view src/share/vm/gc_implementation/parallelScavenge/psScavenge.cpp @ 453:c96030fff130

6684579: SoftReference processing can be made more efficient Summary: For current soft-ref clearing policies, we can decide at marking time if a soft-reference will definitely not be cleared, postponing the decision of whether it will definitely be cleared to the final reference processing phase. This can be especially beneficial in the case of concurrent collectors where the marking is usually concurrent but reference processing is usually not. Reviewed-by: jmasa
author ysr
date Thu, 20 Nov 2008 16:56:09 -0800
parents 850fdf70db2b
children 27a80744a83b
line wrap: on
line source

/*
 * Copyright 2002-2008 Sun Microsystems, Inc.  All Rights Reserved.
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 *
 * This code is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 only, as
 * published by the Free Software Foundation.
 *
 * This code is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 * version 2 for more details (a copy is included in the LICENSE file that
 * accompanied this code).
 *
 * You should have received a copy of the GNU General Public License version
 * 2 along with this work; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 *
 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
 * CA 95054 USA or visit www.sun.com if you need additional information or
 * have any questions.
 *
 */


# include "incls/_precompiled.incl"
# include "incls/_psScavenge.cpp.incl"

HeapWord*                  PSScavenge::_to_space_top_before_gc = NULL;
int                        PSScavenge::_consecutive_skipped_scavenges = 0;
ReferenceProcessor*        PSScavenge::_ref_processor = NULL;
CardTableExtension*        PSScavenge::_card_table = NULL;
bool                       PSScavenge::_survivor_overflow = false;
int                        PSScavenge::_tenuring_threshold = 0;
HeapWord*                  PSScavenge::_young_generation_boundary = NULL;
elapsedTimer               PSScavenge::_accumulated_time;
GrowableArray<markOop>*    PSScavenge::_preserved_mark_stack = NULL;
GrowableArray<oop>*        PSScavenge::_preserved_oop_stack = NULL;
CollectorCounters*         PSScavenge::_counters = NULL;

// Define before use
class PSIsAliveClosure: public BoolObjectClosure {
public:
  void do_object(oop p) {
    assert(false, "Do not call.");
  }
  bool do_object_b(oop p) {
    return (!PSScavenge::is_obj_in_young((HeapWord*) p)) || p->is_forwarded();
  }
};

PSIsAliveClosure PSScavenge::_is_alive_closure;

class PSKeepAliveClosure: public OopClosure {
protected:
  MutableSpace* _to_space;
  PSPromotionManager* _promotion_manager;

public:
  PSKeepAliveClosure(PSPromotionManager* pm) : _promotion_manager(pm) {
    ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
    assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
    _to_space = heap->young_gen()->to_space();

    assert(_promotion_manager != NULL, "Sanity");
  }

  template <class T> void do_oop_work(T* p) {
    assert (!oopDesc::is_null(*p), "expected non-null ref");
    assert ((oopDesc::load_decode_heap_oop_not_null(p))->is_oop(),
            "expected an oop while scanning weak refs");

    // Weak refs may be visited more than once.
    if (PSScavenge::should_scavenge(p, _to_space)) {
      PSScavenge::copy_and_push_safe_barrier(_promotion_manager, p);
    }
  }
  virtual void do_oop(oop* p)       { PSKeepAliveClosure::do_oop_work(p); }
  virtual void do_oop(narrowOop* p) { PSKeepAliveClosure::do_oop_work(p); }
};

class PSEvacuateFollowersClosure: public VoidClosure {
 private:
  PSPromotionManager* _promotion_manager;
 public:
  PSEvacuateFollowersClosure(PSPromotionManager* pm) : _promotion_manager(pm) {}

  virtual void do_void() {
    assert(_promotion_manager != NULL, "Sanity");
    _promotion_manager->drain_stacks(true);
    guarantee(_promotion_manager->stacks_empty(),
              "stacks should be empty at this point");
  }
};

class PSPromotionFailedClosure : public ObjectClosure {
  virtual void do_object(oop obj) {
    if (obj->is_forwarded()) {
      obj->init_mark();
    }
  }
};

class PSRefProcTaskProxy: public GCTask {
  typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask;
  ProcessTask & _rp_task;
  uint          _work_id;
public:
  PSRefProcTaskProxy(ProcessTask & rp_task, uint work_id)
    : _rp_task(rp_task),
      _work_id(work_id)
  { }

private:
  virtual char* name() { return (char *)"Process referents by policy in parallel"; }
  virtual void do_it(GCTaskManager* manager, uint which);
};

void PSRefProcTaskProxy::do_it(GCTaskManager* manager, uint which)
{
  PSPromotionManager* promotion_manager =
    PSPromotionManager::gc_thread_promotion_manager(which);
  assert(promotion_manager != NULL, "sanity check");
  PSKeepAliveClosure keep_alive(promotion_manager);
  PSEvacuateFollowersClosure evac_followers(promotion_manager);
  PSIsAliveClosure is_alive;
  _rp_task.work(_work_id, is_alive, keep_alive, evac_followers);
}

class PSRefEnqueueTaskProxy: public GCTask {
  typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask;
  EnqueueTask& _enq_task;
  uint         _work_id;

public:
  PSRefEnqueueTaskProxy(EnqueueTask& enq_task, uint work_id)
    : _enq_task(enq_task),
      _work_id(work_id)
  { }

  virtual char* name() { return (char *)"Enqueue reference objects in parallel"; }
  virtual void do_it(GCTaskManager* manager, uint which)
  {
    _enq_task.work(_work_id);
  }
};

class PSRefProcTaskExecutor: public AbstractRefProcTaskExecutor {
  virtual void execute(ProcessTask& task);
  virtual void execute(EnqueueTask& task);
};

void PSRefProcTaskExecutor::execute(ProcessTask& task)
{
  GCTaskQueue* q = GCTaskQueue::create();
  for(uint i=0; i<ParallelGCThreads; i++) {
    q->enqueue(new PSRefProcTaskProxy(task, i));
  }
  ParallelTaskTerminator terminator(
    ParallelScavengeHeap::gc_task_manager()->workers(),
    UseDepthFirstScavengeOrder ?
        (TaskQueueSetSuper*) PSPromotionManager::stack_array_depth()
      : (TaskQueueSetSuper*) PSPromotionManager::stack_array_breadth());
  if (task.marks_oops_alive() && ParallelGCThreads > 1) {
    for (uint j=0; j<ParallelGCThreads; j++) {
      q->enqueue(new StealTask(&terminator));
    }
  }
  ParallelScavengeHeap::gc_task_manager()->execute_and_wait(q);
}


void PSRefProcTaskExecutor::execute(EnqueueTask& task)
{
  GCTaskQueue* q = GCTaskQueue::create();
  for(uint i=0; i<ParallelGCThreads; i++) {
    q->enqueue(new PSRefEnqueueTaskProxy(task, i));
  }
  ParallelScavengeHeap::gc_task_manager()->execute_and_wait(q);
}

// This method contains all heap specific policy for invoking scavenge.
// PSScavenge::invoke_no_policy() will do nothing but attempt to
// scavenge. It will not clean up after failed promotions, bail out if
// we've exceeded policy time limits, or any other special behavior.
// All such policy should be placed here.
//
// Note that this method should only be called from the vm_thread while
// at a safepoint!
void PSScavenge::invoke()
{
  assert(SafepointSynchronize::is_at_safepoint(), "should be at safepoint");
  assert(Thread::current() == (Thread*)VMThread::vm_thread(), "should be in vm thread");
  assert(!Universe::heap()->is_gc_active(), "not reentrant");

  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");

  PSAdaptiveSizePolicy* policy = heap->size_policy();

  // Before each allocation/collection attempt, find out from the
  // policy object if GCs are, on the whole, taking too long. If so,
  // bail out without attempting a collection.
  if (!policy->gc_time_limit_exceeded()) {
    IsGCActiveMark mark;

    bool scavenge_was_done = PSScavenge::invoke_no_policy();

    PSGCAdaptivePolicyCounters* counters = heap->gc_policy_counters();
    if (UsePerfData)
      counters->update_full_follows_scavenge(0);
    if (!scavenge_was_done ||
        policy->should_full_GC(heap->old_gen()->free_in_bytes())) {
      if (UsePerfData)
        counters->update_full_follows_scavenge(full_follows_scavenge);

      GCCauseSetter gccs(heap, GCCause::_adaptive_size_policy);
      if (UseParallelOldGC) {
        PSParallelCompact::invoke_no_policy(false);
      } else {
        PSMarkSweep::invoke_no_policy(false);
      }
    }
  }
}

// This method contains no policy. You should probably
// be calling invoke() instead.
bool PSScavenge::invoke_no_policy() {
  assert(SafepointSynchronize::is_at_safepoint(), "should be at safepoint");
  assert(Thread::current() == (Thread*)VMThread::vm_thread(), "should be in vm thread");

  TimeStamp scavenge_entry;
  TimeStamp scavenge_midpoint;
  TimeStamp scavenge_exit;

  scavenge_entry.update();

  if (GC_locker::check_active_before_gc()) {
    return false;
  }

  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
  GCCause::Cause gc_cause = heap->gc_cause();
  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");

  // Check for potential problems.
  if (!should_attempt_scavenge()) {
    return false;
  }

  bool promotion_failure_occurred = false;

  PSYoungGen* young_gen = heap->young_gen();
  PSOldGen* old_gen = heap->old_gen();
  PSPermGen* perm_gen = heap->perm_gen();
  PSAdaptiveSizePolicy* size_policy = heap->size_policy();
  heap->increment_total_collections();

  AdaptiveSizePolicyOutput(size_policy, heap->total_collections());

  if ((gc_cause != GCCause::_java_lang_system_gc) ||
       UseAdaptiveSizePolicyWithSystemGC) {
    // Gather the feedback data for eden occupancy.
    young_gen->eden_space()->accumulate_statistics();
  }

  if (ZapUnusedHeapArea) {
    // Save information needed to minimize mangling
    heap->record_gen_tops_before_GC();
  }

  if (PrintHeapAtGC) {
    Universe::print_heap_before_gc();
  }

  assert(!NeverTenure || _tenuring_threshold == markOopDesc::max_age + 1, "Sanity");
  assert(!AlwaysTenure || _tenuring_threshold == 0, "Sanity");

  size_t prev_used = heap->used();
  assert(promotion_failed() == false, "Sanity");

  // Fill in TLABs
  heap->accumulate_statistics_all_tlabs();
  heap->ensure_parsability(true);  // retire TLABs

  if (VerifyBeforeGC && heap->total_collections() >= VerifyGCStartAt) {
    HandleMark hm;  // Discard invalid handles created during verification
    gclog_or_tty->print(" VerifyBeforeGC:");
    Universe::verify(true);
  }

  {
    ResourceMark rm;
    HandleMark hm;

    gclog_or_tty->date_stamp(PrintGC && PrintGCDateStamps);
    TraceCPUTime tcpu(PrintGCDetails, true, gclog_or_tty);
    TraceTime t1("GC", PrintGC, !PrintGCDetails, gclog_or_tty);
    TraceCollectorStats tcs(counters());
    TraceMemoryManagerStats tms(false /* not full GC */);

    if (TraceGen0Time) accumulated_time()->start();

    // Let the size policy know we're starting
    size_policy->minor_collection_begin();

    // Verify the object start arrays.
    if (VerifyObjectStartArray &&
        VerifyBeforeGC) {
      old_gen->verify_object_start_array();
      perm_gen->verify_object_start_array();
    }

    // Verify no unmarked old->young roots
    if (VerifyRememberedSets) {
      CardTableExtension::verify_all_young_refs_imprecise();
    }

    if (!ScavengeWithObjectsInToSpace) {
      assert(young_gen->to_space()->is_empty(),
             "Attempt to scavenge with live objects in to_space");
      young_gen->to_space()->clear(SpaceDecorator::Mangle);
    } else if (ZapUnusedHeapArea) {
      young_gen->to_space()->mangle_unused_area();
    }
    save_to_space_top_before_gc();

    NOT_PRODUCT(reference_processor()->verify_no_references_recorded());
    COMPILER2_PRESENT(DerivedPointerTable::clear());

    reference_processor()->enable_discovery();
    reference_processor()->snap_policy(false);

    // We track how much was promoted to the next generation for
    // the AdaptiveSizePolicy.
    size_t old_gen_used_before = old_gen->used_in_bytes();

    // For PrintGCDetails
    size_t young_gen_used_before = young_gen->used_in_bytes();

    // Reset our survivor overflow.
    set_survivor_overflow(false);

    // We need to save the old/perm top values before
    // creating the promotion_manager. We pass the top
    // values to the card_table, to prevent it from
    // straying into the promotion labs.
    HeapWord* old_top = old_gen->object_space()->top();
    HeapWord* perm_top = perm_gen->object_space()->top();

    // Release all previously held resources
    gc_task_manager()->release_all_resources();

    PSPromotionManager::pre_scavenge();

    // We'll use the promotion manager again later.
    PSPromotionManager* promotion_manager = PSPromotionManager::vm_thread_promotion_manager();
    {
      // TraceTime("Roots");

      GCTaskQueue* q = GCTaskQueue::create();

      for(uint i=0; i<ParallelGCThreads; i++) {
        q->enqueue(new OldToYoungRootsTask(old_gen, old_top, i));
      }

      q->enqueue(new SerialOldToYoungRootsTask(perm_gen, perm_top));

      q->enqueue(new ScavengeRootsTask(ScavengeRootsTask::universe));
      q->enqueue(new ScavengeRootsTask(ScavengeRootsTask::jni_handles));
      // We scan the thread roots in parallel
      Threads::create_thread_roots_tasks(q);
      q->enqueue(new ScavengeRootsTask(ScavengeRootsTask::object_synchronizer));
      q->enqueue(new ScavengeRootsTask(ScavengeRootsTask::flat_profiler));
      q->enqueue(new ScavengeRootsTask(ScavengeRootsTask::management));
      q->enqueue(new ScavengeRootsTask(ScavengeRootsTask::system_dictionary));
      q->enqueue(new ScavengeRootsTask(ScavengeRootsTask::jvmti));

      ParallelTaskTerminator terminator(
        gc_task_manager()->workers(),
        promotion_manager->depth_first() ?
            (TaskQueueSetSuper*) promotion_manager->stack_array_depth()
          : (TaskQueueSetSuper*) promotion_manager->stack_array_breadth());
      if (ParallelGCThreads>1) {
        for (uint j=0; j<ParallelGCThreads; j++) {
          q->enqueue(new StealTask(&terminator));
        }
      }

      gc_task_manager()->execute_and_wait(q);
    }

    scavenge_midpoint.update();

    // Process reference objects discovered during scavenge
    {
      reference_processor()->snap_policy(false); // not always_clear
      PSKeepAliveClosure keep_alive(promotion_manager);
      PSEvacuateFollowersClosure evac_followers(promotion_manager);
      if (reference_processor()->processing_is_mt()) {
        PSRefProcTaskExecutor task_executor;
        reference_processor()->process_discovered_references(
          &_is_alive_closure, &keep_alive, &evac_followers, &task_executor);
      } else {
        reference_processor()->process_discovered_references(
          &_is_alive_closure, &keep_alive, &evac_followers, NULL);
      }
    }

    // Enqueue reference objects discovered during scavenge.
    if (reference_processor()->processing_is_mt()) {
      PSRefProcTaskExecutor task_executor;
      reference_processor()->enqueue_discovered_references(&task_executor);
    } else {
      reference_processor()->enqueue_discovered_references(NULL);
    }

    // Finally, flush the promotion_manager's labs, and deallocate its stacks.
    assert(promotion_manager->claimed_stack_empty(), "Sanity");
    PSPromotionManager::post_scavenge();

    promotion_failure_occurred = promotion_failed();
    if (promotion_failure_occurred) {
      clean_up_failed_promotion();
      if (PrintGC) {
        gclog_or_tty->print("--");
      }
    }

    // Let the size policy know we're done.  Note that we count promotion
    // failure cleanup time as part of the collection (otherwise, we're
    // implicitly saying it's mutator time).
    size_policy->minor_collection_end(gc_cause);

    if (!promotion_failure_occurred) {
      // Swap the survivor spaces.


      young_gen->eden_space()->clear(SpaceDecorator::Mangle);
      young_gen->from_space()->clear(SpaceDecorator::Mangle);
      young_gen->swap_spaces();

      size_t survived = young_gen->from_space()->used_in_bytes();
      size_t promoted = old_gen->used_in_bytes() - old_gen_used_before;
      size_policy->update_averages(_survivor_overflow, survived, promoted);

      if (UseAdaptiveSizePolicy) {
        // Calculate the new survivor size and tenuring threshold

        if (PrintAdaptiveSizePolicy) {
          gclog_or_tty->print("AdaptiveSizeStart: ");
          gclog_or_tty->stamp();
          gclog_or_tty->print_cr(" collection: %d ",
                         heap->total_collections());

          if (Verbose) {
            gclog_or_tty->print("old_gen_capacity: %d young_gen_capacity: %d"
              " perm_gen_capacity: %d ",
              old_gen->capacity_in_bytes(), young_gen->capacity_in_bytes(),
              perm_gen->capacity_in_bytes());
          }
        }


        if (UsePerfData) {
          PSGCAdaptivePolicyCounters* counters = heap->gc_policy_counters();
          counters->update_old_eden_size(
            size_policy->calculated_eden_size_in_bytes());
          counters->update_old_promo_size(
            size_policy->calculated_promo_size_in_bytes());
          counters->update_old_capacity(old_gen->capacity_in_bytes());
          counters->update_young_capacity(young_gen->capacity_in_bytes());
          counters->update_survived(survived);
          counters->update_promoted(promoted);
          counters->update_survivor_overflowed(_survivor_overflow);
        }

        size_t survivor_limit =
          size_policy->max_survivor_size(young_gen->max_size());
        _tenuring_threshold =
          size_policy->compute_survivor_space_size_and_threshold(
                                                           _survivor_overflow,
                                                           _tenuring_threshold,
                                                           survivor_limit);

       if (PrintTenuringDistribution) {
         gclog_or_tty->cr();
         gclog_or_tty->print_cr("Desired survivor size %ld bytes, new threshold %d (max %d)",
                                size_policy->calculated_survivor_size_in_bytes(),
                                _tenuring_threshold, MaxTenuringThreshold);
       }

        if (UsePerfData) {
          PSGCAdaptivePolicyCounters* counters = heap->gc_policy_counters();
          counters->update_tenuring_threshold(_tenuring_threshold);
          counters->update_survivor_size_counters();
        }

        // Do call at minor collections?
        // Don't check if the size_policy is ready at this
        // level.  Let the size_policy check that internally.
        if (UseAdaptiveSizePolicy &&
            UseAdaptiveGenerationSizePolicyAtMinorCollection &&
            ((gc_cause != GCCause::_java_lang_system_gc) ||
              UseAdaptiveSizePolicyWithSystemGC)) {

          // Calculate optimial free space amounts
          assert(young_gen->max_size() >
            young_gen->from_space()->capacity_in_bytes() +
            young_gen->to_space()->capacity_in_bytes(),
            "Sizes of space in young gen are out-of-bounds");
          size_t max_eden_size = young_gen->max_size() -
            young_gen->from_space()->capacity_in_bytes() -
            young_gen->to_space()->capacity_in_bytes();
          size_policy->compute_generation_free_space(young_gen->used_in_bytes(),
                                   young_gen->eden_space()->used_in_bytes(),
                                   old_gen->used_in_bytes(),
                                   perm_gen->used_in_bytes(),
                                   young_gen->eden_space()->capacity_in_bytes(),
                                   old_gen->max_gen_size(),
                                   max_eden_size,
                                   false  /* full gc*/,
                                   gc_cause);

        }
        // Resize the young generation at every collection
        // even if new sizes have not been calculated.  This is
        // to allow resizes that may have been inhibited by the
        // relative location of the "to" and "from" spaces.

        // Resizing the old gen at minor collects can cause increases
        // that don't feed back to the generation sizing policy until
        // a major collection.  Don't resize the old gen here.

        heap->resize_young_gen(size_policy->calculated_eden_size_in_bytes(),
                        size_policy->calculated_survivor_size_in_bytes());

        if (PrintAdaptiveSizePolicy) {
          gclog_or_tty->print_cr("AdaptiveSizeStop: collection: %d ",
                         heap->total_collections());
        }
      }

      // Update the structure of the eden. With NUMA-eden CPU hotplugging or offlining can
      // cause the change of the heap layout. Make sure eden is reshaped if that's the case.
      // Also update() will case adaptive NUMA chunk resizing.
      assert(young_gen->eden_space()->is_empty(), "eden space should be empty now");
      young_gen->eden_space()->update();

      heap->gc_policy_counters()->update_counters();

      heap->resize_all_tlabs();

      assert(young_gen->to_space()->is_empty(), "to space should be empty now");
    }

    COMPILER2_PRESENT(DerivedPointerTable::update_pointers());

    NOT_PRODUCT(reference_processor()->verify_no_references_recorded());

    // Re-verify object start arrays
    if (VerifyObjectStartArray &&
        VerifyAfterGC) {
      old_gen->verify_object_start_array();
      perm_gen->verify_object_start_array();
    }

    // Verify all old -> young cards are now precise
    if (VerifyRememberedSets) {
      // Precise verification will give false positives. Until this is fixed,
      // use imprecise verification.
      // CardTableExtension::verify_all_young_refs_precise();
      CardTableExtension::verify_all_young_refs_imprecise();
    }

    if (TraceGen0Time) accumulated_time()->stop();

    if (PrintGC) {
      if (PrintGCDetails) {
        // Don't print a GC timestamp here.  This is after the GC so
        // would be confusing.
        young_gen->print_used_change(young_gen_used_before);
      }
      heap->print_heap_change(prev_used);
    }

    // Track memory usage and detect low memory
    MemoryService::track_memory_usage();
    heap->update_counters();
  }

  if (VerifyAfterGC && heap->total_collections() >= VerifyGCStartAt) {
    HandleMark hm;  // Discard invalid handles created during verification
    gclog_or_tty->print(" VerifyAfterGC:");
    Universe::verify(false);
  }

  if (PrintHeapAtGC) {
    Universe::print_heap_after_gc();
  }

  if (ZapUnusedHeapArea) {
    young_gen->eden_space()->check_mangled_unused_area_complete();
    young_gen->from_space()->check_mangled_unused_area_complete();
    young_gen->to_space()->check_mangled_unused_area_complete();
  }

  scavenge_exit.update();

  if (PrintGCTaskTimeStamps) {
    tty->print_cr("VM-Thread " INT64_FORMAT " " INT64_FORMAT " " INT64_FORMAT,
                  scavenge_entry.ticks(), scavenge_midpoint.ticks(),
                  scavenge_exit.ticks());
    gc_task_manager()->print_task_time_stamps();
  }

  return !promotion_failure_occurred;
}

// This method iterates over all objects in the young generation,
// unforwarding markOops. It then restores any preserved mark oops,
// and clears the _preserved_mark_stack.
void PSScavenge::clean_up_failed_promotion() {
  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
  assert(promotion_failed(), "Sanity");

  PSYoungGen* young_gen = heap->young_gen();

  {
    ResourceMark rm;

    // Unforward all pointers in the young gen.
    PSPromotionFailedClosure unforward_closure;
    young_gen->object_iterate(&unforward_closure);

    if (PrintGC && Verbose) {
      gclog_or_tty->print_cr("Restoring %d marks",
                              _preserved_oop_stack->length());
    }

    // Restore any saved marks.
    for (int i=0; i < _preserved_oop_stack->length(); i++) {
      oop obj       = _preserved_oop_stack->at(i);
      markOop mark  = _preserved_mark_stack->at(i);
      obj->set_mark(mark);
    }

    // Deallocate the preserved mark and oop stacks.
    // The stacks were allocated as CHeap objects, so
    // we must call delete to prevent mem leaks.
    delete _preserved_mark_stack;
    _preserved_mark_stack = NULL;
    delete _preserved_oop_stack;
    _preserved_oop_stack = NULL;
  }

  // Reset the PromotionFailureALot counters.
  NOT_PRODUCT(Universe::heap()->reset_promotion_should_fail();)
}

// This method is called whenever an attempt to promote an object
// fails. Some markOops will need preserving, some will not. Note
// that the entire eden is traversed after a failed promotion, with
// all forwarded headers replaced by the default markOop. This means
// it is not neccessary to preserve most markOops.
void PSScavenge::oop_promotion_failed(oop obj, markOop obj_mark) {
  if (_preserved_mark_stack == NULL) {
    ThreadCritical tc; // Lock and retest
    if (_preserved_mark_stack == NULL) {
      assert(_preserved_oop_stack == NULL, "Sanity");
      _preserved_mark_stack = new (ResourceObj::C_HEAP) GrowableArray<markOop>(40, true);
      _preserved_oop_stack = new (ResourceObj::C_HEAP) GrowableArray<oop>(40, true);
    }
  }

  // Because we must hold the ThreadCritical lock before using
  // the stacks, we should be safe from observing partial allocations,
  // which are also guarded by the ThreadCritical lock.
  if (obj_mark->must_be_preserved_for_promotion_failure(obj)) {
    ThreadCritical tc;
    _preserved_oop_stack->push(obj);
    _preserved_mark_stack->push(obj_mark);
  }
}

bool PSScavenge::should_attempt_scavenge() {
  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
  PSGCAdaptivePolicyCounters* counters = heap->gc_policy_counters();

  if (UsePerfData) {
    counters->update_scavenge_skipped(not_skipped);
  }

  PSYoungGen* young_gen = heap->young_gen();
  PSOldGen* old_gen = heap->old_gen();

  if (!ScavengeWithObjectsInToSpace) {
    // Do not attempt to promote unless to_space is empty
    if (!young_gen->to_space()->is_empty()) {
      _consecutive_skipped_scavenges++;
      if (UsePerfData) {
        counters->update_scavenge_skipped(to_space_not_empty);
      }
      return false;
    }
  }

  // Test to see if the scavenge will likely fail.
  PSAdaptiveSizePolicy* policy = heap->size_policy();

  // A similar test is done in the policy's should_full_GC().  If this is
  // changed, decide if that test should also be changed.
  size_t avg_promoted = (size_t) policy->padded_average_promoted_in_bytes();
  size_t promotion_estimate = MIN2(avg_promoted, young_gen->used_in_bytes());
  bool result = promotion_estimate < old_gen->free_in_bytes();

  if (PrintGCDetails && Verbose) {
    gclog_or_tty->print(result ? "  do scavenge: " : "  skip scavenge: ");
    gclog_or_tty->print_cr(" average_promoted " SIZE_FORMAT
      " padded_average_promoted " SIZE_FORMAT
      " free in old gen " SIZE_FORMAT,
      (size_t) policy->average_promoted_in_bytes(),
      (size_t) policy->padded_average_promoted_in_bytes(),
      old_gen->free_in_bytes());
    if (young_gen->used_in_bytes() <
        (size_t) policy->padded_average_promoted_in_bytes()) {
      gclog_or_tty->print_cr(" padded_promoted_average is greater"
        " than maximum promotion = " SIZE_FORMAT, young_gen->used_in_bytes());
    }
  }

  if (result) {
    _consecutive_skipped_scavenges = 0;
  } else {
    _consecutive_skipped_scavenges++;
    if (UsePerfData) {
      counters->update_scavenge_skipped(promoted_too_large);
    }
  }
  return result;
}

  // Used to add tasks
GCTaskManager* const PSScavenge::gc_task_manager() {
  assert(ParallelScavengeHeap::gc_task_manager() != NULL,
   "shouldn't return NULL");
  return ParallelScavengeHeap::gc_task_manager();
}

void PSScavenge::initialize() {
  // Arguments must have been parsed

  if (AlwaysTenure) {
    _tenuring_threshold = 0;
  } else if (NeverTenure) {
    _tenuring_threshold = markOopDesc::max_age + 1;
  } else {
    // We want to smooth out our startup times for the AdaptiveSizePolicy
    _tenuring_threshold = (UseAdaptiveSizePolicy) ? InitialTenuringThreshold :
                                                    MaxTenuringThreshold;
  }

  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");

  PSYoungGen* young_gen = heap->young_gen();
  PSOldGen* old_gen = heap->old_gen();
  PSPermGen* perm_gen = heap->perm_gen();

  // Set boundary between young_gen and old_gen
  assert(perm_gen->reserved().end() <= old_gen->object_space()->bottom(),
         "perm above old");
  assert(old_gen->reserved().end() <= young_gen->eden_space()->bottom(),
         "old above young");
  _young_generation_boundary = young_gen->eden_space()->bottom();

  // Initialize ref handling object for scavenging.
  MemRegion mr = young_gen->reserved();
  _ref_processor = ReferenceProcessor::create_ref_processor(
    mr,                         // span
    true,                       // atomic_discovery
    true,                       // mt_discovery
    NULL,                       // is_alive_non_header
    ParallelGCThreads,
    ParallelRefProcEnabled);

  // Cache the cardtable
  BarrierSet* bs = Universe::heap()->barrier_set();
  assert(bs->kind() == BarrierSet::CardTableModRef, "Wrong barrier set kind");
  _card_table = (CardTableExtension*)bs;

  _counters = new CollectorCounters("PSScavenge", 0);
}