ViewVC Help

View File | Revision Log | Show Annotations | Download File | Root Listing

root/jsr166/jsr166/src/jsr166y/ForkJoinPool.java

Comparing jsr166/src/jsr166y/ForkJoinPool.java (file contents):
Revision 1.63 by dl, Fri Aug 13 16:21:23 2010 UTC vs.
Revision 1.87 by dl, Tue Nov 23 00:10:39 2010 UTC

#	Line 6 | Line 6
6
7		package jsr166y;
8
9	–	import java.util.concurrent.*;
10	–
9		import java.util.ArrayList;
10		import java.util.Arrays;
11		import java.util.Collection;
12		import java.util.Collections;
13		import java.util.List;
14	+	import java.util.concurrent.AbstractExecutorService;
15	+	import java.util.concurrent.Callable;
16	+	import java.util.concurrent.ExecutorService;
17	+	import java.util.concurrent.Future;
18	+	import java.util.concurrent.RejectedExecutionException;
19	+	import java.util.concurrent.RunnableFuture;
20	+	import java.util.concurrent.TimeUnit;
21	+	import java.util.concurrent.TimeoutException;
22	+	import java.util.concurrent.atomic.AtomicInteger;
23		import java.util.concurrent.locks.LockSupport;
24		import java.util.concurrent.locks.ReentrantLock;
18	–	import java.util.concurrent.atomic.AtomicInteger;
19	–	import java.util.concurrent.CountDownLatch;
25
26		/**
27		* An {@link ExecutorService} for running {@link ForkJoinTask}s.
#	Line 69 | Line 74 | import java.util.concurrent.CountDownLat
74		*    <td ALIGN=CENTER> <b>Call from within fork/join computations</b></td>
75		*  </tr>
76		*  <tr>
77	<	*    <td> <b>Arange async execution</td>
77	>	*    <td> <b>Arrange async execution</td>
78		*    <td> {@link #execute(ForkJoinTask)}</td>
79		*    <td> {@link ForkJoinTask#fork}</td>
80		*  </tr>
#	Line 140 | Line 145 | public class ForkJoinPool extends Abstra
145		* Beyond work-stealing support and essential bookkeeping, the
146		* main responsibility of this framework is to take actions when
147		* one worker is waiting to join a task stolen (or always held by)
148	<	* another.  Becauae we are multiplexing many tasks on to a pool
148	>	* another.  Because we are multiplexing many tasks on to a pool
149		* of workers, we can't just let them block (as in Thread.join).
150		* We also cannot just reassign the joiner's run-time stack with
151		* another and replace it later, which would be a form of
#	Line 157 | Line 162 | public class ForkJoinPool extends Abstra
162		*      links to try to find such a task.
163		*
164		*   Compensating: Unless there are already enough live threads,
165	<	*      method helpMaintainParallelism() may create or or
165	>	*      method helpMaintainParallelism() may create or
166		*      re-activate a spare thread to compensate for blocked
167		*      joiners until they unblock.
168		*
169	<	* Because the determining existence of conservatively safe
170	<	* helping targets, the availability of already-created spares,
171	<	* and the apparent need to create new spares are all racy and
172	<	* require heuristic guidance, we rely on multiple retries of
173	<	* each. Further, because it is impossible to keep exactly the
174	<	* target (parallelism) number of threads running at any given
175	<	* time, we allow compensation during joins to fail, and enlist
176	<	* all other threads to help out whenever they are not otherwise
177	<	* occupied (i.e., mainly in method preStep).
169	>	* It is impossible to keep exactly the target (parallelism)
170	>	* number of threads running at any given time.  Determining
171	>	* existence of conservatively safe helping targets, the
172	>	* availability of already-created spares, and the apparent need
173	>	* to create new spares are all racy and require heuristic
174	>	* guidance, so we rely on multiple retries of each.  Compensation
175	>	* occurs in slow-motion. It is triggered only upon timeouts of
176	>	* Object.wait used for joins. This reduces poor decisions that
177	>	* would otherwise be made when threads are waiting for others
178	>	* that are stalled because of unrelated activities such as
179	>	* garbage collection.
180		*
181		* The ManagedBlocker extension API can't use helping so relies
182		* only on compensation in method awaitBlocker.
#	Line 224 | Line 231 | public class ForkJoinPool extends Abstra
231		* ManagedBlocker), we may create or resume others to take their
232		* place until they unblock (see below). Implementing this
233		* requires counts of the number of "running" threads (i.e., those
234	<	* that are neither blocked nor artifically suspended) as well as
234	>	* that are neither blocked nor artificially suspended) as well as
235		* the total number.  These two values are packed into one field,
236		* "workerCounts" because we need accurate snapshots when deciding
237		* to create, resume or suspend.  Note however that the
238	<	* correspondance of these counts to reality is not guaranteed. In
238	>	* correspondence of these counts to reality is not guaranteed. In
239		* particular updates for unblocked threads may lag until they
240		* actually wake up.
241		*
#	Line 259 | Line 266 | public class ForkJoinPool extends Abstra
266		* workers that previously could not find a task to now find one:
267		* Submission of a new task to the pool, or another worker pushing
268		* a task onto a previously empty queue.  (We also use this
269	<	* mechanism for termination actions that require wakeups of idle
270	<	* workers).  Each worker maintains its last known event count,
271	<	* and blocks when a scan for work did not find a task AND its
272	<	* lastEventCount matches the current eventCount. Waiting idle
273	<	* workers are recorded in a variant of Treiber stack headed by
274	<	* field eventWaiters which, when nonzero, encodes the thread
275	<	* index and count awaited for by the worker thread most recently
276	<	* calling eventSync. This thread in turn has a record (field
277	<	* nextEventWaiter) for the next waiting worker.  In addition to
278	<	* allowing simpler decisions about need for wakeup, the event
279	<	* count bits in eventWaiters serve the role of tags to avoid ABA
280	<	* errors in Treiber stacks.  To reduce delays in task diffusion,
281	<	* workers not otherwise occupied may invoke method
282	<	* releaseEventWaiters, that removes and signals (unparks) workers
283	<	* not waiting on current count. To reduce stalls, To minimize
284	<	* task production stalls associate with signalling, any worker
285	<	* pushing a task on an empty queue invokes the weaker method
279	<	* signalWork, that only releases idle workers until it detects
280	<	* interference by other threads trying to release, and lets them
281	<	* take over.  The net effect is a tree-like diffusion of signals,
282	<	* where released threads (and possibly others) help with unparks.
283	<	* To further reduce contention effects a bit, failed CASes to
284	<	* increment field eventCount are tolerated without retries.
269	>	* mechanism for configuration and termination actions that
270	>	* require wakeups of idle workers).  Each worker maintains its
271	>	* last known event count, and blocks when a scan for work did not
272	>	* find a task AND its lastEventCount matches the current
273	>	* eventCount. Waiting idle workers are recorded in a variant of
274	>	* Treiber stack headed by field eventWaiters which, when nonzero,
275	>	* encodes the thread index and count awaited for by the worker
276	>	* thread most recently calling eventSync. This thread in turn has
277	>	* a record (field nextEventWaiter) for the next waiting worker.
278	>	* In addition to allowing simpler decisions about need for
279	>	* wakeup, the event count bits in eventWaiters serve the role of
280	>	* tags to avoid ABA errors in Treiber stacks. Upon any wakeup,
281	>	* released threads also try to release at most two others.  The
282	>	* net effect is a tree-like diffusion of signals, where released
283	>	* threads (and possibly others) help with unparks.  To further
284	>	* reduce contention effects a bit, failed CASes to increment
285	>	* field eventCount are tolerated without retries in signalWork.
286		* Conceptually they are merged into the same event, which is OK
287		* when their only purpose is to enable workers to scan for work.
288		*
289	<	* 5. Managing suspension of extra workers. When a worker is about
290	<	* to block waiting for a join (or via ManagedBlockers), we may
291	<	* create a new thread to maintain parallelism level, or at least
292	<	* avoid starvation. Usually, extra threads are needed for only
293	<	* very short periods, yet join dependencies are such that we
294	<	* sometimes need them in bursts. Rather than create new threads
295	<	* each time this happens, we suspend no-longer-needed extra ones
296	<	* as "spares". For most purposes, we don't distinguish "extra"
297	<	* spare threads from normal "core" threads: On each call to
298	<	* preStep (the only point at which we can do this) a worker
299	<	* checks to see if there are now too many running workers, and if
300	<	* so, suspends itself.  Method helpMaintainParallelism looks for
301	<	* suspended threads to resume before considering creating a new
302	<	* replacement. The spares themselves are encoded on another
303	<	* variant of a Treiber Stack, headed at field "spareWaiters".
304	<	* Note that the use of spares is intrinsically racy.  One thread
305	<	* may become a spare at about the same time as another is
306	<	* needlessly being created. We counteract this and related slop
307	<	* in part by requiring resumed spares to immediately recheck (in
308	<	* preStep) to see whether they they should re-suspend.  To avoid
309	<	* long-term build-up of spares, the oldest spare (see
310	<	* ForkJoinWorkerThread.suspendAsSpare) occasionally wakes up if
311	<	* not signalled and calls tryTrimSpare, which uses two different
312	<	* thresholds: Always killing if the number of spares is greater
313	<	* that 25% of total, and killing others only at a slower rate
314	<	* (UNUSED_SPARE_TRIM_RATE_NANOS).
289	>	* 5. Managing suspension of extra workers. When a worker notices
290	>	* (usually upon timeout of a wait()) that there are too few
291	>	* running threads, we may create a new thread to maintain
292	>	* parallelism level, or at least avoid starvation. Usually, extra
293	>	* threads are needed for only very short periods, yet join
294	>	* dependencies are such that we sometimes need them in
295	>	* bursts. Rather than create new threads each time this happens,
296	>	* we suspend no-longer-needed extra ones as "spares". For most
297	>	* purposes, we don't distinguish "extra" spare threads from
298	>	* normal "core" threads: On each call to preStep (the only point
299	>	* at which we can do this) a worker checks to see if there are
300	>	* now too many running workers, and if so, suspends itself.
301	>	* Method helpMaintainParallelism looks for suspended threads to
302	>	* resume before considering creating a new replacement. The
303	>	* spares themselves are encoded on another variant of a Treiber
304	>	* Stack, headed at field "spareWaiters".  Note that the use of
305	>	* spares is intrinsically racy.  One thread may become a spare at
306	>	* about the same time as another is needlessly being created. We
307	>	* counteract this and related slop in part by requiring resumed
308	>	* spares to immediately recheck (in preStep) to see whether they
309	>	* should re-suspend.
310	>	*
311	>	* 6. Killing off unneeded workers. A timeout mechanism is used to
312	>	* shed unused workers: The oldest (first) event queue waiter uses
313	>	* a timed rather than hard wait. When this wait times out without
314	>	* a normal wakeup, it tries to shutdown any one (for convenience
315	>	* the newest) other spare or event waiter via
316	>	* tryShutdownUnusedWorker. This eventually reduces the number of
317	>	* worker threads to a minimum of one after a long enough period
318	>	* without use.
319		*
320	<	* 6. Deciding when to create new workers. The main dynamic
320	>	* 7. Deciding when to create new workers. The main dynamic
321		* control in this class is deciding when to create extra threads
322		* in method helpMaintainParallelism. We would like to keep
323	<	* exactly #parallelism threads running, which is an impossble
323	>	* exactly #parallelism threads running, which is an impossible
324		* task. We always need to create one when the number of running
325		* threads would become zero and all workers are busy. Beyond
326	<	* this, we must rely on heuristics that work well in the the
327	<	* presence of transients phenomena such as GC stalls, dynamic
326	>	* this, we must rely on heuristics that work well in the
327	>	* presence of transient phenomena such as GC stalls, dynamic
328		* compilation, and wake-up lags. These transients are extremely
329		* common -- we are normally trying to fully saturate the CPUs on
330		* a machine, so almost any activity other than running tasks
331	<	* impedes accuracy. Our main defense is to allow some slack in
332	<	* creation thresholds, using rules that reflect the fact that the
333	<	* more threads we have running, the more likely that we are
334	<	* underestimating the number running threads. The rules also
335	<	* better cope with the fact that some of the methods in this
336	<	* class tend to never become compiled (but are interpreted), so
337	<	* some components of the entire set of controls might execute 100
338	<	* times faster than others. And similarly for cases where the
339	<	* apparent lack of work is just due to GC stalls and other
335	<	* transient system activity.
331	>	* impedes accuracy. Our main defense is to allow parallelism to
332	>	* lapse for a while during joins, and use a timeout to see if,
333	>	* after the resulting settling, there is still a need for
334	>	* additional workers.  This also better copes with the fact that
335	>	* some of the methods in this class tend to never become compiled
336	>	* (but are interpreted), so some components of the entire set of
337	>	* controls might execute 100 times faster than others. And
338	>	* similarly for cases where the apparent lack of work is just due
339	>	* to GC stalls and other transient system activity.
340		*
341		* Beware that there is a lot of representation-level coupling
342		* among classes ForkJoinPool, ForkJoinWorkerThread, and
#	Line 347 | Line 351 | public class ForkJoinPool extends Abstra
351		* "while ((local = field) != 0)") which are usually the simplest
352		* way to ensure the required read orderings (which are sometimes
353		* critical). Also several occurrences of the unusual "do {}
354	<	* while(!cas...)" which is the simplest way to force an update of
354	>	* while (!cas...)" which is the simplest way to force an update of
355		* a CAS'ed variable. There are also other coding oddities that
356		* help some methods perform reasonably even when interpreted (not
357		* compiled), at the expense of some messy constructions that
#	Line 419 | Line 423 | public class ForkJoinPool extends Abstra
423		new AtomicInteger();
424
425		/**
426	+	* The time to block in a join (see awaitJoin) before checking if
427	+	* a new worker should be (re)started to maintain parallelism
428	+	* level. The value should be short enough to maintain global
429	+	* responsiveness and progress but long enough to avoid
430	+	* counterproductive firings during GC stalls or unrelated system
431	+	* activity, and to not bog down systems with continual re-firings
432	+	* on GCs or legitimately long waits.
433	+	*/
434	+	private static final long JOIN_TIMEOUT_MILLIS = 250L; // 4 per second
435	+
436	+	/**
437	+	* The wakeup interval (in nanoseconds) for the oldest worker
438	+	* waiting for an event to invoke tryShutdownUnusedWorker to
439	+	* shrink the number of workers.  The exact value does not matter
440	+	* too much. It must be short enough to release resources during
441	+	* sustained periods of idleness, but not so short that threads
442	+	* are continually re-created.
443	+	*/
444	+	private static final long SHRINK_RATE_NANOS =
445	+	30L * 1000L * 1000L * 1000L; // 2 per minute
446	+
447	+	/**
448		* Absolute bound for parallelism level. Twice this number plus
449		* one (i.e., 0xfff) must fit into a 16bit field to enable
450		* word-packing for some counts and indices.
#	Line 463 | Line 489 | public class ForkJoinPool extends Abstra
489		private volatile long stealCount;
490
491		/**
492	<	* The last nanoTime that a spare thread was trimmed
467	<	*/
468	<	private volatile long trimTime;
469	<
470	<	/**
471	<	* The rate at which to trim unused spares
472	<	*/
473	<	static final long UNUSED_SPARE_TRIM_RATE_NANOS =
474	<	1000L * 1000L * 1000L; // 1 sec
475	<
476	<	/**
477	<	* Encoded record of top of treiber stack of threads waiting for
492	>	* Encoded record of top of Treiber stack of threads waiting for
493		* events. The top 32 bits contain the count being waited for. The
494		* bottom 16 bits contains one plus the pool index of waiting
495		* worker thread. (Bits 16-31 are unused.)
496		*/
497		private volatile long eventWaiters;
498
499	<	private static final int  EVENT_COUNT_SHIFT = 32;
500	<	private static final long WAITER_ID_MASK    = (1L << 16) - 1L;
499	>	private static final int EVENT_COUNT_SHIFT = 32;
500	>	private static final int WAITER_ID_MASK    = (1 << 16) - 1;
501
502		/**
503		* A counter for events that may wake up worker threads:
#	Line 493 | Line 508 | public class ForkJoinPool extends Abstra
508		private volatile int eventCount;
509
510		/**
511	<	* Encoded record of top of treiber stack of spare threads waiting
511	>	* Encoded record of top of Treiber stack of spare threads waiting
512		* for resumption. The top 16 bits contain an arbitrary count to
513		* avoid ABA effects. The bottom 16bits contains one plus the pool
514		* index of waiting worker thread.
#	Line 507 | Line 522 | public class ForkJoinPool extends Abstra
522		* Lifecycle control. The low word contains the number of workers
523		* that are (probably) executing tasks. This value is atomically
524		* incremented before a worker gets a task to run, and decremented
525	<	* when worker has no tasks and cannot find any.  Bits 16-18
525	>	* when a worker has no tasks and cannot find any.  Bits 16-18
526		* contain runLevel value. When all are zero, the pool is
527		* running. Level transitions are monotonic (running -> shutdown
528		* -> terminating -> terminated) so each transition adds a bit.
529		* These are bundled together to ensure consistent read for
530		* termination checks (i.e., that runLevel is at least SHUTDOWN
531		* and active threads is zero).
532	+	*
533	+	* Notes: Most direct CASes are dependent on these bitfield
534	+	* positions.  Also, this field is non-private to enable direct
535	+	* performance-sensitive CASes in ForkJoinWorkerThread.
536		*/
537	<	private volatile int runState;
537	>	volatile int runState;
538
539		// Note: The order among run level values matters.
540		private static final int RUNLEVEL_SHIFT     = 16;
#	Line 523 | Line 542 | public class ForkJoinPool extends Abstra
542		private static final int TERMINATING        = 1 << (RUNLEVEL_SHIFT + 1);
543		private static final int TERMINATED         = 1 << (RUNLEVEL_SHIFT + 2);
544		private static final int ACTIVE_COUNT_MASK  = (1 << RUNLEVEL_SHIFT) - 1;
526	–	private static final int ONE_ACTIVE         = 1; // active update delta
545
546		/**
547		* Holds number of total (i.e., created and not yet terminated)
#	Line 564 | Line 582 | public class ForkJoinPool extends Abstra
582		*/
583		private final int poolNumber;
584
585	<
586	<	// Utilities for CASing fields. Note that several of these
569	<	// are manually inlined by callers
585	>	// Utilities for CASing fields. Note that most of these
586	>	// are usually manually inlined by callers
587
588		/**
589	<	* Increments running count part of workerCounts
589	>	* Increments running count part of workerCounts.
590		*/
591		final void incrementRunningCount() {
592		int c;
#	Line 579 | Line 596 | public class ForkJoinPool extends Abstra
596		}
597
598		/**
599	<	* Tries to decrement running count unless already zero
599	>	* Tries to increment running count part of workerCounts.
600	>	*/
601	>	final boolean tryIncrementRunningCount() {
602	>	int c;
603	>	return UNSAFE.compareAndSwapInt(this, workerCountsOffset,
604	>	c = workerCounts,
605	>	c + ONE_RUNNING);
606	>	}
607	>
608	>	/**
609	>	* Tries to decrement running count unless already zero.
610		*/
611		final boolean tryDecrementRunningCount() {
612		int wc = workerCounts;
#	Line 594 | Line 621 | public class ForkJoinPool extends Abstra
621		* (rarely) necessary when other count updates lag.
622		*
623		* @param dr -- either zero or ONE_RUNNING
624	<	* @param dt == either zero or ONE_TOTAL
624	>	* @param dt -- either zero or ONE_TOTAL
625		*/
626		private void decrementWorkerCounts(int dr, int dt) {
627		for (;;) {
628		int wc = workerCounts;
602	–	if (wc == 0 && (runState & TERMINATED) != 0)
603	–	return; // lagging termination on a backout
629		if ((wc & RUNNING_COUNT_MASK)  - dr < 0 ||
630	<	(wc >>> TOTAL_COUNT_SHIFT) - dt < 0)
630	>	(wc >>> TOTAL_COUNT_SHIFT) - dt < 0) {
631	>	if ((runState & TERMINATED) != 0)
632	>	return; // lagging termination on a backout
633		Thread.yield();
634	+	}
635		if (UNSAFE.compareAndSwapInt(this, workerCountsOffset,
636		wc, wc - (dr + dt)))
637		return;
#	Line 611 | Line 639 | public class ForkJoinPool extends Abstra
639		}
640
641		/**
614	–	* Increments event count
615	–	*/
616	–	private void advanceEventCount() {
617	–	int c;
618	–	do {} while(!UNSAFE.compareAndSwapInt(this, eventCountOffset,
619	–	c = eventCount, c+1));
620	–	}
621	–
622	–	/**
623	–	* Tries incrementing active count; fails on contention.
624	–	* Called by workers before executing tasks.
625	–	*
626	–	* @return true on success
627	–	*/
628	–	final boolean tryIncrementActiveCount() {
629	–	int c;
630	–	return UNSAFE.compareAndSwapInt(this, runStateOffset,
631	–	c = runState, c + ONE_ACTIVE);
632	–	}
633	–
634	–	/**
642		* Tries decrementing active count; fails on contention.
643		* Called when workers cannot find tasks to run.
644		*/
645		final boolean tryDecrementActiveCount() {
646		int c;
647		return UNSAFE.compareAndSwapInt(this, runStateOffset,
648	<	c = runState, c - ONE_ACTIVE);
648	>	c = runState, c - 1);
649		}
650
651		/**
#	Line 672 | Line 679 | public class ForkJoinPool extends Abstra
679		for (k = 0; k < n && ws[k] != null; ++k)
680		;
681		if (k == n)
682	<	ws = Arrays.copyOf(ws, n << 1);
682	>	ws = workers = Arrays.copyOf(ws, n << 1);
683		}
684		ws[k] = w;
685	<	workers = ws; // volatile array write ensures slot visibility
685	>	int c = eventCount; // advance event count to ensure visibility
686	>	UNSAFE.compareAndSwapInt(this, eventCountOffset, c, c+1);
687		} finally {
688		lock.unlock();
689		}
#	Line 683 | Line 691 | public class ForkJoinPool extends Abstra
691		}
692
693		/**
694	<	* Nulls out record of worker in workers array
694	>	* Nulls out record of worker in workers array.
695		*/
696		private void forgetWorker(ForkJoinWorkerThread w) {
697		int idx = w.poolIndex;
698	<	// Locking helps method recordWorker avoid unecessary expansion
698	>	// Locking helps method recordWorker avoid unnecessary expansion
699		final ReentrantLock lock = this.workerLock;
700		lock.lock();
701		try {
#	Line 699 | Line 707 | public class ForkJoinPool extends Abstra
707		}
708		}
709
702	–	// adding and removing workers
703	–
704	–	/**
705	–	* Tries to create and add new worker. Assumes that worker counts
706	–	* are already updated to accommodate the worker, so adjusts on
707	–	* failure.
708	–	*/
709	–	private void addWorker() {
710	–	ForkJoinWorkerThread w = null;
711	–	try {
712	–	w = factory.newThread(this);
713	–	} finally { // Adjust on either null or exceptional factory return
714	–	if (w == null) {
715	–	decrementWorkerCounts(ONE_RUNNING, ONE_TOTAL);
716	–	tryTerminate(false); // in case of failure during shutdown
717	–	}
718	–	}
719	–	if (w != null)
720	–	w.start(recordWorker(w), ueh);
721	–	}
722	–
710		/**
711		* Final callback from terminating worker.  Removes record of
712		* worker from array, and adjusts counts. If pool is shutting
713	<	* down, tries to complete terminatation.
713	>	* down, tries to complete termination.
714		*
715		* @param w the worker
716		*/
717		final void workerTerminated(ForkJoinWorkerThread w) {
718		forgetWorker(w);
719	<	decrementWorkerCounts(w.isTrimmed()? 0 : ONE_RUNNING, ONE_TOTAL);
719	>	decrementWorkerCounts(w.isTrimmed() ? 0 : ONE_RUNNING, ONE_TOTAL);
720		while (w.stealCount != 0) // collect final count
721		tryAccumulateStealCount(w);
722		tryTerminate(false);
#	Line 740 | Line 727 | public class ForkJoinPool extends Abstra
727		/**
728		* Releases workers blocked on a count not equal to current count.
729		* Normally called after precheck that eventWaiters isn't zero to
730	<	* avoid wasted array checks.
731	<	*
745	<	* @param signalling true if caller is a signalling worker so can
746	<	* exit upon (conservatively) detected contention by other threads
747	<	* who will continue to release
730	>	* avoid wasted array checks. Gives up upon a change in count or
731	>	* upon releasing two workers, letting others take over.
732		*/
733	<	private void releaseEventWaiters(boolean signalling) {
733	>	private void releaseEventWaiters() {
734		ForkJoinWorkerThread[] ws = workers;
735		int n = ws.length;
736	<	long h; // head of stack
737	<	ForkJoinWorkerThread w; int id, ec;
738	<	while ((id = ((int)((h = eventWaiters) & WAITER_ID_MASK)) - 1) >= 0 &&
739	<	(int)(h >>> EVENT_COUNT_SHIFT) != (ec = eventCount) &&
736	>	long h = eventWaiters;
737	>	int ec = eventCount;
738	>	boolean releasedOne = false;
739	>	ForkJoinWorkerThread w; int id;
740	>	while ((id = (((int)h) & WAITER_ID_MASK) - 1) >= 0 &&
741	>	(int)(h >>> EVENT_COUNT_SHIFT) != ec &&
742		id < n && (w = ws[id]) != null) {
743		if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
744	<	h, h = w.nextWaiter))
744	>	h,  w.nextWaiter)) {
745		LockSupport.unpark(w);
746	<	if (signalling && (eventCount != ec || eventWaiters != h))
746	>	if (releasedOne) // exit on second release
747	>	break;
748	>	releasedOne = true;
749	>	}
750	>	if (eventCount != ec)
751		break;
752	+	h = eventWaiters;
753		}
754		}
755
#	Line 770 | Line 761 | public class ForkJoinPool extends Abstra
761		int c; // try to increment event count -- CAS failure OK
762		UNSAFE.compareAndSwapInt(this, eventCountOffset, c = eventCount, c+1);
763		if (eventWaiters != 0L)
764	<	releaseEventWaiters(true);
764	>	releaseEventWaiters();
765		}
766
767		/**
768	<	* Blocks worker until terminating or event count
769	<	* advances from last value held by worker
768	>	* Adds the given worker to event queue and blocks until
769	>	* terminating or event count advances from the given value
770		*
771		* @param w the calling worker thread
772	+	* @param ec the count
773		*/
774	<	private void eventSync(ForkJoinWorkerThread w) {
775	<	int wec = w.lastEventCount;
784	<	long nh = (((long)wec) << EVENT_COUNT_SHIFT) | ((long)(w.poolIndex+1));
774	>	private void eventSync(ForkJoinWorkerThread w, int ec) {
775	>	long nh = (((long)ec) << EVENT_COUNT_SHIFT) | ((long)(w.poolIndex+1));
776		long h;
777		while ((runState < SHUTDOWN || !tryTerminate(false)) &&
778	<	((h = eventWaiters) == 0L ||
779	<	(int)(h >>> EVENT_COUNT_SHIFT) == wec) &&
780	<	eventCount == wec) {
778	>	(((int)(h = eventWaiters) & WAITER_ID_MASK) == 0 ||
779	>	(int)(h >>> EVENT_COUNT_SHIFT) == ec) &&
780	>	eventCount == ec) {
781		if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
782		w.nextWaiter = h, nh)) {
783	<	while (runState < TERMINATING && eventCount == wec) {
784	<	if (!tryAccumulateStealCount(w))  // transfer while idle
785	<	continue;
786	<	Thread.interrupted();             // clear/ignore interrupt
787	<	if (eventCount != wec)
788	<	break;
783	>	awaitEvent(w, ec);
784	>	break;
785	>	}
786	>	}
787	>	}
788	>
789	>	/**
790	>	* Blocks the given worker (that has already been entered as an
791	>	* event waiter) until terminating or event count advances from
792	>	* the given value. The oldest (first) waiter uses a timed wait to
793	>	* occasionally one-by-one shrink the number of workers (to a
794	>	* minimum of one) if the pool has not been used for extended
795	>	* periods.
796	>	*
797	>	* @param w the calling worker thread
798	>	* @param ec the count
799	>	*/
800	>	private void awaitEvent(ForkJoinWorkerThread w, int ec) {
801	>	while (eventCount == ec) {
802	>	if (tryAccumulateStealCount(w)) { // transfer while idle
803	>	boolean untimed = (w.nextWaiter != 0L ||
804	>	(workerCounts & RUNNING_COUNT_MASK) <= 1);
805	>	long startTime = untimed ? 0 : System.nanoTime();
806	>	Thread.interrupted();         // clear/ignore interrupt
807	>	if (w.isTerminating() || eventCount != ec)
808	>	break;                    // recheck after clear
809	>	if (untimed)
810		LockSupport.park(w);
811	+	else {
812	+	LockSupport.parkNanos(w, SHRINK_RATE_NANOS);
813	+	if (eventCount != ec || w.isTerminating())
814	+	break;
815	+	if (System.nanoTime() - startTime >= SHRINK_RATE_NANOS)
816	+	tryShutdownUnusedWorker(ec);
817		}
800	–	break;
818		}
819		}
803	–	w.lastEventCount = eventCount;
820		}
821
822	<	// Maintaining spares
822	>	// Maintaining parallelism
823
824		/**
825	<	* Pushes worker onto the spare stack
825	>	* Pushes worker onto the spare stack.
826		*/
827		final void pushSpare(ForkJoinWorkerThread w) {
828	<	int ns = (++w.spareCount << SPARE_COUNT_SHIFT) | (w.poolIndex+1);
828	>	int ns = (++w.spareCount << SPARE_COUNT_SHIFT) | (w.poolIndex + 1);
829		do {} while (!UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
830		w.nextSpare = spareWaiters,ns));
831		}
832
833		/**
834	<	* Tries (once) to resume a spare if running count is less than
835	<	* target parallelism. Fails on contention or stale workers.
834	>	* Tries (once) to resume a spare if the number of running
835	>	* threads is less than target.
836		*/
837		private void tryResumeSpare() {
838		int sw, id;
839	+	ForkJoinWorkerThread[] ws = workers;
840	+	int n = ws.length;
841		ForkJoinWorkerThread w;
842	<	ForkJoinWorkerThread[] ws;
843	<	if ((id = ((sw = spareWaiters) & SPARE_ID_MASK) - 1) >= 0 &&
844	<	id < (ws = workers).length && (w = ws[id]) != null &&
845	<	(workerCounts & RUNNING_COUNT_MASK) < parallelism &&
846	<	eventWaiters == 0L &&
842	>	if ((sw = spareWaiters) != 0 &&
843	>	(id = (sw & SPARE_ID_MASK) - 1) >= 0 &&
844	>	id < n && (w = ws[id]) != null &&
845	>	(runState >= TERMINATING ||
846	>	(workerCounts & RUNNING_COUNT_MASK) < parallelism) &&
847		spareWaiters == sw &&
848		UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
849	<	sw, w.nextSpare) &&
850	<	w.tryUnsuspend()) {
851	<	int c; // try increment; if contended, finish after unpark
852	<	boolean inc = UNSAFE.compareAndSwapInt(this, workerCountsOffset,
853	<	c = workerCounts,
854	<	c + ONE_RUNNING);
855	<	LockSupport.unpark(w);
856	<	if (!inc) {
857	<	do {} while(!UNSAFE.compareAndSwapInt(this, workerCountsOffset,
840	<	c = workerCounts,
841	<	c + ONE_RUNNING));
842	<	}
849	>	sw, w.nextSpare)) {
850	>	int c; // increment running count before resume
851	>	do {} while (!UNSAFE.compareAndSwapInt
852	>	(this, workerCountsOffset,
853	>	c = workerCounts, c + ONE_RUNNING));
854	>	if (w.tryUnsuspend())
855	>	LockSupport.unpark(w);
856	>	else   // back out if w was shutdown
857	>	decrementWorkerCounts(ONE_RUNNING, 0);
858		}
859		}
860
861		/**
862	<	* Callback from oldest spare occasionally waking up.  Tries
863	<	* (once) to shutdown a spare if more than 25% spare overage, or
864	<	* if UNUSED_SPARE_TRIM_RATE_NANOS have elapsed and there are at
865	<	* least #parallelism running threads. Note that we don't need CAS
866	<	* or locks here because the method is called only from the oldest
867	<	* suspended spare occasionally waking (and even misfires are OK).
868	<	*
854	<	* @param now the wake up nanoTime of caller
855	<	*/
856	<	final void tryTrimSpare(long now) {
857	<	long lastTrim = trimTime;
858	<	trimTime = now;
859	<	helpMaintainParallelism(); // first, help wake up any needed spares
860	<	int sw, id;
861	<	ForkJoinWorkerThread w;
862	<	ForkJoinWorkerThread[] ws;
862	>	* Tries to increase the number of running workers if below target
863	>	* parallelism: If a spare exists tries to resume it via
864	>	* tryResumeSpare.  Otherwise, if not enough total workers or all
865	>	* existing workers are busy, adds a new worker. In all cases also
866	>	* helps wake up releasable workers waiting for work.
867	>	*/
868	>	private void helpMaintainParallelism() {
869		int pc = parallelism;
870	<	int wc = workerCounts;
871	<	if ((wc & RUNNING_COUNT_MASK) >= pc &&
872	<	(((wc >>> TOTAL_COUNT_SHIFT) - pc) > (pc >>> 2) + 1 ||// approx 25%
873	<	now - lastTrim >= UNUSED_SPARE_TRIM_RATE_NANOS) &&
874	<	(id = ((sw = spareWaiters) & SPARE_ID_MASK) - 1) >= 0 &&
875	<	id < (ws = workers).length && (w = ws[id]) != null &&
876	<	UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
877	<	sw, w.nextSpare))
878	<	w.shutdown(false);
870	>	int wc, rs, tc;
871	>	while (((wc = workerCounts) & RUNNING_COUNT_MASK) < pc &&
872	>	(rs = runState) < TERMINATING) {
873	>	if (spareWaiters != 0)
874	>	tryResumeSpare();
875	>	else if ((tc = wc >>> TOTAL_COUNT_SHIFT) >= MAX_WORKERS ||
876	>	(tc >= pc && (rs & ACTIVE_COUNT_MASK) != tc))
877	>	break;   // enough total
878	>	else if (runState == rs && workerCounts == wc &&
879	>	UNSAFE.compareAndSwapInt(this, workerCountsOffset, wc,
880	>	wc + (ONE_RUNNING|ONE_TOTAL))) {
881	>	ForkJoinWorkerThread w = null;
882	>	Throwable fail = null;
883	>	try {
884	>	w = factory.newThread(this);
885	>	} catch (Throwable ex) {
886	>	fail = ex;
887	>	}
888	>	if (w == null) { // null or exceptional factory return
889	>	decrementWorkerCounts(ONE_RUNNING, ONE_TOTAL);
890	>	tryTerminate(false); // handle failure during shutdown
891	>	// If originating from an external caller,
892	>	// propagate exception, else ignore
893	>	if (fail != null && runState < TERMINATING &&
894	>	!(Thread.currentThread() instanceof
895	>	ForkJoinWorkerThread))
896	>	UNSAFE.throwException(fail);
897	>	break;
898	>	}
899	>	w.start(recordWorker(w), ueh);
900	>	if ((workerCounts >>> TOTAL_COUNT_SHIFT) >= pc)
901	>	break; // add at most one unless total below target
902	>	}
903	>	}
904	>	if (eventWaiters != 0L)
905	>	releaseEventWaiters();
906		}
907
908		/**
909	<	* Does at most one of:
910	<	*
911	<	* 1. Help wake up existing workers waiting for work via
912	<	*    releaseEventWaiters. (If any exist, then it probably doesn't
913	<	*    matter right now if under target parallelism level.)
914	<	*
915	<	* 2. If below parallelism level and a spare exists, try (once)
883	<	*    to resume it via tryResumeSpare.
909	>	* Callback from the oldest waiter in awaitEvent waking up after a
910	>	* period of non-use. If all workers are idle, tries (once) to
911	>	* shutdown an event waiter or a spare, if one exists. Note that
912	>	* we don't need CAS or locks here because the method is called
913	>	* only from one thread occasionally waking (and even misfires are
914	>	* OK). Note that until the shutdown worker fully terminates,
915	>	* workerCounts will overestimate total count, which is tolerable.
916		*
917	<	* 3. If neither of the above, tries (once) to add a new
918	<	*    worker if either there are not enough total, or if all
887	<	*    existing workers are busy, there are either no running
888	<	*    workers or the deficit is at least twice the surplus.
917	>	* @param ec the event count waited on by caller (to abort
918	>	* attempt if count has since changed).
919		*/
920	<	private void helpMaintainParallelism() {
921	<	// uglified to work better when not compiled
922	<	int pc, wc, rc, tc, rs; long h;
923	<	if ((h = eventWaiters) != 0L) {
924	<	if ((int)(h >>> EVENT_COUNT_SHIFT) != eventCount)
925	<	releaseEventWaiters(false); // avoid useless call
926	<	}
927	<	else if ((pc = parallelism) >
928	<	(rc = ((wc = workerCounts) & RUNNING_COUNT_MASK))) {
929	<	if (spareWaiters != 0)
930	<	tryResumeSpare();
931	<	else if ((rs = runState) < TERMINATING &&
932	<	((tc = wc >>> TOTAL_COUNT_SHIFT) < pc ||
933	<	(tc == (rs & ACTIVE_COUNT_MASK) && // all busy
934	<	(rc == 0 ||                       // must add
935	<	rc < pc - ((tc - pc) << 1)) &&   // within slack
936	<	tc < MAX_WORKERS && runState == rs)) && // recheck busy
937	<	workerCounts == wc &&
938	<	UNSAFE.compareAndSwapInt(this, workerCountsOffset, wc,
939	<	wc + (ONE_RUNNING|ONE_TOTAL)))
940	<	addWorker();
920	>	private void tryShutdownUnusedWorker(int ec) {
921	>	if (runState == 0 && eventCount == ec) { // only trigger if all idle
922	>	ForkJoinWorkerThread[] ws = workers;
923	>	int n = ws.length;
924	>	ForkJoinWorkerThread w = null;
925	>	boolean shutdown = false;
926	>	int sw;
927	>	long h;
928	>	if ((sw = spareWaiters) != 0) { // prefer killing spares
929	>	int id = (sw & SPARE_ID_MASK) - 1;
930	>	if (id >= 0 && id < n && (w = ws[id]) != null &&
931	>	UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
932	>	sw, w.nextSpare))
933	>	shutdown = true;
934	>	}
935	>	else if ((h = eventWaiters) != 0L) {
936	>	long nh;
937	>	int id = (((int)h) & WAITER_ID_MASK) - 1;
938	>	if (id >= 0 && id < n && (w = ws[id]) != null &&
939	>	(nh = w.nextWaiter) != 0L && // keep at least one worker
940	>	UNSAFE.compareAndSwapLong(this, eventWaitersOffset, h, nh))
941	>	shutdown = true;
942	>	}
943	>	if (w != null && shutdown) {
944	>	w.shutdown();
945	>	LockSupport.unpark(w);
946	>	}
947		}
948	+	releaseEventWaiters(); // in case of interference
949		}
950
951		/**
952		* Callback from workers invoked upon each top-level action (i.e.,
953	<	* stealing a task or taking a submission and running
954	<	* it). Performs one or more of the following:
953	>	* stealing a task or taking a submission and running it).
954	>	* Performs one or more of the following:
955		*
956	<	* 1. If the worker cannot find work (misses > 0), updates its
957	<	*    active status to inactive and updates activeCount unless
958	<	*    this is the first miss and there is contention, in which
959	<	*    case it may try again (either in this or a subsequent
960	<	*    call).
961	<	*
962	<	* 2. If there are at least 2 misses, awaits the next task event
963	<	*    via eventSync
964	<	*
965	<	* 3. If there are too many running threads, suspends this worker
966	<	*    (first forcing inactivation if necessary).  If it is not
967	<	*    needed, it may be killed while suspended via
968	<	*    tryTrimSpare. Otherwise, upon resume it rechecks to make
969	<	*    sure that it is still needed.
970	<	*
971	<	* 4. Helps release and/or reactivate other workers via
972	<	*    helpMaintainParallelism
956	>	* 1. If the worker is active and either did not run a task
957	>	*    or there are too many workers, try to set its active status
958	>	*    to inactive and update activeCount. On contention, we may
959	>	*    try again in this or a subsequent call.
960	>	*
961	>	* 2. If not enough total workers, help create some.
962	>	*
963	>	* 3. If there are too many running workers, suspend this worker
964	>	*    (first forcing inactive if necessary).  If it is not needed,
965	>	*    it may be shutdown while suspended (via
966	>	*    tryShutdownUnusedWorker).  Otherwise, upon resume it
967	>	*    rechecks running thread count and need for event sync.
968	>	*
969	>	* 4. If worker did not run a task, await the next task event via
970	>	*    eventSync if necessary (first forcing inactivation), upon
971	>	*    which the worker may be shutdown via
972	>	*    tryShutdownUnusedWorker.  Otherwise, help release any
973	>	*    existing event waiters that are now releasable,
974		*
975		* @param w the worker
976	<	* @param misses the number of scans by caller failing to find work
939	<	* (saturating at 2 just to avoid wraparound)
976	>	* @param ran true if worker ran a task since last call to this method
977		*/
978	<	final void preStep(ForkJoinWorkerThread w, int misses) {
978	>	final void preStep(ForkJoinWorkerThread w, boolean ran) {
979	>	int wec = w.lastEventCount;
980		boolean active = w.active;
981	+	boolean inactivate = false;
982		int pc = parallelism;
983	<	for (;;) {
984	<	int wc = workerCounts;
985	<	int rc = wc & RUNNING_COUNT_MASK;
986	<	if (active && (misses > 0 || rc > pc)) {
987	<	int rs;                      // try inactivate
949	<	if (UNSAFE.compareAndSwapInt(this, runStateOffset,
950	<	rs = runState, rs - ONE_ACTIVE))
951	<	active = w.active = false;
952	<	else if (misses > 1 || rc > pc ||
953	<	(rs & ACTIVE_COUNT_MASK) >= pc)
954	<	continue;                // force inactivate
983	>	while (w.runState == 0) {
984	>	int rs = runState;
985	>	if (rs >= TERMINATING) {           // propagate shutdown
986	>	w.shutdown();
987	>	break;
988		}
989	<	if (misses > 1) {
990	<	misses = 0;                  // don't re-sync
991	<	eventSync(w);                // continue loop to recheck rc
989	>	if ((inactivate || (active && (rs & ACTIVE_COUNT_MASK) >= pc)) &&
990	>	UNSAFE.compareAndSwapInt(this, runStateOffset, rs, --rs)) {
991	>	inactivate = active = w.active = false;
992	>	if (rs == SHUTDOWN) {          // all inactive and shut down
993	>	tryTerminate(false);
994	>	continue;
995	>	}
996		}
997	<	else if (rc > pc) {
998	<	if (workerCounts == wc &&   // try to suspend as spare
997	>	int wc = workerCounts;             // try to suspend as spare
998	>	if ((wc & RUNNING_COUNT_MASK) > pc) {
999	>	if (!(inactivate |= active) && // must inactivate to suspend
1000	>	workerCounts == wc &&
1001		UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1002	<	wc, wc - ONE_RUNNING) &&
1003	<	!w.suspendAsSpare())    // false if killed
965	<	break;
1002	>	wc, wc - ONE_RUNNING))
1003	>	w.suspendAsSpare();
1004		}
1005	<	else {
1006	<	if (rc < pc || eventWaiters != 0L)
1007	<	helpMaintainParallelism();
1005	>	else if ((wc >>> TOTAL_COUNT_SHIFT) < pc)
1006	>	helpMaintainParallelism();     // not enough workers
1007	>	else if (ran)
1008		break;
1009	+	else {
1010	+	long h = eventWaiters;
1011	+	int ec = eventCount;
1012	+	if (h != 0L && (int)(h >>> EVENT_COUNT_SHIFT) != ec)
1013	+	releaseEventWaiters();     // release others before waiting
1014	+	else if (ec != wec) {
1015	+	w.lastEventCount = ec;     // no need to wait
1016	+	break;
1017	+	}
1018	+	else if (!(inactivate |= active))
1019	+	eventSync(w, wec);         // must inactivate before sync
1020		}
1021		}
1022		}
1023
1024		/**
1025		* Helps and/or blocks awaiting join of the given task.
1026	<	* Alternates between helpJoinTask() and helpMaintainParallelism()
978	<	* as many times as there is a deficit in running count (or longer
979	<	* if running count would become zero), then blocks if task still
980	<	* not done.
1026	>	* See above for explanation.
1027		*
1028		* @param joinMe the task to join
1029	<	*/
1030	<	final void awaitJoin(ForkJoinTask<?> joinMe, ForkJoinWorkerThread worker) {
1031	<	int threshold = parallelism;         // descend blocking thresholds
1029	>	* @param worker the current worker thread
1030	>	* @param timed true if wait should time out
1031	>	* @param nanos timeout value if timed
1032	>	*/
1033	>	final void awaitJoin(ForkJoinTask<?> joinMe, ForkJoinWorkerThread worker,
1034	>	boolean timed, long nanos) {
1035	>	long startTime = timed? System.nanoTime() : 0L;
1036	>	int retries = 2 + (parallelism >> 2); // #helpJoins before blocking
1037	>	boolean running = true;               // false when count decremented
1038		while (joinMe.status >= 0) {
1039	<	boolean block; int wc;
1040	<	worker.helpJoinTask(joinMe);
989	<	if (joinMe.status < 0)
1039	>	if (runState >= TERMINATING) {
1040	>	joinMe.cancelIgnoringExceptions();
1041		break;
991	–	if (((wc = workerCounts) & RUNNING_COUNT_MASK) <= threshold) {
992	–	if (threshold > 0)
993	–	--threshold;
994	–	else
995	–	advanceEventCount(); // force release
996	–	block = false;
1042		}
1043	<	else
1044	<	block = UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1000	<	wc, wc - ONE_RUNNING);
1001	<	helpMaintainParallelism();
1002	<	if (block) {
1003	<	int c;
1004	<	joinMe.internalAwaitDone();
1005	<	do {} while (!UNSAFE.compareAndSwapInt
1006	<	(this, workerCountsOffset,
1007	<	c = workerCounts, c + ONE_RUNNING));
1043	>	running = worker.helpJoinTask(joinMe, running);
1044	>	if (joinMe.status < 0)
1045		break;
1046	+	if (retries > 0) {
1047	+	--retries;
1048	+	continue;
1049		}
1050	+	int wc = workerCounts;
1051	+	if ((wc & RUNNING_COUNT_MASK) != 0) {
1052	+	if (running) {
1053	+	if (!UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1054	+	wc, wc - ONE_RUNNING))
1055	+	continue;
1056	+	running = false;
1057	+	}
1058	+	long h = eventWaiters;
1059	+	if (h != 0L && (int)(h >>> EVENT_COUNT_SHIFT) != eventCount)
1060	+	releaseEventWaiters();
1061	+	if ((workerCounts & RUNNING_COUNT_MASK) != 0) {
1062	+	long ms; int ns;
1063	+	if (!timed) {
1064	+	ms = JOIN_TIMEOUT_MILLIS;
1065	+	ns = 0;
1066	+	}
1067	+	else { // at most JOIN_TIMEOUT_MILLIS per wait
1068	+	long nt = nanos - (System.nanoTime() - startTime);
1069	+	if (nt <= 0L)
1070	+	break;
1071	+	ms = nt / 1000000;
1072	+	if (ms > JOIN_TIMEOUT_MILLIS) {
1073	+	ms = JOIN_TIMEOUT_MILLIS;
1074	+	ns = 0;
1075	+	}
1076	+	else
1077	+	ns = (int) (nt % 1000000);
1078	+	}
1079	+	joinMe.internalAwaitDone(ms, ns);
1080	+	}
1081	+	if (joinMe.status < 0)
1082	+	break;
1083	+	}
1084	+	helpMaintainParallelism();
1085	+	}
1086	+	if (!running) {
1087	+	int c;
1088	+	do {} while (!UNSAFE.compareAndSwapInt
1089	+	(this, workerCountsOffset,
1090	+	c = workerCounts, c + ONE_RUNNING));
1091		}
1092		}
1093
1094		/**
1095	<	* Same idea as awaitJoin, but no helping
1095	>	* Same idea as awaitJoin, but no helping, retries, or timeouts.
1096		*/
1097		final void awaitBlocker(ManagedBlocker blocker)
1098		throws InterruptedException {
1018	–	int threshold = parallelism;
1099		while (!blocker.isReleasable()) {
1100	<	boolean block; int wc;
1101	<	if (((wc = workerCounts) & RUNNING_COUNT_MASK) <= threshold) {
1102	<	if (threshold > 0)
1103	<	--threshold;
1104	<	else
1025	<	advanceEventCount();
1026	<	block = false;
1027	<	}
1028	<	else
1029	<	block = UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1030	<	wc, wc - ONE_RUNNING);
1031	<	helpMaintainParallelism();
1032	<	if (block) {
1100	>	int wc = workerCounts;
1101	>	if ((wc & RUNNING_COUNT_MASK) == 0)
1102	>	helpMaintainParallelism();
1103	>	else if (UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1104	>	wc, wc - ONE_RUNNING)) {
1105		try {
1106	<	do {} while (!blocker.isReleasable() && !blocker.block());
1106	>	while (!blocker.isReleasable()) {
1107	>	long h = eventWaiters;
1108	>	if (h != 0L &&
1109	>	(int)(h >>> EVENT_COUNT_SHIFT) != eventCount)
1110	>	releaseEventWaiters();
1111	>	else if ((workerCounts & RUNNING_COUNT_MASK) == 0 &&
1112	>	runState < TERMINATING)
1113	>	helpMaintainParallelism();
1114	>	else if (blocker.block())
1115	>	break;
1116	>	}
1117		} finally {
1118		int c;
1119		do {} while (!UNSAFE.compareAndSwapInt
#	Line 1064 | Line 1146 | public class ForkJoinPool extends Abstra
1146		// Finish now if all threads terminated; else in some subsequent call
1147		if ((workerCounts >>> TOTAL_COUNT_SHIFT) == 0) {
1148		advanceRunLevel(TERMINATED);
1149	<	termination.arrive();
1149	>	termination.forceTermination();
1150		}
1151		return true;
1152		}
#	Line 1073 | Line 1155 | public class ForkJoinPool extends Abstra
1155		* Actions on transition to TERMINATING
1156		*
1157		* Runs up to four passes through workers: (0) shutting down each
1158	<	* quietly (without waking up if parked) to quickly spread
1159	<	* notifications without unnecessary bouncing around event queues
1160	<	* etc (1) wake up and help cancel tasks (2) interrupt (3) mop up
1161	<	* races with interrupted workers
1158	>	* (without waking up if parked) to quickly spread notifications
1159	>	* without unnecessary bouncing around event queues etc (1) wake
1160	>	* up and help cancel tasks (2) interrupt (3) mop up races with
1161	>	* interrupted workers
1162		*/
1163		private void startTerminating() {
1164		cancelSubmissions();
1165		for (int passes = 0; passes < 4 && workerCounts != 0; ++passes) {
1166	<	advanceEventCount();
1166	>	int c; // advance event count
1167	>	UNSAFE.compareAndSwapInt(this, eventCountOffset,
1168	>	c = eventCount, c+1);
1169		eventWaiters = 0L; // clobber lists
1170		spareWaiters = 0;
1171	<	ForkJoinWorkerThread[] ws = workers;
1088	<	int n = ws.length;
1089	<	for (int i = 0; i < n; ++i) {
1090	<	ForkJoinWorkerThread w = ws[i];
1171	>	for (ForkJoinWorkerThread w : workers) {
1172		if (w != null) {
1173	<	w.shutdown(true);
1173	>	w.shutdown();
1174		if (passes > 0 && !w.isTerminated()) {
1175		w.cancelTasks();
1176		LockSupport.unpark(w);
1177	<	if (passes > 1) {
1177	>	if (passes > 1 && !w.isInterrupted()) {
1178		try {
1179		w.interrupt();
1180		} catch (SecurityException ignore) {
#	Line 1106 | Line 1187 | public class ForkJoinPool extends Abstra
1187		}
1188
1189		/**
1190	<	* Clear out and cancel submissions, ignoring exceptions
1190	>	* Clears out and cancels submissions, ignoring exceptions.
1191		*/
1192		private void cancelSubmissions() {
1193		ForkJoinTask<?> task;
#	Line 1121 | Line 1202 | public class ForkJoinPool extends Abstra
1202		// misc support for ForkJoinWorkerThread
1203
1204		/**
1205	<	* Returns pool number
1205	>	* Returns pool number.
1206		*/
1207		final int getPoolNumber() {
1208		return poolNumber;
1209		}
1210
1211		/**
1212	<	* Tries to accumulates steal count from a worker, clearing
1213	<	* the worker's value.
1212	>	* Tries to accumulate steal count from a worker, clearing
1213	>	* the worker's value if successful.
1214		*
1215		* @return true if worker steal count now zero
1216		*/
#	Line 1151 | Line 1232 | public class ForkJoinPool extends Abstra
1232		*/
1233		final int idlePerActive() {
1234		int pc = parallelism; // use parallelism, not rc
1235	<	int ac = runState;    // no mask -- artifically boosts during shutdown
1235	>	int ac = runState;    // no mask -- artificially boosts during shutdown
1236		// Use exact results for small values, saturate past 4
1237	<	return pc <= ac? 0 : pc >>> 1 <= ac? 1 : pc >>> 2 <= ac? 3 : pc >>> 3;
1237	>	return ((pc <= ac) ? 0 :
1238	>	(pc >>> 1 <= ac) ? 1 :
1239	>	(pc >>> 2 <= ac) ? 3 :
1240	>	pc >>> 3);
1241		}
1242
1243		// Public and protected methods
#	Line 1203 | Line 1287 | public class ForkJoinPool extends Abstra
1287		* use {@link #defaultForkJoinWorkerThreadFactory}.
1288		* @param handler the handler for internal worker threads that
1289		* terminate due to unrecoverable errors encountered while executing
1290	<	* tasks. For default value, use <code>null</code>.
1290	>	* tasks. For default value, use {@code null}.
1291		* @param asyncMode if true,
1292		* establishes local first-in-first-out scheduling mode for forked
1293		* tasks that are never joined. This mode may be more appropriate
1294		* than default locally stack-based mode in applications in which
1295		* worker threads only process event-style asynchronous tasks.
1296	<	* For default value, use <code>false</code>.
1296	>	* For default value, use {@code false}.
1297		* @throws IllegalArgumentException if parallelism less than or
1298		*         equal to zero, or greater than implementation limit
1299		* @throws NullPointerException if the factory is null
#	Line 1237 | Line 1321 | public class ForkJoinPool extends Abstra
1321		this.workerLock = new ReentrantLock();
1322		this.termination = new Phaser(1);
1323		this.poolNumber = poolNumberGenerator.incrementAndGet();
1240	–	this.trimTime = System.nanoTime();
1324		}
1325
1326		/**
#	Line 1245 | Line 1328 | public class ForkJoinPool extends Abstra
1328		* @param pc the initial parallelism level
1329		*/
1330		private static int initialArraySizeFor(int pc) {
1331	<	// See Hackers Delight, sec 3.2. We know MAX_WORKERS < (1 >>> 16)
1331	>	// If possible, initially allocate enough space for one spare
1332		int size = pc < MAX_WORKERS ? pc + 1 : MAX_WORKERS;
1333	+	// See Hackers Delight, sec 3.2. We know MAX_WORKERS < (1 >>> 16)
1334		size |= size >>> 1;
1335		size |= size >>> 2;
1336		size |= size >>> 4;
#	Line 1257 | Line 1341 | public class ForkJoinPool extends Abstra
1341		// Execution methods
1342
1343		/**
1344	<	* Common code for execute, invoke and submit
1344	>	* Submits task and creates, starts, or resumes some workers if necessary
1345		*/
1346		private <T> void doSubmit(ForkJoinTask<T> task) {
1263	–	if (task == null)
1264	–	throw new NullPointerException();
1265	–	if (runState >= SHUTDOWN)
1266	–	throw new RejectedExecutionException();
1347		submissionQueue.offer(task);
1348	<	advanceEventCount();
1349	<	helpMaintainParallelism();         // start or wake up workers
1348	>	int c; // try to increment event count -- CAS failure OK
1349	>	UNSAFE.compareAndSwapInt(this, eventCountOffset, c = eventCount, c+1);
1350	>	helpMaintainParallelism();
1351		}
1352
1353		/**
#	Line 1279 | Line 1360 | public class ForkJoinPool extends Abstra
1360		*         scheduled for execution
1361		*/
1362		public <T> T invoke(ForkJoinTask<T> task) {
1363	<	doSubmit(task);
1364	<	return task.join();
1363	>	if (task == null)
1364	>	throw new NullPointerException();
1365	>	if (runState >= SHUTDOWN)
1366	>	throw new RejectedExecutionException();
1367	>	Thread t = Thread.currentThread();
1368	>	if ((t instanceof ForkJoinWorkerThread) &&
1369	>	((ForkJoinWorkerThread)t).pool == this)
1370	>	return task.invoke();  // bypass submit if in same pool
1371	>	else {
1372	>	doSubmit(task);
1373	>	return task.join();
1374	>	}
1375	>	}
1376	>
1377	>	/**
1378	>	* Unless terminating, forks task if within an ongoing FJ
1379	>	* computation in the current pool, else submits as external task.
1380	>	*/
1381	>	private <T> void forkOrSubmit(ForkJoinTask<T> task) {
1382	>	if (runState >= SHUTDOWN)
1383	>	throw new RejectedExecutionException();
1384	>	Thread t = Thread.currentThread();
1385	>	if ((t instanceof ForkJoinWorkerThread) &&
1386	>	((ForkJoinWorkerThread)t).pool == this)
1387	>	task.fork();
1388	>	else
1389	>	doSubmit(task);
1390		}
1391
1392		/**
1393		* Arranges for (asynchronous) execution of the given task.
1288	–	* If the caller is already engaged in a fork/join computation in
1289	–	* the current pool, this method is equivalent in effect to
1290	–	* {@link ForkJoinTask#fork}.
1394		*
1395		* @param task the task
1396		* @throws NullPointerException if the task is null
#	Line 1295 | Line 1398 | public class ForkJoinPool extends Abstra
1398		*         scheduled for execution
1399		*/
1400		public void execute(ForkJoinTask<?> task) {
1401	<	doSubmit(task);
1401	>	if (task == null)
1402	>	throw new NullPointerException();
1403	>	forkOrSubmit(task);
1404		}
1405
1406		// AbstractExecutorService methods
#	Line 1306 | Line 1411 | public class ForkJoinPool extends Abstra
1411		*         scheduled for execution
1412		*/
1413		public void execute(Runnable task) {
1414	+	if (task == null)
1415	+	throw new NullPointerException();
1416		ForkJoinTask<?> job;
1417		if (task instanceof ForkJoinTask<?>) // avoid re-wrap
1418		job = (ForkJoinTask<?>) task;
1419		else
1420		job = ForkJoinTask.adapt(task, null);
1421	<	doSubmit(job);
1421	>	forkOrSubmit(job);
1422		}
1423
1424		/**
1425		* Submits a ForkJoinTask for execution.
1319	–	* If the caller is already engaged in a fork/join computation in
1320	–	* the current pool, this method is equivalent in effect to
1321	–	* {@link ForkJoinTask#fork}.
1426		*
1427		* @param task the task to submit
1428		* @return the task
#	Line 1327 | Line 1431 | public class ForkJoinPool extends Abstra
1431		*         scheduled for execution
1432		*/
1433		public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) {
1434	<	doSubmit(task);
1434	>	if (task == null)
1435	>	throw new NullPointerException();
1436	>	forkOrSubmit(task);
1437		return task;
1438		}
1439
#	Line 1337 | Line 1443 | public class ForkJoinPool extends Abstra
1443		*         scheduled for execution
1444		*/
1445		public <T> ForkJoinTask<T> submit(Callable<T> task) {
1446	+	if (task == null)
1447	+	throw new NullPointerException();
1448		ForkJoinTask<T> job = ForkJoinTask.adapt(task);
1449	<	doSubmit(job);
1449	>	forkOrSubmit(job);
1450		return job;
1451		}
1452
#	Line 1348 | Line 1456 | public class ForkJoinPool extends Abstra
1456		*         scheduled for execution
1457		*/
1458		public <T> ForkJoinTask<T> submit(Runnable task, T result) {
1459	+	if (task == null)
1460	+	throw new NullPointerException();
1461		ForkJoinTask<T> job = ForkJoinTask.adapt(task, result);
1462	<	doSubmit(job);
1462	>	forkOrSubmit(job);
1463		return job;
1464		}
1465
#	Line 1359 | Line 1469 | public class ForkJoinPool extends Abstra
1469		*         scheduled for execution
1470		*/
1471		public ForkJoinTask<?> submit(Runnable task) {
1472	+	if (task == null)
1473	+	throw new NullPointerException();
1474		ForkJoinTask<?> job;
1475		if (task instanceof ForkJoinTask<?>) // avoid re-wrap
1476		job = (ForkJoinTask<?>) task;
1477		else
1478		job = ForkJoinTask.adapt(task, null);
1479	<	doSubmit(job);
1479	>	forkOrSubmit(job);
1480		return job;
1481		}
1482
#	Line 1424 | Line 1536 | public class ForkJoinPool extends Abstra
1536
1537		/**
1538		* Returns the number of worker threads that have started but not
1539	<	* yet terminated.  This result returned by this method may differ
1539	>	* yet terminated.  The result returned by this method may differ
1540		* from {@link #getParallelism} when threads are created to
1541		* maintain parallelism when others are cooperatively blocked.
1542		*
#	Line 1509 | Line 1621 | public class ForkJoinPool extends Abstra
1621		*/
1622		public long getQueuedTaskCount() {
1623		long count = 0;
1624	<	ForkJoinWorkerThread[] ws = workers;
1513	<	int n = ws.length;
1514	<	for (int i = 0; i < n; ++i) {
1515	<	ForkJoinWorkerThread w = ws[i];
1624	>	for (ForkJoinWorkerThread w : workers)
1625		if (w != null)
1626		count += w.getQueueSize();
1518	–	}
1627		return count;
1628		}
1629
#	Line 1570 | Line 1678 | public class ForkJoinPool extends Abstra
1678		*/
1679		protected int drainTasksTo(Collection<? super ForkJoinTask<?>> c) {
1680		int count = submissionQueue.drainTo(c);
1681	<	ForkJoinWorkerThread[] ws = workers;
1574	<	int n = ws.length;
1575	<	for (int i = 0; i < n; ++i) {
1576	<	ForkJoinWorkerThread w = ws[i];
1681	>	for (ForkJoinWorkerThread w : workers)
1682		if (w != null)
1683		count += w.drainTasksTo(c);
1579	–	}
1684		return count;
1685		}
1686
#	Line 1680 | Line 1784 | public class ForkJoinPool extends Abstra
1784		}
1785
1786		/**
1787	+	* Returns true if terminating or terminated. Used by ForkJoinWorkerThread.
1788	+	*/
1789	+	final boolean isAtLeastTerminating() {
1790	+	return runState >= TERMINATING;
1791	+	}
1792	+
1793	+	/**
1794		* Returns {@code true} if this pool has been shut down.
1795		*
1796		* @return {@code true} if this pool has been shut down
#	Line 1702 | Line 1813 | public class ForkJoinPool extends Abstra
1813		public boolean awaitTermination(long timeout, TimeUnit unit)
1814		throws InterruptedException {
1815		try {
1816	<	return termination.awaitAdvanceInterruptibly(0, timeout, unit) > 0;
1817	<	} catch(TimeoutException ex) {
1816	>	termination.awaitAdvanceInterruptibly(0, timeout, unit);
1817	>	} catch (TimeoutException ex) {
1818		return false;
1819		}
1820	+	return true;
1821		}
1822
1823		/**
#	Line 1750 | Line 1862 | public class ForkJoinPool extends Abstra
1862		*   QueueTaker(BlockingQueue<E> q) { this.queue = q; }
1863		*   public boolean block() throws InterruptedException {
1864		*     if (item == null)
1865	<	*       item = queue.take
1865	>	*       item = queue.take();
1866		*     return true;
1867		*   }
1868		*   public boolean isReleasable() {
1869	<	*     return item != null || (item = queue.poll) != null;
1869	>	*     return item != null || (item = queue.poll()) != null;
1870		*   }
1871		*   public E getItem() { // call after pool.managedBlock completes
1872		*     return item;
#	Line 1833 | Line 1945 | public class ForkJoinPool extends Abstra
1945		private static final long eventCountOffset =
1946		objectFieldOffset("eventCount", ForkJoinPool.class);
1947		private static final long eventWaitersOffset =
1948	<	objectFieldOffset("eventWaiters",ForkJoinPool.class);
1948	>	objectFieldOffset("eventWaiters", ForkJoinPool.class);
1949		private static final long stealCountOffset =
1950	<	objectFieldOffset("stealCount",ForkJoinPool.class);
1950	>	objectFieldOffset("stealCount", ForkJoinPool.class);
1951		private static final long spareWaitersOffset =
1952	<	objectFieldOffset("spareWaiters",ForkJoinPool.class);
1952	>	objectFieldOffset("spareWaiters", ForkJoinPool.class);
1953
1954		private static long objectFieldOffset(String field, Class<?> klazz) {
1955		try {

Diff Legend

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