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 1.6 by dl, Tue Oct 28 23:03:24 2008 UTC vs.
Revision 1.21 by jsr166, Sun Jul 26 05:55:34 2009 UTC

#	Line 5 | Line 5
5		*/
6
7		package jsr166y;
8	+
9		import java.util.concurrent.*;
10	<	import java.util.concurrent.atomic.*;
10	>
11	>	import java.util.concurrent.atomic.AtomicReference;
12		import java.util.concurrent.locks.LockSupport;
11	–	import sun.misc.Unsafe;
12	–	import java.lang.reflect.*;
13
14		/**
15		* A reusable synchronization barrier, similar in functionality to a
16	<	* {@link java.util.concurrent.CyclicBarrier} and {@link
17	<	* java.util.concurrent.CountDownLatch} but supporting more flexible
18	<	* usage.
16	>	* {@link java.util.concurrent.CyclicBarrier CyclicBarrier} and
17	>	* {@link java.util.concurrent.CountDownLatch CountDownLatch}
18	>	* but supporting more flexible usage.
19		*
20		* <ul>
21		*
#	Line 25 | Line 25 | import java.lang.reflect.*;
25		* basic synchronization constructs, registration and deregistration
26		* affect only internal counts; they do not establish any further
27		* internal bookkeeping, so tasks cannot query whether they are
28	<	* registered. (However, you can introduce such bookkeeping in by
28	>	* registered. (However, you can introduce such bookkeeping by
29		* subclassing this class.)
30		*
31		* <li> Each generation has an associated phase value, starting at
32		* zero, and advancing when all parties reach the barrier (wrapping
33	<	* around to zero after reaching <tt>Integer.MAX_VALUE</tt>).
33	>	* around to zero after reaching {@code Integer.MAX_VALUE}).
34		*
35		* <li> Like a CyclicBarrier, a Phaser may be repeatedly awaited.
36	<	* Method <tt>arriveAndAwaitAdvance</tt> has effect analogous to
37	<	* <tt>CyclicBarrier.await</tt>.  However, Phasers separate two
36	>	* Method {@code arriveAndAwaitAdvance} has effect analogous to
37	>	* {@code CyclicBarrier.await}.  However, Phasers separate two
38		* aspects of coordination, that may also be invoked independently:
39		*
40		* <ul>
41		*
42	<	*   <li> Arriving at a barrier. Methods <tt>arrive</tt> and
43	<	*       <tt>arriveAndDeregister</tt> do not block, but return
42	>	*   <li> Arriving at a barrier. Methods {@code arrive} and
43	>	*       {@code arriveAndDeregister} do not block, but return
44		*       the phase value current upon entry to the method.
45		*
46	<	*   <li> Awaiting others. Method <tt>awaitAdvance</tt> requires an
46	>	*   <li> Awaiting others. Method {@code awaitAdvance} requires an
47		*       argument indicating the entry phase, and returns when the
48		*       barrier advances to a new phase.
49		* </ul>
#	Line 51 | Line 51 | import java.lang.reflect.*;
51		*
52		* <li> Barrier actions, performed by the task triggering a phase
53		* advance while others may be waiting, are arranged by overriding
54	<	* method <tt>onAdvance</tt>, that also controls termination.
54	>	* method {@code onAdvance}, that also controls termination.
55		* Overriding this method may be used to similar but more flexible
56		* effect as providing a barrier action to a CyclicBarrier.
57		*
58		* <li> Phasers may enter a <em>termination</em> state in which all
59	<	* await actions immediately return, indicating (via a negative phase
60	<	* value) that execution is complete.  Termination is triggered by
61	<	* executing the overridable <tt>onAdvance</tt> method that is invoked
62	<	* each time the barrier is about to be tripped. When a Phaser is
63	<	* controlling an action with a fixed number of iterations, it is
64	<	* often convenient to override this method to cause termination when
65	<	* the current phase number reaches a threshold. Method
66	<	* <tt>forceTermination</tt> is also available to abruptly release
67	<	* waiting threads and allow them to terminate.
59	>	* actions immediately return without updating phaser state or waiting
60	>	* for advance, and indicating (via a negative phase value) that
61	>	* execution is complete.  Termination is triggered by executing the
62	>	* overridable {@code onAdvance} method that is invoked each time the
63	>	* barrier is about to be tripped. When a Phaser is controlling an
64	>	* action with a fixed number of iterations, it is often convenient to
65	>	* override this method to cause termination when the current phase
66	>	* number reaches a threshold. Method {@code forceTermination} is also
67	>	* available to abruptly release waiting threads and allow them to
68	>	* terminate.
69		*
70		* <li> Phasers may be tiered to reduce contention. Phasers with large
71		* numbers of parties that would otherwise experience heavy
#	Line 72 | Line 73 | import java.lang.reflect.*;
73		* This will typically greatly increase throughput even though it
74		* incurs somewhat greater per-operation overhead.
75		*
76	<	* <li> By default, <tt>awaitAdvance</tt> continues to wait even if
76	>	* <li> By default, {@code awaitAdvance} continues to wait even if
77		* the waiting thread is interrupted. And unlike the case in
78		* CyclicBarriers, exceptions encountered while tasks wait
79		* interruptibly or with timeout do not change the state of the
80		* barrier. If necessary, you can perform any associated recovery
81		* within handlers of those exceptions, often after invoking
82	<	* <tt>forceTermination</tt>.
82	>	* {@code forceTermination}.
83	>	*
84	>	* <li>Phasers ensure lack of starvation when used by ForkJoinTasks.
85		*
86		* </ul>
87		*
88		* <p><b>Sample usages:</b>
89		*
90	<	* <p>A Phaser may be used instead of a <tt>CountdownLatch</tt> to control
90	>	* <p>A Phaser may be used instead of a {@code CountDownLatch} to control
91		* a one-shot action serving a variable number of parties. The typical
92		* idiom is for the method setting this up to first register, then
93		* start the actions, then deregister, as in:
94		*
95	<	* <pre>
96	<	*  void runTasks(List&lt;Runnable&gt; list) {
97	<	*    final Phaser phaser = new Phaser(1); // "1" to register self
98	<	*    for (Runnable r : list) {
99	<	*      phaser.register();
100	<	*      new Thread() {
101	<	*        public void run() {
102	<	*          phaser.arriveAndAwaitAdvance(); // await all creation
103	<	*          r.run();
104	<	*          phaser.arriveAndDeregister();   // signal completion
105	<	*        }
106	<	*      }.start();
95	>	*  <pre> {@code
96	>	* void runTasks(List<Runnable> list) {
97	>	*   final Phaser phaser = new Phaser(1); // "1" to register self
98	>	*   for (Runnable r : list) {
99	>	*     phaser.register();
100	>	*     new Thread() {
101	>	*       public void run() {
102	>	*         phaser.arriveAndAwaitAdvance(); // await all creation
103	>	*         r.run();
104	>	*         phaser.arriveAndDeregister();   // signal completion
105	>	*       }
106	>	*     }.start();
107		*   }
108		*
109		*   doSomethingOnBehalfOfWorkers();
#	Line 108 | Line 111 | import java.lang.reflect.*;
111		*   int p = phaser.arriveAndDeregister(); // deregister self  ...
112		*   p = phaser.awaitAdvance(p); // ... and await arrival
113		*   otherActions(); // do other things while tasks execute
114	<	*   phaser.awaitAdvance(p); // awit final completion
115	<	* }
113	<	* </pre>
114	>	*   phaser.awaitAdvance(p); // await final completion
115	>	* }}</pre>
116		*
117		* <p>One way to cause a set of threads to repeatedly perform actions
118	<	* for a given number of iterations is to override <tt>onAdvance</tt>:
118	>	* for a given number of iterations is to override {@code onAdvance}:
119		*
120	<	* <pre>
121	<	*  void startTasks(List&lt;Runnable&gt; list, final int iterations) {
122	<	*    final Phaser phaser = new Phaser() {
123	<	*       public boolean onAdvance(int phase, int registeredParties) {
124	<	*         return phase &gt;= iterations || registeredParties == 0;
120	>	*  <pre> {@code
121	>	* void startTasks(List<Runnable> list, final int iterations) {
122	>	*   final Phaser phaser = new Phaser() {
123	>	*     public boolean onAdvance(int phase, int registeredParties) {
124	>	*       return phase >= iterations || registeredParties == 0;
125	>	*     }
126	>	*   };
127	>	*   phaser.register();
128	>	*   for (Runnable r : list) {
129	>	*     phaser.register();
130	>	*     new Thread() {
131	>	*       public void run() {
132	>	*         do {
133	>	*           r.run();
134	>	*           phaser.arriveAndAwaitAdvance();
135	>	*         } while(!phaser.isTerminated();
136		*       }
137	<	*    };
125	<	*    phaser.register();
126	<	*    for (Runnable r : list) {
127	<	*      phaser.register();
128	<	*      new Thread() {
129	<	*        public void run() {
130	<	*           do {
131	<	*             r.run();
132	<	*             phaser.arriveAndAwaitAdvance();
133	<	*           } while(!phaser.isTerminated();
134	<	*        }
135	<	*      }.start();
137	>	*     }.start();
138		*   }
139		*   phaser.arriveAndDeregister(); // deregister self, don't wait
140	<	* }
139	<	* </pre>
140	>	* }}</pre>
141		*
142		* <p> To create a set of tasks using a tree of Phasers,
143		* you could use code of the following form, assuming a
144		* Task class with a constructor accepting a Phaser that
145		* it registers for upon construction:
146	<	* <pre>
147	<	*  void build(Task[] actions, int lo, int hi, Phaser b) {
148	<	*    int step = (hi - lo) / TASKS_PER_PHASER;
149	<	*    if (step &gt; 1) {
150	<	*       int i = lo;
151	<	*       while (i &lt; hi) {
152	<	*         int r = Math.min(i + step, hi);
153	<	*         build(actions, i, r, new Phaser(b));
154	<	*         i = r;
155	<	*       }
156	<	*    }
157	<	*    else {
158	<	*      for (int i = lo; i &lt; hi; ++i)
159	<	*        actions[i] = new Task(b);
160	<	*        // assumes new Task(b) performs b.register()
161	<	*    }
162	<	*  }
163	<	*  // .. initially called, for n tasks via
163	<	*  build(new Task[n], 0, n, new Phaser());
164	<	* </pre>
146	>	*  <pre> {@code
147	>	* void build(Task[] actions, int lo, int hi, Phaser b) {
148	>	*   int step = (hi - lo) / TASKS_PER_PHASER;
149	>	*   if (step > 1) {
150	>	*     int i = lo;
151	>	*     while (i < hi) {
152	>	*       int r = Math.min(i + step, hi);
153	>	*       build(actions, i, r, new Phaser(b));
154	>	*       i = r;
155	>	*     }
156	>	*   } else {
157	>	*     for (int i = lo; i < hi; ++i)
158	>	*       actions[i] = new Task(b);
159	>	*       // assumes new Task(b) performs b.register()
160	>	*   }
161	>	* }
162	>	* // .. initially called, for n tasks via
163	>	* build(new Task[n], 0, n, new Phaser());}</pre>
164		*
165	<	* The best value of <tt>TASKS_PER_PHASER</tt> depends mainly on
165	>	* The best value of {@code TASKS_PER_PHASER} depends mainly on
166		* expected barrier synchronization rates. A value as low as four may
167		* be appropriate for extremely small per-barrier task bodies (thus
168		* high rates), or up to hundreds for extremely large ones.
#	Line 175 | Line 174 | import java.lang.reflect.*;
174		* parties result in IllegalStateExceptions. However, you can and
175		* should create tiered phasers to accommodate arbitrarily large sets
176		* of participants.
177	+	*
178	+	* @since 1.7
179	+	* @author Doug Lea
180		*/
181		public class Phaser {
182		/*
#	Line 195 | Line 197 | public class Phaser {
197		* However, to efficiently maintain atomicity, these values are
198		* packed into a single (atomic) long. Termination uses the sign
199		* bit of 32 bit representation of phase, so phase is set to -1 on
200	<	* termination. Good performace relies on keeping state decoding
200	>	* termination. Good performance relies on keeping state decoding
201		* and encoding simple, and keeping race windows short.
202		*
203		* Note: there are some cheats in arrive() that rely on unarrived
204	<	* being lowest 16 bits.
204	>	* count being lowest 16 bits.
205		*/
206		private volatile long state;
207
208		private static final int ushortBits = 16;
209	<	private static final int ushortMask =  (1 << ushortBits) - 1;
210	<	private static final int phaseMask = 0x7fffffff;
209	>	private static final int ushortMask = 0xffff;
210	>	private static final int phaseMask  = 0x7fffffff;
211
212		private static int unarrivedOf(long s) {
213	<	return (int)(s & ushortMask);
213	>	return (int) (s & ushortMask);
214		}
215
216		private static int partiesOf(long s) {
217	<	return (int)(s & (ushortMask << 16)) >>> 16;
217	>	return ((int) s) >>> 16;
218		}
219
220		private static int phaseOf(long s) {
221	<	return (int)(s >>> 32);
221	>	return (int) (s >>> 32);
222		}
223
224		private static int arrivedOf(long s) {
#	Line 224 | Line 226 | public class Phaser {
226		}
227
228		private static long stateFor(int phase, int parties, int unarrived) {
229	<	return (((long)phase) << 32) | ((parties << 16) | unarrived);
229	>	return ((((long) phase) << 32) | (((long) parties) << 16) |
230	>	(long) unarrived);
231		}
232
233		private static long trippedStateFor(int phase, int parties) {
234	<	return (((long)phase) << 32) | ((parties << 16) | parties);
234	>	long lp = (long) parties;
235	>	return (((long) phase) << 32) | (lp << 16) | lp;
236		}
237
238	<	private static IllegalStateException badBounds(int parties, int unarrived) {
239	<	return new IllegalStateException
240	<	("Attempt to set " + unarrived +
241	<	" unarrived of " + parties + " parties");
238	>	/**
239	>	* Returns message string for bad bounds exceptions.
240	>	*/
241	>	private static String badBounds(int parties, int unarrived) {
242	>	return ("Attempt to set " + unarrived +
243	>	" unarrived of " + parties + " parties");
244		}
245
246		/**
#	Line 251 | Line 257 | public class Phaser {
257		// Wait queues
258
259		/**
260	<	* Heads of Treiber stacks waiting for nonFJ threads. To eliminate
260	>	* Heads of Treiber stacks for waiting threads. To eliminate
261		* contention while releasing some threads while adding others, we
262		* use two of them, alternating across even and odd phases.
263		*/
#	Line 259 | Line 265 | public class Phaser {
265		private final AtomicReference<QNode> oddQ  = new AtomicReference<QNode>();
266
267		private AtomicReference<QNode> queueFor(int phase) {
268	<	return (phase & 1) == 0? evenQ : oddQ;
268	>	return ((phase & 1) == 0) ? evenQ : oddQ;
269		}
270
271		/**
#	Line 267 | Line 273 | public class Phaser {
273		* root if necessary.
274		*/
275		private long getReconciledState() {
276	<	return parent == null? state : reconcileState();
276	>	return (parent == null) ? state : reconcileState();
277		}
278
279		/**
#	Line 295 | Line 301 | public class Phaser {
301
302		/**
303		* Creates a new Phaser without any initially registered parties,
304	<	* initial phase number 0, and no parent.
304	>	* initial phase number 0, and no parent. Any thread using this
305	>	* Phaser will need to first register for it.
306		*/
307		public Phaser() {
308		this(null);
#	Line 304 | Line 311 | public class Phaser {
311		/**
312		* Creates a new Phaser with the given numbers of registered
313		* unarrived parties, initial phase number 0, and no parent.
314	<	* @param parties the number of parties required to trip barrier.
314	>	*
315	>	* @param parties the number of parties required to trip barrier
316		* @throws IllegalArgumentException if parties less than zero
317	<	* or greater than the maximum number of parties supported.
317	>	* or greater than the maximum number of parties supported
318		*/
319		public Phaser(int parties) {
320		this(null, parties);
#	Line 317 | Line 325 | public class Phaser {
325		* initially registered parties. If parent is non-null this phaser
326		* is registered with the parent and its initial phase number is
327		* the same as that of parent phaser.
328	<	* @param parent the parent phaser.
328	>	*
329	>	* @param parent the parent phaser
330		*/
331		public Phaser(Phaser parent) {
332		int phase = 0;
#	Line 333 | Line 342 | public class Phaser {
342
343		/**
344		* Creates a new Phaser with the given parent and numbers of
345	<	* registered unarrived parties. If parent is non-null this phaser
345	>	* registered unarrived parties. If parent is non-null, this phaser
346		* is registered with the parent and its initial phase number is
347		* the same as that of parent phaser.
348	<	* @param parent the parent phaser.
349	<	* @param parties the number of parties required to trip barrier.
348	>	*
349	>	* @param parent the parent phaser
350	>	* @param parties the number of parties required to trip barrier
351		* @throws IllegalArgumentException if parties less than zero
352	<	* or greater than the maximum number of parties supported.
352	>	* or greater than the maximum number of parties supported
353		*/
354		public Phaser(Phaser parent, int parties) {
355		if (parties < 0 || parties > ushortMask)
#	Line 357 | Line 367 | public class Phaser {
367
368		/**
369		* Adds a new unarrived party to this phaser.
370	+	*
371		* @return the current barrier phase number upon registration
372		* @throws IllegalStateException if attempting to register more
373	<	* than the maximum supported number of parties.
373	>	* than the maximum supported number of parties
374		*/
375		public int register() {
376		return doRegister(1);
#	Line 367 | Line 378 | public class Phaser {
378
379		/**
380		* Adds the given number of new unarrived parties to this phaser.
381	<	* @param parties the number of parties required to trip barrier.
381	>	*
382	>	* @param parties the number of parties required to trip barrier
383		* @return the current barrier phase number upon registration
384		* @throws IllegalStateException if attempting to register more
385	<	* than the maximum supported number of parties.
385	>	* than the maximum supported number of parties
386		*/
387		public int bulkRegister(int parties) {
388		if (parties < 0)
#	Line 393 | Line 405 | public class Phaser {
405		if (phase < 0)
406		break;
407		if (parties > ushortMask || unarrived > ushortMask)
408	<	throw badBounds(parties, unarrived);
408	>	throw new IllegalStateException(badBounds(parties, unarrived));
409		if (phase == phaseOf(root.state) &&
410		casState(s, stateFor(phase, parties, unarrived)))
411		break;
#	Line 406 | Line 418 | public class Phaser {
418		* in turn wait for others via {@link #awaitAdvance}).
419		*
420		* @return the barrier phase number upon entry to this method, or a
421	<	* negative value if terminated;
421	>	* negative value if terminated
422		* @throws IllegalStateException if not terminated and the number
423	<	* of unarrived parties would become negative.
423	>	* of unarrived parties would become negative
424		*/
425		public int arrive() {
426		int phase;
427		for (;;) {
428		long s = state;
429		phase = phaseOf(s);
430	+	if (phase < 0)
431	+	break;
432		int parties = partiesOf(s);
433		int unarrived = unarrivedOf(s) - 1;
434		if (unarrived > 0) {        // Not the last arrival
#	Line 426 | Line 440 | public class Phaser {
440		if (par == null) {      // directly trip
441		if (casState
442		(s,
443	<	trippedStateFor(onAdvance(phase, parties)? -1 :
443	>	trippedStateFor(onAdvance(phase, parties) ? -1 :
444		((phase + 1) & phaseMask), parties))) {
445		releaseWaiters(phase);
446		break;
#	Line 440 | Line 454 | public class Phaser {
454		}
455		}
456		}
443	–	else if (phase < 0) // Don't throw exception if terminated
444	–	break;
457		else if (phase != phaseOf(root.state)) // or if unreconciled
458		reconcileState();
459		else
460	<	throw badBounds(parties, unarrived);
460	>	throw new IllegalStateException(badBounds(parties, unarrived));
461		}
462		return phase;
463		}
#	Line 458 | Line 470 | public class Phaser {
470		* zero parties, this phaser is also deregistered from its parent.
471		*
472		* @return the current barrier phase number upon entry to
473	<	* this method, or a negative value if terminated;
473	>	* this method, or a negative value if terminated
474		* @throws IllegalStateException if not terminated and the number
475	<	* of registered or unarrived parties would become negative.
475	>	* of registered or unarrived parties would become negative
476		*/
477		public int arriveAndDeregister() {
478		// similar code to arrive, but too different to merge
#	Line 469 | Line 481 | public class Phaser {
481		for (;;) {
482		long s = state;
483		phase = phaseOf(s);
484	+	if (phase < 0)
485	+	break;
486		int parties = partiesOf(s) - 1;
487		int unarrived = unarrivedOf(s) - 1;
488		if (parties >= 0) {
#	Line 487 | Line 501 | public class Phaser {
501		if (unarrived == 0) {
502		if (casState
503		(s,
504	<	trippedStateFor(onAdvance(phase, parties)? -1 :
504	>	trippedStateFor(onAdvance(phase, parties) ? -1 :
505		((phase + 1) & phaseMask), parties))) {
506		releaseWaiters(phase);
507		break;
508		}
509		continue;
510		}
497	–	if (phase < 0)
498	–	break;
511		if (par != null && phase != phaseOf(root.state)) {
512		reconcileState();
513		continue;
514		}
515		}
516	<	throw badBounds(parties, unarrived);
516	>	throw new IllegalStateException(badBounds(parties, unarrived));
517		}
518		return phase;
519		}
520
521		/**
522		* Arrives at the barrier and awaits others. Equivalent in effect
523	<	* to <tt>awaitAdvance(arrive())</tt>.  If you instead need to
523	>	* to {@code awaitAdvance(arrive())}.  If you instead need to
524		* await with interruption of timeout, and/or deregister upon
525		* arrival, you can arrange them using analogous constructions.
526	+	*
527		* @return the phase on entry to this method
528		* @throws IllegalStateException if not terminated and the number
529	<	* of unarrived parties would become negative.
529	>	* of unarrived parties would become negative
530		*/
531		public int arriveAndAwaitAdvance() {
532		return awaitAdvance(arrive());
#	Line 523 | Line 536 | public class Phaser {
536		* Awaits the phase of the barrier to advance from the given
537		* value, or returns immediately if argument is negative or this
538		* barrier is terminated.
539	+	*
540		* @param phase the phase on entry to this method
541		* @return the phase on exit from this method
542		*/
#	Line 533 | Line 547 | public class Phaser {
547		int p = phaseOf(s);
548		if (p != phase)
549		return p;
550	<	if (unarrivedOf(s) == 0)
550	>	if (unarrivedOf(s) == 0 && parent != null)
551		parent.awaitAdvance(phase);
552		// Fall here even if parent waited, to reconcile and help release
553		return untimedWait(phase);
#	Line 541 | Line 555 | public class Phaser {
555
556		/**
557		* Awaits the phase of the barrier to advance from the given
558	<	* value, or returns immediately if argumet is negative or this
558	>	* value, or returns immediately if argument is negative or this
559		* barrier is terminated, or throws InterruptedException if
560		* interrupted while waiting.
561	+	*
562		* @param phase the phase on entry to this method
563		* @return the phase on exit from this method
564		* @throws InterruptedException if thread interrupted while waiting
565		*/
566	<	public int awaitAdvanceInterruptibly(int phase) throws InterruptedException {
566	>	public int awaitAdvanceInterruptibly(int phase)
567	>	throws InterruptedException {
568		if (phase < 0)
569		return phase;
570		long s = getReconciledState();
571		int p = phaseOf(s);
572		if (p != phase)
573		return p;
574	<	if (unarrivedOf(s) != 0)
574	>	if (unarrivedOf(s) == 0 && parent != null)
575		parent.awaitAdvanceInterruptibly(phase);
576		return interruptibleWait(phase);
577		}
#	Line 564 | Line 580 | public class Phaser {
580		* Awaits the phase of the barrier to advance from the given value
581		* or the given timeout elapses, or returns immediately if
582		* argument is negative or this barrier is terminated.
583	+	*
584		* @param phase the phase on entry to this method
585		* @return the phase on exit from this method
586		* @throws InterruptedException if thread interrupted while waiting
587		* @throws TimeoutException if timed out while waiting
588		*/
589	<	public int awaitAdvanceInterruptibly(int phase, long timeout, TimeUnit unit)
589	>	public int awaitAdvanceInterruptibly(int phase,
590	>	long timeout, TimeUnit unit)
591		throws InterruptedException, TimeoutException {
592		if (phase < 0)
593		return phase;
#	Line 577 | Line 595 | public class Phaser {
595		int p = phaseOf(s);
596		if (p != phase)
597		return p;
598	<	if (unarrivedOf(s) == 0)
598	>	if (unarrivedOf(s) == 0 && parent != null)
599		parent.awaitAdvanceInterruptibly(phase, timeout, unit);
600		return timedWait(phase, unit.toNanos(timeout));
601		}
#	Line 608 | Line 626 | public class Phaser {
626
627		/**
628		* Returns the current phase number. The maximum phase number is
629	<	* <tt>Integer.MAX_VALUE</tt>, after which it restarts at
629	>	* {@code Integer.MAX_VALUE}, after which it restarts at
630		* zero. Upon termination, the phase number is negative.
631	+	*
632		* @return the phase number, or a negative value if terminated
633		*/
634		public final int getPhase() {
#	Line 617 | Line 636 | public class Phaser {
636		}
637
638		/**
639	<	* Returns true if the current phase number equals the given phase.
639	>	* Returns {@code true} if the current phase number equals the given phase.
640	>	*
641		* @param phase the phase
642	<	* @return true if the current phase number equals the given phase.
642	>	* @return {@code true} if the current phase number equals the given phase
643		*/
644		public final boolean hasPhase(int phase) {
645		return phaseOf(getReconciledState()) == phase;
#	Line 627 | Line 647 | public class Phaser {
647
648		/**
649		* Returns the number of parties registered at this barrier.
650	+	*
651		* @return the number of parties
652		*/
653		public int getRegisteredParties() {
#	Line 636 | Line 657 | public class Phaser {
657		/**
658		* Returns the number of parties that have arrived at the current
659		* phase of this barrier.
660	+	*
661		* @return the number of arrived parties
662		*/
663		public int getArrivedParties() {
#	Line 645 | Line 667 | public class Phaser {
667		/**
668		* Returns the number of registered parties that have not yet
669		* arrived at the current phase of this barrier.
670	+	*
671		* @return the number of unarrived parties
672		*/
673		public int getUnarrivedParties() {
#	Line 653 | Line 676 | public class Phaser {
676
677		/**
678		* Returns the parent of this phaser, or null if none.
679	<	* @return the parent of this phaser, or null if none.
679	>	*
680	>	* @return the parent of this phaser, or null if none
681		*/
682		public Phaser getParent() {
683		return parent;
#	Line 662 | Line 686 | public class Phaser {
686		/**
687		* Returns the root ancestor of this phaser, which is the same as
688		* this phaser if it has no parent.
689	<	* @return the root ancestor of this phaser.
689	>	*
690	>	* @return the root ancestor of this phaser
691		*/
692		public Phaser getRoot() {
693		return root;
694		}
695
696		/**
697	<	* Returns true if this barrier has been terminated.
698	<	* @return true if this barrier has been terminated
697	>	* Returns {@code true} if this barrier has been terminated.
698	>	*
699	>	* @return {@code true} if this barrier has been terminated
700		*/
701		public boolean isTerminated() {
702		return getPhase() < 0;
#	Line 682 | Line 708 | public class Phaser {
708		* barrier is tripped (and thus all other waiting parties are
709		* dormant). If it returns true, then, rather than advance the
710		* phase number, this barrier will be set to a final termination
711	<	* state, and subsequent calls to <tt>isTerminated</tt> will
711	>	* state, and subsequent calls to {@code isTerminated} will
712		* return true.
713		*
714		* <p> The default version returns true when the number of
#	Line 693 | Line 719 | public class Phaser {
719		* <p> You may override this method to perform an action with side
720		* effects visible to participating tasks, but it is in general
721		* only sensible to do so in designs where all parties register
722	<	* before any arrive, and all <tt>awaitAdvance</tt> at each phase.
722	>	* before any arrive, and all {@code awaitAdvance} at each phase.
723		* Otherwise, you cannot ensure lack of interference. In
724		* particular, this method may be invoked more than once per
725		* transition if other parties successfully register while the
#	Line 702 | Line 728 | public class Phaser {
728		* method.
729		*
730		* @param phase the phase number on entering the barrier
731	<	* @param registeredParties the current number of registered
732	<	* parties.
707	<	* @return true if this barrier should terminate
731	>	* @param registeredParties the current number of registered parties
732	>	* @return {@code true} if this barrier should terminate
733		*/
734		protected boolean onAdvance(int phase, int registeredParties) {
735		return registeredParties <= 0;
#	Line 713 | Line 738 | public class Phaser {
738		/**
739		* Returns a string identifying this phaser, as well as its
740		* state.  The state, in brackets, includes the String {@code
741	<	* "phase ="} followed by the phase number, {@code "parties ="}
741	>	* "phase = "} followed by the phase number, {@code "parties = "}
742		* followed by the number of registered parties, and {@code
743	<	* "arrived ="} followed by the number of arrived parties
743	>	* "arrived = "} followed by the number of arrived parties.
744		*
745		* @return a string identifying this barrier, as well as its state
746		*/
747		public String toString() {
748		long s = getReconciledState();
749	<	return super.toString() + "[phase = " + phaseOf(s) + " parties = " + partiesOf(s) + " arrived = " + arrivedOf(s) + "]";
749	>	return super.toString() +
750	>	"[phase = " + phaseOf(s) +
751	>	" parties = " + partiesOf(s) +
752	>	" arrived = " + arrivedOf(s) + "]";
753		}
754
755		// methods for waiting
756
729	–	/** The number of CPUs, for spin control */
730	–	static final int NCPUS = Runtime.getRuntime().availableProcessors();
731	–
732	–	/**
733	–	* The number of times to spin before blocking in timed waits.
734	–	* The value is empirically derived.
735	–	*/
736	–	static final int maxTimedSpins = (NCPUS < 2)? 0 : 32;
737	–
738	–	/**
739	–	* The number of times to spin before blocking in untimed waits.
740	–	* This is greater than timed value because untimed waits spin
741	–	* faster since they don't need to check times on each spin.
742	–	*/
743	–	static final int maxUntimedSpins = maxTimedSpins * 32;
744	–
745	–	/**
746	–	* The number of nanoseconds for which it is faster to spin
747	–	* rather than to use timed park. A rough estimate suffices.
748	–	*/
749	–	static final long spinForTimeoutThreshold = 1000L;
750	–
757		/**
758	<	* Wait nodes for Treiber stack representing wait queue for non-FJ
753	<	* tasks.
758	>	* Wait nodes for Treiber stack representing wait queue
759		*/
760	<	static final class QNode {
761	<	QNode next;
760	>	static final class QNode implements ForkJoinPool.ManagedBlocker {
761	>	final Phaser phaser;
762	>	final int phase;
763	>	final long startTime;
764	>	final long nanos;
765	>	final boolean timed;
766	>	final boolean interruptible;
767	>	volatile boolean wasInterrupted = false;
768		volatile Thread thread; // nulled to cancel wait
769	<	QNode() {
769	>	QNode next;
770	>	QNode(Phaser phaser, int phase, boolean interruptible,
771	>	boolean timed, long startTime, long nanos) {
772	>	this.phaser = phaser;
773	>	this.phase = phase;
774	>	this.timed = timed;
775	>	this.interruptible = interruptible;
776	>	this.startTime = startTime;
777	>	this.nanos = nanos;
778		thread = Thread.currentThread();
779		}
780	+	public boolean isReleasable() {
781	+	return (thread == null ||
782	+	phaser.getPhase() != phase ||
783	+	(interruptible && wasInterrupted) ||
784	+	(timed && (nanos - (System.nanoTime() - startTime)) <= 0));
785	+	}
786	+	public boolean block() {
787	+	if (Thread.interrupted()) {
788	+	wasInterrupted = true;
789	+	if (interruptible)
790	+	return true;
791	+	}
792	+	if (!timed)
793	+	LockSupport.park(this);
794	+	else {
795	+	long waitTime = nanos - (System.nanoTime() - startTime);
796	+	if (waitTime <= 0)
797	+	return true;
798	+	LockSupport.parkNanos(this, waitTime);
799	+	}
800	+	return isReleasable();
801	+	}
802		void signal() {
803		Thread t = thread;
804		if (t != null) {
#	Line 765 | Line 806 | public class Phaser {
806		LockSupport.unpark(t);
807		}
808		}
809	+	boolean doWait() {
810	+	if (thread != null) {
811	+	try {
812	+	ForkJoinPool.managedBlock(this, false);
813	+	} catch (InterruptedException ie) {
814	+	}
815	+	}
816	+	return wasInterrupted;
817	+	}
818	+
819		}
820
821		/**
822	<	* Removes and signals waiting threads from wait queue
822	>	* Removes and signals waiting threads from wait queue.
823		*/
824		private void releaseWaiters(int phase) {
825		AtomicReference<QNode> head = queueFor(phase);
#	Line 780 | Line 831 | public class Phaser {
831		}
832
833		/**
834	+	* Tries to enqueue given node in the appropriate wait queue.
835	+	*
836	+	* @return true if successful
837	+	*/
838	+	private boolean tryEnqueue(QNode node) {
839	+	AtomicReference<QNode> head = queueFor(node.phase);
840	+	return head.compareAndSet(node.next = head.get(), node);
841	+	}
842	+
843	+	/**
844		* Enqueues node and waits unless aborted or signalled.
845	+	*
846	+	* @return current phase
847		*/
848		private int untimedWait(int phase) {
786	–	int spins = maxUntimedSpins;
849		QNode node = null;
788	–	boolean interrupted = false;
850		boolean queued = false;
851	+	boolean interrupted = false;
852		int p;
853		while ((p = getPhase()) == phase) {
854	<	interrupted = Thread.interrupted();
855	<	if (node != null) {
856	<	if (!queued) {
857	<	AtomicReference<QNode> head = queueFor(phase);
858	<	queued = head.compareAndSet(node.next = head.get(), node);
859	<	}
798	<	else if (node.thread != null)
799	<	LockSupport.park(this);
800	<	}
801	<	else if (spins <= 0)
802	<	node = new QNode();
854	>	if (Thread.interrupted())
855	>	interrupted = true;
856	>	else if (node == null)
857	>	node = new QNode(this, phase, false, false, 0, 0);
858	>	else if (!queued)
859	>	queued = tryEnqueue(node);
860		else
861	<	--spins;
861	>	interrupted = node.doWait();
862		}
863		if (node != null)
864		node.thread = null;
865	+	releaseWaiters(phase);
866		if (interrupted)
867		Thread.currentThread().interrupt();
810	–	releaseWaiters(phase);
868		return p;
869		}
870
871		/**
872	<	* Messier interruptible version
872	>	* Interruptible version
873	>	* @return current phase
874		*/
875		private int interruptibleWait(int phase) throws InterruptedException {
818	–	int spins = maxUntimedSpins;
876		QNode node = null;
877		boolean queued = false;
878		boolean interrupted = false;
879		int p;
880	<	while ((p = getPhase()) == phase) {
881	<	if (interrupted = Thread.interrupted())
882	<	break;
883	<	if (node != null) {
884	<	if (!queued) {
885	<	AtomicReference<QNode> head = queueFor(phase);
886	<	queued = head.compareAndSet(node.next = head.get(), node);
830	<	}
831	<	else if (node.thread != null)
832	<	LockSupport.park(this);
833	<	}
834	<	else if (spins <= 0)
835	<	node = new QNode();
880	>	while ((p = getPhase()) == phase && !interrupted) {
881	>	if (Thread.interrupted())
882	>	interrupted = true;
883	>	else if (node == null)
884	>	node = new QNode(this, phase, true, false, 0, 0);
885	>	else if (!queued)
886	>	queued = tryEnqueue(node);
887		else
888	<	--spins;
888	>	interrupted = node.doWait();
889		}
890		if (node != null)
891		node.thread = null;
892	+	if (p != phase || (p = getPhase()) != phase)
893	+	releaseWaiters(phase);
894		if (interrupted)
895		throw new InterruptedException();
843	–	releaseWaiters(phase);
896		return p;
897		}
898
899		/**
900	<	* Even messier timeout version.
900	>	* Timeout version.
901	>	* @return current phase
902		*/
903		private int timedWait(int phase, long nanos)
904		throws InterruptedException, TimeoutException {
905	+	long startTime = System.nanoTime();
906	+	QNode node = null;
907	+	boolean queued = false;
908	+	boolean interrupted = false;
909		int p;
910	<	if ((p = getPhase()) == phase) {
911	<	long lastTime = System.nanoTime();
912	<	int spins = maxTimedSpins;
913	<	QNode node = null;
914	<	boolean queued = false;
915	<	boolean interrupted = false;
916	<	while ((p = getPhase()) == phase) {
917	<	if (interrupted = Thread.interrupted())
918	<	break;
919	<	long now = System.nanoTime();
920	<	if ((nanos -= now - lastTime) <= 0)
864	<	break;
865	<	lastTime = now;
866	<	if (node != null) {
867	<	if (!queued) {
868	<	AtomicReference<QNode> head = queueFor(phase);
869	<	queued = head.compareAndSet(node.next = head.get(), node);
870	<	}
871	<	else if (node.thread != null &&
872	<	nanos > spinForTimeoutThreshold) {
873	<	LockSupport.parkNanos(this, nanos);
874	<	}
875	<	}
876	<	else if (spins <= 0)
877	<	node = new QNode();
878	<	else
879	<	--spins;
880	<	}
881	<	if (node != null)
882	<	node.thread = null;
883	<	if (interrupted)
884	<	throw new InterruptedException();
885	<	if (p == phase && (p = getPhase()) == phase)
886	<	throw new TimeoutException();
910	>	while ((p = getPhase()) == phase && !interrupted) {
911	>	if (Thread.interrupted())
912	>	interrupted = true;
913	>	else if (nanos - (System.nanoTime() - startTime) <= 0)
914	>	break;
915	>	else if (node == null)
916	>	node = new QNode(this, phase, true, true, startTime, nanos);
917	>	else if (!queued)
918	>	queued = tryEnqueue(node);
919	>	else
920	>	interrupted = node.doWait();
921		}
922	<	releaseWaiters(phase);
922	>	if (node != null)
923	>	node.thread = null;
924	>	if (p != phase || (p = getPhase()) != phase)
925	>	releaseWaiters(phase);
926	>	if (interrupted)
927	>	throw new InterruptedException();
928	>	if (p == phase)
929	>	throw new TimeoutException();
930		return p;
931		}
932
933	<	// Temporary Unsafe mechanics for preliminary release
933	>	// Unsafe mechanics for jsr166y 3rd party package.
934	>	private static sun.misc.Unsafe getUnsafe() {
935	>	try {
936	>	return sun.misc.Unsafe.getUnsafe();
937	>	} catch (SecurityException se) {
938	>	try {
939	>	return java.security.AccessController.doPrivileged
940	>	(new java.security.PrivilegedExceptionAction<sun.misc.Unsafe>() {
941	>	public sun.misc.Unsafe run() throws Exception {
942	>	return getUnsafeByReflection();
943	>	}});
944	>	} catch (java.security.PrivilegedActionException e) {
945	>	throw new RuntimeException("Could not initialize intrinsics",
946	>	e.getCause());
947	>	}
948	>	}
949	>	}
950
951	<	static final Unsafe _unsafe;
952	<	static final long stateOffset;
951	>	private static sun.misc.Unsafe getUnsafeByReflection()
952	>	throws NoSuchFieldException, IllegalAccessException {
953	>	java.lang.reflect.Field f =
954	>	sun.misc.Unsafe.class.getDeclaredField("theUnsafe");
955	>	f.setAccessible(true);
956	>	return (sun.misc.Unsafe) f.get(null);
957	>	}
958
959	<	static {
959	>	private static long fieldOffset(String fieldName, Class<?> klazz) {
960		try {
961	<	if (Phaser.class.getClassLoader() != null) {
962	<	Field f = Unsafe.class.getDeclaredField("theUnsafe");
963	<	f.setAccessible(true);
964	<	_unsafe = (Unsafe)f.get(null);
965	<	}
966	<	else
905	<	_unsafe = Unsafe.getUnsafe();
906	<	stateOffset = _unsafe.objectFieldOffset
907	<	(Phaser.class.getDeclaredField("state"));
908	<	} catch (Exception e) {
909	<	throw new RuntimeException("Could not initialize intrinsics", e);
961	>	return UNSAFE.objectFieldOffset(klazz.getDeclaredField(fieldName));
962	>	} catch (NoSuchFieldException e) {
963	>	// Convert Exception to Error
964	>	NoSuchFieldError error = new NoSuchFieldError(fieldName);
965	>	error.initCause(e);
966	>	throw error;
967		}
968		}
969
970	<	final boolean casState(long cmp, long val) {
971	<	return _unsafe.compareAndSwapLong(this, stateOffset, cmp, val);
970	>	private static final sun.misc.Unsafe UNSAFE = getUnsafe();
971	>	private static final long stateOffset =
972	>	fieldOffset("state", Phaser.class);
973	>
974	>	private final boolean casState(long cmp, long val) {
975	>	return UNSAFE.compareAndSwapLong(this, stateOffset, cmp, val);
976		}
977		}

Diff Legend

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