1 /* 2 * Copyright (c) 2001, 2015, Oracle and/or its affiliates. All rights reserved. 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 4 * 5 * This code is free software; you can redistribute it and/or modify it 6 * under the terms of the GNU General Public License version 2 only, as 7 * published by the Free Software Foundation. 8 * 9 * This code is distributed in the hope that it will be useful, but WITHOUT 10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 12 * version 2 for more details (a copy is included in the LICENSE file that 13 * accompanied this code). 14 * 15 * You should have received a copy of the GNU General Public License version 16 * 2 along with this work; if not, write to the Free Software Foundation, 17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 18 * 19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 20 * or visit www.oracle.com if you need additional information or have any 21 * questions. 22 * 23 */ 24 25 #include "precompiled.hpp" 26 #include "classfile/metadataOnStackMark.hpp" 27 #include "classfile/symbolTable.hpp" 28 #include "code/codeCache.hpp" 29 #include "gc_implementation/g1/concurrentMark.inline.hpp" 30 #include "gc_implementation/g1/concurrentMarkThread.inline.hpp" 31 #include "gc_implementation/g1/g1CollectedHeap.inline.hpp" 32 #include "gc_implementation/g1/g1CollectorPolicy.hpp" 33 #include "gc_implementation/g1/g1ErgoVerbose.hpp" 34 #include "gc_implementation/g1/g1Log.hpp" 35 #include "gc_implementation/g1/g1OopClosures.inline.hpp" 36 #include "gc_implementation/g1/g1RemSet.hpp" 37 #include "gc_implementation/g1/heapRegion.inline.hpp" 38 #include "gc_implementation/g1/heapRegionManager.inline.hpp" 39 #include "gc_implementation/g1/heapRegionRemSet.hpp" 40 #include "gc_implementation/g1/heapRegionSet.inline.hpp" 41 #include "gc_implementation/shared/vmGCOperations.hpp" 42 #include "gc_implementation/shared/gcTimer.hpp" 43 #include "gc_implementation/shared/gcTrace.hpp" 44 #include "gc_implementation/shared/gcTraceTime.hpp" 45 #include "memory/allocation.hpp" 46 #include "memory/genOopClosures.inline.hpp" 47 #include "memory/referencePolicy.hpp" 48 #include "memory/resourceArea.hpp" 49 #include "memory/strongRootsScope.hpp" 50 #include "oops/oop.inline.hpp" 51 #include "runtime/handles.inline.hpp" 52 #include "runtime/java.hpp" 53 #include "runtime/atomic.inline.hpp" 54 #include "runtime/prefetch.inline.hpp" 55 #include "services/memTracker.hpp" 56 57 // Concurrent marking bit map wrapper 58 59 CMBitMapRO::CMBitMapRO(int shifter) : 60 _bm(), 61 _shifter(shifter) { 62 _bmStartWord = 0; 63 _bmWordSize = 0; 64 } 65 66 HeapWord* CMBitMapRO::getNextMarkedWordAddress(const HeapWord* addr, 67 const HeapWord* limit) const { 68 // First we must round addr *up* to a possible object boundary. 69 addr = (HeapWord*)align_size_up((intptr_t)addr, 70 HeapWordSize << _shifter); 71 size_t addrOffset = heapWordToOffset(addr); 72 if (limit == NULL) { 73 limit = _bmStartWord + _bmWordSize; 74 } 75 size_t limitOffset = heapWordToOffset(limit); 76 size_t nextOffset = _bm.get_next_one_offset(addrOffset, limitOffset); 77 HeapWord* nextAddr = offsetToHeapWord(nextOffset); 78 assert(nextAddr >= addr, "get_next_one postcondition"); 79 assert(nextAddr == limit || isMarked(nextAddr), 80 "get_next_one postcondition"); 81 return nextAddr; 82 } 83 84 HeapWord* CMBitMapRO::getNextUnmarkedWordAddress(const HeapWord* addr, 85 const HeapWord* limit) const { 86 size_t addrOffset = heapWordToOffset(addr); 87 if (limit == NULL) { 88 limit = _bmStartWord + _bmWordSize; 89 } 90 size_t limitOffset = heapWordToOffset(limit); 91 size_t nextOffset = _bm.get_next_zero_offset(addrOffset, limitOffset); 92 HeapWord* nextAddr = offsetToHeapWord(nextOffset); 93 assert(nextAddr >= addr, "get_next_one postcondition"); 94 assert(nextAddr == limit || !isMarked(nextAddr), 95 "get_next_one postcondition"); 96 return nextAddr; 97 } 98 99 int CMBitMapRO::heapWordDiffToOffsetDiff(size_t diff) const { 100 assert((diff & ((1 << _shifter) - 1)) == 0, "argument check"); 101 return (int) (diff >> _shifter); 102 } 103 104 #ifndef PRODUCT 105 bool CMBitMapRO::covers(MemRegion heap_rs) const { 106 // assert(_bm.map() == _virtual_space.low(), "map inconsistency"); 107 assert(((size_t)_bm.size() * ((size_t)1 << _shifter)) == _bmWordSize, 108 "size inconsistency"); 109 return _bmStartWord == (HeapWord*)(heap_rs.start()) && 110 _bmWordSize == heap_rs.word_size(); 111 } 112 #endif 113 114 void CMBitMapRO::print_on_error(outputStream* st, const char* prefix) const { 115 _bm.print_on_error(st, prefix); 116 } 117 118 size_t CMBitMap::compute_size(size_t heap_size) { 119 return ReservedSpace::allocation_align_size_up(heap_size / mark_distance()); 120 } 121 122 size_t CMBitMap::mark_distance() { 123 return MinObjAlignmentInBytes * BitsPerByte; 124 } 125 126 void CMBitMap::initialize(MemRegion heap, G1RegionToSpaceMapper* storage) { 127 _bmStartWord = heap.start(); 128 _bmWordSize = heap.word_size(); 129 130 _bm.set_map((BitMap::bm_word_t*) storage->reserved().start()); 131 _bm.set_size(_bmWordSize >> _shifter); 132 133 storage->set_mapping_changed_listener(&_listener); 134 } 135 136 void CMBitMapMappingChangedListener::on_commit(uint start_region, size_t num_regions, bool zero_filled) { 137 if (zero_filled) { 138 return; 139 } 140 // We need to clear the bitmap on commit, removing any existing information. 141 MemRegion mr(G1CollectedHeap::heap()->bottom_addr_for_region(start_region), num_regions * HeapRegion::GrainWords); 142 _bm->clearRange(mr); 143 } 144 145 // Closure used for clearing the given mark bitmap. 146 class ClearBitmapHRClosure : public HeapRegionClosure { 147 private: 148 ConcurrentMark* _cm; 149 CMBitMap* _bitmap; 150 bool _may_yield; // The closure may yield during iteration. If yielded, abort the iteration. 151 public: 152 ClearBitmapHRClosure(ConcurrentMark* cm, CMBitMap* bitmap, bool may_yield) : HeapRegionClosure(), _cm(cm), _bitmap(bitmap), _may_yield(may_yield) { 153 assert(!may_yield || cm != NULL, "CM must be non-NULL if this closure is expected to yield."); 154 } 155 156 virtual bool doHeapRegion(HeapRegion* r) { 157 size_t const chunk_size_in_words = M / HeapWordSize; 158 159 HeapWord* cur = r->bottom(); 160 HeapWord* const end = r->end(); 161 162 while (cur < end) { 163 MemRegion mr(cur, MIN2(cur + chunk_size_in_words, end)); 164 _bitmap->clearRange(mr); 165 166 cur += chunk_size_in_words; 167 168 // Abort iteration if after yielding the marking has been aborted. 169 if (_may_yield && _cm->do_yield_check() && _cm->has_aborted()) { 170 return true; 171 } 172 // Repeat the asserts from before the start of the closure. We will do them 173 // as asserts here to minimize their overhead on the product. However, we 174 // will have them as guarantees at the beginning / end of the bitmap 175 // clearing to get some checking in the product. 176 assert(!_may_yield || _cm->cmThread()->during_cycle(), "invariant"); 177 assert(!_may_yield || !G1CollectedHeap::heap()->mark_in_progress(), "invariant"); 178 } 179 180 return false; 181 } 182 }; 183 184 class ParClearNextMarkBitmapTask : public AbstractGangTask { 185 ClearBitmapHRClosure* _cl; 186 HeapRegionClaimer _hrclaimer; 187 bool _suspendible; // If the task is suspendible, workers must join the STS. 188 189 public: 190 ParClearNextMarkBitmapTask(ClearBitmapHRClosure *cl, uint n_workers, bool suspendible) : 191 _cl(cl), _suspendible(suspendible), AbstractGangTask("Parallel Clear Bitmap Task"), _hrclaimer(n_workers) {} 192 193 void work(uint worker_id) { 194 if (_suspendible) { 195 SuspendibleThreadSet::join(); 196 } 197 G1CollectedHeap::heap()->heap_region_par_iterate(_cl, worker_id, &_hrclaimer, true); 198 if (_suspendible) { 199 SuspendibleThreadSet::leave(); 200 } 201 } 202 }; 203 204 void CMBitMap::clearAll() { 205 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 206 ClearBitmapHRClosure cl(NULL, this, false /* may_yield */); 207 uint n_workers = g1h->workers()->active_workers(); 208 ParClearNextMarkBitmapTask task(&cl, n_workers, false); 209 g1h->workers()->run_task(&task); 210 guarantee(cl.complete(), "Must have completed iteration."); 211 return; 212 } 213 214 void CMBitMap::markRange(MemRegion mr) { 215 mr.intersection(MemRegion(_bmStartWord, _bmWordSize)); 216 assert(!mr.is_empty(), "unexpected empty region"); 217 assert((offsetToHeapWord(heapWordToOffset(mr.end())) == 218 ((HeapWord *) mr.end())), 219 "markRange memory region end is not card aligned"); 220 // convert address range into offset range 221 _bm.at_put_range(heapWordToOffset(mr.start()), 222 heapWordToOffset(mr.end()), true); 223 } 224 225 void CMBitMap::clearRange(MemRegion mr) { 226 mr.intersection(MemRegion(_bmStartWord, _bmWordSize)); 227 assert(!mr.is_empty(), "unexpected empty region"); 228 // convert address range into offset range 229 _bm.at_put_range(heapWordToOffset(mr.start()), 230 heapWordToOffset(mr.end()), false); 231 } 232 233 MemRegion CMBitMap::getAndClearMarkedRegion(HeapWord* addr, 234 HeapWord* end_addr) { 235 HeapWord* start = getNextMarkedWordAddress(addr); 236 start = MIN2(start, end_addr); 237 HeapWord* end = getNextUnmarkedWordAddress(start); 238 end = MIN2(end, end_addr); 239 assert(start <= end, "Consistency check"); 240 MemRegion mr(start, end); 241 if (!mr.is_empty()) { 242 clearRange(mr); 243 } 244 return mr; 245 } 246 247 CMMarkStack::CMMarkStack(ConcurrentMark* cm) : 248 _base(NULL), _cm(cm) 249 #ifdef ASSERT 250 , _drain_in_progress(false) 251 , _drain_in_progress_yields(false) 252 #endif 253 {} 254 255 bool CMMarkStack::allocate(size_t capacity) { 256 // allocate a stack of the requisite depth 257 ReservedSpace rs(ReservedSpace::allocation_align_size_up(capacity * sizeof(oop))); 258 if (!rs.is_reserved()) { 259 warning("ConcurrentMark MarkStack allocation failure"); 260 return false; 261 } 262 MemTracker::record_virtual_memory_type((address)rs.base(), mtGC); 263 if (!_virtual_space.initialize(rs, rs.size())) { 264 warning("ConcurrentMark MarkStack backing store failure"); 265 // Release the virtual memory reserved for the marking stack 266 rs.release(); 267 return false; 268 } 269 assert(_virtual_space.committed_size() == rs.size(), 270 "Didn't reserve backing store for all of ConcurrentMark stack?"); 271 _base = (oop*) _virtual_space.low(); 272 setEmpty(); 273 _capacity = (jint) capacity; 274 _saved_index = -1; 275 _should_expand = false; 276 NOT_PRODUCT(_max_depth = 0); 277 return true; 278 } 279 280 void CMMarkStack::expand() { 281 // Called, during remark, if we've overflown the marking stack during marking. 282 assert(isEmpty(), "stack should been emptied while handling overflow"); 283 assert(_capacity <= (jint) MarkStackSizeMax, "stack bigger than permitted"); 284 // Clear expansion flag 285 _should_expand = false; 286 if (_capacity == (jint) MarkStackSizeMax) { 287 if (PrintGCDetails && Verbose) { 288 gclog_or_tty->print_cr(" (benign) Can't expand marking stack capacity, at max size limit"); 289 } 290 return; 291 } 292 // Double capacity if possible 293 jint new_capacity = MIN2(_capacity*2, (jint) MarkStackSizeMax); 294 // Do not give up existing stack until we have managed to 295 // get the double capacity that we desired. 296 ReservedSpace rs(ReservedSpace::allocation_align_size_up(new_capacity * 297 sizeof(oop))); 298 if (rs.is_reserved()) { 299 // Release the backing store associated with old stack 300 _virtual_space.release(); 301 // Reinitialize virtual space for new stack 302 if (!_virtual_space.initialize(rs, rs.size())) { 303 fatal("Not enough swap for expanded marking stack capacity"); 304 } 305 _base = (oop*)(_virtual_space.low()); 306 _index = 0; 307 _capacity = new_capacity; 308 } else { 309 if (PrintGCDetails && Verbose) { 310 // Failed to double capacity, continue; 311 gclog_or_tty->print(" (benign) Failed to expand marking stack capacity from " 312 SIZE_FORMAT"K to " SIZE_FORMAT"K", 313 _capacity / K, new_capacity / K); 314 } 315 } 316 } 317 318 void CMMarkStack::set_should_expand() { 319 // If we're resetting the marking state because of an 320 // marking stack overflow, record that we should, if 321 // possible, expand the stack. 322 _should_expand = _cm->has_overflown(); 323 } 324 325 CMMarkStack::~CMMarkStack() { 326 if (_base != NULL) { 327 _base = NULL; 328 _virtual_space.release(); 329 } 330 } 331 332 void CMMarkStack::par_push(oop ptr) { 333 while (true) { 334 if (isFull()) { 335 _overflow = true; 336 return; 337 } 338 // Otherwise... 339 jint index = _index; 340 jint next_index = index+1; 341 jint res = Atomic::cmpxchg(next_index, &_index, index); 342 if (res == index) { 343 _base[index] = ptr; 344 // Note that we don't maintain this atomically. We could, but it 345 // doesn't seem necessary. 346 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index)); 347 return; 348 } 349 // Otherwise, we need to try again. 350 } 351 } 352 353 void CMMarkStack::par_adjoin_arr(oop* ptr_arr, int n) { 354 while (true) { 355 if (isFull()) { 356 _overflow = true; 357 return; 358 } 359 // Otherwise... 360 jint index = _index; 361 jint next_index = index + n; 362 if (next_index > _capacity) { 363 _overflow = true; 364 return; 365 } 366 jint res = Atomic::cmpxchg(next_index, &_index, index); 367 if (res == index) { 368 for (int i = 0; i < n; i++) { 369 int ind = index + i; 370 assert(ind < _capacity, "By overflow test above."); 371 _base[ind] = ptr_arr[i]; 372 } 373 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index)); 374 return; 375 } 376 // Otherwise, we need to try again. 377 } 378 } 379 380 void CMMarkStack::par_push_arr(oop* ptr_arr, int n) { 381 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); 382 jint start = _index; 383 jint next_index = start + n; 384 if (next_index > _capacity) { 385 _overflow = true; 386 return; 387 } 388 // Otherwise. 389 _index = next_index; 390 for (int i = 0; i < n; i++) { 391 int ind = start + i; 392 assert(ind < _capacity, "By overflow test above."); 393 _base[ind] = ptr_arr[i]; 394 } 395 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index)); 396 } 397 398 bool CMMarkStack::par_pop_arr(oop* ptr_arr, int max, int* n) { 399 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); 400 jint index = _index; 401 if (index == 0) { 402 *n = 0; 403 return false; 404 } else { 405 int k = MIN2(max, index); 406 jint new_ind = index - k; 407 for (int j = 0; j < k; j++) { 408 ptr_arr[j] = _base[new_ind + j]; 409 } 410 _index = new_ind; 411 *n = k; 412 return true; 413 } 414 } 415 416 template<class OopClosureClass> 417 bool CMMarkStack::drain(OopClosureClass* cl, CMBitMap* bm, bool yield_after) { 418 assert(!_drain_in_progress || !_drain_in_progress_yields || yield_after 419 || SafepointSynchronize::is_at_safepoint(), 420 "Drain recursion must be yield-safe."); 421 bool res = true; 422 debug_only(_drain_in_progress = true); 423 debug_only(_drain_in_progress_yields = yield_after); 424 while (!isEmpty()) { 425 oop newOop = pop(); 426 assert(G1CollectedHeap::heap()->is_in_reserved(newOop), "Bad pop"); 427 assert(newOop->is_oop(), "Expected an oop"); 428 assert(bm == NULL || bm->isMarked((HeapWord*)newOop), 429 "only grey objects on this stack"); 430 newOop->oop_iterate(cl); 431 if (yield_after && _cm->do_yield_check()) { 432 res = false; 433 break; 434 } 435 } 436 debug_only(_drain_in_progress = false); 437 return res; 438 } 439 440 void CMMarkStack::note_start_of_gc() { 441 assert(_saved_index == -1, 442 "note_start_of_gc()/end_of_gc() bracketed incorrectly"); 443 _saved_index = _index; 444 } 445 446 void CMMarkStack::note_end_of_gc() { 447 // This is intentionally a guarantee, instead of an assert. If we 448 // accidentally add something to the mark stack during GC, it 449 // will be a correctness issue so it's better if we crash. we'll 450 // only check this once per GC anyway, so it won't be a performance 451 // issue in any way. 452 guarantee(_saved_index == _index, 453 err_msg("saved index: %d index: %d", _saved_index, _index)); 454 _saved_index = -1; 455 } 456 457 void CMMarkStack::oops_do(OopClosure* f) { 458 assert(_saved_index == _index, 459 err_msg("saved index: %d index: %d", _saved_index, _index)); 460 for (int i = 0; i < _index; i += 1) { 461 f->do_oop(&_base[i]); 462 } 463 } 464 465 CMRootRegions::CMRootRegions() : 466 _young_list(NULL), _cm(NULL), _scan_in_progress(false), 467 _should_abort(false), _next_survivor(NULL) { } 468 469 void CMRootRegions::init(G1CollectedHeap* g1h, ConcurrentMark* cm) { 470 _young_list = g1h->young_list(); 471 _cm = cm; 472 } 473 474 void CMRootRegions::prepare_for_scan() { 475 assert(!scan_in_progress(), "pre-condition"); 476 477 // Currently, only survivors can be root regions. 478 assert(_next_survivor == NULL, "pre-condition"); 479 _next_survivor = _young_list->first_survivor_region(); 480 _scan_in_progress = (_next_survivor != NULL); 481 _should_abort = false; 482 } 483 484 HeapRegion* CMRootRegions::claim_next() { 485 if (_should_abort) { 486 // If someone has set the should_abort flag, we return NULL to 487 // force the caller to bail out of their loop. 488 return NULL; 489 } 490 491 // Currently, only survivors can be root regions. 492 HeapRegion* res = _next_survivor; 493 if (res != NULL) { 494 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag); 495 // Read it again in case it changed while we were waiting for the lock. 496 res = _next_survivor; 497 if (res != NULL) { 498 if (res == _young_list->last_survivor_region()) { 499 // We just claimed the last survivor so store NULL to indicate 500 // that we're done. 501 _next_survivor = NULL; 502 } else { 503 _next_survivor = res->get_next_young_region(); 504 } 505 } else { 506 // Someone else claimed the last survivor while we were trying 507 // to take the lock so nothing else to do. 508 } 509 } 510 assert(res == NULL || res->is_survivor(), "post-condition"); 511 512 return res; 513 } 514 515 void CMRootRegions::scan_finished() { 516 assert(scan_in_progress(), "pre-condition"); 517 518 // Currently, only survivors can be root regions. 519 if (!_should_abort) { 520 assert(_next_survivor == NULL, "we should have claimed all survivors"); 521 } 522 _next_survivor = NULL; 523 524 { 525 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag); 526 _scan_in_progress = false; 527 RootRegionScan_lock->notify_all(); 528 } 529 } 530 531 bool CMRootRegions::wait_until_scan_finished() { 532 if (!scan_in_progress()) return false; 533 534 { 535 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag); 536 while (scan_in_progress()) { 537 RootRegionScan_lock->wait(Mutex::_no_safepoint_check_flag); 538 } 539 } 540 return true; 541 } 542 543 #ifdef _MSC_VER // the use of 'this' below gets a warning, make it go away 544 #pragma warning( disable:4355 ) // 'this' : used in base member initializer list 545 #endif // _MSC_VER 546 547 uint ConcurrentMark::scale_parallel_threads(uint n_par_threads) { 548 return MAX2((n_par_threads + 2) / 4, 1U); 549 } 550 551 ConcurrentMark::ConcurrentMark(G1CollectedHeap* g1h, G1RegionToSpaceMapper* prev_bitmap_storage, G1RegionToSpaceMapper* next_bitmap_storage) : 552 _g1h(g1h), 553 _markBitMap1(), 554 _markBitMap2(), 555 _parallel_marking_threads(0), 556 _max_parallel_marking_threads(0), 557 _sleep_factor(0.0), 558 _marking_task_overhead(1.0), 559 _cleanup_sleep_factor(0.0), 560 _cleanup_task_overhead(1.0), 561 _cleanup_list("Cleanup List"), 562 _region_bm((BitMap::idx_t)(g1h->max_regions()), false /* in_resource_area*/), 563 _card_bm((g1h->reserved_region().byte_size() + CardTableModRefBS::card_size - 1) >> 564 CardTableModRefBS::card_shift, 565 false /* in_resource_area*/), 566 567 _prevMarkBitMap(&_markBitMap1), 568 _nextMarkBitMap(&_markBitMap2), 569 570 _markStack(this), 571 // _finger set in set_non_marking_state 572 573 _max_worker_id(MAX2((uint)ParallelGCThreads, 1U)), 574 // _active_tasks set in set_non_marking_state 575 // _tasks set inside the constructor 576 _task_queues(new CMTaskQueueSet((int) _max_worker_id)), 577 _terminator(ParallelTaskTerminator((int) _max_worker_id, _task_queues)), 578 579 _has_overflown(false), 580 _concurrent(false), 581 _has_aborted(false), 582 _aborted_gc_id(GCId::undefined()), 583 _restart_for_overflow(false), 584 _concurrent_marking_in_progress(false), 585 586 // _verbose_level set below 587 588 _init_times(), 589 _remark_times(), _remark_mark_times(), _remark_weak_ref_times(), 590 _cleanup_times(), 591 _total_counting_time(0.0), 592 _total_rs_scrub_time(0.0), 593 594 _parallel_workers(NULL), 595 596 _count_card_bitmaps(NULL), 597 _count_marked_bytes(NULL), 598 _completed_initialization(false) { 599 CMVerboseLevel verbose_level = (CMVerboseLevel) G1MarkingVerboseLevel; 600 if (verbose_level < no_verbose) { 601 verbose_level = no_verbose; 602 } 603 if (verbose_level > high_verbose) { 604 verbose_level = high_verbose; 605 } 606 _verbose_level = verbose_level; 607 608 if (verbose_low()) { 609 gclog_or_tty->print_cr("[global] init, heap start = "PTR_FORMAT", " 610 "heap end = " PTR_FORMAT, p2i(_heap_start), p2i(_heap_end)); 611 } 612 613 _markBitMap1.initialize(g1h->reserved_region(), prev_bitmap_storage); 614 _markBitMap2.initialize(g1h->reserved_region(), next_bitmap_storage); 615 616 // Create & start a ConcurrentMark thread. 617 _cmThread = new ConcurrentMarkThread(this); 618 assert(cmThread() != NULL, "CM Thread should have been created"); 619 assert(cmThread()->cm() != NULL, "CM Thread should refer to this cm"); 620 if (_cmThread->osthread() == NULL) { 621 vm_shutdown_during_initialization("Could not create ConcurrentMarkThread"); 622 } 623 624 assert(CGC_lock != NULL, "Where's the CGC_lock?"); 625 assert(_markBitMap1.covers(g1h->reserved_region()), "_markBitMap1 inconsistency"); 626 assert(_markBitMap2.covers(g1h->reserved_region()), "_markBitMap2 inconsistency"); 627 628 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set(); 629 satb_qs.set_buffer_size(G1SATBBufferSize); 630 631 _root_regions.init(_g1h, this); 632 633 if (ConcGCThreads > ParallelGCThreads) { 634 warning("Can't have more ConcGCThreads (" UINTX_FORMAT ") " 635 "than ParallelGCThreads (" UINTX_FORMAT ").", 636 ConcGCThreads, ParallelGCThreads); 637 return; 638 } 639 if (!FLAG_IS_DEFAULT(ConcGCThreads) && ConcGCThreads > 0) { 640 // Note: ConcGCThreads has precedence over G1MarkingOverheadPercent 641 // if both are set 642 _sleep_factor = 0.0; 643 _marking_task_overhead = 1.0; 644 } else if (G1MarkingOverheadPercent > 0) { 645 // We will calculate the number of parallel marking threads based 646 // on a target overhead with respect to the soft real-time goal 647 double marking_overhead = (double) G1MarkingOverheadPercent / 100.0; 648 double overall_cm_overhead = 649 (double) MaxGCPauseMillis * marking_overhead / 650 (double) GCPauseIntervalMillis; 651 double cpu_ratio = 1.0 / (double) os::processor_count(); 652 double marking_thread_num = ceil(overall_cm_overhead / cpu_ratio); 653 double marking_task_overhead = 654 overall_cm_overhead / marking_thread_num * 655 (double) os::processor_count(); 656 double sleep_factor = 657 (1.0 - marking_task_overhead) / marking_task_overhead; 658 659 FLAG_SET_ERGO(uintx, ConcGCThreads, (uint) marking_thread_num); 660 _sleep_factor = sleep_factor; 661 _marking_task_overhead = marking_task_overhead; 662 } else { 663 // Calculate the number of parallel marking threads by scaling 664 // the number of parallel GC threads. 665 uint marking_thread_num = scale_parallel_threads((uint) ParallelGCThreads); 666 FLAG_SET_ERGO(uintx, ConcGCThreads, marking_thread_num); 667 _sleep_factor = 0.0; 668 _marking_task_overhead = 1.0; 669 } 670 671 assert(ConcGCThreads > 0, "Should have been set"); 672 _parallel_marking_threads = (uint) ConcGCThreads; 673 _max_parallel_marking_threads = _parallel_marking_threads; 674 675 if (parallel_marking_threads() > 1) { 676 _cleanup_task_overhead = 1.0; 677 } else { 678 _cleanup_task_overhead = marking_task_overhead(); 679 } 680 _cleanup_sleep_factor = 681 (1.0 - cleanup_task_overhead()) / cleanup_task_overhead(); 682 683 #if 0 684 gclog_or_tty->print_cr("Marking Threads %d", parallel_marking_threads()); 685 gclog_or_tty->print_cr("CM Marking Task Overhead %1.4lf", marking_task_overhead()); 686 gclog_or_tty->print_cr("CM Sleep Factor %1.4lf", sleep_factor()); 687 gclog_or_tty->print_cr("CL Marking Task Overhead %1.4lf", cleanup_task_overhead()); 688 gclog_or_tty->print_cr("CL Sleep Factor %1.4lf", cleanup_sleep_factor()); 689 #endif 690 691 _parallel_workers = new FlexibleWorkGang("G1 Marker", 692 _max_parallel_marking_threads, false, true); 693 if (_parallel_workers == NULL) { 694 vm_exit_during_initialization("Failed necessary allocation."); 695 } else { 696 _parallel_workers->initialize_workers(); 697 } 698 699 if (FLAG_IS_DEFAULT(MarkStackSize)) { 700 size_t mark_stack_size = 701 MIN2(MarkStackSizeMax, 702 MAX2(MarkStackSize, (size_t) (parallel_marking_threads() * TASKQUEUE_SIZE))); 703 // Verify that the calculated value for MarkStackSize is in range. 704 // It would be nice to use the private utility routine from Arguments. 705 if (!(mark_stack_size >= 1 && mark_stack_size <= MarkStackSizeMax)) { 706 warning("Invalid value calculated for MarkStackSize (" SIZE_FORMAT "): " 707 "must be between 1 and " SIZE_FORMAT, 708 mark_stack_size, MarkStackSizeMax); 709 return; 710 } 711 FLAG_SET_ERGO(size_t, MarkStackSize, mark_stack_size); 712 } else { 713 // Verify MarkStackSize is in range. 714 if (FLAG_IS_CMDLINE(MarkStackSize)) { 715 if (FLAG_IS_DEFAULT(MarkStackSizeMax)) { 716 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) { 717 warning("Invalid value specified for MarkStackSize (" SIZE_FORMAT "): " 718 "must be between 1 and " SIZE_FORMAT, 719 MarkStackSize, MarkStackSizeMax); 720 return; 721 } 722 } else if (FLAG_IS_CMDLINE(MarkStackSizeMax)) { 723 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) { 724 warning("Invalid value specified for MarkStackSize (" SIZE_FORMAT ")" 725 " or for MarkStackSizeMax (" SIZE_FORMAT ")", 726 MarkStackSize, MarkStackSizeMax); 727 return; 728 } 729 } 730 } 731 } 732 733 if (!_markStack.allocate(MarkStackSize)) { 734 warning("Failed to allocate CM marking stack"); 735 return; 736 } 737 738 _tasks = NEW_C_HEAP_ARRAY(CMTask*, _max_worker_id, mtGC); 739 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_worker_id, mtGC); 740 741 _count_card_bitmaps = NEW_C_HEAP_ARRAY(BitMap, _max_worker_id, mtGC); 742 _count_marked_bytes = NEW_C_HEAP_ARRAY(size_t*, _max_worker_id, mtGC); 743 744 BitMap::idx_t card_bm_size = _card_bm.size(); 745 746 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail 747 _active_tasks = _max_worker_id; 748 749 uint max_regions = _g1h->max_regions(); 750 for (uint i = 0; i < _max_worker_id; ++i) { 751 CMTaskQueue* task_queue = new CMTaskQueue(); 752 task_queue->initialize(); 753 _task_queues->register_queue(i, task_queue); 754 755 _count_card_bitmaps[i] = BitMap(card_bm_size, false); 756 _count_marked_bytes[i] = NEW_C_HEAP_ARRAY(size_t, max_regions, mtGC); 757 758 _tasks[i] = new CMTask(i, this, 759 _count_marked_bytes[i], 760 &_count_card_bitmaps[i], 761 task_queue, _task_queues); 762 763 _accum_task_vtime[i] = 0.0; 764 } 765 766 // Calculate the card number for the bottom of the heap. Used 767 // in biasing indexes into the accounting card bitmaps. 768 _heap_bottom_card_num = 769 intptr_t(uintptr_t(_g1h->reserved_region().start()) >> 770 CardTableModRefBS::card_shift); 771 772 // Clear all the liveness counting data 773 clear_all_count_data(); 774 775 // so that the call below can read a sensible value 776 _heap_start = g1h->reserved_region().start(); 777 set_non_marking_state(); 778 _completed_initialization = true; 779 } 780 781 void ConcurrentMark::reset() { 782 // Starting values for these two. This should be called in a STW 783 // phase. 784 MemRegion reserved = _g1h->g1_reserved(); 785 _heap_start = reserved.start(); 786 _heap_end = reserved.end(); 787 788 // Separated the asserts so that we know which one fires. 789 assert(_heap_start != NULL, "heap bounds should look ok"); 790 assert(_heap_end != NULL, "heap bounds should look ok"); 791 assert(_heap_start < _heap_end, "heap bounds should look ok"); 792 793 // Reset all the marking data structures and any necessary flags 794 reset_marking_state(); 795 796 if (verbose_low()) { 797 gclog_or_tty->print_cr("[global] resetting"); 798 } 799 800 // We do reset all of them, since different phases will use 801 // different number of active threads. So, it's easiest to have all 802 // of them ready. 803 for (uint i = 0; i < _max_worker_id; ++i) { 804 _tasks[i]->reset(_nextMarkBitMap); 805 } 806 807 // we need this to make sure that the flag is on during the evac 808 // pause with initial mark piggy-backed 809 set_concurrent_marking_in_progress(); 810 } 811 812 813 void ConcurrentMark::reset_marking_state(bool clear_overflow) { 814 _markStack.set_should_expand(); 815 _markStack.setEmpty(); // Also clears the _markStack overflow flag 816 if (clear_overflow) { 817 clear_has_overflown(); 818 } else { 819 assert(has_overflown(), "pre-condition"); 820 } 821 _finger = _heap_start; 822 823 for (uint i = 0; i < _max_worker_id; ++i) { 824 CMTaskQueue* queue = _task_queues->queue(i); 825 queue->set_empty(); 826 } 827 } 828 829 void ConcurrentMark::set_concurrency(uint active_tasks) { 830 assert(active_tasks <= _max_worker_id, "we should not have more"); 831 832 _active_tasks = active_tasks; 833 // Need to update the three data structures below according to the 834 // number of active threads for this phase. 835 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues); 836 _first_overflow_barrier_sync.set_n_workers((int) active_tasks); 837 _second_overflow_barrier_sync.set_n_workers((int) active_tasks); 838 } 839 840 void ConcurrentMark::set_concurrency_and_phase(uint active_tasks, bool concurrent) { 841 set_concurrency(active_tasks); 842 843 _concurrent = concurrent; 844 // We propagate this to all tasks, not just the active ones. 845 for (uint i = 0; i < _max_worker_id; ++i) 846 _tasks[i]->set_concurrent(concurrent); 847 848 if (concurrent) { 849 set_concurrent_marking_in_progress(); 850 } else { 851 // We currently assume that the concurrent flag has been set to 852 // false before we start remark. At this point we should also be 853 // in a STW phase. 854 assert(!concurrent_marking_in_progress(), "invariant"); 855 assert(out_of_regions(), 856 err_msg("only way to get here: _finger: "PTR_FORMAT", _heap_end: "PTR_FORMAT, 857 p2i(_finger), p2i(_heap_end))); 858 } 859 } 860 861 void ConcurrentMark::set_non_marking_state() { 862 // We set the global marking state to some default values when we're 863 // not doing marking. 864 reset_marking_state(); 865 _active_tasks = 0; 866 clear_concurrent_marking_in_progress(); 867 } 868 869 ConcurrentMark::~ConcurrentMark() { 870 // The ConcurrentMark instance is never freed. 871 ShouldNotReachHere(); 872 } 873 874 void ConcurrentMark::clearNextBitmap() { 875 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 876 877 // Make sure that the concurrent mark thread looks to still be in 878 // the current cycle. 879 guarantee(cmThread()->during_cycle(), "invariant"); 880 881 // We are finishing up the current cycle by clearing the next 882 // marking bitmap and getting it ready for the next cycle. During 883 // this time no other cycle can start. So, let's make sure that this 884 // is the case. 885 guarantee(!g1h->mark_in_progress(), "invariant"); 886 887 ClearBitmapHRClosure cl(this, _nextMarkBitMap, true /* may_yield */); 888 ParClearNextMarkBitmapTask task(&cl, parallel_marking_threads(), true); 889 _parallel_workers->run_task(&task); 890 891 // Clear the liveness counting data. If the marking has been aborted, the abort() 892 // call already did that. 893 if (cl.complete()) { 894 clear_all_count_data(); 895 } 896 897 // Repeat the asserts from above. 898 guarantee(cmThread()->during_cycle(), "invariant"); 899 guarantee(!g1h->mark_in_progress(), "invariant"); 900 } 901 902 class CheckBitmapClearHRClosure : public HeapRegionClosure { 903 CMBitMap* _bitmap; 904 bool _error; 905 public: 906 CheckBitmapClearHRClosure(CMBitMap* bitmap) : _bitmap(bitmap) { 907 } 908 909 virtual bool doHeapRegion(HeapRegion* r) { 910 // This closure can be called concurrently to the mutator, so we must make sure 911 // that the result of the getNextMarkedWordAddress() call is compared to the 912 // value passed to it as limit to detect any found bits. 913 // We can use the region's orig_end() for the limit and the comparison value 914 // as it always contains the "real" end of the region that never changes and 915 // has no side effects. 916 // Due to the latter, there can also be no problem with the compiler generating 917 // reloads of the orig_end() call. 918 HeapWord* end = r->orig_end(); 919 return _bitmap->getNextMarkedWordAddress(r->bottom(), end) != end; 920 } 921 }; 922 923 bool ConcurrentMark::nextMarkBitmapIsClear() { 924 CheckBitmapClearHRClosure cl(_nextMarkBitMap); 925 _g1h->heap_region_iterate(&cl); 926 return cl.complete(); 927 } 928 929 class NoteStartOfMarkHRClosure: public HeapRegionClosure { 930 public: 931 bool doHeapRegion(HeapRegion* r) { 932 if (!r->is_continues_humongous()) { 933 r->note_start_of_marking(); 934 } 935 return false; 936 } 937 }; 938 939 void ConcurrentMark::checkpointRootsInitialPre() { 940 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 941 G1CollectorPolicy* g1p = g1h->g1_policy(); 942 943 _has_aborted = false; 944 945 #ifndef PRODUCT 946 if (G1PrintReachableAtInitialMark) { 947 print_reachable("at-cycle-start", 948 VerifyOption_G1UsePrevMarking, true /* all */); 949 } 950 #endif 951 952 // Initialize marking structures. This has to be done in a STW phase. 953 reset(); 954 955 // For each region note start of marking. 956 NoteStartOfMarkHRClosure startcl; 957 g1h->heap_region_iterate(&startcl); 958 } 959 960 961 void ConcurrentMark::checkpointRootsInitialPost() { 962 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 963 964 // If we force an overflow during remark, the remark operation will 965 // actually abort and we'll restart concurrent marking. If we always 966 // force an overflow during remark we'll never actually complete the 967 // marking phase. So, we initialize this here, at the start of the 968 // cycle, so that at the remaining overflow number will decrease at 969 // every remark and we'll eventually not need to cause one. 970 force_overflow_stw()->init(); 971 972 // Start Concurrent Marking weak-reference discovery. 973 ReferenceProcessor* rp = g1h->ref_processor_cm(); 974 // enable ("weak") refs discovery 975 rp->enable_discovery(); 976 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle 977 978 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 979 // This is the start of the marking cycle, we're expected all 980 // threads to have SATB queues with active set to false. 981 satb_mq_set.set_active_all_threads(true, /* new active value */ 982 false /* expected_active */); 983 984 _root_regions.prepare_for_scan(); 985 986 // update_g1_committed() will be called at the end of an evac pause 987 // when marking is on. So, it's also called at the end of the 988 // initial-mark pause to update the heap end, if the heap expands 989 // during it. No need to call it here. 990 } 991 992 /* 993 * Notice that in the next two methods, we actually leave the STS 994 * during the barrier sync and join it immediately afterwards. If we 995 * do not do this, the following deadlock can occur: one thread could 996 * be in the barrier sync code, waiting for the other thread to also 997 * sync up, whereas another one could be trying to yield, while also 998 * waiting for the other threads to sync up too. 999 * 1000 * Note, however, that this code is also used during remark and in 1001 * this case we should not attempt to leave / enter the STS, otherwise 1002 * we'll either hit an assert (debug / fastdebug) or deadlock 1003 * (product). So we should only leave / enter the STS if we are 1004 * operating concurrently. 1005 * 1006 * Because the thread that does the sync barrier has left the STS, it 1007 * is possible to be suspended for a Full GC or an evacuation pause 1008 * could occur. This is actually safe, since the entering the sync 1009 * barrier is one of the last things do_marking_step() does, and it 1010 * doesn't manipulate any data structures afterwards. 1011 */ 1012 1013 void ConcurrentMark::enter_first_sync_barrier(uint worker_id) { 1014 if (verbose_low()) { 1015 gclog_or_tty->print_cr("[%u] entering first barrier", worker_id); 1016 } 1017 1018 if (concurrent()) { 1019 SuspendibleThreadSet::leave(); 1020 } 1021 1022 bool barrier_aborted = !_first_overflow_barrier_sync.enter(); 1023 1024 if (concurrent()) { 1025 SuspendibleThreadSet::join(); 1026 } 1027 // at this point everyone should have synced up and not be doing any 1028 // more work 1029 1030 if (verbose_low()) { 1031 if (barrier_aborted) { 1032 gclog_or_tty->print_cr("[%u] aborted first barrier", worker_id); 1033 } else { 1034 gclog_or_tty->print_cr("[%u] leaving first barrier", worker_id); 1035 } 1036 } 1037 1038 if (barrier_aborted) { 1039 // If the barrier aborted we ignore the overflow condition and 1040 // just abort the whole marking phase as quickly as possible. 1041 return; 1042 } 1043 1044 // If we're executing the concurrent phase of marking, reset the marking 1045 // state; otherwise the marking state is reset after reference processing, 1046 // during the remark pause. 1047 // If we reset here as a result of an overflow during the remark we will 1048 // see assertion failures from any subsequent set_concurrency_and_phase() 1049 // calls. 1050 if (concurrent()) { 1051 // let the task associated with with worker 0 do this 1052 if (worker_id == 0) { 1053 // task 0 is responsible for clearing the global data structures 1054 // We should be here because of an overflow. During STW we should 1055 // not clear the overflow flag since we rely on it being true when 1056 // we exit this method to abort the pause and restart concurrent 1057 // marking. 1058 reset_marking_state(true /* clear_overflow */); 1059 force_overflow()->update(); 1060 1061 if (G1Log::fine()) { 1062 gclog_or_tty->gclog_stamp(concurrent_gc_id()); 1063 gclog_or_tty->print_cr("[GC concurrent-mark-reset-for-overflow]"); 1064 } 1065 } 1066 } 1067 1068 // after this, each task should reset its own data structures then 1069 // then go into the second barrier 1070 } 1071 1072 void ConcurrentMark::enter_second_sync_barrier(uint worker_id) { 1073 if (verbose_low()) { 1074 gclog_or_tty->print_cr("[%u] entering second barrier", worker_id); 1075 } 1076 1077 if (concurrent()) { 1078 SuspendibleThreadSet::leave(); 1079 } 1080 1081 bool barrier_aborted = !_second_overflow_barrier_sync.enter(); 1082 1083 if (concurrent()) { 1084 SuspendibleThreadSet::join(); 1085 } 1086 // at this point everything should be re-initialized and ready to go 1087 1088 if (verbose_low()) { 1089 if (barrier_aborted) { 1090 gclog_or_tty->print_cr("[%u] aborted second barrier", worker_id); 1091 } else { 1092 gclog_or_tty->print_cr("[%u] leaving second barrier", worker_id); 1093 } 1094 } 1095 } 1096 1097 #ifndef PRODUCT 1098 void ForceOverflowSettings::init() { 1099 _num_remaining = G1ConcMarkForceOverflow; 1100 _force = false; 1101 update(); 1102 } 1103 1104 void ForceOverflowSettings::update() { 1105 if (_num_remaining > 0) { 1106 _num_remaining -= 1; 1107 _force = true; 1108 } else { 1109 _force = false; 1110 } 1111 } 1112 1113 bool ForceOverflowSettings::should_force() { 1114 if (_force) { 1115 _force = false; 1116 return true; 1117 } else { 1118 return false; 1119 } 1120 } 1121 #endif // !PRODUCT 1122 1123 class CMConcurrentMarkingTask: public AbstractGangTask { 1124 private: 1125 ConcurrentMark* _cm; 1126 ConcurrentMarkThread* _cmt; 1127 1128 public: 1129 void work(uint worker_id) { 1130 assert(Thread::current()->is_ConcurrentGC_thread(), 1131 "this should only be done by a conc GC thread"); 1132 ResourceMark rm; 1133 1134 double start_vtime = os::elapsedVTime(); 1135 1136 SuspendibleThreadSet::join(); 1137 1138 assert(worker_id < _cm->active_tasks(), "invariant"); 1139 CMTask* the_task = _cm->task(worker_id); 1140 the_task->record_start_time(); 1141 if (!_cm->has_aborted()) { 1142 do { 1143 double start_vtime_sec = os::elapsedVTime(); 1144 double mark_step_duration_ms = G1ConcMarkStepDurationMillis; 1145 1146 the_task->do_marking_step(mark_step_duration_ms, 1147 true /* do_termination */, 1148 false /* is_serial*/); 1149 1150 double end_vtime_sec = os::elapsedVTime(); 1151 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec; 1152 _cm->clear_has_overflown(); 1153 1154 _cm->do_yield_check(worker_id); 1155 1156 jlong sleep_time_ms; 1157 if (!_cm->has_aborted() && the_task->has_aborted()) { 1158 sleep_time_ms = 1159 (jlong) (elapsed_vtime_sec * _cm->sleep_factor() * 1000.0); 1160 SuspendibleThreadSet::leave(); 1161 os::sleep(Thread::current(), sleep_time_ms, false); 1162 SuspendibleThreadSet::join(); 1163 } 1164 } while (!_cm->has_aborted() && the_task->has_aborted()); 1165 } 1166 the_task->record_end_time(); 1167 guarantee(!the_task->has_aborted() || _cm->has_aborted(), "invariant"); 1168 1169 SuspendibleThreadSet::leave(); 1170 1171 double end_vtime = os::elapsedVTime(); 1172 _cm->update_accum_task_vtime(worker_id, end_vtime - start_vtime); 1173 } 1174 1175 CMConcurrentMarkingTask(ConcurrentMark* cm, 1176 ConcurrentMarkThread* cmt) : 1177 AbstractGangTask("Concurrent Mark"), _cm(cm), _cmt(cmt) { } 1178 1179 ~CMConcurrentMarkingTask() { } 1180 }; 1181 1182 // Calculates the number of active workers for a concurrent 1183 // phase. 1184 uint ConcurrentMark::calc_parallel_marking_threads() { 1185 uint n_conc_workers = 0; 1186 if (!UseDynamicNumberOfGCThreads || 1187 (!FLAG_IS_DEFAULT(ConcGCThreads) && 1188 !ForceDynamicNumberOfGCThreads)) { 1189 n_conc_workers = max_parallel_marking_threads(); 1190 } else { 1191 n_conc_workers = 1192 AdaptiveSizePolicy::calc_default_active_workers( 1193 max_parallel_marking_threads(), 1194 1, /* Minimum workers */ 1195 parallel_marking_threads(), 1196 Threads::number_of_non_daemon_threads()); 1197 // Don't scale down "n_conc_workers" by scale_parallel_threads() because 1198 // that scaling has already gone into "_max_parallel_marking_threads". 1199 } 1200 assert(n_conc_workers > 0, "Always need at least 1"); 1201 return n_conc_workers; 1202 } 1203 1204 void ConcurrentMark::scanRootRegion(HeapRegion* hr, uint worker_id) { 1205 // Currently, only survivors can be root regions. 1206 assert(hr->next_top_at_mark_start() == hr->bottom(), "invariant"); 1207 G1RootRegionScanClosure cl(_g1h, this, worker_id); 1208 1209 const uintx interval = PrefetchScanIntervalInBytes; 1210 HeapWord* curr = hr->bottom(); 1211 const HeapWord* end = hr->top(); 1212 while (curr < end) { 1213 Prefetch::read(curr, interval); 1214 oop obj = oop(curr); 1215 int size = obj->oop_iterate(&cl); 1216 assert(size == obj->size(), "sanity"); 1217 curr += size; 1218 } 1219 } 1220 1221 class CMRootRegionScanTask : public AbstractGangTask { 1222 private: 1223 ConcurrentMark* _cm; 1224 1225 public: 1226 CMRootRegionScanTask(ConcurrentMark* cm) : 1227 AbstractGangTask("Root Region Scan"), _cm(cm) { } 1228 1229 void work(uint worker_id) { 1230 assert(Thread::current()->is_ConcurrentGC_thread(), 1231 "this should only be done by a conc GC thread"); 1232 1233 CMRootRegions* root_regions = _cm->root_regions(); 1234 HeapRegion* hr = root_regions->claim_next(); 1235 while (hr != NULL) { 1236 _cm->scanRootRegion(hr, worker_id); 1237 hr = root_regions->claim_next(); 1238 } 1239 } 1240 }; 1241 1242 void ConcurrentMark::scanRootRegions() { 1243 // Start of concurrent marking. 1244 ClassLoaderDataGraph::clear_claimed_marks(); 1245 1246 // scan_in_progress() will have been set to true only if there was 1247 // at least one root region to scan. So, if it's false, we 1248 // should not attempt to do any further work. 1249 if (root_regions()->scan_in_progress()) { 1250 _parallel_marking_threads = calc_parallel_marking_threads(); 1251 assert(parallel_marking_threads() <= max_parallel_marking_threads(), 1252 "Maximum number of marking threads exceeded"); 1253 uint active_workers = MAX2(1U, parallel_marking_threads()); 1254 1255 CMRootRegionScanTask task(this); 1256 _parallel_workers->set_active_workers(active_workers); 1257 _parallel_workers->run_task(&task); 1258 1259 // It's possible that has_aborted() is true here without actually 1260 // aborting the survivor scan earlier. This is OK as it's 1261 // mainly used for sanity checking. 1262 root_regions()->scan_finished(); 1263 } 1264 } 1265 1266 void ConcurrentMark::markFromRoots() { 1267 // we might be tempted to assert that: 1268 // assert(asynch == !SafepointSynchronize::is_at_safepoint(), 1269 // "inconsistent argument?"); 1270 // However that wouldn't be right, because it's possible that 1271 // a safepoint is indeed in progress as a younger generation 1272 // stop-the-world GC happens even as we mark in this generation. 1273 1274 _restart_for_overflow = false; 1275 force_overflow_conc()->init(); 1276 1277 // _g1h has _n_par_threads 1278 _parallel_marking_threads = calc_parallel_marking_threads(); 1279 assert(parallel_marking_threads() <= max_parallel_marking_threads(), 1280 "Maximum number of marking threads exceeded"); 1281 1282 uint active_workers = MAX2(1U, parallel_marking_threads()); 1283 1284 // Parallel task terminator is set in "set_concurrency_and_phase()" 1285 set_concurrency_and_phase(active_workers, true /* concurrent */); 1286 1287 CMConcurrentMarkingTask markingTask(this, cmThread()); 1288 _parallel_workers->set_active_workers(active_workers); 1289 // Don't set _n_par_threads because it affects MT in process_roots() 1290 // and the decisions on that MT processing is made elsewhere. 1291 assert(_parallel_workers->active_workers() > 0, "Should have been set"); 1292 _parallel_workers->run_task(&markingTask); 1293 print_stats(); 1294 } 1295 1296 // Helper class to get rid of some boilerplate code. 1297 class G1CMTraceTime : public GCTraceTime { 1298 static bool doit_and_prepend(bool doit) { 1299 if (doit) { 1300 gclog_or_tty->put(' '); 1301 } 1302 return doit; 1303 } 1304 1305 public: 1306 G1CMTraceTime(const char* title, bool doit) 1307 : GCTraceTime(title, doit_and_prepend(doit), false, G1CollectedHeap::heap()->gc_timer_cm(), 1308 G1CollectedHeap::heap()->concurrent_mark()->concurrent_gc_id()) { 1309 } 1310 }; 1311 1312 void ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) { 1313 // world is stopped at this checkpoint 1314 assert(SafepointSynchronize::is_at_safepoint(), 1315 "world should be stopped"); 1316 1317 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 1318 1319 // If a full collection has happened, we shouldn't do this. 1320 if (has_aborted()) { 1321 g1h->set_marking_complete(); // So bitmap clearing isn't confused 1322 return; 1323 } 1324 1325 SvcGCMarker sgcm(SvcGCMarker::OTHER); 1326 1327 if (VerifyDuringGC) { 1328 HandleMark hm; // handle scope 1329 Universe::heap()->prepare_for_verify(); 1330 Universe::verify(VerifyOption_G1UsePrevMarking, 1331 " VerifyDuringGC:(before)"); 1332 } 1333 g1h->check_bitmaps("Remark Start"); 1334 1335 G1CollectorPolicy* g1p = g1h->g1_policy(); 1336 g1p->record_concurrent_mark_remark_start(); 1337 1338 double start = os::elapsedTime(); 1339 1340 checkpointRootsFinalWork(); 1341 1342 double mark_work_end = os::elapsedTime(); 1343 1344 weakRefsWork(clear_all_soft_refs); 1345 1346 if (has_overflown()) { 1347 // Oops. We overflowed. Restart concurrent marking. 1348 _restart_for_overflow = true; 1349 if (G1TraceMarkStackOverflow) { 1350 gclog_or_tty->print_cr("\nRemark led to restart for overflow."); 1351 } 1352 1353 // Verify the heap w.r.t. the previous marking bitmap. 1354 if (VerifyDuringGC) { 1355 HandleMark hm; // handle scope 1356 Universe::heap()->prepare_for_verify(); 1357 Universe::verify(VerifyOption_G1UsePrevMarking, 1358 " VerifyDuringGC:(overflow)"); 1359 } 1360 1361 // Clear the marking state because we will be restarting 1362 // marking due to overflowing the global mark stack. 1363 reset_marking_state(); 1364 } else { 1365 { 1366 G1CMTraceTime trace("GC aggregate-data", G1Log::finer()); 1367 1368 // Aggregate the per-task counting data that we have accumulated 1369 // while marking. 1370 aggregate_count_data(); 1371 } 1372 1373 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 1374 // We're done with marking. 1375 // This is the end of the marking cycle, we're expected all 1376 // threads to have SATB queues with active set to true. 1377 satb_mq_set.set_active_all_threads(false, /* new active value */ 1378 true /* expected_active */); 1379 1380 if (VerifyDuringGC) { 1381 HandleMark hm; // handle scope 1382 Universe::heap()->prepare_for_verify(); 1383 Universe::verify(VerifyOption_G1UseNextMarking, 1384 " VerifyDuringGC:(after)"); 1385 } 1386 g1h->check_bitmaps("Remark End"); 1387 assert(!restart_for_overflow(), "sanity"); 1388 // Completely reset the marking state since marking completed 1389 set_non_marking_state(); 1390 } 1391 1392 // Expand the marking stack, if we have to and if we can. 1393 if (_markStack.should_expand()) { 1394 _markStack.expand(); 1395 } 1396 1397 // Statistics 1398 double now = os::elapsedTime(); 1399 _remark_mark_times.add((mark_work_end - start) * 1000.0); 1400 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0); 1401 _remark_times.add((now - start) * 1000.0); 1402 1403 g1p->record_concurrent_mark_remark_end(); 1404 1405 G1CMIsAliveClosure is_alive(g1h); 1406 g1h->gc_tracer_cm()->report_object_count_after_gc(&is_alive); 1407 } 1408 1409 // Base class of the closures that finalize and verify the 1410 // liveness counting data. 1411 class CMCountDataClosureBase: public HeapRegionClosure { 1412 protected: 1413 G1CollectedHeap* _g1h; 1414 ConcurrentMark* _cm; 1415 CardTableModRefBS* _ct_bs; 1416 1417 BitMap* _region_bm; 1418 BitMap* _card_bm; 1419 1420 // Takes a region that's not empty (i.e., it has at least one 1421 // live object in it and sets its corresponding bit on the region 1422 // bitmap to 1. If the region is "starts humongous" it will also set 1423 // to 1 the bits on the region bitmap that correspond to its 1424 // associated "continues humongous" regions. 1425 void set_bit_for_region(HeapRegion* hr) { 1426 assert(!hr->is_continues_humongous(), "should have filtered those out"); 1427 1428 BitMap::idx_t index = (BitMap::idx_t) hr->hrm_index(); 1429 if (!hr->is_starts_humongous()) { 1430 // Normal (non-humongous) case: just set the bit. 1431 _region_bm->par_at_put(index, true); 1432 } else { 1433 // Starts humongous case: calculate how many regions are part of 1434 // this humongous region and then set the bit range. 1435 BitMap::idx_t end_index = (BitMap::idx_t) hr->last_hc_index(); 1436 _region_bm->par_at_put_range(index, end_index, true); 1437 } 1438 } 1439 1440 public: 1441 CMCountDataClosureBase(G1CollectedHeap* g1h, 1442 BitMap* region_bm, BitMap* card_bm): 1443 _g1h(g1h), _cm(g1h->concurrent_mark()), 1444 _ct_bs(barrier_set_cast<CardTableModRefBS>(g1h->barrier_set())), 1445 _region_bm(region_bm), _card_bm(card_bm) { } 1446 }; 1447 1448 // Closure that calculates the # live objects per region. Used 1449 // for verification purposes during the cleanup pause. 1450 class CalcLiveObjectsClosure: public CMCountDataClosureBase { 1451 CMBitMapRO* _bm; 1452 size_t _region_marked_bytes; 1453 1454 public: 1455 CalcLiveObjectsClosure(CMBitMapRO *bm, G1CollectedHeap* g1h, 1456 BitMap* region_bm, BitMap* card_bm) : 1457 CMCountDataClosureBase(g1h, region_bm, card_bm), 1458 _bm(bm), _region_marked_bytes(0) { } 1459 1460 bool doHeapRegion(HeapRegion* hr) { 1461 1462 if (hr->is_continues_humongous()) { 1463 // We will ignore these here and process them when their 1464 // associated "starts humongous" region is processed (see 1465 // set_bit_for_heap_region()). Note that we cannot rely on their 1466 // associated "starts humongous" region to have their bit set to 1467 // 1 since, due to the region chunking in the parallel region 1468 // iteration, a "continues humongous" region might be visited 1469 // before its associated "starts humongous". 1470 return false; 1471 } 1472 1473 HeapWord* ntams = hr->next_top_at_mark_start(); 1474 HeapWord* start = hr->bottom(); 1475 1476 assert(start <= hr->end() && start <= ntams && ntams <= hr->end(), 1477 err_msg("Preconditions not met - " 1478 "start: "PTR_FORMAT", ntams: "PTR_FORMAT", end: "PTR_FORMAT, 1479 p2i(start), p2i(ntams), p2i(hr->end()))); 1480 1481 // Find the first marked object at or after "start". 1482 start = _bm->getNextMarkedWordAddress(start, ntams); 1483 1484 size_t marked_bytes = 0; 1485 1486 while (start < ntams) { 1487 oop obj = oop(start); 1488 int obj_sz = obj->size(); 1489 HeapWord* obj_end = start + obj_sz; 1490 1491 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start); 1492 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(obj_end); 1493 1494 // Note: if we're looking at the last region in heap - obj_end 1495 // could be actually just beyond the end of the heap; end_idx 1496 // will then correspond to a (non-existent) card that is also 1497 // just beyond the heap. 1498 if (_g1h->is_in_g1_reserved(obj_end) && !_ct_bs->is_card_aligned(obj_end)) { 1499 // end of object is not card aligned - increment to cover 1500 // all the cards spanned by the object 1501 end_idx += 1; 1502 } 1503 1504 // Set the bits in the card BM for the cards spanned by this object. 1505 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */); 1506 1507 // Add the size of this object to the number of marked bytes. 1508 marked_bytes += (size_t)obj_sz * HeapWordSize; 1509 1510 // Find the next marked object after this one. 1511 start = _bm->getNextMarkedWordAddress(obj_end, ntams); 1512 } 1513 1514 // Mark the allocated-since-marking portion... 1515 HeapWord* top = hr->top(); 1516 if (ntams < top) { 1517 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams); 1518 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top); 1519 1520 // Note: if we're looking at the last region in heap - top 1521 // could be actually just beyond the end of the heap; end_idx 1522 // will then correspond to a (non-existent) card that is also 1523 // just beyond the heap. 1524 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) { 1525 // end of object is not card aligned - increment to cover 1526 // all the cards spanned by the object 1527 end_idx += 1; 1528 } 1529 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */); 1530 1531 // This definitely means the region has live objects. 1532 set_bit_for_region(hr); 1533 } 1534 1535 // Update the live region bitmap. 1536 if (marked_bytes > 0) { 1537 set_bit_for_region(hr); 1538 } 1539 1540 // Set the marked bytes for the current region so that 1541 // it can be queried by a calling verification routine 1542 _region_marked_bytes = marked_bytes; 1543 1544 return false; 1545 } 1546 1547 size_t region_marked_bytes() const { return _region_marked_bytes; } 1548 }; 1549 1550 // Heap region closure used for verifying the counting data 1551 // that was accumulated concurrently and aggregated during 1552 // the remark pause. This closure is applied to the heap 1553 // regions during the STW cleanup pause. 1554 1555 class VerifyLiveObjectDataHRClosure: public HeapRegionClosure { 1556 G1CollectedHeap* _g1h; 1557 ConcurrentMark* _cm; 1558 CalcLiveObjectsClosure _calc_cl; 1559 BitMap* _region_bm; // Region BM to be verified 1560 BitMap* _card_bm; // Card BM to be verified 1561 bool _verbose; // verbose output? 1562 1563 BitMap* _exp_region_bm; // Expected Region BM values 1564 BitMap* _exp_card_bm; // Expected card BM values 1565 1566 int _failures; 1567 1568 public: 1569 VerifyLiveObjectDataHRClosure(G1CollectedHeap* g1h, 1570 BitMap* region_bm, 1571 BitMap* card_bm, 1572 BitMap* exp_region_bm, 1573 BitMap* exp_card_bm, 1574 bool verbose) : 1575 _g1h(g1h), _cm(g1h->concurrent_mark()), 1576 _calc_cl(_cm->nextMarkBitMap(), g1h, exp_region_bm, exp_card_bm), 1577 _region_bm(region_bm), _card_bm(card_bm), _verbose(verbose), 1578 _exp_region_bm(exp_region_bm), _exp_card_bm(exp_card_bm), 1579 _failures(0) { } 1580 1581 int failures() const { return _failures; } 1582 1583 bool doHeapRegion(HeapRegion* hr) { 1584 if (hr->is_continues_humongous()) { 1585 // We will ignore these here and process them when their 1586 // associated "starts humongous" region is processed (see 1587 // set_bit_for_heap_region()). Note that we cannot rely on their 1588 // associated "starts humongous" region to have their bit set to 1589 // 1 since, due to the region chunking in the parallel region 1590 // iteration, a "continues humongous" region might be visited 1591 // before its associated "starts humongous". 1592 return false; 1593 } 1594 1595 int failures = 0; 1596 1597 // Call the CalcLiveObjectsClosure to walk the marking bitmap for 1598 // this region and set the corresponding bits in the expected region 1599 // and card bitmaps. 1600 bool res = _calc_cl.doHeapRegion(hr); 1601 assert(res == false, "should be continuing"); 1602 1603 MutexLockerEx x((_verbose ? ParGCRareEvent_lock : NULL), 1604 Mutex::_no_safepoint_check_flag); 1605 1606 // Verify the marked bytes for this region. 1607 size_t exp_marked_bytes = _calc_cl.region_marked_bytes(); 1608 size_t act_marked_bytes = hr->next_marked_bytes(); 1609 1610 // We're not OK if expected marked bytes > actual marked bytes. It means 1611 // we have missed accounting some objects during the actual marking. 1612 if (exp_marked_bytes > act_marked_bytes) { 1613 if (_verbose) { 1614 gclog_or_tty->print_cr("Region %u: marked bytes mismatch: " 1615 "expected: " SIZE_FORMAT ", actual: " SIZE_FORMAT, 1616 hr->hrm_index(), exp_marked_bytes, act_marked_bytes); 1617 } 1618 failures += 1; 1619 } 1620 1621 // Verify the bit, for this region, in the actual and expected 1622 // (which was just calculated) region bit maps. 1623 // We're not OK if the bit in the calculated expected region 1624 // bitmap is set and the bit in the actual region bitmap is not. 1625 BitMap::idx_t index = (BitMap::idx_t) hr->hrm_index(); 1626 1627 bool expected = _exp_region_bm->at(index); 1628 bool actual = _region_bm->at(index); 1629 if (expected && !actual) { 1630 if (_verbose) { 1631 gclog_or_tty->print_cr("Region %u: region bitmap mismatch: " 1632 "expected: %s, actual: %s", 1633 hr->hrm_index(), 1634 BOOL_TO_STR(expected), BOOL_TO_STR(actual)); 1635 } 1636 failures += 1; 1637 } 1638 1639 // Verify that the card bit maps for the cards spanned by the current 1640 // region match. We have an error if we have a set bit in the expected 1641 // bit map and the corresponding bit in the actual bitmap is not set. 1642 1643 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(hr->bottom()); 1644 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(hr->top()); 1645 1646 for (BitMap::idx_t i = start_idx; i < end_idx; i+=1) { 1647 expected = _exp_card_bm->at(i); 1648 actual = _card_bm->at(i); 1649 1650 if (expected && !actual) { 1651 if (_verbose) { 1652 gclog_or_tty->print_cr("Region %u: card bitmap mismatch at " SIZE_FORMAT ": " 1653 "expected: %s, actual: %s", 1654 hr->hrm_index(), i, 1655 BOOL_TO_STR(expected), BOOL_TO_STR(actual)); 1656 } 1657 failures += 1; 1658 } 1659 } 1660 1661 if (failures > 0 && _verbose) { 1662 gclog_or_tty->print_cr("Region " HR_FORMAT ", ntams: " PTR_FORMAT ", " 1663 "marked_bytes: calc/actual " SIZE_FORMAT "/" SIZE_FORMAT, 1664 HR_FORMAT_PARAMS(hr), p2i(hr->next_top_at_mark_start()), 1665 _calc_cl.region_marked_bytes(), hr->next_marked_bytes()); 1666 } 1667 1668 _failures += failures; 1669 1670 // We could stop iteration over the heap when we 1671 // find the first violating region by returning true. 1672 return false; 1673 } 1674 }; 1675 1676 class G1ParVerifyFinalCountTask: public AbstractGangTask { 1677 protected: 1678 G1CollectedHeap* _g1h; 1679 ConcurrentMark* _cm; 1680 BitMap* _actual_region_bm; 1681 BitMap* _actual_card_bm; 1682 1683 uint _n_workers; 1684 1685 BitMap* _expected_region_bm; 1686 BitMap* _expected_card_bm; 1687 1688 int _failures; 1689 bool _verbose; 1690 1691 HeapRegionClaimer _hrclaimer; 1692 1693 public: 1694 G1ParVerifyFinalCountTask(G1CollectedHeap* g1h, 1695 BitMap* region_bm, BitMap* card_bm, 1696 BitMap* expected_region_bm, BitMap* expected_card_bm) 1697 : AbstractGangTask("G1 verify final counting"), 1698 _g1h(g1h), _cm(_g1h->concurrent_mark()), 1699 _actual_region_bm(region_bm), _actual_card_bm(card_bm), 1700 _expected_region_bm(expected_region_bm), _expected_card_bm(expected_card_bm), 1701 _failures(0), _verbose(false), 1702 _n_workers(_g1h->workers()->active_workers()), _hrclaimer(_n_workers) { 1703 assert(VerifyDuringGC, "don't call this otherwise"); 1704 assert(_expected_card_bm->size() == _actual_card_bm->size(), "sanity"); 1705 assert(_expected_region_bm->size() == _actual_region_bm->size(), "sanity"); 1706 1707 _verbose = _cm->verbose_medium(); 1708 } 1709 1710 void work(uint worker_id) { 1711 assert(worker_id < _n_workers, "invariant"); 1712 1713 VerifyLiveObjectDataHRClosure verify_cl(_g1h, 1714 _actual_region_bm, _actual_card_bm, 1715 _expected_region_bm, 1716 _expected_card_bm, 1717 _verbose); 1718 1719 _g1h->heap_region_par_iterate(&verify_cl, worker_id, &_hrclaimer); 1720 1721 Atomic::add(verify_cl.failures(), &_failures); 1722 } 1723 1724 int failures() const { return _failures; } 1725 }; 1726 1727 // Closure that finalizes the liveness counting data. 1728 // Used during the cleanup pause. 1729 // Sets the bits corresponding to the interval [NTAMS, top] 1730 // (which contains the implicitly live objects) in the 1731 // card liveness bitmap. Also sets the bit for each region, 1732 // containing live data, in the region liveness bitmap. 1733 1734 class FinalCountDataUpdateClosure: public CMCountDataClosureBase { 1735 public: 1736 FinalCountDataUpdateClosure(G1CollectedHeap* g1h, 1737 BitMap* region_bm, 1738 BitMap* card_bm) : 1739 CMCountDataClosureBase(g1h, region_bm, card_bm) { } 1740 1741 bool doHeapRegion(HeapRegion* hr) { 1742 1743 if (hr->is_continues_humongous()) { 1744 // We will ignore these here and process them when their 1745 // associated "starts humongous" region is processed (see 1746 // set_bit_for_heap_region()). Note that we cannot rely on their 1747 // associated "starts humongous" region to have their bit set to 1748 // 1 since, due to the region chunking in the parallel region 1749 // iteration, a "continues humongous" region might be visited 1750 // before its associated "starts humongous". 1751 return false; 1752 } 1753 1754 HeapWord* ntams = hr->next_top_at_mark_start(); 1755 HeapWord* top = hr->top(); 1756 1757 assert(hr->bottom() <= ntams && ntams <= hr->end(), "Preconditions."); 1758 1759 // Mark the allocated-since-marking portion... 1760 if (ntams < top) { 1761 // This definitely means the region has live objects. 1762 set_bit_for_region(hr); 1763 1764 // Now set the bits in the card bitmap for [ntams, top) 1765 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams); 1766 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top); 1767 1768 // Note: if we're looking at the last region in heap - top 1769 // could be actually just beyond the end of the heap; end_idx 1770 // will then correspond to a (non-existent) card that is also 1771 // just beyond the heap. 1772 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) { 1773 // end of object is not card aligned - increment to cover 1774 // all the cards spanned by the object 1775 end_idx += 1; 1776 } 1777 1778 assert(end_idx <= _card_bm->size(), 1779 err_msg("oob: end_idx= "SIZE_FORMAT", bitmap size= "SIZE_FORMAT, 1780 end_idx, _card_bm->size())); 1781 assert(start_idx < _card_bm->size(), 1782 err_msg("oob: start_idx= "SIZE_FORMAT", bitmap size= "SIZE_FORMAT, 1783 start_idx, _card_bm->size())); 1784 1785 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */); 1786 } 1787 1788 // Set the bit for the region if it contains live data 1789 if (hr->next_marked_bytes() > 0) { 1790 set_bit_for_region(hr); 1791 } 1792 1793 return false; 1794 } 1795 }; 1796 1797 class G1ParFinalCountTask: public AbstractGangTask { 1798 protected: 1799 G1CollectedHeap* _g1h; 1800 ConcurrentMark* _cm; 1801 BitMap* _actual_region_bm; 1802 BitMap* _actual_card_bm; 1803 1804 uint _n_workers; 1805 HeapRegionClaimer _hrclaimer; 1806 1807 public: 1808 G1ParFinalCountTask(G1CollectedHeap* g1h, BitMap* region_bm, BitMap* card_bm) 1809 : AbstractGangTask("G1 final counting"), 1810 _g1h(g1h), _cm(_g1h->concurrent_mark()), 1811 _actual_region_bm(region_bm), _actual_card_bm(card_bm), 1812 _n_workers(_g1h->workers()->active_workers()), _hrclaimer(_n_workers) { 1813 } 1814 1815 void work(uint worker_id) { 1816 assert(worker_id < _n_workers, "invariant"); 1817 1818 FinalCountDataUpdateClosure final_update_cl(_g1h, 1819 _actual_region_bm, 1820 _actual_card_bm); 1821 1822 _g1h->heap_region_par_iterate(&final_update_cl, worker_id, &_hrclaimer); 1823 } 1824 }; 1825 1826 class G1ParNoteEndTask; 1827 1828 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure { 1829 G1CollectedHeap* _g1; 1830 size_t _max_live_bytes; 1831 uint _regions_claimed; 1832 size_t _freed_bytes; 1833 FreeRegionList* _local_cleanup_list; 1834 HeapRegionSetCount _old_regions_removed; 1835 HeapRegionSetCount _humongous_regions_removed; 1836 HRRSCleanupTask* _hrrs_cleanup_task; 1837 double _claimed_region_time; 1838 double _max_region_time; 1839 1840 public: 1841 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1, 1842 FreeRegionList* local_cleanup_list, 1843 HRRSCleanupTask* hrrs_cleanup_task) : 1844 _g1(g1), 1845 _max_live_bytes(0), _regions_claimed(0), 1846 _freed_bytes(0), 1847 _claimed_region_time(0.0), _max_region_time(0.0), 1848 _local_cleanup_list(local_cleanup_list), 1849 _old_regions_removed(), 1850 _humongous_regions_removed(), 1851 _hrrs_cleanup_task(hrrs_cleanup_task) { } 1852 1853 size_t freed_bytes() { return _freed_bytes; } 1854 const HeapRegionSetCount& old_regions_removed() { return _old_regions_removed; } 1855 const HeapRegionSetCount& humongous_regions_removed() { return _humongous_regions_removed; } 1856 1857 bool doHeapRegion(HeapRegion *hr) { 1858 if (hr->is_continues_humongous()) { 1859 return false; 1860 } 1861 // We use a claim value of zero here because all regions 1862 // were claimed with value 1 in the FinalCount task. 1863 _g1->reset_gc_time_stamps(hr); 1864 double start = os::elapsedTime(); 1865 _regions_claimed++; 1866 hr->note_end_of_marking(); 1867 _max_live_bytes += hr->max_live_bytes(); 1868 1869 if (hr->used() > 0 && hr->max_live_bytes() == 0 && !hr->is_young()) { 1870 _freed_bytes += hr->used(); 1871 hr->set_containing_set(NULL); 1872 if (hr->is_humongous()) { 1873 assert(hr->is_starts_humongous(), "we should only see starts humongous"); 1874 _humongous_regions_removed.increment(1u, hr->capacity()); 1875 _g1->free_humongous_region(hr, _local_cleanup_list, true); 1876 } else { 1877 _old_regions_removed.increment(1u, hr->capacity()); 1878 _g1->free_region(hr, _local_cleanup_list, true); 1879 } 1880 } else { 1881 hr->rem_set()->do_cleanup_work(_hrrs_cleanup_task); 1882 } 1883 1884 double region_time = (os::elapsedTime() - start); 1885 _claimed_region_time += region_time; 1886 if (region_time > _max_region_time) { 1887 _max_region_time = region_time; 1888 } 1889 return false; 1890 } 1891 1892 size_t max_live_bytes() { return _max_live_bytes; } 1893 uint regions_claimed() { return _regions_claimed; } 1894 double claimed_region_time_sec() { return _claimed_region_time; } 1895 double max_region_time_sec() { return _max_region_time; } 1896 }; 1897 1898 class G1ParNoteEndTask: public AbstractGangTask { 1899 friend class G1NoteEndOfConcMarkClosure; 1900 1901 protected: 1902 G1CollectedHeap* _g1h; 1903 size_t _max_live_bytes; 1904 size_t _freed_bytes; 1905 FreeRegionList* _cleanup_list; 1906 HeapRegionClaimer _hrclaimer; 1907 1908 public: 1909 G1ParNoteEndTask(G1CollectedHeap* g1h, FreeRegionList* cleanup_list, uint n_workers) : 1910 AbstractGangTask("G1 note end"), _g1h(g1h), _max_live_bytes(0), _freed_bytes(0), _cleanup_list(cleanup_list), _hrclaimer(n_workers) { 1911 } 1912 1913 void work(uint worker_id) { 1914 FreeRegionList local_cleanup_list("Local Cleanup List"); 1915 HRRSCleanupTask hrrs_cleanup_task; 1916 G1NoteEndOfConcMarkClosure g1_note_end(_g1h, &local_cleanup_list, 1917 &hrrs_cleanup_task); 1918 _g1h->heap_region_par_iterate(&g1_note_end, worker_id, &_hrclaimer); 1919 assert(g1_note_end.complete(), "Shouldn't have yielded!"); 1920 1921 // Now update the lists 1922 _g1h->remove_from_old_sets(g1_note_end.old_regions_removed(), g1_note_end.humongous_regions_removed()); 1923 { 1924 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); 1925 _g1h->decrement_summary_bytes(g1_note_end.freed_bytes()); 1926 _max_live_bytes += g1_note_end.max_live_bytes(); 1927 _freed_bytes += g1_note_end.freed_bytes(); 1928 1929 // If we iterate over the global cleanup list at the end of 1930 // cleanup to do this printing we will not guarantee to only 1931 // generate output for the newly-reclaimed regions (the list 1932 // might not be empty at the beginning of cleanup; we might 1933 // still be working on its previous contents). So we do the 1934 // printing here, before we append the new regions to the global 1935 // cleanup list. 1936 1937 G1HRPrinter* hr_printer = _g1h->hr_printer(); 1938 if (hr_printer->is_active()) { 1939 FreeRegionListIterator iter(&local_cleanup_list); 1940 while (iter.more_available()) { 1941 HeapRegion* hr = iter.get_next(); 1942 hr_printer->cleanup(hr); 1943 } 1944 } 1945 1946 _cleanup_list->add_ordered(&local_cleanup_list); 1947 assert(local_cleanup_list.is_empty(), "post-condition"); 1948 1949 HeapRegionRemSet::finish_cleanup_task(&hrrs_cleanup_task); 1950 } 1951 } 1952 size_t max_live_bytes() { return _max_live_bytes; } 1953 size_t freed_bytes() { return _freed_bytes; } 1954 }; 1955 1956 class G1ParScrubRemSetTask: public AbstractGangTask { 1957 protected: 1958 G1RemSet* _g1rs; 1959 BitMap* _region_bm; 1960 BitMap* _card_bm; 1961 HeapRegionClaimer _hrclaimer; 1962 1963 public: 1964 G1ParScrubRemSetTask(G1CollectedHeap* g1h, BitMap* region_bm, BitMap* card_bm, uint n_workers) : 1965 AbstractGangTask("G1 ScrubRS"), _g1rs(g1h->g1_rem_set()), _region_bm(region_bm), _card_bm(card_bm), _hrclaimer(n_workers) { 1966 } 1967 1968 void work(uint worker_id) { 1969 _g1rs->scrub(_region_bm, _card_bm, worker_id, &_hrclaimer); 1970 } 1971 1972 }; 1973 1974 void ConcurrentMark::cleanup() { 1975 // world is stopped at this checkpoint 1976 assert(SafepointSynchronize::is_at_safepoint(), 1977 "world should be stopped"); 1978 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 1979 1980 // If a full collection has happened, we shouldn't do this. 1981 if (has_aborted()) { 1982 g1h->set_marking_complete(); // So bitmap clearing isn't confused 1983 return; 1984 } 1985 1986 g1h->verify_region_sets_optional(); 1987 1988 if (VerifyDuringGC) { 1989 HandleMark hm; // handle scope 1990 Universe::heap()->prepare_for_verify(); 1991 Universe::verify(VerifyOption_G1UsePrevMarking, 1992 " VerifyDuringGC:(before)"); 1993 } 1994 g1h->check_bitmaps("Cleanup Start"); 1995 1996 G1CollectorPolicy* g1p = G1CollectedHeap::heap()->g1_policy(); 1997 g1p->record_concurrent_mark_cleanup_start(); 1998 1999 double start = os::elapsedTime(); 2000 2001 HeapRegionRemSet::reset_for_cleanup_tasks(); 2002 2003 uint n_workers; 2004 2005 // Do counting once more with the world stopped for good measure. 2006 G1ParFinalCountTask g1_par_count_task(g1h, &_region_bm, &_card_bm); 2007 2008 g1h->set_par_threads(); 2009 n_workers = g1h->n_par_threads(); 2010 assert(g1h->n_par_threads() == n_workers, 2011 "Should not have been reset"); 2012 g1h->workers()->run_task(&g1_par_count_task); 2013 // Done with the parallel phase so reset to 0. 2014 g1h->set_par_threads(0); 2015 2016 if (VerifyDuringGC) { 2017 // Verify that the counting data accumulated during marking matches 2018 // that calculated by walking the marking bitmap. 2019 2020 // Bitmaps to hold expected values 2021 BitMap expected_region_bm(_region_bm.size(), true); 2022 BitMap expected_card_bm(_card_bm.size(), true); 2023 2024 G1ParVerifyFinalCountTask g1_par_verify_task(g1h, 2025 &_region_bm, 2026 &_card_bm, 2027 &expected_region_bm, 2028 &expected_card_bm); 2029 2030 g1h->set_par_threads((int)n_workers); 2031 g1h->workers()->run_task(&g1_par_verify_task); 2032 // Done with the parallel phase so reset to 0. 2033 g1h->set_par_threads(0); 2034 2035 guarantee(g1_par_verify_task.failures() == 0, "Unexpected accounting failures"); 2036 } 2037 2038 size_t start_used_bytes = g1h->used(); 2039 g1h->set_marking_complete(); 2040 2041 double count_end = os::elapsedTime(); 2042 double this_final_counting_time = (count_end - start); 2043 _total_counting_time += this_final_counting_time; 2044 2045 if (G1PrintRegionLivenessInfo) { 2046 G1PrintRegionLivenessInfoClosure cl(gclog_or_tty, "Post-Marking"); 2047 _g1h->heap_region_iterate(&cl); 2048 } 2049 2050 // Install newly created mark bitMap as "prev". 2051 swapMarkBitMaps(); 2052 2053 g1h->reset_gc_time_stamp(); 2054 2055 // Note end of marking in all heap regions. 2056 G1ParNoteEndTask g1_par_note_end_task(g1h, &_cleanup_list, n_workers); 2057 g1h->set_par_threads((int)n_workers); 2058 g1h->workers()->run_task(&g1_par_note_end_task); 2059 g1h->set_par_threads(0); 2060 g1h->check_gc_time_stamps(); 2061 2062 if (!cleanup_list_is_empty()) { 2063 // The cleanup list is not empty, so we'll have to process it 2064 // concurrently. Notify anyone else that might be wanting free 2065 // regions that there will be more free regions coming soon. 2066 g1h->set_free_regions_coming(); 2067 } 2068 2069 // call below, since it affects the metric by which we sort the heap 2070 // regions. 2071 if (G1ScrubRemSets) { 2072 double rs_scrub_start = os::elapsedTime(); 2073 G1ParScrubRemSetTask g1_par_scrub_rs_task(g1h, &_region_bm, &_card_bm, n_workers); 2074 g1h->set_par_threads((int)n_workers); 2075 g1h->workers()->run_task(&g1_par_scrub_rs_task); 2076 g1h->set_par_threads(0); 2077 2078 double rs_scrub_end = os::elapsedTime(); 2079 double this_rs_scrub_time = (rs_scrub_end - rs_scrub_start); 2080 _total_rs_scrub_time += this_rs_scrub_time; 2081 } 2082 2083 // this will also free any regions totally full of garbage objects, 2084 // and sort the regions. 2085 g1h->g1_policy()->record_concurrent_mark_cleanup_end((int)n_workers); 2086 2087 // Statistics. 2088 double end = os::elapsedTime(); 2089 _cleanup_times.add((end - start) * 1000.0); 2090 2091 if (G1Log::fine()) { 2092 g1h->g1_policy()->print_heap_transition(start_used_bytes); 2093 } 2094 2095 // Clean up will have freed any regions completely full of garbage. 2096 // Update the soft reference policy with the new heap occupancy. 2097 Universe::update_heap_info_at_gc(); 2098 2099 if (VerifyDuringGC) { 2100 HandleMark hm; // handle scope 2101 Universe::heap()->prepare_for_verify(); 2102 Universe::verify(VerifyOption_G1UsePrevMarking, 2103 " VerifyDuringGC:(after)"); 2104 } 2105 2106 g1h->check_bitmaps("Cleanup End"); 2107 2108 g1h->verify_region_sets_optional(); 2109 2110 // We need to make this be a "collection" so any collection pause that 2111 // races with it goes around and waits for completeCleanup to finish. 2112 g1h->increment_total_collections(); 2113 2114 // Clean out dead classes and update Metaspace sizes. 2115 if (ClassUnloadingWithConcurrentMark) { 2116 ClassLoaderDataGraph::purge(); 2117 } 2118 MetaspaceGC::compute_new_size(); 2119 2120 // We reclaimed old regions so we should calculate the sizes to make 2121 // sure we update the old gen/space data. 2122 g1h->g1mm()->update_sizes(); 2123 g1h->allocation_context_stats().update_after_mark(); 2124 2125 g1h->trace_heap_after_concurrent_cycle(); 2126 } 2127 2128 void ConcurrentMark::completeCleanup() { 2129 if (has_aborted()) return; 2130 2131 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 2132 2133 _cleanup_list.verify_optional(); 2134 FreeRegionList tmp_free_list("Tmp Free List"); 2135 2136 if (G1ConcRegionFreeingVerbose) { 2137 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : " 2138 "cleanup list has %u entries", 2139 _cleanup_list.length()); 2140 } 2141 2142 // No one else should be accessing the _cleanup_list at this point, 2143 // so it is not necessary to take any locks 2144 while (!_cleanup_list.is_empty()) { 2145 HeapRegion* hr = _cleanup_list.remove_region(true /* from_head */); 2146 assert(hr != NULL, "Got NULL from a non-empty list"); 2147 hr->par_clear(); 2148 tmp_free_list.add_ordered(hr); 2149 2150 // Instead of adding one region at a time to the secondary_free_list, 2151 // we accumulate them in the local list and move them a few at a 2152 // time. This also cuts down on the number of notify_all() calls 2153 // we do during this process. We'll also append the local list when 2154 // _cleanup_list is empty (which means we just removed the last 2155 // region from the _cleanup_list). 2156 if ((tmp_free_list.length() % G1SecondaryFreeListAppendLength == 0) || 2157 _cleanup_list.is_empty()) { 2158 if (G1ConcRegionFreeingVerbose) { 2159 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : " 2160 "appending %u entries to the secondary_free_list, " 2161 "cleanup list still has %u entries", 2162 tmp_free_list.length(), 2163 _cleanup_list.length()); 2164 } 2165 2166 { 2167 MutexLockerEx x(SecondaryFreeList_lock, Mutex::_no_safepoint_check_flag); 2168 g1h->secondary_free_list_add(&tmp_free_list); 2169 SecondaryFreeList_lock->notify_all(); 2170 } 2171 #ifndef PRODUCT 2172 if (G1StressConcRegionFreeing) { 2173 for (uintx i = 0; i < G1StressConcRegionFreeingDelayMillis; ++i) { 2174 os::sleep(Thread::current(), (jlong) 1, false); 2175 } 2176 } 2177 #endif 2178 } 2179 } 2180 assert(tmp_free_list.is_empty(), "post-condition"); 2181 } 2182 2183 // Supporting Object and Oop closures for reference discovery 2184 // and processing in during marking 2185 2186 bool G1CMIsAliveClosure::do_object_b(oop obj) { 2187 HeapWord* addr = (HeapWord*)obj; 2188 return addr != NULL && 2189 (!_g1->is_in_g1_reserved(addr) || !_g1->is_obj_ill(obj)); 2190 } 2191 2192 // 'Keep Alive' oop closure used by both serial parallel reference processing. 2193 // Uses the CMTask associated with a worker thread (for serial reference 2194 // processing the CMTask for worker 0 is used) to preserve (mark) and 2195 // trace referent objects. 2196 // 2197 // Using the CMTask and embedded local queues avoids having the worker 2198 // threads operating on the global mark stack. This reduces the risk 2199 // of overflowing the stack - which we would rather avoid at this late 2200 // state. Also using the tasks' local queues removes the potential 2201 // of the workers interfering with each other that could occur if 2202 // operating on the global stack. 2203 2204 class G1CMKeepAliveAndDrainClosure: public OopClosure { 2205 ConcurrentMark* _cm; 2206 CMTask* _task; 2207 int _ref_counter_limit; 2208 int _ref_counter; 2209 bool _is_serial; 2210 public: 2211 G1CMKeepAliveAndDrainClosure(ConcurrentMark* cm, CMTask* task, bool is_serial) : 2212 _cm(cm), _task(task), _is_serial(is_serial), 2213 _ref_counter_limit(G1RefProcDrainInterval) { 2214 assert(_ref_counter_limit > 0, "sanity"); 2215 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code"); 2216 _ref_counter = _ref_counter_limit; 2217 } 2218 2219 virtual void do_oop(narrowOop* p) { do_oop_work(p); } 2220 virtual void do_oop( oop* p) { do_oop_work(p); } 2221 2222 template <class T> void do_oop_work(T* p) { 2223 if (!_cm->has_overflown()) { 2224 oop obj = oopDesc::load_decode_heap_oop(p); 2225 if (_cm->verbose_high()) { 2226 gclog_or_tty->print_cr("\t[%u] we're looking at location " 2227 "*"PTR_FORMAT" = "PTR_FORMAT, 2228 _task->worker_id(), p2i(p), p2i((void*) obj)); 2229 } 2230 2231 _task->deal_with_reference(obj); 2232 _ref_counter--; 2233 2234 if (_ref_counter == 0) { 2235 // We have dealt with _ref_counter_limit references, pushing them 2236 // and objects reachable from them on to the local stack (and 2237 // possibly the global stack). Call CMTask::do_marking_step() to 2238 // process these entries. 2239 // 2240 // We call CMTask::do_marking_step() in a loop, which we'll exit if 2241 // there's nothing more to do (i.e. we're done with the entries that 2242 // were pushed as a result of the CMTask::deal_with_reference() calls 2243 // above) or we overflow. 2244 // 2245 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted() 2246 // flag while there may still be some work to do. (See the comment at 2247 // the beginning of CMTask::do_marking_step() for those conditions - 2248 // one of which is reaching the specified time target.) It is only 2249 // when CMTask::do_marking_step() returns without setting the 2250 // has_aborted() flag that the marking step has completed. 2251 do { 2252 double mark_step_duration_ms = G1ConcMarkStepDurationMillis; 2253 _task->do_marking_step(mark_step_duration_ms, 2254 false /* do_termination */, 2255 _is_serial); 2256 } while (_task->has_aborted() && !_cm->has_overflown()); 2257 _ref_counter = _ref_counter_limit; 2258 } 2259 } else { 2260 if (_cm->verbose_high()) { 2261 gclog_or_tty->print_cr("\t[%u] CM Overflow", _task->worker_id()); 2262 } 2263 } 2264 } 2265 }; 2266 2267 // 'Drain' oop closure used by both serial and parallel reference processing. 2268 // Uses the CMTask associated with a given worker thread (for serial 2269 // reference processing the CMtask for worker 0 is used). Calls the 2270 // do_marking_step routine, with an unbelievably large timeout value, 2271 // to drain the marking data structures of the remaining entries 2272 // added by the 'keep alive' oop closure above. 2273 2274 class G1CMDrainMarkingStackClosure: public VoidClosure { 2275 ConcurrentMark* _cm; 2276 CMTask* _task; 2277 bool _is_serial; 2278 public: 2279 G1CMDrainMarkingStackClosure(ConcurrentMark* cm, CMTask* task, bool is_serial) : 2280 _cm(cm), _task(task), _is_serial(is_serial) { 2281 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code"); 2282 } 2283 2284 void do_void() { 2285 do { 2286 if (_cm->verbose_high()) { 2287 gclog_or_tty->print_cr("\t[%u] Drain: Calling do_marking_step - serial: %s", 2288 _task->worker_id(), BOOL_TO_STR(_is_serial)); 2289 } 2290 2291 // We call CMTask::do_marking_step() to completely drain the local 2292 // and global marking stacks of entries pushed by the 'keep alive' 2293 // oop closure (an instance of G1CMKeepAliveAndDrainClosure above). 2294 // 2295 // CMTask::do_marking_step() is called in a loop, which we'll exit 2296 // if there's nothing more to do (i.e. we've completely drained the 2297 // entries that were pushed as a a result of applying the 'keep alive' 2298 // closure to the entries on the discovered ref lists) or we overflow 2299 // the global marking stack. 2300 // 2301 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted() 2302 // flag while there may still be some work to do. (See the comment at 2303 // the beginning of CMTask::do_marking_step() for those conditions - 2304 // one of which is reaching the specified time target.) It is only 2305 // when CMTask::do_marking_step() returns without setting the 2306 // has_aborted() flag that the marking step has completed. 2307 2308 _task->do_marking_step(1000000000.0 /* something very large */, 2309 true /* do_termination */, 2310 _is_serial); 2311 } while (_task->has_aborted() && !_cm->has_overflown()); 2312 } 2313 }; 2314 2315 // Implementation of AbstractRefProcTaskExecutor for parallel 2316 // reference processing at the end of G1 concurrent marking 2317 2318 class G1CMRefProcTaskExecutor: public AbstractRefProcTaskExecutor { 2319 private: 2320 G1CollectedHeap* _g1h; 2321 ConcurrentMark* _cm; 2322 WorkGang* _workers; 2323 int _active_workers; 2324 2325 public: 2326 G1CMRefProcTaskExecutor(G1CollectedHeap* g1h, 2327 ConcurrentMark* cm, 2328 WorkGang* workers, 2329 int n_workers) : 2330 _g1h(g1h), _cm(cm), 2331 _workers(workers), _active_workers(n_workers) { } 2332 2333 // Executes the given task using concurrent marking worker threads. 2334 virtual void execute(ProcessTask& task); 2335 virtual void execute(EnqueueTask& task); 2336 }; 2337 2338 class G1CMRefProcTaskProxy: public AbstractGangTask { 2339 typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask; 2340 ProcessTask& _proc_task; 2341 G1CollectedHeap* _g1h; 2342 ConcurrentMark* _cm; 2343 2344 public: 2345 G1CMRefProcTaskProxy(ProcessTask& proc_task, 2346 G1CollectedHeap* g1h, 2347 ConcurrentMark* cm) : 2348 AbstractGangTask("Process reference objects in parallel"), 2349 _proc_task(proc_task), _g1h(g1h), _cm(cm) { 2350 ReferenceProcessor* rp = _g1h->ref_processor_cm(); 2351 assert(rp->processing_is_mt(), "shouldn't be here otherwise"); 2352 } 2353 2354 virtual void work(uint worker_id) { 2355 ResourceMark rm; 2356 HandleMark hm; 2357 CMTask* task = _cm->task(worker_id); 2358 G1CMIsAliveClosure g1_is_alive(_g1h); 2359 G1CMKeepAliveAndDrainClosure g1_par_keep_alive(_cm, task, false /* is_serial */); 2360 G1CMDrainMarkingStackClosure g1_par_drain(_cm, task, false /* is_serial */); 2361 2362 _proc_task.work(worker_id, g1_is_alive, g1_par_keep_alive, g1_par_drain); 2363 } 2364 }; 2365 2366 void G1CMRefProcTaskExecutor::execute(ProcessTask& proc_task) { 2367 assert(_workers != NULL, "Need parallel worker threads."); 2368 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT"); 2369 2370 G1CMRefProcTaskProxy proc_task_proxy(proc_task, _g1h, _cm); 2371 2372 // We need to reset the concurrency level before each 2373 // proxy task execution, so that the termination protocol 2374 // and overflow handling in CMTask::do_marking_step() knows 2375 // how many workers to wait for. 2376 _cm->set_concurrency(_active_workers); 2377 _g1h->set_par_threads(_active_workers); 2378 _workers->run_task(&proc_task_proxy); 2379 _g1h->set_par_threads(0); 2380 } 2381 2382 class G1CMRefEnqueueTaskProxy: public AbstractGangTask { 2383 typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask; 2384 EnqueueTask& _enq_task; 2385 2386 public: 2387 G1CMRefEnqueueTaskProxy(EnqueueTask& enq_task) : 2388 AbstractGangTask("Enqueue reference objects in parallel"), 2389 _enq_task(enq_task) { } 2390 2391 virtual void work(uint worker_id) { 2392 _enq_task.work(worker_id); 2393 } 2394 }; 2395 2396 void G1CMRefProcTaskExecutor::execute(EnqueueTask& enq_task) { 2397 assert(_workers != NULL, "Need parallel worker threads."); 2398 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT"); 2399 2400 G1CMRefEnqueueTaskProxy enq_task_proxy(enq_task); 2401 2402 // Not strictly necessary but... 2403 // 2404 // We need to reset the concurrency level before each 2405 // proxy task execution, so that the termination protocol 2406 // and overflow handling in CMTask::do_marking_step() knows 2407 // how many workers to wait for. 2408 _cm->set_concurrency(_active_workers); 2409 _g1h->set_par_threads(_active_workers); 2410 _workers->run_task(&enq_task_proxy); 2411 _g1h->set_par_threads(0); 2412 } 2413 2414 void ConcurrentMark::weakRefsWorkParallelPart(BoolObjectClosure* is_alive, bool purged_classes) { 2415 G1CollectedHeap::heap()->parallel_cleaning(is_alive, true, true, purged_classes); 2416 } 2417 2418 void ConcurrentMark::weakRefsWork(bool clear_all_soft_refs) { 2419 if (has_overflown()) { 2420 // Skip processing the discovered references if we have 2421 // overflown the global marking stack. Reference objects 2422 // only get discovered once so it is OK to not 2423 // de-populate the discovered reference lists. We could have, 2424 // but the only benefit would be that, when marking restarts, 2425 // less reference objects are discovered. 2426 return; 2427 } 2428 2429 ResourceMark rm; 2430 HandleMark hm; 2431 2432 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 2433 2434 // Is alive closure. 2435 G1CMIsAliveClosure g1_is_alive(g1h); 2436 2437 // Inner scope to exclude the cleaning of the string and symbol 2438 // tables from the displayed time. 2439 { 2440 G1CMTraceTime t("GC ref-proc", G1Log::finer()); 2441 2442 ReferenceProcessor* rp = g1h->ref_processor_cm(); 2443 2444 // See the comment in G1CollectedHeap::ref_processing_init() 2445 // about how reference processing currently works in G1. 2446 2447 // Set the soft reference policy 2448 rp->setup_policy(clear_all_soft_refs); 2449 assert(_markStack.isEmpty(), "mark stack should be empty"); 2450 2451 // Instances of the 'Keep Alive' and 'Complete GC' closures used 2452 // in serial reference processing. Note these closures are also 2453 // used for serially processing (by the the current thread) the 2454 // JNI references during parallel reference processing. 2455 // 2456 // These closures do not need to synchronize with the worker 2457 // threads involved in parallel reference processing as these 2458 // instances are executed serially by the current thread (e.g. 2459 // reference processing is not multi-threaded and is thus 2460 // performed by the current thread instead of a gang worker). 2461 // 2462 // The gang tasks involved in parallel reference processing create 2463 // their own instances of these closures, which do their own 2464 // synchronization among themselves. 2465 G1CMKeepAliveAndDrainClosure g1_keep_alive(this, task(0), true /* is_serial */); 2466 G1CMDrainMarkingStackClosure g1_drain_mark_stack(this, task(0), true /* is_serial */); 2467 2468 // We need at least one active thread. If reference processing 2469 // is not multi-threaded we use the current (VMThread) thread, 2470 // otherwise we use the work gang from the G1CollectedHeap and 2471 // we utilize all the worker threads we can. 2472 bool processing_is_mt = rp->processing_is_mt(); 2473 uint active_workers = (processing_is_mt ? g1h->workers()->active_workers() : 1U); 2474 active_workers = MAX2(MIN2(active_workers, _max_worker_id), 1U); 2475 2476 // Parallel processing task executor. 2477 G1CMRefProcTaskExecutor par_task_executor(g1h, this, 2478 g1h->workers(), active_workers); 2479 AbstractRefProcTaskExecutor* executor = (processing_is_mt ? &par_task_executor : NULL); 2480 2481 // Set the concurrency level. The phase was already set prior to 2482 // executing the remark task. 2483 set_concurrency(active_workers); 2484 2485 // Set the degree of MT processing here. If the discovery was done MT, 2486 // the number of threads involved during discovery could differ from 2487 // the number of active workers. This is OK as long as the discovered 2488 // Reference lists are balanced (see balance_all_queues() and balance_queues()). 2489 rp->set_active_mt_degree(active_workers); 2490 2491 // Process the weak references. 2492 const ReferenceProcessorStats& stats = 2493 rp->process_discovered_references(&g1_is_alive, 2494 &g1_keep_alive, 2495 &g1_drain_mark_stack, 2496 executor, 2497 g1h->gc_timer_cm(), 2498 concurrent_gc_id()); 2499 g1h->gc_tracer_cm()->report_gc_reference_stats(stats); 2500 2501 // The do_oop work routines of the keep_alive and drain_marking_stack 2502 // oop closures will set the has_overflown flag if we overflow the 2503 // global marking stack. 2504 2505 assert(_markStack.overflow() || _markStack.isEmpty(), 2506 "mark stack should be empty (unless it overflowed)"); 2507 2508 if (_markStack.overflow()) { 2509 // This should have been done already when we tried to push an 2510 // entry on to the global mark stack. But let's do it again. 2511 set_has_overflown(); 2512 } 2513 2514 assert(rp->num_q() == active_workers, "why not"); 2515 2516 rp->enqueue_discovered_references(executor); 2517 2518 rp->verify_no_references_recorded(); 2519 assert(!rp->discovery_enabled(), "Post condition"); 2520 } 2521 2522 if (has_overflown()) { 2523 // We can not trust g1_is_alive if the marking stack overflowed 2524 return; 2525 } 2526 2527 assert(_markStack.isEmpty(), "Marking should have completed"); 2528 2529 // Unload Klasses, String, Symbols, Code Cache, etc. 2530 { 2531 G1CMTraceTime trace("Unloading", G1Log::finer()); 2532 2533 if (ClassUnloadingWithConcurrentMark) { 2534 bool purged_classes; 2535 2536 { 2537 G1CMTraceTime trace("System Dictionary Unloading", G1Log::finest()); 2538 purged_classes = SystemDictionary::do_unloading(&g1_is_alive, false /* Defer klass cleaning */); 2539 } 2540 2541 { 2542 G1CMTraceTime trace("Parallel Unloading", G1Log::finest()); 2543 weakRefsWorkParallelPart(&g1_is_alive, purged_classes); 2544 } 2545 } 2546 2547 if (G1StringDedup::is_enabled()) { 2548 G1CMTraceTime trace("String Deduplication Unlink", G1Log::finest()); 2549 G1StringDedup::unlink(&g1_is_alive); 2550 } 2551 } 2552 } 2553 2554 void ConcurrentMark::swapMarkBitMaps() { 2555 CMBitMapRO* temp = _prevMarkBitMap; 2556 _prevMarkBitMap = (CMBitMapRO*)_nextMarkBitMap; 2557 _nextMarkBitMap = (CMBitMap*) temp; 2558 } 2559 2560 class CMObjectClosure; 2561 2562 // Closure for iterating over objects, currently only used for 2563 // processing SATB buffers. 2564 class CMObjectClosure : public ObjectClosure { 2565 private: 2566 CMTask* _task; 2567 2568 public: 2569 void do_object(oop obj) { 2570 _task->deal_with_reference(obj); 2571 } 2572 2573 CMObjectClosure(CMTask* task) : _task(task) { } 2574 }; 2575 2576 class G1RemarkThreadsClosure : public ThreadClosure { 2577 CMObjectClosure _cm_obj; 2578 G1CMOopClosure _cm_cl; 2579 MarkingCodeBlobClosure _code_cl; 2580 int _thread_parity; 2581 2582 public: 2583 G1RemarkThreadsClosure(G1CollectedHeap* g1h, CMTask* task) : 2584 _cm_obj(task), _cm_cl(g1h, g1h->concurrent_mark(), task), _code_cl(&_cm_cl, !CodeBlobToOopClosure::FixRelocations), 2585 _thread_parity(Threads::thread_claim_parity()) {} 2586 2587 void do_thread(Thread* thread) { 2588 if (thread->is_Java_thread()) { 2589 if (thread->claim_oops_do(true, _thread_parity)) { 2590 JavaThread* jt = (JavaThread*)thread; 2591 2592 // In theory it should not be neccessary to explicitly walk the nmethods to find roots for concurrent marking 2593 // however the liveness of oops reachable from nmethods have very complex lifecycles: 2594 // * Alive if on the stack of an executing method 2595 // * Weakly reachable otherwise 2596 // Some objects reachable from nmethods, such as the class loader (or klass_holder) of the receiver should be 2597 // live by the SATB invariant but other oops recorded in nmethods may behave differently. 2598 jt->nmethods_do(&_code_cl); 2599 2600 jt->satb_mark_queue().apply_closure_and_empty(&_cm_obj); 2601 } 2602 } else if (thread->is_VM_thread()) { 2603 if (thread->claim_oops_do(true, _thread_parity)) { 2604 JavaThread::satb_mark_queue_set().shared_satb_queue()->apply_closure_and_empty(&_cm_obj); 2605 } 2606 } 2607 } 2608 }; 2609 2610 class CMRemarkTask: public AbstractGangTask { 2611 private: 2612 ConcurrentMark* _cm; 2613 public: 2614 void work(uint worker_id) { 2615 // Since all available tasks are actually started, we should 2616 // only proceed if we're supposed to be active. 2617 if (worker_id < _cm->active_tasks()) { 2618 CMTask* task = _cm->task(worker_id); 2619 task->record_start_time(); 2620 { 2621 ResourceMark rm; 2622 HandleMark hm; 2623 2624 G1RemarkThreadsClosure threads_f(G1CollectedHeap::heap(), task); 2625 Threads::threads_do(&threads_f); 2626 } 2627 2628 do { 2629 task->do_marking_step(1000000000.0 /* something very large */, 2630 true /* do_termination */, 2631 false /* is_serial */); 2632 } while (task->has_aborted() && !_cm->has_overflown()); 2633 // If we overflow, then we do not want to restart. We instead 2634 // want to abort remark and do concurrent marking again. 2635 task->record_end_time(); 2636 } 2637 } 2638 2639 CMRemarkTask(ConcurrentMark* cm, int active_workers) : 2640 AbstractGangTask("Par Remark"), _cm(cm) { 2641 _cm->terminator()->reset_for_reuse(active_workers); 2642 } 2643 }; 2644 2645 void ConcurrentMark::checkpointRootsFinalWork() { 2646 ResourceMark rm; 2647 HandleMark hm; 2648 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 2649 2650 G1CMTraceTime trace("Finalize Marking", G1Log::finer()); 2651 2652 g1h->ensure_parsability(false); 2653 2654 StrongRootsScope srs; 2655 // this is remark, so we'll use up all active threads 2656 uint active_workers = g1h->workers()->active_workers(); 2657 if (active_workers == 0) { 2658 assert(active_workers > 0, "Should have been set earlier"); 2659 active_workers = (uint) ParallelGCThreads; 2660 g1h->workers()->set_active_workers(active_workers); 2661 } 2662 set_concurrency_and_phase(active_workers, false /* concurrent */); 2663 // Leave _parallel_marking_threads at it's 2664 // value originally calculated in the ConcurrentMark 2665 // constructor and pass values of the active workers 2666 // through the gang in the task. 2667 2668 CMRemarkTask remarkTask(this, active_workers); 2669 // We will start all available threads, even if we decide that the 2670 // active_workers will be fewer. The extra ones will just bail out 2671 // immediately. 2672 g1h->set_par_threads(active_workers); 2673 g1h->workers()->run_task(&remarkTask); 2674 g1h->set_par_threads(0); 2675 2676 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 2677 guarantee(has_overflown() || 2678 satb_mq_set.completed_buffers_num() == 0, 2679 err_msg("Invariant: has_overflown = %s, num buffers = %d", 2680 BOOL_TO_STR(has_overflown()), 2681 satb_mq_set.completed_buffers_num())); 2682 2683 print_stats(); 2684 } 2685 2686 #ifndef PRODUCT 2687 2688 class PrintReachableOopClosure: public OopClosure { 2689 private: 2690 G1CollectedHeap* _g1h; 2691 outputStream* _out; 2692 VerifyOption _vo; 2693 bool _all; 2694 2695 public: 2696 PrintReachableOopClosure(outputStream* out, 2697 VerifyOption vo, 2698 bool all) : 2699 _g1h(G1CollectedHeap::heap()), 2700 _out(out), _vo(vo), _all(all) { } 2701 2702 void do_oop(narrowOop* p) { do_oop_work(p); } 2703 void do_oop( oop* p) { do_oop_work(p); } 2704 2705 template <class T> void do_oop_work(T* p) { 2706 oop obj = oopDesc::load_decode_heap_oop(p); 2707 const char* str = NULL; 2708 const char* str2 = ""; 2709 2710 if (obj == NULL) { 2711 str = ""; 2712 } else if (!_g1h->is_in_g1_reserved(obj)) { 2713 str = " O"; 2714 } else { 2715 HeapRegion* hr = _g1h->heap_region_containing(obj); 2716 bool over_tams = _g1h->allocated_since_marking(obj, hr, _vo); 2717 bool marked = _g1h->is_marked(obj, _vo); 2718 2719 if (over_tams) { 2720 str = " >"; 2721 if (marked) { 2722 str2 = " AND MARKED"; 2723 } 2724 } else if (marked) { 2725 str = " M"; 2726 } else { 2727 str = " NOT"; 2728 } 2729 } 2730 2731 _out->print_cr(" "PTR_FORMAT": "PTR_FORMAT"%s%s", 2732 p2i(p), p2i((void*) obj), str, str2); 2733 } 2734 }; 2735 2736 class PrintReachableObjectClosure : public ObjectClosure { 2737 private: 2738 G1CollectedHeap* _g1h; 2739 outputStream* _out; 2740 VerifyOption _vo; 2741 bool _all; 2742 HeapRegion* _hr; 2743 2744 public: 2745 PrintReachableObjectClosure(outputStream* out, 2746 VerifyOption vo, 2747 bool all, 2748 HeapRegion* hr) : 2749 _g1h(G1CollectedHeap::heap()), 2750 _out(out), _vo(vo), _all(all), _hr(hr) { } 2751 2752 void do_object(oop o) { 2753 bool over_tams = _g1h->allocated_since_marking(o, _hr, _vo); 2754 bool marked = _g1h->is_marked(o, _vo); 2755 bool print_it = _all || over_tams || marked; 2756 2757 if (print_it) { 2758 _out->print_cr(" "PTR_FORMAT"%s", 2759 p2i((void *)o), (over_tams) ? " >" : (marked) ? " M" : ""); 2760 PrintReachableOopClosure oopCl(_out, _vo, _all); 2761 o->oop_iterate_no_header(&oopCl); 2762 } 2763 } 2764 }; 2765 2766 class PrintReachableRegionClosure : public HeapRegionClosure { 2767 private: 2768 G1CollectedHeap* _g1h; 2769 outputStream* _out; 2770 VerifyOption _vo; 2771 bool _all; 2772 2773 public: 2774 bool doHeapRegion(HeapRegion* hr) { 2775 HeapWord* b = hr->bottom(); 2776 HeapWord* e = hr->end(); 2777 HeapWord* t = hr->top(); 2778 HeapWord* p = _g1h->top_at_mark_start(hr, _vo); 2779 _out->print_cr("** ["PTR_FORMAT", "PTR_FORMAT"] top: "PTR_FORMAT" " 2780 "TAMS: " PTR_FORMAT, p2i(b), p2i(e), p2i(t), p2i(p)); 2781 _out->cr(); 2782 2783 HeapWord* from = b; 2784 HeapWord* to = t; 2785 2786 if (to > from) { 2787 _out->print_cr("Objects in [" PTR_FORMAT ", " PTR_FORMAT "]", p2i(from), p2i(to)); 2788 _out->cr(); 2789 PrintReachableObjectClosure ocl(_out, _vo, _all, hr); 2790 hr->object_iterate_mem_careful(MemRegion(from, to), &ocl); 2791 _out->cr(); 2792 } 2793 2794 return false; 2795 } 2796 2797 PrintReachableRegionClosure(outputStream* out, 2798 VerifyOption vo, 2799 bool all) : 2800 _g1h(G1CollectedHeap::heap()), _out(out), _vo(vo), _all(all) { } 2801 }; 2802 2803 void ConcurrentMark::print_reachable(const char* str, 2804 VerifyOption vo, 2805 bool all) { 2806 gclog_or_tty->cr(); 2807 gclog_or_tty->print_cr("== Doing heap dump... "); 2808 2809 if (G1PrintReachableBaseFile == NULL) { 2810 gclog_or_tty->print_cr(" #### error: no base file defined"); 2811 return; 2812 } 2813 2814 if (strlen(G1PrintReachableBaseFile) + 1 + strlen(str) > 2815 (JVM_MAXPATHLEN - 1)) { 2816 gclog_or_tty->print_cr(" #### error: file name too long"); 2817 return; 2818 } 2819 2820 char file_name[JVM_MAXPATHLEN]; 2821 sprintf(file_name, "%s.%s", G1PrintReachableBaseFile, str); 2822 gclog_or_tty->print_cr(" dumping to file %s", file_name); 2823 2824 fileStream fout(file_name); 2825 if (!fout.is_open()) { 2826 gclog_or_tty->print_cr(" #### error: could not open file"); 2827 return; 2828 } 2829 2830 outputStream* out = &fout; 2831 out->print_cr("-- USING %s", _g1h->top_at_mark_start_str(vo)); 2832 out->cr(); 2833 2834 out->print_cr("--- ITERATING OVER REGIONS"); 2835 out->cr(); 2836 PrintReachableRegionClosure rcl(out, vo, all); 2837 _g1h->heap_region_iterate(&rcl); 2838 out->cr(); 2839 2840 gclog_or_tty->print_cr(" done"); 2841 gclog_or_tty->flush(); 2842 } 2843 2844 #endif // PRODUCT 2845 2846 void ConcurrentMark::clearRangePrevBitmap(MemRegion mr) { 2847 // Note we are overriding the read-only view of the prev map here, via 2848 // the cast. 2849 ((CMBitMap*)_prevMarkBitMap)->clearRange(mr); 2850 } 2851 2852 void ConcurrentMark::clearRangeNextBitmap(MemRegion mr) { 2853 _nextMarkBitMap->clearRange(mr); 2854 } 2855 2856 HeapRegion* 2857 ConcurrentMark::claim_region(uint worker_id) { 2858 // "checkpoint" the finger 2859 HeapWord* finger = _finger; 2860 2861 // _heap_end will not change underneath our feet; it only changes at 2862 // yield points. 2863 while (finger < _heap_end) { 2864 assert(_g1h->is_in_g1_reserved(finger), "invariant"); 2865 2866 // Note on how this code handles humongous regions. In the 2867 // normal case the finger will reach the start of a "starts 2868 // humongous" (SH) region. Its end will either be the end of the 2869 // last "continues humongous" (CH) region in the sequence, or the 2870 // standard end of the SH region (if the SH is the only region in 2871 // the sequence). That way claim_region() will skip over the CH 2872 // regions. However, there is a subtle race between a CM thread 2873 // executing this method and a mutator thread doing a humongous 2874 // object allocation. The two are not mutually exclusive as the CM 2875 // thread does not need to hold the Heap_lock when it gets 2876 // here. So there is a chance that claim_region() will come across 2877 // a free region that's in the progress of becoming a SH or a CH 2878 // region. In the former case, it will either 2879 // a) Miss the update to the region's end, in which case it will 2880 // visit every subsequent CH region, will find their bitmaps 2881 // empty, and do nothing, or 2882 // b) Will observe the update of the region's end (in which case 2883 // it will skip the subsequent CH regions). 2884 // If it comes across a region that suddenly becomes CH, the 2885 // scenario will be similar to b). So, the race between 2886 // claim_region() and a humongous object allocation might force us 2887 // to do a bit of unnecessary work (due to some unnecessary bitmap 2888 // iterations) but it should not introduce and correctness issues. 2889 HeapRegion* curr_region = _g1h->heap_region_containing_raw(finger); 2890 2891 // Above heap_region_containing_raw may return NULL as we always scan claim 2892 // until the end of the heap. In this case, just jump to the next region. 2893 HeapWord* end = curr_region != NULL ? curr_region->end() : finger + HeapRegion::GrainWords; 2894 2895 // Is the gap between reading the finger and doing the CAS too long? 2896 HeapWord* res = (HeapWord*) Atomic::cmpxchg_ptr(end, &_finger, finger); 2897 if (res == finger && curr_region != NULL) { 2898 // we succeeded 2899 HeapWord* bottom = curr_region->bottom(); 2900 HeapWord* limit = curr_region->next_top_at_mark_start(); 2901 2902 if (verbose_low()) { 2903 gclog_or_tty->print_cr("[%u] curr_region = "PTR_FORMAT" " 2904 "["PTR_FORMAT", "PTR_FORMAT"), " 2905 "limit = "PTR_FORMAT, 2906 worker_id, p2i(curr_region), p2i(bottom), p2i(end), p2i(limit)); 2907 } 2908 2909 // notice that _finger == end cannot be guaranteed here since, 2910 // someone else might have moved the finger even further 2911 assert(_finger >= end, "the finger should have moved forward"); 2912 2913 if (verbose_low()) { 2914 gclog_or_tty->print_cr("[%u] we were successful with region = " 2915 PTR_FORMAT, worker_id, p2i(curr_region)); 2916 } 2917 2918 if (limit > bottom) { 2919 if (verbose_low()) { 2920 gclog_or_tty->print_cr("[%u] region "PTR_FORMAT" is not empty, " 2921 "returning it ", worker_id, p2i(curr_region)); 2922 } 2923 return curr_region; 2924 } else { 2925 assert(limit == bottom, 2926 "the region limit should be at bottom"); 2927 if (verbose_low()) { 2928 gclog_or_tty->print_cr("[%u] region "PTR_FORMAT" is empty, " 2929 "returning NULL", worker_id, p2i(curr_region)); 2930 } 2931 // we return NULL and the caller should try calling 2932 // claim_region() again. 2933 return NULL; 2934 } 2935 } else { 2936 assert(_finger > finger, "the finger should have moved forward"); 2937 if (verbose_low()) { 2938 if (curr_region == NULL) { 2939 gclog_or_tty->print_cr("[%u] found uncommitted region, moving finger, " 2940 "global finger = "PTR_FORMAT", " 2941 "our finger = "PTR_FORMAT, 2942 worker_id, p2i(_finger), p2i(finger)); 2943 } else { 2944 gclog_or_tty->print_cr("[%u] somebody else moved the finger, " 2945 "global finger = "PTR_FORMAT", " 2946 "our finger = "PTR_FORMAT, 2947 worker_id, p2i(_finger), p2i(finger)); 2948 } 2949 } 2950 2951 // read it again 2952 finger = _finger; 2953 } 2954 } 2955 2956 return NULL; 2957 } 2958 2959 #ifndef PRODUCT 2960 enum VerifyNoCSetOopsPhase { 2961 VerifyNoCSetOopsStack, 2962 VerifyNoCSetOopsQueues, 2963 VerifyNoCSetOopsSATBCompleted, 2964 VerifyNoCSetOopsSATBThread 2965 }; 2966 2967 class VerifyNoCSetOopsClosure : public OopClosure, public ObjectClosure { 2968 private: 2969 G1CollectedHeap* _g1h; 2970 VerifyNoCSetOopsPhase _phase; 2971 int _info; 2972 2973 const char* phase_str() { 2974 switch (_phase) { 2975 case VerifyNoCSetOopsStack: return "Stack"; 2976 case VerifyNoCSetOopsQueues: return "Queue"; 2977 case VerifyNoCSetOopsSATBCompleted: return "Completed SATB Buffers"; 2978 case VerifyNoCSetOopsSATBThread: return "Thread SATB Buffers"; 2979 default: ShouldNotReachHere(); 2980 } 2981 return NULL; 2982 } 2983 2984 void do_object_work(oop obj) { 2985 guarantee(!_g1h->obj_in_cs(obj), 2986 err_msg("obj: "PTR_FORMAT" in CSet, phase: %s, info: %d", 2987 p2i((void*) obj), phase_str(), _info)); 2988 } 2989 2990 public: 2991 VerifyNoCSetOopsClosure() : _g1h(G1CollectedHeap::heap()) { } 2992 2993 void set_phase(VerifyNoCSetOopsPhase phase, int info = -1) { 2994 _phase = phase; 2995 _info = info; 2996 } 2997 2998 virtual void do_oop(oop* p) { 2999 oop obj = oopDesc::load_decode_heap_oop(p); 3000 do_object_work(obj); 3001 } 3002 3003 virtual void do_oop(narrowOop* p) { 3004 // We should not come across narrow oops while scanning marking 3005 // stacks and SATB buffers. 3006 ShouldNotReachHere(); 3007 } 3008 3009 virtual void do_object(oop obj) { 3010 do_object_work(obj); 3011 } 3012 }; 3013 3014 void ConcurrentMark::verify_no_cset_oops(bool verify_stacks, 3015 bool verify_enqueued_buffers, 3016 bool verify_thread_buffers, 3017 bool verify_fingers) { 3018 assert(SafepointSynchronize::is_at_safepoint(), "should be at a safepoint"); 3019 if (!G1CollectedHeap::heap()->mark_in_progress()) { 3020 return; 3021 } 3022 3023 VerifyNoCSetOopsClosure cl; 3024 3025 if (verify_stacks) { 3026 // Verify entries on the global mark stack 3027 cl.set_phase(VerifyNoCSetOopsStack); 3028 _markStack.oops_do(&cl); 3029 3030 // Verify entries on the task queues 3031 for (uint i = 0; i < _max_worker_id; i += 1) { 3032 cl.set_phase(VerifyNoCSetOopsQueues, i); 3033 CMTaskQueue* queue = _task_queues->queue(i); 3034 queue->oops_do(&cl); 3035 } 3036 } 3037 3038 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set(); 3039 3040 // Verify entries on the enqueued SATB buffers 3041 if (verify_enqueued_buffers) { 3042 cl.set_phase(VerifyNoCSetOopsSATBCompleted); 3043 satb_qs.iterate_completed_buffers_read_only(&cl); 3044 } 3045 3046 // Verify entries on the per-thread SATB buffers 3047 if (verify_thread_buffers) { 3048 cl.set_phase(VerifyNoCSetOopsSATBThread); 3049 satb_qs.iterate_thread_buffers_read_only(&cl); 3050 } 3051 3052 if (verify_fingers) { 3053 // Verify the global finger 3054 HeapWord* global_finger = finger(); 3055 if (global_finger != NULL && global_finger < _heap_end) { 3056 // The global finger always points to a heap region boundary. We 3057 // use heap_region_containing_raw() to get the containing region 3058 // given that the global finger could be pointing to a free region 3059 // which subsequently becomes continues humongous. If that 3060 // happens, heap_region_containing() will return the bottom of the 3061 // corresponding starts humongous region and the check below will 3062 // not hold any more. 3063 // Since we always iterate over all regions, we might get a NULL HeapRegion 3064 // here. 3065 HeapRegion* global_hr = _g1h->heap_region_containing_raw(global_finger); 3066 guarantee(global_hr == NULL || global_finger == global_hr->bottom(), 3067 err_msg("global finger: "PTR_FORMAT" region: "HR_FORMAT, 3068 p2i(global_finger), HR_FORMAT_PARAMS(global_hr))); 3069 } 3070 3071 // Verify the task fingers 3072 assert(parallel_marking_threads() <= _max_worker_id, "sanity"); 3073 for (int i = 0; i < (int) parallel_marking_threads(); i += 1) { 3074 CMTask* task = _tasks[i]; 3075 HeapWord* task_finger = task->finger(); 3076 if (task_finger != NULL && task_finger < _heap_end) { 3077 // See above note on the global finger verification. 3078 HeapRegion* task_hr = _g1h->heap_region_containing_raw(task_finger); 3079 guarantee(task_hr == NULL || task_finger == task_hr->bottom() || 3080 !task_hr->in_collection_set(), 3081 err_msg("task finger: "PTR_FORMAT" region: "HR_FORMAT, 3082 p2i(task_finger), HR_FORMAT_PARAMS(task_hr))); 3083 } 3084 } 3085 } 3086 } 3087 #endif // PRODUCT 3088 3089 // Aggregate the counting data that was constructed concurrently 3090 // with marking. 3091 class AggregateCountDataHRClosure: public HeapRegionClosure { 3092 G1CollectedHeap* _g1h; 3093 ConcurrentMark* _cm; 3094 CardTableModRefBS* _ct_bs; 3095 BitMap* _cm_card_bm; 3096 uint _max_worker_id; 3097 3098 public: 3099 AggregateCountDataHRClosure(G1CollectedHeap* g1h, 3100 BitMap* cm_card_bm, 3101 uint max_worker_id) : 3102 _g1h(g1h), _cm(g1h->concurrent_mark()), 3103 _ct_bs(barrier_set_cast<CardTableModRefBS>(g1h->barrier_set())), 3104 _cm_card_bm(cm_card_bm), _max_worker_id(max_worker_id) { } 3105 3106 bool doHeapRegion(HeapRegion* hr) { 3107 if (hr->is_continues_humongous()) { 3108 // We will ignore these here and process them when their 3109 // associated "starts humongous" region is processed. 3110 // Note that we cannot rely on their associated 3111 // "starts humongous" region to have their bit set to 1 3112 // since, due to the region chunking in the parallel region 3113 // iteration, a "continues humongous" region might be visited 3114 // before its associated "starts humongous". 3115 return false; 3116 } 3117 3118 HeapWord* start = hr->bottom(); 3119 HeapWord* limit = hr->next_top_at_mark_start(); 3120 HeapWord* end = hr->end(); 3121 3122 assert(start <= limit && limit <= hr->top() && hr->top() <= hr->end(), 3123 err_msg("Preconditions not met - " 3124 "start: "PTR_FORMAT", limit: "PTR_FORMAT", " 3125 "top: "PTR_FORMAT", end: "PTR_FORMAT, 3126 p2i(start), p2i(limit), p2i(hr->top()), p2i(hr->end()))); 3127 3128 assert(hr->next_marked_bytes() == 0, "Precondition"); 3129 3130 if (start == limit) { 3131 // NTAMS of this region has not been set so nothing to do. 3132 return false; 3133 } 3134 3135 // 'start' should be in the heap. 3136 assert(_g1h->is_in_g1_reserved(start) && _ct_bs->is_card_aligned(start), "sanity"); 3137 // 'end' *may* be just beyond the end of the heap (if hr is the last region) 3138 assert(!_g1h->is_in_g1_reserved(end) || _ct_bs->is_card_aligned(end), "sanity"); 3139 3140 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start); 3141 BitMap::idx_t limit_idx = _cm->card_bitmap_index_for(limit); 3142 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(end); 3143 3144 // If ntams is not card aligned then we bump card bitmap index 3145 // for limit so that we get the all the cards spanned by 3146 // the object ending at ntams. 3147 // Note: if this is the last region in the heap then ntams 3148 // could be actually just beyond the end of the the heap; 3149 // limit_idx will then correspond to a (non-existent) card 3150 // that is also outside the heap. 3151 if (_g1h->is_in_g1_reserved(limit) && !_ct_bs->is_card_aligned(limit)) { 3152 limit_idx += 1; 3153 } 3154 3155 assert(limit_idx <= end_idx, "or else use atomics"); 3156 3157 // Aggregate the "stripe" in the count data associated with hr. 3158 uint hrm_index = hr->hrm_index(); 3159 size_t marked_bytes = 0; 3160 3161 for (uint i = 0; i < _max_worker_id; i += 1) { 3162 size_t* marked_bytes_array = _cm->count_marked_bytes_array_for(i); 3163 BitMap* task_card_bm = _cm->count_card_bitmap_for(i); 3164 3165 // Fetch the marked_bytes in this region for task i and 3166 // add it to the running total for this region. 3167 marked_bytes += marked_bytes_array[hrm_index]; 3168 3169 // Now union the bitmaps[0,max_worker_id)[start_idx..limit_idx) 3170 // into the global card bitmap. 3171 BitMap::idx_t scan_idx = task_card_bm->get_next_one_offset(start_idx, limit_idx); 3172 3173 while (scan_idx < limit_idx) { 3174 assert(task_card_bm->at(scan_idx) == true, "should be"); 3175 _cm_card_bm->set_bit(scan_idx); 3176 assert(_cm_card_bm->at(scan_idx) == true, "should be"); 3177 3178 // BitMap::get_next_one_offset() can handle the case when 3179 // its left_offset parameter is greater than its right_offset 3180 // parameter. It does, however, have an early exit if 3181 // left_offset == right_offset. So let's limit the value 3182 // passed in for left offset here. 3183 BitMap::idx_t next_idx = MIN2(scan_idx + 1, limit_idx); 3184 scan_idx = task_card_bm->get_next_one_offset(next_idx, limit_idx); 3185 } 3186 } 3187 3188 // Update the marked bytes for this region. 3189 hr->add_to_marked_bytes(marked_bytes); 3190 3191 // Next heap region 3192 return false; 3193 } 3194 }; 3195 3196 class G1AggregateCountDataTask: public AbstractGangTask { 3197 protected: 3198 G1CollectedHeap* _g1h; 3199 ConcurrentMark* _cm; 3200 BitMap* _cm_card_bm; 3201 uint _max_worker_id; 3202 int _active_workers; 3203 HeapRegionClaimer _hrclaimer; 3204 3205 public: 3206 G1AggregateCountDataTask(G1CollectedHeap* g1h, 3207 ConcurrentMark* cm, 3208 BitMap* cm_card_bm, 3209 uint max_worker_id, 3210 int n_workers) : 3211 AbstractGangTask("Count Aggregation"), 3212 _g1h(g1h), _cm(cm), _cm_card_bm(cm_card_bm), 3213 _max_worker_id(max_worker_id), 3214 _active_workers(n_workers), 3215 _hrclaimer(_active_workers) { 3216 } 3217 3218 void work(uint worker_id) { 3219 AggregateCountDataHRClosure cl(_g1h, _cm_card_bm, _max_worker_id); 3220 3221 _g1h->heap_region_par_iterate(&cl, worker_id, &_hrclaimer); 3222 } 3223 }; 3224 3225 3226 void ConcurrentMark::aggregate_count_data() { 3227 int n_workers = _g1h->workers()->active_workers(); 3228 3229 G1AggregateCountDataTask g1_par_agg_task(_g1h, this, &_card_bm, 3230 _max_worker_id, n_workers); 3231 3232 _g1h->set_par_threads(n_workers); 3233 _g1h->workers()->run_task(&g1_par_agg_task); 3234 _g1h->set_par_threads(0); 3235 } 3236 3237 // Clear the per-worker arrays used to store the per-region counting data 3238 void ConcurrentMark::clear_all_count_data() { 3239 // Clear the global card bitmap - it will be filled during 3240 // liveness count aggregation (during remark) and the 3241 // final counting task. 3242 _card_bm.clear(); 3243 3244 // Clear the global region bitmap - it will be filled as part 3245 // of the final counting task. 3246 _region_bm.clear(); 3247 3248 uint max_regions = _g1h->max_regions(); 3249 assert(_max_worker_id > 0, "uninitialized"); 3250 3251 for (uint i = 0; i < _max_worker_id; i += 1) { 3252 BitMap* task_card_bm = count_card_bitmap_for(i); 3253 size_t* marked_bytes_array = count_marked_bytes_array_for(i); 3254 3255 assert(task_card_bm->size() == _card_bm.size(), "size mismatch"); 3256 assert(marked_bytes_array != NULL, "uninitialized"); 3257 3258 memset(marked_bytes_array, 0, (size_t) max_regions * sizeof(size_t)); 3259 task_card_bm->clear(); 3260 } 3261 } 3262 3263 void ConcurrentMark::print_stats() { 3264 if (verbose_stats()) { 3265 gclog_or_tty->print_cr("---------------------------------------------------------------------"); 3266 for (size_t i = 0; i < _active_tasks; ++i) { 3267 _tasks[i]->print_stats(); 3268 gclog_or_tty->print_cr("---------------------------------------------------------------------"); 3269 } 3270 } 3271 } 3272 3273 // abandon current marking iteration due to a Full GC 3274 void ConcurrentMark::abort() { 3275 // Clear all marks in the next bitmap for the next marking cycle. This will allow us to skip the next 3276 // concurrent bitmap clearing. 3277 _nextMarkBitMap->clearAll(); 3278 3279 // Note we cannot clear the previous marking bitmap here 3280 // since VerifyDuringGC verifies the objects marked during 3281 // a full GC against the previous bitmap. 3282 3283 // Clear the liveness counting data 3284 clear_all_count_data(); 3285 // Empty mark stack 3286 reset_marking_state(); 3287 for (uint i = 0; i < _max_worker_id; ++i) { 3288 _tasks[i]->clear_region_fields(); 3289 } 3290 _first_overflow_barrier_sync.abort(); 3291 _second_overflow_barrier_sync.abort(); 3292 const GCId& gc_id = _g1h->gc_tracer_cm()->gc_id(); 3293 if (!gc_id.is_undefined()) { 3294 // We can do multiple full GCs before ConcurrentMarkThread::run() gets a chance 3295 // to detect that it was aborted. Only keep track of the first GC id that we aborted. 3296 _aborted_gc_id = gc_id; 3297 } 3298 _has_aborted = true; 3299 3300 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 3301 satb_mq_set.abandon_partial_marking(); 3302 // This can be called either during or outside marking, we'll read 3303 // the expected_active value from the SATB queue set. 3304 satb_mq_set.set_active_all_threads( 3305 false, /* new active value */ 3306 satb_mq_set.is_active() /* expected_active */); 3307 3308 _g1h->trace_heap_after_concurrent_cycle(); 3309 _g1h->register_concurrent_cycle_end(); 3310 } 3311 3312 const GCId& ConcurrentMark::concurrent_gc_id() { 3313 if (has_aborted()) { 3314 return _aborted_gc_id; 3315 } 3316 return _g1h->gc_tracer_cm()->gc_id(); 3317 } 3318 3319 static void print_ms_time_info(const char* prefix, const char* name, 3320 NumberSeq& ns) { 3321 gclog_or_tty->print_cr("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).", 3322 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg()); 3323 if (ns.num() > 0) { 3324 gclog_or_tty->print_cr("%s [std. dev = %8.2f ms, max = %8.2f ms]", 3325 prefix, ns.sd(), ns.maximum()); 3326 } 3327 } 3328 3329 void ConcurrentMark::print_summary_info() { 3330 gclog_or_tty->print_cr(" Concurrent marking:"); 3331 print_ms_time_info(" ", "init marks", _init_times); 3332 print_ms_time_info(" ", "remarks", _remark_times); 3333 { 3334 print_ms_time_info(" ", "final marks", _remark_mark_times); 3335 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times); 3336 3337 } 3338 print_ms_time_info(" ", "cleanups", _cleanup_times); 3339 gclog_or_tty->print_cr(" Final counting total time = %8.2f s (avg = %8.2f ms).", 3340 _total_counting_time, 3341 (_cleanup_times.num() > 0 ? _total_counting_time * 1000.0 / 3342 (double)_cleanup_times.num() 3343 : 0.0)); 3344 if (G1ScrubRemSets) { 3345 gclog_or_tty->print_cr(" RS scrub total time = %8.2f s (avg = %8.2f ms).", 3346 _total_rs_scrub_time, 3347 (_cleanup_times.num() > 0 ? _total_rs_scrub_time * 1000.0 / 3348 (double)_cleanup_times.num() 3349 : 0.0)); 3350 } 3351 gclog_or_tty->print_cr(" Total stop_world time = %8.2f s.", 3352 (_init_times.sum() + _remark_times.sum() + 3353 _cleanup_times.sum())/1000.0); 3354 gclog_or_tty->print_cr(" Total concurrent time = %8.2f s " 3355 "(%8.2f s marking).", 3356 cmThread()->vtime_accum(), 3357 cmThread()->vtime_mark_accum()); 3358 } 3359 3360 void ConcurrentMark::print_worker_threads_on(outputStream* st) const { 3361 _parallel_workers->print_worker_threads_on(st); 3362 } 3363 3364 void ConcurrentMark::print_on_error(outputStream* st) const { 3365 st->print_cr("Marking Bits (Prev, Next): (CMBitMap*) " PTR_FORMAT ", (CMBitMap*) " PTR_FORMAT, 3366 p2i(_prevMarkBitMap), p2i(_nextMarkBitMap)); 3367 _prevMarkBitMap->print_on_error(st, " Prev Bits: "); 3368 _nextMarkBitMap->print_on_error(st, " Next Bits: "); 3369 } 3370 3371 // We take a break if someone is trying to stop the world. 3372 bool ConcurrentMark::do_yield_check(uint worker_id) { 3373 if (SuspendibleThreadSet::should_yield()) { 3374 if (worker_id == 0) { 3375 _g1h->g1_policy()->record_concurrent_pause(); 3376 } 3377 SuspendibleThreadSet::yield(); 3378 return true; 3379 } else { 3380 return false; 3381 } 3382 } 3383 3384 #ifndef PRODUCT 3385 // for debugging purposes 3386 void ConcurrentMark::print_finger() { 3387 gclog_or_tty->print_cr("heap ["PTR_FORMAT", "PTR_FORMAT"), global finger = "PTR_FORMAT, 3388 p2i(_heap_start), p2i(_heap_end), p2i(_finger)); 3389 for (uint i = 0; i < _max_worker_id; ++i) { 3390 gclog_or_tty->print(" %u: " PTR_FORMAT, i, p2i(_tasks[i]->finger())); 3391 } 3392 gclog_or_tty->cr(); 3393 } 3394 #endif 3395 3396 void CMTask::scan_object(oop obj) { 3397 assert(_nextMarkBitMap->isMarked((HeapWord*) obj), "invariant"); 3398 3399 if (_cm->verbose_high()) { 3400 gclog_or_tty->print_cr("[%u] we're scanning object "PTR_FORMAT, 3401 _worker_id, p2i((void*) obj)); 3402 } 3403 3404 size_t obj_size = obj->size(); 3405 _words_scanned += obj_size; 3406 3407 obj->oop_iterate(_cm_oop_closure); 3408 statsOnly( ++_objs_scanned ); 3409 check_limits(); 3410 } 3411 3412 // Closure for iteration over bitmaps 3413 class CMBitMapClosure : public BitMapClosure { 3414 private: 3415 // the bitmap that is being iterated over 3416 CMBitMap* _nextMarkBitMap; 3417 ConcurrentMark* _cm; 3418 CMTask* _task; 3419 3420 public: 3421 CMBitMapClosure(CMTask *task, ConcurrentMark* cm, CMBitMap* nextMarkBitMap) : 3422 _task(task), _cm(cm), _nextMarkBitMap(nextMarkBitMap) { } 3423 3424 bool do_bit(size_t offset) { 3425 HeapWord* addr = _nextMarkBitMap->offsetToHeapWord(offset); 3426 assert(_nextMarkBitMap->isMarked(addr), "invariant"); 3427 assert( addr < _cm->finger(), "invariant"); 3428 3429 statsOnly( _task->increase_objs_found_on_bitmap() ); 3430 assert(addr >= _task->finger(), "invariant"); 3431 3432 // We move that task's local finger along. 3433 _task->move_finger_to(addr); 3434 3435 _task->scan_object(oop(addr)); 3436 // we only partially drain the local queue and global stack 3437 _task->drain_local_queue(true); 3438 _task->drain_global_stack(true); 3439 3440 // if the has_aborted flag has been raised, we need to bail out of 3441 // the iteration 3442 return !_task->has_aborted(); 3443 } 3444 }; 3445 3446 G1CMOopClosure::G1CMOopClosure(G1CollectedHeap* g1h, 3447 ConcurrentMark* cm, 3448 CMTask* task) 3449 : _g1h(g1h), _cm(cm), _task(task) { 3450 assert(_ref_processor == NULL, "should be initialized to NULL"); 3451 3452 if (G1UseConcMarkReferenceProcessing) { 3453 _ref_processor = g1h->ref_processor_cm(); 3454 assert(_ref_processor != NULL, "should not be NULL"); 3455 } 3456 } 3457 3458 void CMTask::setup_for_region(HeapRegion* hr) { 3459 assert(hr != NULL, 3460 "claim_region() should have filtered out NULL regions"); 3461 assert(!hr->is_continues_humongous(), 3462 "claim_region() should have filtered out continues humongous regions"); 3463 3464 if (_cm->verbose_low()) { 3465 gclog_or_tty->print_cr("[%u] setting up for region "PTR_FORMAT, 3466 _worker_id, p2i(hr)); 3467 } 3468 3469 _curr_region = hr; 3470 _finger = hr->bottom(); 3471 update_region_limit(); 3472 } 3473 3474 void CMTask::update_region_limit() { 3475 HeapRegion* hr = _curr_region; 3476 HeapWord* bottom = hr->bottom(); 3477 HeapWord* limit = hr->next_top_at_mark_start(); 3478 3479 if (limit == bottom) { 3480 if (_cm->verbose_low()) { 3481 gclog_or_tty->print_cr("[%u] found an empty region " 3482 "["PTR_FORMAT", "PTR_FORMAT")", 3483 _worker_id, p2i(bottom), p2i(limit)); 3484 } 3485 // The region was collected underneath our feet. 3486 // We set the finger to bottom to ensure that the bitmap 3487 // iteration that will follow this will not do anything. 3488 // (this is not a condition that holds when we set the region up, 3489 // as the region is not supposed to be empty in the first place) 3490 _finger = bottom; 3491 } else if (limit >= _region_limit) { 3492 assert(limit >= _finger, "peace of mind"); 3493 } else { 3494 assert(limit < _region_limit, "only way to get here"); 3495 // This can happen under some pretty unusual circumstances. An 3496 // evacuation pause empties the region underneath our feet (NTAMS 3497 // at bottom). We then do some allocation in the region (NTAMS 3498 // stays at bottom), followed by the region being used as a GC 3499 // alloc region (NTAMS will move to top() and the objects 3500 // originally below it will be grayed). All objects now marked in 3501 // the region are explicitly grayed, if below the global finger, 3502 // and we do not need in fact to scan anything else. So, we simply 3503 // set _finger to be limit to ensure that the bitmap iteration 3504 // doesn't do anything. 3505 _finger = limit; 3506 } 3507 3508 _region_limit = limit; 3509 } 3510 3511 void CMTask::giveup_current_region() { 3512 assert(_curr_region != NULL, "invariant"); 3513 if (_cm->verbose_low()) { 3514 gclog_or_tty->print_cr("[%u] giving up region "PTR_FORMAT, 3515 _worker_id, p2i(_curr_region)); 3516 } 3517 clear_region_fields(); 3518 } 3519 3520 void CMTask::clear_region_fields() { 3521 // Values for these three fields that indicate that we're not 3522 // holding on to a region. 3523 _curr_region = NULL; 3524 _finger = NULL; 3525 _region_limit = NULL; 3526 } 3527 3528 void CMTask::set_cm_oop_closure(G1CMOopClosure* cm_oop_closure) { 3529 if (cm_oop_closure == NULL) { 3530 assert(_cm_oop_closure != NULL, "invariant"); 3531 } else { 3532 assert(_cm_oop_closure == NULL, "invariant"); 3533 } 3534 _cm_oop_closure = cm_oop_closure; 3535 } 3536 3537 void CMTask::reset(CMBitMap* nextMarkBitMap) { 3538 guarantee(nextMarkBitMap != NULL, "invariant"); 3539 3540 if (_cm->verbose_low()) { 3541 gclog_or_tty->print_cr("[%u] resetting", _worker_id); 3542 } 3543 3544 _nextMarkBitMap = nextMarkBitMap; 3545 clear_region_fields(); 3546 3547 _calls = 0; 3548 _elapsed_time_ms = 0.0; 3549 _termination_time_ms = 0.0; 3550 _termination_start_time_ms = 0.0; 3551 3552 #if _MARKING_STATS_ 3553 _aborted = 0; 3554 _aborted_overflow = 0; 3555 _aborted_cm_aborted = 0; 3556 _aborted_yield = 0; 3557 _aborted_timed_out = 0; 3558 _aborted_satb = 0; 3559 _aborted_termination = 0; 3560 _steal_attempts = 0; 3561 _steals = 0; 3562 _local_pushes = 0; 3563 _local_pops = 0; 3564 _local_max_size = 0; 3565 _objs_scanned = 0; 3566 _global_pushes = 0; 3567 _global_pops = 0; 3568 _global_max_size = 0; 3569 _global_transfers_to = 0; 3570 _global_transfers_from = 0; 3571 _regions_claimed = 0; 3572 _objs_found_on_bitmap = 0; 3573 _satb_buffers_processed = 0; 3574 #endif // _MARKING_STATS_ 3575 } 3576 3577 bool CMTask::should_exit_termination() { 3578 regular_clock_call(); 3579 // This is called when we are in the termination protocol. We should 3580 // quit if, for some reason, this task wants to abort or the global 3581 // stack is not empty (this means that we can get work from it). 3582 return !_cm->mark_stack_empty() || has_aborted(); 3583 } 3584 3585 void CMTask::reached_limit() { 3586 assert(_words_scanned >= _words_scanned_limit || 3587 _refs_reached >= _refs_reached_limit , 3588 "shouldn't have been called otherwise"); 3589 regular_clock_call(); 3590 } 3591 3592 void CMTask::regular_clock_call() { 3593 if (has_aborted()) return; 3594 3595 // First, we need to recalculate the words scanned and refs reached 3596 // limits for the next clock call. 3597 recalculate_limits(); 3598 3599 // During the regular clock call we do the following 3600 3601 // (1) If an overflow has been flagged, then we abort. 3602 if (_cm->has_overflown()) { 3603 set_has_aborted(); 3604 return; 3605 } 3606 3607 // If we are not concurrent (i.e. we're doing remark) we don't need 3608 // to check anything else. The other steps are only needed during 3609 // the concurrent marking phase. 3610 if (!concurrent()) return; 3611 3612 // (2) If marking has been aborted for Full GC, then we also abort. 3613 if (_cm->has_aborted()) { 3614 set_has_aborted(); 3615 statsOnly( ++_aborted_cm_aborted ); 3616 return; 3617 } 3618 3619 double curr_time_ms = os::elapsedVTime() * 1000.0; 3620 3621 // (3) If marking stats are enabled, then we update the step history. 3622 #if _MARKING_STATS_ 3623 if (_words_scanned >= _words_scanned_limit) { 3624 ++_clock_due_to_scanning; 3625 } 3626 if (_refs_reached >= _refs_reached_limit) { 3627 ++_clock_due_to_marking; 3628 } 3629 3630 double last_interval_ms = curr_time_ms - _interval_start_time_ms; 3631 _interval_start_time_ms = curr_time_ms; 3632 _all_clock_intervals_ms.add(last_interval_ms); 3633 3634 if (_cm->verbose_medium()) { 3635 gclog_or_tty->print_cr("[%u] regular clock, interval = %1.2lfms, " 3636 "scanned = "SIZE_FORMAT"%s, refs reached = "SIZE_FORMAT"%s", 3637 _worker_id, last_interval_ms, 3638 _words_scanned, 3639 (_words_scanned >= _words_scanned_limit) ? " (*)" : "", 3640 _refs_reached, 3641 (_refs_reached >= _refs_reached_limit) ? " (*)" : ""); 3642 } 3643 #endif // _MARKING_STATS_ 3644 3645 // (4) We check whether we should yield. If we have to, then we abort. 3646 if (SuspendibleThreadSet::should_yield()) { 3647 // We should yield. To do this we abort the task. The caller is 3648 // responsible for yielding. 3649 set_has_aborted(); 3650 statsOnly( ++_aborted_yield ); 3651 return; 3652 } 3653 3654 // (5) We check whether we've reached our time quota. If we have, 3655 // then we abort. 3656 double elapsed_time_ms = curr_time_ms - _start_time_ms; 3657 if (elapsed_time_ms > _time_target_ms) { 3658 set_has_aborted(); 3659 _has_timed_out = true; 3660 statsOnly( ++_aborted_timed_out ); 3661 return; 3662 } 3663 3664 // (6) Finally, we check whether there are enough completed STAB 3665 // buffers available for processing. If there are, we abort. 3666 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 3667 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) { 3668 if (_cm->verbose_low()) { 3669 gclog_or_tty->print_cr("[%u] aborting to deal with pending SATB buffers", 3670 _worker_id); 3671 } 3672 // we do need to process SATB buffers, we'll abort and restart 3673 // the marking task to do so 3674 set_has_aborted(); 3675 statsOnly( ++_aborted_satb ); 3676 return; 3677 } 3678 } 3679 3680 void CMTask::recalculate_limits() { 3681 _real_words_scanned_limit = _words_scanned + words_scanned_period; 3682 _words_scanned_limit = _real_words_scanned_limit; 3683 3684 _real_refs_reached_limit = _refs_reached + refs_reached_period; 3685 _refs_reached_limit = _real_refs_reached_limit; 3686 } 3687 3688 void CMTask::decrease_limits() { 3689 // This is called when we believe that we're going to do an infrequent 3690 // operation which will increase the per byte scanned cost (i.e. move 3691 // entries to/from the global stack). It basically tries to decrease the 3692 // scanning limit so that the clock is called earlier. 3693 3694 if (_cm->verbose_medium()) { 3695 gclog_or_tty->print_cr("[%u] decreasing limits", _worker_id); 3696 } 3697 3698 _words_scanned_limit = _real_words_scanned_limit - 3699 3 * words_scanned_period / 4; 3700 _refs_reached_limit = _real_refs_reached_limit - 3701 3 * refs_reached_period / 4; 3702 } 3703 3704 void CMTask::move_entries_to_global_stack() { 3705 // local array where we'll store the entries that will be popped 3706 // from the local queue 3707 oop buffer[global_stack_transfer_size]; 3708 3709 int n = 0; 3710 oop obj; 3711 while (n < global_stack_transfer_size && _task_queue->pop_local(obj)) { 3712 buffer[n] = obj; 3713 ++n; 3714 } 3715 3716 if (n > 0) { 3717 // we popped at least one entry from the local queue 3718 3719 statsOnly( ++_global_transfers_to; _local_pops += n ); 3720 3721 if (!_cm->mark_stack_push(buffer, n)) { 3722 if (_cm->verbose_low()) { 3723 gclog_or_tty->print_cr("[%u] aborting due to global stack overflow", 3724 _worker_id); 3725 } 3726 set_has_aborted(); 3727 } else { 3728 // the transfer was successful 3729 3730 if (_cm->verbose_medium()) { 3731 gclog_or_tty->print_cr("[%u] pushed %d entries to the global stack", 3732 _worker_id, n); 3733 } 3734 statsOnly( size_t tmp_size = _cm->mark_stack_size(); 3735 if (tmp_size > _global_max_size) { 3736 _global_max_size = tmp_size; 3737 } 3738 _global_pushes += n ); 3739 } 3740 } 3741 3742 // this operation was quite expensive, so decrease the limits 3743 decrease_limits(); 3744 } 3745 3746 void CMTask::get_entries_from_global_stack() { 3747 // local array where we'll store the entries that will be popped 3748 // from the global stack. 3749 oop buffer[global_stack_transfer_size]; 3750 int n; 3751 _cm->mark_stack_pop(buffer, global_stack_transfer_size, &n); 3752 assert(n <= global_stack_transfer_size, 3753 "we should not pop more than the given limit"); 3754 if (n > 0) { 3755 // yes, we did actually pop at least one entry 3756 3757 statsOnly( ++_global_transfers_from; _global_pops += n ); 3758 if (_cm->verbose_medium()) { 3759 gclog_or_tty->print_cr("[%u] popped %d entries from the global stack", 3760 _worker_id, n); 3761 } 3762 for (int i = 0; i < n; ++i) { 3763 bool success = _task_queue->push(buffer[i]); 3764 // We only call this when the local queue is empty or under a 3765 // given target limit. So, we do not expect this push to fail. 3766 assert(success, "invariant"); 3767 } 3768 3769 statsOnly( size_t tmp_size = (size_t)_task_queue->size(); 3770 if (tmp_size > _local_max_size) { 3771 _local_max_size = tmp_size; 3772 } 3773 _local_pushes += n ); 3774 } 3775 3776 // this operation was quite expensive, so decrease the limits 3777 decrease_limits(); 3778 } 3779 3780 void CMTask::drain_local_queue(bool partially) { 3781 if (has_aborted()) return; 3782 3783 // Decide what the target size is, depending whether we're going to 3784 // drain it partially (so that other tasks can steal if they run out 3785 // of things to do) or totally (at the very end). 3786 size_t target_size; 3787 if (partially) { 3788 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize); 3789 } else { 3790 target_size = 0; 3791 } 3792 3793 if (_task_queue->size() > target_size) { 3794 if (_cm->verbose_high()) { 3795 gclog_or_tty->print_cr("[%u] draining local queue, target size = " SIZE_FORMAT, 3796 _worker_id, target_size); 3797 } 3798 3799 oop obj; 3800 bool ret = _task_queue->pop_local(obj); 3801 while (ret) { 3802 statsOnly( ++_local_pops ); 3803 3804 if (_cm->verbose_high()) { 3805 gclog_or_tty->print_cr("[%u] popped "PTR_FORMAT, _worker_id, 3806 p2i((void*) obj)); 3807 } 3808 3809 assert(_g1h->is_in_g1_reserved((HeapWord*) obj), "invariant" ); 3810 assert(!_g1h->is_on_master_free_list( 3811 _g1h->heap_region_containing((HeapWord*) obj)), "invariant"); 3812 3813 scan_object(obj); 3814 3815 if (_task_queue->size() <= target_size || has_aborted()) { 3816 ret = false; 3817 } else { 3818 ret = _task_queue->pop_local(obj); 3819 } 3820 } 3821 3822 if (_cm->verbose_high()) { 3823 gclog_or_tty->print_cr("[%u] drained local queue, size = %u", 3824 _worker_id, _task_queue->size()); 3825 } 3826 } 3827 } 3828 3829 void CMTask::drain_global_stack(bool partially) { 3830 if (has_aborted()) return; 3831 3832 // We have a policy to drain the local queue before we attempt to 3833 // drain the global stack. 3834 assert(partially || _task_queue->size() == 0, "invariant"); 3835 3836 // Decide what the target size is, depending whether we're going to 3837 // drain it partially (so that other tasks can steal if they run out 3838 // of things to do) or totally (at the very end). Notice that, 3839 // because we move entries from the global stack in chunks or 3840 // because another task might be doing the same, we might in fact 3841 // drop below the target. But, this is not a problem. 3842 size_t target_size; 3843 if (partially) { 3844 target_size = _cm->partial_mark_stack_size_target(); 3845 } else { 3846 target_size = 0; 3847 } 3848 3849 if (_cm->mark_stack_size() > target_size) { 3850 if (_cm->verbose_low()) { 3851 gclog_or_tty->print_cr("[%u] draining global_stack, target size " SIZE_FORMAT, 3852 _worker_id, target_size); 3853 } 3854 3855 while (!has_aborted() && _cm->mark_stack_size() > target_size) { 3856 get_entries_from_global_stack(); 3857 drain_local_queue(partially); 3858 } 3859 3860 if (_cm->verbose_low()) { 3861 gclog_or_tty->print_cr("[%u] drained global stack, size = " SIZE_FORMAT, 3862 _worker_id, _cm->mark_stack_size()); 3863 } 3864 } 3865 } 3866 3867 // SATB Queue has several assumptions on whether to call the par or 3868 // non-par versions of the methods. this is why some of the code is 3869 // replicated. We should really get rid of the single-threaded version 3870 // of the code to simplify things. 3871 void CMTask::drain_satb_buffers() { 3872 if (has_aborted()) return; 3873 3874 // We set this so that the regular clock knows that we're in the 3875 // middle of draining buffers and doesn't set the abort flag when it 3876 // notices that SATB buffers are available for draining. It'd be 3877 // very counter productive if it did that. :-) 3878 _draining_satb_buffers = true; 3879 3880 CMObjectClosure oc(this); 3881 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 3882 satb_mq_set.set_closure(_worker_id, &oc); 3883 3884 // This keeps claiming and applying the closure to completed buffers 3885 // until we run out of buffers or we need to abort. 3886 while (!has_aborted() && 3887 satb_mq_set.apply_closure_to_completed_buffer(_worker_id)) { 3888 if (_cm->verbose_medium()) { 3889 gclog_or_tty->print_cr("[%u] processed an SATB buffer", _worker_id); 3890 } 3891 statsOnly( ++_satb_buffers_processed ); 3892 regular_clock_call(); 3893 } 3894 3895 _draining_satb_buffers = false; 3896 3897 assert(has_aborted() || 3898 concurrent() || 3899 satb_mq_set.completed_buffers_num() == 0, "invariant"); 3900 3901 satb_mq_set.set_closure(_worker_id, NULL); 3902 3903 // again, this was a potentially expensive operation, decrease the 3904 // limits to get the regular clock call early 3905 decrease_limits(); 3906 } 3907 3908 void CMTask::print_stats() { 3909 gclog_or_tty->print_cr("Marking Stats, task = %u, calls = %d", 3910 _worker_id, _calls); 3911 gclog_or_tty->print_cr(" Elapsed time = %1.2lfms, Termination time = %1.2lfms", 3912 _elapsed_time_ms, _termination_time_ms); 3913 gclog_or_tty->print_cr(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms", 3914 _step_times_ms.num(), _step_times_ms.avg(), 3915 _step_times_ms.sd()); 3916 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms", 3917 _step_times_ms.maximum(), _step_times_ms.sum()); 3918 3919 #if _MARKING_STATS_ 3920 gclog_or_tty->print_cr(" Clock Intervals (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms", 3921 _all_clock_intervals_ms.num(), _all_clock_intervals_ms.avg(), 3922 _all_clock_intervals_ms.sd()); 3923 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms", 3924 _all_clock_intervals_ms.maximum(), 3925 _all_clock_intervals_ms.sum()); 3926 gclog_or_tty->print_cr(" Clock Causes (cum): scanning = " SIZE_FORMAT ", marking = " SIZE_FORMAT, 3927 _clock_due_to_scanning, _clock_due_to_marking); 3928 gclog_or_tty->print_cr(" Objects: scanned = " SIZE_FORMAT ", found on the bitmap = " SIZE_FORMAT, 3929 _objs_scanned, _objs_found_on_bitmap); 3930 gclog_or_tty->print_cr(" Local Queue: pushes = " SIZE_FORMAT ", pops = " SIZE_FORMAT ", max size = " SIZE_FORMAT, 3931 _local_pushes, _local_pops, _local_max_size); 3932 gclog_or_tty->print_cr(" Global Stack: pushes = " SIZE_FORMAT ", pops = " SIZE_FORMAT ", max size = " SIZE_FORMAT, 3933 _global_pushes, _global_pops, _global_max_size); 3934 gclog_or_tty->print_cr(" transfers to = " SIZE_FORMAT ", transfers from = " SIZE_FORMAT, 3935 _global_transfers_to,_global_transfers_from); 3936 gclog_or_tty->print_cr(" Regions: claimed = " SIZE_FORMAT, _regions_claimed); 3937 gclog_or_tty->print_cr(" SATB buffers: processed = " SIZE_FORMAT, _satb_buffers_processed); 3938 gclog_or_tty->print_cr(" Steals: attempts = " SIZE_FORMAT ", successes = " SIZE_FORMAT, 3939 _steal_attempts, _steals); 3940 gclog_or_tty->print_cr(" Aborted: " SIZE_FORMAT ", due to", _aborted); 3941 gclog_or_tty->print_cr(" overflow: " SIZE_FORMAT ", global abort: " SIZE_FORMAT ", yield: " SIZE_FORMAT, 3942 _aborted_overflow, _aborted_cm_aborted, _aborted_yield); 3943 gclog_or_tty->print_cr(" time out: " SIZE_FORMAT ", SATB: " SIZE_FORMAT ", termination: " SIZE_FORMAT, 3944 _aborted_timed_out, _aborted_satb, _aborted_termination); 3945 #endif // _MARKING_STATS_ 3946 } 3947 3948 /***************************************************************************** 3949 3950 The do_marking_step(time_target_ms, ...) method is the building 3951 block of the parallel marking framework. It can be called in parallel 3952 with other invocations of do_marking_step() on different tasks 3953 (but only one per task, obviously) and concurrently with the 3954 mutator threads, or during remark, hence it eliminates the need 3955 for two versions of the code. When called during remark, it will 3956 pick up from where the task left off during the concurrent marking 3957 phase. Interestingly, tasks are also claimable during evacuation 3958 pauses too, since do_marking_step() ensures that it aborts before 3959 it needs to yield. 3960 3961 The data structures that it uses to do marking work are the 3962 following: 3963 3964 (1) Marking Bitmap. If there are gray objects that appear only 3965 on the bitmap (this happens either when dealing with an overflow 3966 or when the initial marking phase has simply marked the roots 3967 and didn't push them on the stack), then tasks claim heap 3968 regions whose bitmap they then scan to find gray objects. A 3969 global finger indicates where the end of the last claimed region 3970 is. A local finger indicates how far into the region a task has 3971 scanned. The two fingers are used to determine how to gray an 3972 object (i.e. whether simply marking it is OK, as it will be 3973 visited by a task in the future, or whether it needs to be also 3974 pushed on a stack). 3975 3976 (2) Local Queue. The local queue of the task which is accessed 3977 reasonably efficiently by the task. Other tasks can steal from 3978 it when they run out of work. Throughout the marking phase, a 3979 task attempts to keep its local queue short but not totally 3980 empty, so that entries are available for stealing by other 3981 tasks. Only when there is no more work, a task will totally 3982 drain its local queue. 3983 3984 (3) Global Mark Stack. This handles local queue overflow. During 3985 marking only sets of entries are moved between it and the local 3986 queues, as access to it requires a mutex and more fine-grain 3987 interaction with it which might cause contention. If it 3988 overflows, then the marking phase should restart and iterate 3989 over the bitmap to identify gray objects. Throughout the marking 3990 phase, tasks attempt to keep the global mark stack at a small 3991 length but not totally empty, so that entries are available for 3992 popping by other tasks. Only when there is no more work, tasks 3993 will totally drain the global mark stack. 3994 3995 (4) SATB Buffer Queue. This is where completed SATB buffers are 3996 made available. Buffers are regularly removed from this queue 3997 and scanned for roots, so that the queue doesn't get too 3998 long. During remark, all completed buffers are processed, as 3999 well as the filled in parts of any uncompleted buffers. 4000 4001 The do_marking_step() method tries to abort when the time target 4002 has been reached. There are a few other cases when the 4003 do_marking_step() method also aborts: 4004 4005 (1) When the marking phase has been aborted (after a Full GC). 4006 4007 (2) When a global overflow (on the global stack) has been 4008 triggered. Before the task aborts, it will actually sync up with 4009 the other tasks to ensure that all the marking data structures 4010 (local queues, stacks, fingers etc.) are re-initialized so that 4011 when do_marking_step() completes, the marking phase can 4012 immediately restart. 4013 4014 (3) When enough completed SATB buffers are available. The 4015 do_marking_step() method only tries to drain SATB buffers right 4016 at the beginning. So, if enough buffers are available, the 4017 marking step aborts and the SATB buffers are processed at 4018 the beginning of the next invocation. 4019 4020 (4) To yield. when we have to yield then we abort and yield 4021 right at the end of do_marking_step(). This saves us from a lot 4022 of hassle as, by yielding we might allow a Full GC. If this 4023 happens then objects will be compacted underneath our feet, the 4024 heap might shrink, etc. We save checking for this by just 4025 aborting and doing the yield right at the end. 4026 4027 From the above it follows that the do_marking_step() method should 4028 be called in a loop (or, otherwise, regularly) until it completes. 4029 4030 If a marking step completes without its has_aborted() flag being 4031 true, it means it has completed the current marking phase (and 4032 also all other marking tasks have done so and have all synced up). 4033 4034 A method called regular_clock_call() is invoked "regularly" (in 4035 sub ms intervals) throughout marking. It is this clock method that 4036 checks all the abort conditions which were mentioned above and 4037 decides when the task should abort. A work-based scheme is used to 4038 trigger this clock method: when the number of object words the 4039 marking phase has scanned or the number of references the marking 4040 phase has visited reach a given limit. Additional invocations to 4041 the method clock have been planted in a few other strategic places 4042 too. The initial reason for the clock method was to avoid calling 4043 vtime too regularly, as it is quite expensive. So, once it was in 4044 place, it was natural to piggy-back all the other conditions on it 4045 too and not constantly check them throughout the code. 4046 4047 If do_termination is true then do_marking_step will enter its 4048 termination protocol. 4049 4050 The value of is_serial must be true when do_marking_step is being 4051 called serially (i.e. by the VMThread) and do_marking_step should 4052 skip any synchronization in the termination and overflow code. 4053 Examples include the serial remark code and the serial reference 4054 processing closures. 4055 4056 The value of is_serial must be false when do_marking_step is 4057 being called by any of the worker threads in a work gang. 4058 Examples include the concurrent marking code (CMMarkingTask), 4059 the MT remark code, and the MT reference processing closures. 4060 4061 *****************************************************************************/ 4062 4063 void CMTask::do_marking_step(double time_target_ms, 4064 bool do_termination, 4065 bool is_serial) { 4066 assert(time_target_ms >= 1.0, "minimum granularity is 1ms"); 4067 assert(concurrent() == _cm->concurrent(), "they should be the same"); 4068 4069 G1CollectorPolicy* g1_policy = _g1h->g1_policy(); 4070 assert(_task_queues != NULL, "invariant"); 4071 assert(_task_queue != NULL, "invariant"); 4072 assert(_task_queues->queue(_worker_id) == _task_queue, "invariant"); 4073 4074 assert(!_claimed, 4075 "only one thread should claim this task at any one time"); 4076 4077 // OK, this doesn't safeguard again all possible scenarios, as it is 4078 // possible for two threads to set the _claimed flag at the same 4079 // time. But it is only for debugging purposes anyway and it will 4080 // catch most problems. 4081 _claimed = true; 4082 4083 _start_time_ms = os::elapsedVTime() * 1000.0; 4084 statsOnly( _interval_start_time_ms = _start_time_ms ); 4085 4086 // If do_stealing is true then do_marking_step will attempt to 4087 // steal work from the other CMTasks. It only makes sense to 4088 // enable stealing when the termination protocol is enabled 4089 // and do_marking_step() is not being called serially. 4090 bool do_stealing = do_termination && !is_serial; 4091 4092 double diff_prediction_ms = 4093 g1_policy->get_new_prediction(&_marking_step_diffs_ms); 4094 _time_target_ms = time_target_ms - diff_prediction_ms; 4095 4096 // set up the variables that are used in the work-based scheme to 4097 // call the regular clock method 4098 _words_scanned = 0; 4099 _refs_reached = 0; 4100 recalculate_limits(); 4101 4102 // clear all flags 4103 clear_has_aborted(); 4104 _has_timed_out = false; 4105 _draining_satb_buffers = false; 4106 4107 ++_calls; 4108 4109 if (_cm->verbose_low()) { 4110 gclog_or_tty->print_cr("[%u] >>>>>>>>>> START, call = %d, " 4111 "target = %1.2lfms >>>>>>>>>>", 4112 _worker_id, _calls, _time_target_ms); 4113 } 4114 4115 // Set up the bitmap and oop closures. Anything that uses them is 4116 // eventually called from this method, so it is OK to allocate these 4117 // statically. 4118 CMBitMapClosure bitmap_closure(this, _cm, _nextMarkBitMap); 4119 G1CMOopClosure cm_oop_closure(_g1h, _cm, this); 4120 set_cm_oop_closure(&cm_oop_closure); 4121 4122 if (_cm->has_overflown()) { 4123 // This can happen if the mark stack overflows during a GC pause 4124 // and this task, after a yield point, restarts. We have to abort 4125 // as we need to get into the overflow protocol which happens 4126 // right at the end of this task. 4127 set_has_aborted(); 4128 } 4129 4130 // First drain any available SATB buffers. After this, we will not 4131 // look at SATB buffers before the next invocation of this method. 4132 // If enough completed SATB buffers are queued up, the regular clock 4133 // will abort this task so that it restarts. 4134 drain_satb_buffers(); 4135 // ...then partially drain the local queue and the global stack 4136 drain_local_queue(true); 4137 drain_global_stack(true); 4138 4139 do { 4140 if (!has_aborted() && _curr_region != NULL) { 4141 // This means that we're already holding on to a region. 4142 assert(_finger != NULL, "if region is not NULL, then the finger " 4143 "should not be NULL either"); 4144 4145 // We might have restarted this task after an evacuation pause 4146 // which might have evacuated the region we're holding on to 4147 // underneath our feet. Let's read its limit again to make sure 4148 // that we do not iterate over a region of the heap that 4149 // contains garbage (update_region_limit() will also move 4150 // _finger to the start of the region if it is found empty). 4151 update_region_limit(); 4152 // We will start from _finger not from the start of the region, 4153 // as we might be restarting this task after aborting half-way 4154 // through scanning this region. In this case, _finger points to 4155 // the address where we last found a marked object. If this is a 4156 // fresh region, _finger points to start(). 4157 MemRegion mr = MemRegion(_finger, _region_limit); 4158 4159 if (_cm->verbose_low()) { 4160 gclog_or_tty->print_cr("[%u] we're scanning part " 4161 "["PTR_FORMAT", "PTR_FORMAT") " 4162 "of region "HR_FORMAT, 4163 _worker_id, p2i(_finger), p2i(_region_limit), 4164 HR_FORMAT_PARAMS(_curr_region)); 4165 } 4166 4167 assert(!_curr_region->is_humongous() || mr.start() == _curr_region->bottom(), 4168 "humongous regions should go around loop once only"); 4169 4170 // Some special cases: 4171 // If the memory region is empty, we can just give up the region. 4172 // If the current region is humongous then we only need to check 4173 // the bitmap for the bit associated with the start of the object, 4174 // scan the object if it's live, and give up the region. 4175 // Otherwise, let's iterate over the bitmap of the part of the region 4176 // that is left. 4177 // If the iteration is successful, give up the region. 4178 if (mr.is_empty()) { 4179 giveup_current_region(); 4180 regular_clock_call(); 4181 } else if (_curr_region->is_humongous() && mr.start() == _curr_region->bottom()) { 4182 if (_nextMarkBitMap->isMarked(mr.start())) { 4183 // The object is marked - apply the closure 4184 BitMap::idx_t offset = _nextMarkBitMap->heapWordToOffset(mr.start()); 4185 bitmap_closure.do_bit(offset); 4186 } 4187 // Even if this task aborted while scanning the humongous object 4188 // we can (and should) give up the current region. 4189 giveup_current_region(); 4190 regular_clock_call(); 4191 } else if (_nextMarkBitMap->iterate(&bitmap_closure, mr)) { 4192 giveup_current_region(); 4193 regular_clock_call(); 4194 } else { 4195 assert(has_aborted(), "currently the only way to do so"); 4196 // The only way to abort the bitmap iteration is to return 4197 // false from the do_bit() method. However, inside the 4198 // do_bit() method we move the _finger to point to the 4199 // object currently being looked at. So, if we bail out, we 4200 // have definitely set _finger to something non-null. 4201 assert(_finger != NULL, "invariant"); 4202 4203 // Region iteration was actually aborted. So now _finger 4204 // points to the address of the object we last scanned. If we 4205 // leave it there, when we restart this task, we will rescan 4206 // the object. It is easy to avoid this. We move the finger by 4207 // enough to point to the next possible object header (the 4208 // bitmap knows by how much we need to move it as it knows its 4209 // granularity). 4210 assert(_finger < _region_limit, "invariant"); 4211 HeapWord* new_finger = _nextMarkBitMap->nextObject(_finger); 4212 // Check if bitmap iteration was aborted while scanning the last object 4213 if (new_finger >= _region_limit) { 4214 giveup_current_region(); 4215 } else { 4216 move_finger_to(new_finger); 4217 } 4218 } 4219 } 4220 // At this point we have either completed iterating over the 4221 // region we were holding on to, or we have aborted. 4222 4223 // We then partially drain the local queue and the global stack. 4224 // (Do we really need this?) 4225 drain_local_queue(true); 4226 drain_global_stack(true); 4227 4228 // Read the note on the claim_region() method on why it might 4229 // return NULL with potentially more regions available for 4230 // claiming and why we have to check out_of_regions() to determine 4231 // whether we're done or not. 4232 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) { 4233 // We are going to try to claim a new region. We should have 4234 // given up on the previous one. 4235 // Separated the asserts so that we know which one fires. 4236 assert(_curr_region == NULL, "invariant"); 4237 assert(_finger == NULL, "invariant"); 4238 assert(_region_limit == NULL, "invariant"); 4239 if (_cm->verbose_low()) { 4240 gclog_or_tty->print_cr("[%u] trying to claim a new region", _worker_id); 4241 } 4242 HeapRegion* claimed_region = _cm->claim_region(_worker_id); 4243 if (claimed_region != NULL) { 4244 // Yes, we managed to claim one 4245 statsOnly( ++_regions_claimed ); 4246 4247 if (_cm->verbose_low()) { 4248 gclog_or_tty->print_cr("[%u] we successfully claimed " 4249 "region "PTR_FORMAT, 4250 _worker_id, p2i(claimed_region)); 4251 } 4252 4253 setup_for_region(claimed_region); 4254 assert(_curr_region == claimed_region, "invariant"); 4255 } 4256 // It is important to call the regular clock here. It might take 4257 // a while to claim a region if, for example, we hit a large 4258 // block of empty regions. So we need to call the regular clock 4259 // method once round the loop to make sure it's called 4260 // frequently enough. 4261 regular_clock_call(); 4262 } 4263 4264 if (!has_aborted() && _curr_region == NULL) { 4265 assert(_cm->out_of_regions(), 4266 "at this point we should be out of regions"); 4267 } 4268 } while ( _curr_region != NULL && !has_aborted()); 4269 4270 if (!has_aborted()) { 4271 // We cannot check whether the global stack is empty, since other 4272 // tasks might be pushing objects to it concurrently. 4273 assert(_cm->out_of_regions(), 4274 "at this point we should be out of regions"); 4275 4276 if (_cm->verbose_low()) { 4277 gclog_or_tty->print_cr("[%u] all regions claimed", _worker_id); 4278 } 4279 4280 // Try to reduce the number of available SATB buffers so that 4281 // remark has less work to do. 4282 drain_satb_buffers(); 4283 } 4284 4285 // Since we've done everything else, we can now totally drain the 4286 // local queue and global stack. 4287 drain_local_queue(false); 4288 drain_global_stack(false); 4289 4290 // Attempt at work stealing from other task's queues. 4291 if (do_stealing && !has_aborted()) { 4292 // We have not aborted. This means that we have finished all that 4293 // we could. Let's try to do some stealing... 4294 4295 // We cannot check whether the global stack is empty, since other 4296 // tasks might be pushing objects to it concurrently. 4297 assert(_cm->out_of_regions() && _task_queue->size() == 0, 4298 "only way to reach here"); 4299 4300 if (_cm->verbose_low()) { 4301 gclog_or_tty->print_cr("[%u] starting to steal", _worker_id); 4302 } 4303 4304 while (!has_aborted()) { 4305 oop obj; 4306 statsOnly( ++_steal_attempts ); 4307 4308 if (_cm->try_stealing(_worker_id, &_hash_seed, obj)) { 4309 if (_cm->verbose_medium()) { 4310 gclog_or_tty->print_cr("[%u] stolen "PTR_FORMAT" successfully", 4311 _worker_id, p2i((void*) obj)); 4312 } 4313 4314 statsOnly( ++_steals ); 4315 4316 assert(_nextMarkBitMap->isMarked((HeapWord*) obj), 4317 "any stolen object should be marked"); 4318 scan_object(obj); 4319 4320 // And since we're towards the end, let's totally drain the 4321 // local queue and global stack. 4322 drain_local_queue(false); 4323 drain_global_stack(false); 4324 } else { 4325 break; 4326 } 4327 } 4328 } 4329 4330 // If we are about to wrap up and go into termination, check if we 4331 // should raise the overflow flag. 4332 if (do_termination && !has_aborted()) { 4333 if (_cm->force_overflow()->should_force()) { 4334 _cm->set_has_overflown(); 4335 regular_clock_call(); 4336 } 4337 } 4338 4339 // We still haven't aborted. Now, let's try to get into the 4340 // termination protocol. 4341 if (do_termination && !has_aborted()) { 4342 // We cannot check whether the global stack is empty, since other 4343 // tasks might be concurrently pushing objects on it. 4344 // Separated the asserts so that we know which one fires. 4345 assert(_cm->out_of_regions(), "only way to reach here"); 4346 assert(_task_queue->size() == 0, "only way to reach here"); 4347 4348 if (_cm->verbose_low()) { 4349 gclog_or_tty->print_cr("[%u] starting termination protocol", _worker_id); 4350 } 4351 4352 _termination_start_time_ms = os::elapsedVTime() * 1000.0; 4353 4354 // The CMTask class also extends the TerminatorTerminator class, 4355 // hence its should_exit_termination() method will also decide 4356 // whether to exit the termination protocol or not. 4357 bool finished = (is_serial || 4358 _cm->terminator()->offer_termination(this)); 4359 double termination_end_time_ms = os::elapsedVTime() * 1000.0; 4360 _termination_time_ms += 4361 termination_end_time_ms - _termination_start_time_ms; 4362 4363 if (finished) { 4364 // We're all done. 4365 4366 if (_worker_id == 0) { 4367 // let's allow task 0 to do this 4368 if (concurrent()) { 4369 assert(_cm->concurrent_marking_in_progress(), "invariant"); 4370 // we need to set this to false before the next 4371 // safepoint. This way we ensure that the marking phase 4372 // doesn't observe any more heap expansions. 4373 _cm->clear_concurrent_marking_in_progress(); 4374 } 4375 } 4376 4377 // We can now guarantee that the global stack is empty, since 4378 // all other tasks have finished. We separated the guarantees so 4379 // that, if a condition is false, we can immediately find out 4380 // which one. 4381 guarantee(_cm->out_of_regions(), "only way to reach here"); 4382 guarantee(_cm->mark_stack_empty(), "only way to reach here"); 4383 guarantee(_task_queue->size() == 0, "only way to reach here"); 4384 guarantee(!_cm->has_overflown(), "only way to reach here"); 4385 guarantee(!_cm->mark_stack_overflow(), "only way to reach here"); 4386 4387 if (_cm->verbose_low()) { 4388 gclog_or_tty->print_cr("[%u] all tasks terminated", _worker_id); 4389 } 4390 } else { 4391 // Apparently there's more work to do. Let's abort this task. It 4392 // will restart it and we can hopefully find more things to do. 4393 4394 if (_cm->verbose_low()) { 4395 gclog_or_tty->print_cr("[%u] apparently there is more work to do", 4396 _worker_id); 4397 } 4398 4399 set_has_aborted(); 4400 statsOnly( ++_aborted_termination ); 4401 } 4402 } 4403 4404 // Mainly for debugging purposes to make sure that a pointer to the 4405 // closure which was statically allocated in this frame doesn't 4406 // escape it by accident. 4407 set_cm_oop_closure(NULL); 4408 double end_time_ms = os::elapsedVTime() * 1000.0; 4409 double elapsed_time_ms = end_time_ms - _start_time_ms; 4410 // Update the step history. 4411 _step_times_ms.add(elapsed_time_ms); 4412 4413 if (has_aborted()) { 4414 // The task was aborted for some reason. 4415 4416 statsOnly( ++_aborted ); 4417 4418 if (_has_timed_out) { 4419 double diff_ms = elapsed_time_ms - _time_target_ms; 4420 // Keep statistics of how well we did with respect to hitting 4421 // our target only if we actually timed out (if we aborted for 4422 // other reasons, then the results might get skewed). 4423 _marking_step_diffs_ms.add(diff_ms); 4424 } 4425 4426 if (_cm->has_overflown()) { 4427 // This is the interesting one. We aborted because a global 4428 // overflow was raised. This means we have to restart the 4429 // marking phase and start iterating over regions. However, in 4430 // order to do this we have to make sure that all tasks stop 4431 // what they are doing and re-initialize in a safe manner. We 4432 // will achieve this with the use of two barrier sync points. 4433 4434 if (_cm->verbose_low()) { 4435 gclog_or_tty->print_cr("[%u] detected overflow", _worker_id); 4436 } 4437 4438 if (!is_serial) { 4439 // We only need to enter the sync barrier if being called 4440 // from a parallel context 4441 _cm->enter_first_sync_barrier(_worker_id); 4442 4443 // When we exit this sync barrier we know that all tasks have 4444 // stopped doing marking work. So, it's now safe to 4445 // re-initialize our data structures. At the end of this method, 4446 // task 0 will clear the global data structures. 4447 } 4448 4449 statsOnly( ++_aborted_overflow ); 4450 4451 // We clear the local state of this task... 4452 clear_region_fields(); 4453 4454 if (!is_serial) { 4455 // ...and enter the second barrier. 4456 _cm->enter_second_sync_barrier(_worker_id); 4457 } 4458 // At this point, if we're during the concurrent phase of 4459 // marking, everything has been re-initialized and we're 4460 // ready to restart. 4461 } 4462 4463 if (_cm->verbose_low()) { 4464 gclog_or_tty->print_cr("[%u] <<<<<<<<<< ABORTING, target = %1.2lfms, " 4465 "elapsed = %1.2lfms <<<<<<<<<<", 4466 _worker_id, _time_target_ms, elapsed_time_ms); 4467 if (_cm->has_aborted()) { 4468 gclog_or_tty->print_cr("[%u] ========== MARKING ABORTED ==========", 4469 _worker_id); 4470 } 4471 } 4472 } else { 4473 if (_cm->verbose_low()) { 4474 gclog_or_tty->print_cr("[%u] <<<<<<<<<< FINISHED, target = %1.2lfms, " 4475 "elapsed = %1.2lfms <<<<<<<<<<", 4476 _worker_id, _time_target_ms, elapsed_time_ms); 4477 } 4478 } 4479 4480 _claimed = false; 4481 } 4482 4483 CMTask::CMTask(uint worker_id, 4484 ConcurrentMark* cm, 4485 size_t* marked_bytes, 4486 BitMap* card_bm, 4487 CMTaskQueue* task_queue, 4488 CMTaskQueueSet* task_queues) 4489 : _g1h(G1CollectedHeap::heap()), 4490 _worker_id(worker_id), _cm(cm), 4491 _claimed(false), 4492 _nextMarkBitMap(NULL), _hash_seed(17), 4493 _task_queue(task_queue), 4494 _task_queues(task_queues), 4495 _cm_oop_closure(NULL), 4496 _marked_bytes_array(marked_bytes), 4497 _card_bm(card_bm) { 4498 guarantee(task_queue != NULL, "invariant"); 4499 guarantee(task_queues != NULL, "invariant"); 4500 4501 statsOnly( _clock_due_to_scanning = 0; 4502 _clock_due_to_marking = 0 ); 4503 4504 _marking_step_diffs_ms.add(0.5); 4505 } 4506 4507 // These are formatting macros that are used below to ensure 4508 // consistent formatting. The *_H_* versions are used to format the 4509 // header for a particular value and they should be kept consistent 4510 // with the corresponding macro. Also note that most of the macros add 4511 // the necessary white space (as a prefix) which makes them a bit 4512 // easier to compose. 4513 4514 // All the output lines are prefixed with this string to be able to 4515 // identify them easily in a large log file. 4516 #define G1PPRL_LINE_PREFIX "###" 4517 4518 #define G1PPRL_ADDR_BASE_FORMAT " "PTR_FORMAT"-"PTR_FORMAT 4519 #ifdef _LP64 4520 #define G1PPRL_ADDR_BASE_H_FORMAT " %37s" 4521 #else // _LP64 4522 #define G1PPRL_ADDR_BASE_H_FORMAT " %21s" 4523 #endif // _LP64 4524 4525 // For per-region info 4526 #define G1PPRL_TYPE_FORMAT " %-4s" 4527 #define G1PPRL_TYPE_H_FORMAT " %4s" 4528 #define G1PPRL_BYTE_FORMAT " "SIZE_FORMAT_W(9) 4529 #define G1PPRL_BYTE_H_FORMAT " %9s" 4530 #define G1PPRL_DOUBLE_FORMAT " %14.1f" 4531 #define G1PPRL_DOUBLE_H_FORMAT " %14s" 4532 4533 // For summary info 4534 #define G1PPRL_SUM_ADDR_FORMAT(tag) " "tag":"G1PPRL_ADDR_BASE_FORMAT 4535 #define G1PPRL_SUM_BYTE_FORMAT(tag) " "tag": "SIZE_FORMAT 4536 #define G1PPRL_SUM_MB_FORMAT(tag) " "tag": %1.2f MB" 4537 #define G1PPRL_SUM_MB_PERC_FORMAT(tag) G1PPRL_SUM_MB_FORMAT(tag)" / %1.2f %%" 4538 4539 G1PrintRegionLivenessInfoClosure:: 4540 G1PrintRegionLivenessInfoClosure(outputStream* out, const char* phase_name) 4541 : _out(out), 4542 _total_used_bytes(0), _total_capacity_bytes(0), 4543 _total_prev_live_bytes(0), _total_next_live_bytes(0), 4544 _hum_used_bytes(0), _hum_capacity_bytes(0), 4545 _hum_prev_live_bytes(0), _hum_next_live_bytes(0), 4546 _total_remset_bytes(0), _total_strong_code_roots_bytes(0) { 4547 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 4548 MemRegion g1_reserved = g1h->g1_reserved(); 4549 double now = os::elapsedTime(); 4550 4551 // Print the header of the output. 4552 _out->cr(); 4553 _out->print_cr(G1PPRL_LINE_PREFIX" PHASE %s @ %1.3f", phase_name, now); 4554 _out->print_cr(G1PPRL_LINE_PREFIX" HEAP" 4555 G1PPRL_SUM_ADDR_FORMAT("reserved") 4556 G1PPRL_SUM_BYTE_FORMAT("region-size"), 4557 p2i(g1_reserved.start()), p2i(g1_reserved.end()), 4558 HeapRegion::GrainBytes); 4559 _out->print_cr(G1PPRL_LINE_PREFIX); 4560 _out->print_cr(G1PPRL_LINE_PREFIX 4561 G1PPRL_TYPE_H_FORMAT 4562 G1PPRL_ADDR_BASE_H_FORMAT 4563 G1PPRL_BYTE_H_FORMAT 4564 G1PPRL_BYTE_H_FORMAT 4565 G1PPRL_BYTE_H_FORMAT 4566 G1PPRL_DOUBLE_H_FORMAT 4567 G1PPRL_BYTE_H_FORMAT 4568 G1PPRL_BYTE_H_FORMAT, 4569 "type", "address-range", 4570 "used", "prev-live", "next-live", "gc-eff", 4571 "remset", "code-roots"); 4572 _out->print_cr(G1PPRL_LINE_PREFIX 4573 G1PPRL_TYPE_H_FORMAT 4574 G1PPRL_ADDR_BASE_H_FORMAT 4575 G1PPRL_BYTE_H_FORMAT 4576 G1PPRL_BYTE_H_FORMAT 4577 G1PPRL_BYTE_H_FORMAT 4578 G1PPRL_DOUBLE_H_FORMAT 4579 G1PPRL_BYTE_H_FORMAT 4580 G1PPRL_BYTE_H_FORMAT, 4581 "", "", 4582 "(bytes)", "(bytes)", "(bytes)", "(bytes/ms)", 4583 "(bytes)", "(bytes)"); 4584 } 4585 4586 // It takes as a parameter a reference to one of the _hum_* fields, it 4587 // deduces the corresponding value for a region in a humongous region 4588 // series (either the region size, or what's left if the _hum_* field 4589 // is < the region size), and updates the _hum_* field accordingly. 4590 size_t G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* hum_bytes) { 4591 size_t bytes = 0; 4592 // The > 0 check is to deal with the prev and next live bytes which 4593 // could be 0. 4594 if (*hum_bytes > 0) { 4595 bytes = MIN2(HeapRegion::GrainBytes, *hum_bytes); 4596 *hum_bytes -= bytes; 4597 } 4598 return bytes; 4599 } 4600 4601 // It deduces the values for a region in a humongous region series 4602 // from the _hum_* fields and updates those accordingly. It assumes 4603 // that that _hum_* fields have already been set up from the "starts 4604 // humongous" region and we visit the regions in address order. 4605 void G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* used_bytes, 4606 size_t* capacity_bytes, 4607 size_t* prev_live_bytes, 4608 size_t* next_live_bytes) { 4609 assert(_hum_used_bytes > 0 && _hum_capacity_bytes > 0, "pre-condition"); 4610 *used_bytes = get_hum_bytes(&_hum_used_bytes); 4611 *capacity_bytes = get_hum_bytes(&_hum_capacity_bytes); 4612 *prev_live_bytes = get_hum_bytes(&_hum_prev_live_bytes); 4613 *next_live_bytes = get_hum_bytes(&_hum_next_live_bytes); 4614 } 4615 4616 bool G1PrintRegionLivenessInfoClosure::doHeapRegion(HeapRegion* r) { 4617 const char* type = r->get_type_str(); 4618 HeapWord* bottom = r->bottom(); 4619 HeapWord* end = r->end(); 4620 size_t capacity_bytes = r->capacity(); 4621 size_t used_bytes = r->used(); 4622 size_t prev_live_bytes = r->live_bytes(); 4623 size_t next_live_bytes = r->next_live_bytes(); 4624 double gc_eff = r->gc_efficiency(); 4625 size_t remset_bytes = r->rem_set()->mem_size(); 4626 size_t strong_code_roots_bytes = r->rem_set()->strong_code_roots_mem_size(); 4627 4628 if (r->is_starts_humongous()) { 4629 assert(_hum_used_bytes == 0 && _hum_capacity_bytes == 0 && 4630 _hum_prev_live_bytes == 0 && _hum_next_live_bytes == 0, 4631 "they should have been zeroed after the last time we used them"); 4632 // Set up the _hum_* fields. 4633 _hum_capacity_bytes = capacity_bytes; 4634 _hum_used_bytes = used_bytes; 4635 _hum_prev_live_bytes = prev_live_bytes; 4636 _hum_next_live_bytes = next_live_bytes; 4637 get_hum_bytes(&used_bytes, &capacity_bytes, 4638 &prev_live_bytes, &next_live_bytes); 4639 end = bottom + HeapRegion::GrainWords; 4640 } else if (r->is_continues_humongous()) { 4641 get_hum_bytes(&used_bytes, &capacity_bytes, 4642 &prev_live_bytes, &next_live_bytes); 4643 assert(end == bottom + HeapRegion::GrainWords, "invariant"); 4644 } 4645 4646 _total_used_bytes += used_bytes; 4647 _total_capacity_bytes += capacity_bytes; 4648 _total_prev_live_bytes += prev_live_bytes; 4649 _total_next_live_bytes += next_live_bytes; 4650 _total_remset_bytes += remset_bytes; 4651 _total_strong_code_roots_bytes += strong_code_roots_bytes; 4652 4653 // Print a line for this particular region. 4654 _out->print_cr(G1PPRL_LINE_PREFIX 4655 G1PPRL_TYPE_FORMAT 4656 G1PPRL_ADDR_BASE_FORMAT 4657 G1PPRL_BYTE_FORMAT 4658 G1PPRL_BYTE_FORMAT 4659 G1PPRL_BYTE_FORMAT 4660 G1PPRL_DOUBLE_FORMAT 4661 G1PPRL_BYTE_FORMAT 4662 G1PPRL_BYTE_FORMAT, 4663 type, p2i(bottom), p2i(end), 4664 used_bytes, prev_live_bytes, next_live_bytes, gc_eff, 4665 remset_bytes, strong_code_roots_bytes); 4666 4667 return false; 4668 } 4669 4670 G1PrintRegionLivenessInfoClosure::~G1PrintRegionLivenessInfoClosure() { 4671 // add static memory usages to remembered set sizes 4672 _total_remset_bytes += HeapRegionRemSet::fl_mem_size() + HeapRegionRemSet::static_mem_size(); 4673 // Print the footer of the output. 4674 _out->print_cr(G1PPRL_LINE_PREFIX); 4675 _out->print_cr(G1PPRL_LINE_PREFIX 4676 " SUMMARY" 4677 G1PPRL_SUM_MB_FORMAT("capacity") 4678 G1PPRL_SUM_MB_PERC_FORMAT("used") 4679 G1PPRL_SUM_MB_PERC_FORMAT("prev-live") 4680 G1PPRL_SUM_MB_PERC_FORMAT("next-live") 4681 G1PPRL_SUM_MB_FORMAT("remset") 4682 G1PPRL_SUM_MB_FORMAT("code-roots"), 4683 bytes_to_mb(_total_capacity_bytes), 4684 bytes_to_mb(_total_used_bytes), 4685 perc(_total_used_bytes, _total_capacity_bytes), 4686 bytes_to_mb(_total_prev_live_bytes), 4687 perc(_total_prev_live_bytes, _total_capacity_bytes), 4688 bytes_to_mb(_total_next_live_bytes), 4689 perc(_total_next_live_bytes, _total_capacity_bytes), 4690 bytes_to_mb(_total_remset_bytes), 4691 bytes_to_mb(_total_strong_code_roots_bytes)); 4692 _out->cr(); 4693 }