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 1.39 by dl, Mon Aug 24 12:15:46 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}, where {@link #getArrivedParties} have
101	>	* arrived at the current phase ({@link #getPhase}). When the
102	>	* remaining {@link #getUnarrivedParties}) arrive, the phase
103	>	* advances. Method {@link #toString} returns snapshots of these state
104	>	* queries in a form convenient for informal monitoring.
105	>	*
106	>	* <p><b>Sample usages:</b>
107	>	*
108	>	* <p>A {@code Phaser} may be used instead of a {@code CountDownLatch}
109	>	* to control a one-shot action serving a variable number of
110	>	* parties. The typical idiom is for the method setting this up to
111	>	* first register, then start the actions, then deregister, as in:
112	>	*
113	>	*  <pre> {@code
114	>	* void runTasks(List<Runnable> tasks) {
115	>	*   final Phaser phaser = new Phaser(1); // "1" to register self
116	>	*   // create and start threads
117	>	*   for (Runnable task : tasks) {
118	>	*     phaser.register();
119	>	*     new Thread() {
120	>	*       public void run() {
121	>	*         phaser.arriveAndAwaitAdvance(); // await all creation
122	>	*         task.run();
123	>	*       }
124	>	*     }.start();
125	>	*   }
126	>	*
127	>	*   // allow threads to start and deregister self
128	>	*   phaser.arriveAndDeregister();
129	>	* }}</pre>
130	>	*
131	>	* <p>One way to cause a set of threads to repeatedly perform actions
132	>	* for a given number of iterations is to override {@code onAdvance}:
133	>	*
134	>	*  <pre> {@code
135	>	* void startTasks(List<Runnable> tasks, final int iterations) {
136	>	*   final Phaser phaser = new Phaser() {
137	>	*     protected boolean onAdvance(int phase, int registeredParties) {
138	>	*       return phase >= iterations || registeredParties == 0;
139	>	*     }
140	>	*   };
141	>	*   phaser.register();
142	>	*   for (Runnable task : tasks) {
143	>	*     phaser.register();
144	>	*     new Thread() {
145	>	*       public void run() {
146	>	*         do {
147	>	*           task.run();
148	>	*           phaser.arriveAndAwaitAdvance();
149	>	*         } while(!phaser.isTerminated();
150	>	*       }
151	>	*     }.start();
152	>	*   }
153	>	*   phaser.arriveAndDeregister(); // deregister self, don't wait
154	>	* }}</pre>
155		*
156	<	* <p><b>Sample usage:</b>
156	>	* If the main task must later await termination, it
157	>	* may re-register and then execute a similar loop:
158	>	* <pre> {@code
159	>	*   // ...
160	>	*   phaser.register();
161	>	*   while (!phaser.isTerminated())
162	>	*     phaser.arriveAndAwaitAdvance();
163	>	* }</pre>
164		*
165	<	* <p>[todo: non-FJ example]
165	>	* Related constructions may be used to await particular phase numbers
166	>	* in contexts where you are sure that the phase will never wrap around
167	>	* {@code Integer.MAX_VALUE}. For example:
168		*
169	<	* <p> A Phaser may be used to support a style of programming in
170	<	* which a task waits for others to complete, without otherwise
171	<	* needing to keep track of which tasks it is waiting for. This is
172	<	* similar to the "sync" construct in Cilk and "clocks" in X10.
173	<	* Special constructions based on such barriers are available using
174	<	* the <tt>LinkedAsyncAction</tt> and <tt>CyclicAction</tt> classes,
175	<	* but they can be useful in other contexts as well.  For a simple
176	<	* (but not very useful) example, here is a variant of Fibonacci:
177	<	*
178	<	* <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();
169	>	* <pre> {@code
170	>	*   void awaitPhase(Phaser phaser, int phase) {
171	>	*     int p = phaser.register(); // assumes caller not already registered
172	>	*     while (p < phase) {
173	>	*       if (phaser.isTerminated())
174	>	*         // ... deal with unexpected termination
175	>	*       else
176	>	*         p = phaser.arriveAndAwaitAdvance();
177	>	*     }
178	>	*     phaser.arriveAndDeregister();
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	>	* }</pre>
181	>	*
182	>	*
183	>	* <p>To create a set of tasks using a tree of phasers,
184	>	* you could use code of the following form, assuming a
185	>	* Task class with a constructor accepting a phaser that
186	>	* it registers for upon construction:
187	>	*  <pre> {@code
188	>	* void build(Task[] actions, int lo, int hi, Phaser b) {
189	>	*   int step = (hi - lo) / TASKS_PER_PHASER;
190	>	*   if (step > 1) {
191	>	*     int i = lo;
192	>	*     while (i < hi) {
193	>	*       int r = Math.min(i + step, hi);
194	>	*       build(actions, i, r, new Phaser(b));
195	>	*       i = r;
196		*     }
197	<	*     parentBarrier.arriveAndDeregister();
197	>	*   } else {
198	>	*     for (int i = lo; i < hi; ++i)
199	>	*       actions[i] = new Task(b);
200	>	*       // assumes new Task(b) performs b.register()
201		*   }
202		* }
203	+	* // .. initially called, for n tasks via
204	+	* build(new Task[n], 0, n, new Phaser());}</pre>
205	+	*
206	+	* The best value of {@code TASKS_PER_PHASER} depends mainly on
207	+	* expected barrier synchronization rates. A value as low as four may
208	+	* be appropriate for extremely small per-barrier task bodies (thus
209	+	* high rates), or up to hundreds for extremely large ones.
210	+	*
211		* </pre>
212		*
213		* <p><b>Implementation notes</b>: This implementation restricts the
214	<	* maximum number of parties to 65535. Attempts to register
215	<	* additional parties result in IllegalStateExceptions.
214	>	* maximum number of parties to 65535. Attempts to register additional
215	>	* parties result in {@code IllegalStateException}. However, you can and
216	>	* should create tiered phasers to accommodate arbitrarily large sets
217	>	* of participants.
218	>	*
219	>	* @since 1.7
220	>	* @author Doug Lea
221		*/
222		public class Phaser {
223		/*
224		* This class implements an extension of X10 "clocks".  Thanks to
225	<	* Vijay Saraswat for the idea of applying it to ForkJoinTasks,
226	<	* and to Vivek Sarkar for enhancements to extend functionality.
225	>	* Vijay Saraswat for the idea, and to Vivek Sarkar for
226	>	* enhancements to extend functionality.
227		*/
228
229		/**
230		* Barrier state representation. Conceptually, a barrier contains
231		* four values:
232	<	*
232	>	*
233		* * parties -- the number of parties to wait (16 bits)
234		* * unarrived -- the number of parties yet to hit barrier (16 bits)
235		* * phase -- the generation of the barrier (31 bits)
236		* * terminated -- set if barrier is terminated (1 bit)
237		*
238		* However, to efficiently maintain atomicity, these values are
239	<	* packed into a single AtomicLong. Termination uses the sign bit
240	<	* of 32 bit representation of phase, so phase is set to -1 on
241	<	* termination.
242	<	*/
243	<	private final AtomicLong state;
244	<
245	<	/**
143	<	* Head of Treiber stack for waiting nonFJ threads.
239	>	* packed into a single (atomic) long. Termination uses the sign
240	>	* bit of 32 bit representation of phase, so phase is set to -1 on
241	>	* termination. Good performance relies on keeping state decoding
242	>	* and encoding simple, and keeping race windows short.
243	>	*
244	>	* Note: there are some cheats in arrive() that rely on unarrived
245	>	* count being lowest 16 bits.
246		*/
247	<	private final AtomicReference<QNode> head = new AtomicReference<QNode>();
247	>	private volatile long state;
248
249		private static final int ushortBits = 16;
250	<	private static final int ushortMask =  (1 << ushortBits) - 1;
251	<	private static final int phaseMask = 0x7fffffff;
250	>	private static final int ushortMask = 0xffff;
251	>	private static final int phaseMask  = 0x7fffffff;
252
253		private static int unarrivedOf(long s) {
254	<	return (int)(s & ushortMask);
254	>	return (int) (s & ushortMask);
255		}
256
257		private static int partiesOf(long s) {
258	<	return (int)(s & (ushortMask << 16)) >>> 16;
258	>	return ((int) s) >>> 16;
259		}
260
261		private static int phaseOf(long s) {
262	<	return (int)(s >>> 32);
262	>	return (int) (s >>> 32);
263		}
264
265		private static int arrivedOf(long s) {
#	Line 165 | Line 267 | public class Phaser {
267		}
268
269		private static long stateFor(int phase, int parties, int unarrived) {
270	<	return (((long)phase) << 32) | ((parties << 16) | unarrived);
270	>	return ((((long) phase) << 32) | (((long) parties) << 16) |
271	>	(long) unarrived);
272		}
273
274	<	private static IllegalStateException badBounds(int parties, int unarrived) {
275	<	return new IllegalStateException("Attempt to set " + unarrived +
276	<	" unarrived of " + parties + " parties");
274	>	private static long trippedStateFor(int phase, int parties) {
275	>	long lp = (long) parties;
276	>	return (((long) phase) << 32) | (lp << 16) | lp;
277		}
278
279		/**
280	<	* Creates a new Phaser without any initially registered parties,
281	<	* and initial phase number 0.
280	>	* Returns message string for bad bounds exceptions.
281	>	*/
282	>	private static String badBounds(int parties, int unarrived) {
283	>	return ("Attempt to set " + unarrived +
284	>	" unarrived of " + parties + " parties");
285	>	}
286	>
287	>	/**
288	>	* The parent of this phaser, or null if none
289	>	*/
290	>	private final Phaser parent;
291	>
292	>	/**
293	>	* The root of phaser tree. Equals this if not in a tree.  Used to
294	>	* support faster state push-down.
295	>	*/
296	>	private final Phaser root;
297	>
298	>	// Wait queues
299	>
300	>	/**
301	>	* Heads of Treiber stacks for waiting threads. To eliminate
302	>	* contention while releasing some threads while adding others, we
303	>	* use two of them, alternating across even and odd phases.
304	>	*/
305	>	private final AtomicReference<QNode> evenQ = new AtomicReference<QNode>();
306	>	private final AtomicReference<QNode> oddQ  = new AtomicReference<QNode>();
307	>
308	>	private AtomicReference<QNode> queueFor(int phase) {
309	>	return ((phase & 1) == 0) ? evenQ : oddQ;
310	>	}
311	>
312	>	/**
313	>	* Returns current state, first resolving lagged propagation from
314	>	* root if necessary.
315	>	*/
316	>	private long getReconciledState() {
317	>	return (parent == null) ? state : reconcileState();
318	>	}
319	>
320	>	/**
321	>	* Recursively resolves state.
322	>	*/
323	>	private long reconcileState() {
324	>	Phaser p = parent;
325	>	long s = state;
326	>	if (p != null) {
327	>	while (unarrivedOf(s) == 0 && phaseOf(s) != phaseOf(root.state)) {
328	>	long parentState = p.getReconciledState();
329	>	int parentPhase = phaseOf(parentState);
330	>	int phase = phaseOf(s = state);
331	>	if (phase != parentPhase) {
332	>	long next = trippedStateFor(parentPhase, partiesOf(s));
333	>	if (casState(s, next)) {
334	>	releaseWaiters(phase);
335	>	s = next;
336	>	}
337	>	}
338	>	}
339	>	}
340	>	return s;
341	>	}
342	>
343	>	/**
344	>	* Creates a new phaser without any initially registered parties,
345	>	* initial phase number 0, and no parent. Any thread using this
346	>	* phaser will need to first register for it.
347		*/
348		public Phaser() {
349	<	state = new AtomicLong(stateFor(0, 0, 0));
349	>	this(null);
350		}
351
352		/**
353	<	* Creates a new Phaser with the given numbers of registered
354	<	* unarrived parties and initial phase number 0.
355	<	* @param parties the number of parties required to trip barrier.
353	>	* Creates a new phaser with the given numbers of registered
354	>	* unarrived parties, initial phase number 0, and no parent.
355	>	*
356	>	* @param parties the number of parties required to trip barrier
357		* @throws IllegalArgumentException if parties less than zero
358	<	* or greater than the maximum number of parties supported.
358	>	* or greater than the maximum number of parties supported
359		*/
360		public Phaser(int parties) {
361	+	this(null, parties);
362	+	}
363	+
364	+	/**
365	+	* Creates a new phaser with the given parent, without any
366	+	* initially registered parties. If parent is non-null this phaser
367	+	* is registered with the parent and its initial phase number is
368	+	* the same as that of parent phaser.
369	+	*
370	+	* @param parent the parent phaser
371	+	*/
372	+	public Phaser(Phaser parent) {
373	+	int phase = 0;
374	+	this.parent = parent;
375	+	if (parent != null) {
376	+	this.root = parent.root;
377	+	phase = parent.register();
378	+	}
379	+	else
380	+	this.root = this;
381	+	this.state = trippedStateFor(phase, 0);
382	+	}
383	+
384	+	/**
385	+	* Creates a new phaser with the given parent and numbers of
386	+	* registered unarrived parties. If parent is non-null, this phaser
387	+	* is registered with the parent and its initial phase number is
388	+	* the same as that of parent phaser.
389	+	*
390	+	* @param parent the parent phaser
391	+	* @param parties the number of parties required to trip barrier
392	+	* @throws IllegalArgumentException if parties less than zero
393	+	* or greater than the maximum number of parties supported
394	+	*/
395	+	public Phaser(Phaser parent, int parties) {
396		if (parties < 0 || parties > ushortMask)
397		throw new IllegalArgumentException("Illegal number of parties");
398	<	state = new AtomicLong(stateFor(0, parties, parties));
398	>	int phase = 0;
399	>	this.parent = parent;
400	>	if (parent != null) {
401	>	this.root = parent.root;
402	>	phase = parent.register();
403	>	}
404	>	else
405	>	this.root = this;
406	>	this.state = trippedStateFor(phase, parties);
407		}
408
409		/**
410		* Adds a new unarrived party to this phaser.
411	<	* @return the current barrier phase number upon registration
411	>	*
412	>	* @return the arrival phase number to which this registration applied
413	>	* @throws IllegalStateException if attempting to register more
414	>	* than the maximum supported number of parties
415	>	*/
416	>	public int register() {
417	>	return doRegister(1);
418	>	}
419	>
420	>	/**
421	>	* Adds the given number of new unarrived parties to this phaser.
422	>	*
423	>	* @param parties the number of parties required to trip barrier
424	>	* @return the arrival phase number to which this registration applied
425		* @throws IllegalStateException if attempting to register more
426	<	* than the maximum supported number of parties.
426	>	* than the maximum supported number of parties
427		*/
428	<	public int register() { // increment both parties and unarrived
429	<	final AtomicLong state = this.state;
428	>	public int bulkRegister(int parties) {
429	>	if (parties < 0)
430	>	throw new IllegalArgumentException();
431	>	if (parties == 0)
432	>	return getPhase();
433	>	return doRegister(parties);
434	>	}
435	>
436	>	/**
437	>	* Shared code for register, bulkRegister
438	>	*/
439	>	private int doRegister(int registrations) {
440	>	int phase;
441		for (;;) {
442	<	long s = state.get();
443	<	int phase = phaseOf(s);
444	<	int parties = partiesOf(s) + 1;
445	<	int unarrived = unarrivedOf(s) + 1;
442	>	long s = getReconciledState();
443	>	phase = phaseOf(s);
444	>	int unarrived = unarrivedOf(s) + registrations;
445	>	int parties = partiesOf(s) + registrations;
446	>	if (phase < 0)
447	>	break;
448		if (parties > ushortMask || unarrived > ushortMask)
449	<	throw badBounds(parties, unarrived);
450	<	if (state.compareAndSet(s, stateFor(phase, parties, unarrived)))
451	<	return phase;
449	>	throw new IllegalStateException(badBounds(parties, unarrived));
450	>	if (phase == phaseOf(root.state) &&
451	>	casState(s, stateFor(phase, parties, unarrived)))
452	>	break;
453		}
454	+	return phase;
455		}
456
457		/**
458		* Arrives at the barrier, but does not wait for others.  (You can
459	<	* in turn wait for others via {@link #awaitAdvance}).
459	>	* in turn wait for others via {@link #awaitAdvance}).  It is an
460	>	* unenforced usage error for an unregistered party to invoke this
461	>	* method.
462		*
463	<	* @return the current barrier phase number upon entry to
464	<	* this method, or a negative value if terminated;
465	<	* @throws IllegalStateException if the number of unarrived
224	<	* parties would become negative.
463	>	* @return the arrival phase number, or a negative value if terminated
464	>	* @throws IllegalStateException if not terminated and the number
465	>	* of unarrived parties would become negative
466		*/
467	<	public int arrive() { // decrement unarrived. If zero, trip
468	<	final AtomicLong state = this.state;
467	>	public int arrive() {
468	>	int phase;
469		for (;;) {
470	<	long s = state.get();
471	<	int phase = phaseOf(s);
470	>	long s = state;
471	>	phase = phaseOf(s);
472	>	if (phase < 0)
473	>	break;
474		int parties = partiesOf(s);
475		int unarrived = unarrivedOf(s) - 1;
476	<	if (unarrived < 0)
477	<	throw badBounds(parties, unarrived);
478	<	if (unarrived == 0 && phase >= 0) {
236	<	trip(phase, parties);
237	<	return phase;
476	>	if (unarrived > 0) {        // Not the last arrival
477	>	if (casState(s, s - 1)) // s-1 adds one arrival
478	>	break;
479		}
480	<	if (state.compareAndSet(s, stateFor(phase, parties, unarrived)))
481	<	return phase;
480	>	else if (unarrived == 0) {  // the last arrival
481	>	Phaser par = parent;
482	>	if (par == null) {      // directly trip
483	>	if (casState
484	>	(s,
485	>	trippedStateFor(onAdvance(phase, parties) ? -1 :
486	>	((phase + 1) & phaseMask), parties))) {
487	>	releaseWaiters(phase);
488	>	break;
489	>	}
490	>	}
491	>	else {                  // cascade to parent
492	>	if (casState(s, s - 1)) { // zeroes unarrived
493	>	par.arrive();
494	>	reconcileState();
495	>	break;
496	>	}
497	>	}
498	>	}
499	>	else if (phase != phaseOf(root.state)) // or if unreconciled
500	>	reconcileState();
501	>	else
502	>	throw new IllegalStateException(badBounds(parties, unarrived));
503		}
504	+	return phase;
505		}
506
507		/**
508	<	* Arrives at the barrier, and deregisters from it, without
509	<	* waiting for others.
508	>	* Arrives at the barrier and deregisters from it without waiting
509	>	* for others. Deregistration reduces the number of parties
510	>	* required to trip the barrier in future phases.  If this phaser
511	>	* has a parent, and deregistration causes this phaser to have
512	>	* zero parties, this phaser also arrives at and is deregistered
513	>	* from its parent.  It is an unenforced usage error for an
514	>	* unregistered party to invoke this method.
515		*
516	<	* @return the current barrier phase number upon entry to
517	<	* this method, or a negative value if terminated;
518	<	* @throws IllegalStateException if the number of registered or
519	<	* unarrived parties would become negative.
520	<	*/
521	<	public int arriveAndDeregister() { // Same as arrive, plus decrement parties
522	<	final AtomicLong state = this.state;
516	>	* @return the arrival phase number, or a negative value if terminated
517	>	* @throws IllegalStateException if not terminated and the number
518	>	* of registered or unarrived parties would become negative
519	>	*/
520	>	public int arriveAndDeregister() {
521	>	// similar code to arrive, but too different to merge
522	>	Phaser par = parent;
523	>	int phase;
524		for (;;) {
525	<	long s = state.get();
526	<	int phase = phaseOf(s);
525	>	long s = state;
526	>	phase = phaseOf(s);
527	>	if (phase < 0)
528	>	break;
529		int parties = partiesOf(s) - 1;
530		int unarrived = unarrivedOf(s) - 1;
531	<	if (parties < 0 || unarrived < 0)
532	<	throw badBounds(parties, unarrived);
533	<	if (unarrived == 0 && phase >= 0) {
534	<	trip(phase, parties);
535	<	return phase;
531	>	if (parties >= 0) {
532	>	if (unarrived > 0 || (unarrived == 0 && par != null)) {
533	>	if (casState
534	>	(s,
535	>	stateFor(phase, parties, unarrived))) {
536	>	if (unarrived == 0) {
537	>	par.arriveAndDeregister();
538	>	reconcileState();
539	>	}
540	>	break;
541	>	}
542	>	continue;
543	>	}
544	>	if (unarrived == 0) {
545	>	if (casState
546	>	(s,
547	>	trippedStateFor(onAdvance(phase, parties) ? -1 :
548	>	((phase + 1) & phaseMask), parties))) {
549	>	releaseWaiters(phase);
550	>	break;
551	>	}
552	>	continue;
553	>	}
554	>	if (par != null && phase != phaseOf(root.state)) {
555	>	reconcileState();
556	>	continue;
557	>	}
558		}
559	<	if (state.compareAndSet(s, stateFor(phase, parties, unarrived)))
267	<	return phase;
559	>	throw new IllegalStateException(badBounds(parties, unarrived));
560		}
561	+	return phase;
562		}
563
564		/**
565	<	* Arrives at the barrier and awaits others. Unlike other arrival
566	<	* methods, this method returns the arrival index of the
567	<	* caller. The caller tripping the barrier returns zero, the
568	<	* previous caller 1, and so on.
569	<	* @return the arrival index
570	<	* @throws IllegalStateException if the number of unarrived
571	<	* parties would become negative.
565	>	* Arrives at the barrier and awaits others. Equivalent in effect
566	>	* to {@code awaitAdvance(arrive())}.  If you need to await with
567	>	* interruption or timeout, you can arrange this with an analogous
568	>	* construction using one of the other forms of the awaitAdvance
569	>	* method.  If instead you need to deregister upon arrival use
570	>	* {@code arriveAndDeregister}. It is an unenforced usage error
571	>	* for an unregistered party to invoke this method.
572	>	*
573	>	* @return the arrival phase number, or a negative number if terminated
574	>	* @throws IllegalStateException if not terminated and the number
575	>	* of unarrived parties would become negative
576		*/
577		public int arriveAndAwaitAdvance() {
578	<	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	<	}
578	>	return awaitAdvance(arrive());
579		}
580
581		/**
582	<	* Awaits the phase of the barrier to advance from the given
583	<	* value, or returns immediately if this barrier is terminated
584	<	* @param phase the phase on entry to this method
585	<	* @return the phase on exit from this method
582	>	* Awaits the phase of the barrier to advance from the given phase
583	>	* value, returning immediately if the current phase of the
584	>	* barrier is not equal to the given phase value or this barrier
585	>	* is terminated.  It is an unenforced usage error for an
586	>	* unregistered party to invoke this method.
587	>	*
588	>	* @param phase an arrival phase number, or negative value if
589	>	* terminated; this argument is normally the value returned by a
590	>	* previous call to {@code arrive} or its variants
591	>	* @return the next arrival phase number, or a negative value
592	>	* if terminated or argument is negative
593		*/
594		public int awaitAdvance(int phase) {
595		if (phase < 0)
596		return phase;
597	<	Thread current = Thread.currentThread();
598	<	if (current instanceof ForkJoinWorkerThread)
599	<	return helpingWait(phase);
600	<	if (untimedWait(current, phase, false))
601	<	current.interrupt();
602	<	return phaseOf(state.get());
597	>	long s = getReconciledState();
598	>	int p = phaseOf(s);
599	>	if (p != phase)
600	>	return p;
601	>	if (unarrivedOf(s) == 0 && parent != null)
602	>	parent.awaitAdvance(phase);
603	>	// Fall here even if parent waited, to reconcile and help release
604	>	return untimedWait(phase);
605		}
606
607		/**
608	<	* Awaits the phase of the barrier to advance from the given
609	<	* value, or returns immediately if this barrier is terminated, or
610	<	* throws InterruptedException if interrupted while waiting.
611	<	* @param phase the phase on entry to this method
612	<	* @return the phase on exit from this method
608	>	* Awaits the phase of the barrier to advance from the given phase
609	>	* value, throwing {@code InterruptedException} if interrupted
610	>	* while waiting, or returning immediately if the current phase of
611	>	* the barrier is not equal to the given phase value or this
612	>	* barrier is terminated. It is an unenforced usage error for an
613	>	* unregistered party to invoke this method.
614	>	*
615	>	* @param phase an arrival phase number, or negative value if
616	>	* terminated; this argument is normally the value returned by a
617	>	* previous call to {@code arrive} or its variants
618	>	* @return the next arrival phase number, or a negative value
619	>	* if terminated or argument is negative
620		* @throws InterruptedException if thread interrupted while waiting
621		*/
622	<	public int awaitAdvanceInterruptibly(int phase) throws InterruptedException {
622	>	public int awaitAdvanceInterruptibly(int phase)
623	>	throws InterruptedException {
624		if (phase < 0)
625		return phase;
626	<	Thread current = Thread.currentThread();
627	<	if (current instanceof ForkJoinWorkerThread)
628	<	return helpingWait(phase);
629	<	else if (Thread.interrupted() || untimedWait(current, phase, true))
630	<	throw new InterruptedException();
631	<	else
632	<	return phaseOf(state.get());
626	>	long s = getReconciledState();
627	>	int p = phaseOf(s);
628	>	if (p != phase)
629	>	return p;
630	>	if (unarrivedOf(s) == 0 && parent != null)
631	>	parent.awaitAdvanceInterruptibly(phase);
632	>	return interruptibleWait(phase);
633		}
634
635		/**
636	<	* Awaits the phase of the barrier to advance from the given value
637	<	* or the given timeout elapses, or returns immediately if this
638	<	* barrier is terminated
639	<	* @param phase the phase on entry to this method
640	<	* @return the phase on exit from this method
636	>	* Awaits the phase of the barrier to advance from the given phase
637	>	* value or the given timeout to elapse, throwing {@code
638	>	* InterruptedException} if interrupted while waiting, or
639	>	* returning immediately if the current phase of the barrier is
640	>	* not equal to the given phase value or this barrier is
641	>	* terminated.  It is an unenforced usage error for an
642	>	* unregistered party to invoke this method.
643	>	*
644	>	* @param phase an arrival phase number, or negative value if
645	>	* terminated; this argument is normally the value returned by a
646	>	* previous call to {@code arrive} or its variants
647	>	* @param timeout how long to wait before giving up, in units of
648	>	*        {@code unit}
649	>	* @param unit a {@code TimeUnit} determining how to interpret the
650	>	*        {@code timeout} parameter
651	>	* @return the next arrival phase number, or a negative value
652	>	* if terminated or argument is negative
653		* @throws InterruptedException if thread interrupted while waiting
654		* @throws TimeoutException if timed out while waiting
655		*/
656	<	public int awaitAdvanceInterruptibly(int phase, long timeout, TimeUnit unit)
656	>	public int awaitAdvanceInterruptibly(int phase,
657	>	long timeout, TimeUnit unit)
658		throws InterruptedException, TimeoutException {
659		if (phase < 0)
660		return phase;
661	<	long nanos = unit.toNanos(timeout);
662	<	Thread current = Thread.currentThread();
663	<	if (current instanceof ForkJoinWorkerThread)
664	<	return timedHelpingWait(phase, nanos);
665	<	timedWait(current, phase, nanos);
666	<	return phaseOf(state.get());
661	>	long s = getReconciledState();
662	>	int p = phaseOf(s);
663	>	if (p != phase)
664	>	return p;
665	>	if (unarrivedOf(s) == 0 && parent != null)
666	>	parent.awaitAdvanceInterruptibly(phase, timeout, unit);
667	>	return timedWait(phase, unit.toNanos(timeout));
668		}
669
670		/**
671		* Forces this barrier to enter termination state. Counts of
672	<	* arrived and registered parties are unaffected. This method may
673	<	* be useful for coordinating recovery after one or more tasks
674	<	* encounter unexpected exceptions.
672	>	* arrived and registered parties are unaffected. If this phaser
673	>	* has a parent, it too is terminated. This method may be useful
674	>	* for coordinating recovery after one or more tasks encounter
675	>	* unexpected exceptions.
676		*/
677		public void forceTermination() {
365	–	final AtomicLong state = this.state;
678		for (;;) {
679	<	long s = state.get();
679	>	long s = getReconciledState();
680		int phase = phaseOf(s);
681		int parties = partiesOf(s);
682		int unarrived = unarrivedOf(s);
683		if (phase < 0 ||
684	<	state.compareAndSet(s, stateFor(-1, parties, unarrived))) {
685	<	if (head.get() != null)
686	<	releaseWaiters(-1);
684	>	casState(s, stateFor(-1, parties, unarrived))) {
685	>	releaseWaiters(0);
686	>	releaseWaiters(1);
687	>	if (parent != null)
688	>	parent.forceTermination();
689		return;
690		}
691		}
692		}
693
694		/**
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	–	/**
695		* Returns the current phase number. The maximum phase number is
696	<	* <tt>Integer.MAX_VALUE</tt>, after which it restarts at
696	>	* {@code Integer.MAX_VALUE}, after which it restarts at
697		* zero. Upon termination, the phase number is negative.
698	+	*
699		* @return the phase number, or a negative value if terminated
700		*/
701	<	public int getPhase() {
702	<	return phaseOf(state.get());
701	>	public final int getPhase() {
702	>	return phaseOf(getReconciledState());
703		}
704
705		/**
706		* Returns the number of parties registered at this barrier.
707	+	*
708		* @return the number of parties
709		*/
710		public int getRegisteredParties() {
711	<	return partiesOf(state.get());
711	>	return partiesOf(state);
712		}
713
714		/**
715	<	* Returns the number of parties that have arrived at the current
716	<	* phase of this barrier.
715	>	* Returns the number of registered parties that have arrived at
716	>	* the current phase of this barrier.
717	>	*
718		* @return the number of arrived parties
719		*/
720		public int getArrivedParties() {
721	<	return arrivedOf(state.get());
721	>	return arrivedOf(state);
722		}
723
724		/**
725		* Returns the number of registered parties that have not yet
726		* arrived at the current phase of this barrier.
727	+	*
728		* @return the number of unarrived parties
729		*/
730		public int getUnarrivedParties() {
731	<	return unarrivedOf(state.get());
731	>	return unarrivedOf(state);
732	>	}
733	>
734	>	/**
735	>	* Returns the parent of this phaser, or {@code null} if none.
736	>	*
737	>	* @return the parent of this phaser, or {@code null} if none
738	>	*/
739	>	public Phaser getParent() {
740	>	return parent;
741	>	}
742	>
743	>	/**
744	>	* Returns the root ancestor of this phaser, which is the same as
745	>	* this phaser if it has no parent.
746	>	*
747	>	* @return the root ancestor of this phaser
748	>	*/
749	>	public Phaser getRoot() {
750	>	return root;
751		}
752
753		/**
754	<	* Returns true if this barrier has been terminated
755	<	* @return true if this barrier has been terminated
754	>	* Returns {@code true} if this barrier has been terminated.
755	>	*
756	>	* @return {@code true} if this barrier has been terminated
757		*/
758		public boolean isTerminated() {
759	<	return phaseOf(state.get()) < 0;
759	>	return getPhase() < 0;
760		}
761
762		/**
763		* Overridable method to perform an action upon phase advance, and
764		* to control termination. This method is invoked whenever the
765		* barrier is tripped (and thus all other waiting parties are
766	<	* dormant). If it returns true, then, rather than advance the
767	<	* phase number, this barrier will be set to a final termination
768	<	* state, and subsequent calls to <tt>isTerminated</tt> will
769	<	* return true.
770	<	*
771	<	* <p> The default version returns true when the number of
766	>	* dormant). If it returns {@code true}, then, rather than advance
767	>	* the phase number, this barrier will be set to a final
768	>	* termination state, and subsequent calls to {@link #isTerminated}
769	>	* will return true.
770	>	*
771	>	* <p>The default version returns {@code true} when the number of
772		* registered parties is zero. Normally, overrides that arrange
773		* termination for other reasons should also preserve this
774		* property.
775		*
776	+	* <p>You may override this method to perform an action with side
777	+	* effects visible to participating tasks, but it is in general
778	+	* only sensible to do so in designs where all parties register
779	+	* before any arrive, and all {@link #awaitAdvance} at each phase.
780	+	* Otherwise, you cannot ensure lack of interference from other
781	+	* parties during the invocation of this method.
782	+	*
783		* @param phase the phase number on entering the barrier
784	<	* @param registeredParties the current number of registered
785	<	* parties.
458	<	* @return true if this barrier should terminate
784	>	* @param registeredParties the current number of registered parties
785	>	* @return {@code true} if this barrier should terminate
786		*/
787		protected boolean onAdvance(int phase, int registeredParties) {
788		return registeredParties <= 0;
789		}
790
791		/**
792	<	* Returns a string identifying this barrier, as well as its
792	>	* Returns a string identifying this phaser, as well as its
793		* state.  The state, in brackets, includes the String {@code
794	<	* "phase ="} followed by the phase number, {@code "parties ="}
794	>	* "phase = "} followed by the phase number, {@code "parties = "}
795		* followed by the number of registered parties, and {@code
796	<	* "arrived ="} followed by the number of arrived parties
796	>	* "arrived = "} followed by the number of arrived parties.
797		*
798		* @return a string identifying this barrier, as well as its state
799		*/
800		public String toString() {
801	<	long s = state.get();
802	<	return super.toString() + "[phase = " + phaseOf(s) + " parties = " + partiesOf(s) + " arrived = " + arrivedOf(s) + "]";
801	>	long s = getReconciledState();
802	>	return super.toString() +
803	>	"[phase = " + phaseOf(s) +
804	>	" parties = " + partiesOf(s) +
805	>	" arrived = " + arrivedOf(s) + "]";
806		}
807
808	<	// methods for tripping and waiting
808	>	// methods for waiting
809
810		/**
811	<	* Advance the current phase (or terminate)
811	>	* Wait nodes for Treiber stack representing wait queue
812		*/
813	<	private void trip(int phase, int parties) {
814	<	int next = onAdvance(phase, parties)? -1 : ((phase + 1) & phaseMask);
815	<	state.set(stateFor(next, parties, parties));
816	<	if (head.get() != null)
817	<	releaseWaiters(next);
818	<	}
819	<
820	<	private int helpingWait(int phase) {
821	<	final AtomicLong state = this.state;
822	<	int p;
823	<	while ((p = phaseOf(state.get())) == phase) {
824	<	ForkJoinTask<?> t = ForkJoinWorkerThread.pollTask();
825	<	if (t != null) {
826	<	if ((p = phaseOf(state.get())) == phase)
827	<	t.exec();
828	<	else {   // push task and exit if barrier advanced
829	<	t.fork();
830	<	break;
831	<	}
502	<	}
813	>	static final class QNode implements ForkJoinPool.ManagedBlocker {
814	>	final Phaser phaser;
815	>	final int phase;
816	>	final long startTime;
817	>	final long nanos;
818	>	final boolean timed;
819	>	final boolean interruptible;
820	>	volatile boolean wasInterrupted = false;
821	>	volatile Thread thread; // nulled to cancel wait
822	>	QNode next;
823	>	QNode(Phaser phaser, int phase, boolean interruptible,
824	>	boolean timed, long startTime, long nanos) {
825	>	this.phaser = phaser;
826	>	this.phase = phase;
827	>	this.timed = timed;
828	>	this.interruptible = interruptible;
829	>	this.startTime = startTime;
830	>	this.nanos = nanos;
831	>	thread = Thread.currentThread();
832		}
833	<	return p;
834	<	}
835	<
836	<	private int timedHelpingWait(int phase, long nanos) throws TimeoutException {
837	<	final AtomicLong state = this.state;
838	<	long lastTime = System.nanoTime();
839	<	int p;
840	<	while ((p = phaseOf(state.get())) == phase) {
841	<	long now = System.nanoTime();
842	<	nanos -= now - lastTime;
843	<	lastTime = now;
844	<	if (nanos <= 0) {
845	<	if ((p = phaseOf(state.get())) == phase)
846	<	throw new TimeoutException();
847	<	else
848	<	break;
833	>	public boolean isReleasable() {
834	>	return (thread == null ||
835	>	phaser.getPhase() != phase ||
836	>	(interruptible && wasInterrupted) ||
837	>	(timed && (nanos - (System.nanoTime() - startTime)) <= 0));
838	>	}
839	>	public boolean block() {
840	>	if (Thread.interrupted()) {
841	>	wasInterrupted = true;
842	>	if (interruptible)
843	>	return true;
844	>	}
845	>	if (!timed)
846	>	LockSupport.park(this);
847	>	else {
848	>	long waitTime = nanos - (System.nanoTime() - startTime);
849	>	if (waitTime <= 0)
850	>	return true;
851	>	LockSupport.parkNanos(this, waitTime);
852		}
853	<	ForkJoinTask<?> t = ForkJoinWorkerThread.pollTask();
853	>	return isReleasable();
854	>	}
855	>	void signal() {
856	>	Thread t = thread;
857		if (t != null) {
858	<	if ((p = phaseOf(state.get())) == phase)
859	<	t.exec();
860	<	else {   // push task and exit if barrier advanced
861	<	t.fork();
862	<	break;
858	>	thread = null;
859	>	LockSupport.unpark(t);
860	>	}
861	>	}
862	>	boolean doWait() {
863	>	if (thread != null) {
864	>	try {
865	>	ForkJoinPool.managedBlock(this, false);
866	>	} catch (InterruptedException ie) {
867		}
868		}
869	+	return wasInterrupted;
870		}
871	<	return p;
871	>
872		}
873
874		/**
875	<	* 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.
875	>	* Removes and signals waiting threads from wait queue.
876		*/
877	<	static final class QNode {
878	<	QNode next;
879	<	volatile Thread thread; // nulled to cancel wait
880	<	final int phase;
881	<	QNode(Thread t, int c) {
882	<	thread = t;
545	<	phase = c;
877	>	private void releaseWaiters(int phase) {
878	>	AtomicReference<QNode> head = queueFor(phase);
879	>	QNode q;
880	>	while ((q = head.get()) != null) {
881	>	if (head.compareAndSet(q, q.next))
882	>	q.signal();
883		}
884		}
885
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	–	}
562	–	}
563	–	}
564	–
565	–	/** The number of CPUs, for spin control */
566	–	static final int NCPUS = Runtime.getRuntime().availableProcessors();
567	–
886		/**
887	<	* The number of times to spin before blocking in timed waits.
888	<	* The value is empirically derived
889	<	*/
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.
887	>	* Tries to enqueue given node in the appropriate wait queue.
888	>	*
889	>	* @return true if successful
890		*/
891	<	static final int maxUntimedSpins = maxTimedSpins * 32;
891	>	private boolean tryEnqueue(QNode node) {
892	>	AtomicReference<QNode> head = queueFor(node.phase);
893	>	return head.compareAndSet(node.next = head.get(), node);
894	>	}
895
896		/**
897	<	* The number of nanoseconds for which it is faster to spin
898	<	* rather than to use timed park. A rough estimate suffices.
897	>	* Enqueues node and waits unless aborted or signalled.
898	>	*
899	>	* @return current phase
900		*/
901	<	static final long spinForTimeoutThreshold = 1000L;
901	>	private int untimedWait(int phase) {
902	>	QNode node = null;
903	>	boolean queued = false;
904	>	boolean interrupted = false;
905	>	int p;
906	>	while ((p = getPhase()) == phase) {
907	>	if (Thread.interrupted())
908	>	interrupted = true;
909	>	else if (node == null)
910	>	node = new QNode(this, phase, false, false, 0, 0);
911	>	else if (!queued)
912	>	queued = tryEnqueue(node);
913	>	else
914	>	interrupted = node.doWait();
915	>	}
916	>	if (node != null)
917	>	node.thread = null;
918	>	releaseWaiters(phase);
919	>	if (interrupted)
920	>	Thread.currentThread().interrupt();
921	>	return p;
922	>	}
923
924		/**
925	<	* Enqueues node and waits unless aborted or signalled.
925	>	* Interruptible version
926	>	* @return current phase
927		*/
928	<	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;
928	>	private int interruptibleWait(int phase) throws InterruptedException {
929		QNode node = null;
930		boolean queued = false;
931	<	int spins = maxUntimedSpins;
932	<	while (phaseOf(state.get()) == currentPhase) {
933	<	QNode h;
934	<	if (node != null && queued) {
935	<	if (node.thread != null) {
936	<	LockSupport.park();
937	<	if (Thread.interrupted()) {
938	<	wasInterrupted = true;
939	<	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);
931	>	boolean interrupted = false;
932	>	int p;
933	>	while ((p = getPhase()) == phase && !interrupted) {
934	>	if (Thread.interrupted())
935	>	interrupted = true;
936	>	else if (node == null)
937	>	node = new QNode(this, phase, true, false, 0, 0);
938	>	else if (!queued)
939	>	queued = tryEnqueue(node);
940		else
941	<	--spins;
941	>	interrupted = node.doWait();
942		}
943		if (node != null)
944		node.thread = null;
945	<	return wasInterrupted;
945	>	if (p != phase || (p = getPhase()) != phase)
946	>	releaseWaiters(phase);
947	>	if (interrupted)
948	>	throw new InterruptedException();
949	>	return p;
950		}
951
952		/**
953	<	* Messier timeout version
953	>	* Timeout version.
954	>	* @return current phase
955		*/
956	<	private void timedWait(Thread thread, int currentPhase, long nanos)
956	>	private int timedWait(int phase, long nanos)
957		throws InterruptedException, TimeoutException {
958	<	final AtomicReference<QNode> head = this.head;
641	<	final AtomicLong state = this.state;
642	<	long lastTime = System.nanoTime();
958	>	long startTime = System.nanoTime();
959		QNode node = null;
960		boolean queued = false;
961	<	int spins = maxTimedSpins;
962	<	while (phaseOf(state.get()) == currentPhase) {
963	<	QNode h;
964	<	long now = System.nanoTime();
965	<	nanos -= now - lastTime;
966	<	lastTime = now;
967	<	if (nanos <= 0) {
968	<	if (node != null)
969	<	node.thread = null;
970	<	if (phaseOf(state.get()) == currentPhase)
971	<	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);
961	>	boolean interrupted = false;
962	>	int p;
963	>	while ((p = getPhase()) == phase && !interrupted) {
964	>	if (Thread.interrupted())
965	>	interrupted = true;
966	>	else if (nanos - (System.nanoTime() - startTime) <= 0)
967	>	break;
968	>	else if (node == null)
969	>	node = new QNode(this, phase, true, true, startTime, nanos);
970	>	else if (!queued)
971	>	queued = tryEnqueue(node);
972		else
973	<	--spins;
973	>	interrupted = node.doWait();
974		}
975		if (node != null)
976		node.thread = null;
977	+	if (p != phase || (p = getPhase()) != phase)
978	+	releaseWaiters(phase);
979	+	if (interrupted)
980	+	throw new InterruptedException();
981	+	if (p == phase)
982	+	throw new TimeoutException();
983	+	return p;
984		}
985
986	<	}
986	>	// Unsafe mechanics
987
988	+	private static final sun.misc.Unsafe UNSAFE = getUnsafe();
989	+	private static final long stateOffset =
990	+	objectFieldOffset("state", Phaser.class);
991	+
992	+	private final boolean casState(long cmp, long val) {
993	+	return UNSAFE.compareAndSwapLong(this, stateOffset, cmp, val);
994	+	}
995	+
996	+	private static long objectFieldOffset(String field, Class<?> klazz) {
997	+	try {
998	+	return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
999	+	} catch (NoSuchFieldException e) {
1000	+	// Convert Exception to corresponding Error
1001	+	NoSuchFieldError error = new NoSuchFieldError(field);
1002	+	error.initCause(e);
1003	+	throw error;
1004	+	}
1005	+	}
1006	+
1007	+	/**
1008	+	* Returns a sun.misc.Unsafe.  Suitable for use in a 3rd party package.
1009	+	* Replace with a simple call to Unsafe.getUnsafe when integrating
1010	+	* into a jdk.
1011	+	*
1012	+	* @return a sun.misc.Unsafe
1013	+	*/
1014	+	private static sun.misc.Unsafe getUnsafe() {
1015	+	try {
1016	+	return sun.misc.Unsafe.getUnsafe();
1017	+	} catch (SecurityException se) {
1018	+	try {
1019	+	return java.security.AccessController.doPrivileged
1020	+	(new java.security
1021	+	.PrivilegedExceptionAction<sun.misc.Unsafe>() {
1022	+	public sun.misc.Unsafe run() throws Exception {
1023	+	java.lang.reflect.Field f = sun.misc
1024	+	.Unsafe.class.getDeclaredField("theUnsafe");
1025	+	f.setAccessible(true);
1026	+	return (sun.misc.Unsafe) f.get(null);
1027	+	}});
1028	+	} catch (java.security.PrivilegedActionException e) {
1029	+	throw new RuntimeException("Could not initialize intrinsics",
1030	+	e.getCause());
1031	+	}
1032	+	}
1033	+	}
1034	+	}

Diff Legend

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