ViewVC Help

View File | Revision Log | Show Annotations | Download File | Root Listing

root/jsr166/jsr166/src/jsr166y/ForkJoinPool.java

Comparing jsr166/src/jsr166y/ForkJoinPool.java (file contents):
Revision 1.41 by jsr166, Mon Aug 3 01:11:58 2009 UTC vs.
Revision 1.67 by jsr166, Wed Sep 1 03:32:03 2010 UTC

#	Line 13 | Line 13 | import java.util.Arrays;
13		import java.util.Collection;
14		import java.util.Collections;
15		import java.util.List;
16	–	import java.util.concurrent.locks.Condition;
16		import java.util.concurrent.locks.LockSupport;
17		import java.util.concurrent.locks.ReentrantLock;
18		import java.util.concurrent.atomic.AtomicInteger;
19	<	import java.util.concurrent.atomic.AtomicLong;
19	>	import java.util.concurrent.CountDownLatch;
20
21		/**
22		* An {@link ExecutorService} for running {@link ForkJoinTask}s.
23		* A {@code ForkJoinPool} provides the entry point for submissions
24	<	* from non-{@code ForkJoinTask}s, as well as management and
25	<	* monitoring operations.  Normally a single {@code ForkJoinPool} is
27	<	* used for a large number of submitted tasks. Otherwise, use would
28	<	* not usually outweigh the construction and bookkeeping overhead of
29	<	* creating a large set of threads.
24	>	* from non-{@code ForkJoinTask} clients, as well as management and
25	>	* monitoring operations.
26		*
27	<	* <p>{@code ForkJoinPool}s differ from other kinds of {@link
28	<	* Executor}s mainly in that they provide <em>work-stealing</em>: all
29	<	* threads in the pool attempt to find and execute subtasks created by
30	<	* other active tasks (eventually blocking if none exist). This makes
31	<	* them efficient when most tasks spawn other subtasks (as do most
32	<	* {@code ForkJoinTask}s), as well as the mixed execution of some
33	<	* plain {@code Runnable}- or {@code Callable}- based activities along
34	<	* with {@code ForkJoinTask}s. When setting {@linkplain #setAsyncMode
35	<	* async mode}, a {@code ForkJoinPool} may also be appropriate for use
40	<	* with fine-grained tasks that are never joined. Otherwise, other
41	<	* {@code ExecutorService} implementations are typically more
42	<	* appropriate choices.
27	>	* <p>A {@code ForkJoinPool} differs from other kinds of {@link
28	>	* ExecutorService} mainly by virtue of employing
29	>	* <em>work-stealing</em>: all threads in the pool attempt to find and
30	>	* execute subtasks created by other active tasks (eventually blocking
31	>	* waiting for work if none exist). This enables efficient processing
32	>	* when most tasks spawn other subtasks (as do most {@code
33	>	* ForkJoinTask}s). When setting <em>asyncMode</em> to true in
34	>	* constructors, {@code ForkJoinPool}s may also be appropriate for use
35	>	* with event-style tasks that are never joined.
36		*
37	<	* <p>A {@code ForkJoinPool} may be constructed with a given
38	<	* parallelism level (target pool size), which it attempts to maintain
39	<	* by dynamically adding, suspending, or resuming threads, even if
40	<	* some tasks are waiting to join others. However, no such adjustments
41	<	* are performed in the face of blocked IO or other unmanaged
42	<	* synchronization. The nested {@link ManagedBlocker} interface
43	<	* enables extension of the kinds of synchronization accommodated.
44	<	* The target parallelism level may also be changed dynamically
45	<	* ({@link #setParallelism}) and thread construction can be limited
53	<	* using methods {@link #setMaximumPoolSize} and/or {@link
54	<	* #setMaintainsParallelism}.
37	>	* <p>A {@code ForkJoinPool} is constructed with a given target
38	>	* parallelism level; by default, equal to the number of available
39	>	* processors. The pool attempts to maintain enough active (or
40	>	* available) threads by dynamically adding, suspending, or resuming
41	>	* internal worker threads, even if some tasks are stalled waiting to
42	>	* join others. However, no such adjustments are guaranteed in the
43	>	* face of blocked IO or other unmanaged synchronization. The nested
44	>	* {@link ManagedBlocker} interface enables extension of the kinds of
45	>	* synchronization accommodated.
46		*
47		* <p>In addition to execution and lifecycle control methods, this
48		* class provides status check methods (for example
#	Line 60 | Line 51 | import java.util.concurrent.atomic.Atomi
51		* {@link #toString} returns indications of pool state in a
52		* convenient form for informal monitoring.
53		*
54	+	* <p> As is the case with other ExecutorServices, there are three
55	+	* main task execution methods summarized in the following
56	+	* table. These are designed to be used by clients not already engaged
57	+	* in fork/join computations in the current pool.  The main forms of
58	+	* these methods accept instances of {@code ForkJoinTask}, but
59	+	* overloaded forms also allow mixed execution of plain {@code
60	+	* Runnable}- or {@code Callable}- based activities as well.  However,
61	+	* tasks that are already executing in a pool should normally
62	+	* <em>NOT</em> use these pool execution methods, but instead use the
63	+	* within-computation forms listed in the table.
64	+	*
65	+	* <table BORDER CELLPADDING=3 CELLSPACING=1>
66	+	*  <tr>
67	+	*    <td></td>
68	+	*    <td ALIGN=CENTER> <b>Call from non-fork/join clients</b></td>
69	+	*    <td ALIGN=CENTER> <b>Call from within fork/join computations</b></td>
70	+	*  </tr>
71	+	*  <tr>
72	+	*    <td> <b>Arrange async execution</td>
73	+	*    <td> {@link #execute(ForkJoinTask)}</td>
74	+	*    <td> {@link ForkJoinTask#fork}</td>
75	+	*  </tr>
76	+	*  <tr>
77	+	*    <td> <b>Await and obtain result</td>
78	+	*    <td> {@link #invoke(ForkJoinTask)}</td>
79	+	*    <td> {@link ForkJoinTask#invoke}</td>
80	+	*  </tr>
81	+	*  <tr>
82	+	*    <td> <b>Arrange exec and obtain Future</td>
83	+	*    <td> {@link #submit(ForkJoinTask)}</td>
84	+	*    <td> {@link ForkJoinTask#fork} (ForkJoinTasks <em>are</em> Futures)</td>
85	+	*  </tr>
86	+	* </table>
87	+	*
88	+	* <p><b>Sample Usage.</b> Normally a single {@code ForkJoinPool} is
89	+	* used for all parallel task execution in a program or subsystem.
90	+	* Otherwise, use would not usually outweigh the construction and
91	+	* bookkeeping overhead of creating a large set of threads. For
92	+	* example, a common pool could be used for the {@code SortTasks}
93	+	* illustrated in {@link RecursiveAction}. Because {@code
94	+	* ForkJoinPool} uses threads in {@linkplain java.lang.Thread#isDaemon
95	+	* daemon} mode, there is typically no need to explicitly {@link
96	+	* #shutdown} such a pool upon program exit.
97	+	*
98	+	* <pre>
99	+	* static final ForkJoinPool mainPool = new ForkJoinPool();
100	+	* ...
101	+	* public void sort(long[] array) {
102	+	*   mainPool.invoke(new SortTask(array, 0, array.length));
103	+	* }
104	+	* </pre>
105	+	*
106		* <p><b>Implementation notes</b>: This implementation restricts the
107		* maximum number of running threads to 32767. Attempts to create
108	<	* pools with greater than the maximum result in
108	>	* pools with greater than the maximum number result in
109		* {@code IllegalArgumentException}.
110		*
111	+	* <p>This implementation rejects submitted tasks (that is, by throwing
112	+	* {@link RejectedExecutionException}) only when the pool is shut down
113	+	* or internal resources have been exhausted.
114	+	*
115		* @since 1.7
116		* @author Doug Lea
117		*/
118		public class ForkJoinPool extends AbstractExecutorService {
119
120		/*
121	<	* See the extended comments interspersed below for design,
122	<	* rationale, and walkthroughs.
121	>	* Implementation Overview
122	>	*
123	>	* This class provides the central bookkeeping and control for a
124	>	* set of worker threads: Submissions from non-FJ threads enter
125	>	* into a submission queue. Workers take these tasks and typically
126	>	* split them into subtasks that may be stolen by other workers.
127	>	* The main work-stealing mechanics implemented in class
128	>	* ForkJoinWorkerThread give first priority to processing tasks
129	>	* from their own queues (LIFO or FIFO, depending on mode), then
130	>	* to randomized FIFO steals of tasks in other worker queues, and
131	>	* lastly to new submissions. These mechanics do not consider
132	>	* affinities, loads, cache localities, etc, so rarely provide the
133	>	* best possible performance on a given machine, but portably
134	>	* provide good throughput by averaging over these factors.
135	>	* (Further, even if we did try to use such information, we do not
136	>	* usually have a basis for exploiting it. For example, some sets
137	>	* of tasks profit from cache affinities, but others are harmed by
138	>	* cache pollution effects.)
139	>	*
140	>	* Beyond work-stealing support and essential bookkeeping, the
141	>	* main responsibility of this framework is to take actions when
142	>	* one worker is waiting to join a task stolen (or always held by)
143	>	* another.  Because we are multiplexing many tasks on to a pool
144	>	* of workers, we can't just let them block (as in Thread.join).
145	>	* We also cannot just reassign the joiner's run-time stack with
146	>	* another and replace it later, which would be a form of
147	>	* "continuation", that even if possible is not necessarily a good
148	>	* idea. Given that the creation costs of most threads on most
149	>	* systems mainly surrounds setting up runtime stacks, thread
150	>	* creation and switching is usually not much more expensive than
151	>	* stack creation and switching, and is more flexible). Instead we
152	>	* combine two tactics:
153	>	*
154	>	*   Helping: Arranging for the joiner to execute some task that it
155	>	*      would be running if the steal had not occurred.  Method
156	>	*      ForkJoinWorkerThread.helpJoinTask tracks joining->stealing
157	>	*      links to try to find such a task.
158	>	*
159	>	*   Compensating: Unless there are already enough live threads,
160	>	*      method helpMaintainParallelism() may create or or
161	>	*      re-activate a spare thread to compensate for blocked
162	>	*      joiners until they unblock.
163	>	*
164	>	* It is impossible to keep exactly the target (parallelism)
165	>	* number of threads running at any given time.  Determining
166	>	* existence of conservatively safe helping targets, the
167	>	* availability of already-created spares, and the apparent need
168	>	* to create new spares are all racy and require heuristic
169	>	* guidance, so we rely on multiple retries of each.  Compensation
170	>	* occurs in slow-motion. It is triggered only upon timeouts of
171	>	* Object.wait used for joins. This reduces poor decisions that
172	>	* would otherwise be made when threads are waiting for others
173	>	* that are stalled because of unrelated activities such as
174	>	* garbage collection.
175	>	*
176	>	* The ManagedBlocker extension API can't use helping so relies
177	>	* only on compensation in method awaitBlocker.
178	>	*
179	>	* The main throughput advantages of work-stealing stem from
180	>	* decentralized control -- workers mostly steal tasks from each
181	>	* other. We do not want to negate this by creating bottlenecks
182	>	* implementing other management responsibilities. So we use a
183	>	* collection of techniques that avoid, reduce, or cope well with
184	>	* contention. These entail several instances of bit-packing into
185	>	* CASable fields to maintain only the minimally required
186	>	* atomicity. To enable such packing, we restrict maximum
187	>	* parallelism to (1<<15)-1 (enabling twice this (to accommodate
188	>	* unbalanced increments and decrements) to fit into a 16 bit
189	>	* field, which is far in excess of normal operating range.  Even
190	>	* though updates to some of these bookkeeping fields do sometimes
191	>	* contend with each other, they don't normally cache-contend with
192	>	* updates to others enough to warrant memory padding or
193	>	* isolation. So they are all held as fields of ForkJoinPool
194	>	* objects.  The main capabilities are as follows:
195	>	*
196	>	* 1. Creating and removing workers. Workers are recorded in the
197	>	* "workers" array. This is an array as opposed to some other data
198	>	* structure to support index-based random steals by workers.
199	>	* Updates to the array recording new workers and unrecording
200	>	* terminated ones are protected from each other by a lock
201	>	* (workerLock) but the array is otherwise concurrently readable,
202	>	* and accessed directly by workers. To simplify index-based
203	>	* operations, the array size is always a power of two, and all
204	>	* readers must tolerate null slots. Currently, all worker thread
205	>	* creation is on-demand, triggered by task submissions,
206	>	* replacement of terminated workers, and/or compensation for
207	>	* blocked workers. However, all other support code is set up to
208	>	* work with other policies.
209	>	*
210	>	* To ensure that we do not hold on to worker references that
211	>	* would prevent GC, ALL accesses to workers are via indices into
212	>	* the workers array (which is one source of some of the unusual
213	>	* code constructions here). In essence, the workers array serves
214	>	* as a WeakReference mechanism. Thus for example the event queue
215	>	* stores worker indices, not worker references. Access to the
216	>	* workers in associated methods (for example releaseEventWaiters)
217	>	* must both index-check and null-check the IDs. All such accesses
218	>	* ignore bad IDs by returning out early from what they are doing,
219	>	* since this can only be associated with shutdown, in which case
220	>	* it is OK to give up. On termination, we just clobber these
221	>	* data structures without trying to use them.
222	>	*
223	>	* 2. Bookkeeping for dynamically adding and removing workers. We
224	>	* aim to approximately maintain the given level of parallelism.
225	>	* When some workers are known to be blocked (on joins or via
226	>	* ManagedBlocker), we may create or resume others to take their
227	>	* place until they unblock (see below). Implementing this
228	>	* requires counts of the number of "running" threads (i.e., those
229	>	* that are neither blocked nor artificially suspended) as well as
230	>	* the total number.  These two values are packed into one field,
231	>	* "workerCounts" because we need accurate snapshots when deciding
232	>	* to create, resume or suspend.  Note however that the
233	>	* correspondence of these counts to reality is not guaranteed. In
234	>	* particular updates for unblocked threads may lag until they
235	>	* actually wake up.
236	>	*
237	>	* 3. Maintaining global run state. The run state of the pool
238	>	* consists of a runLevel (SHUTDOWN, TERMINATING, etc) similar to
239	>	* those in other Executor implementations, as well as a count of
240	>	* "active" workers -- those that are, or soon will be, or
241	>	* recently were executing tasks. The runLevel and active count
242	>	* are packed together in order to correctly trigger shutdown and
243	>	* termination. Without care, active counts can be subject to very
244	>	* high contention.  We substantially reduce this contention by
245	>	* relaxing update rules.  A worker must claim active status
246	>	* prospectively, by activating if it sees that a submitted or
247	>	* stealable task exists (it may find after activating that the
248	>	* task no longer exists). It stays active while processing this
249	>	* task (if it exists) and any other local subtasks it produces,
250	>	* until it cannot find any other tasks. It then tries
251	>	* inactivating (see method preStep), but upon update contention
252	>	* instead scans for more tasks, later retrying inactivation if it
253	>	* doesn't find any.
254	>	*
255	>	* 4. Managing idle workers waiting for tasks. We cannot let
256	>	* workers spin indefinitely scanning for tasks when none are
257	>	* available. On the other hand, we must quickly prod them into
258	>	* action when new tasks are submitted or generated.  We
259	>	* park/unpark these idle workers using an event-count scheme.
260	>	* Field eventCount is incremented upon events that may enable
261	>	* workers that previously could not find a task to now find one:
262	>	* Submission of a new task to the pool, or another worker pushing
263	>	* a task onto a previously empty queue.  (We also use this
264	>	* mechanism for configuration and termination actions that
265	>	* require wakeups of idle workers).  Each worker maintains its
266	>	* last known event count, and blocks when a scan for work did not
267	>	* find a task AND its lastEventCount matches the current
268	>	* eventCount. Waiting idle workers are recorded in a variant of
269	>	* Treiber stack headed by field eventWaiters which, when nonzero,
270	>	* encodes the thread index and count awaited for by the worker
271	>	* thread most recently calling eventSync. This thread in turn has
272	>	* a record (field nextEventWaiter) for the next waiting worker.
273	>	* In addition to allowing simpler decisions about need for
274	>	* wakeup, the event count bits in eventWaiters serve the role of
275	>	* tags to avoid ABA errors in Treiber stacks. Upon any wakeup,
276	>	* released threads also try to release at most two others.  The
277	>	* net effect is a tree-like diffusion of signals, where released
278	>	* threads (and possibly others) help with unparks.  To further
279	>	* reduce contention effects a bit, failed CASes to increment
280	>	* field eventCount are tolerated without retries in signalWork.
281	>	* Conceptually they are merged into the same event, which is OK
282	>	* when their only purpose is to enable workers to scan for work.
283	>	*
284	>	* 5. Managing suspension of extra workers. When a worker notices
285	>	* (usually upon timeout of a wait()) that there are too few
286	>	* running threads, we may create a new thread to maintain
287	>	* parallelism level, or at least avoid starvation. Usually, extra
288	>	* threads are needed for only very short periods, yet join
289	>	* dependencies are such that we sometimes need them in
290	>	* bursts. Rather than create new threads each time this happens,
291	>	* we suspend no-longer-needed extra ones as "spares". For most
292	>	* purposes, we don't distinguish "extra" spare threads from
293	>	* normal "core" threads: On each call to preStep (the only point
294	>	* at which we can do this) a worker checks to see if there are
295	>	* now too many running workers, and if so, suspends itself.
296	>	* Method helpMaintainParallelism looks for suspended threads to
297	>	* resume before considering creating a new replacement. The
298	>	* spares themselves are encoded on another variant of a Treiber
299	>	* Stack, headed at field "spareWaiters".  Note that the use of
300	>	* spares is intrinsically racy.  One thread may become a spare at
301	>	* about the same time as another is needlessly being created. We
302	>	* counteract this and related slop in part by requiring resumed
303	>	* spares to immediately recheck (in preStep) to see whether they
304	>	* they should re-suspend.
305	>	*
306	>	* 6. Killing off unneeded workers. A timeout mechanism is used to
307	>	* shed unused workers: The oldest (first) event queue waiter uses
308	>	* a timed rather than hard wait. When this wait times out without
309	>	* a normal wakeup, it tries to shutdown any one (for convenience
310	>	* the newest) other spare or event waiter via
311	>	* tryShutdownUnusedWorker. This eventually reduces the number of
312	>	* worker threads to a minimum of one after a long enough period
313	>	* without use.
314	>	*
315	>	* 7. Deciding when to create new workers. The main dynamic
316	>	* control in this class is deciding when to create extra threads
317	>	* in method helpMaintainParallelism. We would like to keep
318	>	* exactly #parallelism threads running, which is an impossible
319	>	* task. We always need to create one when the number of running
320	>	* threads would become zero and all workers are busy. Beyond
321	>	* this, we must rely on heuristics that work well in the the
322	>	* presence of transients phenomena such as GC stalls, dynamic
323	>	* compilation, and wake-up lags. These transients are extremely
324	>	* common -- we are normally trying to fully saturate the CPUs on
325	>	* a machine, so almost any activity other than running tasks
326	>	* impedes accuracy. Our main defense is to allow parallelism to
327	>	* lapse for a while during joins, and use a timeout to see if,
328	>	* after the resulting settling, there is still a need for
329	>	* additional workers.  This also better copes with the fact that
330	>	* some of the methods in this class tend to never become compiled
331	>	* (but are interpreted), so some components of the entire set of
332	>	* controls might execute 100 times faster than others. And
333	>	* similarly for cases where the apparent lack of work is just due
334	>	* to GC stalls and other transient system activity.
335	>	*
336	>	* Beware that there is a lot of representation-level coupling
337	>	* among classes ForkJoinPool, ForkJoinWorkerThread, and
338	>	* ForkJoinTask.  For example, direct access to "workers" array by
339	>	* workers, and direct access to ForkJoinTask.status by both
340	>	* ForkJoinPool and ForkJoinWorkerThread.  There is little point
341	>	* trying to reduce this, since any associated future changes in
342	>	* representations will need to be accompanied by algorithmic
343	>	* changes anyway.
344	>	*
345	>	* Style notes: There are lots of inline assignments (of form
346	>	* "while ((local = field) != 0)") which are usually the simplest
347	>	* way to ensure the required read orderings (which are sometimes
348	>	* critical). Also several occurrences of the unusual "do {}
349	>	* while(!cas...)" which is the simplest way to force an update of
350	>	* a CAS'ed variable. There are also other coding oddities that
351	>	* help some methods perform reasonably even when interpreted (not
352	>	* compiled), at the expense of some messy constructions that
353	>	* reduce byte code counts.
354	>	*
355	>	* The order of declarations in this file is: (1) statics (2)
356	>	* fields (along with constants used when unpacking some of them)
357	>	* (3) internal control methods (4) callbacks and other support
358	>	* for ForkJoinTask and ForkJoinWorkerThread classes, (5) exported
359	>	* methods (plus a few little helpers).
360		*/
361
78	–	/** Mask for packing and unpacking shorts */
79	–	private static final int  shortMask = 0xffff;
80	–
81	–	/** Max pool size -- must be a power of two minus 1 */
82	–	private static final int MAX_THREADS =  0x7FFF;
83	–
362		/**
363		* Factory for creating new {@link ForkJoinWorkerThread}s.
364		* A {@code ForkJoinWorkerThreadFactory} must be defined and used
#	Line 92 | Line 370 | public class ForkJoinPool extends Abstra
370		* Returns a new worker thread operating in the given pool.
371		*
372		* @param pool the pool this thread works in
373	<	* @throws NullPointerException if pool is null
373	>	* @throws NullPointerException if the pool is null
374		*/
375		public ForkJoinWorkerThread newThread(ForkJoinPool pool);
376		}
#	Line 101 | Line 379 | public class ForkJoinPool extends Abstra
379		* Default ForkJoinWorkerThreadFactory implementation; creates a
380		* new ForkJoinWorkerThread.
381		*/
382	<	static class  DefaultForkJoinWorkerThreadFactory
382	>	static class DefaultForkJoinWorkerThreadFactory
383		implements ForkJoinWorkerThreadFactory {
384		public ForkJoinWorkerThread newThread(ForkJoinPool pool) {
385	<	try {
108	<	return new ForkJoinWorkerThread(pool);
109	<	} catch (OutOfMemoryError oom)  {
110	<	return null;
111	<	}
385	>	return new ForkJoinWorkerThread(pool);
386		}
387		}
388
#	Line 144 | Line 418 | public class ForkJoinPool extends Abstra
418		new AtomicInteger();
419
420		/**
421	<	* Array holding all worker threads in the pool. Initialized upon
422	<	* first use. Array size must be a power of two.  Updates and
423	<	* replacements are protected by workerLock, but it is always kept
424	<	* in a consistent enough state to be randomly accessed without
425	<	* locking by workers performing work-stealing.
421	>	* The time to block in a join (see awaitJoin) before checking if
422	>	* a new worker should be (re)started to maintain parallelism
423	>	* level. The value should be short enough to maintain global
424	>	* responsiveness and progress but long enough to avoid
425	>	* counterproductive firings during GC stalls or unrelated system
426	>	* activity, and to not bog down systems with continual re-firings
427	>	* on GCs or legitimately long waits.
428		*/
429	<	volatile ForkJoinWorkerThread[] workers;
429	>	private static final long JOIN_TIMEOUT_MILLIS = 250L; // 4 per second
430
431		/**
432	<	* Lock protecting access to workers.
432	>	* The wakeup interval (in nanoseconds) for the oldest worker
433	>	* worker waiting for an event invokes tryShutdownUnusedWorker to shrink
434	>	* the number of workers.  The exact value does not matter too
435	>	* much, but should be long enough to slowly release resources
436	>	* during long periods without use without disrupting normal use.
437		*/
438	<	private final ReentrantLock workerLock;
438	>	private static final long SHRINK_RATE_NANOS =
439	>	30L * 1000L * 1000L * 1000L; // 2 per minute
440
441		/**
442	<	* Condition for awaitTermination.
442	>	* Absolute bound for parallelism level. Twice this number plus
443	>	* one (i.e., 0xfff) must fit into a 16bit field to enable
444	>	* word-packing for some counts and indices.
445		*/
446	<	private final Condition termination;
446	>	private static final int MAX_WORKERS   = 0x7fff;
447
448		/**
449	<	* The uncaught exception handler used when any worker
450	<	* abruptly terminates
449	>	* Array holding all worker threads in the pool.  Array size must
450	>	* be a power of two.  Updates and replacements are protected by
451	>	* workerLock, but the array is always kept in a consistent enough
452	>	* state to be randomly accessed without locking by workers
453	>	* performing work-stealing, as well as other traversal-based
454	>	* methods in this class. All readers must tolerate that some
455	>	* array slots may be null.
456		*/
457	<	private Thread.UncaughtExceptionHandler ueh;
457	>	volatile ForkJoinWorkerThread[] workers;
458
459		/**
460	<	* Creation factory for worker threads.
460	>	* Queue for external submissions.
461		*/
462	<	private final ForkJoinWorkerThreadFactory factory;
462	>	private final LinkedTransferQueue<ForkJoinTask<?>> submissionQueue;
463
464		/**
465	<	* Head of stack of threads that were created to maintain
178	<	* parallelism when other threads blocked, but have since
179	<	* suspended when the parallelism level rose.
465	>	* Lock protecting updates to workers array.
466		*/
467	<	private volatile WaitQueueNode spareStack;
467	>	private final ReentrantLock workerLock;
468
469		/**
470	<	* Sum of per-thread steal counts, updated only when threads are
185	<	* idle or terminating.
470	>	* Latch released upon termination.
471		*/
472	<	private final AtomicLong stealCount;
472	>	private final Phaser termination;
473
474		/**
475	<	* Queue for external submissions.
475	>	* Creation factory for worker threads.
476		*/
477	<	private final LinkedTransferQueue<ForkJoinTask<?>> submissionQueue;
477	>	private final ForkJoinWorkerThreadFactory factory;
478
479		/**
480	<	* Head of Treiber stack for barrier sync. See below for explanation.
480	>	* Sum of per-thread steal counts, updated only when threads are
481	>	* idle or terminating.
482		*/
483	<	private volatile WaitQueueNode syncStack;
483	>	private volatile long stealCount;
484
485		/**
486	<	* The count for event barrier
486	>	* Encoded record of top of Treiber stack of threads waiting for
487	>	* events. The top 32 bits contain the count being waited for. The
488	>	* bottom 16 bits contains one plus the pool index of waiting
489	>	* worker thread. (Bits 16-31 are unused.)
490		*/
491	<	private volatile long eventCount;
491	>	private volatile long eventWaiters;
492
493	<	/**
494	<	* Pool number, just for assigning useful names to worker threads
206	<	*/
207	<	private final int poolNumber;
493	>	private static final int  EVENT_COUNT_SHIFT = 32;
494	>	private static final long WAITER_ID_MASK    = (1L << 16) - 1L;
495
496		/**
497	<	* The maximum allowed pool size
497	>	* A counter for events that may wake up worker threads:
498	>	*   - Submission of a new task to the pool
499	>	*   - A worker pushing a task on an empty queue
500	>	*   - termination
501		*/
502	<	private volatile int maxPoolSize;
502	>	private volatile int eventCount;
503
504		/**
505	<	* The desired parallelism level, updated only under workerLock.
505	>	* Encoded record of top of Treiber stack of spare threads waiting
506	>	* for resumption. The top 16 bits contain an arbitrary count to
507	>	* avoid ABA effects. The bottom 16bits contains one plus the pool
508	>	* index of waiting worker thread.
509		*/
510	<	private volatile int parallelism;
510	>	private volatile int spareWaiters;
511	>
512	>	private static final int SPARE_COUNT_SHIFT = 16;
513	>	private static final int SPARE_ID_MASK     = (1 << 16) - 1;
514
515		/**
516	<	* True if use local fifo, not default lifo, for local polling
516	>	* Lifecycle control. The low word contains the number of workers
517	>	* that are (probably) executing tasks. This value is atomically
518	>	* incremented before a worker gets a task to run, and decremented
519	>	* when worker has no tasks and cannot find any.  Bits 16-18
520	>	* contain runLevel value. When all are zero, the pool is
521	>	* running. Level transitions are monotonic (running -> shutdown
522	>	* -> terminating -> terminated) so each transition adds a bit.
523	>	* These are bundled together to ensure consistent read for
524	>	* termination checks (i.e., that runLevel is at least SHUTDOWN
525	>	* and active threads is zero).
526	>	*
527	>	* Notes: Most direct CASes are dependent on these bitfield
528	>	* positions.  Also, this field is non-private to enable direct
529	>	* performance-sensitive CASes in ForkJoinWorkerThread.
530		*/
531	<	private volatile boolean locallyFifo;
531	>	volatile int runState;
532	>
533	>	// Note: The order among run level values matters.
534	>	private static final int RUNLEVEL_SHIFT     = 16;
535	>	private static final int SHUTDOWN           = 1 << RUNLEVEL_SHIFT;
536	>	private static final int TERMINATING        = 1 << (RUNLEVEL_SHIFT + 1);
537	>	private static final int TERMINATED         = 1 << (RUNLEVEL_SHIFT + 2);
538	>	private static final int ACTIVE_COUNT_MASK  = (1 << RUNLEVEL_SHIFT) - 1;
539
540		/**
541		* Holds number of total (i.e., created and not yet terminated)
542		* and running (i.e., not blocked on joins or other managed sync)
543	<	* threads, packed into one int to ensure consistent snapshot when
543	>	* threads, packed together to ensure consistent snapshot when
544		* making decisions about creating and suspending spare
545	<	* threads. Updated only by CAS.  Note: CASes in
546	<	* updateRunningCount and preJoin assume that running active count
547	<	* is in low word, so need to be modified if this changes.
545	>	* threads. Updated only by CAS. Note that adding a new worker
546	>	* requires incrementing both counts, since workers start off in
547	>	* running state.
548		*/
549		private volatile int workerCounts;
550
551	<	private static int totalCountOf(int s)           { return s >>> 16;  }
552	<	private static int runningCountOf(int s)         { return s & shortMask; }
553	<	private static int workerCountsFor(int t, int r) { return (t << 16) + r; }
551	>	private static final int TOTAL_COUNT_SHIFT  = 16;
552	>	private static final int RUNNING_COUNT_MASK = (1 << TOTAL_COUNT_SHIFT) - 1;
553	>	private static final int ONE_RUNNING        = 1;
554	>	private static final int ONE_TOTAL          = 1 << TOTAL_COUNT_SHIFT;
555
556		/**
557	<	* Adds delta (which may be negative) to running count.  This must
558	<	* be called before (with negative arg) and after (with positive)
242	<	* any managed synchronization (i.e., mainly, joins).
243	<	*
244	<	* @param delta the number to add
557	>	* The target parallelism level.
558	>	* Accessed directly by ForkJoinWorkerThreads.
559		*/
560	<	final void updateRunningCount(int delta) {
247	<	int s;
248	<	do {} while (!casWorkerCounts(s = workerCounts, s + delta));
249	<	}
560	>	final int parallelism;
561
562		/**
563	<	* Adds delta (which may be negative) to both total and running
564	<	* count.  This must be called upon creation and termination of
254	<	* worker threads.
255	<	*
256	<	* @param delta the number to add
563	>	* True if use local fifo, not default lifo, for local polling
564	>	* Read by, and replicated by ForkJoinWorkerThreads
565		*/
566	<	private void updateWorkerCount(int delta) {
259	<	int d = delta + (delta << 16); // add to both lo and hi parts
260	<	int s;
261	<	do {} while (!casWorkerCounts(s = workerCounts, s + d));
262	<	}
566	>	final boolean locallyFifo;
567
568		/**
569	<	* Lifecycle control. High word contains runState, low word
570	<	* contains the number of workers that are (probably) executing
267	<	* tasks. This value is atomically incremented before a worker
268	<	* gets a task to run, and decremented when worker has no tasks
269	<	* and cannot find any. These two fields are bundled together to
270	<	* support correct termination triggering.  Note: activeCount
271	<	* CAS'es cheat by assuming active count is in low word, so need
272	<	* to be modified if this changes
569	>	* The uncaught exception handler used when any worker abruptly
570	>	* terminates.
571		*/
572	<	private volatile int runControl;
572	>	private final Thread.UncaughtExceptionHandler ueh;
573
574	<	// RunState values. Order among values matters
575	<	private static final int RUNNING     = 0;
576	<	private static final int SHUTDOWN    = 1;
577	<	private static final int TERMINATING = 2;
280	<	private static final int TERMINATED  = 3;
574	>	/**
575	>	* Pool number, just for assigning useful names to worker threads
576	>	*/
577	>	private final int poolNumber;
578
579	<	private static int runStateOf(int c)             { return c >>> 16; }
580	<	private static int activeCountOf(int c)          { return c & shortMask; }
284	<	private static int runControlFor(int r, int a)   { return (r << 16) + a; }
579	>	// Utilities for CASing fields. Note that most of these
580	>	// are usually manually inlined by callers
581
582		/**
583	<	* Tries incrementing active count; fails on contention.
288	<	* Called by workers before/during executing tasks.
289	<	*
290	<	* @return true on success
583	>	* Increments running count part of workerCounts
584		*/
585	<	final boolean tryIncrementActiveCount() {
586	<	int c = runControl;
587	<	return casRunControl(c, c+1);
585	>	final void incrementRunningCount() {
586	>	int c;
587	>	do {} while (!UNSAFE.compareAndSwapInt(this, workerCountsOffset,
588	>	c = workerCounts,
589	>	c + ONE_RUNNING));
590		}
591
592		/**
593	<	* Tries decrementing active count; fails on contention.
299	<	* Possibly triggers termination on success.
300	<	* Called by workers when they can't find tasks.
301	<	*
302	<	* @return true on success
593	>	* Tries to decrement running count unless already zero
594		*/
595	<	final boolean tryDecrementActiveCount() {
596	<	int c = runControl;
597	<	int nextc = c - 1;
307	<	if (!casRunControl(c, nextc))
595	>	final boolean tryDecrementRunningCount() {
596	>	int wc = workerCounts;
597	>	if ((wc & RUNNING_COUNT_MASK) == 0)
598		return false;
599	<	if (canTerminateOnShutdown(nextc))
600	<	terminateOnShutdown();
311	<	return true;
599	>	return UNSAFE.compareAndSwapInt(this, workerCountsOffset,
600	>	wc, wc - ONE_RUNNING);
601		}
602
603		/**
604	<	* Returns {@code true} if argument represents zero active count
605	<	* and nonzero runstate, which is the triggering condition for
606	<	* terminating on shutdown.
604	>	* Forces decrement of encoded workerCounts, awaiting nonzero if
605	>	* (rarely) necessary when other count updates lag.
606	>	*
607	>	* @param dr -- either zero or ONE_RUNNING
608	>	* @param dt == either zero or ONE_TOTAL
609		*/
610	<	private static boolean canTerminateOnShutdown(int c) {
611	<	// i.e. least bit is nonzero runState bit
612	<	return ((c & -c) >>> 16) != 0;
610	>	private void decrementWorkerCounts(int dr, int dt) {
611	>	for (;;) {
612	>	int wc = workerCounts;
613	>	if ((wc & RUNNING_COUNT_MASK)  - dr < 0 ||
614	>	(wc >>> TOTAL_COUNT_SHIFT) - dt < 0) {
615	>	if ((runState & TERMINATED) != 0)
616	>	return; // lagging termination on a backout
617	>	Thread.yield();
618	>	}
619	>	if (UNSAFE.compareAndSwapInt(this, workerCountsOffset,
620	>	wc, wc - (dr + dt)))
621	>	return;
622	>	}
623	>	}
624	>
625	>	/**
626	>	* Tries decrementing active count; fails on contention.
627	>	* Called when workers cannot find tasks to run.
628	>	*/
629	>	final boolean tryDecrementActiveCount() {
630	>	int c;
631	>	return UNSAFE.compareAndSwapInt(this, runStateOffset,
632	>	c = runState, c - 1);
633		}
634
635		/**
636	<	* Transition run state to at least the given state. Return true
637	<	* if not already at least given state.
636	>	* Advances to at least the given level. Returns true if not
637	>	* already in at least the given level.
638		*/
639	<	private boolean transitionRunStateTo(int state) {
639	>	private boolean advanceRunLevel(int level) {
640		for (;;) {
641	<	int c = runControl;
642	<	if (runStateOf(c) >= state)
641	>	int s = runState;
642	>	if ((s & level) != 0)
643		return false;
644	<	if (casRunControl(c, runControlFor(state, activeCountOf(c))))
644	>	if (UNSAFE.compareAndSwapInt(this, runStateOffset, s, s | level))
645		return true;
646		}
647		}
648
649	+	// workers array maintenance
650	+
651		/**
652	<	* Controls whether to add spares to maintain parallelism
652	>	* Records and returns a workers array index for new worker.
653		*/
654	<	private volatile boolean maintainsParallelism;
654	>	private int recordWorker(ForkJoinWorkerThread w) {
655	>	// Try using slot totalCount-1. If not available, scan and/or resize
656	>	int k = (workerCounts >>> TOTAL_COUNT_SHIFT) - 1;
657	>	final ReentrantLock lock = this.workerLock;
658	>	lock.lock();
659	>	try {
660	>	ForkJoinWorkerThread[] ws = workers;
661	>	int n = ws.length;
662	>	if (k < 0 || k >= n || ws[k] != null) {
663	>	for (k = 0; k < n && ws[k] != null; ++k)
664	>	;
665	>	if (k == n)
666	>	ws = Arrays.copyOf(ws, n << 1);
667	>	}
668	>	ws[k] = w;
669	>	workers = ws; // volatile array write ensures slot visibility
670	>	} finally {
671	>	lock.unlock();
672	>	}
673	>	return k;
674	>	}
675
676	<	// Constructors
676	>	/**
677	>	* Nulls out record of worker in workers array
678	>	*/
679	>	private void forgetWorker(ForkJoinWorkerThread w) {
680	>	int idx = w.poolIndex;
681	>	// Locking helps method recordWorker avoid unnecessary expansion
682	>	final ReentrantLock lock = this.workerLock;
683	>	lock.lock();
684	>	try {
685	>	ForkJoinWorkerThread[] ws = workers;
686	>	if (idx >= 0 && idx < ws.length && ws[idx] == w) // verify
687	>	ws[idx] = null;
688	>	} finally {
689	>	lock.unlock();
690	>	}
691	>	}
692
693		/**
694	<	* Creates a {@code ForkJoinPool} with a pool size equal to the
695	<	* number of processors available on the system, using the
696	<	* {@linkplain #defaultForkJoinWorkerThreadFactory default thread factory}.
694	>	* Final callback from terminating worker.  Removes record of
695	>	* worker from array, and adjusts counts. If pool is shutting
696	>	* down, tries to complete termination.
697		*
698	<	* @throws SecurityException if a security manager exists and
351	<	*         the caller is not permitted to modify threads
352	<	*         because it does not hold {@link
353	<	*         java.lang.RuntimePermission}{@code ("modifyThread")}
698	>	* @param w the worker
699		*/
700	<	public ForkJoinPool() {
701	<	this(Runtime.getRuntime().availableProcessors(),
702	<	defaultForkJoinWorkerThreadFactory);
700	>	final void workerTerminated(ForkJoinWorkerThread w) {
701	>	forgetWorker(w);
702	>	decrementWorkerCounts(w.isTrimmed()? 0 : ONE_RUNNING, ONE_TOTAL);
703	>	while (w.stealCount != 0) // collect final count
704	>	tryAccumulateStealCount(w);
705	>	tryTerminate(false);
706		}
707
708	+	// Waiting for and signalling events
709	+
710		/**
711	<	* Creates a {@code ForkJoinPool} with the indicated parallelism level
712	<	* threads and using the
713	<	* {@linkplain #defaultForkJoinWorkerThreadFactory default thread factory}.
714	<	*
365	<	* @param parallelism the number of worker threads
366	<	* @throws IllegalArgumentException if parallelism less than or
367	<	* equal to zero
368	<	* @throws SecurityException if a security manager exists and
369	<	*         the caller is not permitted to modify threads
370	<	*         because it does not hold {@link
371	<	*         java.lang.RuntimePermission}{@code ("modifyThread")}
711	>	* Releases workers blocked on a count not equal to current count.
712	>	* Normally called after precheck that eventWaiters isn't zero to
713	>	* avoid wasted array checks. Gives up upon a change in count or
714	>	* upon releasing two workers, letting others take over.
715		*/
716	<	public ForkJoinPool(int parallelism) {
717	<	this(parallelism, defaultForkJoinWorkerThreadFactory);
716	>	private void releaseEventWaiters() {
717	>	ForkJoinWorkerThread[] ws = workers;
718	>	int n = ws.length;
719	>	long h = eventWaiters;
720	>	int ec = eventCount;
721	>	boolean releasedOne = false;
722	>	ForkJoinWorkerThread w; int id;
723	>	while ((id = ((int)(h & WAITER_ID_MASK)) - 1) >= 0 &&
724	>	(int)(h >>> EVENT_COUNT_SHIFT) != ec &&
725	>	id < n && (w = ws[id]) != null) {
726	>	if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
727	>	h,  w.nextWaiter)) {
728	>	LockSupport.unpark(w);
729	>	if (releasedOne) // exit on second release
730	>	break;
731	>	releasedOne = true;
732	>	}
733	>	if (eventCount != ec)
734	>	break;
735	>	h = eventWaiters;
736	>	}
737		}
738
739		/**
740	<	* Creates a {@code ForkJoinPool} with parallelism equal to the
741	<	* number of processors available on the system and using the
380	<	* given thread factory.
381	<	*
382	<	* @param factory the factory for creating new threads
383	<	* @throws NullPointerException if factory is null
384	<	* @throws SecurityException if a security manager exists and
385	<	*         the caller is not permitted to modify threads
386	<	*         because it does not hold {@link
387	<	*         java.lang.RuntimePermission}{@code ("modifyThread")}
740	>	* Tries to advance eventCount and releases waiters. Called only
741	>	* from workers.
742		*/
743	<	public ForkJoinPool(ForkJoinWorkerThreadFactory factory) {
744	<	this(Runtime.getRuntime().availableProcessors(), factory);
743	>	final void signalWork() {
744	>	int c; // try to increment event count -- CAS failure OK
745	>	UNSAFE.compareAndSwapInt(this, eventCountOffset, c = eventCount, c+1);
746	>	if (eventWaiters != 0L)
747	>	releaseEventWaiters();
748		}
749
750		/**
751	<	* Creates a {@code ForkJoinPool} with the given parallelism and
752	<	* thread factory.
751	>	* Adds the given worker to event queue and blocks until
752	>	* terminating or event count advances from the given value
753		*
754	<	* @param parallelism the targeted number of worker threads
755	<	* @param factory the factory for creating new threads
399	<	* @throws IllegalArgumentException if parallelism less than or
400	<	* equal to zero, or greater than implementation limit
401	<	* @throws NullPointerException if factory is null
402	<	* @throws SecurityException if a security manager exists and
403	<	*         the caller is not permitted to modify threads
404	<	*         because it does not hold {@link
405	<	*         java.lang.RuntimePermission}{@code ("modifyThread")}
754	>	* @param w the calling worker thread
755	>	* @param ec the count
756		*/
757	<	public ForkJoinPool(int parallelism, ForkJoinWorkerThreadFactory factory) {
758	<	if (parallelism <= 0 || parallelism > MAX_THREADS)
759	<	throw new IllegalArgumentException();
760	<	if (factory == null)
761	<	throw new NullPointerException();
762	<	checkPermission();
763	<	this.factory = factory;
764	<	this.parallelism = parallelism;
765	<	this.maxPoolSize = MAX_THREADS;
766	<	this.maintainsParallelism = true;
767	<	this.poolNumber = poolNumberGenerator.incrementAndGet();
768	<	this.workerLock = new ReentrantLock();
769	<	this.termination = workerLock.newCondition();
420	<	this.stealCount = new AtomicLong();
421	<	this.submissionQueue = new LinkedTransferQueue<ForkJoinTask<?>>();
422	<	// worker array and workers are lazily constructed
757	>	private void eventSync(ForkJoinWorkerThread w, int ec) {
758	>	long nh = (((long)ec) << EVENT_COUNT_SHIFT) | ((long)(w.poolIndex+1));
759	>	long h;
760	>	while ((runState < SHUTDOWN || !tryTerminate(false)) &&
761	>	(((int)((h = eventWaiters) & WAITER_ID_MASK)) == 0 ||
762	>	(int)(h >>> EVENT_COUNT_SHIFT) == ec) &&
763	>	eventCount == ec) {
764	>	if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
765	>	w.nextWaiter = h, nh)) {
766	>	awaitEvent(w, ec);
767	>	break;
768	>	}
769	>	}
770		}
771
772		/**
773	<	* Creates a new worker thread using factory.
773	>	* Blocks the given worker (that has already been entered as an
774	>	* event waiter) until terminating or event count advances from
775	>	* the given value. The oldest (first) waiter uses a timed wait to
776	>	* occasionally one-by-one shrink the number of workers (to a
777	>	* minimum of one) if the pool has not been used for extended
778	>	* periods.
779		*
780	<	* @param index the index to assign worker
781	<	* @return new worker, or null if factory failed
780	>	* @param w the calling worker thread
781	>	* @param ec the count
782		*/
783	<	private ForkJoinWorkerThread createWorker(int index) {
784	<	Thread.UncaughtExceptionHandler h = ueh;
785	<	ForkJoinWorkerThread w = factory.newThread(this);
786	<	if (w != null) {
787	<	w.poolIndex = index;
788	<	w.setDaemon(true);
789	<	w.setAsyncMode(locallyFifo);
790	<	w.setName("ForkJoinPool-" + poolNumber + "-worker-" + index);
791	<	if (h != null)
792	<	w.setUncaughtExceptionHandler(h);
783	>	private void awaitEvent(ForkJoinWorkerThread w, int ec) {
784	>	while (eventCount == ec) {
785	>	if (tryAccumulateStealCount(w)) { // transfer while idle
786	>	boolean untimed = (w.nextWaiter != 0L ||
787	>	(workerCounts & RUNNING_COUNT_MASK) <= 1);
788	>	long startTime = untimed? 0 : System.nanoTime();
789	>	Thread.interrupted();         // clear/ignore interrupt
790	>	if (eventCount != ec || w.runState != 0 ||
791	>	runState >= TERMINATING)  // recheck after clear
792	>	break;
793	>	if (untimed)
794	>	LockSupport.park(w);
795	>	else {
796	>	LockSupport.parkNanos(w, SHRINK_RATE_NANOS);
797	>	if (eventCount != ec || w.runState != 0 ||
798	>	runState >= TERMINATING)
799	>	break;
800	>	if (System.nanoTime() - startTime >= SHRINK_RATE_NANOS)
801	>	tryShutdownUnusedWorker(ec);
802	>	}
803	>	}
804		}
442	–	return w;
805		}
806
807	+	// Maintaining parallelism
808	+
809		/**
810	<	* Returns a good size for worker array given pool size.
447	<	* Currently requires size to be a power of two.
810	>	* Pushes worker onto the spare stack
811		*/
812	<	private static int arraySizeFor(int poolSize) {
813	<	return (poolSize <= 1) ? 1 :
814	<	(1 << (32 - Integer.numberOfLeadingZeros(poolSize-1)));
812	>	final void pushSpare(ForkJoinWorkerThread w) {
813	>	int ns = (++w.spareCount << SPARE_COUNT_SHIFT) | (w.poolIndex + 1);
814	>	do {} while (!UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
815	>	w.nextSpare = spareWaiters,ns));
816		}
817
818		/**
819	<	* Creates or resizes array if necessary to hold newLength.
820	<	* Call only under exclusion.
457	<	*
458	<	* @return the array
819	>	* Tries (once) to resume a spare if the number of running
820	>	* threads is less than target.
821		*/
822	<	private ForkJoinWorkerThread[] ensureWorkerArrayCapacity(int newLength) {
822	>	private void tryResumeSpare() {
823	>	int sw, id;
824		ForkJoinWorkerThread[] ws = workers;
825	<	if (ws == null)
826	<	return workers = new ForkJoinWorkerThread[arraySizeFor(newLength)];
827	<	else if (newLength > ws.length)
828	<	return workers = Arrays.copyOf(ws, arraySizeFor(newLength));
829	<	else
830	<	return ws;
825	>	int n = ws.length;
826	>	ForkJoinWorkerThread w;
827	>	if ((sw = spareWaiters) != 0 &&
828	>	(id = (sw & SPARE_ID_MASK) - 1) >= 0 &&
829	>	id < n && (w = ws[id]) != null &&
830	>	(workerCounts & RUNNING_COUNT_MASK) < parallelism &&
831	>	spareWaiters == sw &&
832	>	UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
833	>	sw, w.nextSpare)) {
834	>	int c; // increment running count before resume
835	>	do {} while(!UNSAFE.compareAndSwapInt
836	>	(this, workerCountsOffset,
837	>	c = workerCounts, c + ONE_RUNNING));
838	>	if (w.tryUnsuspend())
839	>	LockSupport.unpark(w);
840	>	else   // back out if w was shutdown
841	>	decrementWorkerCounts(ONE_RUNNING, 0);
842	>	}
843		}
844
845		/**
846	<	* Tries to shrink workers into smaller array after one or more terminate.
847	<	*/
848	<	private void tryShrinkWorkerArray() {
849	<	ForkJoinWorkerThread[] ws = workers;
850	<	if (ws != null) {
851	<	int len = ws.length;
852	<	int last = len - 1;
853	<	while (last >= 0 && ws[last] == null)
854	<	--last;
855	<	int newLength = arraySizeFor(last+1);
856	<	if (newLength < len)
857	<	workers = Arrays.copyOf(ws, newLength);
846	>	* Tries to increase the number of running workers if below target
847	>	* parallelism: If a spare exists tries to resume it via
848	>	* tryResumeSpare.  Otherwise, if not enough total workers or all
849	>	* existing workers are busy, adds a new worker. In all cases also
850	>	* helps wake up releasable workers waiting for work.
851	>	*/
852	>	private void helpMaintainParallelism() {
853	>	int pc = parallelism;
854	>	int wc, rs, tc;
855	>	while (((wc = workerCounts) & RUNNING_COUNT_MASK) < pc &&
856	>	(rs = runState) < TERMINATING) {
857	>	if (spareWaiters != 0)
858	>	tryResumeSpare();
859	>	else if ((tc = wc >>> TOTAL_COUNT_SHIFT) >= MAX_WORKERS ||
860	>	(tc >= pc && (rs & ACTIVE_COUNT_MASK) != tc))
861	>	break;   // enough total
862	>	else if (runState == rs && workerCounts == wc &&
863	>	UNSAFE.compareAndSwapInt(this, workerCountsOffset, wc,
864	>	wc + (ONE_RUNNING|ONE_TOTAL))) {
865	>	ForkJoinWorkerThread w = null;
866	>	try {
867	>	w = factory.newThread(this);
868	>	} finally { // adjust on null or exceptional factory return
869	>	if (w == null) {
870	>	decrementWorkerCounts(ONE_RUNNING, ONE_TOTAL);
871	>	tryTerminate(false); // handle failure during shutdown
872	>	}
873	>	}
874	>	if (w == null)
875	>	break;
876	>	w.start(recordWorker(w), ueh);
877	>	if ((workerCounts >>> TOTAL_COUNT_SHIFT) >= pc) {
878	>	int c; // advance event count
879	>	UNSAFE.compareAndSwapInt(this, eventCountOffset,
880	>	c = eventCount, c+1);
881	>	break; // add at most one unless total below target
882	>	}
883	>	}
884		}
885	+	if (eventWaiters != 0L)
886	+	releaseEventWaiters();
887		}
888
889		/**
890	<	* Initializes workers if necessary.
890	>	* Callback from the oldest waiter in awaitEvent waking up after a
891	>	* period of non-use. If all workers are idle, tries (once) to
892	>	* shutdown an event waiter or a spare, if one exists. Note that
893	>	* we don't need CAS or locks here because the method is called
894	>	* only from one thread occasionally waking (and even misfires are
895	>	* OK). Note that until the shutdown worker fully terminates,
896	>	* workerCounts will overestimate total count, which is tolerable.
897	>	*
898	>	* @param ec the event count waited on by caller (to abort
899	>	* attempt if count has since changed).
900		*/
901	<	final void ensureWorkerInitialization() {
902	<	ForkJoinWorkerThread[] ws = workers;
903	<	if (ws == null) {
904	<	final ReentrantLock lock = this.workerLock;
905	<	lock.lock();
906	<	try {
907	<	ws = workers;
908	<	if (ws == null) {
909	<	int ps = parallelism;
910	<	ws = ensureWorkerArrayCapacity(ps);
911	<	for (int i = 0; i < ps; ++i) {
912	<	ForkJoinWorkerThread w = createWorker(i);
913	<	if (w != null) {
914	<	ws[i] = w;
915	<	w.start();
916	<	updateWorkerCount(1);
917	<	}
918	<	}
901	>	private void tryShutdownUnusedWorker(int ec) {
902	>	if (runState == 0 && eventCount == ec) { // only trigger if all idle
903	>	ForkJoinWorkerThread[] ws = workers;
904	>	int n = ws.length;
905	>	ForkJoinWorkerThread w = null;
906	>	boolean shutdown = false;
907	>	int sw;
908	>	long h;
909	>	if ((sw = spareWaiters) != 0) { // prefer killing spares
910	>	int id = (sw & SPARE_ID_MASK) - 1;
911	>	if (id >= 0 && id < n && (w = ws[id]) != null &&
912	>	UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
913	>	sw, w.nextSpare))
914	>	shutdown = true;
915	>	}
916	>	else if ((h = eventWaiters) != 0L) {
917	>	long nh;
918	>	int id = ((int)(h & WAITER_ID_MASK)) - 1;
919	>	if (id >= 0 && id < n && (w = ws[id]) != null &&
920	>	(nh = w.nextWaiter) != 0L && // keep at least one worker
921	>	UNSAFE.compareAndSwapLong(this, eventWaitersOffset, h, nh))
922	>	shutdown = true;
923	>	}
924	>	if (w != null && shutdown) {
925	>	w.shutdown();
926	>	LockSupport.unpark(w);
927	>	}
928	>	}
929	>	releaseEventWaiters(); // in case of interference
930	>	}
931	>
932	>	/**
933	>	* Callback from workers invoked upon each top-level action (i.e.,
934	>	* stealing a task or taking a submission and running it).
935	>	* Performs one or more of the following:
936	>	*
937	>	* 1. If the worker is active and either did not run a task
938	>	*    or there are too many workers, try to set its active status
939	>	*    to inactive and update activeCount. On contention, we may
940	>	*    try again in this or a subsequent call.
941	>	*
942	>	* 2. If not enough total workers, help create some.
943	>	*
944	>	* 3. If there are too many running workers, suspend this worker
945	>	*    (first forcing inactive if necessary).  If it is not needed,
946	>	*    it may be shutdown while suspended (via
947	>	*    tryShutdownUnusedWorker).  Otherwise, upon resume it
948	>	*    rechecks running thread count and need for event sync.
949	>	*
950	>	* 4. If worker did not run a task, await the next task event via
951	>	*    eventSync if necessary (first forcing inactivation), upon
952	>	*    which the worker may be shutdown via
953	>	*    tryShutdownUnusedWorker.  Otherwise, help release any
954	>	*    existing event waiters that are now releasable,
955	>	*
956	>	* @param w the worker
957	>	* @param ran true if worker ran a task since last call to this method
958	>	*/
959	>	final void preStep(ForkJoinWorkerThread w, boolean ran) {
960	>	int wec = w.lastEventCount;
961	>	boolean active = w.active;
962	>	boolean inactivate = false;
963	>	int pc = parallelism;
964	>	int rs;
965	>	while (w.runState == 0 && (rs = runState) < TERMINATING) {
966	>	if ((inactivate || (active && (rs & ACTIVE_COUNT_MASK) >= pc)) &&
967	>	UNSAFE.compareAndSwapInt(this, runStateOffset, rs, rs - 1))
968	>	inactivate = active = w.active = false;
969	>	int wc = workerCounts;
970	>	if ((wc & RUNNING_COUNT_MASK) > pc) {
971	>	if (!(inactivate |= active) && // must inactivate to suspend
972	>	workerCounts == wc &&      // try to suspend as spare
973	>	UNSAFE.compareAndSwapInt(this, workerCountsOffset,
974	>	wc, wc - ONE_RUNNING))
975	>	w.suspendAsSpare();
976	>	}
977	>	else if ((wc >>> TOTAL_COUNT_SHIFT) < pc)
978	>	helpMaintainParallelism();     // not enough workers
979	>	else if (!ran) {
980	>	long h = eventWaiters;
981	>	int ec = eventCount;
982	>	if (h != 0L && (int)(h >>> EVENT_COUNT_SHIFT) != ec)
983	>	releaseEventWaiters();     // release others before waiting
984	>	else if (ec != wec) {
985	>	w.lastEventCount = ec;     // no need to wait
986	>	break;
987		}
988	<	} finally {
989	<	lock.unlock();
988	>	else if (!(inactivate |= active))
989	>	eventSync(w, wec);         // must inactivate before sync
990		}
991	+	else
992	+	break;
993		}
994		}
995
996		/**
997	<	* Worker creation and startup for threads added via setParallelism.
997	>	* Helps and/or blocks awaiting join of the given task.
998	>	* See above for explanation.
999	>	*
1000	>	* @param joinMe the task to join
1001	>	* @param worker the current worker thread
1002		*/
1003	<	private void createAndStartAddedWorkers() {
1004	<	resumeAllSpares();  // Allow spares to convert to nonspare
1005	<	int ps = parallelism;
1006	<	ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(ps);
1007	<	int len = ws.length;
1008	<	// Sweep through slots, to keep lowest indices most populated
1009	<	int k = 0;
1010	<	while (k < len) {
1011	<	if (ws[k] != null) {
1012	<	++k;
1013	<	continue;
1003	>	final void awaitJoin(ForkJoinTask<?> joinMe, ForkJoinWorkerThread worker) {
1004	>	int retries = 2 + (parallelism >> 2); // #helpJoins before blocking
1005	>	while (joinMe.status >= 0) {
1006	>	int wc;
1007	>	worker.helpJoinTask(joinMe);
1008	>	if (joinMe.status < 0)
1009	>	break;
1010	>	else if (retries > 0)
1011	>	--retries;
1012	>	else if (((wc = workerCounts) & RUNNING_COUNT_MASK) != 0 &&
1013	>	UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1014	>	wc, wc - ONE_RUNNING)) {
1015	>	int stat, c; long h;
1016	>	while ((stat = joinMe.status) >= 0 &&
1017	>	(h = eventWaiters) != 0L && // help release others
1018	>	(int)(h >>> EVENT_COUNT_SHIFT) != eventCount)
1019	>	releaseEventWaiters();
1020	>	if (stat >= 0 &&
1021	>	((workerCounts & RUNNING_COUNT_MASK) == 0 ||
1022	>	(stat =
1023	>	joinMe.internalAwaitDone(JOIN_TIMEOUT_MILLIS)) >= 0))
1024	>	helpMaintainParallelism(); // timeout or no running workers
1025	>	do {} while (!UNSAFE.compareAndSwapInt
1026	>	(this, workerCountsOffset,
1027	>	c = workerCounts, c + ONE_RUNNING));
1028	>	if (stat < 0)
1029	>	break;   // else restart
1030		}
1031	<	int s = workerCounts;
1032	<	int tc = totalCountOf(s);
1033	<	int rc = runningCountOf(s);
1034	<	if (rc >= ps || tc >= ps)
1031	>	}
1032	>	}
1033	>
1034	>	/**
1035	>	* Same idea as awaitJoin, but no helping, retries, or timeouts.
1036	>	*/
1037	>	final void awaitBlocker(ManagedBlocker blocker)
1038	>	throws InterruptedException {
1039	>	while (!blocker.isReleasable()) {
1040	>	int wc = workerCounts;
1041	>	if ((wc & RUNNING_COUNT_MASK) != 0 &&
1042	>	UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1043	>	wc, wc - ONE_RUNNING)) {
1044	>	try {
1045	>	while (!blocker.isReleasable()) {
1046	>	long h = eventWaiters;
1047	>	if (h != 0L &&
1048	>	(int)(h >>> EVENT_COUNT_SHIFT) != eventCount)
1049	>	releaseEventWaiters();
1050	>	else if ((workerCounts & RUNNING_COUNT_MASK) == 0 &&
1051	>	runState < TERMINATING)
1052	>	helpMaintainParallelism();
1053	>	else if (blocker.block())
1054	>	break;
1055	>	}
1056	>	} finally {
1057	>	int c;
1058	>	do {} while (!UNSAFE.compareAndSwapInt
1059	>	(this, workerCountsOffset,
1060	>	c = workerCounts, c + ONE_RUNNING));
1061	>	}
1062		break;
1063	<	if (casWorkerCounts (s, workerCountsFor(tc+1, rc+1))) {
1064	<	ForkJoinWorkerThread w = createWorker(k);
1063	>	}
1064	>	}
1065	>	}
1066	>
1067	>	/**
1068	>	* Possibly initiates and/or completes termination.
1069	>	*
1070	>	* @param now if true, unconditionally terminate, else only
1071	>	* if shutdown and empty queue and no active workers
1072	>	* @return true if now terminating or terminated
1073	>	*/
1074	>	private boolean tryTerminate(boolean now) {
1075	>	if (now)
1076	>	advanceRunLevel(SHUTDOWN); // ensure at least SHUTDOWN
1077	>	else if (runState < SHUTDOWN ||
1078	>	!submissionQueue.isEmpty() ||
1079	>	(runState & ACTIVE_COUNT_MASK) != 0)
1080	>	return false;
1081	>
1082	>	if (advanceRunLevel(TERMINATING))
1083	>	startTerminating();
1084	>
1085	>	// Finish now if all threads terminated; else in some subsequent call
1086	>	if ((workerCounts >>> TOTAL_COUNT_SHIFT) == 0) {
1087	>	advanceRunLevel(TERMINATED);
1088	>	termination.arrive();
1089	>	}
1090	>	return true;
1091	>	}
1092	>
1093	>	/**
1094	>	* Actions on transition to TERMINATING
1095	>	*
1096	>	* Runs up to four passes through workers: (0) shutting down each
1097	>	* (without waking up if parked) to quickly spread notifications
1098	>	* without unnecessary bouncing around event queues etc (1) wake
1099	>	* up and help cancel tasks (2) interrupt (3) mop up races with
1100	>	* interrupted workers
1101	>	*/
1102	>	private void startTerminating() {
1103	>	cancelSubmissions();
1104	>	for (int passes = 0; passes < 4 && workerCounts != 0; ++passes) {
1105	>	int c; // advance event count
1106	>	UNSAFE.compareAndSwapInt(this, eventCountOffset,
1107	>	c = eventCount, c+1);
1108	>	eventWaiters = 0L; // clobber lists
1109	>	spareWaiters = 0;
1110	>	ForkJoinWorkerThread[] ws = workers;
1111	>	int n = ws.length;
1112	>	for (int i = 0; i < n; ++i) {
1113	>	ForkJoinWorkerThread w = ws[i];
1114		if (w != null) {
1115	<	ws[k++] = w;
1116	<	w.start();
1117	<	}
1118	<	else {
1119	<	updateWorkerCount(-1); // back out on failed creation
1120	<	break;
1115	>	w.shutdown();
1116	>	if (passes > 0 && !w.isTerminated()) {
1117	>	w.cancelTasks();
1118	>	LockSupport.unpark(w);
1119	>	if (passes > 1) {
1120	>	try {
1121	>	w.interrupt();
1122	>	} catch (SecurityException ignore) {
1123	>	}
1124	>	}
1125	>	}
1126		}
1127		}
1128		}
1129		}
1130
1131	+	/**
1132	+	* Clear out and cancel submissions, ignoring exceptions
1133	+	*/
1134	+	private void cancelSubmissions() {
1135	+	ForkJoinTask<?> task;
1136	+	while ((task = submissionQueue.poll()) != null) {
1137	+	try {
1138	+	task.cancel(false);
1139	+	} catch (Throwable ignore) {
1140	+	}
1141	+	}
1142	+	}
1143	+
1144	+	// misc support for ForkJoinWorkerThread
1145	+
1146	+	/**
1147	+	* Returns pool number
1148	+	*/
1149	+	final int getPoolNumber() {
1150	+	return poolNumber;
1151	+	}
1152	+
1153	+	/**
1154	+	* Tries to accumulates steal count from a worker, clearing
1155	+	* the worker's value.
1156	+	*
1157	+	* @return true if worker steal count now zero
1158	+	*/
1159	+	final boolean tryAccumulateStealCount(ForkJoinWorkerThread w) {
1160	+	int sc = w.stealCount;
1161	+	long c = stealCount;
1162	+	// CAS even if zero, for fence effects
1163	+	if (UNSAFE.compareAndSwapLong(this, stealCountOffset, c, c + sc)) {
1164	+	if (sc != 0)
1165	+	w.stealCount = 0;
1166	+	return true;
1167	+	}
1168	+	return sc == 0;
1169	+	}
1170	+
1171	+	/**
1172	+	* Returns the approximate (non-atomic) number of idle threads per
1173	+	* active thread.
1174	+	*/
1175	+	final int idlePerActive() {
1176	+	int pc = parallelism; // use parallelism, not rc
1177	+	int ac = runState;    // no mask -- artificially boosts during shutdown
1178	+	// Use exact results for small values, saturate past 4
1179	+	return pc <= ac? 0 : pc >>> 1 <= ac? 1 : pc >>> 2 <= ac? 3 : pc >>> 3;
1180	+	}
1181	+
1182	+	// Public and protected methods
1183	+
1184	+	// Constructors
1185	+
1186	+	/**
1187	+	* Creates a {@code ForkJoinPool} with parallelism equal to {@link
1188	+	* java.lang.Runtime#availableProcessors}, using the {@linkplain
1189	+	* #defaultForkJoinWorkerThreadFactory default thread factory},
1190	+	* no UncaughtExceptionHandler, and non-async LIFO processing mode.
1191	+	*
1192	+	* @throws SecurityException if a security manager exists and
1193	+	*         the caller is not permitted to modify threads
1194	+	*         because it does not hold {@link
1195	+	*         java.lang.RuntimePermission}{@code ("modifyThread")}
1196	+	*/
1197	+	public ForkJoinPool() {
1198	+	this(Runtime.getRuntime().availableProcessors(),
1199	+	defaultForkJoinWorkerThreadFactory, null, false);
1200	+	}
1201	+
1202	+	/**
1203	+	* Creates a {@code ForkJoinPool} with the indicated parallelism
1204	+	* level, the {@linkplain
1205	+	* #defaultForkJoinWorkerThreadFactory default thread factory},
1206	+	* no UncaughtExceptionHandler, and non-async LIFO processing mode.
1207	+	*
1208	+	* @param parallelism the parallelism level
1209	+	* @throws IllegalArgumentException if parallelism less than or
1210	+	*         equal to zero, or greater than implementation limit
1211	+	* @throws SecurityException if a security manager exists and
1212	+	*         the caller is not permitted to modify threads
1213	+	*         because it does not hold {@link
1214	+	*         java.lang.RuntimePermission}{@code ("modifyThread")}
1215	+	*/
1216	+	public ForkJoinPool(int parallelism) {
1217	+	this(parallelism, defaultForkJoinWorkerThreadFactory, null, false);
1218	+	}
1219	+
1220	+	/**
1221	+	* Creates a {@code ForkJoinPool} with the given parameters.
1222	+	*
1223	+	* @param parallelism the parallelism level. For default value,
1224	+	* use {@link java.lang.Runtime#availableProcessors}.
1225	+	* @param factory the factory for creating new threads. For default value,
1226	+	* use {@link #defaultForkJoinWorkerThreadFactory}.
1227	+	* @param handler the handler for internal worker threads that
1228	+	* terminate due to unrecoverable errors encountered while executing
1229	+	* tasks. For default value, use <code>null</code>.
1230	+	* @param asyncMode if true,
1231	+	* establishes local first-in-first-out scheduling mode for forked
1232	+	* tasks that are never joined. This mode may be more appropriate
1233	+	* than default locally stack-based mode in applications in which
1234	+	* worker threads only process event-style asynchronous tasks.
1235	+	* For default value, use <code>false</code>.
1236	+	* @throws IllegalArgumentException if parallelism less than or
1237	+	*         equal to zero, or greater than implementation limit
1238	+	* @throws NullPointerException if the factory is null
1239	+	* @throws SecurityException if a security manager exists and
1240	+	*         the caller is not permitted to modify threads
1241	+	*         because it does not hold {@link
1242	+	*         java.lang.RuntimePermission}{@code ("modifyThread")}
1243	+	*/
1244	+	public ForkJoinPool(int parallelism,
1245	+	ForkJoinWorkerThreadFactory factory,
1246	+	Thread.UncaughtExceptionHandler handler,
1247	+	boolean asyncMode) {
1248	+	checkPermission();
1249	+	if (factory == null)
1250	+	throw new NullPointerException();
1251	+	if (parallelism <= 0 || parallelism > MAX_WORKERS)
1252	+	throw new IllegalArgumentException();
1253	+	this.parallelism = parallelism;
1254	+	this.factory = factory;
1255	+	this.ueh = handler;
1256	+	this.locallyFifo = asyncMode;
1257	+	int arraySize = initialArraySizeFor(parallelism);
1258	+	this.workers = new ForkJoinWorkerThread[arraySize];
1259	+	this.submissionQueue = new LinkedTransferQueue<ForkJoinTask<?>>();
1260	+	this.workerLock = new ReentrantLock();
1261	+	this.termination = new Phaser(1);
1262	+	this.poolNumber = poolNumberGenerator.incrementAndGet();
1263	+	}
1264	+
1265	+	/**
1266	+	* Returns initial power of two size for workers array.
1267	+	* @param pc the initial parallelism level
1268	+	*/
1269	+	private static int initialArraySizeFor(int pc) {
1270	+	// If possible, initially allocate enough space for one spare
1271	+	int size = pc < MAX_WORKERS ? pc + 1 : MAX_WORKERS;
1272	+	// See Hackers Delight, sec 3.2. We know MAX_WORKERS < (1 >>> 16)
1273	+	size |= size >>> 1;
1274	+	size |= size >>> 2;
1275	+	size |= size >>> 4;
1276	+	size |= size >>> 8;
1277	+	return size + 1;
1278	+	}
1279	+
1280		// Execution methods
1281
1282		/**
#	Line 553 | Line 1285 | public class ForkJoinPool extends Abstra
1285		private <T> void doSubmit(ForkJoinTask<T> task) {
1286		if (task == null)
1287		throw new NullPointerException();
1288	<	if (isShutdown())
1288	>	if (runState >= SHUTDOWN)
1289		throw new RejectedExecutionException();
558	–	if (workers == null)
559	–	ensureWorkerInitialization();
1290		submissionQueue.offer(task);
1291	<	signalIdleWorkers();
1291	>	int c; // try to increment event count -- CAS failure OK
1292	>	UNSAFE.compareAndSwapInt(this, eventCountOffset, c = eventCount, c+1);
1293	>	helpMaintainParallelism(); // create, start, or resume some workers
1294		}
1295
1296		/**
#	Line 566 | Line 1298 | public class ForkJoinPool extends Abstra
1298		*
1299		* @param task the task
1300		* @return the task's result
1301	<	* @throws NullPointerException if task is null
1302	<	* @throws RejectedExecutionException if pool is shut down
1301	>	* @throws NullPointerException if the task is null
1302	>	* @throws RejectedExecutionException if the task cannot be
1303	>	*         scheduled for execution
1304		*/
1305		public <T> T invoke(ForkJoinTask<T> task) {
1306		doSubmit(task);
#	Line 578 | Line 1311 | public class ForkJoinPool extends Abstra
1311		* Arranges for (asynchronous) execution of the given task.
1312		*
1313		* @param task the task
1314	<	* @throws NullPointerException if task is null
1315	<	* @throws RejectedExecutionException if pool is shut down
1314	>	* @throws NullPointerException if the task is null
1315	>	* @throws RejectedExecutionException if the task cannot be
1316	>	*         scheduled for execution
1317		*/
1318		public void execute(ForkJoinTask<?> task) {
1319		doSubmit(task);
#	Line 587 | Line 1321 | public class ForkJoinPool extends Abstra
1321
1322		// AbstractExecutorService methods
1323
1324	+	/**
1325	+	* @throws NullPointerException if the task is null
1326	+	* @throws RejectedExecutionException if the task cannot be
1327	+	*         scheduled for execution
1328	+	*/
1329		public void execute(Runnable task) {
1330		ForkJoinTask<?> job;
1331		if (task instanceof ForkJoinTask<?>) // avoid re-wrap
#	Line 596 | Line 1335 | public class ForkJoinPool extends Abstra
1335		doSubmit(job);
1336		}
1337
1338	+	/**
1339	+	* Submits a ForkJoinTask for execution.
1340	+	*
1341	+	* @param task the task to submit
1342	+	* @return the task
1343	+	* @throws NullPointerException if the task is null
1344	+	* @throws RejectedExecutionException if the task cannot be
1345	+	*         scheduled for execution
1346	+	*/
1347	+	public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) {
1348	+	doSubmit(task);
1349	+	return task;
1350	+	}
1351	+
1352	+	/**
1353	+	* @throws NullPointerException if the task is null
1354	+	* @throws RejectedExecutionException if the task cannot be
1355	+	*         scheduled for execution
1356	+	*/
1357		public <T> ForkJoinTask<T> submit(Callable<T> task) {
1358		ForkJoinTask<T> job = ForkJoinTask.adapt(task);
1359		doSubmit(job);
1360		return job;
1361		}
1362
1363	+	/**
1364	+	* @throws NullPointerException if the task is null
1365	+	* @throws RejectedExecutionException if the task cannot be
1366	+	*         scheduled for execution
1367	+	*/
1368		public <T> ForkJoinTask<T> submit(Runnable task, T result) {
1369		ForkJoinTask<T> job = ForkJoinTask.adapt(task, result);
1370		doSubmit(job);
1371		return job;
1372		}
1373
1374	+	/**
1375	+	* @throws NullPointerException if the task is null
1376	+	* @throws RejectedExecutionException if the task cannot be
1377	+	*         scheduled for execution
1378	+	*/
1379		public ForkJoinTask<?> submit(Runnable task) {
1380		ForkJoinTask<?> job;
1381		if (task instanceof ForkJoinTask<?>) // avoid re-wrap
#	Line 619 | Line 1387 | public class ForkJoinPool extends Abstra
1387		}
1388
1389		/**
1390	<	* Submits a ForkJoinTask for execution.
1391	<	*
624	<	* @param task the task to submit
625	<	* @return the task
626	<	* @throws RejectedExecutionException if the task cannot be
627	<	*         scheduled for execution
628	<	* @throws NullPointerException if the task is null
1390	>	* @throws NullPointerException       {@inheritDoc}
1391	>	* @throws RejectedExecutionException {@inheritDoc}
1392		*/
630	–	public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) {
631	–	doSubmit(task);
632	–	return task;
633	–	}
634	–
635	–
1393		public <T> List<Future<T>> invokeAll(Collection<? extends Callable<T>> tasks) {
1394		ArrayList<ForkJoinTask<T>> forkJoinTasks =
1395		new ArrayList<ForkJoinTask<T>>(tasks.size());
#	Line 641 | Line 1398 | public class ForkJoinPool extends Abstra
1398		invoke(new InvokeAll<T>(forkJoinTasks));
1399
1400		@SuppressWarnings({"unchecked", "rawtypes"})
1401	<	List<Future<T>> futures = (List<Future<T>>) (List) forkJoinTasks;
1401	>	List<Future<T>> futures = (List<Future<T>>) (List) forkJoinTasks;
1402		return futures;
1403		}
1404
#	Line 655 | Line 1412 | public class ForkJoinPool extends Abstra
1412		private static final long serialVersionUID = -7914297376763021607L;
1413		}
1414
658	–	// Configuration and status settings and queries
659	–
1415		/**
1416		* Returns the factory used for constructing new workers.
1417		*
#	Line 673 | Line 1428 | public class ForkJoinPool extends Abstra
1428		* @return the handler, or {@code null} if none
1429		*/
1430		public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() {
1431	<	Thread.UncaughtExceptionHandler h;
677	<	final ReentrantLock lock = this.workerLock;
678	<	lock.lock();
679	<	try {
680	<	h = ueh;
681	<	} finally {
682	<	lock.unlock();
683	<	}
684	<	return h;
685	<	}
686	<
687	<	/**
688	<	* Sets the handler for internal worker threads that terminate due
689	<	* to unrecoverable errors encountered while executing tasks.
690	<	* Unless set, the current default or ThreadGroup handler is used
691	<	* as handler.
692	<	*
693	<	* @param h the new handler
694	<	* @return the old handler, or {@code null} if none
695	<	* @throws SecurityException if a security manager exists and
696	<	*         the caller is not permitted to modify threads
697	<	*         because it does not hold {@link
698	<	*         java.lang.RuntimePermission}{@code ("modifyThread")}
699	<	*/
700	<	public Thread.UncaughtExceptionHandler
701	<	setUncaughtExceptionHandler(Thread.UncaughtExceptionHandler h) {
702	<	checkPermission();
703	<	Thread.UncaughtExceptionHandler old = null;
704	<	final ReentrantLock lock = this.workerLock;
705	<	lock.lock();
706	<	try {
707	<	old = ueh;
708	<	ueh = h;
709	<	ForkJoinWorkerThread[] ws = workers;
710	<	if (ws != null) {
711	<	for (int i = 0; i < ws.length; ++i) {
712	<	ForkJoinWorkerThread w = ws[i];
713	<	if (w != null)
714	<	w.setUncaughtExceptionHandler(h);
715	<	}
716	<	}
717	<	} finally {
718	<	lock.unlock();
719	<	}
720	<	return old;
1431	>	return ueh;
1432		}
1433
723	–
1434		/**
1435	<	* Sets the target parallelism level of this pool.
1435	>	* Returns the targeted parallelism level of this pool.
1436		*
1437	<	* @param parallelism the target parallelism
728	<	* @throws IllegalArgumentException if parallelism less than or
729	<	* equal to zero or greater than maximum size bounds
730	<	* @throws SecurityException if a security manager exists and
731	<	*         the caller is not permitted to modify threads
732	<	*         because it does not hold {@link
733	<	*         java.lang.RuntimePermission}{@code ("modifyThread")}
734	<	*/
735	<	public void setParallelism(int parallelism) {
736	<	checkPermission();
737	<	if (parallelism <= 0 || parallelism > maxPoolSize)
738	<	throw new IllegalArgumentException();
739	<	final ReentrantLock lock = this.workerLock;
740	<	lock.lock();
741	<	try {
742	<	if (!isTerminating()) {
743	<	int p = this.parallelism;
744	<	this.parallelism = parallelism;
745	<	if (parallelism > p)
746	<	createAndStartAddedWorkers();
747	<	else
748	<	trimSpares();
749	<	}
750	<	} finally {
751	<	lock.unlock();
752	<	}
753	<	signalIdleWorkers();
754	<	}
755	<
756	<	/**
757	<	* Returns the targeted number of worker threads in this pool.
758	<	*
759	<	* @return the targeted number of worker threads in this pool
1437	>	* @return the targeted parallelism level of this pool
1438		*/
1439		public int getParallelism() {
1440		return parallelism;
#	Line 771 | Line 1449 | public class ForkJoinPool extends Abstra
1449		* @return the number of worker threads
1450		*/
1451		public int getPoolSize() {
1452	<	return totalCountOf(workerCounts);
775	<	}
776	<
777	<	/**
778	<	* Returns the maximum number of threads allowed to exist in the
779	<	* pool, even if there are insufficient unblocked running threads.
780	<	*
781	<	* @return the maximum
782	<	*/
783	<	public int getMaximumPoolSize() {
784	<	return maxPoolSize;
785	<	}
786	<
787	<	/**
788	<	* Sets the maximum number of threads allowed to exist in the
789	<	* pool, even if there are insufficient unblocked running threads.
790	<	* Setting this value has no effect on current pool size. It
791	<	* controls construction of new threads.
792	<	*
793	<	* @throws IllegalArgumentException if negative or greater than
794	<	* internal implementation limit
795	<	*/
796	<	public void setMaximumPoolSize(int newMax) {
797	<	if (newMax < 0 || newMax > MAX_THREADS)
798	<	throw new IllegalArgumentException();
799	<	maxPoolSize = newMax;
800	<	}
801	<
802	<
803	<	/**
804	<	* Returns {@code true} if this pool dynamically maintains its
805	<	* target parallelism level. If false, new threads are added only
806	<	* to avoid possible starvation.  This setting is by default true.
807	<	*
808	<	* @return {@code true} if maintains parallelism
809	<	*/
810	<	public boolean getMaintainsParallelism() {
811	<	return maintainsParallelism;
812	<	}
813	<
814	<	/**
815	<	* Sets whether this pool dynamically maintains its target
816	<	* parallelism level. If false, new threads are added only to
817	<	* avoid possible starvation.
818	<	*
819	<	* @param enable {@code true} to maintain parallelism
820	<	*/
821	<	public void setMaintainsParallelism(boolean enable) {
822	<	maintainsParallelism = enable;
823	<	}
824	<
825	<	/**
826	<	* Establishes local first-in-first-out scheduling mode for forked
827	<	* tasks that are never joined. This mode may be more appropriate
828	<	* than default locally stack-based mode in applications in which
829	<	* worker threads only process asynchronous tasks.  This method is
830	<	* designed to be invoked only when the pool is quiescent, and
831	<	* typically only before any tasks are submitted. The effects of
832	<	* invocations at other times may be unpredictable.
833	<	*
834	<	* @param async if {@code true}, use locally FIFO scheduling
835	<	* @return the previous mode
836	<	* @see #getAsyncMode
837	<	*/
838	<	public boolean setAsyncMode(boolean async) {
839	<	boolean oldMode = locallyFifo;
840	<	locallyFifo = async;
841	<	ForkJoinWorkerThread[] ws = workers;
842	<	if (ws != null) {
843	<	for (int i = 0; i < ws.length; ++i) {
844	<	ForkJoinWorkerThread t = ws[i];
845	<	if (t != null)
846	<	t.setAsyncMode(async);
847	<	}
848	<	}
849	<	return oldMode;
1452	>	return workerCounts >>> TOTAL_COUNT_SHIFT;
1453		}
1454
1455		/**
#	Line 854 | Line 1457 | public class ForkJoinPool extends Abstra
1457		* scheduling mode for forked tasks that are never joined.
1458		*
1459		* @return {@code true} if this pool uses async mode
857	–	* @see #setAsyncMode
1460		*/
1461		public boolean getAsyncMode() {
1462		return locallyFifo;
#	Line 863 | Line 1465 | public class ForkJoinPool extends Abstra
1465		/**
1466		* Returns an estimate of the number of worker threads that are
1467		* not blocked waiting to join tasks or for other managed
1468	<	* synchronization.
1468	>	* synchronization. This method may overestimate the
1469	>	* number of running threads.
1470		*
1471		* @return the number of worker threads
1472		*/
1473		public int getRunningThreadCount() {
1474	<	return runningCountOf(workerCounts);
1474	>	return workerCounts & RUNNING_COUNT_MASK;
1475		}
1476
1477		/**
#	Line 879 | Line 1482 | public class ForkJoinPool extends Abstra
1482		* @return the number of active threads
1483		*/
1484		public int getActiveThreadCount() {
1485	<	return activeCountOf(runControl);
883	<	}
884	<
885	<	/**
886	<	* Returns an estimate of the number of threads that are currently
887	<	* idle waiting for tasks. This method may underestimate the
888	<	* number of idle threads.
889	<	*
890	<	* @return the number of idle threads
891	<	*/
892	<	final int getIdleThreadCount() {
893	<	int c = runningCountOf(workerCounts) - activeCountOf(runControl);
894	<	return (c <= 0) ? 0 : c;
1485	>	return runState & ACTIVE_COUNT_MASK;
1486		}
1487
1488		/**
#	Line 906 | Line 1497 | public class ForkJoinPool extends Abstra
1497		* @return {@code true} if all threads are currently idle
1498		*/
1499		public boolean isQuiescent() {
1500	<	return activeCountOf(runControl) == 0;
1500	>	return (runState & ACTIVE_COUNT_MASK) == 0;
1501		}
1502
1503		/**
#	Line 921 | Line 1512 | public class ForkJoinPool extends Abstra
1512		* @return the number of steals
1513		*/
1514		public long getStealCount() {
1515	<	return stealCount.get();
925	<	}
926	<
927	<	/**
928	<	* Accumulates steal count from a worker.
929	<	* Call only when worker known to be idle.
930	<	*/
931	<	private void updateStealCount(ForkJoinWorkerThread w) {
932	<	int sc = w.getAndClearStealCount();
933	<	if (sc != 0)
934	<	stealCount.addAndGet(sc);
1515	>	return stealCount;
1516		}
1517
1518		/**
#	Line 947 | Line 1528 | public class ForkJoinPool extends Abstra
1528		public long getQueuedTaskCount() {
1529		long count = 0;
1530		ForkJoinWorkerThread[] ws = workers;
1531	<	if (ws != null) {
1532	<	for (int i = 0; i < ws.length; ++i) {
1533	<	ForkJoinWorkerThread t = ws[i];
1534	<	if (t != null)
1535	<	count += t.getQueueSize();
955	<	}
1531	>	int n = ws.length;
1532	>	for (int i = 0; i < n; ++i) {
1533	>	ForkJoinWorkerThread w = ws[i];
1534	>	if (w != null)
1535	>	count += w.getQueueSize();
1536		}
1537		return count;
1538		}
#	Line 1007 | Line 1587 | public class ForkJoinPool extends Abstra
1587		* @return the number of elements transferred
1588		*/
1589		protected int drainTasksTo(Collection<? super ForkJoinTask<?>> c) {
1590	<	int n = submissionQueue.drainTo(c);
1590	>	int count = submissionQueue.drainTo(c);
1591		ForkJoinWorkerThread[] ws = workers;
1592	<	if (ws != null) {
1593	<	for (int i = 0; i < ws.length; ++i) {
1594	<	ForkJoinWorkerThread w = ws[i];
1595	<	if (w != null)
1596	<	n += w.drainTasksTo(c);
1017	<	}
1592	>	int n = ws.length;
1593	>	for (int i = 0; i < n; ++i) {
1594	>	ForkJoinWorkerThread w = ws[i];
1595	>	if (w != null)
1596	>	count += w.drainTasksTo(c);
1597		}
1598	<	return n;
1598	>	return count;
1599		}
1600
1601		/**
#	Line 1027 | Line 1606 | public class ForkJoinPool extends Abstra
1606		* @return a string identifying this pool, as well as its state
1607		*/
1608		public String toString() {
1030	–	int ps = parallelism;
1031	–	int wc = workerCounts;
1032	–	int rc = runControl;
1609		long st = getStealCount();
1610		long qt = getQueuedTaskCount();
1611		long qs = getQueuedSubmissionCount();
1612	+	int wc = workerCounts;
1613	+	int tc = wc >>> TOTAL_COUNT_SHIFT;
1614	+	int rc = wc & RUNNING_COUNT_MASK;
1615	+	int pc = parallelism;
1616	+	int rs = runState;
1617	+	int ac = rs & ACTIVE_COUNT_MASK;
1618		return super.toString() +
1619	<	"[" + runStateToString(runStateOf(rc)) +
1620	<	", parallelism = " + ps +
1621	<	", size = " + totalCountOf(wc) +
1622	<	", active = " + activeCountOf(rc) +
1623	<	", running = " + runningCountOf(wc) +
1619	>	"[" + runLevelToString(rs) +
1620	>	", parallelism = " + pc +
1621	>	", size = " + tc +
1622	>	", active = " + ac +
1623	>	", running = " + rc +
1624		", steals = " + st +
1625		", tasks = " + qt +
1626		", submissions = " + qs +
1627		"]";
1628		}
1629
1630	<	private static String runStateToString(int rs) {
1631	<	switch(rs) {
1632	<	case RUNNING: return "Running";
1633	<	case SHUTDOWN: return "Shutting down";
1634	<	case TERMINATING: return "Terminating";
1053	<	case TERMINATED: return "Terminated";
1054	<	default: throw new Error("Unknown run state");
1055	<	}
1630	>	private static String runLevelToString(int s) {
1631	>	return ((s & TERMINATED) != 0 ? "Terminated" :
1632	>	((s & TERMINATING) != 0 ? "Terminating" :
1633	>	((s & SHUTDOWN) != 0 ? "Shutting down" :
1634	>	"Running")));
1635		}
1636
1058	–	// lifecycle control
1059	–
1637		/**
1638		* Initiates an orderly shutdown in which previously submitted
1639		* tasks are executed, but no new tasks will be accepted.
#	Line 1071 | Line 1648 | public class ForkJoinPool extends Abstra
1648		*/
1649		public void shutdown() {
1650		checkPermission();
1651	<	transitionRunStateTo(SHUTDOWN);
1652	<	if (canTerminateOnShutdown(runControl)) {
1076	<	if (workers == null) { // shutting down before workers created
1077	<	final ReentrantLock lock = this.workerLock;
1078	<	lock.lock();
1079	<	try {
1080	<	if (workers == null) {
1081	<	terminate();
1082	<	transitionRunStateTo(TERMINATED);
1083	<	termination.signalAll();
1084	<	}
1085	<	} finally {
1086	<	lock.unlock();
1087	<	}
1088	<	}
1089	<	terminateOnShutdown();
1090	<	}
1651	>	advanceRunLevel(SHUTDOWN);
1652	>	tryTerminate(false);
1653		}
1654
1655		/**
1656	<	* Attempts to stop all actively executing tasks, and cancels all
1657	<	* waiting tasks.  Tasks that are in the process of being
1658	<	* submitted or executed concurrently during the course of this
1659	<	* method may or may not be rejected. Unlike some other executors,
1660	<	* this method cancels rather than collects non-executed tasks
1661	<	* upon termination, so always returns an empty list. However, you
1662	<	* can use method {@link #drainTasksTo} before invoking this
1663	<	* method to transfer unexecuted tasks to another collection.
1656	>	* Attempts to cancel and/or stop all tasks, and reject all
1657	>	* subsequently submitted tasks.  Tasks that are in the process of
1658	>	* being submitted or executed concurrently during the course of
1659	>	* this method may or may not be rejected. This method cancels
1660	>	* both existing and unexecuted tasks, in order to permit
1661	>	* termination in the presence of task dependencies. So the method
1662	>	* always returns an empty list (unlike the case for some other
1663	>	* Executors).
1664		*
1665		* @return an empty list
1666		* @throws SecurityException if a security manager exists and
#	Line 1108 | Line 1670 | public class ForkJoinPool extends Abstra
1670		*/
1671		public List<Runnable> shutdownNow() {
1672		checkPermission();
1673	<	terminate();
1673	>	tryTerminate(true);
1674		return Collections.emptyList();
1675		}
1676
#	Line 1118 | Line 1680 | public class ForkJoinPool extends Abstra
1680		* @return {@code true} if all tasks have completed following shut down
1681		*/
1682		public boolean isTerminated() {
1683	<	return runStateOf(runControl) == TERMINATED;
1683	>	return runState >= TERMINATED;
1684		}
1685
1686		/**
1687		* Returns {@code true} if the process of termination has
1688	<	* commenced but possibly not yet completed.
1688	>	* commenced but not yet completed.  This method may be useful for
1689	>	* debugging. A return of {@code true} reported a sufficient
1690	>	* period after shutdown may indicate that submitted tasks have
1691	>	* ignored or suppressed interruption, causing this executor not
1692	>	* to properly terminate.
1693		*
1694	<	* @return {@code true} if terminating
1694	>	* @return {@code true} if terminating but not yet terminated
1695		*/
1696		public boolean isTerminating() {
1697	<	return runStateOf(runControl) >= TERMINATING;
1697	>	return (runState & (TERMINATING|TERMINATED)) == TERMINATING;
1698		}
1699
1700		/**
#	Line 1137 | Line 1703 | public class ForkJoinPool extends Abstra
1703		* @return {@code true} if this pool has been shut down
1704		*/
1705		public boolean isShutdown() {
1706	<	return runStateOf(runControl) >= SHUTDOWN;
1706	>	return runState >= SHUTDOWN;
1707		}
1708
1709		/**
#	Line 1153 | Line 1719 | public class ForkJoinPool extends Abstra
1719		*/
1720		public boolean awaitTermination(long timeout, TimeUnit unit)
1721		throws InterruptedException {
1156	–	long nanos = unit.toNanos(timeout);
1157	–	final ReentrantLock lock = this.workerLock;
1158	–	lock.lock();
1159	–	try {
1160	–	for (;;) {
1161	–	if (isTerminated())
1162	–	return true;
1163	–	if (nanos <= 0)
1164	–	return false;
1165	–	nanos = termination.awaitNanos(nanos);
1166	–	}
1167	–	} finally {
1168	–	lock.unlock();
1169	–	}
1170	–	}
1171	–
1172	–	// Shutdown and termination support
1173	–
1174	–	/**
1175	–	* Callback from terminating worker. Nulls out the corresponding
1176	–	* workers slot, and if terminating, tries to terminate; else
1177	–	* tries to shrink workers array.
1178	–	*
1179	–	* @param w the worker
1180	–	*/
1181	–	final void workerTerminated(ForkJoinWorkerThread w) {
1182	–	updateStealCount(w);
1183	–	updateWorkerCount(-1);
1184	–	final ReentrantLock lock = this.workerLock;
1185	–	lock.lock();
1186	–	try {
1187	–	ForkJoinWorkerThread[] ws = workers;
1188	–	if (ws != null) {
1189	–	int idx = w.poolIndex;
1190	–	if (idx >= 0 && idx < ws.length && ws[idx] == w)
1191	–	ws[idx] = null;
1192	–	if (totalCountOf(workerCounts) == 0) {
1193	–	terminate(); // no-op if already terminating
1194	–	transitionRunStateTo(TERMINATED);
1195	–	termination.signalAll();
1196	–	}
1197	–	else if (!isTerminating()) {
1198	–	tryShrinkWorkerArray();
1199	–	tryResumeSpare(true); // allow replacement
1200	–	}
1201	–	}
1202	–	} finally {
1203	–	lock.unlock();
1204	–	}
1205	–	signalIdleWorkers();
1206	–	}
1207	–
1208	–	/**
1209	–	* Initiates termination.
1210	–	*/
1211	–	private void terminate() {
1212	–	if (transitionRunStateTo(TERMINATING)) {
1213	–	stopAllWorkers();
1214	–	resumeAllSpares();
1215	–	signalIdleWorkers();
1216	–	cancelQueuedSubmissions();
1217	–	cancelQueuedWorkerTasks();
1218	–	interruptUnterminatedWorkers();
1219	–	signalIdleWorkers(); // resignal after interrupt
1220	–	}
1221	–	}
1222	–
1223	–	/**
1224	–	* Possibly terminates when on shutdown state.
1225	–	*/
1226	–	private void terminateOnShutdown() {
1227	–	if (!hasQueuedSubmissions() && canTerminateOnShutdown(runControl))
1228	–	terminate();
1229	–	}
1230	–
1231	–	/**
1232	–	* Clears out and cancels submissions.
1233	–	*/
1234	–	private void cancelQueuedSubmissions() {
1235	–	ForkJoinTask<?> task;
1236	–	while ((task = pollSubmission()) != null)
1237	–	task.cancel(false);
1238	–	}
1239	–
1240	–	/**
1241	–	* Cleans out worker queues.
1242	–	*/
1243	–	private void cancelQueuedWorkerTasks() {
1244	–	final ReentrantLock lock = this.workerLock;
1245	–	lock.lock();
1722		try {
1723	<	ForkJoinWorkerThread[] ws = workers;
1724	<	if (ws != null) {
1249	<	for (int i = 0; i < ws.length; ++i) {
1250	<	ForkJoinWorkerThread t = ws[i];
1251	<	if (t != null)
1252	<	t.cancelTasks();
1253	<	}
1254	<	}
1255	<	} finally {
1256	<	lock.unlock();
1257	<	}
1258	<	}
1259	<
1260	<	/**
1261	<	* Sets each worker's status to terminating. Requires lock to avoid
1262	<	* conflicts with add/remove.
1263	<	*/
1264	<	private void stopAllWorkers() {
1265	<	final ReentrantLock lock = this.workerLock;
1266	<	lock.lock();
1267	<	try {
1268	<	ForkJoinWorkerThread[] ws = workers;
1269	<	if (ws != null) {
1270	<	for (int i = 0; i < ws.length; ++i) {
1271	<	ForkJoinWorkerThread t = ws[i];
1272	<	if (t != null)
1273	<	t.shutdownNow();
1274	<	}
1275	<	}
1276	<	} finally {
1277	<	lock.unlock();
1278	<	}
1279	<	}
1280	<
1281	<	/**
1282	<	* Interrupts all unterminated workers.  This is not required for
1283	<	* sake of internal control, but may help unstick user code during
1284	<	* shutdown.
1285	<	*/
1286	<	private void interruptUnterminatedWorkers() {
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 && !t.isTerminated()) {
1295	<	try {
1296	<	t.interrupt();
1297	<	} catch (SecurityException ignore) {
1298	<	}
1299	<	}
1300	<	}
1301	<	}
1302	<	} finally {
1303	<	lock.unlock();
1304	<	}
1305	<	}
1306	<
1307	<
1308	<	/*
1309	<	* Nodes for event barrier to manage idle threads.  Queue nodes
1310	<	* are basic Treiber stack nodes, also used for spare stack.
1311	<	*
1312	<	* The event barrier has an event count and a wait queue (actually
1313	<	* a Treiber stack).  Workers are enabled to look for work when
1314	<	* the eventCount is incremented. If they fail to find work, they
1315	<	* may wait for next count. Upon release, threads help others wake
1316	<	* up.
1317	<	*
1318	<	* Synchronization events occur only in enough contexts to
1319	<	* maintain overall liveness:
1320	<	*
1321	<	*   - Submission of a new task to the pool
1322	<	*   - Resizes or other changes to the workers array
1323	<	*   - pool termination
1324	<	*   - A worker pushing a task on an empty queue
1325	<	*
1326	<	* The case of pushing a task occurs often enough, and is heavy
1327	<	* enough compared to simple stack pushes, to require special
1328	<	* handling: Method signalWork returns without advancing count if
1329	<	* the queue appears to be empty.  This would ordinarily result in
1330	<	* races causing some queued waiters not to be woken up. To avoid
1331	<	* this, the first worker enqueued in method sync (see
1332	<	* syncIsReleasable) rescans for tasks after being enqueued, and
1333	<	* helps signal if any are found. This works well because the
1334	<	* worker has nothing better to do, and so might as well help
1335	<	* alleviate the overhead and contention on the threads actually
1336	<	* doing work.  Also, since event counts increments on task
1337	<	* availability exist to maintain liveness (rather than to force
1338	<	* refreshes etc), it is OK for callers to exit early if
1339	<	* contending with another signaller.
1340	<	*/
1341	<	static final class WaitQueueNode {
1342	<	WaitQueueNode next; // only written before enqueued
1343	<	volatile ForkJoinWorkerThread thread; // nulled to cancel wait
1344	<	final long count; // unused for spare stack
1345	<
1346	<	WaitQueueNode(long c, ForkJoinWorkerThread w) {
1347	<	count = c;
1348	<	thread = w;
1349	<	}
1350	<
1351	<	/**
1352	<	* Wakes up waiter, returning false if known to already
1353	<	*/
1354	<	boolean signal() {
1355	<	ForkJoinWorkerThread t = thread;
1356	<	if (t == null)
1357	<	return false;
1358	<	thread = null;
1359	<	LockSupport.unpark(t);
1360	<	return true;
1361	<	}
1362	<
1363	<	/**
1364	<	* Awaits release on sync.
1365	<	*/
1366	<	void awaitSyncRelease(ForkJoinPool p) {
1367	<	while (thread != null && !p.syncIsReleasable(this))
1368	<	LockSupport.park(this);
1369	<	}
1370	<
1371	<	/**
1372	<	* Awaits resumption as spare.
1373	<	*/
1374	<	void awaitSpareRelease() {
1375	<	while (thread != null) {
1376	<	if (!Thread.interrupted())
1377	<	LockSupport.park(this);
1378	<	}
1379	<	}
1380	<	}
1381	<
1382	<	/**
1383	<	* Ensures that no thread is waiting for count to advance from the
1384	<	* current value of eventCount read on entry to this method, by
1385	<	* releasing waiting threads if necessary.
1386	<	*
1387	<	* @return the count
1388	<	*/
1389	<	final long ensureSync() {
1390	<	long c = eventCount;
1391	<	WaitQueueNode q;
1392	<	while ((q = syncStack) != null && q.count < c) {
1393	<	if (casBarrierStack(q, null)) {
1394	<	do {
1395	<	q.signal();
1396	<	} while ((q = q.next) != null);
1397	<	break;
1398	<	}
1399	<	}
1400	<	return c;
1401	<	}
1402	<
1403	<	/**
1404	<	* Increments event count and releases waiting threads.
1405	<	*/
1406	<	private void signalIdleWorkers() {
1407	<	long c;
1408	<	do {} while (!casEventCount(c = eventCount, c+1));
1409	<	ensureSync();
1410	<	}
1411	<
1412	<	/**
1413	<	* Signals threads waiting to poll a task. Because method sync
1414	<	* rechecks availability, it is OK to only proceed if queue
1415	<	* appears to be non-empty, and OK to skip under contention to
1416	<	* increment count (since some other thread succeeded).
1417	<	*/
1418	<	final void signalWork() {
1419	<	long c;
1420	<	WaitQueueNode q;
1421	<	if (syncStack != null &&
1422	<	casEventCount(c = eventCount, c+1) &&
1423	<	(((q = syncStack) != null && q.count <= c) &&
1424	<	(!casBarrierStack(q, q.next) || !q.signal())))
1425	<	ensureSync();
1426	<	}
1427	<
1428	<	/**
1429	<	* Waits until event count advances from last value held by
1430	<	* caller, or if excess threads, caller is resumed as spare, or
1431	<	* caller or pool is terminating. Updates caller's event on exit.
1432	<	*
1433	<	* @param w the calling worker thread
1434	<	*/
1435	<	final void sync(ForkJoinWorkerThread w) {
1436	<	updateStealCount(w); // Transfer w's count while it is idle
1437	<
1438	<	while (!w.isShutdown() && !isTerminating() && !suspendIfSpare(w)) {
1439	<	long prev = w.lastEventCount;
1440	<	WaitQueueNode node = null;
1441	<	WaitQueueNode h;
1442	<	while (eventCount == prev &&
1443	<	((h = syncStack) == null || h.count == prev)) {
1444	<	if (node == null)
1445	<	node = new WaitQueueNode(prev, w);
1446	<	if (casBarrierStack(node.next = h, node)) {
1447	<	node.awaitSyncRelease(this);
1448	<	break;
1449	<	}
1450	<	}
1451	<	long ec = ensureSync();
1452	<	if (ec != prev) {
1453	<	w.lastEventCount = ec;
1454	<	break;
1455	<	}
1456	<	}
1457	<	}
1458	<
1459	<	/**
1460	<	* Returns {@code true} if worker waiting on sync can proceed:
1461	<	*  - on signal (thread == null)
1462	<	*  - on event count advance (winning race to notify vs signaller)
1463	<	*  - on interrupt
1464	<	*  - if the first queued node, we find work available
1465	<	* If node was not signalled and event count not advanced on exit,
1466	<	* then we also help advance event count.
1467	<	*
1468	<	* @return {@code true} if node can be released
1469	<	*/
1470	<	final boolean syncIsReleasable(WaitQueueNode node) {
1471	<	long prev = node.count;
1472	<	if (!Thread.interrupted() && node.thread != null &&
1473	<	(node.next != null ||
1474	<	!ForkJoinWorkerThread.hasQueuedTasks(workers)) &&
1475	<	eventCount == prev)
1476	<	return false;
1477	<	if (node.thread != null) {
1478	<	node.thread = null;
1479	<	long ec = eventCount;
1480	<	if (prev <= ec) // help signal
1481	<	casEventCount(ec, ec+1);
1482	<	}
1483	<	return true;
1484	<	}
1485	<
1486	<	/**
1487	<	* Returns {@code true} if a new sync event occurred since last
1488	<	* call to sync or this method, if so, updating caller's count.
1489	<	*/
1490	<	final boolean hasNewSyncEvent(ForkJoinWorkerThread w) {
1491	<	long lc = w.lastEventCount;
1492	<	long ec = ensureSync();
1493	<	if (ec == lc)
1723	>	return termination.awaitAdvanceInterruptibly(0, timeout, unit) > 0;
1724	>	} catch(TimeoutException ex) {
1725		return false;
1495	–	w.lastEventCount = ec;
1496	–	return true;
1497	–	}
1498	–
1499	–	//  Parallelism maintenance
1500	–
1501	–	/**
1502	–	* Decrements running count; if too low, adds spare.
1503	–	*
1504	–	* Conceptually, all we need to do here is add or resume a
1505	–	* spare thread when one is about to block (and remove or
1506	–	* suspend it later when unblocked -- see suspendIfSpare).
1507	–	* However, implementing this idea requires coping with
1508	–	* several problems: we have imperfect information about the
1509	–	* states of threads. Some count updates can and usually do
1510	–	* lag run state changes, despite arrangements to keep them
1511	–	* accurate (for example, when possible, updating counts
1512	–	* before signalling or resuming), especially when running on
1513	–	* dynamic JVMs that don't optimize the infrequent paths that
1514	–	* update counts. Generating too many threads can make these
1515	–	* problems become worse, because excess threads are more
1516	–	* likely to be context-switched with others, slowing them all
1517	–	* down, especially if there is no work available, so all are
1518	–	* busy scanning or idling.  Also, excess spare threads can
1519	–	* only be suspended or removed when they are idle, not
1520	–	* immediately when they aren't needed. So adding threads will
1521	–	* raise parallelism level for longer than necessary.  Also,
1522	–	* FJ applications often encounter highly transient peaks when
1523	–	* many threads are blocked joining, but for less time than it
1524	–	* takes to create or resume spares.
1525	–	*
1526	–	* @param joinMe if non-null, return early if done
1527	–	* @param maintainParallelism if true, try to stay within
1528	–	* target counts, else create only to avoid starvation
1529	–	* @return true if joinMe known to be done
1530	–	*/
1531	–	final boolean preJoin(ForkJoinTask<?> joinMe,
1532	–	boolean maintainParallelism) {
1533	–	maintainParallelism &= maintainsParallelism; // overrride
1534	–	boolean dec = false;  // true when running count decremented
1535	–	while (spareStack == null || !tryResumeSpare(dec)) {
1536	–	int counts = workerCounts;
1537	–	if (dec || (dec = casWorkerCounts(counts, --counts))) {
1538	–	// CAS cheat
1539	–	if (!needSpare(counts, maintainParallelism))
1540	–	break;
1541	–	if (joinMe.status < 0)
1542	–	return true;
1543	–	if (tryAddSpare(counts))
1544	–	break;
1545	–	}
1546	–	}
1547	–	return false;
1548	–	}
1549	–
1550	–	/**
1551	–	* Same idea as preJoin
1552	–	*/
1553	–	final boolean preBlock(ManagedBlocker blocker,
1554	–	boolean maintainParallelism) {
1555	–	maintainParallelism &= maintainsParallelism;
1556	–	boolean dec = false;
1557	–	while (spareStack == null || !tryResumeSpare(dec)) {
1558	–	int counts = workerCounts;
1559	–	if (dec || (dec = casWorkerCounts(counts, --counts))) {
1560	–	if (!needSpare(counts, maintainParallelism))
1561	–	break;
1562	–	if (blocker.isReleasable())
1563	–	return true;
1564	–	if (tryAddSpare(counts))
1565	–	break;
1566	–	}
1567	–	}
1568	–	return false;
1569	–	}
1570	–
1571	–	/**
1572	–	* Returns {@code true} if a spare thread appears to be needed.
1573	–	* If maintaining parallelism, returns true when the deficit in
1574	–	* running threads is more than the surplus of total threads, and
1575	–	* there is apparently some work to do.  This self-limiting rule
1576	–	* means that the more threads that have already been added, the
1577	–	* less parallelism we will tolerate before adding another.
1578	–	*
1579	–	* @param counts current worker counts
1580	–	* @param maintainParallelism try to maintain parallelism
1581	–	*/
1582	–	private boolean needSpare(int counts, boolean maintainParallelism) {
1583	–	int ps = parallelism;
1584	–	int rc = runningCountOf(counts);
1585	–	int tc = totalCountOf(counts);
1586	–	int runningDeficit = ps - rc;
1587	–	int totalSurplus = tc - ps;
1588	–	return (tc < maxPoolSize &&
1589	–	(rc == 0 || totalSurplus < 0 ||
1590	–	(maintainParallelism &&
1591	–	runningDeficit > totalSurplus &&
1592	–	ForkJoinWorkerThread.hasQueuedTasks(workers))));
1593	–	}
1594	–
1595	–	/**
1596	–	* Adds a spare worker if lock available and no more than the
1597	–	* expected numbers of threads exist.
1598	–	*
1599	–	* @return true if successful
1600	–	*/
1601	–	private boolean tryAddSpare(int expectedCounts) {
1602	–	final ReentrantLock lock = this.workerLock;
1603	–	int expectedRunning = runningCountOf(expectedCounts);
1604	–	int expectedTotal = totalCountOf(expectedCounts);
1605	–	boolean success = false;
1606	–	boolean locked = false;
1607	–	// confirm counts while locking; CAS after obtaining lock
1608	–	try {
1609	–	for (;;) {
1610	–	int s = workerCounts;
1611	–	int tc = totalCountOf(s);
1612	–	int rc = runningCountOf(s);
1613	–	if (rc > expectedRunning || tc > expectedTotal)
1614	–	break;
1615	–	if (!locked && !(locked = lock.tryLock()))
1616	–	break;
1617	–	if (casWorkerCounts(s, workerCountsFor(tc+1, rc+1))) {
1618	–	createAndStartSpare(tc);
1619	–	success = true;
1620	–	break;
1621	–	}
1622	–	}
1623	–	} finally {
1624	–	if (locked)
1625	–	lock.unlock();
1626	–	}
1627	–	return success;
1628	–	}
1629	–
1630	–	/**
1631	–	* Adds the kth spare worker. On entry, pool counts are already
1632	–	* adjusted to reflect addition.
1633	–	*/
1634	–	private void createAndStartSpare(int k) {
1635	–	ForkJoinWorkerThread w = null;
1636	–	ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(k + 1);
1637	–	int len = ws.length;
1638	–	// Probably, we can place at slot k. If not, find empty slot
1639	–	if (k < len && ws[k] != null) {
1640	–	for (k = 0; k < len && ws[k] != null; ++k)
1641	–	;
1642	–	}
1643	–	if (k < len && !isTerminating() && (w = createWorker(k)) != null) {
1644	–	ws[k] = w;
1645	–	w.start();
1646	–	}
1647	–	else
1648	–	updateWorkerCount(-1); // adjust on failure
1649	–	signalIdleWorkers();
1650	–	}
1651	–
1652	–	/**
1653	–	* Suspends calling thread w if there are excess threads.  Called
1654	–	* only from sync.  Spares are enqueued in a Treiber stack using
1655	–	* the same WaitQueueNodes as barriers.  They are resumed mainly
1656	–	* in preJoin, but are also woken on pool events that require all
1657	–	* threads to check run state.
1658	–	*
1659	–	* @param w the caller
1660	–	*/
1661	–	private boolean suspendIfSpare(ForkJoinWorkerThread w) {
1662	–	WaitQueueNode node = null;
1663	–	int s;
1664	–	while (parallelism < runningCountOf(s = workerCounts)) {
1665	–	if (node == null)
1666	–	node = new WaitQueueNode(0, w);
1667	–	if (casWorkerCounts(s, s-1)) { // representation-dependent
1668	–	// push onto stack
1669	–	do {} while (!casSpareStack(node.next = spareStack, node));
1670	–	// block until released by resumeSpare
1671	–	node.awaitSpareRelease();
1672	–	return true;
1673	–	}
1674	–	}
1675	–	return false;
1676	–	}
1677	–
1678	–	/**
1679	–	* Tries to pop and resume a spare thread.
1680	–	*
1681	–	* @param updateCount if true, increment running count on success
1682	–	* @return true if successful
1683	–	*/
1684	–	private boolean tryResumeSpare(boolean updateCount) {
1685	–	WaitQueueNode q;
1686	–	while ((q = spareStack) != null) {
1687	–	if (casSpareStack(q, q.next)) {
1688	–	if (updateCount)
1689	–	updateRunningCount(1);
1690	–	q.signal();
1691	–	return true;
1692	–	}
1693	–	}
1694	–	return false;
1695	–	}
1696	–
1697	–	/**
1698	–	* Pops and resumes all spare threads. Same idea as ensureSync.
1699	–	*
1700	–	* @return true if any spares released
1701	–	*/
1702	–	private boolean resumeAllSpares() {
1703	–	WaitQueueNode q;
1704	–	while ( (q = spareStack) != null) {
1705	–	if (casSpareStack(q, null)) {
1706	–	do {
1707	–	updateRunningCount(1);
1708	–	q.signal();
1709	–	} while ((q = q.next) != null);
1710	–	return true;
1711	–	}
1712	–	}
1713	–	return false;
1714	–	}
1715	–
1716	–	/**
1717	–	* Pops and shuts down excessive spare threads. Call only while
1718	–	* holding lock. This is not guaranteed to eliminate all excess
1719	–	* threads, only those suspended as spares, which are the ones
1720	–	* unlikely to be needed in the future.
1721	–	*/
1722	–	private void trimSpares() {
1723	–	int surplus = totalCountOf(workerCounts) - parallelism;
1724	–	WaitQueueNode q;
1725	–	while (surplus > 0 && (q = spareStack) != null) {
1726	–	if (casSpareStack(q, null)) {
1727	–	do {
1728	–	updateRunningCount(1);
1729	–	ForkJoinWorkerThread w = q.thread;
1730	–	if (w != null && surplus > 0 &&
1731	–	runningCountOf(workerCounts) > 0 && w.shutdown())
1732	–	--surplus;
1733	–	q.signal();
1734	–	} while ((q = q.next) != null);
1735	–	}
1726		}
1727		}
1728
#	Line 1740 | Line 1730 | public class ForkJoinPool extends Abstra
1730		* Interface for extending managed parallelism for tasks running
1731		* in {@link ForkJoinPool}s.
1732		*
1733	<	* <p>A {@code ManagedBlocker} provides two methods.
1734	<	* Method {@code isReleasable} must return {@code true} if
1735	<	* blocking is not necessary. Method {@code block} blocks the
1736	<	* current thread if necessary (perhaps internally invoking
1737	<	* {@code isReleasable} before actually blocking).
1733	>	* <p>A {@code ManagedBlocker} provides two methods.  Method
1734	>	* {@code isReleasable} must return {@code true} if blocking is
1735	>	* not necessary. Method {@code block} blocks the current thread
1736	>	* if necessary (perhaps internally invoking {@code isReleasable}
1737	>	* before actually blocking). The unusual methods in this API
1738	>	* accommodate synchronizers that may, but don't usually, block
1739	>	* for long periods. Similarly, they allow more efficient internal
1740	>	* handling of cases in which additional workers may be, but
1741	>	* usually are not, needed to ensure sufficient parallelism.
1742	>	* Toward this end, implementations of method {@code isReleasable}
1743	>	* must be amenable to repeated invocation.
1744		*
1745		* <p>For example, here is a ManagedBlocker based on a
1746		* ReentrantLock:
#	Line 1762 | Line 1758 | public class ForkJoinPool extends Abstra
1758		*     return hasLock || (hasLock = lock.tryLock());
1759		*   }
1760		* }}</pre>
1761	+	*
1762	+	* <p>Here is a class that possibly blocks waiting for an
1763	+	* item on a given queue:
1764	+	*  <pre> {@code
1765	+	* class QueueTaker<E> implements ManagedBlocker {
1766	+	*   final BlockingQueue<E> queue;
1767	+	*   volatile E item = null;
1768	+	*   QueueTaker(BlockingQueue<E> q) { this.queue = q; }
1769	+	*   public boolean block() throws InterruptedException {
1770	+	*     if (item == null)
1771	+	*       item = queue.take();
1772	+	*     return true;
1773	+	*   }
1774	+	*   public boolean isReleasable() {
1775	+	*     return item != null || (item = queue.poll()) != null;
1776	+	*   }
1777	+	*   public E getItem() { // call after pool.managedBlock completes
1778	+	*     return item;
1779	+	*   }
1780	+	* }}</pre>
1781		*/
1782		public static interface ManagedBlocker {
1783		/**
#	Line 1785 | Line 1801 | public class ForkJoinPool extends Abstra
1801		* Blocks in accord with the given blocker.  If the current thread
1802		* is a {@link ForkJoinWorkerThread}, this method possibly
1803		* arranges for a spare thread to be activated if necessary to
1804	<	* ensure parallelism while the current thread is blocked.
1789	<	*
1790	<	* <p>If {@code maintainParallelism} is {@code true} and the pool
1791	<	* supports it ({@link #getMaintainsParallelism}), this method
1792	<	* attempts to maintain the pool's nominal parallelism. Otherwise
1793	<	* it activates a thread only if necessary to avoid complete
1794	<	* starvation. This option may be preferable when blockages use
1795	<	* timeouts, or are almost always brief.
1804	>	* ensure sufficient parallelism while the current thread is blocked.
1805		*
1806		* <p>If the caller is not a {@link ForkJoinTask}, this method is
1807		* behaviorally equivalent to
#	Line 1806 | Line 1815 | public class ForkJoinPool extends Abstra
1815		* first be expanded to ensure parallelism, and later adjusted.
1816		*
1817		* @param blocker the blocker
1809	–	* @param maintainParallelism if {@code true} and supported by
1810	–	* this pool, attempt to maintain the pool's nominal parallelism;
1811	–	* otherwise activate a thread only if necessary to avoid
1812	–	* complete starvation.
1818		* @throws InterruptedException if blocker.block did so
1819		*/
1820	<	public static void managedBlock(ManagedBlocker blocker,
1816	<	boolean maintainParallelism)
1820	>	public static void managedBlock(ManagedBlocker blocker)
1821		throws InterruptedException {
1822		Thread t = Thread.currentThread();
1823	<	ForkJoinPool pool = ((t instanceof ForkJoinWorkerThread) ?
1824	<	((ForkJoinWorkerThread) t).pool : null);
1825	<	if (!blocker.isReleasable()) {
1826	<	try {
1827	<	if (pool == null ||
1828	<	!pool.preBlock(blocker, maintainParallelism))
1825	<	awaitBlocker(blocker);
1826	<	} finally {
1827	<	if (pool != null)
1828	<	pool.updateRunningCount(1);
1829	<	}
1823	>	if (t instanceof ForkJoinWorkerThread) {
1824	>	ForkJoinWorkerThread w = (ForkJoinWorkerThread) t;
1825	>	w.pool.awaitBlocker(blocker);
1826	>	}
1827	>	else {
1828	>	do {} while (!blocker.isReleasable() && !blocker.block());
1829		}
1831	–	}
1832	–
1833	–	private static void awaitBlocker(ManagedBlocker blocker)
1834	–	throws InterruptedException {
1835	–	do {} while (!blocker.isReleasable() && !blocker.block());
1830		}
1831
1832		// AbstractExecutorService overrides.  These rely on undocumented
#	Line 1850 | Line 1844 | public class ForkJoinPool extends Abstra
1844		// Unsafe mechanics
1845
1846		private static final sun.misc.Unsafe UNSAFE = getUnsafe();
1853	–	private static final long eventCountOffset =
1854	–	objectFieldOffset("eventCount", ForkJoinPool.class);
1847		private static final long workerCountsOffset =
1848		objectFieldOffset("workerCounts", ForkJoinPool.class);
1849	<	private static final long runControlOffset =
1850	<	objectFieldOffset("runControl", ForkJoinPool.class);
1851	<	private static final long syncStackOffset =
1852	<	objectFieldOffset("syncStack",ForkJoinPool.class);
1853	<	private static final long spareStackOffset =
1854	<	objectFieldOffset("spareStack", ForkJoinPool.class);
1855	<
1856	<	private boolean casEventCount(long cmp, long val) {
1857	<	return UNSAFE.compareAndSwapLong(this, eventCountOffset, cmp, val);
1858	<	}
1867	<	private boolean casWorkerCounts(int cmp, int val) {
1868	<	return UNSAFE.compareAndSwapInt(this, workerCountsOffset, cmp, val);
1869	<	}
1870	<	private boolean casRunControl(int cmp, int val) {
1871	<	return UNSAFE.compareAndSwapInt(this, runControlOffset, cmp, val);
1872	<	}
1873	<	private boolean casSpareStack(WaitQueueNode cmp, WaitQueueNode val) {
1874	<	return UNSAFE.compareAndSwapObject(this, spareStackOffset, cmp, val);
1875	<	}
1876	<	private boolean casBarrierStack(WaitQueueNode cmp, WaitQueueNode val) {
1877	<	return UNSAFE.compareAndSwapObject(this, syncStackOffset, cmp, val);
1878	<	}
1849	>	private static final long runStateOffset =
1850	>	objectFieldOffset("runState", ForkJoinPool.class);
1851	>	private static final long eventCountOffset =
1852	>	objectFieldOffset("eventCount", ForkJoinPool.class);
1853	>	private static final long eventWaitersOffset =
1854	>	objectFieldOffset("eventWaiters",ForkJoinPool.class);
1855	>	private static final long stealCountOffset =
1856	>	objectFieldOffset("stealCount",ForkJoinPool.class);
1857	>	private static final long spareWaitersOffset =
1858	>	objectFieldOffset("spareWaiters",ForkJoinPool.class);
1859
1860		private static long objectFieldOffset(String field, Class<?> klazz) {
1861		try {

Diff Legend

–	Removed lines
+	Added lines
<	Changed lines
>	Changed lines