[ViewVC] Diff of: jsr166/jsr166/src/jsr166y/ForkJoinPool.java

Comparing jsr166/src/jsr166y/ForkJoinPool.java (file contents):
Revision 1.136 by dl, Mon Oct 29 17:23:34 2012 UTC vs.
Revision 1.164 by dl, Tue Dec 18 21:46:16 2012 UTC

#	Line 11 \| Line 11 \| import java.util.Arrays;
11		import java.util.Collection;
12		import java.util.Collections;
13		import java.util.List;
14	–	import java.util.Random;
14		import java.util.concurrent.AbstractExecutorService;
15		import java.util.concurrent.Callable;
16		import java.util.concurrent.ExecutorService;
#	Line 19 \| Line 18 \| import java.util.concurrent.Future;
18		import java.util.concurrent.RejectedExecutionException;
19		import java.util.concurrent.RunnableFuture;
20		import java.util.concurrent.TimeUnit;
22	–	import java.util.concurrent.atomic.AtomicInteger;
23	–	import java.util.concurrent.atomic.AtomicLong;
24	–	import java.util.concurrent.locks.AbstractQueuedSynchronizer;
25	–	import java.util.concurrent.locks.Condition;
21
22		/**
23		* An {@link ExecutorService} for running {@link ForkJoinTask}s.
#	Line 42 \| Line 37 \| import java.util.concurrent.locks.Condit
37		* ForkJoinPool}s may also be appropriate for use with event-style
38		* tasks that are never joined.
39		*
40	<	* <p>A static {@link #commonPool} is available and appropriate for
40	>	* <p>A static {@link #commonPool()} is available and appropriate for
41		* most applications. The common pool is used by any ForkJoinTask that
42		* is not explicitly submitted to a specified pool. Using the common
43		* pool normally reduces resource usage (its threads are slowly
44		* reclaimed during periods of non-use, and reinstated upon subsequent
45	<	* use). The common pool is by default constructed with default
51	<	* parameters, but these may be controlled by setting any or all of
52	<	* the three properties {@code
53	<	* java.util.concurrent.ForkJoinPool.common.{parallelism,
54	<	* threadFactory, exceptionHandler}}.
45	>	* use).
46		*
47		* <p>For applications that require separate or custom pools, a {@code
48		* ForkJoinPool} may be constructed with a given target parallelism
#	Line 60 \| Line 51 \| import java.util.concurrent.locks.Condit
51		* dynamically adding, suspending, or resuming internal worker
52		* threads, even if some tasks are stalled waiting to join
53		* others. However, no such adjustments are guaranteed in the face of
54	<	* blocked IO or other unmanaged synchronization. The nested {@link
54	>	* blocked I/O or other unmanaged synchronization. The nested {@link
55		* ManagedBlocker} interface enables extension of the kinds of
56		* synchronization accommodated.
57		*
#	Line 71 \| Line 62 \| import java.util.concurrent.locks.Condit
62		* {@link #toString} returns indications of pool state in a
63		* convenient form for informal monitoring.
64		*
65	<	* <p> As is the case with other ExecutorServices, there are three
65	>	* <p>As is the case with other ExecutorServices, there are three
66		* main task execution methods summarized in the following table.
67		* These are designed to be used primarily by clients not already
68		* engaged in fork/join computations in the current pool. The main
#	Line 106 \| Line 97 \| import java.util.concurrent.locks.Condit
97		* </tr>
98		* </table>
99		*
100	+	* <p>The common pool is by default constructed with default
101	+	* parameters, but these may be controlled by setting three {@link
102	+	* System#getProperty system properties} with prefix {@code
103	+	* java.util.concurrent.ForkJoinPool.common}: {@code parallelism} --
104	+	* an integer greater than zero, {@code threadFactory} -- the class
105	+	* name of a {@link ForkJoinWorkerThreadFactory}, and {@code
106	+	* exceptionHandler} -- the class name of a {@link
107	+	* java.lang.Thread.UncaughtExceptionHandler
108	+	* Thread.UncaughtExceptionHandler}. Upon any error in establishing
109	+	* these settings, default parameters are used.
110	+	*
111		* <p><b>Implementation notes</b>: This implementation restricts the
112		* maximum number of running threads to 32767. Attempts to create
113		* pools with greater than the maximum number result in
#	Line 192 \| Line 194 \| public class ForkJoinPool extends Abstra
194		* WorkQueues are also used in a similar way for tasks submitted
195		* to the pool. We cannot mix these tasks in the same queues used
196		* for work-stealing (this would contaminate lifo/fifo
197	<	* processing). Instead, we loosely associate submission queues
197	>	* processing). Instead, we randomly associate submission queues
198		* with submitting threads, using a form of hashing. The
199		* ThreadLocal Submitter class contains a value initially used as
200		* a hash code for choosing existing queues, but may be randomly
201		* repositioned upon contention with other submitters. In
202	<	* essence, submitters act like workers except that they never
203	<	* take tasks, and they are multiplexed on to a finite number of
204	<	* shared work queues. However, classes are set up so that future
205	<	* extensions could allow submitters to optionally help perform
206	<	* tasks as well. Insertion of tasks in shared mode requires a
207	<	* lock (mainly to protect in the case of resizing) but we use
208	<	* only a simple spinlock (using bits in field runState), because
209	<	* submitters encountering a busy queue move on to try or create
210	<	* other queues -- they block only when creating and registering
211	<	* new queues.
202	>	* essence, submitters act like workers except that they are
203	>	* restricted to executing local tasks that they submitted (or in
204	>	* the case of CountedCompleters, others with the same root task).
205	>	* However, because most shared/external queue operations are more
206	>	* expensive than internal, and because, at steady state, external
207	>	* submitters will compete for CPU with workers, ForkJoinTask.join
208	>	* and related methods disable them from repeatedly helping to
209	>	* process tasks if all workers are active. Insertion of tasks in
210	>	* shared mode requires a lock (mainly to protect in the case of
211	>	* resizing) but we use only a simple spinlock (using bits in
212	>	* field qlock), because submitters encountering a busy queue move
213	>	* on to try or create other queues -- they block only when
214	>	* creating and registering new queues.
215		*
216		* Management
217		* ==========
#	Line 228 \| Line 233 \| public class ForkJoinPool extends Abstra
233		* and their negations (used for thresholding) to fit into 16bit
234		* fields.
235		*
236	<	* Field "runState" contains 32 bits needed to register and
237	<	* deregister WorkQueues, as well as to enable shutdown. It is
238	<	* only modified under a lock (normally briefly held, but
239	<	* occasionally protecting allocations and resizings) but even
240	<	* when locked remains available to check consistency.
236	>	* Field "plock" is a form of sequence lock with a saturating
237	>	* shutdown bit (similarly for per-queue "qlocks"), mainly
238	>	* protecting updates to the workQueues array, as well as to
239	>	* enable shutdown. When used as a lock, it is normally only very
240	>	* briefly held, so is nearly always available after at most a
241	>	* brief spin, but we use a monitor-based backup strategy to
242	>	* block when needed.
243		*
244		* Recording WorkQueues. WorkQueues are recorded in the
245		* "workQueues" array that is created upon first use and expanded
#	Line 241 \| Line 248 \| public class ForkJoinPool extends Abstra
248		* by a lock but the array is otherwise concurrently readable, and
249		* accessed directly. To simplify index-based operations, the
250		* array size is always a power of two, and all readers must
251	<	* tolerate null slots. Shared (submission) queues are at even
252	<	* indices, worker queues at odd indices. Grouping them together
253	<	* in this way simplifies and speeds up task scanning.
251	>	* tolerate null slots. Worker queues are at odd indices. Shared
252	>	* (submission) queues are at even indices, up to a maximum of 64
253	>	* slots, to limit growth even if array needs to expand to add
254	>	* more workers. Grouping them together in this way simplifies and
255	>	* speeds up task scanning.
256		*
257		* All worker thread creation is on-demand, triggered by task
258		* submissions, replacement of terminated workers, and/or
#	Line 304 \| Line 313 \| public class ForkJoinPool extends Abstra
313		*
314		* Signalling. We create or wake up workers only when there
315		* appears to be at least one task they might be able to find and
316	<	* execute. When a submission is added or another worker adds a
317	<	* task to a queue that previously had fewer than two tasks, they
318	<	* signal waiting workers (or trigger creation of new ones if
319	<	* fewer than the given parallelism level -- see signalWork).
320	<	* These primary signals are buttressed by signals during rescans;
321	<	* together these cover the signals needed in cases when more
322	<	* tasks are pushed but untaken, and improve performance compared
323	<	* to having one thread wake up all workers.
316	>	* execute. However, many other threads may notice the same task
317	>	* and each signal to wake up a thread that might take it. So in
318	>	* general, pools will be over-signalled. When a submission is
319	>	* added or another worker adds a task to a queue that has fewer
320	>	* than two tasks, they signal waiting workers (or trigger
321	>	* creation of new ones if fewer than the given parallelism level
322	>	* -- signalWork), and may leave a hint to the unparked worker to
323	>	* help signal others upon wakeup). These primary signals are
324	>	* buttressed by others (see method helpSignal) whenever other
325	>	* threads scan for work or do not have a task to process. On
326	>	* most platforms, signalling (unpark) overhead time is noticeably
327	>	* long, and the time between signalling a thread and it actually
328	>	* making progress can be very noticeably long, so it is worth
329	>	* offloading these delays from critical paths as much as
330	>	* possible.
331		*
332		* Trimming workers. To release resources after periods of lack of
333		* use, a worker starting to wait when the pool is quiescent will
#	Line 322 \| Line 338 \| public class ForkJoinPool extends Abstra
338		* periods of non-use.
339		*
340		* Shutdown and Termination. A call to shutdownNow atomically sets
341	<	* a runState bit and then (non-atomically) sets each worker's
342	<	* runState status, cancels all unprocessed tasks, and wakes up
341	>	* a plock bit and then (non-atomically) sets each worker's
342	>	* qlock status, cancels all unprocessed tasks, and wakes up
343		* all waiting workers. Detecting whether termination should
344		* commence after a non-abrupt shutdown() call requires more work
345		* and bookkeeping. We need consensus about quiescence (i.e., that
#	Line 351 \| Line 367 \| public class ForkJoinPool extends Abstra
367		* method tryCompensate() may create or re-activate a spare
368		* thread to compensate for blocked joiners until they unblock.
369		*
370	<	* A third form (implemented in tryRemoveAndExec and
371	<	* tryPollForAndExec) amounts to helping a hypothetical
372	<	* compensator: If we can readily tell that a possible action of a
373	<	* compensator is to steal and execute the task being joined, the
374	<	* joining thread can do so directly, without the need for a
375	<	* compensation thread (although at the expense of larger run-time
376	<	* stacks, but the tradeoff is typically worthwhile).
370	>	* A third form (implemented in tryRemoveAndExec) amounts to
371	>	* helping a hypothetical compensator: If we can readily tell that
372	>	* a possible action of a compensator is to steal and execute the
373	>	* task being joined, the joining thread can do so directly,
374	>	* without the need for a compensation thread (although at the
375	>	* expense of larger run-time stacks, but the tradeoff is
376	>	* typically worthwhile).
377		*
378		* The ManagedBlocker extension API can't use helping so relies
379		* only on compensation in method awaitBlocker.
#	Line 379 \| Line 395 \| public class ForkJoinPool extends Abstra
395		* steals, rather than use per-task bookkeeping. This sometimes
396		* requires a linear scan of workQueues array to locate stealers,
397		* but often doesn't because stealers leave hints (that may become
398	<	* stale/wrong) of where to locate them. A stealHint is only a
399	<	* hint because a worker might have had multiple steals and the
400	<	* hint records only one of them (usually the most current).
401	<	* Hinting isolates cost to when it is needed, rather than adding
402	<	* to per-task overhead. (2) It is "shallow", ignoring nesting
403	<	* and potentially cyclic mutual steals. (3) It is intentionally
398	>	* stale/wrong) of where to locate them. It is only a hint
399	>	* because a worker might have had multiple steals and the hint
400	>	* records only one of them (usually the most current). Hinting
401	>	* isolates cost to when it is needed, rather than adding to
402	>	* per-task overhead. (2) It is "shallow", ignoring nesting and
403	>	* potentially cyclic mutual steals. (3) It is intentionally
404		* racy: field currentJoin is updated only while actively joining,
405		* which means that we miss links in the chain during long-lived
406		* tasks, GC stalls etc (which is OK since blocking in such cases
#	Line 392 \| Line 408 \| public class ForkJoinPool extends Abstra
408		* to find work (see MAX_HELP) and fall back to suspending the
409		* worker and if necessary replacing it with another.
410		*
411	+	* Helping actions for CountedCompleters are much simpler: Method
412	+	* helpComplete can take and execute any task with the same root
413	+	* as the task being waited on. However, this still entails some
414	+	* traversal of completer chains, so is less efficient than using
415	+	* CountedCompleters without explicit joins.
416	+	*
417		* It is impossible to keep exactly the target parallelism number
418		* of threads running at any given time. Determining the
419		* existence of conservatively safe helping targets, the
#	Line 413 \| Line 435 \| public class ForkJoinPool extends Abstra
435		* intractable) game with an opponent that may choose the worst
436		* (for us) active thread to stall at any time. We take several
437		* precautions to bound losses (and thus bound gains), mainly in
438	<	* methods tryCompensate and awaitJoin: (1) We only try
439	<	* compensation after attempting enough helping steps (measured
440	<	* via counting and timing) that we have already consumed the
441	<	* estimated cost of creating and activating a new thread. (2) We
442	<	* allow up to 50% of threads to be blocked before initially
443	<	* adding any others, and unless completely saturated, check that
444	<	* some work is available for a new worker before adding. Also, we
445	<	* create up to only 50% more threads until entering a mode that
446	<	* only adds a thread if all others are possibly blocked. All
447	<	* together, this means that we might be half as fast to react,
448	<	* and create half as many threads as possible in the ideal case,
449	<	* but present vastly fewer anomalies in all other cases compared
450	<	* to both more aggressive and more conservative alternatives.
451	<	*
452	<	* Style notes: There is a lot of representation-level coupling
453	<	* among classes ForkJoinPool, ForkJoinWorkerThread, and
454	<	* ForkJoinTask. The fields of WorkQueue maintain data structures
455	<	* managed by ForkJoinPool, so are directly accessed. There is
456	<	* little point trying to reduce this, since any associated future
457	<	* changes in representations will need to be accompanied by
458	<	* algorithmic changes anyway. Several methods intrinsically
459	<	* sprawl because they must accumulate sets of consistent reads of
460	<	* volatiles held in local variables. Methods signalWork() and
461	<	* scan() are the main bottlenecks, so are especially heavily
438	>	* methods tryCompensate and awaitJoin.
439	>	*
440	>	* Common Pool
441	>	* ===========
442	>	*
443	>	* The static commonPool always exists after static
444	>	* initialization. Since it (or any other created pool) need
445	>	* never be used, we minimize initial construction overhead and
446	>	* footprint to the setup of about a dozen fields, with no nested
447	>	* allocation. Most bootstrapping occurs within method
448	>	* fullExternalPush during the first submission to the pool.
449	>	*
450	>	* When external threads submit to the common pool, they can
451	>	* perform some subtask processing (see externalHelpJoin and
452	>	* related methods). We do not need to record whether these
453	>	* submissions are to the common pool -- if not, externalHelpJoin
454	>	* returns quickly (at the most helping to signal some common pool
455	>	* workers). These submitters would otherwise be blocked waiting
456	>	* for completion, so the extra effort (with liberally sprinkled
457	>	* task status checks) in inapplicable cases amounts to an odd
458	>	* form of limited spin-wait before blocking in ForkJoinTask.join.
459	>	*
460	>	* Style notes
461	>	* ===========
462	>	*
463	>	* There is a lot of representation-level coupling among classes
464	>	* ForkJoinPool, ForkJoinWorkerThread, and ForkJoinTask. The
465	>	* fields of WorkQueue maintain data structures managed by
466	>	* ForkJoinPool, so are directly accessed. There is little point
467	>	* trying to reduce this, since any associated future changes in
468	>	* representations will need to be accompanied by algorithmic
469	>	* changes anyway. Several methods intrinsically sprawl because
470	>	* they must accumulate sets of consistent reads of volatiles held
471	>	* in local variables. Methods signalWork() and scan() are the
472	>	* main bottlenecks, so are especially heavily
473		* micro-optimized/mangled. There are lots of inline assignments
474		* (of form "while ((local = field) != 0)") which are usually the
475		* simplest way to ensure the required read orderings (which are
#	Line 444 \| Line 477 \| public class ForkJoinPool extends Abstra
477		* declarations of these locals at the heads of methods or blocks.
478		* There are several occurrences of the unusual "do {} while
479		* (!cas...)" which is the simplest way to force an update of a
480	<	* CAS'ed variable. There are also other coding oddities that help
480	>	* CAS'ed variable. There are also other coding oddities (including
481	>	* several unnecessary-looking hoisted null checks) that help
482		* some methods perform reasonably even when interpreted (not
483		* compiled).
484		*
#	Line 493 \| Line 527 \| public class ForkJoinPool extends Abstra
527		* Default ForkJoinWorkerThreadFactory implementation; creates a
528		* new ForkJoinWorkerThread.
529		*/
530	<	static class DefaultForkJoinWorkerThreadFactory
530	>	static final class DefaultForkJoinWorkerThreadFactory
531		implements ForkJoinWorkerThreadFactory {
532	<	public ForkJoinWorkerThread newThread(ForkJoinPool pool) {
532	>	public final ForkJoinWorkerThread newThread(ForkJoinPool pool) {
533		return new ForkJoinWorkerThread(pool);
534		}
535		}
536
537		/**
538	+	* Per-thread records for threads that submit to pools. Currently
539	+	* holds only pseudo-random seed / index that is used to choose
540	+	* submission queues in method externalPush. In the future, this may
541	+	* also incorporate a means to implement different task rejection
542	+	* and resubmission policies.
543	+	*
544	+	* Seeds for submitters and workers/workQueues work in basically
545	+	* the same way but are initialized and updated using slightly
546	+	* different mechanics. Both are initialized using the same
547	+	* approach as in class ThreadLocal, where successive values are
548	+	* unlikely to collide with previous values. Seeds are then
549	+	* randomly modified upon collisions using xorshifts, which
550	+	* requires a non-zero seed.
551	+	*/
552	+	static final class Submitter {
553	+	int seed;
554	+	Submitter(int s) { seed = s; }
555	+	}
556	+
557	+	/**
558		* Class for artificial tasks that are used to replace the target
559		* of local joins if they are removed from an interior queue slot
560		* in WorkQueue.tryRemoveAndExec. We don't need the proxy to
561		* actually do anything beyond having a unique identity.
562		*/
563		static final class EmptyTask extends ForkJoinTask<Void> {
564	+	private static final long serialVersionUID = -7721805057305804111L;
565		EmptyTask() { status = ForkJoinTask.NORMAL; } // force done
566		public final Void getRawResult() { return null; }
567		public final void setRawResult(Void x) {}
#	Line 527 \| Line 582 \| public class ForkJoinPool extends Abstra
582		*
583		* Field "top" is the index (mod array.length) of the next queue
584		* slot to push to or pop from. It is written only by owner thread
585	<	* for push, or under lock for trySharedPush, and accessed by
586	<	* other threads only after reading (volatile) base. Both top and
587	<	* base are allowed to wrap around on overflow, but (top - base)
588	<	* (or more commonly -(base - top) to force volatile read of base
589	<	* before top) still estimates size.
585	>	* for push, or under lock for external/shared push, and accessed
586	>	* by other threads only after reading (volatile) base. Both top
587	>	* and base are allowed to wrap around on overflow, but (top -
588	>	* base) (or more commonly -(base - top) to force volatile read of
589	>	* base before top) still estimates size. The lock ("qlock") is
590	>	* forced to -1 on termination, causing all further lock attempts
591	>	* to fail. (Note: we don't need CAS for termination state because
592	>	* upon pool shutdown, all shared-queues will stop being used
593	>	* anyway.) Nearly all lock bodies are set up so that exceptions
594	>	* within lock bodies are "impossible" (modulo JVM errors that
595	>	* would cause failure anyway.)
596		*
597		* The array slots are read and written using the emulation of
598		* volatiles/atomics provided by Unsafe. Insertions must in
599		* general use putOrderedObject as a form of releasing store to
600		* ensure that all writes to the task object are ordered before
601	<	* its publication in the queue. (Although we can avoid one case
602	<	* of this when locked in trySharedPush.) All removals entail a
603	<	* CAS to null. The array is always a power of two. To ensure
604	<	* safety of Unsafe array operations, all accesses perform
544	<	* explicit null checks and implicit bounds checks via
545	<	* power-of-two masking.
601	>	* its publication in the queue. All removals entail a CAS to
602	>	* null. The array is always a power of two. To ensure safety of
603	>	* Unsafe array operations, all accesses perform explicit null
604	>	* checks and implicit bounds checks via power-of-two masking.
605		*
606		* In addition to basic queuing support, this class contains
607		* fields described elsewhere to control execution. It turns out
608	<	* to work better memory-layout-wise to include them in this
609	<	* class rather than a separate class.
608	>	* to work better memory-layout-wise to include them in this class
609	>	* rather than a separate class.
610		*
611		* Performance on most platforms is very sensitive to placement of
612		* instances of both WorkQueues and their arrays -- we absolutely
#	Line 561 \| Line 620 \| public class ForkJoinPool extends Abstra
620		* trades off slightly slower average field access for the sake of
621		* avoiding really bad worst-case access. (Until better JVM
622		* support is in place, this padding is dependent on transient
623	<	* properties of JVM field layout rules.) We also take care in
623	>	* properties of JVM field layout rules.) We also take care in
624		* allocating, sizing and resizing the array. Non-shared queue
625	<	* arrays are initialized (via method growArray) by workers before
626	<	* use. Others are allocated on first use.
625	>	* arrays are initialized by workers before use. Others are
626	>	* allocated on first use.
627		*/
628		static final class WorkQueue {
629		/**
#	Line 587 \| Line 646 \| public class ForkJoinPool extends Abstra
646		*/
647		static final int MAXIMUM_QUEUE_CAPACITY = 1 << 26; // 64M
648
649	<	volatile long totalSteals; // cumulative number of steals
649	>	// Heuristic padding to ameliorate unfortunate memory placements
650	>	volatile long pad00, pad01, pad02, pad03, pad04, pad05, pad06;
651	>
652		int seed; // for random scanning; initialize nonzero
653		volatile int eventCount; // encoded inactivation count; < 0 if inactive
654		int nextWait; // encoded record of next event waiter
655	<	int rescans; // remaining scans until block
595	<	int nsteals; // top-level task executions since last idle
596	<	final int mode; // lifo, fifo, or shared
655	>	int hint; // steal or signal hint (index)
656		int poolIndex; // index of this queue in pool (or 0)
657	<	int stealHint; // index of most recent known stealer
658	<	volatile int runState; // 1: locked, -1: terminate; else 0
657	>	final int mode; // 0: lifo, > 0: fifo, < 0: shared
658	>	int nsteals; // number of steals
659	>	volatile int qlock; // 1: locked, -1: terminate; else 0
660		volatile int base; // index of next slot for poll
661		int top; // index of next slot for push
662		ForkJoinTask<?>[] array; // the elements (initially unallocated)
#	Line 605 \| Line 665 \| public class ForkJoinPool extends Abstra
665		volatile Thread parker; // == owner during call to park; else null
666		volatile ForkJoinTask<?> currentJoin; // task being joined in awaitJoin
667		ForkJoinTask<?> currentSteal; // current non-local task being executed
608	–	// Heuristic padding to ameliorate unfortunate memory placements
609	–	Object p00, p01, p02, p03, p04, p05, p06, p07;
610	–	Object p08, p09, p0a, p0b, p0c, p0d, p0e;
668
669	<	WorkQueue(ForkJoinPool pool, ForkJoinWorkerThread owner, int mode) {
670	<	this.mode = mode;
669	>	volatile Object pad10, pad11, pad12, pad13, pad14, pad15, pad16, pad17;
670	>	volatile Object pad18, pad19, pad1a, pad1b, pad1c, pad1d;
671	>
672	>	WorkQueue(ForkJoinPool pool, ForkJoinWorkerThread owner, int mode,
673	>	int seed) {
674		this.pool = pool;
675		this.owner = owner;
676	+	this.mode = mode;
677	+	this.seed = seed;
678		// Place indices in the center of array (that is not yet allocated)
679		base = top = INITIAL_QUEUE_CAPACITY >>> 1;
680		}
#	Line 625 \| Line 687 \| public class ForkJoinPool extends Abstra
687		return (n >= 0) ? 0 : -n; // ignore transient negative
688		}
689
690	<	/**
690	>	/**
691		* Provides a more accurate estimate of whether this queue has
692		* any tasks than does queueSize, by checking whether a
693		* near-empty queue has at least one unclaimed task.
#	Line 637 \| Line 699 \| public class ForkJoinPool extends Abstra
699		(n == -1 &&
700		((a = array) == null \|\|
701		(m = a.length - 1) < 0 \|\|
702	<	U.getObjectVolatile
703	<	(a, ((m & (s - 1)) << ASHIFT) + ABASE) == null)));
702	>	U.getObject
703	>	(a, (long)((m & (s - 1)) << ASHIFT) + ABASE) == null)));
704		}
705
706		/**
707	<	* Pushes a task. Call only by owner in unshared queues.
707	>	* Pushes a task. Call only by owner in unshared queues. (The
708	>	* shared-queue version is embedded in method externalPush.)
709		*
710		* @param task the task. Caller must ensure non-null.
711		* @throw RejectedExecutionException if array cannot be resized
#	Line 651 \| Line 714 \| public class ForkJoinPool extends Abstra
714		ForkJoinTask<?>[] a; ForkJoinPool p;
715		int s = top, m, n;
716		if ((a = array) != null) { // ignore if queue removed
717	<	U.putOrderedObject
718	<	(a, (((m = a.length - 1) & s) << ASHIFT) + ABASE, task);
717	>	int j = (((m = a.length - 1) & s) << ASHIFT) + ABASE;
718	>	U.putOrderedObject(a, j, task);
719		if ((n = (top = s + 1) - base) <= 2) {
720		if ((p = pool) != null)
721	<	p.signalWork();
721	>	p.signalWork(this);
722		}
723		else if (n >= m)
724	<	growArray(true);
724	>	growArray();
725		}
726		}
727
728	<	/**
729	<	* Pushes a task if lock is free and array is either big
730	<	* enough or can be resized to be big enough.
731	<	*
669	<	* @param task the task. Caller must ensure non-null.
670	<	* @return true if submitted
728	>	/**
729	>	* Initializes or doubles the capacity of array. Call either
730	>	* by owner or with lock held -- it is OK for base, but not
731	>	* top, to move while resizings are in progress.
732		*/
733	<	final boolean trySharedPush(ForkJoinTask<?> task) {
734	<	boolean submitted = false;
735	<	if (runState == 0 && U.compareAndSwapInt(this, RUNSTATE, 0, 1)) {
736	<	ForkJoinTask<?>[] a = array;
737	<	int s = top;
738	<	try {
739	<	if ((a != null && a.length > s + 1 - base) \|\|
740	<	(a = growArray(false)) != null) { // must presize
741	<	int j = (((a.length - 1) & s) << ASHIFT) + ABASE;
742	<	U.putObject(a, (long)j, task); // don't need "ordered"
743	<	top = s + 1;
744	<	submitted = true;
745	<	}
746	<	} finally {
747	<	runState = 0; // unlock
748	<	}
733	>	final ForkJoinTask<?>[] growArray() {
734	>	ForkJoinTask<?>[] oldA = array;
735	>	int size = oldA != null ? oldA.length << 1 : INITIAL_QUEUE_CAPACITY;
736	>	if (size > MAXIMUM_QUEUE_CAPACITY)
737	>	throw new RejectedExecutionException("Queue capacity exceeded");
738	>	int oldMask, t, b;
739	>	ForkJoinTask<?>[] a = array = new ForkJoinTask<?>[size];
740	>	if (oldA != null && (oldMask = oldA.length - 1) >= 0 &&
741	>	(t = top) - (b = base) > 0) {
742	>	int mask = size - 1;
743	>	do {
744	>	ForkJoinTask<?> x;
745	>	int oldj = ((b & oldMask) << ASHIFT) + ABASE;
746	>	int j = ((b & mask) << ASHIFT) + ABASE;
747	>	x = (ForkJoinTask<?>)U.getObjectVolatile(oldA, oldj);
748	>	if (x != null &&
749	>	U.compareAndSwapObject(oldA, oldj, x, null))
750	>	U.putObjectVolatile(a, j, x);
751	>	} while (++b != t);
752		}
753	<	return submitted;
753	>	return a;
754		}
755
756		/**
757		* Takes next task, if one exists, in LIFO order. Call only
758	<	* by owner in unshared queues. (We do not have a shared
695	<	* version of this method because it is never needed.)
758	>	* by owner in unshared queues.
759		*/
760		final ForkJoinTask<?> pop() {
761		ForkJoinTask<?>[] a; ForkJoinTask<?> t; int m;
#	Line 747 \| Line 810 \| public class ForkJoinPool extends Abstra
810		else if (base == b) {
811		if (b + 1 == top)
812		break;
813	<	Thread.yield(); // wait for lagging update
813	>	Thread.yield(); // wait for lagging update (very rare)
814		}
815		}
816		return null;
#	Line 774 \| Line 837 \| public class ForkJoinPool extends Abstra
837
838		/**
839		* Pops the given task only if it is at the current top.
840	+	* (A shared version is available only via FJP.tryExternalUnpush)
841		*/
842		final boolean tryUnpush(ForkJoinTask<?> t) {
843		ForkJoinTask<?>[] a; int s;
#	Line 787 \| Line 851 \| public class ForkJoinPool extends Abstra
851		}
852
853		/**
790	–	* Version of tryUnpush for shared queues; called by non-FJ
791	–	* submitters after prechecking that task probably exists.
792	–	*/
793	–	final boolean trySharedUnpush(ForkJoinTask<?> t) {
794	–	boolean success = false;
795	–	if (runState == 0 && U.compareAndSwapInt(this, RUNSTATE, 0, 1)) {
796	–	try {
797	–	ForkJoinTask<?>[] a; int s;
798	–	if ((a = array) != null && (s = top) != base &&
799	–	U.compareAndSwapObject
800	–	(a, (((a.length - 1) & --s) << ASHIFT) + ABASE, t, null)) {
801	–	top = s;
802	–	success = true;
803	–	}
804	–	} finally {
805	–	runState = 0; // unlock
806	–	}
807	–	}
808	–	return success;
809	–	}
810	–
811	–	/**
812	–	* Polls the given task only if it is at the current base.
813	–	*/
814	–	final boolean pollFor(ForkJoinTask<?> task) {
815	–	ForkJoinTask<?>[] a; int b;
816	–	if ((b = base) - top < 0 && (a = array) != null) {
817	–	int j = (((a.length - 1) & b) << ASHIFT) + ABASE;
818	–	if (U.getObjectVolatile(a, j) == task && base == b &&
819	–	U.compareAndSwapObject(a, j, task, null)) {
820	–	base = b + 1;
821	–	return true;
822	–	}
823	–	}
824	–	return false;
825	–	}
826	–
827	–	/**
828	–	* Initializes or doubles the capacity of array. Call either
829	–	* by owner or with lock held -- it is OK for base, but not
830	–	* top, to move while resizings are in progress.
831	–	*
832	–	* @param rejectOnFailure if true, throw exception if capacity
833	–	* exceeded (relayed ultimately to user); else return null.
834	–	*/
835	–	final ForkJoinTask<?>[] growArray(boolean rejectOnFailure) {
836	–	ForkJoinTask<?>[] oldA = array;
837	–	int size = oldA != null ? oldA.length << 1 : INITIAL_QUEUE_CAPACITY;
838	–	if (size <= MAXIMUM_QUEUE_CAPACITY) {
839	–	int oldMask, t, b;
840	–	ForkJoinTask<?>[] a = array = new ForkJoinTask<?>[size];
841	–	if (oldA != null && (oldMask = oldA.length - 1) >= 0 &&
842	–	(t = top) - (b = base) > 0) {
843	–	int mask = size - 1;
844	–	do {
845	–	ForkJoinTask<?> x;
846	–	int oldj = ((b & oldMask) << ASHIFT) + ABASE;
847	–	int j = ((b & mask) << ASHIFT) + ABASE;
848	–	x = (ForkJoinTask<?>)U.getObjectVolatile(oldA, oldj);
849	–	if (x != null &&
850	–	U.compareAndSwapObject(oldA, oldj, x, null))
851	–	U.putObjectVolatile(a, j, x);
852	–	} while (++b != t);
853	–	}
854	–	return a;
855	–	}
856	–	else if (!rejectOnFailure)
857	–	return null;
858	–	else
859	–	throw new RejectedExecutionException("Queue capacity exceeded");
860	–	}
861	–
862	–	/**
854		* Removes and cancels all known tasks, ignoring any exceptions.
855		*/
856		final void cancelAll() {
#	Line 883 \| Line 874 \| public class ForkJoinPool extends Abstra
874		return seed = r ^= r << 5;
875		}
876
877	<	// Execution methods
877	>	// Specialized execution methods
878
879		/**
880		* Pops and runs tasks until empty.
#	Line 912 \| Line 903 \| public class ForkJoinPool extends Abstra
903		}
904
905		/**
906	<	* If present, removes from queue and executes the given task, or
907	<	* any other cancelled task. Returns (true) immediately on any CAS
906	>	* If present, removes from queue and executes the given task,
907	>	* or any other cancelled task. Returns (true) on any CAS
908		* or consistency check failure so caller can retry.
909		*
910	<	* @return 0 if no progress can be made, else positive
920	<	* (this unusual convention simplifies use with tryHelpStealer.)
910	>	* @return false if no progress can be made, else true;
911		*/
912	<	final int tryRemoveAndExec(ForkJoinTask<?> task) {
913	<	int stat = 1;
924	<	boolean removed = false, empty = true;
912	>	final boolean tryRemoveAndExec(ForkJoinTask<?> task) {
913	>	boolean stat = true, removed = false, empty = true;
914		ForkJoinTask<?>[] a; int m, s, b, n;
915		if ((a = array) != null && (m = a.length - 1) >= 0 &&
916		(n = (s = top) - (b = base)) > 0) {
#	Line 951 \| Line 940 \| public class ForkJoinPool extends Abstra
940		}
941		if (--n == 0) {
942		if (!empty && base == b)
943	<	stat = 0;
943	>	stat = false;
944		break;
945		}
946		}
#	Line 962 \| Line 951 \| public class ForkJoinPool extends Abstra
951		}
952
953		/**
954	+	* Polls for and executes the given task or any other task in
955	+	* its CountedCompleter computation
956	+	*/
957	+	final boolean pollAndExecCC(ForkJoinTask<?> root) {
958	+	ForkJoinTask<?>[] a; int b; Object o;
959	+	outer: while ((b = base) - top < 0 && (a = array) != null) {
960	+	long j = (((a.length - 1) & b) << ASHIFT) + ABASE;
961	+	if ((o = U.getObject(a, j)) == null \|\|
962	+	!(o instanceof CountedCompleter))
963	+	break;
964	+	for (CountedCompleter<?> t = (CountedCompleter<?>)o, r = t;;) {
965	+	if (r == root) {
966	+	if (base == b &&
967	+	U.compareAndSwapObject(a, j, t, null)) {
968	+	base = b + 1;
969	+	t.doExec();
970	+	return true;
971	+	}
972	+	else
973	+	break; // restart
974	+	}
975	+	if ((r = r.completer) == null)
976	+	break outer; // not part of root computation
977	+	}
978	+	}
979	+	return false;
980	+	}
981	+
982	+	/**
983		* Executes a top-level task and any local tasks remaining
984		* after execution.
985		*/
986		final void runTask(ForkJoinTask<?> t) {
987		if (t != null) {
988	<	currentSteal = t;
989	<	t.doExec();
990	<	if (top != base) { // process remaining local tasks
988	>	(currentSteal = t).doExec();
989	>	currentSteal = null;
990	>	++nsteals;
991	>	if (base - top < 0) { // process remaining local tasks
992		if (mode == 0)
993		popAndExecAll();
994		else
995		pollAndExecAll();
996		}
978	–	++nsteals;
979	–	currentSteal = null;
997		}
998		}
999
#	Line 986 \| Line 1003 \| public class ForkJoinPool extends Abstra
1003		final void runSubtask(ForkJoinTask<?> t) {
1004		if (t != null) {
1005		ForkJoinTask<?> ps = currentSteal;
1006	<	currentSteal = t;
990	<	t.doExec();
1006	>	(currentSteal = t).doExec();
1007		currentSteal = ps;
1008		}
1009		}
#	Line 1004 \| Line 1020 \| public class ForkJoinPool extends Abstra
1020		s != Thread.State.TIMED_WAITING);
1021		}
1022
1007	–	/**
1008	–	* If this owned and is not already interrupted, try to
1009	–	* interrupt and/or unpark, ignoring exceptions.
1010	–	*/
1011	–	final void interruptOwner() {
1012	–	Thread wt, p;
1013	–	if ((wt = owner) != null && !wt.isInterrupted()) {
1014	–	try {
1015	–	wt.interrupt();
1016	–	} catch (SecurityException ignore) {
1017	–	}
1018	–	}
1019	–	if ((p = parker) != null)
1020	–	U.unpark(p);
1021	–	}
1022	–
1023		// Unsafe mechanics
1024		private static final sun.misc.Unsafe U;
1025	<	private static final long RUNSTATE;
1025	>	private static final long QLOCK;
1026		private static final int ABASE;
1027		private static final int ASHIFT;
1028		static {
#	Line 1031 \| Line 1031 \| public class ForkJoinPool extends Abstra
1031		U = getUnsafe();
1032		Class<?> k = WorkQueue.class;
1033		Class<?> ak = ForkJoinTask[].class;
1034	<	RUNSTATE = U.objectFieldOffset
1035	<	(k.getDeclaredField("runState"));
1034	>	QLOCK = U.objectFieldOffset
1035	>	(k.getDeclaredField("qlock"));
1036		ABASE = U.arrayBaseOffset(ak);
1037		s = U.arrayIndexScale(ak);
1038		} catch (Exception e) {
#	Line 1044 \| Line 1044 \| public class ForkJoinPool extends Abstra
1044		}
1045		}
1046
1047	–	/**
1048	–	* Per-thread records for threads that submit to pools. Currently
1049	–	* holds only pseudo-random seed / index that is used to choose
1050	–	* submission queues in method doSubmit. In the future, this may
1051	–	* also incorporate a means to implement different task rejection
1052	–	* and resubmission policies.
1053	–	*
1054	–	* Seeds for submitters and workers/workQueues work in basically
1055	–	* the same way but are initialized and updated using slightly
1056	–	* different mechanics. Both are initialized using the same
1057	–	* approach as in class ThreadLocal, where successive values are
1058	–	* unlikely to collide with previous values. This is done during
1059	–	* registration for workers, but requires a separate AtomicInteger
1060	–	* for submitters. Seeds are then randomly modified upon
1061	–	* collisions using xorshifts, which requires a non-zero seed.
1062	–	*/
1063	–	static final class Submitter {
1064	–	int seed;
1065	–	Submitter() {
1066	–	int s = nextSubmitterSeed.getAndAdd(SEED_INCREMENT);
1067	–	seed = (s == 0) ? 1 : s; // ensure non-zero
1068	–	}
1069	–	}
1070	–
1071	–	/** ThreadLocal class for Submitters */
1072	–	static final class ThreadSubmitter extends ThreadLocal<Submitter> {
1073	–	public Submitter initialValue() { return new Submitter(); }
1074	–	}
1075	–
1047		// static fields (initialized in static initializer below)
1048
1049		/**
#	Line 1082 \| Line 1053 \| public class ForkJoinPool extends Abstra
1053		public static final ForkJoinWorkerThreadFactory
1054		defaultForkJoinWorkerThreadFactory;
1055
1056	+	/**
1057	+	* Per-thread submission bookkeeping. Shared across all pools
1058	+	* to reduce ThreadLocal pollution and because random motion
1059	+	* to avoid contention in one pool is likely to hold for others.
1060	+	* Lazily initialized on first submission (but null-checked
1061	+	* in other contexts to avoid unnecessary initialization).
1062	+	*/
1063	+	static final ThreadLocal<Submitter> submitters;
1064
1065	<	/** Property prefix for constructing common pool */
1066	<	private static final String propPrefix =
1067	<	"java.util.concurrent.ForkJoinPool.common.";
1065	>	/**
1066	>	* Permission required for callers of methods that may start or
1067	>	* kill threads.
1068	>	*/
1069	>	private static final RuntimePermission modifyThreadPermission;
1070
1071		/**
1072		* Common (static) pool. Non-null for public use unless a static
1073	<	* construction exception, but internal usages must null-check on
1074	<	* use.
1073	>	* construction exception, but internal usages null-check on use
1074	>	* to paranoically avoid potential initialization circularities
1075	>	* as well as to simplify generated code.
1076		*/
1077		static final ForkJoinPool commonPool;
1078
#	Line 1100 \| Line 1082 \| public class ForkJoinPool extends Abstra
1082		static final int commonPoolParallelism;
1083
1084		/**
1085	<	* Generator for assigning sequence numbers as pool names.
1085	>	* Sequence number for creating workerNamePrefix.
1086		*/
1087	<	private static final AtomicInteger poolNumberGenerator;
1087	>	private static int poolNumberSequence;
1088
1089		/**
1090	<	* Generator for initial hashes/seeds for submitters. Accessed by
1091	<	* Submitter class constructor.
1090	>	* Return the next sequence number. We don't expect this to
1091	>	* ever contend so use simple builtin sync.
1092		*/
1093	<	static final AtomicInteger nextSubmitterSeed;
1093	>	private static final synchronized int nextPoolId() {
1094	>	return ++poolNumberSequence;
1095	>	}
1096
1097	<	/**
1114	<	* Permission required for callers of methods that may start or
1115	<	* kill threads.
1116	<	*/
1117	<	private static final RuntimePermission modifyThreadPermission;
1097	>	// static constants
1098
1099		/**
1100	<	* Per-thread submission bookkeeping. Shared across all pools
1101	<	* to reduce ThreadLocal pollution and because random motion
1102	<	* to avoid contention in one pool is likely to hold for others.
1100	>	* Initial timeout value (in nanoseconds) for the thread
1101	>	* triggering quiescence to park waiting for new work. On timeout,
1102	>	* the thread will instead try to shrink the number of
1103	>	* workers. The value should be large enough to avoid overly
1104	>	* aggressive shrinkage during most transient stalls (long GCs
1105	>	* etc).
1106		*/
1107	<	private static final ThreadSubmitter submitters;
1125	<
1126	<	// static constants
1107	>	private static final long IDLE_TIMEOUT = 2000L * 1000L * 1000L; // 2sec
1108
1109		/**
1110	<	* Initial timeout value (in nanoseconds) for the thread triggering
1130	<	* quiescence to park waiting for new work. On timeout, the thread
1131	<	* will instead try to shrink the number of workers.
1110	>	* Timeout value when there are more threads than parallelism level
1111		*/
1112	<	private static final long IDLE_TIMEOUT = 1000L * 1000L * 1000L; // 1sec
1112	>	private static final long FAST_IDLE_TIMEOUT = 200L * 1000L * 1000L;
1113
1114		/**
1115	<	* Timeout value when there are more threads than parallelism level
1115	>	* Tolerance for idle timeouts, to cope with timer undershoots
1116		*/
1117	<	private static final long FAST_IDLE_TIMEOUT = 100L * 1000L * 1000L;
1117	>	private static final long TIMEOUT_SLOP = 2000000L;
1118
1119		/**
1120		* The maximum stolen->joining link depth allowed in method
1121	<	* tryHelpStealer. Must be a power of two. This value also
1143	<	* controls the maximum number of times to try to help join a task
1144	<	* without any apparent progress or change in pool state before
1145	<	* giving up and blocking (see awaitJoin). Depths for legitimate
1121	>	* tryHelpStealer. Must be a power of two. Depths for legitimate
1122		* chains are unbounded, but we use a fixed constant to avoid
1123		* (otherwise unchecked) cycles and to bound staleness of
1124		* traversal parameters at the expense of sometimes blocking when
#	Line 1151 \| Line 1127 \| public class ForkJoinPool extends Abstra
1127		private static final int MAX_HELP = 64;
1128
1129		/**
1154	–	* Secondary time-based bound (in nanosecs) for helping attempts
1155	–	* before trying compensated blocking in awaitJoin. Used in
1156	–	* conjunction with MAX_HELP to reduce variance due to different
1157	–	* polling rates associated with different helping options. The
1158	–	* value should roughly approximate the time required to create
1159	–	* and/or activate a worker thread.
1160	–	*/
1161	–	private static final long COMPENSATION_DELAY = 1L << 18; // ~0.25 millisec
1162	–
1163	–	/**
1130		* Increment for seed generators. See class ThreadLocal for
1131		* explanation.
1132		*/
#	Line 1194 \| Line 1160 \| public class ForkJoinPool extends Abstra
1160		* scan for them to avoid queuing races. Note however that
1161		* eventCount updates lag releases so usage requires care.
1162		*
1163	<	* Field runState is an int packed with:
1163	>	* Field plock is an int packed with:
1164		* SHUTDOWN: true if shutdown is enabled (1 bit)
1165	<	* SEQ: a sequence number updated upon (de)registering workers (30 bits)
1166	<	* INIT: set true after workQueues array construction (1 bit)
1165	>	* SEQ: a sequence lock, with PL_LOCK bit set if locked (30 bits)
1166	>	* SIGNAL: set when threads may be waiting on the lock (1 bit)
1167		*
1168		* The sequence number enables simple consistency checks:
1169		* Staleness of read-only operations on the workQueues array can
1170	<	* be checked by comparing runState before vs after the reads.
1170	>	* be checked by comparing plock before vs after the reads.
1171		*/
1172
1173		// bit positions/shifts for fields
#	Line 1213 \| Line 1179 \| public class ForkJoinPool extends Abstra
1179		// bounds
1180		private static final int SMASK = 0xffff; // short bits
1181		private static final int MAX_CAP = 0x7fff; // max #workers - 1
1182	<	private static final int SQMASK = 0xfffe; // even short bits
1182	>	private static final int EVENMASK = 0xfffe; // even short bits
1183	>	private static final int SQMASK = 0x007e; // max 64 (even) slots
1184		private static final int SHORT_SIGN = 1 << 15;
1185		private static final int INT_SIGN = 1 << 31;
1186
#	Line 1238 \| Line 1205 \| public class ForkJoinPool extends Abstra
1205		private static final int E_MASK = 0x7fffffff; // no STOP_BIT
1206		private static final int E_SEQ = 1 << EC_SHIFT;
1207
1208	<	// runState bits
1208	>	// plock bits
1209		private static final int SHUTDOWN = 1 << 31;
1210	+	private static final int PL_LOCK = 2;
1211	+	private static final int PL_SIGNAL = 1;
1212	+	private static final int PL_SPINS = 1 << 8;
1213
1214		// access mode for WorkQueue
1215		static final int LIFO_QUEUE = 0;
1216		static final int FIFO_QUEUE = 1;
1217		static final int SHARED_QUEUE = -1;
1218
1219	+	// bounds for #steps in scan loop -- must be power 2 minus 1
1220	+	private static final int MIN_SCAN = 0x1ff; // cover estimation slop
1221	+	private static final int MAX_SCAN = 0x1ffff; // 4 * max workers
1222	+
1223		// Instance fields
1224
1225		/*
1226	<	* Field layout order in this class tends to matter more than one
1227	<	* would like. Runtime layout order is only loosely related to
1226	>	* Field layout of this class tends to matter more than one would
1227	>	* like. Runtime layout order is only loosely related to
1228		* declaration order and may differ across JVMs, but the following
1229		* empirically works OK on current JVMs.
1230		*/
1231
1232	+	// Heuristic padding to ameliorate unfortunate memory placements
1233	+	volatile long pad00, pad01, pad02, pad03, pad04, pad05, pad06;
1234	+
1235		volatile long stealCount; // collects worker counts
1236		volatile long ctl; // main pool control
1237	<	final int parallelism; // parallelism level
1238	<	final int localMode; // per-worker scheduling mode
1239	<	volatile int nextWorkerNumber; // to create worker name string
1263	<	final int submitMask; // submit queue index bound
1264	<	int nextSeed; // for initializing worker seeds
1265	<	volatile int mainLock; // spinlock for array updates
1266	<	volatile int runState; // shutdown status and seq
1237	>	volatile int plock; // shutdown status and seqLock
1238	>	volatile int indexSeed; // worker/submitter index seed
1239	>	final int config; // mode and parallelism level
1240		WorkQueue[] workQueues; // main registry
1241	<	final ForkJoinWorkerThreadFactory factory; // factory for new workers
1241	>	final ForkJoinWorkerThreadFactory factory;
1242		final Thread.UncaughtExceptionHandler ueh; // per-worker UEH
1243		final String workerNamePrefix; // to create worker name string
1244
1245	<	/*
1246	<	* Mechanics for main lock protecting worker array updates. Uses
1274	<	* the same strategy as ConcurrentHashMap bins -- a spinLock for
1275	<	* normal cases, but falling back to builtin lock when (rarely)
1276	<	* needed. See internal ConcurrentHashMap documentation for
1277	<	* explanation.
1278	<	*/
1245	>	volatile Object pad10, pad11, pad12, pad13, pad14, pad15, pad16, pad17;
1246	>	volatile Object pad18, pad19, pad1a, pad1b;
1247
1248	<	static final int LOCK_WAITING = 2; // bit to indicate need for signal
1249	<	static final int MAX_LOCK_SPINS = 1 << 8;
1250	<
1251	<	private void tryAwaitMainLock() {
1252	<	int spins = MAX_LOCK_SPINS, r = 0, h;
1253	<	while (((h = mainLock) & 1) != 0) {
1254	<	if (r == 0)
1255	<	r = ThreadLocalRandom.current().nextInt(); // randomize spins
1248	>	/*
1249	>	* Acquires the plock lock to protect worker array and related
1250	>	* updates. This method is called only if an initial CAS on plock
1251	>	* fails. This acts as a spinLock for normal cases, but falls back
1252	>	* to builtin monitor to block when (rarely) needed. This would be
1253	>	* a terrible idea for a highly contended lock, but works fine as
1254	>	* a more conservative alternative to a pure spinlock.
1255	>	*/
1256	>	private int acquirePlock() {
1257	>	int spins = PL_SPINS, r = 0, ps, nps;
1258	>	for (;;) {
1259	>	if (((ps = plock) & PL_LOCK) == 0 &&
1260	>	U.compareAndSwapInt(this, PLOCK, ps, nps = ps + PL_LOCK))
1261	>	return nps;
1262	>	else if (r == 0) { // randomize spins if possible
1263	>	Thread t = Thread.currentThread(); WorkQueue w; Submitter z;
1264	>	if ((t instanceof ForkJoinWorkerThread) &&
1265	>	(w = ((ForkJoinWorkerThread)t).workQueue) != null)
1266	>	r = w.seed;
1267	>	else if ((z = submitters.get()) != null)
1268	>	r = z.seed;
1269	>	else
1270	>	r = 1;
1271	>	}
1272		else if (spins >= 0) {
1273		r ^= r << 1; r ^= r >>> 3; r ^= r << 10; // xorshift
1274		if (r >= 0)
1275		--spins;
1276		}
1277	<	else if (U.compareAndSwapInt(this, MAINLOCK, h, h \| LOCK_WAITING)) {
1277	>	else if (U.compareAndSwapInt(this, PLOCK, ps, ps \| PL_SIGNAL)) {
1278		synchronized (this) {
1279	<	if ((mainLock & LOCK_WAITING) != 0) {
1279	>	if ((plock & PL_SIGNAL) != 0) {
1280		try {
1281		wait();
1282		} catch (InterruptedException ie) {
1283	<	Thread.currentThread().interrupt();
1283	>	try {
1284	>	Thread.currentThread().interrupt();
1285	>	} catch (SecurityException ignore) {
1286	>	}
1287		}
1288		}
1289		else
1290	<	notifyAll(); // possibly won race vs signaller
1290	>	notifyAll();
1291		}
1305	–	break;
1292		}
1293		}
1294		}
1295
1310	–	// Creating, registering, and deregistering workers
1311	–
1296		/**
1297	<	* Tries to create and start a worker
1297	>	* Unlocks and signals any thread waiting for plock. Called only
1298	>	* when CAS of seq value for unlock fails.
1299		*/
1300	<	private void addWorker() {
1301	<	Throwable ex = null;
1302	<	ForkJoinWorkerThread wt = null;
1318	<	try {
1319	<	if ((wt = factory.newThread(this)) != null) {
1320	<	wt.start();
1321	<	return;
1322	<	}
1323	<	} catch (Throwable e) {
1324	<	ex = e;
1325	<	}
1326	<	deregisterWorker(wt, ex); // adjust counts etc on failure
1300	>	private void releasePlock(int ps) {
1301	>	plock = ps;
1302	>	synchronized (this) { notifyAll(); }
1303		}
1304
1305		/**
1306	<	* Callback from ForkJoinWorkerThread constructor to assign a
1307	<	* public name. This must be separate from registerWorker because
1308	<	* it is called during the "super" constructor call in
1309	<	* ForkJoinWorkerThread.
1306	>	* Performs secondary initialization, called when plock is zero.
1307	>	* Creates workQueue array and sets plock to a valid value. The
1308	>	* lock body must be exception-free (so no try/finally) so we
1309	>	* optimistically allocate new array outside the lock and throw
1310	>	* away if (very rarely) not needed. (A similar tactic is used in
1311	>	* fullExternalPush.) Because the plock seq value can eventually
1312	>	* wrap around zero, this method harmlessly fails to reinitialize
1313	>	* if workQueues exists, while still advancing plock.
1314	>	*
1315	>	* Additionally tries to create the first worker.
1316		*/
1317	<	final String nextWorkerName() {
1318	<	int n;
1319	<	do {} while(!U.compareAndSwapInt(this, NEXTWORKERNUMBER,
1320	<	n = nextWorkerNumber, ++n));
1321	<	return workerNamePrefix.concat(Integer.toString(n));
1317	>	private void initWorkers() {
1318	>	WorkQueue[] ws, nws; int ps;
1319	>	int p = config & SMASK; // find power of two table size
1320	>	int n = (p > 1) ? p - 1 : 1; // ensure at least 2 slots
1321	>	n \|= n >>> 1; n \|= n >>> 2; n \|= n >>> 4; n \|= n >>> 8; n \|= n >>> 16;
1322	>	n = (n + 1) << 1;
1323	>	if ((ws = workQueues) == null \|\| ws.length == 0)
1324	>	nws = new WorkQueue[n];
1325	>	else
1326	>	nws = null;
1327	>	if (((ps = plock) & PL_LOCK) != 0 \|\|
1328	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
1329	>	ps = acquirePlock();
1330	>	if (((ws = workQueues) == null \|\| ws.length == 0) && nws != null)
1331	>	workQueues = nws;
1332	>	int nps = (ps & SHUTDOWN) \| ((ps + PL_LOCK) & ~SHUTDOWN);
1333	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
1334	>	releasePlock(nps);
1335	>	tryAddWorker();
1336		}
1337
1338		/**
1339	<	* Callback from ForkJoinWorkerThread constructor to establish its
1340	<	* poolIndex and record its WorkQueue. To avoid scanning bias due
1345	<	* to packing entries in front of the workQueues array, we treat
1346	<	* the array as a simple power-of-two hash table using per-thread
1347	<	* seed as hash, expanding as needed.
1348	<	*
1349	<	* @param w the worker's queue
1339	>	* Tries to create and start one worker if fewer than target
1340	>	* parallelism level exist. Adjusts counts etc on failure.
1341		*/
1342	<	final void registerWorker(WorkQueue w) {
1343	<	while (!U.compareAndSwapInt(this, MAINLOCK, 0, 1))
1344	<	tryAwaitMainLock();
1342	>	private void tryAddWorker() {
1343	>	long c; int u;
1344	>	while ((u = (int)((c = ctl) >>> 32)) < 0 &&
1345	>	(u & SHORT_SIGN) != 0 && (int)c == 0) {
1346	>	long nc = (long)(((u + UTC_UNIT) & UTC_MASK) \|
1347	>	((u + UAC_UNIT) & UAC_MASK)) << 32;
1348	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1349	>	ForkJoinWorkerThreadFactory fac;
1350	>	Throwable ex = null;
1351	>	ForkJoinWorkerThread wt = null;
1352	>	try {
1353	>	if ((fac = factory) != null &&
1354	>	(wt = fac.newThread(this)) != null) {
1355	>	wt.start();
1356	>	break;
1357	>	}
1358	>	} catch (Throwable e) {
1359	>	ex = e;
1360	>	}
1361	>	deregisterWorker(wt, ex);
1362	>	break;
1363	>	}
1364	>	}
1365	>	}
1366	>
1367	>	// Registering and deregistering workers
1368	>
1369	>	/**
1370	>	* Callback from ForkJoinWorkerThread to establish and record its
1371	>	* WorkQueue. To avoid scanning bias due to packing entries in
1372	>	* front of the workQueues array, we treat the array as a simple
1373	>	* power-of-two hash table using per-thread seed as hash,
1374	>	* expanding as needed.
1375	>	*
1376	>	* @param wt the worker thread
1377	>	* @return the worker's queue
1378	>	*/
1379	>	final WorkQueue registerWorker(ForkJoinWorkerThread wt) {
1380	>	Thread.UncaughtExceptionHandler handler; WorkQueue[] ws; int s, ps;
1381	>	wt.setDaemon(true);
1382	>	if ((handler = ueh) != null)
1383	>	wt.setUncaughtExceptionHandler(handler);
1384	>	do {} while (!U.compareAndSwapInt(this, INDEXSEED, s = indexSeed,
1385	>	s += SEED_INCREMENT) \|\|
1386	>	s == 0); // skip 0
1387	>	WorkQueue w = new WorkQueue(this, wt, config >>> 16, s);
1388	>	if (((ps = plock) & PL_LOCK) != 0 \|\|
1389	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
1390	>	ps = acquirePlock();
1391	>	int nps = (ps & SHUTDOWN) \| ((ps + PL_LOCK) & ~SHUTDOWN);
1392		try {
1393	<	WorkQueue[] ws;
1394	<	if ((ws = workQueues) == null)
1395	<	ws = workQueues = new WorkQueue[submitMask + 1];
1396	<	if (w != null) {
1397	<	int rs, n = ws.length, m = n - 1;
1398	<	int s = nextSeed += SEED_INCREMENT; // rarely-colliding sequence
1361	<	w.seed = (s == 0) ? 1 : s; // ensure non-zero seed
1362	<	int r = (s << 1) \| 1; // use odd-numbered indices
1363	<	if (ws[r &= m] != null) { // collision
1364	<	int probes = 0; // step by approx half size
1365	<	int step = (n <= 4) ? 2 : ((n >>> 1) & SQMASK) + 2;
1393	>	if ((ws = workQueues) != null) { // skip if shutting down
1394	>	int n = ws.length, m = n - 1;
1395	>	int r = (s << 1) \| 1; // use odd-numbered indices
1396	>	if (ws[r &= m] != null) { // collision
1397	>	int probes = 0; // step by approx half size
1398	>	int step = (n <= 4) ? 2 : ((n >>> 1) & EVENMASK) + 2;
1399		while (ws[r = (r + step) & m] != null) {
1400		if (++probes >= n) {
1401		workQueues = ws = Arrays.copyOf(ws, n <<= 1);
#	Line 1371 \| Line 1404 \| public class ForkJoinPool extends Abstra
1404		}
1405		}
1406		}
1407	<	w.eventCount = w.poolIndex = r; // establish before recording
1408	<	ws[r] = w; // also update seq
1376	<	runState = ((rs = runState) & SHUTDOWN) \| ((rs + 2) & ~SHUTDOWN);
1407	>	w.eventCount = w.poolIndex = r; // volatile write orders
1408	>	ws[r] = w;
1409		}
1410		} finally {
1411	<	if (!U.compareAndSwapInt(this, MAINLOCK, 1, 0)) {
1412	<	mainLock = 0;
1381	<	synchronized (this) { notifyAll(); };
1382	<	}
1411	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
1412	>	releasePlock(nps);
1413		}
1414	<
1414	>	wt.setName(workerNamePrefix.concat(Integer.toString(w.poolIndex)));
1415	>	return w;
1416		}
1417
1418		/**
1419		* Final callback from terminating worker, as well as upon failure
1420	<	* to construct or start a worker in addWorker. Removes record of
1421	<	* worker from array, and adjusts counts. If pool is shutting
1422	<	* down, tries to complete termination.
1420	>	* to construct or start a worker. Removes record of worker from
1421	>	* array, and adjusts counts. If pool is shutting down, tries to
1422	>	* complete termination.
1423		*
1424	<	* @param wt the worker thread or null if addWorker failed
1424	>	* @param wt the worker thread or null if construction failed
1425		* @param ex the exception causing failure, or null if none
1426		*/
1427		final void deregisterWorker(ForkJoinWorkerThread wt, Throwable ex) {
1428		WorkQueue w = null;
1429		if (wt != null && (w = wt.workQueue) != null) {
1430	<	w.runState = -1; // ensure runState is set
1431	<	long steals = w.totalSteals + w.nsteals, sc;
1432	<	do {} while(!U.compareAndSwapLong(this, STEALCOUNT,
1433	<	sc = stealCount, sc + steals));
1434	<	int idx = w.poolIndex;
1435	<	while (!U.compareAndSwapInt(this, MAINLOCK, 0, 1))
1436	<	tryAwaitMainLock();
1430	>	int ps;
1431	>	w.qlock = -1; // ensure set
1432	>	long ns = w.nsteals, sc; // collect steal count
1433	>	do {} while (!U.compareAndSwapLong(this, STEALCOUNT,
1434	>	sc = stealCount, sc + ns));
1435	>	if (((ps = plock) & PL_LOCK) != 0 \|\|
1436	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
1437	>	ps = acquirePlock();
1438	>	int nps = (ps & SHUTDOWN) \| ((ps + PL_LOCK) & ~SHUTDOWN);
1439		try {
1440	+	int idx = w.poolIndex;
1441		WorkQueue[] ws = workQueues;
1442		if (ws != null && idx >= 0 && idx < ws.length && ws[idx] == w)
1443		ws[idx] = null;
1444		} finally {
1445	<	if (!U.compareAndSwapInt(this, MAINLOCK, 1, 0)) {
1446	<	mainLock = 0;
1413	<	synchronized (this) { notifyAll(); };
1414	<	}
1445	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
1446	>	releasePlock(nps);
1447		}
1448		}
1449
1450	<	long c; // adjust ctl counts
1450	>	long c; // adjust ctl counts
1451		do {} while (!U.compareAndSwapLong
1452		(this, CTL, c = ctl, (((c - AC_UNIT) & AC_MASK) \|
1453		((c - TC_UNIT) & TC_MASK) \|
1454		(c & ~(AC_MASK\|TC_MASK)))));
1455
1456	<	if (!tryTerminate(false, false) && w != null) {
1457	<	w.cancelAll(); // cancel remaining tasks
1458	<	if (w.array != null) // suppress signal if never ran
1459	<	signalWork(); // wake up or create replacement
1460	<	if (ex == null) // help clean refs on way out
1461	<	ForkJoinTask.helpExpungeStaleExceptions();
1456	>	if (!tryTerminate(false, false) && w != null && w.array != null) {
1457	>	w.cancelAll(); // cancel remaining tasks
1458	>	WorkQueue[] ws; WorkQueue v; Thread p; int u, i, e;
1459	>	while ((u = (int)((c = ctl) >>> 32)) < 0 && (e = (int)c) >= 0) {
1460	>	if (e > 0) { // activate or create replacement
1461	>	if ((ws = workQueues) == null \|\|
1462	>	(i = e & SMASK) >= ws.length \|\|
1463	>	(v = ws[i]) == null)
1464	>	break;
1465	>	long nc = (((long)(v.nextWait & E_MASK)) \|
1466	>	((long)(u + UAC_UNIT) << 32));
1467	>	if (v.eventCount != (e \| INT_SIGN))
1468	>	break;
1469	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1470	>	v.eventCount = (e + E_SEQ) & E_MASK;
1471	>	if ((p = v.parker) != null)
1472	>	U.unpark(p);
1473	>	break;
1474	>	}
1475	>	}
1476	>	else {
1477	>	if ((short)u < 0)
1478	>	tryAddWorker();
1479	>	break;
1480	>	}
1481	>	}
1482		}
1483	<
1484	<	if (ex != null) // rethrow
1485	<	U.throwException(ex);
1483	>	if (ex == null) // help clean refs on way out
1484	>	ForkJoinTask.helpExpungeStaleExceptions();
1485	>	else // rethrow
1486	>	ForkJoinTask.rethrow(ex);
1487		}
1488
1489		// Submissions
#	Line 1438 \| Line 1491 \| public class ForkJoinPool extends Abstra
1491		/**
1492		* Unless shutting down, adds the given task to a submission queue
1493		* at submitter's current queue index (modulo submission
1494	<	* range). If no queue exists at the index, one is created. If
1495	<	* the queue is busy, another index is randomly chosen. The
1443	<	* submitMask bounds the effective number of queues to the
1444	<	* (nearest power of two for) parallelism level.
1494	>	* range). Only the most common path is directly handled in this
1495	>	* method. All others are relayed to fullExternalPush.
1496		*
1497		* @param task the task. Caller must ensure non-null.
1498		*/
1499	<	private void doSubmit(ForkJoinTask<?> task) {
1500	<	Submitter s = submitters.get();
1501	<	for (int r = s.seed, m = submitMask;;) {
1502	<	WorkQueue[] ws; WorkQueue q;
1503	<	int k = r & m & SQMASK; // use only even indices
1504	<	if (runState < 0)
1505	<	throw new RejectedExecutionException(); // shutting down
1506	<	else if ((ws = workQueues) == null \|\| ws.length <= k) {
1507	<	while (!U.compareAndSwapInt(this, MAINLOCK, 0, 1))
1508	<	tryAwaitMainLock();
1509	<	try {
1510	<	if (workQueues == null)
1511	<	workQueues = new WorkQueue[submitMask + 1];
1512	<	} finally {
1462	<	if (!U.compareAndSwapInt(this, MAINLOCK, 1, 0)) {
1463	<	mainLock = 0;
1464	<	synchronized (this) { notifyAll(); };
1465	<	}
1466	<	}
1467	<	}
1468	<	else if ((q = ws[k]) == null) { // create new queue
1469	<	WorkQueue nq = new WorkQueue(this, null, SHARED_QUEUE);
1470	<	while (!U.compareAndSwapInt(this, MAINLOCK, 0, 1))
1471	<	tryAwaitMainLock();
1472	<	try {
1473	<	int rs = runState; // to update seq
1474	<	if (ws == workQueues && ws[k] == null) {
1475	<	ws[k] = nq;
1476	<	runState = ((rs & SHUTDOWN) \| ((rs + 2) & ~SHUTDOWN));
1477	<	}
1478	<	} finally {
1479	<	if (!U.compareAndSwapInt(this, MAINLOCK, 1, 0)) {
1480	<	mainLock = 0;
1481	<	synchronized (this) { notifyAll(); };
1482	<	}
1483	<	}
1484	<	}
1485	<	else if (q.trySharedPush(task)) {
1486	<	signalWork();
1499	>	final void externalPush(ForkJoinTask<?> task) {
1500	>	WorkQueue[] ws; WorkQueue q; Submitter z; int m; ForkJoinTask<?>[] a;
1501	>	if ((z = submitters.get()) != null && plock > 0 &&
1502	>	(ws = workQueues) != null && (m = (ws.length - 1)) >= 0 &&
1503	>	(q = ws[m & z.seed & SQMASK]) != null &&
1504	>	U.compareAndSwapInt(q, QLOCK, 0, 1)) { // lock
1505	>	int b = q.base, s = q.top, n, an;
1506	>	if ((a = q.array) != null && (an = a.length) > (n = s + 1 - b)) {
1507	>	int j = (((an - 1) & s) << ASHIFT) + ABASE;
1508	>	U.putOrderedObject(a, j, task);
1509	>	q.top = s + 1; // push on to deque
1510	>	q.qlock = 0;
1511	>	if (n <= 2)
1512	>	signalWork(q);
1513		return;
1514		}
1515	<	else if (m > 1) { // move to a different index
1515	>	q.qlock = 0;
1516	>	}
1517	>	fullExternalPush(task);
1518	>	}
1519	>
1520	>	/**
1521	>	* Full version of externalPush. This method is called, among
1522	>	* other times, upon the first submission of the first task to the
1523	>	* pool, so must perform secondary initialization (via
1524	>	* initWorkers). It also detects first submission by an external
1525	>	* thread by looking up its ThreadLocal, and creates a new shared
1526	>	* queue if the one at index if empty or contended. The plock lock
1527	>	* body must be exception-free (so no try/finally) so we
1528	>	* optimistically allocate new queues outside the lock and throw
1529	>	* them away if (very rarely) not needed.
1530	>	*/
1531	>	private void fullExternalPush(ForkJoinTask<?> task) {
1532	>	int r = 0; // random index seed
1533	>	for (Submitter z = submitters.get();;) {
1534	>	WorkQueue[] ws; WorkQueue q; int ps, m, k;
1535	>	if (z == null) {
1536	>	if (U.compareAndSwapInt(this, INDEXSEED, r = indexSeed,
1537	>	r += SEED_INCREMENT) && r != 0)
1538	>	submitters.set(z = new Submitter(r));
1539	>	}
1540	>	else if (r == 0) { // move to a different index
1541	>	r = z.seed;
1542		r ^= r << 13; // same xorshift as WorkQueues
1543		r ^= r >>> 17;
1544	<	s.seed = r ^= r << 5;
1544	>	z.seed = r ^ (r << 5);
1545	>	}
1546	>	else if ((ps = plock) < 0)
1547	>	throw new RejectedExecutionException();
1548	>	else if (ps == 0 \|\| (ws = workQueues) == null \|\|
1549	>	(m = ws.length - 1) < 0)
1550	>	initWorkers();
1551	>	else if ((q = ws[k = r & m & SQMASK]) != null) {
1552	>	if (q.qlock == 0 && U.compareAndSwapInt(q, QLOCK, 0, 1)) {
1553	>	ForkJoinTask<?>[] a = q.array;
1554	>	int s = q.top;
1555	>	boolean submitted = false;
1556	>	try { // locked version of push
1557	>	if ((a != null && a.length > s + 1 - q.base) \|\|
1558	>	(a = q.growArray()) != null) { // must presize
1559	>	int j = (((a.length - 1) & s) << ASHIFT) + ABASE;
1560	>	U.putOrderedObject(a, j, task);
1561	>	q.top = s + 1;
1562	>	submitted = true;
1563	>	}
1564	>	} finally {
1565	>	q.qlock = 0; // unlock
1566	>	}
1567	>	if (submitted) {
1568	>	signalWork(q);
1569	>	return;
1570	>	}
1571	>	}
1572	>	r = 0; // move on failure
1573	>	}
1574	>	else if (((ps = plock) & PL_LOCK) == 0) { // create new queue
1575	>	q = new WorkQueue(this, null, SHARED_QUEUE, r);
1576	>	if (((ps = plock) & PL_LOCK) != 0 \|\|
1577	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
1578	>	ps = acquirePlock();
1579	>	if ((ws = workQueues) != null && k < ws.length && ws[k] == null)
1580	>	ws[k] = q;
1581	>	int nps = (ps & SHUTDOWN) \| ((ps + PL_LOCK) & ~SHUTDOWN);
1582	>	if (!U.compareAndSwapInt(this, PLOCK, ps, nps))
1583	>	releasePlock(nps);
1584		}
1585		else
1586	<	Thread.yield(); // yield if no alternatives
1586	>	r = 0; // try elsewhere while lock held
1587		}
1588		}
1589
1499	–	/**
1500	–	* Submits the given (non-null) task to the common pool, if possible.
1501	–	*/
1502	–	static void submitToCommonPool(ForkJoinTask<?> task) {
1503	–	ForkJoinPool p;
1504	–	if ((p = commonPool) == null)
1505	–	throw new RejectedExecutionException("Common Pool Unavailable");
1506	–	p.doSubmit(task);
1507	–	}
1508	–
1509	–	/**
1510	–	* Returns true if the given task was submitted to common pool
1511	–	* and has not yet commenced execution, and is available for
1512	–	* removal according to execution policies; if so removing the
1513	–	* submission from the pool.
1514	–	*
1515	–	* @param task the task
1516	–	* @return true if successful
1517	–	*/
1518	–	static boolean tryUnsubmitFromCommonPool(ForkJoinTask<?> task) {
1519	–	// Peek, looking for task and eligibility before
1520	–	// using trySharedUnpush to actually take it under lock
1521	–	ForkJoinPool p; WorkQueue[] ws; WorkQueue q;
1522	–	ForkJoinTask<?>[] a; int t, s, n;
1523	–	int k = submitters.get().seed & SQMASK;
1524	–	return ((p = commonPool) != null &&
1525	–	(ws = p.workQueues) != null &&
1526	–	ws.length > (k &= p.submitMask) &&
1527	–	(q = ws[k]) != null &&
1528	–	(a = q.array) != null &&
1529	–	(n = (t = q.top) - q.base) > 0 &&
1530	–	(n > 1 \|\| (int)(p.ctl >> AC_SHIFT) < 0) &&
1531	–	(s = t - 1) >= 0 && s < a.length && a[s] == task &&
1532	–	q.trySharedUnpush(task));
1533	–	}
1534	–
1590		// Maintaining ctl counts
1591
1592		/**
#	Line 1543 \| Line 1598 \| public class ForkJoinPool extends Abstra
1598		}
1599
1600		/**
1601	<	* Tries to create one or activate one or more workers if too few are active.
1601	>	* Tries to create or activate a worker if too few are active.
1602	>	*
1603	>	* @param q the (non-null) queue holding tasks to be signalled
1604		*/
1605	<	final void signalWork() {
1606	<	long c; int u;
1607	<	while ((u = (int)((c = ctl) >>> 32)) < 0) { // too few active
1608	<	WorkQueue[] ws = workQueues; int e, i; WorkQueue w; Thread p;
1609	<	if ((e = (int)c) > 0) { // at least one waiting
1610	<	if (ws != null && (i = e & SMASK) < ws.length &&
1605	>	final void signalWork(WorkQueue q) {
1606	>	int hint = q.poolIndex;
1607	>	long c; int e, u, i, n; WorkQueue[] ws; WorkQueue w; Thread p;
1608	>	while ((u = (int)((c = ctl) >>> 32)) < 0) {
1609	>	if ((e = (int)c) > 0) {
1610	>	if ((ws = workQueues) != null && ws.length > (i = e & SMASK) &&
1611		(w = ws[i]) != null && w.eventCount == (e \| INT_SIGN)) {
1612		long nc = (((long)(w.nextWait & E_MASK)) \|
1613		((long)(u + UAC_UNIT) << 32));
1614		if (U.compareAndSwapLong(this, CTL, c, nc)) {
1615	+	w.hint = hint;
1616		w.eventCount = (e + E_SEQ) & E_MASK;
1617		if ((p = w.parker) != null)
1618	<	U.unpark(p); // activate and release
1618	>	U.unpark(p);
1619		break;
1620		}
1621	+	if (q.top - q.base <= 0)
1622	+	break;
1623		}
1624		else
1625		break;
1626		}
1627	<	else if (e == 0 && (u & SHORT_SIGN) != 0) { // too few total
1628	<	long nc = (long)(((u + UTC_UNIT) & UTC_MASK) \|
1629	<	((u + UAC_UNIT) & UAC_MASK)) << 32;
1570	<	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1571	<	addWorker();
1572	<	break;
1573	<	}
1574	<	}
1575	<	else
1627	>	else {
1628	>	if ((short)u < 0)
1629	>	tryAddWorker();
1630		break;
1631	+	}
1632		}
1633		}
1634
#	Line 1583 \| Line 1638 \| public class ForkJoinPool extends Abstra
1638		* Top-level runloop for workers, called by ForkJoinWorkerThread.run.
1639		*/
1640		final void runWorker(WorkQueue w) {
1641	<	w.growArray(false); // initialize queue array in this thread
1642	<	do { w.runTask(scan(w)); } while (w.runState >= 0);
1641	>	w.growArray(); // allocate queue
1642	>	do { w.runTask(scan(w)); } while (w.qlock >= 0);
1643		}
1644
1645		/**
#	Line 1595 \| Line 1650 \| public class ForkJoinPool extends Abstra
1650		* contention, or state changes that indicate possible success on
1651		* re-invocation.
1652		*
1653	<	* The scan searches for tasks across a random permutation of
1654	<	* queues (starting at a random index and stepping by a random
1655	<	* relative prime, checking each at least once). The scan
1656	<	* terminates upon either finding a non-empty queue, or completing
1657	<	* the sweep. If the worker is not inactivated, it takes and
1658	<	* returns a task from this queue. On failure to find a task, we
1659	<	* take one of the following actions, after which the caller will
1660	<	* retry calling this method unless terminated.
1653	>	* The scan searches for tasks across queues (starting at a random
1654	>	* index, and relying on registerWorker to irregularly scatter
1655	>	* them within array to avoid bias), checking each at least twice.
1656	>	* The scan terminates upon either finding a non-empty queue, or
1657	>	* completing the sweep. If the worker is not inactivated, it
1658	>	* takes and returns a task from this queue. Otherwise, if not
1659	>	* activated, it signals workers (that may include itself) and
1660	>	* returns so caller can retry. Also returns for true if the
1661	>	* worker array may have changed during an empty scan. On failure
1662	>	* to find a task, we take one of the following actions, after
1663	>	* which the caller will retry calling this method unless
1664	>	* terminated.
1665		*
1666		* * If pool is terminating, terminate the worker.
1667		*
1609	–	* * If not a complete sweep, try to release a waiting worker. If
1610	–	* the scan terminated because the worker is inactivated, then the
1611	–	* released worker will often be the calling worker, and it can
1612	–	* succeed obtaining a task on the next call. Or maybe it is
1613	–	* another worker, but with same net effect. Releasing in other
1614	–	* cases as well ensures that we have enough workers running.
1615	–	*
1668		* * If not already enqueued, try to inactivate and enqueue the
1669		* worker on wait queue. Or, if inactivating has caused the pool
1670	<	* to be quiescent, relay to idleAwaitWork to check for
1671	<	* termination and possibly shrink pool.
1670	>	* to be quiescent, relay to idleAwaitWork to possibly shrink
1671	>	* pool.
1672		*
1673	<	* * If already inactive, and the caller has run a task since the
1674	<	* last empty scan, return (to allow rescan) unless others are
1623	<	* also inactivated. Field WorkQueue.rescans counts down on each
1624	<	* scan to ensure eventual inactivation and blocking.
1673	>	* * If already enqueued and none of the above apply, possibly
1674	>	* park awaiting signal, else lingering to help scan and signal.
1675		*
1676	<	* * If already enqueued and none of the above apply, park
1677	<	* awaiting signal,
1676	>	* * If a non-empty queue discovered or left as a hint,
1677	>	* help wake up other workers before return
1678		*
1679		* @param w the worker (via its WorkQueue)
1680		* @return a task or null if none found
1681		*/
1682		private final ForkJoinTask<?> scan(WorkQueue w) {
1683	<	WorkQueue[] ws; // first update random seed
1684	<	int r = w.seed; r ^= r << 13; r ^= r >>> 17; w.seed = r ^= r << 5;
1685	<	int rs = runState, m; // volatile read order matters
1686	<	if ((ws = workQueues) != null && (m = ws.length - 1) > 0) {
1687	<	int ec = w.eventCount; // ec is negative if inactive
1688	<	int step = (r >>> 16) \| 1; // relative prime
1689	<	for (int j = (m + 1) << 2; ; r += step) {
1690	<	WorkQueue q; ForkJoinTask<?> t; ForkJoinTask<?>[] a; int b;
1691	<	if ((q = ws[r & m]) != null && (b = q.base) - q.top < 0 &&
1692	<	(a = q.array) != null) { // probably nonempty
1683	>	WorkQueue[] ws; int m;
1684	>	int ps = plock; // read plock before ws
1685	>	if (w != null && (ws = workQueues) != null && (m = ws.length - 1) >= 0) {
1686	>	int ec = w.eventCount; // ec is negative if inactive
1687	>	int r = w.seed; r ^= r << 13; r ^= r >>> 17; w.seed = r ^= r << 5;
1688	>	w.hint = -1; // update seed and clear hint
1689	>	int j = ((m + m + 1) \| MIN_SCAN) & MAX_SCAN;
1690	>	do {
1691	>	WorkQueue q; ForkJoinTask<?>[] a; int b;
1692	>	if ((q = ws[(r + j) & m]) != null && (b = q.base) - q.top < 0 &&
1693	>	(a = q.array) != null) { // probably nonempty
1694		int i = (((a.length - 1) & b) << ASHIFT) + ABASE;
1695	<	t = (ForkJoinTask<?>)U.getObjectVolatile(a, i);
1695	>	ForkJoinTask<?> t = (ForkJoinTask<?>)
1696	>	U.getObjectVolatile(a, i);
1697		if (q.base == b && ec >= 0 && t != null &&
1698		U.compareAndSwapObject(a, i, t, null)) {
1699	<	if (q.top - (q.base = b + 1) > 0)
1700	<	signalWork(); // help pushes signal
1701	<	return t;
1702	<	}
1703	<	else if (ec < 0 \|\| j <= m) {
1704	<	rs = 0; // mark scan as imcomplete
1705	<	break; // caller can retry after release
1706	<	}
1707	<	}
1708	<	if (--j < 0)
1709	<	break;
1710	<	}
1711	<
1712	<	long c = ctl; int e = (int)c, a = (int)(c >> AC_SHIFT), nr, ns;
1713	<	if (e < 0) // decode ctl on empty scan
1714	<	w.runState = -1; // pool is terminating
1715	<	else if (rs == 0 \|\| rs != runState) { // incomplete scan
1716	<	WorkQueue v; Thread p; // try to release a waiter
1717	<	if (e > 0 && a < 0 && w.eventCount == ec &&
1718	<	(v = ws[e & m]) != null && v.eventCount == (e \| INT_SIGN)) {
1719	<	long nc = ((long)(v.nextWait & E_MASK) \|
1720	<	((c + AC_UNIT) & (AC_MASK\|TC_MASK)));
1721	<	if (ctl == c && U.compareAndSwapLong(this, CTL, c, nc)) {
1699	>	if ((q.base = b + 1) - q.top < 0)
1700	>	signalWork(q);
1701	>	return t; // taken
1702	>	}
1703	>	else if ((ec < 0 \|\| j < m) && (int)(ctl >> AC_SHIFT) <= 0) {
1704	>	w.hint = (r + j) & m; // help signal below
1705	>	break; // cannot take
1706	>	}
1707	>	}
1708	>	} while (--j >= 0);
1709	>
1710	>	int h, e, ns; long c, sc; WorkQueue q;
1711	>	if ((ns = w.nsteals) != 0) {
1712	>	if (U.compareAndSwapLong(this, STEALCOUNT,
1713	>	sc = stealCount, sc + ns))
1714	>	w.nsteals = 0; // collect steals and rescan
1715	>	}
1716	>	else if (plock != ps) // consistency check
1717	>	; // skip
1718	>	else if ((e = (int)(c = ctl)) < 0)
1719	>	w.qlock = -1; // pool is terminating
1720	>	else {
1721	>	if ((h = w.hint) < 0) {
1722	>	if (ec >= 0) { // try to enqueue/inactivate
1723	>	long nc = (((long)ec \|
1724	>	((c - AC_UNIT) & (AC_MASK\|TC_MASK))));
1725	>	w.nextWait = e; // link and mark inactive
1726	>	w.eventCount = ec \| INT_SIGN;
1727	>	if (ctl != c \|\| !U.compareAndSwapLong(this, CTL, c, nc))
1728	>	w.eventCount = ec; // unmark on CAS failure
1729	>	else if ((int)(c >> AC_SHIFT) == 1 - (config & SMASK))
1730	>	idleAwaitWork(w, nc, c);
1731	>	}
1732	>	else if (w.eventCount < 0 && !tryTerminate(false, false) &&
1733	>	ctl == c) { // block
1734	>	Thread wt = Thread.currentThread();
1735	>	Thread.interrupted(); // clear status
1736	>	U.putObject(wt, PARKBLOCKER, this);
1737	>	w.parker = wt; // emulate LockSupport.park
1738	>	if (w.eventCount < 0) // recheck
1739	>	U.park(false, 0L);
1740	>	w.parker = null;
1741	>	U.putObject(wt, PARKBLOCKER, null);
1742	>	}
1743	>	}
1744	>	if ((h >= 0 \|\| (h = w.hint) >= 0) &&
1745	>	(ws = workQueues) != null && h < ws.length &&
1746	>	(q = ws[h]) != null) { // signal others before retry
1747	>	WorkQueue v; Thread p; int u, i, s;
1748	>	for (int n = (config & SMASK) >>> 1;;) {
1749	>	int idleCount = (w.eventCount < 0) ? 0 : -1;
1750	>	if (((s = idleCount - q.base + q.top) <= n &&
1751	>	(n = s) <= 0) \|\|
1752	>	(u = (int)((c = ctl) >>> 32)) >= 0 \|\|
1753	>	(e = (int)c) <= 0 \|\| m < (i = e & SMASK) \|\|
1754	>	(v = ws[i]) == null)
1755	>	break;
1756	>	long nc = (((long)(v.nextWait & E_MASK)) \|
1757	>	((long)(u + UAC_UNIT) << 32));
1758	>	if (v.eventCount != (e \| INT_SIGN) \|\|
1759	>	!U.compareAndSwapLong(this, CTL, c, nc))
1760	>	break;
1761	>	v.hint = h;
1762		v.eventCount = (e + E_SEQ) & E_MASK;
1763		if ((p = v.parker) != null)
1764		U.unpark(p);
1765	+	if (--n <= 0)
1766	+	break;
1767		}
1768		}
1769		}
1676	–	else if (ec >= 0) { // try to enqueue/inactivate
1677	–	long nc = (long)ec \| ((c - AC_UNIT) & (AC_MASK\|TC_MASK));
1678	–	w.nextWait = e;
1679	–	w.eventCount = ec \| INT_SIGN; // mark as inactive
1680	–	if (ctl != c \|\| !U.compareAndSwapLong(this, CTL, c, nc))
1681	–	w.eventCount = ec; // unmark on CAS failure
1682	–	else {
1683	–	if ((ns = w.nsteals) != 0) {
1684	–	w.nsteals = 0; // set rescans if ran task
1685	–	w.rescans = (a > 0) ? 0 : a + parallelism;
1686	–	w.totalSteals += ns;
1687	–	}
1688	–	if (a == 1 - parallelism) // quiescent
1689	–	idleAwaitWork(w, nc, c);
1690	–	}
1691	–	}
1692	–	else if (w.eventCount < 0) { // already queued
1693	–	int ac = a + parallelism;
1694	–	if ((nr = w.rescans) > 0) // continue rescanning
1695	–	w.rescans = (ac < nr) ? ac : nr - 1;
1696	–	else if (((w.seed >>> 16) & ac) == 0) { // randomize park
1697	–	Thread.interrupted(); // clear status
1698	–	Thread wt = Thread.currentThread();
1699	–	U.putObject(wt, PARKBLOCKER, this);
1700	–	w.parker = wt; // emulate LockSupport.park
1701	–	if (w.eventCount < 0) // recheck
1702	–	U.park(false, 0L);
1703	–	w.parker = null;
1704	–	U.putObject(wt, PARKBLOCKER, null);
1705	–	}
1706	–	}
1770		}
1771		return null;
1772		}
#	Line 1721 \| Line 1784 \| public class ForkJoinPool extends Abstra
1784		* @param prevCtl the ctl value to restore if thread is terminated
1785		*/
1786		private void idleAwaitWork(WorkQueue w, long currentCtl, long prevCtl) {
1787	<	if (w.eventCount < 0 && !tryTerminate(false, false) &&
1788	<	(int)prevCtl != 0 && !hasQueuedSubmissions() && ctl == currentCtl) {
1787	>	if (w != null && w.eventCount < 0 &&
1788	>	!tryTerminate(false, false) && (int)prevCtl != 0) {
1789		int dc = -(short)(currentCtl >>> TC_SHIFT);
1790		long parkTime = dc < 0 ? FAST_IDLE_TIMEOUT: (dc + 1) * IDLE_TIMEOUT;
1791	<	long deadline = System.nanoTime() + parkTime - 100000L; // 1ms slop
1791	>	long deadline = System.nanoTime() + parkTime - TIMEOUT_SLOP;
1792		Thread wt = Thread.currentThread();
1793		while (ctl == currentCtl) {
1794		Thread.interrupted(); // timed variant of version in scan()
#	Line 1740 \| Line 1803 \| public class ForkJoinPool extends Abstra
1803		if (deadline - System.nanoTime() <= 0L &&
1804		U.compareAndSwapLong(this, CTL, currentCtl, prevCtl)) {
1805		w.eventCount = (w.eventCount + E_SEQ) \| E_MASK;
1806	<	w.runState = -1; // shrink
1806	>	w.qlock = -1; // shrink
1807		break;
1808		}
1809		}
#	Line 1748 \| Line 1811 \| public class ForkJoinPool extends Abstra
1811		}
1812
1813		/**
1814	+	* Scans through queues looking for work while joining a task; if
1815	+	* any present, signals. May return early if more signalling is
1816	+	* detectably unneeded.
1817	+	*
1818	+	* @param task return early if done
1819	+	* @param origin an index to start scan
1820	+	*/
1821	+	private void helpSignal(ForkJoinTask<?> task, int origin) {
1822	+	WorkQueue[] ws; WorkQueue w; Thread p; long c; int m, u, e, i, s;
1823	+	if (task != null && task.status >= 0 &&
1824	+	(u = (int)(ctl >>> 32)) < 0 && (u >> UAC_SHIFT) < 0 &&
1825	+	(ws = workQueues) != null && (m = ws.length - 1) >= 0) {
1826	+	outer: for (int k = origin, j = m; j >= 0; --j) {
1827	+	WorkQueue q = ws[k++ & m];
1828	+	for (int n = m;;) { // limit to at most m signals
1829	+	if (task.status < 0)
1830	+	break outer;
1831	+	if (q == null \|\|
1832	+	((s = -q.base + q.top) <= n && (n = s) <= 0))
1833	+	break;
1834	+	if ((u = (int)((c = ctl) >>> 32)) >= 0 \|\|
1835	+	(e = (int)c) <= 0 \|\| m < (i = e & SMASK) \|\|
1836	+	(w = ws[i]) == null)
1837	+	break outer;
1838	+	long nc = (((long)(w.nextWait & E_MASK)) \|
1839	+	((long)(u + UAC_UNIT) << 32));
1840	+	if (w.eventCount != (e \| INT_SIGN))
1841	+	break outer;
1842	+	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1843	+	w.eventCount = (e + E_SEQ) & E_MASK;
1844	+	if ((p = w.parker) != null)
1845	+	U.unpark(p);
1846	+	if (--n <= 0)
1847	+	break;
1848	+	}
1849	+	}
1850	+	}
1851	+	}
1852	+	}
1853	+
1854	+	/**
1855		* Tries to locate and execute tasks for a stealer of the given
1856		* task, or in turn one of its stealers, Traces currentSteal ->
1857		* currentJoin links looking for a thread working on a descendant
#	Line 1778 \| Line 1882 \| public class ForkJoinPool extends Abstra
1882		}
1883		if ((ws = workQueues) == null \|\| (m = ws.length - 1) <= 0)
1884		break restart; // shutting down
1885	<	if ((v = ws[h = (j.stealHint \| 1) & m]) == null \|\|
1885	>	if ((v = ws[h = (j.hint \| 1) & m]) == null \|\|
1886		v.currentSteal != subtask) {
1887		for (int origin = h;;) { // find stealer
1888		if (((h = (h + 2) & m) & 15) == 1 &&
#	Line 1786 \| Line 1890 \| public class ForkJoinPool extends Abstra
1890		continue restart; // occasional staleness check
1891		if ((v = ws[h]) != null &&
1892		v.currentSteal == subtask) {
1893	<	j.stealHint = h; // save hint
1893	>	j.hint = h; // save hint
1894		break;
1895		}
1896		if (h == origin)
#	Line 1834 \| Line 1938 \| public class ForkJoinPool extends Abstra
1938		}
1939
1940		/**
1941	<	* If task is at base of some steal queue, steals and executes it.
1941	>	* Analog of tryHelpStealer for CountedCompleters. Tries to steal
1942	>	* and run tasks within the target's computation.
1943	>	*
1944	>	* @param task the task to join
1945	>	* @param mode if shared, exit upon completing any task
1946	>	* if all workers are active
1947		*
1839	–	* @param joiner the joining worker
1840	–	* @param task the task
1948		*/
1949	<	private void tryPollForAndExec(WorkQueue joiner, ForkJoinTask<?> task) {
1950	<	WorkQueue[] ws;
1951	<	if ((ws = workQueues) != null) {
1952	<	for (int j = 1; j < ws.length && task.status >= 0; j += 2) {
1953	<	WorkQueue q = ws[j];
1954	<	if (q != null && q.pollFor(task)) {
1955	<	joiner.runSubtask(task);
1956	<	break;
1949	>	private int helpComplete(ForkJoinTask<?> task, int mode) {
1950	>	WorkQueue[] ws; WorkQueue q; int m, n, s, u;
1951	>	if (task != null && (ws = workQueues) != null &&
1952	>	(m = ws.length - 1) >= 0) {
1953	>	for (int j = 1, origin = j;;) {
1954	>	if ((s = task.status) < 0)
1955	>	return s;
1956	>	if ((q = ws[j & m]) != null && q.pollAndExecCC(task)) {
1957	>	origin = j;
1958	>	if (mode == SHARED_QUEUE &&
1959	>	((u = (int)(ctl >>> 32)) >= 0 \|\| (u >> UAC_SHIFT) >= 0))
1960	>	break;
1961		}
1962	+	else if ((j = (j + 2) & m) == origin)
1963	+	break;
1964		}
1965		}
1966	+	return 0;
1967		}
1968
1969		/**
1970		* Tries to decrement active count (sometimes implicitly) and
1971		* possibly release or create a compensating worker in preparation
1972		* for blocking. Fails on contention or termination. Otherwise,
1973	<	* adds a new thread if no idle workers are available and either
1974	<	* pool would become completely starved or: (at least half
1861	<	* starved, and fewer than 50% spares exist, and there is at least
1862	<	* one task apparently available). Even though the availability
1863	<	* check requires a full scan, it is worthwhile in reducing false
1864	<	* alarms.
1865	<	*
1866	<	* @param task if non-null, a task being waited for
1867	<	* @param blocker if non-null, a blocker being waited for
1868	<	* @return true if the caller can block, else should recheck and retry
1973	>	* adds a new thread if no idle workers are available and pool
1974	>	* may become starved.
1975		*/
1976	<	final boolean tryCompensate(ForkJoinTask<?> task, ManagedBlocker blocker) {
1977	<	int pc = parallelism, e;
1978	<	long c = ctl;
1979	<	WorkQueue[] ws = workQueues;
1980	<	if ((e = (int)c) >= 0 && ws != null) {
1981	<	int u, a, ac, hc;
1982	<	int tc = (short)((u = (int)(c >>> 32)) >>> UTC_SHIFT) + pc;
1983	<	boolean replace = false;
1984	<	if ((a = u >> UAC_SHIFT) <= 0) {
1985	<	if ((ac = a + pc) <= 1)
1986	<	replace = true;
1987	<	else if ((e > 0 \|\| (task != null &&
1988	<	ac <= (hc = pc >>> 1) && tc < pc + hc))) {
1883	<	WorkQueue w;
1884	<	for (int j = 0; j < ws.length; ++j) {
1885	<	if ((w = ws[j]) != null && !w.isEmpty()) {
1886	<	replace = true;
1887	<	break; // in compensation range and tasks available
1888	<	}
1889	<	}
1976	>	final boolean tryCompensate() {
1977	>	int pc = config & SMASK, e, i, tc; long c;
1978	>	WorkQueue[] ws; WorkQueue w; Thread p;
1979	>	if ((ws = workQueues) != null && (e = (int)(c = ctl)) >= 0) {
1980	>	if (e != 0 && (i = e & SMASK) < ws.length &&
1981	>	(w = ws[i]) != null && w.eventCount == (e \| INT_SIGN)) {
1982	>	long nc = ((long)(w.nextWait & E_MASK) \|
1983	>	(c & (AC_MASK\|TC_MASK)));
1984	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1985	>	w.eventCount = (e + E_SEQ) & E_MASK;
1986	>	if ((p = w.parker) != null)
1987	>	U.unpark(p);
1988	>	return true; // replace with idle worker
1989		}
1990		}
1991	<	if ((task == null \|\| task.status >= 0) && // recheck need to block
1992	<	(blocker == null \|\| !blocker.isReleasable()) && ctl == c) {
1993	<	if (!replace) { // no compensation
1994	<	long nc = ((c - AC_UNIT) & AC_MASK) \| (c & ~AC_MASK);
1995	<	if (U.compareAndSwapLong(this, CTL, c, nc))
1996	<	return true;
1997	<	}
1998	<	else if (e != 0) { // release an idle worker
1999	<	WorkQueue w; Thread p; int i;
2000	<	if ((i = e & SMASK) < ws.length && (w = ws[i]) != null) {
2001	<	long nc = ((long)(w.nextWait & E_MASK) \|
2002	<	(c & (AC_MASK\|TC_MASK)));
2003	<	if (w.eventCount == (e \| INT_SIGN) &&
2004	<	U.compareAndSwapLong(this, CTL, c, nc)) {
2005	<	w.eventCount = (e + E_SEQ) & E_MASK;
2006	<	if ((p = w.parker) != null)
1908	<	U.unpark(p);
1991	>	else if ((tc = (short)(c >>> TC_SHIFT)) >= 0 &&
1992	>	(int)(c >> AC_SHIFT) + pc > 1) {
1993	>	long nc = ((c - AC_UNIT) & AC_MASK) \| (c & ~AC_MASK);
1994	>	if (U.compareAndSwapLong(this, CTL, c, nc))
1995	>	return true; // no compensation
1996	>	}
1997	>	else if (tc + pc < MAX_CAP) {
1998	>	long nc = ((c + TC_UNIT) & TC_MASK) \| (c & ~TC_MASK);
1999	>	if (U.compareAndSwapLong(this, CTL, c, nc)) {
2000	>	ForkJoinWorkerThreadFactory fac;
2001	>	Throwable ex = null;
2002	>	ForkJoinWorkerThread wt = null;
2003	>	try {
2004	>	if ((fac = factory) != null &&
2005	>	(wt = fac.newThread(this)) != null) {
2006	>	wt.start();
2007		return true;
2008		}
2009	+	} catch (Throwable rex) {
2010	+	ex = rex;
2011		}
2012	<	}
1913	<	else if (tc < MAX_CAP) { // create replacement
1914	<	long nc = ((c + TC_UNIT) & TC_MASK) \| (c & ~TC_MASK);
1915	<	if (U.compareAndSwapLong(this, CTL, c, nc)) {
1916	<	addWorker();
1917	<	return true;
1918	<	}
2012	>	deregisterWorker(wt, ex); // clean up and return false
2013		}
2014		}
2015		}
#	Line 1930 \| Line 2024 \| public class ForkJoinPool extends Abstra
2024		* @return task status on exit
2025		*/
2026		final int awaitJoin(WorkQueue joiner, ForkJoinTask<?> task) {
2027	<	int s;
2028	<	if ((s = task.status) >= 0) {
2027	>	int s = 0;
2028	>	if (joiner != null && task != null && (s = task.status) >= 0) {
2029		ForkJoinTask<?> prevJoin = joiner.currentJoin;
2030		joiner.currentJoin = task;
2031	<	long startTime = 0L;
2032	<	for (int k = 0;;) {
2033	<	if ((s = (joiner.isEmpty() ? // try to help
2034	<	tryHelpStealer(joiner, task) :
2035	<	joiner.tryRemoveAndExec(task))) == 0 &&
2031	>	do {} while ((s = task.status) >= 0 && !joiner.isEmpty() &&
2032	>	joiner.tryRemoveAndExec(task)); // process local tasks
2033	>	if (s >= 0 && (s = task.status) >= 0) {
2034	>	helpSignal(task, joiner.poolIndex);
2035	>	if ((s = task.status) >= 0 &&
2036	>	(task instanceof CountedCompleter))
2037	>	s = helpComplete(task, LIFO_QUEUE);
2038	>	}
2039	>	while (s >= 0 && (s = task.status) >= 0) {
2040	>	if ((!joiner.isEmpty() \|\| // try helping
2041	>	(s = tryHelpStealer(joiner, task)) == 0) &&
2042		(s = task.status) >= 0) {
2043	<	if (k == 0) {
2044	<	startTime = System.nanoTime();
2045	<	tryPollForAndExec(joiner, task); // check uncommon case
1946	<	}
1947	<	else if ((k & (MAX_HELP - 1)) == 0 &&
1948	<	System.nanoTime() - startTime >=
1949	<	COMPENSATION_DELAY &&
1950	<	tryCompensate(task, null)) {
1951	<	if (task.trySetSignal()) {
2043	>	helpSignal(task, joiner.poolIndex);
2044	>	if ((s = task.status) >= 0 && tryCompensate()) {
2045	>	if (task.trySetSignal() && (s = task.status) >= 0) {
2046		synchronized (task) {
2047		if (task.status >= 0) {
2048		try { // see ForkJoinTask
#	Line 1965 \| Line 2059 \| public class ForkJoinPool extends Abstra
2059		(this, CTL, c = ctl, c + AC_UNIT));
2060		}
2061		}
1968	–	if (s < 0 \|\| (s = task.status) < 0) {
1969	–	joiner.currentJoin = prevJoin;
1970	–	break;
1971	–	}
1972	–	else if ((k++ & (MAX_HELP - 1)) == MAX_HELP >>> 1)
1973	–	Thread.yield(); // for politeness
2062		}
2063	+	joiner.currentJoin = prevJoin;
2064		}
2065		return s;
2066		}
#	Line 1983 \| Line 2072 \| public class ForkJoinPool extends Abstra
2072		*
2073		* @param joiner the joining worker
2074		* @param task the task
1986	–	* @return task status on exit
2075		*/
2076	<	final int helpJoinOnce(WorkQueue joiner, ForkJoinTask<?> task) {
2076	>	final void helpJoinOnce(WorkQueue joiner, ForkJoinTask<?> task) {
2077		int s;
2078	<	while ((s = task.status) >= 0 &&
2079	<	(joiner.isEmpty() ?
2080	<	tryHelpStealer(joiner, task) :
2081	<	joiner.tryRemoveAndExec(task)) != 0)
2082	<	;
2083	<	return s;
2078	>	if (joiner != null && task != null && (s = task.status) >= 0) {
2079	>	ForkJoinTask<?> prevJoin = joiner.currentJoin;
2080	>	joiner.currentJoin = task;
2081	>	do {} while ((s = task.status) >= 0 && !joiner.isEmpty() &&
2082	>	joiner.tryRemoveAndExec(task));
2083	>	if (s >= 0 && (s = task.status) >= 0) {
2084	>	helpSignal(task, joiner.poolIndex);
2085	>	if ((s = task.status) >= 0 &&
2086	>	(task instanceof CountedCompleter))
2087	>	s = helpComplete(task, LIFO_QUEUE);
2088	>	}
2089	>	if (s >= 0 && joiner.isEmpty()) {
2090	>	do {} while (task.status >= 0 &&
2091	>	tryHelpStealer(joiner, task) > 0);
2092	>	}
2093	>	joiner.currentJoin = prevJoin;
2094	>	}
2095		}
2096
2097		/**
#	Line 2000 \| Line 2099 \| public class ForkJoinPool extends Abstra
2099		* during a random, then cyclic scan, else null. This method must
2100		* be retried by caller if, by the time it tries to use the queue,
2101		* it is empty.
2102	+	* @param r a (random) seed for scanning
2103		*/
2104	<	private WorkQueue findNonEmptyStealQueue(WorkQueue w) {
2005	<	// Similar to loop in scan(), but ignoring submissions
2006	<	int r = w.seed; r ^= r << 13; r ^= r >>> 17; w.seed = r ^= r << 5;
2007	<	int step = (r >>> 16) \| 1;
2104	>	private WorkQueue findNonEmptyStealQueue(int r) {
2105		for (WorkQueue[] ws;;) {
2106	<	int rs = runState, m;
2106	>	int ps = plock, m;
2107		if ((ws = workQueues) == null \|\| (m = ws.length - 1) < 1)
2108		return null;
2109	<	for (int j = (m + 1) << 2; ; r += step) {
2110	<	WorkQueue q = ws[((r << 1) \| 1) & m];
2111	<	if (q != null && !q.isEmpty())
2109	>	for (int j = (m + 1) << 2; ;) {
2110	>	WorkQueue q = ws[(((r + j) << 1) \| 1) & m];
2111	>	if (q != null && q.base - q.top < 0)
2112		return q;
2113		else if (--j < 0) {
2114	<	if (runState == rs)
2114	>	if (plock == ps)
2115		return null;
2116		break;
2117		}
#	Line 2033 \| Line 2130 \| public class ForkJoinPool extends Abstra
2130		ForkJoinTask<?> localTask; // exhaust local queue
2131		while ((localTask = w.nextLocalTask()) != null)
2132		localTask.doExec();
2133	<	WorkQueue q = findNonEmptyStealQueue(w);
2133	>	// Similar to loop in scan(), but ignoring submissions
2134	>	WorkQueue q = findNonEmptyStealQueue(w.nextSeed());
2135		if (q != null) {
2136		ForkJoinTask<?> t; int b;
2137		if (!active) { // re-establish active count
#	Line 2042 \| Line 2140 \| public class ForkJoinPool extends Abstra
2140		do {} while (!U.compareAndSwapLong
2141		(this, CTL, c = ctl, c + AC_UNIT));
2142		}
2143	<	if ((b = q.base) - q.top < 0 && (t = q.pollAt(b)) != null)
2143	>	if ((b = q.base) - q.top < 0 && (t = q.pollAt(b)) != null) {
2144	>	if (q.base - q.top < 0)
2145	>	signalWork(q);
2146		w.runSubtask(t);
2147	+	}
2148		}
2149		else {
2150		long c;
#	Line 2054 \| Line 2155 \| public class ForkJoinPool extends Abstra
2155		}
2156		else
2157		c = ctl; // re-increment on exit
2158	<	if ((int)(c >> AC_SHIFT) + parallelism == 0) {
2158	>	if ((int)(c >> AC_SHIFT) + (config & SMASK) == 0) {
2159		do {} while (!U.compareAndSwapLong
2160		(this, CTL, c = ctl, c + AC_UNIT));
2161		break;
#	Line 2064 \| Line 2165 \| public class ForkJoinPool extends Abstra
2165		}
2166
2167		/**
2067	–	* Restricted version of helpQuiescePool for non-FJ callers
2068	–	*/
2069	–	static void externalHelpQuiescePool() {
2070	–	ForkJoinPool p; WorkQueue[] ws; WorkQueue w, q;
2071	–	ForkJoinTask<?> t; int b;
2072	–	int k = submitters.get().seed & SQMASK;
2073	–	if ((p = commonPool) != null &&
2074	–	(ws = p.workQueues) != null &&
2075	–	ws.length > (k &= p.submitMask) &&
2076	–	(w = ws[k]) != null &&
2077	–	(q = p.findNonEmptyStealQueue(w)) != null &&
2078	–	(b = q.base) - q.top < 0 &&
2079	–	(t = q.pollAt(b)) != null)
2080	–	t.doExec();
2081	–	}
2082	–
2083	–	/**
2168		* Gets and removes a local or stolen task for the given worker.
2169		*
2170		* @return a task, if available
#	Line 2090 \| Line 2174 \| public class ForkJoinPool extends Abstra
2174		WorkQueue q; int b;
2175		if ((t = w.nextLocalTask()) != null)
2176		return t;
2177	<	if ((q = findNonEmptyStealQueue(w)) == null)
2177	>	if ((q = findNonEmptyStealQueue(w.nextSeed())) == null)
2178		return null;
2179	<	if ((b = q.base) - q.top < 0 && (t = q.pollAt(b)) != null)
2179	>	if ((b = q.base) - q.top < 0 && (t = q.pollAt(b)) != null) {
2180	>	if (q.base - q.top < 0)
2181	>	signalWork(q);
2182		return t;
2183	+	}
2184		}
2185		}
2186
2187		/**
2188	<	* Returns the approximate (non-atomic) number of idle threads per
2189	<	* active thread to offset steal queue size for method
2190	<	* ForkJoinTask.getSurplusQueuedTaskCount().
2191	<	*/
2192	<	final int idlePerActive() {
2193	<	// Approximate at powers of two for small values, saturate past 4
2194	<	int p = parallelism;
2195	<	int a = p + (int)(ctl >> AC_SHIFT);
2196	<	return (a > (p >>>= 1) ? 0 :
2197	<	a > (p >>>= 1) ? 1 :
2198	<	a > (p >>>= 1) ? 2 :
2199	<	a > (p >>>= 1) ? 4 :
2200	<	8);
2201	<	}
2202	<
2203	<	/**
2204	<	* Returns approximate submission queue length for the given caller
2205	<	*/
2206	<	static int getEstimatedSubmitterQueueLength() {
2207	<	ForkJoinPool p; WorkQueue[] ws; WorkQueue q;
2208	<	int k = submitters.get().seed & SQMASK;
2209	<	return ((p = commonPool) != null &&
2210	<	p.runState >= 0 &&
2211	<	(ws = p.workQueues) != null &&
2212	<	ws.length > (k &= p.submitMask) &&
2213	<	(q = ws[k]) != null) ?
2214	<	q.queueSize() : 0;
2188	>	* Returns a cheap heuristic guide for task partitioning when
2189	>	* programmers, frameworks, tools, or languages have little or no
2190	>	* idea about task granularity. In essence by offering this
2191	>	* method, we ask users only about tradeoffs in overhead vs
2192	>	* expected throughput and its variance, rather than how finely to
2193	>	* partition tasks.
2194	>	*
2195	>	* In a steady state strict (tree-structured) computation, each
2196	>	* thread makes available for stealing enough tasks for other
2197	>	* threads to remain active. Inductively, if all threads play by
2198	>	* the same rules, each thread should make available only a
2199	>	* constant number of tasks.
2200	>	*
2201	>	* The minimum useful constant is just 1. But using a value of 1
2202	>	* would require immediate replenishment upon each steal to
2203	>	* maintain enough tasks, which is infeasible. Further,
2204	>	* partitionings/granularities of offered tasks should minimize
2205	>	* steal rates, which in general means that threads nearer the top
2206	>	* of computation tree should generate more than those nearer the
2207	>	* bottom. In perfect steady state, each thread is at
2208	>	* approximately the same level of computation tree. However,
2209	>	* producing extra tasks amortizes the uncertainty of progress and
2210	>	* diffusion assumptions.
2211	>	*
2212	>	* So, users will want to use values larger, but not much larger
2213	>	* than 1 to both smooth over transient shortages and hedge
2214	>	* against uneven progress; as traded off against the cost of
2215	>	* extra task overhead. We leave the user to pick a threshold
2216	>	* value to compare with the results of this call to guide
2217	>	* decisions, but recommend values such as 3.
2218	>	*
2219	>	* When all threads are active, it is on average OK to estimate
2220	>	* surplus strictly locally. In steady-state, if one thread is
2221	>	* maintaining say 2 surplus tasks, then so are others. So we can
2222	>	* just use estimated queue length. However, this strategy alone
2223	>	* leads to serious mis-estimates in some non-steady-state
2224	>	* conditions (ramp-up, ramp-down, other stalls). We can detect
2225	>	* many of these by further considering the number of "idle"
2226	>	* threads, that are known to have zero queued tasks, so
2227	>	* compensate by a factor of (#idle/#active) threads.
2228	>	*
2229	>	* Note: The approximation of #busy workers as #active workers is
2230	>	* not very good under current signalling scheme, and should be
2231	>	* improved.
2232	>	*/
2233	>	static int getSurplusQueuedTaskCount() {
2234	>	Thread t; ForkJoinWorkerThread wt; ForkJoinPool pool; WorkQueue q;
2235	>	if (((t = Thread.currentThread()) instanceof ForkJoinWorkerThread)) {
2236	>	int p = (pool = (wt = (ForkJoinWorkerThread)t).pool).config & SMASK;
2237	>	int n = (q = wt.workQueue).top - q.base;
2238	>	int a = (int)(pool.ctl >> AC_SHIFT) + p;
2239	>	return n - (a > (p >>>= 1) ? 0 :
2240	>	a > (p >>>= 1) ? 1 :
2241	>	a > (p >>>= 1) ? 2 :
2242	>	a > (p >>>= 1) ? 4 :
2243	>	8);
2244	>	}
2245	>	return 0;
2246		}
2247
2248		// Termination
#	Line 2144 \| Line 2262 \| public class ForkJoinPool extends Abstra
2262		* @return true if now terminating or terminated
2263		*/
2264		private boolean tryTerminate(boolean now, boolean enable) {
2265	+	if (this == commonPool) // cannot shut down
2266	+	return false;
2267		for (long c;;) {
2268		if (((c = ctl) & STOP_BIT) != 0) { // already terminating
2269	<	if ((short)(c >>> TC_SHIFT) == -parallelism) {
2270	<	synchronized(this) {
2269	>	if ((short)(c >>> TC_SHIFT) == -(config & SMASK)) {
2270	>	synchronized (this) {
2271		notifyAll(); // signal when 0 workers
2272		}
2273		}
2274		return true;
2275		}
2276	<	if (runState >= 0) { // not yet enabled
2276	>	if (plock >= 0) { // not yet enabled
2277	>	int ps;
2278		if (!enable)
2279		return false;
2280	<	while (!U.compareAndSwapInt(this, MAINLOCK, 0, 1))
2281	<	tryAwaitMainLock();
2282	<	try {
2283	<	runState \|= SHUTDOWN;
2284	<	} finally {
2164	<	if (!U.compareAndSwapInt(this, MAINLOCK, 1, 0)) {
2165	<	mainLock = 0;
2166	<	synchronized (this) { notifyAll(); };
2167	<	}
2168	<	}
2280	>	if (((ps = plock) & PL_LOCK) != 0 \|\|
2281	>	!U.compareAndSwapInt(this, PLOCK, ps, ps += PL_LOCK))
2282	>	ps = acquirePlock();
2283	>	if (!U.compareAndSwapInt(this, PLOCK, ps, SHUTDOWN))
2284	>	releasePlock(SHUTDOWN);
2285		}
2286		if (!now) { // check if idle & no tasks
2287	<	if ((int)(c >> AC_SHIFT) != -parallelism \|\|
2287	>	if ((int)(c >> AC_SHIFT) != -(config & SMASK) \|\|
2288		hasQueuedSubmissions())
2289		return false;
2290		// Check for unqueued inactive workers. One pass suffices.
#	Line 2184 \| Line 2300 \| public class ForkJoinPool extends Abstra
2300		for (int pass = 0; pass < 3; ++pass) {
2301		WorkQueue[] ws = workQueues;
2302		if (ws != null) {
2303	<	WorkQueue w;
2303	>	WorkQueue w; Thread wt;
2304		int n = ws.length;
2305		for (int i = 0; i < n; ++i) {
2306		if ((w = ws[i]) != null) {
2307	<	w.runState = -1;
2307	>	w.qlock = -1;
2308		if (pass > 0) {
2309		w.cancelAll();
2310	<	if (pass > 1)
2311	<	w.interruptOwner();
2310	>	if (pass > 1 && (wt = w.owner) != null) {
2311	>	if (!wt.isInterrupted()) {
2312	>	try {
2313	>	wt.interrupt();
2314	>	} catch (SecurityException ignore) {
2315	>	}
2316	>	}
2317	>	U.unpark(wt);
2318	>	}
2319		}
2320		}
2321		}
#	Line 2207 \| Line 2330 \| public class ForkJoinPool extends Abstra
2330		if (w.eventCount == (e \| INT_SIGN) &&
2331		U.compareAndSwapLong(this, CTL, cc, nc)) {
2332		w.eventCount = (e + E_SEQ) & E_MASK;
2333	<	w.runState = -1;
2333	>	w.qlock = -1;
2334		if ((p = w.parker) != null)
2335		U.unpark(p);
2336		}
#	Line 2218 \| Line 2341 \| public class ForkJoinPool extends Abstra
2341		}
2342		}
2343
2344	+	// external operations on common pool
2345	+
2346	+	/**
2347	+	* Returns common pool queue for a thread that has submitted at
2348	+	* least one task.
2349	+	*/
2350	+	static WorkQueue commonSubmitterQueue() {
2351	+	ForkJoinPool p; WorkQueue[] ws; int m; Submitter z;
2352	+	return ((z = submitters.get()) != null &&
2353	+	(p = commonPool) != null &&
2354	+	(ws = p.workQueues) != null &&
2355	+	(m = ws.length - 1) >= 0) ?
2356	+	ws[m & z.seed & SQMASK] : null;
2357	+	}
2358	+
2359	+	/**
2360	+	* Tries to pop the given task from submitter's queue in common pool.
2361	+	*/
2362	+	static boolean tryExternalUnpush(ForkJoinTask<?> t) {
2363	+	ForkJoinPool p; WorkQueue[] ws; WorkQueue q; Submitter z;
2364	+	ForkJoinTask<?>[] a; int m, s;
2365	+	if (t != null &&
2366	+	(z = submitters.get()) != null &&
2367	+	(p = commonPool) != null &&
2368	+	(ws = p.workQueues) != null &&
2369	+	(m = ws.length - 1) >= 0 &&
2370	+	(q = ws[m & z.seed & SQMASK]) != null &&
2371	+	(s = q.top) != q.base &&
2372	+	(a = q.array) != null) {
2373	+	long j = (((a.length - 1) & (s - 1)) << ASHIFT) + ABASE;
2374	+	if (U.getObject(a, j) == t &&
2375	+	U.compareAndSwapInt(q, QLOCK, 0, 1)) {
2376	+	if (q.array == a && q.top == s && // recheck
2377	+	U.compareAndSwapObject(a, j, t, null)) {
2378	+	q.top = s - 1;
2379	+	q.qlock = 0;
2380	+	return true;
2381	+	}
2382	+	q.qlock = 0;
2383	+	}
2384	+	}
2385	+	return false;
2386	+	}
2387	+
2388	+	/**
2389	+	* Tries to pop and run local tasks within the same computation
2390	+	* as the given root. On failure, tries to help complete from
2391	+	* other queues via helpComplete.
2392	+	*/
2393	+	private void externalHelpComplete(WorkQueue q, ForkJoinTask<?> root) {
2394	+	ForkJoinTask<?>[] a; int m;
2395	+	if (q != null && (a = q.array) != null && (m = (a.length - 1)) >= 0 &&
2396	+	root != null && root.status >= 0) {
2397	+	for (;;) {
2398	+	int s, u; Object o; CountedCompleter<?> task = null;
2399	+	if ((s = q.top) - q.base > 0) {
2400	+	long j = ((m & (s - 1)) << ASHIFT) + ABASE;
2401	+	if ((o = U.getObject(a, j)) != null &&
2402	+	(o instanceof CountedCompleter)) {
2403	+	CountedCompleter<?> t = (CountedCompleter<?>)o, r = t;
2404	+	do {
2405	+	if (r == root) {
2406	+	if (U.compareAndSwapInt(q, QLOCK, 0, 1)) {
2407	+	if (q.array == a && q.top == s &&
2408	+	U.compareAndSwapObject(a, j, t, null)) {
2409	+	q.top = s - 1;
2410	+	task = t;
2411	+	}
2412	+	q.qlock = 0;
2413	+	}
2414	+	break;
2415	+	}
2416	+	} while ((r = r.completer) != null);
2417	+	}
2418	+	}
2419	+	if (task != null)
2420	+	task.doExec();
2421	+	if (root.status < 0 \|\|
2422	+	(u = (int)(ctl >>> 32)) >= 0 \|\| (u >> UAC_SHIFT) >= 0)
2423	+	break;
2424	+	if (task == null) {
2425	+	helpSignal(root, q.poolIndex);
2426	+	if (root.status >= 0)
2427	+	helpComplete(root, SHARED_QUEUE);
2428	+	break;
2429	+	}
2430	+	}
2431	+	}
2432	+	}
2433	+
2434	+	/**
2435	+	* Tries to help execute or signal availability of the given task
2436	+	* from submitter's queue in common pool.
2437	+	*/
2438	+	static void externalHelpJoin(ForkJoinTask<?> t) {
2439	+	// Some hard-to-avoid overlap with tryExternalUnpush
2440	+	ForkJoinPool p; WorkQueue[] ws; WorkQueue q, w; Submitter z;
2441	+	ForkJoinTask<?>[] a; int m, s, n;
2442	+	if (t != null &&
2443	+	(z = submitters.get()) != null &&
2444	+	(p = commonPool) != null &&
2445	+	(ws = p.workQueues) != null &&
2446	+	(m = ws.length - 1) >= 0 &&
2447	+	(q = ws[m & z.seed & SQMASK]) != null &&
2448	+	(a = q.array) != null) {
2449	+	int am = a.length - 1;
2450	+	if ((s = q.top) != q.base) {
2451	+	long j = ((am & (s - 1)) << ASHIFT) + ABASE;
2452	+	if (U.getObject(a, j) == t &&
2453	+	U.compareAndSwapInt(q, QLOCK, 0, 1)) {
2454	+	if (q.array == a && q.top == s &&
2455	+	U.compareAndSwapObject(a, j, t, null)) {
2456	+	q.top = s - 1;
2457	+	q.qlock = 0;
2458	+	t.doExec();
2459	+	}
2460	+	else
2461	+	q.qlock = 0;
2462	+	}
2463	+	}
2464	+	if (t.status >= 0) {
2465	+	if (t instanceof CountedCompleter)
2466	+	p.externalHelpComplete(q, t);
2467	+	else
2468	+	p.helpSignal(t, q.poolIndex);
2469	+	}
2470	+	}
2471	+	}
2472	+
2473	+	/**
2474	+	* Restricted version of helpQuiescePool for external callers
2475	+	*/
2476	+	static void externalHelpQuiescePool() {
2477	+	ForkJoinPool p; ForkJoinTask<?> t; WorkQueue q; int b;
2478	+	if ((p = commonPool) != null &&
2479	+	(q = p.findNonEmptyStealQueue(1)) != null &&
2480	+	(b = q.base) - q.top < 0 &&
2481	+	(t = q.pollAt(b)) != null) {
2482	+	if (q.base - q.top < 0)
2483	+	p.signalWork(q);
2484	+	t.doExec();
2485	+	}
2486	+	}
2487	+
2488		// Exported methods
2489
2490		// Constructors
#	Line 2289 \| Line 2556 \| public class ForkJoinPool extends Abstra
2556		throw new NullPointerException();
2557		if (parallelism <= 0 \|\| parallelism > MAX_CAP)
2558		throw new IllegalArgumentException();
2292	–	this.parallelism = parallelism;
2559		this.factory = factory;
2560		this.ueh = handler;
2561	<	this.localMode = asyncMode ? FIFO_QUEUE : LIFO_QUEUE;
2561	>	this.config = parallelism \| (asyncMode ? (FIFO_QUEUE << 16) : 0);
2562		long np = (long)(-parallelism); // offset ctl counts
2563		this.ctl = ((np << AC_SHIFT) & AC_MASK) \| ((np << TC_SHIFT) & TC_MASK);
2564	<	// Use nearest power 2 for workQueues size. See Hackers Delight sec 3.2.
2299	<	int n = parallelism - 1;
2300	<	n \|= n >>> 1; n \|= n >>> 2; n \|= n >>> 4; n \|= n >>> 8; n \|= n >>> 16;
2301	<	this.submitMask = ((n + 1) << 1) - 1;
2302	<	int pn = poolNumberGenerator.incrementAndGet();
2564	>	int pn = nextPoolId();
2565		StringBuilder sb = new StringBuilder("ForkJoinPool-");
2566		sb.append(Integer.toString(pn));
2567		sb.append("-worker-");
2568		this.workerNamePrefix = sb.toString();
2307	–	this.runState = 1; // set init flag
2569		}
2570
2571		/**
2572		* Constructor for common pool, suitable only for static initialization.
2573		* Basically the same as above, but uses smallest possible initial footprint.
2574		*/
2575	<	ForkJoinPool(int parallelism, int submitMask,
2575	>	ForkJoinPool(int parallelism, long ctl,
2576		ForkJoinWorkerThreadFactory factory,
2577		Thread.UncaughtExceptionHandler handler) {
2578	+	this.config = parallelism;
2579	+	this.ctl = ctl;
2580		this.factory = factory;
2581		this.ueh = handler;
2319	–	this.submitMask = submitMask;
2320	–	this.parallelism = parallelism;
2321	–	long np = (long)(-parallelism);
2322	–	this.ctl = ((np << AC_SHIFT) & AC_MASK) \| ((np << TC_SHIFT) & TC_MASK);
2323	–	this.localMode = LIFO_QUEUE;
2582		this.workerNamePrefix = "ForkJoinPool.commonPool-worker-";
2325	–	this.runState = 1;
2583		}
2584
2585		/**
2586	<	* Returns the common pool instance.
2586	>	* Returns the common pool instance. This pool is statically
2587	>	* constructed; its run state is unaffected by attempts to
2588	>	* {@link #shutdown} or {@link #shutdownNow}.
2589		*
2590		* @return the common pool instance
2591		*/
2592		public static ForkJoinPool commonPool() {
2593	<	ForkJoinPool p;
2594	<	if ((p = commonPool) == null)
2336	<	throw new Error("Common Pool Unavailable");
2337	<	return p;
2593	>	// assert commonPool != null : "static init error";
2594	>	return commonPool;
2595		}
2596
2597		// Execution methods
#	Line 2358 \| Line 2615 \| public class ForkJoinPool extends Abstra
2615		public <T> T invoke(ForkJoinTask<T> task) {
2616		if (task == null)
2617		throw new NullPointerException();
2618	<	doSubmit(task);
2618	>	externalPush(task);
2619		return task.join();
2620		}
2621
#	Line 2373 \| Line 2630 \| public class ForkJoinPool extends Abstra
2630		public void execute(ForkJoinTask<?> task) {
2631		if (task == null)
2632		throw new NullPointerException();
2633	<	doSubmit(task);
2633	>	externalPush(task);
2634		}
2635
2636		// AbstractExecutorService methods
#	Line 2391 \| Line 2648 \| public class ForkJoinPool extends Abstra
2648		job = (ForkJoinTask<?>) task;
2649		else
2650		job = new ForkJoinTask.AdaptedRunnableAction(task);
2651	<	doSubmit(job);
2651	>	externalPush(job);
2652		}
2653
2654		/**
#	Line 2406 \| Line 2663 \| public class ForkJoinPool extends Abstra
2663		public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) {
2664		if (task == null)
2665		throw new NullPointerException();
2666	<	doSubmit(task);
2666	>	externalPush(task);
2667		return task;
2668		}
2669
#	Line 2417 \| Line 2674 \| public class ForkJoinPool extends Abstra
2674		*/
2675		public <T> ForkJoinTask<T> submit(Callable<T> task) {
2676		ForkJoinTask<T> job = new ForkJoinTask.AdaptedCallable<T>(task);
2677	<	doSubmit(job);
2677	>	externalPush(job);
2678		return job;
2679		}
2680
#	Line 2428 \| Line 2685 \| public class ForkJoinPool extends Abstra
2685		*/
2686		public <T> ForkJoinTask<T> submit(Runnable task, T result) {
2687		ForkJoinTask<T> job = new ForkJoinTask.AdaptedRunnable<T>(task, result);
2688	<	doSubmit(job);
2688	>	externalPush(job);
2689		return job;
2690		}
2691
#	Line 2445 \| Line 2702 \| public class ForkJoinPool extends Abstra
2702		job = (ForkJoinTask<?>) task;
2703		else
2704		job = new ForkJoinTask.AdaptedRunnableAction(task);
2705	<	doSubmit(job);
2705	>	externalPush(job);
2706		return job;
2707		}
2708
#	Line 2467 \| Line 2724 \| public class ForkJoinPool extends Abstra
2724		try {
2725		for (Callable<T> t : tasks) {
2726		ForkJoinTask<T> f = new ForkJoinTask.AdaptedCallable<T>(t);
2727	<	doSubmit(f);
2727	>	externalPush(f);
2728		fs.add(f);
2729		}
2730		for (ForkJoinTask<T> f : fs)
#	Line 2506 \| Line 2763 \| public class ForkJoinPool extends Abstra
2763		* @return the targeted parallelism level of this pool
2764		*/
2765		public int getParallelism() {
2766	<	return parallelism;
2766	>	return config & SMASK;
2767		}
2768
2769		/**
#	Line 2527 \| Line 2784 \| public class ForkJoinPool extends Abstra
2784		* @return the number of worker threads
2785		*/
2786		public int getPoolSize() {
2787	<	return parallelism + (short)(ctl >>> TC_SHIFT);
2787	>	return (config & SMASK) + (short)(ctl >>> TC_SHIFT);
2788		}
2789
2790		/**
#	Line 2537 \| Line 2794 \| public class ForkJoinPool extends Abstra
2794		* @return {@code true} if this pool uses async mode
2795		*/
2796		public boolean getAsyncMode() {
2797	<	return localMode != 0;
2797	>	return (config >>> 16) == FIFO_QUEUE;
2798		}
2799
2800		/**
#	Line 2568 \| Line 2825 \| public class ForkJoinPool extends Abstra
2825		* @return the number of active threads
2826		*/
2827		public int getActiveThreadCount() {
2828	<	int r = parallelism + (int)(ctl >> AC_SHIFT);
2828	>	int r = (config & SMASK) + (int)(ctl >> AC_SHIFT);
2829		return (r <= 0) ? 0 : r; // suppress momentarily negative values
2830		}
2831
#	Line 2584 \| Line 2841 \| public class ForkJoinPool extends Abstra
2841		* @return {@code true} if all threads are currently idle
2842		*/
2843		public boolean isQuiescent() {
2844	<	return (int)(ctl >> AC_SHIFT) + parallelism == 0;
2844	>	return (int)(ctl >> AC_SHIFT) + (config & SMASK) == 0;
2845		}
2846
2847		/**
#	Line 2604 \| Line 2861 \| public class ForkJoinPool extends Abstra
2861		if ((ws = workQueues) != null) {
2862		for (int i = 1; i < ws.length; i += 2) {
2863		if ((w = ws[i]) != null)
2864	<	count += w.totalSteals;
2864	>	count += w.nsteals;
2865		}
2866		}
2867		return count;
#	Line 2740 \| Line 2997 \| public class ForkJoinPool extends Abstra
2997		qs += size;
2998		else {
2999		qt += size;
3000	<	st += w.totalSteals;
3000	>	st += w.nsteals;
3001		if (w.isApparentlyUnblocked())
3002		++rc;
3003		}
3004		}
3005		}
3006		}
3007	<	int pc = parallelism;
3007	>	int pc = (config & SMASK);
3008		int tc = pc + (short)(c >>> TC_SHIFT);
3009		int ac = pc + (int)(c >> AC_SHIFT);
3010		if (ac < 0) // ignore transient negative
#	Line 2756 \| Line 3013 \| public class ForkJoinPool extends Abstra
3013		if ((c & STOP_BIT) != 0)
3014		level = (tc == 0) ? "Terminated" : "Terminating";
3015		else
3016	<	level = runState < 0 ? "Shutting down" : "Running";
3016	>	level = plock < 0 ? "Shutting down" : "Running";
3017		return super.toString() +
3018		"[" + level +
3019		", parallelism = " + pc +
#	Line 2785 \| Line 3042 \| public class ForkJoinPool extends Abstra
3042		*/
3043		public void shutdown() {
3044		checkPermission();
3045	<	if (this != commonPool)
2789	<	tryTerminate(false, true);
3045	>	tryTerminate(false, true);
3046		}
3047
3048		/**
#	Line 2809 \| Line 3065 \| public class ForkJoinPool extends Abstra
3065		*/
3066		public List<Runnable> shutdownNow() {
3067		checkPermission();
3068	<	if (this != commonPool)
2813	<	tryTerminate(true, true);
3068	>	tryTerminate(true, true);
3069		return Collections.emptyList();
3070		}
3071
#	Line 2822 \| Line 3077 \| public class ForkJoinPool extends Abstra
3077		public boolean isTerminated() {
3078		long c = ctl;
3079		return ((c & STOP_BIT) != 0L &&
3080	<	(short)(c >>> TC_SHIFT) == -parallelism);
3080	>	(short)(c >>> TC_SHIFT) == -(config & SMASK));
3081		}
3082
3083		/**
#	Line 2830 \| Line 3085 \| public class ForkJoinPool extends Abstra
3085		* commenced but not yet completed. This method may be useful for
3086		* debugging. A return of {@code true} reported a sufficient
3087		* period after shutdown may indicate that submitted tasks have
3088	<	* ignored or suppressed interruption, or are waiting for IO,
3088	>	* ignored or suppressed interruption, or are waiting for I/O,
3089		* causing this executor not to properly terminate. (See the
3090		* advisory notes for class {@link ForkJoinTask} stating that
3091		* tasks should not normally entail blocking operations. But if
#	Line 2841 \| Line 3096 \| public class ForkJoinPool extends Abstra
3096		public boolean isTerminating() {
3097		long c = ctl;
3098		return ((c & STOP_BIT) != 0L &&
3099	<	(short)(c >>> TC_SHIFT) != -parallelism);
3099	>	(short)(c >>> TC_SHIFT) != -(config & SMASK));
3100		}
3101
3102		/**
#	Line 2850 \| Line 3105 \| public class ForkJoinPool extends Abstra
3105		* @return {@code true} if this pool has been shut down
3106		*/
3107		public boolean isShutdown() {
3108	<	return runState < 0;
3108	>	return plock < 0;
3109		}
3110
3111		/**
3112	<	* Blocks until all tasks have completed execution after a shutdown
3113	<	* request, or the timeout occurs, or the current thread is
3114	<	* interrupted, whichever happens first.
3112	>	* Blocks until all tasks have completed execution after a
3113	>	* shutdown request, or the timeout occurs, or the current thread
3114	>	* is interrupted, whichever happens first. Note that the {@link
3115	>	* #commonPool()} never terminates until program shutdown so
3116	>	* this method will always time out.
3117		*
3118		* @param timeout the maximum time to wait
3119		* @param unit the time unit of the timeout argument
#	Line 2871 \| Line 3128 \| public class ForkJoinPool extends Abstra
3128		return true;
3129		long startTime = System.nanoTime();
3130		boolean terminated = false;
3131	<	synchronized(this) {
3131	>	synchronized (this) {
3132		for (long waitTime = nanos, millis = 0L;;) {
3133		if (terminated = isTerminated() \|\|
3134		waitTime <= 0L \|\|
#	Line 2980 \| Line 3237 \| public class ForkJoinPool extends Abstra
3237		public static void managedBlock(ManagedBlocker blocker)
3238		throws InterruptedException {
3239		Thread t = Thread.currentThread();
3240	<	ForkJoinPool p = ((t instanceof ForkJoinWorkerThread) ?
3241	<	((ForkJoinWorkerThread)t).pool : null);
3242	<	while (!blocker.isReleasable()) {
3243	<	if (p == null \|\| p.tryCompensate(null, blocker)) {
3244	<	try {
3245	<	do {} while (!blocker.isReleasable() && !blocker.block());
3246	<	} finally {
3247	<	if (p != null)
3240	>	if (t instanceof ForkJoinWorkerThread) {
3241	>	ForkJoinPool p = ((ForkJoinWorkerThread)t).pool;
3242	>	while (!blocker.isReleasable()) { // variant of helpSignal
3243	>	WorkQueue[] ws; WorkQueue q; int m, u;
3244	>	if ((ws = p.workQueues) != null && (m = ws.length - 1) >= 0) {
3245	>	for (int i = 0; i <= m; ++i) {
3246	>	if (blocker.isReleasable())
3247	>	return;
3248	>	if ((q = ws[i]) != null && q.base - q.top < 0) {
3249	>	p.signalWork(q);
3250	>	if ((u = (int)(p.ctl >>> 32)) >= 0 \|\|
3251	>	(u >> UAC_SHIFT) >= 0)
3252	>	break;
3253	>	}
3254	>	}
3255	>	}
3256	>	if (p.tryCompensate()) {
3257	>	try {
3258	>	do {} while (!blocker.isReleasable() &&
3259	>	!blocker.block());
3260	>	} finally {
3261		p.incrementActiveCount();
3262	+	}
3263	+	break;
3264		}
2993	–	break;
3265		}
3266		}
3267	+	else {
3268	+	do {} while (!blocker.isReleasable() &&
3269	+	!blocker.block());
3270	+	}
3271		}
3272
3273		// AbstractExecutorService overrides. These rely on undocumented
#	Line 3013 \| Line 3288 \| public class ForkJoinPool extends Abstra
3288		private static final long PARKBLOCKER;
3289		private static final int ABASE;
3290		private static final int ASHIFT;
3016	–	private static final long NEXTWORKERNUMBER;
3291		private static final long STEALCOUNT;
3292	<	private static final long MAINLOCK;
3292	>	private static final long PLOCK;
3293	>	private static final long INDEXSEED;
3294	>	private static final long QLOCK;
3295
3296		static {
3297	<	poolNumberGenerator = new AtomicInteger();
3022	<	nextSubmitterSeed = new AtomicInteger(0x55555555);
3023	<	modifyThreadPermission = new RuntimePermission("modifyThread");
3024	<	defaultForkJoinWorkerThreadFactory =
3025	<	new DefaultForkJoinWorkerThreadFactory();
3026	<	submitters = new ThreadSubmitter();
3027	<	int s;
3297	>	int s; // initialize field offsets for CAS etc
3298		try {
3299		U = getUnsafe();
3300		Class<?> k = ForkJoinPool.class;
3031	–	Class<?> ak = ForkJoinTask[].class;
3301		CTL = U.objectFieldOffset
3302		(k.getDeclaredField("ctl"));
3034	–	NEXTWORKERNUMBER = U.objectFieldOffset
3035	–	(k.getDeclaredField("nextWorkerNumber"));
3303		STEALCOUNT = U.objectFieldOffset
3304		(k.getDeclaredField("stealCount"));
3305	<	MAINLOCK = U.objectFieldOffset
3306	<	(k.getDeclaredField("mainLock"));
3305	>	PLOCK = U.objectFieldOffset
3306	>	(k.getDeclaredField("plock"));
3307	>	INDEXSEED = U.objectFieldOffset
3308	>	(k.getDeclaredField("indexSeed"));
3309		Class<?> tk = Thread.class;
3310		PARKBLOCKER = U.objectFieldOffset
3311		(tk.getDeclaredField("parkBlocker"));
3312	+	Class<?> wk = WorkQueue.class;
3313	+	QLOCK = U.objectFieldOffset
3314	+	(wk.getDeclaredField("qlock"));
3315	+	Class<?> ak = ForkJoinTask[].class;
3316		ABASE = U.arrayBaseOffset(ak);
3317		s = U.arrayIndexScale(ak);
3318		ASHIFT = 31 - Integer.numberOfLeadingZeros(s);
#	Line 3048 \| Line 3321 \| public class ForkJoinPool extends Abstra
3321		}
3322		if ((s & (s-1)) != 0)
3323		throw new Error("data type scale not a power of two");
3324	<	try { // Establish common pool
3325	<	String pp = System.getProperty(propPrefix + "parallelism");
3326	<	String fp = System.getProperty(propPrefix + "threadFactory");
3327	<	String up = System.getProperty(propPrefix + "exceptionHandler");
3328	<	ForkJoinWorkerThreadFactory fac = (fp == null) ?
3329	<	defaultForkJoinWorkerThreadFactory :
3330	<	((ForkJoinWorkerThreadFactory)ClassLoader.
3331	<	getSystemClassLoader().loadClass(fp).newInstance());
3332	<	Thread.UncaughtExceptionHandler ueh = (up == null)? null :
3333	<	((Thread.UncaughtExceptionHandler)ClassLoader.
3334	<	getSystemClassLoader().loadClass(up).newInstance());
3335	<	int par;
3336	<	if ((pp == null \|\| (par = Integer.parseInt(pp)) <= 0))
3337	<	par = Runtime.getRuntime().availableProcessors();
3338	<	if (par > MAX_CAP)
3339	<	par = MAX_CAP;
3340	<	commonPoolParallelism = par;
3341	<	int n = par - 1; // precompute submit mask
3342	<	n \|= n >>> 1; n \|= n >>> 2; n \|= n >>> 4;
3343	<	n \|= n >>> 8; n \|= n >>> 16;
3344	<	int mask = ((n + 1) << 1) - 1;
3345	<	commonPool = new ForkJoinPool(par, mask, fac, ueh);
3346	<	} catch (Exception e) {
3347	<	throw new Error(e);
3348	<	}
3324	>
3325	>	submitters = new ThreadLocal<Submitter>();
3326	>	ForkJoinWorkerThreadFactory fac = defaultForkJoinWorkerThreadFactory =
3327	>	new DefaultForkJoinWorkerThreadFactory();
3328	>	modifyThreadPermission = new RuntimePermission("modifyThread");
3329	>
3330	>	/*
3331	>	* Establish common pool parameters. For extra caution,
3332	>	* computations to set up common pool state are here; the
3333	>	* constructor just assigns these values to fields.
3334	>	*/
3335	>
3336	>	int par = 0;
3337	>	Thread.UncaughtExceptionHandler handler = null;
3338	>	try { // TBD: limit or report ignored exceptions?
3339	>	String pp = System.getProperty
3340	>	("java.util.concurrent.ForkJoinPool.common.parallelism");
3341	>	String hp = System.getProperty
3342	>	("java.util.concurrent.ForkJoinPool.common.exceptionHandler");
3343	>	String fp = System.getProperty
3344	>	("java.util.concurrent.ForkJoinPool.common.threadFactory");
3345	>	if (fp != null)
3346	>	fac = ((ForkJoinWorkerThreadFactory)ClassLoader.
3347	>	getSystemClassLoader().loadClass(fp).newInstance());
3348	>	if (hp != null)
3349	>	handler = ((Thread.UncaughtExceptionHandler)ClassLoader.
3350	>	getSystemClassLoader().loadClass(hp).newInstance());
3351	>	if (pp != null)
3352	>	par = Integer.parseInt(pp);
3353	>	} catch (Exception ignore) {
3354	>	}
3355	>
3356	>	if (par <= 0)
3357	>	par = Runtime.getRuntime().availableProcessors();
3358	>	if (par > MAX_CAP)
3359	>	par = MAX_CAP;
3360	>	commonPoolParallelism = par;
3361	>	long np = (long)(-par); // precompute initial ctl value
3362	>	long ct = ((np << AC_SHIFT) & AC_MASK) \| ((np << TC_SHIFT) & TC_MASK);
3363	>
3364	>	commonPool = new ForkJoinPool(par, ct, fac, handler);
3365		}
3366
3367		/**

Diff Legend

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

Comparing jsr166/src/jsr166y/ForkJoinPool.java (file contents): Revision 1.136 by dl, Mon Oct 29 17:23:34 2012 UTC vs. Revision 1.164 by dl, Tue Dec 18 21:46:16 2012 UTC

Diff Legend

Comparing jsr166/src/jsr166y/ForkJoinPool.java (file contents):
Revision 1.136 by dl, Mon Oct 29 17:23:34 2012 UTC vs.
Revision 1.164 by dl, Tue Dec 18 21:46:16 2012 UTC