Mon, 03 Aug 2009 12:59:30 -0700
6865703: G1: Parallelize hot card cache cleanup
Summary: Have the GC worker threads clear the hot card cache in parallel by having each worker thread claim a chunk of the card cache and process the cards in that chunk. The size of the chunks that each thread will claim is determined at VM initialization from the size of the card cache and the number of worker threads.
Reviewed-by: jmasa, tonyp
1 /*
2 * Copyright 2001-2007 Sun Microsystems, Inc. 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 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
20 * CA 95054 USA or visit www.sun.com if you need additional information or
21 * have any questions.
22 *
23 */
25 class FreeIdSet;
27 // A closure class for processing card table entries. Note that we don't
28 // require these closure objects to be stack-allocated.
29 class CardTableEntryClosure: public CHeapObj {
30 public:
31 // Process the card whose card table entry is "card_ptr". If returns
32 // "false", terminate the iteration early.
33 virtual bool do_card_ptr(jbyte* card_ptr, int worker_i = 0) = 0;
34 };
36 // A ptrQueue whose elements are "oops", pointers to object heads.
37 class DirtyCardQueue: public PtrQueue {
38 public:
39 DirtyCardQueue(PtrQueueSet* qset_, bool perm = false) :
40 PtrQueue(qset_, perm)
41 {
42 // Dirty card queues are always active.
43 _active = true;
44 }
45 // Apply the closure to all elements, and reset the index to make the
46 // buffer empty. If a closure application returns "false", return
47 // "false" immediately, halting the iteration. If "consume" is true,
48 // deletes processed entries from logs.
49 bool apply_closure(CardTableEntryClosure* cl,
50 bool consume = true,
51 size_t worker_i = 0);
53 // Apply the closure to all elements of "buf", down to "index"
54 // (inclusive.) If returns "false", then a closure application returned
55 // "false", and we return immediately. If "consume" is true, entries are
56 // set to NULL as they are processed, so they will not be processed again
57 // later.
58 static bool apply_closure_to_buffer(CardTableEntryClosure* cl,
59 void** buf, size_t index, size_t sz,
60 bool consume = true,
61 int worker_i = 0);
62 void **get_buf() { return _buf;}
63 void set_buf(void **buf) {_buf = buf;}
64 size_t get_index() { return _index;}
65 void reinitialize() { _buf = 0; _sz = 0; _index = 0;}
66 };
70 class DirtyCardQueueSet: public PtrQueueSet {
71 CardTableEntryClosure* _closure;
73 DirtyCardQueue _shared_dirty_card_queue;
75 // Override.
76 bool mut_process_buffer(void** buf);
78 // Protected by the _cbl_mon.
79 FreeIdSet* _free_ids;
81 // The number of completed buffers processed by mutator and rs thread,
82 // respectively.
83 jint _processed_buffers_mut;
84 jint _processed_buffers_rs_thread;
86 public:
87 DirtyCardQueueSet();
89 void initialize(Monitor* cbl_mon, Mutex* fl_lock,
90 int max_completed_queue = 0,
91 Mutex* lock = NULL, PtrQueueSet* fl_owner = NULL);
93 // The number of parallel ids that can be claimed to allow collector or
94 // mutator threads to do card-processing work.
95 static size_t num_par_ids();
97 static void handle_zero_index_for_thread(JavaThread* t);
99 // Register "blk" as "the closure" for all queues. Only one such closure
100 // is allowed. The "apply_closure_to_completed_buffer" method will apply
101 // this closure to a completed buffer, and "iterate_closure_all_threads"
102 // applies it to partially-filled buffers (the latter should only be done
103 // with the world stopped).
104 void set_closure(CardTableEntryClosure* closure);
106 // If there is a registered closure for buffers, apply it to all entries
107 // in all currently-active buffers. This should only be applied at a
108 // safepoint. (Currently must not be called in parallel; this should
109 // change in the future.) If "consume" is true, processed entries are
110 // discarded.
111 void iterate_closure_all_threads(bool consume = true,
112 size_t worker_i = 0);
114 // If there exists some completed buffer, pop it, then apply the
115 // registered closure to all its elements, nulling out those elements
116 // processed. If all elements are processed, returns "true". If no
117 // completed buffers exist, returns false. If a completed buffer exists,
118 // but is only partially completed before a "yield" happens, the
119 // partially completed buffer (with its processed elements set to NULL)
120 // is returned to the completed buffer set, and this call returns false.
121 bool apply_closure_to_completed_buffer(int worker_i = 0,
122 int stop_at = 0,
123 bool with_CAS = false);
124 bool apply_closure_to_completed_buffer_helper(int worker_i,
125 CompletedBufferNode* nd);
127 CompletedBufferNode* get_completed_buffer_CAS();
128 CompletedBufferNode* get_completed_buffer_lock(int stop_at);
129 // Applies the current closure to all completed buffers,
130 // non-consumptively.
131 void apply_closure_to_all_completed_buffers();
133 DirtyCardQueue* shared_dirty_card_queue() {
134 return &_shared_dirty_card_queue;
135 }
137 // If a full collection is happening, reset partial logs, and ignore
138 // completed ones: the full collection will make them all irrelevant.
139 void abandon_logs();
141 // If any threads have partial logs, add them to the global list of logs.
142 void concatenate_logs();
143 void clear_n_completed_buffers() { _n_completed_buffers = 0;}
145 jint processed_buffers_mut() {
146 return _processed_buffers_mut;
147 }
148 jint processed_buffers_rs_thread() {
149 return _processed_buffers_rs_thread;
150 }
152 };