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 1.1 by dl, Mon Jul 7 16:53:30 2008 UTC vs.
Revision 1.42 by dl, Mon Aug 24 18:37:15 2009 UTC

#	Line 5 | Line 5
5		*/
6
7		package jsr166y;
8	<	import jsr166y.forkjoin.*;
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;
13
14		/**
15	<	* A reusable synchronization barrier, similar in functionality to a
16	<	* {@link java.util.concurrent.CyclicBarrier}, but supporting more
17	<	* flexible usage.
15	>	* A reusable synchronization barrier, similar in functionality to
16	>	* {@link java.util.concurrent.CyclicBarrier CyclicBarrier} and
17	>	* {@link java.util.concurrent.CountDownLatch CountDownLatch}
18	>	* but supporting more flexible usage.
19		*
20	<	* <ul>
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
24	>	* constructors establishing initial numbers of parties), and
25	>	* optionally deregistered upon any arrival (using {@link
26	>	* #arriveAndDeregister}).  As is the case with most basic
27	>	* synchronization constructs, registration and deregistration affect
28	>	* only internal counts; they do not establish any further internal
29	>	* bookkeeping, so tasks cannot query whether they are registered.
30	>	* (However, you can introduce such bookkeeping by subclassing this
31	>	* class.)
32		*
33	<	* <li> The number of parties synchronizing on the barrier may vary
34	<	* over time.  A task may register to be a party in a barrier at any
35	<	* time, and may deregister upon arriving at the barrier.  As is the
36	<	* case with most basic synchronization constructs, registration
37	<	* and deregistration affect only internal counts; they do not
38	<	* establish any further internal bookkeeping, so tasks cannot query
39	<	* whether they are registered.
40	<	*
41	<	* <li> Each generation has an associated phase value, starting at
42	<	* zero, and advancing when all parties reach the barrier (wrapping
43	<	* around to zero after reaching <tt>Integer.MAX_VALUE</tt>).
31	<	*
32	<	* <li> Like a CyclicBarrier, a Phaser may be repeatedly awaited.
33	<	* Method <tt>arriveAndAwaitAdvance</tt> has effect analogous to
34	<	* <tt>CyclicBarrier.await</tt>.  However, Phasers separate two
35	<	* aspects of coordination, that may be invoked independently:
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
37	>	* generation of a {@code Phaser} has an associated phase number. The
38	>	* phase number starts at zero, and advances when all parties arrive
39	>	* at the barrier, wrapping around to zero after reaching {@code
40	>	* Integer.MAX_VALUE}. The use of phase numbers enables independent
41	>	* control of actions upon arrival at a barrier and upon awaiting
42	>	* others, via two kinds of methods that may be invoked by any
43	>	* registered party:
44		*
45		* <ul>
46		*
47	<	*   <li> Arriving at a barrier. Methods <tt>arrive</tt> and
48	<	*       <tt>arriveAndDeregister</tt> do not block, but return
49	<	*       the phase value on entry to the method.
50	<	*
51	<	*   <li> Awaiting others. Method <tt>awaitAdvance</tt> requires an
52	<	*       argument indicating the entry phase, and returns when the
53	<	*       barrier advances to a new phase.
47	>	*   <li> <b>Arrival.</b> Methods {@link #arrive} and
48	>	*       {@link #arriveAndDeregister} record arrival at a
49	>	*       barrier. These methods do not block, but return an associated
50	>	*       <em>arrival phase number</em>; that is, the phase number of
51	>	*       the barrier to which the arrival applied. When the final
52	>	*       party for a given phase arrives, an optional barrier action
53	>	*       is performed and the phase advances.  Barrier actions,
54	>	*       performed by the party triggering a phase advance, are
55	>	*       arranged by overriding method {@link #onAdvance(int, int)},
56	>	*       which also controls termination. Overriding this method is
57	>	*       similar to, but more flexible than, providing a barrier
58	>	*       action to a {@code CyclicBarrier}.
59	>	*
60	>	*   <li> <b>Waiting.</b> Method {@link #awaitAdvance} requires an
61	>	*       argument indicating an arrival phase number, and returns when
62	>	*       the barrier advances to (or is already at) a different phase.
63	>	*       Unlike similar constructions using {@code CyclicBarrier},
64	>	*       method {@code awaitAdvance} continues to wait even if the
65	>	*       waiting thread is interrupted. Interruptible and timeout
66	>	*       versions are also available, but exceptions encountered while
67	>	*       tasks wait interruptibly or with timeout do not change the
68	>	*       state of the barrier. If necessary, you can perform any
69	>	*       associated recovery within handlers of those exceptions,
70	>	*       often after invoking {@code forceTermination}.  Phasers may
71	>	*       also be used by tasks executing in a {@link ForkJoinPool},
72	>	*       which will ensure sufficient parallelism to execute tasks
73	>	*       when others are blocked waiting for a phase to advance.
74	>	*
75		* </ul>
76		*
77	+	* <p> <b>Termination.</b> A {@code Phaser} may enter a
78	+	* <em>termination</em> state in which all synchronization methods
79	+	* immediately return without updating phaser state or waiting for
80	+	* advance, and indicating (via a negative phase value) that execution
81	+	* is complete.  Termination is triggered when an invocation of {@code
82	+	* onAdvance} returns {@code true}.  As illustrated below, when
83	+	* phasers control actions with a fixed number of iterations, it is
84	+	* often convenient to override this method to cause termination when
85	+	* the current phase number reaches a threshold. Method {@link
86	+	* #forceTermination} is also available to abruptly release waiting
87	+	* threads and allow them to terminate.
88		*
89	<	* <li> Barrier actions, performed by the task triggering a phase
90	<	* advance while others may be waiting, are arranged by overriding
91	<	* method <tt>onAdvance</tt>, that also controls termination.
92	<	*
93	<	* <li> Phasers may enter a <em>termination</em> state in which all
94	<	* await actions immediately return, indicating (via a negative phase
95	<	* value) that execution is complete.  Termination is triggered by
56	<	* executing the overridable <tt>onAdvance</tt> method that is invoked
57	<	* each time the barrier is tripped. When a Phaser is controlling an
58	<	* action with a fixed number of iterations, it is often convenient to
59	<	* override this method to cause termination when the current phase
60	<	* number reaches a threshold.  Method <tt>forceTermination</tt> is
61	<	* also available to assist recovery actions upon failure.
62	<	*
63	<	* <li> Unlike most synchronizers, a Phaser may also be used with
64	<	* ForkJoinTasks (as well as plain threads).
65	<	*
66	<	* <li> By default, <tt>awaitAdvance</tt> continues to wait even if
67	<	* the current thread is interrupted. And unlike the case in
68	<	* CyclicBarriers, exceptions encountered while tasks wait
69	<	* interruptibly or with timeout do not change the state of the
70	<	* barrier. If necessary, you can perform any associated recovery
71	<	* within handlers of those exceptions.
89	>	* <p> <b>Tiering.</b> Phasers may be <em>tiered</em> (i.e., arranged
90	>	* in tree structures) to reduce contention. Phasers with large
91	>	* numbers of parties that would otherwise experience heavy
92	>	* synchronization contention costs may instead be set up so that
93	>	* groups of sub-phasers share a common parent.  This may greatly
94	>	* increase throughput even though it incurs greater per-operation
95	>	* overhead.
96		*
97	<	* </ul>
97	>	* <p><b>Monitoring.</b> While synchronization methods may be invoked
98	>	* only by registered parties, the current state of a phaser may be
99	>	* monitored by any caller.  At any given moment there are {@link
100	>	* #getRegisteredParties} parties in total, of which {@link
101	>	* #getArrivedParties} have arrived at the current phase ({@link
102	>	* #getPhase}).  When the remaining ({@link #getUnarrivedParties})
103	>	* parties arrive, the phase advances.  The values returned by these
104	>	* methods may reflect transient states and so are not in general
105	>	* useful for synchronization control.  Method {@link #toString}
106	>	* returns snapshots of these state queries in a form convenient for
107	>	* informal monitoring.
108	>	*
109	>	* <p><b>Sample usages:</b>
110	>	*
111	>	* <p>A {@code Phaser} may be used instead of a {@code CountDownLatch}
112	>	* to control a one-shot action serving a variable number of
113	>	* parties. The typical idiom is for the method setting this up to
114	>	* first register, then start the actions, then deregister, as in:
115	>	*
116	>	*  <pre> {@code
117	>	* void runTasks(List<Runnable> tasks) {
118	>	*   final Phaser phaser = new Phaser(1); // "1" to register self
119	>	*   // create and start threads
120	>	*   for (Runnable task : tasks) {
121	>	*     phaser.register();
122	>	*     new Thread() {
123	>	*       public void run() {
124	>	*         phaser.arriveAndAwaitAdvance(); // await all creation
125	>	*         task.run();
126	>	*       }
127	>	*     }.start();
128	>	*   }
129	>	*
130	>	*   // allow threads to start and deregister self
131	>	*   phaser.arriveAndDeregister();
132	>	* }}</pre>
133	>	*
134	>	* <p>One way to cause a set of threads to repeatedly perform actions
135	>	* for a given number of iterations is to override {@code onAdvance}:
136	>	*
137	>	*  <pre> {@code
138	>	* void startTasks(List<Runnable> tasks, final int iterations) {
139	>	*   final Phaser phaser = new Phaser() {
140	>	*     protected boolean onAdvance(int phase, int registeredParties) {
141	>	*       return phase >= iterations || registeredParties == 0;
142	>	*     }
143	>	*   };
144	>	*   phaser.register();
145	>	*   for (Runnable task : tasks) {
146	>	*     phaser.register();
147	>	*     new Thread() {
148	>	*       public void run() {
149	>	*         do {
150	>	*           task.run();
151	>	*           phaser.arriveAndAwaitAdvance();
152	>	*         } while(!phaser.isTerminated();
153	>	*       }
154	>	*     }.start();
155	>	*   }
156	>	*   phaser.arriveAndDeregister(); // deregister self, don't wait
157	>	* }}</pre>
158		*
159	<	* <p><b>Sample usage:</b>
159	>	* If the main task must later await termination, it
160	>	* may re-register and then execute a similar loop:
161	>	* <pre> {@code
162	>	*   // ...
163	>	*   phaser.register();
164	>	*   while (!phaser.isTerminated())
165	>	*     phaser.arriveAndAwaitAdvance();
166	>	* }</pre>
167		*
168	<	* <p>[todo: non-FJ example]
168	>	* Related constructions may be used to await particular phase numbers
169	>	* in contexts where you are sure that the phase will never wrap around
170	>	* {@code Integer.MAX_VALUE}. For example:
171		*
172	<	* <p> A Phaser may be used to support a style of programming in
173	<	* which a task waits for others to complete, without otherwise
174	<	* needing to keep track of which tasks it is waiting for. This is
175	<	* similar to the "sync" construct in Cilk and "clocks" in X10.
176	<	* Special constructions based on such barriers are available using
177	<	* the <tt>LinkedAsyncAction</tt> and <tt>CyclicAction</tt> classes,
178	<	* but they can be useful in other contexts as well.  For a simple
179	<	* (but not very useful) example, here is a variant of Fibonacci:
180	<	*
181	<	* <pre>
89	<	* class BarrierFibonacci extends RecursiveAction {
90	<	*   int argument, result;
91	<	*   final Phaser parentBarrier;
92	<	*   BarrierFibonacci(int n, Phaser parentBarrier) {
93	<	*     this.argument = n;
94	<	*     this.parentBarrier = parentBarrier;
95	<	*     parentBarrier.register();
172	>	* <pre> {@code
173	>	*   void awaitPhase(Phaser phaser, int phase) {
174	>	*     int p = phaser.register(); // assumes caller not already registered
175	>	*     while (p < phase) {
176	>	*       if (phaser.isTerminated())
177	>	*         // ... deal with unexpected termination
178	>	*       else
179	>	*         p = phaser.arriveAndAwaitAdvance();
180	>	*     }
181	>	*     phaser.arriveAndDeregister();
182		*   }
183	<	*   protected void compute() {
184	<	*     int n = argument;
185	<	*     if (n &lt;= 1)
186	<	*        result = n;
187	<	*     else {
188	<	*        Phaser childBarrier = new Phaser(1);
189	<	*        BarrierFibonacci f1 = new BarrierFibonacci(n - 1, childBarrier);
190	<	*        BarrierFibonacci f2 = new BarrierFibonacci(n - 2, childBarrier);
191	<	*        f1.fork();
192	<	*        f2.fork();
193	<	*        childBarrier.arriveAndAwait();
194	<	*        result = f1.result + f2.result;
183	>	* }</pre>
184	>	*
185	>	*
186	>	* <p>To create a set of tasks using a tree of phasers,
187	>	* you could use code of the following form, assuming a
188	>	* Task class with a constructor accepting a phaser that
189	>	* it registers for upon construction:
190	>	*  <pre> {@code
191	>	* void build(Task[] actions, int lo, int hi, Phaser b) {
192	>	*   int step = (hi - lo) / TASKS_PER_PHASER;
193	>	*   if (step > 1) {
194	>	*     int i = lo;
195	>	*     while (i < hi) {
196	>	*       int r = Math.min(i + step, hi);
197	>	*       build(actions, i, r, new Phaser(b));
198	>	*       i = r;
199		*     }
200	<	*     parentBarrier.arriveAndDeregister();
200	>	*   } else {
201	>	*     for (int i = lo; i < hi; ++i)
202	>	*       actions[i] = new Task(b);
203	>	*       // assumes new Task(b) performs b.register()
204		*   }
205		* }
206	+	* // .. initially called, for n tasks via
207	+	* build(new Task[n], 0, n, new Phaser());}</pre>
208	+	*
209	+	* The best value of {@code TASKS_PER_PHASER} depends mainly on
210	+	* expected barrier synchronization rates. A value as low as four may
211	+	* be appropriate for extremely small per-barrier task bodies (thus
212	+	* high rates), or up to hundreds for extremely large ones.
213	+	*
214		* </pre>
215		*
216		* <p><b>Implementation notes</b>: This implementation restricts the
217	<	* maximum number of parties to 65535. Attempts to register
218	<	* additional parties result in IllegalStateExceptions.
217	>	* maximum number of parties to 65535. Attempts to register additional
218	>	* parties result in {@code IllegalStateException}. However, you can and
219	>	* should create tiered phasers to accommodate arbitrarily large sets
220	>	* of participants.
221	>	*
222	>	* @since 1.7
223	>	* @author Doug Lea
224		*/
225		public class Phaser {
226		/*
227		* This class implements an extension of X10 "clocks".  Thanks to
228	<	* Vijay Saraswat for the idea of applying it to ForkJoinTasks,
229	<	* and to Vivek Sarkar for enhancements to extend functionality.
228	>	* Vijay Saraswat for the idea, and to Vivek Sarkar for
229	>	* enhancements to extend functionality.
230		*/
231
232		/**
233		* Barrier state representation. Conceptually, a barrier contains
234		* four values:
235	<	*
235	>	*
236		* * parties -- the number of parties to wait (16 bits)
237		* * unarrived -- the number of parties yet to hit barrier (16 bits)
238		* * phase -- the generation of the barrier (31 bits)
239		* * terminated -- set if barrier is terminated (1 bit)
240		*
241		* However, to efficiently maintain atomicity, these values are
242	<	* packed into a single AtomicLong. Termination uses the sign bit
243	<	* of 32 bit representation of phase, so phase is set to -1 on
244	<	* termination.
245	<	*/
246	<	private final AtomicLong state;
247	<
248	<	/**
143	<	* Head of Treiber stack for waiting nonFJ threads.
242	>	* packed into a single (atomic) long. Termination uses the sign
243	>	* bit of 32 bit representation of phase, so phase is set to -1 on
244	>	* termination. Good performance relies on keeping state decoding
245	>	* and encoding simple, and keeping race windows short.
246	>	*
247	>	* Note: there are some cheats in arrive() that rely on unarrived
248	>	* count being lowest 16 bits.
249		*/
250	<	private final AtomicReference<QNode> head = new AtomicReference<QNode>();
250	>	private volatile long state;
251
252	<	private static final int ushortBits = 16;
253	<	private static final int ushortMask =  (1 << ushortBits) - 1;
149	<	private static final int phaseMask = 0x7fffffff;
252	>	private static final int ushortMask = 0xffff;
253	>	private static final int phaseMask  = 0x7fffffff;
254
255		private static int unarrivedOf(long s) {
256	<	return (int)(s & ushortMask);
256	>	return (int) (s & ushortMask);
257		}
258
259		private static int partiesOf(long s) {
260	<	return (int)(s & (ushortMask << 16)) >>> 16;
260	>	return ((int) s) >>> 16;
261		}
262
263		private static int phaseOf(long s) {
264	<	return (int)(s >>> 32);
264	>	return (int) (s >>> 32);
265		}
266
267		private static int arrivedOf(long s) {
#	Line 165 | Line 269 | public class Phaser {
269		}
270
271		private static long stateFor(int phase, int parties, int unarrived) {
272	<	return (((long)phase) << 32) | ((parties << 16) | unarrived);
272	>	return ((((long) phase) << 32) | (((long) parties) << 16) |
273	>	(long) unarrived);
274		}
275
276	<	private static IllegalStateException badBounds(int parties, int unarrived) {
277	<	return new IllegalStateException("Attempt to set " + unarrived +
278	<	" unarrived of " + parties + " parties");
276	>	private static long trippedStateFor(int phase, int parties) {
277	>	long lp = (long) parties;
278	>	return (((long) phase) << 32) | (lp << 16) | lp;
279		}
280
281		/**
282	<	* Creates a new Phaser without any initially registered parties,
283	<	* and initial phase number 0.
282	>	* Returns message string for bad bounds exceptions.
283	>	*/
284	>	private static String badBounds(int parties, int unarrived) {
285	>	return ("Attempt to set " + unarrived +
286	>	" unarrived of " + parties + " parties");
287	>	}
288	>
289	>	/**
290	>	* The parent of this phaser, or null if none
291	>	*/
292	>	private final Phaser parent;
293	>
294	>	/**
295	>	* The root of phaser tree. Equals this if not in a tree.  Used to
296	>	* support faster state push-down.
297	>	*/
298	>	private final Phaser root;
299	>
300	>	// Wait queues
301	>
302	>	/**
303	>	* Heads of Treiber stacks for waiting threads. To eliminate
304	>	* contention while releasing some threads while adding others, we
305	>	* use two of them, alternating across even and odd phases.
306	>	*/
307	>	private final AtomicReference<QNode> evenQ = new AtomicReference<QNode>();
308	>	private final AtomicReference<QNode> oddQ  = new AtomicReference<QNode>();
309	>
310	>	private AtomicReference<QNode> queueFor(int phase) {
311	>	return ((phase & 1) == 0) ? evenQ : oddQ;
312	>	}
313	>
314	>	/**
315	>	* Returns current state, first resolving lagged propagation from
316	>	* root if necessary.
317	>	*/
318	>	private long getReconciledState() {
319	>	return (parent == null) ? state : reconcileState();
320	>	}
321	>
322	>	/**
323	>	* Recursively resolves state.
324	>	*/
325	>	private long reconcileState() {
326	>	Phaser p = parent;
327	>	long s = state;
328	>	if (p != null) {
329	>	while (unarrivedOf(s) == 0 && phaseOf(s) != phaseOf(root.state)) {
330	>	long parentState = p.getReconciledState();
331	>	int parentPhase = phaseOf(parentState);
332	>	int phase = phaseOf(s = state);
333	>	if (phase != parentPhase) {
334	>	long next = trippedStateFor(parentPhase, partiesOf(s));
335	>	if (casState(s, next)) {
336	>	releaseWaiters(phase);
337	>	s = next;
338	>	}
339	>	}
340	>	}
341	>	}
342	>	return s;
343	>	}
344	>
345	>	/**
346	>	* Creates a new phaser without any initially registered parties,
347	>	* initial phase number 0, and no parent. Any thread using this
348	>	* phaser will need to first register for it.
349		*/
350		public Phaser() {
351	<	state = new AtomicLong(stateFor(0, 0, 0));
351	>	this(null);
352		}
353
354		/**
355	<	* Creates a new Phaser with the given numbers of registered
356	<	* unarrived parties and initial phase number 0.
357	<	* @param parties the number of parties required to trip barrier.
355	>	* Creates a new phaser with the given numbers of registered
356	>	* unarrived parties, initial phase number 0, and no parent.
357	>	*
358	>	* @param parties the number of parties required to trip barrier
359		* @throws IllegalArgumentException if parties less than zero
360	<	* or greater than the maximum number of parties supported.
360	>	* or greater than the maximum number of parties supported
361		*/
362		public Phaser(int parties) {
363	+	this(null, parties);
364	+	}
365	+
366	+	/**
367	+	* Creates a new phaser with the given parent, without any
368	+	* initially registered parties. If parent is non-null this phaser
369	+	* is registered with the parent and its initial phase number is
370	+	* the same as that of parent phaser.
371	+	*
372	+	* @param parent the parent phaser
373	+	*/
374	+	public Phaser(Phaser parent) {
375	+	int phase = 0;
376	+	this.parent = parent;
377	+	if (parent != null) {
378	+	this.root = parent.root;
379	+	phase = parent.register();
380	+	}
381	+	else
382	+	this.root = this;
383	+	this.state = trippedStateFor(phase, 0);
384	+	}
385	+
386	+	/**
387	+	* Creates a new phaser with the given parent and numbers of
388	+	* registered unarrived parties. If parent is non-null, this phaser
389	+	* is registered with the parent and its initial phase number is
390	+	* the same as that of parent phaser.
391	+	*
392	+	* @param parent the parent phaser
393	+	* @param parties the number of parties required to trip barrier
394	+	* @throws IllegalArgumentException if parties less than zero
395	+	* or greater than the maximum number of parties supported
396	+	*/
397	+	public Phaser(Phaser parent, int parties) {
398		if (parties < 0 || parties > ushortMask)
399		throw new IllegalArgumentException("Illegal number of parties");
400	<	state = new AtomicLong(stateFor(0, parties, parties));
400	>	int phase = 0;
401	>	this.parent = parent;
402	>	if (parent != null) {
403	>	this.root = parent.root;
404	>	phase = parent.register();
405	>	}
406	>	else
407	>	this.root = this;
408	>	this.state = trippedStateFor(phase, parties);
409		}
410
411		/**
412		* Adds a new unarrived party to this phaser.
413	<	* @return the current barrier phase number upon registration
413	>	*
414	>	* @return the arrival phase number to which this registration applied
415	>	* @throws IllegalStateException if attempting to register more
416	>	* than the maximum supported number of parties
417	>	*/
418	>	public int register() {
419	>	return doRegister(1);
420	>	}
421	>
422	>	/**
423	>	* Adds the given number of new unarrived parties to this phaser.
424	>	*
425	>	* @param parties the number of parties required to trip barrier
426	>	* @return the arrival phase number to which this registration applied
427		* @throws IllegalStateException if attempting to register more
428	<	* than the maximum supported number of parties.
428	>	* than the maximum supported number of parties
429		*/
430	<	public int register() { // increment both parties and unarrived
431	<	final AtomicLong state = this.state;
430	>	public int bulkRegister(int parties) {
431	>	if (parties < 0)
432	>	throw new IllegalArgumentException();
433	>	if (parties == 0)
434	>	return getPhase();
435	>	return doRegister(parties);
436	>	}
437	>
438	>	/**
439	>	* Shared code for register, bulkRegister
440	>	*/
441	>	private int doRegister(int registrations) {
442	>	int phase;
443		for (;;) {
444	<	long s = state.get();
445	<	int phase = phaseOf(s);
446	<	int parties = partiesOf(s) + 1;
447	<	int unarrived = unarrivedOf(s) + 1;
444	>	long s = getReconciledState();
445	>	phase = phaseOf(s);
446	>	int unarrived = unarrivedOf(s) + registrations;
447	>	int parties = partiesOf(s) + registrations;
448	>	if (phase < 0)
449	>	break;
450		if (parties > ushortMask || unarrived > ushortMask)
451	<	throw badBounds(parties, unarrived);
452	<	if (state.compareAndSet(s, stateFor(phase, parties, unarrived)))
453	<	return phase;
451	>	throw new IllegalStateException(badBounds(parties, unarrived));
452	>	if (phase == phaseOf(root.state) &&
453	>	casState(s, stateFor(phase, parties, unarrived)))
454	>	break;
455		}
456	+	return phase;
457		}
458
459		/**
460		* Arrives at the barrier, but does not wait for others.  (You can
461	<	* in turn wait for others via {@link #awaitAdvance}).
461	>	* in turn wait for others via {@link #awaitAdvance}).  It is an
462	>	* unenforced usage error for an unregistered party to invoke this
463	>	* method.
464		*
465	<	* @return the current barrier phase number upon entry to
466	<	* this method, or a negative value if terminated;
467	<	* @throws IllegalStateException if the number of unarrived
224	<	* parties would become negative.
465	>	* @return the arrival phase number, or a negative value if terminated
466	>	* @throws IllegalStateException if not terminated and the number
467	>	* of unarrived parties would become negative
468		*/
469	<	public int arrive() { // decrement unarrived. If zero, trip
470	<	final AtomicLong state = this.state;
469	>	public int arrive() {
470	>	int phase;
471		for (;;) {
472	<	long s = state.get();
473	<	int phase = phaseOf(s);
472	>	long s = state;
473	>	phase = phaseOf(s);
474	>	if (phase < 0)
475	>	break;
476		int parties = partiesOf(s);
477		int unarrived = unarrivedOf(s) - 1;
478	<	if (unarrived < 0)
479	<	throw badBounds(parties, unarrived);
480	<	if (unarrived == 0 && phase >= 0) {
236	<	trip(phase, parties);
237	<	return phase;
478	>	if (unarrived > 0) {        // Not the last arrival
479	>	if (casState(s, s - 1)) // s-1 adds one arrival
480	>	break;
481		}
482	<	if (state.compareAndSet(s, stateFor(phase, parties, unarrived)))
483	<	return phase;
482	>	else if (unarrived == 0) {  // the last arrival
483	>	Phaser par = parent;
484	>	if (par == null) {      // directly trip
485	>	if (casState
486	>	(s,
487	>	trippedStateFor(onAdvance(phase, parties) ? -1 :
488	>	((phase + 1) & phaseMask), parties))) {
489	>	releaseWaiters(phase);
490	>	break;
491	>	}
492	>	}
493	>	else {                  // cascade to parent
494	>	if (casState(s, s - 1)) { // zeroes unarrived
495	>	par.arrive();
496	>	reconcileState();
497	>	break;
498	>	}
499	>	}
500	>	}
501	>	else if (phase != phaseOf(root.state)) // or if unreconciled
502	>	reconcileState();
503	>	else
504	>	throw new IllegalStateException(badBounds(parties, unarrived));
505		}
506	+	return phase;
507		}
508
509		/**
510	<	* Arrives at the barrier, and deregisters from it, without
511	<	* waiting for others.
510	>	* Arrives at the barrier and deregisters from it without waiting
511	>	* for others. Deregistration reduces the number of parties
512	>	* required to trip the barrier in future phases.  If this phaser
513	>	* has a parent, and deregistration causes this phaser to have
514	>	* zero parties, this phaser also arrives at and is deregistered
515	>	* from its parent.  It is an unenforced usage error for an
516	>	* unregistered party to invoke this method.
517		*
518	<	* @return the current barrier phase number upon entry to
519	<	* this method, or a negative value if terminated;
520	<	* @throws IllegalStateException if the number of registered or
521	<	* unarrived parties would become negative.
522	<	*/
523	<	public int arriveAndDeregister() { // Same as arrive, plus decrement parties
524	<	final AtomicLong state = this.state;
518	>	* @return the arrival phase number, or a negative value if terminated
519	>	* @throws IllegalStateException if not terminated and the number
520	>	* of registered or unarrived parties would become negative
521	>	*/
522	>	public int arriveAndDeregister() {
523	>	// similar code to arrive, but too different to merge
524	>	Phaser par = parent;
525	>	int phase;
526		for (;;) {
527	<	long s = state.get();
528	<	int phase = phaseOf(s);
527	>	long s = state;
528	>	phase = phaseOf(s);
529	>	if (phase < 0)
530	>	break;
531		int parties = partiesOf(s) - 1;
532		int unarrived = unarrivedOf(s) - 1;
533	<	if (parties < 0 || unarrived < 0)
534	<	throw badBounds(parties, unarrived);
535	<	if (unarrived == 0 && phase >= 0) {
536	<	trip(phase, parties);
537	<	return phase;
533	>	if (parties >= 0) {
534	>	if (unarrived > 0 || (unarrived == 0 && par != null)) {
535	>	if (casState
536	>	(s,
537	>	stateFor(phase, parties, unarrived))) {
538	>	if (unarrived == 0) {
539	>	par.arriveAndDeregister();
540	>	reconcileState();
541	>	}
542	>	break;
543	>	}
544	>	continue;
545	>	}
546	>	if (unarrived == 0) {
547	>	if (casState
548	>	(s,
549	>	trippedStateFor(onAdvance(phase, parties) ? -1 :
550	>	((phase + 1) & phaseMask), parties))) {
551	>	releaseWaiters(phase);
552	>	break;
553	>	}
554	>	continue;
555	>	}
556	>	if (par != null && phase != phaseOf(root.state)) {
557	>	reconcileState();
558	>	continue;
559	>	}
560		}
561	<	if (state.compareAndSet(s, stateFor(phase, parties, unarrived)))
267	<	return phase;
561	>	throw new IllegalStateException(badBounds(parties, unarrived));
562		}
563	+	return phase;
564		}
565
566		/**
567	<	* Arrives at the barrier and awaits others. Unlike other arrival
568	<	* methods, this method returns the arrival index of the
569	<	* caller. The caller tripping the barrier returns zero, the
570	<	* previous caller 1, and so on.
571	<	* @return the arrival index
572	<	* @throws IllegalStateException if the number of unarrived
573	<	* parties would become negative.
567	>	* Arrives at the barrier and awaits others. Equivalent in effect
568	>	* to {@code awaitAdvance(arrive())}.  If you need to await with
569	>	* interruption or timeout, you can arrange this with an analogous
570	>	* construction using one of the other forms of the awaitAdvance
571	>	* method.  If instead you need to deregister upon arrival use
572	>	* {@code arriveAndDeregister}. It is an unenforced usage error
573	>	* for an unregistered party to invoke this method.
574	>	*
575	>	* @return the arrival phase number, or a negative number if terminated
576	>	* @throws IllegalStateException if not terminated and the number
577	>	* of unarrived parties would become negative
578		*/
579		public int arriveAndAwaitAdvance() {
580	<	final AtomicLong state = this.state;
282	<	for (;;) {
283	<	long s = state.get();
284	<	int phase = phaseOf(s);
285	<	int parties = partiesOf(s);
286	<	int unarrived = unarrivedOf(s) - 1;
287	<	if (unarrived < 0)
288	<	throw badBounds(parties, unarrived);
289	<	if (unarrived == 0 && phase >= 0) {
290	<	trip(phase, parties);
291	<	return 0;
292	<	}
293	<	if (state.compareAndSet(s, stateFor(phase, parties, unarrived))) {
294	<	awaitAdvance(phase);
295	<	return unarrived;
296	<	}
297	<	}
580	>	return awaitAdvance(arrive());
581		}
582
583		/**
584	<	* Awaits the phase of the barrier to advance from the given
585	<	* value, or returns immediately if this barrier is terminated
586	<	* @param phase the phase on entry to this method
587	<	* @return the phase on exit from this method
584	>	* Awaits the phase of the barrier to advance from the given phase
585	>	* value, returning immediately if the current phase of the
586	>	* barrier is not equal to the given phase value or this barrier
587	>	* is terminated.  It is an unenforced usage error for an
588	>	* unregistered party to invoke this method.
589	>	*
590	>	* @param phase an arrival phase number, or negative value if
591	>	* terminated; this argument is normally the value returned by a
592	>	* previous call to {@code arrive} or its variants
593	>	* @return the next arrival phase number, or a negative value
594	>	* if terminated or argument is negative
595		*/
596		public int awaitAdvance(int phase) {
597		if (phase < 0)
598		return phase;
599	<	Thread current = Thread.currentThread();
600	<	if (current instanceof ForkJoinWorkerThread)
601	<	return helpingWait(phase);
602	<	if (untimedWait(current, phase, false))
603	<	current.interrupt();
604	<	return phaseOf(state.get());
599	>	long s = getReconciledState();
600	>	int p = phaseOf(s);
601	>	if (p != phase)
602	>	return p;
603	>	if (unarrivedOf(s) == 0 && parent != null)
604	>	parent.awaitAdvance(phase);
605	>	// Fall here even if parent waited, to reconcile and help release
606	>	return untimedWait(phase);
607		}
608
609		/**
610	<	* Awaits the phase of the barrier to advance from the given
611	<	* value, or returns immediately if this barrier is terminated, or
612	<	* throws InterruptedException if interrupted while waiting.
613	<	* @param phase the phase on entry to this method
614	<	* @return the phase on exit from this method
610	>	* Awaits the phase of the barrier to advance from the given phase
611	>	* value, throwing {@code InterruptedException} if interrupted
612	>	* while waiting, or returning immediately if the current phase of
613	>	* the barrier is not equal to the given phase value or this
614	>	* barrier is terminated. It is an unenforced usage error for an
615	>	* unregistered party to invoke this method.
616	>	*
617	>	* @param phase an arrival phase number, or negative value if
618	>	* terminated; this argument is normally the value returned by a
619	>	* previous call to {@code arrive} or its variants
620	>	* @return the next arrival phase number, or a negative value
621	>	* if terminated or argument is negative
622		* @throws InterruptedException if thread interrupted while waiting
623		*/
624	<	public int awaitAdvanceInterruptibly(int phase) throws InterruptedException {
624	>	public int awaitAdvanceInterruptibly(int phase)
625	>	throws InterruptedException {
626		if (phase < 0)
627		return phase;
628	<	Thread current = Thread.currentThread();
629	<	if (current instanceof ForkJoinWorkerThread)
630	<	return helpingWait(phase);
631	<	else if (Thread.interrupted() || untimedWait(current, phase, true))
632	<	throw new InterruptedException();
633	<	else
634	<	return phaseOf(state.get());
628	>	long s = getReconciledState();
629	>	int p = phaseOf(s);
630	>	if (p != phase)
631	>	return p;
632	>	if (unarrivedOf(s) == 0 && parent != null)
633	>	parent.awaitAdvanceInterruptibly(phase);
634	>	return interruptibleWait(phase);
635		}
636
637		/**
638	<	* Awaits the phase of the barrier to advance from the given value
639	<	* or the given timeout elapses, or returns immediately if this
640	<	* barrier is terminated
641	<	* @param phase the phase on entry to this method
642	<	* @return the phase on exit from this method
638	>	* Awaits the phase of the barrier to advance from the given phase
639	>	* value or the given timeout to elapse, throwing {@code
640	>	* InterruptedException} if interrupted while waiting, or
641	>	* returning immediately if the current phase of the barrier is
642	>	* not equal to the given phase value or this barrier is
643	>	* terminated.  It is an unenforced usage error for an
644	>	* unregistered party to invoke this method.
645	>	*
646	>	* @param phase an arrival phase number, or negative value if
647	>	* terminated; this argument is normally the value returned by a
648	>	* previous call to {@code arrive} or its variants
649	>	* @param timeout how long to wait before giving up, in units of
650	>	*        {@code unit}
651	>	* @param unit a {@code TimeUnit} determining how to interpret the
652	>	*        {@code timeout} parameter
653	>	* @return the next arrival phase number, or a negative value
654	>	* if terminated or argument is negative
655		* @throws InterruptedException if thread interrupted while waiting
656		* @throws TimeoutException if timed out while waiting
657		*/
658	<	public int awaitAdvanceInterruptibly(int phase, long timeout, TimeUnit unit)
658	>	public int awaitAdvanceInterruptibly(int phase,
659	>	long timeout, TimeUnit unit)
660		throws InterruptedException, TimeoutException {
661		if (phase < 0)
662		return phase;
663	<	long nanos = unit.toNanos(timeout);
664	<	Thread current = Thread.currentThread();
665	<	if (current instanceof ForkJoinWorkerThread)
666	<	return timedHelpingWait(phase, nanos);
667	<	timedWait(current, phase, nanos);
668	<	return phaseOf(state.get());
663	>	long s = getReconciledState();
664	>	int p = phaseOf(s);
665	>	if (p != phase)
666	>	return p;
667	>	if (unarrivedOf(s) == 0 && parent != null)
668	>	parent.awaitAdvanceInterruptibly(phase, timeout, unit);
669	>	return timedWait(phase, unit.toNanos(timeout));
670		}
671
672		/**
673		* Forces this barrier to enter termination state. Counts of
674	<	* arrived and registered parties are unaffected. This method may
675	<	* be useful for coordinating recovery after one or more tasks
676	<	* encounter unexpected exceptions.
674	>	* arrived and registered parties are unaffected. If this phaser
675	>	* has a parent, it too is terminated. This method may be useful
676	>	* for coordinating recovery after one or more tasks encounter
677	>	* unexpected exceptions.
678		*/
679		public void forceTermination() {
365	–	final AtomicLong state = this.state;
680		for (;;) {
681	<	long s = state.get();
681	>	long s = getReconciledState();
682		int phase = phaseOf(s);
683		int parties = partiesOf(s);
684		int unarrived = unarrivedOf(s);
685		if (phase < 0 ||
686	<	state.compareAndSet(s, stateFor(-1, parties, unarrived))) {
687	<	if (head.get() != null)
688	<	releaseWaiters(-1);
686	>	casState(s, stateFor(-1, parties, unarrived))) {
687	>	releaseWaiters(0);
688	>	releaseWaiters(1);
689	>	if (parent != null)
690	>	parent.forceTermination();
691		return;
692		}
693		}
694		}
695
696		/**
381	–	* Resets the barrier with the given numbers of registered unarrived
382	–	* parties and phase number 0. This method allows repeated reuse
383	–	* of this barrier, but only if it is somehow known not to be in
384	–	* use for other purposes.
385	–	* @param parties the number of parties required to trip barrier.
386	–	* @throws IllegalArgumentException if parties less than zero
387	–	* or greater than the maximum number of parties supported.
388	–	*/
389	–	public void reset(int parties) {
390	–	if (parties < 0 || parties > ushortMask)
391	–	throw new IllegalArgumentException("Illegal number of parties");
392	–	state.set(stateFor(0, parties, parties));
393	–	if (head.get() != null)
394	–	releaseWaiters(0);
395	–	}
396	–
397	–	/**
697		* Returns the current phase number. The maximum phase number is
698	<	* <tt>Integer.MAX_VALUE</tt>, after which it restarts at
698	>	* {@code Integer.MAX_VALUE}, after which it restarts at
699		* zero. Upon termination, the phase number is negative.
700	+	*
701		* @return the phase number, or a negative value if terminated
702		*/
703	<	public int getPhase() {
704	<	return phaseOf(state.get());
703	>	public final int getPhase() {
704	>	return phaseOf(getReconciledState());
705		}
706
707		/**
708		* Returns the number of parties registered at this barrier.
709	+	*
710		* @return the number of parties
711		*/
712		public int getRegisteredParties() {
713	<	return partiesOf(state.get());
713	>	return partiesOf(state);
714		}
715
716		/**
717	<	* Returns the number of parties that have arrived at the current
718	<	* phase of this barrier.
717	>	* Returns the number of registered parties that have arrived at
718	>	* the current phase of this barrier.
719	>	*
720		* @return the number of arrived parties
721		*/
722		public int getArrivedParties() {
723	<	return arrivedOf(state.get());
723	>	return arrivedOf(state);
724		}
725
726		/**
727		* Returns the number of registered parties that have not yet
728		* arrived at the current phase of this barrier.
729	+	*
730		* @return the number of unarrived parties
731		*/
732		public int getUnarrivedParties() {
733	<	return unarrivedOf(state.get());
733	>	return unarrivedOf(state);
734	>	}
735	>
736	>	/**
737	>	* Returns the parent of this phaser, or {@code null} if none.
738	>	*
739	>	* @return the parent of this phaser, or {@code null} if none
740	>	*/
741	>	public Phaser getParent() {
742	>	return parent;
743	>	}
744	>
745	>	/**
746	>	* Returns the root ancestor of this phaser, which is the same as
747	>	* this phaser if it has no parent.
748	>	*
749	>	* @return the root ancestor of this phaser
750	>	*/
751	>	public Phaser getRoot() {
752	>	return root;
753		}
754
755		/**
756	<	* Returns true if this barrier has been terminated
757	<	* @return true if this barrier has been terminated
756	>	* Returns {@code true} if this barrier has been terminated.
757	>	*
758	>	* @return {@code true} if this barrier has been terminated
759		*/
760		public boolean isTerminated() {
761	<	return phaseOf(state.get()) < 0;
761	>	return getPhase() < 0;
762		}
763
764		/**
765		* Overridable method to perform an action upon phase advance, and
766	<	* to control termination. This method is invoked whenever the
767	<	* barrier is tripped (and thus all other waiting parties are
768	<	* dormant). If it returns true, then, rather than advance the
769	<	* phase number, this barrier will be set to a final termination
770	<	* state, and subsequent calls to <tt>isTerminated</tt> will
771	<	* return true.
772	<	*
773	<	* <p> The default version returns true when the number of
766	>	* to control termination. This method is invoked upon arrival of
767	>	* the party tripping the barrier (when all other waiting parties
768	>	* are dormant).  If this method returns {@code true}, then,
769	>	* rather than advance the phase number, this barrier will be set
770	>	* to a final termination state, and subsequent calls to {@link
771	>	* #isTerminated} will return true. Any (unchecked) Exception or
772	>	* Error thrown by an invocation of this method is propagated to
773	>	* the party attempting to trip the barrier, in which case no
774	>	* advance occurs.
775	>	*
776	>	* <p>The arguments to this method provide the state of the phaser
777	>	* prevailing for the current transition. (When called from within
778	>	* an implementation of {@code onAdvance} the values returned by
779	>	* methods such as {@code getPhase} may or may not reliably
780	>	* indicate the state to which this transition applies.)
781	>	*
782	>	* <p>The default version returns {@code true} when the number of
783		* registered parties is zero. Normally, overrides that arrange
784		* termination for other reasons should also preserve this
785		* property.
786		*
787	+	* <p>You may override this method to perform an action with side
788	+	* effects visible to participating tasks, but it is in general
789	+	* only sensible to do so in designs where all parties register
790	+	* before any arrive, and all {@link #awaitAdvance} at each phase.
791	+	* Otherwise, you cannot ensure lack of interference from other
792	+	* parties during the invocation of this method.
793	+	*
794		* @param phase the phase number on entering the barrier
795	<	* @param registeredParties the current number of registered
796	<	* parties.
458	<	* @return true if this barrier should terminate
795	>	* @param registeredParties the current number of registered parties
796	>	* @return {@code true} if this barrier should terminate
797		*/
798		protected boolean onAdvance(int phase, int registeredParties) {
799		return registeredParties <= 0;
800		}
801
802		/**
803	<	* Returns a string identifying this barrier, as well as its
803	>	* Returns a string identifying this phaser, as well as its
804		* state.  The state, in brackets, includes the String {@code
805	<	* "phase ="} followed by the phase number, {@code "parties ="}
805	>	* "phase = "} followed by the phase number, {@code "parties = "}
806		* followed by the number of registered parties, and {@code
807	<	* "arrived ="} followed by the number of arrived parties
807	>	* "arrived = "} followed by the number of arrived parties.
808		*
809		* @return a string identifying this barrier, as well as its state
810		*/
811		public String toString() {
812	<	long s = state.get();
813	<	return super.toString() + "[phase = " + phaseOf(s) + " parties = " + partiesOf(s) + " arrived = " + arrivedOf(s) + "]";
812	>	long s = getReconciledState();
813	>	return super.toString() +
814	>	"[phase = " + phaseOf(s) +
815	>	" parties = " + partiesOf(s) +
816	>	" arrived = " + arrivedOf(s) + "]";
817		}
818
819	<	// methods for tripping and waiting
819	>	// methods for waiting
820
821		/**
822	<	* Advance the current phase (or terminate)
822	>	* Wait nodes for Treiber stack representing wait queue
823		*/
824	<	private void trip(int phase, int parties) {
825	<	int next = onAdvance(phase, parties)? -1 : ((phase + 1) & phaseMask);
826	<	state.set(stateFor(next, parties, parties));
827	<	if (head.get() != null)
828	<	releaseWaiters(next);
829	<	}
830	<
831	<	private int helpingWait(int phase) {
832	<	final AtomicLong state = this.state;
833	<	int p;
834	<	while ((p = phaseOf(state.get())) == phase) {
835	<	ForkJoinTask<?> t = ForkJoinWorkerThread.pollTask();
836	<	if (t != null) {
837	<	if ((p = phaseOf(state.get())) == phase)
838	<	t.exec();
839	<	else {   // push task and exit if barrier advanced
840	<	t.fork();
841	<	break;
842	<	}
502	<	}
824	>	static final class QNode implements ForkJoinPool.ManagedBlocker {
825	>	final Phaser phaser;
826	>	final int phase;
827	>	final long startTime;
828	>	final long nanos;
829	>	final boolean timed;
830	>	final boolean interruptible;
831	>	volatile boolean wasInterrupted = false;
832	>	volatile Thread thread; // nulled to cancel wait
833	>	QNode next;
834	>	QNode(Phaser phaser, int phase, boolean interruptible,
835	>	boolean timed, long startTime, long nanos) {
836	>	this.phaser = phaser;
837	>	this.phase = phase;
838	>	this.timed = timed;
839	>	this.interruptible = interruptible;
840	>	this.startTime = startTime;
841	>	this.nanos = nanos;
842	>	thread = Thread.currentThread();
843		}
844	<	return p;
845	<	}
846	<
847	<	private int timedHelpingWait(int phase, long nanos) throws TimeoutException {
848	<	final AtomicLong state = this.state;
849	<	long lastTime = System.nanoTime();
850	<	int p;
851	<	while ((p = phaseOf(state.get())) == phase) {
852	<	long now = System.nanoTime();
853	<	nanos -= now - lastTime;
854	<	lastTime = now;
515	<	if (nanos <= 0) {
516	<	if ((p = phaseOf(state.get())) == phase)
517	<	throw new TimeoutException();
518	<	else
519	<	break;
844	>	public boolean isReleasable() {
845	>	return (thread == null ||
846	>	phaser.getPhase() != phase ||
847	>	(interruptible && wasInterrupted) ||
848	>	(timed && (nanos - (System.nanoTime() - startTime)) <= 0));
849	>	}
850	>	public boolean block() {
851	>	if (Thread.interrupted()) {
852	>	wasInterrupted = true;
853	>	if (interruptible)
854	>	return true;
855		}
856	<	ForkJoinTask<?> t = ForkJoinWorkerThread.pollTask();
856	>	if (!timed)
857	>	LockSupport.park(this);
858	>	else {
859	>	long waitTime = nanos - (System.nanoTime() - startTime);
860	>	if (waitTime <= 0)
861	>	return true;
862	>	LockSupport.parkNanos(this, waitTime);
863	>	}
864	>	return isReleasable();
865	>	}
866	>	void signal() {
867	>	Thread t = thread;
868		if (t != null) {
869	<	if ((p = phaseOf(state.get())) == phase)
870	<	t.exec();
871	<	else {   // push task and exit if barrier advanced
872	<	t.fork();
873	<	break;
869	>	thread = null;
870	>	LockSupport.unpark(t);
871	>	}
872	>	}
873	>	boolean doWait() {
874	>	if (thread != null) {
875	>	try {
876	>	ForkJoinPool.managedBlock(this, false);
877	>	} catch (InterruptedException ie) {
878		}
879		}
880	+	return wasInterrupted;
881		}
882	<	return p;
882	>
883		}
884
885		/**
886	<	* Wait nodes for Treiber stack representing wait queue for non-FJ
536	<	* tasks. The waiting scheme is an adaptation of the one used in
537	<	* forkjoin.PoolBarrier.
886	>	* Removes and signals waiting threads from wait queue.
887		*/
888	<	static final class QNode {
889	<	QNode next;
890	<	volatile Thread thread; // nulled to cancel wait
891	<	final int phase;
892	<	QNode(Thread t, int c) {
893	<	thread = t;
545	<	phase = c;
546	<	}
547	<	}
548	<
549	<	private void releaseWaiters(int currentPhase) {
550	<	final AtomicReference<QNode> head = this.head;
551	<	QNode p;
552	<	while ((p = head.get()) != null && p.phase != currentPhase) {
553	<	if (head.compareAndSet(p, null)) {
554	<	do {
555	<	Thread t = p.thread;
556	<	if (t != null) {
557	<	p.thread = null;
558	<	LockSupport.unpark(t);
559	<	}
560	<	} while ((p = p.next) != null);
561	<	}
888	>	private void releaseWaiters(int phase) {
889	>	AtomicReference<QNode> head = queueFor(phase);
890	>	QNode q;
891	>	while ((q = head.get()) != null) {
892	>	if (head.compareAndSet(q, q.next))
893	>	q.signal();
894		}
895		}
896
565	–	/** The number of CPUs, for spin control */
566	–	static final int NCPUS = Runtime.getRuntime().availableProcessors();
567	–
897		/**
898	<	* The number of times to spin before blocking in timed waits.
899	<	* The value is empirically derived
900	<	*/
572	<	static final int maxTimedSpins = (NCPUS < 2)? 0 : 32;
573	<
574	<	/**
575	<	* The number of times to spin before blocking in untimed waits.
576	<	* This is greater than timed value because untimed waits spin
577	<	* faster since they don't need to check times on each spin.
898	>	* Tries to enqueue given node in the appropriate wait queue.
899	>	*
900	>	* @return true if successful
901		*/
902	<	static final int maxUntimedSpins = maxTimedSpins * 32;
902	>	private boolean tryEnqueue(QNode node) {
903	>	AtomicReference<QNode> head = queueFor(node.phase);
904	>	return head.compareAndSet(node.next = head.get(), node);
905	>	}
906
907		/**
908	<	* The number of nanoseconds for which it is faster to spin
909	<	* rather than to use timed park. A rough estimate suffices.
908	>	* Enqueues node and waits unless aborted or signalled.
909	>	*
910	>	* @return current phase
911		*/
912	<	static final long spinForTimeoutThreshold = 1000L;
912	>	private int untimedWait(int phase) {
913	>	QNode node = null;
914	>	boolean queued = false;
915	>	boolean interrupted = false;
916	>	int p;
917	>	while ((p = getPhase()) == phase) {
918	>	if (Thread.interrupted())
919	>	interrupted = true;
920	>	else if (node == null)
921	>	node = new QNode(this, phase, false, false, 0, 0);
922	>	else if (!queued)
923	>	queued = tryEnqueue(node);
924	>	else
925	>	interrupted = node.doWait();
926	>	}
927	>	if (node != null)
928	>	node.thread = null;
929	>	releaseWaiters(phase);
930	>	if (interrupted)
931	>	Thread.currentThread().interrupt();
932	>	return p;
933	>	}
934
935		/**
936	<	* Enqueues node and waits unless aborted or signalled.
936	>	* Interruptible version
937	>	* @return current phase
938		*/
939	<	private boolean untimedWait(Thread thread, int currentPhase,
591	<	boolean abortOnInterrupt) {
592	<	final AtomicReference<QNode> head = this.head;
593	<	final AtomicLong state = this.state;
594	<	boolean wasInterrupted = false;
939	>	private int interruptibleWait(int phase) throws InterruptedException {
940		QNode node = null;
941		boolean queued = false;
942	<	int spins = maxUntimedSpins;
943	<	while (phaseOf(state.get()) == currentPhase) {
944	<	QNode h;
945	<	if (node != null && queued) {
946	<	if (node.thread != null) {
947	<	LockSupport.park();
948	<	if (Thread.interrupted()) {
949	<	wasInterrupted = true;
950	<	if (abortOnInterrupt)
606	<	break;
607	<	}
608	<	}
609	<	}
610	<	else if ((h = head.get()) != null && h.phase != currentPhase) {
611	<	if (phaseOf(state.get()) == currentPhase) { // must recheck
612	<	if (head.compareAndSet(h, h.next)) {
613	<	Thread t = h.thread; // help clear out old waiters
614	<	if (t != null) {
615	<	h.thread = null;
616	<	LockSupport.unpark(t);
617	<	}
618	<	}
619	<	}
620	<	else
621	<	break;
622	<	}
623	<	else if (node != null)
624	<	queued = head.compareAndSet(node.next = h, node);
625	<	else if (spins <= 0)
626	<	node = new QNode(thread, currentPhase);
942	>	boolean interrupted = false;
943	>	int p;
944	>	while ((p = getPhase()) == phase && !interrupted) {
945	>	if (Thread.interrupted())
946	>	interrupted = true;
947	>	else if (node == null)
948	>	node = new QNode(this, phase, true, false, 0, 0);
949	>	else if (!queued)
950	>	queued = tryEnqueue(node);
951		else
952	<	--spins;
952	>	interrupted = node.doWait();
953		}
954		if (node != null)
955		node.thread = null;
956	<	return wasInterrupted;
956	>	if (p != phase || (p = getPhase()) != phase)
957	>	releaseWaiters(phase);
958	>	if (interrupted)
959	>	throw new InterruptedException();
960	>	return p;
961		}
962
963		/**
964	<	* Messier timeout version
964	>	* Timeout version.
965	>	* @return current phase
966		*/
967	<	private void timedWait(Thread thread, int currentPhase, long nanos)
967	>	private int timedWait(int phase, long nanos)
968		throws InterruptedException, TimeoutException {
969	<	final AtomicReference<QNode> head = this.head;
641	<	final AtomicLong state = this.state;
642	<	long lastTime = System.nanoTime();
969	>	long startTime = System.nanoTime();
970		QNode node = null;
971		boolean queued = false;
972	<	int spins = maxTimedSpins;
973	<	while (phaseOf(state.get()) == currentPhase) {
974	<	QNode h;
975	<	long now = System.nanoTime();
976	<	nanos -= now - lastTime;
977	<	lastTime = now;
978	<	if (nanos <= 0) {
979	<	if (node != null)
980	<	node.thread = null;
981	<	if (phaseOf(state.get()) == currentPhase)
982	<	throw new TimeoutException();
656	<	else
657	<	break;
658	<	}
659	<	else if (node != null && queued) {
660	<	if (node.thread != null &&
661	<	nanos > spinForTimeoutThreshold) {
662	<	//                LockSupport.parkNanos(this, nanos);
663	<	LockSupport.parkNanos(nanos);
664	<	if (Thread.interrupted()) {
665	<	node.thread = null;
666	<	throw new InterruptedException();
667	<	}
668	<	}
669	<	}
670	<	else if ((h = head.get()) != null && h.phase != currentPhase) {
671	<	if (phaseOf(state.get()) == currentPhase) { // must recheck
672	<	if (head.compareAndSet(h, h.next)) {
673	<	Thread t = h.thread; // help clear out old waiters
674	<	if (t != null) {
675	<	h.thread = null;
676	<	LockSupport.unpark(t);
677	<	}
678	<	}
679	<	}
680	<	else
681	<	break;
682	<	}
683	<	else if (node != null)
684	<	queued = head.compareAndSet(node.next = h, node);
685	<	else if (spins <= 0)
686	<	node = new QNode(thread, currentPhase);
972	>	boolean interrupted = false;
973	>	int p;
974	>	while ((p = getPhase()) == phase && !interrupted) {
975	>	if (Thread.interrupted())
976	>	interrupted = true;
977	>	else if (nanos - (System.nanoTime() - startTime) <= 0)
978	>	break;
979	>	else if (node == null)
980	>	node = new QNode(this, phase, true, true, startTime, nanos);
981	>	else if (!queued)
982	>	queued = tryEnqueue(node);
983		else
984	<	--spins;
984	>	interrupted = node.doWait();
985		}
986		if (node != null)
987		node.thread = null;
988	+	if (p != phase || (p = getPhase()) != phase)
989	+	releaseWaiters(phase);
990	+	if (interrupted)
991	+	throw new InterruptedException();
992	+	if (p == phase)
993	+	throw new TimeoutException();
994	+	return p;
995		}
996
997	<	}
997	>	// Unsafe mechanics
998
999	+	private static final sun.misc.Unsafe UNSAFE = getUnsafe();
1000	+	private static final long stateOffset =
1001	+	objectFieldOffset("state", Phaser.class);
1002	+
1003	+	private final boolean casState(long cmp, long val) {
1004	+	return UNSAFE.compareAndSwapLong(this, stateOffset, cmp, val);
1005	+	}
1006	+
1007	+	private static long objectFieldOffset(String field, Class<?> klazz) {
1008	+	try {
1009	+	return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
1010	+	} catch (NoSuchFieldException e) {
1011	+	// Convert Exception to corresponding Error
1012	+	NoSuchFieldError error = new NoSuchFieldError(field);
1013	+	error.initCause(e);
1014	+	throw error;
1015	+	}
1016	+	}
1017	+
1018	+	/**
1019	+	* Returns a sun.misc.Unsafe.  Suitable for use in a 3rd party package.
1020	+	* Replace with a simple call to Unsafe.getUnsafe when integrating
1021	+	* into a jdk.
1022	+	*
1023	+	* @return a sun.misc.Unsafe
1024	+	*/
1025	+	private static sun.misc.Unsafe getUnsafe() {
1026	+	try {
1027	+	return sun.misc.Unsafe.getUnsafe();
1028	+	} catch (SecurityException se) {
1029	+	try {
1030	+	return java.security.AccessController.doPrivileged
1031	+	(new java.security
1032	+	.PrivilegedExceptionAction<sun.misc.Unsafe>() {
1033	+	public sun.misc.Unsafe run() throws Exception {
1034	+	java.lang.reflect.Field f = sun.misc
1035	+	.Unsafe.class.getDeclaredField("theUnsafe");
1036	+	f.setAccessible(true);
1037	+	return (sun.misc.Unsafe) f.get(null);
1038	+	}});
1039	+	} catch (java.security.PrivilegedActionException e) {
1040	+	throw new RuntimeException("Could not initialize intrinsics",
1041	+	e.getCause());
1042	+	}
1043	+	}
1044	+	}
1045	+	}

Diff Legend

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