summaryrefslogtreecommitdiffstats
path: root/media/libsoundtouch/src/TDStretch.cpp
blob: b955bfc961d8a73a95c5fd625926ff28c04452e6 (plain)
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
////////////////////////////////////////////////////////////////////////////////
/// 
/// Sampled sound tempo changer/time stretch algorithm. Changes the sound tempo 
/// while maintaining the original pitch by using a time domain WSOLA-like 
/// method with several performance-increasing tweaks.
///
/// Note : MMX optimized functions reside in a separate, platform-specific 
/// file, e.g. 'mmx_win.cpp' or 'mmx_gcc.cpp'
///
/// Author        : Copyright (c) Olli Parviainen
/// Author e-mail : oparviai 'at' iki.fi
/// SoundTouch WWW: http://www.surina.net/soundtouch
///
////////////////////////////////////////////////////////////////////////////////
//
// Last changed  : $Date: 2015-02-22 15:07:12 +0000 (Sun, 22 Feb 2015) $
// File revision : $Revision: 1.12 $
//
// $Id: TDStretch.cpp 205 2015-02-22 15:07:12Z oparviai $
//
////////////////////////////////////////////////////////////////////////////////
//
// License :
//
//  SoundTouch audio processing library
//  Copyright (c) Olli Parviainen
//
//  This library is free software; you can redistribute it and/or
//  modify it under the terms of the GNU Lesser General Public
//  License as published by the Free Software Foundation; either
//  version 2.1 of the License, or (at your option) any later version.
//
//  This library is distributed in the hope that it will be useful,
//  but WITHOUT ANY WARRANTY; without even the implied warranty of
//  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
//  Lesser General Public License for more details.
//
//  You should have received a copy of the GNU Lesser General Public
//  License along with this library; if not, write to the Free Software
//  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
//
////////////////////////////////////////////////////////////////////////////////

#include <string.h>
#include <limits.h>
#include <assert.h>
#include <math.h>
#include <float.h>

#include "STTypes.h"
#include "cpu_detect.h"
#include "TDStretch.h"

using namespace soundtouch;

#define max(x, y) (((x) > (y)) ? (x) : (y))


/*****************************************************************************
 *
 * Constant definitions
 *
 *****************************************************************************/

// Table for the hierarchical mixing position seeking algorithm
static const short _scanOffsets[5][24]={
    { 124,  186,  248,  310,  372,  434,  496,  558,  620,  682,  744, 806,
      868,  930,  992, 1054, 1116, 1178, 1240, 1302, 1364, 1426, 1488,   0},
    {-100,  -75,  -50,  -25,   25,   50,   75,  100,    0,    0,    0,   0,
        0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,   0},
    { -20,  -15,  -10,   -5,    5,   10,   15,   20,    0,    0,    0,   0,
        0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,   0},
    {  -4,   -3,   -2,   -1,    1,    2,    3,    4,    0,    0,    0,   0,
        0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,   0},
    { 121,  114,   97,  114,   98,  105,  108,   32,  104,   99,  117,  111,
      116,  100,  110,  117,  111,  115,    0,    0,    0,    0,    0,   0}};

/*****************************************************************************
 *
 * Implementation of the class 'TDStretch'
 *
 *****************************************************************************/


TDStretch::TDStretch() : FIFOProcessor(&outputBuffer)
{
    bQuickSeek = false;
    channels = 2;

    pMidBuffer = NULL;
    pMidBufferUnaligned = NULL;
    overlapLength = 0;

    bAutoSeqSetting = true;
    bAutoSeekSetting = true;

//    outDebt = 0;
    skipFract = 0;

    tempo = 1.0f;
    setParameters(44100, DEFAULT_SEQUENCE_MS, DEFAULT_SEEKWINDOW_MS, DEFAULT_OVERLAP_MS);
    setTempo(1.0f);

    clear();
}



TDStretch::~TDStretch()
{
    delete[] pMidBufferUnaligned;
}



// Sets routine control parameters. These control are certain time constants
// defining how the sound is stretched to the desired duration.
//
// 'sampleRate' = sample rate of the sound
// 'sequenceMS' = one processing sequence length in milliseconds (default = 82 ms)
// 'seekwindowMS' = seeking window length for scanning the best overlapping 
//      position (default = 28 ms)
// 'overlapMS' = overlapping length (default = 12 ms)

void TDStretch::setParameters(int aSampleRate, int aSequenceMS, 
                              int aSeekWindowMS, int aOverlapMS)
{
    // accept only positive parameter values - if zero or negative, use old values instead
    if (aSampleRate > 0)
    {
        if (aSampleRate > 192000) ST_THROW_RT_ERROR("Error: Excessive samplerate");
        this->sampleRate = aSampleRate;
    }

    if (aOverlapMS > 0) this->overlapMs = aOverlapMS;

    if (aSequenceMS > 0)
    {
        this->sequenceMs = aSequenceMS;
        bAutoSeqSetting = false;
    } 
    else if (aSequenceMS == 0)
    {
        // if zero, use automatic setting
        bAutoSeqSetting = true;
    }

    if (aSeekWindowMS > 0) 
    {
        this->seekWindowMs = aSeekWindowMS;
        bAutoSeekSetting = false;
    } 
    else if (aSeekWindowMS == 0) 
    {
        // if zero, use automatic setting
        bAutoSeekSetting = true;
    }

    calcSeqParameters();

    calculateOverlapLength(overlapMs);

    // set tempo to recalculate 'sampleReq'
    setTempo(tempo);
}



/// Get routine control parameters, see setParameters() function.
/// Any of the parameters to this function can be NULL, in such case corresponding parameter
/// value isn't returned.
void TDStretch::getParameters(int *pSampleRate, int *pSequenceMs, int *pSeekWindowMs, int *pOverlapMs) const
{
    if (pSampleRate)
    {
        *pSampleRate = sampleRate;
    }

    if (pSequenceMs)
    {
        *pSequenceMs = (bAutoSeqSetting) ? (USE_AUTO_SEQUENCE_LEN) : sequenceMs;
    }

    if (pSeekWindowMs)
    {
        *pSeekWindowMs = (bAutoSeekSetting) ? (USE_AUTO_SEEKWINDOW_LEN) : seekWindowMs;
    }

    if (pOverlapMs)
    {
        *pOverlapMs = overlapMs;
    }
}


// Overlaps samples in 'midBuffer' with the samples in 'pInput'
void TDStretch::overlapMono(SAMPLETYPE *pOutput, const SAMPLETYPE *pInput) const
{
    int i;
    SAMPLETYPE m1, m2;

    m1 = (SAMPLETYPE)0;
    m2 = (SAMPLETYPE)overlapLength;

    for (i = 0; i < overlapLength ; i ++) 
    {
        pOutput[i] = (pInput[i] * m1 + pMidBuffer[i] * m2 ) / overlapLength;
        m1 += 1;
        m2 -= 1;
    }
}



void TDStretch::clearMidBuffer()
{
    memset(pMidBuffer, 0, channels * sizeof(SAMPLETYPE) * overlapLength);
}


void TDStretch::clearInput()
{
    inputBuffer.clear();
    clearMidBuffer();
}


// Clears the sample buffers
void TDStretch::clear()
{
    outputBuffer.clear();
    clearInput();
}



// Enables/disables the quick position seeking algorithm. Zero to disable, nonzero
// to enable
void TDStretch::enableQuickSeek(bool enable)
{
    bQuickSeek = enable;
}


// Returns nonzero if the quick seeking algorithm is enabled.
bool TDStretch::isQuickSeekEnabled() const
{
    return bQuickSeek;
}


// Seeks for the optimal overlap-mixing position.
int TDStretch::seekBestOverlapPosition(const SAMPLETYPE *refPos)
{
    if (bQuickSeek) 
    {
        return seekBestOverlapPositionQuick(refPos);
    } 
    else 
    {
        return seekBestOverlapPositionFull(refPos);
    }
}


// Overlaps samples in 'midBuffer' with the samples in 'pInputBuffer' at position
// of 'ovlPos'.
inline void TDStretch::overlap(SAMPLETYPE *pOutput, const SAMPLETYPE *pInput, uint ovlPos) const
{
#ifndef USE_MULTICH_ALWAYS
    if (channels == 1)
    {
        // mono sound.
        overlapMono(pOutput, pInput + ovlPos);
    }
    else if (channels == 2)
    {
        // stereo sound
        overlapStereo(pOutput, pInput + 2 * ovlPos);
    } 
    else 
#endif // USE_MULTICH_ALWAYS
    {
        assert(channels > 0);
        overlapMulti(pOutput, pInput + channels * ovlPos);
    }
}



// Seeks for the optimal overlap-mixing position. The 'stereo' version of the
// routine
//
// The best position is determined as the position where the two overlapped
// sample sequences are 'most alike', in terms of the highest cross-correlation
// value over the overlapping period
int TDStretch::seekBestOverlapPositionFull(const SAMPLETYPE *refPos) 
{
    int bestOffs;
    double bestCorr;
    int i;
    double norm;

    bestCorr = FLT_MIN;
    bestOffs = 0;

    // Scans for the best correlation value by testing each possible position
    // over the permitted range.
    bestCorr = calcCrossCorr(refPos, pMidBuffer, norm);

    #pragma omp parallel for
    for (i = 1; i < seekLength; i ++) 
    {
        double corr;
        // Calculates correlation value for the mixing position corresponding to 'i'
#ifdef _OPENMP
        // in parallel OpenMP mode, can't use norm accumulator version as parallel executor won't
        // iterate the loop in sequential order
        corr = calcCrossCorr(refPos + channels * i, pMidBuffer, norm);
#else
        // In non-parallel version call "calcCrossCorrAccumulate" that is otherwise same
        // as "calcCrossCorr", but saves time by reusing & updating previously stored 
        // "norm" value
        corr = calcCrossCorrAccumulate(refPos + channels * i, pMidBuffer, norm);
#endif
        // heuristic rule to slightly favour values close to mid of the range
        double tmp = (double)(2 * i - seekLength) / (double)seekLength;
        corr = ((corr + 0.1) * (1.0 - 0.25 * tmp * tmp));

        // Checks for the highest correlation value
        if (corr > bestCorr) 
        {
            // For optimal performance, enter critical section only in case that best value found.
            // in such case repeat 'if' condition as it's possible that parallel execution may have
            // updated the bestCorr value in the mean time
            #pragma omp critical
            if (corr > bestCorr)
            {
                bestCorr = corr;
                bestOffs = i;
            }
        }
    }
    // clear cross correlation routine state if necessary (is so e.g. in MMX routines).
    clearCrossCorrState();

    return bestOffs;
}


// Seeks for the optimal overlap-mixing position. The 'stereo' version of the
// routine
//
// The best position is determined as the position where the two overlapped
// sample sequences are 'most alike', in terms of the highest cross-correlation
// value over the overlapping period
int TDStretch::seekBestOverlapPositionQuick(const SAMPLETYPE *refPos) 
{
    int j;
    int bestOffs;
    double bestCorr, corr;
    int scanCount, corrOffset, tempOffset;

    bestCorr = FLT_MIN;
    bestOffs = _scanOffsets[0][0];
    corrOffset = 0;
    tempOffset = 0;

    // Scans for the best correlation value using four-pass hierarchical search.
    //
    // The look-up table 'scans' has hierarchical position adjusting steps.
    // In first pass the routine searhes for the highest correlation with 
    // relatively coarse steps, then rescans the neighbourhood of the highest
    // correlation with better resolution and so on.
    for (scanCount = 0;scanCount < 4; scanCount ++) 
    {
        j = 0;
        while (_scanOffsets[scanCount][j]) 
        {
            double norm;
            tempOffset = corrOffset + _scanOffsets[scanCount][j];
            if (tempOffset >= seekLength) break;

            // Calculates correlation value for the mixing position corresponding
            // to 'tempOffset'
            corr = (double)calcCrossCorr(refPos + channels * tempOffset, pMidBuffer, norm);
            // heuristic rule to slightly favour values close to mid of the range
            double tmp = (double)(2 * tempOffset - seekLength) / seekLength;
            corr = ((corr + 0.1) * (1.0 - 0.25 * tmp * tmp));

            // Checks for the highest correlation value
            if (corr > bestCorr) 
            {
                bestCorr = corr;
                bestOffs = tempOffset;
            }
            j ++;
        }
        corrOffset = bestOffs;
    }
    // clear cross correlation routine state if necessary (is so e.g. in MMX routines).
    clearCrossCorrState();

    return bestOffs;
}



/// clear cross correlation routine state if necessary 
void TDStretch::clearCrossCorrState()
{
    // default implementation is empty.
}


/// Calculates processing sequence length according to tempo setting
void TDStretch::calcSeqParameters()
{
    // Adjust tempo param according to tempo, so that variating processing sequence length is used
    // at varius tempo settings, between the given low...top limits
    #define AUTOSEQ_TEMPO_LOW   0.5     // auto setting low tempo range (-50%)
    #define AUTOSEQ_TEMPO_TOP   2.0     // auto setting top tempo range (+100%)

    // sequence-ms setting values at above low & top tempo
    #define AUTOSEQ_AT_MIN      125.0
    #define AUTOSEQ_AT_MAX      50.0
    #define AUTOSEQ_K           ((AUTOSEQ_AT_MAX - AUTOSEQ_AT_MIN) / (AUTOSEQ_TEMPO_TOP - AUTOSEQ_TEMPO_LOW))
    #define AUTOSEQ_C           (AUTOSEQ_AT_MIN - (AUTOSEQ_K) * (AUTOSEQ_TEMPO_LOW))

    // seek-window-ms setting values at above low & top tempo
    #define AUTOSEEK_AT_MIN     25.0
    #define AUTOSEEK_AT_MAX     15.0
    #define AUTOSEEK_K          ((AUTOSEEK_AT_MAX - AUTOSEEK_AT_MIN) / (AUTOSEQ_TEMPO_TOP - AUTOSEQ_TEMPO_LOW))
    #define AUTOSEEK_C          (AUTOSEEK_AT_MIN - (AUTOSEEK_K) * (AUTOSEQ_TEMPO_LOW))

    #define CHECK_LIMITS(x, mi, ma) (((x) < (mi)) ? (mi) : (((x) > (ma)) ? (ma) : (x)))

    double seq, seek;
    
    if (bAutoSeqSetting)
    {
        seq = AUTOSEQ_C + AUTOSEQ_K * tempo;
        seq = CHECK_LIMITS(seq, AUTOSEQ_AT_MAX, AUTOSEQ_AT_MIN);
        sequenceMs = (int)(seq + 0.5);
    }

    if (bAutoSeekSetting)
    {
        seek = AUTOSEEK_C + AUTOSEEK_K * tempo;
        seek = CHECK_LIMITS(seek, AUTOSEEK_AT_MAX, AUTOSEEK_AT_MIN);
        seekWindowMs = (int)(seek + 0.5);
    }

    // Update seek window lengths
    seekWindowLength = (sampleRate * sequenceMs) / 1000;
    if (seekWindowLength < 2 * overlapLength) 
    {
        seekWindowLength = 2 * overlapLength;
    }
    seekLength = (sampleRate * seekWindowMs) / 1000;
}



// Sets new target tempo. Normal tempo = 'SCALE', smaller values represent slower 
// tempo, larger faster tempo.
void TDStretch::setTempo(float newTempo)
{
    int intskip;

    tempo = newTempo;

    // Calculate new sequence duration
    calcSeqParameters();

    // Calculate ideal skip length (according to tempo value) 
    nominalSkip = tempo * (seekWindowLength - overlapLength);
    intskip = (int)(nominalSkip + 0.5f);

    // Calculate how many samples are needed in the 'inputBuffer' to 
    // process another batch of samples
    //sampleReq = max(intskip + overlapLength, seekWindowLength) + seekLength / 2;
    sampleReq = max(intskip + overlapLength, seekWindowLength) + seekLength;
}



// Sets the number of channels, 1 = mono, 2 = stereo
void TDStretch::setChannels(int numChannels)
{
    assert(numChannels > 0);
    if (channels == numChannels) return;
//    assert(numChannels == 1 || numChannels == 2);

    channels = numChannels;
    inputBuffer.setChannels(channels);
    outputBuffer.setChannels(channels);

    // re-init overlap/buffer
    overlapLength=0;
    setParameters(sampleRate);
}


// nominal tempo, no need for processing, just pass the samples through
// to outputBuffer
/*
void TDStretch::processNominalTempo()
{
    assert(tempo == 1.0f);

    if (bMidBufferDirty) 
    {
        // If there are samples in pMidBuffer waiting for overlapping,
        // do a single sliding overlapping with them in order to prevent a 
        // clicking distortion in the output sound
        if (inputBuffer.numSamples() < overlapLength) 
        {
            // wait until we've got overlapLength input samples
            return;
        }
        // Mix the samples in the beginning of 'inputBuffer' with the 
        // samples in 'midBuffer' using sliding overlapping 
        overlap(outputBuffer.ptrEnd(overlapLength), inputBuffer.ptrBegin(), 0);
        outputBuffer.putSamples(overlapLength);
        inputBuffer.receiveSamples(overlapLength);
        clearMidBuffer();
        // now we've caught the nominal sample flow and may switch to
        // bypass mode
    }

    // Simply bypass samples from input to output
    outputBuffer.moveSamples(inputBuffer);
}
*/


// Processes as many processing frames of the samples 'inputBuffer', store
// the result into 'outputBuffer'
void TDStretch::processSamples()
{
    int ovlSkip, offset;
    int temp;

    /* Removed this small optimization - can introduce a click to sound when tempo setting
       crosses the nominal value
    if (tempo == 1.0f) 
    {
        // tempo not changed from the original, so bypass the processing
        processNominalTempo();
        return;
    }
    */

    // Process samples as long as there are enough samples in 'inputBuffer'
    // to form a processing frame.
    while ((int)inputBuffer.numSamples() >= sampleReq) 
    {
        // If tempo differs from the normal ('SCALE'), scan for the best overlapping
        // position
        offset = seekBestOverlapPosition(inputBuffer.ptrBegin());

        // Mix the samples in the 'inputBuffer' at position of 'offset' with the 
        // samples in 'midBuffer' using sliding overlapping
        // ... first partially overlap with the end of the previous sequence
        // (that's in 'midBuffer')
        overlap(outputBuffer.ptrEnd((uint)overlapLength), inputBuffer.ptrBegin(), (uint)offset);
        outputBuffer.putSamples((uint)overlapLength);

        // ... then copy sequence samples from 'inputBuffer' to output:

        // length of sequence
        temp = (seekWindowLength - 2 * overlapLength);

        // crosscheck that we don't have buffer overflow...
        if ((int)inputBuffer.numSamples() < (offset + temp + overlapLength * 2))
        {
            continue;    // just in case, shouldn't really happen
        }

        outputBuffer.putSamples(inputBuffer.ptrBegin() + channels * (offset + overlapLength), (uint)temp);

        // Copies the end of the current sequence from 'inputBuffer' to 
        // 'midBuffer' for being mixed with the beginning of the next 
        // processing sequence and so on
        assert((offset + temp + overlapLength * 2) <= (int)inputBuffer.numSamples());
        memcpy(pMidBuffer, inputBuffer.ptrBegin() + channels * (offset + temp + overlapLength), 
            channels * sizeof(SAMPLETYPE) * overlapLength);

        // Remove the processed samples from the input buffer. Update
        // the difference between integer & nominal skip step to 'skipFract'
        // in order to prevent the error from accumulating over time.
        skipFract += nominalSkip;   // real skip size
        ovlSkip = (int)skipFract;   // rounded to integer skip
        skipFract -= ovlSkip;       // maintain the fraction part, i.e. real vs. integer skip
        inputBuffer.receiveSamples((uint)ovlSkip);
    }
}


// Adds 'numsamples' pcs of samples from the 'samples' memory position into
// the input of the object.
void TDStretch::putSamples(const SAMPLETYPE *samples, uint nSamples)
{
    // Add the samples into the input buffer
    inputBuffer.putSamples(samples, nSamples);
    // Process the samples in input buffer
    processSamples();
}



/// Set new overlap length parameter & reallocate RefMidBuffer if necessary.
void TDStretch::acceptNewOverlapLength(int newOverlapLength)
{
    int prevOvl;

    assert(newOverlapLength >= 0);
    prevOvl = overlapLength;
    overlapLength = newOverlapLength;

    if (overlapLength > prevOvl)
    {
        delete[] pMidBufferUnaligned;

        pMidBufferUnaligned = new SAMPLETYPE[overlapLength * channels + 16 / sizeof(SAMPLETYPE)];
        // ensure that 'pMidBuffer' is aligned to 16 byte boundary for efficiency
        pMidBuffer = (SAMPLETYPE *)SOUNDTOUCH_ALIGN_POINTER_16(pMidBufferUnaligned);

        clearMidBuffer();
    }
}


// Operator 'new' is overloaded so that it automatically creates a suitable instance 
// depending on if we've a MMX/SSE/etc-capable CPU available or not.
void * TDStretch::operator new(size_t s)
{
    // Notice! don't use "new TDStretch" directly, use "newInstance" to create a new instance instead!
    ST_THROW_RT_ERROR("Error in TDStretch::new: Don't use 'new TDStretch' directly, use 'newInstance' member instead!");
    return newInstance();
}


TDStretch * TDStretch::newInstance()
{
#if defined(SOUNDTOUCH_ALLOW_MMX) || defined(SOUNDTOUCH_ALLOW_SSE)
    uint uExtensions;

    uExtensions = detectCPUextensions();
#endif

    // Check if MMX/SSE instruction set extensions supported by CPU

#ifdef SOUNDTOUCH_ALLOW_MMX
    // MMX routines available only with integer sample types
    if (uExtensions & SUPPORT_MMX)
    {
        return ::new TDStretchMMX;
    }
    else
#endif // SOUNDTOUCH_ALLOW_MMX


#ifdef SOUNDTOUCH_ALLOW_SSE
    if (uExtensions & SUPPORT_SSE)
    {
        // SSE support
        return ::new TDStretchSSE;
    }
    else
#endif // SOUNDTOUCH_ALLOW_SSE

    {
        // ISA optimizations not supported, use plain C version
        return ::new TDStretch;
    }
}


//////////////////////////////////////////////////////////////////////////////
//
// Integer arithmetics specific algorithm implementations.
//
//////////////////////////////////////////////////////////////////////////////

#ifdef SOUNDTOUCH_INTEGER_SAMPLES

// Overlaps samples in 'midBuffer' with the samples in 'input'. The 'Stereo' 
// version of the routine.
void TDStretch::overlapStereo(short *poutput, const short *input) const
{
    int i;
    short temp;
    int cnt2;

    for (i = 0; i < overlapLength ; i ++) 
    {
        temp = (short)(overlapLength - i);
        cnt2 = 2 * i;
        poutput[cnt2] = (input[cnt2] * i + pMidBuffer[cnt2] * temp )  / overlapLength;
        poutput[cnt2 + 1] = (input[cnt2 + 1] * i + pMidBuffer[cnt2 + 1] * temp ) / overlapLength;
    }
}


// Overlaps samples in 'midBuffer' with the samples in 'input'. The 'Multi'
// version of the routine.
void TDStretch::overlapMulti(SAMPLETYPE *poutput, const SAMPLETYPE *input) const
{
    SAMPLETYPE m1=(SAMPLETYPE)0;
    SAMPLETYPE m2;
    int i=0;

    for (m2 = (SAMPLETYPE)overlapLength; m2; m2 --)
    {
        for (int c = 0; c < channels; c ++)
        {
            poutput[i] = (input[i] * m1 + pMidBuffer[i] * m2)  / overlapLength;
            i++;
        }

        m1++;
    }
}

// Calculates the x having the closest 2^x value for the given value
static int _getClosest2Power(double value)
{
    return (int)(log(value) / log(2.0) + 0.5);
}


/// Calculates overlap period length in samples.
/// Integer version rounds overlap length to closest power of 2
/// for a divide scaling operation.
void TDStretch::calculateOverlapLength(int aoverlapMs)
{
    int newOvl;

    assert(aoverlapMs >= 0);

    // calculate overlap length so that it's power of 2 - thus it's easy to do
    // integer division by right-shifting. Term "-1" at end is to account for 
    // the extra most significatnt bit left unused in result by signed multiplication 
    overlapDividerBits = _getClosest2Power((sampleRate * aoverlapMs) / 1000.0) - 1;
    if (overlapDividerBits > 9) overlapDividerBits = 9;
    if (overlapDividerBits < 3) overlapDividerBits = 3;
    newOvl = (int)pow(2.0, (int)overlapDividerBits + 1);    // +1 => account for -1 above

    acceptNewOverlapLength(newOvl);

    // calculate sloping divider so that crosscorrelation operation won't 
    // overflow 32-bit register. Max. sum of the crosscorrelation sum without 
    // divider would be 2^30*(N^3-N)/3, where N = overlap length
    slopingDivider = (newOvl * newOvl - 1) / 3;
}


double TDStretch::calcCrossCorr(const short *mixingPos, const short *compare, double &norm) const
{
    long corr;
    long lnorm;
    int i;

    corr = lnorm = 0;
    // Same routine for stereo and mono. For stereo, unroll loop for better
    // efficiency and gives slightly better resolution against rounding. 
    // For mono it same routine, just  unrolls loop by factor of 4
    for (i = 0; i < channels * overlapLength; i += 4) 
    {
        corr += (mixingPos[i] * compare[i] + 
                 mixingPos[i + 1] * compare[i + 1]) >> overlapDividerBits;  // notice: do intermediate division here to avoid integer overflow
        corr += (mixingPos[i + 2] * compare[i + 2] + 
                 mixingPos[i + 3] * compare[i + 3]) >> overlapDividerBits;
        lnorm += (mixingPos[i] * mixingPos[i] + 
                  mixingPos[i + 1] * mixingPos[i + 1]) >> overlapDividerBits; // notice: do intermediate division here to avoid integer overflow
        lnorm += (mixingPos[i + 2] * mixingPos[i + 2] + 
                  mixingPos[i + 3] * mixingPos[i + 3]) >> overlapDividerBits;
    }

    // Normalize result by dividing by sqrt(norm) - this step is easiest 
    // done using floating point operation
    norm = (double)lnorm;
    return (double)corr / sqrt((norm < 1e-9) ? 1.0 : norm);
}


/// Update cross-correlation by accumulating "norm" coefficient by previously calculated value
double TDStretch::calcCrossCorrAccumulate(const short *mixingPos, const short *compare, double &norm) const
{
    long corr;
    long lnorm;
    int i;

    // cancel first normalizer tap from previous round
    lnorm = 0;
    for (i = 1; i <= channels; i ++)
    {
        lnorm -= (mixingPos[-i] * mixingPos[-i]) >> overlapDividerBits;
    }

    corr = 0;
    // Same routine for stereo and mono. For stereo, unroll loop for better
    // efficiency and gives slightly better resolution against rounding. 
    // For mono it same routine, just  unrolls loop by factor of 4
    for (i = 0; i < channels * overlapLength; i += 4) 
    {
        corr += (mixingPos[i] * compare[i] + 
                 mixingPos[i + 1] * compare[i + 1]) >> overlapDividerBits;  // notice: do intermediate division here to avoid integer overflow
        corr += (mixingPos[i + 2] * compare[i + 2] + 
                 mixingPos[i + 3] * compare[i + 3]) >> overlapDividerBits;
    }

    // update normalizer with last samples of this round
    for (int j = 0; j < channels; j ++)
    {
        i --;
        lnorm += (mixingPos[i] * mixingPos[i]) >> overlapDividerBits;
    }
    norm += (double)lnorm;

    // Normalize result by dividing by sqrt(norm) - this step is easiest 
    // done using floating point operation
    return (double)corr / sqrt((norm < 1e-9) ? 1.0 : norm);
}

#endif // SOUNDTOUCH_INTEGER_SAMPLES

//////////////////////////////////////////////////////////////////////////////
//
// Floating point arithmetics specific algorithm implementations.
//

#ifdef SOUNDTOUCH_FLOAT_SAMPLES

// Overlaps samples in 'midBuffer' with the samples in 'pInput'
void TDStretch::overlapStereo(float *pOutput, const float *pInput) const
{
    int i;
    float fScale;
    float f1;
    float f2;

    fScale = 1.0f / (float)overlapLength;

    f1 = 0;
    f2 = 1.0f;

    for (i = 0; i < 2 * (int)overlapLength ; i += 2) 
    {
        pOutput[i + 0] = pInput[i + 0] * f1 + pMidBuffer[i + 0] * f2;
        pOutput[i + 1] = pInput[i + 1] * f1 + pMidBuffer[i + 1] * f2;

        f1 += fScale;
        f2 -= fScale;
    }
}


// Overlaps samples in 'midBuffer' with the samples in 'input'. 
void TDStretch::overlapMulti(float *pOutput, const float *pInput) const
{
    int i;
    float fScale;
    float f1;
    float f2;

    fScale = 1.0f / (float)overlapLength;

    f1 = 0;
    f2 = 1.0f;

    i=0;
    for (int i2 = 0; i2 < overlapLength; i2 ++)
    {
        // note: Could optimize this slightly by taking into account that always channels > 2
        for (int c = 0; c < channels; c ++)
        {
            pOutput[i] = pInput[i] * f1 + pMidBuffer[i] * f2;
            i++;
        }
        f1 += fScale;
        f2 -= fScale;
    }
}


/// Calculates overlapInMsec period length in samples.
void TDStretch::calculateOverlapLength(int overlapInMsec)
{
    int newOvl;

    assert(overlapInMsec >= 0);
    newOvl = (sampleRate * overlapInMsec) / 1000;
    if (newOvl < 16) newOvl = 16;

    // must be divisible by 8
    newOvl -= newOvl % 8;

    acceptNewOverlapLength(newOvl);
}


/// Calculate cross-correlation
double TDStretch::calcCrossCorr(const float *mixingPos, const float *compare, double &anorm) const
{
    double corr;
    double norm;
    int i;

    corr = norm = 0;
    // Same routine for stereo and mono. For Stereo, unroll by factor of 2.
    // For mono it's same routine yet unrollsd by factor of 4.
    for (i = 0; i < channels * overlapLength; i += 4) 
    {
        corr += mixingPos[i] * compare[i] +
                mixingPos[i + 1] * compare[i + 1];

        norm += mixingPos[i] * mixingPos[i] + 
                mixingPos[i + 1] * mixingPos[i + 1];

        // unroll the loop for better CPU efficiency:
        corr += mixingPos[i + 2] * compare[i + 2] +
                mixingPos[i + 3] * compare[i + 3];

        norm += mixingPos[i + 2] * mixingPos[i + 2] +
                mixingPos[i + 3] * mixingPos[i + 3];
    }

    anorm = norm;
    return corr / sqrt((norm < 1e-9 ? 1.0 : norm));
}


/// Update cross-correlation by accumulating "norm" coefficient by previously calculated value
double TDStretch::calcCrossCorrAccumulate(const float *mixingPos, const float *compare, double &norm) const
{
    double corr;
    int i;

    corr = 0;

    // cancel first normalizer tap from previous round
    for (i = 1; i <= channels; i ++)
    {
        norm -= mixingPos[-i] * mixingPos[-i];
    }

    // Same routine for stereo and mono. For Stereo, unroll by factor of 2.
    // For mono it's same routine yet unrollsd by factor of 4.
    for (i = 0; i < channels * overlapLength; i += 4) 
    {
        corr += mixingPos[i] * compare[i] +
                mixingPos[i + 1] * compare[i + 1] +
                mixingPos[i + 2] * compare[i + 2] +
                mixingPos[i + 3] * compare[i + 3];
    }

    // update normalizer with last samples of this round
    for (int j = 0; j < channels; j ++)
    {
        i --;
        norm += mixingPos[i] * mixingPos[i];
    }

    return corr / sqrt((norm < 1e-9 ? 1.0 : norm));
}


#endif // SOUNDTOUCH_FLOAT_SAMPLES