ViewVC Help
View File | Revision Log | Show Annotations | Download File | Root Listing
root/jsr166/jsr166/src/jsr166y/Phaser.java
(Generate patch)

Comparing jsr166/src/jsr166y/Phaser.java (file contents):
Revision 1.7 by jsr166, Mon Jan 5 03:53:26 2009 UTC vs.
Revision 1.57 by dl, Fri Nov 19 16:03:24 2010 UTC

# Line 5 | Line 5
5   */
6  
7   package jsr166y;
8 < import java.util.concurrent.*;
9 < import java.util.concurrent.atomic.*;
8 >
9 > import java.util.concurrent.TimeUnit;
10 > import java.util.concurrent.TimeoutException;
11 > import java.util.concurrent.atomic.AtomicReference;
12   import java.util.concurrent.locks.LockSupport;
11 import sun.misc.Unsafe;
12 import java.lang.reflect.*;
13  
14   /**
15 < * A reusable synchronization barrier, similar in functionality to a
16 < * {@link java.util.concurrent.CyclicBarrier} and {@link
17 < * java.util.concurrent.CountDownLatch} but supporting more flexible
18 < * usage.
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 > * <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 > * <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> The number of parties synchronizing on a phaser may vary over
48 < * time.  A task may register to be a party at any time, and may
49 < * deregister upon arriving at the barrier.  As is the case with most
50 < * basic synchronization constructs, registration and deregistration
51 < * affect only internal counts; they do not establish any further
52 < * internal bookkeeping, so tasks cannot query whether they are
53 < * registered. (However, you can introduce such bookkeeping in by
54 < * subclassing this class.)
55 < *
56 < * <li> Each generation has an associated phase value, starting at
57 < * zero, and advancing when all parties reach the barrier (wrapping
58 < * around to zero after reaching {@code Integer.MAX_VALUE}).
59 < *
60 < * <li> Like a CyclicBarrier, a Phaser may be repeatedly awaited.
61 < * Method {@code arriveAndAwaitAdvance} has effect analogous to
62 < * {@code CyclicBarrier.await}.  However, Phasers separate two
63 < * aspects of coordination, that may also be invoked independently:
64 < *
65 < * <ul>
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   *
42 *   <li> Arriving at a barrier. Methods {@code arrive} and
43 *       {@code arriveAndDeregister} do not block, but return
44 *       the phase value current upon entry to the method.
45 *
46 *   <li> Awaiting others. Method {@code awaitAdvance} requires an
47 *       argument indicating the entry phase, and returns when the
48 *       barrier advances to a new phase.
75   * </ul>
76   *
77 < *
78 < * <li> Barrier actions, performed by the task triggering a phase
79 < * advance while others may be waiting, are arranged by overriding
80 < * method {@code onAdvance}, that also controls termination.
81 < * Overriding this method may be used to similar but more flexible
82 < * effect as providing a barrier action to a CyclicBarrier.
83 < *
58 < * <li> Phasers may enter a <em>termination</em> state in which all
59 < * await actions immediately return, indicating (via a negative phase
60 < * value) that execution is complete.  Termination is triggered by
61 < * executing the overridable {@code onAdvance} method that is invoked
62 < * each time the barrier is about to be tripped. When a Phaser is
63 < * controlling an action with a fixed number of iterations, it is
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
86 < * {@code forceTermination} is also available to abruptly release
87 < * waiting threads and allow them to terminate.
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> Phasers may be tiered to reduce contention. Phasers with large
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 arranged in trees.
93 < * This will typically greatly increase throughput even though it
94 < * incurs somewhat greater per-operation overhead.
95 < *
96 < * <li> By default, {@code awaitAdvance} continues to wait even if
97 < * the waiting thread is interrupted. And unlike the case in
98 < * CyclicBarriers, exceptions encountered while tasks wait
99 < * interruptibly or with timeout do not change the state of the
100 < * barrier. If necessary, you can perform any associated recovery
101 < * within handlers of those exceptions, often after invoking
102 < * {@code forceTermination}.
103 < *
104 < * </ul>
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 > * <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 Phaser may be used instead of a {@code CountdownLatch} to control
112 < * a one-shot action serving a variable number of parties. The typical
113 < * idiom is for the method setting this up to first register, then
114 < * start the actions, then deregister, as in:
115 < *
116 < * <pre>
117 < *  void runTasks(List&lt;Runnable&gt; list) {
118 < *    final Phaser phaser = new Phaser(1); // "1" to register self
119 < *    for (Runnable r : list) {
120 < *      phaser.register();
121 < *      new Thread() {
122 < *        public void run() {
123 < *          phaser.arriveAndAwaitAdvance(); // await all creation
124 < *          r.run();
125 < *          phaser.arriveAndDeregister();   // signal completion
126 < *        }
127 < *      }.start();
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 parties.
113 > * The typical idiom is for the method setting this up to first
114 > * 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 < *   doSomethingOnBehalfOfWorkers();
131 < *   phaser.arrive(); // allow threads to start
132 < *   int p = phaser.arriveAndDeregister(); // deregister self  ...
109 < *   p = phaser.awaitAdvance(p); // ... and await arrival
110 < *   otherActions(); // do other things while tasks execute
111 < *   phaser.awaitAdvance(p); // awit final completion
112 < * }
113 < * </pre>
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>
138 < *  void startTasks(List&lt;Runnable&gt; list, final int iterations) {
139 < *    final Phaser phaser = new Phaser() {
140 < *       public boolean onAdvance(int phase, int registeredParties) {
141 < *         return phase &gt;= iterations || registeredParties == 0;
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 < *    };
125 < *    phaser.register();
126 < *    for (Runnable r : list) {
127 < *      phaser.register();
128 < *      new Thread() {
129 < *        public void run() {
130 < *           do {
131 < *             r.run();
132 < *             phaser.arriveAndAwaitAdvance();
133 < *           } while(!phaser.isTerminated();
134 < *        }
135 < *      }.start();
154 > *     }.start();
155   *   }
156   *   phaser.arriveAndDeregister(); // deregister self, don't wait
157 < * }
158 < * </pre>
157 > * }}</pre>
158 > *
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>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 > *  <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 > *   phaser.arriveAndDeregister();
181 > * }}</pre>
182 > *
183   *
184 < * <p> To create a set of tasks using a tree of Phasers,
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 < * <pre>
189 < *  void build(Task[] actions, int lo, int hi, Phaser b) {
190 < *    int step = (hi - lo) / TASKS_PER_PHASER;
191 < *    if (step &gt; 1) {
192 < *       int i = lo;
193 < *       while (i &lt; hi) {
194 < *         int r = Math.min(i + step, hi);
195 < *         build(actions, i, r, new Phaser(b));
196 < *         i = r;
197 < *       }
198 < *    }
199 < *    else {
200 < *      for (int i = lo; i &lt; hi; ++i)
201 < *        actions[i] = new Task(b);
202 < *        // assumes new Task(b) performs b.register()
203 < *    }
161 < *  }
162 < *  // .. initially called, for n tasks via
163 < *  build(new Task[n], 0, n, new Phaser());
164 < * </pre>
186 > * Task class with a constructor accepting a phaser that
187 > * it registers with 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 > *   } 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   *
171 * </pre>
172 *
210   * <p><b>Implementation notes</b>: This implementation restricts the
211   * maximum number of parties to 65535. Attempts to register additional
212 < * parties result in IllegalStateExceptions. However, you can and
212 > * parties result in {@code IllegalStateException}. However, you can and
213   * should create tiered phasers to accommodate arbitrarily large sets
214   * of participants.
215 + *
216 + * @since 1.7
217 + * @author Doug Lea
218   */
219   public class Phaser {
220      /*
# Line 187 | Line 227 | public class Phaser {
227       * Barrier state representation. Conceptually, a barrier contains
228       * four values:
229       *
230 <     * * parties -- the number of parties to wait (16 bits)
231 <     * * unarrived -- the number of parties yet to hit barrier (16 bits)
232 <     * * phase -- the generation of the barrier (31 bits)
233 <     * * terminated -- set if barrier is terminated (1 bit)
230 >     * * unarrived -- the number of parties yet to hit barrier (bits  0-15)
231 >     * * parties -- the number of parties to wait              (bits 16-31)
232 >     * * phase -- the generation of the barrier                (bits 32-62)
233 >     * * terminated -- set if barrier is terminated            (bit  63 / sign)
234       *
235       * However, to efficiently maintain atomicity, these values are
236       * packed into a single (atomic) long. Termination uses the sign
237       * bit of 32 bit representation of phase, so phase is set to -1 on
238 <     * termination. Good performace relies on keeping state decoding
238 >     * termination. Good performance relies on keeping state decoding
239       * and encoding simple, and keeping race windows short.
200     *
201     * Note: there are some cheats in arrive() that rely on unarrived
202     * being lowest 16 bits.
240       */
241      private volatile long state;
242  
243 <    private static final int ushortBits = 16;
244 <    private static final int ushortMask =  (1 << ushortBits) - 1;
245 <    private static final int phaseMask = 0x7fffffff;
243 >    private static final int  MAX_PARTIES    = 0xffff;
244 >    private static final int  MAX_PHASE      = 0x7fffffff;
245 >    private static final int  PARTIES_SHIFT  = 16;
246 >    private static final int  PHASE_SHIFT    = 32;
247 >    private static final int  UNARRIVED_MASK = 0xffff;
248 >    private static final long PARTIES_MASK   = 0xffff0000L; // for masking long
249 >    private static final long ONE_ARRIVAL    = 1L;
250 >    private static final long ONE_PARTY      = 1L << PARTIES_SHIFT;
251 >    private static final long TERMINATION_PHASE  = -1L << PHASE_SHIFT;
252 >
253 >    // The following unpacking methods are usually manually inlined
254  
255      private static int unarrivedOf(long s) {
256 <        return (int)(s & ushortMask);
256 >        return (int)s & UNARRIVED_MASK;
257      }
258  
259      private static int partiesOf(long s) {
260 <        return (int)(s & (ushortMask << 16)) >>> 16;
260 >        return (int)s >>> PARTIES_SHIFT;
261      }
262  
263      private static int phaseOf(long s) {
264 <        return (int)(s >>> 32);
264 >        return (int) (s >>> PHASE_SHIFT);
265      }
266  
267      private static int arrivedOf(long s) {
268          return partiesOf(s) - unarrivedOf(s);
269      }
270  
226    private static long stateFor(int phase, int parties, int unarrived) {
227        return (((long)phase) << 32) | ((parties << 16) | unarrived);
228    }
229
230    private static long trippedStateFor(int phase, int parties) {
231        return (((long)phase) << 32) | ((parties << 16) | parties);
232    }
233
234    private static IllegalStateException badBounds(int parties, int unarrived) {
235        return new IllegalStateException
236            ("Attempt to set " + unarrived +
237             " unarrived of " + parties + " parties");
238    }
239
271      /**
272       * The parent of this phaser, or null if none
273       */
274      private final Phaser parent;
275  
276      /**
277 <     * The root of Phaser tree. Equals this if not in a tree.  Used to
277 >     * The root of phaser tree. Equals this if not in a tree.  Used to
278       * support faster state push-down.
279       */
280      private final Phaser root;
281  
251    // Wait queues
252
282      /**
283 <     * Heads of Treiber stacks waiting for nonFJ threads. To eliminate
284 <     * contention while releasing some threads while adding others, we
283 >     * Heads of Treiber stacks for waiting threads. To eliminate
284 >     * contention when releasing some threads while adding others, we
285       * use two of them, alternating across even and odd phases.
286 +     * Subphasers share queues with root to speed up releases.
287       */
288 <    private final AtomicReference<QNode> evenQ = new AtomicReference<QNode>();
289 <    private final AtomicReference<QNode> oddQ  = new AtomicReference<QNode>();
288 >    private final AtomicReference<QNode> evenQ;
289 >    private final AtomicReference<QNode> oddQ;
290  
291      private AtomicReference<QNode> queueFor(int phase) {
292 <        return (phase & 1) == 0? evenQ : oddQ;
292 >        return ((phase & 1) == 0) ? evenQ : oddQ;
293      }
294  
295      /**
296 <     * Returns current state, first resolving lagged propagation from
297 <     * root if necessary.
296 >     * Main implementation for methods arrive and arriveAndDeregister.
297 >     * Manually tuned to speed up and minimize race windows for the
298 >     * common case of just decrementing unarrived field.
299 >     *
300 >     * @param adj - adjustment to apply to state -- either
301 >     * ONE_ARRIVAL (for arrive) or
302 >     * ONE_ARRIVAL|ONE_PARTY (for arriveAndDeregister)
303       */
304 <    private long getReconciledState() {
305 <        return parent == null? state : reconcileState();
304 >    private int doArrive(long adj) {
305 >        for (;;) {
306 >            long s = state;
307 >            int phase = (int)(s >>> PHASE_SHIFT);
308 >            if (phase < 0)
309 >                return phase;
310 >            int unarrived = (int)s & UNARRIVED_MASK;
311 >            if (unarrived == 0)
312 >                checkBadArrive(s);
313 >            else if (UNSAFE.compareAndSwapLong(this, stateOffset, s, s-=adj)) {
314 >                if (unarrived == 1) {
315 >                    long p = s & PARTIES_MASK; // unshifted parties field
316 >                    long lu = p >>> PARTIES_SHIFT;
317 >                    int u = (int)lu;
318 >                    int nextPhase = (phase + 1) & MAX_PHASE;
319 >                    long next = ((long)nextPhase << PHASE_SHIFT) | p | lu;
320 >                    final Phaser parent = this.parent;
321 >                    if (parent == null) {
322 >                        if (onAdvance(phase, u))
323 >                            next |= TERMINATION_PHASE; // obliterate phase
324 >                        UNSAFE.compareAndSwapLong(this, stateOffset, s, next);
325 >                        releaseWaiters(phase);
326 >                    }
327 >                    else {
328 >                        parent.doArrive((u == 0) ?
329 >                                        ONE_ARRIVAL|ONE_PARTY : ONE_ARRIVAL);
330 >                        if ((int)(parent.state >>> PHASE_SHIFT) != nextPhase ||
331 >                            ((int)(state >>> PHASE_SHIFT) != nextPhase &&
332 >                             !UNSAFE.compareAndSwapLong(this, stateOffset,
333 >                                                        s, next)))
334 >                            reconcileState();
335 >                    }
336 >                }
337 >                return phase;
338 >            }
339 >        }
340 >    }
341 >
342 >    /**
343 >     * Rechecks state and throws bounds exceptions on arrival -- called
344 >     * only if unarrived is apparently zero.
345 >     */
346 >    private void checkBadArrive(long s) {
347 >        if (reconcileState() == s)
348 >            throw new IllegalStateException
349 >                ("Attempted arrival of unregistered party for " +
350 >                 stateToString(s));
351 >    }
352 >
353 >    /**
354 >     * Implementation of register, bulkRegister
355 >     *
356 >     * @param registrations number to add to both parties and unarrived fields
357 >     */
358 >    private int doRegister(int registrations) {
359 >        // assert registrations > 0;
360 >        // adjustment to state
361 >        long adj = ((long)registrations << PARTIES_SHIFT) | registrations;
362 >        final Phaser parent = this.parent;
363 >        for (;;) {
364 >            long s = (parent == null) ? state : reconcileState();
365 >            int phase = (int)(s >>> PHASE_SHIFT);
366 >            if (phase < 0)
367 >                return phase;
368 >            int parties = (int)s >>> PARTIES_SHIFT;
369 >            if (parties != 0 && ((int)s & UNARRIVED_MASK) == 0)
370 >                internalAwaitAdvance(phase, null); // wait for onAdvance
371 >            else if (registrations > MAX_PARTIES - parties)
372 >                throw new IllegalStateException(badRegister(s));
373 >            else if (UNSAFE.compareAndSwapLong(this, stateOffset, s, s + adj))
374 >                return phase;
375 >        }
376      }
377  
378      /**
379 <     * Recursively resolves state.
379 >     * Returns message string for out of bounds exceptions on registration.
380 >     */
381 >    private String badRegister(long s) {
382 >        return "Attempt to register more than " +
383 >            MAX_PARTIES + " parties for " + stateToString(s);
384 >    }
385 >
386 >    /**
387 >     * Recursively resolves lagged phase propagation from root if necessary.
388       */
389      private long reconcileState() {
390 <        Phaser p = parent;
390 >        Phaser par = parent;
391          long s = state;
392 <        if (p != null) {
393 <            while (unarrivedOf(s) == 0 && phaseOf(s) != phaseOf(root.state)) {
394 <                long parentState = p.getReconciledState();
395 <                int parentPhase = phaseOf(parentState);
396 <                int phase = phaseOf(s = state);
397 <                if (phase != parentPhase) {
398 <                    long next = trippedStateFor(parentPhase, partiesOf(s));
399 <                    if (casState(s, next)) {
400 <                        releaseWaiters(phase);
401 <                        s = next;
402 <                    }
392 >        if (par != null) {
393 >            Phaser rt = root;
394 >            int phase, rPhase;
395 >            while ((phase = (int)(s >>> PHASE_SHIFT)) >= 0 &&
396 >                   (rPhase = (int)(rt.state >>> PHASE_SHIFT)) != phase) {
397 >                if ((int)(par.state >>> PHASE_SHIFT) != rPhase)
398 >                    par.reconcileState();
399 >                else if (rPhase < 0 || ((int)s & UNARRIVED_MASK) == 0) {
400 >                    long u = s & PARTIES_MASK; // reset unarrived to parties
401 >                    long next = ((((long) rPhase) << PHASE_SHIFT) | u |
402 >                                 (u >>> PARTIES_SHIFT));
403 >                    if (state == s &&
404 >                        UNSAFE.compareAndSwapLong(this, stateOffset,
405 >                                                  s, s = next))
406 >                        break;
407                  }
408 +                s = state;
409              }
410          }
411          return s;
412      }
413  
414      /**
415 <     * Creates a new Phaser without any initially registered parties,
416 <     * initial phase number 0, and no parent.
415 >     * Creates a new phaser without any initially registered parties,
416 >     * initial phase number 0, and no parent. Any thread using this
417 >     * phaser will need to first register for it.
418       */
419      public Phaser() {
420 <        this(null);
420 >        this(null, 0);
421      }
422  
423      /**
424 <     * Creates a new Phaser with the given numbers of registered
424 >     * Creates a new phaser with the given number of registered
425       * unarrived parties, initial phase number 0, and no parent.
426 <     * @param parties the number of parties required to trip barrier.
426 >     *
427 >     * @param parties the number of parties required to trip barrier
428       * @throws IllegalArgumentException if parties less than zero
429 <     * or greater than the maximum number of parties supported.
429 >     * or greater than the maximum number of parties supported
430       */
431      public Phaser(int parties) {
432          this(null, parties);
433      }
434  
435      /**
436 <     * Creates a new Phaser with the given parent, without any
436 >     * Creates a new phaser with the given parent, without any
437       * initially registered parties. If parent is non-null this phaser
438       * is registered with the parent and its initial phase number is
439       * the same as that of parent phaser.
440 <     * @param parent the parent phaser.
440 >     *
441 >     * @param parent the parent phaser
442       */
443      public Phaser(Phaser parent) {
444 <        int phase = 0;
324 <        this.parent = parent;
325 <        if (parent != null) {
326 <            this.root = parent.root;
327 <            phase = parent.register();
328 <        }
329 <        else
330 <            this.root = this;
331 <        this.state = trippedStateFor(phase, 0);
444 >        this(parent, 0);
445      }
446  
447      /**
448 <     * Creates a new Phaser with the given parent and numbers of
449 <     * registered unarrived parties. If parent is non-null this phaser
448 >     * Creates a new phaser with the given parent and number of
449 >     * registered unarrived parties. If parent is non-null, this phaser
450       * is registered with the parent and its initial phase number is
451       * the same as that of parent phaser.
452 <     * @param parent the parent phaser.
453 <     * @param parties the number of parties required to trip barrier.
452 >     *
453 >     * @param parent the parent phaser
454 >     * @param parties the number of parties required to trip barrier
455       * @throws IllegalArgumentException if parties less than zero
456 <     * or greater than the maximum number of parties supported.
456 >     * or greater than the maximum number of parties supported
457       */
458      public Phaser(Phaser parent, int parties) {
459 <        if (parties < 0 || parties > ushortMask)
459 >        if (parties >>> PARTIES_SHIFT != 0)
460              throw new IllegalArgumentException("Illegal number of parties");
461 <        int phase = 0;
461 >        int phase;
462          this.parent = parent;
463          if (parent != null) {
464 <            this.root = parent.root;
464 >            Phaser r = parent.root;
465 >            this.root = r;
466 >            this.evenQ = r.evenQ;
467 >            this.oddQ = r.oddQ;
468              phase = parent.register();
469          }
470 <        else
470 >        else {
471              this.root = this;
472 <        this.state = trippedStateFor(phase, parties);
472 >            this.evenQ = new AtomicReference<QNode>();
473 >            this.oddQ = new AtomicReference<QNode>();
474 >            phase = 0;
475 >        }
476 >        long p = (long)parties;
477 >        this.state = (((long)phase) << PHASE_SHIFT) | p | (p << PARTIES_SHIFT);
478      }
479  
480      /**
481       * Adds a new unarrived party to this phaser.
482 <     * @return the current barrier phase number upon registration
482 >     * If an ongoing invocation of {@link #onAdvance} is in progress,
483 >     * this method may wait until its completion before registering.
484 >     *
485 >     * @return the arrival phase number to which this registration applied
486       * @throws IllegalStateException if attempting to register more
487 <     * than the maximum supported number of parties.
487 >     * than the maximum supported number of parties
488       */
489      public int register() {
490          return doRegister(1);
# Line 367 | Line 492 | public class Phaser {
492  
493      /**
494       * Adds the given number of new unarrived parties to this phaser.
495 <     * @param parties the number of parties required to trip barrier.
496 <     * @return the current barrier phase number upon registration
495 >     * If an ongoing invocation of {@link #onAdvance} is in progress,
496 >     * this method may wait until its completion before registering.
497 >     *
498 >     * @param parties the number of additional parties required to trip barrier
499 >     * @return the arrival phase number to which this registration applied
500       * @throws IllegalStateException if attempting to register more
501 <     * than the maximum supported number of parties.
501 >     * than the maximum supported number of parties
502 >     * @throws IllegalArgumentException if {@code parties < 0}
503       */
504      public int bulkRegister(int parties) {
505          if (parties < 0)
# Line 381 | Line 510 | public class Phaser {
510      }
511  
512      /**
384     * Shared code for register, bulkRegister
385     */
386    private int doRegister(int registrations) {
387        int phase;
388        for (;;) {
389            long s = getReconciledState();
390            phase = phaseOf(s);
391            int unarrived = unarrivedOf(s) + registrations;
392            int parties = partiesOf(s) + registrations;
393            if (phase < 0)
394                break;
395            if (parties > ushortMask || unarrived > ushortMask)
396                throw badBounds(parties, unarrived);
397            if (phase == phaseOf(root.state) &&
398                casState(s, stateFor(phase, parties, unarrived)))
399                break;
400        }
401        return phase;
402    }
403
404    /**
513       * Arrives at the barrier, but does not wait for others.  (You can
514 <     * in turn wait for others via {@link #awaitAdvance}).
514 >     * in turn wait for others via {@link #awaitAdvance}).  It is an
515 >     * unenforced usage error for an unregistered party to invoke this
516 >     * method.
517       *
518 <     * @return the barrier phase number upon entry to this method, or a
409 <     * negative value if terminated;
518 >     * @return the arrival phase number, or a negative value if terminated
519       * @throws IllegalStateException if not terminated and the number
520 <     * of unarrived parties would become negative.
520 >     * of unarrived parties would become negative
521       */
522      public int arrive() {
523 <        int phase;
415 <        for (;;) {
416 <            long s = state;
417 <            phase = phaseOf(s);
418 <            int parties = partiesOf(s);
419 <            int unarrived = unarrivedOf(s) - 1;
420 <            if (unarrived > 0) {        // Not the last arrival
421 <                if (casState(s, s - 1)) // s-1 adds one arrival
422 <                    break;
423 <            }
424 <            else if (unarrived == 0) {  // the last arrival
425 <                Phaser par = parent;
426 <                if (par == null) {      // directly trip
427 <                    if (casState
428 <                        (s,
429 <                         trippedStateFor(onAdvance(phase, parties)? -1 :
430 <                                         ((phase + 1) & phaseMask), parties))) {
431 <                        releaseWaiters(phase);
432 <                        break;
433 <                    }
434 <                }
435 <                else {                  // cascade to parent
436 <                    if (casState(s, s - 1)) { // zeroes unarrived
437 <                        par.arrive();
438 <                        reconcileState();
439 <                        break;
440 <                    }
441 <                }
442 <            }
443 <            else if (phase < 0) // Don't throw exception if terminated
444 <                break;
445 <            else if (phase != phaseOf(root.state)) // or if unreconciled
446 <                reconcileState();
447 <            else
448 <                throw badBounds(parties, unarrived);
449 <        }
450 <        return phase;
523 >        return doArrive(ONE_ARRIVAL);
524      }
525  
526      /**
527 <     * Arrives at the barrier, and deregisters from it, without
528 <     * waiting for others. Deregistration reduces number of parties
527 >     * Arrives at the barrier and deregisters from it without waiting
528 >     * for others. Deregistration reduces the number of parties
529       * required to trip the barrier in future phases.  If this phaser
530       * has a parent, and deregistration causes this phaser to have
531 <     * zero parties, this phaser is also deregistered from its parent.
531 >     * zero parties, this phaser also arrives at and is deregistered
532 >     * from its parent.  It is an unenforced usage error for an
533 >     * unregistered party to invoke this method.
534       *
535 <     * @return the current barrier phase number upon entry to
461 <     * this method, or a negative value if terminated;
535 >     * @return the arrival phase number, or a negative value if terminated
536       * @throws IllegalStateException if not terminated and the number
537 <     * of registered or unarrived parties would become negative.
537 >     * of registered or unarrived parties would become negative
538       */
539      public int arriveAndDeregister() {
540 <        // similar code to arrive, but too different to merge
467 <        Phaser par = parent;
468 <        int phase;
469 <        for (;;) {
470 <            long s = state;
471 <            phase = phaseOf(s);
472 <            int parties = partiesOf(s) - 1;
473 <            int unarrived = unarrivedOf(s) - 1;
474 <            if (parties >= 0) {
475 <                if (unarrived > 0 || (unarrived == 0 && par != null)) {
476 <                    if (casState
477 <                        (s,
478 <                         stateFor(phase, parties, unarrived))) {
479 <                        if (unarrived == 0) {
480 <                            par.arriveAndDeregister();
481 <                            reconcileState();
482 <                        }
483 <                        break;
484 <                    }
485 <                    continue;
486 <                }
487 <                if (unarrived == 0) {
488 <                    if (casState
489 <                        (s,
490 <                         trippedStateFor(onAdvance(phase, parties)? -1 :
491 <                                         ((phase + 1) & phaseMask), parties))) {
492 <                        releaseWaiters(phase);
493 <                        break;
494 <                    }
495 <                    continue;
496 <                }
497 <                if (phase < 0)
498 <                    break;
499 <                if (par != null && phase != phaseOf(root.state)) {
500 <                    reconcileState();
501 <                    continue;
502 <                }
503 <            }
504 <            throw badBounds(parties, unarrived);
505 <        }
506 <        return phase;
540 >        return doArrive(ONE_ARRIVAL|ONE_PARTY);
541      }
542  
543      /**
544       * Arrives at the barrier and awaits others. Equivalent in effect
545 <     * to {@code awaitAdvance(arrive())}.  If you instead need to
546 <     * await with interruption of timeout, and/or deregister upon
547 <     * arrival, you can arrange them using analogous constructions.
548 <     * @return the phase on entry to this method
545 >     * to {@code awaitAdvance(arrive())}.  If you need to await with
546 >     * interruption or timeout, you can arrange this with an analogous
547 >     * construction using one of the other forms of the {@code
548 >     * awaitAdvance} method.  If instead you need to deregister upon
549 >     * arrival, use {@link #arriveAndDeregister}. It is an unenforced
550 >     * usage error for an unregistered party to invoke this method.
551 >     *
552 >     * @return the arrival phase number, or a negative number if terminated
553       * @throws IllegalStateException if not terminated and the number
554 <     * of unarrived parties would become negative.
554 >     * of unarrived parties would become negative
555       */
556      public int arriveAndAwaitAdvance() {
557          return awaitAdvance(arrive());
558      }
559  
560      /**
561 <     * Awaits the phase of the barrier to advance from the given
562 <     * value, or returns immediately if argument is negative or this
563 <     * barrier is terminated.
564 <     * @param phase the phase on entry to this method
565 <     * @return the phase on exit from this method
561 >     * Awaits the phase of the barrier to advance from the given phase
562 >     * value, returning immediately if the current phase of the
563 >     * barrier is not equal to the given phase value or this barrier
564 >     * is terminated.
565 >     *
566 >     * @param phase an arrival phase number, or negative value if
567 >     * terminated; this argument is normally the value returned by a
568 >     * previous call to {@code arrive} or its variants
569 >     * @return the next arrival phase number, or a negative value
570 >     * if terminated or argument is negative
571       */
572      public int awaitAdvance(int phase) {
573          if (phase < 0)
574              return phase;
575 <        long s = getReconciledState();
576 <        int p = phaseOf(s);
577 <        if (p != phase)
535 <            return p;
536 <        if (unarrivedOf(s) == 0)
537 <            parent.awaitAdvance(phase);
538 <        // Fall here even if parent waited, to reconcile and help release
539 <        return untimedWait(phase);
575 >        long s = (parent == null) ? state : reconcileState();
576 >        int p = (int)(s >>> PHASE_SHIFT);
577 >        return (p != phase) ? p : internalAwaitAdvance(phase, null);
578      }
579  
580      /**
581 <     * Awaits the phase of the barrier to advance from the given
582 <     * value, or returns immediately if argumet is negative or this
583 <     * barrier is terminated, or throws InterruptedException if
584 <     * interrupted while waiting.
585 <     * @param phase the phase on entry to this method
586 <     * @return the phase on exit from this method
581 >     * Awaits the phase of the barrier to advance from the given phase
582 >     * value, throwing {@code InterruptedException} if interrupted
583 >     * while waiting, or returning immediately if the current phase of
584 >     * the barrier is not equal to the given phase value or this
585 >     * barrier is terminated.
586 >     *
587 >     * @param phase an arrival phase number, or negative value if
588 >     * terminated; this argument is normally the value returned by a
589 >     * previous call to {@code arrive} or its variants
590 >     * @return the next arrival phase number, or a negative value
591 >     * if terminated or argument is negative
592       * @throws InterruptedException if thread interrupted while waiting
593       */
594 <    public int awaitAdvanceInterruptibly(int phase) throws InterruptedException {
594 >    public int awaitAdvanceInterruptibly(int phase)
595 >        throws InterruptedException {
596          if (phase < 0)
597              return phase;
598 <        long s = getReconciledState();
599 <        int p = phaseOf(s);
600 <        if (p != phase)
601 <            return p;
602 <        if (unarrivedOf(s) != 0)
603 <            parent.awaitAdvanceInterruptibly(phase);
604 <        return interruptibleWait(phase);
598 >        long s = (parent == null) ? state : reconcileState();
599 >        int p = (int)(s >>> PHASE_SHIFT);
600 >        if (p == phase) {
601 >            QNode node = new QNode(this, phase, true, false, 0L);
602 >            p = internalAwaitAdvance(phase, node);
603 >            if (node.wasInterrupted)
604 >                throw new InterruptedException();
605 >        }
606 >        return p;
607      }
608  
609      /**
610 <     * Awaits the phase of the barrier to advance from the given value
611 <     * or the given timeout elapses, or returns immediately if
612 <     * argument is negative or this barrier is terminated.
613 <     * @param phase the phase on entry to this method
614 <     * @return the phase on exit from this method
610 >     * Awaits the phase of the barrier to advance from the given phase
611 >     * value or the given timeout to elapse, throwing {@code
612 >     * InterruptedException} if interrupted while waiting, or
613 >     * returning immediately if the current phase of the barrier is
614 >     * not equal to the given phase value or this barrier is
615 >     * terminated.
616 >     *
617 >     * @param phase an arrival phase number, or negative value if
618 >     * terminated; this argument is normally the value returned by a
619 >     * previous call to {@code arrive} or its variants
620 >     * @param timeout how long to wait before giving up, in units of
621 >     *        {@code unit}
622 >     * @param unit a {@code TimeUnit} determining how to interpret the
623 >     *        {@code timeout} parameter
624 >     * @return the next arrival phase number, or a negative value
625 >     * if terminated or argument is negative
626       * @throws InterruptedException if thread interrupted while waiting
627       * @throws TimeoutException if timed out while waiting
628       */
629 <    public int awaitAdvanceInterruptibly(int phase, long timeout, TimeUnit unit)
629 >    public int awaitAdvanceInterruptibly(int phase,
630 >                                         long timeout, TimeUnit unit)
631          throws InterruptedException, TimeoutException {
632          if (phase < 0)
633              return phase;
634 <        long s = getReconciledState();
635 <        int p = phaseOf(s);
636 <        if (p != phase)
637 <            return p;
638 <        if (unarrivedOf(s) == 0)
639 <            parent.awaitAdvanceInterruptibly(phase, timeout, unit);
640 <        return timedWait(phase, unit.toNanos(timeout));
634 >        long s = (parent == null) ? state : reconcileState();
635 >        int p = (int)(s >>> PHASE_SHIFT);
636 >        if (p == phase) {
637 >            long nanos = unit.toNanos(timeout);
638 >            QNode node = new QNode(this, phase, true, true, nanos);
639 >            p = internalAwaitAdvance(phase, node);
640 >            if (node.wasInterrupted)
641 >                throw new InterruptedException();
642 >            else if (p == phase)
643 >                throw new TimeoutException();
644 >        }
645 >        return p;
646      }
647  
648      /**
649 <     * Forces this barrier to enter termination state. Counts of
650 <     * arrived and registered parties are unaffected. If this phaser
651 <     * has a parent, it too is terminated. This method may be useful
652 <     * for coordinating recovery after one or more tasks encounter
653 <     * unexpected exceptions.
649 >     * Forces this barrier to enter termination state.  Counts of
650 >     * arrived and registered parties are unaffected.  If this phaser
651 >     * is a member of a tiered set of phasers, then all of the phasers
652 >     * in the set are terminated.  If this phaser is already
653 >     * terminated, this method has no effect.  This method may be
654 >     * useful for coordinating recovery after one or more tasks
655 >     * encounter unexpected exceptions.
656       */
657      public void forceTermination() {
658 <        for (;;) {
659 <            long s = getReconciledState();
660 <            int phase = phaseOf(s);
661 <            int parties = partiesOf(s);
662 <            int unarrived = unarrivedOf(s);
663 <            if (phase < 0 ||
664 <                casState(s, stateFor(-1, parties, unarrived))) {
600 <                releaseWaiters(0);
658 >        // Only need to change root state
659 >        final Phaser root = this.root;
660 >        long s;
661 >        while ((s = root.state) >= 0) {
662 >            if (UNSAFE.compareAndSwapLong(root, stateOffset,
663 >                                          s, s | TERMINATION_PHASE)) {
664 >                releaseWaiters(0); // signal all threads
665                  releaseWaiters(1);
602                if (parent != null)
603                    parent.forceTermination();
666                  return;
667              }
668          }
# Line 610 | Line 672 | public class Phaser {
672       * Returns the current phase number. The maximum phase number is
673       * {@code Integer.MAX_VALUE}, after which it restarts at
674       * zero. Upon termination, the phase number is negative.
675 +     *
676       * @return the phase number, or a negative value if terminated
677       */
678      public final int getPhase() {
679 <        return phaseOf(getReconciledState());
617 <    }
618 <
619 <    /**
620 <     * Returns true if the current phase number equals the given phase.
621 <     * @param phase the phase
622 <     * @return true if the current phase number equals the given phase.
623 <     */
624 <    public final boolean hasPhase(int phase) {
625 <        return phaseOf(getReconciledState()) == phase;
679 >        return (int)(root.state >>> PHASE_SHIFT);
680      }
681  
682      /**
683       * Returns the number of parties registered at this barrier.
684 +     *
685       * @return the number of parties
686       */
687      public int getRegisteredParties() {
# Line 634 | Line 689 | public class Phaser {
689      }
690  
691      /**
692 <     * Returns the number of parties that have arrived at the current
693 <     * phase of this barrier.
692 >     * Returns the number of registered parties that have arrived at
693 >     * the current phase of this barrier.
694 >     *
695       * @return the number of arrived parties
696       */
697      public int getArrivedParties() {
698 <        return arrivedOf(state);
698 >        return arrivedOf(parent==null? state : reconcileState());
699      }
700  
701      /**
702       * Returns the number of registered parties that have not yet
703       * arrived at the current phase of this barrier.
704 +     *
705       * @return the number of unarrived parties
706       */
707      public int getUnarrivedParties() {
708 <        return unarrivedOf(state);
708 >        return unarrivedOf(parent==null? state : reconcileState());
709      }
710  
711      /**
712 <     * Returns the parent of this phaser, or null if none.
713 <     * @return the parent of this phaser, or null if none.
712 >     * Returns the parent of this phaser, or {@code null} if none.
713 >     *
714 >     * @return the parent of this phaser, or {@code null} if none
715       */
716      public Phaser getParent() {
717          return parent;
# Line 662 | Line 720 | public class Phaser {
720      /**
721       * Returns the root ancestor of this phaser, which is the same as
722       * this phaser if it has no parent.
723 <     * @return the root ancestor of this phaser.
723 >     *
724 >     * @return the root ancestor of this phaser
725       */
726      public Phaser getRoot() {
727          return root;
728      }
729  
730      /**
731 <     * Returns true if this barrier has been terminated.
732 <     * @return true if this barrier has been terminated
731 >     * Returns {@code true} if this barrier has been terminated.
732 >     *
733 >     * @return {@code true} if this barrier has been terminated
734       */
735      public boolean isTerminated() {
736 <        return getPhase() < 0;
736 >        return root.state < 0L;
737      }
738  
739      /**
740 <     * Overridable method to perform an action upon phase advance, and
741 <     * to control termination. This method is invoked whenever the
742 <     * barrier is tripped (and thus all other waiting parties are
743 <     * dormant). If it returns true, then, rather than advance the
744 <     * phase number, this barrier will be set to a final termination
745 <     * state, and subsequent calls to {@code isTerminated} will
746 <     * return true.
740 >     * Overridable method to perform an action upon impending phase
741 >     * advance, and to control termination. This method is invoked
742 >     * upon arrival of the party tripping the barrier (when all other
743 >     * waiting parties are dormant).  If this method returns {@code
744 >     * true}, then, rather than advance the phase number, this barrier
745 >     * will be set to a final termination state, and subsequent calls
746 >     * to {@link #isTerminated} will return true. Any (unchecked)
747 >     * Exception or Error thrown by an invocation of this method is
748 >     * propagated to the party attempting to trip the barrier, in
749 >     * which case no advance occurs.
750 >     *
751 >     * <p>The arguments to this method provide the state of the phaser
752 >     * prevailing for the current transition.  The effects of invoking
753 >     * arrival, registration, and waiting methods on this Phaser from
754 >     * within {@code onAdvance} are unspecified and should not be
755 >     * relied on.
756       *
757 <     * <p> The default version returns true when the number of
757 >     * <p>If this Phaser is a member of a tiered set of Phasers, then
758 >     * {@code onAdvance} is invoked only for its root Phaser on each
759 >     * advance.
760 >     *
761 >     * <p>The default version returns {@code true} when the number of
762       * registered parties is zero. Normally, overrides that arrange
763       * termination for other reasons should also preserve this
764       * property.
765       *
693     * <p> You may override this method to perform an action with side
694     * effects visible to participating tasks, but it is in general
695     * only sensible to do so in designs where all parties register
696     * before any arrive, and all {@code awaitAdvance} at each phase.
697     * Otherwise, you cannot ensure lack of interference. In
698     * particular, this method may be invoked more than once per
699     * transition if other parties successfully register while the
700     * invocation of this method is in progress, thus postponing the
701     * transition until those parties also arrive, re-triggering this
702     * method.
703     *
766       * @param phase the phase number on entering the barrier
767 <     * @param registeredParties the current number of registered
768 <     * parties.
707 <     * @return true if this barrier should terminate
767 >     * @param registeredParties the current number of registered parties
768 >     * @return {@code true} if this barrier should terminate
769       */
770      protected boolean onAdvance(int phase, int registeredParties) {
771          return registeredParties <= 0;
# Line 713 | Line 774 | public class Phaser {
774      /**
775       * Returns a string identifying this phaser, as well as its
776       * state.  The state, in brackets, includes the String {@code
777 <     * "phase ="} followed by the phase number, {@code "parties ="}
777 >     * "phase = "} followed by the phase number, {@code "parties = "}
778       * followed by the number of registered parties, and {@code
779 <     * "arrived ="} followed by the number of arrived parties
779 >     * "arrived = "} followed by the number of arrived parties.
780       *
781       * @return a string identifying this barrier, as well as its state
782       */
783      public String toString() {
784 <        long s = getReconciledState();
724 <        return super.toString() + "[phase = " + phaseOf(s) + " parties = " + partiesOf(s) + " arrived = " + arrivedOf(s) + "]";
784 >        return stateToString(reconcileState());
785      }
786  
727    // methods for waiting
728
729    /** The number of CPUs, for spin control */
730    static final int NCPUS = Runtime.getRuntime().availableProcessors();
731
732    /**
733     * The number of times to spin before blocking in timed waits.
734     * The value is empirically derived.
735     */
736    static final int maxTimedSpins = (NCPUS < 2)? 0 : 32;
737
738    /**
739     * The number of times to spin before blocking in untimed waits.
740     * This is greater than timed value because untimed waits spin
741     * faster since they don't need to check times on each spin.
742     */
743    static final int maxUntimedSpins = maxTimedSpins * 32;
744
787      /**
788 <     * The number of nanoseconds for which it is faster to spin
747 <     * rather than to use timed park. A rough estimate suffices.
788 >     * Implementation of toString and string-based error messages
789       */
790 <    static final long spinForTimeoutThreshold = 1000L;
791 <
792 <    /**
793 <     * Wait nodes for Treiber stack representing wait queue for non-FJ
794 <     * tasks.
754 <     */
755 <    static final class QNode {
756 <        QNode next;
757 <        volatile Thread thread; // nulled to cancel wait
758 <        QNode() {
759 <            thread = Thread.currentThread();
760 <        }
761 <        void signal() {
762 <            Thread t = thread;
763 <            if (t != null) {
764 <                thread = null;
765 <                LockSupport.unpark(t);
766 <            }
767 <        }
790 >    private String stateToString(long s) {
791 >        return super.toString() +
792 >            "[phase = " + phaseOf(s) +
793 >            " parties = " + partiesOf(s) +
794 >            " arrived = " + arrivedOf(s) + "]";
795      }
796  
797 +    // Waiting mechanics
798 +
799      /**
800 <     * Removes and signals waiting threads from wait queue
800 >     * Removes and signals threads from queue for phase.
801       */
802      private void releaseWaiters(int phase) {
803          AtomicReference<QNode> head = queueFor(phase);
804          QNode q;
805 <        while ((q = head.get()) != null) {
805 >        int p;
806 >        while ((q = head.get()) != null &&
807 >               ((p = q.phase) == phase ||
808 >                (int)(root.state >>> PHASE_SHIFT) != p)) {
809              if (head.compareAndSet(q, q.next))
810                  q.signal();
811          }
812      }
813  
814 +    /** The number of CPUs, for spin control */
815 +    private static final int NCPU = Runtime.getRuntime().availableProcessors();
816 +
817      /**
818 <     * Enqueues node and waits unless aborted or signalled.
818 >     * The number of times to spin before blocking while waiting for
819 >     * advance, per arrival while waiting. On multiprocessors, fully
820 >     * blocking and waking up a large number of threads all at once is
821 >     * usually a very slow process, so we use rechargeable spins to
822 >     * avoid it when threads regularly arrive: When a thread in
823 >     * internalAwaitAdvance notices another arrival before blocking,
824 >     * and there appear to be enough CPUs available, it spins
825 >     * SPINS_PER_ARRIVAL more times before blocking. Plus, even on
826 >     * uniprocessors, there is at least one intervening Thread.yield
827 >     * before blocking. The value trades off good-citizenship vs big
828 >     * unnecessary slowdowns.
829       */
830 <    private int untimedWait(int phase) {
786 <        int spins = maxUntimedSpins;
787 <        QNode node = null;
788 <        boolean interrupted = false;
789 <        boolean queued = false;
790 <        int p;
791 <        while ((p = getPhase()) == phase) {
792 <            interrupted = Thread.interrupted();
793 <            if (node != null) {
794 <                if (!queued) {
795 <                    AtomicReference<QNode> head = queueFor(phase);
796 <                    queued = head.compareAndSet(node.next = head.get(), node);
797 <                }
798 <                else if (node.thread != null)
799 <                    LockSupport.park(this);
800 <            }
801 <            else if (spins <= 0)
802 <                node = new QNode();
803 <            else
804 <                --spins;
805 <        }
806 <        if (node != null)
807 <            node.thread = null;
808 <        if (interrupted)
809 <            Thread.currentThread().interrupt();
810 <        releaseWaiters(phase);
811 <        return p;
812 <    }
830 >    static final int SPINS_PER_ARRIVAL = (NCPU < 2) ? 1 : 1 << 8;
831  
832      /**
833 <     * Messier interruptible version
834 <     */
835 <    private int interruptibleWait(int phase) throws InterruptedException {
836 <        int spins = maxUntimedSpins;
837 <        QNode node = null;
838 <        boolean queued = false;
839 <        boolean interrupted = false;
833 >     * Possibly blocks and waits for phase to advance unless aborted.
834 >     *
835 >     * @param phase current phase
836 >     * @param node if non-null, the wait node to track interrupt and timeout;
837 >     * if null, denotes noninterruptible wait
838 >     * @return current phase
839 >     */
840 >    private int internalAwaitAdvance(int phase, QNode node) {
841 >        Phaser current = this;       // to eventually wait at root if tiered
842 >        boolean queued = false;      // true when node is enqueued
843 >        int lastUnarrived = -1;      // to increase spins upon change
844 >        int spins = SPINS_PER_ARRIVAL;
845 >        long s;
846          int p;
847 <        while ((p = getPhase()) == phase) {
848 <            if (interrupted = Thread.interrupted())
849 <                break;
850 <            if (node != null) {
851 <                if (!queued) {
852 <                    AtomicReference<QNode> head = queueFor(phase);
853 <                    queued = head.compareAndSet(node.next = head.get(), node);
847 >        while ((p = (int)((s = current.state) >>> PHASE_SHIFT)) == phase) {
848 >            Phaser par;
849 >            int unarrived = (int)s & UNARRIVED_MASK;
850 >            if (unarrived != lastUnarrived) {
851 >                if (lastUnarrived == -1) // ensure old queue clean
852 >                    releaseWaiters(phase-1);
853 >                if ((lastUnarrived = unarrived) < NCPU)
854 >                    spins += SPINS_PER_ARRIVAL;
855 >            }
856 >            else if (unarrived == 0 && (par = current.parent) != null) {
857 >                current = par;       // if all arrived, use parent
858 >                par = par.parent;
859 >                lastUnarrived = -1;
860 >            }
861 >            else if (spins > 0) {
862 >                if (--spins == (SPINS_PER_ARRIVAL >>> 1))
863 >                    Thread.yield();  // yield midway through spin
864 >            }
865 >            else if (node == null)   // must be noninterruptible
866 >                node = new QNode(this, phase, false, false, 0L);
867 >            else if (node.isReleasable()) {
868 >                if ((p = (int)(root.state >>> PHASE_SHIFT)) != phase)
869 >                    break;
870 >                else
871 >                    return phase;    // aborted
872 >            }
873 >            else if (!queued) {      // push onto queue
874 >                AtomicReference<QNode> head = queueFor(phase);
875 >                QNode q = head.get();
876 >                if (q == null || q.phase == phase) {
877 >                    node.next = q;
878 >                    if ((p = (int)(root.state >>> PHASE_SHIFT)) != phase)
879 >                        break;       // recheck to avoid stale enqueue
880 >                    else
881 >                        queued = head.compareAndSet(q, node);
882 >                }
883 >            }
884 >            else {
885 >                try {
886 >                    ForkJoinPool.managedBlock(node);
887 >                } catch (InterruptedException ie) {
888 >                    node.wasInterrupted = true;
889                  }
831                else if (node.thread != null)
832                    LockSupport.park(this);
890              }
834            else if (spins <= 0)
835                node = new QNode();
836            else
837                --spins;
891          }
839        if (node != null)
840            node.thread = null;
841        if (interrupted)
842            throw new InterruptedException();
892          releaseWaiters(phase);
893 +        if (node != null)
894 +            node.onRelease();
895          return p;
896      }
897  
898      /**
899 <     * Even messier timeout version.
899 >     * Wait nodes for Treiber stack representing wait queue
900       */
901 <    private int timedWait(int phase, long nanos)
902 <        throws InterruptedException, TimeoutException {
903 <        int p;
904 <        if ((p = getPhase()) == phase) {
905 <            long lastTime = System.nanoTime();
906 <            int spins = maxTimedSpins;
907 <            QNode node = null;
908 <            boolean queued = false;
909 <            boolean interrupted = false;
910 <            while ((p = getPhase()) == phase) {
911 <                if (interrupted = Thread.interrupted())
912 <                    break;
913 <                long now = System.nanoTime();
914 <                if ((nanos -= now - lastTime) <= 0)
915 <                    break;
916 <                lastTime = now;
917 <                if (node != null) {
918 <                    if (!queued) {
919 <                        AtomicReference<QNode> head = queueFor(phase);
920 <                        queued = head.compareAndSet(node.next = head.get(), node);
921 <                    }
922 <                    else if (node.thread != null &&
923 <                             nanos > spinForTimeoutThreshold) {
924 <                        LockSupport.parkNanos(this, nanos);
901 >    static final class QNode implements ForkJoinPool.ManagedBlocker {
902 >        final Phaser phaser;
903 >        final int phase;
904 >        final boolean interruptible;
905 >        final boolean timed;
906 >        boolean wasInterrupted;
907 >        long nanos;
908 >        long lastTime;
909 >        volatile Thread thread; // nulled to cancel wait
910 >        QNode next;
911 >
912 >        QNode(Phaser phaser, int phase, boolean interruptible,
913 >              boolean timed, long nanos) {
914 >            this.phaser = phaser;
915 >            this.phase = phase;
916 >            this.interruptible = interruptible;
917 >            this.nanos = nanos;
918 >            this.timed = timed;
919 >            this.lastTime = timed? System.nanoTime() : 0L;
920 >            thread = Thread.currentThread();
921 >        }
922 >
923 >        public boolean isReleasable() {
924 >            Thread t = thread;
925 >            if (t != null) {
926 >                if (phaser.getPhase() != phase)
927 >                    t = null;
928 >                else {
929 >                    if (Thread.interrupted())
930 >                        wasInterrupted = true;
931 >                    if (interruptible && wasInterrupted)
932 >                        t = null;
933 >                    else if (timed) {
934 >                        if (nanos > 0) {
935 >                            long now = System.nanoTime();
936 >                            nanos -= now - lastTime;
937 >                            lastTime = now;
938 >                        }
939 >                        if (nanos <= 0)
940 >                            t = null;
941                      }
942                  }
943 <                else if (spins <= 0)
944 <                    node = new QNode();
945 <                else
879 <                    --spins;
943 >                if (t != null)
944 >                    return false;
945 >                thread = null;
946              }
947 <            if (node != null)
882 <                node.thread = null;
883 <            if (interrupted)
884 <                throw new InterruptedException();
885 <            if (p == phase && (p = getPhase()) == phase)
886 <                throw new TimeoutException();
947 >            return true;
948          }
949 <        releaseWaiters(phase);
950 <        return p;
949 >
950 >        public boolean block() {
951 >            if (isReleasable())
952 >                return true;
953 >            else if (!timed)
954 >                LockSupport.park(this);
955 >            else if (nanos > 0)
956 >                LockSupport.parkNanos(this, nanos);
957 >            return isReleasable();
958 >        }
959 >
960 >        void signal() {
961 >            Thread t = thread;
962 >            if (t != null) {
963 >                thread = null;
964 >                LockSupport.unpark(t);
965 >            }
966 >        }
967 >
968 >        void onRelease() { // actions upon return from internalAwaitAdvance
969 >            if (!interruptible && wasInterrupted)
970 >                Thread.currentThread().interrupt();
971 >            if (thread != null)
972 >                thread = null;
973 >        }
974 >
975      }
976  
977 <    // Temporary Unsafe mechanics for preliminary release
977 >    // Unsafe mechanics
978  
979 <    static final Unsafe _unsafe;
980 <    static final long stateOffset;
979 >    private static final sun.misc.Unsafe UNSAFE = getUnsafe();
980 >    private static final long stateOffset =
981 >        objectFieldOffset("state", Phaser.class);
982  
983 <    static {
983 >    private static long objectFieldOffset(String field, Class<?> klazz) {
984          try {
985 <            if (Phaser.class.getClassLoader() != null) {
986 <                Field f = Unsafe.class.getDeclaredField("theUnsafe");
987 <                f.setAccessible(true);
988 <                _unsafe = (Unsafe)f.get(null);
989 <            }
990 <            else
905 <                _unsafe = Unsafe.getUnsafe();
906 <            stateOffset = _unsafe.objectFieldOffset
907 <                (Phaser.class.getDeclaredField("state"));
908 <        } catch (Exception e) {
909 <            throw new RuntimeException("Could not initialize intrinsics", e);
985 >            return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
986 >        } catch (NoSuchFieldException e) {
987 >            // Convert Exception to corresponding Error
988 >            NoSuchFieldError error = new NoSuchFieldError(field);
989 >            error.initCause(e);
990 >            throw error;
991          }
992      }
993  
994 <    final boolean casState(long cmp, long val) {
995 <        return _unsafe.compareAndSwapLong(this, stateOffset, cmp, val);
994 >    /**
995 >     * Returns a sun.misc.Unsafe.  Suitable for use in a 3rd party package.
996 >     * Replace with a simple call to Unsafe.getUnsafe when integrating
997 >     * into a jdk.
998 >     *
999 >     * @return a sun.misc.Unsafe
1000 >     */
1001 >    private static sun.misc.Unsafe getUnsafe() {
1002 >        try {
1003 >            return sun.misc.Unsafe.getUnsafe();
1004 >        } catch (SecurityException se) {
1005 >            try {
1006 >                return java.security.AccessController.doPrivileged
1007 >                    (new java.security
1008 >                     .PrivilegedExceptionAction<sun.misc.Unsafe>() {
1009 >                        public sun.misc.Unsafe run() throws Exception {
1010 >                            java.lang.reflect.Field f = sun.misc
1011 >                                .Unsafe.class.getDeclaredField("theUnsafe");
1012 >                            f.setAccessible(true);
1013 >                            return (sun.misc.Unsafe) f.get(null);
1014 >                        }});
1015 >            } catch (java.security.PrivilegedActionException e) {
1016 >                throw new RuntimeException("Could not initialize intrinsics",
1017 >                                           e.getCause());
1018 >            }
1019 >        }
1020      }
1021   }

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines