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/g1ErgoVerbose.hpp"
32 #include "gc_implementation/g1/heapRegionRemSet.hpp"
33 #include "gc_implementation/shared/gcPolicyCounters.hpp"
34 #include "runtime/arguments.hpp"
35 #include "runtime/java.hpp"
36 #include "runtime/mutexLocker.hpp"
37 #include "utilities/debug.hpp"
38
39 #define PREDICTIONS_VERBOSE 0
40
41 // <NEW PREDICTION>
42
43 // Different defaults for different number of GC threads
44 // They were chosen by running GCOld and SPECjbb on debris with different
45 // numbers of GC threads and choosing them based on the results
46
47 // all the same
48 static double rs_length_diff_defaults[] = {
49 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0
50 };
51
52 static double cost_per_card_ms_defaults[] = {
53 0.01, 0.005, 0.005, 0.003, 0.003, 0.002, 0.002, 0.0015
54 };
55
56 // all the same
57 static double fully_young_cards_per_entry_ratio_defaults[] = {
58 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0
59 };
60
61 static double cost_per_entry_ms_defaults[] = {
62 0.015, 0.01, 0.01, 0.008, 0.008, 0.0055, 0.0055, 0.005
63 };
64
65 static double cost_per_byte_ms_defaults[] = {
66 0.00006, 0.00003, 0.00003, 0.000015, 0.000015, 0.00001, 0.00001, 0.000009
67 };
68
69 // these should be pretty consistent
70 static double constant_other_time_ms_defaults[] = {
71 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0
72 };
73
74
75 static double young_other_cost_per_region_ms_defaults[] = {
76 0.3, 0.2, 0.2, 0.15, 0.15, 0.12, 0.12, 0.1
77 };
78
79 static double non_young_other_cost_per_region_ms_defaults[] = {
80 1.0, 0.7, 0.7, 0.5, 0.5, 0.42, 0.42, 0.30
81 };
82
83 // </NEW PREDICTION>
84
85 // Help class for avoiding interleaved logging
86 class LineBuffer: public StackObj {
87
88 private:
89 static const int BUFFER_LEN = 1024;
90 static const int INDENT_CHARS = 3;
91 char _buffer[BUFFER_LEN];
92 int _indent_level;
93 int _cur;
94
95 void vappend(const char* format, va_list ap) {
96 int res = vsnprintf(&_buffer[_cur], BUFFER_LEN - _cur, format, ap);
97 if (res != -1) {
98 _cur += res;
99 } else {
100 DEBUG_ONLY(warning("buffer too small in LineBuffer");)
101 _buffer[BUFFER_LEN -1] = 0;
102 _cur = BUFFER_LEN; // vsnprintf above should not add to _buffer if we are called again
103 }
104 }
105
106 public:
107 explicit LineBuffer(int indent_level): _indent_level(indent_level), _cur(0) {
108 for (; (_cur < BUFFER_LEN && _cur < (_indent_level * INDENT_CHARS)); _cur++) {
109 _buffer[_cur] = ' ';
110 }
111 }
112
113 #ifndef PRODUCT
114 ~LineBuffer() {
115 assert(_cur == _indent_level * INDENT_CHARS, "pending data in buffer - append_and_print_cr() not called?");
116 }
117 #endif
118
119 void append(const char* format, ...) {
120 va_list ap;
121 va_start(ap, format);
122 vappend(format, ap);
123 va_end(ap);
124 }
125
126 void append_and_print_cr(const char* format, ...) {
127 va_list ap;
128 va_start(ap, format);
129 vappend(format, ap);
130 va_end(ap);
131 gclog_or_tty->print_cr("%s", _buffer);
132 _cur = _indent_level * INDENT_CHARS;
133 }
134 };
135
136 G1CollectorPolicy::G1CollectorPolicy() :
137 _parallel_gc_threads(G1CollectedHeap::use_parallel_gc_threads()
138 ? ParallelGCThreads : 1),
139
140 _n_pauses(0),
141 _recent_rs_scan_times_ms(new TruncatedSeq(NumPrevPausesForHeuristics)),
142 _recent_pause_times_ms(new TruncatedSeq(NumPrevPausesForHeuristics)),
143 _recent_rs_sizes(new TruncatedSeq(NumPrevPausesForHeuristics)),
144 _recent_gc_times_ms(new TruncatedSeq(NumPrevPausesForHeuristics)),
145 _all_pause_times_ms(new NumberSeq()),
146 _stop_world_start(0.0),
147 _all_stop_world_times_ms(new NumberSeq()),
148 _all_yield_times_ms(new NumberSeq()),
149 _using_new_ratio_calculations(false),
150
151 _all_mod_union_times_ms(new NumberSeq()),
152
153 _summary(new Summary()),
154
155 _cur_clear_ct_time_ms(0.0),
156
157 _cur_ref_proc_time_ms(0.0),
158 _cur_ref_enq_time_ms(0.0),
159
160 #ifndef PRODUCT
161 _min_clear_cc_time_ms(-1.0),
162 _max_clear_cc_time_ms(-1.0),
163 _cur_clear_cc_time_ms(0.0),
164 _cum_clear_cc_time_ms(0.0),
165 _num_cc_clears(0L),
166 #endif
167
168 _region_num_young(0),
169 _region_num_tenured(0),
170 _prev_region_num_young(0),
171 _prev_region_num_tenured(0),
172
173 _aux_num(10),
174 _all_aux_times_ms(new NumberSeq[_aux_num]),
175 _cur_aux_start_times_ms(new double[_aux_num]),
176 _cur_aux_times_ms(new double[_aux_num]),
177 _cur_aux_times_set(new bool[_aux_num]),
178
179 _concurrent_mark_remark_times_ms(new TruncatedSeq(NumPrevPausesForHeuristics)),
180 _concurrent_mark_cleanup_times_ms(new TruncatedSeq(NumPrevPausesForHeuristics)),
181
182 // <NEW PREDICTION>
183
184 _alloc_rate_ms_seq(new TruncatedSeq(TruncatedSeqLength)),
185 _prev_collection_pause_end_ms(0.0),
186 _pending_card_diff_seq(new TruncatedSeq(TruncatedSeqLength)),
187 _rs_length_diff_seq(new TruncatedSeq(TruncatedSeqLength)),
188 _cost_per_card_ms_seq(new TruncatedSeq(TruncatedSeqLength)),
189 _fully_young_cards_per_entry_ratio_seq(new TruncatedSeq(TruncatedSeqLength)),
190 _partially_young_cards_per_entry_ratio_seq(
191 new TruncatedSeq(TruncatedSeqLength)),
192 _cost_per_entry_ms_seq(new TruncatedSeq(TruncatedSeqLength)),
193 _partially_young_cost_per_entry_ms_seq(new TruncatedSeq(TruncatedSeqLength)),
194 _cost_per_byte_ms_seq(new TruncatedSeq(TruncatedSeqLength)),
195 _cost_per_byte_ms_during_cm_seq(new TruncatedSeq(TruncatedSeqLength)),
196 _constant_other_time_ms_seq(new TruncatedSeq(TruncatedSeqLength)),
197 _young_other_cost_per_region_ms_seq(new TruncatedSeq(TruncatedSeqLength)),
198 _non_young_other_cost_per_region_ms_seq(
199 new TruncatedSeq(TruncatedSeqLength)),
200
201 _pending_cards_seq(new TruncatedSeq(TruncatedSeqLength)),
202 _scanned_cards_seq(new TruncatedSeq(TruncatedSeqLength)),
203 _rs_lengths_seq(new TruncatedSeq(TruncatedSeqLength)),
204
205 _pause_time_target_ms((double) MaxGCPauseMillis),
206
207 // </NEW PREDICTION>
208
209 _full_young_gcs(true),
210 _full_young_pause_num(0),
211 _partial_young_pause_num(0),
212
213 _during_marking(false),
214 _in_marking_window(false),
215 _in_marking_window_im(false),
216
217 _known_garbage_ratio(0.0),
218 _known_garbage_bytes(0),
219
220 _young_gc_eff_seq(new TruncatedSeq(TruncatedSeqLength)),
221
222 _recent_prev_end_times_for_all_gcs_sec(new TruncatedSeq(NumPrevPausesForHeuristics)),
223
224 _recent_CS_bytes_used_before(new TruncatedSeq(NumPrevPausesForHeuristics)),
225 _recent_CS_bytes_surviving(new TruncatedSeq(NumPrevPausesForHeuristics)),
226
227 _recent_avg_pause_time_ratio(0.0),
228
229 _all_full_gc_times_ms(new NumberSeq()),
230
231 // G1PausesBtwnConcMark defaults to -1
232 // so the hack is to do the cast QQQ FIXME
233 _pauses_btwn_concurrent_mark((size_t)G1PausesBtwnConcMark),
234 _initiate_conc_mark_if_possible(false),
235 _during_initial_mark_pause(false),
236 _should_revert_to_full_young_gcs(false),
237 _last_full_young_gc(false),
238
239 _eden_bytes_before_gc(0),
240 _survivor_bytes_before_gc(0),
241 _capacity_before_gc(0),
242
243 _prev_collection_pause_used_at_end_bytes(0),
244
245 _collection_set(NULL),
246 _collection_set_size(0),
247 _collection_set_bytes_used_before(0),
248
249 // Incremental CSet attributes
250 _inc_cset_build_state(Inactive),
251 _inc_cset_head(NULL),
252 _inc_cset_tail(NULL),
253 _inc_cset_size(0),
254 _inc_cset_young_index(0),
255 _inc_cset_bytes_used_before(0),
256 _inc_cset_max_finger(NULL),
257 _inc_cset_recorded_young_bytes(0),
258 _inc_cset_recorded_rs_lengths(0),
259 _inc_cset_predicted_elapsed_time_ms(0.0),
260 _inc_cset_predicted_bytes_to_copy(0),
261
262 #ifdef _MSC_VER // the use of 'this' below gets a warning, make it go away
263 #pragma warning( disable:4355 ) // 'this' : used in base member initializer list
264 #endif // _MSC_VER
265
266 _short_lived_surv_rate_group(new SurvRateGroup(this, "Short Lived",
267 G1YoungSurvRateNumRegionsSummary)),
268 _survivor_surv_rate_group(new SurvRateGroup(this, "Survivor",
269 G1YoungSurvRateNumRegionsSummary)),
270 // add here any more surv rate groups
271 _recorded_survivor_regions(0),
272 _recorded_survivor_head(NULL),
273 _recorded_survivor_tail(NULL),
274 _survivors_age_table(true),
275
276 _gc_overhead_perc(0.0) {
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 G1ErgoVerbose::initialize();
285 if (PrintAdaptiveSizePolicy) {
286 // Currently, we only use a single switch for all the heuristics.
287 G1ErgoVerbose::set_enabled(true);
288 // Given that we don't currently have a verboseness level
289 // parameter, we'll hardcode this to high. This can be easily
290 // changed in the future.
291 G1ErgoVerbose::set_level(ErgoHigh);
292 } else {
293 G1ErgoVerbose::set_enabled(false);
294 }
295
296 // Verify PLAB sizes
297 const size_t region_size = HeapRegion::GrainWords;
298 if (YoungPLABSize > region_size || OldPLABSize > region_size) {
299 char buffer[128];
300 jio_snprintf(buffer, sizeof(buffer), "%sPLABSize should be at most "SIZE_FORMAT,
301 OldPLABSize > region_size ? "Old" : "Young", region_size);
302 vm_exit_during_initialization(buffer);
303 }
304
305 _recent_prev_end_times_for_all_gcs_sec->add(os::elapsedTime());
306 _prev_collection_pause_end_ms = os::elapsedTime() * 1000.0;
307
308 _par_last_gc_worker_start_times_ms = new double[_parallel_gc_threads];
309 _par_last_ext_root_scan_times_ms = new double[_parallel_gc_threads];
310 _par_last_mark_stack_scan_times_ms = new double[_parallel_gc_threads];
311
312 _par_last_update_rs_times_ms = new double[_parallel_gc_threads];
313 _par_last_update_rs_processed_buffers = new double[_parallel_gc_threads];
314
315 _par_last_scan_rs_times_ms = new double[_parallel_gc_threads];
316
317 _par_last_obj_copy_times_ms = new double[_parallel_gc_threads];
318
319 _par_last_termination_times_ms = new double[_parallel_gc_threads];
320 _par_last_termination_attempts = new double[_parallel_gc_threads];
321 _par_last_gc_worker_end_times_ms = new double[_parallel_gc_threads];
322 _par_last_gc_worker_times_ms = new double[_parallel_gc_threads];
323 _par_last_gc_worker_other_times_ms = new double[_parallel_gc_threads];
324
325 // start conservatively
326 _expensive_region_limit_ms = 0.5 * (double) MaxGCPauseMillis;
327
328 // <NEW PREDICTION>
329
330 int index;
331 if (ParallelGCThreads == 0)
332 index = 0;
333 else if (ParallelGCThreads > 8)
334 index = 7;
335 else
336 index = ParallelGCThreads - 1;
337
338 _pending_card_diff_seq->add(0.0);
339 _rs_length_diff_seq->add(rs_length_diff_defaults[index]);
340 _cost_per_card_ms_seq->add(cost_per_card_ms_defaults[index]);
341 _fully_young_cards_per_entry_ratio_seq->add(
342 fully_young_cards_per_entry_ratio_defaults[index]);
343 _cost_per_entry_ms_seq->add(cost_per_entry_ms_defaults[index]);
344 _cost_per_byte_ms_seq->add(cost_per_byte_ms_defaults[index]);
345 _constant_other_time_ms_seq->add(constant_other_time_ms_defaults[index]);
346 _young_other_cost_per_region_ms_seq->add(
347 young_other_cost_per_region_ms_defaults[index]);
348 _non_young_other_cost_per_region_ms_seq->add(
349 non_young_other_cost_per_region_ms_defaults[index]);
350
351 // </NEW PREDICTION>
352
353 // Below, we might need to calculate the pause time target based on
354 // the pause interval. When we do so we are going to give G1 maximum
355 // flexibility and allow it to do pauses when it needs to. So, we'll
356 // arrange that the pause interval to be pause time target + 1 to
357 // ensure that a) the pause time target is maximized with respect to
358 // the pause interval and b) we maintain the invariant that pause
359 // time target < pause interval. If the user does not want this
360 // maximum flexibility, they will have to set the pause interval
361 // explicitly.
362
363 // First make sure that, if either parameter is set, its value is
364 // reasonable.
365 if (!FLAG_IS_DEFAULT(MaxGCPauseMillis)) {
366 if (MaxGCPauseMillis < 1) {
367 vm_exit_during_initialization("MaxGCPauseMillis should be "
368 "greater than 0");
369 }
370 }
371 if (!FLAG_IS_DEFAULT(GCPauseIntervalMillis)) {
372 if (GCPauseIntervalMillis < 1) {
373 vm_exit_during_initialization("GCPauseIntervalMillis should be "
374 "greater than 0");
375 }
376 }
377
378 // Then, if the pause time target parameter was not set, set it to
379 // the default value.
380 if (FLAG_IS_DEFAULT(MaxGCPauseMillis)) {
381 if (FLAG_IS_DEFAULT(GCPauseIntervalMillis)) {
382 // The default pause time target in G1 is 200ms
383 FLAG_SET_DEFAULT(MaxGCPauseMillis, 200);
384 } else {
385 // We do not allow the pause interval to be set without the
386 // pause time target
387 vm_exit_during_initialization("GCPauseIntervalMillis cannot be set "
388 "without setting MaxGCPauseMillis");
389 }
390 }
391
392 // Then, if the interval parameter was not set, set it according to
393 // the pause time target (this will also deal with the case when the
394 // pause time target is the default value).
395 if (FLAG_IS_DEFAULT(GCPauseIntervalMillis)) {
396 FLAG_SET_DEFAULT(GCPauseIntervalMillis, MaxGCPauseMillis + 1);
397 }
398
399 // Finally, make sure that the two parameters are consistent.
400 if (MaxGCPauseMillis >= GCPauseIntervalMillis) {
401 char buffer[256];
402 jio_snprintf(buffer, 256,
403 "MaxGCPauseMillis (%u) should be less than "
404 "GCPauseIntervalMillis (%u)",
405 MaxGCPauseMillis, GCPauseIntervalMillis);
406 vm_exit_during_initialization(buffer);
407 }
408
409 double max_gc_time = (double) MaxGCPauseMillis / 1000.0;
410 double time_slice = (double) GCPauseIntervalMillis / 1000.0;
411 _mmu_tracker = new G1MMUTrackerQueue(time_slice, max_gc_time);
412 _sigma = (double) G1ConfidencePercent / 100.0;
413
414 // start conservatively (around 50ms is about right)
415 _concurrent_mark_remark_times_ms->add(0.05);
416 _concurrent_mark_cleanup_times_ms->add(0.20);
417 _tenuring_threshold = MaxTenuringThreshold;
418 // _max_survivor_regions will be calculated by
419 // update_young_list_target_length() during initialization.
420 _max_survivor_regions = 0;
421
422 assert(GCTimeRatio > 0,
423 "we should have set it to a default value set_g1_gc_flags() "
424 "if a user set it to 0");
425 _gc_overhead_perc = 100.0 * (1.0 / (1.0 + GCTimeRatio));
426
427 uintx reserve_perc = G1ReservePercent;
428 // Put an artificial ceiling on this so that it's not set to a silly value.
429 if (reserve_perc > 50) {
430 reserve_perc = 50;
431 warning("G1ReservePercent is set to a value that is too large, "
432 "it's been updated to %u", reserve_perc);
433 }
434 _reserve_factor = (double) reserve_perc / 100.0;
435 // This will be set when the heap is expanded
436 // for the first time during initialization.
437 _reserve_regions = 0;
438
439 initialize_all();
440 _collectionSetChooser = new CollectionSetChooser();
441 }
442
443 // Increment "i", mod "len"
444 static void inc_mod(int& i, int len) {
445 i++; if (i == len) i = 0;
446 }
447
448 void G1CollectorPolicy::initialize_flags() {
449 set_min_alignment(HeapRegion::GrainBytes);
450 set_max_alignment(GenRemSet::max_alignment_constraint(rem_set_name()));
451 if (SurvivorRatio < 1) {
452 vm_exit_during_initialization("Invalid survivor ratio specified");
453 }
454 CollectorPolicy::initialize_flags();
455 }
456
457 // The easiest way to deal with the parsing of the NewSize /
458 // MaxNewSize / etc. parameteres is to re-use the code in the
459 // TwoGenerationCollectorPolicy class. This is similar to what
460 // ParallelScavenge does with its GenerationSizer class (see
461 // ParallelScavengeHeap::initialize()). We might change this in the
462 // future, but it's a good start.
463 class G1YoungGenSizer : public TwoGenerationCollectorPolicy {
464 private:
465 size_t size_to_region_num(size_t byte_size) {
466 return MAX2((size_t) 1, byte_size / HeapRegion::GrainBytes);
467 }
468
469 public:
470 G1YoungGenSizer() {
471 initialize_flags();
472 initialize_size_info();
473 }
474 size_t min_young_region_num() {
475 return size_to_region_num(_min_gen0_size);
476 }
477 size_t initial_young_region_num() {
478 return size_to_region_num(_initial_gen0_size);
479 }
480 size_t max_young_region_num() {
481 return size_to_region_num(_max_gen0_size);
482 }
483 };
484
485 void G1CollectorPolicy::update_young_list_size_using_newratio(size_t number_of_heap_regions) {
486 assert(number_of_heap_regions > 0, "Heap must be initialized");
487 size_t young_size = number_of_heap_regions / (NewRatio + 1);
488 _min_desired_young_length = young_size;
489 _max_desired_young_length = young_size;
490 }
491
492 void G1CollectorPolicy::init() {
493 // Set aside an initial future to_space.
494 _g1 = G1CollectedHeap::heap();
495
496 assert(Heap_lock->owned_by_self(), "Locking discipline.");
497
498 initialize_gc_policy_counters();
499
500 G1YoungGenSizer sizer;
501 size_t initial_region_num = sizer.initial_young_region_num();
502 _min_desired_young_length = sizer.min_young_region_num();
503 _max_desired_young_length = sizer.max_young_region_num();
504
505 if (FLAG_IS_CMDLINE(NewRatio)) {
506 if (FLAG_IS_CMDLINE(NewSize) || FLAG_IS_CMDLINE(MaxNewSize)) {
507 warning("-XX:NewSize and -XX:MaxNewSize override -XX:NewRatio");
508 } else {
509 // Treat NewRatio as a fixed size that is only recalculated when the heap size changes
510 update_young_list_size_using_newratio(_g1->n_regions());
511 _using_new_ratio_calculations = true;
512 }
513 }
514
515 // GenCollectorPolicy guarantees that min <= initial <= max.
516 // Asserting here just to state that we rely on this property.
517 assert(_min_desired_young_length <= _max_desired_young_length, "Invalid min/max young gen size values");
518 assert(initial_region_num <= _max_desired_young_length, "Initial young gen size too large");
519 assert(_min_desired_young_length <= initial_region_num, "Initial young gen size too small");
520
521 set_adaptive_young_list_length(_min_desired_young_length < _max_desired_young_length);
522 if (adaptive_young_list_length()) {
523 _young_list_fixed_length = 0;
524 } else {
525 _young_list_fixed_length = initial_region_num;
526 }
527 _free_regions_at_end_of_collection = _g1->free_regions();
528 update_young_list_target_length();
529 _prev_eden_capacity = _young_list_target_length * HeapRegion::GrainBytes;
530
531 // We may immediately start allocating regions and placing them on the
532 // collection set list. Initialize the per-collection set info
533 start_incremental_cset_building();
534 }
535
536 // Create the jstat counters for the policy.
537 void G1CollectorPolicy::initialize_gc_policy_counters() {
538 _gc_policy_counters = new GCPolicyCounters("GarbageFirst", 1, 3);
539 }
540
541 bool G1CollectorPolicy::predict_will_fit(size_t young_length,
542 double base_time_ms,
543 size_t base_free_regions,
544 double target_pause_time_ms) {
545 if (young_length >= base_free_regions) {
546 // end condition 1: not enough space for the young regions
547 return false;
548 }
549
550 double accum_surv_rate = accum_yg_surv_rate_pred((int)(young_length - 1));
551 size_t bytes_to_copy =
552 (size_t) (accum_surv_rate * (double) HeapRegion::GrainBytes);
553 double copy_time_ms = predict_object_copy_time_ms(bytes_to_copy);
554 double young_other_time_ms = predict_young_other_time_ms(young_length);
555 double pause_time_ms = base_time_ms + copy_time_ms + young_other_time_ms;
556 if (pause_time_ms > target_pause_time_ms) {
557 // end condition 2: prediction is over the target pause time
558 return false;
559 }
560
561 size_t free_bytes =
562 (base_free_regions - young_length) * HeapRegion::GrainBytes;
563 if ((2.0 * sigma()) * (double) bytes_to_copy > (double) free_bytes) {
564 // end condition 3: out-of-space (conservatively!)
565 return false;
566 }
567
568 // success!
569 return true;
570 }
571
572 void G1CollectorPolicy::record_new_heap_size(size_t new_number_of_regions) {
573 // re-calculate the necessary reserve
574 double reserve_regions_d = (double) new_number_of_regions * _reserve_factor;
575 // We use ceiling so that if reserve_regions_d is > 0.0 (but
576 // smaller than 1.0) we'll get 1.
577 _reserve_regions = (size_t) ceil(reserve_regions_d);
578
579 if (_using_new_ratio_calculations) {
580 // -XX:NewRatio was specified so we need to update the
581 // young gen length when the heap size has changed.
582 update_young_list_size_using_newratio(new_number_of_regions);
583 }
584 }
585
586 size_t G1CollectorPolicy::calculate_young_list_desired_min_length(
587 size_t base_min_length) {
588 size_t desired_min_length = 0;
589 if (adaptive_young_list_length()) {
590 if (_alloc_rate_ms_seq->num() > 3) {
591 double now_sec = os::elapsedTime();
592 double when_ms = _mmu_tracker->when_max_gc_sec(now_sec) * 1000.0;
593 double alloc_rate_ms = predict_alloc_rate_ms();
594 desired_min_length = (size_t) ceil(alloc_rate_ms * when_ms);
595 } else {
596 // otherwise we don't have enough info to make the prediction
597 }
598 }
599 desired_min_length += base_min_length;
600 // make sure we don't go below any user-defined minimum bound
601 return MAX2(_min_desired_young_length, desired_min_length);
602 }
603
604 size_t G1CollectorPolicy::calculate_young_list_desired_max_length() {
605 // Here, we might want to also take into account any additional
606 // constraints (i.e., user-defined minimum bound). Currently, we
607 // effectively don't set this bound.
608 return _max_desired_young_length;
609 }
610
611 void G1CollectorPolicy::update_young_list_target_length(size_t rs_lengths) {
612 if (rs_lengths == (size_t) -1) {
613 // if it's set to the default value (-1), we should predict it;
614 // otherwise, use the given value.
615 rs_lengths = (size_t) get_new_prediction(_rs_lengths_seq);
616 }
617
618 // Calculate the absolute and desired min bounds.
619
620 // This is how many young regions we already have (currently: the survivors).
621 size_t base_min_length = recorded_survivor_regions();
622 // This is the absolute minimum young length, which ensures that we
623 // can allocate one eden region in the worst-case.
624 size_t absolute_min_length = base_min_length + 1;
625 size_t desired_min_length =
626 calculate_young_list_desired_min_length(base_min_length);
627 if (desired_min_length < absolute_min_length) {
628 desired_min_length = absolute_min_length;
629 }
630
631 // Calculate the absolute and desired max bounds.
632
633 // We will try our best not to "eat" into the reserve.
634 size_t absolute_max_length = 0;
635 if (_free_regions_at_end_of_collection > _reserve_regions) {
636 absolute_max_length = _free_regions_at_end_of_collection - _reserve_regions;
637 }
638 size_t desired_max_length = calculate_young_list_desired_max_length();
639 if (desired_max_length > absolute_max_length) {
640 desired_max_length = absolute_max_length;
641 }
642
643 size_t young_list_target_length = 0;
644 if (adaptive_young_list_length()) {
645 if (full_young_gcs()) {
646 young_list_target_length =
647 calculate_young_list_target_length(rs_lengths,
648 base_min_length,
649 desired_min_length,
650 desired_max_length);
651 _rs_lengths_prediction = rs_lengths;
652 } else {
653 // Don't calculate anything and let the code below bound it to
654 // the desired_min_length, i.e., do the next GC as soon as
655 // possible to maximize how many old regions we can add to it.
656 }
657 } else {
658 if (full_young_gcs()) {
659 young_list_target_length = _young_list_fixed_length;
660 } else {
661 // A bit arbitrary: during partially-young GCs we allocate half
662 // the young regions to try to add old regions to the CSet.
663 young_list_target_length = _young_list_fixed_length / 2;
664 // We choose to accept that we might go under the desired min
665 // length given that we intentionally ask for a smaller young gen.
666 desired_min_length = absolute_min_length;
667 }
668 }
669
670 // Make sure we don't go over the desired max length, nor under the
671 // desired min length. In case they clash, desired_min_length wins
672 // which is why that test is second.
673 if (young_list_target_length > desired_max_length) {
674 young_list_target_length = desired_max_length;
675 }
676 if (young_list_target_length < desired_min_length) {
677 young_list_target_length = desired_min_length;
678 }
679
680 assert(young_list_target_length > recorded_survivor_regions(),
681 "we should be able to allocate at least one eden region");
682 assert(young_list_target_length >= absolute_min_length, "post-condition");
683 _young_list_target_length = young_list_target_length;
684
685 update_max_gc_locker_expansion();
686 }
687
688 size_t
689 G1CollectorPolicy::calculate_young_list_target_length(size_t rs_lengths,
690 size_t base_min_length,
691 size_t desired_min_length,
692 size_t desired_max_length) {
693 assert(adaptive_young_list_length(), "pre-condition");
694 assert(full_young_gcs(), "only call this for fully-young GCs");
695
696 // In case some edge-condition makes the desired max length too small...
697 if (desired_max_length <= desired_min_length) {
698 return desired_min_length;
699 }
700
701 // We'll adjust min_young_length and max_young_length not to include
702 // the already allocated young regions (i.e., so they reflect the
703 // min and max eden regions we'll allocate). The base_min_length
704 // will be reflected in the predictions by the
705 // survivor_regions_evac_time prediction.
706 assert(desired_min_length > base_min_length, "invariant");
707 size_t min_young_length = desired_min_length - base_min_length;
708 assert(desired_max_length > base_min_length, "invariant");
709 size_t max_young_length = desired_max_length - base_min_length;
710
711 double target_pause_time_ms = _mmu_tracker->max_gc_time() * 1000.0;
712 double survivor_regions_evac_time = predict_survivor_regions_evac_time();
713 size_t pending_cards = (size_t) get_new_prediction(_pending_cards_seq);
714 size_t adj_rs_lengths = rs_lengths + predict_rs_length_diff();
715 size_t scanned_cards = predict_young_card_num(adj_rs_lengths);
716 double base_time_ms =
717 predict_base_elapsed_time_ms(pending_cards, scanned_cards) +
718 survivor_regions_evac_time;
719 size_t available_free_regions = _free_regions_at_end_of_collection;
720 size_t base_free_regions = 0;
721 if (available_free_regions > _reserve_regions) {
722 base_free_regions = available_free_regions - _reserve_regions;
723 }
724
725 // Here, we will make sure that the shortest young length that
726 // makes sense fits within the target pause time.
727
728 if (predict_will_fit(min_young_length, base_time_ms,
729 base_free_regions, target_pause_time_ms)) {
730 // The shortest young length will fit into the target pause time;
731 // we'll now check whether the absolute maximum number of young
732 // regions will fit in the target pause time. If not, we'll do
733 // a binary search between min_young_length and max_young_length.
734 if (predict_will_fit(max_young_length, base_time_ms,
735 base_free_regions, target_pause_time_ms)) {
736 // The maximum young length will fit into the target pause time.
737 // We are done so set min young length to the maximum length (as
738 // the result is assumed to be returned in min_young_length).
739 min_young_length = max_young_length;
740 } else {
741 // The maximum possible number of young regions will not fit within
742 // the target pause time so we'll search for the optimal
743 // length. The loop invariants are:
744 //
745 // min_young_length < max_young_length
746 // min_young_length is known to fit into the target pause time
747 // max_young_length is known not to fit into the target pause time
748 //
749 // Going into the loop we know the above hold as we've just
750 // checked them. Every time around the loop we check whether
751 // the middle value between min_young_length and
752 // max_young_length fits into the target pause time. If it
753 // does, it becomes the new min. If it doesn't, it becomes
754 // the new max. This way we maintain the loop invariants.
755
756 assert(min_young_length < max_young_length, "invariant");
757 size_t diff = (max_young_length - min_young_length) / 2;
758 while (diff > 0) {
759 size_t young_length = min_young_length + diff;
760 if (predict_will_fit(young_length, base_time_ms,
761 base_free_regions, target_pause_time_ms)) {
762 min_young_length = young_length;
763 } else {
764 max_young_length = young_length;
765 }
766 assert(min_young_length < max_young_length, "invariant");
767 diff = (max_young_length - min_young_length) / 2;
768 }
769 // The results is min_young_length which, according to the
770 // loop invariants, should fit within the target pause time.
771
772 // These are the post-conditions of the binary search above:
773 assert(min_young_length < max_young_length,
774 "otherwise we should have discovered that max_young_length "
775 "fits into the pause target and not done the binary search");
776 assert(predict_will_fit(min_young_length, base_time_ms,
777 base_free_regions, target_pause_time_ms),
778 "min_young_length, the result of the binary search, should "
779 "fit into the pause target");
780 assert(!predict_will_fit(min_young_length + 1, base_time_ms,
781 base_free_regions, target_pause_time_ms),
782 "min_young_length, the result of the binary search, should be "
783 "optimal, so no larger length should fit into the pause target");
784 }
785 } else {
786 // Even the minimum length doesn't fit into the pause time
787 // target, return it as the result nevertheless.
788 }
789 return base_min_length + min_young_length;
790 }
791
792 double G1CollectorPolicy::predict_survivor_regions_evac_time() {
793 double survivor_regions_evac_time = 0.0;
794 for (HeapRegion * r = _recorded_survivor_head;
795 r != NULL && r != _recorded_survivor_tail->get_next_young_region();
796 r = r->get_next_young_region()) {
797 survivor_regions_evac_time += predict_region_elapsed_time_ms(r, true);
798 }
799 return survivor_regions_evac_time;
800 }
801
802 void G1CollectorPolicy::revise_young_list_target_length_if_necessary() {
803 guarantee( adaptive_young_list_length(), "should not call this otherwise" );
804
805 size_t rs_lengths = _g1->young_list()->sampled_rs_lengths();
806 if (rs_lengths > _rs_lengths_prediction) {
807 // add 10% to avoid having to recalculate often
808 size_t rs_lengths_prediction = rs_lengths * 1100 / 1000;
809 update_young_list_target_length(rs_lengths_prediction);
810 }
811 }
812
813
814
815 HeapWord* G1CollectorPolicy::mem_allocate_work(size_t size,
816 bool is_tlab,
817 bool* gc_overhead_limit_was_exceeded) {
818 guarantee(false, "Not using this policy feature yet.");
819 return NULL;
820 }
821
822 // This method controls how a collector handles one or more
823 // of its generations being fully allocated.
824 HeapWord* G1CollectorPolicy::satisfy_failed_allocation(size_t size,
825 bool is_tlab) {
826 guarantee(false, "Not using this policy feature yet.");
827 return NULL;
828 }
829
830
831 #ifndef PRODUCT
832 bool G1CollectorPolicy::verify_young_ages() {
833 HeapRegion* head = _g1->young_list()->first_region();
834 return
835 verify_young_ages(head, _short_lived_surv_rate_group);
836 // also call verify_young_ages on any additional surv rate groups
837 }
838
839 bool
840 G1CollectorPolicy::verify_young_ages(HeapRegion* head,
841 SurvRateGroup *surv_rate_group) {
842 guarantee( surv_rate_group != NULL, "pre-condition" );
843
844 const char* name = surv_rate_group->name();
845 bool ret = true;
846 int prev_age = -1;
847
848 for (HeapRegion* curr = head;
849 curr != NULL;
850 curr = curr->get_next_young_region()) {
851 SurvRateGroup* group = curr->surv_rate_group();
852 if (group == NULL && !curr->is_survivor()) {
853 gclog_or_tty->print_cr("## %s: encountered NULL surv_rate_group", name);
854 ret = false;
855 }
856
857 if (surv_rate_group == group) {
858 int age = curr->age_in_surv_rate_group();
859
860 if (age < 0) {
861 gclog_or_tty->print_cr("## %s: encountered negative age", name);
862 ret = false;
863 }
864
865 if (age <= prev_age) {
866 gclog_or_tty->print_cr("## %s: region ages are not strictly increasing "
867 "(%d, %d)", name, age, prev_age);
868 ret = false;
869 }
870 prev_age = age;
871 }
872 }
873
874 return ret;
875 }
876 #endif // PRODUCT
877
878 void G1CollectorPolicy::record_full_collection_start() {
879 _cur_collection_start_sec = os::elapsedTime();
880 // Release the future to-space so that it is available for compaction into.
881 _g1->set_full_collection();
882 }
883
884 void G1CollectorPolicy::record_full_collection_end() {
885 // Consider this like a collection pause for the purposes of allocation
886 // since last pause.
887 double end_sec = os::elapsedTime();
888 double full_gc_time_sec = end_sec - _cur_collection_start_sec;
889 double full_gc_time_ms = full_gc_time_sec * 1000.0;
890
891 _all_full_gc_times_ms->add(full_gc_time_ms);
892
893 update_recent_gc_times(end_sec, full_gc_time_ms);
894
895 _g1->clear_full_collection();
896
897 // "Nuke" the heuristics that control the fully/partially young GC
898 // transitions and make sure we start with fully young GCs after the
899 // Full GC.
900 set_full_young_gcs(true);
901 _last_full_young_gc = false;
902 _should_revert_to_full_young_gcs = false;
903 clear_initiate_conc_mark_if_possible();
904 clear_during_initial_mark_pause();
905 _known_garbage_bytes = 0;
906 _known_garbage_ratio = 0.0;
907 _in_marking_window = false;
908 _in_marking_window_im = false;
909
910 _short_lived_surv_rate_group->start_adding_regions();
911 // also call this on any additional surv rate groups
912
913 record_survivor_regions(0, NULL, NULL);
914
915 _prev_region_num_young = _region_num_young;
916 _prev_region_num_tenured = _region_num_tenured;
917
918 _free_regions_at_end_of_collection = _g1->free_regions();
919 // Reset survivors SurvRateGroup.
920 _survivor_surv_rate_group->reset();
921 update_young_list_target_length();
922 _collectionSetChooser->updateAfterFullCollection();
923 }
924
925 void G1CollectorPolicy::record_stop_world_start() {
926 _stop_world_start = os::elapsedTime();
927 }
928
929 void G1CollectorPolicy::record_collection_pause_start(double start_time_sec,
930 size_t start_used) {
931 if (PrintGCDetails) {
932 gclog_or_tty->stamp(PrintGCTimeStamps);
933 gclog_or_tty->print("[GC pause");
934 gclog_or_tty->print(" (%s)", full_young_gcs() ? "young" : "partial");
935 }
936
937 // We only need to do this here as the policy will only be applied
938 // to the GC we're about to start. so, no point is calculating this
939 // every time we calculate / recalculate the target young length.
940 update_survivors_policy();
941
942 assert(_g1->used() == _g1->recalculate_used(),
943 err_msg("sanity, used: "SIZE_FORMAT" recalculate_used: "SIZE_FORMAT,
944 _g1->used(), _g1->recalculate_used()));
945
946 double s_w_t_ms = (start_time_sec - _stop_world_start) * 1000.0;
947 _all_stop_world_times_ms->add(s_w_t_ms);
948 _stop_world_start = 0.0;
949
950 _cur_collection_start_sec = start_time_sec;
951 _cur_collection_pause_used_at_start_bytes = start_used;
952 _cur_collection_pause_used_regions_at_start = _g1->used_regions();
953 _pending_cards = _g1->pending_card_num();
954 _max_pending_cards = _g1->max_pending_card_num();
955
956 _bytes_in_collection_set_before_gc = 0;
957 _bytes_copied_during_gc = 0;
958
959 YoungList* young_list = _g1->young_list();
960 _eden_bytes_before_gc = young_list->eden_used_bytes();
961 _survivor_bytes_before_gc = young_list->survivor_used_bytes();
962 _capacity_before_gc = _g1->capacity();
963
964 #ifdef DEBUG
965 // initialise these to something well known so that we can spot
966 // if they are not set properly
967
968 for (int i = 0; i < _parallel_gc_threads; ++i) {
969 _par_last_gc_worker_start_times_ms[i] = -1234.0;
970 _par_last_ext_root_scan_times_ms[i] = -1234.0;
971 _par_last_mark_stack_scan_times_ms[i] = -1234.0;
972 _par_last_update_rs_times_ms[i] = -1234.0;
973 _par_last_update_rs_processed_buffers[i] = -1234.0;
974 _par_last_scan_rs_times_ms[i] = -1234.0;
975 _par_last_obj_copy_times_ms[i] = -1234.0;
976 _par_last_termination_times_ms[i] = -1234.0;
977 _par_last_termination_attempts[i] = -1234.0;
978 _par_last_gc_worker_end_times_ms[i] = -1234.0;
979 _par_last_gc_worker_times_ms[i] = -1234.0;
980 _par_last_gc_worker_other_times_ms[i] = -1234.0;
981 }
982 #endif
983
984 for (int i = 0; i < _aux_num; ++i) {
985 _cur_aux_times_ms[i] = 0.0;
986 _cur_aux_times_set[i] = false;
987 }
988
989 // These are initialized to zero here and they are set during
990 // the evacuation pause if marking is in progress.
991 _cur_satb_drain_time_ms = 0.0;
992 _last_satb_drain_processed_buffers = 0;
993
994 _last_young_gc_full = false;
995
996 // do that for any other surv rate groups
997 _short_lived_surv_rate_group->stop_adding_regions();
998 _survivors_age_table.clear();
999
1000 assert( verify_young_ages(), "region age verification" );
1001 }
1002
1003 void G1CollectorPolicy::record_mark_closure_time(double mark_closure_time_ms) {
1004 _mark_closure_time_ms = mark_closure_time_ms;
1005 }
1006
1007 void G1CollectorPolicy::record_concurrent_mark_init_end(double
1008 mark_init_elapsed_time_ms) {
1009 _during_marking = true;
1010 assert(!initiate_conc_mark_if_possible(), "we should have cleared it by now");
1011 clear_during_initial_mark_pause();
1012 _cur_mark_stop_world_time_ms = mark_init_elapsed_time_ms;
1013 }
1014
1015 void G1CollectorPolicy::record_concurrent_mark_remark_start() {
1016 _mark_remark_start_sec = os::elapsedTime();
1017 _during_marking = false;
1018 }
1019
1020 void G1CollectorPolicy::record_concurrent_mark_remark_end() {
1021 double end_time_sec = os::elapsedTime();
1022 double elapsed_time_ms = (end_time_sec - _mark_remark_start_sec)*1000.0;
1023 _concurrent_mark_remark_times_ms->add(elapsed_time_ms);
1024 _cur_mark_stop_world_time_ms += elapsed_time_ms;
1025 _prev_collection_pause_end_ms += elapsed_time_ms;
1026
1027 _mmu_tracker->add_pause(_mark_remark_start_sec, end_time_sec, true);
1028 }
1029
1030 void G1CollectorPolicy::record_concurrent_mark_cleanup_start() {
1031 _mark_cleanup_start_sec = os::elapsedTime();
1032 }
1033
1034 void G1CollectorPolicy::record_concurrent_mark_cleanup_completed() {
1035 _should_revert_to_full_young_gcs = false;
1036 _last_full_young_gc = true;
1037 _in_marking_window = false;
1038 }
1039
1040 void G1CollectorPolicy::record_concurrent_pause() {
1041 if (_stop_world_start > 0.0) {
1042 double yield_ms = (os::elapsedTime() - _stop_world_start) * 1000.0;
1043 _all_yield_times_ms->add(yield_ms);
1044 }
1045 }
1046
1047 void G1CollectorPolicy::record_concurrent_pause_end() {
1048 }
1049
1050 template<class T>
1051 T sum_of(T* sum_arr, int start, int n, int N) {
1052 T sum = (T)0;
1053 for (int i = 0; i < n; i++) {
1054 int j = (start + i) % N;
1055 sum += sum_arr[j];
1056 }
1057 return sum;
1058 }
1059
1060 void G1CollectorPolicy::print_par_stats(int level,
1061 const char* str,
1062 double* data) {
1063 double min = data[0], max = data[0];
1064 double total = 0.0;
1065 LineBuffer buf(level);
1066 buf.append("[%s (ms):", str);
1067 for (uint i = 0; i < ParallelGCThreads; ++i) {
1068 double val = data[i];
1069 if (val < min)
1070 min = val;
1071 if (val > max)
1072 max = val;
1073 total += val;
1074 buf.append(" %3.1lf", val);
1075 }
1076 buf.append_and_print_cr("");
1077 double avg = total / (double) ParallelGCThreads;
1078 buf.append_and_print_cr(" Avg: %5.1lf, Min: %5.1lf, Max: %5.1lf, Diff: %5.1lf]",
1079 avg, min, max, max - min);
1080 }
1081
1082 void G1CollectorPolicy::print_par_sizes(int level,
1083 const char* str,
1084 double* data) {
1085 double min = data[0], max = data[0];
1086 double total = 0.0;
1087 LineBuffer buf(level);
1088 buf.append("[%s :", str);
1089 for (uint i = 0; i < ParallelGCThreads; ++i) {
1090 double val = data[i];
1091 if (val < min)
1092 min = val;
1093 if (val > max)
1094 max = val;
1095 total += val;
1096 buf.append(" %d", (int) val);
1097 }
1098 buf.append_and_print_cr("");
1099 double avg = total / (double) ParallelGCThreads;
1100 buf.append_and_print_cr(" Sum: %d, Avg: %d, Min: %d, Max: %d, Diff: %d]",
1101 (int)total, (int)avg, (int)min, (int)max, (int)max - (int)min);
1102 }
1103
1104 void G1CollectorPolicy::print_stats(int level,
1105 const char* str,
1106 double value) {
1107 LineBuffer(level).append_and_print_cr("[%s: %5.1lf ms]", str, value);
1108 }
1109
1110 void G1CollectorPolicy::print_stats(int level,
1111 const char* str,
1112 int value) {
1113 LineBuffer(level).append_and_print_cr("[%s: %d]", str, value);
1114 }
1115
1116 double G1CollectorPolicy::avg_value(double* data) {
1117 if (G1CollectedHeap::use_parallel_gc_threads()) {
1118 double ret = 0.0;
1119 for (uint i = 0; i < ParallelGCThreads; ++i) {
1120 ret += data[i];
1121 }
1122 return ret / (double) ParallelGCThreads;
1123 } else {
1124 return data[0];
1125 }
1126 }
1127
1128 double G1CollectorPolicy::max_value(double* data) {
1129 if (G1CollectedHeap::use_parallel_gc_threads()) {
1130 double ret = data[0];
1131 for (uint i = 1; i < ParallelGCThreads; ++i) {
1132 if (data[i] > ret) {
1133 ret = data[i];
1134 }
1135 }
1136 return ret;
1137 } else {
1138 return data[0];
1139 }
1140 }
1141
1142 double G1CollectorPolicy::sum_of_values(double* data) {
1143 if (G1CollectedHeap::use_parallel_gc_threads()) {
1144 double sum = 0.0;
1145 for (uint i = 0; i < ParallelGCThreads; i++) {
1146 sum += data[i];
1147 }
1148 return sum;
1149 } else {
1150 return data[0];
1151 }
1152 }
1153
1154 double G1CollectorPolicy::max_sum(double* data1, double* data2) {
1155 double ret = data1[0] + data2[0];
1156
1157 if (G1CollectedHeap::use_parallel_gc_threads()) {
1158 for (uint i = 1; i < ParallelGCThreads; ++i) {
1159 double data = data1[i] + data2[i];
1160 if (data > ret) {
1161 ret = data;
1162 }
1163 }
1164 }
1165 return ret;
1166 }
1167
1168 // Anything below that is considered to be zero
1169 #define MIN_TIMER_GRANULARITY 0.0000001
1170
1171 void G1CollectorPolicy::record_collection_pause_end() {
1172 double end_time_sec = os::elapsedTime();
1173 double elapsed_ms = _last_pause_time_ms;
1174 bool parallel = G1CollectedHeap::use_parallel_gc_threads();
1175 size_t rs_size =
1176 _cur_collection_pause_used_regions_at_start - collection_set_size();
1177 size_t cur_used_bytes = _g1->used();
1178 assert(cur_used_bytes == _g1->recalculate_used(), "It should!");
1179 bool last_pause_included_initial_mark = false;
1180 bool update_stats = !_g1->evacuation_failed();
1181
1182 #ifndef PRODUCT
1183 if (G1YoungSurvRateVerbose) {
1184 gclog_or_tty->print_cr("");
1185 _short_lived_surv_rate_group->print();
1186 // do that for any other surv rate groups too
1187 }
1188 #endif // PRODUCT
1189
1190 last_pause_included_initial_mark = during_initial_mark_pause();
1191 if (last_pause_included_initial_mark)
1192 record_concurrent_mark_init_end(0.0);
1193
1194 size_t marking_initiating_used_threshold =
1195 (_g1->capacity() / 100) * InitiatingHeapOccupancyPercent;
1196
1197 if (!_g1->mark_in_progress() && !_last_full_young_gc) {
1198 assert(!last_pause_included_initial_mark, "invariant");
1199 if (cur_used_bytes > marking_initiating_used_threshold) {
1200 if (cur_used_bytes > _prev_collection_pause_used_at_end_bytes) {
1201 assert(!during_initial_mark_pause(), "we should not see this here");
1202
1203 ergo_verbose3(ErgoConcCycles,
1204 "request concurrent cycle initiation",
1205 ergo_format_reason("occupancy higher than threshold")
1206 ergo_format_byte("occupancy")
1207 ergo_format_byte_perc("threshold"),
1208 cur_used_bytes,
1209 marking_initiating_used_threshold,
1210 (double) InitiatingHeapOccupancyPercent);
1211
1212 // Note: this might have already been set, if during the last
1213 // pause we decided to start a cycle but at the beginning of
1214 // this pause we decided to postpone it. That's OK.
1215 set_initiate_conc_mark_if_possible();
1216 } else {
1217 ergo_verbose2(ErgoConcCycles,
1218 "do not request concurrent cycle initiation",
1219 ergo_format_reason("occupancy lower than previous occupancy")
1220 ergo_format_byte("occupancy")
1221 ergo_format_byte("previous occupancy"),
1222 cur_used_bytes,
1223 _prev_collection_pause_used_at_end_bytes);
1224 }
1225 }
1226 }
1227
1228 _prev_collection_pause_used_at_end_bytes = cur_used_bytes;
1229
1230 _mmu_tracker->add_pause(end_time_sec - elapsed_ms/1000.0,
1231 end_time_sec, false);
1232
1233 guarantee(_cur_collection_pause_used_regions_at_start >=
1234 collection_set_size(),
1235 "Negative RS size?");
1236
1237 // This assert is exempted when we're doing parallel collection pauses,
1238 // because the fragmentation caused by the parallel GC allocation buffers
1239 // can lead to more memory being used during collection than was used
1240 // before. Best leave this out until the fragmentation problem is fixed.
1241 // Pauses in which evacuation failed can also lead to negative
1242 // collections, since no space is reclaimed from a region containing an
1243 // object whose evacuation failed.
1244 // Further, we're now always doing parallel collection. But I'm still
1245 // leaving this here as a placeholder for a more precise assertion later.
1246 // (DLD, 10/05.)
1247 assert((true || parallel) // Always using GC LABs now.
1248 || _g1->evacuation_failed()
1249 || _cur_collection_pause_used_at_start_bytes >= cur_used_bytes,
1250 "Negative collection");
1251
1252 size_t freed_bytes =
1253 _cur_collection_pause_used_at_start_bytes - cur_used_bytes;
1254 size_t surviving_bytes = _collection_set_bytes_used_before - freed_bytes;
1255
1256 double survival_fraction =
1257 (double)surviving_bytes/
1258 (double)_collection_set_bytes_used_before;
1259
1260 _n_pauses++;
1261
1262 // These values are used to update the summary information that is
1263 // displayed when TraceGen0Time is enabled, and are output as part
1264 // of the PrintGCDetails output, in the non-parallel case.
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 known_time = ext_root_scan_time +
1276 mark_stack_scan_time +
1277 update_rs_time +
1278 scan_rs_time +
1279 obj_copy_time;
1280
1281 double other_time_ms = elapsed_ms;
1282
1283 // Subtract the SATB drain time. It's initialized to zero at the
1284 // start of the pause and is updated during the pause if marking
1285 // is in progress.
1286 other_time_ms -= _cur_satb_drain_time_ms;
1287
1288 if (parallel) {
1289 other_time_ms -= _cur_collection_par_time_ms;
1290 } else {
1291 other_time_ms -= known_time;
1292 }
1293
1294 // Subtract the time taken to clean the card table from the
1295 // current value of "other time"
1296 other_time_ms -= _cur_clear_ct_time_ms;
1297
1298 // TraceGen0Time and TraceGen1Time summary info updating.
1299 _all_pause_times_ms->add(elapsed_ms);
1300
1301 if (update_stats) {
1302 _recent_rs_scan_times_ms->add(scan_rs_time);
1303 _recent_pause_times_ms->add(elapsed_ms);
1304 _recent_rs_sizes->add(rs_size);
1305
1306 _summary->record_total_time_ms(elapsed_ms);
1307 _summary->record_other_time_ms(other_time_ms);
1308
1309 MainBodySummary* body_summary = _summary->main_body_summary();
1310 assert(body_summary != NULL, "should not be null!");
1311
1312 // This will be non-zero iff marking is currently in progress (i.e.
1313 // _g1->mark_in_progress() == true) and the currrent pause was not
1314 // an initial mark pause. Since the body_summary items are NumberSeqs,
1315 // however, they have to be consistent and updated in lock-step with
1316 // each other. Therefore we unconditionally record the SATB drain
1317 // time - even if it's zero.
1318 body_summary->record_satb_drain_time_ms(_cur_satb_drain_time_ms);
1319
1320 body_summary->record_ext_root_scan_time_ms(ext_root_scan_time);
1321 body_summary->record_mark_stack_scan_time_ms(mark_stack_scan_time);
1322 body_summary->record_update_rs_time_ms(update_rs_time);
1323 body_summary->record_scan_rs_time_ms(scan_rs_time);
1324 body_summary->record_obj_copy_time_ms(obj_copy_time);
1325
1326 if (parallel) {
1327 body_summary->record_parallel_time_ms(_cur_collection_par_time_ms);
1328 body_summary->record_termination_time_ms(termination_time);
1329
1330 double parallel_known_time = known_time + termination_time;
1331 double parallel_other_time = _cur_collection_par_time_ms - parallel_known_time;
1332 body_summary->record_parallel_other_time_ms(parallel_other_time);
1333 }
1334
1335 body_summary->record_mark_closure_time_ms(_mark_closure_time_ms);
1336 body_summary->record_clear_ct_time_ms(_cur_clear_ct_time_ms);
1337
1338 // We exempt parallel collection from this check because Alloc Buffer
1339 // fragmentation can produce negative collections. Same with evac
1340 // failure.
1341 // Further, we're now always doing parallel collection. But I'm still
1342 // leaving this here as a placeholder for a more precise assertion later.
1343 // (DLD, 10/05.
1344 assert((true || parallel)
1345 || _g1->evacuation_failed()
1346 || surviving_bytes <= _collection_set_bytes_used_before,
1347 "Or else negative collection!");
1348
1349 _recent_CS_bytes_used_before->add(_collection_set_bytes_used_before);
1350 _recent_CS_bytes_surviving->add(surviving_bytes);
1351
1352 // this is where we update the allocation rate of the application
1353 double app_time_ms =
1354 (_cur_collection_start_sec * 1000.0 - _prev_collection_pause_end_ms);
1355 if (app_time_ms < MIN_TIMER_GRANULARITY) {
1356 // This usually happens due to the timer not having the required
1357 // granularity. Some Linuxes are the usual culprits.
1358 // We'll just set it to something (arbitrarily) small.
1359 app_time_ms = 1.0;
1360 }
1361 size_t regions_allocated =
1362 (_region_num_young - _prev_region_num_young) +
1363 (_region_num_tenured - _prev_region_num_tenured);
1364 double alloc_rate_ms = (double) regions_allocated / app_time_ms;
1365 _alloc_rate_ms_seq->add(alloc_rate_ms);
1366 _prev_region_num_young = _region_num_young;
1367 _prev_region_num_tenured = _region_num_tenured;
1368
1369 double interval_ms =
1370 (end_time_sec - _recent_prev_end_times_for_all_gcs_sec->oldest()) * 1000.0;
1371 update_recent_gc_times(end_time_sec, elapsed_ms);
1372 _recent_avg_pause_time_ratio = _recent_gc_times_ms->sum()/interval_ms;
1373 if (recent_avg_pause_time_ratio() < 0.0 ||
1374 (recent_avg_pause_time_ratio() - 1.0 > 0.0)) {
1375 #ifndef PRODUCT
1376 // Dump info to allow post-facto debugging
1377 gclog_or_tty->print_cr("recent_avg_pause_time_ratio() out of bounds");
1378 gclog_or_tty->print_cr("-------------------------------------------");
1379 gclog_or_tty->print_cr("Recent GC Times (ms):");
1380 _recent_gc_times_ms->dump();
1381 gclog_or_tty->print_cr("(End Time=%3.3f) Recent GC End Times (s):", end_time_sec);
1382 _recent_prev_end_times_for_all_gcs_sec->dump();
1383 gclog_or_tty->print_cr("GC = %3.3f, Interval = %3.3f, Ratio = %3.3f",
1384 _recent_gc_times_ms->sum(), interval_ms, recent_avg_pause_time_ratio());
1385 // In debug mode, terminate the JVM if the user wants to debug at this point.
1386 assert(!G1FailOnFPError, "Debugging data for CR 6898948 has been dumped above");
1387 #endif // !PRODUCT
1388 // Clip ratio between 0.0 and 1.0, and continue. This will be fixed in
1389 // CR 6902692 by redoing the manner in which the ratio is incrementally computed.
1390 if (_recent_avg_pause_time_ratio < 0.0) {
1391 _recent_avg_pause_time_ratio = 0.0;
1392 } else {
1393 assert(_recent_avg_pause_time_ratio - 1.0 > 0.0, "Ctl-point invariant");
1394 _recent_avg_pause_time_ratio = 1.0;
1395 }
1396 }
1397 }
1398
1399 for (int i = 0; i < _aux_num; ++i) {
1400 if (_cur_aux_times_set[i]) {
1401 _all_aux_times_ms[i].add(_cur_aux_times_ms[i]);
1402 }
1403 }
1404
1405
1406 if (G1PolicyVerbose > 1) {
1407 gclog_or_tty->print_cr(" Recording collection pause(%d)", _n_pauses);
1408 }
1409
1410 if (G1PolicyVerbose > 1) {
1411 gclog_or_tty->print_cr(" ET: %10.6f ms (avg: %10.6f ms)\n"
1412 " ET-RS: %10.6f ms (avg: %10.6f ms)\n"
1413 " |RS|: " SIZE_FORMAT,
1414 elapsed_ms, recent_avg_time_for_pauses_ms(),
1415 scan_rs_time, recent_avg_time_for_rs_scan_ms(),
1416 rs_size);
1417
1418 gclog_or_tty->print_cr(" Used at start: " SIZE_FORMAT"K"
1419 " At end " SIZE_FORMAT "K\n"
1420 " garbage : " SIZE_FORMAT "K"
1421 " of " SIZE_FORMAT "K\n"
1422 " survival : %6.2f%% (%6.2f%% avg)",
1423 _cur_collection_pause_used_at_start_bytes/K,
1424 _g1->used()/K, freed_bytes/K,
1425 _collection_set_bytes_used_before/K,
1426 survival_fraction*100.0,
1427 recent_avg_survival_fraction()*100.0);
1428 gclog_or_tty->print_cr(" Recent %% gc pause time: %6.2f",
1429 recent_avg_pause_time_ratio() * 100.0);
1430 }
1431
1432 // PrintGCDetails output
1433 if (PrintGCDetails) {
1434 bool print_marking_info =
1435 _g1->mark_in_progress() && !last_pause_included_initial_mark;
1436
1437 gclog_or_tty->print_cr("%s, %1.8lf secs]",
1438 (last_pause_included_initial_mark) ? " (initial-mark)" : "",
1439 elapsed_ms / 1000.0);
1440
1441 if (print_marking_info) {
1442 print_stats(1, "SATB Drain Time", _cur_satb_drain_time_ms);
1443 print_stats(2, "Processed Buffers", _last_satb_drain_processed_buffers);
1444 }
1445
1446 if (parallel) {
1447 print_stats(1, "Parallel Time", _cur_collection_par_time_ms);
1448 print_par_stats(2, "GC Worker Start", _par_last_gc_worker_start_times_ms);
1449 print_par_stats(2, "Ext Root Scanning", _par_last_ext_root_scan_times_ms);
1450 if (print_marking_info) {
1451 print_par_stats(2, "Mark Stack Scanning", _par_last_mark_stack_scan_times_ms);
1452 }
1453 print_par_stats(2, "Update RS", _par_last_update_rs_times_ms);
1454 print_par_sizes(3, "Processed Buffers", _par_last_update_rs_processed_buffers);
1455 print_par_stats(2, "Scan RS", _par_last_scan_rs_times_ms);
1456 print_par_stats(2, "Object Copy", _par_last_obj_copy_times_ms);
1457 print_par_stats(2, "Termination", _par_last_termination_times_ms);
1458 print_par_sizes(3, "Termination Attempts", _par_last_termination_attempts);
1459 print_par_stats(2, "GC Worker End", _par_last_gc_worker_end_times_ms);
1460
1461 for (int i = 0; i < _parallel_gc_threads; i++) {
1462 _par_last_gc_worker_times_ms[i] = _par_last_gc_worker_end_times_ms[i] - _par_last_gc_worker_start_times_ms[i];
1463
1464 double worker_known_time = _par_last_ext_root_scan_times_ms[i] +
1465 _par_last_mark_stack_scan_times_ms[i] +
1466 _par_last_update_rs_times_ms[i] +
1467 _par_last_scan_rs_times_ms[i] +
1468 _par_last_obj_copy_times_ms[i] +
1469 _par_last_termination_times_ms[i];
1470
1471 _par_last_gc_worker_other_times_ms[i] = _cur_collection_par_time_ms - worker_known_time;
1472 }
1473 print_par_stats(2, "GC Worker", _par_last_gc_worker_times_ms);
1474 print_par_stats(2, "GC Worker Other", _par_last_gc_worker_other_times_ms);
1475 } else {
1476 print_stats(1, "Ext Root Scanning", ext_root_scan_time);
1477 if (print_marking_info) {
1478 print_stats(1, "Mark Stack Scanning", mark_stack_scan_time);
1479 }
1480 print_stats(1, "Update RS", update_rs_time);
1481 print_stats(2, "Processed Buffers", (int)update_rs_processed_buffers);
1482 print_stats(1, "Scan RS", scan_rs_time);
1483 print_stats(1, "Object Copying", obj_copy_time);
1484 }
1485 print_stats(1, "Clear CT", _cur_clear_ct_time_ms);
1486 #ifndef PRODUCT
1487 print_stats(1, "Cur Clear CC", _cur_clear_cc_time_ms);
1488 print_stats(1, "Cum Clear CC", _cum_clear_cc_time_ms);
1489 print_stats(1, "Min Clear CC", _min_clear_cc_time_ms);
1490 print_stats(1, "Max Clear CC", _max_clear_cc_time_ms);
1491 if (_num_cc_clears > 0) {
1492 print_stats(1, "Avg Clear CC", _cum_clear_cc_time_ms / ((double)_num_cc_clears));
1493 }
1494 #endif
1495 print_stats(1, "Other", other_time_ms);
1496 print_stats(2, "Choose CSet", _recorded_young_cset_choice_time_ms);
1497 print_stats(2, "Ref Proc", _cur_ref_proc_time_ms);
1498 print_stats(2, "Ref Enq", _cur_ref_enq_time_ms);
1499
1500 for (int i = 0; i < _aux_num; ++i) {
1501 if (_cur_aux_times_set[i]) {
1502 char buffer[96];
1503 sprintf(buffer, "Aux%d", i);
1504 print_stats(1, buffer, _cur_aux_times_ms[i]);
1505 }
1506 }
1507 }
1508
1509 // Update the efficiency-since-mark vars.
1510 double proc_ms = elapsed_ms * (double) _parallel_gc_threads;
1511 if (elapsed_ms < MIN_TIMER_GRANULARITY) {
1512 // This usually happens due to the timer not having the required
1513 // granularity. Some Linuxes are the usual culprits.
1514 // We'll just set it to something (arbitrarily) small.
1515 proc_ms = 1.0;
1516 }
1517 double cur_efficiency = (double) freed_bytes / proc_ms;
1518
1519 bool new_in_marking_window = _in_marking_window;
1520 bool new_in_marking_window_im = false;
1521 if (during_initial_mark_pause()) {
1522 new_in_marking_window = true;
1523 new_in_marking_window_im = true;
1524 }
1525
1526 if (_last_full_young_gc) {
1527 if (!last_pause_included_initial_mark) {
1528 ergo_verbose2(ErgoPartiallyYoungGCs,
1529 "start partially-young GCs",
1530 ergo_format_byte_perc("known garbage"),
1531 _known_garbage_bytes, _known_garbage_ratio * 100.0);
1532 set_full_young_gcs(false);
1533 } else {
1534 ergo_verbose0(ErgoPartiallyYoungGCs,
1535 "do not start partially-young GCs",
1536 ergo_format_reason("concurrent cycle is about to start"));
1537 }
1538 _last_full_young_gc = false;
1539 }
1540
1541 if ( !_last_young_gc_full ) {
1542 if (_should_revert_to_full_young_gcs) {
1543 ergo_verbose2(ErgoPartiallyYoungGCs,
1544 "end partially-young GCs",
1545 ergo_format_reason("partially-young GCs end requested")
1546 ergo_format_byte_perc("known garbage"),
1547 _known_garbage_bytes, _known_garbage_ratio * 100.0);
1548 set_full_young_gcs(true);
1549 } else if (_known_garbage_ratio < 0.05) {
1550 ergo_verbose3(ErgoPartiallyYoungGCs,
1551 "end partially-young GCs",
1552 ergo_format_reason("known garbage percent lower than threshold")
1553 ergo_format_byte_perc("known garbage")
1554 ergo_format_perc("threshold"),
1555 _known_garbage_bytes, _known_garbage_ratio * 100.0,
1556 0.05 * 100.0);
1557 set_full_young_gcs(true);
1558 } else if (adaptive_young_list_length() &&
1559 (get_gc_eff_factor() * cur_efficiency < predict_young_gc_eff())) {
1560 ergo_verbose5(ErgoPartiallyYoungGCs,
1561 "end partially-young GCs",
1562 ergo_format_reason("current GC efficiency lower than "
1563 "predicted fully-young GC efficiency")
1564 ergo_format_double("GC efficiency factor")
1565 ergo_format_double("current GC efficiency")
1566 ergo_format_double("predicted fully-young GC efficiency")
1567 ergo_format_byte_perc("known garbage"),
1568 get_gc_eff_factor(), cur_efficiency,
1569 predict_young_gc_eff(),
1570 _known_garbage_bytes, _known_garbage_ratio * 100.0);
1571 set_full_young_gcs(true);
1572 }
1573 }
1574 _should_revert_to_full_young_gcs = false;
1575
1576 if (_last_young_gc_full && !_during_marking) {
1577 _young_gc_eff_seq->add(cur_efficiency);
1578 }
1579
1580 _short_lived_surv_rate_group->start_adding_regions();
1581 // do that for any other surv rate groupsx
1582
1583 // <NEW PREDICTION>
1584
1585 if (update_stats) {
1586 double pause_time_ms = elapsed_ms;
1587
1588 size_t diff = 0;
1589 if (_max_pending_cards >= _pending_cards)
1590 diff = _max_pending_cards - _pending_cards;
1591 _pending_card_diff_seq->add((double) diff);
1592
1593 double cost_per_card_ms = 0.0;
1594 if (_pending_cards > 0) {
1595 cost_per_card_ms = update_rs_time / (double) _pending_cards;
1596 _cost_per_card_ms_seq->add(cost_per_card_ms);
1597 }
1598
1599 size_t cards_scanned = _g1->cards_scanned();
1600
1601 double cost_per_entry_ms = 0.0;
1602 if (cards_scanned > 10) {
1603 cost_per_entry_ms = scan_rs_time / (double) cards_scanned;
1604 if (_last_young_gc_full)
1605 _cost_per_entry_ms_seq->add(cost_per_entry_ms);
1606 else
1607 _partially_young_cost_per_entry_ms_seq->add(cost_per_entry_ms);
1608 }
1609
1610 if (_max_rs_lengths > 0) {
1611 double cards_per_entry_ratio =
1612 (double) cards_scanned / (double) _max_rs_lengths;
1613 if (_last_young_gc_full)
1614 _fully_young_cards_per_entry_ratio_seq->add(cards_per_entry_ratio);
1615 else
1616 _partially_young_cards_per_entry_ratio_seq->add(cards_per_entry_ratio);
1617 }
1618
1619 size_t rs_length_diff = _max_rs_lengths - _recorded_rs_lengths;
1620 if (rs_length_diff >= 0)
1621 _rs_length_diff_seq->add((double) rs_length_diff);
1622
1623 size_t copied_bytes = surviving_bytes;
1624 double cost_per_byte_ms = 0.0;
1625 if (copied_bytes > 0) {
1626 cost_per_byte_ms = obj_copy_time / (double) copied_bytes;
1627 if (_in_marking_window)
1628 _cost_per_byte_ms_during_cm_seq->add(cost_per_byte_ms);
1629 else
1630 _cost_per_byte_ms_seq->add(cost_per_byte_ms);
1631 }
1632
1633 double all_other_time_ms = pause_time_ms -
1634 (update_rs_time + scan_rs_time + obj_copy_time +
1635 _mark_closure_time_ms + termination_time);
1636
1637 double young_other_time_ms = 0.0;
1638 if (_recorded_young_regions > 0) {
1639 young_other_time_ms =
1640 _recorded_young_cset_choice_time_ms +
1641 _recorded_young_free_cset_time_ms;
1642 _young_other_cost_per_region_ms_seq->add(young_other_time_ms /
1643 (double) _recorded_young_regions);
1644 }
1645 double non_young_other_time_ms = 0.0;
1646 if (_recorded_non_young_regions > 0) {
1647 non_young_other_time_ms =
1648 _recorded_non_young_cset_choice_time_ms +
1649 _recorded_non_young_free_cset_time_ms;
1650
1651 _non_young_other_cost_per_region_ms_seq->add(non_young_other_time_ms /
1652 (double) _recorded_non_young_regions);
1653 }
1654
1655 double constant_other_time_ms = all_other_time_ms -
1656 (young_other_time_ms + non_young_other_time_ms);
1657 _constant_other_time_ms_seq->add(constant_other_time_ms);
1658
1659 double survival_ratio = 0.0;
1660 if (_bytes_in_collection_set_before_gc > 0) {
1661 survival_ratio = (double) _bytes_copied_during_gc /
1662 (double) _bytes_in_collection_set_before_gc;
1663 }
1664
1665 _pending_cards_seq->add((double) _pending_cards);
1666 _scanned_cards_seq->add((double) cards_scanned);
1667 _rs_lengths_seq->add((double) _max_rs_lengths);
1668
1669 double expensive_region_limit_ms =
1670 (double) MaxGCPauseMillis - predict_constant_other_time_ms();
1671 if (expensive_region_limit_ms < 0.0) {
1672 // this means that the other time was predicted to be longer than
1673 // than the max pause time
1674 expensive_region_limit_ms = (double) MaxGCPauseMillis;
1675 }
1676 _expensive_region_limit_ms = expensive_region_limit_ms;
1677
1678 if (PREDICTIONS_VERBOSE) {
1679 gclog_or_tty->print_cr("");
1680 gclog_or_tty->print_cr("PREDICTIONS %1.4lf %d "
1681 "REGIONS %d %d %d "
1682 "PENDING_CARDS %d %d "
1683 "CARDS_SCANNED %d %d "
1684 "RS_LENGTHS %d %d "
1685 "RS_UPDATE %1.6lf %1.6lf RS_SCAN %1.6lf %1.6lf "
1686 "SURVIVAL_RATIO %1.6lf %1.6lf "
1687 "OBJECT_COPY %1.6lf %1.6lf OTHER_CONSTANT %1.6lf %1.6lf "
1688 "OTHER_YOUNG %1.6lf %1.6lf "
1689 "OTHER_NON_YOUNG %1.6lf %1.6lf "
1690 "VTIME_DIFF %1.6lf TERMINATION %1.6lf "
1691 "ELAPSED %1.6lf %1.6lf ",
1692 _cur_collection_start_sec,
1693 (!_last_young_gc_full) ? 2 :
1694 (last_pause_included_initial_mark) ? 1 : 0,
1695 _recorded_region_num,
1696 _recorded_young_regions,
1697 _recorded_non_young_regions,
1698 _predicted_pending_cards, _pending_cards,
1699 _predicted_cards_scanned, cards_scanned,
1700 _predicted_rs_lengths, _max_rs_lengths,
1701 _predicted_rs_update_time_ms, update_rs_time,
1702 _predicted_rs_scan_time_ms, scan_rs_time,
1703 _predicted_survival_ratio, survival_ratio,
1704 _predicted_object_copy_time_ms, obj_copy_time,
1705 _predicted_constant_other_time_ms, constant_other_time_ms,
1706 _predicted_young_other_time_ms, young_other_time_ms,
1707 _predicted_non_young_other_time_ms,
1708 non_young_other_time_ms,
1709 _vtime_diff_ms, termination_time,
1710 _predicted_pause_time_ms, elapsed_ms);
1711 }
1712
1713 if (G1PolicyVerbose > 0) {
1714 gclog_or_tty->print_cr("Pause Time, predicted: %1.4lfms (predicted %s), actual: %1.4lfms",
1715 _predicted_pause_time_ms,
1716 (_within_target) ? "within" : "outside",
1717 elapsed_ms);
1718 }
1719
1720 }
1721
1722 _in_marking_window = new_in_marking_window;
1723 _in_marking_window_im = new_in_marking_window_im;
1724 _free_regions_at_end_of_collection = _g1->free_regions();
1725 update_young_list_target_length();
1726
1727 // Note that _mmu_tracker->max_gc_time() returns the time in seconds.
1728 double update_rs_time_goal_ms = _mmu_tracker->max_gc_time() * MILLIUNITS * G1RSetUpdatingPauseTimePercent / 100.0;
1729 adjust_concurrent_refinement(update_rs_time, update_rs_processed_buffers, update_rs_time_goal_ms);
1730 // </NEW PREDICTION>
1731
1732 assert(assertMarkedBytesDataOK(), "Marked regions not OK at pause end.");
1733 }
1734
1735 #define EXT_SIZE_FORMAT "%d%s"
1736 #define EXT_SIZE_PARAMS(bytes) \
1737 byte_size_in_proper_unit((bytes)), \
1738 proper_unit_for_byte_size((bytes))
1739
1740 void G1CollectorPolicy::print_heap_transition() {
1741 if (PrintGCDetails) {
1742 YoungList* young_list = _g1->young_list();
1743 size_t eden_bytes = young_list->eden_used_bytes();
1744 size_t survivor_bytes = young_list->survivor_used_bytes();
1745 size_t used_before_gc = _cur_collection_pause_used_at_start_bytes;
1746 size_t used = _g1->used();
1747 size_t capacity = _g1->capacity();
1748 size_t eden_capacity =
1749 (_young_list_target_length * HeapRegion::GrainBytes) - survivor_bytes;
1750
1751 gclog_or_tty->print_cr(
1752 " [Eden: "EXT_SIZE_FORMAT"("EXT_SIZE_FORMAT")->"EXT_SIZE_FORMAT"("EXT_SIZE_FORMAT") "
1753 "Survivors: "EXT_SIZE_FORMAT"->"EXT_SIZE_FORMAT" "
1754 "Heap: "EXT_SIZE_FORMAT"("EXT_SIZE_FORMAT")->"
1755 EXT_SIZE_FORMAT"("EXT_SIZE_FORMAT")]",
1756 EXT_SIZE_PARAMS(_eden_bytes_before_gc),
1757 EXT_SIZE_PARAMS(_prev_eden_capacity),
1758 EXT_SIZE_PARAMS(eden_bytes),
1759 EXT_SIZE_PARAMS(eden_capacity),
1760 EXT_SIZE_PARAMS(_survivor_bytes_before_gc),
1761 EXT_SIZE_PARAMS(survivor_bytes),
1762 EXT_SIZE_PARAMS(used_before_gc),
1763 EXT_SIZE_PARAMS(_capacity_before_gc),
1764 EXT_SIZE_PARAMS(used),
1765 EXT_SIZE_PARAMS(capacity));
1766
1767 _prev_eden_capacity = eden_capacity;
1768 } else if (PrintGC) {
1769 _g1->print_size_transition(gclog_or_tty,
1770 _cur_collection_pause_used_at_start_bytes,
1771 _g1->used(), _g1->capacity());
1772 }
1773 }
1774
1775 // <NEW PREDICTION>
1776
1777 void G1CollectorPolicy::adjust_concurrent_refinement(double update_rs_time,
1778 double update_rs_processed_buffers,
1779 double goal_ms) {
1780 DirtyCardQueueSet& dcqs = JavaThread::dirty_card_queue_set();
1781 ConcurrentG1Refine *cg1r = G1CollectedHeap::heap()->concurrent_g1_refine();
1782
1783 if (G1UseAdaptiveConcRefinement) {
1784 const int k_gy = 3, k_gr = 6;
1785 const double inc_k = 1.1, dec_k = 0.9;
1786
1787 int g = cg1r->green_zone();
1788 if (update_rs_time > goal_ms) {
1789 g = (int)(g * dec_k); // Can become 0, that's OK. That would mean a mutator-only processing.
1790 } else {
1791 if (update_rs_time < goal_ms && update_rs_processed_buffers > g) {
1792 g = (int)MAX2(g * inc_k, g + 1.0);
1793 }
1794 }
1795 // Change the refinement threads params
1796 cg1r->set_green_zone(g);
1797 cg1r->set_yellow_zone(g * k_gy);
1798 cg1r->set_red_zone(g * k_gr);
1799 cg1r->reinitialize_threads();
1800
1801 int processing_threshold_delta = MAX2((int)(cg1r->green_zone() * sigma()), 1);
1802 int processing_threshold = MIN2(cg1r->green_zone() + processing_threshold_delta,
1803 cg1r->yellow_zone());
1804 // Change the barrier params
1805 dcqs.set_process_completed_threshold(processing_threshold);
1806 dcqs.set_max_completed_queue(cg1r->red_zone());
1807 }
1808
1809 int curr_queue_size = dcqs.completed_buffers_num();
1810 if (curr_queue_size >= cg1r->yellow_zone()) {
1811 dcqs.set_completed_queue_padding(curr_queue_size);
1812 } else {
1813 dcqs.set_completed_queue_padding(0);
1814 }
1815 dcqs.notify_if_necessary();
1816 }
1817
1818 double
1819 G1CollectorPolicy::
1820 predict_young_collection_elapsed_time_ms(size_t adjustment) {
1821 guarantee( adjustment == 0 || adjustment == 1, "invariant" );
1822
1823 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1824 size_t young_num = g1h->young_list()->length();
1825 if (young_num == 0)
1826 return 0.0;
1827
1828 young_num += adjustment;
1829 size_t pending_cards = predict_pending_cards();
1830 size_t rs_lengths = g1h->young_list()->sampled_rs_lengths() +
1831 predict_rs_length_diff();
1832 size_t card_num;
1833 if (full_young_gcs())
1834 card_num = predict_young_card_num(rs_lengths);
1835 else
1836 card_num = predict_non_young_card_num(rs_lengths);
1837 size_t young_byte_size = young_num * HeapRegion::GrainBytes;
1838 double accum_yg_surv_rate =
1839 _short_lived_surv_rate_group->accum_surv_rate(adjustment);
1840
1841 size_t bytes_to_copy =
1842 (size_t) (accum_yg_surv_rate * (double) HeapRegion::GrainBytes);
1843
1844 return
1845 predict_rs_update_time_ms(pending_cards) +
1846 predict_rs_scan_time_ms(card_num) +
1847 predict_object_copy_time_ms(bytes_to_copy) +
1848 predict_young_other_time_ms(young_num) +
1849 predict_constant_other_time_ms();
1850 }
1851
1852 double
1853 G1CollectorPolicy::predict_base_elapsed_time_ms(size_t pending_cards) {
1854 size_t rs_length = predict_rs_length_diff();
1855 size_t card_num;
1856 if (full_young_gcs())
1857 card_num = predict_young_card_num(rs_length);
1858 else
1859 card_num = predict_non_young_card_num(rs_length);
1860 return predict_base_elapsed_time_ms(pending_cards, card_num);
1861 }
1862
1863 double
1864 G1CollectorPolicy::predict_base_elapsed_time_ms(size_t pending_cards,
1865 size_t scanned_cards) {
1866 return
1867 predict_rs_update_time_ms(pending_cards) +
1868 predict_rs_scan_time_ms(scanned_cards) +
1869 predict_constant_other_time_ms();
1870 }
1871
1872 double
1873 G1CollectorPolicy::predict_region_elapsed_time_ms(HeapRegion* hr,
1874 bool young) {
1875 size_t rs_length = hr->rem_set()->occupied();
1876 size_t card_num;
1877 if (full_young_gcs())
1878 card_num = predict_young_card_num(rs_length);
1879 else
1880 card_num = predict_non_young_card_num(rs_length);
1881 size_t bytes_to_copy = predict_bytes_to_copy(hr);
1882
1883 double region_elapsed_time_ms =
1884 predict_rs_scan_time_ms(card_num) +
1885 predict_object_copy_time_ms(bytes_to_copy);
1886
1887 if (young)
1888 region_elapsed_time_ms += predict_young_other_time_ms(1);
1889 else
1890 region_elapsed_time_ms += predict_non_young_other_time_ms(1);
1891
1892 return region_elapsed_time_ms;
1893 }
1894
1895 size_t
1896 G1CollectorPolicy::predict_bytes_to_copy(HeapRegion* hr) {
1897 size_t bytes_to_copy;
1898 if (hr->is_marked())
1899 bytes_to_copy = hr->max_live_bytes();
1900 else {
1901 guarantee( hr->is_young() && hr->age_in_surv_rate_group() != -1,
1902 "invariant" );
1903 int age = hr->age_in_surv_rate_group();
1904 double yg_surv_rate = predict_yg_surv_rate(age, hr->surv_rate_group());
1905 bytes_to_copy = (size_t) ((double) hr->used() * yg_surv_rate);
1906 }
1907
1908 return bytes_to_copy;
1909 }
1910
1911 void
1912 G1CollectorPolicy::start_recording_regions() {
1913 _recorded_rs_lengths = 0;
1914 _recorded_young_regions = 0;
1915 _recorded_non_young_regions = 0;
1916
1917 #if PREDICTIONS_VERBOSE
1918 _recorded_marked_bytes = 0;
1919 _recorded_young_bytes = 0;
1920 _predicted_bytes_to_copy = 0;
1921 _predicted_rs_lengths = 0;
1922 _predicted_cards_scanned = 0;
1923 #endif // PREDICTIONS_VERBOSE
1924 }
1925
1926 void
1927 G1CollectorPolicy::record_cset_region_info(HeapRegion* hr, bool young) {
1928 #if PREDICTIONS_VERBOSE
1929 if (!young) {
1930 _recorded_marked_bytes += hr->max_live_bytes();
1931 }
1932 _predicted_bytes_to_copy += predict_bytes_to_copy(hr);
1933 #endif // PREDICTIONS_VERBOSE
1934
1935 size_t rs_length = hr->rem_set()->occupied();
1936 _recorded_rs_lengths += rs_length;
1937 }
1938
1939 void
1940 G1CollectorPolicy::record_non_young_cset_region(HeapRegion* hr) {
1941 assert(!hr->is_young(), "should not call this");
1942 ++_recorded_non_young_regions;
1943 record_cset_region_info(hr, false);
1944 }
1945
1946 void
1947 G1CollectorPolicy::set_recorded_young_regions(size_t n_regions) {
1948 _recorded_young_regions = n_regions;
1949 }
1950
1951 void G1CollectorPolicy::set_recorded_young_bytes(size_t bytes) {
1952 #if PREDICTIONS_VERBOSE
1953 _recorded_young_bytes = bytes;
1954 #endif // PREDICTIONS_VERBOSE
1955 }
1956
1957 void G1CollectorPolicy::set_recorded_rs_lengths(size_t rs_lengths) {
1958 _recorded_rs_lengths = rs_lengths;
1959 }
1960
1961 void G1CollectorPolicy::set_predicted_bytes_to_copy(size_t bytes) {
1962 _predicted_bytes_to_copy = bytes;
1963 }
1964
1965 void
1966 G1CollectorPolicy::end_recording_regions() {
1967 // The _predicted_pause_time_ms field is referenced in code
1968 // not under PREDICTIONS_VERBOSE. Let's initialize it.
1969 _predicted_pause_time_ms = -1.0;
1970
1971 #if PREDICTIONS_VERBOSE
1972 _predicted_pending_cards = predict_pending_cards();
1973 _predicted_rs_lengths = _recorded_rs_lengths + predict_rs_length_diff();
1974 if (full_young_gcs())
1975 _predicted_cards_scanned += predict_young_card_num(_predicted_rs_lengths);
1976 else
1977 _predicted_cards_scanned +=
1978 predict_non_young_card_num(_predicted_rs_lengths);
1979 _recorded_region_num = _recorded_young_regions + _recorded_non_young_regions;
1980
1981 _predicted_rs_update_time_ms =
1982 predict_rs_update_time_ms(_g1->pending_card_num());
1983 _predicted_rs_scan_time_ms =
1984 predict_rs_scan_time_ms(_predicted_cards_scanned);
1985 _predicted_object_copy_time_ms =
1986 predict_object_copy_time_ms(_predicted_bytes_to_copy);
1987 _predicted_constant_other_time_ms =
1988 predict_constant_other_time_ms();
1989 _predicted_young_other_time_ms =
1990 predict_young_other_time_ms(_recorded_young_regions);
1991 _predicted_non_young_other_time_ms =
1992 predict_non_young_other_time_ms(_recorded_non_young_regions);
1993
1994 _predicted_pause_time_ms =
1995 _predicted_rs_update_time_ms +
1996 _predicted_rs_scan_time_ms +
1997 _predicted_object_copy_time_ms +
1998 _predicted_constant_other_time_ms +
1999 _predicted_young_other_time_ms +
2000 _predicted_non_young_other_time_ms;
2001 #endif // PREDICTIONS_VERBOSE
2002 }
2003
2004 void G1CollectorPolicy::check_if_region_is_too_expensive(double
2005 predicted_time_ms) {
2006 // I don't think we need to do this when in young GC mode since
2007 // marking will be initiated next time we hit the soft limit anyway...
2008 if (predicted_time_ms > _expensive_region_limit_ms) {
2009 ergo_verbose2(ErgoPartiallyYoungGCs,
2010 "request partially-young GCs end",
2011 ergo_format_reason("predicted region time higher than threshold")
2012 ergo_format_ms("predicted region time")
2013 ergo_format_ms("threshold"),
2014 predicted_time_ms, _expensive_region_limit_ms);
2015 // no point in doing another partial one
2016 _should_revert_to_full_young_gcs = true;
2017 }
2018 }
2019
2020 // </NEW PREDICTION>
2021
2022
2023 void G1CollectorPolicy::update_recent_gc_times(double end_time_sec,
2024 double elapsed_ms) {
2025 _recent_gc_times_ms->add(elapsed_ms);
2026 _recent_prev_end_times_for_all_gcs_sec->add(end_time_sec);
2027 _prev_collection_pause_end_ms = end_time_sec * 1000.0;
2028 }
2029
2030 double G1CollectorPolicy::recent_avg_time_for_pauses_ms() {
2031 if (_recent_pause_times_ms->num() == 0) {
2032 return (double) MaxGCPauseMillis;
2033 }
2034 return _recent_pause_times_ms->avg();
2035 }
2036
2037 double G1CollectorPolicy::recent_avg_time_for_rs_scan_ms() {
2038 if (_recent_rs_scan_times_ms->num() == 0) {
2039 return (double)MaxGCPauseMillis/3.0;
2040 }
2041 return _recent_rs_scan_times_ms->avg();
2042 }
2043
2044 int G1CollectorPolicy::number_of_recent_gcs() {
2045 assert(_recent_rs_scan_times_ms->num() ==
2046 _recent_pause_times_ms->num(), "Sequence out of sync");
2047 assert(_recent_pause_times_ms->num() ==
2048 _recent_CS_bytes_used_before->num(), "Sequence out of sync");
2049 assert(_recent_CS_bytes_used_before->num() ==
2050 _recent_CS_bytes_surviving->num(), "Sequence out of sync");
2051
2052 return _recent_pause_times_ms->num();
2053 }
2054
2055 double G1CollectorPolicy::recent_avg_survival_fraction() {
2056 return recent_avg_survival_fraction_work(_recent_CS_bytes_surviving,
2057 _recent_CS_bytes_used_before);
2058 }
2059
2060 double G1CollectorPolicy::last_survival_fraction() {
2061 return last_survival_fraction_work(_recent_CS_bytes_surviving,
2062 _recent_CS_bytes_used_before);
2063 }
2064
2065 double
2066 G1CollectorPolicy::recent_avg_survival_fraction_work(TruncatedSeq* surviving,
2067 TruncatedSeq* before) {
2068 assert(surviving->num() == before->num(), "Sequence out of sync");
2069 if (before->sum() > 0.0) {
2070 double recent_survival_rate = surviving->sum() / before->sum();
2071 // We exempt parallel collection from this check because Alloc Buffer
2072 // fragmentation can produce negative collections.
2073 // Further, we're now always doing parallel collection. But I'm still
2074 // leaving this here as a placeholder for a more precise assertion later.
2075 // (DLD, 10/05.)
2076 assert((true || G1CollectedHeap::use_parallel_gc_threads()) ||
2077 _g1->evacuation_failed() ||
2078 recent_survival_rate <= 1.0, "Or bad frac");
2079 return recent_survival_rate;
2080 } else {
2081 return 1.0; // Be conservative.
2082 }
2083 }
2084
2085 double
2086 G1CollectorPolicy::last_survival_fraction_work(TruncatedSeq* surviving,
2087 TruncatedSeq* before) {
2088 assert(surviving->num() == before->num(), "Sequence out of sync");
2089 if (surviving->num() > 0 && before->last() > 0.0) {
2090 double last_survival_rate = surviving->last() / before->last();
2091 // We exempt parallel collection from this check because Alloc Buffer
2092 // fragmentation can produce negative collections.
2093 // Further, we're now always doing parallel collection. But I'm still
2094 // leaving this here as a placeholder for a more precise assertion later.
2095 // (DLD, 10/05.)
2096 assert((true || G1CollectedHeap::use_parallel_gc_threads()) ||
2097 last_survival_rate <= 1.0, "Or bad frac");
2098 return last_survival_rate;
2099 } else {
2100 return 1.0;
2101 }
2102 }
2103
2104 static const int survival_min_obs = 5;
2105 static double survival_min_obs_limits[] = { 0.9, 0.7, 0.5, 0.3, 0.1 };
2106 static const double min_survival_rate = 0.1;
2107
2108 double
2109 G1CollectorPolicy::conservative_avg_survival_fraction_work(double avg,
2110 double latest) {
2111 double res = avg;
2112 if (number_of_recent_gcs() < survival_min_obs) {
2113 res = MAX2(res, survival_min_obs_limits[number_of_recent_gcs()]);
2114 }
2115 res = MAX2(res, latest);
2116 res = MAX2(res, min_survival_rate);
2117 // In the parallel case, LAB fragmentation can produce "negative
2118 // collections"; so can evac failure. Cap at 1.0
2119 res = MIN2(res, 1.0);
2120 return res;
2121 }
2122
2123 size_t G1CollectorPolicy::expansion_amount() {
2124 double recent_gc_overhead = recent_avg_pause_time_ratio() * 100.0;
2125 double threshold = _gc_overhead_perc;
2126 if (recent_gc_overhead > threshold) {
2127 // We will double the existing space, or take
2128 // G1ExpandByPercentOfAvailable % of the available expansion
2129 // space, whichever is smaller, bounded below by a minimum
2130 // expansion (unless that's all that's left.)
2131 const size_t min_expand_bytes = 1*M;
2132 size_t reserved_bytes = _g1->max_capacity();
2133 size_t committed_bytes = _g1->capacity();
2134 size_t uncommitted_bytes = reserved_bytes - committed_bytes;
2135 size_t expand_bytes;
2136 size_t expand_bytes_via_pct =
2137 uncommitted_bytes * G1ExpandByPercentOfAvailable / 100;
2138 expand_bytes = MIN2(expand_bytes_via_pct, committed_bytes);
2139 expand_bytes = MAX2(expand_bytes, min_expand_bytes);
2140 expand_bytes = MIN2(expand_bytes, uncommitted_bytes);
2141
2142 ergo_verbose5(ErgoHeapSizing,
2143 "attempt heap expansion",
2144 ergo_format_reason("recent GC overhead higher than "
2145 "threshold after GC")
2146 ergo_format_perc("recent GC overhead")
2147 ergo_format_perc("threshold")
2148 ergo_format_byte("uncommitted")
2149 ergo_format_byte_perc("calculated expansion amount"),
2150 recent_gc_overhead, threshold,
2151 uncommitted_bytes,
2152 expand_bytes_via_pct, (double) G1ExpandByPercentOfAvailable);
2153
2154 return expand_bytes;
2155 } else {
2156 return 0;
2157 }
2158 }
2159
2160 class CountCSClosure: public HeapRegionClosure {
2161 G1CollectorPolicy* _g1_policy;
2162 public:
2163 CountCSClosure(G1CollectorPolicy* g1_policy) :
2164 _g1_policy(g1_policy) {}
2165 bool doHeapRegion(HeapRegion* r) {
2166 _g1_policy->_bytes_in_collection_set_before_gc += r->used();
2167 return false;
2168 }
2169 };
2170
2171 void G1CollectorPolicy::count_CS_bytes_used() {
2172 CountCSClosure cs_closure(this);
2173 _g1->collection_set_iterate(&cs_closure);
2174 }
2175
2176 void G1CollectorPolicy::print_summary(int level,
2177 const char* str,
2178 NumberSeq* seq) const {
2179 double sum = seq->sum();
2180 LineBuffer(level + 1).append_and_print_cr("%-24s = %8.2lf s (avg = %8.2lf ms)",
2181 str, sum / 1000.0, seq->avg());
2182 }
2183
2184 void G1CollectorPolicy::print_summary_sd(int level,
2185 const char* str,
2186 NumberSeq* seq) const {
2187 print_summary(level, str, seq);
2188 LineBuffer(level + 6).append_and_print_cr("(num = %5d, std dev = %8.2lf ms, max = %8.2lf ms)",
2189 seq->num(), seq->sd(), seq->maximum());
2190 }
2191
2192 void G1CollectorPolicy::check_other_times(int level,
2193 NumberSeq* other_times_ms,
2194 NumberSeq* calc_other_times_ms) const {
2195 bool should_print = false;
2196 LineBuffer buf(level + 2);
2197
2198 double max_sum = MAX2(fabs(other_times_ms->sum()),
2199 fabs(calc_other_times_ms->sum()));
2200 double min_sum = MIN2(fabs(other_times_ms->sum()),
2201 fabs(calc_other_times_ms->sum()));
2202 double sum_ratio = max_sum / min_sum;
2203 if (sum_ratio > 1.1) {
2204 should_print = true;
2205 buf.append_and_print_cr("## CALCULATED OTHER SUM DOESN'T MATCH RECORDED ###");
2206 }
2207
2208 double max_avg = MAX2(fabs(other_times_ms->avg()),
2209 fabs(calc_other_times_ms->avg()));
2210 double min_avg = MIN2(fabs(other_times_ms->avg()),
2211 fabs(calc_other_times_ms->avg()));
2212 double avg_ratio = max_avg / min_avg;
2213 if (avg_ratio > 1.1) {
2214 should_print = true;
2215 buf.append_and_print_cr("## CALCULATED OTHER AVG DOESN'T MATCH RECORDED ###");
2216 }
2217
2218 if (other_times_ms->sum() < -0.01) {
2219 buf.append_and_print_cr("## RECORDED OTHER SUM IS NEGATIVE ###");
2220 }
2221
2222 if (other_times_ms->avg() < -0.01) {
2223 buf.append_and_print_cr("## RECORDED OTHER AVG IS NEGATIVE ###");
2224 }
2225
2226 if (calc_other_times_ms->sum() < -0.01) {
2227 should_print = true;
2228 buf.append_and_print_cr("## CALCULATED OTHER SUM IS NEGATIVE ###");
2229 }
2230
2231 if (calc_other_times_ms->avg() < -0.01) {
2232 should_print = true;
2233 buf.append_and_print_cr("## CALCULATED OTHER AVG IS NEGATIVE ###");
2234 }
2235
2236 if (should_print)
2237 print_summary(level, "Other(Calc)", calc_other_times_ms);
2238 }
2239
2240 void G1CollectorPolicy::print_summary(PauseSummary* summary) const {
2241 bool parallel = G1CollectedHeap::use_parallel_gc_threads();
2242 MainBodySummary* body_summary = summary->main_body_summary();
2243 if (summary->get_total_seq()->num() > 0) {
2244 print_summary_sd(0, "Evacuation Pauses", summary->get_total_seq());
2245 if (body_summary != NULL) {
2246 print_summary(1, "SATB Drain", body_summary->get_satb_drain_seq());
2247 if (parallel) {
2248 print_summary(1, "Parallel Time", body_summary->get_parallel_seq());
2249 print_summary(2, "Ext Root Scanning", body_summary->get_ext_root_scan_seq());
2250 print_summary(2, "Mark Stack Scanning", body_summary->get_mark_stack_scan_seq());
2251 print_summary(2, "Update RS", body_summary->get_update_rs_seq());
2252 print_summary(2, "Scan RS", body_summary->get_scan_rs_seq());
2253 print_summary(2, "Object Copy", body_summary->get_obj_copy_seq());
2254 print_summary(2, "Termination", body_summary->get_termination_seq());
2255 print_summary(2, "Parallel Other", body_summary->get_parallel_other_seq());
2256 {
2257 NumberSeq* other_parts[] = {
2258 body_summary->get_ext_root_scan_seq(),
2259 body_summary->get_mark_stack_scan_seq(),
2260 body_summary->get_update_rs_seq(),
2261 body_summary->get_scan_rs_seq(),
2262 body_summary->get_obj_copy_seq(),
2263 body_summary->get_termination_seq()
2264 };
2265 NumberSeq calc_other_times_ms(body_summary->get_parallel_seq(),
2266 6, other_parts);
2267 check_other_times(2, body_summary->get_parallel_other_seq(),
2268 &calc_other_times_ms);
2269 }
2270 } else {
2271 print_summary(1, "Ext Root Scanning", body_summary->get_ext_root_scan_seq());
2272 print_summary(1, "Mark Stack Scanning", body_summary->get_mark_stack_scan_seq());
2273 print_summary(1, "Update RS", body_summary->get_update_rs_seq());
2274 print_summary(1, "Scan RS", body_summary->get_scan_rs_seq());
2275 print_summary(1, "Object Copy", body_summary->get_obj_copy_seq());
2276 }
2277 }
2278 print_summary(1, "Mark Closure", body_summary->get_mark_closure_seq());
2279 print_summary(1, "Clear CT", body_summary->get_clear_ct_seq());
2280 print_summary(1, "Other", summary->get_other_seq());
2281 {
2282 if (body_summary != NULL) {
2283 NumberSeq calc_other_times_ms;
2284 if (parallel) {
2285 // parallel
2286 NumberSeq* other_parts[] = {
2287 body_summary->get_satb_drain_seq(),
2288 body_summary->get_parallel_seq(),
2289 body_summary->get_clear_ct_seq()
2290 };
2291 calc_other_times_ms = NumberSeq(summary->get_total_seq(),
2292 3, other_parts);
2293 } else {
2294 // serial
2295 NumberSeq* other_parts[] = {
2296 body_summary->get_satb_drain_seq(),
2297 body_summary->get_update_rs_seq(),
2298 body_summary->get_ext_root_scan_seq(),
2299 body_summary->get_mark_stack_scan_seq(),
2300 body_summary->get_scan_rs_seq(),
2301 body_summary->get_obj_copy_seq()
2302 };
2303 calc_other_times_ms = NumberSeq(summary->get_total_seq(),
2304 6, other_parts);
2305 }
2306 check_other_times(1, summary->get_other_seq(), &calc_other_times_ms);
2307 }
2308 }
2309 } else {
2310 LineBuffer(1).append_and_print_cr("none");
2311 }
2312 LineBuffer(0).append_and_print_cr("");
2313 }
2314
2315 void G1CollectorPolicy::print_tracing_info() const {
2316 if (TraceGen0Time) {
2317 gclog_or_tty->print_cr("ALL PAUSES");
2318 print_summary_sd(0, "Total", _all_pause_times_ms);
2319 gclog_or_tty->print_cr("");
2320 gclog_or_tty->print_cr("");
2321 gclog_or_tty->print_cr(" Full Young GC Pauses: %8d", _full_young_pause_num);
2322 gclog_or_tty->print_cr(" Partial Young GC Pauses: %8d", _partial_young_pause_num);
2323 gclog_or_tty->print_cr("");
2324
2325 gclog_or_tty->print_cr("EVACUATION PAUSES");
2326 print_summary(_summary);
2327
2328 gclog_or_tty->print_cr("MISC");
2329 print_summary_sd(0, "Stop World", _all_stop_world_times_ms);
2330 print_summary_sd(0, "Yields", _all_yield_times_ms);
2331 for (int i = 0; i < _aux_num; ++i) {
2332 if (_all_aux_times_ms[i].num() > 0) {
2333 char buffer[96];
2334 sprintf(buffer, "Aux%d", i);
2335 print_summary_sd(0, buffer, &_all_aux_times_ms[i]);
2336 }
2337 }
2338
2339 size_t all_region_num = _region_num_young + _region_num_tenured;
2340 gclog_or_tty->print_cr(" New Regions %8d, Young %8d (%6.2lf%%), "
2341 "Tenured %8d (%6.2lf%%)",
2342 all_region_num,
2343 _region_num_young,
2344 (double) _region_num_young / (double) all_region_num * 100.0,
2345 _region_num_tenured,
2346 (double) _region_num_tenured / (double) all_region_num * 100.0);
2347 }
2348 if (TraceGen1Time) {
2349 if (_all_full_gc_times_ms->num() > 0) {
2350 gclog_or_tty->print("\n%4d full_gcs: total time = %8.2f s",
2351 _all_full_gc_times_ms->num(),
2352 _all_full_gc_times_ms->sum() / 1000.0);
2353 gclog_or_tty->print_cr(" (avg = %8.2fms).", _all_full_gc_times_ms->avg());
2354 gclog_or_tty->print_cr(" [std. dev = %8.2f ms, max = %8.2f ms]",
2355 _all_full_gc_times_ms->sd(),
2356 _all_full_gc_times_ms->maximum());
2357 }
2358 }
2359 }
2360
2361 void G1CollectorPolicy::print_yg_surv_rate_info() const {
2362 #ifndef PRODUCT
2363 _short_lived_surv_rate_group->print_surv_rate_summary();
2364 // add this call for any other surv rate groups
2365 #endif // PRODUCT
2366 }
2367
2368 void G1CollectorPolicy::update_region_num(bool young) {
2369 if (young) {
2370 ++_region_num_young;
2371 } else {
2372 ++_region_num_tenured;
2373 }
2374 }
2375
2376 #ifndef PRODUCT
2377 // for debugging, bit of a hack...
2378 static char*
2379 region_num_to_mbs(int length) {
2380 static char buffer[64];
2381 double bytes = (double) (length * HeapRegion::GrainBytes);
2382 double mbs = bytes / (double) (1024 * 1024);
2383 sprintf(buffer, "%7.2lfMB", mbs);
2384 return buffer;
2385 }
2386 #endif // PRODUCT
2387
2388 size_t G1CollectorPolicy::max_regions(int purpose) {
2389 switch (purpose) {
2390 case GCAllocForSurvived:
2391 return _max_survivor_regions;
2392 case GCAllocForTenured:
2393 return REGIONS_UNLIMITED;
2394 default:
2395 ShouldNotReachHere();
2396 return REGIONS_UNLIMITED;
2397 };
2398 }
2399
2400 void G1CollectorPolicy::update_max_gc_locker_expansion() {
2401 size_t expansion_region_num = 0;
2402 if (GCLockerEdenExpansionPercent > 0) {
2403 double perc = (double) GCLockerEdenExpansionPercent / 100.0;
2404 double expansion_region_num_d = perc * (double) _young_list_target_length;
2405 // We use ceiling so that if expansion_region_num_d is > 0.0 (but
2406 // less than 1.0) we'll get 1.
2407 expansion_region_num = (size_t) ceil(expansion_region_num_d);
2408 } else {
2409 assert(expansion_region_num == 0, "sanity");
2410 }
2411 _young_list_max_length = _young_list_target_length + expansion_region_num;
2412 assert(_young_list_target_length <= _young_list_max_length, "post-condition");
2413 }
2414
2415 // Calculates survivor space parameters.
2416 void G1CollectorPolicy::update_survivors_policy() {
2417 double max_survivor_regions_d =
2418 (double) _young_list_target_length / (double) SurvivorRatio;
2419 // We use ceiling so that if max_survivor_regions_d is > 0.0 (but
2420 // smaller than 1.0) we'll get 1.
2421 _max_survivor_regions = (size_t) ceil(max_survivor_regions_d);
2422
2423 _tenuring_threshold = _survivors_age_table.compute_tenuring_threshold(
2424 HeapRegion::GrainWords * _max_survivor_regions);
2425 }
2426
2427 #ifndef PRODUCT
2428 class HRSortIndexIsOKClosure: public HeapRegionClosure {
2429 CollectionSetChooser* _chooser;
2430 public:
2431 HRSortIndexIsOKClosure(CollectionSetChooser* chooser) :
2432 _chooser(chooser) {}
2433
2434 bool doHeapRegion(HeapRegion* r) {
2435 if (!r->continuesHumongous()) {
2436 assert(_chooser->regionProperlyOrdered(r), "Ought to be.");
2437 }
2438 return false;
2439 }
2440 };
2441
2442 bool G1CollectorPolicy::assertMarkedBytesDataOK() {
2443 HRSortIndexIsOKClosure cl(_collectionSetChooser);
2444 _g1->heap_region_iterate(&cl);
2445 return true;
2446 }
2447 #endif
2448
2449 bool G1CollectorPolicy::force_initial_mark_if_outside_cycle(
2450 GCCause::Cause gc_cause) {
2451 bool during_cycle = _g1->concurrent_mark()->cmThread()->during_cycle();
2452 if (!during_cycle) {
2453 ergo_verbose1(ErgoConcCycles,
2454 "request concurrent cycle initiation",
2455 ergo_format_reason("requested by GC cause")
2456 ergo_format_str("GC cause"),
2457 GCCause::to_string(gc_cause));
2458 set_initiate_conc_mark_if_possible();
2459 return true;
2460 } else {
2461 ergo_verbose1(ErgoConcCycles,
2462 "do not request concurrent cycle initiation",
2463 ergo_format_reason("concurrent cycle already in progress")
2464 ergo_format_str("GC cause"),
2465 GCCause::to_string(gc_cause));
2466 return false;
2467 }
2468 }
2469
2470 void
2471 G1CollectorPolicy::decide_on_conc_mark_initiation() {
2472 // We are about to decide on whether this pause will be an
2473 // initial-mark pause.
2474
2475 // First, during_initial_mark_pause() should not be already set. We
2476 // will set it here if we have to. However, it should be cleared by
2477 // the end of the pause (it's only set for the duration of an
2478 // initial-mark pause).
2479 assert(!during_initial_mark_pause(), "pre-condition");
2480
2481 if (initiate_conc_mark_if_possible()) {
2482 // We had noticed on a previous pause that the heap occupancy has
2483 // gone over the initiating threshold and we should start a
2484 // concurrent marking cycle. So we might initiate one.
2485
2486 bool during_cycle = _g1->concurrent_mark()->cmThread()->during_cycle();
2487 if (!during_cycle) {
2488 // The concurrent marking thread is not "during a cycle", i.e.,
2489 // it has completed the last one. So we can go ahead and
2490 // initiate a new cycle.
2491
2492 set_during_initial_mark_pause();
2493 // We do not allow non-full young GCs during marking.
2494 if (!full_young_gcs()) {
2495 set_full_young_gcs(true);
2496 ergo_verbose0(ErgoPartiallyYoungGCs,
2497 "end partially-young GCs",
2498 ergo_format_reason("concurrent cycle is about to start"));
2499 }
2500
2501 // And we can now clear initiate_conc_mark_if_possible() as
2502 // we've already acted on it.
2503 clear_initiate_conc_mark_if_possible();
2504
2505 ergo_verbose0(ErgoConcCycles,
2506 "initiate concurrent cycle",
2507 ergo_format_reason("concurrent cycle initiation requested"));
2508 } else {
2509 // The concurrent marking thread is still finishing up the
2510 // previous cycle. If we start one right now the two cycles
2511 // overlap. In particular, the concurrent marking thread might
2512 // be in the process of clearing the next marking bitmap (which
2513 // we will use for the next cycle if we start one). Starting a
2514 // cycle now will be bad given that parts of the marking
2515 // information might get cleared by the marking thread. And we
2516 // cannot wait for the marking thread to finish the cycle as it
2517 // periodically yields while clearing the next marking bitmap
2518 // and, if it's in a yield point, it's waiting for us to
2519 // finish. So, at this point we will not start a cycle and we'll
2520 // let the concurrent marking thread complete the last one.
2521 ergo_verbose0(ErgoConcCycles,
2522 "do not initiate concurrent cycle",
2523 ergo_format_reason("concurrent cycle already in progress"));
2524 }
2525 }
2526 }
2527
2528 class KnownGarbageClosure: public HeapRegionClosure {
2529 CollectionSetChooser* _hrSorted;
2530
2531 public:
2532 KnownGarbageClosure(CollectionSetChooser* hrSorted) :
2533 _hrSorted(hrSorted)
2534 {}
2535
2536 bool doHeapRegion(HeapRegion* r) {
2537 // We only include humongous regions in collection
2538 // sets when concurrent mark shows that their contained object is
2539 // unreachable.
2540
2541 // Do we have any marking information for this region?
2542 if (r->is_marked()) {
2543 // We don't include humongous regions in collection
2544 // sets because we collect them immediately at the end of a marking
2545 // cycle. We also don't include young regions because we *must*
2546 // include them in the next collection pause.
2547 if (!r->isHumongous() && !r->is_young()) {
2548 _hrSorted->addMarkedHeapRegion(r);
2549 }
2550 }
2551 return false;
2552 }
2553 };
2554
2555 class ParKnownGarbageHRClosure: public HeapRegionClosure {
2556 CollectionSetChooser* _hrSorted;
2557 jint _marked_regions_added;
2558 jint _chunk_size;
2559 jint _cur_chunk_idx;
2560 jint _cur_chunk_end; // Cur chunk [_cur_chunk_idx, _cur_chunk_end)
2561 int _worker;
2562 int _invokes;
2563
2564 void get_new_chunk() {
2565 _cur_chunk_idx = _hrSorted->getParMarkedHeapRegionChunk(_chunk_size);
2566 _cur_chunk_end = _cur_chunk_idx + _chunk_size;
2567 }
2568 void add_region(HeapRegion* r) {
2569 if (_cur_chunk_idx == _cur_chunk_end) {
2570 get_new_chunk();
2571 }
2572 assert(_cur_chunk_idx < _cur_chunk_end, "postcondition");
2573 _hrSorted->setMarkedHeapRegion(_cur_chunk_idx, r);
2574 _marked_regions_added++;
2575 _cur_chunk_idx++;
2576 }
2577
2578 public:
2579 ParKnownGarbageHRClosure(CollectionSetChooser* hrSorted,
2580 jint chunk_size,
2581 int worker) :
2582 _hrSorted(hrSorted), _chunk_size(chunk_size), _worker(worker),
2583 _marked_regions_added(0), _cur_chunk_idx(0), _cur_chunk_end(0),
2584 _invokes(0)
2585 {}
2586
2587 bool doHeapRegion(HeapRegion* r) {
2588 // We only include humongous regions in collection
2589 // sets when concurrent mark shows that their contained object is
2590 // unreachable.
2591 _invokes++;
2592
2593 // Do we have any marking information for this region?
2594 if (r->is_marked()) {
2595 // We don't include humongous regions in collection
2596 // sets because we collect them immediately at the end of a marking
2597 // cycle.
2598 // We also do not include young regions in collection sets
2599 if (!r->isHumongous() && !r->is_young()) {
2600 add_region(r);
2601 }
2602 }
2603 return false;
2604 }
2605 jint marked_regions_added() { return _marked_regions_added; }
2606 int invokes() { return _invokes; }
2607 };
2608
2609 class ParKnownGarbageTask: public AbstractGangTask {
2610 CollectionSetChooser* _hrSorted;
2611 jint _chunk_size;
2612 G1CollectedHeap* _g1;
2613 public:
2614 ParKnownGarbageTask(CollectionSetChooser* hrSorted, jint chunk_size) :
2615 AbstractGangTask("ParKnownGarbageTask"),
2616 _hrSorted(hrSorted), _chunk_size(chunk_size),
2617 _g1(G1CollectedHeap::heap())
2618 {}
2619
2620 void work(int i) {
2621 ParKnownGarbageHRClosure parKnownGarbageCl(_hrSorted, _chunk_size, i);
2622 // Back to zero for the claim value.
2623 _g1->heap_region_par_iterate_chunked(&parKnownGarbageCl, i,
2624 HeapRegion::InitialClaimValue);
2625 jint regions_added = parKnownGarbageCl.marked_regions_added();
2626 _hrSorted->incNumMarkedHeapRegions(regions_added);
2627 if (G1PrintParCleanupStats) {
2628 gclog_or_tty->print_cr(" Thread %d called %d times, added %d regions to list.",
2629 i, parKnownGarbageCl.invokes(), regions_added);
2630 }
2631 }
2632 };
2633
2634 void
2635 G1CollectorPolicy::record_concurrent_mark_cleanup_end() {
2636 double start_sec;
2637 if (G1PrintParCleanupStats) {
2638 start_sec = os::elapsedTime();
2639 }
2640
2641 _collectionSetChooser->clearMarkedHeapRegions();
2642 double clear_marked_end_sec;
2643 if (G1PrintParCleanupStats) {
2644 clear_marked_end_sec = os::elapsedTime();
2645 gclog_or_tty->print_cr(" clear marked regions: %8.3f ms.",
2646 (clear_marked_end_sec - start_sec) * 1000.0);
2647 }
2648
2649 if (G1CollectedHeap::use_parallel_gc_threads()) {
2650 const size_t OverpartitionFactor = 4;
2651 const size_t MinWorkUnit = 8;
2652 const size_t WorkUnit =
2653 MAX2(_g1->n_regions() / (ParallelGCThreads * OverpartitionFactor),
2654 MinWorkUnit);
2655 _collectionSetChooser->prepareForAddMarkedHeapRegionsPar(_g1->n_regions(),
2656 WorkUnit);
2657 ParKnownGarbageTask parKnownGarbageTask(_collectionSetChooser,
2658 (int) WorkUnit);
2659 _g1->workers()->run_task(&parKnownGarbageTask);
2660
2661 assert(_g1->check_heap_region_claim_values(HeapRegion::InitialClaimValue),
2662 "sanity check");
2663 } else {
2664 KnownGarbageClosure knownGarbagecl(_collectionSetChooser);
2665 _g1->heap_region_iterate(&knownGarbagecl);
2666 }
2667 double known_garbage_end_sec;
2668 if (G1PrintParCleanupStats) {
2669 known_garbage_end_sec = os::elapsedTime();
2670 gclog_or_tty->print_cr(" compute known garbage: %8.3f ms.",
2671 (known_garbage_end_sec - clear_marked_end_sec) * 1000.0);
2672 }
2673
2674 _collectionSetChooser->sortMarkedHeapRegions();
2675 double end_sec = os::elapsedTime();
2676 if (G1PrintParCleanupStats) {
2677 gclog_or_tty->print_cr(" sorting: %8.3f ms.",
2678 (end_sec - known_garbage_end_sec) * 1000.0);
2679 }
2680
2681 double elapsed_time_ms = (end_sec - _mark_cleanup_start_sec) * 1000.0;
2682 _concurrent_mark_cleanup_times_ms->add(elapsed_time_ms);
2683 _cur_mark_stop_world_time_ms += elapsed_time_ms;
2684 _prev_collection_pause_end_ms += elapsed_time_ms;
2685 _mmu_tracker->add_pause(_mark_cleanup_start_sec, end_sec, true);
2686 }
2687
2688 // Add the heap region at the head of the non-incremental collection set
2689 void G1CollectorPolicy::
2690 add_to_collection_set(HeapRegion* hr) {
2691 assert(_inc_cset_build_state == Active, "Precondition");
2692 assert(!hr->is_young(), "non-incremental add of young region");
2693
2694 if (_g1->mark_in_progress())
2695 _g1->concurrent_mark()->registerCSetRegion(hr);
2696
2697 assert(!hr->in_collection_set(), "should not already be in the CSet");
2698 hr->set_in_collection_set(true);
2699 hr->set_next_in_collection_set(_collection_set);
2700 _collection_set = hr;
2701 _collection_set_size++;
2702 _collection_set_bytes_used_before += hr->used();
2703 _g1->register_region_with_in_cset_fast_test(hr);
2704 }
2705
2706 // Initialize the per-collection-set information
2707 void G1CollectorPolicy::start_incremental_cset_building() {
2708 assert(_inc_cset_build_state == Inactive, "Precondition");
2709
2710 _inc_cset_head = NULL;
2711 _inc_cset_tail = NULL;
2712 _inc_cset_size = 0;
2713 _inc_cset_bytes_used_before = 0;
2714
2715 _inc_cset_young_index = 0;
2716
2717 _inc_cset_max_finger = 0;
2718 _inc_cset_recorded_young_bytes = 0;
2719 _inc_cset_recorded_rs_lengths = 0;
2720 _inc_cset_predicted_elapsed_time_ms = 0;
2721 _inc_cset_predicted_bytes_to_copy = 0;
2722 _inc_cset_build_state = Active;
2723 }
2724
2725 void G1CollectorPolicy::add_to_incremental_cset_info(HeapRegion* hr, size_t rs_length) {
2726 // This routine is used when:
2727 // * adding survivor regions to the incremental cset at the end of an
2728 // evacuation pause,
2729 // * adding the current allocation region to the incremental cset
2730 // when it is retired, and
2731 // * updating existing policy information for a region in the
2732 // incremental cset via young list RSet sampling.
2733 // Therefore this routine may be called at a safepoint by the
2734 // VM thread, or in-between safepoints by mutator threads (when
2735 // retiring the current allocation region) or a concurrent
2736 // refine thread (RSet sampling).
2737
2738 double region_elapsed_time_ms = predict_region_elapsed_time_ms(hr, true);
2739 size_t used_bytes = hr->used();
2740
2741 _inc_cset_recorded_rs_lengths += rs_length;
2742 _inc_cset_predicted_elapsed_time_ms += region_elapsed_time_ms;
2743
2744 _inc_cset_bytes_used_before += used_bytes;
2745
2746 // Cache the values we have added to the aggregated informtion
2747 // in the heap region in case we have to remove this region from
2748 // the incremental collection set, or it is updated by the
2749 // rset sampling code
2750 hr->set_recorded_rs_length(rs_length);
2751 hr->set_predicted_elapsed_time_ms(region_elapsed_time_ms);
2752
2753 #if PREDICTIONS_VERBOSE
2754 size_t bytes_to_copy = predict_bytes_to_copy(hr);
2755 _inc_cset_predicted_bytes_to_copy += bytes_to_copy;
2756
2757 // Record the number of bytes used in this region
2758 _inc_cset_recorded_young_bytes += used_bytes;
2759
2760 // Cache the values we have added to the aggregated informtion
2761 // in the heap region in case we have to remove this region from
2762 // the incremental collection set, or it is updated by the
2763 // rset sampling code
2764 hr->set_predicted_bytes_to_copy(bytes_to_copy);
2765 #endif // PREDICTIONS_VERBOSE
2766 }
2767
2768 void G1CollectorPolicy::remove_from_incremental_cset_info(HeapRegion* hr) {
2769 // This routine is currently only called as part of the updating of
2770 // existing policy information for regions in the incremental cset that
2771 // is performed by the concurrent refine thread(s) as part of young list
2772 // RSet sampling. Therefore we should not be at a safepoint.
2773
2774 assert(!SafepointSynchronize::is_at_safepoint(), "should not be at safepoint");
2775 assert(hr->is_young(), "it should be");
2776
2777 size_t used_bytes = hr->used();
2778 size_t old_rs_length = hr->recorded_rs_length();
2779 double old_elapsed_time_ms = hr->predicted_elapsed_time_ms();
2780
2781 // Subtract the old recorded/predicted policy information for
2782 // the given heap region from the collection set info.
2783 _inc_cset_recorded_rs_lengths -= old_rs_length;
2784 _inc_cset_predicted_elapsed_time_ms -= old_elapsed_time_ms;
2785
2786 _inc_cset_bytes_used_before -= used_bytes;
2787
2788 // Clear the values cached in the heap region
2789 hr->set_recorded_rs_length(0);
2790 hr->set_predicted_elapsed_time_ms(0);
2791
2792 #if PREDICTIONS_VERBOSE
2793 size_t old_predicted_bytes_to_copy = hr->predicted_bytes_to_copy();
2794 _inc_cset_predicted_bytes_to_copy -= old_predicted_bytes_to_copy;
2795
2796 // Subtract the number of bytes used in this region
2797 _inc_cset_recorded_young_bytes -= used_bytes;
2798
2799 // Clear the values cached in the heap region
2800 hr->set_predicted_bytes_to_copy(0);
2801 #endif // PREDICTIONS_VERBOSE
2802 }
2803
2804 void G1CollectorPolicy::update_incremental_cset_info(HeapRegion* hr, size_t new_rs_length) {
2805 // Update the collection set information that is dependent on the new RS length
2806 assert(hr->is_young(), "Precondition");
2807
2808 remove_from_incremental_cset_info(hr);
2809 add_to_incremental_cset_info(hr, new_rs_length);
2810 }
2811
2812 void G1CollectorPolicy::add_region_to_incremental_cset_common(HeapRegion* hr) {
2813 assert( hr->is_young(), "invariant");
2814 assert( hr->young_index_in_cset() == -1, "invariant" );
2815 assert(_inc_cset_build_state == Active, "Precondition");
2816
2817 // We need to clear and set the cached recorded/cached collection set
2818 // information in the heap region here (before the region gets added
2819 // to the collection set). An individual heap region's cached values
2820 // are calculated, aggregated with the policy collection set info,
2821 // and cached in the heap region here (initially) and (subsequently)
2822 // by the Young List sampling code.
2823
2824 size_t rs_length = hr->rem_set()->occupied();
2825 add_to_incremental_cset_info(hr, rs_length);
2826
2827 HeapWord* hr_end = hr->end();
2828 _inc_cset_max_finger = MAX2(_inc_cset_max_finger, hr_end);
2829
2830 assert(!hr->in_collection_set(), "invariant");
2831 hr->set_in_collection_set(true);
2832 assert( hr->next_in_collection_set() == NULL, "invariant");
2833
2834 _inc_cset_size++;
2835 _g1->register_region_with_in_cset_fast_test(hr);
2836
2837 hr->set_young_index_in_cset((int) _inc_cset_young_index);
2838 ++_inc_cset_young_index;
2839 }
2840
2841 // Add the region at the RHS of the incremental cset
2842 void G1CollectorPolicy::add_region_to_incremental_cset_rhs(HeapRegion* hr) {
2843 // We should only ever be appending survivors at the end of a pause
2844 assert( hr->is_survivor(), "Logic");
2845
2846 // Do the 'common' stuff
2847 add_region_to_incremental_cset_common(hr);
2848
2849 // Now add the region at the right hand side
2850 if (_inc_cset_tail == NULL) {
2851 assert(_inc_cset_head == NULL, "invariant");
2852 _inc_cset_head = hr;
2853 } else {
2854 _inc_cset_tail->set_next_in_collection_set(hr);
2855 }
2856 _inc_cset_tail = hr;
2857 }
2858
2859 // Add the region to the LHS of the incremental cset
2860 void G1CollectorPolicy::add_region_to_incremental_cset_lhs(HeapRegion* hr) {
2861 // Survivors should be added to the RHS at the end of a pause
2862 assert(!hr->is_survivor(), "Logic");
2863
2864 // Do the 'common' stuff
2865 add_region_to_incremental_cset_common(hr);
2866
2867 // Add the region at the left hand side
2868 hr->set_next_in_collection_set(_inc_cset_head);
2869 if (_inc_cset_head == NULL) {
2870 assert(_inc_cset_tail == NULL, "Invariant");
2871 _inc_cset_tail = hr;
2872 }
2873 _inc_cset_head = hr;
2874 }
2875
2876 #ifndef PRODUCT
2877 void G1CollectorPolicy::print_collection_set(HeapRegion* list_head, outputStream* st) {
2878 assert(list_head == inc_cset_head() || list_head == collection_set(), "must be");
2879
2880 st->print_cr("\nCollection_set:");
2881 HeapRegion* csr = list_head;
2882 while (csr != NULL) {
2883 HeapRegion* next = csr->next_in_collection_set();
2884 assert(csr->in_collection_set(), "bad CS");
2885 st->print_cr(" [%08x-%08x], t: %08x, P: %08x, N: %08x, C: %08x, "
2886 "age: %4d, y: %d, surv: %d",
2887 csr->bottom(), csr->end(),
2888 csr->top(),
2889 csr->prev_top_at_mark_start(),
2890 csr->next_top_at_mark_start(),
2891 csr->top_at_conc_mark_count(),
2892 csr->age_in_surv_rate_group_cond(),
2893 csr->is_young(),
2894 csr->is_survivor());
2895 csr = next;
2896 }
2897 }
2898 #endif // !PRODUCT
2899
2900 void G1CollectorPolicy::choose_collection_set(double target_pause_time_ms) {
2901 // Set this here - in case we're not doing young collections.
2902 double non_young_start_time_sec = os::elapsedTime();
2903
2904 YoungList* young_list = _g1->young_list();
2905
2906 start_recording_regions();
2907
2908 guarantee(target_pause_time_ms > 0.0,
2909 err_msg("target_pause_time_ms = %1.6lf should be positive",
2910 target_pause_time_ms));
2911 guarantee(_collection_set == NULL, "Precondition");
2912
2913 double base_time_ms = predict_base_elapsed_time_ms(_pending_cards);
2914 double predicted_pause_time_ms = base_time_ms;
2915
2916 double time_remaining_ms = target_pause_time_ms - base_time_ms;
2917
2918 ergo_verbose3(ErgoCSetConstruction | ErgoHigh,
2919 "start choosing CSet",
2920 ergo_format_ms("predicted base time")
2921 ergo_format_ms("remaining time")
2922 ergo_format_ms("target pause time"),
2923 base_time_ms, time_remaining_ms, target_pause_time_ms);
2924
2925 // the 10% and 50% values are arbitrary...
2926 double threshold = 0.10 * target_pause_time_ms;
2927 if (time_remaining_ms < threshold) {
2928 double prev_time_remaining_ms = time_remaining_ms;
2929 time_remaining_ms = 0.50 * target_pause_time_ms;
2930 _within_target = false;
2931 ergo_verbose3(ErgoCSetConstruction,
2932 "adjust remaining time",
2933 ergo_format_reason("remaining time lower than threshold")
2934 ergo_format_ms("remaining time")
2935 ergo_format_ms("threshold")
2936 ergo_format_ms("adjusted remaining time"),
2937 prev_time_remaining_ms, threshold, time_remaining_ms);
2938 } else {
2939 _within_target = true;
2940 }
2941
2942 size_t expansion_bytes = _g1->expansion_regions() * HeapRegion::GrainBytes;
2943
2944 HeapRegion* hr;
2945 double young_start_time_sec = os::elapsedTime();
2946
2947 _collection_set_bytes_used_before = 0;
2948 _collection_set_size = 0;
2949 _young_cset_length = 0;
2950 _last_young_gc_full = full_young_gcs() ? true : false;
2951
2952 if (_last_young_gc_full) {
2953 ++_full_young_pause_num;
2954 } else {
2955 ++_partial_young_pause_num;
2956 }
2957
2958 // The young list is laid with the survivor regions from the previous
2959 // pause are appended to the RHS of the young list, i.e.
2960 // [Newly Young Regions ++ Survivors from last pause].
2961
2962 size_t survivor_region_num = young_list->survivor_length();
2963 size_t eden_region_num = young_list->length() - survivor_region_num;
2964 size_t old_region_num = 0;
2965 hr = young_list->first_survivor_region();
2966 while (hr != NULL) {
2967 assert(hr->is_survivor(), "badly formed young list");
2968 hr->set_young();
2969 hr = hr->get_next_young_region();
2970 }
2971
2972 // Clear the fields that point to the survivor list - they are all young now.
2973 young_list->clear_survivors();
2974
2975 if (_g1->mark_in_progress())
2976 _g1->concurrent_mark()->register_collection_set_finger(_inc_cset_max_finger);
2977
2978 _young_cset_length = _inc_cset_young_index;
2979 _collection_set = _inc_cset_head;
2980 _collection_set_size = _inc_cset_size;
2981 _collection_set_bytes_used_before = _inc_cset_bytes_used_before;
2982 time_remaining_ms -= _inc_cset_predicted_elapsed_time_ms;
2983 predicted_pause_time_ms += _inc_cset_predicted_elapsed_time_ms;
2984
2985 ergo_verbose3(ErgoCSetConstruction | ErgoHigh,
2986 "add young regions to CSet",
2987 ergo_format_region("eden")
2988 ergo_format_region("survivors")
2989 ergo_format_ms("predicted young region time"),
2990 eden_region_num, survivor_region_num,
2991 _inc_cset_predicted_elapsed_time_ms);
2992
2993 // The number of recorded young regions is the incremental
2994 // collection set's current size
2995 set_recorded_young_regions(_inc_cset_size);
2996 set_recorded_rs_lengths(_inc_cset_recorded_rs_lengths);
2997 set_recorded_young_bytes(_inc_cset_recorded_young_bytes);
2998 #if PREDICTIONS_VERBOSE
2999 set_predicted_bytes_to_copy(_inc_cset_predicted_bytes_to_copy);
3000 #endif // PREDICTIONS_VERBOSE
3001
3002 assert(_inc_cset_size == young_list->length(), "Invariant");
3003
3004 double young_end_time_sec = os::elapsedTime();
3005 _recorded_young_cset_choice_time_ms =
3006 (young_end_time_sec - young_start_time_sec) * 1000.0;
3007
3008 // We are doing young collections so reset this.
3009 non_young_start_time_sec = young_end_time_sec;
3010
3011 if (!full_young_gcs()) {
3012 bool should_continue = true;
3013 NumberSeq seq;
3014 double avg_prediction = 100000000000000000.0; // something very large
3015
3016 size_t prev_collection_set_size = _collection_set_size;
3017 double prev_predicted_pause_time_ms = predicted_pause_time_ms;
3018 do {
3019 hr = _collectionSetChooser->getNextMarkedRegion(time_remaining_ms,
3020 avg_prediction);
3021 if (hr != NULL) {
3022 double predicted_time_ms = predict_region_elapsed_time_ms(hr, false);
3023 time_remaining_ms -= predicted_time_ms;
3024 predicted_pause_time_ms += predicted_time_ms;
3025 add_to_collection_set(hr);
3026 record_non_young_cset_region(hr);
3027 seq.add(predicted_time_ms);
3028 avg_prediction = seq.avg() + seq.sd();
3029 }
3030
3031 should_continue = true;
3032 if (hr == NULL) {
3033 // No need for an ergo verbose message here,
3034 // getNextMarkRegion() does this when it returns NULL.
3035 should_continue = false;
3036 } else {
3037 if (adaptive_young_list_length()) {
3038 if (time_remaining_ms < 0.0) {
3039 ergo_verbose1(ErgoCSetConstruction,
3040 "stop adding old regions to CSet",
3041 ergo_format_reason("remaining time is lower than 0")
3042 ergo_format_ms("remaining time"),
3043 time_remaining_ms);
3044 should_continue = false;
3045 }
3046 } else {
3047 if (_collection_set_size >= _young_list_fixed_length) {
3048 ergo_verbose2(ErgoCSetConstruction,
3049 "stop adding old regions to CSet",
3050 ergo_format_reason("CSet length reached target")
3051 ergo_format_region("CSet")
3052 ergo_format_region("young target"),
3053 _collection_set_size, _young_list_fixed_length);
3054 should_continue = false;
3055 }
3056 }
3057 }
3058 } while (should_continue);
3059
3060 if (!adaptive_young_list_length() &&
3061 _collection_set_size < _young_list_fixed_length) {
3062 ergo_verbose2(ErgoCSetConstruction,
3063 "request partially-young GCs end",
3064 ergo_format_reason("CSet length lower than target")
3065 ergo_format_region("CSet")
3066 ergo_format_region("young target"),
3067 _collection_set_size, _young_list_fixed_length);
3068 _should_revert_to_full_young_gcs = true;
3069 }
3070
3071 old_region_num = _collection_set_size - prev_collection_set_size;
3072
3073 ergo_verbose2(ErgoCSetConstruction | ErgoHigh,
3074 "add old regions to CSet",
3075 ergo_format_region("old")
3076 ergo_format_ms("predicted old region time"),
3077 old_region_num,
3078 predicted_pause_time_ms - prev_predicted_pause_time_ms);
3079 }
3080
3081 stop_incremental_cset_building();
3082
3083 count_CS_bytes_used();
3084
3085 end_recording_regions();
3086
3087 ergo_verbose5(ErgoCSetConstruction,
3088 "finish choosing CSet",
3089 ergo_format_region("eden")
3090 ergo_format_region("survivors")
3091 ergo_format_region("old")
3092 ergo_format_ms("predicted pause time")
3093 ergo_format_ms("target pause time"),
3094 eden_region_num, survivor_region_num, old_region_num,
3095 predicted_pause_time_ms, target_pause_time_ms);
3096
3097 double non_young_end_time_sec = os::elapsedTime();
3098 _recorded_non_young_cset_choice_time_ms =
3099 (non_young_end_time_sec - non_young_start_time_sec) * 1000.0;
3100 }