ViewVC Help

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

root/jsr166/jsr166/src/jsr166y/Phaser.java

Comparing jsr166/src/jsr166y/Phaser.java (file contents):
Revision 1.64 by jsr166, Mon Nov 29 20:58:06 2010 UTC vs.
Revision 1.80 by jsr166, Sun Sep 13 16:28:14 2015 UTC

#	Line 1 | Line 1
1		/*
2		* Written by Doug Lea with assistance from members of JCP JSR-166
3		* Expert Group and released to the public domain, as explained at
4	<	* http://creativecommons.org/licenses/publicdomain
4	>	* http://creativecommons.org/publicdomain/zero/1.0/
5		*/
6
7		package jsr166y;
#	Line 17 | Line 17 | import java.util.concurrent.locks.LockSu
17		* {@link java.util.concurrent.CountDownLatch CountDownLatch}
18		* but supporting more flexible usage.
19		*
20	<	* <p> <b>Registration.</b> Unlike the case for other barriers, the
20	>	* <p><b>Registration.</b> Unlike the case for other barriers, the
21		* number of parties <em>registered</em> to synchronize on a phaser
22		* may vary over time.  Tasks may be registered at any time (using
23		* methods {@link #register}, {@link #bulkRegister}, or forms of
#	Line 30 | Line 30 | import java.util.concurrent.locks.LockSu
30		* (However, you can introduce such bookkeeping by subclassing this
31		* class.)
32		*
33	<	* <p> <b>Synchronization.</b> Like a {@code CyclicBarrier}, a {@code
33	>	* <p><b>Synchronization.</b> Like a {@code CyclicBarrier}, a {@code
34		* Phaser} may be repeatedly awaited.  Method {@link
35		* #arriveAndAwaitAdvance} has effect analogous to {@link
36		* java.util.concurrent.CyclicBarrier#await CyclicBarrier.await}. Each
#	Line 44 | Line 44 | import java.util.concurrent.locks.LockSu
44		*
45		* <ul>
46		*
47	<	*   <li> <b>Arrival.</b> Methods {@link #arrive} and
47	>	*   <li><b>Arrival.</b> Methods {@link #arrive} and
48		*       {@link #arriveAndDeregister} record arrival.  These methods
49		*       do not block, but return an associated <em>arrival phase
50		*       number</em>; that is, the phase number of the phaser to which
#	Line 57 | Line 57 | import java.util.concurrent.locks.LockSu
57		*       flexible than, providing a barrier action to a {@code
58		*       CyclicBarrier}.
59		*
60	<	*   <li> <b>Waiting.</b> Method {@link #awaitAdvance} requires an
60	>	*   <li><b>Waiting.</b> Method {@link #awaitAdvance} requires an
61		*       argument indicating an arrival phase number, and returns when
62		*       the phaser advances to (or is already at) a different phase.
63		*       Unlike similar constructions using {@code CyclicBarrier},
#	Line 74 | Line 74 | import java.util.concurrent.locks.LockSu
74		*
75		* </ul>
76		*
77	<	* <p> <b>Termination.</b> A phaser may enter a <em>termination</em>
78	<	* state in which all synchronization methods immediately return
79	<	* without updating phaser state or waiting for advance, and
80	<	* indicating (via a negative phase value) that execution is complete.
77	>	* <p><b>Termination.</b> A phaser may enter a <em>termination</em>
78	>	* state, that may be checked using method {@link #isTerminated}. Upon
79	>	* termination, all synchronization methods immediately return without
80	>	* waiting for advance, as indicated by a negative return value.
81	>	* Similarly, attempts to register upon termination have no effect.
82		* Termination is triggered when an invocation of {@code onAdvance}
83		* returns {@code true}. The default implementation returns {@code
84		* true} if a deregistration has caused the number of registered
#	Line 88 | Line 89 | import java.util.concurrent.locks.LockSu
89		* also available to abruptly release waiting threads and allow them
90		* to terminate.
91		*
92	<	* <p> <b>Tiering.</b> Phasers may be <em>tiered</em> (i.e.,
92	>	* <p><b>Tiering.</b> Phasers may be <em>tiered</em> (i.e.,
93		* constructed in tree structures) to reduce contention. Phasers with
94		* large numbers of parties that would otherwise experience heavy
95		* synchronization contention costs may instead be set up so that
#	Line 96 | Line 97 | import java.util.concurrent.locks.LockSu
97		* increase throughput even though it incurs greater per-operation
98		* overhead.
99		*
100	+	* <p>In a tree of tiered phasers, registration and deregistration of
101	+	* child phasers with their parent are managed automatically.
102	+	* Whenever the number of registered parties of a child phaser becomes
103	+	* non-zero (as established in the {@link #Phaser(Phaser,int)}
104	+	* constructor, {@link #register}, or {@link #bulkRegister}), the
105	+	* child phaser is registered with its parent.  Whenever the number of
106	+	* registered parties becomes zero as the result of an invocation of
107	+	* {@link #arriveAndDeregister}, the child phaser is deregistered
108	+	* from its parent.
109	+	*
110		* <p><b>Monitoring.</b> While synchronization methods may be invoked
111		* only by registered parties, the current state of a phaser may be
112		* monitored by any caller.  At any given moment there are {@link
#	Line 119 | Line 130 | import java.util.concurrent.locks.LockSu
130		* void runTasks(List<Runnable> tasks) {
131		*   final Phaser phaser = new Phaser(1); // "1" to register self
132		*   // create and start threads
133	<	*   for (Runnable task : tasks) {
133	>	*   for (final Runnable task : tasks) {
134		*     phaser.register();
135		*     new Thread() {
136		*       public void run() {
#	Line 183 | Line 194 | import java.util.concurrent.locks.LockSu
194		* }}</pre>
195		*
196		*
197	<	* <p>To create a set of tasks using a tree of phasers,
198	<	* you could use code of the following form, assuming a
199	<	* Task class with a constructor accepting a {@code Phaser} that
200	<	* it registers with upon construction:
197	>	* <p>To create a set of {@code n} tasks using a tree of phasers, you
198	>	* could use code of the following form, assuming a Task class with a
199	>	* constructor accepting a {@code Phaser} that it registers with upon
200	>	* construction. After invocation of {@code build(new Task[n], 0, n,
201	>	* new Phaser())}, these tasks could then be started, for example by
202	>	* submitting to a pool:
203		*
204		*  <pre> {@code
205	<	* void build(Task[] actions, int lo, int hi, Phaser ph) {
205	>	* void build(Task[] tasks, int lo, int hi, Phaser ph) {
206		*   if (hi - lo > TASKS_PER_PHASER) {
207		*     for (int i = lo; i < hi; i += TASKS_PER_PHASER) {
208		*       int j = Math.min(i + TASKS_PER_PHASER, hi);
209	<	*       build(actions, i, j, new Phaser(ph));
209	>	*       build(tasks, i, j, new Phaser(ph));
210		*     }
211		*   } else {
212		*     for (int i = lo; i < hi; ++i)
213	<	*       actions[i] = new Task(ph);
213	>	*       tasks[i] = new Task(ph);
214		*       // assumes new Task(ph) performs ph.register()
215		*   }
216	<	* }
204	<	* // .. initially called, for n tasks via
205	<	* build(new Task[n], 0, n, new Phaser());}</pre>
216	>	* }}</pre>
217		*
218		* The best value of {@code TASKS_PER_PHASER} depends mainly on
219		* expected synchronization rates. A value as low as four may
#	Line 226 | Line 237 | public class Phaser {
237		*/
238
239		/**
240	<	* Primary state representation, holding four fields:
240	>	* Primary state representation, holding four bit-fields:
241		*
242	<	* * unarrived -- the number of parties yet to hit barrier (bits  0-15)
243	<	* * parties -- the number of parties to wait              (bits 16-31)
244	<	* * phase -- the generation of the barrier                (bits 32-62)
245	<	* * terminated -- set if barrier is terminated            (bit  63 / sign)
246	<	*
247	<	* However, to efficiently maintain atomicity, these values are
248	<	* packed into a single (atomic) long. Termination uses the sign
249	<	* bit of 32 bit representation of phase, so phase is set to -1 on
250	<	* termination. Good performance relies on keeping state decoding
251	<	* and encoding simple, and keeping race windows short.
242	>	* unarrived  -- the number of parties yet to hit barrier (bits  0-15)
243	>	* parties    -- the number of parties to wait            (bits 16-31)
244	>	* phase      -- the generation of the barrier            (bits 32-62)
245	>	* terminated -- set if barrier is terminated             (bit  63 / sign)
246	>	*
247	>	* Except that a phaser with no registered parties is
248	>	* distinguished by the otherwise illegal state of having zero
249	>	* parties and one unarrived parties (encoded as EMPTY below).
250	>	*
251	>	* To efficiently maintain atomicity, these values are packed into
252	>	* a single (atomic) long. Good performance relies on keeping
253	>	* state decoding and encoding simple, and keeping race windows
254	>	* short.
255	>	*
256	>	* All state updates are performed via CAS except initial
257	>	* registration of a sub-phaser (i.e., one with a non-null
258	>	* parent).  In this (relatively rare) case, we use built-in
259	>	* synchronization to lock while first registering with its
260	>	* parent.
261	>	*
262	>	* The phase of a subphaser is allowed to lag that of its
263	>	* ancestors until it is actually accessed -- see method
264	>	* reconcileState.
265		*/
266		private volatile long state;
267
268		private static final int  MAX_PARTIES     = 0xffff;
269	<	private static final int  MAX_PHASE       = 0x7fffffff;
269	>	private static final int  MAX_PHASE       = Integer.MAX_VALUE;
270		private static final int  PARTIES_SHIFT   = 16;
271		private static final int  PHASE_SHIFT     = 32;
272		private static final int  UNARRIVED_MASK  = 0xffff;      // to mask ints
273		private static final long PARTIES_MASK    = 0xffff0000L; // to mask longs
274	<	private static final long ONE_ARRIVAL     = 1L;
251	<	private static final long ONE_PARTY       = 1L << PARTIES_SHIFT;
274	>	private static final long COUNTS_MASK     = 0xffffffffL;
275		private static final long TERMINATION_BIT = 1L << 63;
276
277	+	// some special values
278	+	private static final int  ONE_ARRIVAL     = 1;
279	+	private static final int  ONE_PARTY       = 1 << PARTIES_SHIFT;
280	+	private static final int  ONE_DEREGISTER  = ONE_ARRIVAL|ONE_PARTY;
281	+	private static final int  EMPTY           = 1;
282	+
283		// The following unpacking methods are usually manually inlined
284
285		private static int unarrivedOf(long s) {
286	<	return (int)s & UNARRIVED_MASK;
286	>	int counts = (int)s;
287	>	return (counts == EMPTY) ? 0 : (counts & UNARRIVED_MASK);
288		}
289
290		private static int partiesOf(long s) {
#	Line 262 | Line 292 | public class Phaser {
292		}
293
294		private static int phaseOf(long s) {
295	<	return (int) (s >>> PHASE_SHIFT);
295	>	return (int)(s >>> PHASE_SHIFT);
296		}
297
298		private static int arrivedOf(long s) {
299	<	return partiesOf(s) - unarrivedOf(s);
299	>	int counts = (int)s;
300	>	return (counts == EMPTY) ? 0 :
301	>	(counts >>> PARTIES_SHIFT) - (counts & UNARRIVED_MASK);
302		}
303
304		/**
#	Line 275 | Line 307 | public class Phaser {
307		private final Phaser parent;
308
309		/**
310	<	* The root of phaser tree. Equals this if not in a tree.  Used to
279	<	* support faster state push-down.
310	>	* The root of phaser tree. Equals this if not in a tree.
311		*/
312		private final Phaser root;
313
#	Line 314 | Line 345 | public class Phaser {
345		* Manually tuned to speed up and minimize race windows for the
346		* common case of just decrementing unarrived field.
347		*
348	<	* @param adj - adjustment to apply to state -- either
349	<	* ONE_ARRIVAL (for arrive) or
350	<	* ONE_ARRIVAL|ONE_PARTY (for arriveAndDeregister)
348	>	* @param adjust value to subtract from state;
349	>	*               ONE_ARRIVAL for arrive,
350	>	*               ONE_DEREGISTER for arriveAndDeregister
351		*/
352	<	private int doArrive(long adj) {
352	>	private int doArrive(int adjust) {
353	>	final Phaser root = this.root;
354		for (;;) {
355	<	long s = state;
324	<	int unarrived = (int)s & UNARRIVED_MASK;
355	>	long s = (root == this) ? state : reconcileState();
356		int phase = (int)(s >>> PHASE_SHIFT);
357		if (phase < 0)
358		return phase;
359	<	else if (unarrived == 0) {
360	<	if (reconcileState() == s)     // recheck
361	<	throw new IllegalStateException(badArrive(s));
362	<	}
363	<	else if (UNSAFE.compareAndSwapLong(this, stateOffset, s, s-=adj)) {
359	>	int counts = (int)s;
360	>	int unarrived = (counts == EMPTY) ? 0 : (counts & UNARRIVED_MASK);
361	>	if (unarrived <= 0)
362	>	throw new IllegalStateException(badArrive(s));
363	>	if (UNSAFE.compareAndSwapLong(this, stateOffset, s, s-=adjust)) {
364		if (unarrived == 1) {
365	<	long p = s & PARTIES_MASK; // unshifted parties field
366	<	long lu = p >>> PARTIES_SHIFT;
367	<	int u = (int)lu;
368	<	int nextPhase = (phase + 1) & MAX_PHASE;
369	<	long next = ((long)nextPhase << PHASE_SHIFT) | p | lu;
370	<	final Phaser parent = this.parent;
371	<	if (parent == null) {
372	<	if (onAdvance(phase, u))
373	<	next |= TERMINATION_BIT;
374	<	UNSAFE.compareAndSwapLong(this, stateOffset, s, next);
365	>	long n = s & PARTIES_MASK;  // base of next state
366	>	int nextUnarrived = (int)n >>> PARTIES_SHIFT;
367	>	if (root == this) {
368	>	if (onAdvance(phase, nextUnarrived))
369	>	n |= TERMINATION_BIT;
370	>	else if (nextUnarrived == 0)
371	>	n |= EMPTY;
372	>	else
373	>	n |= nextUnarrived;
374	>	int nextPhase = (phase + 1) & MAX_PHASE;
375	>	n |= (long)nextPhase << PHASE_SHIFT;
376	>	UNSAFE.compareAndSwapLong(this, stateOffset, s, n);
377		releaseWaiters(phase);
378		}
379	<	else {
380	<	parent.doArrive((u == 0) ?
381	<	ONE_ARRIVAL|ONE_PARTY : ONE_ARRIVAL);
382	<	if ((int)(parent.state >>> PHASE_SHIFT) != nextPhase)
350	<	reconcileState();
351	<	else if (state == s)
352	<	UNSAFE.compareAndSwapLong(this, stateOffset, s,
353	<	next);
379	>	else if (nextUnarrived == 0) { // propagate deregistration
380	>	phase = parent.doArrive(ONE_DEREGISTER);
381	>	UNSAFE.compareAndSwapLong(this, stateOffset,
382	>	s, s | EMPTY);
383		}
384	+	else
385	+	phase = parent.doArrive(ONE_ARRIVAL);
386		}
387		return phase;
388		}
#	Line 366 | Line 397 | public class Phaser {
397		*/
398		private int doRegister(int registrations) {
399		// adjustment to state
400	<	long adj = ((long)registrations << PARTIES_SHIFT) | registrations;
400	>	long adjust = ((long)registrations << PARTIES_SHIFT) | registrations;
401		final Phaser parent = this.parent;
402	+	int phase;
403		for (;;) {
404		long s = (parent == null) ? state : reconcileState();
405	<	int parties = (int)s >>> PARTIES_SHIFT;
406	<	int phase = (int)(s >>> PHASE_SHIFT);
407	<	if (phase < 0)
408	<	return phase;
377	<	else if (registrations > MAX_PARTIES - parties)
405	>	int counts = (int)s;
406	>	int parties = counts >>> PARTIES_SHIFT;
407	>	int unarrived = counts & UNARRIVED_MASK;
408	>	if (registrations > MAX_PARTIES - parties)
409		throw new IllegalStateException(badRegister(s));
410	<	else if ((parties == 0 && parent == null) || // first reg of root
411	<	((int)s & UNARRIVED_MASK) != 0) {   // not advancing
412	<	if (UNSAFE.compareAndSwapLong(this, stateOffset, s, s + adj))
413	<	return phase;
410	>	phase = (int)(s >>> PHASE_SHIFT);
411	>	if (phase < 0)
412	>	break;
413	>	if (counts != EMPTY) {                  // not 1st registration
414	>	if (parent == null || reconcileState() == s) {
415	>	if (unarrived == 0)             // wait out advance
416	>	root.internalAwaitAdvance(phase, null);
417	>	else if (UNSAFE.compareAndSwapLong(this, stateOffset,
418	>	s, s + adjust))
419	>	break;
420	>	}
421		}
422	<	else if (parties != 0)               // wait for onAdvance
423	<	root.internalAwaitAdvance(phase, null);
424	<	else {                               // 1st registration of child
425	<	synchronized (this) {            // register parent first
426	<	if (reconcileState() == s) { // recheck under lock
427	<	parent.doRegister(1);    // OK if throws IllegalState
428	<	for (;;) {               // simpler form of outer loop
429	<	s = reconcileState();
430	<	phase = (int)(s >>> PHASE_SHIFT);
431	<	if (phase < 0 ||
432	<	UNSAFE.compareAndSwapLong(this, stateOffset,
433	<	s, s + adj))
434	<	return phase;
422	>	else if (parent == null) {              // 1st root registration
423	>	long next = ((long)phase << PHASE_SHIFT) | adjust;
424	>	if (UNSAFE.compareAndSwapLong(this, stateOffset, s, next))
425	>	break;
426	>	}
427	>	else {
428	>	synchronized (this) {               // 1st sub registration
429	>	if (state == s) {               // recheck under lock
430	>	phase = parent.doRegister(1);
431	>	if (phase < 0)
432	>	break;
433	>	// finish registration whenever parent registration
434	>	// succeeded, even when racing with termination,
435	>	// since these are part of the same "transaction".
436	>	while (!UNSAFE.compareAndSwapLong
437	>	(this, stateOffset, s,
438	>	((long)phase << PHASE_SHIFT) | adjust)) {
439	>	s = state;
440	>	phase = (int)(root.state >>> PHASE_SHIFT);
441	>	// assert (int)s == EMPTY;
442		}
443	+	break;
444		}
445		}
446		}
447		}
448	+	return phase;
449		}
450
451		/**
452	<	* Recursively resolves lagged phase propagation from root if necessary.
452	>	* Resolves lagged phase propagation from root if necessary.
453	>	* Reconciliation normally occurs when root has advanced but
454	>	* subphasers have not yet done so, in which case they must finish
455	>	* their own advance by setting unarrived to parties (or if
456	>	* parties is zero, resetting to unregistered EMPTY state).
457	>	*
458	>	* @return reconciled state
459		*/
460		private long reconcileState() {
461	<	Phaser par = parent;
461	>	final Phaser root = this.root;
462		long s = state;
463	<	if (par != null) {
464	<	Phaser rt = root;
465	<	int phase, rPhase;
466	<	while ((phase = (int)(s >>> PHASE_SHIFT)) >= 0 &&
467	<	(rPhase = (int)(rt.state >>> PHASE_SHIFT)) != phase) {
468	<	if (par != rt && (int)(par.state >>> PHASE_SHIFT) != rPhase)
469	<	par.reconcileState();
470	<	else if (rPhase < 0 || ((int)s & UNARRIVED_MASK) == 0) {
471	<	long u = s & PARTIES_MASK; // reset unarrived to parties
472	<	long next = ((((long) rPhase) << PHASE_SHIFT) | u |
473	<	(u >>> PARTIES_SHIFT));
421	<	UNSAFE.compareAndSwapLong(this, stateOffset, s, next);
422	<	}
463	>	if (root != this) {
464	>	int phase, p;
465	>	// CAS to root phase with current parties, tripping unarrived
466	>	while ((phase = (int)(root.state >>> PHASE_SHIFT)) !=
467	>	(int)(s >>> PHASE_SHIFT) &&
468	>	!UNSAFE.compareAndSwapLong
469	>	(this, stateOffset, s,
470	>	s = (((long)phase << PHASE_SHIFT) |
471	>	((phase < 0) ? (s & COUNTS_MASK) :
472	>	(((p = (int)s >>> PARTIES_SHIFT) == 0) ? EMPTY :
473	>	((s & PARTIES_MASK) | p))))))
474		s = state;
424	–	}
475		}
476		return s;
477		}
#	Line 459 | Line 509 | public class Phaser {
509
510		/**
511		* Creates a new phaser with the given parent and number of
512	<	* registered unarrived parties. Registration and deregistration
513	<	* of this child phaser with its parent are managed automatically.
514	<	* If the given parent is non-null, whenever this child phaser has
465	<	* any registered parties (as established in this constructor,
466	<	* {@link #register}, or {@link #bulkRegister}), this child phaser
467	<	* is registered with its parent. Whenever the number of
468	<	* registered parties becomes zero as the result of an invocation
469	<	* of {@link #arriveAndDeregister}, this child phaser is
470	<	* deregistered from its parent.
512	>	* registered unarrived parties.  When the given parent is non-null
513	>	* and the given number of parties is greater than zero, this
514	>	* child phaser is registered with its parent.
515		*
516		* @param parent the parent phaser
517		* @param parties the number of parties required to advance to the
#	Line 478 | Line 522 | public class Phaser {
522		public Phaser(Phaser parent, int parties) {
523		if (parties >>> PARTIES_SHIFT != 0)
524		throw new IllegalArgumentException("Illegal number of parties");
525	<	long s = ((long) parties) | (((long) parties) << PARTIES_SHIFT);
525	>	int phase = 0;
526		this.parent = parent;
527		if (parent != null) {
528	<	Phaser r = parent.root;
529	<	this.root = r;
530	<	this.evenQ = r.evenQ;
531	<	this.oddQ = r.oddQ;
528	>	final Phaser root = parent.root;
529	>	this.root = root;
530	>	this.evenQ = root.evenQ;
531	>	this.oddQ = root.oddQ;
532		if (parties != 0)
533	<	s |= ((long)(parent.doRegister(1))) << PHASE_SHIFT;
533	>	phase = parent.doRegister(1);
534		}
535		else {
536		this.root = this;
537		this.evenQ = new AtomicReference<QNode>();
538		this.oddQ = new AtomicReference<QNode>();
539		}
540	<	this.state = s;
540	>	this.state = (parties == 0) ? (long)EMPTY :
541	>	((long)phase << PHASE_SHIFT) |
542	>	((long)parties << PARTIES_SHIFT) |
543	>	((long)parties);
544		}
545
546		/**
#	Line 501 | Line 548 | public class Phaser {
548		* invocation of {@link #onAdvance} is in progress, this method
549		* may await its completion before returning.  If this phaser has
550		* a parent, and this phaser previously had no registered parties,
551	<	* this phaser is also registered with its parent.
552	<	*
553	<	* @return the arrival phase number to which this registration applied
551	>	* this child phaser is also registered with its parent. If
552	>	* this phaser is terminated, the attempt to register has
553	>	* no effect, and a negative value is returned.
554	>	*
555	>	* @return the arrival phase number to which this registration
556	>	* applied.  If this value is negative, then this phaser has
557	>	* terminated, in which case registration has no effect.
558		* @throws IllegalStateException if attempting to register more
559		* than the maximum supported number of parties
560		*/
#	Line 515 | Line 566 | public class Phaser {
566		* Adds the given number of new unarrived parties to this phaser.
567		* If an ongoing invocation of {@link #onAdvance} is in progress,
568		* this method may await its completion before returning.  If this
569	<	* phaser has a parent, and the given number of parities is
570	<	* greater than zero, and this phaser previously had no registered
571	<	* parties, this phaser is also registered with its parent.
569	>	* phaser has a parent, and the given number of parties is greater
570	>	* than zero, and this phaser previously had no registered
571	>	* parties, this child phaser is also registered with its parent.
572	>	* If this phaser is terminated, the attempt to register has no
573	>	* effect, and a negative value is returned.
574		*
575		* @param parties the number of additional parties required to
576		* advance to the next phase
577	<	* @return the arrival phase number to which this registration applied
577	>	* @return the arrival phase number to which this registration
578	>	* applied.  If this value is negative, then this phaser has
579	>	* terminated, in which case registration has no effect.
580		* @throws IllegalStateException if attempting to register more
581		* than the maximum supported number of parties
582		* @throws IllegalArgumentException if {@code parties < 0}
#	Line 567 | Line 622 | public class Phaser {
622		* of registered or unarrived parties would become negative
623		*/
624		public int arriveAndDeregister() {
625	<	return doArrive(ONE_ARRIVAL|ONE_PARTY);
625	>	return doArrive(ONE_DEREGISTER);
626		}
627
628		/**
#	Line 583 | Line 638 | public class Phaser {
638		* IllegalStateException} only upon some subsequent operation on
639		* this phaser, if ever.
640		*
641	<	* @return the arrival phase number, or a negative number if terminated
641	>	* @return the arrival phase number, or the (negative)
642	>	* {@linkplain #getPhase() current phase} if terminated
643		* @throws IllegalStateException if not terminated and the number
644		* of unarrived parties would become negative
645		*/
646		public int arriveAndAwaitAdvance() {
647	<	return awaitAdvance(doArrive(ONE_ARRIVAL));
647	>	// Specialization of doArrive+awaitAdvance eliminating some reads/paths
648	>	final Phaser root = this.root;
649	>	for (;;) {
650	>	long s = (root == this) ? state : reconcileState();
651	>	int phase = (int)(s >>> PHASE_SHIFT);
652	>	if (phase < 0)
653	>	return phase;
654	>	int counts = (int)s;
655	>	int unarrived = (counts == EMPTY) ? 0 : (counts & UNARRIVED_MASK);
656	>	if (unarrived <= 0)
657	>	throw new IllegalStateException(badArrive(s));
658	>	if (UNSAFE.compareAndSwapLong(this, stateOffset, s,
659	>	s -= ONE_ARRIVAL)) {
660	>	if (unarrived > 1)
661	>	return root.internalAwaitAdvance(phase, null);
662	>	if (root != this)
663	>	return parent.arriveAndAwaitAdvance();
664	>	long n = s & PARTIES_MASK;  // base of next state
665	>	int nextUnarrived = (int)n >>> PARTIES_SHIFT;
666	>	if (onAdvance(phase, nextUnarrived))
667	>	n |= TERMINATION_BIT;
668	>	else if (nextUnarrived == 0)
669	>	n |= EMPTY;
670	>	else
671	>	n |= nextUnarrived;
672	>	int nextPhase = (phase + 1) & MAX_PHASE;
673	>	n |= (long)nextPhase << PHASE_SHIFT;
674	>	if (!UNSAFE.compareAndSwapLong(this, stateOffset, s, n))
675	>	return (int)(state >>> PHASE_SHIFT); // terminated
676	>	releaseWaiters(phase);
677	>	return nextPhase;
678	>	}
679	>	}
680		}
681
682		/**
#	Line 599 | Line 687 | public class Phaser {
687		* @param phase an arrival phase number, or negative value if
688		* terminated; this argument is normally the value returned by a
689		* previous call to {@code arrive} or {@code arriveAndDeregister}.
690	<	* @return the next arrival phase number, or a negative value
691	<	* if terminated or argument is negative
690	>	* @return the next arrival phase number, or the argument if it is
691	>	* negative, or the (negative) {@linkplain #getPhase() current phase}
692	>	* if terminated
693		*/
694		public int awaitAdvance(int phase) {
695	<	Phaser rt;
696	<	int p = (int)(state >>> PHASE_SHIFT);
695	>	final Phaser root = this.root;
696	>	long s = (root == this) ? state : reconcileState();
697	>	int p = (int)(s >>> PHASE_SHIFT);
698		if (phase < 0)
699		return phase;
700	<	if (p == phase &&
701	<	(p = (int)((rt = root).state >>> PHASE_SHIFT)) == phase)
612	<	return rt.internalAwaitAdvance(phase, null);
700	>	if (p == phase)
701	>	return root.internalAwaitAdvance(phase, null);
702		return p;
703		}
704
#	Line 623 | Line 712 | public class Phaser {
712		* @param phase an arrival phase number, or negative value if
713		* terminated; this argument is normally the value returned by a
714		* previous call to {@code arrive} or {@code arriveAndDeregister}.
715	<	* @return the next arrival phase number, or a negative value
716	<	* if terminated or argument is negative
715	>	* @return the next arrival phase number, or the argument if it is
716	>	* negative, or the (negative) {@linkplain #getPhase() current phase}
717	>	* if terminated
718		* @throws InterruptedException if thread interrupted while waiting
719		*/
720		public int awaitAdvanceInterruptibly(int phase)
721		throws InterruptedException {
722	<	Phaser rt;
723	<	int p = (int)(state >>> PHASE_SHIFT);
722	>	final Phaser root = this.root;
723	>	long s = (root == this) ? state : reconcileState();
724	>	int p = (int)(s >>> PHASE_SHIFT);
725		if (phase < 0)
726		return phase;
727	<	if (p == phase &&
637	<	(p = (int)((rt = root).state >>> PHASE_SHIFT)) == phase) {
727	>	if (p == phase) {
728		QNode node = new QNode(this, phase, true, false, 0L);
729	<	p = rt.internalAwaitAdvance(phase, node);
729	>	p = root.internalAwaitAdvance(phase, node);
730		if (node.wasInterrupted)
731		throw new InterruptedException();
732		}
#	Line 657 | Line 747 | public class Phaser {
747		*        {@code unit}
748		* @param unit a {@code TimeUnit} determining how to interpret the
749		*        {@code timeout} parameter
750	<	* @return the next arrival phase number, or a negative value
751	<	* if terminated or argument is negative
750	>	* @return the next arrival phase number, or the argument if it is
751	>	* negative, or the (negative) {@linkplain #getPhase() current phase}
752	>	* if terminated
753		* @throws InterruptedException if thread interrupted while waiting
754		* @throws TimeoutException if timed out while waiting
755		*/
#	Line 666 | Line 757 | public class Phaser {
757		long timeout, TimeUnit unit)
758		throws InterruptedException, TimeoutException {
759		long nanos = unit.toNanos(timeout);
760	<	Phaser rt;
761	<	int p = (int)(state >>> PHASE_SHIFT);
760	>	final Phaser root = this.root;
761	>	long s = (root == this) ? state : reconcileState();
762	>	int p = (int)(s >>> PHASE_SHIFT);
763		if (phase < 0)
764		return phase;
765	<	if (p == phase &&
674	<	(p = (int)((rt = root).state >>> PHASE_SHIFT)) == phase) {
765	>	if (p == phase) {
766		QNode node = new QNode(this, phase, true, true, nanos);
767	<	p = rt.internalAwaitAdvance(phase, node);
767	>	p = root.internalAwaitAdvance(phase, node);
768		if (node.wasInterrupted)
769		throw new InterruptedException();
770		else if (p == phase)
#	Line 684 | Line 775 | public class Phaser {
775
776		/**
777		* Forces this phaser to enter termination state.  Counts of
778	<	* arrived and registered parties are unaffected.  If this phaser
779	<	* is a member of a tiered set of phasers, then all of the phasers
780	<	* in the set are terminated.  If this phaser is already
781	<	* terminated, this method has no effect.  This method may be
782	<	* useful for coordinating recovery after one or more tasks
783	<	* encounter unexpected exceptions.
778	>	* registered parties are unaffected.  If this phaser is a member
779	>	* of a tiered set of phasers, then all of the phasers in the set
780	>	* are terminated.  If this phaser is already terminated, this
781	>	* method has no effect.  This method may be useful for
782	>	* coordinating recovery after one or more tasks encounter
783	>	* unexpected exceptions.
784		*/
785		public void forceTermination() {
786		// Only need to change root state
#	Line 698 | Line 789 | public class Phaser {
789		while ((s = root.state) >= 0) {
790		if (UNSAFE.compareAndSwapLong(root, stateOffset,
791		s, s | TERMINATION_BIT)) {
792	<	releaseWaiters(0); // signal all threads
793	<	releaseWaiters(1);
792	>	// signal all threads
793	>	releaseWaiters(0); // Waiters on evenQ
794	>	releaseWaiters(1); // Waiters on oddQ
795		return;
796		}
797		}
#	Line 729 | Line 821 | public class Phaser {
821
822		/**
823		* Returns the number of registered parties that have arrived at
824	<	* the current phase of this phaser.
824	>	* the current phase of this phaser. If this phaser has terminated,
825	>	* the returned value is meaningless and arbitrary.
826		*
827		* @return the number of arrived parties
828		*/
829		public int getArrivedParties() {
830	<	long s = state;
738	<	int u = unarrivedOf(s); // only reconcile if possibly needed
739	<	return (u != 0 || parent == null) ?
740	<	partiesOf(s) - u :
741	<	arrivedOf(reconcileState());
830	>	return arrivedOf(reconcileState());
831		}
832
833		/**
834		* Returns the number of registered parties that have not yet
835	<	* arrived at the current phase of this phaser.
835	>	* arrived at the current phase of this phaser. If this phaser has
836	>	* terminated, the returned value is meaningless and arbitrary.
837		*
838		* @return the number of unarrived parties
839		*/
840		public int getUnarrivedParties() {
841	<	int u = unarrivedOf(state);
752	<	return (u != 0 || parent == null) ? u : unarrivedOf(reconcileState());
841	>	return unarrivedOf(reconcileState());
842		}
843
844		/**
#	Line 785 | Line 874 | public class Phaser {
874		* advance, and to control termination. This method is invoked
875		* upon arrival of the party advancing this phaser (when all other
876		* waiting parties are dormant).  If this method returns {@code
877	<	* true}, then, rather than advance the phase number, this phaser
878	<	* will be set to a final termination state, and subsequent calls
879	<	* to {@link #isTerminated} will return true. Any (unchecked)
880	<	* Exception or Error thrown by an invocation of this method is
881	<	* propagated to the party attempting to advance this phaser, in
882	<	* which case no advance occurs.
877	>	* true}, this phaser will be set to a final termination state
878	>	* upon advance, and subsequent calls to {@link #isTerminated}
879	>	* will return true. Any (unchecked) Exception or Error thrown by
880	>	* an invocation of this method is propagated to the party
881	>	* attempting to advance this phaser, in which case no advance
882	>	* occurs.
883		*
884		* <p>The arguments to this method provide the state of the phaser
885		* prevailing for the current transition.  The effects of invoking
#	Line 854 | Line 943 | public class Phaser {
943		*/
944		private void releaseWaiters(int phase) {
945		QNode q;   // first element of queue
857	–	int p;     // its phase
946		Thread t;  // its thread
947		AtomicReference<QNode> head = (phase & 1) == 0 ? evenQ : oddQ;
948		while ((q = head.get()) != null &&
949	<	((p = q.phase) == phase ||
862	<	(int)(root.state >>> PHASE_SHIFT) != p)) {
949	>	q.phase != (int)(root.state >>> PHASE_SHIFT)) {
950		if (head.compareAndSet(q, q.next) &&
951		(t = q.thread) != null) {
952		q.thread = null;
#	Line 868 | Line 955 | public class Phaser {
955		}
956		}
957
958	+	/**
959	+	* Variant of releaseWaiters that additionally tries to remove any
960	+	* nodes no longer waiting for advance due to timeout or
961	+	* interrupt. Currently, nodes are removed only if they are at
962	+	* head of queue, which suffices to reduce memory footprint in
963	+	* most usages.
964	+	*
965	+	* @return current phase on exit
966	+	*/
967	+	private int abortWait(int phase) {
968	+	AtomicReference<QNode> head = (phase & 1) == 0 ? evenQ : oddQ;
969	+	for (;;) {
970	+	Thread t;
971	+	QNode q = head.get();
972	+	int p = (int)(root.state >>> PHASE_SHIFT);
973	+	if (q == null || ((t = q.thread) != null && q.phase == p))
974	+	return p;
975	+	if (head.compareAndSet(q, q.next) && t != null) {
976	+	q.thread = null;
977	+	LockSupport.unpark(t);
978	+	}
979	+	}
980	+	}
981	+
982		/** The number of CPUs, for spin control */
983		private static final int NCPU = Runtime.getRuntime().availableProcessors();
984
#	Line 886 | Line 997 | public class Phaser {
997
998		/**
999		* Possibly blocks and waits for phase to advance unless aborted.
1000	<	* Call only from root node.
1000	>	* Call only on root phaser.
1001		*
1002		* @param phase current phase
1003		* @param node if non-null, the wait node to track interrupt and timeout;
#	Line 894 | Line 1005 | public class Phaser {
1005		* @return current phase
1006		*/
1007		private int internalAwaitAdvance(int phase, QNode node) {
1008	+	// assert root == this;
1009		releaseWaiters(phase-1);          // ensure old queue clean
1010		boolean queued = false;           // true when node is enqueued
1011		int lastUnarrived = 0;            // to increase spins upon change
#	Line 935 | Line 1047 | public class Phaser {
1047		node.thread = null;       // avoid need for unpark()
1048		if (node.wasInterrupted && !node.interruptible)
1049		Thread.currentThread().interrupt();
1050	<	if ((p = (int)(state >>> PHASE_SHIFT)) == phase)
1051	<	return p;                 // recheck abort
1050	>	if (p == phase && (p = (int)(state >>> PHASE_SHIFT)) == phase)
1051	>	return abortWait(phase); // possibly clean up on abort
1052		}
1053		releaseWaiters(phase);
1054		return p;
#	Line 1007 | Line 1119 | public class Phaser {
1119
1120		// Unsafe mechanics
1121
1122	<	private static final sun.misc.Unsafe UNSAFE = getUnsafe();
1123	<	private static final long stateOffset =
1124	<	objectFieldOffset("state", Phaser.class);
1013	<
1014	<	private static long objectFieldOffset(String field, Class<?> klazz) {
1122	>	private static final sun.misc.Unsafe UNSAFE;
1123	>	private static final long stateOffset;
1124	>	static {
1125		try {
1126	<	return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
1127	<	} catch (NoSuchFieldException e) {
1128	<	// Convert Exception to corresponding Error
1129	<	NoSuchFieldError error = new NoSuchFieldError(field);
1130	<	error.initCause(e);
1131	<	throw error;
1126	>	UNSAFE = getUnsafe();
1127	>	Class<?> k = Phaser.class;
1128	>	stateOffset = UNSAFE.objectFieldOffset
1129	>	(k.getDeclaredField("state"));
1130	>	} catch (Exception e) {
1131	>	throw new Error(e);
1132		}
1133		}
1134
#	Line 1032 | Line 1142 | public class Phaser {
1142		private static sun.misc.Unsafe getUnsafe() {
1143		try {
1144		return sun.misc.Unsafe.getUnsafe();
1145	<	} catch (SecurityException se) {
1146	<	try {
1147	<	return java.security.AccessController.doPrivileged
1148	<	(new java.security
1149	<	.PrivilegedExceptionAction<sun.misc.Unsafe>() {
1150	<	public sun.misc.Unsafe run() throws Exception {
1151	<	java.lang.reflect.Field f = sun.misc
1152	<	.Unsafe.class.getDeclaredField("theUnsafe");
1153	<	f.setAccessible(true);
1154	<	return (sun.misc.Unsafe) f.get(null);
1155	<	}});
1156	<	} catch (java.security.PrivilegedActionException e) {
1157	<	throw new RuntimeException("Could not initialize intrinsics",
1158	<	e.getCause());
1159	<	}
1145	>	} catch (SecurityException tryReflectionInstead) {}
1146	>	try {
1147	>	return java.security.AccessController.doPrivileged
1148	>	(new java.security.PrivilegedExceptionAction<sun.misc.Unsafe>() {
1149	>	public sun.misc.Unsafe run() throws Exception {
1150	>	Class<sun.misc.Unsafe> k = sun.misc.Unsafe.class;
1151	>	for (java.lang.reflect.Field f : k.getDeclaredFields()) {
1152	>	f.setAccessible(true);
1153	>	Object x = f.get(null);
1154	>	if (k.isInstance(x))
1155	>	return k.cast(x);
1156	>	}
1157	>	throw new NoSuchFieldError("the Unsafe");
1158	>	}});
1159	>	} catch (java.security.PrivilegedActionException e) {
1160	>	throw new RuntimeException("Could not initialize intrinsics",
1161	>	e.getCause());
1162		}
1163		}
1164		}

Diff Legend

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