--- jsr166/src/jsr166y/ForkJoinPool.java 2010/08/13 16:21:23 1.63
+++ jsr166/src/jsr166y/ForkJoinPool.java 2011/03/04 13:29:39 1.95
@@ -6,17 +6,24 @@
package jsr166y;
-import java.util.concurrent.*;
-
import java.util.ArrayList;
import java.util.Arrays;
import java.util.Collection;
import java.util.Collections;
import java.util.List;
+import java.util.Random;
+import java.util.concurrent.AbstractExecutorService;
+import java.util.concurrent.Callable;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Future;
+import java.util.concurrent.RejectedExecutionException;
+import java.util.concurrent.RunnableFuture;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.TimeoutException;
+import java.util.concurrent.atomic.AtomicInteger;
import java.util.concurrent.locks.LockSupport;
import java.util.concurrent.locks.ReentrantLock;
-import java.util.concurrent.atomic.AtomicInteger;
-import java.util.concurrent.CountDownLatch;
+import java.util.concurrent.locks.Condition;
/**
* An {@link ExecutorService} for running {@link ForkJoinTask}s.
@@ -69,7 +76,7 @@ import java.util.concurrent.CountDownLat
*
Call from within fork/join computations |
*
*
- * Arange async execution |
+ * Arrange async execution |
* {@link #execute(ForkJoinTask)} |
* {@link ForkJoinTask#fork} |
*
@@ -124,240 +131,208 @@ public class ForkJoinPool extends Abstra
* 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
+ * Preference rules 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.
+ *
+ * The main throughput advantages of work-stealing stem from
+ * decentralized control -- workers mostly take tasks from
+ * themselves or each other. We cannot negate this in the
+ * implementation of other management responsibilities. The main
+ * tactic for avoiding bottlenecks is packing nearly all
+ * essentially atomic control state into a single 64bit volatile
+ * variable ("ctl"). This variable is read on the order of 10-100
+ * times as often as it is modified (always via CAS). (There is
+ * some additional control state, for example variable "shutdown"
+ * for which we can cope with uncoordinated updates.) This
+ * streamlines synchronization and control at the expense of messy
+ * constructions needed to repack status bits upon updates.
+ * Updates tend not to contend with each other except during
+ * bursts while submitted tasks begin or end. In some cases when
+ * they do contend, threads can instead do something else
+ * (usually, scan for tasks) until contention subsides.
+ *
+ * To enable packing, we restrict maximum parallelism to (1<<15)-1
+ * (which is far in excess of normal operating range) to allow
+ * ids, counts, and their negations (used for thresholding) to fit
+ * into 16bit fields.
+ *
+ * Recording Workers. Workers are recorded in the "workers" array
+ * that is created upon pool construction and expanded if (rarely)
+ * necessary. 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 seqLock
+ * (scanGuard) 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. To avoid flailing during
+ * start-up, the array is presized to hold twice #parallelism
+ * workers (which is unlikely to need further resizing during
+ * execution). But to avoid dealing with so many null slots,
+ * variable scanGuard includes a mask for the nearest power of two
+ * that contains all current workers. 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 messy code constructions here). In essence, the workers
+ * array serves as a weak reference mechanism. Thus for example
+ * the wait queue field of ctl stores worker indices, not worker
+ * references. Access to the workers in associated methods (for
+ * example signalWork) 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 termination, in which case it is OK to give up.
+ *
+ * All uses of the workers array, as well as queue arrays, check
+ * that the array is non-null (even if previously non-null). This
+ * allows nulling during termination, which is currently not
+ * necessary, but remains an option for resource-revocation-based
+ * shutdown schemes.
+ *
+ * Wait Queuing. Unlike HPC work-stealing frameworks, we cannot
+ * let workers spin indefinitely scanning for tasks when none can
+ * be found immediately, and we cannot start/resume workers unless
+ * there appear to be tasks available. On the other hand, we must
+ * quickly prod them into action when new tasks are submitted or
+ * generated. We park/unpark workers after placing in an event
+ * wait queue when they cannot find work. This "queue" is actually
+ * a simple Treiber stack, headed by the "id" field of ctl, plus a
+ * 15bit counter value to both wake up waiters (by advancing their
+ * count) and avoid ABA effects. Successors are held in worker
+ * field "nextWait". Queuing deals with several intrinsic races,
+ * mainly that a task-producing thread can miss seeing (and
+ * signalling) another thread that gave up looking for work but
+ * has not yet entered the wait queue. We solve this by requiring
+ * a full sweep of all workers both before (in scan()) and after
+ * (in tryAwaitWork()) a newly waiting worker is added to the wait
+ * queue. During a rescan, the worker might release some other
+ * queued worker rather than itself, which has the same net
+ * effect. Because enqueued workers may actually be rescanning
+ * rather than waiting, we set and clear the "parked" field of
+ * ForkJoinWorkerThread to reduce unnecessary calls to unpark.
+ * (Use of the parked field requires a secondary recheck to avoid
+ * missed signals.)
+ *
+ * Signalling. We create or wake up workers only when there
+ * appears to be at least one task they might be able to find and
+ * execute. When a submission is added or another worker adds a
+ * task to a queue that previously had two or fewer tasks, they
+ * signal waiting workers (or trigger creation of new ones if
+ * fewer than the given parallelism level -- see signalWork).
+ * These primary signals are buttressed by signals during rescans
+ * as well as those performed when a worker steals a task and
+ * notices that there are more tasks too; together these cover the
+ * signals needed in cases when more than two tasks are pushed
+ * but untaken.
+ *
+ * Trimming workers. To release resources after periods of lack of
+ * use, a worker starting to wait when the pool is quiescent will
+ * time out and terminate if the pool has remained quiescent for
+ * SHRINK_RATE nanosecs. This will slowly propagate, eventually
+ * terminating all workers after long periods of non-use.
+ *
+ * Submissions. External submissions are maintained in an
+ * array-based queue that is structured identically to
+ * ForkJoinWorkerThread queues except for the use of
+ * submissionLock in method addSubmission. Unlike the case for
+ * worker queues, multiple external threads can add new
+ * submissions, so adding requires a lock.
+ *
+ * Compensation. Beyond work-stealing support and lifecycle
+ * control, the main responsibility of this framework is to take
+ * actions when one worker is waiting to join a task stolen (or
+ * always held by) another. Because we are multiplexing many
+ * tasks on to a pool of workers, we can't just let them block (as
+ * in Thread.join). We also cannot just reassign the joiner's
+ * run-time stack with another and replace it later, which would
+ * be a form of "continuation", that even if possible is not
+ * necessarily a good idea since we sometimes need both an
+ * unblocked task and its continuation to progress. 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
+ * ForkJoinWorkerThread.joinTask 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).
+ * method tryPreBlock() may create or re-activate a spare
+ * thread to compensate for blocked joiners until they
+ * unblock.
*
* 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:
+ * It is impossible to keep exactly the target parallelism number
+ * of threads running at any given time. Determining the
+ * existence of conservatively safe helping targets, the
+ * availability of already-created spares, and the apparent need
+ * to create new spares are all racy and require heuristic
+ * guidance, so we rely on multiple retries of each. Currently,
+ * in keeping with on-demand signalling policy, we compensate only
+ * if blocking would leave less than one active (non-waiting,
+ * non-blocked) worker. Additionally, to avoid some false alarms
+ * due to GC, lagging counters, system activity, etc, compensated
+ * blocking for joins is only attempted after rechecks stabilize
+ * (retries are interspersed with Thread.yield, for good
+ * citizenship). The variable blockedCount, incremented before
+ * blocking and decremented after, is sometimes needed to
+ * distinguish cases of waiting for work vs blocking on joins or
+ * other managed sync. Both cases are equivalent for most pool
+ * control, so we can update non-atomically. (Additionally,
+ * contention on blockedCount alleviates some contention on ctl).
+ *
+ * Shutdown and Termination. A call to shutdownNow atomically sets
+ * the ctl stop bit and then (non-atomically) sets each workers
+ * "terminate" status, cancels all unprocessed tasks, and wakes up
+ * all waiting workers. Detecting whether termination should
+ * commence after a non-abrupt shutdown() call requires more work
+ * and bookkeeping. We need consensus about quiesence (i.e., that
+ * there is no more work) which is reflected in active counts so
+ * long as there are no current blockers, as well as possible
+ * re-evaluations during independent changes in blocking or
+ * quiescing workers.
*
- * 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 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. To reduce delays in task diffusion,
- * workers not otherwise occupied may invoke method
- * releaseEventWaiters, that removes and signals (unparks) workers
- * not waiting on current count. To reduce stalls, To minimize
- * task production stalls associate with signalling, any worker
- * pushing a task on an empty queue invokes the weaker method
- * signalWork, that only releases idle workers until it detects
- * interference by other threads trying to release, and lets them
- * take over. 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.
- * 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. To avoid
- * long-term build-up of spares, the oldest spare (see
- * ForkJoinWorkerThread.suspendAsSpare) occasionally wakes up if
- * not signalled and calls tryTrimSpare, which uses two different
- * thresholds: Always killing if the number of spares is greater
- * that 25% of total, and killing others only at a slower rate
- * (UNUSED_SPARE_TRIM_RATE_NANOS).
- *
- * 6. 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. 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
+ * Style notes: There is a lot of representation-level coupling
* among classes ForkJoinPool, ForkJoinWorkerThread, and
- * ForkJoinTask. For example, direct access to "workers" array by
+ * ForkJoinTask. Most fields of ForkJoinWorkerThread maintain
+ * data structures managed by ForkJoinPool, so are directly
+ * accessed. Conversely we allow 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).
+ * changes anyway. All together, these low-level implementation
+ * choices produce as much as a factor of 4 performance
+ * improvement compared to naive implementations, and enable the
+ * processing of billions of tasks per second, at the expense of
+ * some ugliness.
+ *
+ * Methods signalWork() and scan() are the main bottlenecks so are
+ * especially heavily micro-optimized/mangled. 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). This leads to a
+ * "C"-like style of listing declarations of these locals at the
+ * heads of methods or blocks. There are 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).
+ *
+ * The order of declarations in this file is: (1) declarations of
+ * statics (2) fields (along with constants used when unpacking
+ * some of them), listed in an order that tends to reduce
+ * contention among them a bit under most JVMs. (3) internal
+ * control methods (4) callbacks and other support for
+ * ForkJoinTask and ForkJoinWorkerThread classes, (5) exported
+ * methods (plus a few little helpers). (6) static block
+ * initializing all statics in a minimally dependent order.
*/
/**
@@ -392,15 +367,13 @@ public class ForkJoinPool extends Abstra
* overridden in ForkJoinPool constructors.
*/
public static final ForkJoinWorkerThreadFactory
- defaultForkJoinWorkerThreadFactory =
- new DefaultForkJoinWorkerThreadFactory();
+ defaultForkJoinWorkerThreadFactory;
/**
* Permission required for callers of methods that may start or
* kill threads.
*/
- private static final RuntimePermission modifyThreadPermission =
- new RuntimePermission("modifyThread");
+ private static final RuntimePermission modifyThreadPermission;
/**
* If there is a security manager, makes sure caller has
@@ -415,137 +388,163 @@ public class ForkJoinPool extends Abstra
/**
* Generator for assigning sequence numbers as pool names.
*/
- private static final AtomicInteger poolNumberGenerator =
- new AtomicInteger();
+ private static final AtomicInteger poolNumberGenerator;
/**
- * 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.
+ * Generator for initial random seeds for worker victim
+ * selection. This is used only to create initial seeds. Random
+ * steals use a cheaper xorshift generator per steal attempt. We
+ * don't expect much contention on seedGenerator, so just use a
+ * plain Random.
*/
- private static final int MAX_WORKERS = 0x7fff;
+ static final Random workerSeedGenerator;
/**
- * 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.
+ * Array holding all worker threads in the pool. Initialized upon
+ * construction. Array size must be a power of two. Updates and
+ * replacements are protected by scanGuard, 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, so long
+ * as reads memory-acquire by first reading ctl. All readers must
+ * tolerate that some array slots may be null.
*/
- volatile ForkJoinWorkerThread[] workers;
+ ForkJoinWorkerThread[] workers;
/**
- * Queue for external submissions.
+ * Initial size for submission queue array. Must be a power of
+ * two. In many applications, these always stay small so we use a
+ * small initial cap.
*/
- private final LinkedTransferQueue> submissionQueue;
+ private static final int INITIAL_QUEUE_CAPACITY = 8;
/**
- * Lock protecting updates to workers array.
+ * Maximum size for submission queue array. Must be a power of two
+ * less than or equal to 1 << (31 - width of array entry) to
+ * ensure lack of index wraparound, but is capped at a lower
+ * value to help users trap runaway computations.
*/
- private final ReentrantLock workerLock;
+ private static final int MAXIMUM_QUEUE_CAPACITY = 1 << 24; // 16M
/**
- * Latch released upon termination.
+ * Array serving as submission queue. Initialized upon construction.
*/
- private final Phaser termination;
+ private ForkJoinTask>[] submissionQueue;
/**
- * Creation factory for worker threads.
+ * Lock protecting submissions array for addSubmission
*/
- private final ForkJoinWorkerThreadFactory factory;
+ private final ReentrantLock submissionLock;
/**
- * Sum of per-thread steal counts, updated only when threads are
- * idle or terminating.
+ * Condition for awaitTermination, using submissionLock for
+ * convenience.
*/
- private volatile long stealCount;
+ private final Condition termination;
/**
- * The last nanoTime that a spare thread was trimmed
+ * Creation factory for worker threads.
*/
- private volatile long trimTime;
+ private final ForkJoinWorkerThreadFactory factory;
/**
- * The rate at which to trim unused spares
+ * The uncaught exception handler used when any worker abruptly
+ * terminates.
*/
- static final long UNUSED_SPARE_TRIM_RATE_NANOS =
- 1000L * 1000L * 1000L; // 1 sec
+ final Thread.UncaughtExceptionHandler ueh;
/**
- * 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.)
+ * Prefix for assigning names to worker threads
*/
- private volatile long eventWaiters;
-
- private static final int EVENT_COUNT_SHIFT = 32;
- private static final long WAITER_ID_MASK = (1L << 16) - 1L;
+ private final String workerNamePrefix;
/**
- * 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
+ * Sum of per-thread steal counts, updated only when threads are
+ * idle or terminating.
*/
- private volatile int eventCount;
+ private volatile long stealCount;
/**
- * 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 int spareWaiters;
-
- private static final int SPARE_COUNT_SHIFT = 16;
- private static final int SPARE_ID_MASK = (1 << 16) - 1;
+ * Main pool control -- a long packed with:
+ * AC: Number of active running workers minus target parallelism (16 bits)
+ * TC: Number of total workers minus target parallelism (16bits)
+ * ST: true if pool is terminating (1 bit)
+ * EC: the wait count of top waiting thread (15 bits)
+ * ID: ~poolIndex of top of Treiber stack of waiting threads (16 bits)
+ *
+ * When convenient, we can extract the upper 32 bits of counts and
+ * the lower 32 bits of queue state, u = (int)(ctl >>> 32) and e =
+ * (int)ctl. The ec field is never accessed alone, but always
+ * together with id and st. The offsets of counts by the target
+ * parallelism and the positionings of fields makes it possible to
+ * perform the most common checks via sign tests of fields: When
+ * ac is negative, there are not enough active workers, when tc is
+ * negative, there are not enough total workers, when id is
+ * negative, there is at least one waiting worker, and when e is
+ * negative, the pool is terminating. To deal with these possibly
+ * negative fields, we use casts in and out of "short" and/or
+ * signed shifts to maintain signedness.
+ */
+ volatile long ctl;
+
+ // bit positions/shifts for fields
+ private static final int AC_SHIFT = 48;
+ private static final int TC_SHIFT = 32;
+ private static final int ST_SHIFT = 31;
+ private static final int EC_SHIFT = 16;
+
+ // bounds
+ private static final int MAX_ID = 0x7fff; // max poolIndex
+ private static final int SMASK = 0xffff; // mask short bits
+ private static final int SHORT_SIGN = 1 << 15;
+ private static final int INT_SIGN = 1 << 31;
+
+ // masks
+ private static final long STOP_BIT = 0x0001L << ST_SHIFT;
+ private static final long AC_MASK = ((long)SMASK) << AC_SHIFT;
+ private static final long TC_MASK = ((long)SMASK) << TC_SHIFT;
+
+ // units for incrementing and decrementing
+ private static final long TC_UNIT = 1L << TC_SHIFT;
+ private static final long AC_UNIT = 1L << AC_SHIFT;
+
+ // masks and units for dealing with u = (int)(ctl >>> 32)
+ private static final int UAC_SHIFT = AC_SHIFT - 32;
+ private static final int UTC_SHIFT = TC_SHIFT - 32;
+ private static final int UAC_MASK = SMASK << UAC_SHIFT;
+ private static final int UTC_MASK = SMASK << UTC_SHIFT;
+ private static final int UAC_UNIT = 1 << UAC_SHIFT;
+ private static final int UTC_UNIT = 1 << UTC_SHIFT;
+
+ // masks and units for dealing with e = (int)ctl
+ private static final int E_MASK = 0x7fffffff; // no STOP_BIT
+ private static final int EC_UNIT = 1 << EC_SHIFT;
/**
- * 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).
+ * The target parallelism level.
*/
- private 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;
- private static final int ONE_ACTIVE = 1; // active update delta
+ final int parallelism;
/**
- * 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.
+ * Index (mod submission queue length) of next element to take
+ * from submission queue. Usage is identical to that for
+ * per-worker queues -- see ForkJoinWorkerThread internal
+ * documentation.
*/
- private volatile int workerCounts;
+ volatile int queueBase;
- 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;
+ /**
+ * Index (mod submission queue length) of next element to add
+ * in submission queue. Usage is identical to that for
+ * per-worker queues -- see ForkJoinWorkerThread internal
+ * documentation.
+ */
+ int queueTop;
/**
- * The target parallelism level.
- * Accessed directly by ForkJoinWorkerThreads.
+ * True when shutdown() has been called.
*/
- final int parallelism;
+ volatile boolean shutdown;
/**
* True if use local fifo, not default lifo, for local polling
@@ -554,495 +553,665 @@ public class ForkJoinPool extends Abstra
final boolean locallyFifo;
/**
- * The uncaught exception handler used when any worker abruptly
- * terminates.
+ * The number of threads in ForkJoinWorkerThreads.helpQuiescePool.
+ * When non-zero, suppresses automatic shutdown when active
+ * counts become zero.
*/
- private final Thread.UncaughtExceptionHandler ueh;
+ volatile int quiescerCount;
/**
- * Pool number, just for assigning useful names to worker threads
+ * The number of threads blocked in join.
*/
- private final int poolNumber;
+ volatile int blockedCount;
+ /**
+ * Counter for worker Thread names (unrelated to their poolIndex)
+ */
+ private volatile int nextWorkerNumber;
- // Utilities for CASing fields. Note that several of these
- // are manually inlined by callers
+ /**
+ * The index for the next created worker. Accessed under scanGuard.
+ */
+ private int nextWorkerIndex;
/**
- * Increments running count part of workerCounts
+ * SeqLock and index masking for updates to workers array. Locked
+ * when SG_UNIT is set. Unlocking clears bit by adding
+ * SG_UNIT. Staleness of read-only operations can be checked by
+ * comparing scanGuard to value before the reads. The low 16 bits
+ * (i.e, anding with SMASK) hold (the smallest power of two
+ * covering all worker indices, minus one, and is used to avoid
+ * dealing with large numbers of null slots when the workers array
+ * is overallocated.
*/
- final void incrementRunningCount() {
- int c;
- do {} while (!UNSAFE.compareAndSwapInt(this, workerCountsOffset,
- c = workerCounts,
- c + ONE_RUNNING));
- }
+ volatile int scanGuard;
+
+ private static final int SG_UNIT = 1 << 16;
/**
- * Tries to decrement running count unless already zero
+ * The wakeup interval (in nanoseconds) for a worker waiting for a
+ * task when the pool is quiescent to instead try to shrink the
+ * number of workers. The exact value does not matter too
+ * much. It must be short enough to release resources during
+ * sustained periods of idleness, but not so short that threads
+ * are continually re-created.
*/
- final boolean tryDecrementRunningCount() {
- int wc = workerCounts;
- if ((wc & RUNNING_COUNT_MASK) == 0)
- return false;
- return UNSAFE.compareAndSwapInt(this, workerCountsOffset,
- wc, wc - ONE_RUNNING);
- }
+ private static final long SHRINK_RATE =
+ 4L * 1000L * 1000L * 1000L; // 4 seconds
/**
- * Forces decrement of encoded workerCounts, awaiting nonzero if
- * (rarely) necessary when other count updates lag.
+ * Top-level loop for worker threads: On each step: if the
+ * previous step swept through all queues and found no tasks, or
+ * there are excess threads, then possibly blocks. Otherwise,
+ * scans for and, if found, executes a task. Returns when pool
+ * and/or worker terminate.
*
- * @param dr -- either zero or ONE_RUNNING
- * @param dt == either zero or ONE_TOTAL
+ * @param w the worker
*/
- private void decrementWorkerCounts(int dr, int dt) {
- for (;;) {
- int wc = workerCounts;
- if (wc == 0 && (runState & TERMINATED) != 0)
- return; // lagging termination on a backout
- if ((wc & RUNNING_COUNT_MASK) - dr < 0 ||
- (wc >>> TOTAL_COUNT_SHIFT) - dt < 0)
- Thread.yield();
- if (UNSAFE.compareAndSwapInt(this, workerCountsOffset,
- wc, wc - (dr + dt)))
- return;
+ final void work(ForkJoinWorkerThread w) {
+ boolean swept = false; // true on empty scans
+ long c;
+ while (!w.terminate && (int)(c = ctl) >= 0) {
+ int a; // active count
+ if (!swept && (a = (int)(c >> AC_SHIFT)) <= 0)
+ swept = scan(w, a);
+ else if (tryAwaitWork(w, c))
+ swept = false;
}
}
+ // Signalling
+
/**
- * Increments event count
+ * Wakes up or creates a worker.
*/
- private void advanceEventCount() {
- int c;
- do {} while(!UNSAFE.compareAndSwapInt(this, eventCountOffset,
- c = eventCount, c+1));
+ final void signalWork() {
+ /*
+ * The while condition is true if: (there is are too few total
+ * workers OR there is at least one waiter) AND (there are too
+ * few active workers OR the pool is terminating). The value
+ * of e distinguishes the remaining cases: zero (no waiters)
+ * for create, negative if terminating (in which case do
+ * nothing), else release a waiter. The secondary checks for
+ * release (non-null array etc) can fail if the pool begins
+ * terminating after the test, and don't impose any added cost
+ * because JVMs must perform null and bounds checks anyway.
+ */
+ long c; int e, u;
+ while ((((e = (int)(c = ctl)) | (u = (int)(c >>> 32))) &
+ (INT_SIGN|SHORT_SIGN)) == (INT_SIGN|SHORT_SIGN) && e >= 0) {
+ if (e > 0) { // release a waiting worker
+ int i; ForkJoinWorkerThread w; ForkJoinWorkerThread[] ws;
+ if ((ws = workers) == null ||
+ (i = ~e & SMASK) >= ws.length ||
+ (w = ws[i]) == null)
+ break;
+ long nc = (((long)(w.nextWait & E_MASK)) |
+ ((long)(u + UAC_UNIT) << 32));
+ if (w.eventCount == e &&
+ UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc)) {
+ w.eventCount = (e + EC_UNIT) & E_MASK;
+ if (w.parked)
+ UNSAFE.unpark(w);
+ break;
+ }
+ }
+ else if (UNSAFE.compareAndSwapLong
+ (this, ctlOffset, c,
+ (long)(((u + UTC_UNIT) & UTC_MASK) |
+ ((u + UAC_UNIT) & UAC_MASK)) << 32)) {
+ addWorker();
+ break;
+ }
+ }
}
/**
- * Tries incrementing active count; fails on contention.
- * Called by workers before executing tasks.
+ * Variant of signalWork to help release waiters on rescans.
+ * Tries once to release a waiter if active count < 0.
*
- * @return true on success
+ * @return false if failed due to contention, else true
*/
- final boolean tryIncrementActiveCount() {
- int c;
- return UNSAFE.compareAndSwapInt(this, runStateOffset,
- c = runState, c + ONE_ACTIVE);
+ private boolean tryReleaseWaiter() {
+ long c; int e, i; ForkJoinWorkerThread w; ForkJoinWorkerThread[] ws;
+ if ((e = (int)(c = ctl)) > 0 &&
+ (int)(c >> AC_SHIFT) < 0 &&
+ (ws = workers) != null &&
+ (i = ~e & SMASK) < ws.length &&
+ (w = ws[i]) != null) {
+ long nc = ((long)(w.nextWait & E_MASK) |
+ ((c + AC_UNIT) & (AC_MASK|TC_MASK)));
+ if (w.eventCount != e ||
+ !UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc))
+ return false;
+ w.eventCount = (e + EC_UNIT) & E_MASK;
+ if (w.parked)
+ UNSAFE.unpark(w);
+ }
+ return true;
}
+ // Scanning for tasks
+
/**
- * Tries decrementing active count; fails on contention.
- * Called when workers cannot find tasks to run.
+ * Scans for and, if found, executes one task. Scans start at a
+ * random index of workers array, and randomly select the first
+ * (2*#workers)-1 probes, and then, if all empty, resort to 2
+ * circular sweeps, which is necessary to check quiescence. and
+ * taking a submission only if no stealable tasks were found. The
+ * steal code inside the loop is a specialized form of
+ * ForkJoinWorkerThread.deqTask, followed bookkeeping to support
+ * helpJoinTask and signal propagation. The code for submission
+ * queues is almost identical. On each steal, the worker completes
+ * not only the task, but also all local tasks that this task may
+ * have generated. On detecting staleness or contention when
+ * trying to take a task, this method returns without finishing
+ * sweep, which allows global state rechecks before retry.
+ *
+ * @param w the worker
+ * @param a the number of active workers
+ * @return true if swept all queues without finding a task
*/
- final boolean tryDecrementActiveCount() {
- int c;
- return UNSAFE.compareAndSwapInt(this, runStateOffset,
- c = runState, c - ONE_ACTIVE);
+ private boolean scan(ForkJoinWorkerThread w, int a) {
+ int g = scanGuard; // mask 0 avoids useless scans if only one active
+ int m = parallelism == 1 - a? 0 : g & SMASK;
+ ForkJoinWorkerThread[] ws = workers;
+ if (ws == null || ws.length <= m) // staleness check
+ return false;
+ for (int r = w.seed, k = r, j = -(m + m); j <= m + m; ++j) {
+ ForkJoinTask> t; ForkJoinTask>[] q; int b, i;
+ ForkJoinWorkerThread v = ws[k & m];
+ if (v != null && (b = v.queueBase) != v.queueTop &&
+ (q = v.queue) != null && (i = (q.length - 1) & b) >= 0) {
+ long u = (i << ASHIFT) + ABASE;
+ if ((t = q[i]) != null && v.queueBase == b &&
+ UNSAFE.compareAndSwapObject(q, u, t, null)) {
+ int d = (v.queueBase = b + 1) - v.queueTop;
+ v.stealHint = w.poolIndex;
+ if (d != 0)
+ signalWork(); // propagate if nonempty
+ w.execTask(t);
+ }
+ r ^= r << 13; r ^= r >>> 17; w.seed = r ^ (r << 5);
+ return false; // store next seed
+ }
+ else if (j < 0) { // xorshift
+ r ^= r << 13; r ^= r >>> 17; k = r ^= r << 5;
+ }
+ else
+ ++k;
+ }
+ if (scanGuard != g) // staleness check
+ return false;
+ else { // try to take submission
+ ForkJoinTask> t; ForkJoinTask>[] q; int b, i;
+ if ((b = queueBase) != queueTop &&
+ (q = submissionQueue) != null &&
+ (i = (q.length - 1) & b) >= 0) {
+ long u = (i << ASHIFT) + ABASE;
+ if ((t = q[i]) != null && queueBase == b &&
+ UNSAFE.compareAndSwapObject(q, u, t, null)) {
+ queueBase = b + 1;
+ w.execTask(t);
+ }
+ return false;
+ }
+ return true; // all queues empty
+ }
}
/**
- * Advances to at least the given level. Returns true if not
- * already in at least the given level.
- */
- private boolean advanceRunLevel(int level) {
- for (;;) {
- int s = runState;
- if ((s & level) != 0)
- return false;
- if (UNSAFE.compareAndSwapInt(this, runStateOffset, s, s | level))
+ * Tries to enqueue worker w in wait queue and await change in
+ * worker's eventCount. If the pool is quiescent, possibly
+ * terminates worker upon exit. Otherwise, before blocking,
+ * rescans queues to avoid missed signals. Upon finding work,
+ * releases at least one worker (which may be the current
+ * worker). Rescans restart upon detected staleness or failure to
+ * release due to contention. Note the unusual conventions about
+ * Thread.interrupt here and elsewhere: Because interrupts are
+ * used solely to alert threads to check termination, which is
+ * checked here anyway, we clear status (using Thread.interrupted)
+ * before any call to park, so that park does not immediately
+ * return due to status being set via some other unrelated call to
+ * interrupt in user code.
+ *
+ * @param w the calling worker
+ * @param c the ctl value on entry
+ * @return true if waited or another thread was released upon enq
+ */
+ private boolean tryAwaitWork(ForkJoinWorkerThread w, long c) {
+ int v = w.eventCount;
+ w.nextWait = (int)c; // w's successor record
+ long nc = (long)(v & E_MASK) | ((c - AC_UNIT) & (AC_MASK|TC_MASK));
+ if (ctl != c || !UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc)) {
+ long d = ctl; // return true if lost to a deq, to force scan
+ return (int)d != (int)c && ((d - c) & AC_MASK) >= 0L;
+ }
+ for (int sc = w.stealCount; sc != 0;) { // accumulate stealCount
+ long s = stealCount;
+ if (UNSAFE.compareAndSwapLong(this, stealCountOffset, s, s + sc))
+ sc = w.stealCount = 0;
+ else if (w.eventCount != v)
+ return true; // update next time
+ }
+ if (parallelism + (int)(nc >> AC_SHIFT) == 0 &&
+ blockedCount == 0 && quiescerCount == 0)
+ idleAwaitWork(w, nc, c, v); // quiescent
+ for (boolean rescanned = false;;) {
+ if (w.eventCount != v)
return true;
+ if (!rescanned) {
+ int g = scanGuard, m = g & SMASK;
+ ForkJoinWorkerThread[] ws = workers;
+ if (ws != null && m < ws.length) {
+ rescanned = true;
+ for (int i = 0; i <= m; ++i) {
+ ForkJoinWorkerThread u = ws[i];
+ if (u != null) {
+ if (u.queueBase != u.queueTop &&
+ !tryReleaseWaiter())
+ rescanned = false; // contended
+ if (w.eventCount != v)
+ return true;
+ }
+ }
+ }
+ if (scanGuard != g || // stale
+ (queueBase != queueTop && !tryReleaseWaiter()))
+ rescanned = false;
+ if (!rescanned)
+ Thread.yield(); // reduce contention
+ else
+ Thread.interrupted(); // clear before park
+ }
+ else {
+ w.parked = true; // must recheck
+ if (w.eventCount != v) {
+ w.parked = false;
+ return true;
+ }
+ LockSupport.park(this);
+ rescanned = w.parked = false;
+ }
}
}
- // workers array maintenance
-
/**
- * Records and returns a workers array index for new worker.
- */
- 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);
+ * If inactivating worker w has caused pool to become
+ * quiescent, check for pool termination, and wait for event
+ * for up to SHRINK_RATE nanosecs (rescans are unnecessary in
+ * this case because quiescence reflects consensus about lack
+ * of work). On timeout, if ctl has not changed, terminate the
+ * worker. Upon its termination (see deregisterWorker), it may
+ * wake up another worker to possibly repeat this process.
+ *
+ * @param w the calling worker
+ * @param currentCtl the ctl value after enqueuing w
+ * @param prevCtl the ctl value if w terminated
+ * @param v the eventCount w awaits change
+ */
+ private void idleAwaitWork(ForkJoinWorkerThread w, long currentCtl,
+ long prevCtl, int v) {
+ if (w.eventCount == v) {
+ if (shutdown)
+ tryTerminate(false);
+ ForkJoinTask.helpExpungeStaleExceptions(); // help clean weak refs
+ while (ctl == currentCtl) {
+ long startTime = System.nanoTime();
+ w.parked = true;
+ if (w.eventCount == v) // must recheck
+ LockSupport.parkNanos(this, SHRINK_RATE);
+ w.parked = false;
+ if (w.eventCount != v)
+ break;
+ else if (System.nanoTime() - startTime < SHRINK_RATE)
+ Thread.interrupted(); // spurious wakeup
+ else if (UNSAFE.compareAndSwapLong(this, ctlOffset,
+ currentCtl, prevCtl)) {
+ w.terminate = true; // restore previous
+ w.eventCount = ((int)currentCtl + EC_UNIT) & E_MASK;
+ break;
+ }
}
- ws[k] = w;
- workers = ws; // volatile array write ensures slot visibility
- } finally {
- lock.unlock();
}
- return k;
}
+ // Submissions
+
/**
- * Nulls out record of worker in workers array
+ * Enqueues the given task in the submissionQueue. Same idea as
+ * ForkJoinWorkerThread.pushTask except for use of submissionLock.
+ *
+ * @param t the task
*/
- private void forgetWorker(ForkJoinWorkerThread w) {
- int idx = w.poolIndex;
- // Locking helps method recordWorker avoid unecessary expansion
- final ReentrantLock lock = this.workerLock;
+ private void addSubmission(ForkJoinTask> t) {
+ final ReentrantLock lock = this.submissionLock;
lock.lock();
try {
- ForkJoinWorkerThread[] ws = workers;
- if (idx >= 0 && idx < ws.length && ws[idx] == w) // verify
- ws[idx] = null;
+ ForkJoinTask>[] q; int s, m;
+ if ((q = submissionQueue) != null) { // ignore if queue removed
+ long u = (((s = queueTop) & (m = q.length-1)) << ASHIFT)+ABASE;
+ UNSAFE.putOrderedObject(q, u, t);
+ queueTop = s + 1;
+ if (s - queueBase == m)
+ growSubmissionQueue();
+ }
} finally {
lock.unlock();
}
+ signalWork();
}
- // adding and removing workers
+ // (pollSubmission is defined below with exported methods)
/**
- * Tries to create and add new worker. Assumes that worker counts
- * are already updated to accommodate the worker, so adjusts on
- * failure.
+ * Creates or doubles submissionQueue array.
+ * Basically identical to ForkJoinWorkerThread version.
*/
- private void 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
+ private void growSubmissionQueue() {
+ ForkJoinTask>[] oldQ = submissionQueue;
+ int size = oldQ != null ? oldQ.length << 1 : INITIAL_QUEUE_CAPACITY;
+ if (size > MAXIMUM_QUEUE_CAPACITY)
+ throw new RejectedExecutionException("Queue capacity exceeded");
+ if (size < INITIAL_QUEUE_CAPACITY)
+ size = INITIAL_QUEUE_CAPACITY;
+ ForkJoinTask>[] q = submissionQueue = new ForkJoinTask>[size];
+ int mask = size - 1;
+ int top = queueTop;
+ int oldMask;
+ if (oldQ != null && (oldMask = oldQ.length - 1) >= 0) {
+ for (int b = queueBase; b != top; ++b) {
+ long u = ((b & oldMask) << ASHIFT) + ABASE;
+ Object x = UNSAFE.getObjectVolatile(oldQ, u);
+ if (x != null && UNSAFE.compareAndSwapObject(oldQ, u, x, null))
+ UNSAFE.putObjectVolatile
+ (q, ((b & mask) << ASHIFT) + ABASE, x);
}
}
- if (w != null)
- w.start(recordWorker(w), ueh);
}
+ // Blocking support
+
/**
- * 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
- */
- 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);
+ * Tries to increment blockedCount, decrement active count
+ * (sometimes implicitly) and possibly release or create a
+ * compensating worker in preparation for blocking. Fails
+ * on contention or termination.
+ *
+ * @return true if the caller can block, else should recheck and retry
+ */
+ private boolean tryPreBlock() {
+ int b = blockedCount;
+ if (UNSAFE.compareAndSwapInt(this, blockedCountOffset, b, b + 1)) {
+ int pc = parallelism;
+ do {
+ ForkJoinWorkerThread[] ws; ForkJoinWorkerThread w;
+ int e, ac, tc, rc, i;
+ long c = ctl;
+ int u = (int)(c >>> 32);
+ if ((e = (int)c) < 0) {
+ // skip -- terminating
+ }
+ else if ((ac = (u >> UAC_SHIFT)) <= 0 && e != 0 &&
+ (ws = workers) != null &&
+ (i = ~e & SMASK) < ws.length &&
+ (w = ws[i]) != null) {
+ long nc = ((long)(w.nextWait & E_MASK) |
+ (c & (AC_MASK|TC_MASK)));
+ if (w.eventCount == e &&
+ UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc)) {
+ w.eventCount = (e + EC_UNIT) & E_MASK;
+ if (w.parked)
+ UNSAFE.unpark(w);
+ return true; // release an idle worker
+ }
+ }
+ else if ((tc = (short)(u >>> UTC_SHIFT)) >= 0 && ac + pc > 1) {
+ long nc = ((c - AC_UNIT) & AC_MASK) | (c & ~AC_MASK);
+ if (UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc))
+ return true; // no compensation needed
+ }
+ else if (tc + pc < MAX_ID) {
+ long nc = ((c + TC_UNIT) & TC_MASK) | (c & ~TC_MASK);
+ if (UNSAFE.compareAndSwapLong(this, ctlOffset, c, nc)) {
+ addWorker();
+ return true; // create a replacement
+ }
+ }
+ // try to back out on any failure and let caller retry
+ } while (!UNSAFE.compareAndSwapInt(this, blockedCountOffset,
+ b = blockedCount, b - 1));
+ }
+ return false;
}
- // Waiting for and signalling events
-
/**
- * 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.
- *
- * @param signalling true if caller is a signalling worker so can
- * exit upon (conservatively) detected contention by other threads
- * who will continue to release
+ * Decrements blockedCount and increments active count
*/
- private void releaseEventWaiters(boolean signalling) {
- ForkJoinWorkerThread[] ws = workers;
- int n = ws.length;
- long h; // head of stack
- ForkJoinWorkerThread w; int id, ec;
- while ((id = ((int)((h = eventWaiters) & WAITER_ID_MASK)) - 1) >= 0 &&
- (int)(h >>> EVENT_COUNT_SHIFT) != (ec = eventCount) &&
- id < n && (w = ws[id]) != null) {
- if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
- h, h = w.nextWaiter))
- LockSupport.unpark(w);
- if (signalling && (eventCount != ec || eventWaiters != h))
- break;
- }
+ private void postBlock() {
+ long c;
+ do {} while (!UNSAFE.compareAndSwapLong(this, ctlOffset, // no mask
+ c = ctl, c + AC_UNIT));
+ int b;
+ do {} while(!UNSAFE.compareAndSwapInt(this, blockedCountOffset,
+ b = blockedCount, b - 1));
}
/**
- * Tries to advance eventCount and releases waiters. Called only
- * from workers.
+ * Possibly blocks waiting for the given task to complete, or
+ * cancels the task if terminating. Fails to wait if contended.
+ *
+ * @param joinMe the task
*/
- 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(true);
+ final void tryAwaitJoin(ForkJoinTask> joinMe) {
+ int s;
+ Thread.interrupted(); // clear interrupts before checking termination
+ if (joinMe.status >= 0) {
+ if (tryPreBlock()) {
+ joinMe.tryAwaitDone(0L);
+ postBlock();
+ }
+ else if ((ctl & STOP_BIT) != 0L)
+ joinMe.cancelIgnoringExceptions();
+ }
}
/**
- * Blocks worker until terminating or event count
- * advances from last value held by worker
- *
- * @param w the calling worker thread
- */
- private void eventSync(ForkJoinWorkerThread w) {
- int wec = w.lastEventCount;
- long nh = (((long)wec) << EVENT_COUNT_SHIFT) | ((long)(w.poolIndex+1));
- long h;
- while ((runState < SHUTDOWN || !tryTerminate(false)) &&
- ((h = eventWaiters) == 0L ||
- (int)(h >>> EVENT_COUNT_SHIFT) == wec) &&
- eventCount == wec) {
- if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
- w.nextWaiter = h, nh)) {
- while (runState < TERMINATING && eventCount == wec) {
- if (!tryAccumulateStealCount(w)) // transfer while idle
- continue;
- Thread.interrupted(); // clear/ignore interrupt
- if (eventCount != wec)
+ * Possibly blocks the given worker waiting for joinMe to
+ * complete or timeout
+ *
+ * @param joinMe the task
+ * @param millis the wait time for underlying Object.wait
+ */
+ final void timedAwaitJoin(ForkJoinTask> joinMe, long nanos) {
+ while (joinMe.status >= 0) {
+ Thread.interrupted();
+ if ((ctl & STOP_BIT) != 0L) {
+ joinMe.cancelIgnoringExceptions();
+ break;
+ }
+ if (tryPreBlock()) {
+ long last = System.nanoTime();
+ while (joinMe.status >= 0) {
+ long millis = TimeUnit.NANOSECONDS.toMillis(nanos);
+ if (millis <= 0)
+ break;
+ joinMe.tryAwaitDone(millis);
+ if (joinMe.status < 0)
break;
- LockSupport.park(w);
+ if ((ctl & STOP_BIT) != 0L) {
+ joinMe.cancelIgnoringExceptions();
+ break;
+ }
+ long now = System.nanoTime();
+ nanos -= now - last;
+ last = now;
}
+ postBlock();
break;
}
}
- w.lastEventCount = eventCount;
}
- // Maintaining spares
-
/**
- * Pushes worker onto the spare stack
+ * If necessary, compensates for blocker, and blocks
*/
- 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));
- }
-
- /**
- * Tries (once) to resume a spare if running count is less than
- * target parallelism. Fails on contention or stale workers.
- */
- private void tryResumeSpare() {
- 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 &&
- eventWaiters == 0L &&
- spareWaiters == sw &&
- UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
- sw, w.nextSpare) &&
- w.tryUnsuspend()) {
- int c; // try increment; if contended, finish after unpark
- boolean inc = UNSAFE.compareAndSwapInt(this, workerCountsOffset,
- c = workerCounts,
- c + ONE_RUNNING);
- LockSupport.unpark(w);
- if (!inc) {
- do {} while(!UNSAFE.compareAndSwapInt(this, workerCountsOffset,
- c = workerCounts,
- c + ONE_RUNNING));
+ private void awaitBlocker(ManagedBlocker blocker)
+ throws InterruptedException {
+ while (!blocker.isReleasable()) {
+ if (tryPreBlock()) {
+ try {
+ do {} while (!blocker.isReleasable() && !blocker.block());
+ } finally {
+ postBlock();
+ }
+ break;
}
}
}
- /**
- * Callback from oldest spare occasionally waking up. Tries
- * (once) to shutdown a spare if more than 25% spare overage, or
- * if UNUSED_SPARE_TRIM_RATE_NANOS have elapsed and there are at
- * least #parallelism running threads. Note that we don't need CAS
- * or locks here because the method is called only from the oldest
- * suspended spare occasionally waking (and even misfires are OK).
- *
- * @param now the wake up nanoTime of caller
- */
- final void tryTrimSpare(long now) {
- long lastTrim = trimTime;
- trimTime = now;
- helpMaintainParallelism(); // first, help wake up any needed spares
- int sw, id;
- ForkJoinWorkerThread w;
- ForkJoinWorkerThread[] ws;
- int pc = parallelism;
- int wc = workerCounts;
- if ((wc & RUNNING_COUNT_MASK) >= pc &&
- (((wc >>> TOTAL_COUNT_SHIFT) - pc) > (pc >>> 2) + 1 ||// approx 25%
- now - lastTrim >= UNUSED_SPARE_TRIM_RATE_NANOS) &&
- (id = ((sw = spareWaiters) & SPARE_ID_MASK) - 1) >= 0 &&
- id < (ws = workers).length && (w = ws[id]) != null &&
- UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
- sw, w.nextSpare))
- w.shutdown(false);
- }
+ // Creating, registering and deregistring workers
/**
- * Does at most one of:
- *
- * 1. Help wake up existing workers waiting for work via
- * releaseEventWaiters. (If any exist, then it probably doesn't
- * matter right now if under target parallelism level.)
- *
- * 2. If below parallelism level and a spare exists, try (once)
- * to resume it via tryResumeSpare.
- *
- * 3. If neither of the above, tries (once) to add a new
- * worker if either there are not enough total, or if all
- * existing workers are busy, there are either no running
- * workers or the deficit is at least twice the surplus.
- */
- private void helpMaintainParallelism() {
- // uglified to work better when not compiled
- int pc, wc, rc, tc, rs; long h;
- if ((h = eventWaiters) != 0L) {
- if ((int)(h >>> EVENT_COUNT_SHIFT) != eventCount)
- releaseEventWaiters(false); // avoid useless call
- }
- else if ((pc = parallelism) >
- (rc = ((wc = workerCounts) & RUNNING_COUNT_MASK))) {
- if (spareWaiters != 0)
- tryResumeSpare();
- else if ((rs = runState) < TERMINATING &&
- ((tc = wc >>> TOTAL_COUNT_SHIFT) < pc ||
- (tc == (rs & ACTIVE_COUNT_MASK) && // all busy
- (rc == 0 || // must add
- rc < pc - ((tc - pc) << 1)) && // within slack
- tc < MAX_WORKERS && runState == rs)) && // recheck busy
- workerCounts == wc &&
- UNSAFE.compareAndSwapInt(this, workerCountsOffset, wc,
- wc + (ONE_RUNNING|ONE_TOTAL)))
- addWorker();
+ * Tries to create and start a worker; minimally rolls back counts
+ * on failure.
+ */
+ private void addWorker() {
+ Throwable ex = null;
+ ForkJoinWorkerThread t = null;
+ try {
+ t = factory.newThread(this);
+ } catch (Throwable e) {
+ ex = e;
+ }
+ if (t == null) { // null or exceptional factory return
+ long c; // adjust counts
+ do {} while (!UNSAFE.compareAndSwapLong
+ (this, ctlOffset, c = ctl,
+ (((c - AC_UNIT) & AC_MASK) |
+ ((c - TC_UNIT) & TC_MASK) |
+ (c & ~(AC_MASK|TC_MASK)))));
+ // Propagate exception if originating from an external caller
+ if (!tryTerminate(false) && ex != null &&
+ !(Thread.currentThread() instanceof ForkJoinWorkerThread))
+ UNSAFE.throwException(ex);
}
+ else
+ t.start();
}
/**
- * 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 cannot find work (misses > 0), updates its
- * active status to inactive and updates activeCount unless
- * this is the first miss and there is contention, in which
- * case it may try again (either in this or a subsequent
- * call).
- *
- * 2. If there are at least 2 misses, awaits the next task event
- * via eventSync
- *
- * 3. If there are too many running threads, suspends this worker
- * (first forcing inactivation if necessary). If it is not
- * needed, it may be killed while suspended via
- * tryTrimSpare. Otherwise, upon resume it rechecks to make
- * sure that it is still needed.
- *
- * 4. Helps release and/or reactivate other workers via
- * helpMaintainParallelism
- *
- * @param w the worker
- * @param misses the number of scans by caller failing to find work
- * (saturating at 2 just to avoid wraparound)
+ * Callback from ForkJoinWorkerThread constructor to assign a
+ * public name
*/
- final void preStep(ForkJoinWorkerThread w, int misses) {
- boolean active = w.active;
- int pc = parallelism;
- for (;;) {
- int wc = workerCounts;
- int rc = wc & RUNNING_COUNT_MASK;
- if (active && (misses > 0 || rc > pc)) {
- int rs; // try inactivate
- if (UNSAFE.compareAndSwapInt(this, runStateOffset,
- rs = runState, rs - ONE_ACTIVE))
- active = w.active = false;
- else if (misses > 1 || rc > pc ||
- (rs & ACTIVE_COUNT_MASK) >= pc)
- continue; // force inactivate
- }
- if (misses > 1) {
- misses = 0; // don't re-sync
- eventSync(w); // continue loop to recheck rc
- }
- else if (rc > pc) {
- if (workerCounts == wc && // try to suspend as spare
- UNSAFE.compareAndSwapInt(this, workerCountsOffset,
- wc, wc - ONE_RUNNING) &&
- !w.suspendAsSpare()) // false if killed
- break;
- }
- else {
- if (rc < pc || eventWaiters != 0L)
- helpMaintainParallelism();
- break;
- }
+ final String nextWorkerName() {
+ for (int n;;) {
+ if (UNSAFE.compareAndSwapInt(this, nextWorkerNumberOffset,
+ n = nextWorkerNumber, ++n))
+ return workerNamePrefix + n;
}
}
/**
- * 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.
+ * Callback from ForkJoinWorkerThread constructor to
+ * determine its poolIndex and record in workers array.
*
- * @param joinMe the task to join
+ * @param w the worker
+ * @return the worker's pool index
*/
- 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;
+ final int registerWorker(ForkJoinWorkerThread w) {
+ /*
+ * In the typical case, a new worker acquires the lock, uses
+ * next available index and returns quickly. Since we should
+ * not block callers (ultimately from signalWork or
+ * tryPreBlock) waiting for the lock needed to do this, we
+ * instead help release other workers while waiting for the
+ * lock.
+ */
+ for (int g;;) {
+ ForkJoinWorkerThread[] ws;
+ if (((g = scanGuard) & SG_UNIT) == 0 &&
+ UNSAFE.compareAndSwapInt(this, scanGuardOffset,
+ g, g | SG_UNIT)) {
+ int k = nextWorkerIndex;
+ try {
+ if ((ws = workers) != null) { // ignore on shutdown
+ 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 = workers = Arrays.copyOf(ws, n << 1);
+ }
+ ws[k] = w;
+ nextWorkerIndex = k + 1;
+ int m = g & SMASK;
+ g = k >= m? ((m << 1) + 1) & SMASK : g + (SG_UNIT<<1);
+ }
+ } finally {
+ scanGuard = g;
+ }
+ return k;
}
- 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;
+ else if ((ws = workers) != null) { // help release others
+ for (ForkJoinWorkerThread u : ws) {
+ if (u != null && u.queueBase != u.queueTop) {
+ if (tryReleaseWaiter())
+ break;
+ }
+ }
}
}
}
/**
- * Same idea as awaitJoin, but no helping
+ * Final callback from terminating worker. Removes record of
+ * worker from array, and adjusts counts. If pool is shutting
+ * down, tries to complete termination.
+ *
+ * @param w the worker
*/
- 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;
+ final void deregisterWorker(ForkJoinWorkerThread w, Throwable ex) {
+ int idx = w.poolIndex;
+ int sc = w.stealCount;
+ int steps = 0;
+ // Remove from array, adjust worker counts and collect steal count.
+ // We can intermix failed removes or adjusts with steal updates
+ do {
+ long s, c;
+ int g;
+ if (steps == 0 && ((g = scanGuard) & SG_UNIT) == 0 &&
+ UNSAFE.compareAndSwapInt(this, scanGuardOffset,
+ g, g |= SG_UNIT)) {
+ ForkJoinWorkerThread[] ws = workers;
+ if (ws != null && idx >= 0 &&
+ idx < ws.length && ws[idx] == w)
+ ws[idx] = null; // verify
+ nextWorkerIndex = idx;
+ scanGuard = g + SG_UNIT;
+ steps = 1;
}
+ if (steps == 1 &&
+ UNSAFE.compareAndSwapLong(this, ctlOffset, c = ctl,
+ (((c - AC_UNIT) & AC_MASK) |
+ ((c - TC_UNIT) & TC_MASK) |
+ (c & ~(AC_MASK|TC_MASK)))))
+ steps = 2;
+ if (sc != 0 &&
+ UNSAFE.compareAndSwapLong(this, stealCountOffset,
+ s = stealCount, s + sc))
+ sc = 0;
+ } while (steps != 2 || sc != 0);
+ if (!tryTerminate(false)) {
+ if (ex != null) // possibly replace if died abnormally
+ signalWork();
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));
- }
- break;
- }
+ tryReleaseWaiter();
}
}
+ // Shutdown and termination
+
/**
* Possibly initiates and/or completes termination.
*
@@ -1051,98 +1220,132 @@ public class ForkJoinPool extends Abstra
* @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();
+ long c;
+ while (((c = ctl) & STOP_BIT) == 0) {
+ if (!now) {
+ if ((int)(c >> AC_SHIFT) != -parallelism)
+ return false;
+ if (!shutdown || blockedCount != 0 || quiescerCount != 0 ||
+ queueBase != queueTop) {
+ if (ctl == c) // staleness check
+ return false;
+ continue;
+ }
+ }
+ if (UNSAFE.compareAndSwapLong(this, ctlOffset, c, c | STOP_BIT))
+ startTerminating();
+ }
+ if ((short)(c >>> TC_SHIFT) == -parallelism) { // signal when 0 workers
+ final ReentrantLock lock = this.submissionLock;
+ lock.lock();
+ try {
+ termination.signalAll();
+ } finally {
+ lock.unlock();
+ }
}
return true;
}
/**
- * Actions on transition to TERMINATING
- *
- * Runs up to four passes through workers: (0) shutting down each
- * quietly (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
+ * Runs up to three passes through workers: (0) Setting
+ * termination status for each worker, followed by wakeups up to
+ * queued workers; (1) helping cancel tasks; (2) interrupting
+ * lagging threads (likely in external tasks, but possibly also
+ * blocked in joins). Each pass repeats previous steps because of
+ * potential lagging thread creation.
*/
private void startTerminating() {
cancelSubmissions();
- for (int passes = 0; passes < 4 && workerCounts != 0; ++passes) {
- advanceEventCount();
- eventWaiters = 0L; // clobber lists
- spareWaiters = 0;
+ for (int pass = 0; pass < 3; ++pass) {
ForkJoinWorkerThread[] ws = workers;
- int n = ws.length;
- for (int i = 0; i < n; ++i) {
- ForkJoinWorkerThread w = ws[i];
- if (w != null) {
- w.shutdown(true);
- if (passes > 0 && !w.isTerminated()) {
- w.cancelTasks();
- LockSupport.unpark(w);
- if (passes > 1) {
- try {
- w.interrupt();
- } catch (SecurityException ignore) {
+ if (ws != null) {
+ for (ForkJoinWorkerThread w : ws) {
+ if (w != null) {
+ w.terminate = true;
+ if (pass > 0) {
+ w.cancelTasks();
+ if (pass > 1 && !w.isInterrupted()) {
+ try {
+ w.interrupt();
+ } catch (SecurityException ignore) {
+ }
}
}
}
}
+ terminateWaiters();
}
}
}
/**
- * Clear out and cancel submissions, ignoring exceptions
+ * Polls and cancels all submissions. Called only during termination.
*/
private void cancelSubmissions() {
- ForkJoinTask> task;
- while ((task = submissionQueue.poll()) != null) {
- try {
- task.cancel(false);
- } catch (Throwable ignore) {
+ while (queueBase != queueTop) {
+ ForkJoinTask> task = pollSubmission();
+ if (task != null) {
+ try {
+ task.cancel(false);
+ } catch (Throwable ignore) {
+ }
}
}
}
- // misc support for ForkJoinWorkerThread
+ /**
+ * Tries to set the termination status of waiting workers, and
+ * then wakes them up (after which they will terminate).
+ */
+ private void terminateWaiters() {
+ ForkJoinWorkerThread[] ws = workers;
+ if (ws != null) {
+ ForkJoinWorkerThread w; long c; int i, e;
+ int n = ws.length;
+ while ((i = ~(e = (int)(c = ctl)) & SMASK) < n &&
+ (w = ws[i]) != null && w.eventCount == (e & E_MASK)) {
+ if (UNSAFE.compareAndSwapLong(this, ctlOffset, c,
+ (long)(w.nextWait & E_MASK) |
+ ((c + AC_UNIT) & AC_MASK) |
+ (c & (TC_MASK|STOP_BIT)))) {
+ w.terminate = true;
+ w.eventCount = e + EC_UNIT;
+ if (w.parked)
+ UNSAFE.unpark(w);
+ }
+ }
+ }
+ }
+
+ // misc ForkJoinWorkerThread support
/**
- * Returns pool number
+ * Increment or decrement quiescerCount. Needed only to prevent
+ * triggering shutdown if a worker is transiently inactive while
+ * checking quiescence.
+ *
+ * @param delta 1 for increment, -1 for decrement
*/
- final int getPoolNumber() {
- return poolNumber;
+ final void addQuiescerCount(int delta) {
+ int c;
+ do {} while(!UNSAFE.compareAndSwapInt(this, quiescerCountOffset,
+ c = quiescerCount, c + delta));
}
/**
- * Tries to accumulates steal count from a worker, clearing
- * the worker's value.
+ * Directly increment or decrement active count without
+ * queuing. This method is used to transiently assert inactivation
+ * while checking quiescence.
*
- * @return true if worker steal count now zero
+ * @param delta 1 for increment, -1 for decrement
*/
- 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;
+ final void addActiveCount(int delta) {
+ long d = delta < 0 ? -AC_UNIT : AC_UNIT;
+ long c;
+ do {} while (!UNSAFE.compareAndSwapLong(this, ctlOffset, c = ctl,
+ ((c + d) & AC_MASK) |
+ (c & ~AC_MASK)));
}
/**
@@ -1150,13 +1353,17 @@ public class ForkJoinPool extends Abstra
* 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;
+ // Approximate at powers of two for small values, saturate past 4
+ int p = parallelism;
+ int a = p + (int)(ctl >> AC_SHIFT);
+ return (a > (p >>>= 1) ? 0 :
+ a > (p >>>= 1) ? 1 :
+ a > (p >>>= 1) ? 2 :
+ a > (p >>>= 1) ? 4 :
+ 8);
}
- // Public and protected methods
+ // Exported methods
// Constructors
@@ -1203,13 +1410,13 @@ public class ForkJoinPool extends Abstra
* 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
.
+ * tasks. For default value, use {@code 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
.
+ * For default value, use {@code false}.
* @throws IllegalArgumentException if parallelism less than or
* equal to zero, or greater than implementation limit
* @throws NullPointerException if the factory is null
@@ -1225,52 +1432,42 @@ public class ForkJoinPool extends Abstra
checkPermission();
if (factory == null)
throw new NullPointerException();
- if (parallelism <= 0 || parallelism > MAX_WORKERS)
+ if (parallelism <= 0 || parallelism > MAX_ID)
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();
- this.trimTime = System.nanoTime();
- }
-
- /**
- * 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;
+ long np = (long)(-parallelism); // offset ctl counts
+ this.ctl = ((np << AC_SHIFT) & AC_MASK) | ((np << TC_SHIFT) & TC_MASK);
+ this.submissionQueue = new ForkJoinTask>[INITIAL_QUEUE_CAPACITY];
+ // initialize workers array with room for 2*parallelism if possible
+ int n = parallelism << 1;
+ if (n >= MAX_ID)
+ n = MAX_ID;
+ else { // See Hackers Delight, sec 3.2, where n < (1 << 16)
+ n |= n >>> 1; n |= n >>> 2; n |= n >>> 4; n |= n >>> 8;
+ }
+ workers = new ForkJoinWorkerThread[n + 1];
+ this.submissionLock = new ReentrantLock();
+ this.termination = submissionLock.newCondition();
+ StringBuilder sb = new StringBuilder("ForkJoinPool-");
+ sb.append(poolNumberGenerator.incrementAndGet());
+ sb.append("-worker-");
+ this.workerNamePrefix = sb.toString();
}
// Execution methods
/**
- * Common code for execute, invoke and submit
- */
- private void doSubmit(ForkJoinTask task) {
- if (task == null)
- throw new NullPointerException();
- if (runState >= SHUTDOWN)
- throw new RejectedExecutionException();
- submissionQueue.offer(task);
- advanceEventCount();
- helpMaintainParallelism(); // start or wake up workers
- }
-
- /**
* Performs the given task, returning its result upon completion.
+ * If the computation encounters an unchecked Exception or Error,
+ * it is rethrown as the outcome of this invocation. Rethrown
+ * exceptions behave in the same way as regular exceptions, but,
+ * when possible, contain stack traces (as displayed for example
+ * using {@code ex.printStackTrace()}) of both the current thread
+ * as well as the thread actually encountering the exception;
+ * minimally only the latter.
*
* @param task the task
* @return the task's result
@@ -1279,15 +1476,38 @@ public class ForkJoinPool extends Abstra
* scheduled for execution
*/
public T invoke(ForkJoinTask task) {
- doSubmit(task);
- return task.join();
+ Thread t = Thread.currentThread();
+ if (task == null)
+ throw new NullPointerException();
+ if (shutdown)
+ throw new RejectedExecutionException();
+ if ((t instanceof ForkJoinWorkerThread) &&
+ ((ForkJoinWorkerThread)t).pool == this)
+ return task.invoke(); // bypass submit if in same pool
+ else {
+ addSubmission(task);
+ return task.join();
+ }
+ }
+
+ /**
+ * Unless terminating, forks task if within an ongoing FJ
+ * computation in the current pool, else submits as external task.
+ */
+ private void forkOrSubmit(ForkJoinTask task) {
+ ForkJoinWorkerThread w;
+ Thread t = Thread.currentThread();
+ if (shutdown)
+ throw new RejectedExecutionException();
+ if ((t instanceof ForkJoinWorkerThread) &&
+ (w = (ForkJoinWorkerThread)t).pool == this)
+ w.pushTask(task);
+ else
+ addSubmission(task);
}
/**
* Arranges for (asynchronous) execution of the given task.
- * If the caller is already engaged in a fork/join computation in
- * the current pool, this method is equivalent in effect to
- * {@link ForkJoinTask#fork}.
*
* @param task the task
* @throws NullPointerException if the task is null
@@ -1295,7 +1515,9 @@ public class ForkJoinPool extends Abstra
* scheduled for execution
*/
public void execute(ForkJoinTask> task) {
- doSubmit(task);
+ if (task == null)
+ throw new NullPointerException();
+ forkOrSubmit(task);
}
// AbstractExecutorService methods
@@ -1306,19 +1528,18 @@ public class ForkJoinPool extends Abstra
* scheduled for execution
*/
public void execute(Runnable task) {
+ if (task == null)
+ throw new NullPointerException();
ForkJoinTask> job;
if (task instanceof ForkJoinTask>) // avoid re-wrap
job = (ForkJoinTask>) task;
else
job = ForkJoinTask.adapt(task, null);
- doSubmit(job);
+ forkOrSubmit(job);
}
/**
* Submits a ForkJoinTask for execution.
- * If the caller is already engaged in a fork/join computation in
- * the current pool, this method is equivalent in effect to
- * {@link ForkJoinTask#fork}.
*
* @param task the task to submit
* @return the task
@@ -1327,7 +1548,9 @@ public class ForkJoinPool extends Abstra
* scheduled for execution
*/
public ForkJoinTask submit(ForkJoinTask task) {
- doSubmit(task);
+ if (task == null)
+ throw new NullPointerException();
+ forkOrSubmit(task);
return task;
}
@@ -1337,8 +1560,10 @@ public class ForkJoinPool extends Abstra
* scheduled for execution
*/
public ForkJoinTask submit(Callable task) {
+ if (task == null)
+ throw new NullPointerException();
ForkJoinTask job = ForkJoinTask.adapt(task);
- doSubmit(job);
+ forkOrSubmit(job);
return job;
}
@@ -1348,8 +1573,10 @@ public class ForkJoinPool extends Abstra
* scheduled for execution
*/
public ForkJoinTask submit(Runnable task, T result) {
+ if (task == null)
+ throw new NullPointerException();
ForkJoinTask job = ForkJoinTask.adapt(task, result);
- doSubmit(job);
+ forkOrSubmit(job);
return job;
}
@@ -1359,12 +1586,14 @@ public class ForkJoinPool extends Abstra
* scheduled for execution
*/
public ForkJoinTask> submit(Runnable task) {
+ if (task == null)
+ throw new NullPointerException();
ForkJoinTask> job;
if (task instanceof ForkJoinTask>) // avoid re-wrap
job = (ForkJoinTask>) task;
else
job = ForkJoinTask.adapt(task, null);
- doSubmit(job);
+ forkOrSubmit(job);
return job;
}
@@ -1424,14 +1653,14 @@ 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
+ * yet terminated. The result returned by this method may differ
* 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 workerCounts >>> TOTAL_COUNT_SHIFT;
+ return parallelism + (short)(ctl >>> TC_SHIFT);
}
/**
@@ -1453,7 +1682,8 @@ public class ForkJoinPool extends Abstra
* @return the number of worker threads
*/
public int getRunningThreadCount() {
- return workerCounts & RUNNING_COUNT_MASK;
+ int r = parallelism + (int)(ctl >> AC_SHIFT);
+ return r <= 0? 0 : r; // suppress momentarily negative values
}
/**
@@ -1464,7 +1694,8 @@ public class ForkJoinPool extends Abstra
* @return the number of active threads
*/
public int getActiveThreadCount() {
- return runState & ACTIVE_COUNT_MASK;
+ int r = parallelism + (int)(ctl >> AC_SHIFT) + blockedCount;
+ return r <= 0? 0 : r; // suppress momentarily negative values
}
/**
@@ -1479,7 +1710,7 @@ public class ForkJoinPool extends Abstra
* @return {@code true} if all threads are currently idle
*/
public boolean isQuiescent() {
- return (runState & ACTIVE_COUNT_MASK) == 0;
+ return parallelism + (int)(ctl >> AC_SHIFT) + blockedCount == 0;
}
/**
@@ -1509,25 +1740,25 @@ public class ForkJoinPool extends Abstra
*/
public long getQueuedTaskCount() {
long count = 0;
- ForkJoinWorkerThread[] ws = workers;
- int n = ws.length;
- for (int i = 0; i < n; ++i) {
- ForkJoinWorkerThread w = ws[i];
- if (w != null)
- count += w.getQueueSize();
+ ForkJoinWorkerThread[] ws;
+ if ((short)(ctl >>> TC_SHIFT) > -parallelism &&
+ (ws = workers) != null) {
+ for (ForkJoinWorkerThread w : ws)
+ if (w != null)
+ count -= w.queueBase - w.queueTop; // must read base first
}
return count;
}
/**
* 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.
+ * pool that have not yet begun executing. This method may take
+ * time proportional to the number of submissions.
*
* @return the number of queued submissions
*/
public int getQueuedSubmissionCount() {
- return submissionQueue.size();
+ return -queueBase + queueTop;
}
/**
@@ -1537,7 +1768,7 @@ public class ForkJoinPool extends Abstra
* @return {@code true} if there are any queued submissions
*/
public boolean hasQueuedSubmissions() {
- return !submissionQueue.isEmpty();
+ return queueBase != queueTop;
}
/**
@@ -1548,7 +1779,19 @@ public class ForkJoinPool extends Abstra
* @return the next submission, or {@code null} if none
*/
protected ForkJoinTask> pollSubmission() {
- return submissionQueue.poll();
+ ForkJoinTask> t; ForkJoinTask>[] q; int b, i;
+ while ((b = queueBase) != queueTop &&
+ (q = submissionQueue) != null &&
+ (i = (q.length - 1) & b) >= 0) {
+ long u = (i << ASHIFT) + ABASE;
+ if ((t = q[i]) != null &&
+ queueBase == b &&
+ UNSAFE.compareAndSwapObject(q, u, t, null)) {
+ queueBase = b + 1;
+ return t;
+ }
+ }
+ return null;
}
/**
@@ -1569,13 +1812,20 @@ public class ForkJoinPool extends Abstra
* @return the number of elements transferred
*/
protected int drainTasksTo(Collection super ForkJoinTask>> c) {
- int count = submissionQueue.drainTo(c);
- ForkJoinWorkerThread[] ws = workers;
- int n = ws.length;
- for (int i = 0; i < n; ++i) {
- ForkJoinWorkerThread w = ws[i];
- if (w != null)
- count += w.drainTasksTo(c);
+ int count = 0;
+ while (queueBase != queueTop) {
+ ForkJoinTask> t = pollSubmission();
+ if (t != null) {
+ c.add(t);
+ ++count;
+ }
+ }
+ ForkJoinWorkerThread[] ws;
+ if ((short)(ctl >>> TC_SHIFT) > -parallelism &&
+ (ws = workers) != null) {
+ for (ForkJoinWorkerThread w : ws)
+ if (w != null)
+ count += w.drainTasksTo(c);
}
return count;
}
@@ -1591,14 +1841,20 @@ public class ForkJoinPool extends Abstra
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;
+ long c = ctl;
+ int tc = pc + (short)(c >>> TC_SHIFT);
+ int rc = pc + (int)(c >> AC_SHIFT);
+ if (rc < 0) // ignore transient negative
+ rc = 0;
+ int ac = rc + blockedCount;
+ String level;
+ if ((c & STOP_BIT) != 0)
+ level = (tc == 0)? "Terminated" : "Terminating";
+ else
+ level = shutdown? "Shutting down" : "Running";
return super.toString() +
- "[" + runLevelToString(rs) +
+ "[" + level +
", parallelism = " + pc +
", size = " + tc +
", active = " + ac +
@@ -1609,13 +1865,6 @@ public class ForkJoinPool extends Abstra
"]";
}
- private static String runLevelToString(int s) {
- return ((s & TERMINATED) != 0 ? "Terminated" :
- ((s & TERMINATING) != 0 ? "Terminating" :
- ((s & SHUTDOWN) != 0 ? "Shutting down" :
- "Running")));
- }
-
/**
* Initiates an orderly shutdown in which previously submitted
* tasks are executed, but no new tasks will be accepted.
@@ -1630,7 +1879,7 @@ public class ForkJoinPool extends Abstra
*/
public void shutdown() {
checkPermission();
- advanceRunLevel(SHUTDOWN);
+ shutdown = true;
tryTerminate(false);
}
@@ -1652,6 +1901,7 @@ public class ForkJoinPool extends Abstra
*/
public List shutdownNow() {
checkPermission();
+ shutdown = true;
tryTerminate(true);
return Collections.emptyList();
}
@@ -1662,7 +1912,9 @@ public class ForkJoinPool extends Abstra
* @return {@code true} if all tasks have completed following shut down
*/
public boolean isTerminated() {
- return runState >= TERMINATED;
+ long c = ctl;
+ return ((c & STOP_BIT) != 0L &&
+ (short)(c >>> TC_SHIFT) == -parallelism);
}
/**
@@ -1670,13 +1922,25 @@ public class ForkJoinPool extends Abstra
* 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.
+ * ignored or suppressed interruption, or are waiting for IO,
+ * causing this executor not to properly terminate. (See the
+ * advisory notes for class {@link ForkJoinTask} stating that
+ * tasks should not normally entail blocking operations. But if
+ * they do, they must abort them on interrupt.)
*
* @return {@code true} if terminating but not yet terminated
*/
public boolean isTerminating() {
- return (runState & (TERMINATING|TERMINATED)) == TERMINATING;
+ long c = ctl;
+ return ((c & STOP_BIT) != 0L &&
+ (short)(c >>> TC_SHIFT) != -parallelism);
+ }
+
+ /**
+ * Returns true if terminating or terminated. Used by ForkJoinWorkerThread.
+ */
+ final boolean isAtLeastTerminating() {
+ return (ctl & STOP_BIT) != 0L;
}
/**
@@ -1685,7 +1949,7 @@ public class ForkJoinPool extends Abstra
* @return {@code true} if this pool has been shut down
*/
public boolean isShutdown() {
- return runState >= SHUTDOWN;
+ return shutdown;
}
/**
@@ -1701,10 +1965,19 @@ public class ForkJoinPool extends Abstra
*/
public boolean awaitTermination(long timeout, TimeUnit unit)
throws InterruptedException {
+ long nanos = unit.toNanos(timeout);
+ final ReentrantLock lock = this.submissionLock;
+ lock.lock();
try {
- return termination.awaitAdvanceInterruptibly(0, timeout, unit) > 0;
- } catch(TimeoutException ex) {
- return false;
+ for (;;) {
+ if (isTerminated())
+ return true;
+ if (nanos <= 0)
+ return false;
+ nanos = termination.awaitNanos(nanos);
+ }
+ } finally {
+ lock.unlock();
}
}
@@ -1716,13 +1989,15 @@ public class ForkJoinPool extends Abstra
* {@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.
+ * before actually blocking). These actions are performed by any
+ * thread invoking {@link ForkJoinPool#managedBlock}. 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:
@@ -1750,11 +2025,11 @@ public class ForkJoinPool extends Abstra
* QueueTaker(BlockingQueue q) { this.queue = q; }
* public boolean block() throws InterruptedException {
* if (item == null)
- * item = queue.take
+ * item = queue.take();
* return true;
* }
* public boolean isReleasable() {
- * return item != null || (item = queue.poll) != null;
+ * return item != null || (item = queue.poll()) != null;
* }
* public E getItem() { // call after pool.managedBlock completes
* return item;
@@ -1824,30 +2099,47 @@ public class ForkJoinPool extends Abstra
}
// 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);
-
- private static long objectFieldOffset(String field, Class> klazz) {
+ private static final sun.misc.Unsafe UNSAFE;
+ private static final long ctlOffset;
+ private static final long stealCountOffset;
+ private static final long blockedCountOffset;
+ private static final long quiescerCountOffset;
+ private static final long scanGuardOffset;
+ private static final long nextWorkerNumberOffset;
+ private static final long ABASE;
+ private static final int ASHIFT;
+
+ static {
+ poolNumberGenerator = new AtomicInteger();
+ workerSeedGenerator = new Random();
+ modifyThreadPermission = new RuntimePermission("modifyThread");
+ defaultForkJoinWorkerThreadFactory =
+ new DefaultForkJoinWorkerThreadFactory();
+ int s;
try {
- return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
- } catch (NoSuchFieldException e) {
- // Convert Exception to corresponding Error
- NoSuchFieldError error = new NoSuchFieldError(field);
- error.initCause(e);
- throw error;
- }
+ UNSAFE = getUnsafe();
+ Class k = ForkJoinPool.class;
+ ctlOffset = UNSAFE.objectFieldOffset
+ (k.getDeclaredField("ctl"));
+ stealCountOffset = UNSAFE.objectFieldOffset
+ (k.getDeclaredField("stealCount"));
+ blockedCountOffset = UNSAFE.objectFieldOffset
+ (k.getDeclaredField("blockedCount"));
+ quiescerCountOffset = UNSAFE.objectFieldOffset
+ (k.getDeclaredField("quiescerCount"));
+ scanGuardOffset = UNSAFE.objectFieldOffset
+ (k.getDeclaredField("scanGuard"));
+ nextWorkerNumberOffset = UNSAFE.objectFieldOffset
+ (k.getDeclaredField("nextWorkerNumber"));
+ Class a = ForkJoinTask[].class;
+ ABASE = UNSAFE.arrayBaseOffset(a);
+ s = UNSAFE.arrayIndexScale(a);
+ } catch (Exception e) {
+ throw new Error(e);
+ }
+ if ((s & (s-1)) != 0)
+ throw new Error("data type scale not a power of two");
+ ASHIFT = 31 - Integer.numberOfLeadingZeros(s);
}
/**