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 1.7 by jsr166, Mon Jul 20 21:45:06 2009 UTC vs.
Revision 1.189 by jsr166, Sat Sep 12 19:16:45 2015 UTC

#	Line 1 | Line 1
1		/*
2		* Written by Doug Lea with assistance from members of JCP JSR-166
3		* Expert Group and released to the public domain, as explained at
4	<	* http://creativecommons.org/licenses/publicdomain
4	>	* http://creativecommons.org/publicdomain/zero/1.0/
5		*/
6
7		package jsr166y;
8	<	import java.util.*;
9	<	import java.util.concurrent.*;
10	<	import java.util.concurrent.locks.*;
11	<	import java.util.concurrent.atomic.*;
12	<	import sun.misc.Unsafe;
13	<	import java.lang.reflect.*;
8	>
9	>	import java.util.ArrayList;
10	>	import java.util.Arrays;
11	>	import java.util.Collection;
12	>	import java.util.Collections;
13	>	import java.util.List;
14	>	import java.util.concurrent.AbstractExecutorService;
15	>	import java.util.concurrent.Callable;
16	>	import java.util.concurrent.ExecutorService;
17	>	import java.util.concurrent.Future;
18	>	import java.util.concurrent.RejectedExecutionException;
19	>	import java.util.concurrent.RunnableFuture;
20	>	import java.util.concurrent.TimeUnit;
21
22		/**
23	<	* An {@link ExecutorService} for running {@link ForkJoinTask}s.  A
24	<	* ForkJoinPool provides the entry point for submissions from
25	<	* non-ForkJoinTasks, as well as management and monitoring operations.
26	<	* 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.
23	>	* An {@link ExecutorService} for running {@link ForkJoinTask}s.
24	>	* A {@code ForkJoinPool} provides the entry point for submissions
25	>	* from non-{@code ForkJoinTask} clients, as well as management and
26	>	* monitoring operations.
27	>	*
28	>	* <p>A {@code ForkJoinPool} differs from other kinds of {@link
29	>	* ExecutorService} mainly by virtue of employing
30	>	* <em>work-stealing</em>: all threads in the pool attempt to find and
31	>	* execute tasks submitted to the pool and/or created by other active
32	>	* tasks (eventually blocking waiting for work if none exist). This
33	>	* enables efficient processing when most tasks spawn other subtasks
34	>	* (as do most {@code ForkJoinTask}s), as well as when many small
35	>	* tasks are submitted to the pool from external clients.  Especially
36	>	* when setting <em>asyncMode</em> to true in constructors, {@code
37	>	* ForkJoinPool}s may also be appropriate for use with event-style
38	>	* tasks that are never joined.
39		*
40	<	* <p>ForkJoinPools differ from other kinds of Executors mainly in
41	<	* that they provide <em>work-stealing</em>: all threads in the pool
42	<	* attempt to find and execute subtasks created by other active tasks
43	<	* (eventually blocking if none exist). This makes them efficient when
44	<	* most tasks spawn other subtasks (as do most ForkJoinTasks), as well
45	<	* as the mixed execution of some plain Runnable- or Callable- based
30	<	* activities along with ForkJoinTasks. When setting
31	<	* <tt>setAsyncMode</tt>, a ForkJoinPools may also be appropriate for
32	<	* use with fine-grained tasks that are never joined. Otherwise, other
33	<	* ExecutorService implementations are typically more appropriate
34	<	* choices.
40	>	* <p>A static {@link #commonPool()} is available and appropriate for
41	>	* most applications. The common pool is used by any ForkJoinTask that
42	>	* is not explicitly submitted to a specified pool. Using the common
43	>	* pool normally reduces resource usage (its threads are slowly
44	>	* reclaimed during periods of non-use, and reinstated upon subsequent
45	>	* use).
46		*
47	<	* <p>A ForkJoinPool may be constructed with a given parallelism level
48	<	* (target pool size), which it attempts to maintain by dynamically
49	<	* adding, suspending, or resuming threads, even if some tasks are
50	<	* waiting to join others. However, no such adjustments are performed
51	<	* in the face of blocked IO or other unmanaged synchronization. The
52	<	* nested <code>ManagedBlocker</code> interface enables extension of
53	<	* the kinds of synchronization accommodated.  The target parallelism
54	<	* level may also be changed dynamically (<code>setParallelism</code>)
55	<	* and thread construction can be limited using methods
56	<	* <code>setMaximumPoolSize</code> and/or
46	<	* <code>setMaintainsParallelism</code>.
47	>	* <p>For applications that require separate or custom pools, a {@code
48	>	* ForkJoinPool} may be constructed with a given target parallelism
49	>	* level; by default, equal to the number of available processors. The
50	>	* pool attempts to maintain enough active (or available) threads by
51	>	* dynamically adding, suspending, or resuming internal worker
52	>	* threads, even if some tasks are stalled waiting to join
53	>	* others. However, no such adjustments are guaranteed in the face of
54	>	* blocked I/O or other unmanaged synchronization. The nested {@link
55	>	* ManagedBlocker} interface enables extension of the kinds of
56	>	* synchronization accommodated.
57		*
58		* <p>In addition to execution and lifecycle control methods, this
59		* class provides status check methods (for example
60	<	* <code>getStealCount</code>) that are intended to aid in developing,
60	>	* {@link #getStealCount}) that are intended to aid in developing,
61		* tuning, and monitoring fork/join applications. Also, method
62	<	* <code>toString</code> returns indications of pool state in a
62	>	* {@link #toString} returns indications of pool state in a
63		* convenient form for informal monitoring.
64		*
65	+	* <p>As is the case with other ExecutorServices, there are three
66	+	* main task execution methods summarized in the following table.
67	+	* These are designed to be used primarily by clients not already
68	+	* engaged in fork/join computations in the current pool.  The main
69	+	* forms of these methods accept instances of {@code ForkJoinTask},
70	+	* but overloaded forms also allow mixed execution of plain {@code
71	+	* Runnable}- or {@code Callable}- based activities as well.  However,
72	+	* tasks that are already executing in a pool should normally instead
73	+	* use the within-computation forms listed in the table unless using
74	+	* async event-style tasks that are not usually joined, in which case
75	+	* there is little difference among choice of methods.
76	+	*
77	+	* <table BORDER CELLPADDING=3 CELLSPACING=1>
78	+	*  <tr>
79	+	*    <td></td>
80	+	*    <td ALIGN=CENTER> <b>Call from non-fork/join clients</b></td>
81	+	*    <td ALIGN=CENTER> <b>Call from within fork/join computations</b></td>
82	+	*  </tr>
83	+	*  <tr>
84	+	*    <td> <b>Arrange async execution</td>
85	+	*    <td> {@link #execute(ForkJoinTask)}</td>
86	+	*    <td> {@link ForkJoinTask#fork}</td>
87	+	*  </tr>
88	+	*  <tr>
89	+	*    <td> <b>Await and obtain result</td>
90	+	*    <td> {@link #invoke(ForkJoinTask)}</td>
91	+	*    <td> {@link ForkJoinTask#invoke}</td>
92	+	*  </tr>
93	+	*  <tr>
94	+	*    <td> <b>Arrange exec and obtain Future</td>
95	+	*    <td> {@link #submit(ForkJoinTask)}</td>
96	+	*    <td> {@link ForkJoinTask#fork} (ForkJoinTasks <em>are</em> Futures)</td>
97	+	*  </tr>
98	+	* </table>
99	+	*
100	+	* <p>The common pool is by default constructed with default
101	+	* parameters, but these may be controlled by setting three {@link
102	+	* System#getProperty system properties} with prefix {@code
103	+	* java.util.concurrent.ForkJoinPool.common}: {@code parallelism} --
104	+	* an integer greater than zero, {@code threadFactory} -- the class
105	+	* name of a {@link ForkJoinWorkerThreadFactory}, and {@code
106	+	* exceptionHandler} -- the class name of a {@link
107	+	* java.lang.Thread.UncaughtExceptionHandler
108	+	* Thread.UncaughtExceptionHandler}. Upon any error in establishing
109	+	* these settings, default parameters are used.
110	+	*
111		* <p><b>Implementation notes</b>: This implementation restricts the
112		* maximum number of running threads to 32767. Attempts to create
113	<	* pools with greater than the maximum result in
114	<	* IllegalArgumentExceptions.
113	>	* pools with greater than the maximum number result in
114	>	* {@code IllegalArgumentException}.
115	>	*
116	>	* <p>This implementation rejects submitted tasks (that is, by throwing
117	>	* {@link RejectedExecutionException}) only when the pool is shut down
118	>	* or internal resources have been exhausted.
119	>	*
120	>	* @since 1.7
121	>	* @author Doug Lea
122		*/
123		public class ForkJoinPool extends AbstractExecutorService {
124
125		/*
126	<	* See the extended comments interspersed below for design,
127	<	* rationale, and walkthroughs.
126	>	* Implementation Overview
127	>	*
128	>	* This class and its nested classes provide the main
129	>	* functionality and control for a set of worker threads:
130	>	* Submissions from non-FJ threads enter into submission queues.
131	>	* Workers take these tasks and typically split them into subtasks
132	>	* that may be stolen by other workers.  Preference rules give
133	>	* first priority to processing tasks from their own queues (LIFO
134	>	* or FIFO, depending on mode), then to randomized FIFO steals of
135	>	* tasks in other queues.
136	>	*
137	>	* WorkQueues
138	>	* ==========
139	>	*
140	>	* Most operations occur within work-stealing queues (in nested
141	>	* class WorkQueue).  These are special forms of Deques that
142	>	* support only three of the four possible end-operations -- push,
143	>	* pop, and poll (aka steal), under the further constraints that
144	>	* push and pop are called only from the owning thread (or, as
145	>	* extended here, under a lock), while poll may be called from
146	>	* other threads.  (If you are unfamiliar with them, you probably
147	>	* want to read Herlihy and Shavit's book "The Art of
148	>	* Multiprocessor programming", chapter 16 describing these in
149	>	* more detail before proceeding.)  The main work-stealing queue
150	>	* design is roughly similar to those in the papers "Dynamic
151	>	* Circular Work-Stealing Deque" by Chase and Lev, SPAA 2005
152	>	* (http://research.sun.com/scalable/pubs/index.html) and
153	>	* "Idempotent work stealing" by Michael, Saraswat, and Vechev,
154	>	* PPoPP 2009 (http://portal.acm.org/citation.cfm?id=1504186).
155	>	* The main differences ultimately stem from GC requirements that
156	>	* we null out taken slots as soon as we can, to maintain as small
157	>	* a footprint as possible even in programs generating huge
158	>	* numbers of tasks. To accomplish this, we shift the CAS
159	>	* arbitrating pop vs poll (steal) from being on the indices
160	>	* ("base" and "top") to the slots themselves.  So, both a
161	>	* successful pop and poll mainly entail a CAS of a slot from
162	>	* non-null to null.  Because we rely on CASes of references, we
163	>	* do not need tag bits on base or top.  They are simple ints as
164	>	* used in any circular array-based queue (see for example
165	>	* ArrayDeque).  Updates to the indices must still be ordered in a
166	>	* way that guarantees that top == base means the queue is empty,
167	>	* but otherwise may err on the side of possibly making the queue
168	>	* appear nonempty when a push, pop, or poll have not fully
169	>	* committed. Note that this means that the poll operation,
170	>	* considered individually, is not wait-free. One thief cannot
171	>	* successfully continue until another in-progress one (or, if
172	>	* previously empty, a push) completes.  However, in the
173	>	* aggregate, we ensure at least probabilistic non-blockingness.
174	>	* If an attempted steal fails, a thief always chooses a different
175	>	* random victim target to try next. So, in order for one thief to
176	>	* progress, it suffices for any in-progress poll or new push on
177	>	* any empty queue to complete. (This is why we normally use
178	>	* method pollAt and its variants that try once at the apparent
179	>	* base index, else consider alternative actions, rather than
180	>	* method poll.)
181	>	*
182	>	* This approach also enables support of a user mode in which local
183	>	* task processing is in FIFO, not LIFO order, simply by using
184	>	* poll rather than pop.  This can be useful in message-passing
185	>	* frameworks in which tasks are never joined.  However neither
186	>	* mode considers affinities, loads, cache localities, etc, so
187	>	* rarely provide the best possible performance on a given
188	>	* machine, but portably provide good throughput by averaging over
189	>	* these factors.  (Further, even if we did try to use such
190	>	* information, we do not usually have a basis for exploiting it.
191	>	* For example, some sets of tasks profit from cache affinities,
192	>	* but others are harmed by cache pollution effects.)
193	>	*
194	>	* WorkQueues are also used in a similar way for tasks submitted
195	>	* to the pool. We cannot mix these tasks in the same queues used
196	>	* for work-stealing (this would contaminate lifo/fifo
197	>	* processing). Instead, we randomly associate submission queues
198	>	* with submitting threads, using a form of hashing.  The
199	>	* ThreadLocal Submitter class contains a value initially used as
200	>	* a hash code for choosing existing queues, but may be randomly
201	>	* repositioned upon contention with other submitters.  In
202	>	* essence, submitters act like workers except that they are
203	>	* restricted to executing local tasks that they submitted (or in
204	>	* the case of CountedCompleters, others with the same root task).
205	>	* However, because most shared/external queue operations are more
206	>	* expensive than internal, and because, at steady state, external
207	>	* submitters will compete for CPU with workers, ForkJoinTask.join
208	>	* and related methods disable them from repeatedly helping to
209	>	* process tasks if all workers are active.  Insertion of tasks in
210	>	* shared mode requires a lock (mainly to protect in the case of
211	>	* resizing) but we use only a simple spinlock (using bits in
212	>	* field qlock), because submitters encountering a busy queue move
213	>	* on to try or create other queues -- they block only when
214	>	* creating and registering new queues.
215	>	*
216	>	* Management
217	>	* ==========
218	>	*
219	>	* The main throughput advantages of work-stealing stem from
220	>	* decentralized control -- workers mostly take tasks from
221	>	* themselves or each other. We cannot negate this in the
222	>	* implementation of other management responsibilities. The main
223	>	* tactic for avoiding bottlenecks is packing nearly all
224	>	* essentially atomic control state into two volatile variables
225	>	* that are by far most often read (not written) as status and
226	>	* consistency checks.
227	>	*
228	>	* Field "ctl" contains 64 bits holding all the information needed
229	>	* to atomically decide to add, inactivate, enqueue (on an event
230	>	* queue), dequeue, and/or re-activate workers.  To enable this
231	>	* packing, we restrict maximum parallelism to (1<<15)-1 (which is
232	>	* far in excess of normal operating range) to allow ids, counts,
233	>	* and their negations (used for thresholding) to fit into 16bit
234	>	* fields.
235	>	*
236	>	* Field "plock" is a form of sequence lock with a saturating
237	>	* shutdown bit (similarly for per-queue "qlocks"), mainly
238	>	* protecting updates to the workQueues array, as well as to
239	>	* enable shutdown.  When used as a lock, it is normally only very
240	>	* briefly held, so is nearly always available after at most a
241	>	* brief spin, but we use a monitor-based backup strategy to
242	>	* block when needed.
243	>	*
244	>	* Recording WorkQueues.  WorkQueues are recorded in the
245	>	* "workQueues" array that is created upon first use and expanded
246	>	* if necessary.  Updates to the array while recording new workers
247	>	* and unrecording terminated ones are protected from each other
248	>	* by a lock but the array is otherwise concurrently readable, and
249	>	* accessed directly.  To simplify index-based operations, the
250	>	* array size is always a power of two, and all readers must
251	>	* tolerate null slots. Worker queues are at odd indices. Shared
252	>	* (submission) queues are at even indices, up to a maximum of 64
253	>	* slots, to limit growth even if array needs to expand to add
254	>	* more workers. Grouping them together in this way simplifies and
255	>	* speeds up task scanning.
256	>	*
257	>	* All worker thread creation is on-demand, triggered by task
258	>	* submissions, replacement of terminated workers, and/or
259	>	* compensation for blocked workers. However, all other support
260	>	* code is set up to work with other policies.  To ensure that we
261	>	* do not hold on to worker references that would prevent GC, ALL
262	>	* accesses to workQueues are via indices into the workQueues
263	>	* array (which is one source of some of the messy code
264	>	* constructions here). In essence, the workQueues array serves as
265	>	* a weak reference mechanism. Thus for example the wait queue
266	>	* field of ctl stores indices, not references.  Access to the
267	>	* workQueues in associated methods (for example signalWork) must
268	>	* both index-check and null-check the IDs. All such accesses
269	>	* ignore bad IDs by returning out early from what they are doing,
270	>	* since this can only be associated with termination, in which
271	>	* case it is OK to give up.  All uses of the workQueues array
272	>	* also check that it is non-null (even if previously
273	>	* non-null). This allows nulling during termination, which is
274	>	* currently not necessary, but remains an option for
275	>	* resource-revocation-based shutdown schemes. It also helps
276	>	* reduce JIT issuance of uncommon-trap code, which tends to
277	>	* unnecessarily complicate control flow in some methods.
278	>	*
279	>	* Event Queuing. Unlike HPC work-stealing frameworks, we cannot
280	>	* let workers spin indefinitely scanning for tasks when none can
281	>	* be found immediately, and we cannot start/resume workers unless
282	>	* there appear to be tasks available.  On the other hand, we must
283	>	* quickly prod them into action when new tasks are submitted or
284	>	* generated. In many usages, ramp-up time to activate workers is
285	>	* the main limiting factor in overall performance (this is
286	>	* compounded at program start-up by JIT compilation and
287	>	* allocation). So we try to streamline this as much as possible.
288	>	* We park/unpark workers after placing in an event wait queue
289	>	* when they cannot find work. This "queue" is actually a simple
290	>	* Treiber stack, headed by the "id" field of ctl, plus a 15bit
291	>	* counter value (that reflects the number of times a worker has
292	>	* been inactivated) to avoid ABA effects (we need only as many
293	>	* version numbers as worker threads). Successors are held in
294	>	* field WorkQueue.nextWait.  Queuing deals with several intrinsic
295	>	* races, mainly that a task-producing thread can miss seeing (and
296	>	* signalling) another thread that gave up looking for work but
297	>	* has not yet entered the wait queue. We solve this by requiring
298	>	* a full sweep of all workers (via repeated calls to method
299	>	* scan()) both before and after a newly waiting worker is added
300	>	* to the wait queue. During a rescan, the worker might release
301	>	* some other queued worker rather than itself, which has the same
302	>	* net effect. Because enqueued workers may actually be rescanning
303	>	* rather than waiting, we set and clear the "parker" field of
304	>	* WorkQueues to reduce unnecessary calls to unpark.  (This
305	>	* requires a secondary recheck to avoid missed signals.)  Note
306	>	* the unusual conventions about Thread.interrupts surrounding
307	>	* parking and other blocking: Because interrupts are used solely
308	>	* to alert threads to check termination, which is checked anyway
309	>	* upon blocking, we clear status (using Thread.interrupted)
310	>	* before any call to park, so that park does not immediately
311	>	* return due to status being set via some other unrelated call to
312	>	* interrupt in user code.
313	>	*
314	>	* Signalling.  We create or wake up workers only when there
315	>	* appears to be at least one task they might be able to find and
316	>	* execute. However, many other threads may notice the same task
317	>	* and each signal to wake up a thread that might take it. So in
318	>	* general, pools will be over-signalled.  When a submission is
319	>	* added or another worker adds a task to a queue that has fewer
320	>	* than two tasks, they signal waiting workers (or trigger
321	>	* creation of new ones if fewer than the given parallelism level
322	>	* -- signalWork), and may leave a hint to the unparked worker to
323	>	* help signal others upon wakeup).  These primary signals are
324	>	* buttressed by others (see method helpSignal) whenever other
325	>	* threads scan for work or do not have a task to process.  On
326	>	* most platforms, signalling (unpark) overhead time is noticeably
327	>	* long, and the time between signalling a thread and it actually
328	>	* making progress can be very noticeably long, so it is worth
329	>	* offloading these delays from critical paths as much as
330	>	* possible.
331	>	*
332	>	* Trimming workers. To release resources after periods of lack of
333	>	* use, a worker starting to wait when the pool is quiescent will
334	>	* time out and terminate if the pool has remained quiescent for a
335	>	* given period -- a short period if there are more threads than
336	>	* parallelism, longer as the number of threads decreases. This
337	>	* will slowly propagate, eventually terminating all workers after
338	>	* periods of non-use.
339	>	*
340	>	* Shutdown and Termination. A call to shutdownNow atomically sets
341	>	* a plock bit and then (non-atomically) sets each worker's
342	>	* qlock status, cancels all unprocessed tasks, and wakes up
343	>	* all waiting workers.  Detecting whether termination should
344	>	* commence after a non-abrupt shutdown() call requires more work
345	>	* and bookkeeping. We need consensus about quiescence (i.e., that
346	>	* there is no more work). The active count provides a primary
347	>	* indication but non-abrupt shutdown still requires a rechecking
348	>	* scan for any workers that are inactive but not queued.
349	>	*
350	>	* Joining Tasks
351	>	* =============
352	>	*
353	>	* Any of several actions may be taken when one worker is waiting
354	>	* to join a task stolen (or always held) by another.  Because we
355	>	* are multiplexing many tasks on to a pool of workers, we can't
356	>	* just let them block (as in Thread.join).  We also cannot just
357	>	* reassign the joiner's run-time stack with another and replace
358	>	* it later, which would be a form of "continuation", that even if
359	>	* possible is not necessarily a good idea since we sometimes need
360	>	* both an unblocked task and its continuation to progress.
361	>	* Instead we combine two tactics:
362	>	*
363	>	*   Helping: Arranging for the joiner to execute some task that it
364	>	*      would be running if the steal had not occurred.
365	>	*
366	>	*   Compensating: Unless there are already enough live threads,
367	>	*      method tryCompensate() may create or re-activate a spare
368	>	*      thread to compensate for blocked joiners until they unblock.
369	>	*
370	>	* A third form (implemented in tryRemoveAndExec) amounts to
371	>	* helping a hypothetical compensator: If we can readily tell that
372	>	* a possible action of a compensator is to steal and execute the
373	>	* task being joined, the joining thread can do so directly,
374	>	* without the need for a compensation thread (although at the
375	>	* expense of larger run-time stacks, but the tradeoff is
376	>	* typically worthwhile).
377	>	*
378	>	* The ManagedBlocker extension API can't use helping so relies
379	>	* only on compensation in method awaitBlocker.
380	>	*
381	>	* The algorithm in tryHelpStealer entails a form of "linear"
382	>	* helping: Each worker records (in field currentSteal) the most
383	>	* recent task it stole from some other worker. Plus, it records
384	>	* (in field currentJoin) the task it is currently actively
385	>	* joining. Method tryHelpStealer uses these markers to try to
386	>	* find a worker to help (i.e., steal back a task from and execute
387	>	* it) that could hasten completion of the actively joined task.
388	>	* In essence, the joiner executes a task that would be on its own
389	>	* local deque had the to-be-joined task not been stolen. This may
390	>	* be seen as a conservative variant of the approach in Wagner &
391	>	* Calder "Leapfrogging: a portable technique for implementing
392	>	* efficient futures" SIGPLAN Notices, 1993
393	>	* (http://portal.acm.org/citation.cfm?id=155354). It differs in
394	>	* that: (1) We only maintain dependency links across workers upon
395	>	* steals, rather than use per-task bookkeeping.  This sometimes
396	>	* requires a linear scan of workQueues array to locate stealers,
397	>	* but often doesn't because stealers leave hints (that may become
398	>	* stale/wrong) of where to locate them.  It is only a hint
399	>	* because a worker might have had multiple steals and the hint
400	>	* records only one of them (usually the most current).  Hinting
401	>	* isolates cost to when it is needed, rather than adding to
402	>	* per-task overhead.  (2) It is "shallow", ignoring nesting and
403	>	* potentially cyclic mutual steals.  (3) It is intentionally
404	>	* racy: field currentJoin is updated only while actively joining,
405	>	* which means that we miss links in the chain during long-lived
406	>	* tasks, GC stalls etc (which is OK since blocking in such cases
407	>	* is usually a good idea).  (4) We bound the number of attempts
408	>	* to find work (see MAX_HELP) and fall back to suspending the
409	>	* worker and if necessary replacing it with another.
410	>	*
411	>	* Helping actions for CountedCompleters are much simpler: Method
412	>	* helpComplete can take and execute any task with the same root
413	>	* as the task being waited on. However, this still entails some
414	>	* traversal of completer chains, so is less efficient than using
415	>	* CountedCompleters without explicit joins.
416	>	*
417	>	* It is impossible to keep exactly the target parallelism number
418	>	* of threads running at any given time.  Determining the
419	>	* existence of conservatively safe helping targets, the
420	>	* availability of already-created spares, and the apparent need
421	>	* to create new spares are all racy, so we rely on multiple
422	>	* retries of each.  Compensation in the apparent absence of
423	>	* helping opportunities is challenging to control on JVMs, where
424	>	* GC and other activities can stall progress of tasks that in
425	>	* turn stall out many other dependent tasks, without us being
426	>	* able to determine whether they will ever require compensation.
427	>	* Even though work-stealing otherwise encounters little
428	>	* degradation in the presence of more threads than cores,
429	>	* aggressively adding new threads in such cases entails risk of
430	>	* unwanted positive feedback control loops in which more threads
431	>	* cause more dependent stalls (as well as delayed progress of
432	>	* unblocked threads to the point that we know they are available)
433	>	* leading to more situations requiring more threads, and so
434	>	* on. This aspect of control can be seen as an (analytically
435	>	* intractable) game with an opponent that may choose the worst
436	>	* (for us) active thread to stall at any time.  We take several
437	>	* precautions to bound losses (and thus bound gains), mainly in
438	>	* methods tryCompensate and awaitJoin.
439	>	*
440	>	* Common Pool
441	>	* ===========
442	>	*
443	>	* The static common Pool always exists after static
444	>	* initialization.  Since it (or any other created pool) need
445	>	* never be used, we minimize initial construction overhead and
446	>	* footprint to the setup of about a dozen fields, with no nested
447	>	* allocation. Most bootstrapping occurs within method
448	>	* fullExternalPush during the first submission to the pool.
449	>	*
450	>	* When external threads submit to the common pool, they can
451	>	* perform some subtask processing (see externalHelpJoin and
452	>	* related methods).  We do not need to record whether these
453	>	* submissions are to the common pool -- if not, externalHelpJoin
454	>	* returns quickly (at the most helping to signal some common pool
455	>	* workers). These submitters would otherwise be blocked waiting
456	>	* for completion, so the extra effort (with liberally sprinkled
457	>	* task status checks) in inapplicable cases amounts to an odd
458	>	* form of limited spin-wait before blocking in ForkJoinTask.join.
459	>	*
460	>	* Style notes
461	>	* ===========
462	>	*
463	>	* There is a lot of representation-level coupling among classes
464	>	* ForkJoinPool, ForkJoinWorkerThread, and ForkJoinTask.  The
465	>	* fields of WorkQueue maintain data structures managed by
466	>	* ForkJoinPool, so are directly accessed.  There is little point
467	>	* trying to reduce this, since any associated future changes in
468	>	* representations will need to be accompanied by algorithmic
469	>	* changes anyway. Several methods intrinsically sprawl because
470	>	* they must accumulate sets of consistent reads of volatiles held
471	>	* in local variables.  Methods signalWork() and scan() are the
472	>	* main bottlenecks, so are especially heavily
473	>	* micro-optimized/mangled.  There are lots of inline assignments
474	>	* (of form "while ((local = field) != 0)") which are usually the
475	>	* simplest way to ensure the required read orderings (which are
476	>	* sometimes critical). This leads to a "C"-like style of listing
477	>	* declarations of these locals at the heads of methods or blocks.
478	>	* There are several occurrences of the unusual "do {} while
479	>	* (!cas...)"  which is the simplest way to force an update of a
480	>	* CAS'ed variable. There are also other coding oddities (including
481	>	* several unnecessary-looking hoisted null checks) that help
482	>	* some methods perform reasonably even when interpreted (not
483	>	* compiled).
484	>	*
485	>	* The order of declarations in this file is:
486	>	* (1) Static utility functions
487	>	* (2) Nested (static) classes
488	>	* (3) Static fields
489	>	* (4) Fields, along with constants used when unpacking some of them
490	>	* (5) Internal control methods
491	>	* (6) Callbacks and other support for ForkJoinTask methods
492	>	* (7) Exported methods
493	>	* (8) Static block initializing statics in minimally dependent order
494		*/
495
496	<	/** Mask for packing and unpacking shorts */
68	<	private static final int  shortMask = 0xffff;
496	>	// Static utilities
497
498	<	/** Max pool size -- must be a power of two minus 1 */
499	<	private static final int MAX_THREADS =  0x7FFF;
498	>	/**
499	>	* If there is a security manager, makes sure caller has
500	>	* permission to modify threads.
501	>	*/
502	>	private static void checkPermission() {
503	>	SecurityManager security = System.getSecurityManager();
504	>	if (security != null)
505	>	security.checkPermission(modifyThreadPermission);
506	>	}
507	>
508	>	// Nested classes
509
510		/**
511	<	* Factory for creating new ForkJoinWorkerThreads.  A
512	<	* ForkJoinWorkerThreadFactory must be defined and used for
513	<	* ForkJoinWorkerThread subclasses that extend base functionality
514	<	* or initialize threads with different contexts.
511	>	* Factory for creating new {@link ForkJoinWorkerThread}s.
512	>	* A {@code ForkJoinWorkerThreadFactory} must be defined and used
513	>	* for {@code ForkJoinWorkerThread} subclasses that extend base
514	>	* functionality or initialize threads with different contexts.
515		*/
516		public static interface ForkJoinWorkerThreadFactory {
517		/**
518		* Returns a new worker thread operating in the given pool.
519		*
520		* @param pool the pool this thread works in
521	<	* @throws NullPointerException if pool is null;
521	>	* @throws NullPointerException if the pool is null
522		*/
523		public ForkJoinWorkerThread newThread(ForkJoinPool pool);
524		}
525
526		/**
527	<	* Default ForkJoinWorkerThreadFactory implementation, creates a
527	>	* Default ForkJoinWorkerThreadFactory implementation; creates a
528		* new ForkJoinWorkerThread.
529		*/
530	<	static class  DefaultForkJoinWorkerThreadFactory
530	>	static final class DefaultForkJoinWorkerThreadFactory
531		implements ForkJoinWorkerThreadFactory {
532	<	public ForkJoinWorkerThread newThread(ForkJoinPool pool) {
533	<	try {
534	<	return new ForkJoinWorkerThread(pool);
535	<	} catch (OutOfMemoryError oom)  {
532	>	public final ForkJoinWorkerThread newThread(ForkJoinPool pool) {
533	>	return new ForkJoinWorkerThread(pool);
534	>	}
535	>	}
536	>
537	>	/**
538	>	* Per-thread records for threads that submit to pools. Currently
539	>	* holds only pseudo-random seed / index that is used to choose
540	>	* submission queues in method externalPush. In the future, this may
541	>	* also incorporate a means to implement different task rejection
542	>	* and resubmission policies.
543	>	*
544	>	* Seeds for submitters and workers/workQueues work in basically
545	>	* the same way but are initialized and updated using slightly
546	>	* different mechanics. Both are initialized using the same
547	>	* approach as in class ThreadLocal, where successive values are
548	>	* unlikely to collide with previous values. Seeds are then
549	>	* randomly modified upon collisions using xorshifts, which
550	>	* requires a non-zero seed.
551	>	*/
552	>	static final class Submitter {
553	>	int seed;
554	>	Submitter(int s) { seed = s; }
555	>	}
556	>
557	>	/**
558	>	* Class for artificial tasks that are used to replace the target
559	>	* of local joins if they are removed from an interior queue slot
560	>	* in WorkQueue.tryRemoveAndExec. We don't need the proxy to
561	>	* actually do anything beyond having a unique identity.
562	>	*/
563	>	static final class EmptyTask extends ForkJoinTask<Void> {
564	>	private static final long serialVersionUID = -7721805057305804111L;
565	>	EmptyTask() { status = ForkJoinTask.NORMAL; } // force done
566	>	public final Void getRawResult() { return null; }
567	>	public final void setRawResult(Void x) {}
568	>	public final boolean exec() { return true; }
569	>	}
570	>
571	>	/**
572	>	* Queues supporting work-stealing as well as external task
573	>	* submission. See above for main rationale and algorithms.
574	>	* Implementation relies heavily on "Unsafe" intrinsics
575	>	* and selective use of "volatile":
576	>	*
577	>	* Field "base" is the index (mod array.length) of the least valid
578	>	* queue slot, which is always the next position to steal (poll)
579	>	* from if nonempty. Reads and writes require volatile orderings
580	>	* but not CAS, because updates are only performed after slot
581	>	* CASes.
582	>	*
583	>	* Field "top" is the index (mod array.length) of the next queue
584	>	* slot to push to or pop from. It is written only by owner thread
585	>	* for push, or under lock for external/shared push, and accessed
586	>	* by other threads only after reading (volatile) base.  Both top
587	>	* and base are allowed to wrap around on overflow, but (top -
588	>	* base) (or more commonly -(base - top) to force volatile read of
589	>	* base before top) still estimates size. The lock ("qlock") is
590	>	* forced to -1 on termination, causing all further lock attempts
591	>	* to fail. (Note: we don't need CAS for termination state because
592	>	* upon pool shutdown, all shared-queues will stop being used
593	>	* anyway.)  Nearly all lock bodies are set up so that exceptions
594	>	* within lock bodies are "impossible" (modulo JVM errors that
595	>	* would cause failure anyway.)
596	>	*
597	>	* The array slots are read and written using the emulation of
598	>	* volatiles/atomics provided by Unsafe. Insertions must in
599	>	* general use putOrderedObject as a form of releasing store to
600	>	* ensure that all writes to the task object are ordered before
601	>	* its publication in the queue.  All removals entail a CAS to
602	>	* null.  The array is always a power of two. To ensure safety of
603	>	* Unsafe array operations, all accesses perform explicit null
604	>	* checks and implicit bounds checks via power-of-two masking.
605	>	*
606	>	* In addition to basic queuing support, this class contains
607	>	* fields described elsewhere to control execution. It turns out
608	>	* to work better memory-layout-wise to include them in this class
609	>	* rather than a separate class.
610	>	*
611	>	* Performance on most platforms is very sensitive to placement of
612	>	* instances of both WorkQueues and their arrays -- we absolutely
613	>	* do not want multiple WorkQueue instances or multiple queue
614	>	* arrays sharing cache lines. (It would be best for queue objects
615	>	* and their arrays to share, but there is nothing available to
616	>	* help arrange that).  Unfortunately, because they are recorded
617	>	* in a common array, WorkQueue instances are often moved to be
618	>	* adjacent by garbage collectors. To reduce impact, we use field
619	>	* padding that works OK on common platforms; this effectively
620	>	* trades off slightly slower average field access for the sake of
621	>	* avoiding really bad worst-case access. (Until better JVM
622	>	* support is in place, this padding is dependent on transient
623	>	* properties of JVM field layout rules.) We also take care in
624	>	* allocating, sizing and resizing the array. Non-shared queue
625	>	* arrays are initialized by workers before use. Others are
626	>	* allocated on first use.
627	>	*/
628	>	static final class WorkQueue {
629	>	/**
630	>	* Capacity of work-stealing queue array upon initialization.
631	>	* Must be a power of two; at least 4, but should be larger to
632	>	* reduce or eliminate cacheline sharing among queues.
633	>	* Currently, it is much larger, as a partial workaround for
634	>	* the fact that JVMs often place arrays in locations that
635	>	* share GC bookkeeping (especially cardmarks) such that
636	>	* per-write accesses encounter serious memory contention.
637	>	*/
638	>	static final int INITIAL_QUEUE_CAPACITY = 1 << 13;
639	>
640	>	/**
641	>	* Maximum size for queue arrays. Must be a power of two less
642	>	* than or equal to 1 << (31 - width of array entry) to ensure
643	>	* lack of wraparound of index calculations, but defined to a
644	>	* value a bit less than this to help users trap runaway
645	>	* programs before saturating systems.
646	>	*/
647	>	static final int MAXIMUM_QUEUE_CAPACITY = 1 << 26; // 64M
648	>
649	>	// Heuristic padding to ameliorate unfortunate memory placements
650	>	volatile long pad00, pad01, pad02, pad03, pad04, pad05, pad06;
651	>
652	>	int seed;                  // for random scanning; initialize nonzero
653	>	volatile int eventCount;   // encoded inactivation count; < 0 if inactive
654	>	int nextWait;              // encoded record of next event waiter
655	>	int hint;                  // steal or signal hint (index)
656	>	int poolIndex;             // index of this queue in pool (or 0)
657	>	final int mode;            // 0: lifo, > 0: fifo, < 0: shared
658	>	int nsteals;               // number of steals
659	>	volatile int qlock;        // 1: locked, -1: terminate; else 0
660	>	volatile int base;         // index of next slot for poll
661	>	int top;                   // index of next slot for push
662	>	ForkJoinTask<?>[] array;   // the elements (initially unallocated)
663	>	final ForkJoinPool pool;   // the containing pool (may be null)
664	>	final ForkJoinWorkerThread owner; // owning thread or null if shared
665	>	volatile Thread parker;    // == owner during call to park; else null
666	>	volatile ForkJoinTask<?> currentJoin;  // task being joined in awaitJoin
667	>	ForkJoinTask<?> currentSteal; // current non-local task being executed
668	>
669	>	volatile Object pad10, pad11, pad12, pad13, pad14, pad15, pad16, pad17;
670	>	volatile Object pad18, pad19, pad1a, pad1b, pad1c, pad1d;
671	>
672	>	WorkQueue(ForkJoinPool pool, ForkJoinWorkerThread owner, int mode,
673	>	int seed) {
674	>	this.pool = pool;
675	>	this.owner = owner;
676	>	this.mode = mode;
677	>	this.seed = seed;
678	>	// Place indices in the center of array (that is not yet allocated)
679	>	base = top = INITIAL_QUEUE_CAPACITY >>> 1;
680	>	}
681	>
682	>	/**
683	>	* Returns the approximate number of tasks in the queue.
684	>	*/
685	>	final int queueSize() {
686	>	int n = base - top;       // non-owner callers must read base first
687	>	return (n >= 0) ? 0 : -n; // ignore transient negative
688	>	}
689	>
690	>	/**
691	>	* Provides a more accurate estimate of whether this queue has
692	>	* any tasks than does queueSize, by checking whether a
693	>	* near-empty queue has at least one unclaimed task.
694	>	*/
695	>	final boolean isEmpty() {
696	>	ForkJoinTask<?>[] a; int m, s;
697	>	int n = base - (s = top);
698	>	return (n >= 0 ||
699	>	(n == -1 &&
700	>	((a = array) == null ||
701	>	(m = a.length - 1) < 0 ||
702	>	U.getObject
703	>	(a, (long)((m & (s - 1)) << ASHIFT) + ABASE) == null)));
704	>	}
705	>
706	>	/**
707	>	* Pushes a task. Call only by owner in unshared queues.  (The
708	>	* shared-queue version is embedded in method externalPush.)
709	>	*
710	>	* @param task the task. Caller must ensure non-null.
711	>	* @throws RejectedExecutionException if array cannot be resized
712	>	*/
713	>	final void push(ForkJoinTask<?> task) {
714	>	ForkJoinTask<?>[] a; ForkJoinPool p;
715	>	int s = top, m, n;
716	>	if ((a = array) != null) {    // ignore if queue removed
717	>	int j = (((m = a.length - 1) & s) << ASHIFT) + ABASE;
718	>	U.putOrderedObject(a, j, task);
719	>	if ((n = (top = s + 1) - base) <= 2) {
720	>	if ((p = pool) != null)
721	>	p.signalWork(this);
722	>	}
723	>	else if (n >= m)
724	>	growArray();
725	>	}
726	>	}
727	>
728	>	/**
729	>	* Initializes or doubles the capacity of array. Call either
730	>	* by owner or with lock held -- it is OK for base, but not
731	>	* top, to move while resizings are in progress.
732	>	*/
733	>	final ForkJoinTask<?>[] growArray() {
734	>	ForkJoinTask<?>[] oldA = array;
735	>	int size = oldA != null ? oldA.length << 1 : INITIAL_QUEUE_CAPACITY;
736	>	if (size > MAXIMUM_QUEUE_CAPACITY)
737	>	throw new RejectedExecutionException("Queue capacity exceeded");
738	>	int oldMask, t, b;
739	>	ForkJoinTask<?>[] a = array = new ForkJoinTask<?>[size];
740	>	if (oldA != null && (oldMask = oldA.length - 1) >= 0 &&
741	>	(t = top) - (b = base) > 0) {
742	>	int mask = size - 1;
743	>	do {
744	>	ForkJoinTask<?> x;
745	>	int oldj = ((b & oldMask) << ASHIFT) + ABASE;
746	>	int j    = ((b &    mask) << ASHIFT) + ABASE;
747	>	x = (ForkJoinTask<?>)U.getObjectVolatile(oldA, oldj);
748	>	if (x != null &&
749	>	U.compareAndSwapObject(oldA, oldj, x, null))
750	>	U.putObjectVolatile(a, j, x);
751	>	} while (++b != t);
752	>	}
753	>	return a;
754	>	}
755	>
756	>	/**
757	>	* Takes next task, if one exists, in LIFO order.  Call only
758	>	* by owner in unshared queues.
759	>	*/
760	>	final ForkJoinTask<?> pop() {
761	>	ForkJoinTask<?>[] a; ForkJoinTask<?> t; int m;
762	>	if ((a = array) != null && (m = a.length - 1) >= 0) {
763	>	for (int s; (s = top - 1) - base >= 0;) {
764	>	long j = ((m & s) << ASHIFT) + ABASE;
765	>	if ((t = (ForkJoinTask<?>)U.getObject(a, j)) == null)
766	>	break;
767	>	if (U.compareAndSwapObject(a, j, t, null)) {
768	>	top = s;
769	>	return t;
770	>	}
771	>	}
772	>	}
773	>	return null;
774	>	}
775	>
776	>	/**
777	>	* Takes a task in FIFO order if b is base of queue and a task
778	>	* can be claimed without contention. Specialized versions
779	>	* appear in ForkJoinPool methods scan and tryHelpStealer.
780	>	*/
781	>	final ForkJoinTask<?> pollAt(int b) {
782	>	ForkJoinTask<?> t; ForkJoinTask<?>[] a;
783	>	if ((a = array) != null) {
784	>	int j = (((a.length - 1) & b) << ASHIFT) + ABASE;
785	>	if ((t = (ForkJoinTask<?>)U.getObjectVolatile(a, j)) != null &&
786	>	base == b &&
787	>	U.compareAndSwapObject(a, j, t, null)) {
788	>	base = b + 1;
789	>	return t;
790	>	}
791	>	}
792	>	return null;
793	>	}
794	>
795	>	/**
796	>	* Takes next task, if one exists, in FIFO order.
797	>	*/
798	>	final ForkJoinTask<?> poll() {
799	>	ForkJoinTask<?>[] a; int b; ForkJoinTask<?> t;
800	>	while ((b = base) - top < 0 && (a = array) != null) {
801	>	int j = (((a.length - 1) & b) << ASHIFT) + ABASE;
802	>	t = (ForkJoinTask<?>)U.getObjectVolatile(a, j);
803	>	if (t != null) {
804	>	if (base == b &&
805	>	U.compareAndSwapObject(a, j, t, null)) {
806	>	base = b + 1;
807	>	return t;
808	>	}
809	>	}
810	>	else if (base == b) {
811	>	if (b + 1 == top)
812	>	break;
813	>	Thread.yield(); // wait for lagging update (very rare)
814	>	}
815	>	}
816	>	return null;
817	>	}
818	>
819	>	/**
820	>	* Takes next task, if one exists, in order specified by mode.
821	>	*/
822	>	final ForkJoinTask<?> nextLocalTask() {
823	>	return mode == 0 ? pop() : poll();
824	>	}
825	>
826	>	/**
827	>	* Returns next task, if one exists, in order specified by mode.
828	>	*/
829	>	final ForkJoinTask<?> peek() {
830	>	ForkJoinTask<?>[] a = array; int m;
831	>	if (a == null || (m = a.length - 1) < 0)
832		return null;
833	+	int i = mode == 0 ? top - 1 : base;
834	+	int j = ((i & m) << ASHIFT) + ABASE;
835	+	return (ForkJoinTask<?>)U.getObjectVolatile(a, j);
836	+	}
837	+
838	+	/**
839	+	* Pops the given task only if it is at the current top.
840	+	* (A shared version is available only via FJP.tryExternalUnpush)
841	+	*/
842	+	final boolean tryUnpush(ForkJoinTask<?> t) {
843	+	ForkJoinTask<?>[] a; int s;
844	+	if ((a = array) != null && (s = top) != base &&
845	+	U.compareAndSwapObject
846	+	(a, (((a.length - 1) & --s) << ASHIFT) + ABASE, t, null)) {
847	+	top = s;
848	+	return true;
849	+	}
850	+	return false;
851	+	}
852	+
853	+	/**
854	+	* Removes and cancels all known tasks, ignoring any exceptions.
855	+	*/
856	+	final void cancelAll() {
857	+	ForkJoinTask.cancelIgnoringExceptions(currentJoin);
858	+	ForkJoinTask.cancelIgnoringExceptions(currentSteal);
859	+	for (ForkJoinTask<?> t; (t = poll()) != null; )
860	+	ForkJoinTask.cancelIgnoringExceptions(t);
861	+	}
862	+
863	+	/**
864	+	* Computes next value for random probes.  Scans don't require
865	+	* a very high quality generator, but also not a crummy one.
866	+	* Marsaglia xor-shift is cheap and works well enough.  Note:
867	+	* This is manually inlined in its usages in ForkJoinPool to
868	+	* avoid writes inside busy scan loops.
869	+	*/
870	+	final int nextSeed() {
871	+	int r = seed;
872	+	r ^= r << 13;
873	+	r ^= r >>> 17;
874	+	return seed = r ^= r << 5;
875	+	}
876	+
877	+	// Specialized execution methods
878	+
879	+	/**
880	+	* Pops and runs tasks until empty.
881	+	*/
882	+	private void popAndExecAll() {
883	+	// A bit faster than repeated pop calls
884	+	ForkJoinTask<?>[] a; int m, s; long j; ForkJoinTask<?> t;
885	+	while ((a = array) != null && (m = a.length - 1) >= 0 &&
886	+	(s = top - 1) - base >= 0 &&
887	+	(t = ((ForkJoinTask<?>)
888	+	U.getObject(a, j = ((m & s) << ASHIFT) + ABASE)))
889	+	!= null) {
890	+	if (U.compareAndSwapObject(a, j, t, null)) {
891	+	top = s;
892	+	t.doExec();
893	+	}
894	+	}
895	+	}
896	+
897	+	/**
898	+	* Polls and runs tasks until empty.
899	+	*/
900	+	private void pollAndExecAll() {
901	+	for (ForkJoinTask<?> t; (t = poll()) != null;)
902	+	t.doExec();
903	+	}
904	+
905	+	/**
906	+	* If present, removes from queue and executes the given task,
907	+	* or any other cancelled task. Returns (true) on any CAS
908	+	* or consistency check failure so caller can retry.
909	+	*
910	+	* @return false if no progress can be made, else true
911	+	*/
912	+	final boolean tryRemoveAndExec(ForkJoinTask<?> task) {
913	+	boolean stat = true, removed = false, empty = true;
914	+	ForkJoinTask<?>[] a; int m, s, b, n;
915	+	if ((a = array) != null && (m = a.length - 1) >= 0 &&
916	+	(n = (s = top) - (b = base)) > 0) {
917	+	for (ForkJoinTask<?> t;;) {           // traverse from s to b
918	+	int j = ((--s & m) << ASHIFT) + ABASE;
919	+	t = (ForkJoinTask<?>)U.getObjectVolatile(a, j);
920	+	if (t == null)                    // inconsistent length
921	+	break;
922	+	else if (t == task) {
923	+	if (s + 1 == top) {           // pop
924	+	if (!U.compareAndSwapObject(a, j, task, null))
925	+	break;
926	+	top = s;
927	+	removed = true;
928	+	}
929	+	else if (base == b)           // replace with proxy
930	+	removed = U.compareAndSwapObject(a, j, task,
931	+	new EmptyTask());
932	+	break;
933	+	}
934	+	else if (t.status >= 0)
935	+	empty = false;
936	+	else if (s + 1 == top) {          // pop and throw away
937	+	if (U.compareAndSwapObject(a, j, t, null))
938	+	top = s;
939	+	break;
940	+	}
941	+	if (--n == 0) {
942	+	if (!empty && base == b)
943	+	stat = false;
944	+	break;
945	+	}
946	+	}
947	+	}
948	+	if (removed)
949	+	task.doExec();
950	+	return stat;
951	+	}
952	+
953	+	/**
954	+	* Polls for and executes the given task or any other task in
955	+	* its CountedCompleter computation.
956	+	*/
957	+	final boolean pollAndExecCC(ForkJoinTask<?> root) {
958	+	ForkJoinTask<?>[] a; int b; Object o;
959	+	outer: while ((b = base) - top < 0 && (a = array) != null) {
960	+	long j = (((a.length - 1) & b) << ASHIFT) + ABASE;
961	+	if ((o = U.getObject(a, j)) == null ||
962	+	!(o instanceof CountedCompleter))
963	+	break;
964	+	for (CountedCompleter<?> t = (CountedCompleter<?>)o, r = t;;) {
965	+	if (r == root) {
966	+	if (base == b &&
967	+	U.compareAndSwapObject(a, j, t, null)) {
968	+	base = b + 1;
969	+	t.doExec();
970	+	return true;
971	+	}
972	+	else
973	+	break; // restart
974	+	}
975	+	if ((r = r.completer) == null)
976	+	break outer; // not part of root computation
977	+	}
978	+	}
979	+	return false;
980	+	}
981	+
982	+	/**
983	+	* Executes a top-level task and any local tasks remaining
984	+	* after execution.
985	+	*/
986	+	final void runTask(ForkJoinTask<?> t) {
987	+	if (t != null) {
988	+	(currentSteal = t).doExec();
989	+	currentSteal = null;
990	+	++nsteals;
991	+	if (base - top < 0) {       // process remaining local tasks
992	+	if (mode == 0)
993	+	popAndExecAll();
994	+	else
995	+	pollAndExecAll();
996	+	}
997	+	}
998	+	}
999	+
1000	+	/**
1001	+	* Executes a non-top-level (stolen) task.
1002	+	*/
1003	+	final void runSubtask(ForkJoinTask<?> t) {
1004	+	if (t != null) {
1005	+	ForkJoinTask<?> ps = currentSteal;
1006	+	(currentSteal = t).doExec();
1007	+	currentSteal = ps;
1008	+	}
1009	+	}
1010	+
1011	+	/**
1012	+	* Returns true if owned and not known to be blocked.
1013	+	*/
1014	+	final boolean isApparentlyUnblocked() {
1015	+	Thread wt; Thread.State s;
1016	+	return (eventCount >= 0 &&
1017	+	(wt = owner) != null &&
1018	+	(s = wt.getState()) != Thread.State.BLOCKED &&
1019	+	s != Thread.State.WAITING &&
1020	+	s != Thread.State.TIMED_WAITING);
1021	+	}
1022	+
1023	+	// Unsafe mechanics
1024	+	private static final sun.misc.Unsafe U;
1025	+	private static final long QLOCK;
1026	+	private static final int ABASE;
1027	+	private static final int ASHIFT;
1028	+	static {
1029	+	try {
1030	+	U = getUnsafe();
1031	+	Class<?> k = WorkQueue.class;
1032	+	Class<?> ak = ForkJoinTask[].class;
1033	+	QLOCK = U.objectFieldOffset
1034	+	(k.getDeclaredField("qlock"));
1035	+	ABASE = U.arrayBaseOffset(ak);
1036	+	int scale = U.arrayIndexScale(ak);
1037	+	if ((scale & (scale - 1)) != 0)
1038	+	throw new Error("data type scale not a power of two");
1039	+	ASHIFT = 31 - Integer.numberOfLeadingZeros(scale);
1040	+	} catch (Exception e) {
1041	+	throw new Error(e);
1042		}
1043		}
1044		}
1045
1046	+	// static fields (initialized in static initializer below)
1047	+
1048		/**
1049		* Creates a new ForkJoinWorkerThread. This factory is used unless
1050		* overridden in ForkJoinPool constructors.
1051		*/
1052		public static final ForkJoinWorkerThreadFactory
1053	<	defaultForkJoinWorkerThreadFactory =
110	<	new DefaultForkJoinWorkerThreadFactory();
1053	>	defaultForkJoinWorkerThreadFactory;
1054
1055		/**
1056	<	* Permission required for callers of methods that may start or
1057	<	* kill threads.
1056	>	* Per-thread submission bookkeeping. Shared across all pools
1057	>	* to reduce ThreadLocal pollution and because random motion
1058	>	* to avoid contention in one pool is likely to hold for others.
1059	>	* Lazily initialized on first submission (but null-checked
1060	>	* in other contexts to avoid unnecessary initialization).
1061		*/
1062	<	private static final RuntimePermission modifyThreadPermission =
117	<	new RuntimePermission("modifyThread");
1062	>	static final ThreadLocal<Submitter> submitters;
1063
1064		/**
1065	<	* If there is a security manager, makes sure caller has
1066	<	* permission to modify threads.
1065	>	* Permission required for callers of methods that may start or
1066	>	* kill threads.
1067		*/
1068	<	private static void checkPermission() {
124	<	SecurityManager security = System.getSecurityManager();
125	<	if (security != null)
126	<	security.checkPermission(modifyThreadPermission);
127	<	}
1068	>	private static final RuntimePermission modifyThreadPermission;
1069
1070		/**
1071	<	* Generator for assigning sequence numbers as pool names.
1071	>	* Common (static) pool. Non-null for public use unless a static
1072	>	* construction exception, but internal usages null-check on use
1073	>	* to paranoically avoid potential initialization circularities
1074	>	* as well as to simplify generated code.
1075		*/
1076	<	private static final AtomicInteger poolNumberGenerator =
133	<	new AtomicInteger();
1076	>	static final ForkJoinPool common;
1077
1078		/**
1079	<	* Array holding all worker threads in the pool. Initialized upon
137	<	* first use. Array size must be a power of two.  Updates and
138	<	* replacements are protected by workerLock, but it is always kept
139	<	* in a consistent enough state to be randomly accessed without
140	<	* locking by workers performing work-stealing.
1079	>	* Common pool parallelism. Must equal common.parallelism.
1080		*/
1081	<	volatile ForkJoinWorkerThread[] workers;
1081	>	static final int commonParallelism;
1082
1083		/**
1084	<	* Lock protecting access to workers.
1084	>	* Sequence number for creating workerNamePrefix.
1085		*/
1086	<	private final ReentrantLock workerLock;
1086	>	private static int poolNumberSequence;
1087
1088		/**
1089	<	* Condition for awaitTermination.
1089	>	* Returns the next sequence number. We don't expect this to
1090	>	* ever contend, so use simple builtin sync.
1091		*/
1092	<	private final Condition termination;
1092	>	private static final synchronized int nextPoolId() {
1093	>	return ++poolNumberSequence;
1094	>	}
1095
1096	<	/**
155	<	* The uncaught exception handler used when any worker
156	<	* abrupty terminates
157	<	*/
158	<	private Thread.UncaughtExceptionHandler ueh;
1096	>	// static constants
1097
1098		/**
1099	<	* Creation factory for worker threads.
1099	>	* Initial timeout value (in nanoseconds) for the thread
1100	>	* triggering quiescence to park waiting for new work. On timeout,
1101	>	* the thread will instead try to shrink the number of
1102	>	* workers. The value should be large enough to avoid overly
1103	>	* aggressive shrinkage during most transient stalls (long GCs
1104	>	* etc).
1105		*/
1106	<	private final ForkJoinWorkerThreadFactory factory;
1106	>	private static final long IDLE_TIMEOUT      = 2000L * 1000L * 1000L; // 2sec
1107
1108		/**
1109	<	* Head of stack of threads that were created to maintain
167	<	* parallelism when other threads blocked, but have since
168	<	* suspended when the parallelism level rose.
1109	>	* Timeout value when there are more threads than parallelism level
1110		*/
1111	<	private volatile WaitQueueNode spareStack;
1111	>	private static final long FAST_IDLE_TIMEOUT =  200L * 1000L * 1000L;
1112
1113		/**
1114	<	* Sum of per-thread steal counts, updated only when threads are
174	<	* idle or terminating.
1114	>	* Tolerance for idle timeouts, to cope with timer undershoots
1115		*/
1116	<	private final AtomicLong stealCount;
1116	>	private static final long TIMEOUT_SLOP = 2000000L;
1117
1118		/**
1119	<	* Queue for external submissions.
1119	>	* The maximum stolen->joining link depth allowed in method
1120	>	* tryHelpStealer.  Must be a power of two.  Depths for legitimate
1121	>	* chains are unbounded, but we use a fixed constant to avoid
1122	>	* (otherwise unchecked) cycles and to bound staleness of
1123	>	* traversal parameters at the expense of sometimes blocking when
1124	>	* we could be helping.
1125		*/
1126	<	private final LinkedTransferQueue<ForkJoinTask<?>> submissionQueue;
1126	>	private static final int MAX_HELP = 64;
1127
1128		/**
1129	<	* Head of Treiber stack for barrier sync. See below for explanation
1129	>	* Increment for seed generators. See class ThreadLocal for
1130	>	* explanation.
1131		*/
1132	<	private volatile WaitQueueNode syncStack;
1132	>	private static final int SEED_INCREMENT = 0x61c88647;
1133
1134	<	/**
1135	<	* The count for event barrier
1136	<	*/
1137	<	private volatile long eventCount;
1134	>	/*
1135	>	* Bits and masks for control variables
1136	>	*
1137	>	* Field ctl is a long packed with:
1138	>	* AC: Number of active running workers minus target parallelism (16 bits)
1139	>	* TC: Number of total workers minus target parallelism (16 bits)
1140	>	* ST: true if pool is terminating (1 bit)
1141	>	* EC: the wait count of top waiting thread (15 bits)
1142	>	* ID: poolIndex of top of Treiber stack of waiters (16 bits)
1143	>	*
1144	>	* When convenient, we can extract the upper 32 bits of counts and
1145	>	* the lower 32 bits of queue state, u = (int)(ctl >>> 32) and e =
1146	>	* (int)ctl.  The ec field is never accessed alone, but always
1147	>	* together with id and st. The offsets of counts by the target
1148	>	* parallelism and the positionings of fields makes it possible to
1149	>	* perform the most common checks via sign tests of fields: When
1150	>	* ac is negative, there are not enough active workers, when tc is
1151	>	* negative, there are not enough total workers, and when e is
1152	>	* negative, the pool is terminating.  To deal with these possibly
1153	>	* negative fields, we use casts in and out of "short" and/or
1154	>	* signed shifts to maintain signedness.
1155	>	*
1156	>	* When a thread is queued (inactivated), its eventCount field is
1157	>	* set negative, which is the only way to tell if a worker is
1158	>	* prevented from executing tasks, even though it must continue to
1159	>	* scan for them to avoid queuing races. Note however that
1160	>	* eventCount updates lag releases so usage requires care.
1161	>	*
1162	>	* Field plock is an int packed with:
1163	>	* SHUTDOWN: true if shutdown is enabled (1 bit)
1164	>	* SEQ:  a sequence lock, with PL_LOCK bit set if locked (30 bits)
1165	>	* SIGNAL: set when threads may be waiting on the lock (1 bit)
1166	>	*
1167	>	* The sequence number enables simple consistency checks:
1168	>	* Staleness of read-only operations on the workQueues array can
1169	>	* be checked by comparing plock before vs after the reads.
1170	>	*/
1171	>
1172	>	// bit positions/shifts for fields
1173	>	private static final int  AC_SHIFT   = 48;
1174	>	private static final int  TC_SHIFT   = 32;
1175	>	private static final int  ST_SHIFT   = 31;
1176	>	private static final int  EC_SHIFT   = 16;
1177	>
1178	>	// bounds
1179	>	private static final int  SMASK      = 0xffff;  // short bits
1180	>	private static final int  MAX_CAP    = 0x7fff;  // max #workers - 1
1181	>	private static final int  EVENMASK   = 0xfffe;  // even short bits
1182	>	private static final int  SQMASK     = 0x007e;  // max 64 (even) slots
1183	>	private static final int  SHORT_SIGN = 1 << 15;
1184	>	private static final int  INT_SIGN   = 1 << 31;
1185	>
1186	>	// masks
1187	>	private static final long STOP_BIT   = 0x0001L << ST_SHIFT;
1188	>	private static final long AC_MASK    = ((long)SMASK) << AC_SHIFT;
1189	>	private static final long TC_MASK    = ((long)SMASK) << TC_SHIFT;
1190	>
1191	>	// units for incrementing and decrementing
1192	>	private static final long TC_UNIT    = 1L << TC_SHIFT;
1193	>	private static final long AC_UNIT    = 1L << AC_SHIFT;
1194	>
1195	>	// masks and units for dealing with u = (int)(ctl >>> 32)
1196	>	private static final int  UAC_SHIFT  = AC_SHIFT - 32;
1197	>	private static final int  UTC_SHIFT  = TC_SHIFT - 32;
1198	>	private static final int  UAC_MASK   = SMASK << UAC_SHIFT;
1199	>	private static final int  UTC_MASK   = SMASK << UTC_SHIFT;
1200	>	private static final int  UAC_UNIT   = 1 << UAC_SHIFT;
1201	>	private static final int  UTC_UNIT   = 1 << UTC_SHIFT;
1202	>
1203	>	// masks and units for dealing with e = (int)ctl
1204	>	private static final int E_MASK      = 0x7fffffff; // no STOP_BIT
1205	>	private static final int E_SEQ       = 1 << EC_SHIFT;
1206	>
1207	>	// plock bits
1208	>	private static final int SHUTDOWN    = 1 << 31;
1209	>	private static final int PL_LOCK     = 2;
1210	>	private static final int PL_SIGNAL   = 1;
1211	>	private static final int PL_SPINS    = 1 << 8;
1212	>
1213	>	// access mode for WorkQueue
1214	>	static final int LIFO_QUEUE          =  0;
1215	>	static final int FIFO_QUEUE          =  1;
1216	>	static final int SHARED_QUEUE        = -1;
1217	>
1218	>	// bounds for #steps in scan loop -- must be power 2 minus 1
1219	>	private static final int MIN_SCAN    = 0x1ff;   // cover estimation slop
1220	>	private static final int MAX_SCAN    = 0x1ffff; // 4 * max workers
1221
1222	<	/**
1223	<	* Pool number, just for assigning useful names to worker threads
1222	>	// Instance fields
1223	>
1224	>	/*
1225	>	* Field layout of this class tends to matter more than one would
1226	>	* like. Runtime layout order is only loosely related to
1227	>	* declaration order and may differ across JVMs, but the following
1228	>	* empirically works OK on current JVMs.
1229	>	*/
1230	>
1231	>	// Heuristic padding to ameliorate unfortunate memory placements
1232	>	volatile long pad00, pad01, pad02, pad03, pad04, pad05, pad06;
1233	>
1234	>	volatile long stealCount;                  // collects worker counts
1235	>	volatile long ctl;                         // main pool control
1236	>	volatile int plock;                        // shutdown status and seqLock
1237	>	volatile int indexSeed;                    // worker/submitter index seed
1238	>	final int config;                          // mode and parallelism level
1239	>	WorkQueue[] workQueues;                    // main registry
1240	>	final ForkJoinWorkerThreadFactory factory;
1241	>	final Thread.UncaughtExceptionHandler ueh; // per-worker UEH
1242	>	final String workerNamePrefix;             // to create worker name string
1243	>
1244	>	volatile Object pad10, pad11, pad12, pad13, pad14, pad15, pad16, pad17;
1245	>	volatile Object pad18, pad19, pad1a, pad1b;
1246	>
1247	>	/**
1248	>	* Acquires the plock lock to protect worker array and related
1249	>	* updates. This method is called only if an initial CAS on plock
1250	>	* fails. This acts as a spinlock for normal cases, but falls back
1251	>	* to builtin monitor to block when (rarely) needed. This would be
1252	>	* a terrible idea for a highly contended lock, but works fine as
1253	>	* a more conservative alternative to a pure spinlock.
1254		*/
1255	<	private final int poolNumber;
1255	>	private int acquirePlock() {
1256	>	int spins = PL_SPINS, r = 0, ps, nps;
1257	>	for (;;) {
1258	>	if (((ps = plock) & PL_LOCK) == 0 &&
1259	>	U.compareAndSwapInt(this, PLOCK, ps, nps = ps + PL_LOCK))
1260	>	return nps;
1261	>	else if (r == 0) { // randomize spins if possible
1262	>	Thread t = Thread.currentThread(); WorkQueue w; Submitter z;
1263	>	if ((t instanceof ForkJoinWorkerThread) &&
1264	>	(w = ((ForkJoinWorkerThread)t).workQueue) != null)
1265	>	r = w.seed;
1266	>	else if ((z = submitters.get()) != null)
1267	>	r = z.seed;
1268	>	else
1269	>	r = 1;
1270	>	}
1271	>	else if (spins >= 0) {
1272	>	r ^= r << 1; r ^= r >>> 3; r ^= r << 10; // xorshift
1273	>	if (r >= 0)
1274	>	--spins;
1275	>	}
1276	>	else if (U.compareAndSwapInt(this, PLOCK, ps, ps | PL_SIGNAL)) {
1277	>	synchronized (this) {
1278	>	if ((plock & PL_SIGNAL) != 0) {
1279	>	try {
1280	>	wait();
1281	>	} catch (InterruptedException ie) {
1282	>	try {
1283	>	Thread.currentThread().interrupt();
1284	>	} catch (SecurityException ignore) {
1285	>	}
1286	>	}
1287	>	}
1288	>	else
1289	>	notifyAll();
1290	>	}
1291	>	}
1292	>	}
1293	>	}
1294
1295		/**
1296	<	* The maximum allowed pool size
1296	>	* Unlocks and signals any thread waiting for plock. Called only
1297	>	* when CAS of seq value for unlock fails.
1298		*/
1299	<	private volatile int maxPoolSize;
1299	>	private void releasePlock(int ps) {
1300	>	plock = ps;
1301	>	synchronized (this) { notifyAll(); }
1302	>	}
1303
1304		/**
1305	<	* The desired parallelism level, updated only under workerLock.
1306	<	*/
1307	<	private volatile int parallelism;
1305	>	* Tries to create and start one worker if fewer than target
1306	>	* parallelism level exist. Adjusts counts etc on failure.
1307	>	*/
1308	>	private void tryAddWorker() {
1309	>	long c; int u;
1310	>	while ((u = (int)((c = ctl) >>> 32)) < 0 &&
1311	>	(u & SHORT_SIGN) != 0 && (int)c == 0) {
1312	>	long nc = (long)(((u + UTC_UNIT) & UTC_MASK) |
1313	>	((u + UAC_UNIT) & UAC_MASK)) << 32;
1314	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1315	>	ForkJoinWorkerThreadFactory fac;
1316	>	Throwable ex = null;
1317	>	ForkJoinWorkerThread wt = null;
1318	>	try {
1319	>	if ((fac = factory) != null &&
1320	>	(wt = fac.newThread(this)) != null) {
1321	>	wt.start();
1322	>	break;
1323	>	}
1324	>	} catch (Throwable e) {
1325	>	ex = e;
1326	>	}
1327	>	deregisterWorker(wt, ex);
1328	>	break;
1329	>	}
1330	>	}
1331	>	}
1332	>
1333	>	//  Registering and deregistering workers
1334
1335		/**
1336	<	* True if use local fifo, not default lifo, for local polling
1337	<	*/
1338	<	private volatile boolean locallyFifo;
1336	>	* Callback from ForkJoinWorkerThread to establish and record its
1337	>	* WorkQueue. To avoid scanning bias due to packing entries in
1338	>	* front of the workQueues array, we treat the array as a simple
1339	>	* power-of-two hash table using per-thread seed as hash,
1340	>	* expanding as needed.
1341	>	*
1342	>	* @param wt the worker thread
1343	>	* @return the worker's queue
1344	>	*/
1345	>	final WorkQueue registerWorker(ForkJoinWorkerThread wt) {
1346	>	Thread.UncaughtExceptionHandler handler; WorkQueue[] ws; int s, ps;
1347	>	wt.setDaemon(true);
1348	>	if ((handler = ueh) != null)
1349	>	wt.setUncaughtExceptionHandler(handler);
1350	>	do {} while (!U.compareAndSwapInt(this, INDEXSEED, s = indexSeed,
1351	>	s += SEED_INCREMENT) ||
1352	>	s == 0); // skip 0
1353	>	WorkQueue w = new WorkQueue(this, wt, config >>> 16, s);
1354	>	if (((ps = plock) & PL_LOCK) != 0 ||
1355	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
1356	>	ps = acquirePlock();
1357	>	int nps = (ps & SHUTDOWN) | ((ps + PL_LOCK) & ~SHUTDOWN);
1358	>	try {
1359	>	if ((ws = workQueues) != null) {    // skip if shutting down
1360	>	int n = ws.length, m = n - 1;
1361	>	int r = (s << 1) | 1;           // use odd-numbered indices
1362	>	if (ws[r &= m] != null) {       // collision
1363	>	int probes = 0;             // step by approx half size
1364	>	int step = (n <= 4) ? 2 : ((n >>> 1) & EVENMASK) + 2;
1365	>	while (ws[r = (r + step) & m] != null) {
1366	>	if (++probes >= n) {
1367	>	workQueues = ws = Arrays.copyOf(ws, n <<= 1);
1368	>	m = n - 1;
1369	>	probes = 0;
1370	>	}
1371	>	}
1372	>	}
1373	>	w.eventCount = w.poolIndex = r; // volatile write orders
1374	>	ws[r] = w;
1375	>	}
1376	>	} finally {
1377	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
1378	>	releasePlock(nps);
1379	>	}
1380	>	wt.setName(workerNamePrefix.concat(Integer.toString(w.poolIndex)));
1381	>	return w;
1382	>	}
1383
1384		/**
1385	<	* Holds number of total (i.e., created and not yet terminated)
1386	<	* and running (i.e., not blocked on joins or other managed sync)
1387	<	* threads, packed into one int to ensure consistent snapshot when
1388	<	* making decisions about creating and suspending spare
1389	<	* threads. Updated only by CAS.  Note: CASes in
1390	<	* updateRunningCount and preJoin running active count is in low
1391	<	* word, so need to be modified if this changes
1392	<	*/
1393	<	private volatile int workerCounts;
1385	>	* Final callback from terminating worker, as well as upon failure
1386	>	* to construct or start a worker.  Removes record of worker from
1387	>	* array, and adjusts counts. If pool is shutting down, tries to
1388	>	* complete termination.
1389	>	*
1390	>	* @param wt the worker thread or null if construction failed
1391	>	* @param ex the exception causing failure, or null if none
1392	>	*/
1393	>	final void deregisterWorker(ForkJoinWorkerThread wt, Throwable ex) {
1394	>	WorkQueue w = null;
1395	>	if (wt != null && (w = wt.workQueue) != null) {
1396	>	int ps;
1397	>	w.qlock = -1;                // ensure set
1398	>	long ns = w.nsteals, sc;     // collect steal count
1399	>	do {} while (!U.compareAndSwapLong(this, STEALCOUNT,
1400	>	sc = stealCount, sc + ns));
1401	>	if (((ps = plock) & PL_LOCK) != 0 ||
1402	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
1403	>	ps = acquirePlock();
1404	>	int nps = (ps & SHUTDOWN) | ((ps + PL_LOCK) & ~SHUTDOWN);
1405	>	try {
1406	>	int idx = w.poolIndex;
1407	>	WorkQueue[] ws = workQueues;
1408	>	if (ws != null && idx >= 0 && idx < ws.length && ws[idx] == w)
1409	>	ws[idx] = null;
1410	>	} finally {
1411	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
1412	>	releasePlock(nps);
1413	>	}
1414	>	}
1415
1416	<	private static int totalCountOf(int s)           { return s >>> 16;  }
1417	<	private static int runningCountOf(int s)         { return s & shortMask; }
1418	<	private static int workerCountsFor(int t, int r) { return (t << 16) + r; }
1416	>	long c;                          // adjust ctl counts
1417	>	do {} while (!U.compareAndSwapLong
1418	>	(this, CTL, c = ctl, (((c - AC_UNIT) & AC_MASK) |
1419	>	((c - TC_UNIT) & TC_MASK) |
1420	>	(c & ~(AC_MASK|TC_MASK)))));
1421	>
1422	>	if (!tryTerminate(false, false) && w != null && w.array != null) {
1423	>	w.cancelAll();               // cancel remaining tasks
1424	>	WorkQueue[] ws; WorkQueue v; Thread p; int u, i, e;
1425	>	while ((u = (int)((c = ctl) >>> 32)) < 0 && (e = (int)c) >= 0) {
1426	>	if (e > 0) {             // activate or create replacement
1427	>	if ((ws = workQueues) == null ||
1428	>	(i = e & SMASK) >= ws.length ||
1429	>	(v = ws[i]) == null)
1430	>	break;
1431	>	long nc = (((long)(v.nextWait & E_MASK)) |
1432	>	((long)(u + UAC_UNIT) << 32));
1433	>	if (v.eventCount != (e | INT_SIGN))
1434	>	break;
1435	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1436	>	v.eventCount = (e + E_SEQ) & E_MASK;
1437	>	if ((p = v.parker) != null)
1438	>	U.unpark(p);
1439	>	break;
1440	>	}
1441	>	}
1442	>	else {
1443	>	if ((short)u < 0)
1444	>	tryAddWorker();
1445	>	break;
1446	>	}
1447	>	}
1448	>	}
1449	>	if (ex == null)                     // help clean refs on way out
1450	>	ForkJoinTask.helpExpungeStaleExceptions();
1451	>	else                                // rethrow
1452	>	ForkJoinTask.rethrow(ex);
1453	>	}
1454	>
1455	>	// Submissions
1456	>
1457	>	/**
1458	>	* Unless shutting down, adds the given task to a submission queue
1459	>	* at submitter's current queue index (modulo submission
1460	>	* range). Only the most common path is directly handled in this
1461	>	* method. All others are relayed to fullExternalPush.
1462	>	*
1463	>	* @param task the task. Caller must ensure non-null.
1464	>	*/
1465	>	final void externalPush(ForkJoinTask<?> task) {
1466	>	WorkQueue[] ws; WorkQueue q; Submitter z; int m; ForkJoinTask<?>[] a;
1467	>	if ((z = submitters.get()) != null && plock > 0 &&
1468	>	(ws = workQueues) != null && (m = (ws.length - 1)) >= 0 &&
1469	>	(q = ws[m & z.seed & SQMASK]) != null &&
1470	>	U.compareAndSwapInt(q, QLOCK, 0, 1)) { // lock
1471	>	int b = q.base, s = q.top, n, an;
1472	>	if ((a = q.array) != null && (an = a.length) > (n = s + 1 - b)) {
1473	>	int j = (((an - 1) & s) << ASHIFT) + ABASE;
1474	>	U.putOrderedObject(a, j, task);
1475	>	q.top = s + 1;                     // push on to deque
1476	>	q.qlock = 0;
1477	>	if (n <= 2)
1478	>	signalWork(q);
1479	>	return;
1480	>	}
1481	>	q.qlock = 0;
1482	>	}
1483	>	fullExternalPush(task);
1484	>	}
1485	>
1486	>	/**
1487	>	* Full version of externalPush. This method is called, among
1488	>	* other times, upon the first submission of the first task to the
1489	>	* pool, so must perform secondary initialization.  It also
1490	>	* detects first submission by an external thread by looking up
1491	>	* its ThreadLocal, and creates a new shared queue if the one at
1492	>	* index if empty or contended. The plock lock body must be
1493	>	* exception-free (so no try/finally) so we optimistically
1494	>	* allocate new queues outside the lock and throw them away if
1495	>	* (very rarely) not needed.
1496	>	*
1497	>	* Secondary initialization occurs when plock is zero, to create
1498	>	* workQueue array and set plock to a valid value.  This lock body
1499	>	* must also be exception-free. Because the plock seq value can
1500	>	* eventually wrap around zero, this method harmlessly fails to
1501	>	* reinitialize if workQueues exists, while still advancing plock.
1502	>	*/
1503	>	private void fullExternalPush(ForkJoinTask<?> task) {
1504	>	int r = 0; // random index seed
1505	>	for (Submitter z = submitters.get();;) {
1506	>	WorkQueue[] ws; WorkQueue q; int ps, m, k;
1507	>	if (z == null) {
1508	>	if (U.compareAndSwapInt(this, INDEXSEED, r = indexSeed,
1509	>	r += SEED_INCREMENT) && r != 0)
1510	>	submitters.set(z = new Submitter(r));
1511	>	}
1512	>	else if (r == 0) {                  // move to a different index
1513	>	r = z.seed;
1514	>	r ^= r << 13;                   // same xorshift as WorkQueues
1515	>	r ^= r >>> 17;
1516	>	z.seed = r ^ (r << 5);
1517	>	}
1518	>	else if ((ps = plock) < 0)
1519	>	throw new RejectedExecutionException();
1520	>	else if (ps == 0 || (ws = workQueues) == null ||
1521	>	(m = ws.length - 1) < 0) { // initialize workQueues
1522	>	int p = config & SMASK;         // find power of two table size
1523	>	int n = (p > 1) ? p - 1 : 1;    // ensure at least 2 slots
1524	>	n |= n >>> 1;
1525	>	n |= n >>> 2;
1526	>	n |= n >>> 4;
1527	>	n |= n >>> 8;
1528	>	n |= n >>> 16;
1529	>	n = (n + 1) << 1;
1530	>	WorkQueue[] nws = ((ws = workQueues) == null || ws.length == 0 ?
1531	>	new WorkQueue[n] : null);
1532	>	if (((ps = plock) & PL_LOCK) != 0 ||
1533	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
1534	>	ps = acquirePlock();
1535	>	if (((ws = workQueues) == null || ws.length == 0) && nws != null)
1536	>	workQueues = nws;
1537	>	int nps = (ps & SHUTDOWN) | ((ps + PL_LOCK) & ~SHUTDOWN);
1538	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
1539	>	releasePlock(nps);
1540	>	}
1541	>	else if ((q = ws[k = r & m & SQMASK]) != null) {
1542	>	if (q.qlock == 0 && U.compareAndSwapInt(q, QLOCK, 0, 1)) {
1543	>	ForkJoinTask<?>[] a = q.array;
1544	>	int s = q.top;
1545	>	boolean submitted = false;
1546	>	try {                      // locked version of push
1547	>	if ((a != null && a.length > s + 1 - q.base) ||
1548	>	(a = q.growArray()) != null) {   // must presize
1549	>	int j = (((a.length - 1) & s) << ASHIFT) + ABASE;
1550	>	U.putOrderedObject(a, j, task);
1551	>	q.top = s + 1;
1552	>	submitted = true;
1553	>	}
1554	>	} finally {
1555	>	q.qlock = 0;  // unlock
1556	>	}
1557	>	if (submitted) {
1558	>	signalWork(q);
1559	>	return;
1560	>	}
1561	>	}
1562	>	r = 0; // move on failure
1563	>	}
1564	>	else if (((ps = plock) & PL_LOCK) == 0) { // create new queue
1565	>	q = new WorkQueue(this, null, SHARED_QUEUE, r);
1566	>	if (((ps = plock) & PL_LOCK) != 0 ||
1567	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
1568	>	ps = acquirePlock();
1569	>	if ((ws = workQueues) != null && k < ws.length && ws[k] == null)
1570	>	ws[k] = q;
1571	>	int nps = (ps & SHUTDOWN) | ((ps + PL_LOCK) & ~SHUTDOWN);
1572	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
1573	>	releasePlock(nps);
1574	>	}
1575	>	else
1576	>	r = 0; // try elsewhere while lock held
1577	>	}
1578	>	}
1579	>
1580	>	// Maintaining ctl counts
1581
1582		/**
1583	<	* Add delta (which may be negative) to running count.  This must
230	<	* be called before (with negative arg) and after (with positive)
231	<	* any managed synchronization (i.e., mainly, joins)
232	<	* @param delta the number to add
1583	>	* Increments active count; mainly called upon return from blocking.
1584		*/
1585	<	final void updateRunningCount(int delta) {
1586	<	int s;
1587	<	do;while (!casWorkerCounts(s = workerCounts, s + delta));
1585	>	final void incrementActiveCount() {
1586	>	long c;
1587	>	do {} while (!U.compareAndSwapLong(this, CTL, c = ctl, c + AC_UNIT));
1588		}
1589
1590		/**
1591	<	* Add delta (which may be negative) to both total and running
1592	<	* count.  This must be called upon creation and termination of
1593	<	* worker threads.
243	<	* @param delta the number to add
1591	>	* Tries to create or activate a worker if too few are active.
1592	>	*
1593	>	* @param q the (non-null) queue holding tasks to be signalled
1594		*/
1595	<	private void updateWorkerCount(int delta) {
1596	<	int d = delta + (delta << 16); // add to both lo and hi parts
1597	<	int s;
1598	<	do;while (!casWorkerCounts(s = workerCounts, s + d));
1595	>	final void signalWork(WorkQueue q) {
1596	>	int hint = q.poolIndex;
1597	>	long c; int e, u, i, n; WorkQueue[] ws; WorkQueue w; Thread p;
1598	>	while ((u = (int)((c = ctl) >>> 32)) < 0) {
1599	>	if ((e = (int)c) > 0) {
1600	>	if ((ws = workQueues) != null && ws.length > (i = e & SMASK) &&
1601	>	(w = ws[i]) != null && w.eventCount == (e | INT_SIGN)) {
1602	>	long nc = (((long)(w.nextWait & E_MASK)) |
1603	>	((long)(u + UAC_UNIT) << 32));
1604	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1605	>	w.hint = hint;
1606	>	w.eventCount = (e + E_SEQ) & E_MASK;
1607	>	if ((p = w.parker) != null)
1608	>	U.unpark(p);
1609	>	break;
1610	>	}
1611	>	if (q.top - q.base <= 0)
1612	>	break;
1613	>	}
1614	>	else
1615	>	break;
1616	>	}
1617	>	else {
1618	>	if ((short)u < 0)
1619	>	tryAddWorker();
1620	>	break;
1621	>	}
1622	>	}
1623		}
1624
1625	+	// Scanning for tasks
1626	+
1627		/**
1628	<	* Lifecycle control. High word contains runState, low word
253	<	* contains the number of workers that are (probably) executing
254	<	* tasks. This value is atomically incremented before a worker
255	<	* gets a task to run, and decremented when worker has no tasks
256	<	* and cannot find any. These two fields are bundled together to
257	<	* support correct termination triggering.  Note: activeCount
258	<	* CAS'es cheat by assuming active count is in low word, so need
259	<	* to be modified if this changes
1628	>	* Top-level runloop for workers, called by ForkJoinWorkerThread.run.
1629		*/
1630	<	private volatile int runControl;
1630	>	final void runWorker(WorkQueue w) {
1631	>	w.growArray(); // allocate queue
1632	>	do { w.runTask(scan(w)); } while (w.qlock >= 0);
1633	>	}
1634	>
1635	>	/**
1636	>	* Scans for and, if found, returns one task, else possibly
1637	>	* inactivates the worker. This method operates on single reads of
1638	>	* volatile state and is designed to be re-invoked continuously,
1639	>	* in part because it returns upon detecting inconsistencies,
1640	>	* contention, or state changes that indicate possible success on
1641	>	* re-invocation.
1642	>	*
1643	>	* The scan searches for tasks across queues (starting at a random
1644	>	* index, and relying on registerWorker to irregularly scatter
1645	>	* them within array to avoid bias), checking each at least twice.
1646	>	* The scan terminates upon either finding a non-empty queue, or
1647	>	* completing the sweep. If the worker is not inactivated, it
1648	>	* takes and returns a task from this queue. Otherwise, if not
1649	>	* activated, it signals workers (that may include itself) and
1650	>	* returns so caller can retry. Also returns for true if the
1651	>	* worker array may have changed during an empty scan.  On failure
1652	>	* to find a task, we take one of the following actions, after
1653	>	* which the caller will retry calling this method unless
1654	>	* terminated.
1655	>	*
1656	>	* * If pool is terminating, terminate the worker.
1657	>	*
1658	>	* * If not already enqueued, try to inactivate and enqueue the
1659	>	* worker on wait queue. Or, if inactivating has caused the pool
1660	>	* to be quiescent, relay to idleAwaitWork to possibly shrink
1661	>	* pool.
1662	>	*
1663	>	* * If already enqueued and none of the above apply, possibly
1664	>	* park awaiting signal, else lingering to help scan and signal.
1665	>	*
1666	>	* * If a non-empty queue discovered or left as a hint,
1667	>	* help wake up other workers before return.
1668	>	*
1669	>	* @param w the worker (via its WorkQueue)
1670	>	* @return a task or null if none found
1671	>	*/
1672	>	private final ForkJoinTask<?> scan(WorkQueue w) {
1673	>	WorkQueue[] ws; int m;
1674	>	int ps = plock;                          // read plock before ws
1675	>	if (w != null && (ws = workQueues) != null && (m = ws.length - 1) >= 0) {
1676	>	int ec = w.eventCount;               // ec is negative if inactive
1677	>	int r = w.seed; r ^= r << 13; r ^= r >>> 17; w.seed = r ^= r << 5;
1678	>	w.hint = -1;                         // update seed and clear hint
1679	>	int j = ((m + m + 1) | MIN_SCAN) & MAX_SCAN;
1680	>	do {
1681	>	WorkQueue q; ForkJoinTask<?>[] a; int b;
1682	>	if ((q = ws[(r + j) & m]) != null && (b = q.base) - q.top < 0 &&
1683	>	(a = q.array) != null) {     // probably nonempty
1684	>	int i = (((a.length - 1) & b) << ASHIFT) + ABASE;
1685	>	ForkJoinTask<?> t = (ForkJoinTask<?>)
1686	>	U.getObjectVolatile(a, i);
1687	>	if (q.base == b && ec >= 0 && t != null &&
1688	>	U.compareAndSwapObject(a, i, t, null)) {
1689	>	if ((q.base = b + 1) - q.top < 0)
1690	>	signalWork(q);
1691	>	return t;                // taken
1692	>	}
1693	>	else if ((ec < 0 || j < m) && (int)(ctl >> AC_SHIFT) <= 0) {
1694	>	w.hint = (r + j) & m;    // help signal below
1695	>	break;                   // cannot take
1696	>	}
1697	>	}
1698	>	} while (--j >= 0);
1699
1700	<	// RunState values. Order among values matters
1701	<	private static final int RUNNING     = 0;
1702	<	private static final int SHUTDOWN    = 1;
1703	<	private static final int TERMINATING = 2;
1704	<	private static final int TERMINATED  = 3;
1700	>	int h, e, ns; long c, sc; WorkQueue q;
1701	>	if ((ns = w.nsteals) != 0) {
1702	>	if (U.compareAndSwapLong(this, STEALCOUNT,
1703	>	sc = stealCount, sc + ns))
1704	>	w.nsteals = 0;               // collect steals and rescan
1705	>	}
1706	>	else if (plock != ps)                // consistency check
1707	>	;                                // skip
1708	>	else if ((e = (int)(c = ctl)) < 0)
1709	>	w.qlock = -1;                    // pool is terminating
1710	>	else {
1711	>	if ((h = w.hint) < 0) {
1712	>	if (ec >= 0) {               // try to enqueue/inactivate
1713	>	long nc = (((long)ec |
1714	>	((c - AC_UNIT) & (AC_MASK|TC_MASK))));
1715	>	w.nextWait = e;          // link and mark inactive
1716	>	w.eventCount = ec | INT_SIGN;
1717	>	if (ctl != c || !U.compareAndSwapLong(this, CTL, c, nc))
1718	>	w.eventCount = ec;   // unmark on CAS failure
1719	>	else if ((int)(c >> AC_SHIFT) == 1 - (config & SMASK))
1720	>	idleAwaitWork(w, nc, c);
1721	>	}
1722	>	else if (w.eventCount < 0 && ctl == c) {
1723	>	Thread wt = Thread.currentThread();
1724	>	Thread.interrupted();    // clear status
1725	>	U.putObject(wt, PARKBLOCKER, this);
1726	>	w.parker = wt;           // emulate LockSupport.park
1727	>	if (w.eventCount < 0)    // recheck
1728	>	U.park(false, 0L);   // block
1729	>	w.parker = null;
1730	>	U.putObject(wt, PARKBLOCKER, null);
1731	>	}
1732	>	}
1733	>	if ((h >= 0 || (h = w.hint) >= 0) &&
1734	>	(ws = workQueues) != null && h < ws.length &&
1735	>	(q = ws[h]) != null) {      // signal others before retry
1736	>	WorkQueue v; Thread p; int u, i, s;
1737	>	for (int n = (config & SMASK) - 1;;) {
1738	>	int idleCount = (w.eventCount < 0) ? 0 : -1;
1739	>	if (((s = idleCount - q.base + q.top) <= n &&
1740	>	(n = s) <= 0) ||
1741	>	(u = (int)((c = ctl) >>> 32)) >= 0 ||
1742	>	(e = (int)c) <= 0 || m < (i = e & SMASK) ||
1743	>	(v = ws[i]) == null)
1744	>	break;
1745	>	long nc = (((long)(v.nextWait & E_MASK)) |
1746	>	((long)(u + UAC_UNIT) << 32));
1747	>	if (v.eventCount != (e | INT_SIGN) ||
1748	>	!U.compareAndSwapLong(this, CTL, c, nc))
1749	>	break;
1750	>	v.hint = h;
1751	>	v.eventCount = (e + E_SEQ) & E_MASK;
1752	>	if ((p = v.parker) != null)
1753	>	U.unpark(p);
1754	>	if (--n <= 0)
1755	>	break;
1756	>	}
1757	>	}
1758	>	}
1759	>	}
1760	>	return null;
1761	>	}
1762
1763	<	private static int runStateOf(int c)             { return c >>> 16; }
1764	<	private static int activeCountOf(int c)          { return c & shortMask; }
1765	<	private static int runControlFor(int r, int a)   { return (r << 16) + a; }
1763	>	/**
1764	>	* If inactivating worker w has caused the pool to become
1765	>	* quiescent, checks for pool termination, and, so long as this is
1766	>	* not the only worker, waits for event for up to a given
1767	>	* duration.  On timeout, if ctl has not changed, terminates the
1768	>	* worker, which will in turn wake up another worker to possibly
1769	>	* repeat this process.
1770	>	*
1771	>	* @param w the calling worker
1772	>	* @param currentCtl the ctl value triggering possible quiescence
1773	>	* @param prevCtl the ctl value to restore if thread is terminated
1774	>	*/
1775	>	private void idleAwaitWork(WorkQueue w, long currentCtl, long prevCtl) {
1776	>	if (w != null && w.eventCount < 0 &&
1777	>	!tryTerminate(false, false) && (int)prevCtl != 0 &&
1778	>	ctl == currentCtl) {
1779	>	int dc = -(short)(currentCtl >>> TC_SHIFT);
1780	>	long parkTime = dc < 0 ? FAST_IDLE_TIMEOUT: (dc + 1) * IDLE_TIMEOUT;
1781	>	long deadline = System.nanoTime() + parkTime - TIMEOUT_SLOP;
1782	>	Thread wt = Thread.currentThread();
1783	>	while (ctl == currentCtl) {
1784	>	Thread.interrupted();  // timed variant of version in scan()
1785	>	U.putObject(wt, PARKBLOCKER, this);
1786	>	w.parker = wt;
1787	>	if (ctl == currentCtl)
1788	>	U.park(false, parkTime);
1789	>	w.parker = null;
1790	>	U.putObject(wt, PARKBLOCKER, null);
1791	>	if (ctl != currentCtl)
1792	>	break;
1793	>	if (deadline - System.nanoTime() <= 0L &&
1794	>	U.compareAndSwapLong(this, CTL, currentCtl, prevCtl)) {
1795	>	w.eventCount = (w.eventCount + E_SEQ) | E_MASK;
1796	>	w.hint = -1;
1797	>	w.qlock = -1;   // shrink
1798	>	break;
1799	>	}
1800	>	}
1801	>	}
1802	>	}
1803
1804		/**
1805	<	* Try incrementing active count; fail on contention. Called by
1806	<	* workers before/during executing tasks.
1807	<	* @return true on success;
1808	<	*/
1809	<	final boolean tryIncrementActiveCount() {
1810	<	int c = runControl;
1811	<	return casRunControl(c, c+1);
1805	>	* Scans through queues looking for work while joining a task; if
1806	>	* any present, signals. May return early if more signalling is
1807	>	* detectably unneeded.
1808	>	*
1809	>	* @param task return early if done
1810	>	* @param origin an index to start scan
1811	>	*/
1812	>	private void helpSignal(ForkJoinTask<?> task, int origin) {
1813	>	WorkQueue[] ws; WorkQueue w; Thread p; long c; int m, u, e, i, s;
1814	>	if (task != null && task.status >= 0 &&
1815	>	(u = (int)(ctl >>> 32)) < 0 && (u >> UAC_SHIFT) < 0 &&
1816	>	(ws = workQueues) != null && (m = ws.length - 1) >= 0) {
1817	>	outer: for (int k = origin, j = m; j >= 0; --j) {
1818	>	WorkQueue q = ws[k++ & m];
1819	>	for (int n = m;;) { // limit to at most m signals
1820	>	if (task.status < 0)
1821	>	break outer;
1822	>	if (q == null ||
1823	>	((s = -q.base + q.top) <= n && (n = s) <= 0))
1824	>	break;
1825	>	if ((u = (int)((c = ctl) >>> 32)) >= 0 ||
1826	>	(e = (int)c) <= 0 || m < (i = e & SMASK) ||
1827	>	(w = ws[i]) == null)
1828	>	break outer;
1829	>	long nc = (((long)(w.nextWait & E_MASK)) |
1830	>	((long)(u + UAC_UNIT) << 32));
1831	>	if (w.eventCount != (e | INT_SIGN))
1832	>	break outer;
1833	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1834	>	w.eventCount = (e + E_SEQ) & E_MASK;
1835	>	if ((p = w.parker) != null)
1836	>	U.unpark(p);
1837	>	if (--n <= 0)
1838	>	break;
1839	>	}
1840	>	}
1841	>	}
1842	>	}
1843		}
1844
1845		/**
1846	<	* Try decrementing active count; fail on contention.
1847	<	* Possibly trigger termination on success
1848	<	* Called by workers when they can't find tasks.
1849	<	* @return true on success
1850	<	*/
1851	<	final boolean tryDecrementActiveCount() {
1852	<	int c = runControl;
1853	<	int nextc = c - 1;
1854	<	if (!casRunControl(c, nextc))
1855	<	return false;
1856	<	if (canTerminateOnShutdown(nextc))
1857	<	terminateOnShutdown();
1858	<	return true;
1846	>	* Tries to locate and execute tasks for a stealer of the given
1847	>	* task, or in turn one of its stealers, Traces currentSteal ->
1848	>	* currentJoin links looking for a thread working on a descendant
1849	>	* of the given task and with a non-empty queue to steal back and
1850	>	* execute tasks from. The first call to this method upon a
1851	>	* waiting join will often entail scanning/search, (which is OK
1852	>	* because the joiner has nothing better to do), but this method
1853	>	* leaves hints in workers to speed up subsequent calls. The
1854	>	* implementation is very branchy to cope with potential
1855	>	* inconsistencies or loops encountering chains that are stale,
1856	>	* unknown, or so long that they are likely cyclic.
1857	>	*
1858	>	* @param joiner the joining worker
1859	>	* @param task the task to join
1860	>	* @return 0 if no progress can be made, negative if task
1861	>	* known complete, else positive
1862	>	*/
1863	>	private int tryHelpStealer(WorkQueue joiner, ForkJoinTask<?> task) {
1864	>	int stat = 0, steps = 0;                    // bound to avoid cycles
1865	>	if (joiner != null && task != null) {       // hoist null checks
1866	>	restart: for (;;) {
1867	>	ForkJoinTask<?> subtask = task;     // current target
1868	>	for (WorkQueue j = joiner, v;;) {   // v is stealer of subtask
1869	>	WorkQueue[] ws; int m, s, h;
1870	>	if ((s = task.status) < 0) {
1871	>	stat = s;
1872	>	break restart;
1873	>	}
1874	>	if ((ws = workQueues) == null || (m = ws.length - 1) <= 0)
1875	>	break restart;              // shutting down
1876	>	if ((v = ws[h = (j.hint | 1) & m]) == null ||
1877	>	v.currentSteal != subtask) {
1878	>	for (int origin = h;;) {    // find stealer
1879	>	if (((h = (h + 2) & m) & 15) == 1 &&
1880	>	(subtask.status < 0 || j.currentJoin != subtask))
1881	>	continue restart;   // occasional staleness check
1882	>	if ((v = ws[h]) != null &&
1883	>	v.currentSteal == subtask) {
1884	>	j.hint = h;        // save hint
1885	>	break;
1886	>	}
1887	>	if (h == origin)
1888	>	break restart;      // cannot find stealer
1889	>	}
1890	>	}
1891	>	for (;;) { // help stealer or descend to its stealer
1892	>	ForkJoinTask[] a;  int b;
1893	>	if (subtask.status < 0)     // surround probes with
1894	>	continue restart;       //   consistency checks
1895	>	if ((b = v.base) - v.top < 0 && (a = v.array) != null) {
1896	>	int i = (((a.length - 1) & b) << ASHIFT) + ABASE;
1897	>	ForkJoinTask<?> t =
1898	>	(ForkJoinTask<?>)U.getObjectVolatile(a, i);
1899	>	if (subtask.status < 0 || j.currentJoin != subtask ||
1900	>	v.currentSteal != subtask)
1901	>	continue restart;   // stale
1902	>	stat = 1;               // apparent progress
1903	>	if (t != null && v.base == b &&
1904	>	U.compareAndSwapObject(a, i, t, null)) {
1905	>	v.base = b + 1;     // help stealer
1906	>	joiner.runSubtask(t);
1907	>	}
1908	>	else if (v.base == b && ++steps == MAX_HELP)
1909	>	break restart;      // v apparently stalled
1910	>	}
1911	>	else {                      // empty -- try to descend
1912	>	ForkJoinTask<?> next = v.currentJoin;
1913	>	if (subtask.status < 0 || j.currentJoin != subtask ||
1914	>	v.currentSteal != subtask)
1915	>	continue restart;   // stale
1916	>	else if (next == null || ++steps == MAX_HELP)
1917	>	break restart;      // dead-end or maybe cyclic
1918	>	else {
1919	>	subtask = next;
1920	>	j = v;
1921	>	break;
1922	>	}
1923	>	}
1924	>	}
1925	>	}
1926	>	}
1927	>	}
1928	>	return stat;
1929		}
1930
1931		/**
1932	<	* Return true if argument represents zero active count and
1933	<	* nonzero runstate, which is the triggering condition for
1934	<	* terminating on shutdown.
1935	<	*/
1936	<	private static boolean canTerminateOnShutdown(int c) {
1937	<	return ((c & -c) >>> 16) != 0; // i.e. least bit is nonzero runState bit
1932	>	* Analog of tryHelpStealer for CountedCompleters. Tries to steal
1933	>	* and run tasks within the target's computation.
1934	>	*
1935	>	* @param task the task to join
1936	>	* @param mode if shared, exit upon completing any task
1937	>	* if all workers are active
1938	>	*/
1939	>	private int helpComplete(ForkJoinTask<?> task, int mode) {
1940	>	WorkQueue[] ws; WorkQueue q; int m, n, s, u;
1941	>	if (task != null && (ws = workQueues) != null &&
1942	>	(m = ws.length - 1) >= 0) {
1943	>	for (int j = 1, origin = j;;) {
1944	>	if ((s = task.status) < 0)
1945	>	return s;
1946	>	if ((q = ws[j & m]) != null && q.pollAndExecCC(task)) {
1947	>	origin = j;
1948	>	if (mode == SHARED_QUEUE &&
1949	>	((u = (int)(ctl >>> 32)) >= 0 || (u >> UAC_SHIFT) >= 0))
1950	>	break;
1951	>	}
1952	>	else if ((j = (j + 2) & m) == origin)
1953	>	break;
1954	>	}
1955	>	}
1956	>	return 0;
1957		}
1958
1959		/**
1960	<	* Transition run state to at least the given state. Return true
1961	<	* if not already at least given state.
1962	<	*/
1963	<	private boolean transitionRunStateTo(int state) {
1964	<	for (;;) {
1965	<	int c = runControl;
1966	<	if (runStateOf(c) >= state)
1967	<	return false;
1968	<	if (casRunControl(c, runControlFor(state, activeCountOf(c))))
1969	<	return true;
1960	>	* Tries to decrement active count (sometimes implicitly) and
1961	>	* possibly release or create a compensating worker in preparation
1962	>	* for blocking. Fails on contention or termination. Otherwise,
1963	>	* adds a new thread if no idle workers are available and pool
1964	>	* may become starved.
1965	>	*/
1966	>	final boolean tryCompensate() {
1967	>	int pc = config & SMASK, e, i, tc; long c;
1968	>	WorkQueue[] ws; WorkQueue w; Thread p;
1969	>	if ((ws = workQueues) != null && (e = (int)(c = ctl)) >= 0) {
1970	>	if (e != 0 && (i = e & SMASK) < ws.length &&
1971	>	(w = ws[i]) != null && w.eventCount == (e | INT_SIGN)) {
1972	>	long nc = ((long)(w.nextWait & E_MASK) |
1973	>	(c & (AC_MASK|TC_MASK)));
1974	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1975	>	w.eventCount = (e + E_SEQ) & E_MASK;
1976	>	if ((p = w.parker) != null)
1977	>	U.unpark(p);
1978	>	return true;   // replace with idle worker
1979	>	}
1980	>	}
1981	>	else if ((tc = (short)(c >>> TC_SHIFT)) >= 0 &&
1982	>	(int)(c >> AC_SHIFT) + pc > 1) {
1983	>	long nc = ((c - AC_UNIT) & AC_MASK) | (c & ~AC_MASK);
1984	>	if (U.compareAndSwapLong(this, CTL, c, nc))
1985	>	return true;   // no compensation
1986	>	}
1987	>	else if (tc + pc < MAX_CAP) {
1988	>	long nc = ((c + TC_UNIT) & TC_MASK) | (c & ~TC_MASK);
1989	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1990	>	ForkJoinWorkerThreadFactory fac;
1991	>	Throwable ex = null;
1992	>	ForkJoinWorkerThread wt = null;
1993	>	try {
1994	>	if ((fac = factory) != null &&
1995	>	(wt = fac.newThread(this)) != null) {
1996	>	wt.start();
1997	>	return true;
1998	>	}
1999	>	} catch (Throwable rex) {
2000	>	ex = rex;
2001	>	}
2002	>	deregisterWorker(wt, ex); // clean up and return false
2003	>	}
2004	>	}
2005		}
2006	+	return false;
2007		}
2008
2009		/**
2010	<	* Controls whether to add spares to maintain parallelism
2010	>	* Helps and/or blocks until the given task is done.
2011	>	*
2012	>	* @param joiner the joining worker
2013	>	* @param task the task
2014	>	* @return task status on exit
2015		*/
2016	<	private volatile boolean maintainsParallelism;
2017	<
2018	<	// Constructors
2016	>	final int awaitJoin(WorkQueue joiner, ForkJoinTask<?> task) {
2017	>	int s = 0;
2018	>	if (joiner != null && task != null && (s = task.status) >= 0) {
2019	>	ForkJoinTask<?> prevJoin = joiner.currentJoin;
2020	>	joiner.currentJoin = task;
2021	>	do {} while ((s = task.status) >= 0 && !joiner.isEmpty() &&
2022	>	joiner.tryRemoveAndExec(task)); // process local tasks
2023	>	if (s >= 0 && (s = task.status) >= 0) {
2024	>	helpSignal(task, joiner.poolIndex);
2025	>	if ((s = task.status) >= 0 &&
2026	>	(task instanceof CountedCompleter))
2027	>	s = helpComplete(task, LIFO_QUEUE);
2028	>	}
2029	>	while (s >= 0 && (s = task.status) >= 0) {
2030	>	if ((!joiner.isEmpty() ||           // try helping
2031	>	(s = tryHelpStealer(joiner, task)) == 0) &&
2032	>	(s = task.status) >= 0) {
2033	>	helpSignal(task, joiner.poolIndex);
2034	>	if ((s = task.status) >= 0 && tryCompensate()) {
2035	>	if (task.trySetSignal() && (s = task.status) >= 0) {
2036	>	synchronized (task) {
2037	>	if (task.status >= 0) {
2038	>	try {                // see ForkJoinTask
2039	>	task.wait();     //  for explanation
2040	>	} catch (InterruptedException ie) {
2041	>	}
2042	>	}
2043	>	else
2044	>	task.notifyAll();
2045	>	}
2046	>	}
2047	>	long c;                          // re-activate
2048	>	do {} while (!U.compareAndSwapLong
2049	>	(this, CTL, c = ctl, c + AC_UNIT));
2050	>	}
2051	>	}
2052	>	}
2053	>	joiner.currentJoin = prevJoin;
2054	>	}
2055	>	return s;
2056	>	}
2057
2058		/**
2059	<	* Creates a ForkJoinPool with a pool size equal to the number of
2060	<	* processors available on the system and using the default
2061	<	* ForkJoinWorkerThreadFactory,
2062	<	* @throws SecurityException if a security manager exists and
2063	<	*         the caller is not permitted to modify threads
2064	<	*         because it does not hold {@link
336	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
2059	>	* Stripped-down variant of awaitJoin used by timed joins. Tries
2060	>	* to help join only while there is continuous progress. (Caller
2061	>	* will then enter a timed wait.)
2062	>	*
2063	>	* @param joiner the joining worker
2064	>	* @param task the task
2065		*/
2066	<	public ForkJoinPool() {
2067	<	this(Runtime.getRuntime().availableProcessors(),
2068	<	defaultForkJoinWorkerThreadFactory);
2066	>	final void helpJoinOnce(WorkQueue joiner, ForkJoinTask<?> task) {
2067	>	int s;
2068	>	if (joiner != null && task != null && (s = task.status) >= 0) {
2069	>	ForkJoinTask<?> prevJoin = joiner.currentJoin;
2070	>	joiner.currentJoin = task;
2071	>	do {} while ((s = task.status) >= 0 && !joiner.isEmpty() &&
2072	>	joiner.tryRemoveAndExec(task));
2073	>	if (s >= 0 && (s = task.status) >= 0) {
2074	>	helpSignal(task, joiner.poolIndex);
2075	>	if ((s = task.status) >= 0 &&
2076	>	(task instanceof CountedCompleter))
2077	>	s = helpComplete(task, LIFO_QUEUE);
2078	>	}
2079	>	if (s >= 0 && joiner.isEmpty()) {
2080	>	do {} while (task.status >= 0 &&
2081	>	tryHelpStealer(joiner, task) > 0);
2082	>	}
2083	>	joiner.currentJoin = prevJoin;
2084	>	}
2085		}
2086
2087		/**
2088	<	* Creates a ForkJoinPool with the indicated parellelism level
2089	<	* threads, and using the default ForkJoinWorkerThreadFactory,
2090	<	* @param parallelism the number of worker threads
2091	<	* @throws IllegalArgumentException if parallelism less than or
348	<	* equal to zero
349	<	* @throws SecurityException if a security manager exists and
350	<	*         the caller is not permitted to modify threads
351	<	*         because it does not hold {@link
352	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
2088	>	* Returns a (probably) non-empty steal queue, if one is found
2089	>	* during a scan, else null.  This method must be retried by
2090	>	* caller if, by the time it tries to use the queue, it is empty.
2091	>	* @param r a (random) seed for scanning
2092		*/
2093	<	public ForkJoinPool(int parallelism) {
2094	<	this(parallelism, defaultForkJoinWorkerThreadFactory);
2093	>	private WorkQueue findNonEmptyStealQueue(int r) {
2094	>	for (;;) {
2095	>	int ps = plock, m; WorkQueue[] ws; WorkQueue q;
2096	>	if ((ws = workQueues) != null && (m = ws.length - 1) >= 0) {
2097	>	for (int j = (m + 1) << 2; j >= 0; --j) {
2098	>	if ((q = ws[(((r + j) << 1) | 1) & m]) != null &&
2099	>	q.base - q.top < 0)
2100	>	return q;
2101	>	}
2102	>	}
2103	>	if (plock == ps)
2104	>	return null;
2105	>	}
2106		}
2107
2108		/**
2109	<	* Creates a ForkJoinPool with parallelism equal to the number of
2110	<	* processors available on the system and using the given
2111	<	* ForkJoinWorkerThreadFactory,
2112	<	* @param factory the factory for creating new threads
2113	<	* @throws NullPointerException if factory is null
2114	<	* @throws SecurityException if a security manager exists and
2115	<	*         the caller is not permitted to modify threads
2116	<	*         because it does not hold {@link
2117	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
2118	<	*/
2119	<	public ForkJoinPool(ForkJoinWorkerThreadFactory factory) {
2120	<	this(Runtime.getRuntime().availableProcessors(), factory);
2109	>	* Runs tasks until {@code isQuiescent()}. We piggyback on
2110	>	* active count ctl maintenance, but rather than blocking
2111	>	* when tasks cannot be found, we rescan until all others cannot
2112	>	* find tasks either.
2113	>	*/
2114	>	final void helpQuiescePool(WorkQueue w) {
2115	>	for (boolean active = true;;) {
2116	>	long c; WorkQueue q; ForkJoinTask<?> t; int b;
2117	>	while ((t = w.nextLocalTask()) != null) {
2118	>	if (w.base - w.top < 0)
2119	>	signalWork(w);
2120	>	t.doExec();
2121	>	}
2122	>	if ((q = findNonEmptyStealQueue(w.nextSeed())) != null) {
2123	>	if (!active) {      // re-establish active count
2124	>	active = true;
2125	>	do {} while (!U.compareAndSwapLong
2126	>	(this, CTL, c = ctl, c + AC_UNIT));
2127	>	}
2128	>	if ((b = q.base) - q.top < 0 && (t = q.pollAt(b)) != null) {
2129	>	if (q.base - q.top < 0)
2130	>	signalWork(q);
2131	>	w.runSubtask(t);
2132	>	}
2133	>	}
2134	>	else if (active) {      // decrement active count without queuing
2135	>	long nc = (c = ctl) - AC_UNIT;
2136	>	if ((int)(nc >> AC_SHIFT) + (config & SMASK) == 0)
2137	>	return;         // bypass decrement-then-increment
2138	>	if (U.compareAndSwapLong(this, CTL, c, nc))
2139	>	active = false;
2140	>	}
2141	>	else if ((int)((c = ctl) >> AC_SHIFT) + (config & SMASK) == 0 &&
2142	>	U.compareAndSwapLong(this, CTL, c, c + AC_UNIT))
2143	>	return;
2144	>	}
2145		}
2146
2147		/**
2148	<	* Creates a ForkJoinPool with the given parallelism and factory.
2148	>	* Gets and removes a local or stolen task for the given worker.
2149		*
2150	<	* @param parallelism the targeted number of worker threads
377	<	* @param factory the factory for creating new threads
378	<	* @throws IllegalArgumentException if parallelism less than or
379	<	* equal to zero, or greater than implementation limit.
380	<	* @throws NullPointerException if factory is null
381	<	* @throws SecurityException if a security manager exists and
382	<	*         the caller is not permitted to modify threads
383	<	*         because it does not hold {@link
384	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
2150	>	* @return a task, if available
2151		*/
2152	<	public ForkJoinPool(int parallelism, ForkJoinWorkerThreadFactory factory) {
2153	<	if (parallelism <= 0 || parallelism > MAX_THREADS)
2154	<	throw new IllegalArgumentException();
2155	<	if (factory == null)
2156	<	throw new NullPointerException();
2157	<	checkPermission();
2158	<	this.factory = factory;
2159	<	this.parallelism = parallelism;
2160	<	this.maxPoolSize = MAX_THREADS;
2161	<	this.maintainsParallelism = true;
2162	<	this.poolNumber = poolNumberGenerator.incrementAndGet();
2163	<	this.workerLock = new ReentrantLock();
2164	<	this.termination = workerLock.newCondition();
399	<	this.stealCount = new AtomicLong();
400	<	this.submissionQueue = new LinkedTransferQueue<ForkJoinTask<?>>();
401	<	// worker array and workers are lazily constructed
2152	>	final ForkJoinTask<?> nextTaskFor(WorkQueue w) {
2153	>	for (ForkJoinTask<?> t;;) {
2154	>	WorkQueue q; int b;
2155	>	if ((t = w.nextLocalTask()) != null)
2156	>	return t;
2157	>	if ((q = findNonEmptyStealQueue(w.nextSeed())) == null)
2158	>	return null;
2159	>	if ((b = q.base) - q.top < 0 && (t = q.pollAt(b)) != null) {
2160	>	if (q.base - q.top < 0)
2161	>	signalWork(q);
2162	>	return t;
2163	>	}
2164	>	}
2165		}
2166
2167		/**
2168	<	* Create new worker using factory.
2169	<	* @param index the index to assign worker
2170	<	* @return new worker, or null of factory failed
2171	<	*/
2172	<	private ForkJoinWorkerThread createWorker(int index) {
2173	<	Thread.UncaughtExceptionHandler h = ueh;
2174	<	ForkJoinWorkerThread w = factory.newThread(this);
2175	<	if (w != null) {
2176	<	w.poolIndex = index;
2177	<	w.setDaemon(true);
2178	<	w.setAsyncMode(locallyFifo);
2179	<	w.setName("ForkJoinPool-" + poolNumber + "-worker-" + index);
2180	<	if (h != null)
2181	<	w.setUncaughtExceptionHandler(h);
2168	>	* Returns a cheap heuristic guide for task partitioning when
2169	>	* programmers, frameworks, tools, or languages have little or no
2170	>	* idea about task granularity.  In essence by offering this
2171	>	* method, we ask users only about tradeoffs in overhead vs
2172	>	* expected throughput and its variance, rather than how finely to
2173	>	* partition tasks.
2174	>	*
2175	>	* In a steady state strict (tree-structured) computation, each
2176	>	* thread makes available for stealing enough tasks for other
2177	>	* threads to remain active. Inductively, if all threads play by
2178	>	* the same rules, each thread should make available only a
2179	>	* constant number of tasks.
2180	>	*
2181	>	* The minimum useful constant is just 1. But using a value of 1
2182	>	* would require immediate replenishment upon each steal to
2183	>	* maintain enough tasks, which is infeasible.  Further,
2184	>	* partitionings/granularities of offered tasks should minimize
2185	>	* steal rates, which in general means that threads nearer the top
2186	>	* of computation tree should generate more than those nearer the
2187	>	* bottom. In perfect steady state, each thread is at
2188	>	* approximately the same level of computation tree. However,
2189	>	* producing extra tasks amortizes the uncertainty of progress and
2190	>	* diffusion assumptions.
2191	>	*
2192	>	* So, users will want to use values larger (but not much larger)
2193	>	* than 1 to both smooth over transient shortages and hedge
2194	>	* against uneven progress; as traded off against the cost of
2195	>	* extra task overhead. We leave the user to pick a threshold
2196	>	* value to compare with the results of this call to guide
2197	>	* decisions, but recommend values such as 3.
2198	>	*
2199	>	* When all threads are active, it is on average OK to estimate
2200	>	* surplus strictly locally. In steady-state, if one thread is
2201	>	* maintaining say 2 surplus tasks, then so are others. So we can
2202	>	* just use estimated queue length.  However, this strategy alone
2203	>	* leads to serious mis-estimates in some non-steady-state
2204	>	* conditions (ramp-up, ramp-down, other stalls). We can detect
2205	>	* many of these by further considering the number of "idle"
2206	>	* threads, that are known to have zero queued tasks, so
2207	>	* compensate by a factor of (#idle/#active) threads.
2208	>	*
2209	>	* Note: The approximation of #busy workers as #active workers is
2210	>	* not very good under current signalling scheme, and should be
2211	>	* improved.
2212	>	*/
2213	>	static int getSurplusQueuedTaskCount() {
2214	>	Thread t; ForkJoinWorkerThread wt; ForkJoinPool pool; WorkQueue q;
2215	>	if (((t = Thread.currentThread()) instanceof ForkJoinWorkerThread)) {
2216	>	int p = (pool = (wt = (ForkJoinWorkerThread)t).pool).config & SMASK;
2217	>	int n = (q = wt.workQueue).top - q.base;
2218	>	int a = (int)(pool.ctl >> AC_SHIFT) + p;
2219	>	return n - (a > (p >>>= 1) ? 0 :
2220	>	a > (p >>>= 1) ? 1 :
2221	>	a > (p >>>= 1) ? 2 :
2222	>	a > (p >>>= 1) ? 4 :
2223	>	8);
2224	>	}
2225	>	return 0;
2226	>	}
2227	>
2228	>	//  Termination
2229	>
2230	>	/**
2231	>	* Possibly initiates and/or completes termination.  The caller
2232	>	* triggering termination runs three passes through workQueues:
2233	>	* (0) Setting termination status, followed by wakeups of queued
2234	>	* workers; (1) cancelling all tasks; (2) interrupting lagging
2235	>	* threads (likely in external tasks, but possibly also blocked in
2236	>	* joins).  Each pass repeats previous steps because of potential
2237	>	* lagging thread creation.
2238	>	*
2239	>	* @param now if true, unconditionally terminate, else only
2240	>	* if no work and no active workers
2241	>	* @param enable if true, enable shutdown when next possible
2242	>	* @return true if now terminating or terminated
2243	>	*/
2244	>	private boolean tryTerminate(boolean now, boolean enable) {
2245	>	int ps;
2246	>	if (this == common)                    // cannot shut down
2247	>	return false;
2248	>	if ((ps = plock) >= 0) {                   // enable by setting plock
2249	>	if (!enable)
2250	>	return false;
2251	>	if ((ps & PL_LOCK) != 0 ||
2252	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
2253	>	ps = acquirePlock();
2254	>	int nps = ((ps + PL_LOCK) & ~SHUTDOWN) | SHUTDOWN;
2255	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
2256	>	releasePlock(nps);
2257	>	}
2258	>	for (long c;;) {
2259	>	if (((c = ctl) & STOP_BIT) != 0) {     // already terminating
2260	>	if ((short)(c >>> TC_SHIFT) == -(config & SMASK)) {
2261	>	synchronized (this) {
2262	>	notifyAll();               // signal when 0 workers
2263	>	}
2264	>	}
2265	>	return true;
2266	>	}
2267	>	if (!now) {                            // check if idle & no tasks
2268	>	WorkQueue[] ws; WorkQueue w;
2269	>	if ((int)(c >> AC_SHIFT) != -(config & SMASK))
2270	>	return false;
2271	>	if ((ws = workQueues) != null) {
2272	>	for (int i = 0; i < ws.length; ++i) {
2273	>	if ((w = ws[i]) != null) {
2274	>	if (!w.isEmpty()) {    // signal unprocessed tasks
2275	>	signalWork(w);
2276	>	return false;
2277	>	}
2278	>	if ((i & 1) != 0 && w.eventCount >= 0)
2279	>	return false;      // unqueued inactive worker
2280	>	}
2281	>	}
2282	>	}
2283	>	}
2284	>	if (U.compareAndSwapLong(this, CTL, c, c | STOP_BIT)) {
2285	>	for (int pass = 0; pass < 3; ++pass) {
2286	>	WorkQueue[] ws; WorkQueue w; Thread wt;
2287	>	if ((ws = workQueues) != null) {
2288	>	int n = ws.length;
2289	>	for (int i = 0; i < n; ++i) {
2290	>	if ((w = ws[i]) != null) {
2291	>	w.qlock = -1;
2292	>	if (pass > 0) {
2293	>	w.cancelAll();
2294	>	if (pass > 1 && (wt = w.owner) != null) {
2295	>	if (!wt.isInterrupted()) {
2296	>	try {
2297	>	wt.interrupt();
2298	>	} catch (Throwable ignore) {
2299	>	}
2300	>	}
2301	>	U.unpark(wt);
2302	>	}
2303	>	}
2304	>	}
2305	>	}
2306	>	// Wake up workers parked on event queue
2307	>	int i, e; long cc; Thread p;
2308	>	while ((e = (int)(cc = ctl) & E_MASK) != 0 &&
2309	>	(i = e & SMASK) < n && i >= 0 &&
2310	>	(w = ws[i]) != null) {
2311	>	long nc = ((long)(w.nextWait & E_MASK) |
2312	>	((cc + AC_UNIT) & AC_MASK) |
2313	>	(cc & (TC_MASK|STOP_BIT)));
2314	>	if (w.eventCount == (e | INT_SIGN) &&
2315	>	U.compareAndSwapLong(this, CTL, cc, nc)) {
2316	>	w.eventCount = (e + E_SEQ) & E_MASK;
2317	>	w.qlock = -1;
2318	>	if ((p = w.parker) != null)
2319	>	U.unpark(p);
2320	>	}
2321	>	}
2322	>	}
2323	>	}
2324	>	}
2325		}
420	–	return w;
2326		}
2327
2328	<	/**
424	<	* Return a good size for worker array given pool size.
425	<	* Currently requires size to be a power of two.
426	<	*/
427	<	private static int arraySizeFor(int ps) {
428	<	return ps <= 1? 1 : (1 << (32 - Integer.numberOfLeadingZeros(ps-1)));
429	<	}
2328	>	// external operations on common pool
2329
2330		/**
2331	<	* Create or resize array if necessary to hold newLength.
2332	<	* Call only under exlusion or lock
434	<	* @return the array
2331	>	* Returns common pool queue for a thread that has submitted at
2332	>	* least one task.
2333		*/
2334	<	private ForkJoinWorkerThread[] ensureWorkerArrayCapacity(int newLength) {
2335	<	ForkJoinWorkerThread[] ws = workers;
2336	<	if (ws == null)
2337	<	return workers = new ForkJoinWorkerThread[arraySizeFor(newLength)];
2338	<	else if (newLength > ws.length)
2339	<	return workers = Arrays.copyOf(ws, arraySizeFor(newLength));
2340	<	else
443	<	return ws;
2334	>	static WorkQueue commonSubmitterQueue() {
2335	>	ForkJoinPool p; WorkQueue[] ws; int m; Submitter z;
2336	>	return ((z = submitters.get()) != null &&
2337	>	(p = common) != null &&
2338	>	(ws = p.workQueues) != null &&
2339	>	(m = ws.length - 1) >= 0) ?
2340	>	ws[m & z.seed & SQMASK] : null;
2341		}
2342
2343		/**
2344	<	* Try to shrink workers into smaller array after one or more terminate
2345	<	*/
2346	<	private void tryShrinkWorkerArray() {
2347	<	ForkJoinWorkerThread[] ws = workers;
2348	<	if (ws != null) {
2349	<	int len = ws.length;
2350	<	int last = len - 1;
2351	<	while (last >= 0 && ws[last] == null)
2352	<	--last;
2353	<	int newLength = arraySizeFor(last+1);
2354	<	if (newLength < len)
2355	<	workers = Arrays.copyOf(ws, newLength);
2344	>	* Tries to pop the given task from submitter's queue in common pool.
2345	>	*/
2346	>	static boolean tryExternalUnpush(ForkJoinTask<?> t) {
2347	>	ForkJoinPool p; WorkQueue[] ws; WorkQueue q; Submitter z;
2348	>	ForkJoinTask<?>[] a;  int m, s;
2349	>	if (t != null &&
2350	>	(z = submitters.get()) != null &&
2351	>	(p = common) != null &&
2352	>	(ws = p.workQueues) != null &&
2353	>	(m = ws.length - 1) >= 0 &&
2354	>	(q = ws[m & z.seed & SQMASK]) != null &&
2355	>	(s = q.top) != q.base &&
2356	>	(a = q.array) != null) {
2357	>	long j = (((a.length - 1) & (s - 1)) << ASHIFT) + ABASE;
2358	>	if (U.getObject(a, j) == t &&
2359	>	U.compareAndSwapInt(q, QLOCK, 0, 1)) {
2360	>	if (q.array == a && q.top == s && // recheck
2361	>	U.compareAndSwapObject(a, j, t, null)) {
2362	>	q.top = s - 1;
2363	>	q.qlock = 0;
2364	>	return true;
2365	>	}
2366	>	q.qlock = 0;
2367	>	}
2368		}
2369	+	return false;
2370		}
2371
2372		/**
2373	<	* Initialize workers if necessary
2374	<	*/
2375	<	final void ensureWorkerInitialization() {
2376	<	ForkJoinWorkerThread[] ws = workers;
2377	<	if (ws == null) {
2378	<	final ReentrantLock lock = this.workerLock;
2379	<	lock.lock();
2380	<	try {
2381	<	ws = workers;
2382	<	if (ws == null) {
2383	<	int ps = parallelism;
2384	<	ws = ensureWorkerArrayCapacity(ps);
2385	<	for (int i = 0; i < ps; ++i) {
2386	<	ForkJoinWorkerThread w = createWorker(i);
2387	<	if (w != null) {
2388	<	ws[i] = w;
2389	<	w.start();
2390	<	updateWorkerCount(1);
2391	<	}
2373	>	* Tries to pop and run local tasks within the same computation
2374	>	* as the given root. On failure, tries to help complete from
2375	>	* other queues via helpComplete.
2376	>	*/
2377	>	private void externalHelpComplete(WorkQueue q, ForkJoinTask<?> root) {
2378	>	ForkJoinTask<?>[] a; int m;
2379	>	if (q != null && (a = q.array) != null && (m = (a.length - 1)) >= 0 &&
2380	>	root != null && root.status >= 0) {
2381	>	for (;;) {
2382	>	int s, u; Object o; CountedCompleter<?> task = null;
2383	>	if ((s = q.top) - q.base > 0) {
2384	>	long j = ((m & (s - 1)) << ASHIFT) + ABASE;
2385	>	if ((o = U.getObject(a, j)) != null &&
2386	>	(o instanceof CountedCompleter)) {
2387	>	CountedCompleter<?> t = (CountedCompleter<?>)o, r = t;
2388	>	do {
2389	>	if (r == root) {
2390	>	if (U.compareAndSwapInt(q, QLOCK, 0, 1)) {
2391	>	if (q.array == a && q.top == s &&
2392	>	U.compareAndSwapObject(a, j, t, null)) {
2393	>	q.top = s - 1;
2394	>	task = t;
2395	>	}
2396	>	q.qlock = 0;
2397	>	}
2398	>	break;
2399	>	}
2400	>	} while ((r = r.completer) != null);
2401		}
2402		}
2403	<	} finally {
2404	<	lock.unlock();
2403	>	if (task != null)
2404	>	task.doExec();
2405	>	if (root.status < 0 ||
2406	>	(u = (int)(ctl >>> 32)) >= 0 || (u >> UAC_SHIFT) >= 0)
2407	>	break;
2408	>	if (task == null) {
2409	>	helpSignal(root, q.poolIndex);
2410	>	if (root.status >= 0)
2411	>	helpComplete(root, SHARED_QUEUE);
2412	>	break;
2413	>	}
2414		}
2415		}
2416		}
2417
2418		/**
2419	<	* Worker creation and startup for threads added via setParallelism.
2419	>	* Tries to help execute or signal availability of the given task
2420	>	* from submitter's queue in common pool.
2421		*/
2422	<	private void createAndStartAddedWorkers() {
2423	<	resumeAllSpares();  // Allow spares to convert to nonspare
2424	<	int ps = parallelism;
2425	<	ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(ps);
2426	<	int len = ws.length;
2427	<	// Sweep through slots, to keep lowest indices most populated
2428	<	int k = 0;
2429	<	while (k < len) {
2430	<	if (ws[k] != null) {
2431	<	++k;
2432	<	continue;
2433	<	}
2434	<	int s = workerCounts;
2435	<	int tc = totalCountOf(s);
2436	<	int rc = runningCountOf(s);
2437	<	if (rc >= ps || tc >= ps)
2438	<	break;
2439	<	if (casWorkerCounts (s, workerCountsFor(tc+1, rc+1))) {
2440	<	ForkJoinWorkerThread w = createWorker(k);
2441	<	if (w != null) {
2442	<	ws[k++] = w;
2443	<	w.start();
2444	<	}
2445	<	else {
517	<	updateWorkerCount(-1); // back out on failed creation
518	<	break;
2422	>	static void externalHelpJoin(ForkJoinTask<?> t) {
2423	>	// Some hard-to-avoid overlap with tryExternalUnpush
2424	>	ForkJoinPool p; WorkQueue[] ws; WorkQueue q, w; Submitter z;
2425	>	ForkJoinTask<?>[] a;  int m, s, n;
2426	>	if (t != null &&
2427	>	(z = submitters.get()) != null &&
2428	>	(p = common) != null &&
2429	>	(ws = p.workQueues) != null &&
2430	>	(m = ws.length - 1) >= 0 &&
2431	>	(q = ws[m & z.seed & SQMASK]) != null &&
2432	>	(a = q.array) != null) {
2433	>	int am = a.length - 1;
2434	>	if ((s = q.top) != q.base) {
2435	>	long j = ((am & (s - 1)) << ASHIFT) + ABASE;
2436	>	if (U.getObject(a, j) == t &&
2437	>	U.compareAndSwapInt(q, QLOCK, 0, 1)) {
2438	>	if (q.array == a && q.top == s &&
2439	>	U.compareAndSwapObject(a, j, t, null)) {
2440	>	q.top = s - 1;
2441	>	q.qlock = 0;
2442	>	t.doExec();
2443	>	}
2444	>	else
2445	>	q.qlock = 0;
2446		}
2447		}
2448	+	if (t.status >= 0) {
2449	+	if (t instanceof CountedCompleter)
2450	+	p.externalHelpComplete(q, t);
2451	+	else
2452	+	p.helpSignal(t, q.poolIndex);
2453	+	}
2454		}
2455		}
2456
2457	<	// Execution methods
2457	>	// Exported methods
2458	>
2459	>	// Constructors
2460	>
2461	>	/**
2462	>	* Creates a {@code ForkJoinPool} with parallelism equal to {@link
2463	>	* java.lang.Runtime#availableProcessors}, using the {@linkplain
2464	>	* #defaultForkJoinWorkerThreadFactory default thread factory},
2465	>	* no UncaughtExceptionHandler, and non-async LIFO processing mode.
2466	>	*
2467	>	* @throws SecurityException if a security manager exists and
2468	>	*         the caller is not permitted to modify threads
2469	>	*         because it does not hold {@link
2470	>	*         java.lang.RuntimePermission}{@code ("modifyThread")}
2471	>	*/
2472	>	public ForkJoinPool() {
2473	>	this(Math.min(MAX_CAP, Runtime.getRuntime().availableProcessors()),
2474	>	defaultForkJoinWorkerThreadFactory, null, false);
2475	>	}
2476	>
2477	>	/**
2478	>	* Creates a {@code ForkJoinPool} with the indicated parallelism
2479	>	* level, the {@linkplain
2480	>	* #defaultForkJoinWorkerThreadFactory default thread factory},
2481	>	* no UncaughtExceptionHandler, and non-async LIFO processing mode.
2482	>	*
2483	>	* @param parallelism the parallelism level
2484	>	* @throws IllegalArgumentException if parallelism less than or
2485	>	*         equal to zero, or greater than implementation limit
2486	>	* @throws SecurityException if a security manager exists and
2487	>	*         the caller is not permitted to modify threads
2488	>	*         because it does not hold {@link
2489	>	*         java.lang.RuntimePermission}{@code ("modifyThread")}
2490	>	*/
2491	>	public ForkJoinPool(int parallelism) {
2492	>	this(parallelism, defaultForkJoinWorkerThreadFactory, null, false);
2493	>	}
2494
2495		/**
2496	<	* Common code for execute, invoke and submit
2496	>	* Creates a {@code ForkJoinPool} with the given parameters.
2497	>	*
2498	>	* @param parallelism the parallelism level. For default value,
2499	>	* use {@link java.lang.Runtime#availableProcessors}.
2500	>	* @param factory the factory for creating new threads. For default value,
2501	>	* use {@link #defaultForkJoinWorkerThreadFactory}.
2502	>	* @param handler the handler for internal worker threads that
2503	>	* terminate due to unrecoverable errors encountered while executing
2504	>	* tasks. For default value, use {@code null}.
2505	>	* @param asyncMode if true,
2506	>	* establishes local first-in-first-out scheduling mode for forked
2507	>	* tasks that are never joined. This mode may be more appropriate
2508	>	* than default locally stack-based mode in applications in which
2509	>	* worker threads only process event-style asynchronous tasks.
2510	>	* For default value, use {@code false}.
2511	>	* @throws IllegalArgumentException if parallelism less than or
2512	>	*         equal to zero, or greater than implementation limit
2513	>	* @throws NullPointerException if the factory is null
2514	>	* @throws SecurityException if a security manager exists and
2515	>	*         the caller is not permitted to modify threads
2516	>	*         because it does not hold {@link
2517	>	*         java.lang.RuntimePermission}{@code ("modifyThread")}
2518		*/
2519	<	private <T> void doSubmit(ForkJoinTask<T> task) {
2520	<	if (isShutdown())
2521	<	throw new RejectedExecutionException();
2522	<	if (workers == null)
2523	<	ensureWorkerInitialization();
2524	<	submissionQueue.offer(task);
2525	<	signalIdleWorkers();
2519	>	public ForkJoinPool(int parallelism,
2520	>	ForkJoinWorkerThreadFactory factory,
2521	>	Thread.UncaughtExceptionHandler handler,
2522	>	boolean asyncMode) {
2523	>	checkPermission();
2524	>	if (factory == null)
2525	>	throw new NullPointerException();
2526	>	if (parallelism <= 0 || parallelism > MAX_CAP)
2527	>	throw new IllegalArgumentException();
2528	>	this.factory = factory;
2529	>	this.ueh = handler;
2530	>	this.config = parallelism | (asyncMode ? (FIFO_QUEUE << 16) : 0);
2531	>	long np = (long)(-parallelism); // offset ctl counts
2532	>	this.ctl = ((np << AC_SHIFT) & AC_MASK) | ((np << TC_SHIFT) & TC_MASK);
2533	>	int pn = nextPoolId();
2534	>	StringBuilder sb = new StringBuilder("ForkJoinPool-");
2535	>	sb.append(Integer.toString(pn));
2536	>	sb.append("-worker-");
2537	>	this.workerNamePrefix = sb.toString();
2538	>	}
2539	>
2540	>	/**
2541	>	* Constructor for common pool, suitable only for static initialization.
2542	>	* Basically the same as above, but uses smallest possible initial footprint.
2543	>	*/
2544	>	ForkJoinPool(int parallelism, long ctl,
2545	>	ForkJoinWorkerThreadFactory factory,
2546	>	Thread.UncaughtExceptionHandler handler) {
2547	>	this.config = parallelism;
2548	>	this.ctl = ctl;
2549	>	this.factory = factory;
2550	>	this.ueh = handler;
2551	>	this.workerNamePrefix = "ForkJoinPool.commonPool-worker-";
2552	>	}
2553	>
2554	>	/**
2555	>	* Returns the common pool instance. This pool is statically
2556	>	* constructed; its run state is unaffected by attempts to {@link
2557	>	* #shutdown} or {@link #shutdownNow}. However this pool and any
2558	>	* ongoing processing are automatically terminated upon program
2559	>	* {@link System#exit}.  Any program that relies on asynchronous
2560	>	* task processing to complete before program termination should
2561	>	* invoke {@code commonPool().}{@link #awaitQuiescence}, before
2562	>	* exit.
2563	>	*
2564	>	* @return the common pool instance
2565	>	* @since 1.8
2566	>	*/
2567	>	public static ForkJoinPool commonPool() {
2568	>	// assert common != null : "static init error";
2569	>	return common;
2570		}
2571
2572	+	// Execution methods
2573	+
2574		/**
2575	<	* Performs the given task; returning its result upon completion
2575	>	* Performs the given task, returning its result upon completion.
2576	>	* If the computation encounters an unchecked Exception or Error,
2577	>	* it is rethrown as the outcome of this invocation.  Rethrown
2578	>	* exceptions behave in the same way as regular exceptions, but,
2579	>	* when possible, contain stack traces (as displayed for example
2580	>	* using {@code ex.printStackTrace()}) of both the current thread
2581	>	* as well as the thread actually encountering the exception;
2582	>	* minimally only the latter.
2583	>	*
2584		* @param task the task
2585		* @return the task's result
2586	<	* @throws NullPointerException if task is null
2587	<	* @throws RejectedExecutionException if pool is shut down
2586	>	* @throws NullPointerException if the task is null
2587	>	* @throws RejectedExecutionException if the task cannot be
2588	>	*         scheduled for execution
2589		*/
2590		public <T> T invoke(ForkJoinTask<T> task) {
2591	<	doSubmit(task);
2591	>	if (task == null)
2592	>	throw new NullPointerException();
2593	>	externalPush(task);
2594		return task.join();
2595		}
2596
2597		/**
2598		* Arranges for (asynchronous) execution of the given task.
2599	+	*
2600		* @param task the task
2601	<	* @throws NullPointerException if task is null
2602	<	* @throws RejectedExecutionException if pool is shut down
2601	>	* @throws NullPointerException if the task is null
2602	>	* @throws RejectedExecutionException if the task cannot be
2603	>	*         scheduled for execution
2604		*/
2605	<	public <T> void execute(ForkJoinTask<T> task) {
2606	<	doSubmit(task);
2605	>	public void execute(ForkJoinTask<?> task) {
2606	>	if (task == null)
2607	>	throw new NullPointerException();
2608	>	externalPush(task);
2609		}
2610
2611		// AbstractExecutorService methods
2612
2613	+	/**
2614	+	* @throws NullPointerException if the task is null
2615	+	* @throws RejectedExecutionException if the task cannot be
2616	+	*         scheduled for execution
2617	+	*/
2618		public void execute(Runnable task) {
2619	<	doSubmit(new AdaptedRunnable<Void>(task, null));
2619	>	if (task == null)
2620	>	throw new NullPointerException();
2621	>	ForkJoinTask<?> job;
2622	>	if (task instanceof ForkJoinTask<?>) // avoid re-wrap
2623	>	job = (ForkJoinTask<?>) task;
2624	>	else
2625	>	job = new ForkJoinTask.AdaptedRunnableAction(task);
2626	>	externalPush(job);
2627		}
2628
2629	<	public <T> ForkJoinTask<T> submit(Callable<T> task) {
2630	<	ForkJoinTask<T> job = new AdaptedCallable<T>(task);
2631	<	doSubmit(job);
2632	<	return job;
2629	>	/**
2630	>	* Submits a ForkJoinTask for execution.
2631	>	*
2632	>	* @param task the task to submit
2633	>	* @return the task
2634	>	* @throws NullPointerException if the task is null
2635	>	* @throws RejectedExecutionException if the task cannot be
2636	>	*         scheduled for execution
2637	>	*/
2638	>	public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) {
2639	>	if (task == null)
2640	>	throw new NullPointerException();
2641	>	externalPush(task);
2642	>	return task;
2643		}
2644
2645	<	public <T> ForkJoinTask<T> submit(Runnable task, T result) {
2646	<	ForkJoinTask<T> job = new AdaptedRunnable<T>(task, result);
2647	<	doSubmit(job);
2645	>	/**
2646	>	* @throws NullPointerException if the task is null
2647	>	* @throws RejectedExecutionException if the task cannot be
2648	>	*         scheduled for execution
2649	>	*/
2650	>	public <T> ForkJoinTask<T> submit(Callable<T> task) {
2651	>	ForkJoinTask<T> job = new ForkJoinTask.AdaptedCallable<T>(task);
2652	>	externalPush(job);
2653		return job;
2654		}
2655
2656	<	public ForkJoinTask<?> submit(Runnable task) {
2657	<	ForkJoinTask<Void> job = new AdaptedRunnable<Void>(task, null);
2658	<	doSubmit(job);
2656	>	/**
2657	>	* @throws NullPointerException if the task is null
2658	>	* @throws RejectedExecutionException if the task cannot be
2659	>	*         scheduled for execution
2660	>	*/
2661	>	public <T> ForkJoinTask<T> submit(Runnable task, T result) {
2662	>	ForkJoinTask<T> job = new ForkJoinTask.AdaptedRunnable<T>(task, result);
2663	>	externalPush(job);
2664		return job;
2665		}
2666
2667		/**
2668	<	* Adaptor for Runnables. This implements RunnableFuture
2669	<	* to be compliant with AbstractExecutorService constraints
2668	>	* @throws NullPointerException if the task is null
2669	>	* @throws RejectedExecutionException if the task cannot be
2670	>	*         scheduled for execution
2671		*/
2672	<	static final class AdaptedRunnable<T> extends ForkJoinTask<T>
2673	<	implements RunnableFuture<T> {
2674	<	final Runnable runnable;
2675	<	final T resultOnCompletion;
2676	<	T result;
2677	<	AdaptedRunnable(Runnable runnable, T result) {
2678	<	if (runnable == null) throw new NullPointerException();
2679	<	this.runnable = runnable;
2680	<	this.resultOnCompletion = result;
2681	<	}
598	<	public T getRawResult() { return result; }
599	<	public void setRawResult(T v) { result = v; }
600	<	public boolean exec() {
601	<	runnable.run();
602	<	result = resultOnCompletion;
603	<	return true;
604	<	}
605	<	public void run() { invoke(); }
2672	>	public ForkJoinTask<?> submit(Runnable task) {
2673	>	if (task == null)
2674	>	throw new NullPointerException();
2675	>	ForkJoinTask<?> job;
2676	>	if (task instanceof ForkJoinTask<?>) // avoid re-wrap
2677	>	job = (ForkJoinTask<?>) task;
2678	>	else
2679	>	job = new ForkJoinTask.AdaptedRunnableAction(task);
2680	>	externalPush(job);
2681	>	return job;
2682		}
2683
2684		/**
2685	<	* Adaptor for Callables
2685	>	* @throws NullPointerException       {@inheritDoc}
2686	>	* @throws RejectedExecutionException {@inheritDoc}
2687		*/
611	–	static final class AdaptedCallable<T> extends ForkJoinTask<T>
612	–	implements RunnableFuture<T> {
613	–	final Callable<T> callable;
614	–	T result;
615	–	AdaptedCallable(Callable<T> callable) {
616	–	if (callable == null) throw new NullPointerException();
617	–	this.callable = callable;
618	–	}
619	–	public T getRawResult() { return result; }
620	–	public void setRawResult(T v) { result = v; }
621	–	public boolean exec() {
622	–	try {
623	–	result = callable.call();
624	–	return true;
625	–	} catch (Error err) {
626	–	throw err;
627	–	} catch (RuntimeException rex) {
628	–	throw rex;
629	–	} catch (Exception ex) {
630	–	throw new RuntimeException(ex);
631	–	}
632	–	}
633	–	public void run() { invoke(); }
634	–	}
635	–
2688		public <T> List<Future<T>> invokeAll(Collection<? extends Callable<T>> tasks) {
2689	<	ArrayList<ForkJoinTask<T>> ts =
2690	<	new ArrayList<ForkJoinTask<T>>(tasks.size());
2691	<	for (Callable<T> c : tasks)
2692	<	ts.add(new AdaptedCallable<T>(c));
641	<	invoke(new InvokeAll<T>(ts));
642	<	return (List<Future<T>>)(List)ts;
643	<	}
2689	>	// In previous versions of this class, this method constructed
2690	>	// a task to run ForkJoinTask.invokeAll, but now external
2691	>	// invocation of multiple tasks is at least as efficient.
2692	>	ArrayList<Future<T>> futures = new ArrayList<Future<T>>(tasks.size());
2693
2694	<	static final class InvokeAll<T> extends RecursiveAction {
2695	<	final ArrayList<ForkJoinTask<T>> tasks;
2696	<	InvokeAll(ArrayList<ForkJoinTask<T>> tasks) { this.tasks = tasks; }
2697	<	public void compute() {
2698	<	try { invokeAll(tasks); } catch(Exception ignore) {}
2694	>	boolean done = false;
2695	>	try {
2696	>	for (Callable<T> t : tasks) {
2697	>	ForkJoinTask<T> f = new ForkJoinTask.AdaptedCallable<T>(t);
2698	>	futures.add(f);
2699	>	externalPush(f);
2700	>	}
2701	>	for (int i = 0, size = futures.size(); i < size; i++)
2702	>	((ForkJoinTask<?>)futures.get(i)).quietlyJoin();
2703	>	done = true;
2704	>	return futures;
2705	>	} finally {
2706	>	if (!done)
2707	>	for (int i = 0, size = futures.size(); i < size; i++)
2708	>	futures.get(i).cancel(false);
2709		}
2710		}
2711
653	–	// Configuration and status settings and queries
654	–
2712		/**
2713	<	* Returns the factory used for constructing new workers
2713	>	* Returns the factory used for constructing new workers.
2714		*
2715		* @return the factory used for constructing new workers
2716		*/
#	Line 664 | Line 2721 | public class ForkJoinPool extends Abstra
2721		/**
2722		* Returns the handler for internal worker threads that terminate
2723		* due to unrecoverable errors encountered while executing tasks.
2724	<	* @return the handler, or null if none
2724	>	*
2725	>	* @return the handler, or {@code null} if none
2726		*/
2727		public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() {
2728	<	Thread.UncaughtExceptionHandler h;
671	<	final ReentrantLock lock = this.workerLock;
672	<	lock.lock();
673	<	try {
674	<	h = ueh;
675	<	} finally {
676	<	lock.unlock();
677	<	}
678	<	return h;
2728	>	return ueh;
2729		}
2730
2731		/**
2732	<	* Sets the handler for internal worker threads that terminate due
683	<	* to unrecoverable errors encountered while executing tasks.
684	<	* Unless set, the current default or ThreadGroup handler is used
685	<	* as handler.
2732	>	* Returns the targeted parallelism level of this pool.
2733		*
2734	<	* @param h the new handler
688	<	* @return the old handler, or null if none
689	<	* @throws SecurityException if a security manager exists and
690	<	*         the caller is not permitted to modify threads
691	<	*         because it does not hold {@link
692	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
2734	>	* @return the targeted parallelism level of this pool
2735		*/
2736	<	public Thread.UncaughtExceptionHandler
2737	<	setUncaughtExceptionHandler(Thread.UncaughtExceptionHandler h) {
696	<	checkPermission();
697	<	Thread.UncaughtExceptionHandler old = null;
698	<	final ReentrantLock lock = this.workerLock;
699	<	lock.lock();
700	<	try {
701	<	old = ueh;
702	<	ueh = h;
703	<	ForkJoinWorkerThread[] ws = workers;
704	<	if (ws != null) {
705	<	for (int i = 0; i < ws.length; ++i) {
706	<	ForkJoinWorkerThread w = ws[i];
707	<	if (w != null)
708	<	w.setUncaughtExceptionHandler(h);
709	<	}
710	<	}
711	<	} finally {
712	<	lock.unlock();
713	<	}
714	<	return old;
715	<	}
716	<
717	<
718	<	/**
719	<	* Sets the target paralleism level of this pool.
720	<	* @param parallelism the target parallelism
721	<	* @throws IllegalArgumentException if parallelism less than or
722	<	* equal to zero or greater than maximum size bounds.
723	<	* @throws SecurityException if a security manager exists and
724	<	*         the caller is not permitted to modify threads
725	<	*         because it does not hold {@link
726	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
727	<	*/
728	<	public void setParallelism(int parallelism) {
729	<	checkPermission();
730	<	if (parallelism <= 0 || parallelism > maxPoolSize)
731	<	throw new IllegalArgumentException();
732	<	final ReentrantLock lock = this.workerLock;
733	<	lock.lock();
734	<	try {
735	<	if (!isTerminating()) {
736	<	int p = this.parallelism;
737	<	this.parallelism = parallelism;
738	<	if (parallelism > p)
739	<	createAndStartAddedWorkers();
740	<	else
741	<	trimSpares();
742	<	}
743	<	} finally {
744	<	lock.unlock();
745	<	}
746	<	signalIdleWorkers();
2736	>	public int getParallelism() {
2737	>	return config & SMASK;
2738		}
2739
2740		/**
2741	<	* Returns the targeted number of worker threads in this pool.
2741	>	* Returns the targeted parallelism level of the common pool.
2742		*
2743	<	* @return the targeted number of worker threads in this pool
2743	>	* @return the targeted parallelism level of the common pool
2744	>	* @since 1.8
2745		*/
2746	<	public int getParallelism() {
2747	<	return parallelism;
2746	>	public static int getCommonPoolParallelism() {
2747	>	return commonParallelism;
2748		}
2749
2750		/**
2751		* Returns the number of worker threads that have started but not
2752	<	* yet terminated.  This result returned by this method may differ
2753	<	* from <code>getParallelism</code> when threads are created to
2752	>	* yet terminated.  The result returned by this method may differ
2753	>	* from {@link #getParallelism} when threads are created to
2754		* maintain parallelism when others are cooperatively blocked.
2755		*
2756		* @return the number of worker threads
2757		*/
2758		public int getPoolSize() {
2759	<	return totalCountOf(workerCounts);
768	<	}
769	<
770	<	/**
771	<	* Returns the maximum number of threads allowed to exist in the
772	<	* pool, even if there are insufficient unblocked running threads.
773	<	* @return the maximum
774	<	*/
775	<	public int getMaximumPoolSize() {
776	<	return maxPoolSize;
2759	>	return (config & SMASK) + (short)(ctl >>> TC_SHIFT);
2760		}
2761
2762		/**
2763	<	* Sets the maximum number of threads allowed to exist in the
781	<	* pool, even if there are insufficient unblocked running threads.
782	<	* Setting this value has no effect on current pool size. It
783	<	* controls construction of new threads.
784	<	* @throws IllegalArgumentException if negative or greater then
785	<	* internal implementation limit.
786	<	*/
787	<	public void setMaximumPoolSize(int newMax) {
788	<	if (newMax < 0 || newMax > MAX_THREADS)
789	<	throw new IllegalArgumentException();
790	<	maxPoolSize = newMax;
791	<	}
792	<
793	<
794	<	/**
795	<	* Returns true if this pool dynamically maintains its target
796	<	* parallelism level. If false, new threads are added only to
797	<	* avoid possible starvation.
798	<	* This setting is by default true;
799	<	* @return true if maintains parallelism
800	<	*/
801	<	public boolean getMaintainsParallelism() {
802	<	return maintainsParallelism;
803	<	}
804	<
805	<	/**
806	<	* Sets whether this pool dynamically maintains its target
807	<	* parallelism level. If false, new threads are added only to
808	<	* avoid possible starvation.
809	<	* @param enable true to maintains parallelism
810	<	*/
811	<	public void setMaintainsParallelism(boolean enable) {
812	<	maintainsParallelism = enable;
813	<	}
814	<
815	<	/**
816	<	* Establishes local first-in-first-out scheduling mode for forked
817	<	* tasks that are never joined. This mode may be more appropriate
818	<	* than default locally stack-based mode in applications in which
819	<	* worker threads only process asynchronous tasks.  This method is
820	<	* designed to be invoked only when pool is quiescent, and
821	<	* typically only before any tasks are submitted. The effects of
822	<	* invocations at ather times may be unpredictable.
823	<	*
824	<	* @param async if true, use locally FIFO scheduling
825	<	* @return the previous mode.
826	<	*/
827	<	public boolean setAsyncMode(boolean async) {
828	<	boolean oldMode = locallyFifo;
829	<	locallyFifo = async;
830	<	ForkJoinWorkerThread[] ws = workers;
831	<	if (ws != null) {
832	<	for (int i = 0; i < ws.length; ++i) {
833	<	ForkJoinWorkerThread t = ws[i];
834	<	if (t != null)
835	<	t.setAsyncMode(async);
836	<	}
837	<	}
838	<	return oldMode;
839	<	}
840	<
841	<	/**
842	<	* Returns true if this pool uses local first-in-first-out
2763	>	* Returns {@code true} if this pool uses local first-in-first-out
2764		* scheduling mode for forked tasks that are never joined.
2765		*
2766	<	* @return true if this pool uses async mode.
2766	>	* @return {@code true} if this pool uses async mode
2767		*/
2768		public boolean getAsyncMode() {
2769	<	return locallyFifo;
2769	>	return (config >>> 16) == FIFO_QUEUE;
2770		}
2771
2772		/**
2773		* Returns an estimate of the number of worker threads that are
2774		* not blocked waiting to join tasks or for other managed
2775	<	* synchronization.
2775	>	* synchronization. This method may overestimate the
2776	>	* number of running threads.
2777		*
2778		* @return the number of worker threads
2779		*/
2780		public int getRunningThreadCount() {
2781	<	return runningCountOf(workerCounts);
2781	>	int rc = 0;
2782	>	WorkQueue[] ws; WorkQueue w;
2783	>	if ((ws = workQueues) != null) {
2784	>	for (int i = 1; i < ws.length; i += 2) {
2785	>	if ((w = ws[i]) != null && w.isApparentlyUnblocked())
2786	>	++rc;
2787	>	}
2788	>	}
2789	>	return rc;
2790		}
2791
2792		/**
2793		* Returns an estimate of the number of threads that are currently
2794		* stealing or executing tasks. This method may overestimate the
2795		* number of active threads.
2796	<	* @return the number of active threads.
2796	>	*
2797	>	* @return the number of active threads
2798		*/
2799		public int getActiveThreadCount() {
2800	<	return activeCountOf(runControl);
2800	>	int r = (config & SMASK) + (int)(ctl >> AC_SHIFT);
2801	>	return (r <= 0) ? 0 : r; // suppress momentarily negative values
2802		}
2803
2804		/**
2805	<	* Returns an estimate of the number of threads that are currently
2806	<	* idle waiting for tasks. This method may underestimate the
2807	<	* number of idle threads.
2808	<	* @return the number of idle threads.
2809	<	*/
2810	<	final int getIdleThreadCount() {
2811	<	int c = runningCountOf(workerCounts) - activeCountOf(runControl);
2812	<	return (c <= 0)? 0 : c;
2813	<	}
882	<
883	<	/**
884	<	* Returns true if all worker threads are currently idle. An idle
885	<	* worker is one that cannot obtain a task to execute because none
886	<	* are available to steal from other threads, and there are no
887	<	* pending submissions to the pool. This method is conservative:
888	<	* It might not return true immediately upon idleness of all
889	<	* threads, but will eventually become true if threads remain
890	<	* inactive.
891	<	* @return true if all threads are currently idle
2805	>	* Returns {@code true} if all worker threads are currently idle.
2806	>	* An idle worker is one that cannot obtain a task to execute
2807	>	* because none are available to steal from other threads, and
2808	>	* there are no pending submissions to the pool. This method is
2809	>	* conservative; it might not return {@code true} immediately upon
2810	>	* idleness of all threads, but will eventually become true if
2811	>	* threads remain inactive.
2812	>	*
2813	>	* @return {@code true} if all threads are currently idle
2814		*/
2815		public boolean isQuiescent() {
2816	<	return activeCountOf(runControl) == 0;
2816	>	return (int)(ctl >> AC_SHIFT) + (config & SMASK) == 0;
2817		}
2818
2819		/**
#	Line 899 | Line 2821 | public class ForkJoinPool extends Abstra
2821		* one thread's work queue by another. The reported value
2822		* underestimates the actual total number of steals when the pool
2823		* is not quiescent. This value may be useful for monitoring and
2824	<	* tuning fork/join programs: In general, steal counts should be
2824	>	* tuning fork/join programs: in general, steal counts should be
2825		* high enough to keep threads busy, but low enough to avoid
2826		* overhead and contention across threads.
2827	<	* @return the number of steals.
2827	>	*
2828	>	* @return the number of steals
2829		*/
2830		public long getStealCount() {
2831	<	return stealCount.get();
2832	<	}
2833	<
2834	<	/**
2835	<	* Accumulate steal count from a worker. Call only
2836	<	* when worker known to be idle.
2837	<	*/
2838	<	private void updateStealCount(ForkJoinWorkerThread w) {
2839	<	int sc = w.getAndClearStealCount();
917	<	if (sc != 0)
918	<	stealCount.addAndGet(sc);
2831	>	long count = stealCount;
2832	>	WorkQueue[] ws; WorkQueue w;
2833	>	if ((ws = workQueues) != null) {
2834	>	for (int i = 1; i < ws.length; i += 2) {
2835	>	if ((w = ws[i]) != null)
2836	>	count += w.nsteals;
2837	>	}
2838	>	}
2839	>	return count;
2840		}
2841
2842		/**
#	Line 925 | Line 2846 | public class ForkJoinPool extends Abstra
2846		* an approximation, obtained by iterating across all threads in
2847		* the pool. This method may be useful for tuning task
2848		* granularities.
2849	<	* @return the number of queued tasks.
2849	>	*
2850	>	* @return the number of queued tasks
2851		*/
2852		public long getQueuedTaskCount() {
2853		long count = 0;
2854	<	ForkJoinWorkerThread[] ws = workers;
2855	<	if (ws != null) {
2856	<	for (int i = 0; i < ws.length; ++i) {
2857	<	ForkJoinWorkerThread t = ws[i];
2858	<	if (t != null)
937	<	count += t.getQueueSize();
2854	>	WorkQueue[] ws; WorkQueue w;
2855	>	if ((ws = workQueues) != null) {
2856	>	for (int i = 1; i < ws.length; i += 2) {
2857	>	if ((w = ws[i]) != null)
2858	>	count += w.queueSize();
2859		}
2860		}
2861		return count;
2862		}
2863
2864		/**
2865	<	* Returns an estimate of the number tasks submitted to this pool
2866	<	* that have not yet begun executing. This method takes time
2867	<	* proportional to the number of submissions.
2868	<	* @return the number of queued submissions.
2865	>	* Returns an estimate of the number of tasks submitted to this
2866	>	* pool that have not yet begun executing.  This method may take
2867	>	* time proportional to the number of submissions.
2868	>	*
2869	>	* @return the number of queued submissions
2870		*/
2871		public int getQueuedSubmissionCount() {
2872	<	return submissionQueue.size();
2872	>	int count = 0;
2873	>	WorkQueue[] ws; WorkQueue w;
2874	>	if ((ws = workQueues) != null) {
2875	>	for (int i = 0; i < ws.length; i += 2) {
2876	>	if ((w = ws[i]) != null)
2877	>	count += w.queueSize();
2878	>	}
2879	>	}
2880	>	return count;
2881		}
2882
2883		/**
2884	<	* Returns true if there are any tasks submitted to this pool
2885	<	* that have not yet begun executing.
2886	<	* @return <code>true</code> if there are any queued submissions.
2884	>	* Returns {@code true} if there are any tasks submitted to this
2885	>	* pool that have not yet begun executing.
2886	>	*
2887	>	* @return {@code true} if there are any queued submissions
2888		*/
2889		public boolean hasQueuedSubmissions() {
2890	<	return !submissionQueue.isEmpty();
2890	>	WorkQueue[] ws; WorkQueue w;
2891	>	if ((ws = workQueues) != null) {
2892	>	for (int i = 0; i < ws.length; i += 2) {
2893	>	if ((w = ws[i]) != null && !w.isEmpty())
2894	>	return true;
2895	>	}
2896	>	}
2897	>	return false;
2898		}
2899
2900		/**
2901		* Removes and returns the next unexecuted submission if one is
2902		* available.  This method may be useful in extensions to this
2903		* class that re-assign work in systems with multiple pools.
2904	<	* @return the next submission, or null if none
2904	>	*
2905	>	* @return the next submission, or {@code null} if none
2906		*/
2907		protected ForkJoinTask<?> pollSubmission() {
2908	<	return submissionQueue.poll();
2908	>	WorkQueue[] ws; WorkQueue w; ForkJoinTask<?> t;
2909	>	if ((ws = workQueues) != null) {
2910	>	for (int i = 0; i < ws.length; i += 2) {
2911	>	if ((w = ws[i]) != null && (t = w.poll()) != null)
2912	>	return t;
2913	>	}
2914	>	}
2915	>	return null;
2916		}
2917
2918		/**
2919		* Removes all available unexecuted submitted and forked tasks
2920		* from scheduling queues and adds them to the given collection,
2921		* without altering their execution status. These may include
2922	<	* artifically generated or wrapped tasks. This method id designed
2923	<	* to be invoked only when the pool is known to be
2922	>	* artificially generated or wrapped tasks. This method is
2923	>	* designed to be invoked only when the pool is known to be
2924		* quiescent. Invocations at other times may not remove all
2925		* tasks. A failure encountered while attempting to add elements
2926	<	* to collection <tt>c</tt> may result in elements being in
2926	>	* to collection {@code c} may result in elements being in
2927		* neither, either or both collections when the associated
2928		* exception is thrown.  The behavior of this operation is
2929		* undefined if the specified collection is modified while the
2930		* operation is in progress.
2931	+	*
2932		* @param c the collection to transfer elements into
2933		* @return the number of elements transferred
2934		*/
2935	<	protected int drainTasksTo(Collection<ForkJoinTask<?>> c) {
2936	<	int n = submissionQueue.drainTo(c);
2937	<	ForkJoinWorkerThread[] ws = workers;
2938	<	if (ws != null) {
2935	>	protected int drainTasksTo(Collection<? super ForkJoinTask<?>> c) {
2936	>	int count = 0;
2937	>	WorkQueue[] ws; WorkQueue w; ForkJoinTask<?> t;
2938	>	if ((ws = workQueues) != null) {
2939		for (int i = 0; i < ws.length; ++i) {
2940	<	ForkJoinWorkerThread w = ws[i];
2941	<	if (w != null)
2942	<	n += w.drainTasksTo(c);
2940	>	if ((w = ws[i]) != null) {
2941	>	while ((t = w.poll()) != null) {
2942	>	c.add(t);
2943	>	++count;
2944	>	}
2945	>	}
2946		}
2947		}
2948	<	return n;
2948	>	return count;
2949		}
2950
2951		/**
#	Line 1006 | Line 2956 | public class ForkJoinPool extends Abstra
2956		* @return a string identifying this pool, as well as its state
2957		*/
2958		public String toString() {
2959	<	int ps = parallelism;
2960	<	int wc = workerCounts;
2961	<	int rc = runControl;
2962	<	long st = getStealCount();
2963	<	long qt = getQueuedTaskCount();
2964	<	long qs = getQueuedSubmissionCount();
2959	>	// Use a single pass through workQueues to collect counts
2960	>	long qt = 0L, qs = 0L; int rc = 0;
2961	>	long st = stealCount;
2962	>	long c = ctl;
2963	>	WorkQueue[] ws; WorkQueue w;
2964	>	if ((ws = workQueues) != null) {
2965	>	for (int i = 0; i < ws.length; ++i) {
2966	>	if ((w = ws[i]) != null) {
2967	>	int size = w.queueSize();
2968	>	if ((i & 1) == 0)
2969	>	qs += size;
2970	>	else {
2971	>	qt += size;
2972	>	st += w.nsteals;
2973	>	if (w.isApparentlyUnblocked())
2974	>	++rc;
2975	>	}
2976	>	}
2977	>	}
2978	>	}
2979	>	int pc = (config & SMASK);
2980	>	int tc = pc + (short)(c >>> TC_SHIFT);
2981	>	int ac = pc + (int)(c >> AC_SHIFT);
2982	>	if (ac < 0) // ignore transient negative
2983	>	ac = 0;
2984	>	String level;
2985	>	if ((c & STOP_BIT) != 0)
2986	>	level = (tc == 0) ? "Terminated" : "Terminating";
2987	>	else
2988	>	level = plock < 0 ? "Shutting down" : "Running";
2989		return super.toString() +
2990	<	"[" + runStateToString(runStateOf(rc)) +
2991	<	", parallelism = " + ps +
2992	<	", size = " + totalCountOf(wc) +
2993	<	", active = " + activeCountOf(rc) +
2994	<	", running = " + runningCountOf(wc) +
2990	>	"[" + level +
2991	>	", parallelism = " + pc +
2992	>	", size = " + tc +
2993	>	", active = " + ac +
2994	>	", running = " + rc +
2995		", steals = " + st +
2996		", tasks = " + qt +
2997		", submissions = " + qs +
2998		"]";
2999		}
3000
1027	–	private static String runStateToString(int rs) {
1028	–	switch(rs) {
1029	–	case RUNNING: return "Running";
1030	–	case SHUTDOWN: return "Shutting down";
1031	–	case TERMINATING: return "Terminating";
1032	–	case TERMINATED: return "Terminated";
1033	–	default: throw new Error("Unknown run state");
1034	–	}
1035	–	}
1036	–
1037	–	// lifecycle control
1038	–
3001		/**
3002	<	* Initiates an orderly shutdown in which previously submitted
3003	<	* tasks are executed, but no new tasks will be accepted.
3004	<	* Invocation has no additional effect if already shut down.
3005	<	* Tasks that are in the process of being submitted concurrently
3006	<	* during the course of this method may or may not be rejected.
3002	>	* Possibly initiates an orderly shutdown in which previously
3003	>	* submitted tasks are executed, but no new tasks will be
3004	>	* accepted. Invocation has no effect on execution state if this
3005	>	* is the {@link #commonPool()}, and no additional effect if
3006	>	* already shut down.  Tasks that are in the process of being
3007	>	* submitted concurrently during the course of this method may or
3008	>	* may not be rejected.
3009	>	*
3010		* @throws SecurityException if a security manager exists and
3011		*         the caller is not permitted to modify threads
3012		*         because it does not hold {@link
3013	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
3013	>	*         java.lang.RuntimePermission}{@code ("modifyThread")}
3014		*/
3015		public void shutdown() {
3016		checkPermission();
3017	<	transitionRunStateTo(SHUTDOWN);
1053	<	if (canTerminateOnShutdown(runControl))
1054	<	terminateOnShutdown();
3017	>	tryTerminate(false, true);
3018		}
3019
3020		/**
3021	<	* Attempts to stop all actively executing tasks, and cancels all
3022	<	* waiting tasks.  Tasks that are in the process of being
3023	<	* submitted or executed concurrently during the course of this
3024	<	* method may or may not be rejected. Unlike some other executors,
3025	<	* this method cancels rather than collects non-executed tasks
3026	<	* upon termination, so always returns an empty list. However, you
3027	<	* can use method <code>drainTasksTo</code> before invoking this
3028	<	* method to transfer unexecuted tasks to another collection.
3021	>	* Possibly attempts to cancel and/or stop all tasks, and reject
3022	>	* all subsequently submitted tasks.  Invocation has no effect on
3023	>	* execution state if this is the {@link #commonPool()}, and no
3024	>	* additional effect if already shut down. Otherwise, tasks that
3025	>	* are in the process of being submitted or executed concurrently
3026	>	* during the course of this method may or may not be
3027	>	* rejected. This method cancels both existing and unexecuted
3028	>	* tasks, in order to permit termination in the presence of task
3029	>	* dependencies. So the method always returns an empty list
3030	>	* (unlike the case for some other Executors).
3031	>	*
3032		* @return an empty list
3033		* @throws SecurityException if a security manager exists and
3034		*         the caller is not permitted to modify threads
3035		*         because it does not hold {@link
3036	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
3036	>	*         java.lang.RuntimePermission}{@code ("modifyThread")}
3037		*/
3038		public List<Runnable> shutdownNow() {
3039		checkPermission();
3040	<	terminate();
3040	>	tryTerminate(true, true);
3041		return Collections.emptyList();
3042		}
3043
3044		/**
3045	<	* Returns <code>true</code> if all tasks have completed following shut down.
3045	>	* Returns {@code true} if all tasks have completed following shut down.
3046		*
3047	<	* @return <code>true</code> if all tasks have completed following shut down
3047	>	* @return {@code true} if all tasks have completed following shut down
3048		*/
3049		public boolean isTerminated() {
3050	<	return runStateOf(runControl) == TERMINATED;
3050	>	long c = ctl;
3051	>	return ((c & STOP_BIT) != 0L &&
3052	>	(short)(c >>> TC_SHIFT) == -(config & SMASK));
3053		}
3054
3055		/**
3056	<	* Returns <code>true</code> if the process of termination has
3057	<	* commenced but possibly not yet completed.
3056	>	* Returns {@code true} if the process of termination has
3057	>	* commenced but not yet completed.  This method may be useful for
3058	>	* debugging. A return of {@code true} reported a sufficient
3059	>	* period after shutdown may indicate that submitted tasks have
3060	>	* ignored or suppressed interruption, or are waiting for I/O,
3061	>	* causing this executor not to properly terminate. (See the
3062	>	* advisory notes for class {@link ForkJoinTask} stating that
3063	>	* tasks should not normally entail blocking operations.  But if
3064	>	* they do, they must abort them on interrupt.)
3065		*
3066	<	* @return <code>true</code> if terminating
3066	>	* @return {@code true} if terminating but not yet terminated
3067		*/
3068		public boolean isTerminating() {
3069	<	return runStateOf(runControl) >= TERMINATING;
3069	>	long c = ctl;
3070	>	return ((c & STOP_BIT) != 0L &&
3071	>	(short)(c >>> TC_SHIFT) != -(config & SMASK));
3072		}
3073
3074		/**
3075	<	* Returns <code>true</code> if this pool has been shut down.
3075	>	* Returns {@code true} if this pool has been shut down.
3076		*
3077	<	* @return <code>true</code> if this pool has been shut down
3077	>	* @return {@code true} if this pool has been shut down
3078		*/
3079		public boolean isShutdown() {
3080	<	return runStateOf(runControl) >= SHUTDOWN;
3080	>	return plock < 0;
3081		}
3082
3083		/**
3084	<	* Blocks until all tasks have completed execution after a shutdown
3085	<	* request, or the timeout occurs, or the current thread is
3086	<	* interrupted, whichever happens first.
3084	>	* Blocks until all tasks have completed execution after a
3085	>	* shutdown request, or the timeout occurs, or the current thread
3086	>	* is interrupted, whichever happens first. Because the {@link
3087	>	* #commonPool()} never terminates until program shutdown, when
3088	>	* applied to the common pool, this method is equivalent to {@link
3089	>	* #awaitQuiescence} but always returns {@code false}.
3090		*
3091		* @param timeout the maximum time to wait
3092		* @param unit the time unit of the timeout argument
3093	<	* @return <code>true</code> if this executor terminated and
3094	<	*         <code>false</code> if the timeout elapsed before termination
3093	>	* @return {@code true} if this executor terminated and
3094	>	*         {@code false} if the timeout elapsed before termination
3095		* @throws InterruptedException if interrupted while waiting
3096		*/
3097		public boolean awaitTermination(long timeout, TimeUnit unit)
3098		throws InterruptedException {
3099	<	long nanos = unit.toNanos(timeout);
3100	<	final ReentrantLock lock = this.workerLock;
3101	<	lock.lock();
3102	<	try {
3103	<	for (;;) {
1124	<	if (isTerminated())
1125	<	return true;
1126	<	if (nanos <= 0)
1127	<	return false;
1128	<	nanos = termination.awaitNanos(nanos);
1129	<	}
1130	<	} finally {
1131	<	lock.unlock();
1132	<	}
1133	<	}
1134	<
1135	<	// Shutdown and termination support
1136	<
1137	<	/**
1138	<	* Callback from terminating worker. Null out the corresponding
1139	<	* workers slot, and if terminating, try to terminate, else try to
1140	<	* shrink workers array.
1141	<	* @param w the worker
1142	<	*/
1143	<	final void workerTerminated(ForkJoinWorkerThread w) {
1144	<	updateStealCount(w);
1145	<	updateWorkerCount(-1);
1146	<	final ReentrantLock lock = this.workerLock;
1147	<	lock.lock();
1148	<	try {
1149	<	ForkJoinWorkerThread[] ws = workers;
1150	<	if (ws != null) {
1151	<	int idx = w.poolIndex;
1152	<	if (idx >= 0 && idx < ws.length && ws[idx] == w)
1153	<	ws[idx] = null;
1154	<	if (totalCountOf(workerCounts) == 0) {
1155	<	terminate(); // no-op if already terminating
1156	<	transitionRunStateTo(TERMINATED);
1157	<	termination.signalAll();
1158	<	}
1159	<	else if (!isTerminating()) {
1160	<	tryShrinkWorkerArray();
1161	<	tryResumeSpare(true); // allow replacement
1162	<	}
1163	<	}
1164	<	} finally {
1165	<	lock.unlock();
1166	<	}
1167	<	signalIdleWorkers();
1168	<	}
1169	<
1170	<	/**
1171	<	* Initiate termination.
1172	<	*/
1173	<	private void terminate() {
1174	<	if (transitionRunStateTo(TERMINATING)) {
1175	<	stopAllWorkers();
1176	<	resumeAllSpares();
1177	<	signalIdleWorkers();
1178	<	cancelQueuedSubmissions();
1179	<	cancelQueuedWorkerTasks();
1180	<	interruptUnterminatedWorkers();
1181	<	signalIdleWorkers(); // resignal after interrupt
1182	<	}
1183	<	}
1184	<
1185	<	/**
1186	<	* Possibly terminate when on shutdown state
1187	<	*/
1188	<	private void terminateOnShutdown() {
1189	<	if (!hasQueuedSubmissions() && canTerminateOnShutdown(runControl))
1190	<	terminate();
1191	<	}
1192	<
1193	<	/**
1194	<	* Clear out and cancel submissions
1195	<	*/
1196	<	private void cancelQueuedSubmissions() {
1197	<	ForkJoinTask<?> task;
1198	<	while ((task = pollSubmission()) != null)
1199	<	task.cancel(false);
1200	<	}
1201	<
1202	<	/**
1203	<	* Clean out worker queues.
1204	<	*/
1205	<	private void cancelQueuedWorkerTasks() {
1206	<	final ReentrantLock lock = this.workerLock;
1207	<	lock.lock();
1208	<	try {
1209	<	ForkJoinWorkerThread[] ws = workers;
1210	<	if (ws != null) {
1211	<	for (int i = 0; i < ws.length; ++i) {
1212	<	ForkJoinWorkerThread t = ws[i];
1213	<	if (t != null)
1214	<	t.cancelTasks();
1215	<	}
1216	<	}
1217	<	} finally {
1218	<	lock.unlock();
1219	<	}
1220	<	}
1221	<
1222	<	/**
1223	<	* Set each worker's status to terminating. Requires lock to avoid
1224	<	* conflicts with add/remove
1225	<	*/
1226	<	private void stopAllWorkers() {
1227	<	final ReentrantLock lock = this.workerLock;
1228	<	lock.lock();
1229	<	try {
1230	<	ForkJoinWorkerThread[] ws = workers;
1231	<	if (ws != null) {
1232	<	for (int i = 0; i < ws.length; ++i) {
1233	<	ForkJoinWorkerThread t = ws[i];
1234	<	if (t != null)
1235	<	t.shutdownNow();
1236	<	}
1237	<	}
1238	<	} finally {
1239	<	lock.unlock();
1240	<	}
1241	<	}
1242	<
1243	<	/**
1244	<	* Interrupt all unterminated workers.  This is not required for
1245	<	* sake of internal control, but may help unstick user code during
1246	<	* shutdown.
1247	<	*/
1248	<	private void interruptUnterminatedWorkers() {
1249	<	final ReentrantLock lock = this.workerLock;
1250	<	lock.lock();
1251	<	try {
1252	<	ForkJoinWorkerThread[] ws = workers;
1253	<	if (ws != null) {
1254	<	for (int i = 0; i < ws.length; ++i) {
1255	<	ForkJoinWorkerThread t = ws[i];
1256	<	if (t != null && !t.isTerminated()) {
1257	<	try {
1258	<	t.interrupt();
1259	<	} catch (SecurityException ignore) {
1260	<	}
1261	<	}
1262	<	}
1263	<	}
1264	<	} finally {
1265	<	lock.unlock();
1266	<	}
1267	<	}
1268	<
1269	<
1270	<	/*
1271	<	* Nodes for event barrier to manage idle threads.  Queue nodes
1272	<	* are basic Treiber stack nodes, also used for spare stack.
1273	<	*
1274	<	* The event barrier has an event count and a wait queue (actually
1275	<	* a Treiber stack).  Workers are enabled to look for work when
1276	<	* the eventCount is incremented. If they fail to find work, they
1277	<	* may wait for next count. Upon release, threads help others wake
1278	<	* up.
1279	<	*
1280	<	* Synchronization events occur only in enough contexts to
1281	<	* maintain overall liveness:
1282	<	*
1283	<	*   - Submission of a new task to the pool
1284	<	*   - Resizes or other changes to the workers array
1285	<	*   - pool termination
1286	<	*   - A worker pushing a task on an empty queue
1287	<	*
1288	<	* The case of pushing a task occurs often enough, and is heavy
1289	<	* enough compared to simple stack pushes, to require special
1290	<	* handling: Method signalWork returns without advancing count if
1291	<	* the queue appears to be empty.  This would ordinarily result in
1292	<	* races causing some queued waiters not to be woken up. To avoid
1293	<	* this, the first worker enqueued in method sync (see
1294	<	* syncIsReleasable) rescans for tasks after being enqueued, and
1295	<	* helps signal if any are found. This works well because the
1296	<	* worker has nothing better to do, and so might as well help
1297	<	* alleviate the overhead and contention on the threads actually
1298	<	* doing work.  Also, since event counts increments on task
1299	<	* availability exist to maintain liveness (rather than to force
1300	<	* refreshes etc), it is OK for callers to exit early if
1301	<	* contending with another signaller.
1302	<	*/
1303	<	static final class WaitQueueNode {
1304	<	WaitQueueNode next; // only written before enqueued
1305	<	volatile ForkJoinWorkerThread thread; // nulled to cancel wait
1306	<	final long count; // unused for spare stack
1307	<
1308	<	WaitQueueNode(long c, ForkJoinWorkerThread w) {
1309	<	count = c;
1310	<	thread = w;
3099	>	if (Thread.interrupted())
3100	>	throw new InterruptedException();
3101	>	if (this == common) {
3102	>	awaitQuiescence(timeout, unit);
3103	>	return false;
3104		}
3105	<
3106	<	/**
1314	<	* Wake up waiter, returning false if known to already
1315	<	*/
1316	<	boolean signal() {
1317	<	ForkJoinWorkerThread t = thread;
1318	<	if (t == null)
1319	<	return false;
1320	<	thread = null;
1321	<	LockSupport.unpark(t);
3105	>	long nanos = unit.toNanos(timeout);
3106	>	if (isTerminated())
3107		return true;
3108	<	}
3109	<
3110	<	/**
3111	<	* Await release on sync
3112	<	*/
3113	<	void awaitSyncRelease(ForkJoinPool p) {
3114	<	while (thread != null && !p.syncIsReleasable(this))
1330	<	LockSupport.park(this);
1331	<	}
1332	<
1333	<	/**
1334	<	* Await resumption as spare
1335	<	*/
1336	<	void awaitSpareRelease() {
1337	<	while (thread != null) {
1338	<	if (!Thread.interrupted())
1339	<	LockSupport.park(this);
1340	<	}
1341	<	}
1342	<	}
1343	<
1344	<	/**
1345	<	* Ensures that no thread is waiting for count to advance from the
1346	<	* current value of eventCount read on entry to this method, by
1347	<	* releasing waiting threads if necessary.
1348	<	* @return the count
1349	<	*/
1350	<	final long ensureSync() {
1351	<	long c = eventCount;
1352	<	WaitQueueNode q;
1353	<	while ((q = syncStack) != null && q.count < c) {
1354	<	if (casBarrierStack(q, null)) {
1355	<	do {
1356	<	q.signal();
1357	<	} while ((q = q.next) != null);
1358	<	break;
1359	<	}
1360	<	}
1361	<	return c;
1362	<	}
1363	<
1364	<	/**
1365	<	* Increments event count and releases waiting threads.
1366	<	*/
1367	<	private void signalIdleWorkers() {
1368	<	long c;
1369	<	do;while (!casEventCount(c = eventCount, c+1));
1370	<	ensureSync();
1371	<	}
1372	<
1373	<	/**
1374	<	* Signal threads waiting to poll a task. Because method sync
1375	<	* rechecks availability, it is OK to only proceed if queue
1376	<	* appears to be non-empty, and OK to skip under contention to
1377	<	* increment count (since some other thread succeeded).
1378	<	*/
1379	<	final void signalWork() {
1380	<	long c;
1381	<	WaitQueueNode q;
1382	<	if (syncStack != null &&
1383	<	casEventCount(c = eventCount, c+1) &&
1384	<	(((q = syncStack) != null && q.count <= c) &&
1385	<	(!casBarrierStack(q, q.next) || !q.signal())))
1386	<	ensureSync();
1387	<	}
1388	<
1389	<	/**
1390	<	* Waits until event count advances from last value held by
1391	<	* caller, or if excess threads, caller is resumed as spare, or
1392	<	* caller or pool is terminating. Updates caller's event on exit.
1393	<	* @param w the calling worker thread
1394	<	*/
1395	<	final void sync(ForkJoinWorkerThread w) {
1396	<	updateStealCount(w); // Transfer w's count while it is idle
1397	<
1398	<	while (!w.isShutdown() && !isTerminating() && !suspendIfSpare(w)) {
1399	<	long prev = w.lastEventCount;
1400	<	WaitQueueNode node = null;
1401	<	WaitQueueNode h;
1402	<	while (eventCount == prev &&
1403	<	((h = syncStack) == null || h.count == prev)) {
1404	<	if (node == null)
1405	<	node = new WaitQueueNode(prev, w);
1406	<	if (casBarrierStack(node.next = h, node)) {
1407	<	node.awaitSyncRelease(this);
3108	>	long startTime = System.nanoTime();
3109	>	boolean terminated = false;
3110	>	synchronized (this) {
3111	>	for (long waitTime = nanos, millis = 0L;;) {
3112	>	if (terminated = isTerminated() ||
3113	>	waitTime <= 0L ||
3114	>	(millis = unit.toMillis(waitTime)) <= 0L)
3115		break;
3116	<	}
3116	>	wait(millis);
3117	>	waitTime = nanos - (System.nanoTime() - startTime);
3118		}
1411	–	long ec = ensureSync();
1412	–	if (ec != prev) {
1413	–	w.lastEventCount = ec;
1414	–	break;
1415	–	}
1416	–	}
1417	–	}
1418	–
1419	–	/**
1420	–	* Returns true if worker waiting on sync can proceed:
1421	–	*  - on signal (thread == null)
1422	–	*  - on event count advance (winning race to notify vs signaller)
1423	–	*  - on Interrupt
1424	–	*  - if the first queued node, we find work available
1425	–	* If node was not signalled and event count not advanced on exit,
1426	–	* then we also help advance event count.
1427	–	* @return true if node can be released
1428	–	*/
1429	–	final boolean syncIsReleasable(WaitQueueNode node) {
1430	–	long prev = node.count;
1431	–	if (!Thread.interrupted() && node.thread != null &&
1432	–	(node.next != null ||
1433	–	!ForkJoinWorkerThread.hasQueuedTasks(workers)) &&
1434	–	eventCount == prev)
1435	–	return false;
1436	–	if (node.thread != null) {
1437	–	node.thread = null;
1438	–	long ec = eventCount;
1439	–	if (prev <= ec) // help signal
1440	–	casEventCount(ec, ec+1);
3119		}
3120	<	return true;
1443	<	}
1444	<
1445	<	/**
1446	<	* Returns true if a new sync event occurred since last call to
1447	<	* sync or this method, if so, updating caller's count.
1448	<	*/
1449	<	final boolean hasNewSyncEvent(ForkJoinWorkerThread w) {
1450	<	long lc = w.lastEventCount;
1451	<	long ec = ensureSync();
1452	<	if (ec == lc)
1453	<	return false;
1454	<	w.lastEventCount = ec;
1455	<	return true;
3120	>	return terminated;
3121		}
3122
1458	–	//  Parallelism maintenance
1459	–
3123		/**
3124	<	* Decrement running count; if too low, add spare.
3124	>	* If called by a ForkJoinTask operating in this pool, equivalent
3125	>	* in effect to {@link ForkJoinTask#helpQuiesce}. Otherwise,
3126	>	* waits and/or attempts to assist performing tasks until this
3127	>	* pool {@link #isQuiescent} or the indicated timeout elapses.
3128		*
3129	<	* Conceptually, all we need to do here is add or resume a
3130	<	* spare thread when one is about to block (and remove or
3131	<	* suspend it later when unblocked -- see suspendIfSpare).
3132	<	* However, implementing this idea requires coping with
1467	<	* several problems: We have imperfect information about the
1468	<	* states of threads. Some count updates can and usually do
1469	<	* lag run state changes, despite arrangements to keep them
1470	<	* accurate (for example, when possible, updating counts
1471	<	* before signalling or resuming), especially when running on
1472	<	* dynamic JVMs that don't optimize the infrequent paths that
1473	<	* update counts. Generating too many threads can make these
1474	<	* problems become worse, because excess threads are more
1475	<	* likely to be context-switched with others, slowing them all
1476	<	* down, especially if there is no work available, so all are
1477	<	* busy scanning or idling.  Also, excess spare threads can
1478	<	* only be suspended or removed when they are idle, not
1479	<	* immediately when they aren't needed. So adding threads will
1480	<	* raise parallelism level for longer than necessary.  Also,
1481	<	* FJ applications often enounter highly transient peaks when
1482	<	* many threads are blocked joining, but for less time than it
1483	<	* takes to create or resume spares.
1484	<	*
1485	<	* @param joinMe if non-null, return early if done
1486	<	* @param maintainParallelism if true, try to stay within
1487	<	* target counts, else create only to avoid starvation
1488	<	* @return true if joinMe known to be done
1489	<	*/
1490	<	final boolean preJoin(ForkJoinTask<?> joinMe, boolean maintainParallelism) {
1491	<	maintainParallelism &= maintainsParallelism; // overrride
1492	<	boolean dec = false;  // true when running count decremented
1493	<	while (spareStack == null || !tryResumeSpare(dec)) {
1494	<	int counts = workerCounts;
1495	<	if (dec || (dec = casWorkerCounts(counts, --counts))) { // CAS cheat
1496	<	if (!needSpare(counts, maintainParallelism))
1497	<	break;
1498	<	if (joinMe.status < 0)
1499	<	return true;
1500	<	if (tryAddSpare(counts))
1501	<	break;
1502	<	}
1503	<	}
1504	<	return false;
1505	<	}
1506	<
1507	<	/**
1508	<	* Same idea as preJoin
3129	>	* @param timeout the maximum time to wait
3130	>	* @param unit the time unit of the timeout argument
3131	>	* @return {@code true} if quiescent; {@code false} if the
3132	>	* timeout elapsed.
3133		*/
3134	<	final boolean preBlock(ManagedBlocker blocker, boolean maintainParallelism){
3135	<	maintainParallelism &= maintainsParallelism;
3136	<	boolean dec = false;
3137	<	while (spareStack == null || !tryResumeSpare(dec)) {
3138	<	int counts = workerCounts;
3139	<	if (dec || (dec = casWorkerCounts(counts, --counts))) {
3140	<	if (!needSpare(counts, maintainParallelism))
3141	<	break;
1518	<	if (blocker.isReleasable())
1519	<	return true;
1520	<	if (tryAddSpare(counts))
1521	<	break;
1522	<	}
3134	>	public boolean awaitQuiescence(long timeout, TimeUnit unit) {
3135	>	long nanos = unit.toNanos(timeout);
3136	>	ForkJoinWorkerThread wt;
3137	>	Thread thread = Thread.currentThread();
3138	>	if ((thread instanceof ForkJoinWorkerThread) &&
3139	>	(wt = (ForkJoinWorkerThread)thread).pool == this) {
3140	>	helpQuiescePool(wt.workQueue);
3141	>	return true;
3142		}
3143	<	return false;
3144	<	}
3145	<
3146	<	/**
3147	<	* Returns true if a spare thread appears to be needed.  If
3148	<	* maintaining parallelism, returns true when the deficit in
3149	<	* running threads is more than the surplus of total threads, and
3150	<	* there is apparently some work to do.  This self-limiting rule
3151	<	* means that the more threads that have already been added, the
3152	<	* less parallelism we will tolerate before adding another.
3153	<	* @param counts current worker counts
3154	<	* @param maintainParallelism try to maintain parallelism
3155	<	*/
3156	<	private boolean needSpare(int counts, boolean maintainParallelism) {
3157	<	int ps = parallelism;
3158	<	int rc = runningCountOf(counts);
3159	<	int tc = totalCountOf(counts);
3160	<	int runningDeficit = ps - rc;
3161	<	int totalSurplus = tc - ps;
3162	<	return (tc < maxPoolSize &&
3163	<	(rc == 0 || totalSurplus < 0 ||
1545	<	(maintainParallelism &&
1546	<	runningDeficit > totalSurplus &&
1547	<	ForkJoinWorkerThread.hasQueuedTasks(workers))));
1548	<	}
1549	<
1550	<	/**
1551	<	* Add a spare worker if lock available and no more than the
1552	<	* expected numbers of threads exist
1553	<	* @return true if successful
1554	<	*/
1555	<	private boolean tryAddSpare(int expectedCounts) {
1556	<	final ReentrantLock lock = this.workerLock;
1557	<	int expectedRunning = runningCountOf(expectedCounts);
1558	<	int expectedTotal = totalCountOf(expectedCounts);
1559	<	boolean success = false;
1560	<	boolean locked = false;
1561	<	// confirm counts while locking; CAS after obtaining lock
1562	<	try {
1563	<	for (;;) {
1564	<	int s = workerCounts;
1565	<	int tc = totalCountOf(s);
1566	<	int rc = runningCountOf(s);
1567	<	if (rc > expectedRunning || tc > expectedTotal)
1568	<	break;
1569	<	if (!locked && !(locked = lock.tryLock()))
1570	<	break;
1571	<	if (casWorkerCounts(s, workerCountsFor(tc+1, rc+1))) {
1572	<	createAndStartSpare(tc);
1573	<	success = true;
3143	>	long startTime = System.nanoTime();
3144	>	WorkQueue[] ws;
3145	>	int r = 0, m;
3146	>	boolean found = true;
3147	>	while (!isQuiescent() && (ws = workQueues) != null &&
3148	>	(m = ws.length - 1) >= 0) {
3149	>	if (!found) {
3150	>	if ((System.nanoTime() - startTime) > nanos)
3151	>	return false;
3152	>	Thread.yield(); // cannot block
3153	>	}
3154	>	found = false;
3155	>	for (int j = (m + 1) << 2; j >= 0; --j) {
3156	>	ForkJoinTask<?> t; WorkQueue q; int b;
3157	>	if ((q = ws[r++ & m]) != null && (b = q.base) - q.top < 0) {
3158	>	found = true;
3159	>	if ((t = q.pollAt(b)) != null) {
3160	>	if (q.base - q.top < 0)
3161	>	signalWork(q);
3162	>	t.doExec();
3163	>	}
3164		break;
3165		}
3166		}
1577	–	} finally {
1578	–	if (locked)
1579	–	lock.unlock();
1580	–	}
1581	–	return success;
1582	–	}
1583	–
1584	–	/**
1585	–	* Add the kth spare worker. On entry, pool coounts are already
1586	–	* adjusted to reflect addition.
1587	–	*/
1588	–	private void createAndStartSpare(int k) {
1589	–	ForkJoinWorkerThread w = null;
1590	–	ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(k + 1);
1591	–	int len = ws.length;
1592	–	// Probably, we can place at slot k. If not, find empty slot
1593	–	if (k < len && ws[k] != null) {
1594	–	for (k = 0; k < len && ws[k] != null; ++k)
1595	–	;
3167		}
3168	<	if (k < len && !isTerminating() && (w = createWorker(k)) != null) {
1598	<	ws[k] = w;
1599	<	w.start();
1600	<	}
1601	<	else
1602	<	updateWorkerCount(-1); // adjust on failure
1603	<	signalIdleWorkers();
3168	>	return true;
3169		}
3170
3171		/**
3172	<	* Suspend calling thread w if there are excess threads.  Called
3173	<	* only from sync.  Spares are enqueued in a Treiber stack
1609	<	* using the same WaitQueueNodes as barriers.  They are resumed
1610	<	* mainly in preJoin, but are also woken on pool events that
1611	<	* require all threads to check run state.
1612	<	* @param w the caller
3172	>	* Waits and/or attempts to assist performing tasks indefinitely
3173	>	* until the {@link #commonPool()} {@link #isQuiescent}.
3174		*/
3175	<	private boolean suspendIfSpare(ForkJoinWorkerThread w) {
3176	<	WaitQueueNode node = null;
1616	<	int s;
1617	<	while (parallelism < runningCountOf(s = workerCounts)) {
1618	<	if (node == null)
1619	<	node = new WaitQueueNode(0, w);
1620	<	if (casWorkerCounts(s, s-1)) { // representation-dependent
1621	<	// push onto stack
1622	<	do;while (!casSpareStack(node.next = spareStack, node));
1623	<	// block until released by resumeSpare
1624	<	node.awaitSpareRelease();
1625	<	return true;
1626	<	}
1627	<	}
1628	<	return false;
1629	<	}
1630	<
1631	<	/**
1632	<	* Try to pop and resume a spare thread.
1633	<	* @param updateCount if true, increment running count on success
1634	<	* @return true if successful
1635	<	*/
1636	<	private boolean tryResumeSpare(boolean updateCount) {
1637	<	WaitQueueNode q;
1638	<	while ((q = spareStack) != null) {
1639	<	if (casSpareStack(q, q.next)) {
1640	<	if (updateCount)
1641	<	updateRunningCount(1);
1642	<	q.signal();
1643	<	return true;
1644	<	}
1645	<	}
1646	<	return false;
1647	<	}
1648	<
1649	<	/**
1650	<	* Pop and resume all spare threads. Same idea as ensureSync.
1651	<	* @return true if any spares released
1652	<	*/
1653	<	private boolean resumeAllSpares() {
1654	<	WaitQueueNode q;
1655	<	while ( (q = spareStack) != null) {
1656	<	if (casSpareStack(q, null)) {
1657	<	do {
1658	<	updateRunningCount(1);
1659	<	q.signal();
1660	<	} while ((q = q.next) != null);
1661	<	return true;
1662	<	}
1663	<	}
1664	<	return false;
1665	<	}
1666	<
1667	<	/**
1668	<	* Pop and shutdown excessive spare threads. Call only while
1669	<	* holding lock. This is not guaranteed to eliminate all excess
1670	<	* threads, only those suspended as spares, which are the ones
1671	<	* unlikely to be needed in the future.
1672	<	*/
1673	<	private void trimSpares() {
1674	<	int surplus = totalCountOf(workerCounts) - parallelism;
1675	<	WaitQueueNode q;
1676	<	while (surplus > 0 && (q = spareStack) != null) {
1677	<	if (casSpareStack(q, null)) {
1678	<	do {
1679	<	updateRunningCount(1);
1680	<	ForkJoinWorkerThread w = q.thread;
1681	<	if (w != null && surplus > 0 &&
1682	<	runningCountOf(workerCounts) > 0 && w.shutdown())
1683	<	--surplus;
1684	<	q.signal();
1685	<	} while ((q = q.next) != null);
1686	<	}
1687	<	}
3175	>	static void quiesceCommonPool() {
3176	>	common.awaitQuiescence(Long.MAX_VALUE, TimeUnit.NANOSECONDS);
3177		}
3178
3179		/**
3180		* Interface for extending managed parallelism for tasks running
3181	<	* in ForkJoinPools. A ManagedBlocker provides two methods.
3182	<	* Method <code>isReleasable</code> must return true if blocking is not
3183	<	* necessary. Method <code>block</code> blocks the current thread
3184	<	* if necessary (perhaps internally invoking isReleasable before
3185	<	* actually blocking.).
3181	>	* in {@link ForkJoinPool}s.
3182	>	*
3183	>	* <p>A {@code ManagedBlocker} provides two methods.  Method
3184	>	* {@code isReleasable} must return {@code true} if blocking is
3185	>	* not necessary. Method {@code block} blocks the current thread
3186	>	* if necessary (perhaps internally invoking {@code isReleasable}
3187	>	* before actually blocking). These actions are performed by any
3188	>	* thread invoking {@link ForkJoinPool#managedBlock}.  The
3189	>	* unusual methods in this API accommodate synchronizers that may,
3190	>	* but don't usually, block for long periods. Similarly, they
3191	>	* allow more efficient internal handling of cases in which
3192	>	* additional workers may be, but usually are not, needed to
3193	>	* ensure sufficient parallelism.  Toward this end,
3194	>	* implementations of method {@code isReleasable} must be amenable
3195	>	* to repeated invocation.
3196	>	*
3197		* <p>For example, here is a ManagedBlocker based on a
3198		* ReentrantLock:
3199	<	* <pre>
3200	<	*   class ManagedLocker implements ManagedBlocker {
3201	<	*     final ReentrantLock lock;
3202	<	*     boolean hasLock = false;
3203	<	*     ManagedLocker(ReentrantLock lock) { this.lock = lock; }
3204	<	*     public boolean block() {
3205	<	*        if (!hasLock)
3206	<	*           lock.lock();
3207	<	*        return true;
3208	<	*     }
3209	<	*     public boolean isReleasable() {
3210	<	*        return hasLock || (hasLock = lock.tryLock());
3211	<	*     }
3199	>	*  <pre> {@code
3200	>	* class ManagedLocker implements ManagedBlocker {
3201	>	*   final ReentrantLock lock;
3202	>	*   boolean hasLock = false;
3203	>	*   ManagedLocker(ReentrantLock lock) { this.lock = lock; }
3204	>	*   public boolean block() {
3205	>	*     if (!hasLock)
3206	>	*       lock.lock();
3207	>	*     return true;
3208	>	*   }
3209	>	*   public boolean isReleasable() {
3210	>	*     return hasLock || (hasLock = lock.tryLock());
3211	>	*   }
3212	>	* }}</pre>
3213	>	*
3214	>	* <p>Here is a class that possibly blocks waiting for an
3215	>	* item on a given queue:
3216	>	*  <pre> {@code
3217	>	* class QueueTaker<E> implements ManagedBlocker {
3218	>	*   final BlockingQueue<E> queue;
3219	>	*   volatile E item = null;
3220	>	*   QueueTaker(BlockingQueue<E> q) { this.queue = q; }
3221	>	*   public boolean block() throws InterruptedException {
3222	>	*     if (item == null)
3223	>	*       item = queue.take();
3224	>	*     return true;
3225	>	*   }
3226	>	*   public boolean isReleasable() {
3227	>	*     return item != null || (item = queue.poll()) != null;
3228		*   }
3229	<	* </pre>
3229	>	*   public E getItem() { // call after pool.managedBlock completes
3230	>	*     return item;
3231	>	*   }
3232	>	* }}</pre>
3233		*/
3234		public static interface ManagedBlocker {
3235		/**
3236		* Possibly blocks the current thread, for example waiting for
3237		* a lock or condition.
3238	<	* @return true if no additional blocking is necessary (i.e.,
3239	<	* if isReleasable would return true).
3238	>	*
3239	>	* @return {@code true} if no additional blocking is necessary
3240	>	* (i.e., if isReleasable would return true)
3241		* @throws InterruptedException if interrupted while waiting
3242	<	* (the method is not required to do so, but is allowe to).
3242	>	* (the method is not required to do so, but is allowed to)
3243		*/
3244		boolean block() throws InterruptedException;
3245
3246		/**
3247	<	* Returns true if blocking is unnecessary.
3247	>	* Returns {@code true} if blocking is unnecessary.
3248		*/
3249		boolean isReleasable();
3250		}
3251
3252		/**
3253		* Blocks in accord with the given blocker.  If the current thread
3254	<	* is a ForkJoinWorkerThread, this method possibly arranges for a
3255	<	* spare thread to be activated if necessary to ensure parallelism
3256	<	* while the current thread is blocked.  If
3257	<	* <code>maintainParallelism</code> is true and the pool supports
3258	<	* it ({@link #getMaintainsParallelism}), this method attempts to
3259	<	* maintain the pool's nominal parallelism. Otherwise if activates
3260	<	* a thread only if necessary to avoid complete starvation. This
3261	<	* option may be preferable when blockages use timeouts, or are
3262	<	* almost always brief.
3263	<	*
3264	<	* <p> If the caller is not a ForkJoinTask, this method is behaviorally
3265	<	* equivalent to
3266	<	* <pre>
3267	<	*   while (!blocker.isReleasable())
1748	<	*      if (blocker.block())
1749	<	*         return;
1750	<	* </pre>
1751	<	* If the caller is a ForkJoinTask, then the pool may first
1752	<	* be expanded to ensure parallelism, and later adjusted.
3254	>	* is a {@link ForkJoinWorkerThread}, this method possibly
3255	>	* arranges for a spare thread to be activated if necessary to
3256	>	* ensure sufficient parallelism while the current thread is blocked.
3257	>	*
3258	>	* <p>If the caller is not a {@link ForkJoinTask}, this method is
3259	>	* behaviorally equivalent to
3260	>	*  <pre> {@code
3261	>	* while (!blocker.isReleasable())
3262	>	*   if (blocker.block())
3263	>	*     return;
3264	>	* }</pre>
3265	>	*
3266	>	* If the caller is a {@code ForkJoinTask}, then the pool may
3267	>	* first be expanded to ensure parallelism, and later adjusted.
3268		*
3269		* @param blocker the blocker
3270	<	* @param maintainParallelism if true and supported by this pool,
1756	<	* attempt to maintain the pool's nominal parallelism; otherwise
1757	<	* activate a thread only if necessary to avoid complete
1758	<	* starvation.
1759	<	* @throws InterruptedException if blocker.block did so.
3270	>	* @throws InterruptedException if blocker.block did so
3271		*/
3272	<	public static void managedBlock(ManagedBlocker blocker,
1762	<	boolean maintainParallelism)
3272	>	public static void managedBlock(ManagedBlocker blocker)
3273		throws InterruptedException {
3274		Thread t = Thread.currentThread();
3275	<	ForkJoinPool pool = (t instanceof ForkJoinWorkerThread?
3276	<	((ForkJoinWorkerThread)t).pool : null);
3277	<	if (!blocker.isReleasable()) {
3278	<	try {
3279	<	if (pool == null ||
3280	<	!pool.preBlock(blocker, maintainParallelism))
3281	<	awaitBlocker(blocker);
3282	<	} finally {
3283	<	if (pool != null)
3284	<	pool.updateRunningCount(1);
3275	>	if (t instanceof ForkJoinWorkerThread) {
3276	>	ForkJoinPool p = ((ForkJoinWorkerThread)t).pool;
3277	>	while (!blocker.isReleasable()) { // variant of helpSignal
3278	>	WorkQueue[] ws; WorkQueue q; int m, u;
3279	>	if ((ws = p.workQueues) != null && (m = ws.length - 1) >= 0) {
3280	>	for (int i = 0; i <= m; ++i) {
3281	>	if (blocker.isReleasable())
3282	>	return;
3283	>	if ((q = ws[i]) != null && q.base - q.top < 0) {
3284	>	p.signalWork(q);
3285	>	if ((u = (int)(p.ctl >>> 32)) >= 0 ||
3286	>	(u >> UAC_SHIFT) >= 0)
3287	>	break;
3288	>	}
3289	>	}
3290	>	}
3291	>	if (p.tryCompensate()) {
3292	>	try {
3293	>	do {} while (!blocker.isReleasable() &&
3294	>	!blocker.block());
3295	>	} finally {
3296	>	p.incrementActiveCount();
3297	>	}
3298	>	break;
3299	>	}
3300		}
3301		}
3302	+	else {
3303	+	do {} while (!blocker.isReleasable() &&
3304	+	!blocker.block());
3305	+	}
3306		}
3307
3308	<	private static void awaitBlocker(ManagedBlocker blocker)
3309	<	throws InterruptedException {
3310	<	do;while (!blocker.isReleasable() && !blocker.block());
1782	<	}
1783	<
1784	<	// AbstractExecutorService overrides
3308	>	// AbstractExecutorService overrides.  These rely on undocumented
3309	>	// fact that ForkJoinTask.adapt returns ForkJoinTasks that also
3310	>	// implement RunnableFuture.
3311
3312		protected <T> RunnableFuture<T> newTaskFor(Runnable runnable, T value) {
3313	<	return new AdaptedRunnable(runnable, value);
3313	>	return new ForkJoinTask.AdaptedRunnable<T>(runnable, value);
3314		}
3315
3316		protected <T> RunnableFuture<T> newTaskFor(Callable<T> callable) {
3317	<	return new AdaptedCallable(callable);
3317	>	return new ForkJoinTask.AdaptedCallable<T>(callable);
3318		}
3319
3320	+	// Unsafe mechanics
3321	+	private static final sun.misc.Unsafe U;
3322	+	private static final long CTL;
3323	+	private static final long PARKBLOCKER;
3324	+	private static final int ABASE;
3325	+	private static final int ASHIFT;
3326	+	private static final long STEALCOUNT;
3327	+	private static final long PLOCK;
3328	+	private static final long INDEXSEED;
3329	+	private static final long QLOCK;
3330
3331	<	// Temporary Unsafe mechanics for preliminary release
3332	<	private static Unsafe getUnsafe() throws Throwable {
3331	>	static {
3332	>	// initialize field offsets for CAS etc
3333		try {
3334	<	return Unsafe.getUnsafe();
3335	<	} catch (SecurityException se) {
3336	<	try {
3337	<	return java.security.AccessController.doPrivileged
3338	<	(new java.security.PrivilegedExceptionAction<Unsafe>() {
3339	<	public Unsafe run() throws Exception {
3340	<	return getUnsafePrivileged();
3341	<	}});
3342	<	} catch (java.security.PrivilegedActionException e) {
3343	<	throw e.getCause();
3344	<	}
3345	<	}
3346	<	}
3347	<
3348	<	private static Unsafe getUnsafePrivileged()
3349	<	throws NoSuchFieldException, IllegalAccessException {
3350	<	Field f = Unsafe.class.getDeclaredField("theUnsafe");
3351	<	f.setAccessible(true);
3352	<	return (Unsafe) f.get(null);
3353	<	}
3334	>	U = getUnsafe();
3335	>	Class<?> k = ForkJoinPool.class;
3336	>	CTL = U.objectFieldOffset
3337	>	(k.getDeclaredField("ctl"));
3338	>	STEALCOUNT = U.objectFieldOffset
3339	>	(k.getDeclaredField("stealCount"));
3340	>	PLOCK = U.objectFieldOffset
3341	>	(k.getDeclaredField("plock"));
3342	>	INDEXSEED = U.objectFieldOffset
3343	>	(k.getDeclaredField("indexSeed"));
3344	>	Class<?> tk = Thread.class;
3345	>	PARKBLOCKER = U.objectFieldOffset
3346	>	(tk.getDeclaredField("parkBlocker"));
3347	>	Class<?> wk = WorkQueue.class;
3348	>	QLOCK = U.objectFieldOffset
3349	>	(wk.getDeclaredField("qlock"));
3350	>	Class<?> ak = ForkJoinTask[].class;
3351	>	ABASE = U.arrayBaseOffset(ak);
3352	>	int scale = U.arrayIndexScale(ak);
3353	>	if ((scale & (scale - 1)) != 0)
3354	>	throw new Error("data type scale not a power of two");
3355	>	ASHIFT = 31 - Integer.numberOfLeadingZeros(scale);
3356	>	} catch (Exception e) {
3357	>	throw new Error(e);
3358	>	}
3359	>
3360	>	submitters = new ThreadLocal<Submitter>();
3361	>	ForkJoinWorkerThreadFactory fac = defaultForkJoinWorkerThreadFactory =
3362	>	new DefaultForkJoinWorkerThreadFactory();
3363	>	modifyThreadPermission = new RuntimePermission("modifyThread");
3364	>
3365	>	/*
3366	>	* Establish common pool parameters.  For extra caution,
3367	>	* computations to set up common pool state are here; the
3368	>	* constructor just assigns these values to fields.
3369	>	*/
3370
3371	<	private static long fieldOffset(String fieldName)
3372	<	throws NoSuchFieldException {
3373	<	return _unsafe.objectFieldOffset
3374	<	(ForkJoinPool.class.getDeclaredField(fieldName));
3371	>	int par = 0;
3372	>	Thread.UncaughtExceptionHandler handler = null;
3373	>	try {  // TBD: limit or report ignored exceptions?
3374	>	String pp = System.getProperty
3375	>	("java.util.concurrent.ForkJoinPool.common.parallelism");
3376	>	String hp = System.getProperty
3377	>	("java.util.concurrent.ForkJoinPool.common.exceptionHandler");
3378	>	String fp = System.getProperty
3379	>	("java.util.concurrent.ForkJoinPool.common.threadFactory");
3380	>	if (fp != null)
3381	>	fac = ((ForkJoinWorkerThreadFactory)ClassLoader.
3382	>	getSystemClassLoader().loadClass(fp).newInstance());
3383	>	if (hp != null)
3384	>	handler = ((Thread.UncaughtExceptionHandler)ClassLoader.
3385	>	getSystemClassLoader().loadClass(hp).newInstance());
3386	>	if (pp != null)
3387	>	par = Integer.parseInt(pp);
3388	>	} catch (Exception ignore) {
3389	>	}
3390	>
3391	>	if (par <= 0)
3392	>	par = Runtime.getRuntime().availableProcessors();
3393	>	if (par > MAX_CAP)
3394	>	par = MAX_CAP;
3395	>	commonParallelism = par;
3396	>	long np = (long)(-par); // precompute initial ctl value
3397	>	long ct = ((np << AC_SHIFT) & AC_MASK) | ((np << TC_SHIFT) & TC_MASK);
3398	>
3399	>	common = new ForkJoinPool(par, ct, fac, handler);
3400		}
3401
3402	<	static final Unsafe _unsafe;
3403	<	static final long eventCountOffset;
3404	<	static final long workerCountsOffset;
3405	<	static final long runControlOffset;
3406	<	static final long syncStackOffset;
3407	<	static final long spareStackOffset;
3408	<
3409	<	static {
3402	>	/**
3403	>	* Returns a sun.misc.Unsafe.  Suitable for use in a 3rd party package.
3404	>	* Replace with a simple call to Unsafe.getUnsafe when integrating
3405	>	* into a jdk.
3406	>	*
3407	>	* @return a sun.misc.Unsafe
3408	>	*/
3409	>	private static sun.misc.Unsafe getUnsafe() {
3410	>	try {
3411	>	return sun.misc.Unsafe.getUnsafe();
3412	>	} catch (SecurityException tryReflectionInstead) {}
3413		try {
3414	<	_unsafe = getUnsafe();
3415	<	eventCountOffset = fieldOffset("eventCount");
3416	<	workerCountsOffset = fieldOffset("workerCounts");
3417	<	runControlOffset = fieldOffset("runControl");
3418	<	syncStackOffset = fieldOffset("syncStack");
3419	<	spareStackOffset = fieldOffset("spareStack");
3420	<	} catch (Throwable e) {
3421	<	throw new RuntimeException("Could not initialize intrinsics", e);
3414	>	return java.security.AccessController.doPrivileged
3415	>	(new java.security.PrivilegedExceptionAction<sun.misc.Unsafe>() {
3416	>	public sun.misc.Unsafe run() throws Exception {
3417	>	Class<sun.misc.Unsafe> k = sun.misc.Unsafe.class;
3418	>	for (java.lang.reflect.Field f : k.getDeclaredFields()) {
3419	>	f.setAccessible(true);
3420	>	Object x = f.get(null);
3421	>	if (k.isInstance(x))
3422	>	return k.cast(x);
3423	>	}
3424	>	throw new NoSuchFieldError("the Unsafe");
3425	>	}});
3426	>	} catch (java.security.PrivilegedActionException e) {
3427	>	throw new RuntimeException("Could not initialize intrinsics",
3428	>	e.getCause());
3429		}
3430		}
1844	–
1845	–	private boolean casEventCount(long cmp, long val) {
1846	–	return _unsafe.compareAndSwapLong(this, eventCountOffset, cmp, val);
1847	–	}
1848	–	private boolean casWorkerCounts(int cmp, int val) {
1849	–	return _unsafe.compareAndSwapInt(this, workerCountsOffset, cmp, val);
1850	–	}
1851	–	private boolean casRunControl(int cmp, int val) {
1852	–	return _unsafe.compareAndSwapInt(this, runControlOffset, cmp, val);
1853	–	}
1854	–	private boolean casSpareStack(WaitQueueNode cmp, WaitQueueNode val) {
1855	–	return _unsafe.compareAndSwapObject(this, spareStackOffset, cmp, val);
1856	–	}
1857	–	private boolean casBarrierStack(WaitQueueNode cmp, WaitQueueNode val) {
1858	–	return _unsafe.compareAndSwapObject(this, syncStackOffset, cmp, val);
1859	–	}
3431		}

Diff Legend

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