/* * 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/licenses/publicdomain */ package jsr166y; import java.util.concurrent.*; import java.io.Serializable; import java.util.Collection; import java.util.Collections; import java.util.List; import java.util.Map; import java.util.WeakHashMap; /** * Abstract base class for tasks that run within a {@link * ForkJoinPool}. A ForkJoinTask is a thread-like entity that is much * lighter weight than a normal thread. Huge numbers of tasks and * subtasks may be hosted by a small number of actual threads in a * ForkJoinPool, at the price of some usage limitations. * *
A "main" ForkJoinTask begins execution when submitted to a * {@link ForkJoinPool}. Once started, it will usually in turn start * other subtasks. As indicated by the name of this class, many * programs using ForkJoinTasks employ only methods {@code fork} and * {@code join}, or derivatives such as {@code invokeAll}. However, * this class also provides a number of other methods that can come * into play in advanced usages, as well as extension mechanics that * allow support of new forms of fork/join processing. * *
A ForkJoinTask is a lightweight form of {@link Future}. The * efficiency of ForkJoinTasks stems from a set of restrictions (that * are only partially statically enforceable) reflecting their * intended use as computational tasks calculating pure functions or * operating on purely isolated objects. The primary coordination * mechanisms are {@link #fork}, that arranges asynchronous execution, * and {@link #join}, that doesn't proceed until the task's result has * been computed. Computations should avoid {@code synchronized} * methods or blocks, and should minimize other blocking * synchronization apart from joining other tasks or using * synchronizers such as Phasers that are advertised to cooperate with * fork/join scheduling. Tasks should also not perform blocking IO, * and should ideally access variables that are completely independent * of those accessed by other running tasks. Minor breaches of these * restrictions, for example using shared output streams, may be * tolerable in practice, but frequent use may result in poor * performance, and the potential to indefinitely stall if the number * of threads not waiting for IO or other external synchronization * becomes exhausted. This usage restriction is in part enforced by * not permitting checked exceptions such as {@code IOExceptions} * to be thrown. However, computations may still encounter unchecked * exceptions, that are rethrown to callers attempting join * them. These exceptions may additionally include * RejectedExecutionExceptions stemming from internal resource * exhaustion such as failure to allocate internal task queues. * *
The primary method for awaiting completion and extracting * results of a task is {@link #join}, but there are several variants: * The {@link Future#get} methods support interruptible and/or timed * waits for completion and report results using {@code Future} * conventions. Method {@link #helpJoin} enables callers to actively * execute other tasks while awaiting joins, which is sometimes more * efficient but only applies when all subtasks are known to be * strictly tree-structured. Method {@link #invoke} is semantically * equivalent to {@code fork(); join()} but always attempts to * begin execution in the current thread. The "quiet" forms * of these methods do not extract results or report exceptions. These * may be useful when a set of tasks are being executed, and you need * to delay processing of results or exceptions until all complete. * Method {@code invokeAll} (available in multiple versions) * performs the most common form of parallel invocation: forking a set * of tasks and joining them all. * *
The ForkJoinTask class is not usually directly subclassed. * Instead, you subclass one of the abstract classes that support a * particular style of fork/join processing. Normally, a concrete * ForkJoinTask subclass declares fields comprising its parameters, * established in a constructor, and then defines a {@code compute} * method that somehow uses the control methods supplied by this base * class. While these methods have {@code public} access (to allow * instances of different task subclasses to call each others * methods), some of them may only be called from within other * ForkJoinTasks (as may be determined using method {@link * #inForkJoinPool}). Attempts to invoke them in other contexts * result in exceptions or errors, possibly including * ClassCastException. * *
Most base support methods are {@code final} because their * implementations are intrinsically tied to the underlying * lightweight task scheduling framework, and so cannot be overridden. * Developers creating new basic styles of fork/join processing should * minimally implement {@code protected} methods * {@link #exec}, {@link #setRawResult}, and * {@link #getRawResult}, while also introducing an abstract * computational method that can be implemented in its subclasses, * possibly relying on other {@code protected} methods provided * by this class. * *
ForkJoinTasks should perform relatively small amounts of * computations, otherwise splitting into smaller tasks. As a very * rough rule of thumb, a task should perform more than 100 and less * than 10000 basic computational steps. If tasks are too big, then * parallelism cannot improve throughput. If too small, then memory * and internal task maintenance overhead may overwhelm processing. * *
ForkJoinTasks are {@code Serializable}, which enables them
* to be used in extensions such as remote execution frameworks. It is
* in general sensible to serialize tasks only before or after, but
* not during execution. Serialization is not relied on during
* execution itself.
*
* @since 1.7
* @author Doug Lea
*/
public abstract class ForkJoinTask This method may be overridden in subclasses, but if so, must
* still ensure that these minimal properties hold. In particular,
* the cancel method itself must not throw exceptions.
*
* This method is designed to be invoked by other
* tasks. To terminate the current task, you can just return or
* throw an unchecked exception from its computation method, or
* invoke {@link #completeExceptionally}.
*
* @param mayInterruptIfRunning this value is ignored in the
* default implementation because tasks are not in general
* cancelled via interruption
*
* @return {@code true} if this task is now cancelled
*/
public boolean cancel(boolean mayInterruptIfRunning) {
setCompletion(CANCELLED);
return (status & COMPLETION_MASK) == CANCELLED;
}
/**
* Returns {@code true} if this task threw an exception or was cancelled.
*
* @return {@code true} if this task threw an exception or was cancelled
*/
public final boolean isCompletedAbnormally() {
return (status & COMPLETION_MASK) < NORMAL;
}
/**
* Returns the exception thrown by the base computation, or a
* CancellationException if cancelled, or null if none or if the
* method has not yet completed.
*
* @return the exception, or {@code null} if none
*/
public final Throwable getException() {
int s = status & COMPLETION_MASK;
if (s >= NORMAL)
return null;
if (s == CANCELLED)
return new CancellationException();
return exceptionMap.get(this);
}
/**
* Completes this task abnormally, and if not already aborted or
* cancelled, causes it to throw the given exception upon
* {@code join} and related operations. This method may be used
* to induce exceptions in asynchronous tasks, or to force
* completion of tasks that would not otherwise complete. Its use
* in other situations is likely to be wrong. This method is
* overridable, but overridden versions must invoke {@code super}
* implementation to maintain guarantees.
*
* @param ex the exception to throw. If this exception is
* not a RuntimeException or Error, the actual exception thrown
* will be a RuntimeException with cause ex.
*/
public void completeExceptionally(Throwable ex) {
setDoneExceptionally((ex instanceof RuntimeException) ||
(ex instanceof Error) ? ex :
new RuntimeException(ex));
}
/**
* Completes this task, and if not already aborted or cancelled,
* returning a {@code null} result upon {@code join} and related
* operations. This method may be used to provide results for
* asynchronous tasks, or to provide alternative handling for
* tasks that would not otherwise complete normally. Its use in
* other situations is likely to be wrong. This method is
* overridable, but overridden versions must invoke {@code super}
* implementation to maintain guarantees.
*
* @param value the result value for this task
*/
public void complete(V value) {
try {
setRawResult(value);
} catch (Throwable rex) {
setDoneExceptionally(rex);
return;
}
setNormalCompletion();
}
public final V get() throws InterruptedException, ExecutionException {
ForkJoinWorkerThread w = getWorker();
if (w == null || status < 0 || !w.unpushTask(this) || !tryQuietlyInvoke())
awaitDone(w, true);
return reportFutureResult();
}
public final V get(long timeout, TimeUnit unit)
throws InterruptedException, ExecutionException, TimeoutException {
long nanos = unit.toNanos(timeout);
ForkJoinWorkerThread w = getWorker();
if (w == null || status < 0 || !w.unpushTask(this) || !tryQuietlyInvoke())
awaitDone(w, nanos);
return reportTimedFutureResult();
}
/**
* Possibly executes other tasks until this task is ready, then
* returns the result of the computation. This method may be more
* efficient than {@code join}, but is only applicable when
* there are no potential dependencies between continuation of the
* current task and that of any other task that might be executed
* while helping. (This usually holds for pure divide-and-conquer
* tasks). This method may be invoked only from within
* ForkJoinTask computations (as may be determined using method
* {@link #inForkJoinPool}). Attempts to invoke in other contexts
* result in exceptions or errors, possibly including
* ClassCastException.
*
* @return the computed result
*/
public final V helpJoin() {
ForkJoinWorkerThread w = (ForkJoinWorkerThread) Thread.currentThread();
if (status < 0 || !w.unpushTask(this) || !tryExec())
reportException(busyJoin(w));
return getRawResult();
}
/**
* Possibly executes other tasks until this task is ready. This
* method may be invoked only from within ForkJoinTask
* computations (as may be determined using method {@link
* #inForkJoinPool}). Attempts to invoke in other contexts result
* in exceptions or errors, possibly including ClassCastException.
*/
public final void quietlyHelpJoin() {
if (status >= 0) {
ForkJoinWorkerThread w =
(ForkJoinWorkerThread) Thread.currentThread();
if (!w.unpushTask(this) || !tryQuietlyInvoke())
busyJoin(w);
}
}
/**
* Joins this task, without returning its result or throwing an
* exception. This method may be useful when processing
* collections of tasks when some have been cancelled or otherwise
* known to have aborted.
*/
public final void quietlyJoin() {
if (status >= 0) {
ForkJoinWorkerThread w = getWorker();
if (w == null || !w.unpushTask(this) || !tryQuietlyInvoke())
awaitDone(w, true);
}
}
/**
* Commences performing this task and awaits its completion if
* necessary, without returning its result or throwing an
* exception. This method may be useful when processing
* collections of tasks when some have been cancelled or otherwise
* known to have aborted.
*/
public final void quietlyInvoke() {
if (status >= 0 && !tryQuietlyInvoke())
quietlyJoin();
}
/**
* Possibly executes tasks until the pool hosting the current task
* {@link ForkJoinPool#isQuiescent}. This method may be of use in
* designs in which many tasks are forked, but none are explicitly
* joined, instead executing them until all are processed.
*/
public static void helpQuiesce() {
((ForkJoinWorkerThread) Thread.currentThread())
.helpQuiescePool();
}
/**
* Resets the internal bookkeeping state of this task, allowing a
* subsequent {@code fork}. This method allows repeated reuse of
* this task, but only if reuse occurs when this task has either
* never been forked, or has been forked, then completed and all
* outstanding joins of this task have also completed. Effects
* under any other usage conditions are not guaranteed, and are
* almost surely wrong. This method may be useful when executing
* pre-constructed trees of subtasks in loops.
*/
public void reinitialize() {
if ((status & COMPLETION_MASK) == EXCEPTIONAL)
exceptionMap.remove(this);
status = 0;
}
/**
* Returns the pool hosting the current task execution, or null
* if this task is executing outside of any ForkJoinPool.
*
* @return the pool, or {@code null} if none
*/
public static ForkJoinPool getPool() {
Thread t = Thread.currentThread();
return (t instanceof ForkJoinWorkerThread) ?
((ForkJoinWorkerThread) t).pool : null;
}
/**
* Returns {@code true} if the current thread is executing as a
* ForkJoinPool computation.
*
* @return {@code true} if the current thread is executing as a
* ForkJoinPool computation, or false otherwise
*/
public static boolean inForkJoinPool() {
return Thread.currentThread() instanceof ForkJoinWorkerThread;
}
/**
* Tries to unschedule this task for execution. This method will
* typically succeed if this task is the most recently forked task
* by the current thread, and has not commenced executing in
* another thread. This method may be useful when arranging
* alternative local processing of tasks that could have been, but
* were not, stolen. This method may be invoked only from within
* ForkJoinTask computations (as may be determined using method
* {@link #inForkJoinPool}). Attempts to invoke in other contexts
* result in exceptions or errors, possibly including
* ClassCastException.
*
* @return {@code true} if unforked
*/
public boolean tryUnfork() {
return ((ForkJoinWorkerThread) Thread.currentThread())
.unpushTask(this);
}
/**
* Returns an estimate of the number of tasks that have been
* forked by the current worker thread but not yet executed. This
* value may be useful for heuristic decisions about whether to
* fork other tasks.
*
* @return the number of tasks
*/
public static int getQueuedTaskCount() {
return ((ForkJoinWorkerThread) Thread.currentThread())
.getQueueSize();
}
/**
* Returns an estimate of how many more locally queued tasks are
* held by the current worker thread than there are other worker
* threads that might steal them. This value may be useful for
* heuristic decisions about whether to fork other tasks. In many
* usages of ForkJoinTasks, at steady state, each worker should
* aim to maintain a small constant surplus (for example, 3) of
* tasks, and to process computations locally if this threshold is
* exceeded.
*
* @return the surplus number of tasks, which may be negative
*/
public static int getSurplusQueuedTaskCount() {
return ((ForkJoinWorkerThread) Thread.currentThread())
.getEstimatedSurplusTaskCount();
}
// Extension methods
/**
* Returns the result that would be returned by {@link #join}, even
* if this task completed abnormally, or {@code null} if this task
* is not known to have been completed. This method is designed
* to aid debugging, as well as to support extensions. Its use in
* any other context is discouraged.
*
* @return the result, or {@code null} if not completed
*/
public abstract V getRawResult();
/**
* Forces the given value to be returned as a result. This method
* is designed to support extensions, and should not in general be
* called otherwise.
*
* @param value the value
*/
protected abstract void setRawResult(V value);
/**
* Immediately performs the base action of this task. This method
* is designed to support extensions, and should not in general be
* called otherwise. The return value controls whether this task
* is considered to be done normally. It may return false in
* asynchronous actions that require explicit invocations of
* {@link #complete} to become joinable. It may throw exceptions
* to indicate abnormal exit.
*
* @return {@code true} if completed normally
* @throws Error or RuntimeException if encountered during computation
*/
protected abstract boolean exec();
/**
* Returns, but does not unschedule or execute, a task queued by
* the current thread but not yet executed, if one is immediately
* available. There is no guarantee that this task will actually
* be polled or executed next. Conversely, this method may return
* null even if a task exists but cannot be accessed without
* contention with other threads. This method is designed
* primarily to support extensions, and is unlikely to be useful
* otherwise. This method may be invoked only from within
* ForkJoinTask computations (as may be determined using method
* {@link #inForkJoinPool}). Attempts to invoke in other contexts
* result in exceptions or errors, possibly including
* ClassCastException.
*
* @return the next task, or {@code null} if none are available
*/
protected static ForkJoinTask> peekNextLocalTask() {
return ((ForkJoinWorkerThread) Thread.currentThread())
.peekTask();
}
/**
* Unschedules and returns, without executing, the next task
* queued by the current thread but not yet executed. This method
* is designed primarily to support extensions, and is unlikely to
* be useful otherwise. This method may be invoked only from
* within ForkJoinTask computations (as may be determined using
* method {@link #inForkJoinPool}). Attempts to invoke in other
* contexts result in exceptions or errors, possibly including
* ClassCastException.
*
* @return the next task, or {@code null} if none are available
*/
protected static ForkJoinTask> pollNextLocalTask() {
return ((ForkJoinWorkerThread) Thread.currentThread())
.pollLocalTask();
}
/**
* Unschedules and returns, without executing, the next task
* queued by the current thread but not yet executed, if one is
* available, or if not available, a task that was forked by some
* other thread, if available. Availability may be transient, so a
* {@code null} result does not necessarily imply quiescence
* of the pool this task is operating in. This method is designed
* primarily to support extensions, and is unlikely to be useful
* otherwise. This method may be invoked only from within
* ForkJoinTask computations (as may be determined using method
* {@link #inForkJoinPool}). Attempts to invoke in other contexts
* result in exceptions or errors, possibly including
* ClassCastException.
*
* @return a task, or {@code null} if none are available
*/
protected static ForkJoinTask> pollTask() {
return ((ForkJoinWorkerThread) Thread.currentThread())
.pollTask();
}
/**
* Adaptor for Runnables. This implements RunnableFuture
* to be compliant with AbstractExecutorService constraints
* when used in ForkJoinPool.
*/
static final class AdaptedRunnable