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 1.45 by jsr166, Tue Aug 25 16:50:24 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 (final 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();}</pre>
166		*
167	<	* <p>[todo: non-FJ example]
167	>	* <p>Related constructions may be used to await particular phase numbers
168	>	* in contexts where you are sure that the phase will never wrap around
169	>	* {@code Integer.MAX_VALUE}. For example:
170		*
171	<	* <p> A Phaser may be used to support a style of programming in
172	<	* which a task waits for others to complete, without otherwise
173	<	* needing to keep track of which tasks it is waiting for. This is
174	<	* similar to the "sync" construct in Cilk and "clocks" in X10.
175	<	* Special constructions based on such barriers are available using
176	<	* the <tt>LinkedAsyncAction</tt> and <tt>CyclicAction</tt> classes,
177	<	* but they can be useful in other contexts as well.  For a simple
178	<	* (but not very useful) example, here is a variant of Fibonacci:
87	<	*
88	<	* <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();
171	>	*  <pre> {@code
172	>	* void awaitPhase(Phaser phaser, int phase) {
173	>	*   int p = phaser.register(); // assumes caller not already registered
174	>	*   while (p < phase) {
175	>	*     if (phaser.isTerminated())
176	>	*       // ... deal with unexpected termination
177	>	*     else
178	>	*       p = phaser.arriveAndAwaitAdvance();
179		*   }
180	<	*   protected void compute() {
181	<	*     int n = argument;
182	<	*     if (n &lt;= 1)
183	<	*        result = n;
184	<	*     else {
185	<	*        Phaser childBarrier = new Phaser(1);
186	<	*        BarrierFibonacci f1 = new BarrierFibonacci(n - 1, childBarrier);
187	<	*        BarrierFibonacci f2 = new BarrierFibonacci(n - 2, childBarrier);
188	<	*        f1.fork();
189	<	*        f2.fork();
190	<	*        childBarrier.arriveAndAwait();
191	<	*        result = f1.result + f2.result;
180	>	*   phaser.arriveAndDeregister();
181	>	* }}</pre>
182	>	*
183	>	*
184	>	* <p>To create a set of tasks using a tree of phasers,
185	>	* you could use code of the following form, assuming a
186	>	* Task class with a constructor accepting a phaser that
187	>	* it registers for upon construction:
188	>	*
189	>	*  <pre> {@code
190	>	* void build(Task[] actions, int lo, int hi, Phaser ph) {
191	>	*   if (hi - lo > TASKS_PER_PHASER) {
192	>	*     for (int i = lo; i < hi; i += TASKS_PER_PHASER) {
193	>	*       int j = Math.min(i + TASKS_PER_PHASER, hi);
194	>	*       build(actions, i, j, new Phaser(ph));
195		*     }
196	<	*     parentBarrier.arriveAndDeregister();
196	>	*   } else {
197	>	*     for (int i = lo; i < hi; ++i)
198	>	*       actions[i] = new Task(ph);
199	>	*       // assumes new Task(ph) performs ph.register()
200		*   }
201		* }
202	+	* // .. initially called, for n tasks via
203	+	* build(new Task[n], 0, n, new Phaser());}</pre>
204	+	*
205	+	* The best value of {@code TASKS_PER_PHASER} depends mainly on
206	+	* expected barrier synchronization rates. A value as low as four may
207	+	* be appropriate for extremely small per-barrier task bodies (thus
208	+	* high rates), or up to hundreds for extremely large ones.
209	+	*
210		* </pre>
211		*
212		* <p><b>Implementation notes</b>: This implementation restricts the
213	<	* maximum number of parties to 65535. Attempts to register
214	<	* additional parties result in IllegalStateExceptions.
213	>	* maximum number of parties to 65535. Attempts to register additional
214	>	* parties result in {@code IllegalStateException}. However, you can and
215	>	* should create tiered phasers to accommodate arbitrarily large sets
216	>	* of participants.
217	>	*
218	>	* @since 1.7
219	>	* @author Doug Lea
220		*/
221		public class Phaser {
222		/*
223		* This class implements an extension of X10 "clocks".  Thanks to
224	<	* Vijay Saraswat for the idea of applying it to ForkJoinTasks,
225	<	* and to Vivek Sarkar for enhancements to extend functionality.
224	>	* Vijay Saraswat for the idea, and to Vivek Sarkar for
225	>	* enhancements to extend functionality.
226		*/
227
228		/**
229		* Barrier state representation. Conceptually, a barrier contains
230		* four values:
231	<	*
231	>	*
232		* * parties -- the number of parties to wait (16 bits)
233		* * unarrived -- the number of parties yet to hit barrier (16 bits)
234		* * phase -- the generation of the barrier (31 bits)
235		* * terminated -- set if barrier is terminated (1 bit)
236		*
237		* However, to efficiently maintain atomicity, these values are
238	<	* packed into a single AtomicLong. Termination uses the sign bit
239	<	* of 32 bit representation of phase, so phase is set to -1 on
240	<	* termination.
241	<	*/
242	<	private final AtomicLong state;
243	<
244	<	/**
143	<	* Head of Treiber stack for waiting nonFJ threads.
238	>	* packed into a single (atomic) long. Termination uses the sign
239	>	* bit of 32 bit representation of phase, so phase is set to -1 on
240	>	* termination. Good performance relies on keeping state decoding
241	>	* and encoding simple, and keeping race windows short.
242	>	*
243	>	* Note: there are some cheats in arrive() that rely on unarrived
244	>	* count being lowest 16 bits.
245		*/
246	<	private final AtomicReference<QNode> head = new AtomicReference<QNode>();
246	>	private volatile long state;
247
248	<	private static final int ushortBits = 16;
249	<	private static final int ushortMask =  (1 << ushortBits) - 1;
149	<	private static final int phaseMask = 0x7fffffff;
248	>	private static final int ushortMask = 0xffff;
249	>	private static final int phaseMask  = 0x7fffffff;
250
251		private static int unarrivedOf(long s) {
252	<	return (int)(s & ushortMask);
252	>	return (int) (s & ushortMask);
253		}
254
255		private static int partiesOf(long s) {
256	<	return (int)(s & (ushortMask << 16)) >>> 16;
256	>	return ((int) s) >>> 16;
257		}
258
259		private static int phaseOf(long s) {
260	<	return (int)(s >>> 32);
260	>	return (int) (s >>> 32);
261		}
262
263		private static int arrivedOf(long s) {
#	Line 165 | Line 265 | public class Phaser {
265		}
266
267		private static long stateFor(int phase, int parties, int unarrived) {
268	<	return (((long)phase) << 32) | ((parties << 16) | unarrived);
268	>	return ((((long) phase) << 32) | (((long) parties) << 16) |
269	>	(long) unarrived);
270		}
271
272	<	private static IllegalStateException badBounds(int parties, int unarrived) {
273	<	return new IllegalStateException("Attempt to set " + unarrived +
274	<	" unarrived of " + parties + " parties");
272	>	private static long trippedStateFor(int phase, int parties) {
273	>	long lp = (long) parties;
274	>	return (((long) phase) << 32) | (lp << 16) | lp;
275		}
276
277		/**
278	<	* Creates a new Phaser without any initially registered parties,
279	<	* and initial phase number 0.
278	>	* Returns message string for bad bounds exceptions.
279	>	*/
280	>	private static String badBounds(int parties, int unarrived) {
281	>	return ("Attempt to set " + unarrived +
282	>	" unarrived of " + parties + " parties");
283	>	}
284	>
285	>	/**
286	>	* The parent of this phaser, or null if none
287	>	*/
288	>	private final Phaser parent;
289	>
290	>	/**
291	>	* The root of phaser tree. Equals this if not in a tree.  Used to
292	>	* support faster state push-down.
293	>	*/
294	>	private final Phaser root;
295	>
296	>	// Wait queues
297	>
298	>	/**
299	>	* Heads of Treiber stacks for waiting threads. To eliminate
300	>	* contention while releasing some threads while adding others, we
301	>	* use two of them, alternating across even and odd phases.
302	>	*/
303	>	private final AtomicReference<QNode> evenQ = new AtomicReference<QNode>();
304	>	private final AtomicReference<QNode> oddQ  = new AtomicReference<QNode>();
305	>
306	>	private AtomicReference<QNode> queueFor(int phase) {
307	>	return ((phase & 1) == 0) ? evenQ : oddQ;
308	>	}
309	>
310	>	/**
311	>	* Returns current state, first resolving lagged propagation from
312	>	* root if necessary.
313	>	*/
314	>	private long getReconciledState() {
315	>	return (parent == null) ? state : reconcileState();
316	>	}
317	>
318	>	/**
319	>	* Recursively resolves state.
320	>	*/
321	>	private long reconcileState() {
322	>	Phaser p = parent;
323	>	long s = state;
324	>	if (p != null) {
325	>	while (unarrivedOf(s) == 0 && phaseOf(s) != phaseOf(root.state)) {
326	>	long parentState = p.getReconciledState();
327	>	int parentPhase = phaseOf(parentState);
328	>	int phase = phaseOf(s = state);
329	>	if (phase != parentPhase) {
330	>	long next = trippedStateFor(parentPhase, partiesOf(s));
331	>	if (casState(s, next)) {
332	>	releaseWaiters(phase);
333	>	s = next;
334	>	}
335	>	}
336	>	}
337	>	}
338	>	return s;
339	>	}
340	>
341	>	/**
342	>	* Creates a new phaser without any initially registered parties,
343	>	* initial phase number 0, and no parent. Any thread using this
344	>	* phaser will need to first register for it.
345		*/
346		public Phaser() {
347	<	state = new AtomicLong(stateFor(0, 0, 0));
347	>	this(null);
348		}
349
350		/**
351	<	* Creates a new Phaser with the given numbers of registered
352	<	* unarrived parties and initial phase number 0.
353	<	* @param parties the number of parties required to trip barrier.
351	>	* Creates a new phaser with the given numbers of registered
352	>	* unarrived parties, initial phase number 0, and no parent.
353	>	*
354	>	* @param parties the number of parties required to trip barrier
355		* @throws IllegalArgumentException if parties less than zero
356	<	* or greater than the maximum number of parties supported.
356	>	* or greater than the maximum number of parties supported
357		*/
358		public Phaser(int parties) {
359	+	this(null, parties);
360	+	}
361	+
362	+	/**
363	+	* Creates a new phaser with the given parent, without any
364	+	* initially registered parties. If parent is non-null this phaser
365	+	* is registered with the parent and its initial phase number is
366	+	* the same as that of parent phaser.
367	+	*
368	+	* @param parent the parent phaser
369	+	*/
370	+	public Phaser(Phaser parent) {
371	+	int phase = 0;
372	+	this.parent = parent;
373	+	if (parent != null) {
374	+	this.root = parent.root;
375	+	phase = parent.register();
376	+	}
377	+	else
378	+	this.root = this;
379	+	this.state = trippedStateFor(phase, 0);
380	+	}
381	+
382	+	/**
383	+	* Creates a new phaser with the given parent and numbers of
384	+	* registered unarrived parties. If parent is non-null, this phaser
385	+	* is registered with the parent and its initial phase number is
386	+	* the same as that of parent phaser.
387	+	*
388	+	* @param parent the parent phaser
389	+	* @param parties the number of parties required to trip barrier
390	+	* @throws IllegalArgumentException if parties less than zero
391	+	* or greater than the maximum number of parties supported
392	+	*/
393	+	public Phaser(Phaser parent, int parties) {
394		if (parties < 0 || parties > ushortMask)
395		throw new IllegalArgumentException("Illegal number of parties");
396	<	state = new AtomicLong(stateFor(0, parties, parties));
396	>	int phase = 0;
397	>	this.parent = parent;
398	>	if (parent != null) {
399	>	this.root = parent.root;
400	>	phase = parent.register();
401	>	}
402	>	else
403	>	this.root = this;
404	>	this.state = trippedStateFor(phase, parties);
405		}
406
407		/**
408		* Adds a new unarrived party to this phaser.
409	<	* @return the current barrier phase number upon registration
409	>	*
410	>	* @return the arrival phase number to which this registration applied
411	>	* @throws IllegalStateException if attempting to register more
412	>	* than the maximum supported number of parties
413	>	*/
414	>	public int register() {
415	>	return doRegister(1);
416	>	}
417	>
418	>	/**
419	>	* Adds the given number of new unarrived parties to this phaser.
420	>	*
421	>	* @param parties the number of parties required to trip barrier
422	>	* @return the arrival phase number to which this registration applied
423		* @throws IllegalStateException if attempting to register more
424	<	* than the maximum supported number of parties.
424	>	* than the maximum supported number of parties
425		*/
426	<	public int register() { // increment both parties and unarrived
427	<	final AtomicLong state = this.state;
426	>	public int bulkRegister(int parties) {
427	>	if (parties < 0)
428	>	throw new IllegalArgumentException();
429	>	if (parties == 0)
430	>	return getPhase();
431	>	return doRegister(parties);
432	>	}
433	>
434	>	/**
435	>	* Shared code for register, bulkRegister
436	>	*/
437	>	private int doRegister(int registrations) {
438	>	int phase;
439		for (;;) {
440	<	long s = state.get();
441	<	int phase = phaseOf(s);
442	<	int parties = partiesOf(s) + 1;
443	<	int unarrived = unarrivedOf(s) + 1;
440	>	long s = getReconciledState();
441	>	phase = phaseOf(s);
442	>	int unarrived = unarrivedOf(s) + registrations;
443	>	int parties = partiesOf(s) + registrations;
444	>	if (phase < 0)
445	>	break;
446		if (parties > ushortMask || unarrived > ushortMask)
447	<	throw badBounds(parties, unarrived);
448	<	if (state.compareAndSet(s, stateFor(phase, parties, unarrived)))
449	<	return phase;
447	>	throw new IllegalStateException(badBounds(parties, unarrived));
448	>	if (phase == phaseOf(root.state) &&
449	>	casState(s, stateFor(phase, parties, unarrived)))
450	>	break;
451		}
452	+	return phase;
453		}
454
455		/**
456		* Arrives at the barrier, but does not wait for others.  (You can
457	<	* in turn wait for others via {@link #awaitAdvance}).
457	>	* in turn wait for others via {@link #awaitAdvance}).  It is an
458	>	* unenforced usage error for an unregistered party to invoke this
459	>	* method.
460		*
461	<	* @return the current barrier phase number upon entry to
462	<	* this method, or a negative value if terminated;
463	<	* @throws IllegalStateException if the number of unarrived
224	<	* parties would become negative.
461	>	* @return the arrival phase number, or a negative value if terminated
462	>	* @throws IllegalStateException if not terminated and the number
463	>	* of unarrived parties would become negative
464		*/
465	<	public int arrive() { // decrement unarrived. If zero, trip
466	<	final AtomicLong state = this.state;
465	>	public int arrive() {
466	>	int phase;
467		for (;;) {
468	<	long s = state.get();
469	<	int phase = phaseOf(s);
468	>	long s = state;
469	>	phase = phaseOf(s);
470	>	if (phase < 0)
471	>	break;
472		int parties = partiesOf(s);
473		int unarrived = unarrivedOf(s) - 1;
474	<	if (unarrived < 0)
475	<	throw badBounds(parties, unarrived);
476	<	if (unarrived == 0 && phase >= 0) {
236	<	trip(phase, parties);
237	<	return phase;
474	>	if (unarrived > 0) {        // Not the last arrival
475	>	if (casState(s, s - 1)) // s-1 adds one arrival
476	>	break;
477		}
478	<	if (state.compareAndSet(s, stateFor(phase, parties, unarrived)))
479	<	return phase;
478	>	else if (unarrived == 0) {  // the last arrival
479	>	Phaser par = parent;
480	>	if (par == null) {      // directly trip
481	>	if (casState
482	>	(s,
483	>	trippedStateFor(onAdvance(phase, parties) ? -1 :
484	>	((phase + 1) & phaseMask), parties))) {
485	>	releaseWaiters(phase);
486	>	break;
487	>	}
488	>	}
489	>	else {                  // cascade to parent
490	>	if (casState(s, s - 1)) { // zeroes unarrived
491	>	par.arrive();
492	>	reconcileState();
493	>	break;
494	>	}
495	>	}
496	>	}
497	>	else if (phase != phaseOf(root.state)) // or if unreconciled
498	>	reconcileState();
499	>	else
500	>	throw new IllegalStateException(badBounds(parties, unarrived));
501		}
502	+	return phase;
503		}
504
505		/**
506	<	* Arrives at the barrier, and deregisters from it, without
507	<	* waiting for others.
506	>	* Arrives at the barrier and deregisters from it without waiting
507	>	* for others. Deregistration reduces the number of parties
508	>	* required to trip the barrier in future phases.  If this phaser
509	>	* has a parent, and deregistration causes this phaser to have
510	>	* zero parties, this phaser also arrives at and is deregistered
511	>	* from its parent.  It is an unenforced usage error for an
512	>	* unregistered party to invoke this method.
513		*
514	<	* @return the current barrier phase number upon entry to
515	<	* this method, or a negative value if terminated;
516	<	* @throws IllegalStateException if the number of registered or
517	<	* unarrived parties would become negative.
518	<	*/
519	<	public int arriveAndDeregister() { // Same as arrive, plus decrement parties
520	<	final AtomicLong state = this.state;
514	>	* @return the arrival phase number, or a negative value if terminated
515	>	* @throws IllegalStateException if not terminated and the number
516	>	* of registered or unarrived parties would become negative
517	>	*/
518	>	public int arriveAndDeregister() {
519	>	// similar code to arrive, but too different to merge
520	>	Phaser par = parent;
521	>	int phase;
522		for (;;) {
523	<	long s = state.get();
524	<	int phase = phaseOf(s);
523	>	long s = state;
524	>	phase = phaseOf(s);
525	>	if (phase < 0)
526	>	break;
527		int parties = partiesOf(s) - 1;
528		int unarrived = unarrivedOf(s) - 1;
529	<	if (parties < 0 || unarrived < 0)
530	<	throw badBounds(parties, unarrived);
531	<	if (unarrived == 0 && phase >= 0) {
532	<	trip(phase, parties);
533	<	return phase;
529	>	if (parties >= 0) {
530	>	if (unarrived > 0 || (unarrived == 0 && par != null)) {
531	>	if (casState
532	>	(s,
533	>	stateFor(phase, parties, unarrived))) {
534	>	if (unarrived == 0) {
535	>	par.arriveAndDeregister();
536	>	reconcileState();
537	>	}
538	>	break;
539	>	}
540	>	continue;
541	>	}
542	>	if (unarrived == 0) {
543	>	if (casState
544	>	(s,
545	>	trippedStateFor(onAdvance(phase, parties) ? -1 :
546	>	((phase + 1) & phaseMask), parties))) {
547	>	releaseWaiters(phase);
548	>	break;
549	>	}
550	>	continue;
551	>	}
552	>	if (par != null && phase != phaseOf(root.state)) {
553	>	reconcileState();
554	>	continue;
555	>	}
556		}
557	<	if (state.compareAndSet(s, stateFor(phase, parties, unarrived)))
267	<	return phase;
557	>	throw new IllegalStateException(badBounds(parties, unarrived));
558		}
559	+	return phase;
560		}
561
562		/**
563	<	* Arrives at the barrier and awaits others. Unlike other arrival
564	<	* methods, this method returns the arrival index of the
565	<	* caller. The caller tripping the barrier returns zero, the
566	<	* previous caller 1, and so on.
567	<	* @return the arrival index
568	<	* @throws IllegalStateException if the number of unarrived
569	<	* parties would become negative.
563	>	* Arrives at the barrier and awaits others. Equivalent in effect
564	>	* to {@code awaitAdvance(arrive())}.  If you need to await with
565	>	* interruption or timeout, you can arrange this with an analogous
566	>	* construction using one of the other forms of the awaitAdvance
567	>	* method.  If instead you need to deregister upon arrival use
568	>	* {@code arriveAndDeregister}. It is an unenforced usage error
569	>	* for an unregistered party to invoke this method.
570	>	*
571	>	* @return the arrival phase number, or a negative number if terminated
572	>	* @throws IllegalStateException if not terminated and the number
573	>	* of unarrived parties would become negative
574		*/
575		public int arriveAndAwaitAdvance() {
576	<	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	<	}
576	>	return awaitAdvance(arrive());
577		}
578
579		/**
580	<	* Awaits the phase of the barrier to advance from the given
581	<	* value, or returns immediately if this barrier is terminated
582	<	* @param phase the phase on entry to this method
583	<	* @return the phase on exit from this method
580	>	* Awaits the phase of the barrier to advance from the given phase
581	>	* value, returning immediately if the current phase of the
582	>	* barrier is not equal to the given phase value or this barrier
583	>	* is terminated.  It is an unenforced usage error for an
584	>	* unregistered party to invoke this method.
585	>	*
586	>	* @param phase an arrival phase number, or negative value if
587	>	* terminated; this argument is normally the value returned by a
588	>	* previous call to {@code arrive} or its variants
589	>	* @return the next arrival phase number, or a negative value
590	>	* if terminated or argument is negative
591		*/
592		public int awaitAdvance(int phase) {
593		if (phase < 0)
594		return phase;
595	<	Thread current = Thread.currentThread();
596	<	if (current instanceof ForkJoinWorkerThread)
597	<	return helpingWait(phase);
598	<	if (untimedWait(current, phase, false))
599	<	current.interrupt();
600	<	return phaseOf(state.get());
595	>	long s = getReconciledState();
596	>	int p = phaseOf(s);
597	>	if (p != phase)
598	>	return p;
599	>	if (unarrivedOf(s) == 0 && parent != null)
600	>	parent.awaitAdvance(phase);
601	>	// Fall here even if parent waited, to reconcile and help release
602	>	return untimedWait(phase);
603		}
604
605		/**
606	<	* Awaits the phase of the barrier to advance from the given
607	<	* value, or returns immediately if this barrier is terminated, or
608	<	* throws InterruptedException if interrupted while waiting.
609	<	* @param phase the phase on entry to this method
610	<	* @return the phase on exit from this method
606	>	* Awaits the phase of the barrier to advance from the given phase
607	>	* value, throwing {@code InterruptedException} if interrupted
608	>	* while waiting, or returning immediately if the current phase of
609	>	* the barrier is not equal to the given phase value or this
610	>	* barrier is terminated. It is an unenforced usage error for an
611	>	* unregistered party to invoke this method.
612	>	*
613	>	* @param phase an arrival phase number, or negative value if
614	>	* terminated; this argument is normally the value returned by a
615	>	* previous call to {@code arrive} or its variants
616	>	* @return the next arrival phase number, or a negative value
617	>	* if terminated or argument is negative
618		* @throws InterruptedException if thread interrupted while waiting
619		*/
620	<	public int awaitAdvanceInterruptibly(int phase) throws InterruptedException {
620	>	public int awaitAdvanceInterruptibly(int phase)
621	>	throws InterruptedException {
622		if (phase < 0)
623		return phase;
624	<	Thread current = Thread.currentThread();
625	<	if (current instanceof ForkJoinWorkerThread)
626	<	return helpingWait(phase);
627	<	else if (Thread.interrupted() || untimedWait(current, phase, true))
628	<	throw new InterruptedException();
629	<	else
630	<	return phaseOf(state.get());
624	>	long s = getReconciledState();
625	>	int p = phaseOf(s);
626	>	if (p != phase)
627	>	return p;
628	>	if (unarrivedOf(s) == 0 && parent != null)
629	>	parent.awaitAdvanceInterruptibly(phase);
630	>	return interruptibleWait(phase);
631		}
632
633		/**
634	<	* Awaits the phase of the barrier to advance from the given value
635	<	* or the given timeout elapses, or returns immediately if this
636	<	* barrier is terminated
637	<	* @param phase the phase on entry to this method
638	<	* @return the phase on exit from this method
634	>	* Awaits the phase of the barrier to advance from the given phase
635	>	* value or the given timeout to elapse, throwing {@code
636	>	* InterruptedException} if interrupted while waiting, or
637	>	* returning immediately if the current phase of the barrier is
638	>	* not equal to the given phase value or this barrier is
639	>	* terminated.  It is an unenforced usage error for an
640	>	* unregistered party to invoke this method.
641	>	*
642	>	* @param phase an arrival phase number, or negative value if
643	>	* terminated; this argument is normally the value returned by a
644	>	* previous call to {@code arrive} or its variants
645	>	* @param timeout how long to wait before giving up, in units of
646	>	*        {@code unit}
647	>	* @param unit a {@code TimeUnit} determining how to interpret the
648	>	*        {@code timeout} parameter
649	>	* @return the next arrival phase number, or a negative value
650	>	* if terminated or argument is negative
651		* @throws InterruptedException if thread interrupted while waiting
652		* @throws TimeoutException if timed out while waiting
653		*/
654	<	public int awaitAdvanceInterruptibly(int phase, long timeout, TimeUnit unit)
654	>	public int awaitAdvanceInterruptibly(int phase,
655	>	long timeout, TimeUnit unit)
656		throws InterruptedException, TimeoutException {
657		if (phase < 0)
658		return phase;
659	<	long nanos = unit.toNanos(timeout);
660	<	Thread current = Thread.currentThread();
661	<	if (current instanceof ForkJoinWorkerThread)
662	<	return timedHelpingWait(phase, nanos);
663	<	timedWait(current, phase, nanos);
664	<	return phaseOf(state.get());
659	>	long s = getReconciledState();
660	>	int p = phaseOf(s);
661	>	if (p != phase)
662	>	return p;
663	>	if (unarrivedOf(s) == 0 && parent != null)
664	>	parent.awaitAdvanceInterruptibly(phase, timeout, unit);
665	>	return timedWait(phase, unit.toNanos(timeout));
666		}
667
668		/**
669		* Forces this barrier to enter termination state. Counts of
670	<	* arrived and registered parties are unaffected. This method may
671	<	* be useful for coordinating recovery after one or more tasks
672	<	* encounter unexpected exceptions.
670	>	* arrived and registered parties are unaffected. If this phaser
671	>	* has a parent, it too is terminated. This method may be useful
672	>	* for coordinating recovery after one or more tasks encounter
673	>	* unexpected exceptions.
674		*/
675		public void forceTermination() {
365	–	final AtomicLong state = this.state;
676		for (;;) {
677	<	long s = state.get();
677	>	long s = getReconciledState();
678		int phase = phaseOf(s);
679		int parties = partiesOf(s);
680		int unarrived = unarrivedOf(s);
681		if (phase < 0 ||
682	<	state.compareAndSet(s, stateFor(-1, parties, unarrived))) {
683	<	if (head.get() != null)
684	<	releaseWaiters(-1);
682	>	casState(s, stateFor(-1, parties, unarrived))) {
683	>	releaseWaiters(0);
684	>	releaseWaiters(1);
685	>	if (parent != null)
686	>	parent.forceTermination();
687		return;
688		}
689		}
690		}
691
692		/**
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	–	/**
693		* Returns the current phase number. The maximum phase number is
694	<	* <tt>Integer.MAX_VALUE</tt>, after which it restarts at
694	>	* {@code Integer.MAX_VALUE}, after which it restarts at
695		* zero. Upon termination, the phase number is negative.
696	+	*
697		* @return the phase number, or a negative value if terminated
698		*/
699	<	public int getPhase() {
700	<	return phaseOf(state.get());
699	>	public final int getPhase() {
700	>	return phaseOf(getReconciledState());
701		}
702
703		/**
704		* Returns the number of parties registered at this barrier.
705	+	*
706		* @return the number of parties
707		*/
708		public int getRegisteredParties() {
709	<	return partiesOf(state.get());
709	>	return partiesOf(state);
710		}
711
712		/**
713	<	* Returns the number of parties that have arrived at the current
714	<	* phase of this barrier.
713	>	* Returns the number of registered parties that have arrived at
714	>	* the current phase of this barrier.
715	>	*
716		* @return the number of arrived parties
717		*/
718		public int getArrivedParties() {
719	<	return arrivedOf(state.get());
719	>	return arrivedOf(state);
720		}
721
722		/**
723		* Returns the number of registered parties that have not yet
724		* arrived at the current phase of this barrier.
725	+	*
726		* @return the number of unarrived parties
727		*/
728		public int getUnarrivedParties() {
729	<	return unarrivedOf(state.get());
729	>	return unarrivedOf(state);
730	>	}
731	>
732	>	/**
733	>	* Returns the parent of this phaser, or {@code null} if none.
734	>	*
735	>	* @return the parent of this phaser, or {@code null} if none
736	>	*/
737	>	public Phaser getParent() {
738	>	return parent;
739	>	}
740	>
741	>	/**
742	>	* Returns the root ancestor of this phaser, which is the same as
743	>	* this phaser if it has no parent.
744	>	*
745	>	* @return the root ancestor of this phaser
746	>	*/
747	>	public Phaser getRoot() {
748	>	return root;
749		}
750
751		/**
752	<	* Returns true if this barrier has been terminated
753	<	* @return true if this barrier has been terminated
752	>	* Returns {@code true} if this barrier has been terminated.
753	>	*
754	>	* @return {@code true} if this barrier has been terminated
755		*/
756		public boolean isTerminated() {
757	<	return phaseOf(state.get()) < 0;
757	>	return getPhase() < 0;
758		}
759
760		/**
761	<	* Overridable method to perform an action upon phase advance, and
762	<	* to control termination. This method is invoked whenever the
763	<	* barrier is tripped (and thus all other waiting parties are
764	<	* dormant). If it returns true, then, rather than advance the
765	<	* phase number, this barrier will be set to a final termination
766	<	* state, and subsequent calls to <tt>isTerminated</tt> will
767	<	* return true.
768	<	*
769	<	* <p> The default version returns true when the number of
761	>	* Overridable method to perform an action upon impending phase
762	>	* advance, and to control termination. This method is invoked
763	>	* upon arrival of the party tripping the barrier (when all other
764	>	* waiting parties are dormant).  If this method returns {@code
765	>	* true}, then, rather than advance the phase number, this barrier
766	>	* will be set to a final termination state, and subsequent calls
767	>	* to {@link #isTerminated} will return true. Any (unchecked)
768	>	* Exception or Error thrown by an invocation of this method is
769	>	* propagated to the party attempting to trip the barrier, in
770	>	* which case no advance occurs.
771	>	*
772	>	* <p>The arguments to this method provide the state of the phaser
773	>	* prevailing for the current transition. (When called from within
774	>	* an implementation of {@code onAdvance} the values returned by
775	>	* methods such as {@code getPhase} may or may not reliably
776	>	* indicate the state to which this transition applies.)
777	>	*
778	>	* <p>The default version returns {@code true} when the number of
779		* registered parties is zero. Normally, overrides that arrange
780		* termination for other reasons should also preserve this
781		* property.
782		*
783	+	* <p>You may override this method to perform an action with side
784	+	* effects visible to participating tasks, but doing so requires
785	+	* care: Method {@code onAdvance} may be invoked more than once
786	+	* per transition.  Further, unless all parties register before
787	+	* any arrive, and all {@link #awaitAdvance} at each phase, then
788	+	* you cannot ensure lack of interference from other parties
789	+	* during the invocation of this method.
790	+	*
791		* @param phase the phase number on entering the barrier
792	<	* @param registeredParties the current number of registered
793	<	* parties.
458	<	* @return true if this barrier should terminate
792	>	* @param registeredParties the current number of registered parties
793	>	* @return {@code true} if this barrier should terminate
794		*/
795		protected boolean onAdvance(int phase, int registeredParties) {
796		return registeredParties <= 0;
797		}
798
799		/**
800	<	* Returns a string identifying this barrier, as well as its
800	>	* Returns a string identifying this phaser, as well as its
801		* state.  The state, in brackets, includes the String {@code
802	<	* "phase ="} followed by the phase number, {@code "parties ="}
802	>	* "phase = "} followed by the phase number, {@code "parties = "}
803		* followed by the number of registered parties, and {@code
804	<	* "arrived ="} followed by the number of arrived parties
804	>	* "arrived = "} followed by the number of arrived parties.
805		*
806		* @return a string identifying this barrier, as well as its state
807		*/
808		public String toString() {
809	<	long s = state.get();
810	<	return super.toString() + "[phase = " + phaseOf(s) + " parties = " + partiesOf(s) + " arrived = " + arrivedOf(s) + "]";
809	>	long s = getReconciledState();
810	>	return super.toString() +
811	>	"[phase = " + phaseOf(s) +
812	>	" parties = " + partiesOf(s) +
813	>	" arrived = " + arrivedOf(s) + "]";
814		}
815
816	<	// methods for tripping and waiting
816	>	// methods for waiting
817
818		/**
819	<	* Advance the current phase (or terminate)
819	>	* Wait nodes for Treiber stack representing wait queue
820		*/
821	<	private void trip(int phase, int parties) {
822	<	int next = onAdvance(phase, parties)? -1 : ((phase + 1) & phaseMask);
823	<	state.set(stateFor(next, parties, parties));
824	<	if (head.get() != null)
825	<	releaseWaiters(next);
826	<	}
827	<
828	<	private int helpingWait(int phase) {
829	<	final AtomicLong state = this.state;
830	<	int p;
831	<	while ((p = phaseOf(state.get())) == phase) {
832	<	ForkJoinTask<?> t = ForkJoinWorkerThread.pollTask();
833	<	if (t != null) {
834	<	if ((p = phaseOf(state.get())) == phase)
835	<	t.exec();
836	<	else {   // push task and exit if barrier advanced
837	<	t.fork();
838	<	break;
839	<	}
502	<	}
821	>	static final class QNode implements ForkJoinPool.ManagedBlocker {
822	>	final Phaser phaser;
823	>	final int phase;
824	>	final long startTime;
825	>	final long nanos;
826	>	final boolean timed;
827	>	final boolean interruptible;
828	>	volatile boolean wasInterrupted = false;
829	>	volatile Thread thread; // nulled to cancel wait
830	>	QNode next;
831	>	QNode(Phaser phaser, int phase, boolean interruptible,
832	>	boolean timed, long startTime, long nanos) {
833	>	this.phaser = phaser;
834	>	this.phase = phase;
835	>	this.timed = timed;
836	>	this.interruptible = interruptible;
837	>	this.startTime = startTime;
838	>	this.nanos = nanos;
839	>	thread = Thread.currentThread();
840		}
841	<	return p;
842	<	}
843	<
844	<	private int timedHelpingWait(int phase, long nanos) throws TimeoutException {
845	<	final AtomicLong state = this.state;
846	<	long lastTime = System.nanoTime();
847	<	int p;
848	<	while ((p = phaseOf(state.get())) == phase) {
849	<	long now = System.nanoTime();
850	<	nanos -= now - lastTime;
851	<	lastTime = now;
515	<	if (nanos <= 0) {
516	<	if ((p = phaseOf(state.get())) == phase)
517	<	throw new TimeoutException();
518	<	else
519	<	break;
841	>	public boolean isReleasable() {
842	>	return (thread == null ||
843	>	phaser.getPhase() != phase ||
844	>	(interruptible && wasInterrupted) ||
845	>	(timed && (nanos - (System.nanoTime() - startTime)) <= 0));
846	>	}
847	>	public boolean block() {
848	>	if (Thread.interrupted()) {
849	>	wasInterrupted = true;
850	>	if (interruptible)
851	>	return true;
852		}
853	<	ForkJoinTask<?> t = ForkJoinWorkerThread.pollTask();
853	>	if (!timed)
854	>	LockSupport.park(this);
855	>	else {
856	>	long waitTime = nanos - (System.nanoTime() - startTime);
857	>	if (waitTime <= 0)
858	>	return true;
859	>	LockSupport.parkNanos(this, waitTime);
860	>	}
861	>	return isReleasable();
862	>	}
863	>	void signal() {
864	>	Thread t = thread;
865		if (t != null) {
866	<	if ((p = phaseOf(state.get())) == phase)
867	<	t.exec();
868	<	else {   // push task and exit if barrier advanced
869	<	t.fork();
870	<	break;
866	>	thread = null;
867	>	LockSupport.unpark(t);
868	>	}
869	>	}
870	>	boolean doWait() {
871	>	if (thread != null) {
872	>	try {
873	>	ForkJoinPool.managedBlock(this, false);
874	>	} catch (InterruptedException ie) {
875		}
876		}
877	+	return wasInterrupted;
878		}
879	<	return p;
879	>
880		}
881
882		/**
883	<	* 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.
883	>	* Removes and signals waiting threads from wait queue.
884		*/
885	<	static final class QNode {
886	<	QNode next;
887	<	volatile Thread thread; // nulled to cancel wait
888	<	final int phase;
889	<	QNode(Thread t, int c) {
890	<	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	<	}
885	>	private void releaseWaiters(int phase) {
886	>	AtomicReference<QNode> head = queueFor(phase);
887	>	QNode q;
888	>	while ((q = head.get()) != null) {
889	>	if (head.compareAndSet(q, q.next))
890	>	q.signal();
891		}
892		}
893
565	–	/** The number of CPUs, for spin control */
566	–	static final int NCPUS = Runtime.getRuntime().availableProcessors();
567	–
894		/**
895	<	* The number of times to spin before blocking in timed waits.
896	<	* The value is empirically derived
897	<	*/
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.
895	>	* Tries to enqueue given node in the appropriate wait queue.
896	>	*
897	>	* @return true if successful
898		*/
899	<	static final int maxUntimedSpins = maxTimedSpins * 32;
899	>	private boolean tryEnqueue(QNode node) {
900	>	AtomicReference<QNode> head = queueFor(node.phase);
901	>	return head.compareAndSet(node.next = head.get(), node);
902	>	}
903
904		/**
905	<	* The number of nanoseconds for which it is faster to spin
906	<	* rather than to use timed park. A rough estimate suffices.
905	>	* Enqueues node and waits unless aborted or signalled.
906	>	*
907	>	* @return current phase
908		*/
909	<	static final long spinForTimeoutThreshold = 1000L;
909	>	private int untimedWait(int phase) {
910	>	QNode node = null;
911	>	boolean queued = false;
912	>	boolean interrupted = false;
913	>	int p;
914	>	while ((p = getPhase()) == phase) {
915	>	if (Thread.interrupted())
916	>	interrupted = true;
917	>	else if (node == null)
918	>	node = new QNode(this, phase, false, false, 0, 0);
919	>	else if (!queued)
920	>	queued = tryEnqueue(node);
921	>	else
922	>	interrupted = node.doWait();
923	>	}
924	>	if (node != null)
925	>	node.thread = null;
926	>	releaseWaiters(phase);
927	>	if (interrupted)
928	>	Thread.currentThread().interrupt();
929	>	return p;
930	>	}
931
932		/**
933	<	* Enqueues node and waits unless aborted or signalled.
933	>	* Interruptible version
934	>	* @return current phase
935		*/
936	<	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;
936	>	private int interruptibleWait(int phase) throws InterruptedException {
937		QNode node = null;
938		boolean queued = false;
939	<	int spins = maxUntimedSpins;
940	<	while (phaseOf(state.get()) == currentPhase) {
941	<	QNode h;
942	<	if (node != null && queued) {
943	<	if (node.thread != null) {
944	<	LockSupport.park();
945	<	if (Thread.interrupted()) {
946	<	wasInterrupted = true;
947	<	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);
939	>	boolean interrupted = false;
940	>	int p;
941	>	while ((p = getPhase()) == phase && !interrupted) {
942	>	if (Thread.interrupted())
943	>	interrupted = true;
944	>	else if (node == null)
945	>	node = new QNode(this, phase, true, false, 0, 0);
946	>	else if (!queued)
947	>	queued = tryEnqueue(node);
948		else
949	<	--spins;
949	>	interrupted = node.doWait();
950		}
951		if (node != null)
952		node.thread = null;
953	<	return wasInterrupted;
953	>	if (p != phase || (p = getPhase()) != phase)
954	>	releaseWaiters(phase);
955	>	if (interrupted)
956	>	throw new InterruptedException();
957	>	return p;
958		}
959
960		/**
961	<	* Messier timeout version
961	>	* Timeout version.
962	>	* @return current phase
963		*/
964	<	private void timedWait(Thread thread, int currentPhase, long nanos)
964	>	private int timedWait(int phase, long nanos)
965		throws InterruptedException, TimeoutException {
966	<	final AtomicReference<QNode> head = this.head;
641	<	final AtomicLong state = this.state;
642	<	long lastTime = System.nanoTime();
966	>	long startTime = System.nanoTime();
967		QNode node = null;
968		boolean queued = false;
969	<	int spins = maxTimedSpins;
970	<	while (phaseOf(state.get()) == currentPhase) {
971	<	QNode h;
972	<	long now = System.nanoTime();
973	<	nanos -= now - lastTime;
974	<	lastTime = now;
975	<	if (nanos <= 0) {
976	<	if (node != null)
977	<	node.thread = null;
978	<	if (phaseOf(state.get()) == currentPhase)
979	<	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);
969	>	boolean interrupted = false;
970	>	int p;
971	>	while ((p = getPhase()) == phase && !interrupted) {
972	>	if (Thread.interrupted())
973	>	interrupted = true;
974	>	else if (nanos - (System.nanoTime() - startTime) <= 0)
975	>	break;
976	>	else if (node == null)
977	>	node = new QNode(this, phase, true, true, startTime, nanos);
978	>	else if (!queued)
979	>	queued = tryEnqueue(node);
980		else
981	<	--spins;
981	>	interrupted = node.doWait();
982		}
983		if (node != null)
984		node.thread = null;
985	+	if (p != phase || (p = getPhase()) != phase)
986	+	releaseWaiters(phase);
987	+	if (interrupted)
988	+	throw new InterruptedException();
989	+	if (p == phase)
990	+	throw new TimeoutException();
991	+	return p;
992		}
993
994	<	}
994	>	// Unsafe mechanics
995
996	+	private static final sun.misc.Unsafe UNSAFE = getUnsafe();
997	+	private static final long stateOffset =
998	+	objectFieldOffset("state", Phaser.class);
999	+
1000	+	private final boolean casState(long cmp, long val) {
1001	+	return UNSAFE.compareAndSwapLong(this, stateOffset, cmp, val);
1002	+	}
1003	+
1004	+	private static long objectFieldOffset(String field, Class<?> klazz) {
1005	+	try {
1006	+	return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
1007	+	} catch (NoSuchFieldException e) {
1008	+	// Convert Exception to corresponding Error
1009	+	NoSuchFieldError error = new NoSuchFieldError(field);
1010	+	error.initCause(e);
1011	+	throw error;
1012	+	}
1013	+	}
1014	+
1015	+	/**
1016	+	* Returns a sun.misc.Unsafe.  Suitable for use in a 3rd party package.
1017	+	* Replace with a simple call to Unsafe.getUnsafe when integrating
1018	+	* into a jdk.
1019	+	*
1020	+	* @return a sun.misc.Unsafe
1021	+	*/
1022	+	private static sun.misc.Unsafe getUnsafe() {
1023	+	try {
1024	+	return sun.misc.Unsafe.getUnsafe();
1025	+	} catch (SecurityException se) {
1026	+	try {
1027	+	return java.security.AccessController.doPrivileged
1028	+	(new java.security
1029	+	.PrivilegedExceptionAction<sun.misc.Unsafe>() {
1030	+	public sun.misc.Unsafe run() throws Exception {
1031	+	java.lang.reflect.Field f = sun.misc
1032	+	.Unsafe.class.getDeclaredField("theUnsafe");
1033	+	f.setAccessible(true);
1034	+	return (sun.misc.Unsafe) f.get(null);
1035	+	}});
1036	+	} catch (java.security.PrivilegedActionException e) {
1037	+	throw new RuntimeException("Could not initialize intrinsics",
1038	+	e.getCause());
1039	+	}
1040	+	}
1041	+	}
1042	+	}

Diff Legend

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