-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathfqzcomp5.c
2346 lines (2016 loc) · 59.4 KB
/
fqzcomp5.c
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
/* Tests for fqz codec */
/*
* Copyright (c) 2019,2020,2022 Genome Research Ltd.
* Author(s): James Bonfield
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* 1. Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
*
* 2. Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following
* disclaimer in the documentation and/or other materials provided
* with the distribution.
*
* 3. Neither the names Genome Research Ltd and Wellcome Trust Sanger
* Institute nor the names of its contributors may be used to endorse
* or promote products derived from this software without specific
* prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY GENOME RESEARCH LTD AND CONTRIBUTORS "AS
* IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
* PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL GENOME RESEARCH
* LTD OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/*
File format:
[Header] TODO (magic number)
[Block]* Zero or more blocks of records
[Index] TODO: self index + metrics
Where Block is:
4 Num records
4 TODO: Total compressed block size
[Block CRC] TODO
1 Name strategy
4 NU: uncompressed name length
4 NC: compressed name length
NC Name compressed data
1 Read length strategy
? Read length data
1 Sequence strategy (bits 0..2), both_strands (bit 3), level (4..7)
4 SU: Uncompressed sequence size
4 SC: Compressed sequence size
SC Compressed sequence data
1 Quality strategy
4 QU: Uncompressed quality size
4 QC: Compressed quality size
QC Compressed quality data
*/
// TODO
// - Split aux tags into own data series using CRAM's TL + per tag.
// - Seq encoding using STR + copy-number?
// Couldn't get this to work well though, even though it sounds like an
// ideal thing for ONT/PB-CLR. Maybe let it kick in after so many bases
// in a homopolymer?
// - Removal of lengths from fqzqual stream
// - Entropy encoding of read length stream
// - Reuse of memory buffers for speed
// - Also check why memory usage is so high. Is it all in data models?
// - Increase sizes of fqzcomp contexts? Not so useful for CRAM, but maybe
// it's still beneficial to go beyond 16-bit.
// - Improve fqzcomp tables to permit any mapping rather than monotonic.
// - Distinguish explicit method opts (-s1 -S13B -q1 -Q2 etc) from auto
// picked options (-3, -5) which use auto-selected metrics
// - We don't need large input fastq blocks to do large block compression.
// We could separate model reset from block boundaries, so blocks are small
// but reset boundaries less common.
// - Flag for "+" (3rd record) duplicating first "@".
#include <stdio.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <assert.h>
#include <fcntl.h>
#include <ctype.h>
#include <limits.h>
#include <errno.h>
#include <pthread.h>
#include <sys/time.h>
#include "htscodecs/varint.h"
#include "htscodecs/fqzcomp_qual.h"
#include "htscodecs/tokenise_name3.h"
#include "htscodecs/rANS_static4x16.h"
#include "htscodecs/varint.h"
#include "thread_pool.h"
#include "lzp16e.h"
#define BLK_SIZE 512*1000000
#ifndef MIN
# define MIN(a,b) ((a)<(b)?(a):(b))
# define MAX(a,b) ((a)>(b)?(a):(b))
#endif
// Review metrics stats every X blocks for Y trials
#define METRICS_REVIEW 100
#define METRICS_TRIAL 3
typedef enum {
SEC_NAME,
SEC_LEN,
SEC_SEQ,
SEC_QUAL,
SEC_LAST
} sections;
// TODO: add method costs too
typedef enum {
// general purpose
RANS0=1, RANS1, RANS64, RANS65, RANS128, RANS129, RANS192, RANS193,
RANSXN1,
// LZP; differing min lengths? Make len part of format?
LZP3,
TLZP3, // name lzp; fold into LZP3 with better format structuring.
// TODO LZP2, LZP4, LZP16? Needs storing in byte stream too
// Name specific; may just use arg.slevel and ignore multiplicity here?
// Do we ever want to gather starts on multiple compression levels to
// judge if worth it? Maybe, but only for highest probably.
TOK3_3, TOK3_5, TOK3_7, TOK3_9,
TOK3_3_LZP, TOK3_5_LZP, TOK3_7_LZP, TOK3_9_LZP,
// Seq
SEQ10, SEQ12, SEQ12B, SEQ13B, SEQ14B, SEQ_CUSTOM,
// Qual
FQZ0, FQZ1, FQZ2, FQZ3, FQZ4,
M_LAST,
} methods;
int method_costs[] = {
1.0, 1.1, 1.1, 1.2, 0.5, 0.7, 1.0, 1.2, 1.2, // various RANS
1.3, // LZP3
1.3, // TLZP3
1.3, 1.4, 1.5, 1.6, // TOK3
1.3, 1.4, 1.5, 1.6, // TOK3+LZP
1.7, 1.8, 1.9, 2.0, 2.2, 1.5, // FQZ-SEQ
1.3, 1.3, 1.3, 1.3, 1.3, // FQZ-QUAL
};
typedef struct {
uint64_t usize[M_LAST], csize[M_LAST]; // current accumulated sizes
int review, trial;
int count[M_LAST];
} metrics;
pthread_mutex_t metric_m = PTHREAD_MUTEX_INITIALIZER;
static uint32_t method_avail[SEC_LAST];
static methods method_used[SEC_LAST];
static metrics stats[SEC_LAST];
typedef struct {
int num_records; // number of fastq entries
char *name_buf; // concatenated names, \0 separator
char *seq_buf; // concatenated seq, no separator
char *qual_buf; // concatenated qual, no separator
int *name; // index into name_buf
int *seq; // index into seq_buf
int *qual; // index into qual_buf
unsigned int *len; // sequence length
unsigned int *flag; // READ1/READ2 parsed from name
int name_len, seq_len, qual_len; // used size of _buf above
int name_sz, seq_sz, qual_sz; // alloced size of _buf above
int fixed_len; // length of each seq, 0 if not fixed
} fastq;
fastq *fastq_alloc(int nr) {
fastq *fq = calloc(1, sizeof(*fq));
fq->num_records = nr;
fq->name = calloc(nr, sizeof(char *));
fq->seq = calloc(nr, sizeof(char *));
fq->qual = calloc(nr, sizeof(char *));
fq->len = calloc(nr, sizeof(int));
fq->flag = calloc(nr, sizeof(int));
return fq;
}
void fastq_free(fastq *fq) {
if (!fq)
return;
free(fq->name_buf);
free(fq->seq_buf);
free(fq->qual_buf);
free(fq->name);
free(fq->seq);
free(fq->qual);
free(fq->len);
free(fq->flag);
free(fq);
}
#define goto if (fprintf(stderr, "ERR %s:%d\n", __FILE__, __LINE__)) goto
fastq *load_seqs(char *in, int blk_size, int *last_offset) {
fastq *fq = calloc(1, sizeof(*fq));
if (!fq)
goto err;
size_t name_sz = blk_size/10;
size_t seq_sz = blk_size/2;
size_t qual_sz = blk_size/2;
char *name_buf = fq->name_buf = malloc(name_sz);
char *seq_buf = fq->seq_buf = malloc(seq_sz);
char *qual_buf = fq->qual_buf = malloc(qual_sz);
if (!name_buf || !seq_buf || !qual_buf)
goto err;
int c, i = 0, nr = 0, ar = 0;
int last_name = -1;
fq->fixed_len = -1;
int name_i = 0, seq_i = 0, qual_i = 0;
int last_start = 0;
while (i < blk_size) {
if (nr >= ar) {
ar = ar*1.5 + 10000;
fq->name = realloc(fq->name, ar*sizeof(char *));
fq->seq = realloc(fq->seq , ar*sizeof(char *));
fq->qual = realloc(fq->qual, ar*sizeof(char *));
fq->len = realloc(fq->len, ar*sizeof(int));
fq->flag = realloc(fq->flag, ar*sizeof(int));
}
// @name
fq->name[nr] = name_i;
c = in[i++];
if (c != '@')
goto err;
int name_i_ = name_i; // tmp copy so we can unwind a partial decode
while (i < blk_size && (c = in[i++]) && c != '\n') {
if (name_i_+1 >= name_sz) {
name_sz = name_sz * 1.5 + 1000;
name_buf = fq->name_buf = realloc(fq->name_buf, name_sz);
}
name_buf[name_i_++] = c;
}
if (i == blk_size)
break;
name_buf[name_i_++] = 0;
int flag = 0;
if (name_i_ > 3 &&
name_buf[name_i_-2] == '2' &&
name_buf[name_i_-3] == '/')
flag = FQZ_FREAD2;
if (last_name >= 0 &&
strcmp(fq->name_buf + fq->name[nr], fq->name_buf + last_name) == 0)
flag = FQZ_FREAD2;
fq->flag[nr] = flag;
last_name = fq->name[nr];
// seq
fq->seq[nr] = seq_i;
int len = seq_i;
int seq_i_ = seq_i;
while (i < blk_size && (c = in[i++]) && c != '\n') {
if (seq_i_ >= seq_sz) {
// very unlikely given blk_size/2 starting point,
// but not impossible.
seq_sz = seq_sz * 1.5 + 1000;
seq_buf = fq->seq_buf = realloc(fq->seq_buf, seq_sz);
}
seq_buf[seq_i_++] = c;
}
if (i == blk_size)
break;
fq->len[nr] = seq_i_ - len;
//seq_buf[seq_i_++] = '\n'; // use instead of length terminator?
if (fq->fixed_len == -1)
fq->fixed_len = fq->len[nr];
else if (fq->fixed_len > 0)
if (fq->fixed_len != fq->len[nr])
fq->fixed_len = 0;
// +(name)
if (i < blk_size && (c = in[i++]) != '+')
goto err;
while (i < blk_size && (c = in[i++]) && c != '\n')
;
if (i == blk_size)
break;
// qual
fq->qual[nr] = qual_i;
len = qual_i;
int qual_i_ = qual_i;
while (i < blk_size && (c = in[i++]) && c != '\n') {
if (qual_i_ >= qual_sz) {
// very unlikely given blk_size/2 starting point
qual_sz = qual_sz * 1.5 + 1000;
qual_buf = fq->qual_buf = realloc(fq->qual_buf, qual_sz);
}
qual_buf[qual_i_++] = c-33;
}
if (fq->len[nr] != qual_i_ - len) {
if (i == blk_size)
break;
goto err;
} else if (i == blk_size && c != '\n')
break;
name_i = name_i_;
seq_i = seq_i_;
qual_i = qual_i_;
last_start = i;
nr++;
}
// FIXME: need to deal with case where block size is smaller than
// a single record! For now that puts a limit on smallest block size.
*last_offset = last_start; // so we can continue for next block
fq->name_len = name_i;
fq->seq_len = seq_i;
fq->qual_len = qual_i;
fq->num_records = nr;
// Reduce memory wastage
fq->name_buf = realloc(fq->name_buf, name_i);
fq->seq_buf = realloc(fq->seq_buf, seq_i);
fq->qual_buf = realloc(fq->qual_buf, qual_i);
return fq;
err:
fprintf(stderr, "Failed to load fastq input\n");
fastq_free(fq);
return NULL;
}
#if 0
static uint64_t manual_strats[10] = {0};
static int manual_nstrat = 0;
/*
* Manually specified strategies held in global manual_strats[].
*/
static inline
int fqz_manual_parameters(fqz_gparams *gp,
fqz_slice *s,
unsigned char *in,
size_t in_size) {
int i, p;
int dsqr[] = {
0, 1, 1, 1, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3,
4, 4, 4, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 5,
5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
6, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7
};
gp->vers = FQZ_VERS;
gp->nparam = manual_nstrat;
gp->gflags = GFLAG_MULTI_PARAM | GFLAG_HAVE_STAB;
for (i = 0; i < 256; i++)
gp->stab[i] = 0;
// Fill these out later
gp->max_sel = 0;
gp->max_sym = 0;
gp->p = malloc(gp->nparam * sizeof(*gp->p));
for (p = 0; p < gp->nparam; p++) {
fqz_param *pm = &gp->p[p];
uint64_t st = manual_strats[p];
pm->boff = st & 15; st >>= 4;
pm->bloc = st & 15; st >>= 4;
pm->bbits = st & 15; st >>= 4;
pm->do_qa = st & 15; st >>= 4;
pm->do_r2 = st & 15; st >>= 4;
pm->dloc = st & 15; st >>= 4;
pm->ploc = st & 15; st >>= 4;
pm->sloc = st & 15; st >>= 4;
pm->qloc = st & 15; st >>= 4;
pm->dshift = st & 15; st >>= 4;
pm->dbits = st & 15; st >>= 4;
pm->pshift = st & 15; st >>= 4;
pm->pbits = st & 15; st >>= 4;
pm->qshift = st & 15; st >>= 4;
pm->qbits = st & 15; st >>= 4;
// Gather some stats, as per qual_stats func.
// r in rec count.
// i = index to in[]
// j = index within this rec
uint32_t qhist[256] = {0};
// qual stats for seqs using this parameter only
fqz_qual_stats(s, in, in_size, pm, qhist, p);
int max_sel = pm->max_sel;
// Update max_sel running total. Eg with 4 sub-params:
//
// sel param no. => new
// 0 0 0
// 0/1 1 1,2
// 0/1 2 3,4
// 0 3 5
for (i = gp->max_sel; i < gp->max_sel + max_sel+1; i++)
gp->stab[i] = p;
gp->max_sel += max_sel+1;
pm->fixed_len = pm->fixed_len > 0;
pm->use_qtab = 0; // unused by current encoder
pm->store_qmap = pm->nsym <= 8;
// Adjust parameters based on quality stats.
// FIXME: dup from fqz_pick_parameters.
for (i = 0; i < sizeof(dsqr)/sizeof(*dsqr); i++)
if (dsqr[i] > (1<<pm->dbits)-1)
dsqr[i] = (1<<pm->dbits)-1;
if (pm->store_qmap) {
int j;
for (i = j = 0; i < 256; i++)
if (qhist[i])
pm->qmap[i] = j++;
else
pm->qmap[i] = INT_MAX;
pm->max_sym = pm->nsym;
} else {
pm->nsym = 255;
for (i = 0; i < 256; i++)
pm->qmap[i] = i;
}
if (gp->max_sym < pm->max_sym)
gp->max_sym = pm->max_sym;
// Produce ptab from pshift.
if (pm->qbits) {
for (i = 0; i < 256; i++) {
pm->qtab[i] = i; // 1:1
//pm->qtab[i] = (1<<pm->qshift)-i; // 1:1
//pm->qtab[i] = i/4;
//pm->qtab[i] = ((1<<pm->qshift)-i)/3;
// Alternative mappings:
//qtab[i] = i > 30 ? MIN(max_sym,i)-15 : i/2; // eg for 9827 BAM
}
#if 0
// qtab for PacBio CCS data; saves 3%
for (i='~'-33; i<256; i++) {
pm->qtab[i] = 24+i-('~'-33);
}
int x = 0;
for (i = 0; i < 1; i++)
pm->qtab[i] = x,x++;
for (;i < '~'-33; i++)
//pm->qtab[i] = x,x+=(i%4==0);
pm->qtab[i] = x,x+=(i%4==0);
x++;
for (;i <= '~'-33; i++)
pm->qtab[i] = x,x++;
for (;i < 256; i++)
pm->qtab[i] = x,x+=(i%4==0);
#endif
// for (i = 0; i < 128; i++) {
// for (x = i+1; x < 128; x++) {
// if (pm->qtab[i] != pm->qtab[x])
// break;
// }
// x--;
// if (i==x)
// fprintf(stderr, "%d:%d ", pm->qtab[i], i);
// else {
// fprintf(stderr, "%d:%d-%d ", pm->qtab[i], i, x);
// i=x;
// }
// }
// fprintf(stderr, "\n");
//pm->qtab['~'-33]=32;
// pm->use_qtab = 1;
// for (i = 0; i <= 2 ; i++) pm->qtab[i] = 0;
// for ( ; i <= 12; i++) pm->qtab[i] = 1;
// for ( ; i <= 18; i++) pm->qtab[i] = 2;
// for ( ; i <= 36; i++) pm->qtab[i] = 3;
}
//pm->use_qtab = 0;
pm->qmask = (1<<pm->qbits)-1;
if (pm->pbits) {
for (i = 0; i < 1024; i++)
pm->ptab[i] = MIN((1<<pm->pbits)-1, i>>pm->pshift);
// for (i = 0; i < 1024; i++)
// pm->ptab[i] = MIN((1<<pm->pbits)-1, i < 10 ? i : 10 + i/3);
// Alternatively via analysis of quality distributions we
// may select a bunch of positions that are special and
// have a non-uniform ptab[].
// Manual experimentation on a NovaSeq run saved 2.8% here.
}
if (pm->dbits) {
for (i = 0; i < 256; i++)
pm->dtab[i] = dsqr[MIN(sizeof(dsqr)/sizeof(*dsqr)-1, i>>pm->dshift)];
}
pm->use_ptab = (pm->pbits > 0);
pm->use_dtab = (pm->dbits > 0);
pm->pflags =
(pm->use_qtab ?PFLAG_HAVE_QTAB :0)|
(pm->use_dtab ?PFLAG_HAVE_DTAB :0)|
(pm->use_ptab ?PFLAG_HAVE_PTAB :0)|
(pm->do_sel ?PFLAG_DO_SEL :0)|
(pm->fixed_len ?PFLAG_DO_LEN :0)|
(pm->do_dedup ?PFLAG_DO_DEDUP :0)|
(pm->store_qmap ?PFLAG_HAVE_QMAP :0);
}
for (i = gp->max_sel; i < 256; i++)
gp->stab[i] = gp->stab[gp->max_sel-1];
return 0;
}
#endif
#define NSYM 256
#define STEP 8
#include "htscodecs/c_simple_model.h"
#undef NSYM
#undef STEP
#define NSYM 4
#include "htscodecs/c_small_model.h"
#undef NSYM
#define NSYM 2
#include "htscodecs/c_small_model.h"
// An order-N arithmetic encoder, dedicated to sequence contexts.
char *encode_seq(unsigned char *in, unsigned int in_size,
unsigned int *len, int nrecords, int both_strands,
int ctx_size, unsigned int *out_size) {
char *out = malloc(in_size + 100);
if (!out)
return NULL;
const int msize = 1<<(2*ctx_size);
const int mask = msize-1;
int i;
SMALL_MODEL(4,_) *seq_model = malloc(msize * sizeof(*seq_model));
for (i = 0; i < msize; i++)
SMALL_MODEL(4,_init)(&seq_model[i]);
SMALL_MODEL(2,_) state_model[3];
SIMPLE_MODEL(256,_) run_len[3];
for (i = 0; i < 3; i++) {
SMALL_MODEL(2,_init)(&state_model[i]);
SIMPLE_MODEL(256,_init)(&run_len[i], 256);
}
SIMPLE_MODEL(256,_) literal;
SIMPLE_MODEL(256,_init)(&literal, 256);
RangeCoder rc;
RC_SetOutput(&rc, (char *)out);
RC_StartEncode(&rc);
/* Corresponds to a 12-mer word that doesn't occur in human genome. */
int last = 0x007616c7 & mask;
int last2 = (0x2c6b62ff >> (32 - 2*ctx_size)) & mask; // both strands mode
int L[256];
for (int i = 0; i < 256; i++)
L[i] = 4; // N
L['A'] = 0;
L['C'] = 1;
L['G'] = 2;
L['T'] = 3;
L['a'] = 0x80;
L['c'] = 0x81;
L['g'] = 0x82;
L['t'] = 0x83;
// Transition table to stored code:
// uc lc N
// uc - 0 1
// lc 0 - 1
// N 0 1 -
enum { uc_ACGT = 0, lc_ACGT = 1, other = 2 } state = uc_ACGT;
int nseq = 0;
int seq_len = len[nseq++];
for (int i = 0; i < in_size; ) {//i++) {
// Count size of consecutive symbols matching the same state
int j, run, r2;
switch (state) {
case uc_ACGT: // uppercase ACGT symbols
for (j = i; j < in_size; j++) {
if (L[in[j]] >= 4)
break;
}
break;
case lc_ACGT: // lowercase acgt symbols
for (j = i; j < in_size; j++) {
if (L[in[j]] < 0x80)
break;
}
break;
case other: // ambiguity codes
for (j = i; j < in_size; j++) {
if (L[in[j]] != 4)
break;
}
break;
}
// Encode the run length
r2 = run = j-i;
for (;;) {
//fprintf(stderr, "Encode %d of %d for state %d\n",
// MIN(255, r2), run, state);
SIMPLE_MODEL(256, _encodeSymbol)(&run_len[state], &rc,
MIN(255, r2));
if (r2 >= 255)
r2 -= 255;
else
break;
}
// Encode the symbols
switch (state) {
case uc_ACGT:
case lc_ACGT:
for (j = 0; j < run; j++) {
unsigned char b = L[in[i+j]] & 3;
SMALL_MODEL(4, _encodeSymbol)(&seq_model[last], &rc, b);
last = ((last<<2) + b) & mask;
int pf = ((last<<6)&mask)
+(i+j+3<in_size
?L[in[i+j+1]]*16+L[in[i+j+2]]*4+L[in[i+j+3]]
:0);
_mm_prefetch((const char *)&seq_model[pf], _MM_HINT_T0);
// 0.7% and 3.2% smaller for _.FQ and _.fq respectively
// (at ctx_size 12), but 45% more CPU for seq encoding.
if (both_strands) {
int b2 = last2 & 3;
last2 = last2/4 + ((3-b) << (2*ctx_size-2));
SMALL_MODEL(4, _updateSymbol)(&seq_model[last2], b2);
// ~25% speed gain by prefetching bottom strand too
uint32_t i3 = i+j+3 < in_size
? L[in[i+j+1]] + L[in[i+j+2]]*4 + L[in[i+j+3]]*16
: 0;
i3 = (0x3f - i3) << (2*ctx_size-6);
pf = i+j+3 < in_size ? (last2>>6) +i3 : 0;
_mm_prefetch((const char *)&seq_model[pf], _MM_HINT_T0);
}
// In theory we should reset context for each new sequence
// as there is no obvious correlation between one sequence
// and the next. In practice the difference is 1-2%.
// It only costs 1-2% CPU too, so worth doing.
//
// -s5/100MB -s7 -b / 512MB
// Without: 76655812 60731188
// With: 75789505 -1.1% 59638082 -1.8%
// Slowdown 0.2% 1.9%
if (--seq_len == 0 && i+j+1 < in_size) {
if (nseq >= nrecords) {
free(out);
free(seq_model);
return NULL;
}
seq_len = len[nseq++];
last = 0x007616c7 & mask;
last2 = (0x2c6b62ff >> (32 - 2*ctx_size)) & mask;
}
}
break;
case other:
for (j = 0; j < run; j++) {
SIMPLE_MODEL(256, _encodeSymbol)(&literal, &rc, in[i+j]);
if (--seq_len == 0 && i+j+1 < in_size) {
if (nseq >= nrecords) {
free(out);
free(seq_model);
return NULL;
}
seq_len = len[nseq++];
last = 0x007616c7 & mask;
last2 = (0x2c6b62ff >> (32 - 2*ctx_size)) & mask;
}
}
}
i += run;
if (i >= in_size)
break;
// Encode switch to next state
switch(L[in[i]]) {
case 0: case 1: case 2: case 3:
//fprintf(stderr, "state %d to 0, => 0\n", state);
SMALL_MODEL(2, _encodeSymbol)(&state_model[state], &rc, 0);
state = uc_ACGT;
break;
case 0x80: case 0x81: case 0x82: case 0x83:
//fprintf(stderr, "state %d to 1, => %d\n", state, state==other);
SMALL_MODEL(2, _encodeSymbol)(&state_model[state], &rc,
state == other);
state = lc_ACGT;
break;
default:
//fprintf(stderr, "state %d to 2, => 1\n", state);
SMALL_MODEL(2, _encodeSymbol)(&state_model[state], &rc, 1);
state = other;
break;
}
}
RC_FinishEncode(&rc);
*out_size = RC_OutSize(&rc);
free(seq_model);
return out;
}
char *decode_seq(unsigned char *in, unsigned int in_size,
unsigned int *len, int nrecords, int both_strands,
int ctx_size, unsigned int out_size) {
char *out = malloc(out_size);
if (!out)
return NULL;
const int msize = 1<<(2*ctx_size);
const int mask = msize-1;
int i;
SMALL_MODEL(4,_) *seq_model = malloc(msize * sizeof(*seq_model));
// Do histogram to get observed values.
// Then set m to max number of elements in histogram.
for (i = 0; i < msize; i++)
SMALL_MODEL(4,_init)(&seq_model[i]);
SMALL_MODEL(2,_) state_model[3];
SIMPLE_MODEL(256,_) run_len[3];
for (i = 0; i < 3; i++) {
SMALL_MODEL(2,_init)(&state_model[i]);
SIMPLE_MODEL(256,_init)(&run_len[i], 256);
}
SIMPLE_MODEL(256,_) literal;
SIMPLE_MODEL(256,_init)(&literal, 256);
RangeCoder rc;
RC_SetInput(&rc, (char *)in, (char *)in+in_size);
RC_StartDecode(&rc);
/* Corresponds to a 12-mer word that doesn't occur in human genome. */
int last = 0x007616c7 & mask;
int last2 = (0x2c6b62ff >> (32 - 2*ctx_size)) & mask; // both strands mode
// Transition table to stored code:
// uc lc N
// uc - 0 1
// lc 0 - 1
// N 0 1 -
enum { uc_ACGT = 0, lc_ACGT = 1, other = 2 } state = uc_ACGT;
int nseq = 0;
int seq_len = len[nseq++];
for (int i = 0; i < out_size;) {
int j, run = 0, r2;
// Fetch run length
do {
r2 = SIMPLE_MODEL(256, _decodeSymbol)(&run_len[state], &rc);
run += r2;
//fprintf(stderr, "Decode %d of %d for state %d\n", r2, run, state);
} while (r2 == 255);
if (i + run > out_size)
// or error as it's malformed data
run = out_size - i;
// Decode
switch (state) {
case uc_ACGT:
case lc_ACGT: {
char *bases = state==lc_ACGT ? "acgt" : "ACGT";
for (j = 0; j < run; j++) {
unsigned char b =
SMALL_MODEL(4, _decodeSymbol)(&seq_model[last], &rc);
last = ((last<<2) + b) & mask;
_mm_prefetch((const char *)&seq_model[(last<<4)&mask],
_MM_HINT_T0);
out[i+j] = bases[b];
if (both_strands) {
int b2 = last2 & 3;
last2 = last2/4 + ((3-b) << (2*ctx_size-2));
SMALL_MODEL(4, _updateSymbol)(&seq_model[last2], b2);
}
if (--seq_len == 0 && i+j+1 < out_size) {
if (nseq >= nrecords) {
free(out);
free(seq_model);
return NULL;
}
seq_len = len[nseq++];
last = 0x007616c7 & mask;
last2 = (0x2c6b62ff >> (32 - 2*ctx_size)) & mask;
}
}
break;
}
case other: // ambiguity codes
for (j = 0; j < run; j++) {
out[i+j] = SIMPLE_MODEL(256, _decodeSymbol)(&literal, &rc);
if (--seq_len == 0 && i+j+1 < out_size) {
if (nseq >= nrecords) {
free(out);
free(seq_model);
return NULL;
}
seq_len = len[nseq++];
last = 0x007616c7 & mask;
last2 = (0x2c6b62ff >> (32 - 2*ctx_size)) & mask;
}
}
break;
}
i += run;
if (i >= out_size)
break;
// Next state
int nstate = SMALL_MODEL(2, _decodeSymbol)(&state_model[state], &rc);
switch (state) {
case uc_ACGT:
state = nstate ? other : lc_ACGT;
break;
case lc_ACGT:
state = nstate ? other : uc_ACGT;
break;
case other:
state = nstate ? lc_ACGT : uc_ACGT;
break;
}
}
RC_FinishDecode(&rc);
free(seq_model);
return out;
}
static char *encode_names(unsigned char *name_buf, unsigned int name_len,
int strat, int level, unsigned int *out_size) {
// TODO: work out a better maximum bound
char *nout = malloc(name_len*2+1000), *cp = nout;
*(uint32_t *)cp = name_len; cp += 4;
*cp++ = strat; // name method;
if (strat == 0) {
unsigned char *lzp_out = (unsigned char *)cp;
unsigned int clen = lzp(name_buf, name_len, lzp_out);
unsigned char *out = rans_compress_4x16(lzp_out, clen, &clen, 5);
*(uint32_t *)cp = clen; cp += 4;
memcpy(cp, out, clen); cp += clen;
free(out);
} else if (strat == 1) {
int clen;
unsigned char *out = tok3_encode_names((char *)name_buf, name_len,
level, 0, &clen, NULL);
*(uint32_t *)cp = clen; cp += 4;
memcpy(cp, out, clen); cp += clen;
free(out);
} else {
char *n1 = malloc(name_len);
char *n2 = malloc(name_len);
unsigned char *flag = malloc(name_len/2); //Worst case\n
char *cp1 = n1, *cp2 = n2;
int i = 0, nr = 0;
// Flag bit 0: has "/NUM"
// Flag bit 1: /1 vs /2
// Flag bit 2: has a comment
// Flag bit 3: space vs tab before comment
while (i < name_len) {
int j, f = 0;
int w1end = 0;
int w2start = 0;
int w2end = 0;
for (j = i; j < name_len; j++) {
if (name_buf[j] == '\0') {
w2end = j;
break;
}
if (!w2start && (name_buf[j] == ' ' ||
name_buf[j] == '\t')) {
w1end = j;
w2start = j+1;
f |= 4; // FLAG: has comment
}
}
if (!w1end)
w1end = j;
if (!w2end)
w2end = j;
if (w2start)
// FLAG: space vs tab
f |= name_buf[w2start-1] == ' ' ? 0 : 8;
if (w1end>1 && name_buf[w1end-2] == '/') {
// FLAG /1 or /2
if (name_buf[w1end-1] == '1')
f |= 1, w1end -= 2;
else if (name_buf[w1end-1] == '2')
f |= 3, w1end -= 2;
}
flag[nr++] = f;
memcpy(cp1, &name_buf[i], w1end-i);
cp1[w1end-i]=0;
cp1 += w1end-i+1;
if (w2start) {
memcpy(cp2, &name_buf[w2start], w2end-w2start);
cp2[w2end-w2start] = 0;
cp2 += w2end-w2start+1;
}
i = j+1;
}
int clen1;
unsigned int clenf, clen2 = 0;
unsigned char *out = tok3_encode_names(n1, cp1-n1, level, 0, &clen1,
NULL);
unsigned char *outf = rans_compress_4x16(flag, nr, &clenf, 129);
unsigned char *out2 = NULL;