Mercurial > hg > truffle
annotate src/share/vm/gc_implementation/g1/concurrentMark.cpp @ 639:922c573ea67d
6815683: G1: SEGV during marking
Summary: We should mark the regions that continue humongous regions as live if the first region is live.
Reviewed-by: tonyp
| author | iveresov |
|---|---|
| date | Mon, 16 Mar 2009 17:48:01 -0700 |
| parents | 7ea5ca260b28 |
| children | bd441136a5ce |
| rev | line source |
|---|---|
| 342 | 1 /* |
| 470 | 2 * Copyright 2001-2008 Sun Microsystems, Inc. All Rights Reserved. |
| 342 | 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. |
| 4 * | |
| 5 * This code is free software; you can redistribute it and/or modify it | |
| 6 * under the terms of the GNU General Public License version 2 only, as | |
| 7 * published by the Free Software Foundation. | |
| 8 * | |
| 9 * This code is distributed in the hope that it will be useful, but WITHOUT | |
| 10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
| 11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License | |
| 12 * version 2 for more details (a copy is included in the LICENSE file that | |
| 13 * accompanied this code). | |
| 14 * | |
| 15 * You should have received a copy of the GNU General Public License version | |
| 16 * 2 along with this work; if not, write to the Free Software Foundation, | |
| 17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. | |
| 18 * | |
| 19 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, | |
| 20 * CA 95054 USA or visit www.sun.com if you need additional information or | |
| 21 * have any questions. | |
| 22 * | |
| 23 */ | |
| 24 | |
| 25 #include "incls/_precompiled.incl" | |
| 26 #include "incls/_concurrentMark.cpp.incl" | |
| 27 | |
| 28 // | |
| 29 // CMS Bit Map Wrapper | |
| 30 | |
| 31 CMBitMapRO::CMBitMapRO(ReservedSpace rs, int shifter): | |
| 32 _bm((uintptr_t*)NULL,0), | |
| 33 _shifter(shifter) { | |
| 34 _bmStartWord = (HeapWord*)(rs.base()); | |
| 35 _bmWordSize = rs.size()/HeapWordSize; // rs.size() is in bytes | |
| 36 ReservedSpace brs(ReservedSpace::allocation_align_size_up( | |
| 37 (_bmWordSize >> (_shifter + LogBitsPerByte)) + 1)); | |
| 38 | |
| 39 guarantee(brs.is_reserved(), "couldn't allocate CMS bit map"); | |
| 40 // For now we'll just commit all of the bit map up fromt. | |
| 41 // Later on we'll try to be more parsimonious with swap. | |
| 42 guarantee(_virtual_space.initialize(brs, brs.size()), | |
| 43 "couldn't reseve backing store for CMS bit map"); | |
| 44 assert(_virtual_space.committed_size() == brs.size(), | |
| 45 "didn't reserve backing store for all of CMS bit map?"); | |
| 46 _bm.set_map((uintptr_t*)_virtual_space.low()); | |
| 47 assert(_virtual_space.committed_size() << (_shifter + LogBitsPerByte) >= | |
| 48 _bmWordSize, "inconsistency in bit map sizing"); | |
| 49 _bm.set_size(_bmWordSize >> _shifter); | |
| 50 } | |
| 51 | |
| 52 HeapWord* CMBitMapRO::getNextMarkedWordAddress(HeapWord* addr, | |
| 53 HeapWord* limit) const { | |
| 54 // First we must round addr *up* to a possible object boundary. | |
| 55 addr = (HeapWord*)align_size_up((intptr_t)addr, | |
| 56 HeapWordSize << _shifter); | |
| 57 size_t addrOffset = heapWordToOffset(addr); | |
| 58 if (limit == NULL) limit = _bmStartWord + _bmWordSize; | |
| 59 size_t limitOffset = heapWordToOffset(limit); | |
| 60 size_t nextOffset = _bm.get_next_one_offset(addrOffset, limitOffset); | |
| 61 HeapWord* nextAddr = offsetToHeapWord(nextOffset); | |
| 62 assert(nextAddr >= addr, "get_next_one postcondition"); | |
| 63 assert(nextAddr == limit || isMarked(nextAddr), | |
| 64 "get_next_one postcondition"); | |
| 65 return nextAddr; | |
| 66 } | |
| 67 | |
| 68 HeapWord* CMBitMapRO::getNextUnmarkedWordAddress(HeapWord* addr, | |
| 69 HeapWord* limit) const { | |
| 70 size_t addrOffset = heapWordToOffset(addr); | |
| 71 if (limit == NULL) limit = _bmStartWord + _bmWordSize; | |
| 72 size_t limitOffset = heapWordToOffset(limit); | |
| 73 size_t nextOffset = _bm.get_next_zero_offset(addrOffset, limitOffset); | |
| 74 HeapWord* nextAddr = offsetToHeapWord(nextOffset); | |
| 75 assert(nextAddr >= addr, "get_next_one postcondition"); | |
| 76 assert(nextAddr == limit || !isMarked(nextAddr), | |
| 77 "get_next_one postcondition"); | |
| 78 return nextAddr; | |
| 79 } | |
| 80 | |
| 81 int CMBitMapRO::heapWordDiffToOffsetDiff(size_t diff) const { | |
| 82 assert((diff & ((1 << _shifter) - 1)) == 0, "argument check"); | |
| 83 return (int) (diff >> _shifter); | |
| 84 } | |
| 85 | |
| 86 bool CMBitMapRO::iterate(BitMapClosure* cl, MemRegion mr) { | |
| 87 HeapWord* left = MAX2(_bmStartWord, mr.start()); | |
| 88 HeapWord* right = MIN2(_bmStartWord + _bmWordSize, mr.end()); | |
| 89 if (right > left) { | |
| 90 // Right-open interval [leftOffset, rightOffset). | |
| 91 return _bm.iterate(cl, heapWordToOffset(left), heapWordToOffset(right)); | |
| 92 } else { | |
| 93 return true; | |
| 94 } | |
| 95 } | |
| 96 | |
| 97 void CMBitMapRO::mostly_disjoint_range_union(BitMap* from_bitmap, | |
| 98 size_t from_start_index, | |
| 99 HeapWord* to_start_word, | |
| 100 size_t word_num) { | |
| 101 _bm.mostly_disjoint_range_union(from_bitmap, | |
| 102 from_start_index, | |
| 103 heapWordToOffset(to_start_word), | |
| 104 word_num); | |
| 105 } | |
| 106 | |
| 107 #ifndef PRODUCT | |
| 108 bool CMBitMapRO::covers(ReservedSpace rs) const { | |
| 109 // assert(_bm.map() == _virtual_space.low(), "map inconsistency"); | |
| 110 assert(((size_t)_bm.size() * (1 << _shifter)) == _bmWordSize, | |
| 111 "size inconsistency"); | |
| 112 return _bmStartWord == (HeapWord*)(rs.base()) && | |
| 113 _bmWordSize == rs.size()>>LogHeapWordSize; | |
| 114 } | |
| 115 #endif | |
| 116 | |
| 117 void CMBitMap::clearAll() { | |
| 118 _bm.clear(); | |
| 119 return; | |
| 120 } | |
| 121 | |
| 122 void CMBitMap::markRange(MemRegion mr) { | |
| 123 mr.intersection(MemRegion(_bmStartWord, _bmWordSize)); | |
| 124 assert(!mr.is_empty(), "unexpected empty region"); | |
| 125 assert((offsetToHeapWord(heapWordToOffset(mr.end())) == | |
| 126 ((HeapWord *) mr.end())), | |
| 127 "markRange memory region end is not card aligned"); | |
| 128 // convert address range into offset range | |
| 129 _bm.at_put_range(heapWordToOffset(mr.start()), | |
| 130 heapWordToOffset(mr.end()), true); | |
| 131 } | |
| 132 | |
| 133 void CMBitMap::clearRange(MemRegion mr) { | |
| 134 mr.intersection(MemRegion(_bmStartWord, _bmWordSize)); | |
| 135 assert(!mr.is_empty(), "unexpected empty region"); | |
| 136 // convert address range into offset range | |
| 137 _bm.at_put_range(heapWordToOffset(mr.start()), | |
| 138 heapWordToOffset(mr.end()), false); | |
| 139 } | |
| 140 | |
| 141 MemRegion CMBitMap::getAndClearMarkedRegion(HeapWord* addr, | |
| 142 HeapWord* end_addr) { | |
| 143 HeapWord* start = getNextMarkedWordAddress(addr); | |
| 144 start = MIN2(start, end_addr); | |
| 145 HeapWord* end = getNextUnmarkedWordAddress(start); | |
| 146 end = MIN2(end, end_addr); | |
| 147 assert(start <= end, "Consistency check"); | |
| 148 MemRegion mr(start, end); | |
| 149 if (!mr.is_empty()) { | |
| 150 clearRange(mr); | |
| 151 } | |
| 152 return mr; | |
| 153 } | |
| 154 | |
| 155 CMMarkStack::CMMarkStack(ConcurrentMark* cm) : | |
| 156 _base(NULL), _cm(cm) | |
| 157 #ifdef ASSERT | |
| 158 , _drain_in_progress(false) | |
| 159 , _drain_in_progress_yields(false) | |
| 160 #endif | |
| 161 {} | |
| 162 | |
| 163 void CMMarkStack::allocate(size_t size) { | |
| 164 _base = NEW_C_HEAP_ARRAY(oop, size); | |
| 165 if (_base == NULL) | |
| 166 vm_exit_during_initialization("Failed to allocate " | |
| 167 "CM region mark stack"); | |
| 168 _index = 0; | |
| 169 // QQQQ cast ... | |
| 170 _capacity = (jint) size; | |
| 171 _oops_do_bound = -1; | |
| 172 NOT_PRODUCT(_max_depth = 0); | |
| 173 } | |
| 174 | |
| 175 CMMarkStack::~CMMarkStack() { | |
| 176 if (_base != NULL) FREE_C_HEAP_ARRAY(oop, _base); | |
| 177 } | |
| 178 | |
| 179 void CMMarkStack::par_push(oop ptr) { | |
| 180 while (true) { | |
| 181 if (isFull()) { | |
| 182 _overflow = true; | |
| 183 return; | |
| 184 } | |
| 185 // Otherwise... | |
| 186 jint index = _index; | |
| 187 jint next_index = index+1; | |
| 188 jint res = Atomic::cmpxchg(next_index, &_index, index); | |
| 189 if (res == index) { | |
| 190 _base[index] = ptr; | |
| 191 // Note that we don't maintain this atomically. We could, but it | |
| 192 // doesn't seem necessary. | |
| 193 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index)); | |
| 194 return; | |
| 195 } | |
| 196 // Otherwise, we need to try again. | |
| 197 } | |
| 198 } | |
| 199 | |
| 200 void CMMarkStack::par_adjoin_arr(oop* ptr_arr, int n) { | |
| 201 while (true) { | |
| 202 if (isFull()) { | |
| 203 _overflow = true; | |
| 204 return; | |
| 205 } | |
| 206 // Otherwise... | |
| 207 jint index = _index; | |
| 208 jint next_index = index + n; | |
| 209 if (next_index > _capacity) { | |
| 210 _overflow = true; | |
| 211 return; | |
| 212 } | |
| 213 jint res = Atomic::cmpxchg(next_index, &_index, index); | |
| 214 if (res == index) { | |
| 215 for (int i = 0; i < n; i++) { | |
| 216 int ind = index + i; | |
| 217 assert(ind < _capacity, "By overflow test above."); | |
| 218 _base[ind] = ptr_arr[i]; | |
| 219 } | |
| 220 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index)); | |
| 221 return; | |
| 222 } | |
| 223 // Otherwise, we need to try again. | |
| 224 } | |
| 225 } | |
| 226 | |
| 227 | |
| 228 void CMMarkStack::par_push_arr(oop* ptr_arr, int n) { | |
| 229 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); | |
| 230 jint start = _index; | |
| 231 jint next_index = start + n; | |
| 232 if (next_index > _capacity) { | |
| 233 _overflow = true; | |
| 234 return; | |
| 235 } | |
| 236 // Otherwise. | |
| 237 _index = next_index; | |
| 238 for (int i = 0; i < n; i++) { | |
| 239 int ind = start + i; | |
| 240 guarantee(ind < _capacity, "By overflow test above."); | |
| 241 _base[ind] = ptr_arr[i]; | |
| 242 } | |
| 243 } | |
| 244 | |
| 245 | |
| 246 bool CMMarkStack::par_pop_arr(oop* ptr_arr, int max, int* n) { | |
| 247 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); | |
| 248 jint index = _index; | |
| 249 if (index == 0) { | |
| 250 *n = 0; | |
| 251 return false; | |
| 252 } else { | |
| 253 int k = MIN2(max, index); | |
| 254 jint new_ind = index - k; | |
| 255 for (int j = 0; j < k; j++) { | |
| 256 ptr_arr[j] = _base[new_ind + j]; | |
| 257 } | |
| 258 _index = new_ind; | |
| 259 *n = k; | |
| 260 return true; | |
| 261 } | |
| 262 } | |
| 263 | |
| 264 | |
| 265 CMRegionStack::CMRegionStack() : _base(NULL) {} | |
| 266 | |
| 267 void CMRegionStack::allocate(size_t size) { | |
| 268 _base = NEW_C_HEAP_ARRAY(MemRegion, size); | |
| 269 if (_base == NULL) | |
| 270 vm_exit_during_initialization("Failed to allocate " | |
| 271 "CM region mark stack"); | |
| 272 _index = 0; | |
| 273 // QQQQ cast ... | |
| 274 _capacity = (jint) size; | |
| 275 } | |
| 276 | |
| 277 CMRegionStack::~CMRegionStack() { | |
| 278 if (_base != NULL) FREE_C_HEAP_ARRAY(oop, _base); | |
| 279 } | |
| 280 | |
| 281 void CMRegionStack::push(MemRegion mr) { | |
| 282 assert(mr.word_size() > 0, "Precondition"); | |
| 283 while (true) { | |
| 284 if (isFull()) { | |
| 285 _overflow = true; | |
| 286 return; | |
| 287 } | |
| 288 // Otherwise... | |
| 289 jint index = _index; | |
| 290 jint next_index = index+1; | |
| 291 jint res = Atomic::cmpxchg(next_index, &_index, index); | |
| 292 if (res == index) { | |
| 293 _base[index] = mr; | |
| 294 return; | |
| 295 } | |
| 296 // Otherwise, we need to try again. | |
| 297 } | |
| 298 } | |
| 299 | |
| 300 MemRegion CMRegionStack::pop() { | |
| 301 while (true) { | |
| 302 // Otherwise... | |
| 303 jint index = _index; | |
| 304 | |
| 305 if (index == 0) { | |
| 306 return MemRegion(); | |
| 307 } | |
| 308 jint next_index = index-1; | |
| 309 jint res = Atomic::cmpxchg(next_index, &_index, index); | |
| 310 if (res == index) { | |
| 311 MemRegion mr = _base[next_index]; | |
| 312 if (mr.start() != NULL) { | |
| 313 tmp_guarantee_CM( mr.end() != NULL, "invariant" ); | |
| 314 tmp_guarantee_CM( mr.word_size() > 0, "invariant" ); | |
| 315 return mr; | |
| 316 } else { | |
| 317 // that entry was invalidated... let's skip it | |
| 318 tmp_guarantee_CM( mr.end() == NULL, "invariant" ); | |
| 319 } | |
| 320 } | |
| 321 // Otherwise, we need to try again. | |
| 322 } | |
| 323 } | |
| 324 | |
| 325 bool CMRegionStack::invalidate_entries_into_cset() { | |
| 326 bool result = false; | |
| 327 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
| 328 for (int i = 0; i < _oops_do_bound; ++i) { | |
| 329 MemRegion mr = _base[i]; | |
| 330 if (mr.start() != NULL) { | |
| 331 tmp_guarantee_CM( mr.end() != NULL, "invariant"); | |
| 332 tmp_guarantee_CM( mr.word_size() > 0, "invariant" ); | |
| 333 HeapRegion* hr = g1h->heap_region_containing(mr.start()); | |
| 334 tmp_guarantee_CM( hr != NULL, "invariant" ); | |
| 335 if (hr->in_collection_set()) { | |
| 336 // The region points into the collection set | |
| 337 _base[i] = MemRegion(); | |
| 338 result = true; | |
| 339 } | |
| 340 } else { | |
| 341 // that entry was invalidated... let's skip it | |
| 342 tmp_guarantee_CM( mr.end() == NULL, "invariant" ); | |
| 343 } | |
| 344 } | |
| 345 return result; | |
| 346 } | |
| 347 | |
| 348 template<class OopClosureClass> | |
| 349 bool CMMarkStack::drain(OopClosureClass* cl, CMBitMap* bm, bool yield_after) { | |
| 350 assert(!_drain_in_progress || !_drain_in_progress_yields || yield_after | |
| 351 || SafepointSynchronize::is_at_safepoint(), | |
| 352 "Drain recursion must be yield-safe."); | |
| 353 bool res = true; | |
| 354 debug_only(_drain_in_progress = true); | |
| 355 debug_only(_drain_in_progress_yields = yield_after); | |
| 356 while (!isEmpty()) { | |
| 357 oop newOop = pop(); | |
| 358 assert(G1CollectedHeap::heap()->is_in_reserved(newOop), "Bad pop"); | |
| 359 assert(newOop->is_oop(), "Expected an oop"); | |
| 360 assert(bm == NULL || bm->isMarked((HeapWord*)newOop), | |
| 361 "only grey objects on this stack"); | |
| 362 // iterate over the oops in this oop, marking and pushing | |
| 363 // the ones in CMS generation. | |
| 364 newOop->oop_iterate(cl); | |
| 365 if (yield_after && _cm->do_yield_check()) { | |
| 366 res = false; break; | |
| 367 } | |
| 368 } | |
| 369 debug_only(_drain_in_progress = false); | |
| 370 return res; | |
| 371 } | |
| 372 | |
| 373 void CMMarkStack::oops_do(OopClosure* f) { | |
| 374 if (_index == 0) return; | |
| 375 assert(_oops_do_bound != -1 && _oops_do_bound <= _index, | |
| 376 "Bound must be set."); | |
| 377 for (int i = 0; i < _oops_do_bound; i++) { | |
| 378 f->do_oop(&_base[i]); | |
| 379 } | |
| 380 _oops_do_bound = -1; | |
| 381 } | |
| 382 | |
| 383 bool ConcurrentMark::not_yet_marked(oop obj) const { | |
| 384 return (_g1h->is_obj_ill(obj) | |
| 385 || (_g1h->is_in_permanent(obj) | |
| 386 && !nextMarkBitMap()->isMarked((HeapWord*)obj))); | |
| 387 } | |
| 388 | |
| 389 #ifdef _MSC_VER // the use of 'this' below gets a warning, make it go away | |
| 390 #pragma warning( disable:4355 ) // 'this' : used in base member initializer list | |
| 391 #endif // _MSC_VER | |
| 392 | |
| 393 ConcurrentMark::ConcurrentMark(ReservedSpace rs, | |
| 394 int max_regions) : | |
| 395 _markBitMap1(rs, MinObjAlignment - 1), | |
| 396 _markBitMap2(rs, MinObjAlignment - 1), | |
| 397 | |
| 398 _parallel_marking_threads(0), | |
| 399 _sleep_factor(0.0), | |
| 400 _marking_task_overhead(1.0), | |
| 401 _cleanup_sleep_factor(0.0), | |
| 402 _cleanup_task_overhead(1.0), | |
| 403 _region_bm(max_regions, false /* in_resource_area*/), | |
| 404 _card_bm((rs.size() + CardTableModRefBS::card_size - 1) >> | |
| 405 CardTableModRefBS::card_shift, | |
| 406 false /* in_resource_area*/), | |
| 407 _prevMarkBitMap(&_markBitMap1), | |
| 408 _nextMarkBitMap(&_markBitMap2), | |
| 409 _at_least_one_mark_complete(false), | |
| 410 | |
| 411 _markStack(this), | |
| 412 _regionStack(), | |
| 413 // _finger set in set_non_marking_state | |
| 414 | |
| 415 _max_task_num(MAX2(ParallelGCThreads, (size_t)1)), | |
| 416 // _active_tasks set in set_non_marking_state | |
| 417 // _tasks set inside the constructor | |
| 418 _task_queues(new CMTaskQueueSet((int) _max_task_num)), | |
| 419 _terminator(ParallelTaskTerminator((int) _max_task_num, _task_queues)), | |
| 420 | |
| 421 _has_overflown(false), | |
| 422 _concurrent(false), | |
|
619
7ea5ca260b28
6814467: G1: small fixes related to concurrent marking verboseness
tonyp
parents:
470
diff
changeset
|
423 _has_aborted(false), |
|
7ea5ca260b28
6814467: G1: small fixes related to concurrent marking verboseness
tonyp
parents:
470
diff
changeset
|
424 _restart_for_overflow(false), |
|
7ea5ca260b28
6814467: G1: small fixes related to concurrent marking verboseness
tonyp
parents:
470
diff
changeset
|
425 _concurrent_marking_in_progress(false), |
|
7ea5ca260b28
6814467: G1: small fixes related to concurrent marking verboseness
tonyp
parents:
470
diff
changeset
|
426 _should_gray_objects(false), |
| 342 | 427 |
| 428 // _verbose_level set below | |
| 429 | |
| 430 _init_times(), | |
| 431 _remark_times(), _remark_mark_times(), _remark_weak_ref_times(), | |
| 432 _cleanup_times(), | |
| 433 _total_counting_time(0.0), | |
| 434 _total_rs_scrub_time(0.0), | |
| 435 | |
| 436 _parallel_workers(NULL), | |
| 437 _cleanup_co_tracker(G1CLGroup) | |
| 438 { | |
| 439 CMVerboseLevel verbose_level = | |
| 440 (CMVerboseLevel) G1MarkingVerboseLevel; | |
| 441 if (verbose_level < no_verbose) | |
| 442 verbose_level = no_verbose; | |
| 443 if (verbose_level > high_verbose) | |
| 444 verbose_level = high_verbose; | |
| 445 _verbose_level = verbose_level; | |
| 446 | |
| 447 if (verbose_low()) | |
| 448 gclog_or_tty->print_cr("[global] init, heap start = "PTR_FORMAT", " | |
| 449 "heap end = "PTR_FORMAT, _heap_start, _heap_end); | |
| 450 | |
| 451 _markStack.allocate(G1CMStackSize); | |
| 452 _regionStack.allocate(G1CMRegionStackSize); | |
| 453 | |
| 454 // Create & start a ConcurrentMark thread. | |
| 455 if (G1ConcMark) { | |
| 456 _cmThread = new ConcurrentMarkThread(this); | |
| 457 assert(cmThread() != NULL, "CM Thread should have been created"); | |
| 458 assert(cmThread()->cm() != NULL, "CM Thread should refer to this cm"); | |
| 459 } else { | |
| 460 _cmThread = NULL; | |
| 461 } | |
| 462 _g1h = G1CollectedHeap::heap(); | |
| 463 assert(CGC_lock != NULL, "Where's the CGC_lock?"); | |
| 464 assert(_markBitMap1.covers(rs), "_markBitMap1 inconsistency"); | |
| 465 assert(_markBitMap2.covers(rs), "_markBitMap2 inconsistency"); | |
| 466 | |
| 467 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set(); | |
| 468 satb_qs.set_buffer_size(G1SATBLogBufferSize); | |
| 469 | |
| 470 int size = (int) MAX2(ParallelGCThreads, (size_t)1); | |
| 471 _par_cleanup_thread_state = NEW_C_HEAP_ARRAY(ParCleanupThreadState*, size); | |
| 472 for (int i = 0 ; i < size; i++) { | |
| 473 _par_cleanup_thread_state[i] = new ParCleanupThreadState; | |
| 474 } | |
| 475 | |
| 476 _tasks = NEW_C_HEAP_ARRAY(CMTask*, _max_task_num); | |
| 477 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_task_num); | |
| 478 | |
| 479 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail | |
| 480 _active_tasks = _max_task_num; | |
| 481 for (int i = 0; i < (int) _max_task_num; ++i) { | |
| 482 CMTaskQueue* task_queue = new CMTaskQueue(); | |
| 483 task_queue->initialize(); | |
| 484 _task_queues->register_queue(i, task_queue); | |
| 485 | |
| 486 _tasks[i] = new CMTask(i, this, task_queue, _task_queues); | |
| 487 _accum_task_vtime[i] = 0.0; | |
| 488 } | |
| 489 | |
| 490 if (ParallelMarkingThreads > ParallelGCThreads) { | |
| 491 vm_exit_during_initialization("Can't have more ParallelMarkingThreads " | |
| 492 "than ParallelGCThreads."); | |
| 493 } | |
| 494 if (ParallelGCThreads == 0) { | |
| 495 // if we are not running with any parallel GC threads we will not | |
| 496 // spawn any marking threads either | |
| 497 _parallel_marking_threads = 0; | |
| 498 _sleep_factor = 0.0; | |
| 499 _marking_task_overhead = 1.0; | |
| 500 } else { | |
| 501 if (ParallelMarkingThreads > 0) { | |
| 502 // notice that ParallelMarkingThreads overwrites G1MarkingOverheadPerc | |
| 503 // if both are set | |
| 504 | |
| 505 _parallel_marking_threads = ParallelMarkingThreads; | |
| 506 _sleep_factor = 0.0; | |
| 507 _marking_task_overhead = 1.0; | |
| 508 } else if (G1MarkingOverheadPerc > 0) { | |
| 509 // we will calculate the number of parallel marking threads | |
| 510 // based on a target overhead with respect to the soft real-time | |
| 511 // goal | |
| 512 | |
| 513 double marking_overhead = (double) G1MarkingOverheadPerc / 100.0; | |
| 514 double overall_cm_overhead = | |
| 515 (double) G1MaxPauseTimeMS * marking_overhead / (double) G1TimeSliceMS; | |
| 516 double cpu_ratio = 1.0 / (double) os::processor_count(); | |
| 517 double marking_thread_num = ceil(overall_cm_overhead / cpu_ratio); | |
| 518 double marking_task_overhead = | |
| 519 overall_cm_overhead / marking_thread_num * | |
| 520 (double) os::processor_count(); | |
| 521 double sleep_factor = | |
| 522 (1.0 - marking_task_overhead) / marking_task_overhead; | |
| 523 | |
| 524 _parallel_marking_threads = (size_t) marking_thread_num; | |
| 525 _sleep_factor = sleep_factor; | |
| 526 _marking_task_overhead = marking_task_overhead; | |
| 527 } else { | |
| 528 _parallel_marking_threads = MAX2((ParallelGCThreads + 2) / 4, (size_t)1); | |
| 529 _sleep_factor = 0.0; | |
| 530 _marking_task_overhead = 1.0; | |
| 531 } | |
| 532 | |
| 533 if (parallel_marking_threads() > 1) | |
| 534 _cleanup_task_overhead = 1.0; | |
| 535 else | |
| 536 _cleanup_task_overhead = marking_task_overhead(); | |
| 537 _cleanup_sleep_factor = | |
| 538 (1.0 - cleanup_task_overhead()) / cleanup_task_overhead(); | |
| 539 | |
| 540 #if 0 | |
| 541 gclog_or_tty->print_cr("Marking Threads %d", parallel_marking_threads()); | |
| 542 gclog_or_tty->print_cr("CM Marking Task Overhead %1.4lf", marking_task_overhead()); | |
| 543 gclog_or_tty->print_cr("CM Sleep Factor %1.4lf", sleep_factor()); | |
| 544 gclog_or_tty->print_cr("CL Marking Task Overhead %1.4lf", cleanup_task_overhead()); | |
| 545 gclog_or_tty->print_cr("CL Sleep Factor %1.4lf", cleanup_sleep_factor()); | |
| 546 #endif | |
| 547 | |
| 548 guarantee( parallel_marking_threads() > 0, "peace of mind" ); | |
| 549 _parallel_workers = new WorkGang("Parallel Marking Threads", | |
| 550 (int) parallel_marking_threads(), false, true); | |
| 551 if (_parallel_workers == NULL) | |
| 552 vm_exit_during_initialization("Failed necessary allocation."); | |
| 553 } | |
| 554 | |
| 555 // so that the call below can read a sensible value | |
| 556 _heap_start = (HeapWord*) rs.base(); | |
| 557 set_non_marking_state(); | |
| 558 } | |
| 559 | |
| 560 void ConcurrentMark::update_g1_committed(bool force) { | |
| 561 // If concurrent marking is not in progress, then we do not need to | |
| 562 // update _heap_end. This has a subtle and important | |
| 563 // side-effect. Imagine that two evacuation pauses happen between | |
| 564 // marking completion and remark. The first one can grow the | |
| 565 // heap (hence now the finger is below the heap end). Then, the | |
| 566 // second one could unnecessarily push regions on the region | |
| 567 // stack. This causes the invariant that the region stack is empty | |
| 568 // at the beginning of remark to be false. By ensuring that we do | |
| 569 // not observe heap expansions after marking is complete, then we do | |
| 570 // not have this problem. | |
| 571 if (!concurrent_marking_in_progress() && !force) | |
| 572 return; | |
| 573 | |
| 574 MemRegion committed = _g1h->g1_committed(); | |
| 575 tmp_guarantee_CM( committed.start() == _heap_start, | |
| 576 "start shouldn't change" ); | |
| 577 HeapWord* new_end = committed.end(); | |
| 578 if (new_end > _heap_end) { | |
| 579 // The heap has been expanded. | |
| 580 | |
| 581 _heap_end = new_end; | |
| 582 } | |
| 583 // Notice that the heap can also shrink. However, this only happens | |
| 584 // during a Full GC (at least currently) and the entire marking | |
| 585 // phase will bail out and the task will not be restarted. So, let's | |
| 586 // do nothing. | |
| 587 } | |
| 588 | |
| 589 void ConcurrentMark::reset() { | |
| 590 // Starting values for these two. This should be called in a STW | |
| 591 // phase. CM will be notified of any future g1_committed expansions | |
| 592 // will be at the end of evacuation pauses, when tasks are | |
| 593 // inactive. | |
| 594 MemRegion committed = _g1h->g1_committed(); | |
| 595 _heap_start = committed.start(); | |
| 596 _heap_end = committed.end(); | |
| 597 | |
| 598 guarantee( _heap_start != NULL && | |
| 599 _heap_end != NULL && | |
| 600 _heap_start < _heap_end, "heap bounds should look ok" ); | |
| 601 | |
| 602 // reset all the marking data structures and any necessary flags | |
| 603 clear_marking_state(); | |
| 604 | |
| 605 if (verbose_low()) | |
| 606 gclog_or_tty->print_cr("[global] resetting"); | |
| 607 | |
| 608 // We do reset all of them, since different phases will use | |
| 609 // different number of active threads. So, it's easiest to have all | |
| 610 // of them ready. | |
| 611 for (int i = 0; i < (int) _max_task_num; ++i) | |
| 612 _tasks[i]->reset(_nextMarkBitMap); | |
| 613 | |
| 614 // we need this to make sure that the flag is on during the evac | |
| 615 // pause with initial mark piggy-backed | |
| 616 set_concurrent_marking_in_progress(); | |
| 617 } | |
| 618 | |
| 619 void ConcurrentMark::set_phase(size_t active_tasks, bool concurrent) { | |
| 620 guarantee( active_tasks <= _max_task_num, "we should not have more" ); | |
| 621 | |
| 622 _active_tasks = active_tasks; | |
| 623 // Need to update the three data structures below according to the | |
| 624 // number of active threads for this phase. | |
| 625 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues); | |
| 626 _first_overflow_barrier_sync.set_n_workers((int) active_tasks); | |
| 627 _second_overflow_barrier_sync.set_n_workers((int) active_tasks); | |
| 628 | |
| 629 _concurrent = concurrent; | |
| 630 // We propagate this to all tasks, not just the active ones. | |
| 631 for (int i = 0; i < (int) _max_task_num; ++i) | |
| 632 _tasks[i]->set_concurrent(concurrent); | |
| 633 | |
| 634 if (concurrent) { | |
| 635 set_concurrent_marking_in_progress(); | |
| 636 } else { | |
| 637 // We currently assume that the concurrent flag has been set to | |
| 638 // false before we start remark. At this point we should also be | |
| 639 // in a STW phase. | |
| 640 guarantee( !concurrent_marking_in_progress(), "invariant" ); | |
| 641 guarantee( _finger == _heap_end, "only way to get here" ); | |
| 642 update_g1_committed(true); | |
| 643 } | |
| 644 } | |
| 645 | |
| 646 void ConcurrentMark::set_non_marking_state() { | |
| 647 // We set the global marking state to some default values when we're | |
| 648 // not doing marking. | |
| 649 clear_marking_state(); | |
| 650 _active_tasks = 0; | |
| 651 clear_concurrent_marking_in_progress(); | |
| 652 } | |
| 653 | |
| 654 ConcurrentMark::~ConcurrentMark() { | |
| 655 int size = (int) MAX2(ParallelGCThreads, (size_t)1); | |
| 656 for (int i = 0; i < size; i++) delete _par_cleanup_thread_state[i]; | |
| 657 FREE_C_HEAP_ARRAY(ParCleanupThreadState*, | |
| 658 _par_cleanup_thread_state); | |
| 659 | |
| 660 for (int i = 0; i < (int) _max_task_num; ++i) { | |
| 661 delete _task_queues->queue(i); | |
| 662 delete _tasks[i]; | |
| 663 } | |
| 664 delete _task_queues; | |
| 665 FREE_C_HEAP_ARRAY(CMTask*, _max_task_num); | |
| 666 } | |
| 667 | |
| 668 // This closure is used to mark refs into the g1 generation | |
| 669 // from external roots in the CMS bit map. | |
| 670 // Called at the first checkpoint. | |
| 671 // | |
| 672 | |
| 673 #define PRINT_REACHABLE_AT_INITIAL_MARK 0 | |
| 674 #if PRINT_REACHABLE_AT_INITIAL_MARK | |
| 675 static FILE* reachable_file = NULL; | |
| 676 | |
| 677 class PrintReachableClosure: public OopsInGenClosure { | |
| 678 CMBitMap* _bm; | |
| 679 int _level; | |
| 680 public: | |
| 681 PrintReachableClosure(CMBitMap* bm) : | |
| 682 _bm(bm), _level(0) { | |
| 683 guarantee(reachable_file != NULL, "pre-condition"); | |
| 684 } | |
| 685 void do_oop(oop* p) { | |
| 686 oop obj = *p; | |
| 687 HeapWord* obj_addr = (HeapWord*)obj; | |
| 688 if (obj == NULL) return; | |
| 689 fprintf(reachable_file, "%d: "PTR_FORMAT" -> "PTR_FORMAT" (%d)\n", | |
| 690 _level, p, (void*) obj, _bm->isMarked(obj_addr)); | |
| 691 if (!_bm->isMarked(obj_addr)) { | |
| 692 _bm->mark(obj_addr); | |
| 693 _level++; | |
| 694 obj->oop_iterate(this); | |
| 695 _level--; | |
| 696 } | |
| 697 } | |
| 698 }; | |
| 699 #endif // PRINT_REACHABLE_AT_INITIAL_MARK | |
| 700 | |
| 701 #define SEND_HEAP_DUMP_TO_FILE 0 | |
| 702 #if SEND_HEAP_DUMP_TO_FILE | |
| 703 static FILE* heap_dump_file = NULL; | |
| 704 #endif // SEND_HEAP_DUMP_TO_FILE | |
| 705 | |
| 706 void ConcurrentMark::clearNextBitmap() { | |
| 707 guarantee(!G1CollectedHeap::heap()->mark_in_progress(), "Precondition."); | |
| 708 | |
| 709 // clear the mark bitmap (no grey objects to start with). | |
| 710 // We need to do this in chunks and offer to yield in between | |
| 711 // each chunk. | |
| 712 HeapWord* start = _nextMarkBitMap->startWord(); | |
| 713 HeapWord* end = _nextMarkBitMap->endWord(); | |
| 714 HeapWord* cur = start; | |
| 715 size_t chunkSize = M; | |
| 716 while (cur < end) { | |
| 717 HeapWord* next = cur + chunkSize; | |
| 718 if (next > end) | |
| 719 next = end; | |
| 720 MemRegion mr(cur,next); | |
| 721 _nextMarkBitMap->clearRange(mr); | |
| 722 cur = next; | |
| 723 do_yield_check(); | |
| 724 } | |
| 725 } | |
| 726 | |
| 727 class NoteStartOfMarkHRClosure: public HeapRegionClosure { | |
| 728 public: | |
| 729 bool doHeapRegion(HeapRegion* r) { | |
| 730 if (!r->continuesHumongous()) { | |
| 731 r->note_start_of_marking(true); | |
| 732 } | |
| 733 return false; | |
| 734 } | |
| 735 }; | |
| 736 | |
| 737 void ConcurrentMark::checkpointRootsInitialPre() { | |
| 738 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
| 739 G1CollectorPolicy* g1p = g1h->g1_policy(); | |
| 740 | |
| 741 _has_aborted = false; | |
| 742 | |
| 743 // Find all the reachable objects... | |
| 744 #if PRINT_REACHABLE_AT_INITIAL_MARK | |
| 745 guarantee(reachable_file == NULL, "Protocol"); | |
| 746 char fn_buf[100]; | |
| 747 sprintf(fn_buf, "/tmp/reachable.txt.%d", os::current_process_id()); | |
| 748 reachable_file = fopen(fn_buf, "w"); | |
| 749 // clear the mark bitmap (no grey objects to start with) | |
| 750 _nextMarkBitMap->clearAll(); | |
| 751 PrintReachableClosure prcl(_nextMarkBitMap); | |
| 752 g1h->process_strong_roots( | |
| 753 false, // fake perm gen collection | |
| 754 SharedHeap::SO_AllClasses, | |
| 755 &prcl, // Regular roots | |
| 756 &prcl // Perm Gen Roots | |
| 757 ); | |
| 758 // The root iteration above "consumed" dirty cards in the perm gen. | |
| 759 // Therefore, as a shortcut, we dirty all such cards. | |
| 760 g1h->rem_set()->invalidate(g1h->perm_gen()->used_region(), false); | |
| 761 fclose(reachable_file); | |
| 762 reachable_file = NULL; | |
| 763 // clear the mark bitmap again. | |
| 764 _nextMarkBitMap->clearAll(); | |
| 765 COMPILER2_PRESENT(DerivedPointerTable::update_pointers()); | |
| 766 COMPILER2_PRESENT(DerivedPointerTable::clear()); | |
| 767 #endif // PRINT_REACHABLE_AT_INITIAL_MARK | |
| 768 | |
| 769 // Initialise marking structures. This has to be done in a STW phase. | |
| 770 reset(); | |
| 771 } | |
| 772 | |
| 773 class CMMarkRootsClosure: public OopsInGenClosure { | |
| 774 private: | |
| 775 ConcurrentMark* _cm; | |
| 776 G1CollectedHeap* _g1h; | |
| 777 bool _do_barrier; | |
| 778 | |
| 779 public: | |
| 780 CMMarkRootsClosure(ConcurrentMark* cm, | |
| 781 G1CollectedHeap* g1h, | |
| 782 bool do_barrier) : _cm(cm), _g1h(g1h), | |
| 783 _do_barrier(do_barrier) { } | |
| 784 | |
| 785 virtual void do_oop(narrowOop* p) { | |
| 786 guarantee(false, "NYI"); | |
| 787 } | |
| 788 | |
| 789 virtual void do_oop(oop* p) { | |
| 790 oop thisOop = *p; | |
| 791 if (thisOop != NULL) { | |
| 792 assert(thisOop->is_oop() || thisOop->mark() == NULL, | |
| 793 "expected an oop, possibly with mark word displaced"); | |
| 794 HeapWord* addr = (HeapWord*)thisOop; | |
| 795 if (_g1h->is_in_g1_reserved(addr)) { | |
| 796 _cm->grayRoot(thisOop); | |
| 797 } | |
| 798 } | |
| 799 if (_do_barrier) { | |
| 800 assert(!_g1h->is_in_g1_reserved(p), | |
| 801 "Should be called on external roots"); | |
| 802 do_barrier(p); | |
| 803 } | |
| 804 } | |
| 805 }; | |
| 806 | |
| 807 void ConcurrentMark::checkpointRootsInitialPost() { | |
| 808 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
| 809 | |
| 810 // For each region note start of marking. | |
| 811 NoteStartOfMarkHRClosure startcl; | |
| 812 g1h->heap_region_iterate(&startcl); | |
| 813 | |
| 814 // Start weak-reference discovery. | |
| 815 ReferenceProcessor* rp = g1h->ref_processor(); | |
| 816 rp->verify_no_references_recorded(); | |
| 817 rp->enable_discovery(); // enable ("weak") refs discovery | |
|
457
27a80744a83b
6778647: snap(), snap_policy() should be renamed setup(), setup_policy()
ysr
parents:
453
diff
changeset
|
818 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle |
| 342 | 819 |
| 820 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
| 821 satb_mq_set.set_process_completed_threshold(G1SATBProcessCompletedThreshold); | |
| 822 satb_mq_set.set_active_all_threads(true); | |
| 823 | |
| 824 // update_g1_committed() will be called at the end of an evac pause | |
| 825 // when marking is on. So, it's also called at the end of the | |
| 826 // initial-mark pause to update the heap end, if the heap expands | |
| 827 // during it. No need to call it here. | |
| 828 | |
| 829 guarantee( !_cleanup_co_tracker.enabled(), "invariant" ); | |
| 830 | |
| 831 size_t max_marking_threads = | |
| 832 MAX2((size_t) 1, parallel_marking_threads()); | |
| 833 for (int i = 0; i < (int)_max_task_num; ++i) { | |
| 834 _tasks[i]->enable_co_tracker(); | |
| 835 if (i < (int) max_marking_threads) | |
| 836 _tasks[i]->reset_co_tracker(marking_task_overhead()); | |
| 837 else | |
| 838 _tasks[i]->reset_co_tracker(0.0); | |
| 839 } | |
| 840 } | |
| 841 | |
| 842 // Checkpoint the roots into this generation from outside | |
| 843 // this generation. [Note this initial checkpoint need only | |
| 844 // be approximate -- we'll do a catch up phase subsequently.] | |
| 845 void ConcurrentMark::checkpointRootsInitial() { | |
| 846 assert(SafepointSynchronize::is_at_safepoint(), "world should be stopped"); | |
| 847 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
| 848 | |
| 849 double start = os::elapsedTime(); | |
| 850 GCOverheadReporter::recordSTWStart(start); | |
| 851 | |
| 852 // If there has not been a GC[n-1] since last GC[n] cycle completed, | |
| 853 // precede our marking with a collection of all | |
| 854 // younger generations to keep floating garbage to a minimum. | |
| 855 // YSR: we won't do this for now -- it's an optimization to be | |
| 856 // done post-beta. | |
| 857 | |
| 858 // YSR: ignoring weak refs for now; will do at bug fixing stage | |
| 859 // EVM: assert(discoveredRefsAreClear()); | |
| 860 | |
| 861 | |
| 862 G1CollectorPolicy* g1p = G1CollectedHeap::heap()->g1_policy(); | |
| 863 g1p->record_concurrent_mark_init_start(); | |
| 864 checkpointRootsInitialPre(); | |
| 865 | |
| 866 // YSR: when concurrent precleaning is in place, we'll | |
| 867 // need to clear the cached card table here | |
| 868 | |
| 869 ResourceMark rm; | |
| 870 HandleMark hm; | |
| 871 | |
| 872 g1h->ensure_parsability(false); | |
| 873 g1h->perm_gen()->save_marks(); | |
| 874 | |
| 875 CMMarkRootsClosure notOlder(this, g1h, false); | |
| 876 CMMarkRootsClosure older(this, g1h, true); | |
| 877 | |
| 878 g1h->set_marking_started(); | |
| 879 g1h->rem_set()->prepare_for_younger_refs_iterate(false); | |
| 880 | |
| 881 g1h->process_strong_roots(false, // fake perm gen collection | |
| 882 SharedHeap::SO_AllClasses, | |
| 883 ¬Older, // Regular roots | |
| 884 &older // Perm Gen Roots | |
| 885 ); | |
| 886 checkpointRootsInitialPost(); | |
| 887 | |
| 888 // Statistics. | |
| 889 double end = os::elapsedTime(); | |
| 890 _init_times.add((end - start) * 1000.0); | |
| 891 GCOverheadReporter::recordSTWEnd(end); | |
| 892 | |
| 893 g1p->record_concurrent_mark_init_end(); | |
| 894 } | |
| 895 | |
| 896 /* | |
| 897 Notice that in the next two methods, we actually leave the STS | |
| 898 during the barrier sync and join it immediately afterwards. If we | |
| 899 do not do this, this then the following deadlock can occur: one | |
| 900 thread could be in the barrier sync code, waiting for the other | |
| 901 thread to also sync up, whereas another one could be trying to | |
| 902 yield, while also waiting for the other threads to sync up too. | |
| 903 | |
| 904 Because the thread that does the sync barrier has left the STS, it | |
| 905 is possible to be suspended for a Full GC or an evacuation pause | |
| 906 could occur. This is actually safe, since the entering the sync | |
| 907 barrier is one of the last things do_marking_step() does, and it | |
| 908 doesn't manipulate any data structures afterwards. | |
| 909 */ | |
| 910 | |
| 911 void ConcurrentMark::enter_first_sync_barrier(int task_num) { | |
| 912 if (verbose_low()) | |
| 913 gclog_or_tty->print_cr("[%d] entering first barrier", task_num); | |
| 914 | |
| 915 ConcurrentGCThread::stsLeave(); | |
| 916 _first_overflow_barrier_sync.enter(); | |
| 917 ConcurrentGCThread::stsJoin(); | |
| 918 // at this point everyone should have synced up and not be doing any | |
| 919 // more work | |
| 920 | |
| 921 if (verbose_low()) | |
| 922 gclog_or_tty->print_cr("[%d] leaving first barrier", task_num); | |
| 923 | |
| 924 // let task 0 do this | |
| 925 if (task_num == 0) { | |
| 926 // task 0 is responsible for clearing the global data structures | |
| 927 clear_marking_state(); | |
| 928 | |
| 929 if (PrintGC) { | |
| 930 gclog_or_tty->date_stamp(PrintGCDateStamps); | |
| 931 gclog_or_tty->stamp(PrintGCTimeStamps); | |
| 932 gclog_or_tty->print_cr("[GC concurrent-mark-reset-for-overflow]"); | |
| 933 } | |
| 934 } | |
| 935 | |
| 936 // after this, each task should reset its own data structures then | |
| 937 // then go into the second barrier | |
| 938 } | |
| 939 | |
| 940 void ConcurrentMark::enter_second_sync_barrier(int task_num) { | |
| 941 if (verbose_low()) | |
| 942 gclog_or_tty->print_cr("[%d] entering second barrier", task_num); | |
| 943 | |
| 944 ConcurrentGCThread::stsLeave(); | |
| 945 _second_overflow_barrier_sync.enter(); | |
| 946 ConcurrentGCThread::stsJoin(); | |
| 947 // at this point everything should be re-initialised and ready to go | |
| 948 | |
| 949 if (verbose_low()) | |
| 950 gclog_or_tty->print_cr("[%d] leaving second barrier", task_num); | |
| 951 } | |
| 952 | |
| 953 void ConcurrentMark::grayRoot(oop p) { | |
| 954 HeapWord* addr = (HeapWord*) p; | |
| 955 // We can't really check against _heap_start and _heap_end, since it | |
| 956 // is possible during an evacuation pause with piggy-backed | |
| 957 // initial-mark that the committed space is expanded during the | |
| 958 // pause without CM observing this change. So the assertions below | |
| 959 // is a bit conservative; but better than nothing. | |
| 960 tmp_guarantee_CM( _g1h->g1_committed().contains(addr), | |
| 961 "address should be within the heap bounds" ); | |
| 962 | |
| 963 if (!_nextMarkBitMap->isMarked(addr)) | |
| 964 _nextMarkBitMap->parMark(addr); | |
| 965 } | |
| 966 | |
| 967 void ConcurrentMark::grayRegionIfNecessary(MemRegion mr) { | |
| 968 // The objects on the region have already been marked "in bulk" by | |
| 969 // the caller. We only need to decide whether to push the region on | |
| 970 // the region stack or not. | |
| 971 | |
| 972 if (!concurrent_marking_in_progress() || !_should_gray_objects) | |
| 973 // We're done with marking and waiting for remark. We do not need to | |
| 974 // push anything else on the region stack. | |
| 975 return; | |
| 976 | |
| 977 HeapWord* finger = _finger; | |
| 978 | |
| 979 if (verbose_low()) | |
| 980 gclog_or_tty->print_cr("[global] attempting to push " | |
| 981 "region ["PTR_FORMAT", "PTR_FORMAT"), finger is at " | |
| 982 PTR_FORMAT, mr.start(), mr.end(), finger); | |
| 983 | |
| 984 if (mr.start() < finger) { | |
| 985 // The finger is always heap region aligned and it is not possible | |
| 986 // for mr to span heap regions. | |
| 987 tmp_guarantee_CM( mr.end() <= finger, "invariant" ); | |
| 988 | |
| 989 tmp_guarantee_CM( mr.start() <= mr.end() && | |
| 990 _heap_start <= mr.start() && | |
| 991 mr.end() <= _heap_end, | |
| 992 "region boundaries should fall within the committed space" ); | |
| 993 if (verbose_low()) | |
| 994 gclog_or_tty->print_cr("[global] region ["PTR_FORMAT", "PTR_FORMAT") " | |
| 995 "below the finger, pushing it", | |
| 996 mr.start(), mr.end()); | |
| 997 | |
| 998 if (!region_stack_push(mr)) { | |
| 999 if (verbose_low()) | |
| 1000 gclog_or_tty->print_cr("[global] region stack has overflown."); | |
| 1001 } | |
| 1002 } | |
| 1003 } | |
| 1004 | |
| 1005 void ConcurrentMark::markAndGrayObjectIfNecessary(oop p) { | |
| 1006 // The object is not marked by the caller. We need to at least mark | |
| 1007 // it and maybe push in on the stack. | |
| 1008 | |
| 1009 HeapWord* addr = (HeapWord*)p; | |
| 1010 if (!_nextMarkBitMap->isMarked(addr)) { | |
| 1011 // We definitely need to mark it, irrespective whether we bail out | |
| 1012 // because we're done with marking. | |
| 1013 if (_nextMarkBitMap->parMark(addr)) { | |
| 1014 if (!concurrent_marking_in_progress() || !_should_gray_objects) | |
| 1015 // If we're done with concurrent marking and we're waiting for | |
| 1016 // remark, then we're not pushing anything on the stack. | |
| 1017 return; | |
| 1018 | |
| 1019 // No OrderAccess:store_load() is needed. It is implicit in the | |
| 1020 // CAS done in parMark(addr) above | |
| 1021 HeapWord* finger = _finger; | |
| 1022 | |
| 1023 if (addr < finger) { | |
| 1024 if (!mark_stack_push(oop(addr))) { | |
| 1025 if (verbose_low()) | |
| 1026 gclog_or_tty->print_cr("[global] global stack overflow " | |
| 1027 "during parMark"); | |
| 1028 } | |
| 1029 } | |
| 1030 } | |
| 1031 } | |
| 1032 } | |
| 1033 | |
| 1034 class CMConcurrentMarkingTask: public AbstractGangTask { | |
| 1035 private: | |
| 1036 ConcurrentMark* _cm; | |
| 1037 ConcurrentMarkThread* _cmt; | |
| 1038 | |
| 1039 public: | |
| 1040 void work(int worker_i) { | |
| 1041 guarantee( Thread::current()->is_ConcurrentGC_thread(), | |
| 1042 "this should only be done by a conc GC thread" ); | |
| 1043 | |
| 1044 double start_vtime = os::elapsedVTime(); | |
| 1045 | |
| 1046 ConcurrentGCThread::stsJoin(); | |
| 1047 | |
| 1048 guarantee( (size_t)worker_i < _cm->active_tasks(), "invariant" ); | |
| 1049 CMTask* the_task = _cm->task(worker_i); | |
| 1050 the_task->start_co_tracker(); | |
| 1051 the_task->record_start_time(); | |
| 1052 if (!_cm->has_aborted()) { | |
| 1053 do { | |
| 1054 double start_vtime_sec = os::elapsedVTime(); | |
| 1055 double start_time_sec = os::elapsedTime(); | |
| 1056 the_task->do_marking_step(10.0); | |
| 1057 double end_time_sec = os::elapsedTime(); | |
| 1058 double end_vtime_sec = os::elapsedVTime(); | |
| 1059 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec; | |
| 1060 double elapsed_time_sec = end_time_sec - start_time_sec; | |
| 1061 _cm->clear_has_overflown(); | |
| 1062 | |
| 1063 bool ret = _cm->do_yield_check(worker_i); | |
| 1064 | |
| 1065 jlong sleep_time_ms; | |
| 1066 if (!_cm->has_aborted() && the_task->has_aborted()) { | |
| 1067 sleep_time_ms = | |
| 1068 (jlong) (elapsed_vtime_sec * _cm->sleep_factor() * 1000.0); | |
| 1069 ConcurrentGCThread::stsLeave(); | |
| 1070 os::sleep(Thread::current(), sleep_time_ms, false); | |
| 1071 ConcurrentGCThread::stsJoin(); | |
| 1072 } | |
| 1073 double end_time2_sec = os::elapsedTime(); | |
| 1074 double elapsed_time2_sec = end_time2_sec - start_time_sec; | |
| 1075 | |
| 1076 the_task->update_co_tracker(); | |
| 1077 | |
| 1078 #if 0 | |
| 1079 gclog_or_tty->print_cr("CM: elapsed %1.4lf ms, sleep %1.4lf ms, " | |
| 1080 "overhead %1.4lf", | |
| 1081 elapsed_vtime_sec * 1000.0, (double) sleep_time_ms, | |
| 1082 the_task->conc_overhead(os::elapsedTime()) * 8.0); | |
| 1083 gclog_or_tty->print_cr("elapsed time %1.4lf ms, time 2: %1.4lf ms", | |
| 1084 elapsed_time_sec * 1000.0, elapsed_time2_sec * 1000.0); | |
| 1085 #endif | |
| 1086 } while (!_cm->has_aborted() && the_task->has_aborted()); | |
| 1087 } | |
| 1088 the_task->record_end_time(); | |
| 1089 guarantee( !the_task->has_aborted() || _cm->has_aborted(), "invariant" ); | |
| 1090 | |
| 1091 ConcurrentGCThread::stsLeave(); | |
| 1092 | |
| 1093 double end_vtime = os::elapsedVTime(); | |
| 1094 the_task->update_co_tracker(true); | |
| 1095 _cm->update_accum_task_vtime(worker_i, end_vtime - start_vtime); | |
| 1096 } | |
| 1097 | |
| 1098 CMConcurrentMarkingTask(ConcurrentMark* cm, | |
| 1099 ConcurrentMarkThread* cmt) : | |
| 1100 AbstractGangTask("Concurrent Mark"), _cm(cm), _cmt(cmt) { } | |
| 1101 | |
| 1102 ~CMConcurrentMarkingTask() { } | |
| 1103 }; | |
| 1104 | |
| 1105 void ConcurrentMark::markFromRoots() { | |
| 1106 // we might be tempted to assert that: | |
| 1107 // assert(asynch == !SafepointSynchronize::is_at_safepoint(), | |
| 1108 // "inconsistent argument?"); | |
| 1109 // However that wouldn't be right, because it's possible that | |
| 1110 // a safepoint is indeed in progress as a younger generation | |
| 1111 // stop-the-world GC happens even as we mark in this generation. | |
| 1112 | |
| 1113 _restart_for_overflow = false; | |
| 1114 | |
| 1115 set_phase(MAX2((size_t) 1, parallel_marking_threads()), true); | |
| 1116 | |
| 1117 CMConcurrentMarkingTask markingTask(this, cmThread()); | |
| 1118 if (parallel_marking_threads() > 0) | |
| 1119 _parallel_workers->run_task(&markingTask); | |
| 1120 else | |
| 1121 markingTask.work(0); | |
| 1122 print_stats(); | |
| 1123 } | |
| 1124 | |
| 1125 void ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) { | |
| 1126 // world is stopped at this checkpoint | |
| 1127 assert(SafepointSynchronize::is_at_safepoint(), | |
| 1128 "world should be stopped"); | |
| 1129 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
| 1130 | |
| 1131 // If a full collection has happened, we shouldn't do this. | |
| 1132 if (has_aborted()) { | |
| 1133 g1h->set_marking_complete(); // So bitmap clearing isn't confused | |
| 1134 return; | |
| 1135 } | |
| 1136 | |
| 1137 G1CollectorPolicy* g1p = g1h->g1_policy(); | |
| 1138 g1p->record_concurrent_mark_remark_start(); | |
| 1139 | |
| 1140 double start = os::elapsedTime(); | |
| 1141 GCOverheadReporter::recordSTWStart(start); | |
| 1142 | |
| 1143 checkpointRootsFinalWork(); | |
| 1144 | |
| 1145 double mark_work_end = os::elapsedTime(); | |
| 1146 | |
| 1147 weakRefsWork(clear_all_soft_refs); | |
| 1148 | |
| 1149 if (has_overflown()) { | |
| 1150 // Oops. We overflowed. Restart concurrent marking. | |
| 1151 _restart_for_overflow = true; | |
| 1152 // Clear the flag. We do not need it any more. | |
| 1153 clear_has_overflown(); | |
| 1154 if (G1TraceMarkStackOverflow) | |
| 1155 gclog_or_tty->print_cr("\nRemark led to restart for overflow."); | |
| 1156 } else { | |
| 1157 // We're done with marking. | |
| 1158 JavaThread::satb_mark_queue_set().set_active_all_threads(false); | |
| 1159 } | |
| 1160 | |
| 1161 #if VERIFY_OBJS_PROCESSED | |
| 1162 _scan_obj_cl.objs_processed = 0; | |
| 1163 ThreadLocalObjQueue::objs_enqueued = 0; | |
| 1164 #endif | |
| 1165 | |
| 1166 // Statistics | |
| 1167 double now = os::elapsedTime(); | |
| 1168 _remark_mark_times.add((mark_work_end - start) * 1000.0); | |
| 1169 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0); | |
| 1170 _remark_times.add((now - start) * 1000.0); | |
| 1171 | |
| 1172 GCOverheadReporter::recordSTWEnd(now); | |
| 1173 for (int i = 0; i < (int)_max_task_num; ++i) | |
| 1174 _tasks[i]->disable_co_tracker(); | |
| 1175 _cleanup_co_tracker.enable(); | |
| 1176 _cleanup_co_tracker.reset(cleanup_task_overhead()); | |
| 1177 g1p->record_concurrent_mark_remark_end(); | |
| 1178 } | |
| 1179 | |
| 1180 | |
| 1181 #define CARD_BM_TEST_MODE 0 | |
| 1182 | |
| 1183 class CalcLiveObjectsClosure: public HeapRegionClosure { | |
| 1184 | |
| 1185 CMBitMapRO* _bm; | |
| 1186 ConcurrentMark* _cm; | |
| 1187 COTracker* _co_tracker; | |
| 1188 bool _changed; | |
| 1189 bool _yield; | |
| 1190 size_t _words_done; | |
| 1191 size_t _tot_live; | |
| 1192 size_t _tot_used; | |
| 1193 size_t _regions_done; | |
| 1194 double _start_vtime_sec; | |
| 1195 | |
| 1196 BitMap* _region_bm; | |
| 1197 BitMap* _card_bm; | |
| 1198 intptr_t _bottom_card_num; | |
| 1199 bool _final; | |
| 1200 | |
| 1201 void mark_card_num_range(intptr_t start_card_num, intptr_t last_card_num) { | |
| 1202 for (intptr_t i = start_card_num; i <= last_card_num; i++) { | |
| 1203 #if CARD_BM_TEST_MODE | |
| 1204 guarantee(_card_bm->at(i - _bottom_card_num), | |
| 1205 "Should already be set."); | |
| 1206 #else | |
| 1207 _card_bm->par_at_put(i - _bottom_card_num, 1); | |
| 1208 #endif | |
| 1209 } | |
| 1210 } | |
| 1211 | |
| 1212 public: | |
| 1213 CalcLiveObjectsClosure(bool final, | |
| 1214 CMBitMapRO *bm, ConcurrentMark *cm, | |
| 1215 BitMap* region_bm, BitMap* card_bm, | |
| 1216 COTracker* co_tracker) : | |
| 1217 _bm(bm), _cm(cm), _changed(false), _yield(true), | |
| 1218 _words_done(0), _tot_live(0), _tot_used(0), | |
| 1219 _region_bm(region_bm), _card_bm(card_bm), | |
| 1220 _final(final), _co_tracker(co_tracker), | |
| 1221 _regions_done(0), _start_vtime_sec(0.0) | |
| 1222 { | |
| 1223 _bottom_card_num = | |
| 1224 intptr_t(uintptr_t(G1CollectedHeap::heap()->reserved_region().start()) >> | |
| 1225 CardTableModRefBS::card_shift); | |
| 1226 } | |
| 1227 | |
| 1228 bool doHeapRegion(HeapRegion* hr) { | |
| 1229 if (_co_tracker != NULL) | |
| 1230 _co_tracker->update(); | |
| 1231 | |
| 1232 if (!_final && _regions_done == 0) | |
| 1233 _start_vtime_sec = os::elapsedVTime(); | |
| 1234 | |
| 639 | 1235 if (hr->continuesHumongous()) { |
| 1236 HeapRegion* hum_start = hr->humongous_start_region(); | |
| 1237 // If the head region of the humongous region has been determined | |
| 1238 // to be alive, then all the tail regions should be marked | |
| 1239 // such as well. | |
| 1240 if (_region_bm->at(hum_start->hrs_index())) { | |
| 1241 _region_bm->par_at_put(hr->hrs_index(), 1); | |
| 1242 } | |
| 1243 return false; | |
| 1244 } | |
| 342 | 1245 |
| 1246 HeapWord* nextTop = hr->next_top_at_mark_start(); | |
| 1247 HeapWord* start = hr->top_at_conc_mark_count(); | |
| 1248 assert(hr->bottom() <= start && start <= hr->end() && | |
| 1249 hr->bottom() <= nextTop && nextTop <= hr->end() && | |
| 1250 start <= nextTop, | |
| 1251 "Preconditions."); | |
| 1252 // Otherwise, record the number of word's we'll examine. | |
| 1253 size_t words_done = (nextTop - start); | |
| 1254 // Find the first marked object at or after "start". | |
| 1255 start = _bm->getNextMarkedWordAddress(start, nextTop); | |
| 1256 size_t marked_bytes = 0; | |
| 1257 | |
| 1258 // Below, the term "card num" means the result of shifting an address | |
| 1259 // by the card shift -- address 0 corresponds to card number 0. One | |
| 1260 // must subtract the card num of the bottom of the heap to obtain a | |
| 1261 // card table index. | |
| 1262 // The first card num of the sequence of live cards currently being | |
| 1263 // constructed. -1 ==> no sequence. | |
| 1264 intptr_t start_card_num = -1; | |
| 1265 // The last card num of the sequence of live cards currently being | |
| 1266 // constructed. -1 ==> no sequence. | |
| 1267 intptr_t last_card_num = -1; | |
| 1268 | |
| 1269 while (start < nextTop) { | |
| 1270 if (_yield && _cm->do_yield_check()) { | |
| 1271 // We yielded. It might be for a full collection, in which case | |
| 1272 // all bets are off; terminate the traversal. | |
| 1273 if (_cm->has_aborted()) { | |
| 1274 _changed = false; | |
| 1275 return true; | |
| 1276 } else { | |
| 1277 // Otherwise, it might be a collection pause, and the region | |
| 1278 // we're looking at might be in the collection set. We'll | |
| 1279 // abandon this region. | |
| 1280 return false; | |
| 1281 } | |
| 1282 } | |
| 1283 oop obj = oop(start); | |
| 1284 int obj_sz = obj->size(); | |
| 1285 // The card num of the start of the current object. | |
| 1286 intptr_t obj_card_num = | |
| 1287 intptr_t(uintptr_t(start) >> CardTableModRefBS::card_shift); | |
| 1288 | |
| 1289 HeapWord* obj_last = start + obj_sz - 1; | |
| 1290 intptr_t obj_last_card_num = | |
| 1291 intptr_t(uintptr_t(obj_last) >> CardTableModRefBS::card_shift); | |
| 1292 | |
| 1293 if (obj_card_num != last_card_num) { | |
| 1294 if (start_card_num == -1) { | |
| 1295 assert(last_card_num == -1, "Both or neither."); | |
| 1296 start_card_num = obj_card_num; | |
| 1297 } else { | |
| 1298 assert(last_card_num != -1, "Both or neither."); | |
| 1299 assert(obj_card_num >= last_card_num, "Inv"); | |
| 1300 if ((obj_card_num - last_card_num) > 1) { | |
| 1301 // Mark the last run, and start a new one. | |
| 1302 mark_card_num_range(start_card_num, last_card_num); | |
| 1303 start_card_num = obj_card_num; | |
| 1304 } | |
| 1305 } | |
| 1306 #if CARD_BM_TEST_MODE | |
| 1307 /* | |
| 1308 gclog_or_tty->print_cr("Setting bits from %d/%d.", | |
| 1309 obj_card_num - _bottom_card_num, | |
| 1310 obj_last_card_num - _bottom_card_num); | |
| 1311 */ | |
| 1312 for (intptr_t j = obj_card_num; j <= obj_last_card_num; j++) { | |
| 1313 _card_bm->par_at_put(j - _bottom_card_num, 1); | |
| 1314 } | |
| 1315 #endif | |
| 1316 } | |
| 1317 // In any case, we set the last card num. | |
| 1318 last_card_num = obj_last_card_num; | |
| 1319 | |
| 1320 marked_bytes += obj_sz * HeapWordSize; | |
| 1321 // Find the next marked object after this one. | |
| 1322 start = _bm->getNextMarkedWordAddress(start + 1, nextTop); | |
| 1323 _changed = true; | |
| 1324 } | |
| 1325 // Handle the last range, if any. | |
| 1326 if (start_card_num != -1) | |
| 1327 mark_card_num_range(start_card_num, last_card_num); | |
| 1328 if (_final) { | |
| 1329 // Mark the allocated-since-marking portion... | |
| 1330 HeapWord* tp = hr->top(); | |
| 1331 if (nextTop < tp) { | |
| 1332 start_card_num = | |
| 1333 intptr_t(uintptr_t(nextTop) >> CardTableModRefBS::card_shift); | |
| 1334 last_card_num = | |
| 1335 intptr_t(uintptr_t(tp) >> CardTableModRefBS::card_shift); | |
| 1336 mark_card_num_range(start_card_num, last_card_num); | |
| 1337 // This definitely means the region has live objects. | |
| 1338 _region_bm->par_at_put(hr->hrs_index(), 1); | |
| 1339 } | |
| 1340 } | |
| 1341 | |
| 1342 hr->add_to_marked_bytes(marked_bytes); | |
| 1343 // Update the live region bitmap. | |
| 1344 if (marked_bytes > 0) { | |
| 1345 _region_bm->par_at_put(hr->hrs_index(), 1); | |
| 1346 } | |
| 1347 hr->set_top_at_conc_mark_count(nextTop); | |
| 1348 _tot_live += hr->next_live_bytes(); | |
| 1349 _tot_used += hr->used(); | |
| 1350 _words_done = words_done; | |
| 1351 | |
| 1352 if (!_final) { | |
| 1353 ++_regions_done; | |
| 1354 if (_regions_done % 10 == 0) { | |
| 1355 double end_vtime_sec = os::elapsedVTime(); | |
| 1356 double elapsed_vtime_sec = end_vtime_sec - _start_vtime_sec; | |
| 1357 if (elapsed_vtime_sec > (10.0 / 1000.0)) { | |
| 1358 jlong sleep_time_ms = | |
| 1359 (jlong) (elapsed_vtime_sec * _cm->cleanup_sleep_factor() * 1000.0); | |
| 1360 #if 0 | |
| 1361 gclog_or_tty->print_cr("CL: elapsed %1.4lf ms, sleep %1.4lf ms, " | |
| 1362 "overhead %1.4lf", | |
| 1363 elapsed_vtime_sec * 1000.0, (double) sleep_time_ms, | |
| 1364 _co_tracker->concOverhead(os::elapsedTime())); | |
| 1365 #endif | |
| 1366 os::sleep(Thread::current(), sleep_time_ms, false); | |
| 1367 _start_vtime_sec = end_vtime_sec; | |
| 1368 } | |
| 1369 } | |
| 1370 } | |
| 1371 | |
| 1372 return false; | |
| 1373 } | |
| 1374 | |
| 1375 bool changed() { return _changed; } | |
| 1376 void reset() { _changed = false; _words_done = 0; } | |
| 1377 void no_yield() { _yield = false; } | |
| 1378 size_t words_done() { return _words_done; } | |
| 1379 size_t tot_live() { return _tot_live; } | |
| 1380 size_t tot_used() { return _tot_used; } | |
| 1381 }; | |
| 1382 | |
| 1383 | |
| 1384 void ConcurrentMark::calcDesiredRegions() { | |
| 1385 guarantee( _cleanup_co_tracker.enabled(), "invariant" ); | |
| 1386 _cleanup_co_tracker.start(); | |
| 1387 | |
| 1388 _region_bm.clear(); | |
| 1389 _card_bm.clear(); | |
| 1390 CalcLiveObjectsClosure calccl(false /*final*/, | |
| 1391 nextMarkBitMap(), this, | |
| 1392 &_region_bm, &_card_bm, | |
| 1393 &_cleanup_co_tracker); | |
| 1394 G1CollectedHeap *g1h = G1CollectedHeap::heap(); | |
| 1395 g1h->heap_region_iterate(&calccl); | |
| 1396 | |
| 1397 do { | |
| 1398 calccl.reset(); | |
| 1399 g1h->heap_region_iterate(&calccl); | |
| 1400 } while (calccl.changed()); | |
| 1401 | |
| 1402 _cleanup_co_tracker.update(true); | |
| 1403 } | |
| 1404 | |
| 1405 class G1ParFinalCountTask: public AbstractGangTask { | |
| 1406 protected: | |
| 1407 G1CollectedHeap* _g1h; | |
| 1408 CMBitMap* _bm; | |
| 1409 size_t _n_workers; | |
| 1410 size_t *_live_bytes; | |
| 1411 size_t *_used_bytes; | |
| 1412 BitMap* _region_bm; | |
| 1413 BitMap* _card_bm; | |
| 1414 public: | |
| 1415 G1ParFinalCountTask(G1CollectedHeap* g1h, CMBitMap* bm, | |
| 1416 BitMap* region_bm, BitMap* card_bm) : | |
| 1417 AbstractGangTask("G1 final counting"), _g1h(g1h), | |
| 1418 _bm(bm), _region_bm(region_bm), _card_bm(card_bm) | |
| 1419 { | |
| 1420 if (ParallelGCThreads > 0) | |
| 1421 _n_workers = _g1h->workers()->total_workers(); | |
| 1422 else | |
| 1423 _n_workers = 1; | |
| 1424 _live_bytes = NEW_C_HEAP_ARRAY(size_t, _n_workers); | |
| 1425 _used_bytes = NEW_C_HEAP_ARRAY(size_t, _n_workers); | |
| 1426 } | |
| 1427 | |
| 1428 ~G1ParFinalCountTask() { | |
| 1429 FREE_C_HEAP_ARRAY(size_t, _live_bytes); | |
| 1430 FREE_C_HEAP_ARRAY(size_t, _used_bytes); | |
| 1431 } | |
| 1432 | |
| 1433 void work(int i) { | |
| 1434 CalcLiveObjectsClosure calccl(true /*final*/, | |
| 1435 _bm, _g1h->concurrent_mark(), | |
| 1436 _region_bm, _card_bm, | |
| 1437 NULL /* CO tracker */); | |
| 1438 calccl.no_yield(); | |
| 1439 if (ParallelGCThreads > 0) { | |
| 355 | 1440 _g1h->heap_region_par_iterate_chunked(&calccl, i, |
| 1441 HeapRegion::FinalCountClaimValue); | |
| 342 | 1442 } else { |
| 1443 _g1h->heap_region_iterate(&calccl); | |
| 1444 } | |
| 1445 assert(calccl.complete(), "Shouldn't have yielded!"); | |
| 1446 | |
| 1447 guarantee( (size_t)i < _n_workers, "invariant" ); | |
| 1448 _live_bytes[i] = calccl.tot_live(); | |
| 1449 _used_bytes[i] = calccl.tot_used(); | |
| 1450 } | |
| 1451 size_t live_bytes() { | |
| 1452 size_t live_bytes = 0; | |
| 1453 for (size_t i = 0; i < _n_workers; ++i) | |
| 1454 live_bytes += _live_bytes[i]; | |
| 1455 return live_bytes; | |
| 1456 } | |
| 1457 size_t used_bytes() { | |
| 1458 size_t used_bytes = 0; | |
| 1459 for (size_t i = 0; i < _n_workers; ++i) | |
| 1460 used_bytes += _used_bytes[i]; | |
| 1461 return used_bytes; | |
| 1462 } | |
| 1463 }; | |
| 1464 | |
| 1465 class G1ParNoteEndTask; | |
| 1466 | |
| 1467 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure { | |
| 1468 G1CollectedHeap* _g1; | |
| 1469 int _worker_num; | |
| 1470 size_t _max_live_bytes; | |
| 1471 size_t _regions_claimed; | |
| 1472 size_t _freed_bytes; | |
| 1473 size_t _cleared_h_regions; | |
| 1474 size_t _freed_regions; | |
| 1475 UncleanRegionList* _unclean_region_list; | |
| 1476 double _claimed_region_time; | |
| 1477 double _max_region_time; | |
| 1478 | |
| 1479 public: | |
| 1480 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1, | |
| 1481 UncleanRegionList* list, | |
| 1482 int worker_num); | |
| 1483 size_t freed_bytes() { return _freed_bytes; } | |
| 1484 size_t cleared_h_regions() { return _cleared_h_regions; } | |
| 1485 size_t freed_regions() { return _freed_regions; } | |
| 1486 UncleanRegionList* unclean_region_list() { | |
| 1487 return _unclean_region_list; | |
| 1488 } | |
| 1489 | |
| 1490 bool doHeapRegion(HeapRegion *r); | |
| 1491 | |
| 1492 size_t max_live_bytes() { return _max_live_bytes; } | |
| 1493 size_t regions_claimed() { return _regions_claimed; } | |
| 1494 double claimed_region_time_sec() { return _claimed_region_time; } | |
| 1495 double max_region_time_sec() { return _max_region_time; } | |
| 1496 }; | |
| 1497 | |
| 1498 class G1ParNoteEndTask: public AbstractGangTask { | |
| 1499 friend class G1NoteEndOfConcMarkClosure; | |
| 1500 protected: | |
| 1501 G1CollectedHeap* _g1h; | |
| 1502 size_t _max_live_bytes; | |
| 1503 size_t _freed_bytes; | |
| 1504 ConcurrentMark::ParCleanupThreadState** _par_cleanup_thread_state; | |
| 1505 public: | |
| 1506 G1ParNoteEndTask(G1CollectedHeap* g1h, | |
| 1507 ConcurrentMark::ParCleanupThreadState** | |
| 1508 par_cleanup_thread_state) : | |
| 1509 AbstractGangTask("G1 note end"), _g1h(g1h), | |
| 1510 _max_live_bytes(0), _freed_bytes(0), | |
| 1511 _par_cleanup_thread_state(par_cleanup_thread_state) | |
| 1512 {} | |
| 1513 | |
| 1514 void work(int i) { | |
| 1515 double start = os::elapsedTime(); | |
| 1516 G1NoteEndOfConcMarkClosure g1_note_end(_g1h, | |
| 1517 &_par_cleanup_thread_state[i]->list, | |
| 1518 i); | |
| 1519 if (ParallelGCThreads > 0) { | |
| 355 | 1520 _g1h->heap_region_par_iterate_chunked(&g1_note_end, i, |
| 1521 HeapRegion::NoteEndClaimValue); | |
| 342 | 1522 } else { |
| 1523 _g1h->heap_region_iterate(&g1_note_end); | |
| 1524 } | |
| 1525 assert(g1_note_end.complete(), "Shouldn't have yielded!"); | |
| 1526 | |
| 1527 // Now finish up freeing the current thread's regions. | |
| 1528 _g1h->finish_free_region_work(g1_note_end.freed_bytes(), | |
| 1529 g1_note_end.cleared_h_regions(), | |
| 1530 0, NULL); | |
| 1531 { | |
| 1532 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); | |
| 1533 _max_live_bytes += g1_note_end.max_live_bytes(); | |
| 1534 _freed_bytes += g1_note_end.freed_bytes(); | |
| 1535 } | |
| 1536 double end = os::elapsedTime(); | |
| 1537 if (G1PrintParCleanupStats) { | |
| 1538 gclog_or_tty->print(" Worker thread %d [%8.3f..%8.3f = %8.3f ms] " | |
| 1539 "claimed %d regions (tot = %8.3f ms, max = %8.3f ms).\n", | |
| 1540 i, start, end, (end-start)*1000.0, | |
| 1541 g1_note_end.regions_claimed(), | |
| 1542 g1_note_end.claimed_region_time_sec()*1000.0, | |
| 1543 g1_note_end.max_region_time_sec()*1000.0); | |
| 1544 } | |
| 1545 } | |
| 1546 size_t max_live_bytes() { return _max_live_bytes; } | |
| 1547 size_t freed_bytes() { return _freed_bytes; } | |
| 1548 }; | |
| 1549 | |
| 1550 class G1ParScrubRemSetTask: public AbstractGangTask { | |
| 1551 protected: | |
| 1552 G1RemSet* _g1rs; | |
| 1553 BitMap* _region_bm; | |
| 1554 BitMap* _card_bm; | |
| 1555 public: | |
| 1556 G1ParScrubRemSetTask(G1CollectedHeap* g1h, | |
| 1557 BitMap* region_bm, BitMap* card_bm) : | |
| 1558 AbstractGangTask("G1 ScrubRS"), _g1rs(g1h->g1_rem_set()), | |
| 1559 _region_bm(region_bm), _card_bm(card_bm) | |
| 1560 {} | |
| 1561 | |
| 1562 void work(int i) { | |
| 1563 if (ParallelGCThreads > 0) { | |
| 355 | 1564 _g1rs->scrub_par(_region_bm, _card_bm, i, |
| 1565 HeapRegion::ScrubRemSetClaimValue); | |
| 342 | 1566 } else { |
| 1567 _g1rs->scrub(_region_bm, _card_bm); | |
| 1568 } | |
| 1569 } | |
| 1570 | |
| 1571 }; | |
| 1572 | |
| 1573 G1NoteEndOfConcMarkClosure:: | |
| 1574 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1, | |
| 1575 UncleanRegionList* list, | |
| 1576 int worker_num) | |
| 1577 : _g1(g1), _worker_num(worker_num), | |
| 1578 _max_live_bytes(0), _regions_claimed(0), | |
| 1579 _freed_bytes(0), _cleared_h_regions(0), _freed_regions(0), | |
| 1580 _claimed_region_time(0.0), _max_region_time(0.0), | |
| 1581 _unclean_region_list(list) | |
| 1582 {} | |
| 1583 | |
| 1584 bool G1NoteEndOfConcMarkClosure::doHeapRegion(HeapRegion *r) { | |
| 1585 // We use a claim value of zero here because all regions | |
| 1586 // were claimed with value 1 in the FinalCount task. | |
| 1587 r->reset_gc_time_stamp(); | |
| 1588 if (!r->continuesHumongous()) { | |
| 1589 double start = os::elapsedTime(); | |
| 1590 _regions_claimed++; | |
| 1591 r->note_end_of_marking(); | |
| 1592 _max_live_bytes += r->max_live_bytes(); | |
| 1593 _g1->free_region_if_totally_empty_work(r, | |
| 1594 _freed_bytes, | |
| 1595 _cleared_h_regions, | |
| 1596 _freed_regions, | |
| 1597 _unclean_region_list, | |
| 1598 true /*par*/); | |
| 1599 double region_time = (os::elapsedTime() - start); | |
| 1600 _claimed_region_time += region_time; | |
| 1601 if (region_time > _max_region_time) _max_region_time = region_time; | |
| 1602 } | |
| 1603 return false; | |
| 1604 } | |
| 1605 | |
| 1606 void ConcurrentMark::cleanup() { | |
| 1607 // world is stopped at this checkpoint | |
| 1608 assert(SafepointSynchronize::is_at_safepoint(), | |
| 1609 "world should be stopped"); | |
| 1610 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
| 1611 | |
| 1612 // If a full collection has happened, we shouldn't do this. | |
| 1613 if (has_aborted()) { | |
| 1614 g1h->set_marking_complete(); // So bitmap clearing isn't confused | |
| 1615 return; | |
| 1616 } | |
| 1617 | |
| 1618 _cleanup_co_tracker.disable(); | |
| 1619 | |
| 1620 G1CollectorPolicy* g1p = G1CollectedHeap::heap()->g1_policy(); | |
| 1621 g1p->record_concurrent_mark_cleanup_start(); | |
| 1622 | |
| 1623 double start = os::elapsedTime(); | |
| 1624 GCOverheadReporter::recordSTWStart(start); | |
| 1625 | |
| 1626 // Do counting once more with the world stopped for good measure. | |
| 1627 G1ParFinalCountTask g1_par_count_task(g1h, nextMarkBitMap(), | |
| 1628 &_region_bm, &_card_bm); | |
| 1629 if (ParallelGCThreads > 0) { | |
| 355 | 1630 assert(g1h->check_heap_region_claim_values( |
| 1631 HeapRegion::InitialClaimValue), | |
| 1632 "sanity check"); | |
| 1633 | |
| 342 | 1634 int n_workers = g1h->workers()->total_workers(); |
| 1635 g1h->set_par_threads(n_workers); | |
| 1636 g1h->workers()->run_task(&g1_par_count_task); | |
| 1637 g1h->set_par_threads(0); | |
| 355 | 1638 |
| 1639 assert(g1h->check_heap_region_claim_values( | |
| 1640 HeapRegion::FinalCountClaimValue), | |
| 1641 "sanity check"); | |
| 342 | 1642 } else { |
| 1643 g1_par_count_task.work(0); | |
| 1644 } | |
| 1645 | |
| 1646 size_t known_garbage_bytes = | |
| 1647 g1_par_count_task.used_bytes() - g1_par_count_task.live_bytes(); | |
| 1648 #if 0 | |
| 1649 gclog_or_tty->print_cr("used %1.2lf, live %1.2lf, garbage %1.2lf", | |
| 1650 (double) g1_par_count_task.used_bytes() / (double) (1024 * 1024), | |
| 1651 (double) g1_par_count_task.live_bytes() / (double) (1024 * 1024), | |
| 1652 (double) known_garbage_bytes / (double) (1024 * 1024)); | |
| 1653 #endif // 0 | |
| 1654 g1p->set_known_garbage_bytes(known_garbage_bytes); | |
| 1655 | |
| 1656 size_t start_used_bytes = g1h->used(); | |
| 1657 _at_least_one_mark_complete = true; | |
| 1658 g1h->set_marking_complete(); | |
| 1659 | |
| 1660 double count_end = os::elapsedTime(); | |
| 1661 double this_final_counting_time = (count_end - start); | |
| 1662 if (G1PrintParCleanupStats) { | |
| 1663 gclog_or_tty->print_cr("Cleanup:"); | |
| 1664 gclog_or_tty->print_cr(" Finalize counting: %8.3f ms", | |
| 1665 this_final_counting_time*1000.0); | |
| 1666 } | |
| 1667 _total_counting_time += this_final_counting_time; | |
| 1668 | |
| 1669 // Install newly created mark bitMap as "prev". | |
| 1670 swapMarkBitMaps(); | |
| 1671 | |
| 1672 g1h->reset_gc_time_stamp(); | |
| 1673 | |
| 1674 // Note end of marking in all heap regions. | |
| 1675 double note_end_start = os::elapsedTime(); | |
| 1676 G1ParNoteEndTask g1_par_note_end_task(g1h, _par_cleanup_thread_state); | |
| 1677 if (ParallelGCThreads > 0) { | |
| 1678 int n_workers = g1h->workers()->total_workers(); | |
| 1679 g1h->set_par_threads(n_workers); | |
| 1680 g1h->workers()->run_task(&g1_par_note_end_task); | |
| 1681 g1h->set_par_threads(0); | |
| 355 | 1682 |
| 1683 assert(g1h->check_heap_region_claim_values(HeapRegion::NoteEndClaimValue), | |
| 1684 "sanity check"); | |
| 342 | 1685 } else { |
| 1686 g1_par_note_end_task.work(0); | |
| 1687 } | |
| 1688 g1h->set_unclean_regions_coming(true); | |
| 1689 double note_end_end = os::elapsedTime(); | |
| 1690 // Tell the mutators that there might be unclean regions coming... | |
| 1691 if (G1PrintParCleanupStats) { | |
| 1692 gclog_or_tty->print_cr(" note end of marking: %8.3f ms.", | |
| 1693 (note_end_end - note_end_start)*1000.0); | |
| 1694 } | |
| 1695 | |
| 355 | 1696 |
| 342 | 1697 // call below, since it affects the metric by which we sort the heap |
| 1698 // regions. | |
| 1699 if (G1ScrubRemSets) { | |
| 1700 double rs_scrub_start = os::elapsedTime(); | |
| 1701 G1ParScrubRemSetTask g1_par_scrub_rs_task(g1h, &_region_bm, &_card_bm); | |
| 1702 if (ParallelGCThreads > 0) { | |
| 1703 int n_workers = g1h->workers()->total_workers(); | |
| 1704 g1h->set_par_threads(n_workers); | |
| 1705 g1h->workers()->run_task(&g1_par_scrub_rs_task); | |
| 1706 g1h->set_par_threads(0); | |
| 355 | 1707 |
| 1708 assert(g1h->check_heap_region_claim_values( | |
| 1709 HeapRegion::ScrubRemSetClaimValue), | |
| 1710 "sanity check"); | |
| 342 | 1711 } else { |
| 1712 g1_par_scrub_rs_task.work(0); | |
| 1713 } | |
| 1714 | |
| 1715 double rs_scrub_end = os::elapsedTime(); | |
| 1716 double this_rs_scrub_time = (rs_scrub_end - rs_scrub_start); | |
| 1717 _total_rs_scrub_time += this_rs_scrub_time; | |
| 1718 } | |
| 1719 | |
| 1720 // this will also free any regions totally full of garbage objects, | |
| 1721 // and sort the regions. | |
| 1722 g1h->g1_policy()->record_concurrent_mark_cleanup_end( | |
| 1723 g1_par_note_end_task.freed_bytes(), | |
| 1724 g1_par_note_end_task.max_live_bytes()); | |
| 1725 | |
| 1726 // Statistics. | |
| 1727 double end = os::elapsedTime(); | |
| 1728 _cleanup_times.add((end - start) * 1000.0); | |
| 1729 GCOverheadReporter::recordSTWEnd(end); | |
| 1730 | |
| 1731 // G1CollectedHeap::heap()->print(); | |
| 1732 // gclog_or_tty->print_cr("HEAP GC TIME STAMP : %d", | |
| 1733 // G1CollectedHeap::heap()->get_gc_time_stamp()); | |
| 1734 | |
| 1735 if (PrintGC || PrintGCDetails) { | |
| 1736 g1h->print_size_transition(gclog_or_tty, | |
| 1737 start_used_bytes, | |
| 1738 g1h->used(), | |
| 1739 g1h->capacity()); | |
| 1740 } | |
| 1741 | |
| 1742 size_t cleaned_up_bytes = start_used_bytes - g1h->used(); | |
| 1743 g1p->decrease_known_garbage_bytes(cleaned_up_bytes); | |
| 1744 | |
| 1745 // We need to make this be a "collection" so any collection pause that | |
| 1746 // races with it goes around and waits for completeCleanup to finish. | |
| 1747 g1h->increment_total_collections(); | |
| 1748 | |
| 1749 #ifndef PRODUCT | |
| 1750 if (G1VerifyConcMark) { | |
| 1751 G1CollectedHeap::heap()->prepare_for_verify(); | |
| 1752 G1CollectedHeap::heap()->verify(true,false); | |
| 1753 } | |
| 1754 #endif | |
| 1755 } | |
| 1756 | |
| 1757 void ConcurrentMark::completeCleanup() { | |
| 1758 // A full collection intervened. | |
| 1759 if (has_aborted()) return; | |
| 1760 | |
| 1761 int first = 0; | |
| 1762 int last = (int)MAX2(ParallelGCThreads, (size_t)1); | |
| 1763 for (int t = 0; t < last; t++) { | |
| 1764 UncleanRegionList* list = &_par_cleanup_thread_state[t]->list; | |
| 1765 assert(list->well_formed(), "Inv"); | |
| 1766 HeapRegion* hd = list->hd(); | |
| 1767 while (hd != NULL) { | |
| 1768 // Now finish up the other stuff. | |
| 1769 hd->rem_set()->clear(); | |
| 1770 HeapRegion* next_hd = hd->next_from_unclean_list(); | |
| 1771 (void)list->pop(); | |
| 1772 guarantee(list->hd() == next_hd, "how not?"); | |
| 1773 _g1h->put_region_on_unclean_list(hd); | |
| 1774 if (!hd->isHumongous()) { | |
| 1775 // Add this to the _free_regions count by 1. | |
| 1776 _g1h->finish_free_region_work(0, 0, 1, NULL); | |
| 1777 } | |
| 1778 hd = list->hd(); | |
| 1779 guarantee(hd == next_hd, "how not?"); | |
| 1780 } | |
| 1781 } | |
| 1782 } | |
| 1783 | |
| 1784 | |
| 1785 class G1CMIsAliveClosure: public BoolObjectClosure { | |
| 1786 G1CollectedHeap* _g1; | |
| 1787 public: | |
| 1788 G1CMIsAliveClosure(G1CollectedHeap* g1) : | |
| 1789 _g1(g1) | |
| 1790 {} | |
| 1791 | |
| 1792 void do_object(oop obj) { | |
| 1793 assert(false, "not to be invoked"); | |
| 1794 } | |
| 1795 bool do_object_b(oop obj) { | |
| 1796 HeapWord* addr = (HeapWord*)obj; | |
| 1797 return addr != NULL && | |
| 1798 (!_g1->is_in_g1_reserved(addr) || !_g1->is_obj_ill(obj)); | |
| 1799 } | |
| 1800 }; | |
| 1801 | |
| 1802 class G1CMKeepAliveClosure: public OopClosure { | |
| 1803 G1CollectedHeap* _g1; | |
| 1804 ConcurrentMark* _cm; | |
| 1805 CMBitMap* _bitMap; | |
| 1806 public: | |
| 1807 G1CMKeepAliveClosure(G1CollectedHeap* g1, ConcurrentMark* cm, | |
| 1808 CMBitMap* bitMap) : | |
| 1809 _g1(g1), _cm(cm), | |
| 1810 _bitMap(bitMap) {} | |
| 1811 | |
| 1812 void do_oop(narrowOop* p) { | |
| 1813 guarantee(false, "NYI"); | |
| 1814 } | |
| 1815 | |
| 1816 void do_oop(oop* p) { | |
| 1817 oop thisOop = *p; | |
| 1818 HeapWord* addr = (HeapWord*)thisOop; | |
| 1819 if (_g1->is_in_g1_reserved(addr) && _g1->is_obj_ill(thisOop)) { | |
| 1820 _bitMap->mark(addr); | |
| 1821 _cm->mark_stack_push(thisOop); | |
| 1822 } | |
| 1823 } | |
| 1824 }; | |
| 1825 | |
| 1826 class G1CMDrainMarkingStackClosure: public VoidClosure { | |
| 1827 CMMarkStack* _markStack; | |
| 1828 CMBitMap* _bitMap; | |
| 1829 G1CMKeepAliveClosure* _oopClosure; | |
| 1830 public: | |
| 1831 G1CMDrainMarkingStackClosure(CMBitMap* bitMap, CMMarkStack* markStack, | |
| 1832 G1CMKeepAliveClosure* oopClosure) : | |
| 1833 _bitMap(bitMap), | |
| 1834 _markStack(markStack), | |
| 1835 _oopClosure(oopClosure) | |
| 1836 {} | |
| 1837 | |
| 1838 void do_void() { | |
| 1839 _markStack->drain((OopClosure*)_oopClosure, _bitMap, false); | |
| 1840 } | |
| 1841 }; | |
| 1842 | |
| 1843 void ConcurrentMark::weakRefsWork(bool clear_all_soft_refs) { | |
| 1844 ResourceMark rm; | |
| 1845 HandleMark hm; | |
|
453
c96030fff130
6684579: SoftReference processing can be made more efficient
ysr
parents:
355
diff
changeset
|
1846 G1CollectedHeap* g1h = G1CollectedHeap::heap(); |
|
c96030fff130
6684579: SoftReference processing can be made more efficient
ysr
parents:
355
diff
changeset
|
1847 ReferenceProcessor* rp = g1h->ref_processor(); |
| 342 | 1848 |
| 1849 // Process weak references. | |
|
457
27a80744a83b
6778647: snap(), snap_policy() should be renamed setup(), setup_policy()
ysr
parents:
453
diff
changeset
|
1850 rp->setup_policy(clear_all_soft_refs); |
| 342 | 1851 assert(_markStack.isEmpty(), "mark stack should be empty"); |
| 1852 | |
|
453
c96030fff130
6684579: SoftReference processing can be made more efficient
ysr
parents:
355
diff
changeset
|
1853 G1CMIsAliveClosure g1IsAliveClosure (g1h); |
|
c96030fff130
6684579: SoftReference processing can be made more efficient
ysr
parents:
355
diff
changeset
|
1854 G1CMKeepAliveClosure g1KeepAliveClosure(g1h, this, nextMarkBitMap()); |
| 342 | 1855 G1CMDrainMarkingStackClosure |
| 1856 g1DrainMarkingStackClosure(nextMarkBitMap(), &_markStack, | |
| 1857 &g1KeepAliveClosure); | |
| 1858 | |
| 1859 // XXXYYY Also: copy the parallel ref processing code from CMS. | |
|
453
c96030fff130
6684579: SoftReference processing can be made more efficient
ysr
parents:
355
diff
changeset
|
1860 rp->process_discovered_references(&g1IsAliveClosure, |
| 342 | 1861 &g1KeepAliveClosure, |
| 1862 &g1DrainMarkingStackClosure, | |
| 1863 NULL); | |
| 1864 assert(_markStack.overflow() || _markStack.isEmpty(), | |
| 1865 "mark stack should be empty (unless it overflowed)"); | |
| 1866 if (_markStack.overflow()) { | |
| 1867 set_has_overflown(); | |
| 1868 } | |
| 1869 | |
| 1870 rp->enqueue_discovered_references(); | |
| 1871 rp->verify_no_references_recorded(); | |
| 1872 assert(!rp->discovery_enabled(), "should have been disabled"); | |
| 1873 | |
| 1874 // Now clean up stale oops in SymbolTable and StringTable | |
| 1875 SymbolTable::unlink(&g1IsAliveClosure); | |
| 1876 StringTable::unlink(&g1IsAliveClosure); | |
| 1877 } | |
| 1878 | |
| 1879 void ConcurrentMark::swapMarkBitMaps() { | |
| 1880 CMBitMapRO* temp = _prevMarkBitMap; | |
| 1881 _prevMarkBitMap = (CMBitMapRO*)_nextMarkBitMap; | |
| 1882 _nextMarkBitMap = (CMBitMap*) temp; | |
| 1883 } | |
| 1884 | |
| 1885 class CMRemarkTask: public AbstractGangTask { | |
| 1886 private: | |
| 1887 ConcurrentMark *_cm; | |
| 1888 | |
| 1889 public: | |
| 1890 void work(int worker_i) { | |
| 1891 // Since all available tasks are actually started, we should | |
| 1892 // only proceed if we're supposed to be actived. | |
| 1893 if ((size_t)worker_i < _cm->active_tasks()) { | |
| 1894 CMTask* task = _cm->task(worker_i); | |
| 1895 task->record_start_time(); | |
| 1896 do { | |
| 1897 task->do_marking_step(1000000000.0 /* something very large */); | |
| 1898 } while (task->has_aborted() && !_cm->has_overflown()); | |
| 1899 // If we overflow, then we do not want to restart. We instead | |
| 1900 // want to abort remark and do concurrent marking again. | |
| 1901 task->record_end_time(); | |
| 1902 } | |
| 1903 } | |
| 1904 | |
| 1905 CMRemarkTask(ConcurrentMark* cm) : | |
| 1906 AbstractGangTask("Par Remark"), _cm(cm) { } | |
| 1907 }; | |
| 1908 | |
| 1909 void ConcurrentMark::checkpointRootsFinalWork() { | |
| 1910 ResourceMark rm; | |
| 1911 HandleMark hm; | |
| 1912 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
| 1913 | |
| 1914 g1h->ensure_parsability(false); | |
| 1915 | |
| 1916 if (ParallelGCThreads > 0) { | |
| 1917 g1h->change_strong_roots_parity(); | |
| 1918 // this is remark, so we'll use up all available threads | |
| 1919 int active_workers = ParallelGCThreads; | |
| 1920 set_phase(active_workers, false); | |
| 1921 | |
| 1922 CMRemarkTask remarkTask(this); | |
| 1923 // We will start all available threads, even if we decide that the | |
| 1924 // active_workers will be fewer. The extra ones will just bail out | |
| 1925 // immediately. | |
| 1926 int n_workers = g1h->workers()->total_workers(); | |
| 1927 g1h->set_par_threads(n_workers); | |
| 1928 g1h->workers()->run_task(&remarkTask); | |
| 1929 g1h->set_par_threads(0); | |
| 1930 | |
| 1931 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
| 1932 guarantee( satb_mq_set.completed_buffers_num() == 0, "invariant" ); | |
| 1933 } else { | |
| 1934 g1h->change_strong_roots_parity(); | |
| 1935 // this is remark, so we'll use up all available threads | |
| 1936 int active_workers = 1; | |
| 1937 set_phase(active_workers, false); | |
| 1938 | |
| 1939 CMRemarkTask remarkTask(this); | |
| 1940 // We will start all available threads, even if we decide that the | |
| 1941 // active_workers will be fewer. The extra ones will just bail out | |
| 1942 // immediately. | |
| 1943 remarkTask.work(0); | |
| 1944 | |
| 1945 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
| 1946 guarantee( satb_mq_set.completed_buffers_num() == 0, "invariant" ); | |
| 1947 } | |
| 1948 | |
| 1949 print_stats(); | |
| 1950 | |
| 1951 if (!restart_for_overflow()) | |
| 1952 set_non_marking_state(); | |
| 1953 | |
| 1954 #if VERIFY_OBJS_PROCESSED | |
| 1955 if (_scan_obj_cl.objs_processed != ThreadLocalObjQueue::objs_enqueued) { | |
| 1956 gclog_or_tty->print_cr("Processed = %d, enqueued = %d.", | |
| 1957 _scan_obj_cl.objs_processed, | |
| 1958 ThreadLocalObjQueue::objs_enqueued); | |
| 1959 guarantee(_scan_obj_cl.objs_processed == | |
| 1960 ThreadLocalObjQueue::objs_enqueued, | |
| 1961 "Different number of objs processed and enqueued."); | |
| 1962 } | |
| 1963 #endif | |
| 1964 } | |
| 1965 | |
| 1966 class ReachablePrinterOopClosure: public OopClosure { | |
| 1967 private: | |
| 1968 G1CollectedHeap* _g1h; | |
| 1969 CMBitMapRO* _bitmap; | |
| 1970 outputStream* _out; | |
| 1971 | |
| 1972 public: | |
| 1973 ReachablePrinterOopClosure(CMBitMapRO* bitmap, outputStream* out) : | |
| 1974 _bitmap(bitmap), _g1h(G1CollectedHeap::heap()), _out(out) { } | |
| 1975 | |
| 1976 void do_oop(narrowOop* p) { | |
| 1977 guarantee(false, "NYI"); | |
| 1978 } | |
| 1979 | |
| 1980 void do_oop(oop* p) { | |
| 1981 oop obj = *p; | |
| 1982 const char* str = NULL; | |
| 1983 const char* str2 = ""; | |
| 1984 | |
| 1985 if (!_g1h->is_in_g1_reserved(obj)) | |
| 1986 str = "outside G1 reserved"; | |
| 1987 else { | |
| 1988 HeapRegion* hr = _g1h->heap_region_containing(obj); | |
| 1989 guarantee( hr != NULL, "invariant" ); | |
| 1990 if (hr->obj_allocated_since_prev_marking(obj)) { | |
| 1991 str = "over TAMS"; | |
| 1992 if (_bitmap->isMarked((HeapWord*) obj)) | |
| 1993 str2 = " AND MARKED"; | |
| 1994 } else if (_bitmap->isMarked((HeapWord*) obj)) | |
| 1995 str = "marked"; | |
| 1996 else | |
| 1997 str = "#### NOT MARKED ####"; | |
| 1998 } | |
| 1999 | |
| 2000 _out->print_cr(" "PTR_FORMAT" contains "PTR_FORMAT" %s%s", | |
| 2001 p, (void*) obj, str, str2); | |
| 2002 } | |
| 2003 }; | |
| 2004 | |
| 2005 class ReachablePrinterClosure: public BitMapClosure { | |
| 2006 private: | |
| 2007 CMBitMapRO* _bitmap; | |
| 2008 outputStream* _out; | |
| 2009 | |
| 2010 public: | |
| 2011 ReachablePrinterClosure(CMBitMapRO* bitmap, outputStream* out) : | |
| 2012 _bitmap(bitmap), _out(out) { } | |
| 2013 | |
| 2014 bool do_bit(size_t offset) { | |
| 2015 HeapWord* addr = _bitmap->offsetToHeapWord(offset); | |
| 2016 ReachablePrinterOopClosure oopCl(_bitmap, _out); | |
| 2017 | |
| 2018 _out->print_cr(" obj "PTR_FORMAT", offset %10d (marked)", addr, offset); | |
| 2019 oop(addr)->oop_iterate(&oopCl); | |
| 2020 _out->print_cr(""); | |
| 2021 | |
| 2022 return true; | |
| 2023 } | |
| 2024 }; | |
| 2025 | |
| 2026 class ObjInRegionReachablePrinterClosure : public ObjectClosure { | |
| 2027 private: | |
| 2028 CMBitMapRO* _bitmap; | |
| 2029 outputStream* _out; | |
| 2030 | |
| 2031 public: | |
| 2032 void do_object(oop o) { | |
| 2033 ReachablePrinterOopClosure oopCl(_bitmap, _out); | |
| 2034 | |
| 2035 _out->print_cr(" obj "PTR_FORMAT" (over TAMS)", (void*) o); | |
| 2036 o->oop_iterate(&oopCl); | |
| 2037 _out->print_cr(""); | |
| 2038 } | |
| 2039 | |
| 2040 ObjInRegionReachablePrinterClosure(CMBitMapRO* bitmap, outputStream* out) : | |
| 2041 _bitmap(bitmap), _out(out) { } | |
| 2042 }; | |
| 2043 | |
| 2044 class RegionReachablePrinterClosure : public HeapRegionClosure { | |
| 2045 private: | |
| 2046 CMBitMapRO* _bitmap; | |
| 2047 outputStream* _out; | |
| 2048 | |
| 2049 public: | |
| 2050 bool doHeapRegion(HeapRegion* hr) { | |
| 2051 HeapWord* b = hr->bottom(); | |
| 2052 HeapWord* e = hr->end(); | |
| 2053 HeapWord* t = hr->top(); | |
| 2054 HeapWord* p = hr->prev_top_at_mark_start(); | |
| 2055 _out->print_cr("** ["PTR_FORMAT", "PTR_FORMAT"] top: "PTR_FORMAT" " | |
| 2056 "PTAMS: "PTR_FORMAT, b, e, t, p); | |
| 2057 _out->print_cr(""); | |
| 2058 | |
| 2059 ObjInRegionReachablePrinterClosure ocl(_bitmap, _out); | |
| 2060 hr->object_iterate_mem_careful(MemRegion(p, t), &ocl); | |
| 2061 | |
| 2062 return false; | |
| 2063 } | |
| 2064 | |
| 2065 RegionReachablePrinterClosure(CMBitMapRO* bitmap, | |
| 2066 outputStream* out) : | |
| 2067 _bitmap(bitmap), _out(out) { } | |
| 2068 }; | |
| 2069 | |
| 2070 void ConcurrentMark::print_prev_bitmap_reachable() { | |
| 2071 outputStream* out = gclog_or_tty; | |
| 2072 | |
| 2073 #if SEND_HEAP_DUMP_TO_FILE | |
| 2074 guarantee(heap_dump_file == NULL, "Protocol"); | |
| 2075 char fn_buf[100]; | |
| 2076 sprintf(fn_buf, "/tmp/dump.txt.%d", os::current_process_id()); | |
| 2077 heap_dump_file = fopen(fn_buf, "w"); | |
| 2078 fileStream fstream(heap_dump_file); | |
| 2079 out = &fstream; | |
| 2080 #endif // SEND_HEAP_DUMP_TO_FILE | |
| 2081 | |
| 2082 RegionReachablePrinterClosure rcl(_prevMarkBitMap, out); | |
| 2083 out->print_cr("--- ITERATING OVER REGIONS WITH PTAMS < TOP"); | |
| 2084 _g1h->heap_region_iterate(&rcl); | |
| 2085 out->print_cr(""); | |
| 2086 | |
| 2087 ReachablePrinterClosure cl(_prevMarkBitMap, out); | |
| 2088 out->print_cr("--- REACHABLE OBJECTS ON THE BITMAP"); | |
| 2089 _prevMarkBitMap->iterate(&cl); | |
| 2090 out->print_cr(""); | |
| 2091 | |
| 2092 #if SEND_HEAP_DUMP_TO_FILE | |
| 2093 fclose(heap_dump_file); | |
| 2094 heap_dump_file = NULL; | |
| 2095 #endif // SEND_HEAP_DUMP_TO_FILE | |
| 2096 } | |
| 2097 | |
| 2098 // This note is for drainAllSATBBuffers and the code in between. | |
| 2099 // In the future we could reuse a task to do this work during an | |
| 2100 // evacuation pause (since now tasks are not active and can be claimed | |
| 2101 // during an evacuation pause). This was a late change to the code and | |
| 2102 // is currently not being taken advantage of. | |
| 2103 | |
| 2104 class CMGlobalObjectClosure : public ObjectClosure { | |
| 2105 private: | |
| 2106 ConcurrentMark* _cm; | |
| 2107 | |
| 2108 public: | |
| 2109 void do_object(oop obj) { | |
| 2110 _cm->deal_with_reference(obj); | |
| 2111 } | |
| 2112 | |
| 2113 CMGlobalObjectClosure(ConcurrentMark* cm) : _cm(cm) { } | |
| 2114 }; | |
| 2115 | |
| 2116 void ConcurrentMark::deal_with_reference(oop obj) { | |
| 2117 if (verbose_high()) | |
| 2118 gclog_or_tty->print_cr("[global] we're dealing with reference "PTR_FORMAT, | |
| 2119 (void*) obj); | |
| 2120 | |
| 2121 | |
| 2122 HeapWord* objAddr = (HeapWord*) obj; | |
| 2123 if (_g1h->is_in_g1_reserved(objAddr)) { | |
| 2124 tmp_guarantee_CM( obj != NULL, "is_in_g1_reserved should ensure this" ); | |
| 2125 HeapRegion* hr = _g1h->heap_region_containing(obj); | |
| 2126 if (_g1h->is_obj_ill(obj, hr)) { | |
| 2127 if (verbose_high()) | |
| 2128 gclog_or_tty->print_cr("[global] "PTR_FORMAT" is not considered " | |
| 2129 "marked", (void*) obj); | |
| 2130 | |
| 2131 // we need to mark it first | |
| 2132 if (_nextMarkBitMap->parMark(objAddr)) { | |
| 2133 // No OrderAccess:store_load() is needed. It is implicit in the | |
| 2134 // CAS done in parMark(objAddr) above | |
| 2135 HeapWord* finger = _finger; | |
| 2136 if (objAddr < finger) { | |
| 2137 if (verbose_high()) | |
| 2138 gclog_or_tty->print_cr("[global] below the global finger " | |
| 2139 "("PTR_FORMAT"), pushing it", finger); | |
| 2140 if (!mark_stack_push(obj)) { | |
| 2141 if (verbose_low()) | |
| 2142 gclog_or_tty->print_cr("[global] global stack overflow during " | |
| 2143 "deal_with_reference"); | |
| 2144 } | |
| 2145 } | |
| 2146 } | |
| 2147 } | |
| 2148 } | |
| 2149 } | |
| 2150 | |
| 2151 void ConcurrentMark::drainAllSATBBuffers() { | |
| 2152 CMGlobalObjectClosure oc(this); | |
| 2153 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
| 2154 satb_mq_set.set_closure(&oc); | |
| 2155 | |
| 2156 while (satb_mq_set.apply_closure_to_completed_buffer()) { | |
| 2157 if (verbose_medium()) | |
| 2158 gclog_or_tty->print_cr("[global] processed an SATB buffer"); | |
| 2159 } | |
| 2160 | |
| 2161 // no need to check whether we should do this, as this is only | |
| 2162 // called during an evacuation pause | |
| 2163 satb_mq_set.iterate_closure_all_threads(); | |
| 2164 | |
| 2165 satb_mq_set.set_closure(NULL); | |
| 2166 guarantee( satb_mq_set.completed_buffers_num() == 0, "invariant" ); | |
| 2167 } | |
| 2168 | |
| 2169 void ConcurrentMark::markPrev(oop p) { | |
| 2170 // Note we are overriding the read-only view of the prev map here, via | |
| 2171 // the cast. | |
| 2172 ((CMBitMap*)_prevMarkBitMap)->mark((HeapWord*)p); | |
| 2173 } | |
| 2174 | |
| 2175 void ConcurrentMark::clear(oop p) { | |
| 2176 assert(p != NULL && p->is_oop(), "expected an oop"); | |
| 2177 HeapWord* addr = (HeapWord*)p; | |
| 2178 assert(addr >= _nextMarkBitMap->startWord() || | |
| 2179 addr < _nextMarkBitMap->endWord(), "in a region"); | |
| 2180 | |
| 2181 _nextMarkBitMap->clear(addr); | |
| 2182 } | |
| 2183 | |
| 2184 void ConcurrentMark::clearRangeBothMaps(MemRegion mr) { | |
| 2185 // Note we are overriding the read-only view of the prev map here, via | |
| 2186 // the cast. | |
| 2187 ((CMBitMap*)_prevMarkBitMap)->clearRange(mr); | |
| 2188 _nextMarkBitMap->clearRange(mr); | |
| 2189 } | |
| 2190 | |
| 2191 HeapRegion* | |
| 2192 ConcurrentMark::claim_region(int task_num) { | |
| 2193 // "checkpoint" the finger | |
| 2194 HeapWord* finger = _finger; | |
| 2195 | |
| 2196 // _heap_end will not change underneath our feet; it only changes at | |
| 2197 // yield points. | |
| 2198 while (finger < _heap_end) { | |
| 2199 tmp_guarantee_CM( _g1h->is_in_g1_reserved(finger), "invariant" ); | |
| 2200 | |
| 2201 // is the gap between reading the finger and doing the CAS too long? | |
| 2202 | |
| 2203 HeapRegion* curr_region = _g1h->heap_region_containing(finger); | |
| 2204 HeapWord* bottom = curr_region->bottom(); | |
| 2205 HeapWord* end = curr_region->end(); | |
| 2206 HeapWord* limit = curr_region->next_top_at_mark_start(); | |
| 2207 | |
| 2208 if (verbose_low()) | |
| 2209 gclog_or_tty->print_cr("[%d] curr_region = "PTR_FORMAT" " | |
| 2210 "["PTR_FORMAT", "PTR_FORMAT"), " | |
| 2211 "limit = "PTR_FORMAT, | |
| 2212 task_num, curr_region, bottom, end, limit); | |
| 2213 | |
| 2214 HeapWord* res = | |
| 2215 (HeapWord*) Atomic::cmpxchg_ptr(end, &_finger, finger); | |
| 2216 if (res == finger) { | |
| 2217 // we succeeded | |
| 2218 | |
| 2219 // notice that _finger == end cannot be guaranteed here since, | |
| 2220 // someone else might have moved the finger even further | |
| 2221 guarantee( _finger >= end, "the finger should have moved forward" ); | |
| 2222 | |
| 2223 if (verbose_low()) | |
| 2224 gclog_or_tty->print_cr("[%d] we were successful with region = " | |
| 2225 PTR_FORMAT, task_num, curr_region); | |
| 2226 | |
| 2227 if (limit > bottom) { | |
| 2228 if (verbose_low()) | |
| 2229 gclog_or_tty->print_cr("[%d] region "PTR_FORMAT" is not empty, " | |
| 2230 "returning it ", task_num, curr_region); | |
| 2231 return curr_region; | |
| 2232 } else { | |
| 2233 tmp_guarantee_CM( limit == bottom, | |
| 2234 "the region limit should be at bottom" ); | |
| 2235 if (verbose_low()) | |
| 2236 gclog_or_tty->print_cr("[%d] region "PTR_FORMAT" is empty, " | |
| 2237 "returning NULL", task_num, curr_region); | |
| 2238 // we return NULL and the caller should try calling | |
| 2239 // claim_region() again. | |
| 2240 return NULL; | |
| 2241 } | |
| 2242 } else { | |
| 2243 guarantee( _finger > finger, "the finger should have moved forward" ); | |
| 2244 if (verbose_low()) | |
| 2245 gclog_or_tty->print_cr("[%d] somebody else moved the finger, " | |
| 2246 "global finger = "PTR_FORMAT", " | |
| 2247 "our finger = "PTR_FORMAT, | |
| 2248 task_num, _finger, finger); | |
| 2249 | |
| 2250 // read it again | |
| 2251 finger = _finger; | |
| 2252 } | |
| 2253 } | |
| 2254 | |
| 2255 return NULL; | |
| 2256 } | |
| 2257 | |
| 2258 void ConcurrentMark::oops_do(OopClosure* cl) { | |
| 2259 if (_markStack.size() > 0 && verbose_low()) | |
| 2260 gclog_or_tty->print_cr("[global] scanning the global marking stack, " | |
| 2261 "size = %d", _markStack.size()); | |
| 2262 // we first iterate over the contents of the mark stack... | |
| 2263 _markStack.oops_do(cl); | |
| 2264 | |
| 2265 for (int i = 0; i < (int)_max_task_num; ++i) { | |
| 2266 OopTaskQueue* queue = _task_queues->queue((int)i); | |
| 2267 | |
| 2268 if (queue->size() > 0 && verbose_low()) | |
| 2269 gclog_or_tty->print_cr("[global] scanning task queue of task %d, " | |
| 2270 "size = %d", i, queue->size()); | |
| 2271 | |
| 2272 // ...then over the contents of the all the task queues. | |
| 2273 queue->oops_do(cl); | |
| 2274 } | |
| 2275 | |
| 2276 // finally, invalidate any entries that in the region stack that | |
| 2277 // point into the collection set | |
| 2278 if (_regionStack.invalidate_entries_into_cset()) { | |
| 2279 // otherwise, any gray objects copied during the evacuation pause | |
| 2280 // might not be visited. | |
| 2281 guarantee( _should_gray_objects, "invariant" ); | |
| 2282 } | |
| 2283 } | |
| 2284 | |
| 2285 void ConcurrentMark::clear_marking_state() { | |
| 2286 _markStack.setEmpty(); | |
| 2287 _markStack.clear_overflow(); | |
| 2288 _regionStack.setEmpty(); | |
| 2289 _regionStack.clear_overflow(); | |
| 2290 clear_has_overflown(); | |
| 2291 _finger = _heap_start; | |
| 2292 | |
| 2293 for (int i = 0; i < (int)_max_task_num; ++i) { | |
| 2294 OopTaskQueue* queue = _task_queues->queue(i); | |
| 2295 queue->set_empty(); | |
| 2296 } | |
| 2297 } | |
| 2298 | |
| 2299 void ConcurrentMark::print_stats() { | |
| 2300 if (verbose_stats()) { | |
| 2301 gclog_or_tty->print_cr("---------------------------------------------------------------------"); | |
| 2302 for (size_t i = 0; i < _active_tasks; ++i) { | |
| 2303 _tasks[i]->print_stats(); | |
| 2304 gclog_or_tty->print_cr("---------------------------------------------------------------------"); | |
| 2305 } | |
| 2306 } | |
| 2307 } | |
| 2308 | |
| 2309 class CSMarkOopClosure: public OopClosure { | |
| 2310 friend class CSMarkBitMapClosure; | |
| 2311 | |
| 2312 G1CollectedHeap* _g1h; | |
| 2313 CMBitMap* _bm; | |
| 2314 ConcurrentMark* _cm; | |
| 2315 oop* _ms; | |
| 2316 jint* _array_ind_stack; | |
| 2317 int _ms_size; | |
| 2318 int _ms_ind; | |
| 2319 int _array_increment; | |
| 2320 | |
| 2321 bool push(oop obj, int arr_ind = 0) { | |
| 2322 if (_ms_ind == _ms_size) { | |
| 2323 gclog_or_tty->print_cr("Mark stack is full."); | |
| 2324 return false; | |
| 2325 } | |
| 2326 _ms[_ms_ind] = obj; | |
| 2327 if (obj->is_objArray()) _array_ind_stack[_ms_ind] = arr_ind; | |
| 2328 _ms_ind++; | |
| 2329 return true; | |
| 2330 } | |
| 2331 | |
| 2332 oop pop() { | |
| 2333 if (_ms_ind == 0) return NULL; | |
| 2334 else { | |
| 2335 _ms_ind--; | |
| 2336 return _ms[_ms_ind]; | |
| 2337 } | |
| 2338 } | |
| 2339 | |
| 2340 bool drain() { | |
| 2341 while (_ms_ind > 0) { | |
| 2342 oop obj = pop(); | |
| 2343 assert(obj != NULL, "Since index was non-zero."); | |
| 2344 if (obj->is_objArray()) { | |
| 2345 jint arr_ind = _array_ind_stack[_ms_ind]; | |
| 2346 objArrayOop aobj = objArrayOop(obj); | |
| 2347 jint len = aobj->length(); | |
| 2348 jint next_arr_ind = arr_ind + _array_increment; | |
| 2349 if (next_arr_ind < len) { | |
| 2350 push(obj, next_arr_ind); | |
| 2351 } | |
| 2352 // Now process this portion of this one. | |
| 2353 int lim = MIN2(next_arr_ind, len); | |
| 2354 assert(!UseCompressedOops, "This needs to be fixed"); | |
| 2355 for (int j = arr_ind; j < lim; j++) { | |
| 2356 do_oop(aobj->obj_at_addr<oop>(j)); | |
| 2357 } | |
| 2358 | |
| 2359 } else { | |
| 2360 obj->oop_iterate(this); | |
| 2361 } | |
| 2362 if (abort()) return false; | |
| 2363 } | |
| 2364 return true; | |
| 2365 } | |
| 2366 | |
| 2367 public: | |
| 2368 CSMarkOopClosure(ConcurrentMark* cm, int ms_size) : | |
| 2369 _g1h(G1CollectedHeap::heap()), | |
| 2370 _cm(cm), | |
| 2371 _bm(cm->nextMarkBitMap()), | |
| 2372 _ms_size(ms_size), _ms_ind(0), | |
| 2373 _ms(NEW_C_HEAP_ARRAY(oop, ms_size)), | |
| 2374 _array_ind_stack(NEW_C_HEAP_ARRAY(jint, ms_size)), | |
| 2375 _array_increment(MAX2(ms_size/8, 16)) | |
| 2376 {} | |
| 2377 | |
| 2378 ~CSMarkOopClosure() { | |
| 2379 FREE_C_HEAP_ARRAY(oop, _ms); | |
| 2380 FREE_C_HEAP_ARRAY(jint, _array_ind_stack); | |
| 2381 } | |
| 2382 | |
| 2383 void do_oop(narrowOop* p) { | |
| 2384 guarantee(false, "NYI"); | |
| 2385 } | |
| 2386 | |
| 2387 void do_oop(oop* p) { | |
| 2388 oop obj = *p; | |
| 2389 if (obj == NULL) return; | |
| 2390 if (obj->is_forwarded()) { | |
| 2391 // If the object has already been forwarded, we have to make sure | |
| 2392 // that it's marked. So follow the forwarding pointer. Note that | |
| 2393 // this does the right thing for self-forwarding pointers in the | |
| 2394 // evacuation failure case. | |
| 2395 obj = obj->forwardee(); | |
| 2396 } | |
| 2397 HeapRegion* hr = _g1h->heap_region_containing(obj); | |
| 2398 if (hr != NULL) { | |
| 2399 if (hr->in_collection_set()) { | |
| 2400 if (_g1h->is_obj_ill(obj)) { | |
| 2401 _bm->mark((HeapWord*)obj); | |
| 2402 if (!push(obj)) { | |
| 2403 gclog_or_tty->print_cr("Setting abort in CSMarkOopClosure because push failed."); | |
| 2404 set_abort(); | |
| 2405 } | |
| 2406 } | |
| 2407 } else { | |
| 2408 // Outside the collection set; we need to gray it | |
| 2409 _cm->deal_with_reference(obj); | |
| 2410 } | |
| 2411 } | |
| 2412 } | |
| 2413 }; | |
| 2414 | |
| 2415 class CSMarkBitMapClosure: public BitMapClosure { | |
| 2416 G1CollectedHeap* _g1h; | |
| 2417 CMBitMap* _bitMap; | |
| 2418 ConcurrentMark* _cm; | |
| 2419 CSMarkOopClosure _oop_cl; | |
| 2420 public: | |
| 2421 CSMarkBitMapClosure(ConcurrentMark* cm, int ms_size) : | |
| 2422 _g1h(G1CollectedHeap::heap()), | |
| 2423 _bitMap(cm->nextMarkBitMap()), | |
| 2424 _oop_cl(cm, ms_size) | |
| 2425 {} | |
| 2426 | |
| 2427 ~CSMarkBitMapClosure() {} | |
| 2428 | |
| 2429 bool do_bit(size_t offset) { | |
| 2430 // convert offset into a HeapWord* | |
| 2431 HeapWord* addr = _bitMap->offsetToHeapWord(offset); | |
| 2432 assert(_bitMap->endWord() && addr < _bitMap->endWord(), | |
| 2433 "address out of range"); | |
| 2434 assert(_bitMap->isMarked(addr), "tautology"); | |
| 2435 oop obj = oop(addr); | |
| 2436 if (!obj->is_forwarded()) { | |
| 2437 if (!_oop_cl.push(obj)) return false; | |
| 2438 if (!_oop_cl.drain()) return false; | |
| 2439 } | |
| 2440 // Otherwise... | |
| 2441 return true; | |
| 2442 } | |
| 2443 }; | |
| 2444 | |
| 2445 | |
| 2446 class CompleteMarkingInCSHRClosure: public HeapRegionClosure { | |
| 2447 CMBitMap* _bm; | |
| 2448 CSMarkBitMapClosure _bit_cl; | |
| 2449 enum SomePrivateConstants { | |
| 2450 MSSize = 1000 | |
| 2451 }; | |
| 2452 bool _completed; | |
| 2453 public: | |
| 2454 CompleteMarkingInCSHRClosure(ConcurrentMark* cm) : | |
| 2455 _bm(cm->nextMarkBitMap()), | |
| 2456 _bit_cl(cm, MSSize), | |
| 2457 _completed(true) | |
| 2458 {} | |
| 2459 | |
| 2460 ~CompleteMarkingInCSHRClosure() {} | |
| 2461 | |
| 2462 bool doHeapRegion(HeapRegion* r) { | |
| 2463 if (!r->evacuation_failed()) { | |
| 2464 MemRegion mr = MemRegion(r->bottom(), r->next_top_at_mark_start()); | |
| 2465 if (!mr.is_empty()) { | |
| 2466 if (!_bm->iterate(&_bit_cl, mr)) { | |
| 2467 _completed = false; | |
| 2468 return true; | |
| 2469 } | |
| 2470 } | |
| 2471 } | |
| 2472 return false; | |
| 2473 } | |
| 2474 | |
| 2475 bool completed() { return _completed; } | |
| 2476 }; | |
| 2477 | |
| 2478 class ClearMarksInHRClosure: public HeapRegionClosure { | |
| 2479 CMBitMap* _bm; | |
| 2480 public: | |
| 2481 ClearMarksInHRClosure(CMBitMap* bm): _bm(bm) { } | |
| 2482 | |
| 2483 bool doHeapRegion(HeapRegion* r) { | |
| 2484 if (!r->used_region().is_empty() && !r->evacuation_failed()) { | |
| 2485 MemRegion usedMR = r->used_region(); | |
| 2486 _bm->clearRange(r->used_region()); | |
| 2487 } | |
| 2488 return false; | |
| 2489 } | |
| 2490 }; | |
| 2491 | |
| 2492 void ConcurrentMark::complete_marking_in_collection_set() { | |
| 2493 G1CollectedHeap* g1h = G1CollectedHeap::heap(); | |
| 2494 | |
| 2495 if (!g1h->mark_in_progress()) { | |
| 2496 g1h->g1_policy()->record_mark_closure_time(0.0); | |
| 2497 return; | |
| 2498 } | |
| 2499 | |
| 2500 int i = 1; | |
| 2501 double start = os::elapsedTime(); | |
| 2502 while (true) { | |
| 2503 i++; | |
| 2504 CompleteMarkingInCSHRClosure cmplt(this); | |
| 2505 g1h->collection_set_iterate(&cmplt); | |
| 2506 if (cmplt.completed()) break; | |
| 2507 } | |
| 2508 double end_time = os::elapsedTime(); | |
| 2509 double elapsed_time_ms = (end_time - start) * 1000.0; | |
| 2510 g1h->g1_policy()->record_mark_closure_time(elapsed_time_ms); | |
| 2511 if (PrintGCDetails) { | |
| 2512 gclog_or_tty->print_cr("Mark closure took %5.2f ms.", elapsed_time_ms); | |
| 2513 } | |
| 2514 | |
| 2515 ClearMarksInHRClosure clr(nextMarkBitMap()); | |
| 2516 g1h->collection_set_iterate(&clr); | |
| 2517 } | |
| 2518 | |
| 2519 // The next two methods deal with the following optimisation. Some | |
| 2520 // objects are gray by being marked and located above the finger. If | |
| 2521 // they are copied, during an evacuation pause, below the finger then | |
| 2522 // the need to be pushed on the stack. The observation is that, if | |
| 2523 // there are no regions in the collection set located above the | |
| 2524 // finger, then the above cannot happen, hence we do not need to | |
| 2525 // explicitly gray any objects when copying them to below the | |
| 2526 // finger. The global stack will be scanned to ensure that, if it | |
| 2527 // points to objects being copied, it will update their | |
| 2528 // location. There is a tricky situation with the gray objects in | |
| 2529 // region stack that are being coped, however. See the comment in | |
| 2530 // newCSet(). | |
| 2531 | |
| 2532 void ConcurrentMark::newCSet() { | |
| 2533 if (!concurrent_marking_in_progress()) | |
| 2534 // nothing to do if marking is not in progress | |
| 2535 return; | |
| 2536 | |
| 2537 // find what the lowest finger is among the global and local fingers | |
| 2538 _min_finger = _finger; | |
| 2539 for (int i = 0; i < (int)_max_task_num; ++i) { | |
| 2540 CMTask* task = _tasks[i]; | |
| 2541 HeapWord* task_finger = task->finger(); | |
| 2542 if (task_finger != NULL && task_finger < _min_finger) | |
| 2543 _min_finger = task_finger; | |
| 2544 } | |
| 2545 | |
| 2546 _should_gray_objects = false; | |
| 2547 | |
| 2548 // This fixes a very subtle and fustrating bug. It might be the case | |
| 2549 // that, during en evacuation pause, heap regions that contain | |
| 2550 // objects that are gray (by being in regions contained in the | |
| 2551 // region stack) are included in the collection set. Since such gray | |
| 2552 // objects will be moved, and because it's not easy to redirect | |
| 2553 // region stack entries to point to a new location (because objects | |
| 2554 // in one region might be scattered to multiple regions after they | |
| 2555 // are copied), one option is to ensure that all marked objects | |
| 2556 // copied during a pause are pushed on the stack. Notice, however, | |
| 2557 // that this problem can only happen when the region stack is not | |
| 2558 // empty during an evacuation pause. So, we make the fix a bit less | |
| 2559 // conservative and ensure that regions are pushed on the stack, | |
| 2560 // irrespective whether all collection set regions are below the | |
| 2561 // finger, if the region stack is not empty. This is expected to be | |
| 2562 // a rare case, so I don't think it's necessary to be smarted about it. | |
| 2563 if (!region_stack_empty()) | |
| 2564 _should_gray_objects = true; | |
| 2565 } | |
| 2566 | |
| 2567 void ConcurrentMark::registerCSetRegion(HeapRegion* hr) { | |
| 2568 if (!concurrent_marking_in_progress()) | |
| 2569 return; | |
| 2570 | |
| 2571 HeapWord* region_end = hr->end(); | |
| 2572 if (region_end > _min_finger) | |
| 2573 _should_gray_objects = true; | |
| 2574 } | |
| 2575 | |
| 2576 void ConcurrentMark::disable_co_trackers() { | |
| 2577 if (has_aborted()) { | |
| 2578 if (_cleanup_co_tracker.enabled()) | |
| 2579 _cleanup_co_tracker.disable(); | |
| 2580 for (int i = 0; i < (int)_max_task_num; ++i) { | |
| 2581 CMTask* task = _tasks[i]; | |
| 2582 if (task->co_tracker_enabled()) | |
| 2583 task->disable_co_tracker(); | |
| 2584 } | |
| 2585 } else { | |
| 2586 guarantee( !_cleanup_co_tracker.enabled(), "invariant" ); | |
| 2587 for (int i = 0; i < (int)_max_task_num; ++i) { | |
| 2588 CMTask* task = _tasks[i]; | |
| 2589 guarantee( !task->co_tracker_enabled(), "invariant" ); | |
| 2590 } | |
| 2591 } | |
| 2592 } | |
| 2593 | |
| 2594 // abandon current marking iteration due to a Full GC | |
| 2595 void ConcurrentMark::abort() { | |
| 2596 // If we're not marking, nothing to do. | |
| 2597 if (!G1ConcMark) return; | |
| 2598 | |
| 2599 // Clear all marks to force marking thread to do nothing | |
| 2600 _nextMarkBitMap->clearAll(); | |
| 2601 // Empty mark stack | |
| 2602 clear_marking_state(); | |
| 2603 for (int i = 0; i < (int)_max_task_num; ++i) | |
| 2604 _tasks[i]->clear_region_fields(); | |
| 2605 _has_aborted = true; | |
| 2606 | |
| 2607 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
| 2608 satb_mq_set.abandon_partial_marking(); | |
| 2609 satb_mq_set.set_active_all_threads(false); | |
| 2610 } | |
| 2611 | |
| 2612 static void print_ms_time_info(const char* prefix, const char* name, | |
| 2613 NumberSeq& ns) { | |
| 2614 gclog_or_tty->print_cr("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).", | |
| 2615 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg()); | |
| 2616 if (ns.num() > 0) { | |
| 2617 gclog_or_tty->print_cr("%s [std. dev = %8.2f ms, max = %8.2f ms]", | |
| 2618 prefix, ns.sd(), ns.maximum()); | |
| 2619 } | |
| 2620 } | |
| 2621 | |
| 2622 void ConcurrentMark::print_summary_info() { | |
| 2623 gclog_or_tty->print_cr(" Concurrent marking:"); | |
| 2624 print_ms_time_info(" ", "init marks", _init_times); | |
| 2625 print_ms_time_info(" ", "remarks", _remark_times); | |
| 2626 { | |
| 2627 print_ms_time_info(" ", "final marks", _remark_mark_times); | |
| 2628 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times); | |
| 2629 | |
| 2630 } | |
| 2631 print_ms_time_info(" ", "cleanups", _cleanup_times); | |
| 2632 gclog_or_tty->print_cr(" Final counting total time = %8.2f s (avg = %8.2f ms).", | |
| 2633 _total_counting_time, | |
| 2634 (_cleanup_times.num() > 0 ? _total_counting_time * 1000.0 / | |
| 2635 (double)_cleanup_times.num() | |
| 2636 : 0.0)); | |
| 2637 if (G1ScrubRemSets) { | |
| 2638 gclog_or_tty->print_cr(" RS scrub total time = %8.2f s (avg = %8.2f ms).", | |
| 2639 _total_rs_scrub_time, | |
| 2640 (_cleanup_times.num() > 0 ? _total_rs_scrub_time * 1000.0 / | |
| 2641 (double)_cleanup_times.num() | |
| 2642 : 0.0)); | |
| 2643 } | |
| 2644 gclog_or_tty->print_cr(" Total stop_world time = %8.2f s.", | |
| 2645 (_init_times.sum() + _remark_times.sum() + | |
| 2646 _cleanup_times.sum())/1000.0); | |
| 2647 gclog_or_tty->print_cr(" Total concurrent time = %8.2f s " | |
| 2648 "(%8.2f s marking, %8.2f s counting).", | |
| 2649 cmThread()->vtime_accum(), | |
| 2650 cmThread()->vtime_mark_accum(), | |
| 2651 cmThread()->vtime_count_accum()); | |
| 2652 } | |
| 2653 | |
| 2654 // Closures | |
| 2655 // XXX: there seems to be a lot of code duplication here; | |
| 2656 // should refactor and consolidate the shared code. | |
| 2657 | |
| 2658 // This closure is used to mark refs into the CMS generation in | |
| 2659 // the CMS bit map. Called at the first checkpoint. | |
| 2660 | |
| 2661 // We take a break if someone is trying to stop the world. | |
| 2662 bool ConcurrentMark::do_yield_check(int worker_i) { | |
| 2663 if (should_yield()) { | |
| 2664 if (worker_i == 0) | |
| 2665 _g1h->g1_policy()->record_concurrent_pause(); | |
| 2666 cmThread()->yield(); | |
| 2667 if (worker_i == 0) | |
| 2668 _g1h->g1_policy()->record_concurrent_pause_end(); | |
| 2669 return true; | |
| 2670 } else { | |
| 2671 return false; | |
| 2672 } | |
| 2673 } | |
| 2674 | |
| 2675 bool ConcurrentMark::should_yield() { | |
| 2676 return cmThread()->should_yield(); | |
| 2677 } | |
| 2678 | |
| 2679 bool ConcurrentMark::containing_card_is_marked(void* p) { | |
| 2680 size_t offset = pointer_delta(p, _g1h->reserved_region().start(), 1); | |
| 2681 return _card_bm.at(offset >> CardTableModRefBS::card_shift); | |
| 2682 } | |
| 2683 | |
| 2684 bool ConcurrentMark::containing_cards_are_marked(void* start, | |
| 2685 void* last) { | |
| 2686 return | |
| 2687 containing_card_is_marked(start) && | |
| 2688 containing_card_is_marked(last); | |
| 2689 } | |
| 2690 | |
| 2691 #ifndef PRODUCT | |
| 2692 // for debugging purposes | |
| 2693 void ConcurrentMark::print_finger() { | |
| 2694 gclog_or_tty->print_cr("heap ["PTR_FORMAT", "PTR_FORMAT"), global finger = "PTR_FORMAT, | |
| 2695 _heap_start, _heap_end, _finger); | |
| 2696 for (int i = 0; i < (int) _max_task_num; ++i) { | |
| 2697 gclog_or_tty->print(" %d: "PTR_FORMAT, i, _tasks[i]->finger()); | |
| 2698 } | |
| 2699 gclog_or_tty->print_cr(""); | |
| 2700 } | |
| 2701 #endif | |
| 2702 | |
| 2703 // Closure for iteration over bitmaps | |
| 2704 class CMBitMapClosure : public BitMapClosure { | |
| 2705 private: | |
| 2706 // the bitmap that is being iterated over | |
| 2707 CMBitMap* _nextMarkBitMap; | |
| 2708 ConcurrentMark* _cm; | |
| 2709 CMTask* _task; | |
| 2710 // true if we're scanning a heap region claimed by the task (so that | |
| 2711 // we move the finger along), false if we're not, i.e. currently when | |
| 2712 // scanning a heap region popped from the region stack (so that we | |
| 2713 // do not move the task finger along; it'd be a mistake if we did so). | |
| 2714 bool _scanning_heap_region; | |
| 2715 | |
| 2716 public: | |
| 2717 CMBitMapClosure(CMTask *task, | |
| 2718 ConcurrentMark* cm, | |
| 2719 CMBitMap* nextMarkBitMap) | |
| 2720 : _task(task), _cm(cm), _nextMarkBitMap(nextMarkBitMap) { } | |
| 2721 | |
| 2722 void set_scanning_heap_region(bool scanning_heap_region) { | |
| 2723 _scanning_heap_region = scanning_heap_region; | |
| 2724 } | |
| 2725 | |
| 2726 bool do_bit(size_t offset) { | |
| 2727 HeapWord* addr = _nextMarkBitMap->offsetToHeapWord(offset); | |
| 2728 tmp_guarantee_CM( _nextMarkBitMap->isMarked(addr), "invariant" ); | |
| 2729 tmp_guarantee_CM( addr < _cm->finger(), "invariant" ); | |
| 2730 | |
| 2731 if (_scanning_heap_region) { | |
| 2732 statsOnly( _task->increase_objs_found_on_bitmap() ); | |
| 2733 tmp_guarantee_CM( addr >= _task->finger(), "invariant" ); | |
| 2734 // We move that task's local finger along. | |
| 2735 _task->move_finger_to(addr); | |
| 2736 } else { | |
| 2737 // We move the task's region finger along. | |
| 2738 _task->move_region_finger_to(addr); | |
| 2739 } | |
| 2740 | |
| 2741 _task->scan_object(oop(addr)); | |
| 2742 // we only partially drain the local queue and global stack | |
| 2743 _task->drain_local_queue(true); | |
| 2744 _task->drain_global_stack(true); | |
| 2745 | |
| 2746 // if the has_aborted flag has been raised, we need to bail out of | |
| 2747 // the iteration | |
| 2748 return !_task->has_aborted(); | |
| 2749 } | |
| 2750 }; | |
| 2751 | |
| 2752 // Closure for iterating over objects, currently only used for | |
| 2753 // processing SATB buffers. | |
| 2754 class CMObjectClosure : public ObjectClosure { | |
| 2755 private: | |
| 2756 CMTask* _task; | |
| 2757 | |
| 2758 public: | |
| 2759 void do_object(oop obj) { | |
| 2760 _task->deal_with_reference(obj); | |
| 2761 } | |
| 2762 | |
| 2763 CMObjectClosure(CMTask* task) : _task(task) { } | |
| 2764 }; | |
| 2765 | |
| 2766 // Closure for iterating over object fields | |
| 2767 class CMOopClosure : public OopClosure { | |
| 2768 private: | |
| 2769 G1CollectedHeap* _g1h; | |
| 2770 ConcurrentMark* _cm; | |
| 2771 CMTask* _task; | |
| 2772 | |
| 2773 public: | |
| 2774 void do_oop(narrowOop* p) { | |
| 2775 guarantee(false, "NYI"); | |
| 2776 } | |
| 2777 | |
| 2778 void do_oop(oop* p) { | |
| 2779 tmp_guarantee_CM( _g1h->is_in_g1_reserved((HeapWord*) p), "invariant" ); | |
| 2780 | |
| 2781 oop obj = *p; | |
| 2782 if (_cm->verbose_high()) | |
| 2783 gclog_or_tty->print_cr("[%d] we're looking at location " | |
| 2784 "*"PTR_FORMAT" = "PTR_FORMAT, | |
| 2785 _task->task_id(), p, (void*) obj); | |
| 2786 _task->deal_with_reference(obj); | |
| 2787 } | |
| 2788 | |
| 2789 CMOopClosure(G1CollectedHeap* g1h, | |
| 2790 ConcurrentMark* cm, | |
| 2791 CMTask* task) | |
| 2792 : _g1h(g1h), _cm(cm), _task(task) { } | |
| 2793 }; | |
| 2794 | |
| 2795 void CMTask::setup_for_region(HeapRegion* hr) { | |
| 2796 tmp_guarantee_CM( hr != NULL && !hr->continuesHumongous(), | |
| 2797 "claim_region() should have filtered out continues humongous regions" ); | |
| 2798 | |
| 2799 if (_cm->verbose_low()) | |
| 2800 gclog_or_tty->print_cr("[%d] setting up for region "PTR_FORMAT, | |
| 2801 _task_id, hr); | |
| 2802 | |
| 2803 _curr_region = hr; | |
| 2804 _finger = hr->bottom(); | |
| 2805 update_region_limit(); | |
| 2806 } | |
| 2807 | |
| 2808 void CMTask::update_region_limit() { | |
| 2809 HeapRegion* hr = _curr_region; | |
| 2810 HeapWord* bottom = hr->bottom(); | |
| 2811 HeapWord* limit = hr->next_top_at_mark_start(); | |
| 2812 | |
| 2813 if (limit == bottom) { | |
| 2814 if (_cm->verbose_low()) | |
| 2815 gclog_or_tty->print_cr("[%d] found an empty region " | |
| 2816 "["PTR_FORMAT", "PTR_FORMAT")", | |
| 2817 _task_id, bottom, limit); | |
| 2818 // The region was collected underneath our feet. | |
| 2819 // We set the finger to bottom to ensure that the bitmap | |
| 2820 // iteration that will follow this will not do anything. | |
| 2821 // (this is not a condition that holds when we set the region up, | |
| 2822 // as the region is not supposed to be empty in the first place) | |
| 2823 _finger = bottom; | |
| 2824 } else if (limit >= _region_limit) { | |
| 2825 tmp_guarantee_CM( limit >= _finger, "peace of mind" ); | |
| 2826 } else { | |
| 2827 tmp_guarantee_CM( limit < _region_limit, "only way to get here" ); | |
| 2828 // This can happen under some pretty unusual circumstances. An | |
| 2829 // evacuation pause empties the region underneath our feet (NTAMS | |
| 2830 // at bottom). We then do some allocation in the region (NTAMS | |
| 2831 // stays at bottom), followed by the region being used as a GC | |
| 2832 // alloc region (NTAMS will move to top() and the objects | |
| 2833 // originally below it will be grayed). All objects now marked in | |
| 2834 // the region are explicitly grayed, if below the global finger, | |
| 2835 // and we do not need in fact to scan anything else. So, we simply | |
| 2836 // set _finger to be limit to ensure that the bitmap iteration | |
| 2837 // doesn't do anything. | |
| 2838 _finger = limit; | |
| 2839 } | |
| 2840 | |
| 2841 _region_limit = limit; | |
| 2842 } | |
| 2843 | |
| 2844 void CMTask::giveup_current_region() { | |
| 2845 tmp_guarantee_CM( _curr_region != NULL, "invariant" ); | |
| 2846 if (_cm->verbose_low()) | |
| 2847 gclog_or_tty->print_cr("[%d] giving up region "PTR_FORMAT, | |
| 2848 _task_id, _curr_region); | |
| 2849 clear_region_fields(); | |
| 2850 } | |
| 2851 | |
| 2852 void CMTask::clear_region_fields() { | |
| 2853 // Values for these three fields that indicate that we're not | |
| 2854 // holding on to a region. | |
| 2855 _curr_region = NULL; | |
| 2856 _finger = NULL; | |
| 2857 _region_limit = NULL; | |
| 2858 | |
| 2859 _region_finger = NULL; | |
| 2860 } | |
| 2861 | |
| 2862 void CMTask::reset(CMBitMap* nextMarkBitMap) { | |
| 2863 guarantee( nextMarkBitMap != NULL, "invariant" ); | |
| 2864 | |
| 2865 if (_cm->verbose_low()) | |
| 2866 gclog_or_tty->print_cr("[%d] resetting", _task_id); | |
| 2867 | |
| 2868 _nextMarkBitMap = nextMarkBitMap; | |
| 2869 clear_region_fields(); | |
| 2870 | |
| 2871 _calls = 0; | |
| 2872 _elapsed_time_ms = 0.0; | |
| 2873 _termination_time_ms = 0.0; | |
| 2874 _termination_start_time_ms = 0.0; | |
| 2875 | |
| 2876 #if _MARKING_STATS_ | |
| 2877 _local_pushes = 0; | |
| 2878 _local_pops = 0; | |
| 2879 _local_max_size = 0; | |
| 2880 _objs_scanned = 0; | |
| 2881 _global_pushes = 0; | |
| 2882 _global_pops = 0; | |
| 2883 _global_max_size = 0; | |
| 2884 _global_transfers_to = 0; | |
| 2885 _global_transfers_from = 0; | |
| 2886 _region_stack_pops = 0; | |
| 2887 _regions_claimed = 0; | |
| 2888 _objs_found_on_bitmap = 0; | |
| 2889 _satb_buffers_processed = 0; | |
| 2890 _steal_attempts = 0; | |
| 2891 _steals = 0; | |
| 2892 _aborted = 0; | |
| 2893 _aborted_overflow = 0; | |
| 2894 _aborted_cm_aborted = 0; | |
| 2895 _aborted_yield = 0; | |
| 2896 _aborted_timed_out = 0; | |
| 2897 _aborted_satb = 0; | |
| 2898 _aborted_termination = 0; | |
| 2899 #endif // _MARKING_STATS_ | |
| 2900 } | |
| 2901 | |
| 2902 bool CMTask::should_exit_termination() { | |
| 2903 regular_clock_call(); | |
| 2904 // This is called when we are in the termination protocol. We should | |
| 2905 // quit if, for some reason, this task wants to abort or the global | |
| 2906 // stack is not empty (this means that we can get work from it). | |
| 2907 return !_cm->mark_stack_empty() || has_aborted(); | |
| 2908 } | |
| 2909 | |
| 2910 // This determines whether the method below will check both the local | |
| 2911 // and global fingers when determining whether to push on the stack a | |
| 2912 // gray object (value 1) or whether it will only check the global one | |
| 2913 // (value 0). The tradeoffs are that the former will be a bit more | |
| 2914 // accurate and possibly push less on the stack, but it might also be | |
| 2915 // a little bit slower. | |
| 2916 | |
| 2917 #define _CHECK_BOTH_FINGERS_ 1 | |
| 2918 | |
| 2919 void CMTask::deal_with_reference(oop obj) { | |
| 2920 if (_cm->verbose_high()) | |
| 2921 gclog_or_tty->print_cr("[%d] we're dealing with reference = "PTR_FORMAT, | |
| 2922 _task_id, (void*) obj); | |
| 2923 | |
| 2924 ++_refs_reached; | |
| 2925 | |
| 2926 HeapWord* objAddr = (HeapWord*) obj; | |
| 2927 if (_g1h->is_in_g1_reserved(objAddr)) { | |
| 2928 tmp_guarantee_CM( obj != NULL, "is_in_g1_reserved should ensure this" ); | |
| 2929 HeapRegion* hr = _g1h->heap_region_containing(obj); | |
| 2930 if (_g1h->is_obj_ill(obj, hr)) { | |
| 2931 if (_cm->verbose_high()) | |
| 2932 gclog_or_tty->print_cr("[%d] "PTR_FORMAT" is not considered marked", | |
| 2933 _task_id, (void*) obj); | |
| 2934 | |
| 2935 // we need to mark it first | |
| 2936 if (_nextMarkBitMap->parMark(objAddr)) { | |
| 2937 // No OrderAccess:store_load() is needed. It is implicit in the | |
| 2938 // CAS done in parMark(objAddr) above | |
| 2939 HeapWord* global_finger = _cm->finger(); | |
| 2940 | |
| 2941 #if _CHECK_BOTH_FINGERS_ | |
| 2942 // we will check both the local and global fingers | |
| 2943 | |
| 2944 if (_finger != NULL && objAddr < _finger) { | |
| 2945 if (_cm->verbose_high()) | |
| 2946 gclog_or_tty->print_cr("[%d] below the local finger ("PTR_FORMAT"), " | |
| 2947 "pushing it", _task_id, _finger); | |
| 2948 push(obj); | |
| 2949 } else if (_curr_region != NULL && objAddr < _region_limit) { | |
| 2950 // do nothing | |
| 2951 } else if (objAddr < global_finger) { | |
| 2952 // Notice that the global finger might be moving forward | |
| 2953 // concurrently. This is not a problem. In the worst case, we | |
| 2954 // mark the object while it is above the global finger and, by | |
| 2955 // the time we read the global finger, it has moved forward | |
| 2956 // passed this object. In this case, the object will probably | |
| 2957 // be visited when a task is scanning the region and will also | |
| 2958 // be pushed on the stack. So, some duplicate work, but no | |
| 2959 // correctness problems. | |
| 2960 | |
| 2961 if (_cm->verbose_high()) | |
| 2962 gclog_or_tty->print_cr("[%d] below the global finger " | |
| 2963 "("PTR_FORMAT"), pushing it", | |
| 2964 _task_id, global_finger); | |
| 2965 push(obj); | |
| 2966 } else { | |
| 2967 // do nothing | |
| 2968 } | |
| 2969 #else // _CHECK_BOTH_FINGERS_ | |
| 2970 // we will only check the global finger | |
| 2971 | |
| 2972 if (objAddr < global_finger) { | |
| 2973 // see long comment above | |
| 2974 | |
| 2975 if (_cm->verbose_high()) | |
| 2976 gclog_or_tty->print_cr("[%d] below the global finger " | |
| 2977 "("PTR_FORMAT"), pushing it", | |
| 2978 _task_id, global_finger); | |
| 2979 push(obj); | |
| 2980 } | |
| 2981 #endif // _CHECK_BOTH_FINGERS_ | |
| 2982 } | |
| 2983 } | |
| 2984 } | |
| 2985 } | |
| 2986 | |
| 2987 void CMTask::push(oop obj) { | |
| 2988 HeapWord* objAddr = (HeapWord*) obj; | |
| 2989 tmp_guarantee_CM( _g1h->is_in_g1_reserved(objAddr), "invariant" ); | |
| 2990 tmp_guarantee_CM( !_g1h->is_obj_ill(obj), "invariant" ); | |
| 2991 tmp_guarantee_CM( _nextMarkBitMap->isMarked(objAddr), "invariant" ); | |
| 2992 | |
| 2993 if (_cm->verbose_high()) | |
| 2994 gclog_or_tty->print_cr("[%d] pushing "PTR_FORMAT, _task_id, (void*) obj); | |
| 2995 | |
| 2996 if (!_task_queue->push(obj)) { | |
| 2997 // The local task queue looks full. We need to push some entries | |
| 2998 // to the global stack. | |
| 2999 | |
| 3000 if (_cm->verbose_medium()) | |
| 3001 gclog_or_tty->print_cr("[%d] task queue overflow, " | |
| 3002 "moving entries to the global stack", | |
| 3003 _task_id); | |
| 3004 move_entries_to_global_stack(); | |
| 3005 | |
| 3006 // this should succeed since, even if we overflow the global | |
| 3007 // stack, we should have definitely removed some entries from the | |
| 3008 // local queue. So, there must be space on it. | |
| 3009 bool success = _task_queue->push(obj); | |
| 3010 tmp_guarantee_CM( success, "invariant" ); | |
| 3011 } | |
| 3012 | |
| 3013 statsOnly( int tmp_size = _task_queue->size(); | |
| 3014 if (tmp_size > _local_max_size) | |
| 3015 _local_max_size = tmp_size; | |
| 3016 ++_local_pushes ); | |
| 3017 } | |
| 3018 | |
| 3019 void CMTask::reached_limit() { | |
| 3020 tmp_guarantee_CM( _words_scanned >= _words_scanned_limit || | |
| 3021 _refs_reached >= _refs_reached_limit , | |
| 3022 "shouldn't have been called otherwise" ); | |
| 3023 regular_clock_call(); | |
| 3024 } | |
| 3025 | |
| 3026 void CMTask::regular_clock_call() { | |
| 3027 if (has_aborted()) | |
| 3028 return; | |
| 3029 | |
| 3030 // First, we need to recalculate the words scanned and refs reached | |
| 3031 // limits for the next clock call. | |
| 3032 recalculate_limits(); | |
| 3033 | |
| 3034 // During the regular clock call we do the following | |
| 3035 | |
| 3036 // (1) If an overflow has been flagged, then we abort. | |
| 3037 if (_cm->has_overflown()) { | |
| 3038 set_has_aborted(); | |
| 3039 return; | |
| 3040 } | |
| 3041 | |
| 3042 // If we are not concurrent (i.e. we're doing remark) we don't need | |
| 3043 // to check anything else. The other steps are only needed during | |
| 3044 // the concurrent marking phase. | |
| 3045 if (!concurrent()) | |
| 3046 return; | |
| 3047 | |
| 3048 // (2) If marking has been aborted for Full GC, then we also abort. | |
| 3049 if (_cm->has_aborted()) { | |
| 3050 set_has_aborted(); | |
| 3051 statsOnly( ++_aborted_cm_aborted ); | |
| 3052 return; | |
| 3053 } | |
| 3054 | |
| 3055 double curr_time_ms = os::elapsedVTime() * 1000.0; | |
| 3056 | |
| 3057 // (3) If marking stats are enabled, then we update the step history. | |
| 3058 #if _MARKING_STATS_ | |
| 3059 if (_words_scanned >= _words_scanned_limit) | |
| 3060 ++_clock_due_to_scanning; | |
| 3061 if (_refs_reached >= _refs_reached_limit) | |
| 3062 ++_clock_due_to_marking; | |
| 3063 | |
| 3064 double last_interval_ms = curr_time_ms - _interval_start_time_ms; | |
| 3065 _interval_start_time_ms = curr_time_ms; | |
| 3066 _all_clock_intervals_ms.add(last_interval_ms); | |
| 3067 | |
| 3068 if (_cm->verbose_medium()) { | |
| 3069 gclog_or_tty->print_cr("[%d] regular clock, interval = %1.2lfms, " | |
| 3070 "scanned = %d%s, refs reached = %d%s", | |
| 3071 _task_id, last_interval_ms, | |
| 3072 _words_scanned, | |
| 3073 (_words_scanned >= _words_scanned_limit) ? " (*)" : "", | |
| 3074 _refs_reached, | |
| 3075 (_refs_reached >= _refs_reached_limit) ? " (*)" : ""); | |
| 3076 } | |
| 3077 #endif // _MARKING_STATS_ | |
| 3078 | |
| 3079 // (4) We check whether we should yield. If we have to, then we abort. | |
| 3080 if (_cm->should_yield()) { | |
| 3081 // We should yield. To do this we abort the task. The caller is | |
| 3082 // responsible for yielding. | |
| 3083 set_has_aborted(); | |
| 3084 statsOnly( ++_aborted_yield ); | |
| 3085 return; | |
| 3086 } | |
| 3087 | |
| 3088 // (5) We check whether we've reached our time quota. If we have, | |
| 3089 // then we abort. | |
| 3090 double elapsed_time_ms = curr_time_ms - _start_time_ms; | |
| 3091 if (elapsed_time_ms > _time_target_ms) { | |
| 3092 set_has_aborted(); | |
| 3093 _has_aborted_timed_out = true; | |
| 3094 statsOnly( ++_aborted_timed_out ); | |
| 3095 return; | |
| 3096 } | |
| 3097 | |
| 3098 // (6) Finally, we check whether there are enough completed STAB | |
| 3099 // buffers available for processing. If there are, we abort. | |
| 3100 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
| 3101 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) { | |
| 3102 if (_cm->verbose_low()) | |
| 3103 gclog_or_tty->print_cr("[%d] aborting to deal with pending SATB buffers", | |
| 3104 _task_id); | |
| 3105 // we do need to process SATB buffers, we'll abort and restart | |
| 3106 // the marking task to do so | |
| 3107 set_has_aborted(); | |
| 3108 statsOnly( ++_aborted_satb ); | |
| 3109 return; | |
| 3110 } | |
| 3111 } | |
| 3112 | |
| 3113 void CMTask::recalculate_limits() { | |
| 3114 _real_words_scanned_limit = _words_scanned + words_scanned_period; | |
| 3115 _words_scanned_limit = _real_words_scanned_limit; | |
| 3116 | |
| 3117 _real_refs_reached_limit = _refs_reached + refs_reached_period; | |
| 3118 _refs_reached_limit = _real_refs_reached_limit; | |
| 3119 } | |
| 3120 | |
| 3121 void CMTask::decrease_limits() { | |
| 3122 // This is called when we believe that we're going to do an infrequent | |
| 3123 // operation which will increase the per byte scanned cost (i.e. move | |
| 3124 // entries to/from the global stack). It basically tries to decrease the | |
| 3125 // scanning limit so that the clock is called earlier. | |
| 3126 | |
| 3127 if (_cm->verbose_medium()) | |
| 3128 gclog_or_tty->print_cr("[%d] decreasing limits", _task_id); | |
| 3129 | |
| 3130 _words_scanned_limit = _real_words_scanned_limit - | |
| 3131 3 * words_scanned_period / 4; | |
| 3132 _refs_reached_limit = _real_refs_reached_limit - | |
| 3133 3 * refs_reached_period / 4; | |
| 3134 } | |
| 3135 | |
| 3136 void CMTask::move_entries_to_global_stack() { | |
| 3137 // local array where we'll store the entries that will be popped | |
| 3138 // from the local queue | |
| 3139 oop buffer[global_stack_transfer_size]; | |
| 3140 | |
| 3141 int n = 0; | |
| 3142 oop obj; | |
| 3143 while (n < global_stack_transfer_size && _task_queue->pop_local(obj)) { | |
| 3144 buffer[n] = obj; | |
| 3145 ++n; | |
| 3146 } | |
| 3147 | |
| 3148 if (n > 0) { | |
| 3149 // we popped at least one entry from the local queue | |
| 3150 | |
| 3151 statsOnly( ++_global_transfers_to; _local_pops += n ); | |
| 3152 | |
| 3153 if (!_cm->mark_stack_push(buffer, n)) { | |
| 3154 if (_cm->verbose_low()) | |
| 3155 gclog_or_tty->print_cr("[%d] aborting due to global stack overflow", _task_id); | |
| 3156 set_has_aborted(); | |
| 3157 } else { | |
| 3158 // the transfer was successful | |
| 3159 | |
| 3160 if (_cm->verbose_medium()) | |
| 3161 gclog_or_tty->print_cr("[%d] pushed %d entries to the global stack", | |
| 3162 _task_id, n); | |
| 3163 statsOnly( int tmp_size = _cm->mark_stack_size(); | |
| 3164 if (tmp_size > _global_max_size) | |
| 3165 _global_max_size = tmp_size; | |
| 3166 _global_pushes += n ); | |
| 3167 } | |
| 3168 } | |
| 3169 | |
| 3170 // this operation was quite expensive, so decrease the limits | |
| 3171 decrease_limits(); | |
| 3172 } | |
| 3173 | |
| 3174 void CMTask::get_entries_from_global_stack() { | |
| 3175 // local array where we'll store the entries that will be popped | |
| 3176 // from the global stack. | |
| 3177 oop buffer[global_stack_transfer_size]; | |
| 3178 int n; | |
| 3179 _cm->mark_stack_pop(buffer, global_stack_transfer_size, &n); | |
| 3180 tmp_guarantee_CM( n <= global_stack_transfer_size, | |
| 3181 "we should not pop more than the given limit" ); | |
| 3182 if (n > 0) { | |
| 3183 // yes, we did actually pop at least one entry | |
| 3184 | |
| 3185 statsOnly( ++_global_transfers_from; _global_pops += n ); | |
| 3186 if (_cm->verbose_medium()) | |
| 3187 gclog_or_tty->print_cr("[%d] popped %d entries from the global stack", | |
| 3188 _task_id, n); | |
| 3189 for (int i = 0; i < n; ++i) { | |
| 3190 bool success = _task_queue->push(buffer[i]); | |
| 3191 // We only call this when the local queue is empty or under a | |
| 3192 // given target limit. So, we do not expect this push to fail. | |
| 3193 tmp_guarantee_CM( success, "invariant" ); | |
| 3194 } | |
| 3195 | |
| 3196 statsOnly( int tmp_size = _task_queue->size(); | |
| 3197 if (tmp_size > _local_max_size) | |
| 3198 _local_max_size = tmp_size; | |
| 3199 _local_pushes += n ); | |
| 3200 } | |
| 3201 | |
| 3202 // this operation was quite expensive, so decrease the limits | |
| 3203 decrease_limits(); | |
| 3204 } | |
| 3205 | |
| 3206 void CMTask::drain_local_queue(bool partially) { | |
| 3207 if (has_aborted()) | |
| 3208 return; | |
| 3209 | |
| 3210 // Decide what the target size is, depending whether we're going to | |
| 3211 // drain it partially (so that other tasks can steal if they run out | |
| 3212 // of things to do) or totally (at the very end). | |
| 3213 size_t target_size; | |
| 3214 if (partially) | |
| 3215 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize); | |
| 3216 else | |
| 3217 target_size = 0; | |
| 3218 | |
| 3219 if (_task_queue->size() > target_size) { | |
| 3220 if (_cm->verbose_high()) | |
| 3221 gclog_or_tty->print_cr("[%d] draining local queue, target size = %d", | |
| 3222 _task_id, target_size); | |
| 3223 | |
| 3224 oop obj; | |
| 3225 bool ret = _task_queue->pop_local(obj); | |
| 3226 while (ret) { | |
| 3227 statsOnly( ++_local_pops ); | |
| 3228 | |
| 3229 if (_cm->verbose_high()) | |
| 3230 gclog_or_tty->print_cr("[%d] popped "PTR_FORMAT, _task_id, | |
| 3231 (void*) obj); | |
| 3232 | |
| 3233 tmp_guarantee_CM( _g1h->is_in_g1_reserved((HeapWord*) obj), | |
| 3234 "invariant" ); | |
| 3235 | |
| 3236 scan_object(obj); | |
| 3237 | |
| 3238 if (_task_queue->size() <= target_size || has_aborted()) | |
| 3239 ret = false; | |
| 3240 else | |
| 3241 ret = _task_queue->pop_local(obj); | |
| 3242 } | |
| 3243 | |
| 3244 if (_cm->verbose_high()) | |
| 3245 gclog_or_tty->print_cr("[%d] drained local queue, size = %d", | |
| 3246 _task_id, _task_queue->size()); | |
| 3247 } | |
| 3248 } | |
| 3249 | |
| 3250 void CMTask::drain_global_stack(bool partially) { | |
| 3251 if (has_aborted()) | |
| 3252 return; | |
| 3253 | |
| 3254 // We have a policy to drain the local queue before we attempt to | |
| 3255 // drain the global stack. | |
| 3256 tmp_guarantee_CM( partially || _task_queue->size() == 0, "invariant" ); | |
| 3257 | |
| 3258 // Decide what the target size is, depending whether we're going to | |
| 3259 // drain it partially (so that other tasks can steal if they run out | |
| 3260 // of things to do) or totally (at the very end). Notice that, | |
| 3261 // because we move entries from the global stack in chunks or | |
| 3262 // because another task might be doing the same, we might in fact | |
| 3263 // drop below the target. But, this is not a problem. | |
| 3264 size_t target_size; | |
| 3265 if (partially) | |
| 3266 target_size = _cm->partial_mark_stack_size_target(); | |
| 3267 else | |
| 3268 target_size = 0; | |
| 3269 | |
| 3270 if (_cm->mark_stack_size() > target_size) { | |
| 3271 if (_cm->verbose_low()) | |
| 3272 gclog_or_tty->print_cr("[%d] draining global_stack, target size %d", | |
| 3273 _task_id, target_size); | |
| 3274 | |
| 3275 while (!has_aborted() && _cm->mark_stack_size() > target_size) { | |
| 3276 get_entries_from_global_stack(); | |
| 3277 drain_local_queue(partially); | |
| 3278 } | |
| 3279 | |
| 3280 if (_cm->verbose_low()) | |
| 3281 gclog_or_tty->print_cr("[%d] drained global stack, size = %d", | |
| 3282 _task_id, _cm->mark_stack_size()); | |
| 3283 } | |
| 3284 } | |
| 3285 | |
| 3286 // SATB Queue has several assumptions on whether to call the par or | |
| 3287 // non-par versions of the methods. this is why some of the code is | |
| 3288 // replicated. We should really get rid of the single-threaded version | |
| 3289 // of the code to simplify things. | |
| 3290 void CMTask::drain_satb_buffers() { | |
| 3291 if (has_aborted()) | |
| 3292 return; | |
| 3293 | |
| 3294 // We set this so that the regular clock knows that we're in the | |
| 3295 // middle of draining buffers and doesn't set the abort flag when it | |
| 3296 // notices that SATB buffers are available for draining. It'd be | |
| 3297 // very counter productive if it did that. :-) | |
| 3298 _draining_satb_buffers = true; | |
| 3299 | |
| 3300 CMObjectClosure oc(this); | |
| 3301 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); | |
| 3302 if (ParallelGCThreads > 0) | |
| 3303 satb_mq_set.set_par_closure(_task_id, &oc); | |
| 3304 else | |
| 3305 satb_mq_set.set_closure(&oc); | |
| 3306 | |
| 3307 // This keeps claiming and applying the closure to completed buffers | |
| 3308 // until we run out of buffers or we need to abort. | |
| 3309 if (ParallelGCThreads > 0) { | |
| 3310 while (!has_aborted() && | |
| 3311 satb_mq_set.par_apply_closure_to_completed_buffer(_task_id)) { | |
| 3312 if (_cm->verbose_medium()) | |
| 3313 gclog_or_tty->print_cr("[%d] processed an SATB buffer", _task_id); | |
| 3314 statsOnly( ++_satb_buffers_processed ); | |
| 3315 regular_clock_call(); | |
| 3316 } | |
| 3317 } else { | |
| 3318 while (!has_aborted() && | |
| 3319 satb_mq_set.apply_closure_to_completed_buffer()) { | |
| 3320 if (_cm->verbose_medium()) | |
| 3321 gclog_or_tty->print_cr("[%d] processed an SATB buffer", _task_id); | |
| 3322 statsOnly( ++_satb_buffers_processed ); | |
| 3323 regular_clock_call(); | |
| 3324 } | |
| 3325 } | |
| 3326 | |
| 3327 if (!concurrent() && !has_aborted()) { | |
| 3328 // We should only do this during remark. | |
| 3329 if (ParallelGCThreads > 0) | |
| 3330 satb_mq_set.par_iterate_closure_all_threads(_task_id); | |
| 3331 else | |
| 3332 satb_mq_set.iterate_closure_all_threads(); | |
| 3333 } | |
| 3334 | |
| 3335 _draining_satb_buffers = false; | |
| 3336 | |
| 3337 tmp_guarantee_CM( has_aborted() || | |
| 3338 concurrent() || | |
| 3339 satb_mq_set.completed_buffers_num() == 0, "invariant" ); | |
| 3340 | |
| 3341 if (ParallelGCThreads > 0) | |
| 3342 satb_mq_set.set_par_closure(_task_id, NULL); | |
| 3343 else | |
| 3344 satb_mq_set.set_closure(NULL); | |
| 3345 | |
| 3346 // again, this was a potentially expensive operation, decrease the | |
| 3347 // limits to get the regular clock call early | |
| 3348 decrease_limits(); | |
| 3349 } | |
| 3350 | |
| 3351 void CMTask::drain_region_stack(BitMapClosure* bc) { | |
| 3352 if (has_aborted()) | |
| 3353 return; | |
| 3354 | |
| 3355 tmp_guarantee_CM( _region_finger == NULL, | |
| 3356 "it should be NULL when we're not scanning a region" ); | |
| 3357 | |
| 3358 if (!_cm->region_stack_empty()) { | |
| 3359 if (_cm->verbose_low()) | |
| 3360 gclog_or_tty->print_cr("[%d] draining region stack, size = %d", | |
| 3361 _task_id, _cm->region_stack_size()); | |
| 3362 | |
| 3363 MemRegion mr = _cm->region_stack_pop(); | |
| 3364 // it returns MemRegion() if the pop fails | |
| 3365 statsOnly(if (mr.start() != NULL) ++_region_stack_pops ); | |
| 3366 | |
| 3367 while (mr.start() != NULL) { | |
| 3368 if (_cm->verbose_medium()) | |
| 3369 gclog_or_tty->print_cr("[%d] we are scanning region " | |
| 3370 "["PTR_FORMAT", "PTR_FORMAT")", | |
| 3371 _task_id, mr.start(), mr.end()); | |
| 3372 tmp_guarantee_CM( mr.end() <= _cm->finger(), | |
| 3373 "otherwise the region shouldn't be on the stack" ); | |
| 3374 assert(!mr.is_empty(), "Only non-empty regions live on the region stack"); | |
| 3375 if (_nextMarkBitMap->iterate(bc, mr)) { | |
| 3376 tmp_guarantee_CM( !has_aborted(), | |
| 3377 "cannot abort the task without aborting the bitmap iteration" ); | |
| 3378 | |
| 3379 // We finished iterating over the region without aborting. | |
| 3380 regular_clock_call(); | |
| 3381 if (has_aborted()) | |
| 3382 mr = MemRegion(); | |
| 3383 else { | |
| 3384 mr = _cm->region_stack_pop(); | |
| 3385 // it returns MemRegion() if the pop fails | |
| 3386 statsOnly(if (mr.start() != NULL) ++_region_stack_pops ); | |
| 3387 } | |
| 3388 } else { | |
| 3389 guarantee( has_aborted(), "currently the only way to do so" ); | |
| 3390 | |
| 3391 // The only way to abort the bitmap iteration is to return | |
| 3392 // false from the do_bit() method. However, inside the | |
| 3393 // do_bit() method we move the _region_finger to point to the | |
| 3394 // object currently being looked at. So, if we bail out, we | |
| 3395 // have definitely set _region_finger to something non-null. | |
| 3396 guarantee( _region_finger != NULL, "invariant" ); | |
| 3397 | |
| 3398 // The iteration was actually aborted. So now _region_finger | |
| 3399 // points to the address of the object we last scanned. If we | |
| 3400 // leave it there, when we restart this task, we will rescan | |
| 3401 // the object. It is easy to avoid this. We move the finger by | |
| 3402 // enough to point to the next possible object header (the | |
| 3403 // bitmap knows by how much we need to move it as it knows its | |
| 3404 // granularity). | |
| 3405 MemRegion newRegion = | |
| 3406 MemRegion(_nextMarkBitMap->nextWord(_region_finger), mr.end()); | |
| 3407 | |
| 3408 if (!newRegion.is_empty()) { | |
| 3409 if (_cm->verbose_low()) { | |
| 3410 gclog_or_tty->print_cr("[%d] pushing unscanned region" | |
| 3411 "[" PTR_FORMAT "," PTR_FORMAT ") on region stack", | |
| 3412 _task_id, | |
| 3413 newRegion.start(), newRegion.end()); | |
| 3414 } | |
| 3415 // Now push the part of the region we didn't scan on the | |
| 3416 // region stack to make sure a task scans it later. | |
| 3417 _cm->region_stack_push(newRegion); | |
| 3418 } | |
| 3419 // break from while | |
| 3420 mr = MemRegion(); | |
| 3421 } | |
| 3422 _region_finger = NULL; | |
| 3423 } | |
| 3424 | |
| 3425 // We only push regions on the region stack during evacuation | |
| 3426 // pauses. So if we come out the above iteration because we region | |
| 3427 // stack is empty, it will remain empty until the next yield | |
| 3428 // point. So, the guarantee below is safe. | |
| 3429 guarantee( has_aborted() || _cm->region_stack_empty(), | |
| 3430 "only way to exit the loop" ); | |
| 3431 | |
| 3432 if (_cm->verbose_low()) | |
| 3433 gclog_or_tty->print_cr("[%d] drained region stack, size = %d", | |
| 3434 _task_id, _cm->region_stack_size()); | |
| 3435 } | |
| 3436 } | |
| 3437 | |
| 3438 void CMTask::print_stats() { | |
| 3439 gclog_or_tty->print_cr("Marking Stats, task = %d, calls = %d", | |
| 3440 _task_id, _calls); | |
| 3441 gclog_or_tty->print_cr(" Elapsed time = %1.2lfms, Termination time = %1.2lfms", | |
| 3442 _elapsed_time_ms, _termination_time_ms); | |
| 3443 gclog_or_tty->print_cr(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms", | |
| 3444 _step_times_ms.num(), _step_times_ms.avg(), | |
| 3445 _step_times_ms.sd()); | |
| 3446 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms", | |
| 3447 _step_times_ms.maximum(), _step_times_ms.sum()); | |
| 3448 | |
| 3449 #if _MARKING_STATS_ | |
| 3450 gclog_or_tty->print_cr(" Clock Intervals (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms", | |
| 3451 _all_clock_intervals_ms.num(), _all_clock_intervals_ms.avg(), | |
| 3452 _all_clock_intervals_ms.sd()); | |
| 3453 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms", | |
| 3454 _all_clock_intervals_ms.maximum(), | |
| 3455 _all_clock_intervals_ms.sum()); | |
| 3456 gclog_or_tty->print_cr(" Clock Causes (cum): scanning = %d, marking = %d", | |
| 3457 _clock_due_to_scanning, _clock_due_to_marking); | |
| 3458 gclog_or_tty->print_cr(" Objects: scanned = %d, found on the bitmap = %d", | |
| 3459 _objs_scanned, _objs_found_on_bitmap); | |
| 3460 gclog_or_tty->print_cr(" Local Queue: pushes = %d, pops = %d, max size = %d", | |
| 3461 _local_pushes, _local_pops, _local_max_size); | |
| 3462 gclog_or_tty->print_cr(" Global Stack: pushes = %d, pops = %d, max size = %d", | |
| 3463 _global_pushes, _global_pops, _global_max_size); | |
| 3464 gclog_or_tty->print_cr(" transfers to = %d, transfers from = %d", | |
| 3465 _global_transfers_to,_global_transfers_from); | |
| 3466 gclog_or_tty->print_cr(" Regions: claimed = %d, Region Stack: pops = %d", | |
| 3467 _regions_claimed, _region_stack_pops); | |
| 3468 gclog_or_tty->print_cr(" SATB buffers: processed = %d", _satb_buffers_processed); | |
| 3469 gclog_or_tty->print_cr(" Steals: attempts = %d, successes = %d", | |
| 3470 _steal_attempts, _steals); | |
| 3471 gclog_or_tty->print_cr(" Aborted: %d, due to", _aborted); | |
| 3472 gclog_or_tty->print_cr(" overflow: %d, global abort: %d, yield: %d", | |
| 3473 _aborted_overflow, _aborted_cm_aborted, _aborted_yield); | |
| 3474 gclog_or_tty->print_cr(" time out: %d, SATB: %d, termination: %d", | |
| 3475 _aborted_timed_out, _aborted_satb, _aborted_termination); | |
| 3476 #endif // _MARKING_STATS_ | |
| 3477 } | |
| 3478 | |
| 3479 /***************************************************************************** | |
| 3480 | |
| 3481 The do_marking_step(time_target_ms) method is the building block | |
| 3482 of the parallel marking framework. It can be called in parallel | |
| 3483 with other invocations of do_marking_step() on different tasks | |
| 3484 (but only one per task, obviously) and concurrently with the | |
| 3485 mutator threads, or during remark, hence it eliminates the need | |
| 3486 for two versions of the code. When called during remark, it will | |
| 3487 pick up from where the task left off during the concurrent marking | |
| 3488 phase. Interestingly, tasks are also claimable during evacuation | |
| 3489 pauses too, since do_marking_step() ensures that it aborts before | |
| 3490 it needs to yield. | |
| 3491 | |
| 3492 The data structures that is uses to do marking work are the | |
| 3493 following: | |
| 3494 | |
| 3495 (1) Marking Bitmap. If there are gray objects that appear only | |
| 3496 on the bitmap (this happens either when dealing with an overflow | |
| 3497 or when the initial marking phase has simply marked the roots | |
| 3498 and didn't push them on the stack), then tasks claim heap | |
| 3499 regions whose bitmap they then scan to find gray objects. A | |
| 3500 global finger indicates where the end of the last claimed region | |
| 3501 is. A local finger indicates how far into the region a task has | |
| 3502 scanned. The two fingers are used to determine how to gray an | |
| 3503 object (i.e. whether simply marking it is OK, as it will be | |
| 3504 visited by a task in the future, or whether it needs to be also | |
| 3505 pushed on a stack). | |
| 3506 | |
| 3507 (2) Local Queue. The local queue of the task which is accessed | |
| 3508 reasonably efficiently by the task. Other tasks can steal from | |
| 3509 it when they run out of work. Throughout the marking phase, a | |
| 3510 task attempts to keep its local queue short but not totally | |
| 3511 empty, so that entries are available for stealing by other | |
| 3512 tasks. Only when there is no more work, a task will totally | |
| 3513 drain its local queue. | |
| 3514 | |
| 3515 (3) Global Mark Stack. This handles local queue overflow. During | |
| 3516 marking only sets of entries are moved between it and the local | |
| 3517 queues, as access to it requires a mutex and more fine-grain | |
| 3518 interaction with it which might cause contention. If it | |
| 3519 overflows, then the marking phase should restart and iterate | |
| 3520 over the bitmap to identify gray objects. Throughout the marking | |
| 3521 phase, tasks attempt to keep the global mark stack at a small | |
| 3522 length but not totally empty, so that entries are available for | |
| 3523 popping by other tasks. Only when there is no more work, tasks | |
| 3524 will totally drain the global mark stack. | |
| 3525 | |
| 3526 (4) Global Region Stack. Entries on it correspond to areas of | |
| 3527 the bitmap that need to be scanned since they contain gray | |
| 3528 objects. Pushes on the region stack only happen during | |
| 3529 evacuation pauses and typically correspond to areas covered by | |
| 3530 GC LABS. If it overflows, then the marking phase should restart | |
| 3531 and iterate over the bitmap to identify gray objects. Tasks will | |
| 3532 try to totally drain the region stack as soon as possible. | |
| 3533 | |
| 3534 (5) SATB Buffer Queue. This is where completed SATB buffers are | |
| 3535 made available. Buffers are regularly removed from this queue | |
| 3536 and scanned for roots, so that the queue doesn't get too | |
| 3537 long. During remark, all completed buffers are processed, as | |
| 3538 well as the filled in parts of any uncompleted buffers. | |
| 3539 | |
| 3540 The do_marking_step() method tries to abort when the time target | |
| 3541 has been reached. There are a few other cases when the | |
| 3542 do_marking_step() method also aborts: | |
| 3543 | |
| 3544 (1) When the marking phase has been aborted (after a Full GC). | |
| 3545 | |
| 3546 (2) When a global overflow (either on the global stack or the | |
| 3547 region stack) has been triggered. Before the task aborts, it | |
| 3548 will actually sync up with the other tasks to ensure that all | |
| 3549 the marking data structures (local queues, stacks, fingers etc.) | |
| 3550 are re-initialised so that when do_marking_step() completes, | |
| 3551 the marking phase can immediately restart. | |
| 3552 | |
| 3553 (3) When enough completed SATB buffers are available. The | |
| 3554 do_marking_step() method only tries to drain SATB buffers right | |
| 3555 at the beginning. So, if enough buffers are available, the | |
| 3556 marking step aborts and the SATB buffers are processed at | |
| 3557 the beginning of the next invocation. | |
| 3558 | |
| 3559 (4) To yield. when we have to yield then we abort and yield | |
| 3560 right at the end of do_marking_step(). This saves us from a lot | |
| 3561 of hassle as, by yielding we might allow a Full GC. If this | |
| 3562 happens then objects will be compacted underneath our feet, the | |
| 3563 heap might shrink, etc. We save checking for this by just | |
| 3564 aborting and doing the yield right at the end. | |
| 3565 | |
| 3566 From the above it follows that the do_marking_step() method should | |
| 3567 be called in a loop (or, otherwise, regularly) until it completes. | |
| 3568 | |
| 3569 If a marking step completes without its has_aborted() flag being | |
| 3570 true, it means it has completed the current marking phase (and | |
| 3571 also all other marking tasks have done so and have all synced up). | |
| 3572 | |
| 3573 A method called regular_clock_call() is invoked "regularly" (in | |
| 3574 sub ms intervals) throughout marking. It is this clock method that | |
| 3575 checks all the abort conditions which were mentioned above and | |
| 3576 decides when the task should abort. A work-based scheme is used to | |
| 3577 trigger this clock method: when the number of object words the | |
| 3578 marking phase has scanned or the number of references the marking | |
| 3579 phase has visited reach a given limit. Additional invocations to | |
| 3580 the method clock have been planted in a few other strategic places | |
| 3581 too. The initial reason for the clock method was to avoid calling | |
| 3582 vtime too regularly, as it is quite expensive. So, once it was in | |
| 3583 place, it was natural to piggy-back all the other conditions on it | |
| 3584 too and not constantly check them throughout the code. | |
| 3585 | |
| 3586 *****************************************************************************/ | |
| 3587 | |
| 3588 void CMTask::do_marking_step(double time_target_ms) { | |
| 3589 guarantee( time_target_ms >= 1.0, "minimum granularity is 1ms" ); | |
| 3590 guarantee( concurrent() == _cm->concurrent(), "they should be the same" ); | |
| 3591 | |
| 3592 guarantee( concurrent() || _cm->region_stack_empty(), | |
| 3593 "the region stack should have been cleared before remark" ); | |
| 3594 guarantee( _region_finger == NULL, | |
| 3595 "this should be non-null only when a region is being scanned" ); | |
| 3596 | |
| 3597 G1CollectorPolicy* g1_policy = _g1h->g1_policy(); | |
| 3598 guarantee( _task_queues != NULL, "invariant" ); | |
| 3599 guarantee( _task_queue != NULL, "invariant" ); | |
| 3600 guarantee( _task_queues->queue(_task_id) == _task_queue, "invariant" ); | |
| 3601 | |
| 3602 guarantee( !_claimed, | |
| 3603 "only one thread should claim this task at any one time" ); | |
| 3604 | |
| 3605 // OK, this doesn't safeguard again all possible scenarios, as it is | |
| 3606 // possible for two threads to set the _claimed flag at the same | |
| 3607 // time. But it is only for debugging purposes anyway and it will | |
| 3608 // catch most problems. | |
| 3609 _claimed = true; | |
| 3610 | |
| 3611 _start_time_ms = os::elapsedVTime() * 1000.0; | |
| 3612 statsOnly( _interval_start_time_ms = _start_time_ms ); | |
| 3613 | |
| 3614 double diff_prediction_ms = | |
| 3615 g1_policy->get_new_prediction(&_marking_step_diffs_ms); | |
| 3616 _time_target_ms = time_target_ms - diff_prediction_ms; | |
| 3617 | |
| 3618 // set up the variables that are used in the work-based scheme to | |
| 3619 // call the regular clock method | |
| 3620 _words_scanned = 0; | |
| 3621 _refs_reached = 0; | |
| 3622 recalculate_limits(); | |
| 3623 | |
| 3624 // clear all flags | |
| 3625 clear_has_aborted(); | |
| 3626 _has_aborted_timed_out = false; | |
| 3627 _draining_satb_buffers = false; | |
| 3628 | |
| 3629 ++_calls; | |
| 3630 | |
| 3631 if (_cm->verbose_low()) | |
| 3632 gclog_or_tty->print_cr("[%d] >>>>>>>>>> START, call = %d, " | |
| 3633 "target = %1.2lfms >>>>>>>>>>", | |
| 3634 _task_id, _calls, _time_target_ms); | |
| 3635 | |
| 3636 // Set up the bitmap and oop closures. Anything that uses them is | |
| 3637 // eventually called from this method, so it is OK to allocate these | |
| 3638 // statically. | |
| 3639 CMBitMapClosure bitmap_closure(this, _cm, _nextMarkBitMap); | |
| 3640 CMOopClosure oop_closure(_g1h, _cm, this); | |
| 3641 set_oop_closure(&oop_closure); | |
| 3642 | |
| 3643 if (_cm->has_overflown()) { | |
| 3644 // This can happen if the region stack or the mark stack overflows | |
| 3645 // during a GC pause and this task, after a yield point, | |
| 3646 // restarts. We have to abort as we need to get into the overflow | |
| 3647 // protocol which happens right at the end of this task. | |
| 3648 set_has_aborted(); | |
| 3649 } | |
| 3650 | |
| 3651 // First drain any available SATB buffers. After this, we will not | |
| 3652 // look at SATB buffers before the next invocation of this method. | |
| 3653 // If enough completed SATB buffers are queued up, the regular clock | |
| 3654 // will abort this task so that it restarts. | |
| 3655 drain_satb_buffers(); | |
| 3656 // ...then partially drain the local queue and the global stack | |
| 3657 drain_local_queue(true); | |
| 3658 drain_global_stack(true); | |
| 3659 | |
| 3660 // Then totally drain the region stack. We will not look at | |
| 3661 // it again before the next invocation of this method. Entries on | |
| 3662 // the region stack are only added during evacuation pauses, for | |
| 3663 // which we have to yield. When we do, we abort the task anyway so | |
| 3664 // it will look at the region stack again when it restarts. | |
| 3665 bitmap_closure.set_scanning_heap_region(false); | |
| 3666 drain_region_stack(&bitmap_closure); | |
| 3667 // ...then partially drain the local queue and the global stack | |
| 3668 drain_local_queue(true); | |
| 3669 drain_global_stack(true); | |
| 3670 | |
| 3671 do { | |
| 3672 if (!has_aborted() && _curr_region != NULL) { | |
| 3673 // This means that we're already holding on to a region. | |
| 3674 tmp_guarantee_CM( _finger != NULL, | |
| 3675 "if region is not NULL, then the finger " | |
| 3676 "should not be NULL either" ); | |
| 3677 | |
| 3678 // We might have restarted this task after an evacuation pause | |
| 3679 // which might have evacuated the region we're holding on to | |
| 3680 // underneath our feet. Let's read its limit again to make sure | |
| 3681 // that we do not iterate over a region of the heap that | |
| 3682 // contains garbage (update_region_limit() will also move | |
| 3683 // _finger to the start of the region if it is found empty). | |
| 3684 update_region_limit(); | |
| 3685 // We will start from _finger not from the start of the region, | |
| 3686 // as we might be restarting this task after aborting half-way | |
| 3687 // through scanning this region. In this case, _finger points to | |
| 3688 // the address where we last found a marked object. If this is a | |
| 3689 // fresh region, _finger points to start(). | |
| 3690 MemRegion mr = MemRegion(_finger, _region_limit); | |
| 3691 | |
| 3692 if (_cm->verbose_low()) | |
| 3693 gclog_or_tty->print_cr("[%d] we're scanning part " | |
| 3694 "["PTR_FORMAT", "PTR_FORMAT") " | |
| 3695 "of region "PTR_FORMAT, | |
| 3696 _task_id, _finger, _region_limit, _curr_region); | |
| 3697 | |
| 3698 // Let's iterate over the bitmap of the part of the | |
| 3699 // region that is left. | |
| 3700 bitmap_closure.set_scanning_heap_region(true); | |
| 3701 if (mr.is_empty() || | |
| 3702 _nextMarkBitMap->iterate(&bitmap_closure, mr)) { | |
| 3703 // We successfully completed iterating over the region. Now, | |
| 3704 // let's give up the region. | |
| 3705 giveup_current_region(); | |
| 3706 regular_clock_call(); | |
| 3707 } else { | |
| 3708 guarantee( has_aborted(), "currently the only way to do so" ); | |
| 3709 // The only way to abort the bitmap iteration is to return | |
| 3710 // false from the do_bit() method. However, inside the | |
| 3711 // do_bit() method we move the _finger to point to the | |
| 3712 // object currently being looked at. So, if we bail out, we | |
| 3713 // have definitely set _finger to something non-null. | |
| 3714 guarantee( _finger != NULL, "invariant" ); | |
| 3715 | |
| 3716 // Region iteration was actually aborted. So now _finger | |
| 3717 // points to the address of the object we last scanned. If we | |
| 3718 // leave it there, when we restart this task, we will rescan | |
| 3719 // the object. It is easy to avoid this. We move the finger by | |
| 3720 // enough to point to the next possible object header (the | |
| 3721 // bitmap knows by how much we need to move it as it knows its | |
| 3722 // granularity). | |
| 3723 move_finger_to(_nextMarkBitMap->nextWord(_finger)); | |
| 3724 } | |
| 3725 } | |
| 3726 // At this point we have either completed iterating over the | |
| 3727 // region we were holding on to, or we have aborted. | |
| 3728 | |
| 3729 // We then partially drain the local queue and the global stack. | |
| 3730 // (Do we really need this?) | |
| 3731 drain_local_queue(true); | |
| 3732 drain_global_stack(true); | |
| 3733 | |
| 3734 // Read the note on the claim_region() method on why it might | |
| 3735 // return NULL with potentially more regions available for | |
| 3736 // claiming and why we have to check out_of_regions() to determine | |
| 3737 // whether we're done or not. | |
| 3738 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) { | |
| 3739 // We are going to try to claim a new region. We should have | |
| 3740 // given up on the previous one. | |
| 3741 tmp_guarantee_CM( _curr_region == NULL && | |
| 3742 _finger == NULL && | |
| 3743 _region_limit == NULL, "invariant" ); | |
| 3744 if (_cm->verbose_low()) | |
| 3745 gclog_or_tty->print_cr("[%d] trying to claim a new region", _task_id); | |
| 3746 HeapRegion* claimed_region = _cm->claim_region(_task_id); | |
| 3747 if (claimed_region != NULL) { | |
| 3748 // Yes, we managed to claim one | |
| 3749 statsOnly( ++_regions_claimed ); | |
| 3750 | |
| 3751 if (_cm->verbose_low()) | |
| 3752 gclog_or_tty->print_cr("[%d] we successfully claimed " | |
| 3753 "region "PTR_FORMAT, | |
| 3754 _task_id, claimed_region); | |
| 3755 | |
| 3756 setup_for_region(claimed_region); | |
| 3757 tmp_guarantee_CM( _curr_region == claimed_region, "invariant" ); | |
| 3758 } | |
| 3759 // It is important to call the regular clock here. It might take | |
| 3760 // a while to claim a region if, for example, we hit a large | |
| 3761 // block of empty regions. So we need to call the regular clock | |
| 3762 // method once round the loop to make sure it's called | |
| 3763 // frequently enough. | |
| 3764 regular_clock_call(); | |
| 3765 } | |
| 3766 | |
| 3767 if (!has_aborted() && _curr_region == NULL) { | |
| 3768 tmp_guarantee_CM( _cm->out_of_regions(), | |
| 3769 "at this point we should be out of regions" ); | |
| 3770 } | |
| 3771 } while ( _curr_region != NULL && !has_aborted()); | |
| 3772 | |
| 3773 if (!has_aborted()) { | |
| 3774 // We cannot check whether the global stack is empty, since other | |
|
343
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3775 // tasks might be pushing objects to it concurrently. We also cannot |
|
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3776 // check if the region stack is empty because if a thread is aborting |
|
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3777 // it can push a partially done region back. |
|
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3778 tmp_guarantee_CM( _cm->out_of_regions(), |
| 342 | 3779 "at this point we should be out of regions" ); |
| 3780 | |
| 3781 if (_cm->verbose_low()) | |
| 3782 gclog_or_tty->print_cr("[%d] all regions claimed", _task_id); | |
| 3783 | |
| 3784 // Try to reduce the number of available SATB buffers so that | |
| 3785 // remark has less work to do. | |
| 3786 drain_satb_buffers(); | |
| 3787 } | |
| 3788 | |
| 3789 // Since we've done everything else, we can now totally drain the | |
| 3790 // local queue and global stack. | |
| 3791 drain_local_queue(false); | |
| 3792 drain_global_stack(false); | |
| 3793 | |
| 3794 // Attempt at work stealing from other task's queues. | |
| 3795 if (!has_aborted()) { | |
| 3796 // We have not aborted. This means that we have finished all that | |
| 3797 // we could. Let's try to do some stealing... | |
| 3798 | |
| 3799 // We cannot check whether the global stack is empty, since other | |
|
343
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3800 // tasks might be pushing objects to it concurrently. We also cannot |
|
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3801 // check if the region stack is empty because if a thread is aborting |
|
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3802 // it can push a partially done region back. |
| 342 | 3803 guarantee( _cm->out_of_regions() && |
| 3804 _task_queue->size() == 0, "only way to reach here" ); | |
| 3805 | |
| 3806 if (_cm->verbose_low()) | |
| 3807 gclog_or_tty->print_cr("[%d] starting to steal", _task_id); | |
| 3808 | |
| 3809 while (!has_aborted()) { | |
| 3810 oop obj; | |
| 3811 statsOnly( ++_steal_attempts ); | |
| 3812 | |
| 3813 if (_cm->try_stealing(_task_id, &_hash_seed, obj)) { | |
| 3814 if (_cm->verbose_medium()) | |
| 3815 gclog_or_tty->print_cr("[%d] stolen "PTR_FORMAT" successfully", | |
| 3816 _task_id, (void*) obj); | |
| 3817 | |
| 3818 statsOnly( ++_steals ); | |
| 3819 | |
| 3820 tmp_guarantee_CM( _nextMarkBitMap->isMarked((HeapWord*) obj), | |
| 3821 "any stolen object should be marked" ); | |
| 3822 scan_object(obj); | |
| 3823 | |
| 3824 // And since we're towards the end, let's totally drain the | |
| 3825 // local queue and global stack. | |
| 3826 drain_local_queue(false); | |
| 3827 drain_global_stack(false); | |
| 3828 } else { | |
| 3829 break; | |
| 3830 } | |
| 3831 } | |
| 3832 } | |
| 3833 | |
| 3834 // We still haven't aborted. Now, let's try to get into the | |
| 3835 // termination protocol. | |
| 3836 if (!has_aborted()) { | |
| 3837 // We cannot check whether the global stack is empty, since other | |
|
343
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3838 // tasks might be concurrently pushing objects on it. We also cannot |
|
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3839 // check if the region stack is empty because if a thread is aborting |
|
afc1ce1efe66
6710665: G1: guarantee(_cm->out_of_regions() && _cm->region_stack_empty() && _task_queue->size() == 0, ...)
iveresov
parents:
342
diff
changeset
|
3840 // it can push a partially done region back. |
| 342 | 3841 guarantee( _cm->out_of_regions() && |
| 3842 _task_queue->size() == 0, "only way to reach here" ); | |
| 3843 | |
| 3844 if (_cm->verbose_low()) | |
| 3845 gclog_or_tty->print_cr("[%d] starting termination protocol", _task_id); | |
| 3846 | |
| 3847 _termination_start_time_ms = os::elapsedVTime() * 1000.0; | |
| 3848 // The CMTask class also extends the TerminatorTerminator class, | |
| 3849 // hence its should_exit_termination() method will also decide | |
| 3850 // whether to exit the termination protocol or not. | |
| 3851 bool finished = _cm->terminator()->offer_termination(this); | |
| 3852 double termination_end_time_ms = os::elapsedVTime() * 1000.0; | |
| 3853 _termination_time_ms += | |
| 3854 termination_end_time_ms - _termination_start_time_ms; | |
| 3855 | |
| 3856 if (finished) { | |
| 3857 // We're all done. | |
| 3858 | |
| 3859 if (_task_id == 0) { | |
| 3860 // let's allow task 0 to do this | |
| 3861 if (concurrent()) { | |
| 3862 guarantee( _cm->concurrent_marking_in_progress(), "invariant" ); | |
| 3863 // we need to set this to false before the next | |
| 3864 // safepoint. This way we ensure that the marking phase | |
| 3865 // doesn't observe any more heap expansions. | |
| 3866 _cm->clear_concurrent_marking_in_progress(); | |
| 3867 } | |
| 3868 } | |
| 3869 | |
| 3870 // We can now guarantee that the global stack is empty, since | |
| 3871 // all other tasks have finished. | |
| 3872 guarantee( _cm->out_of_regions() && | |
| 3873 _cm->region_stack_empty() && | |
| 3874 _cm->mark_stack_empty() && | |
| 3875 _task_queue->size() == 0 && | |
| 3876 !_cm->has_overflown() && | |
| 3877 !_cm->mark_stack_overflow() && | |
| 3878 !_cm->region_stack_overflow(), | |
| 3879 "only way to reach here" ); | |
| 3880 | |
| 3881 if (_cm->verbose_low()) | |
| 3882 gclog_or_tty->print_cr("[%d] all tasks terminated", _task_id); | |
| 3883 } else { | |
| 3884 // Apparently there's more work to do. Let's abort this task. It | |
| 3885 // will restart it and we can hopefully find more things to do. | |
| 3886 | |
| 3887 if (_cm->verbose_low()) | |
| 3888 gclog_or_tty->print_cr("[%d] apparently there is more work to do", _task_id); | |
| 3889 | |
| 3890 set_has_aborted(); | |
| 3891 statsOnly( ++_aborted_termination ); | |
| 3892 } | |
| 3893 } | |
| 3894 | |
| 3895 // Mainly for debugging purposes to make sure that a pointer to the | |
| 3896 // closure which was statically allocated in this frame doesn't | |
| 3897 // escape it by accident. | |
| 3898 set_oop_closure(NULL); | |
| 3899 double end_time_ms = os::elapsedVTime() * 1000.0; | |
| 3900 double elapsed_time_ms = end_time_ms - _start_time_ms; | |
| 3901 // Update the step history. | |
| 3902 _step_times_ms.add(elapsed_time_ms); | |
| 3903 | |
| 3904 if (has_aborted()) { | |
| 3905 // The task was aborted for some reason. | |
| 3906 | |
| 3907 statsOnly( ++_aborted ); | |
| 3908 | |
| 3909 if (_has_aborted_timed_out) { | |
| 3910 double diff_ms = elapsed_time_ms - _time_target_ms; | |
| 3911 // Keep statistics of how well we did with respect to hitting | |
| 3912 // our target only if we actually timed out (if we aborted for | |
| 3913 // other reasons, then the results might get skewed). | |
| 3914 _marking_step_diffs_ms.add(diff_ms); | |
| 3915 } | |
| 3916 | |
| 3917 if (_cm->has_overflown()) { | |
| 3918 // This is the interesting one. We aborted because a global | |
| 3919 // overflow was raised. This means we have to restart the | |
| 3920 // marking phase and start iterating over regions. However, in | |
| 3921 // order to do this we have to make sure that all tasks stop | |
| 3922 // what they are doing and re-initialise in a safe manner. We | |
| 3923 // will achieve this with the use of two barrier sync points. | |
| 3924 | |
| 3925 if (_cm->verbose_low()) | |
| 3926 gclog_or_tty->print_cr("[%d] detected overflow", _task_id); | |
| 3927 | |
| 3928 _cm->enter_first_sync_barrier(_task_id); | |
| 3929 // When we exit this sync barrier we know that all tasks have | |
| 3930 // stopped doing marking work. So, it's now safe to | |
| 3931 // re-initialise our data structures. At the end of this method, | |
| 3932 // task 0 will clear the global data structures. | |
| 3933 | |
| 3934 statsOnly( ++_aborted_overflow ); | |
| 3935 | |
| 3936 // We clear the local state of this task... | |
| 3937 clear_region_fields(); | |
| 3938 | |
| 3939 // ...and enter the second barrier. | |
| 3940 _cm->enter_second_sync_barrier(_task_id); | |
| 3941 // At this point everything has bee re-initialised and we're | |
| 3942 // ready to restart. | |
| 3943 } | |
| 3944 | |
| 3945 if (_cm->verbose_low()) { | |
| 3946 gclog_or_tty->print_cr("[%d] <<<<<<<<<< ABORTING, target = %1.2lfms, " | |
| 3947 "elapsed = %1.2lfms <<<<<<<<<<", | |
| 3948 _task_id, _time_target_ms, elapsed_time_ms); | |
| 3949 if (_cm->has_aborted()) | |
| 3950 gclog_or_tty->print_cr("[%d] ========== MARKING ABORTED ==========", | |
| 3951 _task_id); | |
| 3952 } | |
| 3953 } else { | |
| 3954 if (_cm->verbose_low()) | |
| 3955 gclog_or_tty->print_cr("[%d] <<<<<<<<<< FINISHED, target = %1.2lfms, " | |
| 3956 "elapsed = %1.2lfms <<<<<<<<<<", | |
| 3957 _task_id, _time_target_ms, elapsed_time_ms); | |
| 3958 } | |
| 3959 | |
| 3960 _claimed = false; | |
| 3961 } | |
| 3962 | |
| 3963 CMTask::CMTask(int task_id, | |
| 3964 ConcurrentMark* cm, | |
| 3965 CMTaskQueue* task_queue, | |
| 3966 CMTaskQueueSet* task_queues) | |
| 3967 : _g1h(G1CollectedHeap::heap()), | |
| 3968 _co_tracker(G1CMGroup), | |
| 3969 _task_id(task_id), _cm(cm), | |
| 3970 _claimed(false), | |
| 3971 _nextMarkBitMap(NULL), _hash_seed(17), | |
| 3972 _task_queue(task_queue), | |
| 3973 _task_queues(task_queues), | |
| 3974 _oop_closure(NULL) { | |
| 3975 guarantee( task_queue != NULL, "invariant" ); | |
| 3976 guarantee( task_queues != NULL, "invariant" ); | |
| 3977 | |
| 3978 statsOnly( _clock_due_to_scanning = 0; | |
| 3979 _clock_due_to_marking = 0 ); | |
| 3980 | |
| 3981 _marking_step_diffs_ms.add(0.5); | |
| 3982 } |