1 // Copyright 2011 the V8 project authors. All rights reserved. 2 // Use of this source code is governed by a BSD-style license that can be 3 // found in the LICENSE file. 4 5 #ifndef V8_HEAP_SPACES_H_ 6 #define V8_HEAP_SPACES_H_ 7 8 #include "src/allocation.h" 9 #include "src/base/atomicops.h" 10 #include "src/base/bits.h" 11 #include "src/base/platform/mutex.h" 12 #include "src/hashmap.h" 13 #include "src/list.h" 14 #include "src/log.h" 15 #include "src/utils.h" 16 17 namespace v8 { 18 namespace internal { 19 20 class Isolate; 21 22 // ----------------------------------------------------------------------------- 23 // Heap structures: 24 // 25 // A JS heap consists of a young generation, an old generation, and a large 26 // object space. The young generation is divided into two semispaces. A 27 // scavenger implements Cheney's copying algorithm. The old generation is 28 // separated into a map space and an old object space. The map space contains 29 // all (and only) map objects, the rest of old objects go into the old space. 30 // The old generation is collected by a mark-sweep-compact collector. 31 // 32 // The semispaces of the young generation are contiguous. The old and map 33 // spaces consists of a list of pages. A page has a page header and an object 34 // area. 35 // 36 // There is a separate large object space for objects larger than 37 // Page::kMaxHeapObjectSize, so that they do not have to move during 38 // collection. The large object space is paged. Pages in large object space 39 // may be larger than the page size. 40 // 41 // A store-buffer based write barrier is used to keep track of intergenerational 42 // references. See heap/store-buffer.h. 43 // 44 // During scavenges and mark-sweep collections we sometimes (after a store 45 // buffer overflow) iterate intergenerational pointers without decoding heap 46 // object maps so if the page belongs to old pointer space or large object 47 // space it is essential to guarantee that the page does not contain any 48 // garbage pointers to new space: every pointer aligned word which satisfies 49 // the Heap::InNewSpace() predicate must be a pointer to a live heap object in 50 // new space. Thus objects in old pointer and large object spaces should have a 51 // special layout (e.g. no bare integer fields). This requirement does not 52 // apply to map space which is iterated in a special fashion. However we still 53 // require pointer fields of dead maps to be cleaned. 54 // 55 // To enable lazy cleaning of old space pages we can mark chunks of the page 56 // as being garbage. Garbage sections are marked with a special map. These 57 // sections are skipped when scanning the page, even if we are otherwise 58 // scanning without regard for object boundaries. Garbage sections are chained 59 // together to form a free list after a GC. Garbage sections created outside 60 // of GCs by object trunctation etc. may not be in the free list chain. Very 61 // small free spaces are ignored, they need only be cleaned of bogus pointers 62 // into new space. 63 // 64 // Each page may have up to one special garbage section. The start of this 65 // section is denoted by the top field in the space. The end of the section 66 // is denoted by the limit field in the space. This special garbage section 67 // is not marked with a free space map in the data. The point of this section 68 // is to enable linear allocation without having to constantly update the byte 69 // array every time the top field is updated and a new object is created. The 70 // special garbage section is not in the chain of garbage sections. 71 // 72 // Since the top and limit fields are in the space, not the page, only one page 73 // has a special garbage section, and if the top and limit are equal then there 74 // is no special garbage section. 75 76 // Some assertion macros used in the debugging mode. 77 78 #define DCHECK_PAGE_ALIGNED(address) \ 79 DCHECK((OffsetFrom(address) & Page::kPageAlignmentMask) == 0) 80 81 #define DCHECK_OBJECT_ALIGNED(address) \ 82 DCHECK((OffsetFrom(address) & kObjectAlignmentMask) == 0) 83 84 #define DCHECK_OBJECT_SIZE(size) \ 85 DCHECK((0 < size) && (size <= Page::kMaxRegularHeapObjectSize)) 86 87 #define DCHECK_PAGE_OFFSET(offset) \ 88 DCHECK((Page::kObjectStartOffset <= offset) && (offset <= Page::kPageSize)) 89 90 #define DCHECK_MAP_PAGE_INDEX(index) \ 91 DCHECK((0 <= index) && (index <= MapSpace::kMaxMapPageIndex)) 92 93 94 class PagedSpace; 95 class MemoryAllocator; 96 class AllocationInfo; 97 class Space; 98 class FreeList; 99 class MemoryChunk; 100 101 class MarkBit { 102 public: 103 typedef uint32_t CellType; 104 105 inline MarkBit(CellType* cell, CellType mask, bool data_only) 106 : cell_(cell), mask_(mask), data_only_(data_only) {} 107 108 inline CellType* cell() { return cell_; } 109 inline CellType mask() { return mask_; } 110 111 #ifdef DEBUG 112 bool operator==(const MarkBit& other) { 113 return cell_ == other.cell_ && mask_ == other.mask_; 114 } 115 #endif 116 117 inline void Set() { *cell_ |= mask_; } 118 inline bool Get() { return (*cell_ & mask_) != 0; } 119 inline void Clear() { *cell_ &= ~mask_; } 120 121 inline bool data_only() { return data_only_; } 122 123 inline MarkBit Next() { 124 CellType new_mask = mask_ << 1; 125 if (new_mask == 0) { 126 return MarkBit(cell_ + 1, 1, data_only_); 127 } else { 128 return MarkBit(cell_, new_mask, data_only_); 129 } 130 } 131 132 private: 133 CellType* cell_; 134 CellType mask_; 135 // This boolean indicates that the object is in a data-only space with no 136 // pointers. This enables some optimizations when marking. 137 // It is expected that this field is inlined and turned into control flow 138 // at the place where the MarkBit object is created. 139 bool data_only_; 140 }; 141 142 143 // Bitmap is a sequence of cells each containing fixed number of bits. 144 class Bitmap { 145 public: 146 static const uint32_t kBitsPerCell = 32; 147 static const uint32_t kBitsPerCellLog2 = 5; 148 static const uint32_t kBitIndexMask = kBitsPerCell - 1; 149 static const uint32_t kBytesPerCell = kBitsPerCell / kBitsPerByte; 150 static const uint32_t kBytesPerCellLog2 = kBitsPerCellLog2 - kBitsPerByteLog2; 151 152 static const size_t kLength = (1 << kPageSizeBits) >> (kPointerSizeLog2); 153 154 static const size_t kSize = 155 (1 << kPageSizeBits) >> (kPointerSizeLog2 + kBitsPerByteLog2); 156 157 158 static int CellsForLength(int length) { 159 return (length + kBitsPerCell - 1) >> kBitsPerCellLog2; 160 } 161 162 int CellsCount() { return CellsForLength(kLength); } 163 164 static int SizeFor(int cells_count) { 165 return sizeof(MarkBit::CellType) * cells_count; 166 } 167 168 INLINE(static uint32_t IndexToCell(uint32_t index)) { 169 return index >> kBitsPerCellLog2; 170 } 171 172 INLINE(static uint32_t CellToIndex(uint32_t index)) { 173 return index << kBitsPerCellLog2; 174 } 175 176 INLINE(static uint32_t CellAlignIndex(uint32_t index)) { 177 return (index + kBitIndexMask) & ~kBitIndexMask; 178 } 179 180 INLINE(MarkBit::CellType* cells()) { 181 return reinterpret_cast<MarkBit::CellType*>(this); 182 } 183 184 INLINE(Address address()) { return reinterpret_cast<Address>(this); } 185 186 INLINE(static Bitmap* FromAddress(Address addr)) { 187 return reinterpret_cast<Bitmap*>(addr); 188 } 189 190 inline MarkBit MarkBitFromIndex(uint32_t index, bool data_only = false) { 191 MarkBit::CellType mask = 1 << (index & kBitIndexMask); 192 MarkBit::CellType* cell = this->cells() + (index >> kBitsPerCellLog2); 193 return MarkBit(cell, mask, data_only); 194 } 195 196 static inline void Clear(MemoryChunk* chunk); 197 198 static void PrintWord(uint32_t word, uint32_t himask = 0) { 199 for (uint32_t mask = 1; mask != 0; mask <<= 1) { 200 if ((mask & himask) != 0) PrintF("["); 201 PrintF((mask & word) ? "1" : "0"); 202 if ((mask & himask) != 0) PrintF("]"); 203 } 204 } 205 206 class CellPrinter { 207 public: 208 CellPrinter() : seq_start(0), seq_type(0), seq_length(0) {} 209 210 void Print(uint32_t pos, uint32_t cell) { 211 if (cell == seq_type) { 212 seq_length++; 213 return; 214 } 215 216 Flush(); 217 218 if (IsSeq(cell)) { 219 seq_start = pos; 220 seq_length = 0; 221 seq_type = cell; 222 return; 223 } 224 225 PrintF("%d: ", pos); 226 PrintWord(cell); 227 PrintF("\n"); 228 } 229 230 void Flush() { 231 if (seq_length > 0) { 232 PrintF("%d: %dx%d\n", seq_start, seq_type == 0 ? 0 : 1, 233 seq_length * kBitsPerCell); 234 seq_length = 0; 235 } 236 } 237 238 static bool IsSeq(uint32_t cell) { return cell == 0 || cell == 0xFFFFFFFF; } 239 240 private: 241 uint32_t seq_start; 242 uint32_t seq_type; 243 uint32_t seq_length; 244 }; 245 246 void Print() { 247 CellPrinter printer; 248 for (int i = 0; i < CellsCount(); i++) { 249 printer.Print(i, cells()[i]); 250 } 251 printer.Flush(); 252 PrintF("\n"); 253 } 254 255 bool IsClean() { 256 for (int i = 0; i < CellsCount(); i++) { 257 if (cells()[i] != 0) { 258 return false; 259 } 260 } 261 return true; 262 } 263 }; 264 265 266 class SkipList; 267 class SlotsBuffer; 268 269 // MemoryChunk represents a memory region owned by a specific space. 270 // It is divided into the header and the body. Chunk start is always 271 // 1MB aligned. Start of the body is aligned so it can accommodate 272 // any heap object. 273 class MemoryChunk { 274 public: 275 // Only works if the pointer is in the first kPageSize of the MemoryChunk. 276 static MemoryChunk* FromAddress(Address a) { 277 return reinterpret_cast<MemoryChunk*>(OffsetFrom(a) & ~kAlignmentMask); 278 } 279 static const MemoryChunk* FromAddress(const byte* a) { 280 return reinterpret_cast<const MemoryChunk*>(OffsetFrom(a) & 281 ~kAlignmentMask); 282 } 283 284 // Only works for addresses in pointer spaces, not data or code spaces. 285 static inline MemoryChunk* FromAnyPointerAddress(Heap* heap, Address addr); 286 287 Address address() { return reinterpret_cast<Address>(this); } 288 289 bool is_valid() { return address() != NULL; } 290 291 MemoryChunk* next_chunk() const { 292 return reinterpret_cast<MemoryChunk*>(base::Acquire_Load(&next_chunk_)); 293 } 294 295 MemoryChunk* prev_chunk() const { 296 return reinterpret_cast<MemoryChunk*>(base::Acquire_Load(&prev_chunk_)); 297 } 298 299 void set_next_chunk(MemoryChunk* next) { 300 base::Release_Store(&next_chunk_, reinterpret_cast<base::AtomicWord>(next)); 301 } 302 303 void set_prev_chunk(MemoryChunk* prev) { 304 base::Release_Store(&prev_chunk_, reinterpret_cast<base::AtomicWord>(prev)); 305 } 306 307 Space* owner() const { 308 if ((reinterpret_cast<intptr_t>(owner_) & kPageHeaderTagMask) == 309 kPageHeaderTag) { 310 return reinterpret_cast<Space*>(reinterpret_cast<intptr_t>(owner_) - 311 kPageHeaderTag); 312 } else { 313 return NULL; 314 } 315 } 316 317 void set_owner(Space* space) { 318 DCHECK((reinterpret_cast<intptr_t>(space) & kPageHeaderTagMask) == 0); 319 owner_ = reinterpret_cast<Address>(space) + kPageHeaderTag; 320 DCHECK((reinterpret_cast<intptr_t>(owner_) & kPageHeaderTagMask) == 321 kPageHeaderTag); 322 } 323 324 base::VirtualMemory* reserved_memory() { return &reservation_; } 325 326 void InitializeReservedMemory() { reservation_.Reset(); } 327 328 void set_reserved_memory(base::VirtualMemory* reservation) { 329 DCHECK_NOT_NULL(reservation); 330 reservation_.TakeControl(reservation); 331 } 332 333 bool scan_on_scavenge() { return IsFlagSet(SCAN_ON_SCAVENGE); } 334 void initialize_scan_on_scavenge(bool scan) { 335 if (scan) { 336 SetFlag(SCAN_ON_SCAVENGE); 337 } else { 338 ClearFlag(SCAN_ON_SCAVENGE); 339 } 340 } 341 inline void set_scan_on_scavenge(bool scan); 342 343 int store_buffer_counter() { return store_buffer_counter_; } 344 void set_store_buffer_counter(int counter) { 345 store_buffer_counter_ = counter; 346 } 347 348 bool Contains(Address addr) { 349 return addr >= area_start() && addr < area_end(); 350 } 351 352 // Checks whether addr can be a limit of addresses in this page. 353 // It's a limit if it's in the page, or if it's just after the 354 // last byte of the page. 355 bool ContainsLimit(Address addr) { 356 return addr >= area_start() && addr <= area_end(); 357 } 358 359 // Every n write barrier invocations we go to runtime even though 360 // we could have handled it in generated code. This lets us check 361 // whether we have hit the limit and should do some more marking. 362 static const int kWriteBarrierCounterGranularity = 500; 363 364 enum MemoryChunkFlags { 365 IS_EXECUTABLE, 366 ABOUT_TO_BE_FREED, 367 POINTERS_TO_HERE_ARE_INTERESTING, 368 POINTERS_FROM_HERE_ARE_INTERESTING, 369 SCAN_ON_SCAVENGE, 370 IN_FROM_SPACE, // Mutually exclusive with IN_TO_SPACE. 371 IN_TO_SPACE, // All pages in new space has one of these two set. 372 NEW_SPACE_BELOW_AGE_MARK, 373 CONTAINS_ONLY_DATA, 374 EVACUATION_CANDIDATE, 375 RESCAN_ON_EVACUATION, 376 377 // WAS_SWEPT indicates that marking bits have been cleared by the sweeper, 378 // otherwise marking bits are still intact. 379 WAS_SWEPT, 380 381 // Large objects can have a progress bar in their page header. These object 382 // are scanned in increments and will be kept black while being scanned. 383 // Even if the mutator writes to them they will be kept black and a white 384 // to grey transition is performed in the value. 385 HAS_PROGRESS_BAR, 386 387 // Last flag, keep at bottom. 388 NUM_MEMORY_CHUNK_FLAGS 389 }; 390 391 392 static const int kPointersToHereAreInterestingMask = 393 1 << POINTERS_TO_HERE_ARE_INTERESTING; 394 395 static const int kPointersFromHereAreInterestingMask = 396 1 << POINTERS_FROM_HERE_ARE_INTERESTING; 397 398 static const int kEvacuationCandidateMask = 1 << EVACUATION_CANDIDATE; 399 400 static const int kSkipEvacuationSlotsRecordingMask = 401 (1 << EVACUATION_CANDIDATE) | (1 << RESCAN_ON_EVACUATION) | 402 (1 << IN_FROM_SPACE) | (1 << IN_TO_SPACE); 403 404 405 void SetFlag(int flag) { flags_ |= static_cast<uintptr_t>(1) << flag; } 406 407 void ClearFlag(int flag) { flags_ &= ~(static_cast<uintptr_t>(1) << flag); } 408 409 void SetFlagTo(int flag, bool value) { 410 if (value) { 411 SetFlag(flag); 412 } else { 413 ClearFlag(flag); 414 } 415 } 416 417 bool IsFlagSet(int flag) { 418 return (flags_ & (static_cast<uintptr_t>(1) << flag)) != 0; 419 } 420 421 // Set or clear multiple flags at a time. The flags in the mask 422 // are set to the value in "flags", the rest retain the current value 423 // in flags_. 424 void SetFlags(intptr_t flags, intptr_t mask) { 425 flags_ = (flags_ & ~mask) | (flags & mask); 426 } 427 428 // Return all current flags. 429 intptr_t GetFlags() { return flags_; } 430 431 432 // SWEEPING_DONE - The page state when sweeping is complete or sweeping must 433 // not be performed on that page. 434 // SWEEPING_FINALIZE - A sweeper thread is done sweeping this page and will 435 // not touch the page memory anymore. 436 // SWEEPING_IN_PROGRESS - This page is currently swept by a sweeper thread. 437 // SWEEPING_PENDING - This page is ready for parallel sweeping. 438 enum ParallelSweepingState { 439 SWEEPING_DONE, 440 SWEEPING_FINALIZE, 441 SWEEPING_IN_PROGRESS, 442 SWEEPING_PENDING 443 }; 444 445 ParallelSweepingState parallel_sweeping() { 446 return static_cast<ParallelSweepingState>( 447 base::Acquire_Load(¶llel_sweeping_)); 448 } 449 450 void set_parallel_sweeping(ParallelSweepingState state) { 451 base::Release_Store(¶llel_sweeping_, state); 452 } 453 454 bool TryParallelSweeping() { 455 return base::Acquire_CompareAndSwap(¶llel_sweeping_, SWEEPING_PENDING, 456 SWEEPING_IN_PROGRESS) == 457 SWEEPING_PENDING; 458 } 459 460 bool SweepingCompleted() { return parallel_sweeping() <= SWEEPING_FINALIZE; } 461 462 // Manage live byte count (count of bytes known to be live, 463 // because they are marked black). 464 void ResetLiveBytes() { 465 if (FLAG_gc_verbose) { 466 PrintF("ResetLiveBytes:%p:%x->0\n", static_cast<void*>(this), 467 live_byte_count_); 468 } 469 live_byte_count_ = 0; 470 } 471 void IncrementLiveBytes(int by) { 472 if (FLAG_gc_verbose) { 473 printf("UpdateLiveBytes:%p:%x%c=%x->%x\n", static_cast<void*>(this), 474 live_byte_count_, ((by < 0) ? '-' : '+'), ((by < 0) ? -by : by), 475 live_byte_count_ + by); 476 } 477 live_byte_count_ += by; 478 DCHECK_LE(static_cast<unsigned>(live_byte_count_), size_); 479 } 480 int LiveBytes() { 481 DCHECK(static_cast<unsigned>(live_byte_count_) <= size_); 482 return live_byte_count_; 483 } 484 485 int write_barrier_counter() { 486 return static_cast<int>(write_barrier_counter_); 487 } 488 489 void set_write_barrier_counter(int counter) { 490 write_barrier_counter_ = counter; 491 } 492 493 int progress_bar() { 494 DCHECK(IsFlagSet(HAS_PROGRESS_BAR)); 495 return progress_bar_; 496 } 497 498 void set_progress_bar(int progress_bar) { 499 DCHECK(IsFlagSet(HAS_PROGRESS_BAR)); 500 progress_bar_ = progress_bar; 501 } 502 503 void ResetProgressBar() { 504 if (IsFlagSet(MemoryChunk::HAS_PROGRESS_BAR)) { 505 set_progress_bar(0); 506 ClearFlag(MemoryChunk::HAS_PROGRESS_BAR); 507 } 508 } 509 510 bool IsLeftOfProgressBar(Object** slot) { 511 Address slot_address = reinterpret_cast<Address>(slot); 512 DCHECK(slot_address > this->address()); 513 return (slot_address - (this->address() + kObjectStartOffset)) < 514 progress_bar(); 515 } 516 517 static void IncrementLiveBytesFromGC(Address address, int by) { 518 MemoryChunk::FromAddress(address)->IncrementLiveBytes(by); 519 } 520 521 static void IncrementLiveBytesFromMutator(Address address, int by); 522 523 static const intptr_t kAlignment = 524 (static_cast<uintptr_t>(1) << kPageSizeBits); 525 526 static const intptr_t kAlignmentMask = kAlignment - 1; 527 528 static const intptr_t kSizeOffset = 0; 529 530 static const intptr_t kLiveBytesOffset = 531 kSizeOffset + kPointerSize + kPointerSize + kPointerSize + kPointerSize + 532 kPointerSize + kPointerSize + kPointerSize + kPointerSize + kIntSize; 533 534 static const size_t kSlotsBufferOffset = kLiveBytesOffset + kIntSize; 535 536 static const size_t kWriteBarrierCounterOffset = 537 kSlotsBufferOffset + kPointerSize + kPointerSize; 538 539 static const size_t kHeaderSize = 540 kWriteBarrierCounterOffset + kPointerSize + kIntSize + kIntSize + 541 kPointerSize + 5 * kPointerSize + kPointerSize + kPointerSize; 542 543 static const int kBodyOffset = 544 CODE_POINTER_ALIGN(kHeaderSize + Bitmap::kSize); 545 546 // The start offset of the object area in a page. Aligned to both maps and 547 // code alignment to be suitable for both. Also aligned to 32 words because 548 // the marking bitmap is arranged in 32 bit chunks. 549 static const int kObjectStartAlignment = 32 * kPointerSize; 550 static const int kObjectStartOffset = 551 kBodyOffset - 1 + 552 (kObjectStartAlignment - (kBodyOffset - 1) % kObjectStartAlignment); 553 554 size_t size() const { return size_; } 555 556 void set_size(size_t size) { size_ = size; } 557 558 void SetArea(Address area_start, Address area_end) { 559 area_start_ = area_start; 560 area_end_ = area_end; 561 } 562 563 Executability executable() { 564 return IsFlagSet(IS_EXECUTABLE) ? EXECUTABLE : NOT_EXECUTABLE; 565 } 566 567 bool ContainsOnlyData() { return IsFlagSet(CONTAINS_ONLY_DATA); } 568 569 bool InNewSpace() { 570 return (flags_ & ((1 << IN_FROM_SPACE) | (1 << IN_TO_SPACE))) != 0; 571 } 572 573 bool InToSpace() { return IsFlagSet(IN_TO_SPACE); } 574 575 bool InFromSpace() { return IsFlagSet(IN_FROM_SPACE); } 576 577 // --------------------------------------------------------------------- 578 // Markbits support 579 580 inline Bitmap* markbits() { 581 return Bitmap::FromAddress(address() + kHeaderSize); 582 } 583 584 void PrintMarkbits() { markbits()->Print(); } 585 586 inline uint32_t AddressToMarkbitIndex(Address addr) { 587 return static_cast<uint32_t>(addr - this->address()) >> kPointerSizeLog2; 588 } 589 590 inline static uint32_t FastAddressToMarkbitIndex(Address addr) { 591 const intptr_t offset = reinterpret_cast<intptr_t>(addr) & kAlignmentMask; 592 593 return static_cast<uint32_t>(offset) >> kPointerSizeLog2; 594 } 595 596 inline Address MarkbitIndexToAddress(uint32_t index) { 597 return this->address() + (index << kPointerSizeLog2); 598 } 599 600 void InsertAfter(MemoryChunk* other); 601 void Unlink(); 602 603 inline Heap* heap() const { return heap_; } 604 605 static const int kFlagsOffset = kPointerSize; 606 607 bool IsEvacuationCandidate() { return IsFlagSet(EVACUATION_CANDIDATE); } 608 609 bool ShouldSkipEvacuationSlotRecording() { 610 return (flags_ & kSkipEvacuationSlotsRecordingMask) != 0; 611 } 612 613 inline SkipList* skip_list() { return skip_list_; } 614 615 inline void set_skip_list(SkipList* skip_list) { skip_list_ = skip_list; } 616 617 inline SlotsBuffer* slots_buffer() { return slots_buffer_; } 618 619 inline SlotsBuffer** slots_buffer_address() { return &slots_buffer_; } 620 621 void MarkEvacuationCandidate() { 622 DCHECK(slots_buffer_ == NULL); 623 SetFlag(EVACUATION_CANDIDATE); 624 } 625 626 void ClearEvacuationCandidate() { 627 DCHECK(slots_buffer_ == NULL); 628 ClearFlag(EVACUATION_CANDIDATE); 629 } 630 631 Address area_start() { return area_start_; } 632 Address area_end() { return area_end_; } 633 int area_size() { return static_cast<int>(area_end() - area_start()); } 634 bool CommitArea(size_t requested); 635 636 // Approximate amount of physical memory committed for this chunk. 637 size_t CommittedPhysicalMemory() { return high_water_mark_; } 638 639 static inline void UpdateHighWaterMark(Address mark); 640 641 protected: 642 size_t size_; 643 intptr_t flags_; 644 645 // Start and end of allocatable memory on this chunk. 646 Address area_start_; 647 Address area_end_; 648 649 // If the chunk needs to remember its memory reservation, it is stored here. 650 base::VirtualMemory reservation_; 651 // The identity of the owning space. This is tagged as a failure pointer, but 652 // no failure can be in an object, so this can be distinguished from any entry 653 // in a fixed array. 654 Address owner_; 655 Heap* heap_; 656 // Used by the store buffer to keep track of which pages to mark scan-on- 657 // scavenge. 658 int store_buffer_counter_; 659 // Count of bytes marked black on page. 660 int live_byte_count_; 661 SlotsBuffer* slots_buffer_; 662 SkipList* skip_list_; 663 intptr_t write_barrier_counter_; 664 // Used by the incremental marker to keep track of the scanning progress in 665 // large objects that have a progress bar and are scanned in increments. 666 int progress_bar_; 667 // Assuming the initial allocation on a page is sequential, 668 // count highest number of bytes ever allocated on the page. 669 int high_water_mark_; 670 671 base::AtomicWord parallel_sweeping_; 672 673 // PagedSpace free-list statistics. 674 intptr_t available_in_small_free_list_; 675 intptr_t available_in_medium_free_list_; 676 intptr_t available_in_large_free_list_; 677 intptr_t available_in_huge_free_list_; 678 intptr_t non_available_small_blocks_; 679 680 static MemoryChunk* Initialize(Heap* heap, Address base, size_t size, 681 Address area_start, Address area_end, 682 Executability executable, Space* owner); 683 684 private: 685 // next_chunk_ holds a pointer of type MemoryChunk 686 base::AtomicWord next_chunk_; 687 // prev_chunk_ holds a pointer of type MemoryChunk 688 base::AtomicWord prev_chunk_; 689 690 friend class MemoryAllocator; 691 }; 692 693 694 STATIC_ASSERT(sizeof(MemoryChunk) <= MemoryChunk::kHeaderSize); 695 696 697 // ----------------------------------------------------------------------------- 698 // A page is a memory chunk of a size 1MB. Large object pages may be larger. 699 // 700 // The only way to get a page pointer is by calling factory methods: 701 // Page* p = Page::FromAddress(addr); or 702 // Page* p = Page::FromAllocationTop(top); 703 class Page : public MemoryChunk { 704 public: 705 // Returns the page containing a given address. The address ranges 706 // from [page_addr .. page_addr + kPageSize[ 707 // This only works if the object is in fact in a page. See also MemoryChunk:: 708 // FromAddress() and FromAnyAddress(). 709 INLINE(static Page* FromAddress(Address a)) { 710 return reinterpret_cast<Page*>(OffsetFrom(a) & ~kPageAlignmentMask); 711 } 712 713 // Returns the page containing an allocation top. Because an allocation 714 // top address can be the upper bound of the page, we need to subtract 715 // it with kPointerSize first. The address ranges from 716 // [page_addr + kObjectStartOffset .. page_addr + kPageSize]. 717 INLINE(static Page* FromAllocationTop(Address top)) { 718 Page* p = FromAddress(top - kPointerSize); 719 return p; 720 } 721 722 // Returns the next page in the chain of pages owned by a space. 723 inline Page* next_page(); 724 inline Page* prev_page(); 725 inline void set_next_page(Page* page); 726 inline void set_prev_page(Page* page); 727 728 // Checks whether an address is page aligned. 729 static bool IsAlignedToPageSize(Address a) { 730 return 0 == (OffsetFrom(a) & kPageAlignmentMask); 731 } 732 733 // Returns the offset of a given address to this page. 734 INLINE(int Offset(Address a)) { 735 int offset = static_cast<int>(a - address()); 736 return offset; 737 } 738 739 // Returns the address for a given offset to the this page. 740 Address OffsetToAddress(int offset) { 741 DCHECK_PAGE_OFFSET(offset); 742 return address() + offset; 743 } 744 745 // --------------------------------------------------------------------- 746 747 // Page size in bytes. This must be a multiple of the OS page size. 748 static const int kPageSize = 1 << kPageSizeBits; 749 750 // Maximum object size that fits in a page. Objects larger than that size 751 // are allocated in large object space and are never moved in memory. This 752 // also applies to new space allocation, since objects are never migrated 753 // from new space to large object space. Takes double alignment into account. 754 static const int kMaxRegularHeapObjectSize = kPageSize - kObjectStartOffset; 755 756 // Page size mask. 757 static const intptr_t kPageAlignmentMask = (1 << kPageSizeBits) - 1; 758 759 inline void ClearGCFields(); 760 761 static inline Page* Initialize(Heap* heap, MemoryChunk* chunk, 762 Executability executable, PagedSpace* owner); 763 764 void InitializeAsAnchor(PagedSpace* owner); 765 766 bool WasSwept() { return IsFlagSet(WAS_SWEPT); } 767 void SetWasSwept() { SetFlag(WAS_SWEPT); } 768 void ClearWasSwept() { ClearFlag(WAS_SWEPT); } 769 770 void ResetFreeListStatistics(); 771 772 #define FRAGMENTATION_STATS_ACCESSORS(type, name) \ 773 type name() { return name##_; } \ 774 void set_##name(type name) { name##_ = name; } \ 775 void add_##name(type name) { name##_ += name; } 776 777 FRAGMENTATION_STATS_ACCESSORS(intptr_t, non_available_small_blocks) 778 FRAGMENTATION_STATS_ACCESSORS(intptr_t, available_in_small_free_list) 779 FRAGMENTATION_STATS_ACCESSORS(intptr_t, available_in_medium_free_list) 780 FRAGMENTATION_STATS_ACCESSORS(intptr_t, available_in_large_free_list) 781 FRAGMENTATION_STATS_ACCESSORS(intptr_t, available_in_huge_free_list) 782 783 #undef FRAGMENTATION_STATS_ACCESSORS 784 785 #ifdef DEBUG 786 void Print(); 787 #endif // DEBUG 788 789 friend class MemoryAllocator; 790 }; 791 792 793 STATIC_ASSERT(sizeof(Page) <= MemoryChunk::kHeaderSize); 794 795 796 class LargePage : public MemoryChunk { 797 public: 798 HeapObject* GetObject() { return HeapObject::FromAddress(area_start()); } 799 800 inline LargePage* next_page() const { 801 return static_cast<LargePage*>(next_chunk()); 802 } 803 804 inline void set_next_page(LargePage* page) { set_next_chunk(page); } 805 806 private: 807 static inline LargePage* Initialize(Heap* heap, MemoryChunk* chunk); 808 809 friend class MemoryAllocator; 810 }; 811 812 STATIC_ASSERT(sizeof(LargePage) <= MemoryChunk::kHeaderSize); 813 814 // ---------------------------------------------------------------------------- 815 // Space is the abstract superclass for all allocation spaces. 816 class Space : public Malloced { 817 public: 818 Space(Heap* heap, AllocationSpace id, Executability executable) 819 : heap_(heap), id_(id), executable_(executable) {} 820 821 virtual ~Space() {} 822 823 Heap* heap() const { return heap_; } 824 825 // Does the space need executable memory? 826 Executability executable() { return executable_; } 827 828 // Identity used in error reporting. 829 AllocationSpace identity() { return id_; } 830 831 // Returns allocated size. 832 virtual intptr_t Size() = 0; 833 834 // Returns size of objects. Can differ from the allocated size 835 // (e.g. see LargeObjectSpace). 836 virtual intptr_t SizeOfObjects() { return Size(); } 837 838 virtual int RoundSizeDownToObjectAlignment(int size) { 839 if (id_ == CODE_SPACE) { 840 return RoundDown(size, kCodeAlignment); 841 } else { 842 return RoundDown(size, kPointerSize); 843 } 844 } 845 846 #ifdef DEBUG 847 virtual void Print() = 0; 848 #endif 849 850 private: 851 Heap* heap_; 852 AllocationSpace id_; 853 Executability executable_; 854 }; 855 856 857 // ---------------------------------------------------------------------------- 858 // All heap objects containing executable code (code objects) must be allocated 859 // from a 2 GB range of memory, so that they can call each other using 32-bit 860 // displacements. This happens automatically on 32-bit platforms, where 32-bit 861 // displacements cover the entire 4GB virtual address space. On 64-bit 862 // platforms, we support this using the CodeRange object, which reserves and 863 // manages a range of virtual memory. 864 class CodeRange { 865 public: 866 explicit CodeRange(Isolate* isolate); 867 ~CodeRange() { TearDown(); } 868 869 // Reserves a range of virtual memory, but does not commit any of it. 870 // Can only be called once, at heap initialization time. 871 // Returns false on failure. 872 bool SetUp(size_t requested_size); 873 874 // Frees the range of virtual memory, and frees the data structures used to 875 // manage it. 876 void TearDown(); 877 878 bool valid() { return code_range_ != NULL; } 879 Address start() { 880 DCHECK(valid()); 881 return static_cast<Address>(code_range_->address()); 882 } 883 bool contains(Address address) { 884 if (!valid()) return false; 885 Address start = static_cast<Address>(code_range_->address()); 886 return start <= address && address < start + code_range_->size(); 887 } 888 889 // Allocates a chunk of memory from the large-object portion of 890 // the code range. On platforms with no separate code range, should 891 // not be called. 892 MUST_USE_RESULT Address AllocateRawMemory(const size_t requested_size, 893 const size_t commit_size, 894 size_t* allocated); 895 bool CommitRawMemory(Address start, size_t length); 896 bool UncommitRawMemory(Address start, size_t length); 897 void FreeRawMemory(Address buf, size_t length); 898 899 private: 900 Isolate* isolate_; 901 902 // The reserved range of virtual memory that all code objects are put in. 903 base::VirtualMemory* code_range_; 904 // Plain old data class, just a struct plus a constructor. 905 class FreeBlock { 906 public: 907 FreeBlock(Address start_arg, size_t size_arg) 908 : start(start_arg), size(size_arg) { 909 DCHECK(IsAddressAligned(start, MemoryChunk::kAlignment)); 910 DCHECK(size >= static_cast<size_t>(Page::kPageSize)); 911 } 912 FreeBlock(void* start_arg, size_t size_arg) 913 : start(static_cast<Address>(start_arg)), size(size_arg) { 914 DCHECK(IsAddressAligned(start, MemoryChunk::kAlignment)); 915 DCHECK(size >= static_cast<size_t>(Page::kPageSize)); 916 } 917 918 Address start; 919 size_t size; 920 }; 921 922 // Freed blocks of memory are added to the free list. When the allocation 923 // list is exhausted, the free list is sorted and merged to make the new 924 // allocation list. 925 List<FreeBlock> free_list_; 926 // Memory is allocated from the free blocks on the allocation list. 927 // The block at current_allocation_block_index_ is the current block. 928 List<FreeBlock> allocation_list_; 929 int current_allocation_block_index_; 930 931 // Finds a block on the allocation list that contains at least the 932 // requested amount of memory. If none is found, sorts and merges 933 // the existing free memory blocks, and searches again. 934 // If none can be found, returns false. 935 bool GetNextAllocationBlock(size_t requested); 936 // Compares the start addresses of two free blocks. 937 static int CompareFreeBlockAddress(const FreeBlock* left, 938 const FreeBlock* right); 939 940 DISALLOW_COPY_AND_ASSIGN(CodeRange); 941 }; 942 943 944 class SkipList { 945 public: 946 SkipList() { Clear(); } 947 948 void Clear() { 949 for (int idx = 0; idx < kSize; idx++) { 950 starts_[idx] = reinterpret_cast<Address>(-1); 951 } 952 } 953 954 Address StartFor(Address addr) { return starts_[RegionNumber(addr)]; } 955 956 void AddObject(Address addr, int size) { 957 int start_region = RegionNumber(addr); 958 int end_region = RegionNumber(addr + size - kPointerSize); 959 for (int idx = start_region; idx <= end_region; idx++) { 960 if (starts_[idx] > addr) starts_[idx] = addr; 961 } 962 } 963 964 static inline int RegionNumber(Address addr) { 965 return (OffsetFrom(addr) & Page::kPageAlignmentMask) >> kRegionSizeLog2; 966 } 967 968 static void Update(Address addr, int size) { 969 Page* page = Page::FromAddress(addr); 970 SkipList* list = page->skip_list(); 971 if (list == NULL) { 972 list = new SkipList(); 973 page->set_skip_list(list); 974 } 975 976 list->AddObject(addr, size); 977 } 978 979 private: 980 static const int kRegionSizeLog2 = 13; 981 static const int kRegionSize = 1 << kRegionSizeLog2; 982 static const int kSize = Page::kPageSize / kRegionSize; 983 984 STATIC_ASSERT(Page::kPageSize % kRegionSize == 0); 985 986 Address starts_[kSize]; 987 }; 988 989 990 // ---------------------------------------------------------------------------- 991 // A space acquires chunks of memory from the operating system. The memory 992 // allocator allocated and deallocates pages for the paged heap spaces and large 993 // pages for large object space. 994 // 995 // Each space has to manage it's own pages. 996 // 997 class MemoryAllocator { 998 public: 999 explicit MemoryAllocator(Isolate* isolate); 1000 1001 // Initializes its internal bookkeeping structures. 1002 // Max capacity of the total space and executable memory limit. 1003 bool SetUp(intptr_t max_capacity, intptr_t capacity_executable); 1004 1005 void TearDown(); 1006 1007 Page* AllocatePage(intptr_t size, PagedSpace* owner, 1008 Executability executable); 1009 1010 LargePage* AllocateLargePage(intptr_t object_size, Space* owner, 1011 Executability executable); 1012 1013 void Free(MemoryChunk* chunk); 1014 1015 // Returns the maximum available bytes of heaps. 1016 intptr_t Available() { return capacity_ < size_ ? 0 : capacity_ - size_; } 1017 1018 // Returns allocated spaces in bytes. 1019 intptr_t Size() { return size_; } 1020 1021 // Returns the maximum available executable bytes of heaps. 1022 intptr_t AvailableExecutable() { 1023 if (capacity_executable_ < size_executable_) return 0; 1024 return capacity_executable_ - size_executable_; 1025 } 1026 1027 // Returns allocated executable spaces in bytes. 1028 intptr_t SizeExecutable() { return size_executable_; } 1029 1030 // Returns maximum available bytes that the old space can have. 1031 intptr_t MaxAvailable() { 1032 return (Available() / Page::kPageSize) * Page::kMaxRegularHeapObjectSize; 1033 } 1034 1035 // Returns an indication of whether a pointer is in a space that has 1036 // been allocated by this MemoryAllocator. 1037 V8_INLINE bool IsOutsideAllocatedSpace(const void* address) const { 1038 return address < lowest_ever_allocated_ || 1039 address >= highest_ever_allocated_; 1040 } 1041 1042 #ifdef DEBUG 1043 // Reports statistic info of the space. 1044 void ReportStatistics(); 1045 #endif 1046 1047 // Returns a MemoryChunk in which the memory region from commit_area_size to 1048 // reserve_area_size of the chunk area is reserved but not committed, it 1049 // could be committed later by calling MemoryChunk::CommitArea. 1050 MemoryChunk* AllocateChunk(intptr_t reserve_area_size, 1051 intptr_t commit_area_size, 1052 Executability executable, Space* space); 1053 1054 Address ReserveAlignedMemory(size_t requested, size_t alignment, 1055 base::VirtualMemory* controller); 1056 Address AllocateAlignedMemory(size_t reserve_size, size_t commit_size, 1057 size_t alignment, Executability executable, 1058 base::VirtualMemory* controller); 1059 1060 bool CommitMemory(Address addr, size_t size, Executability executable); 1061 1062 void FreeMemory(base::VirtualMemory* reservation, Executability executable); 1063 void FreeMemory(Address addr, size_t size, Executability executable); 1064 1065 // Commit a contiguous block of memory from the initial chunk. Assumes that 1066 // the address is not NULL, the size is greater than zero, and that the 1067 // block is contained in the initial chunk. Returns true if it succeeded 1068 // and false otherwise. 1069 bool CommitBlock(Address start, size_t size, Executability executable); 1070 1071 // Uncommit a contiguous block of memory [start..(start+size)[. 1072 // start is not NULL, the size is greater than zero, and the 1073 // block is contained in the initial chunk. Returns true if it succeeded 1074 // and false otherwise. 1075 bool UncommitBlock(Address start, size_t size); 1076 1077 // Zaps a contiguous block of memory [start..(start+size)[ thus 1078 // filling it up with a recognizable non-NULL bit pattern. 1079 void ZapBlock(Address start, size_t size); 1080 1081 void PerformAllocationCallback(ObjectSpace space, AllocationAction action, 1082 size_t size); 1083 1084 void AddMemoryAllocationCallback(MemoryAllocationCallback callback, 1085 ObjectSpace space, AllocationAction action); 1086 1087 void RemoveMemoryAllocationCallback(MemoryAllocationCallback callback); 1088 1089 bool MemoryAllocationCallbackRegistered(MemoryAllocationCallback callback); 1090 1091 static int CodePageGuardStartOffset(); 1092 1093 static int CodePageGuardSize(); 1094 1095 static int CodePageAreaStartOffset(); 1096 1097 static int CodePageAreaEndOffset(); 1098 1099 static int CodePageAreaSize() { 1100 return CodePageAreaEndOffset() - CodePageAreaStartOffset(); 1101 } 1102 1103 MUST_USE_RESULT bool CommitExecutableMemory(base::VirtualMemory* vm, 1104 Address start, size_t commit_size, 1105 size_t reserved_size); 1106 1107 private: 1108 Isolate* isolate_; 1109 1110 // Maximum space size in bytes. 1111 size_t capacity_; 1112 // Maximum subset of capacity_ that can be executable 1113 size_t capacity_executable_; 1114 1115 // Allocated space size in bytes. 1116 size_t size_; 1117 // Allocated executable space size in bytes. 1118 size_t size_executable_; 1119 1120 // We keep the lowest and highest addresses allocated as a quick way 1121 // of determining that pointers are outside the heap. The estimate is 1122 // conservative, i.e. not all addrsses in 'allocated' space are allocated 1123 // to our heap. The range is [lowest, highest[, inclusive on the low end 1124 // and exclusive on the high end. 1125 void* lowest_ever_allocated_; 1126 void* highest_ever_allocated_; 1127 1128 struct MemoryAllocationCallbackRegistration { 1129 MemoryAllocationCallbackRegistration(MemoryAllocationCallback callback, 1130 ObjectSpace space, 1131 AllocationAction action) 1132 : callback(callback), space(space), action(action) {} 1133 MemoryAllocationCallback callback; 1134 ObjectSpace space; 1135 AllocationAction action; 1136 }; 1137 1138 // A List of callback that are triggered when memory is allocated or free'd 1139 List<MemoryAllocationCallbackRegistration> memory_allocation_callbacks_; 1140 1141 // Initializes pages in a chunk. Returns the first page address. 1142 // This function and GetChunkId() are provided for the mark-compact 1143 // collector to rebuild page headers in the from space, which is 1144 // used as a marking stack and its page headers are destroyed. 1145 Page* InitializePagesInChunk(int chunk_id, int pages_in_chunk, 1146 PagedSpace* owner); 1147 1148 void UpdateAllocatedSpaceLimits(void* low, void* high) { 1149 lowest_ever_allocated_ = Min(lowest_ever_allocated_, low); 1150 highest_ever_allocated_ = Max(highest_ever_allocated_, high); 1151 } 1152 1153 DISALLOW_IMPLICIT_CONSTRUCTORS(MemoryAllocator); 1154 }; 1155 1156 1157 // ----------------------------------------------------------------------------- 1158 // Interface for heap object iterator to be implemented by all object space 1159 // object iterators. 1160 // 1161 // NOTE: The space specific object iterators also implements the own next() 1162 // method which is used to avoid using virtual functions 1163 // iterating a specific space. 1164 1165 class ObjectIterator : public Malloced { 1166 public: 1167 virtual ~ObjectIterator() {} 1168 1169 virtual HeapObject* next_object() = 0; 1170 }; 1171 1172 1173 // ----------------------------------------------------------------------------- 1174 // Heap object iterator in new/old/map spaces. 1175 // 1176 // A HeapObjectIterator iterates objects from the bottom of the given space 1177 // to its top or from the bottom of the given page to its top. 1178 // 1179 // If objects are allocated in the page during iteration the iterator may 1180 // or may not iterate over those objects. The caller must create a new 1181 // iterator in order to be sure to visit these new objects. 1182 class HeapObjectIterator : public ObjectIterator { 1183 public: 1184 // Creates a new object iterator in a given space. 1185 // If the size function is not given, the iterator calls the default 1186 // Object::Size(). 1187 explicit HeapObjectIterator(PagedSpace* space); 1188 HeapObjectIterator(PagedSpace* space, HeapObjectCallback size_func); 1189 HeapObjectIterator(Page* page, HeapObjectCallback size_func); 1190 1191 // Advance to the next object, skipping free spaces and other fillers and 1192 // skipping the special garbage section of which there is one per space. 1193 // Returns NULL when the iteration has ended. 1194 inline HeapObject* Next() { 1195 do { 1196 HeapObject* next_obj = FromCurrentPage(); 1197 if (next_obj != NULL) return next_obj; 1198 } while (AdvanceToNextPage()); 1199 return NULL; 1200 } 1201 1202 virtual HeapObject* next_object() { return Next(); } 1203 1204 private: 1205 enum PageMode { kOnePageOnly, kAllPagesInSpace }; 1206 1207 Address cur_addr_; // Current iteration point. 1208 Address cur_end_; // End iteration point. 1209 HeapObjectCallback size_func_; // Size function or NULL. 1210 PagedSpace* space_; 1211 PageMode page_mode_; 1212 1213 // Fast (inlined) path of next(). 1214 inline HeapObject* FromCurrentPage(); 1215 1216 // Slow path of next(), goes into the next page. Returns false if the 1217 // iteration has ended. 1218 bool AdvanceToNextPage(); 1219 1220 // Initializes fields. 1221 inline void Initialize(PagedSpace* owner, Address start, Address end, 1222 PageMode mode, HeapObjectCallback size_func); 1223 }; 1224 1225 1226 // ----------------------------------------------------------------------------- 1227 // A PageIterator iterates the pages in a paged space. 1228 1229 class PageIterator BASE_EMBEDDED { 1230 public: 1231 explicit inline PageIterator(PagedSpace* space); 1232 1233 inline bool has_next(); 1234 inline Page* next(); 1235 1236 private: 1237 PagedSpace* space_; 1238 Page* prev_page_; // Previous page returned. 1239 // Next page that will be returned. Cached here so that we can use this 1240 // iterator for operations that deallocate pages. 1241 Page* next_page_; 1242 }; 1243 1244 1245 // ----------------------------------------------------------------------------- 1246 // A space has a circular list of pages. The next page can be accessed via 1247 // Page::next_page() call. 1248 1249 // An abstraction of allocation and relocation pointers in a page-structured 1250 // space. 1251 class AllocationInfo { 1252 public: 1253 AllocationInfo() : top_(NULL), limit_(NULL) {} 1254 1255 INLINE(void set_top(Address top)) { 1256 SLOW_DCHECK(top == NULL || 1257 (reinterpret_cast<intptr_t>(top) & HeapObjectTagMask()) == 0); 1258 top_ = top; 1259 } 1260 1261 INLINE(Address top()) const { 1262 SLOW_DCHECK(top_ == NULL || 1263 (reinterpret_cast<intptr_t>(top_) & HeapObjectTagMask()) == 0); 1264 return top_; 1265 } 1266 1267 Address* top_address() { return &top_; } 1268 1269 INLINE(void set_limit(Address limit)) { 1270 SLOW_DCHECK(limit == NULL || 1271 (reinterpret_cast<intptr_t>(limit) & HeapObjectTagMask()) == 0); 1272 limit_ = limit; 1273 } 1274 1275 INLINE(Address limit()) const { 1276 SLOW_DCHECK(limit_ == NULL || 1277 (reinterpret_cast<intptr_t>(limit_) & HeapObjectTagMask()) == 1278 0); 1279 return limit_; 1280 } 1281 1282 Address* limit_address() { return &limit_; } 1283 1284 #ifdef DEBUG 1285 bool VerifyPagedAllocation() { 1286 return (Page::FromAllocationTop(top_) == Page::FromAllocationTop(limit_)) && 1287 (top_ <= limit_); 1288 } 1289 #endif 1290 1291 private: 1292 // Current allocation top. 1293 Address top_; 1294 // Current allocation limit. 1295 Address limit_; 1296 }; 1297 1298 1299 // An abstraction of the accounting statistics of a page-structured space. 1300 // The 'capacity' of a space is the number of object-area bytes (i.e., not 1301 // including page bookkeeping structures) currently in the space. The 'size' 1302 // of a space is the number of allocated bytes, the 'waste' in the space is 1303 // the number of bytes that are not allocated and not available to 1304 // allocation without reorganizing the space via a GC (e.g. small blocks due 1305 // to internal fragmentation, top of page areas in map space), and the bytes 1306 // 'available' is the number of unallocated bytes that are not waste. The 1307 // capacity is the sum of size, waste, and available. 1308 // 1309 // The stats are only set by functions that ensure they stay balanced. These 1310 // functions increase or decrease one of the non-capacity stats in 1311 // conjunction with capacity, or else they always balance increases and 1312 // decreases to the non-capacity stats. 1313 class AllocationStats BASE_EMBEDDED { 1314 public: 1315 AllocationStats() { Clear(); } 1316 1317 // Zero out all the allocation statistics (i.e., no capacity). 1318 void Clear() { 1319 capacity_ = 0; 1320 max_capacity_ = 0; 1321 size_ = 0; 1322 waste_ = 0; 1323 } 1324 1325 void ClearSizeWaste() { 1326 size_ = capacity_; 1327 waste_ = 0; 1328 } 1329 1330 // Reset the allocation statistics (i.e., available = capacity with no 1331 // wasted or allocated bytes). 1332 void Reset() { 1333 size_ = 0; 1334 waste_ = 0; 1335 } 1336 1337 // Accessors for the allocation statistics. 1338 intptr_t Capacity() { return capacity_; } 1339 intptr_t MaxCapacity() { return max_capacity_; } 1340 intptr_t Size() { return size_; } 1341 intptr_t Waste() { return waste_; } 1342 1343 // Grow the space by adding available bytes. They are initially marked as 1344 // being in use (part of the size), but will normally be immediately freed, 1345 // putting them on the free list and removing them from size_. 1346 void ExpandSpace(int size_in_bytes) { 1347 capacity_ += size_in_bytes; 1348 size_ += size_in_bytes; 1349 if (capacity_ > max_capacity_) { 1350 max_capacity_ = capacity_; 1351 } 1352 DCHECK(size_ >= 0); 1353 } 1354 1355 // Shrink the space by removing available bytes. Since shrinking is done 1356 // during sweeping, bytes have been marked as being in use (part of the size) 1357 // and are hereby freed. 1358 void ShrinkSpace(int size_in_bytes) { 1359 capacity_ -= size_in_bytes; 1360 size_ -= size_in_bytes; 1361 DCHECK(size_ >= 0); 1362 } 1363 1364 // Allocate from available bytes (available -> size). 1365 void AllocateBytes(intptr_t size_in_bytes) { 1366 size_ += size_in_bytes; 1367 DCHECK(size_ >= 0); 1368 } 1369 1370 // Free allocated bytes, making them available (size -> available). 1371 void DeallocateBytes(intptr_t size_in_bytes) { 1372 size_ -= size_in_bytes; 1373 DCHECK(size_ >= 0); 1374 } 1375 1376 // Waste free bytes (available -> waste). 1377 void WasteBytes(int size_in_bytes) { 1378 DCHECK(size_in_bytes >= 0); 1379 waste_ += size_in_bytes; 1380 } 1381 1382 private: 1383 intptr_t capacity_; 1384 intptr_t max_capacity_; 1385 intptr_t size_; 1386 intptr_t waste_; 1387 }; 1388 1389 1390 // ----------------------------------------------------------------------------- 1391 // Free lists for old object spaces 1392 // 1393 // Free-list nodes are free blocks in the heap. They look like heap objects 1394 // (free-list node pointers have the heap object tag, and they have a map like 1395 // a heap object). They have a size and a next pointer. The next pointer is 1396 // the raw address of the next free list node (or NULL). 1397 class FreeListNode : public HeapObject { 1398 public: 1399 // Obtain a free-list node from a raw address. This is not a cast because 1400 // it does not check nor require that the first word at the address is a map 1401 // pointer. 1402 static FreeListNode* FromAddress(Address address) { 1403 return reinterpret_cast<FreeListNode*>(HeapObject::FromAddress(address)); 1404 } 1405 1406 static inline bool IsFreeListNode(HeapObject* object); 1407 1408 // Set the size in bytes, which can be read with HeapObject::Size(). This 1409 // function also writes a map to the first word of the block so that it 1410 // looks like a heap object to the garbage collector and heap iteration 1411 // functions. 1412 void set_size(Heap* heap, int size_in_bytes); 1413 1414 // Accessors for the next field. 1415 inline FreeListNode* next(); 1416 inline FreeListNode** next_address(); 1417 inline void set_next(FreeListNode* next); 1418 1419 inline void Zap(); 1420 1421 static inline FreeListNode* cast(Object* object) { 1422 return reinterpret_cast<FreeListNode*>(object); 1423 } 1424 1425 private: 1426 static const int kNextOffset = POINTER_SIZE_ALIGN(FreeSpace::kHeaderSize); 1427 1428 DISALLOW_IMPLICIT_CONSTRUCTORS(FreeListNode); 1429 }; 1430 1431 1432 // The free list category holds a pointer to the top element and a pointer to 1433 // the end element of the linked list of free memory blocks. 1434 class FreeListCategory { 1435 public: 1436 FreeListCategory() : top_(0), end_(NULL), available_(0) {} 1437 1438 intptr_t Concatenate(FreeListCategory* category); 1439 1440 void Reset(); 1441 1442 void Free(FreeListNode* node, int size_in_bytes); 1443 1444 FreeListNode* PickNodeFromList(int* node_size); 1445 FreeListNode* PickNodeFromList(int size_in_bytes, int* node_size); 1446 1447 intptr_t EvictFreeListItemsInList(Page* p); 1448 bool ContainsPageFreeListItemsInList(Page* p); 1449 1450 void RepairFreeList(Heap* heap); 1451 1452 FreeListNode* top() const { 1453 return reinterpret_cast<FreeListNode*>(base::NoBarrier_Load(&top_)); 1454 } 1455 1456 void set_top(FreeListNode* top) { 1457 base::NoBarrier_Store(&top_, reinterpret_cast<base::AtomicWord>(top)); 1458 } 1459 1460 FreeListNode** GetEndAddress() { return &end_; } 1461 FreeListNode* end() const { return end_; } 1462 void set_end(FreeListNode* end) { end_ = end; } 1463 1464 int* GetAvailableAddress() { return &available_; } 1465 int available() const { return available_; } 1466 void set_available(int available) { available_ = available; } 1467 1468 base::Mutex* mutex() { return &mutex_; } 1469 1470 bool IsEmpty() { return top() == 0; } 1471 1472 #ifdef DEBUG 1473 intptr_t SumFreeList(); 1474 int FreeListLength(); 1475 #endif 1476 1477 private: 1478 // top_ points to the top FreeListNode* in the free list category. 1479 base::AtomicWord top_; 1480 FreeListNode* end_; 1481 base::Mutex mutex_; 1482 1483 // Total available bytes in all blocks of this free list category. 1484 int available_; 1485 }; 1486 1487 1488 // The free list for the old space. The free list is organized in such a way 1489 // as to encourage objects allocated around the same time to be near each 1490 // other. The normal way to allocate is intended to be by bumping a 'top' 1491 // pointer until it hits a 'limit' pointer. When the limit is hit we need to 1492 // find a new space to allocate from. This is done with the free list, which 1493 // is divided up into rough categories to cut down on waste. Having finer 1494 // categories would scatter allocation more. 1495 1496 // The old space free list is organized in categories. 1497 // 1-31 words: Such small free areas are discarded for efficiency reasons. 1498 // They can be reclaimed by the compactor. However the distance between top 1499 // and limit may be this small. 1500 // 32-255 words: There is a list of spaces this large. It is used for top and 1501 // limit when the object we need to allocate is 1-31 words in size. These 1502 // spaces are called small. 1503 // 256-2047 words: There is a list of spaces this large. It is used for top and 1504 // limit when the object we need to allocate is 32-255 words in size. These 1505 // spaces are called medium. 1506 // 1048-16383 words: There is a list of spaces this large. It is used for top 1507 // and limit when the object we need to allocate is 256-2047 words in size. 1508 // These spaces are call large. 1509 // At least 16384 words. This list is for objects of 2048 words or larger. 1510 // Empty pages are added to this list. These spaces are called huge. 1511 class FreeList { 1512 public: 1513 explicit FreeList(PagedSpace* owner); 1514 1515 intptr_t Concatenate(FreeList* free_list); 1516 1517 // Clear the free list. 1518 void Reset(); 1519 1520 // Return the number of bytes available on the free list. 1521 intptr_t available() { 1522 return small_list_.available() + medium_list_.available() + 1523 large_list_.available() + huge_list_.available(); 1524 } 1525 1526 // Place a node on the free list. The block of size 'size_in_bytes' 1527 // starting at 'start' is placed on the free list. The return value is the 1528 // number of bytes that have been lost due to internal fragmentation by 1529 // freeing the block. Bookkeeping information will be written to the block, 1530 // i.e., its contents will be destroyed. The start address should be word 1531 // aligned, and the size should be a non-zero multiple of the word size. 1532 int Free(Address start, int size_in_bytes); 1533 1534 // This method returns how much memory can be allocated after freeing 1535 // maximum_freed memory. 1536 static inline int GuaranteedAllocatable(int maximum_freed) { 1537 if (maximum_freed < kSmallListMin) { 1538 return 0; 1539 } else if (maximum_freed <= kSmallListMax) { 1540 return kSmallAllocationMax; 1541 } else if (maximum_freed <= kMediumListMax) { 1542 return kMediumAllocationMax; 1543 } else if (maximum_freed <= kLargeListMax) { 1544 return kLargeAllocationMax; 1545 } 1546 return maximum_freed; 1547 } 1548 1549 // Allocate a block of size 'size_in_bytes' from the free list. The block 1550 // is unitialized. A failure is returned if no block is available. The 1551 // number of bytes lost to fragmentation is returned in the output parameter 1552 // 'wasted_bytes'. The size should be a non-zero multiple of the word size. 1553 MUST_USE_RESULT HeapObject* Allocate(int size_in_bytes); 1554 1555 bool IsEmpty() { 1556 return small_list_.IsEmpty() && medium_list_.IsEmpty() && 1557 large_list_.IsEmpty() && huge_list_.IsEmpty(); 1558 } 1559 1560 #ifdef DEBUG 1561 void Zap(); 1562 intptr_t SumFreeLists(); 1563 bool IsVeryLong(); 1564 #endif 1565 1566 // Used after booting the VM. 1567 void RepairLists(Heap* heap); 1568 1569 intptr_t EvictFreeListItems(Page* p); 1570 bool ContainsPageFreeListItems(Page* p); 1571 1572 FreeListCategory* small_list() { return &small_list_; } 1573 FreeListCategory* medium_list() { return &medium_list_; } 1574 FreeListCategory* large_list() { return &large_list_; } 1575 FreeListCategory* huge_list() { return &huge_list_; } 1576 1577 private: 1578 // The size range of blocks, in bytes. 1579 static const int kMinBlockSize = 3 * kPointerSize; 1580 static const int kMaxBlockSize = Page::kMaxRegularHeapObjectSize; 1581 1582 FreeListNode* FindNodeFor(int size_in_bytes, int* node_size); 1583 1584 PagedSpace* owner_; 1585 Heap* heap_; 1586 1587 static const int kSmallListMin = 0x20 * kPointerSize; 1588 static const int kSmallListMax = 0xff * kPointerSize; 1589 static const int kMediumListMax = 0x7ff * kPointerSize; 1590 static const int kLargeListMax = 0x3fff * kPointerSize; 1591 static const int kSmallAllocationMax = kSmallListMin - kPointerSize; 1592 static const int kMediumAllocationMax = kSmallListMax; 1593 static const int kLargeAllocationMax = kMediumListMax; 1594 FreeListCategory small_list_; 1595 FreeListCategory medium_list_; 1596 FreeListCategory large_list_; 1597 FreeListCategory huge_list_; 1598 1599 DISALLOW_IMPLICIT_CONSTRUCTORS(FreeList); 1600 }; 1601 1602 1603 class AllocationResult { 1604 public: 1605 // Implicit constructor from Object*. 1606 AllocationResult(Object* object) // NOLINT 1607 : object_(object), 1608 retry_space_(INVALID_SPACE) {} 1609 1610 AllocationResult() : object_(NULL), retry_space_(INVALID_SPACE) {} 1611 1612 static inline AllocationResult Retry(AllocationSpace space = NEW_SPACE) { 1613 return AllocationResult(space); 1614 } 1615 1616 inline bool IsRetry() { return retry_space_ != INVALID_SPACE; } 1617 1618 template <typename T> 1619 bool To(T** obj) { 1620 if (IsRetry()) return false; 1621 *obj = T::cast(object_); 1622 return true; 1623 } 1624 1625 Object* ToObjectChecked() { 1626 CHECK(!IsRetry()); 1627 return object_; 1628 } 1629 1630 AllocationSpace RetrySpace() { 1631 DCHECK(IsRetry()); 1632 return retry_space_; 1633 } 1634 1635 private: 1636 explicit AllocationResult(AllocationSpace space) 1637 : object_(NULL), retry_space_(space) {} 1638 1639 Object* object_; 1640 AllocationSpace retry_space_; 1641 }; 1642 1643 1644 class PagedSpace : public Space { 1645 public: 1646 // Creates a space with a maximum capacity, and an id. 1647 PagedSpace(Heap* heap, intptr_t max_capacity, AllocationSpace id, 1648 Executability executable); 1649 1650 virtual ~PagedSpace() {} 1651 1652 // Set up the space using the given address range of virtual memory (from 1653 // the memory allocator's initial chunk) if possible. If the block of 1654 // addresses is not big enough to contain a single page-aligned page, a 1655 // fresh chunk will be allocated. 1656 bool SetUp(); 1657 1658 // Returns true if the space has been successfully set up and not 1659 // subsequently torn down. 1660 bool HasBeenSetUp(); 1661 1662 // Cleans up the space, frees all pages in this space except those belonging 1663 // to the initial chunk, uncommits addresses in the initial chunk. 1664 void TearDown(); 1665 1666 // Checks whether an object/address is in this space. 1667 inline bool Contains(Address a); 1668 bool Contains(HeapObject* o) { return Contains(o->address()); } 1669 1670 // Given an address occupied by a live object, return that object if it is 1671 // in this space, or a Smi if it is not. The implementation iterates over 1672 // objects in the page containing the address, the cost is linear in the 1673 // number of objects in the page. It may be slow. 1674 Object* FindObject(Address addr); 1675 1676 // During boot the free_space_map is created, and afterwards we may need 1677 // to write it into the free list nodes that were already created. 1678 void RepairFreeListsAfterBoot(); 1679 1680 // Prepares for a mark-compact GC. 1681 void PrepareForMarkCompact(); 1682 1683 // Current capacity without growing (Size() + Available()). 1684 intptr_t Capacity() { return accounting_stats_.Capacity(); } 1685 1686 // Total amount of memory committed for this space. For paged 1687 // spaces this equals the capacity. 1688 intptr_t CommittedMemory() { return Capacity(); } 1689 1690 // The maximum amount of memory ever committed for this space. 1691 intptr_t MaximumCommittedMemory() { return accounting_stats_.MaxCapacity(); } 1692 1693 // Approximate amount of physical memory committed for this space. 1694 size_t CommittedPhysicalMemory(); 1695 1696 struct SizeStats { 1697 intptr_t Total() { 1698 return small_size_ + medium_size_ + large_size_ + huge_size_; 1699 } 1700 1701 intptr_t small_size_; 1702 intptr_t medium_size_; 1703 intptr_t large_size_; 1704 intptr_t huge_size_; 1705 }; 1706 1707 void ObtainFreeListStatistics(Page* p, SizeStats* sizes); 1708 void ResetFreeListStatistics(); 1709 1710 // Sets the capacity, the available space and the wasted space to zero. 1711 // The stats are rebuilt during sweeping by adding each page to the 1712 // capacity and the size when it is encountered. As free spaces are 1713 // discovered during the sweeping they are subtracted from the size and added 1714 // to the available and wasted totals. 1715 void ClearStats() { 1716 accounting_stats_.ClearSizeWaste(); 1717 ResetFreeListStatistics(); 1718 } 1719 1720 // Increases the number of available bytes of that space. 1721 void AddToAccountingStats(intptr_t bytes) { 1722 accounting_stats_.DeallocateBytes(bytes); 1723 } 1724 1725 // Available bytes without growing. These are the bytes on the free list. 1726 // The bytes in the linear allocation area are not included in this total 1727 // because updating the stats would slow down allocation. New pages are 1728 // immediately added to the free list so they show up here. 1729 intptr_t Available() { return free_list_.available(); } 1730 1731 // Allocated bytes in this space. Garbage bytes that were not found due to 1732 // concurrent sweeping are counted as being allocated! The bytes in the 1733 // current linear allocation area (between top and limit) are also counted 1734 // here. 1735 virtual intptr_t Size() { return accounting_stats_.Size(); } 1736 1737 // As size, but the bytes in lazily swept pages are estimated and the bytes 1738 // in the current linear allocation area are not included. 1739 virtual intptr_t SizeOfObjects(); 1740 1741 // Wasted bytes in this space. These are just the bytes that were thrown away 1742 // due to being too small to use for allocation. They do not include the 1743 // free bytes that were not found at all due to lazy sweeping. 1744 virtual intptr_t Waste() { return accounting_stats_.Waste(); } 1745 1746 // Returns the allocation pointer in this space. 1747 Address top() { return allocation_info_.top(); } 1748 Address limit() { return allocation_info_.limit(); } 1749 1750 // The allocation top address. 1751 Address* allocation_top_address() { return allocation_info_.top_address(); } 1752 1753 // The allocation limit address. 1754 Address* allocation_limit_address() { 1755 return allocation_info_.limit_address(); 1756 } 1757 1758 // Allocate the requested number of bytes in the space if possible, return a 1759 // failure object if not. 1760 MUST_USE_RESULT inline AllocationResult AllocateRaw(int size_in_bytes); 1761 1762 // Give a block of memory to the space's free list. It might be added to 1763 // the free list or accounted as waste. 1764 // If add_to_freelist is false then just accounting stats are updated and 1765 // no attempt to add area to free list is made. 1766 int Free(Address start, int size_in_bytes) { 1767 int wasted = free_list_.Free(start, size_in_bytes); 1768 accounting_stats_.DeallocateBytes(size_in_bytes); 1769 accounting_stats_.WasteBytes(wasted); 1770 return size_in_bytes - wasted; 1771 } 1772 1773 void ResetFreeList() { free_list_.Reset(); } 1774 1775 // Set space allocation info. 1776 void SetTopAndLimit(Address top, Address limit) { 1777 DCHECK(top == limit || 1778 Page::FromAddress(top) == Page::FromAddress(limit - 1)); 1779 MemoryChunk::UpdateHighWaterMark(allocation_info_.top()); 1780 allocation_info_.set_top(top); 1781 allocation_info_.set_limit(limit); 1782 } 1783 1784 // Empty space allocation info, returning unused area to free list. 1785 void EmptyAllocationInfo() { 1786 // Mark the old linear allocation area with a free space map so it can be 1787 // skipped when scanning the heap. 1788 int old_linear_size = static_cast<int>(limit() - top()); 1789 Free(top(), old_linear_size); 1790 SetTopAndLimit(NULL, NULL); 1791 } 1792 1793 void Allocate(int bytes) { accounting_stats_.AllocateBytes(bytes); } 1794 1795 void IncreaseCapacity(int size); 1796 1797 // Releases an unused page and shrinks the space. 1798 void ReleasePage(Page* page); 1799 1800 // The dummy page that anchors the linked list of pages. 1801 Page* anchor() { return &anchor_; } 1802 1803 #ifdef VERIFY_HEAP 1804 // Verify integrity of this space. 1805 virtual void Verify(ObjectVisitor* visitor); 1806 1807 // Overridden by subclasses to verify space-specific object 1808 // properties (e.g., only maps or free-list nodes are in map space). 1809 virtual void VerifyObject(HeapObject* obj) {} 1810 #endif 1811 1812 #ifdef DEBUG 1813 // Print meta info and objects in this space. 1814 virtual void Print(); 1815 1816 // Reports statistics for the space 1817 void ReportStatistics(); 1818 1819 // Report code object related statistics 1820 void CollectCodeStatistics(); 1821 static void ReportCodeStatistics(Isolate* isolate); 1822 static void ResetCodeStatistics(Isolate* isolate); 1823 #endif 1824 1825 // Evacuation candidates are swept by evacuator. Needs to return a valid 1826 // result before _and_ after evacuation has finished. 1827 static bool ShouldBeSweptBySweeperThreads(Page* p) { 1828 return !p->IsEvacuationCandidate() && 1829 !p->IsFlagSet(Page::RESCAN_ON_EVACUATION) && !p->WasSwept(); 1830 } 1831 1832 void IncrementUnsweptFreeBytes(intptr_t by) { unswept_free_bytes_ += by; } 1833 1834 void IncreaseUnsweptFreeBytes(Page* p) { 1835 DCHECK(ShouldBeSweptBySweeperThreads(p)); 1836 unswept_free_bytes_ += (p->area_size() - p->LiveBytes()); 1837 } 1838 1839 void DecrementUnsweptFreeBytes(intptr_t by) { unswept_free_bytes_ -= by; } 1840 1841 void DecreaseUnsweptFreeBytes(Page* p) { 1842 DCHECK(ShouldBeSweptBySweeperThreads(p)); 1843 unswept_free_bytes_ -= (p->area_size() - p->LiveBytes()); 1844 } 1845 1846 void ResetUnsweptFreeBytes() { unswept_free_bytes_ = 0; } 1847 1848 // This function tries to steal size_in_bytes memory from the sweeper threads 1849 // free-lists. If it does not succeed stealing enough memory, it will wait 1850 // for the sweeper threads to finish sweeping. 1851 // It returns true when sweeping is completed and false otherwise. 1852 bool EnsureSweeperProgress(intptr_t size_in_bytes); 1853 1854 void set_end_of_unswept_pages(Page* page) { end_of_unswept_pages_ = page; } 1855 1856 Page* end_of_unswept_pages() { return end_of_unswept_pages_; } 1857 1858 Page* FirstPage() { return anchor_.next_page(); } 1859 Page* LastPage() { return anchor_.prev_page(); } 1860 1861 void EvictEvacuationCandidatesFromFreeLists(); 1862 1863 bool CanExpand(); 1864 1865 // Returns the number of total pages in this space. 1866 int CountTotalPages(); 1867 1868 // Return size of allocatable area on a page in this space. 1869 inline int AreaSize() { return area_size_; } 1870 1871 void CreateEmergencyMemory(); 1872 void FreeEmergencyMemory(); 1873 void UseEmergencyMemory(); 1874 1875 bool HasEmergencyMemory() { return emergency_memory_ != NULL; } 1876 1877 protected: 1878 FreeList* free_list() { return &free_list_; } 1879 1880 int area_size_; 1881 1882 // Maximum capacity of this space. 1883 intptr_t max_capacity_; 1884 1885 intptr_t SizeOfFirstPage(); 1886 1887 // Accounting information for this space. 1888 AllocationStats accounting_stats_; 1889 1890 // The dummy page that anchors the double linked list of pages. 1891 Page anchor_; 1892 1893 // The space's free list. 1894 FreeList free_list_; 1895 1896 // Normal allocation information. 1897 AllocationInfo allocation_info_; 1898 1899 // The number of free bytes which could be reclaimed by advancing the 1900 // concurrent sweeper threads. 1901 intptr_t unswept_free_bytes_; 1902 1903 // The sweeper threads iterate over the list of pointer and data space pages 1904 // and sweep these pages concurrently. They will stop sweeping after the 1905 // end_of_unswept_pages_ page. 1906 Page* end_of_unswept_pages_; 1907 1908 // Emergency memory is the memory of a full page for a given space, allocated 1909 // conservatively before evacuating a page. If compaction fails due to out 1910 // of memory error the emergency memory can be used to complete compaction. 1911 // If not used, the emergency memory is released after compaction. 1912 MemoryChunk* emergency_memory_; 1913 1914 // Expands the space by allocating a fixed number of pages. Returns false if 1915 // it cannot allocate requested number of pages from OS, or if the hard heap 1916 // size limit has been hit. 1917 bool Expand(); 1918 1919 // Generic fast case allocation function that tries linear allocation at the 1920 // address denoted by top in allocation_info_. 1921 inline HeapObject* AllocateLinearly(int size_in_bytes); 1922 1923 // If sweeping is still in progress try to sweep unswept pages. If that is 1924 // not successful, wait for the sweeper threads and re-try free-list 1925 // allocation. 1926 MUST_USE_RESULT HeapObject* WaitForSweeperThreadsAndRetryAllocation( 1927 int size_in_bytes); 1928 1929 // Slow path of AllocateRaw. This function is space-dependent. 1930 MUST_USE_RESULT HeapObject* SlowAllocateRaw(int size_in_bytes); 1931 1932 friend class PageIterator; 1933 friend class MarkCompactCollector; 1934 }; 1935 1936 1937 class NumberAndSizeInfo BASE_EMBEDDED { 1938 public: 1939 NumberAndSizeInfo() : number_(0), bytes_(0) {} 1940 1941 int number() const { return number_; } 1942 void increment_number(int num) { number_ += num; } 1943 1944 int bytes() const { return bytes_; } 1945 void increment_bytes(int size) { bytes_ += size; } 1946 1947 void clear() { 1948 number_ = 0; 1949 bytes_ = 0; 1950 } 1951 1952 private: 1953 int number_; 1954 int bytes_; 1955 }; 1956 1957 1958 // HistogramInfo class for recording a single "bar" of a histogram. This 1959 // class is used for collecting statistics to print to the log file. 1960 class HistogramInfo : public NumberAndSizeInfo { 1961 public: 1962 HistogramInfo() : NumberAndSizeInfo() {} 1963 1964 const char* name() { return name_; } 1965 void set_name(const char* name) { name_ = name; } 1966 1967 private: 1968 const char* name_; 1969 }; 1970 1971 1972 enum SemiSpaceId { kFromSpace = 0, kToSpace = 1 }; 1973 1974 1975 class SemiSpace; 1976 1977 1978 class NewSpacePage : public MemoryChunk { 1979 public: 1980 // GC related flags copied from from-space to to-space when 1981 // flipping semispaces. 1982 static const intptr_t kCopyOnFlipFlagsMask = 1983 (1 << MemoryChunk::POINTERS_TO_HERE_ARE_INTERESTING) | 1984 (1 << MemoryChunk::POINTERS_FROM_HERE_ARE_INTERESTING) | 1985 (1 << MemoryChunk::SCAN_ON_SCAVENGE); 1986 1987 static const int kAreaSize = Page::kMaxRegularHeapObjectSize; 1988 1989 inline NewSpacePage* next_page() const { 1990 return static_cast<NewSpacePage*>(next_chunk()); 1991 } 1992 1993 inline void set_next_page(NewSpacePage* page) { set_next_chunk(page); } 1994 1995 inline NewSpacePage* prev_page() const { 1996 return static_cast<NewSpacePage*>(prev_chunk()); 1997 } 1998 1999 inline void set_prev_page(NewSpacePage* page) { set_prev_chunk(page); } 2000 2001 SemiSpace* semi_space() { return reinterpret_cast<SemiSpace*>(owner()); } 2002 2003 bool is_anchor() { return !this->InNewSpace(); } 2004 2005 static bool IsAtStart(Address addr) { 2006 return (reinterpret_cast<intptr_t>(addr) & Page::kPageAlignmentMask) == 2007 kObjectStartOffset; 2008 } 2009 2010 static bool IsAtEnd(Address addr) { 2011 return (reinterpret_cast<intptr_t>(addr) & Page::kPageAlignmentMask) == 0; 2012 } 2013 2014 Address address() { return reinterpret_cast<Address>(this); } 2015 2016 // Finds the NewSpacePage containing the given address. 2017 static inline NewSpacePage* FromAddress(Address address_in_page) { 2018 Address page_start = 2019 reinterpret_cast<Address>(reinterpret_cast<uintptr_t>(address_in_page) & 2020 ~Page::kPageAlignmentMask); 2021 NewSpacePage* page = reinterpret_cast<NewSpacePage*>(page_start); 2022 return page; 2023 } 2024 2025 // Find the page for a limit address. A limit address is either an address 2026 // inside a page, or the address right after the last byte of a page. 2027 static inline NewSpacePage* FromLimit(Address address_limit) { 2028 return NewSpacePage::FromAddress(address_limit - 1); 2029 } 2030 2031 // Checks if address1 and address2 are on the same new space page. 2032 static inline bool OnSamePage(Address address1, Address address2) { 2033 return NewSpacePage::FromAddress(address1) == 2034 NewSpacePage::FromAddress(address2); 2035 } 2036 2037 private: 2038 // Create a NewSpacePage object that is only used as anchor 2039 // for the doubly-linked list of real pages. 2040 explicit NewSpacePage(SemiSpace* owner) { InitializeAsAnchor(owner); } 2041 2042 static NewSpacePage* Initialize(Heap* heap, Address start, 2043 SemiSpace* semi_space); 2044 2045 // Intialize a fake NewSpacePage used as sentinel at the ends 2046 // of a doubly-linked list of real NewSpacePages. 2047 // Only uses the prev/next links, and sets flags to not be in new-space. 2048 void InitializeAsAnchor(SemiSpace* owner); 2049 2050 friend class SemiSpace; 2051 friend class SemiSpaceIterator; 2052 }; 2053 2054 2055 // ----------------------------------------------------------------------------- 2056 // SemiSpace in young generation 2057 // 2058 // A semispace is a contiguous chunk of memory holding page-like memory 2059 // chunks. The mark-compact collector uses the memory of the first page in 2060 // the from space as a marking stack when tracing live objects. 2061 2062 class SemiSpace : public Space { 2063 public: 2064 // Constructor. 2065 SemiSpace(Heap* heap, SemiSpaceId semispace) 2066 : Space(heap, NEW_SPACE, NOT_EXECUTABLE), 2067 start_(NULL), 2068 age_mark_(NULL), 2069 id_(semispace), 2070 anchor_(this), 2071 current_page_(NULL) {} 2072 2073 // Sets up the semispace using the given chunk. 2074 void SetUp(Address start, int initial_capacity, int maximum_capacity); 2075 2076 // Tear down the space. Heap memory was not allocated by the space, so it 2077 // is not deallocated here. 2078 void TearDown(); 2079 2080 // True if the space has been set up but not torn down. 2081 bool HasBeenSetUp() { return start_ != NULL; } 2082 2083 // Grow the semispace to the new capacity. The new capacity 2084 // requested must be larger than the current capacity and less than 2085 // the maximum capacity. 2086 bool GrowTo(int new_capacity); 2087 2088 // Shrinks the semispace to the new capacity. The new capacity 2089 // requested must be more than the amount of used memory in the 2090 // semispace and less than the current capacity. 2091 bool ShrinkTo(int new_capacity); 2092 2093 // Returns the start address of the first page of the space. 2094 Address space_start() { 2095 DCHECK(anchor_.next_page() != &anchor_); 2096 return anchor_.next_page()->area_start(); 2097 } 2098 2099 // Returns the start address of the current page of the space. 2100 Address page_low() { return current_page_->area_start(); } 2101 2102 // Returns one past the end address of the space. 2103 Address space_end() { return anchor_.prev_page()->area_end(); } 2104 2105 // Returns one past the end address of the current page of the space. 2106 Address page_high() { return current_page_->area_end(); } 2107 2108 bool AdvancePage() { 2109 NewSpacePage* next_page = current_page_->next_page(); 2110 if (next_page == anchor()) return false; 2111 current_page_ = next_page; 2112 return true; 2113 } 2114 2115 // Resets the space to using the first page. 2116 void Reset(); 2117 2118 // Age mark accessors. 2119 Address age_mark() { return age_mark_; } 2120 void set_age_mark(Address mark); 2121 2122 // True if the address is in the address range of this semispace (not 2123 // necessarily below the allocation pointer). 2124 bool Contains(Address a) { 2125 return (reinterpret_cast<uintptr_t>(a) & address_mask_) == 2126 reinterpret_cast<uintptr_t>(start_); 2127 } 2128 2129 // True if the object is a heap object in the address range of this 2130 // semispace (not necessarily below the allocation pointer). 2131 bool Contains(Object* o) { 2132 return (reinterpret_cast<uintptr_t>(o) & object_mask_) == object_expected_; 2133 } 2134 2135 // If we don't have these here then SemiSpace will be abstract. However 2136 // they should never be called. 2137 virtual intptr_t Size() { 2138 UNREACHABLE(); 2139 return 0; 2140 } 2141 2142 bool is_committed() { return committed_; } 2143 bool Commit(); 2144 bool Uncommit(); 2145 2146 NewSpacePage* first_page() { return anchor_.next_page(); } 2147 NewSpacePage* current_page() { return current_page_; } 2148 2149 #ifdef VERIFY_HEAP 2150 virtual void Verify(); 2151 #endif 2152 2153 #ifdef DEBUG 2154 virtual void Print(); 2155 // Validate a range of of addresses in a SemiSpace. 2156 // The "from" address must be on a page prior to the "to" address, 2157 // in the linked page order, or it must be earlier on the same page. 2158 static void AssertValidRange(Address from, Address to); 2159 #else 2160 // Do nothing. 2161 inline static void AssertValidRange(Address from, Address to) {} 2162 #endif 2163 2164 // Returns the current total capacity of the semispace. 2165 int TotalCapacity() { return total_capacity_; } 2166 2167 // Returns the maximum total capacity of the semispace. 2168 int MaximumTotalCapacity() { return maximum_total_capacity_; } 2169 2170 // Returns the initial capacity of the semispace. 2171 int InitialTotalCapacity() { return initial_total_capacity_; } 2172 2173 SemiSpaceId id() { return id_; } 2174 2175 static void Swap(SemiSpace* from, SemiSpace* to); 2176 2177 // Returns the maximum amount of memory ever committed by the semi space. 2178 size_t MaximumCommittedMemory() { return maximum_committed_; } 2179 2180 // Approximate amount of physical memory committed for this space. 2181 size_t CommittedPhysicalMemory(); 2182 2183 private: 2184 // Flips the semispace between being from-space and to-space. 2185 // Copies the flags into the masked positions on all pages in the space. 2186 void FlipPages(intptr_t flags, intptr_t flag_mask); 2187 2188 // Updates Capacity and MaximumCommitted based on new capacity. 2189 void SetCapacity(int new_capacity); 2190 2191 NewSpacePage* anchor() { return &anchor_; } 2192 2193 // The current and maximum total capacity of the space. 2194 int total_capacity_; 2195 int maximum_total_capacity_; 2196 int initial_total_capacity_; 2197 2198 intptr_t maximum_committed_; 2199 2200 // The start address of the space. 2201 Address start_; 2202 // Used to govern object promotion during mark-compact collection. 2203 Address age_mark_; 2204 2205 // Masks and comparison values to test for containment in this semispace. 2206 uintptr_t address_mask_; 2207 uintptr_t object_mask_; 2208 uintptr_t object_expected_; 2209 2210 bool committed_; 2211 SemiSpaceId id_; 2212 2213 NewSpacePage anchor_; 2214 NewSpacePage* current_page_; 2215 2216 friend class SemiSpaceIterator; 2217 friend class NewSpacePageIterator; 2218 2219 public: 2220 TRACK_MEMORY("SemiSpace") 2221 }; 2222 2223 2224 // A SemiSpaceIterator is an ObjectIterator that iterates over the active 2225 // semispace of the heap's new space. It iterates over the objects in the 2226 // semispace from a given start address (defaulting to the bottom of the 2227 // semispace) to the top of the semispace. New objects allocated after the 2228 // iterator is created are not iterated. 2229 class SemiSpaceIterator : public ObjectIterator { 2230 public: 2231 // Create an iterator over the objects in the given space. If no start 2232 // address is given, the iterator starts from the bottom of the space. If 2233 // no size function is given, the iterator calls Object::Size(). 2234 2235 // Iterate over all of allocated to-space. 2236 explicit SemiSpaceIterator(NewSpace* space); 2237 // Iterate over all of allocated to-space, with a custome size function. 2238 SemiSpaceIterator(NewSpace* space, HeapObjectCallback size_func); 2239 // Iterate over part of allocated to-space, from start to the end 2240 // of allocation. 2241 SemiSpaceIterator(NewSpace* space, Address start); 2242 // Iterate from one address to another in the same semi-space. 2243 SemiSpaceIterator(Address from, Address to); 2244 2245 HeapObject* Next() { 2246 if (current_ == limit_) return NULL; 2247 if (NewSpacePage::IsAtEnd(current_)) { 2248 NewSpacePage* page = NewSpacePage::FromLimit(current_); 2249 page = page->next_page(); 2250 DCHECK(!page->is_anchor()); 2251 current_ = page->area_start(); 2252 if (current_ == limit_) return NULL; 2253 } 2254 2255 HeapObject* object = HeapObject::FromAddress(current_); 2256 int size = (size_func_ == NULL) ? object->Size() : size_func_(object); 2257 2258 current_ += size; 2259 return object; 2260 } 2261 2262 // Implementation of the ObjectIterator functions. 2263 virtual HeapObject* next_object() { return Next(); } 2264 2265 private: 2266 void Initialize(Address start, Address end, HeapObjectCallback size_func); 2267 2268 // The current iteration point. 2269 Address current_; 2270 // The end of iteration. 2271 Address limit_; 2272 // The callback function. 2273 HeapObjectCallback size_func_; 2274 }; 2275 2276 2277 // ----------------------------------------------------------------------------- 2278 // A PageIterator iterates the pages in a semi-space. 2279 class NewSpacePageIterator BASE_EMBEDDED { 2280 public: 2281 // Make an iterator that runs over all pages in to-space. 2282 explicit inline NewSpacePageIterator(NewSpace* space); 2283 2284 // Make an iterator that runs over all pages in the given semispace, 2285 // even those not used in allocation. 2286 explicit inline NewSpacePageIterator(SemiSpace* space); 2287 2288 // Make iterator that iterates from the page containing start 2289 // to the page that contains limit in the same semispace. 2290 inline NewSpacePageIterator(Address start, Address limit); 2291 2292 inline bool has_next(); 2293 inline NewSpacePage* next(); 2294 2295 private: 2296 NewSpacePage* prev_page_; // Previous page returned. 2297 // Next page that will be returned. Cached here so that we can use this 2298 // iterator for operations that deallocate pages. 2299 NewSpacePage* next_page_; 2300 // Last page returned. 2301 NewSpacePage* last_page_; 2302 }; 2303 2304 2305 // ----------------------------------------------------------------------------- 2306 // The young generation space. 2307 // 2308 // The new space consists of a contiguous pair of semispaces. It simply 2309 // forwards most functions to the appropriate semispace. 2310 2311 class NewSpace : public Space { 2312 public: 2313 // Constructor. 2314 explicit NewSpace(Heap* heap) 2315 : Space(heap, NEW_SPACE, NOT_EXECUTABLE), 2316 to_space_(heap, kToSpace), 2317 from_space_(heap, kFromSpace), 2318 reservation_(), 2319 inline_allocation_limit_step_(0) {} 2320 2321 // Sets up the new space using the given chunk. 2322 bool SetUp(int reserved_semispace_size_, int max_semi_space_size); 2323 2324 // Tears down the space. Heap memory was not allocated by the space, so it 2325 // is not deallocated here. 2326 void TearDown(); 2327 2328 // True if the space has been set up but not torn down. 2329 bool HasBeenSetUp() { 2330 return to_space_.HasBeenSetUp() && from_space_.HasBeenSetUp(); 2331 } 2332 2333 // Flip the pair of spaces. 2334 void Flip(); 2335 2336 // Grow the capacity of the semispaces. Assumes that they are not at 2337 // their maximum capacity. 2338 void Grow(); 2339 2340 // Shrink the capacity of the semispaces. 2341 void Shrink(); 2342 2343 // True if the address or object lies in the address range of either 2344 // semispace (not necessarily below the allocation pointer). 2345 bool Contains(Address a) { 2346 return (reinterpret_cast<uintptr_t>(a) & address_mask_) == 2347 reinterpret_cast<uintptr_t>(start_); 2348 } 2349 2350 bool Contains(Object* o) { 2351 Address a = reinterpret_cast<Address>(o); 2352 return (reinterpret_cast<uintptr_t>(a) & object_mask_) == object_expected_; 2353 } 2354 2355 // Return the allocated bytes in the active semispace. 2356 virtual intptr_t Size() { 2357 return pages_used_ * NewSpacePage::kAreaSize + 2358 static_cast<int>(top() - to_space_.page_low()); 2359 } 2360 2361 // The same, but returning an int. We have to have the one that returns 2362 // intptr_t because it is inherited, but if we know we are dealing with the 2363 // new space, which can't get as big as the other spaces then this is useful: 2364 int SizeAsInt() { return static_cast<int>(Size()); } 2365 2366 // Return the allocatable capacity of a semispace. 2367 intptr_t Capacity() { 2368 SLOW_DCHECK(to_space_.TotalCapacity() == from_space_.TotalCapacity()); 2369 return (to_space_.TotalCapacity() / Page::kPageSize) * 2370 NewSpacePage::kAreaSize; 2371 } 2372 2373 // Return the current size of a semispace, allocatable and non-allocatable 2374 // memory. 2375 intptr_t TotalCapacity() { 2376 DCHECK(to_space_.TotalCapacity() == from_space_.TotalCapacity()); 2377 return to_space_.TotalCapacity(); 2378 } 2379 2380 // Return the total amount of memory committed for new space. 2381 intptr_t CommittedMemory() { 2382 if (from_space_.is_committed()) return 2 * Capacity(); 2383 return TotalCapacity(); 2384 } 2385 2386 // Return the total amount of memory committed for new space. 2387 intptr_t MaximumCommittedMemory() { 2388 return to_space_.MaximumCommittedMemory() + 2389 from_space_.MaximumCommittedMemory(); 2390 } 2391 2392 // Approximate amount of physical memory committed for this space. 2393 size_t CommittedPhysicalMemory(); 2394 2395 // Return the available bytes without growing. 2396 intptr_t Available() { return Capacity() - Size(); } 2397 2398 // Return the maximum capacity of a semispace. 2399 int MaximumCapacity() { 2400 DCHECK(to_space_.MaximumTotalCapacity() == 2401 from_space_.MaximumTotalCapacity()); 2402 return to_space_.MaximumTotalCapacity(); 2403 } 2404 2405 bool IsAtMaximumCapacity() { return TotalCapacity() == MaximumCapacity(); } 2406 2407 // Returns the initial capacity of a semispace. 2408 int InitialTotalCapacity() { 2409 DCHECK(to_space_.InitialTotalCapacity() == 2410 from_space_.InitialTotalCapacity()); 2411 return to_space_.InitialTotalCapacity(); 2412 } 2413 2414 // Return the address of the allocation pointer in the active semispace. 2415 Address top() { 2416 DCHECK(to_space_.current_page()->ContainsLimit(allocation_info_.top())); 2417 return allocation_info_.top(); 2418 } 2419 2420 void set_top(Address top) { 2421 DCHECK(to_space_.current_page()->ContainsLimit(top)); 2422 allocation_info_.set_top(top); 2423 } 2424 2425 // Return the address of the allocation pointer limit in the active semispace. 2426 Address limit() { 2427 DCHECK(to_space_.current_page()->ContainsLimit(allocation_info_.limit())); 2428 return allocation_info_.limit(); 2429 } 2430 2431 // Return the address of the first object in the active semispace. 2432 Address bottom() { return to_space_.space_start(); } 2433 2434 // Get the age mark of the inactive semispace. 2435 Address age_mark() { return from_space_.age_mark(); } 2436 // Set the age mark in the active semispace. 2437 void set_age_mark(Address mark) { to_space_.set_age_mark(mark); } 2438 2439 // The start address of the space and a bit mask. Anding an address in the 2440 // new space with the mask will result in the start address. 2441 Address start() { return start_; } 2442 uintptr_t mask() { return address_mask_; } 2443 2444 INLINE(uint32_t AddressToMarkbitIndex(Address addr)) { 2445 DCHECK(Contains(addr)); 2446 DCHECK(IsAligned(OffsetFrom(addr), kPointerSize) || 2447 IsAligned(OffsetFrom(addr) - 1, kPointerSize)); 2448 return static_cast<uint32_t>(addr - start_) >> kPointerSizeLog2; 2449 } 2450 2451 INLINE(Address MarkbitIndexToAddress(uint32_t index)) { 2452 return reinterpret_cast<Address>(index << kPointerSizeLog2); 2453 } 2454 2455 // The allocation top and limit address. 2456 Address* allocation_top_address() { return allocation_info_.top_address(); } 2457 2458 // The allocation limit address. 2459 Address* allocation_limit_address() { 2460 return allocation_info_.limit_address(); 2461 } 2462 2463 MUST_USE_RESULT INLINE(AllocationResult AllocateRaw(int size_in_bytes)); 2464 2465 // Reset the allocation pointer to the beginning of the active semispace. 2466 void ResetAllocationInfo(); 2467 2468 void UpdateInlineAllocationLimit(int size_in_bytes); 2469 void LowerInlineAllocationLimit(intptr_t step) { 2470 inline_allocation_limit_step_ = step; 2471 UpdateInlineAllocationLimit(0); 2472 top_on_previous_step_ = allocation_info_.top(); 2473 } 2474 2475 // Get the extent of the inactive semispace (for use as a marking stack, 2476 // or to zap it). Notice: space-addresses are not necessarily on the 2477 // same page, so FromSpaceStart() might be above FromSpaceEnd(). 2478 Address FromSpacePageLow() { return from_space_.page_low(); } 2479 Address FromSpacePageHigh() { return from_space_.page_high(); } 2480 Address FromSpaceStart() { return from_space_.space_start(); } 2481 Address FromSpaceEnd() { return from_space_.space_end(); } 2482 2483 // Get the extent of the active semispace's pages' memory. 2484 Address ToSpaceStart() { return to_space_.space_start(); } 2485 Address ToSpaceEnd() { return to_space_.space_end(); } 2486 2487 inline bool ToSpaceContains(Address address) { 2488 return to_space_.Contains(address); 2489 } 2490 inline bool FromSpaceContains(Address address) { 2491 return from_space_.Contains(address); 2492 } 2493 2494 // True if the object is a heap object in the address range of the 2495 // respective semispace (not necessarily below the allocation pointer of the 2496 // semispace). 2497 inline bool ToSpaceContains(Object* o) { return to_space_.Contains(o); } 2498 inline bool FromSpaceContains(Object* o) { return from_space_.Contains(o); } 2499 2500 // Try to switch the active semispace to a new, empty, page. 2501 // Returns false if this isn't possible or reasonable (i.e., there 2502 // are no pages, or the current page is already empty), or true 2503 // if successful. 2504 bool AddFreshPage(); 2505 2506 #ifdef VERIFY_HEAP 2507 // Verify the active semispace. 2508 virtual void Verify(); 2509 #endif 2510 2511 #ifdef DEBUG 2512 // Print the active semispace. 2513 virtual void Print() { to_space_.Print(); } 2514 #endif 2515 2516 // Iterates the active semispace to collect statistics. 2517 void CollectStatistics(); 2518 // Reports previously collected statistics of the active semispace. 2519 void ReportStatistics(); 2520 // Clears previously collected statistics. 2521 void ClearHistograms(); 2522 2523 // Record the allocation or promotion of a heap object. Note that we don't 2524 // record every single allocation, but only those that happen in the 2525 // to space during a scavenge GC. 2526 void RecordAllocation(HeapObject* obj); 2527 void RecordPromotion(HeapObject* obj); 2528 2529 // Return whether the operation succeded. 2530 bool CommitFromSpaceIfNeeded() { 2531 if (from_space_.is_committed()) return true; 2532 return from_space_.Commit(); 2533 } 2534 2535 bool UncommitFromSpace() { 2536 if (!from_space_.is_committed()) return true; 2537 return from_space_.Uncommit(); 2538 } 2539 2540 inline intptr_t inline_allocation_limit_step() { 2541 return inline_allocation_limit_step_; 2542 } 2543 2544 SemiSpace* active_space() { return &to_space_; } 2545 2546 private: 2547 // Update allocation info to match the current to-space page. 2548 void UpdateAllocationInfo(); 2549 2550 Address chunk_base_; 2551 uintptr_t chunk_size_; 2552 2553 // The semispaces. 2554 SemiSpace to_space_; 2555 SemiSpace from_space_; 2556 base::VirtualMemory reservation_; 2557 int pages_used_; 2558 2559 // Start address and bit mask for containment testing. 2560 Address start_; 2561 uintptr_t address_mask_; 2562 uintptr_t object_mask_; 2563 uintptr_t object_expected_; 2564 2565 // Allocation pointer and limit for normal allocation and allocation during 2566 // mark-compact collection. 2567 AllocationInfo allocation_info_; 2568 2569 // When incremental marking is active we will set allocation_info_.limit 2570 // to be lower than actual limit and then will gradually increase it 2571 // in steps to guarantee that we do incremental marking steps even 2572 // when all allocation is performed from inlined generated code. 2573 intptr_t inline_allocation_limit_step_; 2574 2575 Address top_on_previous_step_; 2576 2577 HistogramInfo* allocated_histogram_; 2578 HistogramInfo* promoted_histogram_; 2579 2580 MUST_USE_RESULT AllocationResult SlowAllocateRaw(int size_in_bytes); 2581 2582 friend class SemiSpaceIterator; 2583 2584 public: 2585 TRACK_MEMORY("NewSpace") 2586 }; 2587 2588 2589 // ----------------------------------------------------------------------------- 2590 // Old object space (excluding map objects) 2591 2592 class OldSpace : public PagedSpace { 2593 public: 2594 // Creates an old space object with a given maximum capacity. 2595 // The constructor does not allocate pages from OS. 2596 OldSpace(Heap* heap, intptr_t max_capacity, AllocationSpace id, 2597 Executability executable) 2598 : PagedSpace(heap, max_capacity, id, executable) {} 2599 2600 public: 2601 TRACK_MEMORY("OldSpace") 2602 }; 2603 2604 2605 // For contiguous spaces, top should be in the space (or at the end) and limit 2606 // should be the end of the space. 2607 #define DCHECK_SEMISPACE_ALLOCATION_INFO(info, space) \ 2608 SLOW_DCHECK((space).page_low() <= (info).top() && \ 2609 (info).top() <= (space).page_high() && \ 2610 (info).limit() <= (space).page_high()) 2611 2612 2613 // ----------------------------------------------------------------------------- 2614 // Old space for all map objects 2615 2616 class MapSpace : public PagedSpace { 2617 public: 2618 // Creates a map space object with a maximum capacity. 2619 MapSpace(Heap* heap, intptr_t max_capacity, AllocationSpace id) 2620 : PagedSpace(heap, max_capacity, id, NOT_EXECUTABLE), 2621 max_map_space_pages_(kMaxMapPageIndex - 1) {} 2622 2623 // Given an index, returns the page address. 2624 // TODO(1600): this limit is artifical just to keep code compilable 2625 static const int kMaxMapPageIndex = 1 << 16; 2626 2627 virtual int RoundSizeDownToObjectAlignment(int size) { 2628 if (base::bits::IsPowerOfTwo32(Map::kSize)) { 2629 return RoundDown(size, Map::kSize); 2630 } else { 2631 return (size / Map::kSize) * Map::kSize; 2632 } 2633 } 2634 2635 protected: 2636 virtual void VerifyObject(HeapObject* obj); 2637 2638 private: 2639 static const int kMapsPerPage = Page::kMaxRegularHeapObjectSize / Map::kSize; 2640 2641 // Do map space compaction if there is a page gap. 2642 int CompactionThreshold() { 2643 return kMapsPerPage * (max_map_space_pages_ - 1); 2644 } 2645 2646 const int max_map_space_pages_; 2647 2648 public: 2649 TRACK_MEMORY("MapSpace") 2650 }; 2651 2652 2653 // ----------------------------------------------------------------------------- 2654 // Old space for simple property cell objects 2655 2656 class CellSpace : public PagedSpace { 2657 public: 2658 // Creates a property cell space object with a maximum capacity. 2659 CellSpace(Heap* heap, intptr_t max_capacity, AllocationSpace id) 2660 : PagedSpace(heap, max_capacity, id, NOT_EXECUTABLE) {} 2661 2662 virtual int RoundSizeDownToObjectAlignment(int size) { 2663 if (base::bits::IsPowerOfTwo32(Cell::kSize)) { 2664 return RoundDown(size, Cell::kSize); 2665 } else { 2666 return (size / Cell::kSize) * Cell::kSize; 2667 } 2668 } 2669 2670 protected: 2671 virtual void VerifyObject(HeapObject* obj); 2672 2673 public: 2674 TRACK_MEMORY("CellSpace") 2675 }; 2676 2677 2678 // ----------------------------------------------------------------------------- 2679 // Old space for all global object property cell objects 2680 2681 class PropertyCellSpace : public PagedSpace { 2682 public: 2683 // Creates a property cell space object with a maximum capacity. 2684 PropertyCellSpace(Heap* heap, intptr_t max_capacity, AllocationSpace id) 2685 : PagedSpace(heap, max_capacity, id, NOT_EXECUTABLE) {} 2686 2687 virtual int RoundSizeDownToObjectAlignment(int size) { 2688 if (base::bits::IsPowerOfTwo32(PropertyCell::kSize)) { 2689 return RoundDown(size, PropertyCell::kSize); 2690 } else { 2691 return (size / PropertyCell::kSize) * PropertyCell::kSize; 2692 } 2693 } 2694 2695 protected: 2696 virtual void VerifyObject(HeapObject* obj); 2697 2698 public: 2699 TRACK_MEMORY("PropertyCellSpace") 2700 }; 2701 2702 2703 // ----------------------------------------------------------------------------- 2704 // Large objects ( > Page::kMaxHeapObjectSize ) are allocated and managed by 2705 // the large object space. A large object is allocated from OS heap with 2706 // extra padding bytes (Page::kPageSize + Page::kObjectStartOffset). 2707 // A large object always starts at Page::kObjectStartOffset to a page. 2708 // Large objects do not move during garbage collections. 2709 2710 class LargeObjectSpace : public Space { 2711 public: 2712 LargeObjectSpace(Heap* heap, intptr_t max_capacity, AllocationSpace id); 2713 virtual ~LargeObjectSpace() {} 2714 2715 // Initializes internal data structures. 2716 bool SetUp(); 2717 2718 // Releases internal resources, frees objects in this space. 2719 void TearDown(); 2720 2721 static intptr_t ObjectSizeFor(intptr_t chunk_size) { 2722 if (chunk_size <= (Page::kPageSize + Page::kObjectStartOffset)) return 0; 2723 return chunk_size - Page::kPageSize - Page::kObjectStartOffset; 2724 } 2725 2726 // Shared implementation of AllocateRaw, AllocateRawCode and 2727 // AllocateRawFixedArray. 2728 MUST_USE_RESULT AllocationResult 2729 AllocateRaw(int object_size, Executability executable); 2730 2731 // Available bytes for objects in this space. 2732 inline intptr_t Available(); 2733 2734 virtual intptr_t Size() { return size_; } 2735 2736 virtual intptr_t SizeOfObjects() { return objects_size_; } 2737 2738 intptr_t MaximumCommittedMemory() { return maximum_committed_; } 2739 2740 intptr_t CommittedMemory() { return Size(); } 2741 2742 // Approximate amount of physical memory committed for this space. 2743 size_t CommittedPhysicalMemory(); 2744 2745 int PageCount() { return page_count_; } 2746 2747 // Finds an object for a given address, returns a Smi if it is not found. 2748 // The function iterates through all objects in this space, may be slow. 2749 Object* FindObject(Address a); 2750 2751 // Finds a large object page containing the given address, returns NULL 2752 // if such a page doesn't exist. 2753 LargePage* FindPage(Address a); 2754 2755 // Frees unmarked objects. 2756 void FreeUnmarkedObjects(); 2757 2758 // Checks whether a heap object is in this space; O(1). 2759 bool Contains(HeapObject* obj); 2760 2761 // Checks whether the space is empty. 2762 bool IsEmpty() { return first_page_ == NULL; } 2763 2764 LargePage* first_page() { return first_page_; } 2765 2766 #ifdef VERIFY_HEAP 2767 virtual void Verify(); 2768 #endif 2769 2770 #ifdef DEBUG 2771 virtual void Print(); 2772 void ReportStatistics(); 2773 void CollectCodeStatistics(); 2774 #endif 2775 // Checks whether an address is in the object area in this space. It 2776 // iterates all objects in the space. May be slow. 2777 bool SlowContains(Address addr) { return FindObject(addr)->IsHeapObject(); } 2778 2779 private: 2780 intptr_t max_capacity_; 2781 intptr_t maximum_committed_; 2782 // The head of the linked list of large object chunks. 2783 LargePage* first_page_; 2784 intptr_t size_; // allocated bytes 2785 int page_count_; // number of chunks 2786 intptr_t objects_size_; // size of objects 2787 // Map MemoryChunk::kAlignment-aligned chunks to large pages covering them 2788 HashMap chunk_map_; 2789 2790 friend class LargeObjectIterator; 2791 2792 public: 2793 TRACK_MEMORY("LargeObjectSpace") 2794 }; 2795 2796 2797 class LargeObjectIterator : public ObjectIterator { 2798 public: 2799 explicit LargeObjectIterator(LargeObjectSpace* space); 2800 LargeObjectIterator(LargeObjectSpace* space, HeapObjectCallback size_func); 2801 2802 HeapObject* Next(); 2803 2804 // implementation of ObjectIterator. 2805 virtual HeapObject* next_object() { return Next(); } 2806 2807 private: 2808 LargePage* current_; 2809 HeapObjectCallback size_func_; 2810 }; 2811 2812 2813 // Iterates over the chunks (pages and large object pages) that can contain 2814 // pointers to new space. 2815 class PointerChunkIterator BASE_EMBEDDED { 2816 public: 2817 inline explicit PointerChunkIterator(Heap* heap); 2818 2819 // Return NULL when the iterator is done. 2820 MemoryChunk* next() { 2821 switch (state_) { 2822 case kOldPointerState: { 2823 if (old_pointer_iterator_.has_next()) { 2824 return old_pointer_iterator_.next(); 2825 } 2826 state_ = kMapState; 2827 // Fall through. 2828 } 2829 case kMapState: { 2830 if (map_iterator_.has_next()) { 2831 return map_iterator_.next(); 2832 } 2833 state_ = kLargeObjectState; 2834 // Fall through. 2835 } 2836 case kLargeObjectState: { 2837 HeapObject* heap_object; 2838 do { 2839 heap_object = lo_iterator_.Next(); 2840 if (heap_object == NULL) { 2841 state_ = kFinishedState; 2842 return NULL; 2843 } 2844 // Fixed arrays are the only pointer-containing objects in large 2845 // object space. 2846 } while (!heap_object->IsFixedArray()); 2847 MemoryChunk* answer = MemoryChunk::FromAddress(heap_object->address()); 2848 return answer; 2849 } 2850 case kFinishedState: 2851 return NULL; 2852 default: 2853 break; 2854 } 2855 UNREACHABLE(); 2856 return NULL; 2857 } 2858 2859 2860 private: 2861 enum State { kOldPointerState, kMapState, kLargeObjectState, kFinishedState }; 2862 State state_; 2863 PageIterator old_pointer_iterator_; 2864 PageIterator map_iterator_; 2865 LargeObjectIterator lo_iterator_; 2866 }; 2867 2868 2869 #ifdef DEBUG 2870 struct CommentStatistic { 2871 const char* comment; 2872 int size; 2873 int count; 2874 void Clear() { 2875 comment = NULL; 2876 size = 0; 2877 count = 0; 2878 } 2879 // Must be small, since an iteration is used for lookup. 2880 static const int kMaxComments = 64; 2881 }; 2882 #endif 2883 } 2884 } // namespace v8::internal 2885 2886 #endif // V8_HEAP_SPACES_H_ 2887