// Copyright (c) 2019, 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.

#include "vm/globals.h"

#define SHOULD_NOT_INCLUDE_RUNTIME

#include "vm/compiler/stub_code_compiler.h"

#if defined(TARGET_ARCH_ARM) && !defined(DART_PRECOMPILED_RUNTIME)

#include "vm/class_id.h"
#include "vm/code_entry_kind.h"
#include "vm/compiler/assembler/assembler.h"
#include "vm/compiler/backend/locations.h"
#include "vm/constants_arm.h"
#include "vm/instructions.h"
#include "vm/static_type_exactness_state.h"
#include "vm/tags.h"

#define __ assembler->

namespace dart {

DEFINE_FLAG(bool, inline_alloc, true, "Inline allocation of objects.");
DEFINE_FLAG(bool,
            use_slow_path,
            false,
            "Set to true for debugging & verifying the slow paths.");
DECLARE_FLAG(bool, precompiled_mode);

namespace compiler {

// Input parameters:
//   LR : return address.
//   SP : address of last argument in argument array.
//   SP + 4*R4 - 4 : address of first argument in argument array.
//   SP + 4*R4 : address of return value.
//   R9 : address of the runtime function to call.
//   R4 : number of arguments to the call.
void StubCodeCompiler::GenerateCallToRuntimeStub(Assembler* assembler) {
  const intptr_t thread_offset = target::NativeArguments::thread_offset();
  const intptr_t argc_tag_offset = target::NativeArguments::argc_tag_offset();
  const intptr_t argv_offset = target::NativeArguments::argv_offset();
  const intptr_t retval_offset = target::NativeArguments::retval_offset();

  __ ldr(CODE_REG, Address(THR, target::Thread::call_to_runtime_stub_offset()));
  __ EnterStubFrame();

  // Save exit frame information to enable stack walking as we are about
  // to transition to Dart VM C++ code.
  __ StoreToOffset(kWord, FP, THR,
                   target::Thread::top_exit_frame_info_offset());

#if defined(DEBUG)
  {
    Label ok;
    // Check that we are always entering from Dart code.
    __ LoadFromOffset(kWord, R8, THR, target::Thread::vm_tag_offset());
    __ CompareImmediate(R8, VMTag::kDartCompiledTagId);
    __ b(&ok, EQ);
    __ Stop("Not coming from Dart code.");
    __ Bind(&ok);
  }
#endif

  // Mark that the thread is executing VM code.
  __ StoreToOffset(kWord, R9, THR, target::Thread::vm_tag_offset());

  // Reserve space for arguments and align frame before entering C++ world.
  // target::NativeArguments are passed in registers.
  ASSERT(target::NativeArguments::StructSize() == 4 * target::kWordSize);
  __ ReserveAlignedFrameSpace(0);

  // Pass target::NativeArguments structure by value and call runtime.
  // Registers R0, R1, R2, and R3 are used.

  ASSERT(thread_offset == 0 * target::kWordSize);
  // Set thread in NativeArgs.
  __ mov(R0, Operand(THR));

  // There are no runtime calls to closures, so we do not need to set the tag
  // bits kClosureFunctionBit and kInstanceFunctionBit in argc_tag_.
  ASSERT(argc_tag_offset == 1 * target::kWordSize);
  __ mov(R1, Operand(R4));  // Set argc in target::NativeArguments.

  ASSERT(argv_offset == 2 * target::kWordSize);
  __ add(R2, FP, Operand(R4, LSL, 2));  // Compute argv.
  // Set argv in target::NativeArguments.
  __ AddImmediate(R2,
                  target::frame_layout.param_end_from_fp * target::kWordSize);

  ASSERT(retval_offset == 3 * target::kWordSize);
  __ add(R3, R2,
         Operand(target::kWordSize));  // Retval is next to 1st argument.

  // Call runtime or redirection via simulator.
  __ blx(R9);

  // Mark that the thread is executing Dart code.
  __ LoadImmediate(R2, VMTag::kDartCompiledTagId);
  __ StoreToOffset(kWord, R2, THR, target::Thread::vm_tag_offset());

  // Reset exit frame information in Isolate structure.
  __ LoadImmediate(R2, 0);
  __ StoreToOffset(kWord, R2, THR,
                   target::Thread::top_exit_frame_info_offset());

  // Restore the global object pool after returning from runtime (old space is
  // moving, so the GOP could have been relocated).
  if (FLAG_precompiled_mode && FLAG_use_bare_instructions) {
    __ ldr(PP, Address(THR, target::Thread::global_object_pool_offset()));
  }

  __ LeaveStubFrame();

  // The following return can jump to a lazy-deopt stub, which assumes R0
  // contains a return value and will save it in a GC-visible way.  We therefore
  // have to ensure R0 does not contain any garbage value left from the C
  // function we called (which has return type "void").
  // (See GenerateDeoptimizationSequence::saved_result_slot_from_fp.)
  __ LoadImmediate(R0, 0);
  __ Ret();
}

void StubCodeCompiler::GenerateSharedStub(
    Assembler* assembler,
    bool save_fpu_registers,
    const RuntimeEntry* target,
    intptr_t self_code_stub_offset_from_thread,
    bool allow_return) {
  // If the target CPU does not support VFP the caller should always use the
  // non-FPU stub.
  if (save_fpu_registers && !TargetCPUFeatures::vfp_supported()) {
    __ Breakpoint();
    return;
  }

  // We want the saved registers to appear like part of the caller's frame, so
  // we push them before calling EnterStubFrame.
  RegisterSet all_registers;
  all_registers.AddAllNonReservedRegisters(save_fpu_registers);

  // To make the stack map calculation architecture independent we do the same
  // as on intel.
  __ Push(LR);

  __ PushRegisters(all_registers);
  __ ldr(CODE_REG, Address(THR, self_code_stub_offset_from_thread));
  __ EnterStubFrame();
  __ CallRuntime(*target, /*argument_count=*/0);
  if (!allow_return) {
    __ Breakpoint();
    return;
  }
  __ LeaveStubFrame();
  __ PopRegisters(all_registers);
  __ Pop(LR);
  __ bx(LR);
}

// R1: The extracted method.
// R4: The type_arguments_field_offset (or 0)
// SP+0: The object from which we are tearing a method off.
void StubCodeCompiler::GenerateBuildMethodExtractorStub(
    Assembler* assembler,
    const Object& closure_allocation_stub,
    const Object& context_allocation_stub) {
  const intptr_t kReceiverOffset =
      compiler::target::frame_layout.param_end_from_fp + 1;

  __ EnterStubFrame();

  // Build type_arguments vector (or null)
  __ cmp(R4, Operand(0));
  __ ldr(R3, Address(THR, target::Thread::object_null_offset()), EQ);
  __ ldr(R0, Address(FP, kReceiverOffset * target::kWordSize), NE);
  __ ldr(R3, Address(R0, R4), NE);

  // Push type arguments & extracted method.
  __ PushList(1 << R3 | 1 << R1);

  // Allocate context.
  {
    Label done, slow_path;
    __ TryAllocateArray(kContextCid, target::Context::InstanceSize(1),
                        &slow_path,
                        R0,  // instance
                        R1,  // end address
                        R2, R3);
    __ ldr(R1, Address(THR, target::Thread::object_null_offset()));
    __ str(R1, FieldAddress(R0, target::Context::parent_offset()));
    __ LoadImmediate(R1, 1);
    __ str(R1, FieldAddress(R0, target::Context::num_variables_offset()));
    __ b(&done);

    __ Bind(&slow_path);

    __ LoadImmediate(/*num_vars=*/R1, 1);
    __ LoadObject(CODE_REG, context_allocation_stub);
    __ ldr(R0, FieldAddress(CODE_REG, target::Code::entry_point_offset()));
    __ blx(R0);

    __ Bind(&done);
  }

  // Store receiver in context
  __ ldr(R1, Address(FP, target::kWordSize * kReceiverOffset));
  __ StoreIntoObject(R0, FieldAddress(R0, target::Context::variable_offset(0)),
                     R1);

  // Push context.
  __ Push(R0);

  // Allocate closure.
  __ LoadObject(CODE_REG, closure_allocation_stub);
  __ ldr(R1, FieldAddress(CODE_REG, target::Code::entry_point_offset(
                                        CodeEntryKind::kUnchecked)));
  __ blx(R1);

  // Populate closure object.
  __ Pop(R1);  // Pop context.
  __ StoreIntoObject(R0, FieldAddress(R0, target::Closure::context_offset()),
                     R1);
  __ PopList(1 << R3 | 1 << R1);  // Pop type arguments & extracted method.
  __ StoreIntoObjectNoBarrier(
      R0, FieldAddress(R0, target::Closure::function_offset()), R1);
  __ StoreIntoObjectNoBarrier(
      R0,
      FieldAddress(R0, target::Closure::instantiator_type_arguments_offset()),
      R3);
  __ LoadObject(R1, EmptyTypeArguments());
  __ StoreIntoObjectNoBarrier(
      R0, FieldAddress(R0, target::Closure::delayed_type_arguments_offset()),
      R1);

  __ LeaveStubFrame();
  __ Ret();
}

void StubCodeCompiler::GenerateNullErrorSharedWithoutFPURegsStub(
    Assembler* assembler) {
  GenerateSharedStub(
      assembler, /*save_fpu_registers=*/false, &kNullErrorRuntimeEntry,
      target::Thread::null_error_shared_without_fpu_regs_stub_offset(),
      /*allow_return=*/false);
}

void StubCodeCompiler::GenerateNullErrorSharedWithFPURegsStub(
    Assembler* assembler) {
  GenerateSharedStub(
      assembler, /*save_fpu_registers=*/true, &kNullErrorRuntimeEntry,
      target::Thread::null_error_shared_with_fpu_regs_stub_offset(),
      /*allow_return=*/false);
}

void StubCodeCompiler::GenerateStackOverflowSharedWithoutFPURegsStub(
    Assembler* assembler) {
  GenerateSharedStub(
      assembler, /*save_fpu_registers=*/false, &kStackOverflowRuntimeEntry,
      target::Thread::stack_overflow_shared_without_fpu_regs_stub_offset(),
      /*allow_return=*/true);
}

void StubCodeCompiler::GenerateStackOverflowSharedWithFPURegsStub(
    Assembler* assembler) {
  GenerateSharedStub(
      assembler, /*save_fpu_registers=*/true, &kStackOverflowRuntimeEntry,
      target::Thread::stack_overflow_shared_with_fpu_regs_stub_offset(),
      /*allow_return=*/true);
}

// Input parameters:
//   R0 : stop message (const char*).
// Must preserve all registers.
void StubCodeCompiler::GeneratePrintStopMessageStub(Assembler* assembler) {
  __ EnterCallRuntimeFrame(0);
  // Call the runtime leaf function. R0 already contains the parameter.
  __ CallRuntime(kPrintStopMessageRuntimeEntry, 1);
  __ LeaveCallRuntimeFrame();
  __ Ret();
}

// Input parameters:
//   LR : return address.
//   SP : address of return value.
//   R9 : address of the native function to call.
//   R2 : address of first argument in argument array.
//   R1 : argc_tag including number of arguments and function kind.
static void GenerateCallNativeWithWrapperStub(Assembler* assembler,
                                              Address wrapper) {
  const intptr_t thread_offset = target::NativeArguments::thread_offset();
  const intptr_t argc_tag_offset = target::NativeArguments::argc_tag_offset();
  const intptr_t argv_offset = target::NativeArguments::argv_offset();
  const intptr_t retval_offset = target::NativeArguments::retval_offset();

  __ EnterStubFrame();

  // Save exit frame information to enable stack walking as we are about
  // to transition to native code.
  __ StoreToOffset(kWord, FP, THR,
                   target::Thread::top_exit_frame_info_offset());

#if defined(DEBUG)
  {
    Label ok;
    // Check that we are always entering from Dart code.
    __ LoadFromOffset(kWord, R8, THR, target::Thread::vm_tag_offset());
    __ CompareImmediate(R8, VMTag::kDartCompiledTagId);
    __ b(&ok, EQ);
    __ Stop("Not coming from Dart code.");
    __ Bind(&ok);
  }
#endif

  // Mark that the thread is executing native code.
  __ StoreToOffset(kWord, R9, THR, target::Thread::vm_tag_offset());

  // Reserve space for the native arguments structure passed on the stack (the
  // outgoing pointer parameter to the native arguments structure is passed in
  // R0) and align frame before entering the C++ world.
  __ ReserveAlignedFrameSpace(target::NativeArguments::StructSize());

  // Initialize target::NativeArguments structure and call native function.
  // Registers R0, R1, R2, and R3 are used.

  ASSERT(thread_offset == 0 * target::kWordSize);
  // Set thread in NativeArgs.
  __ mov(R0, Operand(THR));

  // There are no native calls to closures, so we do not need to set the tag
  // bits kClosureFunctionBit and kInstanceFunctionBit in argc_tag_.
  ASSERT(argc_tag_offset == 1 * target::kWordSize);
  // Set argc in target::NativeArguments: R1 already contains argc.

  ASSERT(argv_offset == 2 * target::kWordSize);
  // Set argv in target::NativeArguments: R2 already contains argv.

  // Set retval in NativeArgs.
  ASSERT(retval_offset == 3 * target::kWordSize);
  __ add(R3, FP, Operand(2 * target::kWordSize));

  // Passing the structure by value as in runtime calls would require changing
  // Dart API for native functions.
  // For now, space is reserved on the stack and we pass a pointer to it.
  __ stm(IA, SP, (1 << R0) | (1 << R1) | (1 << R2) | (1 << R3));
  __ mov(R0, Operand(SP));  // Pass the pointer to the target::NativeArguments.

  __ mov(R1, Operand(R9));  // Pass the function entrypoint to call.

  // Call native function invocation wrapper or redirection via simulator.
  __ ldr(LR, wrapper);
  __ blx(LR);

  // Mark that the thread is executing Dart code.
  __ LoadImmediate(R2, VMTag::kDartCompiledTagId);
  __ StoreToOffset(kWord, R2, THR, target::Thread::vm_tag_offset());

  // Reset exit frame information in Isolate structure.
  __ LoadImmediate(R2, 0);
  __ StoreToOffset(kWord, R2, THR,
                   target::Thread::top_exit_frame_info_offset());

  __ LeaveStubFrame();
  __ Ret();
}

void StubCodeCompiler::GenerateCallNoScopeNativeStub(Assembler* assembler) {
  GenerateCallNativeWithWrapperStub(
      assembler,
      Address(THR,
              target::Thread::no_scope_native_wrapper_entry_point_offset()));
}

void StubCodeCompiler::GenerateCallAutoScopeNativeStub(Assembler* assembler) {
  GenerateCallNativeWithWrapperStub(
      assembler,
      Address(THR,
              target::Thread::auto_scope_native_wrapper_entry_point_offset()));
}

// Input parameters:
//   LR : return address.
//   SP : address of return value.
//   R9 : address of the native function to call.
//   R2 : address of first argument in argument array.
//   R1 : argc_tag including number of arguments and function kind.
void StubCodeCompiler::GenerateCallBootstrapNativeStub(Assembler* assembler) {
  const intptr_t thread_offset = target::NativeArguments::thread_offset();
  const intptr_t argc_tag_offset = target::NativeArguments::argc_tag_offset();
  const intptr_t argv_offset = target::NativeArguments::argv_offset();
  const intptr_t retval_offset = target::NativeArguments::retval_offset();

  __ EnterStubFrame();

  // Save exit frame information to enable stack walking as we are about
  // to transition to native code.
  __ StoreToOffset(kWord, FP, THR,
                   target::Thread::top_exit_frame_info_offset());

#if defined(DEBUG)
  {
    Label ok;
    // Check that we are always entering from Dart code.
    __ LoadFromOffset(kWord, R8, THR, target::Thread::vm_tag_offset());
    __ CompareImmediate(R8, VMTag::kDartCompiledTagId);
    __ b(&ok, EQ);
    __ Stop("Not coming from Dart code.");
    __ Bind(&ok);
  }
#endif

  // Mark that the thread is executing native code.
  __ StoreToOffset(kWord, R9, THR, target::Thread::vm_tag_offset());

  // Reserve space for the native arguments structure passed on the stack (the
  // outgoing pointer parameter to the native arguments structure is passed in
  // R0) and align frame before entering the C++ world.
  __ ReserveAlignedFrameSpace(target::NativeArguments::StructSize());

  // Initialize target::NativeArguments structure and call native function.
  // Registers R0, R1, R2, and R3 are used.

  ASSERT(thread_offset == 0 * target::kWordSize);
  // Set thread in NativeArgs.
  __ mov(R0, Operand(THR));

  // There are no native calls to closures, so we do not need to set the tag
  // bits kClosureFunctionBit and kInstanceFunctionBit in argc_tag_.
  ASSERT(argc_tag_offset == 1 * target::kWordSize);
  // Set argc in target::NativeArguments: R1 already contains argc.

  ASSERT(argv_offset == 2 * target::kWordSize);
  // Set argv in target::NativeArguments: R2 already contains argv.

  // Set retval in NativeArgs.
  ASSERT(retval_offset == 3 * target::kWordSize);
  __ add(R3, FP, Operand(2 * target::kWordSize));

  // Passing the structure by value as in runtime calls would require changing
  // Dart API for native functions.
  // For now, space is reserved on the stack and we pass a pointer to it.
  __ stm(IA, SP, (1 << R0) | (1 << R1) | (1 << R2) | (1 << R3));
  __ mov(R0, Operand(SP));  // Pass the pointer to the target::NativeArguments.

  // Call native function or redirection via simulator.
  __ blx(R9);

  // Mark that the thread is executing Dart code.
  __ LoadImmediate(R2, VMTag::kDartCompiledTagId);
  __ StoreToOffset(kWord, R2, THR, target::Thread::vm_tag_offset());

  // Reset exit frame information in Isolate structure.
  __ LoadImmediate(R2, 0);
  __ StoreToOffset(kWord, R2, THR,
                   target::Thread::top_exit_frame_info_offset());

  __ LeaveStubFrame();
  __ Ret();
}

// Input parameters:
//   R4: arguments descriptor array.
void StubCodeCompiler::GenerateCallStaticFunctionStub(Assembler* assembler) {
  // Create a stub frame as we are pushing some objects on the stack before
  // calling into the runtime.
  __ EnterStubFrame();
  // Setup space on stack for return value and preserve arguments descriptor.
  __ LoadImmediate(R0, 0);
  __ PushList((1 << R0) | (1 << R4));
  __ CallRuntime(kPatchStaticCallRuntimeEntry, 0);
  // Get Code object result and restore arguments descriptor array.
  __ PopList((1 << R0) | (1 << R4));
  // Remove the stub frame.
  __ LeaveStubFrame();
  // Jump to the dart function.
  __ mov(CODE_REG, Operand(R0));
  __ Branch(FieldAddress(R0, target::Code::entry_point_offset()));
}

// Called from a static call only when an invalid code has been entered
// (invalid because its function was optimized or deoptimized).
// R4: arguments descriptor array.
void StubCodeCompiler::GenerateFixCallersTargetStub(Assembler* assembler) {
  // Load code pointer to this stub from the thread:
  // The one that is passed in, is not correct - it points to the code object
  // that needs to be replaced.
  __ ldr(CODE_REG,
         Address(THR, target::Thread::fix_callers_target_code_offset()));
  // Create a stub frame as we are pushing some objects on the stack before
  // calling into the runtime.
  __ EnterStubFrame();
  // Setup space on stack for return value and preserve arguments descriptor.
  __ LoadImmediate(R0, 0);
  __ PushList((1 << R0) | (1 << R4));
  __ CallRuntime(kFixCallersTargetRuntimeEntry, 0);
  // Get Code object result and restore arguments descriptor array.
  __ PopList((1 << R0) | (1 << R4));
  // Remove the stub frame.
  __ LeaveStubFrame();
  // Jump to the dart function.
  __ mov(CODE_REG, Operand(R0));
  __ Branch(FieldAddress(R0, target::Code::entry_point_offset()));
}

// Called from object allocate instruction when the allocation stub has been
// disabled.
void StubCodeCompiler::GenerateFixAllocationStubTargetStub(
    Assembler* assembler) {
  // Load code pointer to this stub from the thread:
  // The one that is passed in, is not correct - it points to the code object
  // that needs to be replaced.
  __ ldr(CODE_REG,
         Address(THR, target::Thread::fix_allocation_stub_code_offset()));
  __ EnterStubFrame();
  // Setup space on stack for return value.
  __ LoadImmediate(R0, 0);
  __ Push(R0);
  __ CallRuntime(kFixAllocationStubTargetRuntimeEntry, 0);
  // Get Code object result.
  __ Pop(R0);
  // Remove the stub frame.
  __ LeaveStubFrame();
  // Jump to the dart function.
  __ mov(CODE_REG, Operand(R0));
  __ Branch(FieldAddress(R0, target::Code::entry_point_offset()));
}

// Input parameters:
//   R2: smi-tagged argument count, may be zero.
//   FP[target::frame_layout.param_end_from_fp + 1]: last argument.
static void PushArrayOfArguments(Assembler* assembler) {
  // Allocate array to store arguments of caller.
  __ LoadObject(R1, NullObject());
  // R1: null element type for raw Array.
  // R2: smi-tagged argument count, may be zero.
  __ BranchLink(StubCodeAllocateArray());
  // R0: newly allocated array.
  // R2: smi-tagged argument count, may be zero (was preserved by the stub).
  __ Push(R0);  // Array is in R0 and on top of stack.
  __ AddImmediate(R1, FP,
                  target::frame_layout.param_end_from_fp * target::kWordSize);
  __ AddImmediate(R3, R0, target::Array::data_offset() - kHeapObjectTag);
  // Copy arguments from stack to array (starting at the end).
  // R1: address just beyond last argument on stack.
  // R3: address of first argument in array.
  Label enter;
  __ b(&enter);
  Label loop;
  __ Bind(&loop);
  __ ldr(R8, Address(R1, target::kWordSize, Address::PreIndex));
  // Generational barrier is needed, array is not necessarily in new space.
  __ StoreIntoObject(R0, Address(R3, R2, LSL, 1), R8);
  __ Bind(&enter);
  __ subs(R2, R2, Operand(target::ToRawSmi(1)));  // R2 is Smi.
  __ b(&loop, PL);
}

// Used by eager and lazy deoptimization. Preserve result in R0 if necessary.
// This stub translates optimized frame into unoptimized frame. The optimized
// frame can contain values in registers and on stack, the unoptimized
// frame contains all values on stack.
// Deoptimization occurs in following steps:
// - Push all registers that can contain values.
// - Call C routine to copy the stack and saved registers into temporary buffer.
// - Adjust caller's frame to correct unoptimized frame size.
// - Fill the unoptimized frame.
// - Materialize objects that require allocation (e.g. Double instances).
// GC can occur only after frame is fully rewritten.
// Stack after EnterFrame(...) below:
//   +------------------+
//   | Saved PP         | <- TOS
//   +------------------+
//   | Saved FP         | <- FP of stub
//   +------------------+
//   | Saved LR         |  (deoptimization point)
//   +------------------+
//   | pc marker        |
//   +------------------+
//   | Saved CODE_REG   |
//   +------------------+
//   | ...              | <- SP of optimized frame
//
// Parts of the code cannot GC, part of the code can GC.
static void GenerateDeoptimizationSequence(Assembler* assembler,
                                           DeoptStubKind kind) {
  // DeoptimizeCopyFrame expects a Dart frame, i.e. EnterDartFrame(0), but there
  // is no need to set the correct PC marker or load PP, since they get patched.
  __ EnterDartFrame(0);
  __ LoadPoolPointer();

  // The code in this frame may not cause GC. kDeoptimizeCopyFrameRuntimeEntry
  // and kDeoptimizeFillFrameRuntimeEntry are leaf runtime calls.
  const intptr_t saved_result_slot_from_fp =
      compiler::target::frame_layout.first_local_from_fp + 1 -
      (kNumberOfCpuRegisters - R0);
  const intptr_t saved_exception_slot_from_fp =
      compiler::target::frame_layout.first_local_from_fp + 1 -
      (kNumberOfCpuRegisters - R0);
  const intptr_t saved_stacktrace_slot_from_fp =
      compiler::target::frame_layout.first_local_from_fp + 1 -
      (kNumberOfCpuRegisters - R1);
  // Result in R0 is preserved as part of pushing all registers below.

  // Push registers in their enumeration order: lowest register number at
  // lowest address.
  for (intptr_t i = kNumberOfCpuRegisters - 1; i >= 0; --i) {
    if (i == CODE_REG) {
      // Save the original value of CODE_REG pushed before invoking this stub
      // instead of the value used to call this stub.
      __ ldr(IP, Address(FP, 2 * target::kWordSize));
      __ Push(IP);
    } else if (i == SP) {
      // Push(SP) has unpredictable behavior.
      __ mov(IP, Operand(SP));
      __ Push(IP);
    } else {
      __ Push(static_cast<Register>(i));
    }
  }

  if (TargetCPUFeatures::vfp_supported()) {
    ASSERT(kFpuRegisterSize == 4 * target::kWordSize);
    if (kNumberOfDRegisters > 16) {
      __ vstmd(DB_W, SP, D16, kNumberOfDRegisters - 16);
      __ vstmd(DB_W, SP, D0, 16);
    } else {
      __ vstmd(DB_W, SP, D0, kNumberOfDRegisters);
    }
  } else {
    __ AddImmediate(SP, -kNumberOfFpuRegisters * kFpuRegisterSize);
  }

  __ mov(R0, Operand(SP));  // Pass address of saved registers block.
  bool is_lazy =
      (kind == kLazyDeoptFromReturn) || (kind == kLazyDeoptFromThrow);
  __ mov(R1, Operand(is_lazy ? 1 : 0));
  __ ReserveAlignedFrameSpace(0);
  __ CallRuntime(kDeoptimizeCopyFrameRuntimeEntry, 2);
  // Result (R0) is stack-size (FP - SP) in bytes.

  if (kind == kLazyDeoptFromReturn) {
    // Restore result into R1 temporarily.
    __ ldr(R1, Address(FP, saved_result_slot_from_fp * target::kWordSize));
  } else if (kind == kLazyDeoptFromThrow) {
    // Restore result into R1 temporarily.
    __ ldr(R1, Address(FP, saved_exception_slot_from_fp * target::kWordSize));
    __ ldr(R2, Address(FP, saved_stacktrace_slot_from_fp * target::kWordSize));
  }

  __ RestoreCodePointer();
  __ LeaveDartFrame();
  __ sub(SP, FP, Operand(R0));

  // DeoptimizeFillFrame expects a Dart frame, i.e. EnterDartFrame(0), but there
  // is no need to set the correct PC marker or load PP, since they get patched.
  __ EnterStubFrame();
  __ mov(R0, Operand(FP));  // Get last FP address.
  if (kind == kLazyDeoptFromReturn) {
    __ Push(R1);  // Preserve result as first local.
  } else if (kind == kLazyDeoptFromThrow) {
    __ Push(R1);  // Preserve exception as first local.
    __ Push(R2);  // Preserve stacktrace as second local.
  }
  __ ReserveAlignedFrameSpace(0);
  __ CallRuntime(kDeoptimizeFillFrameRuntimeEntry, 1);  // Pass last FP in R0.
  if (kind == kLazyDeoptFromReturn) {
    // Restore result into R1.
    __ ldr(R1, Address(FP, compiler::target::frame_layout.first_local_from_fp *
                               target::kWordSize));
  } else if (kind == kLazyDeoptFromThrow) {
    // Restore result into R1.
    __ ldr(R1, Address(FP, compiler::target::frame_layout.first_local_from_fp *
                               target::kWordSize));
    __ ldr(R2, Address(FP, (compiler::target::frame_layout.first_local_from_fp -
                            1) *
                               target::kWordSize));
  }
  // Code above cannot cause GC.
  __ RestoreCodePointer();
  __ LeaveStubFrame();

  // Frame is fully rewritten at this point and it is safe to perform a GC.
  // Materialize any objects that were deferred by FillFrame because they
  // require allocation.
  // Enter stub frame with loading PP. The caller's PP is not materialized yet.
  __ EnterStubFrame();
  if (kind == kLazyDeoptFromReturn) {
    __ Push(R1);  // Preserve result, it will be GC-d here.
  } else if (kind == kLazyDeoptFromThrow) {
    __ Push(R1);  // Preserve exception, it will be GC-d here.
    __ Push(R2);  // Preserve stacktrace, it will be GC-d here.
  }
  __ PushObject(NullObject());  // Space for the result.
  __ CallRuntime(kDeoptimizeMaterializeRuntimeEntry, 0);
  // Result tells stub how many bytes to remove from the expression stack
  // of the bottom-most frame. They were used as materialization arguments.
  __ Pop(R2);
  if (kind == kLazyDeoptFromReturn) {
    __ Pop(R0);  // Restore result.
  } else if (kind == kLazyDeoptFromThrow) {
    __ Pop(R1);  // Restore stacktrace.
    __ Pop(R0);  // Restore exception.
  }
  __ LeaveStubFrame();
  // Remove materialization arguments.
  __ add(SP, SP, Operand(R2, ASR, kSmiTagSize));
  // The caller is responsible for emitting the return instruction.
}

// R0: result, must be preserved
void StubCodeCompiler::GenerateDeoptimizeLazyFromReturnStub(
    Assembler* assembler) {
  // Push zap value instead of CODE_REG for lazy deopt.
  __ LoadImmediate(IP, kZapCodeReg);
  __ Push(IP);
  // Return address for "call" to deopt stub.
  __ LoadImmediate(LR, kZapReturnAddress);
  __ ldr(CODE_REG,
         Address(THR, target::Thread::lazy_deopt_from_return_stub_offset()));
  GenerateDeoptimizationSequence(assembler, kLazyDeoptFromReturn);
  __ Ret();
}

// R0: exception, must be preserved
// R1: stacktrace, must be preserved
void StubCodeCompiler::GenerateDeoptimizeLazyFromThrowStub(
    Assembler* assembler) {
  // Push zap value instead of CODE_REG for lazy deopt.
  __ LoadImmediate(IP, kZapCodeReg);
  __ Push(IP);
  // Return address for "call" to deopt stub.
  __ LoadImmediate(LR, kZapReturnAddress);
  __ ldr(CODE_REG,
         Address(THR, target::Thread::lazy_deopt_from_throw_stub_offset()));
  GenerateDeoptimizationSequence(assembler, kLazyDeoptFromThrow);
  __ Ret();
}

void StubCodeCompiler::GenerateDeoptimizeStub(Assembler* assembler) {
  __ Push(CODE_REG);
  __ ldr(CODE_REG, Address(THR, target::Thread::deoptimize_stub_offset()));
  GenerateDeoptimizationSequence(assembler, kEagerDeopt);
  __ Ret();
}

static void GenerateDispatcherCode(Assembler* assembler,
                                   Label* call_target_function) {
  __ Comment("NoSuchMethodDispatch");
  // When lazily generated invocation dispatchers are disabled, the
  // miss-handler may return null.
  __ CompareObject(R0, NullObject());
  __ b(call_target_function, NE);
  __ EnterStubFrame();
  // Load the receiver.
  __ ldr(R2, FieldAddress(R4, target::ArgumentsDescriptor::count_offset()));
  __ add(IP, FP, Operand(R2, LSL, 1));  // R2 is Smi.
  __ ldr(R8, Address(IP, target::frame_layout.param_end_from_fp *
                             target::kWordSize));
  __ LoadImmediate(IP, 0);
  __ Push(IP);  // Result slot.
  __ Push(R8);  // Receiver.
  __ Push(R9);  // ICData/MegamorphicCache.
  __ Push(R4);  // Arguments descriptor.

  // Adjust arguments count.
  __ ldr(R3,
         FieldAddress(R4, target::ArgumentsDescriptor::type_args_len_offset()));
  __ cmp(R3, Operand(0));
  __ AddImmediate(R2, R2, target::ToRawSmi(1),
                  NE);  // Include the type arguments.

  // R2: Smi-tagged arguments array length.
  PushArrayOfArguments(assembler);
  const intptr_t kNumArgs = 4;
  __ CallRuntime(kInvokeNoSuchMethodDispatcherRuntimeEntry, kNumArgs);
  __ Drop(4);
  __ Pop(R0);  // Return value.
  __ LeaveStubFrame();
  __ Ret();
}

void StubCodeCompiler::GenerateMegamorphicMissStub(Assembler* assembler) {
  __ EnterStubFrame();

  // Load the receiver.
  __ ldr(R2, FieldAddress(R4, target::ArgumentsDescriptor::count_offset()));
  __ add(IP, FP, Operand(R2, LSL, 1));  // R2 is Smi.
  __ ldr(R8, Address(IP, compiler::target::frame_layout.param_end_from_fp *
                             target::kWordSize));

  // Preserve IC data and arguments descriptor.
  __ PushList((1 << R4) | (1 << R9));

  __ LoadImmediate(IP, 0);
  __ Push(IP);  // result slot
  __ Push(R8);  // receiver
  __ Push(R9);  // ICData
  __ Push(R4);  // arguments descriptor
  __ CallRuntime(kMegamorphicCacheMissHandlerRuntimeEntry, 3);
  // Remove arguments.
  __ Drop(3);
  __ Pop(R0);  // Get result into R0 (target function).

  // Restore IC data and arguments descriptor.
  __ PopList((1 << R4) | (1 << R9));

  __ RestoreCodePointer();
  __ LeaveStubFrame();

  if (!FLAG_lazy_dispatchers) {
    Label call_target_function;
    GenerateDispatcherCode(assembler, &call_target_function);
    __ Bind(&call_target_function);
  }

  // Tail-call to target function.
  __ ldr(CODE_REG, FieldAddress(R0, target::Function::code_offset()));
  __ Branch(FieldAddress(R0, target::Function::entry_point_offset()));
}

// Called for inline allocation of arrays.
// Input parameters:
//   LR: return address.
//   R1: array element type (either NULL or an instantiated type).
//   R2: array length as Smi (must be preserved).
// The newly allocated object is returned in R0.
void StubCodeCompiler::GenerateAllocateArrayStub(Assembler* assembler) {
  Label slow_case;
  // Compute the size to be allocated, it is based on the array length
  // and is computed as:
  // RoundedAllocationSize(
  //     (array_length * kwordSize) + target::Array::header_size()).
  __ mov(R3, Operand(R2));  // Array length.
  // Check that length is a positive Smi.
  __ tst(R3, Operand(kSmiTagMask));
  if (FLAG_use_slow_path) {
    __ b(&slow_case);
  } else {
    __ b(&slow_case, NE);
  }
  __ cmp(R3, Operand(0));
  __ b(&slow_case, LT);

  // Check for maximum allowed length.
  const intptr_t max_len =
      target::ToRawSmi(target::Array::kMaxNewSpaceElements);
  __ CompareImmediate(R3, max_len);
  __ b(&slow_case, GT);

  const intptr_t cid = kArrayCid;
  NOT_IN_PRODUCT(__ LoadAllocationStatsAddress(R4, cid));
  NOT_IN_PRODUCT(__ MaybeTraceAllocation(R4, &slow_case));

  const intptr_t fixed_size_plus_alignment_padding =
      target::Array::header_size() + target::ObjectAlignment::kObjectAlignment -
      1;
  __ LoadImmediate(R9, fixed_size_plus_alignment_padding);
  __ add(R9, R9, Operand(R3, LSL, 1));  // R3 is a Smi.
  ASSERT(kSmiTagShift == 1);
  __ bic(R9, R9, Operand(target::ObjectAlignment::kObjectAlignment - 1));

  // R9: Allocation size.
  // Potential new object start.
  __ ldr(R0, Address(THR, target::Thread::top_offset()));
  __ adds(NOTFP, R0, Operand(R9));  // Potential next object start.
  __ b(&slow_case, CS);             // Branch if unsigned overflow.

  // Check if the allocation fits into the remaining space.
  // R0: potential new object start.
  // NOTFP: potential next object start.
  // R9: allocation size.
  __ ldr(R3, Address(THR, target::Thread::end_offset()));
  __ cmp(NOTFP, Operand(R3));
  __ b(&slow_case, CS);

  // Successfully allocated the object(s), now update top to point to
  // next object start and initialize the object.
  NOT_IN_PRODUCT(__ LoadAllocationStatsAddress(R3, cid));
  __ str(NOTFP, Address(THR, target::Thread::top_offset()));
  __ add(R0, R0, Operand(kHeapObjectTag));

  // Initialize the tags.
  // R0: new object start as a tagged pointer.
  // R3: allocation stats address.
  // NOTFP: new object end address.
  // R9: allocation size.
  {
    const intptr_t shift = target::RawObject::kTagBitsSizeTagPos -
                           target::ObjectAlignment::kObjectAlignmentLog2;

    __ CompareImmediate(R9, target::RawObject::kSizeTagMaxSizeTag);
    __ mov(R8, Operand(R9, LSL, shift), LS);
    __ mov(R8, Operand(0), HI);

    // Get the class index and insert it into the tags.
    // R8: size and bit tags.
    const uint32_t tags =
        target::MakeTagWordForNewSpaceObject(cid, /*instance_size=*/0);
    __ LoadImmediate(TMP, tags);
    __ orr(R8, R8, Operand(TMP));
    __ str(R8, FieldAddress(R0, target::Array::tags_offset()));  // Store tags.
  }

  // R0: new object start as a tagged pointer.
  // NOTFP: new object end address.
  // Store the type argument field.
  __ StoreIntoObjectNoBarrier(
      R0, FieldAddress(R0, target::Array::type_arguments_offset()), R1);

  // Set the length field.
  __ StoreIntoObjectNoBarrier(
      R0, FieldAddress(R0, target::Array::length_offset()), R2);

  // Initialize all array elements to raw_null.
  // R0: new object start as a tagged pointer.
  // R3: allocation stats address.
  // R8, R9: null
  // R4: iterator which initially points to the start of the variable
  // data area to be initialized.
  // NOTFP: new object end address.
  // R9: allocation size.
  NOT_IN_PRODUCT(__ IncrementAllocationStatsWithSize(R3, R9));

  __ LoadObject(R8, NullObject());
  __ mov(R9, Operand(R8));
  __ AddImmediate(R4, R0, target::Array::header_size() - kHeapObjectTag);
  __ InitializeFieldsNoBarrier(R0, R4, NOTFP, R8, R9);
  __ Ret();  // Returns the newly allocated object in R0.
  // Unable to allocate the array using the fast inline code, just call
  // into the runtime.
  __ Bind(&slow_case);

  // Create a stub frame as we are pushing some objects on the stack before
  // calling into the runtime.
  __ EnterStubFrame();
  __ LoadImmediate(IP, 0);
  // Setup space on stack for return value.
  // Push array length as Smi and element type.
  __ PushList((1 << R1) | (1 << R2) | (1 << IP));
  __ CallRuntime(kAllocateArrayRuntimeEntry, 2);
  // Pop arguments; result is popped in IP.
  __ PopList((1 << R1) | (1 << R2) | (1 << IP));  // R2 is restored.
  __ mov(R0, Operand(IP));
  __ LeaveStubFrame();
  __ Ret();
}

// Called when invoking Dart code from C++ (VM code).
// Input parameters:
//   LR : points to return address.
//   R0 : code object of the Dart function to call.
//   R1 : arguments descriptor array.
//   R2 : arguments array.
//   R3 : current thread.
void StubCodeCompiler::GenerateInvokeDartCodeStub(Assembler* assembler) {
  __ Push(LR);  // Marker for the profiler.
  __ EnterFrame((1 << FP) | (1 << LR), 0);

  // Push code object to PC marker slot.
  __ ldr(IP, Address(R3, target::Thread::invoke_dart_code_stub_offset()));
  __ Push(IP);

  // Save new context and C++ ABI callee-saved registers.
  __ PushList(kAbiPreservedCpuRegs);

  const DRegister firstd = EvenDRegisterOf(kAbiFirstPreservedFpuReg);
  if (TargetCPUFeatures::vfp_supported()) {
    ASSERT(2 * kAbiPreservedFpuRegCount < 16);
    // Save FPU registers. 2 D registers per Q register.
    __ vstmd(DB_W, SP, firstd, 2 * kAbiPreservedFpuRegCount);
  } else {
    __ sub(SP, SP, Operand(kAbiPreservedFpuRegCount * kFpuRegisterSize));
  }

  // Set up THR, which caches the current thread in Dart code.
  if (THR != R3) {
    __ mov(THR, Operand(R3));
  }

  // Save the current VMTag on the stack.
  __ LoadFromOffset(kWord, R9, THR, target::Thread::vm_tag_offset());
  __ Push(R9);

  // Save top resource and top exit frame info. Use R4-6 as temporary registers.
  // StackFrameIterator reads the top exit frame info saved in this frame.
  __ LoadFromOffset(kWord, R9, THR,
                    target::Thread::top_exit_frame_info_offset());
  __ LoadFromOffset(kWord, R4, THR, target::Thread::top_resource_offset());
  __ LoadImmediate(R8, 0);
  __ StoreToOffset(kWord, R8, THR, target::Thread::top_resource_offset());
  __ StoreToOffset(kWord, R8, THR,
                   target::Thread::top_exit_frame_info_offset());

  // target::frame_layout.exit_link_slot_from_entry_fp must be kept in sync
  // with the code below.
  __ Push(R4);
#if defined(TARGET_OS_MACOS) || defined(TARGET_OS_MACOS_IOS)
  ASSERT(target::frame_layout.exit_link_slot_from_entry_fp == -26);
#else
  ASSERT(target::frame_layout.exit_link_slot_from_entry_fp == -27);
#endif
  __ Push(R9);

  // Mark that the thread is executing Dart code. Do this after initializing the
  // exit link for the profiler.
  __ LoadImmediate(R9, VMTag::kDartCompiledTagId);
  __ StoreToOffset(kWord, R9, THR, target::Thread::vm_tag_offset());

  // Load arguments descriptor array into R4, which is passed to Dart code.
  __ ldr(R4, Address(R1, VMHandles::kOffsetOfRawPtrInHandle));

  // Load number of arguments into R9 and adjust count for type arguments.
  __ ldr(R3,
         FieldAddress(R4, target::ArgumentsDescriptor::type_args_len_offset()));
  __ ldr(R9, FieldAddress(R4, target::ArgumentsDescriptor::count_offset()));
  __ cmp(R3, Operand(0));
  __ AddImmediate(R9, R9, target::ToRawSmi(1),
                  NE);  // Include the type arguments.
  __ SmiUntag(R9);

  // Compute address of 'arguments array' data area into R2.
  __ ldr(R2, Address(R2, VMHandles::kOffsetOfRawPtrInHandle));
  __ AddImmediate(R2, target::Array::data_offset() - kHeapObjectTag);

  // Set up arguments for the Dart call.
  Label push_arguments;
  Label done_push_arguments;
  __ CompareImmediate(R9, 0);  // check if there are arguments.
  __ b(&done_push_arguments, EQ);
  __ LoadImmediate(R1, 0);
  __ Bind(&push_arguments);
  __ ldr(R3, Address(R2));
  __ Push(R3);
  __ AddImmediate(R2, target::kWordSize);
  __ AddImmediate(R1, 1);
  __ cmp(R1, Operand(R9));
  __ b(&push_arguments, LT);
  __ Bind(&done_push_arguments);

  // Call the Dart code entrypoint.
  if (FLAG_precompiled_mode && FLAG_use_bare_instructions) {
    __ ldr(PP, Address(THR, target::Thread::global_object_pool_offset()));
  } else {
    __ LoadImmediate(PP, 0);  // GC safe value into PP.
  }
  __ ldr(CODE_REG, Address(R0, VMHandles::kOffsetOfRawPtrInHandle));
  __ ldr(R0, FieldAddress(CODE_REG, target::Code::entry_point_offset()));
  __ blx(R0);  // R4 is the arguments descriptor array.

  // Get rid of arguments pushed on the stack.
  __ AddImmediate(
      SP, FP,
      target::frame_layout.exit_link_slot_from_entry_fp * target::kWordSize);

  // Restore the saved top exit frame info and top resource back into the
  // Isolate structure. Uses R9 as a temporary register for this.
  __ Pop(R9);
  __ StoreToOffset(kWord, R9, THR,
                   target::Thread::top_exit_frame_info_offset());
  __ Pop(R9);
  __ StoreToOffset(kWord, R9, THR, target::Thread::top_resource_offset());

  // Restore the current VMTag from the stack.
  __ Pop(R4);
  __ StoreToOffset(kWord, R4, THR, target::Thread::vm_tag_offset());

  // Restore C++ ABI callee-saved registers.
  if (TargetCPUFeatures::vfp_supported()) {
    // Restore FPU registers. 2 D registers per Q register.
    __ vldmd(IA_W, SP, firstd, 2 * kAbiPreservedFpuRegCount);
  } else {
    __ AddImmediate(SP, kAbiPreservedFpuRegCount * kFpuRegisterSize);
  }
  // Restore CPU registers.
  __ PopList(kAbiPreservedCpuRegs);
  __ set_constant_pool_allowed(false);

  // Restore the frame pointer and return.
  __ LeaveFrame((1 << FP) | (1 << LR));
  __ Drop(1);
  __ Ret();
}

// Called when invoking compiled Dart code from interpreted Dart code.
// Input parameters:
//   LR : points to return address.
//   R0 : raw code object of the Dart function to call.
//   R1 : arguments raw descriptor array.
//   R2 : address of first argument.
//   R3 : current thread.
void StubCodeCompiler::GenerateInvokeDartCodeFromBytecodeStub(
    Assembler* assembler) {
#if defined(DART_PRECOMPILED_RUNTIME)
  __ Stop("Not using interpreter");
#else
  __ Push(LR);  // Marker for the profiler.
  __ EnterFrame((1 << FP) | (1 << LR), 0);

  // Push code object to PC marker slot.
  __ ldr(IP,
         Address(R3,
                 target::Thread::invoke_dart_code_from_bytecode_stub_offset()));
  __ Push(IP);

  // Save new context and C++ ABI callee-saved registers.
  __ PushList(kAbiPreservedCpuRegs);

  const DRegister firstd = EvenDRegisterOf(kAbiFirstPreservedFpuReg);
  if (TargetCPUFeatures::vfp_supported()) {
    ASSERT(2 * kAbiPreservedFpuRegCount < 16);
    // Save FPU registers. 2 D registers per Q register.
    __ vstmd(DB_W, SP, firstd, 2 * kAbiPreservedFpuRegCount);
  } else {
    __ sub(SP, SP, Operand(kAbiPreservedFpuRegCount * kFpuRegisterSize));
  }

  // Set up THR, which caches the current thread in Dart code.
  if (THR != R3) {
    __ mov(THR, Operand(R3));
  }

  // Save the current VMTag on the stack.
  __ LoadFromOffset(kWord, R9, THR, target::Thread::vm_tag_offset());
  __ Push(R9);

  // Save top resource and top exit frame info. Use R4-6 as temporary registers.
  // StackFrameIterator reads the top exit frame info saved in this frame.
  __ LoadFromOffset(kWord, R9, THR,
                    target::Thread::top_exit_frame_info_offset());
  __ LoadFromOffset(kWord, R4, THR, target::Thread::top_resource_offset());
  __ LoadImmediate(R8, 0);
  __ StoreToOffset(kWord, R8, THR, target::Thread::top_resource_offset());
  __ StoreToOffset(kWord, R8, THR,
                   target::Thread::top_exit_frame_info_offset());

  // target::frame_layout.exit_link_slot_from_entry_fp must be kept in sync
  // with the code below.
  __ Push(R4);
#if defined(TARGET_OS_MACOS) || defined(TARGET_OS_MACOS_IOS)
  ASSERT(target::frame_layout.exit_link_slot_from_entry_fp == -26);
#else
  ASSERT(target::frame_layout.exit_link_slot_from_entry_fp == -27);
#endif
  __ Push(R9);

  // Mark that the thread is executing Dart code. Do this after initializing the
  // exit link for the profiler.
  __ LoadImmediate(R9, VMTag::kDartCompiledTagId);
  __ StoreToOffset(kWord, R9, THR, target::Thread::vm_tag_offset());

  // Load arguments descriptor array into R4, which is passed to Dart code.
  __ mov(R4, Operand(R1));

  // Load number of arguments into R9 and adjust count for type arguments.
  __ ldr(R3,
         FieldAddress(R4, target::ArgumentsDescriptor::type_args_len_offset()));
  __ ldr(R9, FieldAddress(R4, target::ArgumentsDescriptor::count_offset()));
  __ cmp(R3, Operand(0));
  __ AddImmediate(R9, R9, target::ToRawSmi(1),
                  NE);  // Include the type arguments.
  __ SmiUntag(R9);

  // R2 points to first argument.
  // Set up arguments for the Dart call.
  Label push_arguments;
  Label done_push_arguments;
  __ CompareImmediate(R9, 0);  // check if there are arguments.
  __ b(&done_push_arguments, EQ);
  __ LoadImmediate(R1, 0);
  __ Bind(&push_arguments);
  __ ldr(R3, Address(R2));
  __ Push(R3);
  __ AddImmediate(R2, target::kWordSize);
  __ AddImmediate(R1, 1);
  __ cmp(R1, Operand(R9));
  __ b(&push_arguments, LT);
  __ Bind(&done_push_arguments);

  // Call the Dart code entrypoint.
  __ LoadImmediate(PP, 0);  // GC safe value into PP.
  __ mov(CODE_REG, Operand(R0));
  __ ldr(R0, FieldAddress(CODE_REG, target::Code::entry_point_offset()));
  __ blx(R0);  // R4 is the arguments descriptor array.

  // Get rid of arguments pushed on the stack.
  __ AddImmediate(
      SP, FP,
      target::frame_layout.exit_link_slot_from_entry_fp * target::kWordSize);

  // Restore the saved top exit frame info and top resource back into the
  // Isolate structure. Uses R9 as a temporary register for this.
  __ Pop(R9);
  __ StoreToOffset(kWord, R9, THR,
                   target::Thread::top_exit_frame_info_offset());
  __ Pop(R9);
  __ StoreToOffset(kWord, R9, THR, target::Thread::top_resource_offset());

  // Restore the current VMTag from the stack.
  __ Pop(R4);
  __ StoreToOffset(kWord, R4, THR, target::Thread::vm_tag_offset());

  // Restore C++ ABI callee-saved registers.
  if (TargetCPUFeatures::vfp_supported()) {
    // Restore FPU registers. 2 D registers per Q register.
    __ vldmd(IA_W, SP, firstd, 2 * kAbiPreservedFpuRegCount);
  } else {
    __ AddImmediate(SP, kAbiPreservedFpuRegCount * kFpuRegisterSize);
  }
  // Restore CPU registers.
  __ PopList(kAbiPreservedCpuRegs);
  __ set_constant_pool_allowed(false);

  // Restore the frame pointer and return.
  __ LeaveFrame((1 << FP) | (1 << LR));
  __ Drop(1);
  __ Ret();
#endif  // defined(DART_PRECOMPILED_RUNTIME)
}

// Called for inline allocation of contexts.
// Input:
//   R1: number of context variables.
// Output:
//   R0: new allocated RawContext object.
void StubCodeCompiler::GenerateAllocateContextStub(Assembler* assembler) {
  if (FLAG_inline_alloc) {
    Label slow_case;
    // First compute the rounded instance size.
    // R1: number of context variables.
    intptr_t fixed_size_plus_alignment_padding =
        target::Context::header_size() +
        target::ObjectAlignment::kObjectAlignment - 1;
    __ LoadImmediate(R2, fixed_size_plus_alignment_padding);
    __ add(R2, R2, Operand(R1, LSL, 2));
    ASSERT(kSmiTagShift == 1);
    __ bic(R2, R2, Operand(target::ObjectAlignment::kObjectAlignment - 1));

    NOT_IN_PRODUCT(__ LoadAllocationStatsAddress(R8, kContextCid));
    NOT_IN_PRODUCT(__ MaybeTraceAllocation(R8, &slow_case));
    // Now allocate the object.
    // R1: number of context variables.
    // R2: object size.
    const intptr_t cid = kContextCid;
    __ ldr(R0, Address(THR, target::Thread::top_offset()));
    __ add(R3, R2, Operand(R0));
    // Check if the allocation fits into the remaining space.
    // R0: potential new object.
    // R1: number of context variables.
    // R2: object size.
    // R3: potential next object start.
    __ ldr(IP, Address(THR, target::Thread::end_offset()));
    __ cmp(R3, Operand(IP));
    if (FLAG_use_slow_path) {
      __ b(&slow_case);
    } else {
      __ b(&slow_case, CS);  // Branch if unsigned higher or equal.
    }

    // Successfully allocated the object, now update top to point to
    // next object start and initialize the object.
    // R0: new object start (untagged).
    // R1: number of context variables.
    // R2: object size.
    // R3: next object start.
    NOT_IN_PRODUCT(__ LoadAllocationStatsAddress(R4, cid));
    __ str(R3, Address(THR, target::Thread::top_offset()));
    __ add(R0, R0, Operand(kHeapObjectTag));

    // Calculate the size tag.
    // R0: new object (tagged).
    // R1: number of context variables.
    // R2: object size.
    // R3: next object start.
    // R4: allocation stats address.
    const intptr_t shift = target::RawObject::kTagBitsSizeTagPos -
                           target::ObjectAlignment::kObjectAlignmentLog2;
    __ CompareImmediate(R2, target::RawObject::kSizeTagMaxSizeTag);
    // If no size tag overflow, shift R2 left, else set R2 to zero.
    __ mov(R9, Operand(R2, LSL, shift), LS);
    __ mov(R9, Operand(0), HI);

    // Get the class index and insert it into the tags.
    // R9: size and bit tags.
    const uint32_t tags =
        target::MakeTagWordForNewSpaceObject(cid, /*instance_size=*/0);

    __ LoadImmediate(IP, tags);
    __ orr(R9, R9, Operand(IP));
    __ str(R9, FieldAddress(R0, target::Object::tags_offset()));

    // Setup up number of context variables field.
    // R0: new object.
    // R1: number of context variables as integer value (not object).
    // R2: object size.
    // R3: next object start.
    // R4: allocation stats address.
    __ str(R1, FieldAddress(R0, target::Context::num_variables_offset()));

    // Setup the parent field.
    // R0: new object.
    // R1: number of context variables.
    // R2: object size.
    // R3: next object start.
    // R4: allocation stats address.
    __ LoadObject(R8, NullObject());
    __ StoreIntoObjectNoBarrier(
        R0, FieldAddress(R0, target::Context::parent_offset()), R8);

    // Initialize the context variables.
    // R0: new object.
    // R1: number of context variables.
    // R2: object size.
    // R3: next object start.
    // R8, R9: raw null.
    // R4: allocation stats address.
    Label loop;
    __ AddImmediate(NOTFP, R0,
                    target::Context::variable_offset(0) - kHeapObjectTag);
    __ InitializeFieldsNoBarrier(R0, NOTFP, R3, R8, R9);
    NOT_IN_PRODUCT(__ IncrementAllocationStatsWithSize(R4, R2));

    // Done allocating and initializing the context.
    // R0: new object.
    __ Ret();

    __ Bind(&slow_case);
  }
  // Create a stub frame as we are pushing some objects on the stack before
  // calling into the runtime.
  __ EnterStubFrame();
  // Setup space on stack for return value.
  __ LoadImmediate(R2, 0);
  __ SmiTag(R1);
  __ PushList((1 << R1) | (1 << R2));
  __ CallRuntime(kAllocateContextRuntimeEntry, 1);  // Allocate context.
  __ Drop(1);  // Pop number of context variables argument.
  __ Pop(R0);  // Pop the new context object.
  // R0: new object
  // Restore the frame pointer.
  __ LeaveStubFrame();
  __ Ret();
}

void StubCodeCompiler::GenerateWriteBarrierWrappersStub(Assembler* assembler) {
  RegList saved = (1 << LR) | (1 << kWriteBarrierObjectReg);
  for (intptr_t i = 0; i < kNumberOfCpuRegisters; ++i) {
    if ((kDartAvailableCpuRegs & (1 << i)) == 0) continue;

    Register reg = static_cast<Register>(i);
    intptr_t start = __ CodeSize();
    __ PushList(saved);
    __ mov(kWriteBarrierObjectReg, Operand(reg));
    __ ldr(LR,
           Address(THR, target::Thread::write_barrier_entry_point_offset()));
    __ blx(LR);
    __ PopList(saved);
    __ bx(LR);
    intptr_t end = __ CodeSize();

    RELEASE_ASSERT(end - start == kStoreBufferWrapperSize);
  }
}

// Helper stub to implement Assembler::StoreIntoObject.
// Input parameters:
//   R1: Object (old)
//   R0: Value (old or new)
//   R9: Slot
// If R0 is new, add R1 to the store buffer. Otherwise R0 is old, mark R0
// and add it to the mark list.
COMPILE_ASSERT(kWriteBarrierObjectReg == R1);
COMPILE_ASSERT(kWriteBarrierValueReg == R0);
COMPILE_ASSERT(kWriteBarrierSlotReg == R9);
static void GenerateWriteBarrierStubHelper(Assembler* assembler,
                                           Address stub_code,
                                           bool cards) {
  Label add_to_mark_stack, remember_card;
  __ tst(R0, Operand(1 << target::ObjectAlignment::kNewObjectBitPosition));
  __ b(&add_to_mark_stack, ZERO);

  if (cards) {
    __ ldr(TMP, FieldAddress(R1, target::Object::tags_offset()));
    __ tst(TMP, Operand(1 << target::RawObject::kCardRememberedBit));
    __ b(&remember_card, NOT_ZERO);
  } else {
#if defined(DEBUG)
    Label ok;
    __ ldr(TMP, FieldAddress(R1, target::Object::tags_offset()));
    __ tst(TMP, Operand(1 << target::RawObject::kCardRememberedBit));
    __ b(&ok, ZERO);
    __ Stop("Wrong barrier");
    __ Bind(&ok);
#endif
  }

  // Save values being destroyed.
  __ PushList((1 << R2) | (1 << R3) | (1 << R4));

  if (TargetCPUFeatures::arm_version() == ARMv5TE) {
// TODO(21263): Implement 'swp' and use it below.
#if !defined(USING_SIMULATOR)
    ASSERT(OS::NumberOfAvailableProcessors() <= 1);
#endif
    __ ldr(R2, FieldAddress(R1, target::Object::tags_offset()));
    __ bic(R2, R2, Operand(1 << target::RawObject::kOldAndNotRememberedBit));
    __ str(R2, FieldAddress(R1, target::Object::tags_offset()));
  } else {
    // Atomically set the remembered bit of the object header.
    ASSERT(target::Object::tags_offset() == 0);
    __ sub(R3, R1, Operand(kHeapObjectTag));
    // R3: Untagged address of header word (ldrex/strex do not support offsets).
    Label retry;
    __ Bind(&retry);
    __ ldrex(R2, R3);
    __ bic(R2, R2, Operand(1 << target::RawObject::kOldAndNotRememberedBit));
    __ strex(R4, R2, R3);
    __ cmp(R4, Operand(1));
    __ b(&retry, EQ);
  }

  // Load the StoreBuffer block out of the thread. Then load top_ out of the
  // StoreBufferBlock and add the address to the pointers_.
  __ ldr(R4, Address(THR, target::Thread::store_buffer_block_offset()));
  __ ldr(R2, Address(R4, target::StoreBufferBlock::top_offset()));
  __ add(R3, R4, Operand(R2, LSL, target::kWordSizeLog2));
  __ str(R1, Address(R3, target::StoreBufferBlock::pointers_offset()));

  // Increment top_ and check for overflow.
  // R2: top_.
  // R4: StoreBufferBlock.
  Label overflow;
  __ add(R2, R2, Operand(1));
  __ str(R2, Address(R4, target::StoreBufferBlock::top_offset()));
  __ CompareImmediate(R2, target::StoreBufferBlock::kSize);
  // Restore values.
  __ PopList((1 << R2) | (1 << R3) | (1 << R4));
  __ b(&overflow, EQ);
  __ Ret();

  // Handle overflow: Call the runtime leaf function.
  __ Bind(&overflow);
  // Setup frame, push callee-saved registers.

  __ Push(CODE_REG);
  __ ldr(CODE_REG, stub_code);
  __ EnterCallRuntimeFrame(0 * target::kWordSize);
  __ mov(R0, Operand(THR));
  __ CallRuntime(kStoreBufferBlockProcessRuntimeEntry, 1);
  // Restore callee-saved registers, tear down frame.
  __ LeaveCallRuntimeFrame();
  __ Pop(CODE_REG);
  __ Ret();

  __ Bind(&add_to_mark_stack);
  __ PushList((1 << R2) | (1 << R3) | (1 << R4));  // Spill.

  Label marking_retry, lost_race, marking_overflow;
  if (TargetCPUFeatures::arm_version() == ARMv5TE) {
// TODO(21263): Implement 'swp' and use it below.
#if !defined(USING_SIMULATOR)
    ASSERT(OS::NumberOfAvailableProcessors() <= 1);
#endif
    __ ldr(R2, FieldAddress(R0, target::Object::tags_offset()));
    __ bic(R2, R2, Operand(1 << target::RawObject::kOldAndNotMarkedBit));
    __ str(R2, FieldAddress(R0, target::Object::tags_offset()));
  } else {
    // Atomically clear kOldAndNotMarkedBit.
    ASSERT(target::Object::tags_offset() == 0);
    __ sub(R3, R0, Operand(kHeapObjectTag));
    // R3: Untagged address of header word (ldrex/strex do not support offsets).
    __ Bind(&marking_retry);
    __ ldrex(R2, R3);
    __ tst(R2, Operand(1 << target::RawObject::kOldAndNotMarkedBit));
    __ b(&lost_race, ZERO);
    __ bic(R2, R2, Operand(1 << target::RawObject::kOldAndNotMarkedBit));
    __ strex(R4, R2, R3);
    __ cmp(R4, Operand(1));
    __ b(&marking_retry, EQ);
  }

  __ ldr(R4, Address(THR, target::Thread::marking_stack_block_offset()));
  __ ldr(R2, Address(R4, target::MarkingStackBlock::top_offset()));
  __ add(R3, R4, Operand(R2, LSL, target::kWordSizeLog2));
  __ str(R0, Address(R3, target::MarkingStackBlock::pointers_offset()));
  __ add(R2, R2, Operand(1));
  __ str(R2, Address(R4, target::MarkingStackBlock::top_offset()));
  __ CompareImmediate(R2, target::MarkingStackBlock::kSize);
  __ PopList((1 << R4) | (1 << R2) | (1 << R3));  // Unspill.
  __ b(&marking_overflow, EQ);
  __ Ret();

  __ Bind(&marking_overflow);
  __ Push(CODE_REG);
  __ ldr(CODE_REG, stub_code);
  __ EnterCallRuntimeFrame(0 * target::kWordSize);
  __ mov(R0, Operand(THR));
  __ CallRuntime(kMarkingStackBlockProcessRuntimeEntry, 1);
  __ LeaveCallRuntimeFrame();
  __ Pop(CODE_REG);
  __ Ret();

  __ Bind(&lost_race);
  __ PopList((1 << R2) | (1 << R3) | (1 << R4));  // Unspill.
  __ Ret();

  if (cards) {
    Label remember_card_slow;

    // Get card table.
    __ Bind(&remember_card);
    __ AndImmediate(TMP, R1, target::kPageMask);  // HeapPage.
    __ ldr(TMP,
           Address(TMP, target::HeapPage::card_table_offset()));  // Card table.
    __ cmp(TMP, Operand(0));
    __ b(&remember_card_slow, EQ);

    // Dirty the card.
    __ AndImmediate(TMP, R1, target::kPageMask);  // HeapPage.
    __ sub(R9, R9, Operand(TMP));                 // Offset in page.
    __ ldr(TMP,
           Address(TMP, target::HeapPage::card_table_offset()));  // Card table.
    __ add(TMP, TMP,
           Operand(R9, LSR,
                   target::HeapPage::kBytesPerCardLog2));  // Card address.
    __ strb(R1,
            Address(TMP, 0));  // Low byte of R0 is non-zero from object tag.
    __ Ret();

    // Card table not yet allocated.
    __ Bind(&remember_card_slow);
    __ Push(CODE_REG);
    __ Push(R0);
    __ Push(R1);
    __ ldr(CODE_REG, stub_code);
    __ mov(R0, Operand(R1));  // Arg0 = Object
    __ mov(R1, Operand(R9));  // Arg1 = Slot
    __ EnterCallRuntimeFrame(0);
    __ CallRuntime(kRememberCardRuntimeEntry, 2);
    __ LeaveCallRuntimeFrame();
    __ Pop(R1);
    __ Pop(R0);
    __ Pop(CODE_REG);
    __ Ret();
  }
}

void StubCodeCompiler::GenerateWriteBarrierStub(Assembler* assembler) {
  GenerateWriteBarrierStubHelper(
      assembler, Address(THR, target::Thread::write_barrier_code_offset()),
      false);
}

void StubCodeCompiler::GenerateArrayWriteBarrierStub(Assembler* assembler) {
  GenerateWriteBarrierStubHelper(
      assembler,
      Address(THR, target::Thread::array_write_barrier_code_offset()), true);
}

// Called for inline allocation of objects.
// Input parameters:
//   LR : return address.
//   SP + 0 : type arguments object (only if class is parameterized).
void StubCodeCompiler::GenerateAllocationStubForClass(Assembler* assembler,
                                                      const Class& cls) {
  // The generated code is different if the class is parameterized.
  const bool is_cls_parameterized = target::Class::NumTypeArguments(cls) > 0;
  ASSERT(!is_cls_parameterized || target::Class::TypeArgumentsFieldOffset(
                                      cls) != target::Class::kNoTypeArguments);

  const Register kNullReg = R8;
  const Register kOtherNullReg = R9;
  const Register kTypeArgumentsReg = R3;
  const Register kInstanceReg = R0;
  const Register kEndReg = R1;
  const Register kEndOfInstanceReg = R2;

  // kInlineInstanceSize is a constant used as a threshold for determining
  // when the object initialization should be done as a loop or as
  // straight line code.
  const int kInlineInstanceSize = 12;
  const intptr_t instance_size = target::Class::InstanceSize(cls);
  ASSERT(instance_size > 0);
  ASSERT(instance_size % target::ObjectAlignment::kObjectAlignment == 0);
  if (is_cls_parameterized) {
    __ ldr(kTypeArgumentsReg, Address(SP, 0));
  }

  __ LoadObject(kNullReg, NullObject());
  if (FLAG_inline_alloc &&
      target::Heap::IsAllocatableInNewSpace(instance_size) &&
      !target::Class::TraceAllocation(cls)) {
    Label slow_case;

    // Allocate the object and update top to point to
    // next object start and initialize the allocated object.

    RELEASE_ASSERT((target::Thread::top_offset() + target::kWordSize) ==
                   target::Thread::end_offset());
    __ ldrd(kInstanceReg, kEndReg, THR, target::Thread::top_offset());
    __ AddImmediate(kEndOfInstanceReg, kInstanceReg, instance_size);
    __ cmp(kEndOfInstanceReg, Operand(kEndReg));
    if (FLAG_use_slow_path) {
      __ b(&slow_case);
    } else {
      __ b(&slow_case, CS);  // Unsigned higher or equal.
    }
    __ str(kEndOfInstanceReg, Address(THR, target::Thread::top_offset()));

    // Load the address of the allocation stats table. We split up the load
    // and the increment so that the dependent load is not too nearby.
    NOT_IN_PRODUCT(static Register kAllocationStatsReg = R4);
    NOT_IN_PRODUCT(__ LoadAllocationStatsAddress(kAllocationStatsReg,
                                                 target::Class::GetId(cls)));

    // Set the tags.
    ASSERT(target::Class::GetId(cls) != kIllegalCid);
    const uint32_t tags = target::MakeTagWordForNewSpaceObject(
        target::Class::GetId(cls), instance_size);

    __ LoadImmediate(R1, tags);
    __ str(R1, Address(kInstanceReg, target::Object::tags_offset()));
    __ add(kInstanceReg, kInstanceReg, Operand(kHeapObjectTag));

    // First try inlining the initialization without a loop.
    if (instance_size < (kInlineInstanceSize * target::kWordSize)) {
      intptr_t begin_offset =
          target::Instance::first_field_offset() - kHeapObjectTag;
      intptr_t end_offset = instance_size - kHeapObjectTag;
      if ((end_offset - begin_offset) >= (2 * target::kWordSize)) {
        __ mov(kOtherNullReg, Operand(kNullReg));
      }
      __ InitializeFieldsNoBarrierUnrolled(kInstanceReg, kInstanceReg,
                                           begin_offset, end_offset, kNullReg,
                                           kOtherNullReg);
    } else {
      __ add(R1, kInstanceReg,
             Operand(target::Instance::first_field_offset() - kHeapObjectTag));
      __ mov(kOtherNullReg, Operand(kNullReg));
      __ InitializeFieldsNoBarrier(kInstanceReg, R1, kEndOfInstanceReg,
                                   kNullReg, kOtherNullReg);
    }
    if (is_cls_parameterized) {
      const intptr_t offset = target::Class::TypeArgumentsFieldOffset(cls);
      __ StoreIntoObjectNoBarrier(
          kInstanceReg, FieldAddress(kInstanceReg, offset), kTypeArgumentsReg);
    }

    // Update allocation stats.
    NOT_IN_PRODUCT(__ IncrementAllocationStats(kAllocationStatsReg,
                                               target::Class::GetId(cls)));

    __ Ret();
    __ Bind(&slow_case);
  }
  // Create a stub frame as we are pushing some objects on the stack before
  // calling into the runtime.
  __ EnterStubFrame();  // Uses pool pointer to pass cls to runtime.
  __ LoadObject(R1, CastHandle<Object>(cls));
  __ PushList(1 << kNullReg | 1 << R1);  // Pushes cls, result slot.
  __ Push(is_cls_parameterized ? kTypeArgumentsReg : kNullReg);
  __ CallRuntime(kAllocateObjectRuntimeEntry, 2);  // Allocate object.
  __ ldr(
      kInstanceReg,
      Address(SP,
              2 * target::kWordSize));  // Pop result (newly allocated object).
  __ LeaveDartFrameAndReturn();         // Restores correct SP.
}

// Called for invoking "dynamic noSuchMethod(Invocation invocation)" function
// from the entry code of a dart function after an error in passed argument
// name or number is detected.
// Input parameters:
//  LR : return address.
//  SP : address of last argument.
//  R4: arguments descriptor array.
void StubCodeCompiler::GenerateCallClosureNoSuchMethodStub(
    Assembler* assembler) {
  __ EnterStubFrame();

  // Load the receiver.
  __ ldr(R2, FieldAddress(R4, target::ArgumentsDescriptor::count_offset()));
  __ add(IP, FP, Operand(R2, LSL, 1));  // R2 is Smi.
  __ ldr(R8, Address(IP, target::frame_layout.param_end_from_fp *
                             target::kWordSize));

  // Push space for the return value.
  // Push the receiver.
  // Push arguments descriptor array.
  __ LoadImmediate(IP, 0);
  __ PushList((1 << R4) | (1 << R8) | (1 << IP));

  // Adjust arguments count.
  __ ldr(R3,
         FieldAddress(R4, target::ArgumentsDescriptor::type_args_len_offset()));
  __ cmp(R3, Operand(0));
  __ AddImmediate(R2, R2, target::ToRawSmi(1),
                  NE);  // Include the type arguments.

  // R2: Smi-tagged arguments array length.
  PushArrayOfArguments(assembler);

  const intptr_t kNumArgs = 3;
  __ CallRuntime(kInvokeClosureNoSuchMethodRuntimeEntry, kNumArgs);
  // noSuchMethod on closures always throws an error, so it will never return.
  __ bkpt(0);
}

//  R8: function object.
//  R9: inline cache data object.
// Cannot use function object from ICData as it may be the inlined
// function and not the top-scope function.
void StubCodeCompiler::GenerateOptimizedUsageCounterIncrement(
    Assembler* assembler) {
  Register ic_reg = R9;
  Register func_reg = R8;
  if (FLAG_trace_optimized_ic_calls) {
    __ EnterStubFrame();
    __ PushList((1 << R9) | (1 << R8));  // Preserve.
    __ Push(ic_reg);                     // Argument.
    __ Push(func_reg);                   // Argument.
    __ CallRuntime(kTraceICCallRuntimeEntry, 2);
    __ Drop(2);                         // Discard argument;
    __ PopList((1 << R9) | (1 << R8));  // Restore.
    __ LeaveStubFrame();
  }
  __ ldr(NOTFP,
         FieldAddress(func_reg, target::Function::usage_counter_offset()));
  __ add(NOTFP, NOTFP, Operand(1));
  __ str(NOTFP,
         FieldAddress(func_reg, target::Function::usage_counter_offset()));
}

// Loads function into 'temp_reg'.
void StubCodeCompiler::GenerateUsageCounterIncrement(Assembler* assembler,
                                                     Register temp_reg) {
  if (FLAG_optimization_counter_threshold >= 0) {
    Register ic_reg = R9;
    Register func_reg = temp_reg;
    ASSERT(temp_reg == R8);
    __ Comment("Increment function counter");
    __ ldr(func_reg, FieldAddress(ic_reg, target::ICData::owner_offset()));
    __ ldr(NOTFP,
           FieldAddress(func_reg, target::Function::usage_counter_offset()));
    __ add(NOTFP, NOTFP, Operand(1));
    __ str(NOTFP,
           FieldAddress(func_reg, target::Function::usage_counter_offset()));
  }
}

// Note: R9 must be preserved.
// Attempt a quick Smi operation for known operations ('kind'). The ICData
// must have been primed with a Smi/Smi check that will be used for counting
// the invocations.
static void EmitFastSmiOp(Assembler* assembler,
                          Token::Kind kind,
                          intptr_t num_args,
                          Label* not_smi_or_overflow) {
  __ Comment("Fast Smi op");
  __ ldr(R0, Address(SP, 0 * target::kWordSize));
  __ ldr(R1, Address(SP, 1 * target::kWordSize));
  __ orr(TMP, R0, Operand(R1));
  __ tst(TMP, Operand(kSmiTagMask));
  __ b(not_smi_or_overflow, NE);
  switch (kind) {
    case Token::kADD: {
      __ adds(R0, R1, Operand(R0));   // Adds.
      __ b(not_smi_or_overflow, VS);  // Branch if overflow.
      break;
    }
    case Token::kSUB: {
      __ subs(R0, R1, Operand(R0));   // Subtract.
      __ b(not_smi_or_overflow, VS);  // Branch if overflow.
      break;
    }
    case Token::kEQ: {
      __ cmp(R0, Operand(R1));
      __ LoadObject(R0, CastHandle<Object>(TrueObject()), EQ);
      __ LoadObject(R0, CastHandle<Object>(FalseObject()), NE);
      break;
    }
    default:
      UNIMPLEMENTED();
  }
  // R9: IC data object (preserved).
  __ ldr(R8, FieldAddress(R9, target::ICData::entries_offset()));
  // R8: ic_data_array with check entries: classes and target functions.
  __ AddImmediate(R8, target::Array::data_offset() - kHeapObjectTag);
// R8: points directly to the first ic data array element.
#if defined(DEBUG)
  // Check that first entry is for Smi/Smi.
  Label error, ok;
  const intptr_t imm_smi_cid = target::ToRawSmi(kSmiCid);
  __ ldr(R1, Address(R8, 0));
  __ CompareImmediate(R1, imm_smi_cid);
  __ b(&error, NE);
  __ ldr(R1, Address(R8, target::kWordSize));
  __ CompareImmediate(R1, imm_smi_cid);
  __ b(&ok, EQ);
  __ Bind(&error);
  __ Stop("Incorrect IC data");
  __ Bind(&ok);
#endif
  if (FLAG_optimization_counter_threshold >= 0) {
    // Update counter, ignore overflow.
    const intptr_t count_offset =
        target::ICData::CountIndexFor(num_args) * target::kWordSize;
    __ LoadFromOffset(kWord, R1, R8, count_offset);
    __ adds(R1, R1, Operand(target::ToRawSmi(1)));
    __ StoreIntoSmiField(Address(R8, count_offset), R1);
  }
  __ Ret();
}

// Generate inline cache check for 'num_args'.
//  LR: return address.
//  R9: inline cache data object.
// Control flow:
// - If receiver is null -> jump to IC miss.
// - If receiver is Smi -> load Smi class.
// - If receiver is not-Smi -> load receiver's class.
// - Check if 'num_args' (including receiver) match any IC data group.
// - Match found -> jump to target.
// - Match not found -> jump to IC miss.
void StubCodeCompiler::GenerateNArgsCheckInlineCacheStub(
    Assembler* assembler,
    intptr_t num_args,
    const RuntimeEntry& handle_ic_miss,
    Token::Kind kind,
    bool optimized,
    bool exactness_check /* = false */) {
  ASSERT(!exactness_check);
  __ CheckCodePointer();
  ASSERT(num_args == 1 || num_args == 2);
#if defined(DEBUG)
  {
    Label ok;
    // Check that the IC data array has NumArgsTested() == num_args.
    // 'NumArgsTested' is stored in the least significant bits of 'state_bits'.
    __ ldr(R8, FieldAddress(R9, target::ICData::state_bits_offset()));
    ASSERT(target::ICData::NumArgsTestedShift() == 0);  // No shift needed.
    __ and_(R8, R8, Operand(target::ICData::NumArgsTestedMask()));
    __ CompareImmediate(R8, num_args);
    __ b(&ok, EQ);
    __ Stop("Incorrect stub for IC data");
    __ Bind(&ok);
  }
#endif  // DEBUG

#if !defined(PRODUCT)
  Label stepping, done_stepping;
  if (!optimized) {
    __ Comment("Check single stepping");
    __ LoadIsolate(R8);
    __ ldrb(R8, Address(R8, target::Isolate::single_step_offset()));
    __ CompareImmediate(R8, 0);
    __ b(&stepping, NE);
    __ Bind(&done_stepping);
  }
#endif

  Label not_smi_or_overflow;
  if (kind != Token::kILLEGAL) {
    EmitFastSmiOp(assembler, kind, num_args, &not_smi_or_overflow);
  }
  __ Bind(&not_smi_or_overflow);

  __ Comment("Extract ICData initial values and receiver cid");
  // Load arguments descriptor into R4.
  __ ldr(R4, FieldAddress(R9, target::ICData::arguments_descriptor_offset()));
  // Loop that checks if there is an IC data match.
  Label loop, found, miss;
  // R9: IC data object (preserved).
  __ ldr(R8, FieldAddress(R9, target::ICData::entries_offset()));
  // R8: ic_data_array with check entries: classes and target functions.
  const int kIcDataOffset = target::Array::data_offset() - kHeapObjectTag;
  // R8: points at the IC data array.

  // Get the receiver's class ID (first read number of arguments from
  // arguments descriptor array and then access the receiver from the stack).
  __ ldr(NOTFP, FieldAddress(R4, target::ArgumentsDescriptor::count_offset()));
  __ sub(NOTFP, NOTFP, Operand(target::ToRawSmi(1)));
  // NOTFP: argument_count - 1 (smi).

  __ Comment("ICData loop");

  __ ldr(R0, Address(SP, NOTFP, LSL, 1));  // NOTFP (argument_count - 1) is Smi.
  __ LoadTaggedClassIdMayBeSmi(R0, R0);
  if (num_args == 2) {
    __ sub(R1, NOTFP, Operand(target::ToRawSmi(1)));
    __ ldr(R1, Address(SP, R1, LSL, 1));  // R1 (argument_count - 2) is Smi.
    __ LoadTaggedClassIdMayBeSmi(R1, R1);
  }

  // We unroll the generic one that is generated once more than the others.
  const bool optimize = kind == Token::kILLEGAL;

  __ Bind(&loop);
  for (int unroll = optimize ? 4 : 2; unroll >= 0; unroll--) {
    Label update;

    __ ldr(R2, Address(R8, kIcDataOffset));
    __ cmp(R0, Operand(R2));  // Class id match?
    if (num_args == 2) {
      __ b(&update, NE);  // Continue.
      __ ldr(R2, Address(R8, kIcDataOffset + target::kWordSize));
      __ cmp(R1, Operand(R2));  // Class id match?
    }
    __ b(&found, EQ);  // Break.

    __ Bind(&update);

    const intptr_t entry_size =
        target::ICData::TestEntryLengthFor(num_args, exactness_check) *
        target::kWordSize;
    __ AddImmediate(R8, entry_size);  // Next entry.

    __ CompareImmediate(R2, target::ToRawSmi(kIllegalCid));  // Done?
    if (unroll == 0) {
      __ b(&loop, NE);
    } else {
      __ b(&miss, EQ);
    }
  }

  __ Bind(&miss);
  __ Comment("IC miss");
  // Compute address of arguments.
  // NOTFP: argument_count - 1 (smi).
  __ add(NOTFP, SP, Operand(NOTFP, LSL, 1));  // NOTFP is Smi.
  // NOTFP: address of receiver.
  // Create a stub frame as we are pushing some objects on the stack before
  // calling into the runtime.
  __ EnterStubFrame();
  __ LoadImmediate(R0, 0);
  // Preserve IC data object and arguments descriptor array and
  // setup space on stack for result (target code object).
  __ PushList((1 << R0) | (1 << R4) | (1 << R9));
  // Push call arguments.
  for (intptr_t i = 0; i < num_args; i++) {
    __ LoadFromOffset(kWord, IP, NOTFP, -i * target::kWordSize);
    __ Push(IP);
  }
  // Pass IC data object.
  __ Push(R9);
  __ CallRuntime(handle_ic_miss, num_args + 1);
  // Remove the call arguments pushed earlier, including the IC data object.
  __ Drop(num_args + 1);
  // Pop returned function object into R0.
  // Restore arguments descriptor array and IC data array.
  __ PopList((1 << R0) | (1 << R4) | (1 << R9));
  __ RestoreCodePointer();
  __ LeaveStubFrame();
  Label call_target_function;
  if (!FLAG_lazy_dispatchers) {
    GenerateDispatcherCode(assembler, &call_target_function);
  } else {
    __ b(&call_target_function);
  }

  __ Bind(&found);
  // R8: pointer to an IC data check group.
  const intptr_t target_offset =
      target::ICData::TargetIndexFor(num_args) * target::kWordSize;
  const intptr_t count_offset =
      target::ICData::CountIndexFor(num_args) * target::kWordSize;
  __ LoadFromOffset(kWord, R0, R8, kIcDataOffset + target_offset);

  if (FLAG_optimization_counter_threshold >= 0) {
    __ Comment("Update caller's counter");
    __ LoadFromOffset(kWord, R1, R8, kIcDataOffset + count_offset);
    // Ignore overflow.
    __ adds(R1, R1, Operand(target::ToRawSmi(1)));
    __ StoreIntoSmiField(Address(R8, kIcDataOffset + count_offset), R1);
  }

  __ Comment("Call target");
  __ Bind(&call_target_function);
  // R0: target function.
  __ ldr(CODE_REG, FieldAddress(R0, target::Function::code_offset()));
  __ Branch(FieldAddress(R0, target::Function::entry_point_offset()));

#if !defined(PRODUCT)
  if (!optimized) {
    __ Bind(&stepping);
    __ EnterStubFrame();
    __ Push(R9);  // Preserve IC data.
    __ CallRuntime(kSingleStepHandlerRuntimeEntry, 0);
    __ Pop(R9);
    __ RestoreCodePointer();
    __ LeaveStubFrame();
    __ b(&done_stepping);
  }
#endif
}

// Use inline cache data array to invoke the target or continue in inline
// cache miss handler. Stub for 1-argument check (receiver class).
//  LR: return address.
//  R9: inline cache data object.
// Inline cache data object structure:
// 0: function-name
// 1: N, number of arguments checked.
// 2 .. (length - 1): group of checks, each check containing:
//   - N classes.
//   - 1 target function.
void StubCodeCompiler::GenerateOneArgCheckInlineCacheStub(
    Assembler* assembler) {
  GenerateUsageCounterIncrement(assembler, R8);
  GenerateNArgsCheckInlineCacheStub(
      assembler, 1, kInlineCacheMissHandlerOneArgRuntimeEntry, Token::kILLEGAL);
}

void StubCodeCompiler::GenerateOneArgCheckInlineCacheWithExactnessCheckStub(
    Assembler* assembler) {
  __ Stop("Unimplemented");
}

void StubCodeCompiler::GenerateTwoArgsCheckInlineCacheStub(
    Assembler* assembler) {
  GenerateUsageCounterIncrement(assembler, R8);
  GenerateNArgsCheckInlineCacheStub(assembler, 2,
                                    kInlineCacheMissHandlerTwoArgsRuntimeEntry,
                                    Token::kILLEGAL);
}

void StubCodeCompiler::GenerateSmiAddInlineCacheStub(Assembler* assembler) {
  GenerateUsageCounterIncrement(assembler, R8);
  GenerateNArgsCheckInlineCacheStub(
      assembler, 2, kInlineCacheMissHandlerTwoArgsRuntimeEntry, Token::kADD);
}

void StubCodeCompiler::GenerateSmiSubInlineCacheStub(Assembler* assembler) {
  GenerateUsageCounterIncrement(assembler, R8);
  GenerateNArgsCheckInlineCacheStub(
      assembler, 2, kInlineCacheMissHandlerTwoArgsRuntimeEntry, Token::kSUB);
}

void StubCodeCompiler::GenerateSmiEqualInlineCacheStub(Assembler* assembler) {
  GenerateUsageCounterIncrement(assembler, R8);
  GenerateNArgsCheckInlineCacheStub(
      assembler, 2, kInlineCacheMissHandlerTwoArgsRuntimeEntry, Token::kEQ);
}

void StubCodeCompiler::GenerateOneArgOptimizedCheckInlineCacheStub(
    Assembler* assembler) {
  GenerateOptimizedUsageCounterIncrement(assembler);
  GenerateNArgsCheckInlineCacheStub(assembler, 1,
                                    kInlineCacheMissHandlerOneArgRuntimeEntry,
                                    Token::kILLEGAL, true /* optimized */);
}

void StubCodeCompiler::
    GenerateOneArgOptimizedCheckInlineCacheWithExactnessCheckStub(
        Assembler* assembler) {
  __ Stop("Unimplemented");
}

void StubCodeCompiler::GenerateTwoArgsOptimizedCheckInlineCacheStub(
    Assembler* assembler) {
  GenerateOptimizedUsageCounterIncrement(assembler);
  GenerateNArgsCheckInlineCacheStub(assembler, 2,
                                    kInlineCacheMissHandlerTwoArgsRuntimeEntry,
                                    Token::kILLEGAL, true /* optimized */);
}

// Intermediary stub between a static call and its target. ICData contains
// the target function and the call count.
// R9: ICData
void StubCodeCompiler::GenerateZeroArgsUnoptimizedStaticCallStub(
    Assembler* assembler) {
  GenerateUsageCounterIncrement(assembler, R8);
#if defined(DEBUG)
  {
    Label ok;
    // Check that the IC data array has NumArgsTested() == 0.
    // 'NumArgsTested' is stored in the least significant bits of 'state_bits'.
    __ ldr(R8, FieldAddress(R9, target::ICData::state_bits_offset()));
    ASSERT(target::ICData::NumArgsTestedShift() == 0);  // No shift needed.
    __ and_(R8, R8, Operand(target::ICData::NumArgsTestedMask()));
    __ CompareImmediate(R8, 0);
    __ b(&ok, EQ);
    __ Stop("Incorrect IC data for unoptimized static call");
    __ Bind(&ok);
  }
#endif  // DEBUG

#if !defined(PRODUCT)
  // Check single stepping.
  Label stepping, done_stepping;
  __ LoadIsolate(R8);
  __ ldrb(R8, Address(R8, target::Isolate::single_step_offset()));
  __ CompareImmediate(R8, 0);
  __ b(&stepping, NE);
  __ Bind(&done_stepping);
#endif

  // R9: IC data object (preserved).
  __ ldr(R8, FieldAddress(R9, target::ICData::entries_offset()));
  // R8: ic_data_array with entries: target functions and count.
  __ AddImmediate(R8, target::Array::data_offset() - kHeapObjectTag);
  // R8: points directly to the first ic data array element.
  const intptr_t target_offset =
      target::ICData::TargetIndexFor(0) * target::kWordSize;
  const intptr_t count_offset =
      target::ICData::CountIndexFor(0) * target::kWordSize;

  if (FLAG_optimization_counter_threshold >= 0) {
    // Increment count for this call, ignore overflow.
    __ LoadFromOffset(kWord, R1, R8, count_offset);
    __ adds(R1, R1, Operand(target::ToRawSmi(1)));
    __ StoreIntoSmiField(Address(R8, count_offset), R1);
  }

  // Load arguments descriptor into R4.
  __ ldr(R4, FieldAddress(R9, target::ICData::arguments_descriptor_offset()));

  // Get function and call it, if possible.
  __ LoadFromOffset(kWord, R0, R8, target_offset);
  __ ldr(CODE_REG, FieldAddress(R0, target::Function::code_offset()));
  __ Branch(FieldAddress(R0, target::Function::entry_point_offset()));

#if !defined(PRODUCT)
  __ Bind(&stepping);
  __ EnterStubFrame();
  __ Push(R9);  // Preserve IC data.
  __ CallRuntime(kSingleStepHandlerRuntimeEntry, 0);
  __ Pop(R9);
  __ RestoreCodePointer();
  __ LeaveStubFrame();
  __ b(&done_stepping);
#endif
}

void StubCodeCompiler::GenerateOneArgUnoptimizedStaticCallStub(
    Assembler* assembler) {
  GenerateUsageCounterIncrement(assembler, R8);
  GenerateNArgsCheckInlineCacheStub(
      assembler, 1, kStaticCallMissHandlerOneArgRuntimeEntry, Token::kILLEGAL);
}

void StubCodeCompiler::GenerateTwoArgsUnoptimizedStaticCallStub(
    Assembler* assembler) {
  GenerateUsageCounterIncrement(assembler, R8);
  GenerateNArgsCheckInlineCacheStub(
      assembler, 2, kStaticCallMissHandlerTwoArgsRuntimeEntry, Token::kILLEGAL);
}

// Stub for compiling a function and jumping to the compiled code.
// R4: Arguments descriptor.
// R0: Function.
void StubCodeCompiler::GenerateLazyCompileStub(Assembler* assembler) {
  __ EnterStubFrame();
  __ PushList((1 << R0) | (1 << R4));  // Preserve arg desc, pass function.
  __ CallRuntime(kCompileFunctionRuntimeEntry, 1);
  __ PopList((1 << R0) | (1 << R4));
  __ LeaveStubFrame();

  // When using the interpreter, the function's code may now point to the
  // InterpretCall stub. Make sure R0, R4, and R9 are preserved.
  __ ldr(CODE_REG, FieldAddress(R0, target::Function::code_offset()));
  __ Branch(FieldAddress(R0, target::Function::entry_point_offset()));
}

// Stub for interpreting a function call.
// R4: Arguments descriptor.
// R0: Function.
void StubCodeCompiler::GenerateInterpretCallStub(Assembler* assembler) {
#if defined(DART_PRECOMPILED_RUNTIME)
  __ Stop("Not using interpreter")
#else
  __ EnterStubFrame();

#if defined(DEBUG)
  {
    Label ok;
    // Check that we are always entering from Dart code.
    __ LoadFromOffset(kWord, R8, THR, target::Thread::vm_tag_offset());
    __ CompareImmediate(R8, VMTag::kDartCompiledTagId);
    __ b(&ok, EQ);
    __ Stop("Not coming from Dart code.");
    __ Bind(&ok);
  }
#endif

  // Adjust arguments count for type arguments vector.
  __ LoadFieldFromOffset(kWord, R2, R4,
                         target::ArgumentsDescriptor::count_offset());
  __ SmiUntag(R2);
  __ LoadFieldFromOffset(kWord, R1, R4,
                         target::ArgumentsDescriptor::type_args_len_offset());
  __ cmp(R1, Operand(0));
  __ AddImmediate(R2, R2, 1, NE);  // Include the type arguments.

  // Compute argv.
  __ mov(R3, Operand(R2, LSL, 2));
  __ add(R3, FP, Operand(R3));
  __ AddImmediate(R3,
                  target::frame_layout.param_end_from_fp * target::kWordSize);

  // Indicate decreasing memory addresses of arguments with negative argc.
  __ rsb(R2, R2, Operand(0));

  // Align frame before entering C++ world. Fifth argument passed on the stack.
  __ ReserveAlignedFrameSpace(1 * target::kWordSize);

  // Pass arguments in registers.
  // R0: Function.
  __ mov(R1, Operand(R4));  // Arguments descriptor.
  // R2: Negative argc.
  // R3: Argv.
  __ str(THR, Address(SP, 0));  // Fifth argument: Thread.

  // Save exit frame information to enable stack walking as we are about
  // to transition to Dart VM C++ code.
  __ StoreToOffset(kWord, FP, THR,
                   target::Thread::top_exit_frame_info_offset());

  // Mark that the thread is executing VM code.
  __ LoadFromOffset(kWord, R5, THR,
                    target::Thread::interpret_call_entry_point_offset());
  __ StoreToOffset(kWord, R5, THR, target::Thread::vm_tag_offset());

  __ blx(R5);

  // Mark that the thread is executing Dart code.
  __ LoadImmediate(R2, VMTag::kDartCompiledTagId);
  __ StoreToOffset(kWord, R2, THR, target::Thread::vm_tag_offset());

  // Reset exit frame information in Isolate structure.
  __ LoadImmediate(R2, 0);
  __ StoreToOffset(kWord, R2, THR,
                   target::Thread::top_exit_frame_info_offset());

  __ LeaveStubFrame();
  __ Ret();
#endif  // defined(DART_PRECOMPILED_RUNTIME)
}

// R9: Contains an ICData.
void StubCodeCompiler::GenerateICCallBreakpointStub(Assembler* assembler) {
  __ EnterStubFrame();
  __ LoadImmediate(R0, 0);
  // Preserve arguments descriptor and make room for result.
  __ PushList((1 << R0) | (1 << R9));
  __ CallRuntime(kBreakpointRuntimeHandlerRuntimeEntry, 0);
  __ PopList((1 << R0) | (1 << R9));
  __ LeaveStubFrame();
  __ mov(CODE_REG, Operand(R0));
  __ Branch(FieldAddress(CODE_REG, target::Code::entry_point_offset()));
}

void StubCodeCompiler::GenerateRuntimeCallBreakpointStub(Assembler* assembler) {
  __ EnterStubFrame();
  __ LoadImmediate(R0, 0);
  // Make room for result.
  __ PushList((1 << R0));
  __ CallRuntime(kBreakpointRuntimeHandlerRuntimeEntry, 0);
  __ PopList((1 << CODE_REG));
  __ LeaveStubFrame();
  __ Branch(FieldAddress(CODE_REG, target::Code::entry_point_offset()));
}

// Called only from unoptimized code. All relevant registers have been saved.
void StubCodeCompiler::GenerateDebugStepCheckStub(Assembler* assembler) {
#if defined(PRODUCT)
  __ Ret();
#else
  // Check single stepping.
  Label stepping, done_stepping;
  __ LoadIsolate(R1);
  __ ldrb(R1, Address(R1, target::Isolate::single_step_offset()));
  __ CompareImmediate(R1, 0);
  __ b(&stepping, NE);
  __ Bind(&done_stepping);
  __ Ret();

  __ Bind(&stepping);
  __ EnterStubFrame();
  __ CallRuntime(kSingleStepHandlerRuntimeEntry, 0);
  __ LeaveStubFrame();
  __ b(&done_stepping);
#endif  // defined(PRODUCT)
}

// Used to check class and type arguments. Arguments passed in registers:
// LR: return address.
// R0: instance (must be preserved).
// R2: instantiator type arguments (only if n >= 4, can be raw_null).
// R1: function type arguments (only if n >= 4, can be raw_null).
// R3: target::SubtypeTestCache.
//
// Preserves R0/R2
//
// Result in R1: null -> not found, otherwise result (true or false).
static void GenerateSubtypeNTestCacheStub(Assembler* assembler, int n) {
  ASSERT(n == 1 || n == 2 || n == 4 || n == 6);

  const Register kCacheReg = R3;
  const Register kInstanceReg = R0;
  const Register kInstantiatorTypeArgumentsReg = R2;
  const Register kFunctionTypeArgumentsReg = R1;

  const Register kInstanceCidOrFunction = R8;
  const Register kInstanceInstantiatorTypeArgumentsReg = R4;
  const Register kInstanceParentFunctionTypeArgumentsReg = CODE_REG;
  const Register kInstanceDelayedFunctionTypeArgumentsReg = PP;

  const Register kNullReg = NOTFP;

  __ LoadObject(kNullReg, NullObject());

  // Free up these 2 registers to be used for 6-value test.
  if (n >= 6) {
    __ PushList(1 << kInstanceParentFunctionTypeArgumentsReg |
                1 << kInstanceDelayedFunctionTypeArgumentsReg);
  }

  // Loop initialization (moved up here to avoid having all dependent loads
  // after each other).
  __ ldr(kCacheReg,
         FieldAddress(kCacheReg, target::SubtypeTestCache::cache_offset()));
  __ AddImmediate(kCacheReg, target::Array::data_offset() - kHeapObjectTag);

  Label loop, not_closure;
  if (n >= 4) {
    __ LoadClassIdMayBeSmi(kInstanceCidOrFunction, kInstanceReg);
  } else {
    __ LoadClassId(kInstanceCidOrFunction, kInstanceReg);
  }
  __ CompareImmediate(kInstanceCidOrFunction, kClosureCid);
  __ b(&not_closure, NE);

  // Closure handling.
  {
    __ ldr(kInstanceCidOrFunction,
           FieldAddress(kInstanceReg, target::Closure::function_offset()));
    if (n >= 2) {
      __ ldr(
          kInstanceInstantiatorTypeArgumentsReg,
          FieldAddress(kInstanceReg,
                       target::Closure::instantiator_type_arguments_offset()));
      if (n >= 6) {
        ASSERT(n == 6);
        __ ldr(kInstanceParentFunctionTypeArgumentsReg,
               FieldAddress(kInstanceReg,
                            target::Closure::function_type_arguments_offset()));
        __ ldr(kInstanceDelayedFunctionTypeArgumentsReg,
               FieldAddress(kInstanceReg,
                            target::Closure::delayed_type_arguments_offset()));
      }
    }
    __ b(&loop);
  }

  // Non-Closure handling.
  {
    __ Bind(&not_closure);
    if (n >= 2) {
      Label has_no_type_arguments;
      __ LoadClassById(R9, kInstanceCidOrFunction);
      __ mov(kInstanceInstantiatorTypeArgumentsReg, Operand(kNullReg));
      __ ldr(R9,
             FieldAddress(
                 R9,
                 target::Class::type_arguments_field_offset_in_words_offset()));
      __ CompareImmediate(R9, target::Class::kNoTypeArguments);
      __ b(&has_no_type_arguments, EQ);
      __ add(R9, kInstanceReg, Operand(R9, LSL, 2));
      __ ldr(kInstanceInstantiatorTypeArgumentsReg, FieldAddress(R9, 0));
      __ Bind(&has_no_type_arguments);

      if (n >= 6) {
        __ mov(kInstanceParentFunctionTypeArgumentsReg, Operand(kNullReg));
        __ mov(kInstanceDelayedFunctionTypeArgumentsReg, Operand(kNullReg));
      }
    }
    __ SmiTag(kInstanceCidOrFunction);
  }

  Label found, not_found, next_iteration;

  // Loop header.
  __ Bind(&loop);
  __ ldr(R9, Address(kCacheReg,
                     target::kWordSize *
                         target::SubtypeTestCache::kInstanceClassIdOrFunction));
  __ cmp(R9, Operand(kNullReg));
  __ b(&not_found, EQ);
  __ cmp(R9, Operand(kInstanceCidOrFunction));
  if (n == 1) {
    __ b(&found, EQ);
  } else {
    __ b(&next_iteration, NE);
    __ ldr(R9, Address(kCacheReg,
                       target::kWordSize *
                           target::SubtypeTestCache::kInstanceTypeArguments));
    __ cmp(R9, Operand(kInstanceInstantiatorTypeArgumentsReg));
    if (n == 2) {
      __ b(&found, EQ);
    } else {
      __ b(&next_iteration, NE);
      __ ldr(R9,
             Address(kCacheReg,
                     target::kWordSize *
                         target::SubtypeTestCache::kInstantiatorTypeArguments));
      __ cmp(R9, Operand(kInstantiatorTypeArgumentsReg));
      __ b(&next_iteration, NE);
      __ ldr(R9, Address(kCacheReg,
                         target::kWordSize *
                             target::SubtypeTestCache::kFunctionTypeArguments));
      __ cmp(R9, Operand(kFunctionTypeArgumentsReg));
      if (n == 4) {
        __ b(&found, EQ);
      } else {
        ASSERT(n == 6);
        __ b(&next_iteration, NE);

        __ ldr(R9, Address(kCacheReg,
                           target::kWordSize *
                               target::SubtypeTestCache::
                                   kInstanceParentFunctionTypeArguments));
        __ cmp(R9, Operand(kInstanceParentFunctionTypeArgumentsReg));
        __ b(&next_iteration, NE);

        __ ldr(R9, Address(kCacheReg,
                           target::kWordSize *
                               target::SubtypeTestCache::
                                   kInstanceDelayedFunctionTypeArguments));
        __ cmp(R9, Operand(kInstanceDelayedFunctionTypeArgumentsReg));
        __ b(&found, EQ);
      }
    }
  }
  __ Bind(&next_iteration);
  __ AddImmediate(kCacheReg, target::kWordSize *
                                 target::SubtypeTestCache::kTestEntryLength);
  __ b(&loop);

  __ Bind(&found);
  __ ldr(R1, Address(kCacheReg, target::kWordSize *
                                    target::SubtypeTestCache::kTestResult));
  if (n >= 6) {
    __ PopList(1 << kInstanceParentFunctionTypeArgumentsReg |
               1 << kInstanceDelayedFunctionTypeArgumentsReg);
  }
  __ Ret();

  __ Bind(&not_found);
  __ mov(R1, Operand(kNullReg));
  if (n >= 6) {
    __ PopList(1 << kInstanceParentFunctionTypeArgumentsReg |
               1 << kInstanceDelayedFunctionTypeArgumentsReg);
  }
  __ Ret();
}

// See comment on [GenerateSubtypeNTestCacheStub].
void StubCodeCompiler::GenerateSubtype1TestCacheStub(Assembler* assembler) {
  GenerateSubtypeNTestCacheStub(assembler, 1);
}

// See comment on [GenerateSubtypeNTestCacheStub].
void StubCodeCompiler::GenerateSubtype2TestCacheStub(Assembler* assembler) {
  GenerateSubtypeNTestCacheStub(assembler, 2);
}

// See comment on [GenerateSubtypeNTestCacheStub].
void StubCodeCompiler::GenerateSubtype4TestCacheStub(Assembler* assembler) {
  GenerateSubtypeNTestCacheStub(assembler, 4);
}

// See comment on [GenerateSubtypeNTestCacheStub].
void StubCodeCompiler::GenerateSubtype6TestCacheStub(Assembler* assembler) {
  GenerateSubtypeNTestCacheStub(assembler, 6);
}

// Used to test whether a given value is of a given type (different variants,
// all have the same calling convention).
//
// Inputs:
//   - R0 : instance to test against.
//   - R2 : instantiator type arguments (if needed).
//   - R1 : function type arguments (if needed).
//
//   - R3 : subtype test cache.
//
//   - R8 : type to test against.
//   - R4 : name of destination variable.
//
// Preserves R0/R2.
//
// Note of warning: The caller will not populate CODE_REG and we have therefore
// no access to the pool.
void StubCodeCompiler::GenerateDefaultTypeTestStub(Assembler* assembler) {
  Label done;

  const Register kInstanceReg = R0;
  // Fast case for 'null'.
  __ CompareObject(kInstanceReg, NullObject());
  __ BranchIf(EQUAL, &done);

  __ ldr(CODE_REG, Address(THR, target::Thread::slow_type_test_stub_offset()));
  __ Branch(FieldAddress(CODE_REG, target::Code::entry_point_offset()));

  __ Bind(&done);
  __ Ret();
}

void StubCodeCompiler::GenerateTopTypeTypeTestStub(Assembler* assembler) {
  __ Ret();
}

void StubCodeCompiler::GenerateTypeRefTypeTestStub(Assembler* assembler) {
  const Register kTypeRefReg = R8;

  // We dereference the TypeRef and tail-call to it's type testing stub.
  __ ldr(kTypeRefReg,
         FieldAddress(kTypeRefReg, target::TypeRef::type_offset()));
  __ ldr(R9, FieldAddress(
                 kTypeRefReg,
                 target::AbstractType::type_test_stub_entry_point_offset()));
  __ bx(R9);
}

void StubCodeCompiler::GenerateUnreachableTypeTestStub(Assembler* assembler) {
  __ Breakpoint();
}

static void InvokeTypeCheckFromTypeTestStub(Assembler* assembler,
                                            TypeCheckMode mode) {
  const Register kInstanceReg = R0;
  const Register kInstantiatorTypeArgumentsReg = R2;
  const Register kFunctionTypeArgumentsReg = R1;
  const Register kDstTypeReg = R8;
  const Register kSubtypeTestCacheReg = R3;

  __ PushObject(NullObject());  // Make room for result.
  __ Push(kInstanceReg);
  __ Push(kDstTypeReg);
  __ Push(kInstantiatorTypeArgumentsReg);
  __ Push(kFunctionTypeArgumentsReg);
  __ PushObject(NullObject());
  __ Push(kSubtypeTestCacheReg);
  __ PushImmediate(target::ToRawSmi(mode));
  __ CallRuntime(kTypeCheckRuntimeEntry, 7);
  __ Drop(1);  // mode
  __ Pop(kSubtypeTestCacheReg);
  __ Drop(1);  // dst_name
  __ Pop(kFunctionTypeArgumentsReg);
  __ Pop(kInstantiatorTypeArgumentsReg);
  __ Pop(kDstTypeReg);
  __ Pop(kInstanceReg);
  __ Drop(1);  // Discard return value.
}

void StubCodeCompiler::GenerateLazySpecializeTypeTestStub(
    Assembler* assembler) {
  const Register kInstanceReg = R0;
  Label done;

  __ CompareObject(kInstanceReg, NullObject());
  __ BranchIf(EQUAL, &done);

  __ ldr(CODE_REG,
         Address(THR, target::Thread::lazy_specialize_type_test_stub_offset()));
  __ EnterStubFrame();
  InvokeTypeCheckFromTypeTestStub(assembler, kTypeCheckFromLazySpecializeStub);
  __ LeaveStubFrame();

  __ Bind(&done);
  __ Ret();
}

void StubCodeCompiler::GenerateSlowTypeTestStub(Assembler* assembler) {
  Label done, call_runtime;

  const Register kInstanceReg = R0;
  const Register kFunctionTypeArgumentsReg = R1;
  const Register kDstTypeReg = R8;
  const Register kSubtypeTestCacheReg = R3;

  __ EnterStubFrame();

#ifdef DEBUG
  // Guaranteed by caller.
  Label no_error;
  __ CompareObject(kInstanceReg, NullObject());
  __ BranchIf(NOT_EQUAL, &no_error);
  __ Breakpoint();
  __ Bind(&no_error);
#endif

  // If the subtype-cache is null, it needs to be lazily-created by the runtime.
  __ CompareObject(kSubtypeTestCacheReg, NullObject());
  __ BranchIf(EQUAL, &call_runtime);

  const Register kTmp = NOTFP;

  // If this is not a [Type] object, we'll go to the runtime.
  Label is_simple_case, is_complex_case;
  __ LoadClassId(kTmp, kDstTypeReg);
  __ cmp(kTmp, Operand(kTypeCid));
  __ BranchIf(NOT_EQUAL, &is_complex_case);

  // Check whether this [Type] is instantiated/uninstantiated.
  __ ldrb(kTmp, FieldAddress(kDstTypeReg, target::Type::type_state_offset()));
  __ cmp(kTmp,
         Operand(target::RawAbstractType::kTypeStateFinalizedInstantiated));
  __ BranchIf(NOT_EQUAL, &is_complex_case);

  // Check whether this [Type] is a function type.
  __ ldr(kTmp, FieldAddress(kDstTypeReg, target::Type::signature_offset()));
  __ CompareObject(kTmp, NullObject());
  __ BranchIf(NOT_EQUAL, &is_complex_case);

  // This [Type] could be a FutureOr. Subtype2TestCache does not support Smi.
  __ BranchIfSmi(kInstanceReg, &is_complex_case);

  // Fall through to &is_simple_case

  const intptr_t kRegsToSave = (1 << kSubtypeTestCacheReg) |
                               (1 << kDstTypeReg) |
                               (1 << kFunctionTypeArgumentsReg);

  __ Bind(&is_simple_case);
  {
    __ PushList(kRegsToSave);
    __ BranchLink(StubCodeSubtype2TestCache());
    __ CompareObject(R1, CastHandle<Object>(TrueObject()));
    __ PopList(kRegsToSave);
    __ BranchIf(EQUAL, &done);  // Cache said: yes.
    __ Jump(&call_runtime);
  }

  __ Bind(&is_complex_case);
  {
    __ PushList(kRegsToSave);
    __ BranchLink(StubCodeSubtype6TestCache());
    __ CompareObject(R1, CastHandle<Object>(TrueObject()));
    __ PopList(kRegsToSave);
    __ BranchIf(EQUAL, &done);  // Cache said: yes.
    // Fall through to runtime_call
  }

  __ Bind(&call_runtime);

  // We cannot really ensure here that dynamic/Object/void never occur here
  // (though it is guaranteed at dart_precompiled_runtime time).  This is
  // because we do constant evaluation with default stubs and only install
  // optimized versions before writing out the AOT snapshot.
  // So dynamic/Object/void will run with default stub in constant evaluation.
  __ CompareObject(kDstTypeReg, CastHandle<Object>(DynamicType()));
  __ BranchIf(EQUAL, &done);
  __ CompareObject(kDstTypeReg, CastHandle<Object>(ObjectType()));
  __ BranchIf(EQUAL, &done);
  __ CompareObject(kDstTypeReg, CastHandle<Object>(VoidType()));
  __ BranchIf(EQUAL, &done);

  InvokeTypeCheckFromTypeTestStub(assembler, kTypeCheckFromSlowStub);

  __ Bind(&done);
  __ LeaveStubFrame();
  __ Ret();
}

// Return the current stack pointer address, used to do stack alignment checks.
void StubCodeCompiler::GenerateGetCStackPointerStub(Assembler* assembler) {
  __ mov(R0, Operand(SP));
  __ Ret();
}

// Jump to a frame on the call stack.
// LR: return address.
// R0: program_counter.
// R1: stack_pointer.
// R2: frame_pointer.
// R3: thread.
// Does not return.
void StubCodeCompiler::GenerateJumpToFrameStub(Assembler* assembler) {
  ASSERT(kExceptionObjectReg == R0);
  ASSERT(kStackTraceObjectReg == R1);
  __ mov(IP, Operand(R1));   // Copy Stack pointer into IP.
  __ mov(LR, Operand(R0));   // Program counter.
  __ mov(THR, Operand(R3));  // Thread.
  __ mov(FP, Operand(R2));   // Frame_pointer.
  __ mov(SP, Operand(IP));   // Set Stack pointer.
  // Set the tag.
  __ LoadImmediate(R2, VMTag::kDartCompiledTagId);
  __ StoreToOffset(kWord, R2, THR, target::Thread::vm_tag_offset());
  // Clear top exit frame.
  __ LoadImmediate(R2, 0);
  __ StoreToOffset(kWord, R2, THR,
                   target::Thread::top_exit_frame_info_offset());
  // Restore the pool pointer.
  __ RestoreCodePointer();
  if (FLAG_precompiled_mode && FLAG_use_bare_instructions) {
    __ ldr(PP, Address(THR, target::Thread::global_object_pool_offset()));
    __ set_constant_pool_allowed(true);
  } else {
    __ LoadPoolPointer();
  }
  __ bx(LR);  // Jump to continuation point.
}

// Run an exception handler.  Execution comes from JumpToFrame
// stub or from the simulator.
//
// The arguments are stored in the Thread object.
// Does not return.
void StubCodeCompiler::GenerateRunExceptionHandlerStub(Assembler* assembler) {
  __ LoadFromOffset(kWord, LR, THR, target::Thread::resume_pc_offset());

  word offset_from_thread = 0;
  bool ok = target::CanLoadFromThread(NullObject(), &offset_from_thread);
  ASSERT(ok);
  __ LoadFromOffset(kWord, R2, THR, offset_from_thread);

  // Exception object.
  __ LoadFromOffset(kWord, R0, THR, target::Thread::active_exception_offset());
  __ StoreToOffset(kWord, R2, THR, target::Thread::active_exception_offset());

  // StackTrace object.
  __ LoadFromOffset(kWord, R1, THR, target::Thread::active_stacktrace_offset());
  __ StoreToOffset(kWord, R2, THR, target::Thread::active_stacktrace_offset());

  __ bx(LR);  // Jump to the exception handler code.
}

// Deoptimize a frame on the call stack before rewinding.
// The arguments are stored in the Thread object.
// No result.
void StubCodeCompiler::GenerateDeoptForRewindStub(Assembler* assembler) {
  // Push zap value instead of CODE_REG.
  __ LoadImmediate(IP, kZapCodeReg);
  __ Push(IP);

  // Load the deopt pc into LR.
  __ LoadFromOffset(kWord, LR, THR, target::Thread::resume_pc_offset());
  GenerateDeoptimizationSequence(assembler, kEagerDeopt);

  // After we have deoptimized, jump to the correct frame.
  __ EnterStubFrame();
  __ CallRuntime(kRewindPostDeoptRuntimeEntry, 0);
  __ LeaveStubFrame();
  __ bkpt(0);
}

// Calls to the runtime to optimize the given function.
// R8: function to be reoptimized.
// R4: argument descriptor (preserved).
void StubCodeCompiler::GenerateOptimizeFunctionStub(Assembler* assembler) {
  __ EnterStubFrame();
  __ Push(R4);
  __ LoadImmediate(IP, 0);
  __ Push(IP);  // Setup space on stack for return value.
  __ Push(R8);
  __ CallRuntime(kOptimizeInvokedFunctionRuntimeEntry, 1);
  __ Pop(R0);  // Discard argument.
  __ Pop(R0);  // Get Function object
  __ Pop(R4);  // Restore argument descriptor.
  __ LeaveStubFrame();
  __ ldr(CODE_REG, FieldAddress(R0, target::Function::code_offset()));
  __ Branch(FieldAddress(R0, target::Function::entry_point_offset()));
  __ bkpt(0);
}

// Does identical check (object references are equal or not equal) with special
// checks for boxed numbers.
// LR: return address.
// Return Zero condition flag set if equal.
// Note: A Mint cannot contain a value that would fit in Smi.
static void GenerateIdenticalWithNumberCheckStub(Assembler* assembler,
                                                 const Register left,
                                                 const Register right,
                                                 const Register temp) {
  Label reference_compare, done, check_mint;
  // If any of the arguments is Smi do reference compare.
  __ tst(left, Operand(kSmiTagMask));
  __ b(&reference_compare, EQ);
  __ tst(right, Operand(kSmiTagMask));
  __ b(&reference_compare, EQ);

  // Value compare for two doubles.
  __ CompareClassId(left, kDoubleCid, temp);
  __ b(&check_mint, NE);
  __ CompareClassId(right, kDoubleCid, temp);
  __ b(&done, NE);

  // Double values bitwise compare.
  __ ldr(temp, FieldAddress(left, target::Double::value_offset() +
                                      0 * target::kWordSize));
  __ ldr(IP, FieldAddress(right, target::Double::value_offset() +
                                     0 * target::kWordSize));
  __ cmp(temp, Operand(IP));
  __ b(&done, NE);
  __ ldr(temp, FieldAddress(left, target::Double::value_offset() +
                                      1 * target::kWordSize));
  __ ldr(IP, FieldAddress(right, target::Double::value_offset() +
                                     1 * target::kWordSize));
  __ cmp(temp, Operand(IP));
  __ b(&done);

  __ Bind(&check_mint);
  __ CompareClassId(left, kMintCid, temp);
  __ b(&reference_compare, NE);
  __ CompareClassId(right, kMintCid, temp);
  __ b(&done, NE);
  __ ldr(temp, FieldAddress(
                   left, target::Mint::value_offset() + 0 * target::kWordSize));
  __ ldr(IP, FieldAddress(
                 right, target::Mint::value_offset() + 0 * target::kWordSize));
  __ cmp(temp, Operand(IP));
  __ b(&done, NE);
  __ ldr(temp, FieldAddress(
                   left, target::Mint::value_offset() + 1 * target::kWordSize));
  __ ldr(IP, FieldAddress(
                 right, target::Mint::value_offset() + 1 * target::kWordSize));
  __ cmp(temp, Operand(IP));
  __ b(&done);

  __ Bind(&reference_compare);
  __ cmp(left, Operand(right));
  __ Bind(&done);
}

// Called only from unoptimized code. All relevant registers have been saved.
// LR: return address.
// SP + 4: left operand.
// SP + 0: right operand.
// Return Zero condition flag set if equal.
void StubCodeCompiler::GenerateUnoptimizedIdenticalWithNumberCheckStub(
    Assembler* assembler) {
#if !defined(PRODUCT)
  // Check single stepping.
  Label stepping, done_stepping;
  __ LoadIsolate(R1);
  __ ldrb(R1, Address(R1, target::Isolate::single_step_offset()));
  __ CompareImmediate(R1, 0);
  __ b(&stepping, NE);
  __ Bind(&done_stepping);
#endif

  const Register temp = R2;
  const Register left = R1;
  const Register right = R0;
  __ ldr(left, Address(SP, 1 * target::kWordSize));
  __ ldr(right, Address(SP, 0 * target::kWordSize));
  GenerateIdenticalWithNumberCheckStub(assembler, left, right, temp);
  __ Ret();

#if !defined(PRODUCT)
  __ Bind(&stepping);
  __ EnterStubFrame();
  __ CallRuntime(kSingleStepHandlerRuntimeEntry, 0);
  __ RestoreCodePointer();
  __ LeaveStubFrame();
  __ b(&done_stepping);
#endif
}

// Called from optimized code only.
// LR: return address.
// SP + 4: left operand.
// SP + 0: right operand.
// Return Zero condition flag set if equal.
void StubCodeCompiler::GenerateOptimizedIdenticalWithNumberCheckStub(
    Assembler* assembler) {
  const Register temp = R2;
  const Register left = R1;
  const Register right = R0;
  __ ldr(left, Address(SP, 1 * target::kWordSize));
  __ ldr(right, Address(SP, 0 * target::kWordSize));
  GenerateIdenticalWithNumberCheckStub(assembler, left, right, temp);
  __ Ret();
}

// Called from megamorphic calls.
//  R0: receiver
//  R9: MegamorphicCache (preserved)
// Passed to target:
//  CODE_REG: target Code
//  R4: arguments descriptor
void StubCodeCompiler::GenerateMegamorphicCallStub(Assembler* assembler) {
  __ LoadTaggedClassIdMayBeSmi(R0, R0);
  // R0: receiver cid as Smi.
  __ ldr(R2, FieldAddress(R9, target::MegamorphicCache::buckets_offset()));
  __ ldr(R1, FieldAddress(R9, target::MegamorphicCache::mask_offset()));
  // R2: cache buckets array.
  // R1: mask as a smi.

  // Compute the table index.
  ASSERT(target::MegamorphicCache::kSpreadFactor == 7);
  // Use reverse substract to multiply with 7 == 8 - 1.
  __ rsb(R3, R0, Operand(R0, LSL, 3));
  // R3: probe.
  Label loop;
  __ Bind(&loop);
  __ and_(R3, R3, Operand(R1));

  const intptr_t base = target::Array::data_offset();
  // R3 is smi tagged, but table entries are two words, so LSL 2.
  Label probe_failed;
  __ add(IP, R2, Operand(R3, LSL, 2));
  __ ldr(R6, FieldAddress(IP, base));
  __ cmp(R6, Operand(R0));
  __ b(&probe_failed, NE);

  Label load_target;
  __ Bind(&load_target);
  // Call the target found in the cache.  For a class id match, this is a
  // proper target for the given name and arguments descriptor.  If the
  // illegal class id was found, the target is a cache miss handler that can
  // be invoked as a normal Dart function.
  const auto target_address = FieldAddress(IP, base + target::kWordSize);
  if (FLAG_precompiled_mode && FLAG_use_bare_instructions) {
    __ ldr(ARGS_DESC_REG,
           FieldAddress(
               R9, target::MegamorphicCache::arguments_descriptor_offset()));
    __ Branch(target_address);
  } else {
    __ ldr(R0, target_address);
    __ ldr(CODE_REG, FieldAddress(R0, target::Function::code_offset()));
    __ ldr(ARGS_DESC_REG,
           FieldAddress(
               R9, target::MegamorphicCache::arguments_descriptor_offset()));
    __ Branch(FieldAddress(R0, target::Function::entry_point_offset()));
  }

  // Probe failed, check if it is a miss.
  __ Bind(&probe_failed);
  ASSERT(kIllegalCid == 0);
  __ tst(R6, Operand(R6));
  __ b(&load_target, EQ);  // branch if miss.

  // Try next entry in the table.
  __ AddImmediate(R3, target::ToRawSmi(1));
  __ b(&loop);
}

// Called from switchable IC calls.
//  R0: receiver
//  R9: ICData (preserved)
// Passed to target:
//  CODE_REG: target Code object
//  R4: arguments descriptor
void StubCodeCompiler::GenerateICCallThroughFunctionStub(Assembler* assembler) {
  Label loop, found, miss;
  __ ldr(ARGS_DESC_REG,
         FieldAddress(R9, target::ICData::arguments_descriptor_offset()));
  __ ldr(R8, FieldAddress(R9, target::ICData::entries_offset()));
  __ AddImmediate(R8, target::Array::data_offset() - kHeapObjectTag);
  // R8: first IC entry
  __ LoadTaggedClassIdMayBeSmi(R1, R0);
  // R1: receiver cid as Smi

  __ Bind(&loop);
  __ ldr(R2, Address(R8, 0));
  __ cmp(R1, Operand(R2));
  __ b(&found, EQ);
  __ CompareImmediate(R2, target::ToRawSmi(kIllegalCid));
  __ b(&miss, EQ);

  const intptr_t entry_length =
      target::ICData::TestEntryLengthFor(1, /*tracking_exactness=*/false) *
      target::kWordSize;
  __ AddImmediate(R8, entry_length);  // Next entry.
  __ b(&loop);

  __ Bind(&found);
  const intptr_t target_offset =
      target::ICData::TargetIndexFor(1) * target::kWordSize;
  __ LoadFromOffset(kWord, R0, R8, target_offset);
  __ ldr(CODE_REG, FieldAddress(R0, target::Function::code_offset()));
  __ Branch(FieldAddress(R0, target::Function::entry_point_offset()));

  __ Bind(&miss);
  __ LoadIsolate(R2);
  __ ldr(CODE_REG, Address(R2, target::Isolate::ic_miss_code_offset()));
  __ Branch(FieldAddress(CODE_REG, target::Code::entry_point_offset()));
}

void StubCodeCompiler::GenerateICCallThroughCodeStub(Assembler* assembler) {
  Label loop, found, miss;
  __ ldr(R8, FieldAddress(R9, target::ICData::entries_offset()));
  __ ldr(R4, FieldAddress(R9, target::ICData::arguments_descriptor_offset()));
  __ AddImmediate(R8, target::Array::data_offset() - kHeapObjectTag);
  // R8: first IC entry
  __ LoadTaggedClassIdMayBeSmi(R1, R0);
  // R1: receiver cid as Smi

  __ Bind(&loop);
  __ ldr(R2, Address(R8, 0));
  __ cmp(R1, Operand(R2));
  __ b(&found, EQ);
  __ CompareImmediate(R2, target::ToRawSmi(kIllegalCid));
  __ b(&miss, EQ);

  const intptr_t entry_length =
      target::ICData::TestEntryLengthFor(1, /*tracking_exactness=*/false) *
      target::kWordSize;
  __ AddImmediate(R8, entry_length);  // Next entry.
  __ b(&loop);

  __ Bind(&found);
  const intptr_t code_offset =
      target::ICData::CodeIndexFor(1) * target::kWordSize;
  const intptr_t entry_offset =
      target::ICData::EntryPointIndexFor(1) * target::kWordSize;
  if (!(FLAG_precompiled_mode && FLAG_use_bare_instructions)) {
    __ ldr(CODE_REG, Address(R8, code_offset));
  }
  __ Branch(Address(R8, entry_offset));

  __ Bind(&miss);
  __ LoadIsolate(R2);
  __ ldr(CODE_REG, Address(R2, target::Isolate::ic_miss_code_offset()));
  __ Branch(FieldAddress(CODE_REG, target::Code::entry_point_offset()));
}

// Called from switchable IC calls.
//  R0: receiver
//  R9: UnlinkedCall
void StubCodeCompiler::GenerateUnlinkedCallStub(Assembler* assembler) {
  __ EnterStubFrame();
  __ Push(R0);  // Preserve receiver.

  __ LoadImmediate(IP, 0);
  __ Push(IP);  // Result slot
  __ Push(R0);  // Arg0: Receiver
  __ Push(R9);  // Arg1: UnlinkedCall
  __ CallRuntime(kUnlinkedCallRuntimeEntry, 2);
  __ Drop(2);
  __ Pop(R9);  // result = IC

  __ Pop(R0);  // Restore receiver.
  __ LeaveStubFrame();

  __ ldr(CODE_REG,
         Address(THR, target::Thread::ic_lookup_through_code_stub_offset()));
  __ Branch(FieldAddress(
      CODE_REG, target::Code::entry_point_offset(CodeEntryKind::kMonomorphic)));
}

// Called from switchable IC calls.
//  R0: receiver
//  R9: SingleTargetCache
// Passed to target:
//  CODE_REG: target Code object
void StubCodeCompiler::GenerateSingleTargetCallStub(Assembler* assembler) {
  Label miss;
  __ LoadClassIdMayBeSmi(R1, R0);
  __ ldrh(R2,
          FieldAddress(R9, target::SingleTargetCache::lower_limit_offset()));
  __ ldrh(R3,
          FieldAddress(R9, target::SingleTargetCache::upper_limit_offset()));

  __ cmp(R1, Operand(R2));
  __ b(&miss, LT);
  __ cmp(R1, Operand(R3));
  __ b(&miss, GT);

  __ ldr(CODE_REG,
         FieldAddress(R9, target::SingleTargetCache::target_offset()));
  __ Branch(FieldAddress(R9, target::SingleTargetCache::entry_point_offset()));

  __ Bind(&miss);
  __ EnterStubFrame();
  __ Push(R0);  // Preserve receiver.

  __ LoadImmediate(IP, 0);
  __ Push(IP);  // Result slot
  __ Push(R0);  // Arg0: Receiver
  __ CallRuntime(kSingleTargetMissRuntimeEntry, 1);
  __ Drop(1);
  __ Pop(R9);  // result = IC

  __ Pop(R0);  // Restore receiver.
  __ LeaveStubFrame();

  __ ldr(CODE_REG,
         Address(THR, target::Thread::ic_lookup_through_code_stub_offset()));
  __ Branch(FieldAddress(
      CODE_REG, target::Code::entry_point_offset(CodeEntryKind::kMonomorphic)));
}

// Called from the monomorphic checked entry.
//  R0: receiver
void StubCodeCompiler::GenerateMonomorphicMissStub(Assembler* assembler) {
  __ ldr(CODE_REG,
         Address(THR, target::Thread::monomorphic_miss_stub_offset()));
  __ EnterStubFrame();
  __ Push(R0);  // Preserve receiver.

  __ LoadImmediate(IP, 0);
  __ Push(IP);  // Result slot
  __ Push(R0);  // Arg0: Receiver
  __ CallRuntime(kMonomorphicMissRuntimeEntry, 1);
  __ Drop(1);
  __ Pop(R9);  // result = IC

  __ Pop(R0);  // Restore receiver.
  __ LeaveStubFrame();

  __ ldr(CODE_REG,
         Address(THR, target::Thread::ic_lookup_through_code_stub_offset()));
  __ Branch(FieldAddress(
      CODE_REG, target::Code::entry_point_offset(CodeEntryKind::kMonomorphic)));
}

void StubCodeCompiler::GenerateFrameAwaitingMaterializationStub(
    Assembler* assembler) {
  __ bkpt(0);
}

void StubCodeCompiler::GenerateAsynchronousGapMarkerStub(Assembler* assembler) {
  __ bkpt(0);
}

}  // namespace compiler

}  // namespace dart

#endif  // defined(TARGET_ARCH_ARM) && !defined(DART_PRECOMPILED_RUNTIME)