1 /*
2 * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #ifndef SHARE_VM_GC_IMPLEMENTATION_CONCURRENTMARKSWEEP_COMPACTIBLEFREELISTSPACE_HPP
26 #define SHARE_VM_GC_IMPLEMENTATION_CONCURRENTMARKSWEEP_COMPACTIBLEFREELISTSPACE_HPP
27
28 #include "gc_implementation/concurrentMarkSweep/adaptiveFreeList.hpp"
29 #include "gc_implementation/concurrentMarkSweep/promotionInfo.hpp"
30 #include "memory/binaryTreeDictionary.hpp"
31 #include "memory/blockOffsetTable.inline.hpp"
32 #include "memory/freeList.hpp"
33 #include "memory/space.hpp"
34
35 // Classes in support of keeping track of promotions into a non-Contiguous
36 // space, in this case a CompactibleFreeListSpace.
37
38 // Forward declarations
39 class CompactibleFreeListSpace;
40 class BlkClosure;
41 class BlkClosureCareful;
42 class FreeChunk;
43 class UpwardsObjectClosure;
44 class ObjectClosureCareful;
45 class Klass;
46
47 class LinearAllocBlock VALUE_OBJ_CLASS_SPEC {
48 public:
LinearAllocBlock()49 LinearAllocBlock() : _ptr(0), _word_size(0), _refillSize(0),
50 _allocation_size_limit(0) {}
set(HeapWord * ptr,size_t word_size,size_t refill_size,size_t allocation_size_limit)51 void set(HeapWord* ptr, size_t word_size, size_t refill_size,
52 size_t allocation_size_limit) {
53 _ptr = ptr;
54 _word_size = word_size;
55 _refillSize = refill_size;
56 _allocation_size_limit = allocation_size_limit;
57 }
58 HeapWord* _ptr;
59 size_t _word_size;
60 size_t _refillSize;
61 size_t _allocation_size_limit; // largest size that will be allocated
62
63 void print_on(outputStream* st) const;
64 };
65
66 // Concrete subclass of CompactibleSpace that implements
67 // a free list space, such as used in the concurrent mark sweep
68 // generation.
69
70 class CompactibleFreeListSpace: public CompactibleSpace {
71 friend class VMStructs;
72 friend class ConcurrentMarkSweepGeneration;
73 friend class ASConcurrentMarkSweepGeneration;
74 friend class CMSCollector;
75 // Local alloc buffer for promotion into this space.
76 friend class CFLS_LAB;
77
78 // "Size" of chunks of work (executed during parallel remark phases
79 // of CMS collection); this probably belongs in CMSCollector, although
80 // it's cached here because it's used in
81 // initialize_sequential_subtasks_for_rescan() which modifies
82 // par_seq_tasks which also lives in Space. XXX
83 const size_t _rescan_task_size;
84 const size_t _marking_task_size;
85
86 // Yet another sequential tasks done structure. This supports
87 // CMS GC, where we have threads dynamically
88 // claiming sub-tasks from a larger parallel task.
89 SequentialSubTasksDone _conc_par_seq_tasks;
90
91 BlockOffsetArrayNonContigSpace _bt;
92
93 CMSCollector* _collector;
94 ConcurrentMarkSweepGeneration* _gen;
95
96 // Data structures for free blocks (used during allocation/sweeping)
97
98 // Allocation is done linearly from two different blocks depending on
99 // whether the request is small or large, in an effort to reduce
100 // fragmentation. We assume that any locking for allocation is done
101 // by the containing generation. Thus, none of the methods in this
102 // space are re-entrant.
103 enum SomeConstants {
104 SmallForLinearAlloc = 16, // size < this then use _sLAB
105 SmallForDictionary = 257, // size < this then use _indexedFreeList
106 IndexSetSize = SmallForDictionary // keep this odd-sized
107 };
108 static size_t IndexSetStart;
109 static size_t IndexSetStride;
110
111 private:
112 enum FitStrategyOptions {
113 FreeBlockStrategyNone = 0,
114 FreeBlockBestFitFirst
115 };
116
117 PromotionInfo _promoInfo;
118
119 // helps to impose a global total order on freelistLock ranks;
120 // assumes that CFLSpace's are allocated in global total order
121 static int _lockRank;
122
123 // a lock protecting the free lists and free blocks;
124 // mutable because of ubiquity of locking even for otherwise const methods
125 mutable Mutex _freelistLock;
126 // locking verifier convenience function
127 void assert_locked() const PRODUCT_RETURN;
128 void assert_locked(const Mutex* lock) const PRODUCT_RETURN;
129
130 // Linear allocation blocks
131 LinearAllocBlock _smallLinearAllocBlock;
132
133 FreeBlockDictionary<FreeChunk>::DictionaryChoice _dictionaryChoice;
134 AFLBinaryTreeDictionary* _dictionary; // ptr to dictionary for large size blocks
135
136 AdaptiveFreeList<FreeChunk> _indexedFreeList[IndexSetSize];
137 // indexed array for small size blocks
138 // allocation stategy
139 bool _fitStrategy; // Use best fit strategy.
140 bool _adaptive_freelists; // Use adaptive freelists
141
142 // This is an address close to the largest free chunk in the heap.
143 // It is currently assumed to be at the end of the heap. Free
144 // chunks with addresses greater than nearLargestChunk are coalesced
145 // in an effort to maintain a large chunk at the end of the heap.
146 HeapWord* _nearLargestChunk;
147
148 // Used to keep track of limit of sweep for the space
149 HeapWord* _sweep_limit;
150
151 // Stable value of used().
152 size_t _used_stable;
153
154 // Support for compacting cms
155 HeapWord* cross_threshold(HeapWord* start, HeapWord* end);
156 HeapWord* forward(oop q, size_t size, CompactPoint* cp, HeapWord* compact_top);
157
158 // Initialization helpers.
159 void initializeIndexedFreeListArray();
160
161 // Extra stuff to manage promotion parallelism.
162
163 // a lock protecting the dictionary during par promotion allocation.
164 mutable Mutex _parDictionaryAllocLock;
parDictionaryAllocLock() const165 Mutex* parDictionaryAllocLock() const { return &_parDictionaryAllocLock; }
166
167 // Locks protecting the exact lists during par promotion allocation.
168 Mutex* _indexedFreeListParLocks[IndexSetSize];
169
170 // Attempt to obtain up to "n" blocks of the size "word_sz" (which is
171 // required to be smaller than "IndexSetSize".) If successful,
172 // adds them to "fl", which is required to be an empty free list.
173 // If the count of "fl" is negative, it's absolute value indicates a
174 // number of free chunks that had been previously "borrowed" from global
175 // list of size "word_sz", and must now be decremented.
176 void par_get_chunk_of_blocks(size_t word_sz, size_t n, AdaptiveFreeList<FreeChunk>* fl);
177
178 // Used by par_get_chunk_of_blocks() for the chunks from the
179 // indexed_free_lists.
180 bool par_get_chunk_of_blocks_IFL(size_t word_sz, size_t n, AdaptiveFreeList<FreeChunk>* fl);
181
182 // Used by par_get_chunk_of_blocks_dictionary() to get a chunk
183 // evenly splittable into "n" "word_sz" chunks. Returns that
184 // evenly splittable chunk. May split a larger chunk to get the
185 // evenly splittable chunk.
186 FreeChunk* get_n_way_chunk_to_split(size_t word_sz, size_t n);
187
188 // Used by par_get_chunk_of_blocks() for the chunks from the
189 // dictionary.
190 void par_get_chunk_of_blocks_dictionary(size_t word_sz, size_t n, AdaptiveFreeList<FreeChunk>* fl);
191
192 // Allocation helper functions
193 // Allocate using a strategy that takes from the indexed free lists
194 // first. This allocation strategy assumes a companion sweeping
195 // strategy that attempts to keep the needed number of chunks in each
196 // indexed free lists.
197 HeapWord* allocate_adaptive_freelists(size_t size);
198 // Allocate from the linear allocation buffers first. This allocation
199 // strategy assumes maximal coalescing can maintain chunks large enough
200 // to be used as linear allocation buffers.
201 HeapWord* allocate_non_adaptive_freelists(size_t size);
202
203 // Gets a chunk from the linear allocation block (LinAB). If there
204 // is not enough space in the LinAB, refills it.
205 HeapWord* getChunkFromLinearAllocBlock(LinearAllocBlock* blk, size_t size);
206 HeapWord* getChunkFromSmallLinearAllocBlock(size_t size);
207 // Get a chunk from the space remaining in the linear allocation block. Do
208 // not attempt to refill if the space is not available, return NULL. Do the
209 // repairs on the linear allocation block as appropriate.
210 HeapWord* getChunkFromLinearAllocBlockRemainder(LinearAllocBlock* blk, size_t size);
211 inline HeapWord* getChunkFromSmallLinearAllocBlockRemainder(size_t size);
212
213 // Helper function for getChunkFromIndexedFreeList.
214 // Replenish the indexed free list for this "size". Do not take from an
215 // underpopulated size.
216 FreeChunk* getChunkFromIndexedFreeListHelper(size_t size, bool replenish = true);
217
218 // Get a chunk from the indexed free list. If the indexed free list
219 // does not have a free chunk, try to replenish the indexed free list
220 // then get the free chunk from the replenished indexed free list.
221 inline FreeChunk* getChunkFromIndexedFreeList(size_t size);
222
223 // The returned chunk may be larger than requested (or null).
224 FreeChunk* getChunkFromDictionary(size_t size);
225 // The returned chunk is the exact size requested (or null).
226 FreeChunk* getChunkFromDictionaryExact(size_t size);
227
228 // Find a chunk in the indexed free list that is the best
229 // fit for size "numWords".
230 FreeChunk* bestFitSmall(size_t numWords);
231 // For free list "fl" of chunks of size > numWords,
232 // remove a chunk, split off a chunk of size numWords
233 // and return it. The split off remainder is returned to
234 // the free lists. The old name for getFromListGreater
235 // was lookInListGreater.
236 FreeChunk* getFromListGreater(AdaptiveFreeList<FreeChunk>* fl, size_t numWords);
237 // Get a chunk in the indexed free list or dictionary,
238 // by considering a larger chunk and splitting it.
239 FreeChunk* getChunkFromGreater(size_t numWords);
240 // Verify that the given chunk is in the indexed free lists.
241 bool verifyChunkInIndexedFreeLists(FreeChunk* fc) const;
242 // Remove the specified chunk from the indexed free lists.
243 void removeChunkFromIndexedFreeList(FreeChunk* fc);
244 // Remove the specified chunk from the dictionary.
245 void removeChunkFromDictionary(FreeChunk* fc);
246 // Split a free chunk into a smaller free chunk of size "new_size".
247 // Return the smaller free chunk and return the remainder to the
248 // free lists.
249 FreeChunk* splitChunkAndReturnRemainder(FreeChunk* chunk, size_t new_size);
250 // Add a chunk to the free lists.
251 void addChunkToFreeLists(HeapWord* chunk, size_t size);
252 // Add a chunk to the free lists, preferring to suffix it
253 // to the last free chunk at end of space if possible, and
254 // updating the block census stats as well as block offset table.
255 // Take any locks as appropriate if we are multithreaded.
256 void addChunkToFreeListsAtEndRecordingStats(HeapWord* chunk, size_t size);
257 // Add a free chunk to the indexed free lists.
258 void returnChunkToFreeList(FreeChunk* chunk);
259 // Add a free chunk to the dictionary.
260 void returnChunkToDictionary(FreeChunk* chunk);
261
262 // Functions for maintaining the linear allocation buffers (LinAB).
263 // Repairing a linear allocation block refers to operations
264 // performed on the remainder of a LinAB after an allocation
265 // has been made from it.
266 void repairLinearAllocationBlocks();
267 void repairLinearAllocBlock(LinearAllocBlock* blk);
268 void refillLinearAllocBlock(LinearAllocBlock* blk);
269 void refillLinearAllocBlockIfNeeded(LinearAllocBlock* blk);
270 void refillLinearAllocBlocksIfNeeded();
271
272 void verify_objects_initialized() const;
273
274 // Statistics reporting helper functions
275 void reportFreeListStatistics() const;
276 void reportIndexedFreeListStatistics() const;
277 size_t maxChunkSizeInIndexedFreeLists() const;
278 size_t numFreeBlocksInIndexedFreeLists() const;
279 // Accessor
unallocated_block() const280 HeapWord* unallocated_block() const {
281 if (BlockOffsetArrayUseUnallocatedBlock) {
282 HeapWord* ub = _bt.unallocated_block();
283 assert(ub >= bottom() &&
284 ub <= end(), "space invariant");
285 return ub;
286 } else {
287 return end();
288 }
289 }
freed(HeapWord * start,size_t size)290 void freed(HeapWord* start, size_t size) {
291 _bt.freed(start, size);
292 }
293
294 protected:
295 // reset the indexed free list to its initial empty condition.
296 void resetIndexedFreeListArray();
297 // reset to an initial state with a single free block described
298 // by the MemRegion parameter.
299 void reset(MemRegion mr);
300 // Return the total number of words in the indexed free lists.
301 size_t totalSizeInIndexedFreeLists() const;
302
303 public:
304 // Constructor...
305 CompactibleFreeListSpace(BlockOffsetSharedArray* bs, MemRegion mr,
306 bool use_adaptive_freelists,
307 FreeBlockDictionary<FreeChunk>::DictionaryChoice);
308 // accessors
bestFitFirst()309 bool bestFitFirst() { return _fitStrategy == FreeBlockBestFitFirst; }
dictionary() const310 FreeBlockDictionary<FreeChunk>* dictionary() const { return _dictionary; }
nearLargestChunk() const311 HeapWord* nearLargestChunk() const { return _nearLargestChunk; }
set_nearLargestChunk(HeapWord * v)312 void set_nearLargestChunk(HeapWord* v) { _nearLargestChunk = v; }
313
314 // Set CMS global values
315 static void set_cms_values();
316
317 // Return the free chunk at the end of the space. If no such
318 // chunk exists, return NULL.
319 FreeChunk* find_chunk_at_end();
320
adaptive_freelists() const321 bool adaptive_freelists() const { return _adaptive_freelists; }
322
set_collector(CMSCollector * collector)323 void set_collector(CMSCollector* collector) { _collector = collector; }
324
325 // Support for parallelization of rescan and marking
rescan_task_size() const326 const size_t rescan_task_size() const { return _rescan_task_size; }
marking_task_size() const327 const size_t marking_task_size() const { return _marking_task_size; }
conc_par_seq_tasks()328 SequentialSubTasksDone* conc_par_seq_tasks() {return &_conc_par_seq_tasks; }
329 void initialize_sequential_subtasks_for_rescan(int n_threads);
330 void initialize_sequential_subtasks_for_marking(int n_threads,
331 HeapWord* low = NULL);
332
333 // Space enquiries
334 size_t used() const;
335 size_t free() const;
336 size_t max_alloc_in_words() const;
337 // XXX: should have a less conservative used_region() than that of
338 // Space; we could consider keeping track of highest allocated
339 // address and correcting that at each sweep, as the sweeper
340 // goes through the entire allocated part of the generation. We
341 // could also use that information to keep the sweeper from
342 // sweeping more than is necessary. The allocator and sweeper will
343 // of course need to synchronize on this, since the sweeper will
344 // try to bump down the address and the allocator will try to bump it up.
345 // For now, however, we'll just use the default used_region()
346 // which overestimates the region by returning the entire
347 // committed region (this is safe, but inefficient).
348
349 // Returns monotonically increasing stable used space bytes for CMS.
350 // This is required for jstat and other memory monitoring tools
351 // that might otherwise see inconsistent used space values during a garbage
352 // collection, promotion or allocation into compactibleFreeListSpace.
353 // The value returned by this function might be smaller than the
354 // actual value.
355 size_t used_stable() const;
356 // Recalculate and cache the current stable used() value. Only to be called
357 // in places where we can be sure that the result is stable.
358 void recalculate_used_stable();
359
360 // Returns a subregion of the space containing all the objects in
361 // the space.
used_region() const362 MemRegion used_region() const {
363 return MemRegion(bottom(),
364 BlockOffsetArrayUseUnallocatedBlock ?
365 unallocated_block() : end());
366 }
367
368 virtual bool is_free_block(const HeapWord* p) const;
369
370 // Resizing support
371 void set_end(HeapWord* value); // override
372
373 // mutual exclusion support
freelistLock() const374 Mutex* freelistLock() const { return &_freelistLock; }
375
376 // Iteration support
377 void oop_iterate(ExtendedOopClosure* cl);
378
379 void object_iterate(ObjectClosure* blk);
380 // Apply the closure to each object in the space whose references
381 // point to objects in the heap. The usage of CompactibleFreeListSpace
382 // by the ConcurrentMarkSweepGeneration for concurrent GC's allows
383 // objects in the space with references to objects that are no longer
384 // valid. For example, an object may reference another object
385 // that has already been sweep up (collected). This method uses
386 // obj_is_alive() to determine whether it is safe to iterate of
387 // an object.
388 void safe_object_iterate(ObjectClosure* blk);
389
390 // Iterate over all objects that intersect with mr, calling "cl->do_object"
391 // on each. There is an exception to this: if this closure has already
392 // been invoked on an object, it may skip such objects in some cases. This is
393 // Most likely to happen in an "upwards" (ascending address) iteration of
394 // MemRegions.
395 void object_iterate_mem(MemRegion mr, UpwardsObjectClosure* cl);
396
397 // Requires that "mr" be entirely within the space.
398 // Apply "cl->do_object" to all objects that intersect with "mr".
399 // If the iteration encounters an unparseable portion of the region,
400 // terminate the iteration and return the address of the start of the
401 // subregion that isn't done. Return of "NULL" indicates that the
402 // interation completed.
403 HeapWord* object_iterate_careful_m(MemRegion mr,
404 ObjectClosureCareful* cl);
405
406 // Override: provides a DCTO_CL specific to this kind of space.
407 DirtyCardToOopClosure* new_dcto_cl(ExtendedOopClosure* cl,
408 CardTableModRefBS::PrecisionStyle precision,
409 HeapWord* boundary);
410
411 void blk_iterate(BlkClosure* cl);
412 void blk_iterate_careful(BlkClosureCareful* cl);
413 HeapWord* block_start_const(const void* p) const;
414 HeapWord* block_start_careful(const void* p) const;
415 size_t block_size(const HeapWord* p) const;
416 size_t block_size_no_stall(HeapWord* p, const CMSCollector* c) const;
417 bool block_is_obj(const HeapWord* p) const;
418 bool obj_is_alive(const HeapWord* p) const;
419 size_t block_size_nopar(const HeapWord* p) const;
420 bool block_is_obj_nopar(const HeapWord* p) const;
421
422 // iteration support for promotion
423 void save_marks();
424 bool no_allocs_since_save_marks();
425
426 // iteration support for sweeping
save_sweep_limit()427 void save_sweep_limit() {
428 _sweep_limit = BlockOffsetArrayUseUnallocatedBlock ?
429 unallocated_block() : end();
430 if (CMSTraceSweeper) {
431 gclog_or_tty->print_cr(">>>>> Saving sweep limit " PTR_FORMAT
432 " for space [" PTR_FORMAT "," PTR_FORMAT ") <<<<<<",
433 p2i(_sweep_limit), p2i(bottom()), p2i(end()));
434 }
435 }
NOT_PRODUCT(void clear_sweep_limit (){ _sweep_limit = NULL; } )436 NOT_PRODUCT(
437 void clear_sweep_limit() { _sweep_limit = NULL; }
438 )
439 HeapWord* sweep_limit() { return _sweep_limit; }
440
441 // Apply "blk->do_oop" to the addresses of all reference fields in objects
442 // promoted into this generation since the most recent save_marks() call.
443 // Fields in objects allocated by applications of the closure
444 // *are* included in the iteration. Thus, when the iteration completes
445 // there should be no further such objects remaining.
446 #define CFLS_OOP_SINCE_SAVE_MARKS_DECL(OopClosureType, nv_suffix) \
447 void oop_since_save_marks_iterate##nv_suffix(OopClosureType* blk);
448 ALL_SINCE_SAVE_MARKS_CLOSURES(CFLS_OOP_SINCE_SAVE_MARKS_DECL)
449 #undef CFLS_OOP_SINCE_SAVE_MARKS_DECL
450
451 // Allocation support
452 HeapWord* allocate(size_t size);
453 HeapWord* par_allocate(size_t size);
454
455 oop promote(oop obj, size_t obj_size);
456 void gc_prologue();
457 void gc_epilogue();
458
459 // This call is used by a containing CMS generation / collector
460 // to inform the CFLS space that a sweep has been completed
461 // and that the space can do any related house-keeping functions.
462 void sweep_completed();
463
464 // For an object in this space, the mark-word's two
465 // LSB's having the value [11] indicates that it has been
466 // promoted since the most recent call to save_marks() on
467 // this generation and has not subsequently been iterated
468 // over (using oop_since_save_marks_iterate() above).
469 // This property holds only for single-threaded collections,
470 // and is typically used for Cheney scans; for MT scavenges,
471 // the property holds for all objects promoted during that
472 // scavenge for the duration of the scavenge and is used
473 // by card-scanning to avoid scanning objects (being) promoted
474 // during that scavenge.
obj_allocated_since_save_marks(const oop obj) const475 bool obj_allocated_since_save_marks(const oop obj) const {
476 assert(is_in_reserved(obj), "Wrong space?");
477 return ((PromotedObject*)obj)->hasPromotedMark();
478 }
479
480 // A worst-case estimate of the space required (in HeapWords) to expand the
481 // heap when promoting an obj of size obj_size.
482 size_t expansionSpaceRequired(size_t obj_size) const;
483
484 FreeChunk* allocateScratch(size_t size);
485
486 // returns true if either the small or large linear allocation buffer is empty.
487 bool linearAllocationWouldFail() const;
488
489 // Adjust the chunk for the minimum size. This version is called in
490 // most cases in CompactibleFreeListSpace methods.
adjustObjectSize(size_t size)491 inline static size_t adjustObjectSize(size_t size) {
492 return (size_t) align_object_size(MAX2(size, (size_t)MinChunkSize));
493 }
494 // This is a virtual version of adjustObjectSize() that is called
495 // only occasionally when the compaction space changes and the type
496 // of the new compaction space is is only known to be CompactibleSpace.
adjust_object_size_v(size_t size) const497 size_t adjust_object_size_v(size_t size) const {
498 return adjustObjectSize(size);
499 }
500 // Minimum size of a free block.
minimum_free_block_size() const501 virtual size_t minimum_free_block_size() const { return MinChunkSize; }
502 void removeFreeChunkFromFreeLists(FreeChunk* chunk);
503 void addChunkAndRepairOffsetTable(HeapWord* chunk, size_t size,
504 bool coalesced);
505
506 // Support for decisions regarding concurrent collection policy
507 bool should_concurrent_collect() const;
508
509 // Support for compaction
510 void prepare_for_compaction(CompactPoint* cp);
511 void adjust_pointers();
512 void compact();
513 // reset the space to reflect the fact that a compaction of the
514 // space has been done.
515 virtual void reset_after_compaction();
516
517 // Debugging support
518 void print() const;
519 void print_on(outputStream* st) const;
520 void prepare_for_verify();
521 void verify() const;
522 void verifyFreeLists() const PRODUCT_RETURN;
523 void verifyIndexedFreeLists() const;
524 void verifyIndexedFreeList(size_t size) const;
525 // Verify that the given chunk is in the free lists:
526 // i.e. either the binary tree dictionary, the indexed free lists
527 // or the linear allocation block.
528 bool verify_chunk_in_free_list(FreeChunk* fc) const;
529 // Verify that the given chunk is the linear allocation block
530 bool verify_chunk_is_linear_alloc_block(FreeChunk* fc) const;
531 // Do some basic checks on the the free lists.
532 void check_free_list_consistency() const PRODUCT_RETURN;
533
534 // Printing support
535 void dump_at_safepoint_with_locks(CMSCollector* c, outputStream* st);
536 void print_indexed_free_lists(outputStream* st) const;
537 void print_dictionary_free_lists(outputStream* st) const;
538 void print_promo_info_blocks(outputStream* st) const;
539
540 NOT_PRODUCT (
541 void initializeIndexedFreeListArrayReturnedBytes();
542 size_t sumIndexedFreeListArrayReturnedBytes();
543 // Return the total number of chunks in the indexed free lists.
544 size_t totalCountInIndexedFreeLists() const;
545 // Return the total numberof chunks in the space.
546 size_t totalCount();
547 )
548
549 // The census consists of counts of the quantities such as
550 // the current count of the free chunks, number of chunks
551 // created as a result of the split of a larger chunk or
552 // coalescing of smaller chucks, etc. The counts in the
553 // census is used to make decisions on splitting and
554 // coalescing of chunks during the sweep of garbage.
555
556 // Print the statistics for the free lists.
557 void printFLCensus(size_t sweep_count) const;
558
559 // Statistics functions
560 // Initialize census for lists before the sweep.
561 void beginSweepFLCensus(float inter_sweep_current,
562 float inter_sweep_estimate,
563 float intra_sweep_estimate);
564 // Set the surplus for each of the free lists.
565 void setFLSurplus();
566 // Set the hint for each of the free lists.
567 void setFLHints();
568 // Clear the census for each of the free lists.
569 void clearFLCensus();
570 // Perform functions for the census after the end of the sweep.
571 void endSweepFLCensus(size_t sweep_count);
572 // Return true if the count of free chunks is greater
573 // than the desired number of free chunks.
574 bool coalOverPopulated(size_t size);
575
576 // Record (for each size):
577 //
578 // split-births = #chunks added due to splits in (prev-sweep-end,
579 // this-sweep-start)
580 // split-deaths = #chunks removed for splits in (prev-sweep-end,
581 // this-sweep-start)
582 // num-curr = #chunks at start of this sweep
583 // num-prev = #chunks at end of previous sweep
584 //
585 // The above are quantities that are measured. Now define:
586 //
587 // num-desired := num-prev + split-births - split-deaths - num-curr
588 //
589 // Roughly, num-prev + split-births is the supply,
590 // split-deaths is demand due to other sizes
591 // and num-curr is what we have left.
592 //
593 // Thus, num-desired is roughly speaking the "legitimate demand"
594 // for blocks of this size and what we are striving to reach at the
595 // end of the current sweep.
596 //
597 // For a given list, let num-len be its current population.
598 // Define, for a free list of a given size:
599 //
600 // coal-overpopulated := num-len >= num-desired * coal-surplus
601 // (coal-surplus is set to 1.05, i.e. we allow a little slop when
602 // coalescing -- we do not coalesce unless we think that the current
603 // supply has exceeded the estimated demand by more than 5%).
604 //
605 // For the set of sizes in the binary tree, which is neither dense nor
606 // closed, it may be the case that for a particular size we have never
607 // had, or do not now have, or did not have at the previous sweep,
608 // chunks of that size. We need to extend the definition of
609 // coal-overpopulated to such sizes as well:
610 //
611 // For a chunk in/not in the binary tree, extend coal-overpopulated
612 // defined above to include all sizes as follows:
613 //
614 // . a size that is non-existent is coal-overpopulated
615 // . a size that has a num-desired <= 0 as defined above is
616 // coal-overpopulated.
617 //
618 // Also define, for a chunk heap-offset C and mountain heap-offset M:
619 //
620 // close-to-mountain := C >= 0.99 * M
621 //
622 // Now, the coalescing strategy is:
623 //
624 // Coalesce left-hand chunk with right-hand chunk if and
625 // only if:
626 //
627 // EITHER
628 // . left-hand chunk is of a size that is coal-overpopulated
629 // OR
630 // . right-hand chunk is close-to-mountain
631 void smallCoalBirth(size_t size);
632 void smallCoalDeath(size_t size);
633 void coalBirth(size_t size);
634 void coalDeath(size_t size);
635 void smallSplitBirth(size_t size);
636 void smallSplitDeath(size_t size);
637 void split_birth(size_t size);
638 void splitDeath(size_t size);
639 void split(size_t from, size_t to1);
640
641 double flsFrag() const;
642 };
643
644 // A parallel-GC-thread-local allocation buffer for allocation into a
645 // CompactibleFreeListSpace.
646 class CFLS_LAB : public CHeapObj<mtGC> {
647 // The space that this buffer allocates into.
648 CompactibleFreeListSpace* _cfls;
649
650 // Our local free lists.
651 AdaptiveFreeList<FreeChunk> _indexedFreeList[CompactibleFreeListSpace::IndexSetSize];
652
653 // Initialized from a command-line arg.
654
655 // Allocation statistics in support of dynamic adjustment of
656 // #blocks to claim per get_from_global_pool() call below.
657 static AdaptiveWeightedAverage
658 _blocks_to_claim [CompactibleFreeListSpace::IndexSetSize];
659 static size_t _global_num_blocks [CompactibleFreeListSpace::IndexSetSize];
660 static uint _global_num_workers[CompactibleFreeListSpace::IndexSetSize];
661 size_t _num_blocks [CompactibleFreeListSpace::IndexSetSize];
662
663 // Internal work method
664 void get_from_global_pool(size_t word_sz, AdaptiveFreeList<FreeChunk>* fl);
665
666 public:
667 CFLS_LAB(CompactibleFreeListSpace* cfls);
668
669 // Allocate and return a block of the given size, or else return NULL.
670 HeapWord* alloc(size_t word_sz);
671
672 // Return any unused portions of the buffer to the global pool.
673 void retire(int tid);
674
675 // Dynamic OldPLABSize sizing
676 static void compute_desired_plab_size();
677 // When the settings are modified from default static initialization
678 static void modify_initialization(size_t n, unsigned wt);
679 };
680
refillSize() const681 size_t PromotionInfo::refillSize() const {
682 const size_t CMSSpoolBlockSize = 256;
683 const size_t sz = heap_word_size(sizeof(SpoolBlock) + sizeof(markOop)
684 * CMSSpoolBlockSize);
685 return CompactibleFreeListSpace::adjustObjectSize(sz);
686 }
687
688 #endif // SHARE_VM_GC_IMPLEMENTATION_CONCURRENTMARKSWEEP_COMPACTIBLEFREELISTSPACE_HPP
689