/
kmermatcher.cpp
1334 lines (1227 loc) · 61.7 KB
/
kmermatcher.cpp
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
// include xxhash early to avoid incompatibilites with SIMDe
#define XXH_INLINE_ALL
#include "xxhash.h"
#include "kmermatcher.h"
#include "Debug.h"
#include "Indexer.h"
#include "SubstitutionMatrix.h"
#include "ReducedMatrix.h"
#include "ExtendedSubstitutionMatrix.h"
#include "NucleotideMatrix.h"
#include "tantan.h"
#include "QueryMatcher.h"
#include "KmerGenerator.h"
#include "MarkovKmerScore.h"
#include "FileUtil.h"
#include "FastSort.h"
#include "SequenceWeights.h"
#include <sys/stat.h>
#include <sys/mman.h>
#include <fcntl.h>
#include <limits>
#include <algorithm>
#ifdef OPENMP
#include <omp.h>
#endif
#ifndef SIZE_T_MAX
#define SIZE_T_MAX ((size_t) -1)
#endif
uint64_t hashUInt64(uint64_t in, uint64_t seed) {
#if SIMDE_ENDIAN_ORDER == SIMDE_ENDIAN_BIG
in = __builtin_bswap64(in);
#endif
return XXH64(&in, sizeof(uint64_t), seed);
}
template <typename T>
KmerPosition<T> *initKmerPositionMemory(size_t size) {
KmerPosition<T> * hashSeqPair = new(std::nothrow) KmerPosition<T>[size + 1];
Util::checkAllocation(hashSeqPair, "Can not allocate memory");
size_t pageSize = Util::getPageSize()/sizeof(KmerPosition<T>);
#pragma omp parallel
{
#pragma omp for schedule(static)
for (size_t page = 0; page < size+1; page += pageSize) {
size_t readUntil = std::min(size+1, page + pageSize) - page;
memset(hashSeqPair+page, 0xFF, sizeof(KmerPosition<T>)* readUntil);
}
}
return hashSeqPair;
}
void maskSequence(int maskMode, int maskLowerCase, float maskProb, Sequence &seq, int maskLetter, ProbabilityMatrix * probMatrix){
if (maskMode == 1) {
tantan::maskSequences((char*)seq.numSequence,
(char*)(seq.numSequence + seq.L),
50 /*options.maxCycleLength*/,
probMatrix->probMatrixPointers,
0.005 /*options.repeatProb*/,
0.05 /*options.repeatEndProb*/,
0.5 /*options.repeatOffsetProbDecay*/,
0, 0,
maskProb /*options.minMaskProb*/, probMatrix->hardMaskTable);
}
if(maskLowerCase == 1 && (Parameters::isEqualDbtype(seq.getSequenceType(), Parameters::DBTYPE_AMINO_ACIDS) ||
Parameters::isEqualDbtype(seq.getSequenceType(), Parameters::DBTYPE_NUCLEOTIDES))) {
const char * charSeq = seq.getSeqData();
for (int i = 0; i < seq.L; i++) {
seq.numSequence[i] = (islower(charSeq[i])) ? maskLetter : seq.numSequence[i];
}
}
}
template <int TYPE, typename T>
std::pair<size_t, size_t> fillKmerPositionArray(KmerPosition<T> * kmerArray, size_t kmerArraySize, DBReader<unsigned int> &seqDbr,
Parameters & par, BaseMatrix * subMat, bool hashWholeSequence,
size_t hashStartRange, size_t hashEndRange, size_t * hashDistribution){
size_t offset = 0;
int querySeqType = seqDbr.getDbtype();
size_t longestKmer = par.kmerSize;
ProbabilityMatrix *probMatrix = NULL;
if (par.maskMode == 1) {
probMatrix = new ProbabilityMatrix(*subMat);
}
ScoreMatrix two;
ScoreMatrix three;
if (TYPE == Parameters::DBTYPE_HMM_PROFILE) {
two = ExtendedSubstitutionMatrix::calcScoreMatrix(*subMat, 2);
three = ExtendedSubstitutionMatrix::calcScoreMatrix(*subMat, 3);
}
Debug::Progress progress(seqDbr.getSize());
#pragma omp parallel
{
unsigned int thread_idx = 0;
#ifdef OPENMP
thread_idx = static_cast<unsigned int>(omp_get_thread_num());
#endif
unsigned short * scoreDist= new unsigned short[65536];
unsigned int * hierarchicalScoreDist= new unsigned int[128];
const int adjustedKmerSize = (par.adjustKmerLength) ? std::min( par.kmerSize+5, 23) : par.kmerSize;
Sequence seq(par.maxSeqLen, querySeqType, subMat, adjustedKmerSize, par.spacedKmer, false, true, par.spacedKmerPattern);
KmerGenerator* generator;
if (TYPE == Parameters::DBTYPE_HMM_PROFILE) {
generator = new KmerGenerator( par.kmerSize, subMat->alphabetSize, 150);
generator->setDivideStrategy(&three, &two);
}
Indexer idxer(subMat->alphabetSize - 1, par.kmerSize);
const unsigned int BUFFER_SIZE = 1048576;
size_t bufferPos = 0;
KmerPosition<T> * threadKmerBuffer = new KmerPosition<T>[BUFFER_SIZE];
SequencePosition * kmers = (SequencePosition *) malloc((par.pickNbest * (par.maxSeqLen + 1) + 1) * sizeof(SequencePosition));
size_t kmersArraySize = par.maxSeqLen;
const size_t flushSize = 100000000;
size_t iterations = static_cast<size_t>(ceil(static_cast<double>(seqDbr.getSize()) / static_cast<double>(flushSize)));
for (size_t i = 0; i < iterations; i++) {
size_t start = (i * flushSize);
size_t bucketSize = std::min(seqDbr.getSize() - (i * flushSize), flushSize);
#pragma omp for schedule(dynamic, 100)
for (size_t id = start; id < (start + bucketSize); id++) {
progress.updateProgress();
memset(scoreDist, 0, sizeof(unsigned short) * 65536);
memset(hierarchicalScoreDist, 0, sizeof(unsigned int) * 128);
seq.mapSequence(id, seqDbr.getDbKey(id), seqDbr.getData(id, thread_idx), seqDbr.getSeqLen(id));
size_t seqHash = SIZE_T_MAX;
//TODO, how to handle this in reverse?
if(hashWholeSequence){
seqHash = Util::hash(seq.numSequence, seq.L);
seqHash = hashUInt64(seqHash, par.hashShift);
}
maskSequence(par.maskMode, par.maskLowerCaseMode, par.maskProb, seq, subMat->aa2num[static_cast<int>('X')], probMatrix);
size_t seqKmerCount = 0;
unsigned int seqId = seq.getDbKey();
while (seq.hasNextKmer()) {
unsigned char *kmer = (unsigned char*) seq.nextKmer();
if(seq.kmerContainsX()){
continue;
}
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES){
NucleotideMatrix * nuclMatrix = (NucleotideMatrix*)subMat;
size_t kmerLen = par.kmerSize;
size_t kmerIdx = Indexer::computeKmerIdx(kmer, kmerLen);
size_t revkmerIdx = Util::revComplement(kmerIdx, kmerLen);
// skip forward and rev. identical k-mers.
// We can not know how to align these afterwards
if(revkmerIdx == kmerIdx){
continue;
}
bool pickReverseKmer = (revkmerIdx<kmerIdx);
kmerIdx = (pickReverseKmer) ? revkmerIdx : kmerIdx;
const unsigned short hash = hashUInt64(kmerIdx, par.hashShift);
if(par.adjustKmerLength) {
unsigned char revKmer[32];
unsigned char * kmerToHash = kmer;
if(pickReverseKmer){
for(int pos = static_cast<int>(adjustedKmerSize)-1; pos > -1; pos--){
revKmer[(adjustedKmerSize - 1) - pos]=nuclMatrix->reverseResidue(kmer[pos]);
}
kmerToHash = revKmer;
}
kmerLen = MarkovKmerScore::adjustedLength(kmerToHash, adjustedKmerSize,
(par.kmerSize - MarkovScores::MARKOV_ORDER) * MarkovScores::MEDIAN_SCORE);
longestKmer = std::max(kmerLen, longestKmer);
kmerIdx = Indexer::computeKmerIdx(kmerToHash, kmerLen);
}
// set signed bit for normal kmers to make the SIZE_T_MAX logic easier
// reversed kmers do not have a signed bit
size_t kmerRev = (pickReverseKmer) ? BIT_CLEAR(kmerIdx, 63) : BIT_SET(kmerIdx, 63);
(kmers + seqKmerCount)->kmer = kmerRev;
int pos = seq.getCurrentPosition();
(kmers + seqKmerCount)->pos = (pickReverseKmer) ? (seq.L) - pos - kmerLen : pos;
(kmers + seqKmerCount)->score = hash;
scoreDist[hash]++;
hierarchicalScoreDist[hash >> 9]++;
seqKmerCount++;
} else if(TYPE == Parameters::DBTYPE_HMM_PROFILE) {
std::pair<size_t*, size_t> scoreMat = generator->generateKmerList(kmer, true);
// std::cout << scoreMat.elementSize << std::endl;
for(size_t kmerPos = 0; kmerPos < scoreMat.second && kmerPos < static_cast<size_t >(par.pickNbest); kmerPos++){
size_t kmerIdx = scoreMat.first[kmerPos];
(kmers + seqKmerCount)->kmer = kmerIdx;
(kmers + seqKmerCount)->pos = seq.getCurrentPosition();
const unsigned short hash = hashUInt64(kmerIdx, par.hashShift);
(kmers + seqKmerCount)->score = hash;
scoreDist[hash]++;
hierarchicalScoreDist[hash >> 9]++;
seqKmerCount++;
}
} else {
size_t kmerIdx = idxer.int2index(kmer, 0, par.kmerSize);
(kmers + seqKmerCount)->kmer = kmerIdx;
(kmers + seqKmerCount)->pos = seq.getCurrentPosition();
const unsigned short hash = hashUInt64(kmerIdx, par.hashShift);
// (kmers + seqKmerCount)->score = hash;
// const unsigned short hash = circ_hash(kmer, par.kmerSize, 5);
(kmers + seqKmerCount)->score = hash;
scoreDist[hash]++;
hierarchicalScoreDist[hash >> 9]++;
// std::cout << seqId << "\t" << (kmers + seqKmerCount)->score << "\t" << (kmers + seqKmerCount)->pos << std::endl;
seqKmerCount++;
}
if(seqKmerCount >= kmersArraySize){
kmersArraySize = seq.getMaxLen();
kmers = (SequencePosition *) realloc(kmers, (par.pickNbest * (kmersArraySize + 1) + 1) * sizeof(SequencePosition));
}
}
float kmersPerSequenceScale = (TYPE == Parameters::DBTYPE_NUCLEOTIDES) ? par.kmersPerSequenceScale.values.nucleotide()
: par.kmersPerSequenceScale.values.aminoacid();
size_t kmerConsidered = std::min(static_cast<size_t >(par.kmersPerSequence - 1 + (kmersPerSequenceScale * seq.L)), seqKmerCount);
unsigned int threshold = 0;
size_t kmerInBins = 0;
if (seqKmerCount > 0) {
size_t hierarchicaThreshold = 0;
for(hierarchicaThreshold = 0; hierarchicaThreshold < 128 && kmerInBins < kmerConsidered; hierarchicaThreshold++){
kmerInBins += hierarchicalScoreDist[hierarchicaThreshold];
}
hierarchicaThreshold -= (hierarchicaThreshold > 0) ? 1: 0;
kmerInBins -= hierarchicalScoreDist[hierarchicaThreshold];
for(threshold = hierarchicaThreshold*512; threshold <= USHRT_MAX && kmerInBins < kmerConsidered; threshold++){
kmerInBins += scoreDist[threshold];
}
}
int tooMuchElemInLastBin = (kmerInBins - kmerConsidered);
// add k-mer to represent the identity
if (static_cast<unsigned short>(seqHash) >= hashStartRange && static_cast<unsigned short>(seqHash) <= hashEndRange) {
threadKmerBuffer[bufferPos].kmer = seqHash;
threadKmerBuffer[bufferPos].id = seqId;
threadKmerBuffer[bufferPos].pos = 0;
threadKmerBuffer[bufferPos].seqLen = seq.L;
if(hashDistribution != NULL){
__sync_fetch_and_add(&hashDistribution[static_cast<unsigned short>(seqHash)], 1);
}
bufferPos++;
if (bufferPos >= BUFFER_SIZE) {
size_t writeOffset = __sync_fetch_and_add(&offset, bufferPos);
if(writeOffset + bufferPos < kmerArraySize){
if(kmerArray!=NULL){
memcpy(kmerArray + writeOffset, threadKmerBuffer, sizeof(KmerPosition<T>) * bufferPos);
}
} else{
Debug(Debug::ERROR) << "Kmer array overflow. currKmerArrayOffset="<< writeOffset
<< ", kmerBufferPos=" << bufferPos
<< ", kmerArraySize=" << kmerArraySize <<".\n";
EXIT(EXIT_FAILURE);
}
bufferPos = 0;
}
}
if(par.ignoreMultiKmer){
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES) {
SORT_SERIAL(kmers, kmers + seqKmerCount, SequencePosition::compareByScoreReverse);
}else{
SORT_SERIAL(kmers, kmers + seqKmerCount, SequencePosition::compareByScore);
}
}
size_t selectedKmer = 0;
for (size_t kmerIdx = 0; kmerIdx < seqKmerCount && selectedKmer < kmerConsidered; kmerIdx++) {
/* skip repeated kmer */
if (par.ignoreMultiKmer) {
size_t kmer = (kmers + kmerIdx)->kmer;
if (TYPE == Parameters::DBTYPE_NUCLEOTIDES) {
kmer = BIT_SET(kmer, 63);
}
if (kmerIdx + 1 < seqKmerCount) {
size_t nextKmer = (kmers + kmerIdx + 1)->kmer;
if (TYPE == Parameters::DBTYPE_NUCLEOTIDES) {
nextKmer = BIT_SET(nextKmer, 63);
}
if (kmer == nextKmer) {
while (kmer == nextKmer && kmerIdx < seqKmerCount) {
kmerIdx++;
if(kmerIdx >= seqKmerCount)
break;
nextKmer = (kmers + kmerIdx)->kmer;
if (TYPE == Parameters::DBTYPE_NUCLEOTIDES) {
nextKmer = BIT_SET(nextKmer, 63);
}
}
}
}
if(kmerIdx >= seqKmerCount)
break;
}
if ((kmers + kmerIdx)->score < threshold ){
// this if is needed to avoid extracting too much elements in the last bin
if((kmers + kmerIdx)->score == (threshold - 1) && tooMuchElemInLastBin){
tooMuchElemInLastBin--;
threshold -= (tooMuchElemInLastBin == 0) ? 1 : 0;
}
// std::cout << seqId << "\t" << (kmers + kmerIdx)->score << "\t" << (kmers + kmerIdx)->pos << std::endl;
selectedKmer++;
if ((kmers + kmerIdx)->score >= hashStartRange && (kmers + kmerIdx)->score <= hashEndRange)
{
// {
// size_t tmpKmerIdx= (kmers + kmerIdx)->kmer;
// tmpKmerIdx=BIT_CLEAR(tmpKmerIdx, 63);
// std::cout << seqId << "\t" << (kmers + kmerIdx)->score << "\t" << tmpKmerIdx << std::endl;
// }
threadKmerBuffer[bufferPos].kmer = (kmers + kmerIdx)->kmer;
threadKmerBuffer[bufferPos].id = seqId;
threadKmerBuffer[bufferPos].pos = (kmers + kmerIdx)->pos;
threadKmerBuffer[bufferPos].seqLen = seq.L;
bufferPos++;
if(hashDistribution != NULL){
__sync_fetch_and_add(&hashDistribution[(kmers + kmerIdx)->score], 1);
}
if (bufferPos >= BUFFER_SIZE) {
size_t writeOffset = __sync_fetch_and_add(&offset, bufferPos);
if(writeOffset + bufferPos < kmerArraySize){
if(kmerArray!=NULL) {
memcpy(kmerArray + writeOffset, threadKmerBuffer,
sizeof(KmerPosition<T>) * bufferPos);
}
} else{
Debug(Debug::ERROR) << "Kmer array overflow. currKmerArrayOffset="<< writeOffset
<< ", kmerBufferPos=" << bufferPos
<< ", kmerArraySize=" << kmerArraySize <<".\n";
EXIT(EXIT_FAILURE);
}
bufferPos = 0;
}
}
}
}
}
#pragma omp barrier
unsigned int thread_idx = 0;
#ifdef OPENMP
thread_idx = static_cast<unsigned int>(omp_get_thread_num());
#endif
if (thread_idx == 0) {
seqDbr.remapData();
}
#pragma omp barrier
}
if(bufferPos > 0){
size_t writeOffset = __sync_fetch_and_add(&offset, bufferPos);
if(kmerArray != NULL){
memcpy(kmerArray+writeOffset, threadKmerBuffer, sizeof(KmerPosition<T>) * bufferPos);
}
}
free(kmers);
delete[] threadKmerBuffer;
delete[] hierarchicalScoreDist;
delete[] scoreDist;
if (TYPE == Parameters::DBTYPE_HMM_PROFILE) {
delete generator;
}
}
if (TYPE == Parameters::DBTYPE_HMM_PROFILE) {
ExtendedSubstitutionMatrix::freeScoreMatrix(three);
ExtendedSubstitutionMatrix::freeScoreMatrix(two);
}
if (probMatrix != NULL) {
delete probMatrix;
}
return std::make_pair(offset, longestKmer);
}
template <int TYPE, typename T>
void swapCenterSequence(KmerPosition<T> *hashSeqPair, size_t splitKmerCount, SequenceWeights &seqWeights) {
size_t prevHash = hashSeqPair[0].kmer;
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES){
prevHash = BIT_SET(prevHash, 63);
}
size_t repSeqPos = 0;
size_t prevHashStart = 0;
float repSeqWeight = seqWeights.getWeightById(hashSeqPair[repSeqPos].id);
for (size_t elementIdx = 0; elementIdx < splitKmerCount; elementIdx++) {
size_t currKmer = hashSeqPair[elementIdx].kmer;
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES){
currKmer = BIT_SET(currKmer, 63);
}
if (prevHash != currKmer) {
// swap sequence with heighest weigtht to the front of the block
if (repSeqPos != prevHashStart)
std::swap(hashSeqPair[repSeqPos],hashSeqPair[prevHashStart]);
prevHashStart = elementIdx;
prevHash = hashSeqPair[elementIdx].kmer;
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES){
prevHash = BIT_SET(prevHash, 63);
}
repSeqPos = elementIdx;
repSeqWeight = seqWeights.getWeightById(hashSeqPair[repSeqPos].id);
}
else {
float currWeight = seqWeights.getWeightById(hashSeqPair[elementIdx].id);
if (currWeight > repSeqWeight) {
repSeqWeight = currWeight;
repSeqPos = elementIdx;
}
}
if (hashSeqPair[elementIdx].kmer == SIZE_T_MAX) {
break;
}
}
}
template void swapCenterSequence<0, short>(KmerPosition<short> *kmers, size_t splitKmerCount, SequenceWeights &seqWeights);
template void swapCenterSequence<0, int>(KmerPosition<int> *kmers, size_t splitKmerCount, SequenceWeights &seqWeights);
template void swapCenterSequence<1, short>(KmerPosition<short> *kmers, size_t splitKmerCount, SequenceWeights &seqWeights);
template void swapCenterSequence<1, int>(KmerPosition<int> *kmers, size_t splitKmerCount, SequenceWeights &seqWeights);
template <typename T>
KmerPosition<T> * doComputation(size_t totalKmers, size_t hashStartRange, size_t hashEndRange, std::string splitFile,
DBReader<unsigned int> & seqDbr, Parameters & par, BaseMatrix * subMat) {
KmerPosition<T> * hashSeqPair = initKmerPositionMemory<T>(totalKmers);
size_t elementsToSort;
if(Parameters::isEqualDbtype(seqDbr.getDbtype(), Parameters::DBTYPE_NUCLEOTIDES)){
std::pair<size_t, size_t > ret = fillKmerPositionArray<Parameters::DBTYPE_NUCLEOTIDES, T>(hashSeqPair, totalKmers, seqDbr, par, subMat, true, hashStartRange, hashEndRange, NULL);
elementsToSort = ret.first;
par.kmerSize = ret.second;
Debug(Debug::INFO) << "\nAdjusted k-mer length " << par.kmerSize << "\n";
}else{
std::pair<size_t, size_t > ret = fillKmerPositionArray<Parameters::DBTYPE_AMINO_ACIDS, T>(hashSeqPair, totalKmers, seqDbr, par, subMat, true, hashStartRange, hashEndRange, NULL);
elementsToSort = ret.first;
}
if(hashEndRange == SIZE_T_MAX){
seqDbr.unmapData();
}
Debug(Debug::INFO) << "Sort kmer ";
Timer timer;
if(Parameters::isEqualDbtype(seqDbr.getDbtype(), Parameters::DBTYPE_NUCLEOTIDES)) {
SORT_PARALLEL(hashSeqPair, hashSeqPair + elementsToSort, KmerPosition<T>::compareRepSequenceAndIdAndPosReverse);
}else{
SORT_PARALLEL(hashSeqPair, hashSeqPair + elementsToSort, KmerPosition<T>::compareRepSequenceAndIdAndPos);
}
Debug(Debug::INFO) << timer.lap() << "\n";
SequenceWeights *sequenceWeights = NULL;
// use priority information to swap center sequences
if (par.PARAM_WEIGHT_FILE.wasSet) {
sequenceWeights = new SequenceWeights(par.weightFile.c_str());
if (sequenceWeights != NULL) {
if (Parameters::isEqualDbtype(seqDbr.getDbtype(), Parameters::DBTYPE_NUCLEOTIDES)) {
swapCenterSequence<Parameters::DBTYPE_NUCLEOTIDES, T>(hashSeqPair, totalKmers, *sequenceWeights);
} else {
swapCenterSequence<Parameters::DBTYPE_AMINO_ACIDS, T>(hashSeqPair, totalKmers, *sequenceWeights);
}
}
}
// assign rep. sequence to same kmer members
// The longest sequence is the first since we sorted by kmer, seq.Len and id
size_t writePos;
if(Parameters::isEqualDbtype(seqDbr.getDbtype(), Parameters::DBTYPE_NUCLEOTIDES)){
writePos = assignGroup<Parameters::DBTYPE_NUCLEOTIDES, T>(hashSeqPair, totalKmers, par.includeOnlyExtendable, par.covMode, par.covThr, sequenceWeights, par.weightThr);
}else{
writePos = assignGroup<Parameters::DBTYPE_AMINO_ACIDS, T>(hashSeqPair, totalKmers, par.includeOnlyExtendable, par.covMode, par.covThr, sequenceWeights, par.weightThr);
}
delete sequenceWeights;
// sort by rep. sequence (stored in kmer) and sequence id
Debug(Debug::INFO) << "Sort by rep. sequence ";
timer.reset();
if(Parameters::isEqualDbtype(seqDbr.getDbtype(), Parameters::DBTYPE_NUCLEOTIDES)){
SORT_PARALLEL(hashSeqPair, hashSeqPair + writePos, KmerPosition<T>::compareRepSequenceAndIdAndDiagReverse);
}else{
SORT_PARALLEL(hashSeqPair, hashSeqPair + writePos, KmerPosition<T>::compareRepSequenceAndIdAndDiag);
}
//kx::radix_sort(hashSeqPair, hashSeqPair + elementsToSort, SequenceComparision());
// for(size_t i = 0; i < writePos; i++){
// std::cout << BIT_CLEAR(hashSeqPair[i].kmer, 63) << "\t" << hashSeqPair[i].id << "\t" << hashSeqPair[i].pos << std::endl;
// }
Debug(Debug::INFO) << timer.lap() << "\n";
if(hashEndRange != SIZE_T_MAX){
if(Parameters::isEqualDbtype(seqDbr.getDbtype(), Parameters::DBTYPE_NUCLEOTIDES)){
writeKmersToDisk<Parameters::DBTYPE_NUCLEOTIDES, KmerEntryRev, T>(splitFile, hashSeqPair, writePos + 1);
}else{
writeKmersToDisk<Parameters::DBTYPE_AMINO_ACIDS, KmerEntry, T>(splitFile, hashSeqPair, writePos + 1);
}
delete [] hashSeqPair;
hashSeqPair = NULL;
}
return hashSeqPair;
}
template <int TYPE, typename T>
size_t assignGroup(KmerPosition<T> *hashSeqPair, size_t splitKmerCount, bool includeOnlyExtendable, int covMode, float covThr,
SequenceWeights * sequenceWeights, float weightThr) {
size_t writePos=0;
size_t prevHash = hashSeqPair[0].kmer;
size_t repSeqId = hashSeqPair[0].id;
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES){
bool isReverse = (BIT_CHECK(hashSeqPair[0].kmer, 63) == false);
repSeqId = (isReverse) ? BIT_CLEAR(repSeqId, 63) : BIT_SET(repSeqId, 63);
prevHash = BIT_SET(prevHash, 63);
}
size_t prevHashStart = 0;
size_t prevSetSize = 0;
size_t skipByWeightCount = 0;
T queryLen=hashSeqPair[0].seqLen;
bool repIsReverse = false;
T repSeq_i_pos = hashSeqPair[0].pos;
for (size_t elementIdx = 0; elementIdx < splitKmerCount+1; elementIdx++) {
size_t currKmer = hashSeqPair[elementIdx].kmer;
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES){
currKmer = BIT_SET(currKmer, 63);
}
if (prevHash != currKmer) {
for (size_t i = prevHashStart; i < elementIdx; i++) {
// skip target sequences if weight > weightThr
if(i > prevHashStart && sequenceWeights != NULL
&& sequenceWeights->getWeightById(hashSeqPair[i].id) > weightThr)
continue;
size_t kmer = hashSeqPair[i].kmer;
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES) {
kmer = BIT_SET(hashSeqPair[i].kmer, 63);
}
size_t rId = (kmer != SIZE_T_MAX) ? ((prevSetSize-skipByWeightCount == 1) ? SIZE_T_MAX : repSeqId) : SIZE_T_MAX;
// remove singletones from set
if(rId != SIZE_T_MAX){
int diagonal = repSeq_i_pos - hashSeqPair[i].pos;
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES){
// 00 No problem here both are forward
// 01 We can revert the query of target, lets invert the query.
// 10 Same here, we can revert query to match the not inverted target
// 11 Both are reverted so no problem!
// So we need just 1 bit of information to encode all four states
bool targetIsReverse = (BIT_CHECK(hashSeqPair[i].kmer, 63) == false);
bool queryNeedsToBeRev = false;
// we now need 2 byte of information (00),(01),(10),(11)
// we need to flip the coordinates of the query
T queryPos=0;
T targetPos=0;
// revert kmer in query hits normal kmer in target
// we need revert the query
if (repIsReverse == true && targetIsReverse == false){
queryPos = repSeq_i_pos;
targetPos = hashSeqPair[i].pos;
queryNeedsToBeRev = true;
// both k-mers were extracted on the reverse strand
// this is equal to both are extract on the forward strand
// we just need to offset the position to the forward strand
}else if (repIsReverse == true && targetIsReverse == true){
queryPos = (queryLen - 1) - repSeq_i_pos;
targetPos = (hashSeqPair[i].seqLen - 1) - hashSeqPair[i].pos;
queryNeedsToBeRev = false;
// query is not revers but target k-mer is reverse
// instead of reverting the target, we revert the query and offset the the query/target position
}else if (repIsReverse == false && targetIsReverse == true){
queryPos = (queryLen - 1) - repSeq_i_pos;
targetPos = (hashSeqPair[i].seqLen - 1) - hashSeqPair[i].pos;
queryNeedsToBeRev = true;
// both are forward, everything is good here
}else{
queryPos = repSeq_i_pos;
targetPos = hashSeqPair[i].pos;
queryNeedsToBeRev = false;
}
diagonal = queryPos - targetPos;
rId = (queryNeedsToBeRev) ? BIT_CLEAR(rId, 63) : BIT_SET(rId, 63);
}
// std::cout << diagonal << "\t" << repSeq_i_pos << "\t" << hashSeqPair[i].pos << std::endl;
bool canBeExtended = diagonal < 0 || (diagonal > (queryLen - hashSeqPair[i].seqLen));
bool canBecovered = Util::canBeCovered(covThr, covMode,
static_cast<float>(queryLen),
static_cast<float>(hashSeqPair[i].seqLen));
if((includeOnlyExtendable == false && canBecovered) || (canBeExtended && includeOnlyExtendable ==true )){
hashSeqPair[writePos].kmer = rId;
hashSeqPair[writePos].pos = diagonal;
hashSeqPair[writePos].seqLen = hashSeqPair[i].seqLen;
hashSeqPair[writePos].id = hashSeqPair[i].id;
writePos++;
}
}
// hashSeqPair[i].kmer = SIZE_T_MAX;
hashSeqPair[i].kmer = (i != writePos - 1) ? SIZE_T_MAX : hashSeqPair[i].kmer;
}
prevSetSize = 0;
skipByWeightCount = 0;
prevHashStart = elementIdx;
repSeqId = hashSeqPair[elementIdx].id;
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES){
repIsReverse = (BIT_CHECK(hashSeqPair[elementIdx].kmer, 63) == 0);
repSeqId = (repIsReverse) ? repSeqId : BIT_SET(repSeqId, 63);
}
queryLen = hashSeqPair[elementIdx].seqLen;
repSeq_i_pos = hashSeqPair[elementIdx].pos;
}
if (hashSeqPair[elementIdx].kmer == SIZE_T_MAX) {
break;
}
prevSetSize++;
if(prevSetSize > 1 && sequenceWeights != NULL
&& sequenceWeights->getWeightById(hashSeqPair[elementIdx].id) > weightThr)
skipByWeightCount++;
prevHash = hashSeqPair[elementIdx].kmer;
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES){
prevHash = BIT_SET(prevHash, 63);
}
}
return writePos;
}
template size_t assignGroup<0, short>(KmerPosition<short> *kmers, size_t splitKmerCount, bool includeOnlyExtendable, int covMode, float covThr, SequenceWeights *sequenceWeights, float weightThr);
template size_t assignGroup<0, int>(KmerPosition<int> *kmers, size_t splitKmerCount, bool includeOnlyExtendable, int covMode, float covThr, SequenceWeights *sequenceWeights, float weightThr);
template size_t assignGroup<1, short>(KmerPosition<short> *kmers, size_t splitKmerCount, bool includeOnlyExtendable, int covMode, float covThr, SequenceWeights *sequenceWeights, float weightThr);
template size_t assignGroup<1, int>(KmerPosition<int> *kmers, size_t splitKmerCount, bool includeOnlyExtendable, int covMode, float covThr, SequenceWeights *sequenceWeights, float weightThr);
void setLinearFilterDefault(Parameters *p) {
p->covThr = 0.8;
p->maskMode = 0;
p->spacedKmer = 0;
p->kmerSize = Parameters::CLUST_LINEAR_DEFAULT_K;
p->alphabetSize = MultiParam<NuclAA<int>>(NuclAA<int>(Parameters::CLUST_LINEAR_DEFAULT_ALPH_SIZE, 5));
p->kmersPerSequence = Parameters::CLUST_LINEAR_KMER_PER_SEQ;
}
size_t computeKmerCount(DBReader<unsigned int> &reader, size_t KMER_SIZE, size_t chooseTopKmer, float chooseTopKmerScale) {
size_t totalKmers = 0;
for(size_t id = 0; id < reader.getSize(); id++ ){
int seqLen = static_cast<int>(reader.getSeqLen(id));
// we need one for the sequence hash
int kmerAdjustedSeqLen = std::max(1, seqLen - static_cast<int>(KMER_SIZE ) + 2) ;
totalKmers += std::min(kmerAdjustedSeqLen, static_cast<int>( chooseTopKmer + (chooseTopKmerScale * seqLen)));
}
return totalKmers;
}
template <typename T>
size_t computeMemoryNeededLinearfilter(size_t totalKmer) {
return sizeof(KmerPosition<T>) * totalKmer;
}
template <typename T>
int kmermatcherInner(Parameters& par, DBReader<unsigned int>& seqDbr) {
int querySeqType = seqDbr.getDbtype();
BaseMatrix *subMat;
if (Parameters::isEqualDbtype(querySeqType, Parameters::DBTYPE_NUCLEOTIDES)) {
subMat = new NucleotideMatrix(par.scoringMatrixFile.values.nucleotide().c_str(), 1.0, 0.0);
}else {
if (par.alphabetSize.values.aminoacid() == 21) {
subMat = new SubstitutionMatrix(par.scoringMatrixFile.values.aminoacid().c_str(), 2.0, 0.0);
} else {
SubstitutionMatrix sMat(par.scoringMatrixFile.values.aminoacid().c_str(), 8.0, -0.2f);
subMat = new ReducedMatrix(sMat.probMatrix, sMat.subMatrixPseudoCounts, sMat.aa2num, sMat.num2aa, sMat.alphabetSize, par.alphabetSize.values.aminoacid(), 2.0);
}
}
//seqDbr.readMmapedDataInMemory();
// memoryLimit in bytes
size_t memoryLimit=Util::computeMemory(par.splitMemoryLimit);
Debug(Debug::INFO) << "\n";
float kmersPerSequenceScale = (Parameters::isEqualDbtype(querySeqType, Parameters::DBTYPE_NUCLEOTIDES)) ?
par.kmersPerSequenceScale.values.nucleotide() : par.kmersPerSequenceScale.values.aminoacid();
size_t totalKmers = computeKmerCount(seqDbr, par.kmerSize, par.kmersPerSequence, kmersPerSequenceScale);
size_t totalSizeNeeded = computeMemoryNeededLinearfilter<T>(totalKmers);
// compute splits
size_t splits = static_cast<size_t>(std::ceil(static_cast<float>(totalSizeNeeded) / memoryLimit));
size_t totalKmersPerSplit = std::max(static_cast<size_t>(1024+1),
static_cast<size_t>(std::min(totalSizeNeeded, memoryLimit)/sizeof(KmerPosition<T>))+1);
std::vector<std::pair<size_t, size_t>> hashRanges = setupKmerSplits<T>(par, subMat, seqDbr, totalKmersPerSplit, splits);
if(splits > 1){
Debug(Debug::INFO) << "Process file into " << hashRanges.size() << " parts\n";
}
std::vector<std::string> splitFiles;
KmerPosition<T> *hashSeqPair = NULL;
size_t mpiRank = 0;
#ifdef HAVE_MPI
splits = hashRanges.size();
size_t fromSplit = 0;
size_t splitCount = 1;
mpiRank = MMseqsMPI::rank;
// if split size is great than nodes than we have to
// distribute all splits equally over all nodes
unsigned int * splitCntPerProc = new unsigned int[MMseqsMPI::numProc];
memset(splitCntPerProc, 0, sizeof(unsigned int) * MMseqsMPI::numProc);
for(size_t i = 0; i < splits; i++){
splitCntPerProc[i % MMseqsMPI::numProc] += 1;
}
for(int i = 0; i < MMseqsMPI::rank; i++){
fromSplit += splitCntPerProc[i];
}
splitCount = splitCntPerProc[MMseqsMPI::rank];
delete[] splitCntPerProc;
for(size_t split = fromSplit; split < fromSplit+splitCount; split++) {
std::string splitFileName = par.db2 + "_split_" +SSTR(split);
hashSeqPair = doComputation<T>(totalKmers, hashRanges[split].first, hashRanges[split].second, splitFileName, seqDbr, par, subMat);
}
MPI_Barrier(MPI_COMM_WORLD);
if(mpiRank == 0){
for(size_t split = 0; split < splits; split++) {
std::string splitFileName = par.db2 + "_split_" +SSTR(split);
splitFiles.push_back(splitFileName);
}
}
#else
for(size_t split = 0; split < hashRanges.size(); split++) {
std::string splitFileName = par.db2 + "_split_" +SSTR(split);
Debug(Debug::INFO) << "Generate k-mers list for " << (split+1) <<" split\n";
std::string splitFileNameDone = splitFileName + ".done";
if(FileUtil::fileExists(splitFileNameDone.c_str()) == false){
hashSeqPair = doComputation<T>(totalKmersPerSplit, hashRanges[split].first, hashRanges[split].second, splitFileName, seqDbr, par, subMat);
}
splitFiles.push_back(splitFileName);
}
#endif
if(mpiRank == 0){
std::vector<char> repSequence(seqDbr.getLastKey()+1);
std::fill(repSequence.begin(), repSequence.end(), false);
// write result
DBWriter dbw(par.db2.c_str(), par.db2Index.c_str(), 1, par.compressed,
(Parameters::isEqualDbtype(seqDbr.getDbtype(), Parameters::DBTYPE_NUCLEOTIDES)) ? Parameters::DBTYPE_PREFILTER_REV_RES : Parameters::DBTYPE_PREFILTER_RES );
dbw.open();
Timer timer;
if(splits > 1) {
seqDbr.unmapData();
if(Parameters::isEqualDbtype(seqDbr.getDbtype(), Parameters::DBTYPE_NUCLEOTIDES)) {
mergeKmerFilesAndOutput<Parameters::DBTYPE_NUCLEOTIDES, KmerEntryRev>(dbw, splitFiles, repSequence);
}else{
mergeKmerFilesAndOutput<Parameters::DBTYPE_AMINO_ACIDS, KmerEntry>(dbw, splitFiles, repSequence);
}
for(size_t i = 0; i < splitFiles.size(); i++){
FileUtil::remove(splitFiles[i].c_str());
std::string splitFilesDone = splitFiles[i] + ".done";
FileUtil::remove(splitFilesDone.c_str());
}
} else {
if(Parameters::isEqualDbtype(seqDbr.getDbtype(), Parameters::DBTYPE_NUCLEOTIDES)) {
writeKmerMatcherResult<Parameters::DBTYPE_NUCLEOTIDES>(dbw, hashSeqPair, totalKmersPerSplit, repSequence, 1);
}else{
writeKmerMatcherResult<Parameters::DBTYPE_AMINO_ACIDS>(dbw, hashSeqPair, totalKmersPerSplit, repSequence, 1);
}
}
Debug(Debug::INFO) << "Time for fill: " << timer.lap() << "\n";
// add missing entries to the result (needed for clustering)
#pragma omp parallel num_threads(1)
{
unsigned int thread_idx = 0;
#ifdef OPENMP
thread_idx = static_cast<unsigned int>(omp_get_thread_num());
#endif
#pragma omp for
for (size_t id = 0; id < seqDbr.getSize(); id++) {
char buffer[100];
unsigned int dbKey = seqDbr.getDbKey(id);
if (repSequence[dbKey] == false) {
hit_t h;
h.prefScore = 0;
h.diagonal = 0;
h.seqId = dbKey;
int len = QueryMatcher::prefilterHitToBuffer(buffer, h);
dbw.writeData(buffer, len, dbKey, thread_idx);
}
}
}
dbw.close(false, false);
}
// free memory
delete subMat;
if(hashSeqPair){
delete [] hashSeqPair;
}
return EXIT_SUCCESS;
}
template <typename T>
std::vector<std::pair<size_t, size_t>> setupKmerSplits(Parameters &par, BaseMatrix * subMat, DBReader<unsigned int> &seqDbr, size_t totalKmers, size_t splits){
std::vector<std::pair<size_t, size_t>> hashRanges;
if (splits > 1) {
Debug(Debug::INFO) << "Not enough memory to process at once need to split\n";
// compute exact k-mer dist
size_t * hashDist = new size_t[USHRT_MAX+1];
memset(hashDist, 0 , sizeof(size_t) * (USHRT_MAX+1));
if(Parameters::isEqualDbtype(seqDbr.getDbtype(), Parameters::DBTYPE_NUCLEOTIDES)){
fillKmerPositionArray<Parameters::DBTYPE_NUCLEOTIDES, T>(NULL, SIZE_T_MAX, seqDbr, par, subMat, true, 0, SIZE_T_MAX, hashDist);
}else{
fillKmerPositionArray<Parameters::DBTYPE_AMINO_ACIDS, T>(NULL, SIZE_T_MAX, seqDbr, par, subMat, true, 0, SIZE_T_MAX, hashDist);
}
seqDbr.remapData();
// figure out if machine has enough memory to run this job
size_t maxBucketSize = 0;
for(size_t i = 0; i < (USHRT_MAX+1); i++) {
if(maxBucketSize < hashDist[i]){
maxBucketSize = hashDist[i];
}
}
if(maxBucketSize > totalKmers){
Debug(Debug::INFO) << "Not enough memory to run the kmermatcher. Minimum is at least " << maxBucketSize* sizeof(KmerPosition<T>) << " bytes\n";
EXIT(EXIT_FAILURE);
}
// define splits
size_t currBucketSize = 0;
size_t currBucketStart = 0;
for(size_t i = 0; i < (USHRT_MAX+1); i++){
if(currBucketSize+hashDist[i] >= totalKmers){
hashRanges.emplace_back(currBucketStart, i - 1);
currBucketSize = 0;
currBucketStart = i;
}
currBucketSize+=hashDist[i];
}
hashRanges.emplace_back(currBucketStart, (USHRT_MAX+1));
delete [] hashDist;
}else{
hashRanges.emplace_back(0, SIZE_T_MAX);
}
return hashRanges;
}
int kmermatcher(int argc, const char **argv, const Command &command) {
MMseqsMPI::init(argc, argv);
Parameters &par = Parameters::getInstance();
setLinearFilterDefault(&par);
par.parseParameters(argc, argv, command, true, 0, MMseqsParameter::COMMAND_CLUSTLINEAR);
DBReader<unsigned int> seqDbr(par.db1.c_str(), par.db1Index.c_str(), par.threads,
DBReader<unsigned int>::USE_INDEX | DBReader<unsigned int>::USE_DATA);
seqDbr.open(DBReader<unsigned int>::NOSORT);
int querySeqType = seqDbr.getDbtype();
setKmerLengthAndAlphabet(par, seqDbr.getAminoAcidDBSize(), querySeqType);
std::vector<MMseqsParameter *> *params = command.params;
par.printParameters(command.cmd, argc, argv, *params);
Debug(Debug::INFO) << "Database size: " << seqDbr.getSize() << " type: " << seqDbr.getDbTypeName() << "\n";
if (seqDbr.getMaxSeqLen() < SHRT_MAX) {
kmermatcherInner<short>(par, seqDbr);
}
else {
kmermatcherInner<int>(par, seqDbr);
}
seqDbr.close();
return EXIT_SUCCESS;
}
template <int TYPE, typename T>
void writeKmerMatcherResult(DBWriter & dbw,
KmerPosition<T> *hashSeqPair, size_t totalKmers,
std::vector<char> &repSequence, size_t threads) {
std::vector<size_t> threadOffsets;
size_t splitSize = totalKmers/threads;
threadOffsets.push_back(0);
for(size_t thread = 1; thread < threads; thread++){
size_t kmer = hashSeqPair[thread*splitSize].kmer;
size_t repSeqId = static_cast<size_t>(kmer);
repSeqId=BIT_SET(repSeqId, 63);
bool wasSet = false;
for(size_t pos = thread*splitSize; pos < totalKmers; pos++){
size_t currSeqId = hashSeqPair[pos].kmer;
currSeqId=BIT_SET(currSeqId, 63);
if(repSeqId != currSeqId){
wasSet = true;
threadOffsets.push_back(pos);
break;
}
}
if(wasSet == false){
threadOffsets.push_back(totalKmers - 1 );
}
}
threadOffsets.push_back(totalKmers);
#pragma omp parallel for schedule(dynamic, 1) num_threads(threads)
for(size_t thread = 0; thread < threads; thread++){
std::string prefResultsOutString;
prefResultsOutString.reserve(100000000);
char buffer[100];
size_t lastTargetId = SIZE_T_MAX;
unsigned int writeSets = 0;
size_t kmerPos=0;
size_t repSeqId = SIZE_T_MAX;
for(kmerPos = threadOffsets[thread]; kmerPos < threadOffsets[thread+1] && hashSeqPair[kmerPos].kmer != SIZE_T_MAX; kmerPos++){
size_t currKmer = hashSeqPair[kmerPos].kmer;
int reverMask = 0;
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES){
reverMask = BIT_CHECK(currKmer, 63)==false;
currKmer = BIT_CLEAR(currKmer, 63);
}
if(repSeqId != currKmer) {
if (writeSets > 0) {
repSequence[repSeqId] = true;
dbw.writeData(prefResultsOutString.c_str(), prefResultsOutString.length(), repSeqId, thread);
}else{
if(repSeqId != SIZE_T_MAX) {
repSequence[repSeqId] = false;
}
}
lastTargetId = SIZE_T_MAX;
prefResultsOutString.clear();
repSeqId = currKmer;
hit_t h;
h.seqId = repSeqId;
h.prefScore = 0;
h.diagonal = 0;
int len = QueryMatcher::prefilterHitToBuffer(buffer, h);
// TODO: error handling for len
prefResultsOutString.append(buffer, len);
}
unsigned int targetId = hashSeqPair[kmerPos].id;
T diagonal = hashSeqPair[kmerPos].pos;
size_t kmerOffset = 0;
T prevDiagonal = diagonal;
size_t maxDiagonal = 0;
size_t diagonalCnt = 0;
size_t topScore =0;
int bestReverMask = reverMask;
// compute best diagonal and score for every group of target sequences
while(lastTargetId != targetId
&& kmerPos+kmerOffset < threadOffsets[thread+1]
&& hashSeqPair[kmerPos+kmerOffset].id == targetId){
if(prevDiagonal == hashSeqPair[kmerPos+kmerOffset].pos){
diagonalCnt++;
}else{
diagonalCnt = 1;
}
if(diagonalCnt >= maxDiagonal){
diagonal = hashSeqPair[kmerPos+kmerOffset].pos;
maxDiagonal = diagonalCnt;
if(TYPE == Parameters::DBTYPE_NUCLEOTIDES){
bestReverMask = BIT_CHECK(hashSeqPair[kmerPos+kmerOffset].kmer, 63) == false;
}
}
prevDiagonal = hashSeqPair[kmerPos+kmerOffset].pos;
kmerOffset++;
topScore++;
}
// remove similar double sequence hit
if(targetId != repSeqId && lastTargetId != targetId ){
;
}else{
lastTargetId = targetId;
continue;
}
hit_t h;
h.seqId = targetId;
h.prefScore = (bestReverMask) ? -topScore : topScore;
h.diagonal = diagonal;
int len = QueryMatcher::prefilterHitToBuffer(buffer, h);
prefResultsOutString.append(buffer, len);
lastTargetId = targetId;
writeSets++;
}
if (writeSets > 0) {
repSequence[repSeqId] = true;
dbw.writeData(prefResultsOutString.c_str(), prefResultsOutString.length(), repSeqId, thread);
}else{