1 /*
2 * Copyright (c) 2001, 2020, 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 #ifndef SHARE_GC_G1_G1DIRTYCARDQUEUE_HPP
26 #define SHARE_GC_G1_G1DIRTYCARDQUEUE_HPP
27
28 #include "gc/g1/g1BufferNodeList.hpp"
29 #include "gc/g1/g1FreeIdSet.hpp"
30 #include "gc/shared/ptrQueue.hpp"
31 #include "memory/allocation.hpp"
32 #include "memory/padded.hpp"
33
34 class G1ConcurrentRefineThread;
35 class G1DirtyCardQueueSet;
36 class G1RedirtyCardsQueueSet;
37 class Thread;
38
39 // A ptrQueue whose elements are "oops", pointers to object heads.
40 class G1DirtyCardQueue: public PtrQueue {
41 protected:
42 virtual void handle_completed_buffer();
43
44 public:
45 G1DirtyCardQueue(G1DirtyCardQueueSet* qset);
46
47 // Flush before destroying; queue may be used to capture pending work while
48 // doing something else, with auto-flush on completion.
49 ~G1DirtyCardQueue();
50
51 // Process queue entries and release resources.
52 void flush() { flush_impl(); }
53
54 inline G1DirtyCardQueueSet* dirty_card_qset() const;
55
56 // Compiler support.
57 static ByteSize byte_offset_of_index() {
58 return PtrQueue::byte_offset_of_index<G1DirtyCardQueue>();
59 }
60 using PtrQueue::byte_width_of_index;
61
62 static ByteSize byte_offset_of_buf() {
63 return PtrQueue::byte_offset_of_buf<G1DirtyCardQueue>();
64 }
65 using PtrQueue::byte_width_of_buf;
66
67 };
68
69 class G1DirtyCardQueueSet: public PtrQueueSet {
70 G1ConcurrentRefineThread* _primary_refinement_thread;
71 // Add padding for improved performance for shared access. There's only
72 // one instance of this class, so using a little extra space is fine.
73 // _completed_buffers_{head,tail} and _num_cards are isolated to their
74 // own cache lines. Other members are not, even if shared access, because
75 // they aren't as critical to performance.
76 DEFINE_PAD_MINUS_SIZE(1, DEFAULT_CACHE_LINE_SIZE, sizeof(G1ConcurrentRefineThread*));
77 BufferNode* volatile _completed_buffers_head;
78 DEFINE_PAD_MINUS_SIZE(2, DEFAULT_CACHE_LINE_SIZE, sizeof(BufferNode* volatile));
79 BufferNode* volatile _completed_buffers_tail;
80 DEFINE_PAD_MINUS_SIZE(3, DEFAULT_CACHE_LINE_SIZE, sizeof(BufferNode* volatile));
81 volatile size_t _num_cards;
82 DEFINE_PAD_MINUS_SIZE(4, DEFAULT_CACHE_LINE_SIZE, sizeof(volatile size_t));
83
84 DEBUG_ONLY(mutable volatile int _concurrency;)
85 class ConcurrentVerifier;
86 class NonconcurrentVerifier;
87
88 size_t append_buffers(BufferNode* first, BufferNode* last, size_t card_count);
89 // Verify _num_cards == sum of cards in the completed queue.
90 void verify_num_cards() const NOT_DEBUG_RETURN;
91
92 struct PauseList {
93 BufferNode* volatile _head;
94 BufferNode* _tail;
95 PauseList() : _head(NULL), _tail(NULL) {}
96 };
97 PauseList _paused[2];
98
99 void record_paused_buffer(BufferNode* node);
100 void enqueue_paused_buffers_aux(size_t index);
101 void enqueue_previous_paused_buffers();
102 void enqueue_all_paused_buffers();
103
104 void clear_completed_buffers();
105 void abandon_completed_buffers();
106
107 size_t _process_cards_threshold;
108
109 // Refine the cards in "node" from its index to buffer_size.
110 // Stops processing if SuspendibleThreadSet::should_yield() is true.
111 // Returns true if the entire buffer was processed, false if there
112 // is a pending yield request. The node's index is updated to exclude
113 // the processed elements, e.g. up to the element before processing
114 // stopped, or one past the last element if the entire buffer was
115 // processed. Increments *total_refined_cards by the number of cards
116 // processed and removed from the buffer.
117 bool refine_buffer(BufferNode* node, uint worker_id, size_t* total_refined_cards);
118
119 bool mut_process_buffer(BufferNode* node);
120
121 // If the queue contains more cards than configured here, the
122 // mutator must start doing some of the concurrent refinement work.
123 size_t _max_cards;
124 size_t _max_cards_padding;
125 static const size_t MaxCardsUnlimited = SIZE_MAX;
126
127 G1FreeIdSet _free_ids;
128
129 // Array of cumulative dirty cards refined by mutator threads.
130 // Array has an entry per id in _free_ids.
131 size_t* _mutator_refined_cards_counters;
132
133 public:
134 G1DirtyCardQueueSet(BufferNode::Allocator* allocator);
135 ~G1DirtyCardQueueSet();
136
137 void set_primary_refinement_thread(G1ConcurrentRefineThread* thread) {
138 _primary_refinement_thread = thread;
139 }
140
141 // The number of parallel ids that can be claimed to allow collector or
142 // mutator threads to do card-processing work.
143 static uint num_par_ids();
144
145 static void handle_zero_index_for_thread(Thread* t);
146
147 // Either process the entire buffer and return true, or enqueue the
148 // buffer and return false. If the buffer is completely processed,
149 // it can be reused in place.
150 bool process_or_enqueue_completed_buffer(BufferNode* node);
151
152 virtual void enqueue_completed_buffer(BufferNode* node);
153
154 // If the number of completed buffers is > stop_at, then remove and
155 // return a completed buffer from the list. Otherwise, return NULL.
156 BufferNode* get_completed_buffer(size_t stop_at = 0);
157
158 // The number of cards in completed buffers. Read without synchronization.
159 size_t num_cards() const { return _num_cards; }
160
161 // Get/Set the number of cards that triggers log processing.
162 // Log processing should be done when the number of cards exceeds the
163 // threshold.
164 void set_process_cards_threshold(size_t sz) {
165 _process_cards_threshold = sz;
166 }
167 size_t process_cards_threshold() const {
168 return _process_cards_threshold;
169 }
170 static const size_t ProcessCardsThresholdNever = SIZE_MAX;
171
172 // Notify the consumer if the number of buffers crossed the threshold
173 void notify_if_necessary();
174
175 void merge_bufferlists(G1RedirtyCardsQueueSet* src);
176
177 G1BufferNodeList take_all_completed_buffers();
178
179 // If there are more than stop_at cards in the completed buffers, pop
180 // a buffer, refine its contents, and return true. Otherwise return
181 // false.
182 //
183 // Stops processing a buffer if SuspendibleThreadSet::should_yield(),
184 // returning the incompletely processed buffer to the completed buffer
185 // list, for later processing of the remainder.
186 //
187 // Increments *total_refined_cards by the number of cards processed and
188 // removed from the buffer.
189 bool refine_completed_buffer_concurrently(uint worker_id,
190 size_t stop_at,
191 size_t* total_refined_cards);
192
193 // If a full collection is happening, reset partial logs, and release
194 // completed ones: the full collection will make them all irrelevant.
195 void abandon_logs();
196
197 // If any threads have partial logs, add them to the global list of logs.
198 void concatenate_logs();
199
200 void set_max_cards(size_t m) {
201 _max_cards = m;
202 }
203 size_t max_cards() const {
204 return _max_cards;
205 }
206
207 void set_max_cards_padding(size_t padding) {
208 _max_cards_padding = padding;
209 }
210 size_t max_cards_padding() const {
211 return _max_cards_padding;
212 }
213
214 // Total dirty cards refined by mutator threads.
215 size_t total_mutator_refined_cards() const;
216 };
217
218 inline G1DirtyCardQueueSet* G1DirtyCardQueue::dirty_card_qset() const {
219 return static_cast<G1DirtyCardQueueSet*>(qset());
220 }
221
222 #endif // SHARE_GC_G1_G1DIRTYCARDQUEUE_HPP