1 /*
2 * Copyright Amazon.com Inc. 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_GC_SHENANDOAH_SHENANDOAHSCANREMEMBERED_HPP
26 #define SHARE_GC_SHENANDOAH_SHENANDOAHSCANREMEMBERED_HPP
27
28 // Terminology used within this source file:
29 //
30 // Card Entry: This is the information that identifies whether a
31 // particular card-table entry is Clean or Dirty. A clean
32 // card entry denotes that the associated memory does not
33 // hold references to young-gen memory.
34 //
35 // Card Region, aka
36 // Card Memory: This is the region of memory that is assocated with a
37 // particular card entry.
38 //
39 // Card Cluster: A card cluster represents 64 card entries. A card
40 // cluster is the minimal amount of work performed at a
41 // time by a parallel thread. Note that the work required
42 // to scan a card cluster is somewhat variable in that the
43 // required effort depends on how many cards are dirty, how
44 // many references are held within the objects that span a
45 // DIRTY card's memory, and on the size of the object
46 // that spans the end of a DIRTY card's memory (because
47 // that object, if it's not an array, may need to be scanned in
48 // its entirety, when the object is imprecisely dirtied. Imprecise
49 // dirtying is when the card corresponding to the object header
50 // is dirtied, rather than the card on which the updated field lives).
51 // To better balance work amongst them, parallel worker threads dynamically
52 // claim clusters and are flexible in the number of clusters they
53 // process.
54 //
55 // A cluster represents a "natural" quantum of work to be performed by
56 // a parallel GC thread's background remembered set scanning efforts.
57 // The notion of cluster is similar to the notion of stripe in the
58 // implementation of parallel GC card scanning. However, a cluster is
59 // typically smaller than a stripe, enabling finer grain division of
60 // labor between multiple threads, and potentially better load balancing
61 // when dirty cards are not uniformly distributed in the heap, as is often
62 // the case with generational workloads where more recently promoted objects
63 // may be dirtied more frequently that older objects.
64 //
65 // For illustration, consider the following possible JVM configurations:
66 //
67 // Scenario 1:
68 // RegionSize is 128 MB
69 // Span of a card entry is 512 B
70 // Each card table entry consumes 1 B
71 // Assume one long word (8 B)of the card table represents a cluster.
72 // This long word holds 8 card table entries, spanning a
73 // total of 8*512 B = 4 KB of the heap
74 // The number of clusters per region is 128 MB / 4 KB = 32 K
75 //
76 // Scenario 2:
77 // RegionSize is 128 MB
78 // Span of each card entry is 128 B
79 // Each card table entry consumes 1 bit
80 // Assume one int word (4 B) of the card table represents a cluster.
81 // This int word holds 32 b/1 b = 32 card table entries, spanning a
82 // total of 32 * 128 B = 4 KB of the heap
83 // The number of clusters per region is 128 MB / 4 KB = 32 K
84 //
85 // Scenario 3:
86 // RegionSize is 128 MB
87 // Span of each card entry is 512 B
88 // Each card table entry consumes 1 bit
89 // Assume one long word (8 B) of card table represents a cluster.
90 // This long word holds 64 b/ 1 b = 64 card table entries, spanning a
91 // total of 64 * 512 B = 32 KB of the heap
92 // The number of clusters per region is 128 MB / 32 KB = 4 K
93 //
94 // At the start of a new young-gen concurrent mark pass, the gang of
95 // Shenandoah worker threads collaborate in performing the following
96 // actions:
97 //
98 // Let old_regions = number of ShenandoahHeapRegion comprising
99 // old-gen memory
100 // Let region_size = ShenandoahHeapRegion::region_size_bytes()
101 // represent the number of bytes in each region
102 // Let clusters_per_region = region_size / 512
103 // Let rs represent the relevant RememberedSet implementation
104 // (an instance of ShenandoahDirectCardMarkRememberedSet or an instance
105 // of a to-be-implemented ShenandoahBufferWithSATBRememberedSet)
106 //
107 // for each ShenandoahHeapRegion old_region in the whole heap
108 // determine the cluster number of the first cluster belonging
109 // to that region
110 // for each cluster contained within that region
111 // Assure that exactly one worker thread processes each
112 // cluster, each thread making a series of invocations of the
113 // following:
114 //
115 // rs->process_clusters(worker_id, ReferenceProcessor *,
116 // ShenandoahConcurrentMark *, cluster_no, cluster_count,
117 // HeapWord *end_of_range, OopClosure *oops);
118 //
119 // For efficiency, divide up the clusters so that different threads
120 // are responsible for processing different clusters. Processing costs
121 // may vary greatly between clusters for the following reasons:
122 //
123 // a) some clusters contain mostly dirty cards and other
124 // clusters contain mostly clean cards
125 // b) some clusters contain mostly primitive data and other
126 // clusters contain mostly reference data
127 // c) some clusters are spanned by very large non-array objects that
128 // begin in some other cluster. When a large non-array object
129 // beginning in a preceding cluster spans large portions of
130 // this cluster, then because of imprecise dirtying, the
131 // portion of the object in this cluster may be clean, but
132 // will need to be processed by the worker responsible for
133 // this cluster, potentially increasing its work.
134 // d) in the case that the end of this cluster is spanned by a
135 // very large non-array object, the worker for this cluster will
136 // be responsible for processing the portion of the object
137 // in this cluster.
138 //
139 // Though an initial division of labor between marking threads may
140 // assign equal numbers of clusters to be scanned by each thread, it
141 // should be expected that some threads will finish their assigned
142 // work before others. Therefore, some amount of the full remembered
143 // set scanning effort should be held back and assigned incrementally
144 // to the threads that end up with excess capacity. Consider the
145 // following strategy for dividing labor:
146 //
147 // 1. Assume there are 8 marking threads and 1024 remembered
148 // set clusters to be scanned.
149 // 2. Assign each thread to scan 64 clusters. This leaves
150 // 512 (1024 - (8*64)) clusters to still be scanned.
151 // 3. As the 8 server threads complete previous cluster
152 // scanning assignments, issue each of the next 8 scanning
153 // assignments as units of 32 additional cluster each.
154 // In the case that there is high variance in effort
155 // associated with previous cluster scanning assignments,
156 // multiples of these next assignments may be serviced by
157 // the server threads that were previously assigned lighter
158 // workloads.
159 // 4. Make subsequent scanning assignments as follows:
160 // a) 8 assignments of size 16 clusters
161 // b) 8 assignments of size 8 clusters
162 // c) 16 assignments of size 4 clusters
163 //
164 // When there is no more remembered set processing work to be
165 // assigned to a newly idled worker thread, that thread can move
166 // on to work on other tasks associated with root scanning until such
167 // time as all clusters have been examined.
168 //
169 // Remembered set scanning is designed to run concurrently with
170 // mutator threads, with multiple concurrent workers. Furthermore, the
171 // current implementation of remembered set scanning never clears a
172 // card once it has been marked.
173 //
174 // These limitations will be addressed in future enhancements to the
175 // existing implementation.
176
177 #include <stdint.h>
178 #include "gc/shared/workerThread.hpp"
179 #include "gc/shenandoah/shenandoahCardStats.hpp"
180 #include "gc/shenandoah/shenandoahCardTable.hpp"
181 #include "gc/shenandoah/shenandoahNumberSeq.hpp"
182 #include "gc/shenandoah/shenandoahTaskqueue.hpp"
183 #include "memory/iterator.hpp"
184 #include "utilities/globalDefinitions.hpp"
185
186 class ShenandoahReferenceProcessor;
187 class ShenandoahConcurrentMark;
188 class ShenandoahHeap;
189 class ShenandoahHeapRegion;
190 class ShenandoahRegionIterator;
191 class ShenandoahMarkingContext;
192
193 class CardTable;
194 typedef CardTable::CardValue CardValue;
195
196 class ShenandoahDirectCardMarkRememberedSet: public CHeapObj<mtGC> {
197
198 private:
199
200 // Use symbolic constants defined in cardTable.hpp
201 // CardTable::card_shift = 9;
202 // CardTable::card_size = 512;
203 // CardTable::card_size_in_words = 64;
204 // CardTable::clean_card_val()
205 // CardTable::dirty_card_val()
206
207 ShenandoahHeap *_heap;
208 ShenandoahCardTable *_card_table;
209 size_t _card_shift;
210 size_t _total_card_count;
211 size_t _cluster_count;
212 HeapWord *_whole_heap_base; // Points to first HeapWord of data contained within heap memory
213 CardValue* _byte_map; // Points to first entry within the card table
214 CardValue* _byte_map_base; // Points to byte_map minus the bias computed from address of heap memory
215
216 public:
217
218 // count is the number of cards represented by the card table.
219 ShenandoahDirectCardMarkRememberedSet(ShenandoahCardTable *card_table, size_t total_card_count);
220 ~ShenandoahDirectCardMarkRememberedSet();
221
222 // Card index is zero-based relative to _byte_map.
223 size_t last_valid_index() const;
224 size_t total_cards() const;
225 size_t card_index_for_addr(HeapWord *p) const;
226 HeapWord *addr_for_card_index(size_t card_index) const;
227 inline const CardValue* get_card_table_byte_map(bool write_table) const;
228 inline bool is_card_dirty(size_t card_index) const;
229 inline bool is_write_card_dirty(size_t card_index) const;
230 inline void mark_card_as_dirty(size_t card_index);
231 inline void mark_range_as_dirty(size_t card_index, size_t num_cards);
232 inline void mark_card_as_clean(size_t card_index);
233 inline void mark_read_card_as_clean(size_t card_index);
234 inline void mark_range_as_clean(size_t card_index, size_t num_cards);
235 inline bool is_card_dirty(HeapWord *p) const;
236 inline void mark_card_as_dirty(HeapWord *p);
237 inline void mark_range_as_dirty(HeapWord *p, size_t num_heap_words);
238 inline void mark_card_as_clean(HeapWord *p);
239 inline void mark_range_as_clean(HeapWord *p, size_t num_heap_words);
240 inline size_t cluster_count() const;
241
242 // Called by GC thread at start of concurrent mark to exchange roles of read and write remembered sets.
243 // Not currently used because mutator write barrier does not honor changes to the location of card table.
244 void swap_remset() { _card_table->swap_card_tables(); }
245
246 void merge_write_table(HeapWord* start, size_t word_count) {
247 size_t card_index = card_index_for_addr(start);
248 size_t num_cards = word_count / CardTable::card_size_in_words();
249 size_t iterations = num_cards / (sizeof (intptr_t) / sizeof (CardValue));
250 intptr_t* read_table_ptr = (intptr_t*) &(_card_table->read_byte_map())[card_index];
251 intptr_t* write_table_ptr = (intptr_t*) &(_card_table->write_byte_map())[card_index];
252 for (size_t i = 0; i < iterations; i++) {
253 intptr_t card_value = *write_table_ptr;
254 *read_table_ptr++ &= card_value;
255 write_table_ptr++;
256 }
257 }
258
259 // Instead of swap_remset, the current implementation of concurrent remembered set scanning does reset_remset
260 // in parallel threads, each invocation processing one entire HeapRegion at a time. Processing of a region
261 // consists of copying the write table to the read table and cleaning the write table.
262 void reset_remset(HeapWord* start, size_t word_count) {
263 size_t card_index = card_index_for_addr(start);
264 size_t num_cards = word_count / CardTable::card_size_in_words();
265 size_t iterations = num_cards / (sizeof (intptr_t) / sizeof (CardValue));
266 intptr_t* read_table_ptr = (intptr_t*) &(_card_table->read_byte_map())[card_index];
267 intptr_t* write_table_ptr = (intptr_t*) &(_card_table->write_byte_map())[card_index];
268 for (size_t i = 0; i < iterations; i++) {
269 *read_table_ptr++ = *write_table_ptr;
270 *write_table_ptr++ = CardTable::clean_card_row_val();
271 }
272 }
273
274 // Called by GC thread after scanning old remembered set in order to prepare for next GC pass
275 void clear_old_remset() { _card_table->clear_read_table(); }
276
277 };
278
279 // A ShenandoahCardCluster represents the minimal unit of work
280 // performed by independent parallel GC threads during scanning of
281 // remembered sets.
282 //
283 // The GC threads that perform card-table remembered set scanning may
284 // overwrite card-table entries to mark them as clean in the case that
285 // the associated memory no longer holds references to young-gen
286 // memory. Rather than access the card-table entries directly, all GC
287 // thread access to card-table information is made by way of the
288 // ShenandoahCardCluster data abstraction. This abstraction
289 // effectively manages access to multiple possible underlying
290 // remembered set implementations, including a traditional card-table
291 // approach and a SATB-based approach.
292 //
293 // The API services represent a compromise between efficiency and
294 // convenience.
295 //
296 // Multiple GC threads that scan the remembered set
297 // in parallel. The desire is to divide the complete scanning effort
298 // into multiple clusters of work that can be independently processed
299 // by individual threads without need for synchronizing efforts
300 // between the work performed by each task. The term "cluster" of
301 // work is similar to the term "stripe" as used in the implementation
302 // of Parallel GC.
303 //
304 // Complexity arises when an object to be scanned crosses the boundary
305 // between adjacent cluster regions. Here is the protocol that we currently
306 // follow:
307 //
308 // 1. The thread responsible for scanning the cards in a cluster modifies
309 // the associated card-table entries. Only cards that are dirty are
310 // processed, except as described below for the case of objects that
311 // straddle more than one card.
312 // 2. Object Arrays are precisely dirtied, so only the portion of the obj-array
313 // that overlaps the range of dirty cards in its cluster are scanned
314 // by each worker thread. This holds for portions of obj-arrays that extend
315 // over clusters processed by different workers, with each worked responsible
316 // for scanning the portion of the obj-array overlapping the dirty cards in
317 // its cluster.
318 // 3. Non-array objects are precisely dirtied by the interpreter and the compilers
319 // For such objects that extend over multiple cards, or even multiple clusters,
320 // the entire object is scanned by the worker that processes the (dirty) card on
321 // which the object's header lies. (However, GC workers should precisely dirty the
322 // cards with inter-regional/inter-generational pointers in the body of this object,
323 // thus making subsequent scans potentially less expensive.) Such larger non-array
324 // objects are relatively rare.
325 //
326 // A possible criticism:
327 // C. The representation of pointer location descriptive information
328 // within Klass representations is not designed for efficient
329 // "random access". An alternative approach to this design would
330 // be to scan very large objects multiple times, once for each
331 // cluster that is spanned by the object's range. This reduces
332 // unnecessary overscan, but it introduces different sorts of
333 // overhead effort:
334 // i) For each spanned cluster, we have to look up the start of
335 // the crossing object.
336 // ii) Each time we scan the very large object, we have to
337 // sequentially walk through its pointer location
338 // descriptors, skipping over all of the pointers that
339 // precede the start of the range of addresses that we
340 // consider relevant.
341
342
343 // Because old-gen heap memory is not necessarily contiguous, and
344 // because cards are not necessarily maintained for young-gen memory,
345 // consecutive card numbers do not necessarily correspond to consecutive
346 // address ranges. For the traditional direct-card-marking
347 // implementation of this interface, consecutive card numbers are
348 // likely to correspond to contiguous regions of memory, but this
349 // should not be assumed. Instead, rely only upon the following:
350 //
351 // 1. All card numbers for cards pertaining to the same
352 // ShenandoahHeapRegion are consecutively numbered.
353 // 2. In the case that neighboring ShenandoahHeapRegions both
354 // represent old-gen memory, the card regions that span the
355 // boundary between these neighboring heap regions will be
356 // consecutively numbered.
357 // 3. (A corollary) In the case that an old-gen object straddles the
358 // boundary between two heap regions, the card regions that
359 // correspond to the span of this object will be consecutively
360 // numbered.
361 //
362 // ShenandoahCardCluster abstracts access to the remembered set
363 // and also keeps track of crossing map information to allow efficient
364 // resolution of object start addresses.
365 //
366 // ShenandoahCardCluster supports all of the services of
367 // RememberedSet, plus it supports register_object() and lookup_object().
368 // Note that we only need to register the start addresses of the object that
369 // overlays the first address of a card; we need to do this for every card.
370 // In other words, register_object() checks if the object crosses a card boundary,
371 // and updates the offset value for each card that the object crosses into.
372 // For objects that don't straddle cards, nothing needs to be done.
373 //
374 // The RememberedSet template parameter is intended to represent either
375 // ShenandoahDirectCardMarkRememberedSet, or a to-be-implemented
376 // ShenandoahBufferWithSATBRememberedSet.
377 template<typename RememberedSet>
378 class ShenandoahCardCluster: public CHeapObj<mtGC> {
379
380 private:
381 RememberedSet *_rs;
382
383 public:
384 static const size_t CardsPerCluster = 64;
385
386 private:
387 typedef struct cross_map { uint8_t first; uint8_t last; } xmap;
388 typedef union crossing_info { uint16_t short_word; xmap offsets; } crossing_info;
389
390 // ObjectStartsInCardRegion bit is set within a crossing_info.offsets.start iff at least one object starts within
391 // a particular card region. We pack this bit into start byte under assumption that start byte is accessed less
392 // frequently than last byte. This is true when number of clean cards is greater than number of dirty cards.
393 static const uint16_t ObjectStartsInCardRegion = 0x80;
394 static const uint16_t FirstStartBits = 0x7f;
395
396 // Check that we have enough bits to store the largest possible offset into a card for an object start.
397 // The value for maximum card size is based on the constraints for GCCardSizeInBytes in gc_globals.hpp.
398 static const int MaxCardSize = NOT_LP64(512) LP64_ONLY(1024);
399 STATIC_ASSERT((MaxCardSize / HeapWordSize) - 1 <= FirstStartBits);
400
401 crossing_info *object_starts;
402
403 public:
404 // If we're setting first_start, assume the card has an object.
405 inline void set_first_start(size_t card_index, uint8_t value) {
406 object_starts[card_index].offsets.first = ObjectStartsInCardRegion | value;
407 }
408
409 inline void set_last_start(size_t card_index, uint8_t value) {
410 object_starts[card_index].offsets.last = value;
411 }
412
413 inline void set_starts_object_bit(size_t card_index) {
414 object_starts[card_index].offsets.first |= ObjectStartsInCardRegion;
415 }
416
417 inline void clear_starts_object_bit(size_t card_index) {
418 object_starts[card_index].offsets.first &= ~ObjectStartsInCardRegion;
419 }
420
421 // Returns true iff an object is known to start within the card memory associated with card card_index.
422 inline bool starts_object(size_t card_index) const {
423 return (object_starts[card_index].offsets.first & ObjectStartsInCardRegion) != 0;
424 }
425
426 inline void clear_objects_in_range(HeapWord *addr, size_t num_words) {
427 size_t card_index = _rs->card_index_for_addr(addr);
428 size_t last_card_index = _rs->card_index_for_addr(addr + num_words - 1);
429 while (card_index <= last_card_index)
430 object_starts[card_index++].short_word = 0;
431 }
432
433 ShenandoahCardCluster(RememberedSet *rs) {
434 _rs = rs;
435 // TODO: We don't really need object_starts entries for every card entry. We only need these for
436 // the card entries that correspond to old-gen memory. But for now, let's be quick and dirty.
437 object_starts = NEW_C_HEAP_ARRAY(crossing_info, rs->total_cards(), mtGC);
438 for (size_t i = 0; i < rs->total_cards(); i++) {
439 object_starts[i].short_word = 0;
440 }
441 }
442
443 ~ShenandoahCardCluster() {
444 FREE_C_HEAP_ARRAY(crossing_info, object_starts);
445 object_starts = nullptr;
446 }
447
448 // There is one entry within the object_starts array for each card entry.
449 //
450 // Suppose multiple garbage objects are coalesced during GC sweep
451 // into a single larger "free segment". As each two objects are
452 // coalesced together, the start information pertaining to the second
453 // object must be removed from the objects_starts array. If the
454 // second object had been been the first object within card memory,
455 // the new first object is the object that follows that object if
456 // that starts within the same card memory, or NoObject if the
457 // following object starts within the following cluster. If the
458 // second object had been the last object in the card memory,
459 // replace this entry with the newly coalesced object if it starts
460 // within the same card memory, or with NoObject if it starts in a
461 // preceding card's memory.
462 //
463 // Suppose a large free segment is divided into a smaller free
464 // segment and a new object. The second part of the newly divided
465 // memory must be registered as a new object, overwriting at most
466 // one first_start and one last_start entry. Note that one of the
467 // newly divided two objects might be a new GCLAB.
468 //
469 // Suppose postprocessing of a GCLAB finds that the original GCLAB
470 // has been divided into N objects. Each of the N newly allocated
471 // objects will be registered, overwriting at most one first_start
472 // and one last_start entries.
473 //
474 // No object registration operations are linear in the length of
475 // the registered objects.
476 //
477 // Consider further the following observations regarding object
478 // registration costs:
479 //
480 // 1. The cost is paid once for each old-gen object (Except when
481 // an object is demoted and repromoted, in which case we would
482 // pay the cost again).
483 // 2. The cost can be deferred so that there is no urgency during
484 // mutator copy-on-first-access promotion. Background GC
485 // threads will update the object_starts array by post-
486 // processing the contents of retired PLAB buffers.
487 // 3. The bet is that these costs are paid relatively rarely
488 // because:
489 // a) Most objects die young and objects that die in young-gen
490 // memory never need to be registered with the object_starts
491 // array.
492 // b) Most objects that are promoted into old-gen memory live
493 // there without further relocation for a relatively long
494 // time, so we get a lot of benefit from each investment
495 // in registering an object.
496
497 public:
498
499 // The starting locations of objects contained within old-gen memory
500 // are registered as part of the remembered set implementation. This
501 // information is required when scanning dirty card regions that are
502 // spanned by objects beginning within preceding card regions. It
503 // is necessary to find the first and last objects that begin within
504 // this card region. Starting addresses of objects are required to
505 // find the object headers, and object headers provide information
506 // about which fields within the object hold addresses.
507 //
508 // The old-gen memory allocator invokes register_object() for any
509 // object that is allocated within old-gen memory. This identifies
510 // the starting addresses of objects that span boundaries between
511 // card regions.
512 //
513 // It is not necessary to invoke register_object at the very instant
514 // an object is allocated. It is only necessary to invoke it
515 // prior to the next start of a garbage collection concurrent mark
516 // or concurrent update-references phase. An "ideal" time to register
517 // objects is during post-processing of a GCLAB after the GCLAB is
518 // retired due to depletion of its memory.
519 //
520 // register_object() does not perform synchronization. In the case
521 // that multiple threads are registering objects whose starting
522 // addresses are within the same cluster, races between these
523 // threads may result in corruption of the object-start data
524 // structures. Parallel GC threads should avoid registering objects
525 // residing within the same cluster by adhering to the following
526 // coordination protocols:
527 //
528 // 1. Align thread-local GCLAB buffers with some TBD multiple of
529 // card clusters. The card cluster size is 32 KB. If the
530 // desired GCLAB size is 128 KB, align the buffer on a multiple
531 // of 4 card clusters.
532 // 2. Post-process the contents of GCLAB buffers to register the
533 // objects allocated therein. Allow one GC thread at a
534 // time to do the post-processing of each GCLAB.
535 // 3. Since only one GC thread at a time is registering objects
536 // belonging to a particular allocation buffer, no locking
537 // is performed when registering these objects.
538 // 4. Any remnant of unallocated memory within an expended GC
539 // allocation buffer is not returned to the old-gen allocation
540 // pool until after the GC allocation buffer has been post
541 // processed. Before any remnant memory is returned to the
542 // old-gen allocation pool, the GC thread that scanned this GC
543 // allocation buffer performs a write-commit memory barrier.
544 // 5. Background GC threads that perform tenuring of young-gen
545 // objects without a GCLAB use a CAS lock before registering
546 // each tenured object. The CAS lock assures both mutual
547 // exclusion and memory coherency/visibility. Note that an
548 // object tenured by a background GC thread will not overlap
549 // with any of the clusters that are receiving tenured objects
550 // by way of GCLAB buffers. Multiple independent GC threads may
551 // attempt to tenure objects into a shared cluster. This is why
552 // sychronization may be necessary. Consider the following
553 // scenarios:
554 //
555 // a) If two objects are tenured into the same card region, each
556 // registration may attempt to modify the first-start or
557 // last-start information associated with that card region.
558 // Furthermore, because the representations of first-start
559 // and last-start information within the object_starts array
560 // entry uses different bits of a shared uint_16 to represent
561 // each, it is necessary to lock the entire card entry
562 // before modifying either the first-start or last-start
563 // information within the entry.
564 // b) Suppose GC thread X promotes a tenured object into
565 // card region A and this tenured object spans into
566 // neighboring card region B. Suppose GC thread Y (not equal
567 // to X) promotes a tenured object into cluster B. GC thread X
568 // will update the object_starts information for card A. No
569 // synchronization is required.
570 // c) In summary, when background GC threads register objects
571 // newly tenured into old-gen memory, they must acquire a
572 // mutual exclusion lock on the card that holds the starting
573 // address of the newly tenured object. This can be achieved
574 // by using a CAS instruction to assure that the previous
575 // values of first-offset and last-offset have not been
576 // changed since the same thread inquired as to their most
577 // current values.
578 //
579 // One way to minimize the need for synchronization between
580 // background tenuring GC threads is for each tenuring GC thread
581 // to promote young-gen objects into distinct dedicated cluster
582 // ranges.
583 // 6. The object_starts information is only required during the
584 // starting of concurrent marking and concurrent evacuation
585 // phases of GC. Before we start either of these GC phases, the
586 // JVM enters a safe point and all GC threads perform
587 // commit-write barriers to assure that access to the
588 // object_starts information is coherent.
589
590
591 // Notes on synchronization of register_object():
592 //
593 // 1. For efficiency, there is no locking in the implementation of register_object()
594 // 2. Thus, it is required that users of this service assure that concurrent/parallel invocations of
595 // register_object() do pertain to the same card's memory range. See discussion below to understand
596 // the risks.
597 // 3. When allocating from a TLAB or GCLAB, the mutual exclusion can be guaranteed by assuring that each
598 // LAB's start and end are aligned on card memory boundaries.
599 // 4. Use the same lock that guarantees exclusivity when performing free-list allocation within heap regions.
600 //
601 // Register the newly allocated object while we're holding the global lock since there's no synchronization
602 // built in to the implementation of register_object(). There are potential races when multiple independent
603 // threads are allocating objects, some of which might span the same card region. For example, consider
604 // a card table's memory region within which three objects are being allocated by three different threads:
605 //
606 // objects being "concurrently" allocated:
607 // [-----a------][-----b-----][--------------c------------------]
608 // [---- card table memory range --------------]
609 //
610 // Before any objects are allocated, this card's memory range holds no objects. Note that:
611 // allocation of object a wants to set the has-object, first-start, and last-start attributes of the preceding card region.
612 // allocation of object b wants to set the has-object, first-start, and last-start attributes of this card region.
613 // allocation of object c also wants to set the has-object, first-start, and last-start attributes of this card region.
614 //
615 // The thread allocating b and the thread allocating c can "race" in various ways, resulting in confusion, such as last-start
616 // representing object b while first-start represents object c. This is why we need to require all register_object()
617 // invocations associated with objects that are allocated from "free lists" to provide their own mutual exclusion locking
618 // mechanism.
619
620 // Reset the starts_object() information to false for all cards in the range between from and to.
621 void reset_object_range(HeapWord *from, HeapWord *to);
622
623 // register_object() requires that the caller hold the heap lock
624 // before calling it.
625 void register_object(HeapWord* address);
626
627 // register_object_without_lock() does not require that the caller hold
628 // the heap lock before calling it, under the assumption that the
629 // caller has assure no other thread will endeavor to concurrently
630 // register objects that start within the same card's memory region
631 // as address.
632 void register_object_without_lock(HeapWord* address);
633
634 // During the reference updates phase of GC, we walk through each old-gen memory region that was
635 // not part of the collection set and we invalidate all unmarked objects. As part of this effort,
636 // we coalesce neighboring dead objects in order to make future remembered set scanning more
637 // efficient (since future remembered set scanning of any card region containing consecutive
638 // dead objects can skip over all of them at once by reading only a single dead object header
639 // instead of having to read the header of each of the coalesced dead objects.
640 //
641 // At some future time, we may implement a further optimization: satisfy future allocation requests
642 // by carving new objects out of the range of memory that represents the coalesced dead objects.
643 //
644 // Suppose we want to combine several dead objects into a single coalesced object. How does this
645 // impact our representation of crossing map information?
646 // 1. If the newly coalesced range is contained entirely within a card range, that card's last
647 // start entry either remains the same or it is changed to the start of the coalesced region.
648 // 2. For the card that holds the start of the coalesced object, it will not impact the first start
649 // but it may impact the last start.
650 // 3. For following cards spanned entirely by the newly coalesced object, it will change starts_object
651 // to false (and make first-start and last-start "undefined").
652 // 4. For a following card that is spanned patially by the newly coalesced object, it may change
653 // first-start value, but it will not change the last-start value.
654 //
655 // The range of addresses represented by the arguments to coalesce_objects() must represent a range
656 // of memory that was previously occupied exactly by one or more previously registered objects. For
657 // convenience, it is legal to invoke coalesce_objects() with arguments that span a single previously
658 // registered object.
659 //
660 // The role of coalesce_objects is to change the crossing map information associated with all of the coalesced
661 // objects.
662 void coalesce_objects(HeapWord* address, size_t length_in_words);
663
664 // The typical use case is going to look something like this:
665 // for each heapregion that comprises old-gen memory
666 // for each card number that corresponds to this heap region
667 // scan the objects contained therein if the card is dirty
668 // To avoid excessive lookups in a sparse array, the API queries
669 // the card number pertaining to a particular address and then uses the
670 // card number for subsequent information lookups and stores.
671
672 // If starts_object(card_index), this returns the word offset within this card
673 // memory at which the first object begins. If !starts_object(card_index), the
674 // result is a don't care value -- asserts in a debug build.
675 size_t get_first_start(size_t card_index) const;
676
677 // If starts_object(card_index), this returns the word offset within this card
678 // memory at which the last object begins. If !starts_object(card_index), the
679 // result is a don't care value.
680 size_t get_last_start(size_t card_index) const;
681
682
683 // Given a card_index, return the starting address of the first block in the heap
684 // that straddles into the card. If the card is co-initial with an object, then
685 // this would return the starting address of the heap that this card covers.
686 // Expects to be called for a card affiliated with the old generation in
687 // generational mode.
688 HeapWord* block_start(size_t card_index) const;
689 };
690
691 // ShenandoahScanRemembered is a concrete class representing the
692 // ability to scan the old-gen remembered set for references to
693 // objects residing in young-gen memory.
694 //
695 // Scanning normally begins with an invocation of numRegions and ends
696 // after all clusters of all regions have been scanned.
697 //
698 // Throughout the scanning effort, the number of regions does not
699 // change.
700 //
701 // Even though the regions that comprise old-gen memory are not
702 // necessarily contiguous, the abstraction represented by this class
703 // identifies each of the old-gen regions with an integer value
704 // in the range from 0 to (numRegions() - 1) inclusive.
705 //
706
707 template<typename RememberedSet>
708 class ShenandoahScanRemembered: public CHeapObj<mtGC> {
709
710 private:
711 RememberedSet* _rs;
712 ShenandoahCardCluster<RememberedSet>* _scc;
713
714 // Global card stats (cumulative)
715 HdrSeq _card_stats_scan_rs[MAX_CARD_STAT_TYPE];
716 HdrSeq _card_stats_update_refs[MAX_CARD_STAT_TYPE];
717 // Per worker card stats (multiplexed by phase)
718 HdrSeq** _card_stats;
719
720 // The types of card metrics that we gather
721 const char* _card_stats_name[MAX_CARD_STAT_TYPE] = {
722 "dirty_run", "clean_run",
723 "dirty_cards", "clean_cards",
724 "max_dirty_run", "max_clean_run",
725 "dirty_scan_objs",
726 "alternations"
727 };
728
729 // The statistics are collected and logged separately for
730 // card-scans for initial marking, and for updating refs.
731 const char* _card_stat_log_type[MAX_CARD_STAT_LOG_TYPE] = {
732 "Scan Remembered Set", "Update Refs"
733 };
734
735 int _card_stats_log_counter[2] = {0, 0};
736
737 public:
738 // How to instantiate this object?
739 // ShenandoahDirectCardMarkRememberedSet *rs =
740 // new ShenandoahDirectCardMarkRememberedSet();
741 // scr = new
742 // ShenandoahScanRememberd<ShenandoahDirectCardMarkRememberedSet>(rs);
743 //
744 // or, after the planned implementation of
745 // ShenandoahBufferWithSATBRememberedSet has been completed:
746 //
747 // ShenandoahBufferWithSATBRememberedSet *rs =
748 // new ShenandoahBufferWithSATBRememberedSet();
749 // scr = new
750 // ShenandoahScanRememberd<ShenandoahBufferWithSATBRememberedSet>(rs);
751
752
753 ShenandoahScanRemembered(RememberedSet *rs) {
754 _rs = rs;
755 _scc = new ShenandoahCardCluster<RememberedSet>(rs);
756
757 // We allocate ParallelGCThreads worth even though we usually only
758 // use up to ConcGCThreads, because degenerate collections may employ
759 // ParallelGCThreads for remembered set scanning.
760 if (ShenandoahEnableCardStats) {
761 _card_stats = NEW_C_HEAP_ARRAY(HdrSeq*, ParallelGCThreads, mtGC);
762 for (uint i = 0; i < ParallelGCThreads; i++) {
763 _card_stats[i] = new HdrSeq[MAX_CARD_STAT_TYPE];
764 }
765 } else {
766 _card_stats = nullptr;
767 }
768 }
769
770 ~ShenandoahScanRemembered() {
771 delete _scc;
772 if (ShenandoahEnableCardStats) {
773 for (uint i = 0; i < ParallelGCThreads; i++) {
774 delete _card_stats[i];
775 }
776 FREE_C_HEAP_ARRAY(HdrSeq*, _card_stats);
777 _card_stats = nullptr;
778 }
779 assert(_card_stats == nullptr, "Error");
780 }
781
782 HdrSeq* card_stats(uint worker_id) {
783 assert(worker_id < ParallelGCThreads, "Error");
784 assert(ShenandoahEnableCardStats == (_card_stats != nullptr), "Error");
785 return ShenandoahEnableCardStats ? _card_stats[worker_id] : nullptr;
786 }
787
788 HdrSeq* card_stats_for_phase(CardStatLogType t) {
789 switch (t) {
790 case CARD_STAT_SCAN_RS:
791 return _card_stats_scan_rs;
792 case CARD_STAT_UPDATE_REFS:
793 return _card_stats_update_refs;
794 default:
795 guarantee(false, "No such CardStatLogType");
796 }
797 return nullptr; // Quiet compiler
798 }
799
800 // TODO: We really don't want to share all of these APIs with arbitrary consumers of the ShenandoahScanRemembered abstraction.
801 // But in the spirit of quick and dirty for the time being, I'm going to go ahead and publish everything for right now. Some
802 // of existing code already depends on having access to these services (because existing code has not been written to honor
803 // full abstraction of remembered set scanning. In the not too distant future, we want to try to make most, if not all, of
804 // these services private. Two problems with publicizing:
805 // 1. Allowing arbitrary users to reach beneath the hood allows the users to make assumptions about underlying implementation.
806 // This will make it more difficult to change underlying implementation at a future time, such as when we eventually experiment
807 // with SATB-based implementation of remembered set representation.
808 // 2. If we carefully control sharing of certain of these services, we can reduce the overhead of synchronization by assuring
809 // that all users follow protocols that avoid contention that might require synchronization. When we publish these APIs, we
810 // lose control over who and how the data is accessed. As a result, we are required to insert more defensive measures into
811 // the implementation, including synchronization locks.
812
813
814 // Card index is zero-based relative to first spanned card region.
815 size_t last_valid_index();
816 size_t total_cards();
817 size_t card_index_for_addr(HeapWord *p);
818 HeapWord *addr_for_card_index(size_t card_index);
819 bool is_card_dirty(size_t card_index);
820 bool is_write_card_dirty(size_t card_index) { return _rs->is_write_card_dirty(card_index); }
821 void mark_card_as_dirty(size_t card_index);
822 void mark_range_as_dirty(size_t card_index, size_t num_cards);
823 void mark_card_as_clean(size_t card_index);
824 void mark_read_card_as_clean(size_t card_index) { _rs->mark_read_card_clean(card_index); }
825 void mark_range_as_clean(size_t card_index, size_t num_cards);
826 bool is_card_dirty(HeapWord *p);
827 void mark_card_as_dirty(HeapWord *p);
828 void mark_range_as_dirty(HeapWord *p, size_t num_heap_words);
829 void mark_card_as_clean(HeapWord *p);
830 void mark_range_as_clean(HeapWord *p, size_t num_heap_words);
831 size_t cluster_count();
832
833 // Called by GC thread at start of concurrent mark to exchange roles of read and write remembered sets.
834 void swap_remset() { _rs->swap_remset(); }
835
836 void reset_remset(HeapWord* start, size_t word_count) { _rs->reset_remset(start, word_count); }
837
838 void merge_write_table(HeapWord* start, size_t word_count) { _rs->merge_write_table(start, word_count); }
839
840 // Called by GC thread after scanning old remembered set in order to prepare for next GC pass
841 void clear_old_remset() { _rs->clear_old_remset(); }
842
843 size_t cluster_for_addr(HeapWord *addr);
844 HeapWord* addr_for_cluster(size_t cluster_no);
845
846 void reset_object_range(HeapWord *from, HeapWord *to);
847 void register_object(HeapWord *addr);
848 void register_object_without_lock(HeapWord *addr);
849 void coalesce_objects(HeapWord *addr, size_t length_in_words);
850
851 HeapWord* first_object_in_card(size_t card_index) {
852 if (_scc->starts_object(card_index)) {
853 return addr_for_card_index(card_index) + _scc->get_first_start(card_index);
854 } else {
855 return nullptr;
856 }
857 }
858
859 // Return true iff this object is "properly" registered.
860 bool verify_registration(HeapWord* address, ShenandoahMarkingContext* ctx);
861
862 // clear the cards to clean, and clear the object_starts info to no objects
863 void mark_range_as_empty(HeapWord *addr, size_t length_in_words);
864
865 // process_clusters() scans a portion of the remembered set
866 // for references from old gen into young. Several worker threads
867 // scan different portions of the remembered set by making parallel invocations
868 // of process_clusters() with each invocation scanning different
869 // "clusters" of the remembered set.
870 //
871 // An invocation of process_clusters() examines all of the
872 // intergenerational references spanned by `count` clusters starting
873 // with `first_cluster`. The `oops` argument is a worker-thread-local
874 // OopClosure that is applied to all "valid" references in the remembered set.
875 //
876 // A side-effect of executing process_clusters() is to update the remembered
877 // set entries (e.g. marking dirty cards clean if they no longer
878 // hold references to young-gen memory).
879 //
880 // An implementation of process_clusters() may choose to efficiently
881 // address more typical scenarios in the structure of remembered sets. E.g.
882 // in the generational setting, one might expect remembered sets to be very sparse
883 // (low mutation rates in the old generation leading to sparse dirty cards,
884 // each with very few intergenerational pointers). Specific implementations
885 // may choose to degrade gracefully as the sparsity assumption fails to hold,
886 // such as when there are sudden spikes in (premature) promotion or in the
887 // case of an underprovisioned, poorly-tuned, or poorly-shaped heap.
888 //
889 // At the start of a concurrent young generation marking cycle, we invoke process_clusters
890 // with ClosureType ShenandoahInitMarkRootsClosure.
891 //
892 // At the start of a concurrent evacuation phase, we invoke process_clusters with
893 // ClosureType ShenandoahEvacuateUpdateRootsClosure.
894
895 // All template expansions require methods to be defined in the inline.hpp file, but larger
896 // such methods need not be declared as inline.
897 template <typename ClosureType>
898 void process_clusters(size_t first_cluster, size_t count, HeapWord *end_of_range, ClosureType *oops,
899 bool use_write_table, uint worker_id);
900
901 template <typename ClosureType>
902 inline void process_humongous_clusters(ShenandoahHeapRegion* r, size_t first_cluster, size_t count,
903 HeapWord *end_of_range, ClosureType *oops, bool use_write_table);
904
905 template <typename ClosureType>
906 inline void process_region_slice(ShenandoahHeapRegion* region, size_t offset, size_t clusters, HeapWord* end_of_range,
907 ClosureType *cl, bool use_write_table, uint worker_id);
908
909 // To Do:
910 // Create subclasses of ShenandoahInitMarkRootsClosure and
911 // ShenandoahEvacuateUpdateRootsClosure and any other closures
912 // that need to participate in remembered set scanning. Within the
913 // subclasses, add a (probably templated) instance variable that
914 // refers to the associated ShenandoahCardCluster object. Use this
915 // ShenandoahCardCluster instance to "enhance" the do_oops
916 // processing so that we can:
917 //
918 // 1. Avoid processing references that correspond to clean card
919 // regions, and
920 // 2. Set card status to CLEAN when the associated card region no
921 // longer holds inter-generatioanal references.
922 //
923 // To enable efficient implementation of these behaviors, we
924 // probably also want to add a few fields into the
925 // ShenandoahCardCluster object that allow us to precompute and
926 // remember the addresses at which card status is going to change
927 // from dirty to clean and clean to dirty. The do_oops
928 // implementations will want to update this value each time they
929 // cross one of these boundaries.
930 void roots_do(OopIterateClosure* cl);
931
932 // Log stats related to card/RS stats for given phase t
933 void log_card_stats(uint nworkers, CardStatLogType t) PRODUCT_RETURN;
934 private:
935 // Log stats for given worker id related into given summary card/RS stats
936 void log_worker_card_stats(uint worker_id, HdrSeq* sum_stats) PRODUCT_RETURN;
937
938 // Log given stats
939 inline void log_card_stats(HdrSeq* stats) PRODUCT_RETURN;
940
941 // Merge the stats from worked_id into the given summary stats, and clear the worker_id's stats.
942 void merge_worker_card_stats_cumulative(HdrSeq* worker_stats, HdrSeq* sum_stats) PRODUCT_RETURN;
943 };
944
945
946 // A ShenandoahRegionChunk represents a contiguous interval of a ShenandoahHeapRegion, typically representing
947 // work to be done by a worker thread.
948 struct ShenandoahRegionChunk {
949 ShenandoahHeapRegion *_r; // The region of which this represents a chunk
950 size_t _chunk_offset; // HeapWordSize offset
951 size_t _chunk_size; // HeapWordSize qty
952 };
953
954 // ShenandoahRegionChunkIterator divides the total remembered set scanning effort into ShenandoahRegionChunks
955 // that are assigned one at a time to worker threads. (Here, we use the terms `assignments` and `chunks`
956 // interchangeably.) Note that the effort required to scan a range of memory is not necessarily a linear
957 // function of the size of the range. Some memory ranges hold only a small number of live objects.
958 // Some ranges hold primarily primitive (non-pointer) data. We start with larger chunk sizes because larger chunks
959 // reduce coordination overhead. We expect that the GC worker threads that receive more difficult assignments
960 // will work longer on those chunks. Meanwhile, other worker will threads repeatedly accept and complete multiple
961 // easier chunks. As the total amount of work remaining to be completed decreases, we decrease the size of chunks
962 // given to individual threads. This reduces the likelihood of significant imbalance between worker thread assignments
963 // when there is less meaningful work to be performed by the remaining worker threads while they wait for
964 // worker threads with difficult assignments to finish, reducing the overall duration of the phase.
965
966 class ShenandoahRegionChunkIterator : public StackObj {
967 private:
968 // The largest chunk size is 4 MiB, measured in words. Otherwise, remembered set scanning may become too unbalanced.
969 // If the largest chunk size is too small, there is too much overhead sifting out assignments to individual worker threads.
970 static const size_t _maximum_chunk_size_words = (4 * 1024 * 1024) / HeapWordSize;
971
972 static const size_t _clusters_in_smallest_chunk = 4;
973
974 // smallest_chunk_size is 4 clusters. Each cluster spans 128 KiB.
975 // This is computed from CardTable::card_size_in_words() *
976 // ShenandoahCardCluster<ShenandoahDirectCardMarkRememberedSet>::CardsPerCluster;
977 static size_t smallest_chunk_size_words() {
978 return _clusters_in_smallest_chunk * CardTable::card_size_in_words() *
979 ShenandoahCardCluster<ShenandoahDirectCardMarkRememberedSet>::CardsPerCluster;
980 }
981
982 // The total remembered set scanning effort is divided into chunks of work that are assigned to individual worker tasks.
983 // The chunks of assigned work are divided into groups, where the size of the typical group (_regular_group_size) is half the
984 // total number of regions. The first group may be larger than
985 // _regular_group_size in the case that the first group's chunk
986 // size is less than the region size. The last group may be larger
987 // than _regular_group_size because no group is allowed to
988 // have smaller assignments than _smallest_chunk_size, which is 128 KB.
989
990 // Under normal circumstances, no configuration needs more than _maximum_groups (default value of 16).
991 // The first group "effectively" processes chunks of size 1 MiB (or smaller for smaller region sizes).
992 // The last group processes chunks of size 128 KiB. There are four groups total.
993
994 // group[0] is 4 MiB chunk size (_maximum_chunk_size_words)
995 // group[1] is 2 MiB chunk size
996 // group[2] is 1 MiB chunk size
997 // group[3] is 512 KiB chunk size
998 // group[4] is 256 KiB chunk size
999 // group[5] is 128 Kib shunk size (_smallest_chunk_size_words = 4 * 64 * 64
1000 static const size_t _maximum_groups = 6;
1001
1002 const ShenandoahHeap* _heap;
1003
1004 const size_t _regular_group_size; // Number of chunks in each group
1005 const size_t _first_group_chunk_size_b4_rebalance;
1006 const size_t _num_groups; // Number of groups in this configuration
1007 const size_t _total_chunks;
1008
1009 shenandoah_padding(0);
1010 volatile size_t _index;
1011 shenandoah_padding(1);
1012
1013 size_t _region_index[_maximum_groups]; // The region index for the first region spanned by this group
1014 size_t _group_offset[_maximum_groups]; // The offset at which group begins within first region spanned by this group
1015 size_t _group_chunk_size[_maximum_groups]; // The size of each chunk within this group
1016 size_t _group_entries[_maximum_groups]; // Total chunks spanned by this group and the ones before it.
1017
1018 // No implicit copying: iterators should be passed by reference to capture the state
1019 NONCOPYABLE(ShenandoahRegionChunkIterator);
1020
1021 // Makes use of _heap.
1022 size_t calc_regular_group_size();
1023
1024 // Makes use of _regular_group_size, which must be initialized before call.
1025 size_t calc_first_group_chunk_size_b4_rebalance();
1026
1027 // Makes use of _regular_group_size and _first_group_chunk_size_b4_rebalance, both of which must be initialized before call.
1028 size_t calc_num_groups();
1029
1030 // Makes use of _regular_group_size, _first_group_chunk_size_b4_rebalance, which must be initialized before call.
1031 size_t calc_total_chunks();
1032
1033 public:
1034 ShenandoahRegionChunkIterator(size_t worker_count);
1035 ShenandoahRegionChunkIterator(ShenandoahHeap* heap, size_t worker_count);
1036
1037 // Reset iterator to default state
1038 void reset();
1039
1040 // Fills in assignment with next chunk of work and returns true iff there is more work.
1041 // Otherwise, returns false. This is multi-thread-safe.
1042 inline bool next(struct ShenandoahRegionChunk *assignment);
1043
1044 // This is *not* MT safe. However, in the absence of multithreaded access, it
1045 // can be used to determine if there is more work to do.
1046 inline bool has_next() const;
1047 };
1048
1049 typedef ShenandoahScanRemembered<ShenandoahDirectCardMarkRememberedSet> RememberedScanner;
1050
1051 class ShenandoahScanRememberedTask : public WorkerTask {
1052 private:
1053 ShenandoahObjToScanQueueSet* _queue_set;
1054 ShenandoahObjToScanQueueSet* _old_queue_set;
1055 ShenandoahReferenceProcessor* _rp;
1056 ShenandoahRegionChunkIterator* _work_list;
1057 bool _is_concurrent;
1058
1059 public:
1060 ShenandoahScanRememberedTask(ShenandoahObjToScanQueueSet* queue_set,
1061 ShenandoahObjToScanQueueSet* old_queue_set,
1062 ShenandoahReferenceProcessor* rp,
1063 ShenandoahRegionChunkIterator* work_list,
1064 bool is_concurrent);
1065
1066 void work(uint worker_id);
1067 void do_work(uint worker_id);
1068 };
1069
1070 // Verify that the oop doesn't point into the young generation
1071 class ShenandoahVerifyNoYoungRefsClosure: public BasicOopIterateClosure {
1072 ShenandoahHeap* _heap;
1073 template<class T> void work(T* p);
1074
1075 public:
1076 ShenandoahVerifyNoYoungRefsClosure();
1077
1078 virtual void do_oop(narrowOop* p) { work(p); }
1079 virtual void do_oop(oop* p) { work(p); }
1080 };
1081
1082 #endif // SHARE_GC_SHENANDOAH_SHENANDOAHSCANREMEMBERED_HPP