languages/dart-sdk
languages/dart-sdk
1
0
Fork 0
mirror of https://github.com/dart-lang/sdk synced 2024-11-02 15:01:30 +00:00
Code Issues Projects Releases Packages Wiki Activity
dart-sdk/runtime/vm/raw_object.h
Slava Egorov 425a42e3be Revert "[vm, gc] Mark through new-space." 
This reverts commit 5daaa7d9eb.

Reason for revert: internal crashes

Original change's description:
> [vm, gc] Mark through new-space.
>
>   - Initial and final marking no longer visit all of new-space, reducing the STW pause for major GC.
>   - A scavenge during concurrent marking must forward / filter objects in the marking worklist that are moved / collected, increasing the STW pause for minor GC.
>   - Unreachable intergenerational cycles and weak references are collected in the next mark-sweep instead of first requiring enough scavenges to promote the whole cycle or weak target into old-space.
>   - Artificial minor GCs are no longer needed to avoid memory leaks from back-to-back major GCs.
>   - reachabilityBarrier is now just a count of major GCs.
>
> TEST=ci
> Change-Id: I4a6a23273d8ecb78c640f054731d4ceb737bfc4d
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/325840
> Reviewed-by: Siva Annamalai <asiva@google.com>
> Commit-Queue: Ryan Macnak <rmacnak@google.com>

Change-Id: I8a50074db343c63c14f0487ae8b4f5fee2c4ae76
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/330720
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
2023-10-18 07:47:39 +00:00

3760 lines
141 KiB
C++
Raw Blame History

// Copyright (c) 2012, the Dart project authors. Please see the AUTHORS file
// for details. All rights reserved. Use of this source code is governed by a
// BSD-style license that can be found in the LICENSE file.
#ifndef RUNTIME_VM_RAW_OBJECT_H_
#define RUNTIME_VM_RAW_OBJECT_H_
#if defined(SHOULD_NOT_INCLUDE_RUNTIME)
#error "Should not include runtime"
#endif
#include "platform/assert.h"
#include "vm/class_id.h"
#include "vm/compiler/method_recognizer.h"
#include "vm/compiler/runtime_api.h"
#include "vm/exceptions.h"
#include "vm/globals.h"
#include "vm/pointer_tagging.h"
#include "vm/snapshot.h"
#include "vm/tagged_pointer.h"
#include "vm/thread.h"
#include "vm/token.h"
#include "vm/token_position.h"
#include "vm/visitor.h"
// Currently we have two different axes for offset generation:
//
// * Target architecture
// * DART_PRECOMPILED_RUNTIME (i.e, AOT vs. JIT)
//
// That is, fields in UntaggedObject and its subclasses should only be included
// or excluded conditionally based on these factors. Otherwise, the generated
// offsets can be wrong (which should be caught by offset checking in dart.cc).
//
// TODO(dartbug.com/43646): Add DART_PRECOMPILER as another axis.
namespace dart {
// Forward declarations.
class Isolate;
class IsolateGroup;
#define DEFINE_FORWARD_DECLARATION(clazz) class Untagged##clazz;
CLASS_LIST(DEFINE_FORWARD_DECLARATION)
#undef DEFINE_FORWARD_DECLARATION
class CodeStatistics;
class StackFrame;
#define DEFINE_CONTAINS_COMPRESSED(type) \
static constexpr bool kContainsCompressedPointers = \
is_compressed_ptr<type>::value;
#define CHECK_CONTAIN_COMPRESSED(type) \
static_assert( \
kContainsCompressedPointers || is_uncompressed_ptr<type>::value, \
"From declaration uses ObjectPtr"); \
static_assert( \
!kContainsCompressedPointers || is_compressed_ptr<type>::value, \
"From declaration uses CompressedObjectPtr");
#define VISIT_FROM(first) \
DEFINE_CONTAINS_COMPRESSED(decltype(first##_)) \
static constexpr bool kContainsPointerFields = true; \
base_ptr_type<decltype(first##_)>::type* from() { \
return reinterpret_cast<base_ptr_type<decltype(first##_)>::type*>( \
&first##_); \
}
#define VISIT_FROM_PAYLOAD_START(elem_type) \
static_assert(is_uncompressed_ptr<elem_type>::value || \
is_compressed_ptr<elem_type>::value, \
"Payload elements must be object pointers"); \
DEFINE_CONTAINS_COMPRESSED(elem_type) \
static constexpr bool kContainsPointerFields = true; \
base_ptr_type<elem_type>::type* from() { \
const uword payload_start = reinterpret_cast<uword>(this) + sizeof(*this); \
ASSERT(Utils::IsAligned(payload_start, sizeof(elem_type))); \
return reinterpret_cast<base_ptr_type<elem_type>::type*>(payload_start); \
}
#define VISIT_TO(last) \
CHECK_CONTAIN_COMPRESSED(decltype(last##_)); \
static_assert(kContainsPointerFields, \
"Must have a corresponding VISIT_FROM"); \
base_ptr_type<decltype(last##_)>::type* to(intptr_t length = 0) { \
return reinterpret_cast<base_ptr_type<decltype(last##_)>::type*>( \
&last##_); \
}
#define VISIT_TO_PAYLOAD_END(elem_type) \
static_assert(is_uncompressed_ptr<elem_type>::value || \
is_compressed_ptr<elem_type>::value, \
"Payload elements must be object pointers"); \
static_assert(kContainsPointerFields, \
"Must have a corresponding VISIT_FROM"); \
CHECK_CONTAIN_COMPRESSED(elem_type); \
base_ptr_type<elem_type>::type* to(intptr_t length) { \
const uword payload_start = reinterpret_cast<uword>(this) + sizeof(*this); \
ASSERT(Utils::IsAligned(payload_start, sizeof(elem_type))); \
const uword payload_last = \
payload_start + sizeof(elem_type) * (length - 1); \
return reinterpret_cast<base_ptr_type<elem_type>::type*>(payload_last); \
}
#define VISIT_NOTHING() int NothingToVisit();
#if defined(DART_COMPRESSED_POINTERS)
#define ASSERT_UNCOMPRESSED(Type) \
static_assert(!Untagged##Type::kContainsCompressedPointers, \
"Should contain compressed pointers");
#define ASSERT_COMPRESSED(Type) \
static_assert(Untagged##Type::kContainsCompressedPointers, \
"Should not contain compressed pointers");
#else
// Do no checks if there are no compressed pointers.
#define ASSERT_UNCOMPRESSED(Type)
#define ASSERT_COMPRESSED(Type)
#endif
#define ASSERT_NOTHING_TO_VISIT(Type) \
ASSERT(SIZE_OF_RETURNED_VALUE(Untagged##Type, NothingToVisit) == sizeof(int))
enum TypedDataElementType {
#define V(name) k##name##Element,
CLASS_LIST_TYPED_DATA(V)
#undef V
};
#define VISITOR_SUPPORT(object) \
static intptr_t Visit##object##Pointers(object##Ptr raw_obj, \
ObjectPointerVisitor* visitor);
#define RAW_OBJECT_IMPLEMENTATION(object) \
private: /* NOLINT */ \
VISITOR_SUPPORT(object) \
friend class object; \
friend class UntaggedObject; \
friend class OffsetsTable; \
DISALLOW_ALLOCATION(); \
DISALLOW_IMPLICIT_CONSTRUCTORS(Untagged##object)
#define RAW_HEAP_OBJECT_IMPLEMENTATION(object) \
private: \
RAW_OBJECT_IMPLEMENTATION(object); \
friend class object##SerializationCluster; \
friend class object##DeserializationCluster; \
friend class object##MessageSerializationCluster; \
friend class object##MessageDeserializationCluster; \
friend class Serializer; \
friend class Deserializer; \
template <typename Base> \
friend class ObjectCopy; \
friend class Pass2Visitor;
// UntaggedObject is the base class of all raw objects; even though it carries
// the tags_ field not all raw objects are allocated in the heap and thus cannot
// be dereferenced (e.g. UntaggedSmi).
class UntaggedObject {
public:
// The tags field which is a part of the object header uses the following
// bit fields for storing tags.
enum TagBits {
kCardRememberedBit = 0,
kCanonicalBit = 1,
kOldAndNotMarkedBit = 2, // Incremental barrier target.
kNewBit = 3, // Generational barrier target.
kOldBit = 4, // Incremental barrier source.
kOldAndNotRememberedBit = 5, // Generational barrier source.
kImmutableBit = 6,
kReservedBit = 7,
kSizeTagPos = kReservedBit + 1, // = 8
kSizeTagSize = 4,
kClassIdTagPos = kSizeTagPos + kSizeTagSize, // = 12
kClassIdTagSize = 20,
kHashTagPos = kClassIdTagPos + kClassIdTagSize, // = 32
kHashTagSize = 32,
};
static constexpr intptr_t kGenerationalBarrierMask = 1 << kNewBit;
static constexpr intptr_t kIncrementalBarrierMask = 1 << kOldAndNotMarkedBit;
static constexpr intptr_t kBarrierOverlapShift = 2;
COMPILE_ASSERT(kOldAndNotMarkedBit + kBarrierOverlapShift == kOldBit);
COMPILE_ASSERT(kNewBit + kBarrierOverlapShift == kOldAndNotRememberedBit);
// The bit in the Smi tag position must be something that can be set to 0
// for a dead filler object of either generation.
// See Object::MakeUnusedSpaceTraversable.
COMPILE_ASSERT(kCardRememberedBit == 0);
// Encodes the object size in the tag in units of object alignment.
class SizeTag {
public:
typedef intptr_t Type;
static constexpr intptr_t kMaxSizeTagInUnitsOfAlignment =
((1 << UntaggedObject::kSizeTagSize) - 1);
static constexpr intptr_t kMaxSizeTag =
kMaxSizeTagInUnitsOfAlignment * kObjectAlignment;
static constexpr uword encode(intptr_t size) {
return SizeBits::encode(SizeToTagValue(size));
}
static constexpr uword decode(uword tag) {
return TagValueToSize(SizeBits::decode(tag));
}
static constexpr uword update(intptr_t size, uword tag) {
return SizeBits::update(SizeToTagValue(size), tag);
}
static constexpr bool SizeFits(intptr_t size) {
assert(Utils::IsAligned(size, kObjectAlignment));
return (size <= kMaxSizeTag);
}
private:
// The actual unscaled bit field used within the tag field.
class SizeBits
: public BitField<uword, intptr_t, kSizeTagPos, kSizeTagSize> {};
static constexpr intptr_t SizeToTagValue(intptr_t size) {
assert(Utils::IsAligned(size, kObjectAlignment));
return !SizeFits(size) ? 0 : (size >> kObjectAlignmentLog2);
}
static constexpr intptr_t TagValueToSize(intptr_t value) {
return value << kObjectAlignmentLog2;
}
};
class ClassIdTag : public BitField<uword,
ClassIdTagType,
kClassIdTagPos,
kClassIdTagSize> {};
COMPILE_ASSERT(kBitsPerByte * sizeof(ClassIdTagType) >= kClassIdTagSize);
COMPILE_ASSERT(kClassIdTagMax == (1 << kClassIdTagSize) - 1);
#if defined(HASH_IN_OBJECT_HEADER)
class HashTag : public BitField<uword, uint32_t, kHashTagPos, kHashTagSize> {
};
#endif
class CardRememberedBit
: public BitField<uword, bool, kCardRememberedBit, 1> {};
class OldAndNotMarkedBit
: public BitField<uword, bool, kOldAndNotMarkedBit, 1> {};
class NewBit : public BitField<uword, bool, kNewBit, 1> {};
class CanonicalBit : public BitField<uword, bool, kCanonicalBit, 1> {};
class OldBit : public BitField<uword, bool, kOldBit, 1> {};
class OldAndNotRememberedBit
: public BitField<uword, bool, kOldAndNotRememberedBit, 1> {};
// Will be set to 1 iff
// - is unmodifiable typed data view (backing store may be mutable)
// - is transitively immutable
class ImmutableBit : public BitField<uword, bool, kImmutableBit, 1> {};
class ReservedBit : public BitField<uword, intptr_t, kReservedBit, 1> {};
// Assumes this is a heap object.
bool IsNewObject() const {
uword addr = reinterpret_cast<uword>(this);
return (addr & kObjectAlignmentMask) == kNewObjectAlignmentOffset;
}
// Assumes this is a heap object.
bool IsOldObject() const {
uword addr = reinterpret_cast<uword>(this);
return (addr & kObjectAlignmentMask) == kOldObjectAlignmentOffset;
}
uword tags() const { return tags_; }
// Support for GC marking bit. Marked objects are either grey (not yet
// visited) or black (already visited).
static bool IsMarked(uword tags) { return !OldAndNotMarkedBit::decode(tags); }
bool IsMarked() const {
ASSERT(IsOldObject());
return !tags_.Read<OldAndNotMarkedBit>();
}
bool IsMarkedIgnoreRace() const {
ASSERT(IsOldObject());
return !tags_.ReadIgnoreRace<OldAndNotMarkedBit>();
}
void SetMarkBit() {
ASSERT(IsOldObject());
ASSERT(!IsMarked());
tags_.UpdateBool<OldAndNotMarkedBit>(false);
}
void SetMarkBitUnsynchronized() {
ASSERT(IsOldObject());
ASSERT(!IsMarked());
tags_.UpdateUnsynchronized<OldAndNotMarkedBit>(false);
}
void SetMarkBitRelease() {
ASSERT(IsOldObject());
ASSERT(!IsMarked());
tags_.UpdateBool<OldAndNotMarkedBit, std::memory_order_release>(false);
}
void ClearMarkBit() {
ASSERT(IsOldObject());
ASSERT(IsMarked());
tags_.UpdateBool<OldAndNotMarkedBit>(true);
}
// Returns false if the bit was already set.
DART_WARN_UNUSED_RESULT
bool TryAcquireMarkBit() {
ASSERT(IsOldObject());
return tags_.TryClear<OldAndNotMarkedBit>();
}
// Canonical objects have the property that two canonical objects are
// logically equal iff they are the same object (pointer equal).
bool IsCanonical() const { return tags_.Read<CanonicalBit>(); }
void SetCanonical() { tags_.UpdateBool<CanonicalBit>(true); }
void ClearCanonical() { tags_.UpdateBool<CanonicalBit>(false); }
bool IsImmutable() const { return tags_.Read<ImmutableBit>(); }
void SetImmutable() { tags_.UpdateBool<ImmutableBit>(true); }
void ClearImmutable() { tags_.UpdateBool<ImmutableBit>(false); }
bool InVMIsolateHeap() const;
// Support for GC remembered bit.
bool IsRemembered() const {
ASSERT(IsOldObject());
return !tags_.Read<OldAndNotRememberedBit>();
}
bool TryAcquireRememberedBit() {
ASSERT(!IsCardRemembered());
return tags_.TryClear<OldAndNotRememberedBit>();
}
void ClearRememberedBit() {
ASSERT(IsOldObject());
tags_.UpdateBool<OldAndNotRememberedBit>(true);
}
DART_FORCE_INLINE
void EnsureInRememberedSet(Thread* thread) {
if (TryAcquireRememberedBit()) {
thread->StoreBufferAddObject(ObjectPtr(this));
}
}
bool IsCardRemembered() const { return tags_.Read<CardRememberedBit>(); }
void SetCardRememberedBitUnsynchronized() {
ASSERT(!IsRemembered());
ASSERT(!IsCardRemembered());
tags_.UpdateUnsynchronized<CardRememberedBit>(true);
}
intptr_t GetClassId() const { return tags_.Read<ClassIdTag>(); }
#if defined(HASH_IN_OBJECT_HEADER)
uint32_t GetHeaderHash() const { return tags_.Read<HashTag>(); }
uint32_t SetHeaderHashIfNotSet(uint32_t h) {
return tags_.UpdateConditional<HashTag>(h, /*conditional_old_value=*/0);
}
#endif
intptr_t HeapSize() const {
uword tags = tags_;
intptr_t result = SizeTag::decode(tags);
if (result != 0) {
#if defined(DEBUG)
// TODO(22501) Array::MakeFixedLength has a race with this code: we might
// have loaded tags field and then MakeFixedLength could have updated it
// leading to inconsistency between HeapSizeFromClass() and
// SizeTag::decode(tags). We are working around it by reloading tags_ and
// recomputing size from tags.
const intptr_t size_from_class = HeapSizeFromClass(tags);
if ((result > size_from_class) && (GetClassId() == kArrayCid) &&
(tags_ != tags)) {
result = SizeTag::decode(tags_);
}
ASSERT(result == size_from_class);
#endif
return result;
}
result = HeapSizeFromClass(tags);
ASSERT(result > SizeTag::kMaxSizeTag);
return result;
}
// This variant must not deference this->tags_.
intptr_t HeapSize(uword tags) const {
intptr_t result = SizeTag::decode(tags);
if (result != 0) {
return result;
}
result = HeapSizeFromClass(tags);
ASSERT(result > SizeTag::kMaxSizeTag);
return result;
}
bool Contains(uword addr) const {
intptr_t this_size = HeapSize();
uword this_addr = UntaggedObject::ToAddr(this);
return (addr >= this_addr) && (addr < (this_addr + this_size));
}
void Validate(IsolateGroup* isolate_group) const;
// This function may access the class-ID in the header, but it cannot access
// the actual class object, because the sliding compactor uses this function
// while the class objects are being moved.
intptr_t VisitPointers(ObjectPointerVisitor* visitor) {
// Fall back to virtual variant for predefined classes
intptr_t class_id = GetClassId();
if (class_id < kNumPredefinedCids) {
return VisitPointersPredefined(visitor, class_id);
}
// Calculate the first and last raw object pointer fields.
intptr_t instance_size = HeapSize();
uword obj_addr = ToAddr(this);
uword from = obj_addr + sizeof(UntaggedObject);
uword to = obj_addr + instance_size - kCompressedWordSize;
const auto first = reinterpret_cast<CompressedObjectPtr*>(from);
const auto last = reinterpret_cast<CompressedObjectPtr*>(to);
const auto unboxed_fields_bitmap =
visitor->class_table()->GetUnboxedFieldsMapAt(class_id);
if (!unboxed_fields_bitmap.IsEmpty()) {
intptr_t bit = sizeof(UntaggedObject) / kCompressedWordSize;
for (CompressedObjectPtr* current = first; current <= last; current++) {
if (!unboxed_fields_bitmap.Get(bit++)) {
visitor->VisitCompressedPointers(heap_base(), current, current);
}
}
} else {
visitor->VisitCompressedPointers(heap_base(), first, last);
}
return instance_size;
}
template <class V>
DART_FORCE_INLINE intptr_t VisitPointersNonvirtual(V* visitor) {
// Fall back to virtual variant for predefined classes
intptr_t class_id = GetClassId();
if (class_id < kNumPredefinedCids) {
return VisitPointersPredefined(visitor, class_id);
}
// Calculate the first and last raw object pointer fields.
intptr_t instance_size = HeapSize();
uword obj_addr = ToAddr(this);
uword from = obj_addr + sizeof(UntaggedObject);
uword to = obj_addr + instance_size - kCompressedWordSize;
const auto first = reinterpret_cast<CompressedObjectPtr*>(from);
const auto last = reinterpret_cast<CompressedObjectPtr*>(to);
const auto unboxed_fields_bitmap =
visitor->class_table()->GetUnboxedFieldsMapAt(class_id);
if (!unboxed_fields_bitmap.IsEmpty()) {
intptr_t bit = sizeof(UntaggedObject) / kCompressedWordSize;
for (CompressedObjectPtr* current = first; current <= last; current++) {
if (!unboxed_fields_bitmap.Get(bit++)) {
visitor->V::VisitCompressedPointers(heap_base(), current, current);
}
}
} else {
visitor->V::VisitCompressedPointers(heap_base(), first, last);
}
return instance_size;
}
// This variant ensures that we do not visit the extra slot created from
// rounding up instance sizes up to the allocation unit.
void VisitPointersPrecise(ObjectPointerVisitor* visitor);
static ObjectPtr FromAddr(uword addr) {
// We expect the untagged address here.
ASSERT((addr & kSmiTagMask) != kHeapObjectTag);
return static_cast<ObjectPtr>(addr + kHeapObjectTag);
}
static uword ToAddr(const UntaggedObject* raw_obj) {
return reinterpret_cast<uword>(raw_obj);
}
static uword ToAddr(const ObjectPtr raw_obj) {
return static_cast<uword>(raw_obj) - kHeapObjectTag;
}
static bool IsCanonical(intptr_t value) {
return CanonicalBit::decode(value);
}
private:
AtomicBitFieldContainer<uword> tags_; // Various object tags (bits).
intptr_t VisitPointersPredefined(ObjectPointerVisitor* visitor,
intptr_t class_id);
intptr_t HeapSizeFromClass(uword tags) const;
void SetClassId(intptr_t new_cid) { tags_.Update<ClassIdTag>(new_cid); }
void SetClassIdUnsynchronized(intptr_t new_cid) {
tags_.UpdateUnsynchronized<ClassIdTag>(new_cid);
}
protected:
// Automatically inherited by subclasses unless overridden.
static constexpr bool kContainsCompressedPointers = false;
// Automatically inherited by subclasses unless overridden.
static constexpr bool kContainsPointerFields = false;
// The first offset in an allocated object of the given type that contains a
// (possibly compressed) object pointer. Used to initialize object pointer
// fields to Object::null() instead of 0.
//
// Always returns an offset after the object header tags.
template <typename T>
DART_FORCE_INLINE static uword from_offset();
// The last offset in an allocated object of the given untagged type that
// contains a (possibly compressed) object pointer. Used to initialize object
// pointer fields to Object::null() instead of 0.
//
// Takes an optional argument that is the number of elements in the payload,
// which is ignored if the object never contains a payload.
//
// If there are no pointer fields in the object, then
// to_offset<T>() < from_offset<T>().
template <typename T>
DART_FORCE_INLINE static uword to_offset(intptr_t length = 0);
// All writes to heap objects should ultimately pass through one of the
// methods below or their counterparts in Object, to ensure that the
// write barrier is correctly applied.
template <typename type, std::memory_order order = std::memory_order_relaxed>
type LoadPointer(type const* addr) const {
return reinterpret_cast<std::atomic<type>*>(const_cast<type*>(addr))
->load(order);
}
template <typename type,
typename compressed_type,
std::memory_order order = std::memory_order_relaxed>
type LoadCompressedPointer(compressed_type const* addr) const {
compressed_type v = reinterpret_cast<std::atomic<compressed_type>*>(
const_cast<compressed_type*>(addr))
->load(order);
return static_cast<type>(v.Decompress(heap_base()));
}
uword heap_base() const {
return reinterpret_cast<uword>(this) & kHeapBaseMask;
}
template <typename type, std::memory_order order = std::memory_order_relaxed>
void StorePointer(type const* addr, type value) {
reinterpret_cast<std::atomic<type>*>(const_cast<type*>(addr))
->store(value, order);
if (value.IsHeapObject()) {
CheckHeapPointerStore(value, Thread::Current());
}
}
template <typename type,
typename compressed_type,
std::memory_order order = std::memory_order_relaxed>
void StoreCompressedPointer(compressed_type const* addr, type value) {
reinterpret_cast<std::atomic<compressed_type>*>(
const_cast<compressed_type*>(addr))
->store(static_cast<compressed_type>(value), order);
if (value.IsHeapObject()) {
CheckHeapPointerStore(value, Thread::Current());
}
}
template <typename type>
void StorePointer(type const* addr, type value, Thread* thread) {
*const_cast<type*>(addr) = value;
if (value.IsHeapObject()) {
CheckHeapPointerStore(value, thread);
}
}
template <typename type, typename compressed_type>
void StoreCompressedPointer(compressed_type const* addr,
type value,
Thread* thread) {
*const_cast<compressed_type*>(addr) = value;
if (value.IsHeapObject()) {
CheckHeapPointerStore(value, thread);
}
}
template <typename type>
void StorePointerUnaligned(type const* addr, type value, Thread* thread) {
StoreUnaligned(const_cast<type*>(addr), value);
if (value->IsHeapObject()) {
CheckHeapPointerStore(value, thread);
}
}
// Note: StoreArrayPointer won't work if value_type is a compressed pointer.
template <typename type,
std::memory_order order = std::memory_order_relaxed,
typename value_type = type>
void StoreArrayPointer(type const* addr, value_type value) {
reinterpret_cast<std::atomic<type>*>(const_cast<type*>(addr))
->store(type(value), order);
if (value->IsHeapObject()) {
CheckArrayPointerStore(addr, value, Thread::Current());
}
}
template <typename type, typename value_type = type>
void StoreArrayPointer(type const* addr, value_type value, Thread* thread) {
*const_cast<type*>(addr) = value;
if (value->IsHeapObject()) {
CheckArrayPointerStore(addr, value, thread);
}
}
template <typename type, typename compressed_type, std::memory_order order>
void StoreCompressedArrayPointer(compressed_type const* addr, type value) {
reinterpret_cast<std::atomic<compressed_type>*>(
const_cast<compressed_type*>(addr))
->store(static_cast<compressed_type>(value), order);
if (value->IsHeapObject()) {
CheckArrayPointerStore(addr, value, Thread::Current());
}
}
template <typename type, typename compressed_type, std::memory_order order>
void StoreCompressedArrayPointer(compressed_type const* addr,
type value,
Thread* thread) {
reinterpret_cast<std::atomic<compressed_type>*>(
const_cast<compressed_type*>(addr))
->store(static_cast<compressed_type>(value), order);
if (value->IsHeapObject()) {
CheckArrayPointerStore(addr, value, thread);
}
}
template <typename type, typename compressed_type>
void StoreCompressedArrayPointer(compressed_type const* addr,
type value,
Thread* thread) {
*const_cast<compressed_type*>(addr) = value;
if (value->IsHeapObject()) {
CheckArrayPointerStore(addr, value, thread);
}
}
template <typename type,
typename compressed_type,
std::memory_order order = std::memory_order_relaxed>
type ExchangeCompressedPointer(compressed_type const* addr, type value) {
compressed_type previous_value =
reinterpret_cast<std::atomic<compressed_type>*>(
const_cast<compressed_type*>(addr))
->exchange(static_cast<compressed_type>(value), order);
if (value.IsHeapObject()) {
CheckHeapPointerStore(value, Thread::Current());
}
return static_cast<type>(previous_value.Decompress(heap_base()));
}
template <std::memory_order order = std::memory_order_relaxed>
SmiPtr LoadSmi(SmiPtr const* addr) const {
return reinterpret_cast<std::atomic<SmiPtr>*>(const_cast<SmiPtr*>(addr))
->load(order);
}
template <std::memory_order order = std::memory_order_relaxed>
SmiPtr LoadCompressedSmi(CompressedSmiPtr const* addr) const {
return static_cast<SmiPtr>(reinterpret_cast<std::atomic<CompressedSmiPtr>*>(
const_cast<CompressedSmiPtr*>(addr))
->load(order)
.DecompressSmi());
}
// Use for storing into an explicitly Smi-typed field of an object
// (i.e., both the previous and new value are Smis).
template <typename type, std::memory_order order = std::memory_order_relaxed>
void StoreSmi(type const* addr, type value) {
// Can't use Contains, as array length is initialized through this method.
ASSERT(reinterpret_cast<uword>(addr) >= UntaggedObject::ToAddr(this));
reinterpret_cast<std::atomic<type>*>(const_cast<type*>(addr))
->store(value, order);
}
template <std::memory_order order = std::memory_order_relaxed>
void StoreCompressedSmi(CompressedSmiPtr const* addr, SmiPtr value) {
// Can't use Contains, as array length is initialized through this method.
ASSERT(reinterpret_cast<uword>(addr) >= UntaggedObject::ToAddr(this));
reinterpret_cast<std::atomic<CompressedSmiPtr>*>(
const_cast<CompressedSmiPtr*>(addr))
->store(static_cast<CompressedSmiPtr>(value), order);
}
private:
DART_FORCE_INLINE
void CheckHeapPointerStore(ObjectPtr value, Thread* thread) {
uword source_tags = this->tags_;
uword target_tags = value->untag()->tags_;
if (((source_tags >> kBarrierOverlapShift) & target_tags &
thread->write_barrier_mask()) != 0) {
if (value->IsNewObject()) {
// Generational barrier: record when a store creates an
// old-and-not-remembered -> new reference.
EnsureInRememberedSet(thread);
} else {
// Incremental barrier: record when a store creates an
// old -> old-and-not-marked reference.
ASSERT(value->IsOldObject());
if (ClassIdTag::decode(target_tags) == kInstructionsCid) {
// Instruction pages may be non-writable. Defer marking.
thread->DeferredMarkingStackAddObject(value);
return;
}
if (value->untag()->TryAcquireMarkBit()) {
thread->MarkingStackAddObject(value);
}
}
}
}
template <typename type, typename value_type>
DART_FORCE_INLINE void CheckArrayPointerStore(type const* addr,
value_type value,
Thread* thread) {
uword source_tags = this->tags_;
uword target_tags = value->untag()->tags_;
if (((source_tags >> kBarrierOverlapShift) & target_tags &
thread->write_barrier_mask()) != 0) {
if (value->IsNewObject()) {
// Generational barrier: record when a store creates an
// old-and-not-remembered -> new reference.
if (this->IsCardRemembered()) {
RememberCard(addr);
} else if (this->TryAcquireRememberedBit()) {
thread->StoreBufferAddObject(static_cast<ObjectPtr>(this));
}
} else {
// Incremental barrier: record when a store creates an
// old -> old-and-not-marked reference.
ASSERT(value->IsOldObject());
if (ClassIdTag::decode(target_tags) == kInstructionsCid) {
// Instruction pages may be non-writable. Defer marking.
thread->DeferredMarkingStackAddObject(value);
return;
}
if (value->untag()->TryAcquireMarkBit()) {
thread->MarkingStackAddObject(value);
}
}
}
}
friend class StoreBufferUpdateVisitor; // RememberCard
void RememberCard(ObjectPtr const* slot);
#if defined(DART_COMPRESSED_POINTERS)
void RememberCard(CompressedObjectPtr const* slot);
#endif
friend class Array;
friend class ByteBuffer;
friend class CidRewriteVisitor;
friend class Closure;
friend class Code;
friend class Pointer;
friend class Double;
friend class DynamicLibrary;
friend class ForwardPointersVisitor; // StorePointer
friend class FreeListElement;
friend class Function;
friend class GCMarker;
friend class GCSweeper;
friend class ExternalTypedData;
friend class GrowableObjectArray; // StorePointer
template <bool>
friend class MarkingVisitorBase;
friend class Mint;
friend class Object;
friend class OneByteString; // StoreSmi
friend class UntaggedInstance;
friend class Scavenger;
template <bool>
friend class ScavengerVisitorBase;
friend class ImageReader; // tags_ check
friend class ImageWriter;
friend class AssemblyImageWriter;
friend class BlobImageWriter;
friend class Deserializer;
friend class String;
friend class WeakProperty; // StorePointer
friend class Instance; // StorePointer
friend class StackFrame; // GetCodeObject assertion.
friend class CodeLookupTableBuilder; // profiler
friend class ObjectLocator;
friend class WriteBarrierUpdateVisitor; // CheckHeapPointerStore
friend class OffsetsTable;
friend class Object;
friend uword TagsFromUntaggedObject(UntaggedObject*); // tags_
friend void SetNewSpaceTaggingWord(ObjectPtr, classid_t, uint32_t); // tags_
friend class ObjectCopyBase; // LoadPointer/StorePointer
friend void ReportImpossibleNullError(intptr_t cid,
StackFrame* caller_frame,
Thread* thread);
DISALLOW_ALLOCATION();
DISALLOW_IMPLICIT_CONSTRUCTORS(UntaggedObject);
};
// Note that the below templates for from_offset and to_offset for objects
// with pointer fields assume that the range from from() and to() cover all
// pointer fields. If this is not the case (e.g., the next_seen_by_gc_ field
// in WeakArray/WeakProperty/WeakReference), then specialize the definitions.
template <typename T>
DART_FORCE_INLINE uword UntaggedObject::from_offset() {
if constexpr (T::kContainsPointerFields) {
return reinterpret_cast<uword>(reinterpret_cast<T*>(kOffsetOfPtr)->from()) -
kOffsetOfPtr;
} else {
// Non-zero to ensure to_offset() < from_offset() in this case, as
// to_offset() is the offset to the last pointer field, not past it.
return sizeof(UntaggedObject);
}
}
template <typename T>
DART_FORCE_INLINE uword UntaggedObject::to_offset(intptr_t length) {
if constexpr (T::kContainsPointerFields) {
return reinterpret_cast<uword>(
reinterpret_cast<T*>(kOffsetOfPtr)->to(length)) -
kOffsetOfPtr;
} else {
USE(length);
// Zero to ensure to_offset() < from_offset() in this case, as
// from_offset() is guaranteed to return an offset after the header tags.
return 0;
}
}
inline intptr_t ObjectPtr::GetClassId() const {
return untag()->GetClassId();
}
#define POINTER_FIELD(type, name) \
public: \
template <std::memory_order order = std::memory_order_relaxed> \
type name() const { \
return LoadPointer<type, order>(&name##_); \
} \
template <std::memory_order order = std::memory_order_relaxed> \
void set_##name(type value) { \
StorePointer<type, order>(&name##_, value); \
} \
\
protected: \
type name##_;
#define COMPRESSED_POINTER_FIELD(type, name) \
public: \
template <std::memory_order order = std::memory_order_relaxed> \
type name() const { \
return LoadCompressedPointer<type, Compressed##type, order>(&name##_); \
} \
template <std::memory_order order = std::memory_order_relaxed> \
void set_##name(type value) { \
StoreCompressedPointer<type, Compressed##type, order>(&name##_, value); \
} \
\
protected: \
Compressed##type name##_;
#define ARRAY_POINTER_FIELD(type, name) \
public: \
template <std::memory_order order = std::memory_order_relaxed> \
type name() const { \
return LoadPointer<type, order>(&name##_); \
} \
template <std::memory_order order = std::memory_order_relaxed> \
void set_##name(type value) { \
StoreArrayPointer<type, order>(&name##_, value); \
} \
\
protected: \
type name##_;
#define COMPRESSED_ARRAY_POINTER_FIELD(type, name) \
public: \
template <std::memory_order order = std::memory_order_relaxed> \
type name() const { \
return LoadPointer<Compressed##type, order>(&name##_).Decompress( \
heap_base()); \
} \
template <std::memory_order order = std::memory_order_relaxed> \
void set_##name(type value) { \
StoreCompressedArrayPointer<type, Compressed##type, order>(&name##_, \
value); \
} \
\
protected: \
Compressed##type name##_;
#define VARIABLE_POINTER_FIELDS(type, accessor_name, array_name) \
public: \
template <std::memory_order order = std::memory_order_relaxed> \
type accessor_name(intptr_t index) const { \
return LoadPointer<type, order>(&array_name()[index]); \
} \
template <std::memory_order order = std::memory_order_relaxed> \
void set_##accessor_name(intptr_t index, type value) { \
StoreArrayPointer<type, order>(&array_name()[index], value); \
} \
template <std::memory_order order = std::memory_order_relaxed> \
void set_##accessor_name(intptr_t index, type value, Thread* thread) { \
StoreArrayPointer<type, order>(&array_name()[index], value, thread); \
} \
\
protected: \
type* array_name() { OPEN_ARRAY_START(type, type); } \
type const* array_name() const { OPEN_ARRAY_START(type, type); } \
VISIT_TO_PAYLOAD_END(type)
#define COMPRESSED_VARIABLE_POINTER_FIELDS(type, accessor_name, array_name) \
public: \
template <std::memory_order order = std::memory_order_relaxed> \
type accessor_name(intptr_t index) const { \
return LoadCompressedPointer<type, Compressed##type, order>( \
&array_name()[index]); \
} \
template <std::memory_order order = std::memory_order_relaxed> \
void set_##accessor_name(intptr_t index, type value) { \
StoreCompressedArrayPointer<type, Compressed##type, order>( \
&array_name()[index], value); \
} \
template <std::memory_order order = std::memory_order_relaxed> \
void set_##accessor_name(intptr_t index, type value, Thread* thread) { \
StoreCompressedArrayPointer<type, Compressed##type, order>( \
&array_name()[index], value, thread); \
} \
\
protected: \
Compressed##type* array_name() { \
OPEN_ARRAY_START(Compressed##type, Compressed##type); \
} \
Compressed##type const* array_name() const { \
OPEN_ARRAY_START(Compressed##type, Compressed##type); \
} \
VISIT_TO_PAYLOAD_END(Compressed##type)
#define SMI_FIELD(type, name) \
public: \
template <std::memory_order order = std::memory_order_relaxed> \
type name() const { \
type result = LoadSmi<order>(&name##_); \
ASSERT(!result.IsHeapObject()); \
return result; \
} \
template <std::memory_order order = std::memory_order_relaxed> \
void set_##name(type value) { \
ASSERT(!value.IsHeapObject()); \
StoreSmi<type, order>(&name##_, value); \
} \
\
protected: \
type name##_;
#define COMPRESSED_SMI_FIELD(type, name) \
public: \
template <std::memory_order order = std::memory_order_relaxed> \
type name() const { \
type result = LoadCompressedSmi<order>(&name##_); \
ASSERT(!result.IsHeapObject()); \
return result; \
} \
template <std::memory_order order = std::memory_order_relaxed> \
void set_##name(type value) { \
ASSERT(!value.IsHeapObject()); \
StoreCompressedSmi(&name##_, value); \
} \
\
protected: \
Compressed##type name##_;
// Used to define untagged object fields that can have values wrapped in
// WeakSerializationReferences. Since WeakSerializationReferences are only used
// during precompilation, these fields have type CompressedObjectPtr in the
// precompiler and the normally expected type otherwise.
//
// Fields that are defined with WSR_COMPRESSED_POINTER_FIELD should have
// getters and setters that are declared in object.h with
// PRECOMPILER_WSR_FIELD_DECLARATION and defined in object.cc with
// PRECOMPILER_WSR_FIELD_DEFINITION.
#if defined(DART_PRECOMPILER)
#define WSR_COMPRESSED_POINTER_FIELD(Type, Name) \
COMPRESSED_POINTER_FIELD(ObjectPtr, Name)
#else
#define WSR_COMPRESSED_POINTER_FIELD(Type, Name) \
COMPRESSED_POINTER_FIELD(Type, Name)
#endif
class UntaggedClass : public UntaggedObject {
public:
enum ClassFinalizedState {
kAllocated = 0, // Initial state.
kPreFinalized, // VM classes: size precomputed, but no checks done.
kFinalized, // Class parsed, code compiled, not ready for allocation.
kAllocateFinalized, // CHA invalidated, class is ready for allocation.
};
enum ClassLoadingState {
// Class object is created, but it is not filled up.
// At this state class can only be used as a forward reference during
// class loading.
kNameOnly = 0,
// Class declaration information such as type parameters, supertype and
// implemented interfaces are loaded. However, types in the class are
// not finalized yet.
kDeclarationLoaded,
// Types in the class are finalized. At this point, members can be loaded
// and class can be finalized.
kTypeFinalized,
};
classid_t id() const { return id_; }
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(Class);
COMPRESSED_POINTER_FIELD(StringPtr, name)
VISIT_FROM(name)
NOT_IN_PRODUCT(COMPRESSED_POINTER_FIELD(StringPtr, user_name))
COMPRESSED_POINTER_FIELD(ArrayPtr, functions)
COMPRESSED_POINTER_FIELD(ArrayPtr, functions_hash_table)
COMPRESSED_POINTER_FIELD(ArrayPtr, fields)
COMPRESSED_POINTER_FIELD(ArrayPtr, offset_in_words_to_field)
COMPRESSED_POINTER_FIELD(ArrayPtr, interfaces) // Array of AbstractType.
COMPRESSED_POINTER_FIELD(ScriptPtr, script)
COMPRESSED_POINTER_FIELD(LibraryPtr, library)
COMPRESSED_POINTER_FIELD(TypeParametersPtr, type_parameters)
COMPRESSED_POINTER_FIELD(TypePtr, super_type)
// Canonicalized const instances of this class.
COMPRESSED_POINTER_FIELD(ArrayPtr, constants)
// Declaration type for this class.
COMPRESSED_POINTER_FIELD(TypePtr, declaration_type)
// Cache for dispatcher functions.
COMPRESSED_POINTER_FIELD(ArrayPtr, invocation_dispatcher_cache)
#if !defined(PRODUCT) || !defined(DART_PRECOMPILED_RUNTIME)
// Array of Class.
COMPRESSED_POINTER_FIELD(GrowableObjectArrayPtr, direct_implementors)
// Array of Class.
COMPRESSED_POINTER_FIELD(GrowableObjectArrayPtr, direct_subclasses)
#endif // !defined(PRODUCT) || !defined(DART_PRECOMPILED_RUNTIME)
// Cached declaration instance type arguments for this class.
// Not preserved in AOT snapshots.
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr,
declaration_instance_type_arguments)
#if !defined(DART_PRECOMPILED_RUNTIME)
// Stub code for allocation of instances.
COMPRESSED_POINTER_FIELD(CodePtr, allocation_stub)
// CHA optimized codes.
COMPRESSED_POINTER_FIELD(WeakArrayPtr, dependent_code)
#endif // !defined(DART_PRECOMPILED_RUNTIME)
#if defined(DART_PRECOMPILED_RUNTIME)
VISIT_TO(declaration_instance_type_arguments)
#else
VISIT_TO(dependent_code)
#endif // defined(DART_PRECOMPILED_RUNTIME)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) {
switch (kind) {
case Snapshot::kFullAOT:
#if defined(PRODUCT)
return reinterpret_cast<CompressedObjectPtr*>(
&invocation_dispatcher_cache_);
#else
return reinterpret_cast<CompressedObjectPtr*>(&direct_subclasses_);
#endif // defined(PRODUCT)
case Snapshot::kFull:
case Snapshot::kFullCore:
#if !defined(DART_PRECOMPILED_RUNTIME)
return reinterpret_cast<CompressedObjectPtr*>(&allocation_stub_);
#endif
case Snapshot::kFullJIT:
#if !defined(DART_PRECOMPILED_RUNTIME)
return reinterpret_cast<CompressedObjectPtr*>(&dependent_code_);
#endif
case Snapshot::kNone:
case Snapshot::kInvalid:
break;
}
UNREACHABLE();
return nullptr;
}
NOT_IN_PRECOMPILED(TokenPosition token_pos_);
NOT_IN_PRECOMPILED(TokenPosition end_token_pos_);
NOT_IN_PRECOMPILED(classid_t implementor_cid_);
classid_t id_; // Class Id, also index in the class table.
int16_t num_type_arguments_; // Number of type arguments in flattened vector.
uint16_t num_native_fields_;
uint32_t state_bits_;
// Size if fixed len or 0 if variable len.
int32_t host_instance_size_in_words_;
// Offset of type args fld.
int32_t host_type_arguments_field_offset_in_words_;
// Offset of the next instance field.
int32_t host_next_field_offset_in_words_;
#if defined(DART_PRECOMPILER)
// Size if fixed len or 0 if variable len (target).
int32_t target_instance_size_in_words_;
// Offset of type args fld.
int32_t target_type_arguments_field_offset_in_words_;
// Offset of the next instance field (target).
int32_t target_next_field_offset_in_words_;
#endif // defined(DART_PRECOMPILER)
#if !defined(DART_PRECOMPILED_RUNTIME)
uint32_t kernel_offset_;
#endif // !defined(DART_PRECOMPILED_RUNTIME)
friend class Instance;
friend class IsolateGroup;
friend class Object;
friend class UntaggedInstance;
friend class UntaggedInstructions;
friend class UntaggedTypeArguments;
friend class MessageSerializer;
friend class InstanceSerializationCluster;
friend class TypeSerializationCluster;
friend class CidRewriteVisitor;
friend class FinalizeVMIsolateVisitor;
friend class Api;
};
class UntaggedPatchClass : public UntaggedObject {
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(PatchClass);
COMPRESSED_POINTER_FIELD(ClassPtr, wrapped_class)
VISIT_FROM(wrapped_class)
COMPRESSED_POINTER_FIELD(ScriptPtr, script)
#if !defined(DART_PRECOMPILED_RUNTIME)
COMPRESSED_POINTER_FIELD(KernelProgramInfoPtr, kernel_program_info)
VISIT_TO(kernel_program_info)
#else
VISIT_TO(script)
#endif
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) {
switch (kind) {
case Snapshot::kFullAOT:
return reinterpret_cast<CompressedObjectPtr*>(&script_);
case Snapshot::kFull:
case Snapshot::kFullCore:
case Snapshot::kFullJIT:
#if !defined(DART_PRECOMPILED_RUNTIME)
return reinterpret_cast<CompressedObjectPtr*>(&kernel_program_info_);
#else
UNREACHABLE();
return nullptr;
#endif
case Snapshot::kNone:
case Snapshot::kInvalid:
break;
}
UNREACHABLE();
return nullptr;
}
NOT_IN_PRECOMPILED(intptr_t kernel_library_index_);
friend class Function;
};
class UntaggedFunction : public UntaggedObject {
public:
// When you add a new kind, please also update the observatory to account
// for the new string returned by KindToCString().
// - runtime/observatory/lib/src/models/objects/function.dart (FunctionKind)
// - runtime/observatory/lib/src/elements/function_view.dart
// (_functionKindToString)
// - runtime/observatory/lib/src/service/object.dart (stringToFunctionKind)
#define FOR_EACH_RAW_FUNCTION_KIND(V) \
/* an ordinary or operator method */ \
V(RegularFunction) \
/* a user-declared closure function */ \
V(ClosureFunction) \
/* an implicit closure (i.e., tear-off) */ \
V(ImplicitClosureFunction) \
/* a signature only without actual code */ \
V(GetterFunction) \
/* setter functions e.g: set foo(..) { .. } */ \
V(SetterFunction) \
/* a generative (is_static=false) or factory (is_static=true) constructor */ \
V(Constructor) \
/* an implicit getter for instance fields */ \
V(ImplicitGetter) \
/* an implicit setter for instance fields */ \
V(ImplicitSetter) \
/* represents an implicit getter for static fields with initializers */ \
V(ImplicitStaticGetter) \
/* the initialization expression for a static or instance field */ \
V(FieldInitializer) \
/* return a closure on the receiver for tear-offs */ \
V(MethodExtractor) \
/* builds an Invocation and invokes noSuchMethod */ \
V(NoSuchMethodDispatcher) \
/* invokes a field as a closure (i.e., call-through-getter) */ \
V(InvokeFieldDispatcher) \
/* a generated irregexp matcher function. */ \
V(IrregexpFunction) \
/* a forwarder which performs type checks for arguments of a dynamic call */ \
/* (i.e., those checks omitted by the caller for interface calls). */ \
V(DynamicInvocationForwarder) \
/* A `dart:ffi` call or callback trampoline. */ \
V(FfiTrampoline) \
/* getter for a record field */ \
V(RecordFieldGetter)
enum Kind {
#define KIND_DEFN(Name) k##Name,
FOR_EACH_RAW_FUNCTION_KIND(KIND_DEFN)
#undef KIND_DEFN
};
static const char* KindToCString(Kind k) {
switch (k) {
#define KIND_CASE(Name) \
case Kind::k##Name: \
return #Name;
FOR_EACH_RAW_FUNCTION_KIND(KIND_CASE)
#undef KIND_CASE
default:
UNREACHABLE();
return nullptr;
}
}
static bool ParseKind(const char* str, Kind* out) {
#define KIND_CASE(Name) \
if (strcmp(str, #Name) == 0) { \
*out = Kind::k##Name; \
return true; \
}
FOR_EACH_RAW_FUNCTION_KIND(KIND_CASE)
#undef KIND_CASE
return false;
}
enum AsyncModifier {
kNoModifier = 0x0,
kAsyncBit = 0x1,
kGeneratorBit = 0x2,
kAsync = kAsyncBit,
kSyncGen = kGeneratorBit,
kAsyncGen = kAsyncBit | kGeneratorBit,
};
// Wraps a 64-bit integer to represent the bitmap for unboxed parameters and
// return value. Two bits are used for each of them to denote if it is boxed,
// unboxed integer, unboxed double or unboxed record.
// It includes the two bits for the receiver, even though currently we
// do not have information from TFA that allows the receiver to be unboxed.
class alignas(8) UnboxedParameterBitmap {
public:
enum UnboxedState {
kBoxed,
kUnboxedInt,
kUnboxedDouble,
kUnboxedRecord,
};
static constexpr intptr_t kBitsPerElement = 2;
static constexpr uint64_t kElementBitmask = (1 << kBitsPerElement) - 1;
static constexpr intptr_t kCapacity =
(kBitsPerByte * sizeof(uint64_t)) / kBitsPerElement;
UnboxedParameterBitmap() : bitmap_(0) {}
explicit UnboxedParameterBitmap(uint64_t bitmap) : bitmap_(bitmap) {}
UnboxedParameterBitmap(const UnboxedParameterBitmap&) = default;
UnboxedParameterBitmap& operator=(const UnboxedParameterBitmap&) = default;
DART_FORCE_INLINE bool IsUnboxed(intptr_t position) const {
return At(position) != kBoxed;
}
DART_FORCE_INLINE bool IsUnboxedInteger(intptr_t position) const {
return At(position) == kUnboxedInt;
}
DART_FORCE_INLINE bool IsUnboxedDouble(intptr_t position) const {
return At(position) == kUnboxedDouble;
}
DART_FORCE_INLINE bool IsUnboxedRecord(intptr_t position) const {
return At(position) == kUnboxedRecord;
}
DART_FORCE_INLINE void SetUnboxedInteger(intptr_t position) {
SetAt(position, kUnboxedInt);
}
DART_FORCE_INLINE void SetUnboxedDouble(intptr_t position) {
SetAt(position, kUnboxedDouble);
}
DART_FORCE_INLINE void SetUnboxedRecord(intptr_t position) {
SetAt(position, kUnboxedRecord);
}
DART_FORCE_INLINE uint64_t Value() const { return bitmap_; }
DART_FORCE_INLINE bool IsEmpty() const { return bitmap_ == 0; }
DART_FORCE_INLINE void Reset() { bitmap_ = 0; }
DART_FORCE_INLINE bool HasUnboxedParameters() const {
return (bitmap_ >> kBitsPerElement) != 0;
}
private:
DART_FORCE_INLINE UnboxedState At(intptr_t position) const {
if (position >= kCapacity) {
return kBoxed;
}
return static_cast<UnboxedState>(
(bitmap_ >> (kBitsPerElement * position)) & kElementBitmask);
}
DART_FORCE_INLINE void SetAt(intptr_t position, UnboxedState state) {
ASSERT(position < kCapacity);
const intptr_t shift = kBitsPerElement * position;
bitmap_ = (bitmap_ & ~(kElementBitmask << shift)) |
(static_cast<decltype(bitmap_)>(state) << shift);
}
uint64_t bitmap_;
};
private:
friend class Class;
friend class UnitDeserializationRoots;
RAW_HEAP_OBJECT_IMPLEMENTATION(Function);
uword entry_point_; // Accessed from generated code.
uword unchecked_entry_point_; // Accessed from generated code.
COMPRESSED_POINTER_FIELD(StringPtr, name)
VISIT_FROM(name)
// Class or patch class or mixin class where this function is defined.
COMPRESSED_POINTER_FIELD(ObjectPtr, owner)
WSR_COMPRESSED_POINTER_FIELD(FunctionTypePtr, signature)
// Additional data specific to the function kind. See Function::set_data()
// for details.
COMPRESSED_POINTER_FIELD(ObjectPtr, data)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) {
switch (kind) {
case Snapshot::kFullAOT:
case Snapshot::kFull:
case Snapshot::kFullCore:
case Snapshot::kFullJIT:
return reinterpret_cast<CompressedObjectPtr*>(&data_);
case Snapshot::kNone:
case Snapshot::kInvalid:
break;
}
UNREACHABLE();
return nullptr;
}
// ICData of unoptimized code.
COMPRESSED_POINTER_FIELD(ArrayPtr, ic_data_array);
// Currently active code. Accessed from generated code.
COMPRESSED_POINTER_FIELD(CodePtr, code);
#if defined(DART_PRECOMPILED_RUNTIME)
VISIT_TO(code);
#else
// Positional parameter names are not needed in the AOT runtime.
COMPRESSED_POINTER_FIELD(ArrayPtr, positional_parameter_names);
// Unoptimized code, keep it after optimization.
COMPRESSED_POINTER_FIELD(CodePtr, unoptimized_code);
VISIT_TO(unoptimized_code);
UnboxedParameterBitmap unboxed_parameters_info_;
#endif
#if !defined(DART_PRECOMPILED_RUNTIME) || \
(defined(DART_PRECOMPILED_RUNTIME) && !defined(PRODUCT))
TokenPosition token_pos_;
#endif
#if !defined(DART_PRECOMPILED_RUNTIME)
TokenPosition end_token_pos_;
#endif
AtomicBitFieldContainer<uint32_t> kind_tag_; // See Function::KindTagBits.
#define JIT_FUNCTION_COUNTERS(F) \
F(intptr_t, int32_t, usage_counter) \
F(intptr_t, uint16_t, optimized_instruction_count) \
F(intptr_t, uint16_t, optimized_call_site_count) \
F(int8_t, int8_t, deoptimization_counter) \
F(intptr_t, int8_t, state_bits) \
F(int, int8_t, inlining_depth)
#if !defined(DART_PRECOMPILED_RUNTIME)
uint32_t kernel_offset_;
#define DECLARE(return_type, type, name) type name##_;
JIT_FUNCTION_COUNTERS(DECLARE)
#undef DECLARE
AtomicBitFieldContainer<uint8_t> packed_fields_;
static constexpr intptr_t kMaxOptimizableBits = 1;
using PackedOptimizable =
BitField<decltype(packed_fields_), bool, 0, kMaxOptimizableBits>;
#endif // !defined(DART_PRECOMPILED_RUNTIME)
};
class UntaggedClosureData : public UntaggedObject {
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(ClosureData);
COMPRESSED_POINTER_FIELD(ContextScopePtr, context_scope)
VISIT_FROM(context_scope)
// Enclosing function of this local function.
WSR_COMPRESSED_POINTER_FIELD(FunctionPtr, parent_function)
// Closure object for static implicit closures.
COMPRESSED_POINTER_FIELD(ClosurePtr, closure)
VISIT_TO(closure)
enum class DefaultTypeArgumentsKind : uint8_t {
// Only here to make sure it's explicitly set appropriately.
kInvalid = 0,
// Must instantiate the default type arguments before use.
kNeedsInstantiation,
// The default type arguments are already instantiated.
kIsInstantiated,
// Use the instantiator type arguments that would be used to instantiate
// the default type arguments, as instantiating produces the same result.
kSharesInstantiatorTypeArguments,
// Use the function type arguments that would be used to instantiate
// the default type arguments, as instantiating produces the same result.
kSharesFunctionTypeArguments,
};
// kernel_to_il.cc assumes we can load the untagged value and box it in a Smi.
static_assert(sizeof(DefaultTypeArgumentsKind) * kBitsPerByte <=
compiler::target::kSmiBits,
"Default type arguments kind must fit in a Smi");
static constexpr uint8_t kNoAwaiterLinkDepth = 0xFF;
AtomicBitFieldContainer<uint32_t> packed_fields_;
using PackedDefaultTypeArgumentsKind =
BitField<decltype(packed_fields_), DefaultTypeArgumentsKind, 0, 8>;
using PackedAwaiterLinkDepth =
BitField<decltype(packed_fields_),
uint8_t,
PackedDefaultTypeArgumentsKind::kNextBit,
8>;
using PackedAwaiterLinkIndex = BitField<decltype(packed_fields_),
uint8_t,
PackedAwaiterLinkDepth::kNextBit,
8>;
friend class Function;
friend class UnitDeserializationRoots;
};
class UntaggedFfiTrampolineData : public UntaggedObject {
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(FfiTrampolineData);
COMPRESSED_POINTER_FIELD(TypePtr, signature_type)
VISIT_FROM(signature_type)
COMPRESSED_POINTER_FIELD(FunctionTypePtr, c_signature)
// Target Dart method for callbacks, otherwise null.
COMPRESSED_POINTER_FIELD(FunctionPtr, callback_target)
// For callbacks, value to return if Dart target throws an exception.
COMPRESSED_POINTER_FIELD(InstancePtr, callback_exceptional_return)
VISIT_TO(callback_exceptional_return)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
// Callback id for callbacks.
//
// The callbacks ids are used so that native callbacks can lookup their own
// code objects, since native code doesn't pass code objects into function
// calls. The callback id is also used to for verifying that callbacks are
// called on the correct isolate. See DLRT_VerifyCallbackIsolate for details.
//
// Callback id is -1 for non-callbacks or when id is not allocated yet.
// Check 'callback_target_' to determine if this is a callback or not.
int32_t callback_id_;
// Whether this is a leaf call - i.e. one that doesn't call back into Dart.
bool is_leaf_;
// The kind of trampoline this is. See FfiFunctionKind.
uint8_t ffi_function_kind_;
};
class UntaggedField : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(Field);
COMPRESSED_POINTER_FIELD(StringPtr, name)
VISIT_FROM(name)
// Class or patch class or mixin class where this field is defined or original
// field.
COMPRESSED_POINTER_FIELD(ObjectPtr, owner)
COMPRESSED_POINTER_FIELD(AbstractTypePtr, type)
// Static initializer function.
COMPRESSED_POINTER_FIELD(FunctionPtr, initializer_function)
// - for instance fields: offset in words to the value in the class instance.
// - for static fields: index into field_table.
COMPRESSED_POINTER_FIELD(SmiPtr, host_offset_or_field_id)
COMPRESSED_POINTER_FIELD(SmiPtr, guarded_list_length)
COMPRESSED_POINTER_FIELD(WeakArrayPtr, dependent_code)
VISIT_TO(dependent_code);
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) {
switch (kind) {
case Snapshot::kFull:
case Snapshot::kFullCore:
case Snapshot::kFullJIT:
case Snapshot::kFullAOT:
return reinterpret_cast<CompressedObjectPtr*>(&initializer_function_);
case Snapshot::kNone:
case Snapshot::kInvalid:
break;
}
UNREACHABLE();
return nullptr;
}
TokenPosition token_pos_;
TokenPosition end_token_pos_;
ClassIdTagType guarded_cid_;
ClassIdTagType is_nullable_; // kNullCid if field can contain null value and
// kIllegalCid otherwise.
#if !defined(DART_PRECOMPILED_RUNTIME)
uint32_t kernel_offset_;
#endif // !defined(DART_PRECOMPILED_RUNTIME)
// Offset to the guarded length field inside an instance of class matching
// guarded_cid_. Stored corrected by -kHeapObjectTag to simplify code
// generated on platforms with weak addressing modes (ARM).
int8_t guarded_list_length_in_object_offset_;
// Runtime tracking state of exactness of type annotation of this field.
// See StaticTypeExactnessState for the meaning and possible values in this
// field.
int8_t static_type_exactness_state_;
uint16_t kind_bits_; // static, final, const, has initializer....
#if !defined(DART_PRECOMPILED_RUNTIME)
// for instance fields, the offset in words in the target architecture
int32_t target_offset_;
#endif // !defined(DART_PRECOMPILED_RUNTIME)
friend class CidRewriteVisitor;
friend class GuardFieldClassInstr; // For sizeof(guarded_cid_/...)
friend class LoadFieldInstr; // For sizeof(guarded_cid_/...)
friend class StoreFieldInstr; // For sizeof(guarded_cid_/...)
};
class alignas(8) UntaggedScript : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(Script);
COMPRESSED_POINTER_FIELD(StringPtr, url)
VISIT_FROM(url)
COMPRESSED_POINTER_FIELD(StringPtr, resolved_url)
COMPRESSED_POINTER_FIELD(TypedDataPtr, line_starts)
#if !defined(PRODUCT) && !defined(DART_PRECOMPILED_RUNTIME)
COMPRESSED_POINTER_FIELD(TypedDataViewPtr, constant_coverage)
#endif // !defined(PRODUCT) && !defined(DART_PRECOMPILED_RUNTIME)
COMPRESSED_POINTER_FIELD(ArrayPtr, debug_positions)
COMPRESSED_POINTER_FIELD(KernelProgramInfoPtr, kernel_program_info)
COMPRESSED_POINTER_FIELD(StringPtr, source)
VISIT_TO(source)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) {
switch (kind) {
case Snapshot::kFullAOT:
#if defined(PRODUCT)
return reinterpret_cast<CompressedObjectPtr*>(&url_);
#else
return reinterpret_cast<CompressedObjectPtr*>(&resolved_url_);
#endif
case Snapshot::kFull:
case Snapshot::kFullCore:
case Snapshot::kFullJIT:
return reinterpret_cast<CompressedObjectPtr*>(&kernel_program_info_);
case Snapshot::kNone:
case Snapshot::kInvalid:
break;
}
UNREACHABLE();
return nullptr;
}
#if !defined(PRODUCT) && !defined(DART_PRECOMPILED_RUNTIME)
int64_t load_timestamp_;
int32_t kernel_script_index_;
#else
int32_t kernel_script_index_;
int64_t load_timestamp_;
#endif
#if !defined(DART_PRECOMPILED_RUNTIME)
int32_t flags_and_max_position_;
public:
using LazyLookupSourceAndLineStartsBit =
BitField<decltype(flags_and_max_position_), bool, 0, 1>;
using HasCachedMaxPositionBit =
BitField<decltype(flags_and_max_position_),
bool,
LazyLookupSourceAndLineStartsBit::kNextBit,
1>;
using CachedMaxPositionBitField = BitField<decltype(flags_and_max_position_),
intptr_t,
HasCachedMaxPositionBit::kNextBit>;
private:
#endif
};
class UntaggedLibrary : public UntaggedObject {
enum LibraryState {
kAllocated, // Initial state.
kLoadRequested, // Compiler or script requested load of library.
kLoadInProgress, // Library is in the process of being loaded.
kLoaded, // Library is loaded.
};
enum LibraryFlags {
kDartSchemeBit = 0,
kDebuggableBit, // True if debugger can stop in library.
kInFullSnapshotBit, // True if library is in a full snapshot.
kNnbdBit, // True if library is non nullable by default.
kNnbdCompiledModePos, // Encodes nnbd compiled mode of constants in lib.
kNnbdCompiledModeSize = 2,
kNumFlagBits = kNnbdCompiledModePos + kNnbdCompiledModeSize,
};
COMPILE_ASSERT(kNumFlagBits <= (sizeof(uint8_t) * kBitsPerByte));
class DartSchemeBit : public BitField<uint8_t, bool, kDartSchemeBit, 1> {};
class DebuggableBit : public BitField<uint8_t, bool, kDebuggableBit, 1> {};
class InFullSnapshotBit
: public BitField<uint8_t, bool, kInFullSnapshotBit, 1> {};
class NnbdBit : public BitField<uint8_t, bool, kNnbdBit, 1> {};
class NnbdCompiledModeBits : public BitField<uint8_t,
uint8_t,
kNnbdCompiledModePos,
kNnbdCompiledModeSize> {};
RAW_HEAP_OBJECT_IMPLEMENTATION(Library);
COMPRESSED_POINTER_FIELD(StringPtr, name)
VISIT_FROM(name)
COMPRESSED_POINTER_FIELD(StringPtr, url)
COMPRESSED_POINTER_FIELD(StringPtr, private_key)
// Top-level names in this library.
COMPRESSED_POINTER_FIELD(ArrayPtr, dictionary)
// Metadata on classes, methods etc.
COMPRESSED_POINTER_FIELD(ArrayPtr, metadata)
// Class containing top-level elements.
COMPRESSED_POINTER_FIELD(ClassPtr, toplevel_class)
COMPRESSED_POINTER_FIELD(GrowableObjectArrayPtr, used_scripts)
COMPRESSED_POINTER_FIELD(LoadingUnitPtr, loading_unit)
// List of Namespaces imported without prefix.
COMPRESSED_POINTER_FIELD(ArrayPtr, imports)
// List of re-exported Namespaces.
COMPRESSED_POINTER_FIELD(ArrayPtr, exports)
COMPRESSED_POINTER_FIELD(ArrayPtr, dependencies)
#if !defined(DART_PRECOMPILED_RUNTIME)
COMPRESSED_POINTER_FIELD(KernelProgramInfoPtr, kernel_program_info)
#endif
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) {
switch (kind) {
case Snapshot::kFullAOT:
return reinterpret_cast<CompressedObjectPtr*>(&exports_);
case Snapshot::kFull:
case Snapshot::kFullCore:
case Snapshot::kFullJIT:
#if !defined(DART_PRECOMPILED_RUNTIME)
return reinterpret_cast<CompressedObjectPtr*>(&kernel_program_info_);
#else
UNREACHABLE();
return nullptr;
#endif
case Snapshot::kNone:
case Snapshot::kInvalid:
break;
}
UNREACHABLE();
return nullptr;
}
// Array of scripts loaded in this library.
COMPRESSED_POINTER_FIELD(ArrayPtr, loaded_scripts);
VISIT_TO(loaded_scripts);
Dart_NativeEntryResolver native_entry_resolver_; // Resolves natives.
Dart_NativeEntrySymbol native_entry_symbol_resolver_;
Dart_FfiNativeResolver ffi_native_resolver_;
classid_t index_; // Library id number.
uint16_t num_imports_; // Number of entries in imports_.
int8_t load_state_; // Of type LibraryState.
uint8_t flags_; // BitField for LibraryFlags.
#if !defined(DART_PRECOMPILED_RUNTIME)
uint32_t kernel_library_index_;
#endif // !defined(DART_PRECOMPILED_RUNTIME)
friend class Class;
friend class Isolate;
};
class UntaggedNamespace : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(Namespace);
// library with name dictionary.
COMPRESSED_POINTER_FIELD(LibraryPtr, target)
VISIT_FROM(target)
// list of names that are exported.
COMPRESSED_POINTER_FIELD(ArrayPtr, show_names)
// list of names that are hidden.
COMPRESSED_POINTER_FIELD(ArrayPtr, hide_names)
COMPRESSED_POINTER_FIELD(LibraryPtr, owner)
VISIT_TO(owner)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) {
switch (kind) {
case Snapshot::kFullAOT:
return reinterpret_cast<CompressedObjectPtr*>(&target_);
case Snapshot::kFull:
case Snapshot::kFullCore:
case Snapshot::kFullJIT:
return reinterpret_cast<CompressedObjectPtr*>(&owner_);
case Snapshot::kNone:
case Snapshot::kInvalid:
break;
}
UNREACHABLE();
return nullptr;
}
};
// Contains information about a kernel [Component].
//
// Used to access string tables, canonical name tables, constants, metadata, ...
class UntaggedKernelProgramInfo : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(KernelProgramInfo);
COMPRESSED_POINTER_FIELD(TypedDataBasePtr, kernel_component)
VISIT_FROM(kernel_component)
COMPRESSED_POINTER_FIELD(TypedDataPtr, string_offsets)
COMPRESSED_POINTER_FIELD(TypedDataViewPtr, string_data)
COMPRESSED_POINTER_FIELD(TypedDataPtr, canonical_names)
COMPRESSED_POINTER_FIELD(TypedDataViewPtr, metadata_payloads)
COMPRESSED_POINTER_FIELD(TypedDataViewPtr, metadata_mappings)
COMPRESSED_POINTER_FIELD(ArrayPtr, scripts)
COMPRESSED_POINTER_FIELD(ArrayPtr, constants)
COMPRESSED_POINTER_FIELD(TypedDataViewPtr, constants_table)
COMPRESSED_POINTER_FIELD(ArrayPtr, libraries_cache)
COMPRESSED_POINTER_FIELD(ArrayPtr, classes_cache)
VISIT_TO(classes_cache)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) {
return reinterpret_cast<CompressedObjectPtr*>(&constants_table_);
}
};
class UntaggedWeakSerializationReference : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(WeakSerializationReference);
COMPRESSED_POINTER_FIELD(ObjectPtr, target)
VISIT_FROM(target)
COMPRESSED_POINTER_FIELD(ObjectPtr, replacement)
VISIT_TO(replacement)
};
class UntaggedWeakArray : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(WeakArray);
COMPRESSED_POINTER_FIELD(WeakArrayPtr, next_seen_by_gc)
COMPRESSED_SMI_FIELD(SmiPtr, length)
VISIT_FROM(length)
// Variable length data follows here.
COMPRESSED_VARIABLE_POINTER_FIELDS(ObjectPtr, element, data)
template <typename Table, bool kAllCanonicalObjectsAreIncludedIntoSet>
friend class CanonicalSetDeserializationCluster;
template <typename Type, typename PtrType>
friend class GCLinkedList;
template <bool>
friend class MarkingVisitorBase;
template <bool>
friend class ScavengerVisitorBase;
};
// WeakArray is special in that it has a pointer field which is not
// traversed by pointer visitors, and thus not in the range [from(),to()]:
// next_seen_by_gc, which is before the other fields.
template <>
DART_FORCE_INLINE uword UntaggedObject::from_offset<UntaggedWeakArray>() {
return OFFSET_OF(UntaggedWeakArray, next_seen_by_gc_);
}
class UntaggedCode : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(Code);
// When in the precompiled runtime, there is no disabling of Code objects
// and thus no active_instructions_ field. Thus, the entry point caches are
// only set once during deserialization. If not using bare instructions,
// the caches should match the entry points for instructions_.
//
// Otherwise, they should contain entry points for active_instructions_.
uword entry_point_; // Accessed from generated code.
// In AOT this entry-point supports switchable calls. It checks the type of
// the receiver on entry to the function and calls a stub to patch up the
// caller if they mismatch.
uword monomorphic_entry_point_; // Accessed from generated code (AOT only).
// Entry-point used from call-sites with some additional static information.
// The exact behavior of this entry-point depends on the kind of function:
//
// kRegularFunction/kSetter/kGetter:
//
// Call-site is assumed to know that the (type) arguments are invariantly
// type-correct against the actual runtime-type of the receiver. For
// instance, this entry-point is used for invocations against "this" and
// invocations from IC stubs that test the class type arguments.
//
// kClosureFunction:
//
// Call-site is assumed to pass the correct number of positional and type
// arguments (except in the case of partial instantiation, when the type
// arguments are omitted). All (type) arguments are assumed to match the
// corresponding (type) parameter types (bounds).
//
// kImplicitClosureFunction:
//
// Similar to kClosureFunction, except that the types (bounds) of the (type)
// arguments are expected to match the *runtime signature* of the closure,
// which (unlike with kClosureFunction) may have more general (type)
// parameter types (bounds) than the declared type of the forwarded method.
//
// In many cases a distinct static entry-point will not be created for a
// function if it would not be able to skip a lot of work (e.g., no argument
// type checks are necessary or this Code belongs to a stub). In this case
// 'unchecked_entry_point_' will refer to the same position as 'entry_point_'.
//
uword unchecked_entry_point_; // Accessed from generated code.
uword monomorphic_unchecked_entry_point_; // Accessed from generated code.
POINTER_FIELD(ObjectPoolPtr, object_pool) // Accessed from generated code.
VISIT_FROM(object_pool)
POINTER_FIELD(InstructionsPtr,
instructions) // Accessed from generated code.
// If owner_ is Function::null() the owner is a regular stub.
// If owner_ is a Class the owner is the allocation stub for that class.
// Else, owner_ is a regular Dart Function.
POINTER_FIELD(ObjectPtr, owner) // Function, Null, or a Class.
POINTER_FIELD(ExceptionHandlersPtr, exception_handlers)
POINTER_FIELD(PcDescriptorsPtr, pc_descriptors)
// If FLAG_precompiled_mode, then this field contains
// TypedDataPtr catch_entry_moves_maps
// Otherwise, it is
// SmiPtr num_variables
POINTER_FIELD(ObjectPtr, catch_entry)
POINTER_FIELD(CompressedStackMapsPtr, compressed_stackmaps)
POINTER_FIELD(ArrayPtr, inlined_id_to_function)
POINTER_FIELD(CodeSourceMapPtr, code_source_map)
NOT_IN_PRECOMPILED(POINTER_FIELD(InstructionsPtr, active_instructions))
NOT_IN_PRECOMPILED(POINTER_FIELD(ArrayPtr, deopt_info_array))
// (code-offset, function, code) triples.
NOT_IN_PRECOMPILED(POINTER_FIELD(ArrayPtr, static_calls_target_table))
// If return_address_metadata_ is a Smi, it is the offset to the prologue.
// Else, return_address_metadata_ is null.
NOT_IN_PRODUCT(POINTER_FIELD(ObjectPtr, return_address_metadata))
NOT_IN_PRODUCT(POINTER_FIELD(LocalVarDescriptorsPtr, var_descriptors))
NOT_IN_PRODUCT(POINTER_FIELD(ArrayPtr, comments))
#if !defined(PRODUCT)
VISIT_TO(comments);
#elif defined(DART_PRECOMPILED_RUNTIME)
VISIT_TO(code_source_map);
#else
VISIT_TO(static_calls_target_table);
#endif
// Compilation timestamp.
NOT_IN_PRODUCT(alignas(8) int64_t compile_timestamp_);
// state_bits_ is a bitfield with three fields:
// The optimized bit, the alive bit, and a count of the number of pointer
// offsets.
// Alive: If true, the embedded object pointers will be visited during GC.
int32_t state_bits_;
// Caches the unchecked entry point offset for instructions_, in case we need
// to reset the active_instructions_ to instructions_.
NOT_IN_PRECOMPILED(uint32_t unchecked_offset_);
// Stores the instructions length when not using RawInstructions objects.
ONLY_IN_PRECOMPILED(uint32_t instructions_length_);
// Variable length data follows here.
int32_t* data() { OPEN_ARRAY_START(int32_t, int32_t); }
const int32_t* data() const { OPEN_ARRAY_START(int32_t, int32_t); }
static bool ContainsPC(const ObjectPtr raw_obj, uword pc);
friend class Function;
template <bool>
friend class MarkingVisitorBase;
friend class StackFrame;
friend class Profiler;
friend class FunctionDeserializationCluster;
friend class UnitSerializationRoots;
friend class UnitDeserializationRoots;
friend class CallSiteResetter;
};
class UntaggedObjectPool : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(ObjectPool);
intptr_t length_;
struct Entry {
union {
ObjectPtr raw_obj_;
uword raw_value_;
};
};
Entry* data() { OPEN_ARRAY_START(Entry, Entry); }
Entry const* data() const { OPEN_ARRAY_START(Entry, Entry); }
DEFINE_CONTAINS_COMPRESSED(decltype(Entry::raw_obj_));
// The entry bits are located after the last entry. They are encoded versions
// of `ObjectPool::TypeBits() | ObjectPool::PatchabilityBit()`.
uint8_t* entry_bits() { return reinterpret_cast<uint8_t*>(&data()[length_]); }
uint8_t const* entry_bits() const {
return reinterpret_cast<uint8_t const*>(&data()[length_]);
}
friend class Object;
friend class CodeSerializationCluster;
friend class UnitSerializationRoots;
friend class UnitDeserializationRoots;
};
class UntaggedInstructions : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(Instructions);
VISIT_NOTHING();
// Instructions size in bytes and flags.
// Currently, only flag indicates 1 or 2 entry points.
uint32_t size_and_flags_;
// Variable length data follows here.
uint8_t* data() { OPEN_ARRAY_START(uint8_t, uint8_t); }
// Private helper function used while visiting stack frames. The
// code which iterates over dart frames is also called during GC and
// is not allowed to create handles.
static bool ContainsPC(const InstructionsPtr raw_instr, uword pc);
friend class UntaggedCode;
friend class UntaggedFunction;
friend class Code;
friend class StackFrame;
template <bool>
friend class MarkingVisitorBase;
friend class Function;
friend class ImageReader;
friend class ImageWriter;
friend class AssemblyImageWriter;
friend class BlobImageWriter;
};
// Used to carry extra information to the VM without changing the embedder
// interface, to provide memory accounting for the bare instruction payloads
// we serialize, since they are no longer part of RawInstructions objects,
// and to avoid special casing bare instructions payload Images in the GC.
class UntaggedInstructionsSection : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(InstructionsSection);
VISIT_NOTHING();
// Instructions section payload length in bytes.
uword payload_length_;
// The offset of the corresponding BSS section from this text section.
word bss_offset_;
// The relocated address of this text section in the shared object. Properly
// filled for ELF snapshots, always 0 in assembly snapshots. (For the latter,
// we instead get the value during BSS initialization and store it there.)
uword instructions_relocated_address_;
// The offset of the GNU build ID note section from this text section.
word build_id_offset_;
// Variable length data follows here.
uint8_t* data() { OPEN_ARRAY_START(uint8_t, uint8_t); }
friend class Image;
};
class UntaggedPcDescriptors : public UntaggedObject {
public:
// The macro argument V is passed two arguments, the raw name of the enum value
// and the initialization expression used within the enum definition. The uses
// of enum values inside the initialization expression are hardcoded currently,
// so the second argument is useless outside the enum definition and should be
// dropped by other users of this macro.
#define FOR_EACH_RAW_PC_DESCRIPTOR(V) \
/* Deoptimization continuation point. */ \
V(Deopt, 1) \
/* IC call. */ \
V(IcCall, kDeopt << 1) \
/* Call to a known target via stub. */ \
V(UnoptStaticCall, kIcCall << 1) \
/* Runtime call. */ \
V(RuntimeCall, kUnoptStaticCall << 1) \
/* OSR entry point in unopt. code. */ \
V(OsrEntry, kRuntimeCall << 1) \
/* Call rewind target address. */ \
V(Rewind, kOsrEntry << 1) \
/* Target-word-size relocation. */ \
V(BSSRelocation, kRewind << 1) \
V(Other, kBSSRelocation << 1) \
V(AnyKind, -1)
enum Kind {
#define ENUM_DEF(name, init) k##name = init,
FOR_EACH_RAW_PC_DESCRIPTOR(ENUM_DEF)
#undef ENUM_DEF
kLastKind = kOther,
};
static const char* KindToCString(Kind k);
static bool ParseKind(const char* cstr, Kind* out);
// Used to represent the absence of a yield index in PcDescriptors.
static constexpr intptr_t kInvalidYieldIndex = -1;
class KindAndMetadata {
public:
// Most of the time try_index will be small and merged field will fit into
// one byte.
static uint32_t Encode(intptr_t kind,
intptr_t try_index,
intptr_t yield_index) {
return KindShiftBits::encode(Utils::ShiftForPowerOfTwo(kind)) |
TryIndexBits::encode(try_index + 1) |
YieldIndexBits::encode(yield_index + 1);
}
static intptr_t DecodeKind(uint32_t kind_and_metadata) {
return 1 << KindShiftBits::decode(kind_and_metadata);
}
static intptr_t DecodeTryIndex(uint32_t kind_and_metadata) {
return TryIndexBits::decode(kind_and_metadata) - 1;
}
static intptr_t DecodeYieldIndex(uint32_t kind_and_metadata) {
return YieldIndexBits::decode(kind_and_metadata) - 1;
}
private:
static constexpr intptr_t kKindShiftSize = 3;
static constexpr intptr_t kTryIndexSize = 10;
static constexpr intptr_t kYieldIndexSize =
32 - kKindShiftSize - kTryIndexSize;
class KindShiftBits
: public BitField<uint32_t, intptr_t, 0, kKindShiftSize> {};
class TryIndexBits : public BitField<uint32_t,
intptr_t,
KindShiftBits::kNextBit,
kTryIndexSize> {};
class YieldIndexBits : public BitField<uint32_t,
intptr_t,
TryIndexBits::kNextBit,
kYieldIndexSize> {};
};
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(PcDescriptors);
VISIT_NOTHING();
// Number of descriptors. This only needs to be an int32_t, but we make it a
// uword so that the variable length data is 64 bit aligned on 64 bit
// platforms.
uword length_;
// Variable length data follows here.
uint8_t* data() { OPEN_ARRAY_START(uint8_t, intptr_t); }
const uint8_t* data() const { OPEN_ARRAY_START(uint8_t, intptr_t); }
friend class Object;
friend class ImageWriter;
};
// CodeSourceMap encodes a mapping from code PC ranges to source token
// positions and the stack of inlined functions.
class UntaggedCodeSourceMap : public UntaggedObject {
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(CodeSourceMap);
VISIT_NOTHING();
// Length in bytes. This only needs to be an int32_t, but we make it a uword
// so that the variable length data is 64 bit aligned on 64 bit platforms.
uword length_;
// Variable length data follows here.
uint8_t* data() { OPEN_ARRAY_START(uint8_t, intptr_t); }
const uint8_t* data() const { OPEN_ARRAY_START(uint8_t, intptr_t); }
friend class Object;
friend class ImageWriter;
};
// RawCompressedStackMaps is a compressed representation of the stack maps
// for certain PC offsets into a set of instructions, where a stack map is a bit
// map that marks each live object index starting from the base of the frame.
class UntaggedCompressedStackMaps : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(CompressedStackMaps);
VISIT_NOTHING();
public:
// Note: AOT snapshots pack these structures without any padding in between
// so payload structure should not have any alignment requirements.
// alignas(1) is here to trigger a compiler error if we violate this.
struct alignas(1) Payload {
using FlagsAndSizeHeader = uint32_t;
// The most significant bits are the length of the encoded payload, in
// bytes (excluding the header itself). The low bits determine the
// expected payload contents, as described below.
DART_FORCE_INLINE FlagsAndSizeHeader flags_and_size() const {
// Note: |this| does not necessarily satisfy alignment requirements
// of uint32_t so we should use bit_cast.
return bit_copy<FlagsAndSizeHeader, Payload>(*this);
}
DART_FORCE_INLINE void set_flags_and_size(FlagsAndSizeHeader value) {
// Note: |this| does not necessarily satisfy alignment requirements
// of uint32_t hence the byte copy below.
memcpy(reinterpret_cast<void*>(this), &value, sizeof(value)); // NOLINT
}
// Variable length data follows here. The contents of the payload depend on
// the type of CompressedStackMaps (CSM) being represented. There are three
// major types of CSM:
//
// 1) GlobalTableBit = false, UsesTableBit = false: CSMs that include all
// information about the stack maps. The payload for these contain
// tightly packed entries with the following information:
//
// * A header containing the following three pieces of information:
// * An unsigned integer representing the PC offset as a delta from the
// PC offset of the previous entry (from 0 for the first entry).
// * An unsigned integer representing the number of bits used for
// spill slot entries.
// * An unsigned integer representing the number of bits used for other
// entries.
// * The body containing the bits for the stack map. The length of
// the body in bits is the sum of the spill slot and non-spill slot
// bit counts.
//
// 2) GlobalTableBit = false, UsesTableBit = true: CSMs where the majority
// of the stack map information has been offloaded and canonicalized into
// a global table. The payload contains tightly packed entries with the
// following information:
//
// * A header containing just an unsigned integer representing the PC
// offset delta as described above.
// * The body is just an unsigned integer containing the offset into the
// payload for the global table.
//
// 3) GlobalTableBit = true, UsesTableBit = false: A CSM implementing the
// global table. Here, the payload contains tightly packed entries with
// the following information:
//
// * A header containing the following two pieces of information:
// * An unsigned integer representing the number of bits used for
// spill slot entries.
// * An unsigned integer representing the number of bits used for other
// entries.
// * The body containing the bits for the stack map. The length of the
// body in bits is the sum of the spill slot and non-spill slot bit
// counts.
//
// In all types of CSM, each unsigned integer is LEB128 encoded, as
// generally they tend to fit in a single byte or two. Thus, entry headers
// are not a fixed length, and currently there is no random access of
// entries. In addition, PC offsets are currently encoded as deltas, which
// also inhibits random access without accessing previous entries. That
// means to find an entry for a given PC offset, a linear search must be
// done where the payload is decoded up to the entry whose PC offset
// is greater or equal to the given PC.
uint8_t* data() {
return reinterpret_cast<uint8_t*>(this) + sizeof(FlagsAndSizeHeader);
}
const uint8_t* data() const {
return reinterpret_cast<const uint8_t*>(this) +
sizeof(FlagsAndSizeHeader);
}
};
private:
// We are using OPEN_ARRAY_START rather than embedding Payload directly into
// the UntaggedCompressedStackMaps as a field because that would introduce a
// padding at the end of UntaggedCompressedStackMaps - so we would not be
// able to use sizeof(UntaggedCompressedStackMaps) as the size of the header
// anyway.
Payload* payload() { OPEN_ARRAY_START(Payload, uint8_t); }
const Payload* payload() const { OPEN_ARRAY_START(Payload, uint8_t); }
class GlobalTableBit
: public BitField<Payload::FlagsAndSizeHeader, bool, 0, 1> {};
class UsesTableBit : public BitField<Payload::FlagsAndSizeHeader,
bool,
GlobalTableBit::kNextBit,
1> {};
class SizeField
: public BitField<Payload::FlagsAndSizeHeader,
Payload::FlagsAndSizeHeader,
UsesTableBit::kNextBit,
sizeof(Payload::FlagsAndSizeHeader) * kBitsPerByte -
UsesTableBit::kNextBit> {};
friend class Object;
friend class ImageWriter;
friend class StackMapEntry;
};
class UntaggedInstructionsTable : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(InstructionsTable);
POINTER_FIELD(ArrayPtr, code_objects)
VISIT_FROM(code_objects)
VISIT_TO(code_objects)
struct DataEntry {
uint32_t pc_offset;
uint32_t stack_map_offset;
};
static_assert(sizeof(DataEntry) == sizeof(uint32_t) * 2);
struct Data {
uint32_t canonical_stack_map_entries_offset;
uint32_t length;
uint32_t first_entry_with_code;
uint32_t padding;
const DataEntry* entries() const { OPEN_ARRAY_START(DataEntry, uint32_t); }
const UntaggedCompressedStackMaps::Payload* StackMapAt(
intptr_t offset) const {
return reinterpret_cast<UntaggedCompressedStackMaps::Payload*>(
reinterpret_cast<uword>(this) + offset);
}
};
static_assert(sizeof(Data) == sizeof(uint32_t) * 4);
intptr_t length_;
const Data* rodata_;
uword start_pc_;
uword end_pc_;
friend class Deserializer;
};
class UntaggedLocalVarDescriptors : public UntaggedObject {
public:
enum VarInfoKind {
kStackVar = 1,
kContextVar,
kContextLevel,
kSavedCurrentContext,
};
enum {
kKindPos = 0,
kKindSize = 8,
kIndexPos = kKindPos + kKindSize,
// Since there are 24 bits for the stack slot index, Functions can have
// only ~16.7 million stack slots.
kPayloadSize = sizeof(int32_t) * kBitsPerByte,
kIndexSize = kPayloadSize - kIndexPos,
kIndexBias = 1 << (kIndexSize - 1),
kMaxIndex = (1 << (kIndexSize - 1)) - 1,
};
class IndexBits : public BitField<int32_t, int32_t, kIndexPos, kIndexSize> {};
class KindBits : public BitField<int32_t, int8_t, kKindPos, kKindSize> {};
struct VarInfo {
int32_t index_kind = 0; // Bitfield for slot index on stack or in context,
// and Entry kind of type VarInfoKind.
TokenPosition declaration_pos =
TokenPosition::kNoSource; // Token position of declaration.
TokenPosition begin_pos =
TokenPosition::kNoSource; // Token position of scope start.
TokenPosition end_pos =
TokenPosition::kNoSource; // Token position of scope end.
int16_t scope_id; // Scope to which the variable belongs.
VarInfoKind kind() const {
return static_cast<VarInfoKind>(KindBits::decode(index_kind));
}
void set_kind(VarInfoKind kind) {
index_kind = KindBits::update(kind, index_kind);
}
int32_t index() const { return IndexBits::decode(index_kind) - kIndexBias; }
void set_index(int32_t index) {
index_kind = IndexBits::update(index + kIndexBias, index_kind);
}
};
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(LocalVarDescriptors);
// Number of descriptors. This only needs to be an int32_t, but we make it a
// uword so that the variable length data is 64 bit aligned on 64 bit
// platforms.
uword num_entries_;
VISIT_FROM_PAYLOAD_START(CompressedStringPtr)
COMPRESSED_VARIABLE_POINTER_FIELDS(StringPtr, name, names)
CompressedStringPtr* nameAddrAt(intptr_t i) { return &(names()[i]); }
void set_name(intptr_t i, StringPtr value) {
StoreCompressedPointer(nameAddrAt(i), value);
}
// Variable info with [num_entries_] entries.
VarInfo* data() {
return reinterpret_cast<VarInfo*>(nameAddrAt(num_entries_));
}
friend class Object;
};
class UntaggedExceptionHandlers : public UntaggedObject {
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(ExceptionHandlers);
// Number of exception handler entries and
// async handler.
uint32_t packed_fields_;
// Async handler is used in the async/async* functions.
// It's an implicit exception handler (stub) which runs when
// exception is not handled within the function.
using AsyncHandlerBit = BitField<decltype(packed_fields_), bool, 0, 1>;
using NumEntriesBits = BitField<decltype(packed_fields_),
uint32_t,
AsyncHandlerBit::kNextBit,
31>;
intptr_t num_entries() const {
return NumEntriesBits::decode(packed_fields_);
}
// Array with [num_entries] entries. Each entry is an array of all handled
// exception types.
COMPRESSED_POINTER_FIELD(ArrayPtr, handled_types_data)
VISIT_FROM(handled_types_data)
VISIT_TO(handled_types_data)
// Exception handler info of length [num_entries].
const ExceptionHandlerInfo* data() const {
OPEN_ARRAY_START(ExceptionHandlerInfo, intptr_t);
}
ExceptionHandlerInfo* data() {
OPEN_ARRAY_START(ExceptionHandlerInfo, intptr_t);
}
friend class Object;
};
class UntaggedContext : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(Context);
int32_t num_variables_;
COMPRESSED_POINTER_FIELD(ContextPtr, parent)
VISIT_FROM(parent)
// Variable length data follows here.
COMPRESSED_VARIABLE_POINTER_FIELDS(ObjectPtr, element, data)
friend class Object;
friend void UpdateLengthField(intptr_t,
ObjectPtr,
ObjectPtr); // num_variables_
};
#define CONTEXT_SCOPE_VARIABLE_DESC_FLAG_LIST(V) \
V(Final) \
V(Late) \
V(Nullable) \
V(Invisible) \
V(AwaiterLink)
class UntaggedContextScope : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(ContextScope);
// TODO(iposva): Switch to conventional enum offset based structure to avoid
// alignment mishaps.
struct VariableDesc {
CompressedSmiPtr declaration_token_pos;
CompressedSmiPtr token_pos;
CompressedStringPtr name;
CompressedSmiPtr flags;
enum FlagBits {
#define DECLARE_BIT(Name) kIs##Name,
CONTEXT_SCOPE_VARIABLE_DESC_FLAG_LIST(DECLARE_BIT)
#undef DECLARE_BIT
};
CompressedSmiPtr late_init_offset;
CompressedAbstractTypePtr type;
CompressedSmiPtr cid;
CompressedSmiPtr context_index;
CompressedSmiPtr context_level;
CompressedSmiPtr kernel_offset;
};
int32_t num_variables_;
bool is_implicit_; // true, if this context scope is for an implicit closure.
// Just choose one of the fields in VariableDesc, since they should all be
// compressed or not compressed.
DEFINE_CONTAINS_COMPRESSED(decltype(VariableDesc::name));
CompressedObjectPtr* from() {
VariableDesc* begin = const_cast<VariableDesc*>(VariableDescAddr(0));
return reinterpret_cast<CompressedObjectPtr*>(begin);
}
// Variable length data follows here.
CompressedObjectPtr const* data() const {
OPEN_ARRAY_START(CompressedObjectPtr, CompressedObjectPtr);
}
const VariableDesc* VariableDescAddr(intptr_t index) const {
// data() points to the first component of the first descriptor.
return reinterpret_cast<const VariableDesc*>(data()) + index;
}
#define DEFINE_ACCESSOR(type, name) \
type name##_at(intptr_t index) { \
return LoadCompressedPointer<type>(&VariableDescAddr(index)->name); \
} \
void set_##name##_at(intptr_t index, type value) { \
StoreCompressedPointer(&VariableDescAddr(index)->name, value); \
}
DEFINE_ACCESSOR(SmiPtr, declaration_token_pos)
DEFINE_ACCESSOR(SmiPtr, token_pos)
DEFINE_ACCESSOR(StringPtr, name)
DEFINE_ACCESSOR(SmiPtr, flags)
DEFINE_ACCESSOR(SmiPtr, late_init_offset)
DEFINE_ACCESSOR(AbstractTypePtr, type)
DEFINE_ACCESSOR(SmiPtr, cid)
DEFINE_ACCESSOR(SmiPtr, context_index)
DEFINE_ACCESSOR(SmiPtr, context_level)
DEFINE_ACCESSOR(SmiPtr, kernel_offset)
#undef DEFINE_ACCESSOR
CompressedObjectPtr* to(intptr_t num_vars) {
uword end = reinterpret_cast<uword>(VariableDescAddr(num_vars));
// 'end' is the address just beyond the last descriptor, so step back.
return reinterpret_cast<CompressedObjectPtr*>(end -
sizeof(CompressedObjectPtr));
}
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind, intptr_t num_vars) {
return to(num_vars);
}
friend class Object;
friend class UntaggedClosureData;
};
class UntaggedSentinel : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(Sentinel);
VISIT_NOTHING();
};
class UntaggedSingleTargetCache : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(SingleTargetCache);
POINTER_FIELD(CodePtr, target)
VISIT_FROM(target)
VISIT_TO(target)
uword entry_point_;
ClassIdTagType lower_limit_;
ClassIdTagType upper_limit_;
};
class UntaggedMonomorphicSmiableCall : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(MonomorphicSmiableCall);
VISIT_NOTHING();
uword expected_cid_;
uword entrypoint_;
};
// Abstract base class for RawICData/RawMegamorphicCache
class UntaggedCallSiteData : public UntaggedObject {
protected:
POINTER_FIELD(StringPtr, target_name); // Name of target function.
VISIT_FROM(target_name)
// arg_descriptor in RawICData and in RawMegamorphicCache should be
// in the same position so that NoSuchMethod can access it.
POINTER_FIELD(ArrayPtr, args_descriptor); // Arguments descriptor.
VISIT_TO(args_descriptor)
ObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(CallSiteData)
};
class UntaggedUnlinkedCall : public UntaggedCallSiteData {
RAW_HEAP_OBJECT_IMPLEMENTATION(UnlinkedCall);
bool can_patch_to_monomorphic_;
};
class UntaggedICData : public UntaggedCallSiteData {
RAW_HEAP_OBJECT_IMPLEMENTATION(ICData);
POINTER_FIELD(ArrayPtr, entries) // Contains class-ids, target and count.
// Static type of the receiver, if instance call and available.
NOT_IN_PRECOMPILED(POINTER_FIELD(AbstractTypePtr, receivers_static_type))
POINTER_FIELD(ObjectPtr,
owner) // Parent/calling function or original IC of cloned IC.
VISIT_TO(owner)
ObjectPtr* to_snapshot(Snapshot::Kind kind) {
switch (kind) {
case Snapshot::kFullAOT:
return reinterpret_cast<ObjectPtr*>(&entries_);
case Snapshot::kFull:
case Snapshot::kFullCore:
case Snapshot::kFullJIT:
return to();
case Snapshot::kNone:
case Snapshot::kInvalid:
break;
}
UNREACHABLE();
return nullptr;
}
NOT_IN_PRECOMPILED(int32_t deopt_id_);
// Number of arguments tested in IC, deopt reasons.
AtomicBitFieldContainer<uint32_t> state_bits_;
};
class UntaggedMegamorphicCache : public UntaggedCallSiteData {
RAW_HEAP_OBJECT_IMPLEMENTATION(MegamorphicCache);
POINTER_FIELD(ArrayPtr, buckets)
SMI_FIELD(SmiPtr, mask)
VISIT_TO(mask)
ObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
int32_t filled_entry_count_;
};
class UntaggedSubtypeTestCache : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(SubtypeTestCache);
POINTER_FIELD(ArrayPtr, cache)
VISIT_FROM(cache)
VISIT_TO(cache)
uint32_t num_inputs_;
uint32_t num_occupied_;
};
class UntaggedLoadingUnit : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(LoadingUnit);
COMPRESSED_POINTER_FIELD(LoadingUnitPtr, parent)
VISIT_FROM(parent)
COMPRESSED_POINTER_FIELD(ArrayPtr, base_objects)
VISIT_TO(base_objects)
int32_t id_;
bool load_outstanding_;
bool loaded_;
};
class UntaggedError : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(Error);
};
class UntaggedApiError : public UntaggedError {
RAW_HEAP_OBJECT_IMPLEMENTATION(ApiError);
COMPRESSED_POINTER_FIELD(StringPtr, message)
VISIT_FROM(message)
VISIT_TO(message)
};
class UntaggedLanguageError : public UntaggedError {
RAW_HEAP_OBJECT_IMPLEMENTATION(LanguageError);
COMPRESSED_POINTER_FIELD(ErrorPtr, previous_error) // May be null.
VISIT_FROM(previous_error)
COMPRESSED_POINTER_FIELD(ScriptPtr, script)
COMPRESSED_POINTER_FIELD(StringPtr, message)
// Incl. previous error's formatted message.
COMPRESSED_POINTER_FIELD(StringPtr, formatted_message)
VISIT_TO(formatted_message)
TokenPosition token_pos_; // Source position in script_.
bool report_after_token_; // Report message at or after the token.
int8_t kind_; // Of type Report::Kind.
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
};
class UntaggedUnhandledException : public UntaggedError {
RAW_HEAP_OBJECT_IMPLEMENTATION(UnhandledException);
COMPRESSED_POINTER_FIELD(InstancePtr, exception)
VISIT_FROM(exception)
COMPRESSED_POINTER_FIELD(InstancePtr, stacktrace)
VISIT_TO(stacktrace)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
};
class UntaggedUnwindError : public UntaggedError {
RAW_HEAP_OBJECT_IMPLEMENTATION(UnwindError);
COMPRESSED_POINTER_FIELD(StringPtr, message)
VISIT_FROM(message)
VISIT_TO(message)
bool is_user_initiated_;
};
class UntaggedInstance : public UntaggedObject {
RAW_HEAP_OBJECT_IMPLEMENTATION(Instance);
friend class Object;
public:
#if defined(DART_COMPRESSED_POINTERS)
static constexpr bool kContainsCompressedPointers = true;
#else
static constexpr bool kContainsCompressedPointers = false;
#endif
};
class UntaggedLibraryPrefix : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(LibraryPrefix);
// Library prefix name.
COMPRESSED_POINTER_FIELD(StringPtr, name)
VISIT_FROM(name)
// Libraries imported with this prefix.
COMPRESSED_POINTER_FIELD(ArrayPtr, imports)
// Library which declares this prefix.
COMPRESSED_POINTER_FIELD(LibraryPtr, importer)
VISIT_TO(importer)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) {
switch (kind) {
case Snapshot::kFullAOT:
return reinterpret_cast<CompressedObjectPtr*>(&imports_);
case Snapshot::kFull:
case Snapshot::kFullCore:
case Snapshot::kFullJIT:
return reinterpret_cast<CompressedObjectPtr*>(&importer_);
case Snapshot::kNone:
case Snapshot::kInvalid:
break;
}
UNREACHABLE();
return nullptr;
}
uint16_t num_imports_; // Number of library entries in libraries_.
bool is_deferred_load_;
};
class UntaggedTypeArguments : public UntaggedInstance {
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(TypeArguments);
// The instantiations_ array remains empty for instantiated type arguments.
// Of 3-tuple: 2 instantiators, result.
COMPRESSED_POINTER_FIELD(ArrayPtr, instantiations)
VISIT_FROM(instantiations)
COMPRESSED_SMI_FIELD(SmiPtr, length)
COMPRESSED_SMI_FIELD(SmiPtr, hash)
COMPRESSED_SMI_FIELD(SmiPtr, nullability)
// Variable length data follows here.
COMPRESSED_VARIABLE_POINTER_FIELDS(AbstractTypePtr, element, types)
friend class Object;
};
class UntaggedTypeParameters : public UntaggedObject {
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(TypeParameters);
// Length of names reflects the number of type parameters.
COMPRESSED_POINTER_FIELD(ArrayPtr, names)
VISIT_FROM(names)
// flags: isGenericCovariantImpl and (todo) variance.
COMPRESSED_POINTER_FIELD(ArrayPtr, flags)
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr, bounds)
// defaults is the instantiation to bounds (calculated by CFE).
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr, defaults)
VISIT_TO(defaults)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
friend class Object;
};
class UntaggedAbstractType : public UntaggedInstance {
protected:
// Accessed from generated code.
std::atomic<uword> type_test_stub_entry_point_;
// Accessed from generated code.
std::atomic<uint32_t> flags_;
#if defined(DART_COMPRESSED_POINTERS)
uint32_t padding_; // Makes Windows and Posix agree on layout.
#endif
COMPRESSED_POINTER_FIELD(CodePtr, type_test_stub)
COMPRESSED_POINTER_FIELD(SmiPtr, hash)
VISIT_FROM(type_test_stub)
uint32_t flags() const { return flags_.load(std::memory_order_relaxed); }
void set_flags(uint32_t value) {
flags_.store(value, std::memory_order_relaxed);
}
public:
enum TypeState {
kAllocated, // Initial state.
kFinalizedInstantiated, // Instantiated type ready for use.
kFinalizedUninstantiated, // Uninstantiated type ready for use.
};
using NullabilityBits = BitField<uint32_t, uint8_t, 0, 2>;
static constexpr intptr_t kNullabilityMask = NullabilityBits::mask();
static constexpr intptr_t kTypeStateShift = NullabilityBits::kNextBit;
static constexpr intptr_t kTypeStateBits = 2;
using TypeStateBits =
BitField<uint32_t, uint8_t, kTypeStateShift, kTypeStateBits>;
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(AbstractType);
friend class ObjectStore;
friend class StubCode;
};
class UntaggedType : public UntaggedAbstractType {
public:
static constexpr intptr_t kTypeClassIdShift = TypeStateBits::kNextBit;
using TypeClassIdBits =
BitField<uint32_t, ClassIdTagType, kTypeClassIdShift, kClassIdTagSize>;
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(Type);
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr, arguments)
VISIT_TO(arguments)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
ClassIdTagType type_class_id() const {
return TypeClassIdBits::decode(flags());
}
void set_type_class_id(ClassIdTagType value) {
set_flags(TypeClassIdBits::update(value, flags()));
}
friend class compiler::target::UntaggedType;
friend class CidRewriteVisitor;
friend class UntaggedTypeArguments;
};
class UntaggedFunctionType : public UntaggedAbstractType {
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(FunctionType);
COMPRESSED_POINTER_FIELD(TypeParametersPtr, type_parameters)
COMPRESSED_POINTER_FIELD(AbstractTypePtr, result_type)
COMPRESSED_POINTER_FIELD(ArrayPtr, parameter_types)
COMPRESSED_POINTER_FIELD(ArrayPtr, named_parameter_names);
VISIT_TO(named_parameter_names)
AtomicBitFieldContainer<uint32_t> packed_parameter_counts_;
AtomicBitFieldContainer<uint16_t> packed_type_parameter_counts_;
// The bit fields are public for use in kernel_to_il.cc.
public:
// For packed_type_parameter_counts_.
using PackedNumParentTypeArguments =
BitField<decltype(packed_type_parameter_counts_), uint8_t, 0, 8>;
using PackedNumTypeParameters =
BitField<decltype(packed_type_parameter_counts_),
uint8_t,
PackedNumParentTypeArguments::kNextBit,
8>;
// For packed_parameter_counts_.
using PackedNumImplicitParameters =
BitField<decltype(packed_parameter_counts_), uint8_t, 0, 1>;
using PackedHasNamedOptionalParameters =
BitField<decltype(packed_parameter_counts_),
bool,
PackedNumImplicitParameters::kNextBit,
1>;
using PackedNumFixedParameters =
BitField<decltype(packed_parameter_counts_),
uint16_t,
PackedHasNamedOptionalParameters::kNextBit,
14>;
using PackedNumOptionalParameters =
BitField<decltype(packed_parameter_counts_),
uint16_t,
PackedNumFixedParameters::kNextBit,
14>;
static_assert(PackedNumOptionalParameters::kNextBit <=
compiler::target::kSmiBits,
"In-place mask for number of optional parameters cannot fit in "
"a Smi on the target architecture");
private:
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
friend class Function;
};
class UntaggedRecordType : public UntaggedAbstractType {
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(RecordType);
COMPRESSED_SMI_FIELD(SmiPtr, shape)
COMPRESSED_POINTER_FIELD(ArrayPtr, field_types)
VISIT_TO(field_types)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
};
class UntaggedTypeParameter : public UntaggedAbstractType {
public:
static constexpr intptr_t kIsFunctionTypeParameterBit =
TypeStateBits::kNextBit;
using IsFunctionTypeParameter =
BitField<uint32_t, bool, kIsFunctionTypeParameterBit, 1>;
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(TypeParameter);
// FunctionType or Smi (class id).
COMPRESSED_POINTER_FIELD(ObjectPtr, owner)
VISIT_TO(owner)
uint16_t base_; // Number of enclosing function type parameters.
uint16_t index_; // Keep size in sync with BuildTypeParameterTypeTestStub.
private:
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
friend class CidRewriteVisitor;
};
class UntaggedClosure : public UntaggedInstance {
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(Closure);
// No instance fields should be declared before the following fields whose
// offsets must be identical in Dart and C++.
// The following fields are also declared in the Dart source of class
// _Closure.
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr, instantiator_type_arguments)
VISIT_FROM(instantiator_type_arguments)
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr, function_type_arguments)
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr, delayed_type_arguments)
COMPRESSED_POINTER_FIELD(FunctionPtr, function)
COMPRESSED_POINTER_FIELD(ContextPtr, context)
COMPRESSED_POINTER_FIELD(SmiPtr, hash)
VISIT_TO(hash)
// We have an extra word in the object due to alignment rounding, so use it in
// bare instructions mode to cache the entry point from the closure function
// to avoid an extra redirection on call. Closure functions only have
// one entry point, as dynamic calls use dynamic closure call dispatchers.
ONLY_IN_PRECOMPILED(uword entry_point_);
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
// Note that instantiator_type_arguments_, function_type_arguments_ and
// delayed_type_arguments_ are used to instantiate the signature of function_
// when this closure is involved in a type test. In other words, these fields
// define the function type of this closure instance.
//
// function_type_arguments_ and delayed_type_arguments_ may also be used when
// invoking the closure. Whereas the source frontend will save a copy of the
// function's type arguments in the closure's context and only use the
// function_type_arguments_ field for type tests, the kernel frontend will use
// the function_type_arguments_ vector here directly.
//
// If this closure is generic, it can be invoked with function type arguments
// that will be processed in the prolog of the closure function_. For example,
// if the generic closure function_ has a generic parent function, the
// passed-in function type arguments get concatenated to the function type
// arguments of the parent that are found in the context_.
//
// delayed_type_arguments_ is used to support the partial instantiation
// feature. When this field is set to any value other than
// Object::empty_type_arguments(), the types in this vector will be passed as
// type arguments to the closure when invoked. In this case there may not be
// any type arguments passed directly (or NSM will be invoked instead).
friend class UnitDeserializationRoots;
};
class UntaggedNumber : public UntaggedInstance {
RAW_OBJECT_IMPLEMENTATION(Number);
};
class UntaggedInteger : public UntaggedNumber {
RAW_OBJECT_IMPLEMENTATION(Integer);
};
class UntaggedSmi : public UntaggedInteger {
RAW_OBJECT_IMPLEMENTATION(Smi);
};
class UntaggedMint : public UntaggedInteger {
RAW_HEAP_OBJECT_IMPLEMENTATION(Mint);
VISIT_NOTHING();
ALIGN8 int64_t value_;
friend class Api;
friend class Class;
friend class Integer;
};
COMPILE_ASSERT(sizeof(UntaggedMint) == 16);
class UntaggedDouble : public UntaggedNumber {
RAW_HEAP_OBJECT_IMPLEMENTATION(Double);
VISIT_NOTHING();
ALIGN8 double value_;
friend class Api;
friend class Class;
};
COMPILE_ASSERT(sizeof(UntaggedDouble) == 16);
class UntaggedString : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(String);
protected:
#if !defined(HASH_IN_OBJECT_HEADER)
COMPRESSED_SMI_FIELD(SmiPtr, hash)
VISIT_FROM(hash)
#endif
COMPRESSED_SMI_FIELD(SmiPtr, length)
#if defined(HASH_IN_OBJECT_HEADER)
VISIT_FROM(length)
#endif
VISIT_TO(length)
private:
friend class Library;
friend class RODataSerializationCluster;
friend class ImageWriter;
};
class UntaggedOneByteString : public UntaggedString {
RAW_HEAP_OBJECT_IMPLEMENTATION(OneByteString);
VISIT_NOTHING();
// Variable length data follows here.
uint8_t* data() { OPEN_ARRAY_START(uint8_t, uint8_t); }
const uint8_t* data() const { OPEN_ARRAY_START(uint8_t, uint8_t); }
friend class RODataSerializationCluster;
friend class String;
friend class StringDeserializationCluster;
friend class StringSerializationCluster;
};
class UntaggedTwoByteString : public UntaggedString {
RAW_HEAP_OBJECT_IMPLEMENTATION(TwoByteString);
VISIT_NOTHING();
// Variable length data follows here.
uint16_t* data() { OPEN_ARRAY_START(uint16_t, uint16_t); }
const uint16_t* data() const { OPEN_ARRAY_START(uint16_t, uint16_t); }
friend class RODataSerializationCluster;
friend class String;
friend class StringDeserializationCluster;
friend class StringSerializationCluster;
};
// Abstract base class for UntaggedTypedData/UntaggedExternalTypedData/
// UntaggedTypedDataView/Pointer.
//
// TypedData extends this with a length field, while Pointer extends this with
// TypeArguments field.
class UntaggedPointerBase : public UntaggedInstance {
public:
uint8_t* data() { return data_; }
protected:
// The contents of [data_] depends on what concrete subclass is used:
//
// - UntaggedTypedData: Start of the payload.
// - UntaggedExternalTypedData: Start of the C-heap payload.
// - UntaggedTypedDataView: The [data_] field of the backing store for the
// view plus the [offset_in_bytes_] the view has.
// - UntaggedPointer: Pointer into C memory (no length specified).
//
// During allocation or snapshot reading the [data_] can be temporarily
// nullptr (which is the case for views which just got created but haven't
// gotten the backing store set).
uint8_t* data_;
private:
template <typename T>
friend void CopyTypedDataBaseWithSafepointChecks(
Thread*,
const T&,
const T&,
intptr_t); // Access _data for memmove with safepoint checkins.
RAW_HEAP_OBJECT_IMPLEMENTATION(PointerBase);
};
// Abstract base class for UntaggedTypedData/UntaggedExternalTypedData/
// UntaggedTypedDataView.
class UntaggedTypedDataBase : public UntaggedPointerBase {
protected:
#if defined(DART_COMPRESSED_POINTERS)
uint32_t padding_; // Makes Windows and Posix agree on layout.
#endif
// The length of the view in element sizes (obtainable via
// [TypedDataBase::ElementSizeInBytes]).
COMPRESSED_SMI_FIELD(SmiPtr, length);
VISIT_FROM(length)
VISIT_TO(length)
private:
friend class UntaggedTypedDataView;
friend void UpdateLengthField(intptr_t, ObjectPtr, ObjectPtr); // length_
friend void InitializeExternalTypedData(
intptr_t,
ExternalTypedDataPtr,
ExternalTypedDataPtr); // initialize fields.
friend void InitializeExternalTypedDataWithSafepointChecks(
Thread*,
intptr_t,
const ExternalTypedData&,
const ExternalTypedData&); // initialize fields.
RAW_HEAP_OBJECT_IMPLEMENTATION(TypedDataBase);
};
class UntaggedTypedData : public UntaggedTypedDataBase {
RAW_HEAP_OBJECT_IMPLEMENTATION(TypedData);
public:
static intptr_t payload_offset() {
return OFFSET_OF_RETURNED_VALUE(UntaggedTypedData, internal_data);
}
// Recompute [data_] pointer to internal data.
void RecomputeDataField() { data_ = internal_data(); }
protected:
// Variable length data follows here.
uint8_t* internal_data() { OPEN_ARRAY_START(uint8_t, uint8_t); }
const uint8_t* internal_data() const { OPEN_ARRAY_START(uint8_t, uint8_t); }
uint8_t* data() {
ASSERT(data_ == internal_data());
return data_;
}
const uint8_t* data() const {
ASSERT(data_ == internal_data());
return data_;
}
friend class Api;
friend class Instance;
friend class DeltaEncodedTypedDataDeserializationCluster;
friend class NativeEntryData;
friend class Object;
friend class ObjectPool;
friend class ObjectPoolDeserializationCluster;
friend class ObjectPoolSerializationCluster;
friend class UntaggedObjectPool;
};
// All _*ArrayView/_ByteDataView classes share the same layout.
class UntaggedTypedDataView : public UntaggedTypedDataBase {
RAW_HEAP_OBJECT_IMPLEMENTATION(TypedDataView);
public:
// Recompute [data_] based on internal/external [typed_data_].
void RecomputeDataField() {
const intptr_t offset_in_bytes = RawSmiValue(this->offset_in_bytes());
uint8_t* payload = typed_data()->untag()->data_;
data_ = payload + offset_in_bytes;
}
// Recompute [data_] based on internal [typed_data_] - needs to be called by
// GC whenever the backing store moved.
//
// NOTICE: This method assumes [this] is the forwarded object and the
// [typed_data_] pointer points to the new backing store. The backing store's
// fields don't need to be valid - only it's address.
void RecomputeDataFieldForInternalTypedData() {
data_ = DataFieldForInternalTypedData();
}
uint8_t* DataFieldForInternalTypedData() const {
const intptr_t offset_in_bytes = RawSmiValue(this->offset_in_bytes());
uint8_t* payload =
reinterpret_cast<uint8_t*>(UntaggedObject::ToAddr(typed_data()) +
UntaggedTypedData::payload_offset());
return payload + offset_in_bytes;
}
void ValidateInnerPointer() {
if (typed_data()->untag()->GetClassId() == kNullCid) {
// The view object must have gotten just initialized.
if (data_ != nullptr || RawSmiValue(offset_in_bytes()) != 0 ||
RawSmiValue(length()) != 0) {
FATAL("TypedDataView has invalid inner pointer.");
}
} else {
const intptr_t offset_in_bytes = RawSmiValue(this->offset_in_bytes());
uint8_t* payload = typed_data()->untag()->data_;
if ((payload + offset_in_bytes) != data_) {
FATAL("TypedDataView has invalid inner pointer.");
}
}
}
protected:
COMPRESSED_POINTER_FIELD(TypedDataBasePtr, typed_data)
COMPRESSED_SMI_FIELD(SmiPtr, offset_in_bytes)
VISIT_TO(offset_in_bytes)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
friend void InitializeTypedDataView(TypedDataViewPtr);
friend class Api;
friend class Object;
friend class ObjectPoolDeserializationCluster;
friend class ObjectPoolSerializationCluster;
friend class UntaggedObjectPool;
friend class GCCompactor;
template <bool>
friend class ScavengerVisitorBase;
};
class UntaggedExternalOneByteString : public UntaggedString {
RAW_HEAP_OBJECT_IMPLEMENTATION(ExternalOneByteString);
const uint8_t* external_data_;
void* peer_;
friend class Api;
friend class String;
};
class UntaggedExternalTwoByteString : public UntaggedString {
RAW_HEAP_OBJECT_IMPLEMENTATION(ExternalTwoByteString);
const uint16_t* external_data_;
void* peer_;
friend class Api;
friend class String;
};
class UntaggedBool : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(Bool);
VISIT_NOTHING();
bool value_;
friend class Object;
};
class UntaggedArray : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(Array);
COMPRESSED_ARRAY_POINTER_FIELD(TypeArgumentsPtr, type_arguments)
VISIT_FROM(type_arguments)
COMPRESSED_SMI_FIELD(SmiPtr, length)
// Variable length data follows here.
COMPRESSED_VARIABLE_POINTER_FIELDS(ObjectPtr, element, data)
friend class MapSerializationCluster;
friend class MapDeserializationCluster;
friend class SetSerializationCluster;
friend class SetDeserializationCluster;
friend class CodeSerializationCluster;
friend class CodeDeserializationCluster;
friend class Deserializer;
friend class UntaggedCode;
friend class UntaggedImmutableArray;
friend class GrowableObjectArray;
friend class Map;
friend class UntaggedMap;
friend class UntaggedConstMap;
friend class Object;
friend class ICData; // For high performance access.
friend class SubtypeTestCache; // For high performance access.
friend class ReversePc;
template <typename Table, bool kAllCanonicalObjectsAreIncludedIntoSet>
friend class CanonicalSetDeserializationCluster;
friend class Page;
friend class FastObjectCopy; // For initializing fields.
friend void UpdateLengthField(intptr_t, ObjectPtr, ObjectPtr); // length_
};
class UntaggedImmutableArray : public UntaggedArray {
RAW_HEAP_OBJECT_IMPLEMENTATION(ImmutableArray);
};
class UntaggedGrowableObjectArray : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(GrowableObjectArray);
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr, type_arguments)
VISIT_FROM(type_arguments)
COMPRESSED_SMI_FIELD(SmiPtr, length)
COMPRESSED_POINTER_FIELD(ArrayPtr, data)
VISIT_TO(data)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
friend class ReversePc;
};
class UntaggedLinkedHashBase : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(LinkedHashBase);
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr, type_arguments)
VISIT_FROM(type_arguments)
COMPRESSED_POINTER_FIELD(SmiPtr, hash_mask)
COMPRESSED_POINTER_FIELD(ArrayPtr, data)
COMPRESSED_POINTER_FIELD(SmiPtr, used_data)
COMPRESSED_POINTER_FIELD(SmiPtr, deleted_keys)
COMPRESSED_POINTER_FIELD(TypedDataPtr, index)
VISIT_TO(index)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) {
// Do not serialize index.
return reinterpret_cast<CompressedObjectPtr*>(&deleted_keys_);
}
};
class UntaggedMap : public UntaggedLinkedHashBase {
RAW_HEAP_OBJECT_IMPLEMENTATION(Map);
friend class UntaggedConstMap;
};
class UntaggedConstMap : public UntaggedMap {
RAW_HEAP_OBJECT_IMPLEMENTATION(ConstMap);
};
class UntaggedSet : public UntaggedLinkedHashBase {
RAW_HEAP_OBJECT_IMPLEMENTATION(Set);
friend class UntaggedConstSet;
};
class UntaggedConstSet : public UntaggedSet {
RAW_HEAP_OBJECT_IMPLEMENTATION(ConstSet);
};
class UntaggedFloat32x4 : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(Float32x4);
VISIT_NOTHING();
ALIGN8 float value_[4];
friend class Class;
public:
float x() const { return value_[0]; }
float y() const { return value_[1]; }
float z() const { return value_[2]; }
float w() const { return value_[3]; }
};
COMPILE_ASSERT(sizeof(UntaggedFloat32x4) == 24);
class UntaggedInt32x4 : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(Int32x4);
VISIT_NOTHING();
ALIGN8 int32_t value_[4];
friend class Simd128MessageSerializationCluster;
friend class Simd128MessageDeserializationCluster;
public:
int32_t x() const { return value_[0]; }
int32_t y() const { return value_[1]; }
int32_t z() const { return value_[2]; }
int32_t w() const { return value_[3]; }
};
COMPILE_ASSERT(sizeof(UntaggedInt32x4) == 24);
class UntaggedFloat64x2 : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(Float64x2);
VISIT_NOTHING();
ALIGN8 double value_[2];
friend class Class;
public:
double x() const { return value_[0]; }
double y() const { return value_[1]; }
};
COMPILE_ASSERT(sizeof(UntaggedFloat64x2) == 24);
class UntaggedRecord : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(Record);
#if defined(DART_COMPRESSED_POINTERS)
// This explicit padding avoids implicit padding between [shape] and [data].
// Record allocation doesn't initialize the implicit padding but GC scans
// everything between 'from' (shape) and 'to' (end of data),
// so it would see garbage if implicit padding is inserted.
uint32_t padding_;
#endif
COMPRESSED_SMI_FIELD(SmiPtr, shape)
VISIT_FROM(shape)
// Variable length data follows here.
COMPRESSED_VARIABLE_POINTER_FIELDS(ObjectPtr, field, data)
friend void UpdateLengthField(intptr_t, ObjectPtr,
ObjectPtr); // shape_
};
// Define an aliases for intptr_t.
#if defined(ARCH_IS_32_BIT)
#define kIntPtrCid kTypedDataInt32ArrayCid
#define GetIntPtr GetInt32
#define SetIntPtr SetInt32
#define kUintPtrCid kTypedDataUint32ArrayCid
#define GetUintPtr GetUint32
#define SetUintPtr SetUint32
#elif defined(ARCH_IS_64_BIT)
#define kIntPtrCid kTypedDataInt64ArrayCid
#define GetIntPtr GetInt64
#define SetIntPtr SetInt64
#define kUintPtrCid kTypedDataUint64ArrayCid
#define GetUintPtr GetUint64
#define SetUintPtr SetUint64
#else
#error Architecture is not 32-bit or 64-bit.
#endif // ARCH_IS_32_BIT
class UntaggedExternalTypedData : public UntaggedTypedDataBase {
RAW_HEAP_OBJECT_IMPLEMENTATION(ExternalTypedData);
};
class UntaggedPointer : public UntaggedPointerBase {
RAW_HEAP_OBJECT_IMPLEMENTATION(Pointer);
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr, type_arguments)
VISIT_FROM(type_arguments)
VISIT_TO(type_arguments)
friend class Pointer;
};
class UntaggedDynamicLibrary : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(DynamicLibrary);
VISIT_NOTHING();
void* handle_;
bool isClosed_;
bool canBeClosed_;
friend class DynamicLibrary;
};
// VM implementations of the basic types in the isolate.
class alignas(8) UntaggedCapability : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(Capability);
VISIT_NOTHING();
uint64_t id_;
};
class alignas(8) UntaggedSendPort : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(SendPort);
VISIT_NOTHING();
Dart_Port id_;
Dart_Port origin_id_;
friend class ReceivePort;
};
class UntaggedReceivePort : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(ReceivePort);
COMPRESSED_POINTER_FIELD(SendPortPtr, send_port)
VISIT_FROM(send_port)
COMPRESSED_POINTER_FIELD(SmiPtr, bitfield)
COMPRESSED_POINTER_FIELD(InstancePtr, handler)
#if defined(PRODUCT)
VISIT_TO(handler)
#else
COMPRESSED_POINTER_FIELD(StringPtr, debug_name)
COMPRESSED_POINTER_FIELD(StackTracePtr, allocation_location)
VISIT_TO(allocation_location)
#endif // !defined(PRODUCT)
};
class UntaggedTransferableTypedData : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(TransferableTypedData);
VISIT_NOTHING();
};
// VM type for capturing stacktraces when exceptions are thrown,
// Currently we don't have any interface that this object is supposed
// to implement so we just support the 'toString' method which
// converts the stack trace into a string.
class UntaggedStackTrace : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(StackTrace);
// Link to parent async stack trace.
COMPRESSED_POINTER_FIELD(StackTracePtr, async_link);
VISIT_FROM(async_link)
// Code object for each frame in the stack trace.
COMPRESSED_POINTER_FIELD(ArrayPtr, code_array);
// Offset of PC for each frame.
COMPRESSED_POINTER_FIELD(TypedDataPtr, pc_offset_array);
VISIT_TO(pc_offset_array)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
// False for pre-allocated stack trace (used in OOM and Stack overflow).
bool expand_inlined_;
// Whether the link between the stack and the async-link represents a
// synchronous start to an asynchronous function. In this case, we omit the
// <asynchronous suspension> marker when concatenating the stacks.
bool skip_sync_start_in_parent_stack;
};
class UntaggedSuspendState : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(SuspendState);
NOT_IN_PRECOMPILED(intptr_t frame_capacity_);
intptr_t frame_size_;
uword pc_;
// Holds function-specific object which is returned from
// SuspendState.init* method.
// For async functions: _Future instance.
// For async* functions: _AsyncStarStreamController instance.
COMPRESSED_POINTER_FIELD(InstancePtr, function_data)
COMPRESSED_POINTER_FIELD(ClosurePtr, then_callback)
COMPRESSED_POINTER_FIELD(ClosurePtr, error_callback)
VISIT_FROM(function_data)
VISIT_TO(error_callback)
public:
uword pc() const { return pc_; }
intptr_t frame_capacity() const {
#if defined(DART_PRECOMPILED_RUNTIME)
return frame_size_;
#else
return frame_capacity_;
#endif
}
static intptr_t payload_offset() {
return OFFSET_OF_RETURNED_VALUE(UntaggedSuspendState, payload);
}
// Variable length payload follows here.
uint8_t* payload() { OPEN_ARRAY_START(uint8_t, uint8_t); }
const uint8_t* payload() const { OPEN_ARRAY_START(uint8_t, uint8_t); }
};
// VM type for capturing JS regular expressions.
class UntaggedRegExp : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(RegExp);
COMPRESSED_POINTER_FIELD(ArrayPtr, capture_name_map)
VISIT_FROM(capture_name_map)
// Pattern to be used for matching.
COMPRESSED_POINTER_FIELD(StringPtr, pattern)
COMPRESSED_POINTER_FIELD(ObjectPtr, one_byte) // FunctionPtr or TypedDataPtr
COMPRESSED_POINTER_FIELD(ObjectPtr, two_byte)
COMPRESSED_POINTER_FIELD(ObjectPtr, external_one_byte)
COMPRESSED_POINTER_FIELD(ObjectPtr, external_two_byte)
COMPRESSED_POINTER_FIELD(ObjectPtr, one_byte_sticky)
COMPRESSED_POINTER_FIELD(ObjectPtr, two_byte_sticky)
COMPRESSED_POINTER_FIELD(ObjectPtr, external_one_byte_sticky)
COMPRESSED_POINTER_FIELD(ObjectPtr, external_two_byte_sticky)
VISIT_TO(external_two_byte_sticky)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
std::atomic<intptr_t> num_bracket_expressions_;
intptr_t num_bracket_expressions() {
return num_bracket_expressions_.load(std::memory_order_relaxed);
}
void set_num_bracket_expressions(intptr_t value) {
num_bracket_expressions_.store(value, std::memory_order_relaxed);
}
// The same pattern may use different amount of registers if compiled
// for a one-byte target than a two-byte target. For example, we do not
// need to allocate registers to check whether the current position is within
// a surrogate pair when matching a Unicode pattern against a one-byte string.
intptr_t num_one_byte_registers_;
intptr_t num_two_byte_registers_;
// A bitfield with two fields:
// type: Uninitialized, simple or complex.
// flags: Represents global/local, case insensitive, multiline, unicode,
// dotAll.
// It is possible multiple compilers race to update the flags concurrently.
// That should be safe since all updates update to the same values..
AtomicBitFieldContainer<int8_t> type_flags_;
};
class UntaggedWeakProperty : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(WeakProperty);
COMPRESSED_POINTER_FIELD(ObjectPtr, key) // Weak reference.
VISIT_FROM(key)
COMPRESSED_POINTER_FIELD(ObjectPtr, value)
VISIT_TO(value)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
// Linked list is chaining all pending weak properties. Not visited by
// pointer visitors.
COMPRESSED_POINTER_FIELD(WeakPropertyPtr, next_seen_by_gc)
template <typename Type, typename PtrType>
friend class GCLinkedList;
template <bool>
friend class MarkingVisitorBase;
template <bool>
friend class ScavengerVisitorBase;
friend class FastObjectCopy; // For OFFSET_OF
friend class SlowObjectCopy; // For OFFSET_OF
};
// WeakProperty is special in that it has a pointer field which is not
// traversed by pointer visitors, and thus not in the range [from(),to()]:
// next_seen_by_gc, which is after the other fields.
template <>
DART_FORCE_INLINE uword
UntaggedObject::to_offset<UntaggedWeakProperty>(intptr_t length) {
return OFFSET_OF(UntaggedWeakProperty, next_seen_by_gc_);
}
class UntaggedWeakReference : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(WeakReference);
COMPRESSED_POINTER_FIELD(ObjectPtr, target) // Weak reference.
VISIT_FROM(target)
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr, type_arguments)
VISIT_TO(type_arguments)
CompressedObjectPtr* to_snapshot(Snapshot::Kind kind) { return to(); }
// Linked list is chaining all pending weak properties. Not visited by
// pointer visitors.
COMPRESSED_POINTER_FIELD(WeakReferencePtr, next_seen_by_gc)
template <typename Type, typename PtrType>
friend class GCLinkedList;
template <bool>
friend class MarkingVisitorBase;
template <bool>
friend class ScavengerVisitorBase;
friend class ObjectGraph;
friend class FastObjectCopy; // For OFFSET_OF
friend class SlowObjectCopy; // For OFFSET_OF
};
// WeakReference is special in that it has a pointer field which is not
// traversed by pointer visitors, and thus not in the range [from(),to()]:
// next_seen_by_gc, which is after the other fields.
template <>
DART_FORCE_INLINE uword
UntaggedObject::to_offset<UntaggedWeakReference>(intptr_t length) {
return OFFSET_OF(UntaggedWeakReference, next_seen_by_gc_);
}
class UntaggedFinalizerBase : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(FinalizerBase);
// The isolate this finalizer belongs to. Updated on sent and exit and set
// to null on isolate shutdown. See Isolate::finalizers_.
Isolate* isolate_;
// With compressed pointers, the first field in a subclass is at offset 28.
// If the fields would be public, the first field in a subclass is at offset 32.
// On Windows, it is always at offset 32, no matter public/private.
// This makes it 32 for all OSes.
// We can't use ALIGN8 on the first fields of the subclasses because they use
// the COMPRESSED_POINTER_FIELD macro to define it.
// Placed before the first fields so it is not included between from() and to().
#ifdef DART_COMPRESSED_POINTERS
uint32_t align_first_field_in_subclass;
#endif
COMPRESSED_POINTER_FIELD(ObjectPtr, detachments)
VISIT_FROM(detachments)
COMPRESSED_POINTER_FIELD(SetPtr, all_entries)
COMPRESSED_POINTER_FIELD(FinalizerEntryPtr, entries_collected)
template <typename GCVisitorType>
friend void MournFinalizerEntry(GCVisitorType*, FinalizerEntryPtr);
template <bool>
friend class MarkingVisitorBase;
template <bool>
friend class ScavengerVisitorBase;
friend class ObjectGraph;
};
class UntaggedFinalizer : public UntaggedFinalizerBase {
RAW_HEAP_OBJECT_IMPLEMENTATION(Finalizer);
COMPRESSED_POINTER_FIELD(ClosurePtr, callback)
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr, type_arguments)
VISIT_TO(type_arguments)
template <std::memory_order order = std::memory_order_relaxed>
FinalizerEntryPtr exchange_entries_collected(FinalizerEntryPtr value) {
return ExchangeCompressedPointer<FinalizerEntryPtr,
CompressedFinalizerEntryPtr, order>(
&entries_collected_, value);
}
template <typename GCVisitorType>
friend void MournFinalizerEntry(GCVisitorType*, FinalizerEntryPtr);
template <bool>
friend class MarkingVisitorBase;
template <bool>
friend class ScavengerVisitorBase;
};
class UntaggedNativeFinalizer : public UntaggedFinalizerBase {
RAW_HEAP_OBJECT_IMPLEMENTATION(NativeFinalizer);
COMPRESSED_POINTER_FIELD(PointerPtr, callback)
VISIT_TO(callback)
template <bool>
friend class MarkingVisitorBase;
template <bool>
friend class ScavengerVisitorBase;
};
class UntaggedFinalizerEntry : public UntaggedInstance {
public:
intptr_t external_size() { return external_size_; }
void set_external_size(intptr_t value) { external_size_ = value; }
private:
RAW_HEAP_OBJECT_IMPLEMENTATION(FinalizerEntry);
COMPRESSED_POINTER_FIELD(ObjectPtr, value) // Weak reference.
VISIT_FROM(value)
COMPRESSED_POINTER_FIELD(ObjectPtr, detach) // Weak reference.
COMPRESSED_POINTER_FIELD(ObjectPtr, token)
COMPRESSED_POINTER_FIELD(FinalizerBasePtr, finalizer) // Weak reference.
// Used for the linked list in Finalizer::entries_collected_. That cannot be
// an ordinary list because we need to add elements during a GC so we cannot
// modify the heap.
COMPRESSED_POINTER_FIELD(FinalizerEntryPtr, next)
VISIT_TO(next)
// Linked list is chaining all pending. Not visited by pointer visitors.
// Only populated during the GC, otherwise null.
COMPRESSED_POINTER_FIELD(FinalizerEntryPtr, next_seen_by_gc)
intptr_t external_size_;
template <typename Type, typename PtrType>
friend class GCLinkedList;
template <typename GCVisitorType>
friend void MournFinalizerEntry(GCVisitorType*, FinalizerEntryPtr);
template <bool>
friend class MarkingVisitorBase;
template <bool>
friend class ScavengerVisitorBase;
friend class ObjectGraph;
};
// FinalizerEntry is special in that it has a pointer field which is not
// traversed by pointer visitors, and thus not in the range [from(),to()]:
// next_seen_by_gc, which is after the other fields.
template <>
DART_FORCE_INLINE uword
UntaggedObject::to_offset<UntaggedFinalizerEntry>(intptr_t length) {
return OFFSET_OF(UntaggedFinalizerEntry, next_seen_by_gc_);
}
// MirrorReferences are used by mirrors to hold reflectees that are VM
// internal objects, such as libraries, classes, functions or types.
class UntaggedMirrorReference : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(MirrorReference);
COMPRESSED_POINTER_FIELD(ObjectPtr, referent)
VISIT_FROM(referent)
VISIT_TO(referent)
};
// UserTag are used by the profiler to track Dart script state.
class UntaggedUserTag : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(UserTag);
COMPRESSED_POINTER_FIELD(StringPtr, label)
VISIT_FROM(label)
VISIT_TO(label)
// Isolate unique tag.
uword tag_;
// Should CPU samples with this tag be streamed?
bool streamable_;
friend class Object;
public:
uword tag() const { return tag_; }
bool streamable() const { return streamable_; }
};
class UntaggedFutureOr : public UntaggedInstance {
RAW_HEAP_OBJECT_IMPLEMENTATION(FutureOr);
COMPRESSED_POINTER_FIELD(TypeArgumentsPtr, type_arguments)
VISIT_FROM(type_arguments)
VISIT_TO(type_arguments)
};
#undef WSR_COMPRESSED_POINTER_FIELD
} // namespace dart
#endif // RUNTIME_VM_RAW_OBJECT_H_
Reference in a new issue View git blame Copy permalink