blob: 596783b64f17faff82e7da40f127fc1fe745c801 [file] [log] [blame]
// Copyright 2015 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "src/interpreter/interpreter-assembler.h"
#include <limits>
#include <ostream>
#include "src/codegen/code-factory.h"
#include "src/codegen/interface-descriptors.h"
#include "src/codegen/machine-type.h"
#include "src/execution/frames.h"
#include "src/interpreter/bytecodes.h"
#include "src/interpreter/interpreter.h"
#include "src/objects/objects-inl.h"
#include "src/zone/zone.h"
namespace v8 {
namespace internal {
namespace interpreter {
using compiler::CodeAssemblerState;
InterpreterAssembler::InterpreterAssembler(CodeAssemblerState* state,
Bytecode bytecode,
OperandScale operand_scale)
: CodeStubAssembler(state),
bytecode_(bytecode),
operand_scale_(operand_scale),
TVARIABLE_CONSTRUCTOR(interpreted_frame_pointer_),
TVARIABLE_CONSTRUCTOR(bytecode_array_,
Parameter<BytecodeArray>(
InterpreterDispatchDescriptor::kBytecodeArray)),
TVARIABLE_CONSTRUCTOR(
bytecode_offset_,
UncheckedParameter<IntPtrT>(
InterpreterDispatchDescriptor::kBytecodeOffset)),
TVARIABLE_CONSTRUCTOR(dispatch_table_,
UncheckedParameter<ExternalReference>(
InterpreterDispatchDescriptor::kDispatchTable)),
TVARIABLE_CONSTRUCTOR(
accumulator_,
Parameter<Object>(InterpreterDispatchDescriptor::kAccumulator)),
accumulator_use_(AccumulatorUse::kNone),
made_call_(false),
reloaded_frame_ptr_(false),
bytecode_array_valid_(true) {
#ifdef V8_TRACE_IGNITION
TraceBytecode(Runtime::kInterpreterTraceBytecodeEntry);
#endif
RegisterCallGenerationCallbacks([this] { CallPrologue(); },
[this] { CallEpilogue(); });
// Save the bytecode offset immediately if bytecode will make a call along
// the critical path, or it is a return bytecode.
if (Bytecodes::MakesCallAlongCriticalPath(bytecode) ||
Bytecodes::Returns(bytecode)) {
SaveBytecodeOffset();
}
}
InterpreterAssembler::~InterpreterAssembler() {
// If the following check fails the handler does not use the
// accumulator in the way described in the bytecode definitions in
// bytecodes.h.
DCHECK_EQ(accumulator_use_, Bytecodes::GetAccumulatorUse(bytecode_));
UnregisterCallGenerationCallbacks();
}
TNode<RawPtrT> InterpreterAssembler::GetInterpretedFramePointer() {
if (!interpreted_frame_pointer_.IsBound()) {
interpreted_frame_pointer_ = LoadParentFramePointer();
} else if (Bytecodes::MakesCallAlongCriticalPath(bytecode_) && made_call_ &&
!reloaded_frame_ptr_) {
interpreted_frame_pointer_ = LoadParentFramePointer();
reloaded_frame_ptr_ = true;
}
return interpreted_frame_pointer_.value();
}
TNode<IntPtrT> InterpreterAssembler::BytecodeOffset() {
if (Bytecodes::MakesCallAlongCriticalPath(bytecode_) && made_call_ &&
(bytecode_offset_.value() ==
UncheckedParameter<IntPtrT>(
InterpreterDispatchDescriptor::kBytecodeOffset))) {
bytecode_offset_ = ReloadBytecodeOffset();
}
return bytecode_offset_.value();
}
TNode<IntPtrT> InterpreterAssembler::ReloadBytecodeOffset() {
TNode<IntPtrT> offset = LoadAndUntagRegister(Register::bytecode_offset());
if (operand_scale() != OperandScale::kSingle) {
// Add one to the offset such that it points to the actual bytecode rather
// than the Wide / ExtraWide prefix bytecode.
offset = IntPtrAdd(offset, IntPtrConstant(1));
}
return offset;
}
void InterpreterAssembler::SaveBytecodeOffset() {
TNode<IntPtrT> bytecode_offset = BytecodeOffset();
if (operand_scale() != OperandScale::kSingle) {
// Subtract one from the bytecode_offset such that it points to the Wide /
// ExtraWide prefix bytecode.
bytecode_offset = IntPtrSub(BytecodeOffset(), IntPtrConstant(1));
}
int store_offset =
Register::bytecode_offset().ToOperand() * kSystemPointerSize;
TNode<RawPtrT> base = GetInterpretedFramePointer();
if (SmiValuesAre32Bits()) {
int zero_offset = store_offset + 4;
int payload_offset = store_offset;
#if V8_TARGET_LITTLE_ENDIAN
std::swap(zero_offset, payload_offset);
#endif
StoreNoWriteBarrier(MachineRepresentation::kWord32, base,
IntPtrConstant(zero_offset), Int32Constant(0));
StoreNoWriteBarrier(MachineRepresentation::kWord32, base,
IntPtrConstant(payload_offset),
TruncateIntPtrToInt32(bytecode_offset));
} else {
StoreFullTaggedNoWriteBarrier(base, IntPtrConstant(store_offset),
SmiTag(bytecode_offset));
}
}
TNode<BytecodeArray> InterpreterAssembler::BytecodeArrayTaggedPointer() {
// Force a re-load of the bytecode array after every call in case the debugger
// has been activated.
if (!bytecode_array_valid_) {
bytecode_array_ = CAST(LoadRegister(Register::bytecode_array()));
bytecode_array_valid_ = true;
}
return bytecode_array_.value();
}
TNode<ExternalReference> InterpreterAssembler::DispatchTablePointer() {
if (Bytecodes::MakesCallAlongCriticalPath(bytecode_) && made_call_ &&
(dispatch_table_.value() ==
UncheckedParameter<ExternalReference>(
InterpreterDispatchDescriptor::kDispatchTable))) {
dispatch_table_ = ExternalConstant(
ExternalReference::interpreter_dispatch_table_address(isolate()));
}
return dispatch_table_.value();
}
TNode<Object> InterpreterAssembler::GetAccumulatorUnchecked() {
return accumulator_.value();
}
TNode<Object> InterpreterAssembler::GetAccumulator() {
DCHECK(Bytecodes::ReadsAccumulator(bytecode_));
accumulator_use_ = accumulator_use_ | AccumulatorUse::kRead;
return TaggedPoisonOnSpeculation(GetAccumulatorUnchecked());
}
void InterpreterAssembler::SetAccumulator(TNode<Object> value) {
DCHECK(Bytecodes::WritesAccumulator(bytecode_));
accumulator_use_ = accumulator_use_ | AccumulatorUse::kWrite;
accumulator_ = value;
}
TNode<Context> InterpreterAssembler::GetContext() {
return CAST(LoadRegister(Register::current_context()));
}
void InterpreterAssembler::SetContext(TNode<Context> value) {
StoreRegister(value, Register::current_context());
}
TNode<Context> InterpreterAssembler::GetContextAtDepth(TNode<Context> context,
TNode<Uint32T> depth) {
TVARIABLE(Context, cur_context, context);
TVARIABLE(Uint32T, cur_depth, depth);
Label context_found(this);
Label context_search(this, {&cur_depth, &cur_context});
// Fast path if the depth is 0.
Branch(Word32Equal(depth, Int32Constant(0)), &context_found, &context_search);
// Loop until the depth is 0.
BIND(&context_search);
{
cur_depth = Unsigned(Int32Sub(cur_depth.value(), Int32Constant(1)));
cur_context =
CAST(LoadContextElement(cur_context.value(), Context::PREVIOUS_INDEX));
Branch(Word32Equal(cur_depth.value(), Int32Constant(0)), &context_found,
&context_search);
}
BIND(&context_found);
return cur_context.value();
}
void InterpreterAssembler::GotoIfHasContextExtensionUpToDepth(
TNode<Context> context, TNode<Uint32T> depth, Label* target) {
TVARIABLE(Context, cur_context, context);
TVARIABLE(Uint32T, cur_depth, depth);
Label context_search(this, {&cur_depth, &cur_context});
Label no_extension(this);
// Loop until the depth is 0.
Goto(&context_search);
BIND(&context_search);
{
// Check if context has an extension slot.
TNode<BoolT> has_extension =
LoadScopeInfoHasExtensionField(LoadScopeInfo(cur_context.value()));
GotoIfNot(has_extension, &no_extension);
// Jump to the target if the extension slot is not an undefined value.
TNode<Object> extension_slot =
LoadContextElement(cur_context.value(), Context::EXTENSION_INDEX);
Branch(TaggedNotEqual(extension_slot, UndefinedConstant()), target,
&no_extension);
BIND(&no_extension);
{
cur_depth = Unsigned(Int32Sub(cur_depth.value(), Int32Constant(1)));
cur_context = CAST(
LoadContextElement(cur_context.value(), Context::PREVIOUS_INDEX));
GotoIf(Word32NotEqual(cur_depth.value(), Int32Constant(0)),
&context_search);
}
}
}
TNode<IntPtrT> InterpreterAssembler::RegisterLocation(
TNode<IntPtrT> reg_index) {
return Signed(WordPoisonOnSpeculation(
IntPtrAdd(GetInterpretedFramePointer(), RegisterFrameOffset(reg_index))));
}
TNode<IntPtrT> InterpreterAssembler::RegisterLocation(Register reg) {
return RegisterLocation(IntPtrConstant(reg.ToOperand()));
}
TNode<IntPtrT> InterpreterAssembler::RegisterFrameOffset(TNode<IntPtrT> index) {
return TimesSystemPointerSize(index);
}
TNode<Object> InterpreterAssembler::LoadRegister(TNode<IntPtrT> reg_index) {
return LoadFullTagged(GetInterpretedFramePointer(),
RegisterFrameOffset(reg_index),
LoadSensitivity::kCritical);
}
TNode<Object> InterpreterAssembler::LoadRegister(Register reg) {
return LoadFullTagged(GetInterpretedFramePointer(),
IntPtrConstant(reg.ToOperand() * kSystemPointerSize));
}
TNode<IntPtrT> InterpreterAssembler::LoadAndUntagRegister(Register reg) {
TNode<RawPtrT> base = GetInterpretedFramePointer();
int index = reg.ToOperand() * kSystemPointerSize;
if (SmiValuesAre32Bits()) {
#if V8_TARGET_LITTLE_ENDIAN
index += 4;
#endif
return ChangeInt32ToIntPtr(Load<Int32T>(base, IntPtrConstant(index)));
} else {
return SmiToIntPtr(CAST(LoadFullTagged(base, IntPtrConstant(index))));
}
}
TNode<Object> InterpreterAssembler::LoadRegisterAtOperandIndex(
int operand_index) {
return LoadRegister(
BytecodeOperandReg(operand_index, LoadSensitivity::kSafe));
}
std::pair<TNode<Object>, TNode<Object>>
InterpreterAssembler::LoadRegisterPairAtOperandIndex(int operand_index) {
DCHECK_EQ(OperandType::kRegPair,
Bytecodes::GetOperandType(bytecode_, operand_index));
TNode<IntPtrT> first_reg_index =
BytecodeOperandReg(operand_index, LoadSensitivity::kSafe);
TNode<IntPtrT> second_reg_index = NextRegister(first_reg_index);
return std::make_pair(LoadRegister(first_reg_index),
LoadRegister(second_reg_index));
}
InterpreterAssembler::RegListNodePair
InterpreterAssembler::GetRegisterListAtOperandIndex(int operand_index) {
DCHECK(Bytecodes::IsRegisterListOperandType(
Bytecodes::GetOperandType(bytecode_, operand_index)));
DCHECK_EQ(OperandType::kRegCount,
Bytecodes::GetOperandType(bytecode_, operand_index + 1));
TNode<IntPtrT> base_reg = RegisterLocation(
BytecodeOperandReg(operand_index, LoadSensitivity::kSafe));
TNode<Uint32T> reg_count = BytecodeOperandCount(operand_index + 1);
return RegListNodePair(base_reg, reg_count);
}
TNode<Object> InterpreterAssembler::LoadRegisterFromRegisterList(
const RegListNodePair& reg_list, int index) {
TNode<IntPtrT> location = RegisterLocationInRegisterList(reg_list, index);
// Location is already poisoned on speculation, so no need to poison here.
return LoadFullTagged(location);
}
TNode<IntPtrT> InterpreterAssembler::RegisterLocationInRegisterList(
const RegListNodePair& reg_list, int index) {
CSA_ASSERT(this,
Uint32GreaterThan(reg_list.reg_count(), Int32Constant(index)));
TNode<IntPtrT> offset = RegisterFrameOffset(IntPtrConstant(index));
// Register indexes are negative, so subtract index from base location to get
// location.
return Signed(IntPtrSub(reg_list.base_reg_location(), offset));
}
void InterpreterAssembler::StoreRegister(TNode<Object> value, Register reg) {
StoreFullTaggedNoWriteBarrier(
GetInterpretedFramePointer(),
IntPtrConstant(reg.ToOperand() * kSystemPointerSize), value);
}
void InterpreterAssembler::StoreRegister(TNode<Object> value,
TNode<IntPtrT> reg_index) {
StoreFullTaggedNoWriteBarrier(GetInterpretedFramePointer(),
RegisterFrameOffset(reg_index), value);
}
void InterpreterAssembler::StoreRegisterAtOperandIndex(TNode<Object> value,
int operand_index) {
StoreRegister(value,
BytecodeOperandReg(operand_index, LoadSensitivity::kSafe));
}
void InterpreterAssembler::StoreRegisterPairAtOperandIndex(TNode<Object> value1,
TNode<Object> value2,
int operand_index) {
DCHECK_EQ(OperandType::kRegOutPair,
Bytecodes::GetOperandType(bytecode_, operand_index));
TNode<IntPtrT> first_reg_index =
BytecodeOperandReg(operand_index, LoadSensitivity::kSafe);
StoreRegister(value1, first_reg_index);
TNode<IntPtrT> second_reg_index = NextRegister(first_reg_index);
StoreRegister(value2, second_reg_index);
}
void InterpreterAssembler::StoreRegisterTripleAtOperandIndex(
TNode<Object> value1, TNode<Object> value2, TNode<Object> value3,
int operand_index) {
DCHECK_EQ(OperandType::kRegOutTriple,
Bytecodes::GetOperandType(bytecode_, operand_index));
TNode<IntPtrT> first_reg_index =
BytecodeOperandReg(operand_index, LoadSensitivity::kSafe);
StoreRegister(value1, first_reg_index);
TNode<IntPtrT> second_reg_index = NextRegister(first_reg_index);
StoreRegister(value2, second_reg_index);
TNode<IntPtrT> third_reg_index = NextRegister(second_reg_index);
StoreRegister(value3, third_reg_index);
}
TNode<IntPtrT> InterpreterAssembler::NextRegister(TNode<IntPtrT> reg_index) {
// Register indexes are negative, so the next index is minus one.
return Signed(IntPtrAdd(reg_index, IntPtrConstant(-1)));
}
TNode<IntPtrT> InterpreterAssembler::OperandOffset(int operand_index) {
return IntPtrConstant(
Bytecodes::GetOperandOffset(bytecode_, operand_index, operand_scale()));
}
TNode<Uint8T> InterpreterAssembler::BytecodeOperandUnsignedByte(
int operand_index, LoadSensitivity needs_poisoning) {
DCHECK_LT(operand_index, Bytecodes::NumberOfOperands(bytecode_));
DCHECK_EQ(OperandSize::kByte, Bytecodes::GetOperandSize(
bytecode_, operand_index, operand_scale()));
TNode<IntPtrT> operand_offset = OperandOffset(operand_index);
return Load<Uint8T>(BytecodeArrayTaggedPointer(),
IntPtrAdd(BytecodeOffset(), operand_offset),
needs_poisoning);
}
TNode<Int8T> InterpreterAssembler::BytecodeOperandSignedByte(
int operand_index, LoadSensitivity needs_poisoning) {
DCHECK_LT(operand_index, Bytecodes::NumberOfOperands(bytecode_));
DCHECK_EQ(OperandSize::kByte, Bytecodes::GetOperandSize(
bytecode_, operand_index, operand_scale()));
TNode<IntPtrT> operand_offset = OperandOffset(operand_index);
return Load<Int8T>(BytecodeArrayTaggedPointer(),
IntPtrAdd(BytecodeOffset(), operand_offset),
needs_poisoning);
}
TNode<Word32T> InterpreterAssembler::BytecodeOperandReadUnaligned(
int relative_offset, MachineType result_type,
LoadSensitivity needs_poisoning) {
static const int kMaxCount = 4;
DCHECK(!TargetSupportsUnalignedAccess());
int count;
switch (result_type.representation()) {
case MachineRepresentation::kWord16:
count = 2;
break;
case MachineRepresentation::kWord32:
count = 4;
break;
default:
UNREACHABLE();
}
MachineType msb_type =
result_type.IsSigned() ? MachineType::Int8() : MachineType::Uint8();
#if V8_TARGET_LITTLE_ENDIAN
const int kStep = -1;
int msb_offset = count - 1;
#elif V8_TARGET_BIG_ENDIAN
const int kStep = 1;
int msb_offset = 0;
#else
#error "Unknown Architecture"
#endif
// Read the most signicant bytecode into bytes[0] and then in order
// down to least significant in bytes[count - 1].
DCHECK_LE(count, kMaxCount);
TNode<Word32T> bytes[kMaxCount];
for (int i = 0; i < count; i++) {
MachineType machine_type = (i == 0) ? msb_type : MachineType::Uint8();
TNode<IntPtrT> offset =
IntPtrConstant(relative_offset + msb_offset + i * kStep);
TNode<IntPtrT> array_offset = IntPtrAdd(BytecodeOffset(), offset);
bytes[i] =
UncheckedCast<Word32T>(Load(machine_type, BytecodeArrayTaggedPointer(),
array_offset, needs_poisoning));
}
// Pack LSB to MSB.
TNode<Word32T> result = bytes[--count];
for (int i = 1; --count >= 0; i++) {
TNode<Int32T> shift = Int32Constant(i * kBitsPerByte);
TNode<Word32T> value = Word32Shl(bytes[count], shift);
result = Word32Or(value, result);
}
return result;
}
TNode<Uint16T> InterpreterAssembler::BytecodeOperandUnsignedShort(
int operand_index, LoadSensitivity needs_poisoning) {
DCHECK_LT(operand_index, Bytecodes::NumberOfOperands(bytecode_));
DCHECK_EQ(
OperandSize::kShort,
Bytecodes::GetOperandSize(bytecode_, operand_index, operand_scale()));
int operand_offset =
Bytecodes::GetOperandOffset(bytecode_, operand_index, operand_scale());
if (TargetSupportsUnalignedAccess()) {
return Load<Uint16T>(
BytecodeArrayTaggedPointer(),
IntPtrAdd(BytecodeOffset(), IntPtrConstant(operand_offset)),
needs_poisoning);
} else {
return UncheckedCast<Uint16T>(BytecodeOperandReadUnaligned(
operand_offset, MachineType::Uint16(), needs_poisoning));
}
}
TNode<Int16T> InterpreterAssembler::BytecodeOperandSignedShort(
int operand_index, LoadSensitivity needs_poisoning) {
DCHECK_LT(operand_index, Bytecodes::NumberOfOperands(bytecode_));
DCHECK_EQ(
OperandSize::kShort,
Bytecodes::GetOperandSize(bytecode_, operand_index, operand_scale()));
int operand_offset =
Bytecodes::GetOperandOffset(bytecode_, operand_index, operand_scale());
if (TargetSupportsUnalignedAccess()) {
return Load<Int16T>(
BytecodeArrayTaggedPointer(),
IntPtrAdd(BytecodeOffset(), IntPtrConstant(operand_offset)),
needs_poisoning);
} else {
return UncheckedCast<Int16T>(BytecodeOperandReadUnaligned(
operand_offset, MachineType::Int16(), needs_poisoning));
}
}
TNode<Uint32T> InterpreterAssembler::BytecodeOperandUnsignedQuad(
int operand_index, LoadSensitivity needs_poisoning) {
DCHECK_LT(operand_index, Bytecodes::NumberOfOperands(bytecode_));
DCHECK_EQ(OperandSize::kQuad, Bytecodes::GetOperandSize(
bytecode_, operand_index, operand_scale()));
int operand_offset =
Bytecodes::GetOperandOffset(bytecode_, operand_index, operand_scale());
if (TargetSupportsUnalignedAccess()) {
return Load<Uint32T>(
BytecodeArrayTaggedPointer(),
IntPtrAdd(BytecodeOffset(), IntPtrConstant(operand_offset)),
needs_poisoning);
} else {
return UncheckedCast<Uint32T>(BytecodeOperandReadUnaligned(
operand_offset, MachineType::Uint32(), needs_poisoning));
}
}
TNode<Int32T> InterpreterAssembler::BytecodeOperandSignedQuad(
int operand_index, LoadSensitivity needs_poisoning) {
DCHECK_LT(operand_index, Bytecodes::NumberOfOperands(bytecode_));
DCHECK_EQ(OperandSize::kQuad, Bytecodes::GetOperandSize(
bytecode_, operand_index, operand_scale()));
int operand_offset =
Bytecodes::GetOperandOffset(bytecode_, operand_index, operand_scale());
if (TargetSupportsUnalignedAccess()) {
return Load<Int32T>(
BytecodeArrayTaggedPointer(),
IntPtrAdd(BytecodeOffset(), IntPtrConstant(operand_offset)),
needs_poisoning);
} else {
return UncheckedCast<Int32T>(BytecodeOperandReadUnaligned(
operand_offset, MachineType::Int32(), needs_poisoning));
}
}
TNode<Int32T> InterpreterAssembler::BytecodeSignedOperand(
int operand_index, OperandSize operand_size,
LoadSensitivity needs_poisoning) {
DCHECK(!Bytecodes::IsUnsignedOperandType(
Bytecodes::GetOperandType(bytecode_, operand_index)));
switch (operand_size) {
case OperandSize::kByte:
return BytecodeOperandSignedByte(operand_index, needs_poisoning);
case OperandSize::kShort:
return BytecodeOperandSignedShort(operand_index, needs_poisoning);
case OperandSize::kQuad:
return BytecodeOperandSignedQuad(operand_index, needs_poisoning);
case OperandSize::kNone:
UNREACHABLE();
}
}
TNode<Uint32T> InterpreterAssembler::BytecodeUnsignedOperand(
int operand_index, OperandSize operand_size,
LoadSensitivity needs_poisoning) {
DCHECK(Bytecodes::IsUnsignedOperandType(
Bytecodes::GetOperandType(bytecode_, operand_index)));
switch (operand_size) {
case OperandSize::kByte:
return BytecodeOperandUnsignedByte(operand_index, needs_poisoning);
case OperandSize::kShort:
return BytecodeOperandUnsignedShort(operand_index, needs_poisoning);
case OperandSize::kQuad:
return BytecodeOperandUnsignedQuad(operand_index, needs_poisoning);
case OperandSize::kNone:
UNREACHABLE();
}
}
TNode<Uint32T> InterpreterAssembler::BytecodeOperandCount(int operand_index) {
DCHECK_EQ(OperandType::kRegCount,
Bytecodes::GetOperandType(bytecode_, operand_index));
OperandSize operand_size =
Bytecodes::GetOperandSize(bytecode_, operand_index, operand_scale());
return BytecodeUnsignedOperand(operand_index, operand_size);
}
TNode<Uint32T> InterpreterAssembler::BytecodeOperandFlag(int operand_index) {
DCHECK_EQ(OperandType::kFlag8,
Bytecodes::GetOperandType(bytecode_, operand_index));
OperandSize operand_size =
Bytecodes::GetOperandSize(bytecode_, operand_index, operand_scale());
DCHECK_EQ(operand_size, OperandSize::kByte);
return BytecodeUnsignedOperand(operand_index, operand_size);
}
TNode<Uint32T> InterpreterAssembler::BytecodeOperandUImm(int operand_index) {
DCHECK_EQ(OperandType::kUImm,
Bytecodes::GetOperandType(bytecode_, operand_index));
OperandSize operand_size =
Bytecodes::GetOperandSize(bytecode_, operand_index, operand_scale());
return BytecodeUnsignedOperand(operand_index, operand_size);
}
TNode<UintPtrT> InterpreterAssembler::BytecodeOperandUImmWord(
int operand_index) {
return ChangeUint32ToWord(BytecodeOperandUImm(operand_index));
}
TNode<Smi> InterpreterAssembler::BytecodeOperandUImmSmi(int operand_index) {
return SmiFromUint32(BytecodeOperandUImm(operand_index));
}
TNode<Int32T> InterpreterAssembler::BytecodeOperandImm(int operand_index) {
DCHECK_EQ(OperandType::kImm,
Bytecodes::GetOperandType(bytecode_, operand_index));
OperandSize operand_size =
Bytecodes::GetOperandSize(bytecode_, operand_index, operand_scale());
return BytecodeSignedOperand(operand_index, operand_size);
}
TNode<IntPtrT> InterpreterAssembler::BytecodeOperandImmIntPtr(
int operand_index) {
return ChangeInt32ToIntPtr(BytecodeOperandImm(operand_index));
}
TNode<Smi> InterpreterAssembler::BytecodeOperandImmSmi(int operand_index) {
return SmiFromInt32(BytecodeOperandImm(operand_index));
}
TNode<Uint32T> InterpreterAssembler::BytecodeOperandIdxInt32(
int operand_index) {
DCHECK_EQ(OperandType::kIdx,
Bytecodes::GetOperandType(bytecode_, operand_index));
OperandSize operand_size =
Bytecodes::GetOperandSize(bytecode_, operand_index, operand_scale());
return BytecodeUnsignedOperand(operand_index, operand_size);
}
TNode<UintPtrT> InterpreterAssembler::BytecodeOperandIdx(int operand_index) {
return ChangeUint32ToWord(BytecodeOperandIdxInt32(operand_index));
}
TNode<Smi> InterpreterAssembler::BytecodeOperandIdxSmi(int operand_index) {
return SmiTag(Signed(BytecodeOperandIdx(operand_index)));
}
TNode<TaggedIndex> InterpreterAssembler::BytecodeOperandIdxTaggedIndex(
int operand_index) {
TNode<IntPtrT> index =
ChangeInt32ToIntPtr(Signed(BytecodeOperandIdxInt32(operand_index)));
return IntPtrToTaggedIndex(index);
}
TNode<UintPtrT> InterpreterAssembler::BytecodeOperandConstantPoolIdx(
int operand_index, LoadSensitivity needs_poisoning) {
DCHECK_EQ(OperandType::kIdx,
Bytecodes::GetOperandType(bytecode_, operand_index));
OperandSize operand_size =
Bytecodes::GetOperandSize(bytecode_, operand_index, operand_scale());
return ChangeUint32ToWord(
BytecodeUnsignedOperand(operand_index, operand_size, needs_poisoning));
}
TNode<IntPtrT> InterpreterAssembler::BytecodeOperandReg(
int operand_index, LoadSensitivity needs_poisoning) {
DCHECK(Bytecodes::IsRegisterOperandType(
Bytecodes::GetOperandType(bytecode_, operand_index)));
OperandSize operand_size =
Bytecodes::GetOperandSize(bytecode_, operand_index, operand_scale());
return ChangeInt32ToIntPtr(
BytecodeSignedOperand(operand_index, operand_size, needs_poisoning));
}
TNode<Uint32T> InterpreterAssembler::BytecodeOperandRuntimeId(
int operand_index) {
DCHECK_EQ(OperandType::kRuntimeId,
Bytecodes::GetOperandType(bytecode_, operand_index));
OperandSize operand_size =
Bytecodes::GetOperandSize(bytecode_, operand_index, operand_scale());
DCHECK_EQ(operand_size, OperandSize::kShort);
return BytecodeUnsignedOperand(operand_index, operand_size);
}
TNode<UintPtrT> InterpreterAssembler::BytecodeOperandNativeContextIndex(
int operand_index) {
DCHECK_EQ(OperandType::kNativeContextIndex,
Bytecodes::GetOperandType(bytecode_, operand_index));
OperandSize operand_size =
Bytecodes::GetOperandSize(bytecode_, operand_index, operand_scale());
return ChangeUint32ToWord(
BytecodeUnsignedOperand(operand_index, operand_size));
}
TNode<Uint32T> InterpreterAssembler::BytecodeOperandIntrinsicId(
int operand_index) {
DCHECK_EQ(OperandType::kIntrinsicId,
Bytecodes::GetOperandType(bytecode_, operand_index));
OperandSize operand_size =
Bytecodes::GetOperandSize(bytecode_, operand_index, operand_scale());
DCHECK_EQ(operand_size, OperandSize::kByte);
return BytecodeUnsignedOperand(operand_index, operand_size);
}
TNode<Object> InterpreterAssembler::LoadConstantPoolEntry(TNode<WordT> index) {
TNode<FixedArray> constant_pool = CAST(LoadObjectField(
BytecodeArrayTaggedPointer(), BytecodeArray::kConstantPoolOffset));
return UnsafeLoadFixedArrayElement(constant_pool,
UncheckedCast<IntPtrT>(index), 0,
LoadSensitivity::kCritical);
}
TNode<IntPtrT> InterpreterAssembler::LoadAndUntagConstantPoolEntry(
TNode<WordT> index) {
return SmiUntag(CAST(LoadConstantPoolEntry(index)));
}
TNode<Object> InterpreterAssembler::LoadConstantPoolEntryAtOperandIndex(
int operand_index) {
TNode<UintPtrT> index =
BytecodeOperandConstantPoolIdx(operand_index, LoadSensitivity::kSafe);
return LoadConstantPoolEntry(index);
}
TNode<IntPtrT>
InterpreterAssembler::LoadAndUntagConstantPoolEntryAtOperandIndex(
int operand_index) {
return SmiUntag(CAST(LoadConstantPoolEntryAtOperandIndex(operand_index)));
}
TNode<HeapObject> InterpreterAssembler::LoadFeedbackVector() {
TNode<JSFunction> function = CAST(LoadRegister(Register::function_closure()));
return CodeStubAssembler::LoadFeedbackVector(function);
}
void InterpreterAssembler::CallPrologue() {
if (!Bytecodes::MakesCallAlongCriticalPath(bytecode_)) {
// Bytecodes that make a call along the critical path save the bytecode
// offset in the bytecode handler's prologue. For other bytecodes, if
// there are multiple calls in the bytecode handler, you need to spill
// before each of them, unless SaveBytecodeOffset has explicitly been called
// in a path that dominates _all_ of those calls (which we don't track).
SaveBytecodeOffset();
}
bytecode_array_valid_ = false;
made_call_ = true;
}
void InterpreterAssembler::CallEpilogue() {
}
void InterpreterAssembler::CallJSAndDispatch(
TNode<Object> function, TNode<Context> context, const RegListNodePair& args,
ConvertReceiverMode receiver_mode) {
DCHECK(Bytecodes::MakesCallAlongCriticalPath(bytecode_));
DCHECK(Bytecodes::IsCallOrConstruct(bytecode_) ||
bytecode_ == Bytecode::kInvokeIntrinsic);
DCHECK_EQ(Bytecodes::GetReceiverMode(bytecode_), receiver_mode);
TNode<Word32T> args_count;
if (receiver_mode == ConvertReceiverMode::kNullOrUndefined) {
// The receiver is implied, so it is not in the argument list.
args_count = args.reg_count();
} else {
// Subtract the receiver from the argument count.
TNode<Int32T> receiver_count = Int32Constant(1);
args_count = Int32Sub(args.reg_count(), receiver_count);
}
Callable callable = CodeFactory::InterpreterPushArgsThenCall(
isolate(), receiver_mode, InterpreterPushArgsMode::kOther);
TNode<Code> code_target = HeapConstant(callable.code());
TailCallStubThenBytecodeDispatch(callable.descriptor(), code_target, context,
args_count, args.base_reg_location(),
function);
// TailCallStubThenDispatch updates accumulator with result.
accumulator_use_ = accumulator_use_ | AccumulatorUse::kWrite;
}
template <class... TArgs>
void InterpreterAssembler::CallJSAndDispatch(TNode<Object> function,
TNode<Context> context,
TNode<Word32T> arg_count,
ConvertReceiverMode receiver_mode,
TArgs... args) {
DCHECK(Bytecodes::MakesCallAlongCriticalPath(bytecode_));
DCHECK(Bytecodes::IsCallOrConstruct(bytecode_) ||
bytecode_ == Bytecode::kInvokeIntrinsic);
DCHECK_EQ(Bytecodes::GetReceiverMode(bytecode_), receiver_mode);
Callable callable = CodeFactory::Call(isolate());
TNode<Code> code_target = HeapConstant(callable.code());
if (receiver_mode == ConvertReceiverMode::kNullOrUndefined) {
// The first argument parameter (the receiver) is implied to be undefined.
TailCallStubThenBytecodeDispatch(callable.descriptor(), code_target,
context, function, arg_count, args...,
UndefinedConstant());
} else {
TailCallStubThenBytecodeDispatch(callable.descriptor(), code_target,
context, function, arg_count, args...);
}
// TailCallStubThenDispatch updates accumulator with result.
accumulator_use_ = accumulator_use_ | AccumulatorUse::kWrite;
}
// Instantiate CallJSAndDispatch() for argument counts used by interpreter
// generator.
template V8_EXPORT_PRIVATE void InterpreterAssembler::CallJSAndDispatch(
TNode<Object> function, TNode<Context> context, TNode<Word32T> arg_count,
ConvertReceiverMode receiver_mode);
template V8_EXPORT_PRIVATE void InterpreterAssembler::CallJSAndDispatch(
TNode<Object> function, TNode<Context> context, TNode<Word32T> arg_count,
ConvertReceiverMode receiver_mode, TNode<Object>);
template V8_EXPORT_PRIVATE void InterpreterAssembler::CallJSAndDispatch(
TNode<Object> function, TNode<Context> context, TNode<Word32T> arg_count,
ConvertReceiverMode receiver_mode, TNode<Object>, TNode<Object>);
template V8_EXPORT_PRIVATE void InterpreterAssembler::CallJSAndDispatch(
TNode<Object> function, TNode<Context> context, TNode<Word32T> arg_count,
ConvertReceiverMode receiver_mode, TNode<Object>, TNode<Object>,
TNode<Object>);
void InterpreterAssembler::CallJSWithSpreadAndDispatch(
TNode<Object> function, TNode<Context> context, const RegListNodePair& args,
TNode<UintPtrT> slot_id, TNode<HeapObject> maybe_feedback_vector) {
DCHECK(Bytecodes::MakesCallAlongCriticalPath(bytecode_));
DCHECK_EQ(Bytecodes::GetReceiverMode(bytecode_), ConvertReceiverMode::kAny);
CollectCallFeedback(function, context, maybe_feedback_vector, slot_id);
Comment("call using CallWithSpread builtin");
Callable callable = CodeFactory::InterpreterPushArgsThenCall(
isolate(), ConvertReceiverMode::kAny,
InterpreterPushArgsMode::kWithFinalSpread);
TNode<Code> code_target = HeapConstant(callable.code());
TNode<Int32T> receiver_count = Int32Constant(1);
TNode<Word32T> args_count = Int32Sub(args.reg_count(), receiver_count);
TailCallStubThenBytecodeDispatch(callable.descriptor(), code_target, context,
args_count, args.base_reg_location(),
function);
// TailCallStubThenDispatch updates accumulator with result.
accumulator_use_ = accumulator_use_ | AccumulatorUse::kWrite;
}
TNode<Object> InterpreterAssembler::Construct(
TNode<Object> target, TNode<Context> context, TNode<Object> new_target,
const RegListNodePair& args, TNode<UintPtrT> slot_id,
TNode<HeapObject> maybe_feedback_vector) {
DCHECK(Bytecodes::MakesCallAlongCriticalPath(bytecode_));
TVARIABLE(Object, var_result);
TVARIABLE(AllocationSite, var_site);
Label return_result(this), construct_generic(this),
construct_array(this, &var_site);
CollectConstructFeedback(context, target, new_target, maybe_feedback_vector,
slot_id, &construct_generic, &construct_array,
&var_site);
BIND(&construct_generic);
{
// TODO(bmeurer): Remove the generic type_info parameter from the Construct.
Comment("call using Construct builtin");
Callable callable = CodeFactory::InterpreterPushArgsThenConstruct(
isolate(), InterpreterPushArgsMode::kOther);
var_result =
CallStub(callable, context, args.reg_count(), args.base_reg_location(),
target, new_target, UndefinedConstant());
Goto(&return_result);
}
BIND(&construct_array);
{
// TODO(bmeurer): Introduce a dedicated builtin to deal with the Array
// constructor feedback collection inside of Ignition.
Comment("call using ConstructArray builtin");
Callable callable = CodeFactory::InterpreterPushArgsThenConstruct(
isolate(), InterpreterPushArgsMode::kArrayFunction);
var_result =
CallStub(callable, context, args.reg_count(), args.base_reg_location(),
target, new_target, var_site.value());
Goto(&return_result);
}
BIND(&return_result);
return var_result.value();
}
TNode<Object> InterpreterAssembler::ConstructWithSpread(
TNode<Object> target, TNode<Context> context, TNode<Object> new_target,
const RegListNodePair& args, TNode<UintPtrT> slot_id,
TNode<HeapObject> maybe_feedback_vector) {
// TODO(bmeurer): Unify this with the Construct bytecode feedback
// above once we have a way to pass the AllocationSite to the Array
// constructor _and_ spread the last argument at the same time.
DCHECK(Bytecodes::MakesCallAlongCriticalPath(bytecode_));
Label extra_checks(this, Label::kDeferred), construct(this);
GotoIf(IsUndefined(maybe_feedback_vector), &construct);
TNode<FeedbackVector> feedback_vector = CAST(maybe_feedback_vector);
// Increment the call count.
IncrementCallCount(feedback_vector, slot_id);
// Check if we have monomorphic {new_target} feedback already.
TNode<MaybeObject> feedback =
LoadFeedbackVectorSlot(feedback_vector, slot_id);
Branch(IsWeakReferenceToObject(feedback, new_target), &construct,
&extra_checks);
BIND(&extra_checks);
{
Label check_initialized(this), initialize(this), mark_megamorphic(this);
// Check if it is a megamorphic {new_target}.
Comment("check if megamorphic");
TNode<BoolT> is_megamorphic = TaggedEqual(
feedback, HeapConstant(FeedbackVector::MegamorphicSentinel(isolate())));
GotoIf(is_megamorphic, &construct);
Comment("check if weak reference");
GotoIfNot(IsWeakOrCleared(feedback), &check_initialized);
// If the weak reference is cleared, we have a new chance to become
// monomorphic.
Comment("check if weak reference is cleared");
Branch(IsCleared(feedback), &initialize, &mark_megamorphic);
BIND(&check_initialized);
{
// Check if it is uninitialized.
Comment("check if uninitialized");
TNode<BoolT> is_uninitialized =
TaggedEqual(feedback, UninitializedSymbolConstant());
Branch(is_uninitialized, &initialize, &mark_megamorphic);
}
BIND(&initialize);
{
Comment("check if function in same native context");
GotoIf(TaggedIsSmi(new_target), &mark_megamorphic);
// Check if the {new_target} is a JSFunction or JSBoundFunction
// in the current native context.
TVARIABLE(HeapObject, var_current, CAST(new_target));
Label loop(this, &var_current), done_loop(this);
Goto(&loop);
BIND(&loop);
{
Label if_boundfunction(this), if_function(this);
TNode<HeapObject> current = var_current.value();
TNode<Uint16T> current_instance_type = LoadInstanceType(current);
GotoIf(InstanceTypeEqual(current_instance_type, JS_BOUND_FUNCTION_TYPE),
&if_boundfunction);
Branch(InstanceTypeEqual(current_instance_type, JS_FUNCTION_TYPE),
&if_function, &mark_megamorphic);
BIND(&if_function);
{
// Check that the JSFunction {current} is in the current native
// context.
TNode<Context> current_context =
CAST(LoadObjectField(current, JSFunction::kContextOffset));
TNode<NativeContext> current_native_context =
LoadNativeContext(current_context);
Branch(
TaggedEqual(LoadNativeContext(context), current_native_context),
&done_loop, &mark_megamorphic);
}
BIND(&if_boundfunction);
{
// Continue with the [[BoundTargetFunction]] of {current}.
var_current = LoadObjectField<HeapObject>(
current, JSBoundFunction::kBoundTargetFunctionOffset);
Goto(&loop);
}
}
BIND(&done_loop);
StoreWeakReferenceInFeedbackVector(feedback_vector, slot_id,
CAST(new_target));
ReportFeedbackUpdate(feedback_vector, slot_id,
"ConstructWithSpread:Initialize");
Goto(&construct);
}
BIND(&mark_megamorphic);
{
// MegamorphicSentinel is an immortal immovable object so
// write-barrier is not needed.
Comment("transition to megamorphic");
DCHECK(RootsTable::IsImmortalImmovable(RootIndex::kmegamorphic_symbol));
StoreFeedbackVectorSlot(
feedback_vector, slot_id,
HeapConstant(FeedbackVector::MegamorphicSentinel(isolate())),
SKIP_WRITE_BARRIER);
ReportFeedbackUpdate(feedback_vector, slot_id,
"ConstructWithSpread:TransitionMegamorphic");
Goto(&construct);
}
}
BIND(&construct);
Comment("call using ConstructWithSpread builtin");
Callable callable = CodeFactory::InterpreterPushArgsThenConstruct(
isolate(), InterpreterPushArgsMode::kWithFinalSpread);
return CallStub(callable, context, args.reg_count(), args.base_reg_location(),
target, new_target, UndefinedConstant());
}
template <class T>
TNode<T> InterpreterAssembler::CallRuntimeN(TNode<Uint32T> function_id,
TNode<Context> context,
const RegListNodePair& args,
int return_count) {
DCHECK(Bytecodes::MakesCallAlongCriticalPath(bytecode_));
DCHECK(Bytecodes::IsCallRuntime(bytecode_));
Callable callable = CodeFactory::InterpreterCEntry(isolate(), return_count);
TNode<Code> code_target = HeapConstant(callable.code());
// Get the function entry from the function id.
TNode<RawPtrT> function_table = ReinterpretCast<RawPtrT>(ExternalConstant(
ExternalReference::runtime_function_table_address(isolate())));
TNode<Word32T> function_offset =
Int32Mul(function_id, Int32Constant(sizeof(Runtime::Function)));
TNode<WordT> function =
IntPtrAdd(function_table, ChangeUint32ToWord(function_offset));
TNode<RawPtrT> function_entry = Load<RawPtrT>(
function, IntPtrConstant(offsetof(Runtime::Function, entry)));
return CallStub<T>(callable.descriptor(), code_target, context,
args.reg_count(), args.base_reg_location(),
function_entry);
}
template V8_EXPORT_PRIVATE TNode<Object> InterpreterAssembler::CallRuntimeN(
TNode<Uint32T> function_id, TNode<Context> context,
const RegListNodePair& args, int return_count);
template V8_EXPORT_PRIVATE TNode<PairT<Object, Object>>
InterpreterAssembler::CallRuntimeN(TNode<Uint32T> function_id,
TNode<Context> context,
const RegListNodePair& args,
int return_count);
void InterpreterAssembler::UpdateInterruptBudget(TNode<Int32T> weight,
bool backward) {
Comment("[ UpdateInterruptBudget");
// Assert that the weight is positive (negative weights should be implemented
// as backward updates).
CSA_ASSERT(this, Int32GreaterThanOrEqual(weight, Int32Constant(0)));
Label load_budget_from_bytecode(this), load_budget_done(this);
TNode<JSFunction> function = CAST(LoadRegister(Register::function_closure()));
TNode<FeedbackCell> feedback_cell =
LoadObjectField<FeedbackCell>(function, JSFunction::kFeedbackCellOffset);
TNode<Int32T> old_budget = LoadObjectField<Int32T>(
feedback_cell, FeedbackCell::kInterruptBudgetOffset);
// Make sure we include the current bytecode in the budget calculation.
TNode<Int32T> budget_after_bytecode =
Int32Sub(old_budget, Int32Constant(CurrentBytecodeSize()));
Label done(this);
TVARIABLE(Int32T, new_budget);
if (backward) {
// Update budget by |weight| and check if it reaches zero.
new_budget = Int32Sub(budget_after_bytecode, weight);
TNode<BoolT> condition =
Int32GreaterThanOrEqual(new_budget.value(), Int32Constant(0));
Label ok(this), interrupt_check(this, Label::kDeferred);
Branch(condition, &ok, &interrupt_check);
BIND(&interrupt_check);
CallRuntime(Runtime::kBytecodeBudgetInterruptFromBytecode, GetContext(),
function);
Goto(&done);
BIND(&ok);
} else {
// For a forward jump, we know we only increase the interrupt budget, so
// no need to check if it's below zero.
new_budget = Int32Add(budget_after_bytecode, weight);
}
// Update budget.
StoreObjectFieldNoWriteBarrier(
feedback_cell, FeedbackCell::kInterruptBudgetOffset, new_budget.value());
Goto(&done);
BIND(&done);
Comment("] UpdateInterruptBudget");
}
TNode<IntPtrT> InterpreterAssembler::Advance() {
return Advance(CurrentBytecodeSize());
}
TNode<IntPtrT> InterpreterAssembler::Advance(int delta) {
return Advance(IntPtrConstant(delta));
}
TNode<IntPtrT> InterpreterAssembler::Advance(TNode<IntPtrT> delta,
bool backward) {
#ifdef V8_TRACE_IGNITION
TraceBytecode(Runtime::kInterpreterTraceBytecodeExit);
#endif
TNode<IntPtrT> next_offset = backward ? IntPtrSub(BytecodeOffset(), delta)
: IntPtrAdd(BytecodeOffset(), delta);
bytecode_offset_ = next_offset;
return next_offset;
}
void InterpreterAssembler::Jump(TNode<IntPtrT> jump_offset, bool backward) {
DCHECK(!Bytecodes::IsStarLookahead(bytecode_, operand_scale_));
UpdateInterruptBudget(TruncateIntPtrToInt32(jump_offset), backward);
TNode<IntPtrT> new_bytecode_offset = Advance(jump_offset, backward);
TNode<RawPtrT> target_bytecode =
UncheckedCast<RawPtrT>(LoadBytecode(new_bytecode_offset));
DispatchToBytecode(target_bytecode, new_bytecode_offset);
}
void InterpreterAssembler::Jump(TNode<IntPtrT> jump_offset) {
Jump(jump_offset, false);
}
void InterpreterAssembler::JumpBackward(TNode<IntPtrT> jump_offset) {
Jump(jump_offset, true);
}
void InterpreterAssembler::JumpConditional(TNode<BoolT> condition,
TNode<IntPtrT> jump_offset) {
Label match(this), no_match(this);
Branch(condition, &match, &no_match);
BIND(&match);
Jump(jump_offset);
BIND(&no_match);
Dispatch();
}
void InterpreterAssembler::JumpIfTaggedEqual(TNode<Object> lhs,
TNode<Object> rhs,
TNode<IntPtrT> jump_offset) {
JumpConditional(TaggedEqual(lhs, rhs), jump_offset);
}
void InterpreterAssembler::JumpIfTaggedNotEqual(TNode<Object> lhs,
TNode<Object> rhs,
TNode<IntPtrT> jump_offset) {
JumpConditional(TaggedNotEqual(lhs, rhs), jump_offset);
}
TNode<WordT> InterpreterAssembler::LoadBytecode(
TNode<IntPtrT> bytecode_offset) {
TNode<Uint8T> bytecode =
Load<Uint8T>(BytecodeArrayTaggedPointer(), bytecode_offset);
return ChangeUint32ToWord(bytecode);
}
TNode<WordT> InterpreterAssembler::StarDispatchLookahead(
TNode<WordT> target_bytecode) {
Label do_inline_star(this), done(this);
TVARIABLE(WordT, var_bytecode, target_bytecode);
TNode<Int32T> star_bytecode =
Int32Constant(static_cast<int>(Bytecode::kStar));
TNode<BoolT> is_star =
Word32Equal(TruncateWordToInt32(target_bytecode), star_bytecode);
Branch(is_star, &do_inline_star, &done);
BIND(&do_inline_star);
{
InlineStar();
var_bytecode = LoadBytecode(BytecodeOffset());
Goto(&done);
}
BIND(&done);
return var_bytecode.value();
}
void InterpreterAssembler::InlineStar() {
Bytecode previous_bytecode = bytecode_;
AccumulatorUse previous_acc_use = accumulator_use_;
bytecode_ = Bytecode::kStar;
accumulator_use_ = AccumulatorUse::kNone;
#ifdef V8_TRACE_IGNITION
TraceBytecode(Runtime::kInterpreterTraceBytecodeEntry);
#endif
StoreRegister(GetAccumulator(),
BytecodeOperandReg(0, LoadSensitivity::kSafe));
DCHECK_EQ(accumulator_use_, Bytecodes::GetAccumulatorUse(bytecode_));
Advance();
bytecode_ = previous_bytecode;
accumulator_use_ = previous_acc_use;
}
void InterpreterAssembler::Dispatch() {
Comment("========= Dispatch");
DCHECK_IMPLIES(Bytecodes::MakesCallAlongCriticalPath(bytecode_), made_call_);
TNode<IntPtrT> target_offset = Advance();
TNode<WordT> target_bytecode = LoadBytecode(target_offset);
if (Bytecodes::IsStarLookahead(bytecode_, operand_scale_)) {
target_bytecode = StarDispatchLookahead(target_bytecode);
}
DispatchToBytecode(target_bytecode, BytecodeOffset());
}
void InterpreterAssembler::DispatchToBytecode(
TNode<WordT> target_bytecode, TNode<IntPtrT> new_bytecode_offset) {
if (FLAG_trace_ignition_dispatches) {
TraceBytecodeDispatch(target_bytecode);
}
TNode<RawPtrT> target_code_entry = Load<RawPtrT>(
DispatchTablePointer(), TimesSystemPointerSize(target_bytecode));
DispatchToBytecodeHandlerEntry(target_code_entry, new_bytecode_offset);
}
void InterpreterAssembler::DispatchToBytecodeHandlerEntry(
TNode<RawPtrT> handler_entry, TNode<IntPtrT> bytecode_offset) {
// Propagate speculation poisoning.
TNode<RawPtrT> poisoned_handler_entry =
UncheckedCast<RawPtrT>(WordPoisonOnSpeculation(handler_entry));
TailCallBytecodeDispatch(InterpreterDispatchDescriptor{},
poisoned_handler_entry, GetAccumulatorUnchecked(),
bytecode_offset, BytecodeArrayTaggedPointer(),
DispatchTablePointer());
}
void InterpreterAssembler::DispatchWide(OperandScale operand_scale) {
// Dispatching a wide bytecode requires treating the prefix
// bytecode a base pointer into the dispatch table and dispatching
// the bytecode that follows relative to this base.
//
// Indices 0-255 correspond to bytecodes with operand_scale == 0
// Indices 256-511 correspond to bytecodes with operand_scale == 1
// Indices 512-767 correspond to bytecodes with operand_scale == 2
DCHECK_IMPLIES(Bytecodes::MakesCallAlongCriticalPath(bytecode_), made_call_);
TNode<IntPtrT> next_bytecode_offset = Advance(1);
TNode<WordT> next_bytecode = LoadBytecode(next_bytecode_offset);
if (FLAG_trace_ignition_dispatches) {
TraceBytecodeDispatch(next_bytecode);
}
TNode<IntPtrT> base_index;
switch (operand_scale) {
case OperandScale::kDouble:
base_index = IntPtrConstant(1 << kBitsPerByte);
break;
case OperandScale::kQuadruple:
base_index = IntPtrConstant(2 << kBitsPerByte);
break;
default:
UNREACHABLE();
}
TNode<WordT> target_index = IntPtrAdd(base_index, next_bytecode);
TNode<RawPtrT> target_code_entry = Load<RawPtrT>(
DispatchTablePointer(), TimesSystemPointerSize(target_index));
DispatchToBytecodeHandlerEntry(target_code_entry, next_bytecode_offset);
}
void InterpreterAssembler::UpdateInterruptBudgetOnReturn() {
// TODO(rmcilroy): Investigate whether it is worth supporting self
// optimization of primitive functions like FullCodegen.
// Update profiling count by the number of bytes between the end of the
// current bytecode and the start of the first one, to simulate backedge to
// start of function.
//
// With headers and current offset, the bytecode array layout looks like:
//
// <---------- simulated backedge ----------
// | header | first bytecode | .... | return bytecode |
// |<------ current offset ------->
// ^ tagged bytecode array pointer
//
// UpdateInterruptBudget already handles adding the bytecode size to the
// length of the back-edge, so we just have to correct for the non-zero offset
// of the first bytecode.
const int kFirstBytecodeOffset = BytecodeArray::kHeaderSize - kHeapObjectTag;
TNode<Int32T> profiling_weight =
Int32Sub(TruncateIntPtrToInt32(BytecodeOffset()),
Int32Constant(kFirstBytecodeOffset));
UpdateInterruptBudget(profiling_weight, true);
}
TNode<Int8T> InterpreterAssembler::LoadOsrNestingLevel() {
return LoadObjectField<Int8T>(BytecodeArrayTaggedPointer(),
BytecodeArray::kOsrNestingLevelOffset);
}
void InterpreterAssembler::Abort(AbortReason abort_reason) {
TNode<Smi> abort_id = SmiConstant(abort_reason);
CallRuntime(Runtime::kAbort, GetContext(), abort_id);
}
void InterpreterAssembler::AbortIfWordNotEqual(TNode<WordT> lhs,
TNode<WordT> rhs,
AbortReason abort_reason) {
Label ok(this), abort(this, Label::kDeferred);
Branch(WordEqual(lhs, rhs), &ok, &abort);
BIND(&abort);
Abort(abort_reason);
Goto(&ok);
BIND(&ok);
}
void InterpreterAssembler::MaybeDropFrames(TNode<Context> context) {
TNode<ExternalReference> restart_fp_address =
ExternalConstant(ExternalReference::debug_restart_fp_address(isolate()));
TNode<IntPtrT> restart_fp = Load<IntPtrT>(restart_fp_address);
TNode<IntPtrT> null = IntPtrConstant(0);
Label ok(this), drop_frames(this);
Branch(IntPtrEqual(restart_fp, null), &ok, &drop_frames);
BIND(&drop_frames);
// We don't expect this call to return since the frame dropper tears down
// the stack and jumps into the function on the target frame to restart it.
CallStub(CodeFactory::FrameDropperTrampoline(isolate()), context, restart_fp);
Abort(AbortReason::kUnexpectedReturnFromFrameDropper);
Goto(&ok);
BIND(&ok);
}
void InterpreterAssembler::TraceBytecode(Runtime::FunctionId function_id) {
CallRuntime(function_id, GetContext(), BytecodeArrayTaggedPointer(),
SmiTag(BytecodeOffset()), GetAccumulatorUnchecked());
}
void InterpreterAssembler::TraceBytecodeDispatch(TNode<WordT> target_bytecode) {
TNode<ExternalReference> counters_table = ExternalConstant(
ExternalReference::interpreter_dispatch_counters(isolate()));
TNode<IntPtrT> source_bytecode_table_index = IntPtrConstant(
static_cast<int>(bytecode_) * (static_cast<int>(Bytecode::kLast) + 1));
TNode<WordT> counter_offset = TimesSystemPointerSize(
IntPtrAdd(source_bytecode_table_index, target_bytecode));
TNode<IntPtrT> old_counter = Load<IntPtrT>(counters_table, counter_offset);
Label counter_ok(this), counter_saturated(this, Label::kDeferred);
TNode<BoolT> counter_reached_max = WordEqual(
old_counter, IntPtrConstant(std::numeric_limits<uintptr_t>::max()));
Branch(counter_reached_max, &counter_saturated, &counter_ok);
BIND(&counter_ok);
{
TNode<IntPtrT> new_counter = IntPtrAdd(old_counter, IntPtrConstant(1));
StoreNoWriteBarrier(MachineType::PointerRepresentation(), counters_table,
counter_offset, new_counter);
Goto(&counter_saturated);
}
BIND(&counter_saturated);
}
// static
bool InterpreterAssembler::TargetSupportsUnalignedAccess() {
#if V8_TARGET_ARCH_MIPS || V8_TARGET_ARCH_MIPS64
return false;
#elif V8_TARGET_ARCH_IA32 || V8_TARGET_ARCH_X64 || V8_TARGET_ARCH_S390 || \
V8_TARGET_ARCH_ARM || V8_TARGET_ARCH_ARM64 || V8_TARGET_ARCH_PPC || \
V8_TARGET_ARCH_PPC64
return true;
#else
#error "Unknown Architecture"
#endif
}
void InterpreterAssembler::AbortIfRegisterCountInvalid(
TNode<FixedArrayBase> parameters_and_registers,
TNode<IntPtrT> formal_parameter_count, TNode<UintPtrT> register_count) {
TNode<IntPtrT> array_size =
LoadAndUntagFixedArrayBaseLength(parameters_and_registers);
Label ok(this), abort(this, Label::kDeferred);
Branch(UintPtrLessThanOrEqual(
IntPtrAdd(formal_parameter_count, register_count), array_size),
&ok, &abort);
BIND(&abort);
Abort(AbortReason::kInvalidParametersAndRegistersInGenerator);
Goto(&ok);
BIND(&ok);
}
TNode<FixedArray> InterpreterAssembler::ExportParametersAndRegisterFile(
TNode<FixedArray> array, const RegListNodePair& registers,
TNode<Int32T> formal_parameter_count) {
// Store the formal parameters (without receiver) followed by the
// registers into the generator's internal parameters_and_registers field.
TNode<IntPtrT> formal_parameter_count_intptr =
Signed(ChangeUint32ToWord(formal_parameter_count));
TNode<UintPtrT> register_count = ChangeUint32ToWord(registers.reg_count());
if (FLAG_debug_code) {
CSA_ASSERT(this, IntPtrEqual(registers.base_reg_location(),
RegisterLocation(Register(0))));
AbortIfRegisterCountInvalid(array, formal_parameter_count_intptr,
register_count);
}
{
TVARIABLE(IntPtrT, var_index);
var_index = IntPtrConstant(0);
// Iterate over parameters and write them into the array.
Label loop(this, &var_index), done_loop(this);
TNode<IntPtrT> reg_base =
IntPtrConstant(Register::FromParameterIndex(0, 1).ToOperand() + 1);
Goto(&loop);
BIND(&loop);
{
TNode<IntPtrT> index = var_index.value();
GotoIfNot(UintPtrLessThan(index, formal_parameter_count_intptr),
&done_loop);
TNode<IntPtrT> reg_index = IntPtrAdd(reg_base, index);
TNode<Object> value = LoadRegister(reg_index);
StoreFixedArrayElement(array, index, value);
var_index = IntPtrAdd(index, IntPtrConstant(1));
Goto(&loop);
}
BIND(&done_loop);
}
{
// Iterate over register file and write values into array.
// The mapping of register to array index must match that used in
// BytecodeGraphBuilder::VisitResumeGenerator.
TVARIABLE(IntPtrT, var_index);
var_index = IntPtrConstant(0);
Label loop(this, &var_index), done_loop(this);
Goto(&loop);
BIND(&loop);
{
TNode<IntPtrT> index = var_index.value();
GotoIfNot(UintPtrLessThan(index, register_count), &done_loop);
TNode<IntPtrT> reg_index =
IntPtrSub(IntPtrConstant(Register(0).ToOperand()), index);
TNode<Object> value = LoadRegister(reg_index);
TNode<IntPtrT> array_index =
IntPtrAdd(formal_parameter_count_intptr, index);
StoreFixedArrayElement(array, array_index, value);
var_index = IntPtrAdd(index, IntPtrConstant(1));
Goto(&loop);
}
BIND(&done_loop);
}
return array;
}
TNode<FixedArray> InterpreterAssembler::ImportRegisterFile(
TNode<FixedArray> array, const RegListNodePair& registers,
TNode<Int32T> formal_parameter_count) {
TNode<IntPtrT> formal_parameter_count_intptr =
Signed(ChangeUint32ToWord(formal_parameter_count));
TNode<UintPtrT> register_count = ChangeUint32ToWord(registers.reg_count());
if (FLAG_debug_code) {
CSA_ASSERT(this, IntPtrEqual(registers.base_reg_location(),
RegisterLocation(Register(0))));
AbortIfRegisterCountInvalid(array, formal_parameter_count_intptr,
register_count);
}
TVARIABLE(IntPtrT, var_index, IntPtrConstant(0));
// Iterate over array and write values into register file. Also erase the
// array contents to not keep them alive artificially.
Label loop(this, &var_index), done_loop(this);
Goto(&loop);
BIND(&loop);
{
TNode<IntPtrT> index = var_index.value();
GotoIfNot(UintPtrLessThan(index, register_count), &done_loop);
TNode<IntPtrT> array_index =
IntPtrAdd(formal_parameter_count_intptr, index);
TNode<Object> value = LoadFixedArrayElement(array, array_index);
TNode<IntPtrT> reg_index =
IntPtrSub(IntPtrConstant(Register(0).ToOperand()), index);
StoreRegister(value, reg_index);
StoreFixedArrayElement(array, array_index, StaleRegisterConstant());
var_index = IntPtrAdd(index, IntPtrConstant(1));
Goto(&loop);
}
BIND(&done_loop);
return array;
}
int InterpreterAssembler::CurrentBytecodeSize() const {
return Bytecodes::Size(bytecode_, operand_scale_);
}
void InterpreterAssembler::ToNumberOrNumeric(Object::Conversion mode) {
TNode<Object> object = GetAccumulator();
TNode<Context> context = GetContext();
TVARIABLE(Smi, var_type_feedback);
TVARIABLE(Numeric, var_result);
Label if_done(this), if_objectissmi(this), if_objectisheapnumber(this),
if_objectisother(this, Label::kDeferred);
GotoIf(TaggedIsSmi(object), &if_objectissmi);
Branch(IsHeapNumber(CAST(object)), &if_objectisheapnumber, &if_objectisother);
BIND(&if_objectissmi);
{
var_result = CAST(object);
var_type_feedback = SmiConstant(BinaryOperationFeedback::kSignedSmall);
Goto(&if_done);
}
BIND(&if_objectisheapnumber);
{
var_result = CAST(object);
var_type_feedback = SmiConstant(BinaryOperationFeedback::kNumber);
Goto(&if_done);
}
BIND(&if_objectisother);
{
auto builtin = Builtins::kNonNumberToNumber;
if (mode == Object::Conversion::kToNumeric) {
builtin = Builtins::kNonNumberToNumeric;
// Special case for collecting BigInt feedback.
Label not_bigint(this);
GotoIfNot(IsBigInt(CAST(object)), &not_bigint);
{
var_result = CAST(object);
var_type_feedback = SmiConstant(BinaryOperationFeedback::kBigInt);
Goto(&if_done);
}
BIND(&not_bigint);
}
// Convert {object} by calling out to the appropriate builtin.
var_result = CAST(CallBuiltin(builtin, context, object));
var_type_feedback = SmiConstant(BinaryOperationFeedback::kAny);
Goto(&if_done);
}
BIND(&if_done);
// Record the type feedback collected for {object}.
TNode<UintPtrT> slot_index = BytecodeOperandIdx(0);
TNode<HeapObject> maybe_feedback_vector = LoadFeedbackVector();
UpdateFeedback(var_type_feedback.value(), maybe_feedback_vector, slot_index);
SetAccumulator(var_result.value());
Dispatch();
}
} // namespace interpreter
} // namespace internal
} // namespace v8