// Copyright (c) 2013, the Dart project authors.  Please see the AUTHORS file
// for details. All rights reserved. Use of this source code is governed by a
// BSD-style license that can be found in the LICENSE file.

#ifndef RUNTIME_VM_COMPILER_ASSEMBLER_ASSEMBLER_ARM_H_
#define RUNTIME_VM_COMPILER_ASSEMBLER_ASSEMBLER_ARM_H_

#ifndef RUNTIME_VM_COMPILER_ASSEMBLER_ASSEMBLER_H_
#error Do not include assembler_arm.h directly; use assembler.h instead.
#endif

#include <functional>

#include "platform/assert.h"
#include "platform/utils.h"
#include "vm/code_entry_kind.h"
#include "vm/compiler/runtime_api.h"
#include "vm/constants_arm.h"
#include "vm/cpu.h"
#include "vm/hash_map.h"
#include "vm/simulator.h"

namespace dart {

// Forward declarations.
class FlowGraphCompiler;
class RegisterSet;
class RuntimeEntry;

// TODO(vegorov) these enumerations are temporarily moved out of compiler
// namespace to make refactoring easier.
enum OperandSize {
  kByte,
  kUnsignedByte,
  kHalfword,
  kUnsignedHalfword,
  kWord,
  kUnsignedWord,
  kWordPair,
  kSWord,
  kDWord,
  kRegList,
};

// Load/store multiple addressing mode.
enum BlockAddressMode {
  // bit encoding P U W
  DA = (0 | 0 | 0) << 21,    // decrement after
  IA = (0 | 4 | 0) << 21,    // increment after
  DB = (8 | 0 | 0) << 21,    // decrement before
  IB = (8 | 4 | 0) << 21,    // increment before
  DA_W = (0 | 0 | 1) << 21,  // decrement after with writeback to base
  IA_W = (0 | 4 | 1) << 21,  // increment after with writeback to base
  DB_W = (8 | 0 | 1) << 21,  // decrement before with writeback to base
  IB_W = (8 | 4 | 1) << 21   // increment before with writeback to base
};

namespace compiler {

// Instruction encoding bits.
enum {
  H = 1 << 5,   // halfword (or byte)
  L = 1 << 20,  // load (or store)
  S = 1 << 20,  // set condition code (or leave unchanged)
  W = 1 << 21,  // writeback base register (or leave unchanged)
  A = 1 << 21,  // accumulate in multiply instruction (or not)
  B = 1 << 22,  // unsigned byte (or word)
  D = 1 << 22,  // high/lo bit of start of s/d register range
  N = 1 << 22,  // long (or short)
  U = 1 << 23,  // positive (or negative) offset/index
  P = 1 << 24,  // offset/pre-indexed addressing (or post-indexed addressing)
  I = 1 << 25,  // immediate shifter operand (or not)

  B0 = 1,
  B1 = 1 << 1,
  B2 = 1 << 2,
  B3 = 1 << 3,
  B4 = 1 << 4,
  B5 = 1 << 5,
  B6 = 1 << 6,
  B7 = 1 << 7,
  B8 = 1 << 8,
  B9 = 1 << 9,
  B10 = 1 << 10,
  B11 = 1 << 11,
  B12 = 1 << 12,
  B16 = 1 << 16,
  B17 = 1 << 17,
  B18 = 1 << 18,
  B19 = 1 << 19,
  B20 = 1 << 20,
  B21 = 1 << 21,
  B22 = 1 << 22,
  B23 = 1 << 23,
  B24 = 1 << 24,
  B25 = 1 << 25,
  B26 = 1 << 26,
  B27 = 1 << 27,
};

class ArmEncode : public AllStatic {
 public:
  static inline uint32_t Rd(Register rd) {
    ASSERT(rd < 16);
    return static_cast<uint32_t>(rd) << kRdShift;
  }

  static inline uint32_t Rm(Register rm) {
    ASSERT(rm < 16);
    return static_cast<uint32_t>(rm) << kRmShift;
  }

  static inline uint32_t Rn(Register rn) {
    ASSERT(rn < 16);
    return static_cast<uint32_t>(rn) << kRnShift;
  }

  static inline uint32_t Rs(Register rs) {
    ASSERT(rs < 16);
    return static_cast<uint32_t>(rs) << kRsShift;
  }
};

// Encodes Addressing Mode 1 - Data-processing operands.
class Operand : public ValueObject {
 public:
  // Data-processing operands - Uninitialized.
  Operand() : type_(-1), encoding_(-1) {}

  // Data-processing operands - Copy constructor.
  Operand(const Operand& other)
      : ValueObject(), type_(other.type_), encoding_(other.encoding_) {}

  // Data-processing operands - Assignment operator.
  Operand& operator=(const Operand& other) {
    type_ = other.type_;
    encoding_ = other.encoding_;
    return *this;
  }

  // Data-processing operands - Immediate.
  explicit Operand(uint32_t immediate) {
    ASSERT(immediate < (1 << kImmed8Bits));
    type_ = 1;
    encoding_ = immediate;
  }

  // Data-processing operands - Rotated immediate.
  Operand(uint32_t rotate, uint32_t immed8) {
    ASSERT((rotate < (1 << kRotateBits)) && (immed8 < (1 << kImmed8Bits)));
    type_ = 1;
    encoding_ = (rotate << kRotateShift) | (immed8 << kImmed8Shift);
  }

  // Data-processing operands - Register.
  explicit Operand(Register rm) {
    type_ = 0;
    encoding_ = static_cast<uint32_t>(rm);
  }

  // Data-processing operands - Logical shift/rotate by immediate.
  Operand(Register rm, Shift shift, uint32_t shift_imm) {
    ASSERT(shift_imm < (1 << kShiftImmBits));
    type_ = 0;
    encoding_ = shift_imm << kShiftImmShift |
                static_cast<uint32_t>(shift) << kShiftShift |
                static_cast<uint32_t>(rm);
  }

  // Data-processing operands - Logical shift/rotate by register.
  Operand(Register rm, Shift shift, Register rs) {
    type_ = 0;
    encoding_ = static_cast<uint32_t>(rs) << kShiftRegisterShift |
                static_cast<uint32_t>(shift) << kShiftShift | (1 << 4) |
                static_cast<uint32_t>(rm);
  }

  static bool CanHold(uint32_t immediate, Operand* o) {
    // Avoid the more expensive test for frequent small immediate values.
    if (immediate < (1 << kImmed8Bits)) {
      o->type_ = 1;
      o->encoding_ = (0 << kRotateShift) | (immediate << kImmed8Shift);
      return true;
    }
    // Note that immediate must be unsigned for the test to work correctly.
    for (int rot = 0; rot < 16; rot++) {
      uint32_t imm8 = (immediate << 2 * rot) | (immediate >> (32 - 2 * rot));
      if (imm8 < (1 << kImmed8Bits)) {
        o->type_ = 1;
        o->encoding_ = (rot << kRotateShift) | (imm8 << kImmed8Shift);
        return true;
      }
    }
    return false;
  }

 private:
  bool is_valid() const { return (type_ == 0) || (type_ == 1); }

  uint32_t type() const {
    ASSERT(is_valid());
    return type_;
  }

  uint32_t encoding() const {
    ASSERT(is_valid());
    return encoding_;
  }

  uint32_t type_;  // Encodes the type field (bits 27-25) in the instruction.
  uint32_t encoding_;

  friend class Assembler;
  friend class Address;
};

class Address : public ValueObject {
 public:
  enum OffsetKind {
    Immediate,
    IndexRegister,
    ScaledIndexRegister,
  };

  // Memory operand addressing mode
  enum Mode {
    kModeMask = (8 | 4 | 1) << 21,
    // bit encoding P U W
    Offset = (8 | 4 | 0) << 21,       // offset (w/o writeback to base)
    PreIndex = (8 | 4 | 1) << 21,     // pre-indexed addressing with writeback
    PostIndex = (0 | 4 | 0) << 21,    // post-indexed addressing with writeback
    NegOffset = (8 | 0 | 0) << 21,    // negative offset (w/o writeback to base)
    NegPreIndex = (8 | 0 | 1) << 21,  // negative pre-indexed with writeback
    NegPostIndex = (0 | 0 | 0) << 21  // negative post-indexed with writeback
  };

  Address(const Address& other)
      : ValueObject(), encoding_(other.encoding_), kind_(other.kind_) {}

  Address& operator=(const Address& other) {
    encoding_ = other.encoding_;
    kind_ = other.kind_;
    return *this;
  }

  bool Equals(const Address& other) const {
    return (encoding_ == other.encoding_) && (kind_ == other.kind_);
  }

  explicit Address(Register rn, int32_t offset = 0, Mode am = Offset) {
    ASSERT(Utils::IsAbsoluteUint(12, offset));
    kind_ = Immediate;
    if (offset < 0) {
      encoding_ = (am ^ (1 << kUShift)) | -offset;  // Flip U to adjust sign.
    } else {
      encoding_ = am | offset;
    }
    encoding_ |= ArmEncode::Rn(rn);
  }

  // There is no register offset mode unless Mode is Offset, in which case the
  // shifted register case below should be used.
  Address(Register rn, Register r, Mode am);

  Address(Register rn,
          Register rm,
          Shift shift = LSL,
          uint32_t shift_imm = 0,
          Mode am = Offset) {
    Operand o(rm, shift, shift_imm);

    if ((shift == LSL) && (shift_imm == 0)) {
      kind_ = IndexRegister;
    } else {
      kind_ = ScaledIndexRegister;
    }
    encoding_ = o.encoding() | am | ArmEncode::Rn(rn);
  }

  // There is no shifted register mode with a register shift.
  Address(Register rn, Register rm, Shift shift, Register r, Mode am = Offset);

  static OperandSize OperandSizeFor(intptr_t cid);

  static bool CanHoldLoadOffset(OperandSize size,
                                int32_t offset,
                                int32_t* offset_mask);
  static bool CanHoldStoreOffset(OperandSize size,
                                 int32_t offset,
                                 int32_t* offset_mask);
  static bool CanHoldImmediateOffset(bool is_load,
                                     intptr_t cid,
                                     int64_t offset);

 private:
  Register rn() const {
    return Instr::At(reinterpret_cast<uword>(&encoding_))->RnField();
  }

  Register rm() const {
    return ((kind() == IndexRegister) || (kind() == ScaledIndexRegister))
               ? Instr::At(reinterpret_cast<uword>(&encoding_))->RmField()
               : kNoRegister;
  }

  Mode mode() const { return static_cast<Mode>(encoding() & kModeMask); }

  bool has_writeback() const {
    return (mode() == PreIndex) || (mode() == PostIndex) ||
           (mode() == NegPreIndex) || (mode() == NegPostIndex);
  }

  uint32_t encoding() const { return encoding_; }

  // Encoding for addressing mode 3.
  uint32_t encoding3() const;

  // Encoding for vfp load/store addressing.
  uint32_t vencoding() const;

  OffsetKind kind() const { return kind_; }

  uint32_t encoding_;

  OffsetKind kind_;

  friend class Assembler;
};

class FieldAddress : public Address {
 public:
  FieldAddress(Register base, int32_t disp)
      : Address(base, disp - kHeapObjectTag) {}

  // This addressing mode does not exist.
  FieldAddress(Register base, Register r);

  FieldAddress(const FieldAddress& other) : Address(other) {}

  FieldAddress& operator=(const FieldAddress& other) {
    Address::operator=(other);
    return *this;
  }
};

class Assembler : public AssemblerBase {
 public:
  explicit Assembler(ObjectPoolBuilder* object_pool_builder,
                     bool use_far_branches = false);
  ~Assembler() {}

  void PushRegister(Register r) { Push(r); }
  void PopRegister(Register r) { Pop(r); }

  void Bind(Label* label);
  void Jump(Label* label) { b(label); }

  void LoadField(Register dst, FieldAddress address) { ldr(dst, address); }

  void CompareWithFieldValue(Register value, FieldAddress address) {
    ldr(TMP, address);
    cmp(value, Operand(TMP));
  }

  // Misc. functionality
  bool use_far_branches() const {
    return FLAG_use_far_branches || use_far_branches_;
  }

#if defined(TESTING) || defined(DEBUG)
  // Used in unit tests and to ensure predictable verification code size in
  // FlowGraphCompiler::EmitEdgeCounter.
  void set_use_far_branches(bool b) { use_far_branches_ = b; }
#endif  // TESTING || DEBUG

  // Debugging and bringup support.
  void Breakpoint() { bkpt(0); }
  void Stop(const char* message) override;

  static void InitializeMemoryWithBreakpoints(uword data, intptr_t length);

  static const char* RegisterName(Register reg);

  static const char* FpuRegisterName(FpuRegister reg);

  // Data-processing instructions.
  void and_(Register rd, Register rn, Operand o, Condition cond = AL);

  void eor(Register rd, Register rn, Operand o, Condition cond = AL);

  void sub(Register rd, Register rn, Operand o, Condition cond = AL);
  void subs(Register rd, Register rn, Operand o, Condition cond = AL);

  void rsb(Register rd, Register rn, Operand o, Condition cond = AL);
  void rsbs(Register rd, Register rn, Operand o, Condition cond = AL);

  void add(Register rd, Register rn, Operand o, Condition cond = AL);

  void adds(Register rd, Register rn, Operand o, Condition cond = AL);

  void adc(Register rd, Register rn, Operand o, Condition cond = AL);

  void adcs(Register rd, Register rn, Operand o, Condition cond = AL);

  void sbc(Register rd, Register rn, Operand o, Condition cond = AL);

  void sbcs(Register rd, Register rn, Operand o, Condition cond = AL);

  void rsc(Register rd, Register rn, Operand o, Condition cond = AL);

  void tst(Register rn, Operand o, Condition cond = AL);

  void teq(Register rn, Operand o, Condition cond = AL);

  void cmp(Register rn, Operand o, Condition cond = AL);

  void cmn(Register rn, Operand o, Condition cond = AL);

  void orr(Register rd, Register rn, Operand o, Condition cond = AL);
  void orrs(Register rd, Register rn, Operand o, Condition cond = AL);

  void mov(Register rd, Operand o, Condition cond = AL);
  void movs(Register rd, Operand o, Condition cond = AL);

  void bic(Register rd, Register rn, Operand o, Condition cond = AL);
  void bics(Register rd, Register rn, Operand o, Condition cond = AL);

  void mvn(Register rd, Operand o, Condition cond = AL);
  void mvns(Register rd, Operand o, Condition cond = AL);

  // Miscellaneous data-processing instructions.
  void clz(Register rd, Register rm, Condition cond = AL);

  // Multiply instructions.
  void mul(Register rd, Register rn, Register rm, Condition cond = AL);
  void muls(Register rd, Register rn, Register rm, Condition cond = AL);
  void mla(Register rd,
           Register rn,
           Register rm,
           Register ra,
           Condition cond = AL);
  void mls(Register rd,
           Register rn,
           Register rm,
           Register ra,
           Condition cond = AL);
  void smull(Register rd_lo,
             Register rd_hi,
             Register rn,
             Register rm,
             Condition cond = AL);
  void umull(Register rd_lo,
             Register rd_hi,
             Register rn,
             Register rm,
             Condition cond = AL);
  void smlal(Register rd_lo,
             Register rd_hi,
             Register rn,
             Register rm,
             Condition cond = AL);
  void umlal(Register rd_lo,
             Register rd_hi,
             Register rn,
             Register rm,
             Condition cond = AL);

  // Emulation of this instruction uses IP and the condition codes. Therefore,
  // none of the registers can be IP, and the instruction can only be used
  // unconditionally.
  void umaal(Register rd_lo, Register rd_hi, Register rn, Register rm);

  // Division instructions.
  void sdiv(Register rd, Register rn, Register rm, Condition cond = AL);
  void udiv(Register rd, Register rn, Register rm, Condition cond = AL);

  // Load/store instructions.
  void ldr(Register rd, Address ad, Condition cond = AL);
  void str(Register rd, Address ad, Condition cond = AL);

  void ldrb(Register rd, Address ad, Condition cond = AL);
  void strb(Register rd, Address ad, Condition cond = AL);

  void ldrh(Register rd, Address ad, Condition cond = AL);
  void strh(Register rd, Address ad, Condition cond = AL);

  void ldrsb(Register rd, Address ad, Condition cond = AL);
  void ldrsh(Register rd, Address ad, Condition cond = AL);

  // ldrd and strd actually support the full range of addressing modes, but
  // we don't use them, and we need to split them up into two instructions for
  // ARMv5TE, so we only support the base + offset mode.
  // rd must be an even register and rd2 must be rd + 1.
  void ldrd(Register rd,
            Register rd2,
            Register rn,
            int32_t offset,
            Condition cond = AL);
  void strd(Register rd,
            Register rd2,
            Register rn,
            int32_t offset,
            Condition cond = AL);

  void ldm(BlockAddressMode am,
           Register base,
           RegList regs,
           Condition cond = AL);
  void stm(BlockAddressMode am,
           Register base,
           RegList regs,
           Condition cond = AL);

  void ldrex(Register rd, Register rn, Condition cond = AL);
  void strex(Register rd, Register rt, Register rn, Condition cond = AL);

  // Miscellaneous instructions.
  void clrex();
  void nop(Condition cond = AL);

  // Note that gdb sets breakpoints using the undefined instruction 0xe7f001f0.
  void bkpt(uint16_t imm16);

  static int32_t BkptEncoding(uint16_t imm16) {
    // bkpt requires that the cond field is AL.
    return (AL << kConditionShift) | B24 | B21 | ((imm16 >> 4) << 8) | B6 | B5 |
           B4 | (imm16 & 0xf);
  }

  static uword GetBreakInstructionFiller() { return BkptEncoding(0); }

  // Floating point instructions (VFPv3-D16 and VFPv3-D32 profiles).
  void vmovsr(SRegister sn, Register rt, Condition cond = AL);
  void vmovrs(Register rt, SRegister sn, Condition cond = AL);
  void vmovsrr(SRegister sm, Register rt, Register rt2, Condition cond = AL);
  void vmovrrs(Register rt, Register rt2, SRegister sm, Condition cond = AL);
  void vmovdrr(DRegister dm, Register rt, Register rt2, Condition cond = AL);
  void vmovrrd(Register rt, Register rt2, DRegister dm, Condition cond = AL);
  void vmovdr(DRegister dd, int i, Register rt, Condition cond = AL);
  void vmovs(SRegister sd, SRegister sm, Condition cond = AL);
  void vmovd(DRegister dd, DRegister dm, Condition cond = AL);
  void vmovq(QRegister qd, QRegister qm);

  // Returns false if the immediate cannot be encoded.
  bool vmovs(SRegister sd, float s_imm, Condition cond = AL);
  bool vmovd(DRegister dd, double d_imm, Condition cond = AL);

  void vldrs(SRegister sd, Address ad, Condition cond = AL);
  void vstrs(SRegister sd, Address ad, Condition cond = AL);
  void vldrd(DRegister dd, Address ad, Condition cond = AL);
  void vstrd(DRegister dd, Address ad, Condition cond = AL);

  void vldms(BlockAddressMode am,
             Register base,
             SRegister first,
             SRegister last,
             Condition cond = AL);
  void vstms(BlockAddressMode am,
             Register base,
             SRegister first,
             SRegister last,
             Condition cond = AL);

  void vldmd(BlockAddressMode am,
             Register base,
             DRegister first,
             intptr_t count,
             Condition cond = AL);
  void vstmd(BlockAddressMode am,
             Register base,
             DRegister first,
             intptr_t count,
             Condition cond = AL);

  void vadds(SRegister sd, SRegister sn, SRegister sm, Condition cond = AL);
  void vaddd(DRegister dd, DRegister dn, DRegister dm, Condition cond = AL);
  void vaddqi(OperandSize sz, QRegister qd, QRegister qn, QRegister qm);
  void vaddqs(QRegister qd, QRegister qn, QRegister qm);
  void vsubs(SRegister sd, SRegister sn, SRegister sm, Condition cond = AL);
  void vsubd(DRegister dd, DRegister dn, DRegister dm, Condition cond = AL);
  void vsubqi(OperandSize sz, QRegister qd, QRegister qn, QRegister qm);
  void vsubqs(QRegister qd, QRegister qn, QRegister qm);
  void vmuls(SRegister sd, SRegister sn, SRegister sm, Condition cond = AL);
  void vmuld(DRegister dd, DRegister dn, DRegister dm, Condition cond = AL);
  void vmulqi(OperandSize sz, QRegister qd, QRegister qn, QRegister qm);
  void vmulqs(QRegister qd, QRegister qn, QRegister qm);
  void vshlqi(OperandSize sz, QRegister qd, QRegister qm, QRegister qn);
  void vshlqu(OperandSize sz, QRegister qd, QRegister qm, QRegister qn);
  void vmlas(SRegister sd, SRegister sn, SRegister sm, Condition cond = AL);
  void vmlad(DRegister dd, DRegister dn, DRegister dm, Condition cond = AL);
  void vmlss(SRegister sd, SRegister sn, SRegister sm, Condition cond = AL);
  void vmlsd(DRegister dd, DRegister dn, DRegister dm, Condition cond = AL);
  void vdivs(SRegister sd, SRegister sn, SRegister sm, Condition cond = AL);
  void vdivd(DRegister dd, DRegister dn, DRegister dm, Condition cond = AL);
  void vminqs(QRegister qd, QRegister qn, QRegister qm);
  void vmaxqs(QRegister qd, QRegister qn, QRegister qm);
  void vrecpeqs(QRegister qd, QRegister qm);
  void vrecpsqs(QRegister qd, QRegister qn, QRegister qm);
  void vrsqrteqs(QRegister qd, QRegister qm);
  void vrsqrtsqs(QRegister qd, QRegister qn, QRegister qm);

  void veorq(QRegister qd, QRegister qn, QRegister qm);
  void vorrq(QRegister qd, QRegister qn, QRegister qm);
  void vornq(QRegister qd, QRegister qn, QRegister qm);
  void vandq(QRegister qd, QRegister qn, QRegister qm);
  void vmvnq(QRegister qd, QRegister qm);

  void vceqqi(OperandSize sz, QRegister qd, QRegister qn, QRegister qm);
  void vceqqs(QRegister qd, QRegister qn, QRegister qm);
  void vcgeqi(OperandSize sz, QRegister qd, QRegister qn, QRegister qm);
  void vcugeqi(OperandSize sz, QRegister qd, QRegister qn, QRegister qm);
  void vcgeqs(QRegister qd, QRegister qn, QRegister qm);
  void vcgtqi(OperandSize sz, QRegister qd, QRegister qn, QRegister qm);
  void vcugtqi(OperandSize sz, QRegister qd, QRegister qn, QRegister qm);
  void vcgtqs(QRegister qd, QRegister qn, QRegister qm);

  void vabss(SRegister sd, SRegister sm, Condition cond = AL);
  void vabsd(DRegister dd, DRegister dm, Condition cond = AL);
  void vabsqs(QRegister qd, QRegister qm);
  void vnegs(SRegister sd, SRegister sm, Condition cond = AL);
  void vnegd(DRegister dd, DRegister dm, Condition cond = AL);
  void vnegqs(QRegister qd, QRegister qm);
  void vsqrts(SRegister sd, SRegister sm, Condition cond = AL);
  void vsqrtd(DRegister dd, DRegister dm, Condition cond = AL);

  void vcvtsd(SRegister sd, DRegister dm, Condition cond = AL);
  void vcvtds(DRegister dd, SRegister sm, Condition cond = AL);
  void vcvtis(SRegister sd, SRegister sm, Condition cond = AL);
  void vcvtid(SRegister sd, DRegister dm, Condition cond = AL);
  void vcvtsi(SRegister sd, SRegister sm, Condition cond = AL);
  void vcvtdi(DRegister dd, SRegister sm, Condition cond = AL);
  void vcvtus(SRegister sd, SRegister sm, Condition cond = AL);
  void vcvtud(SRegister sd, DRegister dm, Condition cond = AL);
  void vcvtsu(SRegister sd, SRegister sm, Condition cond = AL);
  void vcvtdu(DRegister dd, SRegister sm, Condition cond = AL);

  void vcmps(SRegister sd, SRegister sm, Condition cond = AL);
  void vcmpd(DRegister dd, DRegister dm, Condition cond = AL);
  void vcmpsz(SRegister sd, Condition cond = AL);
  void vcmpdz(DRegister dd, Condition cond = AL);
  void vmrs(Register rd, Condition cond = AL);
  void vmstat(Condition cond = AL);

  // Duplicates the operand of size sz at index idx from dm to all elements of
  // qd. This is a special case of vtbl.
  void vdup(OperandSize sz, QRegister qd, DRegister dm, int idx);

  // Each byte of dm is an index into the table of bytes formed by concatenating
  // a list of 'length' registers starting with dn. The result is placed in dd.
  void vtbl(DRegister dd, DRegister dn, int length, DRegister dm);

  // The words of qd and qm are interleaved with the low words of the result
  // in qd and the high words in qm.
  void vzipqw(QRegister qd, QRegister qm);

  // Branch instructions.
  void b(Label* label, Condition cond = AL);
  void bl(Label* label, Condition cond = AL);
  void bx(Register rm, Condition cond = AL);
  void blx(Register rm, Condition cond = AL);

  void Branch(const Code& code,
              ObjectPoolBuilderEntry::Patchability patchable =
                  ObjectPoolBuilderEntry::kNotPatchable,
              Register pp = PP,
              Condition cond = AL);

  void Branch(const Address& address, Condition cond = AL);

  void BranchLink(const Code& code,
                  ObjectPoolBuilderEntry::Patchability patchable =
                      ObjectPoolBuilderEntry::kNotPatchable,
                  CodeEntryKind entry_kind = CodeEntryKind::kNormal);
  void BranchLinkToRuntime();

  void CallNullErrorShared(bool save_fpu_registers);

  // Branch and link to an entry address. Call sequence can be patched.
  void BranchLinkPatchable(const Code& code,
                           CodeEntryKind entry_kind = CodeEntryKind::kNormal);

  // Emit a call that shares its object pool entries with other calls
  // that have the same equivalence marker.
  void BranchLinkWithEquivalence(
      const Code& code,
      const Object& equivalence,
      CodeEntryKind entry_kind = CodeEntryKind::kNormal);

  // Branch and link to [base + offset]. Call sequence is never patched.
  void BranchLinkOffset(Register base, int32_t offset);

  // Add signed immediate value to rd. May clobber IP.
  void AddImmediate(Register rd, int32_t value, Condition cond = AL) {
    AddImmediate(rd, rd, value, cond);
  }

  // Add signed immediate value. May clobber IP.
  void AddImmediate(Register rd,
                    Register rn,
                    int32_t value,
                    Condition cond = AL);
  void AddImmediateSetFlags(Register rd,
                            Register rn,
                            int32_t value,
                            Condition cond = AL);
  void SubImmediateSetFlags(Register rd,
                            Register rn,
                            int32_t value,
                            Condition cond = AL);
  void AndImmediate(Register rd, Register rs, int32_t imm, Condition cond = AL);

  // Test rn and immediate. May clobber IP.
  void TestImmediate(Register rn, int32_t imm, Condition cond = AL);

  // Compare rn with signed immediate value. May clobber IP.
  void CompareImmediate(Register rn, int32_t value, Condition cond = AL);

  // Signed integer division of left by right. Checks to see if integer
  // division is supported. If not, uses the FPU for division with
  // temporary registers tmpl and tmpr. tmpl and tmpr must be different
  // registers.
  void IntegerDivide(Register result,
                     Register left,
                     Register right,
                     DRegister tmpl,
                     DRegister tmpr);

  // Load and Store.
  // These three do not clobber IP.
  void LoadPatchableImmediate(Register rd, int32_t value, Condition cond = AL);
  void LoadDecodableImmediate(Register rd, int32_t value, Condition cond = AL);
  void LoadImmediate(Register rd, int32_t value, Condition cond = AL);
  // These two may clobber IP.
  void LoadSImmediate(SRegister sd, float value, Condition cond = AL);
  void LoadDImmediate(DRegister dd,
                      double value,
                      Register scratch,
                      Condition cond = AL);

  void MarkExceptionHandler(Label* label);

  void Drop(intptr_t stack_elements);

  void RestoreCodePointer();
  void LoadPoolPointer(Register reg = PP);

  void LoadIsolate(Register rd);

  // Load word from pool from the given offset using encoding that
  // InstructionPattern::DecodeLoadWordFromPool can decode.
  void LoadWordFromPoolOffset(Register rd,
                              int32_t offset,
                              Register pp,
                              Condition cond);

  void LoadObject(Register rd, const Object& object, Condition cond = AL);
  void LoadUniqueObject(Register rd, const Object& object, Condition cond = AL);
  void LoadFunctionFromCalleePool(Register dst,
                                  const Function& function,
                                  Register new_pp);
  void LoadNativeEntry(Register dst,
                       const ExternalLabel* label,
                       ObjectPoolBuilderEntry::Patchability patchable,
                       Condition cond = AL);
  void PushObject(const Object& object);
  void PushImmediate(int32_t immediate) {
    LoadImmediate(TMP, immediate);
    Push(TMP);
  }
  void CompareObject(Register rn, const Object& object);

  enum CanBeSmi {
    kValueIsNotSmi,
    kValueCanBeSmi,
  };

  // Store into a heap object and apply the generational and incremental write
  // barriers. All stores into heap objects must pass through this function or,
  // if the value can be proven either Smi or old-and-premarked, its NoBarrier
  // variants.
  // Preserves object and value registers.
  void StoreIntoObject(Register object,      // Object we are storing into.
                       const Address& dest,  // Where we are storing into.
                       Register value,       // Value we are storing.
                       CanBeSmi can_value_be_smi = kValueCanBeSmi,
                       bool lr_reserved = false);
  void StoreIntoArray(Register object,
                      Register slot,
                      Register value,
                      CanBeSmi can_value_be_smi = kValueCanBeSmi,
                      bool lr_reserved = false);
  void StoreIntoObjectOffset(Register object,
                             int32_t offset,
                             Register value,
                             CanBeSmi can_value_be_smi = kValueCanBeSmi,
                             bool lr_reserved = false);

  void StoreIntoObjectNoBarrier(Register object,
                                const Address& dest,
                                Register value);
  void StoreIntoObjectNoBarrier(Register object,
                                const Address& dest,
                                const Object& value);
  void StoreIntoObjectNoBarrierOffset(Register object,
                                      int32_t offset,
                                      Register value);
  void StoreIntoObjectNoBarrierOffset(Register object,
                                      int32_t offset,
                                      const Object& value);

  // Store value_even, value_odd, value_even, ... into the words in the address
  // range [begin, end), assumed to be uninitialized fields in object (tagged).
  // The stores must not need a generational store barrier (e.g., smi/null),
  // and (value_even, value_odd) must be a valid register pair.
  // Destroys register 'begin'.
  void InitializeFieldsNoBarrier(Register object,
                                 Register begin,
                                 Register end,
                                 Register value_even,
                                 Register value_odd);
  // Like above, for the range [base+begin_offset, base+end_offset), unrolled.
  void InitializeFieldsNoBarrierUnrolled(Register object,
                                         Register base,
                                         intptr_t begin_offset,
                                         intptr_t end_offset,
                                         Register value_even,
                                         Register value_odd);

  // Stores a Smi value into a heap object field that always contains a Smi.
  void StoreIntoSmiField(const Address& dest, Register value);

  void LoadClassId(Register result, Register object, Condition cond = AL);
  void LoadClassById(Register result, Register class_id);
  void CompareClassId(Register object, intptr_t class_id, Register scratch);
  void LoadClassIdMayBeSmi(Register result, Register object);
  void LoadTaggedClassIdMayBeSmi(Register result, Register object);

  intptr_t FindImmediate(int32_t imm);
  bool CanLoadFromObjectPool(const Object& object) const;
  void LoadFromOffset(OperandSize type,
                      Register reg,
                      Register base,
                      int32_t offset,
                      Condition cond = AL);
  void LoadFieldFromOffset(OperandSize type,
                           Register reg,
                           Register base,
                           int32_t offset,
                           Condition cond = AL) {
    LoadFromOffset(type, reg, base, offset - kHeapObjectTag, cond);
  }
  void StoreToOffset(OperandSize type,
                     Register reg,
                     Register base,
                     int32_t offset,
                     Condition cond = AL);
  void LoadSFromOffset(SRegister reg,
                       Register base,
                       int32_t offset,
                       Condition cond = AL);
  void StoreSToOffset(SRegister reg,
                      Register base,
                      int32_t offset,
                      Condition cond = AL);
  void LoadDFromOffset(DRegister reg,
                       Register base,
                       int32_t offset,
                       Condition cond = AL);
  void StoreDToOffset(DRegister reg,
                      Register base,
                      int32_t offset,
                      Condition cond = AL);

  void LoadMultipleDFromOffset(DRegister first,
                               intptr_t count,
                               Register base,
                               int32_t offset);
  void StoreMultipleDToOffset(DRegister first,
                              intptr_t count,
                              Register base,
                              int32_t offset);

  void CopyDoubleField(Register dst,
                       Register src,
                       Register tmp1,
                       Register tmp2,
                       DRegister dtmp);
  void CopyFloat32x4Field(Register dst,
                          Register src,
                          Register tmp1,
                          Register tmp2,
                          DRegister dtmp);
  void CopyFloat64x2Field(Register dst,
                          Register src,
                          Register tmp1,
                          Register tmp2,
                          DRegister dtmp);

  void Push(Register rd, Condition cond = AL);
  void Pop(Register rd, Condition cond = AL);

  void PushList(RegList regs, Condition cond = AL);
  void PopList(RegList regs, Condition cond = AL);

  void PushRegisters(const RegisterSet& regs);
  void PopRegisters(const RegisterSet& regs);

  void CompareRegisters(Register rn, Register rm) { cmp(rn, Operand(rm)); }
  void BranchIf(Condition condition, Label* label) { b(label, condition); }

  void MoveRegister(Register rd, Register rm, Condition cond = AL);

  // Convenience shift instructions. Use mov instruction with shifter operand
  // for variants setting the status flags.
  void Lsl(Register rd,
           Register rm,
           const Operand& shift_imm,
           Condition cond = AL);
  void Lsl(Register rd, Register rm, Register rs, Condition cond = AL);
  void Lsr(Register rd,
           Register rm,
           const Operand& shift_imm,
           Condition cond = AL);
  void Lsr(Register rd, Register rm, Register rs, Condition cond = AL);
  void Asr(Register rd,
           Register rm,
           const Operand& shift_imm,
           Condition cond = AL);
  void Asr(Register rd, Register rm, Register rs, Condition cond = AL);
  void Asrs(Register rd,
            Register rm,
            const Operand& shift_imm,
            Condition cond = AL);
  void Ror(Register rd,
           Register rm,
           const Operand& shift_imm,
           Condition cond = AL);
  void Ror(Register rd, Register rm, Register rs, Condition cond = AL);
  void Rrx(Register rd, Register rm, Condition cond = AL);

  // Fill rd with the sign of rm.
  void SignFill(Register rd, Register rm, Condition cond = AL);

  void Vreciprocalqs(QRegister qd, QRegister qm);
  void VreciprocalSqrtqs(QRegister qd, QRegister qm);
  // If qm must be preserved, then provide a (non-QTMP) temporary.
  void Vsqrtqs(QRegister qd, QRegister qm, QRegister temp);
  void Vdivqs(QRegister qd, QRegister qn, QRegister qm);

  void SmiTag(Register reg, Condition cond = AL) {
    Lsl(reg, reg, Operand(kSmiTagSize), cond);
  }

  void SmiTag(Register dst, Register src, Condition cond = AL) {
    Lsl(dst, src, Operand(kSmiTagSize), cond);
  }

  void SmiUntag(Register reg, Condition cond = AL) {
    Asr(reg, reg, Operand(kSmiTagSize), cond);
  }

  void SmiUntag(Register dst, Register src, Condition cond = AL) {
    Asr(dst, src, Operand(kSmiTagSize), cond);
  }

  // Untag the value in the register assuming it is a smi.
  // Untagging shifts tag bit into the carry flag - if carry is clear
  // assumption was correct. In this case jump to the is_smi label.
  // Otherwise fall-through.
  void SmiUntag(Register dst, Register src, Label* is_smi) {
    ASSERT(kSmiTagSize == 1);
    Asrs(dst, src, Operand(kSmiTagSize));
    b(is_smi, CC);
  }

  void BranchIfNotSmi(Register reg, Label* label) {
    tst(reg, Operand(kSmiTagMask));
    b(label, NE);
  }

  void BranchIfSmi(Register reg, Label* label) {
    tst(reg, Operand(kSmiTagMask));
    b(label, EQ);
  }

  void CheckCodePointer();

  // Function frame setup and tear down.
  void EnterFrame(RegList regs, intptr_t frame_space);
  void LeaveFrame(RegList regs, bool allow_pop_pc = false);
  void Ret();
  void ReserveAlignedFrameSpace(intptr_t frame_space);

  // Create a frame for calling into runtime that preserves all volatile
  // registers.  Frame's SP is guaranteed to be correctly aligned and
  // frame_space bytes are reserved under it.
  void EnterCallRuntimeFrame(intptr_t frame_space);
  void LeaveCallRuntimeFrame();

  void CallRuntime(const RuntimeEntry& entry, intptr_t argument_count);

  // Set up a Dart frame on entry with a frame pointer and PC information to
  // enable easy access to the RawInstruction object of code corresponding
  // to this frame.
  void EnterDartFrame(intptr_t frame_size);

  void LeaveDartFrame();

  // Leaves the frame and returns.
  //
  // The difference to "LeaveDartFrame(); Ret();" is that we return using
  //
  //   ldmia sp!, {fp, pc}
  //
  // instead of
  //
  //   ldmia sp!, {fp, lr}
  //   blx lr
  //
  // This means that our return must go to ARM mode (and not thumb).
  void LeaveDartFrameAndReturn();

  // Set up a Dart frame for a function compiled for on-stack replacement.
  // The frame layout is a normal Dart frame, but the frame is partially set
  // up on entry (it is the frame of the unoptimized code).
  void EnterOsrFrame(intptr_t extra_size);

  // Set up a stub frame so that the stack traversal code can easily identify
  // a stub frame.
  void EnterStubFrame();
  void LeaveStubFrame();

  void MonomorphicCheckedEntry();

  // The register into which the allocation stats table is loaded with
  // LoadAllocationStatsAddress should be passed to MaybeTraceAllocation and
  // IncrementAllocationStats(WithSize) as stats_addr_reg to update the
  // allocation stats. These are separate assembler macros so we can
  // avoid a dependent load too nearby the load of the table address.
  void LoadAllocationStatsAddress(Register dest, intptr_t cid);
  void IncrementAllocationStats(Register stats_addr, intptr_t cid);
  void IncrementAllocationStatsWithSize(Register stats_addr_reg,
                                        Register size_reg);

  Address ElementAddressForIntIndex(bool is_load,
                                    bool is_external,
                                    intptr_t cid,
                                    intptr_t index_scale,
                                    Register array,
                                    intptr_t index,
                                    Register temp);

  void LoadElementAddressForIntIndex(Register address,
                                     bool is_load,
                                     bool is_external,
                                     intptr_t cid,
                                     intptr_t index_scale,
                                     Register array,
                                     intptr_t index);

  Address ElementAddressForRegIndex(bool is_load,
                                    bool is_external,
                                    intptr_t cid,
                                    intptr_t index_scale,
                                    Register array,
                                    Register index);

  void LoadElementAddressForRegIndex(Register address,
                                     bool is_load,
                                     bool is_external,
                                     intptr_t cid,
                                     intptr_t index_scale,
                                     Register array,
                                     Register index);

  void LoadHalfWordUnaligned(Register dst, Register addr, Register tmp);
  void LoadHalfWordUnsignedUnaligned(Register dst, Register addr, Register tmp);
  void StoreHalfWordUnaligned(Register src, Register addr, Register tmp);
  void LoadWordUnaligned(Register dst, Register addr, Register tmp);
  void StoreWordUnaligned(Register src, Register addr, Register tmp);

  // If allocation tracing is enabled, will jump to |trace| label,
  // which will allocate in the runtime where tracing occurs.
  void MaybeTraceAllocation(Register stats_addr_reg, Label* trace);

  // Inlined allocation of an instance of class 'cls', code has no runtime
  // calls. Jump to 'failure' if the instance cannot be allocated here.
  // Allocated instance is returned in 'instance_reg'.
  // Only the tags field of the object is initialized.
  void TryAllocate(const Class& cls,
                   Label* failure,
                   Register instance_reg,
                   Register temp_reg);

  void TryAllocateArray(intptr_t cid,
                        intptr_t instance_size,
                        Label* failure,
                        Register instance,
                        Register end_address,
                        Register temp1,
                        Register temp2);

  // This emits an PC-relative call of the form "blr <offset>".  The offset
  // is not yet known and needs therefore relocation to the right place before
  // the code can be used.
  //
  // The neccessary information for the "linker" (i.e. the relocation
  // information) is stored in [RawCode::static_calls_target_table_]: an entry
  // of the form
  //
  //   (Code::kPcRelativeCall & pc_offset, <target-code>, <target-function>)
  //
  // will be used during relocation to fix the offset.
  //
  // The provided [offset_into_target] will be added to calculate the final
  // destination.  It can be used e.g. for calling into the middle of a
  // function.
  void GenerateUnRelocatedPcRelativeCall(Condition cond = AL,
                                         intptr_t offset_into_target = 0);

  // Emit data (e.g encoded instruction or immediate) in instruction stream.
  void Emit(int32_t value);

  // On some other platforms, we draw a distinction between safe and unsafe
  // smis.
  static bool IsSafe(const Object& object) { return true; }
  static bool IsSafeSmi(const Object& object) { return target::IsSmi(object); }

  bool constant_pool_allowed() const { return constant_pool_allowed_; }
  void set_constant_pool_allowed(bool b) { constant_pool_allowed_ = b; }

  // Whether we can branch to a target which is [distance] bytes away from the
  // beginning of the branch instruction.
  //
  // Use this function for testing whether [distance] can be encoded using the
  // 24-bit offets in the branch instructions, which are multiples of 4.
  static bool CanEncodeBranchDistance(int32_t distance) {
    ASSERT(Utils::IsAligned(distance, 4));
    // The distance is off by 8 due to the way the ARM CPUs read PC.
    distance -= Instr::kPCReadOffset;
    distance >>= 2;
    return Utils::IsInt(24, distance);
  }

  static int32_t EncodeBranchOffset(int32_t offset, int32_t inst);
  static int32_t DecodeBranchOffset(int32_t inst);

 private:
  bool use_far_branches_;

  // If you are thinking of using one or both of these instructions directly,
  // instead LoadImmediate should probably be used.
  void movw(Register rd, uint16_t imm16, Condition cond = AL);
  void movt(Register rd, uint16_t imm16, Condition cond = AL);

  void BindARMv6(Label* label);
  void BindARMv7(Label* label);

  void BranchLink(const ExternalLabel* label);

  bool constant_pool_allowed_;

  void LoadObjectHelper(Register rd,
                        const Object& object,
                        Condition cond,
                        bool is_unique,
                        Register pp);

  void EmitType01(Condition cond,
                  int type,
                  Opcode opcode,
                  int set_cc,
                  Register rn,
                  Register rd,
                  Operand o);

  void EmitType5(Condition cond, int32_t offset, bool link);

  void EmitMemOp(Condition cond, bool load, bool byte, Register rd, Address ad);

  void EmitMemOpAddressMode3(Condition cond,
                             int32_t mode,
                             Register rd,
                             Address ad);

  void EmitMultiMemOp(Condition cond,
                      BlockAddressMode am,
                      bool load,
                      Register base,
                      RegList regs);

  void EmitShiftImmediate(Condition cond,
                          Shift opcode,
                          Register rd,
                          Register rm,
                          Operand o);

  void EmitShiftRegister(Condition cond,
                         Shift opcode,
                         Register rd,
                         Register rm,
                         Operand o);

  void EmitMulOp(Condition cond,
                 int32_t opcode,
                 Register rd,
                 Register rn,
                 Register rm,
                 Register rs);

  void EmitDivOp(Condition cond,
                 int32_t opcode,
                 Register rd,
                 Register rn,
                 Register rm);

  void EmitMultiVSMemOp(Condition cond,
                        BlockAddressMode am,
                        bool load,
                        Register base,
                        SRegister start,
                        uint32_t count);

  void EmitMultiVDMemOp(Condition cond,
                        BlockAddressMode am,
                        bool load,
                        Register base,
                        DRegister start,
                        int32_t count);

  void EmitVFPsss(Condition cond,
                  int32_t opcode,
                  SRegister sd,
                  SRegister sn,
                  SRegister sm);

  void EmitVFPddd(Condition cond,
                  int32_t opcode,
                  DRegister dd,
                  DRegister dn,
                  DRegister dm);

  void EmitVFPsd(Condition cond, int32_t opcode, SRegister sd, DRegister dm);

  void EmitVFPds(Condition cond, int32_t opcode, DRegister dd, SRegister sm);

  void EmitSIMDqqq(int32_t opcode,
                   OperandSize sz,
                   QRegister qd,
                   QRegister qn,
                   QRegister qm);

  void EmitSIMDddd(int32_t opcode,
                   OperandSize sz,
                   DRegister dd,
                   DRegister dn,
                   DRegister dm);

  void EmitFarBranch(Condition cond, int32_t offset, bool link);
  void EmitBranch(Condition cond, Label* label, bool link);
  void BailoutIfInvalidBranchOffset(int32_t offset);
  int32_t EncodeTstOffset(int32_t offset, int32_t inst);
  int32_t DecodeTstOffset(int32_t inst);

  enum BarrierFilterMode {
    // Filter falls through into the barrier update code. Target label
    // is a "after-store" label.
    kJumpToNoUpdate,

    // Filter falls through to the "after-store" code. Target label
    // is barrier update code label.
    kJumpToBarrier,

    // Filter falls through into the conditional barrier update code and does
    // not jump. Target label is unused. The barrier should run if the NE
    // condition is set.
    kNoJump
  };

  void StoreIntoObjectFilter(Register object,
                             Register value,
                             Label* label,
                             CanBeSmi can_be_smi,
                             BarrierFilterMode barrier_filter_mode);

  friend class dart::FlowGraphCompiler;
  std::function<void(Condition, Register)>
      generate_invoke_write_barrier_wrapper_;
  std::function<void(Condition)> generate_invoke_array_write_barrier_;

  DISALLOW_ALLOCATION();
  DISALLOW_COPY_AND_ASSIGN(Assembler);
};

}  // namespace compiler

// TODO(vegorov) temporary export commonly used classes into dart namespace
// to ease migration.
using compiler::Address;
using compiler::FieldAddress;
using compiler::Operand;

}  // namespace dart

#endif  // RUNTIME_VM_COMPILER_ASSEMBLER_ASSEMBLER_ARM_H_