1 /*
2 * Copyright (c) 2001, 2016, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #include "precompiled.hpp"
26 #include "classfile/metadataOnStackMark.hpp"
27 #include "classfile/symbolTable.hpp"
28 #include "code/codeCache.hpp"
29 #include "gc/g1/concurrentMarkThread.inline.hpp"
30 #include "gc/g1/g1CollectedHeap.inline.hpp"
31 #include "gc/g1/g1CollectorPolicy.hpp"
32 #include "gc/g1/g1CollectorState.hpp"
33 #include "gc/g1/g1ConcurrentMark.inline.hpp"
34 #include "gc/g1/g1HeapVerifier.hpp"
35 #include "gc/g1/g1OopClosures.inline.hpp"
36 #include "gc/g1/g1StringDedup.hpp"
37 #include "gc/g1/heapRegion.inline.hpp"
38 #include "gc/g1/heapRegionRemSet.hpp"
39 #include "gc/g1/heapRegionSet.inline.hpp"
40 #include "gc/g1/suspendibleThreadSet.hpp"
41 #include "gc/shared/gcId.hpp"
42 #include "gc/shared/gcTimer.hpp"
43 #include "gc/shared/gcTrace.hpp"
44 #include "gc/shared/gcTraceTime.inline.hpp"
45 #include "gc/shared/genOopClosures.inline.hpp"
46 #include "gc/shared/referencePolicy.hpp"
47 #include "gc/shared/strongRootsScope.hpp"
48 #include "gc/shared/taskqueue.inline.hpp"
49 #include "gc/shared/vmGCOperations.hpp"
50 #include "logging/log.hpp"
51 #include "logging/logTag.hpp"
52 #include "memory/allocation.hpp"
53 #include "memory/resourceArea.hpp"
54 #include "oops/oop.inline.hpp"
55 #include "runtime/atomic.inline.hpp"
56 #include "runtime/handles.inline.hpp"
57 #include "runtime/java.hpp"
58 #include "runtime/prefetch.inline.hpp"
59 #include "services/memTracker.hpp"
60
61 // Concurrent marking bit map wrapper
62
63 G1CMBitMapRO::G1CMBitMapRO(int shifter) :
64 _bm(),
65 _shifter(shifter) {
66 _bmStartWord = 0;
67 _bmWordSize = 0;
68 }
69
70 HeapWord* G1CMBitMapRO::getNextMarkedWordAddress(const HeapWord* addr,
71 const HeapWord* limit) const {
72 // First we must round addr *up* to a possible object boundary.
73 addr = (HeapWord*)align_size_up((intptr_t)addr,
74 HeapWordSize << _shifter);
75 size_t addrOffset = heapWordToOffset(addr);
76 assert(limit != NULL, "limit must not be NULL");
77 size_t limitOffset = heapWordToOffset(limit);
78 size_t nextOffset = _bm.get_next_one_offset(addrOffset, limitOffset);
79 HeapWord* nextAddr = offsetToHeapWord(nextOffset);
80 assert(nextAddr >= addr, "get_next_one postcondition");
81 assert(nextAddr == limit || isMarked(nextAddr),
82 "get_next_one postcondition");
83 return nextAddr;
84 }
85
86 #ifndef PRODUCT
87 bool G1CMBitMapRO::covers(MemRegion heap_rs) const {
88 // assert(_bm.map() == _virtual_space.low(), "map inconsistency");
89 assert(((size_t)_bm.size() * ((size_t)1 << _shifter)) == _bmWordSize,
90 "size inconsistency");
91 return _bmStartWord == (HeapWord*)(heap_rs.start()) &&
92 _bmWordSize == heap_rs.word_size();
93 }
94 #endif
95
96 void G1CMBitMapRO::print_on_error(outputStream* st, const char* prefix) const {
97 _bm.print_on_error(st, prefix);
98 }
99
100 size_t G1CMBitMap::compute_size(size_t heap_size) {
101 return ReservedSpace::allocation_align_size_up(heap_size / mark_distance());
102 }
103
104 size_t G1CMBitMap::mark_distance() {
105 return MinObjAlignmentInBytes * BitsPerByte;
106 }
107
108 void G1CMBitMap::initialize(MemRegion heap, G1RegionToSpaceMapper* storage) {
109 _bmStartWord = heap.start();
110 _bmWordSize = heap.word_size();
111
112 _bm.set_map((BitMap::bm_word_t*) storage->reserved().start());
113 _bm.set_size(_bmWordSize >> _shifter);
114
115 storage->set_mapping_changed_listener(&_listener);
116 }
117
118 void G1CMBitMapMappingChangedListener::on_commit(uint start_region, size_t num_regions, bool zero_filled) {
119 if (zero_filled) {
120 return;
121 }
122 // We need to clear the bitmap on commit, removing any existing information.
123 MemRegion mr(G1CollectedHeap::heap()->bottom_addr_for_region(start_region), num_regions * HeapRegion::GrainWords);
124 _bm->clear_range(mr);
125 }
126
127 void G1CMBitMap::clear_range(MemRegion mr) {
128 mr.intersection(MemRegion(_bmStartWord, _bmWordSize));
129 assert(!mr.is_empty(), "unexpected empty region");
130 // convert address range into offset range
131 _bm.at_put_range(heapWordToOffset(mr.start()),
132 heapWordToOffset(mr.end()), false);
133 }
134
135 G1CMMarkStack::G1CMMarkStack(G1ConcurrentMark* cm) :
136 _base(NULL), _cm(cm)
137 {}
138
139 bool G1CMMarkStack::allocate(size_t capacity) {
140 // allocate a stack of the requisite depth
141 ReservedSpace rs(ReservedSpace::allocation_align_size_up(capacity * sizeof(oop)));
142 if (!rs.is_reserved()) {
143 log_warning(gc)("ConcurrentMark MarkStack allocation failure");
144 return false;
145 }
146 MemTracker::record_virtual_memory_type((address)rs.base(), mtGC);
147 if (!_virtual_space.initialize(rs, rs.size())) {
148 log_warning(gc)("ConcurrentMark MarkStack backing store failure");
149 // Release the virtual memory reserved for the marking stack
150 rs.release();
151 return false;
152 }
153 assert(_virtual_space.committed_size() == rs.size(),
154 "Didn't reserve backing store for all of G1ConcurrentMark stack?");
155 _base = (oop*) _virtual_space.low();
156 setEmpty();
157 _capacity = (jint) capacity;
158 _saved_index = -1;
159 _should_expand = false;
160 return true;
161 }
162
163 void G1CMMarkStack::expand() {
164 // Called, during remark, if we've overflown the marking stack during marking.
165 assert(isEmpty(), "stack should been emptied while handling overflow");
166 assert(_capacity <= (jint) MarkStackSizeMax, "stack bigger than permitted");
167 // Clear expansion flag
168 _should_expand = false;
169 if (_capacity == (jint) MarkStackSizeMax) {
170 log_trace(gc)("(benign) Can't expand marking stack capacity, at max size limit");
171 return;
172 }
173 // Double capacity if possible
174 jint new_capacity = MIN2(_capacity*2, (jint) MarkStackSizeMax);
175 // Do not give up existing stack until we have managed to
176 // get the double capacity that we desired.
177 ReservedSpace rs(ReservedSpace::allocation_align_size_up(new_capacity *
178 sizeof(oop)));
179 if (rs.is_reserved()) {
180 // Release the backing store associated with old stack
181 _virtual_space.release();
182 // Reinitialize virtual space for new stack
183 if (!_virtual_space.initialize(rs, rs.size())) {
184 fatal("Not enough swap for expanded marking stack capacity");
185 }
186 _base = (oop*)(_virtual_space.low());
187 _index = 0;
188 _capacity = new_capacity;
189 } else {
190 // Failed to double capacity, continue;
191 log_trace(gc)("(benign) Failed to expand marking stack capacity from " SIZE_FORMAT "K to " SIZE_FORMAT "K",
192 _capacity / K, new_capacity / K);
193 }
194 }
195
196 void G1CMMarkStack::set_should_expand() {
197 // If we're resetting the marking state because of an
198 // marking stack overflow, record that we should, if
199 // possible, expand the stack.
200 _should_expand = _cm->has_overflown();
201 }
202
203 G1CMMarkStack::~G1CMMarkStack() {
204 if (_base != NULL) {
205 _base = NULL;
206 _virtual_space.release();
207 }
208 }
209
210 void G1CMMarkStack::par_push_arr(oop* ptr_arr, int n) {
211 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
212 jint start = _index;
213 jint next_index = start + n;
214 if (next_index > _capacity) {
215 _overflow = true;
216 return;
217 }
218 // Otherwise.
219 _index = next_index;
220 for (int i = 0; i < n; i++) {
221 int ind = start + i;
222 assert(ind < _capacity, "By overflow test above.");
223 _base[ind] = ptr_arr[i];
224 }
225 }
226
227 bool G1CMMarkStack::par_pop_arr(oop* ptr_arr, int max, int* n) {
228 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
229 jint index = _index;
230 if (index == 0) {
231 *n = 0;
232 return false;
233 } else {
234 int k = MIN2(max, index);
235 jint new_ind = index - k;
236 for (int j = 0; j < k; j++) {
237 ptr_arr[j] = _base[new_ind + j];
238 }
239 _index = new_ind;
240 *n = k;
241 return true;
242 }
243 }
244
245 void G1CMMarkStack::note_start_of_gc() {
246 assert(_saved_index == -1,
247 "note_start_of_gc()/end_of_gc() bracketed incorrectly");
248 _saved_index = _index;
249 }
250
251 void G1CMMarkStack::note_end_of_gc() {
252 // This is intentionally a guarantee, instead of an assert. If we
253 // accidentally add something to the mark stack during GC, it
254 // will be a correctness issue so it's better if we crash. we'll
255 // only check this once per GC anyway, so it won't be a performance
256 // issue in any way.
257 guarantee(_saved_index == _index,
258 "saved index: %d index: %d", _saved_index, _index);
259 _saved_index = -1;
260 }
261
262 G1CMRootRegions::G1CMRootRegions() :
263 _young_list(NULL), _cm(NULL), _scan_in_progress(false),
264 _should_abort(false), _next_survivor(NULL) { }
265
266 void G1CMRootRegions::init(G1CollectedHeap* g1h, G1ConcurrentMark* cm) {
267 _young_list = g1h->young_list();
268 _cm = cm;
269 }
270
271 void G1CMRootRegions::prepare_for_scan() {
272 assert(!scan_in_progress(), "pre-condition");
273
274 // Currently, only survivors can be root regions.
275 assert(_next_survivor == NULL, "pre-condition");
276 _next_survivor = _young_list->first_survivor_region();
277 _scan_in_progress = (_next_survivor != NULL);
278 _should_abort = false;
279 }
280
281 HeapRegion* G1CMRootRegions::claim_next() {
282 if (_should_abort) {
283 // If someone has set the should_abort flag, we return NULL to
284 // force the caller to bail out of their loop.
285 return NULL;
286 }
287
288 // Currently, only survivors can be root regions.
289 HeapRegion* res = _next_survivor;
290 if (res != NULL) {
291 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
292 // Read it again in case it changed while we were waiting for the lock.
293 res = _next_survivor;
294 if (res != NULL) {
295 if (res == _young_list->last_survivor_region()) {
296 // We just claimed the last survivor so store NULL to indicate
297 // that we're done.
298 _next_survivor = NULL;
299 } else {
300 _next_survivor = res->get_next_young_region();
301 }
302 } else {
303 // Someone else claimed the last survivor while we were trying
304 // to take the lock so nothing else to do.
305 }
306 }
307 assert(res == NULL || res->is_survivor(), "post-condition");
308
309 return res;
310 }
311
312 void G1CMRootRegions::notify_scan_done() {
313 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
314 _scan_in_progress = false;
315 RootRegionScan_lock->notify_all();
316 }
317
318 void G1CMRootRegions::cancel_scan() {
319 notify_scan_done();
320 }
321
322 void G1CMRootRegions::scan_finished() {
323 assert(scan_in_progress(), "pre-condition");
324
325 // Currently, only survivors can be root regions.
326 if (!_should_abort) {
327 assert(_next_survivor == NULL, "we should have claimed all survivors");
328 }
329 _next_survivor = NULL;
330
331 notify_scan_done();
332 }
333
334 bool G1CMRootRegions::wait_until_scan_finished() {
335 if (!scan_in_progress()) return false;
336
337 {
338 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
339 while (scan_in_progress()) {
340 RootRegionScan_lock->wait(Mutex::_no_safepoint_check_flag);
341 }
342 }
343 return true;
344 }
345
346 uint G1ConcurrentMark::scale_parallel_threads(uint n_par_threads) {
347 return MAX2((n_par_threads + 2) / 4, 1U);
348 }
349
350 G1ConcurrentMark::G1ConcurrentMark(G1CollectedHeap* g1h, G1RegionToSpaceMapper* prev_bitmap_storage, G1RegionToSpaceMapper* next_bitmap_storage) :
351 _g1h(g1h),
352 _markBitMap1(),
353 _markBitMap2(),
354 _parallel_marking_threads(0),
355 _max_parallel_marking_threads(0),
356 _sleep_factor(0.0),
357 _marking_task_overhead(1.0),
358 _cleanup_list("Cleanup List"),
359 _region_live_bm(),
360 _card_live_bm(),
361
362 _prevMarkBitMap(&_markBitMap1),
363 _nextMarkBitMap(&_markBitMap2),
364
365 _markStack(this),
366 // _finger set in set_non_marking_state
367
368 _max_worker_id(ParallelGCThreads),
369 // _active_tasks set in set_non_marking_state
370 // _tasks set inside the constructor
371 _task_queues(new G1CMTaskQueueSet((int) _max_worker_id)),
372 _terminator(ParallelTaskTerminator((int) _max_worker_id, _task_queues)),
373
374 _has_overflown(false),
375 _concurrent(false),
376 _has_aborted(false),
377 _restart_for_overflow(false),
378 _concurrent_marking_in_progress(false),
379 _gc_timer_cm(new (ResourceObj::C_HEAP, mtGC) ConcurrentGCTimer()),
380 _gc_tracer_cm(new (ResourceObj::C_HEAP, mtGC) G1OldTracer()),
381
382 // _verbose_level set below
383
384 _init_times(),
385 _remark_times(), _remark_mark_times(), _remark_weak_ref_times(),
386 _cleanup_times(),
387 _total_counting_time(0.0),
388 _total_rs_scrub_time(0.0),
389
390 _parallel_workers(NULL),
391
392 _completed_initialization(false) {
393
394 _markBitMap1.initialize(g1h->reserved_region(), prev_bitmap_storage);
395 _markBitMap2.initialize(g1h->reserved_region(), next_bitmap_storage);
396
397 // Create & start a ConcurrentMark thread.
398 _cmThread = new ConcurrentMarkThread(this);
399 assert(cmThread() != NULL, "CM Thread should have been created");
400 assert(cmThread()->cm() != NULL, "CM Thread should refer to this cm");
401 if (_cmThread->osthread() == NULL) {
402 vm_shutdown_during_initialization("Could not create ConcurrentMarkThread");
403 }
404
405 assert(CGC_lock != NULL, "Where's the CGC_lock?");
406 assert(_markBitMap1.covers(g1h->reserved_region()), "_markBitMap1 inconsistency");
407 assert(_markBitMap2.covers(g1h->reserved_region()), "_markBitMap2 inconsistency");
408
409 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set();
410 satb_qs.set_buffer_size(G1SATBBufferSize);
411
412 _root_regions.init(_g1h, this);
413
414 if (ConcGCThreads > ParallelGCThreads) {
415 log_warning(gc)("Can't have more ConcGCThreads (%u) than ParallelGCThreads (%u).",
416 ConcGCThreads, ParallelGCThreads);
417 return;
418 }
419 if (!FLAG_IS_DEFAULT(ConcGCThreads) && ConcGCThreads > 0) {
420 // Note: ConcGCThreads has precedence over G1MarkingOverheadPercent
421 // if both are set
422 _sleep_factor = 0.0;
423 _marking_task_overhead = 1.0;
424 } else if (G1MarkingOverheadPercent > 0) {
425 // We will calculate the number of parallel marking threads based
426 // on a target overhead with respect to the soft real-time goal
427 double marking_overhead = (double) G1MarkingOverheadPercent / 100.0;
428 double overall_cm_overhead =
429 (double) MaxGCPauseMillis * marking_overhead /
430 (double) GCPauseIntervalMillis;
431 double cpu_ratio = 1.0 / (double) os::processor_count();
432 double marking_thread_num = ceil(overall_cm_overhead / cpu_ratio);
433 double marking_task_overhead =
434 overall_cm_overhead / marking_thread_num *
435 (double) os::processor_count();
436 double sleep_factor =
437 (1.0 - marking_task_overhead) / marking_task_overhead;
438
439 FLAG_SET_ERGO(uint, ConcGCThreads, (uint) marking_thread_num);
440 _sleep_factor = sleep_factor;
441 _marking_task_overhead = marking_task_overhead;
442 } else {
443 // Calculate the number of parallel marking threads by scaling
444 // the number of parallel GC threads.
445 uint marking_thread_num = scale_parallel_threads(ParallelGCThreads);
446 FLAG_SET_ERGO(uint, ConcGCThreads, marking_thread_num);
447 _sleep_factor = 0.0;
448 _marking_task_overhead = 1.0;
449 }
450
451 assert(ConcGCThreads > 0, "Should have been set");
452 _parallel_marking_threads = ConcGCThreads;
453 _max_parallel_marking_threads = _parallel_marking_threads;
454
455 _parallel_workers = new WorkGang("G1 Marker",
456 _max_parallel_marking_threads, false, true);
457 if (_parallel_workers == NULL) {
458 vm_exit_during_initialization("Failed necessary allocation.");
459 } else {
460 _parallel_workers->initialize_workers();
461 }
462
463 if (FLAG_IS_DEFAULT(MarkStackSize)) {
464 size_t mark_stack_size =
465 MIN2(MarkStackSizeMax,
466 MAX2(MarkStackSize, (size_t) (parallel_marking_threads() * TASKQUEUE_SIZE)));
467 // Verify that the calculated value for MarkStackSize is in range.
468 // It would be nice to use the private utility routine from Arguments.
469 if (!(mark_stack_size >= 1 && mark_stack_size <= MarkStackSizeMax)) {
470 log_warning(gc)("Invalid value calculated for MarkStackSize (" SIZE_FORMAT "): "
471 "must be between 1 and " SIZE_FORMAT,
472 mark_stack_size, MarkStackSizeMax);
473 return;
474 }
475 FLAG_SET_ERGO(size_t, MarkStackSize, mark_stack_size);
476 } else {
477 // Verify MarkStackSize is in range.
478 if (FLAG_IS_CMDLINE(MarkStackSize)) {
479 if (FLAG_IS_DEFAULT(MarkStackSizeMax)) {
480 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
481 log_warning(gc)("Invalid value specified for MarkStackSize (" SIZE_FORMAT "): "
482 "must be between 1 and " SIZE_FORMAT,
483 MarkStackSize, MarkStackSizeMax);
484 return;
485 }
486 } else if (FLAG_IS_CMDLINE(MarkStackSizeMax)) {
487 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
488 log_warning(gc)("Invalid value specified for MarkStackSize (" SIZE_FORMAT ")"
489 " or for MarkStackSizeMax (" SIZE_FORMAT ")",
490 MarkStackSize, MarkStackSizeMax);
491 return;
492 }
493 }
494 }
495 }
496
497 if (!_markStack.allocate(MarkStackSize)) {
498 log_warning(gc)("Failed to allocate CM marking stack");
499 return;
500 }
501
502 allocate_internal_bitmaps();
503
504 if (G1PretouchAuxiliaryMemory) {
505 pretouch_internal_bitmaps();
506 }
507
508 _tasks = NEW_C_HEAP_ARRAY(G1CMTask*, _max_worker_id, mtGC);
509 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_worker_id, mtGC);
510
511 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail
512 _active_tasks = _max_worker_id;
513
514 for (uint i = 0; i < _max_worker_id; ++i) {
515 G1CMTaskQueue* task_queue = new G1CMTaskQueue();
516 task_queue->initialize();
517 _task_queues->register_queue(i, task_queue);
518
519 _tasks[i] = new G1CMTask(i, this, task_queue, _task_queues);
520
521 _accum_task_vtime[i] = 0.0;
522 }
523
524 // Calculate the card number for the bottom of the heap. Used
525 // in biasing indexes into the accounting card bitmaps.
526 _heap_bottom_card_num =
527 intptr_t(uintptr_t(_g1h->reserved_region().start()) >>
528 CardTableModRefBS::card_shift);
529
530 // so that the call below can read a sensible value
531 _heap_start = g1h->reserved_region().start();
532 set_non_marking_state();
533 _completed_initialization = true;
534 }
535
536 void G1ConcurrentMark::reset() {
537 // Starting values for these two. This should be called in a STW
538 // phase.
539 MemRegion reserved = _g1h->g1_reserved();
540 _heap_start = reserved.start();
541 _heap_end = reserved.end();
542
543 // Separated the asserts so that we know which one fires.
544 assert(_heap_start != NULL, "heap bounds should look ok");
545 assert(_heap_end != NULL, "heap bounds should look ok");
546 assert(_heap_start < _heap_end, "heap bounds should look ok");
547
548 // Reset all the marking data structures and any necessary flags
549 reset_marking_state();
550
551 // We do reset all of them, since different phases will use
552 // different number of active threads. So, it's easiest to have all
553 // of them ready.
554 for (uint i = 0; i < _max_worker_id; ++i) {
555 _tasks[i]->reset(_nextMarkBitMap);
556 }
557
558 // we need this to make sure that the flag is on during the evac
559 // pause with initial mark piggy-backed
560 set_concurrent_marking_in_progress();
561 }
562
563
564 void G1ConcurrentMark::reset_marking_state(bool clear_overflow) {
565 _markStack.set_should_expand();
566 _markStack.setEmpty(); // Also clears the _markStack overflow flag
567 if (clear_overflow) {
568 clear_has_overflown();
569 } else {
570 assert(has_overflown(), "pre-condition");
571 }
572 _finger = _heap_start;
573
574 for (uint i = 0; i < _max_worker_id; ++i) {
575 G1CMTaskQueue* queue = _task_queues->queue(i);
576 queue->set_empty();
577 }
578 }
579
580 void G1ConcurrentMark::set_concurrency(uint active_tasks) {
581 assert(active_tasks <= _max_worker_id, "we should not have more");
582
583 _active_tasks = active_tasks;
584 // Need to update the three data structures below according to the
585 // number of active threads for this phase.
586 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues);
587 _first_overflow_barrier_sync.set_n_workers((int) active_tasks);
588 _second_overflow_barrier_sync.set_n_workers((int) active_tasks);
589 }
590
591 void G1ConcurrentMark::set_concurrency_and_phase(uint active_tasks, bool concurrent) {
592 set_concurrency(active_tasks);
593
594 _concurrent = concurrent;
595 // We propagate this to all tasks, not just the active ones.
596 for (uint i = 0; i < _max_worker_id; ++i)
597 _tasks[i]->set_concurrent(concurrent);
598
599 if (concurrent) {
600 set_concurrent_marking_in_progress();
601 } else {
602 // We currently assume that the concurrent flag has been set to
603 // false before we start remark. At this point we should also be
604 // in a STW phase.
605 assert(!concurrent_marking_in_progress(), "invariant");
606 assert(out_of_regions(),
607 "only way to get here: _finger: " PTR_FORMAT ", _heap_end: " PTR_FORMAT,
608 p2i(_finger), p2i(_heap_end));
609 }
610 }
611
612 void G1ConcurrentMark::set_non_marking_state() {
613 // We set the global marking state to some default values when we're
614 // not doing marking.
615 reset_marking_state();
616 _active_tasks = 0;
617 clear_concurrent_marking_in_progress();
618 }
619
620 G1ConcurrentMark::~G1ConcurrentMark() {
621 // The G1ConcurrentMark instance is never freed.
622 ShouldNotReachHere();
623 }
624
625 class G1ClearBitMapTask : public AbstractGangTask {
626 // Heap region closure used for clearing the given mark bitmap.
627 class G1ClearBitmapHRClosure : public HeapRegionClosure {
628 private:
629 G1CMBitMap* _bitmap;
630 G1ConcurrentMark* _cm;
631 public:
632 G1ClearBitmapHRClosure(G1CMBitMap* bitmap, G1ConcurrentMark* cm) : HeapRegionClosure(), _cm(cm), _bitmap(bitmap) {
633 }
634
635 virtual bool doHeapRegion(HeapRegion* r) {
636 size_t const chunk_size_in_words = M / HeapWordSize;
637
638 HeapWord* cur = r->bottom();
639 HeapWord* const end = r->end();
640
641 while (cur < end) {
642 MemRegion mr(cur, MIN2(cur + chunk_size_in_words, end));
643 _bitmap->clear_range(mr);
644
645 cur += chunk_size_in_words;
646
647 // Abort iteration if after yielding the marking has been aborted.
648 if (_cm != NULL && _cm->do_yield_check() && _cm->has_aborted()) {
649 return true;
650 }
651 // Repeat the asserts from before the start of the closure. We will do them
652 // as asserts here to minimize their overhead on the product. However, we
653 // will have them as guarantees at the beginning / end of the bitmap
654 // clearing to get some checking in the product.
655 assert(_cm == NULL || _cm->cmThread()->during_cycle(), "invariant");
656 assert(_cm == NULL || !G1CollectedHeap::heap()->collector_state()->mark_in_progress(), "invariant");
657 }
658 assert(cur == end, "Must have completed iteration over the bitmap for region %u.", r->hrm_index());
659
660 return false;
661 }
662 };
663
664 G1ClearBitmapHRClosure _cl;
665 HeapRegionClaimer _hr_claimer;
666 bool _suspendible; // If the task is suspendible, workers must join the STS.
667
668 public:
669 G1ClearBitMapTask(G1CMBitMap* bitmap, G1ConcurrentMark* cm, uint n_workers, bool suspendible) :
670 AbstractGangTask("Parallel Clear Bitmap Task"),
671 _cl(bitmap, suspendible ? cm : NULL),
672 _hr_claimer(n_workers),
673 _suspendible(suspendible)
674 { }
675
676 void work(uint worker_id) {
677 SuspendibleThreadSetJoiner sts_join(_suspendible);
678 G1CollectedHeap::heap()->heap_region_par_iterate(&_cl, worker_id, &_hr_claimer, true);
679 }
680
681 bool is_complete() {
682 return _cl.complete();
683 }
684 };
685
686 void G1ConcurrentMark::clear_bitmap(G1CMBitMap* bitmap, WorkGang* workers, bool may_yield) {
687 assert(may_yield || SafepointSynchronize::is_at_safepoint(), "Non-yielding bitmap clear only allowed at safepoint.");
688
689 G1ClearBitMapTask task(bitmap, this, workers->active_workers(), may_yield);
690 workers->run_task(&task);
691 guarantee(!may_yield || task.is_complete(), "Must have completed iteration when not yielding.");
692 }
693
694 void G1ConcurrentMark::cleanup_for_next_mark() {
695 // Make sure that the concurrent mark thread looks to still be in
696 // the current cycle.
697 guarantee(cmThread()->during_cycle(), "invariant");
698
699 // We are finishing up the current cycle by clearing the next
700 // marking bitmap and getting it ready for the next cycle. During
701 // this time no other cycle can start. So, let's make sure that this
702 // is the case.
703 guarantee(!_g1h->collector_state()->mark_in_progress(), "invariant");
704
705 clear_bitmap(_nextMarkBitMap, _parallel_workers, true);
706
707 // Clear the live count data. If the marking has been aborted, the abort()
708 // call already did that.
709 if (!has_aborted()) {
710 clear_all_live_data(_parallel_workers);
711 DEBUG_ONLY(verify_all_live_data());
712 }
713
714 // Repeat the asserts from above.
715 guarantee(cmThread()->during_cycle(), "invariant");
716 guarantee(!_g1h->collector_state()->mark_in_progress(), "invariant");
717 }
718
719 void G1ConcurrentMark::clear_prev_bitmap(WorkGang* workers) {
720 assert(SafepointSynchronize::is_at_safepoint(), "Should only clear the entire prev bitmap at a safepoint.");
721 clear_bitmap((G1CMBitMap*)_prevMarkBitMap, workers, false);
722 }
723
724 class CheckBitmapClearHRClosure : public HeapRegionClosure {
725 G1CMBitMap* _bitmap;
726 bool _error;
727 public:
728 CheckBitmapClearHRClosure(G1CMBitMap* bitmap) : _bitmap(bitmap) {
729 }
730
731 virtual bool doHeapRegion(HeapRegion* r) {
732 // This closure can be called concurrently to the mutator, so we must make sure
733 // that the result of the getNextMarkedWordAddress() call is compared to the
734 // value passed to it as limit to detect any found bits.
735 // end never changes in G1.
736 HeapWord* end = r->end();
737 return _bitmap->getNextMarkedWordAddress(r->bottom(), end) != end;
738 }
739 };
740
741 bool G1ConcurrentMark::nextMarkBitmapIsClear() {
742 CheckBitmapClearHRClosure cl(_nextMarkBitMap);
743 _g1h->heap_region_iterate(&cl);
744 return cl.complete();
745 }
746
747 class NoteStartOfMarkHRClosure: public HeapRegionClosure {
748 public:
749 bool doHeapRegion(HeapRegion* r) {
750 r->note_start_of_marking();
751 return false;
752 }
753 };
754
755 void G1ConcurrentMark::checkpointRootsInitialPre() {
756 G1CollectedHeap* g1h = G1CollectedHeap::heap();
757 G1CollectorPolicy* g1p = g1h->g1_policy();
758
759 _has_aborted = false;
760
761 // Initialize marking structures. This has to be done in a STW phase.
762 reset();
763
764 // For each region note start of marking.
765 NoteStartOfMarkHRClosure startcl;
766 g1h->heap_region_iterate(&startcl);
767 }
768
769
770 void G1ConcurrentMark::checkpointRootsInitialPost() {
771 G1CollectedHeap* g1h = G1CollectedHeap::heap();
772
773 // Start Concurrent Marking weak-reference discovery.
774 ReferenceProcessor* rp = g1h->ref_processor_cm();
775 // enable ("weak") refs discovery
776 rp->enable_discovery();
777 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle
778
779 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
780 // This is the start of the marking cycle, we're expected all
781 // threads to have SATB queues with active set to false.
782 satb_mq_set.set_active_all_threads(true, /* new active value */
783 false /* expected_active */);
784
785 _root_regions.prepare_for_scan();
786
787 // update_g1_committed() will be called at the end of an evac pause
788 // when marking is on. So, it's also called at the end of the
789 // initial-mark pause to update the heap end, if the heap expands
790 // during it. No need to call it here.
791 }
792
793 /*
794 * Notice that in the next two methods, we actually leave the STS
795 * during the barrier sync and join it immediately afterwards. If we
796 * do not do this, the following deadlock can occur: one thread could
797 * be in the barrier sync code, waiting for the other thread to also
798 * sync up, whereas another one could be trying to yield, while also
799 * waiting for the other threads to sync up too.
800 *
801 * Note, however, that this code is also used during remark and in
802 * this case we should not attempt to leave / enter the STS, otherwise
803 * we'll either hit an assert (debug / fastdebug) or deadlock
804 * (product). So we should only leave / enter the STS if we are
805 * operating concurrently.
806 *
807 * Because the thread that does the sync barrier has left the STS, it
808 * is possible to be suspended for a Full GC or an evacuation pause
809 * could occur. This is actually safe, since the entering the sync
810 * barrier is one of the last things do_marking_step() does, and it
811 * doesn't manipulate any data structures afterwards.
812 */
813
814 void G1ConcurrentMark::enter_first_sync_barrier(uint worker_id) {
815 bool barrier_aborted;
816 {
817 SuspendibleThreadSetLeaver sts_leave(concurrent());
818 barrier_aborted = !_first_overflow_barrier_sync.enter();
819 }
820
821 // at this point everyone should have synced up and not be doing any
822 // more work
823
824 if (barrier_aborted) {
825 // If the barrier aborted we ignore the overflow condition and
826 // just abort the whole marking phase as quickly as possible.
827 return;
828 }
829
830 // If we're executing the concurrent phase of marking, reset the marking
831 // state; otherwise the marking state is reset after reference processing,
832 // during the remark pause.
833 // If we reset here as a result of an overflow during the remark we will
834 // see assertion failures from any subsequent set_concurrency_and_phase()
835 // calls.
836 if (concurrent()) {
837 // let the task associated with with worker 0 do this
838 if (worker_id == 0) {
839 // task 0 is responsible for clearing the global data structures
840 // We should be here because of an overflow. During STW we should
841 // not clear the overflow flag since we rely on it being true when
842 // we exit this method to abort the pause and restart concurrent
843 // marking.
844 reset_marking_state(true /* clear_overflow */);
845
846 log_info(gc, marking)("Concurrent Mark reset for overflow");
847 }
848 }
849
850 // after this, each task should reset its own data structures then
851 // then go into the second barrier
852 }
853
854 void G1ConcurrentMark::enter_second_sync_barrier(uint worker_id) {
855 SuspendibleThreadSetLeaver sts_leave(concurrent());
856 _second_overflow_barrier_sync.enter();
857
858 // at this point everything should be re-initialized and ready to go
859 }
860
861 class G1CMConcurrentMarkingTask: public AbstractGangTask {
862 private:
863 G1ConcurrentMark* _cm;
864 ConcurrentMarkThread* _cmt;
865
866 public:
867 void work(uint worker_id) {
868 assert(Thread::current()->is_ConcurrentGC_thread(),
869 "this should only be done by a conc GC thread");
870 ResourceMark rm;
871
872 double start_vtime = os::elapsedVTime();
873
874 {
875 SuspendibleThreadSetJoiner sts_join;
876
877 assert(worker_id < _cm->active_tasks(), "invariant");
878 G1CMTask* the_task = _cm->task(worker_id);
879 the_task->record_start_time();
880 if (!_cm->has_aborted()) {
881 do {
882 double start_vtime_sec = os::elapsedVTime();
883 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
884
885 the_task->do_marking_step(mark_step_duration_ms,
886 true /* do_termination */,
887 false /* is_serial*/);
888
889 double end_vtime_sec = os::elapsedVTime();
890 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec;
891 _cm->clear_has_overflown();
892
893 _cm->do_yield_check(worker_id);
894
895 jlong sleep_time_ms;
896 if (!_cm->has_aborted() && the_task->has_aborted()) {
897 sleep_time_ms =
898 (jlong) (elapsed_vtime_sec * _cm->sleep_factor() * 1000.0);
899 {
900 SuspendibleThreadSetLeaver sts_leave;
901 os::sleep(Thread::current(), sleep_time_ms, false);
902 }
903 }
904 } while (!_cm->has_aborted() && the_task->has_aborted());
905 }
906 the_task->record_end_time();
907 guarantee(!the_task->has_aborted() || _cm->has_aborted(), "invariant");
908 }
909
910 double end_vtime = os::elapsedVTime();
911 _cm->update_accum_task_vtime(worker_id, end_vtime - start_vtime);
912 }
913
914 G1CMConcurrentMarkingTask(G1ConcurrentMark* cm,
915 ConcurrentMarkThread* cmt) :
916 AbstractGangTask("Concurrent Mark"), _cm(cm), _cmt(cmt) { }
917
918 ~G1CMConcurrentMarkingTask() { }
919 };
920
921 // Calculates the number of active workers for a concurrent
922 // phase.
923 uint G1ConcurrentMark::calc_parallel_marking_threads() {
924 uint n_conc_workers = 0;
925 if (!UseDynamicNumberOfGCThreads ||
926 (!FLAG_IS_DEFAULT(ConcGCThreads) &&
927 !ForceDynamicNumberOfGCThreads)) {
928 n_conc_workers = max_parallel_marking_threads();
929 } else {
930 n_conc_workers =
931 AdaptiveSizePolicy::calc_default_active_workers(
932 max_parallel_marking_threads(),
933 1, /* Minimum workers */
934 parallel_marking_threads(),
935 Threads::number_of_non_daemon_threads());
936 // Don't scale down "n_conc_workers" by scale_parallel_threads() because
937 // that scaling has already gone into "_max_parallel_marking_threads".
938 }
939 assert(n_conc_workers > 0, "Always need at least 1");
940 return n_conc_workers;
941 }
942
943 void G1ConcurrentMark::scanRootRegion(HeapRegion* hr, uint worker_id) {
944 // Currently, only survivors can be root regions.
945 assert(hr->next_top_at_mark_start() == hr->bottom(), "invariant");
946 G1RootRegionScanClosure cl(_g1h, this, worker_id);
947
948 const uintx interval = PrefetchScanIntervalInBytes;
949 HeapWord* curr = hr->bottom();
950 const HeapWord* end = hr->top();
951 while (curr < end) {
952 Prefetch::read(curr, interval);
953 oop obj = oop(curr);
954 int size = obj->oop_iterate_size(&cl);
955 assert(size == obj->size(), "sanity");
956 curr += size;
957 }
958 }
959
960 class G1CMRootRegionScanTask : public AbstractGangTask {
961 private:
962 G1ConcurrentMark* _cm;
963
964 public:
965 G1CMRootRegionScanTask(G1ConcurrentMark* cm) :
966 AbstractGangTask("Root Region Scan"), _cm(cm) { }
967
968 void work(uint worker_id) {
969 assert(Thread::current()->is_ConcurrentGC_thread(),
970 "this should only be done by a conc GC thread");
971
972 G1CMRootRegions* root_regions = _cm->root_regions();
973 HeapRegion* hr = root_regions->claim_next();
974 while (hr != NULL) {
975 _cm->scanRootRegion(hr, worker_id);
976 hr = root_regions->claim_next();
977 }
978 }
979 };
980
981 void G1ConcurrentMark::scan_root_regions() {
982 // scan_in_progress() will have been set to true only if there was
983 // at least one root region to scan. So, if it's false, we
984 // should not attempt to do any further work.
985 if (root_regions()->scan_in_progress()) {
986 assert(!has_aborted(), "Aborting before root region scanning is finished not supported.");
987
988 _parallel_marking_threads = calc_parallel_marking_threads();
989 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
990 "Maximum number of marking threads exceeded");
991 uint active_workers = MAX2(1U, parallel_marking_threads());
992
993 G1CMRootRegionScanTask task(this);
994 _parallel_workers->set_active_workers(active_workers);
995 _parallel_workers->run_task(&task);
996
997 // It's possible that has_aborted() is true here without actually
998 // aborting the survivor scan earlier. This is OK as it's
999 // mainly used for sanity checking.
1000 root_regions()->scan_finished();
1001 }
1002 }
1003
1004 void G1ConcurrentMark::concurrent_cycle_start() {
1005 _gc_timer_cm->register_gc_start();
1006
1007 _gc_tracer_cm->report_gc_start(GCCause::_no_gc /* first parameter is not used */, _gc_timer_cm->gc_start());
1008
1009 _g1h->trace_heap_before_gc(_gc_tracer_cm);
1010 }
1011
1012 void G1ConcurrentMark::concurrent_cycle_end() {
1013 _g1h->trace_heap_after_gc(_gc_tracer_cm);
1014
1015 if (has_aborted()) {
1016 _gc_tracer_cm->report_concurrent_mode_failure();
1017 }
1018
1019 _gc_timer_cm->register_gc_end();
1020
1021 _gc_tracer_cm->report_gc_end(_gc_timer_cm->gc_end(), _gc_timer_cm->time_partitions());
1022 }
1023
1024 void G1ConcurrentMark::mark_from_roots() {
1025 // we might be tempted to assert that:
1026 // assert(asynch == !SafepointSynchronize::is_at_safepoint(),
1027 // "inconsistent argument?");
1028 // However that wouldn't be right, because it's possible that
1029 // a safepoint is indeed in progress as a younger generation
1030 // stop-the-world GC happens even as we mark in this generation.
1031
1032 _restart_for_overflow = false;
1033
1034 // _g1h has _n_par_threads
1035 _parallel_marking_threads = calc_parallel_marking_threads();
1036 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
1037 "Maximum number of marking threads exceeded");
1038
1039 uint active_workers = MAX2(1U, parallel_marking_threads());
1040 assert(active_workers > 0, "Should have been set");
1041
1042 // Parallel task terminator is set in "set_concurrency_and_phase()"
1043 set_concurrency_and_phase(active_workers, true /* concurrent */);
1044
1045 G1CMConcurrentMarkingTask markingTask(this, cmThread());
1046 _parallel_workers->set_active_workers(active_workers);
1047 _parallel_workers->run_task(&markingTask);
1048 print_stats();
1049 }
1050
1051 void G1ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) {
1052 // world is stopped at this checkpoint
1053 assert(SafepointSynchronize::is_at_safepoint(),
1054 "world should be stopped");
1055
1056 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1057
1058 // If a full collection has happened, we shouldn't do this.
1059 if (has_aborted()) {
1060 g1h->collector_state()->set_mark_in_progress(false); // So bitmap clearing isn't confused
1061 return;
1062 }
1063
1064 SvcGCMarker sgcm(SvcGCMarker::OTHER);
1065
1066 if (VerifyDuringGC) {
1067 HandleMark hm; // handle scope
1068 g1h->prepare_for_verify();
1069 Universe::verify(VerifyOption_G1UsePrevMarking, "During GC (before)");
1070 }
1071 g1h->verifier()->check_bitmaps("Remark Start");
1072
1073 G1CollectorPolicy* g1p = g1h->g1_policy();
1074 g1p->record_concurrent_mark_remark_start();
1075
1076 double start = os::elapsedTime();
1077
1078 checkpointRootsFinalWork();
1079
1080 double mark_work_end = os::elapsedTime();
1081
1082 weakRefsWork(clear_all_soft_refs);
1083
1084 if (has_overflown()) {
1085 // Oops. We overflowed. Restart concurrent marking.
1086 _restart_for_overflow = true;
1087
1088 // Verify the heap w.r.t. the previous marking bitmap.
1089 if (VerifyDuringGC) {
1090 HandleMark hm; // handle scope
1091 g1h->prepare_for_verify();
1092 Universe::verify(VerifyOption_G1UsePrevMarking, "During GC (overflow)");
1093 }
1094
1095 // Clear the marking state because we will be restarting
1096 // marking due to overflowing the global mark stack.
1097 reset_marking_state();
1098 } else {
1099 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1100 // We're done with marking.
1101 // This is the end of the marking cycle, we're expected all
1102 // threads to have SATB queues with active set to true.
1103 satb_mq_set.set_active_all_threads(false, /* new active value */
1104 true /* expected_active */);
1105
1106 if (VerifyDuringGC) {
1107 HandleMark hm; // handle scope
1108 g1h->prepare_for_verify();
1109 Universe::verify(VerifyOption_G1UseNextMarking, "During GC (after)");
1110 }
1111 g1h->verifier()->check_bitmaps("Remark End");
1112 assert(!restart_for_overflow(), "sanity");
1113 // Completely reset the marking state since marking completed
1114 set_non_marking_state();
1115 }
1116
1117 // Expand the marking stack, if we have to and if we can.
1118 if (_markStack.should_expand()) {
1119 _markStack.expand();
1120 }
1121
1122 // Statistics
1123 double now = os::elapsedTime();
1124 _remark_mark_times.add((mark_work_end - start) * 1000.0);
1125 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0);
1126 _remark_times.add((now - start) * 1000.0);
1127
1128 g1p->record_concurrent_mark_remark_end();
1129
1130 G1CMIsAliveClosure is_alive(g1h);
1131 _gc_tracer_cm->report_object_count_after_gc(&is_alive);
1132 }
1133
1134 // Base class of the closures that finalize and verify the
1135 // liveness count data.
1136 class G1LiveDataClosureBase: public HeapRegionClosure {
1137 protected:
1138 G1ConcurrentMark* _cm;
1139
1140 BitMap* _region_bm;
1141 BitMap* _card_bm;
1142
1143 // Takes a region that's not empty (i.e., it has at least one
1144 // live object in it and sets its corresponding bit on the region
1145 // bitmap to 1.
1146 void set_bit_for_region(HeapRegion* hr) {
1147 BitMap::idx_t index = (BitMap::idx_t) hr->hrm_index();
1148 _region_bm->par_at_put(index, true);
1149 }
1150
1151 // Utility routine to set an exclusive range of cards on the given
1152 // bitmap.
1153 inline void set_card_bitmap_range(BitMap* card_bm,
1154 BitMap::idx_t start_idx,
1155 BitMap::idx_t end_idx) {
1156
1157 // Set the exclusive bit range [start_idx, end_idx).
1158 assert((end_idx - start_idx) > 0, "at least one card");
1159 assert(end_idx <= card_bm->size(), "sanity");
1160
1161 // For small ranges use a simple loop; otherwise use set_range or
1162 // use par_at_put_range (if parallel). The range is made up of the
1163 // cards that are spanned by an object/mem region so 8 cards will
1164 // allow up to object sizes up to 4K to be handled using the loop.
1165 if ((end_idx - start_idx) <= 8) {
1166 for (BitMap::idx_t i = start_idx; i < end_idx; i += 1) {
1167 card_bm->set_bit(i);
1168 }
1169 } else {
1170 card_bm->set_range(start_idx, end_idx);
1171 }
1172 }
1173
1174 void mark_card_bitmap_range(HeapWord* start, HeapWord* end) {
1175 BitMap::idx_t start_idx = _cm->card_live_bitmap_index_for(start);
1176 BitMap::idx_t end_idx = _cm->card_live_bitmap_index_for((HeapWord*)align_ptr_up(end, CardTableModRefBS::card_size));
1177
1178 assert((end_idx - start_idx) > 0, "Trying to mark zero sized range.");
1179
1180 if (start_idx == _last_marked_bit_idx) {
1181 start_idx++;
1182 }
1183 if (start_idx == end_idx) {
1184 return;
1185 }
1186
1187 // Set the bits in the card bitmap for the cards spanned by this object.
1188 set_card_bitmap_range(_card_bm, start_idx, end_idx);
1189 _last_marked_bit_idx = end_idx - 1;
1190 }
1191
1192 // We cache the last mark set. This avoids setting the same bit multiple times.
1193 // This is particularly interesting for dense bitmaps, as this avoids doing
1194 // any work most of the time.
1195 BitMap::idx_t _last_marked_bit_idx;
1196
1197 void reset_mark_cache() {
1198 _last_marked_bit_idx = (BitMap::idx_t)-1;
1199 }
1200
1201 void mark_allocated_since_marking(HeapRegion* hr) {
1202 reset_mark_cache();
1203
1204 HeapWord* ntams = hr->next_top_at_mark_start();
1205 HeapWord* top = hr->top();
1206
1207 assert(hr->bottom() <= ntams && ntams <= hr->end(), "Preconditions.");
1208
1209 // Mark the allocated-since-marking portion...
1210 if (ntams < top) {
1211 mark_card_bitmap_range(ntams, top);
1212 // This definitely means the region has live objects.
1213 set_bit_for_region(hr);
1214 }
1215 }
1216
1217 bool mark_marked_during_marking(HeapRegion* hr, bool may_suspend, size_t* total_bytes_marked) {
1218 reset_mark_cache();
1219
1220 size_t marked_bytes = 0;
1221
1222 HeapWord* ntams = hr->next_top_at_mark_start();
1223 HeapWord* start = hr->bottom();
1224
1225 if (ntams <= start) {
1226 // Empty region (at the start of marking). Nothing to do.
1227 // hr->add_to_marked_bytes(0);
1228 *total_bytes_marked = marked_bytes;
1229 return false;
1230 } else if (hr->is_starts_humongous()) {
1231 // Humongous object: distribute the marked bytes across the humongous object.
1232 do {
1233 mark_card_bitmap_range(start, hr->top());
1234
1235 marked_bytes += pointer_delta(hr->top(), start, 1);
1236 hr->add_to_marked_bytes(marked_bytes);
1237
1238 hr = G1CollectedHeap::heap()->next_region_in_humongous(hr);
1239 } while (hr != NULL);
1240 *total_bytes_marked = marked_bytes;
1241 return false;
1242 } else if (hr->is_continues_humongous()) {
1243 // Humongous continues regions were handled during processing of the start region.
1244 *total_bytes_marked = marked_bytes;
1245 return false;
1246 }
1247
1248 assert(start <= hr->end() && start <= ntams && ntams <= hr->end(),
1249 "Preconditions not met - "
1250 "start: " PTR_FORMAT ", ntams: " PTR_FORMAT ", end: " PTR_FORMAT,
1251 p2i(start), p2i(ntams), p2i(hr->end()));
1252
1253 G1CMBitMap* bitmap = _cm->nextMarkBitMap();
1254 // Find the first marked object at or after "start".
1255 start = bitmap->getNextMarkedWordAddress(start, ntams);
1256 while (start < ntams) {
1257 oop obj = oop(start);
1258 int obj_sz = obj->size();
1259 HeapWord* obj_end = start + obj_sz;
1260
1261 assert(obj_end <= hr->end(), "Humongous objects must have been handled elsewhere.");
1262
1263 mark_card_bitmap_range(start, obj_end);
1264
1265 // Add the size of this object to the number of marked bytes.
1266 marked_bytes += (size_t)obj_sz * HeapWordSize;
1267
1268 // Find the next marked object after this one.
1269 start = bitmap->getNextMarkedWordAddress(obj_end, ntams);
1270 }
1271
1272 // Update the marked bytes for this region.
1273 hr->add_to_marked_bytes(marked_bytes);
1274 *total_bytes_marked = marked_bytes;
1275
1276 // Abort iteration if after yielding the marking has been aborted.
1277 if (may_suspend && _cm->do_yield_check() && _cm->has_aborted()) {
1278 return true;
1279 }
1280 // Next heap region
1281 return false;
1282 }
1283
1284 public:
1285 G1LiveDataClosureBase(G1CollectedHeap* g1h,
1286 BitMap* region_bm,
1287 BitMap* card_bm):
1288 _cm(g1h->concurrent_mark()),
1289 _region_bm(region_bm),
1290 _card_bm(card_bm) { }
1291 };
1292
1293 // Heap region closure used for verifying the live count data
1294 // that was created concurrently and finalized during
1295 // the remark pause. This closure is applied to the heap
1296 // regions during the STW cleanup pause.
1297 class G1VerifyLiveDataHRClosure: public HeapRegionClosure {
1298 // Calculates the # live objects per region.
1299 class G1VerifyLiveDataClosure: public G1LiveDataClosureBase {
1300 size_t _region_marked_bytes;
1301
1302 public:
1303 G1VerifyLiveDataClosure(G1CollectedHeap* g1h,
1304 BitMap* region_bm,
1305 BitMap* card_bm) :
1306 G1LiveDataClosureBase(g1h, region_bm, card_bm),
1307 _region_marked_bytes(0) { }
1308
1309 bool doHeapRegion(HeapRegion* hr) {
1310 mark_marked_during_marking(hr, false, &_region_marked_bytes);
1311 mark_allocated_since_marking(hr);
1312 return false;
1313 }
1314
1315 size_t region_marked_bytes() const { return _region_marked_bytes; }
1316 };
1317
1318 G1CollectedHeap* _g1h;
1319 G1ConcurrentMark* _cm;
1320 G1VerifyLiveDataClosure _calc_cl;
1321 BitMap* _act_region_bm; // Region BM to be verified
1322 BitMap* _act_card_bm; // Card BM to be verified
1323
1324 BitMap* _exp_region_bm; // Expected Region BM values
1325 BitMap* _exp_card_bm; // Expected card BM values
1326
1327 int _failures;
1328
1329 public:
1330 G1VerifyLiveDataHRClosure(G1CollectedHeap* g1h,
1331 BitMap* act_region_bm,
1332 BitMap* act_card_bm,
1333 BitMap* exp_region_bm,
1334 BitMap* exp_card_bm) :
1335 _g1h(g1h),
1336 _cm(g1h->concurrent_mark()),
1337 _calc_cl(g1h, exp_region_bm, exp_card_bm),
1338 _act_region_bm(act_region_bm),
1339 _act_card_bm(act_card_bm),
1340 _exp_region_bm(exp_region_bm),
1341 _exp_card_bm(exp_card_bm),
1342 _failures(0) { }
1343
1344 int failures() const { return _failures; }
1345
1346 bool doHeapRegion(HeapRegion* hr) {
1347 int failures = 0;
1348
1349 // Call the G1VerifyLiveDataClosure to walk the marking bitmap for
1350 // this region and set the corresponding bits in the expected region
1351 // and card bitmaps.
1352 bool res = _calc_cl.doHeapRegion(hr);
1353 assert(!res, "Should be completed.");
1354
1355 // Verify the marked bytes for this region.
1356 size_t exp_marked_bytes = _calc_cl.region_marked_bytes();
1357 size_t act_marked_bytes = hr->next_marked_bytes();
1358
1359 if (exp_marked_bytes > act_marked_bytes) {
1360 if (hr->is_starts_humongous()) {
1361 // For start_humongous regions, the size of the whole object will be
1362 // in exp_marked_bytes.
1363 HeapRegion* region = hr;
1364 int num_regions;
1365 for (num_regions = 0; region != NULL; num_regions++) {
1366 region = _g1h->next_region_in_humongous(region);
1367 }
1368 if ((num_regions-1) * HeapRegion::GrainBytes >= exp_marked_bytes) {
1369 failures += 1;
1370 } else if (num_regions * HeapRegion::GrainBytes < exp_marked_bytes) {
1371 failures += 1;
1372 }
1373 } else {
1374 // We're not OK if expected marked bytes > actual marked bytes. It means
1375 // we have missed accounting some objects during the actual marking.
1376 failures += 1;
1377 }
1378 }
1379
1380 // Verify the bit, for this region, in the actual and expected
1381 // (which was just calculated) region bit maps.
1382 // We're not OK if the bit in the calculated expected region
1383 // bitmap is set and the bit in the actual region bitmap is not.
1384 BitMap::idx_t index = (BitMap::idx_t) hr->hrm_index();
1385
1386 bool expected = _exp_region_bm->at(index);
1387 bool actual = _act_region_bm->at(index);
1388 if (expected && !actual) {
1389 failures += 1;
1390 }
1391
1392 // Verify that the card bit maps for the cards spanned by the current
1393 // region match. We have an error if we have a set bit in the expected
1394 // bit map and the corresponding bit in the actual bitmap is not set.
1395
1396 BitMap::idx_t start_idx = _cm->card_live_bitmap_index_for(hr->bottom());
1397 BitMap::idx_t end_idx = _cm->card_live_bitmap_index_for(hr->top());
1398
1399 for (BitMap::idx_t i = start_idx; i < end_idx; i+=1) {
1400 expected = _exp_card_bm->at(i);
1401 actual = _act_card_bm->at(i);
1402
1403 if (expected && !actual) {
1404 failures += 1;
1405 }
1406 }
1407
1408 _failures += failures;
1409
1410 // We could stop iteration over the heap when we
1411 // find the first violating region by returning true.
1412 return false;
1413 }
1414 };
1415
1416 class G1VerifyLiveDataTask: public AbstractGangTask {
1417 protected:
1418 G1CollectedHeap* _g1h;
1419 BitMap* _actual_region_bm;
1420 BitMap* _actual_card_bm;
1421
1422 BitMap _expected_region_bm;
1423 BitMap _expected_card_bm;
1424
1425 int _failures;
1426
1427 HeapRegionClaimer _hr_claimer;
1428
1429 public:
1430 G1VerifyLiveDataTask(G1CollectedHeap* g1h,
1431 BitMap* region_bm,
1432 BitMap* card_bm,
1433 uint n_workers)
1434 : AbstractGangTask("G1 verify final counting"),
1435 _g1h(g1h),
1436 _actual_region_bm(region_bm),
1437 _actual_card_bm(card_bm),
1438 _expected_region_bm(region_bm->size(), true /* in_resource_area */),
1439 _expected_card_bm(card_bm->size(), true /* in_resource_area */),
1440 _failures(0),
1441 _hr_claimer(n_workers) {
1442 assert(VerifyDuringGC, "don't call this otherwise");
1443 }
1444
1445 void work(uint worker_id) {
1446 G1VerifyLiveDataHRClosure cl(_g1h,
1447 _actual_region_bm,
1448 _actual_card_bm,
1449 &_expected_region_bm,
1450 &_expected_card_bm);
1451 _g1h->heap_region_par_iterate(&cl, worker_id, &_hr_claimer);
1452
1453 Atomic::add(cl.failures(), &_failures);
1454 }
1455
1456 int failures() const { return _failures; }
1457 };
1458
1459 class G1FinalizeLiveDataTask: public AbstractGangTask {
1460 // Finalizes the liveness counting data.
1461 // Sets the bits corresponding to the interval [NTAMS, top]
1462 // (which contains the implicitly live objects) in the
1463 // card liveness bitmap. Also sets the bit for each region
1464 // containing live data, in the region liveness bitmap.
1465 class G1FinalizeCountDataClosure: public G1LiveDataClosureBase {
1466 public:
1467 G1FinalizeCountDataClosure(G1CollectedHeap* g1h,
1468 BitMap* region_bm,
1469 BitMap* card_bm) :
1470 G1LiveDataClosureBase(g1h, region_bm, card_bm) { }
1471
1472 bool doHeapRegion(HeapRegion* hr) {
1473 mark_allocated_since_marking(hr);
1474 // Set the bit for the region if it contains live data
1475 if (hr->next_marked_bytes() > 0) {
1476 set_bit_for_region(hr);
1477 }
1478
1479 return false;
1480 }
1481 };
1482
1483 G1CollectedHeap* _g1h;
1484 BitMap* _actual_region_bm;
1485 BitMap* _actual_card_bm;
1486
1487 HeapRegionClaimer _hr_claimer;
1488
1489 public:
1490 G1FinalizeLiveDataTask(G1CollectedHeap* g1h, BitMap* region_bm, BitMap* card_bm, uint n_workers) :
1491 AbstractGangTask("G1 final counting"),
1492 _g1h(g1h),
1493 _actual_region_bm(region_bm),
1494 _actual_card_bm(card_bm),
1495 _hr_claimer(n_workers) {
1496 }
1497
1498 void work(uint worker_id) {
1499 G1FinalizeCountDataClosure cl(_g1h,
1500 _actual_region_bm,
1501 _actual_card_bm);
1502
1503 _g1h->heap_region_par_iterate(&cl, worker_id, &_hr_claimer);
1504 }
1505 };
1506
1507 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure {
1508 G1CollectedHeap* _g1;
1509 size_t _freed_bytes;
1510 FreeRegionList* _local_cleanup_list;
1511 uint _old_regions_removed;
1512 uint _humongous_regions_removed;
1513 HRRSCleanupTask* _hrrs_cleanup_task;
1514
1515 public:
1516 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1,
1517 FreeRegionList* local_cleanup_list,
1518 HRRSCleanupTask* hrrs_cleanup_task) :
1519 _g1(g1),
1520 _freed_bytes(0),
1521 _local_cleanup_list(local_cleanup_list),
1522 _old_regions_removed(0),
1523 _humongous_regions_removed(0),
1524 _hrrs_cleanup_task(hrrs_cleanup_task) { }
1525
1526 size_t freed_bytes() { return _freed_bytes; }
1527 const uint old_regions_removed() { return _old_regions_removed; }
1528 const uint humongous_regions_removed() { return _humongous_regions_removed; }
1529
1530 bool doHeapRegion(HeapRegion *hr) {
1531 if (hr->is_archive()) {
1532 return false;
1533 }
1534 // We use a claim value of zero here because all regions
1535 // were claimed with value 1 in the FinalCount task.
1536 _g1->reset_gc_time_stamps(hr);
1537 hr->note_end_of_marking();
1538
1539 if (hr->used() > 0 && hr->max_live_bytes() == 0 && !hr->is_young()) {
1540 _freed_bytes += hr->used();
1541 hr->set_containing_set(NULL);
1542 if (hr->is_humongous()) {
1543 _humongous_regions_removed++;
1544 _g1->free_humongous_region(hr, _local_cleanup_list, true);
1545 } else {
1546 _old_regions_removed++;
1547 _g1->free_region(hr, _local_cleanup_list, true);
1548 }
1549 } else {
1550 hr->rem_set()->do_cleanup_work(_hrrs_cleanup_task);
1551 }
1552
1553 return false;
1554 }
1555 };
1556
1557 class G1ParNoteEndTask: public AbstractGangTask {
1558 friend class G1NoteEndOfConcMarkClosure;
1559
1560 protected:
1561 G1CollectedHeap* _g1h;
1562 FreeRegionList* _cleanup_list;
1563 HeapRegionClaimer _hrclaimer;
1564
1565 public:
1566 G1ParNoteEndTask(G1CollectedHeap* g1h, FreeRegionList* cleanup_list, uint n_workers) :
1567 AbstractGangTask("G1 note end"), _g1h(g1h), _cleanup_list(cleanup_list), _hrclaimer(n_workers) {
1568 }
1569
1570 void work(uint worker_id) {
1571 FreeRegionList local_cleanup_list("Local Cleanup List");
1572 HRRSCleanupTask hrrs_cleanup_task;
1573 G1NoteEndOfConcMarkClosure g1_note_end(_g1h, &local_cleanup_list,
1574 &hrrs_cleanup_task);
1575 _g1h->heap_region_par_iterate(&g1_note_end, worker_id, &_hrclaimer);
1576 assert(g1_note_end.complete(), "Shouldn't have yielded!");
1577
1578 // Now update the lists
1579 _g1h->remove_from_old_sets(g1_note_end.old_regions_removed(), g1_note_end.humongous_regions_removed());
1580 {
1581 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
1582 _g1h->decrement_summary_bytes(g1_note_end.freed_bytes());
1583
1584 // If we iterate over the global cleanup list at the end of
1585 // cleanup to do this printing we will not guarantee to only
1586 // generate output for the newly-reclaimed regions (the list
1587 // might not be empty at the beginning of cleanup; we might
1588 // still be working on its previous contents). So we do the
1589 // printing here, before we append the new regions to the global
1590 // cleanup list.
1591
1592 G1HRPrinter* hr_printer = _g1h->hr_printer();
1593 if (hr_printer->is_active()) {
1594 FreeRegionListIterator iter(&local_cleanup_list);
1595 while (iter.more_available()) {
1596 HeapRegion* hr = iter.get_next();
1597 hr_printer->cleanup(hr);
1598 }
1599 }
1600
1601 _cleanup_list->add_ordered(&local_cleanup_list);
1602 assert(local_cleanup_list.is_empty(), "post-condition");
1603
1604 HeapRegionRemSet::finish_cleanup_task(&hrrs_cleanup_task);
1605 }
1606 }
1607 };
1608
1609 void G1ConcurrentMark::cleanup() {
1610 // world is stopped at this checkpoint
1611 assert(SafepointSynchronize::is_at_safepoint(),
1612 "world should be stopped");
1613 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1614
1615 // If a full collection has happened, we shouldn't do this.
1616 if (has_aborted()) {
1617 g1h->collector_state()->set_mark_in_progress(false); // So bitmap clearing isn't confused
1618 return;
1619 }
1620
1621 g1h->verifier()->verify_region_sets_optional();
1622
1623 if (VerifyDuringGC) {
1624 HandleMark hm; // handle scope
1625 g1h->prepare_for_verify();
1626 Universe::verify(VerifyOption_G1UsePrevMarking, "During GC (before)");
1627 }
1628 g1h->verifier()->check_bitmaps("Cleanup Start");
1629
1630 G1CollectorPolicy* g1p = g1h->g1_policy();
1631 g1p->record_concurrent_mark_cleanup_start();
1632
1633 double start = os::elapsedTime();
1634
1635 HeapRegionRemSet::reset_for_cleanup_tasks();
1636
1637 {
1638 // Finalize the live data.
1639 G1FinalizeLiveDataTask cl(g1h,
1640 &_region_live_bm,
1641 &_card_live_bm,
1642 g1h->workers()->active_workers());
1643 g1h->workers()->run_task(&cl);
1644 }
1645
1646 if (VerifyDuringGC) {
1647 // Verify that the liveness count data created concurrently matches one created
1648 // during this safepoint.
1649 ResourceMark rm;
1650 G1VerifyLiveDataTask cl(g1h,
1651 &_region_live_bm,
1652 &_card_live_bm,
1653 g1h->workers()->active_workers());
1654 g1h->workers()->run_task(&cl);
1655
1656 guarantee(cl.failures() == 0, "Unexpected accounting failures");
1657 }
1658
1659 g1h->collector_state()->set_mark_in_progress(false);
1660
1661 double count_end = os::elapsedTime();
1662 double this_final_counting_time = (count_end - start);
1663 _total_counting_time += this_final_counting_time;
1664
1665 if (log_is_enabled(Trace, gc, liveness)) {
1666 G1PrintRegionLivenessInfoClosure cl("Post-Marking");
1667 _g1h->heap_region_iterate(&cl);
1668 }
1669
1670 // Install newly created mark bitMap as "prev".
1671 swapMarkBitMaps();
1672
1673 g1h->reset_gc_time_stamp();
1674
1675 uint n_workers = _g1h->workers()->active_workers();
1676
1677 // Note end of marking in all heap regions.
1678 G1ParNoteEndTask g1_par_note_end_task(g1h, &_cleanup_list, n_workers);
1679 g1h->workers()->run_task(&g1_par_note_end_task);
1680 g1h->check_gc_time_stamps();
1681
1682 if (!cleanup_list_is_empty()) {
1683 // The cleanup list is not empty, so we'll have to process it
1684 // concurrently. Notify anyone else that might be wanting free
1685 // regions that there will be more free regions coming soon.
1686 g1h->set_free_regions_coming();
1687 }
1688
1689 // call below, since it affects the metric by which we sort the heap
1690 // regions.
1691 if (G1ScrubRemSets) {
1692 double rs_scrub_start = os::elapsedTime();
1693 g1h->scrub_rem_set(&_region_live_bm, &_card_live_bm);
1694 _total_rs_scrub_time += (os::elapsedTime() - rs_scrub_start);
1695 }
1696
1697 // this will also free any regions totally full of garbage objects,
1698 // and sort the regions.
1699 g1h->g1_policy()->record_concurrent_mark_cleanup_end();
1700
1701 // Statistics.
1702 double end = os::elapsedTime();
1703 _cleanup_times.add((end - start) * 1000.0);
1704
1705 // Clean up will have freed any regions completely full of garbage.
1706 // Update the soft reference policy with the new heap occupancy.
1707 Universe::update_heap_info_at_gc();
1708
1709 if (VerifyDuringGC) {
1710 HandleMark hm; // handle scope
1711 g1h->prepare_for_verify();
1712 Universe::verify(VerifyOption_G1UsePrevMarking, "During GC (after)");
1713 }
1714
1715 g1h->verifier()->check_bitmaps("Cleanup End");
1716
1717 g1h->verifier()->verify_region_sets_optional();
1718
1719 // We need to make this be a "collection" so any collection pause that
1720 // races with it goes around and waits for completeCleanup to finish.
1721 g1h->increment_total_collections();
1722
1723 // Clean out dead classes and update Metaspace sizes.
1724 if (ClassUnloadingWithConcurrentMark) {
1725 ClassLoaderDataGraph::purge();
1726 }
1727 MetaspaceGC::compute_new_size();
1728
1729 // We reclaimed old regions so we should calculate the sizes to make
1730 // sure we update the old gen/space data.
1731 g1h->g1mm()->update_sizes();
1732 g1h->allocation_context_stats().update_after_mark();
1733 }
1734
1735 void G1ConcurrentMark::complete_cleanup() {
1736 if (has_aborted()) return;
1737
1738 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1739
1740 _cleanup_list.verify_optional();
1741 FreeRegionList tmp_free_list("Tmp Free List");
1742
1743 log_develop_trace(gc, freelist)("G1ConcRegionFreeing [complete cleanup] : "
1744 "cleanup list has %u entries",
1745 _cleanup_list.length());
1746
1747 // No one else should be accessing the _cleanup_list at this point,
1748 // so it is not necessary to take any locks
1749 while (!_cleanup_list.is_empty()) {
1750 HeapRegion* hr = _cleanup_list.remove_region(true /* from_head */);
1751 assert(hr != NULL, "Got NULL from a non-empty list");
1752 hr->par_clear();
1753 tmp_free_list.add_ordered(hr);
1754
1755 // Instead of adding one region at a time to the secondary_free_list,
1756 // we accumulate them in the local list and move them a few at a
1757 // time. This also cuts down on the number of notify_all() calls
1758 // we do during this process. We'll also append the local list when
1759 // _cleanup_list is empty (which means we just removed the last
1760 // region from the _cleanup_list).
1761 if ((tmp_free_list.length() % G1SecondaryFreeListAppendLength == 0) ||
1762 _cleanup_list.is_empty()) {
1763 log_develop_trace(gc, freelist)("G1ConcRegionFreeing [complete cleanup] : "
1764 "appending %u entries to the secondary_free_list, "
1765 "cleanup list still has %u entries",
1766 tmp_free_list.length(),
1767 _cleanup_list.length());
1768
1769 {
1770 MutexLockerEx x(SecondaryFreeList_lock, Mutex::_no_safepoint_check_flag);
1771 g1h->secondary_free_list_add(&tmp_free_list);
1772 SecondaryFreeList_lock->notify_all();
1773 }
1774 #ifndef PRODUCT
1775 if (G1StressConcRegionFreeing) {
1776 for (uintx i = 0; i < G1StressConcRegionFreeingDelayMillis; ++i) {
1777 os::sleep(Thread::current(), (jlong) 1, false);
1778 }
1779 }
1780 #endif
1781 }
1782 }
1783 assert(tmp_free_list.is_empty(), "post-condition");
1784 }
1785
1786 // Supporting Object and Oop closures for reference discovery
1787 // and processing in during marking
1788
1789 bool G1CMIsAliveClosure::do_object_b(oop obj) {
1790 HeapWord* addr = (HeapWord*)obj;
1791 return addr != NULL &&
1792 (!_g1->is_in_g1_reserved(addr) || !_g1->is_obj_ill(obj));
1793 }
1794
1795 // 'Keep Alive' oop closure used by both serial parallel reference processing.
1796 // Uses the G1CMTask associated with a worker thread (for serial reference
1797 // processing the G1CMTask for worker 0 is used) to preserve (mark) and
1798 // trace referent objects.
1799 //
1800 // Using the G1CMTask and embedded local queues avoids having the worker
1801 // threads operating on the global mark stack. This reduces the risk
1802 // of overflowing the stack - which we would rather avoid at this late
1803 // state. Also using the tasks' local queues removes the potential
1804 // of the workers interfering with each other that could occur if
1805 // operating on the global stack.
1806
1807 class G1CMKeepAliveAndDrainClosure: public OopClosure {
1808 G1ConcurrentMark* _cm;
1809 G1CMTask* _task;
1810 int _ref_counter_limit;
1811 int _ref_counter;
1812 bool _is_serial;
1813 public:
1814 G1CMKeepAliveAndDrainClosure(G1ConcurrentMark* cm, G1CMTask* task, bool is_serial) :
1815 _cm(cm), _task(task), _is_serial(is_serial),
1816 _ref_counter_limit(G1RefProcDrainInterval) {
1817 assert(_ref_counter_limit > 0, "sanity");
1818 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
1819 _ref_counter = _ref_counter_limit;
1820 }
1821
1822 virtual void do_oop(narrowOop* p) { do_oop_work(p); }
1823 virtual void do_oop( oop* p) { do_oop_work(p); }
1824
1825 template <class T> void do_oop_work(T* p) {
1826 if (!_cm->has_overflown()) {
1827 oop obj = oopDesc::load_decode_heap_oop(p);
1828 _task->deal_with_reference(obj);
1829 _ref_counter--;
1830
1831 if (_ref_counter == 0) {
1832 // We have dealt with _ref_counter_limit references, pushing them
1833 // and objects reachable from them on to the local stack (and
1834 // possibly the global stack). Call G1CMTask::do_marking_step() to
1835 // process these entries.
1836 //
1837 // We call G1CMTask::do_marking_step() in a loop, which we'll exit if
1838 // there's nothing more to do (i.e. we're done with the entries that
1839 // were pushed as a result of the G1CMTask::deal_with_reference() calls
1840 // above) or we overflow.
1841 //
1842 // Note: G1CMTask::do_marking_step() can set the G1CMTask::has_aborted()
1843 // flag while there may still be some work to do. (See the comment at
1844 // the beginning of G1CMTask::do_marking_step() for those conditions -
1845 // one of which is reaching the specified time target.) It is only
1846 // when G1CMTask::do_marking_step() returns without setting the
1847 // has_aborted() flag that the marking step has completed.
1848 do {
1849 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
1850 _task->do_marking_step(mark_step_duration_ms,
1851 false /* do_termination */,
1852 _is_serial);
1853 } while (_task->has_aborted() && !_cm->has_overflown());
1854 _ref_counter = _ref_counter_limit;
1855 }
1856 }
1857 }
1858 };
1859
1860 // 'Drain' oop closure used by both serial and parallel reference processing.
1861 // Uses the G1CMTask associated with a given worker thread (for serial
1862 // reference processing the G1CMtask for worker 0 is used). Calls the
1863 // do_marking_step routine, with an unbelievably large timeout value,
1864 // to drain the marking data structures of the remaining entries
1865 // added by the 'keep alive' oop closure above.
1866
1867 class G1CMDrainMarkingStackClosure: public VoidClosure {
1868 G1ConcurrentMark* _cm;
1869 G1CMTask* _task;
1870 bool _is_serial;
1871 public:
1872 G1CMDrainMarkingStackClosure(G1ConcurrentMark* cm, G1CMTask* task, bool is_serial) :
1873 _cm(cm), _task(task), _is_serial(is_serial) {
1874 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
1875 }
1876
1877 void do_void() {
1878 do {
1879 // We call G1CMTask::do_marking_step() to completely drain the local
1880 // and global marking stacks of entries pushed by the 'keep alive'
1881 // oop closure (an instance of G1CMKeepAliveAndDrainClosure above).
1882 //
1883 // G1CMTask::do_marking_step() is called in a loop, which we'll exit
1884 // if there's nothing more to do (i.e. we've completely drained the
1885 // entries that were pushed as a a result of applying the 'keep alive'
1886 // closure to the entries on the discovered ref lists) or we overflow
1887 // the global marking stack.
1888 //
1889 // Note: G1CMTask::do_marking_step() can set the G1CMTask::has_aborted()
1890 // flag while there may still be some work to do. (See the comment at
1891 // the beginning of G1CMTask::do_marking_step() for those conditions -
1892 // one of which is reaching the specified time target.) It is only
1893 // when G1CMTask::do_marking_step() returns without setting the
1894 // has_aborted() flag that the marking step has completed.
1895
1896 _task->do_marking_step(1000000000.0 /* something very large */,
1897 true /* do_termination */,
1898 _is_serial);
1899 } while (_task->has_aborted() && !_cm->has_overflown());
1900 }
1901 };
1902
1903 // Implementation of AbstractRefProcTaskExecutor for parallel
1904 // reference processing at the end of G1 concurrent marking
1905
1906 class G1CMRefProcTaskExecutor: public AbstractRefProcTaskExecutor {
1907 private:
1908 G1CollectedHeap* _g1h;
1909 G1ConcurrentMark* _cm;
1910 WorkGang* _workers;
1911 uint _active_workers;
1912
1913 public:
1914 G1CMRefProcTaskExecutor(G1CollectedHeap* g1h,
1915 G1ConcurrentMark* cm,
1916 WorkGang* workers,
1917 uint n_workers) :
1918 _g1h(g1h), _cm(cm),
1919 _workers(workers), _active_workers(n_workers) { }
1920
1921 // Executes the given task using concurrent marking worker threads.
1922 virtual void execute(ProcessTask& task);
1923 virtual void execute(EnqueueTask& task);
1924 };
1925
1926 class G1CMRefProcTaskProxy: public AbstractGangTask {
1927 typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask;
1928 ProcessTask& _proc_task;
1929 G1CollectedHeap* _g1h;
1930 G1ConcurrentMark* _cm;
1931
1932 public:
1933 G1CMRefProcTaskProxy(ProcessTask& proc_task,
1934 G1CollectedHeap* g1h,
1935 G1ConcurrentMark* cm) :
1936 AbstractGangTask("Process reference objects in parallel"),
1937 _proc_task(proc_task), _g1h(g1h), _cm(cm) {
1938 ReferenceProcessor* rp = _g1h->ref_processor_cm();
1939 assert(rp->processing_is_mt(), "shouldn't be here otherwise");
1940 }
1941
1942 virtual void work(uint worker_id) {
1943 ResourceMark rm;
1944 HandleMark hm;
1945 G1CMTask* task = _cm->task(worker_id);
1946 G1CMIsAliveClosure g1_is_alive(_g1h);
1947 G1CMKeepAliveAndDrainClosure g1_par_keep_alive(_cm, task, false /* is_serial */);
1948 G1CMDrainMarkingStackClosure g1_par_drain(_cm, task, false /* is_serial */);
1949
1950 _proc_task.work(worker_id, g1_is_alive, g1_par_keep_alive, g1_par_drain);
1951 }
1952 };
1953
1954 void G1CMRefProcTaskExecutor::execute(ProcessTask& proc_task) {
1955 assert(_workers != NULL, "Need parallel worker threads.");
1956 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
1957
1958 G1CMRefProcTaskProxy proc_task_proxy(proc_task, _g1h, _cm);
1959
1960 // We need to reset the concurrency level before each
1961 // proxy task execution, so that the termination protocol
1962 // and overflow handling in G1CMTask::do_marking_step() knows
1963 // how many workers to wait for.
1964 _cm->set_concurrency(_active_workers);
1965 _workers->run_task(&proc_task_proxy);
1966 }
1967
1968 class G1CMRefEnqueueTaskProxy: public AbstractGangTask {
1969 typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask;
1970 EnqueueTask& _enq_task;
1971
1972 public:
1973 G1CMRefEnqueueTaskProxy(EnqueueTask& enq_task) :
1974 AbstractGangTask("Enqueue reference objects in parallel"),
1975 _enq_task(enq_task) { }
1976
1977 virtual void work(uint worker_id) {
1978 _enq_task.work(worker_id);
1979 }
1980 };
1981
1982 void G1CMRefProcTaskExecutor::execute(EnqueueTask& enq_task) {
1983 assert(_workers != NULL, "Need parallel worker threads.");
1984 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
1985
1986 G1CMRefEnqueueTaskProxy enq_task_proxy(enq_task);
1987
1988 // Not strictly necessary but...
1989 //
1990 // We need to reset the concurrency level before each
1991 // proxy task execution, so that the termination protocol
1992 // and overflow handling in G1CMTask::do_marking_step() knows
1993 // how many workers to wait for.
1994 _cm->set_concurrency(_active_workers);
1995 _workers->run_task(&enq_task_proxy);
1996 }
1997
1998 void G1ConcurrentMark::weakRefsWorkParallelPart(BoolObjectClosure* is_alive, bool purged_classes) {
1999 G1CollectedHeap::heap()->parallel_cleaning(is_alive, true, true, purged_classes);
2000 }
2001
2002 void G1ConcurrentMark::weakRefsWork(bool clear_all_soft_refs) {
2003 if (has_overflown()) {
2004 // Skip processing the discovered references if we have
2005 // overflown the global marking stack. Reference objects
2006 // only get discovered once so it is OK to not
2007 // de-populate the discovered reference lists. We could have,
2008 // but the only benefit would be that, when marking restarts,
2009 // less reference objects are discovered.
2010 return;
2011 }
2012
2013 ResourceMark rm;
2014 HandleMark hm;
2015
2016 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2017
2018 // Is alive closure.
2019 G1CMIsAliveClosure g1_is_alive(g1h);
2020
2021 // Inner scope to exclude the cleaning of the string and symbol
2022 // tables from the displayed time.
2023 {
2024 GCTraceTime(Debug, gc) trace("Reference Processing", _gc_timer_cm);
2025
2026 ReferenceProcessor* rp = g1h->ref_processor_cm();
2027
2028 // See the comment in G1CollectedHeap::ref_processing_init()
2029 // about how reference processing currently works in G1.
2030
2031 // Set the soft reference policy
2032 rp->setup_policy(clear_all_soft_refs);
2033 assert(_markStack.isEmpty(), "mark stack should be empty");
2034
2035 // Instances of the 'Keep Alive' and 'Complete GC' closures used
2036 // in serial reference processing. Note these closures are also
2037 // used for serially processing (by the the current thread) the
2038 // JNI references during parallel reference processing.
2039 //
2040 // These closures do not need to synchronize with the worker
2041 // threads involved in parallel reference processing as these
2042 // instances are executed serially by the current thread (e.g.
2043 // reference processing is not multi-threaded and is thus
2044 // performed by the current thread instead of a gang worker).
2045 //
2046 // The gang tasks involved in parallel reference processing create
2047 // their own instances of these closures, which do their own
2048 // synchronization among themselves.
2049 G1CMKeepAliveAndDrainClosure g1_keep_alive(this, task(0), true /* is_serial */);
2050 G1CMDrainMarkingStackClosure g1_drain_mark_stack(this, task(0), true /* is_serial */);
2051
2052 // We need at least one active thread. If reference processing
2053 // is not multi-threaded we use the current (VMThread) thread,
2054 // otherwise we use the work gang from the G1CollectedHeap and
2055 // we utilize all the worker threads we can.
2056 bool processing_is_mt = rp->processing_is_mt();
2057 uint active_workers = (processing_is_mt ? g1h->workers()->active_workers() : 1U);
2058 active_workers = MAX2(MIN2(active_workers, _max_worker_id), 1U);
2059
2060 // Parallel processing task executor.
2061 G1CMRefProcTaskExecutor par_task_executor(g1h, this,
2062 g1h->workers(), active_workers);
2063 AbstractRefProcTaskExecutor* executor = (processing_is_mt ? &par_task_executor : NULL);
2064
2065 // Set the concurrency level. The phase was already set prior to
2066 // executing the remark task.
2067 set_concurrency(active_workers);
2068
2069 // Set the degree of MT processing here. If the discovery was done MT,
2070 // the number of threads involved during discovery could differ from
2071 // the number of active workers. This is OK as long as the discovered
2072 // Reference lists are balanced (see balance_all_queues() and balance_queues()).
2073 rp->set_active_mt_degree(active_workers);
2074
2075 // Process the weak references.
2076 const ReferenceProcessorStats& stats =
2077 rp->process_discovered_references(&g1_is_alive,
2078 &g1_keep_alive,
2079 &g1_drain_mark_stack,
2080 executor,
2081 _gc_timer_cm);
2082 _gc_tracer_cm->report_gc_reference_stats(stats);
2083
2084 // The do_oop work routines of the keep_alive and drain_marking_stack
2085 // oop closures will set the has_overflown flag if we overflow the
2086 // global marking stack.
2087
2088 assert(_markStack.overflow() || _markStack.isEmpty(),
2089 "mark stack should be empty (unless it overflowed)");
2090
2091 if (_markStack.overflow()) {
2092 // This should have been done already when we tried to push an
2093 // entry on to the global mark stack. But let's do it again.
2094 set_has_overflown();
2095 }
2096
2097 assert(rp->num_q() == active_workers, "why not");
2098
2099 rp->enqueue_discovered_references(executor);
2100
2101 rp->verify_no_references_recorded();
2102 assert(!rp->discovery_enabled(), "Post condition");
2103 }
2104
2105 if (has_overflown()) {
2106 // We can not trust g1_is_alive if the marking stack overflowed
2107 return;
2108 }
2109
2110 assert(_markStack.isEmpty(), "Marking should have completed");
2111
2112 // Unload Klasses, String, Symbols, Code Cache, etc.
2113 {
2114 GCTraceTime(Debug, gc) trace("Unloading", _gc_timer_cm);
2115
2116 if (ClassUnloadingWithConcurrentMark) {
2117 bool purged_classes;
2118
2119 {
2120 GCTraceTime(Trace, gc) trace("System Dictionary Unloading", _gc_timer_cm);
2121 purged_classes = SystemDictionary::do_unloading(&g1_is_alive, false /* Defer klass cleaning */);
2122 }
2123
2124 {
2125 GCTraceTime(Trace, gc) trace("Parallel Unloading", _gc_timer_cm);
2126 weakRefsWorkParallelPart(&g1_is_alive, purged_classes);
2127 }
2128 }
2129
2130 if (G1StringDedup::is_enabled()) {
2131 GCTraceTime(Trace, gc) trace("String Deduplication Unlink", _gc_timer_cm);
2132 G1StringDedup::unlink(&g1_is_alive);
2133 }
2134 }
2135 }
2136
2137 void G1ConcurrentMark::swapMarkBitMaps() {
2138 G1CMBitMapRO* temp = _prevMarkBitMap;
2139 _prevMarkBitMap = (G1CMBitMapRO*)_nextMarkBitMap;
2140 _nextMarkBitMap = (G1CMBitMap*) temp;
2141 }
2142
2143 BitMap G1ConcurrentMark::allocate_large_bitmap(BitMap::idx_t size_in_bits) {
2144 size_t size_in_words = BitMap::size_in_words(size_in_bits);
2145
2146 BitMap::bm_word_t* map = MmapArrayAllocator<BitMap::bm_word_t, mtGC>::allocate(size_in_words);
2147
2148 return BitMap(map, size_in_bits);
2149 }
2150
2151 void G1ConcurrentMark::allocate_internal_bitmaps() {
2152 double start_time = os::elapsedTime();
2153
2154 _region_live_bm = allocate_large_bitmap(_g1h->max_regions());
2155
2156 guarantee(_g1h->max_capacity() % CardTableModRefBS::card_size == 0,
2157 "Heap capacity must be aligned to card size.");
2158 _card_live_bm = allocate_large_bitmap(_g1h->max_capacity() / CardTableModRefBS::card_size);
2159
2160 log_debug(gc, marking)("Allocating internal bitmaps took %1.2f seconds.", os::elapsedTime() - start_time);
2161 }
2162
2163 void G1ConcurrentMark::pretouch_internal_bitmaps() {
2164 double start_time = os::elapsedTime();
2165
2166 _region_live_bm.pretouch();
2167 _card_live_bm.pretouch();
2168
2169 log_debug(gc, marking)("Pre-touching internal bitmaps took %1.2f seconds.", os::elapsedTime() - start_time);
2170 }
2171
2172 // Closure for marking entries in SATB buffers.
2173 class G1CMSATBBufferClosure : public SATBBufferClosure {
2174 private:
2175 G1CMTask* _task;
2176 G1CollectedHeap* _g1h;
2177
2178 // This is very similar to G1CMTask::deal_with_reference, but with
2179 // more relaxed requirements for the argument, so this must be more
2180 // circumspect about treating the argument as an object.
2181 void do_entry(void* entry) const {
2182 _task->increment_refs_reached();
2183 HeapRegion* hr = _g1h->heap_region_containing(entry);
2184 if (entry < hr->next_top_at_mark_start()) {
2185 // Until we get here, we don't know whether entry refers to a valid
2186 // object; it could instead have been a stale reference.
2187 oop obj = static_cast<oop>(entry);
2188 assert(obj->is_oop(true /* ignore mark word */),
2189 "Invalid oop in SATB buffer: " PTR_FORMAT, p2i(obj));
2190 _task->make_reference_grey(obj);
2191 }
2192 }
2193
2194 public:
2195 G1CMSATBBufferClosure(G1CMTask* task, G1CollectedHeap* g1h)
2196 : _task(task), _g1h(g1h) { }
2197
2198 virtual void do_buffer(void** buffer, size_t size) {
2199 for (size_t i = 0; i < size; ++i) {
2200 do_entry(buffer[i]);
2201 }
2202 }
2203 };
2204
2205 class G1RemarkThreadsClosure : public ThreadClosure {
2206 G1CMSATBBufferClosure _cm_satb_cl;
2207 G1CMOopClosure _cm_cl;
2208 MarkingCodeBlobClosure _code_cl;
2209 int _thread_parity;
2210
2211 public:
2212 G1RemarkThreadsClosure(G1CollectedHeap* g1h, G1CMTask* task) :
2213 _cm_satb_cl(task, g1h),
2214 _cm_cl(g1h, g1h->concurrent_mark(), task),
2215 _code_cl(&_cm_cl, !CodeBlobToOopClosure::FixRelocations),
2216 _thread_parity(Threads::thread_claim_parity()) {}
2217
2218 void do_thread(Thread* thread) {
2219 if (thread->is_Java_thread()) {
2220 if (thread->claim_oops_do(true, _thread_parity)) {
2221 JavaThread* jt = (JavaThread*)thread;
2222
2223 // In theory it should not be neccessary to explicitly walk the nmethods to find roots for concurrent marking
2224 // however the liveness of oops reachable from nmethods have very complex lifecycles:
2225 // * Alive if on the stack of an executing method
2226 // * Weakly reachable otherwise
2227 // Some objects reachable from nmethods, such as the class loader (or klass_holder) of the receiver should be
2228 // live by the SATB invariant but other oops recorded in nmethods may behave differently.
2229 jt->nmethods_do(&_code_cl);
2230
2231 jt->satb_mark_queue().apply_closure_and_empty(&_cm_satb_cl);
2232 }
2233 } else if (thread->is_VM_thread()) {
2234 if (thread->claim_oops_do(true, _thread_parity)) {
2235 JavaThread::satb_mark_queue_set().shared_satb_queue()->apply_closure_and_empty(&_cm_satb_cl);
2236 }
2237 }
2238 }
2239 };
2240
2241 class G1CMRemarkTask: public AbstractGangTask {
2242 private:
2243 G1ConcurrentMark* _cm;
2244 public:
2245 void work(uint worker_id) {
2246 // Since all available tasks are actually started, we should
2247 // only proceed if we're supposed to be active.
2248 if (worker_id < _cm->active_tasks()) {
2249 G1CMTask* task = _cm->task(worker_id);
2250 task->record_start_time();
2251 {
2252 ResourceMark rm;
2253 HandleMark hm;
2254
2255 G1RemarkThreadsClosure threads_f(G1CollectedHeap::heap(), task);
2256 Threads::threads_do(&threads_f);
2257 }
2258
2259 do {
2260 task->do_marking_step(1000000000.0 /* something very large */,
2261 true /* do_termination */,
2262 false /* is_serial */);
2263 } while (task->has_aborted() && !_cm->has_overflown());
2264 // If we overflow, then we do not want to restart. We instead
2265 // want to abort remark and do concurrent marking again.
2266 task->record_end_time();
2267 }
2268 }
2269
2270 G1CMRemarkTask(G1ConcurrentMark* cm, uint active_workers) :
2271 AbstractGangTask("Par Remark"), _cm(cm) {
2272 _cm->terminator()->reset_for_reuse(active_workers);
2273 }
2274 };
2275
2276 void G1ConcurrentMark::checkpointRootsFinalWork() {
2277 ResourceMark rm;
2278 HandleMark hm;
2279 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2280
2281 GCTraceTime(Debug, gc) trace("Finalize Marking", _gc_timer_cm);
2282
2283 g1h->ensure_parsability(false);
2284
2285 // this is remark, so we'll use up all active threads
2286 uint active_workers = g1h->workers()->active_workers();
2287 set_concurrency_and_phase(active_workers, false /* concurrent */);
2288 // Leave _parallel_marking_threads at it's
2289 // value originally calculated in the G1ConcurrentMark
2290 // constructor and pass values of the active workers
2291 // through the gang in the task.
2292
2293 {
2294 StrongRootsScope srs(active_workers);
2295
2296 G1CMRemarkTask remarkTask(this, active_workers);
2297 // We will start all available threads, even if we decide that the
2298 // active_workers will be fewer. The extra ones will just bail out
2299 // immediately.
2300 g1h->workers()->run_task(&remarkTask);
2301 }
2302
2303 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2304 guarantee(has_overflown() ||
2305 satb_mq_set.completed_buffers_num() == 0,
2306 "Invariant: has_overflown = %s, num buffers = " SIZE_FORMAT,
2307 BOOL_TO_STR(has_overflown()),
2308 satb_mq_set.completed_buffers_num());
2309
2310 print_stats();
2311 }
2312
2313 void G1ConcurrentMark::clearRangePrevBitmap(MemRegion mr) {
2314 // Note we are overriding the read-only view of the prev map here, via
2315 // the cast.
2316 ((G1CMBitMap*)_prevMarkBitMap)->clear_range(mr);
2317 }
2318
2319 HeapRegion*
2320 G1ConcurrentMark::claim_region(uint worker_id) {
2321 // "checkpoint" the finger
2322 HeapWord* finger = _finger;
2323
2324 // _heap_end will not change underneath our feet; it only changes at
2325 // yield points.
2326 while (finger < _heap_end) {
2327 assert(_g1h->is_in_g1_reserved(finger), "invariant");
2328
2329 HeapRegion* curr_region = _g1h->heap_region_containing(finger);
2330
2331 // Above heap_region_containing may return NULL as we always scan claim
2332 // until the end of the heap. In this case, just jump to the next region.
2333 HeapWord* end = curr_region != NULL ? curr_region->end() : finger + HeapRegion::GrainWords;
2334
2335 // Is the gap between reading the finger and doing the CAS too long?
2336 HeapWord* res = (HeapWord*) Atomic::cmpxchg_ptr(end, &_finger, finger);
2337 if (res == finger && curr_region != NULL) {
2338 // we succeeded
2339 HeapWord* bottom = curr_region->bottom();
2340 HeapWord* limit = curr_region->next_top_at_mark_start();
2341
2342 // notice that _finger == end cannot be guaranteed here since,
2343 // someone else might have moved the finger even further
2344 assert(_finger >= end, "the finger should have moved forward");
2345
2346 if (limit > bottom) {
2347 return curr_region;
2348 } else {
2349 assert(limit == bottom,
2350 "the region limit should be at bottom");
2351 // we return NULL and the caller should try calling
2352 // claim_region() again.
2353 return NULL;
2354 }
2355 } else {
2356 assert(_finger > finger, "the finger should have moved forward");
2357 // read it again
2358 finger = _finger;
2359 }
2360 }
2361
2362 return NULL;
2363 }
2364
2365 #ifndef PRODUCT
2366 class VerifyNoCSetOops VALUE_OBJ_CLASS_SPEC {
2367 private:
2368 G1CollectedHeap* _g1h;
2369 const char* _phase;
2370 int _info;
2371
2372 public:
2373 VerifyNoCSetOops(const char* phase, int info = -1) :
2374 _g1h(G1CollectedHeap::heap()),
2375 _phase(phase),
2376 _info(info)
2377 { }
2378
2379 void operator()(oop obj) const {
2380 guarantee(obj->is_oop(),
2381 "Non-oop " PTR_FORMAT ", phase: %s, info: %d",
2382 p2i(obj), _phase, _info);
2383 guarantee(!_g1h->obj_in_cs(obj),
2384 "obj: " PTR_FORMAT " in CSet, phase: %s, info: %d",
2385 p2i(obj), _phase, _info);
2386 }
2387 };
2388
2389 void G1ConcurrentMark::verify_no_cset_oops() {
2390 assert(SafepointSynchronize::is_at_safepoint(), "should be at a safepoint");
2391 if (!G1CollectedHeap::heap()->collector_state()->mark_in_progress()) {
2392 return;
2393 }
2394
2395 // Verify entries on the global mark stack
2396 _markStack.iterate(VerifyNoCSetOops("Stack"));
2397
2398 // Verify entries on the task queues
2399 for (uint i = 0; i < _max_worker_id; ++i) {
2400 G1CMTaskQueue* queue = _task_queues->queue(i);
2401 queue->iterate(VerifyNoCSetOops("Queue", i));
2402 }
2403
2404 // Verify the global finger
2405 HeapWord* global_finger = finger();
2406 if (global_finger != NULL && global_finger < _heap_end) {
2407 // Since we always iterate over all regions, we might get a NULL HeapRegion
2408 // here.
2409 HeapRegion* global_hr = _g1h->heap_region_containing(global_finger);
2410 guarantee(global_hr == NULL || global_finger == global_hr->bottom(),
2411 "global finger: " PTR_FORMAT " region: " HR_FORMAT,
2412 p2i(global_finger), HR_FORMAT_PARAMS(global_hr));
2413 }
2414
2415 // Verify the task fingers
2416 assert(parallel_marking_threads() <= _max_worker_id, "sanity");
2417 for (uint i = 0; i < parallel_marking_threads(); ++i) {
2418 G1CMTask* task = _tasks[i];
2419 HeapWord* task_finger = task->finger();
2420 if (task_finger != NULL && task_finger < _heap_end) {
2421 // See above note on the global finger verification.
2422 HeapRegion* task_hr = _g1h->heap_region_containing(task_finger);
2423 guarantee(task_hr == NULL || task_finger == task_hr->bottom() ||
2424 !task_hr->in_collection_set(),
2425 "task finger: " PTR_FORMAT " region: " HR_FORMAT,
2426 p2i(task_finger), HR_FORMAT_PARAMS(task_hr));
2427 }
2428 }
2429 }
2430 #endif // PRODUCT
2431
2432 class G1CreateLiveDataTask: public AbstractGangTask {
2433 // Aggregate the counting data that was constructed concurrently
2434 // with marking.
2435 class G1CreateLiveDataHRClosure: public G1LiveDataClosureBase {
2436 public:
2437 G1CreateLiveDataHRClosure(G1CollectedHeap* g1h,
2438 BitMap* cm_card_bm)
2439 : G1LiveDataClosureBase(g1h, NULL, cm_card_bm) { }
2440
2441 bool doHeapRegion(HeapRegion* hr) {
2442 size_t temp;
2443 return mark_marked_during_marking(hr, true, &temp);
2444 }
2445 };
2446
2447 G1CollectedHeap* _g1h;
2448 G1ConcurrentMark* _cm;
2449 BitMap* _cm_card_bm;
2450 HeapRegionClaimer _hrclaimer;
2451
2452 public:
2453 G1CreateLiveDataTask(G1CollectedHeap* g1h,
2454 BitMap* cm_card_bm,
2455 uint n_workers) :
2456 AbstractGangTask("Create Live Data"),
2457 _g1h(g1h),
2458 _cm_card_bm(cm_card_bm),
2459 _hrclaimer(n_workers) {
2460 }
2461
2462 void work(uint worker_id) {
2463 SuspendibleThreadSetJoiner sts_join;
2464
2465 G1CreateLiveDataHRClosure cl(_g1h, _cm_card_bm);
2466 _g1h->heap_region_par_iterate(&cl, worker_id, &_hrclaimer);
2467 }
2468 };
2469
2470
2471 void G1ConcurrentMark::create_live_data() {
2472 uint n_workers = _parallel_workers->active_workers();
2473
2474 G1CreateLiveDataTask cl(_g1h,
2475 &_card_live_bm,
2476 n_workers);
2477 _parallel_workers->run_task(&cl);
2478 }
2479
2480 class G1ClearAllLiveDataTask : public AbstractGangTask {
2481 BitMap* _bitmap;
2482 size_t _num_tasks;
2483 size_t _cur_task;
2484 public:
2485 G1ClearAllLiveDataTask(BitMap* bitmap, size_t num_tasks) :
2486 AbstractGangTask("Clear All Live Data"),
2487 _bitmap(bitmap),
2488 _num_tasks(num_tasks),
2489 _cur_task(0) {
2490 }
2491
2492 virtual void work(uint worker_id) {
2493 while (true) {
2494 size_t to_process = Atomic::add(1, &_cur_task) - 1;
2495 if (to_process >= _num_tasks) {
2496 break;
2497 }
2498
2499 BitMap::idx_t start = M * BitsPerByte * to_process;
2500 BitMap::idx_t end = MIN2(start + M * BitsPerByte, _bitmap->size());
2501 _bitmap->clear_range(start, end);
2502 }
2503 }
2504 };
2505
2506 void G1ConcurrentMark::clear_all_live_data(WorkGang* workers) {
2507 double start_time = os::elapsedTime();
2508
2509 guarantee(Universe::is_fully_initialized(), "Should not call this during initialization.");
2510
2511 size_t const num_chunks = align_size_up(_card_live_bm.size_in_words() * HeapWordSize, M) / M;
2512
2513 G1ClearAllLiveDataTask cl(&_card_live_bm, num_chunks);
2514 workers->run_task(&cl);
2515
2516 // The region live bitmap is always very small, even for huge heaps. Clear
2517 // directly.
2518 _region_live_bm.clear();
2519
2520
2521 log_debug(gc, marking)("Clear Live Data took %.3fms", (os::elapsedTime() - start_time) * 1000.0);
2522 }
2523
2524 void G1ConcurrentMark::verify_all_live_data() {
2525 assert(_card_live_bm.count_one_bits() == 0, "Master card bitmap not clear");
2526 assert(_region_live_bm.count_one_bits() == 0, "Master region bitmap not clear");
2527 }
2528
2529 void G1ConcurrentMark::print_stats() {
2530 if (!log_is_enabled(Debug, gc, stats)) {
2531 return;
2532 }
2533 log_debug(gc, stats)("---------------------------------------------------------------------");
2534 for (size_t i = 0; i < _active_tasks; ++i) {
2535 _tasks[i]->print_stats();
2536 log_debug(gc, stats)("---------------------------------------------------------------------");
2537 }
2538 }
2539
2540 void G1ConcurrentMark::abort() {
2541 if (!cmThread()->during_cycle() || _has_aborted) {
2542 // We haven't started a concurrent cycle or we have already aborted it. No need to do anything.
2543 return;
2544 }
2545
2546 // Clear all marks in the next bitmap for the next marking cycle. This will allow us to skip the next
2547 // concurrent bitmap clearing.
2548 clear_bitmap(_nextMarkBitMap, _g1h->workers(), false);
2549
2550 // Note we cannot clear the previous marking bitmap here
2551 // since VerifyDuringGC verifies the objects marked during
2552 // a full GC against the previous bitmap.
2553
2554 clear_all_live_data(_g1h->workers());
2555 DEBUG_ONLY(verify_all_live_data());
2556 // Empty mark stack
2557 reset_marking_state();
2558 for (uint i = 0; i < _max_worker_id; ++i) {
2559 _tasks[i]->clear_region_fields();
2560 }
2561 _first_overflow_barrier_sync.abort();
2562 _second_overflow_barrier_sync.abort();
2563 _has_aborted = true;
2564
2565 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2566 satb_mq_set.abandon_partial_marking();
2567 // This can be called either during or outside marking, we'll read
2568 // the expected_active value from the SATB queue set.
2569 satb_mq_set.set_active_all_threads(
2570 false, /* new active value */
2571 satb_mq_set.is_active() /* expected_active */);
2572 }
2573
2574 static void print_ms_time_info(const char* prefix, const char* name,
2575 NumberSeq& ns) {
2576 log_trace(gc, marking)("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).",
2577 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg());
2578 if (ns.num() > 0) {
2579 log_trace(gc, marking)("%s [std. dev = %8.2f ms, max = %8.2f ms]",
2580 prefix, ns.sd(), ns.maximum());
2581 }
2582 }
2583
2584 void G1ConcurrentMark::print_summary_info() {
2585 LogHandle(gc, marking) log;
2586 if (!log.is_trace()) {
2587 return;
2588 }
2589
2590 log.trace(" Concurrent marking:");
2591 print_ms_time_info(" ", "init marks", _init_times);
2592 print_ms_time_info(" ", "remarks", _remark_times);
2593 {
2594 print_ms_time_info(" ", "final marks", _remark_mark_times);
2595 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times);
2596
2597 }
2598 print_ms_time_info(" ", "cleanups", _cleanup_times);
2599 log.trace(" Finalize live data total time = %8.2f s (avg = %8.2f ms).",
2600 _total_counting_time, (_cleanup_times.num() > 0 ? _total_counting_time * 1000.0 / (double)_cleanup_times.num() : 0.0));
2601 if (G1ScrubRemSets) {
2602 log.trace(" RS scrub total time = %8.2f s (avg = %8.2f ms).",
2603 _total_rs_scrub_time, (_cleanup_times.num() > 0 ? _total_rs_scrub_time * 1000.0 / (double)_cleanup_times.num() : 0.0));
2604 }
2605 log.trace(" Total stop_world time = %8.2f s.",
2606 (_init_times.sum() + _remark_times.sum() + _cleanup_times.sum())/1000.0);
2607 log.trace(" Total concurrent time = %8.2f s (%8.2f s marking).",
2608 cmThread()->vtime_accum(), cmThread()->vtime_mark_accum());
2609 }
2610
2611 void G1ConcurrentMark::print_worker_threads_on(outputStream* st) const {
2612 _parallel_workers->print_worker_threads_on(st);
2613 }
2614
2615 void G1ConcurrentMark::print_on_error(outputStream* st) const {
2616 st->print_cr("Marking Bits (Prev, Next): (CMBitMap*) " PTR_FORMAT ", (CMBitMap*) " PTR_FORMAT,
2617 p2i(_prevMarkBitMap), p2i(_nextMarkBitMap));
2618 _prevMarkBitMap->print_on_error(st, " Prev Bits: ");
2619 _nextMarkBitMap->print_on_error(st, " Next Bits: ");
2620 }
2621
2622 // We take a break if someone is trying to stop the world.
2623 bool G1ConcurrentMark::do_yield_check(uint worker_id) {
2624 if (SuspendibleThreadSet::should_yield()) {
2625 SuspendibleThreadSet::yield();
2626 return true;
2627 } else {
2628 return false;
2629 }
2630 }
2631
2632 // Closure for iteration over bitmaps
2633 class G1CMBitMapClosure : public BitMapClosure {
2634 private:
2635 // the bitmap that is being iterated over
2636 G1CMBitMap* _nextMarkBitMap;
2637 G1ConcurrentMark* _cm;
2638 G1CMTask* _task;
2639
2640 public:
2641 G1CMBitMapClosure(G1CMTask *task, G1ConcurrentMark* cm, G1CMBitMap* nextMarkBitMap) :
2642 _task(task), _cm(cm), _nextMarkBitMap(nextMarkBitMap) { }
2643
2644 bool do_bit(size_t offset) {
2645 HeapWord* addr = _nextMarkBitMap->offsetToHeapWord(offset);
2646 assert(_nextMarkBitMap->isMarked(addr), "invariant");
2647 assert( addr < _cm->finger(), "invariant");
2648 assert(addr >= _task->finger(), "invariant");
2649
2650 // We move that task's local finger along.
2651 _task->move_finger_to(addr);
2652
2653 _task->scan_object(oop(addr));
2654 // we only partially drain the local queue and global stack
2655 _task->drain_local_queue(true);
2656 _task->drain_global_stack(true);
2657
2658 // if the has_aborted flag has been raised, we need to bail out of
2659 // the iteration
2660 return !_task->has_aborted();
2661 }
2662 };
2663
2664 static ReferenceProcessor* get_cm_oop_closure_ref_processor(G1CollectedHeap* g1h) {
2665 ReferenceProcessor* result = g1h->ref_processor_cm();
2666 assert(result != NULL, "CM reference processor should not be NULL");
2667 return result;
2668 }
2669
2670 G1CMOopClosure::G1CMOopClosure(G1CollectedHeap* g1h,
2671 G1ConcurrentMark* cm,
2672 G1CMTask* task)
2673 : MetadataAwareOopClosure(get_cm_oop_closure_ref_processor(g1h)),
2674 _g1h(g1h), _cm(cm), _task(task)
2675 { }
2676
2677 void G1CMTask::setup_for_region(HeapRegion* hr) {
2678 assert(hr != NULL,
2679 "claim_region() should have filtered out NULL regions");
2680 _curr_region = hr;
2681 _finger = hr->bottom();
2682 update_region_limit();
2683 }
2684
2685 void G1CMTask::update_region_limit() {
2686 HeapRegion* hr = _curr_region;
2687 HeapWord* bottom = hr->bottom();
2688 HeapWord* limit = hr->next_top_at_mark_start();
2689
2690 if (limit == bottom) {
2691 // The region was collected underneath our feet.
2692 // We set the finger to bottom to ensure that the bitmap
2693 // iteration that will follow this will not do anything.
2694 // (this is not a condition that holds when we set the region up,
2695 // as the region is not supposed to be empty in the first place)
2696 _finger = bottom;
2697 } else if (limit >= _region_limit) {
2698 assert(limit >= _finger, "peace of mind");
2699 } else {
2700 assert(limit < _region_limit, "only way to get here");
2701 // This can happen under some pretty unusual circumstances. An
2702 // evacuation pause empties the region underneath our feet (NTAMS
2703 // at bottom). We then do some allocation in the region (NTAMS
2704 // stays at bottom), followed by the region being used as a GC
2705 // alloc region (NTAMS will move to top() and the objects
2706 // originally below it will be grayed). All objects now marked in
2707 // the region are explicitly grayed, if below the global finger,
2708 // and we do not need in fact to scan anything else. So, we simply
2709 // set _finger to be limit to ensure that the bitmap iteration
2710 // doesn't do anything.
2711 _finger = limit;
2712 }
2713
2714 _region_limit = limit;
2715 }
2716
2717 void G1CMTask::giveup_current_region() {
2718 assert(_curr_region != NULL, "invariant");
2719 clear_region_fields();
2720 }
2721
2722 void G1CMTask::clear_region_fields() {
2723 // Values for these three fields that indicate that we're not
2724 // holding on to a region.
2725 _curr_region = NULL;
2726 _finger = NULL;
2727 _region_limit = NULL;
2728 }
2729
2730 void G1CMTask::set_cm_oop_closure(G1CMOopClosure* cm_oop_closure) {
2731 if (cm_oop_closure == NULL) {
2732 assert(_cm_oop_closure != NULL, "invariant");
2733 } else {
2734 assert(_cm_oop_closure == NULL, "invariant");
2735 }
2736 _cm_oop_closure = cm_oop_closure;
2737 }
2738
2739 void G1CMTask::reset(G1CMBitMap* nextMarkBitMap) {
2740 guarantee(nextMarkBitMap != NULL, "invariant");
2741 _nextMarkBitMap = nextMarkBitMap;
2742 clear_region_fields();
2743
2744 _calls = 0;
2745 _elapsed_time_ms = 0.0;
2746 _termination_time_ms = 0.0;
2747 _termination_start_time_ms = 0.0;
2748 }
2749
2750 bool G1CMTask::should_exit_termination() {
2751 regular_clock_call();
2752 // This is called when we are in the termination protocol. We should
2753 // quit if, for some reason, this task wants to abort or the global
2754 // stack is not empty (this means that we can get work from it).
2755 return !_cm->mark_stack_empty() || has_aborted();
2756 }
2757
2758 void G1CMTask::reached_limit() {
2759 assert(_words_scanned >= _words_scanned_limit ||
2760 _refs_reached >= _refs_reached_limit ,
2761 "shouldn't have been called otherwise");
2762 regular_clock_call();
2763 }
2764
2765 void G1CMTask::regular_clock_call() {
2766 if (has_aborted()) return;
2767
2768 // First, we need to recalculate the words scanned and refs reached
2769 // limits for the next clock call.
2770 recalculate_limits();
2771
2772 // During the regular clock call we do the following
2773
2774 // (1) If an overflow has been flagged, then we abort.
2775 if (_cm->has_overflown()) {
2776 set_has_aborted();
2777 return;
2778 }
2779
2780 // If we are not concurrent (i.e. we're doing remark) we don't need
2781 // to check anything else. The other steps are only needed during
2782 // the concurrent marking phase.
2783 if (!concurrent()) return;
2784
2785 // (2) If marking has been aborted for Full GC, then we also abort.
2786 if (_cm->has_aborted()) {
2787 set_has_aborted();
2788 return;
2789 }
2790
2791 double curr_time_ms = os::elapsedVTime() * 1000.0;
2792
2793 // (4) We check whether we should yield. If we have to, then we abort.
2794 if (SuspendibleThreadSet::should_yield()) {
2795 // We should yield. To do this we abort the task. The caller is
2796 // responsible for yielding.
2797 set_has_aborted();
2798 return;
2799 }
2800
2801 // (5) We check whether we've reached our time quota. If we have,
2802 // then we abort.
2803 double elapsed_time_ms = curr_time_ms - _start_time_ms;
2804 if (elapsed_time_ms > _time_target_ms) {
2805 set_has_aborted();
2806 _has_timed_out = true;
2807 return;
2808 }
2809
2810 // (6) Finally, we check whether there are enough completed STAB
2811 // buffers available for processing. If there are, we abort.
2812 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2813 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) {
2814 // we do need to process SATB buffers, we'll abort and restart
2815 // the marking task to do so
2816 set_has_aborted();
2817 return;
2818 }
2819 }
2820
2821 void G1CMTask::recalculate_limits() {
2822 _real_words_scanned_limit = _words_scanned + words_scanned_period;
2823 _words_scanned_limit = _real_words_scanned_limit;
2824
2825 _real_refs_reached_limit = _refs_reached + refs_reached_period;
2826 _refs_reached_limit = _real_refs_reached_limit;
2827 }
2828
2829 void G1CMTask::decrease_limits() {
2830 // This is called when we believe that we're going to do an infrequent
2831 // operation which will increase the per byte scanned cost (i.e. move
2832 // entries to/from the global stack). It basically tries to decrease the
2833 // scanning limit so that the clock is called earlier.
2834
2835 _words_scanned_limit = _real_words_scanned_limit -
2836 3 * words_scanned_period / 4;
2837 _refs_reached_limit = _real_refs_reached_limit -
2838 3 * refs_reached_period / 4;
2839 }
2840
2841 void G1CMTask::move_entries_to_global_stack() {
2842 // local array where we'll store the entries that will be popped
2843 // from the local queue
2844 oop buffer[global_stack_transfer_size];
2845
2846 int n = 0;
2847 oop obj;
2848 while (n < global_stack_transfer_size && _task_queue->pop_local(obj)) {
2849 buffer[n] = obj;
2850 ++n;
2851 }
2852
2853 if (n > 0) {
2854 // we popped at least one entry from the local queue
2855
2856 if (!_cm->mark_stack_push(buffer, n)) {
2857 set_has_aborted();
2858 }
2859 }
2860
2861 // this operation was quite expensive, so decrease the limits
2862 decrease_limits();
2863 }
2864
2865 void G1CMTask::get_entries_from_global_stack() {
2866 // local array where we'll store the entries that will be popped
2867 // from the global stack.
2868 oop buffer[global_stack_transfer_size];
2869 int n;
2870 _cm->mark_stack_pop(buffer, global_stack_transfer_size, &n);
2871 assert(n <= global_stack_transfer_size,
2872 "we should not pop more than the given limit");
2873 if (n > 0) {
2874 // yes, we did actually pop at least one entry
2875 for (int i = 0; i < n; ++i) {
2876 bool success = _task_queue->push(buffer[i]);
2877 // We only call this when the local queue is empty or under a
2878 // given target limit. So, we do not expect this push to fail.
2879 assert(success, "invariant");
2880 }
2881 }
2882
2883 // this operation was quite expensive, so decrease the limits
2884 decrease_limits();
2885 }
2886
2887 void G1CMTask::drain_local_queue(bool partially) {
2888 if (has_aborted()) return;
2889
2890 // Decide what the target size is, depending whether we're going to
2891 // drain it partially (so that other tasks can steal if they run out
2892 // of things to do) or totally (at the very end).
2893 size_t target_size;
2894 if (partially) {
2895 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize);
2896 } else {
2897 target_size = 0;
2898 }
2899
2900 if (_task_queue->size() > target_size) {
2901 oop obj;
2902 bool ret = _task_queue->pop_local(obj);
2903 while (ret) {
2904 assert(_g1h->is_in_g1_reserved((HeapWord*) obj), "invariant" );
2905 assert(!_g1h->is_on_master_free_list(
2906 _g1h->heap_region_containing((HeapWord*) obj)), "invariant");
2907
2908 scan_object(obj);
2909
2910 if (_task_queue->size() <= target_size || has_aborted()) {
2911 ret = false;
2912 } else {
2913 ret = _task_queue->pop_local(obj);
2914 }
2915 }
2916 }
2917 }
2918
2919 void G1CMTask::drain_global_stack(bool partially) {
2920 if (has_aborted()) return;
2921
2922 // We have a policy to drain the local queue before we attempt to
2923 // drain the global stack.
2924 assert(partially || _task_queue->size() == 0, "invariant");
2925
2926 // Decide what the target size is, depending whether we're going to
2927 // drain it partially (so that other tasks can steal if they run out
2928 // of things to do) or totally (at the very end). Notice that,
2929 // because we move entries from the global stack in chunks or
2930 // because another task might be doing the same, we might in fact
2931 // drop below the target. But, this is not a problem.
2932 size_t target_size;
2933 if (partially) {
2934 target_size = _cm->partial_mark_stack_size_target();
2935 } else {
2936 target_size = 0;
2937 }
2938
2939 if (_cm->mark_stack_size() > target_size) {
2940 while (!has_aborted() && _cm->mark_stack_size() > target_size) {
2941 get_entries_from_global_stack();
2942 drain_local_queue(partially);
2943 }
2944 }
2945 }
2946
2947 // SATB Queue has several assumptions on whether to call the par or
2948 // non-par versions of the methods. this is why some of the code is
2949 // replicated. We should really get rid of the single-threaded version
2950 // of the code to simplify things.
2951 void G1CMTask::drain_satb_buffers() {
2952 if (has_aborted()) return;
2953
2954 // We set this so that the regular clock knows that we're in the
2955 // middle of draining buffers and doesn't set the abort flag when it
2956 // notices that SATB buffers are available for draining. It'd be
2957 // very counter productive if it did that. :-)
2958 _draining_satb_buffers = true;
2959
2960 G1CMSATBBufferClosure satb_cl(this, _g1h);
2961 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2962
2963 // This keeps claiming and applying the closure to completed buffers
2964 // until we run out of buffers or we need to abort.
2965 while (!has_aborted() &&
2966 satb_mq_set.apply_closure_to_completed_buffer(&satb_cl)) {
2967 regular_clock_call();
2968 }
2969
2970 _draining_satb_buffers = false;
2971
2972 assert(has_aborted() ||
2973 concurrent() ||
2974 satb_mq_set.completed_buffers_num() == 0, "invariant");
2975
2976 // again, this was a potentially expensive operation, decrease the
2977 // limits to get the regular clock call early
2978 decrease_limits();
2979 }
2980
2981 void G1CMTask::print_stats() {
2982 log_debug(gc, stats)("Marking Stats, task = %u, calls = %d",
2983 _worker_id, _calls);
2984 log_debug(gc, stats)(" Elapsed time = %1.2lfms, Termination time = %1.2lfms",
2985 _elapsed_time_ms, _termination_time_ms);
2986 log_debug(gc, stats)(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms",
2987 _step_times_ms.num(), _step_times_ms.avg(),
2988 _step_times_ms.sd());
2989 log_debug(gc, stats)(" max = %1.2lfms, total = %1.2lfms",
2990 _step_times_ms.maximum(), _step_times_ms.sum());
2991 }
2992
2993 bool G1ConcurrentMark::try_stealing(uint worker_id, int* hash_seed, oop& obj) {
2994 return _task_queues->steal(worker_id, hash_seed, obj);
2995 }
2996
2997 /*****************************************************************************
2998
2999 The do_marking_step(time_target_ms, ...) method is the building
3000 block of the parallel marking framework. It can be called in parallel
3001 with other invocations of do_marking_step() on different tasks
3002 (but only one per task, obviously) and concurrently with the
3003 mutator threads, or during remark, hence it eliminates the need
3004 for two versions of the code. When called during remark, it will
3005 pick up from where the task left off during the concurrent marking
3006 phase. Interestingly, tasks are also claimable during evacuation
3007 pauses too, since do_marking_step() ensures that it aborts before
3008 it needs to yield.
3009
3010 The data structures that it uses to do marking work are the
3011 following:
3012
3013 (1) Marking Bitmap. If there are gray objects that appear only
3014 on the bitmap (this happens either when dealing with an overflow
3015 or when the initial marking phase has simply marked the roots
3016 and didn't push them on the stack), then tasks claim heap
3017 regions whose bitmap they then scan to find gray objects. A
3018 global finger indicates where the end of the last claimed region
3019 is. A local finger indicates how far into the region a task has
3020 scanned. The two fingers are used to determine how to gray an
3021 object (i.e. whether simply marking it is OK, as it will be
3022 visited by a task in the future, or whether it needs to be also
3023 pushed on a stack).
3024
3025 (2) Local Queue. The local queue of the task which is accessed
3026 reasonably efficiently by the task. Other tasks can steal from
3027 it when they run out of work. Throughout the marking phase, a
3028 task attempts to keep its local queue short but not totally
3029 empty, so that entries are available for stealing by other
3030 tasks. Only when there is no more work, a task will totally
3031 drain its local queue.
3032
3033 (3) Global Mark Stack. This handles local queue overflow. During
3034 marking only sets of entries are moved between it and the local
3035 queues, as access to it requires a mutex and more fine-grain
3036 interaction with it which might cause contention. If it
3037 overflows, then the marking phase should restart and iterate
3038 over the bitmap to identify gray objects. Throughout the marking
3039 phase, tasks attempt to keep the global mark stack at a small
3040 length but not totally empty, so that entries are available for
3041 popping by other tasks. Only when there is no more work, tasks
3042 will totally drain the global mark stack.
3043
3044 (4) SATB Buffer Queue. This is where completed SATB buffers are
3045 made available. Buffers are regularly removed from this queue
3046 and scanned for roots, so that the queue doesn't get too
3047 long. During remark, all completed buffers are processed, as
3048 well as the filled in parts of any uncompleted buffers.
3049
3050 The do_marking_step() method tries to abort when the time target
3051 has been reached. There are a few other cases when the
3052 do_marking_step() method also aborts:
3053
3054 (1) When the marking phase has been aborted (after a Full GC).
3055
3056 (2) When a global overflow (on the global stack) has been
3057 triggered. Before the task aborts, it will actually sync up with
3058 the other tasks to ensure that all the marking data structures
3059 (local queues, stacks, fingers etc.) are re-initialized so that
3060 when do_marking_step() completes, the marking phase can
3061 immediately restart.
3062
3063 (3) When enough completed SATB buffers are available. The
3064 do_marking_step() method only tries to drain SATB buffers right
3065 at the beginning. So, if enough buffers are available, the
3066 marking step aborts and the SATB buffers are processed at
3067 the beginning of the next invocation.
3068
3069 (4) To yield. when we have to yield then we abort and yield
3070 right at the end of do_marking_step(). This saves us from a lot
3071 of hassle as, by yielding we might allow a Full GC. If this
3072 happens then objects will be compacted underneath our feet, the
3073 heap might shrink, etc. We save checking for this by just
3074 aborting and doing the yield right at the end.
3075
3076 From the above it follows that the do_marking_step() method should
3077 be called in a loop (or, otherwise, regularly) until it completes.
3078
3079 If a marking step completes without its has_aborted() flag being
3080 true, it means it has completed the current marking phase (and
3081 also all other marking tasks have done so and have all synced up).
3082
3083 A method called regular_clock_call() is invoked "regularly" (in
3084 sub ms intervals) throughout marking. It is this clock method that
3085 checks all the abort conditions which were mentioned above and
3086 decides when the task should abort. A work-based scheme is used to
3087 trigger this clock method: when the number of object words the
3088 marking phase has scanned or the number of references the marking
3089 phase has visited reach a given limit. Additional invocations to
3090 the method clock have been planted in a few other strategic places
3091 too. The initial reason for the clock method was to avoid calling
3092 vtime too regularly, as it is quite expensive. So, once it was in
3093 place, it was natural to piggy-back all the other conditions on it
3094 too and not constantly check them throughout the code.
3095
3096 If do_termination is true then do_marking_step will enter its
3097 termination protocol.
3098
3099 The value of is_serial must be true when do_marking_step is being
3100 called serially (i.e. by the VMThread) and do_marking_step should
3101 skip any synchronization in the termination and overflow code.
3102 Examples include the serial remark code and the serial reference
3103 processing closures.
3104
3105 The value of is_serial must be false when do_marking_step is
3106 being called by any of the worker threads in a work gang.
3107 Examples include the concurrent marking code (CMMarkingTask),
3108 the MT remark code, and the MT reference processing closures.
3109
3110 *****************************************************************************/
3111
3112 void G1CMTask::do_marking_step(double time_target_ms,
3113 bool do_termination,
3114 bool is_serial) {
3115 assert(time_target_ms >= 1.0, "minimum granularity is 1ms");
3116 assert(concurrent() == _cm->concurrent(), "they should be the same");
3117
3118 G1CollectorPolicy* g1_policy = _g1h->g1_policy();
3119 assert(_task_queues != NULL, "invariant");
3120 assert(_task_queue != NULL, "invariant");
3121 assert(_task_queues->queue(_worker_id) == _task_queue, "invariant");
3122
3123 assert(!_claimed,
3124 "only one thread should claim this task at any one time");
3125
3126 // OK, this doesn't safeguard again all possible scenarios, as it is
3127 // possible for two threads to set the _claimed flag at the same
3128 // time. But it is only for debugging purposes anyway and it will
3129 // catch most problems.
3130 _claimed = true;
3131
3132 _start_time_ms = os::elapsedVTime() * 1000.0;
3133
3134 // If do_stealing is true then do_marking_step will attempt to
3135 // steal work from the other G1CMTasks. It only makes sense to
3136 // enable stealing when the termination protocol is enabled
3137 // and do_marking_step() is not being called serially.
3138 bool do_stealing = do_termination && !is_serial;
3139
3140 double diff_prediction_ms = _g1h->g1_policy()->predictor().get_new_prediction(&_marking_step_diffs_ms);
3141 _time_target_ms = time_target_ms - diff_prediction_ms;
3142
3143 // set up the variables that are used in the work-based scheme to
3144 // call the regular clock method
3145 _words_scanned = 0;
3146 _refs_reached = 0;
3147 recalculate_limits();
3148
3149 // clear all flags
3150 clear_has_aborted();
3151 _has_timed_out = false;
3152 _draining_satb_buffers = false;
3153
3154 ++_calls;
3155
3156 // Set up the bitmap and oop closures. Anything that uses them is
3157 // eventually called from this method, so it is OK to allocate these
3158 // statically.
3159 G1CMBitMapClosure bitmap_closure(this, _cm, _nextMarkBitMap);
3160 G1CMOopClosure cm_oop_closure(_g1h, _cm, this);
3161 set_cm_oop_closure(&cm_oop_closure);
3162
3163 if (_cm->has_overflown()) {
3164 // This can happen if the mark stack overflows during a GC pause
3165 // and this task, after a yield point, restarts. We have to abort
3166 // as we need to get into the overflow protocol which happens
3167 // right at the end of this task.
3168 set_has_aborted();
3169 }
3170
3171 // First drain any available SATB buffers. After this, we will not
3172 // look at SATB buffers before the next invocation of this method.
3173 // If enough completed SATB buffers are queued up, the regular clock
3174 // will abort this task so that it restarts.
3175 drain_satb_buffers();
3176 // ...then partially drain the local queue and the global stack
3177 drain_local_queue(true);
3178 drain_global_stack(true);
3179
3180 do {
3181 if (!has_aborted() && _curr_region != NULL) {
3182 // This means that we're already holding on to a region.
3183 assert(_finger != NULL, "if region is not NULL, then the finger "
3184 "should not be NULL either");
3185
3186 // We might have restarted this task after an evacuation pause
3187 // which might have evacuated the region we're holding on to
3188 // underneath our feet. Let's read its limit again to make sure
3189 // that we do not iterate over a region of the heap that
3190 // contains garbage (update_region_limit() will also move
3191 // _finger to the start of the region if it is found empty).
3192 update_region_limit();
3193 // We will start from _finger not from the start of the region,
3194 // as we might be restarting this task after aborting half-way
3195 // through scanning this region. In this case, _finger points to
3196 // the address where we last found a marked object. If this is a
3197 // fresh region, _finger points to start().
3198 MemRegion mr = MemRegion(_finger, _region_limit);
3199
3200 assert(!_curr_region->is_humongous() || mr.start() == _curr_region->bottom(),
3201 "humongous regions should go around loop once only");
3202
3203 // Some special cases:
3204 // If the memory region is empty, we can just give up the region.
3205 // If the current region is humongous then we only need to check
3206 // the bitmap for the bit associated with the start of the object,
3207 // scan the object if it's live, and give up the region.
3208 // Otherwise, let's iterate over the bitmap of the part of the region
3209 // that is left.
3210 // If the iteration is successful, give up the region.
3211 if (mr.is_empty()) {
3212 giveup_current_region();
3213 regular_clock_call();
3214 } else if (_curr_region->is_humongous() && mr.start() == _curr_region->bottom()) {
3215 if (_nextMarkBitMap->isMarked(mr.start())) {
3216 // The object is marked - apply the closure
3217 BitMap::idx_t offset = _nextMarkBitMap->heapWordToOffset(mr.start());
3218 bitmap_closure.do_bit(offset);
3219 }
3220 // Even if this task aborted while scanning the humongous object
3221 // we can (and should) give up the current region.
3222 giveup_current_region();
3223 regular_clock_call();
3224 } else if (_nextMarkBitMap->iterate(&bitmap_closure, mr)) {
3225 giveup_current_region();
3226 regular_clock_call();
3227 } else {
3228 assert(has_aborted(), "currently the only way to do so");
3229 // The only way to abort the bitmap iteration is to return
3230 // false from the do_bit() method. However, inside the
3231 // do_bit() method we move the _finger to point to the
3232 // object currently being looked at. So, if we bail out, we
3233 // have definitely set _finger to something non-null.
3234 assert(_finger != NULL, "invariant");
3235
3236 // Region iteration was actually aborted. So now _finger
3237 // points to the address of the object we last scanned. If we
3238 // leave it there, when we restart this task, we will rescan
3239 // the object. It is easy to avoid this. We move the finger by
3240 // enough to point to the next possible object header (the
3241 // bitmap knows by how much we need to move it as it knows its
3242 // granularity).
3243 assert(_finger < _region_limit, "invariant");
3244 HeapWord* new_finger = _nextMarkBitMap->nextObject(_finger);
3245 // Check if bitmap iteration was aborted while scanning the last object
3246 if (new_finger >= _region_limit) {
3247 giveup_current_region();
3248 } else {
3249 move_finger_to(new_finger);
3250 }
3251 }
3252 }
3253 // At this point we have either completed iterating over the
3254 // region we were holding on to, or we have aborted.
3255
3256 // We then partially drain the local queue and the global stack.
3257 // (Do we really need this?)
3258 drain_local_queue(true);
3259 drain_global_stack(true);
3260
3261 // Read the note on the claim_region() method on why it might
3262 // return NULL with potentially more regions available for
3263 // claiming and why we have to check out_of_regions() to determine
3264 // whether we're done or not.
3265 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) {
3266 // We are going to try to claim a new region. We should have
3267 // given up on the previous one.
3268 // Separated the asserts so that we know which one fires.
3269 assert(_curr_region == NULL, "invariant");
3270 assert(_finger == NULL, "invariant");
3271 assert(_region_limit == NULL, "invariant");
3272 HeapRegion* claimed_region = _cm->claim_region(_worker_id);
3273 if (claimed_region != NULL) {
3274 // Yes, we managed to claim one
3275 setup_for_region(claimed_region);
3276 assert(_curr_region == claimed_region, "invariant");
3277 }
3278 // It is important to call the regular clock here. It might take
3279 // a while to claim a region if, for example, we hit a large
3280 // block of empty regions. So we need to call the regular clock
3281 // method once round the loop to make sure it's called
3282 // frequently enough.
3283 regular_clock_call();
3284 }
3285
3286 if (!has_aborted() && _curr_region == NULL) {
3287 assert(_cm->out_of_regions(),
3288 "at this point we should be out of regions");
3289 }
3290 } while ( _curr_region != NULL && !has_aborted());
3291
3292 if (!has_aborted()) {
3293 // We cannot check whether the global stack is empty, since other
3294 // tasks might be pushing objects to it concurrently.
3295 assert(_cm->out_of_regions(),
3296 "at this point we should be out of regions");
3297 // Try to reduce the number of available SATB buffers so that
3298 // remark has less work to do.
3299 drain_satb_buffers();
3300 }
3301
3302 // Since we've done everything else, we can now totally drain the
3303 // local queue and global stack.
3304 drain_local_queue(false);
3305 drain_global_stack(false);
3306
3307 // Attempt at work stealing from other task's queues.
3308 if (do_stealing && !has_aborted()) {
3309 // We have not aborted. This means that we have finished all that
3310 // we could. Let's try to do some stealing...
3311
3312 // We cannot check whether the global stack is empty, since other
3313 // tasks might be pushing objects to it concurrently.
3314 assert(_cm->out_of_regions() && _task_queue->size() == 0,
3315 "only way to reach here");
3316 while (!has_aborted()) {
3317 oop obj;
3318 if (_cm->try_stealing(_worker_id, &_hash_seed, obj)) {
3319 assert(_nextMarkBitMap->isMarked((HeapWord*) obj),
3320 "any stolen object should be marked");
3321 scan_object(obj);
3322
3323 // And since we're towards the end, let's totally drain the
3324 // local queue and global stack.
3325 drain_local_queue(false);
3326 drain_global_stack(false);
3327 } else {
3328 break;
3329 }
3330 }
3331 }
3332
3333 // We still haven't aborted. Now, let's try to get into the
3334 // termination protocol.
3335 if (do_termination && !has_aborted()) {
3336 // We cannot check whether the global stack is empty, since other
3337 // tasks might be concurrently pushing objects on it.
3338 // Separated the asserts so that we know which one fires.
3339 assert(_cm->out_of_regions(), "only way to reach here");
3340 assert(_task_queue->size() == 0, "only way to reach here");
3341 _termination_start_time_ms = os::elapsedVTime() * 1000.0;
3342
3343 // The G1CMTask class also extends the TerminatorTerminator class,
3344 // hence its should_exit_termination() method will also decide
3345 // whether to exit the termination protocol or not.
3346 bool finished = (is_serial ||
3347 _cm->terminator()->offer_termination(this));
3348 double termination_end_time_ms = os::elapsedVTime() * 1000.0;
3349 _termination_time_ms +=
3350 termination_end_time_ms - _termination_start_time_ms;
3351
3352 if (finished) {
3353 // We're all done.
3354
3355 if (_worker_id == 0) {
3356 // let's allow task 0 to do this
3357 if (concurrent()) {
3358 assert(_cm->concurrent_marking_in_progress(), "invariant");
3359 // we need to set this to false before the next
3360 // safepoint. This way we ensure that the marking phase
3361 // doesn't observe any more heap expansions.
3362 _cm->clear_concurrent_marking_in_progress();
3363 }
3364 }
3365
3366 // We can now guarantee that the global stack is empty, since
3367 // all other tasks have finished. We separated the guarantees so
3368 // that, if a condition is false, we can immediately find out
3369 // which one.
3370 guarantee(_cm->out_of_regions(), "only way to reach here");
3371 guarantee(_cm->mark_stack_empty(), "only way to reach here");
3372 guarantee(_task_queue->size() == 0, "only way to reach here");
3373 guarantee(!_cm->has_overflown(), "only way to reach here");
3374 guarantee(!_cm->mark_stack_overflow(), "only way to reach here");
3375 } else {
3376 // Apparently there's more work to do. Let's abort this task. It
3377 // will restart it and we can hopefully find more things to do.
3378 set_has_aborted();
3379 }
3380 }
3381
3382 // Mainly for debugging purposes to make sure that a pointer to the
3383 // closure which was statically allocated in this frame doesn't
3384 // escape it by accident.
3385 set_cm_oop_closure(NULL);
3386 double end_time_ms = os::elapsedVTime() * 1000.0;
3387 double elapsed_time_ms = end_time_ms - _start_time_ms;
3388 // Update the step history.
3389 _step_times_ms.add(elapsed_time_ms);
3390
3391 if (has_aborted()) {
3392 // The task was aborted for some reason.
3393 if (_has_timed_out) {
3394 double diff_ms = elapsed_time_ms - _time_target_ms;
3395 // Keep statistics of how well we did with respect to hitting
3396 // our target only if we actually timed out (if we aborted for
3397 // other reasons, then the results might get skewed).
3398 _marking_step_diffs_ms.add(diff_ms);
3399 }
3400
3401 if (_cm->has_overflown()) {
3402 // This is the interesting one. We aborted because a global
3403 // overflow was raised. This means we have to restart the
3404 // marking phase and start iterating over regions. However, in
3405 // order to do this we have to make sure that all tasks stop
3406 // what they are doing and re-initialize in a safe manner. We
3407 // will achieve this with the use of two barrier sync points.
3408
3409 if (!is_serial) {
3410 // We only need to enter the sync barrier if being called
3411 // from a parallel context
3412 _cm->enter_first_sync_barrier(_worker_id);
3413
3414 // When we exit this sync barrier we know that all tasks have
3415 // stopped doing marking work. So, it's now safe to
3416 // re-initialize our data structures. At the end of this method,
3417 // task 0 will clear the global data structures.
3418 }
3419
3420 // We clear the local state of this task...
3421 clear_region_fields();
3422
3423 if (!is_serial) {
3424 // ...and enter the second barrier.
3425 _cm->enter_second_sync_barrier(_worker_id);
3426 }
3427 // At this point, if we're during the concurrent phase of
3428 // marking, everything has been re-initialized and we're
3429 // ready to restart.
3430 }
3431 }
3432
3433 _claimed = false;
3434 }
3435
3436 G1CMTask::G1CMTask(uint worker_id,
3437 G1ConcurrentMark* cm,
3438 G1CMTaskQueue* task_queue,
3439 G1CMTaskQueueSet* task_queues)
3440 : _g1h(G1CollectedHeap::heap()),
3441 _worker_id(worker_id), _cm(cm),
3442 _claimed(false),
3443 _nextMarkBitMap(NULL), _hash_seed(17),
3444 _task_queue(task_queue),
3445 _task_queues(task_queues),
3446 _cm_oop_closure(NULL) {
3447 guarantee(task_queue != NULL, "invariant");
3448 guarantee(task_queues != NULL, "invariant");
3449
3450 _marking_step_diffs_ms.add(0.5);
3451 }
3452
3453 // These are formatting macros that are used below to ensure
3454 // consistent formatting. The *_H_* versions are used to format the
3455 // header for a particular value and they should be kept consistent
3456 // with the corresponding macro. Also note that most of the macros add
3457 // the necessary white space (as a prefix) which makes them a bit
3458 // easier to compose.
3459
3460 // All the output lines are prefixed with this string to be able to
3461 // identify them easily in a large log file.
3462 #define G1PPRL_LINE_PREFIX "###"
3463
3464 #define G1PPRL_ADDR_BASE_FORMAT " " PTR_FORMAT "-" PTR_FORMAT
3465 #ifdef _LP64
3466 #define G1PPRL_ADDR_BASE_H_FORMAT " %37s"
3467 #else // _LP64
3468 #define G1PPRL_ADDR_BASE_H_FORMAT " %21s"
3469 #endif // _LP64
3470
3471 // For per-region info
3472 #define G1PPRL_TYPE_FORMAT " %-4s"
3473 #define G1PPRL_TYPE_H_FORMAT " %4s"
3474 #define G1PPRL_BYTE_FORMAT " " SIZE_FORMAT_W(9)
3475 #define G1PPRL_BYTE_H_FORMAT " %9s"
3476 #define G1PPRL_DOUBLE_FORMAT " %14.1f"
3477 #define G1PPRL_DOUBLE_H_FORMAT " %14s"
3478
3479 // For summary info
3480 #define G1PPRL_SUM_ADDR_FORMAT(tag) " " tag ":" G1PPRL_ADDR_BASE_FORMAT
3481 #define G1PPRL_SUM_BYTE_FORMAT(tag) " " tag ": " SIZE_FORMAT
3482 #define G1PPRL_SUM_MB_FORMAT(tag) " " tag ": %1.2f MB"
3483 #define G1PPRL_SUM_MB_PERC_FORMAT(tag) G1PPRL_SUM_MB_FORMAT(tag) " / %1.2f %%"
3484
3485 G1PrintRegionLivenessInfoClosure::
3486 G1PrintRegionLivenessInfoClosure(const char* phase_name)
3487 : _total_used_bytes(0), _total_capacity_bytes(0),
3488 _total_prev_live_bytes(0), _total_next_live_bytes(0),
3489 _total_remset_bytes(0), _total_strong_code_roots_bytes(0) {
3490 G1CollectedHeap* g1h = G1CollectedHeap::heap();
3491 MemRegion g1_reserved = g1h->g1_reserved();
3492 double now = os::elapsedTime();
3493
3494 // Print the header of the output.
3495 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX" PHASE %s @ %1.3f", phase_name, now);
3496 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX" HEAP"
3497 G1PPRL_SUM_ADDR_FORMAT("reserved")
3498 G1PPRL_SUM_BYTE_FORMAT("region-size"),
3499 p2i(g1_reserved.start()), p2i(g1_reserved.end()),
3500 HeapRegion::GrainBytes);
3501 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX);
3502 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
3503 G1PPRL_TYPE_H_FORMAT
3504 G1PPRL_ADDR_BASE_H_FORMAT
3505 G1PPRL_BYTE_H_FORMAT
3506 G1PPRL_BYTE_H_FORMAT
3507 G1PPRL_BYTE_H_FORMAT
3508 G1PPRL_DOUBLE_H_FORMAT
3509 G1PPRL_BYTE_H_FORMAT
3510 G1PPRL_BYTE_H_FORMAT,
3511 "type", "address-range",
3512 "used", "prev-live", "next-live", "gc-eff",
3513 "remset", "code-roots");
3514 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
3515 G1PPRL_TYPE_H_FORMAT
3516 G1PPRL_ADDR_BASE_H_FORMAT
3517 G1PPRL_BYTE_H_FORMAT
3518 G1PPRL_BYTE_H_FORMAT
3519 G1PPRL_BYTE_H_FORMAT
3520 G1PPRL_DOUBLE_H_FORMAT
3521 G1PPRL_BYTE_H_FORMAT
3522 G1PPRL_BYTE_H_FORMAT,
3523 "", "",
3524 "(bytes)", "(bytes)", "(bytes)", "(bytes/ms)",
3525 "(bytes)", "(bytes)");
3526 }
3527
3528 bool G1PrintRegionLivenessInfoClosure::doHeapRegion(HeapRegion* r) {
3529 const char* type = r->get_type_str();
3530 HeapWord* bottom = r->bottom();
3531 HeapWord* end = r->end();
3532 size_t capacity_bytes = r->capacity();
3533 size_t used_bytes = r->used();
3534 size_t prev_live_bytes = r->live_bytes();
3535 size_t next_live_bytes = r->next_live_bytes();
3536 double gc_eff = r->gc_efficiency();
3537 size_t remset_bytes = r->rem_set()->mem_size();
3538 size_t strong_code_roots_bytes = r->rem_set()->strong_code_roots_mem_size();
3539
3540 _total_used_bytes += used_bytes;
3541 _total_capacity_bytes += capacity_bytes;
3542 _total_prev_live_bytes += prev_live_bytes;
3543 _total_next_live_bytes += next_live_bytes;
3544 _total_remset_bytes += remset_bytes;
3545 _total_strong_code_roots_bytes += strong_code_roots_bytes;
3546
3547 // Print a line for this particular region.
3548 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
3549 G1PPRL_TYPE_FORMAT
3550 G1PPRL_ADDR_BASE_FORMAT
3551 G1PPRL_BYTE_FORMAT
3552 G1PPRL_BYTE_FORMAT
3553 G1PPRL_BYTE_FORMAT
3554 G1PPRL_DOUBLE_FORMAT
3555 G1PPRL_BYTE_FORMAT
3556 G1PPRL_BYTE_FORMAT,
3557 type, p2i(bottom), p2i(end),
3558 used_bytes, prev_live_bytes, next_live_bytes, gc_eff,
3559 remset_bytes, strong_code_roots_bytes);
3560
3561 return false;
3562 }
3563
3564 G1PrintRegionLivenessInfoClosure::~G1PrintRegionLivenessInfoClosure() {
3565 // add static memory usages to remembered set sizes
3566 _total_remset_bytes += HeapRegionRemSet::fl_mem_size() + HeapRegionRemSet::static_mem_size();
3567 // Print the footer of the output.
3568 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX);
3569 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
3570 " SUMMARY"
3571 G1PPRL_SUM_MB_FORMAT("capacity")
3572 G1PPRL_SUM_MB_PERC_FORMAT("used")
3573 G1PPRL_SUM_MB_PERC_FORMAT("prev-live")
3574 G1PPRL_SUM_MB_PERC_FORMAT("next-live")
3575 G1PPRL_SUM_MB_FORMAT("remset")
3576 G1PPRL_SUM_MB_FORMAT("code-roots"),
3577 bytes_to_mb(_total_capacity_bytes),
3578 bytes_to_mb(_total_used_bytes),
3579 perc(_total_used_bytes, _total_capacity_bytes),
3580 bytes_to_mb(_total_prev_live_bytes),
3581 perc(_total_prev_live_bytes, _total_capacity_bytes),
3582 bytes_to_mb(_total_next_live_bytes),
3583 perc(_total_next_live_bytes, _total_capacity_bytes),
3584 bytes_to_mb(_total_remset_bytes),
3585 bytes_to_mb(_total_strong_code_roots_bytes));
3586 }