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 1.116 by dl, Fri Jan 27 17:27:28 2012 UTC vs.
Revision 1.128 by dl, Mon Apr 9 13:11:44 2012 UTC

#	Line 5 | Line 5
5		*/
6
7		package jsr166y;
8	–
8		import java.util.ArrayList;
9		import java.util.Arrays;
10		import java.util.Collection;
#	Line 21 | Line 20 | import java.util.concurrent.RunnableFutu
20		import java.util.concurrent.TimeUnit;
21		import java.util.concurrent.atomic.AtomicInteger;
22		import java.util.concurrent.atomic.AtomicLong;
23	<	import java.util.concurrent.locks.ReentrantLock;
23	>	import java.util.concurrent.locks.AbstractQueuedSynchronizer;
24		import java.util.concurrent.locks.Condition;
25
26		/**
#	Line 60 | Line 59 | import java.util.concurrent.locks.Condit
59		* convenient form for informal monitoring.
60		*
61		* <p> As is the case with other ExecutorServices, there are three
62	<	* main task execution methods summarized in the following
63	<	* table. These are designed to be used primarily by clients not
64	<	* already engaged in fork/join computations in the current pool.  The
65	<	* main forms of these methods accept instances of {@code
66	<	* ForkJoinTask}, but overloaded forms also allow mixed execution of
67	<	* plain {@code Runnable}- or {@code Callable}- based activities as
68	<	* well.  However, tasks that are already executing in a pool should
69	<	* normally instead use the within-computation forms listed in the
70	<	* table unless using async event-style tasks that are not usually
71	<	* joined, in which case there is little difference among choice of
73	<	* methods.
62	>	* main task execution methods summarized in the following table.
63	>	* These are designed to be used primarily by clients not already
64	>	* engaged in fork/join computations in the current pool.  The main
65	>	* forms of these methods accept instances of {@code ForkJoinTask},
66	>	* but overloaded forms also allow mixed execution of plain {@code
67	>	* Runnable}- or {@code Callable}- based activities as well.  However,
68	>	* tasks that are already executing in a pool should normally instead
69	>	* use the within-computation forms listed in the table unless using
70	>	* async event-style tasks that are not usually joined, in which case
71	>	* there is little difference among choice of methods.
72		*
73		* <table BORDER CELLPADDING=3 CELLSPACING=1>
74		*  <tr>
#	Line 131 | Line 129 | public class ForkJoinPool extends Abstra
129		*
130		* This class and its nested classes provide the main
131		* functionality and control for a set of worker threads:
132	<	* Submissions from non-FJ threads enter into submission
133	<	* queues. Workers take these tasks and typically split them into
134	<	* subtasks that may be stolen by other workers.  Preference rules
135	<	* give first priority to processing tasks from their own queues
136	<	* (LIFO or FIFO, depending on mode), then to randomized FIFO
137	<	* steals of tasks in other queues.
132	>	* Submissions from non-FJ threads enter into submission queues.
133	>	* Workers take these tasks and typically split them into subtasks
134	>	* that may be stolen by other workers.  Preference rules give
135	>	* first priority to processing tasks from their own queues (LIFO
136	>	* or FIFO, depending on mode), then to randomized FIFO steals of
137	>	* tasks in other queues.
138		*
139	<	* WorkQueues.
139	>	* WorkQueues
140		* ==========
141		*
142		* Most operations occur within work-stealing queues (in nested
#	Line 156 | Line 154 | public class ForkJoinPool extends Abstra
154		* (http://research.sun.com/scalable/pubs/index.html) and
155		* "Idempotent work stealing" by Michael, Saraswat, and Vechev,
156		* PPoPP 2009 (http://portal.acm.org/citation.cfm?id=1504186).
157	<	* The main differences ultimately stem from gc requirements that
157	>	* The main differences ultimately stem from GC requirements that
158		* we null out taken slots as soon as we can, to maintain as small
159		* a footprint as possible even in programs generating huge
160		* numbers of tasks. To accomplish this, we shift the CAS
#	Line 178 | Line 176 | public class ForkJoinPool extends Abstra
176		* If an attempted steal fails, a thief always chooses a different
177		* random victim target to try next. So, in order for one thief to
178		* progress, it suffices for any in-progress poll or new push on
179	<	* any empty queue to complete.
179	>	* any empty queue to complete. (This is why we normally use
180	>	* method pollAt and its variants that try once at the apparent
181	>	* base index, else consider alternative actions, rather than
182	>	* method poll.)
183		*
184		* This approach also enables support of a user mode in which local
185		* task processing is in FIFO, not LIFO order, simply by using
#	Line 188 | Line 189 | public class ForkJoinPool extends Abstra
189		* rarely provide the best possible performance on a given
190		* machine, but portably provide good throughput by averaging over
191		* these factors.  (Further, even if we did try to use such
192	<	* information, we do not usually have a basis for exploiting
193	<	* it. For example, some sets of tasks profit from cache
194	<	* affinities, but others are harmed by cache pollution effects.)
192	>	* information, we do not usually have a basis for exploiting it.
193	>	* For example, some sets of tasks profit from cache affinities,
194	>	* but others are harmed by cache pollution effects.)
195		*
196		* WorkQueues are also used in a similar way for tasks submitted
197		* to the pool. We cannot mix these tasks in the same queues used
#	Line 204 | Line 205 | public class ForkJoinPool extends Abstra
205		* take tasks, and they are multiplexed on to a finite number of
206		* shared work queues. However, classes are set up so that future
207		* extensions could allow submitters to optionally help perform
208	<	* tasks as well. Pool submissions from internal workers are also
209	<	* allowed, but use randomized rather than thread-hashed queue
210	<	* indices to avoid imbalance.  Insertion of tasks in shared mode
211	<	* requires a lock (mainly to protect in the case of resizing) but
212	<	* we use only a simple spinlock (using bits in field runState),
213	<	* because submitters encountering a busy queue try or create
213	<	* others so never block.
208	>	* tasks as well. Insertion of tasks in shared mode requires a
209	>	* lock (mainly to protect in the case of resizing) but we use
210	>	* only a simple spinlock (using bits in field runState), because
211	>	* submitters encountering a busy queue move on to try or create
212	>	* other queues -- they block only when creating and registering
213	>	* new queues.
214		*
215	<	* Management.
215	>	* Management
216		* ==========
217		*
218		* The main throughput advantages of work-stealing stem from
#	Line 222 | Line 222 | public class ForkJoinPool extends Abstra
222		* tactic for avoiding bottlenecks is packing nearly all
223		* essentially atomic control state into two volatile variables
224		* that are by far most often read (not written) as status and
225	<	* consistency checks
225	>	* consistency checks.
226		*
227		* Field "ctl" contains 64 bits holding all the information needed
228		* to atomically decide to add, inactivate, enqueue (on an event
#	Line 248 | Line 248 | public class ForkJoinPool extends Abstra
248		* readers must tolerate null slots. Shared (submission) queues
249		* are at even indices, worker queues at odd indices. Grouping
250		* them together in this way simplifies and speeds up task
251	<	* scanning. To avoid flailing during start-up, the array is
252	<	* presized to hold twice #parallelism workers (which is unlikely
253	<	* to need further resizing during execution). But to avoid
254	<	* dealing with so many null slots, variable runState includes a
255	<	* mask for the nearest power of two that contains all current
256	<	* workers.  All worker thread creation is on-demand, triggered by
257	<	* task submissions, replacement of terminated workers, and/or
251	>	* scanning.
252	>	*
253	>	* All worker thread creation is on-demand, triggered by task
254	>	* submissions, replacement of terminated workers, and/or
255		* compensation for blocked workers. However, all other support
256		* code is set up to work with other policies.  To ensure that we
257		* do not hold on to worker references that would prevent GC, ALL
#	Line 267 | Line 264 | public class ForkJoinPool extends Abstra
264		* both index-check and null-check the IDs. All such accesses
265		* ignore bad IDs by returning out early from what they are doing,
266		* since this can only be associated with termination, in which
267	<	* case it is OK to give up.
268	<	*
269	<	* All uses of the workQueues array check that it is non-null
270	<	* (even if previously non-null). This allows nulling during
271	<	* termination, which is currently not necessary, but remains an
272	<	* option for resource-revocation-based shutdown schemes. It also
276	<	* helps reduce JIT issuance of uncommon-trap code, which tends to
267	>	* case it is OK to give up.  All uses of the workQueues array
268	>	* also check that it is non-null (even if previously
269	>	* non-null). This allows nulling during termination, which is
270	>	* currently not necessary, but remains an option for
271	>	* resource-revocation-based shutdown schemes. It also helps
272	>	* reduce JIT issuance of uncommon-trap code, which tends to
273		* unnecessarily complicate control flow in some methods.
274		*
275		* Event Queuing. Unlike HPC work-stealing frameworks, we cannot
#	Line 301 | Line 297 | public class ForkJoinPool extends Abstra
297		* some other queued worker rather than itself, which has the same
298		* net effect. Because enqueued workers may actually be rescanning
299		* rather than waiting, we set and clear the "parker" field of
300	<	* Workqueues to reduce unnecessary calls to unpark.  (This
300	>	* WorkQueues to reduce unnecessary calls to unpark.  (This
301		* requires a secondary recheck to avoid missed signals.)  Note
302		* the unusual conventions about Thread.interrupts surrounding
303		* parking and other blocking: Because interrupts are used solely
#	Line 329 | Line 325 | public class ForkJoinPool extends Abstra
325		* terminating all workers after long periods of non-use.
326		*
327		* Shutdown and Termination. A call to shutdownNow atomically sets
328	<	* a runState bit and then (non-atomically) sets each workers
328	>	* a runState bit and then (non-atomically) sets each worker's
329		* runState status, cancels all unprocessed tasks, and wakes up
330		* all waiting workers.  Detecting whether termination should
331		* commence after a non-abrupt shutdown() call requires more work
#	Line 338 | Line 334 | public class ForkJoinPool extends Abstra
334		* indication but non-abrupt shutdown still requires a rechecking
335		* scan for any workers that are inactive but not queued.
336		*
337	<	* Joining Tasks.
338	<	* ==============
337	>	* Joining Tasks
338	>	* =============
339		*
340		* Any of several actions may be taken when one worker is waiting
341	<	* to join a task stolen (or always held by) another.  Because we
341	>	* to join a task stolen (or always held) by another.  Because we
342		* are multiplexing many tasks on to a pool of workers, we can't
343		* just let them block (as in Thread.join).  We also cannot just
344		* reassign the joiner's run-time stack with another and replace
345		* it later, which would be a form of "continuation", that even if
346		* possible is not necessarily a good idea since we sometimes need
347	<	* both an unblocked task and its continuation to
348	<	* progress. Instead we combine two tactics:
347	>	* both an unblocked task and its continuation to progress.
348	>	* Instead we combine two tactics:
349		*
350		*   Helping: Arranging for the joiner to execute some task that it
351		*      would be running if the steal had not occurred.
#	Line 384 | Line 380 | public class ForkJoinPool extends Abstra
380		* (http://portal.acm.org/citation.cfm?id=155354). It differs in
381		* that: (1) We only maintain dependency links across workers upon
382		* steals, rather than use per-task bookkeeping.  This sometimes
383	<	* requires a linear scan of workers array to locate stealers, but
384	<	* often doesn't because stealers leave hints (that may become
383	>	* requires a linear scan of workQueues array to locate stealers,
384	>	* but often doesn't because stealers leave hints (that may become
385		* stale/wrong) of where to locate them.  A stealHint is only a
386		* hint because a worker might have had multiple steals and the
387		* hint records only one of them (usually the most current).
#	Line 396 | Line 392 | public class ForkJoinPool extends Abstra
392		* which means that we miss links in the chain during long-lived
393		* tasks, GC stalls etc (which is OK since blocking in such cases
394		* is usually a good idea).  (4) We bound the number of attempts
395	<	* to find work (see MAX_HELP_DEPTH) and fall back to suspending
396	<	* the worker and if necessary replacing it with another.
395	>	* to find work (see MAX_HELP) and fall back to suspending the
396	>	* worker and if necessary replacing it with another.
397		*
398		* It is impossible to keep exactly the target parallelism number
399		* of threads running at any given time.  Determining the
400		* existence of conservatively safe helping targets, the
401		* availability of already-created spares, and the apparent need
402		* to create new spares are all racy, so we rely on multiple
403	<	* retries of each.  Currently, in keeping with on-demand
404	<	* signalling policy, we compensate only if blocking would leave
405	<	* less than one active (non-waiting, non-blocked) worker.
406	<	* Additionally, to avoid some false alarms due to GC, lagging
407	<	* counters, system activity, etc, compensated blocking for joins
408	<	* is only attempted after rechecks stabilize in
409	<	* ForkJoinTask.awaitJoin. (Retries are interspersed with
410	<	* Thread.yield, for good citizenship.)
403	>	* retries of each.  Compensation in the apparent absence of
404	>	* helping opportunities is challenging to control on JVMs, where
405	>	* GC and other activities can stall progress of tasks that in
406	>	* turn stall out many other dependent tasks, without us being
407	>	* able to determine whether they will ever require compensation.
408	>	* Even though work-stealing otherwise encounters little
409	>	* degradation in the presence of more threads than cores,
410	>	* aggressively adding new threads in such cases entails risk of
411	>	* unwanted positive feedback control loops in which more threads
412	>	* cause more dependent stalls (as well as delayed progress of
413	>	* unblocked threads to the point that we know they are available)
414	>	* leading to more situations requiring more threads, and so
415	>	* on. This aspect of control can be seen as an (analytically
416	>	* intractable) game with an opponent that may choose the worst
417	>	* (for us) active thread to stall at any time.  We take several
418	>	* precautions to bound losses (and thus bound gains), mainly in
419	>	* methods tryCompensate and awaitJoin: (1) We only try
420	>	* compensation after attempting enough helping steps (measured
421	>	* via counting and timing) that we have already consumed the
422	>	* estimated cost of creating and activating a new thread.  (2) We
423	>	* allow up to 50% of threads to be blocked before initially
424	>	* adding any others, and unless completely saturated, check that
425	>	* some work is available for a new worker before adding. Also, we
426	>	* create up to only 50% more threads until entering a mode that
427	>	* only adds a thread if all others are possibly blocked.  All
428	>	* together, this means that we might be half as fast to react,
429	>	* and create half as many threads as possible in the ideal case,
430	>	* but present vastly fewer anomalies in all other cases compared
431	>	* to both more aggressive and more conservative alternatives.
432		*
433		* Style notes: There is a lot of representation-level coupling
434		* among classes ForkJoinPool, ForkJoinWorkerThread, and
#	Line 419 | Line 436 | public class ForkJoinPool extends Abstra
436		* managed by ForkJoinPool, so are directly accessed.  There is
437		* little point trying to reduce this, since any associated future
438		* changes in representations will need to be accompanied by
439	<	* algorithmic changes anyway. All together, these low-level
440	<	* implementation choices produce as much as a factor of 4
441	<	* performance improvement compared to naive implementations, and
442	<	* enable the processing of billions of tasks per second, at the
443	<	* expense of some ugliness.
444	<	*
445	<	* Methods signalWork() and scan() are the main bottlenecks so are
446	<	* especially heavily micro-optimized/mangled.  There are lots of
447	<	* inline assignments (of form "while ((local = field) != 0)")
448	<	* which are usually the simplest way to ensure the required read
449	<	* orderings (which are sometimes critical). This leads to a
450	<	* "C"-like style of listing declarations of these locals at the
451	<	* heads of methods or blocks.  There are several occurrences of
452	<	* the unusual "do {} while (!cas...)"  which is the simplest way
453	<	* to force an update of a CAS'ed variable. There are also other
454	<	* coding oddities that help some methods perform reasonably even
455	<	* when interpreted (not compiled).
456	<	*
457	<	* The order of declarations in this file is: (1) declarations of
458	<	* statics (2) fields (along with constants used when unpacking
459	<	* some of them), listed in an order that tends to reduce
460	<	* contention among them a bit under most JVMs; (3) nested
461	<	* classes; (4) internal control methods; (5) callbacks and other
462	<	* support for ForkJoinTask methods; (6) exported methods (plus a
446	<	* few little helpers); (7) static block initializing all statics
447	<	* in a minimally dependent order.
439	>	* algorithmic changes anyway. Several methods intrinsically
440	>	* sprawl because they must accumulate sets of consistent reads of
441	>	* volatiles held in local variables.  Methods signalWork() and
442	>	* scan() are the main bottlenecks, so are especially heavily
443	>	* micro-optimized/mangled.  There are lots of inline assignments
444	>	* (of form "while ((local = field) != 0)") which are usually the
445	>	* simplest way to ensure the required read orderings (which are
446	>	* sometimes critical). This leads to a "C"-like style of listing
447	>	* declarations of these locals at the heads of methods or blocks.
448	>	* There are several occurrences of the unusual "do {} while
449	>	* (!cas...)"  which is the simplest way to force an update of a
450	>	* CAS'ed variable. There are also other coding oddities that help
451	>	* some methods perform reasonably even when interpreted (not
452	>	* compiled).
453	>	*
454	>	* The order of declarations in this file is:
455	>	* (1) Static utility functions
456	>	* (2) Nested (static) classes
457	>	* (3) Static fields
458	>	* (4) Fields, along with constants used when unpacking some of them
459	>	* (5) Internal control methods
460	>	* (6) Callbacks and other support for ForkJoinTask methods
461	>	* (7) Exported methods
462	>	* (8) Static block initializing statics in minimally dependent order
463		*/
464
465	+	// Static utilities
466	+
467	+	/**
468	+	* If there is a security manager, makes sure caller has
469	+	* permission to modify threads.
470	+	*/
471	+	private static void checkPermission() {
472	+	SecurityManager security = System.getSecurityManager();
473	+	if (security != null)
474	+	security.checkPermission(modifyThreadPermission);
475	+	}
476	+
477	+	// Nested classes
478	+
479		/**
480		* Factory for creating new {@link ForkJoinWorkerThread}s.
481		* A {@code ForkJoinWorkerThreadFactory} must be defined and used
#	Line 475 | Line 504 | public class ForkJoinPool extends Abstra
504		}
505
506		/**
507	<	* Creates a new ForkJoinWorkerThread. This factory is used unless
508	<	* overridden in ForkJoinPool constructors.
509	<	*/
510	<	public static final ForkJoinWorkerThreadFactory
511	<	defaultForkJoinWorkerThreadFactory;
512	<
513	<	/**
514	<	* Permission required for callers of methods that may start or
515	<	* kill threads.
516	<	*/
517	<	private static final RuntimePermission modifyThreadPermission;
518	<
519	<	/**
520	<	* If there is a security manager, makes sure caller has
521	<	* permission to modify threads.
522	<	*/
523	<	private static void checkPermission() {
524	<	SecurityManager security = System.getSecurityManager();
525	<	if (security != null)
526	<	security.checkPermission(modifyThreadPermission);
507	>	* A simple non-reentrant lock used for exclusion when managing
508	>	* queues and workers. We use a custom lock so that we can readily
509	>	* probe lock state in constructions that check among alternative
510	>	* actions. The lock is normally only very briefly held, and
511	>	* sometimes treated as a spinlock, but other usages block to
512	>	* reduce overall contention in those cases where locked code
513	>	* bodies perform allocation/resizing.
514	>	*/
515	>	static final class Mutex extends AbstractQueuedSynchronizer {
516	>	public final boolean tryAcquire(int ignore) {
517	>	return compareAndSetState(0, 1);
518	>	}
519	>	public final boolean tryRelease(int ignore) {
520	>	setState(0);
521	>	return true;
522	>	}
523	>	public final void lock() { acquire(0); }
524	>	public final void unlock() { release(0); }
525	>	public final boolean isHeldExclusively() { return getState() == 1; }
526	>	public final Condition newCondition() { return new ConditionObject(); }
527		}
528
529		/**
530	<	* Generator for assigning sequence numbers as pool names.
531	<	*/
532	<	private static final AtomicInteger poolNumberGenerator;
533	<
505	<	/**
506	<	* Bits and masks for control variables
507	<	*
508	<	* Field ctl is a long packed with:
509	<	* AC: Number of active running workers minus target parallelism (16 bits)
510	<	* TC: Number of total workers minus target parallelism (16 bits)
511	<	* ST: true if pool is terminating (1 bit)
512	<	* EC: the wait count of top waiting thread (15 bits)
513	<	* ID: ~(poolIndex >>> 1) of top of Treiber stack of waiters (16 bits)
514	<	*
515	<	* When convenient, we can extract the upper 32 bits of counts and
516	<	* the lower 32 bits of queue state, u = (int)(ctl >>> 32) and e =
517	<	* (int)ctl.  The ec field is never accessed alone, but always
518	<	* together with id and st. The offsets of counts by the target
519	<	* parallelism and the positionings of fields makes it possible to
520	<	* perform the most common checks via sign tests of fields: When
521	<	* ac is negative, there are not enough active workers, when tc is
522	<	* negative, there are not enough total workers, when id is
523	<	* negative, there is at least one waiting worker, and when e is
524	<	* negative, the pool is terminating.  To deal with these possibly
525	<	* negative fields, we use casts in and out of "short" and/or
526	<	* signed shifts to maintain signedness.
527	<	*
528	<	* When a thread is queued (inactivated), its eventCount field is
529	<	* negative, which is the only way to tell if a worker is
530	<	* prevented from executing tasks, even though it must continue to
531	<	* scan for them to avoid queuing races.
532	<	*
533	<	* Field runState is an int packed with:
534	<	* SHUTDOWN: true if shutdown is enabled (1 bit)
535	<	* SEQ:  a sequence number updated upon (de)registering workers (15 bits)
536	<	* MASK: mask (power of 2 - 1) covering all registered poolIndexes (16 bits)
537	<	*
538	<	* The combination of mask and sequence number enables simple
539	<	* consistency checks: Staleness of read-only operations on the
540	<	* workers and queues arrays can be checked by comparing runState
541	<	* before vs after the reads. The low 16 bits (i.e, anding with
542	<	* SMASK) hold (the smallest power of two covering all worker
543	<	* indices, minus one.  The mask for queues (vs workers) is twice
544	<	* this value plus 1.
545	<	*/
546	<
547	<	// bit positions/shifts for fields
548	<	private static final int  AC_SHIFT   = 48;
549	<	private static final int  TC_SHIFT   = 32;
550	<	private static final int  ST_SHIFT   = 31;
551	<	private static final int  EC_SHIFT   = 16;
552	<
553	<	// bounds
554	<	private static final int  MAX_ID     = 0x7fff;  // max poolIndex
555	<	private static final int  SMASK      = 0xffff;  // mask short bits
556	<	private static final int  SHORT_SIGN = 1 << 15;
557	<	private static final int  INT_SIGN   = 1 << 31;
558	<
559	<	// masks
560	<	private static final long STOP_BIT   = 0x0001L << ST_SHIFT;
561	<	private static final long AC_MASK    = ((long)SMASK) << AC_SHIFT;
562	<	private static final long TC_MASK    = ((long)SMASK) << TC_SHIFT;
563	<
564	<	// units for incrementing and decrementing
565	<	private static final long TC_UNIT    = 1L << TC_SHIFT;
566	<	private static final long AC_UNIT    = 1L << AC_SHIFT;
567	<
568	<	// masks and units for dealing with u = (int)(ctl >>> 32)
569	<	private static final int  UAC_SHIFT  = AC_SHIFT - 32;
570	<	private static final int  UTC_SHIFT  = TC_SHIFT - 32;
571	<	private static final int  UAC_MASK   = SMASK << UAC_SHIFT;
572	<	private static final int  UTC_MASK   = SMASK << UTC_SHIFT;
573	<	private static final int  UAC_UNIT   = 1 << UAC_SHIFT;
574	<	private static final int  UTC_UNIT   = 1 << UTC_SHIFT;
575	<
576	<	// masks and units for dealing with e = (int)ctl
577	<	private static final int E_MASK      = 0x7fffffff; // no STOP_BIT
578	<	private static final int E_SEQ       = 1 << EC_SHIFT;
579	<
580	<	// runState bits
581	<	private static final int SHUTDOWN    = 1 << 31;
582	<	private static final int RS_SEQ      = 1 << 16;
583	<	private static final int RS_SEQ_MASK = 0x7fff0000;
584	<
585	<	// access mode for WorkQueue
586	<	static final int LIFO_QUEUE          =  0;
587	<	static final int FIFO_QUEUE          =  1;
588	<	static final int SHARED_QUEUE        = -1;
589	<
590	<	/**
591	<	* The wakeup interval (in nanoseconds) for a worker waiting for a
592	<	* task when the pool is quiescent to instead try to shrink the
593	<	* number of workers.  The exact value does not matter too
594	<	* much. It must be short enough to release resources during
595	<	* sustained periods of idleness, but not so short that threads
596	<	* are continually re-created.
597	<	*/
598	<	private static final long SHRINK_RATE =
599	<	4L * 1000L * 1000L * 1000L; // 4 seconds
600	<
601	<	/**
602	<	* The timeout value for attempted shrinkage, includes
603	<	* some slop to cope with system timer imprecision.
604	<	*/
605	<	private static final long SHRINK_TIMEOUT = SHRINK_RATE - (SHRINK_RATE / 10);
606	<
607	<	/**
608	<	* The maximum stolen->joining link depth allowed in tryHelpStealer.
609	<	* Depths for legitimate chains are unbounded, but we use a fixed
610	<	* constant to avoid (otherwise unchecked) cycles and to bound
611	<	* staleness of traversal parameters at the expense of sometimes
612	<	* blocking when we could be helping.
613	<	*/
614	<	private static final int MAX_HELP_DEPTH = 16;
615	<
616	<	/*
617	<	* Field layout order in this class tends to matter more than one
618	<	* would like. Runtime layout order is only loosely related to
619	<	* declaration order and may differ across JVMs, but the following
620	<	* empirically works OK on current JVMs.
530	>	* Class for artificial tasks that are used to replace the target
531	>	* of local joins if they are removed from an interior queue slot
532	>	* in WorkQueue.tryRemoveAndExec. We don't need the proxy to
533	>	* actually do anything beyond having a unique identity.
534		*/
535	<
536	<	volatile long ctl;                       // main pool control
537	<	final int parallelism;                   // parallelism level
538	<	final int localMode;                     // per-worker scheduling mode
539	<	int nextPoolIndex;                       // hint used in registerWorker
540	<	volatile int runState;                   // shutdown status, seq, and mask
628	<	WorkQueue[] workQueues;                  // main registry
629	<	final ReentrantLock lock;                // for registration
630	<	final Condition termination;             // for awaitTermination
631	<	final ForkJoinWorkerThreadFactory factory; // factory for new workers
632	<	final Thread.UncaughtExceptionHandler ueh; // per-worker UEH
633	<	final AtomicLong stealCount;             // collect counts when terminated
634	<	final AtomicInteger nextWorkerNumber;    // to create worker name string
635	<	final String workerNamePrefix;           // Prefix for assigning worker names
535	>	static final class EmptyTask extends ForkJoinTask<Void> {
536	>	EmptyTask() { status = ForkJoinTask.NORMAL; } // force done
537	>	public final Void getRawResult() { return null; }
538	>	public final void setRawResult(Void x) {}
539	>	public final boolean exec() { return true; }
540	>	}
541
542		/**
543		* Queues supporting work-stealing as well as external task
#	Line 683 | Line 588 | public class ForkJoinPool extends Abstra
588		* avoiding really bad worst-case access. (Until better JVM
589		* support is in place, this padding is dependent on transient
590		* properties of JVM field layout rules.)  We also take care in
591	<	* allocating and sizing and resizing the array. Non-shared queue
591	>	* allocating, sizing and resizing the array. Non-shared queue
592		* arrays are initialized (via method growArray) by workers before
593		* use. Others are allocated on first use.
594		*/
595		static final class WorkQueue {
596		/**
597		* Capacity of work-stealing queue array upon initialization.
598	<	* Must be a power of two; at least 4, but set larger to
599	<	* reduce cacheline sharing among queues.
598	>	* Must be a power of two; at least 4, but should be larger to
599	>	* reduce or eliminate cacheline sharing among queues.
600	>	* Currently, it is much larger, as a partial workaround for
601	>	* the fact that JVMs often place arrays in locations that
602	>	* share GC bookkeeping (especially cardmarks) such that
603	>	* per-write accesses encounter serious memory contention.
604		*/
605	<	static final int INITIAL_QUEUE_CAPACITY = 1 << 8;
605	>	static final int INITIAL_QUEUE_CAPACITY = 1 << 13;
606
607		/**
608		* Maximum size for queue arrays. Must be a power of two less
#	Line 717 | Line 626 | public class ForkJoinPool extends Abstra
626		volatile int base;         // index of next slot for poll
627		int top;                   // index of next slot for push
628		ForkJoinTask<?>[] array;   // the elements (initially unallocated)
629	+	final ForkJoinPool pool;   // the containing pool (may be null)
630		final ForkJoinWorkerThread owner; // owning thread or null if shared
631		volatile Thread parker;    // == owner during call to park; else null
632	<	ForkJoinTask<?> currentJoin;  // task being joined in awaitJoin
632	>	volatile ForkJoinTask<?> currentJoin;  // task being joined in awaitJoin
633		ForkJoinTask<?> currentSteal; // current non-local task being executed
634		// Heuristic padding to ameliorate unfortunate memory placements
635	<	Object p00, p01, p02, p03, p04, p05, p06, p07, p08, p09, p0a;
635	>	Object p00, p01, p02, p03, p04, p05, p06, p07;
636	>	Object p08, p09, p0a, p0b, p0c, p0d, p0e;
637
638	<	WorkQueue(ForkJoinWorkerThread owner, int mode) {
728	<	this.owner = owner;
638	>	WorkQueue(ForkJoinPool pool, ForkJoinWorkerThread owner, int mode) {
639		this.mode = mode;
640	+	this.pool = pool;
641	+	this.owner = owner;
642		// Place indices in the center of array (that is not yet allocated)
643		base = top = INITIAL_QUEUE_CAPACITY >>> 1;
644		}
645
646		/**
647	<	* Returns number of tasks in the queue
647	>	* Returns the approximate number of tasks in the queue.
648		*/
649		final int queueSize() {
650	<	int n = base - top; // non-owner callers must read base first
651	<	return (n >= 0) ? 0 : -n;
650	>	int n = base - top;       // non-owner callers must read base first
651	>	return (n >= 0) ? 0 : -n; // ignore transient negative
652	>	}
653	>
654	>	/**
655	>	* Provides a more accurate estimate of whether this queue has
656	>	* any tasks than does queueSize, by checking whether a
657	>	* near-empty queue has at least one unclaimed task.
658	>	*/
659	>	final boolean isEmpty() {
660	>	ForkJoinTask<?>[] a; int m, s;
661	>	int n = base - (s = top);
662	>	return (n >= 0 ||
663	>	(n == -1 &&
664	>	((a = array) == null ||
665	>	(m = a.length - 1) < 0 ||
666	>	U.getObjectVolatile
667	>	(a, ((m & (s - 1)) << ASHIFT) + ABASE) == null)));
668		}
669
670		/**
671		* Pushes a task. Call only by owner in unshared queues.
672		*
673		* @param task the task. Caller must ensure non-null.
674	<	* @param p, if non-null, pool to signal if necessary
747	<	* @throw RejectedExecutionException if array cannot
748	<	* be resized
674	>	* @throw RejectedExecutionException if array cannot be resized
675		*/
676	<	final void push(ForkJoinTask<?> task, ForkJoinPool p) {
677	<	ForkJoinTask<?>[] a;
676	>	final void push(ForkJoinTask<?> task) {
677	>	ForkJoinTask<?>[] a; ForkJoinPool p;
678		int s = top, m, n;
679		if ((a = array) != null) {    // ignore if queue removed
680		U.putOrderedObject
681		(a, (((m = a.length - 1) & s) << ASHIFT) + ABASE, task);
682		if ((n = (top = s + 1) - base) <= 2) {
683	<	if (p != null)
683	>	if ((p = pool) != null)
684		p.signalWork();
685		}
686		else if (n >= m)
#	Line 773 | Line 699 | public class ForkJoinPool extends Abstra
699		boolean submitted = false;
700		if (runState == 0 && U.compareAndSwapInt(this, RUNSTATE, 0, 1)) {
701		ForkJoinTask<?>[] a = array;
702	<	int s = top, n = s - base;
702	>	int s = top;
703		try {
704	<	if ((a != null && n < a.length - 1) ||
704	>	if ((a != null && a.length > s + 1 - base) ||
705		(a = growArray(false)) != null) { // must presize
706		int j = (((a.length - 1) & s) << ASHIFT) + ABASE;
707		U.putObject(a, (long)j, task);    // don't need "ordered"
#	Line 790 | Line 716 | public class ForkJoinPool extends Abstra
716		}
717
718		/**
719	<	* Takes next task, if one exists, in FIFO order.
719	>	* Takes next task, if one exists, in LIFO order.  Call only
720	>	* by owner in unshared queues. (We do not have a shared
721	>	* version of this method because it is never needed.)
722		*/
723	<	final ForkJoinTask<?> poll() {
724	<	ForkJoinTask<?>[] a; int b, i;
725	<	while ((b = base) - top < 0 && (a = array) != null &&
726	<	(i = (a.length - 1) & b) >= 0) {
727	<	int j = (i << ASHIFT) + ABASE;
728	<	ForkJoinTask<?> t = (ForkJoinTask<?>)U.getObjectVolatile(a, j);
729	<	if (t != null && base == b &&
723	>	final ForkJoinTask<?> pop() {
724	>	ForkJoinTask<?>[] a; ForkJoinTask<?> t; int m;
725	>	if ((a = array) != null && (m = a.length - 1) >= 0) {
726	>	for (int s; (s = top - 1) - base >= 0;) {
727	>	long j = ((m & s) << ASHIFT) + ABASE;
728	>	if ((t = (ForkJoinTask<?>)U.getObject(a, j)) == null)
729	>	break;
730	>	if (U.compareAndSwapObject(a, j, t, null)) {
731	>	top = s;
732	>	return t;
733	>	}
734	>	}
735	>	}
736	>	return null;
737	>	}
738	>
739	>	/**
740	>	* Takes a task in FIFO order if b is base of queue and a task
741	>	* can be claimed without contention. Specialized versions
742	>	* appear in ForkJoinPool methods scan and tryHelpStealer.
743	>	*/
744	>	final ForkJoinTask<?> pollAt(int b) {
745	>	ForkJoinTask<?> t; ForkJoinTask<?>[] a;
746	>	if ((a = array) != null) {
747	>	int j = (((a.length - 1) & b) << ASHIFT) + ABASE;
748	>	if ((t = (ForkJoinTask<?>)U.getObjectVolatile(a, j)) != null &&
749	>	base == b &&
750		U.compareAndSwapObject(a, j, t, null)) {
751		base = b + 1;
752		return t;
#	Line 808 | Line 756 | public class ForkJoinPool extends Abstra
756		}
757
758		/**
759	<	* Takes next task, if one exists, in LIFO order.
812	<	* Call only by owner in unshared queues.
759	>	* Takes next task, if one exists, in FIFO order.
760		*/
761	<	final ForkJoinTask<?> pop() {
762	<	ForkJoinTask<?> t; int m;
763	<	ForkJoinTask<?>[] a = array;
764	<	if (a != null && (m = a.length - 1) >= 0) {
765	<	for (int s; (s = top - 1) - base >= 0;) {
766	<	int j = ((m & s) << ASHIFT) + ABASE;
767	<	if ((t = (ForkJoinTask<?>)U.getObjectVolatile(a, j)) == null)
768	<	break;
769	<	if (U.compareAndSwapObject(a, j, t, null)) {
823	<	top = s;
761	>	final ForkJoinTask<?> poll() {
762	>	ForkJoinTask<?>[] a; int b; ForkJoinTask<?> t;
763	>	while ((b = base) - top < 0 && (a = array) != null) {
764	>	int j = (((a.length - 1) & b) << ASHIFT) + ABASE;
765	>	t = (ForkJoinTask<?>)U.getObjectVolatile(a, j);
766	>	if (t != null) {
767	>	if (base == b &&
768	>	U.compareAndSwapObject(a, j, t, null)) {
769	>	base = b + 1;
770		return t;
771		}
772		}
773	+	else if (base == b) {
774	+	if (b + 1 == top)
775	+	break;
776	+	Thread.yield(); // wait for lagging update
777	+	}
778		}
779		return null;
780		}
#	Line 848 | Line 799 | public class ForkJoinPool extends Abstra
799		}
800
801		/**
851	–	* Returns task at index b if b is current base of queue.
852	–	*/
853	–	final ForkJoinTask<?> pollAt(int b) {
854	–	ForkJoinTask<?>[] a; int i;
855	–	ForkJoinTask<?> task = null;
856	–	if ((a = array) != null && (i = ((a.length - 1) & b)) >= 0) {
857	–	int j = (i << ASHIFT) + ABASE;
858	–	ForkJoinTask<?> t = (ForkJoinTask<?>)U.getObjectVolatile(a, j);
859	–	if (t != null && base == b &&
860	–	U.compareAndSwapObject(a, j, t, null)) {
861	–	base = b + 1;
862	–	task = t;
863	–	}
864	–	}
865	–	return task;
866	–	}
867	–
868	–	/**
802		* Pops the given task only if it is at the current top.
803		*/
804		final boolean tryUnpush(ForkJoinTask<?> t) {
#	Line 883 | Line 816 | public class ForkJoinPool extends Abstra
816		* Polls the given task only if it is at the current base.
817		*/
818		final boolean pollFor(ForkJoinTask<?> task) {
819	<	ForkJoinTask<?>[] a; int b, i;
820	<	if ((b = base) - top < 0 && (a = array) != null &&
821	<	(i = (a.length - 1) & b) >= 0) {
889	<	int j = (i << ASHIFT) + ABASE;
819	>	ForkJoinTask<?>[] a; int b;
820	>	if ((b = base) - top < 0 && (a = array) != null) {
821	>	int j = (((a.length - 1) & b) << ASHIFT) + ABASE;
822		if (U.getObjectVolatile(a, j) == task && base == b &&
823		U.compareAndSwapObject(a, j, task, null)) {
824		base = b + 1;
#	Line 897 | Line 829 | public class ForkJoinPool extends Abstra
829		}
830
831		/**
900	–	* If present, removes from queue and executes the given task, or
901	–	* any other cancelled task. Returns (true) immediately on any CAS
902	–	* or consistency check failure so caller can retry.
903	–	*
904	–	* @return false if no progress can be made
905	–	*/
906	–	final boolean tryRemoveAndExec(ForkJoinTask<?> task) {
907	–	boolean removed = false, empty = true, progress = true;
908	–	ForkJoinTask<?>[] a; int m, s, b, n;
909	–	if ((a = array) != null && (m = a.length - 1) >= 0 &&
910	–	(n = (s = top) - (b = base)) > 0) {
911	–	for (ForkJoinTask<?> t;;) {           // traverse from s to b
912	–	int j = ((--s & m) << ASHIFT) + ABASE;
913	–	t = (ForkJoinTask<?>)U.getObjectVolatile(a, j);
914	–	if (t == null)                    // inconsistent length
915	–	break;
916	–	else if (t == task) {
917	–	if (s + 1 == top) {           // pop
918	–	if (!U.compareAndSwapObject(a, j, task, null))
919	–	break;
920	–	top = s;
921	–	removed = true;
922	–	}
923	–	else if (base == b)           // replace with proxy
924	–	removed = U.compareAndSwapObject(a, j, task,
925	–	new EmptyTask());
926	–	break;
927	–	}
928	–	else if (t.status >= 0)
929	–	empty = false;
930	–	else if (s + 1 == top) {          // pop and throw away
931	–	if (U.compareAndSwapObject(a, j, t, null))
932	–	top = s;
933	–	break;
934	–	}
935	–	if (--n == 0) {
936	–	if (!empty && base == b)
937	–	progress = false;
938	–	break;
939	–	}
940	–	}
941	–	}
942	–	if (removed)
943	–	task.doExec();
944	–	return progress;
945	–	}
946	–
947	–	/**
832		* Initializes or doubles the capacity of array. Call either
833		* by owner or with lock held -- it is OK for base, but not
834		* top, to move while resizings are in progress.
#	Line 980 | Line 864 | public class ForkJoinPool extends Abstra
864		}
865
866		/**
867	<	* Removes and cancels all known tasks, ignoring any exceptions
867	>	* Removes and cancels all known tasks, ignoring any exceptions.
868		*/
869		final void cancelAll() {
870		ForkJoinTask.cancelIgnoringExceptions(currentJoin);
#	Line 989 | Line 873 | public class ForkJoinPool extends Abstra
873		ForkJoinTask.cancelIgnoringExceptions(t);
874		}
875
876	+	/**
877	+	* Computes next value for random probes.  Scans don't require
878	+	* a very high quality generator, but also not a crummy one.
879	+	* Marsaglia xor-shift is cheap and works well enough.  Note:
880	+	* This is manually inlined in its usages in ForkJoinPool to
881	+	* avoid writes inside busy scan loops.
882	+	*/
883	+	final int nextSeed() {
884	+	int r = seed;
885	+	r ^= r << 13;
886	+	r ^= r >>> 17;
887	+	return seed = r ^= r << 5;
888	+	}
889	+
890		// Execution methods
891
892		/**
893	<	* Removes and runs tasks until empty, using local mode
996	<	* ordering.
893	>	* Pops and runs tasks until empty.
894		*/
895	<	final void runLocalTasks() {
896	<	if (base - top < 0) {
897	<	for (ForkJoinTask<?> t; (t = nextLocalTask()) != null; )
895	>	private void popAndExecAll() {
896	>	// A bit faster than repeated pop calls
897	>	ForkJoinTask<?>[] a; int m, s; long j; ForkJoinTask<?> t;
898	>	while ((a = array) != null && (m = a.length - 1) >= 0 &&
899	>	(s = top - 1) - base >= 0 &&
900	>	(t = ((ForkJoinTask<?>)
901	>	U.getObject(a, j = ((m & s) << ASHIFT) + ABASE)))
902	>	!= null) {
903	>	if (U.compareAndSwapObject(a, j, t, null)) {
904	>	top = s;
905		t.doExec();
906	+	}
907		}
908		}
909
910		/**
911	+	* Polls and runs tasks until empty.
912	+	*/
913	+	private void pollAndExecAll() {
914	+	for (ForkJoinTask<?> t; (t = poll()) != null;)
915	+	t.doExec();
916	+	}
917	+
918	+	/**
919	+	* If present, removes from queue and executes the given task, or
920	+	* any other cancelled task. Returns (true) immediately on any CAS
921	+	* or consistency check failure so caller can retry.
922	+	*
923	+	* @return 0 if no progress can be made, else positive
924	+	* (this unusual convention simplifies use with tryHelpStealer.)
925	+	*/
926	+	final int tryRemoveAndExec(ForkJoinTask<?> task) {
927	+	int stat = 1;
928	+	boolean removed = false, empty = true;
929	+	ForkJoinTask<?>[] a; int m, s, b, n;
930	+	if ((a = array) != null && (m = a.length - 1) >= 0 &&
931	+	(n = (s = top) - (b = base)) > 0) {
932	+	for (ForkJoinTask<?> t;;) {           // traverse from s to b
933	+	int j = ((--s & m) << ASHIFT) + ABASE;
934	+	t = (ForkJoinTask<?>)U.getObjectVolatile(a, j);
935	+	if (t == null)                    // inconsistent length
936	+	break;
937	+	else if (t == task) {
938	+	if (s + 1 == top) {           // pop
939	+	if (!U.compareAndSwapObject(a, j, task, null))
940	+	break;
941	+	top = s;
942	+	removed = true;
943	+	}
944	+	else if (base == b)           // replace with proxy
945	+	removed = U.compareAndSwapObject(a, j, task,
946	+	new EmptyTask());
947	+	break;
948	+	}
949	+	else if (t.status >= 0)
950	+	empty = false;
951	+	else if (s + 1 == top) {          // pop and throw away
952	+	if (U.compareAndSwapObject(a, j, t, null))
953	+	top = s;
954	+	break;
955	+	}
956	+	if (--n == 0) {
957	+	if (!empty && base == b)
958	+	stat = 0;
959	+	break;
960	+	}
961	+	}
962	+	}
963	+	if (removed)
964	+	task.doExec();
965	+	return stat;
966	+	}
967	+
968	+	/**
969		* Executes a top-level task and any local tasks remaining
970		* after execution.
1008	–	*
1009	–	* @return true unless terminating
971		*/
972	<	final boolean runTask(ForkJoinTask<?> t) {
1012	<	boolean alive = true;
972	>	final void runTask(ForkJoinTask<?> t) {
973		if (t != null) {
974		currentSteal = t;
975		t.doExec();
976	<	runLocalTasks();
976	>	if (top != base) {       // process remaining local tasks
977	>	if (mode == 0)
978	>	popAndExecAll();
979	>	else
980	>	pollAndExecAll();
981	>	}
982		++nsteals;
983		currentSteal = null;
984		}
1020	–	else if (runState < 0)            // terminating
1021	–	alive = false;
1022	–	return alive;
985		}
986
987		/**
988	<	* Executes a non-top-level (stolen) task
988	>	* Executes a non-top-level (stolen) task.
989		*/
990		final void runSubtask(ForkJoinTask<?> t) {
991		if (t != null) {
#	Line 1035 | Line 997 | public class ForkJoinPool extends Abstra
997		}
998
999		/**
1000	<	* Computes next value for random probes.  Scans don't require
1039	<	* a very high quality generator, but also not a crummy one.
1040	<	* Marsaglia xor-shift is cheap and works well enough.  Note:
1041	<	* This is manually inlined in several usages in ForkJoinPool
1042	<	* to avoid writes inside busy scan loops.
1000	>	* Returns true if owned and not known to be blocked.
1001		*/
1002	<	final int nextSeed() {
1003	<	int r = seed;
1004	<	r ^= r << 13;
1005	<	r ^= r >>> 17;
1006	<	r ^= r << 5;
1007	<	return seed = r;
1002	>	final boolean isApparentlyUnblocked() {
1003	>	Thread wt; Thread.State s;
1004	>	return (eventCount >= 0 &&
1005	>	(wt = owner) != null &&
1006	>	(s = wt.getState()) != Thread.State.BLOCKED &&
1007	>	s != Thread.State.WAITING &&
1008	>	s != Thread.State.TIMED_WAITING);
1009	>	}
1010	>
1011	>	/**
1012	>	* If this owned and is not already interrupted, try to
1013	>	* interrupt and/or unpark, ignoring exceptions.
1014	>	*/
1015	>	final void interruptOwner() {
1016	>	Thread wt, p;
1017	>	if ((wt = owner) != null && !wt.isInterrupted()) {
1018	>	try {
1019	>	wt.interrupt();
1020	>	} catch (SecurityException ignore) {
1021	>	}
1022	>	}
1023	>	if ((p = parker) != null)
1024	>	U.unpark(p);
1025		}
1026
1027		// Unsafe mechanics
#	Line 1072 | Line 1047 | public class ForkJoinPool extends Abstra
1047		ASHIFT = 31 - Integer.numberOfLeadingZeros(s);
1048		}
1049		}
1075	–
1050		/**
1051	<	* Class for artificial tasks that are used to replace the target
1052	<	* of local joins if they are removed from an interior queue slot
1053	<	* in WorkQueue.tryRemoveAndExec. We don't need the proxy to
1054	<	* actually do anything beyond having a unique identity.
1055	<	*/
1056	<	static final class EmptyTask extends ForkJoinTask<Void> {
1057	<	EmptyTask() { status = ForkJoinTask.NORMAL; } // force done
1058	<	public Void getRawResult() { return null; }
1059	<	public void setRawResult(Void x) {}
1060	<	public boolean exec() { return true; }
1061	<	}
1062	<
1063	<	/**
1064	<	<<<<<<< ForkJoinPool.java
1091	<	* Per-thread records for (typically non-FJ) threads that submit
1092	<	* to pools. Cureently holds only psuedo-random seed / index that
1093	<	* is used to chose submission queues in method doSubmit. In the
1094	<	* future, this may incorporate a means to implement different
1095	<	* task rejection and resubmission policies.
1051	>	* Per-thread records for threads that submit to pools. Currently
1052	>	* holds only pseudo-random seed / index that is used to choose
1053	>	* submission queues in method doSubmit. In the future, this may
1054	>	* also incorporate a means to implement different task rejection
1055	>	* and resubmission policies.
1056	>	*
1057	>	* Seeds for submitters and workers/workQueues work in basically
1058	>	* the same way but are initialized and updated using slightly
1059	>	* different mechanics. Both are initialized using the same
1060	>	* approach as in class ThreadLocal, where successive values are
1061	>	* unlikely to collide with previous values. This is done during
1062	>	* registration for workers, but requires a separate AtomicInteger
1063	>	* for submitters. Seeds are then randomly modified upon
1064	>	* collisions using xorshifts, which requires a non-zero seed.
1065		*/
1066		static final class Submitter {
1067	<	int seed; // seed for random submission queue selection
1099	<
1100	<	// Heuristic padding to ameliorate unfortunate memory placements
1101	<	int p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pa, pb, pc, pd, pe;
1102	<
1067	>	int seed;
1068		Submitter() {
1069	<	// Use identityHashCode, forced negative, for seed
1070	<	seed = System.identityHashCode(Thread.currentThread()) | (1 << 31);
1106	<	}
1107	<
1108	<	/**
1109	<	* Computes next value for random probes.  Like method
1110	<	* WorkQueue.nextSeed, this is manually inlined in several
1111	<	* usages to avoid writes inside busy loops.
1112	<	*/
1113	<	final int nextSeed() {
1114	<	int r = seed;
1115	<	r ^= r << 13;
1116	<	r ^= r >>> 17;
1117	<	return seed = r ^= r << 5;
1069	>	int s = nextSubmitterSeed.getAndAdd(SEED_INCREMENT);
1070	>	seed = (s == 0) ? 1 : s; // ensure non-zero
1071		}
1072		}
1073
#	Line 1123 | Line 1076 | public class ForkJoinPool extends Abstra
1076		public Submitter initialValue() { return new Submitter(); }
1077		}
1078
1079	+	// static fields (initialized in static initializer below)
1080	+
1081	+	/**
1082	+	* Creates a new ForkJoinWorkerThread. This factory is used unless
1083	+	* overridden in ForkJoinPool constructors.
1084	+	*/
1085	+	public static final ForkJoinWorkerThreadFactory
1086	+	defaultForkJoinWorkerThreadFactory;
1087	+
1088	+	/**
1089	+	* Generator for assigning sequence numbers as pool names.
1090	+	*/
1091	+	private static final AtomicInteger poolNumberGenerator;
1092	+
1093	+	/**
1094	+	* Generator for initial hashes/seeds for submitters. Accessed by
1095	+	* Submitter class constructor.
1096	+	*/
1097	+	static final AtomicInteger nextSubmitterSeed;
1098	+
1099	+	/**
1100	+	* Permission required for callers of methods that may start or
1101	+	* kill threads.
1102	+	*/
1103	+	private static final RuntimePermission modifyThreadPermission;
1104	+
1105		/**
1106		* Per-thread submission bookeeping. Shared across all pools
1107		* to reduce ThreadLocal pollution and because random motion
1108		* to avoid contention in one pool is likely to hold for others.
1109		*/
1110	<	static final ThreadSubmitter submitters = new ThreadSubmitter();
1110	>	private static final ThreadSubmitter submitters;
1111	>
1112	>	// static constants
1113	>
1114	>	/**
1115	>	* The wakeup interval (in nanoseconds) for a worker waiting for a
1116	>	* task when the pool is quiescent to instead try to shrink the
1117	>	* number of workers.  The exact value does not matter too
1118	>	* much. It must be short enough to release resources during
1119	>	* sustained periods of idleness, but not so short that threads
1120	>	* are continually re-created.
1121	>	*/
1122	>	private static final long SHRINK_RATE =
1123	>	4L * 1000L * 1000L * 1000L; // 4 seconds
1124
1125		/**
1126	<	* Top-level runloop for workers
1126	>	* The timeout value for attempted shrinkage, includes
1127	>	* some slop to cope with system timer imprecision.
1128		*/
1129	<	final void runWorker(ForkJoinWorkerThread wt) {
1137	<	// Initialize queue array and seed in this thread
1138	<	WorkQueue w = wt.workQueue;
1139	<	w.growArray(false);
1140	<	// Same initial hash as Submitters
1141	<	w.seed = System.identityHashCode(Thread.currentThread()) | (1 << 31);
1129	>	private static final long SHRINK_TIMEOUT = SHRINK_RATE - (SHRINK_RATE / 10);
1130
1131	<	do {} while (w.runTask(scan(w)));
1132	<	}
1131	>	/**
1132	>	* The maximum stolen->joining link depth allowed in method
1133	>	* tryHelpStealer.  Must be a power of two. This value also
1134	>	* controls the maximum number of times to try to help join a task
1135	>	* without any apparent progress or change in pool state before
1136	>	* giving up and blocking (see awaitJoin).  Depths for legitimate
1137	>	* chains are unbounded, but we use a fixed constant to avoid
1138	>	* (otherwise unchecked) cycles and to bound staleness of
1139	>	* traversal parameters at the expense of sometimes blocking when
1140	>	* we could be helping.
1141	>	*/
1142	>	private static final int MAX_HELP = 64;
1143	>
1144	>	/**
1145	>	* Secondary time-based bound (in nanosecs) for helping attempts
1146	>	* before trying compensated blocking in awaitJoin. Used in
1147	>	* conjunction with MAX_HELP to reduce variance due to different
1148	>	* polling rates associated with different helping options. The
1149	>	* value should roughly approximate the time required to create
1150	>	* and/or activate a worker thread.
1151	>	*/
1152	>	private static final long COMPENSATION_DELAY = 1L << 18; // ~0.25 millisec
1153
1154	<	// Creating, registering and deregistering workers
1154	>	/**
1155	>	* Increment for seed generators. See class ThreadLocal for
1156	>	* explanation.
1157	>	*/
1158	>	private static final int SEED_INCREMENT = 0x61c88647;
1159	>
1160	>	/**
1161	>	* Bits and masks for control variables
1162	>	*
1163	>	* Field ctl is a long packed with:
1164	>	* AC: Number of active running workers minus target parallelism (16 bits)
1165	>	* TC: Number of total workers minus target parallelism (16 bits)
1166	>	* ST: true if pool is terminating (1 bit)
1167	>	* EC: the wait count of top waiting thread (15 bits)
1168	>	* ID: poolIndex of top of Treiber stack of waiters (16 bits)
1169	>	*
1170	>	* When convenient, we can extract the upper 32 bits of counts and
1171	>	* the lower 32 bits of queue state, u = (int)(ctl >>> 32) and e =
1172	>	* (int)ctl.  The ec field is never accessed alone, but always
1173	>	* together with id and st. The offsets of counts by the target
1174	>	* parallelism and the positionings of fields makes it possible to
1175	>	* perform the most common checks via sign tests of fields: When
1176	>	* ac is negative, there are not enough active workers, when tc is
1177	>	* negative, there are not enough total workers, and when e is
1178	>	* negative, the pool is terminating.  To deal with these possibly
1179	>	* negative fields, we use casts in and out of "short" and/or
1180	>	* signed shifts to maintain signedness.
1181	>	*
1182	>	* When a thread is queued (inactivated), its eventCount field is
1183	>	* set negative, which is the only way to tell if a worker is
1184	>	* prevented from executing tasks, even though it must continue to
1185	>	* scan for them to avoid queuing races. Note however that
1186	>	* eventCount updates lag releases so usage requires care.
1187	>	*
1188	>	* Field runState is an int packed with:
1189	>	* SHUTDOWN: true if shutdown is enabled (1 bit)
1190	>	* SEQ:  a sequence number updated upon (de)registering workers (30 bits)
1191	>	* INIT: set true after workQueues array construction (1 bit)
1192	>	*
1193	>	* The sequence number enables simple consistency checks:
1194	>	* Staleness of read-only operations on the workQueues array can
1195	>	* be checked by comparing runState before vs after the reads.
1196	>	*/
1197	>
1198	>	// bit positions/shifts for fields
1199	>	private static final int  AC_SHIFT   = 48;
1200	>	private static final int  TC_SHIFT   = 32;
1201	>	private static final int  ST_SHIFT   = 31;
1202	>	private static final int  EC_SHIFT   = 16;
1203	>
1204	>	// bounds
1205	>	private static final int  SMASK      = 0xffff;  // short bits
1206	>	private static final int  MAX_CAP    = 0x7fff;  // max #workers - 1
1207	>	private static final int  SQMASK     = 0xfffe;  // even short bits
1208	>	private static final int  SHORT_SIGN = 1 << 15;
1209	>	private static final int  INT_SIGN   = 1 << 31;
1210	>
1211	>	// masks
1212	>	private static final long STOP_BIT   = 0x0001L << ST_SHIFT;
1213	>	private static final long AC_MASK    = ((long)SMASK) << AC_SHIFT;
1214	>	private static final long TC_MASK    = ((long)SMASK) << TC_SHIFT;
1215	>
1216	>	// units for incrementing and decrementing
1217	>	private static final long TC_UNIT    = 1L << TC_SHIFT;
1218	>	private static final long AC_UNIT    = 1L << AC_SHIFT;
1219	>
1220	>	// masks and units for dealing with u = (int)(ctl >>> 32)
1221	>	private static final int  UAC_SHIFT  = AC_SHIFT - 32;
1222	>	private static final int  UTC_SHIFT  = TC_SHIFT - 32;
1223	>	private static final int  UAC_MASK   = SMASK << UAC_SHIFT;
1224	>	private static final int  UTC_MASK   = SMASK << UTC_SHIFT;
1225	>	private static final int  UAC_UNIT   = 1 << UAC_SHIFT;
1226	>	private static final int  UTC_UNIT   = 1 << UTC_SHIFT;
1227	>
1228	>	// masks and units for dealing with e = (int)ctl
1229	>	private static final int E_MASK      = 0x7fffffff; // no STOP_BIT
1230	>	private static final int E_SEQ       = 1 << EC_SHIFT;
1231	>
1232	>	// runState bits
1233	>	private static final int SHUTDOWN    = 1 << 31;
1234	>
1235	>	// access mode for WorkQueue
1236	>	static final int LIFO_QUEUE          =  0;
1237	>	static final int FIFO_QUEUE          =  1;
1238	>	static final int SHARED_QUEUE        = -1;
1239	>
1240	>	// Instance fields
1241	>
1242	>	/*
1243	>	* Field layout order in this class tends to matter more than one
1244	>	* would like. Runtime layout order is only loosely related to
1245	>	* declaration order and may differ across JVMs, but the following
1246	>	* empirically works OK on current JVMs.
1247	>	*/
1248	>
1249	>	volatile long ctl;                         // main pool control
1250	>	final int parallelism;                     // parallelism level
1251	>	final int localMode;                       // per-worker scheduling mode
1252	>	final int submitMask;                      // submit queue index bound
1253	>	int nextSeed;                              // for initializing worker seeds
1254	>	volatile int runState;                     // shutdown status and seq
1255	>	WorkQueue[] workQueues;                    // main registry
1256	>	final Mutex lock;                          // for registration
1257	>	final Condition termination;               // for awaitTermination
1258	>	final ForkJoinWorkerThreadFactory factory; // factory for new workers
1259	>	final Thread.UncaughtExceptionHandler ueh; // per-worker UEH
1260	>	final AtomicLong stealCount;               // collect counts when terminated
1261	>	final AtomicInteger nextWorkerNumber;      // to create worker name string
1262	>	final String workerNamePrefix;             // to create worker name string
1263	>
1264	>	//  Creating, registering, and deregistering workers
1265
1266		/**
1267		* Tries to create and start a worker
1268		*/
1269		private void addWorker() {
1270		Throwable ex = null;
1271	<	ForkJoinWorkerThread w = null;
1271	>	ForkJoinWorkerThread wt = null;
1272		try {
1273	<	if ((w = factory.newThread(this)) != null) {
1274	<	w.start();
1273	>	if ((wt = factory.newThread(this)) != null) {
1274	>	wt.start();
1275		return;
1276		}
1277		} catch (Throwable e) {
1278		ex = e;
1279		}
1280	<	deregisterWorker(w, ex);
1280	>	deregisterWorker(wt, ex); // adjust counts etc on failure
1281		}
1282
1283		/**
#	Line 1174 | Line 1292 | public class ForkJoinPool extends Abstra
1292		}
1293
1294		/**
1295	<	* Callback from ForkJoinWorkerThread constructor to establish and
1296	<	* record its WorkQueue
1295	>	* Callback from ForkJoinWorkerThread constructor to establish its
1296	>	* poolIndex and record its WorkQueue. To avoid scanning bias due
1297	>	* to packing entries in front of the workQueues array, we treat
1298	>	* the array as a simple power-of-two hash table using per-thread
1299	>	* seed as hash, expanding as needed.
1300		*
1301	<	* @param wt the worker thread
1301	>	* @param w the worker's queue
1302		*/
1303	<	final void registerWorker(ForkJoinWorkerThread wt) {
1304	<	WorkQueue w = wt.workQueue;
1305	<	ReentrantLock lock = this.lock;
1303	>
1304	>	final void registerWorker(WorkQueue w) {
1305	>	Mutex lock = this.lock;
1306		lock.lock();
1307		try {
1187	–	int k = nextPoolIndex;
1308		WorkQueue[] ws = workQueues;
1309	<	if (ws != null) {                       // ignore on shutdown
1310	<	int n = ws.length;
1311	<	if (k < 0 || (k & 1) == 0 || k >= n || ws[k] != null) {
1312	<	for (k = 1; k < n && ws[k] != null; k += 2)
1313	<	;                           // workers are at odd indices
1314	<	if (k >= n)                     // resize
1315	<	workQueues = ws = Arrays.copyOf(ws, n << 1);
1316	<	}
1317	<	w.poolIndex = k;
1318	<	w.eventCount = ~(k >>> 1) & SMASK;  // Set up wait count
1319	<	ws[k] = w;                          // record worker
1320	<	nextPoolIndex = k + 2;
1321	<	int rs = runState;
1322	<	int m = rs & SMASK;                 // recalculate runState mask
1323	<	if (k > m)
1324	<	m = (m << 1) + 1;
1325	<	runState = (rs & SHUTDOWN) | ((rs + RS_SEQ) & RS_SEQ_MASK) | m;
1309	>	if (w != null && ws != null) {          // skip on shutdown/failure
1310	>	int rs, n =  ws.length, m = n - 1;
1311	>	int s = nextSeed += SEED_INCREMENT; // rarely-colliding sequence
1312	>	w.seed = (s == 0) ? 1 : s;          // ensure non-zero seed
1313	>	int r = (s << 1) | 1;               // use odd-numbered indices
1314	>	if (ws[r &= m] != null) {           // collision
1315	>	int probes = 0;                 // step by approx half size
1316	>	int step = (n <= 4) ? 2 : ((n >>> 1) & SQMASK) + 2;
1317	>	while (ws[r = (r + step) & m] != null) {
1318	>	if (++probes >= n) {
1319	>	workQueues = ws = Arrays.copyOf(ws, n <<= 1);
1320	>	m = n - 1;
1321	>	probes = 0;
1322	>	}
1323	>	}
1324	>	}
1325	>	w.eventCount = w.poolIndex = r;     // establish before recording
1326	>	ws[r] = w;                          // also update seq
1327	>	runState = ((rs = runState) & SHUTDOWN) | ((rs + 2) & ~SHUTDOWN);
1328		}
1329		} finally {
1330		lock.unlock();
#	Line 1210 | Line 1332 | public class ForkJoinPool extends Abstra
1332		}
1333
1334		/**
1335	<	* Final callback from terminating worker, as well as failure to
1336	<	* construct or start a worker in addWorker.  Removes record of
1335	>	* Final callback from terminating worker, as well as upon failure
1336	>	* to construct or start a worker in addWorker.  Removes record of
1337		* worker from array, and adjusts counts. If pool is shutting
1338		* down, tries to complete termination.
1339		*
#	Line 1219 | Line 1341 | public class ForkJoinPool extends Abstra
1341		* @param ex the exception causing failure, or null if none
1342		*/
1343		final void deregisterWorker(ForkJoinWorkerThread wt, Throwable ex) {
1344	+	Mutex lock = this.lock;
1345		WorkQueue w = null;
1346		if (wt != null && (w = wt.workQueue) != null) {
1347		w.runState = -1;                // ensure runState is set
1348		stealCount.getAndAdd(w.totalSteals + w.nsteals);
1349		int idx = w.poolIndex;
1227	–	ReentrantLock lock = this.lock;
1350		lock.lock();
1351		try {                           // remove record from array
1352		WorkQueue[] ws = workQueues;
1353		if (ws != null && idx >= 0 && idx < ws.length && ws[idx] == w)
1354	<	ws[nextPoolIndex = idx] = null;
1354	>	ws[idx] = null;
1355		} finally {
1356		lock.unlock();
1357		}
#	Line 1241 | Line 1363 | public class ForkJoinPool extends Abstra
1363		((c - TC_UNIT) & TC_MASK) |
1364		(c & ~(AC_MASK|TC_MASK)))));
1365
1366	<	if (!tryTerminate(false) && w != null) {
1366	>	if (!tryTerminate(false, false) && w != null) {
1367		w.cancelAll();                  // cancel remaining tasks
1368		if (w.array != null)            // suppress signal if never ran
1369		signalWork();               // wake up or create replacement
1370	+	if (ex == null)                 // help clean refs on way out
1371	+	ForkJoinTask.helpExpungeStaleExceptions();
1372		}
1373
1374		if (ex != null)                     // rethrow
1375		U.throwException(ex);
1376		}
1377
1378	+
1379	+	// Submissions
1380	+
1381		/**
1382	<	* Tries to add and register a new queue at the given index.
1382	>	* Unless shutting down, adds the given task to a submission queue
1383	>	* at submitter's current queue index (modulo submission
1384	>	* range). If no queue exists at the index, one is created.  If
1385	>	* the queue is busy, another index is randomly chosen. The
1386	>	* submitMask bounds the effective number of queues to the
1387	>	* (nearest power of two for) parallelism level.
1388		*
1389	<	* @param idx the workQueues array index to register the queue
1390	<	* @return the queue, or null if could not add because could
1391	<	* not acquire lock or idx is unusable
1392	<	*/
1393	<	private WorkQueue tryAddSharedQueue(int idx) {
1394	<	WorkQueue q = null;
1395	<	ReentrantLock lock = this.lock;
1396	<	if (idx >= 0 && (idx & 1) == 0 && !lock.isLocked()) {
1397	<	// create queue outside of lock but only if apparently free
1398	<	WorkQueue nq = new WorkQueue(null, SHARED_QUEUE);
1399	<	if (lock.tryLock()) {
1400	<	try {
1401	<	WorkQueue[] ws = workQueues;
1402	<	if (ws != null && idx < ws.length) {
1403	<	if ((q = ws[idx]) == null) {
1404	<	int rs;         // update runState seq
1405	<	ws[idx] = q = nq;
1406	<	runState = (((rs = runState) & SHUTDOWN) |
1275	<	((rs + RS_SEQ) & ~SHUTDOWN));
1276	<	}
1389	>	* @param task the task. Caller must ensure non-null.
1390	>	*/
1391	>	private void doSubmit(ForkJoinTask<?> task) {
1392	>	Submitter s = submitters.get();
1393	>	for (int r = s.seed, m = submitMask;;) {
1394	>	WorkQueue[] ws; WorkQueue q;
1395	>	int k = r & m & SQMASK;          // use only even indices
1396	>	if (runState < 0 || (ws = workQueues) == null || ws.length <= k)
1397	>	throw new RejectedExecutionException(); // shutting down
1398	>	else if ((q = ws[k]) == null) {  // create new queue
1399	>	WorkQueue nq = new WorkQueue(this, null, SHARED_QUEUE);
1400	>	Mutex lock = this.lock;      // construct outside lock
1401	>	lock.lock();
1402	>	try {                        // recheck under lock
1403	>	int rs = runState;       // to update seq
1404	>	if (ws == workQueues && ws[k] == null) {
1405	>	ws[k] = nq;
1406	>	runState = ((rs & SHUTDOWN) | ((rs + 2) & ~SHUTDOWN));
1407		}
1408		} finally {
1409		lock.unlock();
1410		}
1411		}
1412	+	else if (q.trySharedPush(task)) {
1413	+	signalWork();
1414	+	return;
1415	+	}
1416	+	else if (m > 1) {                // move to a different index
1417	+	r ^= r << 13;                // same xorshift as WorkQueues
1418	+	r ^= r >>> 17;
1419	+	s.seed = r ^= r << 5;
1420	+	}
1421	+	else
1422	+	Thread.yield();              // yield if no alternatives
1423		}
1283	–	return q;
1424		}
1425
1426		// Maintaining ctl counts
1427
1428		/**
1429	<	* Increments active count; mainly called upon return from blocking
1429	>	* Increments active count; mainly called upon return from blocking.
1430		*/
1431		final void incrementActiveCount() {
1432		long c;
#	Line 1294 | Line 1434 | public class ForkJoinPool extends Abstra
1434		}
1435
1436		/**
1437	<	* Activates or creates a worker
1437	>	* Tries to activate or create a worker if too few are active.
1438		*/
1439		final void signalWork() {
1440	<	/*
1441	<	* The while condition is true if: (there is are too few total
1442	<	* workers OR there is at least one waiter) AND (there are too
1443	<	* few active workers OR the pool is terminating).  The value
1444	<	* of e distinguishes the remaining cases: zero (no waiters)
1445	<	* for create, negative if terminating (in which case do
1446	<	* nothing), else release a waiter. The secondary checks for
1447	<	* release (non-null array etc) can fail if the pool begins
1448	<	* terminating after the test, and don't impose any added cost
1449	<	* because JVMs must perform null and bounds checks anyway.
1450	<	*/
1451	<	long c; int e, u;
1452	<	while ((((e = (int)(c = ctl)) | (u = (int)(c >>> 32))) &
1453	<	(INT_SIGN|SHORT_SIGN)) == (INT_SIGN|SHORT_SIGN)) {
1314	<	WorkQueue[] ws = workQueues; int i; WorkQueue w; Thread p;
1315	<	if (e == 0) {                    // add a new worker
1316	<	if (U.compareAndSwapLong
1317	<	(this, CTL, c, (long)(((u + UTC_UNIT) & UTC_MASK) |
1318	<	((u + UAC_UNIT) & UAC_MASK)) << 32)) {
1319	<	addWorker();
1320	<	break;
1440	>	long c; int u;
1441	>	while ((u = (int)((c = ctl) >>> 32)) < 0) {     // too few active
1442	>	WorkQueue[] ws = workQueues; int e, i; WorkQueue w; Thread p;
1443	>	if ((e = (int)c) > 0) {                     // at least one waiting
1444	>	if (ws != null && (i = e & SMASK) < ws.length &&
1445	>	(w = ws[i]) != null && w.eventCount == (e | INT_SIGN)) {
1446	>	long nc = (((long)(w.nextWait & E_MASK)) |
1447	>	((long)(u + UAC_UNIT) << 32));
1448	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1449	>	w.eventCount = (e + E_SEQ) & E_MASK;
1450	>	if ((p = w.parker) != null)
1451	>	U.unpark(p);                // activate and release
1452	>	break;
1453	>	}
1454		}
1455	<	}
1323	<	else if (e > 0 && ws != null &&
1324	<	(i = ((~e << 1) | 1) & SMASK) < ws.length &&
1325	<	(w = ws[i]) != null &&
1326	<	w.eventCount == (e | INT_SIGN)) {
1327	<	if (U.compareAndSwapLong
1328	<	(this, CTL, c, (((long)(w.nextWait & E_MASK)) |
1329	<	((long)(u + UAC_UNIT) << 32)))) {
1330	<	w.eventCount = (e + E_SEQ) & E_MASK;
1331	<	if ((p = w.parker) != null)
1332	<	U.unpark(p);         // release a waiting worker
1455	>	else
1456		break;
1334	–	}
1457		}
1458	<	else
1459	<	break;
1460	<	}
1339	<	}
1340	<
1341	<	/**
1342	<	* Tries to decrement active count (sometimes implicitly) and
1343	<	* possibly release or create a compensating worker in preparation
1344	<	* for blocking. Fails on contention or termination.
1345	<	*
1346	<	* @return true if the caller can block, else should recheck and retry
1347	<	*/
1348	<	final boolean tryCompensate() {
1349	<	WorkQueue[] ws; WorkQueue w; Thread p;
1350	<	int pc = parallelism, e, u, ac, tc, i;
1351	<	long c = ctl;
1352	<
1353	<	if ((e = (int)c) >= 0) {
1354	<	if ((ac = ((u = (int)(c >>> 32)) >> UAC_SHIFT)) <= 0 &&
1355	<	e != 0 && (ws = workQueues) != null &&
1356	<	(i = ((~e << 1) | 1) & SMASK) < ws.length &&
1357	<	(w = ws[i]) != null) {
1358	<	if (w.eventCount == (e | INT_SIGN) &&
1359	<	U.compareAndSwapLong
1360	<	(this, CTL, c, ((long)(w.nextWait & E_MASK) |
1361	<	(c & (AC_MASK|TC_MASK))))) {
1362	<	w.eventCount = (e + E_SEQ) & E_MASK;
1363	<	if ((p = w.parker) != null)
1364	<	U.unpark(p);
1365	<	return true;             // release an idle worker
1366	<	}
1367	<	}
1368	<	else if ((tc = (short)(u >>> UTC_SHIFT)) >= 0 && ac + pc > 1) {
1369	<	long nc = ((c - AC_UNIT) & AC_MASK) | (c & ~AC_MASK);
1370	<	if (U.compareAndSwapLong(this, CTL, c, nc))
1371	<	return true;             // no compensation needed
1372	<	}
1373	<	else if (tc + pc < MAX_ID) {
1374	<	long nc = ((c + TC_UNIT) & TC_MASK) | (c & ~TC_MASK);
1458	>	else if (e == 0 && (u & SHORT_SIGN) != 0) { // too few total
1459	>	long nc = (long)(((u + UTC_UNIT) & UTC_MASK) |
1460	>	((u + UAC_UNIT) & UAC_MASK)) << 32;
1461		if (U.compareAndSwapLong(this, CTL, c, nc)) {
1462		addWorker();
1463	<	return true;             // create replacement
1463	>	break;
1464		}
1465		}
1466	+	else
1467	+	break;
1468		}
1381	–	return false;
1469		}
1470
1471	<	// Submissions
1471	>	// Scanning for tasks
1472
1473		/**
1474	<	* Unless shutting down, adds the given task to a submission queue
1388	<	* at submitter's current queue index. If no queue exists at the
1389	<	* index, one is created unless pool lock is busy.  If the queue
1390	<	* and/or lock are busy, another index is randomly chosen.
1474	>	* Top-level runloop for workers, called by ForkJoinWorkerThread.run.
1475		*/
1476	<	private void doSubmit(ForkJoinTask<?> task) {
1477	<	if (task == null)
1478	<	throw new NullPointerException();
1395	<	Submitter s = submitters.get();
1396	<	for (int r = s.seed;;) {
1397	<	WorkQueue q; int k;
1398	<	int rs = runState, m = rs & SMASK;
1399	<	WorkQueue[] ws = workQueues;
1400	<	if (rs < 0 || ws == null)   // shutting down
1401	<	throw new RejectedExecutionException();
1402	<	if (ws.length > m &&        // k must be at index
1403	<	((q = ws[k = (r << 1) & m]) != null ||
1404	<	(q = tryAddSharedQueue(k)) != null) &&
1405	<	q.trySharedPush(task)) {
1406	<	signalWork();
1407	<	return;
1408	<	}
1409	<	r ^= r << 13;               // xorshift seed to new position
1410	<	r ^= r >>> 17;
1411	<	if (((s.seed = r ^= r << 5) & m) == 0)
1412	<	Thread.yield();         // occasionally yield if busy
1413	<	}
1476	>	final void runWorker(WorkQueue w) {
1477	>	w.growArray(false);         // initialize queue array in this thread
1478	>	do { w.runTask(scan(w)); } while (w.runState >= 0);
1479		}
1480
1416	–
1417	–	// Scanning for tasks
1418	–
1481		/**
1482		* Scans for and, if found, returns one task, else possibly
1483		* inactivates the worker. This method operates on single reads of
1484	<	* volatile state and is designed to be re-invoked continuously in
1485	<	* part because it returns upon detecting inconsistencies,
1484	>	* volatile state and is designed to be re-invoked continuously,
1485	>	* in part because it returns upon detecting inconsistencies,
1486		* contention, or state changes that indicate possible success on
1487		* re-invocation.
1488		*
1489	<	* The scan searches for tasks across queues, randomly selecting
1490	<	* the first #queues probes, favoring steals 2:1 over submissions
1491	<	* (by exploiting even/odd indexing), and then performing a
1492	<	* circular sweep of all queues.  The scan terminates upon either
1493	<	* finding a non-empty queue, or completing a full sweep. If the
1494	<	* worker is not inactivated, it takes and returns a task from
1495	<	* this queue.  On failure to find a task, we take one of the
1496	<	* following actions, after which the caller will retry calling
1497	<	* this method unless terminated.
1489	>	* The scan searches for tasks across a random permutation of
1490	>	* queues (starting at a random index and stepping by a random
1491	>	* relative prime, checking each at least once).  The scan
1492	>	* terminates upon either finding a non-empty queue, or completing
1493	>	* the sweep. If the worker is not inactivated, it takes and
1494	>	* returns a task from this queue.  On failure to find a task, we
1495	>	* take one of the following actions, after which the caller will
1496	>	* retry calling this method unless terminated.
1497	>	*
1498	>	* * If pool is terminating, terminate the worker.
1499		*
1500		* * If not a complete sweep, try to release a waiting worker.  If
1501		* the scan terminated because the worker is inactivated, then the
#	Line 1441 | Line 1504 | public class ForkJoinPool extends Abstra
1504		* another worker, but with same net effect. Releasing in other
1505		* cases as well ensures that we have enough workers running.
1506		*
1444	–	* * If the caller has run a task since the the last empty scan,
1445	–	* return (to allow rescan) if other workers are not also yet
1446	–	* enqueued.  Field WorkQueue.rescans counts down on each scan to
1447	–	* ensure eventual inactivation, and occasional calls to
1448	–	* Thread.yield to help avoid interference with more useful
1449	–	* activities on the system.
1450	–	*
1451	–	* * If pool is terminating, terminate the worker
1452	–	*
1507		* * If not already enqueued, try to inactivate and enqueue the
1508	<	* worker on wait queue.
1508	>	* worker on wait queue. Or, if inactivating has caused the pool
1509	>	* to be quiescent, relay to idleAwaitWork to check for
1510	>	* termination and possibly shrink pool.
1511	>	*
1512	>	* * If already inactive, and the caller has run a task since the
1513	>	* last empty scan, return (to allow rescan) unless others are
1514	>	* also inactivated.  Field WorkQueue.rescans counts down on each
1515	>	* scan to ensure eventual inactivation and blocking.
1516		*
1517	<	* * If already enqueued and none of the above apply, either park
1518	<	* awaiting signal, or if this is the most recent waiter and pool
1458	<	* is quiescent, relay to idleAwaitWork to check for termination
1459	<	* and possibly shrink pool.
1517	>	* * If already enqueued and none of the above apply, park
1518	>	* awaiting signal,
1519		*
1520		* @param w the worker (via its WorkQueue)
1521		* @return a task or null of none found
1522		*/
1523		private final ForkJoinTask<?> scan(WorkQueue w) {
1524	<	boolean swept = false;                 // true after full empty scan
1525	<	WorkQueue[] ws;                        // volatile read order matters
1526	<	int r = w.seed, ec = w.eventCount;     // ec is negative if inactive
1527	<	int rs = runState, m = rs & SMASK;
1528	<	if ((ws = workQueues) != null && ws.length > m) {
1529	<	ForkJoinTask<?> task = null;
1530	<	for (int k = 0, j = -2 - m; ; ++j) {
1531	<	WorkQueue q; int b;
1532	<	if (j < 0) {                    // random probes while j negative
1533	<	r ^= r << 13; r ^= r >>> 17; k = (r ^= r << 5) | (j & 1);
1534	<	}                               // worker (not submit) for odd j
1535	<	else                            // cyclic scan when j >= 0
1536	<	k += (m >>> 1) | 1;         // step by half to reduce bias
1537	<
1538	<	if ((q = ws[k & m]) != null && (b = q.base) - q.top < 0) {
1539	<	if (ec >= 0)
1540	<	task = q.pollAt(b);     // steal
1541	<	break;
1524	>	WorkQueue[] ws;                       // first update random seed
1525	>	int r = w.seed; r ^= r << 13; r ^= r >>> 17; w.seed = r ^= r << 5;
1526	>	int rs = runState, m;                 // volatile read order matters
1527	>	if ((ws = workQueues) != null && (m = ws.length - 1) > 0) {
1528	>	int ec = w.eventCount;            // ec is negative if inactive
1529	>	int step = (r >>> 16) | 1;        // relative prime
1530	>	for (int j = (m + 1) << 2; ; r += step) {
1531	>	WorkQueue q; ForkJoinTask<?> t; ForkJoinTask<?>[] a; int b;
1532	>	if ((q = ws[r & m]) != null && (b = q.base) - q.top < 0 &&
1533	>	(a = q.array) != null) {  // probably nonempty
1534	>	int i = (((a.length - 1) & b) << ASHIFT) + ABASE;
1535	>	t = (ForkJoinTask<?>)U.getObjectVolatile(a, i);
1536	>	if (q.base == b && ec >= 0 && t != null &&
1537	>	U.compareAndSwapObject(a, i, t, null)) {
1538	>	q.base = b + 1;       // specialization of pollAt
1539	>	return t;
1540	>	}
1541	>	else if (ec < 0 || j <= m) {
1542	>	rs = 0;               // mark scan as imcomplete
1543	>	break;                // caller can retry after release
1544	>	}
1545		}
1546	<	else if (j > m) {
1485	<	if (rs == runState)        // staleness check
1486	<	swept = true;
1546	>	if (--j < 0)
1547		break;
1548	+	}
1549	+
1550	+	long c = ctl; int e = (int)c, a = (int)(c >> AC_SHIFT), nr, ns;
1551	+	if (e < 0)                        // decode ctl on empty scan
1552	+	w.runState = -1;              // pool is terminating
1553	+	else if (rs == 0 || rs != runState) { // incomplete scan
1554	+	WorkQueue v; Thread p;        // try to release a waiter
1555	+	if (e > 0 && a < 0 && w.eventCount == ec &&
1556	+	(v = ws[e & m]) != null && v.eventCount == (e | INT_SIGN)) {
1557	+	long nc = ((long)(v.nextWait & E_MASK) |
1558	+	((c + AC_UNIT) & (AC_MASK|TC_MASK)));
1559	+	if (ctl == c && U.compareAndSwapLong(this, CTL, c, nc)) {
1560	+	v.eventCount = (e + E_SEQ) & E_MASK;
1561	+	if ((p = v.parker) != null)
1562	+	U.unpark(p);
1563	+	}
1564		}
1565		}
1566	<	w.seed = r;                        // save seed for next scan
1567	<	if (task != null)
1568	<	return task;
1569	<	}
1570	<
1571	<	// Decode ctl on empty scan
1572	<	long c = ctl; int e = (int)c, a = (int)(c >> AC_SHIFT), nr, ns;
1573	<	if (!swept) {                          // try to release a waiter
1574	<	WorkQueue v; Thread p;
1575	<	if (e > 0 && a < 0 && ws != null &&
1576	<	(v = ws[((~e << 1) | 1) & m]) != null &&
1577	<	v.eventCount == (e | INT_SIGN) && U.compareAndSwapLong
1578	<	(this, CTL, c, ((long)(v.nextWait & E_MASK) |
1579	<	((c + AC_UNIT) & (AC_MASK|TC_MASK))))) {
1580	<	v.eventCount = (e + E_SEQ) & E_MASK;
1581	<	if ((p = v.parker) != null)
1582	<	U.unpark(p);
1583	<	}
1584	<	}
1585	<	else if ((nr = w.rescans) > 0) {       // continue rescanning
1586	<	int ac = a + parallelism;
1587	<	if ((w.rescans = (ac < nr) ? ac : nr - 1) > 0 && w.seed < 0 &&
1588	<	w.eventCount == ec)
1589	<	Thread.yield();                // 1 bit randomness for yield call
1590	<	}
1591	<	else if (e < 0)                        // pool is terminating
1592	<	w.runState = -1;
1593	<	else if (ec >= 0) {                    // try to enqueue
1594	<	long nc = (long)ec | ((c - AC_UNIT) & (AC_MASK|TC_MASK));
1595	<	w.nextWait = e;
1596	<	w.eventCount = ec | INT_SIGN;      // mark as inactive
1597	<	if (!U.compareAndSwapLong(this, CTL, c, nc))
1522	<	w.eventCount = ec;             // back out on CAS failure
1523	<	else if ((ns = w.nsteals) != 0) {  // set rescans if ran task
1524	<	if (a <= 0)                    // ... unless too many active
1525	<	w.rescans = a + parallelism;
1526	<	w.nsteals = 0;
1527	<	w.totalSteals += ns;
1528	<	}
1529	<	}
1530	<	else{                                  // already queued
1531	<	if (parallelism == -a)
1532	<	idleAwaitWork(w);              // quiescent
1533	<	if (w.eventCount == ec) {
1534	<	Thread.interrupted();          // clear status
1535	<	ForkJoinWorkerThread wt = w.owner;
1536	<	U.putObject(wt, PARKBLOCKER, this);
1537	<	w.parker = wt;                 // emulate LockSupport.park
1538	<	if (w.eventCount == ec)        // recheck
1539	<	U.park(false, 0L);         // block
1540	<	w.parker = null;
1541	<	U.putObject(wt, PARKBLOCKER, null);
1566	>	else if (ec >= 0) {               // try to enqueue/inactivate
1567	>	long nc = (long)ec | ((c - AC_UNIT) & (AC_MASK|TC_MASK));
1568	>	w.nextWait = e;
1569	>	w.eventCount = ec | INT_SIGN; // mark as inactive
1570	>	if (ctl != c || !U.compareAndSwapLong(this, CTL, c, nc))
1571	>	w.eventCount = ec;        // unmark on CAS failure
1572	>	else {
1573	>	if ((ns = w.nsteals) != 0) {
1574	>	w.nsteals = 0;        // set rescans if ran task
1575	>	w.rescans = (a > 0) ? 0 : a + parallelism;
1576	>	w.totalSteals += ns;
1577	>	}
1578	>	if (a == 1 - parallelism) // quiescent
1579	>	idleAwaitWork(w, nc, c);
1580	>	}
1581	>	}
1582	>	else if (w.eventCount < 0) {      // already queued
1583	>	if ((nr = w.rescans) > 0) {   // continue rescanning
1584	>	int ac = a + parallelism;
1585	>	if (((w.rescans = (ac < nr) ? ac : nr - 1) & 3) == 0)
1586	>	Thread.yield();       // yield before block
1587	>	}
1588	>	else {
1589	>	Thread.interrupted();     // clear status
1590	>	Thread wt = Thread.currentThread();
1591	>	U.putObject(wt, PARKBLOCKER, this);
1592	>	w.parker = wt;            // emulate LockSupport.park
1593	>	if (w.eventCount < 0)     // recheck
1594	>	U.park(false, 0L);
1595	>	w.parker = null;
1596	>	U.putObject(wt, PARKBLOCKER, null);
1597	>	}
1598		}
1599		}
1600		return null;
1601		}
1602
1603		/**
1604	<	* If inactivating worker w has caused pool to become quiescent,
1605	<	* check for pool termination, and, so long as this is not the
1606	<	* only worker, wait for event for up to SHRINK_RATE nanosecs On
1607	<	* timeout, if ctl has not changed, terminate the worker, which
1608	<	* will in turn wake up another worker to possibly repeat this
1609	<	* process.
1604	>	* If inactivating worker w has caused the pool to become
1605	>	* quiescent, checks for pool termination, and, so long as this is
1606	>	* not the only worker, waits for event for up to SHRINK_RATE
1607	>	* nanosecs.  On timeout, if ctl has not changed, terminates the
1608	>	* worker, which will in turn wake up another worker to possibly
1609	>	* repeat this process.
1610		*
1611		* @param w the calling worker
1612	+	* @param currentCtl the ctl value triggering possible quiescence
1613	+	* @param prevCtl the ctl value to restore if thread is terminated
1614		*/
1615	<	private void idleAwaitWork(WorkQueue w) {
1616	<	long c; int nw, ec;
1617	<	if (!tryTerminate(false) &&
1618	<	(int)((c = ctl) >> AC_SHIFT) + parallelism == 0 &&
1619	<	(ec = w.eventCount) == ((int)c | INT_SIGN) &&
1620	<	(nw = w.nextWait) != 0) {
1563	<	long nc = ((long)(nw & E_MASK) | // ctl to restore on timeout
1564	<	((c + AC_UNIT) & AC_MASK) | (c & TC_MASK));
1565	<	ForkJoinTask.helpExpungeStaleExceptions(); // help clean
1566	<	ForkJoinWorkerThread wt = w.owner;
1567	<	while (ctl == c) {
1615	>	private void idleAwaitWork(WorkQueue w, long currentCtl, long prevCtl) {
1616	>	if (w.eventCount < 0 && !tryTerminate(false, false) &&
1617	>	(int)prevCtl != 0 && !hasQueuedSubmissions() && ctl == currentCtl) {
1618	>	Thread wt = Thread.currentThread();
1619	>	Thread.yield();            // yield before block
1620	>	while (ctl == currentCtl) {
1621		long startTime = System.nanoTime();
1622		Thread.interrupted();  // timed variant of version in scan()
1623		U.putObject(wt, PARKBLOCKER, this);
1624		w.parker = wt;
1625	<	if (ctl == c)
1625	>	if (ctl == currentCtl)
1626		U.park(false, SHRINK_RATE);
1627		w.parker = null;
1628		U.putObject(wt, PARKBLOCKER, null);
1629	<	if (ctl != c)
1629	>	if (ctl != currentCtl)
1630		break;
1631		if (System.nanoTime() - startTime >= SHRINK_TIMEOUT &&
1632	<	U.compareAndSwapLong(this, CTL, c, nc)) {
1633	<	w.runState = -1;          // shrink
1634	<	w.eventCount = (ec + E_SEQ) | E_MASK;
1632	>	U.compareAndSwapLong(this, CTL, currentCtl, prevCtl)) {
1633	>	w.eventCount = (w.eventCount + E_SEQ) | E_MASK;
1634	>	w.runState = -1;   // shrink
1635		break;
1636		}
1637		}
#	Line 1596 | Line 1649 | public class ForkJoinPool extends Abstra
1649		* leaves hints in workers to speed up subsequent calls. The
1650		* implementation is very branchy to cope with potential
1651		* inconsistencies or loops encountering chains that are stale,
1652	<	* unknown, or of length greater than MAX_HELP_DEPTH links.  All
1600	<	* of these cases are dealt with by just retrying by caller.
1652	>	* unknown, or so long that they are likely cyclic.
1653		*
1654		* @param joiner the joining worker
1655		* @param task the task to join
1656	<	* @return true if found or ran a task (and so is immediately retryable)
1656	>	* @return 0 if no progress can be made, negative if task
1657	>	* known complete, else positive
1658		*/
1659	<	final boolean tryHelpStealer(WorkQueue joiner, ForkJoinTask<?> task) {
1660	<	ForkJoinTask<?> subtask;    // current target
1661	<	boolean progress = false;
1662	<	int depth = 0;              // current chain depth
1663	<	int m = runState & SMASK;
1664	<	WorkQueue[] ws = workQueues;
1665	<
1666	<	if (ws != null && ws.length > m && (subtask = task).status >= 0) {
1667	<	outer:for (WorkQueue j = joiner;;) {
1668	<	// Try to find the stealer of subtask, by first using hint
1616	<	WorkQueue stealer = null;
1617	<	WorkQueue v = ws[j.stealHint & m];
1618	<	if (v != null && v.currentSteal == subtask)
1619	<	stealer = v;
1620	<	else {
1621	<	for (int i = 1; i <= m; i += 2) {
1622	<	if ((v = ws[i]) != null && v.currentSteal == subtask) {
1623	<	stealer = v;
1624	<	j.stealHint = i; // save hint
1625	<	break;
1626	<	}
1659	>	private int tryHelpStealer(WorkQueue joiner, ForkJoinTask<?> task) {
1660	>	int stat = 0, steps = 0;                    // bound to avoid cycles
1661	>	if (joiner != null && task != null) {       // hoist null checks
1662	>	restart: for (;;) {
1663	>	ForkJoinTask<?> subtask = task;     // current target
1664	>	for (WorkQueue j = joiner, v;;) {   // v is stealer of subtask
1665	>	WorkQueue[] ws; int m, s, h;
1666	>	if ((s = task.status) < 0) {
1667	>	stat = s;
1668	>	break restart;
1669		}
1670	<	if (stealer == null)
1671	<	break;
1672	<	}
1673	<
1674	<	for (WorkQueue q = stealer;;) { // Try to help stealer
1675	<	ForkJoinTask<?> t; int b;
1676	<	if (task.status < 0)
1677	<	break outer;
1678	<	if ((b = q.base) - q.top < 0) {
1679	<	progress = true;
1680	<	if (subtask.status < 0)
1681	<	break outer;               // stale
1682	<	if ((t = q.pollAt(b)) != null) {
1683	<	stealer.stealHint = joiner.poolIndex;
1684	<	joiner.runSubtask(t);
1670	>	if ((ws = workQueues) == null || (m = ws.length - 1) <= 0)
1671	>	break restart;              // shutting down
1672	>	if ((v = ws[h = (j.stealHint | 1) & m]) == null ||
1673	>	v.currentSteal != subtask) {
1674	>	for (int origin = h;;) {    // find stealer
1675	>	if (((h = (h + 2) & m) & 15) == 1 &&
1676	>	(subtask.status < 0 || j.currentJoin != subtask))
1677	>	continue restart;   // occasional staleness check
1678	>	if ((v = ws[h]) != null &&
1679	>	v.currentSteal == subtask) {
1680	>	j.stealHint = h;    // save hint
1681	>	break;
1682	>	}
1683	>	if (h == origin)
1684	>	break restart;      // cannot find stealer
1685		}
1686		}
1687	<	else { // empty - try to descend to find stealer's stealer
1688	<	ForkJoinTask<?> next = stealer.currentJoin;
1689	<	if (++depth == MAX_HELP_DEPTH || subtask.status < 0 ||
1690	<	next == null || next == subtask)
1691	<	break outer;  // max depth, stale, dead-end, cyclic
1692	<	subtask = next;
1693	<	j = stealer;
1694	<	break;
1687	>	for (;;) { // help stealer or descend to its stealer
1688	>	ForkJoinTask[] a;  int b;
1689	>	if (subtask.status < 0)     // surround probes with
1690	>	continue restart;       //   consistency checks
1691	>	if ((b = v.base) - v.top < 0 && (a = v.array) != null) {
1692	>	int i = (((a.length - 1) & b) << ASHIFT) + ABASE;
1693	>	ForkJoinTask<?> t =
1694	>	(ForkJoinTask<?>)U.getObjectVolatile(a, i);
1695	>	if (subtask.status < 0 || j.currentJoin != subtask ||
1696	>	v.currentSteal != subtask)
1697	>	continue restart;   // stale
1698	>	stat = 1;               // apparent progress
1699	>	if (t != null && v.base == b &&
1700	>	U.compareAndSwapObject(a, i, t, null)) {
1701	>	v.base = b + 1;     // help stealer
1702	>	joiner.runSubtask(t);
1703	>	}
1704	>	else if (v.base == b && ++steps == MAX_HELP)
1705	>	break restart;      // v apparently stalled
1706	>	}
1707	>	else {                      // empty -- try to descend
1708	>	ForkJoinTask<?> next = v.currentJoin;
1709	>	if (subtask.status < 0 || j.currentJoin != subtask ||
1710	>	v.currentSteal != subtask)
1711	>	continue restart;   // stale
1712	>	else if (next == null || ++steps == MAX_HELP)
1713	>	break restart;      // dead-end or maybe cyclic
1714	>	else {
1715	>	subtask = next;
1716	>	j = v;
1717	>	break;
1718	>	}
1719	>	}
1720		}
1721		}
1722		}
1723		}
1724	<	return progress;
1724	>	return stat;
1725		}
1726
1727		/**
#	Line 1663 | Line 1730 | public class ForkJoinPool extends Abstra
1730		* @param joiner the joining worker
1731		* @param task the task
1732		*/
1733	<	final void tryPollForAndExec(WorkQueue joiner, ForkJoinTask<?> task) {
1733	>	private void tryPollForAndExec(WorkQueue joiner, ForkJoinTask<?> task) {
1734		WorkQueue[] ws;
1735	<	int m = runState & SMASK;
1736	<	if ((ws = workQueues) != null && ws.length > m) {
1670	<	for (int j = 1; j <= m && task.status >= 0; j += 2) {
1735	>	if ((ws = workQueues) != null) {
1736	>	for (int j = 1; j < ws.length && task.status >= 0; j += 2) {
1737		WorkQueue q = ws[j];
1738		if (q != null && q.pollFor(task)) {
1739		joiner.runSubtask(task);
#	Line 1678 | Line 1744 | public class ForkJoinPool extends Abstra
1744		}
1745
1746		/**
1747	<	* Returns a non-empty steal queue, if one is found during a random,
1748	<	* then cyclic scan, else null.  This method must be retried by
1749	<	* caller if, by the time it tries to use the queue, it is empty.
1747	>	* Tries to decrement active count (sometimes implicitly) and
1748	>	* possibly release or create a compensating worker in preparation
1749	>	* for blocking. Fails on contention or termination. Otherwise,
1750	>	* adds a new thread if no idle workers are available and either
1751	>	* pool would become completely starved or: (at least half
1752	>	* starved, and fewer than 50% spares exist, and there is at least
1753	>	* one task apparently available). Even though the availability
1754	>	* check requires a full scan, it is worthwhile in reducing false
1755	>	* alarms.
1756	>	*
1757	>	* @param task if non-null, a task being waited for
1758	>	* @param blocker if non-null, a blocker being waited for
1759	>	* @return true if the caller can block, else should recheck and retry
1760	>	*/
1761	>	final boolean tryCompensate(ForkJoinTask<?> task, ManagedBlocker blocker) {
1762	>	int pc = parallelism, e;
1763	>	long c = ctl;
1764	>	WorkQueue[] ws = workQueues;
1765	>	if ((e = (int)c) >= 0 && ws != null) {
1766	>	int u, a, ac, hc;
1767	>	int tc = (short)((u = (int)(c >>> 32)) >>> UTC_SHIFT) + pc;
1768	>	boolean replace = false;
1769	>	if ((a = u >> UAC_SHIFT) <= 0) {
1770	>	if ((ac = a + pc) <= 1)
1771	>	replace = true;
1772	>	else if ((e > 0 || (task != null &&
1773	>	ac <= (hc = pc >>> 1) && tc < pc + hc))) {
1774	>	WorkQueue w;
1775	>	for (int j = 0; j < ws.length; ++j) {
1776	>	if ((w = ws[j]) != null && !w.isEmpty()) {
1777	>	replace = true;
1778	>	break;   // in compensation range and tasks available
1779	>	}
1780	>	}
1781	>	}
1782	>	}
1783	>	if ((task == null || task.status >= 0) && // recheck need to block
1784	>	(blocker == null || !blocker.isReleasable()) && ctl == c) {
1785	>	if (!replace) {          // no compensation
1786	>	long nc = ((c - AC_UNIT) & AC_MASK) | (c & ~AC_MASK);
1787	>	if (U.compareAndSwapLong(this, CTL, c, nc))
1788	>	return true;
1789	>	}
1790	>	else if (e != 0) {       // release an idle worker
1791	>	WorkQueue w; Thread p; int i;
1792	>	if ((i = e & SMASK) < ws.length && (w = ws[i]) != null) {
1793	>	long nc = ((long)(w.nextWait & E_MASK) |
1794	>	(c & (AC_MASK|TC_MASK)));
1795	>	if (w.eventCount == (e | INT_SIGN) &&
1796	>	U.compareAndSwapLong(this, CTL, c, nc)) {
1797	>	w.eventCount = (e + E_SEQ) & E_MASK;
1798	>	if ((p = w.parker) != null)
1799	>	U.unpark(p);
1800	>	return true;
1801	>	}
1802	>	}
1803	>	}
1804	>	else if (tc < MAX_CAP) { // create replacement
1805	>	long nc = ((c + TC_UNIT) & TC_MASK) | (c & ~TC_MASK);
1806	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1807	>	addWorker();
1808	>	return true;
1809	>	}
1810	>	}
1811	>	}
1812	>	}
1813	>	return false;
1814	>	}
1815	>
1816	>	/**
1817	>	* Helps and/or blocks until the given task is done.
1818	>	*
1819	>	* @param joiner the joining worker
1820	>	* @param task the task
1821	>	* @return task status on exit
1822	>	*/
1823	>	final int awaitJoin(WorkQueue joiner, ForkJoinTask<?> task) {
1824	>	int s;
1825	>	if ((s = task.status) >= 0) {
1826	>	ForkJoinTask<?> prevJoin = joiner.currentJoin;
1827	>	joiner.currentJoin = task;
1828	>	long startTime = 0L;
1829	>	for (int k = 0;;) {
1830	>	if ((s = (joiner.isEmpty() ?           // try to help
1831	>	tryHelpStealer(joiner, task) :
1832	>	joiner.tryRemoveAndExec(task))) == 0 &&
1833	>	(s = task.status) >= 0) {
1834	>	if (k == 0) {
1835	>	startTime = System.nanoTime();
1836	>	tryPollForAndExec(joiner, task); // check uncommon case
1837	>	}
1838	>	else if ((k & (MAX_HELP - 1)) == 0 &&
1839	>	System.nanoTime() - startTime >=
1840	>	COMPENSATION_DELAY &&
1841	>	tryCompensate(task, null)) {
1842	>	if (task.trySetSignal()) {
1843	>	synchronized (task) {
1844	>	if (task.status >= 0) {
1845	>	try {                // see ForkJoinTask
1846	>	task.wait();     //  for explanation
1847	>	} catch (InterruptedException ie) {
1848	>	}
1849	>	}
1850	>	else
1851	>	task.notifyAll();
1852	>	}
1853	>	}
1854	>	long c;                          // re-activate
1855	>	do {} while (!U.compareAndSwapLong
1856	>	(this, CTL, c = ctl, c + AC_UNIT));
1857	>	}
1858	>	}
1859	>	if (s < 0 || (s = task.status) < 0) {
1860	>	joiner.currentJoin = prevJoin;
1861	>	break;
1862	>	}
1863	>	else if ((k++ & (MAX_HELP - 1)) == MAX_HELP >>> 1)
1864	>	Thread.yield();                     // for politeness
1865	>	}
1866	>	}
1867	>	return s;
1868	>	}
1869	>
1870	>	/**
1871	>	* Stripped-down variant of awaitJoin used by timed joins. Tries
1872	>	* to help join only while there is continuous progress. (Caller
1873	>	* will then enter a timed wait.)
1874	>	*
1875	>	* @param joiner the joining worker
1876	>	* @param task the task
1877	>	* @return task status on exit
1878	>	*/
1879	>	final int helpJoinOnce(WorkQueue joiner, ForkJoinTask<?> task) {
1880	>	int s;
1881	>	while ((s = task.status) >= 0 &&
1882	>	(joiner.isEmpty() ?
1883	>	tryHelpStealer(joiner, task) :
1884	>	joiner.tryRemoveAndExec(task)) != 0)
1885	>	;
1886	>	return s;
1887	>	}
1888	>
1889	>	/**
1890	>	* Returns a (probably) non-empty steal queue, if one is found
1891	>	* during a random, then cyclic scan, else null.  This method must
1892	>	* be retried by caller if, by the time it tries to use the queue,
1893	>	* it is empty.
1894		*/
1895		private WorkQueue findNonEmptyStealQueue(WorkQueue w) {
1896	<	int r = w.seed;    // Same idea as scan(), but ignoring submissions
1896	>	// Similar to loop in scan(), but ignoring submissions
1897	>	int r = w.seed; r ^= r << 13; r ^= r >>> 17; w.seed = r ^= r << 5;
1898	>	int step = (r >>> 16) | 1;
1899		for (WorkQueue[] ws;;) {
1900	<	int m = runState & SMASK;
1901	<	if ((ws = workQueues) == null)
1900	>	int rs = runState, m;
1901	>	if ((ws = workQueues) == null || (m = ws.length - 1) < 1)
1902		return null;
1903	<	if (ws.length > m) {
1904	<	WorkQueue q;
1905	<	for (int n = m << 2, k = r, j = -n;;) {
1906	<	r ^= r << 13; r ^= r >>> 17; r ^= r << 5;
1907	<	if ((q = ws[(k | 1) & m]) != null && q.base - q.top < 0) {
1908	<	w.seed = r;
1697	<	return q;
1698	<	}
1699	<	else if (j > n)
1903	>	for (int j = (m + 1) << 2; ; r += step) {
1904	>	WorkQueue q = ws[((r << 1) | 1) & m];
1905	>	if (q != null && !q.isEmpty())
1906	>	return q;
1907	>	else if (--j < 0) {
1908	>	if (runState == rs)
1909		return null;
1910	<	else
1702	<	k = (j++ < 0) ? r : k + ((m >>> 1) | 1);
1703	<
1910	>	break;
1911		}
1912		}
1913		}
1914		}
1915
1916	+
1917		/**
1918		* Runs tasks until {@code isQuiescent()}. We piggyback on
1919		* active count ctl maintenance, but rather than blocking
#	Line 1714 | Line 1922 | public class ForkJoinPool extends Abstra
1922		*/
1923		final void helpQuiescePool(WorkQueue w) {
1924		for (boolean active = true;;) {
1925	<	w.runLocalTasks();      // exhaust local queue
1925	>	ForkJoinTask<?> localTask; // exhaust local queue
1926	>	while ((localTask = w.nextLocalTask()) != null)
1927	>	localTask.doExec();
1928		WorkQueue q = findNonEmptyStealQueue(w);
1929		if (q != null) {
1930	<	ForkJoinTask<?> t;
1930	>	ForkJoinTask<?> t; int b;
1931		if (!active) {      // re-establish active count
1932		long c;
1933		active = true;
1934		do {} while (!U.compareAndSwapLong
1935		(this, CTL, c = ctl, c + AC_UNIT));
1936		}
1937	<	if ((t = q.poll()) != null)
1937	>	if ((b = q.base) - q.top < 0 && (t = q.pollAt(b)) != null)
1938		w.runSubtask(t);
1939		}
1940		else {
#	Line 1746 | Line 1956 | public class ForkJoinPool extends Abstra
1956		}
1957
1958		/**
1959	<	* Gets and removes a local or stolen task for the given worker
1959	>	* Gets and removes a local or stolen task for the given worker.
1960		*
1961		* @return a task, if available
1962		*/
1963		final ForkJoinTask<?> nextTaskFor(WorkQueue w) {
1964		for (ForkJoinTask<?> t;;) {
1965	<	WorkQueue q;
1965	>	WorkQueue q; int b;
1966		if ((t = w.nextLocalTask()) != null)
1967		return t;
1968		if ((q = findNonEmptyStealQueue(w)) == null)
1969		return null;
1970	<	if ((t = q.poll()) != null)
1970	>	if ((b = q.base) - q.top < 0 && (t = q.pollAt(b)) != null)
1971		return t;
1972		}
1973		}
#	Line 1778 | Line 1988 | public class ForkJoinPool extends Abstra
1988		8);
1989		}
1990
1991	<	// Termination
1991	>	//  Termination
1992
1993		/**
1994	<	* Sets SHUTDOWN bit of runState under lock
1995	<	*/
1996	<	private void enableShutdown() {
1997	<	ReentrantLock lock = this.lock;
1998	<	if (runState >= 0) {
1999	<	lock.lock();                       // don't need try/finally
2000	<	runState |= SHUTDOWN;
1791	<	lock.unlock();
1792	<	}
1793	<	}
1794	<
1795	<	/**
1796	<	* Possibly initiates and/or completes termination.  Upon
1797	<	* termination, cancels all queued tasks and then
1994	>	* Possibly initiates and/or completes termination.  The caller
1995	>	* triggering termination runs three passes through workQueues:
1996	>	* (0) Setting termination status, followed by wakeups of queued
1997	>	* workers; (1) cancelling all tasks; (2) interrupting lagging
1998	>	* threads (likely in external tasks, but possibly also blocked in
1999	>	* joins).  Each pass repeats previous steps because of potential
2000	>	* lagging thread creation.
2001		*
2002		* @param now if true, unconditionally terminate, else only
2003		* if no work and no active workers
2004	+	* @param enable if true, enable shutdown when next possible
2005		* @return true if now terminating or terminated
2006		*/
2007	<	private boolean tryTerminate(boolean now) {
2007	>	private boolean tryTerminate(boolean now, boolean enable) {
2008	>	Mutex lock = this.lock;
2009		for (long c;;) {
2010		if (((c = ctl) & STOP_BIT) != 0) {      // already terminating
2011		if ((short)(c >>> TC_SHIFT) == -parallelism) {
1807	–	ReentrantLock lock = this.lock; // signal when no workers
2012		lock.lock();                    // don't need try/finally
2013		termination.signalAll();        // signal when 0 workers
2014		lock.unlock();
2015		}
2016		return true;
2017		}
2018	<	if (!now) {
2019	<	if ((int)(c >> AC_SHIFT) != -parallelism || runState >= 0 ||
2018	>	if (runState >= 0) {                    // not yet enabled
2019	>	if (!enable)
2020	>	return false;
2021	>	lock.lock();
2022	>	runState |= SHUTDOWN;
2023	>	lock.unlock();
2024	>	}
2025	>	if (!now) {                             // check if idle & no tasks
2026	>	if ((int)(c >> AC_SHIFT) != -parallelism ||
2027		hasQueuedSubmissions())
2028		return false;
2029		// Check for unqueued inactive workers. One pass suffices.
2030		WorkQueue[] ws = workQueues; WorkQueue w;
2031		if (ws != null) {
2032	<	int n = ws.length;
1822	<	for (int i = 1; i < n; i += 2) {
2032	>	for (int i = 1; i < ws.length; i += 2) {
2033		if ((w = ws[i]) != null && w.eventCount >= 0)
2034		return false;
2035		}
2036		}
2037		}
2038	<	if (U.compareAndSwapLong(this, CTL, c, c | STOP_BIT))
2039	<	startTerminating();
2040	<	}
2041	<	}
2042	<
2043	<	/**
2044	<	* Initiates termination: Runs three passes through workQueues:
2045	<	* (0) Setting termination status, followed by wakeups of queued
2046	<	* workers; (1) cancelling all tasks; (2) interrupting lagging
2047	<	* threads (likely in external tasks, but possibly also blocked in
2048	<	* joins).  Each pass repeats previous steps because of potential
2049	<	* lagging thread creation.
2050	<	*/
1841	<	private void startTerminating() {
1842	<	for (int pass = 0; pass < 3; ++pass) {
1843	<	WorkQueue[] ws = workQueues;
1844	<	if (ws != null) {
1845	<	WorkQueue w; Thread wt;
1846	<	int n = ws.length;
1847	<	for (int j = 0; j < n; ++j) {
1848	<	if ((w = ws[j]) != null) {
1849	<	w.runState = -1;
1850	<	if (pass > 0) {
1851	<	w.cancelAll();
1852	<	if (pass > 1 && (wt = w.owner) != null &&
1853	<	!wt.isInterrupted()) {
1854	<	try {
1855	<	wt.interrupt();
1856	<	} catch (SecurityException ignore) {
2038	>	if (U.compareAndSwapLong(this, CTL, c, c | STOP_BIT)) {
2039	>	for (int pass = 0; pass < 3; ++pass) {
2040	>	WorkQueue[] ws = workQueues;
2041	>	if (ws != null) {
2042	>	WorkQueue w;
2043	>	int n = ws.length;
2044	>	for (int i = 0; i < n; ++i) {
2045	>	if ((w = ws[i]) != null) {
2046	>	w.runState = -1;
2047	>	if (pass > 0) {
2048	>	w.cancelAll();
2049	>	if (pass > 1)
2050	>	w.interruptOwner();
2051		}
2052		}
2053		}
2054	<	}
2055	<	}
2056	<	// Wake up workers parked on event queue
2057	<	int i, e; long c; Thread p;
2058	<	while ((i = ((~(e = (int)(c = ctl)) << 1) | 1) & SMASK) < n &&
2059	<	(w = ws[i]) != null &&
2060	<	w.eventCount == (e | INT_SIGN)) {
2061	<	long nc = ((long)(w.nextWait & E_MASK) |
2062	<	((c + AC_UNIT) & AC_MASK) |
2063	<	(c & (TC_MASK|STOP_BIT)));
2064	<	if (U.compareAndSwapLong(this, CTL, c, nc)) {
2065	<	w.eventCount = (e + E_SEQ) & E_MASK;
2066	<	if ((p = w.parker) != null)
2067	<	U.unpark(p);
2054	>	// Wake up workers parked on event queue
2055	>	int i, e; long cc; Thread p;
2056	>	while ((e = (int)(cc = ctl) & E_MASK) != 0 &&
2057	>	(i = e & SMASK) < n &&
2058	>	(w = ws[i]) != null) {
2059	>	long nc = ((long)(w.nextWait & E_MASK) |
2060	>	((cc + AC_UNIT) & AC_MASK) |
2061	>	(cc & (TC_MASK|STOP_BIT)));
2062	>	if (w.eventCount == (e | INT_SIGN) &&
2063	>	U.compareAndSwapLong(this, CTL, cc, nc)) {
2064	>	w.eventCount = (e + E_SEQ) & E_MASK;
2065	>	w.runState = -1;
2066	>	if ((p = w.parker) != null)
2067	>	U.unpark(p);
2068	>	}
2069	>	}
2070		}
2071		}
2072		}
#	Line 1946 | Line 2142 | public class ForkJoinPool extends Abstra
2142		checkPermission();
2143		if (factory == null)
2144		throw new NullPointerException();
2145	<	if (parallelism <= 0 || parallelism > MAX_ID)
2145	>	if (parallelism <= 0 || parallelism > MAX_CAP)
2146		throw new IllegalArgumentException();
2147		this.parallelism = parallelism;
2148		this.factory = factory;
2149		this.ueh = handler;
2150		this.localMode = asyncMode ? FIFO_QUEUE : LIFO_QUEUE;
1955	–	this.nextPoolIndex = 1;
2151		long np = (long)(-parallelism); // offset ctl counts
2152		this.ctl = ((np << AC_SHIFT) & AC_MASK) | ((np << TC_SHIFT) & TC_MASK);
2153	<	// initialize workQueues array with room for 2*parallelism if possible
2154	<	int n = parallelism << 1;
2155	<	if (n >= MAX_ID)
2156	<	n = MAX_ID;
2157	<	else { // See Hackers Delight, sec 3.2, where n < (1 << 16)
2158	<	n |= n >>> 1; n |= n >>> 2; n |= n >>> 4; n |= n >>> 8;
2159	<	}
1965	<	this.workQueues = new WorkQueue[(n + 1) << 1];
1966	<	ReentrantLock lck = this.lock = new ReentrantLock();
1967	<	this.termination = lck.newCondition();
2153	>	// Use nearest power 2 for workQueues size. See Hackers Delight sec 3.2.
2154	>	int n = parallelism - 1;
2155	>	n |= n >>> 1; n |= n >>> 2; n |= n >>> 4; n |= n >>> 8; n |= n >>> 16;
2156	>	int size = (n + 1) << 1;        // #slots = 2*#workers
2157	>	this.submitMask = size - 1;     // room for max # of submit queues
2158	>	this.workQueues = new WorkQueue[size];
2159	>	this.termination = (this.lock = new Mutex()).newCondition();
2160		this.stealCount = new AtomicLong();
2161		this.nextWorkerNumber = new AtomicInteger();
2162	+	int pn = poolNumberGenerator.incrementAndGet();
2163		StringBuilder sb = new StringBuilder("ForkJoinPool-");
2164	<	sb.append(poolNumberGenerator.incrementAndGet());
2164	>	sb.append(Integer.toString(pn));
2165		sb.append("-worker-");
2166		this.workerNamePrefix = sb.toString();
2167	<	// Create initial submission queue
2168	<	WorkQueue sq = tryAddSharedQueue(0);
2169	<	if (sq != null)
1977	<	sq.growArray(false);
2167	>	lock.lock();
2168	>	this.runState = 1;              // set init flag
2169	>	lock.unlock();
2170		}
2171
2172		// Execution methods
#	Line 1996 | Line 2188 | public class ForkJoinPool extends Abstra
2188		*         scheduled for execution
2189		*/
2190		public <T> T invoke(ForkJoinTask<T> task) {
2191	+	if (task == null)
2192	+	throw new NullPointerException();
2193		doSubmit(task);
2194		return task.join();
2195		}
#	Line 2009 | Line 2203 | public class ForkJoinPool extends Abstra
2203		*         scheduled for execution
2204		*/
2205		public void execute(ForkJoinTask<?> task) {
2206	+	if (task == null)
2207	+	throw new NullPointerException();
2208		doSubmit(task);
2209		}
2210
#	Line 2026 | Line 2222 | public class ForkJoinPool extends Abstra
2222		if (task instanceof ForkJoinTask<?>) // avoid re-wrap
2223		job = (ForkJoinTask<?>) task;
2224		else
2225	<	job = ForkJoinTask.adapt(task, null);
2225	>	job = new ForkJoinTask.AdaptedRunnableAction(task);
2226		doSubmit(job);
2227		}
2228
#	Line 2040 | Line 2236 | public class ForkJoinPool extends Abstra
2236		*         scheduled for execution
2237		*/
2238		public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) {
2239	+	if (task == null)
2240	+	throw new NullPointerException();
2241		doSubmit(task);
2242		return task;
2243		}
#	Line 2050 | Line 2248 | public class ForkJoinPool extends Abstra
2248		*         scheduled for execution
2249		*/
2250		public <T> ForkJoinTask<T> submit(Callable<T> task) {
2251	<	if (task == null)
2054	<	throw new NullPointerException();
2055	<	ForkJoinTask<T> job = ForkJoinTask.adapt(task);
2251	>	ForkJoinTask<T> job = new ForkJoinTask.AdaptedCallable<T>(task);
2252		doSubmit(job);
2253		return job;
2254		}
#	Line 2063 | Line 2259 | public class ForkJoinPool extends Abstra
2259		*         scheduled for execution
2260		*/
2261		public <T> ForkJoinTask<T> submit(Runnable task, T result) {
2262	<	if (task == null)
2067	<	throw new NullPointerException();
2068	<	ForkJoinTask<T> job = ForkJoinTask.adapt(task, result);
2262	>	ForkJoinTask<T> job = new ForkJoinTask.AdaptedRunnable<T>(task, result);
2263		doSubmit(job);
2264		return job;
2265		}
#	Line 2082 | Line 2276 | public class ForkJoinPool extends Abstra
2276		if (task instanceof ForkJoinTask<?>) // avoid re-wrap
2277		job = (ForkJoinTask<?>) task;
2278		else
2279	<	job = ForkJoinTask.adapt(task, null);
2279	>	job = new ForkJoinTask.AdaptedRunnableAction(task);
2280		doSubmit(job);
2281		return job;
2282		}
#	Line 2092 | Line 2286 | public class ForkJoinPool extends Abstra
2286		* @throws RejectedExecutionException {@inheritDoc}
2287		*/
2288		public <T> List<Future<T>> invokeAll(Collection<? extends Callable<T>> tasks) {
2289	<	ArrayList<ForkJoinTask<T>> forkJoinTasks =
2290	<	new ArrayList<ForkJoinTask<T>>(tasks.size());
2291	<	for (Callable<T> task : tasks)
2292	<	forkJoinTasks.add(ForkJoinTask.adapt(task));
2293	<	invoke(new InvokeAll<T>(forkJoinTasks));
2294	<
2289	>	// In previous versions of this class, this method constructed
2290	>	// a task to run ForkJoinTask.invokeAll, but now external
2291	>	// invocation of multiple tasks is at least as efficient.
2292	>	List<ForkJoinTask<T>> fs = new ArrayList<ForkJoinTask<T>>(tasks.size());
2293	>	// Workaround needed because method wasn't declared with
2294	>	// wildcards in return type but should have been.
2295		@SuppressWarnings({"unchecked", "rawtypes"})
2296	<	List<Future<T>> futures = (List<Future<T>>) (List) forkJoinTasks;
2103	<	return futures;
2104	<	}
2296	>	List<Future<T>> futures = (List<Future<T>>) (List) fs;
2297
2298	<	static final class InvokeAll<T> extends RecursiveAction {
2299	<	final ArrayList<ForkJoinTask<T>> tasks;
2300	<	InvokeAll(ArrayList<ForkJoinTask<T>> tasks) { this.tasks = tasks; }
2301	<	public void compute() {
2302	<	try { invokeAll(tasks); }
2303	<	catch (Exception ignore) {}
2298	>	boolean done = false;
2299	>	try {
2300	>	for (Callable<T> t : tasks) {
2301	>	ForkJoinTask<T> f = new ForkJoinTask.AdaptedCallable<T>(t);
2302	>	doSubmit(f);
2303	>	fs.add(f);
2304	>	}
2305	>	for (ForkJoinTask<T> f : fs)
2306	>	f.quietlyJoin();
2307	>	done = true;
2308	>	return futures;
2309	>	} finally {
2310	>	if (!done)
2311	>	for (ForkJoinTask<T> f : fs)
2312	>	f.cancel(false);
2313		}
2113	–	private static final long serialVersionUID = -7914297376763021607L;
2314		}
2315
2316		/**
#	Line 2175 | Line 2375 | public class ForkJoinPool extends Abstra
2375		int rc = 0;
2376		WorkQueue[] ws; WorkQueue w;
2377		if ((ws = workQueues) != null) {
2378	<	int n = ws.length;
2379	<	for (int i = 1; i < n; i += 2) {
2180	<	Thread.State s; ForkJoinWorkerThread wt;
2181	<	if ((w = ws[i]) != null && (wt = w.owner) != null &&
2182	<	w.eventCount >= 0 &&
2183	<	(s = wt.getState()) != Thread.State.BLOCKED &&
2184	<	s != Thread.State.WAITING &&
2185	<	s != Thread.State.TIMED_WAITING)
2378	>	for (int i = 1; i < ws.length; i += 2) {
2379	>	if ((w = ws[i]) != null && w.isApparentlyUnblocked())
2380		++rc;
2381		}
2382		}
#	Line 2231 | Line 2425 | public class ForkJoinPool extends Abstra
2425		long count = stealCount.get();
2426		WorkQueue[] ws; WorkQueue w;
2427		if ((ws = workQueues) != null) {
2428	<	int n = ws.length;
2235	<	for (int i = 1; i < n; i += 2) {
2428	>	for (int i = 1; i < ws.length; i += 2) {
2429		if ((w = ws[i]) != null)
2430		count += w.totalSteals;
2431		}
#	Line 2254 | Line 2447 | public class ForkJoinPool extends Abstra
2447		long count = 0;
2448		WorkQueue[] ws; WorkQueue w;
2449		if ((ws = workQueues) != null) {
2450	<	int n = ws.length;
2258	<	for (int i = 1; i < n; i += 2) {
2450	>	for (int i = 1; i < ws.length; i += 2) {
2451		if ((w = ws[i]) != null)
2452		count += w.queueSize();
2453		}
#	Line 2274 | Line 2466 | public class ForkJoinPool extends Abstra
2466		int count = 0;
2467		WorkQueue[] ws; WorkQueue w;
2468		if ((ws = workQueues) != null) {
2469	<	int n = ws.length;
2278	<	for (int i = 0; i < n; i += 2) {
2469	>	for (int i = 0; i < ws.length; i += 2) {
2470		if ((w = ws[i]) != null)
2471		count += w.queueSize();
2472		}
#	Line 2292 | Line 2483 | public class ForkJoinPool extends Abstra
2483		public boolean hasQueuedSubmissions() {
2484		WorkQueue[] ws; WorkQueue w;
2485		if ((ws = workQueues) != null) {
2486	<	int n = ws.length;
2487	<	for (int i = 0; i < n; i += 2) {
2297	<	if ((w = ws[i]) != null && w.queueSize() != 0)
2486	>	for (int i = 0; i < ws.length; i += 2) {
2487	>	if ((w = ws[i]) != null && !w.isEmpty())
2488		return true;
2489		}
2490		}
#	Line 2311 | Line 2501 | public class ForkJoinPool extends Abstra
2501		protected ForkJoinTask<?> pollSubmission() {
2502		WorkQueue[] ws; WorkQueue w; ForkJoinTask<?> t;
2503		if ((ws = workQueues) != null) {
2504	<	int n = ws.length;
2315	<	for (int i = 0; i < n; i += 2) {
2504	>	for (int i = 0; i < ws.length; i += 2) {
2505		if ((w = ws[i]) != null && (t = w.poll()) != null)
2506		return t;
2507		}
#	Line 2341 | Line 2530 | public class ForkJoinPool extends Abstra
2530		int count = 0;
2531		WorkQueue[] ws; WorkQueue w; ForkJoinTask<?> t;
2532		if ((ws = workQueues) != null) {
2533	<	int n = ws.length;
2345	<	for (int i = 0; i < n; ++i) {
2533	>	for (int i = 0; i < ws.length; ++i) {
2534		if ((w = ws[i]) != null) {
2535		while ((t = w.poll()) != null) {
2536		c.add(t);
#	Line 2362 | Line 2550 | public class ForkJoinPool extends Abstra
2550		* @return a string identifying this pool, as well as its state
2551		*/
2552		public String toString() {
2553	<	long st = getStealCount();
2554	<	long qt = getQueuedTaskCount();
2555	<	long qs = getQueuedSubmissionCount();
2368	<	int rc = getRunningThreadCount();
2369	<	int pc = parallelism;
2553	>	// Use a single pass through workQueues to collect counts
2554	>	long qt = 0L, qs = 0L; int rc = 0;
2555	>	long st = stealCount.get();
2556		long c = ctl;
2557	+	WorkQueue[] ws; WorkQueue w;
2558	+	if ((ws = workQueues) != null) {
2559	+	for (int i = 0; i < ws.length; ++i) {
2560	+	if ((w = ws[i]) != null) {
2561	+	int size = w.queueSize();
2562	+	if ((i & 1) == 0)
2563	+	qs += size;
2564	+	else {
2565	+	qt += size;
2566	+	st += w.totalSteals;
2567	+	if (w.isApparentlyUnblocked())
2568	+	++rc;
2569	+	}
2570	+	}
2571	+	}
2572	+	}
2573	+	int pc = parallelism;
2574		int tc = pc + (short)(c >>> TC_SHIFT);
2575		int ac = pc + (int)(c >> AC_SHIFT);
2576		if (ac < 0) // ignore transient negative
#	Line 2403 | Line 2606 | public class ForkJoinPool extends Abstra
2606		*/
2607		public void shutdown() {
2608		checkPermission();
2609	<	enableShutdown();
2407	<	tryTerminate(false);
2609	>	tryTerminate(false, true);
2610		}
2611
2612		/**
#	Line 2425 | Line 2627 | public class ForkJoinPool extends Abstra
2627		*/
2628		public List<Runnable> shutdownNow() {
2629		checkPermission();
2630	<	enableShutdown();
2429	<	tryTerminate(true);
2630	>	tryTerminate(true, true);
2631		return Collections.emptyList();
2632		}
2633
#	Line 2483 | Line 2684 | public class ForkJoinPool extends Abstra
2684		public boolean awaitTermination(long timeout, TimeUnit unit)
2685		throws InterruptedException {
2686		long nanos = unit.toNanos(timeout);
2687	<	final ReentrantLock lock = this.lock;
2687	>	final Mutex lock = this.lock;
2688		lock.lock();
2689		try {
2690		for (;;) {
#	Line 2597 | Line 2798 | public class ForkJoinPool extends Abstra
2798		ForkJoinPool p = ((t instanceof ForkJoinWorkerThread) ?
2799		((ForkJoinWorkerThread)t).pool : null);
2800		while (!blocker.isReleasable()) {
2801	<	if (p == null || p.tryCompensate()) {
2801	>	if (p == null || p.tryCompensate(null, blocker)) {
2802		try {
2803		do {} while (!blocker.isReleasable() && !blocker.block());
2804		} finally {
#	Line 2614 | Line 2815 | public class ForkJoinPool extends Abstra
2815		// implement RunnableFuture.
2816
2817		protected <T> RunnableFuture<T> newTaskFor(Runnable runnable, T value) {
2818	<	return (RunnableFuture<T>) ForkJoinTask.adapt(runnable, value);
2818	>	return new ForkJoinTask.AdaptedRunnable<T>(runnable, value);
2819		}
2820
2821		protected <T> RunnableFuture<T> newTaskFor(Callable<T> callable) {
2822	<	return (RunnableFuture<T>) ForkJoinTask.adapt(callable);
2822	>	return new ForkJoinTask.AdaptedCallable<T>(callable);
2823		}
2824
2825		// Unsafe mechanics
2826		private static final sun.misc.Unsafe U;
2827		private static final long CTL;
2627	–	private static final long RUNSTATE;
2828		private static final long PARKBLOCKER;
2829	+	private static final int ABASE;
2830	+	private static final int ASHIFT;
2831
2832		static {
2833		poolNumberGenerator = new AtomicInteger();
2834	+	nextSubmitterSeed = new AtomicInteger(0x55555555);
2835		modifyThreadPermission = new RuntimePermission("modifyThread");
2836		defaultForkJoinWorkerThreadFactory =
2837		new DefaultForkJoinWorkerThreadFactory();
2838	+	submitters = new ThreadSubmitter();
2839		int s;
2840		try {
2841		U = getUnsafe();
2842		Class<?> k = ForkJoinPool.class;
2843	<	Class<?> tk = Thread.class;
2843	>	Class<?> ak = ForkJoinTask[].class;
2844		CTL = U.objectFieldOffset
2845		(k.getDeclaredField("ctl"));
2846	<	RUNSTATE = U.objectFieldOffset
2643	<	(k.getDeclaredField("runState"));
2846	>	Class<?> tk = Thread.class;
2847		PARKBLOCKER = U.objectFieldOffset
2848		(tk.getDeclaredField("parkBlocker"));
2849	+	ABASE = U.arrayBaseOffset(ak);
2850	+	s = U.arrayIndexScale(ak);
2851		} catch (Exception e) {
2852		throw new Error(e);
2853		}
2854	+	if ((s & (s-1)) != 0)
2855	+	throw new Error("data type scale not a power of two");
2856	+	ASHIFT = 31 - Integer.numberOfLeadingZeros(s);
2857		}
2858
2859		/**

Diff Legend

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