src/jsr166y/CountedCompleter.java

/*
 * Written by Doug Lea with assistance from members of JCP JSR-166
 * Expert Group and released to the public domain, as explained at
 * http://creativecommons.org/publicdomain/zero/1.0/
 */

package jsr166y;

/**
 * A {@link ForkJoinTask} with a completion action
 * performed when triggered and there are no remaining pending
 * actions. Uses of CountedCompleter are similar to those of other
 * completion based components (such as {@link
 * java.nio.channels.CompletionHandler}) except that multiple
 * <em>pending</em> completions may be necessary to trigger the {@link
 * #onCompletion} action, not just one. Unless initialized otherwise,
 * the {@link #getPendingCount pending count} starts at zero, but may
 * be (atomically) changed using methods {@link #setPendingCount},
 * {@link #addToPendingCount}, and {@link
 * #compareAndSetPendingCount}. Upon invocation of {@link
 * #tryComplete}, if the pending action count is nonzero, it is
 * decremented; otherwise, the completion action is performed, and if
 * this completer itself has a completer, the process is continued
 * with its completer.  As is the case with related synchronization
 * components such as {@link Phaser} and {@link
 * java.util.concurrent.Semaphore} these methods affect only internal
 * counts; they do not establish any further internal bookkeeping. In
 * particular, the identities of pending tasks are not maintained. As
 * illustrated below, you can create subclasses that do record some or
 * all pended tasks or their results when needed.
 *
 * <p>A concrete CountedCompleter class must define method {@link
 * #compute}, that should, in almost all use cases, invoke {@code
 * tryComplete()} once before returning. The class may also optionally
 * override method {@link #onCompletion} to perform an action upon
 * normal completion, and method {@link #onExceptionalCompletion} to
 * perform an action upon any exception.
 *
 * <p>CountedCompleters most often do not bear results, in which case
 * they are normally declared as {@code CountedCompleter<Void>}, and
 * will always return {@code null} as a result value.  In other cases,
 * you should override method {@link #getRawResult} to provide a
 * result from {@code join(), invoke()}, and related methods. (Method
 * {@link #setRawResult} by default plays no role in CountedCompleters
 * but may be overridden for example to maintain fields holding result
 * data.)
 *
 * <p>A CountedCompleter that does not itself have a completer (i.e.,
 * one for which {@link #getCompleter} returns {@code null}) can be
 * used as a regular ForkJoinTask with this added functionality.
 * However, any completer that in turn has another completer serves
 * only as an internal helper for other computations, so its own task
 * status (as reported in methods such as {@link ForkJoinTask#isDone})
 * is arbitrary; this status changes only upon explicit invocations of
 * {@link #complete}, {@link ForkJoinTask#cancel}, {@link
 * ForkJoinTask#completeExceptionally} or upon exceptional completion
 * of method {@code compute}. Upon any exceptional completion, the
 * exception may be relayed to a task's completer (and its completer,
 * and so on), if one exists and it has not otherwise already
 * completed.
 *
 * <p><b>Sample Usages.</b>
 *
 * <p><b>Parallel recursive decomposition.</b> CountedCompleters may
 * be arranged in trees similar to those often used with {@link
 * RecursiveAction}s, although the constructions involved in setting
 * them up typically vary. Here, the completer of each task is its
 * parent in the computation tree. Even though they entail a bit more
 * bookkeeping, CountedCompleters may be better choices when applying
 * a possibly time-consuming operation (that cannot be further
 * subdivided) to each element of an array or collection; especially
 * when the operation takes a significantly different amount of time
 * to complete for some elements than others, either because of
 * intrinsic variation (for example IO) or auxiliary effects such as
 * garbage collection.  Because CountedCompleters provide their own
 * continuations, other threads need not block waiting to perform
 * them.
 *
 * <p> For example, here is an initial version of a class that uses
 * divide-by-two recursive decomposition to divide work into single
 * pieces (leaf tasks). Even when work is split into individual calls,
 * tree-based techniques are usually preferable to directly forking
 * leaf tasks, because they reduce inter-thread communication and
 * improve load balancing. In the recursive case, the second of each
 * pair of subtasks to finish triggers completion of its parent
 * (because no result combination is performed, the default no-op
 * implementation of method {@code onCompletion} is not overridden). A
 * static utility method sets up the base task and invokes it:
 *
 * <pre> {@code
 * class MyOperation<E> { void apply(E e) { ... }  }
 *
 * class ForEach<E> extends CountedCompleter<Void> {
 *
 *     public static <E> void forEach(ForkJoinPool pool, E[] array, MyOperation<E> op) {
 *         pool.invoke(new ForEach<E>(null, array, op, 0, array.length));
 *     }
 *
 *     final E[] array; final MyOperation<E> op; final int lo, hi;
 *     ForEach(CountedCompleter<?> p, E[] array, MyOperation<E> op, int lo, int hi) {
 *         super(p);
 *         this.array = array; this.op = op; this.lo = lo; this.hi = hi;
 *     }
 *
 *     public void compute() { // version 1
 *         if (hi - lo >= 2) {
 *             int mid = (lo + hi) >>> 1;
 *             setPendingCount(2); // must set pending count before fork
 *             new ForEach(this, array, op, mid, hi).fork(); // right child
 *             new ForEach(this, array, op, lo, mid).fork(); // left child
 *         }
 *         else if (hi > lo)
 *             op.apply(array[lo]);
 *         tryComplete();
 *     }
 * } }</pre>
 *
 * This design can be improved by noticing that in the recursive case,
 * the task has nothing to do after forking its right task, so can
 * directly invoke its left task before returning. (This is an analog
 * of tail recursion removal.)  Also, because the task returns upon
 * executing its left task (rather than falling through to invoke
 * tryComplete) the pending count is set to one:
 *
 * <pre> {@code
 * class ForEach<E> ...
 *     public void compute() { // version 2
 *         if (hi - lo >= 2) {
 *             int mid = (lo + hi) >>> 1;
 *             setPendingCount(1); // only one pending
 *             new ForEach(this, array, op, mid, hi).fork(); // right child
 *             new ForEach(this, array, op, lo, mid).compute(); // direct invoke
 *         }
 *         else {
 *             if (hi > lo)
 *                 op.apply(array[lo]);
 *             tryComplete();
 *         }
 *     }
 * }</pre>
 *
 * As a further improvement, notice that the left task need not even
 * exist.  Instead of creating a new one, we can iterate using the
 * original task, and add a pending count for each fork:
 *
 * <pre> {@code
 * class ForEach<E> ...
 *     public void compute() { // version 3
 *         int l = lo,  h = hi;
 *         while (h - l >= 2) {
 *             int mid = (l + h) >>> 1;
 *             addToPendingCount(1);
 *             new ForEach(this, array, op, mid, h).fork(); // right child
 *             h = mid;
 *         }
 *         if (h > l)
 *             op.apply(array[l]);
 *         tryComplete();
 *     }
 * }</pre>
 *
 * Additional improvements of such classes might entail precomputing
 * pending counts so that they can be established in constructors,
 * specializing classes for leaf steps, subdividing by say, four,
 * instead of two per iteration, and using an adaptive threshold
 * instead of always subdividing down to single elements.
 *
 * <p><b>Recording subtasks.</b> CountedCompleter tasks that combine
 * results of multiple subtasks usually need to access these results
 * in method {@link #onCompletion}. As illustrated in the following
 * class (that performs a simplified form of map-reduce where mappings
 * and reductions are all of type {@code E}), one way to do this in
 * divide and conquer designs is to have each subtask record its
 * sibling, so that it can be accessed in method {@code onCompletion}.
 * This technique applies to reductions in which the order of
 * combining left and right results does not matter; ordered
 * reductions require explicit left/right designations.  Variants of
 * other streamlinings seen in the above examples may also apply.
 *
 * <pre> {@code
 * class MyMapper<E> { E apply(E v) {  ...  } }
 * class MyReducer<E> { E apply(E x, E y) {  ...  } }
 * class MapReducer<E> extends CountedCompleter<E> {
 *     final E[] array; final MyMapper<E> mapper;
 *     final MyReducer<E> reducer; final int lo, hi;
 *     MapReducer<E> sibling;
 *     E result;
 *     MapReducer(CountedCompleter p, E[] array, MyMapper<E> mapper,
 *                MyReducer<E> reducer, int lo, int hi) {
 *         super(p);
 *         this.array = array; this.mapper = mapper;
 *         this.reducer = reducer; this.lo = lo; this.hi = hi;
 *     }
 *     public void compute() {
 *         if (hi - lo >= 2) {
 *             int mid = (lo + hi) >>> 1;
 *             MapReducer<E> left = new MapReducer(this, array, mapper, reducer, lo, mid);
 *             MapReducer<E> right = new MapReducer(this, array, mapper, reducer, mid, hi);
 *             left.sibling = right;
 *             right.sibling = left;
 *             setPendingCount(1); // only right is pending
 *             right.fork();
 *             left.compute();     // directly execute left
 *         }
 *         else {
 *             if (hi > lo)
 *                 result = mapper.apply(array[lo]);
 *             tryComplete();
 *         }
 *     }
 *     public void onCompletion(CountedCompleter caller) {
 *         if (caller != this) {
 *            MapReducer<E> child = (MapReducer<E>)caller;
 *            MapReducer<E> sib = child.sibling;
 *            if (sib == null || sib.result == null)
 *                result = child.result;
 *            else
 *                result = reducer.apply(child.result, sib.result);
 *         }
 *     }
 *     public E getRawResult() { return result; }
 *
 *     public static <E> E mapReduce(ForkJoinPool pool, E[] array,
 *                                   MyMapper<E> mapper, MyReducer<E> reducer) {
 *         return pool.invoke(new MapReducer<E>(null, array, mapper,
 *                                              reducer, 0, array.length));
 *     }
 * } }</pre>
 *
 * <p><b>Triggers.</b> Some CountedCompleters are themselves never
 * forked, but instead serve as bits of plumbing in other designs;
 * including those in which the completion of one of more async tasks
 * triggers another async task. For example:
 *
 * <pre> {@code
 * class HeaderBuilder extends CountedCompleter<...> { ... }
 * class BodyBuilder extends CountedCompleter<...> { ... }
 * class PacketSender extends CountedCompleter<...> {
 *     PacketSender(...) { super(null, 1); ... } // trigger on second completion
 *     public void compute() { } // never called
 *     public void onCompletion(CountedCompleter<?> caller) { sendPacket(); }
 * }
 * // sample use:
 * PacketSender p = new PacketSender();
 * new HeaderBuilder(p, ...).fork();
 * new BodyBuilder(p, ...).fork();
 * }</pre>
 *
 * @since 1.8
 * @author Doug Lea
 */
public abstract class CountedCompleter<T> extends ForkJoinTask<T> {
    private static final long serialVersionUID = 5232453752276485070L;

    /** This task's completer, or null if none */
    final CountedCompleter<?> completer;
    /** The number of pending tasks until completion */
    volatile int pending;

    /**
     * Creates a new CountedCompleter with the given completer
     * and initial pending count.
     *
     * @param completer this tasks completer, or {@code null} if none
     * @param initialPendingCount the initial pending count
     */
    protected CountedCompleter(CountedCompleter<?> completer,
                               int initialPendingCount) {
        this.completer = completer;
        this.pending = initialPendingCount;
    }

    /**
     * Creates a new CountedCompleter with the given completer
     * and an initial pending count of zero.
     *
     * @param completer this tasks completer, or {@code null} if none
     */
    protected CountedCompleter(CountedCompleter<?> completer) {
        this.completer = completer;
    }

    /**
     * Creates a new CountedCompleter with no completer
     * and an initial pending count of zero.
     */
    protected CountedCompleter() {
        this.completer = null;
    }

    /**
     * The main computation performed by this task.
     */
    public abstract void compute();

    /**
     * Performs an action when method {@link #tryComplete} is invoked
     * and there are no pending counts, or when the unconditional
     * method {@link #complete} is invoked.  By default, this method
     * does nothing.
     *
     * @param caller the task invoking this method (which may
     * be this task itself).
     */
    public void onCompletion(CountedCompleter<?> caller) {
    }

    /**
     * Performs an action when method {@link #completeExceptionally}
     * is invoked or method {@link #compute} throws an exception, and
     * this task has not otherwise already completed normally. On
     * entry to this method, this task {@link
     * ForkJoinTask#isCompletedAbnormally}.  The return value of this
     * method controls further propagation: If {@code true} and this
     * task has a completer, then this completer is also completed
     * exceptionally.  The default implementation of this method does
     * nothing except return {@code true}.
     *
     * @param ex the exception
     * @param caller the task invoking this method (which may
     * be this task itself).
     * @return true if this exception should be propagated to this
     * tasks completer, if one exists.
     */
    public boolean onExceptionalCompletion(Throwable ex, CountedCompleter<?> caller) {
        return true;
    }

    /**
     * Returns the completer established in this task's constructor,
     * or {@code null} if none.
     *
     * @return the completer
     */
    public final CountedCompleter<?> getCompleter() {
        return completer;
    }

    /**
     * Returns the current pending count.
     *
     * @return the current pending count
     */
    public final int getPendingCount() {
        return pending;
    }

    /**
     * Sets the pending count to the given value.
     *
     * @param count the count
     */
    public final void setPendingCount(int count) {
        pending = count;
    }

    /**
     * Adds (atomically) the given value to the pending count.
     *
     * @param delta the value to add
     */
    public final void addToPendingCount(int delta) {
        int c; // note: can replace with intrinsic in jdk8
        do {} while (!U.compareAndSwapInt(this, PENDING, c = pending, c+delta));
    }

    /**
     * Sets (atomically) the pending count to the given count only if
     * it currently holds the given expected value.
     *
     * @param expected the expected value
     * @param count the new value
     * @return true is successful
     */
    public final boolean compareAndSetPendingCount(int expected, int count) {
        return U.compareAndSwapInt(this, PENDING, expected, count);
    }

    /**
     * If the pending count is nonzero, decrements the count;
     * otherwise invokes {@link #onCompletion} and then similarly
     * tries to complete this task's completer, if one exists,
     * else marks this task as complete.
     */
    public final void tryComplete() {
        CountedCompleter<?> a = this, s = a;
        for (int c;;) {
            if ((c = a.pending) == 0) {
                a.onCompletion(s);
                if ((a = (s = a).completer) == null) {
                    s.quietlyComplete();
                    return;
                }
            }
            else if (U.compareAndSwapInt(a, PENDING, c, c - 1))
                return;
        }
    }

    /**
     * Regardless of pending count, invokes {@link #onCompletion},
     * marks this task as complete and further triggers {@link
     * #tryComplete} on this task's completer, if one exists. This
     * method may be useful when forcing completion as soon as any one
     * (versus all) of several subtask results are obtained.  The
     * given rawResult is used as an argument to {@link #setRawResult}
     * before marking this task as complete; its value is meaningful
     * only for classes overriding {@code setRawResult}.
     *
     * @param rawResult the raw result
     */
    public void complete(T rawResult) {
        CountedCompleter<?> p;
        onCompletion(this);
        setRawResult(rawResult);
        quietlyComplete();
        if ((p = completer) != null)
            p.tryComplete();
    }

    /**
     * Support for FJT exception propagation
     */
    void internalPropagateException(Throwable ex) {
        CountedCompleter<?> a = this, s = a;
        while (a.onExceptionalCompletion(ex, s) &&
               (a = (s = a).completer) != null && a.status >= 0)
            a.recordExceptionalCompletion(ex);
    }

    /**
     * Implements execution conventions for CountedCompleters
     */
    protected final boolean exec() {
        compute();
        return false;
    }

    /**
     * Returns the result of the computation. By default
     * returns {@code null}, which is appropriate for {@code Void}
     * actions, but in other cases should be overridden.
     *
     * @return the result of the computation
     */
    public T getRawResult() { return null; }

    /**
     * A method that result-bearing CountedCompleters may optionally
     * use to help maintain result data.  By default, does nothing.
     */
    protected void setRawResult(T t) { }

    // Unsafe mechanics
    private static final sun.misc.Unsafe U;
    private static final long PENDING;
    static {
        try {
            U = getUnsafe();
            PENDING = U.objectFieldOffset
                (CountedCompleter.class.getDeclaredField("pending"));
        } catch (Exception e) {
            throw new Error(e);
        }
    }


    /**
     * 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<sun.misc.Unsafe>() {
                        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());
            }
        }
    }

}
Revision:	1.3
Committed:	Thu Aug 16 12:25:03 2012 UTC (11 years, 8 months ago) by dl
Branch:	MAIN
Changes since 1.2:	+55 -39 lines
Log Message:	Parameterize CountedCompleters
#	User	Rev	Content
1	dl	1.1	/*
2			* Written by Doug Lea with assistance from members of JCP JSR-166
3			* Expert Group and released to the public domain, as explained at
4			* http://creativecommons.org/publicdomain/zero/1.0/
5			*/
6
7			package jsr166y;
8
9			/**
10	dl	1.3	* A {@link ForkJoinTask} with a completion action
11	dl	1.1	* performed when triggered and there are no remaining pending
12			* actions. Uses of CountedCompleter are similar to those of other
13			* completion based components (such as {@link
14			* java.nio.channels.CompletionHandler}) except that multiple
15			* <em>pending</em> completions may be necessary to trigger the {@link
16			* #onCompletion} action, not just one. Unless initialized otherwise,
17			* the {@link #getPendingCount pending count} starts at zero, but may
18			* be (atomically) changed using methods {@link #setPendingCount},
19			* {@link #addToPendingCount}, and {@link
20			* #compareAndSetPendingCount}. Upon invocation of {@link
21			* #tryComplete}, if the pending action count is nonzero, it is
22			* decremented; otherwise, the completion action is performed, and if
23			* this completer itself has a completer, the process is continued
24	dl	1.2	* with its completer. As is the case with related synchronization
25			* components such as {@link Phaser} and {@link
26			* java.util.concurrent.Semaphore} these methods affect only internal
27			* counts; they do not establish any further internal bookkeeping. In
28			* particular, the identities of pending tasks are not maintained. As
29			* illustrated below, you can create subclasses that do record some or
30			* all pended tasks or their results when needed.
31	dl	1.1	*
32			* <p>A concrete CountedCompleter class must define method {@link
33			* #compute}, that should, in almost all use cases, invoke {@code
34	dl	1.2	* tryComplete()} once before returning. The class may also optionally
35	dl	1.1	* override method {@link #onCompletion} to perform an action upon
36	dl	1.2	* normal completion, and method {@link #onExceptionalCompletion} to
37			* perform an action upon any exception.
38	dl	1.1	*
39	dl	1.3	* <p>CountedCompleters most often do not bear results, in which case
40			* they are normally declared as {@code CountedCompleter<Void>}, and
41			* will always return {@code null} as a result value. In other cases,
42			* you should override method {@link #getRawResult} to provide a
43			* result from {@code join(), invoke()}, and related methods. (Method
44			* {@link #setRawResult} by default plays no role in CountedCompleters
45			* but may be overridden for example to maintain fields holding result
46			* data.)
47			*
48	dl	1.1	* <p>A CountedCompleter that does not itself have a completer (i.e.,
49			* one for which {@link #getCompleter} returns {@code null}) can be
50			* used as a regular ForkJoinTask with this added functionality.
51			* However, any completer that in turn has another completer serves
52			* only as an internal helper for other computations, so its own task
53			* status (as reported in methods such as {@link ForkJoinTask#isDone})
54			* is arbitrary; this status changes only upon explicit invocations of
55			* {@link #complete}, {@link ForkJoinTask#cancel}, {@link
56			* ForkJoinTask#completeExceptionally} or upon exceptional completion
57			* of method {@code compute}. Upon any exceptional completion, the
58	dl	1.2	* exception may be relayed to a task's completer (and its completer,
59			* and so on), if one exists and it has not otherwise already
60			* completed.
61	dl	1.1	*
62			* <p><b>Sample Usages.</b>
63			*
64			* <p><b>Parallel recursive decomposition.</b> CountedCompleters may
65			* be arranged in trees similar to those often used with {@link
66			* RecursiveAction}s, although the constructions involved in setting
67	dl	1.3	* them up typically vary. Here, the completer of each task is its
68			* parent in the computation tree. Even though they entail a bit more
69	dl	1.1	* bookkeeping, CountedCompleters may be better choices when applying
70			* a possibly time-consuming operation (that cannot be further
71			* subdivided) to each element of an array or collection; especially
72			* when the operation takes a significantly different amount of time
73			* to complete for some elements than others, either because of
74			* intrinsic variation (for example IO) or auxiliary effects such as
75			* garbage collection. Because CountedCompleters provide their own
76			* continuations, other threads need not block waiting to perform
77			* them.
78			*
79			* <p> For example, here is an initial version of a class that uses
80			* divide-by-two recursive decomposition to divide work into single
81			* pieces (leaf tasks). Even when work is split into individual calls,
82			* tree-based techniques are usually preferable to directly forking
83			* leaf tasks, because they reduce inter-thread communication and
84			* improve load balancing. In the recursive case, the second of each
85			* pair of subtasks to finish triggers completion of its parent
86			* (because no result combination is performed, the default no-op
87			* implementation of method {@code onCompletion} is not overridden). A
88			* static utility method sets up the base task and invokes it:
89			*
90			* <pre> {@code
91			* class MyOperation<E> { void apply(E e) { ... } }
92			*
93	dl	1.3	* class ForEach<E> extends CountedCompleter<Void> {
94	dl	1.1	*
95			* public static <E> void forEach(ForkJoinPool pool, E[] array, MyOperation<E> op) {
96			* pool.invoke(new ForEach<E>(null, array, op, 0, array.length));
97			* }
98			*
99			* final E[] array; final MyOperation<E> op; final int lo, hi;
100	dl	1.3	* ForEach(CountedCompleter<?> p, E[] array, MyOperation<E> op, int lo, int hi) {
101	dl	1.1	* super(p);
102			* this.array = array; this.op = op; this.lo = lo; this.hi = hi;
103			* }
104			*
105			* public void compute() { // version 1
106			* if (hi - lo >= 2) {
107			* int mid = (lo + hi) >>> 1;
108			* setPendingCount(2); // must set pending count before fork
109			* new ForEach(this, array, op, mid, hi).fork(); // right child
110			* new ForEach(this, array, op, lo, mid).fork(); // left child
111			* }
112			* else if (hi > lo)
113			* op.apply(array[lo]);
114			* tryComplete();
115			* }
116			* } }</pre>
117			*
118			* This design can be improved by noticing that in the recursive case,
119			* the task has nothing to do after forking its right task, so can
120			* directly invoke its left task before returning. (This is an analog
121			* of tail recursion removal.) Also, because the task returns upon
122			* executing its left task (rather than falling through to invoke
123			* tryComplete) the pending count is set to one:
124			*
125			* <pre> {@code
126			* class ForEach<E> ...
127			* public void compute() { // version 2
128			* if (hi - lo >= 2) {
129			* int mid = (lo + hi) >>> 1;
130			* setPendingCount(1); // only one pending
131			* new ForEach(this, array, op, mid, hi).fork(); // right child
132			* new ForEach(this, array, op, lo, mid).compute(); // direct invoke
133			* }
134			* else {
135			* if (hi > lo)
136			* op.apply(array[lo]);
137			* tryComplete();
138			* }
139			* }
140			* }</pre>
141			*
142			* As a further improvement, notice that the left task need not even
143			* exist. Instead of creating a new one, we can iterate using the
144			* original task, and add a pending count for each fork:
145			*
146			* <pre> {@code
147			* class ForEach<E> ...
148			* public void compute() { // version 3
149			* int l = lo, h = hi;
150			* while (h - l >= 2) {
151			* int mid = (l + h) >>> 1;
152			* addToPendingCount(1);
153			* new ForEach(this, array, op, mid, h).fork(); // right child
154			* h = mid;
155			* }
156			* if (h > l)
157			* op.apply(array[l]);
158			* tryComplete();
159			* }
160			* }</pre>
161			*
162			* Additional improvements of such classes might entail precomputing
163			* pending counts so that they can be established in constructors,
164			* specializing classes for leaf steps, subdividing by say, four,
165			* instead of two per iteration, and using an adaptive threshold
166			* instead of always subdividing down to single elements.
167			*
168			* <p><b>Recording subtasks.</b> CountedCompleter tasks that combine
169			* results of multiple subtasks usually need to access these results
170			* in method {@link #onCompletion}. As illustrated in the following
171			* class (that performs a simplified form of map-reduce where mappings
172			* and reductions are all of type {@code E}), one way to do this in
173			* divide and conquer designs is to have each subtask record its
174			* sibling, so that it can be accessed in method {@code onCompletion}.
175	dl	1.3	* This technique applies to reductions in which the order of
176			* combining left and right results does not matter; ordered
177			* reductions require explicit left/right designations. Variants of
178			* other streamlinings seen in the above examples may also apply.
179	dl	1.1	*
180			* <pre> {@code
181			* class MyMapper<E> { E apply(E v) { ... } }
182			* class MyReducer<E> { E apply(E x, E y) { ... } }
183	dl	1.3	* class MapReducer<E> extends CountedCompleter<E> {
184	dl	1.1	* final E[] array; final MyMapper<E> mapper;
185			* final MyReducer<E> reducer; final int lo, hi;
186	dl	1.3	* MapReducer<E> sibling;
187	dl	1.1	* E result;
188			* MapReducer(CountedCompleter p, E[] array, MyMapper<E> mapper,
189			* MyReducer<E> reducer, int lo, int hi) {
190			* super(p);
191			* this.array = array; this.mapper = mapper;
192			* this.reducer = reducer; this.lo = lo; this.hi = hi;
193			* }
194			* public void compute() {
195			* if (hi - lo >= 2) {
196			* int mid = (lo + hi) >>> 1;
197			* MapReducer<E> left = new MapReducer(this, array, mapper, reducer, lo, mid);
198			* MapReducer<E> right = new MapReducer(this, array, mapper, reducer, mid, hi);
199			* left.sibling = right;
200			* right.sibling = left;
201			* setPendingCount(1); // only right is pending
202			* right.fork();
203			* left.compute(); // directly execute left
204			* }
205			* else {
206			* if (hi > lo)
207			* result = mapper.apply(array[lo]);
208			* tryComplete();
209			* }
210			* }
211			* public void onCompletion(CountedCompleter caller) {
212			* if (caller != this) {
213			* MapReducer<E> child = (MapReducer<E>)caller;
214			* MapReducer<E> sib = child.sibling;
215			* if (sib == null \|\| sib.result == null)
216			* result = child.result;
217			* else
218			* result = reducer.apply(child.result, sib.result);
219			* }
220			* }
221	dl	1.3	* public E getRawResult() { return result; }
222	dl	1.1	*
223			* public static <E> E mapReduce(ForkJoinPool pool, E[] array,
224			* MyMapper<E> mapper, MyReducer<E> reducer) {
225	dl	1.3	* return pool.invoke(new MapReducer<E>(null, array, mapper,
226			* reducer, 0, array.length));
227	dl	1.1	* }
228			* } }</pre>
229			*
230			* <p><b>Triggers.</b> Some CountedCompleters are themselves never
231			* forked, but instead serve as bits of plumbing in other designs;
232			* including those in which the completion of one of more async tasks
233			* triggers another async task. For example:
234			*
235			* <pre> {@code
236	dl	1.3	* class HeaderBuilder extends CountedCompleter<...> { ... }
237			* class BodyBuilder extends CountedCompleter<...> { ... }
238			* class PacketSender extends CountedCompleter<...> {
239	dl	1.1	* PacketSender(...) { super(null, 1); ... } // trigger on second completion
240			* public void compute() { } // never called
241	dl	1.3	* public void onCompletion(CountedCompleter<?> caller) { sendPacket(); }
242	dl	1.1	* }
243			* // sample use:
244			* PacketSender p = new PacketSender();
245			* new HeaderBuilder(p, ...).fork();
246			* new BodyBuilder(p, ...).fork();
247			* }</pre>
248			*
249			* @since 1.8
250			* @author Doug Lea
251			*/
252	dl	1.3	public abstract class CountedCompleter<T> extends ForkJoinTask<T> {
253	dl	1.2	private static final long serialVersionUID = 5232453752276485070L;
254
255	dl	1.1	/** This task's completer, or null if none */
256	dl	1.3	final CountedCompleter<?> completer;
257	dl	1.1	/** The number of pending tasks until completion */
258			volatile int pending;
259
260			/**
261			* Creates a new CountedCompleter with the given completer
262			* and initial pending count.
263			*
264			* @param completer this tasks completer, or {@code null} if none
265			* @param initialPendingCount the initial pending count
266			*/
267	dl	1.3	protected CountedCompleter(CountedCompleter<?> completer,
268	dl	1.1	int initialPendingCount) {
269			this.completer = completer;
270			this.pending = initialPendingCount;
271			}
272
273			/**
274			* Creates a new CountedCompleter with the given completer
275			* and an initial pending count of zero.
276			*
277			* @param completer this tasks completer, or {@code null} if none
278			*/
279	dl	1.3	protected CountedCompleter(CountedCompleter<?> completer) {
280	dl	1.1	this.completer = completer;
281			}
282
283			/**
284			* Creates a new CountedCompleter with no completer
285			* and an initial pending count of zero.
286			*/
287			protected CountedCompleter() {
288			this.completer = null;
289			}
290
291			/**
292			* The main computation performed by this task.
293			*/
294			public abstract void compute();
295
296			/**
297	dl	1.2	* Performs an action when method {@link #tryComplete} is invoked
298			* and there are no pending counts, or when the unconditional
299			* method {@link #complete} is invoked. By default, this method
300			* does nothing.
301	dl	1.1	*
302			* @param caller the task invoking this method (which may
303			* be this task itself).
304			*/
305	dl	1.3	public void onCompletion(CountedCompleter<?> caller) {
306	dl	1.1	}
307
308			/**
309	dl	1.2	* Performs an action when method {@link #completeExceptionally}
310			* is invoked or method {@link #compute} throws an exception, and
311			* this task has not otherwise already completed normally. On
312			* entry to this method, this task {@link
313			* ForkJoinTask#isCompletedAbnormally}. The return value of this
314			* method controls further propagation: If {@code true} and this
315			* task has a completer, then this completer is also completed
316			* exceptionally. The default implementation of this method does
317			* nothing except return {@code true}.
318			*
319			* @param ex the exception
320			* @param caller the task invoking this method (which may
321			* be this task itself).
322			* @return true if this exception should be propagated to this
323			* tasks completer, if one exists.
324			*/
325	dl	1.3	public boolean onExceptionalCompletion(Throwable ex, CountedCompleter<?> caller) {
326	dl	1.2	return true;
327			}
328
329			/**
330	dl	1.1	* Returns the completer established in this task's constructor,
331			* or {@code null} if none.
332			*
333			* @return the completer
334			*/
335	dl	1.3	public final CountedCompleter<?> getCompleter() {
336	dl	1.1	return completer;
337			}
338
339			/**
340			* Returns the current pending count.
341			*
342			* @return the current pending count
343			*/
344			public final int getPendingCount() {
345			return pending;
346			}
347
348			/**
349			* Sets the pending count to the given value.
350			*
351			* @param count the count
352			*/
353			public final void setPendingCount(int count) {
354			pending = count;
355			}
356
357			/**
358			* Adds (atomically) the given value to the pending count.
359			*
360			* @param delta the value to add
361			*/
362			public final void addToPendingCount(int delta) {
363			int c; // note: can replace with intrinsic in jdk8
364			do {} while (!U.compareAndSwapInt(this, PENDING, c = pending, c+delta));
365			}
366
367			/**
368			* Sets (atomically) the pending count to the given count only if
369			* it currently holds the given expected value.
370			*
371			* @param expected the expected value
372			* @param count the new value
373			* @return true is successful
374			*/
375			public final boolean compareAndSetPendingCount(int expected, int count) {
376			return U.compareAndSwapInt(this, PENDING, expected, count);
377			}
378
379			/**
380			* If the pending count is nonzero, decrements the count;
381			* otherwise invokes {@link #onCompletion} and then similarly
382			* tries to complete this task's completer, if one exists,
383			* else marks this task as complete.
384			*/
385			public final void tryComplete() {
386	dl	1.3	CountedCompleter<?> a = this, s = a;
387	dl	1.2	for (int c;;) {
388	dl	1.1	if ((c = a.pending) == 0) {
389			a.onCompletion(s);
390			if ((a = (s = a).completer) == null) {
391			s.quietlyComplete();
392			return;
393			}
394			}
395			else if (U.compareAndSwapInt(a, PENDING, c, c - 1))
396			return;
397			}
398			}
399
400			/**
401			* Regardless of pending count, invokes {@link #onCompletion},
402	dl	1.3	* marks this task as complete and further triggers {@link
403			* #tryComplete} on this task's completer, if one exists. This
404			* method may be useful when forcing completion as soon as any one
405			* (versus all) of several subtask results are obtained. The
406			* given rawResult is used as an argument to {@link #setRawResult}
407			* before marking this task as complete; its value is meaningful
408			* only for classes overriding {@code setRawResult}.
409	dl	1.1	*
410	dl	1.3	* @param rawResult the raw result
411	dl	1.1	*/
412	dl	1.3	public void complete(T rawResult) {
413			CountedCompleter<?> p;
414	dl	1.1	onCompletion(this);
415	dl	1.3	setRawResult(rawResult);
416	dl	1.1	quietlyComplete();
417			if ((p = completer) != null)
418			p.tryComplete();
419			}
420
421			/**
422	dl	1.2	* Support for FJT exception propagation
423			*/
424			void internalPropagateException(Throwable ex) {
425	dl	1.3	CountedCompleter<?> a = this, s = a;
426	dl	1.2	while (a.onExceptionalCompletion(ex, s) &&
427			(a = (s = a).completer) != null && a.status >= 0)
428			a.recordExceptionalCompletion(ex);
429			}
430
431			/**
432	dl	1.1	* Implements execution conventions for CountedCompleters
433			*/
434			protected final boolean exec() {
435			compute();
436			return false;
437			}
438
439			/**
440	dl	1.3	* Returns the result of the computation. By default
441			* returns {@code null}, which is appropriate for {@code Void}
442			* actions, but in other cases should be overridden.
443	dl	1.1	*
444	dl	1.3	* @return the result of the computation
445	dl	1.1	*/
446	dl	1.3	public T getRawResult() { return null; }
447	dl	1.1
448			/**
449	dl	1.3	* A method that result-bearing CountedCompleters may optionally
450			* use to help maintain result data. By default, does nothing.
451	dl	1.1	*/
452	dl	1.3	protected void setRawResult(T t) { }
453	dl	1.1
454			// Unsafe mechanics
455			private static final sun.misc.Unsafe U;
456			private static final long PENDING;
457			static {
458			try {
459			U = getUnsafe();
460			PENDING = U.objectFieldOffset
461			(CountedCompleter.class.getDeclaredField("pending"));
462			} catch (Exception e) {
463			throw new Error(e);
464			}
465			}
466
467
468			/**
469			* Returns a sun.misc.Unsafe. Suitable for use in a 3rd party package.
470			* Replace with a simple call to Unsafe.getUnsafe when integrating
471			* into a jdk.
472			*
473			* @return a sun.misc.Unsafe
474			*/
475			private static sun.misc.Unsafe getUnsafe() {
476			try {
477			return sun.misc.Unsafe.getUnsafe();
478			} catch (SecurityException se) {
479			try {
480			return java.security.AccessController.doPrivileged
481			(new java.security
482			.PrivilegedExceptionAction<sun.misc.Unsafe>() {
483			public sun.misc.Unsafe run() throws Exception {
484			java.lang.reflect.Field f = sun.misc
485			.Unsafe.class.getDeclaredField("theUnsafe");
486			f.setAccessible(true);
487			return (sun.misc.Unsafe) f.get(null);
488			}});
489			} catch (java.security.PrivilegedActionException e) {
490			throw new RuntimeException("Could not initialize intrinsics",
491			e.getCause());
492			}
493			}
494			}
495
496			}