1 /*
2 * Copyright Amazon.com Inc. 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
27 #include "gc/shenandoah/heuristics/shenandoahOldHeuristics.hpp"
28 #include "gc/shenandoah/heuristics/shenandoahYoungHeuristics.hpp"
29 #include "gc/shenandoah/shenandoahCollectorPolicy.hpp"
30 #include "gc/shenandoah/shenandoahHeap.inline.hpp"
31 #include "gc/shenandoah/shenandoahHeapRegion.inline.hpp"
32 #include "gc/shenandoah/shenandoahYoungGeneration.hpp"
33
34 #include "utilities/quickSort.hpp"
35
36 ShenandoahYoungHeuristics::ShenandoahYoungHeuristics(ShenandoahYoungGeneration* generation)
37 : ShenandoahGenerationalHeuristics(generation) {
38 assert(!generation->is_old(), "Young heuristics only accept the young generation");
39 }
40
41
42 void ShenandoahYoungHeuristics::choose_collection_set_from_regiondata(ShenandoahCollectionSet* cset,
43 RegionData* data, size_t size,
44 size_t actual_free) {
45 // The logic for cset selection in adaptive is as follows:
46 //
47 // 1. We cannot get cset larger than available free space. Otherwise we guarantee OOME
48 // during evacuation, and thus guarantee full GC. In practice, we also want to let
49 // application to allocate something. This is why we limit CSet to some fraction of
50 // available space. In non-overloaded heap, max_cset would contain all plausible candidates
51 // over garbage threshold.
52 //
53 // 2. We should not get cset too low so that free threshold would not be met right
54 // after the cycle. Otherwise we get back-to-back cycles for no reason if heap is
55 // too fragmented. In non-overloaded non-fragmented heap min_garbage would be around zero.
56 //
57 // Therefore, we start by sorting the regions by garbage. Then we unconditionally add the best candidates
58 // before we meet min_garbage. Then we add all candidates that fit with a garbage threshold before
59 // we hit max_cset. When max_cset is hit, we terminate the cset selection. Note that in this scheme,
60 // ShenandoahGarbageThreshold is the soft threshold which would be ignored until min_garbage is hit.
61
62 // In generational mode, the sort order within the data array is not strictly descending amounts of garbage. In
63 // particular, regions that have reached tenure age will be sorted into this array before younger regions that contain
64 // more garbage. This represents one of the reasons why we keep looking at regions even after we decide, for example,
65 // to exclude one of the regions because it might require evacuation of too much live data.
66
67 // Better select garbage-first regions
68 QuickSort::sort<RegionData>(data, (int) size, compare_by_garbage, false);
69
70 size_t cur_young_garbage = add_preselected_regions_to_collection_set(cset, data, size);
71
72 choose_young_collection_set(cset, data, size, actual_free, cur_young_garbage);
73
74 log_cset_composition(cset);
75 }
76
77 void ShenandoahYoungHeuristics::choose_young_collection_set(ShenandoahCollectionSet* cset,
78 const RegionData* data,
79 size_t size, size_t actual_free,
80 size_t cur_young_garbage) const {
81
82 ShenandoahHeap* heap = ShenandoahHeap::heap();
83
84 size_t capacity = heap->young_generation()->max_capacity();
85 size_t garbage_threshold = ShenandoahHeapRegion::region_size_bytes() * ShenandoahGarbageThreshold / 100;
86 size_t ignore_threshold = ShenandoahHeapRegion::region_size_bytes() * ShenandoahIgnoreGarbageThreshold / 100;
87 const uint tenuring_threshold = heap->age_census()->tenuring_threshold();
88
89 // This is young-gen collection or a mixed evacuation.
90 // If this is mixed evacuation, the old-gen candidate regions have already been added.
91 size_t max_cset = (size_t) (heap->get_young_evac_reserve() / ShenandoahEvacWaste);
92 size_t cur_cset = 0;
93 size_t free_target = (capacity * ShenandoahMinFreeThreshold) / 100 + max_cset;
94 size_t min_garbage = (free_target > actual_free) ? (free_target - actual_free) : 0;
95
96
97 log_info(gc, ergo)(
98 "Adaptive CSet Selection for YOUNG. Max Evacuation: " SIZE_FORMAT "%s, Actual Free: " SIZE_FORMAT "%s.",
99 byte_size_in_proper_unit(max_cset), proper_unit_for_byte_size(max_cset),
100 byte_size_in_proper_unit(actual_free), proper_unit_for_byte_size(actual_free));
101
102 for (size_t idx = 0; idx < size; idx++) {
103 ShenandoahHeapRegion* r = data[idx]._region;
104 if (cset->is_preselected(r->index())) {
105 continue;
106 }
107 if (r->age() < tenuring_threshold) {
108 size_t new_cset = cur_cset + r->get_live_data_bytes();
109 size_t region_garbage = r->garbage();
110 size_t new_garbage = cur_young_garbage + region_garbage;
111 bool add_regardless = (region_garbage > ignore_threshold) && (new_garbage < min_garbage);
112 assert(r->is_young(), "Only young candidates expected in the data array");
113 if ((new_cset <= max_cset) && (add_regardless || (region_garbage > garbage_threshold))) {
114 cur_cset = new_cset;
115 cur_young_garbage = new_garbage;
116 cset->add_region(r);
117 }
118 }
119 // Note that we do not add aged regions if they were not pre-selected. The reason they were not preselected
120 // is because there is not sufficient room in old-gen to hold their to-be-promoted live objects or because
121 // they are to be promoted in place.
122 }
123 }
124
125
126 bool ShenandoahYoungHeuristics::should_start_gc() {
127 // inherited triggers have already decided to start a cycle, so no further evaluation is required
128 if (ShenandoahAdaptiveHeuristics::should_start_gc()) {
129 return true;
130 }
131
132 // Get through promotions and mixed evacuations as quickly as possible. These cycles sometimes require significantly
133 // more time than traditional young-generation cycles so start them up as soon as possible. This is a "mitigation"
134 // for the reality that old-gen and young-gen activities are not truly "concurrent". If there is old-gen work to
135 // be done, we start up the young-gen GC threads so they can do some of this old-gen work. As implemented, promotion
136 // gets priority over old-gen marking.
137 ShenandoahHeap* heap = ShenandoahHeap::heap();
138
139 size_t promo_expedite_threshold = (heap->young_generation()->max_capacity() * ShenandoahEvacReserve) / 512;
140 size_t promo_potential = heap->get_promotion_potential();
141 if (promo_potential > promo_expedite_threshold) {
142 // Detect unsigned arithmetic underflow
143 assert(promo_potential < heap->capacity(), "Sanity");
144 log_info(gc)("Trigger (%s): expedite promotion of " SIZE_FORMAT "%s",
145 _space_info->name(),
146 byte_size_in_proper_unit(promo_potential),
147 proper_unit_for_byte_size(promo_potential));
148 return true;
149 }
150
151 ShenandoahOldHeuristics* old_heuristics = heap->old_heuristics();
152 size_t mixed_candidates = old_heuristics->unprocessed_old_collection_candidates();
153 if (mixed_candidates > 0) {
154 // We need to run young GC in order to open up some free heap regions so we can finish mixed evacuations.
155 log_info(gc)("Trigger (%s): expedite mixed evacuation of " SIZE_FORMAT " regions",
156 _space_info->name(), mixed_candidates);
157 return true;
158 }
159
160 return false;
161 }
162
163 // Return a conservative estimate of how much memory can be allocated before we need to start GC. The estimate is based
164 // on memory that is currently available within young generation plus all of the memory that will be added to the young
165 // generation at the end of the current cycle (as represented by young_regions_to_be_reclaimed) and on the anticipated
166 // amount of time required to perform a GC.
167 size_t ShenandoahYoungHeuristics::bytes_of_allocation_runway_before_gc_trigger(size_t young_regions_to_be_reclaimed) {
168 size_t capacity = _space_info->soft_max_capacity();
169 size_t usage = _space_info->used();
170 size_t available = (capacity > usage)? capacity - usage: 0;
171 size_t allocated = _space_info->bytes_allocated_since_gc_start();
172
173 size_t available_young_collected = ShenandoahHeap::heap()->collection_set()->get_young_available_bytes_collected();
174 size_t anticipated_available =
175 available + young_regions_to_be_reclaimed * ShenandoahHeapRegion::region_size_bytes() - available_young_collected;
176 size_t spike_headroom = capacity * ShenandoahAllocSpikeFactor / 100;
177 size_t penalties = capacity * _gc_time_penalties / 100;
178
179 double rate = _allocation_rate.sample(allocated);
180
181 // At what value of available, would avg and spike triggers occur?
182 // if allocation_headroom < avg_cycle_time * avg_alloc_rate, then we experience avg trigger
183 // if allocation_headroom < avg_cycle_time * rate, then we experience spike trigger if is_spiking
184 //
185 // allocation_headroom =
186 // 0, if penalties > available or if penalties + spike_headroom > available
187 // available - penalties - spike_headroom, otherwise
188 //
189 // so we trigger if available - penalties - spike_headroom < avg_cycle_time * avg_alloc_rate, which is to say
190 // available < avg_cycle_time * avg_alloc_rate + penalties + spike_headroom
191 // or if available < penalties + spike_headroom
192 //
193 // since avg_cycle_time * avg_alloc_rate > 0, the first test is sufficient to test both conditions
194 //
195 // thus, evac_slack_avg is MIN2(0, available - avg_cycle_time * avg_alloc_rate + penalties + spike_headroom)
196 //
197 // similarly, evac_slack_spiking is MIN2(0, available - avg_cycle_time * rate + penalties + spike_headroom)
198 // but evac_slack_spiking is only relevant if is_spiking, as defined below.
199
200 double avg_cycle_time = _gc_cycle_time_history->davg() + (_margin_of_error_sd * _gc_cycle_time_history->dsd());
201
202 // TODO: Consider making conservative adjustments to avg_cycle_time, such as: (avg_cycle_time *= 2) in cases where
203 // we expect a longer-than-normal GC duration. This includes mixed evacuations, evacuation that perform promotion
204 // including promotion in place, and OLD GC bootstrap cycles. It has been observed that these cycles sometimes
205 // require twice or more the duration of "normal" GC cycles. We have experimented with this approach. While it
206 // does appear to reduce the frequency of degenerated cycles due to late triggers, it also has the effect of reducing
207 // evacuation slack so that there is less memory available to be transferred to OLD. The result is that we
208 // throttle promotion and it takes too long to move old objects out of the young generation.
209
210 double avg_alloc_rate = _allocation_rate.upper_bound(_margin_of_error_sd);
211 size_t evac_slack_avg;
212 if (anticipated_available > avg_cycle_time * avg_alloc_rate + penalties + spike_headroom) {
213 evac_slack_avg = anticipated_available - (avg_cycle_time * avg_alloc_rate + penalties + spike_headroom);
214 } else {
215 // we have no slack because it's already time to trigger
216 evac_slack_avg = 0;
217 }
218
219 bool is_spiking = _allocation_rate.is_spiking(rate, _spike_threshold_sd);
220 size_t evac_slack_spiking;
221 if (is_spiking) {
222 if (anticipated_available > avg_cycle_time * rate + penalties + spike_headroom) {
223 evac_slack_spiking = anticipated_available - (avg_cycle_time * rate + penalties + spike_headroom);
224 } else {
225 // we have no slack because it's already time to trigger
226 evac_slack_spiking = 0;
227 }
228 } else {
229 evac_slack_spiking = evac_slack_avg;
230 }
231
232 size_t threshold = min_free_threshold();
233 size_t evac_min_threshold = (anticipated_available > threshold)? anticipated_available - threshold: 0;
234 return MIN3(evac_slack_spiking, evac_slack_avg, evac_min_threshold);
235 }
236