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 1.85 by dl, Sun Nov 21 13:55:04 2010 UTC vs.
Revision 1.100 by dl, Fri Apr 1 20:20:37 2011 UTC

#	Line 1 | Line 1
1		/*
2		* Written by Doug Lea with assistance from members of JCP JSR-166
3		* Expert Group and released to the public domain, as explained at
4	<	* http://creativecommons.org/licenses/publicdomain
4	>	* http://creativecommons.org/publicdomain/zero/1.0/
5		*/
6
7		package jsr166y;
#	Line 11 | Line 11 | import java.util.Arrays;
11		import java.util.Collection;
12		import java.util.Collections;
13		import java.util.List;
14	+	import java.util.Random;
15		import java.util.concurrent.AbstractExecutorService;
16		import java.util.concurrent.Callable;
17		import java.util.concurrent.ExecutorService;
#	Line 22 | Line 23 | import java.util.concurrent.TimeoutExcep
23		import java.util.concurrent.atomic.AtomicInteger;
24		import java.util.concurrent.locks.LockSupport;
25		import java.util.concurrent.locks.ReentrantLock;
26	+	import java.util.concurrent.locks.Condition;
27
28		/**
29		* An {@link ExecutorService} for running {@link ForkJoinTask}s.
#	Line 129 | Line 131 | public class ForkJoinPool extends Abstra
131		* set of worker threads: Submissions from non-FJ threads enter
132		* into a submission queue. Workers take these tasks and typically
133		* split them into subtasks that may be stolen by other workers.
134	<	* The main work-stealing mechanics implemented in class
135	<	* ForkJoinWorkerThread give first priority to processing tasks
136	<	* from their own queues (LIFO or FIFO, depending on mode), then
137	<	* to randomized FIFO steals of tasks in other worker queues, and
138	<	* lastly to new submissions. These mechanics do not consider
139	<	* affinities, loads, cache localities, etc, so rarely provide the
140	<	* best possible performance on a given machine, but portably
141	<	* provide good throughput by averaging over these factors.
142	<	* (Further, even if we did try to use such information, we do not
143	<	* usually have a basis for exploiting it. For example, some sets
144	<	* of tasks profit from cache affinities, but others are harmed by
145	<	* cache pollution effects.)
146	<	*
147	<	* Beyond work-stealing support and essential bookkeeping, the
148	<	* main responsibility of this framework is to take actions when
149	<	* one worker is waiting to join a task stolen (or always held by)
150	<	* another.  Because we are multiplexing many tasks on to a pool
151	<	* of workers, we can't just let them block (as in Thread.join).
152	<	* We also cannot just reassign the joiner's run-time stack with
153	<	* another and replace it later, which would be a form of
154	<	* "continuation", that even if possible is not necessarily a good
155	<	* idea. Given that the creation costs of most threads on most
156	<	* systems mainly surrounds setting up runtime stacks, thread
157	<	* creation and switching is usually not much more expensive than
158	<	* stack creation and switching, and is more flexible). Instead we
134	>	* Preference rules give first priority to processing tasks from
135	>	* their own queues (LIFO or FIFO, depending on mode), then to
136	>	* randomized FIFO steals of tasks in other worker queues, and
137	>	* lastly to new submissions.
138	>	*
139	>	* The main throughput advantages of work-stealing stem from
140	>	* decentralized control -- workers mostly take tasks from
141	>	* themselves or each other. We cannot negate this in the
142	>	* implementation of other management responsibilities. The main
143	>	* tactic for avoiding bottlenecks is packing nearly all
144	>	* essentially atomic control state into a single 64bit volatile
145	>	* variable ("ctl"). This variable is read on the order of 10-100
146	>	* times as often as it is modified (always via CAS). (There is
147	>	* some additional control state, for example variable "shutdown"
148	>	* for which we can cope with uncoordinated updates.)  This
149	>	* streamlines synchronization and control at the expense of messy
150	>	* constructions needed to repack status bits upon updates.
151	>	* Updates tend not to contend with each other except during
152	>	* bursts while submitted tasks begin or end.  In some cases when
153	>	* they do contend, threads can instead do something else
154	>	* (usually, scan for tasks) until contention subsides.
155	>	*
156	>	* To enable packing, we restrict maximum parallelism to (1<<15)-1
157	>	* (which is far in excess of normal operating range) to allow
158	>	* ids, counts, and their negations (used for thresholding) to fit
159	>	* into 16bit fields.
160	>	*
161	>	* Recording Workers.  Workers are recorded in the "workers" array
162	>	* that is created upon pool construction and expanded if (rarely)
163	>	* necessary.  This is an array as opposed to some other data
164	>	* structure to support index-based random steals by workers.
165	>	* Updates to the array recording new workers and unrecording
166	>	* terminated ones are protected from each other by a seqLock
167	>	* (scanGuard) but the array is otherwise concurrently readable,
168	>	* and accessed directly by workers. To simplify index-based
169	>	* operations, the array size is always a power of two, and all
170	>	* readers must tolerate null slots. To avoid flailing during
171	>	* start-up, the array is presized to hold twice #parallelism
172	>	* workers (which is unlikely to need further resizing during
173	>	* execution). But to avoid dealing with so many null slots,
174	>	* variable scanGuard includes a mask for the nearest power of two
175	>	* that contains all current workers.  All worker thread creation
176	>	* is on-demand, triggered by task submissions, replacement of
177	>	* terminated workers, and/or compensation for blocked
178	>	* workers. However, all other support code is set up to work with
179	>	* other policies.  To ensure that we do not hold on to worker
180	>	* references that would prevent GC, ALL accesses to workers are
181	>	* via indices into the workers array (which is one source of some
182	>	* of the messy code constructions here). In essence, the workers
183	>	* array serves as a weak reference mechanism. Thus for example
184	>	* the wait queue field of ctl stores worker indices, not worker
185	>	* references.  Access to the workers in associated methods (for
186	>	* example signalWork) must both index-check and null-check the
187	>	* IDs. All such accesses ignore bad IDs by returning out early
188	>	* from what they are doing, since this can only be associated
189	>	* with termination, in which case it is OK to give up.
190	>	*
191	>	* All uses of the workers array, as well as queue arrays, check
192	>	* that the array is non-null (even if previously non-null). This
193	>	* allows nulling during termination, which is currently not
194	>	* necessary, but remains an option for resource-revocation-based
195	>	* shutdown schemes.
196	>	*
197	>	* Wait Queuing. Unlike HPC work-stealing frameworks, we cannot
198	>	* let workers spin indefinitely scanning for tasks when none can
199	>	* be found immediately, and we cannot start/resume workers unless
200	>	* there appear to be tasks available.  On the other hand, we must
201	>	* quickly prod them into action when new tasks are submitted or
202	>	* generated.  We park/unpark workers after placing in an event
203	>	* wait queue when they cannot find work. This "queue" is actually
204	>	* a simple Treiber stack, headed by the "id" field of ctl, plus a
205	>	* 15bit counter value to both wake up waiters (by advancing their
206	>	* count) and avoid ABA effects. Successors are held in worker
207	>	* field "nextWait".  Queuing deals with several intrinsic races,
208	>	* mainly that a task-producing thread can miss seeing (and
209	>	* signalling) another thread that gave up looking for work but
210	>	* has not yet entered the wait queue. We solve this by requiring
211	>	* a full sweep of all workers both before (in scan()) and after
212	>	* (in tryAwaitWork()) a newly waiting worker is added to the wait
213	>	* queue. During a rescan, the worker might release some other
214	>	* queued worker rather than itself, which has the same net
215	>	* effect. Because enqueued workers may actually be rescanning
216	>	* rather than waiting, we set and clear the "parked" field of
217	>	* ForkJoinWorkerThread to reduce unnecessary calls to unpark.
218	>	* (Use of the parked field requires a secondary recheck to avoid
219	>	* missed signals.)
220	>	*
221	>	* Signalling.  We create or wake up workers only when there
222	>	* appears to be at least one task they might be able to find and
223	>	* execute.  When a submission is added or another worker adds a
224	>	* task to a queue that previously had two or fewer tasks, they
225	>	* signal waiting workers (or trigger creation of new ones if
226	>	* fewer than the given parallelism level -- see signalWork).
227	>	* These primary signals are buttressed by signals during rescans
228	>	* as well as those performed when a worker steals a task and
229	>	* notices that there are more tasks too; together these cover the
230	>	* signals needed in cases when more than two tasks are pushed
231	>	* but untaken.
232	>	*
233	>	* Trimming workers. To release resources after periods of lack of
234	>	* use, a worker starting to wait when the pool is quiescent will
235	>	* time out and terminate if the pool has remained quiescent for
236	>	* SHRINK_RATE nanosecs. This will slowly propagate, eventually
237	>	* terminating all workers after long periods of non-use.
238	>	*
239	>	* Submissions. External submissions are maintained in an
240	>	* array-based queue that is structured identically to
241	>	* ForkJoinWorkerThread queues except for the use of
242	>	* submissionLock in method addSubmission. Unlike the case for
243	>	* worker queues, multiple external threads can add new
244	>	* submissions, so adding requires a lock.
245	>	*
246	>	* Compensation. Beyond work-stealing support and lifecycle
247	>	* control, the main responsibility of this framework is to take
248	>	* actions when one worker is waiting to join a task stolen (or
249	>	* always held by) another.  Because we are multiplexing many
250	>	* tasks on to a pool of workers, we can't just let them block (as
251	>	* in Thread.join).  We also cannot just reassign the joiner's
252	>	* run-time stack with another and replace it later, which would
253	>	* be a form of "continuation", that even if possible is not
254	>	* necessarily a good idea since we sometimes need both an
255	>	* unblocked task and its continuation to progress. Instead we
256		* combine two tactics:
257		*
258		*   Helping: Arranging for the joiner to execute some task that it
259		*      would be running if the steal had not occurred.  Method
260	<	*      ForkJoinWorkerThread.helpJoinTask tracks joining->stealing
260	>	*      ForkJoinWorkerThread.joinTask tracks joining->stealing
261		*      links to try to find such a task.
262		*
263		*   Compensating: Unless there are already enough live threads,
264	<	*      method helpMaintainParallelism() may create or
265	<	*      re-activate a spare thread to compensate for blocked
266	<	*      joiners until they unblock.
168	<	*
169	<	* It is impossible to keep exactly the target (parallelism)
170	<	* number of threads running at any given time.  Determining
171	<	* existence of conservatively safe helping targets, the
172	<	* availability of already-created spares, and the apparent need
173	<	* to create new spares are all racy and require heuristic
174	<	* guidance, so we rely on multiple retries of each.  Compensation
175	<	* occurs in slow-motion. It is triggered only upon timeouts of
176	<	* Object.wait used for joins. This reduces poor decisions that
177	<	* would otherwise be made when threads are waiting for others
178	<	* that are stalled because of unrelated activities such as
179	<	* garbage collection.
264	>	*      method tryPreBlock() may create or re-activate a spare
265	>	*      thread to compensate for blocked joiners until they
266	>	*      unblock.
267		*
268		* The ManagedBlocker extension API can't use helping so relies
269		* only on compensation in method awaitBlocker.
270		*
271	<	* The main throughput advantages of work-stealing stem from
272	<	* decentralized control -- workers mostly steal tasks from each
273	<	* other. We do not want to negate this by creating bottlenecks
274	<	* implementing other management responsibilities. So we use a
275	<	* collection of techniques that avoid, reduce, or cope well with
276	<	* contention. These entail several instances of bit-packing into
277	<	* CASable fields to maintain only the minimally required
278	<	* atomicity. To enable such packing, we restrict maximum
279	<	* parallelism to (1<<15)-1 (enabling twice this (to accommodate
280	<	* unbalanced increments and decrements) to fit into a 16 bit
281	<	* field, which is far in excess of normal operating range.  Even
282	<	* though updates to some of these bookkeeping fields do sometimes
283	<	* contend with each other, they don't normally cache-contend with
284	<	* updates to others enough to warrant memory padding or
285	<	* isolation. So they are all held as fields of ForkJoinPool
286	<	* objects.  The main capabilities are as follows:
287	<	*
288	<	* 1. Creating and removing workers. Workers are recorded in the
289	<	* "workers" array. This is an array as opposed to some other data
290	<	* structure to support index-based random steals by workers.
291	<	* Updates to the array recording new workers and unrecording
292	<	* terminated ones are protected from each other by a lock
293	<	* (workerLock) but the array is otherwise concurrently readable,
294	<	* and accessed directly by workers. To simplify index-based
295	<	* operations, the array size is always a power of two, and all
296	<	* readers must tolerate null slots. Currently, all worker thread
297	<	* creation is on-demand, triggered by task submissions,
298	<	* replacement of terminated workers, and/or compensation for
299	<	* blocked workers. However, all other support code is set up to
213	<	* work with other policies.
214	<	*
215	<	* To ensure that we do not hold on to worker references that
216	<	* would prevent GC, ALL accesses to workers are via indices into
217	<	* the workers array (which is one source of some of the unusual
218	<	* code constructions here). In essence, the workers array serves
219	<	* as a WeakReference mechanism. Thus for example the event queue
220	<	* stores worker indices, not worker references. Access to the
221	<	* workers in associated methods (for example releaseEventWaiters)
222	<	* must both index-check and null-check the IDs. All such accesses
223	<	* ignore bad IDs by returning out early from what they are doing,
224	<	* since this can only be associated with shutdown, in which case
225	<	* it is OK to give up. On termination, we just clobber these
226	<	* data structures without trying to use them.
227	<	*
228	<	* 2. Bookkeeping for dynamically adding and removing workers. We
229	<	* aim to approximately maintain the given level of parallelism.
230	<	* When some workers are known to be blocked (on joins or via
231	<	* ManagedBlocker), we may create or resume others to take their
232	<	* place until they unblock (see below). Implementing this
233	<	* requires counts of the number of "running" threads (i.e., those
234	<	* that are neither blocked nor artificially suspended) as well as
235	<	* the total number.  These two values are packed into one field,
236	<	* "workerCounts" because we need accurate snapshots when deciding
237	<	* to create, resume or suspend.  Note however that the
238	<	* correspondence of these counts to reality is not guaranteed. In
239	<	* particular updates for unblocked threads may lag until they
240	<	* actually wake up.
241	<	*
242	<	* 3. Maintaining global run state. The run state of the pool
243	<	* consists of a runLevel (SHUTDOWN, TERMINATING, etc) similar to
244	<	* those in other Executor implementations, as well as a count of
245	<	* "active" workers -- those that are, or soon will be, or
246	<	* recently were executing tasks. The runLevel and active count
247	<	* are packed together in order to correctly trigger shutdown and
248	<	* termination. Without care, active counts can be subject to very
249	<	* high contention.  We substantially reduce this contention by
250	<	* relaxing update rules.  A worker must claim active status
251	<	* prospectively, by activating if it sees that a submitted or
252	<	* stealable task exists (it may find after activating that the
253	<	* task no longer exists). It stays active while processing this
254	<	* task (if it exists) and any other local subtasks it produces,
255	<	* until it cannot find any other tasks. It then tries
256	<	* inactivating (see method preStep), but upon update contention
257	<	* instead scans for more tasks, later retrying inactivation if it
258	<	* doesn't find any.
259	<	*
260	<	* 4. Managing idle workers waiting for tasks. We cannot let
261	<	* workers spin indefinitely scanning for tasks when none are
262	<	* available. On the other hand, we must quickly prod them into
263	<	* action when new tasks are submitted or generated.  We
264	<	* park/unpark these idle workers using an event-count scheme.
265	<	* Field eventCount is incremented upon events that may enable
266	<	* workers that previously could not find a task to now find one:
267	<	* Submission of a new task to the pool, or another worker pushing
268	<	* a task onto a previously empty queue.  (We also use this
269	<	* mechanism for configuration and termination actions that
270	<	* require wakeups of idle workers).  Each worker maintains its
271	<	* last known event count, and blocks when a scan for work did not
272	<	* find a task AND its lastEventCount matches the current
273	<	* eventCount. Waiting idle workers are recorded in a variant of
274	<	* Treiber stack headed by field eventWaiters which, when nonzero,
275	<	* encodes the thread index and count awaited for by the worker
276	<	* thread most recently calling eventSync. This thread in turn has
277	<	* a record (field nextEventWaiter) for the next waiting worker.
278	<	* In addition to allowing simpler decisions about need for
279	<	* wakeup, the event count bits in eventWaiters serve the role of
280	<	* tags to avoid ABA errors in Treiber stacks. Upon any wakeup,
281	<	* released threads also try to release at most two others.  The
282	<	* net effect is a tree-like diffusion of signals, where released
283	<	* threads (and possibly others) help with unparks.  To further
284	<	* reduce contention effects a bit, failed CASes to increment
285	<	* field eventCount are tolerated without retries in signalWork.
286	<	* Conceptually they are merged into the same event, which is OK
287	<	* when their only purpose is to enable workers to scan for work.
288	<	*
289	<	* 5. Managing suspension of extra workers. When a worker notices
290	<	* (usually upon timeout of a wait()) that there are too few
291	<	* running threads, we may create a new thread to maintain
292	<	* parallelism level, or at least avoid starvation. Usually, extra
293	<	* threads are needed for only very short periods, yet join
294	<	* dependencies are such that we sometimes need them in
295	<	* bursts. Rather than create new threads each time this happens,
296	<	* we suspend no-longer-needed extra ones as "spares". For most
297	<	* purposes, we don't distinguish "extra" spare threads from
298	<	* normal "core" threads: On each call to preStep (the only point
299	<	* at which we can do this) a worker checks to see if there are
300	<	* now too many running workers, and if so, suspends itself.
301	<	* Method helpMaintainParallelism looks for suspended threads to
302	<	* resume before considering creating a new replacement. The
303	<	* spares themselves are encoded on another variant of a Treiber
304	<	* Stack, headed at field "spareWaiters".  Note that the use of
305	<	* spares is intrinsically racy.  One thread may become a spare at
306	<	* about the same time as another is needlessly being created. We
307	<	* counteract this and related slop in part by requiring resumed
308	<	* spares to immediately recheck (in preStep) to see whether they
309	<	* should re-suspend.
310	<	*
311	<	* 6. Killing off unneeded workers. A timeout mechanism is used to
312	<	* shed unused workers: The oldest (first) event queue waiter uses
313	<	* a timed rather than hard wait. When this wait times out without
314	<	* a normal wakeup, it tries to shutdown any one (for convenience
315	<	* the newest) other spare or event waiter via
316	<	* tryShutdownUnusedWorker. This eventually reduces the number of
317	<	* worker threads to a minimum of one after a long enough period
318	<	* without use.
319	<	*
320	<	* 7. Deciding when to create new workers. The main dynamic
321	<	* control in this class is deciding when to create extra threads
322	<	* in method helpMaintainParallelism. We would like to keep
323	<	* exactly #parallelism threads running, which is an impossible
324	<	* task. We always need to create one when the number of running
325	<	* threads would become zero and all workers are busy. Beyond
326	<	* this, we must rely on heuristics that work well in the
327	<	* presence of transient phenomena such as GC stalls, dynamic
328	<	* compilation, and wake-up lags. These transients are extremely
329	<	* common -- we are normally trying to fully saturate the CPUs on
330	<	* a machine, so almost any activity other than running tasks
331	<	* impedes accuracy. Our main defense is to allow parallelism to
332	<	* lapse for a while during joins, and use a timeout to see if,
333	<	* after the resulting settling, there is still a need for
334	<	* additional workers.  This also better copes with the fact that
335	<	* some of the methods in this class tend to never become compiled
336	<	* (but are interpreted), so some components of the entire set of
337	<	* controls might execute 100 times faster than others. And
338	<	* similarly for cases where the apparent lack of work is just due
339	<	* to GC stalls and other transient system activity.
271	>	* It is impossible to keep exactly the target parallelism number
272	>	* of threads running at any given time.  Determining the
273	>	* existence of conservatively safe helping targets, the
274	>	* availability of already-created spares, and the apparent need
275	>	* to create new spares are all racy and require heuristic
276	>	* guidance, so we rely on multiple retries of each.  Currently,
277	>	* in keeping with on-demand signalling policy, we compensate only
278	>	* if blocking would leave less than one active (non-waiting,
279	>	* non-blocked) worker. Additionally, to avoid some false alarms
280	>	* due to GC, lagging counters, system activity, etc, compensated
281	>	* blocking for joins is only attempted after rechecks stabilize
282	>	* (retries are interspersed with Thread.yield, for good
283	>	* citizenship).  The variable blockedCount, incremented before
284	>	* blocking and decremented after, is sometimes needed to
285	>	* distinguish cases of waiting for work vs blocking on joins or
286	>	* other managed sync. Both cases are equivalent for most pool
287	>	* control, so we can update non-atomically. (Additionally,
288	>	* contention on blockedCount alleviates some contention on ctl).
289	>	*
290	>	* Shutdown and Termination. A call to shutdownNow atomically sets
291	>	* the ctl stop bit and then (non-atomically) sets each workers
292	>	* "terminate" status, cancels all unprocessed tasks, and wakes up
293	>	* all waiting workers.  Detecting whether termination should
294	>	* commence after a non-abrupt shutdown() call requires more work
295	>	* and bookkeeping. We need consensus about quiesence (i.e., that
296	>	* there is no more work) which is reflected in active counts so
297	>	* long as there are no current blockers, as well as possible
298	>	* re-evaluations during independent changes in blocking or
299	>	* quiescing workers.
300		*
301	<	* Beware that there is a lot of representation-level coupling
301	>	* Style notes: There is a lot of representation-level coupling
302		* among classes ForkJoinPool, ForkJoinWorkerThread, and
303	<	* ForkJoinTask.  For example, direct access to "workers" array by
303	>	* ForkJoinTask.  Most fields of ForkJoinWorkerThread maintain
304	>	* data structures managed by ForkJoinPool, so are directly
305	>	* accessed.  Conversely we allow access to "workers" array by
306		* workers, and direct access to ForkJoinTask.status by both
307		* ForkJoinPool and ForkJoinWorkerThread.  There is little point
308		* trying to reduce this, since any associated future changes in
309		* representations will need to be accompanied by algorithmic
310	<	* changes anyway.
311	<	*
312	<	* Style notes: There are lots of inline assignments (of form
313	<	* "while ((local = field) != 0)") which are usually the simplest
314	<	* way to ensure the required read orderings (which are sometimes
315	<	* critical). Also several occurrences of the unusual "do {}
316	<	* while (!cas...)" which is the simplest way to force an update of
317	<	* a CAS'ed variable. There are also other coding oddities that
318	<	* help some methods perform reasonably even when interpreted (not
319	<	* compiled), at the expense of some messy constructions that
320	<	* reduce byte code counts.
321	<	*
322	<	* The order of declarations in this file is: (1) statics (2)
323	<	* fields (along with constants used when unpacking some of them)
324	<	* (3) internal control methods (4) callbacks and other support
325	<	* for ForkJoinTask and ForkJoinWorkerThread classes, (5) exported
326	<	* methods (plus a few little helpers).
310	>	* changes anyway. All together, these low-level implementation
311	>	* choices produce as much as a factor of 4 performance
312	>	* improvement compared to naive implementations, and enable the
313	>	* processing of billions of tasks per second, at the expense of
314	>	* some ugliness.
315	>	*
316	>	* Methods signalWork() and scan() are the main bottlenecks so are
317	>	* especially heavily micro-optimized/mangled.  There are lots of
318	>	* inline assignments (of form "while ((local = field) != 0)")
319	>	* which are usually the simplest way to ensure the required read
320	>	* orderings (which are sometimes critical). This leads to a
321	>	* "C"-like style of listing declarations of these locals at the
322	>	* heads of methods or blocks.  There are several occurrences of
323	>	* the unusual "do {} while (!cas...)"  which is the simplest way
324	>	* to force an update of a CAS'ed variable. There are also other
325	>	* coding oddities that help some methods perform reasonably even
326	>	* when interpreted (not compiled).
327	>	*
328	>	* The order of declarations in this file is: (1) declarations of
329	>	* statics (2) fields (along with constants used when unpacking
330	>	* some of them), listed in an order that tends to reduce
331	>	* contention among them a bit under most JVMs.  (3) internal
332	>	* control methods (4) callbacks and other support for
333	>	* ForkJoinTask and ForkJoinWorkerThread classes, (5) exported
334	>	* methods (plus a few little helpers). (6) static block
335	>	* initializing all statics in a minimally dependent order.
336		*/
337
338		/**
#	Line 396 | Line 367 | public class ForkJoinPool extends Abstra
367		* overridden in ForkJoinPool constructors.
368		*/
369		public static final ForkJoinWorkerThreadFactory
370	<	defaultForkJoinWorkerThreadFactory =
400	<	new DefaultForkJoinWorkerThreadFactory();
370	>	defaultForkJoinWorkerThreadFactory;
371
372		/**
373		* Permission required for callers of methods that may start or
374		* kill threads.
375		*/
376	<	private static final RuntimePermission modifyThreadPermission =
407	<	new RuntimePermission("modifyThread");
376	>	private static final RuntimePermission modifyThreadPermission;
377
378		/**
379		* If there is a security manager, makes sure caller has
#	Line 419 | Line 388 | public class ForkJoinPool extends Abstra
388		/**
389		* Generator for assigning sequence numbers as pool names.
390		*/
391	<	private static final AtomicInteger poolNumberGenerator =
423	<	new AtomicInteger();
391	>	private static final AtomicInteger poolNumberGenerator;
392
393		/**
394	<	* The time to block in a join (see awaitJoin) before checking if
395	<	* a new worker should be (re)started to maintain parallelism
396	<	* level. The value should be short enough to maintain global
397	<	* responsiveness and progress but long enough to avoid
398	<	* counterproductive firings during GC stalls or unrelated system
431	<	* activity, and to not bog down systems with continual re-firings
432	<	* on GCs or legitimately long waits.
394	>	* Generator for initial random seeds for worker victim
395	>	* selection. This is used only to create initial seeds. Random
396	>	* steals use a cheaper xorshift generator per steal attempt. We
397	>	* don't expect much contention on seedGenerator, so just use a
398	>	* plain Random.
399		*/
400	<	private static final long JOIN_TIMEOUT_MILLIS = 250L; // 4 per second
400	>	static final Random workerSeedGenerator;
401
402		/**
403	<	* The wakeup interval (in nanoseconds) for the oldest worker
404	<	* waiting for an event to invoke tryShutdownUnusedWorker to
405	<	* shrink the number of workers.  The exact value does not matter
406	<	* too much. It must be short enough to release resources during
407	<	* sustained periods of idleness, but not so short that threads
408	<	* are continually re-created.
403	>	* Array holding all worker threads in the pool.  Initialized upon
404	>	* construction. Array size must be a power of two.  Updates and
405	>	* replacements are protected by scanGuard, but the array is
406	>	* always kept in a consistent enough state to be randomly
407	>	* accessed without locking by workers performing work-stealing,
408	>	* as well as other traversal-based methods in this class, so long
409	>	* as reads memory-acquire by first reading ctl. All readers must
410	>	* tolerate that some array slots may be null.
411		*/
412	<	private static final long SHRINK_RATE_NANOS =
445	<	30L * 1000L * 1000L * 1000L; // 2 per minute
412	>	ForkJoinWorkerThread[] workers;
413
414		/**
415	<	* Absolute bound for parallelism level. Twice this number plus
416	<	* one (i.e., 0xfff) must fit into a 16bit field to enable
417	<	* word-packing for some counts and indices.
415	>	* Initial size for submission queue array. Must be a power of
416	>	* two.  In many applications, these always stay small so we use a
417	>	* small initial cap.
418		*/
419	<	private static final int MAX_WORKERS   = 0x7fff;
419	>	private static final int INITIAL_QUEUE_CAPACITY = 8;
420
421		/**
422	<	* Array holding all worker threads in the pool.  Array size must
423	<	* be a power of two.  Updates and replacements are protected by
424	<	* workerLock, but the array is always kept in a consistent enough
425	<	* state to be randomly accessed without locking by workers
459	<	* performing work-stealing, as well as other traversal-based
460	<	* methods in this class. All readers must tolerate that some
461	<	* array slots may be null.
422	>	* Maximum size for submission queue array. Must be a power of two
423	>	* less than or equal to 1 << (31 - width of array entry) to
424	>	* ensure lack of index wraparound, but is capped at a lower
425	>	* value to help users trap runaway computations.
426		*/
427	<	volatile ForkJoinWorkerThread[] workers;
427	>	private static final int MAXIMUM_QUEUE_CAPACITY = 1 << 24; // 16M
428
429		/**
430	<	* Queue for external submissions.
430	>	* Array serving as submission queue. Initialized upon construction.
431		*/
432	<	private final LinkedTransferQueue<ForkJoinTask<?>> submissionQueue;
432	>	private ForkJoinTask<?>[] submissionQueue;
433
434		/**
435	<	* Lock protecting updates to workers array.
435	>	* Lock protecting submissions array for addSubmission
436		*/
437	<	private final ReentrantLock workerLock;
437	>	private final ReentrantLock submissionLock;
438
439		/**
440	<	* Latch released upon termination.
440	>	* Condition for awaitTermination, using submissionLock for
441	>	* convenience.
442		*/
443	<	private final Phaser termination;
443	>	private final Condition termination;
444
445		/**
446		* Creation factory for worker threads.
#	Line 483 | Line 448 | public class ForkJoinPool extends Abstra
448		private final ForkJoinWorkerThreadFactory factory;
449
450		/**
451	<	* Sum of per-thread steal counts, updated only when threads are
452	<	* idle or terminating.
451	>	* The uncaught exception handler used when any worker abruptly
452	>	* terminates.
453		*/
454	<	private volatile long stealCount;
454	>	final Thread.UncaughtExceptionHandler ueh;
455
456		/**
457	<	* Encoded record of top of Treiber stack of threads waiting for
493	<	* events. The top 32 bits contain the count being waited for. The
494	<	* bottom 16 bits contains one plus the pool index of waiting
495	<	* worker thread. (Bits 16-31 are unused.)
457	>	* Prefix for assigning names to worker threads
458		*/
459	<	private volatile long eventWaiters;
498	<
499	<	private static final int  EVENT_COUNT_SHIFT = 32;
500	<	private static final long WAITER_ID_MASK    = (1L << 16) - 1L;
459	>	private final String workerNamePrefix;
460
461		/**
462	<	* A counter for events that may wake up worker threads:
463	<	*   - Submission of a new task to the pool
505	<	*   - A worker pushing a task on an empty queue
506	<	*   - termination
462	>	* Sum of per-thread steal counts, updated only when threads are
463	>	* idle or terminating.
464		*/
465	<	private volatile int eventCount;
465	>	private volatile long stealCount;
466
467		/**
468	<	* Encoded record of top of Treiber stack of spare threads waiting
469	<	* for resumption. The top 16 bits contain an arbitrary count to
470	<	* avoid ABA effects. The bottom 16bits contains one plus the pool
471	<	* index of waiting worker thread.
472	<	*/
473	<	private volatile int spareWaiters;
474	<
475	<	private static final int SPARE_COUNT_SHIFT = 16;
476	<	private static final int SPARE_ID_MASK     = (1 << 16) - 1;
468	>	* Main pool control -- a long packed with:
469	>	* AC: Number of active running workers minus target parallelism (16 bits)
470	>	* TC: Number of total workers minus target parallelism (16bits)
471	>	* ST: true if pool is terminating (1 bit)
472	>	* EC: the wait count of top waiting thread (15 bits)
473	>	* ID: ~poolIndex of top of Treiber stack of waiting threads (16 bits)
474	>	*
475	>	* When convenient, we can extract the upper 32 bits of counts and
476	>	* the lower 32 bits of queue state, u = (int)(ctl >>> 32) and e =
477	>	* (int)ctl.  The ec field is never accessed alone, but always
478	>	* together with id and st. The offsets of counts by the target
479	>	* parallelism and the positionings of fields makes it possible to
480	>	* perform the most common checks via sign tests of fields: When
481	>	* ac is negative, there are not enough active workers, when tc is
482	>	* negative, there are not enough total workers, when id is
483	>	* negative, there is at least one waiting worker, and when e is
484	>	* negative, the pool is terminating.  To deal with these possibly
485	>	* negative fields, we use casts in and out of "short" and/or
486	>	* signed shifts to maintain signedness.
487	>	*/
488	>	volatile long ctl;
489	>
490	>	// bit positions/shifts for fields
491	>	private static final int  AC_SHIFT   = 48;
492	>	private static final int  TC_SHIFT   = 32;
493	>	private static final int  ST_SHIFT   = 31;
494	>	private static final int  EC_SHIFT   = 16;
495	>
496	>	// bounds
497	>	private static final int  MAX_ID     = 0x7fff;  // max poolIndex
498	>	private static final int  SMASK      = 0xffff;  // mask short bits
499	>	private static final int  SHORT_SIGN = 1 << 15;
500	>	private static final int  INT_SIGN   = 1 << 31;
501	>
502	>	// masks
503	>	private static final long STOP_BIT   = 0x0001L << ST_SHIFT;
504	>	private static final long AC_MASK    = ((long)SMASK) << AC_SHIFT;
505	>	private static final long TC_MASK    = ((long)SMASK) << TC_SHIFT;
506	>
507	>	// units for incrementing and decrementing
508	>	private static final long TC_UNIT    = 1L << TC_SHIFT;
509	>	private static final long AC_UNIT    = 1L << AC_SHIFT;
510	>
511	>	// masks and units for dealing with u = (int)(ctl >>> 32)
512	>	private static final int  UAC_SHIFT  = AC_SHIFT - 32;
513	>	private static final int  UTC_SHIFT  = TC_SHIFT - 32;
514	>	private static final int  UAC_MASK   = SMASK << UAC_SHIFT;
515	>	private static final int  UTC_MASK   = SMASK << UTC_SHIFT;
516	>	private static final int  UAC_UNIT   = 1 << UAC_SHIFT;
517	>	private static final int  UTC_UNIT   = 1 << UTC_SHIFT;
518	>
519	>	// masks and units for dealing with e = (int)ctl
520	>	private static final int  E_MASK     = 0x7fffffff; // no STOP_BIT
521	>	private static final int  EC_UNIT    = 1 << EC_SHIFT;
522
523		/**
524	<	* Lifecycle control. The low word contains the number of workers
523	<	* that are (probably) executing tasks. This value is atomically
524	<	* incremented before a worker gets a task to run, and decremented
525	<	* when a worker has no tasks and cannot find any.  Bits 16-18
526	<	* contain runLevel value. When all are zero, the pool is
527	<	* running. Level transitions are monotonic (running -> shutdown
528	<	* -> terminating -> terminated) so each transition adds a bit.
529	<	* These are bundled together to ensure consistent read for
530	<	* termination checks (i.e., that runLevel is at least SHUTDOWN
531	<	* and active threads is zero).
532	<	*
533	<	* Notes: Most direct CASes are dependent on these bitfield
534	<	* positions.  Also, this field is non-private to enable direct
535	<	* performance-sensitive CASes in ForkJoinWorkerThread.
524	>	* The target parallelism level.
525		*/
526	<	volatile int runState;
538	<
539	<	// Note: The order among run level values matters.
540	<	private static final int RUNLEVEL_SHIFT     = 16;
541	<	private static final int SHUTDOWN           = 1 << RUNLEVEL_SHIFT;
542	<	private static final int TERMINATING        = 1 << (RUNLEVEL_SHIFT + 1);
543	<	private static final int TERMINATED         = 1 << (RUNLEVEL_SHIFT + 2);
544	<	private static final int ACTIVE_COUNT_MASK  = (1 << RUNLEVEL_SHIFT) - 1;
526	>	final int parallelism;
527
528		/**
529	<	* Holds number of total (i.e., created and not yet terminated)
530	<	* and running (i.e., not blocked on joins or other managed sync)
531	<	* threads, packed together to ensure consistent snapshot when
532	<	* making decisions about creating and suspending spare
551	<	* threads. Updated only by CAS. Note that adding a new worker
552	<	* requires incrementing both counts, since workers start off in
553	<	* running state.
529	>	* Index (mod submission queue length) of next element to take
530	>	* from submission queue. Usage is identical to that for
531	>	* per-worker queues -- see ForkJoinWorkerThread internal
532	>	* documentation.
533		*/
534	<	private volatile int workerCounts;
534	>	volatile int queueBase;
535
536	<	private static final int TOTAL_COUNT_SHIFT  = 16;
537	<	private static final int RUNNING_COUNT_MASK = (1 << TOTAL_COUNT_SHIFT) - 1;
538	<	private static final int ONE_RUNNING        = 1;
539	<	private static final int ONE_TOTAL          = 1 << TOTAL_COUNT_SHIFT;
536	>	/**
537	>	* Index (mod submission queue length) of next element to add
538	>	* in submission queue. Usage is identical to that for
539	>	* per-worker queues -- see ForkJoinWorkerThread internal
540	>	* documentation.
541	>	*/
542	>	int queueTop;
543
544		/**
545	<	* The target parallelism level.
564	<	* Accessed directly by ForkJoinWorkerThreads.
545	>	* True when shutdown() has been called.
546		*/
547	<	final int parallelism;
547	>	volatile boolean shutdown;
548
549		/**
550		* True if use local fifo, not default lifo, for local polling
#	Line 572 | Line 553 | public class ForkJoinPool extends Abstra
553		final boolean locallyFifo;
554
555		/**
556	<	* The uncaught exception handler used when any worker abruptly
557	<	* terminates.
556	>	* The number of threads in ForkJoinWorkerThreads.helpQuiescePool.
557	>	* When non-zero, suppresses automatic shutdown when active
558	>	* counts become zero.
559		*/
560	<	private final Thread.UncaughtExceptionHandler ueh;
560	>	volatile int quiescerCount;
561
562		/**
563	<	* Pool number, just for assigning useful names to worker threads
563	>	* The number of threads blocked in join.
564		*/
565	<	private final int poolNumber;
584	<
585	<	// Utilities for CASing fields. Note that most of these
586	<	// are usually manually inlined by callers
565	>	volatile int blockedCount;
566
567		/**
568	<	* Increments running count part of workerCounts
568	>	* Counter for worker Thread names (unrelated to their poolIndex)
569		*/
570	<	final void incrementRunningCount() {
592	<	int c;
593	<	do {} while (!UNSAFE.compareAndSwapInt(this, workerCountsOffset,
594	<	c = workerCounts,
595	<	c + ONE_RUNNING));
596	<	}
570	>	private volatile int nextWorkerNumber;
571
572		/**
573	<	* Tries to increment running count part of workerCounts
573	>	* The index for the next created worker. Accessed under scanGuard.
574		*/
575	<	final boolean tryIncrementRunningCount() {
602	<	int c;
603	<	return UNSAFE.compareAndSwapInt(this, workerCountsOffset,
604	<	c = workerCounts,
605	<	c + ONE_RUNNING);
606	<	}
575	>	private int nextWorkerIndex;
576
577		/**
578	<	* Tries to decrement running count unless already zero
578	>	* SeqLock and index masking for updates to workers array.  Locked
579	>	* when SG_UNIT is set. Unlocking clears bit by adding
580	>	* SG_UNIT. Staleness of read-only operations can be checked by
581	>	* comparing scanGuard to value before the reads. The low 16 bits
582	>	* (i.e, anding with SMASK) hold (the smallest power of two
583	>	* covering all worker indices, minus one, and is used to avoid
584	>	* dealing with large numbers of null slots when the workers array
585	>	* is overallocated.
586		*/
587	<	final boolean tryDecrementRunningCount() {
612	<	int wc = workerCounts;
613	<	if ((wc & RUNNING_COUNT_MASK) == 0)
614	<	return false;
615	<	return UNSAFE.compareAndSwapInt(this, workerCountsOffset,
616	<	wc, wc - ONE_RUNNING);
617	<	}
587	>	volatile int scanGuard;
588
589	<	/**
620	<	* Forces decrement of encoded workerCounts, awaiting nonzero if
621	<	* (rarely) necessary when other count updates lag.
622	<	*
623	<	* @param dr -- either zero or ONE_RUNNING
624	<	* @param dt -- either zero or ONE_TOTAL
625	<	*/
626	<	private void decrementWorkerCounts(int dr, int dt) {
627	<	for (;;) {
628	<	int wc = workerCounts;
629	<	if ((wc & RUNNING_COUNT_MASK)  - dr < 0 ||
630	<	(wc >>> TOTAL_COUNT_SHIFT) - dt < 0) {
631	<	if ((runState & TERMINATED) != 0)
632	<	return; // lagging termination on a backout
633	<	Thread.yield();
634	<	}
635	<	if (UNSAFE.compareAndSwapInt(this, workerCountsOffset,
636	<	wc, wc - (dr + dt)))
637	<	return;
638	<	}
639	<	}
589	>	private static final int SG_UNIT = 1 << 16;
590
591		/**
592	<	* Tries decrementing active count; fails on contention.
593	<	* Called when workers cannot find tasks to run.
592	>	* The wakeup interval (in nanoseconds) for a worker waiting for a
593	>	* task when the pool is quiescent to instead try to shrink the
594	>	* number of workers.  The exact value does not matter too
595	>	* much. It must be short enough to release resources during
596	>	* sustained periods of idleness, but not so short that threads
597	>	* are continually re-created.
598		*/
599	<	final boolean tryDecrementActiveCount() {
600	<	int c;
647	<	return UNSAFE.compareAndSwapInt(this, runStateOffset,
648	<	c = runState, c - 1);
649	<	}
599	>	private static final long SHRINK_RATE =
600	>	4L * 1000L * 1000L * 1000L; // 4 seconds
601
602		/**
603	<	* Advances to at least the given level. Returns true if not
604	<	* already in at least the given level.
603	>	* Top-level loop for worker threads: On each step: if the
604	>	* previous step swept through all queues and found no tasks, or
605	>	* there are excess threads, then possibly blocks. Otherwise,
606	>	* scans for and, if found, executes a task. Returns when pool
607	>	* and/or worker terminate.
608	>	*
609	>	* @param w the worker
610		*/
611	<	private boolean advanceRunLevel(int level) {
612	<	for (;;) {
613	<	int s = runState;
614	<	if ((s & level) != 0)
615	<	return false;
616	<	if (UNSAFE.compareAndSwapInt(this, runStateOffset, s, s | level))
617	<	return true;
611	>	final void work(ForkJoinWorkerThread w) {
612	>	boolean swept = false;                // true on empty scans
613	>	long c;
614	>	while (!w.terminate && (int)(c = ctl) >= 0) {
615	>	int a;                            // active count
616	>	if (!swept && (a = (int)(c >> AC_SHIFT)) <= 0)
617	>	swept = scan(w, a);
618	>	else if (tryAwaitWork(w, c))
619	>	swept = false;
620		}
621		}
622
623	<	// workers array maintenance
623	>	// Signalling
624
625		/**
626	<	* Records and returns a workers array index for new worker.
626	>	* Wakes up or creates a worker.
627		*/
628	<	private int recordWorker(ForkJoinWorkerThread w) {
629	<	// Try using slot totalCount-1. If not available, scan and/or resize
630	<	int k = (workerCounts >>> TOTAL_COUNT_SHIFT) - 1;
631	<	final ReentrantLock lock = this.workerLock;
632	<	lock.lock();
633	<	try {
634	<	ForkJoinWorkerThread[] ws = workers;
635	<	int n = ws.length;
636	<	if (k < 0 || k >= n || ws[k] != null) {
637	<	for (k = 0; k < n && ws[k] != null; ++k)
638	<	;
639	<	if (k == n)
640	<	ws = workers = Arrays.copyOf(ws, n << 1);
641	<	}
642	<	ws[k] = w;
643	<	int c = eventCount; // advance event count to ensure visibility
644	<	UNSAFE.compareAndSwapInt(this, eventCountOffset, c, c+1);
645	<	} finally {
646	<	lock.unlock();
628	>	final void signalWork() {
629	>	/*
630	>	* The while condition is true if: (there is are too few total
631	>	* workers OR there is at least one waiter) AND (there are too
632	>	* few active workers OR the pool is terminating).  The value
633	>	* of e distinguishes the remaining cases: zero (no waiters)
634	>	* for create, negative if terminating (in which case do
635	>	* nothing), else release a waiter. The secondary checks for
636	>	* release (non-null array etc) can fail if the pool begins
637	>	* terminating after the test, and don't impose any added cost
638	>	* because JVMs must perform null and bounds checks anyway.
639	>	*/
640	>	long c; int e, u;
641	>	while ((((e = (int)(c = ctl)) | (u = (int)(c >>> 32))) &
642	>	(INT_SIGN|SHORT_SIGN)) == (INT_SIGN|SHORT_SIGN) && e >= 0) {
643	>	if (e > 0) {                         // release a waiting worker
644	>	int i; ForkJoinWorkerThread w; ForkJoinWorkerThread[] ws;
645	>	if ((ws = workers) == null ||
646	>	(i = ~e & SMASK) >= ws.length ||
647	>	(w = ws[i]) == null)
648	>	break;
649	>	long nc = (((long)(w.nextWait & E_MASK)) |
650	>	((long)(u + UAC_UNIT) << 32));
651	>	if (w.eventCount == e &&
652	>	UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc)) {
653	>	w.eventCount = (e + EC_UNIT) & E_MASK;
654	>	if (w.parked)
655	>	UNSAFE.unpark(w);
656	>	break;
657	>	}
658	>	}
659	>	else if (UNSAFE.compareAndSwapLong
660	>	(this, ctlOffset, c,
661	>	(long)(((u + UTC_UNIT) & UTC_MASK) |
662	>	((u + UAC_UNIT) & UAC_MASK)) << 32)) {
663	>	addWorker();
664	>	break;
665	>	}
666		}
690	–	return k;
667		}
668
669		/**
670	<	* Nulls out record of worker in workers array.
670	>	* Variant of signalWork to help release waiters on rescans.
671	>	* Tries once to release a waiter if active count < 0.
672	>	*
673	>	* @return false if failed due to contention, else true
674		*/
675	<	private void forgetWorker(ForkJoinWorkerThread w) {
676	<	int idx = w.poolIndex;
677	<	// Locking helps method recordWorker avoid unnecessary expansion
678	<	final ReentrantLock lock = this.workerLock;
679	<	lock.lock();
680	<	try {
681	<	ForkJoinWorkerThread[] ws = workers;
682	<	if (idx >= 0 && idx < ws.length && ws[idx] == w) // verify
683	<	ws[idx] = null;
684	<	} finally {
685	<	lock.unlock();
675	>	private boolean tryReleaseWaiter() {
676	>	long c; int e, i; ForkJoinWorkerThread w; ForkJoinWorkerThread[] ws;
677	>	if ((e = (int)(c = ctl)) > 0 &&
678	>	(int)(c >> AC_SHIFT) < 0 &&
679	>	(ws = workers) != null &&
680	>	(i = ~e & SMASK) < ws.length &&
681	>	(w = ws[i]) != null) {
682	>	long nc = ((long)(w.nextWait & E_MASK) |
683	>	((c + AC_UNIT) & (AC_MASK|TC_MASK)));
684	>	if (w.eventCount != e ||
685	>	!UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc))
686	>	return false;
687	>	w.eventCount = (e + EC_UNIT) & E_MASK;
688	>	if (w.parked)
689	>	UNSAFE.unpark(w);
690		}
691	+	return true;
692		}
693
694	+	// Scanning for tasks
695	+
696		/**
697	<	* Final callback from terminating worker.  Removes record of
698	<	* worker from array, and adjusts counts. If pool is shutting
699	<	* down, tries to complete termination.
697	>	* Scans for and, if found, executes one task. Scans start at a
698	>	* random index of workers array, and randomly select the first
699	>	* (2*#workers)-1 probes, and then, if all empty, resort to 2
700	>	* circular sweeps, which is necessary to check quiescence. and
701	>	* taking a submission only if no stealable tasks were found.  The
702	>	* steal code inside the loop is a specialized form of
703	>	* ForkJoinWorkerThread.deqTask, followed bookkeeping to support
704	>	* helpJoinTask and signal propagation. The code for submission
705	>	* queues is almost identical. On each steal, the worker completes
706	>	* not only the task, but also all local tasks that this task may
707	>	* have generated. On detecting staleness or contention when
708	>	* trying to take a task, this method returns without finishing
709	>	* sweep, which allows global state rechecks before retry.
710		*
711		* @param w the worker
712	+	* @param a the number of active workers
713	+	* @return true if swept all queues without finding a task
714		*/
715	<	final void workerTerminated(ForkJoinWorkerThread w) {
716	<	forgetWorker(w);
717	<	decrementWorkerCounts(w.isTrimmed() ? 0 : ONE_RUNNING, ONE_TOTAL);
720	<	while (w.stealCount != 0) // collect final count
721	<	tryAccumulateStealCount(w);
722	<	tryTerminate(false);
723	<	}
724	<
725	<	// Waiting for and signalling events
726	<
727	<	/**
728	<	* Releases workers blocked on a count not equal to current count.
729	<	* Normally called after precheck that eventWaiters isn't zero to
730	<	* avoid wasted array checks. Gives up upon a change in count or
731	<	* upon releasing two workers, letting others take over.
732	<	*/
733	<	private void releaseEventWaiters() {
715	>	private boolean scan(ForkJoinWorkerThread w, int a) {
716	>	int g = scanGuard; // mask 0 avoids useless scans if only one active
717	>	int m = (parallelism == 1 - a && blockedCount == 0) ? 0 : g & SMASK;
718		ForkJoinWorkerThread[] ws = workers;
719	<	int n = ws.length;
720	<	long h = eventWaiters;
721	<	int ec = eventCount;
722	<	boolean releasedOne = false;
723	<	ForkJoinWorkerThread w; int id;
724	<	while ((id = ((int)(h & WAITER_ID_MASK)) - 1) >= 0 &&
725	<	(int)(h >>> EVENT_COUNT_SHIFT) != ec &&
726	<	id < n && (w = ws[id]) != null) {
727	<	if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
728	<	h,  w.nextWaiter)) {
729	<	LockSupport.unpark(w);
730	<	if (releasedOne) // exit on second release
731	<	break;
732	<	releasedOne = true;
719	>	if (ws == null || ws.length <= m)         // staleness check
720	>	return false;
721	>	for (int r = w.seed, k = r, j = -(m + m); j <= m + m; ++j) {
722	>	ForkJoinTask<?> t; ForkJoinTask<?>[] q; int b, i;
723	>	ForkJoinWorkerThread v = ws[k & m];
724	>	if (v != null && (b = v.queueBase) != v.queueTop &&
725	>	(q = v.queue) != null && (i = (q.length - 1) & b) >= 0) {
726	>	long u = (i << ASHIFT) + ABASE;
727	>	if ((t = q[i]) != null && v.queueBase == b &&
728	>	UNSAFE.compareAndSwapObject(q, u, t, null)) {
729	>	int d = (v.queueBase = b + 1) - v.queueTop;
730	>	v.stealHint = w.poolIndex;
731	>	if (d != 0)
732	>	signalWork();             // propagate if nonempty
733	>	w.execTask(t);
734	>	}
735	>	r ^= r << 13; r ^= r >>> 17; w.seed = r ^ (r << 5);
736	>	return false;                     // store next seed
737		}
738	<	if (eventCount != ec)
739	<	break;
740	<	h = eventWaiters;
738	>	else if (j < 0) {                     // xorshift
739	>	r ^= r << 13; r ^= r >>> 17; k = r ^= r << 5;
740	>	}
741	>	else
742	>	++k;
743	>	}
744	>	if (scanGuard != g)                       // staleness check
745	>	return false;
746	>	else {                                    // try to take submission
747	>	ForkJoinTask<?> t; ForkJoinTask<?>[] q; int b, i;
748	>	if ((b = queueBase) != queueTop &&
749	>	(q = submissionQueue) != null &&
750	>	(i = (q.length - 1) & b) >= 0) {
751	>	long u = (i << ASHIFT) + ABASE;
752	>	if ((t = q[i]) != null && queueBase == b &&
753	>	UNSAFE.compareAndSwapObject(q, u, t, null)) {
754	>	queueBase = b + 1;
755	>	w.execTask(t);
756	>	}
757	>	return false;
758	>	}
759	>	return true;                         // all queues empty
760		}
761		}
762
763		/**
764	<	* Tries to advance eventCount and releases waiters. Called only
765	<	* from workers.
766	<	*/
767	<	final void signalWork() {
768	<	int c; // try to increment event count -- CAS failure OK
769	<	UNSAFE.compareAndSwapInt(this, eventCountOffset, c = eventCount, c+1);
770	<	if (eventWaiters != 0L)
771	<	releaseEventWaiters();
772	<	}
773	<
774	<	/**
775	<	* Adds the given worker to event queue and blocks until
776	<	* terminating or event count advances from the given value
777	<	*
778	<	* @param w the calling worker thread
779	<	* @param ec the count
780	<	*/
781	<	private void eventSync(ForkJoinWorkerThread w, int ec) {
782	<	long nh = (((long)ec) << EVENT_COUNT_SHIFT) | ((long)(w.poolIndex+1));
783	<	long h;
784	<	while ((runState < SHUTDOWN || !tryTerminate(false)) &&
785	<	(((int)((h = eventWaiters) & WAITER_ID_MASK)) == 0 ||
786	<	(int)(h >>> EVENT_COUNT_SHIFT) == ec) &&
787	<	eventCount == ec) {
788	<	if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
789	<	w.nextWaiter = h, nh)) {
790	<	awaitEvent(w, ec);
791	<	break;
764	>	* Tries to enqueue worker w in wait queue and await change in
765	>	* worker's eventCount.  If the pool is quiescent and there is
766	>	* more than one worker, possibly terminates worker upon exit.
767	>	* Otherwise, before blocking, rescans queues to avoid missed
768	>	* signals.  Upon finding work, releases at least one worker
769	>	* (which may be the current worker). Rescans restart upon
770	>	* detected staleness or failure to release due to
771	>	* contention. Note the unusual conventions about Thread.interrupt
772	>	* here and elsewhere: Because interrupts are used solely to alert
773	>	* threads to check termination, which is checked here anyway, we
774	>	* clear status (using Thread.interrupted) before any call to
775	>	* park, so that park does not immediately return due to status
776	>	* being set via some other unrelated call to interrupt in user
777	>	* code.
778	>	*
779	>	* @param w the calling worker
780	>	* @param c the ctl value on entry
781	>	* @return true if waited or another thread was released upon enq
782	>	*/
783	>	private boolean tryAwaitWork(ForkJoinWorkerThread w, long c) {
784	>	int v = w.eventCount;
785	>	w.nextWait = (int)c;                      // w's successor record
786	>	long nc = (long)(v & E_MASK) | ((c - AC_UNIT) & (AC_MASK|TC_MASK));
787	>	if (ctl != c || !UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc)) {
788	>	long d = ctl; // return true if lost to a deq, to force scan
789	>	return (int)d != (int)c && ((d - c) & AC_MASK) >= 0L;
790	>	}
791	>	for (int sc = w.stealCount; sc != 0;) {   // accumulate stealCount
792	>	long s = stealCount;
793	>	if (UNSAFE.compareAndSwapLong(this, stealCountOffset, s, s + sc))
794	>	sc = w.stealCount = 0;
795	>	else if (w.eventCount != v)
796	>	return true;                      // update next time
797	>	}
798	>	if ((int)c != 0 && parallelism + (int)(nc >> AC_SHIFT) == 0 &&
799	>	blockedCount == 0 && quiescerCount == 0)
800	>	idleAwaitWork(w, nc, c, v);           // quiescent
801	>	for (boolean rescanned = false;;) {
802	>	if (w.eventCount != v)
803	>	return true;
804	>	if (!rescanned) {
805	>	int g = scanGuard, m = g & SMASK;
806	>	ForkJoinWorkerThread[] ws = workers;
807	>	if (ws != null && m < ws.length) {
808	>	rescanned = true;
809	>	for (int i = 0; i <= m; ++i) {
810	>	ForkJoinWorkerThread u = ws[i];
811	>	if (u != null) {
812	>	if (u.queueBase != u.queueTop &&
813	>	!tryReleaseWaiter())
814	>	rescanned = false; // contended
815	>	if (w.eventCount != v)
816	>	return true;
817	>	}
818	>	}
819	>	}
820	>	if (scanGuard != g ||              // stale
821	>	(queueBase != queueTop && !tryReleaseWaiter()))
822	>	rescanned = false;
823	>	if (!rescanned)
824	>	Thread.yield();                // reduce contention
825	>	else
826	>	Thread.interrupted();          // clear before park
827	>	}
828	>	else {
829	>	w.parked = true;                   // must recheck
830	>	if (w.eventCount != v) {
831	>	w.parked = false;
832	>	return true;
833	>	}
834	>	LockSupport.park(this);
835	>	rescanned = w.parked = false;
836		}
837		}
838		}
839
840		/**
841	<	* Blocks the given worker (that has already been entered as an
842	<	* event waiter) until terminating or event count advances from
843	<	* the given value. The oldest (first) waiter uses a timed wait to
844	<	* occasionally one-by-one shrink the number of workers (to a
845	<	* minimum of one) if the pool has not been used for extended
846	<	* periods.
847	<	*
848	<	* @param w the calling worker thread
849	<	* @param ec the count
850	<	*/
851	<	private void awaitEvent(ForkJoinWorkerThread w, int ec) {
852	<	while (eventCount == ec) {
853	<	if (tryAccumulateStealCount(w)) { // transfer while idle
854	<	boolean untimed = (w.nextWaiter != 0L ||
855	<	(workerCounts & RUNNING_COUNT_MASK) <= 1);
856	<	long startTime = untimed ? 0 : System.nanoTime();
857	<	Thread.interrupted();         // clear/ignore interrupt
858	<	if (w.isTerminating() || eventCount != ec)
859	<	break;                    // recheck after clear
860	<	if (untimed)
861	<	LockSupport.park(w);
862	<	else {
863	<	LockSupport.parkNanos(w, SHRINK_RATE_NANOS);
864	<	if (eventCount != ec || w.isTerminating())
865	<	break;
866	<	if (System.nanoTime() - startTime >= SHRINK_RATE_NANOS)
867	<	tryShutdownUnusedWorker(ec);
841	>	* If inactivating worker w has caused pool to become
842	>	* quiescent, check for pool termination, and wait for event
843	>	* for up to SHRINK_RATE nanosecs (rescans are unnecessary in
844	>	* this case because quiescence reflects consensus about lack
845	>	* of work). On timeout, if ctl has not changed, terminate the
846	>	* worker. Upon its termination (see deregisterWorker), it may
847	>	* wake up another worker to possibly repeat this process.
848	>	*
849	>	* @param w the calling worker
850	>	* @param currentCtl the ctl value after enqueuing w
851	>	* @param prevCtl the ctl value if w terminated
852	>	* @param v the eventCount w awaits change
853	>	*/
854	>	private void idleAwaitWork(ForkJoinWorkerThread w, long currentCtl,
855	>	long prevCtl, int v) {
856	>	if (w.eventCount == v) {
857	>	if (shutdown)
858	>	tryTerminate(false);
859	>	ForkJoinTask.helpExpungeStaleExceptions(); // help clean weak refs
860	>	while (ctl == currentCtl) {
861	>	long startTime = System.nanoTime();
862	>	w.parked = true;
863	>	if (w.eventCount == v)             // must recheck
864	>	LockSupport.parkNanos(this, SHRINK_RATE);
865	>	w.parked = false;
866	>	if (w.eventCount != v)
867	>	break;
868	>	else if (System.nanoTime() - startTime <
869	>	SHRINK_RATE - (SHRINK_RATE / 10)) // timing slop
870	>	Thread.interrupted();          // spurious wakeup
871	>	else if (UNSAFE.compareAndSwapLong(this, ctlOffset,
872	>	currentCtl, prevCtl)) {
873	>	w.terminate = true;            // restore previous
874	>	w.eventCount = ((int)currentCtl + EC_UNIT) & E_MASK;
875	>	break;
876		}
877		}
878		}
879		}
880
881	<	// Maintaining parallelism
881	>	// Submissions
882
883		/**
884	<	* Pushes worker onto the spare stack.
884	>	* Enqueues the given task in the submissionQueue.  Same idea as
885	>	* ForkJoinWorkerThread.pushTask except for use of submissionLock.
886	>	*
887	>	* @param t the task
888		*/
889	<	final void pushSpare(ForkJoinWorkerThread w) {
890	<	int ns = (++w.spareCount << SPARE_COUNT_SHIFT) | (w.poolIndex + 1);
891	<	do {} while (!UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
892	<	w.nextSpare = spareWaiters,ns));
889	>	private void addSubmission(ForkJoinTask<?> t) {
890	>	final ReentrantLock lock = this.submissionLock;
891	>	lock.lock();
892	>	try {
893	>	ForkJoinTask<?>[] q; int s, m;
894	>	if ((q = submissionQueue) != null) {    // ignore if queue removed
895	>	long u = (((s = queueTop) & (m = q.length-1)) << ASHIFT)+ABASE;
896	>	UNSAFE.putOrderedObject(q, u, t);
897	>	queueTop = s + 1;
898	>	if (s - queueBase == m)
899	>	growSubmissionQueue();
900	>	}
901	>	} finally {
902	>	lock.unlock();
903	>	}
904	>	signalWork();
905		}
906
907	+	//  (pollSubmission is defined below with exported methods)
908	+
909		/**
910	<	* Tries (once) to resume a spare if the number of running
911	<	* threads is less than target.
910	>	* Creates or doubles submissionQueue array.
911	>	* Basically identical to ForkJoinWorkerThread version.
912		*/
913	<	private void tryResumeSpare() {
914	<	int sw, id;
915	<	ForkJoinWorkerThread[] ws = workers;
916	<	int n = ws.length;
917	<	ForkJoinWorkerThread w;
918	<	if ((sw = spareWaiters) != 0 &&
919	<	(id = (sw & SPARE_ID_MASK) - 1) >= 0 &&
920	<	id < n && (w = ws[id]) != null &&
921	<	(runState >= TERMINATING ||
922	<	(workerCounts & RUNNING_COUNT_MASK) < parallelism) &&
923	<	spareWaiters == sw &&
924	<	UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
925	<	sw, w.nextSpare)) {
926	<	int c; // increment running count before resume
927	<	do {} while (!UNSAFE.compareAndSwapInt
928	<	(this, workerCountsOffset,
929	<	c = workerCounts, c + ONE_RUNNING));
930	<	if (w.tryUnsuspend())
931	<	LockSupport.unpark(w);
856	<	else   // back out if w was shutdown
857	<	decrementWorkerCounts(ONE_RUNNING, 0);
913	>	private void growSubmissionQueue() {
914	>	ForkJoinTask<?>[] oldQ = submissionQueue;
915	>	int size = oldQ != null ? oldQ.length << 1 : INITIAL_QUEUE_CAPACITY;
916	>	if (size > MAXIMUM_QUEUE_CAPACITY)
917	>	throw new RejectedExecutionException("Queue capacity exceeded");
918	>	if (size < INITIAL_QUEUE_CAPACITY)
919	>	size = INITIAL_QUEUE_CAPACITY;
920	>	ForkJoinTask<?>[] q = submissionQueue = new ForkJoinTask<?>[size];
921	>	int mask = size - 1;
922	>	int top = queueTop;
923	>	int oldMask;
924	>	if (oldQ != null && (oldMask = oldQ.length - 1) >= 0) {
925	>	for (int b = queueBase; b != top; ++b) {
926	>	long u = ((b & oldMask) << ASHIFT) + ABASE;
927	>	Object x = UNSAFE.getObjectVolatile(oldQ, u);
928	>	if (x != null && UNSAFE.compareAndSwapObject(oldQ, u, x, null))
929	>	UNSAFE.putObjectVolatile
930	>	(q, ((b & mask) << ASHIFT) + ABASE, x);
931	>	}
932		}
933		}
934
935	+	// Blocking support
936	+
937		/**
938	<	* Tries to increase the number of running workers if below target
939	<	* parallelism: If a spare exists tries to resume it via
940	<	* tryResumeSpare.  Otherwise, if not enough total workers or all
941	<	* existing workers are busy, adds a new worker. In all cases also
942	<	* helps wake up releasable workers waiting for work.
943	<	*/
944	<	private void helpMaintainParallelism() {
945	<	int pc = parallelism;
946	<	int wc, rs, tc;
947	<	while (((wc = workerCounts) & RUNNING_COUNT_MASK) < pc &&
948	<	(rs = runState) < TERMINATING) {
949	<	if (spareWaiters != 0)
950	<	tryResumeSpare();
951	<	else if ((tc = wc >>> TOTAL_COUNT_SHIFT) >= MAX_WORKERS ||
952	<	(tc >= pc && (rs & ACTIVE_COUNT_MASK) != tc))
953	<	break;   // enough total
954	<	else if (runState == rs && workerCounts == wc &&
955	<	UNSAFE.compareAndSwapInt(this, workerCountsOffset, wc,
880	<	wc + (ONE_RUNNING|ONE_TOTAL))) {
881	<	ForkJoinWorkerThread w = null;
882	<	Throwable fail = null;
883	<	try {
884	<	w = factory.newThread(this);
885	<	} catch (Throwable ex) {
886	<	fail = ex;
938	>	* Tries to increment blockedCount, decrement active count
939	>	* (sometimes implicitly) and possibly release or create a
940	>	* compensating worker in preparation for blocking. Fails
941	>	* on contention or termination.
942	>	*
943	>	* @return true if the caller can block, else should recheck and retry
944	>	*/
945	>	private boolean tryPreBlock() {
946	>	int b = blockedCount;
947	>	if (UNSAFE.compareAndSwapInt(this, blockedCountOffset, b, b + 1)) {
948	>	int pc = parallelism;
949	>	do {
950	>	ForkJoinWorkerThread[] ws; ForkJoinWorkerThread w;
951	>	int e, ac, tc, rc, i;
952	>	long c = ctl;
953	>	int u = (int)(c >>> 32);
954	>	if ((e = (int)c) < 0) {
955	>	// skip -- terminating
956		}
957	<	if (w == null) { // null or exceptional factory return
958	<	decrementWorkerCounts(ONE_RUNNING, ONE_TOTAL);
959	<	tryTerminate(false); // handle failure during shutdown
960	<	// If originating from an external caller,
961	<	// propagate exception, else ignore
962	<	if (fail != null && runState < TERMINATING &&
963	<	!(Thread.currentThread() instanceof
964	<	ForkJoinWorkerThread))
965	<	UNSAFE.throwException(fail);
966	<	break;
957	>	else if ((ac = (u >> UAC_SHIFT)) <= 0 && e != 0 &&
958	>	(ws = workers) != null &&
959	>	(i = ~e & SMASK) < ws.length &&
960	>	(w = ws[i]) != null) {
961	>	long nc = ((long)(w.nextWait & E_MASK) |
962	>	(c & (AC_MASK|TC_MASK)));
963	>	if (w.eventCount == e &&
964	>	UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc)) {
965	>	w.eventCount = (e + EC_UNIT) & E_MASK;
966	>	if (w.parked)
967	>	UNSAFE.unpark(w);
968	>	return true;             // release an idle worker
969	>	}
970		}
971	<	w.start(recordWorker(w), ueh);
972	<	if ((workerCounts >>> TOTAL_COUNT_SHIFT) >= pc)
973	<	break; // add at most one unless total below target
974	<	}
971	>	else if ((tc = (short)(u >>> UTC_SHIFT)) >= 0 && ac + pc > 1) {
972	>	long nc = ((c - AC_UNIT) & AC_MASK) | (c & ~AC_MASK);
973	>	if (UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc))
974	>	return true;             // no compensation needed
975	>	}
976	>	else if (tc + pc < MAX_ID) {
977	>	long nc = ((c + TC_UNIT) & TC_MASK) | (c & ~TC_MASK);
978	>	if (UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc)) {
979	>	addWorker();
980	>	return true;            // create a replacement
981	>	}
982	>	}
983	>	// try to back out on any failure and let caller retry
984	>	} while (!UNSAFE.compareAndSwapInt(this, blockedCountOffset,
985	>	b = blockedCount, b - 1));
986		}
987	<	if (eventWaiters != 0L)
905	<	releaseEventWaiters();
987	>	return false;
988		}
989
990		/**
991	<	* Callback from the oldest waiter in awaitEvent waking up after a
910	<	* period of non-use. If all workers are idle, tries (once) to
911	<	* shutdown an event waiter or a spare, if one exists. Note that
912	<	* we don't need CAS or locks here because the method is called
913	<	* only from one thread occasionally waking (and even misfires are
914	<	* OK). Note that until the shutdown worker fully terminates,
915	<	* workerCounts will overestimate total count, which is tolerable.
916	<	*
917	<	* @param ec the event count waited on by caller (to abort
918	<	* attempt if count has since changed).
991	>	* Decrements blockedCount and increments active count
992		*/
993	<	private void tryShutdownUnusedWorker(int ec) {
994	<	if (runState == 0 && eventCount == ec) { // only trigger if all idle
995	<	ForkJoinWorkerThread[] ws = workers;
996	<	int n = ws.length;
997	<	ForkJoinWorkerThread w = null;
998	<	boolean shutdown = false;
999	<	int sw;
927	<	long h;
928	<	if ((sw = spareWaiters) != 0) { // prefer killing spares
929	<	int id = (sw & SPARE_ID_MASK) - 1;
930	<	if (id >= 0 && id < n && (w = ws[id]) != null &&
931	<	UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
932	<	sw, w.nextSpare))
933	<	shutdown = true;
934	<	}
935	<	else if ((h = eventWaiters) != 0L) {
936	<	long nh;
937	<	int id = ((int)(h & WAITER_ID_MASK)) - 1;
938	<	if (id >= 0 && id < n && (w = ws[id]) != null &&
939	<	(nh = w.nextWaiter) != 0L && // keep at least one worker
940	<	UNSAFE.compareAndSwapLong(this, eventWaitersOffset, h, nh))
941	<	shutdown = true;
942	<	}
943	<	if (w != null && shutdown) {
944	<	w.shutdown();
945	<	LockSupport.unpark(w);
946	<	}
947	<	}
948	<	releaseEventWaiters(); // in case of interference
993	>	private void postBlock() {
994	>	long c;
995	>	do {} while (!UNSAFE.compareAndSwapLong(this, ctlOffset,  // no mask
996	>	c = ctl, c + AC_UNIT));
997	>	int b;
998	>	do {} while(!UNSAFE.compareAndSwapInt(this, blockedCountOffset,
999	>	b = blockedCount, b - 1));
1000		}
1001
1002		/**
1003	<	* Callback from workers invoked upon each top-level action (i.e.,
1004	<	* stealing a task or taking a submission and running it).
954	<	* Performs one or more of the following:
955	<	*
956	<	* 1. If the worker is active and either did not run a task
957	<	*    or there are too many workers, try to set its active status
958	<	*    to inactive and update activeCount. On contention, we may
959	<	*    try again in this or a subsequent call.
960	<	*
961	<	* 2. If not enough total workers, help create some.
962	<	*
963	<	* 3. If there are too many running workers, suspend this worker
964	<	*    (first forcing inactive if necessary).  If it is not needed,
965	<	*    it may be shutdown while suspended (via
966	<	*    tryShutdownUnusedWorker).  Otherwise, upon resume it
967	<	*    rechecks running thread count and need for event sync.
968	<	*
969	<	* 4. If worker did not run a task, await the next task event via
970	<	*    eventSync if necessary (first forcing inactivation), upon
971	<	*    which the worker may be shutdown via
972	<	*    tryShutdownUnusedWorker.  Otherwise, help release any
973	<	*    existing event waiters that are now releasable,
1003	>	* Possibly blocks waiting for the given task to complete, or
1004	>	* cancels the task if terminating.  Fails to wait if contended.
1005		*
1006	<	* @param w the worker
976	<	* @param ran true if worker ran a task since last call to this method
1006	>	* @param joinMe the task
1007		*/
1008	<	final void preStep(ForkJoinWorkerThread w, boolean ran) {
1009	<	int wec = w.lastEventCount;
1010	<	boolean active = w.active;
1011	<	boolean inactivate = false;
1012	<	int pc = parallelism;
1013	<	while (w.runState == 0) {
1014	<	int rs = runState;
985	<	if (rs >= TERMINATING) {           // propagate shutdown
986	<	w.shutdown();
987	<	break;
988	<	}
989	<	if ((inactivate || (active && (rs & ACTIVE_COUNT_MASK) >= pc)) &&
990	<	UNSAFE.compareAndSwapInt(this, runStateOffset, rs, --rs)) {
991	<	inactivate = active = w.active = false;
992	<	if (rs == SHUTDOWN) {          // all inactive and shut down
993	<	tryTerminate(false);
994	<	continue;
995	<	}
996	<	}
997	<	int wc = workerCounts;             // try to suspend as spare
998	<	if ((wc & RUNNING_COUNT_MASK) > pc) {
999	<	if (!(inactivate |= active) && // must inactivate to suspend
1000	<	workerCounts == wc &&
1001	<	UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1002	<	wc, wc - ONE_RUNNING))
1003	<	w.suspendAsSpare();
1004	<	}
1005	<	else if ((wc >>> TOTAL_COUNT_SHIFT) < pc)
1006	<	helpMaintainParallelism();     // not enough workers
1007	<	else if (ran)
1008	<	break;
1009	<	else {
1010	<	long h = eventWaiters;
1011	<	int ec = eventCount;
1012	<	if (h != 0L && (int)(h >>> EVENT_COUNT_SHIFT) != ec)
1013	<	releaseEventWaiters();     // release others before waiting
1014	<	else if (ec != wec) {
1015	<	w.lastEventCount = ec;     // no need to wait
1016	<	break;
1017	<	}
1018	<	else if (!(inactivate |= active))
1019	<	eventSync(w, wec);         // must inactivate before sync
1008	>	final void tryAwaitJoin(ForkJoinTask<?> joinMe) {
1009	>	int s;
1010	>	Thread.interrupted(); // clear interrupts before checking termination
1011	>	if (joinMe.status >= 0) {
1012	>	if (tryPreBlock()) {
1013	>	joinMe.tryAwaitDone(0L);
1014	>	postBlock();
1015		}
1016	+	else if ((ctl & STOP_BIT) != 0L)
1017	+	joinMe.cancelIgnoringExceptions();
1018		}
1019		}
1020
1021		/**
1022	<	* Helps and/or blocks awaiting join of the given task.
1023	<	* See above for explanation.
1022	>	* Possibly blocks the given worker waiting for joinMe to
1023	>	* complete or timeout
1024		*
1025	<	* @param joinMe the task to join
1026	<	* @param worker the current worker thread
1030	<	* @param timed true if wait should time out
1031	<	* @param nanos timeout value if timed
1025	>	* @param joinMe the task
1026	>	* @param millis the wait time for underlying Object.wait
1027		*/
1028	<	final void awaitJoin(ForkJoinTask<?> joinMe, ForkJoinWorkerThread worker,
1034	<	boolean timed, long nanos) {
1035	<	long startTime = timed? System.nanoTime() : 0L;
1036	<	int retries = 2 + (parallelism >> 2); // #helpJoins before blocking
1037	<	boolean running = true;               // false when count decremented
1028	>	final void timedAwaitJoin(ForkJoinTask<?> joinMe, long nanos) {
1029		while (joinMe.status >= 0) {
1030	<	if (runState >= TERMINATING) {
1030	>	Thread.interrupted();
1031	>	if ((ctl & STOP_BIT) != 0L) {
1032		joinMe.cancelIgnoringExceptions();
1033		break;
1034		}
1035	<	running = worker.helpJoinTask(joinMe, running);
1036	<	if (joinMe.status < 0)
1037	<	break;
1038	<	if (retries > 0) {
1039	<	--retries;
1040	<	continue;
1041	<	}
1042	<	int wc = workerCounts;
1051	<	if ((wc & RUNNING_COUNT_MASK) != 0) {
1052	<	if (running) {
1053	<	if (!UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1054	<	wc, wc - ONE_RUNNING))
1055	<	continue;
1056	<	running = false;
1057	<	}
1058	<	long h = eventWaiters;
1059	<	if (h != 0L && (int)(h >>> EVENT_COUNT_SHIFT) != eventCount)
1060	<	releaseEventWaiters();
1061	<	if (joinMe.status < 0)
1062	<	break;
1063	<	if ((workerCounts & RUNNING_COUNT_MASK) != 0) {
1064	<	long ms; int ns;
1065	<	if (!timed) {
1066	<	ms = JOIN_TIMEOUT_MILLIS;
1067	<	ns = 0;
1068	<	}
1069	<	else { // at most JOIN_TIMEOUT_MILLIS per wait
1070	<	long nt = nanos - (System.nanoTime() - startTime);
1071	<	if (nt <= 0L)
1072	<	break;
1073	<	ms = nt / 1000000;
1074	<	if (ms > JOIN_TIMEOUT_MILLIS) {
1075	<	ms = JOIN_TIMEOUT_MILLIS;
1076	<	ns = 0;
1077	<	}
1078	<	else
1079	<	ns = (int) (nt % 1000000);
1080	<	}
1081	<	if (joinMe.internalAwaitDone(ms, ns) < 0)
1035	>	if (tryPreBlock()) {
1036	>	long last = System.nanoTime();
1037	>	while (joinMe.status >= 0) {
1038	>	long millis = TimeUnit.NANOSECONDS.toMillis(nanos);
1039	>	if (millis <= 0)
1040	>	break;
1041	>	joinMe.tryAwaitDone(millis);
1042	>	if (joinMe.status < 0)
1043		break;
1044	+	if ((ctl & STOP_BIT) != 0L) {
1045	+	joinMe.cancelIgnoringExceptions();
1046	+	break;
1047	+	}
1048	+	long now = System.nanoTime();
1049	+	nanos -= now - last;
1050	+	last = now;
1051		}
1052	+	postBlock();
1053	+	break;
1054		}
1085	–	helpMaintainParallelism();
1086	–	}
1087	–	if (!running) {
1088	–	int c;
1089	–	do {} while (!UNSAFE.compareAndSwapInt
1090	–	(this, workerCountsOffset,
1091	–	c = workerCounts, c + ONE_RUNNING));
1055		}
1056		}
1057
1058		/**
1059	<	* Same idea as awaitJoin, but no helping, retries, or timeouts.
1059	>	* If necessary, compensates for blocker, and blocks
1060		*/
1061	<	final void awaitBlocker(ManagedBlocker blocker)
1061	>	private void awaitBlocker(ManagedBlocker blocker)
1062		throws InterruptedException {
1063		while (!blocker.isReleasable()) {
1064	<	int wc = workerCounts;
1102	<	if ((wc & RUNNING_COUNT_MASK) == 0)
1103	<	helpMaintainParallelism();
1104	<	else if (UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1105	<	wc, wc - ONE_RUNNING)) {
1064	>	if (tryPreBlock()) {
1065		try {
1066	<	while (!blocker.isReleasable()) {
1108	<	long h = eventWaiters;
1109	<	if (h != 0L &&
1110	<	(int)(h >>> EVENT_COUNT_SHIFT) != eventCount)
1111	<	releaseEventWaiters();
1112	<	else if ((workerCounts & RUNNING_COUNT_MASK) == 0 &&
1113	<	runState < TERMINATING)
1114	<	helpMaintainParallelism();
1115	<	else if (blocker.block())
1116	<	break;
1117	<	}
1066	>	do {} while (!blocker.isReleasable() && !blocker.block());
1067		} finally {
1068	<	int c;
1120	<	do {} while (!UNSAFE.compareAndSwapInt
1121	<	(this, workerCountsOffset,
1122	<	c = workerCounts, c + ONE_RUNNING));
1068	>	postBlock();
1069		}
1070		break;
1071		}
1072		}
1073		}
1074
1075	+	// Creating, registering and deregistring workers
1076	+
1077	+	/**
1078	+	* Tries to create and start a worker; minimally rolls back counts
1079	+	* on failure.
1080	+	*/
1081	+	private void addWorker() {
1082	+	Throwable ex = null;
1083	+	ForkJoinWorkerThread t = null;
1084	+	try {
1085	+	t = factory.newThread(this);
1086	+	} catch (Throwable e) {
1087	+	ex = e;
1088	+	}
1089	+	if (t == null) {  // null or exceptional factory return
1090	+	long c;       // adjust counts
1091	+	do {} while (!UNSAFE.compareAndSwapLong
1092	+	(this, ctlOffset, c = ctl,
1093	+	(((c - AC_UNIT) & AC_MASK) |
1094	+	((c - TC_UNIT) & TC_MASK) |
1095	+	(c & ~(AC_MASK|TC_MASK)))));
1096	+	// Propagate exception if originating from an external caller
1097	+	if (!tryTerminate(false) && ex != null &&
1098	+	!(Thread.currentThread() instanceof ForkJoinWorkerThread))
1099	+	UNSAFE.throwException(ex);
1100	+	}
1101	+	else
1102	+	t.start();
1103	+	}
1104	+
1105	+	/**
1106	+	* Callback from ForkJoinWorkerThread constructor to assign a
1107	+	* public name
1108	+	*/
1109	+	final String nextWorkerName() {
1110	+	for (int n;;) {
1111	+	if (UNSAFE.compareAndSwapInt(this, nextWorkerNumberOffset,
1112	+	n = nextWorkerNumber, ++n))
1113	+	return workerNamePrefix + n;
1114	+	}
1115	+	}
1116	+
1117	+	/**
1118	+	* Callback from ForkJoinWorkerThread constructor to
1119	+	* determine its poolIndex and record in workers array.
1120	+	*
1121	+	* @param w the worker
1122	+	* @return the worker's pool index
1123	+	*/
1124	+	final int registerWorker(ForkJoinWorkerThread w) {
1125	+	/*
1126	+	* In the typical case, a new worker acquires the lock, uses
1127	+	* next available index and returns quickly.  Since we should
1128	+	* not block callers (ultimately from signalWork or
1129	+	* tryPreBlock) waiting for the lock needed to do this, we
1130	+	* instead help release other workers while waiting for the
1131	+	* lock.
1132	+	*/
1133	+	for (int g;;) {
1134	+	ForkJoinWorkerThread[] ws;
1135	+	if (((g = scanGuard) & SG_UNIT) == 0 &&
1136	+	UNSAFE.compareAndSwapInt(this, scanGuardOffset,
1137	+	g, g | SG_UNIT)) {
1138	+	int k = nextWorkerIndex;
1139	+	try {
1140	+	if ((ws = workers) != null) { // ignore on shutdown
1141	+	int n = ws.length;
1142	+	if (k < 0 || k >= n || ws[k] != null) {
1143	+	for (k = 0; k < n && ws[k] != null; ++k)
1144	+	;
1145	+	if (k == n)
1146	+	ws = workers = Arrays.copyOf(ws, n << 1);
1147	+	}
1148	+	ws[k] = w;
1149	+	nextWorkerIndex = k + 1;
1150	+	int m = g & SMASK;
1151	+	g = k > m? ((m << 1) + 1) & SMASK : g + (SG_UNIT<<1);
1152	+	}
1153	+	} finally {
1154	+	scanGuard = g;
1155	+	}
1156	+	return k;
1157	+	}
1158	+	else if ((ws = workers) != null) { // help release others
1159	+	for (ForkJoinWorkerThread u : ws) {
1160	+	if (u != null && u.queueBase != u.queueTop) {
1161	+	if (tryReleaseWaiter())
1162	+	break;
1163	+	}
1164	+	}
1165	+	}
1166	+	}
1167	+	}
1168	+
1169	+	/**
1170	+	* Final callback from terminating worker.  Removes record of
1171	+	* worker from array, and adjusts counts. If pool is shutting
1172	+	* down, tries to complete termination.
1173	+	*
1174	+	* @param w the worker
1175	+	*/
1176	+	final void deregisterWorker(ForkJoinWorkerThread w, Throwable ex) {
1177	+	int idx = w.poolIndex;
1178	+	int sc = w.stealCount;
1179	+	int steps = 0;
1180	+	// Remove from array, adjust worker counts and collect steal count.
1181	+	// We can intermix failed removes or adjusts with steal updates
1182	+	do {
1183	+	long s, c;
1184	+	int g;
1185	+	if (steps == 0 && ((g = scanGuard) & SG_UNIT) == 0 &&
1186	+	UNSAFE.compareAndSwapInt(this, scanGuardOffset,
1187	+	g, g |= SG_UNIT)) {
1188	+	ForkJoinWorkerThread[] ws = workers;
1189	+	if (ws != null && idx >= 0 &&
1190	+	idx < ws.length && ws[idx] == w)
1191	+	ws[idx] = null;    // verify
1192	+	nextWorkerIndex = idx;
1193	+	scanGuard = g + SG_UNIT;
1194	+	steps = 1;
1195	+	}
1196	+	if (steps == 1 &&
1197	+	UNSAFE.compareAndSwapLong(this, ctlOffset, c = ctl,
1198	+	(((c - AC_UNIT) & AC_MASK) |
1199	+	((c - TC_UNIT) & TC_MASK) |
1200	+	(c & ~(AC_MASK|TC_MASK)))))
1201	+	steps = 2;
1202	+	if (sc != 0 &&
1203	+	UNSAFE.compareAndSwapLong(this, stealCountOffset,
1204	+	s = stealCount, s + sc))
1205	+	sc = 0;
1206	+	} while (steps != 2 || sc != 0);
1207	+	if (!tryTerminate(false)) {
1208	+	if (ex != null)   // possibly replace if died abnormally
1209	+	signalWork();
1210	+	else
1211	+	tryReleaseWaiter();
1212	+	}
1213	+	}
1214	+
1215	+	// Shutdown and termination
1216	+
1217		/**
1218		* Possibly initiates and/or completes termination.
1219		*
#	Line 1134 | Line 1222 | public class ForkJoinPool extends Abstra
1222		* @return true if now terminating or terminated
1223		*/
1224		private boolean tryTerminate(boolean now) {
1225	<	if (now)
1226	<	advanceRunLevel(SHUTDOWN); // ensure at least SHUTDOWN
1227	<	else if (runState < SHUTDOWN ||
1228	<	!submissionQueue.isEmpty() ||
1229	<	(runState & ACTIVE_COUNT_MASK) != 0)
1230	<	return false;
1231	<
1232	<	if (advanceRunLevel(TERMINATING))
1233	<	startTerminating();
1234	<
1235	<	// Finish now if all threads terminated; else in some subsequent call
1236	<	if ((workerCounts >>> TOTAL_COUNT_SHIFT) == 0) {
1237	<	advanceRunLevel(TERMINATED);
1238	<	termination.forceTermination();
1225	>	long c;
1226	>	while (((c = ctl) & STOP_BIT) == 0) {
1227	>	if (!now) {
1228	>	if ((int)(c >> AC_SHIFT) != -parallelism)
1229	>	return false;
1230	>	if (!shutdown || blockedCount != 0 || quiescerCount != 0 ||
1231	>	queueBase != queueTop) {
1232	>	if (ctl == c) // staleness check
1233	>	return false;
1234	>	continue;
1235	>	}
1236	>	}
1237	>	if (UNSAFE.compareAndSwapLong(this, ctlOffset, c, c | STOP_BIT))
1238	>	startTerminating();
1239	>	}
1240	>	if ((short)(c >>> TC_SHIFT) == -parallelism) { // signal when 0 workers
1241	>	final ReentrantLock lock = this.submissionLock;
1242	>	lock.lock();
1243	>	try {
1244	>	termination.signalAll();
1245	>	} finally {
1246	>	lock.unlock();
1247	>	}
1248		}
1249		return true;
1250		}
1251
1252		/**
1253	<	* Actions on transition to TERMINATING
1254	<	*
1255	<	* Runs up to four passes through workers: (0) shutting down each
1256	<	* (without waking up if parked) to quickly spread notifications
1257	<	* without unnecessary bouncing around event queues etc (1) wake
1258	<	* up and help cancel tasks (2) interrupt (3) mop up races with
1162	<	* interrupted workers
1253	>	* Runs up to three passes through workers: (0) Setting
1254	>	* termination status for each worker, followed by wakeups up to
1255	>	* queued workers; (1) helping cancel tasks; (2) interrupting
1256	>	* lagging threads (likely in external tasks, but possibly also
1257	>	* blocked in joins).  Each pass repeats previous steps because of
1258	>	* potential lagging thread creation.
1259		*/
1260		private void startTerminating() {
1261		cancelSubmissions();
1262	<	for (int passes = 0; passes < 4 && workerCounts != 0; ++passes) {
1263	<	int c; // advance event count
1264	<	UNSAFE.compareAndSwapInt(this, eventCountOffset,
1265	<	c = eventCount, c+1);
1266	<	eventWaiters = 0L; // clobber lists
1267	<	spareWaiters = 0;
1268	<	for (ForkJoinWorkerThread w : workers) {
1269	<	if (w != null) {
1270	<	w.shutdown();
1271	<	if (passes > 0 && !w.isTerminated()) {
1272	<	w.cancelTasks();
1273	<	LockSupport.unpark(w);
1274	<	if (passes > 1 && !w.isInterrupted()) {
1179	<	try {
1180	<	w.interrupt();
1181	<	} catch (SecurityException ignore) {
1262	>	for (int pass = 0; pass < 3; ++pass) {
1263	>	ForkJoinWorkerThread[] ws = workers;
1264	>	if (ws != null) {
1265	>	for (ForkJoinWorkerThread w : ws) {
1266	>	if (w != null) {
1267	>	w.terminate = true;
1268	>	if (pass > 0) {
1269	>	w.cancelTasks();
1270	>	if (pass > 1 && !w.isInterrupted()) {
1271	>	try {
1272	>	w.interrupt();
1273	>	} catch (SecurityException ignore) {
1274	>	}
1275		}
1276		}
1277		}
1278		}
1279	+	terminateWaiters();
1280		}
1281		}
1282		}
1283
1284		/**
1285	<	* Clears out and cancels submissions, ignoring exceptions.
1285	>	* Polls and cancels all submissions. Called only during termination.
1286		*/
1287		private void cancelSubmissions() {
1288	<	ForkJoinTask<?> task;
1289	<	while ((task = submissionQueue.poll()) != null) {
1290	<	try {
1291	<	task.cancel(false);
1292	<	} catch (Throwable ignore) {
1288	>	while (queueBase != queueTop) {
1289	>	ForkJoinTask<?> task = pollSubmission();
1290	>	if (task != null) {
1291	>	try {
1292	>	task.cancel(false);
1293	>	} catch (Throwable ignore) {
1294	>	}
1295		}
1296		}
1297		}
1298
1299	<	// misc support for ForkJoinWorkerThread
1299	>	/**
1300	>	* Tries to set the termination status of waiting workers, and
1301	>	* then wakes them up (after which they will terminate).
1302	>	*/
1303	>	private void terminateWaiters() {
1304	>	ForkJoinWorkerThread[] ws = workers;
1305	>	if (ws != null) {
1306	>	ForkJoinWorkerThread w; long c; int i, e;
1307	>	int n = ws.length;
1308	>	while ((i = ~(e = (int)(c = ctl)) & SMASK) < n &&
1309	>	(w = ws[i]) != null && w.eventCount == (e & E_MASK)) {
1310	>	if (UNSAFE.compareAndSwapLong(this, ctlOffset, c,
1311	>	(long)(w.nextWait & E_MASK) |
1312	>	((c + AC_UNIT) & AC_MASK) |
1313	>	(c & (TC_MASK|STOP_BIT)))) {
1314	>	w.terminate = true;
1315	>	w.eventCount = e + EC_UNIT;
1316	>	if (w.parked)
1317	>	UNSAFE.unpark(w);
1318	>	}
1319	>	}
1320	>	}
1321	>	}
1322	>
1323	>	// misc ForkJoinWorkerThread support
1324
1325		/**
1326	<	* Returns pool number.
1326	>	* Increment or decrement quiescerCount. Needed only to prevent
1327	>	* triggering shutdown if a worker is transiently inactive while
1328	>	* checking quiescence.
1329	>	*
1330	>	* @param delta 1 for increment, -1 for decrement
1331		*/
1332	<	final int getPoolNumber() {
1333	<	return poolNumber;
1332	>	final void addQuiescerCount(int delta) {
1333	>	int c;
1334	>	do {} while(!UNSAFE.compareAndSwapInt(this, quiescerCountOffset,
1335	>	c = quiescerCount, c + delta));
1336		}
1337
1338		/**
1339	<	* Tries to accumulate steal count from a worker, clearing
1340	<	* the worker's value if successful.
1339	>	* Directly increment or decrement active count without
1340	>	* queuing. This method is used to transiently assert inactivation
1341	>	* while checking quiescence.
1342		*
1343	<	* @return true if worker steal count now zero
1343	>	* @param delta 1 for increment, -1 for decrement
1344		*/
1345	<	final boolean tryAccumulateStealCount(ForkJoinWorkerThread w) {
1346	<	int sc = w.stealCount;
1347	<	long c = stealCount;
1348	<	// CAS even if zero, for fence effects
1349	<	if (UNSAFE.compareAndSwapLong(this, stealCountOffset, c, c + sc)) {
1350	<	if (sc != 0)
1224	<	w.stealCount = 0;
1225	<	return true;
1226	<	}
1227	<	return sc == 0;
1345	>	final void addActiveCount(int delta) {
1346	>	long d = delta < 0 ? -AC_UNIT : AC_UNIT;
1347	>	long c;
1348	>	do {} while (!UNSAFE.compareAndSwapLong(this, ctlOffset, c = ctl,
1349	>	((c + d) & AC_MASK) |
1350	>	(c & ~AC_MASK)));
1351		}
1352
1353		/**
#	Line 1232 | Line 1355 | public class ForkJoinPool extends Abstra
1355		* active thread.
1356		*/
1357		final int idlePerActive() {
1358	<	int pc = parallelism; // use parallelism, not rc
1359	<	int ac = runState;    // no mask -- artificially boosts during shutdown
1360	<	// Use exact results for small values, saturate past 4
1361	<	return ((pc <= ac) ? 0 :
1362	<	(pc >>> 1 <= ac) ? 1 :
1363	<	(pc >>> 2 <= ac) ? 3 :
1364	<	pc >>> 3);
1358	>	// Approximate at powers of two for small values, saturate past 4
1359	>	int p = parallelism;
1360	>	int a = p + (int)(ctl >> AC_SHIFT);
1361	>	return (a > (p >>>= 1) ? 0 :
1362	>	a > (p >>>= 1) ? 1 :
1363	>	a > (p >>>= 1) ? 2 :
1364	>	a > (p >>>= 1) ? 4 :
1365	>	8);
1366		}
1367
1368	<	// Public and protected methods
1368	>	// Exported methods
1369
1370		// Constructors
1371
#	Line 1310 | Line 1434 | public class ForkJoinPool extends Abstra
1434		checkPermission();
1435		if (factory == null)
1436		throw new NullPointerException();
1437	<	if (parallelism <= 0 || parallelism > MAX_WORKERS)
1437	>	if (parallelism <= 0 || parallelism > MAX_ID)
1438		throw new IllegalArgumentException();
1439		this.parallelism = parallelism;
1440		this.factory = factory;
1441		this.ueh = handler;
1442		this.locallyFifo = asyncMode;
1443	<	int arraySize = initialArraySizeFor(parallelism);
1444	<	this.workers = new ForkJoinWorkerThread[arraySize];
1445	<	this.submissionQueue = new LinkedTransferQueue<ForkJoinTask<?>>();
1446	<	this.workerLock = new ReentrantLock();
1447	<	this.termination = new Phaser(1);
1448	<	this.poolNumber = poolNumberGenerator.incrementAndGet();
1449	<	}
1450	<
1451	<	/**
1452	<	* Returns initial power of two size for workers array.
1453	<	* @param pc the initial parallelism level
1454	<	*/
1455	<	private static int initialArraySizeFor(int pc) {
1456	<	// If possible, initially allocate enough space for one spare
1457	<	int size = pc < MAX_WORKERS ? pc + 1 : MAX_WORKERS;
1458	<	// See Hackers Delight, sec 3.2. We know MAX_WORKERS < (1 >>> 16)
1459	<	size |= size >>> 1;
1336	<	size |= size >>> 2;
1337	<	size |= size >>> 4;
1338	<	size |= size >>> 8;
1339	<	return size + 1;
1443	>	long np = (long)(-parallelism); // offset ctl counts
1444	>	this.ctl = ((np << AC_SHIFT) & AC_MASK) | ((np << TC_SHIFT) & TC_MASK);
1445	>	this.submissionQueue = new ForkJoinTask<?>[INITIAL_QUEUE_CAPACITY];
1446	>	// initialize workers array with room for 2*parallelism if possible
1447	>	int n = parallelism << 1;
1448	>	if (n >= MAX_ID)
1449	>	n = MAX_ID;
1450	>	else { // See Hackers Delight, sec 3.2, where n < (1 << 16)
1451	>	n |= n >>> 1; n |= n >>> 2; n |= n >>> 4; n |= n >>> 8;
1452	>	}
1453	>	workers = new ForkJoinWorkerThread[n + 1];
1454	>	this.submissionLock = new ReentrantLock();
1455	>	this.termination = submissionLock.newCondition();
1456	>	StringBuilder sb = new StringBuilder("ForkJoinPool-");
1457	>	sb.append(poolNumberGenerator.incrementAndGet());
1458	>	sb.append("-worker-");
1459	>	this.workerNamePrefix = sb.toString();
1460		}
1461
1462		// Execution methods
1463
1464		/**
1345	–	* Submits task and creates, starts, or resumes some workers if necessary
1346	–	*/
1347	–	private <T> void doSubmit(ForkJoinTask<T> task) {
1348	–	submissionQueue.offer(task);
1349	–	int c; // try to increment event count -- CAS failure OK
1350	–	UNSAFE.compareAndSwapInt(this, eventCountOffset, c = eventCount, c+1);
1351	–	helpMaintainParallelism();
1352	–	}
1353	–
1354	–	/**
1465		* Performs the given task, returning its result upon completion.
1466	+	* If the computation encounters an unchecked Exception or Error,
1467	+	* it is rethrown as the outcome of this invocation.  Rethrown
1468	+	* exceptions behave in the same way as regular exceptions, but,
1469	+	* when possible, contain stack traces (as displayed for example
1470	+	* using {@code ex.printStackTrace()}) of both the current thread
1471	+	* as well as the thread actually encountering the exception;
1472	+	* minimally only the latter.
1473		*
1474		* @param task the task
1475		* @return the task's result
#	Line 1361 | Line 1478 | public class ForkJoinPool extends Abstra
1478		*         scheduled for execution
1479		*/
1480		public <T> T invoke(ForkJoinTask<T> task) {
1481	+	Thread t = Thread.currentThread();
1482		if (task == null)
1483		throw new NullPointerException();
1484	<	if (runState >= SHUTDOWN)
1484	>	if (shutdown)
1485		throw new RejectedExecutionException();
1368	–	Thread t = Thread.currentThread();
1486		if ((t instanceof ForkJoinWorkerThread) &&
1487		((ForkJoinWorkerThread)t).pool == this)
1488		return task.invoke();  // bypass submit if in same pool
1489		else {
1490	<	doSubmit(task);
1490	>	addSubmission(task);
1491		return task.join();
1492		}
1493		}
#	Line 1380 | Line 1497 | public class ForkJoinPool extends Abstra
1497		* computation in the current pool, else submits as external task.
1498		*/
1499		private <T> void forkOrSubmit(ForkJoinTask<T> task) {
1500	<	if (runState >= SHUTDOWN)
1384	<	throw new RejectedExecutionException();
1500	>	ForkJoinWorkerThread w;
1501		Thread t = Thread.currentThread();
1502	+	if (shutdown)
1503	+	throw new RejectedExecutionException();
1504		if ((t instanceof ForkJoinWorkerThread) &&
1505	<	((ForkJoinWorkerThread)t).pool == this)
1506	<	task.fork();
1505	>	(w = (ForkJoinWorkerThread)t).pool == this)
1506	>	w.pushTask(task);
1507		else
1508	<	doSubmit(task);
1508	>	addSubmission(task);
1509		}
1510
1511		/**
#	Line 1544 | Line 1662 | public class ForkJoinPool extends Abstra
1662		* @return the number of worker threads
1663		*/
1664		public int getPoolSize() {
1665	<	return workerCounts >>> TOTAL_COUNT_SHIFT;
1665	>	return parallelism + (short)(ctl >>> TC_SHIFT);
1666		}
1667
1668		/**
#	Line 1566 | Line 1684 | public class ForkJoinPool extends Abstra
1684		* @return the number of worker threads
1685		*/
1686		public int getRunningThreadCount() {
1687	<	return workerCounts & RUNNING_COUNT_MASK;
1687	>	int r = parallelism + (int)(ctl >> AC_SHIFT);
1688	>	return r <= 0? 0 : r; // suppress momentarily negative values
1689		}
1690
1691		/**
#	Line 1577 | Line 1696 | public class ForkJoinPool extends Abstra
1696		* @return the number of active threads
1697		*/
1698		public int getActiveThreadCount() {
1699	<	return runState & ACTIVE_COUNT_MASK;
1699	>	int r = parallelism + (int)(ctl >> AC_SHIFT) + blockedCount;
1700	>	return r <= 0? 0 : r; // suppress momentarily negative values
1701		}
1702
1703		/**
#	Line 1592 | Line 1712 | public class ForkJoinPool extends Abstra
1712		* @return {@code true} if all threads are currently idle
1713		*/
1714		public boolean isQuiescent() {
1715	<	return (runState & ACTIVE_COUNT_MASK) == 0;
1715	>	return parallelism + (int)(ctl >> AC_SHIFT) + blockedCount == 0;
1716		}
1717
1718		/**
#	Line 1622 | Line 1742 | public class ForkJoinPool extends Abstra
1742		*/
1743		public long getQueuedTaskCount() {
1744		long count = 0;
1745	<	for (ForkJoinWorkerThread w : workers)
1746	<	if (w != null)
1747	<	count += w.getQueueSize();
1745	>	ForkJoinWorkerThread[] ws;
1746	>	if ((short)(ctl >>> TC_SHIFT) > -parallelism &&
1747	>	(ws = workers) != null) {
1748	>	for (ForkJoinWorkerThread w : ws)
1749	>	if (w != null)
1750	>	count -= w.queueBase - w.queueTop; // must read base first
1751	>	}
1752		return count;
1753		}
1754
1755		/**
1756		* Returns an estimate of the number of tasks submitted to this
1757	<	* pool that have not yet begun executing.  This method takes time
1758	<	* proportional to the number of submissions.
1757	>	* pool that have not yet begun executing.  This method may take
1758	>	* time proportional to the number of submissions.
1759		*
1760		* @return the number of queued submissions
1761		*/
1762		public int getQueuedSubmissionCount() {
1763	<	return submissionQueue.size();
1763	>	return -queueBase + queueTop;
1764		}
1765
1766		/**
#	Line 1646 | Line 1770 | public class ForkJoinPool extends Abstra
1770		* @return {@code true} if there are any queued submissions
1771		*/
1772		public boolean hasQueuedSubmissions() {
1773	<	return !submissionQueue.isEmpty();
1773	>	return queueBase != queueTop;
1774		}
1775
1776		/**
#	Line 1657 | Line 1781 | public class ForkJoinPool extends Abstra
1781		* @return the next submission, or {@code null} if none
1782		*/
1783		protected ForkJoinTask<?> pollSubmission() {
1784	<	return submissionQueue.poll();
1784	>	ForkJoinTask<?> t; ForkJoinTask<?>[] q; int b, i;
1785	>	while ((b = queueBase) != queueTop &&
1786	>	(q = submissionQueue) != null &&
1787	>	(i = (q.length - 1) & b) >= 0) {
1788	>	long u = (i << ASHIFT) + ABASE;
1789	>	if ((t = q[i]) != null &&
1790	>	queueBase == b &&
1791	>	UNSAFE.compareAndSwapObject(q, u, t, null)) {
1792	>	queueBase = b + 1;
1793	>	return t;
1794	>	}
1795	>	}
1796	>	return null;
1797		}
1798
1799		/**
#	Line 1678 | Line 1814 | public class ForkJoinPool extends Abstra
1814		* @return the number of elements transferred
1815		*/
1816		protected int drainTasksTo(Collection<? super ForkJoinTask<?>> c) {
1817	<	int count = submissionQueue.drainTo(c);
1818	<	for (ForkJoinWorkerThread w : workers)
1819	<	if (w != null)
1820	<	count += w.drainTasksTo(c);
1817	>	int count = 0;
1818	>	while (queueBase != queueTop) {
1819	>	ForkJoinTask<?> t = pollSubmission();
1820	>	if (t != null) {
1821	>	c.add(t);
1822	>	++count;
1823	>	}
1824	>	}
1825	>	ForkJoinWorkerThread[] ws;
1826	>	if ((short)(ctl >>> TC_SHIFT) > -parallelism &&
1827	>	(ws = workers) != null) {
1828	>	for (ForkJoinWorkerThread w : ws)
1829	>	if (w != null)
1830	>	count += w.drainTasksTo(c);
1831	>	}
1832		return count;
1833		}
1834
#	Line 1696 | Line 1843 | public class ForkJoinPool extends Abstra
1843		long st = getStealCount();
1844		long qt = getQueuedTaskCount();
1845		long qs = getQueuedSubmissionCount();
1699	–	int wc = workerCounts;
1700	–	int tc = wc >>> TOTAL_COUNT_SHIFT;
1701	–	int rc = wc & RUNNING_COUNT_MASK;
1846		int pc = parallelism;
1847	<	int rs = runState;
1848	<	int ac = rs & ACTIVE_COUNT_MASK;
1847	>	long c = ctl;
1848	>	int tc = pc + (short)(c >>> TC_SHIFT);
1849	>	int rc = pc + (int)(c >> AC_SHIFT);
1850	>	if (rc < 0) // ignore transient negative
1851	>	rc = 0;
1852	>	int ac = rc + blockedCount;
1853	>	String level;
1854	>	if ((c & STOP_BIT) != 0)
1855	>	level = (tc == 0)? "Terminated" : "Terminating";
1856	>	else
1857	>	level = shutdown? "Shutting down" : "Running";
1858		return super.toString() +
1859	<	"[" + runLevelToString(rs) +
1859	>	"[" + level +
1860		", parallelism = " + pc +
1861		", size = " + tc +
1862		", active = " + ac +
#	Line 1714 | Line 1867 | public class ForkJoinPool extends Abstra
1867		"]";
1868		}
1869
1717	–	private static String runLevelToString(int s) {
1718	–	return ((s & TERMINATED) != 0 ? "Terminated" :
1719	–	((s & TERMINATING) != 0 ? "Terminating" :
1720	–	((s & SHUTDOWN) != 0 ? "Shutting down" :
1721	–	"Running")));
1722	–	}
1723	–
1870		/**
1871		* Initiates an orderly shutdown in which previously submitted
1872		* tasks are executed, but no new tasks will be accepted.
#	Line 1735 | Line 1881 | public class ForkJoinPool extends Abstra
1881		*/
1882		public void shutdown() {
1883		checkPermission();
1884	<	advanceRunLevel(SHUTDOWN);
1884	>	shutdown = true;
1885		tryTerminate(false);
1886		}
1887
#	Line 1757 | Line 1903 | public class ForkJoinPool extends Abstra
1903		*/
1904		public List<Runnable> shutdownNow() {
1905		checkPermission();
1906	+	shutdown = true;
1907		tryTerminate(true);
1908		return Collections.emptyList();
1909		}
#	Line 1767 | Line 1914 | public class ForkJoinPool extends Abstra
1914		* @return {@code true} if all tasks have completed following shut down
1915		*/
1916		public boolean isTerminated() {
1917	<	return runState >= TERMINATED;
1917	>	long c = ctl;
1918	>	return ((c & STOP_BIT) != 0L &&
1919	>	(short)(c >>> TC_SHIFT) == -parallelism);
1920		}
1921
1922		/**
#	Line 1775 | Line 1924 | public class ForkJoinPool extends Abstra
1924		* commenced but not yet completed.  This method may be useful for
1925		* debugging. A return of {@code true} reported a sufficient
1926		* period after shutdown may indicate that submitted tasks have
1927	<	* ignored or suppressed interruption, causing this executor not
1928	<	* to properly terminate.
1927	>	* ignored or suppressed interruption, or are waiting for IO,
1928	>	* causing this executor not to properly terminate. (See the
1929	>	* advisory notes for class {@link ForkJoinTask} stating that
1930	>	* tasks should not normally entail blocking operations.  But if
1931	>	* they do, they must abort them on interrupt.)
1932		*
1933		* @return {@code true} if terminating but not yet terminated
1934		*/
1935		public boolean isTerminating() {
1936	<	return (runState & (TERMINATING|TERMINATED)) == TERMINATING;
1936	>	long c = ctl;
1937	>	return ((c & STOP_BIT) != 0L &&
1938	>	(short)(c >>> TC_SHIFT) != -parallelism);
1939		}
1940
1941		/**
1942		* Returns true if terminating or terminated. Used by ForkJoinWorkerThread.
1943		*/
1944		final boolean isAtLeastTerminating() {
1945	<	return runState >= TERMINATING;
1945	>	return (ctl & STOP_BIT) != 0L;
1946		}
1947
1948		/**
#	Line 1797 | Line 1951 | public class ForkJoinPool extends Abstra
1951		* @return {@code true} if this pool has been shut down
1952		*/
1953		public boolean isShutdown() {
1954	<	return runState >= SHUTDOWN;
1954	>	return shutdown;
1955		}
1956
1957		/**
#	Line 1813 | Line 1967 | public class ForkJoinPool extends Abstra
1967		*/
1968		public boolean awaitTermination(long timeout, TimeUnit unit)
1969		throws InterruptedException {
1970	+	long nanos = unit.toNanos(timeout);
1971	+	final ReentrantLock lock = this.submissionLock;
1972	+	lock.lock();
1973		try {
1974	<	termination.awaitAdvanceInterruptibly(0, timeout, unit);
1975	<	} catch (TimeoutException ex) {
1976	<	return false;
1974	>	for (;;) {
1975	>	if (isTerminated())
1976	>	return true;
1977	>	if (nanos <= 0)
1978	>	return false;
1979	>	nanos = termination.awaitNanos(nanos);
1980	>	}
1981	>	} finally {
1982	>	lock.unlock();
1983		}
1821	–	return true;
1984		}
1985
1986		/**
#	Line 1829 | Line 1991 | public class ForkJoinPool extends Abstra
1991		* {@code isReleasable} must return {@code true} if blocking is
1992		* not necessary. Method {@code block} blocks the current thread
1993		* if necessary (perhaps internally invoking {@code isReleasable}
1994	<	* before actually blocking). The unusual methods in this API
1995	<	* accommodate synchronizers that may, but don't usually, block
1996	<	* for long periods. Similarly, they allow more efficient internal
1997	<	* handling of cases in which additional workers may be, but
1998	<	* usually are not, needed to ensure sufficient parallelism.
1999	<	* Toward this end, implementations of method {@code isReleasable}
2000	<	* must be amenable to repeated invocation.
1994	>	* before actually blocking). These actions are performed by any
1995	>	* thread invoking {@link ForkJoinPool#managedBlock}.  The
1996	>	* unusual methods in this API accommodate synchronizers that may,
1997	>	* but don't usually, block for long periods. Similarly, they
1998	>	* allow more efficient internal handling of cases in which
1999	>	* additional workers may be, but usually are not, needed to
2000	>	* ensure sufficient parallelism.  Toward this end,
2001	>	* implementations of method {@code isReleasable} must be amenable
2002	>	* to repeated invocation.
2003		*
2004		* <p>For example, here is a ManagedBlocker based on a
2005		* ReentrantLock:
#	Line 1937 | Line 2101 | public class ForkJoinPool extends Abstra
2101		}
2102
2103		// Unsafe mechanics
2104	<
2105	<	private static final sun.misc.Unsafe UNSAFE = getUnsafe();
2106	<	private static final long workerCountsOffset =
2107	<	objectFieldOffset("workerCounts", ForkJoinPool.class);
2108	<	private static final long runStateOffset =
2109	<	objectFieldOffset("runState", ForkJoinPool.class);
2110	<	private static final long eventCountOffset =
2111	<	objectFieldOffset("eventCount", ForkJoinPool.class);
2112	<	private static final long eventWaitersOffset =
2113	<	objectFieldOffset("eventWaiters", ForkJoinPool.class);
2114	<	private static final long stealCountOffset =
2115	<	objectFieldOffset("stealCount", ForkJoinPool.class);
2116	<	private static final long spareWaitersOffset =
2117	<	objectFieldOffset("spareWaiters", ForkJoinPool.class);
2118	<
2119	<	private static long objectFieldOffset(String field, Class<?> klazz) {
2104	>	private static final sun.misc.Unsafe UNSAFE;
2105	>	private static final long ctlOffset;
2106	>	private static final long stealCountOffset;
2107	>	private static final long blockedCountOffset;
2108	>	private static final long quiescerCountOffset;
2109	>	private static final long scanGuardOffset;
2110	>	private static final long nextWorkerNumberOffset;
2111	>	private static final long ABASE;
2112	>	private static final int ASHIFT;
2113	>
2114	>	static {
2115	>	poolNumberGenerator = new AtomicInteger();
2116	>	workerSeedGenerator = new Random();
2117	>	modifyThreadPermission = new RuntimePermission("modifyThread");
2118	>	defaultForkJoinWorkerThreadFactory =
2119	>	new DefaultForkJoinWorkerThreadFactory();
2120	>	int s;
2121		try {
2122	<	return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
2123	<	} catch (NoSuchFieldException e) {
2124	<	// Convert Exception to corresponding Error
2125	<	NoSuchFieldError error = new NoSuchFieldError(field);
2126	<	error.initCause(e);
2127	<	throw error;
2128	<	}
2122	>	UNSAFE = getUnsafe();
2123	>	Class k = ForkJoinPool.class;
2124	>	ctlOffset = UNSAFE.objectFieldOffset
2125	>	(k.getDeclaredField("ctl"));
2126	>	stealCountOffset = UNSAFE.objectFieldOffset
2127	>	(k.getDeclaredField("stealCount"));
2128	>	blockedCountOffset = UNSAFE.objectFieldOffset
2129	>	(k.getDeclaredField("blockedCount"));
2130	>	quiescerCountOffset = UNSAFE.objectFieldOffset
2131	>	(k.getDeclaredField("quiescerCount"));
2132	>	scanGuardOffset = UNSAFE.objectFieldOffset
2133	>	(k.getDeclaredField("scanGuard"));
2134	>	nextWorkerNumberOffset = UNSAFE.objectFieldOffset
2135	>	(k.getDeclaredField("nextWorkerNumber"));
2136	>	Class a = ForkJoinTask[].class;
2137	>	ABASE = UNSAFE.arrayBaseOffset(a);
2138	>	s = UNSAFE.arrayIndexScale(a);
2139	>	} catch (Exception e) {
2140	>	throw new Error(e);
2141	>	}
2142	>	if ((s & (s-1)) != 0)
2143	>	throw new Error("data type scale not a power of two");
2144	>	ASHIFT = 31 - Integer.numberOfLeadingZeros(s);
2145		}
2146
2147		/**

Diff Legend

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