1// Copyright 2011 the V8 project authors. All rights reserved.
2// Redistribution and use in source and binary forms, with or without
3// modification, are permitted provided that the following conditions are
4// met:
5//
6//     * Redistributions of source code must retain the above copyright
7//       notice, this list of conditions and the following disclaimer.
8//     * Redistributions in binary form must reproduce the above
9//       copyright notice, this list of conditions and the following
10//       disclaimer in the documentation and/or other materials provided
11//       with the distribution.
12//     * Neither the name of Google Inc. nor the names of its
13//       contributors may be used to endorse or promote products derived
14//       from this software without specific prior written permission.
15//
16// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
17// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
18// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
19// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
20// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
21// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
22// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
26// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27
28#ifndef V8_SPACES_H_
29#define V8_SPACES_H_
30
31#include "allocation.h"
32#include "hashmap.h"
33#include "list.h"
34#include "log.h"
35
36namespace v8 {
37namespace internal {
38
39class Isolate;
40
41// -----------------------------------------------------------------------------
42// Heap structures:
43//
44// A JS heap consists of a young generation, an old generation, and a large
45// object space. The young generation is divided into two semispaces. A
46// scavenger implements Cheney's copying algorithm. The old generation is
47// separated into a map space and an old object space. The map space contains
48// all (and only) map objects, the rest of old objects go into the old space.
49// The old generation is collected by a mark-sweep-compact collector.
50//
51// The semispaces of the young generation are contiguous.  The old and map
52// spaces consists of a list of pages. A page has a page header and an object
53// area.
54//
55// There is a separate large object space for objects larger than
56// Page::kMaxHeapObjectSize, so that they do not have to move during
57// collection. The large object space is paged. Pages in large object space
58// may be larger than the page size.
59//
60// A store-buffer based write barrier is used to keep track of intergenerational
61// references.  See store-buffer.h.
62//
63// During scavenges and mark-sweep collections we sometimes (after a store
64// buffer overflow) iterate intergenerational pointers without decoding heap
65// object maps so if the page belongs to old pointer space or large object
66// space it is essential to guarantee that the page does not contain any
67// garbage pointers to new space: every pointer aligned word which satisfies
68// the Heap::InNewSpace() predicate must be a pointer to a live heap object in
69// new space. Thus objects in old pointer and large object spaces should have a
70// special layout (e.g. no bare integer fields). This requirement does not
71// apply to map space which is iterated in a special fashion. However we still
72// require pointer fields of dead maps to be cleaned.
73//
74// To enable lazy cleaning of old space pages we can mark chunks of the page
75// as being garbage.  Garbage sections are marked with a special map.  These
76// sections are skipped when scanning the page, even if we are otherwise
77// scanning without regard for object boundaries.  Garbage sections are chained
78// together to form a free list after a GC.  Garbage sections created outside
79// of GCs by object trunctation etc. may not be in the free list chain.  Very
80// small free spaces are ignored, they need only be cleaned of bogus pointers
81// into new space.
82//
83// Each page may have up to one special garbage section.  The start of this
84// section is denoted by the top field in the space.  The end of the section
85// is denoted by the limit field in the space.  This special garbage section
86// is not marked with a free space map in the data.  The point of this section
87// is to enable linear allocation without having to constantly update the byte
88// array every time the top field is updated and a new object is created.  The
89// special garbage section is not in the chain of garbage sections.
90//
91// Since the top and limit fields are in the space, not the page, only one page
92// has a special garbage section, and if the top and limit are equal then there
93// is no special garbage section.
94
95// Some assertion macros used in the debugging mode.
96
97#define ASSERT_PAGE_ALIGNED(address)                                           \
98  ASSERT((OffsetFrom(address) & Page::kPageAlignmentMask) == 0)
99
100#define ASSERT_OBJECT_ALIGNED(address)                                         \
101  ASSERT((OffsetFrom(address) & kObjectAlignmentMask) == 0)
102
103#define ASSERT_MAP_ALIGNED(address)                                            \
104  ASSERT((OffsetFrom(address) & kMapAlignmentMask) == 0)
105
106#define ASSERT_OBJECT_SIZE(size)                                               \
107  ASSERT((0 < size) && (size <= Page::kMaxNonCodeHeapObjectSize))
108
109#define ASSERT_PAGE_OFFSET(offset)                                             \
110  ASSERT((Page::kObjectStartOffset <= offset)                                  \
111      && (offset <= Page::kPageSize))
112
113#define ASSERT_MAP_PAGE_INDEX(index)                                           \
114  ASSERT((0 <= index) && (index <= MapSpace::kMaxMapPageIndex))
115
116
117class PagedSpace;
118class MemoryAllocator;
119class AllocationInfo;
120class Space;
121class FreeList;
122class MemoryChunk;
123
124class MarkBit {
125 public:
126  typedef uint32_t CellType;
127
128  inline MarkBit(CellType* cell, CellType mask, bool data_only)
129      : cell_(cell), mask_(mask), data_only_(data_only) { }
130
131  inline CellType* cell() { return cell_; }
132  inline CellType mask() { return mask_; }
133
134#ifdef DEBUG
135  bool operator==(const MarkBit& other) {
136    return cell_ == other.cell_ && mask_ == other.mask_;
137  }
138#endif
139
140  inline void Set() { *cell_ |= mask_; }
141  inline bool Get() { return (*cell_ & mask_) != 0; }
142  inline void Clear() { *cell_ &= ~mask_; }
143
144  inline bool data_only() { return data_only_; }
145
146  inline MarkBit Next() {
147    CellType new_mask = mask_ << 1;
148    if (new_mask == 0) {
149      return MarkBit(cell_ + 1, 1, data_only_);
150    } else {
151      return MarkBit(cell_, new_mask, data_only_);
152    }
153  }
154
155 private:
156  CellType* cell_;
157  CellType mask_;
158  // This boolean indicates that the object is in a data-only space with no
159  // pointers.  This enables some optimizations when marking.
160  // It is expected that this field is inlined and turned into control flow
161  // at the place where the MarkBit object is created.
162  bool data_only_;
163};
164
165
166// Bitmap is a sequence of cells each containing fixed number of bits.
167class Bitmap {
168 public:
169  static const uint32_t kBitsPerCell = 32;
170  static const uint32_t kBitsPerCellLog2 = 5;
171  static const uint32_t kBitIndexMask = kBitsPerCell - 1;
172  static const uint32_t kBytesPerCell = kBitsPerCell / kBitsPerByte;
173  static const uint32_t kBytesPerCellLog2 = kBitsPerCellLog2 - kBitsPerByteLog2;
174
175  static const size_t kLength =
176    (1 << kPageSizeBits) >> (kPointerSizeLog2);
177
178  static const size_t kSize =
179    (1 << kPageSizeBits) >> (kPointerSizeLog2 + kBitsPerByteLog2);
180
181
182  static int CellsForLength(int length) {
183    return (length + kBitsPerCell - 1) >> kBitsPerCellLog2;
184  }
185
186  int CellsCount() {
187    return CellsForLength(kLength);
188  }
189
190  static int SizeFor(int cells_count) {
191    return sizeof(MarkBit::CellType) * cells_count;
192  }
193
194  INLINE(static uint32_t IndexToCell(uint32_t index)) {
195    return index >> kBitsPerCellLog2;
196  }
197
198  INLINE(static uint32_t CellToIndex(uint32_t index)) {
199    return index << kBitsPerCellLog2;
200  }
201
202  INLINE(static uint32_t CellAlignIndex(uint32_t index)) {
203    return (index + kBitIndexMask) & ~kBitIndexMask;
204  }
205
206  INLINE(MarkBit::CellType* cells()) {
207    return reinterpret_cast<MarkBit::CellType*>(this);
208  }
209
210  INLINE(Address address()) {
211    return reinterpret_cast<Address>(this);
212  }
213
214  INLINE(static Bitmap* FromAddress(Address addr)) {
215    return reinterpret_cast<Bitmap*>(addr);
216  }
217
218  inline MarkBit MarkBitFromIndex(uint32_t index, bool data_only = false) {
219    MarkBit::CellType mask = 1 << (index & kBitIndexMask);
220    MarkBit::CellType* cell = this->cells() + (index >> kBitsPerCellLog2);
221    return MarkBit(cell, mask, data_only);
222  }
223
224  static inline void Clear(MemoryChunk* chunk);
225
226  static void PrintWord(uint32_t word, uint32_t himask = 0) {
227    for (uint32_t mask = 1; mask != 0; mask <<= 1) {
228      if ((mask & himask) != 0) PrintF("[");
229      PrintF((mask & word) ? "1" : "0");
230      if ((mask & himask) != 0) PrintF("]");
231    }
232  }
233
234  class CellPrinter {
235   public:
236    CellPrinter() : seq_start(0), seq_type(0), seq_length(0) { }
237
238    void Print(uint32_t pos, uint32_t cell) {
239      if (cell == seq_type) {
240        seq_length++;
241        return;
242      }
243
244      Flush();
245
246      if (IsSeq(cell)) {
247        seq_start = pos;
248        seq_length = 0;
249        seq_type = cell;
250        return;
251      }
252
253      PrintF("%d: ", pos);
254      PrintWord(cell);
255      PrintF("\n");
256    }
257
258    void Flush() {
259      if (seq_length > 0) {
260        PrintF("%d: %dx%d\n",
261               seq_start,
262               seq_type == 0 ? 0 : 1,
263               seq_length * kBitsPerCell);
264        seq_length = 0;
265      }
266    }
267
268    static bool IsSeq(uint32_t cell) { return cell == 0 || cell == 0xFFFFFFFF; }
269
270   private:
271    uint32_t seq_start;
272    uint32_t seq_type;
273    uint32_t seq_length;
274  };
275
276  void Print() {
277    CellPrinter printer;
278    for (int i = 0; i < CellsCount(); i++) {
279      printer.Print(i, cells()[i]);
280    }
281    printer.Flush();
282    PrintF("\n");
283  }
284
285  bool IsClean() {
286    for (int i = 0; i < CellsCount(); i++) {
287      if (cells()[i] != 0) return false;
288    }
289    return true;
290  }
291};
292
293
294class SkipList;
295class SlotsBuffer;
296
297// MemoryChunk represents a memory region owned by a specific space.
298// It is divided into the header and the body. Chunk start is always
299// 1MB aligned. Start of the body is aligned so it can accommodate
300// any heap object.
301class MemoryChunk {
302 public:
303  // Only works if the pointer is in the first kPageSize of the MemoryChunk.
304  static MemoryChunk* FromAddress(Address a) {
305    return reinterpret_cast<MemoryChunk*>(OffsetFrom(a) & ~kAlignmentMask);
306  }
307
308  // Only works for addresses in pointer spaces, not data or code spaces.
309  static inline MemoryChunk* FromAnyPointerAddress(Address addr);
310
311  Address address() { return reinterpret_cast<Address>(this); }
312
313  bool is_valid() { return address() != NULL; }
314
315  MemoryChunk* next_chunk() const { return next_chunk_; }
316  MemoryChunk* prev_chunk() const { return prev_chunk_; }
317
318  void set_next_chunk(MemoryChunk* next) { next_chunk_ = next; }
319  void set_prev_chunk(MemoryChunk* prev) { prev_chunk_ = prev; }
320
321  Space* owner() const {
322    if ((reinterpret_cast<intptr_t>(owner_) & kFailureTagMask) ==
323        kFailureTag) {
324      return reinterpret_cast<Space*>(owner_ - kFailureTag);
325    } else {
326      return NULL;
327    }
328  }
329
330  void set_owner(Space* space) {
331    ASSERT((reinterpret_cast<intptr_t>(space) & kFailureTagMask) == 0);
332    owner_ = reinterpret_cast<Address>(space) + kFailureTag;
333    ASSERT((reinterpret_cast<intptr_t>(owner_) & kFailureTagMask) ==
334           kFailureTag);
335  }
336
337  VirtualMemory* reserved_memory() {
338    return &reservation_;
339  }
340
341  void InitializeReservedMemory() {
342    reservation_.Reset();
343  }
344
345  void set_reserved_memory(VirtualMemory* reservation) {
346    ASSERT_NOT_NULL(reservation);
347    reservation_.TakeControl(reservation);
348  }
349
350  bool scan_on_scavenge() { return IsFlagSet(SCAN_ON_SCAVENGE); }
351  void initialize_scan_on_scavenge(bool scan) {
352    if (scan) {
353      SetFlag(SCAN_ON_SCAVENGE);
354    } else {
355      ClearFlag(SCAN_ON_SCAVENGE);
356    }
357  }
358  inline void set_scan_on_scavenge(bool scan);
359
360  int store_buffer_counter() { return store_buffer_counter_; }
361  void set_store_buffer_counter(int counter) {
362    store_buffer_counter_ = counter;
363  }
364
365  bool Contains(Address addr) {
366    return addr >= area_start() && addr < area_end();
367  }
368
369  // Checks whether addr can be a limit of addresses in this page.
370  // It's a limit if it's in the page, or if it's just after the
371  // last byte of the page.
372  bool ContainsLimit(Address addr) {
373    return addr >= area_start() && addr <= area_end();
374  }
375
376  enum MemoryChunkFlags {
377    IS_EXECUTABLE,
378    ABOUT_TO_BE_FREED,
379    POINTERS_TO_HERE_ARE_INTERESTING,
380    POINTERS_FROM_HERE_ARE_INTERESTING,
381    SCAN_ON_SCAVENGE,
382    IN_FROM_SPACE,  // Mutually exclusive with IN_TO_SPACE.
383    IN_TO_SPACE,    // All pages in new space has one of these two set.
384    NEW_SPACE_BELOW_AGE_MARK,
385    CONTAINS_ONLY_DATA,
386    EVACUATION_CANDIDATE,
387    RESCAN_ON_EVACUATION,
388
389    // Pages swept precisely can be iterated, hitting only the live objects.
390    // Whereas those swept conservatively cannot be iterated over. Both flags
391    // indicate that marking bits have been cleared by the sweeper, otherwise
392    // marking bits are still intact.
393    WAS_SWEPT_PRECISELY,
394    WAS_SWEPT_CONSERVATIVELY,
395
396    // Last flag, keep at bottom.
397    NUM_MEMORY_CHUNK_FLAGS
398  };
399
400
401  static const int kPointersToHereAreInterestingMask =
402      1 << POINTERS_TO_HERE_ARE_INTERESTING;
403
404  static const int kPointersFromHereAreInterestingMask =
405      1 << POINTERS_FROM_HERE_ARE_INTERESTING;
406
407  static const int kEvacuationCandidateMask =
408      1 << EVACUATION_CANDIDATE;
409
410  static const int kSkipEvacuationSlotsRecordingMask =
411      (1 << EVACUATION_CANDIDATE) |
412      (1 << RESCAN_ON_EVACUATION) |
413      (1 << IN_FROM_SPACE) |
414      (1 << IN_TO_SPACE);
415
416
417  void SetFlag(int flag) {
418    flags_ |= static_cast<uintptr_t>(1) << flag;
419  }
420
421  void ClearFlag(int flag) {
422    flags_ &= ~(static_cast<uintptr_t>(1) << flag);
423  }
424
425  void SetFlagTo(int flag, bool value) {
426    if (value) {
427      SetFlag(flag);
428    } else {
429      ClearFlag(flag);
430    }
431  }
432
433  bool IsFlagSet(int flag) {
434    return (flags_ & (static_cast<uintptr_t>(1) << flag)) != 0;
435  }
436
437  // Set or clear multiple flags at a time. The flags in the mask
438  // are set to the value in "flags", the rest retain the current value
439  // in flags_.
440  void SetFlags(intptr_t flags, intptr_t mask) {
441    flags_ = (flags_ & ~mask) | (flags & mask);
442  }
443
444  // Return all current flags.
445  intptr_t GetFlags() { return flags_; }
446
447  // Manage live byte count (count of bytes known to be live,
448  // because they are marked black).
449  void ResetLiveBytes() {
450    if (FLAG_gc_verbose) {
451      PrintF("ResetLiveBytes:%p:%x->0\n",
452             static_cast<void*>(this), live_byte_count_);
453    }
454    live_byte_count_ = 0;
455  }
456  void IncrementLiveBytes(int by) {
457    if (FLAG_gc_verbose) {
458      printf("UpdateLiveBytes:%p:%x%c=%x->%x\n",
459             static_cast<void*>(this), live_byte_count_,
460             ((by < 0) ? '-' : '+'), ((by < 0) ? -by : by),
461             live_byte_count_ + by);
462    }
463    live_byte_count_ += by;
464    ASSERT_LE(static_cast<unsigned>(live_byte_count_), size_);
465  }
466  int LiveBytes() {
467    ASSERT(static_cast<unsigned>(live_byte_count_) <= size_);
468    return live_byte_count_;
469  }
470
471  static void IncrementLiveBytesFromGC(Address address, int by) {
472    MemoryChunk::FromAddress(address)->IncrementLiveBytes(by);
473  }
474
475  static void IncrementLiveBytesFromMutator(Address address, int by);
476
477  static const intptr_t kAlignment =
478      (static_cast<uintptr_t>(1) << kPageSizeBits);
479
480  static const intptr_t kAlignmentMask = kAlignment - 1;
481
482  static const intptr_t kSizeOffset = kPointerSize + kPointerSize;
483
484  static const intptr_t kLiveBytesOffset =
485     kSizeOffset + kPointerSize + kPointerSize + kPointerSize +
486     kPointerSize + kPointerSize +
487     kPointerSize + kPointerSize + kPointerSize + kIntSize;
488
489  static const size_t kSlotsBufferOffset = kLiveBytesOffset + kIntSize;
490
491  static const size_t kHeaderSize =
492      kSlotsBufferOffset + kPointerSize + kPointerSize;
493
494  static const int kBodyOffset =
495    CODE_POINTER_ALIGN(MAP_POINTER_ALIGN(kHeaderSize + Bitmap::kSize));
496
497  // The start offset of the object area in a page. Aligned to both maps and
498  // code alignment to be suitable for both.  Also aligned to 32 words because
499  // the marking bitmap is arranged in 32 bit chunks.
500  static const int kObjectStartAlignment = 32 * kPointerSize;
501  static const int kObjectStartOffset = kBodyOffset - 1 +
502      (kObjectStartAlignment - (kBodyOffset - 1) % kObjectStartAlignment);
503
504  size_t size() const { return size_; }
505
506  void set_size(size_t size) {
507    size_ = size;
508  }
509
510  void SetArea(Address area_start, Address area_end) {
511    area_start_ = area_start;
512    area_end_ = area_end;
513  }
514
515  Executability executable() {
516    return IsFlagSet(IS_EXECUTABLE) ? EXECUTABLE : NOT_EXECUTABLE;
517  }
518
519  bool ContainsOnlyData() {
520    return IsFlagSet(CONTAINS_ONLY_DATA);
521  }
522
523  bool InNewSpace() {
524    return (flags_ & ((1 << IN_FROM_SPACE) | (1 << IN_TO_SPACE))) != 0;
525  }
526
527  bool InToSpace() {
528    return IsFlagSet(IN_TO_SPACE);
529  }
530
531  bool InFromSpace() {
532    return IsFlagSet(IN_FROM_SPACE);
533  }
534
535  // ---------------------------------------------------------------------
536  // Markbits support
537
538  inline Bitmap* markbits() {
539    return Bitmap::FromAddress(address() + kHeaderSize);
540  }
541
542  void PrintMarkbits() { markbits()->Print(); }
543
544  inline uint32_t AddressToMarkbitIndex(Address addr) {
545    return static_cast<uint32_t>(addr - this->address()) >> kPointerSizeLog2;
546  }
547
548  inline static uint32_t FastAddressToMarkbitIndex(Address addr) {
549    const intptr_t offset =
550        reinterpret_cast<intptr_t>(addr) & kAlignmentMask;
551
552    return static_cast<uint32_t>(offset) >> kPointerSizeLog2;
553  }
554
555  inline Address MarkbitIndexToAddress(uint32_t index) {
556    return this->address() + (index << kPointerSizeLog2);
557  }
558
559  void InsertAfter(MemoryChunk* other);
560  void Unlink();
561
562  inline Heap* heap() { return heap_; }
563
564  static const int kFlagsOffset = kPointerSize * 3;
565
566  bool IsEvacuationCandidate() { return IsFlagSet(EVACUATION_CANDIDATE); }
567
568  bool ShouldSkipEvacuationSlotRecording() {
569    return (flags_ & kSkipEvacuationSlotsRecordingMask) != 0;
570  }
571
572  inline SkipList* skip_list() {
573    return skip_list_;
574  }
575
576  inline void set_skip_list(SkipList* skip_list) {
577    skip_list_ = skip_list;
578  }
579
580  inline SlotsBuffer* slots_buffer() {
581    return slots_buffer_;
582  }
583
584  inline SlotsBuffer** slots_buffer_address() {
585    return &slots_buffer_;
586  }
587
588  void MarkEvacuationCandidate() {
589    ASSERT(slots_buffer_ == NULL);
590    SetFlag(EVACUATION_CANDIDATE);
591  }
592
593  void ClearEvacuationCandidate() {
594    ASSERT(slots_buffer_ == NULL);
595    ClearFlag(EVACUATION_CANDIDATE);
596  }
597
598  Address area_start() { return area_start_; }
599  Address area_end() { return area_end_; }
600  int area_size() {
601    return static_cast<int>(area_end() - area_start());
602  }
603
604 protected:
605  MemoryChunk* next_chunk_;
606  MemoryChunk* prev_chunk_;
607  size_t size_;
608  intptr_t flags_;
609
610  // Start and end of allocatable memory on this chunk.
611  Address area_start_;
612  Address area_end_;
613
614  // If the chunk needs to remember its memory reservation, it is stored here.
615  VirtualMemory reservation_;
616  // The identity of the owning space.  This is tagged as a failure pointer, but
617  // no failure can be in an object, so this can be distinguished from any entry
618  // in a fixed array.
619  Address owner_;
620  Heap* heap_;
621  // Used by the store buffer to keep track of which pages to mark scan-on-
622  // scavenge.
623  int store_buffer_counter_;
624  // Count of bytes marked black on page.
625  int live_byte_count_;
626  SlotsBuffer* slots_buffer_;
627  SkipList* skip_list_;
628
629  static MemoryChunk* Initialize(Heap* heap,
630                                 Address base,
631                                 size_t size,
632                                 Address area_start,
633                                 Address area_end,
634                                 Executability executable,
635                                 Space* owner);
636
637  friend class MemoryAllocator;
638};
639
640STATIC_CHECK(sizeof(MemoryChunk) <= MemoryChunk::kHeaderSize);
641
642// -----------------------------------------------------------------------------
643// A page is a memory chunk of a size 1MB. Large object pages may be larger.
644//
645// The only way to get a page pointer is by calling factory methods:
646//   Page* p = Page::FromAddress(addr); or
647//   Page* p = Page::FromAllocationTop(top);
648class Page : public MemoryChunk {
649 public:
650  // Returns the page containing a given address. The address ranges
651  // from [page_addr .. page_addr + kPageSize[
652  // This only works if the object is in fact in a page.  See also MemoryChunk::
653  // FromAddress() and FromAnyAddress().
654  INLINE(static Page* FromAddress(Address a)) {
655    return reinterpret_cast<Page*>(OffsetFrom(a) & ~kPageAlignmentMask);
656  }
657
658  // Returns the page containing an allocation top. Because an allocation
659  // top address can be the upper bound of the page, we need to subtract
660  // it with kPointerSize first. The address ranges from
661  // [page_addr + kObjectStartOffset .. page_addr + kPageSize].
662  INLINE(static Page* FromAllocationTop(Address top)) {
663    Page* p = FromAddress(top - kPointerSize);
664    return p;
665  }
666
667  // Returns the next page in the chain of pages owned by a space.
668  inline Page* next_page();
669  inline Page* prev_page();
670  inline void set_next_page(Page* page);
671  inline void set_prev_page(Page* page);
672
673  // Checks whether an address is page aligned.
674  static bool IsAlignedToPageSize(Address a) {
675    return 0 == (OffsetFrom(a) & kPageAlignmentMask);
676  }
677
678  // Returns the offset of a given address to this page.
679  INLINE(int Offset(Address a)) {
680    int offset = static_cast<int>(a - address());
681    return offset;
682  }
683
684  // Returns the address for a given offset to the this page.
685  Address OffsetToAddress(int offset) {
686    ASSERT_PAGE_OFFSET(offset);
687    return address() + offset;
688  }
689
690  // ---------------------------------------------------------------------
691
692  // Page size in bytes.  This must be a multiple of the OS page size.
693  static const int kPageSize = 1 << kPageSizeBits;
694
695  // Object area size in bytes.
696  static const int kNonCodeObjectAreaSize = kPageSize - kObjectStartOffset;
697
698  // Maximum object size that fits in a page.
699  static const int kMaxNonCodeHeapObjectSize = kNonCodeObjectAreaSize;
700
701  // Page size mask.
702  static const intptr_t kPageAlignmentMask = (1 << kPageSizeBits) - 1;
703
704  inline void ClearGCFields();
705
706  static inline Page* Initialize(Heap* heap,
707                                 MemoryChunk* chunk,
708                                 Executability executable,
709                                 PagedSpace* owner);
710
711  void InitializeAsAnchor(PagedSpace* owner);
712
713  bool WasSweptPrecisely() { return IsFlagSet(WAS_SWEPT_PRECISELY); }
714  bool WasSweptConservatively() { return IsFlagSet(WAS_SWEPT_CONSERVATIVELY); }
715  bool WasSwept() { return WasSweptPrecisely() || WasSweptConservatively(); }
716
717  void MarkSweptPrecisely() { SetFlag(WAS_SWEPT_PRECISELY); }
718  void MarkSweptConservatively() { SetFlag(WAS_SWEPT_CONSERVATIVELY); }
719
720  void ClearSweptPrecisely() { ClearFlag(WAS_SWEPT_PRECISELY); }
721  void ClearSweptConservatively() { ClearFlag(WAS_SWEPT_CONSERVATIVELY); }
722
723#ifdef DEBUG
724  void Print();
725#endif  // DEBUG
726
727  friend class MemoryAllocator;
728};
729
730
731STATIC_CHECK(sizeof(Page) <= MemoryChunk::kHeaderSize);
732
733
734class LargePage : public MemoryChunk {
735 public:
736  HeapObject* GetObject() {
737    return HeapObject::FromAddress(area_start());
738  }
739
740  inline LargePage* next_page() const {
741    return static_cast<LargePage*>(next_chunk());
742  }
743
744  inline void set_next_page(LargePage* page) {
745    set_next_chunk(page);
746  }
747 private:
748  static inline LargePage* Initialize(Heap* heap, MemoryChunk* chunk);
749
750  friend class MemoryAllocator;
751};
752
753STATIC_CHECK(sizeof(LargePage) <= MemoryChunk::kHeaderSize);
754
755// ----------------------------------------------------------------------------
756// Space is the abstract superclass for all allocation spaces.
757class Space : public Malloced {
758 public:
759  Space(Heap* heap, AllocationSpace id, Executability executable)
760      : heap_(heap), id_(id), executable_(executable) {}
761
762  virtual ~Space() {}
763
764  Heap* heap() const { return heap_; }
765
766  // Does the space need executable memory?
767  Executability executable() { return executable_; }
768
769  // Identity used in error reporting.
770  AllocationSpace identity() { return id_; }
771
772  // Returns allocated size.
773  virtual intptr_t Size() = 0;
774
775  // Returns size of objects. Can differ from the allocated size
776  // (e.g. see LargeObjectSpace).
777  virtual intptr_t SizeOfObjects() { return Size(); }
778
779  virtual int RoundSizeDownToObjectAlignment(int size) {
780    if (id_ == CODE_SPACE) {
781      return RoundDown(size, kCodeAlignment);
782    } else {
783      return RoundDown(size, kPointerSize);
784    }
785  }
786
787#ifdef DEBUG
788  virtual void Print() = 0;
789#endif
790
791  // After calling this we can allocate a certain number of bytes using only
792  // linear allocation (with a LinearAllocationScope and an AlwaysAllocateScope)
793  // without using freelists or causing a GC.  This is used by partial
794  // snapshots.  It returns true of space was reserved or false if a GC is
795  // needed.  For paged spaces the space requested must include the space wasted
796  // at the end of each when allocating linearly.
797  virtual bool ReserveSpace(int bytes) = 0;
798
799 private:
800  Heap* heap_;
801  AllocationSpace id_;
802  Executability executable_;
803};
804
805
806// ----------------------------------------------------------------------------
807// All heap objects containing executable code (code objects) must be allocated
808// from a 2 GB range of memory, so that they can call each other using 32-bit
809// displacements.  This happens automatically on 32-bit platforms, where 32-bit
810// displacements cover the entire 4GB virtual address space.  On 64-bit
811// platforms, we support this using the CodeRange object, which reserves and
812// manages a range of virtual memory.
813class CodeRange {
814 public:
815  explicit CodeRange(Isolate* isolate);
816  ~CodeRange() { TearDown(); }
817
818  // Reserves a range of virtual memory, but does not commit any of it.
819  // Can only be called once, at heap initialization time.
820  // Returns false on failure.
821  bool SetUp(const size_t requested_size);
822
823  // Frees the range of virtual memory, and frees the data structures used to
824  // manage it.
825  void TearDown();
826
827  bool exists() { return this != NULL && code_range_ != NULL; }
828  bool contains(Address address) {
829    if (this == NULL || code_range_ == NULL) return false;
830    Address start = static_cast<Address>(code_range_->address());
831    return start <= address && address < start + code_range_->size();
832  }
833
834  // Allocates a chunk of memory from the large-object portion of
835  // the code range.  On platforms with no separate code range, should
836  // not be called.
837  MUST_USE_RESULT Address AllocateRawMemory(const size_t requested,
838                                            size_t* allocated);
839  void FreeRawMemory(Address buf, size_t length);
840
841 private:
842  Isolate* isolate_;
843
844  // The reserved range of virtual memory that all code objects are put in.
845  VirtualMemory* code_range_;
846  // Plain old data class, just a struct plus a constructor.
847  class FreeBlock {
848   public:
849    FreeBlock(Address start_arg, size_t size_arg)
850        : start(start_arg), size(size_arg) {
851      ASSERT(IsAddressAligned(start, MemoryChunk::kAlignment));
852      ASSERT(size >= static_cast<size_t>(Page::kPageSize));
853    }
854    FreeBlock(void* start_arg, size_t size_arg)
855        : start(static_cast<Address>(start_arg)), size(size_arg) {
856      ASSERT(IsAddressAligned(start, MemoryChunk::kAlignment));
857      ASSERT(size >= static_cast<size_t>(Page::kPageSize));
858    }
859
860    Address start;
861    size_t size;
862  };
863
864  // Freed blocks of memory are added to the free list.  When the allocation
865  // list is exhausted, the free list is sorted and merged to make the new
866  // allocation list.
867  List<FreeBlock> free_list_;
868  // Memory is allocated from the free blocks on the allocation list.
869  // The block at current_allocation_block_index_ is the current block.
870  List<FreeBlock> allocation_list_;
871  int current_allocation_block_index_;
872
873  // Finds a block on the allocation list that contains at least the
874  // requested amount of memory.  If none is found, sorts and merges
875  // the existing free memory blocks, and searches again.
876  // If none can be found, terminates V8 with FatalProcessOutOfMemory.
877  void GetNextAllocationBlock(size_t requested);
878  // Compares the start addresses of two free blocks.
879  static int CompareFreeBlockAddress(const FreeBlock* left,
880                                     const FreeBlock* right);
881
882  DISALLOW_COPY_AND_ASSIGN(CodeRange);
883};
884
885
886class SkipList {
887 public:
888  SkipList() {
889    Clear();
890  }
891
892  void Clear() {
893    for (int idx = 0; idx < kSize; idx++) {
894      starts_[idx] = reinterpret_cast<Address>(-1);
895    }
896  }
897
898  Address StartFor(Address addr) {
899    return starts_[RegionNumber(addr)];
900  }
901
902  void AddObject(Address addr, int size) {
903    int start_region = RegionNumber(addr);
904    int end_region = RegionNumber(addr + size - kPointerSize);
905    for (int idx = start_region; idx <= end_region; idx++) {
906      if (starts_[idx] > addr) starts_[idx] = addr;
907    }
908  }
909
910  static inline int RegionNumber(Address addr) {
911    return (OffsetFrom(addr) & Page::kPageAlignmentMask) >> kRegionSizeLog2;
912  }
913
914  static void Update(Address addr, int size) {
915    Page* page = Page::FromAddress(addr);
916    SkipList* list = page->skip_list();
917    if (list == NULL) {
918      list = new SkipList();
919      page->set_skip_list(list);
920    }
921
922    list->AddObject(addr, size);
923  }
924
925 private:
926  static const int kRegionSizeLog2 = 13;
927  static const int kRegionSize = 1 << kRegionSizeLog2;
928  static const int kSize = Page::kPageSize / kRegionSize;
929
930  STATIC_ASSERT(Page::kPageSize % kRegionSize == 0);
931
932  Address starts_[kSize];
933};
934
935
936// ----------------------------------------------------------------------------
937// A space acquires chunks of memory from the operating system. The memory
938// allocator allocated and deallocates pages for the paged heap spaces and large
939// pages for large object space.
940//
941// Each space has to manage it's own pages.
942//
943class MemoryAllocator {
944 public:
945  explicit MemoryAllocator(Isolate* isolate);
946
947  // Initializes its internal bookkeeping structures.
948  // Max capacity of the total space and executable memory limit.
949  bool SetUp(intptr_t max_capacity, intptr_t capacity_executable);
950
951  void TearDown();
952
953  Page* AllocatePage(PagedSpace* owner, Executability executable);
954
955  LargePage* AllocateLargePage(intptr_t object_size,
956                                      Executability executable,
957                                      Space* owner);
958
959  void Free(MemoryChunk* chunk);
960
961  // Returns the maximum available bytes of heaps.
962  intptr_t Available() { return capacity_ < size_ ? 0 : capacity_ - size_; }
963
964  // Returns allocated spaces in bytes.
965  intptr_t Size() { return size_; }
966
967  // Returns the maximum available executable bytes of heaps.
968  intptr_t AvailableExecutable() {
969    if (capacity_executable_ < size_executable_) return 0;
970    return capacity_executable_ - size_executable_;
971  }
972
973  // Returns allocated executable spaces in bytes.
974  intptr_t SizeExecutable() { return size_executable_; }
975
976  // Returns maximum available bytes that the old space can have.
977  intptr_t MaxAvailable() {
978    return (Available() / Page::kPageSize) * Page::kMaxNonCodeHeapObjectSize;
979  }
980
981#ifdef DEBUG
982  // Reports statistic info of the space.
983  void ReportStatistics();
984#endif
985
986  MemoryChunk* AllocateChunk(intptr_t body_size,
987                             Executability executable,
988                             Space* space);
989
990  Address ReserveAlignedMemory(size_t requested,
991                               size_t alignment,
992                               VirtualMemory* controller);
993  Address AllocateAlignedMemory(size_t requested,
994                                size_t alignment,
995                                Executability executable,
996                                VirtualMemory* controller);
997
998  void FreeMemory(VirtualMemory* reservation, Executability executable);
999  void FreeMemory(Address addr, size_t size, Executability executable);
1000
1001  // Commit a contiguous block of memory from the initial chunk.  Assumes that
1002  // the address is not NULL, the size is greater than zero, and that the
1003  // block is contained in the initial chunk.  Returns true if it succeeded
1004  // and false otherwise.
1005  bool CommitBlock(Address start, size_t size, Executability executable);
1006
1007  // Uncommit a contiguous block of memory [start..(start+size)[.
1008  // start is not NULL, the size is greater than zero, and the
1009  // block is contained in the initial chunk.  Returns true if it succeeded
1010  // and false otherwise.
1011  bool UncommitBlock(Address start, size_t size);
1012
1013  // Zaps a contiguous block of memory [start..(start+size)[ thus
1014  // filling it up with a recognizable non-NULL bit pattern.
1015  void ZapBlock(Address start, size_t size);
1016
1017  void PerformAllocationCallback(ObjectSpace space,
1018                                 AllocationAction action,
1019                                 size_t size);
1020
1021  void AddMemoryAllocationCallback(MemoryAllocationCallback callback,
1022                                          ObjectSpace space,
1023                                          AllocationAction action);
1024
1025  void RemoveMemoryAllocationCallback(
1026      MemoryAllocationCallback callback);
1027
1028  bool MemoryAllocationCallbackRegistered(
1029      MemoryAllocationCallback callback);
1030
1031  static int CodePageGuardStartOffset();
1032
1033  static int CodePageGuardSize();
1034
1035  static int CodePageAreaStartOffset();
1036
1037  static int CodePageAreaEndOffset();
1038
1039  static int CodePageAreaSize() {
1040    return CodePageAreaEndOffset() - CodePageAreaStartOffset();
1041  }
1042
1043  MUST_USE_RESULT static bool CommitCodePage(VirtualMemory* vm,
1044                                             Address start,
1045                                             size_t size);
1046
1047 private:
1048  Isolate* isolate_;
1049
1050  // Maximum space size in bytes.
1051  size_t capacity_;
1052  // Maximum subset of capacity_ that can be executable
1053  size_t capacity_executable_;
1054
1055  // Allocated space size in bytes.
1056  size_t size_;
1057  // Allocated executable space size in bytes.
1058  size_t size_executable_;
1059
1060  struct MemoryAllocationCallbackRegistration {
1061    MemoryAllocationCallbackRegistration(MemoryAllocationCallback callback,
1062                                         ObjectSpace space,
1063                                         AllocationAction action)
1064        : callback(callback), space(space), action(action) {
1065    }
1066    MemoryAllocationCallback callback;
1067    ObjectSpace space;
1068    AllocationAction action;
1069  };
1070
1071  // A List of callback that are triggered when memory is allocated or free'd
1072  List<MemoryAllocationCallbackRegistration>
1073      memory_allocation_callbacks_;
1074
1075  // Initializes pages in a chunk. Returns the first page address.
1076  // This function and GetChunkId() are provided for the mark-compact
1077  // collector to rebuild page headers in the from space, which is
1078  // used as a marking stack and its page headers are destroyed.
1079  Page* InitializePagesInChunk(int chunk_id, int pages_in_chunk,
1080                               PagedSpace* owner);
1081
1082  DISALLOW_IMPLICIT_CONSTRUCTORS(MemoryAllocator);
1083};
1084
1085
1086// -----------------------------------------------------------------------------
1087// Interface for heap object iterator to be implemented by all object space
1088// object iterators.
1089//
1090// NOTE: The space specific object iterators also implements the own next()
1091//       method which is used to avoid using virtual functions
1092//       iterating a specific space.
1093
1094class ObjectIterator : public Malloced {
1095 public:
1096  virtual ~ObjectIterator() { }
1097
1098  virtual HeapObject* next_object() = 0;
1099};
1100
1101
1102// -----------------------------------------------------------------------------
1103// Heap object iterator in new/old/map spaces.
1104//
1105// A HeapObjectIterator iterates objects from the bottom of the given space
1106// to its top or from the bottom of the given page to its top.
1107//
1108// If objects are allocated in the page during iteration the iterator may
1109// or may not iterate over those objects.  The caller must create a new
1110// iterator in order to be sure to visit these new objects.
1111class HeapObjectIterator: public ObjectIterator {
1112 public:
1113  // Creates a new object iterator in a given space.
1114  // If the size function is not given, the iterator calls the default
1115  // Object::Size().
1116  explicit HeapObjectIterator(PagedSpace* space);
1117  HeapObjectIterator(PagedSpace* space, HeapObjectCallback size_func);
1118  HeapObjectIterator(Page* page, HeapObjectCallback size_func);
1119
1120  // Advance to the next object, skipping free spaces and other fillers and
1121  // skipping the special garbage section of which there is one per space.
1122  // Returns NULL when the iteration has ended.
1123  inline HeapObject* Next() {
1124    do {
1125      HeapObject* next_obj = FromCurrentPage();
1126      if (next_obj != NULL) return next_obj;
1127    } while (AdvanceToNextPage());
1128    return NULL;
1129  }
1130
1131  virtual HeapObject* next_object() {
1132    return Next();
1133  }
1134
1135 private:
1136  enum PageMode { kOnePageOnly, kAllPagesInSpace };
1137
1138  Address cur_addr_;  // Current iteration point.
1139  Address cur_end_;   // End iteration point.
1140  HeapObjectCallback size_func_;  // Size function or NULL.
1141  PagedSpace* space_;
1142  PageMode page_mode_;
1143
1144  // Fast (inlined) path of next().
1145  inline HeapObject* FromCurrentPage();
1146
1147  // Slow path of next(), goes into the next page.  Returns false if the
1148  // iteration has ended.
1149  bool AdvanceToNextPage();
1150
1151  // Initializes fields.
1152  inline void Initialize(PagedSpace* owner,
1153                         Address start,
1154                         Address end,
1155                         PageMode mode,
1156                         HeapObjectCallback size_func);
1157};
1158
1159
1160// -----------------------------------------------------------------------------
1161// A PageIterator iterates the pages in a paged space.
1162
1163class PageIterator BASE_EMBEDDED {
1164 public:
1165  explicit inline PageIterator(PagedSpace* space);
1166
1167  inline bool has_next();
1168  inline Page* next();
1169
1170 private:
1171  PagedSpace* space_;
1172  Page* prev_page_;  // Previous page returned.
1173  // Next page that will be returned.  Cached here so that we can use this
1174  // iterator for operations that deallocate pages.
1175  Page* next_page_;
1176};
1177
1178
1179// -----------------------------------------------------------------------------
1180// A space has a circular list of pages. The next page can be accessed via
1181// Page::next_page() call.
1182
1183// An abstraction of allocation and relocation pointers in a page-structured
1184// space.
1185class AllocationInfo {
1186 public:
1187  AllocationInfo() : top(NULL), limit(NULL) {
1188  }
1189
1190  Address top;  // Current allocation top.
1191  Address limit;  // Current allocation limit.
1192
1193#ifdef DEBUG
1194  bool VerifyPagedAllocation() {
1195    return (Page::FromAllocationTop(top) == Page::FromAllocationTop(limit))
1196        && (top <= limit);
1197  }
1198#endif
1199};
1200
1201
1202// An abstraction of the accounting statistics of a page-structured space.
1203// The 'capacity' of a space is the number of object-area bytes (i.e., not
1204// including page bookkeeping structures) currently in the space. The 'size'
1205// of a space is the number of allocated bytes, the 'waste' in the space is
1206// the number of bytes that are not allocated and not available to
1207// allocation without reorganizing the space via a GC (e.g. small blocks due
1208// to internal fragmentation, top of page areas in map space), and the bytes
1209// 'available' is the number of unallocated bytes that are not waste.  The
1210// capacity is the sum of size, waste, and available.
1211//
1212// The stats are only set by functions that ensure they stay balanced. These
1213// functions increase or decrease one of the non-capacity stats in
1214// conjunction with capacity, or else they always balance increases and
1215// decreases to the non-capacity stats.
1216class AllocationStats BASE_EMBEDDED {
1217 public:
1218  AllocationStats() { Clear(); }
1219
1220  // Zero out all the allocation statistics (i.e., no capacity).
1221  void Clear() {
1222    capacity_ = 0;
1223    size_ = 0;
1224    waste_ = 0;
1225  }
1226
1227  void ClearSizeWaste() {
1228    size_ = capacity_;
1229    waste_ = 0;
1230  }
1231
1232  // Reset the allocation statistics (i.e., available = capacity with no
1233  // wasted or allocated bytes).
1234  void Reset() {
1235    size_ = 0;
1236    waste_ = 0;
1237  }
1238
1239  // Accessors for the allocation statistics.
1240  intptr_t Capacity() { return capacity_; }
1241  intptr_t Size() { return size_; }
1242  intptr_t Waste() { return waste_; }
1243
1244  // Grow the space by adding available bytes.  They are initially marked as
1245  // being in use (part of the size), but will normally be immediately freed,
1246  // putting them on the free list and removing them from size_.
1247  void ExpandSpace(int size_in_bytes) {
1248    capacity_ += size_in_bytes;
1249    size_ += size_in_bytes;
1250    ASSERT(size_ >= 0);
1251  }
1252
1253  // Shrink the space by removing available bytes.  Since shrinking is done
1254  // during sweeping, bytes have been marked as being in use (part of the size)
1255  // and are hereby freed.
1256  void ShrinkSpace(int size_in_bytes) {
1257    capacity_ -= size_in_bytes;
1258    size_ -= size_in_bytes;
1259    ASSERT(size_ >= 0);
1260  }
1261
1262  // Allocate from available bytes (available -> size).
1263  void AllocateBytes(intptr_t size_in_bytes) {
1264    size_ += size_in_bytes;
1265    ASSERT(size_ >= 0);
1266  }
1267
1268  // Free allocated bytes, making them available (size -> available).
1269  void DeallocateBytes(intptr_t size_in_bytes) {
1270    size_ -= size_in_bytes;
1271    ASSERT(size_ >= 0);
1272  }
1273
1274  // Waste free bytes (available -> waste).
1275  void WasteBytes(int size_in_bytes) {
1276    size_ -= size_in_bytes;
1277    waste_ += size_in_bytes;
1278    ASSERT(size_ >= 0);
1279  }
1280
1281 private:
1282  intptr_t capacity_;
1283  intptr_t size_;
1284  intptr_t waste_;
1285};
1286
1287
1288// -----------------------------------------------------------------------------
1289// Free lists for old object spaces
1290//
1291// Free-list nodes are free blocks in the heap.  They look like heap objects
1292// (free-list node pointers have the heap object tag, and they have a map like
1293// a heap object).  They have a size and a next pointer.  The next pointer is
1294// the raw address of the next free list node (or NULL).
1295class FreeListNode: public HeapObject {
1296 public:
1297  // Obtain a free-list node from a raw address.  This is not a cast because
1298  // it does not check nor require that the first word at the address is a map
1299  // pointer.
1300  static FreeListNode* FromAddress(Address address) {
1301    return reinterpret_cast<FreeListNode*>(HeapObject::FromAddress(address));
1302  }
1303
1304  static inline bool IsFreeListNode(HeapObject* object);
1305
1306  // Set the size in bytes, which can be read with HeapObject::Size().  This
1307  // function also writes a map to the first word of the block so that it
1308  // looks like a heap object to the garbage collector and heap iteration
1309  // functions.
1310  void set_size(Heap* heap, int size_in_bytes);
1311
1312  // Accessors for the next field.
1313  inline FreeListNode* next();
1314  inline FreeListNode** next_address();
1315  inline void set_next(FreeListNode* next);
1316
1317  inline void Zap();
1318
1319 private:
1320  static const int kNextOffset = POINTER_SIZE_ALIGN(FreeSpace::kHeaderSize);
1321
1322  DISALLOW_IMPLICIT_CONSTRUCTORS(FreeListNode);
1323};
1324
1325
1326// The free list for the old space.  The free list is organized in such a way
1327// as to encourage objects allocated around the same time to be near each
1328// other.  The normal way to allocate is intended to be by bumping a 'top'
1329// pointer until it hits a 'limit' pointer.  When the limit is hit we need to
1330// find a new space to allocate from.  This is done with the free list, which
1331// is divided up into rough categories to cut down on waste.  Having finer
1332// categories would scatter allocation more.
1333
1334// The old space free list is organized in categories.
1335// 1-31 words:  Such small free areas are discarded for efficiency reasons.
1336//     They can be reclaimed by the compactor.  However the distance between top
1337//     and limit may be this small.
1338// 32-255 words: There is a list of spaces this large.  It is used for top and
1339//     limit when the object we need to allocate is 1-31 words in size.  These
1340//     spaces are called small.
1341// 256-2047 words: There is a list of spaces this large.  It is used for top and
1342//     limit when the object we need to allocate is 32-255 words in size.  These
1343//     spaces are called medium.
1344// 1048-16383 words: There is a list of spaces this large.  It is used for top
1345//     and limit when the object we need to allocate is 256-2047 words in size.
1346//     These spaces are call large.
1347// At least 16384 words.  This list is for objects of 2048 words or larger.
1348//     Empty pages are added to this list.  These spaces are called huge.
1349class FreeList BASE_EMBEDDED {
1350 public:
1351  explicit FreeList(PagedSpace* owner);
1352
1353  // Clear the free list.
1354  void Reset();
1355
1356  // Return the number of bytes available on the free list.
1357  intptr_t available() { return available_; }
1358
1359  // Place a node on the free list.  The block of size 'size_in_bytes'
1360  // starting at 'start' is placed on the free list.  The return value is the
1361  // number of bytes that have been lost due to internal fragmentation by
1362  // freeing the block.  Bookkeeping information will be written to the block,
1363  // i.e., its contents will be destroyed.  The start address should be word
1364  // aligned, and the size should be a non-zero multiple of the word size.
1365  int Free(Address start, int size_in_bytes);
1366
1367  // Allocate a block of size 'size_in_bytes' from the free list.  The block
1368  // is unitialized.  A failure is returned if no block is available.  The
1369  // number of bytes lost to fragmentation is returned in the output parameter
1370  // 'wasted_bytes'.  The size should be a non-zero multiple of the word size.
1371  MUST_USE_RESULT HeapObject* Allocate(int size_in_bytes);
1372
1373#ifdef DEBUG
1374  void Zap();
1375  static intptr_t SumFreeList(FreeListNode* node);
1376  static int FreeListLength(FreeListNode* cur);
1377  intptr_t SumFreeLists();
1378  bool IsVeryLong();
1379#endif
1380
1381  struct SizeStats {
1382    intptr_t Total() {
1383      return small_size_ + medium_size_ + large_size_ + huge_size_;
1384    }
1385
1386    intptr_t small_size_;
1387    intptr_t medium_size_;
1388    intptr_t large_size_;
1389    intptr_t huge_size_;
1390  };
1391
1392  void CountFreeListItems(Page* p, SizeStats* sizes);
1393
1394  intptr_t EvictFreeListItems(Page* p);
1395
1396 private:
1397  // The size range of blocks, in bytes.
1398  static const int kMinBlockSize = 3 * kPointerSize;
1399  static const int kMaxBlockSize = Page::kMaxNonCodeHeapObjectSize;
1400
1401  FreeListNode* PickNodeFromList(FreeListNode** list, int* node_size);
1402
1403  FreeListNode* FindNodeFor(int size_in_bytes, int* node_size);
1404
1405  PagedSpace* owner_;
1406  Heap* heap_;
1407
1408  // Total available bytes in all blocks on this free list.
1409  int available_;
1410
1411  static const int kSmallListMin = 0x20 * kPointerSize;
1412  static const int kSmallListMax = 0xff * kPointerSize;
1413  static const int kMediumListMax = 0x7ff * kPointerSize;
1414  static const int kLargeListMax = 0x3fff * kPointerSize;
1415  static const int kSmallAllocationMax = kSmallListMin - kPointerSize;
1416  static const int kMediumAllocationMax = kSmallListMax;
1417  static const int kLargeAllocationMax = kMediumListMax;
1418  FreeListNode* small_list_;
1419  FreeListNode* medium_list_;
1420  FreeListNode* large_list_;
1421  FreeListNode* huge_list_;
1422
1423  DISALLOW_IMPLICIT_CONSTRUCTORS(FreeList);
1424};
1425
1426
1427class PagedSpace : public Space {
1428 public:
1429  // Creates a space with a maximum capacity, and an id.
1430  PagedSpace(Heap* heap,
1431             intptr_t max_capacity,
1432             AllocationSpace id,
1433             Executability executable);
1434
1435  virtual ~PagedSpace() {}
1436
1437  // Set up the space using the given address range of virtual memory (from
1438  // the memory allocator's initial chunk) if possible.  If the block of
1439  // addresses is not big enough to contain a single page-aligned page, a
1440  // fresh chunk will be allocated.
1441  bool SetUp();
1442
1443  // Returns true if the space has been successfully set up and not
1444  // subsequently torn down.
1445  bool HasBeenSetUp();
1446
1447  // Cleans up the space, frees all pages in this space except those belonging
1448  // to the initial chunk, uncommits addresses in the initial chunk.
1449  void TearDown();
1450
1451  // Checks whether an object/address is in this space.
1452  inline bool Contains(Address a);
1453  bool Contains(HeapObject* o) { return Contains(o->address()); }
1454
1455  // Given an address occupied by a live object, return that object if it is
1456  // in this space, or Failure::Exception() if it is not. The implementation
1457  // iterates over objects in the page containing the address, the cost is
1458  // linear in the number of objects in the page. It may be slow.
1459  MUST_USE_RESULT MaybeObject* FindObject(Address addr);
1460
1461  // Prepares for a mark-compact GC.
1462  virtual void PrepareForMarkCompact();
1463
1464  // Current capacity without growing (Size() + Available()).
1465  intptr_t Capacity() { return accounting_stats_.Capacity(); }
1466
1467  // Total amount of memory committed for this space.  For paged
1468  // spaces this equals the capacity.
1469  intptr_t CommittedMemory() { return Capacity(); }
1470
1471  // Sets the capacity, the available space and the wasted space to zero.
1472  // The stats are rebuilt during sweeping by adding each page to the
1473  // capacity and the size when it is encountered.  As free spaces are
1474  // discovered during the sweeping they are subtracted from the size and added
1475  // to the available and wasted totals.
1476  void ClearStats() {
1477    accounting_stats_.ClearSizeWaste();
1478  }
1479
1480  // Available bytes without growing.  These are the bytes on the free list.
1481  // The bytes in the linear allocation area are not included in this total
1482  // because updating the stats would slow down allocation.  New pages are
1483  // immediately added to the free list so they show up here.
1484  intptr_t Available() { return free_list_.available(); }
1485
1486  // Allocated bytes in this space.  Garbage bytes that were not found due to
1487  // lazy sweeping are counted as being allocated!  The bytes in the current
1488  // linear allocation area (between top and limit) are also counted here.
1489  virtual intptr_t Size() { return accounting_stats_.Size(); }
1490
1491  // As size, but the bytes in lazily swept pages are estimated and the bytes
1492  // in the current linear allocation area are not included.
1493  virtual intptr_t SizeOfObjects() {
1494    ASSERT(!IsSweepingComplete() || (unswept_free_bytes_ == 0));
1495    return Size() - unswept_free_bytes_ - (limit() - top());
1496  }
1497
1498  // Wasted bytes in this space.  These are just the bytes that were thrown away
1499  // due to being too small to use for allocation.  They do not include the
1500  // free bytes that were not found at all due to lazy sweeping.
1501  virtual intptr_t Waste() { return accounting_stats_.Waste(); }
1502
1503  // Returns the allocation pointer in this space.
1504  Address top() { return allocation_info_.top; }
1505  Address limit() { return allocation_info_.limit; }
1506
1507  // Allocate the requested number of bytes in the space if possible, return a
1508  // failure object if not.
1509  MUST_USE_RESULT inline MaybeObject* AllocateRaw(int size_in_bytes);
1510
1511  virtual bool ReserveSpace(int bytes);
1512
1513  // Give a block of memory to the space's free list.  It might be added to
1514  // the free list or accounted as waste.
1515  // If add_to_freelist is false then just accounting stats are updated and
1516  // no attempt to add area to free list is made.
1517  int Free(Address start, int size_in_bytes) {
1518    int wasted = free_list_.Free(start, size_in_bytes);
1519    accounting_stats_.DeallocateBytes(size_in_bytes - wasted);
1520    return size_in_bytes - wasted;
1521  }
1522
1523  // Set space allocation info.
1524  void SetTop(Address top, Address limit) {
1525    ASSERT(top == limit ||
1526           Page::FromAddress(top) == Page::FromAddress(limit - 1));
1527    allocation_info_.top = top;
1528    allocation_info_.limit = limit;
1529  }
1530
1531  void Allocate(int bytes) {
1532    accounting_stats_.AllocateBytes(bytes);
1533  }
1534
1535  void IncreaseCapacity(int size) {
1536    accounting_stats_.ExpandSpace(size);
1537  }
1538
1539  // Releases an unused page and shrinks the space.
1540  void ReleasePage(Page* page);
1541
1542  // Releases all of the unused pages.
1543  void ReleaseAllUnusedPages();
1544
1545  // The dummy page that anchors the linked list of pages.
1546  Page* anchor() { return &anchor_; }
1547
1548#ifdef DEBUG
1549  // Print meta info and objects in this space.
1550  virtual void Print();
1551
1552  // Verify integrity of this space.
1553  virtual void Verify(ObjectVisitor* visitor);
1554
1555  // Reports statistics for the space
1556  void ReportStatistics();
1557
1558  // Overridden by subclasses to verify space-specific object
1559  // properties (e.g., only maps or free-list nodes are in map space).
1560  virtual void VerifyObject(HeapObject* obj) {}
1561
1562  // Report code object related statistics
1563  void CollectCodeStatistics();
1564  static void ReportCodeStatistics();
1565  static void ResetCodeStatistics();
1566#endif
1567
1568  bool was_swept_conservatively() { return was_swept_conservatively_; }
1569  void set_was_swept_conservatively(bool b) { was_swept_conservatively_ = b; }
1570
1571  // Evacuation candidates are swept by evacuator.  Needs to return a valid
1572  // result before _and_ after evacuation has finished.
1573  static bool ShouldBeSweptLazily(Page* p) {
1574    return !p->IsEvacuationCandidate() &&
1575           !p->IsFlagSet(Page::RESCAN_ON_EVACUATION) &&
1576           !p->WasSweptPrecisely();
1577  }
1578
1579  void SetPagesToSweep(Page* first) {
1580    ASSERT(unswept_free_bytes_ == 0);
1581    if (first == &anchor_) first = NULL;
1582    first_unswept_page_ = first;
1583  }
1584
1585  void IncrementUnsweptFreeBytes(int by) {
1586    unswept_free_bytes_ += by;
1587  }
1588
1589  void IncreaseUnsweptFreeBytes(Page* p) {
1590    ASSERT(ShouldBeSweptLazily(p));
1591    unswept_free_bytes_ += (p->area_size() - p->LiveBytes());
1592  }
1593
1594  void DecreaseUnsweptFreeBytes(Page* p) {
1595    ASSERT(ShouldBeSweptLazily(p));
1596    unswept_free_bytes_ -= (p->area_size() - p->LiveBytes());
1597  }
1598
1599  bool AdvanceSweeper(intptr_t bytes_to_sweep);
1600
1601  bool IsSweepingComplete() {
1602    return !first_unswept_page_->is_valid();
1603  }
1604
1605  Page* FirstPage() { return anchor_.next_page(); }
1606  Page* LastPage() { return anchor_.prev_page(); }
1607
1608  void CountFreeListItems(Page* p, FreeList::SizeStats* sizes) {
1609    free_list_.CountFreeListItems(p, sizes);
1610  }
1611
1612  void EvictEvacuationCandidatesFromFreeLists();
1613
1614  bool CanExpand();
1615
1616  // Returns the number of total pages in this space.
1617  int CountTotalPages();
1618
1619  // Return size of allocatable area on a page in this space.
1620  inline int AreaSize() {
1621    return area_size_;
1622  }
1623
1624 protected:
1625  int area_size_;
1626
1627  // Maximum capacity of this space.
1628  intptr_t max_capacity_;
1629
1630  // Accounting information for this space.
1631  AllocationStats accounting_stats_;
1632
1633  // The dummy page that anchors the double linked list of pages.
1634  Page anchor_;
1635
1636  // The space's free list.
1637  FreeList free_list_;
1638
1639  // Normal allocation information.
1640  AllocationInfo allocation_info_;
1641
1642  // Bytes of each page that cannot be allocated.  Possibly non-zero
1643  // for pages in spaces with only fixed-size objects.  Always zero
1644  // for pages in spaces with variable sized objects (those pages are
1645  // padded with free-list nodes).
1646  int page_extra_;
1647
1648  bool was_swept_conservatively_;
1649
1650  // The first page to be swept when the lazy sweeper advances. Is set
1651  // to NULL when all pages have been swept.
1652  Page* first_unswept_page_;
1653
1654  // The number of free bytes which could be reclaimed by advancing the
1655  // lazy sweeper.  This is only an estimation because lazy sweeping is
1656  // done conservatively.
1657  intptr_t unswept_free_bytes_;
1658
1659  // Expands the space by allocating a fixed number of pages. Returns false if
1660  // it cannot allocate requested number of pages from OS, or if the hard heap
1661  // size limit has been hit.
1662  bool Expand();
1663
1664  // Generic fast case allocation function that tries linear allocation at the
1665  // address denoted by top in allocation_info_.
1666  inline HeapObject* AllocateLinearly(int size_in_bytes);
1667
1668  // Slow path of AllocateRaw.  This function is space-dependent.
1669  MUST_USE_RESULT virtual HeapObject* SlowAllocateRaw(int size_in_bytes);
1670
1671  friend class PageIterator;
1672};
1673
1674
1675class NumberAndSizeInfo BASE_EMBEDDED {
1676 public:
1677  NumberAndSizeInfo() : number_(0), bytes_(0) {}
1678
1679  int number() const { return number_; }
1680  void increment_number(int num) { number_ += num; }
1681
1682  int bytes() const { return bytes_; }
1683  void increment_bytes(int size) { bytes_ += size; }
1684
1685  void clear() {
1686    number_ = 0;
1687    bytes_ = 0;
1688  }
1689
1690 private:
1691  int number_;
1692  int bytes_;
1693};
1694
1695
1696// HistogramInfo class for recording a single "bar" of a histogram.  This
1697// class is used for collecting statistics to print to the log file.
1698class HistogramInfo: public NumberAndSizeInfo {
1699 public:
1700  HistogramInfo() : NumberAndSizeInfo() {}
1701
1702  const char* name() { return name_; }
1703  void set_name(const char* name) { name_ = name; }
1704
1705 private:
1706  const char* name_;
1707};
1708
1709
1710enum SemiSpaceId {
1711  kFromSpace = 0,
1712  kToSpace = 1
1713};
1714
1715
1716class SemiSpace;
1717
1718
1719class NewSpacePage : public MemoryChunk {
1720 public:
1721  // GC related flags copied from from-space to to-space when
1722  // flipping semispaces.
1723  static const intptr_t kCopyOnFlipFlagsMask =
1724    (1 << MemoryChunk::POINTERS_TO_HERE_ARE_INTERESTING) |
1725    (1 << MemoryChunk::POINTERS_FROM_HERE_ARE_INTERESTING) |
1726    (1 << MemoryChunk::SCAN_ON_SCAVENGE);
1727
1728  static const int kAreaSize = Page::kNonCodeObjectAreaSize;
1729
1730  inline NewSpacePage* next_page() const {
1731    return static_cast<NewSpacePage*>(next_chunk());
1732  }
1733
1734  inline void set_next_page(NewSpacePage* page) {
1735    set_next_chunk(page);
1736  }
1737
1738  inline NewSpacePage* prev_page() const {
1739    return static_cast<NewSpacePage*>(prev_chunk());
1740  }
1741
1742  inline void set_prev_page(NewSpacePage* page) {
1743    set_prev_chunk(page);
1744  }
1745
1746  SemiSpace* semi_space() {
1747    return reinterpret_cast<SemiSpace*>(owner());
1748  }
1749
1750  bool is_anchor() { return !this->InNewSpace(); }
1751
1752  static bool IsAtStart(Address addr) {
1753    return (reinterpret_cast<intptr_t>(addr) & Page::kPageAlignmentMask)
1754        == kObjectStartOffset;
1755  }
1756
1757  static bool IsAtEnd(Address addr) {
1758    return (reinterpret_cast<intptr_t>(addr) & Page::kPageAlignmentMask) == 0;
1759  }
1760
1761  Address address() {
1762    return reinterpret_cast<Address>(this);
1763  }
1764
1765  // Finds the NewSpacePage containg the given address.
1766  static inline NewSpacePage* FromAddress(Address address_in_page) {
1767    Address page_start =
1768        reinterpret_cast<Address>(reinterpret_cast<uintptr_t>(address_in_page) &
1769                                  ~Page::kPageAlignmentMask);
1770    NewSpacePage* page = reinterpret_cast<NewSpacePage*>(page_start);
1771    return page;
1772  }
1773
1774  // Find the page for a limit address. A limit address is either an address
1775  // inside a page, or the address right after the last byte of a page.
1776  static inline NewSpacePage* FromLimit(Address address_limit) {
1777    return NewSpacePage::FromAddress(address_limit - 1);
1778  }
1779
1780 private:
1781  // Create a NewSpacePage object that is only used as anchor
1782  // for the doubly-linked list of real pages.
1783  explicit NewSpacePage(SemiSpace* owner) {
1784    InitializeAsAnchor(owner);
1785  }
1786
1787  static NewSpacePage* Initialize(Heap* heap,
1788                                  Address start,
1789                                  SemiSpace* semi_space);
1790
1791  // Intialize a fake NewSpacePage used as sentinel at the ends
1792  // of a doubly-linked list of real NewSpacePages.
1793  // Only uses the prev/next links, and sets flags to not be in new-space.
1794  void InitializeAsAnchor(SemiSpace* owner);
1795
1796  friend class SemiSpace;
1797  friend class SemiSpaceIterator;
1798};
1799
1800
1801// -----------------------------------------------------------------------------
1802// SemiSpace in young generation
1803//
1804// A semispace is a contiguous chunk of memory holding page-like memory
1805// chunks. The mark-compact collector  uses the memory of the first page in
1806// the from space as a marking stack when tracing live objects.
1807
1808class SemiSpace : public Space {
1809 public:
1810  // Constructor.
1811  SemiSpace(Heap* heap, SemiSpaceId semispace)
1812    : Space(heap, NEW_SPACE, NOT_EXECUTABLE),
1813      start_(NULL),
1814      age_mark_(NULL),
1815      id_(semispace),
1816      anchor_(this),
1817      current_page_(NULL) { }
1818
1819  // Sets up the semispace using the given chunk.
1820  void SetUp(Address start, int initial_capacity, int maximum_capacity);
1821
1822  // Tear down the space.  Heap memory was not allocated by the space, so it
1823  // is not deallocated here.
1824  void TearDown();
1825
1826  // True if the space has been set up but not torn down.
1827  bool HasBeenSetUp() { return start_ != NULL; }
1828
1829  // Grow the semispace to the new capacity.  The new capacity
1830  // requested must be larger than the current capacity and less than
1831  // the maximum capacity.
1832  bool GrowTo(int new_capacity);
1833
1834  // Shrinks the semispace to the new capacity.  The new capacity
1835  // requested must be more than the amount of used memory in the
1836  // semispace and less than the current capacity.
1837  bool ShrinkTo(int new_capacity);
1838
1839  // Returns the start address of the first page of the space.
1840  Address space_start() {
1841    ASSERT(anchor_.next_page() != &anchor_);
1842    return anchor_.next_page()->area_start();
1843  }
1844
1845  // Returns the start address of the current page of the space.
1846  Address page_low() {
1847    return current_page_->area_start();
1848  }
1849
1850  // Returns one past the end address of the space.
1851  Address space_end() {
1852    return anchor_.prev_page()->area_end();
1853  }
1854
1855  // Returns one past the end address of the current page of the space.
1856  Address page_high() {
1857    return current_page_->area_end();
1858  }
1859
1860  bool AdvancePage() {
1861    NewSpacePage* next_page = current_page_->next_page();
1862    if (next_page == anchor()) return false;
1863    current_page_ = next_page;
1864    return true;
1865  }
1866
1867  // Resets the space to using the first page.
1868  void Reset();
1869
1870  // Age mark accessors.
1871  Address age_mark() { return age_mark_; }
1872  void set_age_mark(Address mark);
1873
1874  // True if the address is in the address range of this semispace (not
1875  // necessarily below the allocation pointer).
1876  bool Contains(Address a) {
1877    return (reinterpret_cast<uintptr_t>(a) & address_mask_)
1878           == reinterpret_cast<uintptr_t>(start_);
1879  }
1880
1881  // True if the object is a heap object in the address range of this
1882  // semispace (not necessarily below the allocation pointer).
1883  bool Contains(Object* o) {
1884    return (reinterpret_cast<uintptr_t>(o) & object_mask_) == object_expected_;
1885  }
1886
1887  // If we don't have these here then SemiSpace will be abstract.  However
1888  // they should never be called.
1889  virtual intptr_t Size() {
1890    UNREACHABLE();
1891    return 0;
1892  }
1893
1894  virtual bool ReserveSpace(int bytes) {
1895    UNREACHABLE();
1896    return false;
1897  }
1898
1899  bool is_committed() { return committed_; }
1900  bool Commit();
1901  bool Uncommit();
1902
1903  NewSpacePage* first_page() { return anchor_.next_page(); }
1904  NewSpacePage* current_page() { return current_page_; }
1905
1906#ifdef DEBUG
1907  virtual void Print();
1908  virtual void Verify();
1909  // Validate a range of of addresses in a SemiSpace.
1910  // The "from" address must be on a page prior to the "to" address,
1911  // in the linked page order, or it must be earlier on the same page.
1912  static void AssertValidRange(Address from, Address to);
1913#else
1914  // Do nothing.
1915  inline static void AssertValidRange(Address from, Address to) {}
1916#endif
1917
1918  // Returns the current capacity of the semi space.
1919  int Capacity() { return capacity_; }
1920
1921  // Returns the maximum capacity of the semi space.
1922  int MaximumCapacity() { return maximum_capacity_; }
1923
1924  // Returns the initial capacity of the semi space.
1925  int InitialCapacity() { return initial_capacity_; }
1926
1927  SemiSpaceId id() { return id_; }
1928
1929  static void Swap(SemiSpace* from, SemiSpace* to);
1930
1931 private:
1932  // Flips the semispace between being from-space and to-space.
1933  // Copies the flags into the masked positions on all pages in the space.
1934  void FlipPages(intptr_t flags, intptr_t flag_mask);
1935
1936  NewSpacePage* anchor() { return &anchor_; }
1937
1938  // The current and maximum capacity of the space.
1939  int capacity_;
1940  int maximum_capacity_;
1941  int initial_capacity_;
1942
1943  // The start address of the space.
1944  Address start_;
1945  // Used to govern object promotion during mark-compact collection.
1946  Address age_mark_;
1947
1948  // Masks and comparison values to test for containment in this semispace.
1949  uintptr_t address_mask_;
1950  uintptr_t object_mask_;
1951  uintptr_t object_expected_;
1952
1953  bool committed_;
1954  SemiSpaceId id_;
1955
1956  NewSpacePage anchor_;
1957  NewSpacePage* current_page_;
1958
1959  friend class SemiSpaceIterator;
1960  friend class NewSpacePageIterator;
1961 public:
1962  TRACK_MEMORY("SemiSpace")
1963};
1964
1965
1966// A SemiSpaceIterator is an ObjectIterator that iterates over the active
1967// semispace of the heap's new space.  It iterates over the objects in the
1968// semispace from a given start address (defaulting to the bottom of the
1969// semispace) to the top of the semispace.  New objects allocated after the
1970// iterator is created are not iterated.
1971class SemiSpaceIterator : public ObjectIterator {
1972 public:
1973  // Create an iterator over the objects in the given space.  If no start
1974  // address is given, the iterator starts from the bottom of the space.  If
1975  // no size function is given, the iterator calls Object::Size().
1976
1977  // Iterate over all of allocated to-space.
1978  explicit SemiSpaceIterator(NewSpace* space);
1979  // Iterate over all of allocated to-space, with a custome size function.
1980  SemiSpaceIterator(NewSpace* space, HeapObjectCallback size_func);
1981  // Iterate over part of allocated to-space, from start to the end
1982  // of allocation.
1983  SemiSpaceIterator(NewSpace* space, Address start);
1984  // Iterate from one address to another in the same semi-space.
1985  SemiSpaceIterator(Address from, Address to);
1986
1987  HeapObject* Next() {
1988    if (current_ == limit_) return NULL;
1989    if (NewSpacePage::IsAtEnd(current_)) {
1990      NewSpacePage* page = NewSpacePage::FromLimit(current_);
1991      page = page->next_page();
1992      ASSERT(!page->is_anchor());
1993      current_ = page->area_start();
1994      if (current_ == limit_) return NULL;
1995    }
1996
1997    HeapObject* object = HeapObject::FromAddress(current_);
1998    int size = (size_func_ == NULL) ? object->Size() : size_func_(object);
1999
2000    current_ += size;
2001    return object;
2002  }
2003
2004  // Implementation of the ObjectIterator functions.
2005  virtual HeapObject* next_object() { return Next(); }
2006
2007 private:
2008  void Initialize(Address start,
2009                  Address end,
2010                  HeapObjectCallback size_func);
2011
2012  // The current iteration point.
2013  Address current_;
2014  // The end of iteration.
2015  Address limit_;
2016  // The callback function.
2017  HeapObjectCallback size_func_;
2018};
2019
2020
2021// -----------------------------------------------------------------------------
2022// A PageIterator iterates the pages in a semi-space.
2023class NewSpacePageIterator BASE_EMBEDDED {
2024 public:
2025  // Make an iterator that runs over all pages in to-space.
2026  explicit inline NewSpacePageIterator(NewSpace* space);
2027
2028  // Make an iterator that runs over all pages in the given semispace,
2029  // even those not used in allocation.
2030  explicit inline NewSpacePageIterator(SemiSpace* space);
2031
2032  // Make iterator that iterates from the page containing start
2033  // to the page that contains limit in the same semispace.
2034  inline NewSpacePageIterator(Address start, Address limit);
2035
2036  inline bool has_next();
2037  inline NewSpacePage* next();
2038
2039 private:
2040  NewSpacePage* prev_page_;  // Previous page returned.
2041  // Next page that will be returned.  Cached here so that we can use this
2042  // iterator for operations that deallocate pages.
2043  NewSpacePage* next_page_;
2044  // Last page returned.
2045  NewSpacePage* last_page_;
2046};
2047
2048
2049// -----------------------------------------------------------------------------
2050// The young generation space.
2051//
2052// The new space consists of a contiguous pair of semispaces.  It simply
2053// forwards most functions to the appropriate semispace.
2054
2055class NewSpace : public Space {
2056 public:
2057  // Constructor.
2058  explicit NewSpace(Heap* heap)
2059    : Space(heap, NEW_SPACE, NOT_EXECUTABLE),
2060      to_space_(heap, kToSpace),
2061      from_space_(heap, kFromSpace),
2062      reservation_(),
2063      inline_allocation_limit_step_(0) {}
2064
2065  // Sets up the new space using the given chunk.
2066  bool SetUp(int reserved_semispace_size_, int max_semispace_size);
2067
2068  // Tears down the space.  Heap memory was not allocated by the space, so it
2069  // is not deallocated here.
2070  void TearDown();
2071
2072  // True if the space has been set up but not torn down.
2073  bool HasBeenSetUp() {
2074    return to_space_.HasBeenSetUp() && from_space_.HasBeenSetUp();
2075  }
2076
2077  // Flip the pair of spaces.
2078  void Flip();
2079
2080  // Grow the capacity of the semispaces.  Assumes that they are not at
2081  // their maximum capacity.
2082  void Grow();
2083
2084  // Shrink the capacity of the semispaces.
2085  void Shrink();
2086
2087  // True if the address or object lies in the address range of either
2088  // semispace (not necessarily below the allocation pointer).
2089  bool Contains(Address a) {
2090    return (reinterpret_cast<uintptr_t>(a) & address_mask_)
2091        == reinterpret_cast<uintptr_t>(start_);
2092  }
2093
2094  bool Contains(Object* o) {
2095    Address a = reinterpret_cast<Address>(o);
2096    return (reinterpret_cast<uintptr_t>(a) & object_mask_) == object_expected_;
2097  }
2098
2099  // Return the allocated bytes in the active semispace.
2100  virtual intptr_t Size() {
2101    return pages_used_ * NewSpacePage::kAreaSize +
2102        static_cast<int>(top() - to_space_.page_low());
2103  }
2104
2105  // The same, but returning an int.  We have to have the one that returns
2106  // intptr_t because it is inherited, but if we know we are dealing with the
2107  // new space, which can't get as big as the other spaces then this is useful:
2108  int SizeAsInt() { return static_cast<int>(Size()); }
2109
2110  // Return the current capacity of a semispace.
2111  intptr_t EffectiveCapacity() {
2112    SLOW_ASSERT(to_space_.Capacity() == from_space_.Capacity());
2113    return (to_space_.Capacity() / Page::kPageSize) * NewSpacePage::kAreaSize;
2114  }
2115
2116  // Return the current capacity of a semispace.
2117  intptr_t Capacity() {
2118    ASSERT(to_space_.Capacity() == from_space_.Capacity());
2119    return to_space_.Capacity();
2120  }
2121
2122  // Return the total amount of memory committed for new space.
2123  intptr_t CommittedMemory() {
2124    if (from_space_.is_committed()) return 2 * Capacity();
2125    return Capacity();
2126  }
2127
2128  // Return the available bytes without growing.
2129  intptr_t Available() {
2130    return Capacity() - Size();
2131  }
2132
2133  // Return the maximum capacity of a semispace.
2134  int MaximumCapacity() {
2135    ASSERT(to_space_.MaximumCapacity() == from_space_.MaximumCapacity());
2136    return to_space_.MaximumCapacity();
2137  }
2138
2139  // Returns the initial capacity of a semispace.
2140  int InitialCapacity() {
2141    ASSERT(to_space_.InitialCapacity() == from_space_.InitialCapacity());
2142    return to_space_.InitialCapacity();
2143  }
2144
2145  // Return the address of the allocation pointer in the active semispace.
2146  Address top() {
2147    ASSERT(to_space_.current_page()->ContainsLimit(allocation_info_.top));
2148    return allocation_info_.top;
2149  }
2150  // Return the address of the first object in the active semispace.
2151  Address bottom() { return to_space_.space_start(); }
2152
2153  // Get the age mark of the inactive semispace.
2154  Address age_mark() { return from_space_.age_mark(); }
2155  // Set the age mark in the active semispace.
2156  void set_age_mark(Address mark) { to_space_.set_age_mark(mark); }
2157
2158  // The start address of the space and a bit mask. Anding an address in the
2159  // new space with the mask will result in the start address.
2160  Address start() { return start_; }
2161  uintptr_t mask() { return address_mask_; }
2162
2163  INLINE(uint32_t AddressToMarkbitIndex(Address addr)) {
2164    ASSERT(Contains(addr));
2165    ASSERT(IsAligned(OffsetFrom(addr), kPointerSize) ||
2166           IsAligned(OffsetFrom(addr) - 1, kPointerSize));
2167    return static_cast<uint32_t>(addr - start_) >> kPointerSizeLog2;
2168  }
2169
2170  INLINE(Address MarkbitIndexToAddress(uint32_t index)) {
2171    return reinterpret_cast<Address>(index << kPointerSizeLog2);
2172  }
2173
2174  // The allocation top and limit addresses.
2175  Address* allocation_top_address() { return &allocation_info_.top; }
2176  Address* allocation_limit_address() { return &allocation_info_.limit; }
2177
2178  MUST_USE_RESULT INLINE(MaybeObject* AllocateRaw(int size_in_bytes));
2179
2180  // Reset the allocation pointer to the beginning of the active semispace.
2181  void ResetAllocationInfo();
2182
2183  void LowerInlineAllocationLimit(intptr_t step) {
2184    inline_allocation_limit_step_ = step;
2185    if (step == 0) {
2186      allocation_info_.limit = to_space_.page_high();
2187    } else {
2188      allocation_info_.limit = Min(
2189          allocation_info_.top + inline_allocation_limit_step_,
2190          allocation_info_.limit);
2191    }
2192    top_on_previous_step_ = allocation_info_.top;
2193  }
2194
2195  // Get the extent of the inactive semispace (for use as a marking stack,
2196  // or to zap it). Notice: space-addresses are not necessarily on the
2197  // same page, so FromSpaceStart() might be above FromSpaceEnd().
2198  Address FromSpacePageLow() { return from_space_.page_low(); }
2199  Address FromSpacePageHigh() { return from_space_.page_high(); }
2200  Address FromSpaceStart() { return from_space_.space_start(); }
2201  Address FromSpaceEnd() { return from_space_.space_end(); }
2202
2203  // Get the extent of the active semispace's pages' memory.
2204  Address ToSpaceStart() { return to_space_.space_start(); }
2205  Address ToSpaceEnd() { return to_space_.space_end(); }
2206
2207  inline bool ToSpaceContains(Address address) {
2208    return to_space_.Contains(address);
2209  }
2210  inline bool FromSpaceContains(Address address) {
2211    return from_space_.Contains(address);
2212  }
2213
2214  // True if the object is a heap object in the address range of the
2215  // respective semispace (not necessarily below the allocation pointer of the
2216  // semispace).
2217  inline bool ToSpaceContains(Object* o) { return to_space_.Contains(o); }
2218  inline bool FromSpaceContains(Object* o) { return from_space_.Contains(o); }
2219
2220  // Try to switch the active semispace to a new, empty, page.
2221  // Returns false if this isn't possible or reasonable (i.e., there
2222  // are no pages, or the current page is already empty), or true
2223  // if successful.
2224  bool AddFreshPage();
2225
2226  virtual bool ReserveSpace(int bytes);
2227
2228  // Resizes a sequential string which must be the most recent thing that was
2229  // allocated in new space.
2230  template <typename StringType>
2231  inline void ShrinkStringAtAllocationBoundary(String* string, int len);
2232
2233#ifdef DEBUG
2234  // Verify the active semispace.
2235  virtual void Verify();
2236  // Print the active semispace.
2237  virtual void Print() { to_space_.Print(); }
2238#endif
2239
2240  // Iterates the active semispace to collect statistics.
2241  void CollectStatistics();
2242  // Reports previously collected statistics of the active semispace.
2243  void ReportStatistics();
2244  // Clears previously collected statistics.
2245  void ClearHistograms();
2246
2247  // Record the allocation or promotion of a heap object.  Note that we don't
2248  // record every single allocation, but only those that happen in the
2249  // to space during a scavenge GC.
2250  void RecordAllocation(HeapObject* obj);
2251  void RecordPromotion(HeapObject* obj);
2252
2253  // Return whether the operation succeded.
2254  bool CommitFromSpaceIfNeeded() {
2255    if (from_space_.is_committed()) return true;
2256    return from_space_.Commit();
2257  }
2258
2259  bool UncommitFromSpace() {
2260    if (!from_space_.is_committed()) return true;
2261    return from_space_.Uncommit();
2262  }
2263
2264  inline intptr_t inline_allocation_limit_step() {
2265    return inline_allocation_limit_step_;
2266  }
2267
2268  SemiSpace* active_space() { return &to_space_; }
2269
2270 private:
2271  // Update allocation info to match the current to-space page.
2272  void UpdateAllocationInfo();
2273
2274  Address chunk_base_;
2275  uintptr_t chunk_size_;
2276
2277  // The semispaces.
2278  SemiSpace to_space_;
2279  SemiSpace from_space_;
2280  VirtualMemory reservation_;
2281  int pages_used_;
2282
2283  // Start address and bit mask for containment testing.
2284  Address start_;
2285  uintptr_t address_mask_;
2286  uintptr_t object_mask_;
2287  uintptr_t object_expected_;
2288
2289  // Allocation pointer and limit for normal allocation and allocation during
2290  // mark-compact collection.
2291  AllocationInfo allocation_info_;
2292
2293  // When incremental marking is active we will set allocation_info_.limit
2294  // to be lower than actual limit and then will gradually increase it
2295  // in steps to guarantee that we do incremental marking steps even
2296  // when all allocation is performed from inlined generated code.
2297  intptr_t inline_allocation_limit_step_;
2298
2299  Address top_on_previous_step_;
2300
2301  HistogramInfo* allocated_histogram_;
2302  HistogramInfo* promoted_histogram_;
2303
2304  MUST_USE_RESULT MaybeObject* SlowAllocateRaw(int size_in_bytes);
2305
2306  friend class SemiSpaceIterator;
2307
2308 public:
2309  TRACK_MEMORY("NewSpace")
2310};
2311
2312
2313// -----------------------------------------------------------------------------
2314// Old object space (excluding map objects)
2315
2316class OldSpace : public PagedSpace {
2317 public:
2318  // Creates an old space object with a given maximum capacity.
2319  // The constructor does not allocate pages from OS.
2320  OldSpace(Heap* heap,
2321           intptr_t max_capacity,
2322           AllocationSpace id,
2323           Executability executable)
2324      : PagedSpace(heap, max_capacity, id, executable) {
2325    page_extra_ = 0;
2326  }
2327
2328  // The limit of allocation for a page in this space.
2329  virtual Address PageAllocationLimit(Page* page) {
2330    return page->area_end();
2331  }
2332
2333 public:
2334  TRACK_MEMORY("OldSpace")
2335};
2336
2337
2338// For contiguous spaces, top should be in the space (or at the end) and limit
2339// should be the end of the space.
2340#define ASSERT_SEMISPACE_ALLOCATION_INFO(info, space) \
2341  SLOW_ASSERT((space).page_low() <= (info).top             \
2342              && (info).top <= (space).page_high()         \
2343              && (info).limit <= (space).page_high())
2344
2345
2346// -----------------------------------------------------------------------------
2347// Old space for objects of a fixed size
2348
2349class FixedSpace : public PagedSpace {
2350 public:
2351  FixedSpace(Heap* heap,
2352             intptr_t max_capacity,
2353             AllocationSpace id,
2354             int object_size_in_bytes,
2355             const char* name)
2356      : PagedSpace(heap, max_capacity, id, NOT_EXECUTABLE),
2357        object_size_in_bytes_(object_size_in_bytes),
2358        name_(name) {
2359    page_extra_ = Page::kNonCodeObjectAreaSize % object_size_in_bytes;
2360  }
2361
2362  // The limit of allocation for a page in this space.
2363  virtual Address PageAllocationLimit(Page* page) {
2364    return page->area_end() - page_extra_;
2365  }
2366
2367  int object_size_in_bytes() { return object_size_in_bytes_; }
2368
2369  // Prepares for a mark-compact GC.
2370  virtual void PrepareForMarkCompact();
2371
2372 protected:
2373  void ResetFreeList() {
2374    free_list_.Reset();
2375  }
2376
2377 private:
2378  // The size of objects in this space.
2379  int object_size_in_bytes_;
2380
2381  // The name of this space.
2382  const char* name_;
2383};
2384
2385
2386// -----------------------------------------------------------------------------
2387// Old space for all map objects
2388
2389class MapSpace : public FixedSpace {
2390 public:
2391  // Creates a map space object with a maximum capacity.
2392  MapSpace(Heap* heap, intptr_t max_capacity, AllocationSpace id)
2393      : FixedSpace(heap, max_capacity, id, Map::kSize, "map"),
2394        max_map_space_pages_(kMaxMapPageIndex - 1) {
2395  }
2396
2397  // Given an index, returns the page address.
2398  // TODO(1600): this limit is artifical just to keep code compilable
2399  static const int kMaxMapPageIndex = 1 << 16;
2400
2401  virtual int RoundSizeDownToObjectAlignment(int size) {
2402    if (IsPowerOf2(Map::kSize)) {
2403      return RoundDown(size, Map::kSize);
2404    } else {
2405      return (size / Map::kSize) * Map::kSize;
2406    }
2407  }
2408
2409 protected:
2410#ifdef DEBUG
2411  virtual void VerifyObject(HeapObject* obj);
2412#endif
2413
2414 private:
2415  static const int kMapsPerPage = Page::kNonCodeObjectAreaSize / Map::kSize;
2416
2417  // Do map space compaction if there is a page gap.
2418  int CompactionThreshold() {
2419    return kMapsPerPage * (max_map_space_pages_ - 1);
2420  }
2421
2422  const int max_map_space_pages_;
2423
2424 public:
2425  TRACK_MEMORY("MapSpace")
2426};
2427
2428
2429// -----------------------------------------------------------------------------
2430// Old space for all global object property cell objects
2431
2432class CellSpace : public FixedSpace {
2433 public:
2434  // Creates a property cell space object with a maximum capacity.
2435  CellSpace(Heap* heap, intptr_t max_capacity, AllocationSpace id)
2436      : FixedSpace(heap, max_capacity, id, JSGlobalPropertyCell::kSize, "cell")
2437  {}
2438
2439  virtual int RoundSizeDownToObjectAlignment(int size) {
2440    if (IsPowerOf2(JSGlobalPropertyCell::kSize)) {
2441      return RoundDown(size, JSGlobalPropertyCell::kSize);
2442    } else {
2443      return (size / JSGlobalPropertyCell::kSize) * JSGlobalPropertyCell::kSize;
2444    }
2445  }
2446
2447 protected:
2448#ifdef DEBUG
2449  virtual void VerifyObject(HeapObject* obj);
2450#endif
2451
2452 public:
2453  TRACK_MEMORY("CellSpace")
2454};
2455
2456
2457// -----------------------------------------------------------------------------
2458// Large objects ( > Page::kMaxHeapObjectSize ) are allocated and managed by
2459// the large object space. A large object is allocated from OS heap with
2460// extra padding bytes (Page::kPageSize + Page::kObjectStartOffset).
2461// A large object always starts at Page::kObjectStartOffset to a page.
2462// Large objects do not move during garbage collections.
2463
2464class LargeObjectSpace : public Space {
2465 public:
2466  LargeObjectSpace(Heap* heap, intptr_t max_capacity, AllocationSpace id);
2467  virtual ~LargeObjectSpace() {}
2468
2469  // Initializes internal data structures.
2470  bool SetUp();
2471
2472  // Releases internal resources, frees objects in this space.
2473  void TearDown();
2474
2475  static intptr_t ObjectSizeFor(intptr_t chunk_size) {
2476    if (chunk_size <= (Page::kPageSize + Page::kObjectStartOffset)) return 0;
2477    return chunk_size - Page::kPageSize - Page::kObjectStartOffset;
2478  }
2479
2480  // Shared implementation of AllocateRaw, AllocateRawCode and
2481  // AllocateRawFixedArray.
2482  MUST_USE_RESULT MaybeObject* AllocateRaw(int object_size,
2483                                           Executability executable);
2484
2485  // Available bytes for objects in this space.
2486  inline intptr_t Available();
2487
2488  virtual intptr_t Size() {
2489    return size_;
2490  }
2491
2492  virtual intptr_t SizeOfObjects() {
2493    return objects_size_;
2494  }
2495
2496  int PageCount() {
2497    return page_count_;
2498  }
2499
2500  // Finds an object for a given address, returns Failure::Exception()
2501  // if it is not found. The function iterates through all objects in this
2502  // space, may be slow.
2503  MaybeObject* FindObject(Address a);
2504
2505  // Finds a large object page containing the given address, returns NULL
2506  // if such a page doesn't exist.
2507  LargePage* FindPage(Address a);
2508
2509  // Frees unmarked objects.
2510  void FreeUnmarkedObjects();
2511
2512  // Checks whether a heap object is in this space; O(1).
2513  bool Contains(HeapObject* obj);
2514
2515  // Checks whether the space is empty.
2516  bool IsEmpty() { return first_page_ == NULL; }
2517
2518  // See the comments for ReserveSpace in the Space class.  This has to be
2519  // called after ReserveSpace has been called on the paged spaces, since they
2520  // may use some memory, leaving less for large objects.
2521  virtual bool ReserveSpace(int bytes);
2522
2523  LargePage* first_page() { return first_page_; }
2524
2525#ifdef DEBUG
2526  virtual void Verify();
2527  virtual void Print();
2528  void ReportStatistics();
2529  void CollectCodeStatistics();
2530#endif
2531  // Checks whether an address is in the object area in this space.  It
2532  // iterates all objects in the space. May be slow.
2533  bool SlowContains(Address addr) { return !FindObject(addr)->IsFailure(); }
2534
2535 private:
2536  intptr_t max_capacity_;
2537  // The head of the linked list of large object chunks.
2538  LargePage* first_page_;
2539  intptr_t size_;  // allocated bytes
2540  int page_count_;  // number of chunks
2541  intptr_t objects_size_;  // size of objects
2542  // Map MemoryChunk::kAlignment-aligned chunks to large pages covering them
2543  HashMap chunk_map_;
2544
2545  friend class LargeObjectIterator;
2546
2547 public:
2548  TRACK_MEMORY("LargeObjectSpace")
2549};
2550
2551
2552class LargeObjectIterator: public ObjectIterator {
2553 public:
2554  explicit LargeObjectIterator(LargeObjectSpace* space);
2555  LargeObjectIterator(LargeObjectSpace* space, HeapObjectCallback size_func);
2556
2557  HeapObject* Next();
2558
2559  // implementation of ObjectIterator.
2560  virtual HeapObject* next_object() { return Next(); }
2561
2562 private:
2563  LargePage* current_;
2564  HeapObjectCallback size_func_;
2565};
2566
2567
2568// Iterates over the chunks (pages and large object pages) that can contain
2569// pointers to new space.
2570class PointerChunkIterator BASE_EMBEDDED {
2571 public:
2572  inline explicit PointerChunkIterator(Heap* heap);
2573
2574  // Return NULL when the iterator is done.
2575  MemoryChunk* next() {
2576    switch (state_) {
2577      case kOldPointerState: {
2578        if (old_pointer_iterator_.has_next()) {
2579          return old_pointer_iterator_.next();
2580        }
2581        state_ = kMapState;
2582        // Fall through.
2583      }
2584      case kMapState: {
2585        if (map_iterator_.has_next()) {
2586          return map_iterator_.next();
2587        }
2588        state_ = kLargeObjectState;
2589        // Fall through.
2590      }
2591      case kLargeObjectState: {
2592        HeapObject* heap_object;
2593        do {
2594          heap_object = lo_iterator_.Next();
2595          if (heap_object == NULL) {
2596            state_ = kFinishedState;
2597            return NULL;
2598          }
2599          // Fixed arrays are the only pointer-containing objects in large
2600          // object space.
2601        } while (!heap_object->IsFixedArray());
2602        MemoryChunk* answer = MemoryChunk::FromAddress(heap_object->address());
2603        return answer;
2604      }
2605      case kFinishedState:
2606        return NULL;
2607      default:
2608        break;
2609    }
2610    UNREACHABLE();
2611    return NULL;
2612  }
2613
2614
2615 private:
2616  enum State {
2617    kOldPointerState,
2618    kMapState,
2619    kLargeObjectState,
2620    kFinishedState
2621  };
2622  State state_;
2623  PageIterator old_pointer_iterator_;
2624  PageIterator map_iterator_;
2625  LargeObjectIterator lo_iterator_;
2626};
2627
2628
2629#ifdef DEBUG
2630struct CommentStatistic {
2631  const char* comment;
2632  int size;
2633  int count;
2634  void Clear() {
2635    comment = NULL;
2636    size = 0;
2637    count = 0;
2638  }
2639  // Must be small, since an iteration is used for lookup.
2640  static const int kMaxComments = 64;
2641};
2642#endif
2643
2644
2645} }  // namespace v8::internal
2646
2647#endif  // V8_SPACES_H_
2648