Merge
authorjwilhelm
Tue, 23 Dec 2014 14:03:15 +0100
changeset 28218 84ded0ce573d
parent 28217 57791914628c (diff)
parent 28204 344d97fad05b (current diff)
child 28221 000bc11b1078
Merge
hotspot/test/testlibrary/whitebox/Makefile
hotspot/test/testlibrary/whitebox/sun/hotspot/WhiteBox.java
hotspot/test/testlibrary/whitebox/sun/hotspot/code/BlobType.java
hotspot/test/testlibrary/whitebox/sun/hotspot/code/CodeBlob.java
hotspot/test/testlibrary/whitebox/sun/hotspot/code/NMethod.java
hotspot/test/testlibrary/whitebox/sun/hotspot/cpuinfo/CPUInfo.java
hotspot/test/testlibrary/whitebox/sun/hotspot/parser/DiagnosticCommand.java
--- a/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -2560,12 +2560,12 @@
      x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x,   \
      x }
 
-// Initialize with default setting of CMSParPromoteBlocksToClaim, _not_
-// OldPLABSize, whose static default is different; if overridden at the
+// Initialize with default setting for CMS, _not_
+// generic OldPLABSize, whose static default is different; if overridden at the
 // command-line, this will get reinitialized via a call to
 // modify_initialization() below.
 AdaptiveWeightedAverage CFLS_LAB::_blocks_to_claim[]    =
-  VECTOR_257(AdaptiveWeightedAverage(OldPLABWeight, (float)CMSParPromoteBlocksToClaim));
+  VECTOR_257(AdaptiveWeightedAverage(OldPLABWeight, (float)CFLS_LAB::_default_dynamic_old_plab_size));
 size_t CFLS_LAB::_global_num_blocks[]  = VECTOR_257(0);
 uint   CFLS_LAB::_global_num_workers[] = VECTOR_257(0);
 
--- a/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -690,6 +690,9 @@
   void get_from_global_pool(size_t word_sz, AdaptiveFreeList<FreeChunk>* fl);
 
 public:
+  static const int _default_dynamic_old_plab_size = 16;
+  static const int _default_static_old_plab_size  = 50;
+
   CFLS_LAB(CompactibleFreeListSpace* cfls);
 
   // Allocate and return a block of the given size, or else return NULL.
--- a/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -1610,14 +1610,15 @@
 
   // If the collection is being acquired from the background
   // collector, there may be references on the discovered
-  // references lists that have NULL referents (being those
-  // that were concurrently cleared by a mutator) or
-  // that are no longer active (having been enqueued concurrently
-  // by the mutator).
-  // Scrub the list of those references because Mark-Sweep-Compact
-  // code assumes referents are not NULL and that all discovered
-  // Reference objects are active.
-  ref_processor()->clean_up_discovered_references();
+  // references lists.  Abandon those references, since some
+  // of them may have become unreachable after concurrent
+  // discovery; the STW compacting collector will redo discovery
+  // more precisely, without being subject to floating garbage.
+  // Leaving otherwise unreachable references in the discovered
+  // lists would require special handling.
+  ref_processor()->disable_discovery();
+  ref_processor()->abandon_partial_discovery();
+  ref_processor()->verify_no_references_recorded();
 
   if (first_state > Idling) {
     save_heap_summary();
@@ -1683,7 +1684,7 @@
   ReferenceProcessorMTDiscoveryMutator rp_mut_discovery(ref_processor(), false);
 
   ref_processor()->set_enqueuing_is_done(false);
-  ref_processor()->enable_discovery(false /*verify_disabled*/, false /*check_no_refs*/);
+  ref_processor()->enable_discovery();
   ref_processor()->setup_policy(clear_all_soft_refs);
   // If an asynchronous collection finishes, the _modUnionTable is
   // all clear.  If we are assuming the collection from an asynchronous
@@ -3000,7 +3001,7 @@
                     Mutex::_no_safepoint_check_flag);
     checkpointRootsInitialWork();
     // enable ("weak") refs discovery
-    rp->enable_discovery(true /*verify_disabled*/, true /*check_no_refs*/);
+    rp->enable_discovery();
     _collectorState = Marking;
   }
   SpecializationStats::print();
--- a/hotspot/src/share/vm/gc_implementation/g1/concurrentMark.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/concurrentMark.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -971,7 +971,7 @@
   // Start Concurrent Marking weak-reference discovery.
   ReferenceProcessor* rp = g1h->ref_processor_cm();
   // enable ("weak") refs discovery
-  rp->enable_discovery(true /*verify_disabled*/, true /*verify_no_refs*/);
+  rp->enable_discovery();
   rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle
 
   SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
--- a/hotspot/src/share/vm/gc_implementation/g1/g1AllocRegion.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1AllocRegion.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -254,25 +254,23 @@
 HeapRegion* SurvivorGCAllocRegion::allocate_new_region(size_t word_size,
                                                        bool force) {
   assert(!force, "not supported for GC alloc regions");
-  return _g1h->new_gc_alloc_region(word_size, count(), GCAllocForSurvived);
+  return _g1h->new_gc_alloc_region(word_size, count(), InCSetState::Young);
 }
 
 void SurvivorGCAllocRegion::retire_region(HeapRegion* alloc_region,
                                           size_t allocated_bytes) {
-  _g1h->retire_gc_alloc_region(alloc_region, allocated_bytes,
-                               GCAllocForSurvived);
+  _g1h->retire_gc_alloc_region(alloc_region, allocated_bytes, InCSetState::Young);
 }
 
 HeapRegion* OldGCAllocRegion::allocate_new_region(size_t word_size,
                                                   bool force) {
   assert(!force, "not supported for GC alloc regions");
-  return _g1h->new_gc_alloc_region(word_size, count(), GCAllocForTenured);
+  return _g1h->new_gc_alloc_region(word_size, count(), InCSetState::Old);
 }
 
 void OldGCAllocRegion::retire_region(HeapRegion* alloc_region,
                                      size_t allocated_bytes) {
-  _g1h->retire_gc_alloc_region(alloc_region, allocated_bytes,
-                               GCAllocForTenured);
+  _g1h->retire_gc_alloc_region(alloc_region, allocated_bytes, InCSetState::Old);
 }
 
 HeapRegion* OldGCAllocRegion::release() {
--- a/hotspot/src/share/vm/gc_implementation/g1/g1Allocator.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1Allocator.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -113,15 +113,16 @@
 G1ParGCAllocBuffer::G1ParGCAllocBuffer(size_t gclab_word_size) :
   ParGCAllocBuffer(gclab_word_size), _retired(true) { }
 
-HeapWord* G1ParGCAllocator::allocate_slow(GCAllocPurpose purpose, size_t word_sz, AllocationContext_t context) {
-  HeapWord* obj = NULL;
-  size_t gclab_word_size = _g1h->desired_plab_sz(purpose);
+HeapWord* G1ParGCAllocator::allocate_direct_or_new_plab(InCSetState dest,
+                                                        size_t word_sz,
+                                                        AllocationContext_t context) {
+  size_t gclab_word_size = _g1h->desired_plab_sz(dest);
   if (word_sz * 100 < gclab_word_size * ParallelGCBufferWastePct) {
-    G1ParGCAllocBuffer* alloc_buf = alloc_buffer(purpose, context);
+    G1ParGCAllocBuffer* alloc_buf = alloc_buffer(dest, context);
     add_to_alloc_buffer_waste(alloc_buf->words_remaining());
     alloc_buf->retire(false /* end_of_gc */, false /* retain */);
 
-    HeapWord* buf = _g1h->par_allocate_during_gc(purpose, gclab_word_size, context);
+    HeapWord* buf = _g1h->par_allocate_during_gc(dest, gclab_word_size, context);
     if (buf == NULL) {
       return NULL; // Let caller handle allocation failure.
     }
@@ -129,30 +130,33 @@
     alloc_buf->set_word_size(gclab_word_size);
     alloc_buf->set_buf(buf);
 
-    obj = alloc_buf->allocate(word_sz);
+    HeapWord* const obj = alloc_buf->allocate(word_sz);
     assert(obj != NULL, "buffer was definitely big enough...");
+    return obj;
   } else {
-    obj = _g1h->par_allocate_during_gc(purpose, word_sz, context);
+    return _g1h->par_allocate_during_gc(dest, word_sz, context);
   }
-  return obj;
 }
 
 G1DefaultParGCAllocator::G1DefaultParGCAllocator(G1CollectedHeap* g1h) :
-            G1ParGCAllocator(g1h),
-            _surviving_alloc_buffer(g1h->desired_plab_sz(GCAllocForSurvived)),
-            _tenured_alloc_buffer(g1h->desired_plab_sz(GCAllocForTenured)) {
-
-  _alloc_buffers[GCAllocForSurvived] = &_surviving_alloc_buffer;
-  _alloc_buffers[GCAllocForTenured]  = &_tenured_alloc_buffer;
-
+  G1ParGCAllocator(g1h),
+  _surviving_alloc_buffer(g1h->desired_plab_sz(InCSetState::Young)),
+  _tenured_alloc_buffer(g1h->desired_plab_sz(InCSetState::Old)) {
+  for (uint state = 0; state < InCSetState::Num; state++) {
+    _alloc_buffers[state] = NULL;
+  }
+  _alloc_buffers[InCSetState::Young] = &_surviving_alloc_buffer;
+  _alloc_buffers[InCSetState::Old]  = &_tenured_alloc_buffer;
 }
 
 void G1DefaultParGCAllocator::retire_alloc_buffers() {
-  for (int ap = 0; ap < GCAllocPurposeCount; ++ap) {
-    size_t waste = _alloc_buffers[ap]->words_remaining();
-    add_to_alloc_buffer_waste(waste);
-    _alloc_buffers[ap]->flush_stats_and_retire(_g1h->stats_for_purpose((GCAllocPurpose)ap),
-                                               true /* end_of_gc */,
-                                               false /* retain */);
+  for (uint state = 0; state < InCSetState::Num; state++) {
+    G1ParGCAllocBuffer* const buf = _alloc_buffers[state];
+    if (buf != NULL) {
+      add_to_alloc_buffer_waste(buf->words_remaining());
+      buf->flush_stats_and_retire(_g1h->alloc_buffer_stats(state),
+                                  true /* end_of_gc */,
+                                  false /* retain */);
+    }
   }
 }
--- a/hotspot/src/share/vm/gc_implementation/g1/g1Allocator.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1Allocator.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -27,14 +27,9 @@
 
 #include "gc_implementation/g1/g1AllocationContext.hpp"
 #include "gc_implementation/g1/g1AllocRegion.hpp"
+#include "gc_implementation/g1/g1InCSetState.hpp"
 #include "gc_implementation/shared/parGCAllocBuffer.hpp"
 
-enum GCAllocPurpose {
-  GCAllocForTenured,
-  GCAllocForSurvived,
-  GCAllocPurposeCount
-};
-
 // Base class for G1 allocators.
 class G1Allocator : public CHeapObj<mtGC> {
   friend class VMStructs;
@@ -178,20 +173,40 @@
 protected:
   G1CollectedHeap* _g1h;
 
+  // The survivor alignment in effect in bytes.
+  // == 0 : don't align survivors
+  // != 0 : align survivors to that alignment
+  // These values were chosen to favor the non-alignment case since some
+  // architectures have a special compare against zero instructions.
+  const uint _survivor_alignment_bytes;
+
   size_t _alloc_buffer_waste;
   size_t _undo_waste;
 
   void add_to_alloc_buffer_waste(size_t waste) { _alloc_buffer_waste += waste; }
   void add_to_undo_waste(size_t waste)         { _undo_waste += waste; }
 
-  HeapWord* allocate_slow(GCAllocPurpose purpose, size_t word_sz, AllocationContext_t context);
+  virtual void retire_alloc_buffers() = 0;
+  virtual G1ParGCAllocBuffer* alloc_buffer(InCSetState dest, AllocationContext_t context) = 0;
 
-  virtual void retire_alloc_buffers() = 0;
-  virtual G1ParGCAllocBuffer* alloc_buffer(GCAllocPurpose purpose, AllocationContext_t context) = 0;
+  // Calculate the survivor space object alignment in bytes. Returns that or 0 if
+  // there are no restrictions on survivor alignment.
+  static uint calc_survivor_alignment_bytes() {
+    assert(SurvivorAlignmentInBytes >= ObjectAlignmentInBytes, "sanity");
+    if (SurvivorAlignmentInBytes == ObjectAlignmentInBytes) {
+      // No need to align objects in the survivors differently, return 0
+      // which means "survivor alignment is not used".
+      return 0;
+    } else {
+      assert(SurvivorAlignmentInBytes > 0, "sanity");
+      return SurvivorAlignmentInBytes;
+    }
+  }
 
 public:
   G1ParGCAllocator(G1CollectedHeap* g1h) :
-    _g1h(g1h), _alloc_buffer_waste(0), _undo_waste(0) {
+    _g1h(g1h), _survivor_alignment_bytes(calc_survivor_alignment_bytes()),
+    _alloc_buffer_waste(0), _undo_waste(0) {
   }
 
   static G1ParGCAllocator* create_allocator(G1CollectedHeap* g1h);
@@ -199,24 +214,40 @@
   size_t alloc_buffer_waste() { return _alloc_buffer_waste; }
   size_t undo_waste() {return _undo_waste; }
 
-  HeapWord* allocate(GCAllocPurpose purpose, size_t word_sz, AllocationContext_t context) {
-    HeapWord* obj = NULL;
-    if (purpose == GCAllocForSurvived) {
-      obj = alloc_buffer(purpose, context)->allocate_aligned(word_sz, SurvivorAlignmentInBytes);
+  // Allocate word_sz words in dest, either directly into the regions or by
+  // allocating a new PLAB. Returns the address of the allocated memory, NULL if
+  // not successful.
+  HeapWord* allocate_direct_or_new_plab(InCSetState dest,
+                                        size_t word_sz,
+                                        AllocationContext_t context);
+
+  // Allocate word_sz words in the PLAB of dest.  Returns the address of the
+  // allocated memory, NULL if not successful.
+  HeapWord* plab_allocate(InCSetState dest,
+                          size_t word_sz,
+                          AllocationContext_t context) {
+    G1ParGCAllocBuffer* buffer = alloc_buffer(dest, context);
+    if (_survivor_alignment_bytes == 0) {
+      return buffer->allocate(word_sz);
     } else {
-      obj = alloc_buffer(purpose, context)->allocate(word_sz);
+      return buffer->allocate_aligned(word_sz, _survivor_alignment_bytes);
     }
+  }
+
+  HeapWord* allocate(InCSetState dest, size_t word_sz,
+                     AllocationContext_t context) {
+    HeapWord* const obj = plab_allocate(dest, word_sz, context);
     if (obj != NULL) {
       return obj;
     }
-    return allocate_slow(purpose, word_sz, context);
+    return allocate_direct_or_new_plab(dest, word_sz, context);
   }
 
-  void undo_allocation(GCAllocPurpose purpose, HeapWord* obj, size_t word_sz, AllocationContext_t context) {
-    if (alloc_buffer(purpose, context)->contains(obj)) {
-      assert(alloc_buffer(purpose, context)->contains(obj + word_sz - 1),
+  void undo_allocation(InCSetState dest, HeapWord* obj, size_t word_sz, AllocationContext_t context) {
+    if (alloc_buffer(dest, context)->contains(obj)) {
+      assert(alloc_buffer(dest, context)->contains(obj + word_sz - 1),
              "should contain whole object");
-      alloc_buffer(purpose, context)->undo_allocation(obj, word_sz);
+      alloc_buffer(dest, context)->undo_allocation(obj, word_sz);
     } else {
       CollectedHeap::fill_with_object(obj, word_sz);
       add_to_undo_waste(word_sz);
@@ -227,13 +258,17 @@
 class G1DefaultParGCAllocator : public G1ParGCAllocator {
   G1ParGCAllocBuffer  _surviving_alloc_buffer;
   G1ParGCAllocBuffer  _tenured_alloc_buffer;
-  G1ParGCAllocBuffer* _alloc_buffers[GCAllocPurposeCount];
+  G1ParGCAllocBuffer* _alloc_buffers[InCSetState::Num];
 
 public:
   G1DefaultParGCAllocator(G1CollectedHeap* g1h);
 
-  virtual G1ParGCAllocBuffer* alloc_buffer(GCAllocPurpose purpose, AllocationContext_t context) {
-    return _alloc_buffers[purpose];
+  virtual G1ParGCAllocBuffer* alloc_buffer(InCSetState dest, AllocationContext_t context) {
+    assert(dest.is_valid(),
+           err_msg("Allocation buffer index out-of-bounds: " CSETSTATE_FORMAT, dest.value()));
+    assert(_alloc_buffers[dest.value()] != NULL,
+           err_msg("Allocation buffer is NULL: " CSETSTATE_FORMAT, dest.value()));
+    return _alloc_buffers[dest.value()];
   }
 
   virtual void retire_alloc_buffers() ;
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -1301,7 +1301,7 @@
       // Temporarily clear the STW ref processor's _is_alive_non_header field.
       ReferenceProcessorIsAliveMutator stw_rp_is_alive_null(ref_processor_stw(), NULL);
 
-      ref_processor_stw()->enable_discovery(true /*verify_disabled*/, true /*verify_no_refs*/);
+      ref_processor_stw()->enable_discovery();
       ref_processor_stw()->setup_policy(do_clear_all_soft_refs);
 
       // Do collection work
@@ -1886,13 +1886,12 @@
 
   initialize_reserved_region((HeapWord*)heap_rs.base(), (HeapWord*)(heap_rs.base() + heap_rs.size()));
 
-  // Create the gen rem set (and barrier set) for the entire reserved region.
-  _rem_set = collector_policy()->create_rem_set(reserved_region());
-  set_barrier_set(rem_set()->bs());
-  if (!barrier_set()->is_a(BarrierSet::G1SATBCTLogging)) {
-    vm_exit_during_initialization("G1 requires a G1SATBLoggingCardTableModRefBS");
-    return JNI_ENOMEM;
-  }
+  // Create the barrier set for the entire reserved region.
+  G1SATBCardTableLoggingModRefBS* bs
+    = new G1SATBCardTableLoggingModRefBS(reserved_region());
+  bs->initialize();
+  assert(bs->is_a(BarrierSet::G1SATBCTLogging), "sanity");
+  set_barrier_set(bs);
 
   // Also create a G1 rem set.
   _g1_rem_set = new G1RemSet(this, g1_barrier_set());
@@ -3153,8 +3152,6 @@
         failures = true;
       }
     }
-    if (!silent) gclog_or_tty->print("RemSet ");
-    rem_set()->verify();
 
     if (G1StringDedup::is_enabled()) {
       if (!silent) gclog_or_tty->print("StrDedup ");
@@ -3750,8 +3747,7 @@
       // reference processing currently works in G1.
 
       // Enable discovery in the STW reference processor
-      ref_processor_stw()->enable_discovery(true /*verify_disabled*/,
-                                            true /*verify_no_refs*/);
+      ref_processor_stw()->enable_discovery();
 
       {
         // We want to temporarily turn off discovery by the
@@ -3819,6 +3815,8 @@
 
         register_humongous_regions_with_in_cset_fast_test();
 
+        assert(check_cset_fast_test(), "Inconsistency in the InCSetState table.");
+
         _cm->note_start_of_gc();
         // We should not verify the per-thread SATB buffers given that
         // we have not filtered them yet (we'll do so during the
@@ -4048,29 +4046,6 @@
   return true;
 }
 
-size_t G1CollectedHeap::desired_plab_sz(GCAllocPurpose purpose)
-{
-  size_t gclab_word_size;
-  switch (purpose) {
-    case GCAllocForSurvived:
-      gclab_word_size = _survivor_plab_stats.desired_plab_sz();
-      break;
-    case GCAllocForTenured:
-      gclab_word_size = _old_plab_stats.desired_plab_sz();
-      break;
-    default:
-      assert(false, "unknown GCAllocPurpose");
-      gclab_word_size = _old_plab_stats.desired_plab_sz();
-      break;
-  }
-
-  // Prevent humongous PLAB sizes for two reasons:
-  // * PLABs are allocated using a similar paths as oops, but should
-  //   never be in a humongous region
-  // * Allowing humongous PLABs needlessly churns the region free lists
-  return MIN2(_humongous_object_threshold_in_words, gclab_word_size);
-}
-
 void G1CollectedHeap::init_for_evac_failure(OopsInHeapRegionClosure* cl) {
   _drain_in_progress = false;
   set_evac_failure_closure(cl);
@@ -4196,35 +4171,6 @@
   }
 }
 
-HeapWord* G1CollectedHeap::par_allocate_during_gc(GCAllocPurpose purpose,
-                                                  size_t word_size,
-                                                  AllocationContext_t context) {
-  if (purpose == GCAllocForSurvived) {
-    HeapWord* result = survivor_attempt_allocation(word_size, context);
-    if (result != NULL) {
-      return result;
-    } else {
-      // Let's try to allocate in the old gen in case we can fit the
-      // object there.
-      return old_attempt_allocation(word_size, context);
-    }
-  } else {
-    assert(purpose ==  GCAllocForTenured, "sanity");
-    HeapWord* result = old_attempt_allocation(word_size, context);
-    if (result != NULL) {
-      return result;
-    } else {
-      // Let's try to allocate in the survivors in case we can fit the
-      // object there.
-      return survivor_attempt_allocation(word_size, context);
-    }
-  }
-
-  ShouldNotReachHere();
-  // Trying to keep some compilers happy.
-  return NULL;
-}
-
 void G1ParCopyHelper::mark_object(oop obj) {
   assert(!_g1->heap_region_containing(obj)->in_collection_set(), "should not mark objects in the CSet");
 
@@ -4267,15 +4213,14 @@
 
   assert(_worker_id == _par_scan_state->queue_num(), "sanity");
 
-  G1CollectedHeap::in_cset_state_t state = _g1->in_cset_state(obj);
-
-  if (state == G1CollectedHeap::InCSet) {
+  const InCSetState state = _g1->in_cset_state(obj);
+  if (state.is_in_cset()) {
     oop forwardee;
     markOop m = obj->mark();
     if (m->is_marked()) {
       forwardee = (oop) m->decode_pointer();
     } else {
-      forwardee = _par_scan_state->copy_to_survivor_space(obj, m);
+      forwardee = _par_scan_state->copy_to_survivor_space(state, obj, m);
     }
     assert(forwardee != NULL, "forwardee should not be NULL");
     oopDesc::encode_store_heap_oop(p, forwardee);
@@ -4289,7 +4234,7 @@
       do_klass_barrier(p, forwardee);
     }
   } else {
-    if (state == G1CollectedHeap::IsHumongous) {
+    if (state.is_humongous()) {
       _g1->set_humongous_is_live(obj);
     }
     // The object is not in collection set. If we're a root scanning
@@ -4609,7 +4554,7 @@
 G1CollectedHeap::
 g1_process_roots(OopClosure* scan_non_heap_roots,
                  OopClosure* scan_non_heap_weak_roots,
-                 OopsInHeapRegionClosure* scan_rs,
+                 G1ParPushHeapRSClosure* scan_rs,
                  CLDClosure* scan_strong_clds,
                  CLDClosure* scan_weak_clds,
                  CodeBlobClosure* scan_strong_code,
@@ -5145,17 +5090,17 @@
     oop obj = *p;
     assert(obj != NULL, "the caller should have filtered out NULL values");
 
-    G1CollectedHeap::in_cset_state_t cset_state = _g1->in_cset_state(obj);
-    if (cset_state == G1CollectedHeap::InNeither) {
+    const InCSetState cset_state = _g1->in_cset_state(obj);
+    if (!cset_state.is_in_cset_or_humongous()) {
       return;
     }
-    if (cset_state == G1CollectedHeap::InCSet) {
+    if (cset_state.is_in_cset()) {
       assert( obj->is_forwarded(), "invariant" );
       *p = obj->forwardee();
     } else {
       assert(!obj->is_forwarded(), "invariant" );
-      assert(cset_state == G1CollectedHeap::IsHumongous,
-             err_msg("Only allowed InCSet state is IsHumongous, but is %d", cset_state));
+      assert(cset_state.is_humongous(),
+             err_msg("Only allowed InCSet state is IsHumongous, but is %d", cset_state.value()));
       _g1->set_humongous_is_live(obj);
     }
   }
@@ -5640,8 +5585,6 @@
 
   init_for_evac_failure(NULL);
 
-  rem_set()->prepare_for_younger_refs_iterate(true);
-
   assert(dirty_card_queue_set().completed_buffers_num() == 0, "Should be empty");
   double start_par_time_sec = os::elapsedTime();
   double end_par_time_sec;
@@ -5951,6 +5894,70 @@
   heap_region_iterate(&cl);
   guarantee(!cl.failures(), "bitmap verification");
 }
+
+class G1CheckCSetFastTableClosure : public HeapRegionClosure {
+ private:
+  bool _failures;
+ public:
+  G1CheckCSetFastTableClosure() : HeapRegionClosure(), _failures(false) { }
+
+  virtual bool doHeapRegion(HeapRegion* hr) {
+    uint i = hr->hrm_index();
+    InCSetState cset_state = (InCSetState) G1CollectedHeap::heap()->_in_cset_fast_test.get_by_index(i);
+    if (hr->is_humongous()) {
+      if (hr->in_collection_set()) {
+        gclog_or_tty->print_cr("\n## humongous region %u in CSet", i);
+        _failures = true;
+        return true;
+      }
+      if (cset_state.is_in_cset()) {
+        gclog_or_tty->print_cr("\n## inconsistent cset state %d for humongous region %u", cset_state.value(), i);
+        _failures = true;
+        return true;
+      }
+      if (hr->is_continues_humongous() && cset_state.is_humongous()) {
+        gclog_or_tty->print_cr("\n## inconsistent cset state %d for continues humongous region %u", cset_state.value(), i);
+        _failures = true;
+        return true;
+      }
+    } else {
+      if (cset_state.is_humongous()) {
+        gclog_or_tty->print_cr("\n## inconsistent cset state %d for non-humongous region %u", cset_state.value(), i);
+        _failures = true;
+        return true;
+      }
+      if (hr->in_collection_set() != cset_state.is_in_cset()) {
+        gclog_or_tty->print_cr("\n## in CSet %d / cset state %d inconsistency for region %u",
+                               hr->in_collection_set(), cset_state.value(), i);
+        _failures = true;
+        return true;
+      }
+      if (cset_state.is_in_cset()) {
+        if (hr->is_young() != (cset_state.is_young())) {
+          gclog_or_tty->print_cr("\n## is_young %d / cset state %d inconsistency for region %u",
+                                 hr->is_young(), cset_state.value(), i);
+          _failures = true;
+          return true;
+        }
+        if (hr->is_old() != (cset_state.is_old())) {
+          gclog_or_tty->print_cr("\n## is_old %d / cset state %d inconsistency for region %u",
+                                 hr->is_old(), cset_state.value(), i);
+          _failures = true;
+          return true;
+        }
+      }
+    }
+    return false;
+  }
+
+  bool failures() const { return _failures; }
+};
+
+bool G1CollectedHeap::check_cset_fast_test() {
+  G1CheckCSetFastTableClosure cl;
+  _hrm.iterate(&cl);
+  return !cl.failures();
+}
 #endif // PRODUCT
 
 void G1CollectedHeap::cleanUpCardTable() {
@@ -6519,20 +6526,20 @@
 
 HeapRegion* G1CollectedHeap::new_gc_alloc_region(size_t word_size,
                                                  uint count,
-                                                 GCAllocPurpose ap) {
+                                                 InCSetState dest) {
   assert(FreeList_lock->owned_by_self(), "pre-condition");
 
-  if (count < g1_policy()->max_regions(ap)) {
-    bool survivor = (ap == GCAllocForSurvived);
+  if (count < g1_policy()->max_regions(dest)) {
+    const bool is_survivor = (dest.is_young());
     HeapRegion* new_alloc_region = new_region(word_size,
-                                              !survivor,
+                                              !is_survivor,
                                               true /* do_expand */);
     if (new_alloc_region != NULL) {
       // We really only need to do this for old regions given that we
       // should never scan survivors. But it doesn't hurt to do it
       // for survivors too.
       new_alloc_region->record_timestamp();
-      if (survivor) {
+      if (is_survivor) {
         new_alloc_region->set_survivor();
         _hr_printer.alloc(new_alloc_region, G1HRPrinter::Survivor);
         check_bitmaps("Survivor Region Allocation", new_alloc_region);
@@ -6544,8 +6551,6 @@
       bool during_im = g1_policy()->during_initial_mark_pause();
       new_alloc_region->note_start_of_copying(during_im);
       return new_alloc_region;
-    } else {
-      g1_policy()->note_alloc_region_limit_reached(ap);
     }
   }
   return NULL;
@@ -6553,11 +6558,11 @@
 
 void G1CollectedHeap::retire_gc_alloc_region(HeapRegion* alloc_region,
                                              size_t allocated_bytes,
-                                             GCAllocPurpose ap) {
+                                             InCSetState dest) {
   bool during_im = g1_policy()->during_initial_mark_pause();
   alloc_region->note_end_of_copying(during_im);
   g1_policy()->record_bytes_copied_during_gc(allocated_bytes);
-  if (ap == GCAllocForSurvived) {
+  if (dest.is_young()) {
     young_list()->add_survivor_region(alloc_region);
   } else {
     _old_set.add(alloc_region);
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -32,6 +32,7 @@
 #include "gc_implementation/g1/g1AllocRegion.hpp"
 #include "gc_implementation/g1/g1BiasedArray.hpp"
 #include "gc_implementation/g1/g1HRPrinter.hpp"
+#include "gc_implementation/g1/g1InCSetState.hpp"
 #include "gc_implementation/g1/g1MonitoringSupport.hpp"
 #include "gc_implementation/g1/g1SATBCardTableModRefBS.hpp"
 #include "gc_implementation/g1/g1YCTypes.hpp"
@@ -213,6 +214,9 @@
   friend class G1MarkSweep;
   friend class HeapRegionClaimer;
 
+  // Testing classes.
+  friend class G1CheckCSetFastTableClosure;
+
 private:
   // The one and only G1CollectedHeap, so static functions can find it.
   static G1CollectedHeap* _g1h;
@@ -547,15 +551,9 @@
   // allocation region, either by picking one or expanding the
   // heap, and then allocate a block of the given size. The block
   // may not be a humongous - it must fit into a single heap region.
-  HeapWord* par_allocate_during_gc(GCAllocPurpose purpose,
-                                   size_t word_size,
-                                   AllocationContext_t context);
-
-  HeapWord* allocate_during_gc_slow(GCAllocPurpose purpose,
-                                    HeapRegion*    alloc_region,
-                                    bool           par,
-                                    size_t         word_size);
-
+  inline HeapWord* par_allocate_during_gc(InCSetState dest,
+                                          size_t word_size,
+                                          AllocationContext_t context);
   // Ensure that no further allocations can happen in "r", bearing in mind
   // that parallel threads might be attempting allocations.
   void par_allocate_remaining_space(HeapRegion* r);
@@ -577,9 +575,9 @@
 
   // For GC alloc regions.
   HeapRegion* new_gc_alloc_region(size_t word_size, uint count,
-                                  GCAllocPurpose ap);
+                                  InCSetState dest);
   void retire_gc_alloc_region(HeapRegion* alloc_region,
-                              size_t allocated_bytes, GCAllocPurpose ap);
+                              size_t allocated_bytes, InCSetState dest);
 
   // - if explicit_gc is true, the GC is for a System.gc() or a heap
   //   inspection request and should collect the entire heap
@@ -640,26 +638,11 @@
   // (Rounds up to a HeapRegion boundary.)
   bool expand(size_t expand_bytes);
 
-  // Returns the PLAB statistics given a purpose.
-  PLABStats* stats_for_purpose(GCAllocPurpose purpose) {
-    PLABStats* stats = NULL;
+  // Returns the PLAB statistics for a given destination.
+  inline PLABStats* alloc_buffer_stats(InCSetState dest);
 
-    switch (purpose) {
-    case GCAllocForSurvived:
-      stats = &_survivor_plab_stats;
-      break;
-    case GCAllocForTenured:
-      stats = &_old_plab_stats;
-      break;
-    default:
-      assert(false, "unrecognized GCAllocPurpose");
-    }
-
-    return stats;
-  }
-
-  // Determines PLAB size for a particular allocation purpose.
-  size_t desired_plab_sz(GCAllocPurpose purpose);
+  // Determines PLAB size for a given destination.
+  inline size_t desired_plab_sz(InCSetState dest);
 
   inline AllocationContextStats& allocation_context_stats();
 
@@ -683,8 +666,11 @@
   void register_humongous_regions_with_in_cset_fast_test();
   // We register a region with the fast "in collection set" test. We
   // simply set to true the array slot corresponding to this region.
-  void register_region_with_in_cset_fast_test(HeapRegion* r) {
-    _in_cset_fast_test.set_in_cset(r->hrm_index());
+  void register_young_region_with_in_cset_fast_test(HeapRegion* r) {
+    _in_cset_fast_test.set_in_young(r->hrm_index());
+  }
+  void register_old_region_with_in_cset_fast_test(HeapRegion* r) {
+    _in_cset_fast_test.set_in_old(r->hrm_index());
   }
 
   // This is a fast test on whether a reference points into the
@@ -821,7 +807,7 @@
   // In the sequential case this param will be ignored.
   void g1_process_roots(OopClosure* scan_non_heap_roots,
                         OopClosure* scan_non_heap_weak_roots,
-                        OopsInHeapRegionClosure* scan_rs,
+                        G1ParPushHeapRSClosure* scan_rs,
                         CLDClosure* scan_strong_clds,
                         CLDClosure* scan_weak_clds,
                         CodeBlobClosure* scan_strong_code,
@@ -1181,6 +1167,9 @@
   // appropriate error messages and crash.
   void check_bitmaps(const char* caller) PRODUCT_RETURN;
 
+  // Do sanity check on the contents of the in-cset fast test table.
+  bool check_cset_fast_test() PRODUCT_RETURN_( return true; );
+
   // verify_region_sets() performs verification over the region
   // lists. It will be compiled in the product code to be used when
   // necessary (i.e., during heap verification).
@@ -1276,53 +1265,15 @@
 
   inline bool is_in_cset_or_humongous(const oop obj);
 
-  enum in_cset_state_t {
-   InNeither,           // neither in collection set nor humongous
-   InCSet,              // region is in collection set only
-   IsHumongous          // region is a humongous start region
-  };
  private:
-  // Instances of this class are used for quick tests on whether a reference points
-  // into the collection set or is a humongous object (points into a humongous
-  // object).
-  // Each of the array's elements denotes whether the corresponding region is in
-  // the collection set or a humongous region.
-  // We use this to quickly reclaim humongous objects: by making a humongous region
-  // succeed this test, we sort-of add it to the collection set. During the reference
-  // iteration closures, when we see a humongous region, we simply mark it as
-  // referenced, i.e. live.
-  class G1FastCSetBiasedMappedArray : public G1BiasedMappedArray<char> {
-   protected:
-    char default_value() const { return G1CollectedHeap::InNeither; }
-   public:
-    void set_humongous(uintptr_t index) {
-      assert(get_by_index(index) != InCSet, "Should not overwrite InCSet values");
-      set_by_index(index, G1CollectedHeap::IsHumongous);
-    }
-
-    void clear_humongous(uintptr_t index) {
-      set_by_index(index, G1CollectedHeap::InNeither);
-    }
-
-    void set_in_cset(uintptr_t index) {
-      assert(get_by_index(index) != G1CollectedHeap::IsHumongous, "Should not overwrite IsHumongous value");
-      set_by_index(index, G1CollectedHeap::InCSet);
-    }
-
-    bool is_in_cset_or_humongous(HeapWord* addr) const { return get_by_address(addr) != G1CollectedHeap::InNeither; }
-    bool is_in_cset(HeapWord* addr) const { return get_by_address(addr) == G1CollectedHeap::InCSet; }
-    G1CollectedHeap::in_cset_state_t at(HeapWord* addr) const { return (G1CollectedHeap::in_cset_state_t)get_by_address(addr); }
-    void clear() { G1BiasedMappedArray<char>::clear(); }
-  };
-
   // This array is used for a quick test on whether a reference points into
   // the collection set or not. Each of the array's elements denotes whether the
   // corresponding region is in the collection set or not.
-  G1FastCSetBiasedMappedArray _in_cset_fast_test;
+  G1InCSetStateFastTestBiasedMappedArray _in_cset_fast_test;
 
  public:
 
-  inline in_cset_state_t in_cset_state(const oop obj);
+  inline InCSetState in_cset_state(const oop obj);
 
   // Return "TRUE" iff the given object address is in the reserved
   // region of g1.
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.inline.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.inline.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -35,6 +35,41 @@
 #include "runtime/orderAccess.inline.hpp"
 #include "utilities/taskqueue.hpp"
 
+PLABStats* G1CollectedHeap::alloc_buffer_stats(InCSetState dest) {
+  switch (dest.value()) {
+    case InCSetState::Young:
+      return &_survivor_plab_stats;
+    case InCSetState::Old:
+      return &_old_plab_stats;
+    default:
+      ShouldNotReachHere();
+      return NULL; // Keep some compilers happy
+  }
+}
+
+size_t G1CollectedHeap::desired_plab_sz(InCSetState dest) {
+  size_t gclab_word_size = alloc_buffer_stats(dest)->desired_plab_sz();
+  // Prevent humongous PLAB sizes for two reasons:
+  // * PLABs are allocated using a similar paths as oops, but should
+  //   never be in a humongous region
+  // * Allowing humongous PLABs needlessly churns the region free lists
+  return MIN2(_humongous_object_threshold_in_words, gclab_word_size);
+}
+
+HeapWord* G1CollectedHeap::par_allocate_during_gc(InCSetState dest,
+                                                  size_t word_size,
+                                                  AllocationContext_t context) {
+  switch (dest.value()) {
+    case InCSetState::Young:
+      return survivor_attempt_allocation(word_size, context);
+    case InCSetState::Old:
+      return old_attempt_allocation(word_size, context);
+    default:
+      ShouldNotReachHere();
+      return NULL; // Keep some compilers happy
+  }
+}
+
 // Inline functions for G1CollectedHeap
 
 inline AllocationContextStats& G1CollectedHeap::allocation_context_stats() {
@@ -203,7 +238,7 @@
   return _in_cset_fast_test.is_in_cset_or_humongous((HeapWord*)obj);
 }
 
-G1CollectedHeap::in_cset_state_t G1CollectedHeap::in_cset_state(const oop obj) {
+InCSetState G1CollectedHeap::in_cset_state(const oop obj) {
   return _in_cset_fast_test.at((HeapWord*)obj);
 }
 
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -1437,18 +1437,6 @@
   return young_list_length < young_list_max_length;
 }
 
-uint G1CollectorPolicy::max_regions(int purpose) {
-  switch (purpose) {
-    case GCAllocForSurvived:
-      return _max_survivor_regions;
-    case GCAllocForTenured:
-      return REGIONS_UNLIMITED;
-    default:
-      ShouldNotReachHere();
-      return REGIONS_UNLIMITED;
-  };
-}
-
 void G1CollectorPolicy::update_max_gc_locker_expansion() {
   uint expansion_region_num = 0;
   if (GCLockerEdenExpansionPercent > 0) {
@@ -1634,7 +1622,7 @@
   hr->set_next_in_collection_set(_collection_set);
   _collection_set = hr;
   _collection_set_bytes_used_before += hr->used();
-  _g1->register_region_with_in_cset_fast_test(hr);
+  _g1->register_old_region_with_in_cset_fast_test(hr);
   size_t rs_length = hr->rem_set()->occupied();
   _recorded_rs_lengths += rs_length;
   _old_cset_region_length += 1;
@@ -1767,7 +1755,7 @@
   hr->set_in_collection_set(true);
   assert( hr->next_in_collection_set() == NULL, "invariant");
 
-  _g1->register_region_with_in_cset_fast_test(hr);
+  _g1->register_young_region_with_in_cset_fast_test(hr);
 }
 
 // Add the region at the RHS of the incremental cset
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -881,28 +881,20 @@
 public:
   uint tenuring_threshold() const { return _tenuring_threshold; }
 
-  inline GCAllocPurpose
-    evacuation_destination(HeapRegion* src_region, uint age, size_t word_sz) {
-      if (age < _tenuring_threshold && src_region->is_young()) {
-        return GCAllocForSurvived;
-      } else {
-        return GCAllocForTenured;
-      }
-  }
-
-  inline bool track_object_age(GCAllocPurpose purpose) {
-    return purpose == GCAllocForSurvived;
-  }
-
   static const uint REGIONS_UNLIMITED = (uint) -1;
 
-  uint max_regions(int purpose);
-
-  // The limit on regions for a particular purpose is reached.
-  void note_alloc_region_limit_reached(int purpose) {
-    if (purpose == GCAllocForSurvived) {
-      _tenuring_threshold = 0;
+  uint max_regions(InCSetState dest) {
+    switch (dest.value()) {
+      case InCSetState::Young:
+        return _max_survivor_regions;
+      case InCSetState::Old:
+        return REGIONS_UNLIMITED;
+      default:
+        assert(false, err_msg("Unknown dest state: " CSETSTATE_FORMAT, dest.value()));
+        break;
     }
+    // keep some compilers happy
+    return 0;
   }
 
   void note_start_adding_survivor_regions() {
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1InCSetState.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -0,0 +1,132 @@
+/*
+ * Copyright (c) 2014, Oracle and/or its affiliates. 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 Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#ifndef SHARE_VM_GC_IMPLEMENTATION_G1_G1INCSETSTATE_HPP
+#define SHARE_VM_GC_IMPLEMENTATION_G1_G1INCSETSTATE_HPP
+
+#include "gc_implementation/g1/g1BiasedArray.hpp"
+#include "memory/allocation.hpp"
+
+// Per-region state during garbage collection.
+struct InCSetState {
+ public:
+  // We use different types to represent the state value. Particularly SPARC puts
+  // values in structs from "left to right", i.e. MSB to LSB. This results in many
+  // unnecessary shift operations when loading and storing values of this type.
+  // This degrades performance significantly (>10%) on that platform.
+  // Other tested ABIs do not seem to have this problem, and actually tend to
+  // favor smaller types, so we use the smallest usable type there.
+#ifdef SPARC
+  #define CSETSTATE_FORMAT INTPTR_FORMAT
+  typedef intptr_t in_cset_state_t;
+#else
+  #define CSETSTATE_FORMAT "%d"
+  typedef int8_t in_cset_state_t;
+#endif
+ private:
+  in_cset_state_t _value;
+ public:
+  enum {
+    // Selection of the values were driven to micro-optimize the encoding and
+    // frequency of the checks.
+    // The most common check is whether the region is in the collection set or not.
+    // This encoding allows us to use an != 0 check which in some architectures
+    // (x86*) can be encoded slightly more efficently than a normal comparison
+    // against zero.
+    // The same situation occurs when checking whether the region is humongous
+    // or not, which is encoded by values < 0.
+    // The other values are simply encoded in increasing generation order, which
+    // makes getting the next generation fast by a simple increment.
+    Humongous    = -1,    // The region is humongous - note that actually any value < 0 would be possible here.
+    NotInCSet    =  0,    // The region is not in the collection set.
+    Young        =  1,    // The region is in the collection set and a young region.
+    Old          =  2,    // The region is in the collection set and an old region.
+    Num
+  };
+
+  InCSetState(in_cset_state_t value = NotInCSet) : _value(value) {
+    assert(is_valid(), err_msg("Invalid state %d", _value));
+  }
+
+  in_cset_state_t value() const        { return _value; }
+
+  void set_old()                       { _value = Old; }
+
+  bool is_in_cset_or_humongous() const { return _value != NotInCSet; }
+  bool is_in_cset() const              { return _value > NotInCSet; }
+  bool is_humongous() const            { return _value < NotInCSet; }
+  bool is_young() const                { return _value == Young; }
+  bool is_old() const                  { return _value == Old; }
+
+#ifdef ASSERT
+  bool is_default() const              { return !is_in_cset_or_humongous(); }
+  bool is_valid() const                { return (_value >= Humongous) && (_value < Num); }
+  bool is_valid_gen() const            { return (_value >= Young && _value <= Old); }
+#endif
+};
+
+// Instances of this class are used for quick tests on whether a reference points
+// into the collection set and into which generation or is a humongous object
+//
+// Each of the array's elements indicates whether the corresponding region is in
+// the collection set and if so in which generation, or a humongous region.
+//
+// We use this to speed up reference processing during young collection and
+// quickly reclaim humongous objects. For the latter, by making a humongous region
+// succeed this test, we sort-of add it to the collection set. During the reference
+// iteration closures, when we see a humongous region, we then simply mark it as
+// referenced, i.e. live.
+class G1InCSetStateFastTestBiasedMappedArray : public G1BiasedMappedArray<InCSetState> {
+ protected:
+  InCSetState default_value() const { return InCSetState::NotInCSet; }
+ public:
+  void set_humongous(uintptr_t index) {
+    assert(get_by_index(index).is_default(),
+           err_msg("State at index " INTPTR_FORMAT" should be default but is " CSETSTATE_FORMAT, index, get_by_index(index).value()));
+    set_by_index(index, InCSetState::Humongous);
+  }
+
+  void clear_humongous(uintptr_t index) {
+    set_by_index(index, InCSetState::NotInCSet);
+  }
+
+  void set_in_young(uintptr_t index) {
+    assert(get_by_index(index).is_default(),
+           err_msg("State at index " INTPTR_FORMAT" should be default but is " CSETSTATE_FORMAT, index, get_by_index(index).value()));
+    set_by_index(index, InCSetState::Young);
+  }
+
+  void set_in_old(uintptr_t index) {
+    assert(get_by_index(index).is_default(),
+           err_msg("State at index " INTPTR_FORMAT" should be default but is " CSETSTATE_FORMAT, index, get_by_index(index).value()));
+    set_by_index(index, InCSetState::Old);
+  }
+
+  bool is_in_cset_or_humongous(HeapWord* addr) const { return at(addr).is_in_cset_or_humongous(); }
+  bool is_in_cset(HeapWord* addr) const { return at(addr).is_in_cset(); }
+  InCSetState at(HeapWord* addr) const { return get_by_address(addr); }
+  void clear() { G1BiasedMappedArray<InCSetState>::clear(); }
+};
+
+#endif // SHARE_VM_GC_IMPLEMENTATION_G1_G1INCSETSTATE_HPP
--- a/hotspot/src/share/vm/gc_implementation/g1/g1OopClosures.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1OopClosures.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -26,6 +26,7 @@
 #define SHARE_VM_GC_IMPLEMENTATION_G1_G1OOPCLOSURES_HPP
 
 #include "memory/iterator.hpp"
+#include "oops/markOop.hpp"
 
 class HeapRegion;
 class G1CollectedHeap;
@@ -239,14 +240,14 @@
   G1CollectedHeap* _g1;
   G1RemSet* _g1_rem_set;
   HeapRegion* _from;
-  OopsInHeapRegionClosure* _push_ref_cl;
+  G1ParPushHeapRSClosure* _push_ref_cl;
   bool _record_refs_into_cset;
   uint _worker_i;
 
 public:
   G1UpdateRSOrPushRefOopClosure(G1CollectedHeap* g1h,
                                 G1RemSet* rs,
-                                OopsInHeapRegionClosure* push_ref_cl,
+                                G1ParPushHeapRSClosure* push_ref_cl,
                                 bool record_refs_into_cset,
                                 uint worker_i = 0);
 
@@ -256,7 +257,8 @@
   }
 
   bool self_forwarded(oop obj) {
-    bool result = (obj->is_forwarded() && (obj->forwardee()== obj));
+    markOop m = obj->mark();
+    bool result = (m->is_marked() && ((oop)m->decode_pointer() == obj));
     return result;
   }
 
--- a/hotspot/src/share/vm/gc_implementation/g1/g1OopClosures.inline.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1OopClosures.inline.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -67,8 +67,8 @@
 
   if (!oopDesc::is_null(heap_oop)) {
     oop obj = oopDesc::decode_heap_oop_not_null(heap_oop);
-    G1CollectedHeap::in_cset_state_t state = _g1->in_cset_state(obj);
-    if (state == G1CollectedHeap::InCSet) {
+    const InCSetState state = _g1->in_cset_state(obj);
+    if (state.is_in_cset()) {
       // We're not going to even bother checking whether the object is
       // already forwarded or not, as this usually causes an immediate
       // stall. We'll try to prefetch the object (for write, given that
@@ -87,7 +87,7 @@
 
       _par_scan_state->push_on_queue(p);
     } else {
-      if (state == G1CollectedHeap::IsHumongous) {
+      if (state.is_humongous()) {
         _g1->set_humongous_is_live(obj);
       }
       _par_scan_state->update_rs(_from, p, _worker_id);
--- a/hotspot/src/share/vm/gc_implementation/g1/g1PageBasedVirtualSpace.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1PageBasedVirtualSpace.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -131,6 +131,9 @@
   _committed.set_range(start, start + size_in_pages);
 
   MemRegion result((HeapWord*)page_start(start), byte_size_for_pages(size_in_pages) / HeapWordSize);
+  if (AlwaysPreTouch) {
+    os::pretouch_memory((char*)result.start(), (char*)result.end());
+  }
   return result;
 }
 
--- a/hotspot/src/share/vm/gc_implementation/g1/g1ParScanThreadState.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1ParScanThreadState.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -38,6 +38,7 @@
     _g1_rem(g1h->g1_rem_set()),
     _hash_seed(17), _queue_num(queue_num),
     _term_attempts(0),
+    _tenuring_threshold(g1h->g1_policy()->tenuring_threshold()),
     _age_table(false), _scanner(g1h, rp),
     _strong_roots_time(0), _term_time(0) {
   _scanner.set_par_scan_thread_state(this);
@@ -59,6 +60,12 @@
 
   _g1_par_allocator = G1ParGCAllocator::create_allocator(_g1h);
 
+  _dest[InCSetState::NotInCSet]    = InCSetState::NotInCSet;
+  // The dest for Young is used when the objects are aged enough to
+  // need to be moved to the next space.
+  _dest[InCSetState::Young]        = InCSetState::Old;
+  _dest[InCSetState::Old]          = InCSetState::Old;
+
   _start = os::elapsedTime();
 }
 
@@ -150,52 +157,94 @@
   } while (!_refs->is_empty());
 }
 
-oop G1ParScanThreadState::copy_to_survivor_space(oop const old,
+HeapWord* G1ParScanThreadState::allocate_in_next_plab(InCSetState const state,
+                                                      InCSetState* dest,
+                                                      size_t word_sz,
+                                                      AllocationContext_t const context) {
+  assert(state.is_in_cset_or_humongous(), err_msg("Unexpected state: " CSETSTATE_FORMAT, state.value()));
+  assert(dest->is_in_cset_or_humongous(), err_msg("Unexpected dest: " CSETSTATE_FORMAT, dest->value()));
+
+  // Right now we only have two types of regions (young / old) so
+  // let's keep the logic here simple. We can generalize it when necessary.
+  if (dest->is_young()) {
+    HeapWord* const obj_ptr = _g1_par_allocator->allocate(InCSetState::Old,
+                                                          word_sz, context);
+    if (obj_ptr == NULL) {
+      return NULL;
+    }
+    // Make sure that we won't attempt to copy any other objects out
+    // of a survivor region (given that apparently we cannot allocate
+    // any new ones) to avoid coming into this slow path.
+    _tenuring_threshold = 0;
+    dest->set_old();
+    return obj_ptr;
+  } else {
+    assert(dest->is_old(), err_msg("Unexpected dest: " CSETSTATE_FORMAT, dest->value()));
+    // no other space to try.
+    return NULL;
+  }
+}
+
+InCSetState G1ParScanThreadState::next_state(InCSetState const state, markOop const m, uint& age) {
+  if (state.is_young()) {
+    age = !m->has_displaced_mark_helper() ? m->age()
+                                          : m->displaced_mark_helper()->age();
+    if (age < _tenuring_threshold) {
+      return state;
+    }
+  }
+  return dest(state);
+}
+
+oop G1ParScanThreadState::copy_to_survivor_space(InCSetState const state,
+                                                 oop const old,
                                                  markOop const old_mark) {
-  size_t word_sz = old->size();
-  HeapRegion* from_region = _g1h->heap_region_containing_raw(old);
+  const size_t word_sz = old->size();
+  HeapRegion* const from_region = _g1h->heap_region_containing_raw(old);
   // +1 to make the -1 indexes valid...
-  int       young_index = from_region->young_index_in_cset()+1;
+  const int young_index = from_region->young_index_in_cset()+1;
   assert( (from_region->is_young() && young_index >  0) ||
          (!from_region->is_young() && young_index == 0), "invariant" );
-  G1CollectorPolicy* g1p = _g1h->g1_policy();
-  uint age = old_mark->has_displaced_mark_helper() ? old_mark->displaced_mark_helper()->age()
-                                                   : old_mark->age();
-  GCAllocPurpose alloc_purpose = g1p->evacuation_destination(from_region, age,
-                                                             word_sz);
-  AllocationContext_t context = from_region->allocation_context();
-  HeapWord* obj_ptr = _g1_par_allocator->allocate(alloc_purpose, word_sz, context);
+  const AllocationContext_t context = from_region->allocation_context();
+
+  uint age = 0;
+  InCSetState dest_state = next_state(state, old_mark, age);
+  HeapWord* obj_ptr = _g1_par_allocator->plab_allocate(dest_state, word_sz, context);
+
+  // PLAB allocations should succeed most of the time, so we'll
+  // normally check against NULL once and that's it.
+  if (obj_ptr == NULL) {
+    obj_ptr = _g1_par_allocator->allocate_direct_or_new_plab(dest_state, word_sz, context);
+    if (obj_ptr == NULL) {
+      obj_ptr = allocate_in_next_plab(state, &dest_state, word_sz, context);
+      if (obj_ptr == NULL) {
+        // This will either forward-to-self, or detect that someone else has
+        // installed a forwarding pointer.
+        return _g1h->handle_evacuation_failure_par(this, old);
+      }
+    }
+  }
+
+  assert(obj_ptr != NULL, "when we get here, allocation should have succeeded");
 #ifndef PRODUCT
   // Should this evacuation fail?
   if (_g1h->evacuation_should_fail()) {
-    if (obj_ptr != NULL) {
-      _g1_par_allocator->undo_allocation(alloc_purpose, obj_ptr, word_sz, context);
-      obj_ptr = NULL;
-    }
+    // Doing this after all the allocation attempts also tests the
+    // undo_allocation() method too.
+    _g1_par_allocator->undo_allocation(dest_state, obj_ptr, word_sz, context);
+    return _g1h->handle_evacuation_failure_par(this, old);
   }
 #endif // !PRODUCT
 
-  if (obj_ptr == NULL) {
-    // This will either forward-to-self, or detect that someone else has
-    // installed a forwarding pointer.
-    return _g1h->handle_evacuation_failure_par(this, old);
-  }
-
-  oop obj = oop(obj_ptr);
-
   // We're going to allocate linearly, so might as well prefetch ahead.
   Prefetch::write(obj_ptr, PrefetchCopyIntervalInBytes);
 
-  oop forward_ptr = old->forward_to_atomic(obj);
+  const oop obj = oop(obj_ptr);
+  const oop forward_ptr = old->forward_to_atomic(obj);
   if (forward_ptr == NULL) {
     Copy::aligned_disjoint_words((HeapWord*) old, obj_ptr, word_sz);
 
-    // alloc_purpose is just a hint to allocate() above, recheck the type of region
-    // we actually allocated from and update alloc_purpose accordingly
-    HeapRegion* to_region = _g1h->heap_region_containing_raw(obj_ptr);
-    alloc_purpose = to_region->is_young() ? GCAllocForSurvived : GCAllocForTenured;
-
-    if (g1p->track_object_age(alloc_purpose)) {
+    if (dest_state.is_young()) {
       if (age < markOopDesc::max_age) {
         age++;
       }
@@ -215,13 +264,19 @@
     }
 
     if (G1StringDedup::is_enabled()) {
-      G1StringDedup::enqueue_from_evacuation(from_region->is_young(),
-                                             to_region->is_young(),
+      const bool is_from_young = state.is_young();
+      const bool is_to_young = dest_state.is_young();
+      assert(is_from_young == _g1h->heap_region_containing_raw(old)->is_young(),
+             "sanity");
+      assert(is_to_young == _g1h->heap_region_containing_raw(obj)->is_young(),
+             "sanity");
+      G1StringDedup::enqueue_from_evacuation(is_from_young,
+                                             is_to_young,
                                              queue_num(),
                                              obj);
     }
 
-    size_t* surv_young_words = surviving_young_words();
+    size_t* const surv_young_words = surviving_young_words();
     surv_young_words[young_index] += word_sz;
 
     if (obj->is_objArray() && arrayOop(obj)->length() >= ParGCArrayScanChunk) {
@@ -232,14 +287,13 @@
       oop* old_p = set_partial_array_mask(old);
       push_on_queue(old_p);
     } else {
-      // No point in using the slower heap_region_containing() method,
-      // given that we know obj is in the heap.
-      _scanner.set_region(_g1h->heap_region_containing_raw(obj));
+      HeapRegion* const to_region = _g1h->heap_region_containing_raw(obj_ptr);
+      _scanner.set_region(to_region);
       obj->oop_iterate_backwards(&_scanner);
     }
+    return obj;
   } else {
-    _g1_par_allocator->undo_allocation(alloc_purpose, obj_ptr, word_sz, context);
-    obj = forward_ptr;
+    _g1_par_allocator->undo_allocation(dest_state, obj_ptr, word_sz, context);
+    return forward_ptr;
   }
-  return obj;
 }
--- a/hotspot/src/share/vm/gc_implementation/g1/g1ParScanThreadState.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1ParScanThreadState.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -46,14 +46,16 @@
   G1SATBCardTableModRefBS* _ct_bs;
   G1RemSet* _g1_rem;
 
-  G1ParGCAllocator*   _g1_par_allocator;
-
-  ageTable            _age_table;
+  G1ParGCAllocator* _g1_par_allocator;
 
-  G1ParScanClosure    _scanner;
+  ageTable          _age_table;
+  InCSetState       _dest[InCSetState::Num];
+  // Local tenuring threshold.
+  uint              _tenuring_threshold;
+  G1ParScanClosure  _scanner;
 
-  size_t           _alloc_buffer_waste;
-  size_t           _undo_waste;
+  size_t            _alloc_buffer_waste;
+  size_t            _undo_waste;
 
   OopsInHeapRegionClosure*      _evac_failure_cl;
 
@@ -82,6 +84,14 @@
   DirtyCardQueue& dirty_card_queue()             { return _dcq;  }
   G1SATBCardTableModRefBS* ctbs()                { return _ct_bs; }
 
+  InCSetState dest(InCSetState original) const {
+    assert(original.is_valid(),
+           err_msg("Original state invalid: " CSETSTATE_FORMAT, original.value()));
+    assert(_dest[original.value()].is_valid_gen(),
+           err_msg("Dest state is invalid: " CSETSTATE_FORMAT, _dest[original.value()].value()));
+    return _dest[original.value()];
+  }
+
  public:
   G1ParScanThreadState(G1CollectedHeap* g1h, uint queue_num, ReferenceProcessor* rp);
   ~G1ParScanThreadState();
@@ -112,7 +122,6 @@
       }
     }
   }
- public:
 
   void set_evac_failure_closure(OopsInHeapRegionClosure* evac_failure_cl) {
     _evac_failure_cl = evac_failure_cl;
@@ -193,9 +202,20 @@
   template <class T> inline void deal_with_reference(T* ref_to_scan);
 
   inline void dispatch_reference(StarTask ref);
+
+  // Tries to allocate word_sz in the PLAB of the next "generation" after trying to
+  // allocate into dest. State is the original (source) cset state for the object
+  // that is allocated for.
+  // Returns a non-NULL pointer if successful, and updates dest if required.
+  HeapWord* allocate_in_next_plab(InCSetState const state,
+                                  InCSetState* dest,
+                                  size_t word_sz,
+                                  AllocationContext_t const context);
+
+  inline InCSetState next_state(InCSetState const state, markOop const m, uint& age);
  public:
 
-  oop copy_to_survivor_space(oop const obj, markOop const old_mark);
+  oop copy_to_survivor_space(InCSetState const state, oop const obj, markOop const old_mark);
 
   void trim_queue();
 
--- a/hotspot/src/share/vm/gc_implementation/g1/g1ParScanThreadState.inline.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1ParScanThreadState.inline.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -38,21 +38,21 @@
   // set, due to (benign) races in the claim mechanism during RSet scanning more
   // than one thread might claim the same card. So the same card may be
   // processed multiple times. So redo this check.
-  G1CollectedHeap::in_cset_state_t in_cset_state = _g1h->in_cset_state(obj);
-  if (in_cset_state == G1CollectedHeap::InCSet) {
+  const InCSetState in_cset_state = _g1h->in_cset_state(obj);
+  if (in_cset_state.is_in_cset()) {
     oop forwardee;
     markOop m = obj->mark();
     if (m->is_marked()) {
       forwardee = (oop) m->decode_pointer();
     } else {
-      forwardee = copy_to_survivor_space(obj, m);
+      forwardee = copy_to_survivor_space(in_cset_state, obj, m);
     }
     oopDesc::encode_store_heap_oop(p, forwardee);
-  } else if (in_cset_state == G1CollectedHeap::IsHumongous) {
+  } else if (in_cset_state.is_humongous()) {
     _g1h->set_humongous_is_live(obj);
   } else {
-    assert(in_cset_state == G1CollectedHeap::InNeither,
-           err_msg("In_cset_state must be InNeither here, but is %d", in_cset_state));
+    assert(!in_cset_state.is_in_cset_or_humongous(),
+           err_msg("In_cset_state must be NotInCSet here, but is " CSETSTATE_FORMAT, in_cset_state.value()));
   }
 
   assert(obj != NULL, "Must be");
--- a/hotspot/src/share/vm/gc_implementation/g1/g1RemSet.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1RemSet.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -80,7 +80,7 @@
     _prev_period_summary()
 {
   _seq_task = new SubTasksDone(NumSeqTasks);
-  _cset_rs_update_cl = NEW_C_HEAP_ARRAY(OopsInHeapRegionClosure*, n_workers(), mtGC);
+  _cset_rs_update_cl = NEW_C_HEAP_ARRAY(G1ParPushHeapRSClosure*, n_workers(), mtGC);
   for (uint i = 0; i < n_workers(); i++) {
     _cset_rs_update_cl[i] = NULL;
   }
@@ -94,14 +94,14 @@
   for (uint i = 0; i < n_workers(); i++) {
     assert(_cset_rs_update_cl[i] == NULL, "it should be");
   }
-  FREE_C_HEAP_ARRAY(OopsInHeapRegionClosure*, _cset_rs_update_cl);
+  FREE_C_HEAP_ARRAY(G1ParPushHeapRSClosure*, _cset_rs_update_cl);
 }
 
 class ScanRSClosure : public HeapRegionClosure {
   size_t _cards_done, _cards;
   G1CollectedHeap* _g1h;
 
-  OopsInHeapRegionClosure* _oc;
+  G1ParPushHeapRSClosure* _oc;
   CodeBlobClosure* _code_root_cl;
 
   G1BlockOffsetSharedArray* _bot_shared;
@@ -113,7 +113,7 @@
   bool   _try_claimed;
 
 public:
-  ScanRSClosure(OopsInHeapRegionClosure* oc,
+  ScanRSClosure(G1ParPushHeapRSClosure* oc,
                 CodeBlobClosure* code_root_cl,
                 uint worker_i) :
     _oc(oc),
@@ -135,8 +135,7 @@
   void scanCard(size_t index, HeapRegion *r) {
     // Stack allocate the DirtyCardToOopClosure instance
     HeapRegionDCTOC cl(_g1h, r, _oc,
-                       CardTableModRefBS::Precise,
-                       HeapRegionDCTOC::IntoCSFilterKind);
+                       CardTableModRefBS::Precise);
 
     // Set the "from" region in the closure.
     _oc->set_region(r);
@@ -231,7 +230,7 @@
   size_t cards_looked_up() { return _cards;}
 };
 
-void G1RemSet::scanRS(OopsInHeapRegionClosure* oc,
+void G1RemSet::scanRS(G1ParPushHeapRSClosure* oc,
                       CodeBlobClosure* code_root_cl,
                       uint worker_i) {
   double rs_time_start = os::elapsedTime();
@@ -301,7 +300,7 @@
   HeapRegionRemSet::cleanup();
 }
 
-void G1RemSet::oops_into_collection_set_do(OopsInHeapRegionClosure* oc,
+void G1RemSet::oops_into_collection_set_do(G1ParPushHeapRSClosure* oc,
                                            CodeBlobClosure* code_root_cl,
                                            uint worker_i) {
 #if CARD_REPEAT_HISTO
@@ -417,7 +416,7 @@
 G1UpdateRSOrPushRefOopClosure::
 G1UpdateRSOrPushRefOopClosure(G1CollectedHeap* g1h,
                               G1RemSet* rs,
-                              OopsInHeapRegionClosure* push_ref_cl,
+                              G1ParPushHeapRSClosure* push_ref_cl,
                               bool record_refs_into_cset,
                               uint worker_i) :
   _g1(g1h), _g1_rem_set(rs), _from(NULL),
@@ -518,7 +517,7 @@
   ct_freq_note_card(_ct_bs->index_for(start));
 #endif
 
-  OopsInHeapRegionClosure* oops_in_heap_closure = NULL;
+  G1ParPushHeapRSClosure* oops_in_heap_closure = NULL;
   if (check_for_refs_into_cset) {
     // ConcurrentG1RefineThreads have worker numbers larger than what
     // _cset_rs_update_cl[] is set up to handle. But those threads should
--- a/hotspot/src/share/vm/gc_implementation/g1/g1RemSet.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1RemSet.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -33,6 +33,7 @@
 class G1CollectedHeap;
 class CardTableModRefBarrierSet;
 class ConcurrentG1Refine;
+class G1ParPushHeapRSClosure;
 
 // A G1RemSet in which each heap region has a rem set that records the
 // external heap references into it.  Uses a mod ref bs to track updates,
@@ -68,7 +69,7 @@
 
   // Used for caching the closure that is responsible for scanning
   // references into the collection set.
-  OopsInHeapRegionClosure** _cset_rs_update_cl;
+  G1ParPushHeapRSClosure** _cset_rs_update_cl;
 
   // Print the given summary info
   virtual void print_summary_info(G1RemSetSummary * summary, const char * header = NULL);
@@ -95,7 +96,7 @@
   // partitioning the work to be done. It should be the same as
   // the "i" passed to the calling thread's work(i) function.
   // In the sequential case this param will be ignored.
-  void oops_into_collection_set_do(OopsInHeapRegionClosure* blk,
+  void oops_into_collection_set_do(G1ParPushHeapRSClosure* blk,
                                    CodeBlobClosure* code_root_cl,
                                    uint worker_i);
 
@@ -107,7 +108,7 @@
   void prepare_for_oops_into_collection_set_do();
   void cleanup_after_oops_into_collection_set_do();
 
-  void scanRS(OopsInHeapRegionClosure* oc,
+  void scanRS(G1ParPushHeapRSClosure* oc,
               CodeBlobClosure* code_root_cl,
               uint worker_i);
 
--- a/hotspot/src/share/vm/gc_implementation/g1/g1SATBCardTableModRefBS.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1SATBCardTableModRefBS.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2014, Oracle and/or its affiliates. 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
@@ -33,7 +33,7 @@
 #include "runtime/thread.inline.hpp"
 
 G1SATBCardTableModRefBS::G1SATBCardTableModRefBS(MemRegion whole_heap) :
-    CardTableModRefBSForCTRS(whole_heap)
+    CardTableModRefBS(whole_heap)
 {
   _kind = G1SATBCT;
 }
--- a/hotspot/src/share/vm/gc_implementation/g1/g1SATBCardTableModRefBS.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1SATBCardTableModRefBS.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2014, Oracle and/or its affiliates. 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
@@ -37,7 +37,7 @@
 // This barrier is specialized to use a logging barrier to support
 // snapshot-at-the-beginning marking.
 
-class G1SATBCardTableModRefBS: public CardTableModRefBSForCTRS {
+class G1SATBCardTableModRefBS: public CardTableModRefBS {
 protected:
   enum G1CardValues {
     g1_young_gen = CT_MR_BS_last_reserved << 1
--- a/hotspot/src/share/vm/gc_implementation/g1/heapRegion.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/heapRegion.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -48,93 +48,55 @@
 size_t HeapRegion::CardsPerRegion    = 0;
 
 HeapRegionDCTOC::HeapRegionDCTOC(G1CollectedHeap* g1,
-                                 HeapRegion* hr, ExtendedOopClosure* cl,
-                                 CardTableModRefBS::PrecisionStyle precision,
-                                 FilterKind fk) :
+                                 HeapRegion* hr,
+                                 G1ParPushHeapRSClosure* cl,
+                                 CardTableModRefBS::PrecisionStyle precision) :
   DirtyCardToOopClosure(hr, cl, precision, NULL),
-  _hr(hr), _fk(fk), _g1(g1) { }
+  _hr(hr), _rs_scan(cl), _g1(g1) { }
 
 FilterOutOfRegionClosure::FilterOutOfRegionClosure(HeapRegion* r,
                                                    OopClosure* oc) :
   _r_bottom(r->bottom()), _r_end(r->end()), _oc(oc) { }
 
-template<class ClosureType>
-HeapWord* walk_mem_region_loop(ClosureType* cl, G1CollectedHeap* g1h,
-                               HeapRegion* hr,
-                               HeapWord* cur, HeapWord* top) {
-  oop cur_oop = oop(cur);
-  size_t oop_size = hr->block_size(cur);
-  HeapWord* next_obj = cur + oop_size;
-  while (next_obj < top) {
-    // Keep filtering the remembered set.
-    if (!g1h->is_obj_dead(cur_oop, hr)) {
-      // Bottom lies entirely below top, so we can call the
-      // non-memRegion version of oop_iterate below.
-      cur_oop->oop_iterate(cl);
-    }
-    cur = next_obj;
-    cur_oop = oop(cur);
-    oop_size = hr->block_size(cur);
-    next_obj = cur + oop_size;
-  }
-  return cur;
-}
-
 void HeapRegionDCTOC::walk_mem_region(MemRegion mr,
                                       HeapWord* bottom,
                                       HeapWord* top) {
   G1CollectedHeap* g1h = _g1;
   size_t oop_size;
-  ExtendedOopClosure* cl2 = NULL;
-
-  FilterIntoCSClosure intoCSFilt(this, g1h, _cl);
-  FilterOutOfRegionClosure outOfRegionFilt(_hr, _cl);
-
-  switch (_fk) {
-  case NoFilterKind:          cl2 = _cl; break;
-  case IntoCSFilterKind:      cl2 = &intoCSFilt; break;
-  case OutOfRegionFilterKind: cl2 = &outOfRegionFilt; break;
-  default:                    ShouldNotReachHere();
-  }
+  HeapWord* cur = bottom;
 
   // Start filtering what we add to the remembered set. If the object is
   // not considered dead, either because it is marked (in the mark bitmap)
   // or it was allocated after marking finished, then we add it. Otherwise
   // we can safely ignore the object.
-  if (!g1h->is_obj_dead(oop(bottom), _hr)) {
-    oop_size = oop(bottom)->oop_iterate(cl2, mr);
+  if (!g1h->is_obj_dead(oop(cur), _hr)) {
+    oop_size = oop(cur)->oop_iterate(_rs_scan, mr);
   } else {
-    oop_size = _hr->block_size(bottom);
+    oop_size = _hr->block_size(cur);
   }
 
-  bottom += oop_size;
-
-  if (bottom < top) {
-    // We replicate the loop below for several kinds of possible filters.
-    switch (_fk) {
-    case NoFilterKind:
-      bottom = walk_mem_region_loop(_cl, g1h, _hr, bottom, top);
-      break;
+  cur += oop_size;
 
-    case IntoCSFilterKind: {
-      FilterIntoCSClosure filt(this, g1h, _cl);
-      bottom = walk_mem_region_loop(&filt, g1h, _hr, bottom, top);
-      break;
-    }
-
-    case OutOfRegionFilterKind: {
-      FilterOutOfRegionClosure filt(_hr, _cl);
-      bottom = walk_mem_region_loop(&filt, g1h, _hr, bottom, top);
-      break;
-    }
-
-    default:
-      ShouldNotReachHere();
+  if (cur < top) {
+    oop cur_oop = oop(cur);
+    oop_size = _hr->block_size(cur);
+    HeapWord* next_obj = cur + oop_size;
+    while (next_obj < top) {
+      // Keep filtering the remembered set.
+      if (!g1h->is_obj_dead(cur_oop, _hr)) {
+        // Bottom lies entirely below top, so we can call the
+        // non-memRegion version of oop_iterate below.
+        cur_oop->oop_iterate(_rs_scan);
+      }
+      cur = next_obj;
+      cur_oop = oop(cur);
+      oop_size = _hr->block_size(cur);
+      next_obj = cur + oop_size;
     }
 
     // Last object. Need to do dead-obj filtering here too.
-    if (!g1h->is_obj_dead(oop(bottom), _hr)) {
-      oop(bottom)->oop_iterate(cl2, mr);
+    if (!g1h->is_obj_dead(oop(cur), _hr)) {
+      oop(cur)->oop_iterate(_rs_scan, mr);
     }
   }
 }
--- a/hotspot/src/share/vm/gc_implementation/g1/heapRegion.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/heapRegion.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -67,17 +67,9 @@
 // sets.
 
 class HeapRegionDCTOC : public DirtyCardToOopClosure {
-public:
-  // Specification of possible DirtyCardToOopClosure filtering.
-  enum FilterKind {
-    NoFilterKind,
-    IntoCSFilterKind,
-    OutOfRegionFilterKind
-  };
-
-protected:
+private:
   HeapRegion* _hr;
-  FilterKind _fk;
+  G1ParPushHeapRSClosure* _rs_scan;
   G1CollectedHeap* _g1;
 
   // Walk the given memory region from bottom to (actual) top
@@ -90,9 +82,9 @@
 
 public:
   HeapRegionDCTOC(G1CollectedHeap* g1,
-                  HeapRegion* hr, ExtendedOopClosure* cl,
-                  CardTableModRefBS::PrecisionStyle precision,
-                  FilterKind fk);
+                  HeapRegion* hr,
+                  G1ParPushHeapRSClosure* cl,
+                  CardTableModRefBS::PrecisionStyle precision);
 };
 
 // The complicating factor is that BlockOffsetTable diverged
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psMarkSweep.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psMarkSweep.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -195,7 +195,7 @@
 
     COMPILER2_PRESENT(DerivedPointerTable::clear());
 
-    ref_processor()->enable_discovery(true /*verify_disabled*/, true /*verify_no_refs*/);
+    ref_processor()->enable_discovery();
     ref_processor()->setup_policy(clear_all_softrefs);
 
     mark_sweep_phase1(clear_all_softrefs);
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psParallelCompact.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psParallelCompact.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -2069,7 +2069,7 @@
 
     COMPILER2_PRESENT(DerivedPointerTable::clear());
 
-    ref_processor()->enable_discovery(true /*verify_disabled*/, true /*verify_no_refs*/);
+    ref_processor()->enable_discovery();
     ref_processor()->setup_policy(maximum_heap_compaction);
 
     bool marked_for_unloading = false;
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psScavenge.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psScavenge.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -362,7 +362,7 @@
 
     COMPILER2_PRESENT(DerivedPointerTable::clear());
 
-    reference_processor()->enable_discovery(true /*verify_disabled*/, true /*verify_no_refs*/);
+    reference_processor()->enable_discovery();
     reference_processor()->setup_policy(false);
 
     // We track how much was promoted to the next generation for
--- a/hotspot/src/share/vm/gc_implementation/shared/mutableSpace.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/gc_implementation/shared/mutableSpace.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -63,9 +63,7 @@
 }
 
 void MutableSpace::pretouch_pages(MemRegion mr) {
-  for (volatile char *p = (char*)mr.start(); p < (char*)mr.end(); p += os::vm_page_size()) {
-    char t = *p; *p = t;
-  }
+  os::pretouch_memory((char*)mr.start(), (char*)mr.end());
 }
 
 void MutableSpace::initialize(MemRegion mr,
--- a/hotspot/src/share/vm/memory/cardTableRS.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/memory/cardTableRS.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -33,24 +33,13 @@
 #include "runtime/java.hpp"
 #include "runtime/os.hpp"
 #include "utilities/macros.hpp"
-#if INCLUDE_ALL_GCS
-#include "gc_implementation/g1/concurrentMark.hpp"
-#include "gc_implementation/g1/g1SATBCardTableModRefBS.hpp"
-#endif // INCLUDE_ALL_GCS
 
 CardTableRS::CardTableRS(MemRegion whole_heap) :
   GenRemSet(),
   _cur_youngergen_card_val(youngergenP1_card)
 {
-#if INCLUDE_ALL_GCS
-  if (UseG1GC) {
-      _ct_bs = new G1SATBCardTableLoggingModRefBS(whole_heap);
-  } else {
-    _ct_bs = new CardTableModRefBSForCTRS(whole_heap);
-  }
-#else
+  guarantee(Universe::heap()->kind() == CollectedHeap::GenCollectedHeap, "sanity");
   _ct_bs = new CardTableModRefBSForCTRS(whole_heap);
-#endif
   _ct_bs->initialize();
   set_bs(_ct_bs);
   _last_cur_val_in_gen = NEW_C_HEAP_ARRAY3(jbyte, GenCollectedHeap::max_gens + 1,
--- a/hotspot/src/share/vm/memory/genCollectedHeap.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/memory/genCollectedHeap.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -70,6 +70,7 @@
 
 GenCollectedHeap::GenCollectedHeap(GenCollectorPolicy *policy) :
   SharedHeap(policy),
+  _rem_set(NULL),
   _gen_policy(policy),
   _gen_process_roots_tasks(new SubTasksDone(GCH_PS_NumElements)),
   _full_collections_completed(0)
@@ -465,7 +466,7 @@
           // atomic wrt other collectors in this configuration, we
           // are guaranteed to have empty discovered ref lists.
           if (rp->discovery_is_atomic()) {
-            rp->enable_discovery(true /*verify_disabled*/, true /*verify_no_refs*/);
+            rp->enable_discovery();
             rp->setup_policy(do_clear_all_soft_refs);
           } else {
             // collect() below will enable discovery as appropriate
--- a/hotspot/src/share/vm/memory/genCollectedHeap.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/memory/genCollectedHeap.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2014, Oracle and/or its affiliates. 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
@@ -66,6 +66,9 @@
   Generation* _gens[max_gens];
   GenerationSpec** _gen_specs;
 
+  // The singleton Gen Remembered Set.
+  GenRemSet* _rem_set;
+
   // The generational collector policy.
   GenCollectorPolicy* _gen_policy;
 
@@ -383,6 +386,10 @@
     return _n_gens;
   }
 
+  // This function returns the "GenRemSet" object that allows us to scan
+  // generations in a fully generational heap.
+  GenRemSet* rem_set() { return _rem_set; }
+
   // Convenience function to be used in situations where the heap type can be
   // asserted to be this type.
   static GenCollectedHeap* heap();
--- a/hotspot/src/share/vm/memory/genOopClosures.inline.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/memory/genOopClosures.inline.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2014, Oracle and/or its affiliates. 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
@@ -44,7 +44,7 @@
   _gen_boundary = _gen->reserved().start();
   // Barrier set for the heap, must be set after heap is initialized
   if (_rs == NULL) {
-    GenRemSet* rs = SharedHeap::heap()->rem_set();
+    GenRemSet* rs = GenCollectedHeap::heap()->rem_set();
     _rs = (CardTableRS*)rs;
   }
 }
--- a/hotspot/src/share/vm/memory/generation.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/memory/generation.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -297,7 +297,7 @@
 
 void Generation::younger_refs_in_space_iterate(Space* sp,
                                                OopsInGenClosure* cl) {
-  GenRemSet* rs = SharedHeap::heap()->rem_set();
+  GenRemSet* rs = GenCollectedHeap::heap()->rem_set();
   rs->younger_refs_in_space_iterate(sp, cl);
 }
 
--- a/hotspot/src/share/vm/memory/referenceProcessor.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/memory/referenceProcessor.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -68,10 +68,10 @@
   _pending_list_uses_discovered_field = JDK_Version::current().pending_list_uses_discovered_field();
 }
 
-void ReferenceProcessor::enable_discovery(bool verify_disabled, bool check_no_refs) {
+void ReferenceProcessor::enable_discovery(bool check_no_refs) {
 #ifdef ASSERT
   // Verify that we're not currently discovering refs
-  assert(!verify_disabled || !_discovering_refs, "nested call?");
+  assert(!_discovering_refs, "nested call?");
 
   if (check_no_refs) {
     // Verify that the discovered lists are empty
@@ -963,52 +963,6 @@
   return total_list_count;
 }
 
-void ReferenceProcessor::clean_up_discovered_references() {
-  // loop over the lists
-  for (uint i = 0; i < _max_num_q * number_of_subclasses_of_ref(); i++) {
-    if (TraceReferenceGC && PrintGCDetails && ((i % _max_num_q) == 0)) {
-      gclog_or_tty->print_cr(
-        "\nScrubbing %s discovered list of Null referents",
-        list_name(i));
-    }
-    clean_up_discovered_reflist(_discovered_refs[i]);
-  }
-}
-
-void ReferenceProcessor::clean_up_discovered_reflist(DiscoveredList& refs_list) {
-  assert(!discovery_is_atomic(), "Else why call this method?");
-  DiscoveredListIterator iter(refs_list, NULL, NULL);
-  while (iter.has_next()) {
-    iter.load_ptrs(DEBUG_ONLY(true /* allow_null_referent */));
-    oop next = java_lang_ref_Reference::next(iter.obj());
-    assert(next->is_oop_or_null(), err_msg("Expected an oop or NULL for next field at " PTR_FORMAT, p2i(next)));
-    // If referent has been cleared or Reference is not active,
-    // drop it.
-    if (iter.referent() == NULL || next != NULL) {
-      debug_only(
-        if (PrintGCDetails && TraceReferenceGC) {
-          gclog_or_tty->print_cr("clean_up_discovered_list: Dropping Reference: "
-            INTPTR_FORMAT " with next field: " INTPTR_FORMAT
-            " and referent: " INTPTR_FORMAT,
-            (void *)iter.obj(), (void *)next, (void *)iter.referent());
-        }
-      )
-      // Remove Reference object from list
-      iter.remove();
-      iter.move_to_next();
-    } else {
-      iter.next();
-    }
-  }
-  NOT_PRODUCT(
-    if (PrintGCDetails && TraceReferenceGC) {
-      gclog_or_tty->print(
-        " Removed %d Refs with NULL referents out of %d discovered Refs",
-        iter.removed(), iter.processed());
-    }
-  )
-}
-
 inline DiscoveredList* ReferenceProcessor::get_discovered_list(ReferenceType rt) {
   uint id = 0;
   // Determine the queue index to use for this object.
--- a/hotspot/src/share/vm/memory/referenceProcessor.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/memory/referenceProcessor.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -353,19 +353,6 @@
                                       GCTimer*           gc_timer,
                                       GCId               gc_id);
 
-  // Delete entries in the discovered lists that have
-  // either a null referent or are not active. Such
-  // Reference objects can result from the clearing
-  // or enqueueing of Reference objects concurrent
-  // with their discovery by a (concurrent) collector.
-  // For a definition of "active" see java.lang.ref.Reference;
-  // Refs are born active, become inactive when enqueued,
-  // and never become active again. The state of being
-  // active is encoded as follows: A Ref is active
-  // if and only if its "next" field is NULL.
-  void clean_up_discovered_references();
-  void clean_up_discovered_reflist(DiscoveredList& refs_list);
-
   // Returns the name of the discovered reference list
   // occupying the i / _num_q slot.
   const char* list_name(uint i);
@@ -439,7 +426,7 @@
   void      set_span(MemRegion span) { _span = span; }
 
   // start and stop weak ref discovery
-  void enable_discovery(bool verify_disabled, bool check_no_refs);
+  void enable_discovery(bool check_no_refs = true);
   void disable_discovery()  { _discovering_refs = false; }
   bool discovery_enabled()  { return _discovering_refs;  }
 
@@ -517,7 +504,7 @@
 
   ~NoRefDiscovery() {
     if (_was_discovering_refs) {
-      _rp->enable_discovery(true /*verify_disabled*/, false /*check_no_refs*/);
+      _rp->enable_discovery(false /*check_no_refs*/);
     }
   }
 };
--- a/hotspot/src/share/vm/memory/sharedHeap.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/memory/sharedHeap.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -58,7 +58,6 @@
 SharedHeap::SharedHeap(CollectorPolicy* policy_) :
   CollectedHeap(),
   _collector_policy(policy_),
-  _rem_set(NULL),
   _strong_roots_scope(NULL),
   _strong_roots_parity(0),
   _process_strong_tasks(new SubTasksDone(SH_PS_NumElements)),
--- a/hotspot/src/share/vm/memory/sharedHeap.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/memory/sharedHeap.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2014, Oracle and/or its affiliates. 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
@@ -114,10 +114,6 @@
   // set the static pointer "_sh" to that instance.
   static SharedHeap* _sh;
 
-  // and the Gen Remembered Set, at least one good enough to scan the perm
-  // gen.
-  GenRemSet* _rem_set;
-
   // A gc policy, controls global gc resource issues
   CollectorPolicy *_collector_policy;
 
@@ -152,10 +148,6 @@
   // Initialization of ("weak") reference processing support
   virtual void ref_processing_init();
 
-  // This function returns the "GenRemSet" object that allows us to scan
-  // generations in a fully generational heap.
-  GenRemSet* rem_set() { return _rem_set; }
-
   // Iteration functions.
   void oop_iterate(ExtendedOopClosure* cl) = 0;
 
--- a/hotspot/src/share/vm/runtime/arguments.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/runtime/arguments.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -925,9 +925,9 @@
                     "Warning: support for %s was removed in %s\n",
                     fuzzy_matched->_name,
                     version);
-      }
     }
   }
+  }
 
   // allow for commandline "commenting out" options like -XX:#+Verbose
   return arg[0] == '#';
@@ -1382,41 +1382,24 @@
   if (FLAG_IS_DEFAULT(SurvivorRatio) && MaxTenuringThreshold == 0) {
     FLAG_SET_ERGO(uintx, SurvivorRatio, MAX2((uintx)1024, SurvivorRatio));
   }
-  // If OldPLABSize is set and CMSParPromoteBlocksToClaim is not,
-  // set CMSParPromoteBlocksToClaim equal to OldPLABSize.
-  // This is done in order to make ParNew+CMS configuration to work
-  // with YoungPLABSize and OldPLABSize options.
-  // See CR 6362902.
-  if (!FLAG_IS_DEFAULT(OldPLABSize)) {
-    if (FLAG_IS_DEFAULT(CMSParPromoteBlocksToClaim)) {
-      // OldPLABSize is not the default value but CMSParPromoteBlocksToClaim
-      // is.  In this situation let CMSParPromoteBlocksToClaim follow
-      // the value (either from the command line or ergonomics) of
-      // OldPLABSize.  Following OldPLABSize is an ergonomics decision.
-      FLAG_SET_ERGO(uintx, CMSParPromoteBlocksToClaim, OldPLABSize);
+
+  // OldPLABSize is interpreted in CMS as not the size of the PLAB in words,
+  // but rather the number of free blocks of a given size that are used when
+  // replenishing the local per-worker free list caches.
+  if (FLAG_IS_DEFAULT(OldPLABSize)) {
+    if (!FLAG_IS_DEFAULT(ResizeOldPLAB) && !ResizeOldPLAB) {
+      // OldPLAB sizing manually turned off: Use a larger default setting,
+      // unless it was manually specified. This is because a too-low value
+      // will slow down scavenges.
+      FLAG_SET_ERGO(uintx, OldPLABSize, CFLS_LAB::_default_static_old_plab_size); // default value before 6631166
     } else {
-      // OldPLABSize and CMSParPromoteBlocksToClaim are both set.
-      // CMSParPromoteBlocksToClaim is a collector-specific flag, so
-      // we'll let it to take precedence.
-      jio_fprintf(defaultStream::error_stream(),
-                  "Both OldPLABSize and CMSParPromoteBlocksToClaim"
-                  " options are specified for the CMS collector."
-                  " CMSParPromoteBlocksToClaim will take precedence.\n");
+      FLAG_SET_DEFAULT(OldPLABSize, CFLS_LAB::_default_dynamic_old_plab_size); // old CMSParPromoteBlocksToClaim default
     }
   }
-  if (!FLAG_IS_DEFAULT(ResizeOldPLAB) && !ResizeOldPLAB) {
-    // OldPLAB sizing manually turned off: Use a larger default setting,
-    // unless it was manually specified. This is because a too-low value
-    // will slow down scavenges.
-    if (FLAG_IS_DEFAULT(CMSParPromoteBlocksToClaim)) {
-      FLAG_SET_ERGO(uintx, CMSParPromoteBlocksToClaim, 50); // default value before 6631166
-    }
-  }
-  // Overwrite OldPLABSize which is the variable we will internally use everywhere.
-  FLAG_SET_ERGO(uintx, OldPLABSize, CMSParPromoteBlocksToClaim);
+
   // If either of the static initialization defaults have changed, note this
   // modification.
-  if (!FLAG_IS_DEFAULT(CMSParPromoteBlocksToClaim) || !FLAG_IS_DEFAULT(OldPLABWeight)) {
+  if (!FLAG_IS_DEFAULT(OldPLABSize) || !FLAG_IS_DEFAULT(OldPLABWeight)) {
     CFLS_LAB::modify_initialization(OldPLABSize, OldPLABWeight);
   }
   if (PrintGCDetails && Verbose) {
@@ -3225,52 +3208,6 @@
         FLAG_SET_CMDLINE(bool, NeverTenure, false);
         FLAG_SET_CMDLINE(bool, AlwaysTenure, false);
       }
-    } else if (match_option(option, "-XX:+CMSPermGenSweepingEnabled") ||
-               match_option(option, "-XX:-CMSPermGenSweepingEnabled")) {
-      jio_fprintf(defaultStream::error_stream(),
-        "Please use CMSClassUnloadingEnabled in place of "
-        "CMSPermGenSweepingEnabled in the future\n");
-    } else if (match_option(option, "-XX:+UseGCTimeLimit")) {
-      FLAG_SET_CMDLINE(bool, UseGCOverheadLimit, true);
-      jio_fprintf(defaultStream::error_stream(),
-        "Please use -XX:+UseGCOverheadLimit in place of "
-        "-XX:+UseGCTimeLimit in the future\n");
-    } else if (match_option(option, "-XX:-UseGCTimeLimit")) {
-      FLAG_SET_CMDLINE(bool, UseGCOverheadLimit, false);
-      jio_fprintf(defaultStream::error_stream(),
-        "Please use -XX:-UseGCOverheadLimit in place of "
-        "-XX:-UseGCTimeLimit in the future\n");
-    // The TLE options are for compatibility with 1.3 and will be
-    // removed without notice in a future release.  These options
-    // are not to be documented.
-    } else if (match_option(option, "-XX:MaxTLERatio=", &tail)) {
-      // No longer used.
-    } else if (match_option(option, "-XX:+ResizeTLE")) {
-      FLAG_SET_CMDLINE(bool, ResizeTLAB, true);
-    } else if (match_option(option, "-XX:-ResizeTLE")) {
-      FLAG_SET_CMDLINE(bool, ResizeTLAB, false);
-    } else if (match_option(option, "-XX:+PrintTLE")) {
-      FLAG_SET_CMDLINE(bool, PrintTLAB, true);
-    } else if (match_option(option, "-XX:-PrintTLE")) {
-      FLAG_SET_CMDLINE(bool, PrintTLAB, false);
-    } else if (match_option(option, "-XX:TLEFragmentationRatio=", &tail)) {
-      // No longer used.
-    } else if (match_option(option, "-XX:TLESize=", &tail)) {
-      julong long_tlab_size = 0;
-      ArgsRange errcode = parse_memory_size(tail, &long_tlab_size, 1);
-      if (errcode != arg_in_range) {
-        jio_fprintf(defaultStream::error_stream(),
-                    "Invalid TLAB size: %s\n", option->optionString);
-        describe_range_error(errcode);
-        return JNI_EINVAL;
-      }
-      FLAG_SET_CMDLINE(uintx, TLABSize, long_tlab_size);
-    } else if (match_option(option, "-XX:TLEThreadRatio=", &tail)) {
-      // No longer used.
-    } else if (match_option(option, "-XX:+UseTLE")) {
-      FLAG_SET_CMDLINE(bool, UseTLAB, true);
-    } else if (match_option(option, "-XX:-UseTLE")) {
-      FLAG_SET_CMDLINE(bool, UseTLAB, false);
     } else if (match_option(option, "-XX:+DisplayVMOutputToStderr")) {
       FLAG_SET_CMDLINE(bool, DisplayVMOutputToStdout, false);
       FLAG_SET_CMDLINE(bool, DisplayVMOutputToStderr, true);
@@ -3294,44 +3231,6 @@
       // disable scavenge before parallel mark-compact
       FLAG_SET_CMDLINE(bool, ScavengeBeforeFullGC, false);
 #endif
-    } else if (match_option(option, "-XX:CMSParPromoteBlocksToClaim=", &tail)) {
-      julong cms_blocks_to_claim = (julong)atol(tail);
-      FLAG_SET_CMDLINE(uintx, CMSParPromoteBlocksToClaim, cms_blocks_to_claim);
-      jio_fprintf(defaultStream::error_stream(),
-        "Please use -XX:OldPLABSize in place of "
-        "-XX:CMSParPromoteBlocksToClaim in the future\n");
-    } else if (match_option(option, "-XX:ParCMSPromoteBlocksToClaim=", &tail)) {
-      julong cms_blocks_to_claim = (julong)atol(tail);
-      FLAG_SET_CMDLINE(uintx, CMSParPromoteBlocksToClaim, cms_blocks_to_claim);
-      jio_fprintf(defaultStream::error_stream(),
-        "Please use -XX:OldPLABSize in place of "
-        "-XX:ParCMSPromoteBlocksToClaim in the future\n");
-    } else if (match_option(option, "-XX:ParallelGCOldGenAllocBufferSize=", &tail)) {
-      julong old_plab_size = 0;
-      ArgsRange errcode = parse_memory_size(tail, &old_plab_size, 1);
-      if (errcode != arg_in_range) {
-        jio_fprintf(defaultStream::error_stream(),
-                    "Invalid old PLAB size: %s\n", option->optionString);
-        describe_range_error(errcode);
-        return JNI_EINVAL;
-      }
-      FLAG_SET_CMDLINE(uintx, OldPLABSize, old_plab_size);
-      jio_fprintf(defaultStream::error_stream(),
-                  "Please use -XX:OldPLABSize in place of "
-                  "-XX:ParallelGCOldGenAllocBufferSize in the future\n");
-    } else if (match_option(option, "-XX:ParallelGCToSpaceAllocBufferSize=", &tail)) {
-      julong young_plab_size = 0;
-      ArgsRange errcode = parse_memory_size(tail, &young_plab_size, 1);
-      if (errcode != arg_in_range) {
-        jio_fprintf(defaultStream::error_stream(),
-                    "Invalid young PLAB size: %s\n", option->optionString);
-        describe_range_error(errcode);
-        return JNI_EINVAL;
-      }
-      FLAG_SET_CMDLINE(uintx, YoungPLABSize, young_plab_size);
-      jio_fprintf(defaultStream::error_stream(),
-                  "Please use -XX:YoungPLABSize in place of "
-                  "-XX:ParallelGCToSpaceAllocBufferSize in the future\n");
     } else if (match_option(option, "-XX:CMSMarkStackSize=", &tail) ||
                match_option(option, "-XX:G1MarkStackSize=", &tail)) {
       julong stack_size = 0;
@@ -3342,6 +3241,9 @@
         describe_range_error(errcode);
         return JNI_EINVAL;
       }
+      jio_fprintf(defaultStream::error_stream(),
+        "Please use -XX:MarkStackSize in place of "
+        "-XX:CMSMarkStackSize or -XX:G1MarkStackSize in the future\n");
       FLAG_SET_CMDLINE(uintx, MarkStackSize, stack_size);
     } else if (match_option(option, "-XX:CMSMarkStackSizeMax=", &tail)) {
       julong max_stack_size = 0;
@@ -3353,6 +3255,9 @@
         describe_range_error(errcode);
         return JNI_EINVAL;
       }
+      jio_fprintf(defaultStream::error_stream(),
+         "Please use -XX:MarkStackSizeMax in place of "
+         "-XX:CMSMarkStackSizeMax in the future\n");
       FLAG_SET_CMDLINE(uintx, MarkStackSizeMax, max_stack_size);
     } else if (match_option(option, "-XX:ParallelMarkingThreads=", &tail) ||
                match_option(option, "-XX:ParallelCMSThreads=", &tail)) {
@@ -3362,6 +3267,9 @@
                     "Invalid concurrent threads: %s\n", option->optionString);
         return JNI_EINVAL;
       }
+      jio_fprintf(defaultStream::error_stream(),
+        "Please use -XX:ConcGCThreads in place of "
+        "-XX:ParallelMarkingThreads or -XX:ParallelCMSThreads in the future\n");
       FLAG_SET_CMDLINE(uintx, ConcGCThreads, conc_threads);
     } else if (match_option(option, "-XX:MaxDirectMemorySize=", &tail)) {
       julong max_direct_memory_size = 0;
--- a/hotspot/src/share/vm/runtime/globals.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/runtime/globals.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -1472,7 +1472,8 @@
           "Size of young gen promotion LAB's (in HeapWords)")               \
                                                                             \
   product(uintx, OldPLABSize, 1024,                                         \
-          "Size of old gen promotion LAB's (in HeapWords)")                 \
+          "Size of old gen promotion LAB's (in HeapWords), or Number        \
+          of blocks to attempt to claim when refilling CMS LAB's")          \
                                                                             \
   product(uintx, GCTaskTimeStampEntries, 200,                               \
           "Number of time stamp entries per gc worker thread")              \
@@ -1583,14 +1584,10 @@
           "The number of cards in each chunk of the parallel chunks used "  \
           "during card table scanning")                                     \
                                                                             \
-  product(uintx, CMSParPromoteBlocksToClaim, 16,                            \
-          "Number of blocks to attempt to claim when refilling CMS LAB's "  \
-          "for parallel GC")                                                \
-                                                                            \
   product(uintx, OldPLABWeight, 50,                                         \
           "Percentage (0-100) used to weight the current sample when "      \
           "computing exponentially decaying average for resizing "          \
-          "CMSParPromoteBlocksToClaim")                                     \
+          "OldPLABSize")                                                    \
                                                                             \
   product(bool, ResizeOldPLAB, true,                                        \
           "Dynamically resize (old gen) promotion LAB's")                   \
--- a/hotspot/src/share/vm/runtime/os.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/runtime/os.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -1588,6 +1588,11 @@
   return res;
 }
 
+void os::pretouch_memory(char* start, char* end) {
+  for (volatile char *p = start; p < end; p += os::vm_page_size()) {
+    *p = 0;
+  }
+}
 
 char* os::map_memory(int fd, const char* file_name, size_t file_offset,
                            char *addr, size_t bytes, bool read_only,
--- a/hotspot/src/share/vm/runtime/os.hpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/runtime/os.hpp	Tue Dec 23 14:03:15 2014 +0100
@@ -311,6 +311,12 @@
   static bool   uncommit_memory(char* addr, size_t bytes);
   static bool   release_memory(char* addr, size_t bytes);
 
+  // Touch memory pages that cover the memory range from start to end (exclusive)
+  // to make the OS back the memory range with actual memory.
+  // Current implementation may not touch the last page if unaligned addresses
+  // are passed.
+  static void   pretouch_memory(char* start, char* end);
+
   enum ProtType { MEM_PROT_NONE, MEM_PROT_READ, MEM_PROT_RW, MEM_PROT_RWX };
   static bool   protect_memory(char* addr, size_t bytes, ProtType prot,
                                bool is_committed = true);
--- a/hotspot/src/share/vm/runtime/virtualspace.cpp	Mon Dec 22 14:35:40 2014 -0800
+++ b/hotspot/src/share/vm/runtime/virtualspace.cpp	Tue Dec 23 14:03:15 2014 +0100
@@ -615,19 +615,7 @@
   }
 
   if (pre_touch || AlwaysPreTouch) {
-    int vm_ps = os::vm_page_size();
-    for (char* curr = previous_high;
-         curr < unaligned_new_high;
-         curr += vm_ps) {
-      // Note the use of a write here; originally we tried just a read, but
-      // since the value read was unused, the optimizer removed the read.
-      // If we ever have a concurrent touchahead thread, we'll want to use
-      // a read, to avoid the potential of overwriting data (if a mutator
-      // thread beats the touchahead thread to a page).  There are various
-      // ways of making sure this read is not optimized away: for example,
-      // generating the code for a read procedure at runtime.
-      *curr = 0;
-    }
+    os::pretouch_memory(previous_high, unaligned_new_high);
   }
 
   _high += bytes;
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/gc/TestSmallHeap.java	Tue Dec 23 14:03:15 2014 +0100
@@ -0,0 +1,59 @@
+/*
+ * Copyright (c) 2014, Oracle and/or its affiliates. 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 Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/**
+ * @test TestSmallHeap
+ * @bug 8067438
+ * @requires vm.gc=="null"
+ * @summary Verify that starting the VM with a small heap works
+ * @library /testlibrary
+ * @run main/othervm -Xmx4m -XX:+UseParallelGC TestSmallHeap
+ * @run main/othervm -Xmx4m -XX:+UseSerialGC TestSmallHeap
+ * @run main/othervm -Xmx4m -XX:+UseG1GC TestSmallHeap
+ * @run main/othervm -Xmx4m -XX:+UseConcMarkSweepGC -XX:CMSMarkStackSizeMax=1032 TestSmallHeap
+ *
+ * Note: It would be nice to verify the minimal supported heap size here,
+ * but that turns out to be quite tricky since we align the heap size based
+ * on the card table size. And the card table size is aligned based on the
+ * minimal pages size provided by the os. This means that on most platforms,
+ * where the minimal page size is 4k, we get a minimal heap size of 2m but
+ * on Solaris/Sparc we have a page size of 8k and get a minimal heap size
+ * of 8m.
+ * There is also no check in the VM for verifying that the maximum heap size
+ * is larger than the supported minimal heap size. This means that specifying
+ * -Xmx1m on the command line is fine but will give a heap of 2m (or 4m).
+ * To work around these rather strange behaviors this test uses 4m for all
+ * platforms.
+ */
+
+import sun.management.ManagementFactoryHelper;
+import static com.oracle.java.testlibrary.Asserts.*;
+
+public class TestSmallHeap {
+
+    public static void main(String[] args) {
+        String maxHeap = ManagementFactoryHelper.getDiagnosticMXBean().getVMOption("MaxHeapSize").getValue();
+        String expectedMaxHeap = "4194304";
+        assertEQ(maxHeap, expectedMaxHeap);
+    }
+}