--- jsr166/src/jsr166y/Phaser.java 2008/07/25 18:10:41 1.2
+++ jsr166/src/jsr166y/Phaser.java 2009/08/25 16:32:28 1.44
@@ -5,159 +5,260 @@
*/
package jsr166y;
-import jsr166y.forkjoin.*;
+
import java.util.concurrent.*;
-import java.util.concurrent.atomic.*;
+
+import java.util.concurrent.atomic.AtomicReference;
import java.util.concurrent.locks.LockSupport;
/**
- * A reusable synchronization barrier, similar in functionality to a
- * {@link java.util.concurrent.CyclicBarrier}, but supporting more
- * flexible usage.
+ * A reusable synchronization barrier, similar in functionality to
+ * {@link java.util.concurrent.CyclicBarrier CyclicBarrier} and
+ * {@link java.util.concurrent.CountDownLatch CountDownLatch}
+ * but supporting more flexible usage.
*
- *
+ * Registration. Unlike the case for other barriers, the
+ * number of parties registered to synchronize on a phaser
+ * may vary over time. Tasks may be registered at any time (using
+ * methods {@link #register}, {@link #bulkRegister}, or forms of
+ * constructors establishing initial numbers of parties), and
+ * optionally deregistered upon any arrival (using {@link
+ * #arriveAndDeregister}). As is the case with most basic
+ * synchronization constructs, registration and deregistration affect
+ * only internal counts; they do not establish any further internal
+ * bookkeeping, so tasks cannot query whether they are registered.
+ * (However, you can introduce such bookkeeping by subclassing this
+ * class.)
*
- *
- The number of parties synchronizing on the barrier may vary
- * over time. A task may register to be a party in a barrier at any
- * time, and may deregister upon arriving at the barrier. As is the
- * case with most basic synchronization constructs, registration
- * and deregistration affect only internal counts; they do not
- * establish any further internal bookkeeping, so tasks cannot query
- * whether they are registered.
- *
- *
- Each generation has an associated phase value, starting at
- * zero, and advancing when all parties reach the barrier (wrapping
- * around to zero after reaching Integer.MAX_VALUE).
- *
- *
- Like a CyclicBarrier, a Phaser may be repeatedly awaited.
- * Method arriveAndAwaitAdvance has effect analogous to
- * CyclicBarrier.await. However, Phasers separate two
- * aspects of coordination, that may be invoked independently:
+ *
Synchronization. Like a {@code CyclicBarrier}, a {@code
+ * Phaser} may be repeatedly awaited. Method {@link
+ * #arriveAndAwaitAdvance} has effect analogous to {@link
+ * java.util.concurrent.CyclicBarrier#await CyclicBarrier.await}. Each
+ * generation of a {@code Phaser} has an associated phase number. The
+ * phase number starts at zero, and advances when all parties arrive
+ * at the barrier, wrapping around to zero after reaching {@code
+ * Integer.MAX_VALUE}. The use of phase numbers enables independent
+ * control of actions upon arrival at a barrier and upon awaiting
+ * others, via two kinds of methods that may be invoked by any
+ * registered party:
*
*
*
- * - Arriving at a barrier. Methods arrive and
- * arriveAndDeregister do not block, but return
- * the phase value on entry to the method.
- *
- *
- Awaiting others. Method awaitAdvance requires an
- * argument indicating the entry phase, and returns when the
- * barrier advances to a new phase.
+ *
- Arrival. Methods {@link #arrive} and
+ * {@link #arriveAndDeregister} record arrival at a
+ * barrier. These methods do not block, but return an associated
+ * arrival phase number; that is, the phase number of
+ * the barrier to which the arrival applied. When the final
+ * party for a given phase arrives, an optional barrier action
+ * is performed and the phase advances. Barrier actions,
+ * performed by the party triggering a phase advance, are
+ * arranged by overriding method {@link #onAdvance(int, int)},
+ * which also controls termination. Overriding this method is
+ * similar to, but more flexible than, providing a barrier
+ * action to a {@code CyclicBarrier}.
+ *
+ *
- Waiting. Method {@link #awaitAdvance} requires an
+ * argument indicating an arrival phase number, and returns when
+ * the barrier advances to (or is already at) a different phase.
+ * Unlike similar constructions using {@code CyclicBarrier},
+ * method {@code awaitAdvance} continues to wait even if the
+ * waiting thread is interrupted. Interruptible and timeout
+ * versions are also available, but exceptions encountered while
+ * tasks wait interruptibly or with timeout do not change the
+ * state of the barrier. If necessary, you can perform any
+ * associated recovery within handlers of those exceptions,
+ * often after invoking {@code forceTermination}. Phasers may
+ * also be used by tasks executing in a {@link ForkJoinPool},
+ * which will ensure sufficient parallelism to execute tasks
+ * when others are blocked waiting for a phase to advance.
+ *
*
*
+ * Termination. A {@code Phaser} may enter a
+ * termination state in which all synchronization methods
+ * immediately return without updating phaser state or waiting for
+ * advance, and indicating (via a negative phase value) that execution
+ * is complete. Termination is triggered when an invocation of {@code
+ * onAdvance} returns {@code true}. As illustrated below, when
+ * phasers control actions with a fixed number of iterations, it is
+ * often convenient to override this method to cause termination when
+ * the current phase number reaches a threshold. Method {@link
+ * #forceTermination} is also available to abruptly release waiting
+ * threads and allow them to terminate.
*
- *
- Barrier actions, performed by the task triggering a phase
- * advance while others may be waiting, are arranged by overriding
- * method onAdvance, that also controls termination.
- *
- *
- Phasers may enter a termination state in which all
- * await actions immediately return, indicating (via a negative phase
- * value) that execution is complete. Termination is triggered by
- * executing the overridable onAdvance method that is invoked
- * each time the barrier is tripped. When a Phaser is controlling an
- * action with a fixed number of iterations, it is often convenient to
- * override this method to cause termination when the current phase
- * number reaches a threshold. Method forceTermination is
- * also available to assist recovery actions upon failure.
- *
- *
- Unlike most synchronizers, a Phaser may also be used with
- * ForkJoinTasks (as well as plain threads).
- *
- *
- By default, awaitAdvance continues to wait even if
- * the current thread is interrupted. And unlike the case in
- * CyclicBarriers, exceptions encountered while tasks wait
- * interruptibly or with timeout do not change the state of the
- * barrier. If necessary, you can perform any associated recovery
- * within handlers of those exceptions.
+ *
Tiering. Phasers may be tiered (i.e., arranged
+ * in tree structures) to reduce contention. Phasers with large
+ * numbers of parties that would otherwise experience heavy
+ * synchronization contention costs may instead be set up so that
+ * groups of sub-phasers share a common parent. This may greatly
+ * increase throughput even though it incurs greater per-operation
+ * overhead.
*
- *
+ * Monitoring. While synchronization methods may be invoked
+ * only by registered parties, the current state of a phaser may be
+ * monitored by any caller. At any given moment there are {@link
+ * #getRegisteredParties} parties in total, of which {@link
+ * #getArrivedParties} have arrived at the current phase ({@link
+ * #getPhase}). When the remaining ({@link #getUnarrivedParties})
+ * parties arrive, the phase advances. The values returned by these
+ * methods may reflect transient states and so are not in general
+ * useful for synchronization control. Method {@link #toString}
+ * returns snapshots of these state queries in a form convenient for
+ * informal monitoring.
+ *
+ *
Sample usages:
+ *
+ *
A {@code Phaser} may be used instead of a {@code CountDownLatch}
+ * to control a one-shot action serving a variable number of
+ * parties. The typical idiom is for the method setting this up to
+ * first register, then start the actions, then deregister, as in:
+ *
+ *
{@code
+ * void runTasks(List tasks) {
+ * final Phaser phaser = new Phaser(1); // "1" to register self
+ * // create and start threads
+ * for (Runnable task : tasks) {
+ * phaser.register();
+ * new Thread() {
+ * public void run() {
+ * phaser.arriveAndAwaitAdvance(); // await all creation
+ * task.run();
+ * }
+ * }.start();
+ * }
+ *
+ * // allow threads to start and deregister self
+ * phaser.arriveAndDeregister();
+ * }}
+ *
+ * One way to cause a set of threads to repeatedly perform actions
+ * for a given number of iterations is to override {@code onAdvance}:
+ *
+ *
{@code
+ * void startTasks(List tasks, final int iterations) {
+ * final Phaser phaser = new Phaser() {
+ * protected boolean onAdvance(int phase, int registeredParties) {
+ * return phase >= iterations || registeredParties == 0;
+ * }
+ * };
+ * phaser.register();
+ * for (Runnable task : tasks) {
+ * phaser.register();
+ * new Thread() {
+ * public void run() {
+ * do {
+ * task.run();
+ * phaser.arriveAndAwaitAdvance();
+ * } while(!phaser.isTerminated();
+ * }
+ * }.start();
+ * }
+ * phaser.arriveAndDeregister(); // deregister self, don't wait
+ * }}
*
- * Sample usage:
+ * If the main task must later await termination, it
+ * may re-register and then execute a similar loop:
+ *
{@code
+ * // ...
+ * phaser.register();
+ * while (!phaser.isTerminated())
+ * phaser.arriveAndAwaitAdvance();
+ * }
*
- * [todo: non-FJ example]
+ * Related constructions may be used to await particular phase numbers
+ * in contexts where you are sure that the phase will never wrap around
+ * {@code Integer.MAX_VALUE}. For example:
*
- *
A Phaser may be used to support a style of programming in
- * which a task waits for others to complete, without otherwise
- * needing to keep track of which tasks it is waiting for. This is
- * similar to the "sync" construct in Cilk and "clocks" in X10.
- * Special constructions based on such barriers are available using
- * the LinkedAsyncAction and CyclicAction classes,
- * but they can be useful in other contexts as well. For a simple
- * (but not very useful) example, here is a variant of Fibonacci:
- *
- *
- * class BarrierFibonacci extends RecursiveAction {
- * int argument, result;
- * final Phaser parentBarrier;
- * BarrierFibonacci(int n, Phaser parentBarrier) {
- * this.argument = n;
- * this.parentBarrier = parentBarrier;
- * parentBarrier.register();
+ * {@code
+ * void awaitPhase(Phaser phaser, int phase) {
+ * int p = phaser.register(); // assumes caller not already registered
+ * while (p < phase) {
+ * if (phaser.isTerminated())
+ * // ... deal with unexpected termination
+ * else
+ * p = phaser.arriveAndAwaitAdvance();
+ * }
+ * phaser.arriveAndDeregister();
* }
- * protected void compute() {
- * int n = argument;
- * if (n <= 1)
- * result = n;
- * else {
- * Phaser childBarrier = new Phaser(1);
- * BarrierFibonacci f1 = new BarrierFibonacci(n - 1, childBarrier);
- * BarrierFibonacci f2 = new BarrierFibonacci(n - 2, childBarrier);
- * f1.fork();
- * f2.fork();
- * childBarrier.arriveAndAwait();
- * result = f1.result + f2.result;
+ * }
+ *
+ *
+ * To create a set of tasks using a tree of phasers,
+ * you could use code of the following form, assuming a
+ * Task class with a constructor accepting a phaser that
+ * it registers for upon construction:
+ *
{@code
+ * void build(Task[] actions, int lo, int hi, Phaser ph) {
+ * if (hi - lo > TASKS_PER_PHASER) {
+ * for (int i = lo; i < hi; i += TASKS_PER_PHASER) {
+ * int j = Math.min(i + TASKS_PER_PHASER, hi);
+ * build(actions, i, j, new Phaser(ph));
* }
- * parentBarrier.arriveAndDeregister();
+ * } else {
+ * for (int i = lo; i < hi; ++i)
+ * actions[i] = new Task(ph);
+ * // assumes new Task(ph) performs ph.register()
* }
* }
+ * // .. initially called, for n tasks via
+ * build(new Task[n], 0, n, new Phaser());}
+ *
+ * The best value of {@code TASKS_PER_PHASER} depends mainly on
+ * expected barrier synchronization rates. A value as low as four may
+ * be appropriate for extremely small per-barrier task bodies (thus
+ * high rates), or up to hundreds for extremely large ones.
+ *
*
*
* Implementation notes: This implementation restricts the
- * maximum number of parties to 65535. Attempts to register
- * additional parties result in IllegalStateExceptions.
+ * maximum number of parties to 65535. Attempts to register additional
+ * parties result in {@code IllegalStateException}. However, you can and
+ * should create tiered phasers to accommodate arbitrarily large sets
+ * of participants.
+ *
+ * @since 1.7
+ * @author Doug Lea
*/
public class Phaser {
/*
* This class implements an extension of X10 "clocks". Thanks to
- * Vijay Saraswat for the idea of applying it to ForkJoinTasks,
- * and to Vivek Sarkar for enhancements to extend functionality.
+ * Vijay Saraswat for the idea, and to Vivek Sarkar for
+ * enhancements to extend functionality.
*/
/**
* Barrier state representation. Conceptually, a barrier contains
* four values:
- *
+ *
* * parties -- the number of parties to wait (16 bits)
* * unarrived -- the number of parties yet to hit barrier (16 bits)
* * phase -- the generation of the barrier (31 bits)
* * terminated -- set if barrier is terminated (1 bit)
*
* However, to efficiently maintain atomicity, these values are
- * packed into a single AtomicLong. Termination uses the sign bit
- * of 32 bit representation of phase, so phase is set to -1 on
- * termination.
- */
- private final AtomicLong state;
-
- /**
- * Head of Treiber stack for waiting nonFJ threads.
+ * packed into a single (atomic) long. Termination uses the sign
+ * bit of 32 bit representation of phase, so phase is set to -1 on
+ * termination. Good performance relies on keeping state decoding
+ * and encoding simple, and keeping race windows short.
+ *
+ * Note: there are some cheats in arrive() that rely on unarrived
+ * count being lowest 16 bits.
*/
- private final AtomicReference head = new AtomicReference();
+ private volatile long state;
- private static final int ushortBits = 16;
- private static final int ushortMask = (1 << ushortBits) - 1;
- private static final int phaseMask = 0x7fffffff;
+ private static final int ushortMask = 0xffff;
+ private static final int phaseMask = 0x7fffffff;
private static int unarrivedOf(long s) {
- return (int)(s & ushortMask);
+ return (int) (s & ushortMask);
}
private static int partiesOf(long s) {
- return (int)(s & (ushortMask << 16)) >>> 16;
+ return ((int) s) >>> 16;
}
private static int phaseOf(long s) {
- return (int)(s >>> 32);
+ return (int) (s >>> 32);
}
private static int arrivedOf(long s) {
@@ -165,531 +266,778 @@ public class Phaser {
}
private static long stateFor(int phase, int parties, int unarrived) {
- return (((long)phase) << 32) | ((parties << 16) | unarrived);
+ return ((((long) phase) << 32) | (((long) parties) << 16) |
+ (long) unarrived);
}
- private static IllegalStateException badBounds(int parties, int unarrived) {
- return new IllegalStateException("Attempt to set " + unarrived +
- " unarrived of " + parties + " parties");
+ private static long trippedStateFor(int phase, int parties) {
+ long lp = (long) parties;
+ return (((long) phase) << 32) | (lp << 16) | lp;
}
/**
- * Creates a new Phaser without any initially registered parties,
- * and initial phase number 0.
+ * Returns message string for bad bounds exceptions.
+ */
+ private static String badBounds(int parties, int unarrived) {
+ return ("Attempt to set " + unarrived +
+ " unarrived of " + parties + " parties");
+ }
+
+ /**
+ * The parent of this phaser, or null if none
+ */
+ private final Phaser parent;
+
+ /**
+ * The root of phaser tree. Equals this if not in a tree. Used to
+ * support faster state push-down.
+ */
+ private final Phaser root;
+
+ // Wait queues
+
+ /**
+ * Heads of Treiber stacks for waiting threads. To eliminate
+ * contention while releasing some threads while adding others, we
+ * use two of them, alternating across even and odd phases.
+ */
+ private final AtomicReference evenQ = new AtomicReference();
+ private final AtomicReference oddQ = new AtomicReference();
+
+ private AtomicReference queueFor(int phase) {
+ return ((phase & 1) == 0) ? evenQ : oddQ;
+ }
+
+ /**
+ * Returns current state, first resolving lagged propagation from
+ * root if necessary.
+ */
+ private long getReconciledState() {
+ return (parent == null) ? state : reconcileState();
+ }
+
+ /**
+ * Recursively resolves state.
+ */
+ private long reconcileState() {
+ Phaser p = parent;
+ long s = state;
+ if (p != null) {
+ while (unarrivedOf(s) == 0 && phaseOf(s) != phaseOf(root.state)) {
+ long parentState = p.getReconciledState();
+ int parentPhase = phaseOf(parentState);
+ int phase = phaseOf(s = state);
+ if (phase != parentPhase) {
+ long next = trippedStateFor(parentPhase, partiesOf(s));
+ if (casState(s, next)) {
+ releaseWaiters(phase);
+ s = next;
+ }
+ }
+ }
+ }
+ return s;
+ }
+
+ /**
+ * Creates a new phaser without any initially registered parties,
+ * initial phase number 0, and no parent. Any thread using this
+ * phaser will need to first register for it.
*/
public Phaser() {
- state = new AtomicLong(stateFor(0, 0, 0));
+ this(null);
}
/**
- * Creates a new Phaser with the given numbers of registered
- * unarrived parties and initial phase number 0.
- * @param parties the number of parties required to trip barrier.
+ * Creates a new phaser with the given numbers of registered
+ * unarrived parties, initial phase number 0, and no parent.
+ *
+ * @param parties the number of parties required to trip barrier
* @throws IllegalArgumentException if parties less than zero
- * or greater than the maximum number of parties supported.
+ * or greater than the maximum number of parties supported
*/
public Phaser(int parties) {
+ this(null, parties);
+ }
+
+ /**
+ * Creates a new phaser with the given parent, without any
+ * initially registered parties. If parent is non-null this phaser
+ * is registered with the parent and its initial phase number is
+ * the same as that of parent phaser.
+ *
+ * @param parent the parent phaser
+ */
+ public Phaser(Phaser parent) {
+ int phase = 0;
+ this.parent = parent;
+ if (parent != null) {
+ this.root = parent.root;
+ phase = parent.register();
+ }
+ else
+ this.root = this;
+ this.state = trippedStateFor(phase, 0);
+ }
+
+ /**
+ * Creates a new phaser with the given parent and numbers of
+ * registered unarrived parties. If parent is non-null, this phaser
+ * is registered with the parent and its initial phase number is
+ * the same as that of parent phaser.
+ *
+ * @param parent the parent phaser
+ * @param parties the number of parties required to trip barrier
+ * @throws IllegalArgumentException if parties less than zero
+ * or greater than the maximum number of parties supported
+ */
+ public Phaser(Phaser parent, int parties) {
if (parties < 0 || parties > ushortMask)
throw new IllegalArgumentException("Illegal number of parties");
- state = new AtomicLong(stateFor(0, parties, parties));
+ int phase = 0;
+ this.parent = parent;
+ if (parent != null) {
+ this.root = parent.root;
+ phase = parent.register();
+ }
+ else
+ this.root = this;
+ this.state = trippedStateFor(phase, parties);
}
/**
* Adds a new unarrived party to this phaser.
- * @return the current barrier phase number upon registration
+ *
+ * @return the arrival phase number to which this registration applied
+ * @throws IllegalStateException if attempting to register more
+ * than the maximum supported number of parties
+ */
+ public int register() {
+ return doRegister(1);
+ }
+
+ /**
+ * Adds the given number of new unarrived parties to this phaser.
+ *
+ * @param parties the number of parties required to trip barrier
+ * @return the arrival phase number to which this registration applied
* @throws IllegalStateException if attempting to register more
- * than the maximum supported number of parties.
+ * than the maximum supported number of parties
*/
- public int register() { // increment both parties and unarrived
- final AtomicLong state = this.state;
+ public int bulkRegister(int parties) {
+ if (parties < 0)
+ throw new IllegalArgumentException();
+ if (parties == 0)
+ return getPhase();
+ return doRegister(parties);
+ }
+
+ /**
+ * Shared code for register, bulkRegister
+ */
+ private int doRegister(int registrations) {
+ int phase;
for (;;) {
- long s = state.get();
- int phase = phaseOf(s);
- int parties = partiesOf(s) + 1;
- int unarrived = unarrivedOf(s) + 1;
+ long s = getReconciledState();
+ phase = phaseOf(s);
+ int unarrived = unarrivedOf(s) + registrations;
+ int parties = partiesOf(s) + registrations;
+ if (phase < 0)
+ break;
if (parties > ushortMask || unarrived > ushortMask)
- throw badBounds(parties, unarrived);
- if (state.compareAndSet(s, stateFor(phase, parties, unarrived)))
- return phase;
+ throw new IllegalStateException(badBounds(parties, unarrived));
+ if (phase == phaseOf(root.state) &&
+ casState(s, stateFor(phase, parties, unarrived)))
+ break;
}
+ return phase;
}
/**
* Arrives at the barrier, but does not wait for others. (You can
- * in turn wait for others via {@link #awaitAdvance}).
+ * in turn wait for others via {@link #awaitAdvance}). It is an
+ * unenforced usage error for an unregistered party to invoke this
+ * method.
*
- * @return the current barrier phase number upon entry to
- * this method, or a negative value if terminated;
- * @throws IllegalStateException if the number of unarrived
- * parties would become negative.
+ * @return the arrival phase number, or a negative value if terminated
+ * @throws IllegalStateException if not terminated and the number
+ * of unarrived parties would become negative
*/
- public int arrive() { // decrement unarrived. If zero, trip
- final AtomicLong state = this.state;
+ public int arrive() {
+ int phase;
for (;;) {
- long s = state.get();
- int phase = phaseOf(s);
+ long s = state;
+ phase = phaseOf(s);
+ if (phase < 0)
+ break;
int parties = partiesOf(s);
int unarrived = unarrivedOf(s) - 1;
- if (unarrived < 0)
- throw badBounds(parties, unarrived);
- if (unarrived == 0 && phase >= 0) {
- trip(phase, parties);
- return phase;
+ if (unarrived > 0) { // Not the last arrival
+ if (casState(s, s - 1)) // s-1 adds one arrival
+ break;
}
- if (state.compareAndSet(s, stateFor(phase, parties, unarrived)))
- return phase;
+ else if (unarrived == 0) { // the last arrival
+ Phaser par = parent;
+ if (par == null) { // directly trip
+ if (casState
+ (s,
+ trippedStateFor(onAdvance(phase, parties) ? -1 :
+ ((phase + 1) & phaseMask), parties))) {
+ releaseWaiters(phase);
+ break;
+ }
+ }
+ else { // cascade to parent
+ if (casState(s, s - 1)) { // zeroes unarrived
+ par.arrive();
+ reconcileState();
+ break;
+ }
+ }
+ }
+ else if (phase != phaseOf(root.state)) // or if unreconciled
+ reconcileState();
+ else
+ throw new IllegalStateException(badBounds(parties, unarrived));
}
+ return phase;
}
/**
- * Arrives at the barrier, and deregisters from it, without
- * waiting for others.
+ * Arrives at the barrier and deregisters from it without waiting
+ * for others. Deregistration reduces the number of parties
+ * required to trip the barrier in future phases. If this phaser
+ * has a parent, and deregistration causes this phaser to have
+ * zero parties, this phaser also arrives at and is deregistered
+ * from its parent. It is an unenforced usage error for an
+ * unregistered party to invoke this method.
*
- * @return the current barrier phase number upon entry to
- * this method, or a negative value if terminated;
- * @throws IllegalStateException if the number of registered or
- * unarrived parties would become negative.
- */
- public int arriveAndDeregister() { // Same as arrive, plus decrement parties
- final AtomicLong state = this.state;
+ * @return the arrival phase number, or a negative value if terminated
+ * @throws IllegalStateException if not terminated and the number
+ * of registered or unarrived parties would become negative
+ */
+ public int arriveAndDeregister() {
+ // similar code to arrive, but too different to merge
+ Phaser par = parent;
+ int phase;
for (;;) {
- long s = state.get();
- int phase = phaseOf(s);
+ long s = state;
+ phase = phaseOf(s);
+ if (phase < 0)
+ break;
int parties = partiesOf(s) - 1;
int unarrived = unarrivedOf(s) - 1;
- if (parties < 0 || unarrived < 0)
- throw badBounds(parties, unarrived);
- if (unarrived == 0 && phase >= 0) {
- trip(phase, parties);
- return phase;
+ if (parties >= 0) {
+ if (unarrived > 0 || (unarrived == 0 && par != null)) {
+ if (casState
+ (s,
+ stateFor(phase, parties, unarrived))) {
+ if (unarrived == 0) {
+ par.arriveAndDeregister();
+ reconcileState();
+ }
+ break;
+ }
+ continue;
+ }
+ if (unarrived == 0) {
+ if (casState
+ (s,
+ trippedStateFor(onAdvance(phase, parties) ? -1 :
+ ((phase + 1) & phaseMask), parties))) {
+ releaseWaiters(phase);
+ break;
+ }
+ continue;
+ }
+ if (par != null && phase != phaseOf(root.state)) {
+ reconcileState();
+ continue;
+ }
}
- if (state.compareAndSet(s, stateFor(phase, parties, unarrived)))
- return phase;
+ throw new IllegalStateException(badBounds(parties, unarrived));
}
+ return phase;
}
/**
- * Arrives at the barrier and awaits others. Unlike other arrival
- * methods, this method returns the arrival index of the
- * caller. The caller tripping the barrier returns zero, the
- * previous caller 1, and so on.
- * @return the arrival index
- * @throws IllegalStateException if the number of unarrived
- * parties would become negative.
+ * Arrives at the barrier and awaits others. Equivalent in effect
+ * to {@code awaitAdvance(arrive())}. If you need to await with
+ * interruption or timeout, you can arrange this with an analogous
+ * construction using one of the other forms of the awaitAdvance
+ * method. If instead you need to deregister upon arrival use
+ * {@code arriveAndDeregister}. It is an unenforced usage error
+ * for an unregistered party to invoke this method.
+ *
+ * @return the arrival phase number, or a negative number if terminated
+ * @throws IllegalStateException if not terminated and the number
+ * of unarrived parties would become negative
*/
public int arriveAndAwaitAdvance() {
- final AtomicLong state = this.state;
- for (;;) {
- long s = state.get();
- int phase = phaseOf(s);
- int parties = partiesOf(s);
- int unarrived = unarrivedOf(s) - 1;
- if (unarrived < 0)
- throw badBounds(parties, unarrived);
- if (unarrived == 0 && phase >= 0) {
- trip(phase, parties);
- return 0;
- }
- if (state.compareAndSet(s, stateFor(phase, parties, unarrived))) {
- awaitAdvance(phase);
- return unarrived;
- }
- }
+ return awaitAdvance(arrive());
}
/**
- * Awaits the phase of the barrier to advance from the given
- * value, or returns immediately if this barrier is terminated.
- * @param phase the phase on entry to this method
- * @return the phase on exit from this method
+ * Awaits the phase of the barrier to advance from the given phase
+ * value, returning immediately if the current phase of the
+ * barrier is not equal to the given phase value or this barrier
+ * is terminated. It is an unenforced usage error for an
+ * unregistered party to invoke this method.
+ *
+ * @param phase an arrival phase number, or negative value if
+ * terminated; this argument is normally the value returned by a
+ * previous call to {@code arrive} or its variants
+ * @return the next arrival phase number, or a negative value
+ * if terminated or argument is negative
*/
public int awaitAdvance(int phase) {
if (phase < 0)
return phase;
- Thread current = Thread.currentThread();
- if (current instanceof ForkJoinWorkerThread)
- return helpingWait(phase);
- if (untimedWait(current, phase, false))
- current.interrupt();
- return phaseOf(state.get());
+ long s = getReconciledState();
+ int p = phaseOf(s);
+ if (p != phase)
+ return p;
+ if (unarrivedOf(s) == 0 && parent != null)
+ parent.awaitAdvance(phase);
+ // Fall here even if parent waited, to reconcile and help release
+ return untimedWait(phase);
}
/**
- * Awaits the phase of the barrier to advance from the given
- * value, or returns immediately if this barrier is terminated, or
- * throws InterruptedException if interrupted while waiting.
- * @param phase the phase on entry to this method
- * @return the phase on exit from this method
+ * Awaits the phase of the barrier to advance from the given phase
+ * value, throwing {@code InterruptedException} if interrupted
+ * while waiting, or returning immediately if the current phase of
+ * the barrier is not equal to the given phase value or this
+ * barrier is terminated. It is an unenforced usage error for an
+ * unregistered party to invoke this method.
+ *
+ * @param phase an arrival phase number, or negative value if
+ * terminated; this argument is normally the value returned by a
+ * previous call to {@code arrive} or its variants
+ * @return the next arrival phase number, or a negative value
+ * if terminated or argument is negative
* @throws InterruptedException if thread interrupted while waiting
*/
- public int awaitAdvanceInterruptibly(int phase) throws InterruptedException {
+ public int awaitAdvanceInterruptibly(int phase)
+ throws InterruptedException {
if (phase < 0)
return phase;
- Thread current = Thread.currentThread();
- if (current instanceof ForkJoinWorkerThread)
- return helpingWait(phase);
- else if (Thread.interrupted() || untimedWait(current, phase, true))
- throw new InterruptedException();
- else
- return phaseOf(state.get());
+ long s = getReconciledState();
+ int p = phaseOf(s);
+ if (p != phase)
+ return p;
+ if (unarrivedOf(s) == 0 && parent != null)
+ parent.awaitAdvanceInterruptibly(phase);
+ return interruptibleWait(phase);
}
/**
- * Awaits the phase of the barrier to advance from the given value
- * or the given timeout elapses, or returns immediately if this
- * barrier is terminated.
- * @param phase the phase on entry to this method
- * @return the phase on exit from this method
+ * Awaits the phase of the barrier to advance from the given phase
+ * value or the given timeout to elapse, throwing {@code
+ * InterruptedException} if interrupted while waiting, or
+ * returning immediately if the current phase of the barrier is
+ * not equal to the given phase value or this barrier is
+ * terminated. It is an unenforced usage error for an
+ * unregistered party to invoke this method.
+ *
+ * @param phase an arrival phase number, or negative value if
+ * terminated; this argument is normally the value returned by a
+ * previous call to {@code arrive} or its variants
+ * @param timeout how long to wait before giving up, in units of
+ * {@code unit}
+ * @param unit a {@code TimeUnit} determining how to interpret the
+ * {@code timeout} parameter
+ * @return the next arrival phase number, or a negative value
+ * if terminated or argument is negative
* @throws InterruptedException if thread interrupted while waiting
* @throws TimeoutException if timed out while waiting
*/
- public int awaitAdvanceInterruptibly(int phase, long timeout, TimeUnit unit)
+ public int awaitAdvanceInterruptibly(int phase,
+ long timeout, TimeUnit unit)
throws InterruptedException, TimeoutException {
if (phase < 0)
return phase;
- long nanos = unit.toNanos(timeout);
- Thread current = Thread.currentThread();
- if (current instanceof ForkJoinWorkerThread)
- return timedHelpingWait(phase, nanos);
- timedWait(current, phase, nanos);
- return phaseOf(state.get());
+ long s = getReconciledState();
+ int p = phaseOf(s);
+ if (p != phase)
+ return p;
+ if (unarrivedOf(s) == 0 && parent != null)
+ parent.awaitAdvanceInterruptibly(phase, timeout, unit);
+ return timedWait(phase, unit.toNanos(timeout));
}
/**
* Forces this barrier to enter termination state. Counts of
- * arrived and registered parties are unaffected. This method may
- * be useful for coordinating recovery after one or more tasks
- * encounter unexpected exceptions.
+ * arrived and registered parties are unaffected. If this phaser
+ * has a parent, it too is terminated. This method may be useful
+ * for coordinating recovery after one or more tasks encounter
+ * unexpected exceptions.
*/
public void forceTermination() {
- final AtomicLong state = this.state;
for (;;) {
- long s = state.get();
+ long s = getReconciledState();
int phase = phaseOf(s);
int parties = partiesOf(s);
int unarrived = unarrivedOf(s);
if (phase < 0 ||
- state.compareAndSet(s, stateFor(-1, parties, unarrived))) {
- if (head.get() != null)
- releaseWaiters(-1);
+ casState(s, stateFor(-1, parties, unarrived))) {
+ releaseWaiters(0);
+ releaseWaiters(1);
+ if (parent != null)
+ parent.forceTermination();
return;
}
}
}
/**
- * Resets the barrier with the given numbers of registered unarrived
- * parties and phase number 0. This method allows repeated reuse
- * of this barrier, but only if it is somehow known not to be in
- * use for other purposes.
- * @param parties the number of parties required to trip barrier.
- * @throws IllegalArgumentException if parties less than zero
- * or greater than the maximum number of parties supported.
- */
- public void reset(int parties) {
- if (parties < 0 || parties > ushortMask)
- throw new IllegalArgumentException("Illegal number of parties");
- state.set(stateFor(0, parties, parties));
- if (head.get() != null)
- releaseWaiters(0);
- }
-
- /**
* Returns the current phase number. The maximum phase number is
- * Integer.MAX_VALUE, after which it restarts at
+ * {@code Integer.MAX_VALUE}, after which it restarts at
* zero. Upon termination, the phase number is negative.
+ *
* @return the phase number, or a negative value if terminated
*/
- public int getPhase() {
- return phaseOf(state.get());
+ public final int getPhase() {
+ return phaseOf(getReconciledState());
}
/**
* Returns the number of parties registered at this barrier.
+ *
* @return the number of parties
*/
public int getRegisteredParties() {
- return partiesOf(state.get());
+ return partiesOf(state);
}
/**
- * Returns the number of parties that have arrived at the current
- * phase of this barrier.
+ * Returns the number of registered parties that have arrived at
+ * the current phase of this barrier.
+ *
* @return the number of arrived parties
*/
public int getArrivedParties() {
- return arrivedOf(state.get());
+ return arrivedOf(state);
}
/**
* Returns the number of registered parties that have not yet
* arrived at the current phase of this barrier.
+ *
* @return the number of unarrived parties
*/
public int getUnarrivedParties() {
- return unarrivedOf(state.get());
+ return unarrivedOf(state);
+ }
+
+ /**
+ * Returns the parent of this phaser, or {@code null} if none.
+ *
+ * @return the parent of this phaser, or {@code null} if none
+ */
+ public Phaser getParent() {
+ return parent;
+ }
+
+ /**
+ * Returns the root ancestor of this phaser, which is the same as
+ * this phaser if it has no parent.
+ *
+ * @return the root ancestor of this phaser
+ */
+ public Phaser getRoot() {
+ return root;
}
/**
- * Returns true if this barrier has been terminated.
- * @return true if this barrier has been terminated
+ * Returns {@code true} if this barrier has been terminated.
+ *
+ * @return {@code true} if this barrier has been terminated
*/
public boolean isTerminated() {
- return phaseOf(state.get()) < 0;
+ return getPhase() < 0;
}
/**
- * Overridable method to perform an action upon phase advance, and
- * to control termination. This method is invoked whenever the
- * barrier is tripped (and thus all other waiting parties are
- * dormant). If it returns true, then, rather than advance the
- * phase number, this barrier will be set to a final termination
- * state, and subsequent calls to isTerminated will
- * return true.
- *
- * The default version returns true when the number of
+ * Overridable method to perform an action upon impending phase
+ * advance, and to control termination. This method is invoked
+ * upon arrival of the party tripping the barrier (when all other
+ * waiting parties are dormant). If this method returns {@code
+ * true}, then, rather than advance the phase number, this barrier
+ * will be set to a final termination state, and subsequent calls
+ * to {@link #isTerminated} will return true. Any (unchecked)
+ * Exception or Error thrown by an invocation of this method is
+ * propagated to the party attempting to trip the barrier, in
+ * which case no advance occurs.
+ *
+ *
The arguments to this method provide the state of the phaser
+ * prevailing for the current transition. (When called from within
+ * an implementation of {@code onAdvance} the values returned by
+ * methods such as {@code getPhase} may or may not reliably
+ * indicate the state to which this transition applies.)
+ *
+ *
The default version returns {@code true} when the number of
* registered parties is zero. Normally, overrides that arrange
* termination for other reasons should also preserve this
* property.
*
+ *
You may override this method to perform an action with side
+ * effects visible to participating tasks, but doing so requires
+ * care: Method {@code onAdvance} may be invoked more than once
+ * per transition. Further, unless all parties register before
+ * any arrive, and all {@link #awaitAdvance} at each phase, then
+ * you cannot ensure lack of interference from other parties
+ * during the invocation of this method.
+ *
* @param phase the phase number on entering the barrier
- * @param registeredParties the current number of registered
- * parties.
- * @return true if this barrier should terminate
+ * @param registeredParties the current number of registered parties
+ * @return {@code true} if this barrier should terminate
*/
protected boolean onAdvance(int phase, int registeredParties) {
return registeredParties <= 0;
}
/**
- * Returns a string identifying this barrier, as well as its
+ * Returns a string identifying this phaser, as well as its
* state. The state, in brackets, includes the String {@code
- * "phase ="} followed by the phase number, {@code "parties ="}
+ * "phase = "} followed by the phase number, {@code "parties = "}
* followed by the number of registered parties, and {@code
- * "arrived ="} followed by the number of arrived parties
+ * "arrived = "} followed by the number of arrived parties.
*
* @return a string identifying this barrier, as well as its state
*/
public String toString() {
- long s = state.get();
- return super.toString() + "[phase = " + phaseOf(s) + " parties = " + partiesOf(s) + " arrived = " + arrivedOf(s) + "]";
+ long s = getReconciledState();
+ return super.toString() +
+ "[phase = " + phaseOf(s) +
+ " parties = " + partiesOf(s) +
+ " arrived = " + arrivedOf(s) + "]";
}
- // methods for tripping and waiting
+ // methods for waiting
/**
- * Advance the current phase (or terminate)
+ * Wait nodes for Treiber stack representing wait queue
*/
- private void trip(int phase, int parties) {
- int next = onAdvance(phase, parties)? -1 : ((phase + 1) & phaseMask);
- state.set(stateFor(next, parties, parties));
- if (head.get() != null)
- releaseWaiters(next);
- }
-
- private int helpingWait(int phase) {
- final AtomicLong state = this.state;
- int p;
- while ((p = phaseOf(state.get())) == phase) {
- ForkJoinTask t = ForkJoinWorkerThread.pollTask();
- if (t != null) {
- if ((p = phaseOf(state.get())) == phase)
- t.exec();
- else { // push task and exit if barrier advanced
- t.fork();
- break;
- }
- }
+ static final class QNode implements ForkJoinPool.ManagedBlocker {
+ final Phaser phaser;
+ final int phase;
+ final long startTime;
+ final long nanos;
+ final boolean timed;
+ final boolean interruptible;
+ volatile boolean wasInterrupted = false;
+ volatile Thread thread; // nulled to cancel wait
+ QNode next;
+ QNode(Phaser phaser, int phase, boolean interruptible,
+ boolean timed, long startTime, long nanos) {
+ this.phaser = phaser;
+ this.phase = phase;
+ this.timed = timed;
+ this.interruptible = interruptible;
+ this.startTime = startTime;
+ this.nanos = nanos;
+ thread = Thread.currentThread();
}
- return p;
- }
-
- private int timedHelpingWait(int phase, long nanos) throws TimeoutException {
- final AtomicLong state = this.state;
- long lastTime = System.nanoTime();
- int p;
- while ((p = phaseOf(state.get())) == phase) {
- long now = System.nanoTime();
- nanos -= now - lastTime;
- lastTime = now;
- if (nanos <= 0) {
- if ((p = phaseOf(state.get())) == phase)
- throw new TimeoutException();
- else
- break;
+ public boolean isReleasable() {
+ return (thread == null ||
+ phaser.getPhase() != phase ||
+ (interruptible && wasInterrupted) ||
+ (timed && (nanos - (System.nanoTime() - startTime)) <= 0));
+ }
+ public boolean block() {
+ if (Thread.interrupted()) {
+ wasInterrupted = true;
+ if (interruptible)
+ return true;
}
- ForkJoinTask t = ForkJoinWorkerThread.pollTask();
+ if (!timed)
+ LockSupport.park(this);
+ else {
+ long waitTime = nanos - (System.nanoTime() - startTime);
+ if (waitTime <= 0)
+ return true;
+ LockSupport.parkNanos(this, waitTime);
+ }
+ return isReleasable();
+ }
+ void signal() {
+ Thread t = thread;
if (t != null) {
- if ((p = phaseOf(state.get())) == phase)
- t.exec();
- else { // push task and exit if barrier advanced
- t.fork();
- break;
+ thread = null;
+ LockSupport.unpark(t);
+ }
+ }
+ boolean doWait() {
+ if (thread != null) {
+ try {
+ ForkJoinPool.managedBlock(this, false);
+ } catch (InterruptedException ie) {
}
}
+ return wasInterrupted;
}
- return p;
+
}
/**
- * Wait nodes for Treiber stack representing wait queue for non-FJ
- * tasks. The waiting scheme is an adaptation of the one used in
- * forkjoin.PoolBarrier.
+ * Removes and signals waiting threads from wait queue.
*/
- static final class QNode {
- QNode next;
- volatile Thread thread; // nulled to cancel wait
- final int phase;
- QNode(Thread t, int c) {
- thread = t;
- phase = c;
- }
- }
-
- private void releaseWaiters(int currentPhase) {
- final AtomicReference head = this.head;
- QNode p;
- while ((p = head.get()) != null && p.phase != currentPhase) {
- if (head.compareAndSet(p, null)) {
- do {
- Thread t = p.thread;
- if (t != null) {
- p.thread = null;
- LockSupport.unpark(t);
- }
- } while ((p = p.next) != null);
- }
+ private void releaseWaiters(int phase) {
+ AtomicReference head = queueFor(phase);
+ QNode q;
+ while ((q = head.get()) != null) {
+ if (head.compareAndSet(q, q.next))
+ q.signal();
}
}
- /** The number of CPUs, for spin control */
- static final int NCPUS = Runtime.getRuntime().availableProcessors();
-
/**
- * The number of times to spin before blocking in timed waits.
- * The value is empirically derived.
- */
- static final int maxTimedSpins = (NCPUS < 2)? 0 : 32;
-
- /**
- * The number of times to spin before blocking in untimed waits.
- * This is greater than timed value because untimed waits spin
- * faster since they don't need to check times on each spin.
+ * Tries to enqueue given node in the appropriate wait queue.
+ *
+ * @return true if successful
*/
- static final int maxUntimedSpins = maxTimedSpins * 32;
+ private boolean tryEnqueue(QNode node) {
+ AtomicReference head = queueFor(node.phase);
+ return head.compareAndSet(node.next = head.get(), node);
+ }
/**
- * The number of nanoseconds for which it is faster to spin
- * rather than to use timed park. A rough estimate suffices.
+ * Enqueues node and waits unless aborted or signalled.
+ *
+ * @return current phase
*/
- static final long spinForTimeoutThreshold = 1000L;
+ private int untimedWait(int phase) {
+ QNode node = null;
+ boolean queued = false;
+ boolean interrupted = false;
+ int p;
+ while ((p = getPhase()) == phase) {
+ if (Thread.interrupted())
+ interrupted = true;
+ else if (node == null)
+ node = new QNode(this, phase, false, false, 0, 0);
+ else if (!queued)
+ queued = tryEnqueue(node);
+ else
+ interrupted = node.doWait();
+ }
+ if (node != null)
+ node.thread = null;
+ releaseWaiters(phase);
+ if (interrupted)
+ Thread.currentThread().interrupt();
+ return p;
+ }
/**
- * Enqueues node and waits unless aborted or signalled.
+ * Interruptible version
+ * @return current phase
*/
- private boolean untimedWait(Thread thread, int currentPhase,
- boolean abortOnInterrupt) {
- final AtomicReference head = this.head;
- final AtomicLong state = this.state;
- boolean wasInterrupted = false;
+ private int interruptibleWait(int phase) throws InterruptedException {
QNode node = null;
boolean queued = false;
- int spins = maxUntimedSpins;
- while (phaseOf(state.get()) == currentPhase) {
- QNode h;
- if (node != null && queued) {
- if (node.thread != null) {
- LockSupport.park();
- if (Thread.interrupted()) {
- wasInterrupted = true;
- if (abortOnInterrupt)
- break;
- }
- }
- }
- else if ((h = head.get()) != null && h.phase != currentPhase) {
- if (phaseOf(state.get()) == currentPhase) { // must recheck
- if (head.compareAndSet(h, h.next)) {
- Thread t = h.thread; // help clear out old waiters
- if (t != null) {
- h.thread = null;
- LockSupport.unpark(t);
- }
- }
- }
- else
- break;
- }
- else if (node != null)
- queued = head.compareAndSet(node.next = h, node);
- else if (spins <= 0)
- node = new QNode(thread, currentPhase);
+ boolean interrupted = false;
+ int p;
+ while ((p = getPhase()) == phase && !interrupted) {
+ if (Thread.interrupted())
+ interrupted = true;
+ else if (node == null)
+ node = new QNode(this, phase, true, false, 0, 0);
+ else if (!queued)
+ queued = tryEnqueue(node);
else
- --spins;
+ interrupted = node.doWait();
}
if (node != null)
node.thread = null;
- return wasInterrupted;
+ if (p != phase || (p = getPhase()) != phase)
+ releaseWaiters(phase);
+ if (interrupted)
+ throw new InterruptedException();
+ return p;
}
/**
- * Messier timeout version
+ * Timeout version.
+ * @return current phase
*/
- private void timedWait(Thread thread, int currentPhase, long nanos)
+ private int timedWait(int phase, long nanos)
throws InterruptedException, TimeoutException {
- final AtomicReference head = this.head;
- final AtomicLong state = this.state;
- long lastTime = System.nanoTime();
+ long startTime = System.nanoTime();
QNode node = null;
boolean queued = false;
- int spins = maxTimedSpins;
- while (phaseOf(state.get()) == currentPhase) {
- QNode h;
- long now = System.nanoTime();
- nanos -= now - lastTime;
- lastTime = now;
- if (nanos <= 0) {
- if (node != null)
- node.thread = null;
- if (phaseOf(state.get()) == currentPhase)
- throw new TimeoutException();
- else
- break;
- }
- else if (node != null && queued) {
- if (node.thread != null &&
- nanos > spinForTimeoutThreshold) {
- // LockSupport.parkNanos(this, nanos);
- LockSupport.parkNanos(nanos);
- if (Thread.interrupted()) {
- node.thread = null;
- throw new InterruptedException();
- }
- }
- }
- else if ((h = head.get()) != null && h.phase != currentPhase) {
- if (phaseOf(state.get()) == currentPhase) { // must recheck
- if (head.compareAndSet(h, h.next)) {
- Thread t = h.thread; // help clear out old waiters
- if (t != null) {
- h.thread = null;
- LockSupport.unpark(t);
- }
- }
- }
- else
- break;
- }
- else if (node != null)
- queued = head.compareAndSet(node.next = h, node);
- else if (spins <= 0)
- node = new QNode(thread, currentPhase);
+ boolean interrupted = false;
+ int p;
+ while ((p = getPhase()) == phase && !interrupted) {
+ if (Thread.interrupted())
+ interrupted = true;
+ else if (nanos - (System.nanoTime() - startTime) <= 0)
+ break;
+ else if (node == null)
+ node = new QNode(this, phase, true, true, startTime, nanos);
+ else if (!queued)
+ queued = tryEnqueue(node);
else
- --spins;
+ interrupted = node.doWait();
}
if (node != null)
node.thread = null;
+ if (p != phase || (p = getPhase()) != phase)
+ releaseWaiters(phase);
+ if (interrupted)
+ throw new InterruptedException();
+ if (p == phase)
+ throw new TimeoutException();
+ return p;
}
-}
+ // Unsafe mechanics
+ private static final sun.misc.Unsafe UNSAFE = getUnsafe();
+ private static final long stateOffset =
+ objectFieldOffset("state", Phaser.class);
+
+ private final boolean casState(long cmp, long val) {
+ return UNSAFE.compareAndSwapLong(this, stateOffset, cmp, val);
+ }
+
+ private static long objectFieldOffset(String field, Class klazz) {
+ try {
+ return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
+ } catch (NoSuchFieldException e) {
+ // Convert Exception to corresponding Error
+ NoSuchFieldError error = new NoSuchFieldError(field);
+ error.initCause(e);
+ throw error;
+ }
+ }
+
+ /**
+ * Returns a sun.misc.Unsafe. Suitable for use in a 3rd party package.
+ * Replace with a simple call to Unsafe.getUnsafe when integrating
+ * into a jdk.
+ *
+ * @return a sun.misc.Unsafe
+ */
+ private static sun.misc.Unsafe getUnsafe() {
+ try {
+ return sun.misc.Unsafe.getUnsafe();
+ } catch (SecurityException se) {
+ try {
+ return java.security.AccessController.doPrivileged
+ (new java.security
+ .PrivilegedExceptionAction() {
+ public sun.misc.Unsafe run() throws Exception {
+ java.lang.reflect.Field f = sun.misc
+ .Unsafe.class.getDeclaredField("theUnsafe");
+ f.setAccessible(true);
+ return (sun.misc.Unsafe) f.get(null);
+ }});
+ } catch (java.security.PrivilegedActionException e) {
+ throw new RuntimeException("Could not initialize intrinsics",
+ e.getCause());
+ }
+ }
+ }
+}