1 /*
2 * Copyright (c) 2001, 2015, 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/concurrentMark.inline.hpp"
30 #include "gc/g1/concurrentMarkThread.inline.hpp"
31 #include "gc/g1/g1CollectedHeap.inline.hpp"
32 #include "gc/g1/g1CollectorPolicy.hpp"
33 #include "gc/g1/g1CollectorState.hpp"
34 #include "gc/g1/g1ErgoVerbose.hpp"
35 #include "gc/g1/g1Log.hpp"
36 #include "gc/g1/g1OopClosures.inline.hpp"
37 #include "gc/g1/g1RemSet.hpp"
38 #include "gc/g1/g1StringDedup.hpp"
39 #include "gc/g1/heapRegion.inline.hpp"
40 #include "gc/g1/heapRegionManager.inline.hpp"
41 #include "gc/g1/heapRegionRemSet.hpp"
42 #include "gc/g1/heapRegionSet.inline.hpp"
43 #include "gc/g1/suspendibleThreadSet.hpp"
44 #include "gc/shared/gcId.hpp"
45 #include "gc/shared/gcTimer.hpp"
46 #include "gc/shared/gcTrace.hpp"
47 #include "gc/shared/gcTraceTime.hpp"
48 #include "gc/shared/genOopClosures.inline.hpp"
49 #include "gc/shared/referencePolicy.hpp"
50 #include "gc/shared/strongRootsScope.hpp"
51 #include "gc/shared/taskqueue.inline.hpp"
52 #include "gc/shared/vmGCOperations.hpp"
53 #include "memory/allocation.hpp"
54 #include "memory/resourceArea.hpp"
55 #include "oops/oop.inline.hpp"
56 #include "runtime/atomic.inline.hpp"
57 #include "runtime/handles.inline.hpp"
58 #include "runtime/java.hpp"
59 #include "runtime/prefetch.inline.hpp"
60 #include "services/memTracker.hpp"
61
62 // Concurrent marking bit map wrapper
63
64 CMBitMapRO::CMBitMapRO(int shifter) :
65 _bm(),
66 _shifter(shifter) {
67 _bmStartWord = 0;
68 _bmWordSize = 0;
69 }
70
71 HeapWord* CMBitMapRO::getNextMarkedWordAddress(const HeapWord* addr,
72 const HeapWord* limit) const {
73 // First we must round addr *up* to a possible object boundary.
74 addr = (HeapWord*)align_size_up((intptr_t)addr,
75 HeapWordSize << _shifter);
76 size_t addrOffset = heapWordToOffset(addr);
77 assert(limit != NULL, "limit must not be NULL");
78 size_t limitOffset = heapWordToOffset(limit);
79 size_t nextOffset = _bm.get_next_one_offset(addrOffset, limitOffset);
80 HeapWord* nextAddr = offsetToHeapWord(nextOffset);
81 assert(nextAddr >= addr, "get_next_one postcondition");
82 assert(nextAddr == limit || isMarked(nextAddr),
83 "get_next_one postcondition");
84 return nextAddr;
85 }
86
87 #ifndef PRODUCT
88 bool CMBitMapRO::covers(MemRegion heap_rs) const {
89 // assert(_bm.map() == _virtual_space.low(), "map inconsistency");
90 assert(((size_t)_bm.size() * ((size_t)1 << _shifter)) == _bmWordSize,
91 "size inconsistency");
92 return _bmStartWord == (HeapWord*)(heap_rs.start()) &&
93 _bmWordSize == heap_rs.word_size();
94 }
95 #endif
96
97 void CMBitMapRO::print_on_error(outputStream* st, const char* prefix) const {
98 _bm.print_on_error(st, prefix);
99 }
100
101 size_t CMBitMap::compute_size(size_t heap_size) {
102 return ReservedSpace::allocation_align_size_up(heap_size / mark_distance());
103 }
104
105 size_t CMBitMap::mark_distance() {
106 return MinObjAlignmentInBytes * BitsPerByte;
107 }
108
109 void CMBitMap::initialize(MemRegion heap, G1RegionToSpaceMapper* storage) {
110 _bmStartWord = heap.start();
111 _bmWordSize = heap.word_size();
112
113 _bm.set_map((BitMap::bm_word_t*) storage->reserved().start());
114 _bm.set_size(_bmWordSize >> _shifter);
115
116 storage->set_mapping_changed_listener(&_listener);
117 }
118
119 void CMBitMapMappingChangedListener::on_commit(uint start_region, size_t num_regions, bool zero_filled) {
120 if (zero_filled) {
121 return;
122 }
123 // We need to clear the bitmap on commit, removing any existing information.
124 MemRegion mr(G1CollectedHeap::heap()->bottom_addr_for_region(start_region), num_regions * HeapRegion::GrainWords);
125 _bm->clearRange(mr);
126 }
127
128 // Closure used for clearing the given mark bitmap.
129 class ClearBitmapHRClosure : public HeapRegionClosure {
130 private:
131 ConcurrentMark* _cm;
132 CMBitMap* _bitmap;
133 bool _may_yield; // The closure may yield during iteration. If yielded, abort the iteration.
134 public:
135 ClearBitmapHRClosure(ConcurrentMark* cm, CMBitMap* bitmap, bool may_yield) : HeapRegionClosure(), _cm(cm), _bitmap(bitmap), _may_yield(may_yield) {
136 assert(!may_yield || cm != NULL, "CM must be non-NULL if this closure is expected to yield.");
137 }
138
139 virtual bool doHeapRegion(HeapRegion* r) {
140 size_t const chunk_size_in_words = M / HeapWordSize;
141
142 HeapWord* cur = r->bottom();
143 HeapWord* const end = r->end();
144
145 while (cur < end) {
146 MemRegion mr(cur, MIN2(cur + chunk_size_in_words, end));
147 _bitmap->clearRange(mr);
148
149 cur += chunk_size_in_words;
150
151 // Abort iteration if after yielding the marking has been aborted.
152 if (_may_yield && _cm->do_yield_check() && _cm->has_aborted()) {
153 return true;
154 }
155 // Repeat the asserts from before the start of the closure. We will do them
156 // as asserts here to minimize their overhead on the product. However, we
157 // will have them as guarantees at the beginning / end of the bitmap
158 // clearing to get some checking in the product.
159 assert(!_may_yield || _cm->cmThread()->during_cycle(), "invariant");
160 assert(!_may_yield || !G1CollectedHeap::heap()->collector_state()->mark_in_progress(), "invariant");
161 }
162
163 return false;
164 }
165 };
166
167 class ParClearNextMarkBitmapTask : public AbstractGangTask {
168 ClearBitmapHRClosure* _cl;
169 HeapRegionClaimer _hrclaimer;
170 bool _suspendible; // If the task is suspendible, workers must join the STS.
171
172 public:
173 ParClearNextMarkBitmapTask(ClearBitmapHRClosure *cl, uint n_workers, bool suspendible) :
174 _cl(cl), _suspendible(suspendible), AbstractGangTask("Parallel Clear Bitmap Task"), _hrclaimer(n_workers) {}
175
176 void work(uint worker_id) {
177 SuspendibleThreadSetJoiner sts_join(_suspendible);
178 G1CollectedHeap::heap()->heap_region_par_iterate(_cl, worker_id, &_hrclaimer, true);
179 }
180 };
181
182 void CMBitMap::clearAll() {
183 G1CollectedHeap* g1h = G1CollectedHeap::heap();
184 ClearBitmapHRClosure cl(NULL, this, false /* may_yield */);
185 uint n_workers = g1h->workers()->active_workers();
186 ParClearNextMarkBitmapTask task(&cl, n_workers, false);
187 g1h->workers()->run_task(&task);
188 guarantee(cl.complete(), "Must have completed iteration.");
189 return;
190 }
191
192 void CMBitMap::clearRange(MemRegion mr) {
193 mr.intersection(MemRegion(_bmStartWord, _bmWordSize));
194 assert(!mr.is_empty(), "unexpected empty region");
195 // convert address range into offset range
196 _bm.at_put_range(heapWordToOffset(mr.start()),
197 heapWordToOffset(mr.end()), false);
198 }
199
200 CMMarkStack::CMMarkStack(ConcurrentMark* cm) :
201 _base(NULL), _cm(cm)
202 {}
203
204 bool CMMarkStack::allocate(size_t capacity) {
205 // allocate a stack of the requisite depth
206 ReservedSpace rs(ReservedSpace::allocation_align_size_up(capacity * sizeof(oop)));
207 if (!rs.is_reserved()) {
208 warning("ConcurrentMark MarkStack allocation failure");
209 return false;
210 }
211 MemTracker::record_virtual_memory_type((address)rs.base(), mtGC);
212 if (!_virtual_space.initialize(rs, rs.size())) {
213 warning("ConcurrentMark MarkStack backing store failure");
214 // Release the virtual memory reserved for the marking stack
215 rs.release();
216 return false;
217 }
218 assert(_virtual_space.committed_size() == rs.size(),
219 "Didn't reserve backing store for all of ConcurrentMark stack?");
220 _base = (oop*) _virtual_space.low();
221 setEmpty();
222 _capacity = (jint) capacity;
223 _saved_index = -1;
224 _should_expand = false;
225 return true;
226 }
227
228 void CMMarkStack::expand() {
229 // Called, during remark, if we've overflown the marking stack during marking.
230 assert(isEmpty(), "stack should been emptied while handling overflow");
231 assert(_capacity <= (jint) MarkStackSizeMax, "stack bigger than permitted");
232 // Clear expansion flag
233 _should_expand = false;
234 if (_capacity == (jint) MarkStackSizeMax) {
235 if (PrintGCDetails && Verbose) {
236 gclog_or_tty->print_cr(" (benign) Can't expand marking stack capacity, at max size limit");
237 }
238 return;
239 }
240 // Double capacity if possible
241 jint new_capacity = MIN2(_capacity*2, (jint) MarkStackSizeMax);
242 // Do not give up existing stack until we have managed to
243 // get the double capacity that we desired.
244 ReservedSpace rs(ReservedSpace::allocation_align_size_up(new_capacity *
245 sizeof(oop)));
246 if (rs.is_reserved()) {
247 // Release the backing store associated with old stack
248 _virtual_space.release();
249 // Reinitialize virtual space for new stack
250 if (!_virtual_space.initialize(rs, rs.size())) {
251 fatal("Not enough swap for expanded marking stack capacity");
252 }
253 _base = (oop*)(_virtual_space.low());
254 _index = 0;
255 _capacity = new_capacity;
256 } else {
257 if (PrintGCDetails && Verbose) {
258 // Failed to double capacity, continue;
259 gclog_or_tty->print(" (benign) Failed to expand marking stack capacity from "
260 SIZE_FORMAT "K to " SIZE_FORMAT "K",
261 _capacity / K, new_capacity / K);
262 }
263 }
264 }
265
266 void CMMarkStack::set_should_expand() {
267 // If we're resetting the marking state because of an
268 // marking stack overflow, record that we should, if
269 // possible, expand the stack.
270 _should_expand = _cm->has_overflown();
271 }
272
273 CMMarkStack::~CMMarkStack() {
274 if (_base != NULL) {
275 _base = NULL;
276 _virtual_space.release();
277 }
278 }
279
280 void CMMarkStack::par_push_arr(oop* ptr_arr, int n) {
281 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
282 jint start = _index;
283 jint next_index = start + n;
284 if (next_index > _capacity) {
285 _overflow = true;
286 return;
287 }
288 // Otherwise.
289 _index = next_index;
290 for (int i = 0; i < n; i++) {
291 int ind = start + i;
292 assert(ind < _capacity, "By overflow test above.");
293 _base[ind] = ptr_arr[i];
294 }
295 }
296
297 bool CMMarkStack::par_pop_arr(oop* ptr_arr, int max, int* n) {
298 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
299 jint index = _index;
300 if (index == 0) {
301 *n = 0;
302 return false;
303 } else {
304 int k = MIN2(max, index);
305 jint new_ind = index - k;
306 for (int j = 0; j < k; j++) {
307 ptr_arr[j] = _base[new_ind + j];
308 }
309 _index = new_ind;
310 *n = k;
311 return true;
312 }
313 }
314
315 void CMMarkStack::note_start_of_gc() {
316 assert(_saved_index == -1,
317 "note_start_of_gc()/end_of_gc() bracketed incorrectly");
318 _saved_index = _index;
319 }
320
321 void CMMarkStack::note_end_of_gc() {
322 // This is intentionally a guarantee, instead of an assert. If we
323 // accidentally add something to the mark stack during GC, it
324 // will be a correctness issue so it's better if we crash. we'll
325 // only check this once per GC anyway, so it won't be a performance
326 // issue in any way.
327 guarantee(_saved_index == _index,
328 "saved index: %d index: %d", _saved_index, _index);
329 _saved_index = -1;
330 }
331
332 CMRootRegions::CMRootRegions() :
333 _young_list(NULL), _cm(NULL), _scan_in_progress(false),
334 _should_abort(false), _next_survivor(NULL) { }
335
336 void CMRootRegions::init(G1CollectedHeap* g1h, ConcurrentMark* cm) {
337 _young_list = g1h->young_list();
338 _cm = cm;
339 }
340
341 void CMRootRegions::prepare_for_scan() {
342 assert(!scan_in_progress(), "pre-condition");
343
344 // Currently, only survivors can be root regions.
345 assert(_next_survivor == NULL, "pre-condition");
346 _next_survivor = _young_list->first_survivor_region();
347 _scan_in_progress = (_next_survivor != NULL);
348 _should_abort = false;
349 }
350
351 HeapRegion* CMRootRegions::claim_next() {
352 if (_should_abort) {
353 // If someone has set the should_abort flag, we return NULL to
354 // force the caller to bail out of their loop.
355 return NULL;
356 }
357
358 // Currently, only survivors can be root regions.
359 HeapRegion* res = _next_survivor;
360 if (res != NULL) {
361 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
362 // Read it again in case it changed while we were waiting for the lock.
363 res = _next_survivor;
364 if (res != NULL) {
365 if (res == _young_list->last_survivor_region()) {
366 // We just claimed the last survivor so store NULL to indicate
367 // that we're done.
368 _next_survivor = NULL;
369 } else {
370 _next_survivor = res->get_next_young_region();
371 }
372 } else {
373 // Someone else claimed the last survivor while we were trying
374 // to take the lock so nothing else to do.
375 }
376 }
377 assert(res == NULL || res->is_survivor(), "post-condition");
378
379 return res;
380 }
381
382 void CMRootRegions::scan_finished() {
383 assert(scan_in_progress(), "pre-condition");
384
385 // Currently, only survivors can be root regions.
386 if (!_should_abort) {
387 assert(_next_survivor == NULL, "we should have claimed all survivors");
388 }
389 _next_survivor = NULL;
390
391 {
392 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
393 _scan_in_progress = false;
394 RootRegionScan_lock->notify_all();
395 }
396 }
397
398 bool CMRootRegions::wait_until_scan_finished() {
399 if (!scan_in_progress()) return false;
400
401 {
402 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
403 while (scan_in_progress()) {
404 RootRegionScan_lock->wait(Mutex::_no_safepoint_check_flag);
405 }
406 }
407 return true;
408 }
409
410 uint ConcurrentMark::scale_parallel_threads(uint n_par_threads) {
411 return MAX2((n_par_threads + 2) / 4, 1U);
412 }
413
414 ConcurrentMark::ConcurrentMark(G1CollectedHeap* g1h, G1RegionToSpaceMapper* prev_bitmap_storage, G1RegionToSpaceMapper* next_bitmap_storage) :
415 _g1h(g1h),
416 _markBitMap1(),
417 _markBitMap2(),
418 _parallel_marking_threads(0),
419 _max_parallel_marking_threads(0),
420 _sleep_factor(0.0),
421 _marking_task_overhead(1.0),
422 _cleanup_list("Cleanup List"),
423 _region_bm((BitMap::idx_t)(g1h->max_regions()), false /* in_resource_area*/),
424 _card_bm((g1h->reserved_region().byte_size() + CardTableModRefBS::card_size - 1) >>
425 CardTableModRefBS::card_shift,
426 false /* in_resource_area*/),
427
428 _prevMarkBitMap(&_markBitMap1),
429 _nextMarkBitMap(&_markBitMap2),
430
431 _markStack(this),
432 // _finger set in set_non_marking_state
433
434 _max_worker_id(ParallelGCThreads),
435 // _active_tasks set in set_non_marking_state
436 // _tasks set inside the constructor
437 _task_queues(new CMTaskQueueSet((int) _max_worker_id)),
438 _terminator(ParallelTaskTerminator((int) _max_worker_id, _task_queues)),
439
440 _has_overflown(false),
441 _concurrent(false),
442 _has_aborted(false),
443 _restart_for_overflow(false),
444 _concurrent_marking_in_progress(false),
445
446 // _verbose_level set below
447
448 _init_times(),
449 _remark_times(), _remark_mark_times(), _remark_weak_ref_times(),
450 _cleanup_times(),
451 _total_counting_time(0.0),
452 _total_rs_scrub_time(0.0),
453
454 _parallel_workers(NULL),
455
456 _count_card_bitmaps(NULL),
457 _count_marked_bytes(NULL),
458 _completed_initialization(false) {
459
460 _markBitMap1.initialize(g1h->reserved_region(), prev_bitmap_storage);
461 _markBitMap2.initialize(g1h->reserved_region(), next_bitmap_storage);
462
463 // Create & start a ConcurrentMark thread.
464 _cmThread = new ConcurrentMarkThread(this);
465 assert(cmThread() != NULL, "CM Thread should have been created");
466 assert(cmThread()->cm() != NULL, "CM Thread should refer to this cm");
467 if (_cmThread->osthread() == NULL) {
468 vm_shutdown_during_initialization("Could not create ConcurrentMarkThread");
469 }
470
471 assert(CGC_lock != NULL, "Where's the CGC_lock?");
472 assert(_markBitMap1.covers(g1h->reserved_region()), "_markBitMap1 inconsistency");
473 assert(_markBitMap2.covers(g1h->reserved_region()), "_markBitMap2 inconsistency");
474
475 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set();
476 satb_qs.set_buffer_size(G1SATBBufferSize);
477
478 _root_regions.init(_g1h, this);
479
480 if (ConcGCThreads > ParallelGCThreads) {
481 warning("Can't have more ConcGCThreads (%u) "
482 "than ParallelGCThreads (%u).",
483 ConcGCThreads, ParallelGCThreads);
484 return;
485 }
486 if (!FLAG_IS_DEFAULT(ConcGCThreads) && ConcGCThreads > 0) {
487 // Note: ConcGCThreads has precedence over G1MarkingOverheadPercent
488 // if both are set
489 _sleep_factor = 0.0;
490 _marking_task_overhead = 1.0;
491 } else if (G1MarkingOverheadPercent > 0) {
492 // We will calculate the number of parallel marking threads based
493 // on a target overhead with respect to the soft real-time goal
494 double marking_overhead = (double) G1MarkingOverheadPercent / 100.0;
495 double overall_cm_overhead =
496 (double) MaxGCPauseMillis * marking_overhead /
497 (double) GCPauseIntervalMillis;
498 double cpu_ratio = 1.0 / (double) os::processor_count();
499 double marking_thread_num = ceil(overall_cm_overhead / cpu_ratio);
500 double marking_task_overhead =
501 overall_cm_overhead / marking_thread_num *
502 (double) os::processor_count();
503 double sleep_factor =
504 (1.0 - marking_task_overhead) / marking_task_overhead;
505
506 FLAG_SET_ERGO(uint, ConcGCThreads, (uint) marking_thread_num);
507 _sleep_factor = sleep_factor;
508 _marking_task_overhead = marking_task_overhead;
509 } else {
510 // Calculate the number of parallel marking threads by scaling
511 // the number of parallel GC threads.
512 uint marking_thread_num = scale_parallel_threads(ParallelGCThreads);
513 FLAG_SET_ERGO(uint, ConcGCThreads, marking_thread_num);
514 _sleep_factor = 0.0;
515 _marking_task_overhead = 1.0;
516 }
517
518 assert(ConcGCThreads > 0, "Should have been set");
519 _parallel_marking_threads = ConcGCThreads;
520 _max_parallel_marking_threads = _parallel_marking_threads;
521
522 _parallel_workers = new WorkGang("G1 Marker",
523 _max_parallel_marking_threads, false, true);
524 if (_parallel_workers == NULL) {
525 vm_exit_during_initialization("Failed necessary allocation.");
526 } else {
527 _parallel_workers->initialize_workers();
528 }
529
530 if (FLAG_IS_DEFAULT(MarkStackSize)) {
531 size_t mark_stack_size =
532 MIN2(MarkStackSizeMax,
533 MAX2(MarkStackSize, (size_t) (parallel_marking_threads() * TASKQUEUE_SIZE)));
534 // Verify that the calculated value for MarkStackSize is in range.
535 // It would be nice to use the private utility routine from Arguments.
536 if (!(mark_stack_size >= 1 && mark_stack_size <= MarkStackSizeMax)) {
537 warning("Invalid value calculated for MarkStackSize (" SIZE_FORMAT "): "
538 "must be between 1 and " SIZE_FORMAT,
539 mark_stack_size, MarkStackSizeMax);
540 return;
541 }
542 FLAG_SET_ERGO(size_t, MarkStackSize, mark_stack_size);
543 } else {
544 // Verify MarkStackSize is in range.
545 if (FLAG_IS_CMDLINE(MarkStackSize)) {
546 if (FLAG_IS_DEFAULT(MarkStackSizeMax)) {
547 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
548 warning("Invalid value specified for MarkStackSize (" SIZE_FORMAT "): "
549 "must be between 1 and " SIZE_FORMAT,
550 MarkStackSize, MarkStackSizeMax);
551 return;
552 }
553 } else if (FLAG_IS_CMDLINE(MarkStackSizeMax)) {
554 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
555 warning("Invalid value specified for MarkStackSize (" SIZE_FORMAT ")"
556 " or for MarkStackSizeMax (" SIZE_FORMAT ")",
557 MarkStackSize, MarkStackSizeMax);
558 return;
559 }
560 }
561 }
562 }
563
564 if (!_markStack.allocate(MarkStackSize)) {
565 warning("Failed to allocate CM marking stack");
566 return;
567 }
568
569 _tasks = NEW_C_HEAP_ARRAY(CMTask*, _max_worker_id, mtGC);
570 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_worker_id, mtGC);
571
572 _count_card_bitmaps = NEW_C_HEAP_ARRAY(BitMap, _max_worker_id, mtGC);
573 _count_marked_bytes = NEW_C_HEAP_ARRAY(size_t*, _max_worker_id, mtGC);
574
575 BitMap::idx_t card_bm_size = _card_bm.size();
576
577 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail
578 _active_tasks = _max_worker_id;
579
580 uint max_regions = _g1h->max_regions();
581 for (uint i = 0; i < _max_worker_id; ++i) {
582 CMTaskQueue* task_queue = new CMTaskQueue();
583 task_queue->initialize();
584 _task_queues->register_queue(i, task_queue);
585
586 _count_card_bitmaps[i] = BitMap(card_bm_size, false);
587 _count_marked_bytes[i] = NEW_C_HEAP_ARRAY(size_t, max_regions, mtGC);
588
589 _tasks[i] = new CMTask(i, this,
590 _count_marked_bytes[i],
591 &_count_card_bitmaps[i],
592 task_queue, _task_queues);
593
594 _accum_task_vtime[i] = 0.0;
595 }
596
597 // Calculate the card number for the bottom of the heap. Used
598 // in biasing indexes into the accounting card bitmaps.
599 _heap_bottom_card_num =
600 intptr_t(uintptr_t(_g1h->reserved_region().start()) >>
601 CardTableModRefBS::card_shift);
602
603 // Clear all the liveness counting data
604 clear_all_count_data();
605
606 // so that the call below can read a sensible value
607 _heap_start = g1h->reserved_region().start();
608 set_non_marking_state();
609 _completed_initialization = true;
610 }
611
612 void ConcurrentMark::reset() {
613 // Starting values for these two. This should be called in a STW
614 // phase.
615 MemRegion reserved = _g1h->g1_reserved();
616 _heap_start = reserved.start();
617 _heap_end = reserved.end();
618
619 // Separated the asserts so that we know which one fires.
620 assert(_heap_start != NULL, "heap bounds should look ok");
621 assert(_heap_end != NULL, "heap bounds should look ok");
622 assert(_heap_start < _heap_end, "heap bounds should look ok");
623
624 // Reset all the marking data structures and any necessary flags
625 reset_marking_state();
626
627 // We do reset all of them, since different phases will use
628 // different number of active threads. So, it's easiest to have all
629 // of them ready.
630 for (uint i = 0; i < _max_worker_id; ++i) {
631 _tasks[i]->reset(_nextMarkBitMap);
632 }
633
634 // we need this to make sure that the flag is on during the evac
635 // pause with initial mark piggy-backed
636 set_concurrent_marking_in_progress();
637 }
638
639
640 void ConcurrentMark::reset_marking_state(bool clear_overflow) {
641 _markStack.set_should_expand();
642 _markStack.setEmpty(); // Also clears the _markStack overflow flag
643 if (clear_overflow) {
644 clear_has_overflown();
645 } else {
646 assert(has_overflown(), "pre-condition");
647 }
648 _finger = _heap_start;
649
650 for (uint i = 0; i < _max_worker_id; ++i) {
651 CMTaskQueue* queue = _task_queues->queue(i);
652 queue->set_empty();
653 }
654 }
655
656 void ConcurrentMark::set_concurrency(uint active_tasks) {
657 assert(active_tasks <= _max_worker_id, "we should not have more");
658
659 _active_tasks = active_tasks;
660 // Need to update the three data structures below according to the
661 // number of active threads for this phase.
662 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues);
663 _first_overflow_barrier_sync.set_n_workers((int) active_tasks);
664 _second_overflow_barrier_sync.set_n_workers((int) active_tasks);
665 }
666
667 void ConcurrentMark::set_concurrency_and_phase(uint active_tasks, bool concurrent) {
668 set_concurrency(active_tasks);
669
670 _concurrent = concurrent;
671 // We propagate this to all tasks, not just the active ones.
672 for (uint i = 0; i < _max_worker_id; ++i)
673 _tasks[i]->set_concurrent(concurrent);
674
675 if (concurrent) {
676 set_concurrent_marking_in_progress();
677 } else {
678 // We currently assume that the concurrent flag has been set to
679 // false before we start remark. At this point we should also be
680 // in a STW phase.
681 assert(!concurrent_marking_in_progress(), "invariant");
682 assert(out_of_regions(),
683 "only way to get here: _finger: " PTR_FORMAT ", _heap_end: " PTR_FORMAT,
684 p2i(_finger), p2i(_heap_end));
685 }
686 }
687
688 void ConcurrentMark::set_non_marking_state() {
689 // We set the global marking state to some default values when we're
690 // not doing marking.
691 reset_marking_state();
692 _active_tasks = 0;
693 clear_concurrent_marking_in_progress();
694 }
695
696 ConcurrentMark::~ConcurrentMark() {
697 // The ConcurrentMark instance is never freed.
698 ShouldNotReachHere();
699 }
700
701 void ConcurrentMark::clearNextBitmap() {
702 G1CollectedHeap* g1h = G1CollectedHeap::heap();
703
704 // Make sure that the concurrent mark thread looks to still be in
705 // the current cycle.
706 guarantee(cmThread()->during_cycle(), "invariant");
707
708 // We are finishing up the current cycle by clearing the next
709 // marking bitmap and getting it ready for the next cycle. During
710 // this time no other cycle can start. So, let's make sure that this
711 // is the case.
712 guarantee(!g1h->collector_state()->mark_in_progress(), "invariant");
713
714 ClearBitmapHRClosure cl(this, _nextMarkBitMap, true /* may_yield */);
715 ParClearNextMarkBitmapTask task(&cl, parallel_marking_threads(), true);
716 _parallel_workers->run_task(&task);
717
718 // Clear the liveness counting data. If the marking has been aborted, the abort()
719 // call already did that.
720 if (cl.complete()) {
721 clear_all_count_data();
722 }
723
724 // Repeat the asserts from above.
725 guarantee(cmThread()->during_cycle(), "invariant");
726 guarantee(!g1h->collector_state()->mark_in_progress(), "invariant");
727 }
728
729 class CheckBitmapClearHRClosure : public HeapRegionClosure {
730 CMBitMap* _bitmap;
731 bool _error;
732 public:
733 CheckBitmapClearHRClosure(CMBitMap* bitmap) : _bitmap(bitmap) {
734 }
735
736 virtual bool doHeapRegion(HeapRegion* r) {
737 // This closure can be called concurrently to the mutator, so we must make sure
738 // that the result of the getNextMarkedWordAddress() call is compared to the
739 // value passed to it as limit to detect any found bits.
740 // end never changes in G1.
741 HeapWord* end = r->end();
742 return _bitmap->getNextMarkedWordAddress(r->bottom(), end) != end;
743 }
744 };
745
746 bool ConcurrentMark::nextMarkBitmapIsClear() {
747 CheckBitmapClearHRClosure cl(_nextMarkBitMap);
748 _g1h->heap_region_iterate(&cl);
749 return cl.complete();
750 }
751
752 class NoteStartOfMarkHRClosure: public HeapRegionClosure {
753 public:
754 bool doHeapRegion(HeapRegion* r) {
755 r->note_start_of_marking();
756 return false;
757 }
758 };
759
760 void ConcurrentMark::checkpointRootsInitialPre() {
761 G1CollectedHeap* g1h = G1CollectedHeap::heap();
762 G1CollectorPolicy* g1p = g1h->g1_policy();
763
764 _has_aborted = false;
765
766 // Initialize marking structures. This has to be done in a STW phase.
767 reset();
768
769 // For each region note start of marking.
770 NoteStartOfMarkHRClosure startcl;
771 g1h->heap_region_iterate(&startcl);
772 }
773
774
775 void ConcurrentMark::checkpointRootsInitialPost() {
776 G1CollectedHeap* g1h = G1CollectedHeap::heap();
777
778 // Start Concurrent Marking weak-reference discovery.
779 ReferenceProcessor* rp = g1h->ref_processor_cm();
780 // enable ("weak") refs discovery
781 rp->enable_discovery();
782 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle
783
784 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
785 // This is the start of the marking cycle, we're expected all
786 // threads to have SATB queues with active set to false.
787 satb_mq_set.set_active_all_threads(true, /* new active value */
788 false /* expected_active */);
789
790 _root_regions.prepare_for_scan();
791
792 // update_g1_committed() will be called at the end of an evac pause
793 // when marking is on. So, it's also called at the end of the
794 // initial-mark pause to update the heap end, if the heap expands
795 // during it. No need to call it here.
796 }
797
798 /*
799 * Notice that in the next two methods, we actually leave the STS
800 * during the barrier sync and join it immediately afterwards. If we
801 * do not do this, the following deadlock can occur: one thread could
802 * be in the barrier sync code, waiting for the other thread to also
803 * sync up, whereas another one could be trying to yield, while also
804 * waiting for the other threads to sync up too.
805 *
806 * Note, however, that this code is also used during remark and in
807 * this case we should not attempt to leave / enter the STS, otherwise
808 * we'll either hit an assert (debug / fastdebug) or deadlock
809 * (product). So we should only leave / enter the STS if we are
810 * operating concurrently.
811 *
812 * Because the thread that does the sync barrier has left the STS, it
813 * is possible to be suspended for a Full GC or an evacuation pause
814 * could occur. This is actually safe, since the entering the sync
815 * barrier is one of the last things do_marking_step() does, and it
816 * doesn't manipulate any data structures afterwards.
817 */
818
819 void ConcurrentMark::enter_first_sync_barrier(uint worker_id) {
820 bool barrier_aborted;
821 {
822 SuspendibleThreadSetLeaver sts_leave(concurrent());
823 barrier_aborted = !_first_overflow_barrier_sync.enter();
824 }
825
826 // at this point everyone should have synced up and not be doing any
827 // more work
828
829 if (barrier_aborted) {
830 // If the barrier aborted we ignore the overflow condition and
831 // just abort the whole marking phase as quickly as possible.
832 return;
833 }
834
835 // If we're executing the concurrent phase of marking, reset the marking
836 // state; otherwise the marking state is reset after reference processing,
837 // during the remark pause.
838 // If we reset here as a result of an overflow during the remark we will
839 // see assertion failures from any subsequent set_concurrency_and_phase()
840 // calls.
841 if (concurrent()) {
842 // let the task associated with with worker 0 do this
843 if (worker_id == 0) {
844 // task 0 is responsible for clearing the global data structures
845 // We should be here because of an overflow. During STW we should
846 // not clear the overflow flag since we rely on it being true when
847 // we exit this method to abort the pause and restart concurrent
848 // marking.
849 reset_marking_state(true /* clear_overflow */);
850
851 if (G1Log::fine()) {
852 gclog_or_tty->gclog_stamp();
853 gclog_or_tty->print_cr("[GC concurrent-mark-reset-for-overflow]");
854 }
855 }
856 }
857
858 // after this, each task should reset its own data structures then
859 // then go into the second barrier
860 }
861
862 void ConcurrentMark::enter_second_sync_barrier(uint worker_id) {
863 SuspendibleThreadSetLeaver sts_leave(concurrent());
864 _second_overflow_barrier_sync.enter();
865
866 // at this point everything should be re-initialized and ready to go
867 }
868
869 class CMConcurrentMarkingTask: public AbstractGangTask {
870 private:
871 ConcurrentMark* _cm;
872 ConcurrentMarkThread* _cmt;
873
874 public:
875 void work(uint worker_id) {
876 assert(Thread::current()->is_ConcurrentGC_thread(),
877 "this should only be done by a conc GC thread");
878 ResourceMark rm;
879
880 double start_vtime = os::elapsedVTime();
881
882 {
883 SuspendibleThreadSetJoiner sts_join;
884
885 assert(worker_id < _cm->active_tasks(), "invariant");
886 CMTask* the_task = _cm->task(worker_id);
887 the_task->record_start_time();
888 if (!_cm->has_aborted()) {
889 do {
890 double start_vtime_sec = os::elapsedVTime();
891 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
892
893 the_task->do_marking_step(mark_step_duration_ms,
894 true /* do_termination */,
895 false /* is_serial*/);
896
897 double end_vtime_sec = os::elapsedVTime();
898 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec;
899 _cm->clear_has_overflown();
900
901 _cm->do_yield_check(worker_id);
902
903 jlong sleep_time_ms;
904 if (!_cm->has_aborted() && the_task->has_aborted()) {
905 sleep_time_ms =
906 (jlong) (elapsed_vtime_sec * _cm->sleep_factor() * 1000.0);
907 {
908 SuspendibleThreadSetLeaver sts_leave;
909 os::sleep(Thread::current(), sleep_time_ms, false);
910 }
911 }
912 } while (!_cm->has_aborted() && the_task->has_aborted());
913 }
914 the_task->record_end_time();
915 guarantee(!the_task->has_aborted() || _cm->has_aborted(), "invariant");
916 }
917
918 double end_vtime = os::elapsedVTime();
919 _cm->update_accum_task_vtime(worker_id, end_vtime - start_vtime);
920 }
921
922 CMConcurrentMarkingTask(ConcurrentMark* cm,
923 ConcurrentMarkThread* cmt) :
924 AbstractGangTask("Concurrent Mark"), _cm(cm), _cmt(cmt) { }
925
926 ~CMConcurrentMarkingTask() { }
927 };
928
929 // Calculates the number of active workers for a concurrent
930 // phase.
931 uint ConcurrentMark::calc_parallel_marking_threads() {
932 uint n_conc_workers = 0;
933 if (!UseDynamicNumberOfGCThreads ||
934 (!FLAG_IS_DEFAULT(ConcGCThreads) &&
935 !ForceDynamicNumberOfGCThreads)) {
936 n_conc_workers = max_parallel_marking_threads();
937 } else {
938 n_conc_workers =
939 AdaptiveSizePolicy::calc_default_active_workers(
940 max_parallel_marking_threads(),
941 1, /* Minimum workers */
942 parallel_marking_threads(),
943 Threads::number_of_non_daemon_threads());
944 // Don't scale down "n_conc_workers" by scale_parallel_threads() because
945 // that scaling has already gone into "_max_parallel_marking_threads".
946 }
947 assert(n_conc_workers > 0, "Always need at least 1");
948 return n_conc_workers;
949 }
950
951 void ConcurrentMark::scanRootRegion(HeapRegion* hr, uint worker_id) {
952 // Currently, only survivors can be root regions.
953 assert(hr->next_top_at_mark_start() == hr->bottom(), "invariant");
954 G1RootRegionScanClosure cl(_g1h, this, worker_id);
955
956 const uintx interval = PrefetchScanIntervalInBytes;
957 HeapWord* curr = hr->bottom();
958 const HeapWord* end = hr->top();
959 while (curr < end) {
960 Prefetch::read(curr, interval);
961 oop obj = oop(curr);
962 int size = obj->oop_iterate_size(&cl);
963 assert(size == obj->size(), "sanity");
964 curr += size;
965 }
966 }
967
968 class CMRootRegionScanTask : public AbstractGangTask {
969 private:
970 ConcurrentMark* _cm;
971
972 public:
973 CMRootRegionScanTask(ConcurrentMark* cm) :
974 AbstractGangTask("Root Region Scan"), _cm(cm) { }
975
976 void work(uint worker_id) {
977 assert(Thread::current()->is_ConcurrentGC_thread(),
978 "this should only be done by a conc GC thread");
979
980 CMRootRegions* root_regions = _cm->root_regions();
981 HeapRegion* hr = root_regions->claim_next();
982 while (hr != NULL) {
983 _cm->scanRootRegion(hr, worker_id);
984 hr = root_regions->claim_next();
985 }
986 }
987 };
988
989 void ConcurrentMark::scanRootRegions() {
990 double scan_start = os::elapsedTime();
991
992 // Start of concurrent marking.
993 ClassLoaderDataGraph::clear_claimed_marks();
994
995 // scan_in_progress() will have been set to true only if there was
996 // at least one root region to scan. So, if it's false, we
997 // should not attempt to do any further work.
998 if (root_regions()->scan_in_progress()) {
999 if (G1Log::fine()) {
1000 gclog_or_tty->gclog_stamp();
1001 gclog_or_tty->print_cr("[GC concurrent-root-region-scan-start]");
1002 }
1003
1004 _parallel_marking_threads = calc_parallel_marking_threads();
1005 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
1006 "Maximum number of marking threads exceeded");
1007 uint active_workers = MAX2(1U, parallel_marking_threads());
1008
1009 CMRootRegionScanTask task(this);
1010 _parallel_workers->set_active_workers(active_workers);
1011 _parallel_workers->run_task(&task);
1012
1013 if (G1Log::fine()) {
1014 gclog_or_tty->gclog_stamp();
1015 gclog_or_tty->print_cr("[GC concurrent-root-region-scan-end, %1.7lf secs]", os::elapsedTime() - scan_start);
1016 }
1017
1018 // It's possible that has_aborted() is true here without actually
1019 // aborting the survivor scan earlier. This is OK as it's
1020 // mainly used for sanity checking.
1021 root_regions()->scan_finished();
1022 }
1023 }
1024
1025 void ConcurrentMark::markFromRoots() {
1026 // we might be tempted to assert that:
1027 // assert(asynch == !SafepointSynchronize::is_at_safepoint(),
1028 // "inconsistent argument?");
1029 // However that wouldn't be right, because it's possible that
1030 // a safepoint is indeed in progress as a younger generation
1031 // stop-the-world GC happens even as we mark in this generation.
1032
1033 _restart_for_overflow = false;
1034
1035 // _g1h has _n_par_threads
1036 _parallel_marking_threads = calc_parallel_marking_threads();
1037 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
1038 "Maximum number of marking threads exceeded");
1039
1040 uint active_workers = MAX2(1U, parallel_marking_threads());
1041 assert(active_workers > 0, "Should have been set");
1042
1043 // Parallel task terminator is set in "set_concurrency_and_phase()"
1044 set_concurrency_and_phase(active_workers, true /* concurrent */);
1045
1046 CMConcurrentMarkingTask markingTask(this, cmThread());
1047 _parallel_workers->set_active_workers(active_workers);
1048 _parallel_workers->run_task(&markingTask);
1049 print_stats();
1050 }
1051
1052 // Helper class to get rid of some boilerplate code.
1053 class G1CMTraceTime : public StackObj {
1054 GCTraceTimeImpl _gc_trace_time;
1055 static bool doit_and_prepend(bool doit) {
1056 if (doit) {
1057 gclog_or_tty->put(' ');
1058 }
1059 return doit;
1060 }
1061
1062 public:
1063 G1CMTraceTime(const char* title, bool doit)
1064 : _gc_trace_time(title, doit_and_prepend(doit), false, G1CollectedHeap::heap()->gc_timer_cm()) {
1065 }
1066 };
1067
1068 void ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) {
1069 // world is stopped at this checkpoint
1070 assert(SafepointSynchronize::is_at_safepoint(),
1071 "world should be stopped");
1072
1073 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1074
1075 // If a full collection has happened, we shouldn't do this.
1076 if (has_aborted()) {
1077 g1h->collector_state()->set_mark_in_progress(false); // So bitmap clearing isn't confused
1078 return;
1079 }
1080
1081 SvcGCMarker sgcm(SvcGCMarker::OTHER);
1082
1083 if (VerifyDuringGC) {
1084 HandleMark hm; // handle scope
1085 g1h->prepare_for_verify();
1086 Universe::verify(VerifyOption_G1UsePrevMarking,
1087 " VerifyDuringGC:(before)");
1088 }
1089 g1h->check_bitmaps("Remark Start");
1090
1091 G1CollectorPolicy* g1p = g1h->g1_policy();
1092 g1p->record_concurrent_mark_remark_start();
1093
1094 double start = os::elapsedTime();
1095
1096 checkpointRootsFinalWork();
1097
1098 double mark_work_end = os::elapsedTime();
1099
1100 weakRefsWork(clear_all_soft_refs);
1101
1102 if (has_overflown()) {
1103 // Oops. We overflowed. Restart concurrent marking.
1104 _restart_for_overflow = true;
1105 if (G1TraceMarkStackOverflow) {
1106 gclog_or_tty->print_cr("\nRemark led to restart for overflow.");
1107 }
1108
1109 // Verify the heap w.r.t. the previous marking bitmap.
1110 if (VerifyDuringGC) {
1111 HandleMark hm; // handle scope
1112 g1h->prepare_for_verify();
1113 Universe::verify(VerifyOption_G1UsePrevMarking,
1114 " VerifyDuringGC:(overflow)");
1115 }
1116
1117 // Clear the marking state because we will be restarting
1118 // marking due to overflowing the global mark stack.
1119 reset_marking_state();
1120 } else {
1121 {
1122 G1CMTraceTime trace("GC aggregate-data", G1Log::finer());
1123
1124 // Aggregate the per-task counting data that we have accumulated
1125 // while marking.
1126 aggregate_count_data();
1127 }
1128
1129 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1130 // We're done with marking.
1131 // This is the end of the marking cycle, we're expected all
1132 // threads to have SATB queues with active set to true.
1133 satb_mq_set.set_active_all_threads(false, /* new active value */
1134 true /* expected_active */);
1135
1136 if (VerifyDuringGC) {
1137 HandleMark hm; // handle scope
1138 g1h->prepare_for_verify();
1139 Universe::verify(VerifyOption_G1UseNextMarking,
1140 " VerifyDuringGC:(after)");
1141 }
1142 g1h->check_bitmaps("Remark End");
1143 assert(!restart_for_overflow(), "sanity");
1144 // Completely reset the marking state since marking completed
1145 set_non_marking_state();
1146 }
1147
1148 // Expand the marking stack, if we have to and if we can.
1149 if (_markStack.should_expand()) {
1150 _markStack.expand();
1151 }
1152
1153 // Statistics
1154 double now = os::elapsedTime();
1155 _remark_mark_times.add((mark_work_end - start) * 1000.0);
1156 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0);
1157 _remark_times.add((now - start) * 1000.0);
1158
1159 g1p->record_concurrent_mark_remark_end();
1160
1161 G1CMIsAliveClosure is_alive(g1h);
1162 g1h->gc_tracer_cm()->report_object_count_after_gc(&is_alive);
1163 }
1164
1165 // Base class of the closures that finalize and verify the
1166 // liveness counting data.
1167 class CMCountDataClosureBase: public HeapRegionClosure {
1168 protected:
1169 G1CollectedHeap* _g1h;
1170 ConcurrentMark* _cm;
1171 CardTableModRefBS* _ct_bs;
1172
1173 BitMap* _region_bm;
1174 BitMap* _card_bm;
1175
1176 // Takes a region that's not empty (i.e., it has at least one
1177 // live object in it and sets its corresponding bit on the region
1178 // bitmap to 1.
1179 void set_bit_for_region(HeapRegion* hr) {
1180 BitMap::idx_t index = (BitMap::idx_t) hr->hrm_index();
1181 _region_bm->par_at_put(index, true);
1182 }
1183
1184 public:
1185 CMCountDataClosureBase(G1CollectedHeap* g1h,
1186 BitMap* region_bm, BitMap* card_bm):
1187 _g1h(g1h), _cm(g1h->concurrent_mark()),
1188 _ct_bs(barrier_set_cast<CardTableModRefBS>(g1h->barrier_set())),
1189 _region_bm(region_bm), _card_bm(card_bm) { }
1190 };
1191
1192 // Closure that calculates the # live objects per region. Used
1193 // for verification purposes during the cleanup pause.
1194 class CalcLiveObjectsClosure: public CMCountDataClosureBase {
1195 CMBitMapRO* _bm;
1196 size_t _region_marked_bytes;
1197
1198 public:
1199 CalcLiveObjectsClosure(CMBitMapRO *bm, G1CollectedHeap* g1h,
1200 BitMap* region_bm, BitMap* card_bm) :
1201 CMCountDataClosureBase(g1h, region_bm, card_bm),
1202 _bm(bm), _region_marked_bytes(0) { }
1203
1204 bool doHeapRegion(HeapRegion* hr) {
1205 HeapWord* ntams = hr->next_top_at_mark_start();
1206 HeapWord* start = hr->bottom();
1207
1208 assert(start <= hr->end() && start <= ntams && ntams <= hr->end(),
1209 "Preconditions not met - "
1210 "start: " PTR_FORMAT ", ntams: " PTR_FORMAT ", end: " PTR_FORMAT,
1211 p2i(start), p2i(ntams), p2i(hr->end()));
1212
1213 // Find the first marked object at or after "start".
1214 start = _bm->getNextMarkedWordAddress(start, ntams);
1215
1216 size_t marked_bytes = 0;
1217
1218 while (start < ntams) {
1219 oop obj = oop(start);
1220 int obj_sz = obj->size();
1221 HeapWord* obj_end = start + obj_sz;
1222
1223 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start);
1224 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(obj_end);
1225
1226 // Note: if we're looking at the last region in heap - obj_end
1227 // could be actually just beyond the end of the heap; end_idx
1228 // will then correspond to a (non-existent) card that is also
1229 // just beyond the heap.
1230 if (_g1h->is_in_g1_reserved(obj_end) && !_ct_bs->is_card_aligned(obj_end)) {
1231 // end of object is not card aligned - increment to cover
1232 // all the cards spanned by the object
1233 end_idx += 1;
1234 }
1235
1236 // Set the bits in the card BM for the cards spanned by this object.
1237 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1238
1239 // Add the size of this object to the number of marked bytes.
1240 marked_bytes += (size_t)obj_sz * HeapWordSize;
1241
1242 // This will happen if we are handling a humongous object that spans
1243 // several heap regions.
1244 if (obj_end > hr->end()) {
1245 break;
1246 }
1247 // Find the next marked object after this one.
1248 start = _bm->getNextMarkedWordAddress(obj_end, ntams);
1249 }
1250
1251 // Mark the allocated-since-marking portion...
1252 HeapWord* top = hr->top();
1253 if (ntams < top) {
1254 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams);
1255 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top);
1256
1257 // Note: if we're looking at the last region in heap - top
1258 // could be actually just beyond the end of the heap; end_idx
1259 // will then correspond to a (non-existent) card that is also
1260 // just beyond the heap.
1261 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) {
1262 // end of object is not card aligned - increment to cover
1263 // all the cards spanned by the object
1264 end_idx += 1;
1265 }
1266 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1267
1268 // This definitely means the region has live objects.
1269 set_bit_for_region(hr);
1270 }
1271
1272 // Update the live region bitmap.
1273 if (marked_bytes > 0) {
1274 set_bit_for_region(hr);
1275 }
1276
1277 // Set the marked bytes for the current region so that
1278 // it can be queried by a calling verification routine
1279 _region_marked_bytes = marked_bytes;
1280
1281 return false;
1282 }
1283
1284 size_t region_marked_bytes() const { return _region_marked_bytes; }
1285 };
1286
1287 // Heap region closure used for verifying the counting data
1288 // that was accumulated concurrently and aggregated during
1289 // the remark pause. This closure is applied to the heap
1290 // regions during the STW cleanup pause.
1291
1292 class VerifyLiveObjectDataHRClosure: public HeapRegionClosure {
1293 G1CollectedHeap* _g1h;
1294 ConcurrentMark* _cm;
1295 CalcLiveObjectsClosure _calc_cl;
1296 BitMap* _region_bm; // Region BM to be verified
1297 BitMap* _card_bm; // Card BM to be verified
1298
1299 BitMap* _exp_region_bm; // Expected Region BM values
1300 BitMap* _exp_card_bm; // Expected card BM values
1301
1302 int _failures;
1303
1304 public:
1305 VerifyLiveObjectDataHRClosure(G1CollectedHeap* g1h,
1306 BitMap* region_bm,
1307 BitMap* card_bm,
1308 BitMap* exp_region_bm,
1309 BitMap* exp_card_bm) :
1310 _g1h(g1h), _cm(g1h->concurrent_mark()),
1311 _calc_cl(_cm->nextMarkBitMap(), g1h, exp_region_bm, exp_card_bm),
1312 _region_bm(region_bm), _card_bm(card_bm),
1313 _exp_region_bm(exp_region_bm), _exp_card_bm(exp_card_bm),
1314 _failures(0) { }
1315
1316 int failures() const { return _failures; }
1317
1318 bool doHeapRegion(HeapRegion* hr) {
1319 int failures = 0;
1320
1321 // Call the CalcLiveObjectsClosure to walk the marking bitmap for
1322 // this region and set the corresponding bits in the expected region
1323 // and card bitmaps.
1324 bool res = _calc_cl.doHeapRegion(hr);
1325 assert(res == false, "should be continuing");
1326
1327 // Verify the marked bytes for this region.
1328 size_t exp_marked_bytes = _calc_cl.region_marked_bytes();
1329 size_t act_marked_bytes = hr->next_marked_bytes();
1330
1331 if (exp_marked_bytes > act_marked_bytes) {
1332 if (hr->is_starts_humongous()) {
1333 // For start_humongous regions, the size of the whole object will be
1334 // in exp_marked_bytes.
1335 HeapRegion* region = hr;
1336 int num_regions;
1337 for (num_regions = 0; region != NULL; num_regions++) {
1338 region = _g1h->next_region_in_humongous(region);
1339 }
1340 if ((num_regions-1) * HeapRegion::GrainBytes >= exp_marked_bytes) {
1341 failures += 1;
1342 } else if (num_regions * HeapRegion::GrainBytes < exp_marked_bytes) {
1343 failures += 1;
1344 }
1345 } else {
1346 // We're not OK if expected marked bytes > actual marked bytes. It means
1347 // we have missed accounting some objects during the actual marking.
1348 failures += 1;
1349 }
1350 }
1351
1352 // Verify the bit, for this region, in the actual and expected
1353 // (which was just calculated) region bit maps.
1354 // We're not OK if the bit in the calculated expected region
1355 // bitmap is set and the bit in the actual region bitmap is not.
1356 BitMap::idx_t index = (BitMap::idx_t) hr->hrm_index();
1357
1358 bool expected = _exp_region_bm->at(index);
1359 bool actual = _region_bm->at(index);
1360 if (expected && !actual) {
1361 failures += 1;
1362 }
1363
1364 // Verify that the card bit maps for the cards spanned by the current
1365 // region match. We have an error if we have a set bit in the expected
1366 // bit map and the corresponding bit in the actual bitmap is not set.
1367
1368 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(hr->bottom());
1369 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(hr->top());
1370
1371 for (BitMap::idx_t i = start_idx; i < end_idx; i+=1) {
1372 expected = _exp_card_bm->at(i);
1373 actual = _card_bm->at(i);
1374
1375 if (expected && !actual) {
1376 failures += 1;
1377 }
1378 }
1379
1380 _failures += failures;
1381
1382 // We could stop iteration over the heap when we
1383 // find the first violating region by returning true.
1384 return false;
1385 }
1386 };
1387
1388 class G1ParVerifyFinalCountTask: public AbstractGangTask {
1389 protected:
1390 G1CollectedHeap* _g1h;
1391 ConcurrentMark* _cm;
1392 BitMap* _actual_region_bm;
1393 BitMap* _actual_card_bm;
1394
1395 uint _n_workers;
1396
1397 BitMap* _expected_region_bm;
1398 BitMap* _expected_card_bm;
1399
1400 int _failures;
1401
1402 HeapRegionClaimer _hrclaimer;
1403
1404 public:
1405 G1ParVerifyFinalCountTask(G1CollectedHeap* g1h,
1406 BitMap* region_bm, BitMap* card_bm,
1407 BitMap* expected_region_bm, BitMap* expected_card_bm)
1408 : AbstractGangTask("G1 verify final counting"),
1409 _g1h(g1h), _cm(_g1h->concurrent_mark()),
1410 _actual_region_bm(region_bm), _actual_card_bm(card_bm),
1411 _expected_region_bm(expected_region_bm), _expected_card_bm(expected_card_bm),
1412 _failures(0),
1413 _n_workers(_g1h->workers()->active_workers()), _hrclaimer(_n_workers) {
1414 assert(VerifyDuringGC, "don't call this otherwise");
1415 assert(_expected_card_bm->size() == _actual_card_bm->size(), "sanity");
1416 assert(_expected_region_bm->size() == _actual_region_bm->size(), "sanity");
1417 }
1418
1419 void work(uint worker_id) {
1420 assert(worker_id < _n_workers, "invariant");
1421
1422 VerifyLiveObjectDataHRClosure verify_cl(_g1h,
1423 _actual_region_bm, _actual_card_bm,
1424 _expected_region_bm,
1425 _expected_card_bm);
1426
1427 _g1h->heap_region_par_iterate(&verify_cl, worker_id, &_hrclaimer);
1428
1429 Atomic::add(verify_cl.failures(), &_failures);
1430 }
1431
1432 int failures() const { return _failures; }
1433 };
1434
1435 // Closure that finalizes the liveness counting data.
1436 // Used during the cleanup pause.
1437 // Sets the bits corresponding to the interval [NTAMS, top]
1438 // (which contains the implicitly live objects) in the
1439 // card liveness bitmap. Also sets the bit for each region,
1440 // containing live data, in the region liveness bitmap.
1441
1442 class FinalCountDataUpdateClosure: public CMCountDataClosureBase {
1443 public:
1444 FinalCountDataUpdateClosure(G1CollectedHeap* g1h,
1445 BitMap* region_bm,
1446 BitMap* card_bm) :
1447 CMCountDataClosureBase(g1h, region_bm, card_bm) { }
1448
1449 bool doHeapRegion(HeapRegion* hr) {
1450 HeapWord* ntams = hr->next_top_at_mark_start();
1451 HeapWord* top = hr->top();
1452
1453 assert(hr->bottom() <= ntams && ntams <= hr->end(), "Preconditions.");
1454
1455 // Mark the allocated-since-marking portion...
1456 if (ntams < top) {
1457 // This definitely means the region has live objects.
1458 set_bit_for_region(hr);
1459
1460 // Now set the bits in the card bitmap for [ntams, top)
1461 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams);
1462 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top);
1463
1464 // Note: if we're looking at the last region in heap - top
1465 // could be actually just beyond the end of the heap; end_idx
1466 // will then correspond to a (non-existent) card that is also
1467 // just beyond the heap.
1468 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) {
1469 // end of object is not card aligned - increment to cover
1470 // all the cards spanned by the object
1471 end_idx += 1;
1472 }
1473
1474 assert(end_idx <= _card_bm->size(),
1475 "oob: end_idx= " SIZE_FORMAT ", bitmap size= " SIZE_FORMAT,
1476 end_idx, _card_bm->size());
1477 assert(start_idx < _card_bm->size(),
1478 "oob: start_idx= " SIZE_FORMAT ", bitmap size= " SIZE_FORMAT,
1479 start_idx, _card_bm->size());
1480
1481 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1482 }
1483
1484 // Set the bit for the region if it contains live data
1485 if (hr->next_marked_bytes() > 0) {
1486 set_bit_for_region(hr);
1487 }
1488
1489 return false;
1490 }
1491 };
1492
1493 class G1ParFinalCountTask: public AbstractGangTask {
1494 protected:
1495 G1CollectedHeap* _g1h;
1496 ConcurrentMark* _cm;
1497 BitMap* _actual_region_bm;
1498 BitMap* _actual_card_bm;
1499
1500 uint _n_workers;
1501 HeapRegionClaimer _hrclaimer;
1502
1503 public:
1504 G1ParFinalCountTask(G1CollectedHeap* g1h, BitMap* region_bm, BitMap* card_bm)
1505 : AbstractGangTask("G1 final counting"),
1506 _g1h(g1h), _cm(_g1h->concurrent_mark()),
1507 _actual_region_bm(region_bm), _actual_card_bm(card_bm),
1508 _n_workers(_g1h->workers()->active_workers()), _hrclaimer(_n_workers) {
1509 }
1510
1511 void work(uint worker_id) {
1512 assert(worker_id < _n_workers, "invariant");
1513
1514 FinalCountDataUpdateClosure final_update_cl(_g1h,
1515 _actual_region_bm,
1516 _actual_card_bm);
1517
1518 _g1h->heap_region_par_iterate(&final_update_cl, worker_id, &_hrclaimer);
1519 }
1520 };
1521
1522 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure {
1523 G1CollectedHeap* _g1;
1524 size_t _freed_bytes;
1525 FreeRegionList* _local_cleanup_list;
1526 HeapRegionSetCount _old_regions_removed;
1527 HeapRegionSetCount _humongous_regions_removed;
1528 HRRSCleanupTask* _hrrs_cleanup_task;
1529
1530 public:
1531 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1,
1532 FreeRegionList* local_cleanup_list,
1533 HRRSCleanupTask* hrrs_cleanup_task) :
1534 _g1(g1),
1535 _freed_bytes(0),
1536 _local_cleanup_list(local_cleanup_list),
1537 _old_regions_removed(),
1538 _humongous_regions_removed(),
1539 _hrrs_cleanup_task(hrrs_cleanup_task) { }
1540
1541 size_t freed_bytes() { return _freed_bytes; }
1542 const HeapRegionSetCount& old_regions_removed() { return _old_regions_removed; }
1543 const HeapRegionSetCount& humongous_regions_removed() { return _humongous_regions_removed; }
1544
1545 bool doHeapRegion(HeapRegion *hr) {
1546 if (hr->is_archive()) {
1547 return false;
1548 }
1549 // We use a claim value of zero here because all regions
1550 // were claimed with value 1 in the FinalCount task.
1551 _g1->reset_gc_time_stamps(hr);
1552 hr->note_end_of_marking();
1553
1554 if (hr->used() > 0 && hr->max_live_bytes() == 0 && !hr->is_young()) {
1555 _freed_bytes += hr->used();
1556 hr->set_containing_set(NULL);
1557 if (hr->is_humongous()) {
1558 _humongous_regions_removed.increment(1u, hr->capacity());
1559 _g1->free_humongous_region(hr, _local_cleanup_list, true);
1560 } else {
1561 _old_regions_removed.increment(1u, hr->capacity());
1562 _g1->free_region(hr, _local_cleanup_list, true);
1563 }
1564 } else {
1565 hr->rem_set()->do_cleanup_work(_hrrs_cleanup_task);
1566 }
1567
1568 return false;
1569 }
1570 };
1571
1572 class G1ParNoteEndTask: public AbstractGangTask {
1573 friend class G1NoteEndOfConcMarkClosure;
1574
1575 protected:
1576 G1CollectedHeap* _g1h;
1577 FreeRegionList* _cleanup_list;
1578 HeapRegionClaimer _hrclaimer;
1579
1580 public:
1581 G1ParNoteEndTask(G1CollectedHeap* g1h, FreeRegionList* cleanup_list, uint n_workers) :
1582 AbstractGangTask("G1 note end"), _g1h(g1h), _cleanup_list(cleanup_list), _hrclaimer(n_workers) {
1583 }
1584
1585 void work(uint worker_id) {
1586 FreeRegionList local_cleanup_list("Local Cleanup List");
1587 HRRSCleanupTask hrrs_cleanup_task;
1588 G1NoteEndOfConcMarkClosure g1_note_end(_g1h, &local_cleanup_list,
1589 &hrrs_cleanup_task);
1590 _g1h->heap_region_par_iterate(&g1_note_end, worker_id, &_hrclaimer);
1591 assert(g1_note_end.complete(), "Shouldn't have yielded!");
1592
1593 // Now update the lists
1594 _g1h->remove_from_old_sets(g1_note_end.old_regions_removed(), g1_note_end.humongous_regions_removed());
1595 {
1596 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
1597 _g1h->decrement_summary_bytes(g1_note_end.freed_bytes());
1598
1599 // If we iterate over the global cleanup list at the end of
1600 // cleanup to do this printing we will not guarantee to only
1601 // generate output for the newly-reclaimed regions (the list
1602 // might not be empty at the beginning of cleanup; we might
1603 // still be working on its previous contents). So we do the
1604 // printing here, before we append the new regions to the global
1605 // cleanup list.
1606
1607 G1HRPrinter* hr_printer = _g1h->hr_printer();
1608 if (hr_printer->is_active()) {
1609 FreeRegionListIterator iter(&local_cleanup_list);
1610 while (iter.more_available()) {
1611 HeapRegion* hr = iter.get_next();
1612 hr_printer->cleanup(hr);
1613 }
1614 }
1615
1616 _cleanup_list->add_ordered(&local_cleanup_list);
1617 assert(local_cleanup_list.is_empty(), "post-condition");
1618
1619 HeapRegionRemSet::finish_cleanup_task(&hrrs_cleanup_task);
1620 }
1621 }
1622 };
1623
1624 class G1ParScrubRemSetTask: public AbstractGangTask {
1625 protected:
1626 G1RemSet* _g1rs;
1627 BitMap* _region_bm;
1628 BitMap* _card_bm;
1629 HeapRegionClaimer _hrclaimer;
1630
1631 public:
1632 G1ParScrubRemSetTask(G1CollectedHeap* g1h, BitMap* region_bm, BitMap* card_bm, uint n_workers) :
1633 AbstractGangTask("G1 ScrubRS"), _g1rs(g1h->g1_rem_set()), _region_bm(region_bm), _card_bm(card_bm), _hrclaimer(n_workers) {
1634 }
1635
1636 void work(uint worker_id) {
1637 _g1rs->scrub(_region_bm, _card_bm, worker_id, &_hrclaimer);
1638 }
1639
1640 };
1641
1642 void ConcurrentMark::cleanup() {
1643 // world is stopped at this checkpoint
1644 assert(SafepointSynchronize::is_at_safepoint(),
1645 "world should be stopped");
1646 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1647
1648 // If a full collection has happened, we shouldn't do this.
1649 if (has_aborted()) {
1650 g1h->collector_state()->set_mark_in_progress(false); // So bitmap clearing isn't confused
1651 return;
1652 }
1653
1654 g1h->verify_region_sets_optional();
1655
1656 if (VerifyDuringGC) {
1657 HandleMark hm; // handle scope
1658 g1h->prepare_for_verify();
1659 Universe::verify(VerifyOption_G1UsePrevMarking,
1660 " VerifyDuringGC:(before)");
1661 }
1662 g1h->check_bitmaps("Cleanup Start");
1663
1664 G1CollectorPolicy* g1p = g1h->g1_policy();
1665 g1p->record_concurrent_mark_cleanup_start();
1666
1667 double start = os::elapsedTime();
1668
1669 HeapRegionRemSet::reset_for_cleanup_tasks();
1670
1671 // Do counting once more with the world stopped for good measure.
1672 G1ParFinalCountTask g1_par_count_task(g1h, &_region_bm, &_card_bm);
1673
1674 g1h->workers()->run_task(&g1_par_count_task);
1675
1676 if (VerifyDuringGC) {
1677 // Verify that the counting data accumulated during marking matches
1678 // that calculated by walking the marking bitmap.
1679
1680 // Bitmaps to hold expected values
1681 BitMap expected_region_bm(_region_bm.size(), true);
1682 BitMap expected_card_bm(_card_bm.size(), true);
1683
1684 G1ParVerifyFinalCountTask g1_par_verify_task(g1h,
1685 &_region_bm,
1686 &_card_bm,
1687 &expected_region_bm,
1688 &expected_card_bm);
1689
1690 g1h->workers()->run_task(&g1_par_verify_task);
1691
1692 guarantee(g1_par_verify_task.failures() == 0, "Unexpected accounting failures");
1693 }
1694
1695 size_t start_used_bytes = g1h->used();
1696 g1h->collector_state()->set_mark_in_progress(false);
1697
1698 double count_end = os::elapsedTime();
1699 double this_final_counting_time = (count_end - start);
1700 _total_counting_time += this_final_counting_time;
1701
1702 if (G1PrintRegionLivenessInfo) {
1703 G1PrintRegionLivenessInfoClosure cl(gclog_or_tty, "Post-Marking");
1704 _g1h->heap_region_iterate(&cl);
1705 }
1706
1707 // Install newly created mark bitMap as "prev".
1708 swapMarkBitMaps();
1709
1710 g1h->reset_gc_time_stamp();
1711
1712 uint n_workers = _g1h->workers()->active_workers();
1713
1714 // Note end of marking in all heap regions.
1715 G1ParNoteEndTask g1_par_note_end_task(g1h, &_cleanup_list, n_workers);
1716 g1h->workers()->run_task(&g1_par_note_end_task);
1717 g1h->check_gc_time_stamps();
1718
1719 if (!cleanup_list_is_empty()) {
1720 // The cleanup list is not empty, so we'll have to process it
1721 // concurrently. Notify anyone else that might be wanting free
1722 // regions that there will be more free regions coming soon.
1723 g1h->set_free_regions_coming();
1724 }
1725
1726 // call below, since it affects the metric by which we sort the heap
1727 // regions.
1728 if (G1ScrubRemSets) {
1729 double rs_scrub_start = os::elapsedTime();
1730 G1ParScrubRemSetTask g1_par_scrub_rs_task(g1h, &_region_bm, &_card_bm, n_workers);
1731 g1h->workers()->run_task(&g1_par_scrub_rs_task);
1732
1733 double rs_scrub_end = os::elapsedTime();
1734 double this_rs_scrub_time = (rs_scrub_end - rs_scrub_start);
1735 _total_rs_scrub_time += this_rs_scrub_time;
1736 }
1737
1738 // this will also free any regions totally full of garbage objects,
1739 // and sort the regions.
1740 g1h->g1_policy()->record_concurrent_mark_cleanup_end();
1741
1742 // Statistics.
1743 double end = os::elapsedTime();
1744 _cleanup_times.add((end - start) * 1000.0);
1745
1746 if (G1Log::fine()) {
1747 g1h->g1_policy()->print_heap_transition(start_used_bytes);
1748 }
1749
1750 // Clean up will have freed any regions completely full of garbage.
1751 // Update the soft reference policy with the new heap occupancy.
1752 Universe::update_heap_info_at_gc();
1753
1754 if (VerifyDuringGC) {
1755 HandleMark hm; // handle scope
1756 g1h->prepare_for_verify();
1757 Universe::verify(VerifyOption_G1UsePrevMarking,
1758 " VerifyDuringGC:(after)");
1759 }
1760
1761 g1h->check_bitmaps("Cleanup End");
1762
1763 g1h->verify_region_sets_optional();
1764
1765 // We need to make this be a "collection" so any collection pause that
1766 // races with it goes around and waits for completeCleanup to finish.
1767 g1h->increment_total_collections();
1768
1769 // Clean out dead classes and update Metaspace sizes.
1770 if (ClassUnloadingWithConcurrentMark) {
1771 ClassLoaderDataGraph::purge();
1772 }
1773 MetaspaceGC::compute_new_size();
1774
1775 // We reclaimed old regions so we should calculate the sizes to make
1776 // sure we update the old gen/space data.
1777 g1h->g1mm()->update_sizes();
1778 g1h->allocation_context_stats().update_after_mark();
1779
1780 g1h->trace_heap_after_concurrent_cycle();
1781 }
1782
1783 void ConcurrentMark::completeCleanup() {
1784 if (has_aborted()) return;
1785
1786 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1787
1788 _cleanup_list.verify_optional();
1789 FreeRegionList tmp_free_list("Tmp Free List");
1790
1791 if (G1ConcRegionFreeingVerbose) {
1792 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : "
1793 "cleanup list has %u entries",
1794 _cleanup_list.length());
1795 }
1796
1797 // No one else should be accessing the _cleanup_list at this point,
1798 // so it is not necessary to take any locks
1799 while (!_cleanup_list.is_empty()) {
1800 HeapRegion* hr = _cleanup_list.remove_region(true /* from_head */);
1801 assert(hr != NULL, "Got NULL from a non-empty list");
1802 hr->par_clear();
1803 tmp_free_list.add_ordered(hr);
1804
1805 // Instead of adding one region at a time to the secondary_free_list,
1806 // we accumulate them in the local list and move them a few at a
1807 // time. This also cuts down on the number of notify_all() calls
1808 // we do during this process. We'll also append the local list when
1809 // _cleanup_list is empty (which means we just removed the last
1810 // region from the _cleanup_list).
1811 if ((tmp_free_list.length() % G1SecondaryFreeListAppendLength == 0) ||
1812 _cleanup_list.is_empty()) {
1813 if (G1ConcRegionFreeingVerbose) {
1814 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : "
1815 "appending %u entries to the secondary_free_list, "
1816 "cleanup list still has %u entries",
1817 tmp_free_list.length(),
1818 _cleanup_list.length());
1819 }
1820
1821 {
1822 MutexLockerEx x(SecondaryFreeList_lock, Mutex::_no_safepoint_check_flag);
1823 g1h->secondary_free_list_add(&tmp_free_list);
1824 SecondaryFreeList_lock->notify_all();
1825 }
1826 #ifndef PRODUCT
1827 if (G1StressConcRegionFreeing) {
1828 for (uintx i = 0; i < G1StressConcRegionFreeingDelayMillis; ++i) {
1829 os::sleep(Thread::current(), (jlong) 1, false);
1830 }
1831 }
1832 #endif
1833 }
1834 }
1835 assert(tmp_free_list.is_empty(), "post-condition");
1836 }
1837
1838 // Supporting Object and Oop closures for reference discovery
1839 // and processing in during marking
1840
1841 bool G1CMIsAliveClosure::do_object_b(oop obj) {
1842 HeapWord* addr = (HeapWord*)obj;
1843 return addr != NULL &&
1844 (!_g1->is_in_g1_reserved(addr) || !_g1->is_obj_ill(obj));
1845 }
1846
1847 // 'Keep Alive' oop closure used by both serial parallel reference processing.
1848 // Uses the CMTask associated with a worker thread (for serial reference
1849 // processing the CMTask for worker 0 is used) to preserve (mark) and
1850 // trace referent objects.
1851 //
1852 // Using the CMTask and embedded local queues avoids having the worker
1853 // threads operating on the global mark stack. This reduces the risk
1854 // of overflowing the stack - which we would rather avoid at this late
1855 // state. Also using the tasks' local queues removes the potential
1856 // of the workers interfering with each other that could occur if
1857 // operating on the global stack.
1858
1859 class G1CMKeepAliveAndDrainClosure: public OopClosure {
1860 ConcurrentMark* _cm;
1861 CMTask* _task;
1862 int _ref_counter_limit;
1863 int _ref_counter;
1864 bool _is_serial;
1865 public:
1866 G1CMKeepAliveAndDrainClosure(ConcurrentMark* cm, CMTask* task, bool is_serial) :
1867 _cm(cm), _task(task), _is_serial(is_serial),
1868 _ref_counter_limit(G1RefProcDrainInterval) {
1869 assert(_ref_counter_limit > 0, "sanity");
1870 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
1871 _ref_counter = _ref_counter_limit;
1872 }
1873
1874 virtual void do_oop(narrowOop* p) { do_oop_work(p); }
1875 virtual void do_oop( oop* p) { do_oop_work(p); }
1876
1877 template <class T> void do_oop_work(T* p) {
1878 if (!_cm->has_overflown()) {
1879 oop obj = oopDesc::load_decode_heap_oop(p);
1880 _task->deal_with_reference(obj);
1881 _ref_counter--;
1882
1883 if (_ref_counter == 0) {
1884 // We have dealt with _ref_counter_limit references, pushing them
1885 // and objects reachable from them on to the local stack (and
1886 // possibly the global stack). Call CMTask::do_marking_step() to
1887 // process these entries.
1888 //
1889 // We call CMTask::do_marking_step() in a loop, which we'll exit if
1890 // there's nothing more to do (i.e. we're done with the entries that
1891 // were pushed as a result of the CMTask::deal_with_reference() calls
1892 // above) or we overflow.
1893 //
1894 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted()
1895 // flag while there may still be some work to do. (See the comment at
1896 // the beginning of CMTask::do_marking_step() for those conditions -
1897 // one of which is reaching the specified time target.) It is only
1898 // when CMTask::do_marking_step() returns without setting the
1899 // has_aborted() flag that the marking step has completed.
1900 do {
1901 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
1902 _task->do_marking_step(mark_step_duration_ms,
1903 false /* do_termination */,
1904 _is_serial);
1905 } while (_task->has_aborted() && !_cm->has_overflown());
1906 _ref_counter = _ref_counter_limit;
1907 }
1908 }
1909 }
1910 };
1911
1912 // 'Drain' oop closure used by both serial and parallel reference processing.
1913 // Uses the CMTask associated with a given worker thread (for serial
1914 // reference processing the CMtask for worker 0 is used). Calls the
1915 // do_marking_step routine, with an unbelievably large timeout value,
1916 // to drain the marking data structures of the remaining entries
1917 // added by the 'keep alive' oop closure above.
1918
1919 class G1CMDrainMarkingStackClosure: public VoidClosure {
1920 ConcurrentMark* _cm;
1921 CMTask* _task;
1922 bool _is_serial;
1923 public:
1924 G1CMDrainMarkingStackClosure(ConcurrentMark* cm, CMTask* task, bool is_serial) :
1925 _cm(cm), _task(task), _is_serial(is_serial) {
1926 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
1927 }
1928
1929 void do_void() {
1930 do {
1931 // We call CMTask::do_marking_step() to completely drain the local
1932 // and global marking stacks of entries pushed by the 'keep alive'
1933 // oop closure (an instance of G1CMKeepAliveAndDrainClosure above).
1934 //
1935 // CMTask::do_marking_step() is called in a loop, which we'll exit
1936 // if there's nothing more to do (i.e. we've completely drained the
1937 // entries that were pushed as a a result of applying the 'keep alive'
1938 // closure to the entries on the discovered ref lists) or we overflow
1939 // the global marking stack.
1940 //
1941 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted()
1942 // flag while there may still be some work to do. (See the comment at
1943 // the beginning of CMTask::do_marking_step() for those conditions -
1944 // one of which is reaching the specified time target.) It is only
1945 // when CMTask::do_marking_step() returns without setting the
1946 // has_aborted() flag that the marking step has completed.
1947
1948 _task->do_marking_step(1000000000.0 /* something very large */,
1949 true /* do_termination */,
1950 _is_serial);
1951 } while (_task->has_aborted() && !_cm->has_overflown());
1952 }
1953 };
1954
1955 // Implementation of AbstractRefProcTaskExecutor for parallel
1956 // reference processing at the end of G1 concurrent marking
1957
1958 class G1CMRefProcTaskExecutor: public AbstractRefProcTaskExecutor {
1959 private:
1960 G1CollectedHeap* _g1h;
1961 ConcurrentMark* _cm;
1962 WorkGang* _workers;
1963 uint _active_workers;
1964
1965 public:
1966 G1CMRefProcTaskExecutor(G1CollectedHeap* g1h,
1967 ConcurrentMark* cm,
1968 WorkGang* workers,
1969 uint n_workers) :
1970 _g1h(g1h), _cm(cm),
1971 _workers(workers), _active_workers(n_workers) { }
1972
1973 // Executes the given task using concurrent marking worker threads.
1974 virtual void execute(ProcessTask& task);
1975 virtual void execute(EnqueueTask& task);
1976 };
1977
1978 class G1CMRefProcTaskProxy: public AbstractGangTask {
1979 typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask;
1980 ProcessTask& _proc_task;
1981 G1CollectedHeap* _g1h;
1982 ConcurrentMark* _cm;
1983
1984 public:
1985 G1CMRefProcTaskProxy(ProcessTask& proc_task,
1986 G1CollectedHeap* g1h,
1987 ConcurrentMark* cm) :
1988 AbstractGangTask("Process reference objects in parallel"),
1989 _proc_task(proc_task), _g1h(g1h), _cm(cm) {
1990 ReferenceProcessor* rp = _g1h->ref_processor_cm();
1991 assert(rp->processing_is_mt(), "shouldn't be here otherwise");
1992 }
1993
1994 virtual void work(uint worker_id) {
1995 ResourceMark rm;
1996 HandleMark hm;
1997 CMTask* task = _cm->task(worker_id);
1998 G1CMIsAliveClosure g1_is_alive(_g1h);
1999 G1CMKeepAliveAndDrainClosure g1_par_keep_alive(_cm, task, false /* is_serial */);
2000 G1CMDrainMarkingStackClosure g1_par_drain(_cm, task, false /* is_serial */);
2001
2002 _proc_task.work(worker_id, g1_is_alive, g1_par_keep_alive, g1_par_drain);
2003 }
2004 };
2005
2006 void G1CMRefProcTaskExecutor::execute(ProcessTask& proc_task) {
2007 assert(_workers != NULL, "Need parallel worker threads.");
2008 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
2009
2010 G1CMRefProcTaskProxy proc_task_proxy(proc_task, _g1h, _cm);
2011
2012 // We need to reset the concurrency level before each
2013 // proxy task execution, so that the termination protocol
2014 // and overflow handling in CMTask::do_marking_step() knows
2015 // how many workers to wait for.
2016 _cm->set_concurrency(_active_workers);
2017 _workers->run_task(&proc_task_proxy);
2018 }
2019
2020 class G1CMRefEnqueueTaskProxy: public AbstractGangTask {
2021 typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask;
2022 EnqueueTask& _enq_task;
2023
2024 public:
2025 G1CMRefEnqueueTaskProxy(EnqueueTask& enq_task) :
2026 AbstractGangTask("Enqueue reference objects in parallel"),
2027 _enq_task(enq_task) { }
2028
2029 virtual void work(uint worker_id) {
2030 _enq_task.work(worker_id);
2031 }
2032 };
2033
2034 void G1CMRefProcTaskExecutor::execute(EnqueueTask& enq_task) {
2035 assert(_workers != NULL, "Need parallel worker threads.");
2036 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
2037
2038 G1CMRefEnqueueTaskProxy enq_task_proxy(enq_task);
2039
2040 // Not strictly necessary but...
2041 //
2042 // We need to reset the concurrency level before each
2043 // proxy task execution, so that the termination protocol
2044 // and overflow handling in CMTask::do_marking_step() knows
2045 // how many workers to wait for.
2046 _cm->set_concurrency(_active_workers);
2047 _workers->run_task(&enq_task_proxy);
2048 }
2049
2050 void ConcurrentMark::weakRefsWorkParallelPart(BoolObjectClosure* is_alive, bool purged_classes) {
2051 G1CollectedHeap::heap()->parallel_cleaning(is_alive, true, true, purged_classes);
2052 }
2053
2054 void ConcurrentMark::weakRefsWork(bool clear_all_soft_refs) {
2055 if (has_overflown()) {
2056 // Skip processing the discovered references if we have
2057 // overflown the global marking stack. Reference objects
2058 // only get discovered once so it is OK to not
2059 // de-populate the discovered reference lists. We could have,
2060 // but the only benefit would be that, when marking restarts,
2061 // less reference objects are discovered.
2062 return;
2063 }
2064
2065 ResourceMark rm;
2066 HandleMark hm;
2067
2068 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2069
2070 // Is alive closure.
2071 G1CMIsAliveClosure g1_is_alive(g1h);
2072
2073 // Inner scope to exclude the cleaning of the string and symbol
2074 // tables from the displayed time.
2075 {
2076 G1CMTraceTime t("GC ref-proc", G1Log::finer());
2077
2078 ReferenceProcessor* rp = g1h->ref_processor_cm();
2079
2080 // See the comment in G1CollectedHeap::ref_processing_init()
2081 // about how reference processing currently works in G1.
2082
2083 // Set the soft reference policy
2084 rp->setup_policy(clear_all_soft_refs);
2085 assert(_markStack.isEmpty(), "mark stack should be empty");
2086
2087 // Instances of the 'Keep Alive' and 'Complete GC' closures used
2088 // in serial reference processing. Note these closures are also
2089 // used for serially processing (by the the current thread) the
2090 // JNI references during parallel reference processing.
2091 //
2092 // These closures do not need to synchronize with the worker
2093 // threads involved in parallel reference processing as these
2094 // instances are executed serially by the current thread (e.g.
2095 // reference processing is not multi-threaded and is thus
2096 // performed by the current thread instead of a gang worker).
2097 //
2098 // The gang tasks involved in parallel reference processing create
2099 // their own instances of these closures, which do their own
2100 // synchronization among themselves.
2101 G1CMKeepAliveAndDrainClosure g1_keep_alive(this, task(0), true /* is_serial */);
2102 G1CMDrainMarkingStackClosure g1_drain_mark_stack(this, task(0), true /* is_serial */);
2103
2104 // We need at least one active thread. If reference processing
2105 // is not multi-threaded we use the current (VMThread) thread,
2106 // otherwise we use the work gang from the G1CollectedHeap and
2107 // we utilize all the worker threads we can.
2108 bool processing_is_mt = rp->processing_is_mt();
2109 uint active_workers = (processing_is_mt ? g1h->workers()->active_workers() : 1U);
2110 active_workers = MAX2(MIN2(active_workers, _max_worker_id), 1U);
2111
2112 // Parallel processing task executor.
2113 G1CMRefProcTaskExecutor par_task_executor(g1h, this,
2114 g1h->workers(), active_workers);
2115 AbstractRefProcTaskExecutor* executor = (processing_is_mt ? &par_task_executor : NULL);
2116
2117 // Set the concurrency level. The phase was already set prior to
2118 // executing the remark task.
2119 set_concurrency(active_workers);
2120
2121 // Set the degree of MT processing here. If the discovery was done MT,
2122 // the number of threads involved during discovery could differ from
2123 // the number of active workers. This is OK as long as the discovered
2124 // Reference lists are balanced (see balance_all_queues() and balance_queues()).
2125 rp->set_active_mt_degree(active_workers);
2126
2127 // Process the weak references.
2128 const ReferenceProcessorStats& stats =
2129 rp->process_discovered_references(&g1_is_alive,
2130 &g1_keep_alive,
2131 &g1_drain_mark_stack,
2132 executor,
2133 g1h->gc_timer_cm());
2134 g1h->gc_tracer_cm()->report_gc_reference_stats(stats);
2135
2136 // The do_oop work routines of the keep_alive and drain_marking_stack
2137 // oop closures will set the has_overflown flag if we overflow the
2138 // global marking stack.
2139
2140 assert(_markStack.overflow() || _markStack.isEmpty(),
2141 "mark stack should be empty (unless it overflowed)");
2142
2143 if (_markStack.overflow()) {
2144 // This should have been done already when we tried to push an
2145 // entry on to the global mark stack. But let's do it again.
2146 set_has_overflown();
2147 }
2148
2149 assert(rp->num_q() == active_workers, "why not");
2150
2151 rp->enqueue_discovered_references(executor);
2152
2153 rp->verify_no_references_recorded();
2154 assert(!rp->discovery_enabled(), "Post condition");
2155 }
2156
2157 if (has_overflown()) {
2158 // We can not trust g1_is_alive if the marking stack overflowed
2159 return;
2160 }
2161
2162 assert(_markStack.isEmpty(), "Marking should have completed");
2163
2164 // Unload Klasses, String, Symbols, Code Cache, etc.
2165 {
2166 G1CMTraceTime trace("Unloading", G1Log::finer());
2167
2168 if (ClassUnloadingWithConcurrentMark) {
2169 bool purged_classes;
2170
2171 {
2172 G1CMTraceTime trace("System Dictionary Unloading", G1Log::finest());
2173 purged_classes = SystemDictionary::do_unloading(&g1_is_alive, false /* Defer klass cleaning */);
2174 }
2175
2176 {
2177 G1CMTraceTime trace("Parallel Unloading", G1Log::finest());
2178 weakRefsWorkParallelPart(&g1_is_alive, purged_classes);
2179 }
2180 }
2181
2182 if (G1StringDedup::is_enabled()) {
2183 G1CMTraceTime trace("String Deduplication Unlink", G1Log::finest());
2184 G1StringDedup::unlink(&g1_is_alive);
2185 }
2186 }
2187 }
2188
2189 void ConcurrentMark::swapMarkBitMaps() {
2190 CMBitMapRO* temp = _prevMarkBitMap;
2191 _prevMarkBitMap = (CMBitMapRO*)_nextMarkBitMap;
2192 _nextMarkBitMap = (CMBitMap*) temp;
2193 }
2194
2195 // Closure for marking entries in SATB buffers.
2196 class CMSATBBufferClosure : public SATBBufferClosure {
2197 private:
2198 CMTask* _task;
2199 G1CollectedHeap* _g1h;
2200
2201 // This is very similar to CMTask::deal_with_reference, but with
2202 // more relaxed requirements for the argument, so this must be more
2203 // circumspect about treating the argument as an object.
2204 void do_entry(void* entry) const {
2205 _task->increment_refs_reached();
2206 HeapRegion* hr = _g1h->heap_region_containing(entry);
2207 if (entry < hr->next_top_at_mark_start()) {
2208 // Until we get here, we don't know whether entry refers to a valid
2209 // object; it could instead have been a stale reference.
2210 oop obj = static_cast<oop>(entry);
2211 assert(obj->is_oop(true /* ignore mark word */),
2212 "Invalid oop in SATB buffer: " PTR_FORMAT, p2i(obj));
2213 _task->make_reference_grey(obj, hr);
2214 }
2215 }
2216
2217 public:
2218 CMSATBBufferClosure(CMTask* task, G1CollectedHeap* g1h)
2219 : _task(task), _g1h(g1h) { }
2220
2221 virtual void do_buffer(void** buffer, size_t size) {
2222 for (size_t i = 0; i < size; ++i) {
2223 do_entry(buffer[i]);
2224 }
2225 }
2226 };
2227
2228 class G1RemarkThreadsClosure : public ThreadClosure {
2229 CMSATBBufferClosure _cm_satb_cl;
2230 G1CMOopClosure _cm_cl;
2231 MarkingCodeBlobClosure _code_cl;
2232 int _thread_parity;
2233
2234 public:
2235 G1RemarkThreadsClosure(G1CollectedHeap* g1h, CMTask* task) :
2236 _cm_satb_cl(task, g1h),
2237 _cm_cl(g1h, g1h->concurrent_mark(), task),
2238 _code_cl(&_cm_cl, !CodeBlobToOopClosure::FixRelocations),
2239 _thread_parity(Threads::thread_claim_parity()) {}
2240
2241 void do_thread(Thread* thread) {
2242 if (thread->is_Java_thread()) {
2243 if (thread->claim_oops_do(true, _thread_parity)) {
2244 JavaThread* jt = (JavaThread*)thread;
2245
2246 // In theory it should not be neccessary to explicitly walk the nmethods to find roots for concurrent marking
2247 // however the liveness of oops reachable from nmethods have very complex lifecycles:
2248 // * Alive if on the stack of an executing method
2249 // * Weakly reachable otherwise
2250 // Some objects reachable from nmethods, such as the class loader (or klass_holder) of the receiver should be
2251 // live by the SATB invariant but other oops recorded in nmethods may behave differently.
2252 jt->nmethods_do(&_code_cl);
2253
2254 jt->satb_mark_queue().apply_closure_and_empty(&_cm_satb_cl);
2255 }
2256 } else if (thread->is_VM_thread()) {
2257 if (thread->claim_oops_do(true, _thread_parity)) {
2258 JavaThread::satb_mark_queue_set().shared_satb_queue()->apply_closure_and_empty(&_cm_satb_cl);
2259 }
2260 }
2261 }
2262 };
2263
2264 class CMRemarkTask: public AbstractGangTask {
2265 private:
2266 ConcurrentMark* _cm;
2267 public:
2268 void work(uint worker_id) {
2269 // Since all available tasks are actually started, we should
2270 // only proceed if we're supposed to be active.
2271 if (worker_id < _cm->active_tasks()) {
2272 CMTask* task = _cm->task(worker_id);
2273 task->record_start_time();
2274 {
2275 ResourceMark rm;
2276 HandleMark hm;
2277
2278 G1RemarkThreadsClosure threads_f(G1CollectedHeap::heap(), task);
2279 Threads::threads_do(&threads_f);
2280 }
2281
2282 do {
2283 task->do_marking_step(1000000000.0 /* something very large */,
2284 true /* do_termination */,
2285 false /* is_serial */);
2286 } while (task->has_aborted() && !_cm->has_overflown());
2287 // If we overflow, then we do not want to restart. We instead
2288 // want to abort remark and do concurrent marking again.
2289 task->record_end_time();
2290 }
2291 }
2292
2293 CMRemarkTask(ConcurrentMark* cm, uint active_workers) :
2294 AbstractGangTask("Par Remark"), _cm(cm) {
2295 _cm->terminator()->reset_for_reuse(active_workers);
2296 }
2297 };
2298
2299 void ConcurrentMark::checkpointRootsFinalWork() {
2300 ResourceMark rm;
2301 HandleMark hm;
2302 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2303
2304 G1CMTraceTime trace("Finalize Marking", G1Log::finer());
2305
2306 g1h->ensure_parsability(false);
2307
2308 // this is remark, so we'll use up all active threads
2309 uint active_workers = g1h->workers()->active_workers();
2310 set_concurrency_and_phase(active_workers, false /* concurrent */);
2311 // Leave _parallel_marking_threads at it's
2312 // value originally calculated in the ConcurrentMark
2313 // constructor and pass values of the active workers
2314 // through the gang in the task.
2315
2316 {
2317 StrongRootsScope srs(active_workers);
2318
2319 CMRemarkTask remarkTask(this, active_workers);
2320 // We will start all available threads, even if we decide that the
2321 // active_workers will be fewer. The extra ones will just bail out
2322 // immediately.
2323 g1h->workers()->run_task(&remarkTask);
2324 }
2325
2326 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2327 guarantee(has_overflown() ||
2328 satb_mq_set.completed_buffers_num() == 0,
2329 "Invariant: has_overflown = %s, num buffers = %d",
2330 BOOL_TO_STR(has_overflown()),
2331 satb_mq_set.completed_buffers_num());
2332
2333 print_stats();
2334 }
2335
2336 void ConcurrentMark::clearRangePrevBitmap(MemRegion mr) {
2337 // Note we are overriding the read-only view of the prev map here, via
2338 // the cast.
2339 ((CMBitMap*)_prevMarkBitMap)->clearRange(mr);
2340 }
2341
2342 HeapRegion*
2343 ConcurrentMark::claim_region(uint worker_id) {
2344 // "checkpoint" the finger
2345 HeapWord* finger = _finger;
2346
2347 // _heap_end will not change underneath our feet; it only changes at
2348 // yield points.
2349 while (finger < _heap_end) {
2350 assert(_g1h->is_in_g1_reserved(finger), "invariant");
2351
2352 HeapRegion* curr_region = _g1h->heap_region_containing(finger);
2353
2354 // Above heap_region_containing may return NULL as we always scan claim
2355 // until the end of the heap. In this case, just jump to the next region.
2356 HeapWord* end = curr_region != NULL ? curr_region->end() : finger + HeapRegion::GrainWords;
2357
2358 // Is the gap between reading the finger and doing the CAS too long?
2359 HeapWord* res = (HeapWord*) Atomic::cmpxchg_ptr(end, &_finger, finger);
2360 if (res == finger && curr_region != NULL) {
2361 // we succeeded
2362 HeapWord* bottom = curr_region->bottom();
2363 HeapWord* limit = curr_region->next_top_at_mark_start();
2364
2365 // notice that _finger == end cannot be guaranteed here since,
2366 // someone else might have moved the finger even further
2367 assert(_finger >= end, "the finger should have moved forward");
2368
2369 if (limit > bottom) {
2370 return curr_region;
2371 } else {
2372 assert(limit == bottom,
2373 "the region limit should be at bottom");
2374 // we return NULL and the caller should try calling
2375 // claim_region() again.
2376 return NULL;
2377 }
2378 } else {
2379 assert(_finger > finger, "the finger should have moved forward");
2380 // read it again
2381 finger = _finger;
2382 }
2383 }
2384
2385 return NULL;
2386 }
2387
2388 #ifndef PRODUCT
2389 class VerifyNoCSetOops VALUE_OBJ_CLASS_SPEC {
2390 private:
2391 G1CollectedHeap* _g1h;
2392 const char* _phase;
2393 int _info;
2394
2395 public:
2396 VerifyNoCSetOops(const char* phase, int info = -1) :
2397 _g1h(G1CollectedHeap::heap()),
2398 _phase(phase),
2399 _info(info)
2400 { }
2401
2402 void operator()(oop obj) const {
2403 guarantee(obj->is_oop(),
2404 "Non-oop " PTR_FORMAT ", phase: %s, info: %d",
2405 p2i(obj), _phase, _info);
2406 guarantee(!_g1h->obj_in_cs(obj),
2407 "obj: " PTR_FORMAT " in CSet, phase: %s, info: %d",
2408 p2i(obj), _phase, _info);
2409 }
2410 };
2411
2412 void ConcurrentMark::verify_no_cset_oops() {
2413 assert(SafepointSynchronize::is_at_safepoint(), "should be at a safepoint");
2414 if (!G1CollectedHeap::heap()->collector_state()->mark_in_progress()) {
2415 return;
2416 }
2417
2418 // Verify entries on the global mark stack
2419 _markStack.iterate(VerifyNoCSetOops("Stack"));
2420
2421 // Verify entries on the task queues
2422 for (uint i = 0; i < _max_worker_id; ++i) {
2423 CMTaskQueue* queue = _task_queues->queue(i);
2424 queue->iterate(VerifyNoCSetOops("Queue", i));
2425 }
2426
2427 // Verify the global finger
2428 HeapWord* global_finger = finger();
2429 if (global_finger != NULL && global_finger < _heap_end) {
2430 // Since we always iterate over all regions, we might get a NULL HeapRegion
2431 // here.
2432 HeapRegion* global_hr = _g1h->heap_region_containing(global_finger);
2433 guarantee(global_hr == NULL || global_finger == global_hr->bottom(),
2434 "global finger: " PTR_FORMAT " region: " HR_FORMAT,
2435 p2i(global_finger), HR_FORMAT_PARAMS(global_hr));
2436 }
2437
2438 // Verify the task fingers
2439 assert(parallel_marking_threads() <= _max_worker_id, "sanity");
2440 for (uint i = 0; i < parallel_marking_threads(); ++i) {
2441 CMTask* task = _tasks[i];
2442 HeapWord* task_finger = task->finger();
2443 if (task_finger != NULL && task_finger < _heap_end) {
2444 // See above note on the global finger verification.
2445 HeapRegion* task_hr = _g1h->heap_region_containing(task_finger);
2446 guarantee(task_hr == NULL || task_finger == task_hr->bottom() ||
2447 !task_hr->in_collection_set(),
2448 "task finger: " PTR_FORMAT " region: " HR_FORMAT,
2449 p2i(task_finger), HR_FORMAT_PARAMS(task_hr));
2450 }
2451 }
2452 }
2453 #endif // PRODUCT
2454
2455 // Aggregate the counting data that was constructed concurrently
2456 // with marking.
2457 class AggregateCountDataHRClosure: public HeapRegionClosure {
2458 G1CollectedHeap* _g1h;
2459 ConcurrentMark* _cm;
2460 CardTableModRefBS* _ct_bs;
2461 BitMap* _cm_card_bm;
2462 uint _max_worker_id;
2463
2464 public:
2465 AggregateCountDataHRClosure(G1CollectedHeap* g1h,
2466 BitMap* cm_card_bm,
2467 uint max_worker_id) :
2468 _g1h(g1h), _cm(g1h->concurrent_mark()),
2469 _ct_bs(barrier_set_cast<CardTableModRefBS>(g1h->barrier_set())),
2470 _cm_card_bm(cm_card_bm), _max_worker_id(max_worker_id) { }
2471
2472 bool doHeapRegion(HeapRegion* hr) {
2473 HeapWord* start = hr->bottom();
2474 HeapWord* limit = hr->next_top_at_mark_start();
2475 HeapWord* end = hr->end();
2476
2477 assert(start <= limit && limit <= hr->top() && hr->top() <= hr->end(),
2478 "Preconditions not met - "
2479 "start: " PTR_FORMAT ", limit: " PTR_FORMAT ", "
2480 "top: " PTR_FORMAT ", end: " PTR_FORMAT,
2481 p2i(start), p2i(limit), p2i(hr->top()), p2i(hr->end()));
2482
2483 assert(hr->next_marked_bytes() == 0, "Precondition");
2484
2485 if (start == limit) {
2486 // NTAMS of this region has not been set so nothing to do.
2487 return false;
2488 }
2489
2490 // 'start' should be in the heap.
2491 assert(_g1h->is_in_g1_reserved(start) && _ct_bs->is_card_aligned(start), "sanity");
2492 // 'end' *may* be just beyond the end of the heap (if hr is the last region)
2493 assert(!_g1h->is_in_g1_reserved(end) || _ct_bs->is_card_aligned(end), "sanity");
2494
2495 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start);
2496 BitMap::idx_t limit_idx = _cm->card_bitmap_index_for(limit);
2497 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(end);
2498
2499 // If ntams is not card aligned then we bump card bitmap index
2500 // for limit so that we get the all the cards spanned by
2501 // the object ending at ntams.
2502 // Note: if this is the last region in the heap then ntams
2503 // could be actually just beyond the end of the the heap;
2504 // limit_idx will then correspond to a (non-existent) card
2505 // that is also outside the heap.
2506 if (_g1h->is_in_g1_reserved(limit) && !_ct_bs->is_card_aligned(limit)) {
2507 limit_idx += 1;
2508 }
2509
2510 assert(limit_idx <= end_idx, "or else use atomics");
2511
2512 // Aggregate the "stripe" in the count data associated with hr.
2513 uint hrm_index = hr->hrm_index();
2514 size_t marked_bytes = 0;
2515
2516 for (uint i = 0; i < _max_worker_id; i += 1) {
2517 size_t* marked_bytes_array = _cm->count_marked_bytes_array_for(i);
2518 BitMap* task_card_bm = _cm->count_card_bitmap_for(i);
2519
2520 // Fetch the marked_bytes in this region for task i and
2521 // add it to the running total for this region.
2522 marked_bytes += marked_bytes_array[hrm_index];
2523
2524 // Now union the bitmaps[0,max_worker_id)[start_idx..limit_idx)
2525 // into the global card bitmap.
2526 BitMap::idx_t scan_idx = task_card_bm->get_next_one_offset(start_idx, limit_idx);
2527
2528 while (scan_idx < limit_idx) {
2529 assert(task_card_bm->at(scan_idx) == true, "should be");
2530 _cm_card_bm->set_bit(scan_idx);
2531 assert(_cm_card_bm->at(scan_idx) == true, "should be");
2532
2533 // BitMap::get_next_one_offset() can handle the case when
2534 // its left_offset parameter is greater than its right_offset
2535 // parameter. It does, however, have an early exit if
2536 // left_offset == right_offset. So let's limit the value
2537 // passed in for left offset here.
2538 BitMap::idx_t next_idx = MIN2(scan_idx + 1, limit_idx);
2539 scan_idx = task_card_bm->get_next_one_offset(next_idx, limit_idx);
2540 }
2541 }
2542
2543 // Update the marked bytes for this region.
2544 hr->add_to_marked_bytes(marked_bytes);
2545
2546 // Next heap region
2547 return false;
2548 }
2549 };
2550
2551 class G1AggregateCountDataTask: public AbstractGangTask {
2552 protected:
2553 G1CollectedHeap* _g1h;
2554 ConcurrentMark* _cm;
2555 BitMap* _cm_card_bm;
2556 uint _max_worker_id;
2557 uint _active_workers;
2558 HeapRegionClaimer _hrclaimer;
2559
2560 public:
2561 G1AggregateCountDataTask(G1CollectedHeap* g1h,
2562 ConcurrentMark* cm,
2563 BitMap* cm_card_bm,
2564 uint max_worker_id,
2565 uint n_workers) :
2566 AbstractGangTask("Count Aggregation"),
2567 _g1h(g1h), _cm(cm), _cm_card_bm(cm_card_bm),
2568 _max_worker_id(max_worker_id),
2569 _active_workers(n_workers),
2570 _hrclaimer(_active_workers) {
2571 }
2572
2573 void work(uint worker_id) {
2574 AggregateCountDataHRClosure cl(_g1h, _cm_card_bm, _max_worker_id);
2575
2576 _g1h->heap_region_par_iterate(&cl, worker_id, &_hrclaimer);
2577 }
2578 };
2579
2580
2581 void ConcurrentMark::aggregate_count_data() {
2582 uint n_workers = _g1h->workers()->active_workers();
2583
2584 G1AggregateCountDataTask g1_par_agg_task(_g1h, this, &_card_bm,
2585 _max_worker_id, n_workers);
2586
2587 _g1h->workers()->run_task(&g1_par_agg_task);
2588 }
2589
2590 // Clear the per-worker arrays used to store the per-region counting data
2591 void ConcurrentMark::clear_all_count_data() {
2592 // Clear the global card bitmap - it will be filled during
2593 // liveness count aggregation (during remark) and the
2594 // final counting task.
2595 _card_bm.clear();
2596
2597 // Clear the global region bitmap - it will be filled as part
2598 // of the final counting task.
2599 _region_bm.clear();
2600
2601 uint max_regions = _g1h->max_regions();
2602 assert(_max_worker_id > 0, "uninitialized");
2603
2604 for (uint i = 0; i < _max_worker_id; i += 1) {
2605 BitMap* task_card_bm = count_card_bitmap_for(i);
2606 size_t* marked_bytes_array = count_marked_bytes_array_for(i);
2607
2608 assert(task_card_bm->size() == _card_bm.size(), "size mismatch");
2609 assert(marked_bytes_array != NULL, "uninitialized");
2610
2611 memset(marked_bytes_array, 0, (size_t) max_regions * sizeof(size_t));
2612 task_card_bm->clear();
2613 }
2614 }
2615
2616 void ConcurrentMark::print_stats() {
2617 if (G1MarkingVerboseLevel > 0) {
2618 gclog_or_tty->print_cr("---------------------------------------------------------------------");
2619 for (size_t i = 0; i < _active_tasks; ++i) {
2620 _tasks[i]->print_stats();
2621 gclog_or_tty->print_cr("---------------------------------------------------------------------");
2622 }
2623 }
2624 }
2625
2626 // abandon current marking iteration due to a Full GC
2627 void ConcurrentMark::abort() {
2628 if (!cmThread()->during_cycle() || _has_aborted) {
2629 // We haven't started a concurrent cycle or we have already aborted it. No need to do anything.
2630 return;
2631 }
2632
2633 // Clear all marks in the next bitmap for the next marking cycle. This will allow us to skip the next
2634 // concurrent bitmap clearing.
2635 _nextMarkBitMap->clearAll();
2636
2637 // Note we cannot clear the previous marking bitmap here
2638 // since VerifyDuringGC verifies the objects marked during
2639 // a full GC against the previous bitmap.
2640
2641 // Clear the liveness counting data
2642 clear_all_count_data();
2643 // Empty mark stack
2644 reset_marking_state();
2645 for (uint i = 0; i < _max_worker_id; ++i) {
2646 _tasks[i]->clear_region_fields();
2647 }
2648 _first_overflow_barrier_sync.abort();
2649 _second_overflow_barrier_sync.abort();
2650 _has_aborted = true;
2651
2652 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2653 satb_mq_set.abandon_partial_marking();
2654 // This can be called either during or outside marking, we'll read
2655 // the expected_active value from the SATB queue set.
2656 satb_mq_set.set_active_all_threads(
2657 false, /* new active value */
2658 satb_mq_set.is_active() /* expected_active */);
2659
2660 _g1h->trace_heap_after_concurrent_cycle();
2661 _g1h->register_concurrent_cycle_end();
2662 }
2663
2664 static void print_ms_time_info(const char* prefix, const char* name,
2665 NumberSeq& ns) {
2666 gclog_or_tty->print_cr("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).",
2667 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg());
2668 if (ns.num() > 0) {
2669 gclog_or_tty->print_cr("%s [std. dev = %8.2f ms, max = %8.2f ms]",
2670 prefix, ns.sd(), ns.maximum());
2671 }
2672 }
2673
2674 void ConcurrentMark::print_summary_info() {
2675 gclog_or_tty->print_cr(" Concurrent marking:");
2676 print_ms_time_info(" ", "init marks", _init_times);
2677 print_ms_time_info(" ", "remarks", _remark_times);
2678 {
2679 print_ms_time_info(" ", "final marks", _remark_mark_times);
2680 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times);
2681
2682 }
2683 print_ms_time_info(" ", "cleanups", _cleanup_times);
2684 gclog_or_tty->print_cr(" Final counting total time = %8.2f s (avg = %8.2f ms).",
2685 _total_counting_time,
2686 (_cleanup_times.num() > 0 ? _total_counting_time * 1000.0 /
2687 (double)_cleanup_times.num()
2688 : 0.0));
2689 if (G1ScrubRemSets) {
2690 gclog_or_tty->print_cr(" RS scrub total time = %8.2f s (avg = %8.2f ms).",
2691 _total_rs_scrub_time,
2692 (_cleanup_times.num() > 0 ? _total_rs_scrub_time * 1000.0 /
2693 (double)_cleanup_times.num()
2694 : 0.0));
2695 }
2696 gclog_or_tty->print_cr(" Total stop_world time = %8.2f s.",
2697 (_init_times.sum() + _remark_times.sum() +
2698 _cleanup_times.sum())/1000.0);
2699 gclog_or_tty->print_cr(" Total concurrent time = %8.2f s "
2700 "(%8.2f s marking).",
2701 cmThread()->vtime_accum(),
2702 cmThread()->vtime_mark_accum());
2703 }
2704
2705 void ConcurrentMark::print_worker_threads_on(outputStream* st) const {
2706 _parallel_workers->print_worker_threads_on(st);
2707 }
2708
2709 void ConcurrentMark::print_on_error(outputStream* st) const {
2710 st->print_cr("Marking Bits (Prev, Next): (CMBitMap*) " PTR_FORMAT ", (CMBitMap*) " PTR_FORMAT,
2711 p2i(_prevMarkBitMap), p2i(_nextMarkBitMap));
2712 _prevMarkBitMap->print_on_error(st, " Prev Bits: ");
2713 _nextMarkBitMap->print_on_error(st, " Next Bits: ");
2714 }
2715
2716 // We take a break if someone is trying to stop the world.
2717 bool ConcurrentMark::do_yield_check(uint worker_id) {
2718 if (SuspendibleThreadSet::should_yield()) {
2719 if (worker_id == 0) {
2720 _g1h->g1_policy()->record_concurrent_pause();
2721 }
2722 SuspendibleThreadSet::yield();
2723 return true;
2724 } else {
2725 return false;
2726 }
2727 }
2728
2729 // Closure for iteration over bitmaps
2730 class CMBitMapClosure : public BitMapClosure {
2731 private:
2732 // the bitmap that is being iterated over
2733 CMBitMap* _nextMarkBitMap;
2734 ConcurrentMark* _cm;
2735 CMTask* _task;
2736
2737 public:
2738 CMBitMapClosure(CMTask *task, ConcurrentMark* cm, CMBitMap* nextMarkBitMap) :
2739 _task(task), _cm(cm), _nextMarkBitMap(nextMarkBitMap) { }
2740
2741 bool do_bit(size_t offset) {
2742 HeapWord* addr = _nextMarkBitMap->offsetToHeapWord(offset);
2743 assert(_nextMarkBitMap->isMarked(addr), "invariant");
2744 assert( addr < _cm->finger(), "invariant");
2745 assert(addr >= _task->finger(), "invariant");
2746
2747 // We move that task's local finger along.
2748 _task->move_finger_to(addr);
2749
2750 _task->scan_object(oop(addr));
2751 // we only partially drain the local queue and global stack
2752 _task->drain_local_queue(true);
2753 _task->drain_global_stack(true);
2754
2755 // if the has_aborted flag has been raised, we need to bail out of
2756 // the iteration
2757 return !_task->has_aborted();
2758 }
2759 };
2760
2761 static ReferenceProcessor* get_cm_oop_closure_ref_processor(G1CollectedHeap* g1h) {
2762 ReferenceProcessor* result = NULL;
2763 if (G1UseConcMarkReferenceProcessing) {
2764 result = g1h->ref_processor_cm();
2765 assert(result != NULL, "should not be NULL");
2766 }
2767 return result;
2768 }
2769
2770 G1CMOopClosure::G1CMOopClosure(G1CollectedHeap* g1h,
2771 ConcurrentMark* cm,
2772 CMTask* task)
2773 : MetadataAwareOopClosure(get_cm_oop_closure_ref_processor(g1h)),
2774 _g1h(g1h), _cm(cm), _task(task)
2775 { }
2776
2777 void CMTask::setup_for_region(HeapRegion* hr) {
2778 assert(hr != NULL,
2779 "claim_region() should have filtered out NULL regions");
2780 _curr_region = hr;
2781 _finger = hr->bottom();
2782 update_region_limit();
2783 }
2784
2785 void CMTask::update_region_limit() {
2786 HeapRegion* hr = _curr_region;
2787 HeapWord* bottom = hr->bottom();
2788 HeapWord* limit = hr->next_top_at_mark_start();
2789
2790 if (limit == bottom) {
2791 // The region was collected underneath our feet.
2792 // We set the finger to bottom to ensure that the bitmap
2793 // iteration that will follow this will not do anything.
2794 // (this is not a condition that holds when we set the region up,
2795 // as the region is not supposed to be empty in the first place)
2796 _finger = bottom;
2797 } else if (limit >= _region_limit) {
2798 assert(limit >= _finger, "peace of mind");
2799 } else {
2800 assert(limit < _region_limit, "only way to get here");
2801 // This can happen under some pretty unusual circumstances. An
2802 // evacuation pause empties the region underneath our feet (NTAMS
2803 // at bottom). We then do some allocation in the region (NTAMS
2804 // stays at bottom), followed by the region being used as a GC
2805 // alloc region (NTAMS will move to top() and the objects
2806 // originally below it will be grayed). All objects now marked in
2807 // the region are explicitly grayed, if below the global finger,
2808 // and we do not need in fact to scan anything else. So, we simply
2809 // set _finger to be limit to ensure that the bitmap iteration
2810 // doesn't do anything.
2811 _finger = limit;
2812 }
2813
2814 _region_limit = limit;
2815 }
2816
2817 void CMTask::giveup_current_region() {
2818 assert(_curr_region != NULL, "invariant");
2819 clear_region_fields();
2820 }
2821
2822 void CMTask::clear_region_fields() {
2823 // Values for these three fields that indicate that we're not
2824 // holding on to a region.
2825 _curr_region = NULL;
2826 _finger = NULL;
2827 _region_limit = NULL;
2828 }
2829
2830 void CMTask::set_cm_oop_closure(G1CMOopClosure* cm_oop_closure) {
2831 if (cm_oop_closure == NULL) {
2832 assert(_cm_oop_closure != NULL, "invariant");
2833 } else {
2834 assert(_cm_oop_closure == NULL, "invariant");
2835 }
2836 _cm_oop_closure = cm_oop_closure;
2837 }
2838
2839 void CMTask::reset(CMBitMap* nextMarkBitMap) {
2840 guarantee(nextMarkBitMap != NULL, "invariant");
2841 _nextMarkBitMap = nextMarkBitMap;
2842 clear_region_fields();
2843
2844 _calls = 0;
2845 _elapsed_time_ms = 0.0;
2846 _termination_time_ms = 0.0;
2847 _termination_start_time_ms = 0.0;
2848 }
2849
2850 bool CMTask::should_exit_termination() {
2851 regular_clock_call();
2852 // This is called when we are in the termination protocol. We should
2853 // quit if, for some reason, this task wants to abort or the global
2854 // stack is not empty (this means that we can get work from it).
2855 return !_cm->mark_stack_empty() || has_aborted();
2856 }
2857
2858 void CMTask::reached_limit() {
2859 assert(_words_scanned >= _words_scanned_limit ||
2860 _refs_reached >= _refs_reached_limit ,
2861 "shouldn't have been called otherwise");
2862 regular_clock_call();
2863 }
2864
2865 void CMTask::regular_clock_call() {
2866 if (has_aborted()) return;
2867
2868 // First, we need to recalculate the words scanned and refs reached
2869 // limits for the next clock call.
2870 recalculate_limits();
2871
2872 // During the regular clock call we do the following
2873
2874 // (1) If an overflow has been flagged, then we abort.
2875 if (_cm->has_overflown()) {
2876 set_has_aborted();
2877 return;
2878 }
2879
2880 // If we are not concurrent (i.e. we're doing remark) we don't need
2881 // to check anything else. The other steps are only needed during
2882 // the concurrent marking phase.
2883 if (!concurrent()) return;
2884
2885 // (2) If marking has been aborted for Full GC, then we also abort.
2886 if (_cm->has_aborted()) {
2887 set_has_aborted();
2888 return;
2889 }
2890
2891 double curr_time_ms = os::elapsedVTime() * 1000.0;
2892
2893 // (4) We check whether we should yield. If we have to, then we abort.
2894 if (SuspendibleThreadSet::should_yield()) {
2895 // We should yield. To do this we abort the task. The caller is
2896 // responsible for yielding.
2897 set_has_aborted();
2898 return;
2899 }
2900
2901 // (5) We check whether we've reached our time quota. If we have,
2902 // then we abort.
2903 double elapsed_time_ms = curr_time_ms - _start_time_ms;
2904 if (elapsed_time_ms > _time_target_ms) {
2905 set_has_aborted();
2906 _has_timed_out = true;
2907 return;
2908 }
2909
2910 // (6) Finally, we check whether there are enough completed STAB
2911 // buffers available for processing. If there are, we abort.
2912 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2913 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) {
2914 // we do need to process SATB buffers, we'll abort and restart
2915 // the marking task to do so
2916 set_has_aborted();
2917 return;
2918 }
2919 }
2920
2921 void CMTask::recalculate_limits() {
2922 _real_words_scanned_limit = _words_scanned + words_scanned_period;
2923 _words_scanned_limit = _real_words_scanned_limit;
2924
2925 _real_refs_reached_limit = _refs_reached + refs_reached_period;
2926 _refs_reached_limit = _real_refs_reached_limit;
2927 }
2928
2929 void CMTask::decrease_limits() {
2930 // This is called when we believe that we're going to do an infrequent
2931 // operation which will increase the per byte scanned cost (i.e. move
2932 // entries to/from the global stack). It basically tries to decrease the
2933 // scanning limit so that the clock is called earlier.
2934
2935 _words_scanned_limit = _real_words_scanned_limit -
2936 3 * words_scanned_period / 4;
2937 _refs_reached_limit = _real_refs_reached_limit -
2938 3 * refs_reached_period / 4;
2939 }
2940
2941 void CMTask::move_entries_to_global_stack() {
2942 // local array where we'll store the entries that will be popped
2943 // from the local queue
2944 oop buffer[global_stack_transfer_size];
2945
2946 int n = 0;
2947 oop obj;
2948 while (n < global_stack_transfer_size && _task_queue->pop_local(obj)) {
2949 buffer[n] = obj;
2950 ++n;
2951 }
2952
2953 if (n > 0) {
2954 // we popped at least one entry from the local queue
2955
2956 if (!_cm->mark_stack_push(buffer, n)) {
2957 set_has_aborted();
2958 }
2959 }
2960
2961 // this operation was quite expensive, so decrease the limits
2962 decrease_limits();
2963 }
2964
2965 void CMTask::get_entries_from_global_stack() {
2966 // local array where we'll store the entries that will be popped
2967 // from the global stack.
2968 oop buffer[global_stack_transfer_size];
2969 int n;
2970 _cm->mark_stack_pop(buffer, global_stack_transfer_size, &n);
2971 assert(n <= global_stack_transfer_size,
2972 "we should not pop more than the given limit");
2973 if (n > 0) {
2974 // yes, we did actually pop at least one entry
2975 for (int i = 0; i < n; ++i) {
2976 bool success = _task_queue->push(buffer[i]);
2977 // We only call this when the local queue is empty or under a
2978 // given target limit. So, we do not expect this push to fail.
2979 assert(success, "invariant");
2980 }
2981 }
2982
2983 // this operation was quite expensive, so decrease the limits
2984 decrease_limits();
2985 }
2986
2987 void CMTask::drain_local_queue(bool partially) {
2988 if (has_aborted()) return;
2989
2990 // Decide what the target size is, depending whether we're going to
2991 // drain it partially (so that other tasks can steal if they run out
2992 // of things to do) or totally (at the very end).
2993 size_t target_size;
2994 if (partially) {
2995 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize);
2996 } else {
2997 target_size = 0;
2998 }
2999
3000 if (_task_queue->size() > target_size) {
3001 oop obj;
3002 bool ret = _task_queue->pop_local(obj);
3003 while (ret) {
3004 assert(_g1h->is_in_g1_reserved((HeapWord*) obj), "invariant" );
3005 assert(!_g1h->is_on_master_free_list(
3006 _g1h->heap_region_containing((HeapWord*) obj)), "invariant");
3007
3008 scan_object(obj);
3009
3010 if (_task_queue->size() <= target_size || has_aborted()) {
3011 ret = false;
3012 } else {
3013 ret = _task_queue->pop_local(obj);
3014 }
3015 }
3016 }
3017 }
3018
3019 void CMTask::drain_global_stack(bool partially) {
3020 if (has_aborted()) return;
3021
3022 // We have a policy to drain the local queue before we attempt to
3023 // drain the global stack.
3024 assert(partially || _task_queue->size() == 0, "invariant");
3025
3026 // Decide what the target size is, depending whether we're going to
3027 // drain it partially (so that other tasks can steal if they run out
3028 // of things to do) or totally (at the very end). Notice that,
3029 // because we move entries from the global stack in chunks or
3030 // because another task might be doing the same, we might in fact
3031 // drop below the target. But, this is not a problem.
3032 size_t target_size;
3033 if (partially) {
3034 target_size = _cm->partial_mark_stack_size_target();
3035 } else {
3036 target_size = 0;
3037 }
3038
3039 if (_cm->mark_stack_size() > target_size) {
3040 while (!has_aborted() && _cm->mark_stack_size() > target_size) {
3041 get_entries_from_global_stack();
3042 drain_local_queue(partially);
3043 }
3044 }
3045 }
3046
3047 // SATB Queue has several assumptions on whether to call the par or
3048 // non-par versions of the methods. this is why some of the code is
3049 // replicated. We should really get rid of the single-threaded version
3050 // of the code to simplify things.
3051 void CMTask::drain_satb_buffers() {
3052 if (has_aborted()) return;
3053
3054 // We set this so that the regular clock knows that we're in the
3055 // middle of draining buffers and doesn't set the abort flag when it
3056 // notices that SATB buffers are available for draining. It'd be
3057 // very counter productive if it did that. :-)
3058 _draining_satb_buffers = true;
3059
3060 CMSATBBufferClosure satb_cl(this, _g1h);
3061 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3062
3063 // This keeps claiming and applying the closure to completed buffers
3064 // until we run out of buffers or we need to abort.
3065 while (!has_aborted() &&
3066 satb_mq_set.apply_closure_to_completed_buffer(&satb_cl)) {
3067 regular_clock_call();
3068 }
3069
3070 _draining_satb_buffers = false;
3071
3072 assert(has_aborted() ||
3073 concurrent() ||
3074 satb_mq_set.completed_buffers_num() == 0, "invariant");
3075
3076 // again, this was a potentially expensive operation, decrease the
3077 // limits to get the regular clock call early
3078 decrease_limits();
3079 }
3080
3081 void CMTask::print_stats() {
3082 gclog_or_tty->print_cr("Marking Stats, task = %u, calls = %d",
3083 _worker_id, _calls);
3084 gclog_or_tty->print_cr(" Elapsed time = %1.2lfms, Termination time = %1.2lfms",
3085 _elapsed_time_ms, _termination_time_ms);
3086 gclog_or_tty->print_cr(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms",
3087 _step_times_ms.num(), _step_times_ms.avg(),
3088 _step_times_ms.sd());
3089 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms",
3090 _step_times_ms.maximum(), _step_times_ms.sum());
3091 }
3092
3093 bool ConcurrentMark::try_stealing(uint worker_id, int* hash_seed, oop& obj) {
3094 return _task_queues->steal(worker_id, hash_seed, obj);
3095 }
3096
3097 /*****************************************************************************
3098
3099 The do_marking_step(time_target_ms, ...) method is the building
3100 block of the parallel marking framework. It can be called in parallel
3101 with other invocations of do_marking_step() on different tasks
3102 (but only one per task, obviously) and concurrently with the
3103 mutator threads, or during remark, hence it eliminates the need
3104 for two versions of the code. When called during remark, it will
3105 pick up from where the task left off during the concurrent marking
3106 phase. Interestingly, tasks are also claimable during evacuation
3107 pauses too, since do_marking_step() ensures that it aborts before
3108 it needs to yield.
3109
3110 The data structures that it uses to do marking work are the
3111 following:
3112
3113 (1) Marking Bitmap. If there are gray objects that appear only
3114 on the bitmap (this happens either when dealing with an overflow
3115 or when the initial marking phase has simply marked the roots
3116 and didn't push them on the stack), then tasks claim heap
3117 regions whose bitmap they then scan to find gray objects. A
3118 global finger indicates where the end of the last claimed region
3119 is. A local finger indicates how far into the region a task has
3120 scanned. The two fingers are used to determine how to gray an
3121 object (i.e. whether simply marking it is OK, as it will be
3122 visited by a task in the future, or whether it needs to be also
3123 pushed on a stack).
3124
3125 (2) Local Queue. The local queue of the task which is accessed
3126 reasonably efficiently by the task. Other tasks can steal from
3127 it when they run out of work. Throughout the marking phase, a
3128 task attempts to keep its local queue short but not totally
3129 empty, so that entries are available for stealing by other
3130 tasks. Only when there is no more work, a task will totally
3131 drain its local queue.
3132
3133 (3) Global Mark Stack. This handles local queue overflow. During
3134 marking only sets of entries are moved between it and the local
3135 queues, as access to it requires a mutex and more fine-grain
3136 interaction with it which might cause contention. If it
3137 overflows, then the marking phase should restart and iterate
3138 over the bitmap to identify gray objects. Throughout the marking
3139 phase, tasks attempt to keep the global mark stack at a small
3140 length but not totally empty, so that entries are available for
3141 popping by other tasks. Only when there is no more work, tasks
3142 will totally drain the global mark stack.
3143
3144 (4) SATB Buffer Queue. This is where completed SATB buffers are
3145 made available. Buffers are regularly removed from this queue
3146 and scanned for roots, so that the queue doesn't get too
3147 long. During remark, all completed buffers are processed, as
3148 well as the filled in parts of any uncompleted buffers.
3149
3150 The do_marking_step() method tries to abort when the time target
3151 has been reached. There are a few other cases when the
3152 do_marking_step() method also aborts:
3153
3154 (1) When the marking phase has been aborted (after a Full GC).
3155
3156 (2) When a global overflow (on the global stack) has been
3157 triggered. Before the task aborts, it will actually sync up with
3158 the other tasks to ensure that all the marking data structures
3159 (local queues, stacks, fingers etc.) are re-initialized so that
3160 when do_marking_step() completes, the marking phase can
3161 immediately restart.
3162
3163 (3) When enough completed SATB buffers are available. The
3164 do_marking_step() method only tries to drain SATB buffers right
3165 at the beginning. So, if enough buffers are available, the
3166 marking step aborts and the SATB buffers are processed at
3167 the beginning of the next invocation.
3168
3169 (4) To yield. when we have to yield then we abort and yield
3170 right at the end of do_marking_step(). This saves us from a lot
3171 of hassle as, by yielding we might allow a Full GC. If this
3172 happens then objects will be compacted underneath our feet, the
3173 heap might shrink, etc. We save checking for this by just
3174 aborting and doing the yield right at the end.
3175
3176 From the above it follows that the do_marking_step() method should
3177 be called in a loop (or, otherwise, regularly) until it completes.
3178
3179 If a marking step completes without its has_aborted() flag being
3180 true, it means it has completed the current marking phase (and
3181 also all other marking tasks have done so and have all synced up).
3182
3183 A method called regular_clock_call() is invoked "regularly" (in
3184 sub ms intervals) throughout marking. It is this clock method that
3185 checks all the abort conditions which were mentioned above and
3186 decides when the task should abort. A work-based scheme is used to
3187 trigger this clock method: when the number of object words the
3188 marking phase has scanned or the number of references the marking
3189 phase has visited reach a given limit. Additional invocations to
3190 the method clock have been planted in a few other strategic places
3191 too. The initial reason for the clock method was to avoid calling
3192 vtime too regularly, as it is quite expensive. So, once it was in
3193 place, it was natural to piggy-back all the other conditions on it
3194 too and not constantly check them throughout the code.
3195
3196 If do_termination is true then do_marking_step will enter its
3197 termination protocol.
3198
3199 The value of is_serial must be true when do_marking_step is being
3200 called serially (i.e. by the VMThread) and do_marking_step should
3201 skip any synchronization in the termination and overflow code.
3202 Examples include the serial remark code and the serial reference
3203 processing closures.
3204
3205 The value of is_serial must be false when do_marking_step is
3206 being called by any of the worker threads in a work gang.
3207 Examples include the concurrent marking code (CMMarkingTask),
3208 the MT remark code, and the MT reference processing closures.
3209
3210 *****************************************************************************/
3211
3212 void CMTask::do_marking_step(double time_target_ms,
3213 bool do_termination,
3214 bool is_serial) {
3215 assert(time_target_ms >= 1.0, "minimum granularity is 1ms");
3216 assert(concurrent() == _cm->concurrent(), "they should be the same");
3217
3218 G1CollectorPolicy* g1_policy = _g1h->g1_policy();
3219 assert(_task_queues != NULL, "invariant");
3220 assert(_task_queue != NULL, "invariant");
3221 assert(_task_queues->queue(_worker_id) == _task_queue, "invariant");
3222
3223 assert(!_claimed,
3224 "only one thread should claim this task at any one time");
3225
3226 // OK, this doesn't safeguard again all possible scenarios, as it is
3227 // possible for two threads to set the _claimed flag at the same
3228 // time. But it is only for debugging purposes anyway and it will
3229 // catch most problems.
3230 _claimed = true;
3231
3232 _start_time_ms = os::elapsedVTime() * 1000.0;
3233
3234 // If do_stealing is true then do_marking_step will attempt to
3235 // steal work from the other CMTasks. It only makes sense to
3236 // enable stealing when the termination protocol is enabled
3237 // and do_marking_step() is not being called serially.
3238 bool do_stealing = do_termination && !is_serial;
3239
3240 double diff_prediction_ms = _g1h->g1_policy()->predictor().get_new_prediction(&_marking_step_diffs_ms);
3241 _time_target_ms = time_target_ms - diff_prediction_ms;
3242
3243 // set up the variables that are used in the work-based scheme to
3244 // call the regular clock method
3245 _words_scanned = 0;
3246 _refs_reached = 0;
3247 recalculate_limits();
3248
3249 // clear all flags
3250 clear_has_aborted();
3251 _has_timed_out = false;
3252 _draining_satb_buffers = false;
3253
3254 ++_calls;
3255
3256 // Set up the bitmap and oop closures. Anything that uses them is
3257 // eventually called from this method, so it is OK to allocate these
3258 // statically.
3259 CMBitMapClosure bitmap_closure(this, _cm, _nextMarkBitMap);
3260 G1CMOopClosure cm_oop_closure(_g1h, _cm, this);
3261 set_cm_oop_closure(&cm_oop_closure);
3262
3263 if (_cm->has_overflown()) {
3264 // This can happen if the mark stack overflows during a GC pause
3265 // and this task, after a yield point, restarts. We have to abort
3266 // as we need to get into the overflow protocol which happens
3267 // right at the end of this task.
3268 set_has_aborted();
3269 }
3270
3271 // First drain any available SATB buffers. After this, we will not
3272 // look at SATB buffers before the next invocation of this method.
3273 // If enough completed SATB buffers are queued up, the regular clock
3274 // will abort this task so that it restarts.
3275 drain_satb_buffers();
3276 // ...then partially drain the local queue and the global stack
3277 drain_local_queue(true);
3278 drain_global_stack(true);
3279
3280 do {
3281 if (!has_aborted() && _curr_region != NULL) {
3282 // This means that we're already holding on to a region.
3283 assert(_finger != NULL, "if region is not NULL, then the finger "
3284 "should not be NULL either");
3285
3286 // We might have restarted this task after an evacuation pause
3287 // which might have evacuated the region we're holding on to
3288 // underneath our feet. Let's read its limit again to make sure
3289 // that we do not iterate over a region of the heap that
3290 // contains garbage (update_region_limit() will also move
3291 // _finger to the start of the region if it is found empty).
3292 update_region_limit();
3293 // We will start from _finger not from the start of the region,
3294 // as we might be restarting this task after aborting half-way
3295 // through scanning this region. In this case, _finger points to
3296 // the address where we last found a marked object. If this is a
3297 // fresh region, _finger points to start().
3298 MemRegion mr = MemRegion(_finger, _region_limit);
3299
3300 assert(!_curr_region->is_humongous() || mr.start() == _curr_region->bottom(),
3301 "humongous regions should go around loop once only");
3302
3303 // Some special cases:
3304 // If the memory region is empty, we can just give up the region.
3305 // If the current region is humongous then we only need to check
3306 // the bitmap for the bit associated with the start of the object,
3307 // scan the object if it's live, and give up the region.
3308 // Otherwise, let's iterate over the bitmap of the part of the region
3309 // that is left.
3310 // If the iteration is successful, give up the region.
3311 if (mr.is_empty()) {
3312 giveup_current_region();
3313 regular_clock_call();
3314 } else if (_curr_region->is_humongous() && mr.start() == _curr_region->bottom()) {
3315 if (_nextMarkBitMap->isMarked(mr.start())) {
3316 // The object is marked - apply the closure
3317 BitMap::idx_t offset = _nextMarkBitMap->heapWordToOffset(mr.start());
3318 bitmap_closure.do_bit(offset);
3319 }
3320 // Even if this task aborted while scanning the humongous object
3321 // we can (and should) give up the current region.
3322 giveup_current_region();
3323 regular_clock_call();
3324 } else if (_nextMarkBitMap->iterate(&bitmap_closure, mr)) {
3325 giveup_current_region();
3326 regular_clock_call();
3327 } else {
3328 assert(has_aborted(), "currently the only way to do so");
3329 // The only way to abort the bitmap iteration is to return
3330 // false from the do_bit() method. However, inside the
3331 // do_bit() method we move the _finger to point to the
3332 // object currently being looked at. So, if we bail out, we
3333 // have definitely set _finger to something non-null.
3334 assert(_finger != NULL, "invariant");
3335
3336 // Region iteration was actually aborted. So now _finger
3337 // points to the address of the object we last scanned. If we
3338 // leave it there, when we restart this task, we will rescan
3339 // the object. It is easy to avoid this. We move the finger by
3340 // enough to point to the next possible object header (the
3341 // bitmap knows by how much we need to move it as it knows its
3342 // granularity).
3343 assert(_finger < _region_limit, "invariant");
3344 HeapWord* new_finger = _nextMarkBitMap->nextObject(_finger);
3345 // Check if bitmap iteration was aborted while scanning the last object
3346 if (new_finger >= _region_limit) {
3347 giveup_current_region();
3348 } else {
3349 move_finger_to(new_finger);
3350 }
3351 }
3352 }
3353 // At this point we have either completed iterating over the
3354 // region we were holding on to, or we have aborted.
3355
3356 // We then partially drain the local queue and the global stack.
3357 // (Do we really need this?)
3358 drain_local_queue(true);
3359 drain_global_stack(true);
3360
3361 // Read the note on the claim_region() method on why it might
3362 // return NULL with potentially more regions available for
3363 // claiming and why we have to check out_of_regions() to determine
3364 // whether we're done or not.
3365 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) {
3366 // We are going to try to claim a new region. We should have
3367 // given up on the previous one.
3368 // Separated the asserts so that we know which one fires.
3369 assert(_curr_region == NULL, "invariant");
3370 assert(_finger == NULL, "invariant");
3371 assert(_region_limit == NULL, "invariant");
3372 HeapRegion* claimed_region = _cm->claim_region(_worker_id);
3373 if (claimed_region != NULL) {
3374 // Yes, we managed to claim one
3375 setup_for_region(claimed_region);
3376 assert(_curr_region == claimed_region, "invariant");
3377 }
3378 // It is important to call the regular clock here. It might take
3379 // a while to claim a region if, for example, we hit a large
3380 // block of empty regions. So we need to call the regular clock
3381 // method once round the loop to make sure it's called
3382 // frequently enough.
3383 regular_clock_call();
3384 }
3385
3386 if (!has_aborted() && _curr_region == NULL) {
3387 assert(_cm->out_of_regions(),
3388 "at this point we should be out of regions");
3389 }
3390 } while ( _curr_region != NULL && !has_aborted());
3391
3392 if (!has_aborted()) {
3393 // We cannot check whether the global stack is empty, since other
3394 // tasks might be pushing objects to it concurrently.
3395 assert(_cm->out_of_regions(),
3396 "at this point we should be out of regions");
3397 // Try to reduce the number of available SATB buffers so that
3398 // remark has less work to do.
3399 drain_satb_buffers();
3400 }
3401
3402 // Since we've done everything else, we can now totally drain the
3403 // local queue and global stack.
3404 drain_local_queue(false);
3405 drain_global_stack(false);
3406
3407 // Attempt at work stealing from other task's queues.
3408 if (do_stealing && !has_aborted()) {
3409 // We have not aborted. This means that we have finished all that
3410 // we could. Let's try to do some stealing...
3411
3412 // We cannot check whether the global stack is empty, since other
3413 // tasks might be pushing objects to it concurrently.
3414 assert(_cm->out_of_regions() && _task_queue->size() == 0,
3415 "only way to reach here");
3416 while (!has_aborted()) {
3417 oop obj;
3418 if (_cm->try_stealing(_worker_id, &_hash_seed, obj)) {
3419 assert(_nextMarkBitMap->isMarked((HeapWord*) obj),
3420 "any stolen object should be marked");
3421 scan_object(obj);
3422
3423 // And since we're towards the end, let's totally drain the
3424 // local queue and global stack.
3425 drain_local_queue(false);
3426 drain_global_stack(false);
3427 } else {
3428 break;
3429 }
3430 }
3431 }
3432
3433 // We still haven't aborted. Now, let's try to get into the
3434 // termination protocol.
3435 if (do_termination && !has_aborted()) {
3436 // We cannot check whether the global stack is empty, since other
3437 // tasks might be concurrently pushing objects on it.
3438 // Separated the asserts so that we know which one fires.
3439 assert(_cm->out_of_regions(), "only way to reach here");
3440 assert(_task_queue->size() == 0, "only way to reach here");
3441 _termination_start_time_ms = os::elapsedVTime() * 1000.0;
3442
3443 // The CMTask class also extends the TerminatorTerminator class,
3444 // hence its should_exit_termination() method will also decide
3445 // whether to exit the termination protocol or not.
3446 bool finished = (is_serial ||
3447 _cm->terminator()->offer_termination(this));
3448 double termination_end_time_ms = os::elapsedVTime() * 1000.0;
3449 _termination_time_ms +=
3450 termination_end_time_ms - _termination_start_time_ms;
3451
3452 if (finished) {
3453 // We're all done.
3454
3455 if (_worker_id == 0) {
3456 // let's allow task 0 to do this
3457 if (concurrent()) {
3458 assert(_cm->concurrent_marking_in_progress(), "invariant");
3459 // we need to set this to false before the next
3460 // safepoint. This way we ensure that the marking phase
3461 // doesn't observe any more heap expansions.
3462 _cm->clear_concurrent_marking_in_progress();
3463 }
3464 }
3465
3466 // We can now guarantee that the global stack is empty, since
3467 // all other tasks have finished. We separated the guarantees so
3468 // that, if a condition is false, we can immediately find out
3469 // which one.
3470 guarantee(_cm->out_of_regions(), "only way to reach here");
3471 guarantee(_cm->mark_stack_empty(), "only way to reach here");
3472 guarantee(_task_queue->size() == 0, "only way to reach here");
3473 guarantee(!_cm->has_overflown(), "only way to reach here");
3474 guarantee(!_cm->mark_stack_overflow(), "only way to reach here");
3475 } else {
3476 // Apparently there's more work to do. Let's abort this task. It
3477 // will restart it and we can hopefully find more things to do.
3478 set_has_aborted();
3479 }
3480 }
3481
3482 // Mainly for debugging purposes to make sure that a pointer to the
3483 // closure which was statically allocated in this frame doesn't
3484 // escape it by accident.
3485 set_cm_oop_closure(NULL);
3486 double end_time_ms = os::elapsedVTime() * 1000.0;
3487 double elapsed_time_ms = end_time_ms - _start_time_ms;
3488 // Update the step history.
3489 _step_times_ms.add(elapsed_time_ms);
3490
3491 if (has_aborted()) {
3492 // The task was aborted for some reason.
3493 if (_has_timed_out) {
3494 double diff_ms = elapsed_time_ms - _time_target_ms;
3495 // Keep statistics of how well we did with respect to hitting
3496 // our target only if we actually timed out (if we aborted for
3497 // other reasons, then the results might get skewed).
3498 _marking_step_diffs_ms.add(diff_ms);
3499 }
3500
3501 if (_cm->has_overflown()) {
3502 // This is the interesting one. We aborted because a global
3503 // overflow was raised. This means we have to restart the
3504 // marking phase and start iterating over regions. However, in
3505 // order to do this we have to make sure that all tasks stop
3506 // what they are doing and re-initialize in a safe manner. We
3507 // will achieve this with the use of two barrier sync points.
3508
3509 if (!is_serial) {
3510 // We only need to enter the sync barrier if being called
3511 // from a parallel context
3512 _cm->enter_first_sync_barrier(_worker_id);
3513
3514 // When we exit this sync barrier we know that all tasks have
3515 // stopped doing marking work. So, it's now safe to
3516 // re-initialize our data structures. At the end of this method,
3517 // task 0 will clear the global data structures.
3518 }
3519
3520 // We clear the local state of this task...
3521 clear_region_fields();
3522
3523 if (!is_serial) {
3524 // ...and enter the second barrier.
3525 _cm->enter_second_sync_barrier(_worker_id);
3526 }
3527 // At this point, if we're during the concurrent phase of
3528 // marking, everything has been re-initialized and we're
3529 // ready to restart.
3530 }
3531 }
3532
3533 _claimed = false;
3534 }
3535
3536 CMTask::CMTask(uint worker_id,
3537 ConcurrentMark* cm,
3538 size_t* marked_bytes,
3539 BitMap* card_bm,
3540 CMTaskQueue* task_queue,
3541 CMTaskQueueSet* task_queues)
3542 : _g1h(G1CollectedHeap::heap()),
3543 _worker_id(worker_id), _cm(cm),
3544 _claimed(false),
3545 _nextMarkBitMap(NULL), _hash_seed(17),
3546 _task_queue(task_queue),
3547 _task_queues(task_queues),
3548 _cm_oop_closure(NULL),
3549 _marked_bytes_array(marked_bytes),
3550 _card_bm(card_bm) {
3551 guarantee(task_queue != NULL, "invariant");
3552 guarantee(task_queues != NULL, "invariant");
3553
3554 _marking_step_diffs_ms.add(0.5);
3555 }
3556
3557 // These are formatting macros that are used below to ensure
3558 // consistent formatting. The *_H_* versions are used to format the
3559 // header for a particular value and they should be kept consistent
3560 // with the corresponding macro. Also note that most of the macros add
3561 // the necessary white space (as a prefix) which makes them a bit
3562 // easier to compose.
3563
3564 // All the output lines are prefixed with this string to be able to
3565 // identify them easily in a large log file.
3566 #define G1PPRL_LINE_PREFIX "###"
3567
3568 #define G1PPRL_ADDR_BASE_FORMAT " " PTR_FORMAT "-" PTR_FORMAT
3569 #ifdef _LP64
3570 #define G1PPRL_ADDR_BASE_H_FORMAT " %37s"
3571 #else // _LP64
3572 #define G1PPRL_ADDR_BASE_H_FORMAT " %21s"
3573 #endif // _LP64
3574
3575 // For per-region info
3576 #define G1PPRL_TYPE_FORMAT " %-4s"
3577 #define G1PPRL_TYPE_H_FORMAT " %4s"
3578 #define G1PPRL_BYTE_FORMAT " " SIZE_FORMAT_W(9)
3579 #define G1PPRL_BYTE_H_FORMAT " %9s"
3580 #define G1PPRL_DOUBLE_FORMAT " %14.1f"
3581 #define G1PPRL_DOUBLE_H_FORMAT " %14s"
3582
3583 // For summary info
3584 #define G1PPRL_SUM_ADDR_FORMAT(tag) " " tag ":" G1PPRL_ADDR_BASE_FORMAT
3585 #define G1PPRL_SUM_BYTE_FORMAT(tag) " " tag ": " SIZE_FORMAT
3586 #define G1PPRL_SUM_MB_FORMAT(tag) " " tag ": %1.2f MB"
3587 #define G1PPRL_SUM_MB_PERC_FORMAT(tag) G1PPRL_SUM_MB_FORMAT(tag) " / %1.2f %%"
3588
3589 G1PrintRegionLivenessInfoClosure::
3590 G1PrintRegionLivenessInfoClosure(outputStream* out, const char* phase_name)
3591 : _out(out),
3592 _total_used_bytes(0), _total_capacity_bytes(0),
3593 _total_prev_live_bytes(0), _total_next_live_bytes(0),
3594 _hum_used_bytes(0), _hum_capacity_bytes(0),
3595 _hum_prev_live_bytes(0), _hum_next_live_bytes(0),
3596 _total_remset_bytes(0), _total_strong_code_roots_bytes(0) {
3597 G1CollectedHeap* g1h = G1CollectedHeap::heap();
3598 MemRegion g1_reserved = g1h->g1_reserved();
3599 double now = os::elapsedTime();
3600
3601 // Print the header of the output.
3602 _out->cr();
3603 _out->print_cr(G1PPRL_LINE_PREFIX" PHASE %s @ %1.3f", phase_name, now);
3604 _out->print_cr(G1PPRL_LINE_PREFIX" HEAP"
3605 G1PPRL_SUM_ADDR_FORMAT("reserved")
3606 G1PPRL_SUM_BYTE_FORMAT("region-size"),
3607 p2i(g1_reserved.start()), p2i(g1_reserved.end()),
3608 HeapRegion::GrainBytes);
3609 _out->print_cr(G1PPRL_LINE_PREFIX);
3610 _out->print_cr(G1PPRL_LINE_PREFIX
3611 G1PPRL_TYPE_H_FORMAT
3612 G1PPRL_ADDR_BASE_H_FORMAT
3613 G1PPRL_BYTE_H_FORMAT
3614 G1PPRL_BYTE_H_FORMAT
3615 G1PPRL_BYTE_H_FORMAT
3616 G1PPRL_DOUBLE_H_FORMAT
3617 G1PPRL_BYTE_H_FORMAT
3618 G1PPRL_BYTE_H_FORMAT,
3619 "type", "address-range",
3620 "used", "prev-live", "next-live", "gc-eff",
3621 "remset", "code-roots");
3622 _out->print_cr(G1PPRL_LINE_PREFIX
3623 G1PPRL_TYPE_H_FORMAT
3624 G1PPRL_ADDR_BASE_H_FORMAT
3625 G1PPRL_BYTE_H_FORMAT
3626 G1PPRL_BYTE_H_FORMAT
3627 G1PPRL_BYTE_H_FORMAT
3628 G1PPRL_DOUBLE_H_FORMAT
3629 G1PPRL_BYTE_H_FORMAT
3630 G1PPRL_BYTE_H_FORMAT,
3631 "", "",
3632 "(bytes)", "(bytes)", "(bytes)", "(bytes/ms)",
3633 "(bytes)", "(bytes)");
3634 }
3635
3636 // It takes as a parameter a reference to one of the _hum_* fields, it
3637 // deduces the corresponding value for a region in a humongous region
3638 // series (either the region size, or what's left if the _hum_* field
3639 // is < the region size), and updates the _hum_* field accordingly.
3640 size_t G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* hum_bytes) {
3641 size_t bytes = 0;
3642 // The > 0 check is to deal with the prev and next live bytes which
3643 // could be 0.
3644 if (*hum_bytes > 0) {
3645 bytes = MIN2(HeapRegion::GrainBytes, *hum_bytes);
3646 *hum_bytes -= bytes;
3647 }
3648 return bytes;
3649 }
3650
3651 // It deduces the values for a region in a humongous region series
3652 // from the _hum_* fields and updates those accordingly. It assumes
3653 // that that _hum_* fields have already been set up from the "starts
3654 // humongous" region and we visit the regions in address order.
3655 void G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* used_bytes,
3656 size_t* capacity_bytes,
3657 size_t* prev_live_bytes,
3658 size_t* next_live_bytes) {
3659 assert(_hum_used_bytes > 0 && _hum_capacity_bytes > 0, "pre-condition");
3660 *used_bytes = get_hum_bytes(&_hum_used_bytes);
3661 *capacity_bytes = get_hum_bytes(&_hum_capacity_bytes);
3662 *prev_live_bytes = get_hum_bytes(&_hum_prev_live_bytes);
3663 *next_live_bytes = get_hum_bytes(&_hum_next_live_bytes);
3664 }
3665
3666 bool G1PrintRegionLivenessInfoClosure::doHeapRegion(HeapRegion* r) {
3667 const char* type = r->get_type_str();
3668 HeapWord* bottom = r->bottom();
3669 HeapWord* end = r->end();
3670 size_t capacity_bytes = r->capacity();
3671 size_t used_bytes = r->used();
3672 size_t prev_live_bytes = r->live_bytes();
3673 size_t next_live_bytes = r->next_live_bytes();
3674 double gc_eff = r->gc_efficiency();
3675 size_t remset_bytes = r->rem_set()->mem_size();
3676 size_t strong_code_roots_bytes = r->rem_set()->strong_code_roots_mem_size();
3677
3678 if (r->is_starts_humongous()) {
3679 assert(_hum_used_bytes == 0 && _hum_capacity_bytes == 0 &&
3680 _hum_prev_live_bytes == 0 && _hum_next_live_bytes == 0,
3681 "they should have been zeroed after the last time we used them");
3682 // Set up the _hum_* fields.
3683 _hum_capacity_bytes = capacity_bytes;
3684 _hum_used_bytes = used_bytes;
3685 _hum_prev_live_bytes = prev_live_bytes;
3686 _hum_next_live_bytes = next_live_bytes;
3687 get_hum_bytes(&used_bytes, &capacity_bytes,
3688 &prev_live_bytes, &next_live_bytes);
3689 end = bottom + HeapRegion::GrainWords;
3690 } else if (r->is_continues_humongous()) {
3691 get_hum_bytes(&used_bytes, &capacity_bytes,
3692 &prev_live_bytes, &next_live_bytes);
3693 assert(end == bottom + HeapRegion::GrainWords, "invariant");
3694 }
3695
3696 _total_used_bytes += used_bytes;
3697 _total_capacity_bytes += capacity_bytes;
3698 _total_prev_live_bytes += prev_live_bytes;
3699 _total_next_live_bytes += next_live_bytes;
3700 _total_remset_bytes += remset_bytes;
3701 _total_strong_code_roots_bytes += strong_code_roots_bytes;
3702
3703 // Print a line for this particular region.
3704 _out->print_cr(G1PPRL_LINE_PREFIX
3705 G1PPRL_TYPE_FORMAT
3706 G1PPRL_ADDR_BASE_FORMAT
3707 G1PPRL_BYTE_FORMAT
3708 G1PPRL_BYTE_FORMAT
3709 G1PPRL_BYTE_FORMAT
3710 G1PPRL_DOUBLE_FORMAT
3711 G1PPRL_BYTE_FORMAT
3712 G1PPRL_BYTE_FORMAT,
3713 type, p2i(bottom), p2i(end),
3714 used_bytes, prev_live_bytes, next_live_bytes, gc_eff,
3715 remset_bytes, strong_code_roots_bytes);
3716
3717 return false;
3718 }
3719
3720 G1PrintRegionLivenessInfoClosure::~G1PrintRegionLivenessInfoClosure() {
3721 // add static memory usages to remembered set sizes
3722 _total_remset_bytes += HeapRegionRemSet::fl_mem_size() + HeapRegionRemSet::static_mem_size();
3723 // Print the footer of the output.
3724 _out->print_cr(G1PPRL_LINE_PREFIX);
3725 _out->print_cr(G1PPRL_LINE_PREFIX
3726 " SUMMARY"
3727 G1PPRL_SUM_MB_FORMAT("capacity")
3728 G1PPRL_SUM_MB_PERC_FORMAT("used")
3729 G1PPRL_SUM_MB_PERC_FORMAT("prev-live")
3730 G1PPRL_SUM_MB_PERC_FORMAT("next-live")
3731 G1PPRL_SUM_MB_FORMAT("remset")
3732 G1PPRL_SUM_MB_FORMAT("code-roots"),
3733 bytes_to_mb(_total_capacity_bytes),
3734 bytes_to_mb(_total_used_bytes),
3735 perc(_total_used_bytes, _total_capacity_bytes),
3736 bytes_to_mb(_total_prev_live_bytes),
3737 perc(_total_prev_live_bytes, _total_capacity_bytes),
3738 bytes_to_mb(_total_next_live_bytes),
3739 perc(_total_next_live_bytes, _total_capacity_bytes),
3740 bytes_to_mb(_total_remset_bytes),
3741 bytes_to_mb(_total_strong_code_roots_bytes));
3742 _out->cr();
3743 }