1 /* 2 * Copyright (c) 2001, 2011, Oracle and/or its affiliates. All rights reserved. 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 4 * 5 * This code is free software; you can redistribute it and/or modify it 6 * under the terms of the GNU General Public License version 2 only, as 7 * published by the Free Software Foundation. 8 * 9 * This code is distributed in the hope that it will be useful, but WITHOUT 10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 12 * version 2 for more details (a copy is included in the LICENSE file that 13 * accompanied this code). 14 * 15 * You should have received a copy of the GNU General Public License version 16 * 2 along with this work; if not, write to the Free Software Foundation, 17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 18 * 19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 20 * or visit www.oracle.com if you need additional information or have any 21 * questions. 22 * 23 */ 24 25 #include "precompiled.hpp" 26 #include "gc_implementation/g1/concurrentG1Refine.hpp" 27 #include "gc_implementation/g1/concurrentMark.hpp" 28 #include "gc_implementation/g1/concurrentMarkThread.inline.hpp" 29 #include "gc_implementation/g1/g1CollectedHeap.inline.hpp" 30 #include "gc_implementation/g1/g1CollectorPolicy.hpp" 31 #include "gc_implementation/g1/heapRegionRemSet.hpp" 32 #include "gc_implementation/shared/gcPolicyCounters.hpp" 33 #include "runtime/arguments.hpp" 34 #include "runtime/java.hpp" 35 #include "runtime/mutexLocker.hpp" 36 #include "utilities/debug.hpp" 37 38 #define PREDICTIONS_VERBOSE 0 39 40 // <NEW PREDICTION> 41 42 // Different defaults for different number of GC threads 43 // They were chosen by running GCOld and SPECjbb on debris with different 44 // numbers of GC threads and choosing them based on the results 45 46 // all the same 47 static double rs_length_diff_defaults[] = { 48 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0 49 }; 50 51 static double cost_per_card_ms_defaults[] = { 52 0.01, 0.005, 0.005, 0.003, 0.003, 0.002, 0.002, 0.0015 53 }; 54 55 // all the same 56 static double fully_young_cards_per_entry_ratio_defaults[] = { 57 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0 58 }; 59 60 static double cost_per_entry_ms_defaults[] = { 61 0.015, 0.01, 0.01, 0.008, 0.008, 0.0055, 0.0055, 0.005 62 }; 63 64 static double cost_per_byte_ms_defaults[] = { 65 0.00006, 0.00003, 0.00003, 0.000015, 0.000015, 0.00001, 0.00001, 0.000009 66 }; 67 68 // these should be pretty consistent 69 static double constant_other_time_ms_defaults[] = { 70 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0 71 }; 72 73 74 static double young_other_cost_per_region_ms_defaults[] = { 75 0.3, 0.2, 0.2, 0.15, 0.15, 0.12, 0.12, 0.1 76 }; 77 78 static double non_young_other_cost_per_region_ms_defaults[] = { 79 1.0, 0.7, 0.7, 0.5, 0.5, 0.42, 0.42, 0.30 80 }; 81 82 // </NEW PREDICTION> 83 84 // Help class for avoiding interleaved logging 85 class LineBuffer: public StackObj { 86 87 private: 88 static const int BUFFER_LEN = 1024; 89 static const int INDENT_CHARS = 3; 90 char _buffer[BUFFER_LEN]; 91 int _indent_level; 92 int _cur; 93 94 void vappend(const char* format, va_list ap) { 95 int res = vsnprintf(&_buffer[_cur], BUFFER_LEN - _cur, format, ap); 96 if (res != -1) { 97 _cur += res; 98 } else { 99 DEBUG_ONLY(warning("buffer too small in LineBuffer");) 100 _buffer[BUFFER_LEN -1] = 0; 101 _cur = BUFFER_LEN; // vsnprintf above should not add to _buffer if we are called again 102 } 103 } 104 105 public: 106 explicit LineBuffer(int indent_level): _indent_level(indent_level), _cur(0) { 107 for (; (_cur < BUFFER_LEN && _cur < (_indent_level * INDENT_CHARS)); _cur++) { 108 _buffer[_cur] = ' '; 109 } 110 } 111 112 #ifndef PRODUCT 113 ~LineBuffer() { 114 assert(_cur == _indent_level * INDENT_CHARS, "pending data in buffer - append_and_print_cr() not called?"); 115 } 116 #endif 117 118 void append(const char* format, ...) { 119 va_list ap; 120 va_start(ap, format); 121 vappend(format, ap); 122 va_end(ap); 123 } 124 125 void append_and_print_cr(const char* format, ...) { 126 va_list ap; 127 va_start(ap, format); 128 vappend(format, ap); 129 va_end(ap); 130 gclog_or_tty->print_cr("%s", _buffer); 131 _cur = _indent_level * INDENT_CHARS; 132 } 133 }; 134 135 G1CollectorPolicy::G1CollectorPolicy() : 136 _parallel_gc_threads(G1CollectedHeap::use_parallel_gc_threads() 137 ? ParallelGCThreads : 1), 138 139 _n_pauses(0), 140 _recent_rs_scan_times_ms(new TruncatedSeq(NumPrevPausesForHeuristics)), 141 _recent_pause_times_ms(new TruncatedSeq(NumPrevPausesForHeuristics)), 142 _recent_rs_sizes(new TruncatedSeq(NumPrevPausesForHeuristics)), 143 _recent_gc_times_ms(new TruncatedSeq(NumPrevPausesForHeuristics)), 144 _all_pause_times_ms(new NumberSeq()), 145 _stop_world_start(0.0), 146 _all_stop_world_times_ms(new NumberSeq()), 147 _all_yield_times_ms(new NumberSeq()), 148 _using_new_ratio_calculations(false), 149 150 _all_mod_union_times_ms(new NumberSeq()), 151 152 _summary(new Summary()), 153 154 #ifndef PRODUCT 155 _cur_clear_ct_time_ms(0.0), 156 _min_clear_cc_time_ms(-1.0), 157 _max_clear_cc_time_ms(-1.0), 158 _cur_clear_cc_time_ms(0.0), 159 _cum_clear_cc_time_ms(0.0), 160 _num_cc_clears(0L), 161 #endif 162 163 _region_num_young(0), 164 _region_num_tenured(0), 165 _prev_region_num_young(0), 166 _prev_region_num_tenured(0), 167 168 _aux_num(10), 169 _all_aux_times_ms(new NumberSeq[_aux_num]), 170 _cur_aux_start_times_ms(new double[_aux_num]), 171 _cur_aux_times_ms(new double[_aux_num]), 172 _cur_aux_times_set(new bool[_aux_num]), 173 174 _concurrent_mark_remark_times_ms(new TruncatedSeq(NumPrevPausesForHeuristics)), 175 _concurrent_mark_cleanup_times_ms(new TruncatedSeq(NumPrevPausesForHeuristics)), 176 177 // <NEW PREDICTION> 178 179 _alloc_rate_ms_seq(new TruncatedSeq(TruncatedSeqLength)), 180 _prev_collection_pause_end_ms(0.0), 181 _pending_card_diff_seq(new TruncatedSeq(TruncatedSeqLength)), 182 _rs_length_diff_seq(new TruncatedSeq(TruncatedSeqLength)), 183 _cost_per_card_ms_seq(new TruncatedSeq(TruncatedSeqLength)), 184 _fully_young_cards_per_entry_ratio_seq(new TruncatedSeq(TruncatedSeqLength)), 185 _partially_young_cards_per_entry_ratio_seq( 186 new TruncatedSeq(TruncatedSeqLength)), 187 _cost_per_entry_ms_seq(new TruncatedSeq(TruncatedSeqLength)), 188 _partially_young_cost_per_entry_ms_seq(new TruncatedSeq(TruncatedSeqLength)), 189 _cost_per_byte_ms_seq(new TruncatedSeq(TruncatedSeqLength)), 190 _cost_per_byte_ms_during_cm_seq(new TruncatedSeq(TruncatedSeqLength)), 191 _constant_other_time_ms_seq(new TruncatedSeq(TruncatedSeqLength)), 192 _young_other_cost_per_region_ms_seq(new TruncatedSeq(TruncatedSeqLength)), 193 _non_young_other_cost_per_region_ms_seq( 194 new TruncatedSeq(TruncatedSeqLength)), 195 196 _pending_cards_seq(new TruncatedSeq(TruncatedSeqLength)), 197 _scanned_cards_seq(new TruncatedSeq(TruncatedSeqLength)), 198 _rs_lengths_seq(new TruncatedSeq(TruncatedSeqLength)), 199 200 _pause_time_target_ms((double) MaxGCPauseMillis), 201 202 // </NEW PREDICTION> 203 204 _full_young_gcs(true), 205 _full_young_pause_num(0), 206 _partial_young_pause_num(0), 207 208 _during_marking(false), 209 _in_marking_window(false), 210 _in_marking_window_im(false), 211 212 _known_garbage_ratio(0.0), 213 _known_garbage_bytes(0), 214 215 _young_gc_eff_seq(new TruncatedSeq(TruncatedSeqLength)), 216 217 _recent_prev_end_times_for_all_gcs_sec(new TruncatedSeq(NumPrevPausesForHeuristics)), 218 219 _recent_CS_bytes_used_before(new TruncatedSeq(NumPrevPausesForHeuristics)), 220 _recent_CS_bytes_surviving(new TruncatedSeq(NumPrevPausesForHeuristics)), 221 222 _recent_avg_pause_time_ratio(0.0), 223 _num_markings(0), 224 _n_marks(0), 225 _n_pauses_at_mark_end(0), 226 227 _all_full_gc_times_ms(new NumberSeq()), 228 229 // G1PausesBtwnConcMark defaults to -1 230 // so the hack is to do the cast QQQ FIXME 231 _pauses_btwn_concurrent_mark((size_t)G1PausesBtwnConcMark), 232 _n_marks_since_last_pause(0), 233 _initiate_conc_mark_if_possible(false), 234 _during_initial_mark_pause(false), 235 _should_revert_to_full_young_gcs(false), 236 _last_full_young_gc(false), 237 238 _eden_bytes_before_gc(0), 239 _survivor_bytes_before_gc(0), 240 _capacity_before_gc(0), 241 242 _prev_collection_pause_used_at_end_bytes(0), 243 244 _collection_set(NULL), 245 _collection_set_size(0), 246 _collection_set_bytes_used_before(0), 247 248 // Incremental CSet attributes 249 _inc_cset_build_state(Inactive), 250 _inc_cset_head(NULL), 251 _inc_cset_tail(NULL), 252 _inc_cset_size(0), 253 _inc_cset_young_index(0), 254 _inc_cset_bytes_used_before(0), 255 _inc_cset_max_finger(NULL), 256 _inc_cset_recorded_young_bytes(0), 257 _inc_cset_recorded_rs_lengths(0), 258 _inc_cset_predicted_elapsed_time_ms(0.0), 259 _inc_cset_predicted_bytes_to_copy(0), 260 261 #ifdef _MSC_VER // the use of 'this' below gets a warning, make it go away 262 #pragma warning( disable:4355 ) // 'this' : used in base member initializer list 263 #endif // _MSC_VER 264 265 _short_lived_surv_rate_group(new SurvRateGroup(this, "Short Lived", 266 G1YoungSurvRateNumRegionsSummary)), 267 _survivor_surv_rate_group(new SurvRateGroup(this, "Survivor", 268 G1YoungSurvRateNumRegionsSummary)), 269 // add here any more surv rate groups 270 _recorded_survivor_regions(0), 271 _recorded_survivor_head(NULL), 272 _recorded_survivor_tail(NULL), 273 _survivors_age_table(true), 274 275 _gc_overhead_perc(0.0) 276 277 { 278 // Set up the region size and associated fields. Given that the 279 // policy is created before the heap, we have to set this up here, 280 // so it's done as soon as possible. 281 HeapRegion::setup_heap_region_size(Arguments::min_heap_size()); 282 HeapRegionRemSet::setup_remset_size(); 283 284 // Verify PLAB sizes 285 const uint region_size = HeapRegion::GrainWords; 286 if (YoungPLABSize > region_size || OldPLABSize > region_size) { 287 char buffer[128]; 288 jio_snprintf(buffer, sizeof(buffer), "%sPLABSize should be at most %u", 289 OldPLABSize > region_size ? "Old" : "Young", region_size); 290 vm_exit_during_initialization(buffer); 291 } 292 293 _recent_prev_end_times_for_all_gcs_sec->add(os::elapsedTime()); 294 _prev_collection_pause_end_ms = os::elapsedTime() * 1000.0; 295 296 _par_last_gc_worker_start_times_ms = new double[_parallel_gc_threads]; 297 _par_last_ext_root_scan_times_ms = new double[_parallel_gc_threads]; 298 _par_last_mark_stack_scan_times_ms = new double[_parallel_gc_threads]; 299 300 _par_last_update_rs_times_ms = new double[_parallel_gc_threads]; 301 _par_last_update_rs_processed_buffers = new double[_parallel_gc_threads]; 302 303 _par_last_scan_rs_times_ms = new double[_parallel_gc_threads]; 304 305 _par_last_obj_copy_times_ms = new double[_parallel_gc_threads]; 306 307 _par_last_termination_times_ms = new double[_parallel_gc_threads]; 308 _par_last_termination_attempts = new double[_parallel_gc_threads]; 309 _par_last_gc_worker_end_times_ms = new double[_parallel_gc_threads]; 310 _par_last_gc_worker_times_ms = new double[_parallel_gc_threads]; 311 312 // start conservatively 313 _expensive_region_limit_ms = 0.5 * (double) MaxGCPauseMillis; 314 315 // <NEW PREDICTION> 316 317 int index; 318 if (ParallelGCThreads == 0) 319 index = 0; 320 else if (ParallelGCThreads > 8) 321 index = 7; 322 else 323 index = ParallelGCThreads - 1; 324 325 _pending_card_diff_seq->add(0.0); 326 _rs_length_diff_seq->add(rs_length_diff_defaults[index]); 327 _cost_per_card_ms_seq->add(cost_per_card_ms_defaults[index]); 328 _fully_young_cards_per_entry_ratio_seq->add( 329 fully_young_cards_per_entry_ratio_defaults[index]); 330 _cost_per_entry_ms_seq->add(cost_per_entry_ms_defaults[index]); 331 _cost_per_byte_ms_seq->add(cost_per_byte_ms_defaults[index]); 332 _constant_other_time_ms_seq->add(constant_other_time_ms_defaults[index]); 333 _young_other_cost_per_region_ms_seq->add( 334 young_other_cost_per_region_ms_defaults[index]); 335 _non_young_other_cost_per_region_ms_seq->add( 336 non_young_other_cost_per_region_ms_defaults[index]); 337 338 // </NEW PREDICTION> 339 340 // Below, we might need to calculate the pause time target based on 341 // the pause interval. When we do so we are going to give G1 maximum 342 // flexibility and allow it to do pauses when it needs to. So, we'll 343 // arrange that the pause interval to be pause time target + 1 to 344 // ensure that a) the pause time target is maximized with respect to 345 // the pause interval and b) we maintain the invariant that pause 346 // time target < pause interval. If the user does not want this 347 // maximum flexibility, they will have to set the pause interval 348 // explicitly. 349 350 // First make sure that, if either parameter is set, its value is 351 // reasonable. 352 if (!FLAG_IS_DEFAULT(MaxGCPauseMillis)) { 353 if (MaxGCPauseMillis < 1) { 354 vm_exit_during_initialization("MaxGCPauseMillis should be " 355 "greater than 0"); 356 } 357 } 358 if (!FLAG_IS_DEFAULT(GCPauseIntervalMillis)) { 359 if (GCPauseIntervalMillis < 1) { 360 vm_exit_during_initialization("GCPauseIntervalMillis should be " 361 "greater than 0"); 362 } 363 } 364 365 // Then, if the pause time target parameter was not set, set it to 366 // the default value. 367 if (FLAG_IS_DEFAULT(MaxGCPauseMillis)) { 368 if (FLAG_IS_DEFAULT(GCPauseIntervalMillis)) { 369 // The default pause time target in G1 is 200ms 370 FLAG_SET_DEFAULT(MaxGCPauseMillis, 200); 371 } else { 372 // We do not allow the pause interval to be set without the 373 // pause time target 374 vm_exit_during_initialization("GCPauseIntervalMillis cannot be set " 375 "without setting MaxGCPauseMillis"); 376 } 377 } 378 379 // Then, if the interval parameter was not set, set it according to 380 // the pause time target (this will also deal with the case when the 381 // pause time target is the default value). 382 if (FLAG_IS_DEFAULT(GCPauseIntervalMillis)) { 383 FLAG_SET_DEFAULT(GCPauseIntervalMillis, MaxGCPauseMillis + 1); 384 } 385 386 // Finally, make sure that the two parameters are consistent. 387 if (MaxGCPauseMillis >= GCPauseIntervalMillis) { 388 char buffer[256]; 389 jio_snprintf(buffer, 256, 390 "MaxGCPauseMillis (%u) should be less than " 391 "GCPauseIntervalMillis (%u)", 392 MaxGCPauseMillis, GCPauseIntervalMillis); 393 vm_exit_during_initialization(buffer); 394 } 395 396 double max_gc_time = (double) MaxGCPauseMillis / 1000.0; 397 double time_slice = (double) GCPauseIntervalMillis / 1000.0; 398 _mmu_tracker = new G1MMUTrackerQueue(time_slice, max_gc_time); 399 _sigma = (double) G1ConfidencePercent / 100.0; 400 401 // start conservatively (around 50ms is about right) 402 _concurrent_mark_remark_times_ms->add(0.05); 403 _concurrent_mark_cleanup_times_ms->add(0.20); 404 _tenuring_threshold = MaxTenuringThreshold; 405 // _max_survivor_regions will be calculated by 406 // update_young_list_target_length() during initialization. 407 _max_survivor_regions = 0; 408 409 assert(GCTimeRatio > 0, 410 "we should have set it to a default value set_g1_gc_flags() " 411 "if a user set it to 0"); 412 _gc_overhead_perc = 100.0 * (1.0 / (1.0 + GCTimeRatio)); 413 414 uintx reserve_perc = G1ReservePercent; 415 // Put an artificial ceiling on this so that it's not set to a silly value. 416 if (reserve_perc > 50) { 417 reserve_perc = 50; 418 warning("G1ReservePercent is set to a value that is too large, " 419 "it's been updated to %u", G1ReservePercent); 420 } 421 _reserve_factor = (double) reserve_perc / 100.0; 422 // This will be set when the heap is expanded 423 // for the first time during initialization. 424 _reserve_regions = 0; 425 426 initialize_all(); 427 } 428 429 // Increment "i", mod "len" 430 static void inc_mod(int& i, int len) { 431 i++; if (i == len) i = 0; 432 } 433 434 void G1CollectorPolicy::initialize_flags() { 435 set_min_alignment(HeapRegion::GrainBytes); 436 set_max_alignment(GenRemSet::max_alignment_constraint(rem_set_name())); 437 if (SurvivorRatio < 1) { 438 vm_exit_during_initialization("Invalid survivor ratio specified"); 439 } 440 CollectorPolicy::initialize_flags(); 441 } 442 443 // The easiest way to deal with the parsing of the NewSize / 444 // MaxNewSize / etc. parameteres is to re-use the code in the 445 // TwoGenerationCollectorPolicy class. This is similar to what 446 // ParallelScavenge does with its GenerationSizer class (see 447 // ParallelScavengeHeap::initialize()). We might change this in the 448 // future, but it's a good start. 449 class G1YoungGenSizer : public TwoGenerationCollectorPolicy { 450 size_t size_to_region_num(size_t byte_size) { 451 return MAX2((size_t) 1, byte_size / HeapRegion::GrainBytes); 452 } 453 454 public: 455 G1YoungGenSizer() { 456 initialize_flags(); 457 initialize_size_info(); 458 } 459 460 size_t min_young_region_num() { 461 return size_to_region_num(_min_gen0_size); 462 } 463 size_t initial_young_region_num() { 464 return size_to_region_num(_initial_gen0_size); 465 } 466 size_t max_young_region_num() { 467 return size_to_region_num(_max_gen0_size); 468 } 469 }; 470 471 void G1CollectorPolicy::update_young_list_size_using_newratio(size_t number_of_heap_regions) { 472 assert(number_of_heap_regions > 0, "Heap must be initialized"); 473 size_t young_size = number_of_heap_regions / (NewRatio + 1); 474 _min_desired_young_length = young_size; 475 _max_desired_young_length = young_size; 476 } 477 478 void G1CollectorPolicy::init() { 479 // Set aside an initial future to_space. 480 _g1 = G1CollectedHeap::heap(); 481 482 assert(Heap_lock->owned_by_self(), "Locking discipline."); 483 484 initialize_gc_policy_counters(); 485 486 G1YoungGenSizer sizer; 487 size_t initial_region_num = sizer.initial_young_region_num(); 488 _min_desired_young_length = sizer.min_young_region_num(); 489 _max_desired_young_length = sizer.max_young_region_num(); 490 491 if (FLAG_IS_CMDLINE(NewRatio)) { 492 if (FLAG_IS_CMDLINE(NewSize) || FLAG_IS_CMDLINE(MaxNewSize)) { 493 gclog_or_tty->print_cr("-XX:NewSize and -XX:MaxNewSize overrides -XX:NewRatio"); 494 } else { 495 // Treat NewRatio as a fixed size that is only recalculated when the heap size changes 496 size_t heap_regions = sizer.size_to_region_num(_g1->n_regions()); 497 update_young_list_size_using_newratio(heap_regions); 498 _using_new_ratio_calculations = true; 499 } 500 } 501 502 // GenCollectorPolicy guarantees that min <= initial <= max. 503 // Asserting here just to state that we rely on this property. 504 assert(_min_desired_young_length <= _max_desired_young_length, "Invalid min/max young gen size values"); 505 assert(initial_region_num <= _max_desired_young_length, "Initial young gen size too large"); 506 assert(_min_desired_young_length <= initial_region_num, "Initial young gen size too small"); 507 508 set_adaptive_young_list_length(_min_desired_young_length < _max_desired_young_length); 509 if (adaptive_young_list_length()) { 510 _young_list_fixed_length = 0; 511 } else { 512 _young_list_fixed_length = initial_region_num; 513 } 514 _free_regions_at_end_of_collection = _g1->free_regions(); 515 update_young_list_target_length(); 516 _prev_eden_capacity = _young_list_target_length * HeapRegion::GrainBytes; 517 518 // We may immediately start allocating regions and placing them on the 519 // collection set list. Initialize the per-collection set info 520 start_incremental_cset_building(); 521 } 522 523 // Create the jstat counters for the policy. 524 void G1CollectorPolicy::initialize_gc_policy_counters() { 525 _gc_policy_counters = new GCPolicyCounters("GarbageFirst", 1, 3); 526 } 527 528 bool G1CollectorPolicy::predict_will_fit(size_t young_length, 529 double base_time_ms, 530 size_t base_free_regions, 531 double target_pause_time_ms) { 532 if (young_length >= base_free_regions) { 533 // end condition 1: not enough space for the young regions 534 return false; 535 } 536 537 double accum_surv_rate = accum_yg_surv_rate_pred((int)(young_length - 1)); 538 size_t bytes_to_copy = 539 (size_t) (accum_surv_rate * (double) HeapRegion::GrainBytes); 540 double copy_time_ms = predict_object_copy_time_ms(bytes_to_copy); 541 double young_other_time_ms = predict_young_other_time_ms(young_length); 542 double pause_time_ms = base_time_ms + copy_time_ms + young_other_time_ms; 543 if (pause_time_ms > target_pause_time_ms) { 544 // end condition 2: prediction is over the target pause time 545 return false; 546 } 547 548 size_t free_bytes = 549 (base_free_regions - young_length) * HeapRegion::GrainBytes; 550 if ((2.0 * sigma()) * (double) bytes_to_copy > (double) free_bytes) { 551 // end condition 3: out-of-space (conservatively!) 552 return false; 553 } 554 555 // success! 556 return true; 557 } 558 559 void G1CollectorPolicy::record_new_heap_size(size_t new_number_of_regions) { 560 // re-calculate the necessary reserve 561 double reserve_regions_d = (double) new_number_of_regions * _reserve_factor; 562 // We use ceiling so that if reserve_regions_d is > 0.0 (but 563 // smaller than 1.0) we'll get 1. 564 _reserve_regions = (size_t) ceil(reserve_regions_d); 565 566 if (_using_new_ratio_calculations) { 567 // -XX:NewRatio was specified so we need to update the 568 // young gen length when the heap size has changed. 569 update_young_list_size_using_newratio(new_number_of_regions); 570 } 571 } 572 573 size_t G1CollectorPolicy::calculate_young_list_desired_min_length( 574 size_t base_min_length) { 575 size_t desired_min_length = 0; 576 if (adaptive_young_list_length()) { 577 if (_alloc_rate_ms_seq->num() > 3) { 578 double now_sec = os::elapsedTime(); 579 double when_ms = _mmu_tracker->when_max_gc_sec(now_sec) * 1000.0; 580 double alloc_rate_ms = predict_alloc_rate_ms(); 581 desired_min_length = (size_t) ceil(alloc_rate_ms * when_ms); 582 } else { 583 // otherwise we don't have enough info to make the prediction 584 } 585 } 586 desired_min_length += base_min_length; 587 // make sure we don't go below any user-defined minimum bound 588 return MAX2(_min_desired_young_length, desired_min_length); 589 } 590 591 size_t G1CollectorPolicy::calculate_young_list_desired_max_length() { 592 // Here, we might want to also take into account any additional 593 // constraints (i.e., user-defined minimum bound). Currently, we 594 // effectively don't set this bound. 595 return _max_desired_young_length; 596 } 597 598 void G1CollectorPolicy::update_young_list_target_length(size_t rs_lengths) { 599 if (rs_lengths == (size_t) -1) { 600 // if it's set to the default value (-1), we should predict it; 601 // otherwise, use the given value. 602 rs_lengths = (size_t) get_new_prediction(_rs_lengths_seq); 603 } 604 605 // Calculate the absolute and desired min bounds. 606 607 // This is how many young regions we already have (currently: the survivors). 608 size_t base_min_length = recorded_survivor_regions(); 609 // This is the absolute minimum young length, which ensures that we 610 // can allocate one eden region in the worst-case. 611 size_t absolute_min_length = base_min_length + 1; 612 size_t desired_min_length = 613 calculate_young_list_desired_min_length(base_min_length); 614 if (desired_min_length < absolute_min_length) { 615 desired_min_length = absolute_min_length; 616 } 617 618 // Calculate the absolute and desired max bounds. 619 620 // We will try our best not to "eat" into the reserve. 621 size_t absolute_max_length = 0; 622 if (_free_regions_at_end_of_collection > _reserve_regions) { 623 absolute_max_length = _free_regions_at_end_of_collection - _reserve_regions; 624 } 625 size_t desired_max_length = calculate_young_list_desired_max_length(); 626 if (desired_max_length > absolute_max_length) { 627 desired_max_length = absolute_max_length; 628 } 629 630 size_t young_list_target_length = 0; 631 if (adaptive_young_list_length()) { 632 if (full_young_gcs()) { 633 young_list_target_length = 634 calculate_young_list_target_length(rs_lengths, 635 base_min_length, 636 desired_min_length, 637 desired_max_length); 638 _rs_lengths_prediction = rs_lengths; 639 } else { 640 // Don't calculate anything and let the code below bound it to 641 // the desired_min_length, i.e., do the next GC as soon as 642 // possible to maximize how many old regions we can add to it. 643 } 644 } else { 645 if (full_young_gcs()) { 646 young_list_target_length = _young_list_fixed_length; 647 } else { 648 // A bit arbitrary: during partially-young GCs we allocate half 649 // the young regions to try to add old regions to the CSet. 650 young_list_target_length = _young_list_fixed_length / 2; 651 // We choose to accept that we might go under the desired min 652 // length given that we intentionally ask for a smaller young gen. 653 desired_min_length = absolute_min_length; 654 } 655 } 656 657 // Make sure we don't go over the desired max length, nor under the 658 // desired min length. In case they clash, desired_min_length wins 659 // which is why that test is second. 660 if (young_list_target_length > desired_max_length) { 661 young_list_target_length = desired_max_length; 662 } 663 if (young_list_target_length < desired_min_length) { 664 young_list_target_length = desired_min_length; 665 } 666 667 assert(young_list_target_length > recorded_survivor_regions(), 668 "we should be able to allocate at least one eden region"); 669 assert(young_list_target_length >= absolute_min_length, "post-condition"); 670 _young_list_target_length = young_list_target_length; 671 672 update_max_gc_locker_expansion(); 673 } 674 675 size_t 676 G1CollectorPolicy::calculate_young_list_target_length(size_t rs_lengths, 677 size_t base_min_length, 678 size_t desired_min_length, 679 size_t desired_max_length) { 680 assert(adaptive_young_list_length(), "pre-condition"); 681 assert(full_young_gcs(), "only call this for fully-young GCs"); 682 683 // In case some edge-condition makes the desired max length too small... 684 if (desired_max_length <= desired_min_length) { 685 return desired_min_length; 686 } 687 688 // We'll adjust min_young_length and max_young_length not to include 689 // the already allocated young regions (i.e., so they reflect the 690 // min and max eden regions we'll allocate). The base_min_length 691 // will be reflected in the predictions by the 692 // survivor_regions_evac_time prediction. 693 assert(desired_min_length > base_min_length, "invariant"); 694 size_t min_young_length = desired_min_length - base_min_length; 695 assert(desired_max_length > base_min_length, "invariant"); 696 size_t max_young_length = desired_max_length - base_min_length; 697 698 double target_pause_time_ms = _mmu_tracker->max_gc_time() * 1000.0; 699 double survivor_regions_evac_time = predict_survivor_regions_evac_time(); 700 size_t pending_cards = (size_t) get_new_prediction(_pending_cards_seq); 701 size_t adj_rs_lengths = rs_lengths + predict_rs_length_diff(); 702 size_t scanned_cards = predict_young_card_num(adj_rs_lengths); 703 double base_time_ms = 704 predict_base_elapsed_time_ms(pending_cards, scanned_cards) + 705 survivor_regions_evac_time; 706 size_t available_free_regions = _free_regions_at_end_of_collection; 707 size_t base_free_regions = 0; 708 if (available_free_regions > _reserve_regions) { 709 base_free_regions = available_free_regions - _reserve_regions; 710 } 711 712 // Here, we will make sure that the shortest young length that 713 // makes sense fits within the target pause time. 714 715 if (predict_will_fit(min_young_length, base_time_ms, 716 base_free_regions, target_pause_time_ms)) { 717 // The shortest young length will fit into the target pause time; 718 // we'll now check whether the absolute maximum number of young 719 // regions will fit in the target pause time. If not, we'll do 720 // a binary search between min_young_length and max_young_length. 721 if (predict_will_fit(max_young_length, base_time_ms, 722 base_free_regions, target_pause_time_ms)) { 723 // The maximum young length will fit into the target pause time. 724 // We are done so set min young length to the maximum length (as 725 // the result is assumed to be returned in min_young_length). 726 min_young_length = max_young_length; 727 } else { 728 // The maximum possible number of young regions will not fit within 729 // the target pause time so we'll search for the optimal 730 // length. The loop invariants are: 731 // 732 // min_young_length < max_young_length 733 // min_young_length is known to fit into the target pause time 734 // max_young_length is known not to fit into the target pause time 735 // 736 // Going into the loop we know the above hold as we've just 737 // checked them. Every time around the loop we check whether 738 // the middle value between min_young_length and 739 // max_young_length fits into the target pause time. If it 740 // does, it becomes the new min. If it doesn't, it becomes 741 // the new max. This way we maintain the loop invariants. 742 743 assert(min_young_length < max_young_length, "invariant"); 744 size_t diff = (max_young_length - min_young_length) / 2; 745 while (diff > 0) { 746 size_t young_length = min_young_length + diff; 747 if (predict_will_fit(young_length, base_time_ms, 748 base_free_regions, target_pause_time_ms)) { 749 min_young_length = young_length; 750 } else { 751 max_young_length = young_length; 752 } 753 assert(min_young_length < max_young_length, "invariant"); 754 diff = (max_young_length - min_young_length) / 2; 755 } 756 // The results is min_young_length which, according to the 757 // loop invariants, should fit within the target pause time. 758 759 // These are the post-conditions of the binary search above: 760 assert(min_young_length < max_young_length, 761 "otherwise we should have discovered that max_young_length " 762 "fits into the pause target and not done the binary search"); 763 assert(predict_will_fit(min_young_length, base_time_ms, 764 base_free_regions, target_pause_time_ms), 765 "min_young_length, the result of the binary search, should " 766 "fit into the pause target"); 767 assert(!predict_will_fit(min_young_length + 1, base_time_ms, 768 base_free_regions, target_pause_time_ms), 769 "min_young_length, the result of the binary search, should be " 770 "optimal, so no larger length should fit into the pause target"); 771 } 772 } else { 773 // Even the minimum length doesn't fit into the pause time 774 // target, return it as the result nevertheless. 775 } 776 return base_min_length + min_young_length; 777 } 778 779 double G1CollectorPolicy::predict_survivor_regions_evac_time() { 780 double survivor_regions_evac_time = 0.0; 781 for (HeapRegion * r = _recorded_survivor_head; 782 r != NULL && r != _recorded_survivor_tail->get_next_young_region(); 783 r = r->get_next_young_region()) { 784 survivor_regions_evac_time += predict_region_elapsed_time_ms(r, true); 785 } 786 return survivor_regions_evac_time; 787 } 788 789 void G1CollectorPolicy::revise_young_list_target_length_if_necessary() { 790 guarantee( adaptive_young_list_length(), "should not call this otherwise" ); 791 792 size_t rs_lengths = _g1->young_list()->sampled_rs_lengths(); 793 if (rs_lengths > _rs_lengths_prediction) { 794 // add 10% to avoid having to recalculate often 795 size_t rs_lengths_prediction = rs_lengths * 1100 / 1000; 796 update_young_list_target_length(rs_lengths_prediction); 797 } 798 } 799 800 801 802 HeapWord* G1CollectorPolicy::mem_allocate_work(size_t size, 803 bool is_tlab, 804 bool* gc_overhead_limit_was_exceeded) { 805 guarantee(false, "Not using this policy feature yet."); 806 return NULL; 807 } 808 809 // This method controls how a collector handles one or more 810 // of its generations being fully allocated. 811 HeapWord* G1CollectorPolicy::satisfy_failed_allocation(size_t size, 812 bool is_tlab) { 813 guarantee(false, "Not using this policy feature yet."); 814 return NULL; 815 } 816 817 818 #ifndef PRODUCT 819 bool G1CollectorPolicy::verify_young_ages() { 820 HeapRegion* head = _g1->young_list()->first_region(); 821 return 822 verify_young_ages(head, _short_lived_surv_rate_group); 823 // also call verify_young_ages on any additional surv rate groups 824 } 825 826 bool 827 G1CollectorPolicy::verify_young_ages(HeapRegion* head, 828 SurvRateGroup *surv_rate_group) { 829 guarantee( surv_rate_group != NULL, "pre-condition" ); 830 831 const char* name = surv_rate_group->name(); 832 bool ret = true; 833 int prev_age = -1; 834 835 for (HeapRegion* curr = head; 836 curr != NULL; 837 curr = curr->get_next_young_region()) { 838 SurvRateGroup* group = curr->surv_rate_group(); 839 if (group == NULL && !curr->is_survivor()) { 840 gclog_or_tty->print_cr("## %s: encountered NULL surv_rate_group", name); 841 ret = false; 842 } 843 844 if (surv_rate_group == group) { 845 int age = curr->age_in_surv_rate_group(); 846 847 if (age < 0) { 848 gclog_or_tty->print_cr("## %s: encountered negative age", name); 849 ret = false; 850 } 851 852 if (age <= prev_age) { 853 gclog_or_tty->print_cr("## %s: region ages are not strictly increasing " 854 "(%d, %d)", name, age, prev_age); 855 ret = false; 856 } 857 prev_age = age; 858 } 859 } 860 861 return ret; 862 } 863 #endif // PRODUCT 864 865 void G1CollectorPolicy::record_full_collection_start() { 866 _cur_collection_start_sec = os::elapsedTime(); 867 // Release the future to-space so that it is available for compaction into. 868 _g1->set_full_collection(); 869 } 870 871 void G1CollectorPolicy::record_full_collection_end() { 872 // Consider this like a collection pause for the purposes of allocation 873 // since last pause. 874 double end_sec = os::elapsedTime(); 875 double full_gc_time_sec = end_sec - _cur_collection_start_sec; 876 double full_gc_time_ms = full_gc_time_sec * 1000.0; 877 878 _all_full_gc_times_ms->add(full_gc_time_ms); 879 880 update_recent_gc_times(end_sec, full_gc_time_ms); 881 882 _g1->clear_full_collection(); 883 884 // "Nuke" the heuristics that control the fully/partially young GC 885 // transitions and make sure we start with fully young GCs after the 886 // Full GC. 887 set_full_young_gcs(true); 888 _last_full_young_gc = false; 889 _should_revert_to_full_young_gcs = false; 890 clear_initiate_conc_mark_if_possible(); 891 clear_during_initial_mark_pause(); 892 _known_garbage_bytes = 0; 893 _known_garbage_ratio = 0.0; 894 _in_marking_window = false; 895 _in_marking_window_im = false; 896 897 _short_lived_surv_rate_group->start_adding_regions(); 898 // also call this on any additional surv rate groups 899 900 record_survivor_regions(0, NULL, NULL); 901 902 _prev_region_num_young = _region_num_young; 903 _prev_region_num_tenured = _region_num_tenured; 904 905 _free_regions_at_end_of_collection = _g1->free_regions(); 906 // Reset survivors SurvRateGroup. 907 _survivor_surv_rate_group->reset(); 908 update_young_list_target_length(); 909 } 910 911 void G1CollectorPolicy::record_stop_world_start() { 912 _stop_world_start = os::elapsedTime(); 913 } 914 915 void G1CollectorPolicy::record_collection_pause_start(double start_time_sec, 916 size_t start_used) { 917 if (PrintGCDetails) { 918 gclog_or_tty->stamp(PrintGCTimeStamps); 919 gclog_or_tty->print("[GC pause"); 920 gclog_or_tty->print(" (%s)", full_young_gcs() ? "young" : "partial"); 921 } 922 923 // We only need to do this here as the policy will only be applied 924 // to the GC we're about to start. so, no point is calculating this 925 // every time we calculate / recalculate the target young length. 926 update_survivors_policy(); 927 928 assert(_g1->used() == _g1->recalculate_used(), 929 err_msg("sanity, used: "SIZE_FORMAT" recalculate_used: "SIZE_FORMAT, 930 _g1->used(), _g1->recalculate_used())); 931 932 double s_w_t_ms = (start_time_sec - _stop_world_start) * 1000.0; 933 _all_stop_world_times_ms->add(s_w_t_ms); 934 _stop_world_start = 0.0; 935 936 _cur_collection_start_sec = start_time_sec; 937 _cur_collection_pause_used_at_start_bytes = start_used; 938 _cur_collection_pause_used_regions_at_start = _g1->used_regions(); 939 _pending_cards = _g1->pending_card_num(); 940 _max_pending_cards = _g1->max_pending_card_num(); 941 942 _bytes_in_collection_set_before_gc = 0; 943 _bytes_copied_during_gc = 0; 944 945 YoungList* young_list = _g1->young_list(); 946 _eden_bytes_before_gc = young_list->eden_used_bytes(); 947 _survivor_bytes_before_gc = young_list->survivor_used_bytes(); 948 _capacity_before_gc = _g1->capacity(); 949 950 #ifdef DEBUG 951 // initialise these to something well known so that we can spot 952 // if they are not set properly 953 954 for (int i = 0; i < _parallel_gc_threads; ++i) { 955 _par_last_gc_worker_start_times_ms[i] = -1234.0; 956 _par_last_ext_root_scan_times_ms[i] = -1234.0; 957 _par_last_mark_stack_scan_times_ms[i] = -1234.0; 958 _par_last_update_rs_times_ms[i] = -1234.0; 959 _par_last_update_rs_processed_buffers[i] = -1234.0; 960 _par_last_scan_rs_times_ms[i] = -1234.0; 961 _par_last_obj_copy_times_ms[i] = -1234.0; 962 _par_last_termination_times_ms[i] = -1234.0; 963 _par_last_termination_attempts[i] = -1234.0; 964 _par_last_gc_worker_end_times_ms[i] = -1234.0; 965 _par_last_gc_worker_times_ms[i] = -1234.0; 966 } 967 #endif 968 969 for (int i = 0; i < _aux_num; ++i) { 970 _cur_aux_times_ms[i] = 0.0; 971 _cur_aux_times_set[i] = false; 972 } 973 974 _satb_drain_time_set = false; 975 _last_satb_drain_processed_buffers = -1; 976 977 _last_young_gc_full = false; 978 979 // do that for any other surv rate groups 980 _short_lived_surv_rate_group->stop_adding_regions(); 981 _survivors_age_table.clear(); 982 983 assert( verify_young_ages(), "region age verification" ); 984 } 985 986 void G1CollectorPolicy::record_mark_closure_time(double mark_closure_time_ms) { 987 _mark_closure_time_ms = mark_closure_time_ms; 988 } 989 990 void G1CollectorPolicy::record_concurrent_mark_init_end(double 991 mark_init_elapsed_time_ms) { 992 _during_marking = true; 993 assert(!initiate_conc_mark_if_possible(), "we should have cleared it by now"); 994 clear_during_initial_mark_pause(); 995 _cur_mark_stop_world_time_ms = mark_init_elapsed_time_ms; 996 } 997 998 void G1CollectorPolicy::record_concurrent_mark_remark_start() { 999 _mark_remark_start_sec = os::elapsedTime(); 1000 _during_marking = false; 1001 } 1002 1003 void G1CollectorPolicy::record_concurrent_mark_remark_end() { 1004 double end_time_sec = os::elapsedTime(); 1005 double elapsed_time_ms = (end_time_sec - _mark_remark_start_sec)*1000.0; 1006 _concurrent_mark_remark_times_ms->add(elapsed_time_ms); 1007 _cur_mark_stop_world_time_ms += elapsed_time_ms; 1008 _prev_collection_pause_end_ms += elapsed_time_ms; 1009 1010 _mmu_tracker->add_pause(_mark_remark_start_sec, end_time_sec, true); 1011 } 1012 1013 void G1CollectorPolicy::record_concurrent_mark_cleanup_start() { 1014 _mark_cleanup_start_sec = os::elapsedTime(); 1015 } 1016 1017 void 1018 G1CollectorPolicy::record_concurrent_mark_cleanup_end(size_t freed_bytes, 1019 size_t max_live_bytes) { 1020 record_concurrent_mark_cleanup_end_work1(freed_bytes, max_live_bytes); 1021 record_concurrent_mark_cleanup_end_work2(); 1022 } 1023 1024 void 1025 G1CollectorPolicy:: 1026 record_concurrent_mark_cleanup_end_work1(size_t freed_bytes, 1027 size_t max_live_bytes) { 1028 if (_n_marks < 2) _n_marks++; 1029 if (G1PolicyVerbose > 0) 1030 gclog_or_tty->print_cr("At end of marking, max_live is " SIZE_FORMAT " MB " 1031 " (of " SIZE_FORMAT " MB heap).", 1032 max_live_bytes/M, _g1->capacity()/M); 1033 } 1034 1035 // The important thing about this is that it includes "os::elapsedTime". 1036 void G1CollectorPolicy::record_concurrent_mark_cleanup_end_work2() { 1037 double end_time_sec = os::elapsedTime(); 1038 double elapsed_time_ms = (end_time_sec - _mark_cleanup_start_sec)*1000.0; 1039 _concurrent_mark_cleanup_times_ms->add(elapsed_time_ms); 1040 _cur_mark_stop_world_time_ms += elapsed_time_ms; 1041 _prev_collection_pause_end_ms += elapsed_time_ms; 1042 1043 _mmu_tracker->add_pause(_mark_cleanup_start_sec, end_time_sec, true); 1044 1045 _num_markings++; 1046 1047 // We did a marking, so reset the "since_last_mark" variables. 1048 double considerConcMarkCost = 1.0; 1049 // If there are available processors, concurrent activity is free... 1050 if (Threads::number_of_non_daemon_threads() * 2 < 1051 os::active_processor_count()) { 1052 considerConcMarkCost = 0.0; 1053 } 1054 _n_pauses_at_mark_end = _n_pauses; 1055 _n_marks_since_last_pause++; 1056 } 1057 1058 void 1059 G1CollectorPolicy::record_concurrent_mark_cleanup_completed() { 1060 _should_revert_to_full_young_gcs = false; 1061 _last_full_young_gc = true; 1062 _in_marking_window = false; 1063 } 1064 1065 void G1CollectorPolicy::record_concurrent_pause() { 1066 if (_stop_world_start > 0.0) { 1067 double yield_ms = (os::elapsedTime() - _stop_world_start) * 1000.0; 1068 _all_yield_times_ms->add(yield_ms); 1069 } 1070 } 1071 1072 void G1CollectorPolicy::record_concurrent_pause_end() { 1073 } 1074 1075 template<class T> 1076 T sum_of(T* sum_arr, int start, int n, int N) { 1077 T sum = (T)0; 1078 for (int i = 0; i < n; i++) { 1079 int j = (start + i) % N; 1080 sum += sum_arr[j]; 1081 } 1082 return sum; 1083 } 1084 1085 void G1CollectorPolicy::print_par_stats(int level, 1086 const char* str, 1087 double* data) { 1088 double min = data[0], max = data[0]; 1089 double total = 0.0; 1090 LineBuffer buf(level); 1091 buf.append("[%s (ms):", str); 1092 for (uint i = 0; i < ParallelGCThreads; ++i) { 1093 double val = data[i]; 1094 if (val < min) 1095 min = val; 1096 if (val > max) 1097 max = val; 1098 total += val; 1099 buf.append(" %3.1lf", val); 1100 } 1101 buf.append_and_print_cr(""); 1102 double avg = total / (double) ParallelGCThreads; 1103 buf.append_and_print_cr(" Avg: %5.1lf, Min: %5.1lf, Max: %5.1lf, Diff: %5.1lf]", 1104 avg, min, max, max - min); 1105 } 1106 1107 void G1CollectorPolicy::print_par_sizes(int level, 1108 const char* str, 1109 double* data) { 1110 double min = data[0], max = data[0]; 1111 double total = 0.0; 1112 LineBuffer buf(level); 1113 buf.append("[%s :", str); 1114 for (uint i = 0; i < ParallelGCThreads; ++i) { 1115 double val = data[i]; 1116 if (val < min) 1117 min = val; 1118 if (val > max) 1119 max = val; 1120 total += val; 1121 buf.append(" %d", (int) val); 1122 } 1123 buf.append_and_print_cr(""); 1124 double avg = total / (double) ParallelGCThreads; 1125 buf.append_and_print_cr(" Sum: %d, Avg: %d, Min: %d, Max: %d, Diff: %d]", 1126 (int)total, (int)avg, (int)min, (int)max, (int)max - (int)min); 1127 } 1128 1129 void G1CollectorPolicy::print_stats (int level, 1130 const char* str, 1131 double value) { 1132 LineBuffer(level).append_and_print_cr("[%s: %5.1lf ms]", str, value); 1133 } 1134 1135 void G1CollectorPolicy::print_stats (int level, 1136 const char* str, 1137 int value) { 1138 LineBuffer(level).append_and_print_cr("[%s: %d]", str, value); 1139 } 1140 1141 double G1CollectorPolicy::avg_value (double* data) { 1142 if (G1CollectedHeap::use_parallel_gc_threads()) { 1143 double ret = 0.0; 1144 for (uint i = 0; i < ParallelGCThreads; ++i) 1145 ret += data[i]; 1146 return ret / (double) ParallelGCThreads; 1147 } else { 1148 return data[0]; 1149 } 1150 } 1151 1152 double G1CollectorPolicy::max_value (double* data) { 1153 if (G1CollectedHeap::use_parallel_gc_threads()) { 1154 double ret = data[0]; 1155 for (uint i = 1; i < ParallelGCThreads; ++i) 1156 if (data[i] > ret) 1157 ret = data[i]; 1158 return ret; 1159 } else { 1160 return data[0]; 1161 } 1162 } 1163 1164 double G1CollectorPolicy::sum_of_values (double* data) { 1165 if (G1CollectedHeap::use_parallel_gc_threads()) { 1166 double sum = 0.0; 1167 for (uint i = 0; i < ParallelGCThreads; i++) 1168 sum += data[i]; 1169 return sum; 1170 } else { 1171 return data[0]; 1172 } 1173 } 1174 1175 double G1CollectorPolicy::max_sum (double* data1, 1176 double* data2) { 1177 double ret = data1[0] + data2[0]; 1178 1179 if (G1CollectedHeap::use_parallel_gc_threads()) { 1180 for (uint i = 1; i < ParallelGCThreads; ++i) { 1181 double data = data1[i] + data2[i]; 1182 if (data > ret) 1183 ret = data; 1184 } 1185 } 1186 return ret; 1187 } 1188 1189 // Anything below that is considered to be zero 1190 #define MIN_TIMER_GRANULARITY 0.0000001 1191 1192 void G1CollectorPolicy::record_collection_pause_end() { 1193 double end_time_sec = os::elapsedTime(); 1194 double elapsed_ms = _last_pause_time_ms; 1195 bool parallel = G1CollectedHeap::use_parallel_gc_threads(); 1196 size_t rs_size = 1197 _cur_collection_pause_used_regions_at_start - collection_set_size(); 1198 size_t cur_used_bytes = _g1->used(); 1199 assert(cur_used_bytes == _g1->recalculate_used(), "It should!"); 1200 bool last_pause_included_initial_mark = false; 1201 bool update_stats = !_g1->evacuation_failed(); 1202 1203 #ifndef PRODUCT 1204 if (G1YoungSurvRateVerbose) { 1205 gclog_or_tty->print_cr(""); 1206 _short_lived_surv_rate_group->print(); 1207 // do that for any other surv rate groups too 1208 } 1209 #endif // PRODUCT 1210 1211 last_pause_included_initial_mark = during_initial_mark_pause(); 1212 if (last_pause_included_initial_mark) 1213 record_concurrent_mark_init_end(0.0); 1214 1215 size_t min_used_targ = 1216 (_g1->capacity() / 100) * InitiatingHeapOccupancyPercent; 1217 1218 1219 if (!_g1->mark_in_progress() && !_last_full_young_gc) { 1220 assert(!last_pause_included_initial_mark, "invariant"); 1221 if (cur_used_bytes > min_used_targ && 1222 cur_used_bytes > _prev_collection_pause_used_at_end_bytes) { 1223 assert(!during_initial_mark_pause(), "we should not see this here"); 1224 1225 // Note: this might have already been set, if during the last 1226 // pause we decided to start a cycle but at the beginning of 1227 // this pause we decided to postpone it. That's OK. 1228 set_initiate_conc_mark_if_possible(); 1229 } 1230 } 1231 1232 _prev_collection_pause_used_at_end_bytes = cur_used_bytes; 1233 1234 _mmu_tracker->add_pause(end_time_sec - elapsed_ms/1000.0, 1235 end_time_sec, false); 1236 1237 guarantee(_cur_collection_pause_used_regions_at_start >= 1238 collection_set_size(), 1239 "Negative RS size?"); 1240 1241 // This assert is exempted when we're doing parallel collection pauses, 1242 // because the fragmentation caused by the parallel GC allocation buffers 1243 // can lead to more memory being used during collection than was used 1244 // before. Best leave this out until the fragmentation problem is fixed. 1245 // Pauses in which evacuation failed can also lead to negative 1246 // collections, since no space is reclaimed from a region containing an 1247 // object whose evacuation failed. 1248 // Further, we're now always doing parallel collection. But I'm still 1249 // leaving this here as a placeholder for a more precise assertion later. 1250 // (DLD, 10/05.) 1251 assert((true || parallel) // Always using GC LABs now. 1252 || _g1->evacuation_failed() 1253 || _cur_collection_pause_used_at_start_bytes >= cur_used_bytes, 1254 "Negative collection"); 1255 1256 size_t freed_bytes = 1257 _cur_collection_pause_used_at_start_bytes - cur_used_bytes; 1258 size_t surviving_bytes = _collection_set_bytes_used_before - freed_bytes; 1259 1260 double survival_fraction = 1261 (double)surviving_bytes/ 1262 (double)_collection_set_bytes_used_before; 1263 1264 _n_pauses++; 1265 1266 double ext_root_scan_time = avg_value(_par_last_ext_root_scan_times_ms); 1267 double mark_stack_scan_time = avg_value(_par_last_mark_stack_scan_times_ms); 1268 double update_rs_time = avg_value(_par_last_update_rs_times_ms); 1269 double update_rs_processed_buffers = 1270 sum_of_values(_par_last_update_rs_processed_buffers); 1271 double scan_rs_time = avg_value(_par_last_scan_rs_times_ms); 1272 double obj_copy_time = avg_value(_par_last_obj_copy_times_ms); 1273 double termination_time = avg_value(_par_last_termination_times_ms); 1274 1275 double parallel_known_time = update_rs_time + 1276 ext_root_scan_time + 1277 mark_stack_scan_time + 1278 scan_rs_time + 1279 obj_copy_time + 1280 termination_time; 1281 1282 double parallel_other_time = _cur_collection_par_time_ms - parallel_known_time; 1283 1284 PauseSummary* summary = _summary; 1285 1286 if (update_stats) { 1287 _recent_rs_scan_times_ms->add(scan_rs_time); 1288 _recent_pause_times_ms->add(elapsed_ms); 1289 _recent_rs_sizes->add(rs_size); 1290 1291 MainBodySummary* body_summary = summary->main_body_summary(); 1292 guarantee(body_summary != NULL, "should not be null!"); 1293 1294 if (_satb_drain_time_set) 1295 body_summary->record_satb_drain_time_ms(_cur_satb_drain_time_ms); 1296 else 1297 body_summary->record_satb_drain_time_ms(0.0); 1298 1299 body_summary->record_ext_root_scan_time_ms(ext_root_scan_time); 1300 body_summary->record_mark_stack_scan_time_ms(mark_stack_scan_time); 1301 body_summary->record_update_rs_time_ms(update_rs_time); 1302 body_summary->record_scan_rs_time_ms(scan_rs_time); 1303 body_summary->record_obj_copy_time_ms(obj_copy_time); 1304 if (parallel) { 1305 body_summary->record_parallel_time_ms(_cur_collection_par_time_ms); 1306 body_summary->record_clear_ct_time_ms(_cur_clear_ct_time_ms); 1307 body_summary->record_termination_time_ms(termination_time); 1308 body_summary->record_parallel_other_time_ms(parallel_other_time); 1309 } 1310 body_summary->record_mark_closure_time_ms(_mark_closure_time_ms); 1311 1312 // We exempt parallel collection from this check because Alloc Buffer 1313 // fragmentation can produce negative collections. Same with evac 1314 // failure. 1315 // Further, we're now always doing parallel collection. But I'm still 1316 // leaving this here as a placeholder for a more precise assertion later. 1317 // (DLD, 10/05. 1318 assert((true || parallel) 1319 || _g1->evacuation_failed() 1320 || surviving_bytes <= _collection_set_bytes_used_before, 1321 "Or else negative collection!"); 1322 _recent_CS_bytes_used_before->add(_collection_set_bytes_used_before); 1323 _recent_CS_bytes_surviving->add(surviving_bytes); 1324 1325 // this is where we update the allocation rate of the application 1326 double app_time_ms = 1327 (_cur_collection_start_sec * 1000.0 - _prev_collection_pause_end_ms); 1328 if (app_time_ms < MIN_TIMER_GRANULARITY) { 1329 // This usually happens due to the timer not having the required 1330 // granularity. Some Linuxes are the usual culprits. 1331 // We'll just set it to something (arbitrarily) small. 1332 app_time_ms = 1.0; 1333 } 1334 size_t regions_allocated = 1335 (_region_num_young - _prev_region_num_young) + 1336 (_region_num_tenured - _prev_region_num_tenured); 1337 double alloc_rate_ms = (double) regions_allocated / app_time_ms; 1338 _alloc_rate_ms_seq->add(alloc_rate_ms); 1339 _prev_region_num_young = _region_num_young; 1340 _prev_region_num_tenured = _region_num_tenured; 1341 1342 double interval_ms = 1343 (end_time_sec - _recent_prev_end_times_for_all_gcs_sec->oldest()) * 1000.0; 1344 update_recent_gc_times(end_time_sec, elapsed_ms); 1345 _recent_avg_pause_time_ratio = _recent_gc_times_ms->sum()/interval_ms; 1346 if (recent_avg_pause_time_ratio() < 0.0 || 1347 (recent_avg_pause_time_ratio() - 1.0 > 0.0)) { 1348 #ifndef PRODUCT 1349 // Dump info to allow post-facto debugging 1350 gclog_or_tty->print_cr("recent_avg_pause_time_ratio() out of bounds"); 1351 gclog_or_tty->print_cr("-------------------------------------------"); 1352 gclog_or_tty->print_cr("Recent GC Times (ms):"); 1353 _recent_gc_times_ms->dump(); 1354 gclog_or_tty->print_cr("(End Time=%3.3f) Recent GC End Times (s):", end_time_sec); 1355 _recent_prev_end_times_for_all_gcs_sec->dump(); 1356 gclog_or_tty->print_cr("GC = %3.3f, Interval = %3.3f, Ratio = %3.3f", 1357 _recent_gc_times_ms->sum(), interval_ms, recent_avg_pause_time_ratio()); 1358 // In debug mode, terminate the JVM if the user wants to debug at this point. 1359 assert(!G1FailOnFPError, "Debugging data for CR 6898948 has been dumped above"); 1360 #endif // !PRODUCT 1361 // Clip ratio between 0.0 and 1.0, and continue. This will be fixed in 1362 // CR 6902692 by redoing the manner in which the ratio is incrementally computed. 1363 if (_recent_avg_pause_time_ratio < 0.0) { 1364 _recent_avg_pause_time_ratio = 0.0; 1365 } else { 1366 assert(_recent_avg_pause_time_ratio - 1.0 > 0.0, "Ctl-point invariant"); 1367 _recent_avg_pause_time_ratio = 1.0; 1368 } 1369 } 1370 } 1371 1372 if (G1PolicyVerbose > 1) { 1373 gclog_or_tty->print_cr(" Recording collection pause(%d)", _n_pauses); 1374 } 1375 1376 if (G1PolicyVerbose > 1) { 1377 gclog_or_tty->print_cr(" ET: %10.6f ms (avg: %10.6f ms)\n" 1378 " ET-RS: %10.6f ms (avg: %10.6f ms)\n" 1379 " |RS|: " SIZE_FORMAT, 1380 elapsed_ms, recent_avg_time_for_pauses_ms(), 1381 scan_rs_time, recent_avg_time_for_rs_scan_ms(), 1382 rs_size); 1383 1384 gclog_or_tty->print_cr(" Used at start: " SIZE_FORMAT"K" 1385 " At end " SIZE_FORMAT "K\n" 1386 " garbage : " SIZE_FORMAT "K" 1387 " of " SIZE_FORMAT "K\n" 1388 " survival : %6.2f%% (%6.2f%% avg)", 1389 _cur_collection_pause_used_at_start_bytes/K, 1390 _g1->used()/K, freed_bytes/K, 1391 _collection_set_bytes_used_before/K, 1392 survival_fraction*100.0, 1393 recent_avg_survival_fraction()*100.0); 1394 gclog_or_tty->print_cr(" Recent %% gc pause time: %6.2f", 1395 recent_avg_pause_time_ratio() * 100.0); 1396 } 1397 1398 double other_time_ms = elapsed_ms; 1399 1400 if (_satb_drain_time_set) { 1401 other_time_ms -= _cur_satb_drain_time_ms; 1402 } 1403 1404 if (parallel) { 1405 other_time_ms -= _cur_collection_par_time_ms + _cur_clear_ct_time_ms; 1406 } else { 1407 other_time_ms -= 1408 update_rs_time + 1409 ext_root_scan_time + mark_stack_scan_time + 1410 scan_rs_time + obj_copy_time; 1411 } 1412 1413 if (PrintGCDetails) { 1414 gclog_or_tty->print_cr("%s, %1.8lf secs]", 1415 (last_pause_included_initial_mark) ? " (initial-mark)" : "", 1416 elapsed_ms / 1000.0); 1417 1418 if (_satb_drain_time_set) { 1419 print_stats(1, "SATB Drain Time", _cur_satb_drain_time_ms); 1420 } 1421 if (_last_satb_drain_processed_buffers >= 0) { 1422 print_stats(2, "Processed Buffers", _last_satb_drain_processed_buffers); 1423 } 1424 if (parallel) { 1425 print_stats(1, "Parallel Time", _cur_collection_par_time_ms); 1426 print_par_stats(2, "GC Worker Start Time", _par_last_gc_worker_start_times_ms); 1427 print_par_stats(2, "Update RS", _par_last_update_rs_times_ms); 1428 print_par_sizes(3, "Processed Buffers", _par_last_update_rs_processed_buffers); 1429 print_par_stats(2, "Ext Root Scanning", _par_last_ext_root_scan_times_ms); 1430 print_par_stats(2, "Mark Stack Scanning", _par_last_mark_stack_scan_times_ms); 1431 print_par_stats(2, "Scan RS", _par_last_scan_rs_times_ms); 1432 print_par_stats(2, "Object Copy", _par_last_obj_copy_times_ms); 1433 print_par_stats(2, "Termination", _par_last_termination_times_ms); 1434 print_par_sizes(3, "Termination Attempts", _par_last_termination_attempts); 1435 print_par_stats(2, "GC Worker End Time", _par_last_gc_worker_end_times_ms); 1436 1437 for (int i = 0; i < _parallel_gc_threads; i++) { 1438 _par_last_gc_worker_times_ms[i] = _par_last_gc_worker_end_times_ms[i] - _par_last_gc_worker_start_times_ms[i]; 1439 } 1440 print_par_stats(2, "GC Worker Times", _par_last_gc_worker_times_ms); 1441 1442 print_stats(2, "Parallel Other", parallel_other_time); 1443 print_stats(1, "Clear CT", _cur_clear_ct_time_ms); 1444 } else { 1445 print_stats(1, "Update RS", update_rs_time); 1446 print_stats(2, "Processed Buffers", 1447 (int)update_rs_processed_buffers); 1448 print_stats(1, "Ext Root Scanning", ext_root_scan_time); 1449 print_stats(1, "Mark Stack Scanning", mark_stack_scan_time); 1450 print_stats(1, "Scan RS", scan_rs_time); 1451 print_stats(1, "Object Copying", obj_copy_time); 1452 } 1453 #ifndef PRODUCT 1454 print_stats(1, "Cur Clear CC", _cur_clear_cc_time_ms); 1455 print_stats(1, "Cum Clear CC", _cum_clear_cc_time_ms); 1456 print_stats(1, "Min Clear CC", _min_clear_cc_time_ms); 1457 print_stats(1, "Max Clear CC", _max_clear_cc_time_ms); 1458 if (_num_cc_clears > 0) { 1459 print_stats(1, "Avg Clear CC", _cum_clear_cc_time_ms / ((double)_num_cc_clears)); 1460 } 1461 #endif 1462 print_stats(1, "Other", other_time_ms); 1463 print_stats(2, "Choose CSet", _recorded_young_cset_choice_time_ms); 1464 1465 for (int i = 0; i < _aux_num; ++i) { 1466 if (_cur_aux_times_set[i]) { 1467 char buffer[96]; 1468 sprintf(buffer, "Aux%d", i); 1469 print_stats(1, buffer, _cur_aux_times_ms[i]); 1470 } 1471 } 1472 } 1473 1474 _all_pause_times_ms->add(elapsed_ms); 1475 if (update_stats) { 1476 summary->record_total_time_ms(elapsed_ms); 1477 summary->record_other_time_ms(other_time_ms); 1478 } 1479 for (int i = 0; i < _aux_num; ++i) 1480 if (_cur_aux_times_set[i]) 1481 _all_aux_times_ms[i].add(_cur_aux_times_ms[i]); 1482 1483 // Reset marks-between-pauses counter. 1484 _n_marks_since_last_pause = 0; 1485 1486 // Update the efficiency-since-mark vars. 1487 double proc_ms = elapsed_ms * (double) _parallel_gc_threads; 1488 if (elapsed_ms < MIN_TIMER_GRANULARITY) { 1489 // This usually happens due to the timer not having the required 1490 // granularity. Some Linuxes are the usual culprits. 1491 // We'll just set it to something (arbitrarily) small. 1492 proc_ms = 1.0; 1493 } 1494 double cur_efficiency = (double) freed_bytes / proc_ms; 1495 1496 bool new_in_marking_window = _in_marking_window; 1497 bool new_in_marking_window_im = false; 1498 if (during_initial_mark_pause()) { 1499 new_in_marking_window = true; 1500 new_in_marking_window_im = true; 1501 } 1502 1503 if (_last_full_young_gc) { 1504 set_full_young_gcs(false); 1505 _last_full_young_gc = false; 1506 } 1507 1508 if ( !_last_young_gc_full ) { 1509 if ( _should_revert_to_full_young_gcs || 1510 _known_garbage_ratio < 0.05 || 1511 (adaptive_young_list_length() && 1512 (get_gc_eff_factor() * cur_efficiency < predict_young_gc_eff())) ) { 1513 set_full_young_gcs(true); 1514 } 1515 } 1516 _should_revert_to_full_young_gcs = false; 1517 1518 if (_last_young_gc_full && !_during_marking) { 1519 _young_gc_eff_seq->add(cur_efficiency); 1520 } 1521 1522 _short_lived_surv_rate_group->start_adding_regions(); 1523 // do that for any other surv rate groupsx 1524 1525 // <NEW PREDICTION> 1526 1527 if (update_stats) { 1528 double pause_time_ms = elapsed_ms; 1529 1530 size_t diff = 0; 1531 if (_max_pending_cards >= _pending_cards) 1532 diff = _max_pending_cards - _pending_cards; 1533 _pending_card_diff_seq->add((double) diff); 1534 1535 double cost_per_card_ms = 0.0; 1536 if (_pending_cards > 0) { 1537 cost_per_card_ms = update_rs_time / (double) _pending_cards; 1538 _cost_per_card_ms_seq->add(cost_per_card_ms); 1539 } 1540 1541 size_t cards_scanned = _g1->cards_scanned(); 1542 1543 double cost_per_entry_ms = 0.0; 1544 if (cards_scanned > 10) { 1545 cost_per_entry_ms = scan_rs_time / (double) cards_scanned; 1546 if (_last_young_gc_full) 1547 _cost_per_entry_ms_seq->add(cost_per_entry_ms); 1548 else 1549 _partially_young_cost_per_entry_ms_seq->add(cost_per_entry_ms); 1550 } 1551 1552 if (_max_rs_lengths > 0) { 1553 double cards_per_entry_ratio = 1554 (double) cards_scanned / (double) _max_rs_lengths; 1555 if (_last_young_gc_full) 1556 _fully_young_cards_per_entry_ratio_seq->add(cards_per_entry_ratio); 1557 else 1558 _partially_young_cards_per_entry_ratio_seq->add(cards_per_entry_ratio); 1559 } 1560 1561 size_t rs_length_diff = _max_rs_lengths - _recorded_rs_lengths; 1562 if (rs_length_diff >= 0) 1563 _rs_length_diff_seq->add((double) rs_length_diff); 1564 1565 size_t copied_bytes = surviving_bytes; 1566 double cost_per_byte_ms = 0.0; 1567 if (copied_bytes > 0) { 1568 cost_per_byte_ms = obj_copy_time / (double) copied_bytes; 1569 if (_in_marking_window) 1570 _cost_per_byte_ms_during_cm_seq->add(cost_per_byte_ms); 1571 else 1572 _cost_per_byte_ms_seq->add(cost_per_byte_ms); 1573 } 1574 1575 double all_other_time_ms = pause_time_ms - 1576 (update_rs_time + scan_rs_time + obj_copy_time + 1577 _mark_closure_time_ms + termination_time); 1578 1579 double young_other_time_ms = 0.0; 1580 if (_recorded_young_regions > 0) { 1581 young_other_time_ms = 1582 _recorded_young_cset_choice_time_ms + 1583 _recorded_young_free_cset_time_ms; 1584 _young_other_cost_per_region_ms_seq->add(young_other_time_ms / 1585 (double) _recorded_young_regions); 1586 } 1587 double non_young_other_time_ms = 0.0; 1588 if (_recorded_non_young_regions > 0) { 1589 non_young_other_time_ms = 1590 _recorded_non_young_cset_choice_time_ms + 1591 _recorded_non_young_free_cset_time_ms; 1592 1593 _non_young_other_cost_per_region_ms_seq->add(non_young_other_time_ms / 1594 (double) _recorded_non_young_regions); 1595 } 1596 1597 double constant_other_time_ms = all_other_time_ms - 1598 (young_other_time_ms + non_young_other_time_ms); 1599 _constant_other_time_ms_seq->add(constant_other_time_ms); 1600 1601 double survival_ratio = 0.0; 1602 if (_bytes_in_collection_set_before_gc > 0) { 1603 survival_ratio = (double) _bytes_copied_during_gc / 1604 (double) _bytes_in_collection_set_before_gc; 1605 } 1606 1607 _pending_cards_seq->add((double) _pending_cards); 1608 _scanned_cards_seq->add((double) cards_scanned); 1609 _rs_lengths_seq->add((double) _max_rs_lengths); 1610 1611 double expensive_region_limit_ms = 1612 (double) MaxGCPauseMillis - predict_constant_other_time_ms(); 1613 if (expensive_region_limit_ms < 0.0) { 1614 // this means that the other time was predicted to be longer than 1615 // than the max pause time 1616 expensive_region_limit_ms = (double) MaxGCPauseMillis; 1617 } 1618 _expensive_region_limit_ms = expensive_region_limit_ms; 1619 1620 if (PREDICTIONS_VERBOSE) { 1621 gclog_or_tty->print_cr(""); 1622 gclog_or_tty->print_cr("PREDICTIONS %1.4lf %d " 1623 "REGIONS %d %d %d " 1624 "PENDING_CARDS %d %d " 1625 "CARDS_SCANNED %d %d " 1626 "RS_LENGTHS %d %d " 1627 "RS_UPDATE %1.6lf %1.6lf RS_SCAN %1.6lf %1.6lf " 1628 "SURVIVAL_RATIO %1.6lf %1.6lf " 1629 "OBJECT_COPY %1.6lf %1.6lf OTHER_CONSTANT %1.6lf %1.6lf " 1630 "OTHER_YOUNG %1.6lf %1.6lf " 1631 "OTHER_NON_YOUNG %1.6lf %1.6lf " 1632 "VTIME_DIFF %1.6lf TERMINATION %1.6lf " 1633 "ELAPSED %1.6lf %1.6lf ", 1634 _cur_collection_start_sec, 1635 (!_last_young_gc_full) ? 2 : 1636 (last_pause_included_initial_mark) ? 1 : 0, 1637 _recorded_region_num, 1638 _recorded_young_regions, 1639 _recorded_non_young_regions, 1640 _predicted_pending_cards, _pending_cards, 1641 _predicted_cards_scanned, cards_scanned, 1642 _predicted_rs_lengths, _max_rs_lengths, 1643 _predicted_rs_update_time_ms, update_rs_time, 1644 _predicted_rs_scan_time_ms, scan_rs_time, 1645 _predicted_survival_ratio, survival_ratio, 1646 _predicted_object_copy_time_ms, obj_copy_time, 1647 _predicted_constant_other_time_ms, constant_other_time_ms, 1648 _predicted_young_other_time_ms, young_other_time_ms, 1649 _predicted_non_young_other_time_ms, 1650 non_young_other_time_ms, 1651 _vtime_diff_ms, termination_time, 1652 _predicted_pause_time_ms, elapsed_ms); 1653 } 1654 1655 if (G1PolicyVerbose > 0) { 1656 gclog_or_tty->print_cr("Pause Time, predicted: %1.4lfms (predicted %s), actual: %1.4lfms", 1657 _predicted_pause_time_ms, 1658 (_within_target) ? "within" : "outside", 1659 elapsed_ms); 1660 } 1661 1662 } 1663 1664 _in_marking_window = new_in_marking_window; 1665 _in_marking_window_im = new_in_marking_window_im; 1666 _free_regions_at_end_of_collection = _g1->free_regions(); 1667 update_young_list_target_length(); 1668 1669 // Note that _mmu_tracker->max_gc_time() returns the time in seconds. 1670 double update_rs_time_goal_ms = _mmu_tracker->max_gc_time() * MILLIUNITS * G1RSetUpdatingPauseTimePercent / 100.0; 1671 adjust_concurrent_refinement(update_rs_time, update_rs_processed_buffers, update_rs_time_goal_ms); 1672 // </NEW PREDICTION> 1673 } 1674 1675 #define EXT_SIZE_FORMAT "%d%s" 1676 #define EXT_SIZE_PARAMS(bytes) \ 1677 byte_size_in_proper_unit((bytes)), \ 1678 proper_unit_for_byte_size((bytes)) 1679 1680 void G1CollectorPolicy::print_heap_transition() { 1681 if (PrintGCDetails) { 1682 YoungList* young_list = _g1->young_list(); 1683 size_t eden_bytes = young_list->eden_used_bytes(); 1684 size_t survivor_bytes = young_list->survivor_used_bytes(); 1685 size_t used_before_gc = _cur_collection_pause_used_at_start_bytes; 1686 size_t used = _g1->used(); 1687 size_t capacity = _g1->capacity(); 1688 size_t eden_capacity = 1689 (_young_list_target_length * HeapRegion::GrainBytes) - survivor_bytes; 1690 1691 gclog_or_tty->print_cr( 1692 " [Eden: "EXT_SIZE_FORMAT"("EXT_SIZE_FORMAT")->"EXT_SIZE_FORMAT"("EXT_SIZE_FORMAT") " 1693 "Survivors: "EXT_SIZE_FORMAT"->"EXT_SIZE_FORMAT" " 1694 "Heap: "EXT_SIZE_FORMAT"("EXT_SIZE_FORMAT")->" 1695 EXT_SIZE_FORMAT"("EXT_SIZE_FORMAT")]", 1696 EXT_SIZE_PARAMS(_eden_bytes_before_gc), 1697 EXT_SIZE_PARAMS(_prev_eden_capacity), 1698 EXT_SIZE_PARAMS(eden_bytes), 1699 EXT_SIZE_PARAMS(eden_capacity), 1700 EXT_SIZE_PARAMS(_survivor_bytes_before_gc), 1701 EXT_SIZE_PARAMS(survivor_bytes), 1702 EXT_SIZE_PARAMS(used_before_gc), 1703 EXT_SIZE_PARAMS(_capacity_before_gc), 1704 EXT_SIZE_PARAMS(used), 1705 EXT_SIZE_PARAMS(capacity)); 1706 1707 _prev_eden_capacity = eden_capacity; 1708 } else if (PrintGC) { 1709 _g1->print_size_transition(gclog_or_tty, 1710 _cur_collection_pause_used_at_start_bytes, 1711 _g1->used(), _g1->capacity()); 1712 } 1713 } 1714 1715 // <NEW PREDICTION> 1716 1717 void G1CollectorPolicy::adjust_concurrent_refinement(double update_rs_time, 1718 double update_rs_processed_buffers, 1719 double goal_ms) { 1720 DirtyCardQueueSet& dcqs = JavaThread::dirty_card_queue_set(); 1721 ConcurrentG1Refine *cg1r = G1CollectedHeap::heap()->concurrent_g1_refine(); 1722 1723 if (G1UseAdaptiveConcRefinement) { 1724 const int k_gy = 3, k_gr = 6; 1725 const double inc_k = 1.1, dec_k = 0.9; 1726 1727 int g = cg1r->green_zone(); 1728 if (update_rs_time > goal_ms) { 1729 g = (int)(g * dec_k); // Can become 0, that's OK. That would mean a mutator-only processing. 1730 } else { 1731 if (update_rs_time < goal_ms && update_rs_processed_buffers > g) { 1732 g = (int)MAX2(g * inc_k, g + 1.0); 1733 } 1734 } 1735 // Change the refinement threads params 1736 cg1r->set_green_zone(g); 1737 cg1r->set_yellow_zone(g * k_gy); 1738 cg1r->set_red_zone(g * k_gr); 1739 cg1r->reinitialize_threads(); 1740 1741 int processing_threshold_delta = MAX2((int)(cg1r->green_zone() * sigma()), 1); 1742 int processing_threshold = MIN2(cg1r->green_zone() + processing_threshold_delta, 1743 cg1r->yellow_zone()); 1744 // Change the barrier params 1745 dcqs.set_process_completed_threshold(processing_threshold); 1746 dcqs.set_max_completed_queue(cg1r->red_zone()); 1747 } 1748 1749 int curr_queue_size = dcqs.completed_buffers_num(); 1750 if (curr_queue_size >= cg1r->yellow_zone()) { 1751 dcqs.set_completed_queue_padding(curr_queue_size); 1752 } else { 1753 dcqs.set_completed_queue_padding(0); 1754 } 1755 dcqs.notify_if_necessary(); 1756 } 1757 1758 double 1759 G1CollectorPolicy:: 1760 predict_young_collection_elapsed_time_ms(size_t adjustment) { 1761 guarantee( adjustment == 0 || adjustment == 1, "invariant" ); 1762 1763 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 1764 size_t young_num = g1h->young_list()->length(); 1765 if (young_num == 0) 1766 return 0.0; 1767 1768 young_num += adjustment; 1769 size_t pending_cards = predict_pending_cards(); 1770 size_t rs_lengths = g1h->young_list()->sampled_rs_lengths() + 1771 predict_rs_length_diff(); 1772 size_t card_num; 1773 if (full_young_gcs()) 1774 card_num = predict_young_card_num(rs_lengths); 1775 else 1776 card_num = predict_non_young_card_num(rs_lengths); 1777 size_t young_byte_size = young_num * HeapRegion::GrainBytes; 1778 double accum_yg_surv_rate = 1779 _short_lived_surv_rate_group->accum_surv_rate(adjustment); 1780 1781 size_t bytes_to_copy = 1782 (size_t) (accum_yg_surv_rate * (double) HeapRegion::GrainBytes); 1783 1784 return 1785 predict_rs_update_time_ms(pending_cards) + 1786 predict_rs_scan_time_ms(card_num) + 1787 predict_object_copy_time_ms(bytes_to_copy) + 1788 predict_young_other_time_ms(young_num) + 1789 predict_constant_other_time_ms(); 1790 } 1791 1792 double 1793 G1CollectorPolicy::predict_base_elapsed_time_ms(size_t pending_cards) { 1794 size_t rs_length = predict_rs_length_diff(); 1795 size_t card_num; 1796 if (full_young_gcs()) 1797 card_num = predict_young_card_num(rs_length); 1798 else 1799 card_num = predict_non_young_card_num(rs_length); 1800 return predict_base_elapsed_time_ms(pending_cards, card_num); 1801 } 1802 1803 double 1804 G1CollectorPolicy::predict_base_elapsed_time_ms(size_t pending_cards, 1805 size_t scanned_cards) { 1806 return 1807 predict_rs_update_time_ms(pending_cards) + 1808 predict_rs_scan_time_ms(scanned_cards) + 1809 predict_constant_other_time_ms(); 1810 } 1811 1812 double 1813 G1CollectorPolicy::predict_region_elapsed_time_ms(HeapRegion* hr, 1814 bool young) { 1815 size_t rs_length = hr->rem_set()->occupied(); 1816 size_t card_num; 1817 if (full_young_gcs()) 1818 card_num = predict_young_card_num(rs_length); 1819 else 1820 card_num = predict_non_young_card_num(rs_length); 1821 size_t bytes_to_copy = predict_bytes_to_copy(hr); 1822 1823 double region_elapsed_time_ms = 1824 predict_rs_scan_time_ms(card_num) + 1825 predict_object_copy_time_ms(bytes_to_copy); 1826 1827 if (young) 1828 region_elapsed_time_ms += predict_young_other_time_ms(1); 1829 else 1830 region_elapsed_time_ms += predict_non_young_other_time_ms(1); 1831 1832 return region_elapsed_time_ms; 1833 } 1834 1835 size_t 1836 G1CollectorPolicy::predict_bytes_to_copy(HeapRegion* hr) { 1837 size_t bytes_to_copy; 1838 if (hr->is_marked()) 1839 bytes_to_copy = hr->max_live_bytes(); 1840 else { 1841 guarantee( hr->is_young() && hr->age_in_surv_rate_group() != -1, 1842 "invariant" ); 1843 int age = hr->age_in_surv_rate_group(); 1844 double yg_surv_rate = predict_yg_surv_rate(age, hr->surv_rate_group()); 1845 bytes_to_copy = (size_t) ((double) hr->used() * yg_surv_rate); 1846 } 1847 1848 return bytes_to_copy; 1849 } 1850 1851 void 1852 G1CollectorPolicy::start_recording_regions() { 1853 _recorded_rs_lengths = 0; 1854 _recorded_young_regions = 0; 1855 _recorded_non_young_regions = 0; 1856 1857 #if PREDICTIONS_VERBOSE 1858 _recorded_marked_bytes = 0; 1859 _recorded_young_bytes = 0; 1860 _predicted_bytes_to_copy = 0; 1861 _predicted_rs_lengths = 0; 1862 _predicted_cards_scanned = 0; 1863 #endif // PREDICTIONS_VERBOSE 1864 } 1865 1866 void 1867 G1CollectorPolicy::record_cset_region_info(HeapRegion* hr, bool young) { 1868 #if PREDICTIONS_VERBOSE 1869 if (!young) { 1870 _recorded_marked_bytes += hr->max_live_bytes(); 1871 } 1872 _predicted_bytes_to_copy += predict_bytes_to_copy(hr); 1873 #endif // PREDICTIONS_VERBOSE 1874 1875 size_t rs_length = hr->rem_set()->occupied(); 1876 _recorded_rs_lengths += rs_length; 1877 } 1878 1879 void 1880 G1CollectorPolicy::record_non_young_cset_region(HeapRegion* hr) { 1881 assert(!hr->is_young(), "should not call this"); 1882 ++_recorded_non_young_regions; 1883 record_cset_region_info(hr, false); 1884 } 1885 1886 void 1887 G1CollectorPolicy::set_recorded_young_regions(size_t n_regions) { 1888 _recorded_young_regions = n_regions; 1889 } 1890 1891 void G1CollectorPolicy::set_recorded_young_bytes(size_t bytes) { 1892 #if PREDICTIONS_VERBOSE 1893 _recorded_young_bytes = bytes; 1894 #endif // PREDICTIONS_VERBOSE 1895 } 1896 1897 void G1CollectorPolicy::set_recorded_rs_lengths(size_t rs_lengths) { 1898 _recorded_rs_lengths = rs_lengths; 1899 } 1900 1901 void G1CollectorPolicy::set_predicted_bytes_to_copy(size_t bytes) { 1902 _predicted_bytes_to_copy = bytes; 1903 } 1904 1905 void 1906 G1CollectorPolicy::end_recording_regions() { 1907 // The _predicted_pause_time_ms field is referenced in code 1908 // not under PREDICTIONS_VERBOSE. Let's initialize it. 1909 _predicted_pause_time_ms = -1.0; 1910 1911 #if PREDICTIONS_VERBOSE 1912 _predicted_pending_cards = predict_pending_cards(); 1913 _predicted_rs_lengths = _recorded_rs_lengths + predict_rs_length_diff(); 1914 if (full_young_gcs()) 1915 _predicted_cards_scanned += predict_young_card_num(_predicted_rs_lengths); 1916 else 1917 _predicted_cards_scanned += 1918 predict_non_young_card_num(_predicted_rs_lengths); 1919 _recorded_region_num = _recorded_young_regions + _recorded_non_young_regions; 1920 1921 _predicted_rs_update_time_ms = 1922 predict_rs_update_time_ms(_g1->pending_card_num()); 1923 _predicted_rs_scan_time_ms = 1924 predict_rs_scan_time_ms(_predicted_cards_scanned); 1925 _predicted_object_copy_time_ms = 1926 predict_object_copy_time_ms(_predicted_bytes_to_copy); 1927 _predicted_constant_other_time_ms = 1928 predict_constant_other_time_ms(); 1929 _predicted_young_other_time_ms = 1930 predict_young_other_time_ms(_recorded_young_regions); 1931 _predicted_non_young_other_time_ms = 1932 predict_non_young_other_time_ms(_recorded_non_young_regions); 1933 1934 _predicted_pause_time_ms = 1935 _predicted_rs_update_time_ms + 1936 _predicted_rs_scan_time_ms + 1937 _predicted_object_copy_time_ms + 1938 _predicted_constant_other_time_ms + 1939 _predicted_young_other_time_ms + 1940 _predicted_non_young_other_time_ms; 1941 #endif // PREDICTIONS_VERBOSE 1942 } 1943 1944 void G1CollectorPolicy::check_if_region_is_too_expensive(double 1945 predicted_time_ms) { 1946 // I don't think we need to do this when in young GC mode since 1947 // marking will be initiated next time we hit the soft limit anyway... 1948 if (predicted_time_ms > _expensive_region_limit_ms) { 1949 // no point in doing another partial one 1950 _should_revert_to_full_young_gcs = true; 1951 } 1952 } 1953 1954 // </NEW PREDICTION> 1955 1956 1957 void G1CollectorPolicy::update_recent_gc_times(double end_time_sec, 1958 double elapsed_ms) { 1959 _recent_gc_times_ms->add(elapsed_ms); 1960 _recent_prev_end_times_for_all_gcs_sec->add(end_time_sec); 1961 _prev_collection_pause_end_ms = end_time_sec * 1000.0; 1962 } 1963 1964 double G1CollectorPolicy::recent_avg_time_for_pauses_ms() { 1965 if (_recent_pause_times_ms->num() == 0) { 1966 return (double) MaxGCPauseMillis; 1967 } 1968 return _recent_pause_times_ms->avg(); 1969 } 1970 1971 double G1CollectorPolicy::recent_avg_time_for_rs_scan_ms() { 1972 if (_recent_rs_scan_times_ms->num() == 0) { 1973 return (double)MaxGCPauseMillis/3.0; 1974 } 1975 return _recent_rs_scan_times_ms->avg(); 1976 } 1977 1978 int G1CollectorPolicy::number_of_recent_gcs() { 1979 assert(_recent_rs_scan_times_ms->num() == 1980 _recent_pause_times_ms->num(), "Sequence out of sync"); 1981 assert(_recent_pause_times_ms->num() == 1982 _recent_CS_bytes_used_before->num(), "Sequence out of sync"); 1983 assert(_recent_CS_bytes_used_before->num() == 1984 _recent_CS_bytes_surviving->num(), "Sequence out of sync"); 1985 1986 return _recent_pause_times_ms->num(); 1987 } 1988 1989 double G1CollectorPolicy::recent_avg_survival_fraction() { 1990 return recent_avg_survival_fraction_work(_recent_CS_bytes_surviving, 1991 _recent_CS_bytes_used_before); 1992 } 1993 1994 double G1CollectorPolicy::last_survival_fraction() { 1995 return last_survival_fraction_work(_recent_CS_bytes_surviving, 1996 _recent_CS_bytes_used_before); 1997 } 1998 1999 double 2000 G1CollectorPolicy::recent_avg_survival_fraction_work(TruncatedSeq* surviving, 2001 TruncatedSeq* before) { 2002 assert(surviving->num() == before->num(), "Sequence out of sync"); 2003 if (before->sum() > 0.0) { 2004 double recent_survival_rate = surviving->sum() / before->sum(); 2005 // We exempt parallel collection from this check because Alloc Buffer 2006 // fragmentation can produce negative collections. 2007 // Further, we're now always doing parallel collection. But I'm still 2008 // leaving this here as a placeholder for a more precise assertion later. 2009 // (DLD, 10/05.) 2010 assert((true || G1CollectedHeap::use_parallel_gc_threads()) || 2011 _g1->evacuation_failed() || 2012 recent_survival_rate <= 1.0, "Or bad frac"); 2013 return recent_survival_rate; 2014 } else { 2015 return 1.0; // Be conservative. 2016 } 2017 } 2018 2019 double 2020 G1CollectorPolicy::last_survival_fraction_work(TruncatedSeq* surviving, 2021 TruncatedSeq* before) { 2022 assert(surviving->num() == before->num(), "Sequence out of sync"); 2023 if (surviving->num() > 0 && before->last() > 0.0) { 2024 double last_survival_rate = surviving->last() / before->last(); 2025 // We exempt parallel collection from this check because Alloc Buffer 2026 // fragmentation can produce negative collections. 2027 // Further, we're now always doing parallel collection. But I'm still 2028 // leaving this here as a placeholder for a more precise assertion later. 2029 // (DLD, 10/05.) 2030 assert((true || G1CollectedHeap::use_parallel_gc_threads()) || 2031 last_survival_rate <= 1.0, "Or bad frac"); 2032 return last_survival_rate; 2033 } else { 2034 return 1.0; 2035 } 2036 } 2037 2038 static const int survival_min_obs = 5; 2039 static double survival_min_obs_limits[] = { 0.9, 0.7, 0.5, 0.3, 0.1 }; 2040 static const double min_survival_rate = 0.1; 2041 2042 double 2043 G1CollectorPolicy::conservative_avg_survival_fraction_work(double avg, 2044 double latest) { 2045 double res = avg; 2046 if (number_of_recent_gcs() < survival_min_obs) { 2047 res = MAX2(res, survival_min_obs_limits[number_of_recent_gcs()]); 2048 } 2049 res = MAX2(res, latest); 2050 res = MAX2(res, min_survival_rate); 2051 // In the parallel case, LAB fragmentation can produce "negative 2052 // collections"; so can evac failure. Cap at 1.0 2053 res = MIN2(res, 1.0); 2054 return res; 2055 } 2056 2057 size_t G1CollectorPolicy::expansion_amount() { 2058 if ((recent_avg_pause_time_ratio() * 100.0) > _gc_overhead_perc) { 2059 // We will double the existing space, or take 2060 // G1ExpandByPercentOfAvailable % of the available expansion 2061 // space, whichever is smaller, bounded below by a minimum 2062 // expansion (unless that's all that's left.) 2063 const size_t min_expand_bytes = 1*M; 2064 size_t reserved_bytes = _g1->max_capacity(); 2065 size_t committed_bytes = _g1->capacity(); 2066 size_t uncommitted_bytes = reserved_bytes - committed_bytes; 2067 size_t expand_bytes; 2068 size_t expand_bytes_via_pct = 2069 uncommitted_bytes * G1ExpandByPercentOfAvailable / 100; 2070 expand_bytes = MIN2(expand_bytes_via_pct, committed_bytes); 2071 expand_bytes = MAX2(expand_bytes, min_expand_bytes); 2072 expand_bytes = MIN2(expand_bytes, uncommitted_bytes); 2073 if (G1PolicyVerbose > 1) { 2074 gclog_or_tty->print("Decided to expand: ratio = %5.2f, " 2075 "committed = %d%s, uncommited = %d%s, via pct = %d%s.\n" 2076 " Answer = %d.\n", 2077 recent_avg_pause_time_ratio(), 2078 byte_size_in_proper_unit(committed_bytes), 2079 proper_unit_for_byte_size(committed_bytes), 2080 byte_size_in_proper_unit(uncommitted_bytes), 2081 proper_unit_for_byte_size(uncommitted_bytes), 2082 byte_size_in_proper_unit(expand_bytes_via_pct), 2083 proper_unit_for_byte_size(expand_bytes_via_pct), 2084 byte_size_in_proper_unit(expand_bytes), 2085 proper_unit_for_byte_size(expand_bytes)); 2086 } 2087 return expand_bytes; 2088 } else { 2089 return 0; 2090 } 2091 } 2092 2093 void G1CollectorPolicy::note_start_of_mark_thread() { 2094 _mark_thread_startup_sec = os::elapsedTime(); 2095 } 2096 2097 class CountCSClosure: public HeapRegionClosure { 2098 G1CollectorPolicy* _g1_policy; 2099 public: 2100 CountCSClosure(G1CollectorPolicy* g1_policy) : 2101 _g1_policy(g1_policy) {} 2102 bool doHeapRegion(HeapRegion* r) { 2103 _g1_policy->_bytes_in_collection_set_before_gc += r->used(); 2104 return false; 2105 } 2106 }; 2107 2108 void G1CollectorPolicy::count_CS_bytes_used() { 2109 CountCSClosure cs_closure(this); 2110 _g1->collection_set_iterate(&cs_closure); 2111 } 2112 2113 void G1CollectorPolicy::print_summary (int level, 2114 const char* str, 2115 NumberSeq* seq) const { 2116 double sum = seq->sum(); 2117 LineBuffer(level + 1).append_and_print_cr("%-24s = %8.2lf s (avg = %8.2lf ms)", 2118 str, sum / 1000.0, seq->avg()); 2119 } 2120 2121 void G1CollectorPolicy::print_summary_sd (int level, 2122 const char* str, 2123 NumberSeq* seq) const { 2124 print_summary(level, str, seq); 2125 LineBuffer(level + 6).append_and_print_cr("(num = %5d, std dev = %8.2lf ms, max = %8.2lf ms)", 2126 seq->num(), seq->sd(), seq->maximum()); 2127 } 2128 2129 void G1CollectorPolicy::check_other_times(int level, 2130 NumberSeq* other_times_ms, 2131 NumberSeq* calc_other_times_ms) const { 2132 bool should_print = false; 2133 LineBuffer buf(level + 2); 2134 2135 double max_sum = MAX2(fabs(other_times_ms->sum()), 2136 fabs(calc_other_times_ms->sum())); 2137 double min_sum = MIN2(fabs(other_times_ms->sum()), 2138 fabs(calc_other_times_ms->sum())); 2139 double sum_ratio = max_sum / min_sum; 2140 if (sum_ratio > 1.1) { 2141 should_print = true; 2142 buf.append_and_print_cr("## CALCULATED OTHER SUM DOESN'T MATCH RECORDED ###"); 2143 } 2144 2145 double max_avg = MAX2(fabs(other_times_ms->avg()), 2146 fabs(calc_other_times_ms->avg())); 2147 double min_avg = MIN2(fabs(other_times_ms->avg()), 2148 fabs(calc_other_times_ms->avg())); 2149 double avg_ratio = max_avg / min_avg; 2150 if (avg_ratio > 1.1) { 2151 should_print = true; 2152 buf.append_and_print_cr("## CALCULATED OTHER AVG DOESN'T MATCH RECORDED ###"); 2153 } 2154 2155 if (other_times_ms->sum() < -0.01) { 2156 buf.append_and_print_cr("## RECORDED OTHER SUM IS NEGATIVE ###"); 2157 } 2158 2159 if (other_times_ms->avg() < -0.01) { 2160 buf.append_and_print_cr("## RECORDED OTHER AVG IS NEGATIVE ###"); 2161 } 2162 2163 if (calc_other_times_ms->sum() < -0.01) { 2164 should_print = true; 2165 buf.append_and_print_cr("## CALCULATED OTHER SUM IS NEGATIVE ###"); 2166 } 2167 2168 if (calc_other_times_ms->avg() < -0.01) { 2169 should_print = true; 2170 buf.append_and_print_cr("## CALCULATED OTHER AVG IS NEGATIVE ###"); 2171 } 2172 2173 if (should_print) 2174 print_summary(level, "Other(Calc)", calc_other_times_ms); 2175 } 2176 2177 void G1CollectorPolicy::print_summary(PauseSummary* summary) const { 2178 bool parallel = G1CollectedHeap::use_parallel_gc_threads(); 2179 MainBodySummary* body_summary = summary->main_body_summary(); 2180 if (summary->get_total_seq()->num() > 0) { 2181 print_summary_sd(0, "Evacuation Pauses", summary->get_total_seq()); 2182 if (body_summary != NULL) { 2183 print_summary(1, "SATB Drain", body_summary->get_satb_drain_seq()); 2184 if (parallel) { 2185 print_summary(1, "Parallel Time", body_summary->get_parallel_seq()); 2186 print_summary(2, "Update RS", body_summary->get_update_rs_seq()); 2187 print_summary(2, "Ext Root Scanning", 2188 body_summary->get_ext_root_scan_seq()); 2189 print_summary(2, "Mark Stack Scanning", 2190 body_summary->get_mark_stack_scan_seq()); 2191 print_summary(2, "Scan RS", body_summary->get_scan_rs_seq()); 2192 print_summary(2, "Object Copy", body_summary->get_obj_copy_seq()); 2193 print_summary(2, "Termination", body_summary->get_termination_seq()); 2194 print_summary(2, "Other", body_summary->get_parallel_other_seq()); 2195 { 2196 NumberSeq* other_parts[] = { 2197 body_summary->get_update_rs_seq(), 2198 body_summary->get_ext_root_scan_seq(), 2199 body_summary->get_mark_stack_scan_seq(), 2200 body_summary->get_scan_rs_seq(), 2201 body_summary->get_obj_copy_seq(), 2202 body_summary->get_termination_seq() 2203 }; 2204 NumberSeq calc_other_times_ms(body_summary->get_parallel_seq(), 2205 6, other_parts); 2206 check_other_times(2, body_summary->get_parallel_other_seq(), 2207 &calc_other_times_ms); 2208 } 2209 print_summary(1, "Mark Closure", body_summary->get_mark_closure_seq()); 2210 print_summary(1, "Clear CT", body_summary->get_clear_ct_seq()); 2211 } else { 2212 print_summary(1, "Update RS", body_summary->get_update_rs_seq()); 2213 print_summary(1, "Ext Root Scanning", 2214 body_summary->get_ext_root_scan_seq()); 2215 print_summary(1, "Mark Stack Scanning", 2216 body_summary->get_mark_stack_scan_seq()); 2217 print_summary(1, "Scan RS", body_summary->get_scan_rs_seq()); 2218 print_summary(1, "Object Copy", body_summary->get_obj_copy_seq()); 2219 } 2220 } 2221 print_summary(1, "Other", summary->get_other_seq()); 2222 { 2223 if (body_summary != NULL) { 2224 NumberSeq calc_other_times_ms; 2225 if (parallel) { 2226 // parallel 2227 NumberSeq* other_parts[] = { 2228 body_summary->get_satb_drain_seq(), 2229 body_summary->get_parallel_seq(), 2230 body_summary->get_clear_ct_seq() 2231 }; 2232 calc_other_times_ms = NumberSeq(summary->get_total_seq(), 2233 3, other_parts); 2234 } else { 2235 // serial 2236 NumberSeq* other_parts[] = { 2237 body_summary->get_satb_drain_seq(), 2238 body_summary->get_update_rs_seq(), 2239 body_summary->get_ext_root_scan_seq(), 2240 body_summary->get_mark_stack_scan_seq(), 2241 body_summary->get_scan_rs_seq(), 2242 body_summary->get_obj_copy_seq() 2243 }; 2244 calc_other_times_ms = NumberSeq(summary->get_total_seq(), 2245 6, other_parts); 2246 } 2247 check_other_times(1, summary->get_other_seq(), &calc_other_times_ms); 2248 } 2249 } 2250 } else { 2251 LineBuffer(1).append_and_print_cr("none"); 2252 } 2253 LineBuffer(0).append_and_print_cr(""); 2254 } 2255 2256 void G1CollectorPolicy::print_tracing_info() const { 2257 if (TraceGen0Time) { 2258 gclog_or_tty->print_cr("ALL PAUSES"); 2259 print_summary_sd(0, "Total", _all_pause_times_ms); 2260 gclog_or_tty->print_cr(""); 2261 gclog_or_tty->print_cr(""); 2262 gclog_or_tty->print_cr(" Full Young GC Pauses: %8d", _full_young_pause_num); 2263 gclog_or_tty->print_cr(" Partial Young GC Pauses: %8d", _partial_young_pause_num); 2264 gclog_or_tty->print_cr(""); 2265 2266 gclog_or_tty->print_cr("EVACUATION PAUSES"); 2267 print_summary(_summary); 2268 2269 gclog_or_tty->print_cr("MISC"); 2270 print_summary_sd(0, "Stop World", _all_stop_world_times_ms); 2271 print_summary_sd(0, "Yields", _all_yield_times_ms); 2272 for (int i = 0; i < _aux_num; ++i) { 2273 if (_all_aux_times_ms[i].num() > 0) { 2274 char buffer[96]; 2275 sprintf(buffer, "Aux%d", i); 2276 print_summary_sd(0, buffer, &_all_aux_times_ms[i]); 2277 } 2278 } 2279 2280 size_t all_region_num = _region_num_young + _region_num_tenured; 2281 gclog_or_tty->print_cr(" New Regions %8d, Young %8d (%6.2lf%%), " 2282 "Tenured %8d (%6.2lf%%)", 2283 all_region_num, 2284 _region_num_young, 2285 (double) _region_num_young / (double) all_region_num * 100.0, 2286 _region_num_tenured, 2287 (double) _region_num_tenured / (double) all_region_num * 100.0); 2288 } 2289 if (TraceGen1Time) { 2290 if (_all_full_gc_times_ms->num() > 0) { 2291 gclog_or_tty->print("\n%4d full_gcs: total time = %8.2f s", 2292 _all_full_gc_times_ms->num(), 2293 _all_full_gc_times_ms->sum() / 1000.0); 2294 gclog_or_tty->print_cr(" (avg = %8.2fms).", _all_full_gc_times_ms->avg()); 2295 gclog_or_tty->print_cr(" [std. dev = %8.2f ms, max = %8.2f ms]", 2296 _all_full_gc_times_ms->sd(), 2297 _all_full_gc_times_ms->maximum()); 2298 } 2299 } 2300 } 2301 2302 void G1CollectorPolicy::print_yg_surv_rate_info() const { 2303 #ifndef PRODUCT 2304 _short_lived_surv_rate_group->print_surv_rate_summary(); 2305 // add this call for any other surv rate groups 2306 #endif // PRODUCT 2307 } 2308 2309 void 2310 G1CollectorPolicy::update_region_num(bool young) { 2311 if (young) { 2312 ++_region_num_young; 2313 } else { 2314 ++_region_num_tenured; 2315 } 2316 } 2317 2318 #ifndef PRODUCT 2319 // for debugging, bit of a hack... 2320 static char* 2321 region_num_to_mbs(int length) { 2322 static char buffer[64]; 2323 double bytes = (double) (length * HeapRegion::GrainBytes); 2324 double mbs = bytes / (double) (1024 * 1024); 2325 sprintf(buffer, "%7.2lfMB", mbs); 2326 return buffer; 2327 } 2328 #endif // PRODUCT 2329 2330 size_t G1CollectorPolicy::max_regions(int purpose) { 2331 switch (purpose) { 2332 case GCAllocForSurvived: 2333 return _max_survivor_regions; 2334 case GCAllocForTenured: 2335 return REGIONS_UNLIMITED; 2336 default: 2337 ShouldNotReachHere(); 2338 return REGIONS_UNLIMITED; 2339 }; 2340 } 2341 2342 void G1CollectorPolicy::update_max_gc_locker_expansion() { 2343 size_t expansion_region_num = 0; 2344 if (GCLockerEdenExpansionPercent > 0) { 2345 double perc = (double) GCLockerEdenExpansionPercent / 100.0; 2346 double expansion_region_num_d = perc * (double) _young_list_target_length; 2347 // We use ceiling so that if expansion_region_num_d is > 0.0 (but 2348 // less than 1.0) we'll get 1. 2349 expansion_region_num = (size_t) ceil(expansion_region_num_d); 2350 } else { 2351 assert(expansion_region_num == 0, "sanity"); 2352 } 2353 _young_list_max_length = _young_list_target_length + expansion_region_num; 2354 assert(_young_list_target_length <= _young_list_max_length, "post-condition"); 2355 } 2356 2357 // Calculates survivor space parameters. 2358 void G1CollectorPolicy::update_survivors_policy() { 2359 double max_survivor_regions_d = 2360 (double) _young_list_target_length / (double) SurvivorRatio; 2361 // We use ceiling so that if max_survivor_regions_d is > 0.0 (but 2362 // smaller than 1.0) we'll get 1. 2363 _max_survivor_regions = (size_t) ceil(max_survivor_regions_d); 2364 2365 _tenuring_threshold = _survivors_age_table.compute_tenuring_threshold( 2366 HeapRegion::GrainWords * _max_survivor_regions); 2367 } 2368 2369 #ifndef PRODUCT 2370 class HRSortIndexIsOKClosure: public HeapRegionClosure { 2371 CollectionSetChooser* _chooser; 2372 public: 2373 HRSortIndexIsOKClosure(CollectionSetChooser* chooser) : 2374 _chooser(chooser) {} 2375 2376 bool doHeapRegion(HeapRegion* r) { 2377 if (!r->continuesHumongous()) { 2378 assert(_chooser->regionProperlyOrdered(r), "Ought to be."); 2379 } 2380 return false; 2381 } 2382 }; 2383 2384 bool G1CollectorPolicy_BestRegionsFirst::assertMarkedBytesDataOK() { 2385 HRSortIndexIsOKClosure cl(_collectionSetChooser); 2386 _g1->heap_region_iterate(&cl); 2387 return true; 2388 } 2389 #endif 2390 2391 bool 2392 G1CollectorPolicy::force_initial_mark_if_outside_cycle() { 2393 bool during_cycle = _g1->concurrent_mark()->cmThread()->during_cycle(); 2394 if (!during_cycle) { 2395 set_initiate_conc_mark_if_possible(); 2396 return true; 2397 } else { 2398 return false; 2399 } 2400 } 2401 2402 void 2403 G1CollectorPolicy::decide_on_conc_mark_initiation() { 2404 // We are about to decide on whether this pause will be an 2405 // initial-mark pause. 2406 2407 // First, during_initial_mark_pause() should not be already set. We 2408 // will set it here if we have to. However, it should be cleared by 2409 // the end of the pause (it's only set for the duration of an 2410 // initial-mark pause). 2411 assert(!during_initial_mark_pause(), "pre-condition"); 2412 2413 if (initiate_conc_mark_if_possible()) { 2414 // We had noticed on a previous pause that the heap occupancy has 2415 // gone over the initiating threshold and we should start a 2416 // concurrent marking cycle. So we might initiate one. 2417 2418 bool during_cycle = _g1->concurrent_mark()->cmThread()->during_cycle(); 2419 if (!during_cycle) { 2420 // The concurrent marking thread is not "during a cycle", i.e., 2421 // it has completed the last one. So we can go ahead and 2422 // initiate a new cycle. 2423 2424 set_during_initial_mark_pause(); 2425 2426 // And we can now clear initiate_conc_mark_if_possible() as 2427 // we've already acted on it. 2428 clear_initiate_conc_mark_if_possible(); 2429 } else { 2430 // The concurrent marking thread is still finishing up the 2431 // previous cycle. If we start one right now the two cycles 2432 // overlap. In particular, the concurrent marking thread might 2433 // be in the process of clearing the next marking bitmap (which 2434 // we will use for the next cycle if we start one). Starting a 2435 // cycle now will be bad given that parts of the marking 2436 // information might get cleared by the marking thread. And we 2437 // cannot wait for the marking thread to finish the cycle as it 2438 // periodically yields while clearing the next marking bitmap 2439 // and, if it's in a yield point, it's waiting for us to 2440 // finish. So, at this point we will not start a cycle and we'll 2441 // let the concurrent marking thread complete the last one. 2442 } 2443 } 2444 } 2445 2446 void 2447 G1CollectorPolicy_BestRegionsFirst:: 2448 record_collection_pause_start(double start_time_sec, size_t start_used) { 2449 G1CollectorPolicy::record_collection_pause_start(start_time_sec, start_used); 2450 } 2451 2452 class KnownGarbageClosure: public HeapRegionClosure { 2453 CollectionSetChooser* _hrSorted; 2454 2455 public: 2456 KnownGarbageClosure(CollectionSetChooser* hrSorted) : 2457 _hrSorted(hrSorted) 2458 {} 2459 2460 bool doHeapRegion(HeapRegion* r) { 2461 // We only include humongous regions in collection 2462 // sets when concurrent mark shows that their contained object is 2463 // unreachable. 2464 2465 // Do we have any marking information for this region? 2466 if (r->is_marked()) { 2467 // We don't include humongous regions in collection 2468 // sets because we collect them immediately at the end of a marking 2469 // cycle. We also don't include young regions because we *must* 2470 // include them in the next collection pause. 2471 if (!r->isHumongous() && !r->is_young()) { 2472 _hrSorted->addMarkedHeapRegion(r); 2473 } 2474 } 2475 return false; 2476 } 2477 }; 2478 2479 class ParKnownGarbageHRClosure: public HeapRegionClosure { 2480 CollectionSetChooser* _hrSorted; 2481 jint _marked_regions_added; 2482 jint _chunk_size; 2483 jint _cur_chunk_idx; 2484 jint _cur_chunk_end; // Cur chunk [_cur_chunk_idx, _cur_chunk_end) 2485 int _worker; 2486 int _invokes; 2487 2488 void get_new_chunk() { 2489 _cur_chunk_idx = _hrSorted->getParMarkedHeapRegionChunk(_chunk_size); 2490 _cur_chunk_end = _cur_chunk_idx + _chunk_size; 2491 } 2492 void add_region(HeapRegion* r) { 2493 if (_cur_chunk_idx == _cur_chunk_end) { 2494 get_new_chunk(); 2495 } 2496 assert(_cur_chunk_idx < _cur_chunk_end, "postcondition"); 2497 _hrSorted->setMarkedHeapRegion(_cur_chunk_idx, r); 2498 _marked_regions_added++; 2499 _cur_chunk_idx++; 2500 } 2501 2502 public: 2503 ParKnownGarbageHRClosure(CollectionSetChooser* hrSorted, 2504 jint chunk_size, 2505 int worker) : 2506 _hrSorted(hrSorted), _chunk_size(chunk_size), _worker(worker), 2507 _marked_regions_added(0), _cur_chunk_idx(0), _cur_chunk_end(0), 2508 _invokes(0) 2509 {} 2510 2511 bool doHeapRegion(HeapRegion* r) { 2512 // We only include humongous regions in collection 2513 // sets when concurrent mark shows that their contained object is 2514 // unreachable. 2515 _invokes++; 2516 2517 // Do we have any marking information for this region? 2518 if (r->is_marked()) { 2519 // We don't include humongous regions in collection 2520 // sets because we collect them immediately at the end of a marking 2521 // cycle. 2522 // We also do not include young regions in collection sets 2523 if (!r->isHumongous() && !r->is_young()) { 2524 add_region(r); 2525 } 2526 } 2527 return false; 2528 } 2529 jint marked_regions_added() { return _marked_regions_added; } 2530 int invokes() { return _invokes; } 2531 }; 2532 2533 class ParKnownGarbageTask: public AbstractGangTask { 2534 CollectionSetChooser* _hrSorted; 2535 jint _chunk_size; 2536 G1CollectedHeap* _g1; 2537 public: 2538 ParKnownGarbageTask(CollectionSetChooser* hrSorted, jint chunk_size) : 2539 AbstractGangTask("ParKnownGarbageTask"), 2540 _hrSorted(hrSorted), _chunk_size(chunk_size), 2541 _g1(G1CollectedHeap::heap()) 2542 {} 2543 2544 void work(int i) { 2545 ParKnownGarbageHRClosure parKnownGarbageCl(_hrSorted, _chunk_size, i); 2546 // Back to zero for the claim value. 2547 _g1->heap_region_par_iterate_chunked(&parKnownGarbageCl, i, 2548 HeapRegion::InitialClaimValue); 2549 jint regions_added = parKnownGarbageCl.marked_regions_added(); 2550 _hrSorted->incNumMarkedHeapRegions(regions_added); 2551 if (G1PrintParCleanupStats) { 2552 gclog_or_tty->print_cr(" Thread %d called %d times, added %d regions to list.", 2553 i, parKnownGarbageCl.invokes(), regions_added); 2554 } 2555 } 2556 }; 2557 2558 void 2559 G1CollectorPolicy_BestRegionsFirst:: 2560 record_concurrent_mark_cleanup_end(size_t freed_bytes, 2561 size_t max_live_bytes) { 2562 double start; 2563 if (G1PrintParCleanupStats) start = os::elapsedTime(); 2564 record_concurrent_mark_cleanup_end_work1(freed_bytes, max_live_bytes); 2565 2566 _collectionSetChooser->clearMarkedHeapRegions(); 2567 double clear_marked_end; 2568 if (G1PrintParCleanupStats) { 2569 clear_marked_end = os::elapsedTime(); 2570 gclog_or_tty->print_cr(" clear marked regions + work1: %8.3f ms.", 2571 (clear_marked_end - start)*1000.0); 2572 } 2573 if (G1CollectedHeap::use_parallel_gc_threads()) { 2574 const size_t OverpartitionFactor = 4; 2575 const size_t MinWorkUnit = 8; 2576 const size_t WorkUnit = 2577 MAX2(_g1->n_regions() / (ParallelGCThreads * OverpartitionFactor), 2578 MinWorkUnit); 2579 _collectionSetChooser->prepareForAddMarkedHeapRegionsPar(_g1->n_regions(), 2580 WorkUnit); 2581 ParKnownGarbageTask parKnownGarbageTask(_collectionSetChooser, 2582 (int) WorkUnit); 2583 _g1->workers()->run_task(&parKnownGarbageTask); 2584 2585 assert(_g1->check_heap_region_claim_values(HeapRegion::InitialClaimValue), 2586 "sanity check"); 2587 } else { 2588 KnownGarbageClosure knownGarbagecl(_collectionSetChooser); 2589 _g1->heap_region_iterate(&knownGarbagecl); 2590 } 2591 double known_garbage_end; 2592 if (G1PrintParCleanupStats) { 2593 known_garbage_end = os::elapsedTime(); 2594 gclog_or_tty->print_cr(" compute known garbage: %8.3f ms.", 2595 (known_garbage_end - clear_marked_end)*1000.0); 2596 } 2597 _collectionSetChooser->sortMarkedHeapRegions(); 2598 double sort_end; 2599 if (G1PrintParCleanupStats) { 2600 sort_end = os::elapsedTime(); 2601 gclog_or_tty->print_cr(" sorting: %8.3f ms.", 2602 (sort_end - known_garbage_end)*1000.0); 2603 } 2604 2605 record_concurrent_mark_cleanup_end_work2(); 2606 double work2_end; 2607 if (G1PrintParCleanupStats) { 2608 work2_end = os::elapsedTime(); 2609 gclog_or_tty->print_cr(" work2: %8.3f ms.", 2610 (work2_end - sort_end)*1000.0); 2611 } 2612 } 2613 2614 // Add the heap region at the head of the non-incremental collection set 2615 void G1CollectorPolicy:: 2616 add_to_collection_set(HeapRegion* hr) { 2617 assert(_inc_cset_build_state == Active, "Precondition"); 2618 assert(!hr->is_young(), "non-incremental add of young region"); 2619 2620 if (_g1->mark_in_progress()) 2621 _g1->concurrent_mark()->registerCSetRegion(hr); 2622 2623 assert(!hr->in_collection_set(), "should not already be in the CSet"); 2624 hr->set_in_collection_set(true); 2625 hr->set_next_in_collection_set(_collection_set); 2626 _collection_set = hr; 2627 _collection_set_size++; 2628 _collection_set_bytes_used_before += hr->used(); 2629 _g1->register_region_with_in_cset_fast_test(hr); 2630 } 2631 2632 // Initialize the per-collection-set information 2633 void G1CollectorPolicy::start_incremental_cset_building() { 2634 assert(_inc_cset_build_state == Inactive, "Precondition"); 2635 2636 _inc_cset_head = NULL; 2637 _inc_cset_tail = NULL; 2638 _inc_cset_size = 0; 2639 _inc_cset_bytes_used_before = 0; 2640 2641 _inc_cset_young_index = 0; 2642 2643 _inc_cset_max_finger = 0; 2644 _inc_cset_recorded_young_bytes = 0; 2645 _inc_cset_recorded_rs_lengths = 0; 2646 _inc_cset_predicted_elapsed_time_ms = 0; 2647 _inc_cset_predicted_bytes_to_copy = 0; 2648 _inc_cset_build_state = Active; 2649 } 2650 2651 void G1CollectorPolicy::add_to_incremental_cset_info(HeapRegion* hr, size_t rs_length) { 2652 // This routine is used when: 2653 // * adding survivor regions to the incremental cset at the end of an 2654 // evacuation pause, 2655 // * adding the current allocation region to the incremental cset 2656 // when it is retired, and 2657 // * updating existing policy information for a region in the 2658 // incremental cset via young list RSet sampling. 2659 // Therefore this routine may be called at a safepoint by the 2660 // VM thread, or in-between safepoints by mutator threads (when 2661 // retiring the current allocation region) or a concurrent 2662 // refine thread (RSet sampling). 2663 2664 double region_elapsed_time_ms = predict_region_elapsed_time_ms(hr, true); 2665 size_t used_bytes = hr->used(); 2666 2667 _inc_cset_recorded_rs_lengths += rs_length; 2668 _inc_cset_predicted_elapsed_time_ms += region_elapsed_time_ms; 2669 2670 _inc_cset_bytes_used_before += used_bytes; 2671 2672 // Cache the values we have added to the aggregated informtion 2673 // in the heap region in case we have to remove this region from 2674 // the incremental collection set, or it is updated by the 2675 // rset sampling code 2676 hr->set_recorded_rs_length(rs_length); 2677 hr->set_predicted_elapsed_time_ms(region_elapsed_time_ms); 2678 2679 #if PREDICTIONS_VERBOSE 2680 size_t bytes_to_copy = predict_bytes_to_copy(hr); 2681 _inc_cset_predicted_bytes_to_copy += bytes_to_copy; 2682 2683 // Record the number of bytes used in this region 2684 _inc_cset_recorded_young_bytes += used_bytes; 2685 2686 // Cache the values we have added to the aggregated informtion 2687 // in the heap region in case we have to remove this region from 2688 // the incremental collection set, or it is updated by the 2689 // rset sampling code 2690 hr->set_predicted_bytes_to_copy(bytes_to_copy); 2691 #endif // PREDICTIONS_VERBOSE 2692 } 2693 2694 void G1CollectorPolicy::remove_from_incremental_cset_info(HeapRegion* hr) { 2695 // This routine is currently only called as part of the updating of 2696 // existing policy information for regions in the incremental cset that 2697 // is performed by the concurrent refine thread(s) as part of young list 2698 // RSet sampling. Therefore we should not be at a safepoint. 2699 2700 assert(!SafepointSynchronize::is_at_safepoint(), "should not be at safepoint"); 2701 assert(hr->is_young(), "it should be"); 2702 2703 size_t used_bytes = hr->used(); 2704 size_t old_rs_length = hr->recorded_rs_length(); 2705 double old_elapsed_time_ms = hr->predicted_elapsed_time_ms(); 2706 2707 // Subtract the old recorded/predicted policy information for 2708 // the given heap region from the collection set info. 2709 _inc_cset_recorded_rs_lengths -= old_rs_length; 2710 _inc_cset_predicted_elapsed_time_ms -= old_elapsed_time_ms; 2711 2712 _inc_cset_bytes_used_before -= used_bytes; 2713 2714 // Clear the values cached in the heap region 2715 hr->set_recorded_rs_length(0); 2716 hr->set_predicted_elapsed_time_ms(0); 2717 2718 #if PREDICTIONS_VERBOSE 2719 size_t old_predicted_bytes_to_copy = hr->predicted_bytes_to_copy(); 2720 _inc_cset_predicted_bytes_to_copy -= old_predicted_bytes_to_copy; 2721 2722 // Subtract the number of bytes used in this region 2723 _inc_cset_recorded_young_bytes -= used_bytes; 2724 2725 // Clear the values cached in the heap region 2726 hr->set_predicted_bytes_to_copy(0); 2727 #endif // PREDICTIONS_VERBOSE 2728 } 2729 2730 void G1CollectorPolicy::update_incremental_cset_info(HeapRegion* hr, size_t new_rs_length) { 2731 // Update the collection set information that is dependent on the new RS length 2732 assert(hr->is_young(), "Precondition"); 2733 2734 remove_from_incremental_cset_info(hr); 2735 add_to_incremental_cset_info(hr, new_rs_length); 2736 } 2737 2738 void G1CollectorPolicy::add_region_to_incremental_cset_common(HeapRegion* hr) { 2739 assert( hr->is_young(), "invariant"); 2740 assert( hr->young_index_in_cset() == -1, "invariant" ); 2741 assert(_inc_cset_build_state == Active, "Precondition"); 2742 2743 // We need to clear and set the cached recorded/cached collection set 2744 // information in the heap region here (before the region gets added 2745 // to the collection set). An individual heap region's cached values 2746 // are calculated, aggregated with the policy collection set info, 2747 // and cached in the heap region here (initially) and (subsequently) 2748 // by the Young List sampling code. 2749 2750 size_t rs_length = hr->rem_set()->occupied(); 2751 add_to_incremental_cset_info(hr, rs_length); 2752 2753 HeapWord* hr_end = hr->end(); 2754 _inc_cset_max_finger = MAX2(_inc_cset_max_finger, hr_end); 2755 2756 assert(!hr->in_collection_set(), "invariant"); 2757 hr->set_in_collection_set(true); 2758 assert( hr->next_in_collection_set() == NULL, "invariant"); 2759 2760 _inc_cset_size++; 2761 _g1->register_region_with_in_cset_fast_test(hr); 2762 2763 hr->set_young_index_in_cset((int) _inc_cset_young_index); 2764 ++_inc_cset_young_index; 2765 } 2766 2767 // Add the region at the RHS of the incremental cset 2768 void G1CollectorPolicy::add_region_to_incremental_cset_rhs(HeapRegion* hr) { 2769 // We should only ever be appending survivors at the end of a pause 2770 assert( hr->is_survivor(), "Logic"); 2771 2772 // Do the 'common' stuff 2773 add_region_to_incremental_cset_common(hr); 2774 2775 // Now add the region at the right hand side 2776 if (_inc_cset_tail == NULL) { 2777 assert(_inc_cset_head == NULL, "invariant"); 2778 _inc_cset_head = hr; 2779 } else { 2780 _inc_cset_tail->set_next_in_collection_set(hr); 2781 } 2782 _inc_cset_tail = hr; 2783 } 2784 2785 // Add the region to the LHS of the incremental cset 2786 void G1CollectorPolicy::add_region_to_incremental_cset_lhs(HeapRegion* hr) { 2787 // Survivors should be added to the RHS at the end of a pause 2788 assert(!hr->is_survivor(), "Logic"); 2789 2790 // Do the 'common' stuff 2791 add_region_to_incremental_cset_common(hr); 2792 2793 // Add the region at the left hand side 2794 hr->set_next_in_collection_set(_inc_cset_head); 2795 if (_inc_cset_head == NULL) { 2796 assert(_inc_cset_tail == NULL, "Invariant"); 2797 _inc_cset_tail = hr; 2798 } 2799 _inc_cset_head = hr; 2800 } 2801 2802 #ifndef PRODUCT 2803 void G1CollectorPolicy::print_collection_set(HeapRegion* list_head, outputStream* st) { 2804 assert(list_head == inc_cset_head() || list_head == collection_set(), "must be"); 2805 2806 st->print_cr("\nCollection_set:"); 2807 HeapRegion* csr = list_head; 2808 while (csr != NULL) { 2809 HeapRegion* next = csr->next_in_collection_set(); 2810 assert(csr->in_collection_set(), "bad CS"); 2811 st->print_cr(" [%08x-%08x], t: %08x, P: %08x, N: %08x, C: %08x, " 2812 "age: %4d, y: %d, surv: %d", 2813 csr->bottom(), csr->end(), 2814 csr->top(), 2815 csr->prev_top_at_mark_start(), 2816 csr->next_top_at_mark_start(), 2817 csr->top_at_conc_mark_count(), 2818 csr->age_in_surv_rate_group_cond(), 2819 csr->is_young(), 2820 csr->is_survivor()); 2821 csr = next; 2822 } 2823 } 2824 #endif // !PRODUCT 2825 2826 void 2827 G1CollectorPolicy_BestRegionsFirst::choose_collection_set( 2828 double target_pause_time_ms) { 2829 // Set this here - in case we're not doing young collections. 2830 double non_young_start_time_sec = os::elapsedTime(); 2831 2832 start_recording_regions(); 2833 2834 guarantee(target_pause_time_ms > 0.0, 2835 err_msg("target_pause_time_ms = %1.6lf should be positive", 2836 target_pause_time_ms)); 2837 guarantee(_collection_set == NULL, "Precondition"); 2838 2839 double base_time_ms = predict_base_elapsed_time_ms(_pending_cards); 2840 double predicted_pause_time_ms = base_time_ms; 2841 2842 double time_remaining_ms = target_pause_time_ms - base_time_ms; 2843 2844 // the 10% and 50% values are arbitrary... 2845 if (time_remaining_ms < 0.10 * target_pause_time_ms) { 2846 time_remaining_ms = 0.50 * target_pause_time_ms; 2847 _within_target = false; 2848 } else { 2849 _within_target = true; 2850 } 2851 2852 // We figure out the number of bytes available for future to-space. 2853 // For new regions without marking information, we must assume the 2854 // worst-case of complete survival. If we have marking information for a 2855 // region, we can bound the amount of live data. We can add a number of 2856 // such regions, as long as the sum of the live data bounds does not 2857 // exceed the available evacuation space. 2858 size_t max_live_bytes = _g1->free_regions() * HeapRegion::GrainBytes; 2859 2860 size_t expansion_bytes = 2861 _g1->expansion_regions() * HeapRegion::GrainBytes; 2862 2863 _collection_set_bytes_used_before = 0; 2864 _collection_set_size = 0; 2865 2866 // Adjust for expansion and slop. 2867 max_live_bytes = max_live_bytes + expansion_bytes; 2868 2869 HeapRegion* hr; 2870 double young_start_time_sec = os::elapsedTime(); 2871 2872 if (G1PolicyVerbose > 0) { 2873 gclog_or_tty->print_cr("Adding %d young regions to the CSet", 2874 _g1->young_list()->length()); 2875 } 2876 2877 _young_cset_length = 0; 2878 _last_young_gc_full = full_young_gcs() ? true : false; 2879 2880 if (_last_young_gc_full) 2881 ++_full_young_pause_num; 2882 else 2883 ++_partial_young_pause_num; 2884 2885 // The young list is laid with the survivor regions from the previous 2886 // pause are appended to the RHS of the young list, i.e. 2887 // [Newly Young Regions ++ Survivors from last pause]. 2888 2889 hr = _g1->young_list()->first_survivor_region(); 2890 while (hr != NULL) { 2891 assert(hr->is_survivor(), "badly formed young list"); 2892 hr->set_young(); 2893 hr = hr->get_next_young_region(); 2894 } 2895 2896 // Clear the fields that point to the survivor list - they are 2897 // all young now. 2898 _g1->young_list()->clear_survivors(); 2899 2900 if (_g1->mark_in_progress()) 2901 _g1->concurrent_mark()->register_collection_set_finger(_inc_cset_max_finger); 2902 2903 _young_cset_length = _inc_cset_young_index; 2904 _collection_set = _inc_cset_head; 2905 _collection_set_size = _inc_cset_size; 2906 _collection_set_bytes_used_before = _inc_cset_bytes_used_before; 2907 2908 // For young regions in the collection set, we assume the worst 2909 // case of complete survival 2910 max_live_bytes -= _inc_cset_size * HeapRegion::GrainBytes; 2911 2912 time_remaining_ms -= _inc_cset_predicted_elapsed_time_ms; 2913 predicted_pause_time_ms += _inc_cset_predicted_elapsed_time_ms; 2914 2915 // The number of recorded young regions is the incremental 2916 // collection set's current size 2917 set_recorded_young_regions(_inc_cset_size); 2918 set_recorded_rs_lengths(_inc_cset_recorded_rs_lengths); 2919 set_recorded_young_bytes(_inc_cset_recorded_young_bytes); 2920 #if PREDICTIONS_VERBOSE 2921 set_predicted_bytes_to_copy(_inc_cset_predicted_bytes_to_copy); 2922 #endif // PREDICTIONS_VERBOSE 2923 2924 if (G1PolicyVerbose > 0) { 2925 gclog_or_tty->print_cr(" Added " PTR_FORMAT " Young Regions to CS.", 2926 _inc_cset_size); 2927 gclog_or_tty->print_cr(" (" SIZE_FORMAT " KB left in heap.)", 2928 max_live_bytes/K); 2929 } 2930 2931 assert(_inc_cset_size == _g1->young_list()->length(), "Invariant"); 2932 2933 double young_end_time_sec = os::elapsedTime(); 2934 _recorded_young_cset_choice_time_ms = 2935 (young_end_time_sec - young_start_time_sec) * 1000.0; 2936 2937 // We are doing young collections so reset this. 2938 non_young_start_time_sec = young_end_time_sec; 2939 2940 if (!full_young_gcs()) { 2941 bool should_continue = true; 2942 NumberSeq seq; 2943 double avg_prediction = 100000000000000000.0; // something very large 2944 2945 do { 2946 hr = _collectionSetChooser->getNextMarkedRegion(time_remaining_ms, 2947 avg_prediction); 2948 if (hr != NULL) { 2949 double predicted_time_ms = predict_region_elapsed_time_ms(hr, false); 2950 time_remaining_ms -= predicted_time_ms; 2951 predicted_pause_time_ms += predicted_time_ms; 2952 add_to_collection_set(hr); 2953 record_non_young_cset_region(hr); 2954 max_live_bytes -= MIN2(hr->max_live_bytes(), max_live_bytes); 2955 if (G1PolicyVerbose > 0) { 2956 gclog_or_tty->print_cr(" (" SIZE_FORMAT " KB left in heap.)", 2957 max_live_bytes/K); 2958 } 2959 seq.add(predicted_time_ms); 2960 avg_prediction = seq.avg() + seq.sd(); 2961 } 2962 should_continue = 2963 ( hr != NULL) && 2964 ( (adaptive_young_list_length()) ? time_remaining_ms > 0.0 2965 : _collection_set_size < _young_list_fixed_length ); 2966 } while (should_continue); 2967 2968 if (!adaptive_young_list_length() && 2969 _collection_set_size < _young_list_fixed_length) 2970 _should_revert_to_full_young_gcs = true; 2971 } 2972 2973 stop_incremental_cset_building(); 2974 2975 count_CS_bytes_used(); 2976 2977 end_recording_regions(); 2978 2979 double non_young_end_time_sec = os::elapsedTime(); 2980 _recorded_non_young_cset_choice_time_ms = 2981 (non_young_end_time_sec - non_young_start_time_sec) * 1000.0; 2982 } 2983 2984 void G1CollectorPolicy_BestRegionsFirst::record_full_collection_end() { 2985 G1CollectorPolicy::record_full_collection_end(); 2986 _collectionSetChooser->updateAfterFullCollection(); 2987 } 2988 2989 void G1CollectorPolicy_BestRegionsFirst:: 2990 expand_if_possible(size_t numRegions) { 2991 size_t expansion_bytes = numRegions * HeapRegion::GrainBytes; 2992 _g1->expand(expansion_bytes); 2993 } 2994 2995 void G1CollectorPolicy_BestRegionsFirst:: 2996 record_collection_pause_end() { 2997 G1CollectorPolicy::record_collection_pause_end(); 2998 assert(assertMarkedBytesDataOK(), "Marked regions not OK at pause end."); 2999 }