ViewVC Help
View File | Revision Log | Show Annotations | Download File | Root Listing
root/jsr166/jsr166/src/jsr166y/ForkJoinPool.java
(Generate patch)

Comparing jsr166/src/jsr166y/ForkJoinPool.java (file contents):
Revision 1.46 by dl, Tue Aug 4 14:19:00 2009 UTC vs.
Revision 1.109 by jsr166, Fri Jul 1 18:20:11 2011 UTC

# Line 1 | Line 1
1   /*
2   * Written by Doug Lea with assistance from members of JCP JSR-166
3   * Expert Group and released to the public domain, as explained at
4 < * http://creativecommons.org/licenses/publicdomain
4 > * http://creativecommons.org/publicdomain/zero/1.0/
5   */
6  
7   package jsr166y;
8  
9 import java.util.concurrent.*;
10
9   import java.util.ArrayList;
10   import java.util.Arrays;
11   import java.util.Collection;
12   import java.util.Collections;
13   import java.util.List;
14 < import java.util.concurrent.locks.Condition;
14 > import java.util.Random;
15 > import java.util.concurrent.AbstractExecutorService;
16 > import java.util.concurrent.Callable;
17 > import java.util.concurrent.ExecutorService;
18 > import java.util.concurrent.Future;
19 > import java.util.concurrent.RejectedExecutionException;
20 > import java.util.concurrent.RunnableFuture;
21 > import java.util.concurrent.TimeUnit;
22 > import java.util.concurrent.atomic.AtomicInteger;
23   import java.util.concurrent.locks.LockSupport;
24   import java.util.concurrent.locks.ReentrantLock;
25 < import java.util.concurrent.atomic.AtomicInteger;
20 < import java.util.concurrent.atomic.AtomicLong;
25 > import java.util.concurrent.locks.Condition;
26  
27   /**
28   * An {@link ExecutorService} for running {@link ForkJoinTask}s.
29   * A {@code ForkJoinPool} provides the entry point for submissions
30 < * from non-{@code ForkJoinTask}s, as well as management and
31 < * monitoring operations.  
30 > * from non-{@code ForkJoinTask} clients, as well as management and
31 > * monitoring operations.
32   *
33   * <p>A {@code ForkJoinPool} differs from other kinds of {@link
34   * ExecutorService} mainly by virtue of employing
# Line 31 | Line 36 | import java.util.concurrent.atomic.Atomi
36   * execute subtasks created by other active tasks (eventually blocking
37   * waiting for work if none exist). This enables efficient processing
38   * when most tasks spawn other subtasks (as do most {@code
39 < * ForkJoinTask}s). A {@code ForkJoinPool} may also be used for mixed
40 < * execution of some plain {@code Runnable}- or {@code Callable}-
41 < * based activities along with {@code ForkJoinTask}s. When setting
37 < * {@linkplain #setAsyncMode async mode}, a {@code ForkJoinPool} may
38 < * also be appropriate for use with fine-grained tasks of any form
39 < * that are never joined. Otherwise, other {@code ExecutorService}
40 < * implementations are typically more appropriate choices.
39 > * ForkJoinTask}s). When setting <em>asyncMode</em> to true in
40 > * constructors, {@code ForkJoinPool}s may also be appropriate for use
41 > * with event-style tasks that are never joined.
42   *
43   * <p>A {@code ForkJoinPool} is constructed with a given target
44   * parallelism level; by default, equal to the number of available
45 < * processors. Unless configured otherwise via {@link
46 < * #setMaintainsParallelism}, the pool attempts to maintain this
47 < * number of active (or available) threads by dynamically adding,
48 < * suspending, or resuming internal worker threads, even if some tasks
49 < * are stalled waiting to join others. However, no such adjustments
50 < * are performed in the face of blocked IO or other unmanaged
51 < * synchronization. The nested {@link ManagedBlocker} interface
51 < * enables extension of the kinds of synchronization accommodated.
52 < * The target parallelism level may also be changed dynamically
53 < * ({@link #setParallelism}). The total number of threads may be
54 < * limited using method {@link #setMaximumPoolSize}, in which case it
55 < * may become possible for the activities of a pool to stall due to
56 < * the lack of available threads to process new tasks.
45 > * processors. The pool attempts to maintain enough active (or
46 > * available) threads by dynamically adding, suspending, or resuming
47 > * internal worker threads, even if some tasks are stalled waiting to
48 > * join others. However, no such adjustments are guaranteed in the
49 > * face of blocked IO or other unmanaged synchronization. The nested
50 > * {@link ManagedBlocker} interface enables extension of the kinds of
51 > * synchronization accommodated.
52   *
53   * <p>In addition to execution and lifecycle control methods, this
54   * class provides status check methods (for example
# Line 62 | Line 57 | import java.util.concurrent.atomic.Atomi
57   * {@link #toString} returns indications of pool state in a
58   * convenient form for informal monitoring.
59   *
60 + * <p> As is the case with other ExecutorServices, there are three
61 + * main task execution methods summarized in the following
62 + * table. These are designed to be used by clients not already engaged
63 + * in fork/join computations in the current pool.  The main forms of
64 + * these methods accept instances of {@code ForkJoinTask}, but
65 + * overloaded forms also allow mixed execution of plain {@code
66 + * Runnable}- or {@code Callable}- based activities as well.  However,
67 + * tasks that are already executing in a pool should normally
68 + * <em>NOT</em> use these pool execution methods, but instead use the
69 + * within-computation forms listed in the table.
70 + *
71 + * <table BORDER CELLPADDING=3 CELLSPACING=1>
72 + *  <tr>
73 + *    <td></td>
74 + *    <td ALIGN=CENTER> <b>Call from non-fork/join clients</b></td>
75 + *    <td ALIGN=CENTER> <b>Call from within fork/join computations</b></td>
76 + *  </tr>
77 + *  <tr>
78 + *    <td> <b>Arrange async execution</td>
79 + *    <td> {@link #execute(ForkJoinTask)}</td>
80 + *    <td> {@link ForkJoinTask#fork}</td>
81 + *  </tr>
82 + *  <tr>
83 + *    <td> <b>Await and obtain result</td>
84 + *    <td> {@link #invoke(ForkJoinTask)}</td>
85 + *    <td> {@link ForkJoinTask#invoke}</td>
86 + *  </tr>
87 + *  <tr>
88 + *    <td> <b>Arrange exec and obtain Future</td>
89 + *    <td> {@link #submit(ForkJoinTask)}</td>
90 + *    <td> {@link ForkJoinTask#fork} (ForkJoinTasks <em>are</em> Futures)</td>
91 + *  </tr>
92 + * </table>
93 + *
94   * <p><b>Sample Usage.</b> Normally a single {@code ForkJoinPool} is
95   * used for all parallel task execution in a program or subsystem.
96   * Otherwise, use would not usually outweigh the construction and
# Line 72 | Line 101 | import java.util.concurrent.atomic.Atomi
101   * daemon} mode, there is typically no need to explicitly {@link
102   * #shutdown} such a pool upon program exit.
103   *
104 < * <pre>
104 > *  <pre> {@code
105   * static final ForkJoinPool mainPool = new ForkJoinPool();
106   * ...
107   * public void sort(long[] array) {
108   *   mainPool.invoke(new SortTask(array, 0, array.length));
109 < * }
81 < * </pre>
109 > * }}</pre>
110   *
111   * <p><b>Implementation notes</b>: This implementation restricts the
112   * maximum number of running threads to 32767. Attempts to create
113 < * pools with greater than the maximum result in
113 > * pools with greater than the maximum number result in
114   * {@code IllegalArgumentException}.
115   *
116 + * <p>This implementation rejects submitted tasks (that is, by throwing
117 + * {@link RejectedExecutionException}) only when the pool is shut down
118 + * or internal resources have been exhausted.
119 + *
120   * @since 1.7
121   * @author Doug Lea
122   */
123   public class ForkJoinPool extends AbstractExecutorService {
124  
125      /*
126 <     * See the extended comments interspersed below for design,
127 <     * rationale, and walkthroughs.
126 >     * Implementation Overview
127 >     *
128 >     * This class provides the central bookkeeping and control for a
129 >     * set of worker threads: Submissions from non-FJ threads enter
130 >     * into a submission queue. Workers take these tasks and typically
131 >     * split them into subtasks that may be stolen by other workers.
132 >     * Preference rules give first priority to processing tasks from
133 >     * their own queues (LIFO or FIFO, depending on mode), then to
134 >     * randomized FIFO steals of tasks in other worker queues, and
135 >     * lastly to new submissions.
136 >     *
137 >     * The main throughput advantages of work-stealing stem from
138 >     * decentralized control -- workers mostly take tasks from
139 >     * themselves or each other. We cannot negate this in the
140 >     * implementation of other management responsibilities. The main
141 >     * tactic for avoiding bottlenecks is packing nearly all
142 >     * essentially atomic control state into a single 64bit volatile
143 >     * variable ("ctl"). This variable is read on the order of 10-100
144 >     * times as often as it is modified (always via CAS). (There is
145 >     * some additional control state, for example variable "shutdown"
146 >     * for which we can cope with uncoordinated updates.)  This
147 >     * streamlines synchronization and control at the expense of messy
148 >     * constructions needed to repack status bits upon updates.
149 >     * Updates tend not to contend with each other except during
150 >     * bursts while submitted tasks begin or end.  In some cases when
151 >     * they do contend, threads can instead do something else
152 >     * (usually, scan for tasks) until contention subsides.
153 >     *
154 >     * To enable packing, we restrict maximum parallelism to (1<<15)-1
155 >     * (which is far in excess of normal operating range) to allow
156 >     * ids, counts, and their negations (used for thresholding) to fit
157 >     * into 16bit fields.
158 >     *
159 >     * Recording Workers.  Workers are recorded in the "workers" array
160 >     * that is created upon pool construction and expanded if (rarely)
161 >     * necessary.  This is an array as opposed to some other data
162 >     * structure to support index-based random steals by workers.
163 >     * Updates to the array recording new workers and unrecording
164 >     * terminated ones are protected from each other by a seqLock
165 >     * (scanGuard) but the array is otherwise concurrently readable,
166 >     * and accessed directly by workers. To simplify index-based
167 >     * operations, the array size is always a power of two, and all
168 >     * readers must tolerate null slots. To avoid flailing during
169 >     * start-up, the array is presized to hold twice #parallelism
170 >     * workers (which is unlikely to need further resizing during
171 >     * execution). But to avoid dealing with so many null slots,
172 >     * variable scanGuard includes a mask for the nearest power of two
173 >     * that contains all current workers.  All worker thread creation
174 >     * is on-demand, triggered by task submissions, replacement of
175 >     * terminated workers, and/or compensation for blocked
176 >     * workers. However, all other support code is set up to work with
177 >     * other policies.  To ensure that we do not hold on to worker
178 >     * references that would prevent GC, ALL accesses to workers are
179 >     * via indices into the workers array (which is one source of some
180 >     * of the messy code constructions here). In essence, the workers
181 >     * array serves as a weak reference mechanism. Thus for example
182 >     * the wait queue field of ctl stores worker indices, not worker
183 >     * references.  Access to the workers in associated methods (for
184 >     * example signalWork) must both index-check and null-check the
185 >     * IDs. All such accesses ignore bad IDs by returning out early
186 >     * from what they are doing, since this can only be associated
187 >     * with termination, in which case it is OK to give up.
188 >     *
189 >     * All uses of the workers array, as well as queue arrays, check
190 >     * that the array is non-null (even if previously non-null). This
191 >     * allows nulling during termination, which is currently not
192 >     * necessary, but remains an option for resource-revocation-based
193 >     * shutdown schemes.
194 >     *
195 >     * Wait Queuing. Unlike HPC work-stealing frameworks, we cannot
196 >     * let workers spin indefinitely scanning for tasks when none can
197 >     * be found immediately, and we cannot start/resume workers unless
198 >     * there appear to be tasks available.  On the other hand, we must
199 >     * quickly prod them into action when new tasks are submitted or
200 >     * generated.  We park/unpark workers after placing in an event
201 >     * wait queue when they cannot find work. This "queue" is actually
202 >     * a simple Treiber stack, headed by the "id" field of ctl, plus a
203 >     * 15bit counter value to both wake up waiters (by advancing their
204 >     * count) and avoid ABA effects. Successors are held in worker
205 >     * field "nextWait".  Queuing deals with several intrinsic races,
206 >     * mainly that a task-producing thread can miss seeing (and
207 >     * signalling) another thread that gave up looking for work but
208 >     * has not yet entered the wait queue. We solve this by requiring
209 >     * a full sweep of all workers both before (in scan()) and after
210 >     * (in tryAwaitWork()) a newly waiting worker is added to the wait
211 >     * queue. During a rescan, the worker might release some other
212 >     * queued worker rather than itself, which has the same net
213 >     * effect. Because enqueued workers may actually be rescanning
214 >     * rather than waiting, we set and clear the "parked" field of
215 >     * ForkJoinWorkerThread to reduce unnecessary calls to unpark.
216 >     * (Use of the parked field requires a secondary recheck to avoid
217 >     * missed signals.)
218 >     *
219 >     * Signalling.  We create or wake up workers only when there
220 >     * appears to be at least one task they might be able to find and
221 >     * execute.  When a submission is added or another worker adds a
222 >     * task to a queue that previously had two or fewer tasks, they
223 >     * signal waiting workers (or trigger creation of new ones if
224 >     * fewer than the given parallelism level -- see signalWork).
225 >     * These primary signals are buttressed by signals during rescans
226 >     * as well as those performed when a worker steals a task and
227 >     * notices that there are more tasks too; together these cover the
228 >     * signals needed in cases when more than two tasks are pushed
229 >     * but untaken.
230 >     *
231 >     * Trimming workers. To release resources after periods of lack of
232 >     * use, a worker starting to wait when the pool is quiescent will
233 >     * time out and terminate if the pool has remained quiescent for
234 >     * SHRINK_RATE nanosecs. This will slowly propagate, eventually
235 >     * terminating all workers after long periods of non-use.
236 >     *
237 >     * Submissions. External submissions are maintained in an
238 >     * array-based queue that is structured identically to
239 >     * ForkJoinWorkerThread queues except for the use of
240 >     * submissionLock in method addSubmission. Unlike the case for
241 >     * worker queues, multiple external threads can add new
242 >     * submissions, so adding requires a lock.
243 >     *
244 >     * Compensation. Beyond work-stealing support and lifecycle
245 >     * control, the main responsibility of this framework is to take
246 >     * actions when one worker is waiting to join a task stolen (or
247 >     * always held by) another.  Because we are multiplexing many
248 >     * tasks on to a pool of workers, we can't just let them block (as
249 >     * in Thread.join).  We also cannot just reassign the joiner's
250 >     * run-time stack with another and replace it later, which would
251 >     * be a form of "continuation", that even if possible is not
252 >     * necessarily a good idea since we sometimes need both an
253 >     * unblocked task and its continuation to progress. Instead we
254 >     * combine two tactics:
255 >     *
256 >     *   Helping: Arranging for the joiner to execute some task that it
257 >     *      would be running if the steal had not occurred.  Method
258 >     *      ForkJoinWorkerThread.joinTask tracks joining->stealing
259 >     *      links to try to find such a task.
260 >     *
261 >     *   Compensating: Unless there are already enough live threads,
262 >     *      method tryPreBlock() may create or re-activate a spare
263 >     *      thread to compensate for blocked joiners until they
264 >     *      unblock.
265 >     *
266 >     * The ManagedBlocker extension API can't use helping so relies
267 >     * only on compensation in method awaitBlocker.
268 >     *
269 >     * It is impossible to keep exactly the target parallelism number
270 >     * of threads running at any given time.  Determining the
271 >     * existence of conservatively safe helping targets, the
272 >     * availability of already-created spares, and the apparent need
273 >     * to create new spares are all racy and require heuristic
274 >     * guidance, so we rely on multiple retries of each.  Currently,
275 >     * in keeping with on-demand signalling policy, we compensate only
276 >     * if blocking would leave less than one active (non-waiting,
277 >     * non-blocked) worker. Additionally, to avoid some false alarms
278 >     * due to GC, lagging counters, system activity, etc, compensated
279 >     * blocking for joins is only attempted after rechecks stabilize
280 >     * (retries are interspersed with Thread.yield, for good
281 >     * citizenship).  The variable blockedCount, incremented before
282 >     * blocking and decremented after, is sometimes needed to
283 >     * distinguish cases of waiting for work vs blocking on joins or
284 >     * other managed sync. Both cases are equivalent for most pool
285 >     * control, so we can update non-atomically. (Additionally,
286 >     * contention on blockedCount alleviates some contention on ctl).
287 >     *
288 >     * Shutdown and Termination. A call to shutdownNow atomically sets
289 >     * the ctl stop bit and then (non-atomically) sets each workers
290 >     * "terminate" status, cancels all unprocessed tasks, and wakes up
291 >     * all waiting workers.  Detecting whether termination should
292 >     * commence after a non-abrupt shutdown() call requires more work
293 >     * and bookkeeping. We need consensus about quiescence (i.e., that
294 >     * there is no more work) which is reflected in active counts so
295 >     * long as there are no current blockers, as well as possible
296 >     * re-evaluations during independent changes in blocking or
297 >     * quiescing workers.
298 >     *
299 >     * Style notes: There is a lot of representation-level coupling
300 >     * among classes ForkJoinPool, ForkJoinWorkerThread, and
301 >     * ForkJoinTask.  Most fields of ForkJoinWorkerThread maintain
302 >     * data structures managed by ForkJoinPool, so are directly
303 >     * accessed.  Conversely we allow access to "workers" array by
304 >     * workers, and direct access to ForkJoinTask.status by both
305 >     * ForkJoinPool and ForkJoinWorkerThread.  There is little point
306 >     * trying to reduce this, since any associated future changes in
307 >     * representations will need to be accompanied by algorithmic
308 >     * changes anyway. All together, these low-level implementation
309 >     * choices produce as much as a factor of 4 performance
310 >     * improvement compared to naive implementations, and enable the
311 >     * processing of billions of tasks per second, at the expense of
312 >     * some ugliness.
313 >     *
314 >     * Methods signalWork() and scan() are the main bottlenecks so are
315 >     * especially heavily micro-optimized/mangled.  There are lots of
316 >     * inline assignments (of form "while ((local = field) != 0)")
317 >     * which are usually the simplest way to ensure the required read
318 >     * orderings (which are sometimes critical). This leads to a
319 >     * "C"-like style of listing declarations of these locals at the
320 >     * heads of methods or blocks.  There are several occurrences of
321 >     * the unusual "do {} while (!cas...)"  which is the simplest way
322 >     * to force an update of a CAS'ed variable. There are also other
323 >     * coding oddities that help some methods perform reasonably even
324 >     * when interpreted (not compiled).
325 >     *
326 >     * The order of declarations in this file is: (1) declarations of
327 >     * statics (2) fields (along with constants used when unpacking
328 >     * some of them), listed in an order that tends to reduce
329 >     * contention among them a bit under most JVMs.  (3) internal
330 >     * control methods (4) callbacks and other support for
331 >     * ForkJoinTask and ForkJoinWorkerThread classes, (5) exported
332 >     * methods (plus a few little helpers). (6) static block
333 >     * initializing all statics in a minimally dependent order.
334       */
335  
98    /** Mask for packing and unpacking shorts */
99    private static final int  shortMask = 0xffff;
100
101    /** Max pool size -- must be a power of two minus 1 */
102    private static final int MAX_THREADS =  0x7FFF;
103
336      /**
337       * Factory for creating new {@link ForkJoinWorkerThread}s.
338       * A {@code ForkJoinWorkerThreadFactory} must be defined and used
# Line 112 | Line 344 | public class ForkJoinPool extends Abstra
344           * Returns a new worker thread operating in the given pool.
345           *
346           * @param pool the pool this thread works in
347 <         * @throws NullPointerException if pool is null
347 >         * @throws NullPointerException if the pool is null
348           */
349          public ForkJoinWorkerThread newThread(ForkJoinPool pool);
350      }
# Line 121 | Line 353 | public class ForkJoinPool extends Abstra
353       * Default ForkJoinWorkerThreadFactory implementation; creates a
354       * new ForkJoinWorkerThread.
355       */
356 <    static class  DefaultForkJoinWorkerThreadFactory
356 >    static class DefaultForkJoinWorkerThreadFactory
357          implements ForkJoinWorkerThreadFactory {
358          public ForkJoinWorkerThread newThread(ForkJoinPool pool) {
359 <            try {
128 <                return new ForkJoinWorkerThread(pool);
129 <            } catch (OutOfMemoryError oom)  {
130 <                return null;
131 <            }
359 >            return new ForkJoinWorkerThread(pool);
360          }
361      }
362  
# Line 137 | Line 365 | public class ForkJoinPool extends Abstra
365       * overridden in ForkJoinPool constructors.
366       */
367      public static final ForkJoinWorkerThreadFactory
368 <        defaultForkJoinWorkerThreadFactory =
141 <        new DefaultForkJoinWorkerThreadFactory();
368 >        defaultForkJoinWorkerThreadFactory;
369  
370      /**
371       * Permission required for callers of methods that may start or
372       * kill threads.
373       */
374 <    private static final RuntimePermission modifyThreadPermission =
148 <        new RuntimePermission("modifyThread");
374 >    private static final RuntimePermission modifyThreadPermission;
375  
376      /**
377       * If there is a security manager, makes sure caller has
# Line 160 | Line 386 | public class ForkJoinPool extends Abstra
386      /**
387       * Generator for assigning sequence numbers as pool names.
388       */
389 <    private static final AtomicInteger poolNumberGenerator =
164 <        new AtomicInteger();
389 >    private static final AtomicInteger poolNumberGenerator;
390  
391      /**
392 <     * Array holding all worker threads in the pool. Initialized upon
393 <     * first use. Array size must be a power of two.  Updates and
394 <     * replacements are protected by workerLock, but it is always kept
395 <     * in a consistent enough state to be randomly accessed without
396 <     * locking by workers performing work-stealing.
392 >     * Generator for initial random seeds for worker victim
393 >     * selection. This is used only to create initial seeds. Random
394 >     * steals use a cheaper xorshift generator per steal attempt. We
395 >     * don't expect much contention on seedGenerator, so just use a
396 >     * plain Random.
397       */
398 <    volatile ForkJoinWorkerThread[] workers;
398 >    static final Random workerSeedGenerator;
399  
400      /**
401 <     * Lock protecting access to workers.
401 >     * Array holding all worker threads in the pool.  Initialized upon
402 >     * construction. Array size must be a power of two.  Updates and
403 >     * replacements are protected by scanGuard, but the array is
404 >     * always kept in a consistent enough state to be randomly
405 >     * accessed without locking by workers performing work-stealing,
406 >     * as well as other traversal-based methods in this class, so long
407 >     * as reads memory-acquire by first reading ctl. All readers must
408 >     * tolerate that some array slots may be null.
409       */
410 <    private final ReentrantLock workerLock;
410 >    ForkJoinWorkerThread[] workers;
411  
412      /**
413 <     * Condition for awaitTermination.
413 >     * Initial size for submission queue array. Must be a power of
414 >     * two.  In many applications, these always stay small so we use a
415 >     * small initial cap.
416       */
417 <    private final Condition termination;
417 >    private static final int INITIAL_QUEUE_CAPACITY = 8;
418 >
419 >    /**
420 >     * Maximum size for submission queue array. Must be a power of two
421 >     * less than or equal to 1 << (31 - width of array entry) to
422 >     * ensure lack of index wraparound, but is capped at a lower
423 >     * value to help users trap runaway computations.
424 >     */
425 >    private static final int MAXIMUM_QUEUE_CAPACITY = 1 << 24; // 16M
426  
427      /**
428 <     * The uncaught exception handler used when any worker
187 <     * abruptly terminates
428 >     * Array serving as submission queue. Initialized upon construction.
429       */
430 <    private Thread.UncaughtExceptionHandler ueh;
430 >    private ForkJoinTask<?>[] submissionQueue;
431 >
432 >    /**
433 >     * Lock protecting submissions array for addSubmission
434 >     */
435 >    private final ReentrantLock submissionLock;
436 >
437 >    /**
438 >     * Condition for awaitTermination, using submissionLock for
439 >     * convenience.
440 >     */
441 >    private final Condition termination;
442  
443      /**
444       * Creation factory for worker threads.
# Line 194 | Line 446 | public class ForkJoinPool extends Abstra
446      private final ForkJoinWorkerThreadFactory factory;
447  
448      /**
449 <     * Head of stack of threads that were created to maintain
450 <     * parallelism when other threads blocked, but have since
451 <     * suspended when the parallelism level rose.
449 >     * The uncaught exception handler used when any worker abruptly
450 >     * terminates.
451 >     */
452 >    final Thread.UncaughtExceptionHandler ueh;
453 >
454 >    /**
455 >     * Prefix for assigning names to worker threads
456       */
457 <    private volatile WaitQueueNode spareStack;
457 >    private final String workerNamePrefix;
458  
459      /**
460       * Sum of per-thread steal counts, updated only when threads are
461       * idle or terminating.
462       */
463 <    private final AtomicLong stealCount;
463 >    private volatile long stealCount;
464  
465      /**
466 <     * Queue for external submissions.
466 >     * Main pool control -- a long packed with:
467 >     * AC: Number of active running workers minus target parallelism (16 bits)
468 >     * TC: Number of total workers minus target parallelism (16 bits)
469 >     * ST: true if pool is terminating (1 bit)
470 >     * EC: the wait count of top waiting thread (15 bits)
471 >     * ID: ~poolIndex of top of Treiber stack of waiting threads (16 bits)
472 >     *
473 >     * When convenient, we can extract the upper 32 bits of counts and
474 >     * the lower 32 bits of queue state, u = (int)(ctl >>> 32) and e =
475 >     * (int)ctl.  The ec field is never accessed alone, but always
476 >     * together with id and st. The offsets of counts by the target
477 >     * parallelism and the positionings of fields makes it possible to
478 >     * perform the most common checks via sign tests of fields: When
479 >     * ac is negative, there are not enough active workers, when tc is
480 >     * negative, there are not enough total workers, when id is
481 >     * negative, there is at least one waiting worker, and when e is
482 >     * negative, the pool is terminating.  To deal with these possibly
483 >     * negative fields, we use casts in and out of "short" and/or
484 >     * signed shifts to maintain signedness.
485       */
486 <    private final LinkedTransferQueue<ForkJoinTask<?>> submissionQueue;
486 >    volatile long ctl;
487 >
488 >    // bit positions/shifts for fields
489 >    private static final int  AC_SHIFT   = 48;
490 >    private static final int  TC_SHIFT   = 32;
491 >    private static final int  ST_SHIFT   = 31;
492 >    private static final int  EC_SHIFT   = 16;
493 >
494 >    // bounds
495 >    private static final int  MAX_ID     = 0x7fff;  // max poolIndex
496 >    private static final int  SMASK      = 0xffff;  // mask short bits
497 >    private static final int  SHORT_SIGN = 1 << 15;
498 >    private static final int  INT_SIGN   = 1 << 31;
499 >
500 >    // masks
501 >    private static final long STOP_BIT   = 0x0001L << ST_SHIFT;
502 >    private static final long AC_MASK    = ((long)SMASK) << AC_SHIFT;
503 >    private static final long TC_MASK    = ((long)SMASK) << TC_SHIFT;
504 >
505 >    // units for incrementing and decrementing
506 >    private static final long TC_UNIT    = 1L << TC_SHIFT;
507 >    private static final long AC_UNIT    = 1L << AC_SHIFT;
508 >
509 >    // masks and units for dealing with u = (int)(ctl >>> 32)
510 >    private static final int  UAC_SHIFT  = AC_SHIFT - 32;
511 >    private static final int  UTC_SHIFT  = TC_SHIFT - 32;
512 >    private static final int  UAC_MASK   = SMASK << UAC_SHIFT;
513 >    private static final int  UTC_MASK   = SMASK << UTC_SHIFT;
514 >    private static final int  UAC_UNIT   = 1 << UAC_SHIFT;
515 >    private static final int  UTC_UNIT   = 1 << UTC_SHIFT;
516 >
517 >    // masks and units for dealing with e = (int)ctl
518 >    private static final int  E_MASK     = 0x7fffffff; // no STOP_BIT
519 >    private static final int  EC_UNIT    = 1 << EC_SHIFT;
520  
521      /**
522 <     * Head of Treiber stack for barrier sync. See below for explanation.
522 >     * The target parallelism level.
523       */
524 <    private volatile WaitQueueNode syncStack;
524 >    final int parallelism;
525  
526      /**
527 <     * The count for event barrier
527 >     * Index (mod submission queue length) of next element to take
528 >     * from submission queue. Usage is identical to that for
529 >     * per-worker queues -- see ForkJoinWorkerThread internal
530 >     * documentation.
531       */
532 <    private volatile long eventCount;
532 >    volatile int queueBase;
533  
534      /**
535 <     * Pool number, just for assigning useful names to worker threads
535 >     * Index (mod submission queue length) of next element to add
536 >     * in submission queue. Usage is identical to that for
537 >     * per-worker queues -- see ForkJoinWorkerThread internal
538 >     * documentation.
539       */
540 <    private final int poolNumber;
540 >    int queueTop;
541  
542      /**
543 <     * The maximum allowed pool size
543 >     * True when shutdown() has been called.
544       */
545 <    private volatile int maxPoolSize;
545 >    volatile boolean shutdown;
546  
547      /**
548 <     * The desired parallelism level, updated only under workerLock.
548 >     * True if use local fifo, not default lifo, for local polling.
549 >     * Read by, and replicated by ForkJoinWorkerThreads.
550       */
551 <    private volatile int parallelism;
551 >    final boolean locallyFifo;
552  
553      /**
554 <     * True if use local fifo, not default lifo, for local polling
554 >     * The number of threads in ForkJoinWorkerThreads.helpQuiescePool.
555 >     * When non-zero, suppresses automatic shutdown when active
556 >     * counts become zero.
557       */
558 <    private volatile boolean locallyFifo;
558 >    volatile int quiescerCount;
559  
560      /**
561 <     * Holds number of total (i.e., created and not yet terminated)
246 <     * and running (i.e., not blocked on joins or other managed sync)
247 <     * threads, packed into one int to ensure consistent snapshot when
248 <     * making decisions about creating and suspending spare
249 <     * threads. Updated only by CAS.  Note: CASes in
250 <     * updateRunningCount and preJoin assume that running active count
251 <     * is in low word, so need to be modified if this changes.
561 >     * The number of threads blocked in join.
562       */
563 <    private volatile int workerCounts;
254 <
255 <    private static int totalCountOf(int s)           { return s >>> 16;  }
256 <    private static int runningCountOf(int s)         { return s & shortMask; }
257 <    private static int workerCountsFor(int t, int r) { return (t << 16) + r; }
563 >    volatile int blockedCount;
564  
565      /**
566 <     * Adds delta (which may be negative) to running count.  This must
261 <     * be called before (with negative arg) and after (with positive)
262 <     * any managed synchronization (i.e., mainly, joins).
263 <     *
264 <     * @param delta the number to add
566 >     * Counter for worker Thread names (unrelated to their poolIndex)
567       */
568 <    final void updateRunningCount(int delta) {
267 <        int s;
268 <        do {} while (!casWorkerCounts(s = workerCounts, s + delta));
269 <    }
568 >    private volatile int nextWorkerNumber;
569  
570      /**
571 <     * Adds delta (which may be negative) to both total and running
273 <     * count.  This must be called upon creation and termination of
274 <     * worker threads.
275 <     *
276 <     * @param delta the number to add
571 >     * The index for the next created worker. Accessed under scanGuard.
572       */
573 <    private void updateWorkerCount(int delta) {
279 <        int d = delta + (delta << 16); // add to both lo and hi parts
280 <        int s;
281 <        do {} while (!casWorkerCounts(s = workerCounts, s + d));
282 <    }
573 >    private int nextWorkerIndex;
574  
575      /**
576 <     * Lifecycle control. High word contains runState, low word
577 <     * contains the number of workers that are (probably) executing
578 <     * tasks. This value is atomically incremented before a worker
579 <     * gets a task to run, and decremented when worker has no tasks
580 <     * and cannot find any. These two fields are bundled together to
581 <     * support correct termination triggering.  Note: activeCount
582 <     * CAS'es cheat by assuming active count is in low word, so need
583 <     * to be modified if this changes
576 >     * SeqLock and index masking for updates to workers array.  Locked
577 >     * when SG_UNIT is set. Unlocking clears bit by adding
578 >     * SG_UNIT. Staleness of read-only operations can be checked by
579 >     * comparing scanGuard to value before the reads. The low 16 bits
580 >     * (i.e, anding with SMASK) hold (the smallest power of two
581 >     * covering all worker indices, minus one, and is used to avoid
582 >     * dealing with large numbers of null slots when the workers array
583 >     * is overallocated.
584       */
585 <    private volatile int runControl;
585 >    volatile int scanGuard;
586  
587 <    // RunState values. Order among values matters
297 <    private static final int RUNNING     = 0;
298 <    private static final int SHUTDOWN    = 1;
299 <    private static final int TERMINATING = 2;
300 <    private static final int TERMINATED  = 3;
587 >    private static final int SG_UNIT = 1 << 16;
588  
589 <    private static int runStateOf(int c)             { return c >>> 16; }
590 <    private static int activeCountOf(int c)          { return c & shortMask; }
591 <    private static int runControlFor(int r, int a)   { return (r << 16) + a; }
589 >    /**
590 >     * The wakeup interval (in nanoseconds) for a worker waiting for a
591 >     * task when the pool is quiescent to instead try to shrink the
592 >     * number of workers.  The exact value does not matter too
593 >     * much. It must be short enough to release resources during
594 >     * sustained periods of idleness, but not so short that threads
595 >     * are continually re-created.
596 >     */
597 >    private static final long SHRINK_RATE =
598 >        4L * 1000L * 1000L * 1000L; // 4 seconds
599  
600      /**
601 <     * Tries incrementing active count; fails on contention.
602 <     * Called by workers before/during executing tasks.
601 >     * Top-level loop for worker threads: On each step: if the
602 >     * previous step swept through all queues and found no tasks, or
603 >     * there are excess threads, then possibly blocks. Otherwise,
604 >     * scans for and, if found, executes a task. Returns when pool
605 >     * and/or worker terminate.
606       *
607 <     * @return true on success
607 >     * @param w the worker
608       */
609 <    final boolean tryIncrementActiveCount() {
610 <        int c = runControl;
611 <        return casRunControl(c, c+1);
609 >    final void work(ForkJoinWorkerThread w) {
610 >        boolean swept = false;                // true on empty scans
611 >        long c;
612 >        while (!w.terminate && (int)(c = ctl) >= 0) {
613 >            int a;                            // active count
614 >            if (!swept && (a = (int)(c >> AC_SHIFT)) <= 0)
615 >                swept = scan(w, a);
616 >            else if (tryAwaitWork(w, c))
617 >                swept = false;
618 >        }
619      }
620  
621 +    // Signalling
622 +
623      /**
624 <     * Tries decrementing active count; fails on contention.
319 <     * Possibly triggers termination on success.
320 <     * Called by workers when they can't find tasks.
321 <     *
322 <     * @return true on success
624 >     * Wakes up or creates a worker.
625       */
626 <    final boolean tryDecrementActiveCount() {
627 <        int c = runControl;
628 <        int nextc = c - 1;
629 <        if (!casRunControl(c, nextc))
630 <            return false;
631 <        if (canTerminateOnShutdown(nextc))
632 <            terminateOnShutdown();
633 <        return true;
626 >    final void signalWork() {
627 >        /*
628 >         * The while condition is true if: (there is are too few total
629 >         * workers OR there is at least one waiter) AND (there are too
630 >         * few active workers OR the pool is terminating).  The value
631 >         * of e distinguishes the remaining cases: zero (no waiters)
632 >         * for create, negative if terminating (in which case do
633 >         * nothing), else release a waiter. The secondary checks for
634 >         * release (non-null array etc) can fail if the pool begins
635 >         * terminating after the test, and don't impose any added cost
636 >         * because JVMs must perform null and bounds checks anyway.
637 >         */
638 >        long c; int e, u;
639 >        while ((((e = (int)(c = ctl)) | (u = (int)(c >>> 32))) &
640 >                (INT_SIGN|SHORT_SIGN)) == (INT_SIGN|SHORT_SIGN) && e >= 0) {
641 >            if (e > 0) {                         // release a waiting worker
642 >                int i; ForkJoinWorkerThread w; ForkJoinWorkerThread[] ws;
643 >                if ((ws = workers) == null ||
644 >                    (i = ~e & SMASK) >= ws.length ||
645 >                    (w = ws[i]) == null)
646 >                    break;
647 >                long nc = (((long)(w.nextWait & E_MASK)) |
648 >                           ((long)(u + UAC_UNIT) << 32));
649 >                if (w.eventCount == e &&
650 >                    UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc)) {
651 >                    w.eventCount = (e + EC_UNIT) & E_MASK;
652 >                    if (w.parked)
653 >                        UNSAFE.unpark(w);
654 >                    break;
655 >                }
656 >            }
657 >            else if (UNSAFE.compareAndSwapLong
658 >                     (this, ctlOffset, c,
659 >                      (long)(((u + UTC_UNIT) & UTC_MASK) |
660 >                             ((u + UAC_UNIT) & UAC_MASK)) << 32)) {
661 >                addWorker();
662 >                break;
663 >            }
664 >        }
665      }
666  
667      /**
668 <     * Returns {@code true} if argument represents zero active count
669 <     * and nonzero runstate, which is the triggering condition for
670 <     * terminating on shutdown.
668 >     * Variant of signalWork to help release waiters on rescans.
669 >     * Tries once to release a waiter if active count < 0.
670 >     *
671 >     * @return false if failed due to contention, else true
672       */
673 <    private static boolean canTerminateOnShutdown(int c) {
674 <        // i.e. least bit is nonzero runState bit
675 <        return ((c & -c) >>> 16) != 0;
673 >    private boolean tryReleaseWaiter() {
674 >        long c; int e, i; ForkJoinWorkerThread w; ForkJoinWorkerThread[] ws;
675 >        if ((e = (int)(c = ctl)) > 0 &&
676 >            (int)(c >> AC_SHIFT) < 0 &&
677 >            (ws = workers) != null &&
678 >            (i = ~e & SMASK) < ws.length &&
679 >            (w = ws[i]) != null) {
680 >            long nc = ((long)(w.nextWait & E_MASK) |
681 >                       ((c + AC_UNIT) & (AC_MASK|TC_MASK)));
682 >            if (w.eventCount != e ||
683 >                !UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc))
684 >                return false;
685 >            w.eventCount = (e + EC_UNIT) & E_MASK;
686 >            if (w.parked)
687 >                UNSAFE.unpark(w);
688 >        }
689 >        return true;
690      }
691  
692 +    // Scanning for tasks
693 +
694      /**
695 <     * Transition run state to at least the given state. Return true
696 <     * if not already at least given state.
695 >     * Scans for and, if found, executes one task. Scans start at a
696 >     * random index of workers array, and randomly select the first
697 >     * (2*#workers)-1 probes, and then, if all empty, resort to 2
698 >     * circular sweeps, which is necessary to check quiescence. and
699 >     * taking a submission only if no stealable tasks were found.  The
700 >     * steal code inside the loop is a specialized form of
701 >     * ForkJoinWorkerThread.deqTask, followed bookkeeping to support
702 >     * helpJoinTask and signal propagation. The code for submission
703 >     * queues is almost identical. On each steal, the worker completes
704 >     * not only the task, but also all local tasks that this task may
705 >     * have generated. On detecting staleness or contention when
706 >     * trying to take a task, this method returns without finishing
707 >     * sweep, which allows global state rechecks before retry.
708 >     *
709 >     * @param w the worker
710 >     * @param a the number of active workers
711 >     * @return true if swept all queues without finding a task
712       */
713 <    private boolean transitionRunStateTo(int state) {
714 <        for (;;) {
715 <            int c = runControl;
716 <            if (runStateOf(c) >= state)
713 >    private boolean scan(ForkJoinWorkerThread w, int a) {
714 >        int g = scanGuard; // mask 0 avoids useless scans if only one active
715 >        int m = (parallelism == 1 - a && blockedCount == 0) ? 0 : g & SMASK;
716 >        ForkJoinWorkerThread[] ws = workers;
717 >        if (ws == null || ws.length <= m)         // staleness check
718 >            return false;
719 >        for (int r = w.seed, k = r, j = -(m + m); j <= m + m; ++j) {
720 >            ForkJoinTask<?> t; ForkJoinTask<?>[] q; int b, i;
721 >            ForkJoinWorkerThread v = ws[k & m];
722 >            if (v != null && (b = v.queueBase) != v.queueTop &&
723 >                (q = v.queue) != null && (i = (q.length - 1) & b) >= 0) {
724 >                long u = (i << ASHIFT) + ABASE;
725 >                if ((t = q[i]) != null && v.queueBase == b &&
726 >                    UNSAFE.compareAndSwapObject(q, u, t, null)) {
727 >                    int d = (v.queueBase = b + 1) - v.queueTop;
728 >                    v.stealHint = w.poolIndex;
729 >                    if (d != 0)
730 >                        signalWork();             // propagate if nonempty
731 >                    w.execTask(t);
732 >                }
733 >                r ^= r << 13; r ^= r >>> 17; w.seed = r ^ (r << 5);
734 >                return false;                     // store next seed
735 >            }
736 >            else if (j < 0) {                     // xorshift
737 >                r ^= r << 13; r ^= r >>> 17; k = r ^= r << 5;
738 >            }
739 >            else
740 >                ++k;
741 >        }
742 >        if (scanGuard != g)                       // staleness check
743 >            return false;
744 >        else {                                    // try to take submission
745 >            ForkJoinTask<?> t; ForkJoinTask<?>[] q; int b, i;
746 >            if ((b = queueBase) != queueTop &&
747 >                (q = submissionQueue) != null &&
748 >                (i = (q.length - 1) & b) >= 0) {
749 >                long u = (i << ASHIFT) + ABASE;
750 >                if ((t = q[i]) != null && queueBase == b &&
751 >                    UNSAFE.compareAndSwapObject(q, u, t, null)) {
752 >                    queueBase = b + 1;
753 >                    w.execTask(t);
754 >                }
755                  return false;
756 <            if (casRunControl(c, runControlFor(state, activeCountOf(c))))
756 >            }
757 >            return true;                         // all queues empty
758 >        }
759 >    }
760 >
761 >    /**
762 >     * Tries to enqueue worker w in wait queue and await change in
763 >     * worker's eventCount.  If the pool is quiescent and there is
764 >     * more than one worker, possibly terminates worker upon exit.
765 >     * Otherwise, before blocking, rescans queues to avoid missed
766 >     * signals.  Upon finding work, releases at least one worker
767 >     * (which may be the current worker). Rescans restart upon
768 >     * detected staleness or failure to release due to
769 >     * contention. Note the unusual conventions about Thread.interrupt
770 >     * here and elsewhere: Because interrupts are used solely to alert
771 >     * threads to check termination, which is checked here anyway, we
772 >     * clear status (using Thread.interrupted) before any call to
773 >     * park, so that park does not immediately return due to status
774 >     * being set via some other unrelated call to interrupt in user
775 >     * code.
776 >     *
777 >     * @param w the calling worker
778 >     * @param c the ctl value on entry
779 >     * @return true if waited or another thread was released upon enq
780 >     */
781 >    private boolean tryAwaitWork(ForkJoinWorkerThread w, long c) {
782 >        int v = w.eventCount;
783 >        w.nextWait = (int)c;                      // w's successor record
784 >        long nc = (long)(v & E_MASK) | ((c - AC_UNIT) & (AC_MASK|TC_MASK));
785 >        if (ctl != c || !UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc)) {
786 >            long d = ctl; // return true if lost to a deq, to force scan
787 >            return (int)d != (int)c && (d & AC_MASK) >= (c & AC_MASK);
788 >        }
789 >        for (int sc = w.stealCount; sc != 0;) {   // accumulate stealCount
790 >            long s = stealCount;
791 >            if (UNSAFE.compareAndSwapLong(this, stealCountOffset, s, s + sc))
792 >                sc = w.stealCount = 0;
793 >            else if (w.eventCount != v)
794 >                return true;                      // update next time
795 >        }
796 >        if ((!shutdown || !tryTerminate(false)) &&
797 >            (int)c != 0 && parallelism + (int)(nc >> AC_SHIFT) == 0 &&
798 >            blockedCount == 0 && quiescerCount == 0)
799 >            idleAwaitWork(w, nc, c, v);           // quiescent
800 >        for (boolean rescanned = false;;) {
801 >            if (w.eventCount != v)
802                  return true;
803 +            if (!rescanned) {
804 +                int g = scanGuard, m = g & SMASK;
805 +                ForkJoinWorkerThread[] ws = workers;
806 +                if (ws != null && m < ws.length) {
807 +                    rescanned = true;
808 +                    for (int i = 0; i <= m; ++i) {
809 +                        ForkJoinWorkerThread u = ws[i];
810 +                        if (u != null) {
811 +                            if (u.queueBase != u.queueTop &&
812 +                                !tryReleaseWaiter())
813 +                                rescanned = false; // contended
814 +                            if (w.eventCount != v)
815 +                                return true;
816 +                        }
817 +                    }
818 +                }
819 +                if (scanGuard != g ||              // stale
820 +                    (queueBase != queueTop && !tryReleaseWaiter()))
821 +                    rescanned = false;
822 +                if (!rescanned)
823 +                    Thread.yield();                // reduce contention
824 +                else
825 +                    Thread.interrupted();          // clear before park
826 +            }
827 +            else {
828 +                w.parked = true;                   // must recheck
829 +                if (w.eventCount != v) {
830 +                    w.parked = false;
831 +                    return true;
832 +                }
833 +                LockSupport.park(this);
834 +                rescanned = w.parked = false;
835 +            }
836          }
837      }
838  
839      /**
840 <     * Controls whether to add spares to maintain parallelism
841 <     */
842 <    private volatile boolean maintainsParallelism;
840 >     * If inactivating worker w has caused pool to become
841 >     * quiescent, check for pool termination, and wait for event
842 >     * for up to SHRINK_RATE nanosecs (rescans are unnecessary in
843 >     * this case because quiescence reflects consensus about lack
844 >     * of work). On timeout, if ctl has not changed, terminate the
845 >     * worker. Upon its termination (see deregisterWorker), it may
846 >     * wake up another worker to possibly repeat this process.
847 >     *
848 >     * @param w the calling worker
849 >     * @param currentCtl the ctl value after enqueuing w
850 >     * @param prevCtl the ctl value if w terminated
851 >     * @param v the eventCount w awaits change
852 >     */
853 >    private void idleAwaitWork(ForkJoinWorkerThread w, long currentCtl,
854 >                               long prevCtl, int v) {
855 >        if (w.eventCount == v) {
856 >            if (shutdown)
857 >                tryTerminate(false);
858 >            ForkJoinTask.helpExpungeStaleExceptions(); // help clean weak refs
859 >            while (ctl == currentCtl) {
860 >                long startTime = System.nanoTime();
861 >                w.parked = true;
862 >                if (w.eventCount == v)             // must recheck
863 >                    LockSupport.parkNanos(this, SHRINK_RATE);
864 >                w.parked = false;
865 >                if (w.eventCount != v)
866 >                    break;
867 >                else if (System.nanoTime() - startTime <
868 >                         SHRINK_RATE - (SHRINK_RATE / 10)) // timing slop
869 >                    Thread.interrupted();          // spurious wakeup
870 >                else if (UNSAFE.compareAndSwapLong(this, ctlOffset,
871 >                                                   currentCtl, prevCtl)) {
872 >                    w.terminate = true;            // restore previous
873 >                    w.eventCount = ((int)currentCtl + EC_UNIT) & E_MASK;
874 >                    break;
875 >                }
876 >            }
877 >        }
878 >    }
879  
880 <    // Constructors
880 >    // Submissions
881  
882      /**
883 <     * Creates a {@code ForkJoinPool} with parallelism equal to {@link
884 <     * java.lang.Runtime#availableProcessors}, and using the {@linkplain
368 <     * #defaultForkJoinWorkerThreadFactory default thread factory}.
883 >     * Enqueues the given task in the submissionQueue.  Same idea as
884 >     * ForkJoinWorkerThread.pushTask except for use of submissionLock.
885       *
886 <     * @throws SecurityException if a security manager exists and
371 <     *         the caller is not permitted to modify threads
372 <     *         because it does not hold {@link
373 <     *         java.lang.RuntimePermission}{@code ("modifyThread")}
886 >     * @param t the task
887       */
888 <    public ForkJoinPool() {
889 <        this(Runtime.getRuntime().availableProcessors(),
890 <             defaultForkJoinWorkerThreadFactory);
888 >    private void addSubmission(ForkJoinTask<?> t) {
889 >        final ReentrantLock lock = this.submissionLock;
890 >        lock.lock();
891 >        try {
892 >            ForkJoinTask<?>[] q; int s, m;
893 >            if ((q = submissionQueue) != null) {    // ignore if queue removed
894 >                long u = (((s = queueTop) & (m = q.length-1)) << ASHIFT)+ABASE;
895 >                UNSAFE.putOrderedObject(q, u, t);
896 >                queueTop = s + 1;
897 >                if (s - queueBase == m)
898 >                    growSubmissionQueue();
899 >            }
900 >        } finally {
901 >            lock.unlock();
902 >        }
903 >        signalWork();
904      }
905  
906 +    //  (pollSubmission is defined below with exported methods)
907 +
908      /**
909 <     * Creates a {@code ForkJoinPool} with the indicated parallelism
910 <     * level and using the {@linkplain
383 <     * #defaultForkJoinWorkerThreadFactory default thread factory}.
384 <     *
385 <     * @param parallelism the parallelism level
386 <     * @throws IllegalArgumentException if parallelism less than or
387 <     * equal to zero
388 <     * @throws SecurityException if a security manager exists and
389 <     *         the caller is not permitted to modify threads
390 <     *         because it does not hold {@link
391 <     *         java.lang.RuntimePermission}{@code ("modifyThread")}
909 >     * Creates or doubles submissionQueue array.
910 >     * Basically identical to ForkJoinWorkerThread version.
911       */
912 <    public ForkJoinPool(int parallelism) {
913 <        this(parallelism, defaultForkJoinWorkerThreadFactory);
912 >    private void growSubmissionQueue() {
913 >        ForkJoinTask<?>[] oldQ = submissionQueue;
914 >        int size = oldQ != null ? oldQ.length << 1 : INITIAL_QUEUE_CAPACITY;
915 >        if (size > MAXIMUM_QUEUE_CAPACITY)
916 >            throw new RejectedExecutionException("Queue capacity exceeded");
917 >        if (size < INITIAL_QUEUE_CAPACITY)
918 >            size = INITIAL_QUEUE_CAPACITY;
919 >        ForkJoinTask<?>[] q = submissionQueue = new ForkJoinTask<?>[size];
920 >        int mask = size - 1;
921 >        int top = queueTop;
922 >        int oldMask;
923 >        if (oldQ != null && (oldMask = oldQ.length - 1) >= 0) {
924 >            for (int b = queueBase; b != top; ++b) {
925 >                long u = ((b & oldMask) << ASHIFT) + ABASE;
926 >                Object x = UNSAFE.getObjectVolatile(oldQ, u);
927 >                if (x != null && UNSAFE.compareAndSwapObject(oldQ, u, x, null))
928 >                    UNSAFE.putObjectVolatile
929 >                        (q, ((b & mask) << ASHIFT) + ABASE, x);
930 >            }
931 >        }
932 >    }
933 >
934 >    // Blocking support
935 >
936 >    /**
937 >     * Tries to increment blockedCount, decrement active count
938 >     * (sometimes implicitly) and possibly release or create a
939 >     * compensating worker in preparation for blocking. Fails
940 >     * on contention or termination.
941 >     *
942 >     * @return true if the caller can block, else should recheck and retry
943 >     */
944 >    private boolean tryPreBlock() {
945 >        int b = blockedCount;
946 >        if (UNSAFE.compareAndSwapInt(this, blockedCountOffset, b, b + 1)) {
947 >            int pc = parallelism;
948 >            do {
949 >                ForkJoinWorkerThread[] ws; ForkJoinWorkerThread w;
950 >                int e, ac, tc, i;
951 >                long c = ctl;
952 >                int u = (int)(c >>> 32);
953 >                if ((e = (int)c) < 0) {
954 >                                                 // skip -- terminating
955 >                }
956 >                else if ((ac = (u >> UAC_SHIFT)) <= 0 && e != 0 &&
957 >                         (ws = workers) != null &&
958 >                         (i = ~e & SMASK) < ws.length &&
959 >                         (w = ws[i]) != null) {
960 >                    long nc = ((long)(w.nextWait & E_MASK) |
961 >                               (c & (AC_MASK|TC_MASK)));
962 >                    if (w.eventCount == e &&
963 >                        UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc)) {
964 >                        w.eventCount = (e + EC_UNIT) & E_MASK;
965 >                        if (w.parked)
966 >                            UNSAFE.unpark(w);
967 >                        return true;             // release an idle worker
968 >                    }
969 >                }
970 >                else if ((tc = (short)(u >>> UTC_SHIFT)) >= 0 && ac + pc > 1) {
971 >                    long nc = ((c - AC_UNIT) & AC_MASK) | (c & ~AC_MASK);
972 >                    if (UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc))
973 >                        return true;             // no compensation needed
974 >                }
975 >                else if (tc + pc < MAX_ID) {
976 >                    long nc = ((c + TC_UNIT) & TC_MASK) | (c & ~TC_MASK);
977 >                    if (UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc)) {
978 >                        addWorker();
979 >                        return true;            // create a replacement
980 >                    }
981 >                }
982 >                // try to back out on any failure and let caller retry
983 >            } while (!UNSAFE.compareAndSwapInt(this, blockedCountOffset,
984 >                                               b = blockedCount, b - 1));
985 >        }
986 >        return false;
987      }
988  
989      /**
990 <     * Creates a {@code ForkJoinPool} with parallelism equal to {@link
399 <     * java.lang.Runtime#availableProcessors}, and using the given
400 <     * thread factory.
401 <     *
402 <     * @param factory the factory for creating new threads
403 <     * @throws NullPointerException if factory is null
404 <     * @throws SecurityException if a security manager exists and
405 <     *         the caller is not permitted to modify threads
406 <     *         because it does not hold {@link
407 <     *         java.lang.RuntimePermission}{@code ("modifyThread")}
990 >     * Decrements blockedCount and increments active count.
991       */
992 <    public ForkJoinPool(ForkJoinWorkerThreadFactory factory) {
993 <        this(Runtime.getRuntime().availableProcessors(), factory);
992 >    private void postBlock() {
993 >        long c;
994 >        do {} while (!UNSAFE.compareAndSwapLong(this, ctlOffset,  // no mask
995 >                                                c = ctl, c + AC_UNIT));
996 >        int b;
997 >        do {} while (!UNSAFE.compareAndSwapInt(this, blockedCountOffset,
998 >                                               b = blockedCount, b - 1));
999      }
1000  
1001      /**
1002 <     * Creates a {@code ForkJoinPool} with the given parallelism and
1003 <     * thread factory.
1002 >     * Possibly blocks waiting for the given task to complete, or
1003 >     * cancels the task if terminating.  Fails to wait if contended.
1004       *
1005 <     * @param parallelism the parallelism level
418 <     * @param factory the factory for creating new threads
419 <     * @throws IllegalArgumentException if parallelism less than or
420 <     * equal to zero, or greater than implementation limit
421 <     * @throws NullPointerException if factory is null
422 <     * @throws SecurityException if a security manager exists and
423 <     *         the caller is not permitted to modify threads
424 <     *         because it does not hold {@link
425 <     *         java.lang.RuntimePermission}{@code ("modifyThread")}
1005 >     * @param joinMe the task
1006       */
1007 <    public ForkJoinPool(int parallelism, ForkJoinWorkerThreadFactory factory) {
1008 <        if (parallelism <= 0 || parallelism > MAX_THREADS)
1009 <            throw new IllegalArgumentException();
1010 <        if (factory == null)
1011 <            throw new NullPointerException();
1012 <        checkPermission();
1013 <        this.factory = factory;
1014 <        this.parallelism = parallelism;
1015 <        this.maxPoolSize = MAX_THREADS;
1016 <        this.maintainsParallelism = true;
437 <        this.poolNumber = poolNumberGenerator.incrementAndGet();
438 <        this.workerLock = new ReentrantLock();
439 <        this.termination = workerLock.newCondition();
440 <        this.stealCount = new AtomicLong();
441 <        this.submissionQueue = new LinkedTransferQueue<ForkJoinTask<?>>();
442 <        // worker array and workers are lazily constructed
1007 >    final void tryAwaitJoin(ForkJoinTask<?> joinMe) {
1008 >        Thread.interrupted(); // clear interrupts before checking termination
1009 >        if (joinMe.status >= 0) {
1010 >            if (tryPreBlock()) {
1011 >                joinMe.tryAwaitDone(0L);
1012 >                postBlock();
1013 >            }
1014 >            else if ((ctl & STOP_BIT) != 0L)
1015 >                joinMe.cancelIgnoringExceptions();
1016 >        }
1017      }
1018  
1019      /**
1020 <     * Creates a new worker thread using factory.
1020 >     * Possibly blocks the given worker waiting for joinMe to
1021 >     * complete or timeout.
1022       *
1023 <     * @param index the index to assign worker
1024 <     * @return new worker, or null if factory failed
1023 >     * @param joinMe the task
1024 >     * @param millis the wait time for underlying Object.wait
1025       */
1026 <    private ForkJoinWorkerThread createWorker(int index) {
1027 <        Thread.UncaughtExceptionHandler h = ueh;
1028 <        ForkJoinWorkerThread w = factory.newThread(this);
1029 <        if (w != null) {
1030 <            w.poolIndex = index;
1031 <            w.setDaemon(true);
1032 <            w.setAsyncMode(locallyFifo);
1033 <            w.setName("ForkJoinPool-" + poolNumber + "-worker-" + index);
1034 <            if (h != null)
1035 <                w.setUncaughtExceptionHandler(h);
1026 >    final void timedAwaitJoin(ForkJoinTask<?> joinMe, long nanos) {
1027 >        while (joinMe.status >= 0) {
1028 >            Thread.interrupted();
1029 >            if ((ctl & STOP_BIT) != 0L) {
1030 >                joinMe.cancelIgnoringExceptions();
1031 >                break;
1032 >            }
1033 >            if (tryPreBlock()) {
1034 >                long last = System.nanoTime();
1035 >                while (joinMe.status >= 0) {
1036 >                    long millis = TimeUnit.NANOSECONDS.toMillis(nanos);
1037 >                    if (millis <= 0)
1038 >                        break;
1039 >                    joinMe.tryAwaitDone(millis);
1040 >                    if (joinMe.status < 0)
1041 >                        break;
1042 >                    if ((ctl & STOP_BIT) != 0L) {
1043 >                        joinMe.cancelIgnoringExceptions();
1044 >                        break;
1045 >                    }
1046 >                    long now = System.nanoTime();
1047 >                    nanos -= now - last;
1048 >                    last = now;
1049 >                }
1050 >                postBlock();
1051 >                break;
1052 >            }
1053          }
462        return w;
1054      }
1055  
1056      /**
1057 <     * Returns a good size for worker array given pool size.
467 <     * Currently requires size to be a power of two.
1057 >     * If necessary, compensates for blocker, and blocks.
1058       */
1059 <    private static int arraySizeFor(int poolSize) {
1060 <        if (poolSize <= 1)
1061 <            return 1;
1062 <        // See Hackers Delight, sec 3.2
1063 <        int c = poolSize >= MAX_THREADS ? MAX_THREADS : (poolSize - 1);
1064 <        c |= c >>>  1;
1065 <        c |= c >>>  2;
1066 <        c |= c >>>  4;
1067 <        c |= c >>>  8;
1068 <        c |= c >>> 16;
1069 <        return c + 1;
1059 >    private void awaitBlocker(ManagedBlocker blocker)
1060 >        throws InterruptedException {
1061 >        while (!blocker.isReleasable()) {
1062 >            if (tryPreBlock()) {
1063 >                try {
1064 >                    do {} while (!blocker.isReleasable() && !blocker.block());
1065 >                } finally {
1066 >                    postBlock();
1067 >                }
1068 >                break;
1069 >            }
1070 >        }
1071      }
1072  
1073 +    // Creating, registering and deregistring workers
1074 +
1075      /**
1076 <     * Creates or resizes array if necessary to hold newLength.
1077 <     * Call only under exclusion.
485 <     *
486 <     * @return the array
1076 >     * Tries to create and start a worker; minimally rolls back counts
1077 >     * on failure.
1078       */
1079 <    private ForkJoinWorkerThread[] ensureWorkerArrayCapacity(int newLength) {
1080 <        ForkJoinWorkerThread[] ws = workers;
1081 <        if (ws == null)
1082 <            return workers = new ForkJoinWorkerThread[arraySizeFor(newLength)];
1083 <        else if (newLength > ws.length)
1084 <            return workers = Arrays.copyOf(ws, arraySizeFor(newLength));
1079 >    private void addWorker() {
1080 >        Throwable ex = null;
1081 >        ForkJoinWorkerThread t = null;
1082 >        try {
1083 >            t = factory.newThread(this);
1084 >        } catch (Throwable e) {
1085 >            ex = e;
1086 >        }
1087 >        if (t == null) {  // null or exceptional factory return
1088 >            long c;       // adjust counts
1089 >            do {} while (!UNSAFE.compareAndSwapLong
1090 >                         (this, ctlOffset, c = ctl,
1091 >                          (((c - AC_UNIT) & AC_MASK) |
1092 >                           ((c - TC_UNIT) & TC_MASK) |
1093 >                           (c & ~(AC_MASK|TC_MASK)))));
1094 >            // Propagate exception if originating from an external caller
1095 >            if (!tryTerminate(false) && ex != null &&
1096 >                !(Thread.currentThread() instanceof ForkJoinWorkerThread))
1097 >                UNSAFE.throwException(ex);
1098 >        }
1099          else
1100 <            return ws;
1100 >            t.start();
1101      }
1102  
1103      /**
1104 <     * Tries to shrink workers into smaller array after one or more terminate.
1104 >     * Callback from ForkJoinWorkerThread constructor to assign a
1105 >     * public name
1106       */
1107 <    private void tryShrinkWorkerArray() {
1108 <        ForkJoinWorkerThread[] ws = workers;
1109 <        if (ws != null) {
1110 <            int len = ws.length;
1111 <            int last = len - 1;
506 <            while (last >= 0 && ws[last] == null)
507 <                --last;
508 <            int newLength = arraySizeFor(last+1);
509 <            if (newLength < len)
510 <                workers = Arrays.copyOf(ws, newLength);
1107 >    final String nextWorkerName() {
1108 >        for (int n;;) {
1109 >            if (UNSAFE.compareAndSwapInt(this, nextWorkerNumberOffset,
1110 >                                         n = nextWorkerNumber, ++n))
1111 >                return workerNamePrefix + n;
1112          }
1113      }
1114  
1115      /**
1116 <     * Initializes workers if necessary.
1116 >     * Callback from ForkJoinWorkerThread constructor to
1117 >     * determine its poolIndex and record in workers array.
1118 >     *
1119 >     * @param w the worker
1120 >     * @return the worker's pool index
1121       */
1122 <    final void ensureWorkerInitialization() {
1123 <        ForkJoinWorkerThread[] ws = workers;
1124 <        if (ws == null) {
1125 <            final ReentrantLock lock = this.workerLock;
1126 <            lock.lock();
1127 <            try {
1128 <                ws = workers;
1129 <                if (ws == null) {
1130 <                    int ps = parallelism;
1131 <                    ws = ensureWorkerArrayCapacity(ps);
1132 <                    for (int i = 0; i < ps; ++i) {
1133 <                        ForkJoinWorkerThread w = createWorker(i);
1134 <                        if (w != null) {
1135 <                            ws[i] = w;
1136 <                            w.start();
1137 <                            updateWorkerCount(1);
1122 >    final int registerWorker(ForkJoinWorkerThread w) {
1123 >        /*
1124 >         * In the typical case, a new worker acquires the lock, uses
1125 >         * next available index and returns quickly.  Since we should
1126 >         * not block callers (ultimately from signalWork or
1127 >         * tryPreBlock) waiting for the lock needed to do this, we
1128 >         * instead help release other workers while waiting for the
1129 >         * lock.
1130 >         */
1131 >        for (int g;;) {
1132 >            ForkJoinWorkerThread[] ws;
1133 >            if (((g = scanGuard) & SG_UNIT) == 0 &&
1134 >                UNSAFE.compareAndSwapInt(this, scanGuardOffset,
1135 >                                         g, g | SG_UNIT)) {
1136 >                int k = nextWorkerIndex;
1137 >                try {
1138 >                    if ((ws = workers) != null) { // ignore on shutdown
1139 >                        int n = ws.length;
1140 >                        if (k < 0 || k >= n || ws[k] != null) {
1141 >                            for (k = 0; k < n && ws[k] != null; ++k)
1142 >                                ;
1143 >                            if (k == n)
1144 >                                ws = workers = Arrays.copyOf(ws, n << 1);
1145                          }
1146 +                        ws[k] = w;
1147 +                        nextWorkerIndex = k + 1;
1148 +                        int m = g & SMASK;
1149 +                        g = (k > m) ? ((m << 1) + 1) & SMASK : g + (SG_UNIT<<1);
1150 +                    }
1151 +                } finally {
1152 +                    scanGuard = g;
1153 +                }
1154 +                return k;
1155 +            }
1156 +            else if ((ws = workers) != null) { // help release others
1157 +                for (ForkJoinWorkerThread u : ws) {
1158 +                    if (u != null && u.queueBase != u.queueTop) {
1159 +                        if (tryReleaseWaiter())
1160 +                            break;
1161                      }
1162                  }
1163 +            }
1164 +        }
1165 +    }
1166 +
1167 +    /**
1168 +     * Final callback from terminating worker.  Removes record of
1169 +     * worker from array, and adjusts counts. If pool is shutting
1170 +     * down, tries to complete termination.
1171 +     *
1172 +     * @param w the worker
1173 +     */
1174 +    final void deregisterWorker(ForkJoinWorkerThread w, Throwable ex) {
1175 +        int idx = w.poolIndex;
1176 +        int sc = w.stealCount;
1177 +        int steps = 0;
1178 +        // Remove from array, adjust worker counts and collect steal count.
1179 +        // We can intermix failed removes or adjusts with steal updates
1180 +        do {
1181 +            long s, c;
1182 +            int g;
1183 +            if (steps == 0 && ((g = scanGuard) & SG_UNIT) == 0 &&
1184 +                UNSAFE.compareAndSwapInt(this, scanGuardOffset,
1185 +                                         g, g |= SG_UNIT)) {
1186 +                ForkJoinWorkerThread[] ws = workers;
1187 +                if (ws != null && idx >= 0 &&
1188 +                    idx < ws.length && ws[idx] == w)
1189 +                    ws[idx] = null;    // verify
1190 +                nextWorkerIndex = idx;
1191 +                scanGuard = g + SG_UNIT;
1192 +                steps = 1;
1193 +            }
1194 +            if (steps == 1 &&
1195 +                UNSAFE.compareAndSwapLong(this, ctlOffset, c = ctl,
1196 +                                          (((c - AC_UNIT) & AC_MASK) |
1197 +                                           ((c - TC_UNIT) & TC_MASK) |
1198 +                                           (c & ~(AC_MASK|TC_MASK)))))
1199 +                steps = 2;
1200 +            if (sc != 0 &&
1201 +                UNSAFE.compareAndSwapLong(this, stealCountOffset,
1202 +                                          s = stealCount, s + sc))
1203 +                sc = 0;
1204 +        } while (steps != 2 || sc != 0);
1205 +        if (!tryTerminate(false)) {
1206 +            if (ex != null)   // possibly replace if died abnormally
1207 +                signalWork();
1208 +            else
1209 +                tryReleaseWaiter();
1210 +        }
1211 +    }
1212 +
1213 +    // Shutdown and termination
1214 +
1215 +    /**
1216 +     * Possibly initiates and/or completes termination.
1217 +     *
1218 +     * @param now if true, unconditionally terminate, else only
1219 +     * if shutdown and empty queue and no active workers
1220 +     * @return true if now terminating or terminated
1221 +     */
1222 +    private boolean tryTerminate(boolean now) {
1223 +        long c;
1224 +        while (((c = ctl) & STOP_BIT) == 0) {
1225 +            if (!now) {
1226 +                if ((int)(c >> AC_SHIFT) != -parallelism)
1227 +                    return false;
1228 +                if (!shutdown || blockedCount != 0 || quiescerCount != 0 ||
1229 +                    queueBase != queueTop) {
1230 +                    if (ctl == c) // staleness check
1231 +                        return false;
1232 +                    continue;
1233 +                }
1234 +            }
1235 +            if (UNSAFE.compareAndSwapLong(this, ctlOffset, c, c | STOP_BIT))
1236 +                startTerminating();
1237 +        }
1238 +        if ((short)(c >>> TC_SHIFT) == -parallelism) { // signal when 0 workers
1239 +            final ReentrantLock lock = this.submissionLock;
1240 +            lock.lock();
1241 +            try {
1242 +                termination.signalAll();
1243              } finally {
1244                  lock.unlock();
1245              }
1246          }
1247 +        return true;
1248      }
1249  
1250      /**
1251 <     * Worker creation and startup for threads added via setParallelism.
1252 <     */
1253 <    private void createAndStartAddedWorkers() {
1254 <        resumeAllSpares();  // Allow spares to convert to nonspare
1255 <        int ps = parallelism;
1256 <        ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(ps);
1257 <        int len = ws.length;
1258 <        // Sweep through slots, to keep lowest indices most populated
1259 <        int k = 0;
1260 <        while (k < len) {
1261 <            if (ws[k] != null) {
1262 <                ++k;
1263 <                continue;
1251 >     * Runs up to three passes through workers: (0) Setting
1252 >     * termination status for each worker, followed by wakeups up to
1253 >     * queued workers; (1) helping cancel tasks; (2) interrupting
1254 >     * lagging threads (likely in external tasks, but possibly also
1255 >     * blocked in joins).  Each pass repeats previous steps because of
1256 >     * potential lagging thread creation.
1257 >     */
1258 >    private void startTerminating() {
1259 >        cancelSubmissions();
1260 >        for (int pass = 0; pass < 3; ++pass) {
1261 >            ForkJoinWorkerThread[] ws = workers;
1262 >            if (ws != null) {
1263 >                for (ForkJoinWorkerThread w : ws) {
1264 >                    if (w != null) {
1265 >                        w.terminate = true;
1266 >                        if (pass > 0) {
1267 >                            w.cancelTasks();
1268 >                            if (pass > 1 && !w.isInterrupted()) {
1269 >                                try {
1270 >                                    w.interrupt();
1271 >                                } catch (SecurityException ignore) {
1272 >                                }
1273 >                            }
1274 >                        }
1275 >                    }
1276 >                }
1277 >                terminateWaiters();
1278              }
1279 <            int s = workerCounts;
1280 <            int tc = totalCountOf(s);
1281 <            int rc = runningCountOf(s);
1282 <            if (rc >= ps || tc >= ps)
1283 <                break;
1284 <            if (casWorkerCounts (s, workerCountsFor(tc+1, rc+1))) {
1285 <                ForkJoinWorkerThread w = createWorker(k);
1286 <                if (w != null) {
1287 <                    ws[k++] = w;
1288 <                    w.start();
1279 >        }
1280 >    }
1281 >
1282 >    /**
1283 >     * Polls and cancels all submissions. Called only during termination.
1284 >     */
1285 >    private void cancelSubmissions() {
1286 >        while (queueBase != queueTop) {
1287 >            ForkJoinTask<?> task = pollSubmission();
1288 >            if (task != null) {
1289 >                try {
1290 >                    task.cancel(false);
1291 >                } catch (Throwable ignore) {
1292                  }
1293 <                else {
1294 <                    updateWorkerCount(-1); // back out on failed creation
1295 <                    break;
1293 >            }
1294 >        }
1295 >    }
1296 >
1297 >    /**
1298 >     * Tries to set the termination status of waiting workers, and
1299 >     * then wakes them up (after which they will terminate).
1300 >     */
1301 >    private void terminateWaiters() {
1302 >        ForkJoinWorkerThread[] ws = workers;
1303 >        if (ws != null) {
1304 >            ForkJoinWorkerThread w; long c; int i, e;
1305 >            int n = ws.length;
1306 >            while ((i = ~(e = (int)(c = ctl)) & SMASK) < n &&
1307 >                   (w = ws[i]) != null && w.eventCount == (e & E_MASK)) {
1308 >                if (UNSAFE.compareAndSwapLong(this, ctlOffset, c,
1309 >                                              (long)(w.nextWait & E_MASK) |
1310 >                                              ((c + AC_UNIT) & AC_MASK) |
1311 >                                              (c & (TC_MASK|STOP_BIT)))) {
1312 >                    w.terminate = true;
1313 >                    w.eventCount = e + EC_UNIT;
1314 >                    if (w.parked)
1315 >                        UNSAFE.unpark(w);
1316                  }
1317              }
1318          }
1319      }
1320  
1321 <    // Execution methods
1321 >    // misc ForkJoinWorkerThread support
1322  
1323      /**
1324 <     * Common code for execute, invoke and submit
1324 >     * Increments or decrements quiescerCount. Needed only to prevent
1325 >     * triggering shutdown if a worker is transiently inactive while
1326 >     * checking quiescence.
1327 >     *
1328 >     * @param delta 1 for increment, -1 for decrement
1329       */
1330 <    private <T> void doSubmit(ForkJoinTask<T> task) {
1331 <        if (task == null)
1330 >    final void addQuiescerCount(int delta) {
1331 >        int c;
1332 >        do {} while (!UNSAFE.compareAndSwapInt(this, quiescerCountOffset,
1333 >                                               c = quiescerCount, c + delta));
1334 >    }
1335 >
1336 >    /**
1337 >     * Directly increments or decrements active count without queuing.
1338 >     * This method is used to transiently assert inactivation while
1339 >     * checking quiescence.
1340 >     *
1341 >     * @param delta 1 for increment, -1 for decrement
1342 >     */
1343 >    final void addActiveCount(int delta) {
1344 >        long d = (long)delta << AC_SHIFT;
1345 >        long c;
1346 >        do {} while (!UNSAFE.compareAndSwapLong(this, ctlOffset,
1347 >                                                c = ctl, c + d));
1348 >    }
1349 >
1350 >    /**
1351 >     * Returns the approximate (non-atomic) number of idle threads per
1352 >     * active thread.
1353 >     */
1354 >    final int idlePerActive() {
1355 >        // Approximate at powers of two for small values, saturate past 4
1356 >        int p = parallelism;
1357 >        int a = p + (int)(ctl >> AC_SHIFT);
1358 >        return (a > (p >>>= 1) ? 0 :
1359 >                a > (p >>>= 1) ? 1 :
1360 >                a > (p >>>= 1) ? 2 :
1361 >                a > (p >>>= 1) ? 4 :
1362 >                8);
1363 >    }
1364 >
1365 >    // Exported methods
1366 >
1367 >    // Constructors
1368 >
1369 >    /**
1370 >     * Creates a {@code ForkJoinPool} with parallelism equal to {@link
1371 >     * java.lang.Runtime#availableProcessors}, using the {@linkplain
1372 >     * #defaultForkJoinWorkerThreadFactory default thread factory},
1373 >     * no UncaughtExceptionHandler, and non-async LIFO processing mode.
1374 >     *
1375 >     * @throws SecurityException if a security manager exists and
1376 >     *         the caller is not permitted to modify threads
1377 >     *         because it does not hold {@link
1378 >     *         java.lang.RuntimePermission}{@code ("modifyThread")}
1379 >     */
1380 >    public ForkJoinPool() {
1381 >        this(Runtime.getRuntime().availableProcessors(),
1382 >             defaultForkJoinWorkerThreadFactory, null, false);
1383 >    }
1384 >
1385 >    /**
1386 >     * Creates a {@code ForkJoinPool} with the indicated parallelism
1387 >     * level, the {@linkplain
1388 >     * #defaultForkJoinWorkerThreadFactory default thread factory},
1389 >     * no UncaughtExceptionHandler, and non-async LIFO processing mode.
1390 >     *
1391 >     * @param parallelism the parallelism level
1392 >     * @throws IllegalArgumentException if parallelism less than or
1393 >     *         equal to zero, or greater than implementation limit
1394 >     * @throws SecurityException if a security manager exists and
1395 >     *         the caller is not permitted to modify threads
1396 >     *         because it does not hold {@link
1397 >     *         java.lang.RuntimePermission}{@code ("modifyThread")}
1398 >     */
1399 >    public ForkJoinPool(int parallelism) {
1400 >        this(parallelism, defaultForkJoinWorkerThreadFactory, null, false);
1401 >    }
1402 >
1403 >    /**
1404 >     * Creates a {@code ForkJoinPool} with the given parameters.
1405 >     *
1406 >     * @param parallelism the parallelism level. For default value,
1407 >     * use {@link java.lang.Runtime#availableProcessors}.
1408 >     * @param factory the factory for creating new threads. For default value,
1409 >     * use {@link #defaultForkJoinWorkerThreadFactory}.
1410 >     * @param handler the handler for internal worker threads that
1411 >     * terminate due to unrecoverable errors encountered while executing
1412 >     * tasks. For default value, use {@code null}.
1413 >     * @param asyncMode if true,
1414 >     * establishes local first-in-first-out scheduling mode for forked
1415 >     * tasks that are never joined. This mode may be more appropriate
1416 >     * than default locally stack-based mode in applications in which
1417 >     * worker threads only process event-style asynchronous tasks.
1418 >     * For default value, use {@code false}.
1419 >     * @throws IllegalArgumentException if parallelism less than or
1420 >     *         equal to zero, or greater than implementation limit
1421 >     * @throws NullPointerException if the factory is null
1422 >     * @throws SecurityException if a security manager exists and
1423 >     *         the caller is not permitted to modify threads
1424 >     *         because it does not hold {@link
1425 >     *         java.lang.RuntimePermission}{@code ("modifyThread")}
1426 >     */
1427 >    public ForkJoinPool(int parallelism,
1428 >                        ForkJoinWorkerThreadFactory factory,
1429 >                        Thread.UncaughtExceptionHandler handler,
1430 >                        boolean asyncMode) {
1431 >        checkPermission();
1432 >        if (factory == null)
1433              throw new NullPointerException();
1434 <        if (isShutdown())
1435 <            throw new RejectedExecutionException();
1436 <        if (workers == null)
1437 <            ensureWorkerInitialization();
1438 <        submissionQueue.offer(task);
1439 <        signalIdleWorkers();
1434 >        if (parallelism <= 0 || parallelism > MAX_ID)
1435 >            throw new IllegalArgumentException();
1436 >        this.parallelism = parallelism;
1437 >        this.factory = factory;
1438 >        this.ueh = handler;
1439 >        this.locallyFifo = asyncMode;
1440 >        long np = (long)(-parallelism); // offset ctl counts
1441 >        this.ctl = ((np << AC_SHIFT) & AC_MASK) | ((np << TC_SHIFT) & TC_MASK);
1442 >        this.submissionQueue = new ForkJoinTask<?>[INITIAL_QUEUE_CAPACITY];
1443 >        // initialize workers array with room for 2*parallelism if possible
1444 >        int n = parallelism << 1;
1445 >        if (n >= MAX_ID)
1446 >            n = MAX_ID;
1447 >        else { // See Hackers Delight, sec 3.2, where n < (1 << 16)
1448 >            n |= n >>> 1; n |= n >>> 2; n |= n >>> 4; n |= n >>> 8;
1449 >        }
1450 >        workers = new ForkJoinWorkerThread[n + 1];
1451 >        this.submissionLock = new ReentrantLock();
1452 >        this.termination = submissionLock.newCondition();
1453 >        StringBuilder sb = new StringBuilder("ForkJoinPool-");
1454 >        sb.append(poolNumberGenerator.incrementAndGet());
1455 >        sb.append("-worker-");
1456 >        this.workerNamePrefix = sb.toString();
1457      }
1458  
1459 +    // Execution methods
1460 +
1461      /**
1462       * Performs the given task, returning its result upon completion.
1463 +     * If the computation encounters an unchecked Exception or Error,
1464 +     * it is rethrown as the outcome of this invocation.  Rethrown
1465 +     * exceptions behave in the same way as regular exceptions, but,
1466 +     * when possible, contain stack traces (as displayed for example
1467 +     * using {@code ex.printStackTrace()}) of both the current thread
1468 +     * as well as the thread actually encountering the exception;
1469 +     * minimally only the latter.
1470       *
1471       * @param task the task
1472       * @return the task's result
1473 <     * @throws NullPointerException if task is null
1474 <     * @throws RejectedExecutionException if pool is shut down
1473 >     * @throws NullPointerException if the task is null
1474 >     * @throws RejectedExecutionException if the task cannot be
1475 >     *         scheduled for execution
1476       */
1477      public <T> T invoke(ForkJoinTask<T> task) {
1478 <        doSubmit(task);
1479 <        return task.join();
1478 >        Thread t = Thread.currentThread();
1479 >        if (task == null)
1480 >            throw new NullPointerException();
1481 >        if (shutdown)
1482 >            throw new RejectedExecutionException();
1483 >        if ((t instanceof ForkJoinWorkerThread) &&
1484 >            ((ForkJoinWorkerThread)t).pool == this)
1485 >            return task.invoke();  // bypass submit if in same pool
1486 >        else {
1487 >            addSubmission(task);
1488 >            return task.join();
1489 >        }
1490 >    }
1491 >
1492 >    /**
1493 >     * Unless terminating, forks task if within an ongoing FJ
1494 >     * computation in the current pool, else submits as external task.
1495 >     */
1496 >    private <T> void forkOrSubmit(ForkJoinTask<T> task) {
1497 >        ForkJoinWorkerThread w;
1498 >        Thread t = Thread.currentThread();
1499 >        if (shutdown)
1500 >            throw new RejectedExecutionException();
1501 >        if ((t instanceof ForkJoinWorkerThread) &&
1502 >            (w = (ForkJoinWorkerThread)t).pool == this)
1503 >            w.pushTask(task);
1504 >        else
1505 >            addSubmission(task);
1506      }
1507  
1508      /**
1509       * Arranges for (asynchronous) execution of the given task.
1510       *
1511       * @param task the task
1512 <     * @throws NullPointerException if task is null
1513 <     * @throws RejectedExecutionException if pool is shut down
1512 >     * @throws NullPointerException if the task is null
1513 >     * @throws RejectedExecutionException if the task cannot be
1514 >     *         scheduled for execution
1515       */
1516      public void execute(ForkJoinTask<?> task) {
1517 <        doSubmit(task);
1517 >        if (task == null)
1518 >            throw new NullPointerException();
1519 >        forkOrSubmit(task);
1520      }
1521  
1522      // AbstractExecutorService methods
1523  
1524 +    /**
1525 +     * @throws NullPointerException if the task is null
1526 +     * @throws RejectedExecutionException if the task cannot be
1527 +     *         scheduled for execution
1528 +     */
1529      public void execute(Runnable task) {
1530 +        if (task == null)
1531 +            throw new NullPointerException();
1532          ForkJoinTask<?> job;
1533          if (task instanceof ForkJoinTask<?>) // avoid re-wrap
1534              job = (ForkJoinTask<?>) task;
1535          else
1536              job = ForkJoinTask.adapt(task, null);
1537 <        doSubmit(job);
1537 >        forkOrSubmit(job);
1538 >    }
1539 >
1540 >    /**
1541 >     * Submits a ForkJoinTask for execution.
1542 >     *
1543 >     * @param task the task to submit
1544 >     * @return the task
1545 >     * @throws NullPointerException if the task is null
1546 >     * @throws RejectedExecutionException if the task cannot be
1547 >     *         scheduled for execution
1548 >     */
1549 >    public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) {
1550 >        if (task == null)
1551 >            throw new NullPointerException();
1552 >        forkOrSubmit(task);
1553 >        return task;
1554      }
1555  
1556 +    /**
1557 +     * @throws NullPointerException if the task is null
1558 +     * @throws RejectedExecutionException if the task cannot be
1559 +     *         scheduled for execution
1560 +     */
1561      public <T> ForkJoinTask<T> submit(Callable<T> task) {
1562 +        if (task == null)
1563 +            throw new NullPointerException();
1564          ForkJoinTask<T> job = ForkJoinTask.adapt(task);
1565 <        doSubmit(job);
1565 >        forkOrSubmit(job);
1566          return job;
1567      }
1568  
1569 +    /**
1570 +     * @throws NullPointerException if the task is null
1571 +     * @throws RejectedExecutionException if the task cannot be
1572 +     *         scheduled for execution
1573 +     */
1574      public <T> ForkJoinTask<T> submit(Runnable task, T result) {
1575 +        if (task == null)
1576 +            throw new NullPointerException();
1577          ForkJoinTask<T> job = ForkJoinTask.adapt(task, result);
1578 <        doSubmit(job);
1578 >        forkOrSubmit(job);
1579          return job;
1580      }
1581  
1582 +    /**
1583 +     * @throws NullPointerException if the task is null
1584 +     * @throws RejectedExecutionException if the task cannot be
1585 +     *         scheduled for execution
1586 +     */
1587      public ForkJoinTask<?> submit(Runnable task) {
1588 +        if (task == null)
1589 +            throw new NullPointerException();
1590          ForkJoinTask<?> job;
1591          if (task instanceof ForkJoinTask<?>) // avoid re-wrap
1592              job = (ForkJoinTask<?>) task;
1593          else
1594              job = ForkJoinTask.adapt(task, null);
1595 <        doSubmit(job);
1595 >        forkOrSubmit(job);
1596          return job;
1597      }
1598  
1599      /**
1600 <     * Submits a ForkJoinTask for execution.
1601 <     *
652 <     * @param task the task to submit
653 <     * @return the task
654 <     * @throws RejectedExecutionException if the task cannot be
655 <     *         scheduled for execution
656 <     * @throws NullPointerException if the task is null
1600 >     * @throws NullPointerException       {@inheritDoc}
1601 >     * @throws RejectedExecutionException {@inheritDoc}
1602       */
658    public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) {
659        doSubmit(task);
660        return task;
661    }
662
663
1603      public <T> List<Future<T>> invokeAll(Collection<? extends Callable<T>> tasks) {
1604          ArrayList<ForkJoinTask<T>> forkJoinTasks =
1605              new ArrayList<ForkJoinTask<T>>(tasks.size());
# Line 669 | Line 1608 | public class ForkJoinPool extends Abstra
1608          invoke(new InvokeAll<T>(forkJoinTasks));
1609  
1610          @SuppressWarnings({"unchecked", "rawtypes"})
1611 <        List<Future<T>> futures = (List<Future<T>>) (List) forkJoinTasks;
1611 >            List<Future<T>> futures = (List<Future<T>>) (List) forkJoinTasks;
1612          return futures;
1613      }
1614  
# Line 683 | Line 1622 | public class ForkJoinPool extends Abstra
1622          private static final long serialVersionUID = -7914297376763021607L;
1623      }
1624  
686    // Configuration and status settings and queries
687
1625      /**
1626       * Returns the factory used for constructing new workers.
1627       *
# Line 701 | Line 1638 | public class ForkJoinPool extends Abstra
1638       * @return the handler, or {@code null} if none
1639       */
1640      public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() {
1641 <        Thread.UncaughtExceptionHandler h;
705 <        final ReentrantLock lock = this.workerLock;
706 <        lock.lock();
707 <        try {
708 <            h = ueh;
709 <        } finally {
710 <            lock.unlock();
711 <        }
712 <        return h;
713 <    }
714 <
715 <    /**
716 <     * Sets the handler for internal worker threads that terminate due
717 <     * to unrecoverable errors encountered while executing tasks.
718 <     * Unless set, the current default or ThreadGroup handler is used
719 <     * as handler.
720 <     *
721 <     * @param h the new handler
722 <     * @return the old handler, or {@code null} if none
723 <     * @throws SecurityException if a security manager exists and
724 <     *         the caller is not permitted to modify threads
725 <     *         because it does not hold {@link
726 <     *         java.lang.RuntimePermission}{@code ("modifyThread")}
727 <     */
728 <    public Thread.UncaughtExceptionHandler
729 <        setUncaughtExceptionHandler(Thread.UncaughtExceptionHandler h) {
730 <        checkPermission();
731 <        Thread.UncaughtExceptionHandler old = null;
732 <        final ReentrantLock lock = this.workerLock;
733 <        lock.lock();
734 <        try {
735 <            old = ueh;
736 <            ueh = h;
737 <            ForkJoinWorkerThread[] ws = workers;
738 <            if (ws != null) {
739 <                for (int i = 0; i < ws.length; ++i) {
740 <                    ForkJoinWorkerThread w = ws[i];
741 <                    if (w != null)
742 <                        w.setUncaughtExceptionHandler(h);
743 <                }
744 <            }
745 <        } finally {
746 <            lock.unlock();
747 <        }
748 <        return old;
749 <    }
750 <
751 <
752 <    /**
753 <     * Sets the target parallelism level of this pool.
754 <     *
755 <     * @param parallelism the target parallelism
756 <     * @throws IllegalArgumentException if parallelism less than or
757 <     * equal to zero or greater than maximum size bounds
758 <     * @throws SecurityException if a security manager exists and
759 <     *         the caller is not permitted to modify threads
760 <     *         because it does not hold {@link
761 <     *         java.lang.RuntimePermission}{@code ("modifyThread")}
762 <     */
763 <    public void setParallelism(int parallelism) {
764 <        checkPermission();
765 <        if (parallelism <= 0 || parallelism > maxPoolSize)
766 <            throw new IllegalArgumentException();
767 <        final ReentrantLock lock = this.workerLock;
768 <        lock.lock();
769 <        try {
770 <            if (isProcessingTasks()) {
771 <                int p = this.parallelism;
772 <                this.parallelism = parallelism;
773 <                if (parallelism > p)
774 <                    createAndStartAddedWorkers();
775 <                else
776 <                    trimSpares();
777 <            }
778 <        } finally {
779 <            lock.unlock();
780 <        }
781 <        signalIdleWorkers();
1641 >        return ueh;
1642      }
1643  
1644      /**
# Line 792 | Line 1652 | public class ForkJoinPool extends Abstra
1652  
1653      /**
1654       * Returns the number of worker threads that have started but not
1655 <     * yet terminated.  This result returned by this method may differ
1655 >     * yet terminated.  The result returned by this method may differ
1656       * from {@link #getParallelism} when threads are created to
1657       * maintain parallelism when others are cooperatively blocked.
1658       *
1659       * @return the number of worker threads
1660       */
1661      public int getPoolSize() {
1662 <        return totalCountOf(workerCounts);
803 <    }
804 <
805 <    /**
806 <     * Returns the maximum number of threads allowed to exist in the
807 <     * pool. Unless set using {@link #setMaximumPoolSize}, the
808 <     * maximum is an implementation-defined value designed only to
809 <     * prevent runaway growth.
810 <     *
811 <     * @return the maximum
812 <     */
813 <    public int getMaximumPoolSize() {
814 <        return maxPoolSize;
815 <    }
816 <
817 <    /**
818 <     * Sets the maximum number of threads allowed to exist in the
819 <     * pool. The given value should normally be greater than or equal
820 <     * to the {@link #getParallelism parallelism} level. Setting this
821 <     * value has no effect on current pool size. It controls
822 <     * construction of new threads.
823 <     *
824 <     * @throws IllegalArgumentException if negative or greater than
825 <     * internal implementation limit
826 <     */
827 <    public void setMaximumPoolSize(int newMax) {
828 <        if (newMax < 0 || newMax > MAX_THREADS)
829 <            throw new IllegalArgumentException();
830 <        maxPoolSize = newMax;
831 <    }
832 <
833 <
834 <    /**
835 <     * Returns {@code true} if this pool dynamically maintains its
836 <     * target parallelism level. If false, new threads are added only
837 <     * to avoid possible starvation.  This setting is by default true.
838 <     *
839 <     * @return {@code true} if maintains parallelism
840 <     */
841 <    public boolean getMaintainsParallelism() {
842 <        return maintainsParallelism;
843 <    }
844 <
845 <    /**
846 <     * Sets whether this pool dynamically maintains its target
847 <     * parallelism level. If false, new threads are added only to
848 <     * avoid possible starvation.
849 <     *
850 <     * @param enable {@code true} to maintain parallelism
851 <     */
852 <    public void setMaintainsParallelism(boolean enable) {
853 <        maintainsParallelism = enable;
854 <    }
855 <
856 <    /**
857 <     * Establishes local first-in-first-out scheduling mode for forked
858 <     * tasks that are never joined. This mode may be more appropriate
859 <     * than default locally stack-based mode in applications in which
860 <     * worker threads only process asynchronous tasks.  This method is
861 <     * designed to be invoked only when the pool is quiescent, and
862 <     * typically only before any tasks are submitted. The effects of
863 <     * invocations at other times may be unpredictable.
864 <     *
865 <     * @param async if {@code true}, use locally FIFO scheduling
866 <     * @return the previous mode
867 <     * @see #getAsyncMode
868 <     */
869 <    public boolean setAsyncMode(boolean async) {
870 <        boolean oldMode = locallyFifo;
871 <        locallyFifo = async;
872 <        ForkJoinWorkerThread[] ws = workers;
873 <        if (ws != null) {
874 <            for (int i = 0; i < ws.length; ++i) {
875 <                ForkJoinWorkerThread t = ws[i];
876 <                if (t != null)
877 <                    t.setAsyncMode(async);
878 <            }
879 <        }
880 <        return oldMode;
1662 >        return parallelism + (short)(ctl >>> TC_SHIFT);
1663      }
1664  
1665      /**
# Line 885 | Line 1667 | public class ForkJoinPool extends Abstra
1667       * scheduling mode for forked tasks that are never joined.
1668       *
1669       * @return {@code true} if this pool uses async mode
888     * @see #setAsyncMode
1670       */
1671      public boolean getAsyncMode() {
1672          return locallyFifo;
# Line 894 | Line 1675 | public class ForkJoinPool extends Abstra
1675      /**
1676       * Returns an estimate of the number of worker threads that are
1677       * not blocked waiting to join tasks or for other managed
1678 <     * synchronization.
1678 >     * synchronization. This method may overestimate the
1679 >     * number of running threads.
1680       *
1681       * @return the number of worker threads
1682       */
1683      public int getRunningThreadCount() {
1684 <        return runningCountOf(workerCounts);
1684 >        int r = parallelism + (int)(ctl >> AC_SHIFT);
1685 >        return (r <= 0) ? 0 : r; // suppress momentarily negative values
1686      }
1687  
1688      /**
# Line 910 | Line 1693 | public class ForkJoinPool extends Abstra
1693       * @return the number of active threads
1694       */
1695      public int getActiveThreadCount() {
1696 <        return activeCountOf(runControl);
1697 <    }
915 <
916 <    /**
917 <     * Returns an estimate of the number of threads that are currently
918 <     * idle waiting for tasks. This method may underestimate the
919 <     * number of idle threads.
920 <     *
921 <     * @return the number of idle threads
922 <     */
923 <    final int getIdleThreadCount() {
924 <        int c = runningCountOf(workerCounts) - activeCountOf(runControl);
925 <        return (c <= 0) ? 0 : c;
1696 >        int r = parallelism + (int)(ctl >> AC_SHIFT) + blockedCount;
1697 >        return (r <= 0) ? 0 : r; // suppress momentarily negative values
1698      }
1699  
1700      /**
# Line 937 | Line 1709 | public class ForkJoinPool extends Abstra
1709       * @return {@code true} if all threads are currently idle
1710       */
1711      public boolean isQuiescent() {
1712 <        return activeCountOf(runControl) == 0;
1712 >        return parallelism + (int)(ctl >> AC_SHIFT) + blockedCount == 0;
1713      }
1714  
1715      /**
# Line 952 | Line 1724 | public class ForkJoinPool extends Abstra
1724       * @return the number of steals
1725       */
1726      public long getStealCount() {
1727 <        return stealCount.get();
956 <    }
957 <
958 <    /**
959 <     * Accumulates steal count from a worker.
960 <     * Call only when worker known to be idle.
961 <     */
962 <    private void updateStealCount(ForkJoinWorkerThread w) {
963 <        int sc = w.getAndClearStealCount();
964 <        if (sc != 0)
965 <            stealCount.addAndGet(sc);
1727 >        return stealCount;
1728      }
1729  
1730      /**
# Line 977 | Line 1739 | public class ForkJoinPool extends Abstra
1739       */
1740      public long getQueuedTaskCount() {
1741          long count = 0;
1742 <        ForkJoinWorkerThread[] ws = workers;
1743 <        if (ws != null) {
1744 <            for (int i = 0; i < ws.length; ++i) {
1745 <                ForkJoinWorkerThread t = ws[i];
1746 <                if (t != null)
1747 <                    count += t.getQueueSize();
986 <            }
1742 >        ForkJoinWorkerThread[] ws;
1743 >        if ((short)(ctl >>> TC_SHIFT) > -parallelism &&
1744 >            (ws = workers) != null) {
1745 >            for (ForkJoinWorkerThread w : ws)
1746 >                if (w != null)
1747 >                    count -= w.queueBase - w.queueTop; // must read base first
1748          }
1749          return count;
1750      }
1751  
1752      /**
1753       * Returns an estimate of the number of tasks submitted to this
1754 <     * pool that have not yet begun executing.  This method takes time
1755 <     * proportional to the number of submissions.
1754 >     * pool that have not yet begun executing.  This method may take
1755 >     * time proportional to the number of submissions.
1756       *
1757       * @return the number of queued submissions
1758       */
1759      public int getQueuedSubmissionCount() {
1760 <        return submissionQueue.size();
1760 >        return -queueBase + queueTop;
1761      }
1762  
1763      /**
# Line 1006 | Line 1767 | public class ForkJoinPool extends Abstra
1767       * @return {@code true} if there are any queued submissions
1768       */
1769      public boolean hasQueuedSubmissions() {
1770 <        return !submissionQueue.isEmpty();
1770 >        return queueBase != queueTop;
1771      }
1772  
1773      /**
# Line 1017 | Line 1778 | public class ForkJoinPool extends Abstra
1778       * @return the next submission, or {@code null} if none
1779       */
1780      protected ForkJoinTask<?> pollSubmission() {
1781 <        return submissionQueue.poll();
1781 >        ForkJoinTask<?> t; ForkJoinTask<?>[] q; int b, i;
1782 >        while ((b = queueBase) != queueTop &&
1783 >               (q = submissionQueue) != null &&
1784 >               (i = (q.length - 1) & b) >= 0) {
1785 >            long u = (i << ASHIFT) + ABASE;
1786 >            if ((t = q[i]) != null &&
1787 >                queueBase == b &&
1788 >                UNSAFE.compareAndSwapObject(q, u, t, null)) {
1789 >                queueBase = b + 1;
1790 >                return t;
1791 >            }
1792 >        }
1793 >        return null;
1794      }
1795  
1796      /**
# Line 1038 | Line 1811 | public class ForkJoinPool extends Abstra
1811       * @return the number of elements transferred
1812       */
1813      protected int drainTasksTo(Collection<? super ForkJoinTask<?>> c) {
1814 <        int n = submissionQueue.drainTo(c);
1815 <        ForkJoinWorkerThread[] ws = workers;
1816 <        if (ws != null) {
1817 <            for (int i = 0; i < ws.length; ++i) {
1818 <                ForkJoinWorkerThread w = ws[i];
1819 <                if (w != null)
1047 <                    n += w.drainTasksTo(c);
1814 >        int count = 0;
1815 >        while (queueBase != queueTop) {
1816 >            ForkJoinTask<?> t = pollSubmission();
1817 >            if (t != null) {
1818 >                c.add(t);
1819 >                ++count;
1820              }
1821          }
1822 <        return n;
1822 >        ForkJoinWorkerThread[] ws;
1823 >        if ((short)(ctl >>> TC_SHIFT) > -parallelism &&
1824 >            (ws = workers) != null) {
1825 >            for (ForkJoinWorkerThread w : ws)
1826 >                if (w != null)
1827 >                    count += w.drainTasksTo(c);
1828 >        }
1829 >        return count;
1830      }
1831  
1832      /**
# Line 1058 | Line 1837 | public class ForkJoinPool extends Abstra
1837       * @return a string identifying this pool, as well as its state
1838       */
1839      public String toString() {
1061        int ps = parallelism;
1062        int wc = workerCounts;
1063        int rc = runControl;
1840          long st = getStealCount();
1841          long qt = getQueuedTaskCount();
1842          long qs = getQueuedSubmissionCount();
1843 +        int pc = parallelism;
1844 +        long c = ctl;
1845 +        int tc = pc + (short)(c >>> TC_SHIFT);
1846 +        int rc = pc + (int)(c >> AC_SHIFT);
1847 +        if (rc < 0) // ignore transient negative
1848 +            rc = 0;
1849 +        int ac = rc + blockedCount;
1850 +        String level;
1851 +        if ((c & STOP_BIT) != 0)
1852 +            level = (tc == 0) ? "Terminated" : "Terminating";
1853 +        else
1854 +            level = shutdown ? "Shutting down" : "Running";
1855          return super.toString() +
1856 <            "[" + runStateToString(runStateOf(rc)) +
1857 <            ", parallelism = " + ps +
1858 <            ", size = " + totalCountOf(wc) +
1859 <            ", active = " + activeCountOf(rc) +
1860 <            ", running = " + runningCountOf(wc) +
1856 >            "[" + level +
1857 >            ", parallelism = " + pc +
1858 >            ", size = " + tc +
1859 >            ", active = " + ac +
1860 >            ", running = " + rc +
1861              ", steals = " + st +
1862              ", tasks = " + qt +
1863              ", submissions = " + qs +
1864              "]";
1865      }
1866  
1079    private static String runStateToString(int rs) {
1080        switch(rs) {
1081        case RUNNING: return "Running";
1082        case SHUTDOWN: return "Shutting down";
1083        case TERMINATING: return "Terminating";
1084        case TERMINATED: return "Terminated";
1085        default: throw new Error("Unknown run state");
1086        }
1087    }
1088
1089    // lifecycle control
1090
1867      /**
1868       * Initiates an orderly shutdown in which previously submitted
1869       * tasks are executed, but no new tasks will be accepted.
# Line 1102 | Line 1878 | public class ForkJoinPool extends Abstra
1878       */
1879      public void shutdown() {
1880          checkPermission();
1881 <        transitionRunStateTo(SHUTDOWN);
1882 <        if (canTerminateOnShutdown(runControl)) {
1107 <            if (workers == null) { // shutting down before workers created
1108 <                final ReentrantLock lock = this.workerLock;
1109 <                lock.lock();
1110 <                try {
1111 <                    if (workers == null) {
1112 <                        terminate();
1113 <                        transitionRunStateTo(TERMINATED);
1114 <                        termination.signalAll();
1115 <                    }
1116 <                } finally {
1117 <                    lock.unlock();
1118 <                }
1119 <            }
1120 <            terminateOnShutdown();
1121 <        }
1881 >        shutdown = true;
1882 >        tryTerminate(false);
1883      }
1884  
1885      /**
# Line 1139 | Line 1900 | public class ForkJoinPool extends Abstra
1900       */
1901      public List<Runnable> shutdownNow() {
1902          checkPermission();
1903 <        terminate();
1903 >        shutdown = true;
1904 >        tryTerminate(true);
1905          return Collections.emptyList();
1906      }
1907  
# Line 1149 | Line 1911 | public class ForkJoinPool extends Abstra
1911       * @return {@code true} if all tasks have completed following shut down
1912       */
1913      public boolean isTerminated() {
1914 <        return runStateOf(runControl) == TERMINATED;
1914 >        long c = ctl;
1915 >        return ((c & STOP_BIT) != 0L &&
1916 >                (short)(c >>> TC_SHIFT) == -parallelism);
1917      }
1918  
1919      /**
# Line 1157 | Line 1921 | public class ForkJoinPool extends Abstra
1921       * commenced but not yet completed.  This method may be useful for
1922       * debugging. A return of {@code true} reported a sufficient
1923       * period after shutdown may indicate that submitted tasks have
1924 <     * ignored or suppressed interruption, causing this executor not
1925 <     * to properly terminate.
1924 >     * ignored or suppressed interruption, or are waiting for IO,
1925 >     * causing this executor not to properly terminate. (See the
1926 >     * advisory notes for class {@link ForkJoinTask} stating that
1927 >     * tasks should not normally entail blocking operations.  But if
1928 >     * they do, they must abort them on interrupt.)
1929       *
1930       * @return {@code true} if terminating but not yet terminated
1931       */
1932      public boolean isTerminating() {
1933 <        return runStateOf(runControl) == TERMINATING;
1933 >        long c = ctl;
1934 >        return ((c & STOP_BIT) != 0L &&
1935 >                (short)(c >>> TC_SHIFT) != -parallelism);
1936      }
1937  
1938      /**
1939 <     * Returns {@code true} if this pool has been shut down.
1171 <     *
1172 <     * @return {@code true} if this pool has been shut down
1939 >     * Returns true if terminating or terminated. Used by ForkJoinWorkerThread.
1940       */
1941 <    public boolean isShutdown() {
1942 <        return runStateOf(runControl) >= SHUTDOWN;
1941 >    final boolean isAtLeastTerminating() {
1942 >        return (ctl & STOP_BIT) != 0L;
1943      }
1944  
1945      /**
1946 <     * Returns true if pool is not terminating or terminated.
1947 <     * Used internally to suppress execution when terminating.
1946 >     * Returns {@code true} if this pool has been shut down.
1947 >     *
1948 >     * @return {@code true} if this pool has been shut down
1949       */
1950 <    final boolean isProcessingTasks() {
1951 <        return runStateOf(runControl) < TERMINATING;
1950 >    public boolean isShutdown() {
1951 >        return shutdown;
1952      }
1953  
1954      /**
# Line 1197 | Line 1965 | public class ForkJoinPool extends Abstra
1965      public boolean awaitTermination(long timeout, TimeUnit unit)
1966          throws InterruptedException {
1967          long nanos = unit.toNanos(timeout);
1968 <        final ReentrantLock lock = this.workerLock;
1968 >        final ReentrantLock lock = this.submissionLock;
1969          lock.lock();
1970          try {
1971              for (;;) {
# Line 1212 | Line 1980 | public class ForkJoinPool extends Abstra
1980          }
1981      }
1982  
1215    // Shutdown and termination support
1216
1217    /**
1218     * Callback from terminating worker. Nulls out the corresponding
1219     * workers slot, and if terminating, tries to terminate; else
1220     * tries to shrink workers array.
1221     *
1222     * @param w the worker
1223     */
1224    final void workerTerminated(ForkJoinWorkerThread w) {
1225        updateStealCount(w);
1226        updateWorkerCount(-1);
1227        final ReentrantLock lock = this.workerLock;
1228        lock.lock();
1229        try {
1230            ForkJoinWorkerThread[] ws = workers;
1231            if (ws != null) {
1232                int idx = w.poolIndex;
1233                if (idx >= 0 && idx < ws.length && ws[idx] == w)
1234                    ws[idx] = null;
1235                if (totalCountOf(workerCounts) == 0) {
1236                    terminate(); // no-op if already terminating
1237                    transitionRunStateTo(TERMINATED);
1238                    termination.signalAll();
1239                }
1240                else if (isProcessingTasks()) {
1241                    tryShrinkWorkerArray();
1242                    tryResumeSpare(true); // allow replacement
1243                }
1244            }
1245        } finally {
1246            lock.unlock();
1247        }
1248        signalIdleWorkers();
1249    }
1250
1251    /**
1252     * Initiates termination.
1253     */
1254    private void terminate() {
1255        if (transitionRunStateTo(TERMINATING)) {
1256            stopAllWorkers();
1257            resumeAllSpares();
1258            signalIdleWorkers();
1259            cancelQueuedSubmissions();
1260            cancelQueuedWorkerTasks();
1261            interruptUnterminatedWorkers();
1262            signalIdleWorkers(); // resignal after interrupt
1263        }
1264    }
1265
1266    /**
1267     * Possibly terminates when on shutdown state.
1268     */
1269    private void terminateOnShutdown() {
1270        if (!hasQueuedSubmissions() && canTerminateOnShutdown(runControl))
1271            terminate();
1272    }
1273
1274    /**
1275     * Clears out and cancels submissions.
1276     */
1277    private void cancelQueuedSubmissions() {
1278        ForkJoinTask<?> task;
1279        while ((task = pollSubmission()) != null)
1280            task.cancel(false);
1281    }
1282
1283    /**
1284     * Cleans out worker queues.
1285     */
1286    private void cancelQueuedWorkerTasks() {
1287        final ReentrantLock lock = this.workerLock;
1288        lock.lock();
1289        try {
1290            ForkJoinWorkerThread[] ws = workers;
1291            if (ws != null) {
1292                for (int i = 0; i < ws.length; ++i) {
1293                    ForkJoinWorkerThread t = ws[i];
1294                    if (t != null)
1295                        t.cancelTasks();
1296                }
1297            }
1298        } finally {
1299            lock.unlock();
1300        }
1301    }
1302
1303    /**
1304     * Sets each worker's status to terminating. Requires lock to avoid
1305     * conflicts with add/remove.
1306     */
1307    private void stopAllWorkers() {
1308        final ReentrantLock lock = this.workerLock;
1309        lock.lock();
1310        try {
1311            ForkJoinWorkerThread[] ws = workers;
1312            if (ws != null) {
1313                for (int i = 0; i < ws.length; ++i) {
1314                    ForkJoinWorkerThread t = ws[i];
1315                    if (t != null)
1316                        t.shutdownNow();
1317                }
1318            }
1319        } finally {
1320            lock.unlock();
1321        }
1322    }
1323
1324    /**
1325     * Interrupts all unterminated workers.  This is not required for
1326     * sake of internal control, but may help unstick user code during
1327     * shutdown.
1328     */
1329    private void interruptUnterminatedWorkers() {
1330        final ReentrantLock lock = this.workerLock;
1331        lock.lock();
1332        try {
1333            ForkJoinWorkerThread[] ws = workers;
1334            if (ws != null) {
1335                for (int i = 0; i < ws.length; ++i) {
1336                    ForkJoinWorkerThread t = ws[i];
1337                    if (t != null && !t.isTerminated()) {
1338                        try {
1339                            t.interrupt();
1340                        } catch (SecurityException ignore) {
1341                        }
1342                    }
1343                }
1344            }
1345        } finally {
1346            lock.unlock();
1347        }
1348    }
1349
1350
1351    /*
1352     * Nodes for event barrier to manage idle threads.  Queue nodes
1353     * are basic Treiber stack nodes, also used for spare stack.
1354     *
1355     * The event barrier has an event count and a wait queue (actually
1356     * a Treiber stack).  Workers are enabled to look for work when
1357     * the eventCount is incremented. If they fail to find work, they
1358     * may wait for next count. Upon release, threads help others wake
1359     * up.
1360     *
1361     * Synchronization events occur only in enough contexts to
1362     * maintain overall liveness:
1363     *
1364     *   - Submission of a new task to the pool
1365     *   - Resizes or other changes to the workers array
1366     *   - pool termination
1367     *   - A worker pushing a task on an empty queue
1368     *
1369     * The case of pushing a task occurs often enough, and is heavy
1370     * enough compared to simple stack pushes, to require special
1371     * handling: Method signalWork returns without advancing count if
1372     * the queue appears to be empty.  This would ordinarily result in
1373     * races causing some queued waiters not to be woken up. To avoid
1374     * this, the first worker enqueued in method sync (see
1375     * syncIsReleasable) rescans for tasks after being enqueued, and
1376     * helps signal if any are found. This works well because the
1377     * worker has nothing better to do, and so might as well help
1378     * alleviate the overhead and contention on the threads actually
1379     * doing work.  Also, since event counts increments on task
1380     * availability exist to maintain liveness (rather than to force
1381     * refreshes etc), it is OK for callers to exit early if
1382     * contending with another signaller.
1383     */
1384    static final class WaitQueueNode {
1385        WaitQueueNode next; // only written before enqueued
1386        volatile ForkJoinWorkerThread thread; // nulled to cancel wait
1387        final long count; // unused for spare stack
1388
1389        WaitQueueNode(long c, ForkJoinWorkerThread w) {
1390            count = c;
1391            thread = w;
1392        }
1393
1394        /**
1395         * Wakes up waiter, returning false if known to already
1396         */
1397        boolean signal() {
1398            ForkJoinWorkerThread t = thread;
1399            if (t == null)
1400                return false;
1401            thread = null;
1402            LockSupport.unpark(t);
1403            return true;
1404        }
1405
1406        /**
1407         * Awaits release on sync.
1408         */
1409        void awaitSyncRelease(ForkJoinPool p) {
1410            while (thread != null && !p.syncIsReleasable(this))
1411                LockSupport.park(this);
1412        }
1413
1414        /**
1415         * Awaits resumption as spare.
1416         */
1417        void awaitSpareRelease() {
1418            while (thread != null) {
1419                if (!Thread.interrupted())
1420                    LockSupport.park(this);
1421            }
1422        }
1423    }
1424
1425    /**
1426     * Ensures that no thread is waiting for count to advance from the
1427     * current value of eventCount read on entry to this method, by
1428     * releasing waiting threads if necessary.
1429     *
1430     * @return the count
1431     */
1432    final long ensureSync() {
1433        long c = eventCount;
1434        WaitQueueNode q;
1435        while ((q = syncStack) != null && q.count < c) {
1436            if (casBarrierStack(q, null)) {
1437                do {
1438                    q.signal();
1439                } while ((q = q.next) != null);
1440                break;
1441            }
1442        }
1443        return c;
1444    }
1445
1446    /**
1447     * Increments event count and releases waiting threads.
1448     */
1449    private void signalIdleWorkers() {
1450        long c;
1451        do {} while (!casEventCount(c = eventCount, c+1));
1452        ensureSync();
1453    }
1454
1455    /**
1456     * Signals threads waiting to poll a task. Because method sync
1457     * rechecks availability, it is OK to only proceed if queue
1458     * appears to be non-empty, and OK to skip under contention to
1459     * increment count (since some other thread succeeded).
1460     */
1461    final void signalWork() {
1462        long c;
1463        WaitQueueNode q;
1464        if (syncStack != null &&
1465            casEventCount(c = eventCount, c+1) &&
1466            (((q = syncStack) != null && q.count <= c) &&
1467             (!casBarrierStack(q, q.next) || !q.signal())))
1468            ensureSync();
1469    }
1470
1471    /**
1472     * Waits until event count advances from last value held by
1473     * caller, or if excess threads, caller is resumed as spare, or
1474     * caller or pool is terminating. Updates caller's event on exit.
1475     *
1476     * @param w the calling worker thread
1477     */
1478    final void sync(ForkJoinWorkerThread w) {
1479        updateStealCount(w); // Transfer w's count while it is idle
1480
1481        while (!w.isShutdown() && isProcessingTasks() && !suspendIfSpare(w)) {
1482            long prev = w.lastEventCount;
1483            WaitQueueNode node = null;
1484            WaitQueueNode h;
1485            while (eventCount == prev &&
1486                   ((h = syncStack) == null || h.count == prev)) {
1487                if (node == null)
1488                    node = new WaitQueueNode(prev, w);
1489                if (casBarrierStack(node.next = h, node)) {
1490                    node.awaitSyncRelease(this);
1491                    break;
1492                }
1493            }
1494            long ec = ensureSync();
1495            if (ec != prev) {
1496                w.lastEventCount = ec;
1497                break;
1498            }
1499        }
1500    }
1501
1502    /**
1503     * Returns {@code true} if worker waiting on sync can proceed:
1504     *  - on signal (thread == null)
1505     *  - on event count advance (winning race to notify vs signaller)
1506     *  - on interrupt
1507     *  - if the first queued node, we find work available
1508     * If node was not signalled and event count not advanced on exit,
1509     * then we also help advance event count.
1510     *
1511     * @return {@code true} if node can be released
1512     */
1513    final boolean syncIsReleasable(WaitQueueNode node) {
1514        long prev = node.count;
1515        if (!Thread.interrupted() && node.thread != null &&
1516            (node.next != null ||
1517             !ForkJoinWorkerThread.hasQueuedTasks(workers)) &&
1518            eventCount == prev)
1519            return false;
1520        if (node.thread != null) {
1521            node.thread = null;
1522            long ec = eventCount;
1523            if (prev <= ec) // help signal
1524                casEventCount(ec, ec+1);
1525        }
1526        return true;
1527    }
1528
1529    /**
1530     * Returns {@code true} if a new sync event occurred since last
1531     * call to sync or this method, if so, updating caller's count.
1532     */
1533    final boolean hasNewSyncEvent(ForkJoinWorkerThread w) {
1534        long lc = w.lastEventCount;
1535        long ec = ensureSync();
1536        if (ec == lc)
1537            return false;
1538        w.lastEventCount = ec;
1539        return true;
1540    }
1541
1542    //  Parallelism maintenance
1543
1544    /**
1545     * Decrements running count; if too low, adds spare.
1546     *
1547     * Conceptually, all we need to do here is add or resume a
1548     * spare thread when one is about to block (and remove or
1549     * suspend it later when unblocked -- see suspendIfSpare).
1550     * However, implementing this idea requires coping with
1551     * several problems: we have imperfect information about the
1552     * states of threads. Some count updates can and usually do
1553     * lag run state changes, despite arrangements to keep them
1554     * accurate (for example, when possible, updating counts
1555     * before signalling or resuming), especially when running on
1556     * dynamic JVMs that don't optimize the infrequent paths that
1557     * update counts. Generating too many threads can make these
1558     * problems become worse, because excess threads are more
1559     * likely to be context-switched with others, slowing them all
1560     * down, especially if there is no work available, so all are
1561     * busy scanning or idling.  Also, excess spare threads can
1562     * only be suspended or removed when they are idle, not
1563     * immediately when they aren't needed. So adding threads will
1564     * raise parallelism level for longer than necessary.  Also,
1565     * FJ applications often encounter highly transient peaks when
1566     * many threads are blocked joining, but for less time than it
1567     * takes to create or resume spares.
1568     *
1569     * @param joinMe if non-null, return early if done
1570     * @param maintainParallelism if true, try to stay within
1571     * target counts, else create only to avoid starvation
1572     * @return true if joinMe known to be done
1573     */
1574    final boolean preJoin(ForkJoinTask<?> joinMe,
1575                          boolean maintainParallelism) {
1576        maintainParallelism &= maintainsParallelism; // overrride
1577        boolean dec = false;  // true when running count decremented
1578        while (spareStack == null || !tryResumeSpare(dec)) {
1579            int counts = workerCounts;
1580            if (dec || (dec = casWorkerCounts(counts, --counts))) {
1581                if (!needSpare(counts, maintainParallelism))
1582                    break;
1583                if (joinMe.status < 0)
1584                    return true;
1585                if (tryAddSpare(counts))
1586                    break;
1587            }
1588        }
1589        return false;
1590    }
1591
1592    /**
1593     * Same idea as preJoin
1594     */
1595    final boolean preBlock(ManagedBlocker blocker,
1596                           boolean maintainParallelism) {
1597        maintainParallelism &= maintainsParallelism;
1598        boolean dec = false;
1599        while (spareStack == null || !tryResumeSpare(dec)) {
1600            int counts = workerCounts;
1601            if (dec || (dec = casWorkerCounts(counts, --counts))) {
1602                if (!needSpare(counts, maintainParallelism))
1603                    break;
1604                if (blocker.isReleasable())
1605                    return true;
1606                if (tryAddSpare(counts))
1607                    break;
1608            }
1609        }
1610        return false;
1611    }
1612
1613    /**
1614     * Returns {@code true} if a spare thread appears to be needed.
1615     * If maintaining parallelism, returns true when the deficit in
1616     * running threads is more than the surplus of total threads, and
1617     * there is apparently some work to do.  This self-limiting rule
1618     * means that the more threads that have already been added, the
1619     * less parallelism we will tolerate before adding another.
1620     *
1621     * @param counts current worker counts
1622     * @param maintainParallelism try to maintain parallelism
1623     */
1624    private boolean needSpare(int counts, boolean maintainParallelism) {
1625        int ps = parallelism;
1626        int rc = runningCountOf(counts);
1627        int tc = totalCountOf(counts);
1628        int runningDeficit = ps - rc;
1629        int totalSurplus = tc - ps;
1630        return (tc < maxPoolSize &&
1631                (rc == 0 || totalSurplus < 0 ||
1632                 (maintainParallelism &&
1633                  runningDeficit > totalSurplus &&
1634                  ForkJoinWorkerThread.hasQueuedTasks(workers))));
1635    }
1636
1637    /**
1638     * Adds a spare worker if lock available and no more than the
1639     * expected numbers of threads exist.
1640     *
1641     * @return true if successful
1642     */
1643    private boolean tryAddSpare(int expectedCounts) {
1644        final ReentrantLock lock = this.workerLock;
1645        int expectedRunning = runningCountOf(expectedCounts);
1646        int expectedTotal = totalCountOf(expectedCounts);
1647        boolean success = false;
1648        boolean locked = false;
1649        // confirm counts while locking; CAS after obtaining lock
1650        try {
1651            for (;;) {
1652                int s = workerCounts;
1653                int tc = totalCountOf(s);
1654                int rc = runningCountOf(s);
1655                if (rc > expectedRunning || tc > expectedTotal)
1656                    break;
1657                if (!locked && !(locked = lock.tryLock()))
1658                    break;
1659                if (casWorkerCounts(s, workerCountsFor(tc+1, rc+1))) {
1660                    createAndStartSpare(tc);
1661                    success = true;
1662                    break;
1663                }
1664            }
1665        } finally {
1666            if (locked)
1667                lock.unlock();
1668        }
1669        return success;
1670    }
1671
1672    /**
1673     * Adds the kth spare worker. On entry, pool counts are already
1674     * adjusted to reflect addition.
1675     */
1676    private void createAndStartSpare(int k) {
1677        ForkJoinWorkerThread w = null;
1678        ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(k + 1);
1679        int len = ws.length;
1680        // Probably, we can place at slot k. If not, find empty slot
1681        if (k < len && ws[k] != null) {
1682            for (k = 0; k < len && ws[k] != null; ++k)
1683                ;
1684        }
1685        if (k < len && isProcessingTasks() && (w = createWorker(k)) != null) {
1686            ws[k] = w;
1687            w.start();
1688        }
1689        else
1690            updateWorkerCount(-1); // adjust on failure
1691        signalIdleWorkers();
1692    }
1693
1694    /**
1695     * Suspends calling thread w if there are excess threads.  Called
1696     * only from sync.  Spares are enqueued in a Treiber stack using
1697     * the same WaitQueueNodes as barriers.  They are resumed mainly
1698     * in preJoin, but are also woken on pool events that require all
1699     * threads to check run state.
1700     *
1701     * @param w the caller
1702     */
1703    private boolean suspendIfSpare(ForkJoinWorkerThread w) {
1704        WaitQueueNode node = null;
1705        int s;
1706        while (parallelism < runningCountOf(s = workerCounts)) {
1707            if (node == null)
1708                node = new WaitQueueNode(0, w);
1709            if (casWorkerCounts(s, s-1)) { // representation-dependent
1710                // push onto stack
1711                do {} while (!casSpareStack(node.next = spareStack, node));
1712                // block until released by resumeSpare
1713                node.awaitSpareRelease();
1714                return true;
1715            }
1716        }
1717        return false;
1718    }
1719
1720    /**
1721     * Tries to pop and resume a spare thread.
1722     *
1723     * @param updateCount if true, increment running count on success
1724     * @return true if successful
1725     */
1726    private boolean tryResumeSpare(boolean updateCount) {
1727        WaitQueueNode q;
1728        while ((q = spareStack) != null) {
1729            if (casSpareStack(q, q.next)) {
1730                if (updateCount)
1731                    updateRunningCount(1);
1732                q.signal();
1733                return true;
1734            }
1735        }
1736        return false;
1737    }
1738
1739    /**
1740     * Pops and resumes all spare threads. Same idea as ensureSync.
1741     *
1742     * @return true if any spares released
1743     */
1744    private boolean resumeAllSpares() {
1745        WaitQueueNode q;
1746        while ( (q = spareStack) != null) {
1747            if (casSpareStack(q, null)) {
1748                do {
1749                    updateRunningCount(1);
1750                    q.signal();
1751                } while ((q = q.next) != null);
1752                return true;
1753            }
1754        }
1755        return false;
1756    }
1757
1758    /**
1759     * Pops and shuts down excessive spare threads. Call only while
1760     * holding lock. This is not guaranteed to eliminate all excess
1761     * threads, only those suspended as spares, which are the ones
1762     * unlikely to be needed in the future.
1763     */
1764    private void trimSpares() {
1765        int surplus = totalCountOf(workerCounts) - parallelism;
1766        WaitQueueNode q;
1767        while (surplus > 0 && (q = spareStack) != null) {
1768            if (casSpareStack(q, null)) {
1769                do {
1770                    updateRunningCount(1);
1771                    ForkJoinWorkerThread w = q.thread;
1772                    if (w != null && surplus > 0 &&
1773                        runningCountOf(workerCounts) > 0 && w.shutdown())
1774                        --surplus;
1775                    q.signal();
1776                } while ((q = q.next) != null);
1777            }
1778        }
1779    }
1780
1983      /**
1984       * Interface for extending managed parallelism for tasks running
1985       * in {@link ForkJoinPool}s.
1986       *
1987 <     * <p>A {@code ManagedBlocker} provides two methods.
1988 <     * Method {@code isReleasable} must return {@code true} if
1989 <     * blocking is not necessary. Method {@code block} blocks the
1990 <     * current thread if necessary (perhaps internally invoking
1991 <     * {@code isReleasable} before actually blocking).
1987 >     * <p>A {@code ManagedBlocker} provides two methods.  Method
1988 >     * {@code isReleasable} must return {@code true} if blocking is
1989 >     * not necessary. Method {@code block} blocks the current thread
1990 >     * if necessary (perhaps internally invoking {@code isReleasable}
1991 >     * before actually blocking). These actions are performed by any
1992 >     * thread invoking {@link ForkJoinPool#managedBlock}.  The
1993 >     * unusual methods in this API accommodate synchronizers that may,
1994 >     * but don't usually, block for long periods. Similarly, they
1995 >     * allow more efficient internal handling of cases in which
1996 >     * additional workers may be, but usually are not, needed to
1997 >     * ensure sufficient parallelism.  Toward this end,
1998 >     * implementations of method {@code isReleasable} must be amenable
1999 >     * to repeated invocation.
2000       *
2001       * <p>For example, here is a ManagedBlocker based on a
2002       * ReentrantLock:
# Line 1804 | Line 2014 | public class ForkJoinPool extends Abstra
2014       *     return hasLock || (hasLock = lock.tryLock());
2015       *   }
2016       * }}</pre>
2017 +     *
2018 +     * <p>Here is a class that possibly blocks waiting for an
2019 +     * item on a given queue:
2020 +     *  <pre> {@code
2021 +     * class QueueTaker<E> implements ManagedBlocker {
2022 +     *   final BlockingQueue<E> queue;
2023 +     *   volatile E item = null;
2024 +     *   QueueTaker(BlockingQueue<E> q) { this.queue = q; }
2025 +     *   public boolean block() throws InterruptedException {
2026 +     *     if (item == null)
2027 +     *       item = queue.take();
2028 +     *     return true;
2029 +     *   }
2030 +     *   public boolean isReleasable() {
2031 +     *     return item != null || (item = queue.poll()) != null;
2032 +     *   }
2033 +     *   public E getItem() { // call after pool.managedBlock completes
2034 +     *     return item;
2035 +     *   }
2036 +     * }}</pre>
2037       */
2038      public static interface ManagedBlocker {
2039          /**
# Line 1827 | Line 2057 | public class ForkJoinPool extends Abstra
2057       * Blocks in accord with the given blocker.  If the current thread
2058       * is a {@link ForkJoinWorkerThread}, this method possibly
2059       * arranges for a spare thread to be activated if necessary to
2060 <     * ensure parallelism while the current thread is blocked.
1831 <     *
1832 <     * <p>If {@code maintainParallelism} is {@code true} and the pool
1833 <     * supports it ({@link #getMaintainsParallelism}), this method
1834 <     * attempts to maintain the pool's nominal parallelism. Otherwise
1835 <     * it activates a thread only if necessary to avoid complete
1836 <     * starvation. This option may be preferable when blockages use
1837 <     * timeouts, or are almost always brief.
2060 >     * ensure sufficient parallelism while the current thread is blocked.
2061       *
2062       * <p>If the caller is not a {@link ForkJoinTask}, this method is
2063       * behaviorally equivalent to
# Line 1848 | Line 2071 | public class ForkJoinPool extends Abstra
2071       * first be expanded to ensure parallelism, and later adjusted.
2072       *
2073       * @param blocker the blocker
1851     * @param maintainParallelism if {@code true} and supported by
1852     * this pool, attempt to maintain the pool's nominal parallelism;
1853     * otherwise activate a thread only if necessary to avoid
1854     * complete starvation.
2074       * @throws InterruptedException if blocker.block did so
2075       */
2076 <    public static void managedBlock(ManagedBlocker blocker,
1858 <                                    boolean maintainParallelism)
2076 >    public static void managedBlock(ManagedBlocker blocker)
2077          throws InterruptedException {
2078          Thread t = Thread.currentThread();
2079 <        ForkJoinPool pool = ((t instanceof ForkJoinWorkerThread) ?
2080 <                             ((ForkJoinWorkerThread) t).pool : null);
2081 <        if (!blocker.isReleasable()) {
2082 <            try {
2083 <                if (pool == null ||
2084 <                    !pool.preBlock(blocker, maintainParallelism))
1867 <                    awaitBlocker(blocker);
1868 <            } finally {
1869 <                if (pool != null)
1870 <                    pool.updateRunningCount(1);
1871 <            }
2079 >        if (t instanceof ForkJoinWorkerThread) {
2080 >            ForkJoinWorkerThread w = (ForkJoinWorkerThread) t;
2081 >            w.pool.awaitBlocker(blocker);
2082 >        }
2083 >        else {
2084 >            do {} while (!blocker.isReleasable() && !blocker.block());
2085          }
1873    }
1874
1875    private static void awaitBlocker(ManagedBlocker blocker)
1876        throws InterruptedException {
1877        do {} while (!blocker.isReleasable() && !blocker.block());
2086      }
2087  
2088      // AbstractExecutorService overrides.  These rely on undocumented
# Line 1890 | Line 2098 | public class ForkJoinPool extends Abstra
2098      }
2099  
2100      // Unsafe mechanics
2101 <
2102 <    private static final sun.misc.Unsafe UNSAFE = getUnsafe();
2103 <    private static final long eventCountOffset =
2104 <        objectFieldOffset("eventCount", ForkJoinPool.class);
2105 <    private static final long workerCountsOffset =
2106 <        objectFieldOffset("workerCounts", ForkJoinPool.class);
2107 <    private static final long runControlOffset =
2108 <        objectFieldOffset("runControl", ForkJoinPool.class);
2109 <    private static final long syncStackOffset =
2110 <        objectFieldOffset("syncStack",ForkJoinPool.class);
2111 <    private static final long spareStackOffset =
2112 <        objectFieldOffset("spareStack", ForkJoinPool.class);
2113 <
2114 <    private boolean casEventCount(long cmp, long val) {
2115 <        return UNSAFE.compareAndSwapLong(this, eventCountOffset, cmp, val);
2116 <    }
1909 <    private boolean casWorkerCounts(int cmp, int val) {
1910 <        return UNSAFE.compareAndSwapInt(this, workerCountsOffset, cmp, val);
1911 <    }
1912 <    private boolean casRunControl(int cmp, int val) {
1913 <        return UNSAFE.compareAndSwapInt(this, runControlOffset, cmp, val);
1914 <    }
1915 <    private boolean casSpareStack(WaitQueueNode cmp, WaitQueueNode val) {
1916 <        return UNSAFE.compareAndSwapObject(this, spareStackOffset, cmp, val);
1917 <    }
1918 <    private boolean casBarrierStack(WaitQueueNode cmp, WaitQueueNode val) {
1919 <        return UNSAFE.compareAndSwapObject(this, syncStackOffset, cmp, val);
1920 <    }
1921 <
1922 <    private static long objectFieldOffset(String field, Class<?> klazz) {
2101 >    private static final sun.misc.Unsafe UNSAFE;
2102 >    private static final long ctlOffset;
2103 >    private static final long stealCountOffset;
2104 >    private static final long blockedCountOffset;
2105 >    private static final long quiescerCountOffset;
2106 >    private static final long scanGuardOffset;
2107 >    private static final long nextWorkerNumberOffset;
2108 >    private static final long ABASE;
2109 >    private static final int ASHIFT;
2110 >
2111 >    static {
2112 >        poolNumberGenerator = new AtomicInteger();
2113 >        workerSeedGenerator = new Random();
2114 >        modifyThreadPermission = new RuntimePermission("modifyThread");
2115 >        defaultForkJoinWorkerThreadFactory =
2116 >            new DefaultForkJoinWorkerThreadFactory();
2117          try {
2118 <            return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
2119 <        } catch (NoSuchFieldException e) {
2120 <            // Convert Exception to corresponding Error
2121 <            NoSuchFieldError error = new NoSuchFieldError(field);
2122 <            error.initCause(e);
2123 <            throw error;
2124 <        }
2118 >            UNSAFE = getUnsafe();
2119 >            Class<?> k = ForkJoinPool.class;
2120 >            ctlOffset = UNSAFE.objectFieldOffset
2121 >                (k.getDeclaredField("ctl"));
2122 >            stealCountOffset = UNSAFE.objectFieldOffset
2123 >                (k.getDeclaredField("stealCount"));
2124 >            blockedCountOffset = UNSAFE.objectFieldOffset
2125 >                (k.getDeclaredField("blockedCount"));
2126 >            quiescerCountOffset = UNSAFE.objectFieldOffset
2127 >                (k.getDeclaredField("quiescerCount"));
2128 >            scanGuardOffset = UNSAFE.objectFieldOffset
2129 >                (k.getDeclaredField("scanGuard"));
2130 >            nextWorkerNumberOffset = UNSAFE.objectFieldOffset
2131 >                (k.getDeclaredField("nextWorkerNumber"));
2132 >        } catch (Exception e) {
2133 >            throw new Error(e);
2134 >        }
2135 >        Class<?> a = ForkJoinTask[].class;
2136 >        ABASE = UNSAFE.arrayBaseOffset(a);
2137 >        int s = UNSAFE.arrayIndexScale(a);
2138 >        if ((s & (s-1)) != 0)
2139 >            throw new Error("data type scale not a power of two");
2140 >        ASHIFT = 31 - Integer.numberOfLeadingZeros(s);
2141      }
2142  
2143      /**

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines