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 1.5 by jsr166, Thu Mar 19 05:10:42 2009 UTC vs.
Revision 1.155 by jsr166, Sat Dec 8 20:49:24 2012 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. Otherwise, other
31	<	* ExecutorService implementations are typically more appropriate
32	<	* 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 dynamically thread construction can be limited using methods
56	<	* <code>setMaximumPoolSize</code> and/or
44	<	* <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 commonPool 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 */
497	<	private static final int  shortMask = 0xffff;
496	>	// Static utilities
497	>
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	<	/** Max pool size -- must be a power of two minus 1 */
69	<	private static final int MAX_THREADS =  0x7FFF;
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	>	* @throw 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	+	if (base - top < 0) {       // process remaining local tasks
991	+	if (mode == 0)
992	+	popAndExecAll();
993	+	else
994	+	pollAndExecAll();
995	+	}
996	+	++nsteals;
997	+	hint = -1;
998		}
999		}
1000	+
1001	+	/**
1002	+	* Executes a non-top-level (stolen) task.
1003	+	*/
1004	+	final void runSubtask(ForkJoinTask<?> t) {
1005	+	if (t != null) {
1006	+	ForkJoinTask<?> ps = currentSteal;
1007	+	(currentSteal = t).doExec();
1008	+	currentSteal = ps;
1009	+	}
1010	+	}
1011	+
1012	+	/**
1013	+	* Returns true if owned and not known to be blocked.
1014	+	*/
1015	+	final boolean isApparentlyUnblocked() {
1016	+	Thread wt; Thread.State s;
1017	+	return (eventCount >= 0 &&
1018	+	(wt = owner) != null &&
1019	+	(s = wt.getState()) != Thread.State.BLOCKED &&
1020	+	s != Thread.State.WAITING &&
1021	+	s != Thread.State.TIMED_WAITING);
1022	+	}
1023	+
1024	+	/**
1025	+	* If this owned and is not already interrupted, try to
1026	+	* interrupt and/or unpark, ignoring exceptions.
1027	+	*/
1028	+	final void interruptOwner() {
1029	+	Thread wt, p;
1030	+	if ((wt = owner) != null && !wt.isInterrupted()) {
1031	+	try {
1032	+	wt.interrupt();
1033	+	} catch (SecurityException ignore) {
1034	+	}
1035	+	}
1036	+	if ((p = parker) != null)
1037	+	U.unpark(p);
1038	+	}
1039	+
1040	+	// Unsafe mechanics
1041	+	private static final sun.misc.Unsafe U;
1042	+	private static final long QLOCK;
1043	+	private static final int ABASE;
1044	+	private static final int ASHIFT;
1045	+	static {
1046	+	int s;
1047	+	try {
1048	+	U = getUnsafe();
1049	+	Class<?> k = WorkQueue.class;
1050	+	Class<?> ak = ForkJoinTask[].class;
1051	+	QLOCK = U.objectFieldOffset
1052	+	(k.getDeclaredField("qlock"));
1053	+	ABASE = U.arrayBaseOffset(ak);
1054	+	s = U.arrayIndexScale(ak);
1055	+	} catch (Exception e) {
1056	+	throw new Error(e);
1057	+	}
1058	+	if ((s & (s-1)) != 0)
1059	+	throw new Error("data type scale not a power of two");
1060	+	ASHIFT = 31 - Integer.numberOfLeadingZeros(s);
1061	+	}
1062		}
1063
1064	+	// static fields (initialized in static initializer below)
1065	+
1066		/**
1067		* Creates a new ForkJoinWorkerThread. This factory is used unless
1068		* overridden in ForkJoinPool constructors.
1069		*/
1070		public static final ForkJoinWorkerThreadFactory
1071	<	defaultForkJoinWorkerThreadFactory =
108	<	new DefaultForkJoinWorkerThreadFactory();
1071	>	defaultForkJoinWorkerThreadFactory;
1072
1073		/**
1074	<	* Permission required for callers of methods that may start or
1075	<	* kill threads.
1074	>	* Per-thread submission bookkeeping. Shared across all pools
1075	>	* to reduce ThreadLocal pollution and because random motion
1076	>	* to avoid contention in one pool is likely to hold for others.
1077	>	* Lazily initialized on first submission (but null-checked
1078	>	* in other contexts to avoid unnecessary initialization).
1079		*/
1080	<	private static final RuntimePermission modifyThreadPermission =
115	<	new RuntimePermission("modifyThread");
1080	>	static final ThreadLocal<Submitter> submitters;
1081
1082		/**
1083	<	* If there is a security manager, makes sure caller has
1084	<	* permission to modify threads.
1083	>	* Permission required for callers of methods that may start or
1084	>	* kill threads.
1085		*/
1086	<	private static void checkPermission() {
122	<	SecurityManager security = System.getSecurityManager();
123	<	if (security != null)
124	<	security.checkPermission(modifyThreadPermission);
125	<	}
1086	>	private static final RuntimePermission modifyThreadPermission;
1087
1088		/**
1089	<	* Generator for assigning sequence numbers as pool names.
1089	>	* Common (static) pool. Non-null for public use unless a static
1090	>	* construction exception, but internal usages null-check on use
1091	>	* to paranoically avoid potential initialization circularities
1092	>	* as well as to simplify generated code.
1093		*/
1094	<	private static final AtomicInteger poolNumberGenerator =
131	<	new AtomicInteger();
1094	>	static final ForkJoinPool commonPool;
1095
1096		/**
1097	<	* Array holding all worker threads in the pool. Array size must
135	<	* be a power of two.  Updates and replacements are protected by
136	<	* workerLock, but it is always kept in a consistent enough state
137	<	* to be randomly accessed without locking by workers performing
138	<	* work-stealing.
1097	>	* Common pool parallelism. Must equal commonPool.parallelism.
1098		*/
1099	<	volatile ForkJoinWorkerThread[] workers;
1099	>	static final int commonPoolParallelism;
1100
1101		/**
1102	<	* Lock protecting access to workers.
1102	>	* Sequence number for creating workerNamePrefix.
1103		*/
1104	<	private final ReentrantLock workerLock;
1104	>	private static int poolNumberSequence;
1105
1106		/**
1107	<	* Condition for awaitTermination.
1107	>	* Return the next sequence number. We don't expect this to
1108	>	* ever contend so use simple builtin sync.
1109		*/
1110	<	private final Condition termination;
1110	>	private static final synchronized int nextPoolId() {
1111	>	return ++poolNumberSequence;
1112	>	}
1113	>
1114	>	// static constants
1115
1116		/**
1117	<	* The uncaught exception handler used when any worker
1118	<	* abrupty terminates
1117	>	* Initial timeout value (in nanoseconds) for the thread
1118	>	* triggering quiescence to park waiting for new work. On timeout,
1119	>	* the thread will instead try to shrink the number of
1120	>	* workers. The value should be large enough to avoid overly
1121	>	* aggressive shrinkage during most transient stalls (long GCs
1122	>	* etc).
1123		*/
1124	<	private Thread.UncaughtExceptionHandler ueh;
1124	>	private static final long IDLE_TIMEOUT      = 2000L * 1000L * 1000L; // 2sec
1125
1126		/**
1127	<	* Creation factory for worker threads.
1127	>	* Timeout value when there are more threads than parallelism level
1128		*/
1129	<	private final ForkJoinWorkerThreadFactory factory;
1129	>	private static final long FAST_IDLE_TIMEOUT =  200L * 1000L * 1000L;
1130
1131		/**
1132	<	* Head of stack of threads that were created to maintain
165	<	* parallelism when other threads blocked, but have since
166	<	* suspended when the parallelism level rose.
1132	>	* Tolerance for idle timeouts, to cope with timer undershoots
1133		*/
1134	<	private volatile WaitQueueNode spareStack;
1134	>	private static final long TIMEOUT_SLOP = 2000000L; // 20ms
1135
1136		/**
1137	<	* Sum of per-thread steal counts, updated only when threads are
1138	<	* idle or terminating.
1137	>	* The maximum stolen->joining link depth allowed in method
1138	>	* tryHelpStealer.  Must be a power of two.  Depths for legitimate
1139	>	* chains are unbounded, but we use a fixed constant to avoid
1140	>	* (otherwise unchecked) cycles and to bound staleness of
1141	>	* traversal parameters at the expense of sometimes blocking when
1142	>	* we could be helping.
1143		*/
1144	<	private final AtomicLong stealCount;
1144	>	private static final int MAX_HELP = 64;
1145
1146		/**
1147	<	* Queue for external submissions.
1147	>	* Increment for seed generators. See class ThreadLocal for
1148	>	* explanation.
1149		*/
1150	<	private final LinkedTransferQueue<ForkJoinTask<?>> submissionQueue;
1150	>	private static final int SEED_INCREMENT = 0x61c88647;
1151
1152		/**
1153	<	* Head of Treiber stack for barrier sync. See below for explanation
1153	>	* Bits and masks for control variables
1154	>	*
1155	>	* Field ctl is a long packed with:
1156	>	* AC: Number of active running workers minus target parallelism (16 bits)
1157	>	* TC: Number of total workers minus target parallelism (16 bits)
1158	>	* ST: true if pool is terminating (1 bit)
1159	>	* EC: the wait count of top waiting thread (15 bits)
1160	>	* ID: poolIndex of top of Treiber stack of waiters (16 bits)
1161	>	*
1162	>	* When convenient, we can extract the upper 32 bits of counts and
1163	>	* the lower 32 bits of queue state, u = (int)(ctl >>> 32) and e =
1164	>	* (int)ctl.  The ec field is never accessed alone, but always
1165	>	* together with id and st. The offsets of counts by the target
1166	>	* parallelism and the positionings of fields makes it possible to
1167	>	* perform the most common checks via sign tests of fields: When
1168	>	* ac is negative, there are not enough active workers, when tc is
1169	>	* negative, there are not enough total workers, and when e is
1170	>	* negative, the pool is terminating.  To deal with these possibly
1171	>	* negative fields, we use casts in and out of "short" and/or
1172	>	* signed shifts to maintain signedness.
1173	>	*
1174	>	* When a thread is queued (inactivated), its eventCount field is
1175	>	* set negative, which is the only way to tell if a worker is
1176	>	* prevented from executing tasks, even though it must continue to
1177	>	* scan for them to avoid queuing races. Note however that
1178	>	* eventCount updates lag releases so usage requires care.
1179	>	*
1180	>	* Field plock is an int packed with:
1181	>	* SHUTDOWN: true if shutdown is enabled (1 bit)
1182	>	* SEQ:  a sequence lock, with PL_LOCK bit set if locked (30 bits)
1183	>	* SIGNAL: set when threads may be waiting on the lock (1 bit)
1184	>	*
1185	>	* The sequence number enables simple consistency checks:
1186	>	* Staleness of read-only operations on the workQueues array can
1187	>	* be checked by comparing plock before vs after the reads.
1188	>	*/
1189	>
1190	>	// bit positions/shifts for fields
1191	>	private static final int  AC_SHIFT   = 48;
1192	>	private static final int  TC_SHIFT   = 32;
1193	>	private static final int  ST_SHIFT   = 31;
1194	>	private static final int  EC_SHIFT   = 16;
1195	>
1196	>	// bounds
1197	>	private static final int  SMASK      = 0xffff;  // short bits
1198	>	private static final int  MAX_CAP    = 0x7fff;  // max #workers - 1
1199	>	private static final int  EVENMASK   = 0xfffe;  // even short bits
1200	>	private static final int  SQMASK     = 0x007e;  // max 64 (even) slots
1201	>	private static final int  SHORT_SIGN = 1 << 15;
1202	>	private static final int  INT_SIGN   = 1 << 31;
1203	>
1204	>	// masks
1205	>	private static final long STOP_BIT   = 0x0001L << ST_SHIFT;
1206	>	private static final long AC_MASK    = ((long)SMASK) << AC_SHIFT;
1207	>	private static final long TC_MASK    = ((long)SMASK) << TC_SHIFT;
1208	>
1209	>	// units for incrementing and decrementing
1210	>	private static final long TC_UNIT    = 1L << TC_SHIFT;
1211	>	private static final long AC_UNIT    = 1L << AC_SHIFT;
1212	>
1213	>	// masks and units for dealing with u = (int)(ctl >>> 32)
1214	>	private static final int  UAC_SHIFT  = AC_SHIFT - 32;
1215	>	private static final int  UTC_SHIFT  = TC_SHIFT - 32;
1216	>	private static final int  UAC_MASK   = SMASK << UAC_SHIFT;
1217	>	private static final int  UTC_MASK   = SMASK << UTC_SHIFT;
1218	>	private static final int  UAC_UNIT   = 1 << UAC_SHIFT;
1219	>	private static final int  UTC_UNIT   = 1 << UTC_SHIFT;
1220	>
1221	>	// masks and units for dealing with e = (int)ctl
1222	>	private static final int E_MASK      = 0x7fffffff; // no STOP_BIT
1223	>	private static final int E_SEQ       = 1 << EC_SHIFT;
1224	>
1225	>	// plock bits
1226	>	private static final int SHUTDOWN    = 1 << 31;
1227	>	private static final int PL_LOCK     = 2;
1228	>	private static final int PL_SIGNAL   = 1;
1229	>	private static final int PL_SPINS    = 1 << 8;
1230	>
1231	>	// access mode for WorkQueue
1232	>	static final int LIFO_QUEUE          =  0;
1233	>	static final int FIFO_QUEUE          =  1;
1234	>	static final int SHARED_QUEUE        = -1;
1235	>
1236	>	// bounds for #steps in scan loop -- must be power 2 minus 1
1237	>	private static final int MIN_SCAN    = 0x1ff;   // cover estimation slop
1238	>	private static final int MAX_SCAN    = 0x1ffff; // 4 * max workers
1239	>
1240	>	// Instance fields
1241	>
1242	>	/*
1243	>	* Field layout of this class tends to matter more than one would
1244	>	* like. Runtime layout order is only loosely related to
1245	>	* declaration order and may differ across JVMs, but the following
1246	>	* empirically works OK on current JVMs.
1247	>	*/
1248	>
1249	>	// Heuristic padding to ameliorate unfortunate memory placements
1250	>	volatile long pad00, pad01, pad02, pad03, pad04, pad05, pad06;
1251	>
1252	>	volatile long stealCount;                  // collects worker counts
1253	>	volatile long ctl;                         // main pool control
1254	>	volatile int plock;                        // shutdown status and seqLock
1255	>	volatile int indexSeed;                    // worker/submitter index seed
1256	>	final int config;                          // mode and parallelism level
1257	>	WorkQueue[] workQueues;                    // main registry
1258	>	final ForkJoinWorkerThreadFactory factory;
1259	>	final Thread.UncaughtExceptionHandler ueh; // per-worker UEH
1260	>	final String workerNamePrefix;             // to create worker name string
1261	>
1262	>	volatile Object pad10, pad11, pad12, pad13, pad14, pad15, pad16, pad17;
1263	>	volatile Object pad18, pad19, pad1a, pad1b;
1264	>
1265	>	/*
1266	>	* Acquires the plock lock to protect worker array and related
1267	>	* updates. This method is called only if an initial CAS on plock
1268	>	* fails. This acts as a spinLock for normal cases, but falls back
1269	>	* to builtin monitor to block when (rarely) needed. This would be
1270	>	* a terrible idea for a highly contended lock, but works fine as
1271	>	* a more conservative alternative to a pure spinlock.  See
1272	>	* internal ConcurrentHashMap documentation for further
1273	>	* explanation of nearly the same construction.
1274		*/
1275	<	private volatile WaitQueueNode syncStack;
1275	>	private int acquirePlock() {
1276	>	int spins = PL_SPINS, r = 0, ps, nps;
1277	>	for (;;) {
1278	>	if (((ps = plock) & PL_LOCK) == 0 &&
1279	>	U.compareAndSwapInt(this, PLOCK, ps, nps = ps + PL_LOCK))
1280	>	return nps;
1281	>	else if (r == 0) { // randomize spins if possible
1282	>	Thread t = Thread.currentThread(); WorkQueue w; Submitter z;
1283	>	if ((t instanceof ForkJoinWorkerThread) &&
1284	>	(w = ((ForkJoinWorkerThread)t).workQueue) != null)
1285	>	r = w.seed;
1286	>	else if ((z = submitters.get()) != null)
1287	>	r = z.seed;
1288	>	else
1289	>	r = 1;
1290	>	}
1291	>	else if (spins >= 0) {
1292	>	r ^= r << 1; r ^= r >>> 3; r ^= r << 10; // xorshift
1293	>	if (r >= 0)
1294	>	--spins;
1295	>	}
1296	>	else if (U.compareAndSwapInt(this, PLOCK, ps, ps | PL_SIGNAL)) {
1297	>	synchronized (this) {
1298	>	if ((plock & PL_SIGNAL) != 0) {
1299	>	try {
1300	>	wait();
1301	>	} catch (InterruptedException ie) {
1302	>	try {
1303	>	Thread.currentThread().interrupt();
1304	>	} catch (SecurityException ignore) {
1305	>	}
1306	>	}
1307	>	}
1308	>	else
1309	>	notifyAll();
1310	>	}
1311	>	}
1312	>	}
1313	>	}
1314
1315		/**
1316	<	* The count for event barrier
1316	>	* Unlocks and signals any thread waiting for plock. Called only
1317	>	* when CAS of seq value for unlock fails.
1318		*/
1319	<	private volatile long eventCount;
1319	>	private void releasePlock(int ps) {
1320	>	plock = ps;
1321	>	synchronized (this) { notifyAll(); }
1322	>	}
1323
1324		/**
1325	<	* Pool number, just for assigning useful names to worker threads
1325	>	* Performs secondary initialization, called when plock is zero.
1326	>	* Creates workQueue array and sets plock to a valid value.  The
1327	>	* lock body must be exception-free (so no try/finally) so we
1328	>	* optimistically allocate new array outside the lock and throw
1329	>	* away if (very rarely) not needed. (A similar tactic is used in
1330	>	* fullExternalPush.)  Because the plock seq value can eventually
1331	>	* wrap around zero, this method harmlessly fails to reinitialize
1332	>	* if workQueues exists, while still advancing plock.
1333	>	*
1334	>	* Additionally tries to create the first worker.
1335		*/
1336	<	private final int poolNumber;
1336	>	private void initWorkers() {
1337	>	WorkQueue[] ws, nws; int ps;
1338	>	int p = config & SMASK;        // find power of two table size
1339	>	int n = (p > 1) ? p - 1 : 1;   // ensure at least 2 slots
1340	>	n |= n >>> 1; n |= n >>> 2; n |= n >>> 4; n |= n >>> 8; n |= n >>> 16;
1341	>	n = (n + 1) << 1;
1342	>	if ((ws = workQueues) == null || ws.length == 0)
1343	>	nws = new WorkQueue[n];
1344	>	else
1345	>	nws = null;
1346	>	if (((ps = plock) & PL_LOCK) != 0 ||
1347	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
1348	>	ps = acquirePlock();
1349	>	if (((ws = workQueues) == null || ws.length == 0) && nws != null)
1350	>	workQueues = nws;
1351	>	int nps = (ps & SHUTDOWN) | ((ps + PL_LOCK) & ~SHUTDOWN);
1352	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
1353	>	releasePlock(nps);
1354	>	tryAddWorker();
1355	>	}
1356
1357		/**
1358	<	* The maximum allowed pool size
1359	<	*/
1360	<	private volatile int maxPoolSize;
1358	>	* Tries to create and start one worker if fewer than target
1359	>	* parallelism level exist. Adjusts counts etc on failure.
1360	>	*/
1361	>	private void tryAddWorker() {
1362	>	long c; int u;
1363	>	while ((u = (int)((c = ctl) >>> 32)) < 0 &&
1364	>	(u & SHORT_SIGN) != 0 && (int)c == 0) {
1365	>	long nc = (long)(((u + UTC_UNIT) & UTC_MASK) |
1366	>	((u + UAC_UNIT) & UAC_MASK)) << 32;
1367	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1368	>	ForkJoinWorkerThreadFactory fac;
1369	>	Throwable ex = null;
1370	>	ForkJoinWorkerThread wt = null;
1371	>	try {
1372	>	if ((fac = factory) != null &&
1373	>	(wt = fac.newThread(this)) != null) {
1374	>	wt.start();
1375	>	break;
1376	>	}
1377	>	} catch (Throwable e) {
1378	>	ex = e;
1379	>	}
1380	>	deregisterWorker(wt, ex);
1381	>	break;
1382	>	}
1383	>	}
1384	>	}
1385	>
1386	>	//  Registering and deregistering workers
1387
1388		/**
1389	<	* The desired parallelism level, updated only under workerLock.
1389	>	* Callback from ForkJoinWorkerThread to establish and record its
1390	>	* WorkQueue. To avoid scanning bias due to packing entries in
1391	>	* front of the workQueues array, we treat the array as a simple
1392	>	* power-of-two hash table using per-thread seed as hash,
1393	>	* expanding as needed.
1394	>	*
1395	>	* @param wt the worker thread
1396	>	* @return the worker's queue
1397		*/
1398	<	private volatile int parallelism;
1398	>	final WorkQueue registerWorker(ForkJoinWorkerThread wt) {
1399	>	Thread.UncaughtExceptionHandler handler; WorkQueue[] ws; int s, ps;
1400	>	wt.setDaemon(true);
1401	>	if ((handler = ueh) != null)
1402	>	wt.setUncaughtExceptionHandler(handler);
1403	>	do {} while (!U.compareAndSwapInt(this, INDEXSEED, s = indexSeed,
1404	>	s += SEED_INCREMENT) ||
1405	>	s == 0); // skip 0
1406	>	WorkQueue w = new WorkQueue(this, wt, config >>> 16, s);
1407	>	if (((ps = plock) & PL_LOCK) != 0 ||
1408	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
1409	>	ps = acquirePlock();
1410	>	int nps = (ps & SHUTDOWN) | ((ps + PL_LOCK) & ~SHUTDOWN);
1411	>	try {
1412	>	if ((ws = workQueues) != null) {    // skip if shutting down
1413	>	int n = ws.length, m = n - 1;
1414	>	int r = (s << 1) | 1;           // use odd-numbered indices
1415	>	if (ws[r &= m] != null) {       // collision
1416	>	int probes = 0;             // step by approx half size
1417	>	int step = (n <= 4) ? 2 : ((n >>> 1) & EVENMASK) + 2;
1418	>	while (ws[r = (r + step) & m] != null) {
1419	>	if (++probes >= n) {
1420	>	workQueues = ws = Arrays.copyOf(ws, n <<= 1);
1421	>	m = n - 1;
1422	>	probes = 0;
1423	>	}
1424	>	}
1425	>	}
1426	>	w.eventCount = w.poolIndex = r; // volatile write orders
1427	>	ws[r] = w;
1428	>	}
1429	>	} finally {
1430	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
1431	>	releasePlock(nps);
1432	>	}
1433	>	wt.setName(workerNamePrefix.concat(Integer.toString(w.poolIndex)));
1434	>	return w;
1435	>	}
1436
1437		/**
1438	<	* Holds number of total (i.e., created and not yet terminated)
1439	<	* and running (i.e., not blocked on joins or other managed sync)
1440	<	* threads, packed into one int to ensure consistent snapshot when
1441	<	* making decisions about creating and suspending spare
1442	<	* threads. Updated only by CAS.  Note: CASes in
1443	<	* updateRunningCount and preJoin running active count is in low
1444	<	* word, so need to be modified if this changes
1438	>	* Final callback from terminating worker, as well as upon failure
1439	>	* to construct or start a worker.  Removes record of worker from
1440	>	* array, and adjusts counts. If pool is shutting down, tries to
1441	>	* complete termination.
1442	>	*
1443	>	* @param wt the worker thread or null if construction failed
1444	>	* @param ex the exception causing failure, or null if none
1445		*/
1446	<	private volatile int workerCounts;
1446	>	final void deregisterWorker(ForkJoinWorkerThread wt, Throwable ex) {
1447	>	WorkQueue w = null;
1448	>	if (wt != null && (w = wt.workQueue) != null) {
1449	>	int ps;
1450	>	w.qlock = -1;                // ensure set
1451	>	long ns = w.nsteals, sc;     // collect steal count
1452	>	do {} while (!U.compareAndSwapLong(this, STEALCOUNT,
1453	>	sc = stealCount, sc + ns));
1454	>	if (((ps = plock) & PL_LOCK) != 0 ||
1455	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
1456	>	ps = acquirePlock();
1457	>	int nps = (ps & SHUTDOWN) | ((ps + PL_LOCK) & ~SHUTDOWN);
1458	>	try {
1459	>	int idx = w.poolIndex;
1460	>	WorkQueue[] ws = workQueues;
1461	>	if (ws != null && idx >= 0 && idx < ws.length && ws[idx] == w)
1462	>	ws[idx] = null;
1463	>	} finally {
1464	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
1465	>	releasePlock(nps);
1466	>	}
1467	>	}
1468	>
1469	>	long c;                             // adjust ctl counts
1470	>	do {} while (!U.compareAndSwapLong
1471	>	(this, CTL, c = ctl, (((c - AC_UNIT) & AC_MASK) |
1472	>	((c - TC_UNIT) & TC_MASK) |
1473	>	(c & ~(AC_MASK|TC_MASK)))));
1474	>
1475	>	if (!tryTerminate(false, false) && w != null && w.array != null) {
1476	>	w.cancelAll();                  // cancel remaining tasks
1477	>	int e, u, i, n; WorkQueue[] ws; WorkQueue v; Thread p;
1478	>	while ((u = (int)((c = ctl) >>> 32)) < 0) {
1479	>	if ((e = (int)c) > 0) {     // activate or create replacement
1480	>	if ((ws = workQueues) != null &&
1481	>	ws.length > (i = e & SMASK) &&
1482	>	(v = ws[i]) != null && v.eventCount == (e | INT_SIGN)) {
1483	>	long nc = (((long)(v.nextWait & E_MASK)) |
1484	>	((long)(u + UAC_UNIT) << 32));
1485	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1486	>	v.eventCount = (e + E_SEQ) & E_MASK;
1487	>	if ((p = v.parker) != null)
1488	>	U.unpark(p);
1489	>	break;
1490	>	}
1491	>	}
1492	>	else
1493	>	break;
1494	>	}
1495	>	else {
1496	>	if ((short)u < 0)
1497	>	tryAddWorker();
1498	>	break;
1499	>	}
1500	>	}
1501	>	}
1502	>	if (ex == null)                     // help clean refs on way out
1503	>	ForkJoinTask.helpExpungeStaleExceptions();
1504	>	else                                // rethrow
1505	>	ForkJoinTask.rethrow(ex);
1506	>	}
1507
1508	<	private static int totalCountOf(int s)           { return s >>> 16;  }
218	<	private static int runningCountOf(int s)         { return s & shortMask; }
219	<	private static int workerCountsFor(int t, int r) { return (t << 16) + r; }
1508	>	// Submissions
1509
1510		/**
1511	<	* Add delta (which may be negative) to running count.  This must
1512	<	* be called before (with negative arg) and after (with positive)
1513	<	* any managed synchronization (i.e., mainly, joins)
1514	<	* @param delta the number to add
1511	>	* Unless shutting down, adds the given task to a submission queue
1512	>	* at submitter's current queue index (modulo submission
1513	>	* range). Only the most common path is directly handled in this
1514	>	* method. All others are relayed to fullExternalPush.
1515	>	*
1516	>	* @param task the task. Caller must ensure non-null.
1517		*/
1518	<	final void updateRunningCount(int delta) {
1519	<	int s;
1520	<	do;while (!casWorkerCounts(s = workerCounts, s + delta));
1518	>	final void externalPush(ForkJoinTask<?> task) {
1519	>	WorkQueue[] ws; WorkQueue q; Submitter z; int m; ForkJoinTask<?>[] a;
1520	>	if ((z = submitters.get()) != null && plock > 0 &&
1521	>	(ws = workQueues) != null && (m = (ws.length - 1)) >= 0 &&
1522	>	(q = ws[m & z.seed & SQMASK]) != null &&
1523	>	U.compareAndSwapInt(q, QLOCK, 0, 1)) { // lock
1524	>	int b = q.base, s = q.top, n, an;
1525	>	if ((a = q.array) != null && (an = a.length) > (n = s + 1 - b)) {
1526	>	int j = (((an - 1) & s) << ASHIFT) + ABASE;
1527	>	U.putOrderedObject(a, j, task);
1528	>	q.top = s + 1;                     // push on to deque
1529	>	q.qlock = 0;
1530	>	if (n <= 2)
1531	>	signalWork(q);
1532	>	return;
1533	>	}
1534	>	q.qlock = 0;
1535	>	}
1536	>	fullExternalPush(task);
1537	>	}
1538	>
1539	>	/**
1540	>	* Full version of externalPush. This method is called, among
1541	>	* other times, upon the first submission of the first task to the
1542	>	* pool, so must perform secondary initialization (via
1543	>	* initWorkers). It also detects first submission by an external
1544	>	* thread by looking up its ThreadLocal, and creates a new shared
1545	>	* queue if the one at index if empty or contended. The plock lock
1546	>	* body must be exception-free (so no try/finally) so we
1547	>	* optimistically allocate new queues outside the lock and throw
1548	>	* them away if (very rarely) not needed.
1549	>	*/
1550	>	private void fullExternalPush(ForkJoinTask<?> task) {
1551	>	int r = 0; // random index seed
1552	>	for (Submitter z = submitters.get();;) {
1553	>	WorkQueue[] ws; WorkQueue q; int ps, m, k;
1554	>	if (z == null) {
1555	>	if (U.compareAndSwapInt(this, INDEXSEED, r = indexSeed,
1556	>	r += SEED_INCREMENT) && r != 0)
1557	>	submitters.set(z = new Submitter(r));
1558	>	}
1559	>	else if (r == 0) {               // move to a different index
1560	>	r = z.seed;
1561	>	r ^= r << 13;                // same xorshift as WorkQueues
1562	>	r ^= r >>> 17;
1563	>	z.seed = r ^ (r << 5);
1564	>	}
1565	>	else if ((ps = plock) < 0)
1566	>	throw new RejectedExecutionException();
1567	>	else if (ps == 0 || (ws = workQueues) == null ||
1568	>	(m = ws.length - 1) < 0)
1569	>	initWorkers();
1570	>	else if ((q = ws[k = r & m & SQMASK]) != null) {
1571	>	if (q.qlock == 0 && U.compareAndSwapInt(q, QLOCK, 0, 1)) {
1572	>	ForkJoinTask<?>[] a = q.array;
1573	>	int s = q.top;
1574	>	boolean submitted = false;
1575	>	try {                      // locked version of push
1576	>	if ((a != null && a.length > s + 1 - q.base) ||
1577	>	(a = q.growArray()) != null) {   // must presize
1578	>	int j = (((a.length - 1) & s) << ASHIFT) + ABASE;
1579	>	U.putOrderedObject(a, j, task);
1580	>	q.top = s + 1;
1581	>	submitted = true;
1582	>	}
1583	>	} finally {
1584	>	q.qlock = 0;  // unlock
1585	>	}
1586	>	if (submitted) {
1587	>	signalWork(q);
1588	>	return;
1589	>	}
1590	>	}
1591	>	r = 0; // move on failure
1592	>	}
1593	>	else if (((ps = plock) & PL_LOCK) == 0) { // create new queue
1594	>	q = new WorkQueue(this, null, SHARED_QUEUE, r);
1595	>	if (((ps = plock) & PL_LOCK) != 0 ||
1596	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
1597	>	ps = acquirePlock();
1598	>	if ((ws = workQueues) != null && k < ws.length && ws[k] == null)
1599	>	ws[k] = q;
1600	>	int nps = (ps & SHUTDOWN) | ((ps + PL_LOCK) & ~SHUTDOWN);
1601	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
1602	>	releasePlock(nps);
1603	>	}
1604	>	else
1605	>	r = 0; // try elsewhere while lock held
1606	>	}
1607		}
1608
1609	+	// Maintaining ctl counts
1610	+
1611		/**
1612	<	* Add delta (which may be negative) to both total and running
234	<	* count.  This must be called upon creation and termination of
235	<	* worker threads.
236	<	* @param delta the number to add
1612	>	* Increments active count; mainly called upon return from blocking.
1613		*/
1614	<	private void updateWorkerCount(int delta) {
1615	<	int d = delta + (delta << 16); // add to both lo and hi parts
1616	<	int s;
241	<	do;while (!casWorkerCounts(s = workerCounts, s + d));
1614	>	final void incrementActiveCount() {
1615	>	long c;
1616	>	do {} while (!U.compareAndSwapLong(this, CTL, c = ctl, c + AC_UNIT));
1617		}
1618
1619		/**
1620	<	* Lifecycle control. High word contains runState, low word
1621	<	* contains the number of workers that are (probably) executing
1622	<	* tasks. This value is atomically incremented before a worker
248	<	* gets a task to run, and decremented when worker has no tasks
249	<	* and cannot find any. These two fields are bundled together to
250	<	* support correct termination triggering.  Note: activeCount
251	<	* CAS'es cheat by assuming active count is in low word, so need
252	<	* to be modified if this changes
1620	>	* Tries to create or activate a worker if too few are active.
1621	>	*
1622	>	* @param q the (non-null) queue holding tasks to be signalled
1623		*/
1624	<	private volatile int runControl;
1625	<
1626	<	// RunState values. Order among values matters
1627	<	private static final int RUNNING     = 0;
1628	<	private static final int SHUTDOWN    = 1;
1629	<	private static final int TERMINATING = 2;
1630	<	private static final int TERMINATED  = 3;
1624	>	final void signalWork(WorkQueue q) {
1625	>	int hint = q.poolIndex;
1626	>	long c; int e, u, i, n; WorkQueue[] ws; WorkQueue w; Thread p;
1627	>	while ((u = (int)((c = ctl) >>> 32)) < 0) {
1628	>	if ((e = (int)c) > 0) {
1629	>	if ((ws = workQueues) != null && ws.length > (i = e & SMASK) &&
1630	>	(w = ws[i]) != null && w.eventCount == (e | INT_SIGN)) {
1631	>	long nc = (((long)(w.nextWait & E_MASK)) |
1632	>	((long)(u + UAC_UNIT) << 32));
1633	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1634	>	w.hint = hint;
1635	>	w.eventCount = (e + E_SEQ) & E_MASK;
1636	>	if ((p = w.parker) != null)
1637	>	U.unpark(p);
1638	>	break;
1639	>	}
1640	>	if (q.top - q.base <= 0)
1641	>	break;
1642	>	}
1643	>	else
1644	>	break;
1645	>	}
1646	>	else {
1647	>	if ((short)u < 0)
1648	>	tryAddWorker();
1649	>	break;
1650	>	}
1651	>	}
1652	>	}
1653
1654	<	private static int runStateOf(int c)             { return c >>> 16; }
263	<	private static int activeCountOf(int c)          { return c & shortMask; }
264	<	private static int runControlFor(int r, int a)   { return (r << 16) + a; }
1654	>	// Scanning for tasks
1655
1656		/**
1657	<	* Try incrementing active count; fail on contention. Called by
268	<	* workers before/during executing tasks.
269	<	* @return true on success;
1657	>	* Top-level runloop for workers, called by ForkJoinWorkerThread.run.
1658		*/
1659	<	final boolean tryIncrementActiveCount() {
1660	<	int c = runControl;
1661	<	return casRunControl(c, c+1);
1659	>	final void runWorker(WorkQueue w) {
1660	>	w.growArray(); // allocate queue
1661	>	do { w.runTask(scan(w)); } while (w.qlock >= 0);
1662		}
1663
1664		/**
1665	<	* Try decrementing active count; fail on contention.
1666	<	* Possibly trigger termination on success
1667	<	* Called by workers when they can't find tasks.
1668	<	* @return true on success
1665	>	* Scans for and, if found, returns one task, else possibly
1666	>	* inactivates the worker. This method operates on single reads of
1667	>	* volatile state and is designed to be re-invoked continuously,
1668	>	* in part because it returns upon detecting inconsistencies,
1669	>	* contention, or state changes that indicate possible success on
1670	>	* re-invocation.
1671	>	*
1672	>	* The scan searches for tasks across queues (starting at a random
1673	>	* index, and relying on registerWorker to irregularly scatter
1674	>	* them within array to avoid bias), checking each at least twice.
1675	>	* The scan terminates upon either finding a non-empty queue, or
1676	>	* completing the sweep. If the worker is not inactivated, it
1677	>	* takes and returns a task from this queue. Otherwise, if not
1678	>	* activated, it signals workers (that may include itself) and
1679	>	* returns so caller can retry. Also returns for true if the
1680	>	* worker array may have changed during an empty scan.  On failure
1681	>	* to find a task, we take one of the following actions, after
1682	>	* which the caller will retry calling this method unless
1683	>	* terminated.
1684	>	*
1685	>	* * If pool is terminating, terminate the worker.
1686	>	*
1687	>	* * If not already enqueued, try to inactivate and enqueue the
1688	>	* worker on wait queue. Or, if inactivating has caused the pool
1689	>	* to be quiescent, relay to idleAwaitWork to check for
1690	>	* termination and possibly shrink pool.
1691	>	*
1692	>	* * If already enqueued and none of the above apply, possibly
1693	>	* (with 1/2 probability) park awaiting signal, else lingering to
1694	>	* help scan and signal.
1695	>	*
1696	>	* @param w the worker (via its WorkQueue)
1697	>	* @return a task or null if none found
1698		*/
1699	<	final boolean tryDecrementActiveCount() {
1700	<	int c = runControl;
1701	<	int nextc = c - 1;
1702	<	if (!casRunControl(c, nextc))
1703	<	return false;
1704	<	if (canTerminateOnShutdown(nextc))
1705	<	terminateOnShutdown();
1706	<	return true;
1699	>	private final ForkJoinTask<?> scan(WorkQueue w) {
1700	>	WorkQueue[] ws; int m;
1701	>	int ps = plock;                          // read plock before ws
1702	>	if (w != null && (ws = workQueues) != null && (m = ws.length - 1) >= 0) {
1703	>	int ec = w.eventCount;               // ec is negative if inactive
1704	>	int r = w.seed; r ^= r << 13; r ^= r >>> 17; w.seed = r ^= r << 5;
1705	>	int j = ((m + m + 1) | MIN_SCAN) & MAX_SCAN;
1706	>	do {
1707	>	WorkQueue q; ForkJoinTask<?>[] a; int b;
1708	>	if ((q = ws[(r + j) & m]) != null && (b = q.base) - q.top < 0 &&
1709	>	(a = q.array) != null) {     // probably nonempty
1710	>	int i = (((a.length - 1) & b) << ASHIFT) + ABASE;
1711	>	ForkJoinTask<?> t = (ForkJoinTask<?>)
1712	>	U.getObjectVolatile(a, i);
1713	>	if (q.base == b && ec >= 0 && t != null &&
1714	>	U.compareAndSwapObject(a, i, t, null)) {
1715	>	if ((q.base = b + 1) - q.top < 0)
1716	>	signalWork(q);
1717	>	return t;                // taken
1718	>	}
1719	>	else if ((ec < 0 || j < m) && (int)(ctl >> AC_SHIFT) <= 0) {
1720	>	w.hint = (r + j) & m;    // help signal below
1721	>	break;                   // cannot take
1722	>	}
1723	>	}
1724	>	} while (--j >= 0);
1725	>
1726	>	long c, sc; int e, ns, h;
1727	>	if ((h = w.hint) < 0) {
1728	>	if ((ns = w.nsteals) != 0) {
1729	>	if (U.compareAndSwapLong(this, STEALCOUNT,
1730	>	sc = stealCount, sc + ns))
1731	>	w.nsteals = 0;           // collect steals
1732	>	}
1733	>	else if (plock != ps)            // consistency check
1734	>	;                            // skip
1735	>	else if ((e = (int)(c = ctl)) < 0)
1736	>	w.qlock = -1;                // pool is terminating
1737	>	else if (ec >= 0) {              // try to enqueue/inactivate
1738	>	long nc = ((long)ec | ((c - AC_UNIT) & (AC_MASK|TC_MASK)));
1739	>	w.nextWait = e;              // link and mark inactive
1740	>	w.eventCount = ec | INT_SIGN;
1741	>	if (ctl != c || !U.compareAndSwapLong(this, CTL, c, nc))
1742	>	w.eventCount = ec;       // unmark on CAS failure
1743	>	else if ((int)(c >> AC_SHIFT) == 1 - (config & SMASK))
1744	>	idleAwaitWork(w, nc, c);
1745	>	}
1746	>	else if (w.eventCount < 0) {     // block
1747	>	Thread wt = Thread.currentThread();
1748	>	Thread.interrupted();        // clear status
1749	>	U.putObject(wt, PARKBLOCKER, this);
1750	>	w.parker = wt;               // emulate LockSupport.park
1751	>	if (w.eventCount < 0)        // recheck
1752	>	U.park(false, 0L);
1753	>	w.parker = null;
1754	>	U.putObject(wt, PARKBLOCKER, null);
1755	>	}
1756	>	}
1757	>	if (h >= 0 || w.hint >= 0)           // signal others before retry
1758	>	helpSignalHint(w);
1759	>	}
1760	>	return null;
1761		}
1762
1763		/**
1764	<	* Return true if argument represents zero active count and
1765	<	* nonzero runstate, which is the triggering condition for
1766	<	* terminating on shutdown.
1767	<	*/
1768	<	private static boolean canTerminateOnShutdown(int c) {
1769	<	return ((c & -c) >>> 16) != 0; // i.e. least bit is nonzero runState bit
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	>	int dc = -(short)(currentCtl >>> TC_SHIFT);
1779	>	long parkTime = dc < 0 ? FAST_IDLE_TIMEOUT: (dc + 1) * IDLE_TIMEOUT;
1780	>	long deadline = System.nanoTime() + parkTime - TIMEOUT_SLOP;
1781	>	Thread wt = Thread.currentThread();
1782	>	while (ctl == currentCtl) {
1783	>	Thread.interrupted();  // timed variant of version in scan()
1784	>	U.putObject(wt, PARKBLOCKER, this);
1785	>	w.parker = wt;
1786	>	if (ctl == currentCtl)
1787	>	U.park(false, parkTime);
1788	>	w.parker = null;
1789	>	U.putObject(wt, PARKBLOCKER, null);
1790	>	if (ctl != currentCtl)
1791	>	break;
1792	>	if (deadline - System.nanoTime() <= 0L &&
1793	>	U.compareAndSwapLong(this, CTL, currentCtl, prevCtl)) {
1794	>	w.eventCount = (w.eventCount + E_SEQ) | E_MASK;
1795	>	w.qlock = -1;   // shrink
1796	>	break;
1797	>	}
1798	>	}
1799	>	}
1800		}
1801
1802		/**
1803	<	* Transition run state to at least the given state. Return true
1804	<	* if not already at least given state.
1803	>	* Scans through queues looking for work while joining a task; if
1804	>	* any present, signals. May return early if more signalling is
1805	>	* detectably unneeded.
1806	>	*
1807	>	* @param task return early if done
1808	>	* @param origin an index to start scan
1809		*/
1810	<	private boolean transitionRunStateTo(int state) {
1811	<	for (;;) {
1812	<	int c = runControl;
1813	<	if (runStateOf(c) >= state)
1814	<	return false;
1815	<	if (casRunControl(c, runControlFor(state, activeCountOf(c))))
1816	<	return true;
1810	>	private void helpSignal(ForkJoinTask<?> task, int origin) {
1811	>	WorkQueue[] ws; WorkQueue w; Thread p; long c; int m, u, e, i, s;
1812	>	if (task != null && task.status >= 0 &&
1813	>	(u = (int)(ctl >>> 32)) < 0 && (u >> UAC_SHIFT) < 0 &&
1814	>	(ws = workQueues) != null && (m = ws.length - 1) >= 0) {
1815	>	outer: for (int k = origin, j = m; j >= 0; --j) {
1816	>	WorkQueue q = ws[k++ & m];
1817	>	for (int n = m;;) { // limit to at most m signals
1818	>	if (task.status < 0)
1819	>	break outer;
1820	>	if (q == null ||
1821	>	((s = -q.base + q.top) <= n && (n = s) <= 0))
1822	>	break;
1823	>	if ((u = (int)((c = ctl) >>> 32)) >= 0 ||
1824	>	(e = (int)c) <= 0 || m < (i = e & SMASK) ||
1825	>	(w = ws[i]) == null)
1826	>	break outer;
1827	>	long nc = (((long)(w.nextWait & E_MASK)) |
1828	>	((long)(u + UAC_UNIT) << 32));
1829	>	if (w.eventCount == (e | INT_SIGN) &&
1830	>	U.compareAndSwapLong(this, CTL, c, nc)) {
1831	>	w.eventCount = (e + E_SEQ) & E_MASK;
1832	>	if ((p = w.parker) != null)
1833	>	U.unpark(p);
1834	>	if (--n <= 0)
1835	>	break;
1836	>	}
1837	>	}
1838	>	}
1839		}
1840		}
1841
1842		/**
1843	<	* Controls whether to add spares to maintain parallelism
1843	>	* Signals other workers if tasks are present in hinted queue.
1844	>	*
1845	>	* @param caller the worker with the hint
1846		*/
1847	<	private volatile boolean maintainsParallelism;
1847	>	private void helpSignalHint(WorkQueue caller) {
1848	>	WorkQueue[] ws; WorkQueue q, w; Thread p; long c; int h, m, u, e, i, s;
1849	>	if (caller != null && (h = caller.hint) >= 0 &&
1850	>	(u = (int)(ctl >>> 32)) < 0 && (u >> UAC_SHIFT) < 0 &&
1851	>	(ws = workQueues) != null && (m = ws.length - 1) >= 0 &&
1852	>	(q = ws[h & m]) != null) {
1853	>	caller.hint = -1;
1854	>	for (int n = 2;;) { // limit to at most 2 signals
1855	>	int idleCount = (caller.eventCount < 0) ? 0 : -1;
1856	>	if (((s = idleCount - q.base + q.top) <= n && (n = s) <= 0) ||
1857	>	(u = (int)((c = ctl) >>> 32)) >= 0 ||
1858	>	(e = (int)c) <= 0 || m < (i = e & SMASK) ||
1859	>	(w = ws[i]) == null)
1860	>	break;
1861	>	long nc = (((long)(w.nextWait & E_MASK)) |
1862	>	((long)(u + UAC_UNIT) << 32));
1863	>	if (w.eventCount == (e | INT_SIGN) &&
1864	>	U.compareAndSwapLong(this, CTL, c, nc)) {
1865	>	w.hint = h;
1866	>	w.eventCount = (e + E_SEQ) & E_MASK;
1867	>	if ((p = w.parker) != null)
1868	>	U.unpark(p);
1869	>	if (--n <= 0)
1870	>	break;
1871	>	}
1872	>	}
1873	>	}
1874	>	}
1875
1876	<	// Constructors
1876	>	/**
1877	>	* Tries to locate and execute tasks for a stealer of the given
1878	>	* task, or in turn one of its stealers, Traces currentSteal ->
1879	>	* currentJoin links looking for a thread working on a descendant
1880	>	* of the given task and with a non-empty queue to steal back and
1881	>	* execute tasks from. The first call to this method upon a
1882	>	* waiting join will often entail scanning/search, (which is OK
1883	>	* because the joiner has nothing better to do), but this method
1884	>	* leaves hints in workers to speed up subsequent calls. The
1885	>	* implementation is very branchy to cope with potential
1886	>	* inconsistencies or loops encountering chains that are stale,
1887	>	* unknown, or so long that they are likely cyclic.
1888	>	*
1889	>	* @param joiner the joining worker
1890	>	* @param task the task to join
1891	>	* @return 0 if no progress can be made, negative if task
1892	>	* known complete, else positive
1893	>	*/
1894	>	private int tryHelpStealer(WorkQueue joiner, ForkJoinTask<?> task) {
1895	>	int stat = 0, steps = 0;                    // bound to avoid cycles
1896	>	if (joiner != null && task != null) {       // hoist null checks
1897	>	restart: for (;;) {
1898	>	ForkJoinTask<?> subtask = task;     // current target
1899	>	for (WorkQueue j = joiner, v;;) {   // v is stealer of subtask
1900	>	WorkQueue[] ws; int m, s, h;
1901	>	if ((s = task.status) < 0) {
1902	>	stat = s;
1903	>	break restart;
1904	>	}
1905	>	if ((ws = workQueues) == null || (m = ws.length - 1) <= 0)
1906	>	break restart;              // shutting down
1907	>	if ((v = ws[h = (j.hint | 1) & m]) == null ||
1908	>	v.currentSteal != subtask) {
1909	>	for (int origin = h;;) {    // find stealer
1910	>	if (((h = (h + 2) & m) & 15) == 1 &&
1911	>	(subtask.status < 0 || j.currentJoin != subtask))
1912	>	continue restart;   // occasional staleness check
1913	>	if ((v = ws[h]) != null &&
1914	>	v.currentSteal == subtask) {
1915	>	j.hint = h;        // save hint
1916	>	break;
1917	>	}
1918	>	if (h == origin)
1919	>	break restart;      // cannot find stealer
1920	>	}
1921	>	}
1922	>	for (;;) { // help stealer or descend to its stealer
1923	>	ForkJoinTask[] a;  int b;
1924	>	if (subtask.status < 0)     // surround probes with
1925	>	continue restart;       //   consistency checks
1926	>	if ((b = v.base) - v.top < 0 && (a = v.array) != null) {
1927	>	int i = (((a.length - 1) & b) << ASHIFT) + ABASE;
1928	>	ForkJoinTask<?> t =
1929	>	(ForkJoinTask<?>)U.getObjectVolatile(a, i);
1930	>	if (subtask.status < 0 || j.currentJoin != subtask ||
1931	>	v.currentSteal != subtask)
1932	>	continue restart;   // stale
1933	>	stat = 1;               // apparent progress
1934	>	if (t != null && v.base == b &&
1935	>	U.compareAndSwapObject(a, i, t, null)) {
1936	>	v.base = b + 1;     // help stealer
1937	>	joiner.runSubtask(t);
1938	>	}
1939	>	else if (v.base == b && ++steps == MAX_HELP)
1940	>	break restart;      // v apparently stalled
1941	>	}
1942	>	else {                      // empty -- try to descend
1943	>	ForkJoinTask<?> next = v.currentJoin;
1944	>	if (subtask.status < 0 || j.currentJoin != subtask ||
1945	>	v.currentSteal != subtask)
1946	>	continue restart;   // stale
1947	>	else if (next == null || ++steps == MAX_HELP)
1948	>	break restart;      // dead-end or maybe cyclic
1949	>	else {
1950	>	subtask = next;
1951	>	j = v;
1952	>	break;
1953	>	}
1954	>	}
1955	>	}
1956	>	}
1957	>	}
1958	>	}
1959	>	return stat;
1960	>	}
1961
1962		/**
1963	<	* Creates a ForkJoinPool with a pool size equal to the number of
1964	<	* processors available on the system and using the default
1965	<	* ForkJoinWorkerThreadFactory,
1966	<	* @throws SecurityException if a security manager exists and
1967	<	*         the caller is not permitted to modify threads
1968	<	*         because it does not hold {@link
1969	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
1963	>	* Analog of tryHelpStealer for CountedCompleters. Tries to steal
1964	>	* and run tasks within the target's computation.
1965	>	*
1966	>	* @param task the task to join
1967	>	* @param mode if shared, exit upon completing any task
1968	>	* if all workers are active
1969	>	*
1970		*/
1971	<	public ForkJoinPool() {
1972	<	this(Runtime.getRuntime().availableProcessors(),
1973	<	defaultForkJoinWorkerThreadFactory);
1971	>	private int helpComplete(ForkJoinTask<?> task, int mode) {
1972	>	WorkQueue[] ws; WorkQueue q; int m, n, s, u;
1973	>	if (task != null && (ws = workQueues) != null &&
1974	>	(m = ws.length - 1) >= 0) {
1975	>	for (int j = 1, origin = j;;) {
1976	>	if ((s = task.status) < 0)
1977	>	return s;
1978	>	if ((q = ws[j & m]) != null && q.pollAndExecCC(task)) {
1979	>	origin = j;
1980	>	if (mode == SHARED_QUEUE &&
1981	>	((u = (int)(ctl >>> 32)) >= 0 || (u >> UAC_SHIFT) >= 0))
1982	>	break;
1983	>	}
1984	>	else if ((j = (j + 2) & m) == origin)
1985	>	break;
1986	>	}
1987	>	}
1988	>	return 0;
1989		}
1990
1991		/**
1992	<	* Creates a ForkJoinPool with the indicated parellelism level
1993	<	* threads, and using the default ForkJoinWorkerThreadFactory,
1994	<	* @param parallelism the number of worker threads
1995	<	* @throws IllegalArgumentException if parallelism less than or
1996	<	* equal to zero
1997	<	* @throws SecurityException if a security manager exists and
1998	<	*         the caller is not permitted to modify threads
1999	<	*         because it does not hold {@link
2000	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
2001	<	*/
2002	<	public ForkJoinPool(int parallelism) {
2003	<	this(parallelism, defaultForkJoinWorkerThreadFactory);
1992	>	* Tries to decrement active count (sometimes implicitly) and
1993	>	* possibly release or create a compensating worker in preparation
1994	>	* for blocking. Fails on contention or termination. Otherwise,
1995	>	* adds a new thread if no idle workers are available and pool
1996	>	* may become starved.
1997	>	*/
1998	>	final boolean tryCompensate() {
1999	>	int pc = config & SMASK, e, i, tc; long c;
2000	>	WorkQueue[] ws; WorkQueue w; Thread p;
2001	>	if ((ws = workQueues) != null && (e = (int)(c = ctl)) >= 0) {
2002	>	if (e != 0 && (i = e & SMASK) < ws.length &&
2003	>	(w = ws[i]) != null && w.eventCount == (e | INT_SIGN)) {
2004	>	long nc = ((long)(w.nextWait & E_MASK) |
2005	>	(c & (AC_MASK|TC_MASK)));
2006	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
2007	>	w.eventCount = (e + E_SEQ) & E_MASK;
2008	>	if ((p = w.parker) != null)
2009	>	U.unpark(p);
2010	>	return true;   // replace with idle worker
2011	>	}
2012	>	}
2013	>	else if ((tc = (short)(c >>> TC_SHIFT)) >= 0 &&
2014	>	(int)(c >> AC_SHIFT) + pc > 1) {
2015	>	long nc = ((c - AC_UNIT) & AC_MASK) | (c & ~AC_MASK);
2016	>	if (U.compareAndSwapLong(this, CTL, c, nc))
2017	>	return true;   // no compensation
2018	>	}
2019	>	else if (tc + pc < MAX_CAP) {
2020	>	long nc = ((c + TC_UNIT) & TC_MASK) | (c & ~TC_MASK);
2021	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
2022	>	ForkJoinWorkerThreadFactory fac;
2023	>	Throwable ex = null;
2024	>	ForkJoinWorkerThread wt = null;
2025	>	try {
2026	>	if ((fac = factory) != null &&
2027	>	(wt = fac.newThread(this)) != null) {
2028	>	wt.start();
2029	>	return true;
2030	>	}
2031	>	} catch (Throwable rex) {
2032	>	ex = rex;
2033	>	}
2034	>	deregisterWorker(wt, ex); // clean up and return false
2035	>	}
2036	>	}
2037	>	}
2038	>	return false;
2039		}
2040
2041		/**
2042	<	* Creates a ForkJoinPool with parallelism equal to the number of
2043	<	* processors available on the system and using the given
2044	<	* ForkJoinWorkerThreadFactory,
2045	<	* @param factory the factory for creating new threads
2046	<	* @throws NullPointerException if factory is null
357	<	* @throws SecurityException if a security manager exists and
358	<	*         the caller is not permitted to modify threads
359	<	*         because it does not hold {@link
360	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
2042	>	* Helps and/or blocks until the given task is done.
2043	>	*
2044	>	* @param joiner the joining worker
2045	>	* @param task the task
2046	>	* @return task status on exit
2047		*/
2048	<	public ForkJoinPool(ForkJoinWorkerThreadFactory factory) {
2049	<	this(Runtime.getRuntime().availableProcessors(), factory);
2048	>	final int awaitJoin(WorkQueue joiner, ForkJoinTask<?> task) {
2049	>	int s = 0;
2050	>	if (joiner != null && task != null && (s = task.status) >= 0) {
2051	>	ForkJoinTask<?> prevJoin = joiner.currentJoin;
2052	>	joiner.currentJoin = task;
2053	>	do {} while ((s = task.status) >= 0 && !joiner.isEmpty() &&
2054	>	joiner.tryRemoveAndExec(task)); // process local tasks
2055	>	if (s >= 0 && (s = task.status) >= 0) {
2056	>	helpSignal(task, joiner.poolIndex);
2057	>	if ((s = task.status) >= 0 &&
2058	>	(task instanceof CountedCompleter))
2059	>	s = helpComplete(task, LIFO_QUEUE);
2060	>	}
2061	>	while (s >= 0 && (s = task.status) >= 0) {
2062	>	if ((!joiner.isEmpty() ||           // try helping
2063	>	(s = tryHelpStealer(joiner, task)) == 0) &&
2064	>	(s = task.status) >= 0) {
2065	>	helpSignal(task, joiner.poolIndex);
2066	>	if ((s = task.status) >= 0 && tryCompensate()) {
2067	>	if (task.trySetSignal() && (s = task.status) >= 0) {
2068	>	synchronized (task) {
2069	>	if (task.status >= 0) {
2070	>	try {                // see ForkJoinTask
2071	>	task.wait();     //  for explanation
2072	>	} catch (InterruptedException ie) {
2073	>	}
2074	>	}
2075	>	else
2076	>	task.notifyAll();
2077	>	}
2078	>	}
2079	>	long c;                          // re-activate
2080	>	do {} while (!U.compareAndSwapLong
2081	>	(this, CTL, c = ctl, c + AC_UNIT));
2082	>	}
2083	>	}
2084	>	}
2085	>	joiner.currentJoin = prevJoin;
2086	>	}
2087	>	return s;
2088		}
2089
2090		/**
2091	<	* Creates a ForkJoinPool with the given parallelism and factory.
2091	>	* Stripped-down variant of awaitJoin used by timed joins. Tries
2092	>	* to help join only while there is continuous progress. (Caller
2093	>	* will then enter a timed wait.)
2094		*
2095	<	* @param parallelism the targeted number of worker threads
2096	<	* @param factory the factory for creating new threads
371	<	* @throws IllegalArgumentException if parallelism less than or
372	<	* equal to zero, or greater than implementation limit.
373	<	* @throws NullPointerException if factory is null
374	<	* @throws SecurityException if a security manager exists and
375	<	*         the caller is not permitted to modify threads
376	<	*         because it does not hold {@link
377	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
2095	>	* @param joiner the joining worker
2096	>	* @param task the task
2097		*/
2098	<	public ForkJoinPool(int parallelism, ForkJoinWorkerThreadFactory factory) {
2099	<	if (parallelism <= 0 || parallelism > MAX_THREADS)
2100	<	throw new IllegalArgumentException();
2101	<	if (factory == null)
2102	<	throw new NullPointerException();
2103	<	checkPermission();
2104	<	this.factory = factory;
2105	<	this.parallelism = parallelism;
2106	<	this.maxPoolSize = MAX_THREADS;
2107	<	this.maintainsParallelism = true;
2108	<	this.poolNumber = poolNumberGenerator.incrementAndGet();
2109	<	this.workerLock = new ReentrantLock();
2110	<	this.termination = workerLock.newCondition();
2111	<	this.stealCount = new AtomicLong();
2112	<	this.submissionQueue = new LinkedTransferQueue<ForkJoinTask<?>>();
2113	<	createAndStartInitialWorkers(parallelism);
2098	>	final void helpJoinOnce(WorkQueue joiner, ForkJoinTask<?> task) {
2099	>	int s;
2100	>	if (joiner != null && task != null && (s = task.status) >= 0) {
2101	>	ForkJoinTask<?> prevJoin = joiner.currentJoin;
2102	>	joiner.currentJoin = task;
2103	>	do {} while ((s = task.status) >= 0 && !joiner.isEmpty() &&
2104	>	joiner.tryRemoveAndExec(task));
2105	>	if (s >= 0 && (s = task.status) >= 0) {
2106	>	helpSignal(task, joiner.poolIndex);
2107	>	if ((s = task.status) >= 0 &&
2108	>	(task instanceof CountedCompleter))
2109	>	s = helpComplete(task, LIFO_QUEUE);
2110	>	}
2111	>	if (s >= 0 && joiner.isEmpty()) {
2112	>	do {} while (task.status >= 0 &&
2113	>	tryHelpStealer(joiner, task) > 0);
2114	>	}
2115	>	joiner.currentJoin = prevJoin;
2116	>	}
2117	>	}
2118	>
2119	>	/**
2120	>	* Returns a (probably) non-empty steal queue, if one is found
2121	>	* during a random, then cyclic scan, else null.  This method must
2122	>	* be retried by caller if, by the time it tries to use the queue,
2123	>	* it is empty.
2124	>	* @param r a (random) seed for scanning
2125	>	*/
2126	>	private WorkQueue findNonEmptyStealQueue(int r) {
2127	>	for (WorkQueue[] ws;;) {
2128	>	int ps = plock, m, n;
2129	>	if ((ws = workQueues) == null || (m = ws.length - 1) < 1)
2130	>	return null;
2131	>	for (int j = (m + 1) << 2; ;) {
2132	>	WorkQueue q = ws[(((r + j) << 1) | 1) & m];
2133	>	if (q != null && (n = q.base - q.top) < 0) {
2134	>	if (n < -1)
2135	>	signalWork(q);
2136	>	return q;
2137	>	}
2138	>	else if (--j < 0) {
2139	>	if (plock == ps)
2140	>	return null;
2141	>	break;
2142	>	}
2143	>	}
2144	>	}
2145		}
2146
2147		/**
2148	<	* Create new worker using factory.
2149	<	* @param index the index to assign worker
2150	<	* @return new worker, or null of factory failed
2151	<	*/
2152	<	private ForkJoinWorkerThread createWorker(int index) {
2153	<	Thread.UncaughtExceptionHandler h = ueh;
2154	<	ForkJoinWorkerThread w = factory.newThread(this);
2155	<	if (w != null) {
2156	<	w.poolIndex = index;
2157	<	w.setDaemon(true);
2158	<	w.setName("ForkJoinPool-" + poolNumber + "-worker-" + index);
2159	<	if (h != null)
2160	<	w.setUncaughtExceptionHandler(h);
2148	>	* Runs tasks until {@code isQuiescent()}. We piggyback on
2149	>	* active count ctl maintenance, but rather than blocking
2150	>	* when tasks cannot be found, we rescan until all others cannot
2151	>	* find tasks either.
2152	>	*/
2153	>	final void helpQuiescePool(WorkQueue w) {
2154	>	for (boolean active = true;;) {
2155	>	ForkJoinTask<?> localTask; // exhaust local queue
2156	>	while ((localTask = w.nextLocalTask()) != null)
2157	>	localTask.doExec();
2158	>	// Similar to loop in scan(), but ignoring submissions
2159	>	WorkQueue q = findNonEmptyStealQueue(w.nextSeed());
2160	>	if (q != null) {
2161	>	ForkJoinTask<?> t; int b;
2162	>	if (!active) {      // re-establish active count
2163	>	long c;
2164	>	active = true;
2165	>	do {} while (!U.compareAndSwapLong
2166	>	(this, CTL, c = ctl, c + AC_UNIT));
2167	>	}
2168	>	if ((b = q.base) - q.top < 0 && (t = q.pollAt(b)) != null)
2169	>	w.runSubtask(t);
2170	>	}
2171	>	else {
2172	>	long c;
2173	>	if (active) {       // decrement active count without queuing
2174	>	active = false;
2175	>	do {} while (!U.compareAndSwapLong
2176	>	(this, CTL, c = ctl, c -= AC_UNIT));
2177	>	}
2178	>	else
2179	>	c = ctl;        // re-increment on exit
2180	>	if ((int)(c >> AC_SHIFT) + (config & SMASK) == 0) {
2181	>	do {} while (!U.compareAndSwapLong
2182	>	(this, CTL, c = ctl, c + AC_UNIT));
2183	>	break;
2184	>	}
2185	>	}
2186		}
412	–	return w;
2187		}
2188
2189		/**
2190	<	* Return a good size for worker array given pool size.
2191	<	* Currently requires size to be a power of two.
2190	>	* Gets and removes a local or stolen task for the given worker.
2191	>	*
2192	>	* @return a task, if available
2193		*/
2194	<	private static int arraySizeFor(int ps) {
2195	<	return ps <= 1? 1 : (1 << (32 - Integer.numberOfLeadingZeros(ps-1)));
2194	>	final ForkJoinTask<?> nextTaskFor(WorkQueue w) {
2195	>	for (ForkJoinTask<?> t;;) {
2196	>	WorkQueue q; int b;
2197	>	if ((t = w.nextLocalTask()) != null)
2198	>	return t;
2199	>	if ((q = findNonEmptyStealQueue(w.nextSeed())) == null)
2200	>	return null;
2201	>	if ((b = q.base) - q.top < 0 && (t = q.pollAt(b)) != null)
2202	>	return t;
2203	>	}
2204		}
2205
2206		/**
2207	<	* Create or resize array if necessary to hold newLength
2208	<	* @return the array
2207	>	* Returns a cheap heuristic guide for task partitioning when
2208	>	* programmers, frameworks, tools, or languages have little or no
2209	>	* idea about task granularity.  In essence by offering this
2210	>	* method, we ask users only about tradeoffs in overhead vs
2211	>	* expected throughput and its variance, rather than how finely to
2212	>	* partition tasks.
2213	>	*
2214	>	* In a steady state strict (tree-structured) computation, each
2215	>	* thread makes available for stealing enough tasks for other
2216	>	* threads to remain active. Inductively, if all threads play by
2217	>	* the same rules, each thread should make available only a
2218	>	* constant number of tasks.
2219	>	*
2220	>	* The minimum useful constant is just 1. But using a value of 1
2221	>	* would require immediate replenishment upon each steal to
2222	>	* maintain enough tasks, which is infeasible.  Further,
2223	>	* partitionings/granularities of offered tasks should minimize
2224	>	* steal rates, which in general means that threads nearer the top
2225	>	* of computation tree should generate more than those nearer the
2226	>	* bottom. In perfect steady state, each thread is at
2227	>	* approximately the same level of computation tree. However,
2228	>	* producing extra tasks amortizes the uncertainty of progress and
2229	>	* diffusion assumptions.
2230	>	*
2231	>	* So, users will want to use values larger, but not much larger
2232	>	* than 1 to both smooth over transient shortages and hedge
2233	>	* against uneven progress; as traded off against the cost of
2234	>	* extra task overhead. We leave the user to pick a threshold
2235	>	* value to compare with the results of this call to guide
2236	>	* decisions, but recommend values such as 3.
2237	>	*
2238	>	* When all threads are active, it is on average OK to estimate
2239	>	* surplus strictly locally. In steady-state, if one thread is
2240	>	* maintaining say 2 surplus tasks, then so are others. So we can
2241	>	* just use estimated queue length.  However, this strategy alone
2242	>	* leads to serious mis-estimates in some non-steady-state
2243	>	* conditions (ramp-up, ramp-down, other stalls). We can detect
2244	>	* many of these by further considering the number of "idle"
2245	>	* threads, that are known to have zero queued tasks, so
2246	>	* compensate by a factor of (#idle/#active) threads.
2247	>	*
2248	>	* Note: The approximation of #busy workers as #active workers is
2249	>	* not very good under current signalling scheme, and should be
2250	>	* improved.
2251	>	*/
2252	>	static int getSurplusQueuedTaskCount() {
2253	>	Thread t; ForkJoinWorkerThread wt; ForkJoinPool pool; WorkQueue q;
2254	>	if (((t = Thread.currentThread()) instanceof ForkJoinWorkerThread)) {
2255	>	int p = (pool = (wt = (ForkJoinWorkerThread)t).pool).config & SMASK;
2256	>	int n = (q = wt.workQueue).top - q.base;
2257	>	int a = (int)(pool.ctl >> AC_SHIFT) + p;
2258	>	return n - (a > (p >>>= 1) ? 0 :
2259	>	a > (p >>>= 1) ? 1 :
2260	>	a > (p >>>= 1) ? 2 :
2261	>	a > (p >>>= 1) ? 4 :
2262	>	8);
2263	>	}
2264	>	return 0;
2265	>	}
2266	>
2267	>	//  Termination
2268	>
2269	>	/**
2270	>	* Possibly initiates and/or completes termination.  The caller
2271	>	* triggering termination runs three passes through workQueues:
2272	>	* (0) Setting termination status, followed by wakeups of queued
2273	>	* workers; (1) cancelling all tasks; (2) interrupting lagging
2274	>	* threads (likely in external tasks, but possibly also blocked in
2275	>	* joins).  Each pass repeats previous steps because of potential
2276	>	* lagging thread creation.
2277	>	*
2278	>	* @param now if true, unconditionally terminate, else only
2279	>	* if no work and no active workers
2280	>	* @param enable if true, enable shutdown when next possible
2281	>	* @return true if now terminating or terminated
2282		*/
2283	<	private ForkJoinWorkerThread[] ensureWorkerArrayCapacity(int newLength) {
2284	<	ForkJoinWorkerThread[] ws = workers;
2285	<	if (ws == null)
2286	<	return workers = new ForkJoinWorkerThread[arraySizeFor(newLength)];
2287	<	else if (newLength > ws.length)
2288	<	return workers = Arrays.copyOf(ws, arraySizeFor(newLength));
2289	<	else
2290	<	return ws;
2283	>	private boolean tryTerminate(boolean now, boolean enable) {
2284	>	if (this == commonPool)                     // cannot shut down
2285	>	return false;
2286	>	for (long c;;) {
2287	>	if (((c = ctl) & STOP_BIT) != 0) {      // already terminating
2288	>	if ((short)(c >>> TC_SHIFT) == -(config & SMASK)) {
2289	>	synchronized (this) {
2290	>	notifyAll();                // signal when 0 workers
2291	>	}
2292	>	}
2293	>	return true;
2294	>	}
2295	>	if (plock >= 0) {                       // not yet enabled
2296	>	int ps;
2297	>	if (!enable)
2298	>	return false;
2299	>	if (((ps = plock) & PL_LOCK) != 0 ||
2300	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
2301	>	ps = acquirePlock();
2302	>	int nps = SHUTDOWN;
2303	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
2304	>	releasePlock(nps);
2305	>	}
2306	>	if (!now) {                             // check if idle & no tasks
2307	>	if ((int)(c >> AC_SHIFT) != -(config & SMASK) ||
2308	>	hasQueuedSubmissions())
2309	>	return false;
2310	>	// Check for unqueued inactive workers. One pass suffices.
2311	>	WorkQueue[] ws = workQueues; WorkQueue w;
2312	>	if (ws != null) {
2313	>	for (int i = 1; i < ws.length; i += 2) {
2314	>	if ((w = ws[i]) != null && w.eventCount >= 0)
2315	>	return false;
2316	>	}
2317	>	}
2318	>	}
2319	>	if (U.compareAndSwapLong(this, CTL, c, c | STOP_BIT)) {
2320	>	for (int pass = 0; pass < 3; ++pass) {
2321	>	WorkQueue[] ws = workQueues;
2322	>	if (ws != null) {
2323	>	WorkQueue w;
2324	>	int n = ws.length;
2325	>	for (int i = 0; i < n; ++i) {
2326	>	if ((w = ws[i]) != null) {
2327	>	w.qlock = -1;
2328	>	if (pass > 0) {
2329	>	w.cancelAll();
2330	>	if (pass > 1)
2331	>	w.interruptOwner();
2332	>	}
2333	>	}
2334	>	}
2335	>	// Wake up workers parked on event queue
2336	>	int i, e; long cc; Thread p;
2337	>	while ((e = (int)(cc = ctl) & E_MASK) != 0 &&
2338	>	(i = e & SMASK) < n &&
2339	>	(w = ws[i]) != null) {
2340	>	long nc = ((long)(w.nextWait & E_MASK) |
2341	>	((cc + AC_UNIT) & AC_MASK) |
2342	>	(cc & (TC_MASK|STOP_BIT)));
2343	>	if (w.eventCount == (e | INT_SIGN) &&
2344	>	U.compareAndSwapLong(this, CTL, cc, nc)) {
2345	>	w.eventCount = (e + E_SEQ) & E_MASK;
2346	>	w.qlock = -1;
2347	>	if ((p = w.parker) != null)
2348	>	U.unpark(p);
2349	>	}
2350	>	}
2351	>	}
2352	>	}
2353	>	}
2354	>	}
2355		}
2356
2357	+	// external operations on common pool
2358	+
2359		/**
2360	<	* Try to shrink workers into smaller array after one or more terminate
2360	>	* Returns common pool queue for a thread that has submitted at
2361	>	* least one task.
2362		*/
2363	<	private void tryShrinkWorkerArray() {
2364	<	ForkJoinWorkerThread[] ws = workers;
2365	<	int len = ws.length;
2366	<	int last = len - 1;
2367	<	while (last >= 0 && ws[last] == null)
2368	<	--last;
2369	<	int newLength = arraySizeFor(last+1);
447	<	if (newLength < len)
448	<	workers = Arrays.copyOf(ws, newLength);
2363	>	static WorkQueue commonSubmitterQueue() {
2364	>	ForkJoinPool p; WorkQueue[] ws; int m; Submitter z;
2365	>	return ((z = submitters.get()) != null &&
2366	>	(p = commonPool) != null &&
2367	>	(ws = p.workQueues) != null &&
2368	>	(m = ws.length - 1) >= 0) ?
2369	>	ws[m & z.seed & SQMASK] : null;
2370		}
2371
2372		/**
2373	<	* Initial worker array and worker creation and startup. (This
2374	<	* must be done under lock to avoid interference by some of the
2375	<	* newly started threads while creating others.)
2376	<	*/
2377	<	private void createAndStartInitialWorkers(int ps) {
2378	<	final ReentrantLock lock = this.workerLock;
2379	<	lock.lock();
2380	<	try {
2381	<	ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(ps);
2382	<	for (int i = 0; i < ps; ++i) {
2383	<	ForkJoinWorkerThread w = createWorker(i);
2384	<	if (w != null) {
2385	<	ws[i] = w;
2386	<	w.start();
2387	<	updateWorkerCount(1);
2373	>	* Tries to pop the given task from submitter's queue in common pool.
2374	>	*/
2375	>	static boolean tryExternalUnpush(ForkJoinTask<?> t) {
2376	>	ForkJoinPool p; WorkQueue[] ws; WorkQueue q; Submitter z;
2377	>	ForkJoinTask<?>[] a;  int m, s;
2378	>	if (t != null &&
2379	>	(z = submitters.get()) != null &&
2380	>	(p = commonPool) != null &&
2381	>	(ws = p.workQueues) != null &&
2382	>	(m = ws.length - 1) >= 0 &&
2383	>	(q = ws[m & z.seed & SQMASK]) != null &&
2384	>	(s = q.top) != q.base &&
2385	>	(a = q.array) != null) {
2386	>	long j = (((a.length - 1) & (s - 1)) << ASHIFT) + ABASE;
2387	>	if (U.getObject(a, j) == t &&
2388	>	U.compareAndSwapInt(q, QLOCK, 0, 1)) {
2389	>	if (q.array == a && q.top == s && // recheck
2390	>	U.compareAndSwapObject(a, j, t, null)) {
2391	>	q.top = s - 1;
2392	>	q.qlock = 0;
2393	>	return true;
2394		}
2395	+	q.qlock = 0;
2396		}
469	–	} finally {
470	–	lock.unlock();
2397		}
2398	+	return false;
2399		}
2400
2401		/**
2402	<	* Worker creation and startup for threads added via setParallelism.
2403	<	*/
2404	<	private void createAndStartAddedWorkers() {
2405	<	resumeAllSpares();  // Allow spares to convert to nonspare
2406	<	int ps = parallelism;
2407	<	ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(ps);
2408	<	int len = ws.length;
2409	<	// Sweep through slots, to keep lowest indices most populated
2410	<	int k = 0;
2411	<	while (k < len) {
2412	<	if (ws[k] != null) {
2413	<	++k;
2414	<	continue;
2415	<	}
2416	<	int s = workerCounts;
2417	<	int tc = totalCountOf(s);
2418	<	int rc = runningCountOf(s);
2419	<	if (rc >= ps || tc >= ps)
2420	<	break;
2421	<	if (casWorkerCounts (s, workerCountsFor(tc+1, rc+1))) {
2422	<	ForkJoinWorkerThread w = createWorker(k);
2423	<	if (w != null) {
2424	<	ws[k++] = w;
2425	<	w.start();
2402	>	* Tries to pop and run local tasks within the same computation
2403	>	* as the given root. On failure, tries to help complete from
2404	>	* other queues via helpComplete.
2405	>	*/
2406	>	private void externalHelpComplete(WorkQueue q, ForkJoinTask<?> root) {
2407	>	ForkJoinTask<?>[] a; int m;
2408	>	if (q != null && (a = q.array) != null && (m = (a.length - 1)) >= 0 &&
2409	>	root != null && root.status >= 0) {
2410	>	for (;;) {
2411	>	int s, u; Object o; CountedCompleter<?> task = null;
2412	>	if ((s = q.top) - q.base > 0) {
2413	>	long j = ((m & (s - 1)) << ASHIFT) + ABASE;
2414	>	if ((o = U.getObject(a, j)) != null &&
2415	>	(o instanceof CountedCompleter)) {
2416	>	CountedCompleter<?> t = (CountedCompleter<?>)o, r = t;
2417	>	do {
2418	>	if (r == root) {
2419	>	if (U.compareAndSwapInt(q, QLOCK, 0, 1)) {
2420	>	if (q.array == a && q.top == s &&
2421	>	U.compareAndSwapObject(a, j, t, null)) {
2422	>	q.top = s - 1;
2423	>	task = t;
2424	>	}
2425	>	q.qlock = 0;
2426	>	}
2427	>	break;
2428	>	}
2429	>	} while ((r = r.completer) != null);
2430	>	}
2431		}
2432	<	else {
2433	<	updateWorkerCount(-1); // back out on failed creation
2432	>	if (task != null)
2433	>	task.doExec();
2434	>	if (root.status < 0 ||
2435	>	(u = (int)(ctl >>> 32)) >= 0 || (u >> UAC_SHIFT) >= 0)
2436	>	break;
2437	>	if (task == null) {
2438	>	helpSignal(root, q.poolIndex);
2439	>	if (root.status >= 0)
2440	>	helpComplete(root, SHARED_QUEUE);
2441		break;
2442		}
2443		}
2444		}
2445		}
2446
2447	<	// Execution methods
2447	>	/**
2448	>	* Tries to help execute or signal availability of the given task
2449	>	* from submitter's queue in common pool.
2450	>	*/
2451	>	static void externalHelpJoin(ForkJoinTask<?> t) {
2452	>	// Some hard-to-avoid overlap with tryExternalUnpush
2453	>	ForkJoinPool p; WorkQueue[] ws; WorkQueue q, w; Submitter z;
2454	>	ForkJoinTask<?>[] a;  int m, s, n;
2455	>	if (t != null &&
2456	>	(z = submitters.get()) != null &&
2457	>	(p = commonPool) != null &&
2458	>	(ws = p.workQueues) != null &&
2459	>	(m = ws.length - 1) >= 0 &&
2460	>	(q = ws[m & z.seed & SQMASK]) != null &&
2461	>	(a = q.array) != null) {
2462	>	int am = a.length - 1;
2463	>	if ((s = q.top) != q.base) {
2464	>	long j = ((am & (s - 1)) << ASHIFT) + ABASE;
2465	>	if (U.getObject(a, j) == t &&
2466	>	U.compareAndSwapInt(q, QLOCK, 0, 1)) {
2467	>	if (q.array == a && q.top == s &&
2468	>	U.compareAndSwapObject(a, j, t, null)) {
2469	>	q.top = s - 1;
2470	>	q.qlock = 0;
2471	>	t.doExec();
2472	>	}
2473	>	else
2474	>	q.qlock = 0;
2475	>	}
2476	>	}
2477	>	if (t.status >= 0) {
2478	>	if (t instanceof CountedCompleter)
2479	>	p.externalHelpComplete(q, t);
2480	>	else
2481	>	p.helpSignal(t, q.poolIndex);
2482	>	}
2483	>	}
2484	>	}
2485	>
2486	>	/**
2487	>	* Restricted version of helpQuiescePool for external callers
2488	>	*/
2489	>	static void externalHelpQuiescePool() {
2490	>	ForkJoinPool p; ForkJoinTask<?> t; WorkQueue q; int b;
2491	>	if ((p = commonPool) != null &&
2492	>	(q = p.findNonEmptyStealQueue(1)) != null &&
2493	>	(b = q.base) - q.top < 0 &&
2494	>	(t = q.pollAt(b)) != null)
2495	>	t.doExec();
2496	>	}
2497	>
2498	>	// Exported methods
2499	>
2500	>	// Constructors
2501	>
2502	>	/**
2503	>	* Creates a {@code ForkJoinPool} with parallelism equal to {@link
2504	>	* java.lang.Runtime#availableProcessors}, using the {@linkplain
2505	>	* #defaultForkJoinWorkerThreadFactory default thread factory},
2506	>	* no UncaughtExceptionHandler, and non-async LIFO processing mode.
2507	>	*
2508	>	* @throws SecurityException if a security manager exists and
2509	>	*         the caller is not permitted to modify threads
2510	>	*         because it does not hold {@link
2511	>	*         java.lang.RuntimePermission}{@code ("modifyThread")}
2512	>	*/
2513	>	public ForkJoinPool() {
2514	>	this(Runtime.getRuntime().availableProcessors(),
2515	>	defaultForkJoinWorkerThreadFactory, null, false);
2516	>	}
2517	>
2518	>	/**
2519	>	* Creates a {@code ForkJoinPool} with the indicated parallelism
2520	>	* level, the {@linkplain
2521	>	* #defaultForkJoinWorkerThreadFactory default thread factory},
2522	>	* no UncaughtExceptionHandler, and non-async LIFO processing mode.
2523	>	*
2524	>	* @param parallelism the parallelism level
2525	>	* @throws IllegalArgumentException if parallelism less than or
2526	>	*         equal to zero, or greater than implementation limit
2527	>	* @throws SecurityException if a security manager exists and
2528	>	*         the caller is not permitted to modify threads
2529	>	*         because it does not hold {@link
2530	>	*         java.lang.RuntimePermission}{@code ("modifyThread")}
2531	>	*/
2532	>	public ForkJoinPool(int parallelism) {
2533	>	this(parallelism, defaultForkJoinWorkerThreadFactory, null, false);
2534	>	}
2535	>
2536	>	/**
2537	>	* Creates a {@code ForkJoinPool} with the given parameters.
2538	>	*
2539	>	* @param parallelism the parallelism level. For default value,
2540	>	* use {@link java.lang.Runtime#availableProcessors}.
2541	>	* @param factory the factory for creating new threads. For default value,
2542	>	* use {@link #defaultForkJoinWorkerThreadFactory}.
2543	>	* @param handler the handler for internal worker threads that
2544	>	* terminate due to unrecoverable errors encountered while executing
2545	>	* tasks. For default value, use {@code null}.
2546	>	* @param asyncMode if true,
2547	>	* establishes local first-in-first-out scheduling mode for forked
2548	>	* tasks that are never joined. This mode may be more appropriate
2549	>	* than default locally stack-based mode in applications in which
2550	>	* worker threads only process event-style asynchronous tasks.
2551	>	* For default value, use {@code false}.
2552	>	* @throws IllegalArgumentException if parallelism less than or
2553	>	*         equal to zero, or greater than implementation limit
2554	>	* @throws NullPointerException if the factory is null
2555	>	* @throws SecurityException if a security manager exists and
2556	>	*         the caller is not permitted to modify threads
2557	>	*         because it does not hold {@link
2558	>	*         java.lang.RuntimePermission}{@code ("modifyThread")}
2559	>	*/
2560	>	public ForkJoinPool(int parallelism,
2561	>	ForkJoinWorkerThreadFactory factory,
2562	>	Thread.UncaughtExceptionHandler handler,
2563	>	boolean asyncMode) {
2564	>	checkPermission();
2565	>	if (factory == null)
2566	>	throw new NullPointerException();
2567	>	if (parallelism <= 0 || parallelism > MAX_CAP)
2568	>	throw new IllegalArgumentException();
2569	>	this.factory = factory;
2570	>	this.ueh = handler;
2571	>	this.config = parallelism | (asyncMode ? (FIFO_QUEUE << 16) : 0);
2572	>	long np = (long)(-parallelism); // offset ctl counts
2573	>	this.ctl = ((np << AC_SHIFT) & AC_MASK) | ((np << TC_SHIFT) & TC_MASK);
2574	>	int pn = nextPoolId();
2575	>	StringBuilder sb = new StringBuilder("ForkJoinPool-");
2576	>	sb.append(Integer.toString(pn));
2577	>	sb.append("-worker-");
2578	>	this.workerNamePrefix = sb.toString();
2579	>	}
2580	>
2581	>	/**
2582	>	* Constructor for common pool, suitable only for static initialization.
2583	>	* Basically the same as above, but uses smallest possible initial footprint.
2584	>	*/
2585	>	ForkJoinPool(int parallelism, long ctl,
2586	>	ForkJoinWorkerThreadFactory factory,
2587	>	Thread.UncaughtExceptionHandler handler) {
2588	>	this.config = parallelism;
2589	>	this.ctl = ctl;
2590	>	this.factory = factory;
2591	>	this.ueh = handler;
2592	>	this.workerNamePrefix = "ForkJoinPool.commonPool-worker-";
2593	>	}
2594
2595		/**
2596	<	* Common code for execute, invoke and submit
2596	>	* Returns the common pool instance.
2597	>	*
2598	>	* @return the common pool instance
2599		*/
2600	<	private <T> void doSubmit(ForkJoinTask<T> task) {
2601	<	if (isShutdown())
2602	<	throw new RejectedExecutionException();
516	<	submissionQueue.offer(task);
517	<	signalIdleWorkers();
2600	>	public static ForkJoinPool commonPool() {
2601	>	// assert commonPool != null : "static init error";
2602	>	return commonPool;
2603		}
2604
2605	+	// Execution methods
2606	+
2607		/**
2608	<	* Performs the given task; returning its result upon completion
2608	>	* Performs the given task, returning its result upon completion.
2609	>	* If the computation encounters an unchecked Exception or Error,
2610	>	* it is rethrown as the outcome of this invocation.  Rethrown
2611	>	* exceptions behave in the same way as regular exceptions, but,
2612	>	* when possible, contain stack traces (as displayed for example
2613	>	* using {@code ex.printStackTrace()}) of both the current thread
2614	>	* as well as the thread actually encountering the exception;
2615	>	* minimally only the latter.
2616	>	*
2617		* @param task the task
2618		* @return the task's result
2619	<	* @throws NullPointerException if task is null
2620	<	* @throws RejectedExecutionException if pool is shut down
2619	>	* @throws NullPointerException if the task is null
2620	>	* @throws RejectedExecutionException if the task cannot be
2621	>	*         scheduled for execution
2622		*/
2623		public <T> T invoke(ForkJoinTask<T> task) {
2624	<	doSubmit(task);
2624	>	if (task == null)
2625	>	throw new NullPointerException();
2626	>	externalPush(task);
2627		return task.join();
2628		}
2629
2630		/**
2631		* Arranges for (asynchronous) execution of the given task.
2632	+	*
2633		* @param task the task
2634	<	* @throws NullPointerException if task is null
2635	<	* @throws RejectedExecutionException if pool is shut down
2634	>	* @throws NullPointerException if the task is null
2635	>	* @throws RejectedExecutionException if the task cannot be
2636	>	*         scheduled for execution
2637		*/
2638	<	public <T> void execute(ForkJoinTask<T> task) {
2639	<	doSubmit(task);
2638	>	public void execute(ForkJoinTask<?> task) {
2639	>	if (task == null)
2640	>	throw new NullPointerException();
2641	>	externalPush(task);
2642		}
2643
2644		// AbstractExecutorService methods
2645
2646	+	/**
2647	+	* @throws NullPointerException if the task is null
2648	+	* @throws RejectedExecutionException if the task cannot be
2649	+	*         scheduled for execution
2650	+	*/
2651		public void execute(Runnable task) {
2652	<	doSubmit(new AdaptedRunnable<Void>(task, null));
2652	>	if (task == null)
2653	>	throw new NullPointerException();
2654	>	ForkJoinTask<?> job;
2655	>	if (task instanceof ForkJoinTask<?>) // avoid re-wrap
2656	>	job = (ForkJoinTask<?>) task;
2657	>	else
2658	>	job = new ForkJoinTask.AdaptedRunnableAction(task);
2659	>	externalPush(job);
2660		}
2661
2662	<	public <T> ForkJoinTask<T> submit(Callable<T> task) {
2663	<	ForkJoinTask<T> job = new AdaptedCallable<T>(task);
2664	<	doSubmit(job);
2665	<	return job;
2662	>	/**
2663	>	* Submits a ForkJoinTask for execution.
2664	>	*
2665	>	* @param task the task to submit
2666	>	* @return the task
2667	>	* @throws NullPointerException if the task is null
2668	>	* @throws RejectedExecutionException if the task cannot be
2669	>	*         scheduled for execution
2670	>	*/
2671	>	public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) {
2672	>	if (task == null)
2673	>	throw new NullPointerException();
2674	>	externalPush(task);
2675	>	return task;
2676		}
2677
2678	<	public <T> ForkJoinTask<T> submit(Runnable task, T result) {
2679	<	ForkJoinTask<T> job = new AdaptedRunnable<T>(task, result);
2680	<	doSubmit(job);
2678	>	/**
2679	>	* @throws NullPointerException if the task is null
2680	>	* @throws RejectedExecutionException if the task cannot be
2681	>	*         scheduled for execution
2682	>	*/
2683	>	public <T> ForkJoinTask<T> submit(Callable<T> task) {
2684	>	ForkJoinTask<T> job = new ForkJoinTask.AdaptedCallable<T>(task);
2685	>	externalPush(job);
2686		return job;
2687		}
2688
2689	<	public ForkJoinTask<?> submit(Runnable task) {
2690	<	ForkJoinTask<Void> job = new AdaptedRunnable<Void>(task, null);
2691	<	doSubmit(job);
2689	>	/**
2690	>	* @throws NullPointerException if the task is null
2691	>	* @throws RejectedExecutionException if the task cannot be
2692	>	*         scheduled for execution
2693	>	*/
2694	>	public <T> ForkJoinTask<T> submit(Runnable task, T result) {
2695	>	ForkJoinTask<T> job = new ForkJoinTask.AdaptedRunnable<T>(task, result);
2696	>	externalPush(job);
2697		return job;
2698		}
2699
2700		/**
2701	<	* Adaptor for Runnables. This implements RunnableFuture
2702	<	* to be compliant with AbstractExecutorService constraints
2701	>	* @throws NullPointerException if the task is null
2702	>	* @throws RejectedExecutionException if the task cannot be
2703	>	*         scheduled for execution
2704		*/
2705	<	static final class AdaptedRunnable<T> extends ForkJoinTask<T>
2706	<	implements RunnableFuture<T> {
2707	<	final Runnable runnable;
2708	<	final T resultOnCompletion;
2709	<	T result;
2710	<	AdaptedRunnable(Runnable runnable, T result) {
2711	<	if (runnable == null) throw new NullPointerException();
2712	<	this.runnable = runnable;
2713	<	this.resultOnCompletion = result;
2714	<	}
580	<	public T getRawResult() { return result; }
581	<	public void setRawResult(T v) { result = v; }
582	<	public boolean exec() {
583	<	runnable.run();
584	<	result = resultOnCompletion;
585	<	return true;
586	<	}
587	<	public void run() { invoke(); }
2705	>	public ForkJoinTask<?> submit(Runnable task) {
2706	>	if (task == null)
2707	>	throw new NullPointerException();
2708	>	ForkJoinTask<?> job;
2709	>	if (task instanceof ForkJoinTask<?>) // avoid re-wrap
2710	>	job = (ForkJoinTask<?>) task;
2711	>	else
2712	>	job = new ForkJoinTask.AdaptedRunnableAction(task);
2713	>	externalPush(job);
2714	>	return job;
2715		}
2716
2717		/**
2718	<	* Adaptor for Callables
2718	>	* @throws NullPointerException       {@inheritDoc}
2719	>	* @throws RejectedExecutionException {@inheritDoc}
2720		*/
593	–	static final class AdaptedCallable<T> extends ForkJoinTask<T>
594	–	implements RunnableFuture<T> {
595	–	final Callable<T> callable;
596	–	T result;
597	–	AdaptedCallable(Callable<T> callable) {
598	–	if (callable == null) throw new NullPointerException();
599	–	this.callable = callable;
600	–	}
601	–	public T getRawResult() { return result; }
602	–	public void setRawResult(T v) { result = v; }
603	–	public boolean exec() {
604	–	try {
605	–	result = callable.call();
606	–	return true;
607	–	} catch (Error err) {
608	–	throw err;
609	–	} catch (RuntimeException rex) {
610	–	throw rex;
611	–	} catch (Exception ex) {
612	–	throw new RuntimeException(ex);
613	–	}
614	–	}
615	–	public void run() { invoke(); }
616	–	}
617	–
2721		public <T> List<Future<T>> invokeAll(Collection<? extends Callable<T>> tasks) {
2722	<	ArrayList<ForkJoinTask<T>> ts =
2723	<	new ArrayList<ForkJoinTask<T>>(tasks.size());
2724	<	for (Callable<T> c : tasks)
2725	<	ts.add(new AdaptedCallable<T>(c));
2726	<	invoke(new InvokeAll<T>(ts));
2727	<	return (List<Future<T>>)(List)ts;
2728	<	}
2722	>	// In previous versions of this class, this method constructed
2723	>	// a task to run ForkJoinTask.invokeAll, but now external
2724	>	// invocation of multiple tasks is at least as efficient.
2725	>	List<ForkJoinTask<T>> fs = new ArrayList<ForkJoinTask<T>>(tasks.size());
2726	>	// Workaround needed because method wasn't declared with
2727	>	// wildcards in return type but should have been.
2728	>	@SuppressWarnings({"unchecked", "rawtypes"})
2729	>	List<Future<T>> futures = (List<Future<T>>) (List) fs;
2730
2731	<	static final class InvokeAll<T> extends RecursiveAction {
2732	<	final ArrayList<ForkJoinTask<T>> tasks;
2733	<	InvokeAll(ArrayList<ForkJoinTask<T>> tasks) { this.tasks = tasks; }
2734	<	public void compute() {
2735	<	try { invokeAll(tasks); } catch(Exception ignore) {}
2731	>	boolean done = false;
2732	>	try {
2733	>	for (Callable<T> t : tasks) {
2734	>	ForkJoinTask<T> f = new ForkJoinTask.AdaptedCallable<T>(t);
2735	>	externalPush(f);
2736	>	fs.add(f);
2737	>	}
2738	>	for (ForkJoinTask<T> f : fs)
2739	>	f.quietlyJoin();
2740	>	done = true;
2741	>	return futures;
2742	>	} finally {
2743	>	if (!done)
2744	>	for (ForkJoinTask<T> f : fs)
2745	>	f.cancel(false);
2746		}
2747		}
2748
635	–	// Configuration and status settings and queries
636	–
2749		/**
2750	<	* Returns the factory used for constructing new workers
2750	>	* Returns the factory used for constructing new workers.
2751		*
2752		* @return the factory used for constructing new workers
2753		*/
#	Line 646 | Line 2758 | public class ForkJoinPool extends Abstra
2758		/**
2759		* Returns the handler for internal worker threads that terminate
2760		* due to unrecoverable errors encountered while executing tasks.
2761	<	* @return the handler, or null if none
2761	>	*
2762	>	* @return the handler, or {@code null} if none
2763		*/
2764		public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() {
2765	<	Thread.UncaughtExceptionHandler h;
653	<	final ReentrantLock lock = this.workerLock;
654	<	lock.lock();
655	<	try {
656	<	h = ueh;
657	<	} finally {
658	<	lock.unlock();
659	<	}
660	<	return h;
2765	>	return ueh;
2766		}
2767
2768		/**
2769	<	* Sets the handler for internal worker threads that terminate due
665	<	* to unrecoverable errors encountered while executing tasks.
666	<	* Unless set, the current default or ThreadGroup handler is used
667	<	* as handler.
2769	>	* Returns the targeted parallelism level of this pool.
2770		*
2771	<	* @param h the new handler
670	<	* @return the old handler, or null if none
671	<	* @throws SecurityException if a security manager exists and
672	<	*         the caller is not permitted to modify threads
673	<	*         because it does not hold {@link
674	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
675	<	*/
676	<	public Thread.UncaughtExceptionHandler
677	<	setUncaughtExceptionHandler(Thread.UncaughtExceptionHandler h) {
678	<	checkPermission();
679	<	Thread.UncaughtExceptionHandler old = null;
680	<	final ReentrantLock lock = this.workerLock;
681	<	lock.lock();
682	<	try {
683	<	old = ueh;
684	<	ueh = h;
685	<	ForkJoinWorkerThread[] ws = workers;
686	<	for (int i = 0; i < ws.length; ++i) {
687	<	ForkJoinWorkerThread w = ws[i];
688	<	if (w != null)
689	<	w.setUncaughtExceptionHandler(h);
690	<	}
691	<	} finally {
692	<	lock.unlock();
693	<	}
694	<	return old;
695	<	}
696	<
697	<
698	<	/**
699	<	* Sets the target paralleism level of this pool.
700	<	* @param parallelism the target parallelism
701	<	* @throws IllegalArgumentException if parallelism less than or
702	<	* equal to zero or greater than maximum size bounds.
703	<	* @throws SecurityException if a security manager exists and
704	<	*         the caller is not permitted to modify threads
705	<	*         because it does not hold {@link
706	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
2771	>	* @return the targeted parallelism level of this pool
2772		*/
2773	<	public void setParallelism(int parallelism) {
2774	<	checkPermission();
710	<	if (parallelism <= 0 || parallelism > maxPoolSize)
711	<	throw new IllegalArgumentException();
712	<	final ReentrantLock lock = this.workerLock;
713	<	lock.lock();
714	<	try {
715	<	if (!isTerminating()) {
716	<	int p = this.parallelism;
717	<	this.parallelism = parallelism;
718	<	if (parallelism > p)
719	<	createAndStartAddedWorkers();
720	<	else
721	<	trimSpares();
722	<	}
723	<	} finally {
724	<	lock.unlock();
725	<	}
726	<	signalIdleWorkers();
2773	>	public int getParallelism() {
2774	>	return config & SMASK;
2775		}
2776
2777		/**
2778	<	* Returns the targeted number of worker threads in this pool.
2778	>	* Returns the targeted parallelism level of the common pool.
2779		*
2780	<	* @return the targeted number of worker threads in this pool
2780	>	* @return the targeted parallelism level of the common pool
2781		*/
2782	<	public int getParallelism() {
2783	<	return parallelism;
2782	>	public static int getCommonPoolParallelism() {
2783	>	return commonPoolParallelism;
2784		}
2785
2786		/**
2787		* Returns the number of worker threads that have started but not
2788	<	* yet terminated.  This result returned by this method may differ
2789	<	* from <code>getParallelism</code> when threads are created to
2788	>	* yet terminated.  The result returned by this method may differ
2789	>	* from {@link #getParallelism} when threads are created to
2790		* maintain parallelism when others are cooperatively blocked.
2791		*
2792		* @return the number of worker threads
2793		*/
2794		public int getPoolSize() {
2795	<	return totalCountOf(workerCounts);
2795	>	return (config & SMASK) + (short)(ctl >>> TC_SHIFT);
2796		}
2797
2798		/**
2799	<	* Returns the maximum number of threads allowed to exist in the
2800	<	* pool, even if there are insufficient unblocked running threads.
2801	<	* @return the maximum
2802	<	*/
755	<	public int getMaximumPoolSize() {
756	<	return maxPoolSize;
757	<	}
758	<
759	<	/**
760	<	* Sets the maximum number of threads allowed to exist in the
761	<	* pool, even if there are insufficient unblocked running threads.
762	<	* Setting this value has no effect on current pool size. It
763	<	* controls construction of new threads.
764	<	* @throws IllegalArgumentException if negative or greater then
765	<	* internal implementation limit.
766	<	*/
767	<	public void setMaximumPoolSize(int newMax) {
768	<	if (newMax < 0 || newMax > MAX_THREADS)
769	<	throw new IllegalArgumentException();
770	<	maxPoolSize = newMax;
771	<	}
772	<
773	<
774	<	/**
775	<	* Returns true if this pool dynamically maintains its target
776	<	* parallelism level. If false, new threads are added only to
777	<	* avoid possible starvation.
778	<	* This setting is by default true;
779	<	* @return true if maintains parallelism
780	<	*/
781	<	public boolean getMaintainsParallelism() {
782	<	return maintainsParallelism;
783	<	}
784	<
785	<	/**
786	<	* Sets whether this pool dynamically maintains its target
787	<	* parallelism level. If false, new threads are added only to
788	<	* avoid possible starvation.
789	<	* @param enable true to maintains parallelism
2799	>	* Returns {@code true} if this pool uses local first-in-first-out
2800	>	* scheduling mode for forked tasks that are never joined.
2801	>	*
2802	>	* @return {@code true} if this pool uses async mode
2803		*/
2804	<	public void setMaintainsParallelism(boolean enable) {
2805	<	maintainsParallelism = enable;
2804	>	public boolean getAsyncMode() {
2805	>	return (config >>> 16) == FIFO_QUEUE;
2806		}
2807
2808		/**
2809		* Returns an estimate of the number of worker threads that are
2810		* not blocked waiting to join tasks or for other managed
2811	<	* synchronization.
2811	>	* synchronization. This method may overestimate the
2812	>	* number of running threads.
2813		*
2814		* @return the number of worker threads
2815		*/
2816		public int getRunningThreadCount() {
2817	<	return runningCountOf(workerCounts);
2817	>	int rc = 0;
2818	>	WorkQueue[] ws; WorkQueue w;
2819	>	if ((ws = workQueues) != null) {
2820	>	for (int i = 1; i < ws.length; i += 2) {
2821	>	if ((w = ws[i]) != null && w.isApparentlyUnblocked())
2822	>	++rc;
2823	>	}
2824	>	}
2825	>	return rc;
2826		}
2827
2828		/**
2829		* Returns an estimate of the number of threads that are currently
2830		* stealing or executing tasks. This method may overestimate the
2831		* number of active threads.
2832	<	* @return the number of active threads.
2832	>	*
2833	>	* @return the number of active threads
2834		*/
2835		public int getActiveThreadCount() {
2836	<	return activeCountOf(runControl);
2837	<	}
815	<
816	<	/**
817	<	* Returns an estimate of the number of threads that are currently
818	<	* idle waiting for tasks. This method may underestimate the
819	<	* number of idle threads.
820	<	* @return the number of idle threads.
821	<	*/
822	<	final int getIdleThreadCount() {
823	<	int c = runningCountOf(workerCounts) - activeCountOf(runControl);
824	<	return (c <= 0)? 0 : c;
2836	>	int r = (config & SMASK) + (int)(ctl >> AC_SHIFT);
2837	>	return (r <= 0) ? 0 : r; // suppress momentarily negative values
2838		}
2839
2840		/**
2841	<	* Returns true if all worker threads are currently idle. An idle
2842	<	* worker is one that cannot obtain a task to execute because none
2843	<	* are available to steal from other threads, and there are no
2844	<	* pending submissions to the pool. This method is conservative:
2845	<	* It might not return true immediately upon idleness of all
2846	<	* threads, but will eventually become true if threads remain
2847	<	* inactive.
2848	<	* @return true if all threads are currently idle
2841	>	* Returns {@code true} if all worker threads are currently idle.
2842	>	* An idle worker is one that cannot obtain a task to execute
2843	>	* because none are available to steal from other threads, and
2844	>	* there are no pending submissions to the pool. This method is
2845	>	* conservative; it might not return {@code true} immediately upon
2846	>	* idleness of all threads, but will eventually become true if
2847	>	* threads remain inactive.
2848	>	*
2849	>	* @return {@code true} if all threads are currently idle
2850		*/
2851		public boolean isQuiescent() {
2852	<	return activeCountOf(runControl) == 0;
2852	>	return (int)(ctl >> AC_SHIFT) + (config & SMASK) == 0;
2853		}
2854
2855		/**
#	Line 843 | Line 2857 | public class ForkJoinPool extends Abstra
2857		* one thread's work queue by another. The reported value
2858		* underestimates the actual total number of steals when the pool
2859		* is not quiescent. This value may be useful for monitoring and
2860	<	* tuning fork/join programs: In general, steal counts should be
2860	>	* tuning fork/join programs: in general, steal counts should be
2861		* high enough to keep threads busy, but low enough to avoid
2862		* overhead and contention across threads.
2863	<	* @return the number of steals.
2863	>	*
2864	>	* @return the number of steals
2865		*/
2866		public long getStealCount() {
2867	<	return stealCount.get();
2868	<	}
2869	<
2870	<	/**
2871	<	* Accumulate steal count from a worker. Call only
2872	<	* when worker known to be idle.
2873	<	*/
2874	<	private void updateStealCount(ForkJoinWorkerThread w) {
2875	<	int sc = w.getAndClearStealCount();
861	<	if (sc != 0)
862	<	stealCount.addAndGet(sc);
2867	>	long count = stealCount;
2868	>	WorkQueue[] ws; WorkQueue w;
2869	>	if ((ws = workQueues) != null) {
2870	>	for (int i = 1; i < ws.length; i += 2) {
2871	>	if ((w = ws[i]) != null)
2872	>	count += w.nsteals;
2873	>	}
2874	>	}
2875	>	return count;
2876		}
2877
2878		/**
#	Line 869 | Line 2882 | public class ForkJoinPool extends Abstra
2882		* an approximation, obtained by iterating across all threads in
2883		* the pool. This method may be useful for tuning task
2884		* granularities.
2885	<	* @return the number of queued tasks.
2885	>	*
2886	>	* @return the number of queued tasks
2887		*/
2888		public long getQueuedTaskCount() {
2889		long count = 0;
2890	<	ForkJoinWorkerThread[] ws = workers;
2891	<	for (int i = 0; i < ws.length; ++i) {
2892	<	ForkJoinWorkerThread t = ws[i];
2893	<	if (t != null)
2894	<	count += t.getQueueSize();
2890	>	WorkQueue[] ws; WorkQueue w;
2891	>	if ((ws = workQueues) != null) {
2892	>	for (int i = 1; i < ws.length; i += 2) {
2893	>	if ((w = ws[i]) != null)
2894	>	count += w.queueSize();
2895	>	}
2896		}
2897		return count;
2898		}
2899
2900		/**
2901	<	* Returns an estimate of the number tasks submitted to this pool
2902	<	* that have not yet begun executing. This method takes time
2903	<	* proportional to the number of submissions.
2904	<	* @return the number of queued submissions.
2901	>	* Returns an estimate of the number of tasks submitted to this
2902	>	* pool that have not yet begun executing.  This method may take
2903	>	* time proportional to the number of submissions.
2904	>	*
2905	>	* @return the number of queued submissions
2906		*/
2907		public int getQueuedSubmissionCount() {
2908	<	return submissionQueue.size();
2908	>	int count = 0;
2909	>	WorkQueue[] ws; WorkQueue w;
2910	>	if ((ws = workQueues) != null) {
2911	>	for (int i = 0; i < ws.length; i += 2) {
2912	>	if ((w = ws[i]) != null)
2913	>	count += w.queueSize();
2914	>	}
2915	>	}
2916	>	return count;
2917		}
2918
2919		/**
2920	<	* Returns true if there are any tasks submitted to this pool
2921	<	* that have not yet begun executing.
2922	<	* @return <code>true</code> if there are any queued submissions.
2920	>	* Returns {@code true} if there are any tasks submitted to this
2921	>	* pool that have not yet begun executing.
2922	>	*
2923	>	* @return {@code true} if there are any queued submissions
2924		*/
2925		public boolean hasQueuedSubmissions() {
2926	<	return !submissionQueue.isEmpty();
2926	>	WorkQueue[] ws; WorkQueue w;
2927	>	if ((ws = workQueues) != null) {
2928	>	for (int i = 0; i < ws.length; i += 2) {
2929	>	if ((w = ws[i]) != null && !w.isEmpty())
2930	>	return true;
2931	>	}
2932	>	}
2933	>	return false;
2934		}
2935
2936		/**
2937		* Removes and returns the next unexecuted submission if one is
2938		* available.  This method may be useful in extensions to this
2939		* class that re-assign work in systems with multiple pools.
2940	<	* @return the next submission, or null if none
2940	>	*
2941	>	* @return the next submission, or {@code null} if none
2942		*/
2943		protected ForkJoinTask<?> pollSubmission() {
2944	<	return submissionQueue.poll();
2944	>	WorkQueue[] ws; WorkQueue w; ForkJoinTask<?> t;
2945	>	if ((ws = workQueues) != null) {
2946	>	for (int i = 0; i < ws.length; i += 2) {
2947	>	if ((w = ws[i]) != null && (t = w.poll()) != null)
2948	>	return t;
2949	>	}
2950	>	}
2951	>	return null;
2952	>	}
2953	>
2954	>	/**
2955	>	* Removes all available unexecuted submitted and forked tasks
2956	>	* from scheduling queues and adds them to the given collection,
2957	>	* without altering their execution status. These may include
2958	>	* artificially generated or wrapped tasks. This method is
2959	>	* designed to be invoked only when the pool is known to be
2960	>	* quiescent. Invocations at other times may not remove all
2961	>	* tasks. A failure encountered while attempting to add elements
2962	>	* to collection {@code c} may result in elements being in
2963	>	* neither, either or both collections when the associated
2964	>	* exception is thrown.  The behavior of this operation is
2965	>	* undefined if the specified collection is modified while the
2966	>	* operation is in progress.
2967	>	*
2968	>	* @param c the collection to transfer elements into
2969	>	* @return the number of elements transferred
2970	>	*/
2971	>	protected int drainTasksTo(Collection<? super ForkJoinTask<?>> c) {
2972	>	int count = 0;
2973	>	WorkQueue[] ws; WorkQueue w; ForkJoinTask<?> t;
2974	>	if ((ws = workQueues) != null) {
2975	>	for (int i = 0; i < ws.length; ++i) {
2976	>	if ((w = ws[i]) != null) {
2977	>	while ((t = w.poll()) != null) {
2978	>	c.add(t);
2979	>	++count;
2980	>	}
2981	>	}
2982	>	}
2983	>	}
2984	>	return count;
2985		}
2986
2987		/**
#	Line 919 | Line 2992 | public class ForkJoinPool extends Abstra
2992		* @return a string identifying this pool, as well as its state
2993		*/
2994		public String toString() {
2995	<	int ps = parallelism;
2996	<	int wc = workerCounts;
2997	<	int rc = runControl;
2998	<	long st = getStealCount();
2999	<	long qt = getQueuedTaskCount();
3000	<	long qs = getQueuedSubmissionCount();
2995	>	// Use a single pass through workQueues to collect counts
2996	>	long qt = 0L, qs = 0L; int rc = 0;
2997	>	long st = stealCount;
2998	>	long c = ctl;
2999	>	WorkQueue[] ws; WorkQueue w;
3000	>	if ((ws = workQueues) != null) {
3001	>	for (int i = 0; i < ws.length; ++i) {
3002	>	if ((w = ws[i]) != null) {
3003	>	int size = w.queueSize();
3004	>	if ((i & 1) == 0)
3005	>	qs += size;
3006	>	else {
3007	>	qt += size;
3008	>	st += w.nsteals;
3009	>	if (w.isApparentlyUnblocked())
3010	>	++rc;
3011	>	}
3012	>	}
3013	>	}
3014	>	}
3015	>	int pc = (config & SMASK);
3016	>	int tc = pc + (short)(c >>> TC_SHIFT);
3017	>	int ac = pc + (int)(c >> AC_SHIFT);
3018	>	if (ac < 0) // ignore transient negative
3019	>	ac = 0;
3020	>	String level;
3021	>	if ((c & STOP_BIT) != 0)
3022	>	level = (tc == 0) ? "Terminated" : "Terminating";
3023	>	else
3024	>	level = plock < 0 ? "Shutting down" : "Running";
3025		return super.toString() +
3026	<	"[" + runStateToString(runStateOf(rc)) +
3027	<	", parallelism = " + ps +
3028	<	", size = " + totalCountOf(wc) +
3029	<	", active = " + activeCountOf(rc) +
3030	<	", running = " + runningCountOf(wc) +
3026	>	"[" + level +
3027	>	", parallelism = " + pc +
3028	>	", size = " + tc +
3029	>	", active = " + ac +
3030	>	", running = " + rc +
3031		", steals = " + st +
3032		", tasks = " + qt +
3033		", submissions = " + qs +
3034		"]";
3035		}
3036
940	–	private static String runStateToString(int rs) {
941	–	switch(rs) {
942	–	case RUNNING: return "Running";
943	–	case SHUTDOWN: return "Shutting down";
944	–	case TERMINATING: return "Terminating";
945	–	case TERMINATED: return "Terminated";
946	–	default: throw new Error("Unknown run state");
947	–	}
948	–	}
949	–
950	–	// lifecycle control
951	–
3037		/**
3038	<	* Initiates an orderly shutdown in which previously submitted
3039	<	* tasks are executed, but no new tasks will be accepted.
3040	<	* Invocation has no additional effect if already shut down.
3041	<	* Tasks that are in the process of being submitted concurrently
3042	<	* during the course of this method may or may not be rejected.
3038	>	* Possibly initiates an orderly shutdown in which previously
3039	>	* submitted tasks are executed, but no new tasks will be
3040	>	* accepted. Invocation has no effect on execution state if this
3041	>	* is the {@link #commonPool}, and no additional effect if
3042	>	* already shut down.  Tasks that are in the process of being
3043	>	* submitted concurrently during the course of this method may or
3044	>	* may not be rejected.
3045	>	*
3046		* @throws SecurityException if a security manager exists and
3047		*         the caller is not permitted to modify threads
3048		*         because it does not hold {@link
3049	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
3049	>	*         java.lang.RuntimePermission}{@code ("modifyThread")}
3050		*/
3051		public void shutdown() {
3052		checkPermission();
3053	<	transitionRunStateTo(SHUTDOWN);
966	<	if (canTerminateOnShutdown(runControl))
967	<	terminateOnShutdown();
3053	>	tryTerminate(false, true);
3054		}
3055
3056		/**
3057	<	* Attempts to stop all actively executing tasks, and cancels all
3058	<	* waiting tasks.  Tasks that are in the process of being
3059	<	* submitted or executed concurrently during the course of this
3060	<	* method may or may not be rejected. Unlike some other executors,
3061	<	* this method cancels rather than collects non-executed tasks,
3062	<	* so always returns an empty list.
3057	>	* Possibly attempts to cancel and/or stop all tasks, and reject
3058	>	* all subsequently submitted tasks.  Invocation has no effect on
3059	>	* execution state if this is the {@link #commonPool}, and no
3060	>	* additional effect if already shut down. Otherwise, tasks that
3061	>	* are in the process of being submitted or executed concurrently
3062	>	* during the course of this method may or may not be
3063	>	* rejected. This method cancels both existing and unexecuted
3064	>	* tasks, in order to permit termination in the presence of task
3065	>	* dependencies. So the method always returns an empty list
3066	>	* (unlike the case for some other Executors).
3067	>	*
3068		* @return an empty list
3069		* @throws SecurityException if a security manager exists and
3070		*         the caller is not permitted to modify threads
3071		*         because it does not hold {@link
3072	<	*         java.lang.RuntimePermission}<code>("modifyThread")</code>,
3072	>	*         java.lang.RuntimePermission}{@code ("modifyThread")}
3073		*/
3074		public List<Runnable> shutdownNow() {
3075		checkPermission();
3076	<	terminate();
3076	>	tryTerminate(true, true);
3077		return Collections.emptyList();
3078		}
3079
3080		/**
3081	<	* Returns <code>true</code> if all tasks have completed following shut down.
3081	>	* Returns {@code true} if all tasks have completed following shut down.
3082		*
3083	<	* @return <code>true</code> if all tasks have completed following shut down
3083	>	* @return {@code true} if all tasks have completed following shut down
3084		*/
3085		public boolean isTerminated() {
3086	<	return runStateOf(runControl) == TERMINATED;
3086	>	long c = ctl;
3087	>	return ((c & STOP_BIT) != 0L &&
3088	>	(short)(c >>> TC_SHIFT) == -(config & SMASK));
3089		}
3090
3091		/**
3092	<	* Returns <code>true</code> if the process of termination has
3093	<	* commenced but possibly not yet completed.
3092	>	* Returns {@code true} if the process of termination has
3093	>	* commenced but not yet completed.  This method may be useful for
3094	>	* debugging. A return of {@code true} reported a sufficient
3095	>	* period after shutdown may indicate that submitted tasks have
3096	>	* ignored or suppressed interruption, or are waiting for I/O,
3097	>	* causing this executor not to properly terminate. (See the
3098	>	* advisory notes for class {@link ForkJoinTask} stating that
3099	>	* tasks should not normally entail blocking operations.  But if
3100	>	* they do, they must abort them on interrupt.)
3101		*
3102	<	* @return <code>true</code> if terminating
3102	>	* @return {@code true} if terminating but not yet terminated
3103		*/
3104		public boolean isTerminating() {
3105	<	return runStateOf(runControl) >= TERMINATING;
3105	>	long c = ctl;
3106	>	return ((c & STOP_BIT) != 0L &&
3107	>	(short)(c >>> TC_SHIFT) != -(config & SMASK));
3108		}
3109
3110		/**
3111	<	* Returns <code>true</code> if this pool has been shut down.
3111	>	* Returns {@code true} if this pool has been shut down.
3112		*
3113	<	* @return <code>true</code> if this pool has been shut down
3113	>	* @return {@code true} if this pool has been shut down
3114		*/
3115		public boolean isShutdown() {
3116	<	return runStateOf(runControl) >= SHUTDOWN;
3116	>	return plock < 0;
3117		}
3118
3119		/**
3120	<	* Blocks until all tasks have completed execution after a shutdown
3121	<	* request, or the timeout occurs, or the current thread is
3122	<	* interrupted, whichever happens first.
3120	>	* Blocks until all tasks have completed execution after a
3121	>	* shutdown request, or the timeout occurs, or the current thread
3122	>	* is interrupted, whichever happens first. Note that the {@link
3123	>	* #commonPool()} never terminates until program shutdown so
3124	>	* this method will always time out.
3125		*
3126		* @param timeout the maximum time to wait
3127		* @param unit the time unit of the timeout argument
3128	<	* @return <code>true</code> if this executor terminated and
3129	<	*         <code>false</code> if the timeout elapsed before termination
3128	>	* @return {@code true} if this executor terminated and
3129	>	*         {@code false} if the timeout elapsed before termination
3130		* @throws InterruptedException if interrupted while waiting
3131		*/
3132		public boolean awaitTermination(long timeout, TimeUnit unit)
3133		throws InterruptedException {
3134		long nanos = unit.toNanos(timeout);
3135	<	final ReentrantLock lock = this.workerLock;
1032	<	lock.lock();
1033	<	try {
1034	<	for (;;) {
1035	<	if (isTerminated())
1036	<	return true;
1037	<	if (nanos <= 0)
1038	<	return false;
1039	<	nanos = termination.awaitNanos(nanos);
1040	<	}
1041	<	} finally {
1042	<	lock.unlock();
1043	<	}
1044	<	}
1045	<
1046	<	// Shutdown and termination support
1047	<
1048	<	/**
1049	<	* Callback from terminating worker. Null out the corresponding
1050	<	* workers slot, and if terminating, try to terminate, else try to
1051	<	* shrink workers array.
1052	<	* @param w the worker
1053	<	*/
1054	<	final void workerTerminated(ForkJoinWorkerThread w) {
1055	<	updateStealCount(w);
1056	<	updateWorkerCount(-1);
1057	<	final ReentrantLock lock = this.workerLock;
1058	<	lock.lock();
1059	<	try {
1060	<	ForkJoinWorkerThread[] ws = workers;
1061	<	int idx = w.poolIndex;
1062	<	if (idx >= 0 && idx < ws.length && ws[idx] == w)
1063	<	ws[idx] = null;
1064	<	if (totalCountOf(workerCounts) == 0) {
1065	<	terminate(); // no-op if already terminating
1066	<	transitionRunStateTo(TERMINATED);
1067	<	termination.signalAll();
1068	<	}
1069	<	else if (!isTerminating()) {
1070	<	tryShrinkWorkerArray();
1071	<	tryResumeSpare(true); // allow replacement
1072	<	}
1073	<	} finally {
1074	<	lock.unlock();
1075	<	}
1076	<	signalIdleWorkers();
1077	<	}
1078	<
1079	<	/**
1080	<	* Initiate termination.
1081	<	*/
1082	<	private void terminate() {
1083	<	if (transitionRunStateTo(TERMINATING)) {
1084	<	stopAllWorkers();
1085	<	resumeAllSpares();
1086	<	signalIdleWorkers();
1087	<	cancelQueuedSubmissions();
1088	<	cancelQueuedWorkerTasks();
1089	<	interruptUnterminatedWorkers();
1090	<	signalIdleWorkers(); // resignal after interrupt
1091	<	}
1092	<	}
1093	<
1094	<	/**
1095	<	* Possibly terminate when on shutdown state
1096	<	*/
1097	<	private void terminateOnShutdown() {
1098	<	if (!hasQueuedSubmissions() && canTerminateOnShutdown(runControl))
1099	<	terminate();
1100	<	}
1101	<
1102	<	/**
1103	<	* Clear out and cancel submissions
1104	<	*/
1105	<	private void cancelQueuedSubmissions() {
1106	<	ForkJoinTask<?> task;
1107	<	while ((task = pollSubmission()) != null)
1108	<	task.cancel(false);
1109	<	}
1110	<
1111	<	/**
1112	<	* Clean out worker queues.
1113	<	*/
1114	<	private void cancelQueuedWorkerTasks() {
1115	<	final ReentrantLock lock = this.workerLock;
1116	<	lock.lock();
1117	<	try {
1118	<	ForkJoinWorkerThread[] ws = workers;
1119	<	for (int i = 0; i < ws.length; ++i) {
1120	<	ForkJoinWorkerThread t = ws[i];
1121	<	if (t != null)
1122	<	t.cancelTasks();
1123	<	}
1124	<	} finally {
1125	<	lock.unlock();
1126	<	}
1127	<	}
1128	<
1129	<	/**
1130	<	* Set each worker's status to terminating. Requires lock to avoid
1131	<	* conflicts with add/remove
1132	<	*/
1133	<	private void stopAllWorkers() {
1134	<	final ReentrantLock lock = this.workerLock;
1135	<	lock.lock();
1136	<	try {
1137	<	ForkJoinWorkerThread[] ws = workers;
1138	<	for (int i = 0; i < ws.length; ++i) {
1139	<	ForkJoinWorkerThread t = ws[i];
1140	<	if (t != null)
1141	<	t.shutdownNow();
1142	<	}
1143	<	} finally {
1144	<	lock.unlock();
1145	<	}
1146	<	}
1147	<
1148	<	/**
1149	<	* Interrupt all unterminated workers.  This is not required for
1150	<	* sake of internal control, but may help unstick user code during
1151	<	* shutdown.
1152	<	*/
1153	<	private void interruptUnterminatedWorkers() {
1154	<	final ReentrantLock lock = this.workerLock;
1155	<	lock.lock();
1156	<	try {
1157	<	ForkJoinWorkerThread[] ws = workers;
1158	<	for (int i = 0; i < ws.length; ++i) {
1159	<	ForkJoinWorkerThread t = ws[i];
1160	<	if (t != null && !t.isTerminated()) {
1161	<	try {
1162	<	t.interrupt();
1163	<	} catch (SecurityException ignore) {
1164	<	}
1165	<	}
1166	<	}
1167	<	} finally {
1168	<	lock.unlock();
1169	<	}
1170	<	}
1171	<
1172	<
1173	<	/*
1174	<	* Nodes for event barrier to manage idle threads.  Queue nodes
1175	<	* are basic Treiber stack nodes, also used for spare stack.
1176	<	*
1177	<	* The event barrier has an event count and a wait queue (actually
1178	<	* a Treiber stack).  Workers are enabled to look for work when
1179	<	* the eventCount is incremented. If they fail to find work, they
1180	<	* may wait for next count. Upon release, threads help others wake
1181	<	* up.
1182	<	*
1183	<	* Synchronization events occur only in enough contexts to
1184	<	* maintain overall liveness:
1185	<	*
1186	<	*   - Submission of a new task to the pool
1187	<	*   - Resizes or other changes to the workers array
1188	<	*   - pool termination
1189	<	*   - A worker pushing a task on an empty queue
1190	<	*
1191	<	* The case of pushing a task occurs often enough, and is heavy
1192	<	* enough compared to simple stack pushes, to require special
1193	<	* handling: Method signalWork returns without advancing count if
1194	<	* the queue appears to be empty.  This would ordinarily result in
1195	<	* races causing some queued waiters not to be woken up. To avoid
1196	<	* this, the first worker enqueued in method sync (see
1197	<	* syncIsReleasable) rescans for tasks after being enqueued, and
1198	<	* helps signal if any are found. This works well because the
1199	<	* worker has nothing better to do, and so might as well help
1200	<	* alleviate the overhead and contention on the threads actually
1201	<	* doing work.  Also, since event counts increments on task
1202	<	* availability exist to maintain liveness (rather than to force
1203	<	* refreshes etc), it is OK for callers to exit early if
1204	<	* contending with another signaller.
1205	<	*/
1206	<	static final class WaitQueueNode {
1207	<	WaitQueueNode next; // only written before enqueued
1208	<	volatile ForkJoinWorkerThread thread; // nulled to cancel wait
1209	<	final long count; // unused for spare stack
1210	<
1211	<	WaitQueueNode(long c, ForkJoinWorkerThread w) {
1212	<	count = c;
1213	<	thread = w;
1214	<	}
1215	<
1216	<	/**
1217	<	* Wake up waiter, returning false if known to already
1218	<	*/
1219	<	boolean signal() {
1220	<	ForkJoinWorkerThread t = thread;
1221	<	if (t == null)
1222	<	return false;
1223	<	thread = null;
1224	<	LockSupport.unpark(t);
3135	>	if (isTerminated())
3136		return true;
3137	<	}
3138	<
3139	<	/**
3140	<	* Await release on sync
3141	<	*/
3142	<	void awaitSyncRelease(ForkJoinPool p) {
3143	<	while (thread != null && !p.syncIsReleasable(this))
1233	<	LockSupport.park(this);
1234	<	}
1235	<
1236	<	/**
1237	<	* Await resumption as spare
1238	<	*/
1239	<	void awaitSpareRelease() {
1240	<	while (thread != null) {
1241	<	if (!Thread.interrupted())
1242	<	LockSupport.park(this);
1243	<	}
1244	<	}
1245	<	}
1246	<
1247	<	/**
1248	<	* Ensures that no thread is waiting for count to advance from the
1249	<	* current value of eventCount read on entry to this method, by
1250	<	* releasing waiting threads if necessary.
1251	<	* @return the count
1252	<	*/
1253	<	final long ensureSync() {
1254	<	long c = eventCount;
1255	<	WaitQueueNode q;
1256	<	while ((q = syncStack) != null && q.count < c) {
1257	<	if (casBarrierStack(q, null)) {
1258	<	do {
1259	<	q.signal();
1260	<	} while ((q = q.next) != null);
1261	<	break;
1262	<	}
1263	<	}
1264	<	return c;
1265	<	}
1266	<
1267	<	/**
1268	<	* Increments event count and releases waiting threads.
1269	<	*/
1270	<	private void signalIdleWorkers() {
1271	<	long c;
1272	<	do;while (!casEventCount(c = eventCount, c+1));
1273	<	ensureSync();
1274	<	}
1275	<
1276	<	/**
1277	<	* Signal threads waiting to poll a task. Because method sync
1278	<	* rechecks availability, it is OK to only proceed if queue
1279	<	* appears to be non-empty, and OK to skip under contention to
1280	<	* increment count (since some other thread succeeded).
1281	<	*/
1282	<	final void signalWork() {
1283	<	long c;
1284	<	WaitQueueNode q;
1285	<	if (syncStack != null &&
1286	<	casEventCount(c = eventCount, c+1) &&
1287	<	(((q = syncStack) != null && q.count <= c) &&
1288	<	(!casBarrierStack(q, q.next) || !q.signal())))
1289	<	ensureSync();
1290	<	}
1291	<
1292	<	/**
1293	<	* Waits until event count advances from last value held by
1294	<	* caller, or if excess threads, caller is resumed as spare, or
1295	<	* caller or pool is terminating. Updates caller's event on exit.
1296	<	* @param w the calling worker thread
1297	<	*/
1298	<	final void sync(ForkJoinWorkerThread w) {
1299	<	updateStealCount(w); // Transfer w's count while it is idle
1300	<
1301	<	while (!w.isShutdown() && !isTerminating() && !suspendIfSpare(w)) {
1302	<	long prev = w.lastEventCount;
1303	<	WaitQueueNode node = null;
1304	<	WaitQueueNode h;
1305	<	while (eventCount == prev &&
1306	<	((h = syncStack) == null || h.count == prev)) {
1307	<	if (node == null)
1308	<	node = new WaitQueueNode(prev, w);
1309	<	if (casBarrierStack(node.next = h, node)) {
1310	<	node.awaitSyncRelease(this);
1311	<	break;
1312	<	}
1313	<	}
1314	<	long ec = ensureSync();
1315	<	if (ec != prev) {
1316	<	w.lastEventCount = ec;
1317	<	break;
1318	<	}
1319	<	}
1320	<	}
1321	<
1322	<	/**
1323	<	* Returns true if worker waiting on sync can proceed:
1324	<	*  - on signal (thread == null)
1325	<	*  - on event count advance (winning race to notify vs signaller)
1326	<	*  - on Interrupt
1327	<	*  - if the first queued node, we find work available
1328	<	* If node was not signalled and event count not advanced on exit,
1329	<	* then we also help advance event count.
1330	<	* @return true if node can be released
1331	<	*/
1332	<	final boolean syncIsReleasable(WaitQueueNode node) {
1333	<	long prev = node.count;
1334	<	if (!Thread.interrupted() && node.thread != null &&
1335	<	(node.next != null ||
1336	<	!ForkJoinWorkerThread.hasQueuedTasks(workers)) &&
1337	<	eventCount == prev)
1338	<	return false;
1339	<	if (node.thread != null) {
1340	<	node.thread = null;
1341	<	long ec = eventCount;
1342	<	if (prev <= ec) // help signal
1343	<	casEventCount(ec, ec+1);
1344	<	}
1345	<	return true;
1346	<	}
1347	<
1348	<	/**
1349	<	* Returns true if a new sync event occurred since last call to
1350	<	* sync or this method, if so, updating caller's count.
1351	<	*/
1352	<	final boolean hasNewSyncEvent(ForkJoinWorkerThread w) {
1353	<	long lc = w.lastEventCount;
1354	<	long ec = ensureSync();
1355	<	if (ec == lc)
1356	<	return false;
1357	<	w.lastEventCount = ec;
1358	<	return true;
1359	<	}
1360	<
1361	<	//  Parallelism maintenance
1362	<
1363	<	/**
1364	<	* Decrement running count; if too low, add spare.
1365	<	*
1366	<	* Conceptually, all we need to do here is add or resume a
1367	<	* spare thread when one is about to block (and remove or
1368	<	* suspend it later when unblocked -- see suspendIfSpare).
1369	<	* However, implementing this idea requires coping with
1370	<	* several problems: We have imperfect information about the
1371	<	* states of threads. Some count updates can and usually do
1372	<	* lag run state changes, despite arrangements to keep them
1373	<	* accurate (for example, when possible, updating counts
1374	<	* before signalling or resuming), especially when running on
1375	<	* dynamic JVMs that don't optimize the infrequent paths that
1376	<	* update counts. Generating too many threads can make these
1377	<	* problems become worse, because excess threads are more
1378	<	* likely to be context-switched with others, slowing them all
1379	<	* down, especially if there is no work available, so all are
1380	<	* busy scanning or idling.  Also, excess spare threads can
1381	<	* only be suspended or removed when they are idle, not
1382	<	* immediately when they aren't needed. So adding threads will
1383	<	* raise parallelism level for longer than necessary.  Also,
1384	<	* FJ applications often enounter highly transient peaks when
1385	<	* many threads are blocked joining, but for less time than it
1386	<	* takes to create or resume spares.
1387	<	*
1388	<	* @param joinMe if non-null, return early if done
1389	<	* @param maintainParallelism if true, try to stay within
1390	<	* target counts, else create only to avoid starvation
1391	<	* @return true if joinMe known to be done
1392	<	*/
1393	<	final boolean preJoin(ForkJoinTask<?> joinMe, boolean maintainParallelism) {
1394	<	maintainParallelism &= maintainsParallelism; // overrride
1395	<	boolean dec = false;  // true when running count decremented
1396	<	while (spareStack == null || !tryResumeSpare(dec)) {
1397	<	int counts = workerCounts;
1398	<	if (dec || (dec = casWorkerCounts(counts, --counts))) { // CAS cheat
1399	<	if (!needSpare(counts, maintainParallelism))
1400	<	break;
1401	<	if (joinMe.status < 0)
1402	<	return true;
1403	<	if (tryAddSpare(counts))
3137	>	long startTime = System.nanoTime();
3138	>	boolean terminated = false;
3139	>	synchronized (this) {
3140	>	for (long waitTime = nanos, millis = 0L;;) {
3141	>	if (terminated = isTerminated() ||
3142	>	waitTime <= 0L ||
3143	>	(millis = unit.toMillis(waitTime)) <= 0L)
3144		break;
3145	+	wait(millis);
3146	+	waitTime = nanos - (System.nanoTime() - startTime);
3147		}
3148		}
3149	<	return false;
1408	<	}
1409	<
1410	<	/**
1411	<	* Same idea as preJoin
1412	<	*/
1413	<	final boolean preBlock(ManagedBlocker blocker, boolean maintainParallelism){
1414	<	maintainParallelism &= maintainsParallelism;
1415	<	boolean dec = false;
1416	<	while (spareStack == null || !tryResumeSpare(dec)) {
1417	<	int counts = workerCounts;
1418	<	if (dec || (dec = casWorkerCounts(counts, --counts))) {
1419	<	if (!needSpare(counts, maintainParallelism))
1420	<	break;
1421	<	if (blocker.isReleasable())
1422	<	return true;
1423	<	if (tryAddSpare(counts))
1424	<	break;
1425	<	}
1426	<	}
1427	<	return false;
1428	<	}
1429	<
1430	<	/**
1431	<	* Returns true if a spare thread appears to be needed.  If
1432	<	* maintaining parallelism, returns true when the deficit in
1433	<	* running threads is more than the surplus of total threads, and
1434	<	* there is apparently some work to do.  This self-limiting rule
1435	<	* means that the more threads that have already been added, the
1436	<	* less parallelism we will tolerate before adding another.
1437	<	* @param counts current worker counts
1438	<	* @param maintainParallelism try to maintain parallelism
1439	<	*/
1440	<	private boolean needSpare(int counts, boolean maintainParallelism) {
1441	<	int ps = parallelism;
1442	<	int rc = runningCountOf(counts);
1443	<	int tc = totalCountOf(counts);
1444	<	int runningDeficit = ps - rc;
1445	<	int totalSurplus = tc - ps;
1446	<	return (tc < maxPoolSize &&
1447	<	(rc == 0 || totalSurplus < 0 ||
1448	<	(maintainParallelism &&
1449	<	runningDeficit > totalSurplus &&
1450	<	ForkJoinWorkerThread.hasQueuedTasks(workers))));
1451	<	}
1452	<
1453	<	/**
1454	<	* Add a spare worker if lock available and no more than the
1455	<	* expected numbers of threads exist
1456	<	* @return true if successful
1457	<	*/
1458	<	private boolean tryAddSpare(int expectedCounts) {
1459	<	final ReentrantLock lock = this.workerLock;
1460	<	int expectedRunning = runningCountOf(expectedCounts);
1461	<	int expectedTotal = totalCountOf(expectedCounts);
1462	<	boolean success = false;
1463	<	boolean locked = false;
1464	<	// confirm counts while locking; CAS after obtaining lock
1465	<	try {
1466	<	for (;;) {
1467	<	int s = workerCounts;
1468	<	int tc = totalCountOf(s);
1469	<	int rc = runningCountOf(s);
1470	<	if (rc > expectedRunning || tc > expectedTotal)
1471	<	break;
1472	<	if (!locked && !(locked = lock.tryLock()))
1473	<	break;
1474	<	if (casWorkerCounts(s, workerCountsFor(tc+1, rc+1))) {
1475	<	createAndStartSpare(tc);
1476	<	success = true;
1477	<	break;
1478	<	}
1479	<	}
1480	<	} finally {
1481	<	if (locked)
1482	<	lock.unlock();
1483	<	}
1484	<	return success;
1485	<	}
1486	<
1487	<	/**
1488	<	* Add the kth spare worker. On entry, pool coounts are already
1489	<	* adjusted to reflect addition.
1490	<	*/
1491	<	private void createAndStartSpare(int k) {
1492	<	ForkJoinWorkerThread w = null;
1493	<	ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(k + 1);
1494	<	int len = ws.length;
1495	<	// Probably, we can place at slot k. If not, find empty slot
1496	<	if (k < len && ws[k] != null) {
1497	<	for (k = 0; k < len && ws[k] != null; ++k)
1498	<	;
1499	<	}
1500	<	if (k < len && !isTerminating() && (w = createWorker(k)) != null) {
1501	<	ws[k] = w;
1502	<	w.start();
1503	<	}
1504	<	else
1505	<	updateWorkerCount(-1); // adjust on failure
1506	<	signalIdleWorkers();
1507	<	}
1508	<
1509	<	/**
1510	<	* Suspend calling thread w if there are excess threads.  Called
1511	<	* only from sync.  Spares are enqueued in a Treiber stack
1512	<	* using the same WaitQueueNodes as barriers.  They are resumed
1513	<	* mainly in preJoin, but are also woken on pool events that
1514	<	* require all threads to check run state.
1515	<	* @param w the caller
1516	<	*/
1517	<	private boolean suspendIfSpare(ForkJoinWorkerThread w) {
1518	<	WaitQueueNode node = null;
1519	<	int s;
1520	<	while (parallelism < runningCountOf(s = workerCounts)) {
1521	<	if (node == null)
1522	<	node = new WaitQueueNode(0, w);
1523	<	if (casWorkerCounts(s, s-1)) { // representation-dependent
1524	<	// push onto stack
1525	<	do;while (!casSpareStack(node.next = spareStack, node));
1526	<	// block until released by resumeSpare
1527	<	node.awaitSpareRelease();
1528	<	return true;
1529	<	}
1530	<	}
1531	<	return false;
1532	<	}
1533	<
1534	<	/**
1535	<	* Try to pop and resume a spare thread.
1536	<	* @param updateCount if true, increment running count on success
1537	<	* @return true if successful
1538	<	*/
1539	<	private boolean tryResumeSpare(boolean updateCount) {
1540	<	WaitQueueNode q;
1541	<	while ((q = spareStack) != null) {
1542	<	if (casSpareStack(q, q.next)) {
1543	<	if (updateCount)
1544	<	updateRunningCount(1);
1545	<	q.signal();
1546	<	return true;
1547	<	}
1548	<	}
1549	<	return false;
1550	<	}
1551	<
1552	<	/**
1553	<	* Pop and resume all spare threads. Same idea as ensureSync.
1554	<	* @return true if any spares released
1555	<	*/
1556	<	private boolean resumeAllSpares() {
1557	<	WaitQueueNode q;
1558	<	while ( (q = spareStack) != null) {
1559	<	if (casSpareStack(q, null)) {
1560	<	do {
1561	<	updateRunningCount(1);
1562	<	q.signal();
1563	<	} while ((q = q.next) != null);
1564	<	return true;
1565	<	}
1566	<	}
1567	<	return false;
1568	<	}
1569	<
1570	<	/**
1571	<	* Pop and shutdown excessive spare threads. Call only while
1572	<	* holding lock. This is not guaranteed to eliminate all excess
1573	<	* threads, only those suspended as spares, which are the ones
1574	<	* unlikely to be needed in the future.
1575	<	*/
1576	<	private void trimSpares() {
1577	<	int surplus = totalCountOf(workerCounts) - parallelism;
1578	<	WaitQueueNode q;
1579	<	while (surplus > 0 && (q = spareStack) != null) {
1580	<	if (casSpareStack(q, null)) {
1581	<	do {
1582	<	updateRunningCount(1);
1583	<	ForkJoinWorkerThread w = q.thread;
1584	<	if (w != null && surplus > 0 &&
1585	<	runningCountOf(workerCounts) > 0 && w.shutdown())
1586	<	--surplus;
1587	<	q.signal();
1588	<	} while ((q = q.next) != null);
1589	<	}
1590	<	}
3149	>	return terminated;
3150		}
3151
3152		/**
3153		* Interface for extending managed parallelism for tasks running
3154	<	* in ForkJoinPools. A ManagedBlocker provides two methods.
3155	<	* Method <code>isReleasable</code> must return true if blocking is not
3156	<	* necessary. Method <code>block</code> blocks the current thread
3157	<	* if necessary (perhaps internally invoking isReleasable before
3158	<	* actually blocking.).
3154	>	* in {@link ForkJoinPool}s.
3155	>	*
3156	>	* <p>A {@code ManagedBlocker} provides two methods.  Method
3157	>	* {@code isReleasable} must return {@code true} if blocking is
3158	>	* not necessary. Method {@code block} blocks the current thread
3159	>	* if necessary (perhaps internally invoking {@code isReleasable}
3160	>	* before actually blocking). These actions are performed by any
3161	>	* thread invoking {@link ForkJoinPool#managedBlock}.  The
3162	>	* unusual methods in this API accommodate synchronizers that may,
3163	>	* but don't usually, block for long periods. Similarly, they
3164	>	* allow more efficient internal handling of cases in which
3165	>	* additional workers may be, but usually are not, needed to
3166	>	* ensure sufficient parallelism.  Toward this end,
3167	>	* implementations of method {@code isReleasable} must be amenable
3168	>	* to repeated invocation.
3169	>	*
3170		* <p>For example, here is a ManagedBlocker based on a
3171		* ReentrantLock:
3172	<	* <pre>
3173	<	*   class ManagedLocker implements ManagedBlocker {
3174	<	*     final ReentrantLock lock;
3175	<	*     boolean hasLock = false;
3176	<	*     ManagedLocker(ReentrantLock lock) { this.lock = lock; }
3177	<	*     public boolean block() {
3178	<	*        if (!hasLock)
3179	<	*           lock.lock();
3180	<	*        return true;
3181	<	*     }
3182	<	*     public boolean isReleasable() {
3183	<	*        return hasLock || (hasLock = lock.tryLock());
3184	<	*     }
3172	>	*  <pre> {@code
3173	>	* class ManagedLocker implements ManagedBlocker {
3174	>	*   final ReentrantLock lock;
3175	>	*   boolean hasLock = false;
3176	>	*   ManagedLocker(ReentrantLock lock) { this.lock = lock; }
3177	>	*   public boolean block() {
3178	>	*     if (!hasLock)
3179	>	*       lock.lock();
3180	>	*     return true;
3181	>	*   }
3182	>	*   public boolean isReleasable() {
3183	>	*     return hasLock || (hasLock = lock.tryLock());
3184	>	*   }
3185	>	* }}</pre>
3186	>	*
3187	>	* <p>Here is a class that possibly blocks waiting for an
3188	>	* item on a given queue:
3189	>	*  <pre> {@code
3190	>	* class QueueTaker<E> implements ManagedBlocker {
3191	>	*   final BlockingQueue<E> queue;
3192	>	*   volatile E item = null;
3193	>	*   QueueTaker(BlockingQueue<E> q) { this.queue = q; }
3194	>	*   public boolean block() throws InterruptedException {
3195	>	*     if (item == null)
3196	>	*       item = queue.take();
3197	>	*     return true;
3198	>	*   }
3199	>	*   public boolean isReleasable() {
3200	>	*     return item != null || (item = queue.poll()) != null;
3201	>	*   }
3202	>	*   public E getItem() { // call after pool.managedBlock completes
3203	>	*     return item;
3204		*   }
3205	<	* </pre>
3205	>	* }}</pre>
3206		*/
3207		public static interface ManagedBlocker {
3208		/**
3209		* Possibly blocks the current thread, for example waiting for
3210		* a lock or condition.
3211	<	* @return true if no additional blocking is necessary (i.e.,
3212	<	* if isReleasable would return true).
3211	>	*
3212	>	* @return {@code true} if no additional blocking is necessary
3213	>	* (i.e., if isReleasable would return true)
3214		* @throws InterruptedException if interrupted while waiting
3215	<	* (the method is not required to do so, but is allowe to).
3215	>	* (the method is not required to do so, but is allowed to)
3216		*/
3217		boolean block() throws InterruptedException;
3218
3219		/**
3220	<	* Returns true if blocking is unnecessary.
3220	>	* Returns {@code true} if blocking is unnecessary.
3221		*/
3222		boolean isReleasable();
3223		}
3224
3225		/**
3226		* Blocks in accord with the given blocker.  If the current thread
3227	<	* is a ForkJoinWorkerThread, this method possibly arranges for a
3228	<	* spare thread to be activated if necessary to ensure parallelism
3229	<	* while the current thread is blocked.  If
3230	<	* <code>maintainParallelism</code> is true and the pool supports
3231	<	* it ({@link #getMaintainsParallelism}), this method attempts to
3232	<	* maintain the pool's nominal parallelism. Otherwise if activates
3233	<	* a thread only if necessary to avoid complete starvation. This
3234	<	* option may be preferable when blockages use timeouts, or are
3235	<	* almost always brief.
3236	<	*
3237	<	* <p> If the caller is not a ForkJoinTask, this method is behaviorally
3238	<	* equivalent to
3239	<	* <pre>
3240	<	*   while (!blocker.isReleasable())
1651	<	*      if (blocker.block())
1652	<	*         return;
1653	<	* </pre>
1654	<	* If the caller is a ForkJoinTask, then the pool may first
1655	<	* be expanded to ensure parallelism, and later adjusted.
3227	>	* is a {@link ForkJoinWorkerThread}, this method possibly
3228	>	* arranges for a spare thread to be activated if necessary to
3229	>	* ensure sufficient parallelism while the current thread is blocked.
3230	>	*
3231	>	* <p>If the caller is not a {@link ForkJoinTask}, this method is
3232	>	* behaviorally equivalent to
3233	>	*  <pre> {@code
3234	>	* while (!blocker.isReleasable())
3235	>	*   if (blocker.block())
3236	>	*     return;
3237	>	* }</pre>
3238	>	*
3239	>	* If the caller is a {@code ForkJoinTask}, then the pool may
3240	>	* first be expanded to ensure parallelism, and later adjusted.
3241		*
3242		* @param blocker the blocker
3243	<	* @param maintainParallelism if true and supported by this pool,
1659	<	* attempt to maintain the pool's nominal parallelism; otherwise
1660	<	* activate a thread only if necessary to avoid complete
1661	<	* starvation.
1662	<	* @throws InterruptedException if blocker.block did so.
3243	>	* @throws InterruptedException if blocker.block did so
3244		*/
3245	<	public static void managedBlock(ManagedBlocker blocker,
1665	<	boolean maintainParallelism)
3245	>	public static void managedBlock(ManagedBlocker blocker)
3246		throws InterruptedException {
3247		Thread t = Thread.currentThread();
3248	<	ForkJoinPool pool = (t instanceof ForkJoinWorkerThread?
3249	<	((ForkJoinWorkerThread)t).pool : null);
3250	<	if (!blocker.isReleasable()) {
3251	<	try {
3252	<	if (pool == null ||
3253	<	!pool.preBlock(blocker, maintainParallelism))
3254	<	awaitBlocker(blocker);
3255	<	} finally {
3256	<	if (pool != null)
3257	<	pool.updateRunningCount(1);
3248	>	if (t instanceof ForkJoinWorkerThread) {
3249	>	ForkJoinPool p = ((ForkJoinWorkerThread)t).pool;
3250	>	while (!blocker.isReleasable()) { // variant of helpSignal
3251	>	WorkQueue[] ws; WorkQueue q; int m, u;
3252	>	if ((ws = p.workQueues) != null && (m = ws.length - 1) >= 0) {
3253	>	for (int i = 0; i <= m; ++i) {
3254	>	if (blocker.isReleasable())
3255	>	return;
3256	>	if ((q = ws[i]) != null && q.base - q.top < 0) {
3257	>	p.signalWork(q);
3258	>	if ((u = (int)(p.ctl >>> 32)) >= 0 ||
3259	>	(u >> UAC_SHIFT) >= 0)
3260	>	break;
3261	>	}
3262	>	}
3263	>	}
3264	>	if (p.tryCompensate()) {
3265	>	try {
3266	>	do {} while (!blocker.isReleasable() &&
3267	>	!blocker.block());
3268	>	} finally {
3269	>	p.incrementActiveCount();
3270	>	}
3271	>	break;
3272	>	}
3273		}
3274		}
3275	+	else {
3276	+	do {} while (!blocker.isReleasable() &&
3277	+	!blocker.block());
3278	+	}
3279		}
3280
3281	<	private static void awaitBlocker(ManagedBlocker blocker)
3282	<	throws InterruptedException {
3283	<	do;while (!blocker.isReleasable() && !blocker.block());
1685	<	}
1686	<
1687	<	// AbstractExecutorService overrides
3281	>	// AbstractExecutorService overrides.  These rely on undocumented
3282	>	// fact that ForkJoinTask.adapt returns ForkJoinTasks that also
3283	>	// implement RunnableFuture.
3284
3285		protected <T> RunnableFuture<T> newTaskFor(Runnable runnable, T value) {
3286	<	return new AdaptedRunnable(runnable, value);
3286	>	return new ForkJoinTask.AdaptedRunnable<T>(runnable, value);
3287		}
3288
3289		protected <T> RunnableFuture<T> newTaskFor(Callable<T> callable) {
3290	<	return new AdaptedCallable(callable);
3290	>	return new ForkJoinTask.AdaptedCallable<T>(callable);
3291		}
3292
3293	+	// Unsafe mechanics
3294	+	private static final sun.misc.Unsafe U;
3295	+	private static final long CTL;
3296	+	private static final long PARKBLOCKER;
3297	+	private static final int ABASE;
3298	+	private static final int ASHIFT;
3299	+	private static final long STEALCOUNT;
3300	+	private static final long PLOCK;
3301	+	private static final long INDEXSEED;
3302	+	private static final long QLOCK;
3303
3304	<	// Temporary Unsafe mechanics for preliminary release
3305	<	private static Unsafe getUnsafe() throws Throwable {
3304	>	static {
3305	>	int s; // initialize field offsets for CAS etc
3306		try {
3307	<	return Unsafe.getUnsafe();
3307	>	U = getUnsafe();
3308	>	Class<?> k = ForkJoinPool.class;
3309	>	CTL = U.objectFieldOffset
3310	>	(k.getDeclaredField("ctl"));
3311	>	STEALCOUNT = U.objectFieldOffset
3312	>	(k.getDeclaredField("stealCount"));
3313	>	PLOCK = U.objectFieldOffset
3314	>	(k.getDeclaredField("plock"));
3315	>	INDEXSEED = U.objectFieldOffset
3316	>	(k.getDeclaredField("indexSeed"));
3317	>	Class<?> tk = Thread.class;
3318	>	PARKBLOCKER = U.objectFieldOffset
3319	>	(tk.getDeclaredField("parkBlocker"));
3320	>	Class<?> wk = WorkQueue.class;
3321	>	QLOCK = U.objectFieldOffset
3322	>	(wk.getDeclaredField("qlock"));
3323	>	Class<?> ak = ForkJoinTask[].class;
3324	>	ABASE = U.arrayBaseOffset(ak);
3325	>	s = U.arrayIndexScale(ak);
3326	>	ASHIFT = 31 - Integer.numberOfLeadingZeros(s);
3327	>	} catch (Exception e) {
3328	>	throw new Error(e);
3329	>	}
3330	>	if ((s & (s-1)) != 0)
3331	>	throw new Error("data type scale not a power of two");
3332	>
3333	>	submitters = new ThreadLocal<Submitter>();
3334	>	ForkJoinWorkerThreadFactory fac = defaultForkJoinWorkerThreadFactory =
3335	>	new DefaultForkJoinWorkerThreadFactory();
3336	>	modifyThreadPermission = new RuntimePermission("modifyThread");
3337	>
3338	>	/*
3339	>	* Establish common pool parameters.  For extra caution,
3340	>	* computations to set up common pool state are here; the
3341	>	* constructor just assigns these values to fields.
3342	>	*/
3343	>
3344	>	int par = 0;
3345	>	Thread.UncaughtExceptionHandler handler = null;
3346	>	try {  // TBD: limit or report ignored exceptions?
3347	>	String pp = System.getProperty
3348	>	("java.util.concurrent.ForkJoinPool.common.parallelism");
3349	>	String hp = System.getProperty
3350	>	("java.util.concurrent.ForkJoinPool.common.exceptionHandler");
3351	>	String fp = System.getProperty
3352	>	("java.util.concurrent.ForkJoinPool.common.threadFactory");
3353	>	if (fp != null)
3354	>	fac = ((ForkJoinWorkerThreadFactory)ClassLoader.
3355	>	getSystemClassLoader().loadClass(fp).newInstance());
3356	>	if (hp != null)
3357	>	handler = ((Thread.UncaughtExceptionHandler)ClassLoader.
3358	>	getSystemClassLoader().loadClass(hp).newInstance());
3359	>	if (pp != null)
3360	>	par = Integer.parseInt(pp);
3361	>	} catch (Exception ignore) {
3362	>	}
3363	>
3364	>	if (par <= 0)
3365	>	par = Runtime.getRuntime().availableProcessors();
3366	>	if (par > MAX_CAP)
3367	>	par = MAX_CAP;
3368	>	commonPoolParallelism = par;
3369	>	long np = (long)(-par); // precompute initial ctl value
3370	>	long ct = ((np << AC_SHIFT) & AC_MASK) | ((np << TC_SHIFT) & TC_MASK);
3371	>
3372	>	commonPool = new ForkJoinPool(par, ct, fac, handler);
3373	>	}
3374	>
3375	>	/**
3376	>	* Returns a sun.misc.Unsafe.  Suitable for use in a 3rd party package.
3377	>	* Replace with a simple call to Unsafe.getUnsafe when integrating
3378	>	* into a jdk.
3379	>	*
3380	>	* @return a sun.misc.Unsafe
3381	>	*/
3382	>	private static sun.misc.Unsafe getUnsafe() {
3383	>	try {
3384	>	return sun.misc.Unsafe.getUnsafe();
3385		} catch (SecurityException se) {
3386		try {
3387		return java.security.AccessController.doPrivileged
3388	<	(new java.security.PrivilegedExceptionAction<Unsafe>() {
3389	<	public Unsafe run() throws Exception {
3390	<	return getUnsafePrivileged();
3388	>	(new java.security
3389	>	.PrivilegedExceptionAction<sun.misc.Unsafe>() {
3390	>	public sun.misc.Unsafe run() throws Exception {
3391	>	java.lang.reflect.Field f = sun.misc
3392	>	.Unsafe.class.getDeclaredField("theUnsafe");
3393	>	f.setAccessible(true);
3394	>	return (sun.misc.Unsafe) f.get(null);
3395		}});
3396		} catch (java.security.PrivilegedActionException e) {
3397	<	throw e.getCause();
3397	>	throw new RuntimeException("Could not initialize intrinsics",
3398	>	e.getCause());
3399		}
3400		}
3401		}
3402
1715	–	private static Unsafe getUnsafePrivileged()
1716	–	throws NoSuchFieldException, IllegalAccessException {
1717	–	Field f = Unsafe.class.getDeclaredField("theUnsafe");
1718	–	f.setAccessible(true);
1719	–	return (Unsafe) f.get(null);
1720	–	}
1721	–
1722	–	private static long fieldOffset(String fieldName)
1723	–	throws NoSuchFieldException {
1724	–	return _unsafe.objectFieldOffset
1725	–	(ForkJoinPool.class.getDeclaredField(fieldName));
1726	–	}
1727	–
1728	–	static final Unsafe _unsafe;
1729	–	static final long eventCountOffset;
1730	–	static final long workerCountsOffset;
1731	–	static final long runControlOffset;
1732	–	static final long syncStackOffset;
1733	–	static final long spareStackOffset;
1734	–
1735	–	static {
1736	–	try {
1737	–	_unsafe = getUnsafe();
1738	–	eventCountOffset = fieldOffset("eventCount");
1739	–	workerCountsOffset = fieldOffset("workerCounts");
1740	–	runControlOffset = fieldOffset("runControl");
1741	–	syncStackOffset = fieldOffset("syncStack");
1742	–	spareStackOffset = fieldOffset("spareStack");
1743	–	} catch (Throwable e) {
1744	–	throw new RuntimeException("Could not initialize intrinsics", e);
1745	–	}
1746	–	}
1747	–
1748	–	private boolean casEventCount(long cmp, long val) {
1749	–	return _unsafe.compareAndSwapLong(this, eventCountOffset, cmp, val);
1750	–	}
1751	–	private boolean casWorkerCounts(int cmp, int val) {
1752	–	return _unsafe.compareAndSwapInt(this, workerCountsOffset, cmp, val);
1753	–	}
1754	–	private boolean casRunControl(int cmp, int val) {
1755	–	return _unsafe.compareAndSwapInt(this, runControlOffset, cmp, val);
1756	–	}
1757	–	private boolean casSpareStack(WaitQueueNode cmp, WaitQueueNode val) {
1758	–	return _unsafe.compareAndSwapObject(this, spareStackOffset, cmp, val);
1759	–	}
1760	–	private boolean casBarrierStack(WaitQueueNode cmp, WaitQueueNode val) {
1761	–	return _unsafe.compareAndSwapObject(this, syncStackOffset, cmp, val);
1762	–	}
3403		}

Diff Legend

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