13 |
|
import java.util.Collection; |
14 |
|
import java.util.Collections; |
15 |
|
import java.util.List; |
16 |
– |
import java.util.concurrent.locks.Condition; |
16 |
|
import java.util.concurrent.locks.LockSupport; |
17 |
|
import java.util.concurrent.locks.ReentrantLock; |
18 |
|
import java.util.concurrent.atomic.AtomicInteger; |
19 |
< |
import java.util.concurrent.atomic.AtomicLong; |
19 |
> |
import java.util.concurrent.CountDownLatch; |
20 |
|
|
21 |
|
/** |
22 |
|
* An {@link ExecutorService} for running {@link ForkJoinTask}s. |
23 |
|
* A {@code ForkJoinPool} provides the entry point for submissions |
24 |
< |
* from non-{@code ForkJoinTask}s, as well as management and |
25 |
< |
* monitoring operations. |
24 |
> |
* from non-{@code ForkJoinTask} clients, as well as management and |
25 |
> |
* monitoring operations. |
26 |
|
* |
27 |
|
* <p>A {@code ForkJoinPool} differs from other kinds of {@link |
28 |
|
* ExecutorService} mainly by virtue of employing |
30 |
|
* execute subtasks created by other active tasks (eventually blocking |
31 |
|
* waiting for work if none exist). This enables efficient processing |
32 |
|
* when most tasks spawn other subtasks (as do most {@code |
33 |
< |
* ForkJoinTask}s). A {@code ForkJoinPool} may also be used for mixed |
34 |
< |
* execution of some plain {@code Runnable}- or {@code Callable}- |
35 |
< |
* based activities along with {@code ForkJoinTask}s. When setting |
37 |
< |
* {@linkplain #setAsyncMode async mode}, a {@code ForkJoinPool} may |
38 |
< |
* also be appropriate for use with fine-grained tasks of any form |
39 |
< |
* that are never joined. Otherwise, other {@code ExecutorService} |
40 |
< |
* implementations are typically more appropriate choices. |
33 |
> |
* ForkJoinTask}s). When setting <em>asyncMode</em> to true in |
34 |
> |
* constructors, {@code ForkJoinPool}s may also be appropriate for use |
35 |
> |
* with event-style tasks that are never joined. |
36 |
|
* |
37 |
|
* <p>A {@code ForkJoinPool} is constructed with a given target |
38 |
|
* parallelism level; by default, equal to the number of available |
39 |
< |
* processors. Unless configured otherwise via {@link |
40 |
< |
* #setMaintainsParallelism}, the pool attempts to maintain this |
41 |
< |
* number of active (or available) threads by dynamically adding, |
42 |
< |
* suspending, or resuming internal worker threads, even if some tasks |
43 |
< |
* are stalled waiting to join others. However, no such adjustments |
44 |
< |
* are performed in the face of blocked IO or other unmanaged |
45 |
< |
* synchronization. The nested {@link ManagedBlocker} interface |
51 |
< |
* enables extension of the kinds of synchronization accommodated. |
52 |
< |
* The target parallelism level may also be changed dynamically |
53 |
< |
* ({@link #setParallelism}). The total number of threads may be |
54 |
< |
* limited using method {@link #setMaximumPoolSize}, in which case it |
55 |
< |
* may become possible for the activities of a pool to stall due to |
56 |
< |
* the lack of available threads to process new tasks. |
39 |
> |
* processors. The pool attempts to maintain enough active (or |
40 |
> |
* available) threads by dynamically adding, suspending, or resuming |
41 |
> |
* internal worker threads, even if some tasks are stalled waiting to |
42 |
> |
* join others. However, no such adjustments are guaranteed in the |
43 |
> |
* face of blocked IO or other unmanaged synchronization. The nested |
44 |
> |
* {@link ManagedBlocker} interface enables extension of the kinds of |
45 |
> |
* synchronization accommodated. |
46 |
|
* |
47 |
|
* <p>In addition to execution and lifecycle control methods, this |
48 |
|
* class provides status check methods (for example |
51 |
|
* {@link #toString} returns indications of pool state in a |
52 |
|
* convenient form for informal monitoring. |
53 |
|
* |
54 |
+ |
* <p> As is the case with other ExecutorServices, there are three |
55 |
+ |
* main task execution methods summarized in the following |
56 |
+ |
* table. These are designed to be used by clients not already engaged |
57 |
+ |
* in fork/join computations in the current pool. The main forms of |
58 |
+ |
* these methods accept instances of {@code ForkJoinTask}, but |
59 |
+ |
* overloaded forms also allow mixed execution of plain {@code |
60 |
+ |
* Runnable}- or {@code Callable}- based activities as well. However, |
61 |
+ |
* tasks that are already executing in a pool should normally |
62 |
+ |
* <em>NOT</em> use these pool execution methods, but instead use the |
63 |
+ |
* within-computation forms listed in the table. |
64 |
+ |
* |
65 |
+ |
* <table BORDER CELLPADDING=3 CELLSPACING=1> |
66 |
+ |
* <tr> |
67 |
+ |
* <td></td> |
68 |
+ |
* <td ALIGN=CENTER> <b>Call from non-fork/join clients</b></td> |
69 |
+ |
* <td ALIGN=CENTER> <b>Call from within fork/join computations</b></td> |
70 |
+ |
* </tr> |
71 |
+ |
* <tr> |
72 |
+ |
* <td> <b>Arange async execution</td> |
73 |
+ |
* <td> {@link #execute(ForkJoinTask)}</td> |
74 |
+ |
* <td> {@link ForkJoinTask#fork}</td> |
75 |
+ |
* </tr> |
76 |
+ |
* <tr> |
77 |
+ |
* <td> <b>Await and obtain result</td> |
78 |
+ |
* <td> {@link #invoke(ForkJoinTask)}</td> |
79 |
+ |
* <td> {@link ForkJoinTask#invoke}</td> |
80 |
+ |
* </tr> |
81 |
+ |
* <tr> |
82 |
+ |
* <td> <b>Arrange exec and obtain Future</td> |
83 |
+ |
* <td> {@link #submit(ForkJoinTask)}</td> |
84 |
+ |
* <td> {@link ForkJoinTask#fork} (ForkJoinTasks <em>are</em> Futures)</td> |
85 |
+ |
* </tr> |
86 |
+ |
* </table> |
87 |
+ |
* |
88 |
|
* <p><b>Sample Usage.</b> Normally a single {@code ForkJoinPool} is |
89 |
|
* used for all parallel task execution in a program or subsystem. |
90 |
|
* Otherwise, use would not usually outweigh the construction and |
105 |
|
* |
106 |
|
* <p><b>Implementation notes</b>: This implementation restricts the |
107 |
|
* maximum number of running threads to 32767. Attempts to create |
108 |
< |
* pools with greater than the maximum result in |
108 |
> |
* pools with greater than the maximum number result in |
109 |
|
* {@code IllegalArgumentException}. |
110 |
|
* |
111 |
+ |
* <p>This implementation rejects submitted tasks (that is, by throwing |
112 |
+ |
* {@link RejectedExecutionException}) only when the pool is shut down |
113 |
+ |
* or internal resources have been exhausted. |
114 |
+ |
* |
115 |
|
* @since 1.7 |
116 |
|
* @author Doug Lea |
117 |
|
*/ |
118 |
|
public class ForkJoinPool extends AbstractExecutorService { |
119 |
|
|
120 |
|
/* |
121 |
< |
* See the extended comments interspersed below for design, |
122 |
< |
* rationale, and walkthroughs. |
121 |
> |
* Implementation Overview |
122 |
> |
* |
123 |
> |
* This class provides the central bookkeeping and control for a |
124 |
> |
* set of worker threads: Submissions from non-FJ threads enter |
125 |
> |
* into a submission queue. Workers take these tasks and typically |
126 |
> |
* split them into subtasks that may be stolen by other workers. |
127 |
> |
* The main work-stealing mechanics implemented in class |
128 |
> |
* ForkJoinWorkerThread give first priority to processing tasks |
129 |
> |
* from their own queues (LIFO or FIFO, depending on mode), then |
130 |
> |
* to randomized FIFO steals of tasks in other worker queues, and |
131 |
> |
* lastly to new submissions. These mechanics do not consider |
132 |
> |
* affinities, loads, cache localities, etc, so rarely provide the |
133 |
> |
* best possible performance on a given machine, but portably |
134 |
> |
* provide good throughput by averaging over these factors. |
135 |
> |
* (Further, even if we did try to use such information, we do not |
136 |
> |
* usually have a basis for exploiting it. For example, some sets |
137 |
> |
* of tasks profit from cache affinities, but others are harmed by |
138 |
> |
* cache pollution effects.) |
139 |
> |
* |
140 |
> |
* Beyond work-stealing support and essential bookkeeping, the |
141 |
> |
* main responsibility of this framework is to take actions when |
142 |
> |
* one worker is waiting to join a task stolen (or always held by) |
143 |
> |
* another. Becauae we are multiplexing many tasks on to a pool |
144 |
> |
* of workers, we can't just let them block (as in Thread.join). |
145 |
> |
* We also cannot just reassign the joiner's run-time stack with |
146 |
> |
* another and replace it later, which would be a form of |
147 |
> |
* "continuation", that even if possible is not necessarily a good |
148 |
> |
* idea. Given that the creation costs of most threads on most |
149 |
> |
* systems mainly surrounds setting up runtime stacks, thread |
150 |
> |
* creation and switching is usually not much more expensive than |
151 |
> |
* stack creation and switching, and is more flexible). Instead we |
152 |
> |
* combine two tactics: |
153 |
> |
* |
154 |
> |
* Helping: Arranging for the joiner to execute some task that it |
155 |
> |
* would be running if the steal had not occurred. Method |
156 |
> |
* ForkJoinWorkerThread.helpJoinTask tracks joining->stealing |
157 |
> |
* links to try to find such a task. |
158 |
> |
* |
159 |
> |
* Compensating: Unless there are already enough live threads, |
160 |
> |
* method helpMaintainParallelism() may create or or |
161 |
> |
* re-activate a spare thread to compensate for blocked |
162 |
> |
* joiners until they unblock. |
163 |
> |
* |
164 |
> |
* Because the determining existence of conservatively safe |
165 |
> |
* helping targets, the availability of already-created spares, |
166 |
> |
* and the apparent need to create new spares are all racy and |
167 |
> |
* require heuristic guidance, we rely on multiple retries of |
168 |
> |
* each. Further, because it is impossible to keep exactly the |
169 |
> |
* target (parallelism) number of threads running at any given |
170 |
> |
* time, we allow compensation during joins to fail, and enlist |
171 |
> |
* all other threads to help out whenever they are not otherwise |
172 |
> |
* occupied (i.e., mainly in method preStep). |
173 |
> |
* |
174 |
> |
* The ManagedBlocker extension API can't use helping so relies |
175 |
> |
* only on compensation in method awaitBlocker. |
176 |
> |
* |
177 |
> |
* The main throughput advantages of work-stealing stem from |
178 |
> |
* decentralized control -- workers mostly steal tasks from each |
179 |
> |
* other. We do not want to negate this by creating bottlenecks |
180 |
> |
* implementing other management responsibilities. So we use a |
181 |
> |
* collection of techniques that avoid, reduce, or cope well with |
182 |
> |
* contention. These entail several instances of bit-packing into |
183 |
> |
* CASable fields to maintain only the minimally required |
184 |
> |
* atomicity. To enable such packing, we restrict maximum |
185 |
> |
* parallelism to (1<<15)-1 (enabling twice this (to accommodate |
186 |
> |
* unbalanced increments and decrements) to fit into a 16 bit |
187 |
> |
* field, which is far in excess of normal operating range. Even |
188 |
> |
* though updates to some of these bookkeeping fields do sometimes |
189 |
> |
* contend with each other, they don't normally cache-contend with |
190 |
> |
* updates to others enough to warrant memory padding or |
191 |
> |
* isolation. So they are all held as fields of ForkJoinPool |
192 |
> |
* objects. The main capabilities are as follows: |
193 |
> |
* |
194 |
> |
* 1. Creating and removing workers. Workers are recorded in the |
195 |
> |
* "workers" array. This is an array as opposed to some other data |
196 |
> |
* structure to support index-based random steals by workers. |
197 |
> |
* Updates to the array recording new workers and unrecording |
198 |
> |
* terminated ones are protected from each other by a lock |
199 |
> |
* (workerLock) but the array is otherwise concurrently readable, |
200 |
> |
* and accessed directly by workers. To simplify index-based |
201 |
> |
* operations, the array size is always a power of two, and all |
202 |
> |
* readers must tolerate null slots. Currently, all worker thread |
203 |
> |
* creation is on-demand, triggered by task submissions, |
204 |
> |
* replacement of terminated workers, and/or compensation for |
205 |
> |
* blocked workers. However, all other support code is set up to |
206 |
> |
* work with other policies. |
207 |
> |
* |
208 |
> |
* To ensure that we do not hold on to worker references that |
209 |
> |
* would prevent GC, ALL accesses to workers are via indices into |
210 |
> |
* the workers array (which is one source of some of the unusual |
211 |
> |
* code constructions here). In essence, the workers array serves |
212 |
> |
* as a WeakReference mechanism. Thus for example the event queue |
213 |
> |
* stores worker indices, not worker references. Access to the |
214 |
> |
* workers in associated methods (for example releaseEventWaiters) |
215 |
> |
* must both index-check and null-check the IDs. All such accesses |
216 |
> |
* ignore bad IDs by returning out early from what they are doing, |
217 |
> |
* since this can only be associated with shutdown, in which case |
218 |
> |
* it is OK to give up. On termination, we just clobber these |
219 |
> |
* data structures without trying to use them. |
220 |
> |
* |
221 |
> |
* 2. Bookkeeping for dynamically adding and removing workers. We |
222 |
> |
* aim to approximately maintain the given level of parallelism. |
223 |
> |
* When some workers are known to be blocked (on joins or via |
224 |
> |
* ManagedBlocker), we may create or resume others to take their |
225 |
> |
* place until they unblock (see below). Implementing this |
226 |
> |
* requires counts of the number of "running" threads (i.e., those |
227 |
> |
* that are neither blocked nor artifically suspended) as well as |
228 |
> |
* the total number. These two values are packed into one field, |
229 |
> |
* "workerCounts" because we need accurate snapshots when deciding |
230 |
> |
* to create, resume or suspend. Note however that the |
231 |
> |
* correspondance of these counts to reality is not guaranteed. In |
232 |
> |
* particular updates for unblocked threads may lag until they |
233 |
> |
* actually wake up. |
234 |
> |
* |
235 |
> |
* 3. Maintaining global run state. The run state of the pool |
236 |
> |
* consists of a runLevel (SHUTDOWN, TERMINATING, etc) similar to |
237 |
> |
* those in other Executor implementations, as well as a count of |
238 |
> |
* "active" workers -- those that are, or soon will be, or |
239 |
> |
* recently were executing tasks. The runLevel and active count |
240 |
> |
* are packed together in order to correctly trigger shutdown and |
241 |
> |
* termination. Without care, active counts can be subject to very |
242 |
> |
* high contention. We substantially reduce this contention by |
243 |
> |
* relaxing update rules. A worker must claim active status |
244 |
> |
* prospectively, by activating if it sees that a submitted or |
245 |
> |
* stealable task exists (it may find after activating that the |
246 |
> |
* task no longer exists). It stays active while processing this |
247 |
> |
* task (if it exists) and any other local subtasks it produces, |
248 |
> |
* until it cannot find any other tasks. It then tries |
249 |
> |
* inactivating (see method preStep), but upon update contention |
250 |
> |
* instead scans for more tasks, later retrying inactivation if it |
251 |
> |
* doesn't find any. |
252 |
> |
* |
253 |
> |
* 4. Managing idle workers waiting for tasks. We cannot let |
254 |
> |
* workers spin indefinitely scanning for tasks when none are |
255 |
> |
* available. On the other hand, we must quickly prod them into |
256 |
> |
* action when new tasks are submitted or generated. We |
257 |
> |
* park/unpark these idle workers using an event-count scheme. |
258 |
> |
* Field eventCount is incremented upon events that may enable |
259 |
> |
* workers that previously could not find a task to now find one: |
260 |
> |
* Submission of a new task to the pool, or another worker pushing |
261 |
> |
* a task onto a previously empty queue. (We also use this |
262 |
> |
* mechanism for termination actions that require wakeups of idle |
263 |
> |
* workers). Each worker maintains its last known event count, |
264 |
> |
* and blocks when a scan for work did not find a task AND its |
265 |
> |
* lastEventCount matches the current eventCount. Waiting idle |
266 |
> |
* workers are recorded in a variant of Treiber stack headed by |
267 |
> |
* field eventWaiters which, when nonzero, encodes the thread |
268 |
> |
* index and count awaited for by the worker thread most recently |
269 |
> |
* calling eventSync. This thread in turn has a record (field |
270 |
> |
* nextEventWaiter) for the next waiting worker. In addition to |
271 |
> |
* allowing simpler decisions about need for wakeup, the event |
272 |
> |
* count bits in eventWaiters serve the role of tags to avoid ABA |
273 |
> |
* errors in Treiber stacks. To reduce delays in task diffusion, |
274 |
> |
* workers not otherwise occupied may invoke method |
275 |
> |
* releaseEventWaiters, that removes and signals (unparks) workers |
276 |
> |
* not waiting on current count. To reduce stalls, To minimize |
277 |
> |
* task production stalls associate with signalling, any worker |
278 |
> |
* pushing a task on an empty queue invokes the weaker method |
279 |
> |
* signalWork, that only releases idle workers until it detects |
280 |
> |
* interference by other threads trying to release, and lets them |
281 |
> |
* take over. The net effect is a tree-like diffusion of signals, |
282 |
> |
* where released threads (and possibly others) help with unparks. |
283 |
> |
* To further reduce contention effects a bit, failed CASes to |
284 |
> |
* increment field eventCount are tolerated without retries. |
285 |
> |
* Conceptually they are merged into the same event, which is OK |
286 |
> |
* when their only purpose is to enable workers to scan for work. |
287 |
> |
* |
288 |
> |
* 5. Managing suspension of extra workers. When a worker is about |
289 |
> |
* to block waiting for a join (or via ManagedBlockers), we may |
290 |
> |
* create a new thread to maintain parallelism level, or at least |
291 |
> |
* avoid starvation. Usually, extra threads are needed for only |
292 |
> |
* very short periods, yet join dependencies are such that we |
293 |
> |
* sometimes need them in bursts. Rather than create new threads |
294 |
> |
* each time this happens, we suspend no-longer-needed extra ones |
295 |
> |
* as "spares". For most purposes, we don't distinguish "extra" |
296 |
> |
* spare threads from normal "core" threads: On each call to |
297 |
> |
* preStep (the only point at which we can do this) a worker |
298 |
> |
* checks to see if there are now too many running workers, and if |
299 |
> |
* so, suspends itself. Method helpMaintainParallelism looks for |
300 |
> |
* suspended threads to resume before considering creating a new |
301 |
> |
* replacement. The spares themselves are encoded on another |
302 |
> |
* variant of a Treiber Stack, headed at field "spareWaiters". |
303 |
> |
* Note that the use of spares is intrinsically racy. One thread |
304 |
> |
* may become a spare at about the same time as another is |
305 |
> |
* needlessly being created. We counteract this and related slop |
306 |
> |
* in part by requiring resumed spares to immediately recheck (in |
307 |
> |
* preStep) to see whether they they should re-suspend. To avoid |
308 |
> |
* long-term build-up of spares, the oldest spare (see |
309 |
> |
* ForkJoinWorkerThread.suspendAsSpare) occasionally wakes up if |
310 |
> |
* not signalled and calls tryTrimSpare, which uses two different |
311 |
> |
* thresholds: Always killing if the number of spares is greater |
312 |
> |
* that 25% of total, and killing others only at a slower rate |
313 |
> |
* (UNUSED_SPARE_TRIM_RATE_NANOS). |
314 |
> |
* |
315 |
> |
* 6. Deciding when to create new workers. The main dynamic |
316 |
> |
* control in this class is deciding when to create extra threads |
317 |
> |
* in method helpMaintainParallelism. We would like to keep |
318 |
> |
* exactly #parallelism threads running, which is an impossble |
319 |
> |
* task. We always need to create one when the number of running |
320 |
> |
* threads would become zero and all workers are busy. Beyond |
321 |
> |
* this, we must rely on heuristics that work well in the the |
322 |
> |
* presence of transients phenomena such as GC stalls, dynamic |
323 |
> |
* compilation, and wake-up lags. These transients are extremely |
324 |
> |
* common -- we are normally trying to fully saturate the CPUs on |
325 |
> |
* a machine, so almost any activity other than running tasks |
326 |
> |
* impedes accuracy. Our main defense is to allow some slack in |
327 |
> |
* creation thresholds, using rules that reflect the fact that the |
328 |
> |
* more threads we have running, the more likely that we are |
329 |
> |
* underestimating the number running threads. The rules also |
330 |
> |
* better cope with the fact that some of the methods in this |
331 |
> |
* class tend to never become compiled (but are interpreted), so |
332 |
> |
* some components of the entire set of controls might execute 100 |
333 |
> |
* times faster than others. And similarly for cases where the |
334 |
> |
* apparent lack of work is just due to GC stalls and other |
335 |
> |
* transient system activity. |
336 |
> |
* |
337 |
> |
* Beware that there is a lot of representation-level coupling |
338 |
> |
* among classes ForkJoinPool, ForkJoinWorkerThread, and |
339 |
> |
* ForkJoinTask. For example, direct access to "workers" array by |
340 |
> |
* workers, and direct access to ForkJoinTask.status by both |
341 |
> |
* ForkJoinPool and ForkJoinWorkerThread. There is little point |
342 |
> |
* trying to reduce this, since any associated future changes in |
343 |
> |
* representations will need to be accompanied by algorithmic |
344 |
> |
* changes anyway. |
345 |
> |
* |
346 |
> |
* Style notes: There are lots of inline assignments (of form |
347 |
> |
* "while ((local = field) != 0)") which are usually the simplest |
348 |
> |
* way to ensure the required read orderings (which are sometimes |
349 |
> |
* critical). Also several occurrences of the unusual "do {} |
350 |
> |
* while(!cas...)" which is the simplest way to force an update of |
351 |
> |
* a CAS'ed variable. There are also other coding oddities that |
352 |
> |
* help some methods perform reasonably even when interpreted (not |
353 |
> |
* compiled), at the expense of some messy constructions that |
354 |
> |
* reduce byte code counts. |
355 |
> |
* |
356 |
> |
* The order of declarations in this file is: (1) statics (2) |
357 |
> |
* fields (along with constants used when unpacking some of them) |
358 |
> |
* (3) internal control methods (4) callbacks and other support |
359 |
> |
* for ForkJoinTask and ForkJoinWorkerThread classes, (5) exported |
360 |
> |
* methods (plus a few little helpers). |
361 |
|
*/ |
362 |
|
|
98 |
– |
/** Mask for packing and unpacking shorts */ |
99 |
– |
private static final int shortMask = 0xffff; |
100 |
– |
|
101 |
– |
/** Max pool size -- must be a power of two minus 1 */ |
102 |
– |
private static final int MAX_THREADS = 0x7FFF; |
103 |
– |
|
363 |
|
/** |
364 |
|
* Factory for creating new {@link ForkJoinWorkerThread}s. |
365 |
|
* A {@code ForkJoinWorkerThreadFactory} must be defined and used |
371 |
|
* Returns a new worker thread operating in the given pool. |
372 |
|
* |
373 |
|
* @param pool the pool this thread works in |
374 |
< |
* @throws NullPointerException if pool is null |
374 |
> |
* @throws NullPointerException if the pool is null |
375 |
|
*/ |
376 |
|
public ForkJoinWorkerThread newThread(ForkJoinPool pool); |
377 |
|
} |
380 |
|
* Default ForkJoinWorkerThreadFactory implementation; creates a |
381 |
|
* new ForkJoinWorkerThread. |
382 |
|
*/ |
383 |
< |
static class DefaultForkJoinWorkerThreadFactory |
383 |
> |
static class DefaultForkJoinWorkerThreadFactory |
384 |
|
implements ForkJoinWorkerThreadFactory { |
385 |
|
public ForkJoinWorkerThread newThread(ForkJoinPool pool) { |
386 |
< |
try { |
128 |
< |
return new ForkJoinWorkerThread(pool); |
129 |
< |
} catch (OutOfMemoryError oom) { |
130 |
< |
return null; |
131 |
< |
} |
386 |
> |
return new ForkJoinWorkerThread(pool); |
387 |
|
} |
388 |
|
} |
389 |
|
|
419 |
|
new AtomicInteger(); |
420 |
|
|
421 |
|
/** |
422 |
< |
* Array holding all worker threads in the pool. Initialized upon |
423 |
< |
* first use. Array size must be a power of two. Updates and |
424 |
< |
* replacements are protected by workerLock, but it is always kept |
425 |
< |
* in a consistent enough state to be randomly accessed without |
426 |
< |
* locking by workers performing work-stealing. |
422 |
> |
* Absolute bound for parallelism level. Twice this number plus |
423 |
> |
* one (i.e., 0xfff) must fit into a 16bit field to enable |
424 |
> |
* word-packing for some counts and indices. |
425 |
> |
*/ |
426 |
> |
private static final int MAX_WORKERS = 0x7fff; |
427 |
> |
|
428 |
> |
/** |
429 |
> |
* Array holding all worker threads in the pool. Array size must |
430 |
> |
* be a power of two. Updates and replacements are protected by |
431 |
> |
* workerLock, but the array is always kept in a consistent enough |
432 |
> |
* state to be randomly accessed without locking by workers |
433 |
> |
* performing work-stealing, as well as other traversal-based |
434 |
> |
* methods in this class. All readers must tolerate that some |
435 |
> |
* array slots may be null. |
436 |
|
*/ |
437 |
|
volatile ForkJoinWorkerThread[] workers; |
438 |
|
|
439 |
|
/** |
440 |
< |
* Lock protecting access to workers. |
440 |
> |
* Queue for external submissions. |
441 |
|
*/ |
442 |
< |
private final ReentrantLock workerLock; |
442 |
> |
private final LinkedTransferQueue<ForkJoinTask<?>> submissionQueue; |
443 |
|
|
444 |
|
/** |
445 |
< |
* Condition for awaitTermination. |
445 |
> |
* Lock protecting updates to workers array. |
446 |
|
*/ |
447 |
< |
private final Condition termination; |
447 |
> |
private final ReentrantLock workerLock; |
448 |
|
|
449 |
|
/** |
450 |
< |
* The uncaught exception handler used when any worker |
187 |
< |
* abruptly terminates |
450 |
> |
* Latch released upon termination. |
451 |
|
*/ |
452 |
< |
private Thread.UncaughtExceptionHandler ueh; |
452 |
> |
private final Phaser termination; |
453 |
|
|
454 |
|
/** |
455 |
|
* Creation factory for worker threads. |
457 |
|
private final ForkJoinWorkerThreadFactory factory; |
458 |
|
|
459 |
|
/** |
197 |
– |
* Head of stack of threads that were created to maintain |
198 |
– |
* parallelism when other threads blocked, but have since |
199 |
– |
* suspended when the parallelism level rose. |
200 |
– |
*/ |
201 |
– |
private volatile WaitQueueNode spareStack; |
202 |
– |
|
203 |
– |
/** |
460 |
|
* Sum of per-thread steal counts, updated only when threads are |
461 |
|
* idle or terminating. |
462 |
|
*/ |
463 |
< |
private final AtomicLong stealCount; |
463 |
> |
private volatile long stealCount; |
464 |
|
|
465 |
|
/** |
466 |
< |
* Queue for external submissions. |
466 |
> |
* The last nanoTime that a spare thread was trimmed |
467 |
|
*/ |
468 |
< |
private final LinkedTransferQueue<ForkJoinTask<?>> submissionQueue; |
468 |
> |
private volatile long trimTime; |
469 |
|
|
470 |
|
/** |
471 |
< |
* Head of Treiber stack for barrier sync. See below for explanation. |
471 |
> |
* The rate at which to trim unused spares |
472 |
|
*/ |
473 |
< |
private volatile WaitQueueNode syncStack; |
473 |
> |
static final long UNUSED_SPARE_TRIM_RATE_NANOS = |
474 |
> |
1000L * 1000L * 1000L; // 1 sec |
475 |
|
|
476 |
|
/** |
477 |
< |
* The count for event barrier |
477 |
> |
* Encoded record of top of treiber stack of threads waiting for |
478 |
> |
* events. The top 32 bits contain the count being waited for. The |
479 |
> |
* bottom 16 bits contains one plus the pool index of waiting |
480 |
> |
* worker thread. (Bits 16-31 are unused.) |
481 |
|
*/ |
482 |
< |
private volatile long eventCount; |
482 |
> |
private volatile long eventWaiters; |
483 |
> |
|
484 |
> |
private static final int EVENT_COUNT_SHIFT = 32; |
485 |
> |
private static final long WAITER_ID_MASK = (1L << 16) - 1L; |
486 |
|
|
487 |
|
/** |
488 |
< |
* Pool number, just for assigning useful names to worker threads |
488 |
> |
* A counter for events that may wake up worker threads: |
489 |
> |
* - Submission of a new task to the pool |
490 |
> |
* - A worker pushing a task on an empty queue |
491 |
> |
* - termination |
492 |
|
*/ |
493 |
< |
private final int poolNumber; |
493 |
> |
private volatile int eventCount; |
494 |
|
|
495 |
|
/** |
496 |
< |
* The maximum allowed pool size |
496 |
> |
* Encoded record of top of treiber stack of spare threads waiting |
497 |
> |
* for resumption. The top 16 bits contain an arbitrary count to |
498 |
> |
* avoid ABA effects. The bottom 16bits contains one plus the pool |
499 |
> |
* index of waiting worker thread. |
500 |
> |
*/ |
501 |
> |
private volatile int spareWaiters; |
502 |
> |
|
503 |
> |
private static final int SPARE_COUNT_SHIFT = 16; |
504 |
> |
private static final int SPARE_ID_MASK = (1 << 16) - 1; |
505 |
> |
|
506 |
> |
/** |
507 |
> |
* Lifecycle control. The low word contains the number of workers |
508 |
> |
* that are (probably) executing tasks. This value is atomically |
509 |
> |
* incremented before a worker gets a task to run, and decremented |
510 |
> |
* when worker has no tasks and cannot find any. Bits 16-18 |
511 |
> |
* contain runLevel value. When all are zero, the pool is |
512 |
> |
* running. Level transitions are monotonic (running -> shutdown |
513 |
> |
* -> terminating -> terminated) so each transition adds a bit. |
514 |
> |
* These are bundled together to ensure consistent read for |
515 |
> |
* termination checks (i.e., that runLevel is at least SHUTDOWN |
516 |
> |
* and active threads is zero). |
517 |
> |
*/ |
518 |
> |
private volatile int runState; |
519 |
> |
|
520 |
> |
// Note: The order among run level values matters. |
521 |
> |
private static final int RUNLEVEL_SHIFT = 16; |
522 |
> |
private static final int SHUTDOWN = 1 << RUNLEVEL_SHIFT; |
523 |
> |
private static final int TERMINATING = 1 << (RUNLEVEL_SHIFT + 1); |
524 |
> |
private static final int TERMINATED = 1 << (RUNLEVEL_SHIFT + 2); |
525 |
> |
private static final int ACTIVE_COUNT_MASK = (1 << RUNLEVEL_SHIFT) - 1; |
526 |
> |
private static final int ONE_ACTIVE = 1; // active update delta |
527 |
> |
|
528 |
> |
/** |
529 |
> |
* Holds number of total (i.e., created and not yet terminated) |
530 |
> |
* and running (i.e., not blocked on joins or other managed sync) |
531 |
> |
* threads, packed together to ensure consistent snapshot when |
532 |
> |
* making decisions about creating and suspending spare |
533 |
> |
* threads. Updated only by CAS. Note that adding a new worker |
534 |
> |
* requires incrementing both counts, since workers start off in |
535 |
> |
* running state. |
536 |
|
*/ |
537 |
< |
private volatile int maxPoolSize; |
537 |
> |
private volatile int workerCounts; |
538 |
> |
|
539 |
> |
private static final int TOTAL_COUNT_SHIFT = 16; |
540 |
> |
private static final int RUNNING_COUNT_MASK = (1 << TOTAL_COUNT_SHIFT) - 1; |
541 |
> |
private static final int ONE_RUNNING = 1; |
542 |
> |
private static final int ONE_TOTAL = 1 << TOTAL_COUNT_SHIFT; |
543 |
|
|
544 |
|
/** |
545 |
< |
* The desired parallelism level, updated only under workerLock. |
545 |
> |
* The target parallelism level. |
546 |
> |
* Accessed directly by ForkJoinWorkerThreads. |
547 |
|
*/ |
548 |
< |
private volatile int parallelism; |
548 |
> |
final int parallelism; |
549 |
|
|
550 |
|
/** |
551 |
|
* True if use local fifo, not default lifo, for local polling |
552 |
+ |
* Read by, and replicated by ForkJoinWorkerThreads |
553 |
|
*/ |
554 |
< |
private volatile boolean locallyFifo; |
554 |
> |
final boolean locallyFifo; |
555 |
|
|
556 |
|
/** |
557 |
< |
* Holds number of total (i.e., created and not yet terminated) |
558 |
< |
* and running (i.e., not blocked on joins or other managed sync) |
247 |
< |
* threads, packed into one int to ensure consistent snapshot when |
248 |
< |
* making decisions about creating and suspending spare |
249 |
< |
* threads. Updated only by CAS. Note: CASes in |
250 |
< |
* updateRunningCount and preJoin assume that running active count |
251 |
< |
* is in low word, so need to be modified if this changes. |
557 |
> |
* The uncaught exception handler used when any worker abruptly |
558 |
> |
* terminates. |
559 |
|
*/ |
560 |
< |
private volatile int workerCounts; |
560 |
> |
private final Thread.UncaughtExceptionHandler ueh; |
561 |
> |
|
562 |
> |
/** |
563 |
> |
* Pool number, just for assigning useful names to worker threads |
564 |
> |
*/ |
565 |
> |
private final int poolNumber; |
566 |
|
|
567 |
< |
private static int totalCountOf(int s) { return s >>> 16; } |
568 |
< |
private static int runningCountOf(int s) { return s & shortMask; } |
569 |
< |
private static int workerCountsFor(int t, int r) { return (t << 16) + r; } |
567 |
> |
|
568 |
> |
// Utilities for CASing fields. Note that several of these |
569 |
> |
// are manually inlined by callers |
570 |
|
|
571 |
|
/** |
572 |
< |
* Adds delta (which may be negative) to running count. This must |
261 |
< |
* be called before (with negative arg) and after (with positive) |
262 |
< |
* any managed synchronization (i.e., mainly, joins). |
263 |
< |
* |
264 |
< |
* @param delta the number to add |
572 |
> |
* Increments running count part of workerCounts |
573 |
|
*/ |
574 |
< |
final void updateRunningCount(int delta) { |
575 |
< |
int s; |
576 |
< |
do {} while (!casWorkerCounts(s = workerCounts, s + delta)); |
574 |
> |
final void incrementRunningCount() { |
575 |
> |
int c; |
576 |
> |
do {} while (!UNSAFE.compareAndSwapInt(this, workerCountsOffset, |
577 |
> |
c = workerCounts, |
578 |
> |
c + ONE_RUNNING)); |
579 |
|
} |
580 |
|
|
581 |
|
/** |
582 |
< |
* Adds delta (which may be negative) to both total and running |
273 |
< |
* count. This must be called upon creation and termination of |
274 |
< |
* worker threads. |
275 |
< |
* |
276 |
< |
* @param delta the number to add |
582 |
> |
* Tries to decrement running count unless already zero |
583 |
|
*/ |
584 |
< |
private void updateWorkerCount(int delta) { |
585 |
< |
int d = delta + (delta << 16); // add to both lo and hi parts |
586 |
< |
int s; |
587 |
< |
do {} while (!casWorkerCounts(s = workerCounts, s + d)); |
584 |
> |
final boolean tryDecrementRunningCount() { |
585 |
> |
int wc = workerCounts; |
586 |
> |
if ((wc & RUNNING_COUNT_MASK) == 0) |
587 |
> |
return false; |
588 |
> |
return UNSAFE.compareAndSwapInt(this, workerCountsOffset, |
589 |
> |
wc, wc - ONE_RUNNING); |
590 |
|
} |
591 |
|
|
592 |
|
/** |
593 |
< |
* Lifecycle control. High word contains runState, low word |
594 |
< |
* contains the number of workers that are (probably) executing |
595 |
< |
* tasks. This value is atomically incremented before a worker |
596 |
< |
* gets a task to run, and decremented when worker has no tasks |
597 |
< |
* and cannot find any. These two fields are bundled together to |
290 |
< |
* support correct termination triggering. Note: activeCount |
291 |
< |
* CAS'es cheat by assuming active count is in low word, so need |
292 |
< |
* to be modified if this changes |
593 |
> |
* Forces decrement of encoded workerCounts, awaiting nonzero if |
594 |
> |
* (rarely) necessary when other count updates lag. |
595 |
> |
* |
596 |
> |
* @param dr -- either zero or ONE_RUNNING |
597 |
> |
* @param dt == either zero or ONE_TOTAL |
598 |
|
*/ |
599 |
< |
private volatile int runControl; |
600 |
< |
|
601 |
< |
// RunState values. Order among values matters |
602 |
< |
private static final int RUNNING = 0; |
603 |
< |
private static final int SHUTDOWN = 1; |
604 |
< |
private static final int TERMINATING = 2; |
605 |
< |
private static final int TERMINATED = 3; |
599 |
> |
private void decrementWorkerCounts(int dr, int dt) { |
600 |
> |
for (;;) { |
601 |
> |
int wc = workerCounts; |
602 |
> |
if (wc == 0 && (runState & TERMINATED) != 0) |
603 |
> |
return; // lagging termination on a backout |
604 |
> |
if ((wc & RUNNING_COUNT_MASK) - dr < 0 || |
605 |
> |
(wc >>> TOTAL_COUNT_SHIFT) - dt < 0) |
606 |
> |
Thread.yield(); |
607 |
> |
if (UNSAFE.compareAndSwapInt(this, workerCountsOffset, |
608 |
> |
wc, wc - (dr + dt))) |
609 |
> |
return; |
610 |
> |
} |
611 |
> |
} |
612 |
|
|
613 |
< |
private static int runStateOf(int c) { return c >>> 16; } |
614 |
< |
private static int activeCountOf(int c) { return c & shortMask; } |
615 |
< |
private static int runControlFor(int r, int a) { return (r << 16) + a; } |
613 |
> |
/** |
614 |
> |
* Increments event count |
615 |
> |
*/ |
616 |
> |
private void advanceEventCount() { |
617 |
> |
int c; |
618 |
> |
do {} while(!UNSAFE.compareAndSwapInt(this, eventCountOffset, |
619 |
> |
c = eventCount, c+1)); |
620 |
> |
} |
621 |
|
|
622 |
|
/** |
623 |
|
* Tries incrementing active count; fails on contention. |
624 |
< |
* Called by workers before/during executing tasks. |
624 |
> |
* Called by workers before executing tasks. |
625 |
|
* |
626 |
|
* @return true on success |
627 |
|
*/ |
628 |
|
final boolean tryIncrementActiveCount() { |
629 |
< |
int c = runControl; |
630 |
< |
return casRunControl(c, c+1); |
629 |
> |
int c; |
630 |
> |
return UNSAFE.compareAndSwapInt(this, runStateOffset, |
631 |
> |
c = runState, c + ONE_ACTIVE); |
632 |
|
} |
633 |
|
|
634 |
|
/** |
635 |
|
* Tries decrementing active count; fails on contention. |
636 |
< |
* Possibly triggers termination on success. |
320 |
< |
* Called by workers when they can't find tasks. |
321 |
< |
* |
322 |
< |
* @return true on success |
636 |
> |
* Called when workers cannot find tasks to run. |
637 |
|
*/ |
638 |
|
final boolean tryDecrementActiveCount() { |
639 |
< |
int c = runControl; |
640 |
< |
int nextc = c - 1; |
641 |
< |
if (!casRunControl(c, nextc)) |
328 |
< |
return false; |
329 |
< |
if (canTerminateOnShutdown(nextc)) |
330 |
< |
terminateOnShutdown(); |
331 |
< |
return true; |
639 |
> |
int c; |
640 |
> |
return UNSAFE.compareAndSwapInt(this, runStateOffset, |
641 |
> |
c = runState, c - ONE_ACTIVE); |
642 |
|
} |
643 |
|
|
644 |
|
/** |
645 |
< |
* Returns {@code true} if argument represents zero active count |
646 |
< |
* and nonzero runstate, which is the triggering condition for |
337 |
< |
* terminating on shutdown. |
645 |
> |
* Advances to at least the given level. Returns true if not |
646 |
> |
* already in at least the given level. |
647 |
|
*/ |
648 |
< |
private static boolean canTerminateOnShutdown(int c) { |
649 |
< |
// i.e. least bit is nonzero runState bit |
650 |
< |
return ((c & -c) >>> 16) != 0; |
648 |
> |
private boolean advanceRunLevel(int level) { |
649 |
> |
for (;;) { |
650 |
> |
int s = runState; |
651 |
> |
if ((s & level) != 0) |
652 |
> |
return false; |
653 |
> |
if (UNSAFE.compareAndSwapInt(this, runStateOffset, s, s | level)) |
654 |
> |
return true; |
655 |
> |
} |
656 |
|
} |
657 |
|
|
658 |
+ |
// workers array maintenance |
659 |
+ |
|
660 |
|
/** |
661 |
< |
* Transition run state to at least the given state. Return true |
346 |
< |
* if not already at least given state. |
661 |
> |
* Records and returns a workers array index for new worker. |
662 |
|
*/ |
663 |
< |
private boolean transitionRunStateTo(int state) { |
664 |
< |
for (;;) { |
665 |
< |
int c = runControl; |
666 |
< |
if (runStateOf(c) >= state) |
667 |
< |
return false; |
668 |
< |
if (casRunControl(c, runControlFor(state, activeCountOf(c)))) |
669 |
< |
return true; |
663 |
> |
private int recordWorker(ForkJoinWorkerThread w) { |
664 |
> |
// Try using slot totalCount-1. If not available, scan and/or resize |
665 |
> |
int k = (workerCounts >>> TOTAL_COUNT_SHIFT) - 1; |
666 |
> |
final ReentrantLock lock = this.workerLock; |
667 |
> |
lock.lock(); |
668 |
> |
try { |
669 |
> |
ForkJoinWorkerThread[] ws = workers; |
670 |
> |
int n = ws.length; |
671 |
> |
if (k < 0 || k >= n || ws[k] != null) { |
672 |
> |
for (k = 0; k < n && ws[k] != null; ++k) |
673 |
> |
; |
674 |
> |
if (k == n) |
675 |
> |
ws = Arrays.copyOf(ws, n << 1); |
676 |
> |
} |
677 |
> |
ws[k] = w; |
678 |
> |
workers = ws; // volatile array write ensures slot visibility |
679 |
> |
} finally { |
680 |
> |
lock.unlock(); |
681 |
|
} |
682 |
+ |
return k; |
683 |
|
} |
684 |
|
|
685 |
|
/** |
686 |
< |
* Controls whether to add spares to maintain parallelism |
686 |
> |
* Nulls out record of worker in workers array |
687 |
|
*/ |
688 |
< |
private volatile boolean maintainsParallelism; |
688 |
> |
private void forgetWorker(ForkJoinWorkerThread w) { |
689 |
> |
int idx = w.poolIndex; |
690 |
> |
// Locking helps method recordWorker avoid unecessary expansion |
691 |
> |
final ReentrantLock lock = this.workerLock; |
692 |
> |
lock.lock(); |
693 |
> |
try { |
694 |
> |
ForkJoinWorkerThread[] ws = workers; |
695 |
> |
if (idx >= 0 && idx < ws.length && ws[idx] == w) // verify |
696 |
> |
ws[idx] = null; |
697 |
> |
} finally { |
698 |
> |
lock.unlock(); |
699 |
> |
} |
700 |
> |
} |
701 |
|
|
702 |
< |
// Constructors |
702 |
> |
// adding and removing workers |
703 |
|
|
704 |
|
/** |
705 |
< |
* Creates a {@code ForkJoinPool} with parallelism equal to {@link |
706 |
< |
* java.lang.Runtime#availableProcessors}, and using the {@linkplain |
707 |
< |
* #defaultForkJoinWorkerThreadFactory default thread factory}. |
369 |
< |
* |
370 |
< |
* @throws SecurityException if a security manager exists and |
371 |
< |
* the caller is not permitted to modify threads |
372 |
< |
* because it does not hold {@link |
373 |
< |
* java.lang.RuntimePermission}{@code ("modifyThread")} |
705 |
> |
* Tries to create and add new worker. Assumes that worker counts |
706 |
> |
* are already updated to accommodate the worker, so adjusts on |
707 |
> |
* failure. |
708 |
|
*/ |
709 |
< |
public ForkJoinPool() { |
710 |
< |
this(Runtime.getRuntime().availableProcessors(), |
711 |
< |
defaultForkJoinWorkerThreadFactory); |
709 |
> |
private void addWorker() { |
710 |
> |
ForkJoinWorkerThread w = null; |
711 |
> |
try { |
712 |
> |
w = factory.newThread(this); |
713 |
> |
} finally { // Adjust on either null or exceptional factory return |
714 |
> |
if (w == null) { |
715 |
> |
decrementWorkerCounts(ONE_RUNNING, ONE_TOTAL); |
716 |
> |
tryTerminate(false); // in case of failure during shutdown |
717 |
> |
} |
718 |
> |
} |
719 |
> |
if (w != null) |
720 |
> |
w.start(recordWorker(w), ueh); |
721 |
|
} |
722 |
|
|
723 |
|
/** |
724 |
< |
* Creates a {@code ForkJoinPool} with the indicated parallelism |
725 |
< |
* level and using the {@linkplain |
726 |
< |
* #defaultForkJoinWorkerThreadFactory default thread factory}. |
724 |
> |
* Final callback from terminating worker. Removes record of |
725 |
> |
* worker from array, and adjusts counts. If pool is shutting |
726 |
> |
* down, tries to complete terminatation. |
727 |
|
* |
728 |
< |
* @param parallelism the parallelism level |
386 |
< |
* @throws IllegalArgumentException if parallelism less than or |
387 |
< |
* equal to zero |
388 |
< |
* @throws SecurityException if a security manager exists and |
389 |
< |
* the caller is not permitted to modify threads |
390 |
< |
* because it does not hold {@link |
391 |
< |
* java.lang.RuntimePermission}{@code ("modifyThread")} |
728 |
> |
* @param w the worker |
729 |
|
*/ |
730 |
< |
public ForkJoinPool(int parallelism) { |
731 |
< |
this(parallelism, defaultForkJoinWorkerThreadFactory); |
730 |
> |
final void workerTerminated(ForkJoinWorkerThread w) { |
731 |
> |
forgetWorker(w); |
732 |
> |
decrementWorkerCounts(w.isTrimmed()? 0 : ONE_RUNNING, ONE_TOTAL); |
733 |
> |
while (w.stealCount != 0) // collect final count |
734 |
> |
tryAccumulateStealCount(w); |
735 |
> |
tryTerminate(false); |
736 |
|
} |
737 |
|
|
738 |
+ |
// Waiting for and signalling events |
739 |
+ |
|
740 |
|
/** |
741 |
< |
* Creates a {@code ForkJoinPool} with parallelism equal to {@link |
742 |
< |
* java.lang.Runtime#availableProcessors}, and using the given |
743 |
< |
* thread factory. |
741 |
> |
* Releases workers blocked on a count not equal to current count. |
742 |
> |
* Normally called after precheck that eventWaiters isn't zero to |
743 |
> |
* avoid wasted array checks. |
744 |
|
* |
745 |
< |
* @param factory the factory for creating new threads |
746 |
< |
* @throws NullPointerException if factory is null |
747 |
< |
* @throws SecurityException if a security manager exists and |
405 |
< |
* the caller is not permitted to modify threads |
406 |
< |
* because it does not hold {@link |
407 |
< |
* java.lang.RuntimePermission}{@code ("modifyThread")} |
745 |
> |
* @param signalling true if caller is a signalling worker so can |
746 |
> |
* exit upon (conservatively) detected contention by other threads |
747 |
> |
* who will continue to release |
748 |
|
*/ |
749 |
< |
public ForkJoinPool(ForkJoinWorkerThreadFactory factory) { |
750 |
< |
this(Runtime.getRuntime().availableProcessors(), factory); |
749 |
> |
private void releaseEventWaiters(boolean signalling) { |
750 |
> |
ForkJoinWorkerThread[] ws = workers; |
751 |
> |
int n = ws.length; |
752 |
> |
long h; // head of stack |
753 |
> |
ForkJoinWorkerThread w; int id, ec; |
754 |
> |
while ((id = ((int)((h = eventWaiters) & WAITER_ID_MASK)) - 1) >= 0 && |
755 |
> |
(int)(h >>> EVENT_COUNT_SHIFT) != (ec = eventCount) && |
756 |
> |
id < n && (w = ws[id]) != null) { |
757 |
> |
if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset, |
758 |
> |
h, h = w.nextWaiter)) |
759 |
> |
LockSupport.unpark(w); |
760 |
> |
if (signalling && (eventCount != ec || eventWaiters != h)) |
761 |
> |
break; |
762 |
> |
} |
763 |
|
} |
764 |
|
|
765 |
|
/** |
766 |
< |
* Creates a {@code ForkJoinPool} with the given parallelism and |
767 |
< |
* thread factory. |
416 |
< |
* |
417 |
< |
* @param parallelism the parallelism level |
418 |
< |
* @param factory the factory for creating new threads |
419 |
< |
* @throws IllegalArgumentException if parallelism less than or |
420 |
< |
* equal to zero, or greater than implementation limit |
421 |
< |
* @throws NullPointerException if factory is null |
422 |
< |
* @throws SecurityException if a security manager exists and |
423 |
< |
* the caller is not permitted to modify threads |
424 |
< |
* because it does not hold {@link |
425 |
< |
* java.lang.RuntimePermission}{@code ("modifyThread")} |
766 |
> |
* Tries to advance eventCount and releases waiters. Called only |
767 |
> |
* from workers. |
768 |
|
*/ |
769 |
< |
public ForkJoinPool(int parallelism, ForkJoinWorkerThreadFactory factory) { |
770 |
< |
if (parallelism <= 0 || parallelism > MAX_THREADS) |
771 |
< |
throw new IllegalArgumentException(); |
772 |
< |
if (factory == null) |
773 |
< |
throw new NullPointerException(); |
432 |
< |
checkPermission(); |
433 |
< |
this.factory = factory; |
434 |
< |
this.parallelism = parallelism; |
435 |
< |
this.maxPoolSize = MAX_THREADS; |
436 |
< |
this.maintainsParallelism = true; |
437 |
< |
this.poolNumber = poolNumberGenerator.incrementAndGet(); |
438 |
< |
this.workerLock = new ReentrantLock(); |
439 |
< |
this.termination = workerLock.newCondition(); |
440 |
< |
this.stealCount = new AtomicLong(); |
441 |
< |
this.submissionQueue = new LinkedTransferQueue<ForkJoinTask<?>>(); |
442 |
< |
// worker array and workers are lazily constructed |
769 |
> |
final void signalWork() { |
770 |
> |
int c; // try to increment event count -- CAS failure OK |
771 |
> |
UNSAFE.compareAndSwapInt(this, eventCountOffset, c = eventCount, c+1); |
772 |
> |
if (eventWaiters != 0L) |
773 |
> |
releaseEventWaiters(true); |
774 |
|
} |
775 |
|
|
776 |
|
/** |
777 |
< |
* Creates a new worker thread using factory. |
777 |
> |
* Blocks worker until terminating or event count |
778 |
> |
* advances from last value held by worker |
779 |
|
* |
780 |
< |
* @param index the index to assign worker |
449 |
< |
* @return new worker, or null if factory failed |
780 |
> |
* @param w the calling worker thread |
781 |
|
*/ |
782 |
< |
private ForkJoinWorkerThread createWorker(int index) { |
783 |
< |
Thread.UncaughtExceptionHandler h = ueh; |
784 |
< |
ForkJoinWorkerThread w = factory.newThread(this); |
785 |
< |
if (w != null) { |
786 |
< |
w.poolIndex = index; |
787 |
< |
w.setDaemon(true); |
788 |
< |
w.setAsyncMode(locallyFifo); |
789 |
< |
w.setName("ForkJoinPool-" + poolNumber + "-worker-" + index); |
790 |
< |
if (h != null) |
791 |
< |
w.setUncaughtExceptionHandler(h); |
782 |
> |
private void eventSync(ForkJoinWorkerThread w) { |
783 |
> |
int wec = w.lastEventCount; |
784 |
> |
long nh = (((long)wec) << EVENT_COUNT_SHIFT) | ((long)(w.poolIndex+1)); |
785 |
> |
long h; |
786 |
> |
while ((runState < SHUTDOWN || !tryTerminate(false)) && |
787 |
> |
((h = eventWaiters) == 0L || |
788 |
> |
(int)(h >>> EVENT_COUNT_SHIFT) == wec) && |
789 |
> |
eventCount == wec) { |
790 |
> |
if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset, |
791 |
> |
w.nextWaiter = h, nh)) { |
792 |
> |
while (runState < TERMINATING && eventCount == wec) { |
793 |
> |
if (!tryAccumulateStealCount(w)) // transfer while idle |
794 |
> |
continue; |
795 |
> |
Thread.interrupted(); // clear/ignore interrupt |
796 |
> |
if (eventCount != wec) |
797 |
> |
break; |
798 |
> |
LockSupport.park(w); |
799 |
> |
} |
800 |
> |
break; |
801 |
> |
} |
802 |
|
} |
803 |
< |
return w; |
803 |
> |
w.lastEventCount = eventCount; |
804 |
|
} |
805 |
|
|
806 |
+ |
// Maintaining spares |
807 |
+ |
|
808 |
|
/** |
809 |
< |
* Returns a good size for worker array given pool size. |
467 |
< |
* Currently requires size to be a power of two. |
809 |
> |
* Pushes worker onto the spare stack |
810 |
|
*/ |
811 |
< |
private static int arraySizeFor(int poolSize) { |
812 |
< |
if (poolSize <= 1) |
813 |
< |
return 1; |
814 |
< |
// See Hackers Delight, sec 3.2 |
473 |
< |
int c = poolSize >= MAX_THREADS ? MAX_THREADS : (poolSize - 1); |
474 |
< |
c |= c >>> 1; |
475 |
< |
c |= c >>> 2; |
476 |
< |
c |= c >>> 4; |
477 |
< |
c |= c >>> 8; |
478 |
< |
c |= c >>> 16; |
479 |
< |
return c + 1; |
811 |
> |
final void pushSpare(ForkJoinWorkerThread w) { |
812 |
> |
int ns = (++w.spareCount << SPARE_COUNT_SHIFT) | (w.poolIndex+1); |
813 |
> |
do {} while (!UNSAFE.compareAndSwapInt(this, spareWaitersOffset, |
814 |
> |
w.nextSpare = spareWaiters,ns)); |
815 |
|
} |
816 |
|
|
817 |
|
/** |
818 |
< |
* Creates or resizes array if necessary to hold newLength. |
819 |
< |
* Call only under exclusion. |
485 |
< |
* |
486 |
< |
* @return the array |
818 |
> |
* Tries (once) to resume a spare if running count is less than |
819 |
> |
* target parallelism. Fails on contention or stale workers. |
820 |
|
*/ |
821 |
< |
private ForkJoinWorkerThread[] ensureWorkerArrayCapacity(int newLength) { |
822 |
< |
ForkJoinWorkerThread[] ws = workers; |
823 |
< |
if (ws == null) |
824 |
< |
return workers = new ForkJoinWorkerThread[arraySizeFor(newLength)]; |
825 |
< |
else if (newLength > ws.length) |
826 |
< |
return workers = Arrays.copyOf(ws, arraySizeFor(newLength)); |
827 |
< |
else |
828 |
< |
return ws; |
821 |
> |
private void tryResumeSpare() { |
822 |
> |
int sw, id; |
823 |
> |
ForkJoinWorkerThread w; |
824 |
> |
ForkJoinWorkerThread[] ws; |
825 |
> |
if ((id = ((sw = spareWaiters) & SPARE_ID_MASK) - 1) >= 0 && |
826 |
> |
id < (ws = workers).length && (w = ws[id]) != null && |
827 |
> |
(workerCounts & RUNNING_COUNT_MASK) < parallelism && |
828 |
> |
eventWaiters == 0L && |
829 |
> |
spareWaiters == sw && |
830 |
> |
UNSAFE.compareAndSwapInt(this, spareWaitersOffset, |
831 |
> |
sw, w.nextSpare) && |
832 |
> |
w.tryUnsuspend()) { |
833 |
> |
int c; // try increment; if contended, finish after unpark |
834 |
> |
boolean inc = UNSAFE.compareAndSwapInt(this, workerCountsOffset, |
835 |
> |
c = workerCounts, |
836 |
> |
c + ONE_RUNNING); |
837 |
> |
LockSupport.unpark(w); |
838 |
> |
if (!inc) { |
839 |
> |
do {} while(!UNSAFE.compareAndSwapInt(this, workerCountsOffset, |
840 |
> |
c = workerCounts, |
841 |
> |
c + ONE_RUNNING)); |
842 |
> |
} |
843 |
> |
} |
844 |
|
} |
845 |
|
|
846 |
|
/** |
847 |
< |
* Tries to shrink workers into smaller array after one or more terminate. |
847 |
> |
* Callback from oldest spare occasionally waking up. Tries |
848 |
> |
* (once) to shutdown a spare if more than 25% spare overage, or |
849 |
> |
* if UNUSED_SPARE_TRIM_RATE_NANOS have elapsed and there are at |
850 |
> |
* least #parallelism running threads. Note that we don't need CAS |
851 |
> |
* or locks here because the method is called only from the oldest |
852 |
> |
* suspended spare occasionally waking (and even misfires are OK). |
853 |
> |
* |
854 |
> |
* @param now the wake up nanoTime of caller |
855 |
|
*/ |
856 |
< |
private void tryShrinkWorkerArray() { |
857 |
< |
ForkJoinWorkerThread[] ws = workers; |
858 |
< |
if (ws != null) { |
859 |
< |
int len = ws.length; |
860 |
< |
int last = len - 1; |
861 |
< |
while (last >= 0 && ws[last] == null) |
862 |
< |
--last; |
863 |
< |
int newLength = arraySizeFor(last+1); |
864 |
< |
if (newLength < len) |
865 |
< |
workers = Arrays.copyOf(ws, newLength); |
856 |
> |
final void tryTrimSpare(long now) { |
857 |
> |
long lastTrim = trimTime; |
858 |
> |
trimTime = now; |
859 |
> |
helpMaintainParallelism(); // first, help wake up any needed spares |
860 |
> |
int sw, id; |
861 |
> |
ForkJoinWorkerThread w; |
862 |
> |
ForkJoinWorkerThread[] ws; |
863 |
> |
int pc = parallelism; |
864 |
> |
int wc = workerCounts; |
865 |
> |
if ((wc & RUNNING_COUNT_MASK) >= pc && |
866 |
> |
(((wc >>> TOTAL_COUNT_SHIFT) - pc) > (pc >>> 2) + 1 ||// approx 25% |
867 |
> |
now - lastTrim >= UNUSED_SPARE_TRIM_RATE_NANOS) && |
868 |
> |
(id = ((sw = spareWaiters) & SPARE_ID_MASK) - 1) >= 0 && |
869 |
> |
id < (ws = workers).length && (w = ws[id]) != null && |
870 |
> |
UNSAFE.compareAndSwapInt(this, spareWaitersOffset, |
871 |
> |
sw, w.nextSpare)) |
872 |
> |
w.shutdown(false); |
873 |
> |
} |
874 |
> |
|
875 |
> |
/** |
876 |
> |
* Does at most one of: |
877 |
> |
* |
878 |
> |
* 1. Help wake up existing workers waiting for work via |
879 |
> |
* releaseEventWaiters. (If any exist, then it probably doesn't |
880 |
> |
* matter right now if under target parallelism level.) |
881 |
> |
* |
882 |
> |
* 2. If below parallelism level and a spare exists, try (once) |
883 |
> |
* to resume it via tryResumeSpare. |
884 |
> |
* |
885 |
> |
* 3. If neither of the above, tries (once) to add a new |
886 |
> |
* worker if either there are not enough total, or if all |
887 |
> |
* existing workers are busy, there are either no running |
888 |
> |
* workers or the deficit is at least twice the surplus. |
889 |
> |
*/ |
890 |
> |
private void helpMaintainParallelism() { |
891 |
> |
// uglified to work better when not compiled |
892 |
> |
int pc, wc, rc, tc, rs; long h; |
893 |
> |
if ((h = eventWaiters) != 0L) { |
894 |
> |
if ((int)(h >>> EVENT_COUNT_SHIFT) != eventCount) |
895 |
> |
releaseEventWaiters(false); // avoid useless call |
896 |
> |
} |
897 |
> |
else if ((pc = parallelism) > |
898 |
> |
(rc = ((wc = workerCounts) & RUNNING_COUNT_MASK))) { |
899 |
> |
if (spareWaiters != 0) |
900 |
> |
tryResumeSpare(); |
901 |
> |
else if ((rs = runState) < TERMINATING && |
902 |
> |
((tc = wc >>> TOTAL_COUNT_SHIFT) < pc || |
903 |
> |
(tc == (rs & ACTIVE_COUNT_MASK) && // all busy |
904 |
> |
(rc == 0 || // must add |
905 |
> |
rc < pc - ((tc - pc) << 1)) && // within slack |
906 |
> |
tc < MAX_WORKERS && runState == rs)) && // recheck busy |
907 |
> |
workerCounts == wc && |
908 |
> |
UNSAFE.compareAndSwapInt(this, workerCountsOffset, wc, |
909 |
> |
wc + (ONE_RUNNING|ONE_TOTAL))) |
910 |
> |
addWorker(); |
911 |
> |
} |
912 |
> |
} |
913 |
> |
|
914 |
> |
/** |
915 |
> |
* Callback from workers invoked upon each top-level action (i.e., |
916 |
> |
* stealing a task or taking a submission and running |
917 |
> |
* it). Performs one or more of the following: |
918 |
> |
* |
919 |
> |
* 1. If the worker cannot find work (misses > 0), updates its |
920 |
> |
* active status to inactive and updates activeCount unless |
921 |
> |
* this is the first miss and there is contention, in which |
922 |
> |
* case it may try again (either in this or a subsequent |
923 |
> |
* call). |
924 |
> |
* |
925 |
> |
* 2. If there are at least 2 misses, awaits the next task event |
926 |
> |
* via eventSync |
927 |
> |
* |
928 |
> |
* 3. If there are too many running threads, suspends this worker |
929 |
> |
* (first forcing inactivation if necessary). If it is not |
930 |
> |
* needed, it may be killed while suspended via |
931 |
> |
* tryTrimSpare. Otherwise, upon resume it rechecks to make |
932 |
> |
* sure that it is still needed. |
933 |
> |
* |
934 |
> |
* 4. Helps release and/or reactivate other workers via |
935 |
> |
* helpMaintainParallelism |
936 |
> |
* |
937 |
> |
* @param w the worker |
938 |
> |
* @param misses the number of scans by caller failing to find work |
939 |
> |
* (saturating at 2 just to avoid wraparound) |
940 |
> |
*/ |
941 |
> |
final void preStep(ForkJoinWorkerThread w, int misses) { |
942 |
> |
boolean active = w.active; |
943 |
> |
int pc = parallelism; |
944 |
> |
for (;;) { |
945 |
> |
int wc = workerCounts; |
946 |
> |
int rc = wc & RUNNING_COUNT_MASK; |
947 |
> |
if (active && (misses > 0 || rc > pc)) { |
948 |
> |
int rs; // try inactivate |
949 |
> |
if (UNSAFE.compareAndSwapInt(this, runStateOffset, |
950 |
> |
rs = runState, rs - ONE_ACTIVE)) |
951 |
> |
active = w.active = false; |
952 |
> |
else if (misses > 1 || rc > pc || |
953 |
> |
(rs & ACTIVE_COUNT_MASK) >= pc) |
954 |
> |
continue; // force inactivate |
955 |
> |
} |
956 |
> |
if (misses > 1) { |
957 |
> |
misses = 0; // don't re-sync |
958 |
> |
eventSync(w); // continue loop to recheck rc |
959 |
> |
} |
960 |
> |
else if (rc > pc) { |
961 |
> |
if (workerCounts == wc && // try to suspend as spare |
962 |
> |
UNSAFE.compareAndSwapInt(this, workerCountsOffset, |
963 |
> |
wc, wc - ONE_RUNNING) && |
964 |
> |
!w.suspendAsSpare()) // false if killed |
965 |
> |
break; |
966 |
> |
} |
967 |
> |
else { |
968 |
> |
if (rc < pc || eventWaiters != 0L) |
969 |
> |
helpMaintainParallelism(); |
970 |
> |
break; |
971 |
> |
} |
972 |
|
} |
973 |
|
} |
974 |
|
|
975 |
|
/** |
976 |
< |
* Initializes workers if necessary. |
977 |
< |
*/ |
978 |
< |
final void ensureWorkerInitialization() { |
979 |
< |
ForkJoinWorkerThread[] ws = workers; |
980 |
< |
if (ws == null) { |
981 |
< |
final ReentrantLock lock = this.workerLock; |
982 |
< |
lock.lock(); |
983 |
< |
try { |
984 |
< |
ws = workers; |
985 |
< |
if (ws == null) { |
986 |
< |
int ps = parallelism; |
987 |
< |
ws = ensureWorkerArrayCapacity(ps); |
988 |
< |
for (int i = 0; i < ps; ++i) { |
989 |
< |
ForkJoinWorkerThread w = createWorker(i); |
990 |
< |
if (w != null) { |
991 |
< |
ws[i] = w; |
992 |
< |
w.start(); |
993 |
< |
updateWorkerCount(1); |
994 |
< |
} |
995 |
< |
} |
996 |
< |
} |
997 |
< |
} finally { |
998 |
< |
lock.unlock(); |
976 |
> |
* Helps and/or blocks awaiting join of the given task. |
977 |
> |
* Alternates between helpJoinTask() and helpMaintainParallelism() |
978 |
> |
* as many times as there is a deficit in running count (or longer |
979 |
> |
* if running count would become zero), then blocks if task still |
980 |
> |
* not done. |
981 |
> |
* |
982 |
> |
* @param joinMe the task to join |
983 |
> |
*/ |
984 |
> |
final void awaitJoin(ForkJoinTask<?> joinMe, ForkJoinWorkerThread worker) { |
985 |
> |
int threshold = parallelism; // descend blocking thresholds |
986 |
> |
while (joinMe.status >= 0) { |
987 |
> |
boolean block; int wc; |
988 |
> |
worker.helpJoinTask(joinMe); |
989 |
> |
if (joinMe.status < 0) |
990 |
> |
break; |
991 |
> |
if (((wc = workerCounts) & RUNNING_COUNT_MASK) <= threshold) { |
992 |
> |
if (threshold > 0) |
993 |
> |
--threshold; |
994 |
> |
else |
995 |
> |
advanceEventCount(); // force release |
996 |
> |
block = false; |
997 |
> |
} |
998 |
> |
else |
999 |
> |
block = UNSAFE.compareAndSwapInt(this, workerCountsOffset, |
1000 |
> |
wc, wc - ONE_RUNNING); |
1001 |
> |
helpMaintainParallelism(); |
1002 |
> |
if (block) { |
1003 |
> |
int c; |
1004 |
> |
joinMe.internalAwaitDone(); |
1005 |
> |
do {} while (!UNSAFE.compareAndSwapInt |
1006 |
> |
(this, workerCountsOffset, |
1007 |
> |
c = workerCounts, c + ONE_RUNNING)); |
1008 |
> |
break; |
1009 |
|
} |
1010 |
|
} |
1011 |
|
} |
1012 |
|
|
1013 |
|
/** |
1014 |
< |
* Worker creation and startup for threads added via setParallelism. |
1014 |
> |
* Same idea as awaitJoin, but no helping |
1015 |
|
*/ |
1016 |
< |
private void createAndStartAddedWorkers() { |
1017 |
< |
resumeAllSpares(); // Allow spares to convert to nonspare |
1018 |
< |
int ps = parallelism; |
1019 |
< |
ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(ps); |
1020 |
< |
int len = ws.length; |
1021 |
< |
// Sweep through slots, to keep lowest indices most populated |
1022 |
< |
int k = 0; |
1023 |
< |
while (k < len) { |
1024 |
< |
if (ws[k] != null) { |
1025 |
< |
++k; |
1026 |
< |
continue; |
1016 |
> |
final void awaitBlocker(ManagedBlocker blocker) |
1017 |
> |
throws InterruptedException { |
1018 |
> |
int threshold = parallelism; |
1019 |
> |
while (!blocker.isReleasable()) { |
1020 |
> |
boolean block; int wc; |
1021 |
> |
if (((wc = workerCounts) & RUNNING_COUNT_MASK) <= threshold) { |
1022 |
> |
if (threshold > 0) |
1023 |
> |
--threshold; |
1024 |
> |
else |
1025 |
> |
advanceEventCount(); |
1026 |
> |
block = false; |
1027 |
|
} |
1028 |
< |
int s = workerCounts; |
1029 |
< |
int tc = totalCountOf(s); |
1030 |
< |
int rc = runningCountOf(s); |
1031 |
< |
if (rc >= ps || tc >= ps) |
1028 |
> |
else |
1029 |
> |
block = UNSAFE.compareAndSwapInt(this, workerCountsOffset, |
1030 |
> |
wc, wc - ONE_RUNNING); |
1031 |
> |
helpMaintainParallelism(); |
1032 |
> |
if (block) { |
1033 |
> |
try { |
1034 |
> |
do {} while (!blocker.isReleasable() && !blocker.block()); |
1035 |
> |
} finally { |
1036 |
> |
int c; |
1037 |
> |
do {} while (!UNSAFE.compareAndSwapInt |
1038 |
> |
(this, workerCountsOffset, |
1039 |
> |
c = workerCounts, c + ONE_RUNNING)); |
1040 |
> |
} |
1041 |
|
break; |
1042 |
< |
if (casWorkerCounts (s, workerCountsFor(tc+1, rc+1))) { |
1043 |
< |
ForkJoinWorkerThread w = createWorker(k); |
1042 |
> |
} |
1043 |
> |
} |
1044 |
> |
} |
1045 |
> |
|
1046 |
> |
/** |
1047 |
> |
* Possibly initiates and/or completes termination. |
1048 |
> |
* |
1049 |
> |
* @param now if true, unconditionally terminate, else only |
1050 |
> |
* if shutdown and empty queue and no active workers |
1051 |
> |
* @return true if now terminating or terminated |
1052 |
> |
*/ |
1053 |
> |
private boolean tryTerminate(boolean now) { |
1054 |
> |
if (now) |
1055 |
> |
advanceRunLevel(SHUTDOWN); // ensure at least SHUTDOWN |
1056 |
> |
else if (runState < SHUTDOWN || |
1057 |
> |
!submissionQueue.isEmpty() || |
1058 |
> |
(runState & ACTIVE_COUNT_MASK) != 0) |
1059 |
> |
return false; |
1060 |
> |
|
1061 |
> |
if (advanceRunLevel(TERMINATING)) |
1062 |
> |
startTerminating(); |
1063 |
> |
|
1064 |
> |
// Finish now if all threads terminated; else in some subsequent call |
1065 |
> |
if ((workerCounts >>> TOTAL_COUNT_SHIFT) == 0) { |
1066 |
> |
advanceRunLevel(TERMINATED); |
1067 |
> |
termination.arrive(); |
1068 |
> |
} |
1069 |
> |
return true; |
1070 |
> |
} |
1071 |
> |
|
1072 |
> |
/** |
1073 |
> |
* Actions on transition to TERMINATING |
1074 |
> |
* |
1075 |
> |
* Runs up to four passes through workers: (0) shutting down each |
1076 |
> |
* quietly (without waking up if parked) to quickly spread |
1077 |
> |
* notifications without unnecessary bouncing around event queues |
1078 |
> |
* etc (1) wake up and help cancel tasks (2) interrupt (3) mop up |
1079 |
> |
* races with interrupted workers |
1080 |
> |
*/ |
1081 |
> |
private void startTerminating() { |
1082 |
> |
cancelSubmissions(); |
1083 |
> |
for (int passes = 0; passes < 4 && workerCounts != 0; ++passes) { |
1084 |
> |
advanceEventCount(); |
1085 |
> |
eventWaiters = 0L; // clobber lists |
1086 |
> |
spareWaiters = 0; |
1087 |
> |
ForkJoinWorkerThread[] ws = workers; |
1088 |
> |
int n = ws.length; |
1089 |
> |
for (int i = 0; i < n; ++i) { |
1090 |
> |
ForkJoinWorkerThread w = ws[i]; |
1091 |
|
if (w != null) { |
1092 |
< |
ws[k++] = w; |
1093 |
< |
w.start(); |
1094 |
< |
} |
1095 |
< |
else { |
1096 |
< |
updateWorkerCount(-1); // back out on failed creation |
1097 |
< |
break; |
1092 |
> |
w.shutdown(true); |
1093 |
> |
if (passes > 0 && !w.isTerminated()) { |
1094 |
> |
w.cancelTasks(); |
1095 |
> |
LockSupport.unpark(w); |
1096 |
> |
if (passes > 1) { |
1097 |
> |
try { |
1098 |
> |
w.interrupt(); |
1099 |
> |
} catch (SecurityException ignore) { |
1100 |
> |
} |
1101 |
> |
} |
1102 |
> |
} |
1103 |
|
} |
1104 |
|
} |
1105 |
|
} |
1106 |
|
} |
1107 |
|
|
1108 |
+ |
/** |
1109 |
+ |
* Clear out and cancel submissions, ignoring exceptions |
1110 |
+ |
*/ |
1111 |
+ |
private void cancelSubmissions() { |
1112 |
+ |
ForkJoinTask<?> task; |
1113 |
+ |
while ((task = submissionQueue.poll()) != null) { |
1114 |
+ |
try { |
1115 |
+ |
task.cancel(false); |
1116 |
+ |
} catch (Throwable ignore) { |
1117 |
+ |
} |
1118 |
+ |
} |
1119 |
+ |
} |
1120 |
+ |
|
1121 |
+ |
// misc support for ForkJoinWorkerThread |
1122 |
+ |
|
1123 |
+ |
/** |
1124 |
+ |
* Returns pool number |
1125 |
+ |
*/ |
1126 |
+ |
final int getPoolNumber() { |
1127 |
+ |
return poolNumber; |
1128 |
+ |
} |
1129 |
+ |
|
1130 |
+ |
/** |
1131 |
+ |
* Tries to accumulates steal count from a worker, clearing |
1132 |
+ |
* the worker's value. |
1133 |
+ |
* |
1134 |
+ |
* @return true if worker steal count now zero |
1135 |
+ |
*/ |
1136 |
+ |
final boolean tryAccumulateStealCount(ForkJoinWorkerThread w) { |
1137 |
+ |
int sc = w.stealCount; |
1138 |
+ |
long c = stealCount; |
1139 |
+ |
// CAS even if zero, for fence effects |
1140 |
+ |
if (UNSAFE.compareAndSwapLong(this, stealCountOffset, c, c + sc)) { |
1141 |
+ |
if (sc != 0) |
1142 |
+ |
w.stealCount = 0; |
1143 |
+ |
return true; |
1144 |
+ |
} |
1145 |
+ |
return sc == 0; |
1146 |
+ |
} |
1147 |
+ |
|
1148 |
+ |
/** |
1149 |
+ |
* Returns the approximate (non-atomic) number of idle threads per |
1150 |
+ |
* active thread. |
1151 |
+ |
*/ |
1152 |
+ |
final int idlePerActive() { |
1153 |
+ |
int pc = parallelism; // use parallelism, not rc |
1154 |
+ |
int ac = runState; // no mask -- artifically boosts during shutdown |
1155 |
+ |
// Use exact results for small values, saturate past 4 |
1156 |
+ |
return pc <= ac? 0 : pc >>> 1 <= ac? 1 : pc >>> 2 <= ac? 3 : pc >>> 3; |
1157 |
+ |
} |
1158 |
+ |
|
1159 |
+ |
// Public and protected methods |
1160 |
+ |
|
1161 |
+ |
// Constructors |
1162 |
+ |
|
1163 |
+ |
/** |
1164 |
+ |
* Creates a {@code ForkJoinPool} with parallelism equal to {@link |
1165 |
+ |
* java.lang.Runtime#availableProcessors}, using the {@linkplain |
1166 |
+ |
* #defaultForkJoinWorkerThreadFactory default thread factory}, |
1167 |
+ |
* no UncaughtExceptionHandler, and non-async LIFO processing mode. |
1168 |
+ |
* |
1169 |
+ |
* @throws SecurityException if a security manager exists and |
1170 |
+ |
* the caller is not permitted to modify threads |
1171 |
+ |
* because it does not hold {@link |
1172 |
+ |
* java.lang.RuntimePermission}{@code ("modifyThread")} |
1173 |
+ |
*/ |
1174 |
+ |
public ForkJoinPool() { |
1175 |
+ |
this(Runtime.getRuntime().availableProcessors(), |
1176 |
+ |
defaultForkJoinWorkerThreadFactory, null, false); |
1177 |
+ |
} |
1178 |
+ |
|
1179 |
+ |
/** |
1180 |
+ |
* Creates a {@code ForkJoinPool} with the indicated parallelism |
1181 |
+ |
* level, the {@linkplain |
1182 |
+ |
* #defaultForkJoinWorkerThreadFactory default thread factory}, |
1183 |
+ |
* no UncaughtExceptionHandler, and non-async LIFO processing mode. |
1184 |
+ |
* |
1185 |
+ |
* @param parallelism the parallelism level |
1186 |
+ |
* @throws IllegalArgumentException if parallelism less than or |
1187 |
+ |
* equal to zero, or greater than implementation limit |
1188 |
+ |
* @throws SecurityException if a security manager exists and |
1189 |
+ |
* the caller is not permitted to modify threads |
1190 |
+ |
* because it does not hold {@link |
1191 |
+ |
* java.lang.RuntimePermission}{@code ("modifyThread")} |
1192 |
+ |
*/ |
1193 |
+ |
public ForkJoinPool(int parallelism) { |
1194 |
+ |
this(parallelism, defaultForkJoinWorkerThreadFactory, null, false); |
1195 |
+ |
} |
1196 |
+ |
|
1197 |
+ |
/** |
1198 |
+ |
* Creates a {@code ForkJoinPool} with the given parameters. |
1199 |
+ |
* |
1200 |
+ |
* @param parallelism the parallelism level. For default value, |
1201 |
+ |
* use {@link java.lang.Runtime#availableProcessors}. |
1202 |
+ |
* @param factory the factory for creating new threads. For default value, |
1203 |
+ |
* use {@link #defaultForkJoinWorkerThreadFactory}. |
1204 |
+ |
* @param handler the handler for internal worker threads that |
1205 |
+ |
* terminate due to unrecoverable errors encountered while executing |
1206 |
+ |
* tasks. For default value, use <code>null</code>. |
1207 |
+ |
* @param asyncMode if true, |
1208 |
+ |
* establishes local first-in-first-out scheduling mode for forked |
1209 |
+ |
* tasks that are never joined. This mode may be more appropriate |
1210 |
+ |
* than default locally stack-based mode in applications in which |
1211 |
+ |
* worker threads only process event-style asynchronous tasks. |
1212 |
+ |
* For default value, use <code>false</code>. |
1213 |
+ |
* @throws IllegalArgumentException if parallelism less than or |
1214 |
+ |
* equal to zero, or greater than implementation limit |
1215 |
+ |
* @throws NullPointerException if the factory is null |
1216 |
+ |
* @throws SecurityException if a security manager exists and |
1217 |
+ |
* the caller is not permitted to modify threads |
1218 |
+ |
* because it does not hold {@link |
1219 |
+ |
* java.lang.RuntimePermission}{@code ("modifyThread")} |
1220 |
+ |
*/ |
1221 |
+ |
public ForkJoinPool(int parallelism, |
1222 |
+ |
ForkJoinWorkerThreadFactory factory, |
1223 |
+ |
Thread.UncaughtExceptionHandler handler, |
1224 |
+ |
boolean asyncMode) { |
1225 |
+ |
checkPermission(); |
1226 |
+ |
if (factory == null) |
1227 |
+ |
throw new NullPointerException(); |
1228 |
+ |
if (parallelism <= 0 || parallelism > MAX_WORKERS) |
1229 |
+ |
throw new IllegalArgumentException(); |
1230 |
+ |
this.parallelism = parallelism; |
1231 |
+ |
this.factory = factory; |
1232 |
+ |
this.ueh = handler; |
1233 |
+ |
this.locallyFifo = asyncMode; |
1234 |
+ |
int arraySize = initialArraySizeFor(parallelism); |
1235 |
+ |
this.workers = new ForkJoinWorkerThread[arraySize]; |
1236 |
+ |
this.submissionQueue = new LinkedTransferQueue<ForkJoinTask<?>>(); |
1237 |
+ |
this.workerLock = new ReentrantLock(); |
1238 |
+ |
this.termination = new Phaser(1); |
1239 |
+ |
this.poolNumber = poolNumberGenerator.incrementAndGet(); |
1240 |
+ |
this.trimTime = System.nanoTime(); |
1241 |
+ |
} |
1242 |
+ |
|
1243 |
+ |
/** |
1244 |
+ |
* Returns initial power of two size for workers array. |
1245 |
+ |
* @param pc the initial parallelism level |
1246 |
+ |
*/ |
1247 |
+ |
private static int initialArraySizeFor(int pc) { |
1248 |
+ |
// See Hackers Delight, sec 3.2. We know MAX_WORKERS < (1 >>> 16) |
1249 |
+ |
int size = pc < MAX_WORKERS ? pc + 1 : MAX_WORKERS; |
1250 |
+ |
size |= size >>> 1; |
1251 |
+ |
size |= size >>> 2; |
1252 |
+ |
size |= size >>> 4; |
1253 |
+ |
size |= size >>> 8; |
1254 |
+ |
return size + 1; |
1255 |
+ |
} |
1256 |
+ |
|
1257 |
|
// Execution methods |
1258 |
|
|
1259 |
|
/** |
1262 |
|
private <T> void doSubmit(ForkJoinTask<T> task) { |
1263 |
|
if (task == null) |
1264 |
|
throw new NullPointerException(); |
1265 |
< |
if (isShutdown()) |
1265 |
> |
if (runState >= SHUTDOWN) |
1266 |
|
throw new RejectedExecutionException(); |
586 |
– |
if (workers == null) |
587 |
– |
ensureWorkerInitialization(); |
1267 |
|
submissionQueue.offer(task); |
1268 |
< |
signalIdleWorkers(); |
1268 |
> |
advanceEventCount(); |
1269 |
> |
helpMaintainParallelism(); // start or wake up workers |
1270 |
|
} |
1271 |
|
|
1272 |
|
/** |
1273 |
|
* Performs the given task, returning its result upon completion. |
1274 |
+ |
* If the caller is already engaged in a fork/join computation in |
1275 |
+ |
* the current pool, this method is equivalent in effect to |
1276 |
+ |
* {@link ForkJoinTask#invoke}. |
1277 |
|
* |
1278 |
|
* @param task the task |
1279 |
|
* @return the task's result |
1280 |
< |
* @throws NullPointerException if task is null |
1281 |
< |
* @throws RejectedExecutionException if pool is shut down |
1280 |
> |
* @throws NullPointerException if the task is null |
1281 |
> |
* @throws RejectedExecutionException if the task cannot be |
1282 |
> |
* scheduled for execution |
1283 |
|
*/ |
1284 |
|
public <T> T invoke(ForkJoinTask<T> task) { |
1285 |
|
doSubmit(task); |
1288 |
|
|
1289 |
|
/** |
1290 |
|
* Arranges for (asynchronous) execution of the given task. |
1291 |
+ |
* If the caller is already engaged in a fork/join computation in |
1292 |
+ |
* the current pool, this method is equivalent in effect to |
1293 |
+ |
* {@link ForkJoinTask#fork}. |
1294 |
|
* |
1295 |
|
* @param task the task |
1296 |
< |
* @throws NullPointerException if task is null |
1297 |
< |
* @throws RejectedExecutionException if pool is shut down |
1296 |
> |
* @throws NullPointerException if the task is null |
1297 |
> |
* @throws RejectedExecutionException if the task cannot be |
1298 |
> |
* scheduled for execution |
1299 |
|
*/ |
1300 |
|
public void execute(ForkJoinTask<?> task) { |
1301 |
|
doSubmit(task); |
1303 |
|
|
1304 |
|
// AbstractExecutorService methods |
1305 |
|
|
1306 |
+ |
/** |
1307 |
+ |
* @throws NullPointerException if the task is null |
1308 |
+ |
* @throws RejectedExecutionException if the task cannot be |
1309 |
+ |
* scheduled for execution |
1310 |
+ |
*/ |
1311 |
|
public void execute(Runnable task) { |
1312 |
|
ForkJoinTask<?> job; |
1313 |
|
if (task instanceof ForkJoinTask<?>) // avoid re-wrap |
1317 |
|
doSubmit(job); |
1318 |
|
} |
1319 |
|
|
1320 |
+ |
/** |
1321 |
+ |
* Submits a ForkJoinTask for execution. |
1322 |
+ |
* If the caller is already engaged in a fork/join computation in |
1323 |
+ |
* the current pool, this method is equivalent in effect to |
1324 |
+ |
* {@link ForkJoinTask#fork}. |
1325 |
+ |
* |
1326 |
+ |
* @param task the task to submit |
1327 |
+ |
* @return the task |
1328 |
+ |
* @throws NullPointerException if the task is null |
1329 |
+ |
* @throws RejectedExecutionException if the task cannot be |
1330 |
+ |
* scheduled for execution |
1331 |
+ |
*/ |
1332 |
+ |
public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) { |
1333 |
+ |
doSubmit(task); |
1334 |
+ |
return task; |
1335 |
+ |
} |
1336 |
+ |
|
1337 |
+ |
/** |
1338 |
+ |
* @throws NullPointerException if the task is null |
1339 |
+ |
* @throws RejectedExecutionException if the task cannot be |
1340 |
+ |
* scheduled for execution |
1341 |
+ |
*/ |
1342 |
|
public <T> ForkJoinTask<T> submit(Callable<T> task) { |
1343 |
|
ForkJoinTask<T> job = ForkJoinTask.adapt(task); |
1344 |
|
doSubmit(job); |
1345 |
|
return job; |
1346 |
|
} |
1347 |
|
|
1348 |
+ |
/** |
1349 |
+ |
* @throws NullPointerException if the task is null |
1350 |
+ |
* @throws RejectedExecutionException if the task cannot be |
1351 |
+ |
* scheduled for execution |
1352 |
+ |
*/ |
1353 |
|
public <T> ForkJoinTask<T> submit(Runnable task, T result) { |
1354 |
|
ForkJoinTask<T> job = ForkJoinTask.adapt(task, result); |
1355 |
|
doSubmit(job); |
1356 |
|
return job; |
1357 |
|
} |
1358 |
|
|
1359 |
+ |
/** |
1360 |
+ |
* @throws NullPointerException if the task is null |
1361 |
+ |
* @throws RejectedExecutionException if the task cannot be |
1362 |
+ |
* scheduled for execution |
1363 |
+ |
*/ |
1364 |
|
public ForkJoinTask<?> submit(Runnable task) { |
1365 |
|
ForkJoinTask<?> job; |
1366 |
|
if (task instanceof ForkJoinTask<?>) // avoid re-wrap |
1372 |
|
} |
1373 |
|
|
1374 |
|
/** |
1375 |
< |
* Submits a ForkJoinTask for execution. |
1376 |
< |
* |
652 |
< |
* @param task the task to submit |
653 |
< |
* @return the task |
654 |
< |
* @throws RejectedExecutionException if the task cannot be |
655 |
< |
* scheduled for execution |
656 |
< |
* @throws NullPointerException if the task is null |
1375 |
> |
* @throws NullPointerException {@inheritDoc} |
1376 |
> |
* @throws RejectedExecutionException {@inheritDoc} |
1377 |
|
*/ |
658 |
– |
public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) { |
659 |
– |
doSubmit(task); |
660 |
– |
return task; |
661 |
– |
} |
662 |
– |
|
663 |
– |
|
1378 |
|
public <T> List<Future<T>> invokeAll(Collection<? extends Callable<T>> tasks) { |
1379 |
|
ArrayList<ForkJoinTask<T>> forkJoinTasks = |
1380 |
|
new ArrayList<ForkJoinTask<T>>(tasks.size()); |
1383 |
|
invoke(new InvokeAll<T>(forkJoinTasks)); |
1384 |
|
|
1385 |
|
@SuppressWarnings({"unchecked", "rawtypes"}) |
1386 |
< |
List<Future<T>> futures = (List<Future<T>>) (List) forkJoinTasks; |
1386 |
> |
List<Future<T>> futures = (List<Future<T>>) (List) forkJoinTasks; |
1387 |
|
return futures; |
1388 |
|
} |
1389 |
|
|
1397 |
|
private static final long serialVersionUID = -7914297376763021607L; |
1398 |
|
} |
1399 |
|
|
686 |
– |
// Configuration and status settings and queries |
687 |
– |
|
1400 |
|
/** |
1401 |
|
* Returns the factory used for constructing new workers. |
1402 |
|
* |
1413 |
|
* @return the handler, or {@code null} if none |
1414 |
|
*/ |
1415 |
|
public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() { |
1416 |
< |
Thread.UncaughtExceptionHandler h; |
705 |
< |
final ReentrantLock lock = this.workerLock; |
706 |
< |
lock.lock(); |
707 |
< |
try { |
708 |
< |
h = ueh; |
709 |
< |
} finally { |
710 |
< |
lock.unlock(); |
711 |
< |
} |
712 |
< |
return h; |
713 |
< |
} |
714 |
< |
|
715 |
< |
/** |
716 |
< |
* Sets the handler for internal worker threads that terminate due |
717 |
< |
* to unrecoverable errors encountered while executing tasks. |
718 |
< |
* Unless set, the current default or ThreadGroup handler is used |
719 |
< |
* as handler. |
720 |
< |
* |
721 |
< |
* @param h the new handler |
722 |
< |
* @return the old handler, or {@code null} if none |
723 |
< |
* @throws SecurityException if a security manager exists and |
724 |
< |
* the caller is not permitted to modify threads |
725 |
< |
* because it does not hold {@link |
726 |
< |
* java.lang.RuntimePermission}{@code ("modifyThread")} |
727 |
< |
*/ |
728 |
< |
public Thread.UncaughtExceptionHandler |
729 |
< |
setUncaughtExceptionHandler(Thread.UncaughtExceptionHandler h) { |
730 |
< |
checkPermission(); |
731 |
< |
Thread.UncaughtExceptionHandler old = null; |
732 |
< |
final ReentrantLock lock = this.workerLock; |
733 |
< |
lock.lock(); |
734 |
< |
try { |
735 |
< |
old = ueh; |
736 |
< |
ueh = h; |
737 |
< |
ForkJoinWorkerThread[] ws = workers; |
738 |
< |
if (ws != null) { |
739 |
< |
for (int i = 0; i < ws.length; ++i) { |
740 |
< |
ForkJoinWorkerThread w = ws[i]; |
741 |
< |
if (w != null) |
742 |
< |
w.setUncaughtExceptionHandler(h); |
743 |
< |
} |
744 |
< |
} |
745 |
< |
} finally { |
746 |
< |
lock.unlock(); |
747 |
< |
} |
748 |
< |
return old; |
749 |
< |
} |
750 |
< |
|
751 |
< |
|
752 |
< |
/** |
753 |
< |
* Sets the target parallelism level of this pool. |
754 |
< |
* |
755 |
< |
* @param parallelism the target parallelism |
756 |
< |
* @throws IllegalArgumentException if parallelism less than or |
757 |
< |
* equal to zero or greater than maximum size bounds |
758 |
< |
* @throws SecurityException if a security manager exists and |
759 |
< |
* the caller is not permitted to modify threads |
760 |
< |
* because it does not hold {@link |
761 |
< |
* java.lang.RuntimePermission}{@code ("modifyThread")} |
762 |
< |
*/ |
763 |
< |
public void setParallelism(int parallelism) { |
764 |
< |
checkPermission(); |
765 |
< |
if (parallelism <= 0 || parallelism > maxPoolSize) |
766 |
< |
throw new IllegalArgumentException(); |
767 |
< |
final ReentrantLock lock = this.workerLock; |
768 |
< |
lock.lock(); |
769 |
< |
try { |
770 |
< |
if (isProcessingTasks()) { |
771 |
< |
int p = this.parallelism; |
772 |
< |
this.parallelism = parallelism; |
773 |
< |
if (parallelism > p) |
774 |
< |
createAndStartAddedWorkers(); |
775 |
< |
else |
776 |
< |
trimSpares(); |
777 |
< |
} |
778 |
< |
} finally { |
779 |
< |
lock.unlock(); |
780 |
< |
} |
781 |
< |
signalIdleWorkers(); |
1416 |
> |
return ueh; |
1417 |
|
} |
1418 |
|
|
1419 |
|
/** |
1434 |
|
* @return the number of worker threads |
1435 |
|
*/ |
1436 |
|
public int getPoolSize() { |
1437 |
< |
return totalCountOf(workerCounts); |
803 |
< |
} |
804 |
< |
|
805 |
< |
/** |
806 |
< |
* Returns the maximum number of threads allowed to exist in the |
807 |
< |
* pool. Unless set using {@link #setMaximumPoolSize}, the |
808 |
< |
* maximum is an implementation-defined value designed only to |
809 |
< |
* prevent runaway growth. |
810 |
< |
* |
811 |
< |
* @return the maximum |
812 |
< |
*/ |
813 |
< |
public int getMaximumPoolSize() { |
814 |
< |
return maxPoolSize; |
815 |
< |
} |
816 |
< |
|
817 |
< |
/** |
818 |
< |
* Sets the maximum number of threads allowed to exist in the |
819 |
< |
* pool. The given value should normally be greater than or equal |
820 |
< |
* to the {@link #getParallelism parallelism} level. Setting this |
821 |
< |
* value has no effect on current pool size. It controls |
822 |
< |
* construction of new threads. |
823 |
< |
* |
824 |
< |
* @throws IllegalArgumentException if negative or greater than |
825 |
< |
* internal implementation limit |
826 |
< |
*/ |
827 |
< |
public void setMaximumPoolSize(int newMax) { |
828 |
< |
if (newMax < 0 || newMax > MAX_THREADS) |
829 |
< |
throw new IllegalArgumentException(); |
830 |
< |
maxPoolSize = newMax; |
831 |
< |
} |
832 |
< |
|
833 |
< |
|
834 |
< |
/** |
835 |
< |
* Returns {@code true} if this pool dynamically maintains its |
836 |
< |
* target parallelism level. If false, new threads are added only |
837 |
< |
* to avoid possible starvation. This setting is by default true. |
838 |
< |
* |
839 |
< |
* @return {@code true} if maintains parallelism |
840 |
< |
*/ |
841 |
< |
public boolean getMaintainsParallelism() { |
842 |
< |
return maintainsParallelism; |
843 |
< |
} |
844 |
< |
|
845 |
< |
/** |
846 |
< |
* Sets whether this pool dynamically maintains its target |
847 |
< |
* parallelism level. If false, new threads are added only to |
848 |
< |
* avoid possible starvation. |
849 |
< |
* |
850 |
< |
* @param enable {@code true} to maintain parallelism |
851 |
< |
*/ |
852 |
< |
public void setMaintainsParallelism(boolean enable) { |
853 |
< |
maintainsParallelism = enable; |
854 |
< |
} |
855 |
< |
|
856 |
< |
/** |
857 |
< |
* Establishes local first-in-first-out scheduling mode for forked |
858 |
< |
* tasks that are never joined. This mode may be more appropriate |
859 |
< |
* than default locally stack-based mode in applications in which |
860 |
< |
* worker threads only process asynchronous tasks. This method is |
861 |
< |
* designed to be invoked only when the pool is quiescent, and |
862 |
< |
* typically only before any tasks are submitted. The effects of |
863 |
< |
* invocations at other times may be unpredictable. |
864 |
< |
* |
865 |
< |
* @param async if {@code true}, use locally FIFO scheduling |
866 |
< |
* @return the previous mode |
867 |
< |
* @see #getAsyncMode |
868 |
< |
*/ |
869 |
< |
public boolean setAsyncMode(boolean async) { |
870 |
< |
boolean oldMode = locallyFifo; |
871 |
< |
locallyFifo = async; |
872 |
< |
ForkJoinWorkerThread[] ws = workers; |
873 |
< |
if (ws != null) { |
874 |
< |
for (int i = 0; i < ws.length; ++i) { |
875 |
< |
ForkJoinWorkerThread t = ws[i]; |
876 |
< |
if (t != null) |
877 |
< |
t.setAsyncMode(async); |
878 |
< |
} |
879 |
< |
} |
880 |
< |
return oldMode; |
1437 |
> |
return workerCounts >>> TOTAL_COUNT_SHIFT; |
1438 |
|
} |
1439 |
|
|
1440 |
|
/** |
1442 |
|
* scheduling mode for forked tasks that are never joined. |
1443 |
|
* |
1444 |
|
* @return {@code true} if this pool uses async mode |
888 |
– |
* @see #setAsyncMode |
1445 |
|
*/ |
1446 |
|
public boolean getAsyncMode() { |
1447 |
|
return locallyFifo; |
1450 |
|
/** |
1451 |
|
* Returns an estimate of the number of worker threads that are |
1452 |
|
* not blocked waiting to join tasks or for other managed |
1453 |
< |
* synchronization. |
1453 |
> |
* synchronization. This method may overestimate the |
1454 |
> |
* number of running threads. |
1455 |
|
* |
1456 |
|
* @return the number of worker threads |
1457 |
|
*/ |
1458 |
|
public int getRunningThreadCount() { |
1459 |
< |
return runningCountOf(workerCounts); |
1459 |
> |
return workerCounts & RUNNING_COUNT_MASK; |
1460 |
|
} |
1461 |
|
|
1462 |
|
/** |
1467 |
|
* @return the number of active threads |
1468 |
|
*/ |
1469 |
|
public int getActiveThreadCount() { |
1470 |
< |
return activeCountOf(runControl); |
914 |
< |
} |
915 |
< |
|
916 |
< |
/** |
917 |
< |
* Returns an estimate of the number of threads that are currently |
918 |
< |
* idle waiting for tasks. This method may underestimate the |
919 |
< |
* number of idle threads. |
920 |
< |
* |
921 |
< |
* @return the number of idle threads |
922 |
< |
*/ |
923 |
< |
final int getIdleThreadCount() { |
924 |
< |
int c = runningCountOf(workerCounts) - activeCountOf(runControl); |
925 |
< |
return (c <= 0) ? 0 : c; |
1470 |
> |
return runState & ACTIVE_COUNT_MASK; |
1471 |
|
} |
1472 |
|
|
1473 |
|
/** |
1482 |
|
* @return {@code true} if all threads are currently idle |
1483 |
|
*/ |
1484 |
|
public boolean isQuiescent() { |
1485 |
< |
return activeCountOf(runControl) == 0; |
1485 |
> |
return (runState & ACTIVE_COUNT_MASK) == 0; |
1486 |
|
} |
1487 |
|
|
1488 |
|
/** |
1497 |
|
* @return the number of steals |
1498 |
|
*/ |
1499 |
|
public long getStealCount() { |
1500 |
< |
return stealCount.get(); |
956 |
< |
} |
957 |
< |
|
958 |
< |
/** |
959 |
< |
* Accumulates steal count from a worker. |
960 |
< |
* Call only when worker known to be idle. |
961 |
< |
*/ |
962 |
< |
private void updateStealCount(ForkJoinWorkerThread w) { |
963 |
< |
int sc = w.getAndClearStealCount(); |
964 |
< |
if (sc != 0) |
965 |
< |
stealCount.addAndGet(sc); |
1500 |
> |
return stealCount; |
1501 |
|
} |
1502 |
|
|
1503 |
|
/** |
1513 |
|
public long getQueuedTaskCount() { |
1514 |
|
long count = 0; |
1515 |
|
ForkJoinWorkerThread[] ws = workers; |
1516 |
< |
if (ws != null) { |
1517 |
< |
for (int i = 0; i < ws.length; ++i) { |
1518 |
< |
ForkJoinWorkerThread t = ws[i]; |
1519 |
< |
if (t != null) |
1520 |
< |
count += t.getQueueSize(); |
986 |
< |
} |
1516 |
> |
int n = ws.length; |
1517 |
> |
for (int i = 0; i < n; ++i) { |
1518 |
> |
ForkJoinWorkerThread w = ws[i]; |
1519 |
> |
if (w != null) |
1520 |
> |
count += w.getQueueSize(); |
1521 |
|
} |
1522 |
|
return count; |
1523 |
|
} |
1572 |
|
* @return the number of elements transferred |
1573 |
|
*/ |
1574 |
|
protected int drainTasksTo(Collection<? super ForkJoinTask<?>> c) { |
1575 |
< |
int n = submissionQueue.drainTo(c); |
1575 |
> |
int count = submissionQueue.drainTo(c); |
1576 |
|
ForkJoinWorkerThread[] ws = workers; |
1577 |
< |
if (ws != null) { |
1578 |
< |
for (int i = 0; i < ws.length; ++i) { |
1579 |
< |
ForkJoinWorkerThread w = ws[i]; |
1580 |
< |
if (w != null) |
1581 |
< |
n += w.drainTasksTo(c); |
1048 |
< |
} |
1577 |
> |
int n = ws.length; |
1578 |
> |
for (int i = 0; i < n; ++i) { |
1579 |
> |
ForkJoinWorkerThread w = ws[i]; |
1580 |
> |
if (w != null) |
1581 |
> |
count += w.drainTasksTo(c); |
1582 |
|
} |
1583 |
< |
return n; |
1583 |
> |
return count; |
1584 |
|
} |
1585 |
|
|
1586 |
|
/** |
1591 |
|
* @return a string identifying this pool, as well as its state |
1592 |
|
*/ |
1593 |
|
public String toString() { |
1061 |
– |
int ps = parallelism; |
1062 |
– |
int wc = workerCounts; |
1063 |
– |
int rc = runControl; |
1594 |
|
long st = getStealCount(); |
1595 |
|
long qt = getQueuedTaskCount(); |
1596 |
|
long qs = getQueuedSubmissionCount(); |
1597 |
+ |
int wc = workerCounts; |
1598 |
+ |
int tc = wc >>> TOTAL_COUNT_SHIFT; |
1599 |
+ |
int rc = wc & RUNNING_COUNT_MASK; |
1600 |
+ |
int pc = parallelism; |
1601 |
+ |
int rs = runState; |
1602 |
+ |
int ac = rs & ACTIVE_COUNT_MASK; |
1603 |
|
return super.toString() + |
1604 |
< |
"[" + runStateToString(runStateOf(rc)) + |
1605 |
< |
", parallelism = " + ps + |
1606 |
< |
", size = " + totalCountOf(wc) + |
1607 |
< |
", active = " + activeCountOf(rc) + |
1608 |
< |
", running = " + runningCountOf(wc) + |
1604 |
> |
"[" + runLevelToString(rs) + |
1605 |
> |
", parallelism = " + pc + |
1606 |
> |
", size = " + tc + |
1607 |
> |
", active = " + ac + |
1608 |
> |
", running = " + rc + |
1609 |
|
", steals = " + st + |
1610 |
|
", tasks = " + qt + |
1611 |
|
", submissions = " + qs + |
1612 |
|
"]"; |
1613 |
|
} |
1614 |
|
|
1615 |
< |
private static String runStateToString(int rs) { |
1616 |
< |
switch(rs) { |
1617 |
< |
case RUNNING: return "Running"; |
1618 |
< |
case SHUTDOWN: return "Shutting down"; |
1619 |
< |
case TERMINATING: return "Terminating"; |
1084 |
< |
case TERMINATED: return "Terminated"; |
1085 |
< |
default: throw new Error("Unknown run state"); |
1086 |
< |
} |
1615 |
> |
private static String runLevelToString(int s) { |
1616 |
> |
return ((s & TERMINATED) != 0 ? "Terminated" : |
1617 |
> |
((s & TERMINATING) != 0 ? "Terminating" : |
1618 |
> |
((s & SHUTDOWN) != 0 ? "Shutting down" : |
1619 |
> |
"Running"))); |
1620 |
|
} |
1621 |
|
|
1089 |
– |
// lifecycle control |
1090 |
– |
|
1622 |
|
/** |
1623 |
|
* Initiates an orderly shutdown in which previously submitted |
1624 |
|
* tasks are executed, but no new tasks will be accepted. |
1633 |
|
*/ |
1634 |
|
public void shutdown() { |
1635 |
|
checkPermission(); |
1636 |
< |
transitionRunStateTo(SHUTDOWN); |
1637 |
< |
if (canTerminateOnShutdown(runControl)) { |
1107 |
< |
if (workers == null) { // shutting down before workers created |
1108 |
< |
final ReentrantLock lock = this.workerLock; |
1109 |
< |
lock.lock(); |
1110 |
< |
try { |
1111 |
< |
if (workers == null) { |
1112 |
< |
terminate(); |
1113 |
< |
transitionRunStateTo(TERMINATED); |
1114 |
< |
termination.signalAll(); |
1115 |
< |
} |
1116 |
< |
} finally { |
1117 |
< |
lock.unlock(); |
1118 |
< |
} |
1119 |
< |
} |
1120 |
< |
terminateOnShutdown(); |
1121 |
< |
} |
1636 |
> |
advanceRunLevel(SHUTDOWN); |
1637 |
> |
tryTerminate(false); |
1638 |
|
} |
1639 |
|
|
1640 |
|
/** |
1655 |
|
*/ |
1656 |
|
public List<Runnable> shutdownNow() { |
1657 |
|
checkPermission(); |
1658 |
< |
terminate(); |
1658 |
> |
tryTerminate(true); |
1659 |
|
return Collections.emptyList(); |
1660 |
|
} |
1661 |
|
|
1665 |
|
* @return {@code true} if all tasks have completed following shut down |
1666 |
|
*/ |
1667 |
|
public boolean isTerminated() { |
1668 |
< |
return runStateOf(runControl) == TERMINATED; |
1668 |
> |
return runState >= TERMINATED; |
1669 |
|
} |
1670 |
|
|
1671 |
|
/** |
1679 |
|
* @return {@code true} if terminating but not yet terminated |
1680 |
|
*/ |
1681 |
|
public boolean isTerminating() { |
1682 |
< |
return runStateOf(runControl) == TERMINATING; |
1682 |
> |
return (runState & (TERMINATING|TERMINATED)) == TERMINATING; |
1683 |
|
} |
1684 |
|
|
1685 |
|
/** |
1688 |
|
* @return {@code true} if this pool has been shut down |
1689 |
|
*/ |
1690 |
|
public boolean isShutdown() { |
1691 |
< |
return runStateOf(runControl) >= SHUTDOWN; |
1176 |
< |
} |
1177 |
< |
|
1178 |
< |
/** |
1179 |
< |
* Returns true if pool is not terminating or terminated. |
1180 |
< |
* Used internally to suppress execution when terminating. |
1181 |
< |
*/ |
1182 |
< |
final boolean isProcessingTasks() { |
1183 |
< |
return runStateOf(runControl) < TERMINATING; |
1691 |
> |
return runState >= SHUTDOWN; |
1692 |
|
} |
1693 |
|
|
1694 |
|
/** |
1704 |
|
*/ |
1705 |
|
public boolean awaitTermination(long timeout, TimeUnit unit) |
1706 |
|
throws InterruptedException { |
1199 |
– |
long nanos = unit.toNanos(timeout); |
1200 |
– |
final ReentrantLock lock = this.workerLock; |
1201 |
– |
lock.lock(); |
1202 |
– |
try { |
1203 |
– |
for (;;) { |
1204 |
– |
if (isTerminated()) |
1205 |
– |
return true; |
1206 |
– |
if (nanos <= 0) |
1207 |
– |
return false; |
1208 |
– |
nanos = termination.awaitNanos(nanos); |
1209 |
– |
} |
1210 |
– |
} finally { |
1211 |
– |
lock.unlock(); |
1212 |
– |
} |
1213 |
– |
} |
1214 |
– |
|
1215 |
– |
// Shutdown and termination support |
1216 |
– |
|
1217 |
– |
/** |
1218 |
– |
* Callback from terminating worker. Nulls out the corresponding |
1219 |
– |
* workers slot, and if terminating, tries to terminate; else |
1220 |
– |
* tries to shrink workers array. |
1221 |
– |
* |
1222 |
– |
* @param w the worker |
1223 |
– |
*/ |
1224 |
– |
final void workerTerminated(ForkJoinWorkerThread w) { |
1225 |
– |
updateStealCount(w); |
1226 |
– |
updateWorkerCount(-1); |
1227 |
– |
final ReentrantLock lock = this.workerLock; |
1228 |
– |
lock.lock(); |
1229 |
– |
try { |
1230 |
– |
ForkJoinWorkerThread[] ws = workers; |
1231 |
– |
if (ws != null) { |
1232 |
– |
int idx = w.poolIndex; |
1233 |
– |
if (idx >= 0 && idx < ws.length && ws[idx] == w) |
1234 |
– |
ws[idx] = null; |
1235 |
– |
if (totalCountOf(workerCounts) == 0) { |
1236 |
– |
terminate(); // no-op if already terminating |
1237 |
– |
transitionRunStateTo(TERMINATED); |
1238 |
– |
termination.signalAll(); |
1239 |
– |
} |
1240 |
– |
else if (isProcessingTasks()) { |
1241 |
– |
tryShrinkWorkerArray(); |
1242 |
– |
tryResumeSpare(true); // allow replacement |
1243 |
– |
} |
1244 |
– |
} |
1245 |
– |
} finally { |
1246 |
– |
lock.unlock(); |
1247 |
– |
} |
1248 |
– |
signalIdleWorkers(); |
1249 |
– |
} |
1250 |
– |
|
1251 |
– |
/** |
1252 |
– |
* Initiates termination. |
1253 |
– |
*/ |
1254 |
– |
private void terminate() { |
1255 |
– |
if (transitionRunStateTo(TERMINATING)) { |
1256 |
– |
stopAllWorkers(); |
1257 |
– |
resumeAllSpares(); |
1258 |
– |
signalIdleWorkers(); |
1259 |
– |
cancelQueuedSubmissions(); |
1260 |
– |
cancelQueuedWorkerTasks(); |
1261 |
– |
interruptUnterminatedWorkers(); |
1262 |
– |
signalIdleWorkers(); // resignal after interrupt |
1263 |
– |
} |
1264 |
– |
} |
1265 |
– |
|
1266 |
– |
/** |
1267 |
– |
* Possibly terminates when on shutdown state. |
1268 |
– |
*/ |
1269 |
– |
private void terminateOnShutdown() { |
1270 |
– |
if (!hasQueuedSubmissions() && canTerminateOnShutdown(runControl)) |
1271 |
– |
terminate(); |
1272 |
– |
} |
1273 |
– |
|
1274 |
– |
/** |
1275 |
– |
* Clears out and cancels submissions. |
1276 |
– |
*/ |
1277 |
– |
private void cancelQueuedSubmissions() { |
1278 |
– |
ForkJoinTask<?> task; |
1279 |
– |
while ((task = pollSubmission()) != null) |
1280 |
– |
task.cancel(false); |
1281 |
– |
} |
1282 |
– |
|
1283 |
– |
/** |
1284 |
– |
* Cleans out worker queues. |
1285 |
– |
*/ |
1286 |
– |
private void cancelQueuedWorkerTasks() { |
1287 |
– |
final ReentrantLock lock = this.workerLock; |
1288 |
– |
lock.lock(); |
1707 |
|
try { |
1708 |
< |
ForkJoinWorkerThread[] ws = workers; |
1709 |
< |
if (ws != null) { |
1292 |
< |
for (int i = 0; i < ws.length; ++i) { |
1293 |
< |
ForkJoinWorkerThread t = ws[i]; |
1294 |
< |
if (t != null) |
1295 |
< |
t.cancelTasks(); |
1296 |
< |
} |
1297 |
< |
} |
1298 |
< |
} finally { |
1299 |
< |
lock.unlock(); |
1300 |
< |
} |
1301 |
< |
} |
1302 |
< |
|
1303 |
< |
/** |
1304 |
< |
* Sets each worker's status to terminating. Requires lock to avoid |
1305 |
< |
* conflicts with add/remove. |
1306 |
< |
*/ |
1307 |
< |
private void stopAllWorkers() { |
1308 |
< |
final ReentrantLock lock = this.workerLock; |
1309 |
< |
lock.lock(); |
1310 |
< |
try { |
1311 |
< |
ForkJoinWorkerThread[] ws = workers; |
1312 |
< |
if (ws != null) { |
1313 |
< |
for (int i = 0; i < ws.length; ++i) { |
1314 |
< |
ForkJoinWorkerThread t = ws[i]; |
1315 |
< |
if (t != null) |
1316 |
< |
t.shutdownNow(); |
1317 |
< |
} |
1318 |
< |
} |
1319 |
< |
} finally { |
1320 |
< |
lock.unlock(); |
1321 |
< |
} |
1322 |
< |
} |
1323 |
< |
|
1324 |
< |
/** |
1325 |
< |
* Interrupts all unterminated workers. This is not required for |
1326 |
< |
* sake of internal control, but may help unstick user code during |
1327 |
< |
* shutdown. |
1328 |
< |
*/ |
1329 |
< |
private void interruptUnterminatedWorkers() { |
1330 |
< |
final ReentrantLock lock = this.workerLock; |
1331 |
< |
lock.lock(); |
1332 |
< |
try { |
1333 |
< |
ForkJoinWorkerThread[] ws = workers; |
1334 |
< |
if (ws != null) { |
1335 |
< |
for (int i = 0; i < ws.length; ++i) { |
1336 |
< |
ForkJoinWorkerThread t = ws[i]; |
1337 |
< |
if (t != null && !t.isTerminated()) { |
1338 |
< |
try { |
1339 |
< |
t.interrupt(); |
1340 |
< |
} catch (SecurityException ignore) { |
1341 |
< |
} |
1342 |
< |
} |
1343 |
< |
} |
1344 |
< |
} |
1345 |
< |
} finally { |
1346 |
< |
lock.unlock(); |
1347 |
< |
} |
1348 |
< |
} |
1349 |
< |
|
1350 |
< |
|
1351 |
< |
/* |
1352 |
< |
* Nodes for event barrier to manage idle threads. Queue nodes |
1353 |
< |
* are basic Treiber stack nodes, also used for spare stack. |
1354 |
< |
* |
1355 |
< |
* The event barrier has an event count and a wait queue (actually |
1356 |
< |
* a Treiber stack). Workers are enabled to look for work when |
1357 |
< |
* the eventCount is incremented. If they fail to find work, they |
1358 |
< |
* may wait for next count. Upon release, threads help others wake |
1359 |
< |
* up. |
1360 |
< |
* |
1361 |
< |
* Synchronization events occur only in enough contexts to |
1362 |
< |
* maintain overall liveness: |
1363 |
< |
* |
1364 |
< |
* - Submission of a new task to the pool |
1365 |
< |
* - Resizes or other changes to the workers array |
1366 |
< |
* - pool termination |
1367 |
< |
* - A worker pushing a task on an empty queue |
1368 |
< |
* |
1369 |
< |
* The case of pushing a task occurs often enough, and is heavy |
1370 |
< |
* enough compared to simple stack pushes, to require special |
1371 |
< |
* handling: Method signalWork returns without advancing count if |
1372 |
< |
* the queue appears to be empty. This would ordinarily result in |
1373 |
< |
* races causing some queued waiters not to be woken up. To avoid |
1374 |
< |
* this, the first worker enqueued in method sync (see |
1375 |
< |
* syncIsReleasable) rescans for tasks after being enqueued, and |
1376 |
< |
* helps signal if any are found. This works well because the |
1377 |
< |
* worker has nothing better to do, and so might as well help |
1378 |
< |
* alleviate the overhead and contention on the threads actually |
1379 |
< |
* doing work. Also, since event counts increments on task |
1380 |
< |
* availability exist to maintain liveness (rather than to force |
1381 |
< |
* refreshes etc), it is OK for callers to exit early if |
1382 |
< |
* contending with another signaller. |
1383 |
< |
*/ |
1384 |
< |
static final class WaitQueueNode { |
1385 |
< |
WaitQueueNode next; // only written before enqueued |
1386 |
< |
volatile ForkJoinWorkerThread thread; // nulled to cancel wait |
1387 |
< |
final long count; // unused for spare stack |
1388 |
< |
|
1389 |
< |
WaitQueueNode(long c, ForkJoinWorkerThread w) { |
1390 |
< |
count = c; |
1391 |
< |
thread = w; |
1392 |
< |
} |
1393 |
< |
|
1394 |
< |
/** |
1395 |
< |
* Wakes up waiter, returning false if known to already |
1396 |
< |
*/ |
1397 |
< |
boolean signal() { |
1398 |
< |
ForkJoinWorkerThread t = thread; |
1399 |
< |
if (t == null) |
1400 |
< |
return false; |
1401 |
< |
thread = null; |
1402 |
< |
LockSupport.unpark(t); |
1403 |
< |
return true; |
1404 |
< |
} |
1405 |
< |
|
1406 |
< |
/** |
1407 |
< |
* Awaits release on sync. |
1408 |
< |
*/ |
1409 |
< |
void awaitSyncRelease(ForkJoinPool p) { |
1410 |
< |
while (thread != null && !p.syncIsReleasable(this)) |
1411 |
< |
LockSupport.park(this); |
1412 |
< |
} |
1413 |
< |
|
1414 |
< |
/** |
1415 |
< |
* Awaits resumption as spare. |
1416 |
< |
*/ |
1417 |
< |
void awaitSpareRelease() { |
1418 |
< |
while (thread != null) { |
1419 |
< |
if (!Thread.interrupted()) |
1420 |
< |
LockSupport.park(this); |
1421 |
< |
} |
1422 |
< |
} |
1423 |
< |
} |
1424 |
< |
|
1425 |
< |
/** |
1426 |
< |
* Ensures that no thread is waiting for count to advance from the |
1427 |
< |
* current value of eventCount read on entry to this method, by |
1428 |
< |
* releasing waiting threads if necessary. |
1429 |
< |
* |
1430 |
< |
* @return the count |
1431 |
< |
*/ |
1432 |
< |
final long ensureSync() { |
1433 |
< |
long c = eventCount; |
1434 |
< |
WaitQueueNode q; |
1435 |
< |
while ((q = syncStack) != null && q.count < c) { |
1436 |
< |
if (casBarrierStack(q, null)) { |
1437 |
< |
do { |
1438 |
< |
q.signal(); |
1439 |
< |
} while ((q = q.next) != null); |
1440 |
< |
break; |
1441 |
< |
} |
1442 |
< |
} |
1443 |
< |
return c; |
1444 |
< |
} |
1445 |
< |
|
1446 |
< |
/** |
1447 |
< |
* Increments event count and releases waiting threads. |
1448 |
< |
*/ |
1449 |
< |
private void signalIdleWorkers() { |
1450 |
< |
long c; |
1451 |
< |
do {} while (!casEventCount(c = eventCount, c+1)); |
1452 |
< |
ensureSync(); |
1453 |
< |
} |
1454 |
< |
|
1455 |
< |
/** |
1456 |
< |
* Signals threads waiting to poll a task. Because method sync |
1457 |
< |
* rechecks availability, it is OK to only proceed if queue |
1458 |
< |
* appears to be non-empty, and OK to skip under contention to |
1459 |
< |
* increment count (since some other thread succeeded). |
1460 |
< |
*/ |
1461 |
< |
final void signalWork() { |
1462 |
< |
long c; |
1463 |
< |
WaitQueueNode q; |
1464 |
< |
if (syncStack != null && |
1465 |
< |
casEventCount(c = eventCount, c+1) && |
1466 |
< |
(((q = syncStack) != null && q.count <= c) && |
1467 |
< |
(!casBarrierStack(q, q.next) || !q.signal()))) |
1468 |
< |
ensureSync(); |
1469 |
< |
} |
1470 |
< |
|
1471 |
< |
/** |
1472 |
< |
* Waits until event count advances from last value held by |
1473 |
< |
* caller, or if excess threads, caller is resumed as spare, or |
1474 |
< |
* caller or pool is terminating. Updates caller's event on exit. |
1475 |
< |
* |
1476 |
< |
* @param w the calling worker thread |
1477 |
< |
*/ |
1478 |
< |
final void sync(ForkJoinWorkerThread w) { |
1479 |
< |
updateStealCount(w); // Transfer w's count while it is idle |
1480 |
< |
|
1481 |
< |
while (!w.isShutdown() && isProcessingTasks() && !suspendIfSpare(w)) { |
1482 |
< |
long prev = w.lastEventCount; |
1483 |
< |
WaitQueueNode node = null; |
1484 |
< |
WaitQueueNode h; |
1485 |
< |
while (eventCount == prev && |
1486 |
< |
((h = syncStack) == null || h.count == prev)) { |
1487 |
< |
if (node == null) |
1488 |
< |
node = new WaitQueueNode(prev, w); |
1489 |
< |
if (casBarrierStack(node.next = h, node)) { |
1490 |
< |
node.awaitSyncRelease(this); |
1491 |
< |
break; |
1492 |
< |
} |
1493 |
< |
} |
1494 |
< |
long ec = ensureSync(); |
1495 |
< |
if (ec != prev) { |
1496 |
< |
w.lastEventCount = ec; |
1497 |
< |
break; |
1498 |
< |
} |
1499 |
< |
} |
1500 |
< |
} |
1501 |
< |
|
1502 |
< |
/** |
1503 |
< |
* Returns {@code true} if worker waiting on sync can proceed: |
1504 |
< |
* - on signal (thread == null) |
1505 |
< |
* - on event count advance (winning race to notify vs signaller) |
1506 |
< |
* - on interrupt |
1507 |
< |
* - if the first queued node, we find work available |
1508 |
< |
* If node was not signalled and event count not advanced on exit, |
1509 |
< |
* then we also help advance event count. |
1510 |
< |
* |
1511 |
< |
* @return {@code true} if node can be released |
1512 |
< |
*/ |
1513 |
< |
final boolean syncIsReleasable(WaitQueueNode node) { |
1514 |
< |
long prev = node.count; |
1515 |
< |
if (!Thread.interrupted() && node.thread != null && |
1516 |
< |
(node.next != null || |
1517 |
< |
!ForkJoinWorkerThread.hasQueuedTasks(workers)) && |
1518 |
< |
eventCount == prev) |
1708 |
> |
return termination.awaitAdvanceInterruptibly(0, timeout, unit) > 0; |
1709 |
> |
} catch(TimeoutException ex) { |
1710 |
|
return false; |
1520 |
– |
if (node.thread != null) { |
1521 |
– |
node.thread = null; |
1522 |
– |
long ec = eventCount; |
1523 |
– |
if (prev <= ec) // help signal |
1524 |
– |
casEventCount(ec, ec+1); |
1525 |
– |
} |
1526 |
– |
return true; |
1527 |
– |
} |
1528 |
– |
|
1529 |
– |
/** |
1530 |
– |
* Returns {@code true} if a new sync event occurred since last |
1531 |
– |
* call to sync or this method, if so, updating caller's count. |
1532 |
– |
*/ |
1533 |
– |
final boolean hasNewSyncEvent(ForkJoinWorkerThread w) { |
1534 |
– |
long lc = w.lastEventCount; |
1535 |
– |
long ec = ensureSync(); |
1536 |
– |
if (ec == lc) |
1537 |
– |
return false; |
1538 |
– |
w.lastEventCount = ec; |
1539 |
– |
return true; |
1540 |
– |
} |
1541 |
– |
|
1542 |
– |
// Parallelism maintenance |
1543 |
– |
|
1544 |
– |
/** |
1545 |
– |
* Decrements running count; if too low, adds spare. |
1546 |
– |
* |
1547 |
– |
* Conceptually, all we need to do here is add or resume a |
1548 |
– |
* spare thread when one is about to block (and remove or |
1549 |
– |
* suspend it later when unblocked -- see suspendIfSpare). |
1550 |
– |
* However, implementing this idea requires coping with |
1551 |
– |
* several problems: we have imperfect information about the |
1552 |
– |
* states of threads. Some count updates can and usually do |
1553 |
– |
* lag run state changes, despite arrangements to keep them |
1554 |
– |
* accurate (for example, when possible, updating counts |
1555 |
– |
* before signalling or resuming), especially when running on |
1556 |
– |
* dynamic JVMs that don't optimize the infrequent paths that |
1557 |
– |
* update counts. Generating too many threads can make these |
1558 |
– |
* problems become worse, because excess threads are more |
1559 |
– |
* likely to be context-switched with others, slowing them all |
1560 |
– |
* down, especially if there is no work available, so all are |
1561 |
– |
* busy scanning or idling. Also, excess spare threads can |
1562 |
– |
* only be suspended or removed when they are idle, not |
1563 |
– |
* immediately when they aren't needed. So adding threads will |
1564 |
– |
* raise parallelism level for longer than necessary. Also, |
1565 |
– |
* FJ applications often encounter highly transient peaks when |
1566 |
– |
* many threads are blocked joining, but for less time than it |
1567 |
– |
* takes to create or resume spares. |
1568 |
– |
* |
1569 |
– |
* @param joinMe if non-null, return early if done |
1570 |
– |
* @param maintainParallelism if true, try to stay within |
1571 |
– |
* target counts, else create only to avoid starvation |
1572 |
– |
* @return true if joinMe known to be done |
1573 |
– |
*/ |
1574 |
– |
final boolean preJoin(ForkJoinTask<?> joinMe, |
1575 |
– |
boolean maintainParallelism) { |
1576 |
– |
maintainParallelism &= maintainsParallelism; // overrride |
1577 |
– |
boolean dec = false; // true when running count decremented |
1578 |
– |
while (spareStack == null || !tryResumeSpare(dec)) { |
1579 |
– |
int counts = workerCounts; |
1580 |
– |
if (dec || (dec = casWorkerCounts(counts, --counts))) { |
1581 |
– |
// CAS cheat |
1582 |
– |
if (!needSpare(counts, maintainParallelism)) |
1583 |
– |
break; |
1584 |
– |
if (joinMe.status < 0) |
1585 |
– |
return true; |
1586 |
– |
if (tryAddSpare(counts)) |
1587 |
– |
break; |
1588 |
– |
} |
1589 |
– |
} |
1590 |
– |
return false; |
1591 |
– |
} |
1592 |
– |
|
1593 |
– |
/** |
1594 |
– |
* Same idea as preJoin |
1595 |
– |
*/ |
1596 |
– |
final boolean preBlock(ManagedBlocker blocker, |
1597 |
– |
boolean maintainParallelism) { |
1598 |
– |
maintainParallelism &= maintainsParallelism; |
1599 |
– |
boolean dec = false; |
1600 |
– |
while (spareStack == null || !tryResumeSpare(dec)) { |
1601 |
– |
int counts = workerCounts; |
1602 |
– |
if (dec || (dec = casWorkerCounts(counts, --counts))) { |
1603 |
– |
if (!needSpare(counts, maintainParallelism)) |
1604 |
– |
break; |
1605 |
– |
if (blocker.isReleasable()) |
1606 |
– |
return true; |
1607 |
– |
if (tryAddSpare(counts)) |
1608 |
– |
break; |
1609 |
– |
} |
1610 |
– |
} |
1611 |
– |
return false; |
1612 |
– |
} |
1613 |
– |
|
1614 |
– |
/** |
1615 |
– |
* Returns {@code true} if a spare thread appears to be needed. |
1616 |
– |
* If maintaining parallelism, returns true when the deficit in |
1617 |
– |
* running threads is more than the surplus of total threads, and |
1618 |
– |
* there is apparently some work to do. This self-limiting rule |
1619 |
– |
* means that the more threads that have already been added, the |
1620 |
– |
* less parallelism we will tolerate before adding another. |
1621 |
– |
* |
1622 |
– |
* @param counts current worker counts |
1623 |
– |
* @param maintainParallelism try to maintain parallelism |
1624 |
– |
*/ |
1625 |
– |
private boolean needSpare(int counts, boolean maintainParallelism) { |
1626 |
– |
int ps = parallelism; |
1627 |
– |
int rc = runningCountOf(counts); |
1628 |
– |
int tc = totalCountOf(counts); |
1629 |
– |
int runningDeficit = ps - rc; |
1630 |
– |
int totalSurplus = tc - ps; |
1631 |
– |
return (tc < maxPoolSize && |
1632 |
– |
(rc == 0 || totalSurplus < 0 || |
1633 |
– |
(maintainParallelism && |
1634 |
– |
runningDeficit > totalSurplus && |
1635 |
– |
ForkJoinWorkerThread.hasQueuedTasks(workers)))); |
1636 |
– |
} |
1637 |
– |
|
1638 |
– |
/** |
1639 |
– |
* Adds a spare worker if lock available and no more than the |
1640 |
– |
* expected numbers of threads exist. |
1641 |
– |
* |
1642 |
– |
* @return true if successful |
1643 |
– |
*/ |
1644 |
– |
private boolean tryAddSpare(int expectedCounts) { |
1645 |
– |
final ReentrantLock lock = this.workerLock; |
1646 |
– |
int expectedRunning = runningCountOf(expectedCounts); |
1647 |
– |
int expectedTotal = totalCountOf(expectedCounts); |
1648 |
– |
boolean success = false; |
1649 |
– |
boolean locked = false; |
1650 |
– |
// confirm counts while locking; CAS after obtaining lock |
1651 |
– |
try { |
1652 |
– |
for (;;) { |
1653 |
– |
int s = workerCounts; |
1654 |
– |
int tc = totalCountOf(s); |
1655 |
– |
int rc = runningCountOf(s); |
1656 |
– |
if (rc > expectedRunning || tc > expectedTotal) |
1657 |
– |
break; |
1658 |
– |
if (!locked && !(locked = lock.tryLock())) |
1659 |
– |
break; |
1660 |
– |
if (casWorkerCounts(s, workerCountsFor(tc+1, rc+1))) { |
1661 |
– |
createAndStartSpare(tc); |
1662 |
– |
success = true; |
1663 |
– |
break; |
1664 |
– |
} |
1665 |
– |
} |
1666 |
– |
} finally { |
1667 |
– |
if (locked) |
1668 |
– |
lock.unlock(); |
1669 |
– |
} |
1670 |
– |
return success; |
1671 |
– |
} |
1672 |
– |
|
1673 |
– |
/** |
1674 |
– |
* Adds the kth spare worker. On entry, pool counts are already |
1675 |
– |
* adjusted to reflect addition. |
1676 |
– |
*/ |
1677 |
– |
private void createAndStartSpare(int k) { |
1678 |
– |
ForkJoinWorkerThread w = null; |
1679 |
– |
ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(k + 1); |
1680 |
– |
int len = ws.length; |
1681 |
– |
// Probably, we can place at slot k. If not, find empty slot |
1682 |
– |
if (k < len && ws[k] != null) { |
1683 |
– |
for (k = 0; k < len && ws[k] != null; ++k) |
1684 |
– |
; |
1685 |
– |
} |
1686 |
– |
if (k < len && isProcessingTasks() && (w = createWorker(k)) != null) { |
1687 |
– |
ws[k] = w; |
1688 |
– |
w.start(); |
1689 |
– |
} |
1690 |
– |
else |
1691 |
– |
updateWorkerCount(-1); // adjust on failure |
1692 |
– |
signalIdleWorkers(); |
1693 |
– |
} |
1694 |
– |
|
1695 |
– |
/** |
1696 |
– |
* Suspends calling thread w if there are excess threads. Called |
1697 |
– |
* only from sync. Spares are enqueued in a Treiber stack using |
1698 |
– |
* the same WaitQueueNodes as barriers. They are resumed mainly |
1699 |
– |
* in preJoin, but are also woken on pool events that require all |
1700 |
– |
* threads to check run state. |
1701 |
– |
* |
1702 |
– |
* @param w the caller |
1703 |
– |
*/ |
1704 |
– |
private boolean suspendIfSpare(ForkJoinWorkerThread w) { |
1705 |
– |
WaitQueueNode node = null; |
1706 |
– |
int s; |
1707 |
– |
while (parallelism < runningCountOf(s = workerCounts)) { |
1708 |
– |
if (node == null) |
1709 |
– |
node = new WaitQueueNode(0, w); |
1710 |
– |
if (casWorkerCounts(s, s-1)) { // representation-dependent |
1711 |
– |
// push onto stack |
1712 |
– |
do {} while (!casSpareStack(node.next = spareStack, node)); |
1713 |
– |
// block until released by resumeSpare |
1714 |
– |
node.awaitSpareRelease(); |
1715 |
– |
return true; |
1716 |
– |
} |
1717 |
– |
} |
1718 |
– |
return false; |
1719 |
– |
} |
1720 |
– |
|
1721 |
– |
/** |
1722 |
– |
* Tries to pop and resume a spare thread. |
1723 |
– |
* |
1724 |
– |
* @param updateCount if true, increment running count on success |
1725 |
– |
* @return true if successful |
1726 |
– |
*/ |
1727 |
– |
private boolean tryResumeSpare(boolean updateCount) { |
1728 |
– |
WaitQueueNode q; |
1729 |
– |
while ((q = spareStack) != null) { |
1730 |
– |
if (casSpareStack(q, q.next)) { |
1731 |
– |
if (updateCount) |
1732 |
– |
updateRunningCount(1); |
1733 |
– |
q.signal(); |
1734 |
– |
return true; |
1735 |
– |
} |
1736 |
– |
} |
1737 |
– |
return false; |
1738 |
– |
} |
1739 |
– |
|
1740 |
– |
/** |
1741 |
– |
* Pops and resumes all spare threads. Same idea as ensureSync. |
1742 |
– |
* |
1743 |
– |
* @return true if any spares released |
1744 |
– |
*/ |
1745 |
– |
private boolean resumeAllSpares() { |
1746 |
– |
WaitQueueNode q; |
1747 |
– |
while ( (q = spareStack) != null) { |
1748 |
– |
if (casSpareStack(q, null)) { |
1749 |
– |
do { |
1750 |
– |
updateRunningCount(1); |
1751 |
– |
q.signal(); |
1752 |
– |
} while ((q = q.next) != null); |
1753 |
– |
return true; |
1754 |
– |
} |
1755 |
– |
} |
1756 |
– |
return false; |
1757 |
– |
} |
1758 |
– |
|
1759 |
– |
/** |
1760 |
– |
* Pops and shuts down excessive spare threads. Call only while |
1761 |
– |
* holding lock. This is not guaranteed to eliminate all excess |
1762 |
– |
* threads, only those suspended as spares, which are the ones |
1763 |
– |
* unlikely to be needed in the future. |
1764 |
– |
*/ |
1765 |
– |
private void trimSpares() { |
1766 |
– |
int surplus = totalCountOf(workerCounts) - parallelism; |
1767 |
– |
WaitQueueNode q; |
1768 |
– |
while (surplus > 0 && (q = spareStack) != null) { |
1769 |
– |
if (casSpareStack(q, null)) { |
1770 |
– |
do { |
1771 |
– |
updateRunningCount(1); |
1772 |
– |
ForkJoinWorkerThread w = q.thread; |
1773 |
– |
if (w != null && surplus > 0 && |
1774 |
– |
runningCountOf(workerCounts) > 0 && w.shutdown()) |
1775 |
– |
--surplus; |
1776 |
– |
q.signal(); |
1777 |
– |
} while ((q = q.next) != null); |
1778 |
– |
} |
1711 |
|
} |
1712 |
|
} |
1713 |
|
|
1715 |
|
* Interface for extending managed parallelism for tasks running |
1716 |
|
* in {@link ForkJoinPool}s. |
1717 |
|
* |
1718 |
< |
* <p>A {@code ManagedBlocker} provides two methods. |
1719 |
< |
* Method {@code isReleasable} must return {@code true} if |
1720 |
< |
* blocking is not necessary. Method {@code block} blocks the |
1721 |
< |
* current thread if necessary (perhaps internally invoking |
1722 |
< |
* {@code isReleasable} before actually blocking). |
1718 |
> |
* <p>A {@code ManagedBlocker} provides two methods. Method |
1719 |
> |
* {@code isReleasable} must return {@code true} if blocking is |
1720 |
> |
* not necessary. Method {@code block} blocks the current thread |
1721 |
> |
* if necessary (perhaps internally invoking {@code isReleasable} |
1722 |
> |
* before actually blocking). The unusual methods in this API |
1723 |
> |
* accommodate synchronizers that may, but don't usually, block |
1724 |
> |
* for long periods. Similarly, they allow more efficient internal |
1725 |
> |
* handling of cases in which additional workers may be, but |
1726 |
> |
* usually are not, needed to ensure sufficient parallelism. |
1727 |
> |
* Toward this end, implementations of method {@code isReleasable} |
1728 |
> |
* must be amenable to repeated invocation. |
1729 |
|
* |
1730 |
|
* <p>For example, here is a ManagedBlocker based on a |
1731 |
|
* ReentrantLock: |
1743 |
|
* return hasLock || (hasLock = lock.tryLock()); |
1744 |
|
* } |
1745 |
|
* }}</pre> |
1746 |
+ |
* |
1747 |
+ |
* <p>Here is a class that possibly blocks waiting for an |
1748 |
+ |
* item on a given queue: |
1749 |
+ |
* <pre> {@code |
1750 |
+ |
* class QueueTaker<E> implements ManagedBlocker { |
1751 |
+ |
* final BlockingQueue<E> queue; |
1752 |
+ |
* volatile E item = null; |
1753 |
+ |
* QueueTaker(BlockingQueue<E> q) { this.queue = q; } |
1754 |
+ |
* public boolean block() throws InterruptedException { |
1755 |
+ |
* if (item == null) |
1756 |
+ |
* item = queue.take |
1757 |
+ |
* return true; |
1758 |
+ |
* } |
1759 |
+ |
* public boolean isReleasable() { |
1760 |
+ |
* return item != null || (item = queue.poll) != null; |
1761 |
+ |
* } |
1762 |
+ |
* public E getItem() { // call after pool.managedBlock completes |
1763 |
+ |
* return item; |
1764 |
+ |
* } |
1765 |
+ |
* }}</pre> |
1766 |
|
*/ |
1767 |
|
public static interface ManagedBlocker { |
1768 |
|
/** |
1786 |
|
* Blocks in accord with the given blocker. If the current thread |
1787 |
|
* is a {@link ForkJoinWorkerThread}, this method possibly |
1788 |
|
* arranges for a spare thread to be activated if necessary to |
1789 |
< |
* ensure parallelism while the current thread is blocked. |
1832 |
< |
* |
1833 |
< |
* <p>If {@code maintainParallelism} is {@code true} and the pool |
1834 |
< |
* supports it ({@link #getMaintainsParallelism}), this method |
1835 |
< |
* attempts to maintain the pool's nominal parallelism. Otherwise |
1836 |
< |
* it activates a thread only if necessary to avoid complete |
1837 |
< |
* starvation. This option may be preferable when blockages use |
1838 |
< |
* timeouts, or are almost always brief. |
1789 |
> |
* ensure sufficient parallelism while the current thread is blocked. |
1790 |
|
* |
1791 |
|
* <p>If the caller is not a {@link ForkJoinTask}, this method is |
1792 |
|
* behaviorally equivalent to |
1800 |
|
* first be expanded to ensure parallelism, and later adjusted. |
1801 |
|
* |
1802 |
|
* @param blocker the blocker |
1852 |
– |
* @param maintainParallelism if {@code true} and supported by |
1853 |
– |
* this pool, attempt to maintain the pool's nominal parallelism; |
1854 |
– |
* otherwise activate a thread only if necessary to avoid |
1855 |
– |
* complete starvation. |
1803 |
|
* @throws InterruptedException if blocker.block did so |
1804 |
|
*/ |
1805 |
< |
public static void managedBlock(ManagedBlocker blocker, |
1859 |
< |
boolean maintainParallelism) |
1805 |
> |
public static void managedBlock(ManagedBlocker blocker) |
1806 |
|
throws InterruptedException { |
1807 |
|
Thread t = Thread.currentThread(); |
1808 |
< |
ForkJoinPool pool = ((t instanceof ForkJoinWorkerThread) ? |
1809 |
< |
((ForkJoinWorkerThread) t).pool : null); |
1810 |
< |
if (!blocker.isReleasable()) { |
1811 |
< |
try { |
1812 |
< |
if (pool == null || |
1813 |
< |
!pool.preBlock(blocker, maintainParallelism)) |
1868 |
< |
awaitBlocker(blocker); |
1869 |
< |
} finally { |
1870 |
< |
if (pool != null) |
1871 |
< |
pool.updateRunningCount(1); |
1872 |
< |
} |
1808 |
> |
if (t instanceof ForkJoinWorkerThread) { |
1809 |
> |
ForkJoinWorkerThread w = (ForkJoinWorkerThread) t; |
1810 |
> |
w.pool.awaitBlocker(blocker); |
1811 |
> |
} |
1812 |
> |
else { |
1813 |
> |
do {} while (!blocker.isReleasable() && !blocker.block()); |
1814 |
|
} |
1874 |
– |
} |
1875 |
– |
|
1876 |
– |
private static void awaitBlocker(ManagedBlocker blocker) |
1877 |
– |
throws InterruptedException { |
1878 |
– |
do {} while (!blocker.isReleasable() && !blocker.block()); |
1815 |
|
} |
1816 |
|
|
1817 |
|
// AbstractExecutorService overrides. These rely on undocumented |
1829 |
|
// Unsafe mechanics |
1830 |
|
|
1831 |
|
private static final sun.misc.Unsafe UNSAFE = getUnsafe(); |
1896 |
– |
private static final long eventCountOffset = |
1897 |
– |
objectFieldOffset("eventCount", ForkJoinPool.class); |
1832 |
|
private static final long workerCountsOffset = |
1833 |
|
objectFieldOffset("workerCounts", ForkJoinPool.class); |
1834 |
< |
private static final long runControlOffset = |
1835 |
< |
objectFieldOffset("runControl", ForkJoinPool.class); |
1836 |
< |
private static final long syncStackOffset = |
1837 |
< |
objectFieldOffset("syncStack",ForkJoinPool.class); |
1838 |
< |
private static final long spareStackOffset = |
1839 |
< |
objectFieldOffset("spareStack", ForkJoinPool.class); |
1840 |
< |
|
1841 |
< |
private boolean casEventCount(long cmp, long val) { |
1842 |
< |
return UNSAFE.compareAndSwapLong(this, eventCountOffset, cmp, val); |
1843 |
< |
} |
1910 |
< |
private boolean casWorkerCounts(int cmp, int val) { |
1911 |
< |
return UNSAFE.compareAndSwapInt(this, workerCountsOffset, cmp, val); |
1912 |
< |
} |
1913 |
< |
private boolean casRunControl(int cmp, int val) { |
1914 |
< |
return UNSAFE.compareAndSwapInt(this, runControlOffset, cmp, val); |
1915 |
< |
} |
1916 |
< |
private boolean casSpareStack(WaitQueueNode cmp, WaitQueueNode val) { |
1917 |
< |
return UNSAFE.compareAndSwapObject(this, spareStackOffset, cmp, val); |
1918 |
< |
} |
1919 |
< |
private boolean casBarrierStack(WaitQueueNode cmp, WaitQueueNode val) { |
1920 |
< |
return UNSAFE.compareAndSwapObject(this, syncStackOffset, cmp, val); |
1921 |
< |
} |
1834 |
> |
private static final long runStateOffset = |
1835 |
> |
objectFieldOffset("runState", ForkJoinPool.class); |
1836 |
> |
private static final long eventCountOffset = |
1837 |
> |
objectFieldOffset("eventCount", ForkJoinPool.class); |
1838 |
> |
private static final long eventWaitersOffset = |
1839 |
> |
objectFieldOffset("eventWaiters",ForkJoinPool.class); |
1840 |
> |
private static final long stealCountOffset = |
1841 |
> |
objectFieldOffset("stealCount",ForkJoinPool.class); |
1842 |
> |
private static final long spareWaitersOffset = |
1843 |
> |
objectFieldOffset("spareWaiters",ForkJoinPool.class); |
1844 |
|
|
1845 |
|
private static long objectFieldOffset(String field, Class<?> klazz) { |
1846 |
|
try { |