--- jsr166/src/jsr166y/ForkJoinPool.java 2009/07/20 22:26:03 1.9
+++ jsr166/src/jsr166y/ForkJoinPool.java 2010/08/17 18:30:32 1.64
@@ -5,99 +5,386 @@
*/
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. When setting
- * {@code setAsyncMode}, a ForkJoinPools may also be appropriate for
- * use with fine-grained tasks that are never joined. 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 {@code ManagedBlocker} interface enables extension of
- * the kinds of synchronization accommodated. The target parallelism
- * level may also be changed dynamically ({@code setParallelism})
- * and thread construction can be limited using methods
- * {@code setMaximumPoolSize} and/or
- * {@code 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
- * {@code 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
- * {@code 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 |
+ *
+ *
+ * | Arange 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.
- */
-
- /** 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.
+ * 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. Becauae 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 or
+ * re-activate a spare thread to compensate for blocked
+ * joiners until they unblock.
+ *
+ * Because the 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, we rely on multiple retries of
+ * each. Further, because it is impossible to keep exactly the
+ * target (parallelism) number of threads running at any given
+ * time, we allow compensation during joins to fail, and enlist
+ * all other threads to help out whenever they are not otherwise
+ * occupied (i.e., mainly in method preStep).
+ *
+ * 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 artifically 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
+ * correspondance 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 others (but give up upon
+ * contention to reduce useless flailing). 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 is about
+ * to block waiting for a join (or via ManagedBlockers), 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. The Spare and Event queues use
+ * similar mechanisms to shed unused workers: The oldest (first)
+ * 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 waiter via tryShutdownSpare or
+ * tryShutdownWaiter, respectively. The wakeup rates for spares
+ * are much shorter than for waiters. Together, they will
+ * eventually reduce 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 impossble
+ * 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 the
+ * presence of transients 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 some slack in
+ * creation thresholds, using rules that reflect the fact that the
+ * more threads we have running, the more likely that we are
+ * underestimating the number running threads. (We also include
+ * some heuristic use of Thread.yield when all workers appear to
+ * be busy, to improve likelihood of counts settling.) The rules
+ * also better cope 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).
+ */
+
+ /**
+ * 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);
}
}
@@ -133,29 +420,47 @@ public class ForkJoinPool extends Abstra
new AtomicInteger();
/**
- * Array holding all worker threads in the pool. Initialized upon
- * first use. 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 wakeup interval (in nanoseconds) for the oldest worker
+ * worker waiting for an event invokes tryShutdownWaiter 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 static final long SHRINK_RATE_NANOS =
+ 60L * 1000L * 1000L * 1000L; // one minute
+
+ /**
+ * 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 static final int MAX_WORKERS = 0x7fff;
+
+ /**
+ * 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.
*/
volatile ForkJoinWorkerThread[] workers;
/**
- * Lock protecting access to workers.
+ * Queue for external submissions.
*/
- private final ReentrantLock workerLock;
+ private final LinkedTransferQueue> submissionQueue;
/**
- * Condition for awaitTermination.
+ * Lock protecting updates to workers array.
*/
- private final Condition termination;
+ private final ReentrantLock workerLock;
/**
- * The uncaught exception handler used when any worker
- * abruptly terminates
+ * Latch released upon termination.
*/
- private Thread.UncaughtExceptionHandler ueh;
+ private final Phaser termination;
/**
* Creation factory for worker threads.
@@ -163,384 +468,926 @@ public class ForkJoinPool extends Abstra
private final ForkJoinWorkerThreadFactory factory;
/**
- * Head of stack of threads that were created to maintain
- * parallelism when other threads blocked, but have since
- * suspended when the parallelism level rose.
- */
- private volatile WaitQueueNode spareStack;
-
- /**
* Sum of per-thread steal counts, updated only when threads are
* idle or terminating.
*/
- private final AtomicLong stealCount;
+ private volatile long stealCount;
/**
- * Queue for external submissions.
+ * Encoded record of top of treiber stack of threads waiting for
+ * events. The top 32 bits contain the count being waited for. The
+ * bottom 16 bits contains one plus the pool index of waiting
+ * worker thread. (Bits 16-31 are unused.)
*/
- private final LinkedTransferQueue> submissionQueue;
+ private volatile long eventWaiters;
+
+ private static final int EVENT_COUNT_SHIFT = 32;
+ private static final long WAITER_ID_MASK = (1L << 16) - 1L;
/**
- * Head of Treiber stack for barrier sync. See below for explanation
+ * A counter for events that may wake up worker threads:
+ * - Submission of a new task to the pool
+ * - A worker pushing a task on an empty queue
+ * - termination
*/
- private volatile WaitQueueNode syncStack;
+ private volatile int eventCount;
/**
- * The count for event barrier
+ * Encoded record of top of treiber stack of spare threads waiting
+ * for resumption. The top 16 bits contain an arbitrary count to
+ * avoid ABA effects. The bottom 16bits contains one plus the pool
+ * index of waiting worker thread.
*/
- private volatile long eventCount;
+ private volatile int spareWaiters;
+
+ private static final int SPARE_COUNT_SHIFT = 16;
+ private static final int SPARE_ID_MASK = (1 << 16) - 1;
/**
- * Pool number, just for assigning useful names to worker threads
+ * Lifecycle control. The low word contains the number of workers
+ * that are (probably) executing tasks. This value is atomically
+ * incremented before a worker gets a task to run, and decremented
+ * when worker has no tasks and cannot find any. Bits 16-18
+ * contain runLevel value. When all are zero, the pool is
+ * running. Level transitions are monotonic (running -> shutdown
+ * -> terminating -> terminated) so each transition adds a bit.
+ * These are bundled together to ensure consistent read for
+ * termination checks (i.e., that runLevel is at least SHUTDOWN
+ * and active threads is zero).
+ *
+ * Notes: Most direct CASes are dependent on these bitfield
+ * positions. Also, this field is non-private to enable direct
+ * performance-sensitive CASes in ForkJoinWorkerThread.
*/
- private final int poolNumber;
+ volatile int runState;
+
+ // Note: The order among run level values matters.
+ private static final int RUNLEVEL_SHIFT = 16;
+ private static final int SHUTDOWN = 1 << RUNLEVEL_SHIFT;
+ private static final int TERMINATING = 1 << (RUNLEVEL_SHIFT + 1);
+ private static final int TERMINATED = 1 << (RUNLEVEL_SHIFT + 2);
+ private static final int ACTIVE_COUNT_MASK = (1 << RUNLEVEL_SHIFT) - 1;
/**
- * The maximum allowed pool size
+ * Holds number of total (i.e., created and not yet terminated)
+ * and running (i.e., not blocked on joins or other managed sync)
+ * threads, packed together to ensure consistent snapshot when
+ * making decisions about creating and suspending spare
+ * threads. Updated only by CAS. Note that adding a new worker
+ * requires incrementing both counts, since workers start off in
+ * running state.
*/
- private volatile int maxPoolSize;
+ private volatile int workerCounts;
+
+ private static final int TOTAL_COUNT_SHIFT = 16;
+ private static final int RUNNING_COUNT_MASK = (1 << TOTAL_COUNT_SHIFT) - 1;
+ private static final int ONE_RUNNING = 1;
+ private static final int ONE_TOTAL = 1 << TOTAL_COUNT_SHIFT;
/**
- * The desired parallelism level, updated only under workerLock.
+ * The target parallelism level.
+ * Accessed directly by ForkJoinWorkerThreads.
*/
- private volatile int parallelism;
+ final int parallelism;
/**
* True if use local fifo, not default lifo, for local polling
+ * Read by, and replicated by ForkJoinWorkerThreads
*/
- private volatile boolean locallyFifo;
+ final boolean locallyFifo;
/**
- * Holds number of total (i.e., created and not yet terminated)
- * and running (i.e., not blocked on joins or other managed sync)
- * threads, packed into one int to ensure consistent snapshot when
- * making decisions about creating and suspending spare
- * threads. Updated only by CAS. Note: CASes in
- * updateRunningCount and preJoin running active count is in low
- * word, so need to be modified if this changes
+ * The uncaught exception handler used when any worker abruptly
+ * terminates.
*/
- private volatile int workerCounts;
+ private final Thread.UncaughtExceptionHandler ueh;
- private static int totalCountOf(int s) { return s >>> 16; }
- private static int runningCountOf(int s) { return s & shortMask; }
- private static int workerCountsFor(int t, int r) { return (t << 16) + r; }
+ /**
+ * Pool number, just for assigning useful names to worker threads
+ */
+ private final int poolNumber;
+
+
+ // Utilities for CASing fields. Note that most of these
+ // are usually manually inlined by callers
/**
- * Add delta (which may be negative) to running count. This must
- * be called before (with negative arg) and after (with positive)
- * any managed synchronization (i.e., mainly, joins)
- * @param delta the number to add
+ * Increments running count part of workerCounts
*/
- final void updateRunningCount(int delta) {
- int s;
- do;while (!casWorkerCounts(s = workerCounts, s + delta));
+ final void incrementRunningCount() {
+ int c;
+ do {} while (!UNSAFE.compareAndSwapInt(this, workerCountsOffset,
+ c = workerCounts,
+ c + ONE_RUNNING));
}
/**
- * Add delta (which may be negative) to both total and running
- * count. This must be called upon creation and termination of
- * worker threads.
- * @param delta the number to add
+ * Tries to decrement running count unless already zero
*/
- private void updateWorkerCount(int delta) {
- int d = delta + (delta << 16); // add to both lo and hi parts
- int s;
- do;while (!casWorkerCounts(s = workerCounts, s + d));
+ final boolean tryDecrementRunningCount() {
+ int wc = workerCounts;
+ if ((wc & RUNNING_COUNT_MASK) == 0)
+ return false;
+ return UNSAFE.compareAndSwapInt(this, workerCountsOffset,
+ wc, wc - ONE_RUNNING);
}
/**
- * Lifecycle control. High word contains runState, low word
- * contains the number of workers that are (probably) executing
- * tasks. This value is atomically incremented before a worker
- * gets a task to run, and decremented when worker has no tasks
- * and cannot find any. These two fields are bundled together to
- * support correct termination triggering. Note: activeCount
- * CAS'es cheat by assuming active count is in low word, so need
- * to be modified if this changes
+ * Forces decrement of encoded workerCounts, awaiting nonzero if
+ * (rarely) necessary when other count updates lag.
+ *
+ * @param dr -- either zero or ONE_RUNNING
+ * @param dt == either zero or ONE_TOTAL
*/
- private volatile int runControl;
-
- // RunState values. Order among values matters
- private static final int RUNNING = 0;
- private static final int SHUTDOWN = 1;
- private static final int TERMINATING = 2;
- private static final int TERMINATED = 3;
-
- private static int runStateOf(int c) { return c >>> 16; }
- private static int activeCountOf(int c) { return c & shortMask; }
- private static int runControlFor(int r, int a) { return (r << 16) + a; }
+ private void decrementWorkerCounts(int dr, int dt) {
+ for (;;) {
+ int wc = workerCounts;
+ if ((wc & RUNNING_COUNT_MASK) - dr < 0 ||
+ (wc >>> TOTAL_COUNT_SHIFT) - dt < 0) {
+ if ((runState & TERMINATED) != 0)
+ return; // lagging termination on a backout
+ Thread.yield();
+ }
+ if (UNSAFE.compareAndSwapInt(this, workerCountsOffset,
+ wc, wc - (dr + dt)))
+ return;
+ }
+ }
/**
- * Try incrementing active count; fail on contention. Called by
- * workers before/during executing tasks.
- * @return true on success;
+ * Increments event count
*/
- final boolean tryIncrementActiveCount() {
- int c = runControl;
- return casRunControl(c, c+1);
+ private void advanceEventCount() {
+ int c;
+ do {} while(!UNSAFE.compareAndSwapInt(this, eventCountOffset,
+ c = eventCount, c+1));
}
/**
- * Try decrementing active count; fail on contention.
- * Possibly trigger termination on success
- * Called by workers when they can't find tasks.
+ * Tries incrementing active count; fails on contention.
+ * Called by workers before executing tasks.
+ *
* @return true on success
*/
- final boolean tryDecrementActiveCount() {
- int c = runControl;
- int nextc = c - 1;
- if (!casRunControl(c, nextc))
- return false;
- if (canTerminateOnShutdown(nextc))
- terminateOnShutdown();
- return true;
+ final boolean tryIncrementActiveCount() {
+ int c;
+ return UNSAFE.compareAndSwapInt(this, runStateOffset,
+ c = runState, c + 1);
}
/**
- * Return true if argument represents zero active count and
- * nonzero runstate, which is the triggering condition for
- * terminating on shutdown.
+ * Tries decrementing active count; fails on contention.
+ * Called when workers cannot find tasks to run.
*/
- private static boolean canTerminateOnShutdown(int c) {
- return ((c & -c) >>> 16) != 0; // i.e. least bit is nonzero runState bit
+ final boolean tryDecrementActiveCount() {
+ int c;
+ return UNSAFE.compareAndSwapInt(this, runStateOffset,
+ c = runState, c - 1);
}
/**
- * Transition run state to at least the given state. Return true
- * if not already at least given state.
+ * Advances to at least the given level. Returns true if not
+ * already in at least the given level.
*/
- private boolean transitionRunStateTo(int state) {
+ private boolean advanceRunLevel(int level) {
for (;;) {
- int c = runControl;
- if (runStateOf(c) >= state)
+ int s = runState;
+ if ((s & level) != 0)
return false;
- if (casRunControl(c, runControlFor(state, activeCountOf(c))))
+ if (UNSAFE.compareAndSwapInt(this, runStateOffset, s, s | level))
return true;
}
}
+ // workers array maintenance
+
/**
- * Controls whether to add spares to maintain parallelism
+ * Records and returns a workers array index for new worker.
*/
- private volatile boolean maintainsParallelism;
+ private int recordWorker(ForkJoinWorkerThread w) {
+ // Try using slot totalCount-1. If not available, scan and/or resize
+ int k = (workerCounts >>> TOTAL_COUNT_SHIFT) - 1;
+ final ReentrantLock lock = this.workerLock;
+ lock.lock();
+ try {
+ ForkJoinWorkerThread[] ws = workers;
+ int n = ws.length;
+ if (k < 0 || k >= n || ws[k] != null) {
+ for (k = 0; k < n && ws[k] != null; ++k)
+ ;
+ if (k == n)
+ ws = Arrays.copyOf(ws, n << 1);
+ }
+ ws[k] = w;
+ workers = ws; // volatile array write ensures slot visibility
+ } finally {
+ lock.unlock();
+ }
+ return k;
+ }
- // Constructors
+ /**
+ * Nulls out record of worker in workers array
+ */
+ private void forgetWorker(ForkJoinWorkerThread w) {
+ int idx = w.poolIndex;
+ // Locking helps method recordWorker avoid unecessary expansion
+ final ReentrantLock lock = this.workerLock;
+ lock.lock();
+ try {
+ ForkJoinWorkerThread[] ws = workers;
+ if (idx >= 0 && idx < ws.length && ws[idx] == w) // verify
+ ws[idx] = null;
+ } finally {
+ lock.unlock();
+ }
+ }
+
+ // adding and removing workers
/**
- * Creates a ForkJoinPool with a pool size equal to the number of
- * processors available on the system and using the default
- * ForkJoinWorkerThreadFactory,
- * @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")},
+ * Tries to create and add new worker. Assumes that worker counts
+ * are already updated to accommodate the worker, so adjusts on
+ * failure.
+ *
+ * @return the worker, or null on failure
*/
- public ForkJoinPool() {
- this(Runtime.getRuntime().availableProcessors(),
- defaultForkJoinWorkerThreadFactory);
+ private ForkJoinWorkerThread addWorker() {
+ ForkJoinWorkerThread w = null;
+ try {
+ w = factory.newThread(this);
+ } finally { // Adjust on either null or exceptional factory return
+ if (w == null) {
+ decrementWorkerCounts(ONE_RUNNING, ONE_TOTAL);
+ tryTerminate(false); // in case of failure during shutdown
+ }
+ }
+ if (w != null) {
+ w.start(recordWorker(w), ueh);
+ advanceEventCount();
+ }
+ return w;
}
/**
- * Creates a ForkJoinPool with the indicated parallelism 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}{@code ("modifyThread")},
+ * Final callback from terminating worker. Removes record of
+ * worker from array, and adjusts counts. If pool is shutting
+ * down, tries to complete terminatation.
+ *
+ * @param w the worker
*/
- public ForkJoinPool(int parallelism) {
- this(parallelism, 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 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}{@code ("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
+ * contention, letting other workers take over.
*/
- public ForkJoinPool(ForkJoinWorkerThreadFactory factory) {
- this(Runtime.getRuntime().availableProcessors(), factory);
+ private void releaseEventWaiters() {
+ ForkJoinWorkerThread[] ws = workers;
+ int n = ws.length;
+ long h = eventWaiters;
+ int ec = eventCount;
+ ForkJoinWorkerThread w; int id;
+ while ((int)(h >>> EVENT_COUNT_SHIFT) != ec &&
+ (id = ((int)(h & WAITER_ID_MASK)) - 1) >= 0 &&
+ id < n && (w = ws[id]) != null &&
+ UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
+ h, h = w.nextWaiter)) {
+ LockSupport.unpark(w);
+ if (eventWaiters != h || eventCount != ec)
+ break;
+ }
}
/**
- * Creates a ForkJoinPool with the given parallelism and factory.
+ * Tries to advance eventCount and releases waiters. Called only
+ * from workers.
+ */
+ 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();
+ }
+
+ /**
+ * Adds the given worker to event queue and blocks until
+ * terminating or event count advances from the workers
+ * lastEventCount 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}{@code ("modifyThread")},
+ * @param w the calling worker thread
*/
- 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 LinkedTransferQueue>();
- // worker array and workers are lazily constructed
+ private void eventSync(ForkJoinWorkerThread w) {
+ int ec = w.lastEventCount;
+ long nh = (((long)ec) << EVENT_COUNT_SHIFT) | ((long)(w.poolIndex+1));
+ long h;
+ while ((runState < SHUTDOWN || !tryTerminate(false)) &&
+ (((int)((h = eventWaiters) & WAITER_ID_MASK)) == 0 ||
+ (int)(h >>> EVENT_COUNT_SHIFT) == ec) &&
+ eventCount == ec) {
+ if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
+ w.nextWaiter = h, nh)) {
+ awaitEvent(w, ec);
+ break;
+ }
+ }
}
/**
- * Create new worker using factory.
- * @param index the index to assign worker
- * @return new worker, or null of factory failed
+ * Blocks the given worker (that has already been entered as an
+ * event waiter) until terminating or event count advances from
+ * the given value. The oldest (first) waiter uses a timed wait to
+ * occasionally one-by-one shrink the number of workers (to a
+ * minumum of one) if the pool has not been used for extended
+ * periods.
+ *
+ * @param w the calling worker thread
+ * @param ec the count
*/
- private ForkJoinWorkerThread createWorker(int index) {
- Thread.UncaughtExceptionHandler h = ueh;
- ForkJoinWorkerThread w = factory.newThread(this);
- if (w != null) {
- w.poolIndex = index;
- w.setDaemon(true);
- w.setAsyncMode(locallyFifo);
- w.setName("ForkJoinPool-" + poolNumber + "-worker-" + index);
- if (h != null)
- w.setUncaughtExceptionHandler(h);
+ private void awaitEvent(ForkJoinWorkerThread w, int ec) {
+ while (eventCount == ec) {
+ if (tryAccumulateStealCount(w)) { // transfer while idle
+ boolean untimed = (w.nextWaiter != 0L ||
+ (workerCounts & RUNNING_COUNT_MASK) <= 1);
+ long startTime = untimed? 0 : System.nanoTime();
+ Thread.interrupted(); // clear/ignore interrupt
+ if (eventCount != ec || !w.isRunning() ||
+ runState >= TERMINATING) // recheck after clear
+ break;
+ if (untimed)
+ LockSupport.park(w);
+ else {
+ LockSupport.parkNanos(w, SHRINK_RATE_NANOS);
+ if (eventCount != ec || !w.isRunning() ||
+ runState >= TERMINATING)
+ break;
+ if (System.nanoTime() - startTime >= SHRINK_RATE_NANOS)
+ tryShutdownWaiter(ec);
+ }
+ }
}
- return w;
}
/**
- * Return a good size for worker array given pool size.
- * Currently requires size to be a power of two.
+ * Callback from the oldest waiter in awaitEvent waking up after a
+ * period of non-use. Tries (once) to shutdown an event waiter (or
+ * a spare, if one exists). Note that we don't need CAS or locks
+ * here because the method is called only from one thread
+ * occasionally waking (and even misfires are OK). Note that
+ * until the shutdown worker fully terminates, workerCounts
+ * will overestimate total count, which is tolerable.
+ *
+ * @param ec the event count waited on by caller (to abort
+ * attempt if count has since changed).
+ */
+ private void tryShutdownWaiter(int ec) {
+ if (spareWaiters != 0) { // prefer killing spares
+ tryShutdownSpare();
+ return;
+ }
+ ForkJoinWorkerThread[] ws = workers;
+ int n = ws.length;
+ long h = eventWaiters;
+ ForkJoinWorkerThread w; int id; long nh;
+ if (runState == 0 &&
+ submissionQueue.isEmpty() &&
+ eventCount == ec &&
+ (id = ((int)(h & WAITER_ID_MASK)) - 1) >= 0 &&
+ id < n && (w = ws[id]) != null &&
+ (nh = w.nextWaiter) != 0L && // keep at least one worker
+ UNSAFE.compareAndSwapLong(this, eventWaitersOffset, h, nh)) {
+ w.shutdown();
+ LockSupport.unpark(w);
+ }
+ releaseEventWaiters();
+ }
+
+ // Maintaining spares
+
+ /**
+ * Pushes worker onto the spare stack
*/
- private static int arraySizeFor(int ps) {
- return ps <= 1? 1 : (1 << (32 - Integer.numberOfLeadingZeros(ps-1)));
+ final void pushSpare(ForkJoinWorkerThread w) {
+ int ns = (++w.spareCount << SPARE_COUNT_SHIFT) | (w.poolIndex + 1);
+ do {} while (!UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
+ w.nextSpare = spareWaiters,ns));
}
/**
- * Create or resize array if necessary to hold newLength.
- * Call only under exclusion or lock
- * @return the array
+ * Callback from oldest spare occasionally waking up. Tries
+ * (once) to shutdown a spare. Same idea as tryShutdownWaiter.
*/
- private ForkJoinWorkerThread[] ensureWorkerArrayCapacity(int newLength) {
- ForkJoinWorkerThread[] ws = workers;
- if (ws == null)
- return workers = new ForkJoinWorkerThread[arraySizeFor(newLength)];
- else if (newLength > ws.length)
- return workers = Arrays.copyOf(ws, arraySizeFor(newLength));
- else
- return ws;
+ final void tryShutdownSpare() {
+ int sw, id;
+ ForkJoinWorkerThread w;
+ ForkJoinWorkerThread[] ws;
+ if ((id = ((sw = spareWaiters) & SPARE_ID_MASK) - 1) >= 0 &&
+ id < (ws = workers).length && (w = ws[id]) != null &&
+ (workerCounts & RUNNING_COUNT_MASK) >= parallelism &&
+ UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
+ sw, w.nextSpare)) {
+ w.shutdown();
+ LockSupport.unpark(w);
+ advanceEventCount();
+ }
}
/**
- * Try to shrink workers into smaller array after one or more terminate
+ * Tries (once) to resume a spare if worker counts match
+ * the given count.
+ *
+ * @param wc workerCounts value on invocation of this method
*/
- private void tryShrinkWorkerArray() {
+ private void tryResumeSpare(int wc) {
ForkJoinWorkerThread[] ws = workers;
- if (ws != null) {
- int len = ws.length;
- int last = len - 1;
- while (last >= 0 && ws[last] == null)
- --last;
- int newLength = arraySizeFor(last+1);
- if (newLength < len)
- workers = Arrays.copyOf(ws, newLength);
+ int n = ws.length;
+ int sw, id, rs; ForkJoinWorkerThread w;
+ if ((id = ((sw = spareWaiters) & SPARE_ID_MASK) - 1) >= 0 &&
+ id < n && (w = ws[id]) != null &&
+ (rs = runState) < TERMINATING &&
+ eventWaiters == 0L && workerCounts == wc) {
+ // In case all workers busy, heuristically back off to let settle
+ Thread.yield();
+ if (eventWaiters == 0L && runState == rs && // recheck
+ workerCounts == wc && spareWaiters == sw &&
+ UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
+ sw, w.nextSpare)) {
+ int c; // increment running count before resume
+ do {} while(!UNSAFE.compareAndSwapInt
+ (this, workerCountsOffset,
+ c = workerCounts, c + ONE_RUNNING));
+ if (w.tryUnsuspend())
+ LockSupport.unpark(w);
+ else // back out if w was shutdown
+ decrementWorkerCounts(ONE_RUNNING, 0);
+ }
+ }
+ }
+
+ // adding workers on demand
+
+ /**
+ * Adds one or more workers if needed to establish target parallelism.
+ * Retries upon contention.
+ */
+ private void addWorkerIfBelowTarget() {
+ int pc = parallelism;
+ int wc;
+ while (((wc = workerCounts) >>> TOTAL_COUNT_SHIFT) < pc &&
+ runState < TERMINATING) {
+ if (UNSAFE.compareAndSwapInt(this, workerCountsOffset, wc,
+ wc + (ONE_RUNNING|ONE_TOTAL))) {
+ if (addWorker() == null)
+ break;
+ }
}
}
/**
- * Initialize workers if necessary
+ * Tries (once) to add a new worker if all existing workers are
+ * busy, and there are either no running workers or the deficit is
+ * at least twice the surplus.
+ *
+ * @param wc workerCounts value on invocation of this method
*/
- final void ensureWorkerInitialization() {
- ForkJoinWorkerThread[] ws = workers;
- if (ws == null) {
- final ReentrantLock lock = this.workerLock;
- lock.lock();
- try {
- ws = workers;
- if (ws == null) {
- int ps = parallelism;
- ws = ensureWorkerArrayCapacity(ps);
- for (int i = 0; i < ps; ++i) {
- ForkJoinWorkerThread w = createWorker(i);
- if (w != null) {
- ws[i] = w;
- w.start();
- updateWorkerCount(1);
- }
+ private void tryAddWorkerIfBusy(int wc) {
+ int tc, rc, rs;
+ int pc = parallelism;
+ if ((tc = wc >>> TOTAL_COUNT_SHIFT) < MAX_WORKERS &&
+ ((rc = wc & RUNNING_COUNT_MASK) == 0 ||
+ rc < pc - ((tc - pc) << 1)) &&
+ (rs = runState) < TERMINATING &&
+ (rs & ACTIVE_COUNT_MASK) == tc) {
+ // Since all workers busy, heuristically back off to let settle
+ Thread.yield();
+ if (eventWaiters == 0L && spareWaiters == 0 && // recheck
+ runState == rs && workerCounts == wc &&
+ UNSAFE.compareAndSwapInt(this, workerCountsOffset, wc,
+ wc + (ONE_RUNNING|ONE_TOTAL)))
+ addWorker();
+ }
+ }
+
+ /**
+ * Does at most one of:
+ *
+ * 1. Help wake up existing workers waiting for work via
+ * releaseEventWaiters. (If any exist, then it doesn't
+ * matter right now if under target parallelism level.)
+ *
+ * 2. If a spare exists, try (once) to resume it via tryResumeSpare.
+ *
+ * 3. If there are not enough total workers, add some
+ * via addWorkerIfBelowTarget;
+ *
+ * 4. Try (once) to add a new worker if all existing workers
+ * are busy, via tryAddWorkerIfBusy
+ */
+ private void helpMaintainParallelism() {
+ long h; int pc, wc;
+ if (((int)((h = eventWaiters) & WAITER_ID_MASK)) != 0) {
+ if ((int)(h >>> EVENT_COUNT_SHIFT) != eventCount)
+ releaseEventWaiters(); // avoid useless call
+ }
+ else if ((pc = parallelism) >
+ ((wc = workerCounts) & RUNNING_COUNT_MASK)) {
+ if (spareWaiters != 0)
+ tryResumeSpare(wc);
+ else if ((wc >>> TOTAL_COUNT_SHIFT) < pc)
+ addWorkerIfBelowTarget();
+ else
+ tryAddWorkerIfBusy(wc);
+ }
+ }
+
+ /**
+ * Callback from workers invoked upon each top-level action (i.e.,
+ * stealing a task or taking a submission and running it).
+ * Performs one or more of the following:
+ *
+ * 1. If the worker is active, try to set its active status to
+ * inactive and update activeCount. On contention, we may try
+ * again on this or subsequent call.
+ *
+ * 2. Release any existing event waiters that are now relesable
+ *
+ * 3. If there are too many running threads, suspend this worker
+ * (first forcing inactive if necessary). If it is not
+ * needed, it may be killed while suspended via
+ * tryShutdownSpare. Otherwise, upon resume it rechecks to make
+ * sure that it is still needed.
+ *
+ * 4. If more than 1 miss, await the next task event via
+ * eventSync (first forcing inactivation if necessary), upon
+ * which worker may also be killed, via tryShutdownWaiter.
+ *
+ * 5. Help reactivate other workers via helpMaintainParallelism
+ *
+ * @param w the worker
+ * @param misses the number of scans by caller failing to find work
+ * (saturating at 2 to avoid wraparound)
+ */
+ final void preStep(ForkJoinWorkerThread w, int misses) {
+ boolean active = w.active;
+ int pc = parallelism;
+ for (;;) {
+ int rs, wc, rc, ec; long h;
+ if (active && UNSAFE.compareAndSwapInt(this, runStateOffset,
+ rs = runState, rs - 1))
+ active = w.active = false;
+ if (((int)((h = eventWaiters) & WAITER_ID_MASK)) != 0 &&
+ (int)(h >>> EVENT_COUNT_SHIFT) != eventCount) {
+ releaseEventWaiters();
+ if (misses > 1)
+ continue; // clear before sync below
+ }
+ if ((rc = ((wc = workerCounts) & RUNNING_COUNT_MASK)) > pc) {
+ if (!active && // must inactivate to suspend
+ workerCounts == wc && // try to suspend as spare
+ UNSAFE.compareAndSwapInt(this, workerCountsOffset,
+ wc, wc - ONE_RUNNING)) {
+ w.suspendAsSpare();
+ if (!w.isRunning())
+ break; // was killed while spare
+ }
+ continue;
+ }
+ if (misses > 0) {
+ if ((ec = eventCount) == w.lastEventCount && misses > 1) {
+ if (!active) { // must inactivate to sync
+ eventSync(w);
+ if (w.isRunning())
+ misses = 1; // don't re-sync
+ else
+ break; // was killed while waiting
}
+ continue;
}
- } finally {
- lock.unlock();
+ w.lastEventCount = ec;
}
+ if (rc < pc)
+ helpMaintainParallelism();
+ break;
}
}
/**
- * Worker creation and startup for threads added via setParallelism.
+ * Helps and/or blocks awaiting join of the given task.
+ * Alternates between helpJoinTask() and helpMaintainParallelism()
+ * as many times as there is a deficit in running count (or longer
+ * if running count would become zero), then blocks if task still
+ * not done.
+ *
+ * @param joinMe the task to join
*/
- private void createAndStartAddedWorkers() {
- resumeAllSpares(); // Allow spares to convert to nonspare
- int ps = parallelism;
- ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(ps);
- int len = ws.length;
- // Sweep through slots, to keep lowest indices most populated
- int k = 0;
- while (k < len) {
- if (ws[k] != null) {
- ++k;
- continue;
+ final void awaitJoin(ForkJoinTask joinMe, ForkJoinWorkerThread worker) {
+ int threshold = parallelism; // descend blocking thresholds
+ while (joinMe.status >= 0) {
+ boolean block; int wc;
+ worker.helpJoinTask(joinMe);
+ if (joinMe.status < 0)
+ break;
+ if (((wc = workerCounts) & RUNNING_COUNT_MASK) <= threshold) {
+ if (threshold > 0)
+ --threshold;
+ else
+ advanceEventCount(); // force release
+ block = false;
}
- int s = workerCounts;
- int tc = totalCountOf(s);
- int rc = runningCountOf(s);
- if (rc >= ps || tc >= ps)
+ else
+ block = UNSAFE.compareAndSwapInt(this, workerCountsOffset,
+ wc, wc - ONE_RUNNING);
+ helpMaintainParallelism();
+ if (block) {
+ int c;
+ joinMe.internalAwaitDone();
+ do {} while (!UNSAFE.compareAndSwapInt
+ (this, workerCountsOffset,
+ c = workerCounts, c + ONE_RUNNING));
break;
- if (casWorkerCounts (s, workerCountsFor(tc+1, rc+1))) {
- ForkJoinWorkerThread w = createWorker(k);
- if (w != null) {
- ws[k++] = w;
- w.start();
+ }
+ }
+ }
+
+ /**
+ * Same idea as awaitJoin, but no helping
+ */
+ final void awaitBlocker(ManagedBlocker blocker)
+ throws InterruptedException {
+ int threshold = parallelism;
+ while (!blocker.isReleasable()) {
+ boolean block; int wc;
+ if (((wc = workerCounts) & RUNNING_COUNT_MASK) <= threshold) {
+ if (threshold > 0)
+ --threshold;
+ else
+ advanceEventCount();
+ block = false;
+ }
+ else
+ block = UNSAFE.compareAndSwapInt(this, workerCountsOffset,
+ wc, wc - ONE_RUNNING);
+ helpMaintainParallelism();
+ if (block) {
+ try {
+ do {} while (!blocker.isReleasable() && !blocker.block());
+ } finally {
+ int c;
+ do {} while (!UNSAFE.compareAndSwapInt
+ (this, workerCountsOffset,
+ c = workerCounts, c + ONE_RUNNING));
}
- else {
- updateWorkerCount(-1); // back out on failed creation
- break;
+ break;
+ }
+ }
+ }
+
+ /**
+ * Possibly initiates and/or completes termination.
+ *
+ * @param now if true, unconditionally terminate, else only
+ * if shutdown and empty queue and no active workers
+ * @return true if now terminating or terminated
+ */
+ private boolean tryTerminate(boolean now) {
+ if (now)
+ advanceRunLevel(SHUTDOWN); // ensure at least SHUTDOWN
+ else if (runState < SHUTDOWN ||
+ !submissionQueue.isEmpty() ||
+ (runState & ACTIVE_COUNT_MASK) != 0)
+ return false;
+
+ if (advanceRunLevel(TERMINATING))
+ startTerminating();
+
+ // Finish now if all threads terminated; else in some subsequent call
+ if ((workerCounts >>> TOTAL_COUNT_SHIFT) == 0) {
+ advanceRunLevel(TERMINATED);
+ termination.arrive();
+ }
+ return true;
+ }
+
+ /**
+ * Actions on transition to TERMINATING
+ *
+ * Runs up to four passes through workers: (0) shutting down each
+ * (without waking up if parked) to quickly spread notifications
+ * without unnecessary bouncing around event queues etc (1) wake
+ * up and help cancel tasks (2) interrupt (3) mop up races with
+ * interrupted workers
+ */
+ private void startTerminating() {
+ cancelSubmissions();
+ for (int passes = 0; passes < 4 && workerCounts != 0; ++passes) {
+ advanceEventCount();
+ eventWaiters = 0L; // clobber lists
+ spareWaiters = 0;
+ ForkJoinWorkerThread[] ws = workers;
+ int n = ws.length;
+ for (int i = 0; i < n; ++i) {
+ ForkJoinWorkerThread w = ws[i];
+ if (w != null) {
+ w.shutdown();
+ if (passes > 0 && !w.isTerminated()) {
+ w.cancelTasks();
+ LockSupport.unpark(w);
+ if (passes > 1) {
+ try {
+ w.interrupt();
+ } catch (SecurityException ignore) {
+ }
+ }
+ }
}
}
}
}
+ /**
+ * Clear out and cancel submissions, ignoring exceptions
+ */
+ private void cancelSubmissions() {
+ ForkJoinTask task;
+ while ((task = submissionQueue.poll()) != null) {
+ try {
+ task.cancel(false);
+ } catch (Throwable ignore) {
+ }
+ }
+ }
+
+ // misc support for ForkJoinWorkerThread
+
+ /**
+ * Returns pool number
+ */
+ final int getPoolNumber() {
+ return poolNumber;
+ }
+
+ /**
+ * Tries to accumulates steal count from a worker, clearing
+ * the worker's value.
+ *
+ * @return true if worker steal count now zero
+ */
+ final boolean tryAccumulateStealCount(ForkJoinWorkerThread w) {
+ int sc = w.stealCount;
+ long c = stealCount;
+ // CAS even if zero, for fence effects
+ if (UNSAFE.compareAndSwapLong(this, stealCountOffset, c, c + sc)) {
+ if (sc != 0)
+ w.stealCount = 0;
+ return true;
+ }
+ return sc == 0;
+ }
+
+ /**
+ * Returns the approximate (non-atomic) number of idle threads per
+ * active thread.
+ */
+ final int idlePerActive() {
+ int pc = parallelism; // use parallelism, not rc
+ int ac = runState; // no mask -- artifically boosts during shutdown
+ // Use exact results for small values, saturate past 4
+ return pc <= ac? 0 : pc >>> 1 <= ac? 1 : pc >>> 2 <= ac? 3 : pc >>> 3;
+ }
+
+ // Public and protected methods
+
+ // Constructors
+
+ /**
+ * Creates a {@code ForkJoinPool} with parallelism equal to {@link
+ * java.lang.Runtime#availableProcessors}, using the {@linkplain
+ * #defaultForkJoinWorkerThreadFactory default thread factory},
+ * no UncaughtExceptionHandler, and non-async LIFO processing mode.
+ *
+ * @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() {
+ this(Runtime.getRuntime().availableProcessors(),
+ defaultForkJoinWorkerThreadFactory, null, false);
+ }
+
+ /**
+ * Creates a {@code ForkJoinPool} with the indicated parallelism
+ * level, the {@linkplain
+ * #defaultForkJoinWorkerThreadFactory default thread factory},
+ * no UncaughtExceptionHandler, and non-async LIFO processing mode.
+ *
+ * @param parallelism the parallelism level
+ * @throws IllegalArgumentException if parallelism less than or
+ * equal to zero, or greater than implementation limit
+ * @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) {
+ this(parallelism, defaultForkJoinWorkerThreadFactory, null, false);
+ }
+
+ /**
+ * Creates a {@code ForkJoinPool} with the given parameters.
+ *
+ * @param parallelism the parallelism level. For default value,
+ * use {@link java.lang.Runtime#availableProcessors}.
+ * @param factory the factory for creating new threads. For default value,
+ * use {@link #defaultForkJoinWorkerThreadFactory}.
+ * @param handler the handler for internal worker threads that
+ * terminate due to unrecoverable errors encountered while executing
+ * tasks. For default value, use null.
+ * @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>();
+ this.workerLock = new ReentrantLock();
+ this.termination = new Phaser(1);
+ this.poolNumber = poolNumberGenerator.incrementAndGet();
+ }
+
+ /**
+ * Returns initial power of two size for workers array.
+ * @param pc the initial parallelism level
+ */
+ private static int initialArraySizeFor(int pc) {
+ // See Hackers Delight, sec 3.2. We know MAX_WORKERS < (1 >>> 16)
+ int size = pc < MAX_WORKERS ? pc + 1 : MAX_WORKERS;
+ size |= size >>> 1;
+ size |= size >>> 2;
+ size |= size >>> 4;
+ size |= size >>> 8;
+ return size + 1;
+ }
+
// Execution methods
/**
* Common code for execute, invoke and submit
*/
private void doSubmit(ForkJoinTask task) {
- if (isShutdown())
+ if (task == null)
+ throw new NullPointerException();
+ if (runState >= SHUTDOWN)
throw new RejectedExecutionException();
- if (workers == null)
- ensureWorkerInitialization();
submissionQueue.offer(task);
- signalIdleWorkers();
+ advanceEventCount();
+ if (eventWaiters != 0L)
+ releaseEventWaiters();
+ if ((workerCounts >>> TOTAL_COUNT_SHIFT) < parallelism)
+ addWorkerIfBelowTarget();
}
/**
- * Performs the given task; returning its result upon completion
+ * Performs the given task, returning its result upon completion.
+ *
* @param task the task
* @return the task's result
- * @throws NullPointerException if task is null
- * @throws RejectedExecutionException if pool is shut down
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
*/
public T invoke(ForkJoinTask task) {
doSubmit(task);
@@ -549,111 +1396,111 @@ public class ForkJoinPool extends Abstra
/**
* Arranges for (asynchronous) execution of the given task.
+ *
* @param task the task
- * @throws NullPointerException if task is null
- * @throws RejectedExecutionException if pool is shut down
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
*/
- public void execute(ForkJoinTask task) {
+ public void execute(ForkJoinTask task) {
doSubmit(task);
}
// AbstractExecutorService methods
+ /**
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
+ */
public void execute(Runnable task) {
- doSubmit(new AdaptedRunnable(task, null));
+ ForkJoinTask job;
+ if (task instanceof ForkJoinTask) // avoid re-wrap
+ job = (ForkJoinTask) task;
+ else
+ job = ForkJoinTask.adapt(task, null);
+ doSubmit(job);
+ }
+
+ /**
+ * Submits a ForkJoinTask for execution.
+ *
+ * @param task the task to submit
+ * @return the task
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
+ */
+ public ForkJoinTask submit(ForkJoinTask task) {
+ doSubmit(task);
+ return task;
}
+ /**
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
+ */
public ForkJoinTask submit(Callable task) {
- ForkJoinTask job = new AdaptedCallable(task);
+ ForkJoinTask job = ForkJoinTask.adapt(task);
doSubmit(job);
return job;
}
+ /**
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
+ */
public ForkJoinTask submit(Runnable task, T result) {
- ForkJoinTask job = new AdaptedRunnable(task, result);
+ ForkJoinTask job = ForkJoinTask.adapt(task, result);
doSubmit(job);
return job;
}
+ /**
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
+ */
public ForkJoinTask submit(Runnable task) {
- ForkJoinTask job = new AdaptedRunnable(task, null);
+ ForkJoinTask job;
+ if (task instanceof ForkJoinTask) // avoid re-wrap
+ job = (ForkJoinTask) task;
+ else
+ job = ForkJoinTask.adapt(task, null);
doSubmit(job);
return job;
}
/**
- * Adaptor for Runnables. This implements RunnableFuture
- * to be compliant with AbstractExecutorService constraints
- */
- static final class AdaptedRunnable extends ForkJoinTask
- implements RunnableFuture {
- final Runnable runnable;
- final T resultOnCompletion;
- T result;
- AdaptedRunnable(Runnable runnable, T result) {
- if (runnable == null) throw new NullPointerException();
- this.runnable = runnable;
- this.resultOnCompletion = result;
- }
- public T getRawResult() { return result; }
- public void setRawResult(T v) { result = v; }
- public boolean exec() {
- runnable.run();
- result = resultOnCompletion;
- return true;
- }
- public void run() { invoke(); }
- }
-
- /**
- * Adaptor for Callables
+ * @throws NullPointerException {@inheritDoc}
+ * @throws RejectedExecutionException {@inheritDoc}
*/
- static final class AdaptedCallable extends ForkJoinTask
- implements RunnableFuture {
- final Callable callable;
- T result;
- AdaptedCallable(Callable callable) {
- if (callable == null) throw new NullPointerException();
- this.callable = callable;
- }
- public T getRawResult() { return result; }
- public void setRawResult(T v) { result = v; }
- public boolean exec() {
- try {
- result = callable.call();
- return true;
- } catch (Error err) {
- throw err;
- } catch (RuntimeException rex) {
- throw rex;
- } catch (Exception ex) {
- throw new RuntimeException(ex);
- }
- }
- public void run() { invoke(); }
- }
-
public List> invokeAll(Collection> tasks) {
- ArrayList> ts =
+ ArrayList> forkJoinTasks =
new ArrayList>(tasks.size());
- for (Callable c : tasks)
- ts.add(new AdaptedCallable(c));
- invoke(new InvokeAll(ts));
- return (List>)(List)ts;
+ for (Callable task : tasks)
+ forkJoinTasks.add(ForkJoinTask.adapt(task));
+ invoke(new InvokeAll(forkJoinTasks));
+
+ @SuppressWarnings({"unchecked", "rawtypes"})
+ List> futures = (List>) (List) forkJoinTasks;
+ return futures;
}
static final class InvokeAll extends RecursiveAction {
final ArrayList> tasks;
InvokeAll(ArrayList> tasks) { this.tasks = tasks; }
public void compute() {
- try { invokeAll(tasks); } catch(Exception ignore) {}
+ try { invokeAll(tasks); }
+ catch (Exception ignore) {}
}
+ private static final long serialVersionUID = -7914297376763021607L;
}
- // Configuration and status settings and queries
-
/**
- * Returns the factory used for constructing new workers
+ * Returns the factory used for constructing new workers.
*
* @return the factory used for constructing new workers
*/
@@ -664,92 +1511,17 @@ public class ForkJoinPool extends Abstra
/**
* Returns the handler for internal worker threads that terminate
* due to unrecoverable errors encountered while executing tasks.
- * @return the handler, or null if none
- */
- public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() {
- Thread.UncaughtExceptionHandler h;
- final ReentrantLock lock = this.workerLock;
- lock.lock();
- try {
- h = ueh;
- } finally {
- lock.unlock();
- }
- return h;
- }
-
- /**
- * Sets the handler for internal worker threads that terminate due
- * to unrecoverable errors encountered while executing tasks.
- * Unless set, the current default or ThreadGroup handler is used
- * as handler.
*
- * @param h the new handler
- * @return the old handler, or null if none
- * @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")},
+ * @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;
- if (ws != null) {
- 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 parallelism 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}{@code ("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();
+ 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;
@@ -758,91 +1530,20 @@ 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 {@code 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);
+ return workerCounts >>> TOTAL_COUNT_SHIFT;
}
/**
- * 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;
- }
-
- /**
- * 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
- */
- public void setMaintainsParallelism(boolean enable) {
- maintainsParallelism = enable;
- }
-
- /**
- * 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 asynchronous tasks. This method is
- * designed to be invoked only when pool is quiescent, and
- * typically only before any tasks are submitted. The effects of
- * invocations at other times may be unpredictable.
- *
- * @param async if true, use locally FIFO scheduling
- * @return the previous mode.
- */
- public boolean setAsyncMode(boolean async) {
- boolean oldMode = locallyFifo;
- locallyFifo = async;
- ForkJoinWorkerThread[] ws = workers;
- if (ws != null) {
- for (int i = 0; i < ws.length; ++i) {
- ForkJoinWorkerThread t = ws[i];
- if (t != null)
- t.setAsyncMode(async);
- }
- }
- return oldMode;
- }
-
- /**
- * Returns true if this pool uses local first-in-first-out
+ * Returns {@code true} if this pool uses local first-in-first-out
* scheduling mode for forked tasks that are never joined.
*
- * @return true if this pool uses async mode.
+ * @return {@code true} if this pool uses async mode
*/
public boolean getAsyncMode() {
return locallyFifo;
@@ -851,47 +1552,39 @@ public class ForkJoinPool extends Abstra
/**
* 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);
+ return runState & ACTIVE_COUNT_MASK;
}
/**
- * 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;
- }
-
- /**
- * 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;
}
/**
@@ -899,23 +1592,14 @@ 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;
}
/**
@@ -925,35 +1609,37 @@ 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;
- if (ws != null) {
- 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 {@code 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();
@@ -963,7 +1649,8 @@ 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();
@@ -973,8 +1660,8 @@ public class ForkJoinPool extends Abstra
* 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
+ * 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
@@ -982,20 +1669,20 @@ public class ForkJoinPool extends Abstra
* 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> c) {
- int n = submissionQueue.drainTo(c);
+ protected int drainTasksTo(Collection> c) {
+ int count = submissionQueue.drainTo(c);
ForkJoinWorkerThread[] ws = workers;
- if (ws != null) {
- for (int i = 0; i < ws.length; ++i) {
- ForkJoinWorkerThread w = ws[i];
- if (w != null)
- n += w.drainTasksTo(c);
- }
+ int n = ws.length;
+ for (int i = 0; i < n; ++i) {
+ ForkJoinWorkerThread w = ws[i];
+ if (w != null)
+ count += w.drainTasksTo(c);
}
- return n;
+ return count;
}
/**
@@ -1006,72 +1693,71 @@ 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}{@code ("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
- * upon termination, so always returns an empty list. However, you
- * can use method {@code drainTasksTo} before invoking this
- * method to transfer unexecuted tasks to another collection.
+ * 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}{@code ("modifyThread")},
+ * java.lang.RuntimePermission}{@code ("modifyThread")}
*/
public List shutdownNow() {
checkPermission();
- terminate();
+ tryTerminate(true);
return Collections.emptyList();
}
@@ -1081,17 +1767,21 @@ public class ForkJoinPool extends Abstra
* @return {@code true} if all tasks have completed following shut down
*/
public boolean isTerminated() {
- return runStateOf(runControl) == TERMINATED;
+ return runState >= TERMINATED;
}
/**
* Returns {@code true} if the process of termination has
- * commenced but possibly not yet completed.
+ * 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 {@code true} if terminating
+ * @return {@code true} if terminating but not yet terminated
*/
public boolean isTerminating() {
- return runStateOf(runControl) >= TERMINATING;
+ return (runState & (TERMINATING|TERMINATED)) == TERMINATING;
}
/**
@@ -1100,7 +1790,7 @@ public class ForkJoinPool extends Abstra
* @return {@code true} if this pool has been shut down
*/
public boolean isShutdown() {
- return runStateOf(runControl) >= SHUTDOWN;
+ return runState >= SHUTDOWN;
}
/**
@@ -1116,745 +1806,180 @@ public class ForkJoinPool extends Abstra
*/
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;
- if (ws != null) {
- 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();
- }
-
- /**
- * Initiate termination.
- */
- private void terminate() {
- if (transitionRunStateTo(TERMINATING)) {
- stopAllWorkers();
- resumeAllSpares();
- signalIdleWorkers();
- cancelQueuedSubmissions();
- cancelQueuedWorkerTasks();
- interruptUnterminatedWorkers();
- signalIdleWorkers(); // 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;
- if (ws != null) {
- 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;
- if (ws != null) {
- 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;
- if (ws != null) {
- 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. Queue nodes
- * are basic Treiber stack nodes, also used for spare stack.
- *
- * 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 work, they
- * may wait for next count. Upon release, threads help others wake
- * up.
- *
- * Synchronization events occur only in enough contexts to
- * maintain overall liveness:
- *
- * - Submission of a new task to the pool
- * - Resizes or other changes to the workers array
- * - pool termination
- * - A worker pushing a task on an empty queue
- *
- * The case of pushing a task occurs often enough, and is heavy
- * enough compared to simple stack pushes, to require special
- * handling: Method signalWork 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, the first worker enqueued in method sync (see
- * syncIsReleasable) rescans for tasks after being enqueued, and
- * helps signal if any are found. This works well because the
- * worker has nothing better to do, and so might as well help
- * alleviate the overhead and contention on the threads actually
- * doing work. Also, since event counts increments on task
- * availability exist to maintain liveness (rather than to force
- * refreshes etc), it is OK for callers to exit early if
- * contending with another signaller.
- */
- 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(long c, ForkJoinWorkerThread w) {
- count = c;
- thread = w;
- }
-
- /**
- * Wake up waiter, returning false if known to already
- */
- boolean signal() {
- ForkJoinWorkerThread t = thread;
- if (t == null)
- return false;
- thread = null;
- LockSupport.unpark(t);
- return true;
- }
-
- /**
- * Await release on sync
- */
- void awaitSyncRelease(ForkJoinPool p) {
- while (thread != null && !p.syncIsReleasable(this))
- LockSupport.park(this);
- }
-
- /**
- * Await resumption as spare
- */
- void awaitSpareRelease() {
- while (thread != null) {
- if (!Thread.interrupted())
- LockSupport.park(this);
- }
- }
- }
-
- /**
- * Ensures that no thread is waiting for count to advance from the
- * current value of eventCount read on entry to this method, by
- * releasing waiting threads if necessary.
- * @return the count
- */
- final long ensureSync() {
- long c = eventCount;
- WaitQueueNode q;
- while ((q = syncStack) != null && q.count < c) {
- if (casBarrierStack(q, null)) {
- do {
- q.signal();
- } while ((q = q.next) != null);
- break;
- }
- }
- return c;
- }
-
- /**
- * Increments event count and releases waiting threads.
- */
- private void signalIdleWorkers() {
- long c;
- do;while (!casEventCount(c = eventCount, c+1));
- ensureSync();
- }
-
- /**
- * Signal threads waiting to poll a task. Because method sync
- * rechecks availability, it is OK to only proceed if queue
- * appears to be non-empty, and OK to skip under contention to
- * increment count (since some other thread succeeded).
- */
- final void signalWork() {
- long c;
- WaitQueueNode q;
- if (syncStack != null &&
- casEventCount(c = eventCount, c+1) &&
- (((q = syncStack) != null && q.count <= c) &&
- (!casBarrierStack(q, q.next) || !q.signal())))
- ensureSync();
- }
-
- /**
- * Waits until event count advances from last value held by
- * caller, or if excess threads, caller is resumed as spare, or
- * caller or pool is terminating. Updates caller's event on exit.
- * @param w the calling worker thread
- */
- final void sync(ForkJoinWorkerThread w) {
- updateStealCount(w); // Transfer w's count while it is idle
-
- while (!w.isShutdown() && !isTerminating() && !suspendIfSpare(w)) {
- long prev = w.lastEventCount;
- WaitQueueNode node = null;
- WaitQueueNode h;
- while (eventCount == prev &&
- ((h = syncStack) == null || h.count == prev)) {
- if (node == null)
- node = new WaitQueueNode(prev, w);
- if (casBarrierStack(node.next = h, node)) {
- node.awaitSyncRelease(this);
- break;
- }
- }
- long ec = ensureSync();
- if (ec != prev) {
- w.lastEventCount = ec;
- break;
- }
- }
- }
-
- /**
- * Returns true if worker waiting on sync can proceed:
- * - on signal (thread == null)
- * - on event count advance (winning race to notify vs signaller)
- * - on Interrupt
- * - if the first queued node, we find work available
- * If node was not signalled and event count not advanced on exit,
- * then we also help advance event count.
- * @return true if node can be released
- */
- final boolean syncIsReleasable(WaitQueueNode node) {
- long prev = node.count;
- if (!Thread.interrupted() && node.thread != null &&
- (node.next != null ||
- !ForkJoinWorkerThread.hasQueuedTasks(workers)) &&
- eventCount == prev)
- return false;
- if (node.thread != null) {
- node.thread = null;
- long ec = eventCount;
- if (prev <= ec) // help signal
- casEventCount(ec, ec+1);
- }
- return true;
- }
-
- /**
- * Returns true if a new sync event occurred since last call to
- * sync or this method, if so, updating caller's count.
- */
- final boolean hasNewSyncEvent(ForkJoinWorkerThread w) {
- long lc = w.lastEventCount;
- long ec = ensureSync();
- if (ec == lc)
+ return termination.awaitAdvanceInterruptibly(0, timeout, unit) > 0;
+ } catch(TimeoutException ex) {
return false;
- w.lastEventCount = ec;
- return true;
- }
-
- // 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 encounter 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 &&
- ForkJoinWorkerThread.hasQueuedTasks(workers))));
- }
-
- /**
- * 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 counts 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();
- }
-
- /**
- * 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(0, w);
- if (casWorkerCounts(s, s-1)) { // representation-dependent
- // push onto stack
- do;while (!casSpareStack(node.next = spareStack, node));
- // block until released by resumeSpare
- node.awaitSpareRelease();
- 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 ensureSync.
- * @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);
- }
}
}
/**
* Interface for extending managed parallelism for tasks running
- * in ForkJoinPools. A ManagedBlocker provides two methods.
- * Method {@code isReleasable} must return true if blocking is not
- * necessary. Method {@code block} blocks the current thread
- * if necessary (perhaps internally invoking isReleasable before
- * actually blocking.).
+ * in {@link ForkJoinPool}s.
+ *
+ * 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:
- *
- * 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 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());
* }
- *
+ * }}
+ *
+ * Here is a class that possibly blocks waiting for an
+ * item on a given queue:
+ *
{@code
+ * class QueueTaker implements ManagedBlocker {
+ * final BlockingQueue queue;
+ * volatile E item = null;
+ * QueueTaker(BlockingQueue q) { this.queue = q; }
+ * public boolean block() throws InterruptedException {
+ * if (item == null)
+ * item = queue.take
+ * return true;
+ * }
+ * public boolean isReleasable() {
+ * return item != null || (item = queue.poll) != null;
+ * }
+ * public E getItem() { // call after pool.managedBlock completes
+ * return item;
+ * }
+ * }}
*/
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 allowed 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 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.
- *
- * If the caller is not a ForkJoinTask, this method is behaviorally
- * equivalent to
- *
- * 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.
+ *
+ * If the caller is not a {@link ForkJoinTask}, this method is
+ * behaviorally equivalent to
+ *
{@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 RunnableFuture newTaskFor(Runnable runnable, T value) {
- return new AdaptedRunnable(runnable, value);
+ return (RunnableFuture) ForkJoinTask.adapt(runnable, value);
}
protected RunnableFuture newTaskFor(Callable callable) {
- return new AdaptedCallable(callable);
+ return (RunnableFuture) ForkJoinTask.adapt(callable);
}
+ // Unsafe mechanics
+
+ private static final sun.misc.Unsafe UNSAFE = getUnsafe();
+ private static final long workerCountsOffset =
+ objectFieldOffset("workerCounts", ForkJoinPool.class);
+ private static final long runStateOffset =
+ objectFieldOffset("runState", ForkJoinPool.class);
+ private static final long eventCountOffset =
+ objectFieldOffset("eventCount", ForkJoinPool.class);
+ private static final long eventWaitersOffset =
+ objectFieldOffset("eventWaiters",ForkJoinPool.class);
+ private static final long stealCountOffset =
+ objectFieldOffset("stealCount",ForkJoinPool.class);
+ private static final long spareWaitersOffset =
+ objectFieldOffset("spareWaiters",ForkJoinPool.class);
- // Temporary Unsafe mechanics for preliminary release
- private static Unsafe getUnsafe() throws Throwable {
+ private static long objectFieldOffset(String field, Class klazz) {
try {
- return Unsafe.getUnsafe();
+ return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
+ } catch (NoSuchFieldException e) {
+ // Convert Exception to corresponding Error
+ NoSuchFieldError error = new NoSuchFieldError(field);
+ error.initCause(e);
+ throw error;
+ }
+ }
+
+ /**
+ * Returns a sun.misc.Unsafe. Suitable for use in a 3rd party package.
+ * Replace with a simple call to Unsafe.getUnsafe when integrating
+ * into a jdk.
+ *
+ * @return a sun.misc.Unsafe
+ */
+ private static sun.misc.Unsafe getUnsafe() {
+ try {
+ return sun.misc.Unsafe.getUnsafe();
} catch (SecurityException se) {
try {
return java.security.AccessController.doPrivileged
- (new java.security.PrivilegedExceptionAction() {
- public Unsafe run() throws Exception {
- return getUnsafePrivileged();
+ (new java.security
+ .PrivilegedExceptionAction() {
+ public sun.misc.Unsafe run() throws Exception {
+ java.lang.reflect.Field f = sun.misc
+ .Unsafe.class.getDeclaredField("theUnsafe");
+ f.setAccessible(true);
+ return (sun.misc.Unsafe) f.get(null);
}});
} catch (java.security.PrivilegedActionException e) {
- throw e.getCause();
+ throw new RuntimeException("Could not initialize intrinsics",
+ e.getCause());
}
}
}
-
- private static Unsafe getUnsafePrivileged()
- throws NoSuchFieldException, IllegalAccessException {
- Field f = Unsafe.class.getDeclaredField("theUnsafe");
- f.setAccessible(true);
- return (Unsafe) f.get(null);
- }
-
- private static long fieldOffset(String fieldName)
- throws NoSuchFieldException {
- return _unsafe.objectFieldOffset
- (ForkJoinPool.class.getDeclaredField(fieldName));
- }
-
- static final Unsafe _unsafe;
- static final long eventCountOffset;
- static final long workerCountsOffset;
- static final long runControlOffset;
- static final long syncStackOffset;
- static final long spareStackOffset;
-
- static {
- try {
- _unsafe = getUnsafe();
- eventCountOffset = fieldOffset("eventCount");
- workerCountsOffset = fieldOffset("workerCounts");
- runControlOffset = fieldOffset("runControl");
- syncStackOffset = fieldOffset("syncStack");
- spareStackOffset = fieldOffset("spareStack");
- } catch (Throwable e) {
- throw new RuntimeException("Could not initialize intrinsics", e);
- }
- }
-
- private boolean casEventCount(long cmp, long val) {
- return _unsafe.compareAndSwapLong(this, eventCountOffset, cmp, val);
- }
- private boolean casWorkerCounts(int cmp, int val) {
- return _unsafe.compareAndSwapInt(this, workerCountsOffset, cmp, val);
- }
- private boolean casRunControl(int cmp, int val) {
- return _unsafe.compareAndSwapInt(this, runControlOffset, cmp, val);
- }
- private boolean casSpareStack(WaitQueueNode cmp, WaitQueueNode val) {
- return _unsafe.compareAndSwapObject(this, spareStackOffset, cmp, val);
- }
- private boolean casBarrierStack(WaitQueueNode cmp, WaitQueueNode val) {
- return _unsafe.compareAndSwapObject(this, syncStackOffset, cmp, val);
- }
}