x double>, <1 x double> } @llvm.arm.neon.vld3.v1f64(i8* %1, i32 8)
+  %vld3.fca.0.extract = extractvalue { <1 x double>, <1 x double>, <1 x double> } %vld3, 0
+  %vld3.fca.1.extract = extractvalue { <1 x double>, <1 x double>, <1 x double> } %vld3, 1
+  %vld3.fca.2.extract = extractvalue { <1 x double>, <1 x double>, <1 x double> } %vld3, 2
+  %.fca.0.0.insert = insertvalue %struct.float64x1x3_t undef, <1 x double> %vld3.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.float64x1x3_t %.fca.0.0.insert, <1 x double> %vld3.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.float64x1x3_t %.fca.0.1.insert, <1 x double> %vld3.fca.2.extract, 0, 2
+  ret %struct.float64x1x3_t %.fca.0.2.insert
+}
+
+define %struct.int8x16x4_t @test_vld4q_s8(i8* readonly %a) {
+; CHECK: test_vld4q_s8
+; CHECK: ld4 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [x{{[0-9]+|sp}}]
+  %vld4 = tail call { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.arm.neon.vld4.v16i8(i8* %a, i32 1)
+  %vld4.fca.0.extract = extractvalue { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } %vld4, 0
+  %vld4.fca.1.extract = extractvalue { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } %vld4, 1
+  %vld4.fca.2.extract = extractvalue { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } %vld4, 2
+  %vld4.fca.3.extract = extractvalue { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } %vld4, 3
+  %.fca.0.0.insert = insertvalue %struct.int8x16x4_t undef, <16 x i8> %vld4.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int8x16x4_t %.fca.0.0.insert, <16 x i8> %vld4.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int8x16x4_t %.fca.0.1.insert, <16 x i8> %vld4.fca.2.extract, 0, 2
+  %.fca.0.3.insert = insertvalue %struct.int8x16x4_t %.fca.0.2.insert, <16 x i8> %vld4.fca.3.extract, 0, 3
+  ret %struct.int8x16x4_t %.fca.0.3.insert
+}
+
+define %struct.int16x8x4_t @test_vld4q_s16(i16* readonly %a) {
+; CHECK: test_vld4q_s16
+; CHECK: ld4 {v{{[0-9]+}}.8h, v{{[0-9]+}}.8h, v{{[0-9]+}}.8h, v{{[0-9]+}}.8h}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i16* %a to i8*
+  %vld4 = tail call { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } @llvm.arm.neon.vld4.v8i16(i8* %1, i32 2)
+  %vld4.fca.0.extract = extractvalue { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } %vld4, 0
+  %vld4.fca.1.extract = extractvalue { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } %vld4, 1
+  %vld4.fca.2.extract = extractvalue { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } %vld4, 2
+  %vld4.fca.3.extract = extractvalue { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } %vld4, 3
+  %.fca.0.0.insert = insertvalue %struct.int16x8x4_t undef, <8 x i16> %vld4.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int16x8x4_t %.fca.0.0.insert, <8 x i16> %vld4.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int16x8x4_t %.fca.0.1.insert, <8 x i16> %vld4.fca.2.extract, 0, 2
+  %.fca.0.3.insert = insertvalue %struct.int16x8x4_t %.fca.0.2.insert, <8 x i16> %vld4.fca.3.extract, 0, 3
+  ret %struct.int16x8x4_t %.fca.0.3.insert
+}
+
+define %struct.int32x4x4_t @test_vld4q_s32(i32* readonly %a) {
+; CHECK: test_vld4q_s32
+; CHECK: ld4 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i32* %a to i8*
+  %vld4 = tail call { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } @llvm.arm.neon.vld4.v4i32(i8* %1, i32 4)
+  %vld4.fca.0.extract = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } %vld4, 0
+  %vld4.fca.1.extract = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } %vld4, 1
+  %vld4.fca.2.extract = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } %vld4, 2
+  %vld4.fca.3.extract = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } %vld4, 3
+  %.fca.0.0.insert = insertvalue %struct.int32x4x4_t undef, <4 x i32> %vld4.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int32x4x4_t %.fca.0.0.insert, <4 x i32> %vld4.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int32x4x4_t %.fca.0.1.insert, <4 x i32> %vld4.fca.2.extract, 0, 2
+  %.fca.0.3.insert = insertvalue %struct.int32x4x4_t %.fca.0.2.insert, <4 x i32> %vld4.fca.3.extract, 0, 3
+  ret %struct.int32x4x4_t %.fca.0.3.insert
+}
+
+define %struct.int64x2x4_t @test_vld4q_s64(i64* readonly %a) {
+; CHECK: test_vld4q_s64
+; CHECK: ld4 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i64* %a to i8*
+  %vld4 = tail call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.arm.neon.vld4.v2i64(i8* %1, i32 8)
+  %vld4.fca.0.extract = extractvalue { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %vld4, 0
+  %vld4.fca.1.extract = extractvalue { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %vld4, 1
+  %vld4.fca.2.extract = extractvalue { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %vld4, 2
+  %vld4.fca.3.extract = extractvalue { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } %vld4, 3
+  %.fca.0.0.insert = insertvalue %struct.int64x2x4_t undef, <2 x i64> %vld4.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int64x2x4_t %.fca.0.0.insert, <2 x i64> %vld4.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int64x2x4_t %.fca.0.1.insert, <2 x i64> %vld4.fca.2.extract, 0, 2
+  %.fca.0.3.insert = insertvalue %struct.int64x2x4_t %.fca.0.2.insert, <2 x i64> %vld4.fca.3.extract, 0, 3
+  ret %struct.int64x2x4_t %.fca.0.3.insert
+}
+
+define %struct.float32x4x4_t @test_vld4q_f32(float* readonly %a) {
+; CHECK: test_vld4q_f32
+; CHECK: ld4 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast float* %a to i8*
+  %vld4 = tail call { <4 x float>, <4 x float>, <4 x float>, <4 x float> } @llvm.arm.neon.vld4.v4f32(i8* %1, i32 4)
+  %vld4.fca.0.extract = extractvalue { <4 x float>, <4 x float>, <4 x float>, <4 x float> } %vld4, 0
+  %vld4.fca.1.extract = extractvalue { <4 x float>, <4 x float>, <4 x float>, <4 x float> } %vld4, 1
+  %vld4.fca.2.extract = extractvalue { <4 x float>, <4 x float>, <4 x float>, <4 x float> } %vld4, 2
+  %vld4.fca.3.extract = extractvalue { <4 x float>, <4 x float>, <4 x float>, <4 x float> } %vld4, 3
+  %.fca.0.0.insert = insertvalue %struct.float32x4x4_t undef, <4 x float> %vld4.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.float32x4x4_t %.fca.0.0.insert, <4 x float> %vld4.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.float32x4x4_t %.fca.0.1.insert, <4 x float> %vld4.fca.2.extract, 0, 2
+  %.fca.0.3.insert = insertvalue %struct.float32x4x4_t %.fca.0.2.insert, <4 x float> %vld4.fca.3.extract, 0, 3
+  ret %struct.float32x4x4_t %.fca.0.3.insert
+}
+
+define %struct.float64x2x4_t @test_vld4q_f64(double* readonly %a) {
+; CHECK: test_vld4q_f64
+; CHECK: ld4 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast double* %a to i8*
+  %vld4 = tail call { <2 x double>, <2 x double>, <2 x double>, <2 x double> } @llvm.arm.neon.vld4.v2f64(i8* %1, i32 8)
+  %vld4.fca.0.extract = extractvalue { <2 x double>, <2 x double>, <2 x double>, <2 x double> } %vld4, 0
+  %vld4.fca.1.extract = extractvalue { <2 x double>, <2 x double>, <2 x double>, <2 x double> } %vld4, 1
+  %vld4.fca.2.extract = extractvalue { <2 x double>, <2 x double>, <2 x double>, <2 x double> } %vld4, 2
+  %vld4.fca.3.extract = extractvalue { <2 x double>, <2 x double>, <2 x double>, <2 x double> } %vld4, 3
+  %.fca.0.0.insert = insertvalue %struct.float64x2x4_t undef, <2 x double> %vld4.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.float64x2x4_t %.fca.0.0.insert, <2 x double> %vld4.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.float64x2x4_t %.fca.0.1.insert, <2 x double> %vld4.fca.2.extract, 0, 2
+  %.fca.0.3.insert = insertvalue %struct.float64x2x4_t %.fca.0.2.insert, <2 x double> %vld4.fca.3.extract, 0, 3
+  ret %struct.float64x2x4_t %.fca.0.3.insert
+}
+
+define %struct.int8x8x4_t @test_vld4_s8(i8* readonly %a) {
+; CHECK: test_vld4_s8
+; CHECK: ld4 {v{{[0-9]+}}.8b, v{{[0-9]+}}.8b, v{{[0-9]+}}.8b, v{{[0-9]+}}.8b}, [x{{[0-9]+|sp}}]
+  %vld4 = tail call { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.arm.neon.vld4.v8i8(i8* %a, i32 1)
+  %vld4.fca.0.extract = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %vld4, 0
+  %vld4.fca.1.extract = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %vld4, 1
+  %vld4.fca.2.extract = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %vld4, 2
+  %vld4.fca.3.extract = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %vld4, 3
+  %.fca.0.0.insert = insertvalue %struct.int8x8x4_t undef, <8 x i8> %vld4.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int8x8x4_t %.fca.0.0.insert, <8 x i8> %vld4.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int8x8x4_t %.fca.0.1.insert, <8 x i8> %vld4.fca.2.extract, 0, 2
+  %.fca.0.3.insert = insertvalue %struct.int8x8x4_t %.fca.0.2.insert, <8 x i8> %vld4.fca.3.extract, 0, 3
+  ret %struct.int8x8x4_t %.fca.0.3.insert
+}
+
+define %struct.int16x4x4_t @test_vld4_s16(i16* readonly %a) {
+; CHECK: test_vld4_s16
+; CHECK: ld4 {v{{[0-9]+}}.4h, v{{[0-9]+}}.4h, v{{[0-9]+}}.4h, v{{[0-9]+}}.4h}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i16* %a to i8*
+  %vld4 = tail call { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.arm.neon.vld4.v4i16(i8* %1, i32 2)
+  %vld4.fca.0.extract = extractvalue { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } %vld4, 0
+  %vld4.fca.1.extract = extractvalue { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } %vld4, 1
+  %vld4.fca.2.extract = extractvalue { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } %vld4, 2
+  %vld4.fca.3.extract = extractvalue { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } %vld4, 3
+  %.fca.0.0.insert = insertvalue %struct.int16x4x4_t undef, <4 x i16> %vld4.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int16x4x4_t %.fca.0.0.insert, <4 x i16> %vld4.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int16x4x4_t %.fca.0.1.insert, <4 x i16> %vld4.fca.2.extract, 0, 2
+  %.fca.0.3.insert = insertvalue %struct.int16x4x4_t %.fca.0.2.insert, <4 x i16> %vld4.fca.3.extract, 0, 3
+  ret %struct.int16x4x4_t %.fca.0.3.insert
+}
+
+define %struct.int32x2x4_t @test_vld4_s32(i32* readonly %a) {
+; CHECK: test_vld4_s32
+; CHECK: ld4 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s, v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i32* %a to i8*
+  %vld4 = tail call { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } @llvm.arm.neon.vld4.v2i32(i8* %1, i32 4)
+  %vld4.fca.0.extract = extractvalue { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } %vld4, 0
+  %vld4.fca.1.extract = extractvalue { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } %vld4, 1
+  %vld4.fca.2.extract = extractvalue { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } %vld4, 2
+  %vld4.fca.3.extract = extractvalue { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } %vld4, 3
+  %.fca.0.0.insert = insertvalue %struct.int32x2x4_t undef, <2 x i32> %vld4.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int32x2x4_t %.fca.0.0.insert, <2 x i32> %vld4.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int32x2x4_t %.fca.0.1.insert, <2 x i32> %vld4.fca.2.extract, 0, 2
+  %.fca.0.3.insert = insertvalue %struct.int32x2x4_t %.fca.0.2.insert, <2 x i32> %vld4.fca.3.extract, 0, 3
+  ret %struct.int32x2x4_t %.fca.0.3.insert
+}
+
+define %struct.int64x1x4_t @test_vld4_s64(i64* readonly %a) {
+; CHECK: test_vld4_s64
+; CHECK: ld1 {v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast i64* %a to i8*
+  %vld4 = tail call { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } @llvm.arm.neon.vld4.v1i64(i8* %1, i32 8)
+  %vld4.fca.0.extract = extractvalue { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } %vld4, 0
+  %vld4.fca.1.extract = extractvalue { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } %vld4, 1
+  %vld4.fca.2.extract = extractvalue { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } %vld4, 2
+  %vld4.fca.3.extract = extractvalue { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } %vld4, 3
+  %.fca.0.0.insert = insertvalue %struct.int64x1x4_t undef, <1 x i64> %vld4.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.int64x1x4_t %.fca.0.0.insert, <1 x i64> %vld4.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.int64x1x4_t %.fca.0.1.insert, <1 x i64> %vld4.fca.2.extract, 0, 2
+  %.fca.0.3.insert = insertvalue %struct.int64x1x4_t %.fca.0.2.insert, <1 x i64> %vld4.fca.3.extract, 0, 3
+  ret %struct.int64x1x4_t %.fca.0.3.insert
+}
+
+define %struct.float32x2x4_t @test_vld4_f32(float* readonly %a) {
+; CHECK: test_vld4_f32
+; CHECK: ld4 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s, v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [x{{[0-9]+|sp}}]
+  %1 = bitcast float* %a to i8*
+  %vld4 = tail call { <2 x float>, <2 x float>, <2 x float>, <2 x float> } @llvm.arm.neon.vld4.v2f32(i8* %1, i32 4)
+  %vld4.fca.0.extract = extractvalue { <2 x float>, <2 x float>, <2 x float>, <2 x float> } %vld4, 0
+  %vld4.fca.1.extract = extractvalue { <2 x float>, <2 x float>, <2 x float>, <2 x float> } %vld4, 1
+  %vld4.fca.2.extract = extractvalue { <2 x float>, <2 x float>, <2 x float>, <2 x float> } %vld4, 2
+  %vld4.fca.3.extract = extractvalue { <2 x float>, <2 x float>, <2 x float>, <2 x float> } %vld4, 3
+  %.fca.0.0.insert = insertvalue %struct.float32x2x4_t undef, <2 x float> %vld4.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.float32x2x4_t %.fca.0.0.insert, <2 x float> %vld4.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.float32x2x4_t %.fca.0.1.insert, <2 x float> %vld4.fca.2.extract, 0, 2
+  %.fca.0.3.insert = insertvalue %struct.float32x2x4_t %.fca.0.2.insert, <2 x float> %vld4.fca.3.extract, 0, 3
+  ret %struct.float32x2x4_t %.fca.0.3.insert
+}
+
+define %struct.float64x1x4_t @test_vld4_f64(double* readonly %a) {
+; CHECK: test_vld4_f64
+; CHECK: ld1 {v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d}, [x{{[0-9]+|sp}}]
+  %1 = bitcast double* %a to i8*
+  %vld4 = tail call { <1 x double>, <1 x double>, <1 x double>, <1 x double> } @llvm.arm.neon.vld4.v1f64(i8* %1, i32 8)
+  %vld4.fca.0.extract = extractvalue { <1 x double>, <1 x double>, <1 x double>, <1 x double> } %vld4, 0
+  %vld4.fca.1.extract = extractvalue { <1 x double>, <1 x double>, <1 x double>, <1 x double> } %vld4, 1
+  %vld4.fca.2.extract = extractvalue { <1 x double>, <1 x double>, <1 x double>, <1 x double> } %vld4, 2
+  %vld4.fca.3.extract = extractvalue { <1 x double>, <1 x double>, <1 x double>, <1 x double> } %vld4, 3
+  %.fca.0.0.insert = insertvalue %struct.float64x1x4_t undef, <1 x double> %vld4.fca.0.extract, 0, 0
+  %.fca.0.1.insert = insertvalue %struct.float64x1x4_t %.fca.0.0.insert, <1 x double> %vld4.fca.1.extract, 0, 1
+  %.fca.0.2.insert = insertvalue %struct.float64x1x4_t %.fca.0.1.insert, <1 x double> %vld4.fca.2.extract, 0, 2
+  %.fca.0.3.insert = insertvalue %struct.float64x1x4_t %.fca.0.2.insert, <1 x double> %vld4.fca.3.extract, 0, 3
+  ret %struct.float64x1x4_t %.fca.0.3.insert
+}
+
+declare <16 x i8> @llvm.arm.neon.vld1.v16i8(i8*, i32)
+declare <8 x i16> @llvm.arm.neon.vld1.v8i16(i8*, i32)
+declare <4 x i32> @llvm.arm.neon.vld1.v4i32(i8*, i32)
+declare <2 x i64> @llvm.arm.neon.vld1.v2i64(i8*, i32)
+declare <4 x float> @llvm.arm.neon.vld1.v4f32(i8*, i32)
+declare <2 x double> @llvm.arm.neon.vld1.v2f64(i8*, i32)
+declare <8 x i8> @llvm.arm.neon.vld1.v8i8(i8*, i32)
+declare <4 x i16> @llvm.arm.neon.vld1.v4i16(i8*, i32)
+declare <2 x i32> @llvm.arm.neon.vld1.v2i32(i8*, i32)
+declare <1 x i64> @llvm.arm.neon.vld1.v1i64(i8*, i32)
+declare <2 x float> @llvm.arm.neon.vld1.v2f32(i8*, i32)
+declare <1 x double> @llvm.arm.neon.vld1.v1f64(i8*, i32)
+declare { <16 x i8>, <16 x i8> } @llvm.arm.neon.vld2.v16i8(i8*, i32)
+declare { <8 x i16>, <8 x i16> } @llvm.arm.neon.vld2.v8i16(i8*, i32)
+declare { <4 x i32>, <4 x i32> } @llvm.arm.neon.vld2.v4i32(i8*, i32)
+declare { <2 x i64>, <2 x i64> } @llvm.arm.neon.vld2.v2i64(i8*, i32)
+declare { <4 x float>, <4 x float> } @llvm.arm.neon.vld2.v4f32(i8*, i32)
+declare { <2 x double>, <2 x double> } @llvm.arm.neon.vld2.v2f64(i8*, i32)
+declare { <8 x i8>, <8 x i8> } @llvm.arm.neon.vld2.v8i8(i8*, i32)
+declare { <4 x i16>, <4 x i16> } @llvm.arm.neon.vld2.v4i16(i8*, i32)
+declare { <2 x i32>, <2 x i32> } @llvm.arm.neon.vld2.v2i32(i8*, i32)
+declare { <1 x i64>, <1 x i64> } @llvm.arm.neon.vld2.v1i64(i8*, i32)
+declare { <2 x float>, <2 x float> } @llvm.arm.neon.vld2.v2f32(i8*, i32)
+declare { <1 x double>, <1 x double> } @llvm.arm.neon.vld2.v1f64(i8*, i32)
+declare { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.arm.neon.vld3.v16i8(i8*, i32)
+declare { <8 x i16>, <8 x i16>, <8 x i16> } @llvm.arm.neon.vld3.v8i16(i8*, i32)
+declare { <4 x i32>, <4 x i32>, <4 x i32> } @llvm.arm.neon.vld3.v4i32(i8*, i32)
+declare { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.arm.neon.vld3.v2i64(i8*, i32)
+declare { <4 x float>, <4 x float>, <4 x float> } @llvm.arm.neon.vld3.v4f32(i8*, i32)
+declare { <2 x double>, <2 x double>, <2 x double> } @llvm.arm.neon.vld3.v2f64(i8*, i32)
+declare { <8 x i8>, <8 x i8>, <8 x i8> } @llvm.arm.neon.vld3.v8i8(i8*, i32)
+declare { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.arm.neon.vld3.v4i16(i8*, i32)
+declare { <2 x i32>, <2 x i32>, <2 x i32> } @llvm.arm.neon.vld3.v2i32(i8*, i32)
+declare { <1 x i64>, <1 x i64>, <1 x i64> } @llvm.arm.neon.vld3.v1i64(i8*, i32)
+declare { <2 x float>, <2 x float>, <2 x float> } @llvm.arm.neon.vld3.v2f32(i8*, i32)
+declare { <1 x double>, <1 x double>, <1 x double> } @llvm.arm.neon.vld3.v1f64(i8*, i32)
+declare { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.arm.neon.vld4.v16i8(i8*, i32)
+declare { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } @llvm.arm.neon.vld4.v8i16(i8*, i32)
+declare { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } @llvm.arm.neon.vld4.v4i32(i8*, i32)
+declare { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.arm.neon.vld4.v2i64(i8*, i32)
+declare { <4 x float>, <4 x float>, <4 x float>, <4 x float> } @llvm.arm.neon.vld4.v4f32(i8*, i32)
+declare { <2 x double>, <2 x double>, <2 x double>, <2 x double> } @llvm.arm.neon.vld4.v2f64(i8*, i32)
+declare { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.arm.neon.vld4.v8i8(i8*, i32)
+declare { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.arm.neon.vld4.v4i16(i8*, i32)
+declare { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } @llvm.arm.neon.vld4.v2i32(i8*, i32)
+declare { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } @llvm.arm.neon.vld4.v1i64(i8*, i32)
+declare { <2 x float>, <2 x float>, <2 x float>, <2 x float> } @llvm.arm.neon.vld4.v2f32(i8*, i32)
+declare { <1 x double>, <1 x double>, <1 x double>, <1 x double> } @llvm.arm.neon.vld4.v1f64(i8*, i32)
+
+define void @test_vst1q_s8(i8* %a, <16 x i8> %b) {
+; CHECK: test_vst1q_s8
+; CHECK: st1 {v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
+  tail call void @llvm.arm.neon.vst1.v16i8(i8* %a, <16 x i8> %b, i32 1)
+  ret void
+}
+
+define void @test_vst1q_s16(i16* %a, <8 x i16> %b) {
+; CHECK: test_vst1q_s16
+; CHECK: st1 {v{{[0-9]+}}.8h}, [{{x[0-9]+|sp}}]
+  %1 = bitcast i16* %a to i8*
+  tail call void @llvm.arm.neon.vst1.v8i16(i8* %1, <8 x i16> %b, i32 2)
+  ret void
+}
+
+define void @test_vst1q_s32(i32* %a, <4 x i32> %b) {
+; CHECK: test_vst1q_s32
+; CHECK: st1 {v{{[0-9]+}}.4s}, [{{x[0-9]+|sp}}]
+  %1 = bitcast i32* %a to i8*
+  tail call void @llvm.arm.neon.vst1.v4i32(i8* %1, <4 x i32> %b, i32 4)
+  ret void
+}
+
+define void @test_vst1q_s64(i64* %a, <2 x i64> %b) {
+; CHECK: test_vst1q_s64
+; CHECK: st1 {v{{[0-9]+}}.2d}, [{{x[0-9]+|sp}}]
+  %1 = bitcast i64* %a to i8*
+  tail call void @llvm.arm.neon.vst1.v2i64(i8* %1, <2 x i64> %b, i32 8)
+  ret void
+}
+
+define void @test_vst1q_f32(float* %a, <4 x float> %b) {
+; CHECK: test_vst1q_f32
+; CHECK: st1 {v{{[0-9]+}}.4s}, [{{x[0-9]+|sp}}]
+  %1 = bitcast float* %a to i8*
+  tail call void @llvm.arm.neon.vst1.v4f32(i8* %1, <4 x float> %b, i32 4)
+  ret void
+}
+
+define void @test_vst1q_f64(double* %a, <2 x double> %b) {
+; CHECK: test_vst1q_f64
+; CHECK: st1 {v{{[0-9]+}}.2d}, [{{x[0-9]+|sp}}]
+  %1 = bitcast double* %a to i8*
+  tail call void @llvm.arm.neon.vst1.v2f64(i8* %1, <2 x double> %b, i32 8)
+  ret void
+}
+
+define void @test_vst1_s8(i8* %a, <8 x i8> %b) {
+; CHECK: test_vst1_s8
+; CHECK: st1 {v{{[0-9]+}}.8b}, [{{x[0-9]+|sp}}]
+  tail call void @llvm.arm.neon.vst1.v8i8(i8* %a, <8 x i8> %b, i32 1)
+  ret void
+}
+
+define void @test_vst1_s16(i16* %a, <4 x i16> %b) {
+; CHECK: test_vst1_s16
+; CHECK: st1 {v{{[0-9]+}}.4h}, [{{x[0-9]+|sp}}]
+  %1 = bitcast i16* %a to i8*
+  tail call void @llvm.arm.neon.vst1.v4i16(i8* %1, <4 x i16> %b, i32 2)
+  ret void
+}
+
+define void @test_vst1_s32(i32* %a, <2 x i32> %b) {
+; CHECK: test_vst1_s32
+; CHECK: st1 {v{{[0-9]+}}.2s}, [{{x[0-9]+|sp}}]
+  %1 = bitcast i32* %a to i8*
+  tail call void @llvm.arm.neon.vst1.v2i32(i8* %1, <2 x i32> %b, i32 4)
+  ret void
+}
+
+define void @test_vst1_s64(i64* %a, <1 x i64> %b) {
+; CHECK: test_vst1_s64
+; CHECK: st1 {v{{[0-9]+}}.1d}, [{{x[0-9]+|sp}}]
+  %1 = bitcast i64* %a to i8*
+  tail call void @llvm.arm.neon.vst1.v1i64(i8* %1, <1 x i64> %b, i32 8)
+  ret void
+}
+
+define void @test_vst1_f32(float* %a, <2 x float> %b) {
+; CHECK: test_vst1_f32
+; CHECK: st1 {v{{[0-9]+}}.2s}, [{{x[0-9]+|sp}}]
+  %1 = bitcast float* %a to i8*
+  tail call void @llvm.arm.neon.vst1.v2f32(i8* %1, <2 x float> %b, i32 4)
+  ret void
+}
+
+define void @test_vst1_f64(double* %a, <1 x double> %b) {
+; CHECK: test_vst1_f64
+; CHECK: st1 {v{{[0-9]+}}.1d}, [{{x[0-9]+|sp}}]
+  %1 = bitcast double* %a to i8*
+  tail call void @llvm.arm.neon.vst1.v1f64(i8* %1, <1 x double> %b, i32 8)
+  ret void
+}
+
+define void @test_vst2q_s8(i8* %a, [2 x <16 x i8>] %b.coerce) {
+; CHECK: test_vst2q_s8
+; CHECK: st2 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [2 x <16 x i8>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [2 x <16 x i8>] %b.coerce, 1
+  tail call void @llvm.arm.neon.vst2.v16i8(i8* %a, <16 x i8> %b.coerce.fca.0.extract, <16 x i8> %b.coerce.fca.1.extract, i32 1)
+  ret void
+}
+
+define void @test_vst2q_s16(i16* %a, [2 x <8 x i16>] %b.coerce) {
+; CHECK: test_vst2q_s16
+; CHECK: st2 {v{{[0-9]+}}.8h, v{{[0-9]+}}.8h}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [2 x <8 x i16>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [2 x <8 x i16>] %b.coerce, 1
+  %1 = bitcast i16* %a to i8*
+  tail call void @llvm.arm.neon.vst2.v8i16(i8* %1, <8 x i16> %b.coerce.fca.0.extract, <8 x i16> %b.coerce.fca.1.extract, i32 2)
+  ret void
+}
+
+define void @test_vst2q_s32(i32* %a, [2 x <4 x i32>] %b.coerce) {
+; CHECK: test_vst2q_s32
+; CHECK: st2 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [2 x <4 x i32>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [2 x <4 x i32>] %b.coerce, 1
+  %1 = bitcast i32* %a to i8*
+  tail call void @llvm.arm.neon.vst2.v4i32(i8* %1, <4 x i32> %b.coerce.fca.0.extract, <4 x i32> %b.coerce.fca.1.extract, i32 4)
+  ret void
+}
+
+define void @test_vst2q_s64(i64* %a, [2 x <2 x i64>] %b.coerce) {
+; CHECK: test_vst2q_s64
+; CHECK: st2 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [2 x <2 x i64>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [2 x <2 x i64>] %b.coerce, 1
+  %1 = bitcast i64* %a to i8*
+  tail call void @llvm.arm.neon.vst2.v2i64(i8* %1, <2 x i64> %b.coerce.fca.0.extract, <2 x i64> %b.coerce.fca.1.extract, i32 8)
+  ret void
+}
+
+define void @test_vst2q_f32(float* %a, [2 x <4 x float>] %b.coerce) {
+; CHECK: test_vst2q_f32
+; CHECK: st2 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [2 x <4 x float>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [2 x <4 x float>] %b.coerce, 1
+  %1 = bitcast float* %a to i8*
+  tail call void @llvm.arm.neon.vst2.v4f32(i8* %1, <4 x float> %b.coerce.fca.0.extract, <4 x float> %b.coerce.fca.1.extract, i32 4)
+  ret void
+}
+
+define void @test_vst2q_f64(double* %a, [2 x <2 x double>] %b.coerce) {
+; CHECK: test_vst2q_f64
+; CHECK: st2 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [2 x <2 x double>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [2 x <2 x double>] %b.coerce, 1
+  %1 = bitcast double* %a to i8*
+  tail call void @llvm.arm.neon.vst2.v2f64(i8* %1, <2 x double> %b.coerce.fca.0.extract, <2 x double> %b.coerce.fca.1.extract, i32 8)
+  ret void
+}
+
+define void @test_vst2_s8(i8* %a, [2 x <8 x i8>] %b.coerce) {
+; CHECK: test_vst2_s8
+; CHECK: st2 {v{{[0-9]+}}.8b, v{{[0-9]+}}.8b}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [2 x <8 x i8>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [2 x <8 x i8>] %b.coerce, 1
+  tail call void @llvm.arm.neon.vst2.v8i8(i8* %a, <8 x i8> %b.coerce.fca.0.extract, <8 x i8> %b.coerce.fca.1.extract, i32 1)
+  ret void
+}
+
+define void @test_vst2_s16(i16* %a, [2 x <4 x i16>] %b.coerce) {
+; CHECK: test_vst2_s16
+; CHECK: st2 {v{{[0-9]+}}.4h, v{{[0-9]+}}.4h}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [2 x <4 x i16>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [2 x <4 x i16>] %b.coerce, 1
+  %1 = bitcast i16* %a to i8*
+  tail call void @llvm.arm.neon.vst2.v4i16(i8* %1, <4 x i16> %b.coerce.fca.0.extract, <4 x i16> %b.coerce.fca.1.extract, i32 2)
+  ret void
+}
+
+define void @test_vst2_s32(i32* %a, [2 x <2 x i32>] %b.coerce) {
+; CHECK: test_vst2_s32
+; CHECK: st2 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [2 x <2 x i32>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [2 x <2 x i32>] %b.coerce, 1
+  %1 = bitcast i32* %a to i8*
+  tail call void @llvm.arm.neon.vst2.v2i32(i8* %1, <2 x i32> %b.coerce.fca.0.extract, <2 x i32> %b.coerce.fca.1.extract, i32 4)
+  ret void
+}
+
+define void @test_vst2_s64(i64* %a, [2 x <1 x i64>] %b.coerce) {
+; CHECK: test_vst2_s64
+; CHECK: st1 {v{{[0-9]+}}.1d, v{{[0-9]+}}.1d}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [2 x <1 x i64>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [2 x <1 x i64>] %b.coerce, 1
+  %1 = bitcast i64* %a to i8*
+  tail call void @llvm.arm.neon.vst2.v1i64(i8* %1, <1 x i64> %b.coerce.fca.0.extract, <1 x i64> %b.coerce.fca.1.extract, i32 8)
+  ret void
+}
+
+define void @test_vst2_f32(float* %a, [2 x <2 x float>] %b.coerce) {
+; CHECK: test_vst2_f32
+; CHECK: st2 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [2 x <2 x float>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [2 x <2 x float>] %b.coerce, 1
+  %1 = bitcast float* %a to i8*
+  tail call void @llvm.arm.neon.vst2.v2f32(i8* %1, <2 x float> %b.coerce.fca.0.extract, <2 x float> %b.coerce.fca.1.extract, i32 4)
+  ret void
+}
+
+define void @test_vst2_f64(double* %a, [2 x <1 x double>] %b.coerce) {
+; CHECK: test_vst2_f64
+; CHECK: st1 {v{{[0-9]+}}.1d, v{{[0-9]+}}.1d}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [2 x <1 x double>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [2 x <1 x double>] %b.coerce, 1
+  %1 = bitcast double* %a to i8*
+  tail call void @llvm.arm.neon.vst2.v1f64(i8* %1, <1 x double> %b.coerce.fca.0.extract, <1 x double> %b.coerce.fca.1.extract, i32 8)
+  ret void
+}
+
+define void @test_vst3q_s8(i8* %a, [3 x <16 x i8>] %b.coerce) {
+; CHECK: test_vst3q_s8
+; CHECK: st3 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [3 x <16 x i8>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [3 x <16 x i8>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [3 x <16 x i8>] %b.coerce, 2
+  tail call void @llvm.arm.neon.vst3.v16i8(i8* %a, <16 x i8> %b.coerce.fca.0.extract, <16 x i8> %b.coerce.fca.1.extract, <16 x i8> %b.coerce.fca.2.extract, i32 1)
+  ret void
+}
+
+define void @test_vst3q_s16(i16* %a, [3 x <8 x i16>] %b.coerce) {
+; CHECK: test_vst3q_s16
+; CHECK: st3 {v{{[0-9]+}}.8h, v{{[0-9]+}}.8h, v{{[0-9]+}}.8h}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [3 x <8 x i16>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [3 x <8 x i16>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [3 x <8 x i16>] %b.coerce, 2
+  %1 = bitcast i16* %a to i8*
+  tail call void @llvm.arm.neon.vst3.v8i16(i8* %1, <8 x i16> %b.coerce.fca.0.extract, <8 x i16> %b.coerce.fca.1.extract, <8 x i16> %b.coerce.fca.2.extract, i32 2)
+  ret void
+}
+
+define void @test_vst3q_s32(i32* %a, [3 x <4 x i32>] %b.coerce) {
+; CHECK: test_vst3q_s32
+; CHECK: st3 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [3 x <4 x i32>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [3 x <4 x i32>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [3 x <4 x i32>] %b.coerce, 2
+  %1 = bitcast i32* %a to i8*
+  tail call void @llvm.arm.neon.vst3.v4i32(i8* %1, <4 x i32> %b.coerce.fca.0.extract, <4 x i32> %b.coerce.fca.1.extract, <4 x i32> %b.coerce.fca.2.extract, i32 4)
+  ret void
+}
+
+define void @test_vst3q_s64(i64* %a, [3 x <2 x i64>] %b.coerce) {
+; CHECK: test_vst3q_s64
+; CHECK: st3 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [3 x <2 x i64>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [3 x <2 x i64>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [3 x <2 x i64>] %b.coerce, 2
+  %1 = bitcast i64* %a to i8*
+  tail call void @llvm.arm.neon.vst3.v2i64(i8* %1, <2 x i64> %b.coerce.fca.0.extract, <2 x i64> %b.coerce.fca.1.extract, <2 x i64> %b.coerce.fca.2.extract, i32 8)
+  ret void
+}
+
+define void @test_vst3q_f32(float* %a, [3 x <4 x float>] %b.coerce) {
+; CHECK: test_vst3q_f32
+; CHECK: st3 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [3 x <4 x float>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [3 x <4 x float>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [3 x <4 x float>] %b.coerce, 2
+  %1 = bitcast float* %a to i8*
+  tail call void @llvm.arm.neon.vst3.v4f32(i8* %1, <4 x float> %b.coerce.fca.0.extract, <4 x float> %b.coerce.fca.1.extract, <4 x float> %b.coerce.fca.2.extract, i32 4)
+  ret void
+}
+
+define void @test_vst3q_f64(double* %a, [3 x <2 x double>] %b.coerce) {
+; CHECK: test_vst3q_f64
+; CHECK: st3 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [3 x <2 x double>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [3 x <2 x double>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [3 x <2 x double>] %b.coerce, 2
+  %1 = bitcast double* %a to i8*
+  tail call void @llvm.arm.neon.vst3.v2f64(i8* %1, <2 x double> %b.coerce.fca.0.extract, <2 x double> %b.coerce.fca.1.extract, <2 x double> %b.coerce.fca.2.extract, i32 8)
+  ret void
+}
+
+define void @test_vst3_s8(i8* %a, [3 x <8 x i8>] %b.coerce) {
+; CHECK: test_vst3_s8
+; CHECK: st3 {v{{[0-9]+}}.8b, v{{[0-9]+}}.8b, v{{[0-9]+}}.8b}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [3 x <8 x i8>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [3 x <8 x i8>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [3 x <8 x i8>] %b.coerce, 2
+  tail call void @llvm.arm.neon.vst3.v8i8(i8* %a, <8 x i8> %b.coerce.fca.0.extract, <8 x i8> %b.coerce.fca.1.extract, <8 x i8> %b.coerce.fca.2.extract, i32 1)
+  ret void
+}
+
+define void @test_vst3_s16(i16* %a, [3 x <4 x i16>] %b.coerce) {
+; CHECK: test_vst3_s16
+; CHECK: st3 {v{{[0-9]+}}.4h, v{{[0-9]+}}.4h, v{{[0-9]+}}.4h}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [3 x <4 x i16>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [3 x <4 x i16>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [3 x <4 x i16>] %b.coerce, 2
+  %1 = bitcast i16* %a to i8*
+  tail call void @llvm.arm.neon.vst3.v4i16(i8* %1, <4 x i16> %b.coerce.fca.0.extract, <4 x i16> %b.coerce.fca.1.extract, <4 x i16> %b.coerce.fca.2.extract, i32 2)
+  ret void
+}
+
+define void @test_vst3_s32(i32* %a, [3 x <2 x i32>] %b.coerce) {
+; CHECK: test_vst3_s32
+; CHECK: st3 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [3 x <2 x i32>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [3 x <2 x i32>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [3 x <2 x i32>] %b.coerce, 2
+  %1 = bitcast i32* %a to i8*
+  tail call void @llvm.arm.neon.vst3.v2i32(i8* %1, <2 x i32> %b.coerce.fca.0.extract, <2 x i32> %b.coerce.fca.1.extract, <2 x i32> %b.coerce.fca.2.extract, i32 4)
+  ret void
+}
+
+define void @test_vst3_s64(i64* %a, [3 x <1 x i64>] %b.coerce) {
+; CHECK: test_vst3_s64
+; CHECK: st1 {v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [3 x <1 x i64>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [3 x <1 x i64>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [3 x <1 x i64>] %b.coerce, 2
+  %1 = bitcast i64* %a to i8*
+  tail call void @llvm.arm.neon.vst3.v1i64(i8* %1, <1 x i64> %b.coerce.fca.0.extract, <1 x i64> %b.coerce.fca.1.extract, <1 x i64> %b.coerce.fca.2.extract, i32 8)
+  ret void
+}
+
+define void @test_vst3_f32(float* %a, [3 x <2 x float>] %b.coerce) {
+; CHECK: test_vst3_f32
+; CHECK: st3 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [3 x <2 x float>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [3 x <2 x float>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [3 x <2 x float>] %b.coerce, 2
+  %1 = bitcast float* %a to i8*
+  tail call void @llvm.arm.neon.vst3.v2f32(i8* %1, <2 x float> %b.coerce.fca.0.extract, <2 x float> %b.coerce.fca.1.extract, <2 x float> %b.coerce.fca.2.extract, i32 4)
+  ret void
+}
+
+define void @test_vst3_f64(double* %a, [3 x <1 x double>] %b.coerce) {
+; CHECK: test_vst3_f64
+; CHECK: st1 {v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [3 x <1 x double>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [3 x <1 x double>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [3 x <1 x double>] %b.coerce, 2
+  %1 = bitcast double* %a to i8*
+  tail call void @llvm.arm.neon.vst3.v1f64(i8* %1, <1 x double> %b.coerce.fca.0.extract, <1 x double> %b.coerce.fca.1.extract, <1 x double> %b.coerce.fca.2.extract, i32 8)
+  ret void
+}
+
+define void @test_vst4q_s8(i8* %a, [4 x <16 x i8>] %b.coerce) {
+; CHECK: test_vst4q_s8
+; CHECK: st4 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [4 x <16 x i8>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [4 x <16 x i8>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [4 x <16 x i8>] %b.coerce, 2
+  %b.coerce.fca.3.extract = extractvalue [4 x <16 x i8>] %b.coerce, 3
+  tail call void @llvm.arm.neon.vst4.v16i8(i8* %a, <16 x i8> %b.coerce.fca.0.extract, <16 x i8> %b.coerce.fca.1.extract, <16 x i8> %b.coerce.fca.2.extract, <16 x i8> %b.coerce.fca.3.extract, i32 1)
+  ret void
+}
+
+define void @test_vst4q_s16(i16* %a, [4 x <8 x i16>] %b.coerce) {
+; CHECK: test_vst4q_s16
+; CHECK: st4 {v{{[0-9]+}}.8h, v{{[0-9]+}}.8h, v{{[0-9]+}}.8h, v{{[0-9]+}}.8h}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [4 x <8 x i16>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [4 x <8 x i16>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [4 x <8 x i16>] %b.coerce, 2
+  %b.coerce.fca.3.extract = extractvalue [4 x <8 x i16>] %b.coerce, 3
+  %1 = bitcast i16* %a to i8*
+  tail call void @llvm.arm.neon.vst4.v8i16(i8* %1, <8 x i16> %b.coerce.fca.0.extract, <8 x i16> %b.coerce.fca.1.extract, <8 x i16> %b.coerce.fca.2.extract, <8 x i16> %b.coerce.fca.3.extract, i32 2)
+  ret void
+}
+
+define void @test_vst4q_s32(i32* %a, [4 x <4 x i32>] %b.coerce) {
+; CHECK: test_vst4q_s32
+; CHECK: st4 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [4 x <4 x i32>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [4 x <4 x i32>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [4 x <4 x i32>] %b.coerce, 2
+  %b.coerce.fca.3.extract = extractvalue [4 x <4 x i32>] %b.coerce, 3
+  %1 = bitcast i32* %a to i8*
+  tail call void @llvm.arm.neon.vst4.v4i32(i8* %1, <4 x i32> %b.coerce.fca.0.extract, <4 x i32> %b.coerce.fca.1.extract, <4 x i32> %b.coerce.fca.2.extract, <4 x i32> %b.coerce.fca.3.extract, i32 4)
+  ret void
+}
+
+define void @test_vst4q_s64(i64* %a, [4 x <2 x i64>] %b.coerce) {
+; CHECK: test_vst4q_s64
+; CHECK: st4 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [4 x <2 x i64>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [4 x <2 x i64>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [4 x <2 x i64>] %b.coerce, 2
+  %b.coerce.fca.3.extract = extractvalue [4 x <2 x i64>] %b.coerce, 3
+  %1 = bitcast i64* %a to i8*
+  tail call void @llvm.arm.neon.vst4.v2i64(i8* %1, <2 x i64> %b.coerce.fca.0.extract, <2 x i64> %b.coerce.fca.1.extract, <2 x i64> %b.coerce.fca.2.extract, <2 x i64> %b.coerce.fca.3.extract, i32 8)
+  ret void
+}
+
+define void @test_vst4q_f32(float* %a, [4 x <4 x float>] %b.coerce) {
+; CHECK: test_vst4q_f32
+; CHECK: st4 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [4 x <4 x float>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [4 x <4 x float>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [4 x <4 x float>] %b.coerce, 2
+  %b.coerce.fca.3.extract = extractvalue [4 x <4 x float>] %b.coerce, 3
+  %1 = bitcast float* %a to i8*
+  tail call void @llvm.arm.neon.vst4.v4f32(i8* %1, <4 x float> %b.coerce.fca.0.extract, <4 x float> %b.coerce.fca.1.extract, <4 x float> %b.coerce.fca.2.extract, <4 x float> %b.coerce.fca.3.extract, i32 4)
+  ret void
+}
+
+define void @test_vst4q_f64(double* %a, [4 x <2 x double>] %b.coerce) {
+; CHECK: test_vst4q_f64
+; CHECK: st4 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [4 x <2 x double>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [4 x <2 x double>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [4 x <2 x double>] %b.coerce, 2
+  %b.coerce.fca.3.extract = extractvalue [4 x <2 x double>] %b.coerce, 3
+  %1 = bitcast double* %a to i8*
+  tail call void @llvm.arm.neon.vst4.v2f64(i8* %1, <2 x double> %b.coerce.fca.0.extract, <2 x double> %b.coerce.fca.1.extract, <2 x double> %b.coerce.fca.2.extract, <2 x double> %b.coerce.fca.3.extract, i32 8)
+  ret void
+}
+
+define void @test_vst4_s8(i8* %a, [4 x <8 x i8>] %b.coerce) {
+; CHECK: test_vst4_s8
+; CHECK: st4 {v{{[0-9]+}}.8b, v{{[0-9]+}}.8b, v{{[0-9]+}}.8b, v{{[0-9]+}}.8b}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [4 x <8 x i8>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [4 x <8 x i8>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [4 x <8 x i8>] %b.coerce, 2
+  %b.coerce.fca.3.extract = extractvalue [4 x <8 x i8>] %b.coerce, 3
+  tail call void @llvm.arm.neon.vst4.v8i8(i8* %a, <8 x i8> %b.coerce.fca.0.extract, <8 x i8> %b.coerce.fca.1.extract, <8 x i8> %b.coerce.fca.2.extract, <8 x i8> %b.coerce.fca.3.extract, i32 1)
+  ret void
+}
+
+define void @test_vst4_s16(i16* %a, [4 x <4 x i16>] %b.coerce) {
+; CHECK: test_vst4_s16
+; CHECK: st4 {v{{[0-9]+}}.4h, v{{[0-9]+}}.4h, v{{[0-9]+}}.4h, v{{[0-9]+}}.4h}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [4 x <4 x i16>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [4 x <4 x i16>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [4 x <4 x i16>] %b.coerce, 2
+  %b.coerce.fca.3.extract = extractvalue [4 x <4 x i16>] %b.coerce, 3
+  %1 = bitcast i16* %a to i8*
+  tail call void @llvm.arm.neon.vst4.v4i16(i8* %1, <4 x i16> %b.coerce.fca.0.extract, <4 x i16> %b.coerce.fca.1.extract, <4 x i16> %b.coerce.fca.2.extract, <4 x i16> %b.coerce.fca.3.extract, i32 2)
+  ret void
+}
+
+define void @test_vst4_s32(i32* %a, [4 x <2 x i32>] %b.coerce) {
+; CHECK: test_vst4_s32
+; CHECK: st4 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s, v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [4 x <2 x i32>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [4 x <2 x i32>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [4 x <2 x i32>] %b.coerce, 2
+  %b.coerce.fca.3.extract = extractvalue [4 x <2 x i32>] %b.coerce, 3
+  %1 = bitcast i32* %a to i8*
+  tail call void @llvm.arm.neon.vst4.v2i32(i8* %1, <2 x i32> %b.coerce.fca.0.extract, <2 x i32> %b.coerce.fca.1.extract, <2 x i32> %b.coerce.fca.2.extract, <2 x i32> %b.coerce.fca.3.extract, i32 4)
+  ret void
+}
+
+define void @test_vst4_s64(i64* %a, [4 x <1 x i64>] %b.coerce) {
+; CHECK: test_vst4_s64
+; CHECK: st1 {v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [4 x <1 x i64>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [4 x <1 x i64>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [4 x <1 x i64>] %b.coerce, 2
+  %b.coerce.fca.3.extract = extractvalue [4 x <1 x i64>] %b.coerce, 3
+  %1 = bitcast i64* %a to i8*
+  tail call void @llvm.arm.neon.vst4.v1i64(i8* %1, <1 x i64> %b.coerce.fca.0.extract, <1 x i64> %b.coerce.fca.1.extract, <1 x i64> %b.coerce.fca.2.extract, <1 x i64> %b.coerce.fca.3.extract, i32 8)
+  ret void
+}
+
+define void @test_vst4_f32(float* %a, [4 x <2 x float>] %b.coerce) {
+; CHECK: test_vst4_f32
+; CHECK: st4 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s, v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [4 x <2 x float>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [4 x <2 x float>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [4 x <2 x float>] %b.coerce, 2
+  %b.coerce.fca.3.extract = extractvalue [4 x <2 x float>] %b.coerce, 3
+  %1 = bitcast float* %a to i8*
+  tail call void @llvm.arm.neon.vst4.v2f32(i8* %1, <2 x float> %b.coerce.fca.0.extract, <2 x float> %b.coerce.fca.1.extract, <2 x float> %b.coerce.fca.2.extract, <2 x float> %b.coerce.fca.3.extract, i32 4)
+  ret void
+}
+
+define void @test_vst4_f64(double* %a, [4 x <1 x double>] %b.coerce) {
+; CHECK: test_vst4_f64
+; CHECK: st1 {v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d}, [{{x[0-9]+|sp}}]
+  %b.coerce.fca.0.extract = extractvalue [4 x <1 x double>] %b.coerce, 0
+  %b.coerce.fca.1.extract = extractvalue [4 x <1 x double>] %b.coerce, 1
+  %b.coerce.fca.2.extract = extractvalue [4 x <1 x double>] %b.coerce, 2
+  %b.coerce.fca.3.extract = extractvalue [4 x <1 x double>] %b.coerce, 3
+  %1 = bitcast double* %a to i8*
+  tail call void @llvm.arm.neon.vst4.v1f64(i8* %1, <1 x double> %b.coerce.fca.0.extract, <1 x double> %b.coerce.fca.1.extract, <1 x double> %b.coerce.fca.2.extract, <1 x double> %b.coerce.fca.3.extract, i32 8)
+  ret void
+}
+
+declare void @llvm.arm.neon.vst1.v16i8(i8*, <16 x i8>, i32)
+declare void @llvm.arm.neon.vst1.v8i16(i8*, <8 x i16>, i32)
+declare void @llvm.arm.neon.vst1.v4i32(i8*, <4 x i32>, i32)
+declare void @llvm.arm.neon.vst1.v2i64(i8*, <2 x i64>, i32)
+declare void @llvm.arm.neon.vst1.v4f32(i8*, <4 x float>, i32)
+declare void @llvm.arm.neon.vst1.v2f64(i8*, <2 x double>, i32)
+declare void @llvm.arm.neon.vst1.v8i8(i8*, <8 x i8>, i32)
+declare void @llvm.arm.neon.vst1.v4i16(i8*, <4 x i16>, i32)
+declare void @llvm.arm.neon.vst1.v2i32(i8*, <2 x i32>, i32)
+declare void @llvm.arm.neon.vst1.v1i64(i8*, <1 x i64>, i32)
+declare void @llvm.arm.neon.vst1.v2f32(i8*, <2 x float>, i32)
+declare void @llvm.arm.neon.vst1.v1f64(i8*, <1 x double>, i32)
+declare void @llvm.arm.neon.vst2.v16i8(i8*, <16 x i8>, <16 x i8>, i32)
+declare void @llvm.arm.neon.vst2.v8i16(i8*, <8 x i16>, <8 x i16>, i32)
+declare void @llvm.arm.neon.vst2.v4i32(i8*, <4 x i32>, <4 x i32>, i32)
+declare void @llvm.arm.neon.vst2.v2i64(i8*, <2 x i64>, <2 x i64>, i32)
+declare void @llvm.arm.neon.vst2.v4f32(i8*, <4 x float>, <4 x float>, i32)
+declare void @llvm.arm.neon.vst2.v2f64(i8*, <2 x double>, <2 x double>, i32)
+declare void @llvm.arm.neon.vst2.v8i8(i8*, <8 x i8>, <8 x i8>, i32)
+declare void @llvm.arm.neon.vst2.v4i16(i8*, <4 x i16>, <4 x i16>, i32)
+declare void @llvm.arm.neon.vst2.v2i32(i8*, <2 x i32>, <2 x i32>, i32)
+declare void @llvm.arm.neon.vst2.v1i64(i8*, <1 x i64>, <1 x i64>, i32)
+declare void @llvm.arm.neon.vst2.v2f32(i8*, <2 x float>, <2 x float>, i32)
+declare void @llvm.arm.neon.vst2.v1f64(i8*, <1 x double>, <1 x double>, i32)
+declare void @llvm.arm.neon.vst3.v16i8(i8*, <16 x i8>, <16 x i8>, <16 x i8>, i32)
+declare void @llvm.arm.neon.vst3.v8i16(i8*, <8 x i16>, <8 x i16>, <8 x i16>, i32)
+declare void @llvm.arm.neon.vst3.v4i32(i8*, <4 x i32>, <4 x i32>, <4 x i32>, i32)
+declare void @llvm.arm.neon.vst3.v2i64(i8*, <2 x i64>, <2 x i64>, <2 x i64>, i32)
+declare void @llvm.arm.neon.vst3.v4f32(i8*, <4 x float>, <4 x float>, <4 x float>, i32)
+declare void @llvm.arm.neon.vst3.v2f64(i8*, <2 x double>, <2 x double>, <2 x double>, i32)
+declare void @llvm.arm.neon.vst3.v8i8(i8*, <8 x i8>, <8 x i8>, <8 x i8>, i32)
+declare void @llvm.arm.neon.vst3.v4i16(i8*, <4 x i16>, <4 x i16>, <4 x i16>, i32)
+declare void @llvm.arm.neon.vst3.v2i32(i8*, <2 x i32>, <2 x i32>, <2 x i32>, i32)
+declare void @llvm.arm.neon.vst3.v1i64(i8*, <1 x i64>, <1 x i64>, <1 x i64>, i32)
+declare void @llvm.arm.neon.vst3.v2f32(i8*, <2 x float>, <2 x float>, <2 x float>, i32)
+declare void @llvm.arm.neon.vst3.v1f64(i8*, <1 x double>, <1 x double>, <1 x double>, i32)
+declare void @llvm.arm.neon.vst4.v16i8(i8*, <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8>, i32)
+declare void @llvm.arm.neon.vst4.v8i16(i8*, <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16>, i32)
+declare void @llvm.arm.neon.vst4.v4i32(i8*, <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32>, i32)
+declare void @llvm.arm.neon.vst4.v2i64(i8*, <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, i32)
+declare void @llvm.arm.neon.vst4.v4f32(i8*, <4 x float>, <4 x float>, <4 x float>, <4 x float>, i32)
+declare void @llvm.arm.neon.vst4.v2f64(i8*, <2 x double>, <2 x double>, <2 x double>, <2 x double>, i32)
+declare void @llvm.arm.neon.vst4.v8i8(i8*, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, i32)
+declare void @llvm.arm.neon.vst4.v4i16(i8*, <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16>, i32)
+declare void @llvm.arm.neon.vst4.v2i32(i8*, <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, i32)
+declare void @llvm.arm.neon.vst4.v1i64(i8*, <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64>, i32)
+declare void @llvm.arm.neon.vst4.v2f32(i8*, <2 x float>, <2 x float>, <2 x float>, <2 x float>, i32)
+declare void @llvm.arm.neon.vst4.v1f64(i8*, <1 x double>, <1 x double>, <1 x double>, <1 x double>, i32)
\ No newline at end of file
diff --git a/test/MC/AArch64/neon-diagnostics.s b/test/MC/AArch64/neon-diagnostics.s
index 9127ed840b5..086d4876e35 100644
--- a/test/MC/AArch64/neon-diagnostics.s
+++ b/test/MC/AArch64/neon-diagnostics.s
@@ -3880,3 +3880,224 @@
 // CHECK-ERROR: error: invalid operand for instruction
 // CHECK-ERROR:          frsqrts d8, s22, d18
 // CHECK-ERROR:                      ^
+
+//----------------------------------------------------------------------
+// Vector load/store multiple N-element structure (class SIMD lselem)
+//----------------------------------------------------------------------
+         ld1 {x3}, [x2]
+         ld1 {v4}, [x0]
+         ld1 {v32.16b}, [x0]
+         ld1 {v15.8h}, [x32]
+// CHECK-ERROR: error: expected vector type register
+// CHECK-ERROR:        ld1 {x3}, [x2]
+// CHECK-ERROR:             ^
+// CHECK-ERROR: error: expected vector type register
+// CHECK-ERROR:        ld1 {v4}, [x0]
+// CHECK-ERROR:             ^
+// CHECK-ERROR: error: expected vector type register
+// CHECK-ERROR:        ld1 {v32.16b}, [x0]
+// CHECK-ERROR:             ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        ld1 {v15.8h}, [x32]
+// CHECK-ERROR:                       ^
+
+         ld1 {v0.16b, v2.16b}, [x0]
+         ld1 {v0.8h, v1.8h, v2.8h, v3.8h, v4.8h}, [x0]
+         ld1 v0.8b, v1.8b}, [x0]
+         ld1 {v0.8h-v4.8h}, [x0]
+         ld1 {v1.8h-v1.8h}, [x0]
+         ld1 {v15.8h-v17.4h}, [x15]
+         ld1 {v0.8b-v2.8b, [x0]
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        ld1 {v0.16b, v2.16b}, [x0]
+// CHECK-ERROR:                     ^
+// CHECK-ERROR: error: invalid number of vectors
+// CHECK-ERROR:        ld1 {v0.8h, v1.8h, v2.8h, v3.8h, v4.8h}, [x0]
+// CHECK-ERROR:                                         ^
+// CHECK-ERROR: error: '{' expected
+// CHECK-ERROR:        ld1 v0.8b, v1.8b}, [x0]
+// CHECK-ERROR:            ^
+// CHECK-ERROR: error: invalid number of vectors
+// CHECK-ERROR:        ld1 {v0.8h-v4.8h}, [x0]
+// CHECK-ERROR:                   ^
+// CHECK-ERROR: error: invalid number of vectors
+// CHECK-ERROR:        ld1 {v1.8h-v1.8h}, [x0]
+// CHECK-ERROR:                   ^
+// CHECK-ERROR: error: expected the same vector layout
+// CHECK-ERROR:        ld1 {v15.8h-v17.4h}, [x15]
+// CHECK-ERROR:                        ^
+// CHECK-ERROR: error: '}' expected
+// CHECK-ERROR:        ld1 {v0.8b-v2.8b, [x0]
+// CHECK-ERROR:                        ^
+
+         ld2 {v15.8h, v16.4h}, [x15]
+         ld2 {v0.8b, v2.8b}, [x0]
+         ld2 {v15.4h, v16.4h, v17.4h}, [x32]
+         ld2 {v15.8h-v16.4h}, [x15]
+         ld2 {v0.2d-v2.2d}, [x0]
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        ld2 {v15.8h, v16.4h}, [x15]
+// CHECK-ERROR:                     ^
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        ld2 {v0.8b, v2.8b}, [x0]
+// CHECK-ERROR:                    ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        ld2 {v15.4h, v16.4h, v17.4h}, [x32]
+// CHECK-ERROR:            ^
+// CHECK-ERROR: error: expected the same vector layout
+// CHECK-ERROR:        ld2 {v15.8h-v16.4h}, [x15]
+// CHECK-ERROR:                        ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        ld2 {v0.2d-v2.2d}, [x0]
+// CHECK-ERROR:            ^
+
+         ld3 {v15.8h, v16.8h, v17.4h}, [x15]
+         ld3 {v0.8b, v1,8b, v2.8b, v3.8b}, [x0]
+         ld3 {v0.8b, v2.8b, v3.8b}, [x0]
+         ld3 {v15.8h-v17.4h}, [x15]
+         ld3 {v31.4s-v2.4s}, [sp]
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        ld3 {v15.8h, v16.8h, v17.4h}, [x15]
+// CHECK-ERROR:                             ^
+// CHECK-ERROR: error: expected vector type register
+// CHECK-ERROR:        ld3 {v0.8b, v1,8b, v2.8b, v3.8b}, [x0]
+// CHECK-ERROR:                    ^
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        ld3 {v0.8b, v2.8b, v3.8b}, [x0]
+// CHECK-ERROR:                    ^
+// CHECK-ERROR: error: expected the same vector layout
+// CHECK-ERROR:        ld3 {v15.8h-v17.4h}, [x15]
+// CHECK-ERROR:                        ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        ld3 {v31.4s-v2.4s}, [sp]
+// CHECK-ERROR:            ^
+
+         ld4 {v15.8h, v16.8h, v17.4h, v18.8h}, [x15]
+         ld4 {v0.8b, v2.8b, v3.8b, v4.8b}, [x0]
+         ld4 {v15.4h, v16.4h, v17.4h, v18.4h, v19.4h}, [x31]
+         ld4 {v15.8h-v18.4h}, [x15]
+         ld4 {v31.2s-v1.2s}, [x31]
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        ld4 {v15.8h, v16.8h, v17.4h, v18.8h}, [x15]
+// CHECK-ERROR:                             ^
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        ld4 {v0.8b, v2.8b, v3.8b, v4.8b}, [x0]
+// CHECK-ERROR:                    ^
+// CHECK-ERROR: error: invalid number of vectors
+// CHECK-ERROR:        ld4 {v15.4h, v16.4h, v17.4h, v18.4h, v19.4h}, [x31]
+// CHECK-ERROR:                                             ^
+// CHECK-ERROR: error: expected the same vector layout
+// CHECK-ERROR:        ld4 {v15.8h-v18.4h}, [x15]
+// CHECK-ERROR:                        ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        ld4 {v31.2s-v1.2s}, [x31]
+// CHECK-ERROR:            ^
+
+         st1 {x3}, [x2]
+         st1 {v4}, [x0]
+         st1 {v32.16b}, [x0]
+         st1 {v15.8h}, [x32]
+// CHECK-ERROR: error: expected vector type register
+// CHECK-ERROR:        st1 {x3}, [x2]
+// CHECK-ERROR:             ^
+// CHECK-ERROR: error: expected vector type register
+// CHECK-ERROR:        st1 {v4}, [x0]
+// CHECK-ERROR:             ^
+// CHECK-ERROR: error: expected vector type register
+// CHECK-ERROR:        st1 {v32.16b}, [x0]
+// CHECK-ERROR:             ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        st1 {v15.8h}, [x32]
+// CHECK-ERROR:                       ^
+
+         st1 {v0.16b, v2.16b}, [x0]
+         st1 {v0.8h, v1.8h, v2.8h, v3.8h, v4.8h}, [x0]
+         st1 v0.8b, v1.8b}, [x0]
+         st1 {v0.8h-v4.8h}, [x0]
+         st1 {v1.8h-v1.8h}, [x0]
+         st1 {v15.8h-v17.4h}, [x15]
+         st1 {v0.8b-v2.8b, [x0]
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        st1 {v0.16b, v2.16b}, [x0]
+// CHECK-ERROR:                     ^
+// CHECK-ERROR: error: invalid number of vectors
+// CHECK-ERROR:        st1 {v0.8h, v1.8h, v2.8h, v3.8h, v4.8h}, [x0]
+// CHECK-ERROR:                                         ^
+// CHECK-ERROR: error: '{' expected
+// CHECK-ERROR:        st1 v0.8b, v1.8b}, [x0]
+// CHECK-ERROR:            ^
+// CHECK-ERROR: error: invalid number of vectors
+// CHECK-ERROR:        st1 {v0.8h-v4.8h}, [x0]
+// CHECK-ERROR:                   ^
+// CHECK-ERROR: error: invalid number of vectors
+// CHECK-ERROR:        st1 {v1.8h-v1.8h}, [x0]
+// CHECK-ERROR:                   ^
+// CHECK-ERROR: error: expected the same vector layout
+// CHECK-ERROR:        st1 {v15.8h-v17.4h}, [x15]
+// CHECK-ERROR:                        ^
+// CHECK-ERROR: error: '}' expected
+// CHECK-ERROR:        st1 {v0.8b-v2.8b, [x0]
+// CHECK-ERROR:                        ^
+
+         st2 {v15.8h, v16.4h}, [x15]
+         st2 {v0.8b, v2.8b}, [x0]
+         st2 {v15.4h, v16.4h, v17.4h}, [x30]
+         st2 {v15.8h-v16.4h}, [x15]
+         st2 {v0.2d-v2.2d}, [x0]
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        st2 {v15.8h, v16.4h}, [x15]
+// CHECK-ERROR:                     ^
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        st2 {v0.8b, v2.8b}, [x0]
+// CHECK-ERROR:                    ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        st2 {v15.4h, v16.4h, v17.4h}, [x30]
+// CHECK-ERROR:            ^
+// CHECK-ERROR: error: expected the same vector layout
+// CHECK-ERROR:        st2 {v15.8h-v16.4h}, [x15]
+// CHECK-ERROR:                        ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        st2 {v0.2d-v2.2d}, [x0]
+// CHECK-ERROR:            ^
+
+         st3 {v15.8h, v16.8h, v17.4h}, [x15]
+         st3 {v0.8b, v1,8b, v2.8b, v3.8b}, [x0]
+         st3 {v0.8b, v2.8b, v3.8b}, [x0]
+         st3 {v15.8h-v17.4h}, [x15]
+         st3 {v31.4s-v2.4s}, [sp]
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        st3 {v15.8h, v16.8h, v17.4h}, [x15]
+// CHECK-ERROR:                             ^
+// CHECK-ERROR: error: expected vector type register
+// CHECK-ERROR:        st3 {v0.8b, v1,8b, v2.8b, v3.8b}, [x0]
+// CHECK-ERROR:                    ^
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        st3 {v0.8b, v2.8b, v3.8b}, [x0]
+// CHECK-ERROR:                    ^
+// CHECK-ERROR: error: expected the same vector layout
+// CHECK-ERROR:        st3 {v15.8h-v17.4h}, [x15]
+// CHECK-ERROR:                        ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        st3 {v31.4s-v2.4s}, [sp]
+// CHECK-ERROR:            ^
+
+         st4 {v15.8h, v16.8h, v17.4h, v18.8h}, [x15]
+         st4 {v0.8b, v2.8b, v3.8b, v4.8b}, [x0]
+         st4 {v15.4h, v16.4h, v17.4h, v18.4h, v19.4h}, [x31]
+         st4 {v15.8h-v18.4h}, [x15]
+         st4 {v31.2s-v1.2s}, [x31]
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        st4 {v15.8h, v16.8h, v17.4h, v18.8h}, [x15]
+// CHECK-ERROR:                             ^
+// CHECK-ERROR: error: invalid space between two vectors
+// CHECK-ERROR:        st4 {v0.8b, v2.8b, v3.8b, v4.8b}, [x0]
+// CHECK-ERROR:                    ^
+// CHECK-ERROR: error: invalid number of vectors
+// CHECK-ERROR:        st4 {v15.4h, v16.4h, v17.4h, v18.4h, v19.4h}, [x31]
+// CHECK-ERROR:                                             ^
+// CHECK-ERROR: error: expected the same vector layout
+// CHECK-ERROR:        st4 {v15.8h-v18.4h}, [x15]
+// CHECK-ERROR:                        ^
+// CHECK-ERROR: error: invalid operand for instruction
+// CHECK-ERROR:        st4 {v31.2s-v1.2s}, [x31]
+// CHECK-ERROR:            ^
diff --git a/test/MC/AArch64/neon-simd-ldst-multi-elem.s b/test/MC/AArch64/neon-simd-ldst-multi-elem.s
new file mode 100644
index 00000000000..05fe4dac913
--- /dev/null
+++ b/test/MC/AArch64/neon-simd-ldst-multi-elem.s
@@ -0,0 +1,463 @@
+// RUN: llvm-mc -triple=aarch64 -mattr=+neon -show-encoding < %s | FileCheck %s
+
+// Check that the assembler can handle the documented syntax for AArch64
+
+//------------------------------------------------------------------------------
+// Store multiple 1-element structures from one register
+//------------------------------------------------------------------------------
+         st1 {v0.16b}, [x0]
+         st1 {v15.8h}, [x15]
+         st1 {v31.4s}, [sp]
+         st1 {v0.2d}, [x0]
+         st1 {v0.8b}, [x0]
+         st1 {v15.4h}, [x15]
+         st1 {v31.2s}, [sp]
+         st1 {v0.1d}, [x0]
+// CHECK:	st1	{v0.16b}, [x0]          // encoding: [0x00,0x70,0x00,0x4c]
+// CHECK:	st1	{v15.8h}, [x15]         // encoding: [0xef,0x75,0x00,0x4c]
+// CHECK:	st1	{v31.4s}, [sp]          // encoding: [0xff,0x7b,0x00,0x4c]
+// CHECK:	st1	{v0.2d}, [x0]           // encoding: [0x00,0x7c,0x00,0x4c]
+// CHECK:	st1	{v0.8b}, [x0]           // encoding: [0x00,0x70,0x00,0x0c]
+// CHECK:	st1	{v15.4h}, [x15]         // encoding: [0xef,0x75,0x00,0x0c]
+// CHECK:	st1	{v31.2s}, [sp]          // encoding: [0xff,0x7b,0x00,0x0c]
+// CHECK:	st1	{v0.1d}, [x0]           // encoding: [0x00,0x7c,0x00,0x0c]
+
+//------------------------------------------------------------------------------
+// Store multiple 1-element structures from two consecutive registers
+//------------------------------------------------------------------------------
+         st1 {v0.16b, v1.16b}, [x0]
+         st1 {v15.8h, v16.8h}, [x15]
+         st1 {v31.4s, v0.4s}, [sp]
+         st1 {v0.2d, v1.2d}, [x0]
+         st1 {v0.8b, v1.8b}, [x0]
+         st1 {v15.4h, v16.4h}, [x15]
+         st1 {v31.2s, v0.2s}, [sp]
+         st1 {v0.1d, v1.1d}, [x0]
+// CHECK:	st1	{v0.16b, v1.16b}, [x0]  // encoding: [0x00,0xa0,0x00,0x4c]
+// CHECK:	st1	{v15.8h, v16.8h}, [x15] // encoding: [0xef,0xa5,0x00,0x4c]
+// CHECK:	st1	{v31.4s, v0.4s}, [sp]   // encoding: [0xff,0xab,0x00,0x4c]
+// CHECK:	st1	{v0.2d, v1.2d}, [x0]    // encoding: [0x00,0xac,0x00,0x4c]
+// CHECK:	st1	{v0.8b, v1.8b}, [x0]    // encoding: [0x00,0xa0,0x00,0x0c]
+// CHECK:	st1	{v15.4h, v16.4h}, [x15] // encoding: [0xef,0xa5,0x00,0x0c]
+// CHECK:	st1	{v31.2s, v0.2s}, [sp]   // encoding: [0xff,0xab,0x00,0x0c]
+// CHECK:	st1	{v0.1d, v1.1d}, [x0]    // encoding: [0x00,0xac,0x00,0x0c]
+
+         st1 {v0.16b-v1.16b}, [x0]
+         st1 {v15.8h-v16.8h}, [x15]
+         st1 {v31.4s-v0.4s}, [sp]
+         st1 {v0.2d-v1.2d}, [x0]
+         st1 {v0.8b-v1.8b}, [x0]
+         st1 {v15.4h-v16.4h}, [x15]
+         st1 {v31.2s-v0.2s}, [sp]
+         st1 {v0.1d-v1.1d}, [x0]
+// CHECK:	st1	{v0.16b, v1.16b}, [x0]  // encoding: [0x00,0xa0,0x00,0x4c]
+// CHECK:	st1	{v15.8h, v16.8h}, [x15] // encoding: [0xef,0xa5,0x00,0x4c]
+// CHECK:	st1	{v31.4s, v0.4s}, [sp]   // encoding: [0xff,0xab,0x00,0x4c]
+// CHECK:	st1	{v0.2d, v1.2d}, [x0]    // encoding: [0x00,0xac,0x00,0x4c]
+// CHECK:	st1	{v0.8b, v1.8b}, [x0]    // encoding: [0x00,0xa0,0x00,0x0c]
+// CHECK:	st1	{v15.4h, v16.4h}, [x15] // encoding: [0xef,0xa5,0x00,0x0c]
+// CHECK:	st1	{v31.2s, v0.2s}, [sp]   // encoding: [0xff,0xab,0x00,0x0c]
+// CHECK:	st1	{v0.1d, v1.1d}, [x0]    // encoding: [0x00,0xac,0x00,0x0c]
+
+//------------------------------------------------------------------------------
+// Store multiple 1-element structures from three consecutive registers
+//------------------------------------------------------------------------------
+         st1 {v0.16b, v1.16b, v2.16b}, [x0]
+         st1 {v15.8h, v16.8h, v17.8h}, [x15]
+         st1 {v31.4s, v0.4s, v1.4s}, [sp]
+         st1 {v0.2d, v1.2d, v2.2d}, [x0]
+         st1 {v0.8b, v1.8b, v2.8b}, [x0]
+         st1 {v15.4h, v16.4h, v17.4h}, [x15]
+         st1 {v31.2s, v0.2s, v1.2s}, [sp]
+         st1 {v0.1d, v1.1d, v2.1d}, [x0]
+// CHECK:	st1	{v0.16b, v1.16b, v2.16b}, [x0] // encoding: [0x00,0x60,0x00,0x4c]
+// CHECK:	st1	{v15.8h, v16.8h, v17.8h}, [x15] // encoding: [0xef,0x65,0x00,0x4c]
+// CHECK:	st1	{v31.4s, v0.4s, v1.4s}, [sp] // encoding: [0xff,0x6b,0x00,0x4c]
+// CHECK:	st1	{v0.2d, v1.2d, v2.2d}, [x0] // encoding: [0x00,0x6c,0x00,0x4c]
+// CHECK:	st1	{v0.8b, v1.8b, v2.8b}, [x0] // encoding: [0x00,0x60,0x00,0x0c]
+// CHECK:	st1	{v15.4h, v16.4h, v17.4h}, [x15] // encoding: [0xef,0x65,0x00,0x0c]
+// CHECK:	st1	{v31.2s, v0.2s, v1.2s}, [sp] // encoding: [0xff,0x6b,0x00,0x0c]
+// CHECK:	st1	{v0.1d, v1.1d, v2.1d}, [x0] // encoding: [0x00,0x6c,0x00,0x0c]
+
+         st1 {v0.16b-v2.16b}, [x0]
+         st1 {v15.8h-v17.8h}, [x15]
+         st1 {v31.4s-v1.4s}, [sp]
+         st1 {v0.2d-v2.2d}, [x0]
+         st1 {v0.8b-v2.8b}, [x0]
+         st1 {v15.4h-v17.4h}, [x15]
+         st1 {v31.2s-v1.2s}, [sp]
+         st1 {v0.1d-v2.1d}, [x0]
+// CHECK:	st1	{v0.16b, v1.16b, v2.16b}, [x0] // encoding: [0x00,0x60,0x00,0x4c]
+// CHECK:	st1	{v15.8h, v16.8h, v17.8h}, [x15] // encoding: [0xef,0x65,0x00,0x4c]
+// CHECK:	st1	{v31.4s, v0.4s, v1.4s}, [sp] // encoding: [0xff,0x6b,0x00,0x4c]
+// CHECK:	st1	{v0.2d, v1.2d, v2.2d}, [x0] // encoding: [0x00,0x6c,0x00,0x4c]
+// CHECK:	st1	{v0.8b, v1.8b, v2.8b}, [x0] // encoding: [0x00,0x60,0x00,0x0c]
+// CHECK:	st1	{v15.4h, v16.4h, v17.4h}, [x15] // encoding: [0xef,0x65,0x00,0x0c]
+// CHECK:	st1	{v31.2s, v0.2s, v1.2s}, [sp] // encoding: [0xff,0x6b,0x00,0x0c]
+// CHECK:	st1	{v0.1d, v1.1d, v2.1d}, [x0] // encoding: [0x00,0x6c,0x00,0x0c]
+
+//------------------------------------------------------------------------------
+// Store multiple 1-element structures from four consecutive registers
+//------------------------------------------------------------------------------
+         st1 {v0.16b, v1.16b, v2.16b, v3.16b}, [x0]
+         st1 {v15.8h, v16.8h, v17.8h, v18.8h}, [x15]
+         st1 {v31.4s, v0.4s, v1.4s, v2.4s}, [sp]
+         st1 {v0.2d, v1.2d, v2.2d, v3.2d}, [x0]
+         st1 {v0.8b, v1.8b, v2.8b, v3.8b}, [x0]
+         st1 {v15.4h, v16.4h, v17.4h, v18.4h}, [x15]
+         st1 {v31.2s, v0.2s, v1.2s, v2.2s}, [sp]
+         st1 {v0.1d, v1.1d, v2.1d, v3.1d}, [x0]
+// CHECK:	st1	{v0.16b, v1.16b, v2.16b, v3.16b}, [x0] // encoding: [0x00,0x20,0x00,0x4c]
+// CHECK:	st1	{v15.8h, v16.8h, v17.8h, v18.8h}, [x15] // encoding: [0xef,0x25,0x00,0x4c]
+// CHECK:	st1	{v31.4s, v0.4s, v1.4s, v2.4s}, [sp] // encoding: [0xff,0x2b,0x00,0x4c]
+// CHECK:	st1	{v0.2d, v1.2d, v2.2d, v3.2d}, [x0] // encoding: [0x00,0x2c,0x00,0x4c]
+// CHECK:	st1	{v0.8b, v1.8b, v2.8b, v3.8b}, [x0] // encoding: [0x00,0x20,0x00,0x0c]
+// CHECK:	st1	{v15.4h, v16.4h, v17.4h, v18.4h}, [x15] // encoding: [0xef,0x25,0x00,0x0c]
+// CHECK:	st1	{v31.2s, v0.2s, v1.2s, v2.2s}, [sp] // encoding: [0xff,0x2b,0x00,0x0c]
+// CHECK:	st1	{v0.1d, v1.1d, v2.1d, v3.1d}, [x0] // encoding: [0x00,0x2c,0x00,0x0c]
+
+         st1 {v0.16b-v3.16b}, [x0]
+         st1 {v15.8h-v18.8h}, [x15]
+         st1 {v31.4s-v2.4s}, [sp]
+         st1 {v0.2d-v3.2d}, [x0]
+         st1 {v0.8b-v3.8b}, [x0]
+         st1 {v15.4h-v18.4h}, [x15]
+         st1 {v31.2s-v2.2s}, [sp]
+         st1 {v0.1d-v3.1d}, [x0]
+// CHECK:	st1	{v0.16b, v1.16b, v2.16b, v3.16b}, [x0] // encoding: [0x00,0x20,0x00,0x4c]
+// CHECK:	st1	{v15.8h, v16.8h, v17.8h, v18.8h}, [x15] // encoding: [0xef,0x25,0x00,0x4c]
+// CHECK:	st1	{v31.4s, v0.4s, v1.4s, v2.4s}, [sp] // encoding: [0xff,0x2b,0x00,0x4c]
+// CHECK:	st1	{v0.2d, v1.2d, v2.2d, v3.2d}, [x0] // encoding: [0x00,0x2c,0x00,0x4c]
+// CHECK:	st1	{v0.8b, v1.8b, v2.8b, v3.8b}, [x0] // encoding: [0x00,0x20,0x00,0x0c]
+// CHECK:	st1	{v15.4h, v16.4h, v17.4h, v18.4h}, [x15] // encoding: [0xef,0x25,0x00,0x0c]
+// CHECK:	st1	{v31.2s, v0.2s, v1.2s, v2.2s}, [sp] // encoding: [0xff,0x2b,0x00,0x0c]
+// CHECK:	st1	{v0.1d, v1.1d, v2.1d, v3.1d}, [x0] // encoding: [0x00,0x2c,0x00,0x0c]
+
+//------------------------------------------------------------------------------
+// Store multiple 2-element structures from two consecutive registers
+//------------------------------------------------------------------------------
+         st2 {v0.16b, v1.16b}, [x0]
+         st2 {v15.8h, v16.8h}, [x15]
+         st2 {v31.4s, v0.4s}, [sp]
+         st2 {v0.2d, v1.2d}, [x0]
+         st2 {v0.8b, v1.8b}, [x0]
+         st2 {v15.4h, v16.4h}, [x15]
+         st2 {v31.2s, v0.2s}, [sp]
+// CHECK:	st2	{v0.16b, v1.16b}, [x0]  // encoding: [0x00,0x80,0x00,0x4c]
+// CHECK:	st2	{v15.8h, v16.8h}, [x15] // encoding: [0xef,0x85,0x00,0x4c]
+// CHECK:	st2	{v31.4s, v0.4s}, [sp]   // encoding: [0xff,0x8b,0x00,0x4c]
+// CHECK:	st2	{v0.2d, v1.2d}, [x0]    // encoding: [0x00,0x8c,0x00,0x4c]
+// CHECK:	st2	{v0.8b, v1.8b}, [x0]    // encoding: [0x00,0x80,0x00,0x0c]
+// CHECK:	st2	{v15.4h, v16.4h}, [x15] // encoding: [0xef,0x85,0x00,0x0c]
+// CHECK:	st2	{v31.2s, v0.2s}, [sp]   // encoding: [0xff,0x8b,0x00,0x0c]
+
+         st2 {v0.16b-v1.16b}, [x0]
+         st2 {v15.8h-v16.8h}, [x15]
+         st2 {v31.4s-v0.4s}, [sp]
+         st2 {v0.2d-v1.2d}, [x0]
+         st2 {v0.8b-v1.8b}, [x0]
+         st2 {v15.4h-v16.4h}, [x15]
+         st2 {v31.2s-v0.2s}, [sp]
+// CHECK:	st2	{v0.16b, v1.16b}, [x0]  // encoding: [0x00,0x80,0x00,0x4c]
+// CHECK:	st2	{v15.8h, v16.8h}, [x15] // encoding: [0xef,0x85,0x00,0x4c]
+// CHECK:	st2	{v31.4s, v0.4s}, [sp]   // encoding: [0xff,0x8b,0x00,0x4c]
+// CHECK:	st2	{v0.2d, v1.2d}, [x0]    // encoding: [0x00,0x8c,0x00,0x4c]
+// CHECK:	st2	{v0.8b, v1.8b}, [x0]    // encoding: [0x00,0x80,0x00,0x0c]
+// CHECK:	st2	{v15.4h, v16.4h}, [x15] // encoding: [0xef,0x85,0x00,0x0c]
+// CHECK:	st2	{v31.2s, v0.2s}, [sp]   // encoding: [0xff,0x8b,0x00,0x0c]
+
+//------------------------------------------------------------------------------
+// Store multiple 3-element structures from three consecutive registers
+//------------------------------------------------------------------------------
+         st3 {v0.16b, v1.16b, v2.16b}, [x0]
+         st3 {v15.8h, v16.8h, v17.8h}, [x15]
+         st3 {v31.4s, v0.4s, v1.4s}, [sp]
+         st3 {v0.2d, v1.2d, v2.2d}, [x0]
+         st3 {v0.8b, v1.8b, v2.8b}, [x0]
+         st3 {v15.4h, v16.4h, v17.4h}, [x15]
+         st3 {v31.2s, v0.2s, v1.2s}, [sp]
+// CHECK:	st3	{v0.16b, v1.16b, v2.16b}, [x0] // encoding: [0x00,0x40,0x00,0x4c]
+// CHECK:	st3	{v15.8h, v16.8h, v17.8h}, [x15] // encoding: [0xef,0x45,0x00,0x4c]
+// CHECK:	st3	{v31.4s, v0.4s, v1.4s}, [sp] // encoding: [0xff,0x4b,0x00,0x4c]
+// CHECK:	st3	{v0.2d, v1.2d, v2.2d}, [x0] // encoding: [0x00,0x4c,0x00,0x4c]
+// CHECK:	st3	{v0.8b, v1.8b, v2.8b}, [x0] // encoding: [0x00,0x40,0x00,0x0c]
+// CHECK:	st3	{v15.4h, v16.4h, v17.4h}, [x15] // encoding: [0xef,0x45,0x00,0x0c]
+// CHECK:	st3	{v31.2s, v0.2s, v1.2s}, [sp] // encoding: [0xff,0x4b,0x00,0x0c]
+
+         st3 {v0.16b-v2.16b}, [x0]
+         st3 {v15.8h-v17.8h}, [x15]
+         st3 {v31.4s-v1.4s}, [sp]
+         st3 {v0.2d-v2.2d}, [x0]
+         st3 {v0.8b-v2.8b}, [x0]
+         st3 {v15.4h-v17.4h}, [x15]
+         st3 {v31.2s-v1.2s}, [sp]
+// CHECK:	st3	{v0.16b, v1.16b, v2.16b}, [x0] // encoding: [0x00,0x40,0x00,0x4c]
+// CHECK:	st3	{v15.8h, v16.8h, v17.8h}, [x15] // encoding: [0xef,0x45,0x00,0x4c]
+// CHECK:	st3	{v31.4s, v0.4s, v1.4s}, [sp] // encoding: [0xff,0x4b,0x00,0x4c]
+// CHECK:	st3	{v0.2d, v1.2d, v2.2d}, [x0] // encoding: [0x00,0x4c,0x00,0x4c]
+// CHECK:	st3	{v0.8b, v1.8b, v2.8b}, [x0] // encoding: [0x00,0x40,0x00,0x0c]
+// CHECK:	st3	{v15.4h, v16.4h, v17.4h}, [x15] // encoding: [0xef,0x45,0x00,0x0c]
+// CHECK:	st3	{v31.2s, v0.2s, v1.2s}, [sp] // encoding: [0xff,0x4b,0x00,0x0c]
+
+//------------------------------------------------------------------------------
+// Store multiple 4-element structures from four consecutive registers
+//------------------------------------------------------------------------------
+         st4 {v0.16b, v1.16b, v2.16b, v3.16b}, [x0]
+         st4 {v15.8h, v16.8h, v17.8h, v18.8h}, [x15]
+         st4 {v31.4s, v0.4s, v1.4s, v2.4s}, [sp]
+         st4 {v0.2d, v1.2d, v2.2d, v3.2d}, [x0]
+         st4 {v0.8b, v1.8b, v2.8b, v3.8b}, [x0]
+         st4 {v15.4h, v16.4h, v17.4h, v18.4h}, [x15]
+         st4 {v31.2s, v0.2s, v1.2s, v2.2s}, [sp]
+// CHECK:	st4	{v0.16b, v1.16b, v2.16b, v3.16b}, [x0] // encoding: [0x00,0x00,0x00,0x4c]
+// CHECK:	st4	{v15.8h, v16.8h, v17.8h, v18.8h}, [x15] // encoding: [0xef,0x05,0x00,0x4c]
+// CHECK:	st4	{v31.4s, v0.4s, v1.4s, v2.4s}, [sp] // encoding: [0xff,0x0b,0x00,0x4c]
+// CHECK:	st4	{v0.2d, v1.2d, v2.2d, v3.2d}, [x0] // encoding: [0x00,0x0c,0x00,0x4c]
+// CHECK:	st4	{v0.8b, v1.8b, v2.8b, v3.8b}, [x0] // encoding: [0x00,0x00,0x00,0x0c]
+// CHECK:	st4	{v15.4h, v16.4h, v17.4h, v18.4h}, [x15] // encoding: [0xef,0x05,0x00,0x0c]
+// CHECK:	st4	{v31.2s, v0.2s, v1.2s, v2.2s}, [sp] // encoding: [0xff,0x0b,0x00,0x0c]
+
+         st4 {v0.16b-v3.16b}, [x0]
+         st4 {v15.8h-v18.8h}, [x15]
+         st4 {v31.4s-v2.4s}, [sp]
+         st4 {v0.2d-v3.2d}, [x0]
+         st4 {v0.8b-v3.8b}, [x0]
+         st4 {v15.4h-v18.4h}, [x15]
+         st4 {v31.2s-v2.2s}, [sp]
+// CHECK:	st4	{v0.16b, v1.16b, v2.16b, v3.16b}, [x0] // encoding: [0x00,0x00,0x00,0x4c]
+// CHECK:	st4	{v15.8h, v16.8h, v17.8h, v18.8h}, [x15] // encoding: [0xef,0x05,0x00,0x4c]
+// CHECK:	st4	{v31.4s, v0.4s, v1.4s, v2.4s}, [sp] // encoding: [0xff,0x0b,0x00,0x4c]
+// CHECK:	st4	{v0.2d, v1.2d, v2.2d, v3.2d}, [x0] // encoding: [0x00,0x0c,0x00,0x4c]
+// CHECK:	st4	{v0.8b, v1.8b, v2.8b, v3.8b}, [x0] // encoding: [0x00,0x00,0x00,0x0c]
+// CHECK:	st4	{v15.4h, v16.4h, v17.4h, v18.4h}, [x15] // encoding: [0xef,0x05,0x00,0x0c]
+// CHECK:	st4	{v31.2s, v0.2s, v1.2s, v2.2s}, [sp] // encoding: [0xff,0x0b,0x00,0x0c]
+
+//------------------------------------------------------------------------------
+// Load multiple 1-element structures to one register
+//------------------------------------------------------------------------------
+         ld1 {v0.16b}, [x0]
+         ld1 {v15.8h}, [x15]
+         ld1 {v31.4s}, [sp]
+         ld1 {v0.2d}, [x0]
+         ld1 {v0.8b}, [x0]
+         ld1 {v15.4h}, [x15]
+         ld1 {v31.2s}, [sp]
+         ld1 {v0.1d}, [x0]
+// CHECK:	ld1	{v0.16b}, [x0]          // encoding: [0x00,0x70,0x40,0x4c]
+// CHECK:	ld1	{v15.8h}, [x15]         // encoding: [0xef,0x75,0x40,0x4c]
+// CHECK:	ld1	{v31.4s}, [sp]          // encoding: [0xff,0x7b,0x40,0x4c]
+// CHECK:	ld1	{v0.2d}, [x0]           // encoding: [0x00,0x7c,0x40,0x4c]
+// CHECK:	ld1	{v0.8b}, [x0]           // encoding: [0x00,0x70,0x40,0x0c]
+// CHECK:	ld1	{v15.4h}, [x15]         // encoding: [0xef,0x75,0x40,0x0c]
+// CHECK:	ld1	{v31.2s}, [sp]          // encoding: [0xff,0x7b,0x40,0x0c]
+// CHECK:	ld1	{v0.1d}, [x0]           // encoding: [0x00,0x7c,0x40,0x0c]
+
+//------------------------------------------------------------------------------
+// Load multiple 1-element structures to two consecutive registers
+//------------------------------------------------------------------------------
+         ld1 {v0.16b, v1.16b}, [x0]
+         ld1 {v15.8h, v16.8h}, [x15]
+         ld1 {v31.4s, v0.4s}, [sp]
+         ld1 {v0.2d, v1.2d}, [x0]
+         ld1 {v0.8b, v1.8b}, [x0]
+         ld1 {v15.4h, v16.4h}, [x15]
+         ld1 {v31.2s, v0.2s}, [sp]
+         ld1 {v0.1d, v1.1d}, [x0]
+// CHECK:	ld1	{v0.16b, v1.16b}, [x0]  // encoding: [0x00,0xa0,0x40,0x4c]
+// CHECK:	ld1	{v15.8h, v16.8h}, [x15] // encoding: [0xef,0xa5,0x40,0x4c]
+// CHECK:	ld1	{v31.4s, v0.4s}, [sp]   // encoding: [0xff,0xab,0x40,0x4c]
+// CHECK:	ld1	{v0.2d, v1.2d}, [x0]    // encoding: [0x00,0xac,0x40,0x4c]
+// CHECK:	ld1	{v0.8b, v1.8b}, [x0]    // encoding: [0x00,0xa0,0x40,0x0c]
+// CHECK:	ld1	{v15.4h, v16.4h}, [x15] // encoding: [0xef,0xa5,0x40,0x0c]
+// CHECK:	ld1	{v31.2s, v0.2s}, [sp]   // encoding: [0xff,0xab,0x40,0x0c]
+// CHECK:	ld1	{v0.1d, v1.1d}, [x0]    // encoding: [0x00,0xac,0x40,0x0c]
+
+         ld1 {v0.16b-v1.16b}, [x0]
+         ld1 {v15.8h-v16.8h}, [x15]
+         ld1 {v31.4s-v0.4s}, [sp]
+         ld1 {v0.2d-v1.2d}, [x0]
+         ld1 {v0.8b-v1.8b}, [x0]
+         ld1 {v15.4h-v16.4h}, [x15]
+         ld1 {v31.2s-v0.2s}, [sp]
+         ld1 {v0.1d-v1.1d}, [x0]
+// CHECK:	ld1	{v0.16b, v1.16b}, [x0]  // encoding: [0x00,0xa0,0x40,0x4c]
+// CHECK:	ld1	{v15.8h, v16.8h}, [x15] // encoding: [0xef,0xa5,0x40,0x4c]
+// CHECK:	ld1	{v31.4s, v0.4s}, [sp]   // encoding: [0xff,0xab,0x40,0x4c]
+// CHECK:	ld1	{v0.2d, v1.2d}, [x0]    // encoding: [0x00,0xac,0x40,0x4c]
+// CHECK:	ld1	{v0.8b, v1.8b}, [x0]    // encoding: [0x00,0xa0,0x40,0x0c]
+// CHECK:	ld1	{v15.4h, v16.4h}, [x15] // encoding: [0xef,0xa5,0x40,0x0c]
+// CHECK:	ld1	{v31.2s, v0.2s}, [sp]   // encoding: [0xff,0xab,0x40,0x0c]
+// CHECK:	ld1	{v0.1d, v1.1d}, [x0]    // encoding: [0x00,0xac,0x40,0x0c]
+
+//------------------------------------------------------------------------------
+// Load multiple 1-element structures to three consecutive registers
+//------------------------------------------------------------------------------
+         ld1 {v0.16b, v1.16b, v2.16b}, [x0]
+         ld1 {v15.8h, v16.8h, v17.8h}, [x15]
+         ld1 {v31.4s, v0.4s, v1.4s}, [sp]
+         ld1 {v0.2d, v1.2d, v2.2d}, [x0]
+         ld1 {v0.8b, v1.8b, v2.8b}, [x0]
+         ld1 {v15.4h, v16.4h, v17.4h}, [x15]
+         ld1 {v31.2s, v0.2s, v1.2s}, [sp]
+         ld1 {v0.1d, v1.1d, v2.1d}, [x0]
+// CHECK:	ld1	{v0.16b, v1.16b, v2.16b}, [x0] // encoding: [0x00,0x60,0x40,0x4c]
+// CHECK:	ld1	{v15.8h, v16.8h, v17.8h}, [x15] // encoding: [0xef,0x65,0x40,0x4c]
+// CHECK:	ld1	{v31.4s, v0.4s, v1.4s}, [sp] // encoding: [0xff,0x6b,0x40,0x4c]
+// CHECK:	ld1	{v0.2d, v1.2d, v2.2d}, [x0] // encoding: [0x00,0x6c,0x40,0x4c]
+// CHECK:	ld1	{v0.8b, v1.8b, v2.8b}, [x0] // encoding: [0x00,0x60,0x40,0x0c]
+// CHECK:	ld1	{v15.4h, v16.4h, v17.4h}, [x15] // encoding: [0xef,0x65,0x40,0x0c]
+// CHECK:	ld1	{v31.2s, v0.2s, v1.2s}, [sp] // encoding: [0xff,0x6b,0x40,0x0c]
+// CHECK:	ld1	{v0.1d, v1.1d, v2.1d}, [x0] // encoding: [0x00,0x6c,0x40,0x0c]
+
+         ld1 {v0.16b-v2.16b}, [x0]
+         ld1 {v15.8h-v17.8h}, [x15]
+         ld1 {v31.4s-v1.4s}, [sp]
+         ld1 {v0.2d-v2.2d}, [x0]
+         ld1 {v0.8b-v2.8b}, [x0]
+         ld1 {v15.4h-v17.4h}, [x15]
+         ld1 {v31.2s-v1.2s}, [sp]
+         ld1 {v0.1d-v2.1d}, [x0]
+// CHECK:	ld1	{v0.16b, v1.16b, v2.16b}, [x0] // encoding: [0x00,0x60,0x40,0x4c]
+// CHECK:	ld1	{v15.8h, v16.8h, v17.8h}, [x15] // encoding: [0xef,0x65,0x40,0x4c]
+// CHECK:	ld1	{v31.4s, v0.4s, v1.4s}, [sp] // encoding: [0xff,0x6b,0x40,0x4c]
+// CHECK:	ld1	{v0.2d, v1.2d, v2.2d}, [x0] // encoding: [0x00,0x6c,0x40,0x4c]
+// CHECK:	ld1	{v0.8b, v1.8b, v2.8b}, [x0] // encoding: [0x00,0x60,0x40,0x0c]
+// CHECK:	ld1	{v15.4h, v16.4h, v17.4h}, [x15] // encoding: [0xef,0x65,0x40,0x0c]
+// CHECK:	ld1	{v31.2s, v0.2s, v1.2s}, [sp] // encoding: [0xff,0x6b,0x40,0x0c]
+// CHECK:	ld1	{v0.1d, v1.1d, v2.1d}, [x0] // encoding: [0x00,0x6c,0x40,0x0c]
+
+//------------------------------------------------------------------------------
+// Load multiple 1-element structures to four consecutive registers
+//------------------------------------------------------------------------------
+         ld1 {v0.16b, v1.16b, v2.16b, v3.16b}, [x0]
+         ld1 {v15.8h, v16.8h, v17.8h, v18.8h}, [x15]
+         ld1 {v31.4s, v0.4s, v1.4s, v2.4s}, [sp]
+         ld1 {v0.2d, v1.2d, v2.2d, v3.2d}, [x0]
+         ld1 {v0.8b, v1.8b, v2.8b, v3.8b}, [x0]
+         ld1 {v15.4h, v16.4h, v17.4h, v18.4h}, [x15]
+         ld1 {v31.2s, v0.2s, v1.2s, v2.2s}, [sp]
+         ld1 {v0.1d, v1.1d, v2.1d, v3.1d}, [x0]
+// CHECK:	ld1	{v0.16b, v1.16b, v2.16b, v3.16b}, [x0] // encoding: [0x00,0x20,0x40,0x4c]
+// CHECK:	ld1	{v15.8h, v16.8h, v17.8h, v18.8h}, [x15] // encoding: [0xef,0x25,0x40,0x4c]
+// CHECK:	ld1	{v31.4s, v0.4s, v1.4s, v2.4s}, [sp] // encoding: [0xff,0x2b,0x40,0x4c]
+// CHECK:	ld1	{v0.2d, v1.2d, v2.2d, v3.2d}, [x0] // encoding: [0x00,0x2c,0x40,0x4c]
+// CHECK:	ld1	{v0.8b, v1.8b, v2.8b, v3.8b}, [x0] // encoding: [0x00,0x20,0x40,0x0c]
+// CHECK:	ld1	{v15.4h, v16.4h, v17.4h, v18.4h}, [x15] // encoding: [0xef,0x25,0x40,0x0c]
+// CHECK:	ld1	{v31.2s, v0.2s, v1.2s, v2.2s}, [sp] // encoding: [0xff,0x2b,0x40,0x0c]
+// CHECK:	ld1	{v0.1d, v1.1d, v2.1d, v3.1d}, [x0] // encoding: [0x00,0x2c,0x40,0x0c]
+
+         ld1 {v0.16b-v3.16b}, [x0]
+         ld1 {v15.8h-v18.8h}, [x15]
+         ld1 {v31.4s-v2.4s}, [sp]
+         ld1 {v0.2d-v3.2d}, [x0]
+         ld1 {v0.8b-v3.8b}, [x0]
+         ld1 {v15.4h-v18.4h}, [x15]
+         ld1 {v31.2s-v2.2s}, [sp]
+         ld1 {v0.1d-v3.1d}, [x0]
+// CHECK:	ld1	{v0.16b, v1.16b, v2.16b, v3.16b}, [x0] // encoding: [0x00,0x20,0x40,0x4c]
+// CHECK:	ld1	{v15.8h, v16.8h, v17.8h, v18.8h}, [x15] // encoding: [0xef,0x25,0x40,0x4c]
+// CHECK:	ld1	{v31.4s, v0.4s, v1.4s, v2.4s}, [sp] // encoding: [0xff,0x2b,0x40,0x4c]
+// CHECK:	ld1	{v0.2d, v1.2d, v2.2d, v3.2d}, [x0] // encoding: [0x00,0x2c,0x40,0x4c]
+// CHECK:	ld1	{v0.8b, v1.8b, v2.8b, v3.8b}, [x0] // encoding: [0x00,0x20,0x40,0x0c]
+// CHECK:	ld1	{v15.4h, v16.4h, v17.4h, v18.4h}, [x15] // encoding: [0xef,0x25,0x40,0x0c]
+// CHECK:	ld1	{v31.2s, v0.2s, v1.2s, v2.2s}, [sp] // encoding: [0xff,0x2b,0x40,0x0c]
+// CHECK:	ld1	{v0.1d, v1.1d, v2.1d, v3.1d}, [x0] // encoding: [0x00,0x2c,0x40,0x0c]
+
+//------------------------------------------------------------------------------
+// Load multiple 4-element structures to two consecutive registers
+//------------------------------------------------------------------------------
+         ld2 {v0.16b, v1.16b}, [x0]
+         ld2 {v15.8h, v16.8h}, [x15]
+         ld2 {v31.4s, v0.4s}, [sp]
+         ld2 {v0.2d, v1.2d}, [x0]
+         ld2 {v0.8b, v1.8b}, [x0]
+         ld2 {v15.4h, v16.4h}, [x15]
+         ld2 {v31.2s, v0.2s}, [sp]
+// CHECK:	ld2	{v0.16b, v1.16b}, [x0]  // encoding: [0x00,0x80,0x40,0x4c]
+// CHECK:	ld2	{v15.8h, v16.8h}, [x15] // encoding: [0xef,0x85,0x40,0x4c]
+// CHECK:	ld2	{v31.4s, v0.4s}, [sp]   // encoding: [0xff,0x8b,0x40,0x4c]
+// CHECK:	ld2	{v0.2d, v1.2d}, [x0]    // encoding: [0x00,0x8c,0x40,0x4c]
+// CHECK:	ld2	{v0.8b, v1.8b}, [x0]    // encoding: [0x00,0x80,0x40,0x0c]
+// CHECK:	ld2	{v15.4h, v16.4h}, [x15] // encoding: [0xef,0x85,0x40,0x0c]
+// CHECK:	ld2	{v31.2s, v0.2s}, [sp]   // encoding: [0xff,0x8b,0x40,0x0c]
+
+         ld2 {v0.16b-v1.16b}, [x0]
+         ld2 {v15.8h-v16.8h}, [x15]
+         ld2 {v31.4s-v0.4s}, [sp]
+         ld2 {v0.2d-v1.2d}, [x0]
+         ld2 {v0.8b-v1.8b}, [x0]
+         ld2 {v15.4h-v16.4h}, [x15]
+         ld2 {v31.2s-v0.2s}, [sp]
+// CHECK:	ld2	{v0.16b, v1.16b}, [x0]  // encoding: [0x00,0x80,0x40,0x4c]
+// CHECK:	ld2	{v15.8h, v16.8h}, [x15] // encoding: [0xef,0x85,0x40,0x4c]
+// CHECK:	ld2	{v31.4s, v0.4s}, [sp]   // encoding: [0xff,0x8b,0x40,0x4c]
+// CHECK:	ld2	{v0.2d, v1.2d}, [x0]    // encoding: [0x00,0x8c,0x40,0x4c]
+// CHECK:	ld2	{v0.8b, v1.8b}, [x0]    // encoding: [0x00,0x80,0x40,0x0c]
+// CHECK:	ld2	{v15.4h, v16.4h}, [x15] // encoding: [0xef,0x85,0x40,0x0c]
+// CHECK:	ld2	{v31.2s, v0.2s}, [sp]   // encoding: [0xff,0x8b,0x40,0x0c]
+
+//------------------------------------------------------------------------------
+// Load multiple 3-element structures to three consecutive registers
+//------------------------------------------------------------------------------
+         ld3 {v0.16b, v1.16b, v2.16b}, [x0]
+         ld3 {v15.8h, v16.8h, v17.8h}, [x15]
+         ld3 {v31.4s, v0.4s, v1.4s}, [sp]
+         ld3 {v0.2d, v1.2d, v2.2d}, [x0]
+         ld3 {v0.8b, v1.8b, v2.8b}, [x0]
+         ld3 {v15.4h, v16.4h, v17.4h}, [x15]
+         ld3 {v31.2s, v0.2s, v1.2s}, [sp]
+// CHECK:	ld3	{v0.16b, v1.16b, v2.16b}, [x0] // encoding: [0x00,0x40,0x40,0x4c]
+// CHECK:	ld3	{v15.8h, v16.8h, v17.8h}, [x15] // encoding: [0xef,0x45,0x40,0x4c]
+// CHECK:	ld3	{v31.4s, v0.4s, v1.4s}, [sp] // encoding: [0xff,0x4b,0x40,0x4c]
+// CHECK:	ld3	{v0.2d, v1.2d, v2.2d}, [x0] // encoding: [0x00,0x4c,0x40,0x4c]
+// CHECK:	ld3	{v0.8b, v1.8b, v2.8b}, [x0] // encoding: [0x00,0x40,0x40,0x0c]
+// CHECK:	ld3	{v15.4h, v16.4h, v17.4h}, [x15] // encoding: [0xef,0x45,0x40,0x0c]
+// CHECK:	ld3	{v31.2s, v0.2s, v1.2s}, [sp] // encoding: [0xff,0x4b,0x40,0x0c]
+
+         ld3 {v0.16b-v2.16b}, [x0]
+         ld3 {v15.8h-v17.8h}, [x15]
+         ld3 {v31.4s-v1.4s}, [sp]
+         ld3 {v0.2d-v2.2d}, [x0]
+         ld3 {v0.8b-v2.8b}, [x0]
+         ld3 {v15.4h-v17.4h}, [x15]
+         ld3 {v31.2s-v1.2s}, [sp]
+// CHECK:	ld3	{v0.16b, v1.16b, v2.16b}, [x0] // encoding: [0x00,0x40,0x40,0x4c]
+// CHECK:	ld3	{v15.8h, v16.8h, v17.8h}, [x15] // encoding: [0xef,0x45,0x40,0x4c]
+// CHECK:	ld3	{v31.4s, v0.4s, v1.4s}, [sp] // encoding: [0xff,0x4b,0x40,0x4c]
+// CHECK:	ld3	{v0.2d, v1.2d, v2.2d}, [x0] // encoding: [0x00,0x4c,0x40,0x4c]
+// CHECK:	ld3	{v0.8b, v1.8b, v2.8b}, [x0] // encoding: [0x00,0x40,0x40,0x0c]
+// CHECK:	ld3	{v15.4h, v16.4h, v17.4h}, [x15] // encoding: [0xef,0x45,0x40,0x0c]
+// CHECK:	ld3	{v31.2s, v0.2s, v1.2s}, [sp] // encoding: [0xff,0x4b,0x40,0x0c]
+
+//------------------------------------------------------------------------------
+// Load multiple 4-element structures to four consecutive registers
+//------------------------------------------------------------------------------
+         ld4 {v0.16b, v1.16b, v2.16b, v3.16b}, [x0]
+         ld4 {v15.8h, v16.8h, v17.8h, v18.8h}, [x15]
+         ld4 {v31.4s, v0.4s, v1.4s, v2.4s}, [sp]
+         ld4 {v0.2d, v1.2d, v2.2d, v3.2d}, [x0]
+         ld4 {v0.8b, v1.8b, v2.8b, v3.8b}, [x0]
+         ld4 {v15.4h, v16.4h, v17.4h, v18.4h}, [x15]
+         ld4 {v31.2s, v0.2s, v1.2s, v2.2s}, [sp]
+// CHECK:	ld4	{v0.16b, v1.16b, v2.16b, v3.16b}, [x0] // encoding: [0x00,0x00,0x40,0x4c]
+// CHECK:	ld4	{v15.8h, v16.8h, v17.8h, v18.8h}, [x15] // encoding: [0xef,0x05,0x40,0x4c]
+// CHECK:	ld4	{v31.4s, v0.4s, v1.4s, v2.4s}, [sp] // encoding: [0xff,0x0b,0x40,0x4c]
+// CHECK:	ld4	{v0.2d, v1.2d, v2.2d, v3.2d}, [x0] // encoding: [0x00,0x0c,0x40,0x4c]
+// CHECK:	ld4	{v0.8b, v1.8b, v2.8b, v3.8b}, [x0] // encoding: [0x00,0x00,0x40,0x0c]
+// CHECK:	ld4	{v15.4h, v16.4h, v17.4h, v18.4h}, [x15] // encoding: [0xef,0x05,0x40,0x0c]
+// CHECK:	ld4	{v31.2s, v0.2s, v1.2s, v2.2s}, [sp] // encoding: [0xff,0x0b,0x40,0x0c]
+
+         ld4 {v0.16b-v3.16b}, [x0]
+         ld4 {v15.8h-v18.8h}, [x15]
+         ld4 {v31.4s-v2.4s}, [sp]
+         ld4 {v0.2d-v3.2d}, [x0]
+         ld4 {v0.8b-v3.8b}, [x0]
+         ld4 {v15.4h-v18.4h}, [x15]
+         ld4 {v31.2s-v2.2s}, [sp]
+// CHECK:	ld4	{v0.16b, v1.16b, v2.16b, v3.16b}, [x0] // encoding: [0x00,0x00,0x40,0x4c]
+// CHECK:	ld4	{v15.8h, v16.8h, v17.8h, v18.8h}, [x15] // encoding: [0xef,0x05,0x40,0x4c]
+// CHECK:	ld4	{v31.4s, v0.4s, v1.4s, v2.4s}, [sp] // encoding: [0xff,0x0b,0x40,0x4c]
+// CHECK:	ld4	{v0.2d, v1.2d, v2.2d, v3.2d}, [x0] // encoding: [0x00,0x0c,0x40,0x4c]
+// CHECK:	ld4	{v0.8b, v1.8b, v2.8b, v3.8b}, [x0] // encoding: [0x00,0x00,0x40,0x0c]
+// CHECK:	ld4	{v15.4h, v16.4h, v17.4h, v18.4h}, [x15] // encoding: [0xef,0x05,0x40,0x0c]
+// CHECK:	ld4	{v31.2s, v0.2s, v1.2s, v2.2s}, [sp] // encoding: [0xff,0x0b,0x40,0x0c]