1 /*
2 * Copyright (c) 2018, 2019, 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_UTILITIES_CONCURRENTHASHTABLE_INLINE_HPP
26 #define SHARE_UTILITIES_CONCURRENTHASHTABLE_INLINE_HPP
27
28 #include "memory/allocation.inline.hpp"
29 #include "runtime/atomic.hpp"
30 #include "runtime/orderAccess.hpp"
31 #include "runtime/prefetch.inline.hpp"
32 #include "utilities/concurrentHashTable.hpp"
33 #include "utilities/globalCounter.inline.hpp"
34 #include "utilities/numberSeq.hpp"
35 #include "utilities/spinYield.hpp"
36
37 // 2^30 = 1G buckets
38 #define SIZE_BIG_LOG2 30
39 // 2^5 = 32 buckets
40 #define SIZE_SMALL_LOG2 5
41
42 // Number from spinYield.hpp. In some loops SpinYield would be unfair.
43 #define SPINPAUSES_PER_YIELD 8192
44
45 #ifdef ASSERT
46 #ifdef _LP64
47 // Two low bits are not usable.
48 static const void* POISON_PTR = (void*)UCONST64(0xfbadbadbadbadbac);
49 #else
50 // Two low bits are not usable.
51 static const void* POISON_PTR = (void*)0xffbadbac;
52 #endif
53 #endif
54
55 // Node
56 template <typename VALUE, typename CONFIG, MEMFLAGS F>
57 inline typename ConcurrentHashTable<VALUE, CONFIG, F>::Node*
58 ConcurrentHashTable<VALUE, CONFIG, F>::
59 Node::next() const
60 {
61 return OrderAccess::load_acquire(&_next);
62 }
63
64 // Bucket
65 template <typename VALUE, typename CONFIG, MEMFLAGS F>
66 inline typename ConcurrentHashTable<VALUE, CONFIG, F>::Node*
67 ConcurrentHashTable<VALUE, CONFIG, F>::
68 Bucket::first_raw() const
69 {
70 return OrderAccess::load_acquire(&_first);
71 }
72
73 template <typename VALUE, typename CONFIG, MEMFLAGS F>
74 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
75 Bucket::release_assign_node_ptr(
76 typename ConcurrentHashTable<VALUE, CONFIG, F>::Node* const volatile * dst,
77 typename ConcurrentHashTable<VALUE, CONFIG, F>::Node* node) const
78 {
79 // Due to this assert this methods is not static.
80 assert(is_locked(), "Must be locked.");
81 Node** tmp = (Node**)dst;
82 OrderAccess::release_store(tmp, clear_set_state(node, *dst));
83 }
84
85 template <typename VALUE, typename CONFIG, MEMFLAGS F>
86 inline typename ConcurrentHashTable<VALUE, CONFIG, F>::Node*
87 ConcurrentHashTable<VALUE, CONFIG, F>::
88 Bucket::first() const
89 {
90 // We strip the states bit before returning the ptr.
91 return clear_state(OrderAccess::load_acquire(&_first));
92 }
93
94 template <typename VALUE, typename CONFIG, MEMFLAGS F>
95 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
96 Bucket::have_redirect() const
97 {
98 return is_state(first_raw(), STATE_REDIRECT_BIT);
99 }
100
101 template <typename VALUE, typename CONFIG, MEMFLAGS F>
102 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
103 Bucket::is_locked() const
104 {
105 return is_state(first_raw(), STATE_LOCK_BIT);
106 }
107
108 template <typename VALUE, typename CONFIG, MEMFLAGS F>
109 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
110 Bucket::lock()
111 {
112 int i = 0;
113 // SpinYield would be unfair here
114 while (!this->trylock()) {
115 if ((++i) == SPINPAUSES_PER_YIELD) {
116 // On contemporary OS yielding will give CPU to another runnable thread if
117 // there is no CPU available.
118 os::naked_yield();
119 i = 0;
120 } else {
121 SpinPause();
122 }
123 }
124 }
125
126 template <typename VALUE, typename CONFIG, MEMFLAGS F>
127 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
128 Bucket::release_assign_last_node_next(
129 typename ConcurrentHashTable<VALUE, CONFIG, F>::Node* node)
130 {
131 assert(is_locked(), "Must be locked.");
132 Node* const volatile * ret = first_ptr();
133 while (clear_state(*ret) != NULL) {
134 ret = clear_state(*ret)->next_ptr();
135 }
136 release_assign_node_ptr(ret, node);
137 }
138
139 template <typename VALUE, typename CONFIG, MEMFLAGS F>
140 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
141 Bucket::cas_first(typename ConcurrentHashTable<VALUE, CONFIG, F>::Node* node,
142 typename ConcurrentHashTable<VALUE, CONFIG, F>::Node* expect
143 )
144 {
145 if (is_locked()) {
146 return false;
147 }
148 if (Atomic::cmpxchg(node, &_first, expect) == expect) {
149 return true;
150 }
151 return false;
152 }
153
154 template <typename VALUE, typename CONFIG, MEMFLAGS F>
155 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
156 Bucket::trylock()
157 {
158 if (is_locked()) {
159 return false;
160 }
161 // We will expect a clean first pointer.
162 Node* tmp = first();
163 if (Atomic::cmpxchg(set_state(tmp, STATE_LOCK_BIT), &_first, tmp) == tmp) {
164 return true;
165 }
166 return false;
167 }
168
169 template <typename VALUE, typename CONFIG, MEMFLAGS F>
170 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
171 Bucket::unlock()
172 {
173 assert(is_locked(), "Must be locked.");
174 assert(!have_redirect(),
175 "Unlocking a bucket after it has reached terminal state.");
176 OrderAccess::release_store(&_first, clear_state(first()));
177 }
178
179 template <typename VALUE, typename CONFIG, MEMFLAGS F>
180 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
181 Bucket::redirect()
182 {
183 assert(is_locked(), "Must be locked.");
184 OrderAccess::release_store(&_first, set_state(_first, STATE_REDIRECT_BIT));
185 }
186
187 // InternalTable
188 template <typename VALUE, typename CONFIG, MEMFLAGS F>
189 inline ConcurrentHashTable<VALUE, CONFIG, F>::
190 InternalTable::InternalTable(size_t log2_size)
191 : _log2_size(log2_size), _size(((size_t)1ul) << _log2_size),
192 _hash_mask(~(~((size_t)0) << _log2_size))
193 {
194 assert(_log2_size >= SIZE_SMALL_LOG2 && _log2_size <= SIZE_BIG_LOG2,
195 "Bad size");
196 _buckets = NEW_C_HEAP_ARRAY(Bucket, _size, F);
197 // Use placement new for each element instead of new[] which could use more
198 // memory than allocated.
199 for (size_t i = 0; i < _size; ++i) {
200 new (_buckets + i) Bucket();
201 }
202 }
203
204 template <typename VALUE, typename CONFIG, MEMFLAGS F>
205 inline ConcurrentHashTable<VALUE, CONFIG, F>::
206 InternalTable::~InternalTable()
207 {
208 FREE_C_HEAP_ARRAY(Bucket, _buckets);
209 }
210
211 // ScopedCS
212 template <typename VALUE, typename CONFIG, MEMFLAGS F>
213 inline ConcurrentHashTable<VALUE, CONFIG, F>::
214 ScopedCS::ScopedCS(Thread* thread, ConcurrentHashTable<VALUE, CONFIG, F>* cht)
215 : _thread(thread),
216 _cht(cht),
217 _cs_context(GlobalCounter::critical_section_begin(_thread))
218 {
219 // This version is published now.
220 if (OrderAccess::load_acquire(&_cht->_invisible_epoch) != NULL) {
221 OrderAccess::release_store_fence(&_cht->_invisible_epoch, (Thread*)NULL);
222 }
223 }
224
225 template <typename VALUE, typename CONFIG, MEMFLAGS F>
226 inline ConcurrentHashTable<VALUE, CONFIG, F>::
227 ScopedCS::~ScopedCS()
228 {
229 GlobalCounter::critical_section_end(_thread, _cs_context);
230 }
231
232 // BaseConfig
233 template <typename VALUE, typename CONFIG, MEMFLAGS F>
234 inline void* ConcurrentHashTable<VALUE, CONFIG, F>::
235 BaseConfig::allocate_node(size_t size, const VALUE& value)
236 {
237 return AllocateHeap(size, F);
238 }
239
240 template <typename VALUE, typename CONFIG, MEMFLAGS F>
241 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
242 BaseConfig::free_node(void* memory, const VALUE& value)
243 {
244 FreeHeap(memory);
245 }
246
247 template <typename VALUE, typename CONFIG, MEMFLAGS F>
248 template <typename LOOKUP_FUNC>
249 inline VALUE* ConcurrentHashTable<VALUE, CONFIG, F>::
250 MultiGetHandle::get(LOOKUP_FUNC& lookup_f, bool* grow_hint)
251 {
252 return ScopedCS::_cht->internal_get(ScopedCS::_thread, lookup_f, grow_hint);
253 }
254
255 // HaveDeletables
256 template <typename VALUE, typename CONFIG, MEMFLAGS F>
257 template <typename EVALUATE_FUNC>
258 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
259 HaveDeletables<true, EVALUATE_FUNC>::have_deletable(Bucket* bucket,
260 EVALUATE_FUNC& eval_f,
261 Bucket* prefetch_bucket)
262 {
263 // Instantiated for pointer type (true), so we can use prefetch.
264 // When visiting all Nodes doing this prefetch give around 30%.
265 Node* pref = prefetch_bucket != NULL ? prefetch_bucket->first() : NULL;
266 for (Node* next = bucket->first(); next != NULL ; next = next->next()) {
267 if (pref != NULL) {
268 Prefetch::read(*pref->value(), 0);
269 pref = pref->next();
270 }
271 // Read next() Node* once. May be racing with a thread moving the next
272 // pointers.
273 Node* next_pref = next->next();
274 if (next_pref != NULL) {
275 Prefetch::read(*next_pref->value(), 0);
276 }
277 if (eval_f(next->value())) {
278 return true;
279 }
280 }
281 return false;
282 }
283
284 template <typename VALUE, typename CONFIG, MEMFLAGS F>
285 template <bool b, typename EVALUATE_FUNC>
286 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
287 HaveDeletables<b, EVALUATE_FUNC>::have_deletable(Bucket* bucket,
288 EVALUATE_FUNC& eval_f,
289 Bucket* preb)
290 {
291 for (Node* next = bucket->first(); next != NULL ; next = next->next()) {
292 if (eval_f(next->value())) {
293 return true;
294 }
295 }
296 return false;
297 }
298
299 // ConcurrentHashTable
300 template <typename VALUE, typename CONFIG, MEMFLAGS F>
301 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
302 write_synchonize_on_visible_epoch(Thread* thread)
303 {
304 assert(_resize_lock_owner == thread, "Re-size lock not held");
305 OrderAccess::fence(); // Prevent below load from floating up.
306 // If no reader saw this version we can skip write_synchronize.
307 if (OrderAccess::load_acquire(&_invisible_epoch) == thread) {
308 return;
309 }
310 assert(_invisible_epoch == NULL, "Two thread doing bulk operations");
311 // We set this/next version that we are synchronizing for to not published.
312 // A reader will zero this flag if it reads this/next version.
313 OrderAccess::release_store(&_invisible_epoch, thread);
314 GlobalCounter::write_synchronize();
315 }
316
317 template <typename VALUE, typename CONFIG, MEMFLAGS F>
318 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
319 try_resize_lock(Thread* locker)
320 {
321 if (_resize_lock->try_lock()) {
322 if (_resize_lock_owner != NULL) {
323 assert(locker != _resize_lock_owner, "Already own lock");
324 // We got mutex but internal state is locked.
325 _resize_lock->unlock();
326 return false;
327 }
328 } else {
329 return false;
330 }
331 _invisible_epoch = 0;
332 _resize_lock_owner = locker;
333 return true;
334 }
335
336 template <typename VALUE, typename CONFIG, MEMFLAGS F>
337 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
338 lock_resize_lock(Thread* locker)
339 {
340 size_t i = 0;
341 // If lock is hold by some other thread, the chances that it is return quick
342 // is low. So we will prefer yielding.
343 SpinYield yield(1, 512);
344 do {
345 _resize_lock->lock_without_safepoint_check();
346 // If holder of lock dropped mutex for safepoint mutex might be unlocked,
347 // and _resize_lock_owner will contain the owner.
348 if (_resize_lock_owner != NULL) {
349 assert(locker != _resize_lock_owner, "Already own lock");
350 // We got mutex but internal state is locked.
351 _resize_lock->unlock();
352 yield.wait();
353 } else {
354 break;
355 }
356 } while(true);
357 _resize_lock_owner = locker;
358 _invisible_epoch = 0;
359 }
360
361 template <typename VALUE, typename CONFIG, MEMFLAGS F>
362 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
363 unlock_resize_lock(Thread* locker)
364 {
365 _invisible_epoch = 0;
366 assert(locker == _resize_lock_owner, "Not unlocked by locker.");
367 _resize_lock_owner = NULL;
368 _resize_lock->unlock();
369 }
370
371 template <typename VALUE, typename CONFIG, MEMFLAGS F>
372 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
373 free_nodes()
374 {
375 // We assume we are not MT during freeing.
376 for (size_t node_it = 0; node_it < _table->_size; node_it++) {
377 Bucket* bucket = _table->get_buckets() + node_it;
378 Node* node = bucket->first();
379 while (node != NULL) {
380 Node* free_node = node;
381 node = node->next();
382 Node::destroy_node(free_node);
383 }
384 }
385 }
386
387 template <typename VALUE, typename CONFIG, MEMFLAGS F>
388 inline typename ConcurrentHashTable<VALUE, CONFIG, F>::InternalTable*
389 ConcurrentHashTable<VALUE, CONFIG, F>::
390 get_table() const
391 {
392 return OrderAccess::load_acquire(&_table);
393 }
394
395 template <typename VALUE, typename CONFIG, MEMFLAGS F>
396 inline typename ConcurrentHashTable<VALUE, CONFIG, F>::InternalTable*
397 ConcurrentHashTable<VALUE, CONFIG, F>::
398 get_new_table() const
399 {
400 return OrderAccess::load_acquire(&_new_table);
401 }
402
403 template <typename VALUE, typename CONFIG, MEMFLAGS F>
404 inline typename ConcurrentHashTable<VALUE, CONFIG, F>::InternalTable*
405 ConcurrentHashTable<VALUE, CONFIG, F>::
406 set_table_from_new()
407 {
408 InternalTable* old_table = _table;
409 // Publish the new table.
410 OrderAccess::release_store(&_table, _new_table);
411 // All must see this.
412 GlobalCounter::write_synchronize();
413 // _new_table not read any more.
414 _new_table = NULL;
415 DEBUG_ONLY(_new_table = (InternalTable*)POISON_PTR;)
416 return old_table;
417 }
418
419 template <typename VALUE, typename CONFIG, MEMFLAGS F>
420 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
421 internal_grow_range(Thread* thread, size_t start, size_t stop)
422 {
423 assert(stop <= _table->_size, "Outside backing array");
424 assert(_new_table != NULL, "Grow not proper setup before start");
425 // The state is also copied here. Hence all buckets in new table will be
426 // locked. I call the siblings odd/even, where even have high bit 0 and odd
427 // have high bit 1.
428 for (size_t even_index = start; even_index < stop; even_index++) {
429 Bucket* bucket = _table->get_bucket(even_index);
430
431 bucket->lock();
432
433 size_t odd_index = even_index + _table->_size;
434 _new_table->get_buckets()[even_index] = *bucket;
435 _new_table->get_buckets()[odd_index] = *bucket;
436
437 // Moves lockers go to new table, where they will wait until unlock() below.
438 bucket->redirect(); /* Must release stores above */
439
440 // When this is done we have separated the nodes into corresponding buckets
441 // in new table.
442 if (!unzip_bucket(thread, _table, _new_table, even_index, odd_index)) {
443 // If bucket is empty, unzip does nothing.
444 // We must make sure readers go to new table before we poison the bucket.
445 DEBUG_ONLY(GlobalCounter::write_synchronize();)
446 }
447
448 // Unlock for writes into the new table buckets.
449 _new_table->get_bucket(even_index)->unlock();
450 _new_table->get_bucket(odd_index)->unlock();
451
452 DEBUG_ONLY(
453 bucket->release_assign_node_ptr(
454 _table->get_bucket(even_index)->first_ptr(), (Node*)POISON_PTR);
455 )
456 }
457 }
458
459 template <typename VALUE, typename CONFIG, MEMFLAGS F>
460 template <typename LOOKUP_FUNC, typename DELETE_FUNC>
461 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
462 internal_remove(Thread* thread, LOOKUP_FUNC& lookup_f, DELETE_FUNC& delete_f)
463 {
464 Bucket* bucket = get_bucket_locked(thread, lookup_f.get_hash());
465 assert(bucket->is_locked(), "Must be locked.");
466 Node* const volatile * rem_n_prev = bucket->first_ptr();
467 Node* rem_n = bucket->first();
468 bool have_dead = false;
469 while (rem_n != NULL) {
470 if (lookup_f.equals(rem_n->value(), &have_dead)) {
471 bucket->release_assign_node_ptr(rem_n_prev, rem_n->next());
472 break;
473 } else {
474 rem_n_prev = rem_n->next_ptr();
475 rem_n = rem_n->next();
476 }
477 }
478
479 bucket->unlock();
480
481 if (rem_n == NULL) {
482 return false;
483 }
484 // Publish the deletion.
485 GlobalCounter::write_synchronize();
486 delete_f(rem_n->value());
487 Node::destroy_node(rem_n);
488 return true;
489 }
490
491 template <typename VALUE, typename CONFIG, MEMFLAGS F>
492 template <typename EVALUATE_FUNC, typename DELETE_FUNC>
493 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
494 do_bulk_delete_locked_for(Thread* thread, size_t start_idx, size_t stop_idx,
495 EVALUATE_FUNC& eval_f, DELETE_FUNC& del_f, bool is_mt)
496 {
497 // Here we have resize lock so table is SMR safe, and there is no new
498 // table. Can do this in parallel if we want.
499 assert((is_mt && _resize_lock_owner != NULL) ||
500 (!is_mt && _resize_lock_owner == thread), "Re-size lock not held");
501 Node* ndel[BULK_DELETE_LIMIT];
502 InternalTable* table = get_table();
503 assert(start_idx < stop_idx, "Must be");
504 assert(stop_idx <= _table->_size, "Must be");
505 // Here manual do critical section since we don't want to take the cost of
506 // locking the bucket if there is nothing to delete. But we can have
507 // concurrent single deletes. The _invisible_epoch can only be used by the
508 // owner of _resize_lock, us here. There we should not changed it in our
509 // own read-side.
510 GlobalCounter::CSContext cs_context = GlobalCounter::critical_section_begin(thread);
511 for (size_t bucket_it = start_idx; bucket_it < stop_idx; bucket_it++) {
512 Bucket* bucket = table->get_bucket(bucket_it);
513 Bucket* prefetch_bucket = (bucket_it+1) < stop_idx ?
514 table->get_bucket(bucket_it+1) : NULL;
515
516 if (!HaveDeletables<IsPointer<VALUE>::value, EVALUATE_FUNC>::
517 have_deletable(bucket, eval_f, prefetch_bucket)) {
518 // Nothing to remove in this bucket.
519 continue;
520 }
521
522 GlobalCounter::critical_section_end(thread, cs_context);
523 // We left critical section but the bucket cannot be removed while we hold
524 // the _resize_lock.
525 bucket->lock();
526 size_t nd = delete_check_nodes(bucket, eval_f, BULK_DELETE_LIMIT, ndel);
527 bucket->unlock();
528 if (is_mt) {
529 GlobalCounter::write_synchronize();
530 } else {
531 write_synchonize_on_visible_epoch(thread);
532 }
533 for (size_t node_it = 0; node_it < nd; node_it++) {
534 del_f(ndel[node_it]->value());
535 Node::destroy_node(ndel[node_it]);
536 DEBUG_ONLY(ndel[node_it] = (Node*)POISON_PTR;)
537 }
538 cs_context = GlobalCounter::critical_section_begin(thread);
539 }
540 GlobalCounter::critical_section_end(thread, cs_context);
541 }
542
543 template <typename VALUE, typename CONFIG, MEMFLAGS F>
544 template <typename LOOKUP_FUNC>
545 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
546 delete_in_bucket(Thread* thread, Bucket* bucket, LOOKUP_FUNC& lookup_f)
547 {
548 assert(bucket->is_locked(), "Must be locked.");
549
550 size_t dels = 0;
551 Node* ndel[BULK_DELETE_LIMIT];
552 Node* const volatile * rem_n_prev = bucket->first_ptr();
553 Node* rem_n = bucket->first();
554 while (rem_n != NULL) {
555 bool is_dead = false;
556 lookup_f.equals(rem_n->value(), &is_dead);
557 if (is_dead) {
558 ndel[dels++] = rem_n;
559 Node* next_node = rem_n->next();
560 bucket->release_assign_node_ptr(rem_n_prev, next_node);
561 rem_n = next_node;
562 if (dels == BULK_DELETE_LIMIT) {
563 break;
564 }
565 } else {
566 rem_n_prev = rem_n->next_ptr();
567 rem_n = rem_n->next();
568 }
569 }
570 if (dels > 0) {
571 GlobalCounter::write_synchronize();
572 for (size_t node_it = 0; node_it < dels; node_it++) {
573 Node::destroy_node(ndel[node_it]);
574 DEBUG_ONLY(ndel[node_it] = (Node*)POISON_PTR;)
575 }
576 }
577 }
578
579 template <typename VALUE, typename CONFIG, MEMFLAGS F>
580 inline typename ConcurrentHashTable<VALUE, CONFIG, F>::Bucket*
581 ConcurrentHashTable<VALUE, CONFIG, F>::
582 get_bucket(uintx hash) const
583 {
584 InternalTable* table = get_table();
585 Bucket* bucket = get_bucket_in(table, hash);
586 if (bucket->have_redirect()) {
587 table = get_new_table();
588 bucket = get_bucket_in(table, hash);
589 }
590 return bucket;
591 }
592
593 template <typename VALUE, typename CONFIG, MEMFLAGS F>
594 inline typename ConcurrentHashTable<VALUE, CONFIG, F>::Bucket*
595 ConcurrentHashTable<VALUE, CONFIG, F>::
596 get_bucket_locked(Thread* thread, const uintx hash)
597 {
598 Bucket* bucket;
599 int i = 0;
600 // SpinYield would be unfair here
601 while(true) {
602 {
603 // We need a critical section to protect the table itself. But if we fail
604 // we must leave critical section otherwise we would deadlock.
605 ScopedCS cs(thread, this);
606 bucket = get_bucket(hash);
607 if (bucket->trylock()) {
608 break; /* ends critical section */
609 }
610 } /* ends critical section */
611 if ((++i) == SPINPAUSES_PER_YIELD) {
612 // On contemporary OS yielding will give CPU to another runnable thread if
613 // there is no CPU available.
614 os::naked_yield();
615 i = 0;
616 } else {
617 SpinPause();
618 }
619 }
620 return bucket;
621 }
622
623 // Always called within critical section
624 template <typename VALUE, typename CONFIG, MEMFLAGS F>
625 template <typename LOOKUP_FUNC>
626 typename ConcurrentHashTable<VALUE, CONFIG, F>::Node*
627 ConcurrentHashTable<VALUE, CONFIG, F>::
628 get_node(const Bucket* const bucket, LOOKUP_FUNC& lookup_f,
629 bool* have_dead, size_t* loops) const
630 {
631 size_t loop_count = 0;
632 Node* node = bucket->first();
633 while (node != NULL) {
634 bool is_dead = false;
635 ++loop_count;
636 if (lookup_f.equals(node->value(), &is_dead)) {
637 break;
638 }
639 if (is_dead && !(*have_dead)) {
640 *have_dead = true;
641 }
642 node = node->next();
643 }
644 if (loops != NULL) {
645 *loops = loop_count;
646 }
647 return node;
648 }
649
650 template <typename VALUE, typename CONFIG, MEMFLAGS F>
651 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
652 unzip_bucket(Thread* thread, InternalTable* old_table,
653 InternalTable* new_table, size_t even_index, size_t odd_index)
654 {
655 Node* aux = old_table->get_bucket(even_index)->first();
656 if (aux == NULL) {
657 // This is an empty bucket and in debug we poison first ptr in bucket.
658 // Therefore we must make sure no readers are looking at this bucket.
659 // If we don't do a write_synch here, caller must do it.
660 return false;
661 }
662 Node* delete_me = NULL;
663 Node* const volatile * even = new_table->get_bucket(even_index)->first_ptr();
664 Node* const volatile * odd = new_table->get_bucket(odd_index)->first_ptr();
665 while (aux != NULL) {
666 bool dead_hash = false;
667 size_t aux_hash = CONFIG::get_hash(*aux->value(), &dead_hash);
668 Node* aux_next = aux->next();
669 if (dead_hash) {
670 delete_me = aux;
671 // This item is dead, move both list to next
672 new_table->get_bucket(odd_index)->release_assign_node_ptr(odd,
673 aux_next);
674 new_table->get_bucket(even_index)->release_assign_node_ptr(even,
675 aux_next);
676 } else {
677 size_t aux_index = bucket_idx_hash(new_table, aux_hash);
678 if (aux_index == even_index) {
679 // This is a even, so move odd to aux/even next
680 new_table->get_bucket(odd_index)->release_assign_node_ptr(odd,
681 aux_next);
682 // Keep in even list
683 even = aux->next_ptr();
684 } else if (aux_index == odd_index) {
685 // This is a odd, so move odd to aux/odd next
686 new_table->get_bucket(even_index)->release_assign_node_ptr(even,
687 aux_next);
688 // Keep in odd list
689 odd = aux->next_ptr();
690 } else {
691 fatal("aux_index does not match even or odd indices");
692 }
693 }
694 aux = aux_next;
695
696 // We can only move 1 pointer otherwise a reader might be moved to the wrong
697 // chain. E.g. looking for even hash value but got moved to the odd bucket
698 // chain.
699 write_synchonize_on_visible_epoch(thread);
700 if (delete_me != NULL) {
701 Node::destroy_node(delete_me);
702 delete_me = NULL;
703 }
704 }
705 return true;
706 }
707
708 template <typename VALUE, typename CONFIG, MEMFLAGS F>
709 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
710 internal_shrink_prolog(Thread* thread, size_t log2_size)
711 {
712 if (!try_resize_lock(thread)) {
713 return false;
714 }
715 assert(_resize_lock_owner == thread, "Re-size lock not held");
716 if (_table->_log2_size == _log2_start_size ||
717 _table->_log2_size <= log2_size) {
718 unlock_resize_lock(thread);
719 return false;
720 }
721 _new_table = new InternalTable(_table->_log2_size - 1);
722 return true;
723 }
724
725 template <typename VALUE, typename CONFIG, MEMFLAGS F>
726 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
727 internal_shrink_epilog(Thread* thread)
728 {
729 assert(_resize_lock_owner == thread, "Re-size lock not held");
730
731 InternalTable* old_table = set_table_from_new();
732 _size_limit_reached = false;
733 unlock_resize_lock(thread);
734 #ifdef ASSERT
735 for (size_t i = 0; i < old_table->_size; i++) {
736 assert(old_table->get_bucket(i++)->first() == POISON_PTR,
737 "No poison found");
738 }
739 #endif
740 // ABA safe, old_table not visible to any other threads.
741 delete old_table;
742 }
743
744 template <typename VALUE, typename CONFIG, MEMFLAGS F>
745 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
746 internal_shrink_range(Thread* thread, size_t start, size_t stop)
747 {
748 // The state is also copied here.
749 // Hence all buckets in new table will be locked.
750 for (size_t bucket_it = start; bucket_it < stop; bucket_it++) {
751 size_t even_hash_index = bucket_it; // High bit 0
752 size_t odd_hash_index = bucket_it + _new_table->_size; // High bit 1
753
754 Bucket* b_old_even = _table->get_bucket(even_hash_index);
755 Bucket* b_old_odd = _table->get_bucket(odd_hash_index);
756
757 b_old_even->lock();
758 b_old_odd->lock();
759
760 _new_table->get_buckets()[bucket_it] = *b_old_even;
761
762 // Put chains together.
763 _new_table->get_bucket(bucket_it)->
764 release_assign_last_node_next(*(b_old_odd->first_ptr()));
765
766 b_old_even->redirect();
767 b_old_odd->redirect();
768
769 write_synchonize_on_visible_epoch(thread);
770
771 // Unlock for writes into new smaller table.
772 _new_table->get_bucket(bucket_it)->unlock();
773
774 DEBUG_ONLY(b_old_even->release_assign_node_ptr(b_old_even->first_ptr(),
775 (Node*)POISON_PTR);)
776 DEBUG_ONLY(b_old_odd->release_assign_node_ptr(b_old_odd->first_ptr(),
777 (Node*)POISON_PTR);)
778 }
779 }
780
781 template <typename VALUE, typename CONFIG, MEMFLAGS F>
782 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
783 internal_shrink(Thread* thread, size_t log2_size)
784 {
785 if (!internal_shrink_prolog(thread, log2_size)) {
786 assert(_resize_lock_owner != thread, "Re-size lock held");
787 return false;
788 }
789 assert(_resize_lock_owner == thread, "Should be locked by me");
790 internal_shrink_range(thread, 0, _new_table->_size);
791 internal_shrink_epilog(thread);
792 assert(_resize_lock_owner != thread, "Re-size lock held");
793 return true;
794 }
795
796 template <typename VALUE, typename CONFIG, MEMFLAGS F>
797 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
798 internal_grow_prolog(Thread* thread, size_t log2_size)
799 {
800 // This double checking of _size_limit_reached/is_max_size_reached()
801 // we only do in grow path, since grow means high load on table
802 // while shrink means low load.
803 if (is_max_size_reached()) {
804 return false;
805 }
806 if (!try_resize_lock(thread)) {
807 // Either we have an ongoing resize or an operation which doesn't want us
808 // to resize now.
809 return false;
810 }
811 if (is_max_size_reached() || _table->_log2_size >= log2_size) {
812 unlock_resize_lock(thread);
813 return false;
814 }
815
816 _new_table = new InternalTable(_table->_log2_size + 1);
817
818 if (_new_table->_log2_size == _log2_size_limit) {
819 _size_limit_reached = true;
820 }
821
822 return true;
823 }
824
825 template <typename VALUE, typename CONFIG, MEMFLAGS F>
826 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
827 internal_grow_epilog(Thread* thread)
828 {
829 assert(_resize_lock_owner == thread, "Should be locked");
830
831 InternalTable* old_table = set_table_from_new();
832 unlock_resize_lock(thread);
833 #ifdef ASSERT
834 for (size_t i = 0; i < old_table->_size; i++) {
835 assert(old_table->get_bucket(i++)->first() == POISON_PTR,
836 "No poison found");
837 }
838 #endif
839 // ABA safe, old_table not visible to any other threads.
840 delete old_table;
841 }
842
843 template <typename VALUE, typename CONFIG, MEMFLAGS F>
844 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
845 internal_grow(Thread* thread, size_t log2_size)
846 {
847 if (!internal_grow_prolog(thread, log2_size)) {
848 assert(_resize_lock_owner != thread, "Re-size lock held");
849 return false;
850 }
851 assert(_resize_lock_owner == thread, "Should be locked by me");
852 internal_grow_range(thread, 0, _table->_size);
853 internal_grow_epilog(thread);
854 assert(_resize_lock_owner != thread, "Re-size lock held");
855 return true;
856 }
857
858 // Always called within critical section
859 template <typename VALUE, typename CONFIG, MEMFLAGS F>
860 template <typename LOOKUP_FUNC>
861 inline VALUE* ConcurrentHashTable<VALUE, CONFIG, F>::
862 internal_get(Thread* thread, LOOKUP_FUNC& lookup_f, bool* grow_hint)
863 {
864 bool clean = false;
865 size_t loops = 0;
866 VALUE* ret = NULL;
867
868 const Bucket* bucket = get_bucket(lookup_f.get_hash());
869 Node* node = get_node(bucket, lookup_f, &clean, &loops);
870 if (node != NULL) {
871 ret = node->value();
872 }
873 if (grow_hint != NULL) {
874 *grow_hint = loops > _grow_hint;
875 }
876
877 return ret;
878 }
879
880 template <typename VALUE, typename CONFIG, MEMFLAGS F>
881 template <typename LOOKUP_FUNC>
882 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
883 internal_insert(Thread* thread, LOOKUP_FUNC& lookup_f, const VALUE& value,
884 bool* grow_hint, bool* clean_hint)
885 {
886 bool ret = false;
887 bool clean = false;
888 bool locked;
889 size_t loops = 0;
890 size_t i = 0;
891 uintx hash = lookup_f.get_hash();
892 Node* new_node = Node::create_node(value, NULL);
893
894 while (true) {
895 {
896 ScopedCS cs(thread, this); /* protected the table/bucket */
897 Bucket* bucket = get_bucket(hash);
898 Node* first_at_start = bucket->first();
899 Node* old = get_node(bucket, lookup_f, &clean, &loops);
900 if (old == NULL) {
901 new_node->set_next(first_at_start);
902 if (bucket->cas_first(new_node, first_at_start)) {
903 new_node = NULL;
904 ret = true;
905 break; /* leave critical section */
906 }
907 // CAS failed we must leave critical section and retry.
908 locked = bucket->is_locked();
909 } else {
910 // There is a duplicate.
911 break; /* leave critical section */
912 }
913 } /* leave critical section */
914 i++;
915 if (locked) {
916 os::naked_yield();
917 } else {
918 SpinPause();
919 }
920 }
921
922 if (new_node != NULL) {
923 // CAS failed and a duplicate was inserted, we must free this node.
924 Node::destroy_node(new_node);
925 } else if (i == 0 && clean) {
926 // We only do cleaning on fast inserts.
927 Bucket* bucket = get_bucket_locked(thread, lookup_f.get_hash());
928 delete_in_bucket(thread, bucket, lookup_f);
929 bucket->unlock();
930 clean = false;
931 }
932
933 if (grow_hint != NULL) {
934 *grow_hint = loops > _grow_hint;
935 }
936
937 if (clean_hint != NULL) {
938 *clean_hint = clean;
939 }
940
941 return ret;
942 }
943
944 template <typename VALUE, typename CONFIG, MEMFLAGS F>
945 template <typename FUNC>
946 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
947 visit_nodes(Bucket* bucket, FUNC& visitor_f)
948 {
949 Node* current_node = bucket->first();
950 while (current_node != NULL) {
951 if (!visitor_f(current_node->value())) {
952 return false;
953 }
954 current_node = current_node->next();
955 }
956 return true;
957 }
958
959 template <typename VALUE, typename CONFIG, MEMFLAGS F>
960 template <typename FUNC>
961 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
962 do_scan_locked(Thread* thread, FUNC& scan_f)
963 {
964 assert(_resize_lock_owner == thread, "Re-size lock not held");
965 // We can do a critical section over the entire loop but that would block
966 // updates for a long time. Instead we choose to block resizes.
967 InternalTable* table = get_table();
968 for (size_t bucket_it = 0; bucket_it < table->_size; bucket_it++) {
969 ScopedCS cs(thread, this);
970 if (!visit_nodes(table->get_bucket(bucket_it), scan_f)) {
971 break; /* ends critical section */
972 }
973 } /* ends critical section */
974 }
975
976 template <typename VALUE, typename CONFIG, MEMFLAGS F>
977 template <typename EVALUATE_FUNC>
978 inline size_t ConcurrentHashTable<VALUE, CONFIG, F>::
979 delete_check_nodes(Bucket* bucket, EVALUATE_FUNC& eval_f,
980 size_t num_del, Node** ndel)
981 {
982 size_t dels = 0;
983 Node* const volatile * rem_n_prev = bucket->first_ptr();
984 Node* rem_n = bucket->first();
985 while (rem_n != NULL) {
986 if (eval_f(rem_n->value())) {
987 ndel[dels++] = rem_n;
988 Node* next_node = rem_n->next();
989 bucket->release_assign_node_ptr(rem_n_prev, next_node);
990 rem_n = next_node;
991 if (dels == num_del) {
992 break;
993 }
994 } else {
995 rem_n_prev = rem_n->next_ptr();
996 rem_n = rem_n->next();
997 }
998 }
999 return dels;
1000 }
1001
1002 // Constructor
1003 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1004 inline ConcurrentHashTable<VALUE, CONFIG, F>::
1005 ConcurrentHashTable(size_t log2size, size_t log2size_limit, size_t grow_hint)
1006 : _new_table(NULL), _log2_size_limit(log2size_limit),
1007 _log2_start_size(log2size), _grow_hint(grow_hint),
1008 _size_limit_reached(false), _resize_lock_owner(NULL),
1009 _invisible_epoch(0)
1010 {
1011 _resize_lock =
1012 new Mutex(Mutex::leaf, "ConcurrentHashTable", false,
1013 Monitor::_safepoint_check_never);
1014 _table = new InternalTable(log2size);
1015 assert(log2size_limit >= log2size, "bad ergo");
1016 _size_limit_reached = _table->_log2_size == _log2_size_limit;
1017 }
1018
1019 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1020 inline ConcurrentHashTable<VALUE, CONFIG, F>::
1021 ~ConcurrentHashTable()
1022 {
1023 delete _resize_lock;
1024 free_nodes();
1025 delete _table;
1026 }
1027
1028 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1029 inline size_t ConcurrentHashTable<VALUE, CONFIG, F>::
1030 get_size_log2(Thread* thread)
1031 {
1032 ScopedCS cs(thread, this);
1033 return _table->_log2_size;
1034 }
1035
1036 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1037 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
1038 shrink(Thread* thread, size_t size_limit_log2)
1039 {
1040 size_t tmp = size_limit_log2 == 0 ? _log2_start_size : size_limit_log2;
1041 bool ret = internal_shrink(thread, tmp);
1042 return ret;
1043 }
1044
1045 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1046 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
1047 grow(Thread* thread, size_t size_limit_log2)
1048 {
1049 size_t tmp = size_limit_log2 == 0 ? _log2_size_limit : size_limit_log2;
1050 return internal_grow(thread, tmp);
1051 }
1052
1053 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1054 template <typename LOOKUP_FUNC, typename FOUND_FUNC>
1055 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
1056 get(Thread* thread, LOOKUP_FUNC& lookup_f, FOUND_FUNC& found_f, bool* grow_hint)
1057 {
1058 bool ret = false;
1059 ScopedCS cs(thread, this);
1060 VALUE* val = internal_get(thread, lookup_f, grow_hint);
1061 if (val != NULL) {
1062 found_f(val);
1063 ret = true;
1064 }
1065 return ret;
1066 }
1067
1068 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1069 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
1070 unsafe_insert(const VALUE& value) {
1071 bool dead_hash = false;
1072 size_t hash = CONFIG::get_hash(value, &dead_hash);
1073 if (dead_hash) {
1074 return false;
1075 }
1076 // This is an unsafe operation.
1077 InternalTable* table = get_table();
1078 Bucket* bucket = get_bucket_in(table, hash);
1079 assert(!bucket->have_redirect() && !bucket->is_locked(), "bad");
1080 Node* new_node = Node::create_node(value, bucket->first());
1081 if (!bucket->cas_first(new_node, bucket->first())) {
1082 assert(false, "bad");
1083 }
1084 return true;
1085 }
1086
1087 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1088 template <typename SCAN_FUNC>
1089 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
1090 try_scan(Thread* thread, SCAN_FUNC& scan_f)
1091 {
1092 if (!try_resize_lock(thread)) {
1093 return false;
1094 }
1095 do_scan_locked(thread, scan_f);
1096 unlock_resize_lock(thread);
1097 return true;
1098 }
1099
1100 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1101 template <typename SCAN_FUNC>
1102 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
1103 do_scan(Thread* thread, SCAN_FUNC& scan_f)
1104 {
1105 assert(!SafepointSynchronize::is_at_safepoint(),
1106 "must be outside a safepoint");
1107 assert(_resize_lock_owner != thread, "Re-size lock held");
1108 lock_resize_lock(thread);
1109 do_scan_locked(thread, scan_f);
1110 unlock_resize_lock(thread);
1111 assert(_resize_lock_owner != thread, "Re-size lock held");
1112 }
1113
1114 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1115 template <typename SCAN_FUNC>
1116 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
1117 do_safepoint_scan(SCAN_FUNC& scan_f)
1118 {
1119 // We only allow this method to be used during a safepoint.
1120 assert(SafepointSynchronize::is_at_safepoint(),
1121 "must only be called in a safepoint");
1122 assert(Thread::current()->is_VM_thread(),
1123 "should be in vm thread");
1124
1125 // Here we skip protection,
1126 // thus no other thread may use this table at the same time.
1127 InternalTable* table = get_table();
1128 for (size_t bucket_it = 0; bucket_it < table->_size; bucket_it++) {
1129 Bucket* bucket = table->get_bucket(bucket_it);
1130 // If bucket have a redirect the items will be in the new table.
1131 // We must visit them there since the new table will contain any
1132 // concurrent inserts done after this bucket was resized.
1133 // If the bucket don't have redirect flag all items is in this table.
1134 if (!bucket->have_redirect()) {
1135 if(!visit_nodes(bucket, scan_f)) {
1136 return;
1137 }
1138 } else {
1139 assert(bucket->is_locked(), "Bucket must be locked.");
1140 }
1141 }
1142 // If there is a paused resize we also need to visit the already resized items.
1143 table = get_new_table();
1144 if (table == NULL) {
1145 return;
1146 }
1147 DEBUG_ONLY(if (table == POISON_PTR) { return; })
1148 for (size_t bucket_it = 0; bucket_it < table->_size; bucket_it++) {
1149 Bucket* bucket = table->get_bucket(bucket_it);
1150 assert(!bucket->is_locked(), "Bucket must be unlocked.");
1151 if (!visit_nodes(bucket, scan_f)) {
1152 return;
1153 }
1154 }
1155 }
1156
1157 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1158 template <typename EVALUATE_FUNC, typename DELETE_FUNC>
1159 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
1160 try_bulk_delete(Thread* thread, EVALUATE_FUNC& eval_f, DELETE_FUNC& del_f)
1161 {
1162 if (!try_resize_lock(thread)) {
1163 return false;
1164 }
1165 do_bulk_delete_locked(thread, eval_f, del_f);
1166 unlock_resize_lock(thread);
1167 assert(_resize_lock_owner != thread, "Re-size lock held");
1168 return true;
1169 }
1170
1171 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1172 template <typename EVALUATE_FUNC, typename DELETE_FUNC>
1173 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
1174 bulk_delete(Thread* thread, EVALUATE_FUNC& eval_f, DELETE_FUNC& del_f)
1175 {
1176 assert(!SafepointSynchronize::is_at_safepoint(),
1177 "must be outside a safepoint");
1178 lock_resize_lock(thread);
1179 do_bulk_delete_locked(thread, eval_f, del_f);
1180 unlock_resize_lock(thread);
1181 }
1182
1183 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1184 template <typename VALUE_SIZE_FUNC>
1185 inline void ConcurrentHashTable<VALUE, CONFIG, F>::
1186 statistics_to(Thread* thread, VALUE_SIZE_FUNC& vs_f,
1187 outputStream* st, const char* table_name)
1188 {
1189 NumberSeq summary;
1190 size_t literal_bytes = 0;
1191 if (!try_resize_lock(thread)) {
1192 st->print_cr("statistics unavailable at this moment");
1193 return;
1194 }
1195
1196 InternalTable* table = get_table();
1197 for (size_t bucket_it = 0; bucket_it < table->_size; bucket_it++) {
1198 ScopedCS cs(thread, this);
1199 size_t count = 0;
1200 Bucket* bucket = table->get_bucket(bucket_it);
1201 if (bucket->have_redirect() || bucket->is_locked()) {
1202 continue;
1203 }
1204 Node* current_node = bucket->first();
1205 while (current_node != NULL) {
1206 ++count;
1207 literal_bytes += vs_f(current_node->value());
1208 current_node = current_node->next();
1209 }
1210 summary.add((double)count);
1211 }
1212
1213 double num_buckets = summary.num();
1214 double num_entries = summary.sum();
1215
1216 size_t bucket_bytes = num_buckets * sizeof(Bucket);
1217 size_t entry_bytes = num_entries * sizeof(Node);
1218 size_t total_bytes = literal_bytes + bucket_bytes + entry_bytes;
1219
1220 size_t bucket_size = (num_buckets <= 0) ? 0 : (bucket_bytes / num_buckets);
1221 size_t entry_size = (num_entries <= 0) ? 0 : (entry_bytes / num_entries);
1222
1223 st->print_cr("%s statistics:", table_name);
1224 st->print_cr("Number of buckets : %9" PRIuPTR " = %9" PRIuPTR
1225 " bytes, each " SIZE_FORMAT,
1226 (size_t)num_buckets, bucket_bytes, bucket_size);
1227 st->print_cr("Number of entries : %9" PRIuPTR " = %9" PRIuPTR
1228 " bytes, each " SIZE_FORMAT,
1229 (size_t)num_entries, entry_bytes, entry_size);
1230 if (literal_bytes != 0) {
1231 double literal_avg = (num_entries <= 0) ? 0 : (literal_bytes / num_entries);
1232 st->print_cr("Number of literals : %9" PRIuPTR " = %9" PRIuPTR
1233 " bytes, avg %7.3f",
1234 (size_t)num_entries, literal_bytes, literal_avg);
1235 }
1236 st->print_cr("Total footprsize_t : %9s = %9" PRIuPTR " bytes", ""
1237 , total_bytes);
1238 st->print_cr("Average bucket size : %9.3f", summary.avg());
1239 st->print_cr("Variance of bucket size : %9.3f", summary.variance());
1240 st->print_cr("Std. dev. of bucket size: %9.3f", summary.sd());
1241 st->print_cr("Maximum bucket size : %9" PRIuPTR,
1242 (size_t)summary.maximum());
1243 unlock_resize_lock(thread);
1244 }
1245
1246 template <typename VALUE, typename CONFIG, MEMFLAGS F>
1247 inline bool ConcurrentHashTable<VALUE, CONFIG, F>::
1248 try_move_nodes_to(Thread* thread, ConcurrentHashTable<VALUE, CONFIG, F>* to_cht)
1249 {
1250 if (!try_resize_lock(thread)) {
1251 return false;
1252 }
1253 assert(_new_table == NULL || _new_table == POISON_PTR, "Must be NULL");
1254 for (size_t bucket_it = 0; bucket_it < _table->_size; bucket_it++) {
1255 Bucket* bucket = _table->get_bucket(bucket_it);
1256 assert(!bucket->have_redirect() && !bucket->is_locked(), "Table must be uncontended");
1257 while (bucket->first() != NULL) {
1258 Node* move_node = bucket->first();
1259 bool ok = bucket->cas_first(move_node->next(), move_node);
1260 assert(ok, "Uncontended cas must work");
1261 bool dead_hash = false;
1262 size_t insert_hash = CONFIG::get_hash(*move_node->value(), &dead_hash);
1263 if (!dead_hash) {
1264 Bucket* insert_bucket = to_cht->get_bucket(insert_hash);
1265 assert(!bucket->have_redirect() && !bucket->is_locked(), "Not bit should be present");
1266 move_node->set_next(insert_bucket->first());
1267 ok = insert_bucket->cas_first(move_node, insert_bucket->first());
1268 assert(ok, "Uncontended cas must work");
1269 }
1270 }
1271 }
1272 unlock_resize_lock(thread);
1273 return true;
1274 }
1275
1276 #endif // SHARE_UTILITIES_CONCURRENTHASHTABLE_INLINE_HPP