--- jsr166/src/jsr166y/ForkJoinPool.java 2009/01/07 19:12:36 1.3 +++ jsr166/src/jsr166y/ForkJoinPool.java 2010/09/01 20:12:39 1.69 @@ -5,97 +5,384 @@ */ package jsr166y; -import java.util.*; + import java.util.concurrent.*; -import java.util.concurrent.locks.*; -import java.util.concurrent.atomic.*; -import sun.misc.Unsafe; -import java.lang.reflect.*; + +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collection; +import java.util.Collections; +import java.util.List; +import java.util.concurrent.locks.LockSupport; +import java.util.concurrent.locks.ReentrantLock; +import java.util.concurrent.atomic.AtomicInteger; +import java.util.concurrent.CountDownLatch; /** - * An {@link ExecutorService} for running {@link ForkJoinTask}s. A - * ForkJoinPool provides the entry point for submissions from - * non-ForkJoinTasks, as well as management and monitoring operations. - * Normally a single ForkJoinPool is used for a large number of - * submitted tasks. Otherwise, use would not usually outweigh the - * construction and bookkeeping overhead of creating a large set of - * threads. + * An {@link ExecutorService} for running {@link ForkJoinTask}s. + * A {@code ForkJoinPool} provides the entry point for submissions + * from non-{@code ForkJoinTask} clients, as well as management and + * monitoring operations. * - *
ForkJoinPools differ from other kinds of Executors mainly in - * that they provide work-stealing: all threads in the pool - * attempt to find and execute subtasks created by other active tasks - * (eventually blocking if none exist). This makes them efficient when - * most tasks spawn other subtasks (as do most ForkJoinTasks), as well - * as the mixed execution of some plain Runnable- or Callable- based - * activities along with ForkJoinTasks. Otherwise, other - * ExecutorService implementations are typically more appropriate - * choices. + *
A {@code ForkJoinPool} differs from other kinds of {@link + * ExecutorService} mainly by virtue of employing + * work-stealing: all threads in the pool attempt to find and + * execute subtasks created by other active tasks (eventually blocking + * waiting for work if none exist). This enables efficient processing + * when most tasks spawn other subtasks (as do most {@code + * ForkJoinTask}s). When setting asyncMode to true in + * constructors, {@code ForkJoinPool}s may also be appropriate for use + * with event-style tasks that are never joined. * - *
A ForkJoinPool may be constructed with a given parallelism level
- * (target pool size), which it attempts to maintain by dynamically
- * adding, suspending, or resuming threads, even if some tasks are
- * waiting to join others. However, no such adjustments are performed
- * in the face of blocked IO or other unmanaged synchronization. The
- * nested ManagedBlocker
interface enables extension of
- * the kinds of synchronization accommodated. The target parallelism
- * level may also be changed dynamically (setParallelism
)
- * and dynamically thread construction can be limited using methods
- * setMaximumPoolSize
and/or
- * setMaintainsParallelism
.
+ *
A {@code ForkJoinPool} is constructed with a given target + * parallelism level; by default, equal to the number of available + * processors. The pool attempts to maintain enough active (or + * available) threads by dynamically adding, suspending, or resuming + * internal worker threads, even if some tasks are stalled waiting to + * join others. However, no such adjustments are guaranteed in the + * face of blocked IO or other unmanaged synchronization. The nested + * {@link ManagedBlocker} interface enables extension of the kinds of + * synchronization accommodated. * *
In addition to execution and lifecycle control methods, this
* class provides status check methods (for example
- * getStealCount
) that are intended to aid in developing,
+ * {@link #getStealCount}) that are intended to aid in developing,
* tuning, and monitoring fork/join applications. Also, method
- * toString
returns indications of pool state in a
+ * {@link #toString} returns indications of pool state in a
* convenient form for informal monitoring.
*
+ *
As is the case with other ExecutorServices, there are three + * main task execution methods summarized in the following + * table. These are designed to be used by clients not already engaged + * in fork/join computations in the current pool. The main forms of + * these methods accept instances of {@code ForkJoinTask}, but + * overloaded forms also allow mixed execution of plain {@code + * Runnable}- or {@code Callable}- based activities as well. However, + * tasks that are already executing in a pool should normally + * NOT use these pool execution methods, but instead use the + * within-computation forms listed in the table. + * + *
+ * | Call from non-fork/join clients | + *Call from within fork/join computations | + *
Arrange async execution | + *{@link #execute(ForkJoinTask)} | + *{@link ForkJoinTask#fork} | + *
Await and obtain result | + *{@link #invoke(ForkJoinTask)} | + *{@link ForkJoinTask#invoke} | + *
Arrange exec and obtain Future | + *{@link #submit(ForkJoinTask)} | + *{@link ForkJoinTask#fork} (ForkJoinTasks are Futures) | + *
Sample Usage. Normally a single {@code ForkJoinPool} is + * used for all parallel task execution in a program or subsystem. + * Otherwise, use would not usually outweigh the construction and + * bookkeeping overhead of creating a large set of threads. For + * example, a common pool could be used for the {@code SortTasks} + * illustrated in {@link RecursiveAction}. Because {@code + * ForkJoinPool} uses threads in {@linkplain java.lang.Thread#isDaemon + * daemon} mode, there is typically no need to explicitly {@link + * #shutdown} such a pool upon program exit. + * + *
+ * static final ForkJoinPool mainPool = new ForkJoinPool(); + * ... + * public void sort(long[] array) { + * mainPool.invoke(new SortTask(array, 0, array.length)); + * } + *+ * *
Implementation notes: This implementation restricts the * maximum number of running threads to 32767. Attempts to create - * pools with greater than the maximum result in - * IllegalArgumentExceptions. + * pools with greater than the maximum number result in + * {@code IllegalArgumentException}. + * + *
This implementation rejects submitted tasks (that is, by throwing
+ * {@link RejectedExecutionException}) only when the pool is shut down
+ * or internal resources have been exhausted.
+ *
+ * @since 1.7
+ * @author Doug Lea
*/
public class ForkJoinPool extends AbstractExecutorService {
/*
- * See the extended comments interspersed below for design,
- * rationale, and walkthroughs.
+ * Implementation Overview
+ *
+ * This class provides the central bookkeeping and control for a
+ * set of worker threads: Submissions from non-FJ threads enter
+ * into a submission queue. Workers take these tasks and typically
+ * split them into subtasks that may be stolen by other workers.
+ * The main work-stealing mechanics implemented in class
+ * ForkJoinWorkerThread give first priority to processing tasks
+ * from their own queues (LIFO or FIFO, depending on mode), then
+ * to randomized FIFO steals of tasks in other worker queues, and
+ * lastly to new submissions. These mechanics do not consider
+ * affinities, loads, cache localities, etc, so rarely provide the
+ * best possible performance on a given machine, but portably
+ * provide good throughput by averaging over these factors.
+ * (Further, even if we did try to use such information, we do not
+ * usually have a basis for exploiting it. For example, some sets
+ * of tasks profit from cache affinities, but others are harmed by
+ * cache pollution effects.)
+ *
+ * Beyond work-stealing support and essential bookkeeping, the
+ * main responsibility of this framework is to take actions when
+ * one worker is waiting to join a task stolen (or always held by)
+ * another. Because we are multiplexing many tasks on to a pool
+ * of workers, we can't just let them block (as in Thread.join).
+ * We also cannot just reassign the joiner's run-time stack with
+ * another and replace it later, which would be a form of
+ * "continuation", that even if possible is not necessarily a good
+ * idea. Given that the creation costs of most threads on most
+ * systems mainly surrounds setting up runtime stacks, thread
+ * creation and switching is usually not much more expensive than
+ * stack creation and switching, and is more flexible). Instead we
+ * combine two tactics:
+ *
+ * Helping: Arranging for the joiner to execute some task that it
+ * would be running if the steal had not occurred. Method
+ * ForkJoinWorkerThread.helpJoinTask tracks joining->stealing
+ * links to try to find such a task.
+ *
+ * Compensating: Unless there are already enough live threads,
+ * method helpMaintainParallelism() may create or
+ * re-activate a spare thread to compensate for blocked
+ * joiners until they unblock.
+ *
+ * It is impossible to keep exactly the target (parallelism)
+ * number of threads running at any given time. Determining
+ * existence of conservatively safe helping targets, the
+ * availability of already-created spares, and the apparent need
+ * to create new spares are all racy and require heuristic
+ * guidance, so we rely on multiple retries of each. Compensation
+ * occurs in slow-motion. It is triggered only upon timeouts of
+ * Object.wait used for joins. This reduces poor decisions that
+ * would otherwise be made when threads are waiting for others
+ * that are stalled because of unrelated activities such as
+ * garbage collection.
+ *
+ * The ManagedBlocker extension API can't use helping so relies
+ * only on compensation in method awaitBlocker.
+ *
+ * The main throughput advantages of work-stealing stem from
+ * decentralized control -- workers mostly steal tasks from each
+ * other. We do not want to negate this by creating bottlenecks
+ * implementing other management responsibilities. So we use a
+ * collection of techniques that avoid, reduce, or cope well with
+ * contention. These entail several instances of bit-packing into
+ * CASable fields to maintain only the minimally required
+ * atomicity. To enable such packing, we restrict maximum
+ * parallelism to (1<<15)-1 (enabling twice this (to accommodate
+ * unbalanced increments and decrements) to fit into a 16 bit
+ * field, which is far in excess of normal operating range. Even
+ * though updates to some of these bookkeeping fields do sometimes
+ * contend with each other, they don't normally cache-contend with
+ * updates to others enough to warrant memory padding or
+ * isolation. So they are all held as fields of ForkJoinPool
+ * objects. The main capabilities are as follows:
+ *
+ * 1. Creating and removing workers. Workers are recorded in the
+ * "workers" array. This is an array as opposed to some other data
+ * structure to support index-based random steals by workers.
+ * Updates to the array recording new workers and unrecording
+ * terminated ones are protected from each other by a lock
+ * (workerLock) but the array is otherwise concurrently readable,
+ * and accessed directly by workers. To simplify index-based
+ * operations, the array size is always a power of two, and all
+ * readers must tolerate null slots. Currently, all worker thread
+ * creation is on-demand, triggered by task submissions,
+ * replacement of terminated workers, and/or compensation for
+ * blocked workers. However, all other support code is set up to
+ * work with other policies.
+ *
+ * To ensure that we do not hold on to worker references that
+ * would prevent GC, ALL accesses to workers are via indices into
+ * the workers array (which is one source of some of the unusual
+ * code constructions here). In essence, the workers array serves
+ * as a WeakReference mechanism. Thus for example the event queue
+ * stores worker indices, not worker references. Access to the
+ * workers in associated methods (for example releaseEventWaiters)
+ * must both index-check and null-check the IDs. All such accesses
+ * ignore bad IDs by returning out early from what they are doing,
+ * since this can only be associated with shutdown, in which case
+ * it is OK to give up. On termination, we just clobber these
+ * data structures without trying to use them.
+ *
+ * 2. Bookkeeping for dynamically adding and removing workers. We
+ * aim to approximately maintain the given level of parallelism.
+ * When some workers are known to be blocked (on joins or via
+ * ManagedBlocker), we may create or resume others to take their
+ * place until they unblock (see below). Implementing this
+ * requires counts of the number of "running" threads (i.e., those
+ * that are neither blocked nor artificially suspended) as well as
+ * the total number. These two values are packed into one field,
+ * "workerCounts" because we need accurate snapshots when deciding
+ * to create, resume or suspend. Note however that the
+ * correspondence of these counts to reality is not guaranteed. In
+ * particular updates for unblocked threads may lag until they
+ * actually wake up.
+ *
+ * 3. Maintaining global run state. The run state of the pool
+ * consists of a runLevel (SHUTDOWN, TERMINATING, etc) similar to
+ * those in other Executor implementations, as well as a count of
+ * "active" workers -- those that are, or soon will be, or
+ * recently were executing tasks. The runLevel and active count
+ * are packed together in order to correctly trigger shutdown and
+ * termination. Without care, active counts can be subject to very
+ * high contention. We substantially reduce this contention by
+ * relaxing update rules. A worker must claim active status
+ * prospectively, by activating if it sees that a submitted or
+ * stealable task exists (it may find after activating that the
+ * task no longer exists). It stays active while processing this
+ * task (if it exists) and any other local subtasks it produces,
+ * until it cannot find any other tasks. It then tries
+ * inactivating (see method preStep), but upon update contention
+ * instead scans for more tasks, later retrying inactivation if it
+ * doesn't find any.
+ *
+ * 4. Managing idle workers waiting for tasks. We cannot let
+ * workers spin indefinitely scanning for tasks when none are
+ * available. On the other hand, we must quickly prod them into
+ * action when new tasks are submitted or generated. We
+ * park/unpark these idle workers using an event-count scheme.
+ * Field eventCount is incremented upon events that may enable
+ * workers that previously could not find a task to now find one:
+ * Submission of a new task to the pool, or another worker pushing
+ * a task onto a previously empty queue. (We also use this
+ * mechanism for configuration and termination actions that
+ * require wakeups of idle workers). Each worker maintains its
+ * last known event count, and blocks when a scan for work did not
+ * find a task AND its lastEventCount matches the current
+ * eventCount. Waiting idle workers are recorded in a variant of
+ * Treiber stack headed by field eventWaiters which, when nonzero,
+ * encodes the thread index and count awaited for by the worker
+ * thread most recently calling eventSync. This thread in turn has
+ * a record (field nextEventWaiter) for the next waiting worker.
+ * In addition to allowing simpler decisions about need for
+ * wakeup, the event count bits in eventWaiters serve the role of
+ * tags to avoid ABA errors in Treiber stacks. Upon any wakeup,
+ * released threads also try to release at most two others. The
+ * net effect is a tree-like diffusion of signals, where released
+ * threads (and possibly others) help with unparks. To further
+ * reduce contention effects a bit, failed CASes to increment
+ * field eventCount are tolerated without retries in signalWork.
+ * Conceptually they are merged into the same event, which is OK
+ * when their only purpose is to enable workers to scan for work.
+ *
+ * 5. Managing suspension of extra workers. When a worker notices
+ * (usually upon timeout of a wait()) that there are too few
+ * running threads, we may create a new thread to maintain
+ * parallelism level, or at least avoid starvation. Usually, extra
+ * threads are needed for only very short periods, yet join
+ * dependencies are such that we sometimes need them in
+ * bursts. Rather than create new threads each time this happens,
+ * we suspend no-longer-needed extra ones as "spares". For most
+ * purposes, we don't distinguish "extra" spare threads from
+ * normal "core" threads: On each call to preStep (the only point
+ * at which we can do this) a worker checks to see if there are
+ * now too many running workers, and if so, suspends itself.
+ * Method helpMaintainParallelism looks for suspended threads to
+ * resume before considering creating a new replacement. The
+ * spares themselves are encoded on another variant of a Treiber
+ * Stack, headed at field "spareWaiters". Note that the use of
+ * spares is intrinsically racy. One thread may become a spare at
+ * about the same time as another is needlessly being created. We
+ * counteract this and related slop in part by requiring resumed
+ * spares to immediately recheck (in preStep) to see whether they
+ * they should re-suspend.
+ *
+ * 6. Killing off unneeded workers. A timeout mechanism is used to
+ * shed unused workers: The oldest (first) event queue waiter uses
+ * a timed rather than hard wait. When this wait times out without
+ * a normal wakeup, it tries to shutdown any one (for convenience
+ * the newest) other spare or event waiter via
+ * tryShutdownUnusedWorker. This eventually reduces the number of
+ * worker threads to a minimum of one after a long enough period
+ * without use.
+ *
+ * 7. Deciding when to create new workers. The main dynamic
+ * control in this class is deciding when to create extra threads
+ * in method helpMaintainParallelism. We would like to keep
+ * exactly #parallelism threads running, which is an impossible
+ * task. We always need to create one when the number of running
+ * threads would become zero and all workers are busy. Beyond
+ * this, we must rely on heuristics that work well in the
+ * presence of transient phenomena such as GC stalls, dynamic
+ * compilation, and wake-up lags. These transients are extremely
+ * common -- we are normally trying to fully saturate the CPUs on
+ * a machine, so almost any activity other than running tasks
+ * impedes accuracy. Our main defense is to allow parallelism to
+ * lapse for a while during joins, and use a timeout to see if,
+ * after the resulting settling, there is still a need for
+ * additional workers. This also better copes with the fact that
+ * some of the methods in this class tend to never become compiled
+ * (but are interpreted), so some components of the entire set of
+ * controls might execute 100 times faster than others. And
+ * similarly for cases where the apparent lack of work is just due
+ * to GC stalls and other transient system activity.
+ *
+ * Beware that there is a lot of representation-level coupling
+ * among classes ForkJoinPool, ForkJoinWorkerThread, and
+ * ForkJoinTask. For example, direct access to "workers" array by
+ * workers, and direct access to ForkJoinTask.status by both
+ * ForkJoinPool and ForkJoinWorkerThread. There is little point
+ * trying to reduce this, since any associated future changes in
+ * representations will need to be accompanied by algorithmic
+ * changes anyway.
+ *
+ * Style notes: There are lots of inline assignments (of form
+ * "while ((local = field) != 0)") which are usually the simplest
+ * way to ensure the required read orderings (which are sometimes
+ * critical). Also several occurrences of the unusual "do {}
+ * while (!cas...)" which is the simplest way to force an update of
+ * a CAS'ed variable. There are also other coding oddities that
+ * help some methods perform reasonably even when interpreted (not
+ * compiled), at the expense of some messy constructions that
+ * reduce byte code counts.
+ *
+ * The order of declarations in this file is: (1) statics (2)
+ * fields (along with constants used when unpacking some of them)
+ * (3) internal control methods (4) callbacks and other support
+ * for ForkJoinTask and ForkJoinWorkerThread classes, (5) exported
+ * methods (plus a few little helpers).
*/
- /** Mask for packing and unpacking shorts */
- private static final int shortMask = 0xffff;
-
- /** Max pool size -- must be a power of two minus 1 */
- private static final int MAX_THREADS = 0x7FFF;
-
/**
- * Factory for creating new ForkJoinWorkerThreads. A
- * ForkJoinWorkerThreadFactory must be defined and used for
- * ForkJoinWorkerThread subclasses that extend base functionality
- * or initialize threads with different contexts.
+ * Factory for creating new {@link ForkJoinWorkerThread}s.
+ * A {@code ForkJoinWorkerThreadFactory} must be defined and used
+ * for {@code ForkJoinWorkerThread} subclasses that extend base
+ * functionality or initialize threads with different contexts.
*/
public static interface ForkJoinWorkerThreadFactory {
/**
* Returns a new worker thread operating in the given pool.
*
* @param pool the pool this thread works in
- * @throws NullPointerException if pool is null;
+ * @throws NullPointerException if the pool is null
*/
public ForkJoinWorkerThread newThread(ForkJoinPool pool);
}
/**
- * Default ForkJoinWorkerThreadFactory implementation, creates a
+ * Default ForkJoinWorkerThreadFactory implementation; creates a
* new ForkJoinWorkerThread.
*/
- static class DefaultForkJoinWorkerThreadFactory
+ static class DefaultForkJoinWorkerThreadFactory
implements ForkJoinWorkerThreadFactory {
public ForkJoinWorkerThread newThread(ForkJoinPool pool) {
- try {
- return new ForkJoinWorkerThread(pool);
- } catch (OutOfMemoryError oom) {
- return null;
- }
+ return new ForkJoinWorkerThread(pool);
}
}
@@ -131,392 +418,889 @@ public class ForkJoinPool extends Abstra
new AtomicInteger();
/**
- * Array holding all worker threads in the pool. Array size must
- * be a power of two. Updates and replacements are protected by
- * workerLock, but it is always kept in a consistent enough state
- * to be randomly accessed without locking by workers performing
- * work-stealing.
+ * The time to block in a join (see awaitJoin) before checking if
+ * a new worker should be (re)started to maintain parallelism
+ * level. The value should be short enough to maintain global
+ * responsiveness and progress but long enough to avoid
+ * counterproductive firings during GC stalls or unrelated system
+ * activity, and to not bog down systems with continual re-firings
+ * on GCs or legitimately long waits.
*/
- volatile ForkJoinWorkerThread[] workers;
+ private static final long JOIN_TIMEOUT_MILLIS = 250L; // 4 per second
/**
- * Lock protecting access to workers.
+ * The wakeup interval (in nanoseconds) for the oldest worker
+ * worker waiting for an event invokes tryShutdownUnusedWorker to shrink
+ * the number of workers. The exact value does not matter too
+ * much, but should be long enough to slowly release resources
+ * during long periods without use without disrupting normal use.
*/
- private final ReentrantLock workerLock;
+ private static final long SHRINK_RATE_NANOS =
+ 30L * 1000L * 1000L * 1000L; // 2 per minute
/**
- * Condition for awaitTermination.
+ * Absolute bound for parallelism level. Twice this number plus
+ * one (i.e., 0xfff) must fit into a 16bit field to enable
+ * word-packing for some counts and indices.
*/
- private final Condition termination;
+ private static final int MAX_WORKERS = 0x7fff;
/**
- * The uncaught exception handler used when any worker
- * abrupty terminates
+ * Array holding all worker threads in the pool. Array size must
+ * be a power of two. Updates and replacements are protected by
+ * workerLock, but the array is always kept in a consistent enough
+ * state to be randomly accessed without locking by workers
+ * performing work-stealing, as well as other traversal-based
+ * methods in this class. All readers must tolerate that some
+ * array slots may be null.
*/
- private Thread.UncaughtExceptionHandler ueh;
+ volatile ForkJoinWorkerThread[] workers;
/**
- * Creation factory for worker threads.
+ * Queue for external submissions.
*/
- private final ForkJoinWorkerThreadFactory factory;
+ private final LinkedTransferQueue A {@code ManagedBlocker} provides two methods. Method
+ * {@code isReleasable} must return {@code true} if blocking is
+ * not necessary. Method {@code block} blocks the current thread
+ * if necessary (perhaps internally invoking {@code isReleasable}
+ * before actually blocking). The unusual methods in this API
+ * accommodate synchronizers that may, but don't usually, block
+ * for long periods. Similarly, they allow more efficient internal
+ * handling of cases in which additional workers may be, but
+ * usually are not, needed to ensure sufficient parallelism.
+ * Toward this end, implementations of method {@code isReleasable}
+ * must be amenable to repeated invocation.
+ *
* For example, here is a ManagedBlocker based on a
* ReentrantLock:
- * Here is a class that possibly blocks waiting for an
+ * item on a given queue:
+ * If the caller is not a ForkJoinTask, this method is behaviorally
- * equivalent to
- * If the caller is not a {@link ForkJoinTask}, this method is
+ * behaviorally equivalent to
+ * ("modifyThread")
,
+ * Final callback from terminating worker. Removes record of
+ * worker from array, and adjusts counts. If pool is shutting
+ * down, tries to complete termination.
+ *
+ * @param w the worker
*/
- public ForkJoinPool() {
- this(Runtime.getRuntime().availableProcessors(),
- defaultForkJoinWorkerThreadFactory);
+ final void workerTerminated(ForkJoinWorkerThread w) {
+ forgetWorker(w);
+ decrementWorkerCounts(w.isTrimmed()? 0 : ONE_RUNNING, ONE_TOTAL);
+ while (w.stealCount != 0) // collect final count
+ tryAccumulateStealCount(w);
+ tryTerminate(false);
}
+ // Waiting for and signalling events
+
/**
- * Creates a ForkJoinPool with the indicated parellelism level
- * threads, and using the default ForkJoinWorkerThreadFactory,
- * @param parallelism the number of worker threads
- * @throws IllegalArgumentException if parallelism less than or
- * equal to zero
- * @throws SecurityException if a security manager exists and
- * the caller is not permitted to modify threads
- * because it does not hold {@link
- * java.lang.RuntimePermission}("modifyThread")
,
+ * Releases workers blocked on a count not equal to current count.
+ * Normally called after precheck that eventWaiters isn't zero to
+ * avoid wasted array checks. Gives up upon a change in count or
+ * upon releasing two workers, letting others take over.
*/
- public ForkJoinPool(int parallelism) {
- this(parallelism, defaultForkJoinWorkerThreadFactory);
+ private void releaseEventWaiters() {
+ ForkJoinWorkerThread[] ws = workers;
+ int n = ws.length;
+ long h = eventWaiters;
+ int ec = eventCount;
+ boolean releasedOne = false;
+ ForkJoinWorkerThread w; int id;
+ while ((id = ((int)(h & WAITER_ID_MASK)) - 1) >= 0 &&
+ (int)(h >>> EVENT_COUNT_SHIFT) != ec &&
+ id < n && (w = ws[id]) != null) {
+ if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
+ h, w.nextWaiter)) {
+ LockSupport.unpark(w);
+ if (releasedOne) // exit on second release
+ break;
+ releasedOne = true;
+ }
+ if (eventCount != ec)
+ break;
+ h = eventWaiters;
+ }
}
/**
- * Creates a ForkJoinPool with parallelism equal to the number of
- * processors available on the system and using the given
- * ForkJoinWorkerThreadFactory,
- * @param factory the factory for creating new threads
- * @throws NullPointerException if factory is null
- * @throws SecurityException if a security manager exists and
- * the caller is not permitted to modify threads
- * because it does not hold {@link
- * java.lang.RuntimePermission}("modifyThread")
,
+ * Tries to advance eventCount and releases waiters. Called only
+ * from workers.
*/
- public ForkJoinPool(ForkJoinWorkerThreadFactory factory) {
- this(Runtime.getRuntime().availableProcessors(), factory);
+ final void signalWork() {
+ int c; // try to increment event count -- CAS failure OK
+ UNSAFE.compareAndSwapInt(this, eventCountOffset, c = eventCount, c+1);
+ if (eventWaiters != 0L)
+ releaseEventWaiters();
}
/**
- * Creates a ForkJoinPool with the given parallelism and factory.
+ * Adds the given worker to event queue and blocks until
+ * terminating or event count advances from the given value
*
- * @param parallelism the targeted number of worker threads
- * @param factory the factory for creating new threads
- * @throws IllegalArgumentException if parallelism less than or
- * equal to zero, or greater than implementation limit.
- * @throws NullPointerException if factory is null
- * @throws SecurityException if a security manager exists and
- * the caller is not permitted to modify threads
- * because it does not hold {@link
- * java.lang.RuntimePermission}("modifyThread")
,
+ * @param w the calling worker thread
+ * @param ec the count
*/
- public ForkJoinPool(int parallelism, ForkJoinWorkerThreadFactory factory) {
- if (parallelism <= 0 || parallelism > MAX_THREADS)
- throw new IllegalArgumentException();
- if (factory == null)
- throw new NullPointerException();
- checkPermission();
- this.factory = factory;
- this.parallelism = parallelism;
- this.maxPoolSize = MAX_THREADS;
- this.maintainsParallelism = true;
- this.poolNumber = poolNumberGenerator.incrementAndGet();
- this.workerLock = new ReentrantLock();
- this.termination = workerLock.newCondition();
- this.stealCount = new AtomicLong();
- this.submissionQueue = new LinkedTransferQueuenull
.
+ * @param asyncMode if true,
+ * establishes local first-in-first-out scheduling mode for forked
+ * tasks that are never joined. This mode may be more appropriate
+ * than default locally stack-based mode in applications in which
+ * worker threads only process event-style asynchronous tasks.
+ * For default value, use false
.
+ * @throws IllegalArgumentException if parallelism less than or
+ * equal to zero, or greater than implementation limit
+ * @throws NullPointerException if the factory is null
+ * @throws SecurityException if a security manager exists and
+ * the caller is not permitted to modify threads
+ * because it does not hold {@link
+ * java.lang.RuntimePermission}{@code ("modifyThread")}
+ */
+ public ForkJoinPool(int parallelism,
+ ForkJoinWorkerThreadFactory factory,
+ Thread.UncaughtExceptionHandler handler,
+ boolean asyncMode) {
+ checkPermission();
+ if (factory == null)
+ throw new NullPointerException();
+ if (parallelism <= 0 || parallelism > MAX_WORKERS)
+ throw new IllegalArgumentException();
+ this.parallelism = parallelism;
+ this.factory = factory;
+ this.ueh = handler;
+ this.locallyFifo = asyncMode;
+ int arraySize = initialArraySizeFor(parallelism);
+ this.workers = new ForkJoinWorkerThread[arraySize];
+ this.submissionQueue = new LinkedTransferQueue("modifyThread")
,
+ * @return the handler, or {@code null} if none
*/
- public Thread.UncaughtExceptionHandler
- setUncaughtExceptionHandler(Thread.UncaughtExceptionHandler h) {
- checkPermission();
- Thread.UncaughtExceptionHandler old = null;
- final ReentrantLock lock = this.workerLock;
- lock.lock();
- try {
- old = ueh;
- ueh = h;
- ForkJoinWorkerThread[] ws = workers;
- for (int i = 0; i < ws.length; ++i) {
- ForkJoinWorkerThread w = ws[i];
- if (w != null)
- w.setUncaughtExceptionHandler(h);
- }
- } finally {
- lock.unlock();
- }
- return old;
- }
-
-
- /**
- * Sets the target paralleism level of this pool.
- * @param parallelism the target parallelism
- * @throws IllegalArgumentException if parallelism less than or
- * equal to zero or greater than maximum size bounds.
- * @throws SecurityException if a security manager exists and
- * the caller is not permitted to modify threads
- * because it does not hold {@link
- * java.lang.RuntimePermission}("modifyThread")
,
- */
- public void setParallelism(int parallelism) {
- checkPermission();
- if (parallelism <= 0 || parallelism > maxPoolSize)
- throw new IllegalArgumentException();
- final ReentrantLock lock = this.workerLock;
- lock.lock();
- try {
- if (!isTerminating()) {
- int p = this.parallelism;
- this.parallelism = parallelism;
- if (parallelism > p)
- createAndStartAddedWorkers();
- else
- trimSpares();
- }
- } finally {
- lock.unlock();
- }
- signalIdleWorkers(false);
+ public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() {
+ return ueh;
}
/**
- * Returns the targeted number of worker threads in this pool.
+ * Returns the targeted parallelism level of this pool.
*
- * @return the targeted number of worker threads in this pool
+ * @return the targeted parallelism level of this pool
*/
public int getParallelism() {
return parallelism;
@@ -732,104 +1443,61 @@ public class ForkJoinPool extends Abstra
/**
* Returns the number of worker threads that have started but not
* yet terminated. This result returned by this method may differ
- * from getParallelism
when threads are created to
+ * from {@link #getParallelism} when threads are created to
* maintain parallelism when others are cooperatively blocked.
*
* @return the number of worker threads
*/
public int getPoolSize() {
- return totalCountOf(workerCounts);
- }
-
- /**
- * Returns the maximum number of threads allowed to exist in the
- * pool, even if there are insufficient unblocked running threads.
- * @return the maximum
- */
- public int getMaximumPoolSize() {
- return maxPoolSize;
+ return workerCounts >>> TOTAL_COUNT_SHIFT;
}
/**
- * Sets the maximum number of threads allowed to exist in the
- * pool, even if there are insufficient unblocked running threads.
- * Setting this value has no effect on current pool size. It
- * controls construction of new threads.
- * @throws IllegalArgumentException if negative or greater then
- * internal implementation limit.
- */
- public void setMaximumPoolSize(int newMax) {
- if (newMax < 0 || newMax > MAX_THREADS)
- throw new IllegalArgumentException();
- maxPoolSize = newMax;
- }
-
-
- /**
- * Returns true if this pool dynamically maintains its target
- * parallelism level. If false, new threads are added only to
- * avoid possible starvation.
- * This setting is by default true;
- * @return true if maintains parallelism
- */
- public boolean getMaintainsParallelism() {
- return maintainsParallelism;
- }
-
- /**
- * Sets whether this pool dynamically maintains its target
- * parallelism level. If false, new threads are added only to
- * avoid possible starvation.
- * @param enable true to maintains parallelism
+ * Returns {@code true} if this pool uses local first-in-first-out
+ * scheduling mode for forked tasks that are never joined.
+ *
+ * @return {@code true} if this pool uses async mode
*/
- public void setMaintainsParallelism(boolean enable) {
- maintainsParallelism = enable;
+ public boolean getAsyncMode() {
+ return locallyFifo;
}
/**
* Returns an estimate of the number of worker threads that are
* not blocked waiting to join tasks or for other managed
- * synchronization.
+ * synchronization. This method may overestimate the
+ * number of running threads.
*
* @return the number of worker threads
*/
public int getRunningThreadCount() {
- return runningCountOf(workerCounts);
+ return workerCounts & RUNNING_COUNT_MASK;
}
/**
* Returns an estimate of the number of threads that are currently
* stealing or executing tasks. This method may overestimate the
* number of active threads.
- * @return the number of active threads.
+ *
+ * @return the number of active threads
*/
public int getActiveThreadCount() {
- return activeCountOf(runControl);
- }
-
- /**
- * Returns an estimate of the number of threads that are currently
- * idle waiting for tasks. This method may underestimate the
- * number of idle threads.
- * @return the number of idle threads.
- */
- final int getIdleThreadCount() {
- int c = runningCountOf(workerCounts) - activeCountOf(runControl);
- return (c <= 0)? 0 : c;
+ return runState & ACTIVE_COUNT_MASK;
}
/**
- * Returns true if all worker threads are currently idle. An idle
- * worker is one that cannot obtain a task to execute because none
- * are available to steal from other threads, and there are no
- * pending submissions to the pool. This method is conservative:
- * It might not return true immediately upon idleness of all
- * threads, but will eventually become true if threads remain
- * inactive.
- * @return true if all threads are currently idle
+ * Returns {@code true} if all worker threads are currently idle.
+ * An idle worker is one that cannot obtain a task to execute
+ * because none are available to steal from other threads, and
+ * there are no pending submissions to the pool. This method is
+ * conservative; it might not return {@code true} immediately upon
+ * idleness of all threads, but will eventually become true if
+ * threads remain inactive.
+ *
+ * @return {@code true} if all threads are currently idle
*/
public boolean isQuiescent() {
- return activeCountOf(runControl) == 0;
+ return (runState & ACTIVE_COUNT_MASK) == 0;
}
/**
@@ -837,23 +1505,14 @@ public class ForkJoinPool extends Abstra
* one thread's work queue by another. The reported value
* underestimates the actual total number of steals when the pool
* is not quiescent. This value may be useful for monitoring and
- * tuning fork/join programs: In general, steal counts should be
+ * tuning fork/join programs: in general, steal counts should be
* high enough to keep threads busy, but low enough to avoid
* overhead and contention across threads.
- * @return the number of steals.
+ *
+ * @return the number of steals
*/
public long getStealCount() {
- return stealCount.get();
- }
-
- /**
- * Accumulate steal count from a worker. Call only
- * when worker known to be idle.
- */
- private void updateStealCount(ForkJoinWorkerThread w) {
- int sc = w.getAndClearStealCount();
- if (sc != 0)
- stealCount.addAndGet(sc);
+ return stealCount;
}
/**
@@ -863,33 +1522,37 @@ public class ForkJoinPool extends Abstra
* an approximation, obtained by iterating across all threads in
* the pool. This method may be useful for tuning task
* granularities.
- * @return the number of queued tasks.
+ *
+ * @return the number of queued tasks
*/
public long getQueuedTaskCount() {
long count = 0;
ForkJoinWorkerThread[] ws = workers;
- for (int i = 0; i < ws.length; ++i) {
- ForkJoinWorkerThread t = ws[i];
- if (t != null)
- count += t.getQueueSize();
+ int n = ws.length;
+ for (int i = 0; i < n; ++i) {
+ ForkJoinWorkerThread w = ws[i];
+ if (w != null)
+ count += w.getQueueSize();
}
return count;
}
/**
- * Returns an estimate of the number tasks submitted to this pool
- * that have not yet begun executing. This method takes time
+ * Returns an estimate of the number of tasks submitted to this
+ * pool that have not yet begun executing. This method takes time
* proportional to the number of submissions.
- * @return the number of queued submissions.
+ *
+ * @return the number of queued submissions
*/
public int getQueuedSubmissionCount() {
return submissionQueue.size();
}
/**
- * Returns true if there are any tasks submitted to this pool
- * that have not yet begun executing.
- * @return true
if there are any queued submissions.
+ * Returns {@code true} if there are any tasks submitted to this
+ * pool that have not yet begun executing.
+ *
+ * @return {@code true} if there are any queued submissions
*/
public boolean hasQueuedSubmissions() {
return !submissionQueue.isEmpty();
@@ -899,13 +1562,43 @@ public class ForkJoinPool extends Abstra
* Removes and returns the next unexecuted submission if one is
* available. This method may be useful in extensions to this
* class that re-assign work in systems with multiple pools.
- * @return the next submission, or null if none
+ *
+ * @return the next submission, or {@code null} if none
*/
protected ForkJoinTask> pollSubmission() {
return submissionQueue.poll();
}
/**
+ * Removes all available unexecuted submitted and forked tasks
+ * from scheduling queues and adds them to the given collection,
+ * without altering their execution status. These may include
+ * artificially generated or wrapped tasks. This method is
+ * designed to be invoked only when the pool is known to be
+ * quiescent. Invocations at other times may not remove all
+ * tasks. A failure encountered while attempting to add elements
+ * to collection {@code c} may result in elements being in
+ * neither, either or both collections when the associated
+ * exception is thrown. The behavior of this operation is
+ * undefined if the specified collection is modified while the
+ * operation is in progress.
+ *
+ * @param c the collection to transfer elements into
+ * @return the number of elements transferred
+ */
+ protected int drainTasksTo(Collection super ForkJoinTask>> c) {
+ int count = submissionQueue.drainTo(c);
+ ForkJoinWorkerThread[] ws = workers;
+ int n = ws.length;
+ for (int i = 0; i < n; ++i) {
+ ForkJoinWorkerThread w = ws[i];
+ if (w != null)
+ count += w.drainTasksTo(c);
+ }
+ return count;
+ }
+
+ /**
* Returns a string identifying this pool, as well as its state,
* including indications of run state, parallelism level, and
* worker and task counts.
@@ -913,99 +1606,104 @@ public class ForkJoinPool extends Abstra
* @return a string identifying this pool, as well as its state
*/
public String toString() {
- int ps = parallelism;
- int wc = workerCounts;
- int rc = runControl;
long st = getStealCount();
long qt = getQueuedTaskCount();
long qs = getQueuedSubmissionCount();
+ int wc = workerCounts;
+ int tc = wc >>> TOTAL_COUNT_SHIFT;
+ int rc = wc & RUNNING_COUNT_MASK;
+ int pc = parallelism;
+ int rs = runState;
+ int ac = rs & ACTIVE_COUNT_MASK;
return super.toString() +
- "[" + runStateToString(runStateOf(rc)) +
- ", parallelism = " + ps +
- ", size = " + totalCountOf(wc) +
- ", active = " + activeCountOf(rc) +
- ", running = " + runningCountOf(wc) +
+ "[" + runLevelToString(rs) +
+ ", parallelism = " + pc +
+ ", size = " + tc +
+ ", active = " + ac +
+ ", running = " + rc +
", steals = " + st +
", tasks = " + qt +
", submissions = " + qs +
"]";
}
- private static String runStateToString(int rs) {
- switch(rs) {
- case RUNNING: return "Running";
- case SHUTDOWN: return "Shutting down";
- case TERMINATING: return "Terminating";
- case TERMINATED: return "Terminated";
- default: throw new Error("Unknown run state");
- }
+ private static String runLevelToString(int s) {
+ return ((s & TERMINATED) != 0 ? "Terminated" :
+ ((s & TERMINATING) != 0 ? "Terminating" :
+ ((s & SHUTDOWN) != 0 ? "Shutting down" :
+ "Running")));
}
- // lifecycle control
-
/**
* Initiates an orderly shutdown in which previously submitted
* tasks are executed, but no new tasks will be accepted.
* Invocation has no additional effect if already shut down.
* Tasks that are in the process of being submitted concurrently
* during the course of this method may or may not be rejected.
+ *
* @throws SecurityException if a security manager exists and
* the caller is not permitted to modify threads
* because it does not hold {@link
- * java.lang.RuntimePermission}("modifyThread")
,
+ * java.lang.RuntimePermission}{@code ("modifyThread")}
*/
public void shutdown() {
checkPermission();
- transitionRunStateTo(SHUTDOWN);
- if (canTerminateOnShutdown(runControl))
- terminateOnShutdown();
+ advanceRunLevel(SHUTDOWN);
+ tryTerminate(false);
}
/**
- * Attempts to stop all actively executing tasks, and cancels all
- * waiting tasks. Tasks that are in the process of being
- * submitted or executed concurrently during the course of this
- * method may or may not be rejected. Unlike some other executors,
- * this method cancels rather than collects non-executed tasks,
- * so always returns an empty list.
+ * Attempts to cancel and/or stop all tasks, and reject all
+ * subsequently submitted tasks. Tasks that are in the process of
+ * being submitted or executed concurrently during the course of
+ * this method may or may not be rejected. This method cancels
+ * both existing and unexecuted tasks, in order to permit
+ * termination in the presence of task dependencies. So the method
+ * always returns an empty list (unlike the case for some other
+ * Executors).
+ *
* @return an empty list
* @throws SecurityException if a security manager exists and
* the caller is not permitted to modify threads
* because it does not hold {@link
- * java.lang.RuntimePermission}("modifyThread")
,
+ * java.lang.RuntimePermission}{@code ("modifyThread")}
*/
public Listtrue
if all tasks have completed following shut down.
+ * Returns {@code true} if all tasks have completed following shut down.
*
- * @return true
if all tasks have completed following shut down
+ * @return {@code true} if all tasks have completed following shut down
*/
public boolean isTerminated() {
- return runStateOf(runControl) == TERMINATED;
+ return runState >= TERMINATED;
}
/**
- * Returns true
if the process of termination has
- * commenced but possibly not yet completed.
+ * Returns {@code true} if the process of termination has
+ * commenced but not yet completed. This method may be useful for
+ * debugging. A return of {@code true} reported a sufficient
+ * period after shutdown may indicate that submitted tasks have
+ * ignored or suppressed interruption, causing this executor not
+ * to properly terminate.
*
- * @return true
if terminating
+ * @return {@code true} if terminating but not yet terminated
*/
public boolean isTerminating() {
- return runStateOf(runControl) >= TERMINATING;
+ return (runState & (TERMINATING|TERMINATED)) == TERMINATING;
}
/**
- * Returns true
if this pool has been shut down.
+ * Returns {@code true} if this pool has been shut down.
*
- * @return true
if this pool has been shut down
+ * @return {@code true} if this pool has been shut down
*/
public boolean isShutdown() {
- return runStateOf(runControl) >= SHUTDOWN;
+ return runState >= SHUTDOWN;
}
/**
@@ -1015,728 +1713,186 @@ public class ForkJoinPool extends Abstra
*
* @param timeout the maximum time to wait
* @param unit the time unit of the timeout argument
- * @return true
if this executor terminated and
- * false
if the timeout elapsed before termination
+ * @return {@code true} if this executor terminated and
+ * {@code false} if the timeout elapsed before termination
* @throws InterruptedException if interrupted while waiting
*/
public boolean awaitTermination(long timeout, TimeUnit unit)
throws InterruptedException {
- long nanos = unit.toNanos(timeout);
- final ReentrantLock lock = this.workerLock;
- lock.lock();
try {
- for (;;) {
- if (isTerminated())
- return true;
- if (nanos <= 0)
- return false;
- nanos = termination.awaitNanos(nanos);
- }
- } finally {
- lock.unlock();
- }
- }
-
- // Shutdown and termination support
-
- /**
- * Callback from terminating worker. Null out the corresponding
- * workers slot, and if terminating, try to terminate, else try to
- * shrink workers array.
- * @param w the worker
- */
- final void workerTerminated(ForkJoinWorkerThread w) {
- updateStealCount(w);
- updateWorkerCount(-1);
- final ReentrantLock lock = this.workerLock;
- lock.lock();
- try {
- ForkJoinWorkerThread[] ws = workers;
- int idx = w.poolIndex;
- if (idx >= 0 && idx < ws.length && ws[idx] == w)
- ws[idx] = null;
- if (totalCountOf(workerCounts) == 0) {
- terminate(); // no-op if already terminating
- transitionRunStateTo(TERMINATED);
- termination.signalAll();
- }
- else if (!isTerminating()) {
- tryShrinkWorkerArray();
- tryResumeSpare(true); // allow replacement
- }
- } finally {
- lock.unlock();
- }
- signalIdleWorkers(false);
- }
-
- /**
- * Initiate termination.
- */
- private void terminate() {
- if (transitionRunStateTo(TERMINATING)) {
- stopAllWorkers();
- resumeAllSpares();
- signalIdleWorkers(true);
- cancelQueuedSubmissions();
- cancelQueuedWorkerTasks();
- interruptUnterminatedWorkers();
- signalIdleWorkers(true); // resignal after interrupt
- }
- }
-
- /**
- * Possibly terminate when on shutdown state
- */
- private void terminateOnShutdown() {
- if (!hasQueuedSubmissions() && canTerminateOnShutdown(runControl))
- terminate();
- }
-
- /**
- * Clear out and cancel submissions
- */
- private void cancelQueuedSubmissions() {
- ForkJoinTask> task;
- while ((task = pollSubmission()) != null)
- task.cancel(false);
- }
-
- /**
- * Clean out worker queues.
- */
- private void cancelQueuedWorkerTasks() {
- final ReentrantLock lock = this.workerLock;
- lock.lock();
- try {
- ForkJoinWorkerThread[] ws = workers;
- for (int i = 0; i < ws.length; ++i) {
- ForkJoinWorkerThread t = ws[i];
- if (t != null)
- t.cancelTasks();
- }
- } finally {
- lock.unlock();
- }
- }
-
- /**
- * Set each worker's status to terminating. Requires lock to avoid
- * conflicts with add/remove
- */
- private void stopAllWorkers() {
- final ReentrantLock lock = this.workerLock;
- lock.lock();
- try {
- ForkJoinWorkerThread[] ws = workers;
- for (int i = 0; i < ws.length; ++i) {
- ForkJoinWorkerThread t = ws[i];
- if (t != null)
- t.shutdownNow();
- }
- } finally {
- lock.unlock();
- }
- }
-
- /**
- * Interrupt all unterminated workers. This is not required for
- * sake of internal control, but may help unstick user code during
- * shutdown.
- */
- private void interruptUnterminatedWorkers() {
- final ReentrantLock lock = this.workerLock;
- lock.lock();
- try {
- ForkJoinWorkerThread[] ws = workers;
- for (int i = 0; i < ws.length; ++i) {
- ForkJoinWorkerThread t = ws[i];
- if (t != null && !t.isTerminated()) {
- try {
- t.interrupt();
- } catch (SecurityException ignore) {
- }
- }
- }
- } finally {
- lock.unlock();
- }
- }
-
-
- /*
- * Nodes for event barrier to manage idle threads.
- *
- * The event barrier has an event count and a wait queue (actually
- * a Treiber stack). Workers are enabled to look for work when
- * the eventCount is incremented. If they fail to find some,
- * they may wait for next count. Synchronization events occur only
- * in enough contexts to maintain overall liveness:
- *
- * - Submission of a new task to the pool
- * - Creation or termination of a worker
- * - pool termination
- * - A worker pushing a task on an empty queue
- *
- * The last case (pushing a task) occurs often enough, and is
- * heavy enough compared to simple stack pushes to require some
- * special handling: Method signalNonEmptyWorkerQueue returns
- * without advancing count if the queue appears to be empty. This
- * would ordinarily result in races causing some queued waiters
- * not to be woken up. To avoid this, a worker in sync
- * rescans for tasks after being enqueued if it was the first to
- * enqueue, and aborts the wait if finding one, also helping to
- * signal others. This works well because the worker has nothing
- * better to do anyway, and so might as well help alleviate the
- * overhead and contention on the threads actually doing work.
- *
- * Queue nodes are basic Treiber stack nodes, also used for spare
- * stack.
- */
- static final class WaitQueueNode {
- WaitQueueNode next; // only written before enqueued
- volatile ForkJoinWorkerThread thread; // nulled to cancel wait
- final long count; // unused for spare stack
- WaitQueueNode(ForkJoinWorkerThread w, long c) {
- count = c;
- thread = w;
- }
- final boolean signal() {
- ForkJoinWorkerThread t = thread;
- thread = null;
- if (t != null) {
- LockSupport.unpark(t);
- return true;
- }
+ return termination.awaitAdvanceInterruptibly(0, timeout, unit) > 0;
+ } catch (TimeoutException ex) {
return false;
}
}
/**
- * Release at least one thread waiting for event count to advance,
- * if one exists. If initial attempt fails, release all threads.
- * @param all if false, at first try to only release one thread
- * @return current event
- */
- private long releaseIdleWorkers(boolean all) {
- long c;
- for (;;) {
- WaitQueueNode q = barrierStack;
- c = eventCount;
- long qc;
- if (q == null || (qc = q.count) >= c)
- break;
- if (!all) {
- if (casBarrierStack(q, q.next) && q.signal())
- break;
- all = true;
- }
- else if (casBarrierStack(q, null)) {
- do {
- q.signal();
- } while ((q = q.next) != null);
- break;
- }
- }
- return c;
- }
-
- /**
- * Returns current barrier event count
- * @return current barrier event count
- */
- final long getEventCount() {
- long ec = eventCount;
- releaseIdleWorkers(true); // release to ensure accurate result
- return ec;
- }
-
- /**
- * Increment event count and release at least one waiting thread,
- * if one exists (released threads will in turn wake up others).
- * @param all if true, try to wake up all
- */
- final void signalIdleWorkers(boolean all) {
- long c;
- do;while (!casEventCount(c = eventCount, c+1));
- releaseIdleWorkers(all);
- }
-
- /**
- * Wake up threads waiting to steal a task. Because method
- * sync rechecks availability, it is OK to only proceed if
- * queue appears to be non-empty.
- */
- final void signalNonEmptyWorkerQueue() {
- // If CAS fails another signaller must have succeeded
- long c;
- if (barrierStack != null && casEventCount(c = eventCount, c+1))
- releaseIdleWorkers(false);
- }
-
- /**
- * Waits until event count advances from count, or some thread is
- * waiting on a previous count, or there is stealable work
- * available. Help wake up others on release.
- * @param w the calling worker thread
- * @param prev previous value returned by sync (or 0)
- * @return current event count
- */
- final long sync(ForkJoinWorkerThread w, long prev) {
- updateStealCount(w);
-
- while (!w.isShutdown() && !isTerminating() &&
- (parallelism >= runningCountOf(workerCounts) ||
- !suspendIfSpare(w))) { // prefer suspend to waiting here
- WaitQueueNode node = null;
- boolean queued = false;
- for (;;) {
- if (!queued) {
- if (eventCount != prev)
- break;
- WaitQueueNode h = barrierStack;
- if (h != null && h.count != prev)
- break; // release below and maybe retry
- if (node == null)
- node = new WaitQueueNode(w, prev);
- queued = casBarrierStack(node.next = h, node);
- }
- else if (Thread.interrupted() ||
- node.thread == null ||
- (node.next == null && w.prescan()) ||
- eventCount != prev) {
- node.thread = null;
- if (eventCount == prev) // help trigger
- casEventCount(prev, prev+1);
- break;
- }
- else
- LockSupport.park(this);
- }
- long ec = eventCount;
- if (releaseIdleWorkers(false) != prev)
- return ec;
- }
- return prev; // return old count if aborted
- }
-
- // Parallelism maintenance
-
- /**
- * Decrement running count; if too low, add spare.
- *
- * Conceptually, all we need to do here is add or resume a
- * spare thread when one is about to block (and remove or
- * suspend it later when unblocked -- see suspendIfSpare).
- * However, implementing this idea requires coping with
- * several problems: We have imperfect information about the
- * states of threads. Some count updates can and usually do
- * lag run state changes, despite arrangements to keep them
- * accurate (for example, when possible, updating counts
- * before signalling or resuming), especially when running on
- * dynamic JVMs that don't optimize the infrequent paths that
- * update counts. Generating too many threads can make these
- * problems become worse, because excess threads are more
- * likely to be context-switched with others, slowing them all
- * down, especially if there is no work available, so all are
- * busy scanning or idling. Also, excess spare threads can
- * only be suspended or removed when they are idle, not
- * immediately when they aren't needed. So adding threads will
- * raise parallelism level for longer than necessary. Also,
- * FJ applications often enounter highly transient peaks when
- * many threads are blocked joining, but for less time than it
- * takes to create or resume spares.
- *
- * @param joinMe if non-null, return early if done
- * @param maintainParallelism if true, try to stay within
- * target counts, else create only to avoid starvation
- * @return true if joinMe known to be done
- */
- final boolean preJoin(ForkJoinTask> joinMe, boolean maintainParallelism) {
- maintainParallelism &= maintainsParallelism; // overrride
- boolean dec = false; // true when running count decremented
- while (spareStack == null || !tryResumeSpare(dec)) {
- int counts = workerCounts;
- if (dec || (dec = casWorkerCounts(counts, --counts))) { // CAS cheat
- if (!needSpare(counts, maintainParallelism))
- break;
- if (joinMe.status < 0)
- return true;
- if (tryAddSpare(counts))
- break;
- }
- }
- return false;
- }
-
- /**
- * Same idea as preJoin
- */
- final boolean preBlock(ManagedBlocker blocker, boolean maintainParallelism){
- maintainParallelism &= maintainsParallelism;
- boolean dec = false;
- while (spareStack == null || !tryResumeSpare(dec)) {
- int counts = workerCounts;
- if (dec || (dec = casWorkerCounts(counts, --counts))) {
- if (!needSpare(counts, maintainParallelism))
- break;
- if (blocker.isReleasable())
- return true;
- if (tryAddSpare(counts))
- break;
- }
- }
- return false;
- }
-
- /**
- * Returns true if a spare thread appears to be needed. If
- * maintaining parallelism, returns true when the deficit in
- * running threads is more than the surplus of total threads, and
- * there is apparently some work to do. This self-limiting rule
- * means that the more threads that have already been added, the
- * less parallelism we will tolerate before adding another.
- * @param counts current worker counts
- * @param maintainParallelism try to maintain parallelism
- */
- private boolean needSpare(int counts, boolean maintainParallelism) {
- int ps = parallelism;
- int rc = runningCountOf(counts);
- int tc = totalCountOf(counts);
- int runningDeficit = ps - rc;
- int totalSurplus = tc - ps;
- return (tc < maxPoolSize &&
- (rc == 0 || totalSurplus < 0 ||
- (maintainParallelism &&
- runningDeficit > totalSurplus && mayHaveQueuedWork())));
- }
-
- /**
- * Returns true if at least one worker queue appears to be
- * nonempty. This is expensive but not often called. It is not
- * critical that this be accurate, but if not, more or fewer
- * running threads than desired might be maintained.
- */
- private boolean mayHaveQueuedWork() {
- ForkJoinWorkerThread[] ws = workers;
- int len = ws.length;
- ForkJoinWorkerThread v;
- for (int i = 0; i < len; ++i) {
- if ((v = ws[i]) != null && v.getRawQueueSize() > 0) {
- releaseIdleWorkers(false); // help wake up stragglers
- return true;
- }
- }
- return false;
- }
-
- /**
- * Add a spare worker if lock available and no more than the
- * expected numbers of threads exist
- * @return true if successful
- */
- private boolean tryAddSpare(int expectedCounts) {
- final ReentrantLock lock = this.workerLock;
- int expectedRunning = runningCountOf(expectedCounts);
- int expectedTotal = totalCountOf(expectedCounts);
- boolean success = false;
- boolean locked = false;
- // confirm counts while locking; CAS after obtaining lock
- try {
- for (;;) {
- int s = workerCounts;
- int tc = totalCountOf(s);
- int rc = runningCountOf(s);
- if (rc > expectedRunning || tc > expectedTotal)
- break;
- if (!locked && !(locked = lock.tryLock()))
- break;
- if (casWorkerCounts(s, workerCountsFor(tc+1, rc+1))) {
- createAndStartSpare(tc);
- success = true;
- break;
- }
- }
- } finally {
- if (locked)
- lock.unlock();
- }
- return success;
- }
-
- /**
- * Add the kth spare worker. On entry, pool coounts are already
- * adjusted to reflect addition.
- */
- private void createAndStartSpare(int k) {
- ForkJoinWorkerThread w = null;
- ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(k + 1);
- int len = ws.length;
- // Probably, we can place at slot k. If not, find empty slot
- if (k < len && ws[k] != null) {
- for (k = 0; k < len && ws[k] != null; ++k)
- ;
- }
- if (k < len && !isTerminating() && (w = createWorker(k)) != null) {
- ws[k] = w;
- w.start();
- }
- else
- updateWorkerCount(-1); // adjust on failure
- signalIdleWorkers(false);
- }
-
- /**
- * Suspend calling thread w if there are excess threads. Called
- * only from sync. Spares are enqueued in a Treiber stack
- * using the same WaitQueueNodes as barriers. They are resumed
- * mainly in preJoin, but are also woken on pool events that
- * require all threads to check run state.
- * @param w the caller
- */
- private boolean suspendIfSpare(ForkJoinWorkerThread w) {
- WaitQueueNode node = null;
- int s;
- while (parallelism < runningCountOf(s = workerCounts)) {
- if (node == null)
- node = new WaitQueueNode(w, 0);
- if (casWorkerCounts(s, s-1)) { // representation-dependent
- // push onto stack
- do;while (!casSpareStack(node.next = spareStack, node));
-
- // block until released by resumeSpare
- while (node.thread != null) {
- if (!Thread.interrupted())
- LockSupport.park(this);
- }
- w.activate(); // help warm up
- return true;
- }
- }
- return false;
- }
-
- /**
- * Try to pop and resume a spare thread.
- * @param updateCount if true, increment running count on success
- * @return true if successful
- */
- private boolean tryResumeSpare(boolean updateCount) {
- WaitQueueNode q;
- while ((q = spareStack) != null) {
- if (casSpareStack(q, q.next)) {
- if (updateCount)
- updateRunningCount(1);
- q.signal();
- return true;
- }
- }
- return false;
- }
-
- /**
- * Pop and resume all spare threads. Same idea as
- * releaseIdleWorkers.
- * @return true if any spares released
- */
- private boolean resumeAllSpares() {
- WaitQueueNode q;
- while ( (q = spareStack) != null) {
- if (casSpareStack(q, null)) {
- do {
- updateRunningCount(1);
- q.signal();
- } while ((q = q.next) != null);
- return true;
- }
- }
- return false;
- }
-
- /**
- * Pop and shutdown excessive spare threads. Call only while
- * holding lock. This is not guaranteed to eliminate all excess
- * threads, only those suspended as spares, which are the ones
- * unlikely to be needed in the future.
- */
- private void trimSpares() {
- int surplus = totalCountOf(workerCounts) - parallelism;
- WaitQueueNode q;
- while (surplus > 0 && (q = spareStack) != null) {
- if (casSpareStack(q, null)) {
- do {
- updateRunningCount(1);
- ForkJoinWorkerThread w = q.thread;
- if (w != null && surplus > 0 &&
- runningCountOf(workerCounts) > 0 && w.shutdown())
- --surplus;
- q.signal();
- } while ((q = q.next) != null);
- }
- }
- }
-
- /**
- * Returns approximate number of spares, just for diagnostics.
- */
- private int countSpares() {
- int sum = 0;
- for (WaitQueueNode q = spareStack; q != null; q = q.next)
- ++sum;
- return sum;
- }
-
- /**
* Interface for extending managed parallelism for tasks running
- * in ForkJoinPools. A ManagedBlocker provides two methods.
- * Method isReleasable
must return true if blocking is not
- * necessary. Method block
blocks the current thread
- * if necessary (perhaps internally invoking isReleasable before
- * actually blocking.).
+ * in {@link ForkJoinPool}s.
+ *
+ *
- * class ManagedLocker implements ManagedBlocker {
- * final ReentrantLock lock;
- * boolean hasLock = false;
- * ManagedLocker(ReentrantLock lock) { this.lock = lock; }
- * public boolean block() {
- * if (!hasLock)
- * lock.lock();
- * return true;
- * }
- * public boolean isReleasable() {
- * return hasLock || (hasLock = lock.tryLock());
- * }
+ *
*/
public static interface ManagedBlocker {
/**
* Possibly blocks the current thread, for example waiting for
* a lock or condition.
- * @return true if no additional blocking is necessary (i.e.,
- * if isReleasable would return true).
+ *
+ * @return {@code true} if no additional blocking is necessary
+ * (i.e., if isReleasable would return true)
* @throws InterruptedException if interrupted while waiting
- * (the method is not required to do so, but is allowe to).
+ * (the method is not required to do so, but is allowed to)
*/
boolean block() throws InterruptedException;
/**
- * Returns true if blocking is unnecessary.
+ * Returns {@code true} if blocking is unnecessary.
*/
boolean isReleasable();
}
/**
* Blocks in accord with the given blocker. If the current thread
- * is a ForkJoinWorkerThread, this method possibly arranges for a
- * spare thread to be activated if necessary to ensure parallelism
- * while the current thread is blocked. If
- * {@code
+ * class ManagedLocker implements ManagedBlocker {
+ * final ReentrantLock lock;
+ * boolean hasLock = false;
+ * ManagedLocker(ReentrantLock lock) { this.lock = lock; }
+ * public boolean block() {
+ * if (!hasLock)
+ * lock.lock();
+ * return true;
+ * }
+ * public boolean isReleasable() {
+ * return hasLock || (hasLock = lock.tryLock());
+ * }
+ * }}
+ *
+ * {@code
+ * class QueueTaker
+ * public E getItem() { // call after pool.managedBlock completes
+ * return item;
+ * }
+ * }}maintainParallelism
is true and the pool supports
- * it ({@link #getMaintainsParallelism}), this method attempts to
- * maintain the pool's nominal parallelism. Otherwise if activates
- * a thread only if necessary to avoid complete starvation. This
- * option may be preferable when blockages use timeouts, or are
- * almost always brief.
- *
- *
- * while (!blocker.isReleasable())
- * if (blocker.block())
- * return;
- *
- * If the caller is a ForkJoinTask, then the pool may first
- * be expanded to ensure parallelism, and later adjusted.
+ * is a {@link ForkJoinWorkerThread}, this method possibly
+ * arranges for a spare thread to be activated if necessary to
+ * ensure sufficient parallelism while the current thread is blocked.
+ *
+ * {@code
+ * while (!blocker.isReleasable())
+ * if (blocker.block())
+ * return;
+ * }
+ *
+ * If the caller is a {@code ForkJoinTask}, then the pool may
+ * first be expanded to ensure parallelism, and later adjusted.
*
* @param blocker the blocker
- * @param maintainParallelism if true and supported by this pool,
- * attempt to maintain the pool's nominal parallelism; otherwise
- * activate a thread only if necessary to avoid complete
- * starvation.
- * @throws InterruptedException if blocker.block did so.
+ * @throws InterruptedException if blocker.block did so
*/
- public static void managedBlock(ManagedBlocker blocker,
- boolean maintainParallelism)
+ public static void managedBlock(ManagedBlocker blocker)
throws InterruptedException {
Thread t = Thread.currentThread();
- ForkJoinPool pool = (t instanceof ForkJoinWorkerThread?
- ((ForkJoinWorkerThread)t).pool : null);
- if (!blocker.isReleasable()) {
- try {
- if (pool == null ||
- !pool.preBlock(blocker, maintainParallelism))
- awaitBlocker(blocker);
- } finally {
- if (pool != null)
- pool.updateRunningCount(1);
- }
+ if (t instanceof ForkJoinWorkerThread) {
+ ForkJoinWorkerThread w = (ForkJoinWorkerThread) t;
+ w.pool.awaitBlocker(blocker);
+ }
+ else {
+ do {} while (!blocker.isReleasable() && !blocker.block());
}
}
- private static void awaitBlocker(ManagedBlocker blocker)
- throws InterruptedException {
- do;while (!blocker.isReleasable() && !blocker.block());
- }
-
- // AbstractExecutorService overrides
+ // AbstractExecutorService overrides. These rely on undocumented
+ // fact that ForkJoinTask.adapt returns ForkJoinTasks that also
+ // implement RunnableFuture.
protected