17 |
|
* {@link java.util.concurrent.CountDownLatch CountDownLatch} |
18 |
|
* but supporting more flexible usage. |
19 |
|
* |
20 |
< |
* <ul> |
21 |
< |
* |
22 |
< |
* <li> The number of parties <em>registered</em> to synchronize on a |
23 |
< |
* phaser may vary over time. Tasks may be registered at any time |
24 |
< |
* (using methods {@link #register}, {@link #bulkRegister}, or forms |
25 |
< |
* of constructors establishing initial numbers of parties), and may |
26 |
< |
* optionally be deregistered upon any arrival (using {@link |
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 |
30 |
|
* (However, you can introduce such bookkeeping by subclassing this |
31 |
|
* class.) |
32 |
|
* |
33 |
< |
* <li> Each generation has an associated phase number. The phase |
34 |
< |
* number starts at zero, amd advances when all parties arrive at the |
35 |
< |
* barrier, wrapping around to zero after reaching {@code |
36 |
< |
* Integer.MAX_VALUE}. |
37 |
< |
* |
38 |
< |
* <li> Like a {@code CyclicBarrier}, a phaser may be repeatedly |
39 |
< |
* awaited. Method {@link #arriveAndAwaitAdvance} has effect |
40 |
< |
* analogous to {@link java.util.concurrent.CyclicBarrier#await |
41 |
< |
* CyclicBarrier.await}. However, phasers separate two aspects of |
42 |
< |
* coordination, which may also 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 {@link #arrive} and |
48 |
< |
* {@link #arriveAndDeregister} do not block, but return |
49 |
< |
* an associated <em>arrival phase number</em>; |
50 |
< |
* that is, the phase number of the barrier to which the |
51 |
< |
* arrival applied. |
52 |
< |
* |
53 |
< |
* <li> Awaiting others. Method {@link #awaitAdvance} requires an |
54 |
< |
* argument indicating an arrival phase number, and returns |
55 |
< |
* when the 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 |
< |
* <li> Barrier actions, performed by the task triggering a phase |
78 |
< |
* advance, are arranged by overriding method {@link #onAdvance(int, |
79 |
< |
* int)}, which also controls termination. Overriding this method is |
80 |
< |
* similar to, but more flexible than, providing a barrier action to a |
81 |
< |
* {@code CyclicBarrier}. |
82 |
< |
* |
83 |
< |
* <li> Phasers may enter a <em>termination</em> state in which all |
65 |
< |
* actions immediately return without updating phaser state or waiting |
66 |
< |
* for advance, and indicating (via a negative phase value) that |
67 |
< |
* execution is complete. Termination is triggered when an invocation |
68 |
< |
* of {@code onAdvance} returns {@code true}. When a phaser is |
69 |
< |
* 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 {@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 |
< |
* {@code CyclicBarrier}, 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 |
86 |
< |
* within handlers of those exceptions, often after invoking |
87 |
< |
* {@code forceTermination}. |
88 |
< |
* |
89 |
< |
* <li>Phasers may be used to coordinate tasks executing in a {@link |
90 |
< |
* ForkJoinPool}, which will ensure sufficient parallelism to execute |
91 |
< |
* tasks when others are blocked waiting for a phase to advance. |
92 |
< |
* |
93 |
< |
* <li>The current state of a phaser may be monitored. At any given |
94 |
< |
* moment there are {@link #getRegisteredParties}, where {@link |
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 |
< |
* arrive, the phase advances. Method {@link #toString} returns |
104 |
< |
* snapshots of these state queries in a form convenient for |
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 |
|
* |
101 |
– |
* </ul> |
102 |
– |
* |
109 |
|
* <p><b>Sample usages:</b> |
110 |
|
* |
111 |
|
* <p>A {@code Phaser} may be used instead of a {@code CountDownLatch} |
137 |
|
* <pre> {@code |
138 |
|
* void startTasks(List<Runnable> tasks, final int iterations) { |
139 |
|
* final Phaser phaser = new Phaser() { |
140 |
< |
* public boolean onAdvance(int phase, int registeredParties) { |
140 |
> |
* protected boolean onAdvance(int phase, int registeredParties) { |
141 |
|
* return phase >= iterations || registeredParties == 0; |
142 |
|
* } |
143 |
|
* }; |
156 |
|
* phaser.arriveAndDeregister(); // deregister self, don't wait |
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(); |
166 |
+ |
* }</pre> |
167 |
+ |
* |
168 |
+ |
* Related constructions may be used to await particular phase numbers |
169 |
+ |
* in contexts where you are sure that the phase will never wrap around |
170 |
+ |
* {@code Integer.MAX_VALUE}. For example: |
171 |
+ |
* |
172 |
+ |
* <pre> {@code |
173 |
+ |
* void awaitPhase(Phaser phaser, int phase) { |
174 |
+ |
* int p = phaser.register(); // assumes caller not already registered |
175 |
+ |
* while (p < phase) { |
176 |
+ |
* if (phaser.isTerminated()) |
177 |
+ |
* // ... deal with unexpected termination |
178 |
+ |
* else |
179 |
+ |
* p = phaser.arriveAndAwaitAdvance(); |
180 |
+ |
* } |
181 |
+ |
* phaser.arriveAndDeregister(); |
182 |
+ |
* } |
183 |
+ |
* }</pre> |
184 |
+ |
* |
185 |
+ |
* |
186 |
|
* <p>To create a set of tasks using a tree of phasers, |
187 |
|
* you could use code of the following form, assuming a |
188 |
|
* Task class with a constructor accepting a phaser that |
189 |
|
* it registers for upon construction: |
190 |
|
* <pre> {@code |
191 |
< |
* void build(Task[] actions, int lo, int hi, Phaser b) { |
192 |
< |
* int step = (hi - lo) / TASKS_PER_PHASER; |
193 |
< |
* if (step > 1) { |
194 |
< |
* int i = lo; |
195 |
< |
* while (i < hi) { |
163 |
< |
* int r = Math.min(i + step, hi); |
164 |
< |
* build(actions, i, r, new Phaser(b)); |
165 |
< |
* i = r; |
191 |
> |
* void build(Task[] actions, int lo, int hi, Phaser ph) { |
192 |
> |
* if (hi - lo > TASKS_PER_PHASER) { |
193 |
> |
* for (int i = lo; i < hi; i += TASKS_PER_PHASER) { |
194 |
> |
* int j = Math.min(i + TASKS_PER_PHASER, hi); |
195 |
> |
* build(actions, i, j, new Phaser(ph)); |
196 |
|
* } |
197 |
|
* } else { |
198 |
|
* for (int i = lo; i < hi; ++i) |
199 |
< |
* actions[i] = new Task(b); |
200 |
< |
* // assumes new Task(b) performs b.register() |
199 |
> |
* actions[i] = new Task(ph); |
200 |
> |
* // assumes new Task(ph) performs ph.register() |
201 |
|
* } |
202 |
|
* } |
203 |
|
* // .. initially called, for n tasks via |
246 |
|
*/ |
247 |
|
private volatile long state; |
248 |
|
|
219 |
– |
private static final int ushortBits = 16; |
249 |
|
private static final int ushortMask = 0xffff; |
250 |
|
private static final int phaseMask = 0x7fffffff; |
251 |
|
|
455 |
|
|
456 |
|
/** |
457 |
|
* Arrives at the barrier, but does not wait for others. (You can |
458 |
< |
* in turn wait for others via {@link #awaitAdvance}). |
458 |
> |
* in turn wait for others via {@link #awaitAdvance}). It is an |
459 |
> |
* unenforced usage error for an unregistered party to invoke this |
460 |
> |
* method. |
461 |
|
* |
462 |
|
* @return the arrival phase number, or a negative value if terminated |
463 |
|
* @throws IllegalStateException if not terminated and the number |
509 |
|
* required to trip the barrier in future phases. If this phaser |
510 |
|
* has a parent, and deregistration causes this phaser to have |
511 |
|
* zero parties, this phaser also arrives at and is deregistered |
512 |
< |
* from its parent. |
512 |
> |
* from its parent. It is an unenforced usage error for an |
513 |
> |
* unregistered party to invoke this method. |
514 |
|
* |
515 |
|
* @return the arrival phase number, or a negative value if terminated |
516 |
|
* @throws IllegalStateException if not terminated and the number |
566 |
|
* interruption or timeout, you can arrange this with an analogous |
567 |
|
* construction using one of the other forms of the awaitAdvance |
568 |
|
* method. If instead you need to deregister upon arrival use |
569 |
< |
* {@code arriveAndDeregister}. |
569 |
> |
* {@code arriveAndDeregister}. It is an unenforced usage error |
570 |
> |
* for an unregistered party to invoke this method. |
571 |
|
* |
572 |
|
* @return the arrival phase number, or a negative number if terminated |
573 |
|
* @throws IllegalStateException if not terminated and the number |
581 |
|
* Awaits the phase of the barrier to advance from the given phase |
582 |
|
* value, returning immediately if the current phase of the |
583 |
|
* barrier is not equal to the given phase value or this barrier |
584 |
< |
* is terminated. |
584 |
> |
* is terminated. It is an unenforced usage error for an |
585 |
> |
* unregistered party to invoke this method. |
586 |
|
* |
587 |
|
* @param phase an arrival phase number, or negative value if |
588 |
|
* terminated; this argument is normally the value returned by a |
605 |
|
|
606 |
|
/** |
607 |
|
* Awaits the phase of the barrier to advance from the given phase |
608 |
< |
* value, throwing {@code InterruptedException} if interrupted while |
609 |
< |
* waiting, or returning immediately if the current phase of the |
610 |
< |
* barrier is not equal to the given phase value or this barrier |
611 |
< |
* is terminated. |
608 |
> |
* value, throwing {@code InterruptedException} if interrupted |
609 |
> |
* while waiting, or returning immediately if the current phase of |
610 |
> |
* the barrier is not equal to the given phase value or this |
611 |
> |
* barrier is terminated. It is an unenforced usage error for an |
612 |
> |
* unregistered party to invoke this method. |
613 |
|
* |
614 |
|
* @param phase an arrival phase number, or negative value if |
615 |
|
* terminated; this argument is normally the value returned by a |
633 |
|
|
634 |
|
/** |
635 |
|
* Awaits the phase of the barrier to advance from the given phase |
636 |
< |
* value or the given timeout to elapse, throwing |
637 |
< |
* {@code InterruptedException} if interrupted while waiting, or |
638 |
< |
* returning immediately if the current phase of the barrier is not |
639 |
< |
* equal to the given phase value or this barrier is terminated. |
636 |
> |
* value or the given timeout to elapse, throwing {@code |
637 |
> |
* InterruptedException} if interrupted while waiting, or |
638 |
> |
* returning immediately if the current phase of the barrier is |
639 |
> |
* not equal to the given phase value or this barrier is |
640 |
> |
* terminated. It is an unenforced usage error for an |
641 |
> |
* unregistered party to invoke this method. |
642 |
|
* |
643 |
|
* @param phase an arrival phase number, or negative value if |
644 |
|
* terminated; this argument is normally the value returned by a |
759 |
|
} |
760 |
|
|
761 |
|
/** |
762 |
< |
* Overridable method to perform an action upon phase advance, and |
763 |
< |
* to control termination. This method is invoked whenever the |
764 |
< |
* barrier is tripped (and thus all other waiting parties are |
765 |
< |
* dormant). If it returns {@code true}, then, rather than advance |
766 |
< |
* the phase number, this barrier will be set to a final |
767 |
< |
* termination state, and subsequent calls to {@link #isTerminated} |
768 |
< |
* will return true. |
762 |
> |
* Overridable method to perform an action upon impending phase |
763 |
> |
* advance, and to control termination. This method is invoked |
764 |
> |
* upon arrival of the party tripping the barrier (when all other |
765 |
> |
* waiting parties are dormant). If this method returns {@code |
766 |
> |
* true}, then, rather than advance the phase number, this barrier |
767 |
> |
* will be set to a final termination state, and subsequent calls |
768 |
> |
* to {@link #isTerminated} will return true. Any (unchecked) |
769 |
> |
* Exception or Error thrown by an invocation of this method is |
770 |
> |
* propagated to the party attempting to trip the barrier, in |
771 |
> |
* which case no advance occurs. |
772 |
> |
* |
773 |
> |
* <p>The arguments to this method provide the state of the phaser |
774 |
> |
* prevailing for the current transition. (When called from within |
775 |
> |
* an implementation of {@code onAdvance} the values returned by |
776 |
> |
* methods such as {@code getPhase} may or may not reliably |
777 |
> |
* indicate the state to which this transition applies.) |
778 |
|
* |
779 |
|
* <p>The default version returns {@code true} when the number of |
780 |
|
* registered parties is zero. Normally, overrides that arrange |
782 |
|
* property. |
783 |
|
* |
784 |
|
* <p>You may override this method to perform an action with side |
785 |
< |
* effects visible to participating tasks, but it is in general |
786 |
< |
* only sensible to do so in designs where all parties register |
787 |
< |
* before any arrive, and all {@link #awaitAdvance} at each phase. |
788 |
< |
* Otherwise, you cannot ensure lack of interference from other |
789 |
< |
* parties during the invocation of this method. |
785 |
> |
* effects visible to participating tasks, but doing so requires |
786 |
> |
* care: Method {@code onAdvance} may be invoked more than once |
787 |
> |
* per transition. Further, unless all parties register before |
788 |
> |
* any arrive, and all {@link #awaitAdvance} at each phase, then |
789 |
> |
* you cannot ensure lack of interference from other parties |
790 |
> |
* during the invocation of this method. |
791 |
|
* |
792 |
|
* @param phase the phase number on entering the barrier |
793 |
|
* @param registeredParties the current number of registered parties |