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 1.23 by dl, Sat Jul 25 15:50:57 2009 UTC vs.
Revision 1.56 by dl, Thu May 27 16:46:48 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.  A
23	<	* ForkJoinPool provides the entry point for submissions from
24	<	* non-ForkJoinTasks, as well as management and monitoring operations.
25	<	* Normally a single ForkJoinPool is used for a large number of
27	<	* submitted tasks. Otherwise, use would not usually outweigh the
28	<	* construction and bookkeeping overhead of creating a large set of
29	<	* threads.
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.
26		*
27	<	* <p>ForkJoinPools differ from other kinds of Executors mainly in
28	<	* that they provide <em>work-stealing</em>: all threads in the pool
29	<	* attempt to find and execute subtasks created by other active tasks
30	<	* (eventually blocking if none exist). This makes them efficient when
31	<	* most tasks spawn other subtasks (as do most ForkJoinTasks), as well
32	<	* as the mixed execution of some plain Runnable- or Callable- based
33	<	* activities along with ForkJoinTasks. When setting
34	<	* {@code setAsyncMode}, a ForkJoinPools may also be appropriate for
35	<	* use with fine-grained tasks that are never joined. Otherwise, other
36	<	* ExecutorService implementations are typically more appropriate
37	<	* 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). A {@code ForkJoinPool} may also be used for mixed
34	>	* execution of some plain {@code Runnable}- or {@code Callable}-
35	>	* based activities along with {@code ForkJoinTask}s. When setting
36	>	* {@linkplain #setAsyncMode async mode}, a {@code ForkJoinPool} may
37	>	* also be appropriate for use with fine-grained tasks of any form
38	>	* that are never joined. Otherwise, other {@code ExecutorService}
39	>	* implementations are typically more appropriate choices.
40		*
41	<	* <p>A ForkJoinPool may be constructed with a given parallelism level
42	<	* (target pool size), which it attempts to maintain by dynamically
43	<	* adding, suspending, or resuming threads, even if some tasks are
44	<	* waiting to join others. However, no such adjustments are performed
45	<	* in the face of blocked IO or other unmanaged synchronization. The
46	<	* nested {@code ManagedBlocker} interface enables extension of
47	<	* the kinds of synchronization accommodated.  The target parallelism
48	<	* level may also be changed dynamically ({@code setParallelism})
49	<	* and thread construction can be limited using methods
50	<	* {@code setMaximumPoolSize} and/or
51	<	* {@code setMaintainsParallelism}.
41	>	* <p>A {@code ForkJoinPool} is constructed with a given target
42	>	* parallelism level; by default, equal to the number of available
43	>	* processors. Unless configured otherwise via {@link
44	>	* #setMaintainsParallelism}, the pool attempts to maintain this
45	>	* number of active (or available) threads by dynamically adding,
46	>	* suspending, or resuming internal worker threads, even if some tasks
47	>	* are stalled waiting to join others. However, no such adjustments
48	>	* are performed in the face of blocked IO or other unmanaged
49	>	* synchronization. The nested {@link ManagedBlocker} interface
50	>	* enables extension of the kinds of synchronization accommodated.
51	>	* The target parallelism level may also be changed dynamically
52	>	* ({@link #setParallelism}). The total number of threads may be
53	>	* limited using method {@link #setMaximumPoolSize}, in which case it
54	>	* may become possible for the activities of a pool to stall due to
55	>	* the lack of available threads to process new tasks. When the pool
56	>	* is executing tasks, these and other configuration setting methods
57	>	* may only gradually affect actual pool sizes. It is normally best
58	>	* practice to invoke these methods only when the pool is known to be
59	>	* quiescent.
60		*
61		* <p>In addition to execution and lifecycle control methods, this
62		* class provides status check methods (for example
63	<	* {@code getStealCount}) that are intended to aid in developing,
63	>	* {@link #getStealCount}) that are intended to aid in developing,
64		* tuning, and monitoring fork/join applications. Also, method
65	<	* {@code toString} returns indications of pool state in a
65	>	* {@link #toString} returns indications of pool state in a
66		* convenient form for informal monitoring.
67		*
68	+	* <p><b>Sample Usage.</b> Normally a single {@code ForkJoinPool} is
69	+	* used for all parallel task execution in a program or subsystem.
70	+	* Otherwise, use would not usually outweigh the construction and
71	+	* bookkeeping overhead of creating a large set of threads. For
72	+	* example, a common pool could be used for the {@code SortTasks}
73	+	* illustrated in {@link RecursiveAction}. Because {@code
74	+	* ForkJoinPool} uses threads in {@linkplain java.lang.Thread#isDaemon
75	+	* daemon} mode, there is typically no need to explicitly {@link
76	+	* #shutdown} such a pool upon program exit.
77	+	*
78	+	* <pre>
79	+	* static final ForkJoinPool mainPool = new ForkJoinPool();
80	+	* ...
81	+	* public void sort(long[] array) {
82	+	*   mainPool.invoke(new SortTask(array, 0, array.length));
83	+	* }
84	+	* </pre>
85	+	*
86		* <p><b>Implementation notes</b>: This implementation restricts the
87		* maximum number of running threads to 32767. Attempts to create
88	<	* pools with greater than the maximum result in
89	<	* IllegalArgumentExceptions.
88	>	* pools with greater than the maximum number result in
89	>	* {@code IllegalArgumentException}.
90	>	*
91	>	* <p>This implementation rejects submitted tasks (that is, by throwing
92	>	* {@link RejectedExecutionException}) only when the pool is shut down.
93		*
94		* @since 1.7
95		* @author Doug Lea
#	Line 70 | Line 97 | import java.util.concurrent.atomic.Atomi
97		public class ForkJoinPool extends AbstractExecutorService {
98
99		/*
100	<	* See the extended comments interspersed below for design,
101	<	* rationale, and walkthroughs.
102	<	*/
103	<
104	<	/** Mask for packing and unpacking shorts */
105	<	private static final int  shortMask = 0xffff;
106	<
107	<	/** Max pool size -- must be a power of two minus 1 */
108	<	private static final int MAX_THREADS =  0x7FFF;
109	<
110	<	/**
111	<	* Factory for creating new ForkJoinWorkerThreads.  A
112	<	* ForkJoinWorkerThreadFactory must be defined and used for
113	<	* ForkJoinWorkerThread subclasses that extend base functionality
114	<	* or initialize threads with different contexts.
100	>	* Implementation Overview
101	>	*
102	>	* This class provides the central bookkeeping and control for a
103	>	* set of worker threads: Submissions from non-FJ threads enter
104	>	* into a submission queue. Workers take these tasks and typically
105	>	* split them into subtasks that may be stolen by other workers.
106	>	* The main work-stealing mechanics implemented in class
107	>	* ForkJoinWorkerThread give first priority to processing tasks
108	>	* from their own queues (LIFO or FIFO, depending on mode), then
109	>	* to randomized FIFO steals of tasks in other worker queues, and
110	>	* lastly to new submissions. These mechanics do not consider
111	>	* affinities, loads, cache localities, etc, so rarely provide the
112	>	* best possible performance on a given machine, but portably
113	>	* provide good throughput by averaging over these factors.
114	>	* (Further, even if we did try to use such information, we do not
115	>	* usually have a basis for exploiting it. For example, some sets
116	>	* of tasks profit from cache affinities, but others are harmed by
117	>	* cache pollution effects.)
118	>	*
119	>	* The main throughput advantages of work-stealing stem from
120	>	* decentralized control -- workers mostly steal tasks from each
121	>	* other. We do not want to negate this by creating bottlenecks
122	>	* implementing the management responsibilities of this class. So
123	>	* we use a collection of techniques that avoid, reduce, or cope
124	>	* well with contention. These entail several instances of
125	>	* bit-packing into CASable fields to maintain only the minimally
126	>	* required atomicity. To enable such packing, we restrict maximum
127	>	* parallelism to (1<<15)-1 (enabling twice this to fit into a 16
128	>	* bit field), which is far in excess of normal operating range.
129	>	* Even though updates to some of these bookkeeping fields do
130	>	* sometimes contend with each other, they don't normally
131	>	* cache-contend with updates to others enough to warrant memory
132	>	* padding or isolation. So they are all held as fields of
133	>	* ForkJoinPool objects.  The main capabilities are as follows:
134	>	*
135	>	* 1. Creating and removing workers. Workers are recorded in the
136	>	* "workers" array. This is an array as opposed to some other data
137	>	* structure to support index-based random steals by workers.
138	>	* Updates to the array recording new workers and unrecording
139	>	* terminated ones are protected from each other by a lock
140	>	* (workerLock) but the array is otherwise concurrently readable,
141	>	* and accessed directly by workers. To simplify index-based
142	>	* operations, the array size is always a power of two, and all
143	>	* readers must tolerate null slots. Currently, all worker thread
144	>	* creation is on-demand, triggered by task submissions,
145	>	* replacement of terminated workers, and/or compensation for
146	>	* blocked workers. However, all other support code is set up to
147	>	* work with other policies.
148	>	*
149	>	* 2. Bookkeeping for dynamically adding and removing workers. We
150	>	* maintain a given level of parallelism (or, if
151	>	* maintainsParallelism is false, at least avoid starvation). When
152	>	* some workers are known to be blocked (on joins or via
153	>	* ManagedBlocker), we may create or resume others to take their
154	>	* place until they unblock (see below). Implementing this
155	>	* requires counts of the number of "running" threads (i.e., those
156	>	* that are neither blocked nor artifically suspended) as well as
157	>	* the total number.  These two values are packed into one field,
158	>	* "workerCounts" because we need accurate snapshots when deciding
159	>	* to create, resume or suspend.  To support these decisions,
160	>	* updates to spare counts must be prospective (not
161	>	* retrospective).  For example, the running count is decremented
162	>	* before blocking by a thread about to block as a spare, but
163	>	* incremented by the thread about to unblock it. Updates upon
164	>	* resumption ofr threads blocking in awaitJoin or awaitBlocker
165	>	* cannot usually be prospective, so the running count is in
166	>	* general an upper bound of the number of productively running
167	>	* threads Updates to the workerCounts field sometimes transiently
168	>	* encounter a fair amount of contention when join dependencies
169	>	* are such that many threads block or unblock at about the same
170	>	* time. We alleviate this by sometimes bundling updates (for
171	>	* example blocking one thread on join and resuming a spare cancel
172	>	* each other out), and in most other cases performing an
173	>	* alternative action like releasing waiters or locating spares.
174	>	*
175	>	* 3. Maintaining global run state. The run state of the pool
176	>	* consists of a runLevel (SHUTDOWN, TERMINATING, etc) similar to
177	>	* those in other Executor implementations, as well as a count of
178	>	* "active" workers -- those that are, or soon will be, or
179	>	* recently were executing tasks. The runLevel and active count
180	>	* are packed together in order to correctly trigger shutdown and
181	>	* termination. Without care, active counts can be subject to very
182	>	* high contention.  We substantially reduce this contention by
183	>	* relaxing update rules.  A worker must claim active status
184	>	* prospectively, by activating if it sees that a submitted or
185	>	* stealable task exists (it may find after activating that the
186	>	* task no longer exists). It stays active while processing this
187	>	* task (if it exists) and any other local subtasks it produces,
188	>	* until it cannot find any other tasks. It then tries
189	>	* inactivating (see method preStep), but upon update contention
190	>	* instead scans for more tasks, later retrying inactivation if it
191	>	* doesn't find any.
192	>	*
193	>	* 4. Managing idle workers waiting for tasks. We cannot let
194	>	* workers spin indefinitely scanning for tasks when none are
195	>	* available. On the other hand, we must quickly prod them into
196	>	* action when new tasks are submitted or generated.  We
197	>	* park/unpark these idle workers using an event-count scheme.
198	>	* Field eventCount is incremented upon events that may enable
199	>	* workers that previously could not find a task to now find one:
200	>	* Submission of a new task to the pool, or another worker pushing
201	>	* a task onto a previously empty queue.  (We also use this
202	>	* mechanism for termination and reconfiguration actions that
203	>	* require wakeups of idle workers).  Each worker maintains its
204	>	* last known event count, and blocks when a scan for work did not
205	>	* find a task AND its lastEventCount matches the current
206	>	* eventCount. Waiting idle workers are recorded in a variant of
207	>	* Treiber stack headed by field eventWaiters which, when nonzero,
208	>	* encodes the thread index and count awaited for by the worker
209	>	* thread most recently calling eventSync. This thread in turn has
210	>	* a record (field nextEventWaiter) for the next waiting worker.
211	>	* In addition to allowing simpler decisions about need for
212	>	* wakeup, the event count bits in eventWaiters serve the role of
213	>	* tags to avoid ABA errors in Treiber stacks.  To reduce delays
214	>	* in task diffusion, workers not otherwise occupied may invoke
215	>	* method releaseWaiters, that removes and signals (unparks)
216	>	* workers not waiting on current count. To minimize task
217	>	* production stalls associate with signalling, any worker pushing
218	>	* a task on an empty queue invokes the weaker method signalWork,
219	>	* that only releases idle workers until it detects interference
220	>	* by other threads trying to release, and lets them take
221	>	* over. The net effect is a tree-like diffusion of signals, where
222	>	* released threads (and possibly others) help with unparks.  To
223	>	* further reduce contention effects a bit, failed CASes to
224	>	* increment field eventCount are tolerated without retries.
225	>	* Conceptually they are merged into the same event, which is OK
226	>	* when their only purpose is to enable workers to scan for work.
227	>	*
228	>	* 5. Managing suspension of extra workers. When a worker is about
229	>	* to block waiting for a join (or via ManagedBlockers), we may
230	>	* create a new thread to maintain parallelism level, or at least
231	>	* avoid starvation (see below). Usually, extra threads are needed
232	>	* for only very short periods, yet join dependencies are such
233	>	* that we sometimes need them in bursts. Rather than create new
234	>	* threads each time this happens, we suspend no-longer-needed
235	>	* extra ones as "spares". For most purposes, we don't distinguish
236	>	* "extra" spare threads from normal "core" threads: On each call
237	>	* to preStep (the only point at which we can do this) a worker
238	>	* checks to see if there are now too many running workers, and if
239	>	* so, suspends itself.  Methods awaitJoin and awaitBlocker look
240	>	* for suspended threads to resume before considering creating a
241	>	* new replacement. We don't need a special data structure to
242	>	* maintain spares; simply scanning the workers array looking for
243	>	* worker.isSuspended() is fine because the calling thread is
244	>	* otherwise not doing anything useful anyway; we are at least as
245	>	* happy if after locating a spare, the caller doesn't actually
246	>	* block because the join is ready before we try to adjust and
247	>	* compensate.  Note that this is intrinsically racy.  One thread
248	>	* may become a spare at about the same time as another is
249	>	* needlessly being created. We counteract this and related slop
250	>	* in part by requiring resumed spares to immediately recheck (in
251	>	* preStep) to see whether they they should re-suspend. The only
252	>	* effective difference between "extra" and "core" threads is that
253	>	* we allow the "extra" ones to time out and die if they are not
254	>	* resumed within a keep-alive interval of a few seconds. This is
255	>	* implemented mainly within ForkJoinWorkerThread, but requires
256	>	* some coordination (isTrimmed() -- meaning killed while
257	>	* suspended) to correctly maintain pool counts.
258	>	*
259	>	* 6. Deciding when to create new workers. The main dynamic
260	>	* control in this class is deciding when to create extra threads,
261	>	* in methods awaitJoin and awaitBlocker. We always
262	>	* need to create one when the number of running threads becomes
263	>	* zero. But because blocked joins are typically dependent, we
264	>	* don't necessarily need or want one-to-one replacement. Using a
265	>	* one-to-one compensation rule often leads to enough useless
266	>	* overhead creating, suspending, resuming, and/or killing threads
267	>	* to signficantly degrade throughput.  We use a rule reflecting
268	>	* the idea that, the more spare threads you already have, the
269	>	* more evidence you need to create another one. The "evidence"
270	>	* here takes two forms: (1) Using a creation threshold expressed
271	>	* in terms of the current deficit -- target minus running
272	>	* threads. To reduce flickering and drift around target values,
273	>	* the relation is quadratic: adding a spare if (dc*dc)>=(sc*pc)
274	>	* (where dc is deficit, sc is number of spare threads and pc is
275	>	* target parallelism.)  (2) Using a form of adaptive
276	>	* spionning. requiring a number of threshold checks proportional
277	>	* to the number of spare threads.  This effectively reduces churn
278	>	* at the price of systematically undershooting target parallelism
279	>	* when many threads are blocked.  However, biasing toward
280	>	* undeshooting partially compensates for the above mechanics to
281	>	* suspend extra threads, that normally lead to overshoot because
282	>	* we can only suspend workers in-between top-level actions. It
283	>	* also better copes with the fact that some of the methods in
284	>	* this class tend to never become compiled (but are interpreted),
285	>	* so some components of the entire set of controls might execute
286	>	* many times faster than others. And similarly for cases where
287	>	* the apparent lack of work is just due to GC stalls and other
288	>	* transient system activity.
289	>	*
290	>	* 7. Maintaining other configuration parameters and monitoring
291	>	* statistics. Updates to fields controlling parallelism level,
292	>	* max size, etc can only meaningfully take effect for individual
293	>	* threads upon their next top-level actions; i.e., between
294	>	* stealing/running tasks/submission, which are separated by calls
295	>	* to preStep.  Memory ordering for these (assumed infrequent)
296	>	* reconfiguration calls is ensured by using reads and writes to
297	>	* volatile field workerCounts (that must be read in preStep anyway)
298	>	* as "fences" -- user-level reads are preceded by reads of
299	>	* workCounts, and writes are followed by no-op CAS to
300	>	* workerCounts. The values reported by other management and
301	>	* monitoring methods are either computed on demand, or are kept
302	>	* in fields that are only updated when threads are otherwise
303	>	* idle.
304	>	*
305	>	* Beware that there is a lot of representation-level coupling
306	>	* among classes ForkJoinPool, ForkJoinWorkerThread, and
307	>	* ForkJoinTask.  For example, direct access to "workers" array by
308	>	* workers, and direct access to ForkJoinTask.status by both
309	>	* ForkJoinPool and ForkJoinWorkerThread.  There is little point
310	>	* trying to reduce this, since any associated future changes in
311	>	* representations will need to be accompanied by algorithmic
312	>	* changes anyway.
313	>	*
314	>	* Style notes: There are lots of inline assignments (of form
315	>	* "while ((local = field) != 0)") which are usually the simplest
316	>	* way to ensure read orderings. Also several occurrences of the
317	>	* unusual "do {} while(!cas...)" which is the simplest way to
318	>	* force an update of a CAS'ed variable. There are also a few
319	>	* other coding oddities that help some methods perform reasonably
320	>	* even when interpreted (not compiled).
321	>	*
322	>	* The order of declarations in this file is: (1) statics (2)
323	>	* fields (along with constants used when unpacking some of them)
324	>	* (3) internal control methods (4) callbacks and other support
325	>	* for ForkJoinTask and ForkJoinWorkerThread classes, (5) exported
326	>	* methods (plus a few little helpers).
327	>	*/
328	>
329	>	/**
330	>	* Factory for creating new {@link ForkJoinWorkerThread}s.
331	>	* A {@code ForkJoinWorkerThreadFactory} must be defined and used
332	>	* for {@code ForkJoinWorkerThread} subclasses that extend base
333	>	* functionality or initialize threads with different contexts.
334		*/
335		public static interface ForkJoinWorkerThreadFactory {
336		/**
337		* Returns a new worker thread operating in the given pool.
338		*
339		* @param pool the pool this thread works in
340	<	* @throws NullPointerException if pool is null
340	>	* @throws NullPointerException if the pool is null
341		*/
342		public ForkJoinWorkerThread newThread(ForkJoinPool pool);
343		}
#	Line 103 | Line 349 | public class ForkJoinPool extends Abstra
349		static class  DefaultForkJoinWorkerThreadFactory
350		implements ForkJoinWorkerThreadFactory {
351		public ForkJoinWorkerThread newThread(ForkJoinPool pool) {
352	<	try {
107	<	return new ForkJoinWorkerThread(pool);
108	<	} catch (OutOfMemoryError oom)  {
109	<	return null;
110	<	}
352	>	return new ForkJoinWorkerThread(pool);
353		}
354		}
355
#	Line 143 | Line 385 | public class ForkJoinPool extends Abstra
385		new AtomicInteger();
386
387		/**
388	<	* Array holding all worker threads in the pool. Initialized upon
389	<	* first use. Array size must be a power of two.  Updates and
148	<	* replacements are protected by workerLock, but it is always kept
149	<	* in a consistent enough state to be randomly accessed without
150	<	* locking by workers performing work-stealing.
388	>	* Absolute bound for parallelism level. Twice this number must
389	>	* fit into a 16bit field to enable word-packing for some counts.
390		*/
391	<	volatile ForkJoinWorkerThread[] workers;
391	>	private static final int MAX_THREADS = 0x7fff;
392
393		/**
394	<	* Lock protecting access to workers.
394	>	* Array holding all worker threads in the pool.  Array size must
395	>	* be a power of two.  Updates and replacements are protected by
396	>	* workerLock, but the array is always kept in a consistent enough
397	>	* state to be randomly accessed without locking by workers
398	>	* performing work-stealing, as well as other traversal-based
399	>	* methods in this class. All readers must tolerate that some
400	>	* array slots may be null.
401		*/
402	<	private final ReentrantLock workerLock;
402	>	volatile ForkJoinWorkerThread[] workers;
403
404		/**
405	<	* Condition for awaitTermination.
405	>	* Queue for external submissions.
406		*/
407	<	private final Condition termination;
407	>	private final LinkedTransferQueue<ForkJoinTask<?>> submissionQueue;
408
409		/**
410	<	* The uncaught exception handler used when any worker
166	<	* abruptly terminates
410	>	* Lock protecting updates to workers array.
411		*/
412	<	private Thread.UncaughtExceptionHandler ueh;
412	>	private final ReentrantLock workerLock;
413
414		/**
415	<	* Creation factory for worker threads.
415	>	* Latch released upon termination.
416		*/
417	<	private final ForkJoinWorkerThreadFactory factory;
417	>	private final CountDownLatch terminationLatch;
418
419		/**
420	<	* Head of stack of threads that were created to maintain
177	<	* parallelism when other threads blocked, but have since
178	<	* suspended when the parallelism level rose.
420	>	* Creation factory for worker threads.
421		*/
422	<	private volatile WaitQueueNode spareStack;
422	>	private final ForkJoinWorkerThreadFactory factory;
423
424		/**
425		* Sum of per-thread steal counts, updated only when threads are
426		* idle or terminating.
427		*/
428	<	private final AtomicLong stealCount;
428	>	private volatile long stealCount;
429
430		/**
431	<	* Queue for external submissions.
431	>	* Encoded record of top of treiber stack of threads waiting for
432	>	* events. The top 32 bits contain the count being waited for. The
433	>	* bottom word contains one plus the pool index of waiting worker
434	>	* thread.
435		*/
436	<	private final LinkedTransferQueue<ForkJoinTask<?>> submissionQueue;
436	>	private volatile long eventWaiters;
437	>
438	>	private static final int  EVENT_COUNT_SHIFT = 32;
439	>	private static final long WAITER_INDEX_MASK = (1L << EVENT_COUNT_SHIFT)-1L;
440
441		/**
442	<	* Head of Treiber stack for barrier sync. See below for explanation.
442	>	* A counter for events that may wake up worker threads:
443	>	*   - Submission of a new task to the pool
444	>	*   - A worker pushing a task on an empty queue
445	>	*   - termination and reconfiguration
446		*/
447	<	private volatile WaitQueueNode syncStack;
447	>	private volatile int eventCount;
448
449		/**
450	<	* The count for event barrier
451	<	*/
452	<	private volatile long eventCount;
450	>	* Lifecycle control. The low word contains the number of workers
451	>	* that are (probably) executing tasks. This value is atomically
452	>	* incremented before a worker gets a task to run, and decremented
453	>	* when worker has no tasks and cannot find any.  Bits 16-18
454	>	* contain runLevel value. When all are zero, the pool is
455	>	* running. Level transitions are monotonic (running -> shutdown
456	>	* -> terminating -> terminated) so each transition adds a bit.
457	>	* These are bundled together to ensure consistent read for
458	>	* termination checks (i.e., that runLevel is at least SHUTDOWN
459	>	* and active threads is zero).
460	>	*/
461	>	private volatile int runState;
462	>
463	>	// Note: The order among run level values matters.
464	>	private static final int RUNLEVEL_SHIFT     = 16;
465	>	private static final int SHUTDOWN           = 1 << RUNLEVEL_SHIFT;
466	>	private static final int TERMINATING        = 1 << (RUNLEVEL_SHIFT + 1);
467	>	private static final int TERMINATED         = 1 << (RUNLEVEL_SHIFT + 2);
468	>	private static final int ACTIVE_COUNT_MASK  = (1 << RUNLEVEL_SHIFT) - 1;
469	>	private static final int ONE_ACTIVE         = 1; // active update delta
470
471		/**
472	<	* Pool number, just for assigning useful names to worker threads
472	>	* Holds number of total (i.e., created and not yet terminated)
473	>	* and running (i.e., not blocked on joins or other managed sync)
474	>	* threads, packed together to ensure consistent snapshot when
475	>	* making decisions about creating and suspending spare
476	>	* threads. Updated only by CAS. Note that adding a new worker
477	>	* requires incrementing both counts, since workers start off in
478	>	* running state.  This field is also used for memory-fencing
479	>	* configuration parameters.
480	>	*/
481	>	private volatile int workerCounts;
482	>
483	>	private static final int TOTAL_COUNT_SHIFT  = 16;
484	>	private static final int RUNNING_COUNT_MASK = (1 << TOTAL_COUNT_SHIFT) - 1;
485	>	private static final int ONE_RUNNING        = 1;
486	>	private static final int ONE_TOTAL          = 1 << TOTAL_COUNT_SHIFT;
487	>
488	>	/*
489	>	* Fields parallelism. maxPoolSize, and maintainsParallelism are
490	>	* non-volatile, but external reads/writes use workerCount fences
491	>	* to ensure visability.
492		*/
206	–	private final int poolNumber;
493
494		/**
495	<	* The maximum allowed pool size
495	>	* The target parallelism level.
496		*/
497	<	private volatile int maxPoolSize;
497	>	private int parallelism;
498
499		/**
500	<	* The desired parallelism level, updated only under workerLock.
500	>	* The maximum allowed pool size.
501		*/
502	<	private volatile int parallelism;
502	>	private int maxPoolSize;
503
504		/**
505		* True if use local fifo, not default lifo, for local polling
506	+	* Replicated by ForkJoinWorkerThreads
507		*/
508		private volatile boolean locallyFifo;
509
510		/**
511	<	* Holds number of total (i.e., created and not yet terminated)
225	<	* and running (i.e., not blocked on joins or other managed sync)
226	<	* threads, packed into one int to ensure consistent snapshot when
227	<	* making decisions about creating and suspending spare
228	<	* threads. Updated only by CAS.  Note: CASes in
229	<	* updateRunningCount and preJoin assume that running active count
230	<	* is in low word, so need to be modified if this changes.
511	>	* Controls whether to add spares to maintain parallelism
512		*/
513	<	private volatile int workerCounts;
513	>	private boolean maintainsParallelism;
514
515	<	private static int totalCountOf(int s)           { return s >>> 16;  }
516	<	private static int runningCountOf(int s)         { return s & shortMask; }
517	<	private static int workerCountsFor(int t, int r) { return (t << 16) + r; }
515	>	/**
516	>	* The uncaught exception handler used when any worker
517	>	* abruptly terminates
518	>	*/
519	>	private volatile Thread.UncaughtExceptionHandler ueh;
520
521		/**
522	<	* Adds delta (which may be negative) to running count.  This must
523	<	* be called before (with negative arg) and after (with positive)
524	<	* any managed synchronization (i.e., mainly, joins).
525	<	*
526	<	* @param delta the number to add
522	>	* Pool number, just for assigning useful names to worker threads
523	>	*/
524	>	private final int poolNumber;
525	>
526	>	// utilities for updating fields
527	>
528	>	/**
529	>	* Adds delta to running count.  Used mainly by ForkJoinTask.
530		*/
531		final void updateRunningCount(int delta) {
532	<	int s;
533	<	do {} while (!casWorkerCounts(s = workerCounts, s + delta));
532	>	int wc;
533	>	do {} while (!UNSAFE.compareAndSwapInt(this, workerCountsOffset,
534	>	wc = workerCounts,
535	>	wc + delta));
536		}
537
538		/**
539	<	* Adds delta (which may be negative) to both total and running
252	<	* count.  This must be called upon creation and termination of
253	<	* worker threads.
254	<	*
255	<	* @param delta the number to add
539	>	* Decrements running count unless already zero
540		*/
541	<	private void updateWorkerCount(int delta) {
542	<	int d = delta + (delta << 16); // add to both lo and hi parts
543	<	int s;
544	<	do {} while (!casWorkerCounts(s = workerCounts, s + d));
541	>	final boolean tryDecrementRunningCount() {
542	>	int wc = workerCounts;
543	>	if ((wc & RUNNING_COUNT_MASK) == 0)
544	>	return false;
545	>	return UNSAFE.compareAndSwapInt(this, workerCountsOffset,
546	>	wc, wc - ONE_RUNNING);
547		}
548
549		/**
550	<	* Lifecycle control. High word contains runState, low word
551	<	* contains the number of workers that are (probably) executing
266	<	* tasks. This value is atomically incremented before a worker
267	<	* gets a task to run, and decremented when worker has no tasks
268	<	* and cannot find any. These two fields are bundled together to
269	<	* support correct termination triggering.  Note: activeCount
270	<	* CAS'es cheat by assuming active count is in low word, so need
271	<	* to be modified if this changes
550	>	* Write fence for user modifications of pool parameters
551	>	* (parallelism. etc).  Note that it doesn't matter if CAS fails.
552		*/
553	<	private volatile int runControl;
554	<
555	<	// RunState values. Order among values matters
556	<	private static final int RUNNING     = 0;
557	<	private static final int SHUTDOWN    = 1;
278	<	private static final int TERMINATING = 2;
279	<	private static final int TERMINATED  = 3;
553	>	private void workerCountWriteFence() {
554	>	int wc;
555	>	UNSAFE.compareAndSwapInt(this, workerCountsOffset,
556	>	wc = workerCounts, wc);
557	>	}
558
559	<	private static int runStateOf(int c)             { return c >>> 16; }
560	<	private static int activeCountOf(int c)          { return c & shortMask; }
561	<	private static int runControlFor(int r, int a)   { return (r << 16) + a; }
559	>	/**
560	>	* Read fence for external reads of pool parameters
561	>	* (parallelism. maxPoolSize, etc).
562	>	*/
563	>	private void workerCountReadFence() {
564	>	int ignore = workerCounts;
565	>	}
566
567		/**
568		* Tries incrementing active count; fails on contention.
569	<	* Called by workers before/during executing tasks.
569	>	* Called by workers before executing tasks.
570		*
571		* @return true on success
572		*/
573		final boolean tryIncrementActiveCount() {
574	<	int c = runControl;
575	<	return casRunControl(c, c+1);
574	>	int c;
575	>	return UNSAFE.compareAndSwapInt(this, runStateOffset,
576	>	c = runState, c + ONE_ACTIVE);
577		}
578
579		/**
580		* Tries decrementing active count; fails on contention.
581	<	* Possibly triggers termination on success.
299	<	* Called by workers when they can't find tasks.
300	<	*
301	<	* @return true on success
581	>	* Called when workers cannot find tasks to run.
582		*/
583		final boolean tryDecrementActiveCount() {
584	<	int c = runControl;
585	<	int nextc = c - 1;
586	<	if (!casRunControl(c, nextc))
307	<	return false;
308	<	if (canTerminateOnShutdown(nextc))
309	<	terminateOnShutdown();
310	<	return true;
584	>	int c;
585	>	return UNSAFE.compareAndSwapInt(this, runStateOffset,
586	>	c = runState, c - ONE_ACTIVE);
587		}
588
589		/**
590	<	* Returns true if argument represents zero active count and
591	<	* nonzero runstate, which is the triggering condition for
316	<	* terminating on shutdown.
590	>	* Advances to at least the given level. Returns true if not
591	>	* already in at least the given level.
592		*/
593	<	private static boolean canTerminateOnShutdown(int c) {
594	<	// i.e. least bit is nonzero runState bit
595	<	return ((c & -c) >>> 16) != 0;
593	>	private boolean advanceRunLevel(int level) {
594	>	for (;;) {
595	>	int s = runState;
596	>	if ((s & level) != 0)
597	>	return false;
598	>	if (UNSAFE.compareAndSwapInt(this, runStateOffset, s, s | level))
599	>	return true;
600	>	}
601		}
602
603	+	// workers array maintenance
604	+
605		/**
606	<	* Transition run state to at least the given state. Return true
325	<	* if not already at least given state.
606	>	* Records and returns a workers array index for new worker.
607		*/
608	<	private boolean transitionRunStateTo(int state) {
608	>	private int recordWorker(ForkJoinWorkerThread w) {
609	>	// Try using slot totalCount-1. If not available, scan and/or resize
610	>	int k = (workerCounts >>> TOTAL_COUNT_SHIFT) - 1;
611	>	final ReentrantLock lock = this.workerLock;
612	>	lock.lock();
613	>	try {
614	>	ForkJoinWorkerThread[] ws = workers;
615	>	int nws = ws.length;
616	>	if (k < 0 || k >= nws || ws[k] != null) {
617	>	for (k = 0; k < nws && ws[k] != null; ++k)
618	>	;
619	>	if (k == nws)
620	>	ws = Arrays.copyOf(ws, nws << 1);
621	>	}
622	>	ws[k] = w;
623	>	workers = ws; // volatile array write ensures slot visibility
624	>	} finally {
625	>	lock.unlock();
626	>	}
627	>	return k;
628	>	}
629	>
630	>	/**
631	>	* Nulls out record of worker in workers array
632	>	*/
633	>	private void forgetWorker(ForkJoinWorkerThread w) {
634	>	int idx = w.poolIndex;
635	>	// Locking helps method recordWorker avoid unecessary expansion
636	>	final ReentrantLock lock = this.workerLock;
637	>	lock.lock();
638	>	try {
639	>	ForkJoinWorkerThread[] ws = workers;
640	>	if (idx >= 0 && idx < ws.length && ws[idx] == w) // verify
641	>	ws[idx] = null;
642	>	} finally {
643	>	lock.unlock();
644	>	}
645	>	}
646	>
647	>	// adding and removing workers
648	>
649	>	/**
650	>	* Tries to create and add new worker. Assumes that worker counts
651	>	* are already updated to accommodate the worker, so adjusts on
652	>	* failure.
653	>	*
654	>	* @return new worker or null if creation failed
655	>	*/
656	>	private ForkJoinWorkerThread addWorker() {
657	>	ForkJoinWorkerThread w = null;
658	>	try {
659	>	w = factory.newThread(this);
660	>	} finally { // Adjust on either null or exceptional factory return
661	>	if (w == null) {
662	>	onWorkerCreationFailure();
663	>	return null;
664	>	}
665	>	}
666	>	w.start(recordWorker(w), locallyFifo, ueh);
667	>	return w;
668	>	}
669	>
670	>	/**
671	>	* Adjusts counts upon failure to create worker
672	>	*/
673	>	private void onWorkerCreationFailure() {
674		for (;;) {
675	<	int c = runControl;
676	<	if (runStateOf(c) >= state)
677	<	return false;
678	<	if (casRunControl(c, runControlFor(state, activeCountOf(c))))
675	>	int wc = workerCounts;
676	>	if ((wc >>> TOTAL_COUNT_SHIFT) > 0 &&
677	>	UNSAFE.compareAndSwapInt(this, workerCountsOffset,
678	>	wc, wc - (ONE_RUNNING|ONE_TOTAL)))
679	>	break;
680	>	}
681	>	tryTerminate(false); // in case of failure during shutdown
682	>	}
683	>
684	>	/**
685	>	* Create enough total workers to establish target parallelism,
686	>	* giving up if terminating or addWorker fails
687	>	*/
688	>	private void ensureEnoughTotalWorkers() {
689	>	int wc;
690	>	while (((wc = workerCounts) >>> TOTAL_COUNT_SHIFT) < parallelism &&
691	>	runState < TERMINATING) {
692	>	if ((UNSAFE.compareAndSwapInt(this, workerCountsOffset,
693	>	wc, wc + (ONE_RUNNING|ONE_TOTAL)) &&
694	>	addWorker() == null))
695	>	break;
696	>	}
697	>	}
698	>
699	>	/**
700	>	* Final callback from terminating worker.  Removes record of
701	>	* worker from array, and adjusts counts. If pool is shutting
702	>	* down, tries to complete terminatation, else possibly replaces
703	>	* the worker.
704	>	*
705	>	* @param w the worker
706	>	*/
707	>	final void workerTerminated(ForkJoinWorkerThread w) {
708	>	if (w.active) { // force inactive
709	>	w.active = false;
710	>	do {} while (!tryDecrementActiveCount());
711	>	}
712	>	forgetWorker(w);
713	>
714	>	// Decrement total count, and if was running, running count
715	>	// Spin (waiting for other updates) if either would be negative
716	>	int nr = w.isTrimmed() ? 0 : ONE_RUNNING;
717	>	int unit = ONE_TOTAL + nr;
718	>	for (;;) {
719	>	int wc = workerCounts;
720	>	int rc = wc & RUNNING_COUNT_MASK;
721	>	if (rc - nr < 0 || (wc >>> TOTAL_COUNT_SHIFT) == 0)
722	>	Thread.yield(); // back off if waiting for other updates
723	>	else if (UNSAFE.compareAndSwapInt(this, workerCountsOffset,
724	>	wc, wc - unit))
725	>	break;
726	>	}
727	>
728	>	accumulateStealCount(w); // collect final count
729	>	if (!tryTerminate(false))
730	>	ensureEnoughTotalWorkers();
731	>	}
732	>
733	>	// Waiting for and signalling events
734	>
735	>	/**
736	>	* Ensures eventCount on exit is different (mod 2^32) than on
737	>	* entry.  CAS failures are OK -- any change in count suffices.
738	>	*/
739	>	private void advanceEventCount() {
740	>	int c;
741	>	UNSAFE.compareAndSwapInt(this, eventCountOffset, c = eventCount, c+1);
742	>	}
743	>
744	>	/**
745	>	* Releases workers blocked on a count not equal to current count.
746	>	*/
747	>	final void releaseWaiters() {
748	>	long top;
749	>	int id;
750	>	while ((id = (int)((top = eventWaiters) & WAITER_INDEX_MASK)) > 0 &&
751	>	(int)(top >>> EVENT_COUNT_SHIFT) != eventCount) {
752	>	ForkJoinWorkerThread[] ws = workers;
753	>	ForkJoinWorkerThread w;
754	>	if (ws.length >= id && (w = ws[id - 1]) != null &&
755	>	UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
756	>	top, w.nextWaiter))
757	>	LockSupport.unpark(w);
758	>	}
759	>	}
760	>
761	>	/**
762	>	* Advances eventCount and releases waiters until interference by
763	>	* other releasing threads is detected.
764	>	*/
765	>	final void signalWork() {
766	>	int ec;
767	>	UNSAFE.compareAndSwapInt(this, eventCountOffset, ec=eventCount, ec+1);
768	>	outer:for (;;) {
769	>	long top = eventWaiters;
770	>	ec = eventCount;
771	>	for (;;) {
772	>	ForkJoinWorkerThread[] ws; ForkJoinWorkerThread w;
773	>	int id = (int)(top & WAITER_INDEX_MASK);
774	>	if (id <= 0 || (int)(top >>> EVENT_COUNT_SHIFT) == ec)
775	>	return;
776	>	if ((ws = workers).length < id || (w = ws[id - 1]) == null ||
777	>	!UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
778	>	top, top = w.nextWaiter))
779	>	continue outer;      // possibly stale; reread
780	>	LockSupport.unpark(w);
781	>	if (top != eventWaiters) // let someone else take over
782	>	return;
783	>	}
784	>	}
785	>	}
786	>
787	>	/**
788	>	* If worker is inactive, blocks until terminating or event count
789	>	* advances from last value held by worker; in any case helps
790	>	* release others.
791	>	*
792	>	* @param w the calling worker thread
793	>	*/
794	>	private void eventSync(ForkJoinWorkerThread w) {
795	>	if (!w.active) {
796	>	int prev = w.lastEventCount;
797	>	long nextTop = (((long)prev << EVENT_COUNT_SHIFT) |
798	>	((long)(w.poolIndex + 1)));
799	>	long top;
800	>	while ((runState < SHUTDOWN || !tryTerminate(false)) &&
801	>	(((int)(top = eventWaiters) & WAITER_INDEX_MASK) == 0 ||
802	>	(int)(top >>> EVENT_COUNT_SHIFT) == prev) &&
803	>	eventCount == prev) {
804	>	if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
805	>	w.nextWaiter = top, nextTop)) {
806	>	accumulateStealCount(w); // transfer steals while idle
807	>	Thread.interrupted();    // clear/ignore interrupt
808	>	while (eventCount == prev)
809	>	w.doPark();
810	>	break;
811	>	}
812	>	}
813	>	w.lastEventCount = eventCount;
814	>	}
815	>	releaseWaiters();
816	>	}
817	>
818	>	/**
819	>	* Callback from workers invoked upon each top-level action (i.e.,
820	>	* stealing a task or taking a submission and running
821	>	* it). Performs one or both of the following:
822	>	*
823	>	* * If the worker cannot find work, updates its active status to
824	>	* inactive and updates activeCount unless there is contention, in
825	>	* which case it may try again (either in this or a subsequent
826	>	* call).  Additionally, awaits the next task event and/or helps
827	>	* wake up other releasable waiters.
828	>	*
829	>	* * If there are too many running threads, suspends this worker
830	>	* (first forcing inactivation if necessary).  If it is not
831	>	* resumed before a keepAlive elapses, the worker may be "trimmed"
832	>	* -- killed while suspended within suspendAsSpare. Otherwise,
833	>	* upon resume it rechecks to make sure that it is still needed.
834	>	*
835	>	* @param w the worker
836	>	* @param worked false if the worker scanned for work but didn't
837	>	* find any (in which case it may block waiting for work).
838	>	*/
839	>	final void preStep(ForkJoinWorkerThread w, boolean worked) {
840	>	boolean active = w.active;
841	>	boolean inactivate = !worked & active;
842	>	for (;;) {
843	>	if (inactivate) {
844	>	int c = runState;
845	>	if (UNSAFE.compareAndSwapInt(this, runStateOffset,
846	>	c, c - ONE_ACTIVE))
847	>	inactivate = active = w.active = false;
848	>	}
849	>	int wc = workerCounts;
850	>	if ((wc & RUNNING_COUNT_MASK) <= parallelism) {
851	>	if (!worked)
852	>	eventSync(w);
853	>	return;
854	>	}
855	>	if (!(inactivate |= active) &&  // must inactivate to suspend
856	>	UNSAFE.compareAndSwapInt(this, workerCountsOffset,
857	>	wc, wc - ONE_RUNNING) &&
858	>	!w.suspendAsSpare())        // false if trimmed
859	>	return;
860	>	}
861	>	}
862	>
863	>	/**
864	>	* Adjusts counts and creates or resumes compensating threads for
865	>	* a worker blocking on task joinMe.  First tries resuming an
866	>	* existing spare (which usually also avoids any count
867	>	* adjustment), but must then decrement running count to determine
868	>	* whether a new thread is needed. See above for fuller
869	>	* explanation. This code is sprawled out non-modularly mainly
870	>	* because adaptive spinning works best if the entire method is
871	>	* either interpreted or compiled vs having only some pieces of it
872	>	* compiled.
873	>	*
874	>	* @param joinMe the task to join
875	>	* @return task status on exit (to simplify usage by callers)
876	>	*/
877	>	final int awaitJoin(ForkJoinTask<?> joinMe) {
878	>	int pc = parallelism;
879	>	boolean adj = false;        // true when running count adjusted
880	>	int scans = 0;
881	>
882	>	while (joinMe.status >= 0) {
883	>	ForkJoinWorkerThread spare = null;
884	>	if ((workerCounts & RUNNING_COUNT_MASK) < pc) {
885	>	ForkJoinWorkerThread[] ws = workers;
886	>	int nws = ws.length;
887	>	for (int i = 0; i < nws; ++i) {
888	>	ForkJoinWorkerThread w = ws[i];
889	>	if (w != null && w.isSuspended()) {
890	>	spare = w;
891	>	break;
892	>	}
893	>	}
894	>	if (joinMe.status < 0)
895	>	break;
896	>	}
897	>	int wc = workerCounts;
898	>	int rc = wc & RUNNING_COUNT_MASK;
899	>	int dc = pc - rc;
900	>	if (dc > 0 && spare != null && spare.tryUnsuspend()) {
901	>	if (adj) {
902	>	int c;
903	>	do {} while (!UNSAFE.compareAndSwapInt
904	>	(this, workerCountsOffset,
905	>	c = workerCounts, c + ONE_RUNNING));
906	>	}
907	>	adj = true;
908	>	LockSupport.unpark(spare);
909	>	}
910	>	else if (adj) {
911	>	if (dc <= 0)
912	>	break;
913	>	int tc = wc >>> TOTAL_COUNT_SHIFT;
914	>	if (scans > tc) {
915	>	int ts = (tc - pc) * pc;
916	>	if (rc != 0 &&  (dc * dc < ts || !maintainsParallelism))
917	>	break;
918	>	if (scans > ts && tc < maxPoolSize &&
919	>	UNSAFE.compareAndSwapInt(this, workerCountsOffset, wc,
920	>	wc+(ONE_RUNNING|ONE_TOTAL))){
921	>	addWorker();
922	>	break;
923	>	}
924	>	}
925	>	}
926	>	else if (rc != 0)
927	>	adj = UNSAFE.compareAndSwapInt (this, workerCountsOffset,
928	>	wc, wc - ONE_RUNNING);
929	>	if ((scans++ & 1) == 0)
930	>	releaseWaiters();   // help others progress
931	>	else
932	>	Thread.yield();     // avoid starving productive threads
933	>	}
934	>
935	>	if (adj) {
936	>	joinMe.internalAwaitDone();
937	>	int c;
938	>	do {} while (!UNSAFE.compareAndSwapInt
939	>	(this, workerCountsOffset,
940	>	c = workerCounts, c + ONE_RUNNING));
941	>	}
942	>	return joinMe.status;
943	>	}
944	>
945	>	/**
946	>	* Same idea as awaitJoin
947	>	*/
948	>	final void awaitBlocker(ManagedBlocker blocker, boolean maintainPar)
949	>	throws InterruptedException {
950	>	maintainPar &= maintainsParallelism;
951	>	int pc = parallelism;
952	>	boolean adj = false;        // true when running count adjusted
953	>	int scans = 0;
954	>	boolean done;
955	>
956	>	for (;;) {
957	>	if (done = blocker.isReleasable())
958	>	break;
959	>	ForkJoinWorkerThread spare = null;
960	>	if ((workerCounts & RUNNING_COUNT_MASK) < pc) {
961	>	ForkJoinWorkerThread[] ws = workers;
962	>	int nws = ws.length;
963	>	for (int i = 0; i < nws; ++i) {
964	>	ForkJoinWorkerThread w = ws[i];
965	>	if (w != null && w.isSuspended()) {
966	>	spare = w;
967	>	break;
968	>	}
969	>	}
970	>	if (done = blocker.isReleasable())
971	>	break;
972	>	}
973	>	int wc = workerCounts;
974	>	int rc = wc & RUNNING_COUNT_MASK;
975	>	int dc = pc - rc;
976	>	if (dc > 0 && spare != null && spare.tryUnsuspend()) {
977	>	if (adj) {
978	>	int c;
979	>	do {} while (!UNSAFE.compareAndSwapInt
980	>	(this, workerCountsOffset,
981	>	c = workerCounts, c + ONE_RUNNING));
982	>	}
983	>	adj = true;
984	>	LockSupport.unpark(spare);
985	>	}
986	>	else if (adj) {
987	>	if (dc <= 0)
988	>	break;
989	>	int tc = wc >>> TOTAL_COUNT_SHIFT;
990	>	if (scans > tc) {
991	>	int ts = (tc - pc) * pc;
992	>	if (rc != 0 &&  (dc * dc < ts || !maintainPar))
993	>	break;
994	>	if (scans > ts && tc < maxPoolSize &&
995	>	UNSAFE.compareAndSwapInt(this, workerCountsOffset, wc,
996	>	wc+(ONE_RUNNING|ONE_TOTAL))){
997	>	addWorker();
998	>	break;
999	>	}
1000	>	}
1001	>	}
1002	>	else if (rc != 0)
1003	>	adj = UNSAFE.compareAndSwapInt (this, workerCountsOffset,
1004	>	wc, wc - ONE_RUNNING);
1005	>	if ((++scans & 1) == 0)
1006	>	releaseWaiters();   // help others progress
1007	>	else
1008	>	Thread.yield();     // avoid starving productive threads
1009	>	}
1010	>
1011	>	try {
1012	>	if (!done)
1013	>	do {} while (!blocker.isReleasable() && !blocker.block());
1014	>	} finally {
1015	>	if (adj) {
1016	>	int c;
1017	>	do {} while (!UNSAFE.compareAndSwapInt
1018	>	(this, workerCountsOffset,
1019	>	c = workerCounts, c + ONE_RUNNING));
1020	>	}
1021	>	}
1022	>	}
1023	>
1024	>	/**
1025	>	* Unless there are not enough other running threads, adjusts
1026	>	* counts and blocks a worker performing helpJoin that cannot find
1027	>	* any work.
1028	>	*
1029	>	* @return true if joinMe now done
1030	>	*/
1031	>	final boolean tryAwaitBusyJoin(ForkJoinTask<?> joinMe) {
1032	>	int pc = parallelism;
1033	>	outer:for (;;) {
1034	>	releaseWaiters();
1035	>	if ((workerCounts & RUNNING_COUNT_MASK) < pc) {
1036	>	ForkJoinWorkerThread[] ws = workers;
1037	>	int nws = ws.length;
1038	>	for (int i = 0; i < nws; ++i) {
1039	>	ForkJoinWorkerThread w = ws[i];
1040	>	if (w != null && w.isSuspended()) {
1041	>	if (joinMe.status < 0)
1042	>	return true;
1043	>	if ((workerCounts & RUNNING_COUNT_MASK) > pc)
1044	>	break;
1045	>	if (w.tryUnsuspend()) {
1046	>	LockSupport.unpark(w);
1047	>	break outer;
1048	>	}
1049	>	continue outer;
1050	>	}
1051	>	}
1052	>	}
1053	>	if (joinMe.status < 0)
1054		return true;
1055	+	int wc = workerCounts;
1056	+	if ((wc & RUNNING_COUNT_MASK) <= 2 ||
1057	+	(wc >>> TOTAL_COUNT_SHIFT) < pc)
1058	+	return false;  // keep this thread alive
1059	+	if (UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1060	+	wc, wc - ONE_RUNNING))
1061	+	break;
1062		}
1063	+
1064	+	joinMe.internalAwaitDone();
1065	+	int c;
1066	+	do {} while (!UNSAFE.compareAndSwapInt
1067	+	(this, workerCountsOffset,
1068	+	c = workerCounts, c + ONE_RUNNING));
1069	+	return true;
1070		}
1071
1072		/**
1073	<	* Controls whether to add spares to maintain parallelism
1073	>	* Possibly initiates and/or completes termination.
1074	>	*
1075	>	* @param now if true, unconditionally terminate, else only
1076	>	* if shutdown and empty queue and no active workers
1077	>	* @return true if now terminating or terminated
1078	>	*/
1079	>	private boolean tryTerminate(boolean now) {
1080	>	if (now)
1081	>	advanceRunLevel(SHUTDOWN); // ensure at least SHUTDOWN
1082	>	else if (runState < SHUTDOWN ||
1083	>	!submissionQueue.isEmpty() ||
1084	>	(runState & ACTIVE_COUNT_MASK) != 0)
1085	>	return false;
1086	>
1087	>	if (advanceRunLevel(TERMINATING))
1088	>	startTerminating();
1089	>
1090	>	// Finish now if all threads terminated; else in some subsequent call
1091	>	if ((workerCounts >>> TOTAL_COUNT_SHIFT) == 0) {
1092	>	advanceRunLevel(TERMINATED);
1093	>	terminationLatch.countDown();
1094	>	}
1095	>	return true;
1096	>	}
1097	>
1098	>	/**
1099	>	* Actions on transition to TERMINATING
1100	>	*/
1101	>	private void startTerminating() {
1102	>	for (int i = 0; i < 2; ++i) { // twice to mop up newly created workers
1103	>	cancelSubmissions();
1104	>	shutdownWorkers();
1105	>	cancelWorkerTasks();
1106	>	advanceEventCount();
1107	>	releaseWaiters();
1108	>	interruptWorkers();
1109	>	}
1110	>	}
1111	>
1112	>	/**
1113	>	* Clear out and cancel submissions, ignoring exceptions
1114		*/
1115	<	private volatile boolean maintainsParallelism;
1115	>	private void cancelSubmissions() {
1116	>	ForkJoinTask<?> task;
1117	>	while ((task = submissionQueue.poll()) != null) {
1118	>	try {
1119	>	task.cancel(false);
1120	>	} catch (Throwable ignore) {
1121	>	}
1122	>	}
1123	>	}
1124	>
1125	>	/**
1126	>	* Sets all worker run states to at least shutdown,
1127	>	* also resuming suspended workers
1128	>	*/
1129	>	private void shutdownWorkers() {
1130	>	ForkJoinWorkerThread[] ws = workers;
1131	>	int nws = ws.length;
1132	>	for (int i = 0; i < nws; ++i) {
1133	>	ForkJoinWorkerThread w = ws[i];
1134	>	if (w != null)
1135	>	w.shutdown();
1136	>	}
1137	>	}
1138	>
1139	>	/**
1140	>	* Clears out and cancels all locally queued tasks
1141	>	*/
1142	>	private void cancelWorkerTasks() {
1143	>	ForkJoinWorkerThread[] ws = workers;
1144	>	int nws = ws.length;
1145	>	for (int i = 0; i < nws; ++i) {
1146	>	ForkJoinWorkerThread w = ws[i];
1147	>	if (w != null)
1148	>	w.cancelTasks();
1149	>	}
1150	>	}
1151	>
1152	>	/**
1153	>	* Unsticks all workers blocked on joins etc
1154	>	*/
1155	>	private void interruptWorkers() {
1156	>	ForkJoinWorkerThread[] ws = workers;
1157	>	int nws = ws.length;
1158	>	for (int i = 0; i < nws; ++i) {
1159	>	ForkJoinWorkerThread w = ws[i];
1160	>	if (w != null && !w.isTerminated()) {
1161	>	try {
1162	>	w.interrupt();
1163	>	} catch (SecurityException ignore) {
1164	>	}
1165	>	}
1166	>	}
1167	>	}
1168	>
1169	>	// misc support for ForkJoinWorkerThread
1170	>
1171	>	/**
1172	>	* Returns pool number
1173	>	*/
1174	>	final int getPoolNumber() {
1175	>	return poolNumber;
1176	>	}
1177	>
1178	>	/**
1179	>	* Accumulates steal count from a worker, clearing
1180	>	* the worker's value
1181	>	*/
1182	>	final void accumulateStealCount(ForkJoinWorkerThread w) {
1183	>	int sc = w.stealCount;
1184	>	if (sc != 0) {
1185	>	long c;
1186	>	w.stealCount = 0;
1187	>	do {} while (!UNSAFE.compareAndSwapLong(this, stealCountOffset,
1188	>	c = stealCount, c + sc));
1189	>	}
1190	>	}
1191	>
1192	>	/**
1193	>	* Returns the approximate (non-atomic) number of idle threads per
1194	>	* active thread.
1195	>	*/
1196	>	final int idlePerActive() {
1197	>	int ac = runState;    // no mask -- artifically boosts during shutdown
1198	>	int pc = parallelism; // use targeted parallelism, not rc
1199	>	// Use exact results for small values, saturate past 4
1200	>	return pc <= ac? 0 : pc >>> 1 <= ac? 1 : pc >>> 2 <= ac? 3 : pc >>> 3;
1201	>	}
1202	>
1203	>	// Public and protected methods
1204
1205		// Constructors
1206
1207		/**
1208	<	* Creates a ForkJoinPool with a pool size equal to the number of
1209	<	* processors available on the system, using the default
1210	<	* ForkJoinWorkerThreadFactory.
1208	>	* Creates a {@code ForkJoinPool} with parallelism equal to {@link
1209	>	* java.lang.Runtime#availableProcessors}, and using the {@linkplain
1210	>	* #defaultForkJoinWorkerThreadFactory default thread factory}.
1211		*
1212		* @throws SecurityException if a security manager exists and
1213		*         the caller is not permitted to modify threads
#	Line 357 | Line 1220 | public class ForkJoinPool extends Abstra
1220		}
1221
1222		/**
1223	<	* Creates a ForkJoinPool with the indicated parallelism level
1224	<	* threads and using the default ForkJoinWorkerThreadFactory.
1223	>	* Creates a {@code ForkJoinPool} with the indicated parallelism
1224	>	* level and using the {@linkplain
1225	>	* #defaultForkJoinWorkerThreadFactory default thread factory}.
1226		*
1227	<	* @param parallelism the number of worker threads
1227	>	* @param parallelism the parallelism level
1228		* @throws IllegalArgumentException if parallelism less than or
1229	<	* equal to zero
1229	>	*         equal to zero, or greater than implementation limit
1230		* @throws SecurityException if a security manager exists and
1231		*         the caller is not permitted to modify threads
1232		*         because it does not hold {@link
#	Line 373 | Line 1237 | public class ForkJoinPool extends Abstra
1237		}
1238
1239		/**
1240	<	* Creates a ForkJoinPool with parallelism equal to the number of
1241	<	* processors available on the system and using the given
1242	<	* ForkJoinWorkerThreadFactory.
1240	>	* Creates a {@code ForkJoinPool} with parallelism equal to {@link
1241	>	* java.lang.Runtime#availableProcessors}, and using the given
1242	>	* thread factory.
1243		*
1244		* @param factory the factory for creating new threads
1245	<	* @throws NullPointerException if factory is null
1245	>	* @throws NullPointerException if the factory is null
1246		* @throws SecurityException if a security manager exists and
1247		*         the caller is not permitted to modify threads
1248		*         because it does not hold {@link
#	Line 389 | Line 1253 | public class ForkJoinPool extends Abstra
1253		}
1254
1255		/**
1256	<	* Creates a ForkJoinPool with the given parallelism and factory.
1256	>	* Creates a {@code ForkJoinPool} with the given parallelism and
1257	>	* thread factory.
1258		*
1259	<	* @param parallelism the targeted number of worker threads
1259	>	* @param parallelism the parallelism level
1260		* @param factory the factory for creating new threads
1261		* @throws IllegalArgumentException if parallelism less than or
1262	<	* equal to zero, or greater than implementation limit
1263	<	* @throws NullPointerException if factory is null
1262	>	*         equal to zero, or greater than implementation limit
1263	>	* @throws NullPointerException if the factory is null
1264		* @throws SecurityException if a security manager exists and
1265		*         the caller is not permitted to modify threads
1266		*         because it does not hold {@link
1267		*         java.lang.RuntimePermission}{@code ("modifyThread")}
1268		*/
1269		public ForkJoinPool(int parallelism, ForkJoinWorkerThreadFactory factory) {
1270	<	if (parallelism <= 0 || parallelism > MAX_THREADS)
406	<	throw new IllegalArgumentException();
1270	>	checkPermission();
1271		if (factory == null)
1272		throw new NullPointerException();
1273	<	checkPermission();
1274	<	this.factory = factory;
1273	>	if (parallelism <= 0 || parallelism > MAX_THREADS)
1274	>	throw new IllegalArgumentException();
1275	>	this.poolNumber = poolNumberGenerator.incrementAndGet();
1276	>	int arraySize = initialArraySizeFor(parallelism);
1277		this.parallelism = parallelism;
1278	+	this.factory = factory;
1279		this.maxPoolSize = MAX_THREADS;
1280		this.maintainsParallelism = true;
1281	<	this.poolNumber = poolNumberGenerator.incrementAndGet();
415	<	this.workerLock = new ReentrantLock();
416	<	this.termination = workerLock.newCondition();
417	<	this.stealCount = new AtomicLong();
1281	>	this.workers = new ForkJoinWorkerThread[arraySize];
1282		this.submissionQueue = new LinkedTransferQueue<ForkJoinTask<?>>();
1283	<	// worker array and workers are lazily constructed
1284	<	}
421	<
422	<	/**
423	<	* Creates a new worker thread using factory.
424	<	*
425	<	* @param index the index to assign worker
426	<	* @return new worker, or null of factory failed
427	<	*/
428	<	private ForkJoinWorkerThread createWorker(int index) {
429	<	Thread.UncaughtExceptionHandler h = ueh;
430	<	ForkJoinWorkerThread w = factory.newThread(this);
431	<	if (w != null) {
432	<	w.poolIndex = index;
433	<	w.setDaemon(true);
434	<	w.setAsyncMode(locallyFifo);
435	<	w.setName("ForkJoinPool-" + poolNumber + "-worker-" + index);
436	<	if (h != null)
437	<	w.setUncaughtExceptionHandler(h);
438	<	}
439	<	return w;
440	<	}
441	<
442	<	/**
443	<	* Returns a good size for worker array given pool size.
444	<	* Currently requires size to be a power of two.
445	<	*/
446	<	private static int arraySizeFor(int poolSize) {
447	<	return (poolSize <= 1) ? 1 :
448	<	(1 << (32 - Integer.numberOfLeadingZeros(poolSize-1)));
449	<	}
450	<
451	<	/**
452	<	* Creates or resizes array if necessary to hold newLength.
453	<	* Call only under exclusion.
454	<	*
455	<	* @return the array
456	<	*/
457	<	private ForkJoinWorkerThread[] ensureWorkerArrayCapacity(int newLength) {
458	<	ForkJoinWorkerThread[] ws = workers;
459	<	if (ws == null)
460	<	return workers = new ForkJoinWorkerThread[arraySizeFor(newLength)];
461	<	else if (newLength > ws.length)
462	<	return workers = Arrays.copyOf(ws, arraySizeFor(newLength));
463	<	else
464	<	return ws;
465	<	}
466	<
467	<	/**
468	<	* Tries to shrink workers into smaller array after one or more terminate.
469	<	*/
470	<	private void tryShrinkWorkerArray() {
471	<	ForkJoinWorkerThread[] ws = workers;
472	<	if (ws != null) {
473	<	int len = ws.length;
474	<	int last = len - 1;
475	<	while (last >= 0 && ws[last] == null)
476	<	--last;
477	<	int newLength = arraySizeFor(last+1);
478	<	if (newLength < len)
479	<	workers = Arrays.copyOf(ws, newLength);
480	<	}
481	<	}
482	<
483	<	/**
484	<	* Initializes workers if necessary.
485	<	*/
486	<	final void ensureWorkerInitialization() {
487	<	ForkJoinWorkerThread[] ws = workers;
488	<	if (ws == null) {
489	<	final ReentrantLock lock = this.workerLock;
490	<	lock.lock();
491	<	try {
492	<	ws = workers;
493	<	if (ws == null) {
494	<	int ps = parallelism;
495	<	ws = ensureWorkerArrayCapacity(ps);
496	<	for (int i = 0; i < ps; ++i) {
497	<	ForkJoinWorkerThread w = createWorker(i);
498	<	if (w != null) {
499	<	ws[i] = w;
500	<	w.start();
501	<	updateWorkerCount(1);
502	<	}
503	<	}
504	<	}
505	<	} finally {
506	<	lock.unlock();
507	<	}
508	<	}
1283	>	this.workerLock = new ReentrantLock();
1284	>	this.terminationLatch = new CountDownLatch(1);
1285		}
1286
1287		/**
1288	<	* Worker creation and startup for threads added via setParallelism.
1288	>	* Returns initial power of two size for workers array.
1289	>	* @param pc the initial parallelism level
1290		*/
1291	<	private void createAndStartAddedWorkers() {
1292	<	resumeAllSpares();  // Allow spares to convert to nonspare
1293	<	int ps = parallelism;
1294	<	ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(ps);
1295	<	int len = ws.length;
1296	<	// Sweep through slots, to keep lowest indices most populated
1297	<	int k = 0;
1298	<	while (k < len) {
522	<	if (ws[k] != null) {
523	<	++k;
524	<	continue;
525	<	}
526	<	int s = workerCounts;
527	<	int tc = totalCountOf(s);
528	<	int rc = runningCountOf(s);
529	<	if (rc >= ps || tc >= ps)
530	<	break;
531	<	if (casWorkerCounts (s, workerCountsFor(tc+1, rc+1))) {
532	<	ForkJoinWorkerThread w = createWorker(k);
533	<	if (w != null) {
534	<	ws[k++] = w;
535	<	w.start();
536	<	}
537	<	else {
538	<	updateWorkerCount(-1); // back out on failed creation
539	<	break;
540	<	}
541	<	}
542	<	}
1291	>	private static int initialArraySizeFor(int pc) {
1292	>	// See Hackers Delight, sec 3.2. We know MAX_THREADS < (1 >>> 16)
1293	>	int size = pc < MAX_THREADS ? pc + 1 : MAX_THREADS;
1294	>	size |= size >>> 1;
1295	>	size |= size >>> 2;
1296	>	size |= size >>> 4;
1297	>	size |= size >>> 8;
1298	>	return size + 1;
1299		}
1300
1301		// Execution methods
#	Line 550 | Line 1306 | public class ForkJoinPool extends Abstra
1306		private <T> void doSubmit(ForkJoinTask<T> task) {
1307		if (task == null)
1308		throw new NullPointerException();
1309	<	if (isShutdown())
1309	>	if (runState >= SHUTDOWN)
1310		throw new RejectedExecutionException();
555	–	if (workers == null)
556	–	ensureWorkerInitialization();
1311		submissionQueue.offer(task);
1312	<	signalIdleWorkers();
1312	>	advanceEventCount();
1313	>	releaseWaiters();
1314	>	ensureEnoughTotalWorkers();
1315		}
1316
1317		/**
#	Line 563 | Line 1319 | public class ForkJoinPool extends Abstra
1319		*
1320		* @param task the task
1321		* @return the task's result
1322	<	* @throws NullPointerException if task is null
1323	<	* @throws RejectedExecutionException if pool is shut down
1322	>	* @throws NullPointerException if the task is null
1323	>	* @throws RejectedExecutionException if the task cannot be
1324	>	*         scheduled for execution
1325		*/
1326		public <T> T invoke(ForkJoinTask<T> task) {
1327		doSubmit(task);
#	Line 575 | Line 1332 | public class ForkJoinPool extends Abstra
1332		* Arranges for (asynchronous) execution of the given task.
1333		*
1334		* @param task the task
1335	<	* @throws NullPointerException if task is null
1336	<	* @throws RejectedExecutionException if pool is shut down
1335	>	* @throws NullPointerException if the task is null
1336	>	* @throws RejectedExecutionException if the task cannot be
1337	>	*         scheduled for execution
1338		*/
1339	<	public <T> void execute(ForkJoinTask<T> task) {
1339	>	public void execute(ForkJoinTask<?> task) {
1340		doSubmit(task);
1341		}
1342
1343		// AbstractExecutorService methods
1344
1345	+	/**
1346	+	* @throws NullPointerException if the task is null
1347	+	* @throws RejectedExecutionException if the task cannot be
1348	+	*         scheduled for execution
1349	+	*/
1350		public void execute(Runnable task) {
1351		ForkJoinTask<?> job;
1352	<	if (task instanceof AdaptedCallable) // avoid re-wrap
1353	<	job = (AdaptedCallable<?>)task;
591	<	else if (task instanceof AdaptedRunnable)
592	<	job = (AdaptedRunnable<?>)task;
1352	>	if (task instanceof ForkJoinTask<?>) // avoid re-wrap
1353	>	job = (ForkJoinTask<?>) task;
1354		else
1355	<	job = new AdaptedRunnable<Void>(task, null);
1355	>	job = ForkJoinTask.adapt(task, null);
1356		doSubmit(job);
1357		}
1358
1359	+	/**
1360	+	* @throws NullPointerException if the task is null
1361	+	* @throws RejectedExecutionException if the task cannot be
1362	+	*         scheduled for execution
1363	+	*/
1364		public <T> ForkJoinTask<T> submit(Callable<T> task) {
1365	<	ForkJoinTask<T> job = new AdaptedCallable<T>(task);
1365	>	ForkJoinTask<T> job = ForkJoinTask.adapt(task);
1366		doSubmit(job);
1367		return job;
1368		}
1369
1370	+	/**
1371	+	* @throws NullPointerException if the task is null
1372	+	* @throws RejectedExecutionException if the task cannot be
1373	+	*         scheduled for execution
1374	+	*/
1375		public <T> ForkJoinTask<T> submit(Runnable task, T result) {
1376	<	ForkJoinTask<T> job = new AdaptedRunnable<T>(task, result);
1376	>	ForkJoinTask<T> job = ForkJoinTask.adapt(task, result);
1377		doSubmit(job);
1378		return job;
1379		}
1380
1381	+	/**
1382	+	* @throws NullPointerException if the task is null
1383	+	* @throws RejectedExecutionException if the task cannot be
1384	+	*         scheduled for execution
1385	+	*/
1386		public ForkJoinTask<?> submit(Runnable task) {
1387		ForkJoinTask<?> job;
1388	<	if (task instanceof AdaptedCallable) // avoid re-wrap
1389	<	job = (AdaptedCallable<?>)task;
614	<	else if (task instanceof AdaptedRunnable)
615	<	job = (AdaptedRunnable<?>)task;
1388	>	if (task instanceof ForkJoinTask<?>) // avoid re-wrap
1389	>	job = (ForkJoinTask<?>) task;
1390		else
1391	<	job = new AdaptedRunnable<Void>(task, null);
1391	>	job = ForkJoinTask.adapt(task, null);
1392		doSubmit(job);
1393		return job;
1394		}
#	Line 624 | Line 1398 | public class ForkJoinPool extends Abstra
1398		*
1399		* @param task the task to submit
1400		* @return the task
1401	+	* @throws NullPointerException if the task is null
1402		* @throws RejectedExecutionException if the task cannot be
1403		*         scheduled for execution
629	–	* @throws NullPointerException if the task is null
1404		*/
1405		public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) {
1406		doSubmit(task);
#	Line 634 | Line 1408 | public class ForkJoinPool extends Abstra
1408		}
1409
1410		/**
1411	<	* Adaptor for Runnables. This implements RunnableFuture
1412	<	* to be compliant with AbstractExecutorService constraints.
1411	>	* @throws NullPointerException       {@inheritDoc}
1412	>	* @throws RejectedExecutionException {@inheritDoc}
1413		*/
640	–	static final class AdaptedRunnable<T> extends ForkJoinTask<T>
641	–	implements RunnableFuture<T> {
642	–	final Runnable runnable;
643	–	final T resultOnCompletion;
644	–	T result;
645	–	AdaptedRunnable(Runnable runnable, T result) {
646	–	if (runnable == null) throw new NullPointerException();
647	–	this.runnable = runnable;
648	–	this.resultOnCompletion = result;
649	–	}
650	–	public T getRawResult() { return result; }
651	–	public void setRawResult(T v) { result = v; }
652	–	public boolean exec() {
653	–	runnable.run();
654	–	result = resultOnCompletion;
655	–	return true;
656	–	}
657	–	public void run() { invoke(); }
658	–	private static final long serialVersionUID = 5232453952276885070L;
659	–	}
660	–
661	–	/**
662	–	* Adaptor for Callables
663	–	*/
664	–	static final class AdaptedCallable<T> extends ForkJoinTask<T>
665	–	implements RunnableFuture<T> {
666	–	final Callable<T> callable;
667	–	T result;
668	–	AdaptedCallable(Callable<T> callable) {
669	–	if (callable == null) throw new NullPointerException();
670	–	this.callable = callable;
671	–	}
672	–	public T getRawResult() { return result; }
673	–	public void setRawResult(T v) { result = v; }
674	–	public boolean exec() {
675	–	try {
676	–	result = callable.call();
677	–	return true;
678	–	} catch (Error err) {
679	–	throw err;
680	–	} catch (RuntimeException rex) {
681	–	throw rex;
682	–	} catch (Exception ex) {
683	–	throw new RuntimeException(ex);
684	–	}
685	–	}
686	–	public void run() { invoke(); }
687	–	private static final long serialVersionUID = 2838392045355241008L;
688	–	}
689	–
1414		public <T> List<Future<T>> invokeAll(Collection<? extends Callable<T>> tasks) {
1415		ArrayList<ForkJoinTask<T>> forkJoinTasks =
1416		new ArrayList<ForkJoinTask<T>>(tasks.size());
1417		for (Callable<T> task : tasks)
1418	<	forkJoinTasks.add(new AdaptedCallable<T>(task));
1418	>	forkJoinTasks.add(ForkJoinTask.adapt(task));
1419		invoke(new InvokeAll<T>(forkJoinTasks));
1420
1421		@SuppressWarnings({"unchecked", "rawtypes"})
1422	<	List<Future<T>> futures = (List<Future<T>>) (List) forkJoinTasks;
1422	>	List<Future<T>> futures = (List<Future<T>>) (List) forkJoinTasks;
1423		return futures;
1424		}
1425
#	Line 709 | Line 1433 | public class ForkJoinPool extends Abstra
1433		private static final long serialVersionUID = -7914297376763021607L;
1434		}
1435
712	–	// Configuration and status settings and queries
713	–
1436		/**
1437		* Returns the factory used for constructing new workers.
1438		*
#	Line 724 | Line 1446 | public class ForkJoinPool extends Abstra
1446		* Returns the handler for internal worker threads that terminate
1447		* due to unrecoverable errors encountered while executing tasks.
1448		*
1449	<	* @return the handler, or null if none
1449	>	* @return the handler, or {@code null} if none
1450		*/
1451		public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() {
1452	<	Thread.UncaughtExceptionHandler h;
1453	<	final ReentrantLock lock = this.workerLock;
732	<	lock.lock();
733	<	try {
734	<	h = ueh;
735	<	} finally {
736	<	lock.unlock();
737	<	}
738	<	return h;
1452	>	workerCountReadFence();
1453	>	return ueh;
1454		}
1455
1456		/**
#	Line 745 | Line 1460 | public class ForkJoinPool extends Abstra
1460		* as handler.
1461		*
1462		* @param h the new handler
1463	<	* @return the old handler, or null if none
1463	>	* @return the old handler, or {@code null} if none
1464		* @throws SecurityException if a security manager exists and
1465		*         the caller is not permitted to modify threads
1466		*         because it does not hold {@link
#	Line 754 | Line 1469 | public class ForkJoinPool extends Abstra
1469		public Thread.UncaughtExceptionHandler
1470		setUncaughtExceptionHandler(Thread.UncaughtExceptionHandler h) {
1471		checkPermission();
1472	<	Thread.UncaughtExceptionHandler old = null;
1473	<	final ReentrantLock lock = this.workerLock;
759	<	lock.lock();
760	<	try {
761	<	old = ueh;
1472	>	Thread.UncaughtExceptionHandler old = ueh;
1473	>	if (h != old) {
1474		ueh = h;
1475		ForkJoinWorkerThread[] ws = workers;
1476	<	if (ws != null) {
1477	<	for (int i = 0; i < ws.length; ++i) {
1478	<	ForkJoinWorkerThread w = ws[i];
1479	<	if (w != null)
1480	<	w.setUncaughtExceptionHandler(h);
769	<	}
1476	>	int nws = ws.length;
1477	>	for (int i = 0; i < nws; ++i) {
1478	>	ForkJoinWorkerThread w = ws[i];
1479	>	if (w != null)
1480	>	w.setUncaughtExceptionHandler(h);
1481		}
771	–	} finally {
772	–	lock.unlock();
1482		}
1483		return old;
1484		}
1485
777	–
1486		/**
1487		* Sets the target parallelism level of this pool.
1488		*
#	Line 790 | Line 1498 | public class ForkJoinPool extends Abstra
1498		checkPermission();
1499		if (parallelism <= 0 || parallelism > maxPoolSize)
1500		throw new IllegalArgumentException();
1501	<	final ReentrantLock lock = this.workerLock;
1502	<	lock.lock();
1503	<	try {
1504	<	if (!isTerminating()) {
1505	<	int p = this.parallelism;
1506	<	this.parallelism = parallelism;
1507	<	if (parallelism > p)
1508	<	createAndStartAddedWorkers();
1509	<	else
1510	<	trimSpares();
1501	>	workerCountReadFence();
1502	>	int pc = this.parallelism;
1503	>	if (pc != parallelism) {
1504	>	this.parallelism = parallelism;
1505	>	workerCountWriteFence();
1506	>	// Release spares. If too many, some will die after re-suspend
1507	>	ForkJoinWorkerThread[] ws = workers;
1508	>	int nws = ws.length;
1509	>	for (int i = 0; i < nws; ++i) {
1510	>	ForkJoinWorkerThread w = ws[i];
1511	>	if (w != null && w.tryUnsuspend()) {
1512	>	int c;
1513	>	do {} while (!UNSAFE.compareAndSwapInt
1514	>	(this, workerCountsOffset,
1515	>	c = workerCounts, c + ONE_RUNNING));
1516	>	LockSupport.unpark(w);
1517	>	}
1518		}
1519	<	} finally {
1520	<	lock.unlock();
1519	>	ensureEnoughTotalWorkers();
1520	>	advanceEventCount();
1521	>	releaseWaiters(); // force config recheck by existing workers
1522		}
807	–	signalIdleWorkers();
1523		}
1524
1525		/**
1526	<	* Returns the targeted number of worker threads in this pool.
1526	>	* Returns the targeted parallelism level of this pool.
1527		*
1528	<	* @return the targeted number of worker threads in this pool
1528	>	* @return the targeted parallelism level of this pool
1529		*/
1530		public int getParallelism() {
1531	+	//        workerCountReadFence(); // inlined below
1532	+	int ignore = workerCounts;
1533		return parallelism;
1534		}
1535
1536		/**
1537		* Returns the number of worker threads that have started but not
1538		* yet terminated.  This result returned by this method may differ
1539	<	* from {@code getParallelism} when threads are created to
1539	>	* from {@link #getParallelism} when threads are created to
1540		* maintain parallelism when others are cooperatively blocked.
1541		*
1542		* @return the number of worker threads
1543		*/
1544		public int getPoolSize() {
1545	<	return totalCountOf(workerCounts);
1545	>	return workerCounts >>> TOTAL_COUNT_SHIFT;
1546		}
1547
1548		/**
1549		* Returns the maximum number of threads allowed to exist in the
1550	<	* pool, even if there are insufficient unblocked running threads.
1550	>	* pool. Unless set using {@link #setMaximumPoolSize}, the
1551	>	* maximum is an implementation-defined value designed only to
1552	>	* prevent runaway growth.
1553		*
1554		* @return the maximum
1555		*/
1556		public int getMaximumPoolSize() {
1557	+	workerCountReadFence();
1558		return maxPoolSize;
1559		}
1560
1561		/**
1562		* Sets the maximum number of threads allowed to exist in the
1563	<	* pool, even if there are insufficient unblocked running threads.
1564	<	* Setting this value has no effect on current pool size. It
1565	<	* controls construction of new threads.
1563	>	* pool. The given value should normally be greater than or equal
1564	>	* to the {@link #getParallelism parallelism} level. Setting this
1565	>	* value has no effect on current pool size. It controls
1566	>	* construction of new threads. The use of this method may cause
1567	>	* tasks that intrinsically require extra threads for dependent
1568	>	* computations to indefinitely stall. If you are instead trying
1569	>	* to minimize internal thread creation, consider setting {@link
1570	>	* #setMaintainsParallelism} as false.
1571		*
1572	<	* @throws IllegalArgumentException if negative or greater then
1572	>	* @throws IllegalArgumentException if negative or greater than
1573		* internal implementation limit
1574		*/
1575		public void setMaximumPoolSize(int newMax) {
1576		if (newMax < 0 || newMax > MAX_THREADS)
1577		throw new IllegalArgumentException();
1578		maxPoolSize = newMax;
1579	+	workerCountWriteFence();
1580		}
1581
856	–
1582		/**
1583	<	* Returns true if this pool dynamically maintains its target
1584	<	* parallelism level. If false, new threads are added only to
1585	<	* avoid possible starvation.
861	<	* This setting is by default true.
1583	>	* Returns {@code true} if this pool dynamically maintains its
1584	>	* target parallelism level. If false, new threads are added only
1585	>	* to avoid possible starvation.  This setting is by default true.
1586		*
1587	<	* @return true if maintains parallelism
1587	>	* @return {@code true} if maintains parallelism
1588		*/
1589		public boolean getMaintainsParallelism() {
1590	+	workerCountReadFence();
1591		return maintainsParallelism;
1592		}
1593
#	Line 871 | Line 1596 | public class ForkJoinPool extends Abstra
1596		* parallelism level. If false, new threads are added only to
1597		* avoid possible starvation.
1598		*
1599	<	* @param enable true to maintains parallelism
1599	>	* @param enable {@code true} to maintain parallelism
1600		*/
1601		public void setMaintainsParallelism(boolean enable) {
1602		maintainsParallelism = enable;
1603	+	workerCountWriteFence();
1604		}
1605
1606		/**
#	Line 882 | Line 1608 | public class ForkJoinPool extends Abstra
1608		* tasks that are never joined. This mode may be more appropriate
1609		* than default locally stack-based mode in applications in which
1610		* worker threads only process asynchronous tasks.  This method is
1611	<	* designed to be invoked only when pool is quiescent, and
1611	>	* designed to be invoked only when the pool is quiescent, and
1612		* typically only before any tasks are submitted. The effects of
1613		* invocations at other times may be unpredictable.
1614		*
1615	<	* @param async if true, use locally FIFO scheduling
1615	>	* @param async if {@code true}, use locally FIFO scheduling
1616		* @return the previous mode
1617	+	* @see #getAsyncMode
1618		*/
1619		public boolean setAsyncMode(boolean async) {
1620	+	workerCountReadFence();
1621		boolean oldMode = locallyFifo;
1622	<	locallyFifo = async;
1623	<	ForkJoinWorkerThread[] ws = workers;
1624	<	if (ws != null) {
1625	<	for (int i = 0; i < ws.length; ++i) {
1626	<	ForkJoinWorkerThread t = ws[i];
1627	<	if (t != null)
1628	<	t.setAsyncMode(async);
1622	>	if (oldMode != async) {
1623	>	locallyFifo = async;
1624	>	workerCountWriteFence();
1625	>	ForkJoinWorkerThread[] ws = workers;
1626	>	int nws = ws.length;
1627	>	for (int i = 0; i < nws; ++i) {
1628	>	ForkJoinWorkerThread w = ws[i];
1629	>	if (w != null)
1630	>	w.setAsyncMode(async);
1631		}
1632		}
1633		return oldMode;
1634		}
1635
1636		/**
1637	<	* Returns true if this pool uses local first-in-first-out
1637	>	* Returns {@code true} if this pool uses local first-in-first-out
1638		* scheduling mode for forked tasks that are never joined.
1639		*
1640	<	* @return true if this pool uses async mode
1640	>	* @return {@code true} if this pool uses async mode
1641	>	* @see #setAsyncMode
1642		*/
1643		public boolean getAsyncMode() {
1644	+	workerCountReadFence();
1645		return locallyFifo;
1646		}
1647
1648		/**
1649		* Returns an estimate of the number of worker threads that are
1650		* not blocked waiting to join tasks or for other managed
1651	<	* synchronization.
1651	>	* synchronization. This method may overestimate the
1652	>	* number of running threads.
1653		*
1654		* @return the number of worker threads
1655		*/
1656		public int getRunningThreadCount() {
1657	<	return runningCountOf(workerCounts);
1657	>	return workerCounts & RUNNING_COUNT_MASK;
1658		}
1659
1660		/**
#	Line 932 | Line 1665 | public class ForkJoinPool extends Abstra
1665		* @return the number of active threads
1666		*/
1667		public int getActiveThreadCount() {
1668	<	return activeCountOf(runControl);
1668	>	return runState & ACTIVE_COUNT_MASK;
1669		}
1670
1671		/**
1672	<	* Returns an estimate of the number of threads that are currently
1673	<	* idle waiting for tasks. This method may underestimate the
1674	<	* number of idle threads.
1675	<	*
1676	<	* @return the number of idle threads
1677	<	*/
1678	<	final int getIdleThreadCount() {
946	<	int c = runningCountOf(workerCounts) - activeCountOf(runControl);
947	<	return (c <= 0) ? 0 : c;
948	<	}
949	<
950	<	/**
951	<	* Returns true if all worker threads are currently idle. An idle
952	<	* worker is one that cannot obtain a task to execute because none
953	<	* are available to steal from other threads, and there are no
954	<	* pending submissions to the pool. This method is conservative;
955	<	* it might not return true immediately upon idleness of all
956	<	* threads, but will eventually become true if threads remain
957	<	* inactive.
1672	>	* Returns {@code true} if all worker threads are currently idle.
1673	>	* An idle worker is one that cannot obtain a task to execute
1674	>	* because none are available to steal from other threads, and
1675	>	* there are no pending submissions to the pool. This method is
1676	>	* conservative; it might not return {@code true} immediately upon
1677	>	* idleness of all threads, but will eventually become true if
1678	>	* threads remain inactive.
1679		*
1680	<	* @return true if all threads are currently idle
1680	>	* @return {@code true} if all threads are currently idle
1681		*/
1682		public boolean isQuiescent() {
1683	<	return activeCountOf(runControl) == 0;
1683	>	return (runState & ACTIVE_COUNT_MASK) == 0;
1684		}
1685
1686		/**
#	Line 974 | Line 1695 | public class ForkJoinPool extends Abstra
1695		* @return the number of steals
1696		*/
1697		public long getStealCount() {
1698	<	return stealCount.get();
978	<	}
979	<
980	<	/**
981	<	* Accumulates steal count from a worker.
982	<	* Call only when worker known to be idle.
983	<	*/
984	<	private void updateStealCount(ForkJoinWorkerThread w) {
985	<	int sc = w.getAndClearStealCount();
986	<	if (sc != 0)
987	<	stealCount.addAndGet(sc);
1698	>	return stealCount;
1699		}
1700
1701		/**
#	Line 1000 | Line 1711 | public class ForkJoinPool extends Abstra
1711		public long getQueuedTaskCount() {
1712		long count = 0;
1713		ForkJoinWorkerThread[] ws = workers;
1714	<	if (ws != null) {
1715	<	for (int i = 0; i < ws.length; ++i) {
1716	<	ForkJoinWorkerThread t = ws[i];
1717	<	if (t != null)
1718	<	count += t.getQueueSize();
1008	<	}
1714	>	int nws = ws.length;
1715	>	for (int i = 0; i < nws; ++i) {
1716	>	ForkJoinWorkerThread w = ws[i];
1717	>	if (w != null)
1718	>	count += w.getQueueSize();
1719		}
1720		return count;
1721		}
1722
1723		/**
1724	<	* Returns an estimate of the number tasks submitted to this pool
1725	<	* that have not yet begun executing. This method takes time
1724	>	* Returns an estimate of the number of tasks submitted to this
1725	>	* pool that have not yet begun executing.  This method takes time
1726		* proportional to the number of submissions.
1727		*
1728		* @return the number of queued submissions
#	Line 1022 | Line 1732 | public class ForkJoinPool extends Abstra
1732		}
1733
1734		/**
1735	<	* Returns true if there are any tasks submitted to this pool
1736	<	* that have not yet begun executing.
1735	>	* Returns {@code true} if there are any tasks submitted to this
1736	>	* pool that have not yet begun executing.
1737		*
1738		* @return {@code true} if there are any queued submissions
1739		*/
#	Line 1036 | Line 1746 | public class ForkJoinPool extends Abstra
1746		* available.  This method may be useful in extensions to this
1747		* class that re-assign work in systems with multiple pools.
1748		*
1749	<	* @return the next submission, or null if none
1749	>	* @return the next submission, or {@code null} if none
1750		*/
1751		protected ForkJoinTask<?> pollSubmission() {
1752		return submissionQueue.poll();
#	Line 1046 | Line 1756 | public class ForkJoinPool extends Abstra
1756		* Removes all available unexecuted submitted and forked tasks
1757		* from scheduling queues and adds them to the given collection,
1758		* without altering their execution status. These may include
1759	<	* artificially generated or wrapped tasks. This method is designed
1760	<	* to be invoked only when the pool is known to be
1759	>	* artificially generated or wrapped tasks. This method is
1760	>	* designed to be invoked only when the pool is known to be
1761		* quiescent. Invocations at other times may not remove all
1762		* tasks. A failure encountered while attempting to add elements
1763		* to collection {@code c} may result in elements being in
#	Line 1059 | Line 1769 | public class ForkJoinPool extends Abstra
1769		* @param c the collection to transfer elements into
1770		* @return the number of elements transferred
1771		*/
1772	<	protected int drainTasksTo(Collection<ForkJoinTask<?>> c) {
1772	>	protected int drainTasksTo(Collection<? super ForkJoinTask<?>> c) {
1773		int n = submissionQueue.drainTo(c);
1774		ForkJoinWorkerThread[] ws = workers;
1775	<	if (ws != null) {
1776	<	for (int i = 0; i < ws.length; ++i) {
1777	<	ForkJoinWorkerThread w = ws[i];
1778	<	if (w != null)
1779	<	n += w.drainTasksTo(c);
1070	<	}
1775	>	int nws = ws.length;
1776	>	for (int i = 0; i < nws; ++i) {
1777	>	ForkJoinWorkerThread w = ws[i];
1778	>	if (w != null)
1779	>	n += w.drainTasksTo(c);
1780		}
1781		return n;
1782		}
#	Line 1080 | Line 1789 | public class ForkJoinPool extends Abstra
1789		* @return a string identifying this pool, as well as its state
1790		*/
1791		public String toString() {
1083	–	int ps = parallelism;
1084	–	int wc = workerCounts;
1085	–	int rc = runControl;
1792		long st = getStealCount();
1793		long qt = getQueuedTaskCount();
1794		long qs = getQueuedSubmissionCount();
1795	+	int wc = workerCounts;
1796	+	int tc = wc >>> TOTAL_COUNT_SHIFT;
1797	+	int rc = wc & RUNNING_COUNT_MASK;
1798	+	int pc = parallelism;
1799	+	int rs = runState;
1800	+	int ac = rs & ACTIVE_COUNT_MASK;
1801		return super.toString() +
1802	<	"[" + runStateToString(runStateOf(rc)) +
1803	<	", parallelism = " + ps +
1804	<	", size = " + totalCountOf(wc) +
1805	<	", active = " + activeCountOf(rc) +
1806	<	", running = " + runningCountOf(wc) +
1802	>	"[" + runLevelToString(rs) +
1803	>	", parallelism = " + pc +
1804	>	", size = " + tc +
1805	>	", active = " + ac +
1806	>	", running = " + rc +
1807		", steals = " + st +
1808		", tasks = " + qt +
1809		", submissions = " + qs +
1810		"]";
1811		}
1812
1813	<	private static String runStateToString(int rs) {
1814	<	switch(rs) {
1815	<	case RUNNING: return "Running";
1816	<	case SHUTDOWN: return "Shutting down";
1817	<	case TERMINATING: return "Terminating";
1106	<	case TERMINATED: return "Terminated";
1107	<	default: throw new Error("Unknown run state");
1108	<	}
1813	>	private static String runLevelToString(int s) {
1814	>	return ((s & TERMINATED) != 0 ? "Terminated" :
1815	>	((s & TERMINATING) != 0 ? "Terminating" :
1816	>	((s & SHUTDOWN) != 0 ? "Shutting down" :
1817	>	"Running")));
1818		}
1819
1111	–	// lifecycle control
1112	–
1820		/**
1821		* Initiates an orderly shutdown in which previously submitted
1822		* tasks are executed, but no new tasks will be accepted.
#	Line 1124 | Line 1831 | public class ForkJoinPool extends Abstra
1831		*/
1832		public void shutdown() {
1833		checkPermission();
1834	<	transitionRunStateTo(SHUTDOWN);
1835	<	if (canTerminateOnShutdown(runControl))
1129	<	terminateOnShutdown();
1834	>	advanceRunLevel(SHUTDOWN);
1835	>	tryTerminate(false);
1836		}
1837
1838		/**
1839	<	* Attempts to stop all actively executing tasks, and cancels all
1840	<	* waiting tasks.  Tasks that are in the process of being
1841	<	* submitted or executed concurrently during the course of this
1842	<	* method may or may not be rejected. Unlike some other executors,
1843	<	* this method cancels rather than collects non-executed tasks
1844	<	* upon termination, so always returns an empty list. However, you
1845	<	* can use method {@code drainTasksTo} before invoking this
1846	<	* method to transfer unexecuted tasks to another collection.
1839	>	* Attempts to cancel and/or stop all tasks, and reject all
1840	>	* subsequently submitted tasks.  Tasks that are in the process of
1841	>	* being submitted or executed concurrently during the course of
1842	>	* this method may or may not be rejected. This method cancels
1843	>	* both existing and unexecuted tasks, in order to permit
1844	>	* termination in the presence of task dependencies. So the method
1845	>	* always returns an empty list (unlike the case for some other
1846	>	* Executors).
1847		*
1848		* @return an empty list
1849		* @throws SecurityException if a security manager exists and
#	Line 1147 | Line 1853 | public class ForkJoinPool extends Abstra
1853		*/
1854		public List<Runnable> shutdownNow() {
1855		checkPermission();
1856	<	terminate();
1856	>	tryTerminate(true);
1857		return Collections.emptyList();
1858		}
1859
#	Line 1157 | Line 1863 | public class ForkJoinPool extends Abstra
1863		* @return {@code true} if all tasks have completed following shut down
1864		*/
1865		public boolean isTerminated() {
1866	<	return runStateOf(runControl) == TERMINATED;
1866	>	return runState >= TERMINATED;
1867		}
1868
1869		/**
1870		* Returns {@code true} if the process of termination has
1871	<	* commenced but possibly not yet completed.
1871	>	* commenced but not yet completed.  This method may be useful for
1872	>	* debugging. A return of {@code true} reported a sufficient
1873	>	* period after shutdown may indicate that submitted tasks have
1874	>	* ignored or suppressed interruption, causing this executor not
1875	>	* to properly terminate.
1876		*
1877	<	* @return {@code true} if terminating
1877	>	* @return {@code true} if terminating but not yet terminated
1878		*/
1879		public boolean isTerminating() {
1880	<	return runStateOf(runControl) >= TERMINATING;
1880	>	return (runState & (TERMINATING|TERMINATED)) == TERMINATING;
1881		}
1882
1883		/**
#	Line 1176 | Line 1886 | public class ForkJoinPool extends Abstra
1886		* @return {@code true} if this pool has been shut down
1887		*/
1888		public boolean isShutdown() {
1889	<	return runStateOf(runControl) >= SHUTDOWN;
1889	>	return runState >= SHUTDOWN;
1890		}
1891
1892		/**
#	Line 1192 | Line 1902 | public class ForkJoinPool extends Abstra
1902		*/
1903		public boolean awaitTermination(long timeout, TimeUnit unit)
1904		throws InterruptedException {
1905	<	long nanos = unit.toNanos(timeout);
1196	<	final ReentrantLock lock = this.workerLock;
1197	<	lock.lock();
1198	<	try {
1199	<	for (;;) {
1200	<	if (isTerminated())
1201	<	return true;
1202	<	if (nanos <= 0)
1203	<	return false;
1204	<	nanos = termination.awaitNanos(nanos);
1205	<	}
1206	<	} finally {
1207	<	lock.unlock();
1208	<	}
1209	<	}
1210	<
1211	<	// Shutdown and termination support
1212	<
1213	<	/**
1214	<	* Callback from terminating worker. Nulls out the corresponding
1215	<	* workers slot, and if terminating, tries to terminate; else
1216	<	* tries to shrink workers array.
1217	<	*
1218	<	* @param w the worker
1219	<	*/
1220	<	final void workerTerminated(ForkJoinWorkerThread w) {
1221	<	updateStealCount(w);
1222	<	updateWorkerCount(-1);
1223	<	final ReentrantLock lock = this.workerLock;
1224	<	lock.lock();
1225	<	try {
1226	<	ForkJoinWorkerThread[] ws = workers;
1227	<	if (ws != null) {
1228	<	int idx = w.poolIndex;
1229	<	if (idx >= 0 && idx < ws.length && ws[idx] == w)
1230	<	ws[idx] = null;
1231	<	if (totalCountOf(workerCounts) == 0) {
1232	<	terminate(); // no-op if already terminating
1233	<	transitionRunStateTo(TERMINATED);
1234	<	termination.signalAll();
1235	<	}
1236	<	else if (!isTerminating()) {
1237	<	tryShrinkWorkerArray();
1238	<	tryResumeSpare(true); // allow replacement
1239	<	}
1240	<	}
1241	<	} finally {
1242	<	lock.unlock();
1243	<	}
1244	<	signalIdleWorkers();
1245	<	}
1246	<
1247	<	/**
1248	<	* Initiates termination.
1249	<	*/
1250	<	private void terminate() {
1251	<	if (transitionRunStateTo(TERMINATING)) {
1252	<	stopAllWorkers();
1253	<	resumeAllSpares();
1254	<	signalIdleWorkers();
1255	<	cancelQueuedSubmissions();
1256	<	cancelQueuedWorkerTasks();
1257	<	interruptUnterminatedWorkers();
1258	<	signalIdleWorkers(); // resignal after interrupt
1259	<	}
1260	<	}
1261	<
1262	<	/**
1263	<	* Possibly terminates when on shutdown state.
1264	<	*/
1265	<	private void terminateOnShutdown() {
1266	<	if (!hasQueuedSubmissions() && canTerminateOnShutdown(runControl))
1267	<	terminate();
1268	<	}
1269	<
1270	<	/**
1271	<	* Clears out and cancels submissions.
1272	<	*/
1273	<	private void cancelQueuedSubmissions() {
1274	<	ForkJoinTask<?> task;
1275	<	while ((task = pollSubmission()) != null)
1276	<	task.cancel(false);
1277	<	}
1278	<
1279	<	/**
1280	<	* Cleans out worker queues.
1281	<	*/
1282	<	private void cancelQueuedWorkerTasks() {
1283	<	final ReentrantLock lock = this.workerLock;
1284	<	lock.lock();
1285	<	try {
1286	<	ForkJoinWorkerThread[] ws = workers;
1287	<	if (ws != null) {
1288	<	for (int i = 0; i < ws.length; ++i) {
1289	<	ForkJoinWorkerThread t = ws[i];
1290	<	if (t != null)
1291	<	t.cancelTasks();
1292	<	}
1293	<	}
1294	<	} finally {
1295	<	lock.unlock();
1296	<	}
1297	<	}
1298	<
1299	<	/**
1300	<	* Sets each worker's status to terminating. Requires lock to avoid
1301	<	* conflicts with add/remove.
1302	<	*/
1303	<	private void stopAllWorkers() {
1304	<	final ReentrantLock lock = this.workerLock;
1305	<	lock.lock();
1306	<	try {
1307	<	ForkJoinWorkerThread[] ws = workers;
1308	<	if (ws != null) {
1309	<	for (int i = 0; i < ws.length; ++i) {
1310	<	ForkJoinWorkerThread t = ws[i];
1311	<	if (t != null)
1312	<	t.shutdownNow();
1313	<	}
1314	<	}
1315	<	} finally {
1316	<	lock.unlock();
1317	<	}
1318	<	}
1319	<
1320	<	/**
1321	<	* Interrupts all unterminated workers.  This is not required for
1322	<	* sake of internal control, but may help unstick user code during
1323	<	* shutdown.
1324	<	*/
1325	<	private void interruptUnterminatedWorkers() {
1326	<	final ReentrantLock lock = this.workerLock;
1327	<	lock.lock();
1328	<	try {
1329	<	ForkJoinWorkerThread[] ws = workers;
1330	<	if (ws != null) {
1331	<	for (int i = 0; i < ws.length; ++i) {
1332	<	ForkJoinWorkerThread t = ws[i];
1333	<	if (t != null && !t.isTerminated()) {
1334	<	try {
1335	<	t.interrupt();
1336	<	} catch (SecurityException ignore) {
1337	<	}
1338	<	}
1339	<	}
1340	<	}
1341	<	} finally {
1342	<	lock.unlock();
1343	<	}
1344	<	}
1345	<
1346	<
1347	<	/*
1348	<	* Nodes for event barrier to manage idle threads.  Queue nodes
1349	<	* are basic Treiber stack nodes, also used for spare stack.
1350	<	*
1351	<	* The event barrier has an event count and a wait queue (actually
1352	<	* a Treiber stack).  Workers are enabled to look for work when
1353	<	* the eventCount is incremented. If they fail to find work, they
1354	<	* may wait for next count. Upon release, threads help others wake
1355	<	* up.
1356	<	*
1357	<	* Synchronization events occur only in enough contexts to
1358	<	* maintain overall liveness:
1359	<	*
1360	<	*   - Submission of a new task to the pool
1361	<	*   - Resizes or other changes to the workers array
1362	<	*   - pool termination
1363	<	*   - A worker pushing a task on an empty queue
1364	<	*
1365	<	* The case of pushing a task occurs often enough, and is heavy
1366	<	* enough compared to simple stack pushes, to require special
1367	<	* handling: Method signalWork returns without advancing count if
1368	<	* the queue appears to be empty.  This would ordinarily result in
1369	<	* races causing some queued waiters not to be woken up. To avoid
1370	<	* this, the first worker enqueued in method sync (see
1371	<	* syncIsReleasable) rescans for tasks after being enqueued, and
1372	<	* helps signal if any are found. This works well because the
1373	<	* worker has nothing better to do, and so might as well help
1374	<	* alleviate the overhead and contention on the threads actually
1375	<	* doing work.  Also, since event counts increments on task
1376	<	* availability exist to maintain liveness (rather than to force
1377	<	* refreshes etc), it is OK for callers to exit early if
1378	<	* contending with another signaller.
1379	<	*/
1380	<	static final class WaitQueueNode {
1381	<	WaitQueueNode next; // only written before enqueued
1382	<	volatile ForkJoinWorkerThread thread; // nulled to cancel wait
1383	<	final long count; // unused for spare stack
1384	<
1385	<	WaitQueueNode(long c, ForkJoinWorkerThread w) {
1386	<	count = c;
1387	<	thread = w;
1388	<	}
1389	<
1390	<	/**
1391	<	* Wakes up waiter, returning false if known to already
1392	<	*/
1393	<	boolean signal() {
1394	<	ForkJoinWorkerThread t = thread;
1395	<	if (t == null)
1396	<	return false;
1397	<	thread = null;
1398	<	LockSupport.unpark(t);
1399	<	return true;
1400	<	}
1401	<
1402	<	/**
1403	<	* Awaits release on sync.
1404	<	*/
1405	<	void awaitSyncRelease(ForkJoinPool p) {
1406	<	while (thread != null && !p.syncIsReleasable(this))
1407	<	LockSupport.park(this);
1408	<	}
1409	<
1410	<	/**
1411	<	* Awaits resumption as spare.
1412	<	*/
1413	<	void awaitSpareRelease() {
1414	<	while (thread != null) {
1415	<	if (!Thread.interrupted())
1416	<	LockSupport.park(this);
1417	<	}
1418	<	}
1419	<	}
1420	<
1421	<	/**
1422	<	* Ensures that no thread is waiting for count to advance from the
1423	<	* current value of eventCount read on entry to this method, by
1424	<	* releasing waiting threads if necessary.
1425	<	*
1426	<	* @return the count
1427	<	*/
1428	<	final long ensureSync() {
1429	<	long c = eventCount;
1430	<	WaitQueueNode q;
1431	<	while ((q = syncStack) != null && q.count < c) {
1432	<	if (casBarrierStack(q, null)) {
1433	<	do {
1434	<	q.signal();
1435	<	} while ((q = q.next) != null);
1436	<	break;
1437	<	}
1438	<	}
1439	<	return c;
1440	<	}
1441	<
1442	<	/**
1443	<	* Increments event count and releases waiting threads.
1444	<	*/
1445	<	private void signalIdleWorkers() {
1446	<	long c;
1447	<	do {} while (!casEventCount(c = eventCount, c+1));
1448	<	ensureSync();
1449	<	}
1450	<
1451	<	/**
1452	<	* Signals threads waiting to poll a task. Because method sync
1453	<	* rechecks availability, it is OK to only proceed if queue
1454	<	* appears to be non-empty, and OK to skip under contention to
1455	<	* increment count (since some other thread succeeded).
1456	<	*/
1457	<	final void signalWork() {
1458	<	long c;
1459	<	WaitQueueNode q;
1460	<	if (syncStack != null &&
1461	<	casEventCount(c = eventCount, c+1) &&
1462	<	(((q = syncStack) != null && q.count <= c) &&
1463	<	(!casBarrierStack(q, q.next) || !q.signal())))
1464	<	ensureSync();
1465	<	}
1466	<
1467	<	/**
1468	<	* Waits until event count advances from last value held by
1469	<	* caller, or if excess threads, caller is resumed as spare, or
1470	<	* caller or pool is terminating. Updates caller's event on exit.
1471	<	*
1472	<	* @param w the calling worker thread
1473	<	*/
1474	<	final void sync(ForkJoinWorkerThread w) {
1475	<	updateStealCount(w); // Transfer w's count while it is idle
1476	<
1477	<	while (!w.isShutdown() && !isTerminating() && !suspendIfSpare(w)) {
1478	<	long prev = w.lastEventCount;
1479	<	WaitQueueNode node = null;
1480	<	WaitQueueNode h;
1481	<	while (eventCount == prev &&
1482	<	((h = syncStack) == null || h.count == prev)) {
1483	<	if (node == null)
1484	<	node = new WaitQueueNode(prev, w);
1485	<	if (casBarrierStack(node.next = h, node)) {
1486	<	node.awaitSyncRelease(this);
1487	<	break;
1488	<	}
1489	<	}
1490	<	long ec = ensureSync();
1491	<	if (ec != prev) {
1492	<	w.lastEventCount = ec;
1493	<	break;
1494	<	}
1495	<	}
1496	<	}
1497	<
1498	<	/**
1499	<	* Returns true if worker waiting on sync can proceed:
1500	<	*  - on signal (thread == null)
1501	<	*  - on event count advance (winning race to notify vs signaller)
1502	<	*  - on interrupt
1503	<	*  - if the first queued node, we find work available
1504	<	* If node was not signalled and event count not advanced on exit,
1505	<	* then we also help advance event count.
1506	<	*
1507	<	* @return true if node can be released
1508	<	*/
1509	<	final boolean syncIsReleasable(WaitQueueNode node) {
1510	<	long prev = node.count;
1511	<	if (!Thread.interrupted() && node.thread != null &&
1512	<	(node.next != null ||
1513	<	!ForkJoinWorkerThread.hasQueuedTasks(workers)) &&
1514	<	eventCount == prev)
1515	<	return false;
1516	<	if (node.thread != null) {
1517	<	node.thread = null;
1518	<	long ec = eventCount;
1519	<	if (prev <= ec) // help signal
1520	<	casEventCount(ec, ec+1);
1521	<	}
1522	<	return true;
1523	<	}
1524	<
1525	<	/**
1526	<	* Returns true if a new sync event occurred since last call to
1527	<	* sync or this method, if so, updating caller's count.
1528	<	*/
1529	<	final boolean hasNewSyncEvent(ForkJoinWorkerThread w) {
1530	<	long lc = w.lastEventCount;
1531	<	long ec = ensureSync();
1532	<	if (ec == lc)
1533	<	return false;
1534	<	w.lastEventCount = ec;
1535	<	return true;
1536	<	}
1537	<
1538	<	//  Parallelism maintenance
1539	<
1540	<	/**
1541	<	* Decrements running count; if too low, adds spare.
1542	<	*
1543	<	* Conceptually, all we need to do here is add or resume a
1544	<	* spare thread when one is about to block (and remove or
1545	<	* suspend it later when unblocked -- see suspendIfSpare).
1546	<	* However, implementing this idea requires coping with
1547	<	* several problems: we have imperfect information about the
1548	<	* states of threads. Some count updates can and usually do
1549	<	* lag run state changes, despite arrangements to keep them
1550	<	* accurate (for example, when possible, updating counts
1551	<	* before signalling or resuming), especially when running on
1552	<	* dynamic JVMs that don't optimize the infrequent paths that
1553	<	* update counts. Generating too many threads can make these
1554	<	* problems become worse, because excess threads are more
1555	<	* likely to be context-switched with others, slowing them all
1556	<	* down, especially if there is no work available, so all are
1557	<	* busy scanning or idling.  Also, excess spare threads can
1558	<	* only be suspended or removed when they are idle, not
1559	<	* immediately when they aren't needed. So adding threads will
1560	<	* raise parallelism level for longer than necessary.  Also,
1561	<	* FJ applications often encounter highly transient peaks when
1562	<	* many threads are blocked joining, but for less time than it
1563	<	* takes to create or resume spares.
1564	<	*
1565	<	* @param joinMe if non-null, return early if done
1566	<	* @param maintainParallelism if true, try to stay within
1567	<	* target counts, else create only to avoid starvation
1568	<	* @return true if joinMe known to be done
1569	<	*/
1570	<	final boolean preJoin(ForkJoinTask<?> joinMe,
1571	<	boolean maintainParallelism) {
1572	<	maintainParallelism &= maintainsParallelism; // overrride
1573	<	boolean dec = false;  // true when running count decremented
1574	<	while (spareStack == null || !tryResumeSpare(dec)) {
1575	<	int counts = workerCounts;
1576	<	if (dec || (dec = casWorkerCounts(counts, --counts))) {
1577	<	// CAS cheat
1578	<	if (!needSpare(counts, maintainParallelism))
1579	<	break;
1580	<	if (joinMe.status < 0)
1581	<	return true;
1582	<	if (tryAddSpare(counts))
1583	<	break;
1584	<	}
1585	<	}
1586	<	return false;
1587	<	}
1588	<
1589	<	/**
1590	<	* Same idea as preJoin
1591	<	*/
1592	<	final boolean preBlock(ManagedBlocker blocker,
1593	<	boolean maintainParallelism) {
1594	<	maintainParallelism &= maintainsParallelism;
1595	<	boolean dec = false;
1596	<	while (spareStack == null || !tryResumeSpare(dec)) {
1597	<	int counts = workerCounts;
1598	<	if (dec || (dec = casWorkerCounts(counts, --counts))) {
1599	<	if (!needSpare(counts, maintainParallelism))
1600	<	break;
1601	<	if (blocker.isReleasable())
1602	<	return true;
1603	<	if (tryAddSpare(counts))
1604	<	break;
1605	<	}
1606	<	}
1607	<	return false;
1608	<	}
1609	<
1610	<	/**
1611	<	* Returns true if a spare thread appears to be needed.  If
1612	<	* maintaining parallelism, returns true when the deficit in
1613	<	* running threads is more than the surplus of total threads, and
1614	<	* there is apparently some work to do.  This self-limiting rule
1615	<	* means that the more threads that have already been added, the
1616	<	* less parallelism we will tolerate before adding another.
1617	<	*
1618	<	* @param counts current worker counts
1619	<	* @param maintainParallelism try to maintain parallelism
1620	<	*/
1621	<	private boolean needSpare(int counts, boolean maintainParallelism) {
1622	<	int ps = parallelism;
1623	<	int rc = runningCountOf(counts);
1624	<	int tc = totalCountOf(counts);
1625	<	int runningDeficit = ps - rc;
1626	<	int totalSurplus = tc - ps;
1627	<	return (tc < maxPoolSize &&
1628	<	(rc == 0 || totalSurplus < 0 ||
1629	<	(maintainParallelism &&
1630	<	runningDeficit > totalSurplus &&
1631	<	ForkJoinWorkerThread.hasQueuedTasks(workers))));
1632	<	}
1633	<
1634	<	/**
1635	<	* Adds a spare worker if lock available and no more than the
1636	<	* expected numbers of threads exist.
1637	<	*
1638	<	* @return true if successful
1639	<	*/
1640	<	private boolean tryAddSpare(int expectedCounts) {
1641	<	final ReentrantLock lock = this.workerLock;
1642	<	int expectedRunning = runningCountOf(expectedCounts);
1643	<	int expectedTotal = totalCountOf(expectedCounts);
1644	<	boolean success = false;
1645	<	boolean locked = false;
1646	<	// confirm counts while locking; CAS after obtaining lock
1647	<	try {
1648	<	for (;;) {
1649	<	int s = workerCounts;
1650	<	int tc = totalCountOf(s);
1651	<	int rc = runningCountOf(s);
1652	<	if (rc > expectedRunning || tc > expectedTotal)
1653	<	break;
1654	<	if (!locked && !(locked = lock.tryLock()))
1655	<	break;
1656	<	if (casWorkerCounts(s, workerCountsFor(tc+1, rc+1))) {
1657	<	createAndStartSpare(tc);
1658	<	success = true;
1659	<	break;
1660	<	}
1661	<	}
1662	<	} finally {
1663	<	if (locked)
1664	<	lock.unlock();
1665	<	}
1666	<	return success;
1667	<	}
1668	<
1669	<	/**
1670	<	* Adds the kth spare worker. On entry, pool counts are already
1671	<	* adjusted to reflect addition.
1672	<	*/
1673	<	private void createAndStartSpare(int k) {
1674	<	ForkJoinWorkerThread w = null;
1675	<	ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(k + 1);
1676	<	int len = ws.length;
1677	<	// Probably, we can place at slot k. If not, find empty slot
1678	<	if (k < len && ws[k] != null) {
1679	<	for (k = 0; k < len && ws[k] != null; ++k)
1680	<	;
1681	<	}
1682	<	if (k < len && !isTerminating() && (w = createWorker(k)) != null) {
1683	<	ws[k] = w;
1684	<	w.start();
1685	<	}
1686	<	else
1687	<	updateWorkerCount(-1); // adjust on failure
1688	<	signalIdleWorkers();
1689	<	}
1690	<
1691	<	/**
1692	<	* Suspends calling thread w if there are excess threads.  Called
1693	<	* only from sync.  Spares are enqueued in a Treiber stack using
1694	<	* the same WaitQueueNodes as barriers.  They are resumed mainly
1695	<	* in preJoin, but are also woken on pool events that require all
1696	<	* threads to check run state.
1697	<	*
1698	<	* @param w the caller
1699	<	*/
1700	<	private boolean suspendIfSpare(ForkJoinWorkerThread w) {
1701	<	WaitQueueNode node = null;
1702	<	int s;
1703	<	while (parallelism < runningCountOf(s = workerCounts)) {
1704	<	if (node == null)
1705	<	node = new WaitQueueNode(0, w);
1706	<	if (casWorkerCounts(s, s-1)) { // representation-dependent
1707	<	// push onto stack
1708	<	do {} while (!casSpareStack(node.next = spareStack, node));
1709	<	// block until released by resumeSpare
1710	<	node.awaitSpareRelease();
1711	<	return true;
1712	<	}
1713	<	}
1714	<	return false;
1715	<	}
1716	<
1717	<	/**
1718	<	* Tries to pop and resume a spare thread.
1719	<	*
1720	<	* @param updateCount if true, increment running count on success
1721	<	* @return true if successful
1722	<	*/
1723	<	private boolean tryResumeSpare(boolean updateCount) {
1724	<	WaitQueueNode q;
1725	<	while ((q = spareStack) != null) {
1726	<	if (casSpareStack(q, q.next)) {
1727	<	if (updateCount)
1728	<	updateRunningCount(1);
1729	<	q.signal();
1730	<	return true;
1731	<	}
1732	<	}
1733	<	return false;
1734	<	}
1735	<
1736	<	/**
1737	<	* Pops and resumes all spare threads. Same idea as ensureSync.
1738	<	*
1739	<	* @return true if any spares released
1740	<	*/
1741	<	private boolean resumeAllSpares() {
1742	<	WaitQueueNode q;
1743	<	while ( (q = spareStack) != null) {
1744	<	if (casSpareStack(q, null)) {
1745	<	do {
1746	<	updateRunningCount(1);
1747	<	q.signal();
1748	<	} while ((q = q.next) != null);
1749	<	return true;
1750	<	}
1751	<	}
1752	<	return false;
1753	<	}
1754	<
1755	<	/**
1756	<	* Pops and shuts down excessive spare threads. Call only while
1757	<	* holding lock. This is not guaranteed to eliminate all excess
1758	<	* threads, only those suspended as spares, which are the ones
1759	<	* unlikely to be needed in the future.
1760	<	*/
1761	<	private void trimSpares() {
1762	<	int surplus = totalCountOf(workerCounts) - parallelism;
1763	<	WaitQueueNode q;
1764	<	while (surplus > 0 && (q = spareStack) != null) {
1765	<	if (casSpareStack(q, null)) {
1766	<	do {
1767	<	updateRunningCount(1);
1768	<	ForkJoinWorkerThread w = q.thread;
1769	<	if (w != null && surplus > 0 &&
1770	<	runningCountOf(workerCounts) > 0 && w.shutdown())
1771	<	--surplus;
1772	<	q.signal();
1773	<	} while ((q = q.next) != null);
1774	<	}
1775	<	}
1905	>	return terminationLatch.await(timeout, unit);
1906		}
1907
1908		/**
1909		* Interface for extending managed parallelism for tasks running
1910	<	* in ForkJoinPools. A ManagedBlocker provides two methods.
1911	<	* Method {@code isReleasable} must return true if blocking is not
1912	<	* necessary. Method {@code block} blocks the current thread if
1913	<	* necessary (perhaps internally invoking {@code isReleasable}
1914	<	* before actually blocking.).
1910	>	* in {@link ForkJoinPool}s.
1911	>	*
1912	>	* <p>A {@code ManagedBlocker} provides two methods.
1913	>	* Method {@code isReleasable} must return {@code true} if
1914	>	* blocking is not necessary. Method {@code block} blocks the
1915	>	* current thread if necessary (perhaps internally invoking
1916	>	* {@code isReleasable} before actually blocking).
1917		*
1918		* <p>For example, here is a ManagedBlocker based on a
1919		* ReentrantLock:
#	Line 1805 | Line 1937 | public class ForkJoinPool extends Abstra
1937		* Possibly blocks the current thread, for example waiting for
1938		* a lock or condition.
1939		*
1940	<	* @return true if no additional blocking is necessary (i.e.,
1941	<	* if isReleasable would return true)
1940	>	* @return {@code true} if no additional blocking is necessary
1941	>	* (i.e., if isReleasable would return true)
1942		* @throws InterruptedException if interrupted while waiting
1943		* (the method is not required to do so, but is allowed to)
1944		*/
1945		boolean block() throws InterruptedException;
1946
1947		/**
1948	<	* Returns true if blocking is unnecessary.
1948	>	* Returns {@code true} if blocking is unnecessary.
1949		*/
1950		boolean isReleasable();
1951		}
1952
1953		/**
1954		* Blocks in accord with the given blocker.  If the current thread
1955	<	* is a ForkJoinWorkerThread, this method possibly arranges for a
1956	<	* spare thread to be activated if necessary to ensure parallelism
1957	<	* while the current thread is blocked.  If
1958	<	* {@code maintainParallelism} is true and the pool supports
1959	<	* it ({@link #getMaintainsParallelism}), this method attempts to
1960	<	* maintain the pool's nominal parallelism. Otherwise it activates
1961	<	* a thread only if necessary to avoid complete starvation. This
1962	<	* option may be preferable when blockages use timeouts, or are
1963	<	* almost always brief.
1955	>	* is a {@link ForkJoinWorkerThread}, this method possibly
1956	>	* arranges for a spare thread to be activated if necessary to
1957	>	* ensure parallelism while the current thread is blocked.
1958	>	*
1959	>	* <p>If {@code maintainParallelism} is {@code true} and the pool
1960	>	* supports it ({@link #getMaintainsParallelism}), this method
1961	>	* attempts to maintain the pool's nominal parallelism. Otherwise
1962	>	* it activates a thread only if necessary to avoid complete
1963	>	* starvation. This option may be preferable when blockages use
1964	>	* timeouts, or are almost always brief.
1965		*
1966	<	* <p> If the caller is not a ForkJoinTask, this method is behaviorally
1967	<	* equivalent to
1966	>	* <p>If the caller is not a {@link ForkJoinTask}, this method is
1967	>	* behaviorally equivalent to
1968		*  <pre> {@code
1969		* while (!blocker.isReleasable())
1970		*   if (blocker.block())
1971		*     return;
1972		* }</pre>
1973	<	* If the caller is a ForkJoinTask, then the pool may first
1974	<	* be expanded to ensure parallelism, and later adjusted.
1973	>	*
1974	>	* If the caller is a {@code ForkJoinTask}, then the pool may
1975	>	* first be expanded to ensure parallelism, and later adjusted.
1976		*
1977		* @param blocker the blocker
1978	<	* @param maintainParallelism if true and supported by this pool,
1979	<	* attempt to maintain the pool's nominal parallelism; otherwise
1980	<	* activate a thread only if necessary to avoid complete
1981	<	* starvation.
1978	>	* @param maintainParallelism if {@code true} and supported by
1979	>	* this pool, attempt to maintain the pool's nominal parallelism;
1980	>	* otherwise activate a thread only if necessary to avoid
1981	>	* complete starvation.
1982		* @throws InterruptedException if blocker.block did so
1983		*/
1984		public static void managedBlock(ManagedBlocker blocker,
1985		boolean maintainParallelism)
1986		throws InterruptedException {
1987		Thread t = Thread.currentThread();
1988	<	ForkJoinPool pool = ((t instanceof ForkJoinWorkerThread) ?
1989	<	((ForkJoinWorkerThread) t).pool : null);
1990	<	if (!blocker.isReleasable()) {
1991	<	try {
1992	<	if (pool == null ||
1859	<	!pool.preBlock(blocker, maintainParallelism))
1860	<	awaitBlocker(blocker);
1861	<	} finally {
1862	<	if (pool != null)
1863	<	pool.updateRunningCount(1);
1864	<	}
1865	<	}
1988	>	if (t instanceof ForkJoinWorkerThread)
1989	>	((ForkJoinWorkerThread) t).pool.
1990	>	awaitBlocker(blocker, maintainParallelism);
1991	>	else
1992	>	awaitBlocker(blocker);
1993		}
1994
1995	+	/**
1996	+	* Performs Non-FJ blocking
1997	+	*/
1998		private static void awaitBlocker(ManagedBlocker blocker)
1999		throws InterruptedException {
2000		do {} while (!blocker.isReleasable() && !blocker.block());
2001		}
2002
2003	<	// AbstractExecutorService overrides
2003	>	// AbstractExecutorService overrides.  These rely on undocumented
2004	>	// fact that ForkJoinTask.adapt returns ForkJoinTasks that also
2005	>	// implement RunnableFuture.
2006
2007		protected <T> RunnableFuture<T> newTaskFor(Runnable runnable, T value) {
2008	<	return new AdaptedRunnable<T>(runnable, value);
2008	>	return (RunnableFuture<T>) ForkJoinTask.adapt(runnable, value);
2009		}
2010
2011		protected <T> RunnableFuture<T> newTaskFor(Callable<T> callable) {
2012	<	return new AdaptedCallable<T>(callable);
2012	>	return (RunnableFuture<T>) ForkJoinTask.adapt(callable);
2013		}
2014
2015	+	// Unsafe mechanics
2016	+
2017	+	private static final sun.misc.Unsafe UNSAFE = getUnsafe();
2018	+	private static final long workerCountsOffset =
2019	+	objectFieldOffset("workerCounts", ForkJoinPool.class);
2020	+	private static final long runStateOffset =
2021	+	objectFieldOffset("runState", ForkJoinPool.class);
2022	+	private static final long eventCountOffset =
2023	+	objectFieldOffset("eventCount", ForkJoinPool.class);
2024	+	private static final long eventWaitersOffset =
2025	+	objectFieldOffset("eventWaiters",ForkJoinPool.class);
2026	+	private static final long stealCountOffset =
2027	+	objectFieldOffset("stealCount",ForkJoinPool.class);
2028	+
2029	+
2030	+	private static long objectFieldOffset(String field, Class<?> klazz) {
2031	+	try {
2032	+	return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
2033	+	} catch (NoSuchFieldException e) {
2034	+	// Convert Exception to corresponding Error
2035	+	NoSuchFieldError error = new NoSuchFieldError(field);
2036	+	error.initCause(e);
2037	+	throw error;
2038	+	}
2039	+	}
2040
2041	<	// Unsafe mechanics for jsr166y 3rd party package.
2041	>	/**
2042	>	* Returns a sun.misc.Unsafe.  Suitable for use in a 3rd party package.
2043	>	* Replace with a simple call to Unsafe.getUnsafe when integrating
2044	>	* into a jdk.
2045	>	*
2046	>	* @return a sun.misc.Unsafe
2047	>	*/
2048		private static sun.misc.Unsafe getUnsafe() {
2049		try {
2050		return sun.misc.Unsafe.getUnsafe();
2051		} catch (SecurityException se) {
2052		try {
2053		return java.security.AccessController.doPrivileged
2054	<	(new java.security.PrivilegedExceptionAction<sun.misc.Unsafe>() {
2054	>	(new java.security
2055	>	.PrivilegedExceptionAction<sun.misc.Unsafe>() {
2056		public sun.misc.Unsafe run() throws Exception {
2057	<	return getUnsafeByReflection();
2057	>	java.lang.reflect.Field f = sun.misc
2058	>	.Unsafe.class.getDeclaredField("theUnsafe");
2059	>	f.setAccessible(true);
2060	>	return (sun.misc.Unsafe) f.get(null);
2061		}});
2062		} catch (java.security.PrivilegedActionException e) {
2063		throw new RuntimeException("Could not initialize intrinsics",
#	Line 1898 | Line 2065 | public class ForkJoinPool extends Abstra
2065		}
2066		}
2067		}
1901	–
1902	–	private static sun.misc.Unsafe getUnsafeByReflection()
1903	–	throws NoSuchFieldException, IllegalAccessException {
1904	–	java.lang.reflect.Field f =
1905	–	sun.misc.Unsafe.class.getDeclaredField("theUnsafe");
1906	–	f.setAccessible(true);
1907	–	return (sun.misc.Unsafe) f.get(null);
1908	–	}
1909	–
1910	–	private static long fieldOffset(String fieldName, Class<?> klazz) {
1911	–	try {
1912	–	return UNSAFE.objectFieldOffset(klazz.getDeclaredField(fieldName));
1913	–	} catch (NoSuchFieldException e) {
1914	–	// Convert Exception to Error
1915	–	NoSuchFieldError error = new NoSuchFieldError(fieldName);
1916	–	error.initCause(e);
1917	–	throw error;
1918	–	}
1919	–	}
1920	–
1921	–	private static final sun.misc.Unsafe UNSAFE = getUnsafe();
1922	–	static final long eventCountOffset =
1923	–	fieldOffset("eventCount", ForkJoinPool.class);
1924	–	static final long workerCountsOffset =
1925	–	fieldOffset("workerCounts", ForkJoinPool.class);
1926	–	static final long runControlOffset =
1927	–	fieldOffset("runControl", ForkJoinPool.class);
1928	–	static final long syncStackOffset =
1929	–	fieldOffset("syncStack",ForkJoinPool.class);
1930	–	static final long spareStackOffset =
1931	–	fieldOffset("spareStack", ForkJoinPool.class);
1932	–
1933	–	private boolean casEventCount(long cmp, long val) {
1934	–	return UNSAFE.compareAndSwapLong(this, eventCountOffset, cmp, val);
1935	–	}
1936	–	private boolean casWorkerCounts(int cmp, int val) {
1937	–	return UNSAFE.compareAndSwapInt(this, workerCountsOffset, cmp, val);
1938	–	}
1939	–	private boolean casRunControl(int cmp, int val) {
1940	–	return UNSAFE.compareAndSwapInt(this, runControlOffset, cmp, val);
1941	–	}
1942	–	private boolean casSpareStack(WaitQueueNode cmp, WaitQueueNode val) {
1943	–	return UNSAFE.compareAndSwapObject(this, spareStackOffset, cmp, val);
1944	–	}
1945	–	private boolean casBarrierStack(WaitQueueNode cmp, WaitQueueNode val) {
1946	–	return UNSAFE.compareAndSwapObject(this, syncStackOffset, cmp, val);
1947	–	}
2068		}

Diff Legend

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