xref: /art/runtime/gc/collector/mark_sweep.cc

/*
 * Copyright (C) 2011 The Android Open Source Project
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *      http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#include "mark_sweep.h"

#include <functional>
#include <numeric>
#include <climits>
#include <vector>

#include "base/bounded_fifo.h"
#include "base/logging.h"
#include "base/macros.h"
#include "base/mutex-inl.h"
#include "base/timing_logger.h"
#include "gc/accounting/card_table-inl.h"
#include "gc/accounting/heap_bitmap.h"
#include "gc/accounting/mod_union_table.h"
#include "gc/accounting/space_bitmap-inl.h"
#include "gc/heap.h"
#include "gc/space/image_space.h"
#include "gc/space/large_object_space.h"
#include "gc/space/space-inl.h"
#include "indirect_reference_table.h"
#include "intern_table.h"
#include "jni_internal.h"
#include "monitor.h"
#include "mark_sweep-inl.h"
#include "mirror/art_field.h"
#include "mirror/art_field-inl.h"
#include "mirror/class-inl.h"
#include "mirror/class_loader.h"
#include "mirror/dex_cache.h"
#include "mirror/object-inl.h"
#include "mirror/object_array.h"
#include "mirror/object_array-inl.h"
#include "runtime.h"
#include "thread-inl.h"
#include "thread_list.h"
#include "verifier/method_verifier.h"

using ::art::mirror::ArtField;
using ::art::mirror::Class;
using ::art::mirror::Object;
using ::art::mirror::ObjectArray;

namespace art {
namespace gc {
namespace collector {

// Performance options.
constexpr bool kUseRecursiveMark = false;
constexpr bool kUseMarkStackPrefetch = true;
constexpr size_t kSweepArrayChunkFreeSize = 1024;

// Parallelism options.
constexpr bool kParallelCardScan = true;
constexpr bool kParallelRecursiveMark = true;
// Don't attempt to parallelize mark stack processing unless the mark stack is at least n
// elements. This is temporary until we reduce the overhead caused by allocating tasks, etc.. Not
// having this can add overhead in ProcessReferences since we may end up doing many calls of
// ProcessMarkStack with very small mark stacks.
constexpr size_t kMinimumParallelMarkStackSize = 128;
constexpr bool kParallelProcessMarkStack = true;

// Profiling and information flags.
constexpr bool kCountClassesMarked = false;
constexpr bool kProfileLargeObjects = false;
constexpr bool kMeasureOverhead = false;
constexpr bool kCountTasks = false;
constexpr bool kCountJavaLangRefs = false;

// Turn off kCheckLocks when profiling the GC since it slows the GC down by up to 40%.
constexpr bool kCheckLocks = kDebugLocking;

void MarkSweep::ImmuneSpace(space::ContinuousSpace* space) {
  // Bind live to mark bitmap if necessary.
  if (space->GetLiveBitmap() != space->GetMarkBitmap()) {
    CHECK(space->IsContinuousMemMapAllocSpace());
    space->AsContinuousMemMapAllocSpace()->BindLiveToMarkBitmap();
  }

  // Add the space to the immune region.
  // TODO: Use space limits instead of current end_ since the end_ can be changed by dlmalloc
  // callbacks.
  if (immune_begin_ == NULL) {
    DCHECK(immune_end_ == NULL);
    SetImmuneRange(reinterpret_cast<Object*>(space->Begin()),
                   reinterpret_cast<Object*>(space->End()));
  } else {
    const space::ContinuousSpace* prev_space = nullptr;
    // Find out if the previous space is immune.
    for (const space::ContinuousSpace* cur_space : GetHeap()->GetContinuousSpaces()) {
      if (cur_space == space) {
        break;
      }
      prev_space = cur_space;
    }
    // If previous space was immune, then extend the immune region. Relies on continuous spaces
    // being sorted by Heap::AddContinuousSpace.
    if (prev_space != nullptr && IsImmuneSpace(prev_space)) {
      immune_begin_ = std::min(reinterpret_cast<Object*>(space->Begin()), immune_begin_);
      immune_end_ = std::max(reinterpret_cast<Object*>(space->End()), immune_end_);
    }
  }
}

bool MarkSweep::IsImmuneSpace(const space::ContinuousSpace* space) const {
  return
      immune_begin_ <= reinterpret_cast<Object*>(space->Begin()) &&
      immune_end_ >= reinterpret_cast<Object*>(space->End());
}

void MarkSweep::BindBitmaps() {
  timings_.StartSplit("BindBitmaps");
  WriterMutexLock mu(Thread::Current(), *Locks::heap_bitmap_lock_);
  // Mark all of the spaces we never collect as immune.
  for (const auto& space : GetHeap()->GetContinuousSpaces()) {
    if (space->GetGcRetentionPolicy() == space::kGcRetentionPolicyNeverCollect) {
      ImmuneSpace(space);
    }
  }
  timings_.EndSplit();
}

MarkSweep::MarkSweep(Heap* heap, bool is_concurrent, const std::string& name_prefix)
    : GarbageCollector(heap,
                       name_prefix +
                       (is_concurrent ? "concurrent mark sweep": "mark sweep")),
      current_mark_bitmap_(NULL),
      mark_stack_(NULL),
      immune_begin_(NULL),
      immune_end_(NULL),
      live_stack_freeze_size_(0),
      gc_barrier_(new Barrier(0)),
      large_object_lock_("mark sweep large object lock", kMarkSweepLargeObjectLock),
      mark_stack_lock_("mark sweep mark stack lock", kMarkSweepMarkStackLock),
      is_concurrent_(is_concurrent) {
}

void MarkSweep::InitializePhase() {
  timings_.Reset();
  TimingLogger::ScopedSplit split("InitializePhase", &timings_);
  mark_stack_ = heap_->mark_stack_.get();
  DCHECK(mark_stack_ != nullptr);
  SetImmuneRange(nullptr, nullptr);
  class_count_ = 0;
  array_count_ = 0;
  other_count_ = 0;
  large_object_test_ = 0;
  large_object_mark_ = 0;
  classes_marked_ = 0;
  overhead_time_ = 0;
  work_chunks_created_ = 0;
  work_chunks_deleted_ = 0;
  reference_count_ = 0;

  FindDefaultMarkBitmap();

  // Do any pre GC verification.
  timings_.NewSplit("PreGcVerification");
  heap_->PreGcVerification(this);
}

void MarkSweep::ProcessReferences(Thread* self) {
  TimingLogger::ScopedSplit split("ProcessReferences", &timings_);
  WriterMutexLock mu(self, *Locks::heap_bitmap_lock_);
  GetHeap()->ProcessReferences(timings_, clear_soft_references_, &IsMarkedCallback,
                               &RecursiveMarkObjectCallback, this);
}

bool MarkSweep::HandleDirtyObjectsPhase() {
  TimingLogger::ScopedSplit split("HandleDirtyObjectsPhase", &timings_);
  Thread* self = Thread::Current();
  Locks::mutator_lock_->AssertExclusiveHeld(self);

  {
    WriterMutexLock mu(self, *Locks::heap_bitmap_lock_);

    // Re-mark root set.
    ReMarkRoots();

    // Scan dirty objects, this is only required if we are not doing concurrent GC.
    RecursiveMarkDirtyObjects(true, accounting::CardTable::kCardDirty);
  }

  ProcessReferences(self);

  // Only need to do this if we have the card mark verification on, and only during concurrent GC.
  if (GetHeap()->verify_missing_card_marks_ || GetHeap()->verify_pre_gc_heap_||
      GetHeap()->verify_post_gc_heap_) {
    WriterMutexLock mu(self, *Locks::heap_bitmap_lock_);
    // This second sweep makes sure that we don't have any objects in the live stack which point to
    // freed objects. These cause problems since their references may be previously freed objects.
    SweepArray(GetHeap()->allocation_stack_.get(), false);
  }

  timings_.StartSplit("PreSweepingGcVerification");
  heap_->PreSweepingGcVerification(this);
  timings_.EndSplit();

  // Ensure that nobody inserted items in the live stack after we swapped the stacks.
  ReaderMutexLock mu(self, *Locks::heap_bitmap_lock_);
  CHECK_GE(live_stack_freeze_size_, GetHeap()->GetLiveStack()->Size());

  // Disallow new system weaks to prevent a race which occurs when someone adds a new system
  // weak before we sweep them. Since this new system weak may not be marked, the GC may
  // incorrectly sweep it. This also fixes a race where interning may attempt to return a strong
  // reference to a string that is about to be swept.
  Runtime::Current()->DisallowNewSystemWeaks();
  return true;
}

bool MarkSweep::IsConcurrent() const {
  return is_concurrent_;
}

void MarkSweep::MarkingPhase() {
  TimingLogger::ScopedSplit split("MarkingPhase", &timings_);
  Thread* self = Thread::Current();

  BindBitmaps();
  FindDefaultMarkBitmap();

  // Process dirty cards and add dirty cards to mod union tables.
  heap_->ProcessCards(timings_);

  // Need to do this before the checkpoint since we don't want any threads to add references to
  // the live stack during the recursive mark.
  timings_.NewSplit("SwapStacks");
  heap_->SwapStacks();

  WriterMutexLock mu(self, *Locks::heap_bitmap_lock_);
  if (Locks::mutator_lock_->IsExclusiveHeld(self)) {
    // If we exclusively hold the mutator lock, all threads must be suspended.
    MarkRoots();
  } else {
    MarkThreadRoots(self);
    // At this point the live stack should no longer have any mutators which push into it.
    MarkNonThreadRoots();
  }
  live_stack_freeze_size_ = heap_->GetLiveStack()->Size();
  MarkConcurrentRoots();
  UpdateAndMarkModUnion();
  MarkReachableObjects();
}

void MarkSweep::UpdateAndMarkModUnion() {
  for (const auto& space : heap_->GetContinuousSpaces()) {
    if (IsImmuneSpace(space)) {
      const char* name = space->IsZygoteSpace() ? "UpdateAndMarkZygoteModUnionTable" :
          "UpdateAndMarkImageModUnionTable";
      TimingLogger::ScopedSplit split(name, &timings_);
      accounting::ModUnionTable* mod_union_table = heap_->FindModUnionTableFromSpace(space);
      CHECK(mod_union_table != nullptr);
      mod_union_table->UpdateAndMarkReferences(MarkRootCallback, this);
    }
  }
}

void MarkSweep::MarkThreadRoots(Thread* self) {
  MarkRootsCheckpoint(self);
}

void MarkSweep::MarkReachableObjects() {
  // Mark everything allocated since the last as GC live so that we can sweep concurrently,
  // knowing that new allocations won't be marked as live.
  timings_.StartSplit("MarkStackAsLive");
  accounting::ObjectStack* live_stack = heap_->GetLiveStack();
  heap_->MarkAllocStackAsLive(live_stack);
  live_stack->Reset();
  timings_.EndSplit();
  // Recursively mark all the non-image bits set in the mark bitmap.
  RecursiveMark();
}

void MarkSweep::ReclaimPhase() {
  TimingLogger::ScopedSplit split("ReclaimPhase", &timings_);
  Thread* self = Thread::Current();

  if (!IsConcurrent()) {
    ProcessReferences(self);
  }

  {
    WriterMutexLock mu(self, *Locks::heap_bitmap_lock_);
    SweepSystemWeaks();
  }

  if (IsConcurrent()) {
    Runtime::Current()->AllowNewSystemWeaks();

    TimingLogger::ScopedSplit split("UnMarkAllocStack", &timings_);
    WriterMutexLock mu(self, *Locks::heap_bitmap_lock_);
    accounting::ObjectStack* allocation_stack = GetHeap()->allocation_stack_.get();
    // The allocation stack contains things allocated since the start of the GC. These may have been
    // marked during this GC meaning they won't be eligible for reclaiming in the next sticky GC.
    // Remove these objects from the mark bitmaps so that they will be eligible for sticky
    // collection.
    // There is a race here which is safely handled. Another thread such as the hprof could
    // have flushed the alloc stack after we resumed the threads. This is safe however, since
    // reseting the allocation stack zeros it out with madvise. This means that we will either
    // read NULLs or attempt to unmark a newly allocated object which will not be marked in the
    // first place.
    mirror::Object** end = allocation_stack->End();
    for (mirror::Object** it = allocation_stack->Begin(); it != end; ++it) {
      const Object* obj = *it;
      if (obj != NULL) {
        UnMarkObjectNonNull(obj);
      }
    }
  }

  {
    WriterMutexLock mu(self, *Locks::heap_bitmap_lock_);

    // Reclaim unmarked objects.
    Sweep(false);

    // Swap the live and mark bitmaps for each space which we modified space. This is an
    // optimization that enables us to not clear live bits inside of the sweep. Only swaps unbound
    // bitmaps.
    timings_.StartSplit("SwapBitmaps");
    SwapBitmaps();
    timings_.EndSplit();

    // Unbind the live and mark bitmaps.
    TimingLogger::ScopedSplit split("UnBindBitmaps", &timings_);
    GetHeap()->UnBindBitmaps();
  }
}

void MarkSweep::SetImmuneRange(Object* begin, Object* end) {
  immune_begin_ = begin;
  immune_end_ = end;
}

void MarkSweep::FindDefaultMarkBitmap() {
  TimingLogger::ScopedSplit split("FindDefaultMarkBitmap", &timings_);
  for (const auto& space : GetHeap()->GetContinuousSpaces()) {
    accounting::SpaceBitmap* bitmap = space->GetMarkBitmap();
    if (bitmap != nullptr &&
        space->GetGcRetentionPolicy() == space::kGcRetentionPolicyAlwaysCollect) {
      current_mark_bitmap_ = bitmap;
      CHECK(current_mark_bitmap_ != NULL);
      return;
    }
  }
  GetHeap()->DumpSpaces();
  LOG(FATAL) << "Could not find a default mark bitmap";
}

void MarkSweep::ExpandMarkStack() {
  ResizeMarkStack(mark_stack_->Capacity() * 2);
}

void MarkSweep::ResizeMarkStack(size_t new_size) {
  // Rare case, no need to have Thread::Current be a parameter.
  if (UNLIKELY(mark_stack_->Size() < mark_stack_->Capacity())) {
    // Someone else acquired the lock and expanded the mark stack before us.
    return;
  }
  std::vector<Object*> temp(mark_stack_->Begin(), mark_stack_->End());
  CHECK_LE(mark_stack_->Size(), new_size);
  mark_stack_->Resize(new_size);
  for (const auto& obj : temp) {
    mark_stack_->PushBack(obj);
  }
}

inline void MarkSweep::MarkObjectNonNullParallel(const Object* obj) {
  DCHECK(obj != NULL);
  if (MarkObjectParallel(obj)) {
    MutexLock mu(Thread::Current(), mark_stack_lock_);
    if (UNLIKELY(mark_stack_->Size() >= mark_stack_->Capacity())) {
      ExpandMarkStack();
    }
    // The object must be pushed on to the mark stack.
    mark_stack_->PushBack(const_cast<Object*>(obj));
  }
}

mirror::Object* MarkSweep::RecursiveMarkObjectCallback(mirror::Object* obj, void* arg) {
  MarkSweep* mark_sweep = reinterpret_cast<MarkSweep*>(arg);
  mark_sweep->MarkObject(obj);
  mark_sweep->ProcessMarkStack(true);
  return obj;
}

inline void MarkSweep::UnMarkObjectNonNull(const Object* obj) {
  DCHECK(!IsImmune(obj));
  // Try to take advantage of locality of references within a space, failing this find the space
  // the hard way.
  accounting::SpaceBitmap* object_bitmap = current_mark_bitmap_;
  if (UNLIKELY(!object_bitmap->HasAddress(obj))) {
    accounting::SpaceBitmap* new_bitmap = heap_->GetMarkBitmap()->GetContinuousSpaceBitmap(obj);
    if (LIKELY(new_bitmap != NULL)) {
      object_bitmap = new_bitmap;
    } else {
      MarkLargeObject(obj, false);
      return;
    }
  }

  DCHECK(object_bitmap->HasAddress(obj));
  object_bitmap->Clear(obj);
}

inline void MarkSweep::MarkObjectNonNull(const Object* obj) {
  DCHECK(obj != NULL);

  if (IsImmune(obj)) {
    DCHECK(IsMarked(obj));
    return;
  }

  // Try to take advantage of locality of references within a space, failing this find the space
  // the hard way.
  accounting::SpaceBitmap* object_bitmap = current_mark_bitmap_;
  if (UNLIKELY(!object_bitmap->HasAddress(obj))) {
    accounting::SpaceBitmap* new_bitmap = heap_->GetMarkBitmap()->GetContinuousSpaceBitmap(obj);
    if (LIKELY(new_bitmap != NULL)) {
      object_bitmap = new_bitmap;
    } else {
      MarkLargeObject(obj, true);
      return;
    }
  }

  // This object was not previously marked.
  if (!object_bitmap->Test(obj)) {
    object_bitmap->Set(obj);
    if (UNLIKELY(mark_stack_->Size() >= mark_stack_->Capacity())) {
      // Lock is not needed but is here anyways to please annotalysis.
      MutexLock mu(Thread::Current(), mark_stack_lock_);
      ExpandMarkStack();
    }
    // The object must be pushed on to the mark stack.
    mark_stack_->PushBack(const_cast<Object*>(obj));
  }
}

// Rare case, probably not worth inlining since it will increase instruction cache miss rate.
bool MarkSweep::MarkLargeObject(const Object* obj, bool set) {
  // TODO: support >1 discontinuous space.
  space::LargeObjectSpace* large_object_space = GetHeap()->GetLargeObjectsSpace();
  accounting::ObjectSet* large_objects = large_object_space->GetMarkObjects();
  if (kProfileLargeObjects) {
    ++large_object_test_;
  }
  if (UNLIKELY(!large_objects->Test(obj))) {
    if (!large_object_space->Contains(obj)) {
      LOG(ERROR) << "Tried to mark " << obj << " not contained by any spaces";
      LOG(ERROR) << "Attempting see if it's a bad root";
      VerifyRoots();
      LOG(FATAL) << "Can't mark bad root";
    }
    if (kProfileLargeObjects) {
      ++large_object_mark_;
    }
    if (set) {
      large_objects->Set(obj);
    } else {
      large_objects->Clear(obj);
    }
    return true;
  }
  return false;
}

inline bool MarkSweep::MarkObjectParallel(const Object* obj) {
  DCHECK(obj != NULL);

  if (IsImmune(obj)) {
    DCHECK(IsMarked(obj));
    return false;
  }

  // Try to take advantage of locality of references within a space, failing this find the space
  // the hard way.
  accounting::SpaceBitmap* object_bitmap = current_mark_bitmap_;
  if (UNLIKELY(!object_bitmap->HasAddress(obj))) {
    accounting::SpaceBitmap* new_bitmap = heap_->GetMarkBitmap()->GetContinuousSpaceBitmap(obj);
    if (new_bitmap != NULL) {
      object_bitmap = new_bitmap;
    } else {
      // TODO: Remove the Thread::Current here?
      // TODO: Convert this to some kind of atomic marking?
      MutexLock mu(Thread::Current(), large_object_lock_);
      return MarkLargeObject(obj, true);
    }
  }

  // Return true if the object was not previously marked.
  return !object_bitmap->AtomicTestAndSet(obj);
}

// Used to mark objects when recursing.  Recursion is done by moving
// the finger across the bitmaps in address order and marking child
// objects.  Any newly-marked objects whose addresses are lower than
// the finger won't be visited by the bitmap scan, so those objects
// need to be added to the mark stack.
inline void MarkSweep::MarkObject(const Object* obj) {
  if (obj != NULL) {
    MarkObjectNonNull(obj);
  }
}

void MarkSweep::MarkRoot(const Object* obj) {
  if (obj != NULL) {
    MarkObjectNonNull(obj);
  }
}

Object* MarkSweep::MarkRootParallelCallback(Object* root, void* arg) {
  DCHECK(root != NULL);
  DCHECK(arg != NULL);
  reinterpret_cast<MarkSweep*>(arg)->MarkObjectNonNullParallel(root);
  return root;
}

Object* MarkSweep::MarkRootCallback(Object* root, void* arg) {
  DCHECK(root != nullptr);
  DCHECK(arg != nullptr);
  reinterpret_cast<MarkSweep*>(arg)->MarkObjectNonNull(root);
  return root;
}

void MarkSweep::VerifyRootCallback(const Object* root, void* arg, size_t vreg,
                                   const StackVisitor* visitor) {
  reinterpret_cast<MarkSweep*>(arg)->VerifyRoot(root, vreg, visitor);
}

void MarkSweep::VerifyRoot(const Object* root, size_t vreg, const StackVisitor* visitor) {
  // See if the root is on any space bitmap.
  if (GetHeap()->GetLiveBitmap()->GetContinuousSpaceBitmap(root) == NULL) {
    space::LargeObjectSpace* large_object_space = GetHeap()->GetLargeObjectsSpace();
    if (!large_object_space->Contains(root)) {
      LOG(ERROR) << "Found invalid root: " << root;
      if (visitor != NULL) {
        LOG(ERROR) << visitor->DescribeLocation() << " in VReg: " << vreg;
      }
    }
  }
}

void MarkSweep::VerifyRoots() {
  Runtime::Current()->GetThreadList()->VerifyRoots(VerifyRootCallback, this);
}

// Marks all objects in the root set.
void MarkSweep::MarkRoots() {
  timings_.StartSplit("MarkRoots");
  Runtime::Current()->VisitNonConcurrentRoots(MarkRootCallback, this);
  timings_.EndSplit();
}

void MarkSweep::MarkNonThreadRoots() {
  timings_.StartSplit("MarkNonThreadRoots");
  Runtime::Current()->VisitNonThreadRoots(MarkRootCallback, this);
  timings_.EndSplit();
}

void MarkSweep::MarkConcurrentRoots() {
  timings_.StartSplit("MarkConcurrentRoots");
  // Visit all runtime roots and clear dirty flags.
  Runtime::Current()->VisitConcurrentRoots(MarkRootCallback, this, false, true);
  timings_.EndSplit();
}

class ScanObjectVisitor {
 public:
  explicit ScanObjectVisitor(MarkSweep* const mark_sweep) ALWAYS_INLINE
      : mark_sweep_(mark_sweep) {}

  // TODO: Fixme when anotatalysis works with visitors.
  void operator()(Object* obj) const ALWAYS_INLINE NO_THREAD_SAFETY_ANALYSIS {
    if (kCheckLocks) {
      Locks::mutator_lock_->AssertSharedHeld(Thread::Current());
      Locks::heap_bitmap_lock_->AssertExclusiveHeld(Thread::Current());
    }
    mark_sweep_->ScanObject(obj);
  }

 private:
  MarkSweep* const mark_sweep_;
};

template <bool kUseFinger = false>
class MarkStackTask : public Task {
 public:
  MarkStackTask(ThreadPool* thread_pool, MarkSweep* mark_sweep, size_t mark_stack_size,
                const Object** mark_stack)
      : mark_sweep_(mark_sweep),
        thread_pool_(thread_pool),
        mark_stack_pos_(mark_stack_size) {
    // We may have to copy part of an existing mark stack when another mark stack overflows.
    if (mark_stack_size != 0) {
      DCHECK(mark_stack != NULL);
      // TODO: Check performance?
      std::copy(mark_stack, mark_stack + mark_stack_size, mark_stack_);
    }
    if (kCountTasks) {
      ++mark_sweep_->work_chunks_created_;
    }
  }

  static const size_t kMaxSize = 1 * KB;

 protected:
  class ScanObjectParallelVisitor {
   public:
    explicit ScanObjectParallelVisitor(MarkStackTask<kUseFinger>* chunk_task) ALWAYS_INLINE
        : chunk_task_(chunk_task) {}

    void operator()(Object* obj) const {
      MarkSweep* mark_sweep = chunk_task_->mark_sweep_;
      mark_sweep->ScanObjectVisit(obj,
          [mark_sweep, this](Object* /* obj */, Object* ref, const MemberOffset& /* offset */,
              bool /* is_static */) ALWAYS_INLINE_LAMBDA {
        if (ref != nullptr && mark_sweep->MarkObjectParallel(ref)) {
          if (kUseFinger) {
            android_memory_barrier();
            if (reinterpret_cast<uintptr_t>(ref) >=
                static_cast<uintptr_t>(mark_sweep->atomic_finger_)) {
              return;
            }
          }
          chunk_task_->MarkStackPush(ref);
        }
      });
    }

   private:
    MarkStackTask<kUseFinger>* const chunk_task_;
  };

  virtual ~MarkStackTask() {
    // Make sure that we have cleared our mark stack.
    DCHECK_EQ(mark_stack_pos_, 0U);
    if (kCountTasks) {
      ++mark_sweep_->work_chunks_deleted_;
    }
  }

  MarkSweep* const mark_sweep_;
  ThreadPool* const thread_pool_;
  // Thread local mark stack for this task.
  const Object* mark_stack_[kMaxSize];
  // Mark stack position.
  size_t mark_stack_pos_;

  void MarkStackPush(const Object* obj) ALWAYS_INLINE {
    if (UNLIKELY(mark_stack_pos_ == kMaxSize)) {
      // Mark stack overflow, give 1/2 the stack to the thread pool as a new work task.
      mark_stack_pos_ /= 2;
      auto* task = new MarkStackTask(thread_pool_, mark_sweep_, kMaxSize - mark_stack_pos_,
                                     mark_stack_ + mark_stack_pos_);
      thread_pool_->AddTask(Thread::Current(), task);
    }
    DCHECK(obj != nullptr);
    DCHECK(mark_stack_pos_ < kMaxSize);
    mark_stack_[mark_stack_pos_++] = obj;
  }

  virtual void Finalize() {
    delete this;
  }

  // Scans all of the objects
  virtual void Run(Thread* self) {
    ScanObjectParallelVisitor visitor(this);
    // TODO: Tune this.
    static const size_t kFifoSize = 4;
    BoundedFifoPowerOfTwo<const Object*, kFifoSize> prefetch_fifo;
    for (;;) {
      const Object* obj = nullptr;
      if (kUseMarkStackPrefetch) {
        while (mark_stack_pos_ != 0 && prefetch_fifo.size() < kFifoSize) {
          const Object* obj = mark_stack_[--mark_stack_pos_];
          DCHECK(obj != nullptr);
          __builtin_prefetch(obj);
          prefetch_fifo.push_back(obj);
        }
        if (UNLIKELY(prefetch_fifo.empty())) {
          break;
        }
        obj = prefetch_fifo.front();
        prefetch_fifo.pop_front();
      } else {
        if (UNLIKELY(mark_stack_pos_ == 0)) {
          break;
        }
        obj = mark_stack_[--mark_stack_pos_];
      }
      DCHECK(obj != nullptr);
      visitor(const_cast<mirror::Object*>(obj));
    }
  }
};

class CardScanTask : public MarkStackTask<false> {
 public:
  CardScanTask(ThreadPool* thread_pool, MarkSweep* mark_sweep, accounting::SpaceBitmap* bitmap,
               byte* begin, byte* end, byte minimum_age, size_t mark_stack_size,
               const Object** mark_stack_obj)
      : MarkStackTask<false>(thread_pool, mark_sweep, mark_stack_size, mark_stack_obj),
        bitmap_(bitmap),
        begin_(begin),
        end_(end),
        minimum_age_(minimum_age) {
  }

 protected:
  accounting::SpaceBitmap* const bitmap_;
  byte* const begin_;
  byte* const end_;
  const byte minimum_age_;

  virtual void Finalize() {
    delete this;
  }

  virtual void Run(Thread* self) NO_THREAD_SAFETY_ANALYSIS {
    ScanObjectParallelVisitor visitor(this);
    accounting::CardTable* card_table = mark_sweep_->GetHeap()->GetCardTable();
    size_t cards_scanned = card_table->Scan(bitmap_, begin_, end_, visitor, minimum_age_);
    VLOG(heap) << "Parallel scanning cards " << reinterpret_cast<void*>(begin_) << " - "
        << reinterpret_cast<void*>(end_) << " = " << cards_scanned;
    // Finish by emptying our local mark stack.
    MarkStackTask::Run(self);
  }
};

size_t MarkSweep::GetThreadCount(bool paused) const {
  if (heap_->GetThreadPool() == nullptr || !heap_->CareAboutPauseTimes()) {
    return 0;
  }
  if (paused) {
    return heap_->GetParallelGCThreadCount() + 1;
  } else {
    return heap_->GetConcGCThreadCount() + 1;
  }
}

void MarkSweep::ScanGrayObjects(bool paused, byte minimum_age) {
  accounting::CardTable* card_table = GetHeap()->GetCardTable();
  ThreadPool* thread_pool = GetHeap()->GetThreadPool();
  size_t thread_count = GetThreadCount(paused);
  // The parallel version with only one thread is faster for card scanning, TODO: fix.
  if (kParallelCardScan && thread_count > 0) {
    Thread* self = Thread::Current();
    // Can't have a different split for each space since multiple spaces can have their cards being
    // scanned at the same time.
    timings_.StartSplit(paused ? "(Paused)ScanGrayObjects" : "ScanGrayObjects");
    // Try to take some of the mark stack since we can pass this off to the worker tasks.
    const Object** mark_stack_begin = const_cast<const Object**>(mark_stack_->Begin());
    const Object** mark_stack_end = const_cast<const Object**>(mark_stack_->End());
    const size_t mark_stack_size = mark_stack_end - mark_stack_begin;
    // Estimated number of work tasks we will create.
    const size_t mark_stack_tasks = GetHeap()->GetContinuousSpaces().size() * thread_count;
    DCHECK_NE(mark_stack_tasks, 0U);
    const size_t mark_stack_delta = std::min(CardScanTask::kMaxSize / 2,
                                             mark_stack_size / mark_stack_tasks + 1);
    for (const auto& space : GetHeap()->GetContinuousSpaces()) {
      if (space->GetMarkBitmap() == nullptr) {
        continue;
      }
      byte* card_begin = space->Begin();
      byte* card_end = space->End();
      // Align up the end address. For example, the image space's end
      // may not be card-size-aligned.
      card_end = AlignUp(card_end, accounting::CardTable::kCardSize);
      DCHECK(IsAligned<accounting::CardTable::kCardSize>(card_begin));
      DCHECK(IsAligned<accounting::CardTable::kCardSize>(card_end));
      // Calculate how many bytes of heap we will scan,
      const size_t address_range = card_end - card_begin;
      // Calculate how much address range each task gets.
      const size_t card_delta = RoundUp(address_range / thread_count + 1,
                                        accounting::CardTable::kCardSize);
      // Create the worker tasks for this space.
      while (card_begin != card_end) {
        // Add a range of cards.
        size_t addr_remaining = card_end - card_begin;
        size_t card_increment = std::min(card_delta, addr_remaining);
        // Take from the back of the mark stack.
        size_t mark_stack_remaining = mark_stack_end - mark_stack_begin;
        size_t mark_stack_increment = std::min(mark_stack_delta, mark_stack_remaining);
        mark_stack_end -= mark_stack_increment;
        mark_stack_->PopBackCount(static_cast<int32_t>(mark_stack_increment));
        DCHECK_EQ(mark_stack_end, mark_stack_->End());
        // Add the new task to the thread pool.
        auto* task = new CardScanTask(thread_pool, this, space->GetMarkBitmap(), card_begin,
                                      card_begin + card_increment, minimum_age,
                                      mark_stack_increment, mark_stack_end);
        thread_pool->AddTask(self, task);
        card_begin += card_increment;
      }
    }

    // Note: the card scan below may dirty new cards (and scan them)
    // as a side effect when a Reference object is encountered and
    // queued during the marking. See b/11465268.
    thread_pool->SetMaxActiveWorkers(thread_count - 1);
    thread_pool->StartWorkers(self);
    thread_pool->Wait(self, true, true);
    thread_pool->StopWorkers(self);
    timings_.EndSplit();
  } else {
    for (const auto& space : GetHeap()->GetContinuousSpaces()) {
      if (space->GetMarkBitmap() != nullptr) {
        // Image spaces are handled properly since live == marked for them.
        switch (space->GetGcRetentionPolicy()) {
          case space::kGcRetentionPolicyNeverCollect:
            timings_.StartSplit(paused ? "(Paused)ScanGrayImageSpaceObjects" :
                "ScanGrayImageSpaceObjects");
            break;
          case space::kGcRetentionPolicyFullCollect:
            timings_.StartSplit(paused ? "(Paused)ScanGrayZygoteSpaceObjects" :
                "ScanGrayZygoteSpaceObjects");
            break;
          case space::kGcRetentionPolicyAlwaysCollect:
            timings_.StartSplit(paused ? "(Paused)ScanGrayAllocSpaceObjects" :
                "ScanGrayAllocSpaceObjects");
            break;
          }
        ScanObjectVisitor visitor(this);
        card_table->Scan(space->GetMarkBitmap(), space->Begin(), space->End(), visitor, minimum_age);
        timings_.EndSplit();
      }
    }
  }
}

class RecursiveMarkTask : public MarkStackTask<false> {
 public:
  RecursiveMarkTask(ThreadPool* thread_pool, MarkSweep* mark_sweep,
                    accounting::SpaceBitmap* bitmap, uintptr_t begin, uintptr_t end)
      : MarkStackTask<false>(thread_pool, mark_sweep, 0, NULL),
        bitmap_(bitmap),
        begin_(begin),
        end_(end) {
  }

 protected:
  accounting::SpaceBitmap* const bitmap_;
  const uintptr_t begin_;
  const uintptr_t end_;

  virtual void Finalize() {
    delete this;
  }

  // Scans all of the objects
  virtual void Run(Thread* self) NO_THREAD_SAFETY_ANALYSIS {
    ScanObjectParallelVisitor visitor(this);
    bitmap_->VisitMarkedRange(begin_, end_, visitor);
    // Finish by emptying our local mark stack.
    MarkStackTask::Run(self);
  }
};

// Populates the mark stack based on the set of marked objects and
// recursively marks until the mark stack is emptied.
void MarkSweep::RecursiveMark() {
  TimingLogger::ScopedSplit split("RecursiveMark", &timings_);
  // RecursiveMark will build the lists of known instances of the Reference classes. See
  // DelayReferenceReferent for details.
  if (kUseRecursiveMark) {
    const bool partial = GetGcType() == kGcTypePartial;
    ScanObjectVisitor scan_visitor(this);
    auto* self = Thread::Current();
    ThreadPool* thread_pool = heap_->GetThreadPool();
    size_t thread_count = GetThreadCount(false);
    const bool parallel = kParallelRecursiveMark && thread_count > 1;
    mark_stack_->Reset();
    for (const auto& space : GetHeap()->GetContinuousSpaces()) {
      if ((space->GetGcRetentionPolicy() == space::kGcRetentionPolicyAlwaysCollect) ||
          (!partial && space->GetGcRetentionPolicy() == space::kGcRetentionPolicyFullCollect)) {
        current_mark_bitmap_ = space->GetMarkBitmap();
        if (current_mark_bitmap_ == nullptr) {
          continue;
        }
        if (parallel) {
          // We will use the mark stack the future.
          // CHECK(mark_stack_->IsEmpty());
          // This function does not handle heap end increasing, so we must use the space end.
          uintptr_t begin = reinterpret_cast<uintptr_t>(space->Begin());
          uintptr_t end = reinterpret_cast<uintptr_t>(space->End());
          atomic_finger_ = static_cast<int32_t>(0xFFFFFFFF);

          // Create a few worker tasks.
          const size_t n = thread_count * 2;
          while (begin != end) {
            uintptr_t start = begin;
            uintptr_t delta = (end - begin) / n;
            delta = RoundUp(delta, KB);
            if (delta < 16 * KB) delta = end - begin;
            begin += delta;
            auto* task = new RecursiveMarkTask(thread_pool, this, current_mark_bitmap_, start,
                                               begin);
            thread_pool->AddTask(self, task);
          }
          thread_pool->SetMaxActiveWorkers(thread_count - 1);
          thread_pool->StartWorkers(self);
          thread_pool->Wait(self, true, true);
          thread_pool->StopWorkers(self);
        } else {
          // This function does not handle heap end increasing, so we must use the space end.
          uintptr_t begin = reinterpret_cast<uintptr_t>(space->Begin());
          uintptr_t end = reinterpret_cast<uintptr_t>(space->End());
          current_mark_bitmap_->VisitMarkedRange(begin, end, scan_visitor);
        }
      }
    }
  }
  ProcessMarkStack(false);
}

mirror::Object* MarkSweep::IsMarkedCallback(Object* object, void* arg) {
  if (reinterpret_cast<MarkSweep*>(arg)->IsMarked(object)) {
    return object;
  }
  return nullptr;
}

void MarkSweep::RecursiveMarkDirtyObjects(bool paused, byte minimum_age) {
  ScanGrayObjects(paused, minimum_age);
  ProcessMarkStack(paused);
}

void MarkSweep::ReMarkRoots() {
  timings_.StartSplit("ReMarkRoots");
  Runtime::Current()->VisitRoots(MarkRootCallback, this, true, true);
  timings_.EndSplit();
}

void MarkSweep::SweepSystemWeaks() {
  Runtime* runtime = Runtime::Current();
  timings_.StartSplit("SweepSystemWeaks");
  runtime->SweepSystemWeaks(IsMarkedCallback, this);
  timings_.EndSplit();
}

mirror::Object* MarkSweep::VerifySystemWeakIsLiveCallback(Object* obj, void* arg) {
  reinterpret_cast<MarkSweep*>(arg)->VerifyIsLive(obj);
  // We don't actually want to sweep the object, so lets return "marked"
  return obj;
}

void MarkSweep::VerifyIsLive(const Object* obj) {
  Heap* heap = GetHeap();
  if (!heap->GetLiveBitmap()->Test(obj)) {
    space::LargeObjectSpace* large_object_space = GetHeap()->GetLargeObjectsSpace();
    if (!large_object_space->GetLiveObjects()->Test(obj)) {
      if (std::find(heap->allocation_stack_->Begin(), heap->allocation_stack_->End(), obj) ==
          heap->allocation_stack_->End()) {
        // Object not found!
        heap->DumpSpaces();
        LOG(FATAL) << "Found dead object " << obj;
      }
    }
  }
}

void MarkSweep::VerifySystemWeaks() {
  // Verify system weaks, uses a special object visitor which returns the input object.
  Runtime::Current()->SweepSystemWeaks(VerifySystemWeakIsLiveCallback, this);
}

class CheckpointMarkThreadRoots : public Closure {
 public:
  explicit CheckpointMarkThreadRoots(MarkSweep* mark_sweep) : mark_sweep_(mark_sweep) {}

  virtual void Run(Thread* thread) NO_THREAD_SAFETY_ANALYSIS {
    ATRACE_BEGIN("Marking thread roots");
    // Note: self is not necessarily equal to thread since thread may be suspended.
    Thread* self = Thread::Current();
    CHECK(thread == self || thread->IsSuspended() || thread->GetState() == kWaitingPerformingGc)
        << thread->GetState() << " thread " << thread << " self " << self;
    thread->VisitRoots(MarkSweep::MarkRootParallelCallback, mark_sweep_);
    ATRACE_END();
    mark_sweep_->GetBarrier().Pass(self);
  }

 private:
  MarkSweep* mark_sweep_;
};

void MarkSweep::MarkRootsCheckpoint(Thread* self) {
  CheckpointMarkThreadRoots check_point(this);
  timings_.StartSplit("MarkRootsCheckpoint");
  ThreadList* thread_list = Runtime::Current()->GetThreadList();
  // Request the check point is run on all threads returning a count of the threads that must
  // run through the barrier including self.
  size_t barrier_count = thread_list->RunCheckpoint(&check_point);
  // Release locks then wait for all mutator threads to pass the barrier.
  // TODO: optimize to not release locks when there are no threads to wait for.
  Locks::heap_bitmap_lock_->ExclusiveUnlock(self);
  Locks::mutator_lock_->SharedUnlock(self);
  ThreadState old_state = self->SetState(kWaitingForCheckPointsToRun);
  CHECK_EQ(old_state, kWaitingPerformingGc);
  gc_barrier_->Increment(self, barrier_count);
  self->SetState(kWaitingPerformingGc);
  Locks::mutator_lock_->SharedLock(self);
  Locks::heap_bitmap_lock_->ExclusiveLock(self);
  timings_.EndSplit();
}

void MarkSweep::SweepArray(accounting::ObjectStack* allocations, bool swap_bitmaps) {
  timings_.StartSplit("SweepArray");
  Thread* self = Thread::Current();
  mirror::Object* chunk_free_buffer[kSweepArrayChunkFreeSize];
  size_t chunk_free_pos = 0;
  size_t freed_bytes = 0;
  size_t freed_large_object_bytes = 0;
  size_t freed_objects = 0;
  size_t freed_large_objects = 0;
  // How many objects are left in the array, modified after each space is swept.
  Object** objects = const_cast<Object**>(allocations->Begin());
  size_t count = allocations->Size();
  // Change the order to ensure that the non-moving space last swept as an optimization.
  std::vector<space::ContinuousSpace*> sweep_spaces;
  space::ContinuousSpace* non_moving_space = nullptr;
  for (space::ContinuousSpace* space : heap_->GetContinuousSpaces()) {
    if (space->IsAllocSpace() && !IsImmuneSpace(space) && space->GetLiveBitmap() != nullptr) {
      if (space == heap_->GetNonMovingSpace()) {
        non_moving_space = space;
      } else {
        sweep_spaces.push_back(space);
      }
    }
  }
  // Unlikely to sweep a significant amount of non_movable objects, so we do these after the after
  // the other alloc spaces as an optimization.
  if (non_moving_space != nullptr) {
    sweep_spaces.push_back(non_moving_space);
  }
  // Start by sweeping the continuous spaces.
  for (space::ContinuousSpace* space : sweep_spaces) {
    space::AllocSpace* alloc_space = space->AsAllocSpace();
    accounting::SpaceBitmap* live_bitmap = space->GetLiveBitmap();
    accounting::SpaceBitmap* mark_bitmap = space->GetMarkBitmap();
    if (swap_bitmaps) {
      std::swap(live_bitmap, mark_bitmap);
    }
    Object** out = objects;
    for (size_t i = 0; i < count; ++i) {
      Object* obj = objects[i];
      if (space->HasAddress(obj)) {
        // This object is in the space, remove it from the array and add it to the sweep buffer
        // if needed.
        if (!mark_bitmap->Test(obj)) {
          if (chunk_free_pos >= kSweepArrayChunkFreeSize) {
            timings_.StartSplit("FreeList");
            freed_objects += chunk_free_pos;
            freed_bytes += alloc_space->FreeList(self, chunk_free_pos, chunk_free_buffer);
            timings_.EndSplit();
            chunk_free_pos = 0;
          }
          chunk_free_buffer[chunk_free_pos++] = obj;
        }
      } else {
        *(out++) = obj;
      }
    }
    if (chunk_free_pos > 0) {
      timings_.StartSplit("FreeList");
      freed_objects += chunk_free_pos;
      freed_bytes += alloc_space->FreeList(self, chunk_free_pos, chunk_free_buffer);
      timings_.EndSplit();
      chunk_free_pos = 0;
    }
    // All of the references which space contained are no longer in the allocation stack, update
    // the count.
    count = out - objects;
  }
  // Handle the large object space.
  space::LargeObjectSpace* large_object_space = GetHeap()->GetLargeObjectsSpace();
  accounting::ObjectSet* large_live_objects = large_object_space->GetLiveObjects();
  accounting::ObjectSet* large_mark_objects = large_object_space->GetMarkObjects();
  if (swap_bitmaps) {
    std::swap(large_live_objects, large_mark_objects);
  }
  for (size_t i = 0; i < count; ++i) {
    Object* obj = objects[i];
    // Handle large objects.
    if (!large_mark_objects->Test(obj)) {
      ++freed_large_objects;
      freed_large_object_bytes += large_object_space->Free(self, obj);
    }
  }
  timings_.EndSplit();

  timings_.StartSplit("RecordFree");
  VLOG(heap) << "Freed " << freed_objects << "/" << count
             << " objects with size " << PrettySize(freed_bytes);
  heap_->RecordFree(freed_objects + freed_large_objects, freed_bytes + freed_large_object_bytes);
  freed_objects_.FetchAndAdd(freed_objects);
  freed_large_objects_.FetchAndAdd(freed_large_objects);
  freed_bytes_.FetchAndAdd(freed_bytes);
  freed_large_object_bytes_.FetchAndAdd(freed_large_object_bytes);
  timings_.EndSplit();

  timings_.StartSplit("ResetStack");
  allocations->Reset();
  timings_.EndSplit();
}

void MarkSweep::Sweep(bool swap_bitmaps) {
  DCHECK(mark_stack_->IsEmpty());
  TimingLogger::ScopedSplit("Sweep", &timings_);
  for (const auto& space : GetHeap()->GetContinuousSpaces()) {
    if (space->IsContinuousMemMapAllocSpace()) {
      space::ContinuousMemMapAllocSpace* alloc_space = space->AsContinuousMemMapAllocSpace();
      TimingLogger::ScopedSplit split(
          alloc_space->IsZygoteSpace() ? "SweepZygoteSpace" : "SweepMallocSpace", &timings_);
      size_t freed_objects = 0;
      size_t freed_bytes = 0;
      alloc_space->Sweep(swap_bitmaps, &freed_objects, &freed_bytes);
      heap_->RecordFree(freed_objects, freed_bytes);
      freed_objects_.FetchAndAdd(freed_objects);
      freed_bytes_.FetchAndAdd(freed_bytes);
    }
  }
  SweepLargeObjects(swap_bitmaps);
}

void MarkSweep::SweepLargeObjects(bool swap_bitmaps) {
  TimingLogger::ScopedSplit("SweepLargeObjects", &timings_);
  size_t freed_objects = 0;
  size_t freed_bytes = 0;
  GetHeap()->GetLargeObjectsSpace()->Sweep(swap_bitmaps, &freed_objects, &freed_bytes);
  freed_large_objects_.FetchAndAdd(freed_objects);
  freed_large_object_bytes_.FetchAndAdd(freed_bytes);
  GetHeap()->RecordFree(freed_objects, freed_bytes);
}

// Process the "referent" field in a java.lang.ref.Reference.  If the
// referent has not yet been marked, put it on the appropriate list in
// the heap for later processing.
void MarkSweep::DelayReferenceReferent(mirror::Class* klass, Object* obj) {
  DCHECK(klass != nullptr);
  DCHECK(klass->IsReferenceClass());
  DCHECK(obj != NULL);
  heap_->DelayReferenceReferent(klass, obj, IsMarkedCallback, this);
}

class MarkObjectVisitor {
 public:
  explicit MarkObjectVisitor(MarkSweep* const mark_sweep) ALWAYS_INLINE : mark_sweep_(mark_sweep) {}

  // TODO: Fixme when anotatalysis works with visitors.
  void operator()(const Object* /* obj */, const Object* ref, const MemberOffset& /* offset */,
                  bool /* is_static */) const ALWAYS_INLINE
      NO_THREAD_SAFETY_ANALYSIS {
    if (kCheckLocks) {
      Locks::mutator_lock_->AssertSharedHeld(Thread::Current());
      Locks::heap_bitmap_lock_->AssertExclusiveHeld(Thread::Current());
    }
    mark_sweep_->MarkObject(ref);
  }

 private:
  MarkSweep* const mark_sweep_;
};

// Scans an object reference.  Determines the type of the reference
// and dispatches to a specialized scanning routine.
void MarkSweep::ScanObject(Object* obj) {
  MarkObjectVisitor visitor(this);
  ScanObjectVisit(obj, visitor);
}

void MarkSweep::ProcessMarkStackParallel(size_t thread_count) {
  Thread* self = Thread::Current();
  ThreadPool* thread_pool = GetHeap()->GetThreadPool();
  const size_t chunk_size = std::min(mark_stack_->Size() / thread_count + 1,
                                     static_cast<size_t>(MarkStackTask<false>::kMaxSize));
  CHECK_GT(chunk_size, 0U);
  // Split the current mark stack up into work tasks.
  for (mirror::Object **it = mark_stack_->Begin(), **end = mark_stack_->End(); it < end; ) {
    const size_t delta = std::min(static_cast<size_t>(end - it), chunk_size);
    thread_pool->AddTask(self, new MarkStackTask<false>(thread_pool, this, delta,
                                                        const_cast<const mirror::Object**>(it)));
    it += delta;
  }
  thread_pool->SetMaxActiveWorkers(thread_count - 1);
  thread_pool->StartWorkers(self);
  thread_pool->Wait(self, true, true);
  thread_pool->StopWorkers(self);
  mark_stack_->Reset();
  CHECK_EQ(work_chunks_created_, work_chunks_deleted_) << " some of the work chunks were leaked";
}

// Scan anything that's on the mark stack.
void MarkSweep::ProcessMarkStack(bool paused) {
  timings_.StartSplit("ProcessMarkStack");
  size_t thread_count = GetThreadCount(paused);
  if (kParallelProcessMarkStack && thread_count > 1 &&
      mark_stack_->Size() >= kMinimumParallelMarkStackSize) {
    ProcessMarkStackParallel(thread_count);
  } else {
    // TODO: Tune this.
    static const size_t kFifoSize = 4;
    BoundedFifoPowerOfTwo<Object*, kFifoSize> prefetch_fifo;
    for (;;) {
      Object* obj = NULL;
      if (kUseMarkStackPrefetch) {
        while (!mark_stack_->IsEmpty() && prefetch_fifo.size() < kFifoSize) {
          Object* obj = mark_stack_->PopBack();
          DCHECK(obj != NULL);
          __builtin_prefetch(obj);
          prefetch_fifo.push_back(obj);
        }
        if (prefetch_fifo.empty()) {
          break;
        }
        obj = prefetch_fifo.front();
        prefetch_fifo.pop_front();
      } else {
        if (mark_stack_->IsEmpty()) {
          break;
        }
        obj = mark_stack_->PopBack();
      }
      DCHECK(obj != NULL);
      ScanObject(obj);
    }
  }
  timings_.EndSplit();
}

inline bool MarkSweep::IsMarked(const Object* object) const
    SHARED_LOCKS_REQUIRED(Locks::heap_bitmap_lock_) {
  if (IsImmune(object)) {
    return true;
  }
  DCHECK(current_mark_bitmap_ != NULL);
  if (current_mark_bitmap_->HasAddress(object)) {
    return current_mark_bitmap_->Test(object);
  }
  return heap_->GetMarkBitmap()->Test(object);
}

void MarkSweep::FinishPhase() {
  TimingLogger::ScopedSplit split("FinishPhase", &timings_);
  // Can't enqueue references if we hold the mutator lock.
  Heap* heap = GetHeap();
  timings_.NewSplit("PostGcVerification");
  heap->PostGcVerification(this);

  timings_.NewSplit("RequestHeapTrim");
  heap->RequestHeapTrim();

  // Update the cumulative statistics
  total_freed_objects_ += GetFreedObjects() + GetFreedLargeObjects();
  total_freed_bytes_ += GetFreedBytes() + GetFreedLargeObjectBytes();

  // Ensure that the mark stack is empty.
  CHECK(mark_stack_->IsEmpty());

  if (kCountScannedTypes) {
    VLOG(gc) << "MarkSweep scanned classes=" << class_count_ << " arrays=" << array_count_
             << " other=" << other_count_;
  }

  if (kCountTasks) {
    VLOG(gc) << "Total number of work chunks allocated: " << work_chunks_created_;
  }

  if (kMeasureOverhead) {
    VLOG(gc) << "Overhead time " << PrettyDuration(overhead_time_);
  }

  if (kProfileLargeObjects) {
    VLOG(gc) << "Large objects tested " << large_object_test_ << " marked " << large_object_mark_;
  }

  if (kCountClassesMarked) {
    VLOG(gc) << "Classes marked " << classes_marked_;
  }

  if (kCountJavaLangRefs) {
    VLOG(gc) << "References scanned " << reference_count_;
  }

  // Update the cumulative loggers.
  cumulative_timings_.Start();
  cumulative_timings_.AddLogger(timings_);
  cumulative_timings_.End();

  // Clear all of the spaces' mark bitmaps.
  for (const auto& space : GetHeap()->GetContinuousSpaces()) {
    accounting::SpaceBitmap* bitmap = space->GetMarkBitmap();
    if (bitmap != nullptr &&
        space->GetGcRetentionPolicy() != space::kGcRetentionPolicyNeverCollect) {
      bitmap->Clear();
    }
  }
  mark_stack_->Reset();

  // Reset the marked large objects.
  space::LargeObjectSpace* large_objects = GetHeap()->GetLargeObjectsSpace();
  large_objects->GetMarkObjects()->Clear();
}

}  // namespace collector
}  // namespace gc
}  // namespace art