1 /*
2 * Copyright (c) 2001, 2018, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #include "precompiled.hpp"
26 #include "classfile/metadataOnStackMark.hpp"
27 #include "classfile/symbolTable.hpp"
28 #include "code/codeCache.hpp"
29 #include "gc/g1/concurrentMarkThread.inline.hpp"
30 #include "gc/g1/g1CollectedHeap.inline.hpp"
31 #include "gc/g1/g1CollectorState.hpp"
32 #include "gc/g1/g1ConcurrentMark.inline.hpp"
33 #include "gc/g1/g1OopClosures.inline.hpp"
34 #include "gc/g1/g1Policy.hpp"
35 #include "gc/g1/g1RegionMarkStatsCache.inline.hpp"
36 #include "gc/g1/g1StringDedup.hpp"
37 #include "gc/g1/heapRegion.inline.hpp"
38 #include "gc/g1/heapRegionRemSet.hpp"
39 #include "gc/g1/heapRegionSet.inline.hpp"
40 #include "gc/shared/adaptiveSizePolicy.hpp"
41 #include "gc/shared/gcId.hpp"
42 #include "gc/shared/gcTimer.hpp"
43 #include "gc/shared/gcTrace.hpp"
44 #include "gc/shared/gcTraceTime.inline.hpp"
45 #include "gc/shared/genOopClosures.inline.hpp"
46 #include "gc/shared/referencePolicy.hpp"
47 #include "gc/shared/strongRootsScope.hpp"
48 #include "gc/shared/suspendibleThreadSet.hpp"
49 #include "gc/shared/taskqueue.inline.hpp"
50 #include "gc/shared/vmGCOperations.hpp"
51 #include "gc/shared/weakProcessor.hpp"
52 #include "include/jvm.h"
53 #include "logging/log.hpp"
54 #include "memory/allocation.hpp"
55 #include "memory/resourceArea.hpp"
56 #include "oops/access.inline.hpp"
57 #include "oops/oop.inline.hpp"
58 #include "runtime/atomic.hpp"
59 #include "runtime/handles.inline.hpp"
60 #include "runtime/java.hpp"
61 #include "runtime/prefetch.inline.hpp"
62 #include "services/memTracker.hpp"
63 #include "utilities/align.hpp"
64 #include "utilities/growableArray.hpp"
65
66 bool G1CMBitMapClosure::do_addr(HeapWord* const addr) {
67 assert(addr < _cm->finger(), "invariant");
68 assert(addr >= _task->finger(), "invariant");
69
70 // We move that task's local finger along.
71 _task->move_finger_to(addr);
72
73 _task->scan_task_entry(G1TaskQueueEntry::from_oop(oop(addr)));
74 // we only partially drain the local queue and global stack
75 _task->drain_local_queue(true);
76 _task->drain_global_stack(true);
77
78 // if the has_aborted flag has been raised, we need to bail out of
79 // the iteration
80 return !_task->has_aborted();
81 }
82
83 G1CMMarkStack::G1CMMarkStack() :
84 _max_chunk_capacity(0),
85 _base(NULL),
86 _chunk_capacity(0) {
87 set_empty();
88 }
89
90 bool G1CMMarkStack::resize(size_t new_capacity) {
91 assert(is_empty(), "Only resize when stack is empty.");
92 assert(new_capacity <= _max_chunk_capacity,
93 "Trying to resize stack to " SIZE_FORMAT " chunks when the maximum is " SIZE_FORMAT, new_capacity, _max_chunk_capacity);
94
95 TaskQueueEntryChunk* new_base = MmapArrayAllocator<TaskQueueEntryChunk>::allocate_or_null(new_capacity, mtGC);
96
97 if (new_base == NULL) {
98 log_warning(gc)("Failed to reserve memory for new overflow mark stack with " SIZE_FORMAT " chunks and size " SIZE_FORMAT "B.", new_capacity, new_capacity * sizeof(TaskQueueEntryChunk));
99 return false;
100 }
101 // Release old mapping.
102 if (_base != NULL) {
103 MmapArrayAllocator<TaskQueueEntryChunk>::free(_base, _chunk_capacity);
104 }
105
106 _base = new_base;
107 _chunk_capacity = new_capacity;
108 set_empty();
109
110 return true;
111 }
112
113 size_t G1CMMarkStack::capacity_alignment() {
114 return (size_t)lcm(os::vm_allocation_granularity(), sizeof(TaskQueueEntryChunk)) / sizeof(G1TaskQueueEntry);
115 }
116
117 bool G1CMMarkStack::initialize(size_t initial_capacity, size_t max_capacity) {
118 guarantee(_max_chunk_capacity == 0, "G1CMMarkStack already initialized.");
119
120 size_t const TaskEntryChunkSizeInVoidStar = sizeof(TaskQueueEntryChunk) / sizeof(G1TaskQueueEntry);
121
122 _max_chunk_capacity = align_up(max_capacity, capacity_alignment()) / TaskEntryChunkSizeInVoidStar;
123 size_t initial_chunk_capacity = align_up(initial_capacity, capacity_alignment()) / TaskEntryChunkSizeInVoidStar;
124
125 guarantee(initial_chunk_capacity <= _max_chunk_capacity,
126 "Maximum chunk capacity " SIZE_FORMAT " smaller than initial capacity " SIZE_FORMAT,
127 _max_chunk_capacity,
128 initial_chunk_capacity);
129
130 log_debug(gc)("Initialize mark stack with " SIZE_FORMAT " chunks, maximum " SIZE_FORMAT,
131 initial_chunk_capacity, _max_chunk_capacity);
132
133 return resize(initial_chunk_capacity);
134 }
135
136 void G1CMMarkStack::expand() {
137 if (_chunk_capacity == _max_chunk_capacity) {
138 log_debug(gc)("Can not expand overflow mark stack further, already at maximum capacity of " SIZE_FORMAT " chunks.", _chunk_capacity);
139 return;
140 }
141 size_t old_capacity = _chunk_capacity;
142 // Double capacity if possible
143 size_t new_capacity = MIN2(old_capacity * 2, _max_chunk_capacity);
144
145 if (resize(new_capacity)) {
146 log_debug(gc)("Expanded mark stack capacity from " SIZE_FORMAT " to " SIZE_FORMAT " chunks",
147 old_capacity, new_capacity);
148 } else {
149 log_warning(gc)("Failed to expand mark stack capacity from " SIZE_FORMAT " to " SIZE_FORMAT " chunks",
150 old_capacity, new_capacity);
151 }
152 }
153
154 G1CMMarkStack::~G1CMMarkStack() {
155 if (_base != NULL) {
156 MmapArrayAllocator<TaskQueueEntryChunk>::free(_base, _chunk_capacity);
157 }
158 }
159
160 void G1CMMarkStack::add_chunk_to_list(TaskQueueEntryChunk* volatile* list, TaskQueueEntryChunk* elem) {
161 elem->next = *list;
162 *list = elem;
163 }
164
165 void G1CMMarkStack::add_chunk_to_chunk_list(TaskQueueEntryChunk* elem) {
166 MutexLockerEx x(MarkStackChunkList_lock, Mutex::_no_safepoint_check_flag);
167 add_chunk_to_list(&_chunk_list, elem);
168 _chunks_in_chunk_list++;
169 }
170
171 void G1CMMarkStack::add_chunk_to_free_list(TaskQueueEntryChunk* elem) {
172 MutexLockerEx x(MarkStackFreeList_lock, Mutex::_no_safepoint_check_flag);
173 add_chunk_to_list(&_free_list, elem);
174 }
175
176 G1CMMarkStack::TaskQueueEntryChunk* G1CMMarkStack::remove_chunk_from_list(TaskQueueEntryChunk* volatile* list) {
177 TaskQueueEntryChunk* result = *list;
178 if (result != NULL) {
179 *list = (*list)->next;
180 }
181 return result;
182 }
183
184 G1CMMarkStack::TaskQueueEntryChunk* G1CMMarkStack::remove_chunk_from_chunk_list() {
185 MutexLockerEx x(MarkStackChunkList_lock, Mutex::_no_safepoint_check_flag);
186 TaskQueueEntryChunk* result = remove_chunk_from_list(&_chunk_list);
187 if (result != NULL) {
188 _chunks_in_chunk_list--;
189 }
190 return result;
191 }
192
193 G1CMMarkStack::TaskQueueEntryChunk* G1CMMarkStack::remove_chunk_from_free_list() {
194 MutexLockerEx x(MarkStackFreeList_lock, Mutex::_no_safepoint_check_flag);
195 return remove_chunk_from_list(&_free_list);
196 }
197
198 G1CMMarkStack::TaskQueueEntryChunk* G1CMMarkStack::allocate_new_chunk() {
199 // This dirty read of _hwm is okay because we only ever increase the _hwm in parallel code.
200 // Further this limits _hwm to a value of _chunk_capacity + #threads, avoiding
201 // wraparound of _hwm.
202 if (_hwm >= _chunk_capacity) {
203 return NULL;
204 }
205
206 size_t cur_idx = Atomic::add(1u, &_hwm) - 1;
207 if (cur_idx >= _chunk_capacity) {
208 return NULL;
209 }
210
211 TaskQueueEntryChunk* result = ::new (&_base[cur_idx]) TaskQueueEntryChunk;
212 result->next = NULL;
213 return result;
214 }
215
216 bool G1CMMarkStack::par_push_chunk(G1TaskQueueEntry* ptr_arr) {
217 // Get a new chunk.
218 TaskQueueEntryChunk* new_chunk = remove_chunk_from_free_list();
219
220 if (new_chunk == NULL) {
221 // Did not get a chunk from the free list. Allocate from backing memory.
222 new_chunk = allocate_new_chunk();
223
224 if (new_chunk == NULL) {
225 return false;
226 }
227 }
228
229 Copy::conjoint_memory_atomic(ptr_arr, new_chunk->data, EntriesPerChunk * sizeof(G1TaskQueueEntry));
230
231 add_chunk_to_chunk_list(new_chunk);
232
233 return true;
234 }
235
236 bool G1CMMarkStack::par_pop_chunk(G1TaskQueueEntry* ptr_arr) {
237 TaskQueueEntryChunk* cur = remove_chunk_from_chunk_list();
238
239 if (cur == NULL) {
240 return false;
241 }
242
243 Copy::conjoint_memory_atomic(cur->data, ptr_arr, EntriesPerChunk * sizeof(G1TaskQueueEntry));
244
245 add_chunk_to_free_list(cur);
246 return true;
247 }
248
249 void G1CMMarkStack::set_empty() {
250 _chunks_in_chunk_list = 0;
251 _hwm = 0;
252 _chunk_list = NULL;
253 _free_list = NULL;
254 }
255
256 G1CMRootRegions::G1CMRootRegions() :
257 _survivors(NULL), _cm(NULL), _scan_in_progress(false),
258 _should_abort(false), _claimed_survivor_index(0) { }
259
260 void G1CMRootRegions::init(const G1SurvivorRegions* survivors, G1ConcurrentMark* cm) {
261 _survivors = survivors;
262 _cm = cm;
263 }
264
265 void G1CMRootRegions::prepare_for_scan() {
266 assert(!scan_in_progress(), "pre-condition");
267
268 // Currently, only survivors can be root regions.
269 _claimed_survivor_index = 0;
270 _scan_in_progress = _survivors->regions()->is_nonempty();
271 _should_abort = false;
272 }
273
274 HeapRegion* G1CMRootRegions::claim_next() {
275 if (_should_abort) {
276 // If someone has set the should_abort flag, we return NULL to
277 // force the caller to bail out of their loop.
278 return NULL;
279 }
280
281 // Currently, only survivors can be root regions.
282 const GrowableArray<HeapRegion*>* survivor_regions = _survivors->regions();
283
284 int claimed_index = Atomic::add(1, &_claimed_survivor_index) - 1;
285 if (claimed_index < survivor_regions->length()) {
286 return survivor_regions->at(claimed_index);
287 }
288 return NULL;
289 }
290
291 uint G1CMRootRegions::num_root_regions() const {
292 return (uint)_survivors->regions()->length();
293 }
294
295 void G1CMRootRegions::notify_scan_done() {
296 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
297 _scan_in_progress = false;
298 RootRegionScan_lock->notify_all();
299 }
300
301 void G1CMRootRegions::cancel_scan() {
302 notify_scan_done();
303 }
304
305 void G1CMRootRegions::scan_finished() {
306 assert(scan_in_progress(), "pre-condition");
307
308 // Currently, only survivors can be root regions.
309 if (!_should_abort) {
310 assert(_claimed_survivor_index >= 0, "otherwise comparison is invalid: %d", _claimed_survivor_index);
311 assert((uint)_claimed_survivor_index >= _survivors->length(),
312 "we should have claimed all survivors, claimed index = %u, length = %u",
313 (uint)_claimed_survivor_index, _survivors->length());
314 }
315
316 notify_scan_done();
317 }
318
319 bool G1CMRootRegions::wait_until_scan_finished() {
320 if (!scan_in_progress()) {
321 return false;
322 }
323
324 {
325 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
326 while (scan_in_progress()) {
327 RootRegionScan_lock->wait(Mutex::_no_safepoint_check_flag);
328 }
329 }
330 return true;
331 }
332
333 // Returns the maximum number of workers to be used in a concurrent
334 // phase based on the number of GC workers being used in a STW
335 // phase.
336 static uint scale_concurrent_worker_threads(uint num_gc_workers) {
337 return MAX2((num_gc_workers + 2) / 4, 1U);
338 }
339
340 G1ConcurrentMark::G1ConcurrentMark(G1CollectedHeap* g1h,
341 G1RegionToSpaceMapper* prev_bitmap_storage,
342 G1RegionToSpaceMapper* next_bitmap_storage) :
343 // _cm_thread set inside the constructor
344 _g1h(g1h),
345 _completed_initialization(false),
346
347 _mark_bitmap_1(),
348 _mark_bitmap_2(),
349 _prev_mark_bitmap(&_mark_bitmap_1),
350 _next_mark_bitmap(&_mark_bitmap_2),
351
352 _heap(_g1h->reserved_region()),
353
354 _root_regions(),
355
356 _global_mark_stack(),
357
358 // _finger set in set_non_marking_state
359
360 _worker_id_offset(DirtyCardQueueSet::num_par_ids() + G1ConcRefinementThreads),
361 _max_num_tasks(ParallelGCThreads),
362 // _num_active_tasks set in set_non_marking_state()
363 // _tasks set inside the constructor
364
365 _task_queues(new G1CMTaskQueueSet((int) _max_num_tasks)),
366 _terminator(ParallelTaskTerminator((int) _max_num_tasks, _task_queues)),
367
368 _first_overflow_barrier_sync(),
369 _second_overflow_barrier_sync(),
370
371 _has_overflown(false),
372 _concurrent(false),
373 _has_aborted(false),
374 _restart_for_overflow(false),
375 _gc_timer_cm(new (ResourceObj::C_HEAP, mtGC) ConcurrentGCTimer()),
376 _gc_tracer_cm(new (ResourceObj::C_HEAP, mtGC) G1OldTracer()),
377
378 // _verbose_level set below
379
380 _init_times(),
381 _remark_times(),
382 _remark_mark_times(),
383 _remark_weak_ref_times(),
384 _cleanup_times(),
385 _total_cleanup_time(0.0),
386
387 _accum_task_vtime(NULL),
388
389 _concurrent_workers(NULL),
390 _num_concurrent_workers(0),
391 _max_concurrent_workers(0),
392
393 _region_mark_stats(NEW_C_HEAP_ARRAY(G1RegionMarkStats, _g1h->max_regions(), mtGC)),
394 _top_at_rebuild_starts(NEW_C_HEAP_ARRAY(HeapWord*, _g1h->max_regions(), mtGC))
395 {
396 _mark_bitmap_1.initialize(g1h->reserved_region(), prev_bitmap_storage);
397 _mark_bitmap_2.initialize(g1h->reserved_region(), next_bitmap_storage);
398
399 // Create & start ConcurrentMark thread.
400 _cm_thread = new ConcurrentMarkThread(this);
401 if (_cm_thread->osthread() == NULL) {
402 vm_shutdown_during_initialization("Could not create ConcurrentMarkThread");
403 }
404
405 assert(CGC_lock != NULL, "CGC_lock must be initialized");
406
407 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set();
408 satb_qs.set_buffer_size(G1SATBBufferSize);
409
410 _root_regions.init(_g1h->survivor(), this);
411
412 if (FLAG_IS_DEFAULT(ConcGCThreads) || ConcGCThreads == 0) {
413 // Calculate the number of concurrent worker threads by scaling
414 // the number of parallel GC threads.
415 uint marking_thread_num = scale_concurrent_worker_threads(ParallelGCThreads);
416 FLAG_SET_ERGO(uint, ConcGCThreads, marking_thread_num);
417 }
418
419 assert(ConcGCThreads > 0, "ConcGCThreads have been set.");
420 if (ConcGCThreads > ParallelGCThreads) {
421 log_warning(gc)("More ConcGCThreads (%u) than ParallelGCThreads (%u).",
422 ConcGCThreads, ParallelGCThreads);
423 return;
424 }
425
426 log_debug(gc)("ConcGCThreads: %u offset %u", ConcGCThreads, _worker_id_offset);
427 log_debug(gc)("ParallelGCThreads: %u", ParallelGCThreads);
428
429 _num_concurrent_workers = ConcGCThreads;
430 _max_concurrent_workers = _num_concurrent_workers;
431
432 _concurrent_workers = new WorkGang("G1 Conc", _max_concurrent_workers, false, true);
433 _concurrent_workers->initialize_workers();
434
435 if (FLAG_IS_DEFAULT(MarkStackSize)) {
436 size_t mark_stack_size =
437 MIN2(MarkStackSizeMax,
438 MAX2(MarkStackSize, (size_t) (_max_concurrent_workers * TASKQUEUE_SIZE)));
439 // Verify that the calculated value for MarkStackSize is in range.
440 // It would be nice to use the private utility routine from Arguments.
441 if (!(mark_stack_size >= 1 && mark_stack_size <= MarkStackSizeMax)) {
442 log_warning(gc)("Invalid value calculated for MarkStackSize (" SIZE_FORMAT "): "
443 "must be between 1 and " SIZE_FORMAT,
444 mark_stack_size, MarkStackSizeMax);
445 return;
446 }
447 FLAG_SET_ERGO(size_t, MarkStackSize, mark_stack_size);
448 } else {
449 // Verify MarkStackSize is in range.
450 if (FLAG_IS_CMDLINE(MarkStackSize)) {
451 if (FLAG_IS_DEFAULT(MarkStackSizeMax)) {
452 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
453 log_warning(gc)("Invalid value specified for MarkStackSize (" SIZE_FORMAT "): "
454 "must be between 1 and " SIZE_FORMAT,
455 MarkStackSize, MarkStackSizeMax);
456 return;
457 }
458 } else if (FLAG_IS_CMDLINE(MarkStackSizeMax)) {
459 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
460 log_warning(gc)("Invalid value specified for MarkStackSize (" SIZE_FORMAT ")"
461 " or for MarkStackSizeMax (" SIZE_FORMAT ")",
462 MarkStackSize, MarkStackSizeMax);
463 return;
464 }
465 }
466 }
467 }
468
469 if (!_global_mark_stack.initialize(MarkStackSize, MarkStackSizeMax)) {
470 vm_exit_during_initialization("Failed to allocate initial concurrent mark overflow mark stack.");
471 }
472
473 _tasks = NEW_C_HEAP_ARRAY(G1CMTask*, _max_num_tasks, mtGC);
474 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_num_tasks, mtGC);
475
476 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail
477 _num_active_tasks = _max_num_tasks;
478
479 for (uint i = 0; i < _max_num_tasks; ++i) {
480 G1CMTaskQueue* task_queue = new G1CMTaskQueue();
481 task_queue->initialize();
482 _task_queues->register_queue(i, task_queue);
483
484 _tasks[i] = new G1CMTask(i, this, task_queue, _region_mark_stats, _g1h->max_regions());
485
486 _accum_task_vtime[i] = 0.0;
487 }
488
489 reset_at_marking_complete();
490 _completed_initialization = true;
491 }
492
493 void G1ConcurrentMark::reset() {
494 _has_aborted = false;
495
496 reset_marking_for_restart();
497
498 // Reset all tasks, since different phases will use different number of active
499 // threads. So, it's easiest to have all of them ready.
500 for (uint i = 0; i < _max_num_tasks; ++i) {
501 _tasks[i]->reset(_next_mark_bitmap);
502 }
503
504 uint max_regions = _g1h->max_regions();
505 for (uint i = 0; i < max_regions; i++) {
506 _top_at_rebuild_starts[i] = NULL;
507 _region_mark_stats[i].clear();
508 }
509 }
510
511 void G1ConcurrentMark::clear_statistics_in_region(uint region_idx) {
512 for (uint j = 0; j < _max_num_tasks; ++j) {
513 _tasks[j]->clear_mark_stats_cache(region_idx);
514 }
515 _top_at_rebuild_starts[region_idx] = NULL;
516 _region_mark_stats[region_idx].clear();
517 }
518
519 void G1ConcurrentMark::clear_statistics(HeapRegion* r) {
520 uint const region_idx = r->hrm_index();
521 if (r->is_humongous()) {
522 assert(r->is_starts_humongous(), "Got humongous continues region here");
523 uint const size_in_regions = (uint)_g1h->humongous_obj_size_in_regions(oop(r->humongous_start_region()->bottom())->size());
524 for (uint j = region_idx; j < (region_idx + size_in_regions); j++) {
525 clear_statistics_in_region(j);
526 }
527 } else {
528 clear_statistics_in_region(region_idx);
529 }
530 }
531
532 void G1ConcurrentMark::humongous_object_eagerly_reclaimed(HeapRegion* r) {
533 assert_at_safepoint_on_vm_thread();
534
535 // Need to clear mark bit of the humongous object.
536 if (_next_mark_bitmap->is_marked(r->bottom())) {
537 _next_mark_bitmap->clear(r->bottom());
538 }
539
540 if (!_g1h->collector_state()->mark_or_rebuild_in_progress()) {
541 return;
542 }
543
544 // Clear any statistics about the region gathered so far.
545 clear_statistics(r);
546 }
547
548 void G1ConcurrentMark::reset_marking_for_restart() {
549 _global_mark_stack.set_empty();
550
551 // Expand the marking stack, if we have to and if we can.
552 if (has_overflown()) {
553 _global_mark_stack.expand();
554
555 uint max_regions = _g1h->max_regions();
556 for (uint i = 0; i < max_regions; i++) {
557 _region_mark_stats[i].clear_during_overflow();
558 }
559 }
560
561 clear_has_overflown();
562 _finger = _heap.start();
563
564 for (uint i = 0; i < _max_num_tasks; ++i) {
565 G1CMTaskQueue* queue = _task_queues->queue(i);
566 queue->set_empty();
567 }
568 }
569
570 void G1ConcurrentMark::set_concurrency(uint active_tasks) {
571 assert(active_tasks <= _max_num_tasks, "we should not have more");
572
573 _num_active_tasks = active_tasks;
574 // Need to update the three data structures below according to the
575 // number of active threads for this phase.
576 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues);
577 _first_overflow_barrier_sync.set_n_workers((int) active_tasks);
578 _second_overflow_barrier_sync.set_n_workers((int) active_tasks);
579 }
580
581 void G1ConcurrentMark::set_concurrency_and_phase(uint active_tasks, bool concurrent) {
582 set_concurrency(active_tasks);
583
584 _concurrent = concurrent;
585
586 if (!concurrent) {
587 // At this point we should be in a STW phase, and completed marking.
588 assert_at_safepoint_on_vm_thread();
589 assert(out_of_regions(),
590 "only way to get here: _finger: " PTR_FORMAT ", _heap_end: " PTR_FORMAT,
591 p2i(_finger), p2i(_heap.end()));
592 }
593 }
594
595 void G1ConcurrentMark::reset_at_marking_complete() {
596 // We set the global marking state to some default values when we're
597 // not doing marking.
598 reset_marking_for_restart();
599 _num_active_tasks = 0;
600 }
601
602 G1ConcurrentMark::~G1ConcurrentMark() {
603 FREE_C_HEAP_ARRAY(HeapWord*, _top_at_rebuild_starts);
604 FREE_C_HEAP_ARRAY(G1RegionMarkStats, _region_mark_stats);
605 // The G1ConcurrentMark instance is never freed.
606 ShouldNotReachHere();
607 }
608
609 class G1ClearBitMapTask : public AbstractGangTask {
610 public:
611 static size_t chunk_size() { return M; }
612
613 private:
614 // Heap region closure used for clearing the given mark bitmap.
615 class G1ClearBitmapHRClosure : public HeapRegionClosure {
616 private:
617 G1CMBitMap* _bitmap;
618 G1ConcurrentMark* _cm;
619 public:
620 G1ClearBitmapHRClosure(G1CMBitMap* bitmap, G1ConcurrentMark* cm) : HeapRegionClosure(), _cm(cm), _bitmap(bitmap) {
621 }
622
623 virtual bool do_heap_region(HeapRegion* r) {
624 size_t const chunk_size_in_words = G1ClearBitMapTask::chunk_size() / HeapWordSize;
625
626 HeapWord* cur = r->bottom();
627 HeapWord* const end = r->end();
628
629 while (cur < end) {
630 MemRegion mr(cur, MIN2(cur + chunk_size_in_words, end));
631 _bitmap->clear_range(mr);
632
633 cur += chunk_size_in_words;
634
635 // Abort iteration if after yielding the marking has been aborted.
636 if (_cm != NULL && _cm->do_yield_check() && _cm->has_aborted()) {
637 return true;
638 }
639 // Repeat the asserts from before the start of the closure. We will do them
640 // as asserts here to minimize their overhead on the product. However, we
641 // will have them as guarantees at the beginning / end of the bitmap
642 // clearing to get some checking in the product.
643 assert(_cm == NULL || _cm->cm_thread()->during_cycle(), "invariant");
644 assert(_cm == NULL || !G1CollectedHeap::heap()->collector_state()->mark_or_rebuild_in_progress(), "invariant");
645 }
646 assert(cur == end, "Must have completed iteration over the bitmap for region %u.", r->hrm_index());
647
648 return false;
649 }
650 };
651
652 G1ClearBitmapHRClosure _cl;
653 HeapRegionClaimer _hr_claimer;
654 bool _suspendible; // If the task is suspendible, workers must join the STS.
655
656 public:
657 G1ClearBitMapTask(G1CMBitMap* bitmap, G1ConcurrentMark* cm, uint n_workers, bool suspendible) :
658 AbstractGangTask("G1 Clear Bitmap"),
659 _cl(bitmap, suspendible ? cm : NULL),
660 _hr_claimer(n_workers),
661 _suspendible(suspendible)
662 { }
663
664 void work(uint worker_id) {
665 SuspendibleThreadSetJoiner sts_join(_suspendible);
666 G1CollectedHeap::heap()->heap_region_par_iterate_from_worker_offset(&_cl, &_hr_claimer, worker_id);
667 }
668
669 bool is_complete() {
670 return _cl.is_complete();
671 }
672 };
673
674 void G1ConcurrentMark::clear_bitmap(G1CMBitMap* bitmap, WorkGang* workers, bool may_yield) {
675 assert(may_yield || SafepointSynchronize::is_at_safepoint(), "Non-yielding bitmap clear only allowed at safepoint.");
676
677 size_t const num_bytes_to_clear = (HeapRegion::GrainBytes * _g1h->num_regions()) / G1CMBitMap::heap_map_factor();
678 size_t const num_chunks = align_up(num_bytes_to_clear, G1ClearBitMapTask::chunk_size()) / G1ClearBitMapTask::chunk_size();
679
680 uint const num_workers = (uint)MIN2(num_chunks, (size_t)workers->active_workers());
681
682 G1ClearBitMapTask cl(bitmap, this, num_workers, may_yield);
683
684 log_debug(gc, ergo)("Running %s with %u workers for " SIZE_FORMAT " work units.", cl.name(), num_workers, num_chunks);
685 workers->run_task(&cl, num_workers);
686 guarantee(!may_yield || cl.is_complete(), "Must have completed iteration when not yielding.");
687 }
688
689 void G1ConcurrentMark::cleanup_for_next_mark() {
690 // Make sure that the concurrent mark thread looks to still be in
691 // the current cycle.
692 guarantee(cm_thread()->during_cycle(), "invariant");
693
694 // We are finishing up the current cycle by clearing the next
695 // marking bitmap and getting it ready for the next cycle. During
696 // this time no other cycle can start. So, let's make sure that this
697 // is the case.
698 guarantee(!_g1h->collector_state()->mark_or_rebuild_in_progress(), "invariant");
699
700 clear_bitmap(_next_mark_bitmap, _concurrent_workers, true);
701
702 // Repeat the asserts from above.
703 guarantee(cm_thread()->during_cycle(), "invariant");
704 guarantee(!_g1h->collector_state()->mark_or_rebuild_in_progress(), "invariant");
705 }
706
707 void G1ConcurrentMark::clear_prev_bitmap(WorkGang* workers) {
708 assert_at_safepoint_on_vm_thread();
709 clear_bitmap(_prev_mark_bitmap, workers, false);
710 }
711
712 class CheckBitmapClearHRClosure : public HeapRegionClosure {
713 G1CMBitMap* _bitmap;
714 public:
715 CheckBitmapClearHRClosure(G1CMBitMap* bitmap) : _bitmap(bitmap) {
716 }
717
718 virtual bool do_heap_region(HeapRegion* r) {
719 // This closure can be called concurrently to the mutator, so we must make sure
720 // that the result of the getNextMarkedWordAddress() call is compared to the
721 // value passed to it as limit to detect any found bits.
722 // end never changes in G1.
723 HeapWord* end = r->end();
724 return _bitmap->get_next_marked_addr(r->bottom(), end) != end;
725 }
726 };
727
728 bool G1ConcurrentMark::next_mark_bitmap_is_clear() {
729 CheckBitmapClearHRClosure cl(_next_mark_bitmap);
730 _g1h->heap_region_iterate(&cl);
731 return cl.is_complete();
732 }
733
734 class NoteStartOfMarkHRClosure : public HeapRegionClosure {
735 public:
736 bool do_heap_region(HeapRegion* r) {
737 r->note_start_of_marking();
738 return false;
739 }
740 };
741
742 void G1ConcurrentMark::pre_initial_mark() {
743 // Initialize marking structures. This has to be done in a STW phase.
744 reset();
745
746 // For each region note start of marking.
747 NoteStartOfMarkHRClosure startcl;
748 _g1h->heap_region_iterate(&startcl);
749 }
750
751
752 void G1ConcurrentMark::post_initial_mark() {
753 // Start Concurrent Marking weak-reference discovery.
754 ReferenceProcessor* rp = _g1h->ref_processor_cm();
755 // enable ("weak") refs discovery
756 rp->enable_discovery();
757 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle
758
759 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
760 // This is the start of the marking cycle, we're expected all
761 // threads to have SATB queues with active set to false.
762 satb_mq_set.set_active_all_threads(true, /* new active value */
763 false /* expected_active */);
764
765 _root_regions.prepare_for_scan();
766
767 // update_g1_committed() will be called at the end of an evac pause
768 // when marking is on. So, it's also called at the end of the
769 // initial-mark pause to update the heap end, if the heap expands
770 // during it. No need to call it here.
771 }
772
773 /*
774 * Notice that in the next two methods, we actually leave the STS
775 * during the barrier sync and join it immediately afterwards. If we
776 * do not do this, the following deadlock can occur: one thread could
777 * be in the barrier sync code, waiting for the other thread to also
778 * sync up, whereas another one could be trying to yield, while also
779 * waiting for the other threads to sync up too.
780 *
781 * Note, however, that this code is also used during remark and in
782 * this case we should not attempt to leave / enter the STS, otherwise
783 * we'll either hit an assert (debug / fastdebug) or deadlock
784 * (product). So we should only leave / enter the STS if we are
785 * operating concurrently.
786 *
787 * Because the thread that does the sync barrier has left the STS, it
788 * is possible to be suspended for a Full GC or an evacuation pause
789 * could occur. This is actually safe, since the entering the sync
790 * barrier is one of the last things do_marking_step() does, and it
791 * doesn't manipulate any data structures afterwards.
792 */
793
794 void G1ConcurrentMark::enter_first_sync_barrier(uint worker_id) {
795 bool barrier_aborted;
796 {
797 SuspendibleThreadSetLeaver sts_leave(concurrent());
798 barrier_aborted = !_first_overflow_barrier_sync.enter();
799 }
800
801 // at this point everyone should have synced up and not be doing any
802 // more work
803
804 if (barrier_aborted) {
805 // If the barrier aborted we ignore the overflow condition and
806 // just abort the whole marking phase as quickly as possible.
807 return;
808 }
809 }
810
811 void G1ConcurrentMark::enter_second_sync_barrier(uint worker_id) {
812 SuspendibleThreadSetLeaver sts_leave(concurrent());
813 _second_overflow_barrier_sync.enter();
814
815 // at this point everything should be re-initialized and ready to go
816 }
817
818 class G1CMConcurrentMarkingTask : public AbstractGangTask {
819 G1ConcurrentMark* _cm;
820 public:
821 void work(uint worker_id) {
822 assert(Thread::current()->is_ConcurrentGC_thread(), "Not a concurrent GC thread");
823 ResourceMark rm;
824
825 double start_vtime = os::elapsedVTime();
826
827 {
828 SuspendibleThreadSetJoiner sts_join;
829
830 assert(worker_id < _cm->active_tasks(), "invariant");
831
832 G1CMTask* task = _cm->task(worker_id);
833 task->record_start_time();
834 if (!_cm->has_aborted()) {
835 do {
836 task->do_marking_step(G1ConcMarkStepDurationMillis,
837 true /* do_termination */,
838 false /* is_serial*/);
839
840 _cm->do_yield_check();
841 } while (!_cm->has_aborted() && task->has_aborted());
842 }
843 task->record_end_time();
844 guarantee(!task->has_aborted() || _cm->has_aborted(), "invariant");
845 }
846
847 double end_vtime = os::elapsedVTime();
848 _cm->update_accum_task_vtime(worker_id, end_vtime - start_vtime);
849 }
850
851 G1CMConcurrentMarkingTask(G1ConcurrentMark* cm) :
852 AbstractGangTask("Concurrent Mark"), _cm(cm) { }
853
854 ~G1CMConcurrentMarkingTask() { }
855 };
856
857 uint G1ConcurrentMark::calc_active_marking_workers() {
858 uint result = 0;
859 if (!UseDynamicNumberOfGCThreads ||
860 (!FLAG_IS_DEFAULT(ConcGCThreads) &&
861 !ForceDynamicNumberOfGCThreads)) {
862 result = _max_concurrent_workers;
863 } else {
864 result =
865 AdaptiveSizePolicy::calc_default_active_workers(_max_concurrent_workers,
866 1, /* Minimum workers */
867 _num_concurrent_workers,
868 Threads::number_of_non_daemon_threads());
869 // Don't scale the result down by scale_concurrent_workers() because
870 // that scaling has already gone into "_max_concurrent_workers".
871 }
872 assert(result > 0 && result <= _max_concurrent_workers,
873 "Calculated number of marking workers must be larger than zero and at most the maximum %u, but is %u",
874 _max_concurrent_workers, result);
875 return result;
876 }
877
878 void G1ConcurrentMark::scan_root_region(HeapRegion* hr, uint worker_id) {
879 // Currently, only survivors can be root regions.
880 assert(hr->next_top_at_mark_start() == hr->bottom(), "invariant");
881 G1RootRegionScanClosure cl(_g1h, this, worker_id);
882
883 const uintx interval = PrefetchScanIntervalInBytes;
884 HeapWord* curr = hr->bottom();
885 const HeapWord* end = hr->top();
886 while (curr < end) {
887 Prefetch::read(curr, interval);
888 oop obj = oop(curr);
889 int size = obj->oop_iterate_size(&cl);
890 assert(size == obj->size(), "sanity");
891 curr += size;
892 }
893 }
894
895 class G1CMRootRegionScanTask : public AbstractGangTask {
896 G1ConcurrentMark* _cm;
897 public:
898 G1CMRootRegionScanTask(G1ConcurrentMark* cm) :
899 AbstractGangTask("G1 Root Region Scan"), _cm(cm) { }
900
901 void work(uint worker_id) {
902 assert(Thread::current()->is_ConcurrentGC_thread(),
903 "this should only be done by a conc GC thread");
904
905 G1CMRootRegions* root_regions = _cm->root_regions();
906 HeapRegion* hr = root_regions->claim_next();
907 while (hr != NULL) {
908 _cm->scan_root_region(hr, worker_id);
909 hr = root_regions->claim_next();
910 }
911 }
912 };
913
914 void G1ConcurrentMark::scan_root_regions() {
915 // scan_in_progress() will have been set to true only if there was
916 // at least one root region to scan. So, if it's false, we
917 // should not attempt to do any further work.
918 if (root_regions()->scan_in_progress()) {
919 assert(!has_aborted(), "Aborting before root region scanning is finished not supported.");
920
921 _num_concurrent_workers = MIN2(calc_active_marking_workers(),
922 // We distribute work on a per-region basis, so starting
923 // more threads than that is useless.
924 root_regions()->num_root_regions());
925 assert(_num_concurrent_workers <= _max_concurrent_workers,
926 "Maximum number of marking threads exceeded");
927
928 G1CMRootRegionScanTask task(this);
929 log_debug(gc, ergo)("Running %s using %u workers for %u work units.",
930 task.name(), _num_concurrent_workers, root_regions()->num_root_regions());
931 _concurrent_workers->run_task(&task, _num_concurrent_workers);
932
933 // It's possible that has_aborted() is true here without actually
934 // aborting the survivor scan earlier. This is OK as it's
935 // mainly used for sanity checking.
936 root_regions()->scan_finished();
937 }
938 }
939
940 void G1ConcurrentMark::concurrent_cycle_start() {
941 _gc_timer_cm->register_gc_start();
942
943 _gc_tracer_cm->report_gc_start(GCCause::_no_gc /* first parameter is not used */, _gc_timer_cm->gc_start());
944
945 _g1h->trace_heap_before_gc(_gc_tracer_cm);
946 }
947
948 void G1ConcurrentMark::concurrent_cycle_end() {
949 _g1h->collector_state()->set_clearing_next_bitmap(false);
950
951 _g1h->trace_heap_after_gc(_gc_tracer_cm);
952
953 if (has_aborted()) {
954 log_info(gc, marking)("Concurrent Mark Abort");
955 _gc_tracer_cm->report_concurrent_mode_failure();
956 }
957
958 _gc_timer_cm->register_gc_end();
959
960 _gc_tracer_cm->report_gc_end(_gc_timer_cm->gc_end(), _gc_timer_cm->time_partitions());
961 }
962
963 void G1ConcurrentMark::mark_from_roots() {
964 _restart_for_overflow = false;
965
966 _num_concurrent_workers = calc_active_marking_workers();
967
968 uint active_workers = MAX2(1U, _num_concurrent_workers);
969
970 // Setting active workers is not guaranteed since fewer
971 // worker threads may currently exist and more may not be
972 // available.
973 active_workers = _concurrent_workers->update_active_workers(active_workers);
974 log_info(gc, task)("Using %u workers of %u for marking", active_workers, _concurrent_workers->total_workers());
975
976 // Parallel task terminator is set in "set_concurrency_and_phase()"
977 set_concurrency_and_phase(active_workers, true /* concurrent */);
978
979 G1CMConcurrentMarkingTask marking_task(this);
980 _concurrent_workers->run_task(&marking_task);
981 print_stats();
982 }
983
984 void G1ConcurrentMark::verify_during_pause(G1HeapVerifier::G1VerifyType type, VerifyOption vo, const char* caller) {
985 G1HeapVerifier* verifier = _g1h->verifier();
986
987 verifier->verify_region_sets_optional();
988
989 if (VerifyDuringGC) {
990 GCTraceTime(Debug, gc, phases) trace(caller, _gc_timer_cm);
991
992 size_t const BufLen = 512;
993 char buffer[BufLen];
994
995 jio_snprintf(buffer, BufLen, "During GC (%s)", caller);
996 verifier->verify(type, vo, buffer);
997 }
998
999 verifier->check_bitmaps(caller);
1000 }
1001
1002 class G1UpdateRemSetTrackingBeforeRebuild : public HeapRegionClosure {
1003 G1CollectedHeap* _g1h;
1004 G1ConcurrentMark* _cm;
1005
1006 uint _num_regions_selected_for_rebuild; // The number of regions actually selected for rebuild.
1007
1008 void update_remset_before_rebuild(HeapRegion * hr) {
1009 G1RemSetTrackingPolicy* tracking_policy = _g1h->g1_policy()->remset_tracker();
1010
1011 size_t live_bytes = _cm->liveness(hr->hrm_index()) * HeapWordSize;
1012 bool selected_for_rebuild = tracking_policy->update_before_rebuild(hr, live_bytes);
1013 if (selected_for_rebuild) {
1014 _num_regions_selected_for_rebuild++;
1015 }
1016 _cm->update_top_at_rebuild_start(hr);
1017 }
1018
1019 public:
1020 G1UpdateRemSetTrackingBeforeRebuild(G1CollectedHeap* g1h, G1ConcurrentMark* cm) :
1021 _g1h(g1h), _cm(cm), _num_regions_selected_for_rebuild(0) { }
1022
1023 virtual bool do_heap_region(HeapRegion* r) {
1024 update_remset_before_rebuild(r);
1025 return false;
1026 }
1027
1028 uint num_selected_for_rebuild() const { return _num_regions_selected_for_rebuild; }
1029 };
1030
1031 class G1UpdateRemSetTrackingAfterRebuild : public HeapRegionClosure {
1032 G1CollectedHeap* _g1h;
1033 public:
1034 G1UpdateRemSetTrackingAfterRebuild(G1CollectedHeap* g1h) : _g1h(g1h) { }
1035
1036 virtual bool do_heap_region(HeapRegion* r) {
1037 _g1h->g1_policy()->remset_tracker()->update_after_rebuild(r);
1038 return false;
1039 }
1040 };
1041
1042 void G1ConcurrentMark::remark() {
1043 assert_at_safepoint_on_vm_thread();
1044
1045 // If a full collection has happened, we should not continue. However we might
1046 // have ended up here as the Remark VM operation has been scheduled already.
1047 if (has_aborted()) {
1048 return;
1049 }
1050
1051 G1Policy* g1p = _g1h->g1_policy();
1052 g1p->record_concurrent_mark_remark_start();
1053
1054 double start = os::elapsedTime();
1055
1056 verify_during_pause(G1HeapVerifier::G1VerifyRemark, VerifyOption_G1UsePrevMarking, "Remark before");
1057
1058 {
1059 GCTraceTime(Debug, gc, phases) trace("Finalize Marking", _gc_timer_cm);
1060 finalize_marking();
1061 }
1062
1063 double mark_work_end = os::elapsedTime();
1064
1065 bool const mark_finished = !has_overflown();
1066 if (mark_finished) {
1067 weak_refs_work(false /* clear_all_soft_refs */);
1068
1069 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1070 // We're done with marking.
1071 // This is the end of the marking cycle, we're expected all
1072 // threads to have SATB queues with active set to true.
1073 satb_mq_set.set_active_all_threads(false, /* new active value */
1074 true /* expected_active */);
1075
1076 {
1077 GCTraceTime(Debug, gc, phases)("Flush Task Caches");
1078 flush_all_task_caches();
1079 }
1080
1081 {
1082 GCTraceTime(Debug, gc, phases)("Update Remembered Set Tracking Before Rebuild");
1083 G1UpdateRemSetTrackingBeforeRebuild cl(_g1h, this);
1084 _g1h->heap_region_iterate(&cl);
1085 log_debug(gc, remset, tracking)("Remembered Set Tracking update regions total %u, selected %u",
1086 _g1h->num_regions(), cl.num_selected_for_rebuild());
1087 }
1088
1089 verify_during_pause(G1HeapVerifier::G1VerifyRemark, VerifyOption_G1UseNextMarking, "Remark after");
1090
1091 assert(!restart_for_overflow(), "sanity");
1092 // Completely reset the marking state since marking completed
1093 reset_at_marking_complete();
1094 } else {
1095 // We overflowed. Restart concurrent marking.
1096 _restart_for_overflow = true;
1097
1098 verify_during_pause(G1HeapVerifier::G1VerifyRemark, VerifyOption_G1UsePrevMarking, "Remark overflow");
1099
1100 // Clear the marking state because we will be restarting
1101 // marking due to overflowing the global mark stack.
1102 reset_marking_for_restart();
1103 }
1104
1105 {
1106 GCTraceTime(Debug, gc, phases)("Report Object Count");
1107 report_object_count();
1108 }
1109
1110 // Statistics
1111 double now = os::elapsedTime();
1112 _remark_mark_times.add((mark_work_end - start) * 1000.0);
1113 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0);
1114 _remark_times.add((now - start) * 1000.0);
1115
1116 g1p->record_concurrent_mark_remark_end();
1117 }
1118
1119 class G1CleanupTask : public AbstractGangTask {
1120 // Per-region work during the Cleanup pause.
1121 class G1CleanupRegionsClosure : public HeapRegionClosure {
1122 G1CollectedHeap* _g1h;
1123 size_t _freed_bytes;
1124 FreeRegionList* _local_cleanup_list;
1125 uint _old_regions_removed;
1126 uint _humongous_regions_removed;
1127 HRRSCleanupTask* _hrrs_cleanup_task;
1128
1129 public:
1130 G1CleanupRegionsClosure(G1CollectedHeap* g1,
1131 FreeRegionList* local_cleanup_list,
1132 HRRSCleanupTask* hrrs_cleanup_task) :
1133 _g1h(g1),
1134 _freed_bytes(0),
1135 _local_cleanup_list(local_cleanup_list),
1136 _old_regions_removed(0),
1137 _humongous_regions_removed(0),
1138 _hrrs_cleanup_task(hrrs_cleanup_task) { }
1139
1140 size_t freed_bytes() { return _freed_bytes; }
1141 const uint old_regions_removed() { return _old_regions_removed; }
1142 const uint humongous_regions_removed() { return _humongous_regions_removed; }
1143
1144 bool do_heap_region(HeapRegion *hr) {
1145 hr->note_end_of_marking();
1146
1147 if (hr->used() > 0 && hr->max_live_bytes() == 0 && !hr->is_young() && !hr->is_archive()) {
1148 _freed_bytes += hr->used();
1149 hr->set_containing_set(NULL);
1150 if (hr->is_humongous()) {
1151 _humongous_regions_removed++;
1152 _g1h->free_humongous_region(hr, _local_cleanup_list);
1153 } else {
1154 _old_regions_removed++;
1155 _g1h->free_region(hr, _local_cleanup_list, false /* skip_remset */, false /* skip_hcc */, true /* locked */);
1156 }
1157 hr->clear_cardtable();
1158 _g1h->concurrent_mark()->clear_statistics_in_region(hr->hrm_index());
1159 log_trace(gc)("Reclaimed empty region %u (%s) bot " PTR_FORMAT, hr->hrm_index(), hr->get_short_type_str(), p2i(hr->bottom()));
1160 } else {
1161 hr->rem_set()->do_cleanup_work(_hrrs_cleanup_task);
1162 }
1163
1164 return false;
1165 }
1166 };
1167
1168 G1CollectedHeap* _g1h;
1169 FreeRegionList* _cleanup_list;
1170 HeapRegionClaimer _hrclaimer;
1171
1172 public:
1173 G1CleanupTask(G1CollectedHeap* g1h, FreeRegionList* cleanup_list, uint n_workers) :
1174 AbstractGangTask("G1 Cleanup"),
1175 _g1h(g1h),
1176 _cleanup_list(cleanup_list),
1177 _hrclaimer(n_workers) {
1178
1179 HeapRegionRemSet::reset_for_cleanup_tasks();
1180 }
1181
1182 void work(uint worker_id) {
1183 FreeRegionList local_cleanup_list("Local Cleanup List");
1184 HRRSCleanupTask hrrs_cleanup_task;
1185 G1CleanupRegionsClosure cl(_g1h,
1186 &local_cleanup_list,
1187 &hrrs_cleanup_task);
1188 _g1h->heap_region_par_iterate_from_worker_offset(&cl, &_hrclaimer, worker_id);
1189 assert(cl.is_complete(), "Shouldn't have aborted!");
1190
1191 // Now update the old/humongous region sets
1192 _g1h->remove_from_old_sets(cl.old_regions_removed(), cl.humongous_regions_removed());
1193 {
1194 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
1195 _g1h->decrement_summary_bytes(cl.freed_bytes());
1196
1197 _cleanup_list->add_ordered(&local_cleanup_list);
1198 assert(local_cleanup_list.is_empty(), "post-condition");
1199
1200 HeapRegionRemSet::finish_cleanup_task(&hrrs_cleanup_task);
1201 }
1202 }
1203 };
1204
1205 void G1ConcurrentMark::reclaim_empty_regions() {
1206 WorkGang* workers = _g1h->workers();
1207 FreeRegionList empty_regions_list("Empty Regions After Mark List");
1208
1209 G1CleanupTask cl(_g1h, &empty_regions_list, workers->active_workers());
1210 workers->run_task(&cl);
1211
1212 if (!empty_regions_list.is_empty()) {
1213 log_debug(gc)("Reclaimed %u empty regions", empty_regions_list.length());
1214 // Now print the empty regions list.
1215 G1HRPrinter* hrp = _g1h->hr_printer();
1216 if (hrp->is_active()) {
1217 FreeRegionListIterator iter(&empty_regions_list);
1218 while (iter.more_available()) {
1219 HeapRegion* hr = iter.get_next();
1220 hrp->cleanup(hr);
1221 }
1222 }
1223 // And actually make them available.
1224 _g1h->prepend_to_freelist(&empty_regions_list);
1225 }
1226 }
1227
1228 void G1ConcurrentMark::cleanup() {
1229 assert_at_safepoint_on_vm_thread();
1230
1231 // If a full collection has happened, we shouldn't do this.
1232 if (has_aborted()) {
1233 return;
1234 }
1235
1236 G1Policy* g1p = _g1h->g1_policy();
1237 g1p->record_concurrent_mark_cleanup_start();
1238
1239 double start = os::elapsedTime();
1240
1241 verify_during_pause(G1HeapVerifier::G1VerifyCleanup, VerifyOption_G1UseNextMarking, "Cleanup before");
1242
1243 {
1244 GCTraceTime(Debug, gc, phases)("Update Remembered Set Tracking After Rebuild");
1245 G1UpdateRemSetTrackingAfterRebuild cl(_g1h);
1246 _g1h->heap_region_iterate(&cl);
1247 }
1248
1249 if (log_is_enabled(Trace, gc, liveness)) {
1250 G1PrintRegionLivenessInfoClosure cl("Post-Cleanup");
1251 _g1h->heap_region_iterate(&cl);
1252 }
1253
1254 // Install newly created mark bitmap as "prev".
1255 swap_mark_bitmaps();
1256 {
1257 GCTraceTime(Debug, gc, phases)("Reclaim Empty Regions");
1258 reclaim_empty_regions();
1259 }
1260
1261 // Cleanup will have freed any regions completely full of garbage.
1262 // Update the soft reference policy with the new heap occupancy.
1263 Universe::update_heap_info_at_gc();
1264
1265 // Clean out dead classes and update Metaspace sizes.
1266 if (ClassUnloadingWithConcurrentMark) {
1267 GCTraceTime(Debug, gc, phases)("Purge Metaspace");
1268 ClassLoaderDataGraph::purge();
1269 }
1270 MetaspaceGC::compute_new_size();
1271
1272 // We reclaimed old regions so we should calculate the sizes to make
1273 // sure we update the old gen/space data.
1274 _g1h->g1mm()->update_sizes();
1275
1276 verify_during_pause(G1HeapVerifier::G1VerifyCleanup, VerifyOption_G1UsePrevMarking, "Cleanup after");
1277
1278 // We need to make this be a "collection" so any collection pause that
1279 // races with it goes around and waits for Cleanup to finish.
1280 _g1h->increment_total_collections();
1281
1282 // Local statistics
1283 double recent_cleanup_time = (os::elapsedTime() - start);
1284 _total_cleanup_time += recent_cleanup_time;
1285 _cleanup_times.add(recent_cleanup_time);
1286
1287 {
1288 GCTraceTime(Debug, gc, phases)("Finalize Concurrent Mark Cleanup");
1289 _g1h->g1_policy()->record_concurrent_mark_cleanup_end();
1290 }
1291 }
1292
1293 // Supporting Object and Oop closures for reference discovery
1294 // and processing in during marking
1295
1296 bool G1CMIsAliveClosure::do_object_b(oop obj) {
1297 HeapWord* addr = (HeapWord*)obj;
1298 return addr != NULL &&
1299 (!_g1h->is_in_g1_reserved(addr) || !_g1h->is_obj_ill(obj));
1300 }
1301
1302 // 'Keep Alive' oop closure used by both serial parallel reference processing.
1303 // Uses the G1CMTask associated with a worker thread (for serial reference
1304 // processing the G1CMTask for worker 0 is used) to preserve (mark) and
1305 // trace referent objects.
1306 //
1307 // Using the G1CMTask and embedded local queues avoids having the worker
1308 // threads operating on the global mark stack. This reduces the risk
1309 // of overflowing the stack - which we would rather avoid at this late
1310 // state. Also using the tasks' local queues removes the potential
1311 // of the workers interfering with each other that could occur if
1312 // operating on the global stack.
1313
1314 class G1CMKeepAliveAndDrainClosure : public OopClosure {
1315 G1ConcurrentMark* _cm;
1316 G1CMTask* _task;
1317 int _ref_counter_limit;
1318 int _ref_counter;
1319 bool _is_serial;
1320 public:
1321 G1CMKeepAliveAndDrainClosure(G1ConcurrentMark* cm, G1CMTask* task, bool is_serial) :
1322 _cm(cm), _task(task), _is_serial(is_serial),
1323 _ref_counter_limit(G1RefProcDrainInterval) {
1324 assert(_ref_counter_limit > 0, "sanity");
1325 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
1326 _ref_counter = _ref_counter_limit;
1327 }
1328
1329 virtual void do_oop(narrowOop* p) { do_oop_work(p); }
1330 virtual void do_oop( oop* p) { do_oop_work(p); }
1331
1332 template <class T> void do_oop_work(T* p) {
1333 if (!_cm->has_overflown()) {
1334 _task->deal_with_reference(p);
1335 _ref_counter--;
1336
1337 if (_ref_counter == 0) {
1338 // We have dealt with _ref_counter_limit references, pushing them
1339 // and objects reachable from them on to the local stack (and
1340 // possibly the global stack). Call G1CMTask::do_marking_step() to
1341 // process these entries.
1342 //
1343 // We call G1CMTask::do_marking_step() in a loop, which we'll exit if
1344 // there's nothing more to do (i.e. we're done with the entries that
1345 // were pushed as a result of the G1CMTask::deal_with_reference() calls
1346 // above) or we overflow.
1347 //
1348 // Note: G1CMTask::do_marking_step() can set the G1CMTask::has_aborted()
1349 // flag while there may still be some work to do. (See the comment at
1350 // the beginning of G1CMTask::do_marking_step() for those conditions -
1351 // one of which is reaching the specified time target.) It is only
1352 // when G1CMTask::do_marking_step() returns without setting the
1353 // has_aborted() flag that the marking step has completed.
1354 do {
1355 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
1356 _task->do_marking_step(mark_step_duration_ms,
1357 false /* do_termination */,
1358 _is_serial);
1359 } while (_task->has_aborted() && !_cm->has_overflown());
1360 _ref_counter = _ref_counter_limit;
1361 }
1362 }
1363 }
1364 };
1365
1366 // 'Drain' oop closure used by both serial and parallel reference processing.
1367 // Uses the G1CMTask associated with a given worker thread (for serial
1368 // reference processing the G1CMtask for worker 0 is used). Calls the
1369 // do_marking_step routine, with an unbelievably large timeout value,
1370 // to drain the marking data structures of the remaining entries
1371 // added by the 'keep alive' oop closure above.
1372
1373 class G1CMDrainMarkingStackClosure : public VoidClosure {
1374 G1ConcurrentMark* _cm;
1375 G1CMTask* _task;
1376 bool _is_serial;
1377 public:
1378 G1CMDrainMarkingStackClosure(G1ConcurrentMark* cm, G1CMTask* task, bool is_serial) :
1379 _cm(cm), _task(task), _is_serial(is_serial) {
1380 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
1381 }
1382
1383 void do_void() {
1384 do {
1385 // We call G1CMTask::do_marking_step() to completely drain the local
1386 // and global marking stacks of entries pushed by the 'keep alive'
1387 // oop closure (an instance of G1CMKeepAliveAndDrainClosure above).
1388 //
1389 // G1CMTask::do_marking_step() is called in a loop, which we'll exit
1390 // if there's nothing more to do (i.e. we've completely drained the
1391 // entries that were pushed as a a result of applying the 'keep alive'
1392 // closure to the entries on the discovered ref lists) or we overflow
1393 // the global marking stack.
1394 //
1395 // Note: G1CMTask::do_marking_step() can set the G1CMTask::has_aborted()
1396 // flag while there may still be some work to do. (See the comment at
1397 // the beginning of G1CMTask::do_marking_step() for those conditions -
1398 // one of which is reaching the specified time target.) It is only
1399 // when G1CMTask::do_marking_step() returns without setting the
1400 // has_aborted() flag that the marking step has completed.
1401
1402 _task->do_marking_step(1000000000.0 /* something very large */,
1403 true /* do_termination */,
1404 _is_serial);
1405 } while (_task->has_aborted() && !_cm->has_overflown());
1406 }
1407 };
1408
1409 // Implementation of AbstractRefProcTaskExecutor for parallel
1410 // reference processing at the end of G1 concurrent marking
1411
1412 class G1CMRefProcTaskExecutor : public AbstractRefProcTaskExecutor {
1413 private:
1414 G1CollectedHeap* _g1h;
1415 G1ConcurrentMark* _cm;
1416 WorkGang* _workers;
1417 uint _active_workers;
1418
1419 public:
1420 G1CMRefProcTaskExecutor(G1CollectedHeap* g1h,
1421 G1ConcurrentMark* cm,
1422 WorkGang* workers,
1423 uint n_workers) :
1424 _g1h(g1h), _cm(cm),
1425 _workers(workers), _active_workers(n_workers) { }
1426
1427 // Executes the given task using concurrent marking worker threads.
1428 virtual void execute(ProcessTask& task);
1429 virtual void execute(EnqueueTask& task);
1430 };
1431
1432 class G1CMRefProcTaskProxy : public AbstractGangTask {
1433 typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask;
1434 ProcessTask& _proc_task;
1435 G1CollectedHeap* _g1h;
1436 G1ConcurrentMark* _cm;
1437
1438 public:
1439 G1CMRefProcTaskProxy(ProcessTask& proc_task,
1440 G1CollectedHeap* g1h,
1441 G1ConcurrentMark* cm) :
1442 AbstractGangTask("Process reference objects in parallel"),
1443 _proc_task(proc_task), _g1h(g1h), _cm(cm) {
1444 ReferenceProcessor* rp = _g1h->ref_processor_cm();
1445 assert(rp->processing_is_mt(), "shouldn't be here otherwise");
1446 }
1447
1448 virtual void work(uint worker_id) {
1449 ResourceMark rm;
1450 HandleMark hm;
1451 G1CMTask* task = _cm->task(worker_id);
1452 G1CMIsAliveClosure g1_is_alive(_g1h);
1453 G1CMKeepAliveAndDrainClosure g1_par_keep_alive(_cm, task, false /* is_serial */);
1454 G1CMDrainMarkingStackClosure g1_par_drain(_cm, task, false /* is_serial */);
1455
1456 _proc_task.work(worker_id, g1_is_alive, g1_par_keep_alive, g1_par_drain);
1457 }
1458 };
1459
1460 void G1CMRefProcTaskExecutor::execute(ProcessTask& proc_task) {
1461 assert(_workers != NULL, "Need parallel worker threads.");
1462 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
1463
1464 G1CMRefProcTaskProxy proc_task_proxy(proc_task, _g1h, _cm);
1465
1466 // We need to reset the concurrency level before each
1467 // proxy task execution, so that the termination protocol
1468 // and overflow handling in G1CMTask::do_marking_step() knows
1469 // how many workers to wait for.
1470 _cm->set_concurrency(_active_workers);
1471 _workers->run_task(&proc_task_proxy);
1472 }
1473
1474 class G1CMRefEnqueueTaskProxy : public AbstractGangTask {
1475 typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask;
1476 EnqueueTask& _enq_task;
1477
1478 public:
1479 G1CMRefEnqueueTaskProxy(EnqueueTask& enq_task) :
1480 AbstractGangTask("Enqueue reference objects in parallel"),
1481 _enq_task(enq_task) { }
1482
1483 virtual void work(uint worker_id) {
1484 _enq_task.work(worker_id);
1485 }
1486 };
1487
1488 void G1CMRefProcTaskExecutor::execute(EnqueueTask& enq_task) {
1489 assert(_workers != NULL, "Need parallel worker threads.");
1490 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
1491
1492 G1CMRefEnqueueTaskProxy enq_task_proxy(enq_task);
1493
1494 // Not strictly necessary but...
1495 //
1496 // We need to reset the concurrency level before each
1497 // proxy task execution, so that the termination protocol
1498 // and overflow handling in G1CMTask::do_marking_step() knows
1499 // how many workers to wait for.
1500 _cm->set_concurrency(_active_workers);
1501 _workers->run_task(&enq_task_proxy);
1502 }
1503
1504 void G1ConcurrentMark::weak_refs_work(bool clear_all_soft_refs) {
1505 ResourceMark rm;
1506 HandleMark hm;
1507
1508 // Is alive closure.
1509 G1CMIsAliveClosure g1_is_alive(_g1h);
1510
1511 // Inner scope to exclude the cleaning of the string and symbol
1512 // tables from the displayed time.
1513 {
1514 GCTraceTime(Debug, gc, phases) trace("Reference Processing", _gc_timer_cm);
1515
1516 ReferenceProcessor* rp = _g1h->ref_processor_cm();
1517
1518 // See the comment in G1CollectedHeap::ref_processing_init()
1519 // about how reference processing currently works in G1.
1520
1521 // Set the soft reference policy
1522 rp->setup_policy(clear_all_soft_refs);
1523 assert(_global_mark_stack.is_empty(), "mark stack should be empty");
1524
1525 // Instances of the 'Keep Alive' and 'Complete GC' closures used
1526 // in serial reference processing. Note these closures are also
1527 // used for serially processing (by the the current thread) the
1528 // JNI references during parallel reference processing.
1529 //
1530 // These closures do not need to synchronize with the worker
1531 // threads involved in parallel reference processing as these
1532 // instances are executed serially by the current thread (e.g.
1533 // reference processing is not multi-threaded and is thus
1534 // performed by the current thread instead of a gang worker).
1535 //
1536 // The gang tasks involved in parallel reference processing create
1537 // their own instances of these closures, which do their own
1538 // synchronization among themselves.
1539 G1CMKeepAliveAndDrainClosure g1_keep_alive(this, task(0), true /* is_serial */);
1540 G1CMDrainMarkingStackClosure g1_drain_mark_stack(this, task(0), true /* is_serial */);
1541
1542 // We need at least one active thread. If reference processing
1543 // is not multi-threaded we use the current (VMThread) thread,
1544 // otherwise we use the work gang from the G1CollectedHeap and
1545 // we utilize all the worker threads we can.
1546 bool processing_is_mt = rp->processing_is_mt();
1547 uint active_workers = (processing_is_mt ? _g1h->workers()->active_workers() : 1U);
1548 active_workers = MAX2(MIN2(active_workers, _max_num_tasks), 1U);
1549
1550 // Parallel processing task executor.
1551 G1CMRefProcTaskExecutor par_task_executor(_g1h, this,
1552 _g1h->workers(), active_workers);
1553 AbstractRefProcTaskExecutor* executor = (processing_is_mt ? &par_task_executor : NULL);
1554
1555 // Set the concurrency level. The phase was already set prior to
1556 // executing the remark task.
1557 set_concurrency(active_workers);
1558
1559 // Set the degree of MT processing here. If the discovery was done MT,
1560 // the number of threads involved during discovery could differ from
1561 // the number of active workers. This is OK as long as the discovered
1562 // Reference lists are balanced (see balance_all_queues() and balance_queues()).
1563 rp->set_active_mt_degree(active_workers);
1564
1565 ReferenceProcessorPhaseTimes pt(_gc_timer_cm, rp->num_q());
1566
1567 // Process the weak references.
1568 const ReferenceProcessorStats& stats =
1569 rp->process_discovered_references(&g1_is_alive,
1570 &g1_keep_alive,
1571 &g1_drain_mark_stack,
1572 executor,
1573 &pt);
1574 _gc_tracer_cm->report_gc_reference_stats(stats);
1575 pt.print_all_references();
1576
1577 // The do_oop work routines of the keep_alive and drain_marking_stack
1578 // oop closures will set the has_overflown flag if we overflow the
1579 // global marking stack.
1580
1581 assert(has_overflown() || _global_mark_stack.is_empty(),
1582 "Mark stack should be empty (unless it has overflown)");
1583
1584 assert(rp->num_q() == active_workers, "why not");
1585
1586 rp->enqueue_discovered_references(executor, &pt);
1587
1588 rp->verify_no_references_recorded();
1589
1590 pt.print_enqueue_phase();
1591
1592 assert(!rp->discovery_enabled(), "Post condition");
1593 }
1594
1595 assert(has_overflown() || _global_mark_stack.is_empty(),
1596 "Mark stack should be empty (unless it has overflown)");
1597
1598 {
1599 GCTraceTime(Debug, gc, phases) debug("Weak Processing", _gc_timer_cm);
1600 WeakProcessor::weak_oops_do(&g1_is_alive, &do_nothing_cl);
1601 }
1602
1603 if (has_overflown()) {
1604 // We can not trust g1_is_alive if the marking stack overflowed
1605 return;
1606 }
1607
1608 assert(_global_mark_stack.is_empty(), "Marking should have completed");
1609
1610 // Unload Klasses, String, Symbols, Code Cache, etc.
1611 if (ClassUnloadingWithConcurrentMark) {
1612 GCTraceTime(Debug, gc, phases) debug("Class Unloading", _gc_timer_cm);
1613 bool purged_classes = SystemDictionary::do_unloading(&g1_is_alive, _gc_timer_cm, false /* Defer cleaning */);
1614 _g1h->complete_cleaning(&g1_is_alive, purged_classes);
1615 } else {
1616 GCTraceTime(Debug, gc, phases) debug("Cleanup", _gc_timer_cm);
1617 // No need to clean string table and symbol table as they are treated as strong roots when
1618 // class unloading is disabled.
1619 _g1h->partial_cleaning(&g1_is_alive, false, false, G1StringDedup::is_enabled());
1620 }
1621 }
1622
1623 void G1ConcurrentMark::report_object_count() {
1624 G1CMIsAliveClosure is_alive(_g1h);
1625 _gc_tracer_cm->report_object_count_after_gc(&is_alive);
1626 }
1627
1628 void G1ConcurrentMark::swap_mark_bitmaps() {
1629 G1CMBitMap* temp = _prev_mark_bitmap;
1630 _prev_mark_bitmap = _next_mark_bitmap;
1631 _next_mark_bitmap = temp;
1632 _g1h->collector_state()->set_clearing_next_bitmap(true);
1633 }
1634
1635 // Closure for marking entries in SATB buffers.
1636 class G1CMSATBBufferClosure : public SATBBufferClosure {
1637 private:
1638 G1CMTask* _task;
1639 G1CollectedHeap* _g1h;
1640
1641 // This is very similar to G1CMTask::deal_with_reference, but with
1642 // more relaxed requirements for the argument, so this must be more
1643 // circumspect about treating the argument as an object.
1644 void do_entry(void* entry) const {
1645 _task->increment_refs_reached();
1646 oop const obj = static_cast<oop>(entry);
1647 _task->make_reference_grey(obj);
1648 }
1649
1650 public:
1651 G1CMSATBBufferClosure(G1CMTask* task, G1CollectedHeap* g1h)
1652 : _task(task), _g1h(g1h) { }
1653
1654 virtual void do_buffer(void** buffer, size_t size) {
1655 for (size_t i = 0; i < size; ++i) {
1656 do_entry(buffer[i]);
1657 }
1658 }
1659 };
1660
1661 class G1RemarkThreadsClosure : public ThreadClosure {
1662 G1CMSATBBufferClosure _cm_satb_cl;
1663 G1CMOopClosure _cm_cl;
1664 MarkingCodeBlobClosure _code_cl;
1665 int _thread_parity;
1666
1667 public:
1668 G1RemarkThreadsClosure(G1CollectedHeap* g1h, G1CMTask* task) :
1669 _cm_satb_cl(task, g1h),
1670 _cm_cl(g1h, task),
1671 _code_cl(&_cm_cl, !CodeBlobToOopClosure::FixRelocations),
1672 _thread_parity(Threads::thread_claim_parity()) {}
1673
1674 void do_thread(Thread* thread) {
1675 if (thread->is_Java_thread()) {
1676 if (thread->claim_oops_do(true, _thread_parity)) {
1677 JavaThread* jt = (JavaThread*)thread;
1678
1679 // In theory it should not be neccessary to explicitly walk the nmethods to find roots for concurrent marking
1680 // however the liveness of oops reachable from nmethods have very complex lifecycles:
1681 // * Alive if on the stack of an executing method
1682 // * Weakly reachable otherwise
1683 // Some objects reachable from nmethods, such as the class loader (or klass_holder) of the receiver should be
1684 // live by the SATB invariant but other oops recorded in nmethods may behave differently.
1685 jt->nmethods_do(&_code_cl);
1686
1687 jt->satb_mark_queue().apply_closure_and_empty(&_cm_satb_cl);
1688 }
1689 } else if (thread->is_VM_thread()) {
1690 if (thread->claim_oops_do(true, _thread_parity)) {
1691 JavaThread::satb_mark_queue_set().shared_satb_queue()->apply_closure_and_empty(&_cm_satb_cl);
1692 }
1693 }
1694 }
1695 };
1696
1697 class G1CMRemarkTask : public AbstractGangTask {
1698 G1ConcurrentMark* _cm;
1699 public:
1700 void work(uint worker_id) {
1701 G1CMTask* task = _cm->task(worker_id);
1702 task->record_start_time();
1703 {
1704 ResourceMark rm;
1705 HandleMark hm;
1706
1707 G1RemarkThreadsClosure threads_f(G1CollectedHeap::heap(), task);
1708 Threads::threads_do(&threads_f);
1709 }
1710
1711 do {
1712 task->do_marking_step(1000000000.0 /* something very large */,
1713 true /* do_termination */,
1714 false /* is_serial */);
1715 } while (task->has_aborted() && !_cm->has_overflown());
1716 // If we overflow, then we do not want to restart. We instead
1717 // want to abort remark and do concurrent marking again.
1718 task->record_end_time();
1719 }
1720
1721 G1CMRemarkTask(G1ConcurrentMark* cm, uint active_workers) :
1722 AbstractGangTask("Par Remark"), _cm(cm) {
1723 _cm->terminator()->reset_for_reuse(active_workers);
1724 }
1725 };
1726
1727 void G1ConcurrentMark::finalize_marking() {
1728 ResourceMark rm;
1729 HandleMark hm;
1730
1731 _g1h->ensure_parsability(false);
1732
1733 // this is remark, so we'll use up all active threads
1734 uint active_workers = _g1h->workers()->active_workers();
1735 set_concurrency_and_phase(active_workers, false /* concurrent */);
1736 // Leave _parallel_marking_threads at it's
1737 // value originally calculated in the G1ConcurrentMark
1738 // constructor and pass values of the active workers
1739 // through the gang in the task.
1740
1741 {
1742 StrongRootsScope srs(active_workers);
1743
1744 G1CMRemarkTask remarkTask(this, active_workers);
1745 // We will start all available threads, even if we decide that the
1746 // active_workers will be fewer. The extra ones will just bail out
1747 // immediately.
1748 _g1h->workers()->run_task(&remarkTask);
1749 }
1750
1751 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1752 guarantee(has_overflown() ||
1753 satb_mq_set.completed_buffers_num() == 0,
1754 "Invariant: has_overflown = %s, num buffers = " SIZE_FORMAT,
1755 BOOL_TO_STR(has_overflown()),
1756 satb_mq_set.completed_buffers_num());
1757
1758 print_stats();
1759 }
1760
1761 void G1ConcurrentMark::flush_all_task_caches() {
1762 size_t hits = 0;
1763 size_t misses = 0;
1764 for (uint i = 0; i < _max_num_tasks; i++) {
1765 Pair<size_t, size_t> stats = _tasks[i]->flush_mark_stats_cache();
1766 hits += stats.first;
1767 misses += stats.second;
1768 }
1769 size_t sum = hits + misses;
1770 log_debug(gc, stats)("Mark stats cache hits " SIZE_FORMAT " misses " SIZE_FORMAT " ratio %1.3lf",
1771 hits, misses, percent_of(hits, sum));
1772 }
1773
1774 void G1ConcurrentMark::clear_range_in_prev_bitmap(MemRegion mr) {
1775 _prev_mark_bitmap->clear_range(mr);
1776 }
1777
1778 HeapRegion*
1779 G1ConcurrentMark::claim_region(uint worker_id) {
1780 // "checkpoint" the finger
1781 HeapWord* finger = _finger;
1782
1783 while (finger < _heap.end()) {
1784 assert(_g1h->is_in_g1_reserved(finger), "invariant");
1785
1786 HeapRegion* curr_region = _g1h->heap_region_containing(finger);
1787 // Make sure that the reads below do not float before loading curr_region.
1788 OrderAccess::loadload();
1789 // Above heap_region_containing may return NULL as we always scan claim
1790 // until the end of the heap. In this case, just jump to the next region.
1791 HeapWord* end = curr_region != NULL ? curr_region->end() : finger + HeapRegion::GrainWords;
1792
1793 // Is the gap between reading the finger and doing the CAS too long?
1794 HeapWord* res = Atomic::cmpxchg(end, &_finger, finger);
1795 if (res == finger && curr_region != NULL) {
1796 // we succeeded
1797 HeapWord* bottom = curr_region->bottom();
1798 HeapWord* limit = curr_region->next_top_at_mark_start();
1799
1800 // notice that _finger == end cannot be guaranteed here since,
1801 // someone else might have moved the finger even further
1802 assert(_finger >= end, "the finger should have moved forward");
1803
1804 if (limit > bottom) {
1805 return curr_region;
1806 } else {
1807 assert(limit == bottom,
1808 "the region limit should be at bottom");
1809 // we return NULL and the caller should try calling
1810 // claim_region() again.
1811 return NULL;
1812 }
1813 } else {
1814 assert(_finger > finger, "the finger should have moved forward");
1815 // read it again
1816 finger = _finger;
1817 }
1818 }
1819
1820 return NULL;
1821 }
1822
1823 #ifndef PRODUCT
1824 class VerifyNoCSetOops {
1825 G1CollectedHeap* _g1h;
1826 const char* _phase;
1827 int _info;
1828
1829 public:
1830 VerifyNoCSetOops(const char* phase, int info = -1) :
1831 _g1h(G1CollectedHeap::heap()),
1832 _phase(phase),
1833 _info(info)
1834 { }
1835
1836 void operator()(G1TaskQueueEntry task_entry) const {
1837 if (task_entry.is_array_slice()) {
1838 guarantee(_g1h->is_in_reserved(task_entry.slice()), "Slice " PTR_FORMAT " must be in heap.", p2i(task_entry.slice()));
1839 return;
1840 }
1841 guarantee(oopDesc::is_oop(task_entry.obj()),
1842 "Non-oop " PTR_FORMAT ", phase: %s, info: %d",
1843 p2i(task_entry.obj()), _phase, _info);
1844 guarantee(!_g1h->is_in_cset(task_entry.obj()),
1845 "obj: " PTR_FORMAT " in CSet, phase: %s, info: %d",
1846 p2i(task_entry.obj()), _phase, _info);
1847 }
1848 };
1849
1850 void G1ConcurrentMark::verify_no_cset_oops() {
1851 assert(SafepointSynchronize::is_at_safepoint(), "should be at a safepoint");
1852 if (!_g1h->collector_state()->mark_or_rebuild_in_progress()) {
1853 return;
1854 }
1855
1856 // Verify entries on the global mark stack
1857 _global_mark_stack.iterate(VerifyNoCSetOops("Stack"));
1858
1859 // Verify entries on the task queues
1860 for (uint i = 0; i < _max_num_tasks; ++i) {
1861 G1CMTaskQueue* queue = _task_queues->queue(i);
1862 queue->iterate(VerifyNoCSetOops("Queue", i));
1863 }
1864
1865 // Verify the global finger
1866 HeapWord* global_finger = finger();
1867 if (global_finger != NULL && global_finger < _heap.end()) {
1868 // Since we always iterate over all regions, we might get a NULL HeapRegion
1869 // here.
1870 HeapRegion* global_hr = _g1h->heap_region_containing(global_finger);
1871 guarantee(global_hr == NULL || global_finger == global_hr->bottom(),
1872 "global finger: " PTR_FORMAT " region: " HR_FORMAT,
1873 p2i(global_finger), HR_FORMAT_PARAMS(global_hr));
1874 }
1875
1876 // Verify the task fingers
1877 assert(_num_concurrent_workers <= _max_num_tasks, "sanity");
1878 for (uint i = 0; i < _num_concurrent_workers; ++i) {
1879 G1CMTask* task = _tasks[i];
1880 HeapWord* task_finger = task->finger();
1881 if (task_finger != NULL && task_finger < _heap.end()) {
1882 // See above note on the global finger verification.
1883 HeapRegion* task_hr = _g1h->heap_region_containing(task_finger);
1884 guarantee(task_hr == NULL || task_finger == task_hr->bottom() ||
1885 !task_hr->in_collection_set(),
1886 "task finger: " PTR_FORMAT " region: " HR_FORMAT,
1887 p2i(task_finger), HR_FORMAT_PARAMS(task_hr));
1888 }
1889 }
1890 }
1891 #endif // PRODUCT
1892
1893 void G1ConcurrentMark::rebuild_rem_set_concurrently() {
1894 _g1h->g1_rem_set()->rebuild_rem_set(this, _concurrent_workers, _worker_id_offset);
1895 }
1896
1897 void G1ConcurrentMark::print_stats() {
1898 if (!log_is_enabled(Debug, gc, stats)) {
1899 return;
1900 }
1901 log_debug(gc, stats)("---------------------------------------------------------------------");
1902 for (size_t i = 0; i < _num_active_tasks; ++i) {
1903 _tasks[i]->print_stats();
1904 log_debug(gc, stats)("---------------------------------------------------------------------");
1905 }
1906 }
1907
1908 void G1ConcurrentMark::concurrent_cycle_abort() {
1909 if (!cm_thread()->during_cycle() || _has_aborted) {
1910 // We haven't started a concurrent cycle or we have already aborted it. No need to do anything.
1911 return;
1912 }
1913
1914 // Clear all marks in the next bitmap for the next marking cycle. This will allow us to skip the next
1915 // concurrent bitmap clearing.
1916 {
1917 GCTraceTime(Debug, gc)("Clear Next Bitmap");
1918 clear_bitmap(_next_mark_bitmap, _g1h->workers(), false);
1919 }
1920 // Note we cannot clear the previous marking bitmap here
1921 // since VerifyDuringGC verifies the objects marked during
1922 // a full GC against the previous bitmap.
1923
1924 // Empty mark stack
1925 reset_marking_for_restart();
1926 for (uint i = 0; i < _max_num_tasks; ++i) {
1927 _tasks[i]->clear_region_fields();
1928 }
1929 _first_overflow_barrier_sync.abort();
1930 _second_overflow_barrier_sync.abort();
1931 _has_aborted = true;
1932
1933 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1934 satb_mq_set.abandon_partial_marking();
1935 // This can be called either during or outside marking, we'll read
1936 // the expected_active value from the SATB queue set.
1937 satb_mq_set.set_active_all_threads(
1938 false, /* new active value */
1939 satb_mq_set.is_active() /* expected_active */);
1940 }
1941
1942 static void print_ms_time_info(const char* prefix, const char* name,
1943 NumberSeq& ns) {
1944 log_trace(gc, marking)("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).",
1945 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg());
1946 if (ns.num() > 0) {
1947 log_trace(gc, marking)("%s [std. dev = %8.2f ms, max = %8.2f ms]",
1948 prefix, ns.sd(), ns.maximum());
1949 }
1950 }
1951
1952 void G1ConcurrentMark::print_summary_info() {
1953 Log(gc, marking) log;
1954 if (!log.is_trace()) {
1955 return;
1956 }
1957
1958 log.trace(" Concurrent marking:");
1959 print_ms_time_info(" ", "init marks", _init_times);
1960 print_ms_time_info(" ", "remarks", _remark_times);
1961 {
1962 print_ms_time_info(" ", "final marks", _remark_mark_times);
1963 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times);
1964
1965 }
1966 print_ms_time_info(" ", "cleanups", _cleanup_times);
1967 log.trace(" Finalize live data total time = %8.2f s (avg = %8.2f ms).",
1968 _total_cleanup_time, (_cleanup_times.num() > 0 ? _total_cleanup_time * 1000.0 / (double)_cleanup_times.num() : 0.0));
1969 log.trace(" Total stop_world time = %8.2f s.",
1970 (_init_times.sum() + _remark_times.sum() + _cleanup_times.sum())/1000.0);
1971 log.trace(" Total concurrent time = %8.2f s (%8.2f s marking).",
1972 cm_thread()->vtime_accum(), cm_thread()->vtime_mark_accum());
1973 }
1974
1975 void G1ConcurrentMark::print_worker_threads_on(outputStream* st) const {
1976 _concurrent_workers->print_worker_threads_on(st);
1977 }
1978
1979 void G1ConcurrentMark::threads_do(ThreadClosure* tc) const {
1980 _concurrent_workers->threads_do(tc);
1981 }
1982
1983 void G1ConcurrentMark::print_on_error(outputStream* st) const {
1984 st->print_cr("Marking Bits (Prev, Next): (CMBitMap*) " PTR_FORMAT ", (CMBitMap*) " PTR_FORMAT,
1985 p2i(_prev_mark_bitmap), p2i(_next_mark_bitmap));
1986 _prev_mark_bitmap->print_on_error(st, " Prev Bits: ");
1987 _next_mark_bitmap->print_on_error(st, " Next Bits: ");
1988 }
1989
1990 static ReferenceProcessor* get_cm_oop_closure_ref_processor(G1CollectedHeap* g1h) {
1991 ReferenceProcessor* result = g1h->ref_processor_cm();
1992 assert(result != NULL, "CM reference processor should not be NULL");
1993 return result;
1994 }
1995
1996 G1CMOopClosure::G1CMOopClosure(G1CollectedHeap* g1h,
1997 G1CMTask* task)
1998 : MetadataAwareOopClosure(get_cm_oop_closure_ref_processor(g1h)),
1999 _g1h(g1h), _task(task)
2000 { }
2001
2002 void G1CMTask::setup_for_region(HeapRegion* hr) {
2003 assert(hr != NULL,
2004 "claim_region() should have filtered out NULL regions");
2005 _curr_region = hr;
2006 _finger = hr->bottom();
2007 update_region_limit();
2008 }
2009
2010 void G1CMTask::update_region_limit() {
2011 HeapRegion* hr = _curr_region;
2012 HeapWord* bottom = hr->bottom();
2013 HeapWord* limit = hr->next_top_at_mark_start();
2014
2015 if (limit == bottom) {
2016 // The region was collected underneath our feet.
2017 // We set the finger to bottom to ensure that the bitmap
2018 // iteration that will follow this will not do anything.
2019 // (this is not a condition that holds when we set the region up,
2020 // as the region is not supposed to be empty in the first place)
2021 _finger = bottom;
2022 } else if (limit >= _region_limit) {
2023 assert(limit >= _finger, "peace of mind");
2024 } else {
2025 assert(limit < _region_limit, "only way to get here");
2026 // This can happen under some pretty unusual circumstances. An
2027 // evacuation pause empties the region underneath our feet (NTAMS
2028 // at bottom). We then do some allocation in the region (NTAMS
2029 // stays at bottom), followed by the region being used as a GC
2030 // alloc region (NTAMS will move to top() and the objects
2031 // originally below it will be grayed). All objects now marked in
2032 // the region are explicitly grayed, if below the global finger,
2033 // and we do not need in fact to scan anything else. So, we simply
2034 // set _finger to be limit to ensure that the bitmap iteration
2035 // doesn't do anything.
2036 _finger = limit;
2037 }
2038
2039 _region_limit = limit;
2040 }
2041
2042 void G1CMTask::giveup_current_region() {
2043 assert(_curr_region != NULL, "invariant");
2044 clear_region_fields();
2045 }
2046
2047 void G1CMTask::clear_region_fields() {
2048 // Values for these three fields that indicate that we're not
2049 // holding on to a region.
2050 _curr_region = NULL;
2051 _finger = NULL;
2052 _region_limit = NULL;
2053 }
2054
2055 void G1CMTask::set_cm_oop_closure(G1CMOopClosure* cm_oop_closure) {
2056 if (cm_oop_closure == NULL) {
2057 assert(_cm_oop_closure != NULL, "invariant");
2058 } else {
2059 assert(_cm_oop_closure == NULL, "invariant");
2060 }
2061 _cm_oop_closure = cm_oop_closure;
2062 }
2063
2064 void G1CMTask::reset(G1CMBitMap* next_mark_bitmap) {
2065 guarantee(next_mark_bitmap != NULL, "invariant");
2066 _next_mark_bitmap = next_mark_bitmap;
2067 clear_region_fields();
2068
2069 _calls = 0;
2070 _elapsed_time_ms = 0.0;
2071 _termination_time_ms = 0.0;
2072 _termination_start_time_ms = 0.0;
2073
2074 _mark_stats_cache.reset();
2075 }
2076
2077 bool G1CMTask::should_exit_termination() {
2078 regular_clock_call();
2079 // This is called when we are in the termination protocol. We should
2080 // quit if, for some reason, this task wants to abort or the global
2081 // stack is not empty (this means that we can get work from it).
2082 return !_cm->mark_stack_empty() || has_aborted();
2083 }
2084
2085 void G1CMTask::reached_limit() {
2086 assert(_words_scanned >= _words_scanned_limit ||
2087 _refs_reached >= _refs_reached_limit ,
2088 "shouldn't have been called otherwise");
2089 regular_clock_call();
2090 }
2091
2092 void G1CMTask::regular_clock_call() {
2093 if (has_aborted()) {
2094 return;
2095 }
2096
2097 // First, we need to recalculate the words scanned and refs reached
2098 // limits for the next clock call.
2099 recalculate_limits();
2100
2101 // During the regular clock call we do the following
2102
2103 // (1) If an overflow has been flagged, then we abort.
2104 if (_cm->has_overflown()) {
2105 set_has_aborted();
2106 return;
2107 }
2108
2109 // If we are not concurrent (i.e. we're doing remark) we don't need
2110 // to check anything else. The other steps are only needed during
2111 // the concurrent marking phase.
2112 if (!_cm->concurrent()) {
2113 return;
2114 }
2115
2116 // (2) If marking has been aborted for Full GC, then we also abort.
2117 if (_cm->has_aborted()) {
2118 set_has_aborted();
2119 return;
2120 }
2121
2122 double curr_time_ms = os::elapsedVTime() * 1000.0;
2123
2124 // (4) We check whether we should yield. If we have to, then we abort.
2125 if (SuspendibleThreadSet::should_yield()) {
2126 // We should yield. To do this we abort the task. The caller is
2127 // responsible for yielding.
2128 set_has_aborted();
2129 return;
2130 }
2131
2132 // (5) We check whether we've reached our time quota. If we have,
2133 // then we abort.
2134 double elapsed_time_ms = curr_time_ms - _start_time_ms;
2135 if (elapsed_time_ms > _time_target_ms) {
2136 set_has_aborted();
2137 _has_timed_out = true;
2138 return;
2139 }
2140
2141 // (6) Finally, we check whether there are enough completed STAB
2142 // buffers available for processing. If there are, we abort.
2143 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2144 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) {
2145 // we do need to process SATB buffers, we'll abort and restart
2146 // the marking task to do so
2147 set_has_aborted();
2148 return;
2149 }
2150 }
2151
2152 void G1CMTask::recalculate_limits() {
2153 _real_words_scanned_limit = _words_scanned + words_scanned_period;
2154 _words_scanned_limit = _real_words_scanned_limit;
2155
2156 _real_refs_reached_limit = _refs_reached + refs_reached_period;
2157 _refs_reached_limit = _real_refs_reached_limit;
2158 }
2159
2160 void G1CMTask::decrease_limits() {
2161 // This is called when we believe that we're going to do an infrequent
2162 // operation which will increase the per byte scanned cost (i.e. move
2163 // entries to/from the global stack). It basically tries to decrease the
2164 // scanning limit so that the clock is called earlier.
2165
2166 _words_scanned_limit = _real_words_scanned_limit - 3 * words_scanned_period / 4;
2167 _refs_reached_limit = _real_refs_reached_limit - 3 * refs_reached_period / 4;
2168 }
2169
2170 void G1CMTask::move_entries_to_global_stack() {
2171 // Local array where we'll store the entries that will be popped
2172 // from the local queue.
2173 G1TaskQueueEntry buffer[G1CMMarkStack::EntriesPerChunk];
2174
2175 size_t n = 0;
2176 G1TaskQueueEntry task_entry;
2177 while (n < G1CMMarkStack::EntriesPerChunk && _task_queue->pop_local(task_entry)) {
2178 buffer[n] = task_entry;
2179 ++n;
2180 }
2181 if (n < G1CMMarkStack::EntriesPerChunk) {
2182 buffer[n] = G1TaskQueueEntry();
2183 }
2184
2185 if (n > 0) {
2186 if (!_cm->mark_stack_push(buffer)) {
2187 set_has_aborted();
2188 }
2189 }
2190
2191 // This operation was quite expensive, so decrease the limits.
2192 decrease_limits();
2193 }
2194
2195 bool G1CMTask::get_entries_from_global_stack() {
2196 // Local array where we'll store the entries that will be popped
2197 // from the global stack.
2198 G1TaskQueueEntry buffer[G1CMMarkStack::EntriesPerChunk];
2199
2200 if (!_cm->mark_stack_pop(buffer)) {
2201 return false;
2202 }
2203
2204 // We did actually pop at least one entry.
2205 for (size_t i = 0; i < G1CMMarkStack::EntriesPerChunk; ++i) {
2206 G1TaskQueueEntry task_entry = buffer[i];
2207 if (task_entry.is_null()) {
2208 break;
2209 }
2210 assert(task_entry.is_array_slice() || oopDesc::is_oop(task_entry.obj()), "Element " PTR_FORMAT " must be an array slice or oop", p2i(task_entry.obj()));
2211 bool success = _task_queue->push(task_entry);
2212 // We only call this when the local queue is empty or under a
2213 // given target limit. So, we do not expect this push to fail.
2214 assert(success, "invariant");
2215 }
2216
2217 // This operation was quite expensive, so decrease the limits
2218 decrease_limits();
2219 return true;
2220 }
2221
2222 void G1CMTask::drain_local_queue(bool partially) {
2223 if (has_aborted()) {
2224 return;
2225 }
2226
2227 // Decide what the target size is, depending whether we're going to
2228 // drain it partially (so that other tasks can steal if they run out
2229 // of things to do) or totally (at the very end).
2230 size_t target_size;
2231 if (partially) {
2232 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize);
2233 } else {
2234 target_size = 0;
2235 }
2236
2237 if (_task_queue->size() > target_size) {
2238 G1TaskQueueEntry entry;
2239 bool ret = _task_queue->pop_local(entry);
2240 while (ret) {
2241 scan_task_entry(entry);
2242 if (_task_queue->size() <= target_size || has_aborted()) {
2243 ret = false;
2244 } else {
2245 ret = _task_queue->pop_local(entry);
2246 }
2247 }
2248 }
2249 }
2250
2251 void G1CMTask::drain_global_stack(bool partially) {
2252 if (has_aborted()) {
2253 return;
2254 }
2255
2256 // We have a policy to drain the local queue before we attempt to
2257 // drain the global stack.
2258 assert(partially || _task_queue->size() == 0, "invariant");
2259
2260 // Decide what the target size is, depending whether we're going to
2261 // drain it partially (so that other tasks can steal if they run out
2262 // of things to do) or totally (at the very end).
2263 // Notice that when draining the global mark stack partially, due to the racyness
2264 // of the mark stack size update we might in fact drop below the target. But,
2265 // this is not a problem.
2266 // In case of total draining, we simply process until the global mark stack is
2267 // totally empty, disregarding the size counter.
2268 if (partially) {
2269 size_t const target_size = _cm->partial_mark_stack_size_target();
2270 while (!has_aborted() && _cm->mark_stack_size() > target_size) {
2271 if (get_entries_from_global_stack()) {
2272 drain_local_queue(partially);
2273 }
2274 }
2275 } else {
2276 while (!has_aborted() && get_entries_from_global_stack()) {
2277 drain_local_queue(partially);
2278 }
2279 }
2280 }
2281
2282 // SATB Queue has several assumptions on whether to call the par or
2283 // non-par versions of the methods. this is why some of the code is
2284 // replicated. We should really get rid of the single-threaded version
2285 // of the code to simplify things.
2286 void G1CMTask::drain_satb_buffers() {
2287 if (has_aborted()) {
2288 return;
2289 }
2290
2291 // We set this so that the regular clock knows that we're in the
2292 // middle of draining buffers and doesn't set the abort flag when it
2293 // notices that SATB buffers are available for draining. It'd be
2294 // very counter productive if it did that. :-)
2295 _draining_satb_buffers = true;
2296
2297 G1CMSATBBufferClosure satb_cl(this, _g1h);
2298 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2299
2300 // This keeps claiming and applying the closure to completed buffers
2301 // until we run out of buffers or we need to abort.
2302 while (!has_aborted() &&
2303 satb_mq_set.apply_closure_to_completed_buffer(&satb_cl)) {
2304 regular_clock_call();
2305 }
2306
2307 _draining_satb_buffers = false;
2308
2309 assert(has_aborted() ||
2310 _cm->concurrent() ||
2311 satb_mq_set.completed_buffers_num() == 0, "invariant");
2312
2313 // again, this was a potentially expensive operation, decrease the
2314 // limits to get the regular clock call early
2315 decrease_limits();
2316 }
2317
2318 void G1CMTask::clear_mark_stats_cache(uint region_idx) {
2319 _mark_stats_cache.reset(region_idx);
2320 }
2321
2322 Pair<size_t, size_t> G1CMTask::flush_mark_stats_cache() {
2323 return _mark_stats_cache.evict_all();
2324 }
2325
2326 void G1CMTask::print_stats() {
2327 log_debug(gc, stats)("Marking Stats, task = %u, calls = %u", _worker_id, _calls);
2328 log_debug(gc, stats)(" Elapsed time = %1.2lfms, Termination time = %1.2lfms",
2329 _elapsed_time_ms, _termination_time_ms);
2330 log_debug(gc, stats)(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms max = %1.2lfms, total = %1.2lfms",
2331 _step_times_ms.num(),
2332 _step_times_ms.avg(),
2333 _step_times_ms.sd(),
2334 _step_times_ms.maximum(),
2335 _step_times_ms.sum());
2336 size_t const hits = _mark_stats_cache.hits();
2337 size_t const misses = _mark_stats_cache.misses();
2338 log_debug(gc, stats)(" Mark Stats Cache: hits " SIZE_FORMAT " misses " SIZE_FORMAT " ratio %.3f",
2339 hits, misses, percent_of(hits, hits + misses));
2340 }
2341
2342 bool G1ConcurrentMark::try_stealing(uint worker_id, int* hash_seed, G1TaskQueueEntry& task_entry) {
2343 return _task_queues->steal(worker_id, hash_seed, task_entry);
2344 }
2345
2346 /*****************************************************************************
2347
2348 The do_marking_step(time_target_ms, ...) method is the building
2349 block of the parallel marking framework. It can be called in parallel
2350 with other invocations of do_marking_step() on different tasks
2351 (but only one per task, obviously) and concurrently with the
2352 mutator threads, or during remark, hence it eliminates the need
2353 for two versions of the code. When called during remark, it will
2354 pick up from where the task left off during the concurrent marking
2355 phase. Interestingly, tasks are also claimable during evacuation
2356 pauses too, since do_marking_step() ensures that it aborts before
2357 it needs to yield.
2358
2359 The data structures that it uses to do marking work are the
2360 following:
2361
2362 (1) Marking Bitmap. If there are gray objects that appear only
2363 on the bitmap (this happens either when dealing with an overflow
2364 or when the initial marking phase has simply marked the roots
2365 and didn't push them on the stack), then tasks claim heap
2366 regions whose bitmap they then scan to find gray objects. A
2367 global finger indicates where the end of the last claimed region
2368 is. A local finger indicates how far into the region a task has
2369 scanned. The two fingers are used to determine how to gray an
2370 object (i.e. whether simply marking it is OK, as it will be
2371 visited by a task in the future, or whether it needs to be also
2372 pushed on a stack).
2373
2374 (2) Local Queue. The local queue of the task which is accessed
2375 reasonably efficiently by the task. Other tasks can steal from
2376 it when they run out of work. Throughout the marking phase, a
2377 task attempts to keep its local queue short but not totally
2378 empty, so that entries are available for stealing by other
2379 tasks. Only when there is no more work, a task will totally
2380 drain its local queue.
2381
2382 (3) Global Mark Stack. This handles local queue overflow. During
2383 marking only sets of entries are moved between it and the local
2384 queues, as access to it requires a mutex and more fine-grain
2385 interaction with it which might cause contention. If it
2386 overflows, then the marking phase should restart and iterate
2387 over the bitmap to identify gray objects. Throughout the marking
2388 phase, tasks attempt to keep the global mark stack at a small
2389 length but not totally empty, so that entries are available for
2390 popping by other tasks. Only when there is no more work, tasks
2391 will totally drain the global mark stack.
2392
2393 (4) SATB Buffer Queue. This is where completed SATB buffers are
2394 made available. Buffers are regularly removed from this queue
2395 and scanned for roots, so that the queue doesn't get too
2396 long. During remark, all completed buffers are processed, as
2397 well as the filled in parts of any uncompleted buffers.
2398
2399 The do_marking_step() method tries to abort when the time target
2400 has been reached. There are a few other cases when the
2401 do_marking_step() method also aborts:
2402
2403 (1) When the marking phase has been aborted (after a Full GC).
2404
2405 (2) When a global overflow (on the global stack) has been
2406 triggered. Before the task aborts, it will actually sync up with
2407 the other tasks to ensure that all the marking data structures
2408 (local queues, stacks, fingers etc.) are re-initialized so that
2409 when do_marking_step() completes, the marking phase can
2410 immediately restart.
2411
2412 (3) When enough completed SATB buffers are available. The
2413 do_marking_step() method only tries to drain SATB buffers right
2414 at the beginning. So, if enough buffers are available, the
2415 marking step aborts and the SATB buffers are processed at
2416 the beginning of the next invocation.
2417
2418 (4) To yield. when we have to yield then we abort and yield
2419 right at the end of do_marking_step(). This saves us from a lot
2420 of hassle as, by yielding we might allow a Full GC. If this
2421 happens then objects will be compacted underneath our feet, the
2422 heap might shrink, etc. We save checking for this by just
2423 aborting and doing the yield right at the end.
2424
2425 From the above it follows that the do_marking_step() method should
2426 be called in a loop (or, otherwise, regularly) until it completes.
2427
2428 If a marking step completes without its has_aborted() flag being
2429 true, it means it has completed the current marking phase (and
2430 also all other marking tasks have done so and have all synced up).
2431
2432 A method called regular_clock_call() is invoked "regularly" (in
2433 sub ms intervals) throughout marking. It is this clock method that
2434 checks all the abort conditions which were mentioned above and
2435 decides when the task should abort. A work-based scheme is used to
2436 trigger this clock method: when the number of object words the
2437 marking phase has scanned or the number of references the marking
2438 phase has visited reach a given limit. Additional invocations to
2439 the method clock have been planted in a few other strategic places
2440 too. The initial reason for the clock method was to avoid calling
2441 vtime too regularly, as it is quite expensive. So, once it was in
2442 place, it was natural to piggy-back all the other conditions on it
2443 too and not constantly check them throughout the code.
2444
2445 If do_termination is true then do_marking_step will enter its
2446 termination protocol.
2447
2448 The value of is_serial must be true when do_marking_step is being
2449 called serially (i.e. by the VMThread) and do_marking_step should
2450 skip any synchronization in the termination and overflow code.
2451 Examples include the serial remark code and the serial reference
2452 processing closures.
2453
2454 The value of is_serial must be false when do_marking_step is
2455 being called by any of the worker threads in a work gang.
2456 Examples include the concurrent marking code (CMMarkingTask),
2457 the MT remark code, and the MT reference processing closures.
2458
2459 *****************************************************************************/
2460
2461 void G1CMTask::do_marking_step(double time_target_ms,
2462 bool do_termination,
2463 bool is_serial) {
2464 assert(time_target_ms >= 1.0, "minimum granularity is 1ms");
2465
2466 _start_time_ms = os::elapsedVTime() * 1000.0;
2467
2468 // If do_stealing is true then do_marking_step will attempt to
2469 // steal work from the other G1CMTasks. It only makes sense to
2470 // enable stealing when the termination protocol is enabled
2471 // and do_marking_step() is not being called serially.
2472 bool do_stealing = do_termination && !is_serial;
2473
2474 double diff_prediction_ms = _g1h->g1_policy()->predictor().get_new_prediction(&_marking_step_diffs_ms);
2475 _time_target_ms = time_target_ms - diff_prediction_ms;
2476
2477 // set up the variables that are used in the work-based scheme to
2478 // call the regular clock method
2479 _words_scanned = 0;
2480 _refs_reached = 0;
2481 recalculate_limits();
2482
2483 // clear all flags
2484 clear_has_aborted();
2485 _has_timed_out = false;
2486 _draining_satb_buffers = false;
2487
2488 ++_calls;
2489
2490 // Set up the bitmap and oop closures. Anything that uses them is
2491 // eventually called from this method, so it is OK to allocate these
2492 // statically.
2493 G1CMBitMapClosure bitmap_closure(this, _cm);
2494 G1CMOopClosure cm_oop_closure(_g1h, this);
2495 set_cm_oop_closure(&cm_oop_closure);
2496
2497 if (_cm->has_overflown()) {
2498 // This can happen if the mark stack overflows during a GC pause
2499 // and this task, after a yield point, restarts. We have to abort
2500 // as we need to get into the overflow protocol which happens
2501 // right at the end of this task.
2502 set_has_aborted();
2503 }
2504
2505 // First drain any available SATB buffers. After this, we will not
2506 // look at SATB buffers before the next invocation of this method.
2507 // If enough completed SATB buffers are queued up, the regular clock
2508 // will abort this task so that it restarts.
2509 drain_satb_buffers();
2510 // ...then partially drain the local queue and the global stack
2511 drain_local_queue(true);
2512 drain_global_stack(true);
2513
2514 do {
2515 if (!has_aborted() && _curr_region != NULL) {
2516 // This means that we're already holding on to a region.
2517 assert(_finger != NULL, "if region is not NULL, then the finger "
2518 "should not be NULL either");
2519
2520 // We might have restarted this task after an evacuation pause
2521 // which might have evacuated the region we're holding on to
2522 // underneath our feet. Let's read its limit again to make sure
2523 // that we do not iterate over a region of the heap that
2524 // contains garbage (update_region_limit() will also move
2525 // _finger to the start of the region if it is found empty).
2526 update_region_limit();
2527 // We will start from _finger not from the start of the region,
2528 // as we might be restarting this task after aborting half-way
2529 // through scanning this region. In this case, _finger points to
2530 // the address where we last found a marked object. If this is a
2531 // fresh region, _finger points to start().
2532 MemRegion mr = MemRegion(_finger, _region_limit);
2533
2534 assert(!_curr_region->is_humongous() || mr.start() == _curr_region->bottom(),
2535 "humongous regions should go around loop once only");
2536
2537 // Some special cases:
2538 // If the memory region is empty, we can just give up the region.
2539 // If the current region is humongous then we only need to check
2540 // the bitmap for the bit associated with the start of the object,
2541 // scan the object if it's live, and give up the region.
2542 // Otherwise, let's iterate over the bitmap of the part of the region
2543 // that is left.
2544 // If the iteration is successful, give up the region.
2545 if (mr.is_empty()) {
2546 giveup_current_region();
2547 regular_clock_call();
2548 } else if (_curr_region->is_humongous() && mr.start() == _curr_region->bottom()) {
2549 if (_next_mark_bitmap->is_marked(mr.start())) {
2550 // The object is marked - apply the closure
2551 bitmap_closure.do_addr(mr.start());
2552 }
2553 // Even if this task aborted while scanning the humongous object
2554 // we can (and should) give up the current region.
2555 giveup_current_region();
2556 regular_clock_call();
2557 } else if (_next_mark_bitmap->iterate(&bitmap_closure, mr)) {
2558 giveup_current_region();
2559 regular_clock_call();
2560 } else {
2561 assert(has_aborted(), "currently the only way to do so");
2562 // The only way to abort the bitmap iteration is to return
2563 // false from the do_bit() method. However, inside the
2564 // do_bit() method we move the _finger to point to the
2565 // object currently being looked at. So, if we bail out, we
2566 // have definitely set _finger to something non-null.
2567 assert(_finger != NULL, "invariant");
2568
2569 // Region iteration was actually aborted. So now _finger
2570 // points to the address of the object we last scanned. If we
2571 // leave it there, when we restart this task, we will rescan
2572 // the object. It is easy to avoid this. We move the finger by
2573 // enough to point to the next possible object header.
2574 assert(_finger < _region_limit, "invariant");
2575 HeapWord* const new_finger = _finger + ((oop)_finger)->size();
2576 // Check if bitmap iteration was aborted while scanning the last object
2577 if (new_finger >= _region_limit) {
2578 giveup_current_region();
2579 } else {
2580 move_finger_to(new_finger);
2581 }
2582 }
2583 }
2584 // At this point we have either completed iterating over the
2585 // region we were holding on to, or we have aborted.
2586
2587 // We then partially drain the local queue and the global stack.
2588 // (Do we really need this?)
2589 drain_local_queue(true);
2590 drain_global_stack(true);
2591
2592 // Read the note on the claim_region() method on why it might
2593 // return NULL with potentially more regions available for
2594 // claiming and why we have to check out_of_regions() to determine
2595 // whether we're done or not.
2596 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) {
2597 // We are going to try to claim a new region. We should have
2598 // given up on the previous one.
2599 // Separated the asserts so that we know which one fires.
2600 assert(_curr_region == NULL, "invariant");
2601 assert(_finger == NULL, "invariant");
2602 assert(_region_limit == NULL, "invariant");
2603 HeapRegion* claimed_region = _cm->claim_region(_worker_id);
2604 if (claimed_region != NULL) {
2605 // Yes, we managed to claim one
2606 setup_for_region(claimed_region);
2607 assert(_curr_region == claimed_region, "invariant");
2608 }
2609 // It is important to call the regular clock here. It might take
2610 // a while to claim a region if, for example, we hit a large
2611 // block of empty regions. So we need to call the regular clock
2612 // method once round the loop to make sure it's called
2613 // frequently enough.
2614 regular_clock_call();
2615 }
2616
2617 if (!has_aborted() && _curr_region == NULL) {
2618 assert(_cm->out_of_regions(),
2619 "at this point we should be out of regions");
2620 }
2621 } while ( _curr_region != NULL && !has_aborted());
2622
2623 if (!has_aborted()) {
2624 // We cannot check whether the global stack is empty, since other
2625 // tasks might be pushing objects to it concurrently.
2626 assert(_cm->out_of_regions(),
2627 "at this point we should be out of regions");
2628 // Try to reduce the number of available SATB buffers so that
2629 // remark has less work to do.
2630 drain_satb_buffers();
2631 }
2632
2633 // Since we've done everything else, we can now totally drain the
2634 // local queue and global stack.
2635 drain_local_queue(false);
2636 drain_global_stack(false);
2637
2638 // Attempt at work stealing from other task's queues.
2639 if (do_stealing && !has_aborted()) {
2640 // We have not aborted. This means that we have finished all that
2641 // we could. Let's try to do some stealing...
2642
2643 // We cannot check whether the global stack is empty, since other
2644 // tasks might be pushing objects to it concurrently.
2645 assert(_cm->out_of_regions() && _task_queue->size() == 0,
2646 "only way to reach here");
2647 while (!has_aborted()) {
2648 G1TaskQueueEntry entry;
2649 if (_cm->try_stealing(_worker_id, &_hash_seed, entry)) {
2650 scan_task_entry(entry);
2651
2652 // And since we're towards the end, let's totally drain the
2653 // local queue and global stack.
2654 drain_local_queue(false);
2655 drain_global_stack(false);
2656 } else {
2657 break;
2658 }
2659 }
2660 }
2661
2662 // We still haven't aborted. Now, let's try to get into the
2663 // termination protocol.
2664 if (do_termination && !has_aborted()) {
2665 // We cannot check whether the global stack is empty, since other
2666 // tasks might be concurrently pushing objects on it.
2667 // Separated the asserts so that we know which one fires.
2668 assert(_cm->out_of_regions(), "only way to reach here");
2669 assert(_task_queue->size() == 0, "only way to reach here");
2670 _termination_start_time_ms = os::elapsedVTime() * 1000.0;
2671
2672 // The G1CMTask class also extends the TerminatorTerminator class,
2673 // hence its should_exit_termination() method will also decide
2674 // whether to exit the termination protocol or not.
2675 bool finished = (is_serial ||
2676 _cm->terminator()->offer_termination(this));
2677 double termination_end_time_ms = os::elapsedVTime() * 1000.0;
2678 _termination_time_ms +=
2679 termination_end_time_ms - _termination_start_time_ms;
2680
2681 if (finished) {
2682 // We're all done.
2683
2684 // We can now guarantee that the global stack is empty, since
2685 // all other tasks have finished. We separated the guarantees so
2686 // that, if a condition is false, we can immediately find out
2687 // which one.
2688 guarantee(_cm->out_of_regions(), "only way to reach here");
2689 guarantee(_cm->mark_stack_empty(), "only way to reach here");
2690 guarantee(_task_queue->size() == 0, "only way to reach here");
2691 guarantee(!_cm->has_overflown(), "only way to reach here");
2692 } else {
2693 // Apparently there's more work to do. Let's abort this task. It
2694 // will restart it and we can hopefully find more things to do.
2695 set_has_aborted();
2696 }
2697 }
2698
2699 // Mainly for debugging purposes to make sure that a pointer to the
2700 // closure which was statically allocated in this frame doesn't
2701 // escape it by accident.
2702 set_cm_oop_closure(NULL);
2703 double end_time_ms = os::elapsedVTime() * 1000.0;
2704 double elapsed_time_ms = end_time_ms - _start_time_ms;
2705 // Update the step history.
2706 _step_times_ms.add(elapsed_time_ms);
2707
2708 if (has_aborted()) {
2709 // The task was aborted for some reason.
2710 if (_has_timed_out) {
2711 double diff_ms = elapsed_time_ms - _time_target_ms;
2712 // Keep statistics of how well we did with respect to hitting
2713 // our target only if we actually timed out (if we aborted for
2714 // other reasons, then the results might get skewed).
2715 _marking_step_diffs_ms.add(diff_ms);
2716 }
2717
2718 if (_cm->has_overflown()) {
2719 // This is the interesting one. We aborted because a global
2720 // overflow was raised. This means we have to restart the
2721 // marking phase and start iterating over regions. However, in
2722 // order to do this we have to make sure that all tasks stop
2723 // what they are doing and re-initialize in a safe manner. We
2724 // will achieve this with the use of two barrier sync points.
2725
2726 if (!is_serial) {
2727 // We only need to enter the sync barrier if being called
2728 // from a parallel context
2729 _cm->enter_first_sync_barrier(_worker_id);
2730
2731 // When we exit this sync barrier we know that all tasks have
2732 // stopped doing marking work. So, it's now safe to
2733 // re-initialize our data structures.
2734 }
2735
2736 clear_region_fields();
2737 flush_mark_stats_cache();
2738
2739 if (!is_serial) {
2740 // If we're executing the concurrent phase of marking, reset the marking
2741 // state; otherwise the marking state is reset after reference processing,
2742 // during the remark pause.
2743 // If we reset here as a result of an overflow during the remark we will
2744 // see assertion failures from any subsequent set_concurrency_and_phase()
2745 // calls.
2746 if (_cm->concurrent() && _worker_id == 0) {
2747 // Worker 0 is responsible for clearing the global data structures because
2748 // of an overflow. During STW we should not clear the overflow flag (in
2749 // G1ConcurrentMark::reset_marking_state()) since we rely on it being true when we exit
2750 // method to abort the pause and restart concurrent marking.
2751 _cm->reset_marking_for_restart();
2752
2753 log_info(gc, marking)("Concurrent Mark reset for overflow");
2754 }
2755
2756 // ...and enter the second barrier.
2757 _cm->enter_second_sync_barrier(_worker_id);
2758 }
2759 // At this point, if we're during the concurrent phase of
2760 // marking, everything has been re-initialized and we're
2761 // ready to restart.
2762 }
2763 }
2764 }
2765
2766 G1CMTask::G1CMTask(uint worker_id,
2767 G1ConcurrentMark* cm,
2768 G1CMTaskQueue* task_queue,
2769 G1RegionMarkStats* mark_stats,
2770 uint max_regions) :
2771 _objArray_processor(this),
2772 _worker_id(worker_id),
2773 _g1h(G1CollectedHeap::heap()),
2774 _cm(cm),
2775 _next_mark_bitmap(NULL),
2776 _task_queue(task_queue),
2777 _mark_stats_cache(mark_stats, max_regions, RegionMarkStatsCacheSize),
2778 _calls(0),
2779 _time_target_ms(0.0),
2780 _start_time_ms(0.0),
2781 _cm_oop_closure(NULL),
2782 _curr_region(NULL),
2783 _finger(NULL),
2784 _region_limit(NULL),
2785 _words_scanned(0),
2786 _words_scanned_limit(0),
2787 _real_words_scanned_limit(0),
2788 _refs_reached(0),
2789 _refs_reached_limit(0),
2790 _real_refs_reached_limit(0),
2791 _hash_seed(17),
2792 _has_aborted(false),
2793 _has_timed_out(false),
2794 _draining_satb_buffers(false),
2795 _step_times_ms(),
2796 _elapsed_time_ms(0.0),
2797 _termination_time_ms(0.0),
2798 _termination_start_time_ms(0.0),
2799 _marking_step_diffs_ms()
2800 {
2801 guarantee(task_queue != NULL, "invariant");
2802
2803 _marking_step_diffs_ms.add(0.5);
2804 }
2805
2806 // These are formatting macros that are used below to ensure
2807 // consistent formatting. The *_H_* versions are used to format the
2808 // header for a particular value and they should be kept consistent
2809 // with the corresponding macro. Also note that most of the macros add
2810 // the necessary white space (as a prefix) which makes them a bit
2811 // easier to compose.
2812
2813 // All the output lines are prefixed with this string to be able to
2814 // identify them easily in a large log file.
2815 #define G1PPRL_LINE_PREFIX "###"
2816
2817 #define G1PPRL_ADDR_BASE_FORMAT " " PTR_FORMAT "-" PTR_FORMAT
2818 #ifdef _LP64
2819 #define G1PPRL_ADDR_BASE_H_FORMAT " %37s"
2820 #else // _LP64
2821 #define G1PPRL_ADDR_BASE_H_FORMAT " %21s"
2822 #endif // _LP64
2823
2824 // For per-region info
2825 #define G1PPRL_TYPE_FORMAT " %-4s"
2826 #define G1PPRL_TYPE_H_FORMAT " %4s"
2827 #define G1PPRL_STATE_FORMAT " %-5s"
2828 #define G1PPRL_STATE_H_FORMAT " %5s"
2829 #define G1PPRL_BYTE_FORMAT " " SIZE_FORMAT_W(9)
2830 #define G1PPRL_BYTE_H_FORMAT " %9s"
2831 #define G1PPRL_DOUBLE_FORMAT " %14.1f"
2832 #define G1PPRL_DOUBLE_H_FORMAT " %14s"
2833
2834 // For summary info
2835 #define G1PPRL_SUM_ADDR_FORMAT(tag) " " tag ":" G1PPRL_ADDR_BASE_FORMAT
2836 #define G1PPRL_SUM_BYTE_FORMAT(tag) " " tag ": " SIZE_FORMAT
2837 #define G1PPRL_SUM_MB_FORMAT(tag) " " tag ": %1.2f MB"
2838 #define G1PPRL_SUM_MB_PERC_FORMAT(tag) G1PPRL_SUM_MB_FORMAT(tag) " / %1.2f %%"
2839
2840 G1PrintRegionLivenessInfoClosure::G1PrintRegionLivenessInfoClosure(const char* phase_name) :
2841 _total_used_bytes(0), _total_capacity_bytes(0),
2842 _total_prev_live_bytes(0), _total_next_live_bytes(0),
2843 _total_remset_bytes(0), _total_strong_code_roots_bytes(0)
2844 {
2845 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2846 MemRegion g1_reserved = g1h->g1_reserved();
2847 double now = os::elapsedTime();
2848
2849 // Print the header of the output.
2850 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX" PHASE %s @ %1.3f", phase_name, now);
2851 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX" HEAP"
2852 G1PPRL_SUM_ADDR_FORMAT("reserved")
2853 G1PPRL_SUM_BYTE_FORMAT("region-size"),
2854 p2i(g1_reserved.start()), p2i(g1_reserved.end()),
2855 HeapRegion::GrainBytes);
2856 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX);
2857 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
2858 G1PPRL_TYPE_H_FORMAT
2859 G1PPRL_ADDR_BASE_H_FORMAT
2860 G1PPRL_BYTE_H_FORMAT
2861 G1PPRL_BYTE_H_FORMAT
2862 G1PPRL_BYTE_H_FORMAT
2863 G1PPRL_DOUBLE_H_FORMAT
2864 G1PPRL_BYTE_H_FORMAT
2865 G1PPRL_STATE_H_FORMAT
2866 G1PPRL_BYTE_H_FORMAT,
2867 "type", "address-range",
2868 "used", "prev-live", "next-live", "gc-eff",
2869 "remset", "state", "code-roots");
2870 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
2871 G1PPRL_TYPE_H_FORMAT
2872 G1PPRL_ADDR_BASE_H_FORMAT
2873 G1PPRL_BYTE_H_FORMAT
2874 G1PPRL_BYTE_H_FORMAT
2875 G1PPRL_BYTE_H_FORMAT
2876 G1PPRL_DOUBLE_H_FORMAT
2877 G1PPRL_BYTE_H_FORMAT
2878 G1PPRL_STATE_H_FORMAT
2879 G1PPRL_BYTE_H_FORMAT,
2880 "", "",
2881 "(bytes)", "(bytes)", "(bytes)", "(bytes/ms)",
2882 "(bytes)", "", "(bytes)");
2883 }
2884
2885 bool G1PrintRegionLivenessInfoClosure::do_heap_region(HeapRegion* r) {
2886 const char* type = r->get_type_str();
2887 HeapWord* bottom = r->bottom();
2888 HeapWord* end = r->end();
2889 size_t capacity_bytes = r->capacity();
2890 size_t used_bytes = r->used();
2891 size_t prev_live_bytes = r->live_bytes();
2892 size_t next_live_bytes = r->next_live_bytes();
2893 double gc_eff = r->gc_efficiency();
2894 size_t remset_bytes = r->rem_set()->mem_size();
2895 size_t strong_code_roots_bytes = r->rem_set()->strong_code_roots_mem_size();
2896 const char* remset_type = r->rem_set()->get_short_state_str();
2897
2898 _total_used_bytes += used_bytes;
2899 _total_capacity_bytes += capacity_bytes;
2900 _total_prev_live_bytes += prev_live_bytes;
2901 _total_next_live_bytes += next_live_bytes;
2902 _total_remset_bytes += remset_bytes;
2903 _total_strong_code_roots_bytes += strong_code_roots_bytes;
2904
2905 // Print a line for this particular region.
2906 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
2907 G1PPRL_TYPE_FORMAT
2908 G1PPRL_ADDR_BASE_FORMAT
2909 G1PPRL_BYTE_FORMAT
2910 G1PPRL_BYTE_FORMAT
2911 G1PPRL_BYTE_FORMAT
2912 G1PPRL_DOUBLE_FORMAT
2913 G1PPRL_BYTE_FORMAT
2914 G1PPRL_STATE_FORMAT
2915 G1PPRL_BYTE_FORMAT,
2916 type, p2i(bottom), p2i(end),
2917 used_bytes, prev_live_bytes, next_live_bytes, gc_eff,
2918 remset_bytes, remset_type, strong_code_roots_bytes);
2919
2920 return false;
2921 }
2922
2923 G1PrintRegionLivenessInfoClosure::~G1PrintRegionLivenessInfoClosure() {
2924 // add static memory usages to remembered set sizes
2925 _total_remset_bytes += HeapRegionRemSet::fl_mem_size() + HeapRegionRemSet::static_mem_size();
2926 // Print the footer of the output.
2927 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX);
2928 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
2929 " SUMMARY"
2930 G1PPRL_SUM_MB_FORMAT("capacity")
2931 G1PPRL_SUM_MB_PERC_FORMAT("used")
2932 G1PPRL_SUM_MB_PERC_FORMAT("prev-live")
2933 G1PPRL_SUM_MB_PERC_FORMAT("next-live")
2934 G1PPRL_SUM_MB_FORMAT("remset")
2935 G1PPRL_SUM_MB_FORMAT("code-roots"),
2936 bytes_to_mb(_total_capacity_bytes),
2937 bytes_to_mb(_total_used_bytes),
2938 percent_of(_total_used_bytes, _total_capacity_bytes),
2939 bytes_to_mb(_total_prev_live_bytes),
2940 percent_of(_total_prev_live_bytes, _total_capacity_bytes),
2941 bytes_to_mb(_total_next_live_bytes),
2942 percent_of(_total_next_live_bytes, _total_capacity_bytes),
2943 bytes_to_mb(_total_remset_bytes),
2944 bytes_to_mb(_total_strong_code_roots_bytes));
2945 }