1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
|
/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* Copyright by The HDF Group. *
* Copyright by the Board of Trustees of the University of Illinois. *
* All rights reserved. *
* *
* This file is part of HDF5. The full HDF5 copyright notice, including *
* terms governing use, modification, and redistribution, is contained in *
* the files COPYING and Copyright.html. COPYING can be found at the root *
* of the source code distribution tree; Copyright.html can be found at the *
* root level of an installed copy of the electronic HDF5 document set and *
* is linked from the top-level documents page. It can also be found at *
* http://hdfgroup.org/HDF5/doc/Copyright.html. If you do not have *
* access to either file, you may request a copy from help@hdfgroup.org. *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
/*-------------------------------------------------------------------------
*
* Created: H5C1private.h
* 6/3/04
* John Mainzer
*
* Purpose: Constants and typedefs available to the rest of the
* library.
*
* Modifications:
*
*-------------------------------------------------------------------------
*/
#ifndef _H5C1private_H
#define _H5C1private_H
#include "H5C1public.h" /*public prototypes */
/* Pivate headers needed by this header */
#include "H5private.h" /* Generic Functions */
#include "H5Fprivate.h" /* File access */
#define H5C1_DO_SANITY_CHECKS 0
#define H5C1_DO_EXTREME_SANITY_CHECKS 0
/* This sanity checking constant was picked out of the air. Increase
* or decrease it if appropriate. Its purposes is to detect corrupt
* object sizes, so it probably doesn't matter if it is a bit big.
*
* JRM - 5/17/04
*/
#define H5C1_MAX_ENTRY_SIZE ((size_t)(10 * 1024 * 1024))
/* H5C1_COLLECT_CACHE_STATS controls overall collection of statistics
* on cache activity. In general, this #define should be set to 0.
*/
#define H5C1_COLLECT_CACHE_STATS 0
/* H5C1_COLLECT_CACHE_ENTRY_STATS controls collection of statistics
* in individual cache entries.
*
* H5C1_COLLECT_CACHE_ENTRY_STATS should only be defined to true if
* H5C1_COLLECT_CACHE_STATS is also defined to true.
*/
#if H5C1_COLLECT_CACHE_STATS
#define H5C1_COLLECT_CACHE_ENTRY_STATS 1
#else
#define H5C1_COLLECT_CACHE_ENTRY_STATS 0
#endif /* H5C1_COLLECT_CACHE_STATS */
#ifdef H5_HAVE_PARALLEL
/* we must maintain the clean and dirty LRU lists when we are compiled
* with parallel support.
*/
#define H5C1_MAINTAIN_CLEAN_AND_DIRTY_LRU_LISTS 1
#else /* H5_HAVE_PARALLEL */
/* The clean and dirty LRU lists don't buy us anything here -- we may
* want them on for testing on occasion, but in general they should be
* off.
*/
#define H5C1_MAINTAIN_CLEAN_AND_DIRTY_LRU_LISTS 0
#endif /* H5_HAVE_PARALLEL */
/* Typedef for the main structure for the cache (defined in H5Cpkg.h) */
typedef struct H5C1_t H5C1_t;
/*
* Class methods pertaining to caching. Each type of cached object will
* have a constant variable with permanent life-span that describes how
* to cache the object. That variable will be of type H5C1_class_t and
* have the following required fields...
*
* LOAD: Loads an object from disk to memory. The function
* should allocate some data structure and return it.
*
* FLUSH: Writes some data structure back to disk. It would be
* wise for the data structure to include dirty flags to
* indicate whether it really needs to be written. This
* function is also responsible for freeing memory allocated
* by the LOAD method if the DEST argument is non-zero (by
* calling the DEST method).
*
* DEST: Just frees memory allocated by the LOAD method.
*
* CLEAR: Just marks object as non-dirty.
*
* SIZE: Report the size (on disk) of the specified cache object.
* Note that the space allocated on disk may not be contiguous.
*/
#define H5C1_CALLBACK__NO_FLAGS_SET 0x0
#define H5C1_CALLBACK__SIZE_CHANGED_FLAG 0x1
#define H5C1_CALLBACK__RENAMED_FLAG 0x2
typedef void *(*H5C1_load_func_t)(H5F_t *f,
hid_t dxpl_id,
haddr_t addr,
const void *udata1,
void *udata2);
typedef herr_t (*H5C1_flush_func_t)(H5F_t *f,
hid_t dxpl_id,
hbool_t dest,
haddr_t addr,
void *thing,
unsigned * flags_ptr);
typedef herr_t (*H5C1_dest_func_t)(H5F_t *f,
void *thing);
typedef herr_t (*H5C1_clear_func_t)(H5F_t *f,
void *thing,
hbool_t dest);
typedef herr_t (*H5C1_size_func_t)(const H5F_t *f,
const void *thing,
size_t *size_ptr);
typedef struct H5C1_class_t {
int id;
H5C1_load_func_t load;
H5C1_flush_func_t flush;
H5C1_dest_func_t dest;
H5C1_clear_func_t clear;
H5C1_size_func_t size;
} H5C1_class_t;
/* Type defintions of call back functions used by the cache as a whole */
typedef herr_t (*H5C1_write_permitted_func_t)(const H5F_t *f,
hid_t dxpl_id,
hbool_t * write_permitted_ptr);
typedef herr_t (*H5C1_log_flush_func_t)(H5C1_t * cache_ptr,
haddr_t addr,
hbool_t was_dirty,
unsigned flags,
int type_id);
/* Upper and lower limits on cache size. These limits are picked
* out of a hat -- you should be able to change them as necessary.
*
* However, if you need a very big cache, you should also increase the
* size of the hash table (H5C1__HASH_TABLE_LEN in H5Cpkg.h). The current
* upper bound on cache size is rather large for the current hash table
* size.
*/
#define H5C1__MAX_MAX_CACHE_SIZE ((size_t)(128 * 1024 * 1024))
#define H5C1__MIN_MAX_CACHE_SIZE ((size_t)(1024))
/* Default max cache size and min clean size are give here to make
* them generally accessable.
*/
#define H5C1__DEFAULT_MAX_CACHE_SIZE ((size_t)(4 * 1024 * 1024))
#define H5C1__DEFAULT_MIN_CLEAN_SIZE ((size_t)(2 * 1024 * 1024))
/****************************************************************************
*
* structure H5C1_cache_entry_t
*
* Instances of the H5C1_cache_entry_t structure are used to store cache
* entries in a hash table and sometimes in a skip list.
* See H5SL.c for the particulars of the skip list.
*
* In typical application, this structure is the first field in a
* structure to be cached. For historical reasons, the external module
* is responsible for managing the is_dirty field (this is no longer
* completely true. See the comment on the is_dirty field for details).
* All other fields are managed by the cache.
*
* The fields of this structure are discussed individually below:
*
* JRM - 4/26/04
*
* magic: Unsigned 32 bit integer that must always be set to
* H5C1__H5C1_CACHE_ENTRY_T_MAGIC when the entry is valid.
* The field must be set to H5C1__H5C1_CACHE_ENTRY_T_BAD_MAGIC
* just before the entry is freed.
*
* This is necessary, as the LRU list can be changed out
* from under H5C1_make_space_in_cache() by the flush
* callback which may change the size of an existing entry,
* and/or load a new entry while serializing the target entry.
*
* This in turn can cause a recursive call to
* H5C1_make_space_in_cache() which may either flush or evict
* the next entry that the first invocation of that function
* was about to examine.
*
* The magic field allows H5C1_make_space_in_cache() to
* detect this case, and re-start its scan from the bottom
* of the LRU when this situation occurs.
*
* This field is only compiled in debug mode.
*
* addr: Base address of the cache entry on disk.
*
* size: Length of the cache entry on disk. Note that unlike normal
* caches, the entries in this cache are of variable length.
* The entries should never overlap, and when we do writebacks,
* we will want to writeback adjacent entries where possible.
*
* NB: At present, entries need not be contiguous on disk. Until
* we fix this, we can't do much with writing back adjacent
* entries.
*
* type: Pointer to the instance of H5C1_class_t containing pointers
* to the methods for cache entries of the current type. This
* field should be NULL when the instance of H5C1_cache_entry_t
* is not in use.
*
* The name is not particularly descriptive, but is retained
* to avoid changes in existing code.
*
* is_dirty: Boolean flag indicating whether the contents of the cache
* entry has been modified since the last time it was written
* to disk.
*
* NOTE: For historical reasons, this field is not maintained
* by the cache. Instead, the module using the cache
* sets this flag when it modifies the entry, and the
* flush and clear functions supplied by that module
* reset the dirty when appropriate.
*
* This is a bit quirky, so we may want to change this
* someday. However it will require a change in the
* cache interface.
*
* Update: Management of the is_dirty field has been largely
* moved into the cache. The only remaining exceptions
* are the flush and clear functions supplied by the
* modules using the cache. These still clear the
* is_dirty field as before. -- JRM 7/5/05
*
* dirtied: Boolean flag used to indicate that the entry has been
* dirtied while protected.
*
* This field is set to FALSE in the protect call, and may
* be set to TRUE by the
* H5C1_mark_pinned_or_protected_entry_dirty()
* call at an time prior to the unprotect call.
*
* The H5C1_mark_pinned_or_protected_entry_dirty() call exists
* as a convenience function for the fractal heap code which
* may not know if an entry is protected or pinned, but knows
* that is either protected or pinned. The dirtied field was
* added as in the parallel case, it is necessary to know
* whether a protected entry was dirty prior to the protect call.
*
* is_protected: Boolean flag indicating whether this entry is protected
* (or locked, to use more conventional terms). When it is
* protected, the entry cannot be flushed or accessed until
* it is unprotected (or unlocked -- again to use more
* conventional terms).
*
* Note that protected entries are removed from the LRU lists
* and inserted on the protected list.
*
* is_read_only: Boolean flag that is only meaningful if is_protected is
* TRUE. In this circumstance, it indicates whether the
* entry has been protected read only, or read/write.
*
* If the entry has been protected read only (i.e. is_protected
* and is_read_only are both TRUE), we allow the entry to be
* protected more than once.
*
* In this case, the number of readers is maintained in the
* ro_ref_count field (see below), and unprotect calls simply
* decrement that field until it drops to zero, at which point
* the entry is actually unprotected.
*
* ro_ref_count: Integer field used to maintain a count of the number of
* outstanding read only protects on this entry. This field
* must be zero whenever either is_protected or is_read_only
* are TRUE.
*
* is_pinned: Boolean flag indicating whether the entry has been pinned
* in the cache.
*
* For very hot entries, the protect / unprotect overhead
* can become excessive. Thus the cache has been extended
* to allow an entry to be "pinned" in the cache.
*
* Pinning an entry in the cache has several implications:
*
* 1) A pinned entry cannot be evicted. Thus unprotected
* pinned entries must be stored in the pinned entry
* list, instead of being managed by the replacement
* policy code (LRU at present).
*
* 2) A pinned entry can be accessed or modified at any time.
* Therefore, the cache must check with the entry owner
* before flushing it. If permission is denied, the
* cache does not flush the entry.
*
* 3) A pinned entry can be marked as dirty (and possibly
* change size) while it is unprotected.
*
* 4) The flush-destroy code must allow pinned entries to
* be unpinned (and possibly unprotected) during the
* flush.
*
* JRM -- 3/16/06
*
* in_slist: Boolean flag indicating whether the entry is in the skip list
* As a general rule, entries are placed in the list when they
* are marked dirty. However they may remain in the list after
* being flushed.
*
* Update: Dirty entries are now removed from the skip list
* when they are flushed.
*
* flush_marker: Boolean flag indicating that the entry is to be flushed
* the next time H5C1_flush_cache() is called with the
* H5C1__FLUSH_MARKED_ENTRIES_FLAG. The flag is reset when
* the entry is flushed for whatever reason.
*
* clear_on_unprotect: Boolean flag used only in PHDF5. When H5C is used
* to implement the metadata cache In the parallel case, only
* the cache with mpi rank 0 is allowed to actually write to
* file -- all other caches must retain dirty entries until they
* are advised that the entry is clean.
*
* This flag is used in the case that such an advisory is
* received when the entry is protected. If it is set when an
* entry is unprotected, and the dirtied flag is not set in
* the unprotect, the entry's is_dirty flag is reset by flushing
* it with the H5C1__FLUSH_CLEAR_ONLY_FLAG.
*
* flush_in_progress: Boolean flag that is set to true iff the entry
* is in the process of being flushed. This allows the cache
* to detect when a call is the result of a flush callback.
*
* destroy_in_progress: Boolean flag that is set to true iff the entry
* is in the process of being flushed and destroyed.
*
*
* Fields supporting the hash table:
*
* Fields in the cache are indexed by a more or less conventional hash table.
* If there are multiple entries in any hash bin, they are stored in a doubly
* linked list.
*
* ht_next: Next pointer used by the hash table to store multiple
* entries in a single hash bin. This field points to the
* next entry in the doubly linked list of entries in the
* hash bin, or NULL if there is no next entry.
*
* ht_prev: Prev pointer used by the hash table to store multiple
* entries in a single hash bin. This field points to the
* previous entry in the doubly linked list of entries in
* the hash bin, or NULL if there is no previuos entry.
*
*
* Fields supporting replacement policies:
*
* The cache must have a replacement policy, and it will usually be
* necessary for this structure to contain fields supporting that policy.
*
* While there has been interest in several replacement policies for
* this cache, the initial development schedule is tight. Thus I have
* elected to support only a modified LRU policy for the first cut.
*
* When additional replacement policies are added, the fields in this
* section will be used in different ways or not at all. Thus the
* documentation of these fields is repeated for each replacement policy.
*
* Modified LRU:
*
* When operating in parallel mode, we must ensure that a read does not
* cause a write. If it does, the process will hang, as the write will
* be collective and the other processes will not know to participate.
*
* To deal with this issue, I have modified the usual LRU policy by adding
* clean and dirty LRU lists to the usual LRU list. When reading in
* parallel mode, we evict from the clean LRU list only. This implies
* that we must try to ensure that the clean LRU list is reasonably well
* stocked. See the comments on H5C1_t in H5Cpkg.h for more details.
*
* Note that even if we start with a completely clean cache, a sequence
* of protects without unprotects can empty the clean LRU list. In this
* case, the cache must grow temporarily. At the next write, we will
* attempt to evict enough entries to get the cache down to its nominal
* maximum size.
*
* The use of the replacement policy fields under the Modified LRU policy
* is discussed below:
*
* next: Next pointer in either the LRU or the protected list,
* depending on the current value of protected. If there
* is no next entry on the list, this field should be set
* to NULL.
*
* prev: Prev pointer in either the LRU or the protected list,
* depending on the current value of protected. If there
* is no previous entry on the list, this field should be
* set to NULL.
*
* aux_next: Next pointer on either the clean or dirty LRU lists.
* This entry should be NULL when protected is true. When
* protected is false, and dirty is true, it should point
* to the next item on the dirty LRU list. When protected
* is false, and dirty is false, it should point to the
* next item on the clean LRU list. In either case, when
* there is no next item, it should be NULL.
*
* aux_prev: Previous pointer on either the clean or dirty LRU lists.
* This entry should be NULL when protected is true. When
* protected is false, and dirty is true, it should point
* to the previous item on the dirty LRU list. When protected
* is false, and dirty is false, it should point to the
* previous item on the clean LRU list. In either case, when
* there is no previous item, it should be NULL.
*
*
* Cache entry stats collection fields:
*
* These fields should only be compiled in when both H5C1_COLLECT_CACHE_STATS
* and H5C1_COLLECT_CACHE_ENTRY_STATS are true. When present, they allow
* collection of statistics on individual cache entries.
*
* accesses: int32_t containing the number of times this cache entry has
* been referenced in its lifetime.
*
* clears: int32_t containing the number of times this cache entry has
* been cleared in its life time.
*
* flushes: int32_t containing the number of times this cache entry has
* been flushed to file in its life time.
*
* pins: int32_t containing the number of times this cache entry has
* been pinned in cache in its life time.
*
****************************************************************************/
#ifndef NDEBUG
#define H5C1__H5C1_CACHE_ENTRY_T_MAGIC 0x005CAC0A
#define H5C1__H5C1_CACHE_ENTRY_T_BAD_MAGIC 0xDeadBeef
#endif /* NDEBUG */
typedef struct H5C1_cache_entry_t
{
#ifndef NDEBUG
uint32_t magic;
#endif /* NDEBUG */
haddr_t addr;
size_t size;
const H5C1_class_t * type;
hbool_t is_dirty;
hbool_t dirtied;
hbool_t is_protected;
hbool_t is_read_only;
int ro_ref_count;
hbool_t is_pinned;
hbool_t in_slist;
hbool_t flush_marker;
#ifdef H5_HAVE_PARALLEL
hbool_t clear_on_unprotect;
#endif /* H5_HAVE_PARALLEL */
hbool_t flush_in_progress;
hbool_t destroy_in_progress;
/* fields supporting the hash table: */
struct H5C1_cache_entry_t * ht_next;
struct H5C1_cache_entry_t * ht_prev;
/* fields supporting replacement policies: */
struct H5C1_cache_entry_t * next;
struct H5C1_cache_entry_t * prev;
struct H5C1_cache_entry_t * aux_next;
struct H5C1_cache_entry_t * aux_prev;
#if H5C1_COLLECT_CACHE_ENTRY_STATS
/* cache entry stats fields */
int32_t accesses;
int32_t clears;
int32_t flushes;
int32_t pins;
#endif /* H5C1_COLLECT_CACHE_ENTRY_STATS */
} H5C1_cache_entry_t;
/****************************************************************************
*
* structure H5C1_auto_size_ctl_t
*
* Instances of H5C1_auto_size_ctl_t are used to get and set the control
* fields for automatic cache re-sizing.
*
* The fields of the structure are discussed individually below:
*
* version: Integer field containing the version number of this version
* of the H5C1_auto_size_ctl_t structure. Any instance of
* H5C1_auto_size_ctl_t passed to the cache must have a known
* version number, or an error will be flagged.
*
* report_fcn: Pointer to the function that is to be called to report
* activities each time the auto cache resize code is executed. If the
* field is NULL, no call is made.
*
* If the field is not NULL, it must contain the address of a function
* of type H5C1_auto_resize_report_fcn.
*
* set_initial_size: Boolean flag indicating whether the size of the
* initial size of the cache is to be set to the value given in
* the initial_size field. If set_initial_size is FALSE, the
* initial_size field is ignored.
*
* initial_size: If enabled, this field contain the size the cache is
* to be set to upon receipt of this structure. Needless to say,
* initial_size must lie in the closed interval [min_size, max_size].
*
* min_clean_fraction: double in the range 0 to 1 indicating the fraction
* of the cache that is to be kept clean. This field is only used
* in parallel mode. Typical values are 0.1 to 0.5.
*
* max_size: Maximum size to which the cache can be adjusted. The
* supplied value must fall in the closed interval
* [MIN_MAX_CACHE_SIZE, MAX_MAX_CACHE_SIZE]. Also, max_size must
* be greater than or equal to min_size.
*
* min_size: Minimum size to which the cache can be adjusted. The
* supplied value must fall in the closed interval
* [MIN_MAX_CACHE_SIZE, MAX_MAX_CACHE_SIZE]. Also, min_size must
* be less than or equal to max_size.
*
* epoch_length: Number of accesses on the cache over which to collect
* hit rate stats before running the automatic cache resize code,
* if it is enabled.
*
* At the end of an epoch, we discard prior hit rate data and start
* collecting afresh. The epoch_length must lie in the closed
* interval [H5C1__MIN_AR_EPOCH_LENGTH, H5C1__MAX_AR_EPOCH_LENGTH].
*
*
* Cache size increase control fields:
*
* incr_mode: Instance of the H5C1_cache_incr_mode enumerated type whose
* value indicates how we determine whether the cache size should be
* increased. At present there are two possible values:
*
* H5C1_incr__off: Don't attempt to increase the size of the cache
* automatically.
*
* When this increment mode is selected, the remaining fields
* in the cache size increase section ar ignored.
*
* H5C1_incr__threshold: Attempt to increase the size of the cache
* whenever the average hit rate over the last epoch drops
* below the value supplied in the lower_hr_threshold
* field.
*
* Note that this attempt will fail if the cache is already
* at its maximum size, or if the cache is not already using
* all available space.
*
* lower_hr_threshold: Lower hit rate threshold. If the increment mode
* (incr_mode) is H5C1_incr__threshold and the hit rate drops below the
* value supplied in this field in an epoch, increment the cache size by
* size_increment. Note that cache size may not be incremented above
* max_size, and that the increment may be further restricted by the
* max_increment field if it is enabled.
*
* When enabled, this field must contain a value in the range [0.0, 1.0].
* Depending on the incr_mode selected, it may also have to be less than
* upper_hr_threshold.
*
* increment: Double containing the multiplier used to derive the new
* cache size from the old if a cache size increment is triggered.
* The increment must be greater than 1.0, and should not exceed 2.0.
*
* The new cache size is obtained by multiplying the current max cache
* size by the increment, and then clamping to max_size and to stay
* within the max_increment as necessary.
*
* apply_max_increment: Boolean flag indicating whether the max_increment
* field should be used to limit the maximum cache size increment.
*
* max_increment: If enabled by the apply_max_increment field described
* above, this field contains the maximum number of bytes by which the
* cache size can be increased in a single re-size.
*
* flash_incr_mode: Instance of the H5C1_cache_flash_incr_mode enumerated
* type whose value indicates whether and by what algorithm we should
* make flash increases in the size of the cache to accomodate insertion
* of large entries and large increases in the size of a single entry.
*
* The addition of the flash increment mode was occasioned by performance
* problems that appear when a local heap is increased to a size in excess
* of the current cache size. While the existing re-size code dealt with
* this eventually, performance was very bad for the remainder of the
* epoch.
*
* At present, there are two possible values for the flash_incr_mode:
*
* H5C1_flash_incr__off: Don't perform flash increases in the size of
* the cache.
*
* H5C1_flash_incr__add_space: Let x be either the size of a newly
* newly inserted entry, or the number of bytes by which the
* size of an existing entry has been increased.
*
* If
* x > flash_threshold * current max cache size,
*
* increase the current maximum cache size by x * flash_multiple
* less any free space in the cache, and start a new epoch. For
* now at least, pay no attention to the maximum increment.
*
*
* With a little thought, it should be obvious that the above flash
* cache size increase algorithm is not sufficient for all circumstances --
* for example, suppose the user round robins through
* (1/flash_threshold) +1 groups, adding one data set to each on each
* pass. Then all will increase in size at about the same time, requiring
* the max cache size to at least double to maintain acceptable
* performance, however the above flash increment algorithm will not be
* triggered.
*
* Hopefully, the add space algorithm detailed above will be sufficient
* for the performance problems encountered to date. However, we should
* expect to revisit the issue.
*
* flash_multiple: Double containing the multiple described above in the
* H5C1_flash_incr__add_space section of the discussion of the
* flash_incr_mode section. This field is ignored unless flash_incr_mode
* is H5C1_flash_incr__add_space.
*
* flash_threshold: Double containing the factor by which current max cache size
* is multiplied to obtain the size threshold for the add_space flash
* increment algorithm. The field is ignored unless flash_incr_mode is
* H5C1_flash_incr__add_space.
*
*
* Cache size decrease control fields:
*
* decr_mode: Instance of the H5C1_cache_decr_mode enumerated type whose
* value indicates how we determine whether the cache size should be
* decreased. At present there are four possibilities.
*
* H5C1_decr__off: Don't attempt to decrease the size of the cache
* automatically.
*
* When this increment mode is selected, the remaining fields
* in the cache size decrease section are ignored.
*
* H5C1_decr__threshold: Attempt to decrease the size of the cache
* whenever the average hit rate over the last epoch rises
* above the value supplied in the upper_hr_threshold
* field.
*
* H5C1_decr__age_out: At the end of each epoch, search the cache for
* entries that have not been accessed for at least the number
* of epochs specified in the epochs_before_eviction field, and
* evict these entries. Conceptually, the maximum cache size
* is then decreased to match the new actual cache size. However,
* this reduction may be modified by the min_size, the
* max_decrement, and/or the empty_reserve.
*
* H5C1_decr__age_out_with_threshold: Same as age_out, but we only
* attempt to reduce the cache size when the hit rate observed
* over the last epoch exceeds the value provided in the
* upper_hr_threshold field.
*
* upper_hr_threshold: Upper hit rate threshold. The use of this field
* varies according to the current decr_mode:
*
* H5C1_decr__off or H5C1_decr__age_out: The value of this field is
* ignored.
*
* H5C1_decr__threshold: If the hit rate exceeds this threshold in any
* epoch, attempt to decrement the cache size by size_decrement.
*
* Note that cache size may not be decremented below min_size.
*
* Note also that if the upper_threshold is 1.0, the cache size
* will never be reduced.
*
* H5C1_decr__age_out_with_threshold: If the hit rate exceeds this
* threshold in any epoch, attempt to reduce the cache size
* by evicting entries that have not been accessed for more
* than the specified number of epochs.
*
* decrement: This field is only used when the decr_mode is
* H5C1_decr__threshold.
*
* The field is a double containing the multiplier used to derive the
* new cache size from the old if a cache size decrement is triggered.
* The decrement must be in the range 0.0 (in which case the cache will
* try to contract to its minimum size) to 1.0 (in which case the
* cache will never shrink).
*
* apply_max_decrement: Boolean flag used to determine whether decrements
* in cache size are to be limited by the max_decrement field.
*
* max_decrement: Maximum number of bytes by which the cache size can be
* decreased in a single re-size. Note that decrements may also be
* restricted by the min_size of the cache, and (in age out modes) by
* the empty_reserve field.
*
* epochs_before_eviction: Integer field used in H5C1_decr__age_out and
* H5C1_decr__age_out_with_threshold decrement modes.
*
* This field contains the number of epochs an entry must remain
* unaccessed before it is evicted in an attempt to reduce the
* cache size. If applicable, this field must lie in the range
* [1, H5C1__MAX_EPOCH_MARKERS].
*
* apply_empty_reserve: Boolean field controlling whether the empty_reserve
* field is to be used in computing the new cache size when the
* decr_mode is H5C1_decr__age_out or H5C1_decr__age_out_with_threshold.
*
* empty_reserve: To avoid a constant racheting down of cache size by small
* amounts in the H5C1_decr__age_out and H5C1_decr__age_out_with_threshold
* modes, this field allows one to require that any cache size
* reductions leave the specified fraction of unused space in the cache.
*
* The value of this field must be in the range [0.0, 1.0]. I would
* expect typical values to be in the range of 0.01 to 0.1.
*
****************************************************************************/
#define H5C1_RESIZE_CFG__VALIDATE_GENERAL 0x1
#define H5C1_RESIZE_CFG__VALIDATE_INCREMENT 0x2
#define H5C1_RESIZE_CFG__VALIDATE_DECREMENT 0x4
#define H5C1_RESIZE_CFG__VALIDATE_INTERACTIONS 0x8
#define H5C1_RESIZE_CFG__VALIDATE_ALL \
( \
H5C1_RESIZE_CFG__VALIDATE_GENERAL | \
H5C1_RESIZE_CFG__VALIDATE_INCREMENT | \
H5C1_RESIZE_CFG__VALIDATE_DECREMENT | \
H5C1_RESIZE_CFG__VALIDATE_INTERACTIONS \
)
#define H5C1__CURR_AUTO_SIZE_CTL_VER 1
#define H5C1__CURR_AUTO_RESIZE_RPT_FCN_VER 1
#define H5C1__MAX_EPOCH_MARKERS 10
#define H5C1__DEF_AR_UPPER_THRESHHOLD 0.9999
#define H5C1__DEF_AR_LOWER_THRESHHOLD 0.9
#define H5C1__DEF_AR_MAX_SIZE ((size_t)(16 * 1024 * 1024))
#define H5C1__DEF_AR_INIT_SIZE ((size_t)( 1 * 1024 * 1024))
#define H5C1__DEF_AR_MIN_SIZE ((size_t)( 1 * 1024 * 1024))
#define H5C1__DEF_AR_MIN_CLEAN_FRAC 0.5
#define H5C1__DEF_AR_INCREMENT 2.0
#define H5C1__DEF_AR_MAX_INCREMENT ((size_t)( 2 * 1024 * 1024))
#define H5C1__DEF_AR_FLASH_MULTIPLE 1.0
#define H5C1__DEV_AR_FLASH_THRESHOLD 0.25
#define H5C1__DEF_AR_DECREMENT 0.9
#define H5C1__DEF_AR_MAX_DECREMENT ((size_t)( 1 * 1024 * 1024))
#define H5C1__DEF_AR_EPCHS_B4_EVICT 3
#define H5C1__DEF_AR_EMPTY_RESERVE 0.05
#define H5C1__MIN_AR_EPOCH_LENGTH 100
#define H5C1__DEF_AR_EPOCH_LENGTH 50000
#define H5C1__MAX_AR_EPOCH_LENGTH 1000000
enum H5C1_resize_status
{
in_spec1,
increase1,
flash_increase1,
decrease1,
at_max_size1,
at_min_size1,
increase_disabled1,
decrease_disabled1,
not_full1
}; /* enum H5C1_resize_conditions */
typedef void (*H5C1_auto_resize_rpt_fcn)(H5C1_t * cache_ptr,
int32_t version,
double hit_rate,
enum H5C1_resize_status status,
size_t old_max_cache_size,
size_t new_max_cache_size,
size_t old_min_clean_size,
size_t new_min_clean_size);
typedef struct H5C1_auto_size_ctl_t
{
/* general configuration fields: */
int32_t version;
H5C1_auto_resize_rpt_fcn rpt_fcn;
hbool_t set_initial_size;
size_t initial_size;
double min_clean_fraction;
size_t max_size;
size_t min_size;
int64_t epoch_length;
/* size increase control fields: */
enum H5C1_cache_incr_mode incr_mode;
double lower_hr_threshold;
double increment;
hbool_t apply_max_increment;
size_t max_increment;
enum H5C1_cache_flash_incr_mode flash_incr_mode;
double flash_multiple;
double flash_threshold;
/* size decrease control fields: */
enum H5C1_cache_decr_mode decr_mode;
double upper_hr_threshold;
double decrement;
hbool_t apply_max_decrement;
size_t max_decrement;
int32_t epochs_before_eviction;
hbool_t apply_empty_reserve;
double empty_reserve;
} H5C1_auto_size_ctl_t;
/*
* Library prototypes.
*/
/* #defines of flags used in the flags parameters in some of the
* following function calls. Note that not all flags are applicable
* to all function calls. Flags that don't apply to a particular
* function are ignored in that function.
*
* These flags apply to all function calls:
*
* H5C1__NO_FLAGS_SET (generic "no flags set" for all fcn calls)
*
*
* These flags apply to H5C1_insert_entry():
*
* H5C1__SET_FLUSH_MARKER_FLAG
* H5C1__PIN_ENTRY_FLAG
*
* These flags apply to H5C1_protect()
*
* H5C1__READ_ONLY_FLAG
*
* These flags apply to H5C1_unprotect():
*
* H5C1__SET_FLUSH_MARKER_FLAG
* H5C1__DELETED_FLAG
* H5C1__DIRTIED_FLAG
* H5C1__SIZE_CHANGED_FLAG
* H5C1__PIN_ENTRY_FLAG
* H5C1__UNPIN_ENTRY_FLAG
*
*
* These flags apply to H5C1_flush_cache():
*
* H5C1__FLUSH_INVALIDATE_FLAG
* H5C1__FLUSH_CLEAR_ONLY_FLAG
* H5C1__FLUSH_MARKED_ENTRIES_FLAG
* H5C1__FLUSH_IGNORE_PROTECTED_FLAG (can't use this flag in combination
* with H5C1__FLUSH_INVALIDATE_FLAG)
*
* These flags apply to H5C1_flush_single_entry():
*
* H5C1__FLUSH_INVALIDATE_FLAG
* H5C1__FLUSH_CLEAR_ONLY_FLAG
* H5C1__FLUSH_MARKED_ENTRIES_FLAG
*/
#define H5C1__NO_FLAGS_SET 0x0000
#define H5C1__SET_FLUSH_MARKER_FLAG 0x0001
#define H5C1__DELETED_FLAG 0x0002
#define H5C1__DIRTIED_FLAG 0x0004
#define H5C1__SIZE_CHANGED_FLAG 0x0008
#define H5C1__PIN_ENTRY_FLAG 0x0010
#define H5C1__UNPIN_ENTRY_FLAG 0x0020
#define H5C1__FLUSH_INVALIDATE_FLAG 0x0040
#define H5C1__FLUSH_CLEAR_ONLY_FLAG 0x0080
#define H5C1__FLUSH_MARKED_ENTRIES_FLAG 0x0100
#define H5C1__FLUSH_IGNORE_PROTECTED_FLAG 0x0200
#define H5C1__READ_ONLY_FLAG 0x0400
H5_DLL H5C1_t * H5C1_create(size_t max_cache_size,
size_t min_clean_size,
int max_type_id,
const char * (* type_name_table_ptr),
H5C1_write_permitted_func_t check_write_permitted,
hbool_t write_permitted,
H5C1_log_flush_func_t log_flush,
void * aux_ptr);
H5_DLL void H5C1_def_auto_resize_rpt_fcn(H5C1_t * cache_ptr,
int32_t version,
double hit_rate,
enum H5C1_resize_status status,
size_t old_max_cache_size,
size_t new_max_cache_size,
size_t old_min_clean_size,
size_t new_min_clean_size);
H5_DLL herr_t H5C1_dest(H5F_t * f,
hid_t primary_dxpl_id,
hid_t secondary_dxpl_id,
H5C1_t * cache_ptr);
H5_DLL herr_t H5C1_dest_empty(H5C1_t * cache_ptr);
H5_DLL herr_t H5C1_expunge_entry(H5F_t * f,
hid_t primary_dxpl_id,
hid_t secondary_dxpl_id,
H5C1_t * cache_ptr,
const H5C1_class_t * type,
haddr_t addr);
H5_DLL herr_t H5C1_flush_cache(H5F_t * f,
hid_t primary_dxpl_id,
hid_t secondary_dxpl_id,
H5C1_t * cache_ptr,
unsigned flags);
H5_DLL herr_t H5C1_flush_to_min_clean(H5F_t * f,
hid_t primary_dxpl_id,
hid_t secondary_dxpl_id,
H5C1_t * cache_ptr);
H5_DLL herr_t H5C1_get_cache_auto_resize_config(H5C1_t * cache_ptr,
H5C1_auto_size_ctl_t *config_ptr);
H5_DLL herr_t H5C1_get_cache_size(H5C1_t * cache_ptr,
size_t * max_size_ptr,
size_t * min_clean_size_ptr,
size_t * cur_size_ptr,
int32_t * cur_num_entries_ptr);
H5_DLL herr_t H5C1_get_cache_hit_rate(H5C1_t * cache_ptr,
double * hit_rate_ptr);
H5_DLL herr_t H5C1_get_entry_status(H5C1_t * cache_ptr,
haddr_t addr,
size_t * size_ptr,
hbool_t * in_cache_ptr,
hbool_t * is_dirty_ptr,
hbool_t * is_protected_ptr,
hbool_t * is_pinned_ptr);
H5_DLL herr_t H5C1_get_evictions_enabled(H5C1_t * cache_ptr,
hbool_t * evictions_enabled_ptr);
H5_DLL herr_t H5C1_get_trace_file_ptr(H5C1_t * cache_ptr,
FILE ** trace_file_ptr_ptr);
H5_DLL herr_t H5C1_insert_entry(H5F_t * f,
hid_t primary_dxpl_id,
hid_t secondary_dxpl_id,
H5C1_t * cache_ptr,
const H5C1_class_t * type,
haddr_t addr,
void * thing,
unsigned int flags);
H5_DLL herr_t H5C1_mark_entries_as_clean(H5F_t * f,
hid_t primary_dxpl_id,
hid_t secondary_dxpl_id,
H5C1_t * cache_ptr,
int32_t ce_array_len,
haddr_t * ce_array_ptr);
H5_DLL herr_t H5C1_mark_pinned_entry_dirty(H5C1_t * cache_ptr,
void * thing,
hbool_t size_changed,
size_t new_size);
H5_DLL herr_t H5C1_mark_pinned_or_protected_entry_dirty(H5C1_t * cache_ptr,
void * thing);
H5_DLL herr_t H5C1_rename_entry(H5C1_t * cache_ptr,
const H5C1_class_t * type,
haddr_t old_addr,
haddr_t new_addr);
H5_DLL herr_t H5C1_pin_protected_entry(H5C1_t * cache_ptr,
void * thing);
H5_DLL void * H5C1_protect(H5F_t * f,
hid_t primary_dxpl_id,
hid_t secondary_dxpl_id,
H5C1_t * cache_ptr,
const H5C1_class_t * type,
haddr_t addr,
const void * udata1,
void * udata2,
unsigned flags);
H5_DLL herr_t H5C1_reset_cache_hit_rate_stats(H5C1_t * cache_ptr);
H5_DLL herr_t H5C1_resize_pinned_entry(H5C1_t * cache_ptr,
void * thing,
size_t new_size);
H5_DLL herr_t H5C1_set_cache_auto_resize_config(H5C1_t * cache_ptr,
H5C1_auto_size_ctl_t *config_ptr);
H5_DLL herr_t H5C1_set_evictions_enabled(H5C1_t * cache_ptr,
hbool_t evictions_enabled);
H5_DLL herr_t H5C1_set_prefix(H5C1_t * cache_ptr, char * prefix);
H5_DLL herr_t H5C1_set_skip_flags(H5C1_t * cache_ptr,
hbool_t skip_file_checks,
hbool_t skip_dxpl_id_checks);
H5_DLL herr_t H5C1_set_trace_file_ptr(H5C1_t * cache_ptr,
FILE * trace_file_ptr);
H5_DLL herr_t H5C1_stats(H5C1_t * cache_ptr,
const char * cache_name,
hbool_t display_detailed_stats);
H5_DLL void H5C1_stats__reset(H5C1_t * cache_ptr);
H5_DLL herr_t H5C1_unpin_entry(H5C1_t * cache_ptr, void * thing);
H5_DLL herr_t H5C1_unprotect(H5F_t * f,
hid_t primary_dxpl_id,
hid_t secondary_dxpl_id,
H5C1_t * cache_ptr,
const H5C1_class_t * type,
haddr_t addr,
void * thing,
unsigned int flags,
size_t new_size);
H5_DLL herr_t H5C1_validate_resize_config(H5C1_auto_size_ctl_t * config_ptr,
unsigned int tests);
#endif /* !_H5C1private_H */
|