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/g1CollectedHeap.inline.hpp"
30 #include "gc/g1/g1CollectorState.hpp"
31 #include "gc/g1/g1ConcurrentMark.inline.hpp"
32 #include "gc/g1/g1ConcurrentMarkThread.inline.hpp"
33 #include "gc/g1/g1HeapVerifier.hpp"
34 #include "gc/g1/g1OopClosures.inline.hpp"
35 #include "gc/g1/g1Policy.hpp"
36 #include "gc/g1/g1RegionMarkStatsCache.inline.hpp"
37 #include "gc/g1/g1StringDedup.hpp"
38 #include "gc/g1/heapRegion.inline.hpp"
39 #include "gc/g1/heapRegionRemSet.hpp"
40 #include "gc/g1/heapRegionSet.inline.hpp"
41 #include "gc/shared/adaptiveSizePolicy.hpp"
42 #include "gc/shared/gcId.hpp"
43 #include "gc/shared/gcTimer.hpp"
44 #include "gc/shared/gcTrace.hpp"
45 #include "gc/shared/gcTraceTime.inline.hpp"
46 #include "gc/shared/genOopClosures.inline.hpp"
47 #include "gc/shared/referencePolicy.hpp"
48 #include "gc/shared/strongRootsScope.hpp"
49 #include "gc/shared/suspendibleThreadSet.hpp"
50 #include "gc/shared/taskqueue.inline.hpp"
51 #include "gc/shared/vmGCOperations.hpp"
52 #include "gc/shared/weakProcessor.hpp"
53 #include "include/jvm.h"
54 #include "logging/log.hpp"
55 #include "memory/allocation.hpp"
56 #include "memory/resourceArea.hpp"
57 #include "oops/access.inline.hpp"
58 #include "oops/oop.inline.hpp"
59 #include "runtime/atomic.hpp"
60 #include "runtime/handles.inline.hpp"
61 #include "runtime/java.hpp"
62 #include "runtime/prefetch.inline.hpp"
63 #include "services/memTracker.hpp"
64 #include "utilities/align.hpp"
65 #include "utilities/growableArray.hpp"
66
67 bool G1CMBitMapClosure::do_addr(HeapWord* const addr) {
68 assert(addr < _cm->finger(), "invariant");
69 assert(addr >= _task->finger(), "invariant");
70
71 // We move that task's local finger along.
72 _task->move_finger_to(addr);
73
74 _task->scan_task_entry(G1TaskQueueEntry::from_oop(oop(addr)));
75 // we only partially drain the local queue and global stack
76 _task->drain_local_queue(true);
77 _task->drain_global_stack(true);
78
79 // if the has_aborted flag has been raised, we need to bail out of
80 // the iteration
81 return !_task->has_aborted();
82 }
83
84 G1CMMarkStack::G1CMMarkStack() :
85 _max_chunk_capacity(0),
86 _base(NULL),
87 _chunk_capacity(0) {
88 set_empty();
89 }
90
91 bool G1CMMarkStack::resize(size_t new_capacity) {
92 assert(is_empty(), "Only resize when stack is empty.");
93 assert(new_capacity <= _max_chunk_capacity,
94 "Trying to resize stack to " SIZE_FORMAT " chunks when the maximum is " SIZE_FORMAT, new_capacity, _max_chunk_capacity);
95
96 TaskQueueEntryChunk* new_base = MmapArrayAllocator<TaskQueueEntryChunk>::allocate_or_null(new_capacity, mtGC);
97
98 if (new_base == NULL) {
99 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));
100 return false;
101 }
102 // Release old mapping.
103 if (_base != NULL) {
104 MmapArrayAllocator<TaskQueueEntryChunk>::free(_base, _chunk_capacity);
105 }
106
107 _base = new_base;
108 _chunk_capacity = new_capacity;
109 set_empty();
110
111 return true;
112 }
113
114 size_t G1CMMarkStack::capacity_alignment() {
115 return (size_t)lcm(os::vm_allocation_granularity(), sizeof(TaskQueueEntryChunk)) / sizeof(G1TaskQueueEntry);
116 }
117
118 bool G1CMMarkStack::initialize(size_t initial_capacity, size_t max_capacity) {
119 guarantee(_max_chunk_capacity == 0, "G1CMMarkStack already initialized.");
120
121 size_t const TaskEntryChunkSizeInVoidStar = sizeof(TaskQueueEntryChunk) / sizeof(G1TaskQueueEntry);
122
123 _max_chunk_capacity = align_up(max_capacity, capacity_alignment()) / TaskEntryChunkSizeInVoidStar;
124 size_t initial_chunk_capacity = align_up(initial_capacity, capacity_alignment()) / TaskEntryChunkSizeInVoidStar;
125
126 guarantee(initial_chunk_capacity <= _max_chunk_capacity,
127 "Maximum chunk capacity " SIZE_FORMAT " smaller than initial capacity " SIZE_FORMAT,
128 _max_chunk_capacity,
129 initial_chunk_capacity);
130
131 log_debug(gc)("Initialize mark stack with " SIZE_FORMAT " chunks, maximum " SIZE_FORMAT,
132 initial_chunk_capacity, _max_chunk_capacity);
133
134 return resize(initial_chunk_capacity);
135 }
136
137 void G1CMMarkStack::expand() {
138 if (_chunk_capacity == _max_chunk_capacity) {
139 log_debug(gc)("Can not expand overflow mark stack further, already at maximum capacity of " SIZE_FORMAT " chunks.", _chunk_capacity);
140 return;
141 }
142 size_t old_capacity = _chunk_capacity;
143 // Double capacity if possible
144 size_t new_capacity = MIN2(old_capacity * 2, _max_chunk_capacity);
145
146 if (resize(new_capacity)) {
147 log_debug(gc)("Expanded mark stack capacity from " SIZE_FORMAT " to " SIZE_FORMAT " chunks",
148 old_capacity, new_capacity);
149 } else {
150 log_warning(gc)("Failed to expand mark stack capacity from " SIZE_FORMAT " to " SIZE_FORMAT " chunks",
151 old_capacity, new_capacity);
152 }
153 }
154
155 G1CMMarkStack::~G1CMMarkStack() {
156 if (_base != NULL) {
157 MmapArrayAllocator<TaskQueueEntryChunk>::free(_base, _chunk_capacity);
158 }
159 }
160
161 void G1CMMarkStack::add_chunk_to_list(TaskQueueEntryChunk* volatile* list, TaskQueueEntryChunk* elem) {
162 elem->next = *list;
163 *list = elem;
164 }
165
166 void G1CMMarkStack::add_chunk_to_chunk_list(TaskQueueEntryChunk* elem) {
167 MutexLockerEx x(MarkStackChunkList_lock, Mutex::_no_safepoint_check_flag);
168 add_chunk_to_list(&_chunk_list, elem);
169 _chunks_in_chunk_list++;
170 }
171
172 void G1CMMarkStack::add_chunk_to_free_list(TaskQueueEntryChunk* elem) {
173 MutexLockerEx x(MarkStackFreeList_lock, Mutex::_no_safepoint_check_flag);
174 add_chunk_to_list(&_free_list, elem);
175 }
176
177 G1CMMarkStack::TaskQueueEntryChunk* G1CMMarkStack::remove_chunk_from_list(TaskQueueEntryChunk* volatile* list) {
178 TaskQueueEntryChunk* result = *list;
179 if (result != NULL) {
180 *list = (*list)->next;
181 }
182 return result;
183 }
184
185 G1CMMarkStack::TaskQueueEntryChunk* G1CMMarkStack::remove_chunk_from_chunk_list() {
186 MutexLockerEx x(MarkStackChunkList_lock, Mutex::_no_safepoint_check_flag);
187 TaskQueueEntryChunk* result = remove_chunk_from_list(&_chunk_list);
188 if (result != NULL) {
189 _chunks_in_chunk_list--;
190 }
191 return result;
192 }
193
194 G1CMMarkStack::TaskQueueEntryChunk* G1CMMarkStack::remove_chunk_from_free_list() {
195 MutexLockerEx x(MarkStackFreeList_lock, Mutex::_no_safepoint_check_flag);
196 return remove_chunk_from_list(&_free_list);
197 }
198
199 G1CMMarkStack::TaskQueueEntryChunk* G1CMMarkStack::allocate_new_chunk() {
200 // This dirty read of _hwm is okay because we only ever increase the _hwm in parallel code.
201 // Further this limits _hwm to a value of _chunk_capacity + #threads, avoiding
202 // wraparound of _hwm.
203 if (_hwm >= _chunk_capacity) {
204 return NULL;
205 }
206
207 size_t cur_idx = Atomic::add(1u, &_hwm) - 1;
208 if (cur_idx >= _chunk_capacity) {
209 return NULL;
210 }
211
212 TaskQueueEntryChunk* result = ::new (&_base[cur_idx]) TaskQueueEntryChunk;
213 result->next = NULL;
214 return result;
215 }
216
217 bool G1CMMarkStack::par_push_chunk(G1TaskQueueEntry* ptr_arr) {
218 // Get a new chunk.
219 TaskQueueEntryChunk* new_chunk = remove_chunk_from_free_list();
220
221 if (new_chunk == NULL) {
222 // Did not get a chunk from the free list. Allocate from backing memory.
223 new_chunk = allocate_new_chunk();
224
225 if (new_chunk == NULL) {
226 return false;
227 }
228 }
229
230 Copy::conjoint_memory_atomic(ptr_arr, new_chunk->data, EntriesPerChunk * sizeof(G1TaskQueueEntry));
231
232 add_chunk_to_chunk_list(new_chunk);
233
234 return true;
235 }
236
237 bool G1CMMarkStack::par_pop_chunk(G1TaskQueueEntry* ptr_arr) {
238 TaskQueueEntryChunk* cur = remove_chunk_from_chunk_list();
239
240 if (cur == NULL) {
241 return false;
242 }
243
244 Copy::conjoint_memory_atomic(cur->data, ptr_arr, EntriesPerChunk * sizeof(G1TaskQueueEntry));
245
246 add_chunk_to_free_list(cur);
247 return true;
248 }
249
250 void G1CMMarkStack::set_empty() {
251 _chunks_in_chunk_list = 0;
252 _hwm = 0;
253 _chunk_list = NULL;
254 _free_list = NULL;
255 }
256
257 G1CMRootRegions::G1CMRootRegions() :
258 _survivors(NULL), _cm(NULL), _scan_in_progress(false),
259 _should_abort(false), _claimed_survivor_index(0) { }
260
261 void G1CMRootRegions::init(const G1SurvivorRegions* survivors, G1ConcurrentMark* cm) {
262 _survivors = survivors;
263 _cm = cm;
264 }
265
266 void G1CMRootRegions::prepare_for_scan() {
267 assert(!scan_in_progress(), "pre-condition");
268
269 // Currently, only survivors can be root regions.
270 _claimed_survivor_index = 0;
271 _scan_in_progress = _survivors->regions()->is_nonempty();
272 _should_abort = false;
273 }
274
275 HeapRegion* G1CMRootRegions::claim_next() {
276 if (_should_abort) {
277 // If someone has set the should_abort flag, we return NULL to
278 // force the caller to bail out of their loop.
279 return NULL;
280 }
281
282 // Currently, only survivors can be root regions.
283 const GrowableArray<HeapRegion*>* survivor_regions = _survivors->regions();
284
285 int claimed_index = Atomic::add(1, &_claimed_survivor_index) - 1;
286 if (claimed_index < survivor_regions->length()) {
287 return survivor_regions->at(claimed_index);
288 }
289 return NULL;
290 }
291
292 uint G1CMRootRegions::num_root_regions() const {
293 return (uint)_survivors->regions()->length();
294 }
295
296 void G1CMRootRegions::notify_scan_done() {
297 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
298 _scan_in_progress = false;
299 RootRegionScan_lock->notify_all();
300 }
301
302 void G1CMRootRegions::cancel_scan() {
303 notify_scan_done();
304 }
305
306 void G1CMRootRegions::scan_finished() {
307 assert(scan_in_progress(), "pre-condition");
308
309 // Currently, only survivors can be root regions.
310 if (!_should_abort) {
311 assert(_claimed_survivor_index >= 0, "otherwise comparison is invalid: %d", _claimed_survivor_index);
312 assert((uint)_claimed_survivor_index >= _survivors->length(),
313 "we should have claimed all survivors, claimed index = %u, length = %u",
314 (uint)_claimed_survivor_index, _survivors->length());
315 }
316
317 notify_scan_done();
318 }
319
320 bool G1CMRootRegions::wait_until_scan_finished() {
321 if (!scan_in_progress()) {
322 return false;
323 }
324
325 {
326 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
327 while (scan_in_progress()) {
328 RootRegionScan_lock->wait(Mutex::_no_safepoint_check_flag);
329 }
330 }
331 return true;
332 }
333
334 // Returns the maximum number of workers to be used in a concurrent
335 // phase based on the number of GC workers being used in a STW
336 // phase.
337 static uint scale_concurrent_worker_threads(uint num_gc_workers) {
338 return MAX2((num_gc_workers + 2) / 4, 1U);
339 }
340
341 G1ConcurrentMark::G1ConcurrentMark(G1CollectedHeap* g1h,
342 G1RegionToSpaceMapper* prev_bitmap_storage,
343 G1RegionToSpaceMapper* next_bitmap_storage) :
344 // _cm_thread set inside the constructor
345 _g1h(g1h),
346 _completed_initialization(false),
347
348 _mark_bitmap_1(),
349 _mark_bitmap_2(),
350 _prev_mark_bitmap(&_mark_bitmap_1),
351 _next_mark_bitmap(&_mark_bitmap_2),
352
353 _heap(_g1h->reserved_region()),
354
355 _root_regions(),
356
357 _global_mark_stack(),
358
359 // _finger set in set_non_marking_state
360
361 _worker_id_offset(DirtyCardQueueSet::num_par_ids() + G1ConcRefinementThreads),
362 _max_num_tasks(ParallelGCThreads),
363 // _num_active_tasks set in set_non_marking_state()
364 // _tasks set inside the constructor
365
366 _task_queues(new G1CMTaskQueueSet((int) _max_num_tasks)),
367 _terminator(ParallelTaskTerminator((int) _max_num_tasks, _task_queues)),
368
369 _first_overflow_barrier_sync(),
370 _second_overflow_barrier_sync(),
371
372 _has_overflown(false),
373 _concurrent(false),
374 _has_aborted(false),
375 _restart_for_overflow(false),
376 _gc_timer_cm(new (ResourceObj::C_HEAP, mtGC) ConcurrentGCTimer()),
377 _gc_tracer_cm(new (ResourceObj::C_HEAP, mtGC) G1OldTracer()),
378
379 // _verbose_level set below
380
381 _init_times(),
382 _remark_times(),
383 _remark_mark_times(),
384 _remark_weak_ref_times(),
385 _cleanup_times(),
386 _total_cleanup_time(0.0),
387
388 _accum_task_vtime(NULL),
389
390 _concurrent_workers(NULL),
391 _num_concurrent_workers(0),
392 _max_concurrent_workers(0),
393
394 _region_mark_stats(NEW_C_HEAP_ARRAY(G1RegionMarkStats, _g1h->max_regions(), mtGC)),
395 _top_at_rebuild_starts(NEW_C_HEAP_ARRAY(HeapWord*, _g1h->max_regions(), mtGC))
396 {
397 _mark_bitmap_1.initialize(g1h->reserved_region(), prev_bitmap_storage);
398 _mark_bitmap_2.initialize(g1h->reserved_region(), next_bitmap_storage);
399
400 // Create & start ConcurrentMark thread.
401 _cm_thread = new G1ConcurrentMarkThread(this);
402 if (_cm_thread->osthread() == NULL) {
403 vm_shutdown_during_initialization("Could not create ConcurrentMarkThread");
404 }
405
406 assert(CGC_lock != NULL, "CGC_lock must be initialized");
407
408 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set();
409 satb_qs.set_buffer_size(G1SATBBufferSize);
410
411 _root_regions.init(_g1h->survivor(), this);
412
413 if (FLAG_IS_DEFAULT(ConcGCThreads) || ConcGCThreads == 0) {
414 // Calculate the number of concurrent worker threads by scaling
415 // the number of parallel GC threads.
416 uint marking_thread_num = scale_concurrent_worker_threads(ParallelGCThreads);
417 FLAG_SET_ERGO(uint, ConcGCThreads, marking_thread_num);
418 }
419
420 assert(ConcGCThreads > 0, "ConcGCThreads have been set.");
421 if (ConcGCThreads > ParallelGCThreads) {
422 log_warning(gc)("More ConcGCThreads (%u) than ParallelGCThreads (%u).",
423 ConcGCThreads, ParallelGCThreads);
424 return;
425 }
426
427 log_debug(gc)("ConcGCThreads: %u offset %u", ConcGCThreads, _worker_id_offset);
428 log_debug(gc)("ParallelGCThreads: %u", ParallelGCThreads);
429
430 _num_concurrent_workers = ConcGCThreads;
431 _max_concurrent_workers = _num_concurrent_workers;
432
433 _concurrent_workers = new WorkGang("G1 Conc", _max_concurrent_workers, false, true);
434 _concurrent_workers->initialize_workers();
435
436 if (FLAG_IS_DEFAULT(MarkStackSize)) {
437 size_t mark_stack_size =
438 MIN2(MarkStackSizeMax,
439 MAX2(MarkStackSize, (size_t) (_max_concurrent_workers * TASKQUEUE_SIZE)));
440 // Verify that the calculated value for MarkStackSize is in range.
441 // It would be nice to use the private utility routine from Arguments.
442 if (!(mark_stack_size >= 1 && mark_stack_size <= MarkStackSizeMax)) {
443 log_warning(gc)("Invalid value calculated for MarkStackSize (" SIZE_FORMAT "): "
444 "must be between 1 and " SIZE_FORMAT,
445 mark_stack_size, MarkStackSizeMax);
446 return;
447 }
448 FLAG_SET_ERGO(size_t, MarkStackSize, mark_stack_size);
449 } else {
450 // Verify MarkStackSize is in range.
451 if (FLAG_IS_CMDLINE(MarkStackSize)) {
452 if (FLAG_IS_DEFAULT(MarkStackSizeMax)) {
453 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
454 log_warning(gc)("Invalid value specified for MarkStackSize (" SIZE_FORMAT "): "
455 "must be between 1 and " SIZE_FORMAT,
456 MarkStackSize, MarkStackSizeMax);
457 return;
458 }
459 } else if (FLAG_IS_CMDLINE(MarkStackSizeMax)) {
460 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
461 log_warning(gc)("Invalid value specified for MarkStackSize (" SIZE_FORMAT ")"
462 " or for MarkStackSizeMax (" SIZE_FORMAT ")",
463 MarkStackSize, MarkStackSizeMax);
464 return;
465 }
466 }
467 }
468 }
469
470 if (!_global_mark_stack.initialize(MarkStackSize, MarkStackSizeMax)) {
471 vm_exit_during_initialization("Failed to allocate initial concurrent mark overflow mark stack.");
472 }
473
474 _tasks = NEW_C_HEAP_ARRAY(G1CMTask*, _max_num_tasks, mtGC);
475 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_num_tasks, mtGC);
476
477 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail
478 _num_active_tasks = _max_num_tasks;
479
480 for (uint i = 0; i < _max_num_tasks; ++i) {
481 G1CMTaskQueue* task_queue = new G1CMTaskQueue();
482 task_queue->initialize();
483 _task_queues->register_queue(i, task_queue);
484
485 _tasks[i] = new G1CMTask(i, this, task_queue, _region_mark_stats, _g1h->max_regions());
486
487 _accum_task_vtime[i] = 0.0;
488 }
489
490 reset_at_marking_complete();
491 _completed_initialization = true;
492 }
493
494 void G1ConcurrentMark::reset() {
495 _has_aborted = false;
496
497 reset_marking_for_restart();
498
499 // Reset all tasks, since different phases will use different number of active
500 // threads. So, it's easiest to have all of them ready.
501 for (uint i = 0; i < _max_num_tasks; ++i) {
502 _tasks[i]->reset(_next_mark_bitmap);
503 }
504
505 uint max_regions = _g1h->max_regions();
506 for (uint i = 0; i < max_regions; i++) {
507 _top_at_rebuild_starts[i] = NULL;
508 _region_mark_stats[i].clear();
509 }
510 }
511
512 void G1ConcurrentMark::clear_statistics_in_region(uint region_idx) {
513 for (uint j = 0; j < _max_num_tasks; ++j) {
514 _tasks[j]->clear_mark_stats_cache(region_idx);
515 }
516 _top_at_rebuild_starts[region_idx] = NULL;
517 _region_mark_stats[region_idx].clear();
518 }
519
520 void G1ConcurrentMark::clear_statistics(HeapRegion* r) {
521 uint const region_idx = r->hrm_index();
522 if (r->is_humongous()) {
523 assert(r->is_starts_humongous(), "Got humongous continues region here");
524 uint const size_in_regions = (uint)_g1h->humongous_obj_size_in_regions(oop(r->humongous_start_region()->bottom())->size());
525 for (uint j = region_idx; j < (region_idx + size_in_regions); j++) {
526 clear_statistics_in_region(j);
527 }
528 } else {
529 clear_statistics_in_region(region_idx);
530 }
531 }
532
533 void G1ConcurrentMark::humongous_object_eagerly_reclaimed(HeapRegion* r) {
534 assert_at_safepoint_on_vm_thread();
535
536 // Need to clear mark bit of the humongous object.
537 if (_next_mark_bitmap->is_marked(r->bottom())) {
538 _next_mark_bitmap->clear(r->bottom());
539 }
540
541 if (!_g1h->collector_state()->mark_or_rebuild_in_progress()) {
542 return;
543 }
544
545 // Clear any statistics about the region gathered so far.
546 clear_statistics(r);
547 }
548
549 void G1ConcurrentMark::reset_marking_for_restart() {
550 _global_mark_stack.set_empty();
551
552 // Expand the marking stack, if we have to and if we can.
553 if (has_overflown()) {
554 _global_mark_stack.expand();
555
556 uint max_regions = _g1h->max_regions();
557 for (uint i = 0; i < max_regions; i++) {
558 _region_mark_stats[i].clear_during_overflow();
559 }
560 }
561
562 clear_has_overflown();
563 _finger = _heap.start();
564
565 for (uint i = 0; i < _max_num_tasks; ++i) {
566 G1CMTaskQueue* queue = _task_queues->queue(i);
567 queue->set_empty();
568 }
569 }
570
571 void G1ConcurrentMark::set_concurrency(uint active_tasks) {
572 assert(active_tasks <= _max_num_tasks, "we should not have more");
573
574 _num_active_tasks = active_tasks;
575 // Need to update the three data structures below according to the
576 // number of active threads for this phase.
577 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues);
578 _first_overflow_barrier_sync.set_n_workers((int) active_tasks);
579 _second_overflow_barrier_sync.set_n_workers((int) active_tasks);
580 }
581
582 void G1ConcurrentMark::set_concurrency_and_phase(uint active_tasks, bool concurrent) {
583 set_concurrency(active_tasks);
584
585 _concurrent = concurrent;
586
587 if (!concurrent) {
588 // At this point we should be in a STW phase, and completed marking.
589 assert_at_safepoint_on_vm_thread();
590 assert(out_of_regions(),
591 "only way to get here: _finger: " PTR_FORMAT ", _heap_end: " PTR_FORMAT,
592 p2i(_finger), p2i(_heap.end()));
593 }
594 }
595
596 void G1ConcurrentMark::reset_at_marking_complete() {
597 // We set the global marking state to some default values when we're
598 // not doing marking.
599 reset_marking_for_restart();
600 _num_active_tasks = 0;
601 }
602
603 G1ConcurrentMark::~G1ConcurrentMark() {
604 FREE_C_HEAP_ARRAY(HeapWord*, _top_at_rebuild_starts);
605 FREE_C_HEAP_ARRAY(G1RegionMarkStats, _region_mark_stats);
606 // The G1ConcurrentMark instance is never freed.
607 ShouldNotReachHere();
608 }
609
610 class G1ClearBitMapTask : public AbstractGangTask {
611 public:
612 static size_t chunk_size() { return M; }
613
614 private:
615 // Heap region closure used for clearing the given mark bitmap.
616 class G1ClearBitmapHRClosure : public HeapRegionClosure {
617 private:
618 G1CMBitMap* _bitmap;
619 G1ConcurrentMark* _cm;
620 public:
621 G1ClearBitmapHRClosure(G1CMBitMap* bitmap, G1ConcurrentMark* cm) : HeapRegionClosure(), _cm(cm), _bitmap(bitmap) {
622 }
623
624 virtual bool do_heap_region(HeapRegion* r) {
625 size_t const chunk_size_in_words = G1ClearBitMapTask::chunk_size() / HeapWordSize;
626
627 HeapWord* cur = r->bottom();
628 HeapWord* const end = r->end();
629
630 while (cur < end) {
631 MemRegion mr(cur, MIN2(cur + chunk_size_in_words, end));
632 _bitmap->clear_range(mr);
633
634 cur += chunk_size_in_words;
635
636 // Abort iteration if after yielding the marking has been aborted.
637 if (_cm != NULL && _cm->do_yield_check() && _cm->has_aborted()) {
638 return true;
639 }
640 // Repeat the asserts from before the start of the closure. We will do them
641 // as asserts here to minimize their overhead on the product. However, we
642 // will have them as guarantees at the beginning / end of the bitmap
643 // clearing to get some checking in the product.
644 assert(_cm == NULL || _cm->cm_thread()->during_cycle(), "invariant");
645 assert(_cm == NULL || !G1CollectedHeap::heap()->collector_state()->mark_or_rebuild_in_progress(), "invariant");
646 }
647 assert(cur == end, "Must have completed iteration over the bitmap for region %u.", r->hrm_index());
648
649 return false;
650 }
651 };
652
653 G1ClearBitmapHRClosure _cl;
654 HeapRegionClaimer _hr_claimer;
655 bool _suspendible; // If the task is suspendible, workers must join the STS.
656
657 public:
658 G1ClearBitMapTask(G1CMBitMap* bitmap, G1ConcurrentMark* cm, uint n_workers, bool suspendible) :
659 AbstractGangTask("G1 Clear Bitmap"),
660 _cl(bitmap, suspendible ? cm : NULL),
661 _hr_claimer(n_workers),
662 _suspendible(suspendible)
663 { }
664
665 void work(uint worker_id) {
666 SuspendibleThreadSetJoiner sts_join(_suspendible);
667 G1CollectedHeap::heap()->heap_region_par_iterate_from_worker_offset(&_cl, &_hr_claimer, worker_id);
668 }
669
670 bool is_complete() {
671 return _cl.is_complete();
672 }
673 };
674
675 void G1ConcurrentMark::clear_bitmap(G1CMBitMap* bitmap, WorkGang* workers, bool may_yield) {
676 assert(may_yield || SafepointSynchronize::is_at_safepoint(), "Non-yielding bitmap clear only allowed at safepoint.");
677
678 size_t const num_bytes_to_clear = (HeapRegion::GrainBytes * _g1h->num_regions()) / G1CMBitMap::heap_map_factor();
679 size_t const num_chunks = align_up(num_bytes_to_clear, G1ClearBitMapTask::chunk_size()) / G1ClearBitMapTask::chunk_size();
680
681 uint const num_workers = (uint)MIN2(num_chunks, (size_t)workers->active_workers());
682
683 G1ClearBitMapTask cl(bitmap, this, num_workers, may_yield);
684
685 log_debug(gc, ergo)("Running %s with %u workers for " SIZE_FORMAT " work units.", cl.name(), num_workers, num_chunks);
686 workers->run_task(&cl, num_workers);
687 guarantee(!may_yield || cl.is_complete(), "Must have completed iteration when not yielding.");
688 }
689
690 void G1ConcurrentMark::cleanup_for_next_mark() {
691 // Make sure that the concurrent mark thread looks to still be in
692 // the current cycle.
693 guarantee(cm_thread()->during_cycle(), "invariant");
694
695 // We are finishing up the current cycle by clearing the next
696 // marking bitmap and getting it ready for the next cycle. During
697 // this time no other cycle can start. So, let's make sure that this
698 // is the case.
699 guarantee(!_g1h->collector_state()->mark_or_rebuild_in_progress(), "invariant");
700
701 clear_bitmap(_next_mark_bitmap, _concurrent_workers, true);
702
703 // Repeat the asserts from above.
704 guarantee(cm_thread()->during_cycle(), "invariant");
705 guarantee(!_g1h->collector_state()->mark_or_rebuild_in_progress(), "invariant");
706 }
707
708 void G1ConcurrentMark::clear_prev_bitmap(WorkGang* workers) {
709 assert_at_safepoint_on_vm_thread();
710 clear_bitmap(_prev_mark_bitmap, workers, false);
711 }
712
713 class CheckBitmapClearHRClosure : public HeapRegionClosure {
714 G1CMBitMap* _bitmap;
715 public:
716 CheckBitmapClearHRClosure(G1CMBitMap* bitmap) : _bitmap(bitmap) {
717 }
718
719 virtual bool do_heap_region(HeapRegion* r) {
720 // This closure can be called concurrently to the mutator, so we must make sure
721 // that the result of the getNextMarkedWordAddress() call is compared to the
722 // value passed to it as limit to detect any found bits.
723 // end never changes in G1.
724 HeapWord* end = r->end();
725 return _bitmap->get_next_marked_addr(r->bottom(), end) != end;
726 }
727 };
728
729 bool G1ConcurrentMark::next_mark_bitmap_is_clear() {
730 CheckBitmapClearHRClosure cl(_next_mark_bitmap);
731 _g1h->heap_region_iterate(&cl);
732 return cl.is_complete();
733 }
734
735 class NoteStartOfMarkHRClosure : public HeapRegionClosure {
736 public:
737 bool do_heap_region(HeapRegion* r) {
738 r->note_start_of_marking();
739 return false;
740 }
741 };
742
743 void G1ConcurrentMark::pre_initial_mark() {
744 // Initialize marking structures. This has to be done in a STW phase.
745 reset();
746
747 // For each region note start of marking.
748 NoteStartOfMarkHRClosure startcl;
749 _g1h->heap_region_iterate(&startcl);
750 }
751
752
753 void G1ConcurrentMark::post_initial_mark() {
754 // Start Concurrent Marking weak-reference discovery.
755 ReferenceProcessor* rp = _g1h->ref_processor_cm();
756 // enable ("weak") refs discovery
757 rp->enable_discovery();
758 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle
759
760 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
761 // This is the start of the marking cycle, we're expected all
762 // threads to have SATB queues with active set to false.
763 satb_mq_set.set_active_all_threads(true, /* new active value */
764 false /* expected_active */);
765
766 _root_regions.prepare_for_scan();
767
768 // update_g1_committed() will be called at the end of an evac pause
769 // when marking is on. So, it's also called at the end of the
770 // initial-mark pause to update the heap end, if the heap expands
771 // during it. No need to call it here.
772 }
773
774 /*
775 * Notice that in the next two methods, we actually leave the STS
776 * during the barrier sync and join it immediately afterwards. If we
777 * do not do this, the following deadlock can occur: one thread could
778 * be in the barrier sync code, waiting for the other thread to also
779 * sync up, whereas another one could be trying to yield, while also
780 * waiting for the other threads to sync up too.
781 *
782 * Note, however, that this code is also used during remark and in
783 * this case we should not attempt to leave / enter the STS, otherwise
784 * we'll either hit an assert (debug / fastdebug) or deadlock
785 * (product). So we should only leave / enter the STS if we are
786 * operating concurrently.
787 *
788 * Because the thread that does the sync barrier has left the STS, it
789 * is possible to be suspended for a Full GC or an evacuation pause
790 * could occur. This is actually safe, since the entering the sync
791 * barrier is one of the last things do_marking_step() does, and it
792 * doesn't manipulate any data structures afterwards.
793 */
794
795 void G1ConcurrentMark::enter_first_sync_barrier(uint worker_id) {
796 bool barrier_aborted;
797 {
798 SuspendibleThreadSetLeaver sts_leave(concurrent());
799 barrier_aborted = !_first_overflow_barrier_sync.enter();
800 }
801
802 // at this point everyone should have synced up and not be doing any
803 // more work
804
805 if (barrier_aborted) {
806 // If the barrier aborted we ignore the overflow condition and
807 // just abort the whole marking phase as quickly as possible.
808 return;
809 }
810 }
811
812 void G1ConcurrentMark::enter_second_sync_barrier(uint worker_id) {
813 SuspendibleThreadSetLeaver sts_leave(concurrent());
814 _second_overflow_barrier_sync.enter();
815
816 // at this point everything should be re-initialized and ready to go
817 }
818
819 class G1CMConcurrentMarkingTask : public AbstractGangTask {
820 G1ConcurrentMark* _cm;
821
822 public:
823 void work(uint worker_id) {
824 assert(Thread::current()->is_ConcurrentGC_thread(), "Not a concurrent GC thread");
825 ResourceMark rm;
826
827 double start_vtime = os::elapsedVTime();
828
829 {
830 SuspendibleThreadSetJoiner sts_join;
831
832 assert(worker_id < _cm->active_tasks(), "invariant");
833
834 G1CMTask* task = _cm->task(worker_id);
835 task->record_start_time();
836 if (!_cm->has_aborted()) {
837 do {
838 task->do_marking_step(G1ConcMarkStepDurationMillis,
839 true /* do_termination */,
840 false /* is_serial*/);
841
842 _cm->do_yield_check();
843 } while (!_cm->has_aborted() && task->has_aborted());
844 }
845 task->record_end_time();
846 guarantee(!task->has_aborted() || _cm->has_aborted(), "invariant");
847 }
848
849 double end_vtime = os::elapsedVTime();
850 _cm->update_accum_task_vtime(worker_id, end_vtime - start_vtime);
851 }
852
853 G1CMConcurrentMarkingTask(G1ConcurrentMark* cm) :
854 AbstractGangTask("Concurrent Mark"), _cm(cm) { }
855
856 ~G1CMConcurrentMarkingTask() { }
857 };
858
859 uint G1ConcurrentMark::calc_active_marking_workers() {
860 uint result = 0;
861 if (!UseDynamicNumberOfGCThreads ||
862 (!FLAG_IS_DEFAULT(ConcGCThreads) &&
863 !ForceDynamicNumberOfGCThreads)) {
864 result = _max_concurrent_workers;
865 } else {
866 result =
867 AdaptiveSizePolicy::calc_default_active_workers(_max_concurrent_workers,
868 1, /* Minimum workers */
869 _num_concurrent_workers,
870 Threads::number_of_non_daemon_threads());
871 // Don't scale the result down by scale_concurrent_workers() because
872 // that scaling has already gone into "_max_concurrent_workers".
873 }
874 assert(result > 0 && result <= _max_concurrent_workers,
875 "Calculated number of marking workers must be larger than zero and at most the maximum %u, but is %u",
876 _max_concurrent_workers, result);
877 return result;
878 }
879
880 void G1ConcurrentMark::scan_root_region(HeapRegion* hr, uint worker_id) {
881 // Currently, only survivors can be root regions.
882 assert(hr->next_top_at_mark_start() == hr->bottom(), "invariant");
883 G1RootRegionScanClosure cl(_g1h, this, worker_id);
884
885 const uintx interval = PrefetchScanIntervalInBytes;
886 HeapWord* curr = hr->bottom();
887 const HeapWord* end = hr->top();
888 while (curr < end) {
889 Prefetch::read(curr, interval);
890 oop obj = oop(curr);
891 int size = obj->oop_iterate_size(&cl);
892 assert(size == obj->size(), "sanity");
893 curr += size;
894 }
895 }
896
897 class G1CMRootRegionScanTask : public AbstractGangTask {
898 G1ConcurrentMark* _cm;
899 public:
900 G1CMRootRegionScanTask(G1ConcurrentMark* cm) :
901 AbstractGangTask("G1 Root Region Scan"), _cm(cm) { }
902
903 void work(uint worker_id) {
904 assert(Thread::current()->is_ConcurrentGC_thread(),
905 "this should only be done by a conc GC thread");
906
907 G1CMRootRegions* root_regions = _cm->root_regions();
908 HeapRegion* hr = root_regions->claim_next();
909 while (hr != NULL) {
910 _cm->scan_root_region(hr, worker_id);
911 hr = root_regions->claim_next();
912 }
913 }
914 };
915
916 void G1ConcurrentMark::scan_root_regions() {
917 // scan_in_progress() will have been set to true only if there was
918 // at least one root region to scan. So, if it's false, we
919 // should not attempt to do any further work.
920 if (root_regions()->scan_in_progress()) {
921 assert(!has_aborted(), "Aborting before root region scanning is finished not supported.");
922
923 _num_concurrent_workers = MIN2(calc_active_marking_workers(),
924 // We distribute work on a per-region basis, so starting
925 // more threads than that is useless.
926 root_regions()->num_root_regions());
927 assert(_num_concurrent_workers <= _max_concurrent_workers,
928 "Maximum number of marking threads exceeded");
929
930 G1CMRootRegionScanTask task(this);
931 log_debug(gc, ergo)("Running %s using %u workers for %u work units.",
932 task.name(), _num_concurrent_workers, root_regions()->num_root_regions());
933 _concurrent_workers->run_task(&task, _num_concurrent_workers);
934
935 // It's possible that has_aborted() is true here without actually
936 // aborting the survivor scan earlier. This is OK as it's
937 // mainly used for sanity checking.
938 root_regions()->scan_finished();
939 }
940 }
941
942 void G1ConcurrentMark::concurrent_cycle_start() {
943 _gc_timer_cm->register_gc_start();
944
945 _gc_tracer_cm->report_gc_start(GCCause::_no_gc /* first parameter is not used */, _gc_timer_cm->gc_start());
946
947 _g1h->trace_heap_before_gc(_gc_tracer_cm);
948 }
949
950 void G1ConcurrentMark::concurrent_cycle_end() {
951 _g1h->collector_state()->set_clearing_next_bitmap(false);
952
953 _g1h->trace_heap_after_gc(_gc_tracer_cm);
954
955 if (has_aborted()) {
956 log_info(gc, marking)("Concurrent Mark Abort");
957 _gc_tracer_cm->report_concurrent_mode_failure();
958 }
959
960 _gc_timer_cm->register_gc_end();
961
962 _gc_tracer_cm->report_gc_end(_gc_timer_cm->gc_end(), _gc_timer_cm->time_partitions());
963 }
964
965 void G1ConcurrentMark::mark_from_roots() {
966 _restart_for_overflow = false;
967
968 _num_concurrent_workers = calc_active_marking_workers();
969
970 uint active_workers = MAX2(1U, _num_concurrent_workers);
971
972 // Setting active workers is not guaranteed since fewer
973 // worker threads may currently exist and more may not be
974 // available.
975 active_workers = _concurrent_workers->update_active_workers(active_workers);
976 log_info(gc, task)("Using %u workers of %u for marking", active_workers, _concurrent_workers->total_workers());
977
978 // Parallel task terminator is set in "set_concurrency_and_phase()"
979 set_concurrency_and_phase(active_workers, true /* concurrent */);
980
981 G1CMConcurrentMarkingTask marking_task(this);
982 _concurrent_workers->run_task(&marking_task);
983 print_stats();
984 }
985
986 void G1ConcurrentMark::verify_during_pause(G1HeapVerifier::G1VerifyType type, VerifyOption vo, const char* caller) {
987 G1HeapVerifier* verifier = _g1h->verifier();
988
989 verifier->verify_region_sets_optional();
990
991 if (VerifyDuringGC) {
992 GCTraceTime(Debug, gc, phases) trace(caller, _gc_timer_cm);
993
994 size_t const BufLen = 512;
995 char buffer[BufLen];
996
997 jio_snprintf(buffer, BufLen, "During GC (%s)", caller);
998 verifier->verify(type, vo, buffer);
999 }
1000
1001 verifier->check_bitmaps(caller);
1002 }
1003
1004 class G1UpdateRemSetTrackingBeforeRebuild : public HeapRegionClosure {
1005 G1CollectedHeap* _g1h;
1006 G1ConcurrentMark* _cm;
1007
1008 uint _num_regions_selected_for_rebuild; // The number of regions actually selected for rebuild.
1009
1010 void update_remset_before_rebuild(HeapRegion * hr) {
1011 G1RemSetTrackingPolicy* tracking_policy = _g1h->g1_policy()->remset_tracker();
1012
1013 size_t live_bytes = _cm->liveness(hr->hrm_index()) * HeapWordSize;
1014 bool selected_for_rebuild = tracking_policy->update_before_rebuild(hr, live_bytes);
1015 if (selected_for_rebuild) {
1016 _num_regions_selected_for_rebuild++;
1017 }
1018 _cm->update_top_at_rebuild_start(hr);
1019 }
1020
1021 public:
1022 G1UpdateRemSetTrackingBeforeRebuild(G1CollectedHeap* g1h, G1ConcurrentMark* cm) :
1023 _g1h(g1h), _cm(cm), _num_regions_selected_for_rebuild(0) { }
1024
1025 virtual bool do_heap_region(HeapRegion* r) {
1026 update_remset_before_rebuild(r);
1027 return false;
1028 }
1029
1030 uint num_selected_for_rebuild() const { return _num_regions_selected_for_rebuild; }
1031 };
1032
1033 class G1UpdateRemSetTrackingAfterRebuild : public HeapRegionClosure {
1034 G1CollectedHeap* _g1h;
1035 public:
1036 G1UpdateRemSetTrackingAfterRebuild(G1CollectedHeap* g1h) : _g1h(g1h) { }
1037
1038 virtual bool do_heap_region(HeapRegion* r) {
1039 _g1h->g1_policy()->remset_tracker()->update_after_rebuild(r);
1040 return false;
1041 }
1042 };
1043
1044 void G1ConcurrentMark::remark() {
1045 assert_at_safepoint_on_vm_thread();
1046
1047 // If a full collection has happened, we should not continue. However we might
1048 // have ended up here as the Remark VM operation has been scheduled already.
1049 if (has_aborted()) {
1050 return;
1051 }
1052
1053 G1Policy* g1p = _g1h->g1_policy();
1054 g1p->record_concurrent_mark_remark_start();
1055
1056 double start = os::elapsedTime();
1057
1058 verify_during_pause(G1HeapVerifier::G1VerifyRemark, VerifyOption_G1UsePrevMarking, "Remark before");
1059
1060 {
1061 GCTraceTime(Debug, gc, phases) trace("Finalize Marking", _gc_timer_cm);
1062 finalize_marking();
1063 }
1064
1065 double mark_work_end = os::elapsedTime();
1066
1067 bool const mark_finished = !has_overflown();
1068 if (mark_finished) {
1069 weak_refs_work(false /* clear_all_soft_refs */);
1070
1071 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1072 // We're done with marking.
1073 // This is the end of the marking cycle, we're expected all
1074 // threads to have SATB queues with active set to true.
1075 satb_mq_set.set_active_all_threads(false, /* new active value */
1076 true /* expected_active */);
1077
1078 {
1079 GCTraceTime(Debug, gc, phases)("Flush Task Caches");
1080 flush_all_task_caches();
1081 }
1082
1083 {
1084 GCTraceTime(Debug, gc, phases)("Update Remembered Set Tracking Before Rebuild");
1085 G1UpdateRemSetTrackingBeforeRebuild cl(_g1h, this);
1086 _g1h->heap_region_iterate(&cl);
1087 log_debug(gc, remset, tracking)("Remembered Set Tracking update regions total %u, selected %u",
1088 _g1h->num_regions(), cl.num_selected_for_rebuild());
1089 }
1090
1091 verify_during_pause(G1HeapVerifier::G1VerifyRemark, VerifyOption_G1UseNextMarking, "Remark after");
1092
1093 assert(!restart_for_overflow(), "sanity");
1094 // Completely reset the marking state since marking completed
1095 reset_at_marking_complete();
1096 } else {
1097 // We overflowed. Restart concurrent marking.
1098 _restart_for_overflow = true;
1099
1100 verify_during_pause(G1HeapVerifier::G1VerifyRemark, VerifyOption_G1UsePrevMarking, "Remark overflow");
1101
1102 // Clear the marking state because we will be restarting
1103 // marking due to overflowing the global mark stack.
1104 reset_marking_for_restart();
1105 }
1106
1107 {
1108 GCTraceTime(Debug, gc, phases)("Report Object Count");
1109 report_object_count();
1110 }
1111
1112 // Statistics
1113 double now = os::elapsedTime();
1114 _remark_mark_times.add((mark_work_end - start) * 1000.0);
1115 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0);
1116 _remark_times.add((now - start) * 1000.0);
1117
1118 g1p->record_concurrent_mark_remark_end();
1119 }
1120
1121 class G1CleanupTask : public AbstractGangTask {
1122 // Per-region work during the Cleanup pause.
1123 class G1CleanupRegionsClosure : public HeapRegionClosure {
1124 G1CollectedHeap* _g1h;
1125 size_t _freed_bytes;
1126 FreeRegionList* _local_cleanup_list;
1127 uint _old_regions_removed;
1128 uint _humongous_regions_removed;
1129 HRRSCleanupTask* _hrrs_cleanup_task;
1130
1131 public:
1132 G1CleanupRegionsClosure(G1CollectedHeap* g1,
1133 FreeRegionList* local_cleanup_list,
1134 HRRSCleanupTask* hrrs_cleanup_task) :
1135 _g1h(g1),
1136 _freed_bytes(0),
1137 _local_cleanup_list(local_cleanup_list),
1138 _old_regions_removed(0),
1139 _humongous_regions_removed(0),
1140 _hrrs_cleanup_task(hrrs_cleanup_task) { }
1141
1142 size_t freed_bytes() { return _freed_bytes; }
1143 const uint old_regions_removed() { return _old_regions_removed; }
1144 const uint humongous_regions_removed() { return _humongous_regions_removed; }
1145
1146 bool do_heap_region(HeapRegion *hr) {
1147 hr->note_end_of_marking();
1148
1149 if (hr->used() > 0 && hr->max_live_bytes() == 0 && !hr->is_young() && !hr->is_archive()) {
1150 _freed_bytes += hr->used();
1151 hr->set_containing_set(NULL);
1152 if (hr->is_humongous()) {
1153 _humongous_regions_removed++;
1154 _g1h->free_humongous_region(hr, _local_cleanup_list);
1155 } else {
1156 _old_regions_removed++;
1157 _g1h->free_region(hr, _local_cleanup_list, false /* skip_remset */, false /* skip_hcc */, true /* locked */);
1158 }
1159 hr->clear_cardtable();
1160 _g1h->concurrent_mark()->clear_statistics_in_region(hr->hrm_index());
1161 log_trace(gc)("Reclaimed empty region %u (%s) bot " PTR_FORMAT, hr->hrm_index(), hr->get_short_type_str(), p2i(hr->bottom()));
1162 } else {
1163 hr->rem_set()->do_cleanup_work(_hrrs_cleanup_task);
1164 }
1165
1166 return false;
1167 }
1168 };
1169
1170 G1CollectedHeap* _g1h;
1171 FreeRegionList* _cleanup_list;
1172 HeapRegionClaimer _hrclaimer;
1173
1174 public:
1175 G1CleanupTask(G1CollectedHeap* g1h, FreeRegionList* cleanup_list, uint n_workers) :
1176 AbstractGangTask("G1 Cleanup"),
1177 _g1h(g1h),
1178 _cleanup_list(cleanup_list),
1179 _hrclaimer(n_workers) {
1180
1181 HeapRegionRemSet::reset_for_cleanup_tasks();
1182 }
1183
1184 void work(uint worker_id) {
1185 FreeRegionList local_cleanup_list("Local Cleanup List");
1186 HRRSCleanupTask hrrs_cleanup_task;
1187 G1CleanupRegionsClosure cl(_g1h,
1188 &local_cleanup_list,
1189 &hrrs_cleanup_task);
1190 _g1h->heap_region_par_iterate_from_worker_offset(&cl, &_hrclaimer, worker_id);
1191 assert(cl.is_complete(), "Shouldn't have aborted!");
1192
1193 // Now update the old/humongous region sets
1194 _g1h->remove_from_old_sets(cl.old_regions_removed(), cl.humongous_regions_removed());
1195 {
1196 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
1197 _g1h->decrement_summary_bytes(cl.freed_bytes());
1198
1199 _cleanup_list->add_ordered(&local_cleanup_list);
1200 assert(local_cleanup_list.is_empty(), "post-condition");
1201
1202 HeapRegionRemSet::finish_cleanup_task(&hrrs_cleanup_task);
1203 }
1204 }
1205 };
1206
1207 void G1ConcurrentMark::reclaim_empty_regions() {
1208 WorkGang* workers = _g1h->workers();
1209 FreeRegionList empty_regions_list("Empty Regions After Mark List");
1210
1211 G1CleanupTask cl(_g1h, &empty_regions_list, workers->active_workers());
1212 workers->run_task(&cl);
1213
1214 if (!empty_regions_list.is_empty()) {
1215 log_debug(gc)("Reclaimed %u empty regions", empty_regions_list.length());
1216 // Now print the empty regions list.
1217 G1HRPrinter* hrp = _g1h->hr_printer();
1218 if (hrp->is_active()) {
1219 FreeRegionListIterator iter(&empty_regions_list);
1220 while (iter.more_available()) {
1221 HeapRegion* hr = iter.get_next();
1222 hrp->cleanup(hr);
1223 }
1224 }
1225 // And actually make them available.
1226 _g1h->prepend_to_freelist(&empty_regions_list);
1227 }
1228 }
1229
1230 void G1ConcurrentMark::cleanup() {
1231 assert_at_safepoint_on_vm_thread();
1232
1233 // If a full collection has happened, we shouldn't do this.
1234 if (has_aborted()) {
1235 return;
1236 }
1237
1238 G1Policy* g1p = _g1h->g1_policy();
1239 g1p->record_concurrent_mark_cleanup_start();
1240
1241 double start = os::elapsedTime();
1242
1243 verify_during_pause(G1HeapVerifier::G1VerifyCleanup, VerifyOption_G1UseNextMarking, "Cleanup before");
1244
1245 {
1246 GCTraceTime(Debug, gc, phases)("Update Remembered Set Tracking After Rebuild");
1247 G1UpdateRemSetTrackingAfterRebuild cl(_g1h);
1248 _g1h->heap_region_iterate(&cl);
1249 }
1250
1251 if (log_is_enabled(Trace, gc, liveness)) {
1252 G1PrintRegionLivenessInfoClosure cl("Post-Cleanup");
1253 _g1h->heap_region_iterate(&cl);
1254 }
1255
1256 // Install newly created mark bitmap as "prev".
1257 swap_mark_bitmaps();
1258 {
1259 GCTraceTime(Debug, gc, phases)("Reclaim Empty Regions");
1260 reclaim_empty_regions();
1261 }
1262
1263 // Cleanup will have freed any regions completely full of garbage.
1264 // Update the soft reference policy with the new heap occupancy.
1265 Universe::update_heap_info_at_gc();
1266
1267 // Clean out dead classes and update Metaspace sizes.
1268 if (ClassUnloadingWithConcurrentMark) {
1269 GCTraceTime(Debug, gc, phases)("Purge Metaspace");
1270 ClassLoaderDataGraph::purge();
1271 }
1272 MetaspaceGC::compute_new_size();
1273
1274 // We reclaimed old regions so we should calculate the sizes to make
1275 // sure we update the old gen/space data.
1276 _g1h->g1mm()->update_sizes();
1277
1278 verify_during_pause(G1HeapVerifier::G1VerifyCleanup, VerifyOption_G1UsePrevMarking, "Cleanup after");
1279
1280 // We need to make this be a "collection" so any collection pause that
1281 // races with it goes around and waits for Cleanup to finish.
1282 _g1h->increment_total_collections();
1283
1284 // Local statistics
1285 double recent_cleanup_time = (os::elapsedTime() - start);
1286 _total_cleanup_time += recent_cleanup_time;
1287 _cleanup_times.add(recent_cleanup_time);
1288
1289 {
1290 GCTraceTime(Debug, gc, phases)("Finalize Concurrent Mark Cleanup");
1291 _g1h->g1_policy()->record_concurrent_mark_cleanup_end();
1292 }
1293 }
1294
1295 // Supporting Object and Oop closures for reference discovery
1296 // and processing in during marking
1297
1298 bool G1CMIsAliveClosure::do_object_b(oop obj) {
1299 HeapWord* addr = (HeapWord*)obj;
1300 return addr != NULL &&
1301 (!_g1h->is_in_g1_reserved(addr) || !_g1h->is_obj_ill(obj));
1302 }
1303
1304 // 'Keep Alive' oop closure used by both serial parallel reference processing.
1305 // Uses the G1CMTask associated with a worker thread (for serial reference
1306 // processing the G1CMTask for worker 0 is used) to preserve (mark) and
1307 // trace referent objects.
1308 //
1309 // Using the G1CMTask and embedded local queues avoids having the worker
1310 // threads operating on the global mark stack. This reduces the risk
1311 // of overflowing the stack - which we would rather avoid at this late
1312 // state. Also using the tasks' local queues removes the potential
1313 // of the workers interfering with each other that could occur if
1314 // operating on the global stack.
1315
1316 class G1CMKeepAliveAndDrainClosure : public OopClosure {
1317 G1ConcurrentMark* _cm;
1318 G1CMTask* _task;
1319 int _ref_counter_limit;
1320 int _ref_counter;
1321 bool _is_serial;
1322 public:
1323 G1CMKeepAliveAndDrainClosure(G1ConcurrentMark* cm, G1CMTask* task, bool is_serial) :
1324 _cm(cm), _task(task), _is_serial(is_serial),
1325 _ref_counter_limit(G1RefProcDrainInterval) {
1326 assert(_ref_counter_limit > 0, "sanity");
1327 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
1328 _ref_counter = _ref_counter_limit;
1329 }
1330
1331 virtual void do_oop(narrowOop* p) { do_oop_work(p); }
1332 virtual void do_oop( oop* p) { do_oop_work(p); }
1333
1334 template <class T> void do_oop_work(T* p) {
1335 if (!_cm->has_overflown()) {
1336 _task->deal_with_reference(p);
1337 _ref_counter--;
1338
1339 if (_ref_counter == 0) {
1340 // We have dealt with _ref_counter_limit references, pushing them
1341 // and objects reachable from them on to the local stack (and
1342 // possibly the global stack). Call G1CMTask::do_marking_step() to
1343 // process these entries.
1344 //
1345 // We call G1CMTask::do_marking_step() in a loop, which we'll exit if
1346 // there's nothing more to do (i.e. we're done with the entries that
1347 // were pushed as a result of the G1CMTask::deal_with_reference() calls
1348 // above) or we overflow.
1349 //
1350 // Note: G1CMTask::do_marking_step() can set the G1CMTask::has_aborted()
1351 // flag while there may still be some work to do. (See the comment at
1352 // the beginning of G1CMTask::do_marking_step() for those conditions -
1353 // one of which is reaching the specified time target.) It is only
1354 // when G1CMTask::do_marking_step() returns without setting the
1355 // has_aborted() flag that the marking step has completed.
1356 do {
1357 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
1358 _task->do_marking_step(mark_step_duration_ms,
1359 false /* do_termination */,
1360 _is_serial);
1361 } while (_task->has_aborted() && !_cm->has_overflown());
1362 _ref_counter = _ref_counter_limit;
1363 }
1364 }
1365 }
1366 };
1367
1368 // 'Drain' oop closure used by both serial and parallel reference processing.
1369 // Uses the G1CMTask associated with a given worker thread (for serial
1370 // reference processing the G1CMtask for worker 0 is used). Calls the
1371 // do_marking_step routine, with an unbelievably large timeout value,
1372 // to drain the marking data structures of the remaining entries
1373 // added by the 'keep alive' oop closure above.
1374
1375 class G1CMDrainMarkingStackClosure : public VoidClosure {
1376 G1ConcurrentMark* _cm;
1377 G1CMTask* _task;
1378 bool _is_serial;
1379 public:
1380 G1CMDrainMarkingStackClosure(G1ConcurrentMark* cm, G1CMTask* task, bool is_serial) :
1381 _cm(cm), _task(task), _is_serial(is_serial) {
1382 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
1383 }
1384
1385 void do_void() {
1386 do {
1387 // We call G1CMTask::do_marking_step() to completely drain the local
1388 // and global marking stacks of entries pushed by the 'keep alive'
1389 // oop closure (an instance of G1CMKeepAliveAndDrainClosure above).
1390 //
1391 // G1CMTask::do_marking_step() is called in a loop, which we'll exit
1392 // if there's nothing more to do (i.e. we've completely drained the
1393 // entries that were pushed as a a result of applying the 'keep alive'
1394 // closure to the entries on the discovered ref lists) or we overflow
1395 // the global marking stack.
1396 //
1397 // Note: G1CMTask::do_marking_step() can set the G1CMTask::has_aborted()
1398 // flag while there may still be some work to do. (See the comment at
1399 // the beginning of G1CMTask::do_marking_step() for those conditions -
1400 // one of which is reaching the specified time target.) It is only
1401 // when G1CMTask::do_marking_step() returns without setting the
1402 // has_aborted() flag that the marking step has completed.
1403
1404 _task->do_marking_step(1000000000.0 /* something very large */,
1405 true /* do_termination */,
1406 _is_serial);
1407 } while (_task->has_aborted() && !_cm->has_overflown());
1408 }
1409 };
1410
1411 // Implementation of AbstractRefProcTaskExecutor for parallel
1412 // reference processing at the end of G1 concurrent marking
1413
1414 class G1CMRefProcTaskExecutor : public AbstractRefProcTaskExecutor {
1415 private:
1416 G1CollectedHeap* _g1h;
1417 G1ConcurrentMark* _cm;
1418 WorkGang* _workers;
1419 uint _active_workers;
1420
1421 public:
1422 G1CMRefProcTaskExecutor(G1CollectedHeap* g1h,
1423 G1ConcurrentMark* cm,
1424 WorkGang* workers,
1425 uint n_workers) :
1426 _g1h(g1h), _cm(cm),
1427 _workers(workers), _active_workers(n_workers) { }
1428
1429 // Executes the given task using concurrent marking worker threads.
1430 virtual void execute(ProcessTask& task);
1431 virtual void execute(EnqueueTask& task);
1432 };
1433
1434 class G1CMRefProcTaskProxy : public AbstractGangTask {
1435 typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask;
1436 ProcessTask& _proc_task;
1437 G1CollectedHeap* _g1h;
1438 G1ConcurrentMark* _cm;
1439
1440 public:
1441 G1CMRefProcTaskProxy(ProcessTask& proc_task,
1442 G1CollectedHeap* g1h,
1443 G1ConcurrentMark* cm) :
1444 AbstractGangTask("Process reference objects in parallel"),
1445 _proc_task(proc_task), _g1h(g1h), _cm(cm) {
1446 ReferenceProcessor* rp = _g1h->ref_processor_cm();
1447 assert(rp->processing_is_mt(), "shouldn't be here otherwise");
1448 }
1449
1450 virtual void work(uint worker_id) {
1451 ResourceMark rm;
1452 HandleMark hm;
1453 G1CMTask* task = _cm->task(worker_id);
1454 G1CMIsAliveClosure g1_is_alive(_g1h);
1455 G1CMKeepAliveAndDrainClosure g1_par_keep_alive(_cm, task, false /* is_serial */);
1456 G1CMDrainMarkingStackClosure g1_par_drain(_cm, task, false /* is_serial */);
1457
1458 _proc_task.work(worker_id, g1_is_alive, g1_par_keep_alive, g1_par_drain);
1459 }
1460 };
1461
1462 void G1CMRefProcTaskExecutor::execute(ProcessTask& proc_task) {
1463 assert(_workers != NULL, "Need parallel worker threads.");
1464 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
1465
1466 G1CMRefProcTaskProxy proc_task_proxy(proc_task, _g1h, _cm);
1467
1468 // We need to reset the concurrency level before each
1469 // proxy task execution, so that the termination protocol
1470 // and overflow handling in G1CMTask::do_marking_step() knows
1471 // how many workers to wait for.
1472 _cm->set_concurrency(_active_workers);
1473 _workers->run_task(&proc_task_proxy);
1474 }
1475
1476 class G1CMRefEnqueueTaskProxy : public AbstractGangTask {
1477 typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask;
1478 EnqueueTask& _enq_task;
1479
1480 public:
1481 G1CMRefEnqueueTaskProxy(EnqueueTask& enq_task) :
1482 AbstractGangTask("Enqueue reference objects in parallel"),
1483 _enq_task(enq_task) { }
1484
1485 virtual void work(uint worker_id) {
1486 _enq_task.work(worker_id);
1487 }
1488 };
1489
1490 void G1CMRefProcTaskExecutor::execute(EnqueueTask& enq_task) {
1491 assert(_workers != NULL, "Need parallel worker threads.");
1492 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
1493
1494 G1CMRefEnqueueTaskProxy enq_task_proxy(enq_task);
1495
1496 // Not strictly necessary but...
1497 //
1498 // We need to reset the concurrency level before each
1499 // proxy task execution, so that the termination protocol
1500 // and overflow handling in G1CMTask::do_marking_step() knows
1501 // how many workers to wait for.
1502 _cm->set_concurrency(_active_workers);
1503 _workers->run_task(&enq_task_proxy);
1504 }
1505
1506 void G1ConcurrentMark::weak_refs_work(bool clear_all_soft_refs) {
1507 ResourceMark rm;
1508 HandleMark hm;
1509
1510 // Is alive closure.
1511 G1CMIsAliveClosure g1_is_alive(_g1h);
1512
1513 // Inner scope to exclude the cleaning of the string and symbol
1514 // tables from the displayed time.
1515 {
1516 GCTraceTime(Debug, gc, phases) trace("Reference Processing", _gc_timer_cm);
1517
1518 ReferenceProcessor* rp = _g1h->ref_processor_cm();
1519
1520 // See the comment in G1CollectedHeap::ref_processing_init()
1521 // about how reference processing currently works in G1.
1522
1523 // Set the soft reference policy
1524 rp->setup_policy(clear_all_soft_refs);
1525 assert(_global_mark_stack.is_empty(), "mark stack should be empty");
1526
1527 // Instances of the 'Keep Alive' and 'Complete GC' closures used
1528 // in serial reference processing. Note these closures are also
1529 // used for serially processing (by the the current thread) the
1530 // JNI references during parallel reference processing.
1531 //
1532 // These closures do not need to synchronize with the worker
1533 // threads involved in parallel reference processing as these
1534 // instances are executed serially by the current thread (e.g.
1535 // reference processing is not multi-threaded and is thus
1536 // performed by the current thread instead of a gang worker).
1537 //
1538 // The gang tasks involved in parallel reference processing create
1539 // their own instances of these closures, which do their own
1540 // synchronization among themselves.
1541 G1CMKeepAliveAndDrainClosure g1_keep_alive(this, task(0), true /* is_serial */);
1542 G1CMDrainMarkingStackClosure g1_drain_mark_stack(this, task(0), true /* is_serial */);
1543
1544 // We need at least one active thread. If reference processing
1545 // is not multi-threaded we use the current (VMThread) thread,
1546 // otherwise we use the work gang from the G1CollectedHeap and
1547 // we utilize all the worker threads we can.
1548 bool processing_is_mt = rp->processing_is_mt();
1549 uint active_workers = (processing_is_mt ? _g1h->workers()->active_workers() : 1U);
1550 active_workers = MAX2(MIN2(active_workers, _max_num_tasks), 1U);
1551
1552 // Parallel processing task executor.
1553 G1CMRefProcTaskExecutor par_task_executor(_g1h, this,
1554 _g1h->workers(), active_workers);
1555 AbstractRefProcTaskExecutor* executor = (processing_is_mt ? &par_task_executor : NULL);
1556
1557 // Set the concurrency level. The phase was already set prior to
1558 // executing the remark task.
1559 set_concurrency(active_workers);
1560
1561 // Set the degree of MT processing here. If the discovery was done MT,
1562 // the number of threads involved during discovery could differ from
1563 // the number of active workers. This is OK as long as the discovered
1564 // Reference lists are balanced (see balance_all_queues() and balance_queues()).
1565 rp->set_active_mt_degree(active_workers);
1566
1567 ReferenceProcessorPhaseTimes pt(_gc_timer_cm, rp->num_q());
1568
1569 // Process the weak references.
1570 const ReferenceProcessorStats& stats =
1571 rp->process_discovered_references(&g1_is_alive,
1572 &g1_keep_alive,
1573 &g1_drain_mark_stack,
1574 executor,
1575 &pt);
1576 _gc_tracer_cm->report_gc_reference_stats(stats);
1577 pt.print_all_references();
1578
1579 // The do_oop work routines of the keep_alive and drain_marking_stack
1580 // oop closures will set the has_overflown flag if we overflow the
1581 // global marking stack.
1582
1583 assert(has_overflown() || _global_mark_stack.is_empty(),
1584 "Mark stack should be empty (unless it has overflown)");
1585
1586 assert(rp->num_q() == active_workers, "why not");
1587
1588 rp->enqueue_discovered_references(executor, &pt);
1589
1590 rp->verify_no_references_recorded();
1591
1592 pt.print_enqueue_phase();
1593
1594 assert(!rp->discovery_enabled(), "Post condition");
1595 }
1596
1597 assert(has_overflown() || _global_mark_stack.is_empty(),
1598 "Mark stack should be empty (unless it has overflown)");
1599
1600 {
1601 GCTraceTime(Debug, gc, phases) debug("Weak Processing", _gc_timer_cm);
1602 WeakProcessor::weak_oops_do(&g1_is_alive, &do_nothing_cl);
1603 }
1604
1605 if (has_overflown()) {
1606 // We can not trust g1_is_alive if the marking stack overflowed
1607 return;
1608 }
1609
1610 assert(_global_mark_stack.is_empty(), "Marking should have completed");
1611
1612 // Unload Klasses, String, Symbols, Code Cache, etc.
1613 if (ClassUnloadingWithConcurrentMark) {
1614 GCTraceTime(Debug, gc, phases) debug("Class Unloading", _gc_timer_cm);
1615 bool purged_classes = SystemDictionary::do_unloading(&g1_is_alive, _gc_timer_cm, false /* Defer cleaning */);
1616 _g1h->complete_cleaning(&g1_is_alive, purged_classes);
1617 } else {
1618 GCTraceTime(Debug, gc, phases) debug("Cleanup", _gc_timer_cm);
1619 // No need to clean string table and symbol table as they are treated as strong roots when
1620 // class unloading is disabled.
1621 _g1h->partial_cleaning(&g1_is_alive, false, false, G1StringDedup::is_enabled());
1622 }
1623 }
1624
1625 void G1ConcurrentMark::report_object_count() {
1626 G1CMIsAliveClosure is_alive(_g1h);
1627 _gc_tracer_cm->report_object_count_after_gc(&is_alive);
1628 }
1629
1630 void G1ConcurrentMark::swap_mark_bitmaps() {
1631 G1CMBitMap* temp = _prev_mark_bitmap;
1632 _prev_mark_bitmap = _next_mark_bitmap;
1633 _next_mark_bitmap = temp;
1634 _g1h->collector_state()->set_clearing_next_bitmap(true);
1635 }
1636
1637 // Closure for marking entries in SATB buffers.
1638 class G1CMSATBBufferClosure : public SATBBufferClosure {
1639 private:
1640 G1CMTask* _task;
1641 G1CollectedHeap* _g1h;
1642
1643 // This is very similar to G1CMTask::deal_with_reference, but with
1644 // more relaxed requirements for the argument, so this must be more
1645 // circumspect about treating the argument as an object.
1646 void do_entry(void* entry) const {
1647 _task->increment_refs_reached();
1648 oop const obj = static_cast<oop>(entry);
1649 _task->make_reference_grey(obj);
1650 }
1651
1652 public:
1653 G1CMSATBBufferClosure(G1CMTask* task, G1CollectedHeap* g1h)
1654 : _task(task), _g1h(g1h) { }
1655
1656 virtual void do_buffer(void** buffer, size_t size) {
1657 for (size_t i = 0; i < size; ++i) {
1658 do_entry(buffer[i]);
1659 }
1660 }
1661 };
1662
1663 class G1RemarkThreadsClosure : public ThreadClosure {
1664 G1CMSATBBufferClosure _cm_satb_cl;
1665 G1CMOopClosure _cm_cl;
1666 MarkingCodeBlobClosure _code_cl;
1667 int _thread_parity;
1668
1669 public:
1670 G1RemarkThreadsClosure(G1CollectedHeap* g1h, G1CMTask* task) :
1671 _cm_satb_cl(task, g1h),
1672 _cm_cl(g1h, task),
1673 _code_cl(&_cm_cl, !CodeBlobToOopClosure::FixRelocations),
1674 _thread_parity(Threads::thread_claim_parity()) {}
1675
1676 void do_thread(Thread* thread) {
1677 if (thread->is_Java_thread()) {
1678 if (thread->claim_oops_do(true, _thread_parity)) {
1679 JavaThread* jt = (JavaThread*)thread;
1680
1681 // In theory it should not be neccessary to explicitly walk the nmethods to find roots for concurrent marking
1682 // however the liveness of oops reachable from nmethods have very complex lifecycles:
1683 // * Alive if on the stack of an executing method
1684 // * Weakly reachable otherwise
1685 // Some objects reachable from nmethods, such as the class loader (or klass_holder) of the receiver should be
1686 // live by the SATB invariant but other oops recorded in nmethods may behave differently.
1687 jt->nmethods_do(&_code_cl);
1688
1689 jt->satb_mark_queue().apply_closure_and_empty(&_cm_satb_cl);
1690 }
1691 } else if (thread->is_VM_thread()) {
1692 if (thread->claim_oops_do(true, _thread_parity)) {
1693 JavaThread::satb_mark_queue_set().shared_satb_queue()->apply_closure_and_empty(&_cm_satb_cl);
1694 }
1695 }
1696 }
1697 };
1698
1699 class G1CMRemarkTask : public AbstractGangTask {
1700 G1ConcurrentMark* _cm;
1701 public:
1702 void work(uint worker_id) {
1703 G1CMTask* task = _cm->task(worker_id);
1704 task->record_start_time();
1705 {
1706 ResourceMark rm;
1707 HandleMark hm;
1708
1709 G1RemarkThreadsClosure threads_f(G1CollectedHeap::heap(), task);
1710 Threads::threads_do(&threads_f);
1711 }
1712
1713 do {
1714 task->do_marking_step(1000000000.0 /* something very large */,
1715 true /* do_termination */,
1716 false /* is_serial */);
1717 } while (task->has_aborted() && !_cm->has_overflown());
1718 // If we overflow, then we do not want to restart. We instead
1719 // want to abort remark and do concurrent marking again.
1720 task->record_end_time();
1721 }
1722
1723 G1CMRemarkTask(G1ConcurrentMark* cm, uint active_workers) :
1724 AbstractGangTask("Par Remark"), _cm(cm) {
1725 _cm->terminator()->reset_for_reuse(active_workers);
1726 }
1727 };
1728
1729 void G1ConcurrentMark::finalize_marking() {
1730 ResourceMark rm;
1731 HandleMark hm;
1732
1733 _g1h->ensure_parsability(false);
1734
1735 // this is remark, so we'll use up all active threads
1736 uint active_workers = _g1h->workers()->active_workers();
1737 set_concurrency_and_phase(active_workers, false /* concurrent */);
1738 // Leave _parallel_marking_threads at it's
1739 // value originally calculated in the G1ConcurrentMark
1740 // constructor and pass values of the active workers
1741 // through the gang in the task.
1742
1743 {
1744 StrongRootsScope srs(active_workers);
1745
1746 G1CMRemarkTask remarkTask(this, active_workers);
1747 // We will start all available threads, even if we decide that the
1748 // active_workers will be fewer. The extra ones will just bail out
1749 // immediately.
1750 _g1h->workers()->run_task(&remarkTask);
1751 }
1752
1753 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1754 guarantee(has_overflown() ||
1755 satb_mq_set.completed_buffers_num() == 0,
1756 "Invariant: has_overflown = %s, num buffers = " SIZE_FORMAT,
1757 BOOL_TO_STR(has_overflown()),
1758 satb_mq_set.completed_buffers_num());
1759
1760 print_stats();
1761 }
1762
1763 void G1ConcurrentMark::flush_all_task_caches() {
1764 size_t hits = 0;
1765 size_t misses = 0;
1766 for (uint i = 0; i < _max_num_tasks; i++) {
1767 Pair<size_t, size_t> stats = _tasks[i]->flush_mark_stats_cache();
1768 hits += stats.first;
1769 misses += stats.second;
1770 }
1771 size_t sum = hits + misses;
1772 log_debug(gc, stats)("Mark stats cache hits " SIZE_FORMAT " misses " SIZE_FORMAT " ratio %1.3lf",
1773 hits, misses, percent_of(hits, sum));
1774 }
1775
1776 void G1ConcurrentMark::clear_range_in_prev_bitmap(MemRegion mr) {
1777 _prev_mark_bitmap->clear_range(mr);
1778 }
1779
1780 HeapRegion*
1781 G1ConcurrentMark::claim_region(uint worker_id) {
1782 // "checkpoint" the finger
1783 HeapWord* finger = _finger;
1784
1785 while (finger < _heap.end()) {
1786 assert(_g1h->is_in_g1_reserved(finger), "invariant");
1787
1788 HeapRegion* curr_region = _g1h->heap_region_containing(finger);
1789 // Make sure that the reads below do not float before loading curr_region.
1790 OrderAccess::loadload();
1791 // Above heap_region_containing may return NULL as we always scan claim
1792 // until the end of the heap. In this case, just jump to the next region.
1793 HeapWord* end = curr_region != NULL ? curr_region->end() : finger + HeapRegion::GrainWords;
1794
1795 // Is the gap between reading the finger and doing the CAS too long?
1796 HeapWord* res = Atomic::cmpxchg(end, &_finger, finger);
1797 if (res == finger && curr_region != NULL) {
1798 // we succeeded
1799 HeapWord* bottom = curr_region->bottom();
1800 HeapWord* limit = curr_region->next_top_at_mark_start();
1801
1802 // notice that _finger == end cannot be guaranteed here since,
1803 // someone else might have moved the finger even further
1804 assert(_finger >= end, "the finger should have moved forward");
1805
1806 if (limit > bottom) {
1807 return curr_region;
1808 } else {
1809 assert(limit == bottom,
1810 "the region limit should be at bottom");
1811 // we return NULL and the caller should try calling
1812 // claim_region() again.
1813 return NULL;
1814 }
1815 } else {
1816 assert(_finger > finger, "the finger should have moved forward");
1817 // read it again
1818 finger = _finger;
1819 }
1820 }
1821
1822 return NULL;
1823 }
1824
1825 #ifndef PRODUCT
1826 class VerifyNoCSetOops {
1827 G1CollectedHeap* _g1h;
1828 const char* _phase;
1829 int _info;
1830
1831 public:
1832 VerifyNoCSetOops(const char* phase, int info = -1) :
1833 _g1h(G1CollectedHeap::heap()),
1834 _phase(phase),
1835 _info(info)
1836 { }
1837
1838 void operator()(G1TaskQueueEntry task_entry) const {
1839 if (task_entry.is_array_slice()) {
1840 guarantee(_g1h->is_in_reserved(task_entry.slice()), "Slice " PTR_FORMAT " must be in heap.", p2i(task_entry.slice()));
1841 return;
1842 }
1843 guarantee(oopDesc::is_oop(task_entry.obj()),
1844 "Non-oop " PTR_FORMAT ", phase: %s, info: %d",
1845 p2i(task_entry.obj()), _phase, _info);
1846 guarantee(!_g1h->is_in_cset(task_entry.obj()),
1847 "obj: " PTR_FORMAT " in CSet, phase: %s, info: %d",
1848 p2i(task_entry.obj()), _phase, _info);
1849 }
1850 };
1851
1852 void G1ConcurrentMark::verify_no_cset_oops() {
1853 assert(SafepointSynchronize::is_at_safepoint(), "should be at a safepoint");
1854 if (!_g1h->collector_state()->mark_or_rebuild_in_progress()) {
1855 return;
1856 }
1857
1858 // Verify entries on the global mark stack
1859 _global_mark_stack.iterate(VerifyNoCSetOops("Stack"));
1860
1861 // Verify entries on the task queues
1862 for (uint i = 0; i < _max_num_tasks; ++i) {
1863 G1CMTaskQueue* queue = _task_queues->queue(i);
1864 queue->iterate(VerifyNoCSetOops("Queue", i));
1865 }
1866
1867 // Verify the global finger
1868 HeapWord* global_finger = finger();
1869 if (global_finger != NULL && global_finger < _heap.end()) {
1870 // Since we always iterate over all regions, we might get a NULL HeapRegion
1871 // here.
1872 HeapRegion* global_hr = _g1h->heap_region_containing(global_finger);
1873 guarantee(global_hr == NULL || global_finger == global_hr->bottom(),
1874 "global finger: " PTR_FORMAT " region: " HR_FORMAT,
1875 p2i(global_finger), HR_FORMAT_PARAMS(global_hr));
1876 }
1877
1878 // Verify the task fingers
1879 assert(_num_concurrent_workers <= _max_num_tasks, "sanity");
1880 for (uint i = 0; i < _num_concurrent_workers; ++i) {
1881 G1CMTask* task = _tasks[i];
1882 HeapWord* task_finger = task->finger();
1883 if (task_finger != NULL && task_finger < _heap.end()) {
1884 // See above note on the global finger verification.
1885 HeapRegion* task_hr = _g1h->heap_region_containing(task_finger);
1886 guarantee(task_hr == NULL || task_finger == task_hr->bottom() ||
1887 !task_hr->in_collection_set(),
1888 "task finger: " PTR_FORMAT " region: " HR_FORMAT,
1889 p2i(task_finger), HR_FORMAT_PARAMS(task_hr));
1890 }
1891 }
1892 }
1893 #endif // PRODUCT
1894
1895 void G1ConcurrentMark::rebuild_rem_set_concurrently() {
1896 _g1h->g1_rem_set()->rebuild_rem_set(this, _concurrent_workers, _worker_id_offset);
1897 }
1898
1899 void G1ConcurrentMark::print_stats() {
1900 if (!log_is_enabled(Debug, gc, stats)) {
1901 return;
1902 }
1903 log_debug(gc, stats)("---------------------------------------------------------------------");
1904 for (size_t i = 0; i < _num_active_tasks; ++i) {
1905 _tasks[i]->print_stats();
1906 log_debug(gc, stats)("---------------------------------------------------------------------");
1907 }
1908 }
1909
1910 void G1ConcurrentMark::concurrent_cycle_abort() {
1911 if (!cm_thread()->during_cycle() || _has_aborted) {
1912 // We haven't started a concurrent cycle or we have already aborted it. No need to do anything.
1913 return;
1914 }
1915
1916 // Clear all marks in the next bitmap for the next marking cycle. This will allow us to skip the next
1917 // concurrent bitmap clearing.
1918 {
1919 GCTraceTime(Debug, gc)("Clear Next Bitmap");
1920 clear_bitmap(_next_mark_bitmap, _g1h->workers(), false);
1921 }
1922 // Note we cannot clear the previous marking bitmap here
1923 // since VerifyDuringGC verifies the objects marked during
1924 // a full GC against the previous bitmap.
1925
1926 // Empty mark stack
1927 reset_marking_for_restart();
1928 for (uint i = 0; i < _max_num_tasks; ++i) {
1929 _tasks[i]->clear_region_fields();
1930 }
1931 _first_overflow_barrier_sync.abort();
1932 _second_overflow_barrier_sync.abort();
1933 _has_aborted = true;
1934
1935 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1936 satb_mq_set.abandon_partial_marking();
1937 // This can be called either during or outside marking, we'll read
1938 // the expected_active value from the SATB queue set.
1939 satb_mq_set.set_active_all_threads(
1940 false, /* new active value */
1941 satb_mq_set.is_active() /* expected_active */);
1942 }
1943
1944 static void print_ms_time_info(const char* prefix, const char* name,
1945 NumberSeq& ns) {
1946 log_trace(gc, marking)("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).",
1947 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg());
1948 if (ns.num() > 0) {
1949 log_trace(gc, marking)("%s [std. dev = %8.2f ms, max = %8.2f ms]",
1950 prefix, ns.sd(), ns.maximum());
1951 }
1952 }
1953
1954 void G1ConcurrentMark::print_summary_info() {
1955 Log(gc, marking) log;
1956 if (!log.is_trace()) {
1957 return;
1958 }
1959
1960 log.trace(" Concurrent marking:");
1961 print_ms_time_info(" ", "init marks", _init_times);
1962 print_ms_time_info(" ", "remarks", _remark_times);
1963 {
1964 print_ms_time_info(" ", "final marks", _remark_mark_times);
1965 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times);
1966
1967 }
1968 print_ms_time_info(" ", "cleanups", _cleanup_times);
1969 log.trace(" Finalize live data total time = %8.2f s (avg = %8.2f ms).",
1970 _total_cleanup_time, (_cleanup_times.num() > 0 ? _total_cleanup_time * 1000.0 / (double)_cleanup_times.num() : 0.0));
1971 log.trace(" Total stop_world time = %8.2f s.",
1972 (_init_times.sum() + _remark_times.sum() + _cleanup_times.sum())/1000.0);
1973 log.trace(" Total concurrent time = %8.2f s (%8.2f s marking).",
1974 cm_thread()->vtime_accum(), cm_thread()->vtime_mark_accum());
1975 }
1976
1977 void G1ConcurrentMark::print_worker_threads_on(outputStream* st) const {
1978 _concurrent_workers->print_worker_threads_on(st);
1979 }
1980
1981 void G1ConcurrentMark::threads_do(ThreadClosure* tc) const {
1982 _concurrent_workers->threads_do(tc);
1983 }
1984
1985 void G1ConcurrentMark::print_on_error(outputStream* st) const {
1986 st->print_cr("Marking Bits (Prev, Next): (CMBitMap*) " PTR_FORMAT ", (CMBitMap*) " PTR_FORMAT,
1987 p2i(_prev_mark_bitmap), p2i(_next_mark_bitmap));
1988 _prev_mark_bitmap->print_on_error(st, " Prev Bits: ");
1989 _next_mark_bitmap->print_on_error(st, " Next Bits: ");
1990 }
1991
1992 static ReferenceProcessor* get_cm_oop_closure_ref_processor(G1CollectedHeap* g1h) {
1993 ReferenceProcessor* result = g1h->ref_processor_cm();
1994 assert(result != NULL, "CM reference processor should not be NULL");
1995 return result;
1996 }
1997
1998 G1CMOopClosure::G1CMOopClosure(G1CollectedHeap* g1h,
1999 G1CMTask* task)
2000 : MetadataAwareOopClosure(get_cm_oop_closure_ref_processor(g1h)),
2001 _g1h(g1h), _task(task)
2002 { }
2003
2004 void G1CMTask::setup_for_region(HeapRegion* hr) {
2005 assert(hr != NULL,
2006 "claim_region() should have filtered out NULL regions");
2007 _curr_region = hr;
2008 _finger = hr->bottom();
2009 update_region_limit();
2010 }
2011
2012 void G1CMTask::update_region_limit() {
2013 HeapRegion* hr = _curr_region;
2014 HeapWord* bottom = hr->bottom();
2015 HeapWord* limit = hr->next_top_at_mark_start();
2016
2017 if (limit == bottom) {
2018 // The region was collected underneath our feet.
2019 // We set the finger to bottom to ensure that the bitmap
2020 // iteration that will follow this will not do anything.
2021 // (this is not a condition that holds when we set the region up,
2022 // as the region is not supposed to be empty in the first place)
2023 _finger = bottom;
2024 } else if (limit >= _region_limit) {
2025 assert(limit >= _finger, "peace of mind");
2026 } else {
2027 assert(limit < _region_limit, "only way to get here");
2028 // This can happen under some pretty unusual circumstances. An
2029 // evacuation pause empties the region underneath our feet (NTAMS
2030 // at bottom). We then do some allocation in the region (NTAMS
2031 // stays at bottom), followed by the region being used as a GC
2032 // alloc region (NTAMS will move to top() and the objects
2033 // originally below it will be grayed). All objects now marked in
2034 // the region are explicitly grayed, if below the global finger,
2035 // and we do not need in fact to scan anything else. So, we simply
2036 // set _finger to be limit to ensure that the bitmap iteration
2037 // doesn't do anything.
2038 _finger = limit;
2039 }
2040
2041 _region_limit = limit;
2042 }
2043
2044 void G1CMTask::giveup_current_region() {
2045 assert(_curr_region != NULL, "invariant");
2046 clear_region_fields();
2047 }
2048
2049 void G1CMTask::clear_region_fields() {
2050 // Values for these three fields that indicate that we're not
2051 // holding on to a region.
2052 _curr_region = NULL;
2053 _finger = NULL;
2054 _region_limit = NULL;
2055 }
2056
2057 void G1CMTask::set_cm_oop_closure(G1CMOopClosure* cm_oop_closure) {
2058 if (cm_oop_closure == NULL) {
2059 assert(_cm_oop_closure != NULL, "invariant");
2060 } else {
2061 assert(_cm_oop_closure == NULL, "invariant");
2062 }
2063 _cm_oop_closure = cm_oop_closure;
2064 }
2065
2066 void G1CMTask::reset(G1CMBitMap* next_mark_bitmap) {
2067 guarantee(next_mark_bitmap != NULL, "invariant");
2068 _next_mark_bitmap = next_mark_bitmap;
2069 clear_region_fields();
2070
2071 _calls = 0;
2072 _elapsed_time_ms = 0.0;
2073 _termination_time_ms = 0.0;
2074 _termination_start_time_ms = 0.0;
2075
2076 _mark_stats_cache.reset();
2077 }
2078
2079 bool G1CMTask::should_exit_termination() {
2080 regular_clock_call();
2081 // This is called when we are in the termination protocol. We should
2082 // quit if, for some reason, this task wants to abort or the global
2083 // stack is not empty (this means that we can get work from it).
2084 return !_cm->mark_stack_empty() || has_aborted();
2085 }
2086
2087 void G1CMTask::reached_limit() {
2088 assert(_words_scanned >= _words_scanned_limit ||
2089 _refs_reached >= _refs_reached_limit ,
2090 "shouldn't have been called otherwise");
2091 regular_clock_call();
2092 }
2093
2094 void G1CMTask::regular_clock_call() {
2095 if (has_aborted()) {
2096 return;
2097 }
2098
2099 // First, we need to recalculate the words scanned and refs reached
2100 // limits for the next clock call.
2101 recalculate_limits();
2102
2103 // During the regular clock call we do the following
2104
2105 // (1) If an overflow has been flagged, then we abort.
2106 if (_cm->has_overflown()) {
2107 set_has_aborted();
2108 return;
2109 }
2110
2111 // If we are not concurrent (i.e. we're doing remark) we don't need
2112 // to check anything else. The other steps are only needed during
2113 // the concurrent marking phase.
2114 if (!_cm->concurrent()) {
2115 return;
2116 }
2117
2118 // (2) If marking has been aborted for Full GC, then we also abort.
2119 if (_cm->has_aborted()) {
2120 set_has_aborted();
2121 return;
2122 }
2123
2124 double curr_time_ms = os::elapsedVTime() * 1000.0;
2125
2126 // (4) We check whether we should yield. If we have to, then we abort.
2127 if (SuspendibleThreadSet::should_yield()) {
2128 // We should yield. To do this we abort the task. The caller is
2129 // responsible for yielding.
2130 set_has_aborted();
2131 return;
2132 }
2133
2134 // (5) We check whether we've reached our time quota. If we have,
2135 // then we abort.
2136 double elapsed_time_ms = curr_time_ms - _start_time_ms;
2137 if (elapsed_time_ms > _time_target_ms) {
2138 set_has_aborted();
2139 _has_timed_out = true;
2140 return;
2141 }
2142
2143 // (6) Finally, we check whether there are enough completed STAB
2144 // buffers available for processing. If there are, we abort.
2145 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2146 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) {
2147 // we do need to process SATB buffers, we'll abort and restart
2148 // the marking task to do so
2149 set_has_aborted();
2150 return;
2151 }
2152 }
2153
2154 void G1CMTask::recalculate_limits() {
2155 _real_words_scanned_limit = _words_scanned + words_scanned_period;
2156 _words_scanned_limit = _real_words_scanned_limit;
2157
2158 _real_refs_reached_limit = _refs_reached + refs_reached_period;
2159 _refs_reached_limit = _real_refs_reached_limit;
2160 }
2161
2162 void G1CMTask::decrease_limits() {
2163 // This is called when we believe that we're going to do an infrequent
2164 // operation which will increase the per byte scanned cost (i.e. move
2165 // entries to/from the global stack). It basically tries to decrease the
2166 // scanning limit so that the clock is called earlier.
2167
2168 _words_scanned_limit = _real_words_scanned_limit - 3 * words_scanned_period / 4;
2169 _refs_reached_limit = _real_refs_reached_limit - 3 * refs_reached_period / 4;
2170 }
2171
2172 void G1CMTask::move_entries_to_global_stack() {
2173 // Local array where we'll store the entries that will be popped
2174 // from the local queue.
2175 G1TaskQueueEntry buffer[G1CMMarkStack::EntriesPerChunk];
2176
2177 size_t n = 0;
2178 G1TaskQueueEntry task_entry;
2179 while (n < G1CMMarkStack::EntriesPerChunk && _task_queue->pop_local(task_entry)) {
2180 buffer[n] = task_entry;
2181 ++n;
2182 }
2183 if (n < G1CMMarkStack::EntriesPerChunk) {
2184 buffer[n] = G1TaskQueueEntry();
2185 }
2186
2187 if (n > 0) {
2188 if (!_cm->mark_stack_push(buffer)) {
2189 set_has_aborted();
2190 }
2191 }
2192
2193 // This operation was quite expensive, so decrease the limits.
2194 decrease_limits();
2195 }
2196
2197 bool G1CMTask::get_entries_from_global_stack() {
2198 // Local array where we'll store the entries that will be popped
2199 // from the global stack.
2200 G1TaskQueueEntry buffer[G1CMMarkStack::EntriesPerChunk];
2201
2202 if (!_cm->mark_stack_pop(buffer)) {
2203 return false;
2204 }
2205
2206 // We did actually pop at least one entry.
2207 for (size_t i = 0; i < G1CMMarkStack::EntriesPerChunk; ++i) {
2208 G1TaskQueueEntry task_entry = buffer[i];
2209 if (task_entry.is_null()) {
2210 break;
2211 }
2212 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()));
2213 bool success = _task_queue->push(task_entry);
2214 // We only call this when the local queue is empty or under a
2215 // given target limit. So, we do not expect this push to fail.
2216 assert(success, "invariant");
2217 }
2218
2219 // This operation was quite expensive, so decrease the limits
2220 decrease_limits();
2221 return true;
2222 }
2223
2224 void G1CMTask::drain_local_queue(bool partially) {
2225 if (has_aborted()) {
2226 return;
2227 }
2228
2229 // Decide what the target size is, depending whether we're going to
2230 // drain it partially (so that other tasks can steal if they run out
2231 // of things to do) or totally (at the very end).
2232 size_t target_size;
2233 if (partially) {
2234 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize);
2235 } else {
2236 target_size = 0;
2237 }
2238
2239 if (_task_queue->size() > target_size) {
2240 G1TaskQueueEntry entry;
2241 bool ret = _task_queue->pop_local(entry);
2242 while (ret) {
2243 scan_task_entry(entry);
2244 if (_task_queue->size() <= target_size || has_aborted()) {
2245 ret = false;
2246 } else {
2247 ret = _task_queue->pop_local(entry);
2248 }
2249 }
2250 }
2251 }
2252
2253 void G1CMTask::drain_global_stack(bool partially) {
2254 if (has_aborted()) {
2255 return;
2256 }
2257
2258 // We have a policy to drain the local queue before we attempt to
2259 // drain the global stack.
2260 assert(partially || _task_queue->size() == 0, "invariant");
2261
2262 // Decide what the target size is, depending whether we're going to
2263 // drain it partially (so that other tasks can steal if they run out
2264 // of things to do) or totally (at the very end).
2265 // Notice that when draining the global mark stack partially, due to the racyness
2266 // of the mark stack size update we might in fact drop below the target. But,
2267 // this is not a problem.
2268 // In case of total draining, we simply process until the global mark stack is
2269 // totally empty, disregarding the size counter.
2270 if (partially) {
2271 size_t const target_size = _cm->partial_mark_stack_size_target();
2272 while (!has_aborted() && _cm->mark_stack_size() > target_size) {
2273 if (get_entries_from_global_stack()) {
2274 drain_local_queue(partially);
2275 }
2276 }
2277 } else {
2278 while (!has_aborted() && get_entries_from_global_stack()) {
2279 drain_local_queue(partially);
2280 }
2281 }
2282 }
2283
2284 // SATB Queue has several assumptions on whether to call the par or
2285 // non-par versions of the methods. this is why some of the code is
2286 // replicated. We should really get rid of the single-threaded version
2287 // of the code to simplify things.
2288 void G1CMTask::drain_satb_buffers() {
2289 if (has_aborted()) {
2290 return;
2291 }
2292
2293 // We set this so that the regular clock knows that we're in the
2294 // middle of draining buffers and doesn't set the abort flag when it
2295 // notices that SATB buffers are available for draining. It'd be
2296 // very counter productive if it did that. :-)
2297 _draining_satb_buffers = true;
2298
2299 G1CMSATBBufferClosure satb_cl(this, _g1h);
2300 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2301
2302 // This keeps claiming and applying the closure to completed buffers
2303 // until we run out of buffers or we need to abort.
2304 while (!has_aborted() &&
2305 satb_mq_set.apply_closure_to_completed_buffer(&satb_cl)) {
2306 regular_clock_call();
2307 }
2308
2309 _draining_satb_buffers = false;
2310
2311 assert(has_aborted() ||
2312 _cm->concurrent() ||
2313 satb_mq_set.completed_buffers_num() == 0, "invariant");
2314
2315 // again, this was a potentially expensive operation, decrease the
2316 // limits to get the regular clock call early
2317 decrease_limits();
2318 }
2319
2320 void G1CMTask::clear_mark_stats_cache(uint region_idx) {
2321 _mark_stats_cache.reset(region_idx);
2322 }
2323
2324 Pair<size_t, size_t> G1CMTask::flush_mark_stats_cache() {
2325 return _mark_stats_cache.evict_all();
2326 }
2327
2328 void G1CMTask::print_stats() {
2329 log_debug(gc, stats)("Marking Stats, task = %u, calls = %u", _worker_id, _calls);
2330 log_debug(gc, stats)(" Elapsed time = %1.2lfms, Termination time = %1.2lfms",
2331 _elapsed_time_ms, _termination_time_ms);
2332 log_debug(gc, stats)(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms max = %1.2lfms, total = %1.2lfms",
2333 _step_times_ms.num(),
2334 _step_times_ms.avg(),
2335 _step_times_ms.sd(),
2336 _step_times_ms.maximum(),
2337 _step_times_ms.sum());
2338 size_t const hits = _mark_stats_cache.hits();
2339 size_t const misses = _mark_stats_cache.misses();
2340 log_debug(gc, stats)(" Mark Stats Cache: hits " SIZE_FORMAT " misses " SIZE_FORMAT " ratio %.3f",
2341 hits, misses, percent_of(hits, hits + misses));
2342 }
2343
2344 bool G1ConcurrentMark::try_stealing(uint worker_id, int* hash_seed, G1TaskQueueEntry& task_entry) {
2345 return _task_queues->steal(worker_id, hash_seed, task_entry);
2346 }
2347
2348 /*****************************************************************************
2349
2350 The do_marking_step(time_target_ms, ...) method is the building
2351 block of the parallel marking framework. It can be called in parallel
2352 with other invocations of do_marking_step() on different tasks
2353 (but only one per task, obviously) and concurrently with the
2354 mutator threads, or during remark, hence it eliminates the need
2355 for two versions of the code. When called during remark, it will
2356 pick up from where the task left off during the concurrent marking
2357 phase. Interestingly, tasks are also claimable during evacuation
2358 pauses too, since do_marking_step() ensures that it aborts before
2359 it needs to yield.
2360
2361 The data structures that it uses to do marking work are the
2362 following:
2363
2364 (1) Marking Bitmap. If there are gray objects that appear only
2365 on the bitmap (this happens either when dealing with an overflow
2366 or when the initial marking phase has simply marked the roots
2367 and didn't push them on the stack), then tasks claim heap
2368 regions whose bitmap they then scan to find gray objects. A
2369 global finger indicates where the end of the last claimed region
2370 is. A local finger indicates how far into the region a task has
2371 scanned. The two fingers are used to determine how to gray an
2372 object (i.e. whether simply marking it is OK, as it will be
2373 visited by a task in the future, or whether it needs to be also
2374 pushed on a stack).
2375
2376 (2) Local Queue. The local queue of the task which is accessed
2377 reasonably efficiently by the task. Other tasks can steal from
2378 it when they run out of work. Throughout the marking phase, a
2379 task attempts to keep its local queue short but not totally
2380 empty, so that entries are available for stealing by other
2381 tasks. Only when there is no more work, a task will totally
2382 drain its local queue.
2383
2384 (3) Global Mark Stack. This handles local queue overflow. During
2385 marking only sets of entries are moved between it and the local
2386 queues, as access to it requires a mutex and more fine-grain
2387 interaction with it which might cause contention. If it
2388 overflows, then the marking phase should restart and iterate
2389 over the bitmap to identify gray objects. Throughout the marking
2390 phase, tasks attempt to keep the global mark stack at a small
2391 length but not totally empty, so that entries are available for
2392 popping by other tasks. Only when there is no more work, tasks
2393 will totally drain the global mark stack.
2394
2395 (4) SATB Buffer Queue. This is where completed SATB buffers are
2396 made available. Buffers are regularly removed from this queue
2397 and scanned for roots, so that the queue doesn't get too
2398 long. During remark, all completed buffers are processed, as
2399 well as the filled in parts of any uncompleted buffers.
2400
2401 The do_marking_step() method tries to abort when the time target
2402 has been reached. There are a few other cases when the
2403 do_marking_step() method also aborts:
2404
2405 (1) When the marking phase has been aborted (after a Full GC).
2406
2407 (2) When a global overflow (on the global stack) has been
2408 triggered. Before the task aborts, it will actually sync up with
2409 the other tasks to ensure that all the marking data structures
2410 (local queues, stacks, fingers etc.) are re-initialized so that
2411 when do_marking_step() completes, the marking phase can
2412 immediately restart.
2413
2414 (3) When enough completed SATB buffers are available. The
2415 do_marking_step() method only tries to drain SATB buffers right
2416 at the beginning. So, if enough buffers are available, the
2417 marking step aborts and the SATB buffers are processed at
2418 the beginning of the next invocation.
2419
2420 (4) To yield. when we have to yield then we abort and yield
2421 right at the end of do_marking_step(). This saves us from a lot
2422 of hassle as, by yielding we might allow a Full GC. If this
2423 happens then objects will be compacted underneath our feet, the
2424 heap might shrink, etc. We save checking for this by just
2425 aborting and doing the yield right at the end.
2426
2427 From the above it follows that the do_marking_step() method should
2428 be called in a loop (or, otherwise, regularly) until it completes.
2429
2430 If a marking step completes without its has_aborted() flag being
2431 true, it means it has completed the current marking phase (and
2432 also all other marking tasks have done so and have all synced up).
2433
2434 A method called regular_clock_call() is invoked "regularly" (in
2435 sub ms intervals) throughout marking. It is this clock method that
2436 checks all the abort conditions which were mentioned above and
2437 decides when the task should abort. A work-based scheme is used to
2438 trigger this clock method: when the number of object words the
2439 marking phase has scanned or the number of references the marking
2440 phase has visited reach a given limit. Additional invocations to
2441 the method clock have been planted in a few other strategic places
2442 too. The initial reason for the clock method was to avoid calling
2443 vtime too regularly, as it is quite expensive. So, once it was in
2444 place, it was natural to piggy-back all the other conditions on it
2445 too and not constantly check them throughout the code.
2446
2447 If do_termination is true then do_marking_step will enter its
2448 termination protocol.
2449
2450 The value of is_serial must be true when do_marking_step is being
2451 called serially (i.e. by the VMThread) and do_marking_step should
2452 skip any synchronization in the termination and overflow code.
2453 Examples include the serial remark code and the serial reference
2454 processing closures.
2455
2456 The value of is_serial must be false when do_marking_step is
2457 being called by any of the worker threads in a work gang.
2458 Examples include the concurrent marking code (CMMarkingTask),
2459 the MT remark code, and the MT reference processing closures.
2460
2461 *****************************************************************************/
2462
2463 void G1CMTask::do_marking_step(double time_target_ms,
2464 bool do_termination,
2465 bool is_serial) {
2466 assert(time_target_ms >= 1.0, "minimum granularity is 1ms");
2467
2468 _start_time_ms = os::elapsedVTime() * 1000.0;
2469
2470 // If do_stealing is true then do_marking_step will attempt to
2471 // steal work from the other G1CMTasks. It only makes sense to
2472 // enable stealing when the termination protocol is enabled
2473 // and do_marking_step() is not being called serially.
2474 bool do_stealing = do_termination && !is_serial;
2475
2476 double diff_prediction_ms = _g1h->g1_policy()->predictor().get_new_prediction(&_marking_step_diffs_ms);
2477 _time_target_ms = time_target_ms - diff_prediction_ms;
2478
2479 // set up the variables that are used in the work-based scheme to
2480 // call the regular clock method
2481 _words_scanned = 0;
2482 _refs_reached = 0;
2483 recalculate_limits();
2484
2485 // clear all flags
2486 clear_has_aborted();
2487 _has_timed_out = false;
2488 _draining_satb_buffers = false;
2489
2490 ++_calls;
2491
2492 // Set up the bitmap and oop closures. Anything that uses them is
2493 // eventually called from this method, so it is OK to allocate these
2494 // statically.
2495 G1CMBitMapClosure bitmap_closure(this, _cm);
2496 G1CMOopClosure cm_oop_closure(_g1h, this);
2497 set_cm_oop_closure(&cm_oop_closure);
2498
2499 if (_cm->has_overflown()) {
2500 // This can happen if the mark stack overflows during a GC pause
2501 // and this task, after a yield point, restarts. We have to abort
2502 // as we need to get into the overflow protocol which happens
2503 // right at the end of this task.
2504 set_has_aborted();
2505 }
2506
2507 // First drain any available SATB buffers. After this, we will not
2508 // look at SATB buffers before the next invocation of this method.
2509 // If enough completed SATB buffers are queued up, the regular clock
2510 // will abort this task so that it restarts.
2511 drain_satb_buffers();
2512 // ...then partially drain the local queue and the global stack
2513 drain_local_queue(true);
2514 drain_global_stack(true);
2515
2516 do {
2517 if (!has_aborted() && _curr_region != NULL) {
2518 // This means that we're already holding on to a region.
2519 assert(_finger != NULL, "if region is not NULL, then the finger "
2520 "should not be NULL either");
2521
2522 // We might have restarted this task after an evacuation pause
2523 // which might have evacuated the region we're holding on to
2524 // underneath our feet. Let's read its limit again to make sure
2525 // that we do not iterate over a region of the heap that
2526 // contains garbage (update_region_limit() will also move
2527 // _finger to the start of the region if it is found empty).
2528 update_region_limit();
2529 // We will start from _finger not from the start of the region,
2530 // as we might be restarting this task after aborting half-way
2531 // through scanning this region. In this case, _finger points to
2532 // the address where we last found a marked object. If this is a
2533 // fresh region, _finger points to start().
2534 MemRegion mr = MemRegion(_finger, _region_limit);
2535
2536 assert(!_curr_region->is_humongous() || mr.start() == _curr_region->bottom(),
2537 "humongous regions should go around loop once only");
2538
2539 // Some special cases:
2540 // If the memory region is empty, we can just give up the region.
2541 // If the current region is humongous then we only need to check
2542 // the bitmap for the bit associated with the start of the object,
2543 // scan the object if it's live, and give up the region.
2544 // Otherwise, let's iterate over the bitmap of the part of the region
2545 // that is left.
2546 // If the iteration is successful, give up the region.
2547 if (mr.is_empty()) {
2548 giveup_current_region();
2549 regular_clock_call();
2550 } else if (_curr_region->is_humongous() && mr.start() == _curr_region->bottom()) {
2551 if (_next_mark_bitmap->is_marked(mr.start())) {
2552 // The object is marked - apply the closure
2553 bitmap_closure.do_addr(mr.start());
2554 }
2555 // Even if this task aborted while scanning the humongous object
2556 // we can (and should) give up the current region.
2557 giveup_current_region();
2558 regular_clock_call();
2559 } else if (_next_mark_bitmap->iterate(&bitmap_closure, mr)) {
2560 giveup_current_region();
2561 regular_clock_call();
2562 } else {
2563 assert(has_aborted(), "currently the only way to do so");
2564 // The only way to abort the bitmap iteration is to return
2565 // false from the do_bit() method. However, inside the
2566 // do_bit() method we move the _finger to point to the
2567 // object currently being looked at. So, if we bail out, we
2568 // have definitely set _finger to something non-null.
2569 assert(_finger != NULL, "invariant");
2570
2571 // Region iteration was actually aborted. So now _finger
2572 // points to the address of the object we last scanned. If we
2573 // leave it there, when we restart this task, we will rescan
2574 // the object. It is easy to avoid this. We move the finger by
2575 // enough to point to the next possible object header.
2576 assert(_finger < _region_limit, "invariant");
2577 HeapWord* const new_finger = _finger + ((oop)_finger)->size();
2578 // Check if bitmap iteration was aborted while scanning the last object
2579 if (new_finger >= _region_limit) {
2580 giveup_current_region();
2581 } else {
2582 move_finger_to(new_finger);
2583 }
2584 }
2585 }
2586 // At this point we have either completed iterating over the
2587 // region we were holding on to, or we have aborted.
2588
2589 // We then partially drain the local queue and the global stack.
2590 // (Do we really need this?)
2591 drain_local_queue(true);
2592 drain_global_stack(true);
2593
2594 // Read the note on the claim_region() method on why it might
2595 // return NULL with potentially more regions available for
2596 // claiming and why we have to check out_of_regions() to determine
2597 // whether we're done or not.
2598 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) {
2599 // We are going to try to claim a new region. We should have
2600 // given up on the previous one.
2601 // Separated the asserts so that we know which one fires.
2602 assert(_curr_region == NULL, "invariant");
2603 assert(_finger == NULL, "invariant");
2604 assert(_region_limit == NULL, "invariant");
2605 HeapRegion* claimed_region = _cm->claim_region(_worker_id);
2606 if (claimed_region != NULL) {
2607 // Yes, we managed to claim one
2608 setup_for_region(claimed_region);
2609 assert(_curr_region == claimed_region, "invariant");
2610 }
2611 // It is important to call the regular clock here. It might take
2612 // a while to claim a region if, for example, we hit a large
2613 // block of empty regions. So we need to call the regular clock
2614 // method once round the loop to make sure it's called
2615 // frequently enough.
2616 regular_clock_call();
2617 }
2618
2619 if (!has_aborted() && _curr_region == NULL) {
2620 assert(_cm->out_of_regions(),
2621 "at this point we should be out of regions");
2622 }
2623 } while ( _curr_region != NULL && !has_aborted());
2624
2625 if (!has_aborted()) {
2626 // We cannot check whether the global stack is empty, since other
2627 // tasks might be pushing objects to it concurrently.
2628 assert(_cm->out_of_regions(),
2629 "at this point we should be out of regions");
2630 // Try to reduce the number of available SATB buffers so that
2631 // remark has less work to do.
2632 drain_satb_buffers();
2633 }
2634
2635 // Since we've done everything else, we can now totally drain the
2636 // local queue and global stack.
2637 drain_local_queue(false);
2638 drain_global_stack(false);
2639
2640 // Attempt at work stealing from other task's queues.
2641 if (do_stealing && !has_aborted()) {
2642 // We have not aborted. This means that we have finished all that
2643 // we could. Let's try to do some stealing...
2644
2645 // We cannot check whether the global stack is empty, since other
2646 // tasks might be pushing objects to it concurrently.
2647 assert(_cm->out_of_regions() && _task_queue->size() == 0,
2648 "only way to reach here");
2649 while (!has_aborted()) {
2650 G1TaskQueueEntry entry;
2651 if (_cm->try_stealing(_worker_id, &_hash_seed, entry)) {
2652 scan_task_entry(entry);
2653
2654 // And since we're towards the end, let's totally drain the
2655 // local queue and global stack.
2656 drain_local_queue(false);
2657 drain_global_stack(false);
2658 } else {
2659 break;
2660 }
2661 }
2662 }
2663
2664 // We still haven't aborted. Now, let's try to get into the
2665 // termination protocol.
2666 if (do_termination && !has_aborted()) {
2667 // We cannot check whether the global stack is empty, since other
2668 // tasks might be concurrently pushing objects on it.
2669 // Separated the asserts so that we know which one fires.
2670 assert(_cm->out_of_regions(), "only way to reach here");
2671 assert(_task_queue->size() == 0, "only way to reach here");
2672 _termination_start_time_ms = os::elapsedVTime() * 1000.0;
2673
2674 // The G1CMTask class also extends the TerminatorTerminator class,
2675 // hence its should_exit_termination() method will also decide
2676 // whether to exit the termination protocol or not.
2677 bool finished = (is_serial ||
2678 _cm->terminator()->offer_termination(this));
2679 double termination_end_time_ms = os::elapsedVTime() * 1000.0;
2680 _termination_time_ms +=
2681 termination_end_time_ms - _termination_start_time_ms;
2682
2683 if (finished) {
2684 // We're all done.
2685
2686 // We can now guarantee that the global stack is empty, since
2687 // all other tasks have finished. We separated the guarantees so
2688 // that, if a condition is false, we can immediately find out
2689 // which one.
2690 guarantee(_cm->out_of_regions(), "only way to reach here");
2691 guarantee(_cm->mark_stack_empty(), "only way to reach here");
2692 guarantee(_task_queue->size() == 0, "only way to reach here");
2693 guarantee(!_cm->has_overflown(), "only way to reach here");
2694 } else {
2695 // Apparently there's more work to do. Let's abort this task. It
2696 // will restart it and we can hopefully find more things to do.
2697 set_has_aborted();
2698 }
2699 }
2700
2701 // Mainly for debugging purposes to make sure that a pointer to the
2702 // closure which was statically allocated in this frame doesn't
2703 // escape it by accident.
2704 set_cm_oop_closure(NULL);
2705 double end_time_ms = os::elapsedVTime() * 1000.0;
2706 double elapsed_time_ms = end_time_ms - _start_time_ms;
2707 // Update the step history.
2708 _step_times_ms.add(elapsed_time_ms);
2709
2710 if (has_aborted()) {
2711 // The task was aborted for some reason.
2712 if (_has_timed_out) {
2713 double diff_ms = elapsed_time_ms - _time_target_ms;
2714 // Keep statistics of how well we did with respect to hitting
2715 // our target only if we actually timed out (if we aborted for
2716 // other reasons, then the results might get skewed).
2717 _marking_step_diffs_ms.add(diff_ms);
2718 }
2719
2720 if (_cm->has_overflown()) {
2721 // This is the interesting one. We aborted because a global
2722 // overflow was raised. This means we have to restart the
2723 // marking phase and start iterating over regions. However, in
2724 // order to do this we have to make sure that all tasks stop
2725 // what they are doing and re-initialize in a safe manner. We
2726 // will achieve this with the use of two barrier sync points.
2727
2728 if (!is_serial) {
2729 // We only need to enter the sync barrier if being called
2730 // from a parallel context
2731 _cm->enter_first_sync_barrier(_worker_id);
2732
2733 // When we exit this sync barrier we know that all tasks have
2734 // stopped doing marking work. So, it's now safe to
2735 // re-initialize our data structures.
2736 }
2737
2738 clear_region_fields();
2739 flush_mark_stats_cache();
2740
2741 if (!is_serial) {
2742 // If we're executing the concurrent phase of marking, reset the marking
2743 // state; otherwise the marking state is reset after reference processing,
2744 // during the remark pause.
2745 // If we reset here as a result of an overflow during the remark we will
2746 // see assertion failures from any subsequent set_concurrency_and_phase()
2747 // calls.
2748 if (_cm->concurrent() && _worker_id == 0) {
2749 // Worker 0 is responsible for clearing the global data structures because
2750 // of an overflow. During STW we should not clear the overflow flag (in
2751 // G1ConcurrentMark::reset_marking_state()) since we rely on it being true when we exit
2752 // method to abort the pause and restart concurrent marking.
2753 _cm->reset_marking_for_restart();
2754
2755 log_info(gc, marking)("Concurrent Mark reset for overflow");
2756 }
2757
2758 // ...and enter the second barrier.
2759 _cm->enter_second_sync_barrier(_worker_id);
2760 }
2761 // At this point, if we're during the concurrent phase of
2762 // marking, everything has been re-initialized and we're
2763 // ready to restart.
2764 }
2765 }
2766 }
2767
2768 G1CMTask::G1CMTask(uint worker_id,
2769 G1ConcurrentMark* cm,
2770 G1CMTaskQueue* task_queue,
2771 G1RegionMarkStats* mark_stats,
2772 uint max_regions) :
2773 _objArray_processor(this),
2774 _worker_id(worker_id),
2775 _g1h(G1CollectedHeap::heap()),
2776 _cm(cm),
2777 _next_mark_bitmap(NULL),
2778 _task_queue(task_queue),
2779 _mark_stats_cache(mark_stats, max_regions, RegionMarkStatsCacheSize),
2780 _calls(0),
2781 _time_target_ms(0.0),
2782 _start_time_ms(0.0),
2783 _cm_oop_closure(NULL),
2784 _curr_region(NULL),
2785 _finger(NULL),
2786 _region_limit(NULL),
2787 _words_scanned(0),
2788 _words_scanned_limit(0),
2789 _real_words_scanned_limit(0),
2790 _refs_reached(0),
2791 _refs_reached_limit(0),
2792 _real_refs_reached_limit(0),
2793 _hash_seed(17),
2794 _has_aborted(false),
2795 _has_timed_out(false),
2796 _draining_satb_buffers(false),
2797 _step_times_ms(),
2798 _elapsed_time_ms(0.0),
2799 _termination_time_ms(0.0),
2800 _termination_start_time_ms(0.0),
2801 _marking_step_diffs_ms()
2802 {
2803 guarantee(task_queue != NULL, "invariant");
2804
2805 _marking_step_diffs_ms.add(0.5);
2806 }
2807
2808 // These are formatting macros that are used below to ensure
2809 // consistent formatting. The *_H_* versions are used to format the
2810 // header for a particular value and they should be kept consistent
2811 // with the corresponding macro. Also note that most of the macros add
2812 // the necessary white space (as a prefix) which makes them a bit
2813 // easier to compose.
2814
2815 // All the output lines are prefixed with this string to be able to
2816 // identify them easily in a large log file.
2817 #define G1PPRL_LINE_PREFIX "###"
2818
2819 #define G1PPRL_ADDR_BASE_FORMAT " " PTR_FORMAT "-" PTR_FORMAT
2820 #ifdef _LP64
2821 #define G1PPRL_ADDR_BASE_H_FORMAT " %37s"
2822 #else // _LP64
2823 #define G1PPRL_ADDR_BASE_H_FORMAT " %21s"
2824 #endif // _LP64
2825
2826 // For per-region info
2827 #define G1PPRL_TYPE_FORMAT " %-4s"
2828 #define G1PPRL_TYPE_H_FORMAT " %4s"
2829 #define G1PPRL_STATE_FORMAT " %-5s"
2830 #define G1PPRL_STATE_H_FORMAT " %5s"
2831 #define G1PPRL_BYTE_FORMAT " " SIZE_FORMAT_W(9)
2832 #define G1PPRL_BYTE_H_FORMAT " %9s"
2833 #define G1PPRL_DOUBLE_FORMAT " %14.1f"
2834 #define G1PPRL_DOUBLE_H_FORMAT " %14s"
2835
2836 // For summary info
2837 #define G1PPRL_SUM_ADDR_FORMAT(tag) " " tag ":" G1PPRL_ADDR_BASE_FORMAT
2838 #define G1PPRL_SUM_BYTE_FORMAT(tag) " " tag ": " SIZE_FORMAT
2839 #define G1PPRL_SUM_MB_FORMAT(tag) " " tag ": %1.2f MB"
2840 #define G1PPRL_SUM_MB_PERC_FORMAT(tag) G1PPRL_SUM_MB_FORMAT(tag) " / %1.2f %%"
2841
2842 G1PrintRegionLivenessInfoClosure::G1PrintRegionLivenessInfoClosure(const char* phase_name) :
2843 _total_used_bytes(0), _total_capacity_bytes(0),
2844 _total_prev_live_bytes(0), _total_next_live_bytes(0),
2845 _total_remset_bytes(0), _total_strong_code_roots_bytes(0)
2846 {
2847 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2848 MemRegion g1_reserved = g1h->g1_reserved();
2849 double now = os::elapsedTime();
2850
2851 // Print the header of the output.
2852 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX" PHASE %s @ %1.3f", phase_name, now);
2853 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX" HEAP"
2854 G1PPRL_SUM_ADDR_FORMAT("reserved")
2855 G1PPRL_SUM_BYTE_FORMAT("region-size"),
2856 p2i(g1_reserved.start()), p2i(g1_reserved.end()),
2857 HeapRegion::GrainBytes);
2858 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX);
2859 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
2860 G1PPRL_TYPE_H_FORMAT
2861 G1PPRL_ADDR_BASE_H_FORMAT
2862 G1PPRL_BYTE_H_FORMAT
2863 G1PPRL_BYTE_H_FORMAT
2864 G1PPRL_BYTE_H_FORMAT
2865 G1PPRL_DOUBLE_H_FORMAT
2866 G1PPRL_BYTE_H_FORMAT
2867 G1PPRL_STATE_H_FORMAT
2868 G1PPRL_BYTE_H_FORMAT,
2869 "type", "address-range",
2870 "used", "prev-live", "next-live", "gc-eff",
2871 "remset", "state", "code-roots");
2872 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
2873 G1PPRL_TYPE_H_FORMAT
2874 G1PPRL_ADDR_BASE_H_FORMAT
2875 G1PPRL_BYTE_H_FORMAT
2876 G1PPRL_BYTE_H_FORMAT
2877 G1PPRL_BYTE_H_FORMAT
2878 G1PPRL_DOUBLE_H_FORMAT
2879 G1PPRL_BYTE_H_FORMAT
2880 G1PPRL_STATE_H_FORMAT
2881 G1PPRL_BYTE_H_FORMAT,
2882 "", "",
2883 "(bytes)", "(bytes)", "(bytes)", "(bytes/ms)",
2884 "(bytes)", "", "(bytes)");
2885 }
2886
2887 bool G1PrintRegionLivenessInfoClosure::do_heap_region(HeapRegion* r) {
2888 const char* type = r->get_type_str();
2889 HeapWord* bottom = r->bottom();
2890 HeapWord* end = r->end();
2891 size_t capacity_bytes = r->capacity();
2892 size_t used_bytes = r->used();
2893 size_t prev_live_bytes = r->live_bytes();
2894 size_t next_live_bytes = r->next_live_bytes();
2895 double gc_eff = r->gc_efficiency();
2896 size_t remset_bytes = r->rem_set()->mem_size();
2897 size_t strong_code_roots_bytes = r->rem_set()->strong_code_roots_mem_size();
2898 const char* remset_type = r->rem_set()->get_short_state_str();
2899
2900 _total_used_bytes += used_bytes;
2901 _total_capacity_bytes += capacity_bytes;
2902 _total_prev_live_bytes += prev_live_bytes;
2903 _total_next_live_bytes += next_live_bytes;
2904 _total_remset_bytes += remset_bytes;
2905 _total_strong_code_roots_bytes += strong_code_roots_bytes;
2906
2907 // Print a line for this particular region.
2908 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
2909 G1PPRL_TYPE_FORMAT
2910 G1PPRL_ADDR_BASE_FORMAT
2911 G1PPRL_BYTE_FORMAT
2912 G1PPRL_BYTE_FORMAT
2913 G1PPRL_BYTE_FORMAT
2914 G1PPRL_DOUBLE_FORMAT
2915 G1PPRL_BYTE_FORMAT
2916 G1PPRL_STATE_FORMAT
2917 G1PPRL_BYTE_FORMAT,
2918 type, p2i(bottom), p2i(end),
2919 used_bytes, prev_live_bytes, next_live_bytes, gc_eff,
2920 remset_bytes, remset_type, strong_code_roots_bytes);
2921
2922 return false;
2923 }
2924
2925 G1PrintRegionLivenessInfoClosure::~G1PrintRegionLivenessInfoClosure() {
2926 // add static memory usages to remembered set sizes
2927 _total_remset_bytes += HeapRegionRemSet::fl_mem_size() + HeapRegionRemSet::static_mem_size();
2928 // Print the footer of the output.
2929 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX);
2930 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
2931 " SUMMARY"
2932 G1PPRL_SUM_MB_FORMAT("capacity")
2933 G1PPRL_SUM_MB_PERC_FORMAT("used")
2934 G1PPRL_SUM_MB_PERC_FORMAT("prev-live")
2935 G1PPRL_SUM_MB_PERC_FORMAT("next-live")
2936 G1PPRL_SUM_MB_FORMAT("remset")
2937 G1PPRL_SUM_MB_FORMAT("code-roots"),
2938 bytes_to_mb(_total_capacity_bytes),
2939 bytes_to_mb(_total_used_bytes),
2940 percent_of(_total_used_bytes, _total_capacity_bytes),
2941 bytes_to_mb(_total_prev_live_bytes),
2942 percent_of(_total_prev_live_bytes, _total_capacity_bytes),
2943 bytes_to_mb(_total_next_live_bytes),
2944 percent_of(_total_next_live_bytes, _total_capacity_bytes),
2945 bytes_to_mb(_total_remset_bytes),
2946 bytes_to_mb(_total_strong_code_roots_bytes));
2947 }