Mercurial > hg > truffle
annotate src/share/vm/gc_implementation/g1/concurrentMark.cpp @ 912:308762b2bf14
6872000: G1: compilation fails on linux/older gcc
Reviewed-by: jcoomes, tonyp
author | apetrusenko |
---|---|
date | Fri, 14 Aug 2009 13:44:15 -0700 |
parents | 18f526145aea |
children | e1fdf4fd34dc 148e5441d916 |
rev | line source |
---|---|
342 | 1 /* |
844 | 2 * Copyright 2001-2009 Sun Microsystems, Inc. All Rights Reserved. |
342 | 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 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, | |
20 * CA 95054 USA or visit www.sun.com if you need additional information or | |
21 * have any questions. | |
22 * | |
23 */ | |
24 | |
25 #include "incls/_precompiled.incl" | |
26 #include "incls/_concurrentMark.cpp.incl" | |
27 | |
28 // | |
29 // CMS Bit Map Wrapper | |
30 | |
31 CMBitMapRO::CMBitMapRO(ReservedSpace rs, int shifter): | |
32 _bm((uintptr_t*)NULL,0), | |
33 _shifter(shifter) { | |
34 _bmStartWord = (HeapWord*)(rs.base()); | |
35 _bmWordSize = rs.size()/HeapWordSize; // rs.size() is in bytes | |
36 ReservedSpace brs(ReservedSpace::allocation_align_size_up( | |
37 (_bmWordSize >> (_shifter + LogBitsPerByte)) + 1)); | |
38 | |
39 guarantee(brs.is_reserved(), "couldn't allocate CMS bit map"); | |
40 // For now we'll just commit all of the bit map up fromt. | |
41 // Later on we'll try to be more parsimonious with swap. | |
42 guarantee(_virtual_space.initialize(brs, brs.size()), | |
43 "couldn't reseve backing store for CMS bit map"); | |
44 assert(_virtual_space.committed_size() == brs.size(), | |
45 "didn't reserve backing store for all of CMS bit map?"); | |
46 _bm.set_map((uintptr_t*)_virtual_space.low()); | |
47 assert(_virtual_space.committed_size() << (_shifter + LogBitsPerByte) >= | |
48 _bmWordSize, "inconsistency in bit map sizing"); | |
49 _bm.set_size(_bmWordSize >> _shifter); | |
50 } | |
51 | |
52 HeapWord* CMBitMapRO::getNextMarkedWordAddress(HeapWord* addr, | |
53 HeapWord* limit) const { | |
54 // First we must round addr *up* to a possible object boundary. | |
55 addr = (HeapWord*)align_size_up((intptr_t)addr, | |
56 HeapWordSize << _shifter); | |
57 size_t addrOffset = heapWordToOffset(addr); | |
58 if (limit == NULL) limit = _bmStartWord + _bmWordSize; | |
59 size_t limitOffset = heapWordToOffset(limit); | |
60 size_t nextOffset = _bm.get_next_one_offset(addrOffset, limitOffset); | |
61 HeapWord* nextAddr = offsetToHeapWord(nextOffset); | |
62 assert(nextAddr >= addr, "get_next_one postcondition"); | |
63 assert(nextAddr == limit || isMarked(nextAddr), | |
64 "get_next_one postcondition"); | |
65 return nextAddr; | |
66 } | |
67 | |
68 HeapWord* CMBitMapRO::getNextUnmarkedWordAddress(HeapWord* addr, | |
69 HeapWord* limit) const { | |
70 size_t addrOffset = heapWordToOffset(addr); | |
71 if (limit == NULL) limit = _bmStartWord + _bmWordSize; | |
72 size_t limitOffset = heapWordToOffset(limit); | |
73 size_t nextOffset = _bm.get_next_zero_offset(addrOffset, limitOffset); | |
74 HeapWord* nextAddr = offsetToHeapWord(nextOffset); | |
75 assert(nextAddr >= addr, "get_next_one postcondition"); | |
76 assert(nextAddr == limit || !isMarked(nextAddr), | |
77 "get_next_one postcondition"); | |
78 return nextAddr; | |
79 } | |
80 | |
81 int CMBitMapRO::heapWordDiffToOffsetDiff(size_t diff) const { | |
82 assert((diff & ((1 << _shifter) - 1)) == 0, "argument check"); | |
83 return (int) (diff >> _shifter); | |
84 } | |
85 | |
86 bool CMBitMapRO::iterate(BitMapClosure* cl, MemRegion mr) { | |
87 HeapWord* left = MAX2(_bmStartWord, mr.start()); | |
88 HeapWord* right = MIN2(_bmStartWord + _bmWordSize, mr.end()); | |
89 if (right > left) { | |
90 // Right-open interval [leftOffset, rightOffset). | |
91 return _bm.iterate(cl, heapWordToOffset(left), heapWordToOffset(right)); | |
92 } else { | |
93 return true; | |
94 } | |
95 } | |
96 | |
97 void CMBitMapRO::mostly_disjoint_range_union(BitMap* from_bitmap, | |
98 size_t from_start_index, | |
99 HeapWord* to_start_word, | |
100 size_t word_num) { | |
101 _bm.mostly_disjoint_range_union(from_bitmap, | |
102 from_start_index, | |
103 heapWordToOffset(to_start_word), | |
104 word_num); | |
105 } | |
106 | |
107 #ifndef PRODUCT | |
108 bool CMBitMapRO::covers(ReservedSpace rs) const { | |
109 // assert(_bm.map() == _virtual_space.low(), "map inconsistency"); | |
645
c3a720eefe82
6816308: Changes to allow builds with latest Windows SDK 6.1 on 64bit Windows 2003
kvn
parents:
619
diff
changeset
|
110 assert(((size_t)_bm.size() * (size_t)(1 << _shifter)) == _bmWordSize, |
342 | 111 "size inconsistency"); |
112 return _bmStartWord == (HeapWord*)(rs.base()) && | |
113 _bmWordSize == rs.size()>>LogHeapWordSize; | |
114 } | |
115 #endif | |
116 | |
117 void CMBitMap::clearAll() { | |
118 _bm.clear(); | |
119 return; | |
120 } | |
121 | |
122 void CMBitMap::markRange(MemRegion mr) { | |
123 mr.intersection(MemRegion(_bmStartWord, _bmWordSize)); | |
124 assert(!mr.is_empty(), "unexpected empty region"); | |
125 assert((offsetToHeapWord(heapWordToOffset(mr.end())) == | |
126 ((HeapWord *) mr.end())), | |
127 "markRange memory region end is not card aligned"); | |
128 // convert address range into offset range | |
129 _bm.at_put_range(heapWordToOffset(mr.start()), | |
130 heapWordToOffset(mr.end()), true); | |
131 } | |
132 | |
133 void CMBitMap::clearRange(MemRegion mr) { | |
134 mr.intersection(MemRegion(_bmStartWord, _bmWordSize)); | |
135 assert(!mr.is_empty(), "unexpected empty region"); | |
136 // convert address range into offset range | |
137 _bm.at_put_range(heapWordToOffset(mr.start()), | |
138 heapWordToOffset(mr.end()), false); | |
139 } | |
140 | |
141 MemRegion CMBitMap::getAndClearMarkedRegion(HeapWord* addr, | |
142 HeapWord* end_addr) { | |
143 HeapWord* start = getNextMarkedWordAddress(addr); | |
144 start = MIN2(start, end_addr); | |
145 HeapWord* end = getNextUnmarkedWordAddress(start); | |
146 end = MIN2(end, end_addr); | |
147 assert(start <= end, "Consistency check"); | |
148 MemRegion mr(start, end); | |
149 if (!mr.is_empty()) { | |
150 clearRange(mr); | |
151 } | |
152 return mr; | |
153 } | |
154 | |
155 CMMarkStack::CMMarkStack(ConcurrentMark* cm) : | |
156 _base(NULL), _cm(cm) | |
157 #ifdef ASSERT | |
158 , _drain_in_progress(false) | |
159 , _drain_in_progress_yields(false) | |
160 #endif | |
161 {} | |
162 | |
163 void CMMarkStack::allocate(size_t size) { | |
164 _base = NEW_C_HEAP_ARRAY(oop, size); | |
165 if (_base == NULL) | |
166 vm_exit_during_initialization("Failed to allocate " | |
167 "CM region mark stack"); | |
168 _index = 0; | |
169 // QQQQ cast ... | |
170 _capacity = (jint) size; | |
171 _oops_do_bound = -1; | |
172 NOT_PRODUCT(_max_depth = 0); | |
173 } | |
174 | |
175 CMMarkStack::~CMMarkStack() { | |
176 if (_base != NULL) FREE_C_HEAP_ARRAY(oop, _base); | |
177 } | |
178 | |
179 void CMMarkStack::par_push(oop ptr) { | |
180 while (true) { | |
181 if (isFull()) { | |
182 _overflow = true; | |
183 return; | |
184 } | |
185 // Otherwise... | |
186 jint index = _index; | |
187 jint next_index = index+1; | |
188 jint res = Atomic::cmpxchg(next_index, &_index, index); | |
189 if (res == index) { | |
190 _base[index] = ptr; | |
191 // Note that we don't maintain this atomically. We could, but it | |
192 // doesn't seem necessary. | |
193 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index)); | |
194 return; | |
195 } | |
196 // Otherwise, we need to try again. | |
197 } | |
198 } | |
199 | |
200 void CMMarkStack::par_adjoin_arr(oop* ptr_arr, int n) { | |
201 while (true) { | |
202 if (isFull()) { | |
203 _overflow = true; | |
204 return; | |
205 } | |
206 // Otherwise... | |
207 jint index = _index; | |
208 jint next_index = index + n; | |
209 if (next_index > _capacity) { | |
210 _overflow = true; | |
211 return; | |
212 } | |
213 jint res = Atomic::cmpxchg(next_index, &_index, index); | |
214 if (res == index) { | |
215 for (int i = 0; i < n; i++) { | |
216 int ind = index + i; | |
217 assert(ind < _capacity, "By overflow test above."); | |
218 _base[ind] = ptr_arr[i]; | |
219 } | |
220 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index)); | |
221 return; | |
222 } | |
223 // Otherwise, we need to try again. | |
224 } | |
225 } | |
226 | |
227 | |
228 void CMMarkStack::par_push_arr(oop* ptr_arr, int n) { | |
229 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); | |
230 jint start = _index; | |
231 jint next_index = start + n; | |
232 if (next_index > _capacity) { | |
233 _overflow = true; | |
234 return; | |
235 } | |
236 // Otherwise. | |
237 _index = next_index; | |
238 for (int i = 0; i < n; i++) { | |
239 int ind = start + i; | |
240 guarantee(ind < _capacity, "By overflow test above."); | |
241 _base[ind] = ptr_arr[i]; | |
242 } | |
243 } | |
244 | |
245 | |
246 bool CMMarkStack::par_pop_arr(oop* ptr_arr, int max, int* n) { | |
247 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); | |
248 jint index = _index; | |
249 if (index == 0) { | |
250 *n = 0; | |
251 return false; | |
252 } else { | |
253 int k = MIN2(max, index); | |
254 jint new_ind = index - k; | |
255 for (int j = 0; j < k; j++) { | |
256 ptr_arr[j] = _base[new_ind + j]; | |
257 } | |
258 _index = new_ind; | |
259 *n = k; | |
260 return true; | |
261 } | |
262 } | |
263 | |
264 | |
265 CMRegionStack::CMRegionStack() : _base(NULL) {} | |
266 | |
267 void CMRegionStack::allocate(size_t size) { | |
268 _base = NEW_C_HEAP_ARRAY(MemRegion, size); | |
269 if (_base == NULL) | |
270 vm_exit_during_initialization("Failed to allocate " | |
271 "CM region mark stack"); | |
272 _index = 0; | |
273 // QQQQ cast ... | |
274 _capacity = (jint) size; | |
275 } | |
276 | |
277 CMRegionStack::~CMRegionStack() { | |
278 if (_base != NULL) FREE_C_HEAP_ARRAY(oop, _base); | |
279 } | |
280 | |
281 void CMRegionStack::push(MemRegion mr) { | |
282 assert(mr.word_size() > 0, "Precondition"); | |
283 while (true) { | |
284 if (isFull()) { | |
285 _overflow = true; | |
286 return; | |
287 } | |
288 // Otherwise... | |
289 jint index = _index; | |
290 jint next_index = index+1; | |
291 jint res = Atomic::cmpxchg(next_index, &_index, index); | |
292 if (res == index) { | |
293 _base[index] = mr; | |
294 return; | |
295 } | |
296 // Otherwise, we need to try again. | |
297 } | |
298 } | |
299 | |
300 MemRegion CMRegionStack::pop() { | |
301 while (true) { | |
302 // Otherwise... | |
303 jint index = _index; | |
304 | |
305 if (index == 0) { | |
306 return MemRegion(); | |
307 } | |
308 jint next_index = index-1; | |
309 jint res = Atomic::cmpxchg(next_index, &_index, index); | |
310 if (res == index) { | |
311 MemRegion mr = _base[next_index]; | |
312 if (mr.start() != NULL) { | |
313 tmp_guarantee_CM( mr.end() != NULL, "invariant" ); | |
314 tmp_guarantee_CM( mr.word_size() > 0, "invariant" ); | |
315 return mr; | |
316 } else { | |
317 // that entry was invalidated... let's skip it | |
318 tmp_guarantee_CM( mr.end() == NULL, "invariant" ); | |
319 } | |
320 } | |
321 // Otherwise, we need to try again. | |
322 } | |
323 } | |
324 | |
325 bool CMRegionStack::invalidate_entries_into_cset() { | |
326 bool result = false; | |
327 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
328 for (int i = 0; i < _oops_do_bound; ++i) { | |
329 MemRegion mr = _base[i]; | |
330 if (mr.start() != NULL) { | |
331 tmp_guarantee_CM( mr.end() != NULL, "invariant"); | |
332 tmp_guarantee_CM( mr.word_size() > 0, "invariant" ); | |
333 HeapRegion* hr = g1h->heap_region_containing(mr.start()); | |
334 tmp_guarantee_CM( hr != NULL, "invariant" ); | |
335 if (hr->in_collection_set()) { | |
336 // The region points into the collection set | |
337 _base[i] = MemRegion(); | |
338 result = true; | |
339 } | |
340 } else { | |
341 // that entry was invalidated... let's skip it | |
342 tmp_guarantee_CM( mr.end() == NULL, "invariant" ); | |
343 } | |
344 } | |
345 return result; | |
346 } | |
347 | |
348 template<class OopClosureClass> | |
349 bool CMMarkStack::drain(OopClosureClass* cl, CMBitMap* bm, bool yield_after) { | |
350 assert(!_drain_in_progress || !_drain_in_progress_yields || yield_after | |
351 || SafepointSynchronize::is_at_safepoint(), | |
352 "Drain recursion must be yield-safe."); | |
353 bool res = true; | |
354 debug_only(_drain_in_progress = true); | |
355 debug_only(_drain_in_progress_yields = yield_after); | |
356 while (!isEmpty()) { | |
357 oop newOop = pop(); | |
358 assert(G1CollectedHeap::heap()->is_in_reserved(newOop), "Bad pop"); | |
359 assert(newOop->is_oop(), "Expected an oop"); | |
360 assert(bm == NULL || bm->isMarked((HeapWord*)newOop), | |
361 "only grey objects on this stack"); | |
362 // iterate over the oops in this oop, marking and pushing | |
363 // the ones in CMS generation. | |
364 newOop->oop_iterate(cl); | |
365 if (yield_after && _cm->do_yield_check()) { | |
366 res = false; break; | |
367 } | |
368 } | |
369 debug_only(_drain_in_progress = false); | |
370 return res; | |
371 } | |
372 | |
373 void CMMarkStack::oops_do(OopClosure* f) { | |
374 if (_index == 0) return; | |
375 assert(_oops_do_bound != -1 && _oops_do_bound <= _index, | |
376 "Bound must be set."); | |
377 for (int i = 0; i < _oops_do_bound; i++) { | |
378 f->do_oop(&_base[i]); | |
379 } | |
380 _oops_do_bound = -1; | |
381 } | |
382 | |
383 bool ConcurrentMark::not_yet_marked(oop obj) const { | |
384 return (_g1h->is_obj_ill(obj) | |
385 || (_g1h->is_in_permanent(obj) | |
386 && !nextMarkBitMap()->isMarked((HeapWord*)obj))); | |
387 } | |
388 | |
389 #ifdef _MSC_VER // the use of 'this' below gets a warning, make it go away | |
390 #pragma warning( disable:4355 ) // 'this' : used in base member initializer list | |
391 #endif // _MSC_VER | |
392 | |
393 ConcurrentMark::ConcurrentMark(ReservedSpace rs, | |
394 int max_regions) : | |
395 _markBitMap1(rs, MinObjAlignment - 1), | |
396 _markBitMap2(rs, MinObjAlignment - 1), | |
397 | |
398 _parallel_marking_threads(0), | |
399 _sleep_factor(0.0), | |
400 _marking_task_overhead(1.0), | |
401 _cleanup_sleep_factor(0.0), | |
402 _cleanup_task_overhead(1.0), | |
403 _region_bm(max_regions, false /* in_resource_area*/), | |
404 _card_bm((rs.size() + CardTableModRefBS::card_size - 1) >> | |
405 CardTableModRefBS::card_shift, | |
406 false /* in_resource_area*/), | |
407 _prevMarkBitMap(&_markBitMap1), | |
408 _nextMarkBitMap(&_markBitMap2), | |
409 _at_least_one_mark_complete(false), | |
410 | |
411 _markStack(this), | |
412 _regionStack(), | |
413 // _finger set in set_non_marking_state | |
414 | |
415 _max_task_num(MAX2(ParallelGCThreads, (size_t)1)), | |
416 // _active_tasks set in set_non_marking_state | |
417 // _tasks set inside the constructor | |
418 _task_queues(new CMTaskQueueSet((int) _max_task_num)), | |
419 _terminator(ParallelTaskTerminator((int) _max_task_num, _task_queues)), | |
420 | |
421 _has_overflown(false), | |
422 _concurrent(false), | |
619
7ea5ca260b28
6814467: G1: small fixes related to concurrent marking verboseness
tonyp
parents:
470
diff
changeset
|
423 _has_aborted(false), |
7ea5ca260b28
6814467: G1: small fixes related to concurrent marking verboseness
tonyp
parents:
470
diff
changeset
|
424 _restart_for_overflow(false), |
7ea5ca260b28
6814467: G1: small fixes related to concurrent marking verboseness
tonyp
parents:
470
diff
changeset
|
425 _concurrent_marking_in_progress(false), |
7ea5ca260b28
6814467: G1: small fixes related to concurrent marking verboseness
tonyp
parents:
470
diff
changeset
|
426 _should_gray_objects(false), |
342 | 427 |
428 // _verbose_level set below | |
429 | |
430 _init_times(), | |
431 _remark_times(), _remark_mark_times(), _remark_weak_ref_times(), | |
432 _cleanup_times(), | |
433 _total_counting_time(0.0), | |
434 _total_rs_scrub_time(0.0), | |
435 | |
436 _parallel_workers(NULL), | |
437 _cleanup_co_tracker(G1CLGroup) | |
438 { | |
439 CMVerboseLevel verbose_level = | |
440 (CMVerboseLevel) G1MarkingVerboseLevel; | |
441 if (verbose_level < no_verbose) | |
442 verbose_level = no_verbose; | |
443 if (verbose_level > high_verbose) | |
444 verbose_level = high_verbose; | |
445 _verbose_level = verbose_level; | |
446 | |
447 if (verbose_low()) | |
448 gclog_or_tty->print_cr("[global] init, heap start = "PTR_FORMAT", " | |
449 "heap end = "PTR_FORMAT, _heap_start, _heap_end); | |
450 | |
751 | 451 _markStack.allocate(G1MarkStackSize); |
452 _regionStack.allocate(G1MarkRegionStackSize); | |
342 | 453 |
454 // Create & start a ConcurrentMark thread. | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
455 _cmThread = new ConcurrentMarkThread(this); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
456 assert(cmThread() != NULL, "CM Thread should have been created"); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
457 assert(cmThread()->cm() != NULL, "CM Thread should refer to this cm"); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
458 |
342 | 459 _g1h = G1CollectedHeap::heap(); |
460 assert(CGC_lock != NULL, "Where's the CGC_lock?"); | |
461 assert(_markBitMap1.covers(rs), "_markBitMap1 inconsistency"); | |
462 assert(_markBitMap2.covers(rs), "_markBitMap2 inconsistency"); | |
463 | |
464 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set(); | |
465 satb_qs.set_buffer_size(G1SATBLogBufferSize); | |
466 | |
467 int size = (int) MAX2(ParallelGCThreads, (size_t)1); | |
468 _par_cleanup_thread_state = NEW_C_HEAP_ARRAY(ParCleanupThreadState*, size); | |
469 for (int i = 0 ; i < size; i++) { | |
470 _par_cleanup_thread_state[i] = new ParCleanupThreadState; | |
471 } | |
472 | |
473 _tasks = NEW_C_HEAP_ARRAY(CMTask*, _max_task_num); | |
474 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_task_num); | |
475 | |
476 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail | |
477 _active_tasks = _max_task_num; | |
478 for (int i = 0; i < (int) _max_task_num; ++i) { | |
479 CMTaskQueue* task_queue = new CMTaskQueue(); | |
480 task_queue->initialize(); | |
481 _task_queues->register_queue(i, task_queue); | |
482 | |
483 _tasks[i] = new CMTask(i, this, task_queue, _task_queues); | |
484 _accum_task_vtime[i] = 0.0; | |
485 } | |
486 | |
487 if (ParallelMarkingThreads > ParallelGCThreads) { | |
488 vm_exit_during_initialization("Can't have more ParallelMarkingThreads " | |
489 "than ParallelGCThreads."); | |
490 } | |
491 if (ParallelGCThreads == 0) { | |
492 // if we are not running with any parallel GC threads we will not | |
493 // spawn any marking threads either | |
494 _parallel_marking_threads = 0; | |
495 _sleep_factor = 0.0; | |
496 _marking_task_overhead = 1.0; | |
497 } else { | |
498 if (ParallelMarkingThreads > 0) { | |
751 | 499 // notice that ParallelMarkingThreads overwrites G1MarkingOverheadPercent |
342 | 500 // if both are set |
501 | |
502 _parallel_marking_threads = ParallelMarkingThreads; | |
503 _sleep_factor = 0.0; | |
504 _marking_task_overhead = 1.0; | |
751 | 505 } else if (G1MarkingOverheadPercent > 0) { |
342 | 506 // we will calculate the number of parallel marking threads |
507 // based on a target overhead with respect to the soft real-time | |
508 // goal | |
509 | |
751 | 510 double marking_overhead = (double) G1MarkingOverheadPercent / 100.0; |
342 | 511 double overall_cm_overhead = |
751 | 512 (double) MaxGCPauseMillis * marking_overhead / |
513 (double) GCPauseIntervalMillis; | |
342 | 514 double cpu_ratio = 1.0 / (double) os::processor_count(); |
515 double marking_thread_num = ceil(overall_cm_overhead / cpu_ratio); | |
516 double marking_task_overhead = | |
517 overall_cm_overhead / marking_thread_num * | |
518 (double) os::processor_count(); | |
519 double sleep_factor = | |
520 (1.0 - marking_task_overhead) / marking_task_overhead; | |
521 | |
522 _parallel_marking_threads = (size_t) marking_thread_num; | |
523 _sleep_factor = sleep_factor; | |
524 _marking_task_overhead = marking_task_overhead; | |
525 } else { | |
526 _parallel_marking_threads = MAX2((ParallelGCThreads + 2) / 4, (size_t)1); | |
527 _sleep_factor = 0.0; | |
528 _marking_task_overhead = 1.0; | |
529 } | |
530 | |
531 if (parallel_marking_threads() > 1) | |
532 _cleanup_task_overhead = 1.0; | |
533 else | |
534 _cleanup_task_overhead = marking_task_overhead(); | |
535 _cleanup_sleep_factor = | |
536 (1.0 - cleanup_task_overhead()) / cleanup_task_overhead(); | |
537 | |
538 #if 0 | |
539 gclog_or_tty->print_cr("Marking Threads %d", parallel_marking_threads()); | |
540 gclog_or_tty->print_cr("CM Marking Task Overhead %1.4lf", marking_task_overhead()); | |
541 gclog_or_tty->print_cr("CM Sleep Factor %1.4lf", sleep_factor()); | |
542 gclog_or_tty->print_cr("CL Marking Task Overhead %1.4lf", cleanup_task_overhead()); | |
543 gclog_or_tty->print_cr("CL Sleep Factor %1.4lf", cleanup_sleep_factor()); | |
544 #endif | |
545 | |
546 guarantee( parallel_marking_threads() > 0, "peace of mind" ); | |
547 _parallel_workers = new WorkGang("Parallel Marking Threads", | |
548 (int) parallel_marking_threads(), false, true); | |
549 if (_parallel_workers == NULL) | |
550 vm_exit_during_initialization("Failed necessary allocation."); | |
551 } | |
552 | |
553 // so that the call below can read a sensible value | |
554 _heap_start = (HeapWord*) rs.base(); | |
555 set_non_marking_state(); | |
556 } | |
557 | |
558 void ConcurrentMark::update_g1_committed(bool force) { | |
559 // If concurrent marking is not in progress, then we do not need to | |
560 // update _heap_end. This has a subtle and important | |
561 // side-effect. Imagine that two evacuation pauses happen between | |
562 // marking completion and remark. The first one can grow the | |
563 // heap (hence now the finger is below the heap end). Then, the | |
564 // second one could unnecessarily push regions on the region | |
565 // stack. This causes the invariant that the region stack is empty | |
566 // at the beginning of remark to be false. By ensuring that we do | |
567 // not observe heap expansions after marking is complete, then we do | |
568 // not have this problem. | |
569 if (!concurrent_marking_in_progress() && !force) | |
570 return; | |
571 | |
572 MemRegion committed = _g1h->g1_committed(); | |
573 tmp_guarantee_CM( committed.start() == _heap_start, | |
574 "start shouldn't change" ); | |
575 HeapWord* new_end = committed.end(); | |
576 if (new_end > _heap_end) { | |
577 // The heap has been expanded. | |
578 | |
579 _heap_end = new_end; | |
580 } | |
581 // Notice that the heap can also shrink. However, this only happens | |
582 // during a Full GC (at least currently) and the entire marking | |
583 // phase will bail out and the task will not be restarted. So, let's | |
584 // do nothing. | |
585 } | |
586 | |
587 void ConcurrentMark::reset() { | |
588 // Starting values for these two. This should be called in a STW | |
589 // phase. CM will be notified of any future g1_committed expansions | |
590 // will be at the end of evacuation pauses, when tasks are | |
591 // inactive. | |
592 MemRegion committed = _g1h->g1_committed(); | |
593 _heap_start = committed.start(); | |
594 _heap_end = committed.end(); | |
595 | |
596 guarantee( _heap_start != NULL && | |
597 _heap_end != NULL && | |
598 _heap_start < _heap_end, "heap bounds should look ok" ); | |
599 | |
600 // reset all the marking data structures and any necessary flags | |
601 clear_marking_state(); | |
602 | |
603 if (verbose_low()) | |
604 gclog_or_tty->print_cr("[global] resetting"); | |
605 | |
606 // We do reset all of them, since different phases will use | |
607 // different number of active threads. So, it's easiest to have all | |
608 // of them ready. | |
609 for (int i = 0; i < (int) _max_task_num; ++i) | |
610 _tasks[i]->reset(_nextMarkBitMap); | |
611 | |
612 // we need this to make sure that the flag is on during the evac | |
613 // pause with initial mark piggy-backed | |
614 set_concurrent_marking_in_progress(); | |
615 } | |
616 | |
617 void ConcurrentMark::set_phase(size_t active_tasks, bool concurrent) { | |
618 guarantee( active_tasks <= _max_task_num, "we should not have more" ); | |
619 | |
620 _active_tasks = active_tasks; | |
621 // Need to update the three data structures below according to the | |
622 // number of active threads for this phase. | |
623 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues); | |
624 _first_overflow_barrier_sync.set_n_workers((int) active_tasks); | |
625 _second_overflow_barrier_sync.set_n_workers((int) active_tasks); | |
626 | |
627 _concurrent = concurrent; | |
628 // We propagate this to all tasks, not just the active ones. | |
629 for (int i = 0; i < (int) _max_task_num; ++i) | |
630 _tasks[i]->set_concurrent(concurrent); | |
631 | |
632 if (concurrent) { | |
633 set_concurrent_marking_in_progress(); | |
634 } else { | |
635 // We currently assume that the concurrent flag has been set to | |
636 // false before we start remark. At this point we should also be | |
637 // in a STW phase. | |
638 guarantee( !concurrent_marking_in_progress(), "invariant" ); | |
639 guarantee( _finger == _heap_end, "only way to get here" ); | |
640 update_g1_committed(true); | |
641 } | |
642 } | |
643 | |
644 void ConcurrentMark::set_non_marking_state() { | |
645 // We set the global marking state to some default values when we're | |
646 // not doing marking. | |
647 clear_marking_state(); | |
648 _active_tasks = 0; | |
649 clear_concurrent_marking_in_progress(); | |
650 } | |
651 | |
652 ConcurrentMark::~ConcurrentMark() { | |
653 int size = (int) MAX2(ParallelGCThreads, (size_t)1); | |
654 for (int i = 0; i < size; i++) delete _par_cleanup_thread_state[i]; | |
655 FREE_C_HEAP_ARRAY(ParCleanupThreadState*, | |
656 _par_cleanup_thread_state); | |
657 | |
658 for (int i = 0; i < (int) _max_task_num; ++i) { | |
659 delete _task_queues->queue(i); | |
660 delete _tasks[i]; | |
661 } | |
662 delete _task_queues; | |
663 FREE_C_HEAP_ARRAY(CMTask*, _max_task_num); | |
664 } | |
665 | |
666 // This closure is used to mark refs into the g1 generation | |
667 // from external roots in the CMS bit map. | |
668 // Called at the first checkpoint. | |
669 // | |
670 | |
671 #define PRINT_REACHABLE_AT_INITIAL_MARK 0 | |
672 #if PRINT_REACHABLE_AT_INITIAL_MARK | |
673 static FILE* reachable_file = NULL; | |
674 | |
675 class PrintReachableClosure: public OopsInGenClosure { | |
676 CMBitMap* _bm; | |
677 int _level; | |
678 public: | |
679 PrintReachableClosure(CMBitMap* bm) : | |
680 _bm(bm), _level(0) { | |
681 guarantee(reachable_file != NULL, "pre-condition"); | |
682 } | |
683 void do_oop(oop* p) { | |
684 oop obj = *p; | |
685 HeapWord* obj_addr = (HeapWord*)obj; | |
686 if (obj == NULL) return; | |
687 fprintf(reachable_file, "%d: "PTR_FORMAT" -> "PTR_FORMAT" (%d)\n", | |
688 _level, p, (void*) obj, _bm->isMarked(obj_addr)); | |
689 if (!_bm->isMarked(obj_addr)) { | |
690 _bm->mark(obj_addr); | |
691 _level++; | |
692 obj->oop_iterate(this); | |
693 _level--; | |
694 } | |
695 } | |
696 }; | |
697 #endif // PRINT_REACHABLE_AT_INITIAL_MARK | |
698 | |
699 #define SEND_HEAP_DUMP_TO_FILE 0 | |
700 #if SEND_HEAP_DUMP_TO_FILE | |
701 static FILE* heap_dump_file = NULL; | |
702 #endif // SEND_HEAP_DUMP_TO_FILE | |
703 | |
704 void ConcurrentMark::clearNextBitmap() { | |
705 guarantee(!G1CollectedHeap::heap()->mark_in_progress(), "Precondition."); | |
706 | |
707 // clear the mark bitmap (no grey objects to start with). | |
708 // We need to do this in chunks and offer to yield in between | |
709 // each chunk. | |
710 HeapWord* start = _nextMarkBitMap->startWord(); | |
711 HeapWord* end = _nextMarkBitMap->endWord(); | |
712 HeapWord* cur = start; | |
713 size_t chunkSize = M; | |
714 while (cur < end) { | |
715 HeapWord* next = cur + chunkSize; | |
716 if (next > end) | |
717 next = end; | |
718 MemRegion mr(cur,next); | |
719 _nextMarkBitMap->clearRange(mr); | |
720 cur = next; | |
721 do_yield_check(); | |
722 } | |
723 } | |
724 | |
725 class NoteStartOfMarkHRClosure: public HeapRegionClosure { | |
726 public: | |
727 bool doHeapRegion(HeapRegion* r) { | |
728 if (!r->continuesHumongous()) { | |
729 r->note_start_of_marking(true); | |
730 } | |
731 return false; | |
732 } | |
733 }; | |
734 | |
735 void ConcurrentMark::checkpointRootsInitialPre() { | |
736 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
737 G1CollectorPolicy* g1p = g1h->g1_policy(); | |
738 | |
739 _has_aborted = false; | |
740 | |
741 // Find all the reachable objects... | |
742 #if PRINT_REACHABLE_AT_INITIAL_MARK | |
743 guarantee(reachable_file == NULL, "Protocol"); | |
744 char fn_buf[100]; | |
745 sprintf(fn_buf, "/tmp/reachable.txt.%d", os::current_process_id()); | |
746 reachable_file = fopen(fn_buf, "w"); | |
747 // clear the mark bitmap (no grey objects to start with) | |
748 _nextMarkBitMap->clearAll(); | |
749 PrintReachableClosure prcl(_nextMarkBitMap); | |
750 g1h->process_strong_roots( | |
751 false, // fake perm gen collection | |
752 SharedHeap::SO_AllClasses, | |
753 &prcl, // Regular roots | |
754 &prcl // Perm Gen Roots | |
755 ); | |
756 // The root iteration above "consumed" dirty cards in the perm gen. | |
757 // Therefore, as a shortcut, we dirty all such cards. | |
758 g1h->rem_set()->invalidate(g1h->perm_gen()->used_region(), false); | |
759 fclose(reachable_file); | |
760 reachable_file = NULL; | |
761 // clear the mark bitmap again. | |
762 _nextMarkBitMap->clearAll(); | |
763 COMPILER2_PRESENT(DerivedPointerTable::update_pointers()); | |
764 COMPILER2_PRESENT(DerivedPointerTable::clear()); | |
765 #endif // PRINT_REACHABLE_AT_INITIAL_MARK | |
766 | |
767 // Initialise marking structures. This has to be done in a STW phase. | |
768 reset(); | |
769 } | |
770 | |
771 class CMMarkRootsClosure: public OopsInGenClosure { | |
772 private: | |
773 ConcurrentMark* _cm; | |
774 G1CollectedHeap* _g1h; | |
775 bool _do_barrier; | |
776 | |
777 public: | |
778 CMMarkRootsClosure(ConcurrentMark* cm, | |
779 G1CollectedHeap* g1h, | |
780 bool do_barrier) : _cm(cm), _g1h(g1h), | |
781 _do_barrier(do_barrier) { } | |
782 | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
783 virtual void do_oop(narrowOop* p) { do_oop_work(p); } |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
784 virtual void do_oop( oop* p) { do_oop_work(p); } |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
785 |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
786 template <class T> void do_oop_work(T* p) { |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
787 T heap_oop = oopDesc::load_heap_oop(p); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
788 if (!oopDesc::is_null(heap_oop)) { |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
789 oop obj = oopDesc::decode_heap_oop_not_null(heap_oop); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
790 assert(obj->is_oop() || obj->mark() == NULL, |
342 | 791 "expected an oop, possibly with mark word displaced"); |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
792 HeapWord* addr = (HeapWord*)obj; |
342 | 793 if (_g1h->is_in_g1_reserved(addr)) { |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
794 _cm->grayRoot(obj); |
342 | 795 } |
796 } | |
797 if (_do_barrier) { | |
798 assert(!_g1h->is_in_g1_reserved(p), | |
799 "Should be called on external roots"); | |
800 do_barrier(p); | |
801 } | |
802 } | |
803 }; | |
804 | |
805 void ConcurrentMark::checkpointRootsInitialPost() { | |
806 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
807 | |
808 // For each region note start of marking. | |
809 NoteStartOfMarkHRClosure startcl; | |
810 g1h->heap_region_iterate(&startcl); | |
811 | |
812 // Start weak-reference discovery. | |
813 ReferenceProcessor* rp = g1h->ref_processor(); | |
814 rp->verify_no_references_recorded(); | |
815 rp->enable_discovery(); // enable ("weak") refs discovery | |
457
27a80744a83b
6778647: snap(), snap_policy() should be renamed setup(), setup_policy()
ysr
parents:
453
diff
changeset
|
816 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle |
342 | 817 |
818 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
819 satb_mq_set.set_process_completed_threshold(G1SATBProcessCompletedThreshold); | |
820 satb_mq_set.set_active_all_threads(true); | |
821 | |
822 // update_g1_committed() will be called at the end of an evac pause | |
823 // when marking is on. So, it's also called at the end of the | |
824 // initial-mark pause to update the heap end, if the heap expands | |
825 // during it. No need to call it here. | |
826 | |
827 guarantee( !_cleanup_co_tracker.enabled(), "invariant" ); | |
828 | |
829 size_t max_marking_threads = | |
830 MAX2((size_t) 1, parallel_marking_threads()); | |
831 for (int i = 0; i < (int)_max_task_num; ++i) { | |
832 _tasks[i]->enable_co_tracker(); | |
833 if (i < (int) max_marking_threads) | |
834 _tasks[i]->reset_co_tracker(marking_task_overhead()); | |
835 else | |
836 _tasks[i]->reset_co_tracker(0.0); | |
837 } | |
838 } | |
839 | |
840 // Checkpoint the roots into this generation from outside | |
841 // this generation. [Note this initial checkpoint need only | |
842 // be approximate -- we'll do a catch up phase subsequently.] | |
843 void ConcurrentMark::checkpointRootsInitial() { | |
844 assert(SafepointSynchronize::is_at_safepoint(), "world should be stopped"); | |
845 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
846 | |
847 double start = os::elapsedTime(); | |
848 GCOverheadReporter::recordSTWStart(start); | |
849 | |
850 G1CollectorPolicy* g1p = G1CollectedHeap::heap()->g1_policy(); | |
851 g1p->record_concurrent_mark_init_start(); | |
852 checkpointRootsInitialPre(); | |
853 | |
854 // YSR: when concurrent precleaning is in place, we'll | |
855 // need to clear the cached card table here | |
856 | |
857 ResourceMark rm; | |
858 HandleMark hm; | |
859 | |
860 g1h->ensure_parsability(false); | |
861 g1h->perm_gen()->save_marks(); | |
862 | |
863 CMMarkRootsClosure notOlder(this, g1h, false); | |
864 CMMarkRootsClosure older(this, g1h, true); | |
865 | |
866 g1h->set_marking_started(); | |
867 g1h->rem_set()->prepare_for_younger_refs_iterate(false); | |
868 | |
869 g1h->process_strong_roots(false, // fake perm gen collection | |
870 SharedHeap::SO_AllClasses, | |
871 ¬Older, // Regular roots | |
872 &older // Perm Gen Roots | |
873 ); | |
874 checkpointRootsInitialPost(); | |
875 | |
876 // Statistics. | |
877 double end = os::elapsedTime(); | |
878 _init_times.add((end - start) * 1000.0); | |
879 GCOverheadReporter::recordSTWEnd(end); | |
880 | |
881 g1p->record_concurrent_mark_init_end(); | |
882 } | |
883 | |
884 /* | |
885 Notice that in the next two methods, we actually leave the STS | |
886 during the barrier sync and join it immediately afterwards. If we | |
887 do not do this, this then the following deadlock can occur: one | |
888 thread could be in the barrier sync code, waiting for the other | |
889 thread to also sync up, whereas another one could be trying to | |
890 yield, while also waiting for the other threads to sync up too. | |
891 | |
892 Because the thread that does the sync barrier has left the STS, it | |
893 is possible to be suspended for a Full GC or an evacuation pause | |
894 could occur. This is actually safe, since the entering the sync | |
895 barrier is one of the last things do_marking_step() does, and it | |
896 doesn't manipulate any data structures afterwards. | |
897 */ | |
898 | |
899 void ConcurrentMark::enter_first_sync_barrier(int task_num) { | |
900 if (verbose_low()) | |
901 gclog_or_tty->print_cr("[%d] entering first barrier", task_num); | |
902 | |
903 ConcurrentGCThread::stsLeave(); | |
904 _first_overflow_barrier_sync.enter(); | |
905 ConcurrentGCThread::stsJoin(); | |
906 // at this point everyone should have synced up and not be doing any | |
907 // more work | |
908 | |
909 if (verbose_low()) | |
910 gclog_or_tty->print_cr("[%d] leaving first barrier", task_num); | |
911 | |
912 // let task 0 do this | |
913 if (task_num == 0) { | |
914 // task 0 is responsible for clearing the global data structures | |
915 clear_marking_state(); | |
916 | |
917 if (PrintGC) { | |
918 gclog_or_tty->date_stamp(PrintGCDateStamps); | |
919 gclog_or_tty->stamp(PrintGCTimeStamps); | |
920 gclog_or_tty->print_cr("[GC concurrent-mark-reset-for-overflow]"); | |
921 } | |
922 } | |
923 | |
924 // after this, each task should reset its own data structures then | |
925 // then go into the second barrier | |
926 } | |
927 | |
928 void ConcurrentMark::enter_second_sync_barrier(int task_num) { | |
929 if (verbose_low()) | |
930 gclog_or_tty->print_cr("[%d] entering second barrier", task_num); | |
931 | |
932 ConcurrentGCThread::stsLeave(); | |
933 _second_overflow_barrier_sync.enter(); | |
934 ConcurrentGCThread::stsJoin(); | |
935 // at this point everything should be re-initialised and ready to go | |
936 | |
937 if (verbose_low()) | |
938 gclog_or_tty->print_cr("[%d] leaving second barrier", task_num); | |
939 } | |
940 | |
941 void ConcurrentMark::grayRoot(oop p) { | |
942 HeapWord* addr = (HeapWord*) p; | |
943 // We can't really check against _heap_start and _heap_end, since it | |
944 // is possible during an evacuation pause with piggy-backed | |
945 // initial-mark that the committed space is expanded during the | |
946 // pause without CM observing this change. So the assertions below | |
947 // is a bit conservative; but better than nothing. | |
948 tmp_guarantee_CM( _g1h->g1_committed().contains(addr), | |
949 "address should be within the heap bounds" ); | |
950 | |
951 if (!_nextMarkBitMap->isMarked(addr)) | |
952 _nextMarkBitMap->parMark(addr); | |
953 } | |
954 | |
955 void ConcurrentMark::grayRegionIfNecessary(MemRegion mr) { | |
956 // The objects on the region have already been marked "in bulk" by | |
957 // the caller. We only need to decide whether to push the region on | |
958 // the region stack or not. | |
959 | |
960 if (!concurrent_marking_in_progress() || !_should_gray_objects) | |
961 // We're done with marking and waiting for remark. We do not need to | |
962 // push anything else on the region stack. | |
963 return; | |
964 | |
965 HeapWord* finger = _finger; | |
966 | |
967 if (verbose_low()) | |
968 gclog_or_tty->print_cr("[global] attempting to push " | |
969 "region ["PTR_FORMAT", "PTR_FORMAT"), finger is at " | |
970 PTR_FORMAT, mr.start(), mr.end(), finger); | |
971 | |
972 if (mr.start() < finger) { | |
973 // The finger is always heap region aligned and it is not possible | |
974 // for mr to span heap regions. | |
975 tmp_guarantee_CM( mr.end() <= finger, "invariant" ); | |
976 | |
977 tmp_guarantee_CM( mr.start() <= mr.end() && | |
978 _heap_start <= mr.start() && | |
979 mr.end() <= _heap_end, | |
980 "region boundaries should fall within the committed space" ); | |
981 if (verbose_low()) | |
982 gclog_or_tty->print_cr("[global] region ["PTR_FORMAT", "PTR_FORMAT") " | |
983 "below the finger, pushing it", | |
984 mr.start(), mr.end()); | |
985 | |
986 if (!region_stack_push(mr)) { | |
987 if (verbose_low()) | |
988 gclog_or_tty->print_cr("[global] region stack has overflown."); | |
989 } | |
990 } | |
991 } | |
992 | |
993 void ConcurrentMark::markAndGrayObjectIfNecessary(oop p) { | |
994 // The object is not marked by the caller. We need to at least mark | |
995 // it and maybe push in on the stack. | |
996 | |
997 HeapWord* addr = (HeapWord*)p; | |
998 if (!_nextMarkBitMap->isMarked(addr)) { | |
999 // We definitely need to mark it, irrespective whether we bail out | |
1000 // because we're done with marking. | |
1001 if (_nextMarkBitMap->parMark(addr)) { | |
1002 if (!concurrent_marking_in_progress() || !_should_gray_objects) | |
1003 // If we're done with concurrent marking and we're waiting for | |
1004 // remark, then we're not pushing anything on the stack. | |
1005 return; | |
1006 | |
1007 // No OrderAccess:store_load() is needed. It is implicit in the | |
1008 // CAS done in parMark(addr) above | |
1009 HeapWord* finger = _finger; | |
1010 | |
1011 if (addr < finger) { | |
1012 if (!mark_stack_push(oop(addr))) { | |
1013 if (verbose_low()) | |
1014 gclog_or_tty->print_cr("[global] global stack overflow " | |
1015 "during parMark"); | |
1016 } | |
1017 } | |
1018 } | |
1019 } | |
1020 } | |
1021 | |
1022 class CMConcurrentMarkingTask: public AbstractGangTask { | |
1023 private: | |
1024 ConcurrentMark* _cm; | |
1025 ConcurrentMarkThread* _cmt; | |
1026 | |
1027 public: | |
1028 void work(int worker_i) { | |
1029 guarantee( Thread::current()->is_ConcurrentGC_thread(), | |
1030 "this should only be done by a conc GC thread" ); | |
1031 | |
1032 double start_vtime = os::elapsedVTime(); | |
1033 | |
1034 ConcurrentGCThread::stsJoin(); | |
1035 | |
1036 guarantee( (size_t)worker_i < _cm->active_tasks(), "invariant" ); | |
1037 CMTask* the_task = _cm->task(worker_i); | |
1038 the_task->start_co_tracker(); | |
1039 the_task->record_start_time(); | |
1040 if (!_cm->has_aborted()) { | |
1041 do { | |
1042 double start_vtime_sec = os::elapsedVTime(); | |
1043 double start_time_sec = os::elapsedTime(); | |
1044 the_task->do_marking_step(10.0); | |
1045 double end_time_sec = os::elapsedTime(); | |
1046 double end_vtime_sec = os::elapsedVTime(); | |
1047 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec; | |
1048 double elapsed_time_sec = end_time_sec - start_time_sec; | |
1049 _cm->clear_has_overflown(); | |
1050 | |
1051 bool ret = _cm->do_yield_check(worker_i); | |
1052 | |
1053 jlong sleep_time_ms; | |
1054 if (!_cm->has_aborted() && the_task->has_aborted()) { | |
1055 sleep_time_ms = | |
1056 (jlong) (elapsed_vtime_sec * _cm->sleep_factor() * 1000.0); | |
1057 ConcurrentGCThread::stsLeave(); | |
1058 os::sleep(Thread::current(), sleep_time_ms, false); | |
1059 ConcurrentGCThread::stsJoin(); | |
1060 } | |
1061 double end_time2_sec = os::elapsedTime(); | |
1062 double elapsed_time2_sec = end_time2_sec - start_time_sec; | |
1063 | |
1064 the_task->update_co_tracker(); | |
1065 | |
1066 #if 0 | |
1067 gclog_or_tty->print_cr("CM: elapsed %1.4lf ms, sleep %1.4lf ms, " | |
1068 "overhead %1.4lf", | |
1069 elapsed_vtime_sec * 1000.0, (double) sleep_time_ms, | |
1070 the_task->conc_overhead(os::elapsedTime()) * 8.0); | |
1071 gclog_or_tty->print_cr("elapsed time %1.4lf ms, time 2: %1.4lf ms", | |
1072 elapsed_time_sec * 1000.0, elapsed_time2_sec * 1000.0); | |
1073 #endif | |
1074 } while (!_cm->has_aborted() && the_task->has_aborted()); | |
1075 } | |
1076 the_task->record_end_time(); | |
1077 guarantee( !the_task->has_aborted() || _cm->has_aborted(), "invariant" ); | |
1078 | |
1079 ConcurrentGCThread::stsLeave(); | |
1080 | |
1081 double end_vtime = os::elapsedVTime(); | |
1082 the_task->update_co_tracker(true); | |
1083 _cm->update_accum_task_vtime(worker_i, end_vtime - start_vtime); | |
1084 } | |
1085 | |
1086 CMConcurrentMarkingTask(ConcurrentMark* cm, | |
1087 ConcurrentMarkThread* cmt) : | |
1088 AbstractGangTask("Concurrent Mark"), _cm(cm), _cmt(cmt) { } | |
1089 | |
1090 ~CMConcurrentMarkingTask() { } | |
1091 }; | |
1092 | |
1093 void ConcurrentMark::markFromRoots() { | |
1094 // we might be tempted to assert that: | |
1095 // assert(asynch == !SafepointSynchronize::is_at_safepoint(), | |
1096 // "inconsistent argument?"); | |
1097 // However that wouldn't be right, because it's possible that | |
1098 // a safepoint is indeed in progress as a younger generation | |
1099 // stop-the-world GC happens even as we mark in this generation. | |
1100 | |
1101 _restart_for_overflow = false; | |
1102 | |
1103 set_phase(MAX2((size_t) 1, parallel_marking_threads()), true); | |
1104 | |
1105 CMConcurrentMarkingTask markingTask(this, cmThread()); | |
1106 if (parallel_marking_threads() > 0) | |
1107 _parallel_workers->run_task(&markingTask); | |
1108 else | |
1109 markingTask.work(0); | |
1110 print_stats(); | |
1111 } | |
1112 | |
1113 void ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) { | |
1114 // world is stopped at this checkpoint | |
1115 assert(SafepointSynchronize::is_at_safepoint(), | |
1116 "world should be stopped"); | |
1117 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
1118 | |
1119 // If a full collection has happened, we shouldn't do this. | |
1120 if (has_aborted()) { | |
1121 g1h->set_marking_complete(); // So bitmap clearing isn't confused | |
1122 return; | |
1123 } | |
1124 | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1125 if (VerifyDuringGC) { |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1126 HandleMark hm; // handle scope |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1127 gclog_or_tty->print(" VerifyDuringGC:(before)"); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1128 Universe::heap()->prepare_for_verify(); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1129 Universe::verify(true, false, true); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1130 } |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1131 |
342 | 1132 G1CollectorPolicy* g1p = g1h->g1_policy(); |
1133 g1p->record_concurrent_mark_remark_start(); | |
1134 | |
1135 double start = os::elapsedTime(); | |
1136 GCOverheadReporter::recordSTWStart(start); | |
1137 | |
1138 checkpointRootsFinalWork(); | |
1139 | |
1140 double mark_work_end = os::elapsedTime(); | |
1141 | |
1142 weakRefsWork(clear_all_soft_refs); | |
1143 | |
1144 if (has_overflown()) { | |
1145 // Oops. We overflowed. Restart concurrent marking. | |
1146 _restart_for_overflow = true; | |
1147 // Clear the flag. We do not need it any more. | |
1148 clear_has_overflown(); | |
1149 if (G1TraceMarkStackOverflow) | |
1150 gclog_or_tty->print_cr("\nRemark led to restart for overflow."); | |
1151 } else { | |
1152 // We're done with marking. | |
1153 JavaThread::satb_mark_queue_set().set_active_all_threads(false); | |
811 | 1154 |
1155 if (VerifyDuringGC) { | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1156 HandleMark hm; // handle scope |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1157 gclog_or_tty->print(" VerifyDuringGC:(after)"); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1158 Universe::heap()->prepare_for_verify(); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1159 Universe::heap()->verify(/* allow_dirty */ true, |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1160 /* silent */ false, |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1161 /* use_prev_marking */ false); |
811 | 1162 } |
342 | 1163 } |
1164 | |
1165 #if VERIFY_OBJS_PROCESSED | |
1166 _scan_obj_cl.objs_processed = 0; | |
1167 ThreadLocalObjQueue::objs_enqueued = 0; | |
1168 #endif | |
1169 | |
1170 // Statistics | |
1171 double now = os::elapsedTime(); | |
1172 _remark_mark_times.add((mark_work_end - start) * 1000.0); | |
1173 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0); | |
1174 _remark_times.add((now - start) * 1000.0); | |
1175 | |
1176 GCOverheadReporter::recordSTWEnd(now); | |
1177 for (int i = 0; i < (int)_max_task_num; ++i) | |
1178 _tasks[i]->disable_co_tracker(); | |
1179 _cleanup_co_tracker.enable(); | |
1180 _cleanup_co_tracker.reset(cleanup_task_overhead()); | |
1181 g1p->record_concurrent_mark_remark_end(); | |
1182 } | |
1183 | |
1184 | |
1185 #define CARD_BM_TEST_MODE 0 | |
1186 | |
1187 class CalcLiveObjectsClosure: public HeapRegionClosure { | |
1188 | |
1189 CMBitMapRO* _bm; | |
1190 ConcurrentMark* _cm; | |
1191 COTracker* _co_tracker; | |
1192 bool _changed; | |
1193 bool _yield; | |
1194 size_t _words_done; | |
1195 size_t _tot_live; | |
1196 size_t _tot_used; | |
1197 size_t _regions_done; | |
1198 double _start_vtime_sec; | |
1199 | |
1200 BitMap* _region_bm; | |
1201 BitMap* _card_bm; | |
1202 intptr_t _bottom_card_num; | |
1203 bool _final; | |
1204 | |
1205 void mark_card_num_range(intptr_t start_card_num, intptr_t last_card_num) { | |
1206 for (intptr_t i = start_card_num; i <= last_card_num; i++) { | |
1207 #if CARD_BM_TEST_MODE | |
1208 guarantee(_card_bm->at(i - _bottom_card_num), | |
1209 "Should already be set."); | |
1210 #else | |
1211 _card_bm->par_at_put(i - _bottom_card_num, 1); | |
1212 #endif | |
1213 } | |
1214 } | |
1215 | |
1216 public: | |
1217 CalcLiveObjectsClosure(bool final, | |
1218 CMBitMapRO *bm, ConcurrentMark *cm, | |
1219 BitMap* region_bm, BitMap* card_bm, | |
1220 COTracker* co_tracker) : | |
1221 _bm(bm), _cm(cm), _changed(false), _yield(true), | |
1222 _words_done(0), _tot_live(0), _tot_used(0), | |
1223 _region_bm(region_bm), _card_bm(card_bm), | |
1224 _final(final), _co_tracker(co_tracker), | |
1225 _regions_done(0), _start_vtime_sec(0.0) | |
1226 { | |
1227 _bottom_card_num = | |
1228 intptr_t(uintptr_t(G1CollectedHeap::heap()->reserved_region().start()) >> | |
1229 CardTableModRefBS::card_shift); | |
1230 } | |
1231 | |
829 | 1232 // It takes a region that's not empty (i.e., it has at least one |
1233 // live object in it and sets its corresponding bit on the region | |
1234 // bitmap to 1. If the region is "starts humongous" it will also set | |
1235 // to 1 the bits on the region bitmap that correspond to its | |
1236 // associated "continues humongous" regions. | |
1237 void set_bit_for_region(HeapRegion* hr) { | |
1238 assert(!hr->continuesHumongous(), "should have filtered those out"); | |
1239 | |
1240 size_t index = hr->hrs_index(); | |
1241 if (!hr->startsHumongous()) { | |
1242 // Normal (non-humongous) case: just set the bit. | |
1243 _region_bm->par_at_put((BitMap::idx_t) index, true); | |
1244 } else { | |
1245 // Starts humongous case: calculate how many regions are part of | |
1246 // this humongous region and then set the bit range. It might | |
1247 // have been a bit more efficient to look at the object that | |
1248 // spans these humongous regions to calculate their number from | |
1249 // the object's size. However, it's a good idea to calculate | |
1250 // this based on the metadata itself, and not the region | |
1251 // contents, so that this code is not aware of what goes into | |
1252 // the humongous regions (in case this changes in the future). | |
1253 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
1254 size_t end_index = index + 1; | |
831 | 1255 while (end_index < g1h->n_regions()) { |
1256 HeapRegion* chr = g1h->region_at(end_index); | |
829 | 1257 if (!chr->continuesHumongous()) { |
1258 break; | |
1259 } | |
1260 end_index += 1; | |
1261 } | |
1262 _region_bm->par_at_put_range((BitMap::idx_t) index, | |
1263 (BitMap::idx_t) end_index, true); | |
1264 } | |
1265 } | |
1266 | |
342 | 1267 bool doHeapRegion(HeapRegion* hr) { |
1268 if (_co_tracker != NULL) | |
1269 _co_tracker->update(); | |
1270 | |
1271 if (!_final && _regions_done == 0) | |
1272 _start_vtime_sec = os::elapsedVTime(); | |
1273 | |
639 | 1274 if (hr->continuesHumongous()) { |
829 | 1275 // We will ignore these here and process them when their |
1276 // associated "starts humongous" region is processed (see | |
1277 // set_bit_for_heap_region()). Note that we cannot rely on their | |
1278 // associated "starts humongous" region to have their bit set to | |
1279 // 1 since, due to the region chunking in the parallel region | |
1280 // iteration, a "continues humongous" region might be visited | |
1281 // before its associated "starts humongous". | |
639 | 1282 return false; |
1283 } | |
342 | 1284 |
1285 HeapWord* nextTop = hr->next_top_at_mark_start(); | |
1286 HeapWord* start = hr->top_at_conc_mark_count(); | |
1287 assert(hr->bottom() <= start && start <= hr->end() && | |
1288 hr->bottom() <= nextTop && nextTop <= hr->end() && | |
1289 start <= nextTop, | |
1290 "Preconditions."); | |
1291 // Otherwise, record the number of word's we'll examine. | |
1292 size_t words_done = (nextTop - start); | |
1293 // Find the first marked object at or after "start". | |
1294 start = _bm->getNextMarkedWordAddress(start, nextTop); | |
1295 size_t marked_bytes = 0; | |
1296 | |
1297 // Below, the term "card num" means the result of shifting an address | |
1298 // by the card shift -- address 0 corresponds to card number 0. One | |
1299 // must subtract the card num of the bottom of the heap to obtain a | |
1300 // card table index. | |
1301 // The first card num of the sequence of live cards currently being | |
1302 // constructed. -1 ==> no sequence. | |
1303 intptr_t start_card_num = -1; | |
1304 // The last card num of the sequence of live cards currently being | |
1305 // constructed. -1 ==> no sequence. | |
1306 intptr_t last_card_num = -1; | |
1307 | |
1308 while (start < nextTop) { | |
1309 if (_yield && _cm->do_yield_check()) { | |
1310 // We yielded. It might be for a full collection, in which case | |
1311 // all bets are off; terminate the traversal. | |
1312 if (_cm->has_aborted()) { | |
1313 _changed = false; | |
1314 return true; | |
1315 } else { | |
1316 // Otherwise, it might be a collection pause, and the region | |
1317 // we're looking at might be in the collection set. We'll | |
1318 // abandon this region. | |
1319 return false; | |
1320 } | |
1321 } | |
1322 oop obj = oop(start); | |
1323 int obj_sz = obj->size(); | |
1324 // The card num of the start of the current object. | |
1325 intptr_t obj_card_num = | |
1326 intptr_t(uintptr_t(start) >> CardTableModRefBS::card_shift); | |
1327 | |
1328 HeapWord* obj_last = start + obj_sz - 1; | |
1329 intptr_t obj_last_card_num = | |
1330 intptr_t(uintptr_t(obj_last) >> CardTableModRefBS::card_shift); | |
1331 | |
1332 if (obj_card_num != last_card_num) { | |
1333 if (start_card_num == -1) { | |
1334 assert(last_card_num == -1, "Both or neither."); | |
1335 start_card_num = obj_card_num; | |
1336 } else { | |
1337 assert(last_card_num != -1, "Both or neither."); | |
1338 assert(obj_card_num >= last_card_num, "Inv"); | |
1339 if ((obj_card_num - last_card_num) > 1) { | |
1340 // Mark the last run, and start a new one. | |
1341 mark_card_num_range(start_card_num, last_card_num); | |
1342 start_card_num = obj_card_num; | |
1343 } | |
1344 } | |
1345 #if CARD_BM_TEST_MODE | |
1346 /* | |
1347 gclog_or_tty->print_cr("Setting bits from %d/%d.", | |
1348 obj_card_num - _bottom_card_num, | |
1349 obj_last_card_num - _bottom_card_num); | |
1350 */ | |
1351 for (intptr_t j = obj_card_num; j <= obj_last_card_num; j++) { | |
1352 _card_bm->par_at_put(j - _bottom_card_num, 1); | |
1353 } | |
1354 #endif | |
1355 } | |
1356 // In any case, we set the last card num. | |
1357 last_card_num = obj_last_card_num; | |
1358 | |
1359 marked_bytes += obj_sz * HeapWordSize; | |
1360 // Find the next marked object after this one. | |
1361 start = _bm->getNextMarkedWordAddress(start + 1, nextTop); | |
1362 _changed = true; | |
1363 } | |
1364 // Handle the last range, if any. | |
1365 if (start_card_num != -1) | |
1366 mark_card_num_range(start_card_num, last_card_num); | |
1367 if (_final) { | |
1368 // Mark the allocated-since-marking portion... | |
1369 HeapWord* tp = hr->top(); | |
1370 if (nextTop < tp) { | |
1371 start_card_num = | |
1372 intptr_t(uintptr_t(nextTop) >> CardTableModRefBS::card_shift); | |
1373 last_card_num = | |
1374 intptr_t(uintptr_t(tp) >> CardTableModRefBS::card_shift); | |
1375 mark_card_num_range(start_card_num, last_card_num); | |
1376 // This definitely means the region has live objects. | |
829 | 1377 set_bit_for_region(hr); |
342 | 1378 } |
1379 } | |
1380 | |
1381 hr->add_to_marked_bytes(marked_bytes); | |
1382 // Update the live region bitmap. | |
1383 if (marked_bytes > 0) { | |
829 | 1384 set_bit_for_region(hr); |
342 | 1385 } |
1386 hr->set_top_at_conc_mark_count(nextTop); | |
1387 _tot_live += hr->next_live_bytes(); | |
1388 _tot_used += hr->used(); | |
1389 _words_done = words_done; | |
1390 | |
1391 if (!_final) { | |
1392 ++_regions_done; | |
1393 if (_regions_done % 10 == 0) { | |
1394 double end_vtime_sec = os::elapsedVTime(); | |
1395 double elapsed_vtime_sec = end_vtime_sec - _start_vtime_sec; | |
1396 if (elapsed_vtime_sec > (10.0 / 1000.0)) { | |
1397 jlong sleep_time_ms = | |
1398 (jlong) (elapsed_vtime_sec * _cm->cleanup_sleep_factor() * 1000.0); | |
1399 #if 0 | |
1400 gclog_or_tty->print_cr("CL: elapsed %1.4lf ms, sleep %1.4lf ms, " | |
1401 "overhead %1.4lf", | |
1402 elapsed_vtime_sec * 1000.0, (double) sleep_time_ms, | |
1403 _co_tracker->concOverhead(os::elapsedTime())); | |
1404 #endif | |
1405 os::sleep(Thread::current(), sleep_time_ms, false); | |
1406 _start_vtime_sec = end_vtime_sec; | |
1407 } | |
1408 } | |
1409 } | |
1410 | |
1411 return false; | |
1412 } | |
1413 | |
1414 bool changed() { return _changed; } | |
1415 void reset() { _changed = false; _words_done = 0; } | |
1416 void no_yield() { _yield = false; } | |
1417 size_t words_done() { return _words_done; } | |
1418 size_t tot_live() { return _tot_live; } | |
1419 size_t tot_used() { return _tot_used; } | |
1420 }; | |
1421 | |
1422 | |
1423 void ConcurrentMark::calcDesiredRegions() { | |
1424 guarantee( _cleanup_co_tracker.enabled(), "invariant" ); | |
1425 _cleanup_co_tracker.start(); | |
1426 | |
1427 _region_bm.clear(); | |
1428 _card_bm.clear(); | |
1429 CalcLiveObjectsClosure calccl(false /*final*/, | |
1430 nextMarkBitMap(), this, | |
1431 &_region_bm, &_card_bm, | |
1432 &_cleanup_co_tracker); | |
1433 G1CollectedHeap *g1h = G1CollectedHeap::heap(); | |
1434 g1h->heap_region_iterate(&calccl); | |
1435 | |
1436 do { | |
1437 calccl.reset(); | |
1438 g1h->heap_region_iterate(&calccl); | |
1439 } while (calccl.changed()); | |
1440 | |
1441 _cleanup_co_tracker.update(true); | |
1442 } | |
1443 | |
1444 class G1ParFinalCountTask: public AbstractGangTask { | |
1445 protected: | |
1446 G1CollectedHeap* _g1h; | |
1447 CMBitMap* _bm; | |
1448 size_t _n_workers; | |
1449 size_t *_live_bytes; | |
1450 size_t *_used_bytes; | |
1451 BitMap* _region_bm; | |
1452 BitMap* _card_bm; | |
1453 public: | |
1454 G1ParFinalCountTask(G1CollectedHeap* g1h, CMBitMap* bm, | |
1455 BitMap* region_bm, BitMap* card_bm) : | |
1456 AbstractGangTask("G1 final counting"), _g1h(g1h), | |
1457 _bm(bm), _region_bm(region_bm), _card_bm(card_bm) | |
1458 { | |
1459 if (ParallelGCThreads > 0) | |
1460 _n_workers = _g1h->workers()->total_workers(); | |
1461 else | |
1462 _n_workers = 1; | |
1463 _live_bytes = NEW_C_HEAP_ARRAY(size_t, _n_workers); | |
1464 _used_bytes = NEW_C_HEAP_ARRAY(size_t, _n_workers); | |
1465 } | |
1466 | |
1467 ~G1ParFinalCountTask() { | |
1468 FREE_C_HEAP_ARRAY(size_t, _live_bytes); | |
1469 FREE_C_HEAP_ARRAY(size_t, _used_bytes); | |
1470 } | |
1471 | |
1472 void work(int i) { | |
1473 CalcLiveObjectsClosure calccl(true /*final*/, | |
1474 _bm, _g1h->concurrent_mark(), | |
1475 _region_bm, _card_bm, | |
1476 NULL /* CO tracker */); | |
1477 calccl.no_yield(); | |
1478 if (ParallelGCThreads > 0) { | |
355 | 1479 _g1h->heap_region_par_iterate_chunked(&calccl, i, |
1480 HeapRegion::FinalCountClaimValue); | |
342 | 1481 } else { |
1482 _g1h->heap_region_iterate(&calccl); | |
1483 } | |
1484 assert(calccl.complete(), "Shouldn't have yielded!"); | |
1485 | |
1486 guarantee( (size_t)i < _n_workers, "invariant" ); | |
1487 _live_bytes[i] = calccl.tot_live(); | |
1488 _used_bytes[i] = calccl.tot_used(); | |
1489 } | |
1490 size_t live_bytes() { | |
1491 size_t live_bytes = 0; | |
1492 for (size_t i = 0; i < _n_workers; ++i) | |
1493 live_bytes += _live_bytes[i]; | |
1494 return live_bytes; | |
1495 } | |
1496 size_t used_bytes() { | |
1497 size_t used_bytes = 0; | |
1498 for (size_t i = 0; i < _n_workers; ++i) | |
1499 used_bytes += _used_bytes[i]; | |
1500 return used_bytes; | |
1501 } | |
1502 }; | |
1503 | |
1504 class G1ParNoteEndTask; | |
1505 | |
1506 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure { | |
1507 G1CollectedHeap* _g1; | |
1508 int _worker_num; | |
1509 size_t _max_live_bytes; | |
1510 size_t _regions_claimed; | |
1511 size_t _freed_bytes; | |
1512 size_t _cleared_h_regions; | |
1513 size_t _freed_regions; | |
1514 UncleanRegionList* _unclean_region_list; | |
1515 double _claimed_region_time; | |
1516 double _max_region_time; | |
1517 | |
1518 public: | |
1519 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1, | |
1520 UncleanRegionList* list, | |
1521 int worker_num); | |
1522 size_t freed_bytes() { return _freed_bytes; } | |
1523 size_t cleared_h_regions() { return _cleared_h_regions; } | |
1524 size_t freed_regions() { return _freed_regions; } | |
1525 UncleanRegionList* unclean_region_list() { | |
1526 return _unclean_region_list; | |
1527 } | |
1528 | |
1529 bool doHeapRegion(HeapRegion *r); | |
1530 | |
1531 size_t max_live_bytes() { return _max_live_bytes; } | |
1532 size_t regions_claimed() { return _regions_claimed; } | |
1533 double claimed_region_time_sec() { return _claimed_region_time; } | |
1534 double max_region_time_sec() { return _max_region_time; } | |
1535 }; | |
1536 | |
1537 class G1ParNoteEndTask: public AbstractGangTask { | |
1538 friend class G1NoteEndOfConcMarkClosure; | |
1539 protected: | |
1540 G1CollectedHeap* _g1h; | |
1541 size_t _max_live_bytes; | |
1542 size_t _freed_bytes; | |
1543 ConcurrentMark::ParCleanupThreadState** _par_cleanup_thread_state; | |
1544 public: | |
1545 G1ParNoteEndTask(G1CollectedHeap* g1h, | |
1546 ConcurrentMark::ParCleanupThreadState** | |
1547 par_cleanup_thread_state) : | |
1548 AbstractGangTask("G1 note end"), _g1h(g1h), | |
1549 _max_live_bytes(0), _freed_bytes(0), | |
1550 _par_cleanup_thread_state(par_cleanup_thread_state) | |
1551 {} | |
1552 | |
1553 void work(int i) { | |
1554 double start = os::elapsedTime(); | |
1555 G1NoteEndOfConcMarkClosure g1_note_end(_g1h, | |
1556 &_par_cleanup_thread_state[i]->list, | |
1557 i); | |
1558 if (ParallelGCThreads > 0) { | |
355 | 1559 _g1h->heap_region_par_iterate_chunked(&g1_note_end, i, |
1560 HeapRegion::NoteEndClaimValue); | |
342 | 1561 } else { |
1562 _g1h->heap_region_iterate(&g1_note_end); | |
1563 } | |
1564 assert(g1_note_end.complete(), "Shouldn't have yielded!"); | |
1565 | |
1566 // Now finish up freeing the current thread's regions. | |
1567 _g1h->finish_free_region_work(g1_note_end.freed_bytes(), | |
1568 g1_note_end.cleared_h_regions(), | |
1569 0, NULL); | |
1570 { | |
1571 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); | |
1572 _max_live_bytes += g1_note_end.max_live_bytes(); | |
1573 _freed_bytes += g1_note_end.freed_bytes(); | |
1574 } | |
1575 double end = os::elapsedTime(); | |
1576 if (G1PrintParCleanupStats) { | |
1577 gclog_or_tty->print(" Worker thread %d [%8.3f..%8.3f = %8.3f ms] " | |
1578 "claimed %d regions (tot = %8.3f ms, max = %8.3f ms).\n", | |
1579 i, start, end, (end-start)*1000.0, | |
1580 g1_note_end.regions_claimed(), | |
1581 g1_note_end.claimed_region_time_sec()*1000.0, | |
1582 g1_note_end.max_region_time_sec()*1000.0); | |
1583 } | |
1584 } | |
1585 size_t max_live_bytes() { return _max_live_bytes; } | |
1586 size_t freed_bytes() { return _freed_bytes; } | |
1587 }; | |
1588 | |
1589 class G1ParScrubRemSetTask: public AbstractGangTask { | |
1590 protected: | |
1591 G1RemSet* _g1rs; | |
1592 BitMap* _region_bm; | |
1593 BitMap* _card_bm; | |
1594 public: | |
1595 G1ParScrubRemSetTask(G1CollectedHeap* g1h, | |
1596 BitMap* region_bm, BitMap* card_bm) : | |
1597 AbstractGangTask("G1 ScrubRS"), _g1rs(g1h->g1_rem_set()), | |
1598 _region_bm(region_bm), _card_bm(card_bm) | |
1599 {} | |
1600 | |
1601 void work(int i) { | |
1602 if (ParallelGCThreads > 0) { | |
355 | 1603 _g1rs->scrub_par(_region_bm, _card_bm, i, |
1604 HeapRegion::ScrubRemSetClaimValue); | |
342 | 1605 } else { |
1606 _g1rs->scrub(_region_bm, _card_bm); | |
1607 } | |
1608 } | |
1609 | |
1610 }; | |
1611 | |
1612 G1NoteEndOfConcMarkClosure:: | |
1613 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1, | |
1614 UncleanRegionList* list, | |
1615 int worker_num) | |
1616 : _g1(g1), _worker_num(worker_num), | |
1617 _max_live_bytes(0), _regions_claimed(0), | |
1618 _freed_bytes(0), _cleared_h_regions(0), _freed_regions(0), | |
1619 _claimed_region_time(0.0), _max_region_time(0.0), | |
1620 _unclean_region_list(list) | |
1621 {} | |
1622 | |
1623 bool G1NoteEndOfConcMarkClosure::doHeapRegion(HeapRegion *r) { | |
1624 // We use a claim value of zero here because all regions | |
1625 // were claimed with value 1 in the FinalCount task. | |
1626 r->reset_gc_time_stamp(); | |
1627 if (!r->continuesHumongous()) { | |
1628 double start = os::elapsedTime(); | |
1629 _regions_claimed++; | |
1630 r->note_end_of_marking(); | |
1631 _max_live_bytes += r->max_live_bytes(); | |
1632 _g1->free_region_if_totally_empty_work(r, | |
1633 _freed_bytes, | |
1634 _cleared_h_regions, | |
1635 _freed_regions, | |
1636 _unclean_region_list, | |
1637 true /*par*/); | |
1638 double region_time = (os::elapsedTime() - start); | |
1639 _claimed_region_time += region_time; | |
1640 if (region_time > _max_region_time) _max_region_time = region_time; | |
1641 } | |
1642 return false; | |
1643 } | |
1644 | |
1645 void ConcurrentMark::cleanup() { | |
1646 // world is stopped at this checkpoint | |
1647 assert(SafepointSynchronize::is_at_safepoint(), | |
1648 "world should be stopped"); | |
1649 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
1650 | |
1651 // If a full collection has happened, we shouldn't do this. | |
1652 if (has_aborted()) { | |
1653 g1h->set_marking_complete(); // So bitmap clearing isn't confused | |
1654 return; | |
1655 } | |
1656 | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1657 if (VerifyDuringGC) { |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1658 HandleMark hm; // handle scope |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1659 gclog_or_tty->print(" VerifyDuringGC:(before)"); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1660 Universe::heap()->prepare_for_verify(); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1661 Universe::verify(/* allow dirty */ true, |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1662 /* silent */ false, |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1663 /* prev marking */ true); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1664 } |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1665 |
342 | 1666 _cleanup_co_tracker.disable(); |
1667 | |
1668 G1CollectorPolicy* g1p = G1CollectedHeap::heap()->g1_policy(); | |
1669 g1p->record_concurrent_mark_cleanup_start(); | |
1670 | |
1671 double start = os::elapsedTime(); | |
1672 GCOverheadReporter::recordSTWStart(start); | |
1673 | |
1674 // Do counting once more with the world stopped for good measure. | |
1675 G1ParFinalCountTask g1_par_count_task(g1h, nextMarkBitMap(), | |
1676 &_region_bm, &_card_bm); | |
1677 if (ParallelGCThreads > 0) { | |
355 | 1678 assert(g1h->check_heap_region_claim_values( |
1679 HeapRegion::InitialClaimValue), | |
1680 "sanity check"); | |
1681 | |
342 | 1682 int n_workers = g1h->workers()->total_workers(); |
1683 g1h->set_par_threads(n_workers); | |
1684 g1h->workers()->run_task(&g1_par_count_task); | |
1685 g1h->set_par_threads(0); | |
355 | 1686 |
1687 assert(g1h->check_heap_region_claim_values( | |
1688 HeapRegion::FinalCountClaimValue), | |
1689 "sanity check"); | |
342 | 1690 } else { |
1691 g1_par_count_task.work(0); | |
1692 } | |
1693 | |
1694 size_t known_garbage_bytes = | |
1695 g1_par_count_task.used_bytes() - g1_par_count_task.live_bytes(); | |
1696 #if 0 | |
1697 gclog_or_tty->print_cr("used %1.2lf, live %1.2lf, garbage %1.2lf", | |
1698 (double) g1_par_count_task.used_bytes() / (double) (1024 * 1024), | |
1699 (double) g1_par_count_task.live_bytes() / (double) (1024 * 1024), | |
1700 (double) known_garbage_bytes / (double) (1024 * 1024)); | |
1701 #endif // 0 | |
1702 g1p->set_known_garbage_bytes(known_garbage_bytes); | |
1703 | |
1704 size_t start_used_bytes = g1h->used(); | |
1705 _at_least_one_mark_complete = true; | |
1706 g1h->set_marking_complete(); | |
1707 | |
1708 double count_end = os::elapsedTime(); | |
1709 double this_final_counting_time = (count_end - start); | |
1710 if (G1PrintParCleanupStats) { | |
1711 gclog_or_tty->print_cr("Cleanup:"); | |
1712 gclog_or_tty->print_cr(" Finalize counting: %8.3f ms", | |
1713 this_final_counting_time*1000.0); | |
1714 } | |
1715 _total_counting_time += this_final_counting_time; | |
1716 | |
1717 // Install newly created mark bitMap as "prev". | |
1718 swapMarkBitMaps(); | |
1719 | |
1720 g1h->reset_gc_time_stamp(); | |
1721 | |
1722 // Note end of marking in all heap regions. | |
1723 double note_end_start = os::elapsedTime(); | |
1724 G1ParNoteEndTask g1_par_note_end_task(g1h, _par_cleanup_thread_state); | |
1725 if (ParallelGCThreads > 0) { | |
1726 int n_workers = g1h->workers()->total_workers(); | |
1727 g1h->set_par_threads(n_workers); | |
1728 g1h->workers()->run_task(&g1_par_note_end_task); | |
1729 g1h->set_par_threads(0); | |
355 | 1730 |
1731 assert(g1h->check_heap_region_claim_values(HeapRegion::NoteEndClaimValue), | |
1732 "sanity check"); | |
342 | 1733 } else { |
1734 g1_par_note_end_task.work(0); | |
1735 } | |
1736 g1h->set_unclean_regions_coming(true); | |
1737 double note_end_end = os::elapsedTime(); | |
1738 // Tell the mutators that there might be unclean regions coming... | |
1739 if (G1PrintParCleanupStats) { | |
1740 gclog_or_tty->print_cr(" note end of marking: %8.3f ms.", | |
1741 (note_end_end - note_end_start)*1000.0); | |
1742 } | |
1743 | |
355 | 1744 |
342 | 1745 // call below, since it affects the metric by which we sort the heap |
1746 // regions. | |
1747 if (G1ScrubRemSets) { | |
1748 double rs_scrub_start = os::elapsedTime(); | |
1749 G1ParScrubRemSetTask g1_par_scrub_rs_task(g1h, &_region_bm, &_card_bm); | |
1750 if (ParallelGCThreads > 0) { | |
1751 int n_workers = g1h->workers()->total_workers(); | |
1752 g1h->set_par_threads(n_workers); | |
1753 g1h->workers()->run_task(&g1_par_scrub_rs_task); | |
1754 g1h->set_par_threads(0); | |
355 | 1755 |
1756 assert(g1h->check_heap_region_claim_values( | |
1757 HeapRegion::ScrubRemSetClaimValue), | |
1758 "sanity check"); | |
342 | 1759 } else { |
1760 g1_par_scrub_rs_task.work(0); | |
1761 } | |
1762 | |
1763 double rs_scrub_end = os::elapsedTime(); | |
1764 double this_rs_scrub_time = (rs_scrub_end - rs_scrub_start); | |
1765 _total_rs_scrub_time += this_rs_scrub_time; | |
1766 } | |
1767 | |
1768 // this will also free any regions totally full of garbage objects, | |
1769 // and sort the regions. | |
1770 g1h->g1_policy()->record_concurrent_mark_cleanup_end( | |
1771 g1_par_note_end_task.freed_bytes(), | |
1772 g1_par_note_end_task.max_live_bytes()); | |
1773 | |
1774 // Statistics. | |
1775 double end = os::elapsedTime(); | |
1776 _cleanup_times.add((end - start) * 1000.0); | |
1777 GCOverheadReporter::recordSTWEnd(end); | |
1778 | |
1779 // G1CollectedHeap::heap()->print(); | |
1780 // gclog_or_tty->print_cr("HEAP GC TIME STAMP : %d", | |
1781 // G1CollectedHeap::heap()->get_gc_time_stamp()); | |
1782 | |
1783 if (PrintGC || PrintGCDetails) { | |
1784 g1h->print_size_transition(gclog_or_tty, | |
1785 start_used_bytes, | |
1786 g1h->used(), | |
1787 g1h->capacity()); | |
1788 } | |
1789 | |
1790 size_t cleaned_up_bytes = start_used_bytes - g1h->used(); | |
1791 g1p->decrease_known_garbage_bytes(cleaned_up_bytes); | |
1792 | |
1793 // We need to make this be a "collection" so any collection pause that | |
1794 // races with it goes around and waits for completeCleanup to finish. | |
1795 g1h->increment_total_collections(); | |
1796 | |
751 | 1797 if (VerifyDuringGC) { |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1798 HandleMark hm; // handle scope |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1799 gclog_or_tty->print(" VerifyDuringGC:(after)"); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1800 Universe::heap()->prepare_for_verify(); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1801 Universe::verify(/* allow dirty */ true, |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1802 /* silent */ false, |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1803 /* prev marking */ true); |
342 | 1804 } |
1805 } | |
1806 | |
1807 void ConcurrentMark::completeCleanup() { | |
1808 // A full collection intervened. | |
1809 if (has_aborted()) return; | |
1810 | |
1811 int first = 0; | |
1812 int last = (int)MAX2(ParallelGCThreads, (size_t)1); | |
1813 for (int t = 0; t < last; t++) { | |
1814 UncleanRegionList* list = &_par_cleanup_thread_state[t]->list; | |
1815 assert(list->well_formed(), "Inv"); | |
1816 HeapRegion* hd = list->hd(); | |
1817 while (hd != NULL) { | |
1818 // Now finish up the other stuff. | |
1819 hd->rem_set()->clear(); | |
1820 HeapRegion* next_hd = hd->next_from_unclean_list(); | |
1821 (void)list->pop(); | |
1822 guarantee(list->hd() == next_hd, "how not?"); | |
1823 _g1h->put_region_on_unclean_list(hd); | |
1824 if (!hd->isHumongous()) { | |
1825 // Add this to the _free_regions count by 1. | |
1826 _g1h->finish_free_region_work(0, 0, 1, NULL); | |
1827 } | |
1828 hd = list->hd(); | |
1829 guarantee(hd == next_hd, "how not?"); | |
1830 } | |
1831 } | |
1832 } | |
1833 | |
1834 | |
1835 class G1CMIsAliveClosure: public BoolObjectClosure { | |
1836 G1CollectedHeap* _g1; | |
1837 public: | |
1838 G1CMIsAliveClosure(G1CollectedHeap* g1) : | |
1839 _g1(g1) | |
1840 {} | |
1841 | |
1842 void do_object(oop obj) { | |
1843 assert(false, "not to be invoked"); | |
1844 } | |
1845 bool do_object_b(oop obj) { | |
1846 HeapWord* addr = (HeapWord*)obj; | |
1847 return addr != NULL && | |
1848 (!_g1->is_in_g1_reserved(addr) || !_g1->is_obj_ill(obj)); | |
1849 } | |
1850 }; | |
1851 | |
1852 class G1CMKeepAliveClosure: public OopClosure { | |
1853 G1CollectedHeap* _g1; | |
1854 ConcurrentMark* _cm; | |
1855 CMBitMap* _bitMap; | |
1856 public: | |
1857 G1CMKeepAliveClosure(G1CollectedHeap* g1, ConcurrentMark* cm, | |
1858 CMBitMap* bitMap) : | |
1859 _g1(g1), _cm(cm), | |
1860 _bitMap(bitMap) {} | |
1861 | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1862 virtual void do_oop(narrowOop* p) { do_oop_work(p); } |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1863 virtual void do_oop( oop* p) { do_oop_work(p); } |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1864 |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1865 template <class T> void do_oop_work(T* p) { |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
1866 oop thisOop = oopDesc::load_decode_heap_oop(p); |
342 | 1867 HeapWord* addr = (HeapWord*)thisOop; |
1868 if (_g1->is_in_g1_reserved(addr) && _g1->is_obj_ill(thisOop)) { | |
1869 _bitMap->mark(addr); | |
1870 _cm->mark_stack_push(thisOop); | |
1871 } | |
1872 } | |
1873 }; | |
1874 | |
1875 class G1CMDrainMarkingStackClosure: public VoidClosure { | |
1876 CMMarkStack* _markStack; | |
1877 CMBitMap* _bitMap; | |
1878 G1CMKeepAliveClosure* _oopClosure; | |
1879 public: | |
1880 G1CMDrainMarkingStackClosure(CMBitMap* bitMap, CMMarkStack* markStack, | |
1881 G1CMKeepAliveClosure* oopClosure) : | |
1882 _bitMap(bitMap), | |
1883 _markStack(markStack), | |
1884 _oopClosure(oopClosure) | |
1885 {} | |
1886 | |
1887 void do_void() { | |
1888 _markStack->drain((OopClosure*)_oopClosure, _bitMap, false); | |
1889 } | |
1890 }; | |
1891 | |
1892 void ConcurrentMark::weakRefsWork(bool clear_all_soft_refs) { | |
1893 ResourceMark rm; | |
1894 HandleMark hm; | |
453
c96030fff130
6684579: SoftReference processing can be made more efficient
ysr
parents:
355
diff
changeset
|
1895 G1CollectedHeap* g1h = G1CollectedHeap::heap(); |
c96030fff130
6684579: SoftReference processing can be made more efficient
ysr
parents:
355
diff
changeset
|
1896 ReferenceProcessor* rp = g1h->ref_processor(); |
342 | 1897 |
1898 // Process weak references. | |
457
27a80744a83b
6778647: snap(), snap_policy() should be renamed setup(), setup_policy()
ysr
parents:
453
diff
changeset
|
1899 rp->setup_policy(clear_all_soft_refs); |
342 | 1900 assert(_markStack.isEmpty(), "mark stack should be empty"); |
1901 | |
453
c96030fff130
6684579: SoftReference processing can be made more efficient
ysr
parents:
355
diff
changeset
|
1902 G1CMIsAliveClosure g1IsAliveClosure (g1h); |
c96030fff130
6684579: SoftReference processing can be made more efficient
ysr
parents:
355
diff
changeset
|
1903 G1CMKeepAliveClosure g1KeepAliveClosure(g1h, this, nextMarkBitMap()); |
342 | 1904 G1CMDrainMarkingStackClosure |
1905 g1DrainMarkingStackClosure(nextMarkBitMap(), &_markStack, | |
1906 &g1KeepAliveClosure); | |
1907 | |
1908 // XXXYYY Also: copy the parallel ref processing code from CMS. | |
453
c96030fff130
6684579: SoftReference processing can be made more efficient
ysr
parents:
355
diff
changeset
|
1909 rp->process_discovered_references(&g1IsAliveClosure, |
342 | 1910 &g1KeepAliveClosure, |
1911 &g1DrainMarkingStackClosure, | |
1912 NULL); | |
1913 assert(_markStack.overflow() || _markStack.isEmpty(), | |
1914 "mark stack should be empty (unless it overflowed)"); | |
1915 if (_markStack.overflow()) { | |
1916 set_has_overflown(); | |
1917 } | |
1918 | |
1919 rp->enqueue_discovered_references(); | |
1920 rp->verify_no_references_recorded(); | |
1921 assert(!rp->discovery_enabled(), "should have been disabled"); | |
1922 | |
1923 // Now clean up stale oops in SymbolTable and StringTable | |
1924 SymbolTable::unlink(&g1IsAliveClosure); | |
1925 StringTable::unlink(&g1IsAliveClosure); | |
1926 } | |
1927 | |
1928 void ConcurrentMark::swapMarkBitMaps() { | |
1929 CMBitMapRO* temp = _prevMarkBitMap; | |
1930 _prevMarkBitMap = (CMBitMapRO*)_nextMarkBitMap; | |
1931 _nextMarkBitMap = (CMBitMap*) temp; | |
1932 } | |
1933 | |
1934 class CMRemarkTask: public AbstractGangTask { | |
1935 private: | |
1936 ConcurrentMark *_cm; | |
1937 | |
1938 public: | |
1939 void work(int worker_i) { | |
1940 // Since all available tasks are actually started, we should | |
1941 // only proceed if we're supposed to be actived. | |
1942 if ((size_t)worker_i < _cm->active_tasks()) { | |
1943 CMTask* task = _cm->task(worker_i); | |
1944 task->record_start_time(); | |
1945 do { | |
1946 task->do_marking_step(1000000000.0 /* something very large */); | |
1947 } while (task->has_aborted() && !_cm->has_overflown()); | |
1948 // If we overflow, then we do not want to restart. We instead | |
1949 // want to abort remark and do concurrent marking again. | |
1950 task->record_end_time(); | |
1951 } | |
1952 } | |
1953 | |
1954 CMRemarkTask(ConcurrentMark* cm) : | |
1955 AbstractGangTask("Par Remark"), _cm(cm) { } | |
1956 }; | |
1957 | |
1958 void ConcurrentMark::checkpointRootsFinalWork() { | |
1959 ResourceMark rm; | |
1960 HandleMark hm; | |
1961 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
1962 | |
1963 g1h->ensure_parsability(false); | |
1964 | |
1965 if (ParallelGCThreads > 0) { | |
1966 g1h->change_strong_roots_parity(); | |
1967 // this is remark, so we'll use up all available threads | |
1968 int active_workers = ParallelGCThreads; | |
1969 set_phase(active_workers, false); | |
1970 | |
1971 CMRemarkTask remarkTask(this); | |
1972 // We will start all available threads, even if we decide that the | |
1973 // active_workers will be fewer. The extra ones will just bail out | |
1974 // immediately. | |
1975 int n_workers = g1h->workers()->total_workers(); | |
1976 g1h->set_par_threads(n_workers); | |
1977 g1h->workers()->run_task(&remarkTask); | |
1978 g1h->set_par_threads(0); | |
1979 | |
1980 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
1981 guarantee( satb_mq_set.completed_buffers_num() == 0, "invariant" ); | |
1982 } else { | |
1983 g1h->change_strong_roots_parity(); | |
1984 // this is remark, so we'll use up all available threads | |
1985 int active_workers = 1; | |
1986 set_phase(active_workers, false); | |
1987 | |
1988 CMRemarkTask remarkTask(this); | |
1989 // We will start all available threads, even if we decide that the | |
1990 // active_workers will be fewer. The extra ones will just bail out | |
1991 // immediately. | |
1992 remarkTask.work(0); | |
1993 | |
1994 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
1995 guarantee( satb_mq_set.completed_buffers_num() == 0, "invariant" ); | |
1996 } | |
1997 | |
1998 print_stats(); | |
1999 | |
2000 if (!restart_for_overflow()) | |
2001 set_non_marking_state(); | |
2002 | |
2003 #if VERIFY_OBJS_PROCESSED | |
2004 if (_scan_obj_cl.objs_processed != ThreadLocalObjQueue::objs_enqueued) { | |
2005 gclog_or_tty->print_cr("Processed = %d, enqueued = %d.", | |
2006 _scan_obj_cl.objs_processed, | |
2007 ThreadLocalObjQueue::objs_enqueued); | |
2008 guarantee(_scan_obj_cl.objs_processed == | |
2009 ThreadLocalObjQueue::objs_enqueued, | |
2010 "Different number of objs processed and enqueued."); | |
2011 } | |
2012 #endif | |
2013 } | |
2014 | |
2015 class ReachablePrinterOopClosure: public OopClosure { | |
2016 private: | |
2017 G1CollectedHeap* _g1h; | |
2018 CMBitMapRO* _bitmap; | |
2019 outputStream* _out; | |
2020 | |
2021 public: | |
2022 ReachablePrinterOopClosure(CMBitMapRO* bitmap, outputStream* out) : | |
2023 _bitmap(bitmap), _g1h(G1CollectedHeap::heap()), _out(out) { } | |
2024 | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2025 void do_oop(narrowOop* p) { do_oop_work(p); } |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2026 void do_oop( oop* p) { do_oop_work(p); } |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2027 |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2028 template <class T> void do_oop_work(T* p) { |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2029 oop obj = oopDesc::load_decode_heap_oop(p); |
342 | 2030 const char* str = NULL; |
2031 const char* str2 = ""; | |
2032 | |
2033 if (!_g1h->is_in_g1_reserved(obj)) | |
2034 str = "outside G1 reserved"; | |
2035 else { | |
2036 HeapRegion* hr = _g1h->heap_region_containing(obj); | |
2037 guarantee( hr != NULL, "invariant" ); | |
2038 if (hr->obj_allocated_since_prev_marking(obj)) { | |
2039 str = "over TAMS"; | |
2040 if (_bitmap->isMarked((HeapWord*) obj)) | |
2041 str2 = " AND MARKED"; | |
2042 } else if (_bitmap->isMarked((HeapWord*) obj)) | |
2043 str = "marked"; | |
2044 else | |
2045 str = "#### NOT MARKED ####"; | |
2046 } | |
2047 | |
2048 _out->print_cr(" "PTR_FORMAT" contains "PTR_FORMAT" %s%s", | |
2049 p, (void*) obj, str, str2); | |
2050 } | |
2051 }; | |
2052 | |
2053 class ReachablePrinterClosure: public BitMapClosure { | |
2054 private: | |
2055 CMBitMapRO* _bitmap; | |
2056 outputStream* _out; | |
2057 | |
2058 public: | |
2059 ReachablePrinterClosure(CMBitMapRO* bitmap, outputStream* out) : | |
2060 _bitmap(bitmap), _out(out) { } | |
2061 | |
2062 bool do_bit(size_t offset) { | |
2063 HeapWord* addr = _bitmap->offsetToHeapWord(offset); | |
2064 ReachablePrinterOopClosure oopCl(_bitmap, _out); | |
2065 | |
2066 _out->print_cr(" obj "PTR_FORMAT", offset %10d (marked)", addr, offset); | |
2067 oop(addr)->oop_iterate(&oopCl); | |
2068 _out->print_cr(""); | |
2069 | |
2070 return true; | |
2071 } | |
2072 }; | |
2073 | |
2074 class ObjInRegionReachablePrinterClosure : public ObjectClosure { | |
2075 private: | |
2076 CMBitMapRO* _bitmap; | |
2077 outputStream* _out; | |
2078 | |
2079 public: | |
2080 void do_object(oop o) { | |
2081 ReachablePrinterOopClosure oopCl(_bitmap, _out); | |
2082 | |
2083 _out->print_cr(" obj "PTR_FORMAT" (over TAMS)", (void*) o); | |
2084 o->oop_iterate(&oopCl); | |
2085 _out->print_cr(""); | |
2086 } | |
2087 | |
2088 ObjInRegionReachablePrinterClosure(CMBitMapRO* bitmap, outputStream* out) : | |
2089 _bitmap(bitmap), _out(out) { } | |
2090 }; | |
2091 | |
2092 class RegionReachablePrinterClosure : public HeapRegionClosure { | |
2093 private: | |
2094 CMBitMapRO* _bitmap; | |
2095 outputStream* _out; | |
2096 | |
2097 public: | |
2098 bool doHeapRegion(HeapRegion* hr) { | |
2099 HeapWord* b = hr->bottom(); | |
2100 HeapWord* e = hr->end(); | |
2101 HeapWord* t = hr->top(); | |
2102 HeapWord* p = hr->prev_top_at_mark_start(); | |
2103 _out->print_cr("** ["PTR_FORMAT", "PTR_FORMAT"] top: "PTR_FORMAT" " | |
2104 "PTAMS: "PTR_FORMAT, b, e, t, p); | |
2105 _out->print_cr(""); | |
2106 | |
2107 ObjInRegionReachablePrinterClosure ocl(_bitmap, _out); | |
2108 hr->object_iterate_mem_careful(MemRegion(p, t), &ocl); | |
2109 | |
2110 return false; | |
2111 } | |
2112 | |
2113 RegionReachablePrinterClosure(CMBitMapRO* bitmap, | |
2114 outputStream* out) : | |
2115 _bitmap(bitmap), _out(out) { } | |
2116 }; | |
2117 | |
2118 void ConcurrentMark::print_prev_bitmap_reachable() { | |
2119 outputStream* out = gclog_or_tty; | |
2120 | |
2121 #if SEND_HEAP_DUMP_TO_FILE | |
2122 guarantee(heap_dump_file == NULL, "Protocol"); | |
2123 char fn_buf[100]; | |
2124 sprintf(fn_buf, "/tmp/dump.txt.%d", os::current_process_id()); | |
2125 heap_dump_file = fopen(fn_buf, "w"); | |
2126 fileStream fstream(heap_dump_file); | |
2127 out = &fstream; | |
2128 #endif // SEND_HEAP_DUMP_TO_FILE | |
2129 | |
2130 RegionReachablePrinterClosure rcl(_prevMarkBitMap, out); | |
2131 out->print_cr("--- ITERATING OVER REGIONS WITH PTAMS < TOP"); | |
2132 _g1h->heap_region_iterate(&rcl); | |
2133 out->print_cr(""); | |
2134 | |
2135 ReachablePrinterClosure cl(_prevMarkBitMap, out); | |
2136 out->print_cr("--- REACHABLE OBJECTS ON THE BITMAP"); | |
2137 _prevMarkBitMap->iterate(&cl); | |
2138 out->print_cr(""); | |
2139 | |
2140 #if SEND_HEAP_DUMP_TO_FILE | |
2141 fclose(heap_dump_file); | |
2142 heap_dump_file = NULL; | |
2143 #endif // SEND_HEAP_DUMP_TO_FILE | |
2144 } | |
2145 | |
2146 // This note is for drainAllSATBBuffers and the code in between. | |
2147 // In the future we could reuse a task to do this work during an | |
2148 // evacuation pause (since now tasks are not active and can be claimed | |
2149 // during an evacuation pause). This was a late change to the code and | |
2150 // is currently not being taken advantage of. | |
2151 | |
2152 class CMGlobalObjectClosure : public ObjectClosure { | |
2153 private: | |
2154 ConcurrentMark* _cm; | |
2155 | |
2156 public: | |
2157 void do_object(oop obj) { | |
2158 _cm->deal_with_reference(obj); | |
2159 } | |
2160 | |
2161 CMGlobalObjectClosure(ConcurrentMark* cm) : _cm(cm) { } | |
2162 }; | |
2163 | |
2164 void ConcurrentMark::deal_with_reference(oop obj) { | |
2165 if (verbose_high()) | |
2166 gclog_or_tty->print_cr("[global] we're dealing with reference "PTR_FORMAT, | |
2167 (void*) obj); | |
2168 | |
2169 | |
2170 HeapWord* objAddr = (HeapWord*) obj; | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2171 assert(obj->is_oop_or_null(true /* ignore mark word */), "Error"); |
342 | 2172 if (_g1h->is_in_g1_reserved(objAddr)) { |
2173 tmp_guarantee_CM( obj != NULL, "is_in_g1_reserved should ensure this" ); | |
2174 HeapRegion* hr = _g1h->heap_region_containing(obj); | |
2175 if (_g1h->is_obj_ill(obj, hr)) { | |
2176 if (verbose_high()) | |
2177 gclog_or_tty->print_cr("[global] "PTR_FORMAT" is not considered " | |
2178 "marked", (void*) obj); | |
2179 | |
2180 // we need to mark it first | |
2181 if (_nextMarkBitMap->parMark(objAddr)) { | |
2182 // No OrderAccess:store_load() is needed. It is implicit in the | |
2183 // CAS done in parMark(objAddr) above | |
2184 HeapWord* finger = _finger; | |
2185 if (objAddr < finger) { | |
2186 if (verbose_high()) | |
2187 gclog_or_tty->print_cr("[global] below the global finger " | |
2188 "("PTR_FORMAT"), pushing it", finger); | |
2189 if (!mark_stack_push(obj)) { | |
2190 if (verbose_low()) | |
2191 gclog_or_tty->print_cr("[global] global stack overflow during " | |
2192 "deal_with_reference"); | |
2193 } | |
2194 } | |
2195 } | |
2196 } | |
2197 } | |
2198 } | |
2199 | |
2200 void ConcurrentMark::drainAllSATBBuffers() { | |
2201 CMGlobalObjectClosure oc(this); | |
2202 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
2203 satb_mq_set.set_closure(&oc); | |
2204 | |
2205 while (satb_mq_set.apply_closure_to_completed_buffer()) { | |
2206 if (verbose_medium()) | |
2207 gclog_or_tty->print_cr("[global] processed an SATB buffer"); | |
2208 } | |
2209 | |
2210 // no need to check whether we should do this, as this is only | |
2211 // called during an evacuation pause | |
2212 satb_mq_set.iterate_closure_all_threads(); | |
2213 | |
2214 satb_mq_set.set_closure(NULL); | |
2215 guarantee( satb_mq_set.completed_buffers_num() == 0, "invariant" ); | |
2216 } | |
2217 | |
2218 void ConcurrentMark::markPrev(oop p) { | |
2219 // Note we are overriding the read-only view of the prev map here, via | |
2220 // the cast. | |
2221 ((CMBitMap*)_prevMarkBitMap)->mark((HeapWord*)p); | |
2222 } | |
2223 | |
2224 void ConcurrentMark::clear(oop p) { | |
2225 assert(p != NULL && p->is_oop(), "expected an oop"); | |
2226 HeapWord* addr = (HeapWord*)p; | |
2227 assert(addr >= _nextMarkBitMap->startWord() || | |
2228 addr < _nextMarkBitMap->endWord(), "in a region"); | |
2229 | |
2230 _nextMarkBitMap->clear(addr); | |
2231 } | |
2232 | |
2233 void ConcurrentMark::clearRangeBothMaps(MemRegion mr) { | |
2234 // Note we are overriding the read-only view of the prev map here, via | |
2235 // the cast. | |
2236 ((CMBitMap*)_prevMarkBitMap)->clearRange(mr); | |
2237 _nextMarkBitMap->clearRange(mr); | |
2238 } | |
2239 | |
2240 HeapRegion* | |
2241 ConcurrentMark::claim_region(int task_num) { | |
2242 // "checkpoint" the finger | |
2243 HeapWord* finger = _finger; | |
2244 | |
2245 // _heap_end will not change underneath our feet; it only changes at | |
2246 // yield points. | |
2247 while (finger < _heap_end) { | |
2248 tmp_guarantee_CM( _g1h->is_in_g1_reserved(finger), "invariant" ); | |
2249 | |
2250 // is the gap between reading the finger and doing the CAS too long? | |
2251 | |
2252 HeapRegion* curr_region = _g1h->heap_region_containing(finger); | |
2253 HeapWord* bottom = curr_region->bottom(); | |
2254 HeapWord* end = curr_region->end(); | |
2255 HeapWord* limit = curr_region->next_top_at_mark_start(); | |
2256 | |
2257 if (verbose_low()) | |
2258 gclog_or_tty->print_cr("[%d] curr_region = "PTR_FORMAT" " | |
2259 "["PTR_FORMAT", "PTR_FORMAT"), " | |
2260 "limit = "PTR_FORMAT, | |
2261 task_num, curr_region, bottom, end, limit); | |
2262 | |
2263 HeapWord* res = | |
2264 (HeapWord*) Atomic::cmpxchg_ptr(end, &_finger, finger); | |
2265 if (res == finger) { | |
2266 // we succeeded | |
2267 | |
2268 // notice that _finger == end cannot be guaranteed here since, | |
2269 // someone else might have moved the finger even further | |
2270 guarantee( _finger >= end, "the finger should have moved forward" ); | |
2271 | |
2272 if (verbose_low()) | |
2273 gclog_or_tty->print_cr("[%d] we were successful with region = " | |
2274 PTR_FORMAT, task_num, curr_region); | |
2275 | |
2276 if (limit > bottom) { | |
2277 if (verbose_low()) | |
2278 gclog_or_tty->print_cr("[%d] region "PTR_FORMAT" is not empty, " | |
2279 "returning it ", task_num, curr_region); | |
2280 return curr_region; | |
2281 } else { | |
2282 tmp_guarantee_CM( limit == bottom, | |
2283 "the region limit should be at bottom" ); | |
2284 if (verbose_low()) | |
2285 gclog_or_tty->print_cr("[%d] region "PTR_FORMAT" is empty, " | |
2286 "returning NULL", task_num, curr_region); | |
2287 // we return NULL and the caller should try calling | |
2288 // claim_region() again. | |
2289 return NULL; | |
2290 } | |
2291 } else { | |
2292 guarantee( _finger > finger, "the finger should have moved forward" ); | |
2293 if (verbose_low()) | |
2294 gclog_or_tty->print_cr("[%d] somebody else moved the finger, " | |
2295 "global finger = "PTR_FORMAT", " | |
2296 "our finger = "PTR_FORMAT, | |
2297 task_num, _finger, finger); | |
2298 | |
2299 // read it again | |
2300 finger = _finger; | |
2301 } | |
2302 } | |
2303 | |
2304 return NULL; | |
2305 } | |
2306 | |
2307 void ConcurrentMark::oops_do(OopClosure* cl) { | |
2308 if (_markStack.size() > 0 && verbose_low()) | |
2309 gclog_or_tty->print_cr("[global] scanning the global marking stack, " | |
2310 "size = %d", _markStack.size()); | |
2311 // we first iterate over the contents of the mark stack... | |
2312 _markStack.oops_do(cl); | |
2313 | |
2314 for (int i = 0; i < (int)_max_task_num; ++i) { | |
2315 OopTaskQueue* queue = _task_queues->queue((int)i); | |
2316 | |
2317 if (queue->size() > 0 && verbose_low()) | |
2318 gclog_or_tty->print_cr("[global] scanning task queue of task %d, " | |
2319 "size = %d", i, queue->size()); | |
2320 | |
2321 // ...then over the contents of the all the task queues. | |
2322 queue->oops_do(cl); | |
2323 } | |
2324 | |
2325 // finally, invalidate any entries that in the region stack that | |
2326 // point into the collection set | |
2327 if (_regionStack.invalidate_entries_into_cset()) { | |
2328 // otherwise, any gray objects copied during the evacuation pause | |
2329 // might not be visited. | |
2330 guarantee( _should_gray_objects, "invariant" ); | |
2331 } | |
2332 } | |
2333 | |
2334 void ConcurrentMark::clear_marking_state() { | |
2335 _markStack.setEmpty(); | |
2336 _markStack.clear_overflow(); | |
2337 _regionStack.setEmpty(); | |
2338 _regionStack.clear_overflow(); | |
2339 clear_has_overflown(); | |
2340 _finger = _heap_start; | |
2341 | |
2342 for (int i = 0; i < (int)_max_task_num; ++i) { | |
2343 OopTaskQueue* queue = _task_queues->queue(i); | |
2344 queue->set_empty(); | |
2345 } | |
2346 } | |
2347 | |
2348 void ConcurrentMark::print_stats() { | |
2349 if (verbose_stats()) { | |
2350 gclog_or_tty->print_cr("---------------------------------------------------------------------"); | |
2351 for (size_t i = 0; i < _active_tasks; ++i) { | |
2352 _tasks[i]->print_stats(); | |
2353 gclog_or_tty->print_cr("---------------------------------------------------------------------"); | |
2354 } | |
2355 } | |
2356 } | |
2357 | |
2358 class CSMarkOopClosure: public OopClosure { | |
2359 friend class CSMarkBitMapClosure; | |
2360 | |
2361 G1CollectedHeap* _g1h; | |
2362 CMBitMap* _bm; | |
2363 ConcurrentMark* _cm; | |
2364 oop* _ms; | |
2365 jint* _array_ind_stack; | |
2366 int _ms_size; | |
2367 int _ms_ind; | |
2368 int _array_increment; | |
2369 | |
2370 bool push(oop obj, int arr_ind = 0) { | |
2371 if (_ms_ind == _ms_size) { | |
2372 gclog_or_tty->print_cr("Mark stack is full."); | |
2373 return false; | |
2374 } | |
2375 _ms[_ms_ind] = obj; | |
2376 if (obj->is_objArray()) _array_ind_stack[_ms_ind] = arr_ind; | |
2377 _ms_ind++; | |
2378 return true; | |
2379 } | |
2380 | |
2381 oop pop() { | |
2382 if (_ms_ind == 0) return NULL; | |
2383 else { | |
2384 _ms_ind--; | |
2385 return _ms[_ms_ind]; | |
2386 } | |
2387 } | |
2388 | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2389 template <class T> bool drain() { |
342 | 2390 while (_ms_ind > 0) { |
2391 oop obj = pop(); | |
2392 assert(obj != NULL, "Since index was non-zero."); | |
2393 if (obj->is_objArray()) { | |
2394 jint arr_ind = _array_ind_stack[_ms_ind]; | |
2395 objArrayOop aobj = objArrayOop(obj); | |
2396 jint len = aobj->length(); | |
2397 jint next_arr_ind = arr_ind + _array_increment; | |
2398 if (next_arr_ind < len) { | |
2399 push(obj, next_arr_ind); | |
2400 } | |
2401 // Now process this portion of this one. | |
2402 int lim = MIN2(next_arr_ind, len); | |
2403 for (int j = arr_ind; j < lim; j++) { | |
912
308762b2bf14
6872000: G1: compilation fails on linux/older gcc
apetrusenko
parents:
866
diff
changeset
|
2404 do_oop(aobj->objArrayOopDesc::obj_at_addr<T>(j)); |
342 | 2405 } |
2406 | |
2407 } else { | |
2408 obj->oop_iterate(this); | |
2409 } | |
2410 if (abort()) return false; | |
2411 } | |
2412 return true; | |
2413 } | |
2414 | |
2415 public: | |
2416 CSMarkOopClosure(ConcurrentMark* cm, int ms_size) : | |
2417 _g1h(G1CollectedHeap::heap()), | |
2418 _cm(cm), | |
2419 _bm(cm->nextMarkBitMap()), | |
2420 _ms_size(ms_size), _ms_ind(0), | |
2421 _ms(NEW_C_HEAP_ARRAY(oop, ms_size)), | |
2422 _array_ind_stack(NEW_C_HEAP_ARRAY(jint, ms_size)), | |
2423 _array_increment(MAX2(ms_size/8, 16)) | |
2424 {} | |
2425 | |
2426 ~CSMarkOopClosure() { | |
2427 FREE_C_HEAP_ARRAY(oop, _ms); | |
2428 FREE_C_HEAP_ARRAY(jint, _array_ind_stack); | |
2429 } | |
2430 | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2431 virtual void do_oop(narrowOop* p) { do_oop_work(p); } |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2432 virtual void do_oop( oop* p) { do_oop_work(p); } |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2433 |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2434 template <class T> void do_oop_work(T* p) { |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2435 T heap_oop = oopDesc::load_heap_oop(p); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2436 if (oopDesc::is_null(heap_oop)) return; |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2437 oop obj = oopDesc::decode_heap_oop_not_null(heap_oop); |
342 | 2438 if (obj->is_forwarded()) { |
2439 // If the object has already been forwarded, we have to make sure | |
2440 // that it's marked. So follow the forwarding pointer. Note that | |
2441 // this does the right thing for self-forwarding pointers in the | |
2442 // evacuation failure case. | |
2443 obj = obj->forwardee(); | |
2444 } | |
2445 HeapRegion* hr = _g1h->heap_region_containing(obj); | |
2446 if (hr != NULL) { | |
2447 if (hr->in_collection_set()) { | |
2448 if (_g1h->is_obj_ill(obj)) { | |
2449 _bm->mark((HeapWord*)obj); | |
2450 if (!push(obj)) { | |
2451 gclog_or_tty->print_cr("Setting abort in CSMarkOopClosure because push failed."); | |
2452 set_abort(); | |
2453 } | |
2454 } | |
2455 } else { | |
2456 // Outside the collection set; we need to gray it | |
2457 _cm->deal_with_reference(obj); | |
2458 } | |
2459 } | |
2460 } | |
2461 }; | |
2462 | |
2463 class CSMarkBitMapClosure: public BitMapClosure { | |
2464 G1CollectedHeap* _g1h; | |
2465 CMBitMap* _bitMap; | |
2466 ConcurrentMark* _cm; | |
2467 CSMarkOopClosure _oop_cl; | |
2468 public: | |
2469 CSMarkBitMapClosure(ConcurrentMark* cm, int ms_size) : | |
2470 _g1h(G1CollectedHeap::heap()), | |
2471 _bitMap(cm->nextMarkBitMap()), | |
2472 _oop_cl(cm, ms_size) | |
2473 {} | |
2474 | |
2475 ~CSMarkBitMapClosure() {} | |
2476 | |
2477 bool do_bit(size_t offset) { | |
2478 // convert offset into a HeapWord* | |
2479 HeapWord* addr = _bitMap->offsetToHeapWord(offset); | |
2480 assert(_bitMap->endWord() && addr < _bitMap->endWord(), | |
2481 "address out of range"); | |
2482 assert(_bitMap->isMarked(addr), "tautology"); | |
2483 oop obj = oop(addr); | |
2484 if (!obj->is_forwarded()) { | |
2485 if (!_oop_cl.push(obj)) return false; | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2486 if (UseCompressedOops) { |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2487 if (!_oop_cl.drain<narrowOop>()) return false; |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2488 } else { |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2489 if (!_oop_cl.drain<oop>()) return false; |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2490 } |
342 | 2491 } |
2492 // Otherwise... | |
2493 return true; | |
2494 } | |
2495 }; | |
2496 | |
2497 | |
2498 class CompleteMarkingInCSHRClosure: public HeapRegionClosure { | |
2499 CMBitMap* _bm; | |
2500 CSMarkBitMapClosure _bit_cl; | |
2501 enum SomePrivateConstants { | |
2502 MSSize = 1000 | |
2503 }; | |
2504 bool _completed; | |
2505 public: | |
2506 CompleteMarkingInCSHRClosure(ConcurrentMark* cm) : | |
2507 _bm(cm->nextMarkBitMap()), | |
2508 _bit_cl(cm, MSSize), | |
2509 _completed(true) | |
2510 {} | |
2511 | |
2512 ~CompleteMarkingInCSHRClosure() {} | |
2513 | |
2514 bool doHeapRegion(HeapRegion* r) { | |
2515 if (!r->evacuation_failed()) { | |
2516 MemRegion mr = MemRegion(r->bottom(), r->next_top_at_mark_start()); | |
2517 if (!mr.is_empty()) { | |
2518 if (!_bm->iterate(&_bit_cl, mr)) { | |
2519 _completed = false; | |
2520 return true; | |
2521 } | |
2522 } | |
2523 } | |
2524 return false; | |
2525 } | |
2526 | |
2527 bool completed() { return _completed; } | |
2528 }; | |
2529 | |
2530 class ClearMarksInHRClosure: public HeapRegionClosure { | |
2531 CMBitMap* _bm; | |
2532 public: | |
2533 ClearMarksInHRClosure(CMBitMap* bm): _bm(bm) { } | |
2534 | |
2535 bool doHeapRegion(HeapRegion* r) { | |
2536 if (!r->used_region().is_empty() && !r->evacuation_failed()) { | |
2537 MemRegion usedMR = r->used_region(); | |
2538 _bm->clearRange(r->used_region()); | |
2539 } | |
2540 return false; | |
2541 } | |
2542 }; | |
2543 | |
2544 void ConcurrentMark::complete_marking_in_collection_set() { | |
2545 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
2546 | |
2547 if (!g1h->mark_in_progress()) { | |
2548 g1h->g1_policy()->record_mark_closure_time(0.0); | |
2549 return; | |
2550 } | |
2551 | |
2552 int i = 1; | |
2553 double start = os::elapsedTime(); | |
2554 while (true) { | |
2555 i++; | |
2556 CompleteMarkingInCSHRClosure cmplt(this); | |
2557 g1h->collection_set_iterate(&cmplt); | |
2558 if (cmplt.completed()) break; | |
2559 } | |
2560 double end_time = os::elapsedTime(); | |
2561 double elapsed_time_ms = (end_time - start) * 1000.0; | |
2562 g1h->g1_policy()->record_mark_closure_time(elapsed_time_ms); | |
2563 if (PrintGCDetails) { | |
2564 gclog_or_tty->print_cr("Mark closure took %5.2f ms.", elapsed_time_ms); | |
2565 } | |
2566 | |
2567 ClearMarksInHRClosure clr(nextMarkBitMap()); | |
2568 g1h->collection_set_iterate(&clr); | |
2569 } | |
2570 | |
2571 // The next two methods deal with the following optimisation. Some | |
2572 // objects are gray by being marked and located above the finger. If | |
2573 // they are copied, during an evacuation pause, below the finger then | |
2574 // the need to be pushed on the stack. The observation is that, if | |
2575 // there are no regions in the collection set located above the | |
2576 // finger, then the above cannot happen, hence we do not need to | |
2577 // explicitly gray any objects when copying them to below the | |
2578 // finger. The global stack will be scanned to ensure that, if it | |
2579 // points to objects being copied, it will update their | |
2580 // location. There is a tricky situation with the gray objects in | |
2581 // region stack that are being coped, however. See the comment in | |
2582 // newCSet(). | |
2583 | |
2584 void ConcurrentMark::newCSet() { | |
2585 if (!concurrent_marking_in_progress()) | |
2586 // nothing to do if marking is not in progress | |
2587 return; | |
2588 | |
2589 // find what the lowest finger is among the global and local fingers | |
2590 _min_finger = _finger; | |
2591 for (int i = 0; i < (int)_max_task_num; ++i) { | |
2592 CMTask* task = _tasks[i]; | |
2593 HeapWord* task_finger = task->finger(); | |
2594 if (task_finger != NULL && task_finger < _min_finger) | |
2595 _min_finger = task_finger; | |
2596 } | |
2597 | |
2598 _should_gray_objects = false; | |
2599 | |
2600 // This fixes a very subtle and fustrating bug. It might be the case | |
2601 // that, during en evacuation pause, heap regions that contain | |
2602 // objects that are gray (by being in regions contained in the | |
2603 // region stack) are included in the collection set. Since such gray | |
2604 // objects will be moved, and because it's not easy to redirect | |
2605 // region stack entries to point to a new location (because objects | |
2606 // in one region might be scattered to multiple regions after they | |
2607 // are copied), one option is to ensure that all marked objects | |
2608 // copied during a pause are pushed on the stack. Notice, however, | |
2609 // that this problem can only happen when the region stack is not | |
2610 // empty during an evacuation pause. So, we make the fix a bit less | |
2611 // conservative and ensure that regions are pushed on the stack, | |
2612 // irrespective whether all collection set regions are below the | |
2613 // finger, if the region stack is not empty. This is expected to be | |
2614 // a rare case, so I don't think it's necessary to be smarted about it. | |
2615 if (!region_stack_empty()) | |
2616 _should_gray_objects = true; | |
2617 } | |
2618 | |
2619 void ConcurrentMark::registerCSetRegion(HeapRegion* hr) { | |
2620 if (!concurrent_marking_in_progress()) | |
2621 return; | |
2622 | |
2623 HeapWord* region_end = hr->end(); | |
2624 if (region_end > _min_finger) | |
2625 _should_gray_objects = true; | |
2626 } | |
2627 | |
2628 void ConcurrentMark::disable_co_trackers() { | |
2629 if (has_aborted()) { | |
2630 if (_cleanup_co_tracker.enabled()) | |
2631 _cleanup_co_tracker.disable(); | |
2632 for (int i = 0; i < (int)_max_task_num; ++i) { | |
2633 CMTask* task = _tasks[i]; | |
2634 if (task->co_tracker_enabled()) | |
2635 task->disable_co_tracker(); | |
2636 } | |
2637 } else { | |
2638 guarantee( !_cleanup_co_tracker.enabled(), "invariant" ); | |
2639 for (int i = 0; i < (int)_max_task_num; ++i) { | |
2640 CMTask* task = _tasks[i]; | |
2641 guarantee( !task->co_tracker_enabled(), "invariant" ); | |
2642 } | |
2643 } | |
2644 } | |
2645 | |
2646 // abandon current marking iteration due to a Full GC | |
2647 void ConcurrentMark::abort() { | |
2648 // Clear all marks to force marking thread to do nothing | |
2649 _nextMarkBitMap->clearAll(); | |
2650 // Empty mark stack | |
2651 clear_marking_state(); | |
2652 for (int i = 0; i < (int)_max_task_num; ++i) | |
2653 _tasks[i]->clear_region_fields(); | |
2654 _has_aborted = true; | |
2655 | |
2656 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
2657 satb_mq_set.abandon_partial_marking(); | |
2658 satb_mq_set.set_active_all_threads(false); | |
2659 } | |
2660 | |
2661 static void print_ms_time_info(const char* prefix, const char* name, | |
2662 NumberSeq& ns) { | |
2663 gclog_or_tty->print_cr("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).", | |
2664 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg()); | |
2665 if (ns.num() > 0) { | |
2666 gclog_or_tty->print_cr("%s [std. dev = %8.2f ms, max = %8.2f ms]", | |
2667 prefix, ns.sd(), ns.maximum()); | |
2668 } | |
2669 } | |
2670 | |
2671 void ConcurrentMark::print_summary_info() { | |
2672 gclog_or_tty->print_cr(" Concurrent marking:"); | |
2673 print_ms_time_info(" ", "init marks", _init_times); | |
2674 print_ms_time_info(" ", "remarks", _remark_times); | |
2675 { | |
2676 print_ms_time_info(" ", "final marks", _remark_mark_times); | |
2677 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times); | |
2678 | |
2679 } | |
2680 print_ms_time_info(" ", "cleanups", _cleanup_times); | |
2681 gclog_or_tty->print_cr(" Final counting total time = %8.2f s (avg = %8.2f ms).", | |
2682 _total_counting_time, | |
2683 (_cleanup_times.num() > 0 ? _total_counting_time * 1000.0 / | |
2684 (double)_cleanup_times.num() | |
2685 : 0.0)); | |
2686 if (G1ScrubRemSets) { | |
2687 gclog_or_tty->print_cr(" RS scrub total time = %8.2f s (avg = %8.2f ms).", | |
2688 _total_rs_scrub_time, | |
2689 (_cleanup_times.num() > 0 ? _total_rs_scrub_time * 1000.0 / | |
2690 (double)_cleanup_times.num() | |
2691 : 0.0)); | |
2692 } | |
2693 gclog_or_tty->print_cr(" Total stop_world time = %8.2f s.", | |
2694 (_init_times.sum() + _remark_times.sum() + | |
2695 _cleanup_times.sum())/1000.0); | |
2696 gclog_or_tty->print_cr(" Total concurrent time = %8.2f s " | |
2697 "(%8.2f s marking, %8.2f s counting).", | |
2698 cmThread()->vtime_accum(), | |
2699 cmThread()->vtime_mark_accum(), | |
2700 cmThread()->vtime_count_accum()); | |
2701 } | |
2702 | |
2703 // Closures | |
2704 // XXX: there seems to be a lot of code duplication here; | |
2705 // should refactor and consolidate the shared code. | |
2706 | |
2707 // This closure is used to mark refs into the CMS generation in | |
2708 // the CMS bit map. Called at the first checkpoint. | |
2709 | |
2710 // We take a break if someone is trying to stop the world. | |
2711 bool ConcurrentMark::do_yield_check(int worker_i) { | |
2712 if (should_yield()) { | |
2713 if (worker_i == 0) | |
2714 _g1h->g1_policy()->record_concurrent_pause(); | |
2715 cmThread()->yield(); | |
2716 if (worker_i == 0) | |
2717 _g1h->g1_policy()->record_concurrent_pause_end(); | |
2718 return true; | |
2719 } else { | |
2720 return false; | |
2721 } | |
2722 } | |
2723 | |
2724 bool ConcurrentMark::should_yield() { | |
2725 return cmThread()->should_yield(); | |
2726 } | |
2727 | |
2728 bool ConcurrentMark::containing_card_is_marked(void* p) { | |
2729 size_t offset = pointer_delta(p, _g1h->reserved_region().start(), 1); | |
2730 return _card_bm.at(offset >> CardTableModRefBS::card_shift); | |
2731 } | |
2732 | |
2733 bool ConcurrentMark::containing_cards_are_marked(void* start, | |
2734 void* last) { | |
2735 return | |
2736 containing_card_is_marked(start) && | |
2737 containing_card_is_marked(last); | |
2738 } | |
2739 | |
2740 #ifndef PRODUCT | |
2741 // for debugging purposes | |
2742 void ConcurrentMark::print_finger() { | |
2743 gclog_or_tty->print_cr("heap ["PTR_FORMAT", "PTR_FORMAT"), global finger = "PTR_FORMAT, | |
2744 _heap_start, _heap_end, _finger); | |
2745 for (int i = 0; i < (int) _max_task_num; ++i) { | |
2746 gclog_or_tty->print(" %d: "PTR_FORMAT, i, _tasks[i]->finger()); | |
2747 } | |
2748 gclog_or_tty->print_cr(""); | |
2749 } | |
2750 #endif | |
2751 | |
2752 // Closure for iteration over bitmaps | |
2753 class CMBitMapClosure : public BitMapClosure { | |
2754 private: | |
2755 // the bitmap that is being iterated over | |
2756 CMBitMap* _nextMarkBitMap; | |
2757 ConcurrentMark* _cm; | |
2758 CMTask* _task; | |
2759 // true if we're scanning a heap region claimed by the task (so that | |
2760 // we move the finger along), false if we're not, i.e. currently when | |
2761 // scanning a heap region popped from the region stack (so that we | |
2762 // do not move the task finger along; it'd be a mistake if we did so). | |
2763 bool _scanning_heap_region; | |
2764 | |
2765 public: | |
2766 CMBitMapClosure(CMTask *task, | |
2767 ConcurrentMark* cm, | |
2768 CMBitMap* nextMarkBitMap) | |
2769 : _task(task), _cm(cm), _nextMarkBitMap(nextMarkBitMap) { } | |
2770 | |
2771 void set_scanning_heap_region(bool scanning_heap_region) { | |
2772 _scanning_heap_region = scanning_heap_region; | |
2773 } | |
2774 | |
2775 bool do_bit(size_t offset) { | |
2776 HeapWord* addr = _nextMarkBitMap->offsetToHeapWord(offset); | |
2777 tmp_guarantee_CM( _nextMarkBitMap->isMarked(addr), "invariant" ); | |
2778 tmp_guarantee_CM( addr < _cm->finger(), "invariant" ); | |
2779 | |
2780 if (_scanning_heap_region) { | |
2781 statsOnly( _task->increase_objs_found_on_bitmap() ); | |
2782 tmp_guarantee_CM( addr >= _task->finger(), "invariant" ); | |
2783 // We move that task's local finger along. | |
2784 _task->move_finger_to(addr); | |
2785 } else { | |
2786 // We move the task's region finger along. | |
2787 _task->move_region_finger_to(addr); | |
2788 } | |
2789 | |
2790 _task->scan_object(oop(addr)); | |
2791 // we only partially drain the local queue and global stack | |
2792 _task->drain_local_queue(true); | |
2793 _task->drain_global_stack(true); | |
2794 | |
2795 // if the has_aborted flag has been raised, we need to bail out of | |
2796 // the iteration | |
2797 return !_task->has_aborted(); | |
2798 } | |
2799 }; | |
2800 | |
2801 // Closure for iterating over objects, currently only used for | |
2802 // processing SATB buffers. | |
2803 class CMObjectClosure : public ObjectClosure { | |
2804 private: | |
2805 CMTask* _task; | |
2806 | |
2807 public: | |
2808 void do_object(oop obj) { | |
2809 _task->deal_with_reference(obj); | |
2810 } | |
2811 | |
2812 CMObjectClosure(CMTask* task) : _task(task) { } | |
2813 }; | |
2814 | |
2815 // Closure for iterating over object fields | |
2816 class CMOopClosure : public OopClosure { | |
2817 private: | |
2818 G1CollectedHeap* _g1h; | |
2819 ConcurrentMark* _cm; | |
2820 CMTask* _task; | |
2821 | |
2822 public: | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2823 virtual void do_oop(narrowOop* p) { do_oop_work(p); } |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2824 virtual void do_oop( oop* p) { do_oop_work(p); } |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2825 |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2826 template <class T> void do_oop_work(T* p) { |
342 | 2827 tmp_guarantee_CM( _g1h->is_in_g1_reserved((HeapWord*) p), "invariant" ); |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2828 tmp_guarantee_CM( !_g1h->heap_region_containing((HeapWord*) p)->is_on_free_list(), "invariant" ); |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2829 |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2830 oop obj = oopDesc::load_decode_heap_oop(p); |
342 | 2831 if (_cm->verbose_high()) |
2832 gclog_or_tty->print_cr("[%d] we're looking at location " | |
2833 "*"PTR_FORMAT" = "PTR_FORMAT, | |
2834 _task->task_id(), p, (void*) obj); | |
2835 _task->deal_with_reference(obj); | |
2836 } | |
2837 | |
2838 CMOopClosure(G1CollectedHeap* g1h, | |
2839 ConcurrentMark* cm, | |
2840 CMTask* task) | |
2841 : _g1h(g1h), _cm(cm), _task(task) { } | |
2842 }; | |
2843 | |
2844 void CMTask::setup_for_region(HeapRegion* hr) { | |
2845 tmp_guarantee_CM( hr != NULL && !hr->continuesHumongous(), | |
2846 "claim_region() should have filtered out continues humongous regions" ); | |
2847 | |
2848 if (_cm->verbose_low()) | |
2849 gclog_or_tty->print_cr("[%d] setting up for region "PTR_FORMAT, | |
2850 _task_id, hr); | |
2851 | |
2852 _curr_region = hr; | |
2853 _finger = hr->bottom(); | |
2854 update_region_limit(); | |
2855 } | |
2856 | |
2857 void CMTask::update_region_limit() { | |
2858 HeapRegion* hr = _curr_region; | |
2859 HeapWord* bottom = hr->bottom(); | |
2860 HeapWord* limit = hr->next_top_at_mark_start(); | |
2861 | |
2862 if (limit == bottom) { | |
2863 if (_cm->verbose_low()) | |
2864 gclog_or_tty->print_cr("[%d] found an empty region " | |
2865 "["PTR_FORMAT", "PTR_FORMAT")", | |
2866 _task_id, bottom, limit); | |
2867 // The region was collected underneath our feet. | |
2868 // We set the finger to bottom to ensure that the bitmap | |
2869 // iteration that will follow this will not do anything. | |
2870 // (this is not a condition that holds when we set the region up, | |
2871 // as the region is not supposed to be empty in the first place) | |
2872 _finger = bottom; | |
2873 } else if (limit >= _region_limit) { | |
2874 tmp_guarantee_CM( limit >= _finger, "peace of mind" ); | |
2875 } else { | |
2876 tmp_guarantee_CM( limit < _region_limit, "only way to get here" ); | |
2877 // This can happen under some pretty unusual circumstances. An | |
2878 // evacuation pause empties the region underneath our feet (NTAMS | |
2879 // at bottom). We then do some allocation in the region (NTAMS | |
2880 // stays at bottom), followed by the region being used as a GC | |
2881 // alloc region (NTAMS will move to top() and the objects | |
2882 // originally below it will be grayed). All objects now marked in | |
2883 // the region are explicitly grayed, if below the global finger, | |
2884 // and we do not need in fact to scan anything else. So, we simply | |
2885 // set _finger to be limit to ensure that the bitmap iteration | |
2886 // doesn't do anything. | |
2887 _finger = limit; | |
2888 } | |
2889 | |
2890 _region_limit = limit; | |
2891 } | |
2892 | |
2893 void CMTask::giveup_current_region() { | |
2894 tmp_guarantee_CM( _curr_region != NULL, "invariant" ); | |
2895 if (_cm->verbose_low()) | |
2896 gclog_or_tty->print_cr("[%d] giving up region "PTR_FORMAT, | |
2897 _task_id, _curr_region); | |
2898 clear_region_fields(); | |
2899 } | |
2900 | |
2901 void CMTask::clear_region_fields() { | |
2902 // Values for these three fields that indicate that we're not | |
2903 // holding on to a region. | |
2904 _curr_region = NULL; | |
2905 _finger = NULL; | |
2906 _region_limit = NULL; | |
2907 | |
2908 _region_finger = NULL; | |
2909 } | |
2910 | |
2911 void CMTask::reset(CMBitMap* nextMarkBitMap) { | |
2912 guarantee( nextMarkBitMap != NULL, "invariant" ); | |
2913 | |
2914 if (_cm->verbose_low()) | |
2915 gclog_or_tty->print_cr("[%d] resetting", _task_id); | |
2916 | |
2917 _nextMarkBitMap = nextMarkBitMap; | |
2918 clear_region_fields(); | |
2919 | |
2920 _calls = 0; | |
2921 _elapsed_time_ms = 0.0; | |
2922 _termination_time_ms = 0.0; | |
2923 _termination_start_time_ms = 0.0; | |
2924 | |
2925 #if _MARKING_STATS_ | |
2926 _local_pushes = 0; | |
2927 _local_pops = 0; | |
2928 _local_max_size = 0; | |
2929 _objs_scanned = 0; | |
2930 _global_pushes = 0; | |
2931 _global_pops = 0; | |
2932 _global_max_size = 0; | |
2933 _global_transfers_to = 0; | |
2934 _global_transfers_from = 0; | |
2935 _region_stack_pops = 0; | |
2936 _regions_claimed = 0; | |
2937 _objs_found_on_bitmap = 0; | |
2938 _satb_buffers_processed = 0; | |
2939 _steal_attempts = 0; | |
2940 _steals = 0; | |
2941 _aborted = 0; | |
2942 _aborted_overflow = 0; | |
2943 _aborted_cm_aborted = 0; | |
2944 _aborted_yield = 0; | |
2945 _aborted_timed_out = 0; | |
2946 _aborted_satb = 0; | |
2947 _aborted_termination = 0; | |
2948 #endif // _MARKING_STATS_ | |
2949 } | |
2950 | |
2951 bool CMTask::should_exit_termination() { | |
2952 regular_clock_call(); | |
2953 // This is called when we are in the termination protocol. We should | |
2954 // quit if, for some reason, this task wants to abort or the global | |
2955 // stack is not empty (this means that we can get work from it). | |
2956 return !_cm->mark_stack_empty() || has_aborted(); | |
2957 } | |
2958 | |
2959 // This determines whether the method below will check both the local | |
2960 // and global fingers when determining whether to push on the stack a | |
2961 // gray object (value 1) or whether it will only check the global one | |
2962 // (value 0). The tradeoffs are that the former will be a bit more | |
2963 // accurate and possibly push less on the stack, but it might also be | |
2964 // a little bit slower. | |
2965 | |
2966 #define _CHECK_BOTH_FINGERS_ 1 | |
2967 | |
2968 void CMTask::deal_with_reference(oop obj) { | |
2969 if (_cm->verbose_high()) | |
2970 gclog_or_tty->print_cr("[%d] we're dealing with reference = "PTR_FORMAT, | |
2971 _task_id, (void*) obj); | |
2972 | |
2973 ++_refs_reached; | |
2974 | |
2975 HeapWord* objAddr = (HeapWord*) obj; | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
2976 assert(obj->is_oop_or_null(true /* ignore mark word */), "Error"); |
342 | 2977 if (_g1h->is_in_g1_reserved(objAddr)) { |
2978 tmp_guarantee_CM( obj != NULL, "is_in_g1_reserved should ensure this" ); | |
2979 HeapRegion* hr = _g1h->heap_region_containing(obj); | |
2980 if (_g1h->is_obj_ill(obj, hr)) { | |
2981 if (_cm->verbose_high()) | |
2982 gclog_or_tty->print_cr("[%d] "PTR_FORMAT" is not considered marked", | |
2983 _task_id, (void*) obj); | |
2984 | |
2985 // we need to mark it first | |
2986 if (_nextMarkBitMap->parMark(objAddr)) { | |
2987 // No OrderAccess:store_load() is needed. It is implicit in the | |
2988 // CAS done in parMark(objAddr) above | |
2989 HeapWord* global_finger = _cm->finger(); | |
2990 | |
2991 #if _CHECK_BOTH_FINGERS_ | |
2992 // we will check both the local and global fingers | |
2993 | |
2994 if (_finger != NULL && objAddr < _finger) { | |
2995 if (_cm->verbose_high()) | |
2996 gclog_or_tty->print_cr("[%d] below the local finger ("PTR_FORMAT"), " | |
2997 "pushing it", _task_id, _finger); | |
2998 push(obj); | |
2999 } else if (_curr_region != NULL && objAddr < _region_limit) { | |
3000 // do nothing | |
3001 } else if (objAddr < global_finger) { | |
3002 // Notice that the global finger might be moving forward | |
3003 // concurrently. This is not a problem. In the worst case, we | |
3004 // mark the object while it is above the global finger and, by | |
3005 // the time we read the global finger, it has moved forward | |
3006 // passed this object. In this case, the object will probably | |
3007 // be visited when a task is scanning the region and will also | |
3008 // be pushed on the stack. So, some duplicate work, but no | |
3009 // correctness problems. | |
3010 | |
3011 if (_cm->verbose_high()) | |
3012 gclog_or_tty->print_cr("[%d] below the global finger " | |
3013 "("PTR_FORMAT"), pushing it", | |
3014 _task_id, global_finger); | |
3015 push(obj); | |
3016 } else { | |
3017 // do nothing | |
3018 } | |
3019 #else // _CHECK_BOTH_FINGERS_ | |
3020 // we will only check the global finger | |
3021 | |
3022 if (objAddr < global_finger) { | |
3023 // see long comment above | |
3024 | |
3025 if (_cm->verbose_high()) | |
3026 gclog_or_tty->print_cr("[%d] below the global finger " | |
3027 "("PTR_FORMAT"), pushing it", | |
3028 _task_id, global_finger); | |
3029 push(obj); | |
3030 } | |
3031 #endif // _CHECK_BOTH_FINGERS_ | |
3032 } | |
3033 } | |
3034 } | |
3035 } | |
3036 | |
3037 void CMTask::push(oop obj) { | |
3038 HeapWord* objAddr = (HeapWord*) obj; | |
3039 tmp_guarantee_CM( _g1h->is_in_g1_reserved(objAddr), "invariant" ); | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
3040 tmp_guarantee_CM( !_g1h->heap_region_containing(objAddr)->is_on_free_list(), "invariant" ); |
342 | 3041 tmp_guarantee_CM( !_g1h->is_obj_ill(obj), "invariant" ); |
3042 tmp_guarantee_CM( _nextMarkBitMap->isMarked(objAddr), "invariant" ); | |
3043 | |
3044 if (_cm->verbose_high()) | |
3045 gclog_or_tty->print_cr("[%d] pushing "PTR_FORMAT, _task_id, (void*) obj); | |
3046 | |
3047 if (!_task_queue->push(obj)) { | |
3048 // The local task queue looks full. We need to push some entries | |
3049 // to the global stack. | |
3050 | |
3051 if (_cm->verbose_medium()) | |
3052 gclog_or_tty->print_cr("[%d] task queue overflow, " | |
3053 "moving entries to the global stack", | |
3054 _task_id); | |
3055 move_entries_to_global_stack(); | |
3056 | |
3057 // this should succeed since, even if we overflow the global | |
3058 // stack, we should have definitely removed some entries from the | |
3059 // local queue. So, there must be space on it. | |
3060 bool success = _task_queue->push(obj); | |
3061 tmp_guarantee_CM( success, "invariant" ); | |
3062 } | |
3063 | |
3064 statsOnly( int tmp_size = _task_queue->size(); | |
3065 if (tmp_size > _local_max_size) | |
3066 _local_max_size = tmp_size; | |
3067 ++_local_pushes ); | |
3068 } | |
3069 | |
3070 void CMTask::reached_limit() { | |
3071 tmp_guarantee_CM( _words_scanned >= _words_scanned_limit || | |
3072 _refs_reached >= _refs_reached_limit , | |
3073 "shouldn't have been called otherwise" ); | |
3074 regular_clock_call(); | |
3075 } | |
3076 | |
3077 void CMTask::regular_clock_call() { | |
3078 if (has_aborted()) | |
3079 return; | |
3080 | |
3081 // First, we need to recalculate the words scanned and refs reached | |
3082 // limits for the next clock call. | |
3083 recalculate_limits(); | |
3084 | |
3085 // During the regular clock call we do the following | |
3086 | |
3087 // (1) If an overflow has been flagged, then we abort. | |
3088 if (_cm->has_overflown()) { | |
3089 set_has_aborted(); | |
3090 return; | |
3091 } | |
3092 | |
3093 // If we are not concurrent (i.e. we're doing remark) we don't need | |
3094 // to check anything else. The other steps are only needed during | |
3095 // the concurrent marking phase. | |
3096 if (!concurrent()) | |
3097 return; | |
3098 | |
3099 // (2) If marking has been aborted for Full GC, then we also abort. | |
3100 if (_cm->has_aborted()) { | |
3101 set_has_aborted(); | |
3102 statsOnly( ++_aborted_cm_aborted ); | |
3103 return; | |
3104 } | |
3105 | |
3106 double curr_time_ms = os::elapsedVTime() * 1000.0; | |
3107 | |
3108 // (3) If marking stats are enabled, then we update the step history. | |
3109 #if _MARKING_STATS_ | |
3110 if (_words_scanned >= _words_scanned_limit) | |
3111 ++_clock_due_to_scanning; | |
3112 if (_refs_reached >= _refs_reached_limit) | |
3113 ++_clock_due_to_marking; | |
3114 | |
3115 double last_interval_ms = curr_time_ms - _interval_start_time_ms; | |
3116 _interval_start_time_ms = curr_time_ms; | |
3117 _all_clock_intervals_ms.add(last_interval_ms); | |
3118 | |
3119 if (_cm->verbose_medium()) { | |
3120 gclog_or_tty->print_cr("[%d] regular clock, interval = %1.2lfms, " | |
3121 "scanned = %d%s, refs reached = %d%s", | |
3122 _task_id, last_interval_ms, | |
3123 _words_scanned, | |
3124 (_words_scanned >= _words_scanned_limit) ? " (*)" : "", | |
3125 _refs_reached, | |
3126 (_refs_reached >= _refs_reached_limit) ? " (*)" : ""); | |
3127 } | |
3128 #endif // _MARKING_STATS_ | |
3129 | |
3130 // (4) We check whether we should yield. If we have to, then we abort. | |
3131 if (_cm->should_yield()) { | |
3132 // We should yield. To do this we abort the task. The caller is | |
3133 // responsible for yielding. | |
3134 set_has_aborted(); | |
3135 statsOnly( ++_aborted_yield ); | |
3136 return; | |
3137 } | |
3138 | |
3139 // (5) We check whether we've reached our time quota. If we have, | |
3140 // then we abort. | |
3141 double elapsed_time_ms = curr_time_ms - _start_time_ms; | |
3142 if (elapsed_time_ms > _time_target_ms) { | |
3143 set_has_aborted(); | |
3144 _has_aborted_timed_out = true; | |
3145 statsOnly( ++_aborted_timed_out ); | |
3146 return; | |
3147 } | |
3148 | |
3149 // (6) Finally, we check whether there are enough completed STAB | |
3150 // buffers available for processing. If there are, we abort. | |
3151 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
3152 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) { | |
3153 if (_cm->verbose_low()) | |
3154 gclog_or_tty->print_cr("[%d] aborting to deal with pending SATB buffers", | |
3155 _task_id); | |
3156 // we do need to process SATB buffers, we'll abort and restart | |
3157 // the marking task to do so | |
3158 set_has_aborted(); | |
3159 statsOnly( ++_aborted_satb ); | |
3160 return; | |
3161 } | |
3162 } | |
3163 | |
3164 void CMTask::recalculate_limits() { | |
3165 _real_words_scanned_limit = _words_scanned + words_scanned_period; | |
3166 _words_scanned_limit = _real_words_scanned_limit; | |
3167 | |
3168 _real_refs_reached_limit = _refs_reached + refs_reached_period; | |
3169 _refs_reached_limit = _real_refs_reached_limit; | |
3170 } | |
3171 | |
3172 void CMTask::decrease_limits() { | |
3173 // This is called when we believe that we're going to do an infrequent | |
3174 // operation which will increase the per byte scanned cost (i.e. move | |
3175 // entries to/from the global stack). It basically tries to decrease the | |
3176 // scanning limit so that the clock is called earlier. | |
3177 | |
3178 if (_cm->verbose_medium()) | |
3179 gclog_or_tty->print_cr("[%d] decreasing limits", _task_id); | |
3180 | |
3181 _words_scanned_limit = _real_words_scanned_limit - | |
3182 3 * words_scanned_period / 4; | |
3183 _refs_reached_limit = _real_refs_reached_limit - | |
3184 3 * refs_reached_period / 4; | |
3185 } | |
3186 | |
3187 void CMTask::move_entries_to_global_stack() { | |
3188 // local array where we'll store the entries that will be popped | |
3189 // from the local queue | |
3190 oop buffer[global_stack_transfer_size]; | |
3191 | |
3192 int n = 0; | |
3193 oop obj; | |
3194 while (n < global_stack_transfer_size && _task_queue->pop_local(obj)) { | |
3195 buffer[n] = obj; | |
3196 ++n; | |
3197 } | |
3198 | |
3199 if (n > 0) { | |
3200 // we popped at least one entry from the local queue | |
3201 | |
3202 statsOnly( ++_global_transfers_to; _local_pops += n ); | |
3203 | |
3204 if (!_cm->mark_stack_push(buffer, n)) { | |
3205 if (_cm->verbose_low()) | |
3206 gclog_or_tty->print_cr("[%d] aborting due to global stack overflow", _task_id); | |
3207 set_has_aborted(); | |
3208 } else { | |
3209 // the transfer was successful | |
3210 | |
3211 if (_cm->verbose_medium()) | |
3212 gclog_or_tty->print_cr("[%d] pushed %d entries to the global stack", | |
3213 _task_id, n); | |
3214 statsOnly( int tmp_size = _cm->mark_stack_size(); | |
3215 if (tmp_size > _global_max_size) | |
3216 _global_max_size = tmp_size; | |
3217 _global_pushes += n ); | |
3218 } | |
3219 } | |
3220 | |
3221 // this operation was quite expensive, so decrease the limits | |
3222 decrease_limits(); | |
3223 } | |
3224 | |
3225 void CMTask::get_entries_from_global_stack() { | |
3226 // local array where we'll store the entries that will be popped | |
3227 // from the global stack. | |
3228 oop buffer[global_stack_transfer_size]; | |
3229 int n; | |
3230 _cm->mark_stack_pop(buffer, global_stack_transfer_size, &n); | |
3231 tmp_guarantee_CM( n <= global_stack_transfer_size, | |
3232 "we should not pop more than the given limit" ); | |
3233 if (n > 0) { | |
3234 // yes, we did actually pop at least one entry | |
3235 | |
3236 statsOnly( ++_global_transfers_from; _global_pops += n ); | |
3237 if (_cm->verbose_medium()) | |
3238 gclog_or_tty->print_cr("[%d] popped %d entries from the global stack", | |
3239 _task_id, n); | |
3240 for (int i = 0; i < n; ++i) { | |
3241 bool success = _task_queue->push(buffer[i]); | |
3242 // We only call this when the local queue is empty or under a | |
3243 // given target limit. So, we do not expect this push to fail. | |
3244 tmp_guarantee_CM( success, "invariant" ); | |
3245 } | |
3246 | |
3247 statsOnly( int tmp_size = _task_queue->size(); | |
3248 if (tmp_size > _local_max_size) | |
3249 _local_max_size = tmp_size; | |
3250 _local_pushes += n ); | |
3251 } | |
3252 | |
3253 // this operation was quite expensive, so decrease the limits | |
3254 decrease_limits(); | |
3255 } | |
3256 | |
3257 void CMTask::drain_local_queue(bool partially) { | |
3258 if (has_aborted()) | |
3259 return; | |
3260 | |
3261 // Decide what the target size is, depending whether we're going to | |
3262 // drain it partially (so that other tasks can steal if they run out | |
3263 // of things to do) or totally (at the very end). | |
3264 size_t target_size; | |
3265 if (partially) | |
3266 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize); | |
3267 else | |
3268 target_size = 0; | |
3269 | |
3270 if (_task_queue->size() > target_size) { | |
3271 if (_cm->verbose_high()) | |
3272 gclog_or_tty->print_cr("[%d] draining local queue, target size = %d", | |
3273 _task_id, target_size); | |
3274 | |
3275 oop obj; | |
3276 bool ret = _task_queue->pop_local(obj); | |
3277 while (ret) { | |
3278 statsOnly( ++_local_pops ); | |
3279 | |
3280 if (_cm->verbose_high()) | |
3281 gclog_or_tty->print_cr("[%d] popped "PTR_FORMAT, _task_id, | |
3282 (void*) obj); | |
3283 | |
3284 tmp_guarantee_CM( _g1h->is_in_g1_reserved((HeapWord*) obj), | |
3285 "invariant" ); | |
845
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
3286 tmp_guarantee_CM( !_g1h->heap_region_containing(obj)->is_on_free_list(), |
df6caf649ff7
6700789: G1: Enable use of compressed oops with G1 heaps
ysr
parents:
831
diff
changeset
|
3287 "invariant" ); |
342 | 3288 |
3289 scan_object(obj); | |
3290 | |
3291 if (_task_queue->size() <= target_size || has_aborted()) | |
3292 ret = false; | |
3293 else | |
3294 ret = _task_queue->pop_local(obj); | |
3295 } | |
3296 | |
3297 if (_cm->verbose_high()) | |
3298 gclog_or_tty->print_cr("[%d] drained local queue, size = %d", | |
3299 _task_id, _task_queue->size()); | |
3300 } | |
3301 } | |
3302 | |
3303 void CMTask::drain_global_stack(bool partially) { | |
3304 if (has_aborted()) | |
3305 return; | |
3306 | |
3307 // We have a policy to drain the local queue before we attempt to | |
3308 // drain the global stack. | |
3309 tmp_guarantee_CM( partially || _task_queue->size() == 0, "invariant" ); | |
3310 | |
3311 // Decide what the target size is, depending whether we're going to | |
3312 // drain it partially (so that other tasks can steal if they run out | |
3313 // of things to do) or totally (at the very end). Notice that, | |
3314 // because we move entries from the global stack in chunks or | |
3315 // because another task might be doing the same, we might in fact | |
3316 // drop below the target. But, this is not a problem. | |
3317 size_t target_size; | |
3318 if (partially) | |
3319 target_size = _cm->partial_mark_stack_size_target(); | |
3320 else | |
3321 target_size = 0; | |
3322 | |
3323 if (_cm->mark_stack_size() > target_size) { | |
3324 if (_cm->verbose_low()) | |
3325 gclog_or_tty->print_cr("[%d] draining global_stack, target size %d", | |
3326 _task_id, target_size); | |
3327 | |
3328 while (!has_aborted() && _cm->mark_stack_size() > target_size) { | |
3329 get_entries_from_global_stack(); | |
3330 drain_local_queue(partially); | |
3331 } | |
3332 | |
3333 if (_cm->verbose_low()) | |
3334 gclog_or_tty->print_cr("[%d] drained global stack, size = %d", | |
3335 _task_id, _cm->mark_stack_size()); | |
3336 } | |
3337 } | |
3338 | |
3339 // SATB Queue has several assumptions on whether to call the par or | |
3340 // non-par versions of the methods. this is why some of the code is | |
3341 // replicated. We should really get rid of the single-threaded version | |
3342 // of the code to simplify things. | |
3343 void CMTask::drain_satb_buffers() { | |
3344 if (has_aborted()) | |
3345 return; | |
3346 | |
3347 // We set this so that the regular clock knows that we're in the | |
3348 // middle of draining buffers and doesn't set the abort flag when it | |
3349 // notices that SATB buffers are available for draining. It'd be | |
3350 // very counter productive if it did that. :-) | |
3351 _draining_satb_buffers = true; | |
3352 | |
3353 CMObjectClosure oc(this); | |
3354 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
3355 if (ParallelGCThreads > 0) | |
3356 satb_mq_set.set_par_closure(_task_id, &oc); | |
3357 else | |
3358 satb_mq_set.set_closure(&oc); | |
3359 | |
3360 // This keeps claiming and applying the closure to completed buffers | |
3361 // until we run out of buffers or we need to abort. | |
3362 if (ParallelGCThreads > 0) { | |
3363 while (!has_aborted() && | |
3364 satb_mq_set.par_apply_closure_to_completed_buffer(_task_id)) { | |
3365 if (_cm->verbose_medium()) | |
3366 gclog_or_tty->print_cr("[%d] processed an SATB buffer", _task_id); | |
3367 statsOnly( ++_satb_buffers_processed ); | |
3368 regular_clock_call(); | |
3369 } | |
3370 } else { | |
3371 while (!has_aborted() && | |
3372 satb_mq_set.apply_closure_to_completed_buffer()) { | |
3373 if (_cm->verbose_medium()) | |
3374 gclog_or_tty->print_cr("[%d] processed an SATB buffer", _task_id); | |
3375 statsOnly( ++_satb_buffers_processed ); | |
3376 regular_clock_call(); | |
3377 } | |
3378 } | |
3379 | |
3380 if (!concurrent() && !has_aborted()) { | |
3381 // We should only do this during remark. | |
3382 if (ParallelGCThreads > 0) | |
3383 satb_mq_set.par_iterate_closure_all_threads(_task_id); | |
3384 else | |
3385 satb_mq_set.iterate_closure_all_threads(); | |
3386 } | |
3387 | |
3388 _draining_satb_buffers = false; | |
3389 | |
3390 tmp_guarantee_CM( has_aborted() || | |
3391 concurrent() || | |
3392 satb_mq_set.completed_buffers_num() == 0, "invariant" ); | |
3393 | |
3394 if (ParallelGCThreads > 0) | |
3395 satb_mq_set.set_par_closure(_task_id, NULL); | |
3396 else | |
3397 satb_mq_set.set_closure(NULL); | |
3398 | |
3399 // again, this was a potentially expensive operation, decrease the | |
3400 // limits to get the regular clock call early | |
3401 decrease_limits(); | |
3402 } | |
3403 | |
3404 void CMTask::drain_region_stack(BitMapClosure* bc) { | |
3405 if (has_aborted()) | |
3406 return; | |
3407 | |
3408 tmp_guarantee_CM( _region_finger == NULL, | |
3409 "it should be NULL when we're not scanning a region" ); | |
3410 | |
3411 if (!_cm->region_stack_empty()) { | |
3412 if (_cm->verbose_low()) | |
3413 gclog_or_tty->print_cr("[%d] draining region stack, size = %d", | |
3414 _task_id, _cm->region_stack_size()); | |
3415 | |
3416 MemRegion mr = _cm->region_stack_pop(); | |
3417 // it returns MemRegion() if the pop fails | |
3418 statsOnly(if (mr.start() != NULL) ++_region_stack_pops ); | |
3419 | |
3420 while (mr.start() != NULL) { | |
3421 if (_cm->verbose_medium()) | |
3422 gclog_or_tty->print_cr("[%d] we are scanning region " | |
3423 "["PTR_FORMAT", "PTR_FORMAT")", | |
3424 _task_id, mr.start(), mr.end()); | |
3425 tmp_guarantee_CM( mr.end() <= _cm->finger(), | |
3426 "otherwise the region shouldn't be on the stack" ); | |
3427 assert(!mr.is_empty(), "Only non-empty regions live on the region stack"); | |
3428 if (_nextMarkBitMap->iterate(bc, mr)) { | |
3429 tmp_guarantee_CM( !has_aborted(), | |
3430 "cannot abort the task without aborting the bitmap iteration" ); | |
3431 | |
3432 // We finished iterating over the region without aborting. | |
3433 regular_clock_call(); | |
3434 if (has_aborted()) | |
3435 mr = MemRegion(); | |
3436 else { | |
3437 mr = _cm->region_stack_pop(); | |
3438 // it returns MemRegion() if the pop fails | |
3439 statsOnly(if (mr.start() != NULL) ++_region_stack_pops ); | |
3440 } | |
3441 } else { | |
3442 guarantee( has_aborted(), "currently the only way to do so" ); | |
3443 | |
3444 // The only way to abort the bitmap iteration is to return | |
3445 // false from the do_bit() method. However, inside the | |
3446 // do_bit() method we move the _region_finger to point to the | |
3447 // object currently being looked at. So, if we bail out, we | |
3448 // have definitely set _region_finger to something non-null. | |
3449 guarantee( _region_finger != NULL, "invariant" ); | |
3450 | |
3451 // The iteration was actually aborted. So now _region_finger | |
3452 // points to the address of the object we last scanned. If we | |
3453 // leave it there, when we restart this task, we will rescan | |
3454 // the object. It is easy to avoid this. We move the finger by | |
3455 // enough to point to the next possible object header (the | |
3456 // bitmap knows by how much we need to move it as it knows its | |
3457 // granularity). | |
3458 MemRegion newRegion = | |
3459 MemRegion(_nextMarkBitMap->nextWord(_region_finger), mr.end()); | |
3460 | |
3461 if (!newRegion.is_empty()) { | |
3462 if (_cm->verbose_low()) { | |
3463 gclog_or_tty->print_cr("[%d] pushing unscanned region" | |
3464 "[" PTR_FORMAT "," PTR_FORMAT ") on region stack", | |
3465 _task_id, | |
3466 newRegion.start(), newRegion.end()); | |
3467 } | |
3468 // Now push the part of the region we didn't scan on the | |
3469 // region stack to make sure a task scans it later. | |
3470 _cm->region_stack_push(newRegion); | |
3471 } | |
3472 // break from while | |
3473 mr = MemRegion(); | |
3474 } | |
3475 _region_finger = NULL; | |
3476 } | |
3477 | |
3478 // We only push regions on the region stack during evacuation | |
3479 // pauses. So if we come out the above iteration because we region | |
3480 // stack is empty, it will remain empty until the next yield | |
3481 // point. So, the guarantee below is safe. | |
3482 guarantee( has_aborted() || _cm->region_stack_empty(), | |
3483 "only way to exit the loop" ); | |
3484 | |
3485 if (_cm->verbose_low()) | |
3486 gclog_or_tty->print_cr("[%d] drained region stack, size = %d", | |
3487 _task_id, _cm->region_stack_size()); | |
3488 } | |
3489 } | |
3490 | |
3491 void CMTask::print_stats() { | |
3492 gclog_or_tty->print_cr("Marking Stats, task = %d, calls = %d", | |
3493 _task_id, _calls); | |
3494 gclog_or_tty->print_cr(" Elapsed time = %1.2lfms, Termination time = %1.2lfms", | |
3495 _elapsed_time_ms, _termination_time_ms); | |
3496 gclog_or_tty->print_cr(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms", | |
3497 _step_times_ms.num(), _step_times_ms.avg(), | |
3498 _step_times_ms.sd()); | |
3499 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms", | |
3500 _step_times_ms.maximum(), _step_times_ms.sum()); | |
3501 | |
3502 #if _MARKING_STATS_ | |
3503 gclog_or_tty->print_cr(" Clock Intervals (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms", | |
3504 _all_clock_intervals_ms.num(), _all_clock_intervals_ms.avg(), | |
3505 _all_clock_intervals_ms.sd()); | |
3506 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms", | |
3507 _all_clock_intervals_ms.maximum(), | |
3508 _all_clock_intervals_ms.sum()); | |
3509 gclog_or_tty->print_cr(" Clock Causes (cum): scanning = %d, marking = %d", | |
3510 _clock_due_to_scanning, _clock_due_to_marking); | |
3511 gclog_or_tty->print_cr(" Objects: scanned = %d, found on the bitmap = %d", | |
3512 _objs_scanned, _objs_found_on_bitmap); | |
3513 gclog_or_tty->print_cr(" Local Queue: pushes = %d, pops = %d, max size = %d", | |
3514 _local_pushes, _local_pops, _local_max_size); | |
3515 gclog_or_tty->print_cr(" Global Stack: pushes = %d, pops = %d, max size = %d", | |
3516 _global_pushes, _global_pops, _global_max_size); | |
3517 gclog_or_tty->print_cr(" transfers to = %d, transfers from = %d", | |
3518 _global_transfers_to,_global_transfers_from); | |
3519 gclog_or_tty->print_cr(" Regions: claimed = %d, Region Stack: pops = %d", | |
3520 _regions_claimed, _region_stack_pops); | |
3521 gclog_or_tty->print_cr(" SATB buffers: processed = %d", _satb_buffers_processed); | |
3522 gclog_or_tty->print_cr(" Steals: attempts = %d, successes = %d", | |
3523 _steal_attempts, _steals); | |
3524 gclog_or_tty->print_cr(" Aborted: %d, due to", _aborted); | |
3525 gclog_or_tty->print_cr(" overflow: %d, global abort: %d, yield: %d", | |
3526 _aborted_overflow, _aborted_cm_aborted, _aborted_yield); | |
3527 gclog_or_tty->print_cr(" time out: %d, SATB: %d, termination: %d", | |
3528 _aborted_timed_out, _aborted_satb, _aborted_termination); | |
3529 #endif // _MARKING_STATS_ | |
3530 } | |
3531 | |
3532 /***************************************************************************** | |
3533 | |
3534 The do_marking_step(time_target_ms) method is the building block | |
3535 of the parallel marking framework. It can be called in parallel | |
3536 with other invocations of do_marking_step() on different tasks | |
3537 (but only one per task, obviously) and concurrently with the | |
3538 mutator threads, or during remark, hence it eliminates the need | |
3539 for two versions of the code. When called during remark, it will | |
3540 pick up from where the task left off during the concurrent marking | |
3541 phase. Interestingly, tasks are also claimable during evacuation | |
3542 pauses too, since do_marking_step() ensures that it aborts before | |
3543 it needs to yield. | |
3544 | |
3545 The data structures that is uses to do marking work are the | |
3546 following: | |
3547 | |
3548 (1) Marking Bitmap. If there are gray objects that appear only | |
3549 on the bitmap (this happens either when dealing with an overflow | |
3550 or when the initial marking phase has simply marked the roots | |
3551 and didn't push them on the stack), then tasks claim heap | |
3552 regions whose bitmap they then scan to find gray objects. A | |
3553 global finger indicates where the end of the last claimed region | |
3554 is. A local finger indicates how far into the region a task has | |
3555 scanned. The two fingers are used to determine how to gray an | |
3556 object (i.e. whether simply marking it is OK, as it will be | |
3557 visited by a task in the future, or whether it needs to be also | |
3558 pushed on a stack). | |
3559 | |
3560 (2) Local Queue. The local queue of the task which is accessed | |
3561 reasonably efficiently by the task. Other tasks can steal from | |
3562 it when they run out of work. Throughout the marking phase, a | |
3563 task attempts to keep its local queue short but not totally | |
3564 empty, so that entries are available for stealing by other | |
3565 tasks. Only when there is no more work, a task will totally | |
3566 drain its local queue. | |
3567 | |
3568 (3) Global Mark Stack. This handles local queue overflow. During | |
3569 marking only sets of entries are moved between it and the local | |
3570 queues, as access to it requires a mutex and more fine-grain | |
3571 interaction with it which might cause contention. If it | |
3572 overflows, then the marking phase should restart and iterate | |
3573 over the bitmap to identify gray objects. Throughout the marking | |
3574 phase, tasks attempt to keep the global mark stack at a small | |
3575 length but not totally empty, so that entries are available for | |
3576 popping by other tasks. Only when there is no more work, tasks | |
3577 will totally drain the global mark stack. | |
3578 | |
3579 (4) Global Region Stack. Entries on it correspond to areas of | |
3580 the bitmap that need to be scanned since they contain gray | |
3581 objects. Pushes on the region stack only happen during | |
3582 evacuation pauses and typically correspond to areas covered by | |
3583 GC LABS. If it overflows, then the marking phase should restart | |
3584 and iterate over the bitmap to identify gray objects. Tasks will | |
3585 try to totally drain the region stack as soon as possible. | |
3586 | |
3587 (5) SATB Buffer Queue. This is where completed SATB buffers are | |
3588 made available. Buffers are regularly removed from this queue | |
3589 and scanned for roots, so that the queue doesn't get too | |
3590 long. During remark, all completed buffers are processed, as | |
3591 well as the filled in parts of any uncompleted buffers. | |
3592 | |
3593 The do_marking_step() method tries to abort when the time target | |
3594 has been reached. There are a few other cases when the | |
3595 do_marking_step() method also aborts: | |
3596 | |
3597 (1) When the marking phase has been aborted (after a Full GC). | |
3598 | |
3599 (2) When a global overflow (either on the global stack or the | |
3600 region stack) has been triggered. Before the task aborts, it | |
3601 will actually sync up with the other tasks to ensure that all | |
3602 the marking data structures (local queues, stacks, fingers etc.) | |
3603 are re-initialised so that when do_marking_step() completes, | |
3604 the marking phase can immediately restart. | |
3605 | |
3606 (3) When enough completed SATB buffers are available. The | |
3607 do_marking_step() method only tries to drain SATB buffers right | |
3608 at the beginning. So, if enough buffers are available, the | |
3609 marking step aborts and the SATB buffers are processed at | |
3610 the beginning of the next invocation. | |
3611 | |
3612 (4) To yield. when we have to yield then we abort and yield | |
3613 right at the end of do_marking_step(). This saves us from a lot | |
3614 of hassle as, by yielding we might allow a Full GC. If this | |
3615 happens then objects will be compacted underneath our feet, the | |
3616 heap might shrink, etc. We save checking for this by just | |
3617 aborting and doing the yield right at the end. | |
3618 | |
3619 From the above it follows that the do_marking_step() method should | |
3620 be called in a loop (or, otherwise, regularly) until it completes. | |
3621 | |
3622 If a marking step completes without its has_aborted() flag being | |
3623 true, it means it has completed the current marking phase (and | |
3624 also all other marking tasks have done so and have all synced up). | |
3625 | |
3626 A method called regular_clock_call() is invoked "regularly" (in | |
3627 sub ms intervals) throughout marking. It is this clock method that | |
3628 checks all the abort conditions which were mentioned above and | |
3629 decides when the task should abort. A work-based scheme is used to | |
3630 trigger this clock method: when the number of object words the | |
3631 marking phase has scanned or the number of references the marking | |
3632 phase has visited reach a given limit. Additional invocations to | |
3633 the method clock have been planted in a few other strategic places | |
3634 too. The initial reason for the clock method was to avoid calling | |
3635 vtime too regularly, as it is quite expensive. So, once it was in | |
3636 place, it was natural to piggy-back all the other conditions on it | |
3637 too and not constantly check them throughout the code. | |
3638 | |
3639 *****************************************************************************/ | |
3640 | |
3641 void CMTask::do_marking_step(double time_target_ms) { | |
3642 guarantee( time_target_ms >= 1.0, "minimum granularity is 1ms" ); | |
3643 guarantee( concurrent() == _cm->concurrent(), "they should be the same" ); | |
3644 | |
3645 guarantee( concurrent() || _cm->region_stack_empty(), | |
3646 "the region stack should have been cleared before remark" ); | |
3647 guarantee( _region_finger == NULL, | |
3648 "this should be non-null only when a region is being scanned" ); | |
3649 | |
3650 G1CollectorPolicy* g1_policy = _g1h->g1_policy(); | |
3651 guarantee( _task_queues != NULL, "invariant" ); | |
3652 guarantee( _task_queue != NULL, "invariant" ); | |
3653 guarantee( _task_queues->queue(_task_id) == _task_queue, "invariant" ); | |
3654 | |
3655 guarantee( !_claimed, | |
3656 "only one thread should claim this task at any one time" ); | |
3657 | |
3658 // OK, this doesn't safeguard again all possible scenarios, as it is | |
3659 // possible for two threads to set the _claimed flag at the same | |
3660 // time. But it is only for debugging purposes anyway and it will | |
3661 // catch most problems. | |
3662 _claimed = true; | |
3663 | |
3664 _start_time_ms = os::elapsedVTime() * 1000.0; | |
3665 statsOnly( _interval_start_time_ms = _start_time_ms ); | |
3666 | |
3667 double diff_prediction_ms = | |
3668 g1_policy->get_new_prediction(&_marking_step_diffs_ms); | |
3669 _time_target_ms = time_target_ms - diff_prediction_ms; | |
3670 | |
3671 // set up the variables that are used in the work-based scheme to | |
3672 // call the regular clock method | |
3673 _words_scanned = 0; | |
3674 _refs_reached = 0; | |
3675 recalculate_limits(); | |
3676 | |
3677 // clear all flags | |
3678 clear_has_aborted(); | |
3679 _has_aborted_timed_out = false; | |
3680 _draining_satb_buffers = false; | |
3681 | |
3682 ++_calls; | |
3683 | |
3684 if (_cm->verbose_low()) | |
3685 gclog_or_tty->print_cr("[%d] >>>>>>>>>> START, call = %d, " | |
3686 "target = %1.2lfms >>>>>>>>>>", | |
3687 _task_id, _calls, _time_target_ms); | |
3688 | |
3689 // Set up the bitmap and oop closures. Anything that uses them is | |
3690 // eventually called from this method, so it is OK to allocate these | |
3691 // statically. | |
3692 CMBitMapClosure bitmap_closure(this, _cm, _nextMarkBitMap); | |
3693 CMOopClosure oop_closure(_g1h, _cm, this); | |
3694 set_oop_closure(&oop_closure); | |
3695 | |
3696 if (_cm->has_overflown()) { | |
3697 // This can happen if the region stack or the mark stack overflows | |
3698 // during a GC pause and this task, after a yield point, | |
3699 // restarts. We have to abort as we need to get into the overflow | |
3700 // protocol which happens right at the end of this task. | |
3701 set_has_aborted(); | |
3702 } | |
3703 | |
3704 // First drain any available SATB buffers. After this, we will not | |
3705 // look at SATB buffers before the next invocation of this method. | |
3706 // If enough completed SATB buffers are queued up, the regular clock | |
3707 // will abort this task so that it restarts. | |
3708 drain_satb_buffers(); | |
3709 // ...then partially drain the local queue and the global stack | |
3710 drain_local_queue(true); | |
3711 drain_global_stack(true); | |
3712 | |
3713 // Then totally drain the region stack. We will not look at | |
3714 // it again before the next invocation of this method. Entries on | |
3715 // the region stack are only added during evacuation pauses, for | |
3716 // which we have to yield. When we do, we abort the task anyway so | |
3717 // it will look at the region stack again when it restarts. | |
3718 bitmap_closure.set_scanning_heap_region(false); | |
3719 drain_region_stack(&bitmap_closure); | |
3720 // ...then partially drain the local queue and the global stack | |
3721 drain_local_queue(true); | |
3722 drain_global_stack(true); | |
3723 | |
3724 do { | |
3725 if (!has_aborted() && _curr_region != NULL) { | |
3726 // This means that we're already holding on to a region. | |
3727 tmp_guarantee_CM( _finger != NULL, | |
3728 "if region is not NULL, then the finger " | |
3729 "should not be NULL either" ); | |
3730 | |
3731 // We might have restarted this task after an evacuation pause | |
3732 // which might have evacuated the region we're holding on to | |
3733 // underneath our feet. Let's read its limit again to make sure | |
3734 // that we do not iterate over a region of the heap that | |
3735 // contains garbage (update_region_limit() will also move | |
3736 // _finger to the start of the region if it is found empty). | |
3737 update_region_limit(); | |
3738 // We will start from _finger not from the start of the region, | |
3739 // as we might be restarting this task after aborting half-way | |
3740 // through scanning this region. In this case, _finger points to | |
3741 // the address where we last found a marked object. If this is a | |
3742 // fresh region, _finger points to start(). | |
3743 MemRegion mr = MemRegion(_finger, _region_limit); | |
3744 | |
3745 if (_cm->verbose_low()) | |
3746 gclog_or_tty->print_cr("[%d] we're scanning part " | |
3747 "["PTR_FORMAT", "PTR_FORMAT") " | |
3748 "of region "PTR_FORMAT, | |
3749 _task_id, _finger, _region_limit, _curr_region); | |
3750 | |
3751 // Let's iterate over the bitmap of the part of the | |
3752 // region that is left. | |
3753 bitmap_closure.set_scanning_heap_region(true); | |
3754 if (mr.is_empty() || | |
3755 _nextMarkBitMap->iterate(&bitmap_closure, mr)) { | |
3756 // We successfully completed iterating over the region. Now, | |
3757 // let's give up the region. | |
3758 giveup_current_region(); | |
3759 regular_clock_call(); | |
3760 } else { | |
3761 guarantee( has_aborted(), "currently the only way to do so" ); | |
3762 // The only way to abort the bitmap iteration is to return | |
3763 // false from the do_bit() method. However, inside the | |
3764 // do_bit() method we move the _finger to point to the | |
3765 // object currently being looked at. So, if we bail out, we | |
3766 // have definitely set _finger to something non-null. | |
3767 guarantee( _finger != NULL, "invariant" ); | |
3768 | |
3769 // Region iteration was actually aborted. So now _finger | |
3770 // points to the address of the object we last scanned. If we | |
3771 // leave it there, when we restart this task, we will rescan | |
3772 // the object. It is easy to avoid this. We move the finger by | |
3773 // enough to point to the next possible object header (the | |
3774 // bitmap knows by how much we need to move it as it knows its | |
3775 // granularity). | |
3776 move_finger_to(_nextMarkBitMap->nextWord(_finger)); | |
3777 } | |
3778 } | |
3779 // At this point we have either completed iterating over the | |
3780 // region we were holding on to, or we have aborted. | |
3781 | |
3782 // We then partially drain the local queue and the global stack. | |
3783 // (Do we really need this?) | |
3784 drain_local_queue(true); | |
3785 drain_global_stack(true); | |
3786 | |
3787 // Read the note on the claim_region() method on why it might | |
3788 // return NULL with potentially more regions available for | |
3789 // claiming and why we have to check out_of_regions() to determine | |
3790 // whether we're done or not. | |
3791 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) { | |
3792 // We are going to try to claim a new region. We should have | |
3793 // given up on the previous one. | |
3794 tmp_guarantee_CM( _curr_region == NULL && | |
3795 _finger == NULL && | |
3796 _region_limit == NULL, "invariant" ); | |
3797 if (_cm->verbose_low()) | |
3798 gclog_or_tty->print_cr("[%d] trying to claim a new region", _task_id); | |
3799 HeapRegion* claimed_region = _cm->claim_region(_task_id); | |
3800 if (claimed_region != NULL) { | |
3801 // Yes, we managed to claim one | |
3802 statsOnly( ++_regions_claimed ); | |
3803 | |
3804 if (_cm->verbose_low()) | |
3805 gclog_or_tty->print_cr("[%d] we successfully claimed " | |
3806 "region "PTR_FORMAT, | |
3807 _task_id, claimed_region); | |
3808 | |
3809 setup_for_region(claimed_region); | |
3810 tmp_guarantee_CM( _curr_region == claimed_region, "invariant" ); | |
3811 } | |
3812 // It is important to call the regular clock here. It might take | |
3813 // a while to claim a region if, for example, we hit a large | |
3814 // block of empty regions. So we need to call the regular clock | |
3815 // method once round the loop to make sure it's called | |
3816 // frequently enough. | |
3817 regular_clock_call(); | |
3818 } | |
3819 | |
3820 if (!has_aborted() && _curr_region == NULL) { | |
3821 tmp_guarantee_CM( _cm->out_of_regions(), | |
3822 "at this point we should be out of regions" ); | |
3823 } | |
3824 } while ( _curr_region != NULL && !has_aborted()); | |
3825 | |
3826 if (!has_aborted()) { | |
3827 // We cannot check whether the global stack is empty, since other | |
343
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3828 // tasks might be pushing objects to it concurrently. We also cannot |
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3829 // check if the region stack is empty because if a thread is aborting |
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3830 // it can push a partially done region back. |
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3831 tmp_guarantee_CM( _cm->out_of_regions(), |
342 | 3832 "at this point we should be out of regions" ); |
3833 | |
3834 if (_cm->verbose_low()) | |
3835 gclog_or_tty->print_cr("[%d] all regions claimed", _task_id); | |
3836 | |
3837 // Try to reduce the number of available SATB buffers so that | |
3838 // remark has less work to do. | |
3839 drain_satb_buffers(); | |
3840 } | |
3841 | |
3842 // Since we've done everything else, we can now totally drain the | |
3843 // local queue and global stack. | |
3844 drain_local_queue(false); | |
3845 drain_global_stack(false); | |
3846 | |
3847 // Attempt at work stealing from other task's queues. | |
3848 if (!has_aborted()) { | |
3849 // We have not aborted. This means that we have finished all that | |
3850 // we could. Let's try to do some stealing... | |
3851 | |
3852 // We cannot check whether the global stack is empty, since other | |
343
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3853 // tasks might be pushing objects to it concurrently. We also cannot |
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3854 // check if the region stack is empty because if a thread is aborting |
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3855 // it can push a partially done region back. |
342 | 3856 guarantee( _cm->out_of_regions() && |
3857 _task_queue->size() == 0, "only way to reach here" ); | |
3858 | |
3859 if (_cm->verbose_low()) | |
3860 gclog_or_tty->print_cr("[%d] starting to steal", _task_id); | |
3861 | |
3862 while (!has_aborted()) { | |
3863 oop obj; | |
3864 statsOnly( ++_steal_attempts ); | |
3865 | |
3866 if (_cm->try_stealing(_task_id, &_hash_seed, obj)) { | |
3867 if (_cm->verbose_medium()) | |
3868 gclog_or_tty->print_cr("[%d] stolen "PTR_FORMAT" successfully", | |
3869 _task_id, (void*) obj); | |
3870 | |
3871 statsOnly( ++_steals ); | |
3872 | |
3873 tmp_guarantee_CM( _nextMarkBitMap->isMarked((HeapWord*) obj), | |
3874 "any stolen object should be marked" ); | |
3875 scan_object(obj); | |
3876 | |
3877 // And since we're towards the end, let's totally drain the | |
3878 // local queue and global stack. | |
3879 drain_local_queue(false); | |
3880 drain_global_stack(false); | |
3881 } else { | |
3882 break; | |
3883 } | |
3884 } | |
3885 } | |
3886 | |
3887 // We still haven't aborted. Now, let's try to get into the | |
3888 // termination protocol. | |
3889 if (!has_aborted()) { | |
3890 // We cannot check whether the global stack is empty, since other | |
343
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3891 // tasks might be concurrently pushing objects on it. We also cannot |
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3892 // check if the region stack is empty because if a thread is aborting |
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3893 // it can push a partially done region back. |
342 | 3894 guarantee( _cm->out_of_regions() && |
3895 _task_queue->size() == 0, "only way to reach here" ); | |
3896 | |
3897 if (_cm->verbose_low()) | |
3898 gclog_or_tty->print_cr("[%d] starting termination protocol", _task_id); | |
3899 | |
3900 _termination_start_time_ms = os::elapsedVTime() * 1000.0; | |
3901 // The CMTask class also extends the TerminatorTerminator class, | |
3902 // hence its should_exit_termination() method will also decide | |
3903 // whether to exit the termination protocol or not. | |
3904 bool finished = _cm->terminator()->offer_termination(this); | |
3905 double termination_end_time_ms = os::elapsedVTime() * 1000.0; | |
3906 _termination_time_ms += | |
3907 termination_end_time_ms - _termination_start_time_ms; | |
3908 | |
3909 if (finished) { | |
3910 // We're all done. | |
3911 | |
3912 if (_task_id == 0) { | |
3913 // let's allow task 0 to do this | |
3914 if (concurrent()) { | |
3915 guarantee( _cm->concurrent_marking_in_progress(), "invariant" ); | |
3916 // we need to set this to false before the next | |
3917 // safepoint. This way we ensure that the marking phase | |
3918 // doesn't observe any more heap expansions. | |
3919 _cm->clear_concurrent_marking_in_progress(); | |
3920 } | |
3921 } | |
3922 | |
3923 // We can now guarantee that the global stack is empty, since | |
3924 // all other tasks have finished. | |
3925 guarantee( _cm->out_of_regions() && | |
3926 _cm->region_stack_empty() && | |
3927 _cm->mark_stack_empty() && | |
3928 _task_queue->size() == 0 && | |
3929 !_cm->has_overflown() && | |
3930 !_cm->mark_stack_overflow() && | |
3931 !_cm->region_stack_overflow(), | |
3932 "only way to reach here" ); | |
3933 | |
3934 if (_cm->verbose_low()) | |
3935 gclog_or_tty->print_cr("[%d] all tasks terminated", _task_id); | |
3936 } else { | |
3937 // Apparently there's more work to do. Let's abort this task. It | |
3938 // will restart it and we can hopefully find more things to do. | |
3939 | |
3940 if (_cm->verbose_low()) | |
3941 gclog_or_tty->print_cr("[%d] apparently there is more work to do", _task_id); | |
3942 | |
3943 set_has_aborted(); | |
3944 statsOnly( ++_aborted_termination ); | |
3945 } | |
3946 } | |
3947 | |
3948 // Mainly for debugging purposes to make sure that a pointer to the | |
3949 // closure which was statically allocated in this frame doesn't | |
3950 // escape it by accident. | |
3951 set_oop_closure(NULL); | |
3952 double end_time_ms = os::elapsedVTime() * 1000.0; | |
3953 double elapsed_time_ms = end_time_ms - _start_time_ms; | |
3954 // Update the step history. | |
3955 _step_times_ms.add(elapsed_time_ms); | |
3956 | |
3957 if (has_aborted()) { | |
3958 // The task was aborted for some reason. | |
3959 | |
3960 statsOnly( ++_aborted ); | |
3961 | |
3962 if (_has_aborted_timed_out) { | |
3963 double diff_ms = elapsed_time_ms - _time_target_ms; | |
3964 // Keep statistics of how well we did with respect to hitting | |
3965 // our target only if we actually timed out (if we aborted for | |
3966 // other reasons, then the results might get skewed). | |
3967 _marking_step_diffs_ms.add(diff_ms); | |
3968 } | |
3969 | |
3970 if (_cm->has_overflown()) { | |
3971 // This is the interesting one. We aborted because a global | |
3972 // overflow was raised. This means we have to restart the | |
3973 // marking phase and start iterating over regions. However, in | |
3974 // order to do this we have to make sure that all tasks stop | |
3975 // what they are doing and re-initialise in a safe manner. We | |
3976 // will achieve this with the use of two barrier sync points. | |
3977 | |
3978 if (_cm->verbose_low()) | |
3979 gclog_or_tty->print_cr("[%d] detected overflow", _task_id); | |
3980 | |
3981 _cm->enter_first_sync_barrier(_task_id); | |
3982 // When we exit this sync barrier we know that all tasks have | |
3983 // stopped doing marking work. So, it's now safe to | |
3984 // re-initialise our data structures. At the end of this method, | |
3985 // task 0 will clear the global data structures. | |
3986 | |
3987 statsOnly( ++_aborted_overflow ); | |
3988 | |
3989 // We clear the local state of this task... | |
3990 clear_region_fields(); | |
3991 | |
3992 // ...and enter the second barrier. | |
3993 _cm->enter_second_sync_barrier(_task_id); | |
3994 // At this point everything has bee re-initialised and we're | |
3995 // ready to restart. | |
3996 } | |
3997 | |
3998 if (_cm->verbose_low()) { | |
3999 gclog_or_tty->print_cr("[%d] <<<<<<<<<< ABORTING, target = %1.2lfms, " | |
4000 "elapsed = %1.2lfms <<<<<<<<<<", | |
4001 _task_id, _time_target_ms, elapsed_time_ms); | |
4002 if (_cm->has_aborted()) | |
4003 gclog_or_tty->print_cr("[%d] ========== MARKING ABORTED ==========", | |
4004 _task_id); | |
4005 } | |
4006 } else { | |
4007 if (_cm->verbose_low()) | |
4008 gclog_or_tty->print_cr("[%d] <<<<<<<<<< FINISHED, target = %1.2lfms, " | |
4009 "elapsed = %1.2lfms <<<<<<<<<<", | |
4010 _task_id, _time_target_ms, elapsed_time_ms); | |
4011 } | |
4012 | |
4013 _claimed = false; | |
4014 } | |
4015 | |
4016 CMTask::CMTask(int task_id, | |
4017 ConcurrentMark* cm, | |
4018 CMTaskQueue* task_queue, | |
4019 CMTaskQueueSet* task_queues) | |
4020 : _g1h(G1CollectedHeap::heap()), | |
4021 _co_tracker(G1CMGroup), | |
4022 _task_id(task_id), _cm(cm), | |
4023 _claimed(false), | |
4024 _nextMarkBitMap(NULL), _hash_seed(17), | |
4025 _task_queue(task_queue), | |
4026 _task_queues(task_queues), | |
4027 _oop_closure(NULL) { | |
4028 guarantee( task_queue != NULL, "invariant" ); | |
4029 guarantee( task_queues != NULL, "invariant" ); | |
4030 | |
4031 statsOnly( _clock_due_to_scanning = 0; | |
4032 _clock_due_to_marking = 0 ); | |
4033 | |
4034 _marking_step_diffs_ms.add(0.5); | |
4035 } |