1 /*
2 * Copyright (c) 2001, 2013, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #ifndef SHARE_VM_UTILITIES_TASKQUEUE_HPP
26 #define SHARE_VM_UTILITIES_TASKQUEUE_HPP
27
28 #include "memory/allocation.hpp"
29 #include "memory/allocation.inline.hpp"
30 #include "runtime/mutex.hpp"
31 #include "utilities/stack.hpp"
32 #ifdef TARGET_OS_ARCH_linux_x86
33 # include "orderAccess_linux_x86.inline.hpp"
34 #endif
35 #ifdef TARGET_OS_ARCH_linux_sparc
36 # include "orderAccess_linux_sparc.inline.hpp"
37 #endif
38 #ifdef TARGET_OS_ARCH_linux_zero
39 # include "orderAccess_linux_zero.inline.hpp"
40 #endif
41 #ifdef TARGET_OS_ARCH_solaris_x86
42 # include "orderAccess_solaris_x86.inline.hpp"
43 #endif
44 #ifdef TARGET_OS_ARCH_solaris_sparc
45 # include "orderAccess_solaris_sparc.inline.hpp"
46 #endif
47 #ifdef TARGET_OS_ARCH_windows_x86
48 # include "orderAccess_windows_x86.inline.hpp"
49 #endif
50 #ifdef TARGET_OS_ARCH_linux_arm
51 # include "orderAccess_linux_arm.inline.hpp"
52 #endif
53 #ifdef TARGET_OS_ARCH_linux_ppc
54 # include "orderAccess_linux_ppc.inline.hpp"
55 #endif
56 #ifdef TARGET_OS_ARCH_bsd_x86
57 # include "orderAccess_bsd_x86.inline.hpp"
58 #endif
59 #ifdef TARGET_OS_ARCH_bsd_zero
60 # include "orderAccess_bsd_zero.inline.hpp"
61 #endif
62
63 // Simple TaskQueue stats that are collected by default in debug builds.
64
65 #if !defined(TASKQUEUE_STATS) && defined(ASSERT)
66 #define TASKQUEUE_STATS 1
67 #elif !defined(TASKQUEUE_STATS)
68 #define TASKQUEUE_STATS 0
69 #endif
70
71 #if TASKQUEUE_STATS
72 #define TASKQUEUE_STATS_ONLY(code) code
73 #else
74 #define TASKQUEUE_STATS_ONLY(code)
75 #endif // TASKQUEUE_STATS
76
77 #if TASKQUEUE_STATS
78 class TaskQueueStats {
79 public:
80 enum StatId {
81 push, // number of taskqueue pushes
82 pop, // number of taskqueue pops
83 pop_slow, // subset of taskqueue pops that were done slow-path
84 steal_attempt, // number of taskqueue steal attempts
85 steal, // number of taskqueue steals
86 overflow, // number of overflow pushes
87 overflow_max_len, // max length of overflow stack
88 last_stat_id
89 };
90
91 public:
92 inline TaskQueueStats() { reset(); }
93
94 inline void record_push() { ++_stats[push]; }
95 inline void record_pop() { ++_stats[pop]; }
96 inline void record_pop_slow() { record_pop(); ++_stats[pop_slow]; }
97 inline void record_steal(bool success);
98 inline void record_overflow(size_t new_length);
99
100 TaskQueueStats & operator +=(const TaskQueueStats & addend);
101
102 inline size_t get(StatId id) const { return _stats[id]; }
103 inline const size_t* get() const { return _stats; }
104
105 inline void reset();
106
107 // Print the specified line of the header (does not include a line separator).
108 static void print_header(unsigned int line, outputStream* const stream = tty,
109 unsigned int width = 10);
110 // Print the statistics (does not include a line separator).
111 void print(outputStream* const stream = tty, unsigned int width = 10) const;
112
113 DEBUG_ONLY(void verify() const;)
114
115 private:
116 size_t _stats[last_stat_id];
117 static const char * const _names[last_stat_id];
118 };
119
120 void TaskQueueStats::record_steal(bool success) {
121 ++_stats[steal_attempt];
122 if (success) ++_stats[steal];
123 }
124
125 void TaskQueueStats::record_overflow(size_t new_len) {
126 ++_stats[overflow];
127 if (new_len > _stats[overflow_max_len]) _stats[overflow_max_len] = new_len;
128 }
129
130 void TaskQueueStats::reset() {
131 memset(_stats, 0, sizeof(_stats));
132 }
133 #endif // TASKQUEUE_STATS
134
135 // TaskQueueSuper collects functionality common to all GenericTaskQueue instances.
136
137 template <unsigned int N, MEMFLAGS F>
138 class TaskQueueSuper: public CHeapObj<F> {
139 protected:
140 // Internal type for indexing the queue; also used for the tag.
141 typedef NOT_LP64(uint16_t) LP64_ONLY(uint32_t) idx_t;
142
143 // The first free element after the last one pushed (mod N).
144 volatile uint _bottom;
145
146 enum { MOD_N_MASK = N - 1 };
147
148 class Age {
149 public:
150 Age(size_t data = 0) { _data = data; }
151 Age(const Age& age) { _data = age._data; }
152 Age(idx_t top, idx_t tag) { _fields._top = top; _fields._tag = tag; }
153
154 Age get() const volatile { return _data; }
155 void set(Age age) volatile { _data = age._data; }
156
157 idx_t top() const volatile { return _fields._top; }
158 idx_t tag() const volatile { return _fields._tag; }
159
160 // Increment top; if it wraps, increment tag also.
161 void increment() {
162 _fields._top = increment_index(_fields._top);
163 if (_fields._top == 0) ++_fields._tag;
164 }
165
166 Age cmpxchg(const Age new_age, const Age old_age) volatile {
167 return (size_t) Atomic::cmpxchg_ptr((intptr_t)new_age._data,
168 (volatile intptr_t *)&_data,
169 (intptr_t)old_age._data);
170 }
171
172 bool operator ==(const Age& other) const { return _data == other._data; }
173
174 private:
175 struct fields {
176 idx_t _top;
177 idx_t _tag;
178 };
179 union {
180 size_t _data;
181 fields _fields;
182 };
183 };
184
185 volatile Age _age;
186
187 // These both operate mod N.
188 static uint increment_index(uint ind) {
189 return (ind + 1) & MOD_N_MASK;
190 }
191 static uint decrement_index(uint ind) {
192 return (ind - 1) & MOD_N_MASK;
193 }
194
195 // Returns a number in the range [0..N). If the result is "N-1", it should be
196 // interpreted as 0.
197 uint dirty_size(uint bot, uint top) const {
198 return (bot - top) & MOD_N_MASK;
199 }
200
201 // Returns the size corresponding to the given "bot" and "top".
202 uint size(uint bot, uint top) const {
203 uint sz = dirty_size(bot, top);
204 // Has the queue "wrapped", so that bottom is less than top? There's a
205 // complicated special case here. A pair of threads could perform pop_local
206 // and pop_global operations concurrently, starting from a state in which
207 // _bottom == _top+1. The pop_local could succeed in decrementing _bottom,
208 // and the pop_global in incrementing _top (in which case the pop_global
209 // will be awarded the contested queue element.) The resulting state must
210 // be interpreted as an empty queue. (We only need to worry about one such
211 // event: only the queue owner performs pop_local's, and several concurrent
212 // threads attempting to perform the pop_global will all perform the same
213 // CAS, and only one can succeed.) Any stealing thread that reads after
214 // either the increment or decrement will see an empty queue, and will not
215 // join the competitors. The "sz == -1 || sz == N-1" state will not be
216 // modified by concurrent queues, so the owner thread can reset the state to
217 // _bottom == top so subsequent pushes will be performed normally.
218 return (sz == N - 1) ? 0 : sz;
219 }
220
221 public:
222 TaskQueueSuper() : _bottom(0), _age() {}
223
224 // Return true if the TaskQueue contains/does not contain any tasks.
225 bool peek() const { return _bottom != _age.top(); }
226 bool is_empty() const { return size() == 0; }
227
228 // Return an estimate of the number of elements in the queue.
229 // The "careful" version admits the possibility of pop_local/pop_global
230 // races.
231 uint size() const {
232 return size(_bottom, _age.top());
233 }
234
235 uint dirty_size() const {
236 return dirty_size(_bottom, _age.top());
237 }
238
239 void set_empty() {
240 _bottom = 0;
241 _age.set(0);
242 }
243
244 // Maximum number of elements allowed in the queue. This is two less
245 // than the actual queue size, for somewhat complicated reasons.
246 uint max_elems() const { return N - 2; }
247
248 // Total size of queue.
249 static const uint total_size() { return N; }
250
251 TASKQUEUE_STATS_ONLY(TaskQueueStats stats;)
252 };
253
254 //
255 // GenericTaskQueue implements an ABP, Aurora-Blumofe-Plaxton, double-
256 // ended-queue (deque), intended for use in work stealing. Queue operations
257 // are non-blocking.
258 //
259 // A queue owner thread performs push() and pop_local() operations on one end
260 // of the queue, while other threads may steal work using the pop_global()
261 // method.
262 //
263 // The main difference to the original algorithm is that this
264 // implementation allows wrap-around at the end of its allocated
265 // storage, which is an array.
266 //
267 // The original paper is:
268 //
269 // Arora, N. S., Blumofe, R. D., and Plaxton, C. G.
270 // Thread scheduling for multiprogrammed multiprocessors.
271 // Theory of Computing Systems 34, 2 (2001), 115-144.
272 //
273 // The following paper provides an correctness proof and an
274 // implementation for weakly ordered memory models including (pseudo-)
275 // code containing memory barriers for a Chase-Lev deque. Chase-Lev is
276 // similar to ABP, with the main difference that it allows resizing of the
277 // underlying storage:
278 //
279 // Le, N. M., Pop, A., Cohen A., and Nardell, F. Z.
280 // Correct and efficient work-stealing for weak memory models
281 // Proceedings of the 18th ACM SIGPLAN symposium on Principles and
282 // practice of parallel programming (PPoPP 2013), 69-80
283 //
284
285 template <class E, MEMFLAGS F, unsigned int N = TASKQUEUE_SIZE>
286 class GenericTaskQueue: public TaskQueueSuper<N, F> {
287 ArrayAllocator<E, F> _array_allocator;
288 protected:
289 typedef typename TaskQueueSuper<N, F>::Age Age;
290 typedef typename TaskQueueSuper<N, F>::idx_t idx_t;
291
292 using TaskQueueSuper<N, F>::_bottom;
293 using TaskQueueSuper<N, F>::_age;
294 using TaskQueueSuper<N, F>::increment_index;
295 using TaskQueueSuper<N, F>::decrement_index;
296 using TaskQueueSuper<N, F>::dirty_size;
297
298 public:
299 using TaskQueueSuper<N, F>::max_elems;
300 using TaskQueueSuper<N, F>::size;
301
302 #if TASKQUEUE_STATS
303 using TaskQueueSuper<N, F>::stats;
304 #endif
305
306 private:
307 // Slow paths for push, pop_local. (pop_global has no fast path.)
308 bool push_slow(E t, uint dirty_n_elems);
309 bool pop_local_slow(uint localBot, Age oldAge);
310
311 public:
312 typedef E element_type;
313
314 // Initializes the queue to empty.
315 GenericTaskQueue();
316
317 void initialize();
318
319 // Push the task "t" on the queue. Returns "false" iff the queue is full.
320 inline bool push(E t);
321
322 // Attempts to claim a task from the "local" end of the queue (the most
323 // recently pushed). If successful, returns true and sets t to the task;
324 // otherwise, returns false (the queue is empty).
325 inline bool pop_local(E& t);
326
327 // Like pop_local(), but uses the "global" end of the queue (the least
328 // recently pushed).
329 bool pop_global(E& t);
330
331 // Delete any resource associated with the queue.
332 ~GenericTaskQueue();
333
334 // apply the closure to all elements in the task queue
335 void oops_do(OopClosure* f);
336
337 private:
338 // Element array.
339 volatile E* _elems;
340 };
341
342 template<class E, MEMFLAGS F, unsigned int N>
343 GenericTaskQueue<E, F, N>::GenericTaskQueue() {
344 assert(sizeof(Age) == sizeof(size_t), "Depends on this.");
345 }
346
347 template<class E, MEMFLAGS F, unsigned int N>
348 void GenericTaskQueue<E, F, N>::initialize() {
349 _elems = _array_allocator.allocate(N);
350 }
351
352 template<class E, MEMFLAGS F, unsigned int N>
353 void GenericTaskQueue<E, F, N>::oops_do(OopClosure* f) {
354 // tty->print_cr("START OopTaskQueue::oops_do");
355 uint iters = size();
356 uint index = _bottom;
357 for (uint i = 0; i < iters; ++i) {
358 index = decrement_index(index);
359 // tty->print_cr(" doing entry %d," INTPTR_T " -> " INTPTR_T,
360 // index, &_elems[index], _elems[index]);
361 E* t = (E*)&_elems[index]; // cast away volatility
362 oop* p = (oop*)t;
363 assert((*t)->is_oop_or_null(), "Not an oop or null");
364 f->do_oop(p);
365 }
366 // tty->print_cr("END OopTaskQueue::oops_do");
367 }
368
369 template<class E, MEMFLAGS F, unsigned int N>
370 bool GenericTaskQueue<E, F, N>::push_slow(E t, uint dirty_n_elems) {
371 if (dirty_n_elems == N - 1) {
372 // Actually means 0, so do the push.
373 uint localBot = _bottom;
374 // g++ complains if the volatile result of the assignment is
375 // unused, so we cast the volatile away. We cannot cast directly
376 // to void, because gcc treats that as not using the result of the
377 // assignment. However, casting to E& means that we trigger an
378 // unused-value warning. So, we cast the E& to void.
379 (void)const_cast<E&>(_elems[localBot] = t);
380 OrderAccess::release_store(&_bottom, increment_index(localBot));
381 TASKQUEUE_STATS_ONLY(stats.record_push());
382 return true;
383 }
384 return false;
385 }
386
387 // pop_local_slow() is done by the owning thread and is trying to
388 // get the last task in the queue. It will compete with pop_global()
389 // that will be used by other threads. The tag age is incremented
390 // whenever the queue goes empty which it will do here if this thread
391 // gets the last task or in pop_global() if the queue wraps (top == 0
392 // and pop_global() succeeds, see pop_global()).
393 template<class E, MEMFLAGS F, unsigned int N>
394 bool GenericTaskQueue<E, F, N>::pop_local_slow(uint localBot, Age oldAge) {
395 // This queue was observed to contain exactly one element; either this
396 // thread will claim it, or a competing "pop_global". In either case,
397 // the queue will be logically empty afterwards. Create a new Age value
398 // that represents the empty queue for the given value of "_bottom". (We
399 // must also increment "tag" because of the case where "bottom == 1",
400 // "top == 0". A pop_global could read the queue element in that case,
401 // then have the owner thread do a pop followed by another push. Without
402 // the incrementing of "tag", the pop_global's CAS could succeed,
403 // allowing it to believe it has claimed the stale element.)
404 Age newAge((idx_t)localBot, oldAge.tag() + 1);
405 // Perhaps a competing pop_global has already incremented "top", in which
406 // case it wins the element.
407 if (localBot == oldAge.top()) {
408 // No competing pop_global has yet incremented "top"; we'll try to
409 // install new_age, thus claiming the element.
410 Age tempAge = _age.cmpxchg(newAge, oldAge);
411 if (tempAge == oldAge) {
412 // We win.
413 assert(dirty_size(localBot, _age.top()) != N - 1, "sanity");
414 TASKQUEUE_STATS_ONLY(stats.record_pop_slow());
415 return true;
416 }
417 }
418 // We lose; a completing pop_global gets the element. But the queue is empty
419 // and top is greater than bottom. Fix this representation of the empty queue
420 // to become the canonical one.
421 _age.set(newAge);
422 assert(dirty_size(localBot, _age.top()) != N - 1, "sanity");
423 return false;
424 }
425
426 template<class E, MEMFLAGS F, unsigned int N>
427 bool GenericTaskQueue<E, F, N>::pop_global(E& t) {
428 Age oldAge = _age.get();
429 // Architectures with weak memory model require a barrier here
430 // to guarantee that bottom is not older than age,
431 // which is crucial for the correctness of the algorithm.
432 #if !(defined SPARC || defined IA32 || defined AMD64)
433 OrderAccess::fence();
434 #endif
435 uint localBot = OrderAccess::load_acquire((volatile juint*)&_bottom);
436 uint n_elems = size(localBot, oldAge.top());
437 if (n_elems == 0) {
438 return false;
439 }
440
441 // g++ complains if the volatile result of the assignment is
442 // unused, so we cast the volatile away. We cannot cast directly
443 // to void, because gcc treats that as not using the result of the
444 // assignment. However, casting to E& means that we trigger an
445 // unused-value warning. So, we cast the E& to void.
446 (void) const_cast<E&>(t = _elems[oldAge.top()]);
447 Age newAge(oldAge);
448 newAge.increment();
449 Age resAge = _age.cmpxchg(newAge, oldAge);
450
451 // Note that using "_bottom" here might fail, since a pop_local might
452 // have decremented it.
453 assert(dirty_size(localBot, newAge.top()) != N - 1, "sanity");
454 return resAge == oldAge;
455 }
456
457 template<class E, MEMFLAGS F, unsigned int N>
458 GenericTaskQueue<E, F, N>::~GenericTaskQueue() {
459 FREE_C_HEAP_ARRAY(E, _elems, F);
460 }
461
462 // OverflowTaskQueue is a TaskQueue that also includes an overflow stack for
463 // elements that do not fit in the TaskQueue.
464 //
465 // This class hides two methods from super classes:
466 //
467 // push() - push onto the task queue or, if that fails, onto the overflow stack
468 // is_empty() - return true if both the TaskQueue and overflow stack are empty
469 //
470 // Note that size() is not hidden--it returns the number of elements in the
471 // TaskQueue, and does not include the size of the overflow stack. This
472 // simplifies replacement of GenericTaskQueues with OverflowTaskQueues.
473 template<class E, MEMFLAGS F, unsigned int N = TASKQUEUE_SIZE>
474 class OverflowTaskQueue: public GenericTaskQueue<E, F, N>
475 {
476 public:
477 typedef Stack<E, F> overflow_t;
478 typedef GenericTaskQueue<E, F, N> taskqueue_t;
479
480 TASKQUEUE_STATS_ONLY(using taskqueue_t::stats;)
481
482 // Push task t onto the queue or onto the overflow stack. Return true.
483 inline bool push(E t);
484
485 // Attempt to pop from the overflow stack; return true if anything was popped.
486 inline bool pop_overflow(E& t);
487
488 inline overflow_t* overflow_stack() { return &_overflow_stack; }
489
490 inline bool taskqueue_empty() const { return taskqueue_t::is_empty(); }
491 inline bool overflow_empty() const { return _overflow_stack.is_empty(); }
492 inline bool is_empty() const {
493 return taskqueue_empty() && overflow_empty();
494 }
495
496 private:
497 overflow_t _overflow_stack;
498 };
499
500 template <class E, MEMFLAGS F, unsigned int N>
501 bool OverflowTaskQueue<E, F, N>::push(E t)
502 {
503 if (!taskqueue_t::push(t)) {
504 overflow_stack()->push(t);
505 TASKQUEUE_STATS_ONLY(stats.record_overflow(overflow_stack()->size()));
506 }
507 return true;
508 }
509
510 template <class E, MEMFLAGS F, unsigned int N>
511 bool OverflowTaskQueue<E, F, N>::pop_overflow(E& t)
512 {
513 if (overflow_empty()) return false;
514 t = overflow_stack()->pop();
515 return true;
516 }
517
518 class TaskQueueSetSuper {
519 protected:
520 static int randomParkAndMiller(int* seed0);
521 public:
522 // Returns "true" if some TaskQueue in the set contains a task.
523 virtual bool peek() = 0;
524 };
525
526 template <MEMFLAGS F> class TaskQueueSetSuperImpl: public CHeapObj<F>, public TaskQueueSetSuper {
527 };
528
529 template<class T, MEMFLAGS F>
530 class GenericTaskQueueSet: public TaskQueueSetSuperImpl<F> {
531 private:
532 uint _n;
533 T** _queues;
534
535 public:
536 typedef typename T::element_type E;
537
538 GenericTaskQueueSet(int n) : _n(n) {
539 typedef T* GenericTaskQueuePtr;
540 _queues = NEW_C_HEAP_ARRAY(GenericTaskQueuePtr, n, F);
541 for (int i = 0; i < n; i++) {
542 _queues[i] = NULL;
543 }
544 }
545
546 bool steal_best_of_2(uint queue_num, int* seed, E& t);
547
548 void register_queue(uint i, T* q);
549
550 T* queue(uint n);
551
552 // The thread with queue number "queue_num" (and whose random number seed is
553 // at "seed") is trying to steal a task from some other queue. (It may try
554 // several queues, according to some configuration parameter.) If some steal
555 // succeeds, returns "true" and sets "t" to the stolen task, otherwise returns
556 // false.
557 bool steal(uint queue_num, int* seed, E& t);
558
559 bool peek();
560 };
561
562 template<class T, MEMFLAGS F> void
563 GenericTaskQueueSet<T, F>::register_queue(uint i, T* q) {
564 assert(i < _n, "index out of range.");
565 _queues[i] = q;
566 }
567
568 template<class T, MEMFLAGS F> T*
569 GenericTaskQueueSet<T, F>::queue(uint i) {
570 return _queues[i];
571 }
572
573 template<class T, MEMFLAGS F> bool
574 GenericTaskQueueSet<T, F>::steal(uint queue_num, int* seed, E& t) {
575 for (uint i = 0; i < 2 * _n; i++) {
576 if (steal_best_of_2(queue_num, seed, t)) {
577 TASKQUEUE_STATS_ONLY(queue(queue_num)->stats.record_steal(true));
578 return true;
579 }
580 }
581 TASKQUEUE_STATS_ONLY(queue(queue_num)->stats.record_steal(false));
582 return false;
583 }
584
585 template<class T, MEMFLAGS F> bool
586 GenericTaskQueueSet<T, F>::steal_best_of_2(uint queue_num, int* seed, E& t) {
587 if (_n > 2) {
588 uint k1 = queue_num;
589 while (k1 == queue_num) k1 = TaskQueueSetSuper::randomParkAndMiller(seed) % _n;
590 uint k2 = queue_num;
591 while (k2 == queue_num || k2 == k1) k2 = TaskQueueSetSuper::randomParkAndMiller(seed) % _n;
592 // Sample both and try the larger.
593 uint sz1 = _queues[k1]->size();
594 uint sz2 = _queues[k2]->size();
595 if (sz2 > sz1) return _queues[k2]->pop_global(t);
596 else return _queues[k1]->pop_global(t);
597 } else if (_n == 2) {
598 // Just try the other one.
599 uint k = (queue_num + 1) % 2;
600 return _queues[k]->pop_global(t);
601 } else {
602 assert(_n == 1, "can't be zero.");
603 return false;
604 }
605 }
606
607 template<class T, MEMFLAGS F>
608 bool GenericTaskQueueSet<T, F>::peek() {
609 // Try all the queues.
610 for (uint j = 0; j < _n; j++) {
611 if (_queues[j]->peek())
612 return true;
613 }
614 return false;
615 }
616
617 // When to terminate from the termination protocol.
618 class TerminatorTerminator: public CHeapObj<mtInternal> {
619 public:
620 virtual bool should_exit_termination() = 0;
621 };
622
623 // A class to aid in the termination of a set of parallel tasks using
624 // TaskQueueSet's for work stealing.
625
626 #undef TRACESPINNING
627
628 class ParallelTaskTerminator: public StackObj {
629 private:
630 int _n_threads;
631 TaskQueueSetSuper* _queue_set;
632 int _offered_termination;
633
634 #ifdef TRACESPINNING
635 static uint _total_yields;
636 static uint _total_spins;
637 static uint _total_peeks;
638 #endif
639
640 bool peek_in_queue_set();
641 protected:
642 virtual void yield();
643 void sleep(uint millis);
644
645 public:
646
647 // "n_threads" is the number of threads to be terminated. "queue_set" is a
648 // queue sets of work queues of other threads.
649 ParallelTaskTerminator(int n_threads, TaskQueueSetSuper* queue_set);
650
651 // The current thread has no work, and is ready to terminate if everyone
652 // else is. If returns "true", all threads are terminated. If returns
653 // "false", available work has been observed in one of the task queues,
654 // so the global task is not complete.
655 bool offer_termination() {
656 return offer_termination(NULL);
657 }
658
659 // As above, but it also terminates if the should_exit_termination()
660 // method of the terminator parameter returns true. If terminator is
661 // NULL, then it is ignored.
662 bool offer_termination(TerminatorTerminator* terminator);
663
664 // Reset the terminator, so that it may be reused again.
665 // The caller is responsible for ensuring that this is done
666 // in an MT-safe manner, once the previous round of use of
667 // the terminator is finished.
668 void reset_for_reuse();
669 // Same as above but the number of parallel threads is set to the
670 // given number.
671 void reset_for_reuse(int n_threads);
672
673 #ifdef TRACESPINNING
674 static uint total_yields() { return _total_yields; }
675 static uint total_spins() { return _total_spins; }
676 static uint total_peeks() { return _total_peeks; }
677 static void print_termination_counts();
678 #endif
679 };
680
681 template<class E, MEMFLAGS F, unsigned int N> inline bool
682 GenericTaskQueue<E, F, N>::push(E t) {
683 uint localBot = _bottom;
684 assert(localBot < N, "_bottom out of range.");
685 idx_t top = _age.top();
686 uint dirty_n_elems = dirty_size(localBot, top);
687 assert(dirty_n_elems < N, "n_elems out of range.");
688 if (dirty_n_elems < max_elems()) {
689 // g++ complains if the volatile result of the assignment is
690 // unused, so we cast the volatile away. We cannot cast directly
691 // to void, because gcc treats that as not using the result of the
692 // assignment. However, casting to E& means that we trigger an
693 // unused-value warning. So, we cast the E& to void.
694 (void) const_cast<E&>(_elems[localBot] = t);
695 OrderAccess::release_store(&_bottom, increment_index(localBot));
696 TASKQUEUE_STATS_ONLY(stats.record_push());
697 return true;
698 } else {
699 return push_slow(t, dirty_n_elems);
700 }
701 }
702
703 template<class E, MEMFLAGS F, unsigned int N> inline bool
704 GenericTaskQueue<E, F, N>::pop_local(E& t) {
705 uint localBot = _bottom;
706 // This value cannot be N-1. That can only occur as a result of
707 // the assignment to bottom in this method. If it does, this method
708 // resets the size to 0 before the next call (which is sequential,
709 // since this is pop_local.)
710 uint dirty_n_elems = dirty_size(localBot, _age.top());
711 assert(dirty_n_elems != N - 1, "Shouldn't be possible...");
712 if (dirty_n_elems == 0) return false;
713 localBot = decrement_index(localBot);
714 _bottom = localBot;
715 // This is necessary to prevent any read below from being reordered
716 // before the store just above.
717 OrderAccess::fence();
718 // g++ complains if the volatile result of the assignment is
719 // unused, so we cast the volatile away. We cannot cast directly
720 // to void, because gcc treats that as not using the result of the
721 // assignment. However, casting to E& means that we trigger an
722 // unused-value warning. So, we cast the E& to void.
723 (void) const_cast<E&>(t = _elems[localBot]);
724 // This is a second read of "age"; the "size()" above is the first.
725 // If there's still at least one element in the queue, based on the
726 // "_bottom" and "age" we've read, then there can be no interference with
727 // a "pop_global" operation, and we're done.
728 idx_t tp = _age.top(); // XXX
729 if (size(localBot, tp) > 0) {
730 assert(dirty_size(localBot, tp) != N - 1, "sanity");
731 TASKQUEUE_STATS_ONLY(stats.record_pop());
732 return true;
733 } else {
734 // Otherwise, the queue contained exactly one element; we take the slow
735 // path.
736 return pop_local_slow(localBot, _age.get());
737 }
738 }
739
740 typedef GenericTaskQueue<oop, mtGC> OopTaskQueue;
741 typedef GenericTaskQueueSet<OopTaskQueue, mtGC> OopTaskQueueSet;
742
743 #ifdef _MSC_VER
744 #pragma warning(push)
745 // warning C4522: multiple assignment operators specified
746 #pragma warning(disable:4522)
747 #endif
748
749 // This is a container class for either an oop* or a narrowOop*.
750 // Both are pushed onto a task queue and the consumer will test is_narrow()
751 // to determine which should be processed.
752 class StarTask {
753 void* _holder; // either union oop* or narrowOop*
754
755 enum { COMPRESSED_OOP_MASK = 1 };
756
757 public:
758 StarTask(narrowOop* p) {
759 assert(((uintptr_t)p & COMPRESSED_OOP_MASK) == 0, "Information loss!");
760 _holder = (void *)((uintptr_t)p | COMPRESSED_OOP_MASK);
761 }
762 StarTask(oop* p) {
763 assert(((uintptr_t)p & COMPRESSED_OOP_MASK) == 0, "Information loss!");
764 _holder = (void*)p;
765 }
766 StarTask() { _holder = NULL; }
767 operator oop*() { return (oop*)_holder; }
768 operator narrowOop*() {
769 return (narrowOop*)((uintptr_t)_holder & ~COMPRESSED_OOP_MASK);
770 }
771
772 StarTask& operator=(const StarTask& t) {
773 _holder = t._holder;
774 return *this;
775 }
776 volatile StarTask& operator=(const volatile StarTask& t) volatile {
777 _holder = t._holder;
778 return *this;
779 }
780
781 bool is_narrow() const {
782 return (((uintptr_t)_holder & COMPRESSED_OOP_MASK) != 0);
783 }
784 };
785
786 class ObjArrayTask
787 {
788 public:
789 ObjArrayTask(oop o = NULL, int idx = 0): _obj(o), _index(idx) { }
790 ObjArrayTask(oop o, size_t idx): _obj(o), _index(int(idx)) {
791 assert(idx <= size_t(max_jint), "too big");
792 }
793 ObjArrayTask(const ObjArrayTask& t): _obj(t._obj), _index(t._index) { }
794
795 ObjArrayTask& operator =(const ObjArrayTask& t) {
796 _obj = t._obj;
797 _index = t._index;
798 return *this;
799 }
800 volatile ObjArrayTask&
801 operator =(const volatile ObjArrayTask& t) volatile {
802 _obj = t._obj;
803 _index = t._index;
804 return *this;
805 }
806
807 inline oop obj() const { return _obj; }
808 inline int index() const { return _index; }
809
810 DEBUG_ONLY(bool is_valid() const); // Tasks to be pushed/popped must be valid.
811
812 private:
813 oop _obj;
814 int _index;
815 };
816
817 #ifdef _MSC_VER
818 #pragma warning(pop)
819 #endif
820
821 typedef OverflowTaskQueue<StarTask, mtClass> OopStarTaskQueue;
822 typedef GenericTaskQueueSet<OopStarTaskQueue, mtClass> OopStarTaskQueueSet;
823
824 typedef OverflowTaskQueue<size_t, mtInternal> RegionTaskQueue;
825 typedef GenericTaskQueueSet<RegionTaskQueue, mtClass> RegionTaskQueueSet;
826
827
828 #endif // SHARE_VM_UTILITIES_TASKQUEUE_HPP