ViewVC Help
View File | Revision Log | Show Annotations | Download File | Root Listing
root/jsr166/jsr166/src/jsr166y/ForkJoinWorkerThread.java
Revision: 1.3
Committed: Wed Jan 7 19:12:36 2009 UTC (15 years, 4 months ago) by dl
Branch: MAIN
Changes since 1.2: +15 -25 lines
Log Message:
More misc cleanup

File Contents

# User Rev Content
1 dl 1.1 /*
2     * Written by Doug Lea with assistance from members of JCP JSR-166
3     * Expert Group and released to the public domain, as explained at
4     * http://creativecommons.org/licenses/publicdomain
5     */
6    
7     package jsr166y;
8     import java.util.*;
9     import java.util.concurrent.*;
10     import java.util.concurrent.atomic.*;
11     import java.util.concurrent.locks.*;
12     import sun.misc.Unsafe;
13     import java.lang.reflect.*;
14    
15     /**
16 dl 1.2 * A thread managed by a {@link ForkJoinPool}. This class is
17     * subclassable solely for the sake of adding functionality -- there
18     * are no overridable methods dealing with scheduling or
19     * execution. However, you can override initialization and termination
20     * cleanup methods surrounding the main task processing loop. If you
21     * do create such a subclass, you will also need to supply a custom
22     * ForkJoinWorkerThreadFactory to use it in a ForkJoinPool.
23     *
24     * <p>This class also provides methods for generating per-thread
25     * random numbers, with the same properties as {@link
26     * java.util.Random} but with each generator isolated from those of
27     * other threads.
28 dl 1.1 */
29     public class ForkJoinWorkerThread extends Thread {
30     /*
31     * Algorithm overview:
32     *
33     * 1. Work-Stealing: Work-stealing queues are special forms of
34     * Deques that support only three of the four possible
35     * end-operations -- push, pop, and deq (aka steal), and only do
36     * so under the constraints that push and pop are called only from
37     * the owning thread, while deq may be called from other threads.
38     * (If you are unfamiliar with them, you probably want to read
39     * Herlihy and Shavit's book "The Art of Multiprocessor
40     * programming", chapter 16 describing these in more detail before
41     * proceeding.) The main work-stealing queue design is roughly
42     * similar to "Dynamic Circular Work-Stealing Deque" by David
43     * Chase and Yossi Lev, SPAA 2005
44     * (http://research.sun.com/scalable/pubs/index.html). The main
45     * difference ultimately stems from gc requirements that we null
46     * out taken slots as soon as we can, to maintain as small a
47     * footprint as possible even in programs generating huge numbers
48     * of tasks. To accomplish this, we shift the CAS arbitrating pop
49     * vs deq (steal) from being on the indices ("base" and "sp") to
50     * the slots themselves (mainly via method "casSlotNull()"). So,
51     * both a successful pop and deq mainly entail CAS'ing a nonnull
52     * slot to null. Because we rely on CASes of references, we do
53     * not need tag bits on base or sp. They are simple ints as used
54     * in any circular array-based queue (see for example ArrayDeque).
55     * Updates to the indices must still be ordered in a way that
56     * guarantees that (sp - base) > 0 means the queue is empty, but
57     * otherwise may err on the side of possibly making the queue
58     * appear nonempty when a push, pop, or deq have not fully
59     * committed. Note that this means that the deq operation,
60     * considered individually, is not wait-free. One thief cannot
61     * successfully continue until another in-progress one (or, if
62     * previously empty, a push) completes. However, in the
63     * aggregate, we ensure at least probablistic non-blockingness. If
64     * an attempted steal fails, a thief always chooses a different
65     * random victim target to try next. So, in order for one thief to
66     * progress, it suffices for any in-progress deq or new push on
67     * any empty queue to complete. One reason this works well here is
68     * that apparently-nonempty often means soon-to-be-stealable,
69     * which gives threads a chance to activate if necessary before
70     * stealing (see below).
71     *
72     * Efficient implementation of this approach currently relies on
73     * an uncomfortable amount of "Unsafe" mechanics. To maintain
74     * correct orderings, reads and writes of variable base require
75     * volatile ordering. Variable sp does not require volatile write
76     * but needs cheaper store-ordering on writes. Because they are
77     * protected by volatile base reads, reads of the queue array and
78     * its slots do not need volatile load semantics, but writes (in
79     * push) require store order and CASes (in pop and deq) require
80     * (volatile) CAS semantics. Since these combinations aren't
81     * supported using ordinary volatiles, the only way to accomplish
82     * these effciently is to use direct Unsafe calls. (Using external
83     * AtomicIntegers and AtomicReferenceArrays for the indices and
84     * array is significantly slower because of memory locality and
85     * indirection effects.) Further, performance on most platforms is
86     * very sensitive to placement and sizing of the (resizable) queue
87     * array. Even though these queues don't usually become all that
88     * big, the initial size must be large enough to counteract cache
89     * contention effects across multiple queues (especially in the
90     * presence of GC cardmarking). Also, to improve thread-locality,
91     * queues are currently initialized immediately after the thread
92     * gets the initial signal to start processing tasks. However,
93     * all queue-related methods except pushTask are written in a way
94     * that allows them to instead be lazily allocated and/or disposed
95     * of when empty. All together, these low-level implementation
96     * choices produce as much as a factor of 4 performance
97     * improvement compared to naive implementations, and enable the
98     * processing of billions of tasks per second, sometimes at the
99     * expense of ugliness.
100     *
101     * 2. Run control: The primary run control is based on a global
102     * counter (activeCount) held by the pool. It uses an algorithm
103     * similar to that in Herlihy and Shavit section 17.6 to cause
104     * threads to eventually block when all threads declare they are
105     * inactive. (See variable "scans".) For this to work, threads
106     * must be declared active when executing tasks, and before
107     * stealing a task. They must be inactive before blocking on the
108     * Pool Barrier (awaiting a new submission or other Pool
109     * event). In between, there is some free play which we take
110     * advantage of to avoid contention and rapid flickering of the
111     * global activeCount: If inactive, we activate only if a victim
112     * queue appears to be nonempty (see above). Similarly, a thread
113     * tries to inactivate only after a full scan of other threads.
114     * The net effect is that contention on activeCount is rarely a
115     * measurable performance issue. (There are also a few other cases
116     * where we scan for work rather than retry/block upon
117     * contention.)
118     *
119     * 3. Selection control. We maintain policy of always choosing to
120     * run local tasks rather than stealing, and always trying to
121     * steal tasks before trying to run a new submission. All steals
122     * are currently performed in randomly-chosen deq-order. It may be
123     * worthwhile to bias these with locality / anti-locality
124     * information, but doing this well probably requires more
125     * lower-level information from JVMs than currently provided.
126     */
127    
128     /**
129     * Capacity of work-stealing queue array upon initialization.
130     * Must be a power of two. Initial size must be at least 2, but is
131     * padded to minimize cache effects.
132     */
133     private static final int INITIAL_QUEUE_CAPACITY = 1 << 13;
134    
135     /**
136     * Maximum work-stealing queue array size. Must be less than or
137     * equal to 1 << 30 to ensure lack of index wraparound.
138     */
139     private static final int MAXIMUM_QUEUE_CAPACITY = 1 << 30;
140    
141     /**
142     * Generator of seeds for per-thread random numbers.
143     */
144     private static final Random randomSeedGenerator = new Random();
145    
146     /**
147     * The work-stealing queue array. Size must be a power of two.
148     */
149     private ForkJoinTask<?>[] queue;
150    
151     /**
152     * Index (mod queue.length) of next queue slot to push to or pop
153     * from. It is written only by owner thread, via ordered store.
154     * Both sp and base are allowed to wrap around on overflow, but
155     * (sp - base) still estimates size.
156     */
157     private volatile int sp;
158    
159     /**
160     * Index (mod queue.length) of least valid queue slot, which is
161     * always the next position to steal from if nonempty.
162     */
163     private volatile int base;
164    
165     /**
166     * The pool this thread works in.
167     */
168     final ForkJoinPool pool;
169    
170     /**
171     * Index of this worker in pool array. Set once by pool before
172     * running, and accessed directly by pool during cleanup etc
173     */
174     int poolIndex;
175    
176     /**
177     * Run state of this worker. Supports simple versions of the usual
178     * shutdown/shutdownNow control.
179     */
180     private volatile int runState;
181    
182     // Runstate values. Order matters
183     private static final int RUNNING = 0;
184     private static final int SHUTDOWN = 1;
185     private static final int TERMINATING = 2;
186     private static final int TERMINATED = 3;
187    
188     /**
189     * Activity status. When true, this worker is considered active.
190     * Must be false upon construction. It must be true when executing
191     * tasks, and BEFORE stealing a task. It must be false before
192     * blocking on the Pool Barrier.
193     */
194     private boolean active;
195    
196     /**
197     * Number of steals, transferred to pool when idle
198     */
199     private int stealCount;
200    
201     /**
202     * Seed for random number generator for choosing steal victims
203     */
204     private int randomVictimSeed;
205    
206     /**
207     * Seed for embedded Jurandom
208     */
209     private long juRandomSeed;
210    
211     /**
212     * The last barrier event waited for
213     */
214     private long eventCount;
215    
216     /**
217     * Creates a ForkJoinWorkerThread operating in the given pool.
218     * @param pool the pool this thread works in
219     * @throws NullPointerException if pool is null
220     */
221     protected ForkJoinWorkerThread(ForkJoinPool pool) {
222     if (pool == null) throw new NullPointerException();
223     this.pool = pool;
224     // remaining initialization deferred to onStart
225     }
226    
227 dl 1.2 // public access methods
228    
229     /**
230     * Returns the pool hosting the current task execution.
231     * @return the pool
232     */
233     public static ForkJoinPool getPool() {
234     return ((ForkJoinWorkerThread)(Thread.currentThread())).pool;
235     }
236    
237     /**
238     * Returns the index number of the current worker thread in its
239     * pool. The returned value ranges from zero to the maximum
240     * number of threads (minus one) that have ever been created in
241     * the pool. This method may be useful for applications that
242     * track status or collect results on a per-worker basis.
243     * @return the index number.
244     */
245     public static int getPoolIndex() {
246     return ((ForkJoinWorkerThread)(Thread.currentThread())).poolIndex;
247     }
248    
249 dl 1.1 // Access methods used by Pool
250    
251     /**
252     * Get and clear steal count for accumulation by pool. Called
253     * only when known to be idle (in pool.sync and termination).
254     */
255     final int getAndClearStealCount() {
256     int sc = stealCount;
257     stealCount = 0;
258     return sc;
259     }
260    
261     /**
262     * Returns estimate of the number of tasks in the queue, without
263     * correcting for transient negative values
264     */
265     final int getRawQueueSize() {
266     return sp - base;
267     }
268    
269     // Intrinsics-based support for queue operations.
270     // Currently these three (setSp, setSlot, casSlotNull) are
271     // usually manually inlined to improve performance
272    
273     /**
274     * Sets sp in store-order.
275     */
276     private void setSp(int s) {
277     _unsafe.putOrderedInt(this, spOffset, s);
278     }
279    
280     /**
281     * Add in store-order the given task at given slot of q to
282     * null. Caller must ensure q is nonnull and index is in range.
283     */
284     private static void setSlot(ForkJoinTask<?>[] q, int i,
285     ForkJoinTask<?> t){
286     _unsafe.putOrderedObject(q, (i << qShift) + qBase, t);
287     }
288    
289     /**
290     * CAS given slot of q to null. Caller must ensure q is nonnull
291     * and index is in range.
292     */
293     private static boolean casSlotNull(ForkJoinTask<?>[] q, int i,
294     ForkJoinTask<?> t) {
295     return _unsafe.compareAndSwapObject(q, (i << qShift) + qBase, t, null);
296     }
297    
298     // Main queue methods
299    
300     /**
301     * Pushes a task. Called only by current thread.
302     * @param t the task. Caller must ensure nonnull
303     */
304     final void pushTask(ForkJoinTask<?> t) {
305     ForkJoinTask<?>[] q = queue;
306     int mask = q.length - 1;
307     int s = sp;
308     _unsafe.putOrderedObject(q, ((s & mask) << qShift) + qBase, t);
309     _unsafe.putOrderedInt(this, spOffset, ++s);
310     if ((s -= base) == 1)
311     pool.signalNonEmptyWorkerQueue();
312     else if (s >= mask)
313     growQueue();
314     }
315    
316     /**
317     * Tries to take a task from the base of the queue, failing if
318     * either empty or contended.
319     * @return a task, or null if none or contended.
320     */
321     private ForkJoinTask<?> deqTask() {
322     ForkJoinTask<?>[] q;
323     ForkJoinTask<?> t;
324     int i;
325     int b;
326     if (sp != (b = base) &&
327     (q = queue) != null && // must read q after b
328     (t = q[i = (q.length - 1) & b]) != null &&
329     _unsafe.compareAndSwapObject(q, (i << qShift) + qBase, t, null)) {
330     base = b + 1;
331     return t;
332     }
333     return null;
334     }
335    
336     /**
337     * Returns a popped task, or null if empty. Called only by
338     * current thread.
339     */
340     final ForkJoinTask<?> popTask() {
341     ForkJoinTask<?> t;
342     int i;
343     ForkJoinTask<?>[] q = queue;
344     int mask = q.length - 1;
345     int s = sp;
346     if (s != base &&
347     (t = q[i = (s - 1) & mask]) != null &&
348     _unsafe.compareAndSwapObject(q, (i << qShift) + qBase, t, null)) {
349     _unsafe.putOrderedInt(this, spOffset, s - 1);
350     return t;
351     }
352     return null;
353     }
354    
355     /**
356     * Specialized version of popTask to pop only if
357     * topmost element is the given task. Called only
358     * by current thread.
359     * @param t the task. Caller must ensure nonnull
360     */
361     final boolean unpushTask(ForkJoinTask<?> t) {
362     ForkJoinTask<?>[] q = queue;
363     int mask = q.length - 1;
364     int s = sp - 1;
365     if (_unsafe.compareAndSwapObject(q, ((s & mask) << qShift) + qBase,
366     t, null)) {
367     _unsafe.putOrderedInt(this, spOffset, s);
368     return true;
369     }
370     return false;
371     }
372    
373     /**
374     * Returns next task to pop.
375     */
376 dl 1.2 final ForkJoinTask<?> peekTask() {
377 dl 1.1 ForkJoinTask<?>[] q = queue;
378     return q == null? null : q[(sp - 1) & (q.length - 1)];
379     }
380    
381     /**
382     * Doubles queue array size. Transfers elements by emulating
383     * steals (deqs) from old array and placing, oldest first, into
384     * new array.
385     */
386     private void growQueue() {
387     ForkJoinTask<?>[] oldQ = queue;
388     int oldSize = oldQ.length;
389     int newSize = oldSize << 1;
390     if (newSize > MAXIMUM_QUEUE_CAPACITY)
391     throw new RejectedExecutionException("Queue capacity exceeded");
392     ForkJoinTask<?>[] newQ = queue = new ForkJoinTask<?>[newSize];
393    
394     int b = base;
395     int bf = b + oldSize;
396     int oldMask = oldSize - 1;
397     int newMask = newSize - 1;
398     do {
399     int oldIndex = b & oldMask;
400     ForkJoinTask<?> t = oldQ[oldIndex];
401     if (t != null && !casSlotNull(oldQ, oldIndex, t))
402     t = null;
403     setSlot(newQ, b & newMask, t);
404     } while (++b != bf);
405     pool.signalIdleWorkers(false);
406     }
407    
408     // Runstate management
409    
410     final boolean isShutdown() { return runState >= SHUTDOWN; }
411     final boolean isTerminating() { return runState >= TERMINATING; }
412     final boolean isTerminated() { return runState == TERMINATED; }
413     final boolean shutdown() { return transitionRunStateTo(SHUTDOWN); }
414     final boolean shutdownNow() { return transitionRunStateTo(TERMINATING); }
415    
416     /**
417     * Transition to at least the given state. Return true if not
418     * already at least given state.
419     */
420     private boolean transitionRunStateTo(int state) {
421     for (;;) {
422     int s = runState;
423     if (s >= state)
424     return false;
425     if (_unsafe.compareAndSwapInt(this, runStateOffset, s, state))
426     return true;
427     }
428     }
429    
430     /**
431     * Ensure status is active and if necessary adjust pool active count
432     */
433     final void activate() {
434     if (!active) {
435     active = true;
436     pool.incrementActiveCount();
437     }
438     }
439    
440     /**
441     * Ensure status is inactive and if necessary adjust pool active count
442     */
443     final void inactivate() {
444     if (active) {
445     active = false;
446     pool.decrementActiveCount();
447     }
448     }
449    
450     // Lifecycle methods
451    
452     /**
453     * Initializes internal state after construction but before
454     * processing any tasks. If you override this method, you must
455     * invoke super.onStart() at the beginning of the method.
456     * Initialization requires care: Most fields must have legal
457     * default values, to ensure that attempted accesses from other
458     * threads work correctly even before this thread starts
459     * processing tasks.
460     */
461     protected void onStart() {
462     juRandomSeed = randomSeedGenerator.nextLong();
463     do;while((randomVictimSeed = nextRandomInt()) == 0); // must be nonzero
464     if (queue == null)
465     queue = new ForkJoinTask<?>[INITIAL_QUEUE_CAPACITY];
466    
467     // Heuristically allow one initial thread to warm up; others wait
468     if (poolIndex < pool.getParallelism() - 1) {
469     eventCount = pool.sync(this, 0);
470     activate();
471     }
472     }
473    
474     /**
475     * Perform cleanup associated with termination of this worker
476     * thread. If you override this method, you must invoke
477     * super.onTermination at the end of the overridden method.
478     *
479     * @param exception the exception causing this thread to abort due
480     * to an unrecoverable error, or null if completed normally.
481     */
482     protected void onTermination(Throwable exception) {
483     try {
484     clearLocalTasks();
485     inactivate();
486     cancelTasks();
487     } finally {
488     terminate(exception);
489     }
490     }
491    
492     /**
493     * Notify pool of termination and, if exception is nonnull,
494     * rethrow it to trigger this thread's uncaughtExceptionHandler
495     */
496     private void terminate(Throwable exception) {
497     transitionRunStateTo(TERMINATED);
498     try {
499     pool.workerTerminated(this);
500     } finally {
501     if (exception != null)
502     ForkJoinTask.rethrowException(exception);
503     }
504     }
505    
506     /**
507     * Run local tasks on exit from main.
508     */
509     private void clearLocalTasks() {
510     while (base != sp && !pool.isTerminating()) {
511     ForkJoinTask<?> t = popTask();
512     if (t != null) {
513     activate(); // ensure active status
514     t.quietlyExec();
515     }
516     }
517     }
518    
519     /**
520     * Removes and cancels all tasks in queue. Can be called from any
521     * thread.
522     */
523     final void cancelTasks() {
524     while (base != sp) {
525     ForkJoinTask<?> t = deqTask();
526     if (t != null)
527 dl 1.3 t.cancelIgnoringExceptions();
528 dl 1.1 }
529     }
530    
531     /**
532     * This method is required to be public, but should never be
533     * called explicitly. It performs the main run loop to execute
534     * ForkJoinTasks.
535     */
536     public void run() {
537     Throwable exception = null;
538     try {
539     onStart();
540     while (!isShutdown())
541     step();
542     } catch (Throwable ex) {
543     exception = ex;
544     } finally {
545     onTermination(exception);
546     }
547     }
548    
549     /**
550     * Main top-level action.
551     */
552     private void step() {
553     ForkJoinTask<?> t = sp != base? popTask() : null;
554     if (t != null || (t = scan(null, true)) != null) {
555     activate();
556     t.quietlyExec();
557     }
558     else {
559     inactivate();
560     eventCount = pool.sync(this, eventCount);
561     }
562     }
563    
564     // scanning for and stealing tasks
565    
566     /**
567     * Computes next value for random victim probe. Scans don't
568     * require a very high quality generator, but also not a crummy
569     * one. Marsaglia xor-shift is cheap and works well.
570     *
571     * This is currently unused, and manually inlined
572     */
573     private static int xorShift(int r) {
574     r ^= r << 1;
575     r ^= r >>> 3;
576     r ^= r << 10;
577     return r;
578     }
579    
580     /**
581     * Tries to steal a task from another worker and/or, if enabled,
582     * submission queue. Starts at a random index of workers array,
583     * and probes workers until finding one with non-empty queue or
584     * finding that all are empty. It randomly selects the first n-1
585     * probes. If these are empty, it resorts to full circular
586     * traversal, which is necessary to accurately set active status
587     * by caller. Also restarts if pool barrier has tripped since last
588     * scan, which forces refresh of workers array, in case barrier
589     * was associated with resize.
590     *
591     * This method must be both fast and quiet -- usually avoiding
592     * memory accesses that could disrupt cache sharing etc other than
593     * those needed to check for and take tasks. This accounts for,
594     * among other things, updating random seed in place without
595     * storing it until exit. (Note that we only need to store it if
596     * we found a task; otherwise it doesn't matter if we start at the
597     * same place next time.)
598     *
599     * @param joinMe if non null; exit early if done
600     * @param checkSubmissions true if OK to take submissions
601     * @return a task, or null if none found
602     */
603     private ForkJoinTask<?> scan(ForkJoinTask<?> joinMe,
604     boolean checkSubmissions) {
605     ForkJoinPool p = pool;
606     if (p == null) // Never null, but avoids
607     return null; // implicit nullchecks below
608     int r = randomVictimSeed; // extract once to keep scan quiet
609     restart: // outer loop refreshes ws array
610     while (joinMe == null || joinMe.status >= 0) {
611     int mask;
612     ForkJoinWorkerThread[] ws = p.workers;
613     if (ws != null && (mask = ws.length - 1) > 0) {
614     int probes = -mask; // use random index while negative
615     int idx = r;
616     for (;;) {
617     ForkJoinWorkerThread v;
618     // inlined xorshift to update seed
619     r ^= r << 1; r ^= r >>> 3; r ^= r << 10;
620     if ((v = ws[mask & idx]) != null && v.sp != v.base) {
621     ForkJoinTask<?> t;
622     activate();
623     if ((joinMe == null || joinMe.status >= 0) &&
624     (t = v.deqTask()) != null) {
625     randomVictimSeed = r;
626     ++stealCount;
627     return t;
628     }
629     continue restart; // restart on contention
630     }
631     if ((probes >> 1) <= mask) // n-1 random then circular
632     idx = (probes++ < 0)? r : (idx + 1);
633     else
634     break;
635     }
636     }
637     if (checkSubmissions && p.hasQueuedSubmissions()) {
638     activate();
639     ForkJoinTask<?> t = p.pollSubmission();
640     if (t != null)
641     return t;
642     }
643     else {
644     long ec = eventCount; // restart on pool event
645     if ((eventCount = p.getEventCount()) == ec)
646     break;
647     }
648     }
649     return null;
650     }
651    
652     /**
653     * Callback from pool.sync to rescan before blocking. If a
654     * task is found, it is pushed so it can be executed upon return.
655     * @return true if found and pushed a task
656     */
657     final boolean prescan() {
658     ForkJoinTask<?> t = scan(null, true);
659     if (t != null) {
660     pushTask(t);
661     return true;
662     }
663     else {
664     inactivate();
665     return false;
666     }
667     }
668    
669 dl 1.2 // Support for ForkJoinTask methods
670    
671 dl 1.1 /**
672 dl 1.3 * Scan, returning early if joinMe done
673 dl 1.1 */
674 dl 1.3 final ForkJoinTask<?> scanWhileJoining(ForkJoinTask<?> joinMe) {
675     ForkJoinTask<?> t = scan(joinMe, false);
676     if (t != null && joinMe.status < 0 && sp == base) {
677     pushTask(t); // unsteal if done and this task would be stealable
678     t = null;
679 dl 1.1 }
680 dl 1.3 return t;
681 dl 1.1 }
682 dl 1.3
683 dl 1.1 /**
684     * Pops or steals a task
685     * @return task, or null if none available
686     */
687 dl 1.3 final ForkJoinTask<?> pollLocalOrStolenTask() {
688     ForkJoinTask<?> t;
689     return (t = popTask()) == null? scan(null, false) : t;
690 dl 1.1 }
691    
692     /**
693     * Runs tasks until pool isQuiescent
694     */
695     final void helpQuiescePool() {
696     for (;;) {
697 dl 1.3 ForkJoinTask<?> t = pollLocalOrStolenTask();
698 dl 1.2 if (t != null) {
699     activate();
700     t.quietlyExec();
701     }
702     else {
703 dl 1.1 inactivate();
704     if (pool.isQuiescent()) {
705     activate(); // re-activate on exit
706     break;
707     }
708     }
709     }
710     }
711    
712     /**
713 dl 1.2 * Returns an estimate of the number of tasks in the queue.
714     */
715     final int getQueueSize() {
716 dl 1.3 int n = sp - base;
717 dl 1.2 return n <= 0? 0 : n; // suppress momentarily negative values
718     }
719    
720     /**
721 dl 1.1 * Returns an estimate of the number of tasks, offset by a
722     * function of number of idle workers.
723     */
724     final int getEstimatedSurplusTaskCount() {
725 dl 1.3 // The halving approximates weighting idle vs non-idle workers
726 dl 1.1 return (sp - base) - (pool.getIdleThreadCount() >>> 1);
727     }
728    
729     // Per-worker exported random numbers
730    
731     // Same constants as java.util.Random
732     final static long JURandomMultiplier = 0x5DEECE66DL;
733     final static long JURandomAddend = 0xBL;
734     final static long JURandomMask = (1L << 48) - 1;
735    
736     private final int nextJURandom(int bits) {
737     long next = (juRandomSeed * JURandomMultiplier + JURandomAddend) &
738     JURandomMask;
739     juRandomSeed = next;
740     return (int)(next >>> (48 - bits));
741     }
742    
743     private final int nextJURandomInt(int n) {
744     if (n <= 0)
745     throw new IllegalArgumentException("n must be positive");
746     int bits = nextJURandom(31);
747     if ((n & -n) == n)
748     return (int)((n * (long)bits) >> 31);
749    
750     for (;;) {
751     int val = bits % n;
752     if (bits - val + (n-1) >= 0)
753     return val;
754     bits = nextJURandom(31);
755     }
756     }
757    
758     private final long nextJURandomLong() {
759     return ((long)(nextJURandom(32)) << 32) + nextJURandom(32);
760     }
761    
762     private final long nextJURandomLong(long n) {
763     if (n <= 0)
764     throw new IllegalArgumentException("n must be positive");
765     long offset = 0;
766     while (n >= Integer.MAX_VALUE) { // randomly pick half range
767     int bits = nextJURandom(2); // 2nd bit for odd vs even split
768     long half = n >>> 1;
769     long nextn = ((bits & 2) == 0)? half : n - half;
770     if ((bits & 1) == 0)
771     offset += n - nextn;
772     n = nextn;
773     }
774     return offset + nextJURandomInt((int)n);
775     }
776    
777     private final double nextJURandomDouble() {
778     return (((long)(nextJURandom(26)) << 27) + nextJURandom(27))
779     / (double)(1L << 53);
780     }
781    
782     /**
783     * Returns a random integer using a per-worker random
784     * number generator with the same properties as
785     * {@link java.util.Random#nextInt}
786     * @return the next pseudorandom, uniformly distributed {@code int}
787     * value from this worker's random number generator's sequence
788     */
789     public static int nextRandomInt() {
790     return ((ForkJoinWorkerThread)(Thread.currentThread())).
791     nextJURandom(32);
792     }
793    
794     /**
795     * Returns a random integer using a per-worker random
796     * number generator with the same properties as
797     * {@link java.util.Random#nextInt(int)}
798     * @param n the bound on the random number to be returned. Must be
799     * positive.
800     * @return the next pseudorandom, uniformly distributed {@code int}
801     * value between {@code 0} (inclusive) and {@code n} (exclusive)
802     * from this worker's random number generator's sequence
803     * @throws IllegalArgumentException if n is not positive
804     */
805     public static int nextRandomInt(int n) {
806     return ((ForkJoinWorkerThread)(Thread.currentThread())).
807     nextJURandomInt(n);
808     }
809    
810     /**
811     * Returns a random long using a per-worker random
812     * number generator with the same properties as
813     * {@link java.util.Random#nextLong}
814     * @return the next pseudorandom, uniformly distributed {@code long}
815     * value from this worker's random number generator's sequence
816     */
817     public static long nextRandomLong() {
818     return ((ForkJoinWorkerThread)(Thread.currentThread())).
819     nextJURandomLong();
820     }
821    
822     /**
823     * Returns a random integer using a per-worker random
824     * number generator with the same properties as
825     * {@link java.util.Random#nextInt(int)}
826     * @param n the bound on the random number to be returned. Must be
827     * positive.
828     * @return the next pseudorandom, uniformly distributed {@code int}
829     * value between {@code 0} (inclusive) and {@code n} (exclusive)
830     * from this worker's random number generator's sequence
831     * @throws IllegalArgumentException if n is not positive
832     */
833     public static long nextRandomLong(long n) {
834     return ((ForkJoinWorkerThread)(Thread.currentThread())).
835     nextJURandomLong(n);
836     }
837    
838     /**
839     * Returns a random double using a per-worker random
840     * number generator with the same properties as
841     * {@link java.util.Random#nextDouble}
842     * @return the next pseudorandom, uniformly distributed {@code double}
843     * value between {@code 0.0} and {@code 1.0} from this
844     * worker's random number generator's sequence
845     */
846     public static double nextRandomDouble() {
847     return ((ForkJoinWorkerThread)(Thread.currentThread())).
848     nextJURandomDouble();
849     }
850    
851     // Temporary Unsafe mechanics for preliminary release
852    
853     static final Unsafe _unsafe;
854     static final long baseOffset;
855     static final long spOffset;
856     static final long qBase;
857     static final int qShift;
858     static final long runStateOffset;
859     static {
860     try {
861     if (ForkJoinWorkerThread.class.getClassLoader() != null) {
862     Field f = Unsafe.class.getDeclaredField("theUnsafe");
863     f.setAccessible(true);
864     _unsafe = (Unsafe)f.get(null);
865     }
866     else
867     _unsafe = Unsafe.getUnsafe();
868     baseOffset = _unsafe.objectFieldOffset
869     (ForkJoinWorkerThread.class.getDeclaredField("base"));
870     spOffset = _unsafe.objectFieldOffset
871     (ForkJoinWorkerThread.class.getDeclaredField("sp"));
872     runStateOffset = _unsafe.objectFieldOffset
873     (ForkJoinWorkerThread.class.getDeclaredField("runState"));
874     qBase = _unsafe.arrayBaseOffset(ForkJoinTask[].class);
875     int s = _unsafe.arrayIndexScale(ForkJoinTask[].class);
876     if ((s & (s-1)) != 0)
877     throw new Error("data type scale not a power of two");
878     qShift = 31 - Integer.numberOfLeadingZeros(s);
879     } catch (Exception e) {
880     throw new RuntimeException("Could not initialize intrinsics", e);
881     }
882     }
883     }