aboutsummaryrefslogtreecommitdiffhomepage
path: root/libfaad2/sbr_qmf.c
blob: cf17b64792a9ec181fcdc49a929bf380a02d1db0 (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
/*
** FAAD2 - Freeware Advanced Audio (AAC) Decoder including SBR decoding
** Copyright (C) 2003-2004 M. Bakker, Ahead Software AG, http://www.nero.com
**
** This program is free software; you can redistribute it and/or modify
** it under the terms of the GNU General Public License as published by
** the Free Software Foundation; either version 2 of the License, or
** (at your option) any later version.
**
** This program 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 General Public License for more details.
**
** You should have received a copy of the GNU General Public License
** along with this program; if not, write to the Free Software
** Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
**
** Any non-GPL usage of this software or parts of this software is strictly
** forbidden.
**
** Commercial non-GPL licensing of this software is possible.
** For more info contact Ahead Software through Mpeg4AAClicense@nero.com.
**
** $Id: sbr_qmf.c,v 1.25 2004/06/30 12:45:57 menno Exp $
**/

#include "common.h"
#include "structs.h"

#ifdef SBR_DEC


#include <stdlib.h>
#include <string.h>
#include "sbr_dct.h"
#include "sbr_qmf.h"
#include "sbr_qmf_c.h"
#include "sbr_syntax.h"

qmfa_info *qmfa_init(uint8_t channels)
{
    qmfa_info *qmfa = (qmfa_info*)faad_malloc(sizeof(qmfa_info));

	/* x is implemented as double ringbuffer */
    qmfa->x = (real_t*)faad_malloc(2 * channels * 10 * sizeof(real_t));
    memset(qmfa->x, 0, 2 * channels * 10 * sizeof(real_t));

	/* ringbuffer index */
	qmfa->x_index = 0;

    qmfa->channels = channels;

    return qmfa;
}

void qmfa_end(qmfa_info *qmfa)
{
    if (qmfa)
    {
        if (qmfa->x) faad_free(qmfa->x);
        faad_free(qmfa);
    }
}

void sbr_qmf_analysis_32(sbr_info *sbr, qmfa_info *qmfa, const real_t *input,
                         qmf_t X[MAX_NTSRHFG][64], uint8_t offset, uint8_t kx)
{
    ALIGN real_t u[64];
#ifndef SBR_LOW_POWER
    ALIGN real_t in_real[32], in_imag[32], out_real[32], out_imag[32];
#else
    ALIGN real_t y[32];
#endif
    uint16_t in = 0;
    uint8_t l;

    /* qmf subsample l */
    for (l = 0; l < sbr->numTimeSlotsRate; l++)
    {
        int16_t n;

        /* shift input buffer x */
		/* input buffer is not shifted anymore, x is implemented as double ringbuffer */
        //memmove(qmfa->x + 32, qmfa->x, (320-32)*sizeof(real_t));

        /* add new samples to input buffer x */
        for (n = 32 - 1; n >= 0; n--)
        {
#ifdef FIXED_POINT
            qmfa->x[qmfa->x_index + n] = qmfa->x[qmfa->x_index + n + 320] = (input[in++]) >> 4;
#else
            qmfa->x[qmfa->x_index + n] = qmfa->x[qmfa->x_index + n + 320] = input[in++];
#endif
        }

        /* window and summation to create array u */
        for (n = 0; n < 64; n++)
        {
            u[n] = MUL_F(qmfa->x[qmfa->x_index + n], qmf_c[2*n]) +
                MUL_F(qmfa->x[qmfa->x_index + n + 64], qmf_c[2*(n + 64)]) +
                MUL_F(qmfa->x[qmfa->x_index + n + 128], qmf_c[2*(n + 128)]) +
                MUL_F(qmfa->x[qmfa->x_index + n + 192], qmf_c[2*(n + 192)]) +
                MUL_F(qmfa->x[qmfa->x_index + n + 256], qmf_c[2*(n + 256)]);
        }

		/* update ringbuffer index */
		qmfa->x_index -= 32;
		if (qmfa->x_index < 0)
			qmfa->x_index = (320-32);

        /* calculate 32 subband samples by introducing X */
#ifdef SBR_LOW_POWER
        y[0] = u[48];
        for (n = 1; n < 16; n++)
            y[n] = u[n+48] + u[48-n];
        for (n = 16; n < 32; n++)
            y[n] = -u[n-16] + u[48-n];

        DCT3_32_unscaled(u, y);

        for (n = 0; n < 32; n++)
        {
            if (n < kx)
            {
#ifdef FIXED_POINT
                QMF_RE(X[l + offset][n]) = u[n] /*<< 1*/;
#else
                QMF_RE(X[l + offset][n]) = 2. * u[n];
#endif
            } else {
                QMF_RE(X[l + offset][n]) = 0;
            }
        }
#else

        // Reordering of data moved from DCT_IV to here
        in_imag[31] = u[1];
        in_real[0] = u[0];
        for (n = 1; n < 31; n++)
        {
            in_imag[31 - n] = u[n+1];
            in_real[n] = -u[64-n];
        }
        in_imag[0] = u[32];
        in_real[31] = -u[33];

        // dct4_kernel is DCT_IV without reordering which is done before and after FFT
        dct4_kernel(in_real, in_imag, out_real, out_imag);

        // Reordering of data moved from DCT_IV to here
        for (n = 0; n < 16; n++) {
            if (2*n+1 < kx) {
#ifdef FIXED_POINT
                QMF_RE(X[l + offset][2*n])   = out_real[n];
                QMF_IM(X[l + offset][2*n])   = out_imag[n];
                QMF_RE(X[l + offset][2*n+1]) = -out_imag[31-n];
                QMF_IM(X[l + offset][2*n+1]) = -out_real[31-n];
#else
                QMF_RE(X[l + offset][2*n])   = 2. * out_real[n];
                QMF_IM(X[l + offset][2*n])   = 2. * out_imag[n];
                QMF_RE(X[l + offset][2*n+1]) = -2. * out_imag[31-n];
                QMF_IM(X[l + offset][2*n+1]) = -2. * out_real[31-n];
#endif
            } else {
                if (2*n < kx) {
#ifdef FIXED_POINT
                    QMF_RE(X[l + offset][2*n])   = out_real[n];
                    QMF_IM(X[l + offset][2*n])   = out_imag[n];
#else
                    QMF_RE(X[l + offset][2*n])   = 2. * out_real[n];
                    QMF_IM(X[l + offset][2*n])   = 2. * out_imag[n];
#endif
                }
                else {
                    QMF_RE(X[l + offset][2*n]) = 0;
                    QMF_IM(X[l + offset][2*n]) = 0;
                }
                QMF_RE(X[l + offset][2*n+1]) = 0;
                QMF_IM(X[l + offset][2*n+1]) = 0;
            }
        }
#endif
    }
}

static const complex_t qmf32_pre_twiddle[] =
{
    { FRAC_CONST(0.999924701839145), FRAC_CONST(-0.012271538285720) },
    { FRAC_CONST(0.999322384588350), FRAC_CONST(-0.036807222941359) },
    { FRAC_CONST(0.998118112900149), FRAC_CONST(-0.061320736302209) },
    { FRAC_CONST(0.996312612182778), FRAC_CONST(-0.085797312344440) },
    { FRAC_CONST(0.993906970002356), FRAC_CONST(-0.110222207293883) },
    { FRAC_CONST(0.990902635427780), FRAC_CONST(-0.134580708507126) },
    { FRAC_CONST(0.987301418157858), FRAC_CONST(-0.158858143333861) },
    { FRAC_CONST(0.983105487431216), FRAC_CONST(-0.183039887955141) },
    { FRAC_CONST(0.978317370719628), FRAC_CONST(-0.207111376192219) },
    { FRAC_CONST(0.972939952205560), FRAC_CONST(-0.231058108280671) },
    { FRAC_CONST(0.966976471044852), FRAC_CONST(-0.254865659604515) },
    { FRAC_CONST(0.960430519415566), FRAC_CONST(-0.278519689385053) },
    { FRAC_CONST(0.953306040354194), FRAC_CONST(-0.302005949319228) },
    { FRAC_CONST(0.945607325380521), FRAC_CONST(-0.325310292162263) },
    { FRAC_CONST(0.937339011912575), FRAC_CONST(-0.348418680249435) },
    { FRAC_CONST(0.928506080473216), FRAC_CONST(-0.371317193951838) },
    { FRAC_CONST(0.919113851690058), FRAC_CONST(-0.393992040061048) },
    { FRAC_CONST(0.909167983090522), FRAC_CONST(-0.416429560097637) },
    { FRAC_CONST(0.898674465693954), FRAC_CONST(-0.438616238538528) },
    { FRAC_CONST(0.887639620402854), FRAC_CONST(-0.460538710958240) },
    { FRAC_CONST(0.876070094195407), FRAC_CONST(-0.482183772079123) },
    { FRAC_CONST(0.863972856121587), FRAC_CONST(-0.503538383725718) },
    { FRAC_CONST(0.851355193105265), FRAC_CONST(-0.524589682678469) },
    { FRAC_CONST(0.838224705554838), FRAC_CONST(-0.545324988422046) },
    { FRAC_CONST(0.824589302785025), FRAC_CONST(-0.565731810783613) },
    { FRAC_CONST(0.810457198252595), FRAC_CONST(-0.585797857456439) },
    { FRAC_CONST(0.795836904608884), FRAC_CONST(-0.605511041404326) },
    { FRAC_CONST(0.780737228572094), FRAC_CONST(-0.624859488142386) },
    { FRAC_CONST(0.765167265622459), FRAC_CONST(-0.643831542889791) },
    { FRAC_CONST(0.749136394523459), FRAC_CONST(-0.662415777590172) },
    { FRAC_CONST(0.732654271672413), FRAC_CONST(-0.680600997795453) },
    { FRAC_CONST(0.715730825283819), FRAC_CONST(-0.698376249408973) }
};

qmfs_info *qmfs_init(uint8_t channels)
{
    qmfs_info *qmfs = (qmfs_info*)faad_malloc(sizeof(qmfs_info));

	/* v is a double ringbuffer */
    qmfs->v = (real_t*)faad_malloc(2 * channels * 20 * sizeof(real_t));
    memset(qmfs->v, 0, 2 * channels * 20 * sizeof(real_t));

    qmfs->v_index = 0;

    qmfs->channels = channels;

    return qmfs;
}

void qmfs_end(qmfs_info *qmfs)
{
    if (qmfs)
    {
        if (qmfs->v) faad_free(qmfs->v);
        faad_free(qmfs);
    }
}

#ifdef SBR_LOW_POWER

void sbr_qmf_synthesis_32(sbr_info *sbr, qmfs_info *qmfs, qmf_t X[MAX_NTSRHFG][64],
                          real_t *output)
{
    ALIGN real_t x[16];
    ALIGN real_t y[16];
    int16_t n, k, out = 0;
    uint8_t l;

    /* qmf subsample l */
    for (l = 0; l < sbr->numTimeSlotsRate; l++)
    {
        /* shift buffers */
        /* we are not shifting v, it is a double ringbuffer */
        //memmove(qmfs->v + 64, qmfs->v, (640-64)*sizeof(real_t));

        /* calculate 64 samples */
        for (k = 0; k < 16; k++)
        {
#ifdef FIXED_POINT
            y[k] = (QMF_RE(X[l][k]) - QMF_RE(X[l][31 - k]));
            x[k] = (QMF_RE(X[l][k]) + QMF_RE(X[l][31 - k]));
#else
            y[k] = (QMF_RE(X[l][k]) - QMF_RE(X[l][31 - k])) / 32.0;
            x[k] = (QMF_RE(X[l][k]) + QMF_RE(X[l][31 - k])) / 32.0;
#endif
        }

        /* even n samples */
        DCT2_16_unscaled(x, x);
        /* odd n samples */
        DCT4_16(y, y);

        for (n = 8; n < 24; n++)
        {
            qmfs->v[qmfs->v_index + n*2] = qmfs->v[qmfs->v_index + 640 + n*2] = x[n-8];
            qmfs->v[qmfs->v_index + n*2+1] = qmfs->v[qmfs->v_index + 640 + n*2+1] = y[n-8];
        }
        for (n = 0; n < 16; n++)
        {
            qmfs->v[qmfs->v_index + n] = qmfs->v[qmfs->v_index + 640 + n] = qmfs->v[qmfs->v_index + 32-n];
        }
        qmfs->v[qmfs->v_index + 48] = qmfs->v[qmfs->v_index + 640 + 48] = 0;
        for (n = 1; n < 16; n++)
        {
            qmfs->v[qmfs->v_index + 48+n] = qmfs->v[qmfs->v_index + 640 + 48+n] = -qmfs->v[qmfs->v_index + 48-n];
        }

        /* calculate 32 output samples and window */
        for (k = 0; k < 32; k++)
        {
            output[out++] = MUL_F(qmfs->v[qmfs->v_index + k], qmf_c[2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 96 + k], qmf_c[64 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 128 + k], qmf_c[128 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 224 + k], qmf_c[192 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 256 + k], qmf_c[256 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 352 + k], qmf_c[320 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 384 + k], qmf_c[384 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 480 + k], qmf_c[448 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 512 + k], qmf_c[512 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 608 + k], qmf_c[576 + 2*k]);
        }

        /* update the ringbuffer index */
        qmfs->v_index -= 64;
        if (qmfs->v_index < 0)
            qmfs->v_index = (640-64);
    }
}

void sbr_qmf_synthesis_64(sbr_info *sbr, qmfs_info *qmfs, qmf_t X[MAX_NTSRHFG][64],
                          real_t *output)
{
    ALIGN real_t x[64];
    ALIGN real_t y[64];
    int16_t n, k, out = 0;
    uint8_t l;


    /* qmf subsample l */
    for (l = 0; l < sbr->numTimeSlotsRate; l++)
    {
        /* shift buffers */
        /* we are not shifting v, it is a double ringbuffer */
        //memmove(qmfs->v + 128, qmfs->v, (1280-128)*sizeof(real_t));

        /* calculate 128 samples */
        for (k = 0; k < 32; k++)
        {
#ifdef FIXED_POINT
            y[k] = (QMF_RE(X[l][k]) - QMF_RE(X[l][63 - k]));
            x[k] = (QMF_RE(X[l][k]) + QMF_RE(X[l][63 - k]));
#else
            y[k] = (QMF_RE(X[l][k]) - QMF_RE(X[l][63 - k])) / 32.0;
            x[k] = (QMF_RE(X[l][k]) + QMF_RE(X[l][63 - k])) / 32.0;
#endif
        }

        /* even n samples */
        DCT2_32_unscaled(x, x);
        /* odd n samples */
        DCT4_32(y, y);

        for (n = 16; n < 48; n++)
        {
            qmfs->v[qmfs->v_index + n*2]   = qmfs->v[qmfs->v_index + 1280 + n*2]   = x[n-16];
            qmfs->v[qmfs->v_index + n*2+1] = qmfs->v[qmfs->v_index + 1280 + n*2+1] = y[n-16];
        }
        for (n = 0; n < 32; n++)
        {
            qmfs->v[qmfs->v_index + n] = qmfs->v[qmfs->v_index + 1280 + n] = qmfs->v[qmfs->v_index + 64-n];
        }
        qmfs->v[qmfs->v_index + 96] = qmfs->v[qmfs->v_index + 1280 + 96] = 0;
        for (n = 1; n < 32; n++)
        {
            qmfs->v[qmfs->v_index + 96+n] = qmfs->v[qmfs->v_index + 1280 + 96+n] = -qmfs->v[qmfs->v_index + 96-n];
        }

        /* calculate 64 output samples and window */
        for (k = 0; k < 64; k++)
        {
            output[out++] = MUL_F(qmfs->v[qmfs->v_index + k], qmf_c[k]) +
                MUL_F(qmfs->v[qmfs->v_index + 192 + k], qmf_c[64 + k]) +
                MUL_F(qmfs->v[qmfs->v_index + 256 + k], qmf_c[128 + k]) +
                MUL_F(qmfs->v[qmfs->v_index + 256 + 192 + k], qmf_c[128 + 64 + k]) +
                MUL_F(qmfs->v[qmfs->v_index + 512 + k], qmf_c[256 + k]) +
                MUL_F(qmfs->v[qmfs->v_index + 512 + 192 + k], qmf_c[256 + 64 + k]) +
                MUL_F(qmfs->v[qmfs->v_index + 768 + k], qmf_c[384 + k]) +
                MUL_F(qmfs->v[qmfs->v_index + 768 + 192 + k], qmf_c[384 + 64 + k]) +
                MUL_F(qmfs->v[qmfs->v_index + 1024 + k], qmf_c[512 + k]) +
                MUL_F(qmfs->v[qmfs->v_index + 1024 + 192 + k], qmf_c[512 + 64 + k]);
        }

        /* update the ringbuffer index */
        qmfs->v_index -= 128;
        if (qmfs->v_index < 0)
            qmfs->v_index = (1280-128);
    }
}
#else
void sbr_qmf_synthesis_32(sbr_info *sbr, qmfs_info *qmfs, qmf_t X[MAX_NTSRHFG][64],
                          real_t *output)
{
    ALIGN real_t x1[32], x2[32];
#ifndef FIXED_POINT
    real_t scale = 1.f/64.f;
#endif
    int16_t n, k, out = 0;
    uint8_t l;


    /* qmf subsample l */
    for (l = 0; l < sbr->numTimeSlotsRate; l++)
    {
        /* shift buffer v */
        /* buffer is not shifted, we are using a ringbuffer */
        //memmove(qmfs->v + 64, qmfs->v, (640-64)*sizeof(real_t));

        /* calculate 64 samples */
        /* complex pre-twiddle */
        for (k = 0; k < 32; k++)
        {
            x1[k] = MUL_F(QMF_RE(X[l][k]), RE(qmf32_pre_twiddle[k])) - MUL_F(QMF_IM(X[l][k]), IM(qmf32_pre_twiddle[k]));
            x2[k] = MUL_F(QMF_IM(X[l][k]), RE(qmf32_pre_twiddle[k])) + MUL_F(QMF_RE(X[l][k]), IM(qmf32_pre_twiddle[k]));

#ifndef FIXED_POINT
            x1[k] *= scale;
            x2[k] *= scale;
#else
            x1[k] >>= 1;
            x2[k] >>= 1;
#endif
        }

        /* transform */
        DCT4_32(x1, x1);
        DST4_32(x2, x2);

        for (n = 0; n < 32; n++)
        {
            qmfs->v[qmfs->v_index + n]      = qmfs->v[qmfs->v_index + 640 + n]      = -x1[n] + x2[n];
            qmfs->v[qmfs->v_index + 63 - n] = qmfs->v[qmfs->v_index + 640 + 63 - n] =  x1[n] + x2[n];
        }

        /* calculate 32 output samples and window */
        for (k = 0; k < 32; k++)
        {
            output[out++] = MUL_F(qmfs->v[qmfs->v_index + k], qmf_c[2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 96 + k], qmf_c[64 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 128 + k], qmf_c[128 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 224 + k], qmf_c[192 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 256 + k], qmf_c[256 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 352 + k], qmf_c[320 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 384 + k], qmf_c[384 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 480 + k], qmf_c[448 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 512 + k], qmf_c[512 + 2*k]) +
                MUL_F(qmfs->v[qmfs->v_index + 608 + k], qmf_c[576 + 2*k]);
        }

        /* update ringbuffer index */
        qmfs->v_index -= 64;
        if (qmfs->v_index < 0)
            qmfs->v_index = (640 - 64);
    }
}

void sbr_qmf_synthesis_64(sbr_info *sbr, qmfs_info *qmfs, qmf_t X[MAX_NTSRHFG][64],
                          real_t *output)
{
//    ALIGN real_t x1[64], x2[64];
#ifndef SBR_LOW_POWER
    ALIGN real_t in_real1[32], in_imag1[32], out_real1[32], out_imag1[32];
    ALIGN real_t in_real2[32], in_imag2[32], out_real2[32], out_imag2[32];
#endif
    qmf_t * pX;
    real_t * pring_buffer_1, * pring_buffer_3;
//    real_t * ptemp_1, * ptemp_2;
#ifdef PREFER_POINTERS
    // These pointers are used if target platform has autoinc address generators
    real_t * pring_buffer_2, * pring_buffer_4;
    real_t * pring_buffer_5, * pring_buffer_6;
    real_t * pring_buffer_7, * pring_buffer_8;
    real_t * pring_buffer_9, * pring_buffer_10;
    const real_t * pqmf_c_1, * pqmf_c_2, * pqmf_c_3, * pqmf_c_4;
    const real_t * pqmf_c_5, * pqmf_c_6, * pqmf_c_7, * pqmf_c_8;
    const real_t * pqmf_c_9, * pqmf_c_10;
#endif // #ifdef PREFER_POINTERS
#ifndef FIXED_POINT
    real_t scale = 1.f/64.f;
#endif
    int16_t n, k, out = 0;
    uint8_t l;


    /* qmf subsample l */
    for (l = 0; l < sbr->numTimeSlotsRate; l++)
    {
        /* shift buffer v */
		/* buffer is not shifted, we use double ringbuffer */
		//memmove(qmfs->v + 128, qmfs->v, (1280-128)*sizeof(real_t));

        /* calculate 128 samples */
#ifndef FIXED_POINT

        pX = X[l];

        in_imag1[31] = scale*QMF_RE(pX[1]);
        in_real1[0]  = scale*QMF_RE(pX[0]);
        in_imag2[31] = scale*QMF_IM(pX[63-1]);
        in_real2[0]  = scale*QMF_IM(pX[63-0]);
        for (k = 1; k < 31; k++)
        {
            in_imag1[31 - k] = scale*QMF_RE(pX[2*k + 1]);
            in_real1[     k] = scale*QMF_RE(pX[2*k    ]);
            in_imag2[31 - k] = scale*QMF_IM(pX[63 - (2*k + 1)]);
            in_real2[     k] = scale*QMF_IM(pX[63 - (2*k    )]);
        }
        in_imag1[0]  = scale*QMF_RE(pX[63]);
        in_real1[31] = scale*QMF_RE(pX[62]);
        in_imag2[0]  = scale*QMF_IM(pX[63-63]);
        in_real2[31] = scale*QMF_IM(pX[63-62]);

#else

        pX = X[l];

        in_imag1[31] = QMF_RE(pX[1]) >> 1;
        in_real1[0]  = QMF_RE(pX[0]) >> 1;
        in_imag2[31] = QMF_IM(pX[62]) >> 1;
        in_real2[0]  = QMF_IM(pX[63]) >> 1;
        for (k = 1; k < 31; k++)
        {
            in_imag1[31 - k] = QMF_RE(pX[2*k + 1]) >> 1;
            in_real1[     k] = QMF_RE(pX[2*k    ]) >> 1;
            in_imag2[31 - k] = QMF_IM(pX[63 - (2*k + 1)]) >> 1;
            in_real2[     k] = QMF_IM(pX[63 - (2*k    )]) >> 1;
        }
        in_imag1[0]  = QMF_RE(pX[63]) >> 1;
        in_real1[31] = QMF_RE(pX[62]) >> 1;
        in_imag2[0]  = QMF_IM(pX[0]) >> 1;
        in_real2[31] = QMF_IM(pX[1]) >> 1;

#endif


        // dct4_kernel is DCT_IV without reordering which is done before and after FFT
        dct4_kernel(in_real1, in_imag1, out_real1, out_imag1);
        dct4_kernel(in_real2, in_imag2, out_real2, out_imag2);


        pring_buffer_1 = qmfs->v + qmfs->v_index;
        pring_buffer_3 = pring_buffer_1 + 1280;
#ifdef PREFER_POINTERS
        pring_buffer_2 = pring_buffer_1 + 127;
        pring_buffer_4 = pring_buffer_1 + (1280 + 127);
#endif // #ifdef PREFER_POINTERS
//        ptemp_1 = x1;
//        ptemp_2 = x2;
#ifdef PREFER_POINTERS
        for (n = 0; n < 32; n ++)
        {
            //real_t x1 = *ptemp_1++;
            //real_t x2 = *ptemp_2++;
            // pring_buffer_3 and pring_buffer_4 are needed only for double ring buffer
            *pring_buffer_1++ = *pring_buffer_3++ = out_real2[n] - out_real1[n];
            *pring_buffer_2-- = *pring_buffer_4-- = out_real2[n] + out_real1[n];
            //x1 = *ptemp_1++;
            //x2 = *ptemp_2++;
            *pring_buffer_1++ = *pring_buffer_3++ = out_imag2[31-n] + out_imag1[31-n];
            *pring_buffer_2-- = *pring_buffer_4-- = out_imag2[31-n] - out_imag1[31-n];
        }
#else // #ifdef PREFER_POINTERS

        for (n = 0; n < 32; n++)
        {
            // pring_buffer_3 and pring_buffer_4 are needed only for double ring buffer
            pring_buffer_1[2*n]         = pring_buffer_3[2*n]         = out_real2[n] - out_real1[n];
            pring_buffer_1[127-2*n]     = pring_buffer_3[127-2*n]     = out_real2[n] + out_real1[n];
            pring_buffer_1[2*n+1]       = pring_buffer_3[2*n+1]       = out_imag2[31-n] + out_imag1[31-n];
            pring_buffer_1[127-(2*n+1)] = pring_buffer_3[127-(2*n+1)] = out_imag2[31-n] - out_imag1[31-n];
        }

#endif // #ifdef PREFER_POINTERS

        pring_buffer_1 = qmfs->v + qmfs->v_index;
#ifdef PREFER_POINTERS
        pring_buffer_2 = pring_buffer_1 + 192;
        pring_buffer_3 = pring_buffer_1 + 256;
        pring_buffer_4 = pring_buffer_1 + (256 + 192);
        pring_buffer_5 = pring_buffer_1 + 512;
        pring_buffer_6 = pring_buffer_1 + (512 + 192);
        pring_buffer_7 = pring_buffer_1 + 768;
        pring_buffer_8 = pring_buffer_1 + (768 + 192);
        pring_buffer_9 = pring_buffer_1 + 1024;
        pring_buffer_10 = pring_buffer_1 + (1024 + 192);
        pqmf_c_1 = qmf_c;
        pqmf_c_2 = qmf_c + 64;
        pqmf_c_3 = qmf_c + 128;
        pqmf_c_4 = qmf_c + 192;
        pqmf_c_5 = qmf_c + 256;
        pqmf_c_6 = qmf_c + 320;
        pqmf_c_7 = qmf_c + 384;
        pqmf_c_8 = qmf_c + 448;
        pqmf_c_9 = qmf_c + 512;
        pqmf_c_10 = qmf_c + 576;
#endif // #ifdef PREFER_POINTERS

        /* calculate 64 output samples and window */
        for (k = 0; k < 64; k++)
        {
#ifdef PREFER_POINTERS
            output[out++] =
                MUL_F(*pring_buffer_1++,  *pqmf_c_1++) +
                MUL_F(*pring_buffer_2++,  *pqmf_c_2++) +
                MUL_F(*pring_buffer_3++,  *pqmf_c_3++) +
                MUL_F(*pring_buffer_4++,  *pqmf_c_4++) +
                MUL_F(*pring_buffer_5++,  *pqmf_c_5++) +
                MUL_F(*pring_buffer_6++,  *pqmf_c_6++) +
                MUL_F(*pring_buffer_7++,  *pqmf_c_7++) +
                MUL_F(*pring_buffer_8++,  *pqmf_c_8++) +
                MUL_F(*pring_buffer_9++,  *pqmf_c_9++) +
                MUL_F(*pring_buffer_10++, *pqmf_c_10++);
#else // #ifdef PREFER_POINTERS
            output[out++] =
                MUL_F(pring_buffer_1[k+0],          qmf_c[k+0])   +
                MUL_F(pring_buffer_1[k+192],        qmf_c[k+64])  +
                MUL_F(pring_buffer_1[k+256],        qmf_c[k+128]) +
                MUL_F(pring_buffer_1[k+(256+192)],  qmf_c[k+192]) +
                MUL_F(pring_buffer_1[k+512],        qmf_c[k+256]) +
                MUL_F(pring_buffer_1[k+(512+192)],  qmf_c[k+320]) +
                MUL_F(pring_buffer_1[k+768],        qmf_c[k+384]) +
                MUL_F(pring_buffer_1[k+(768+192)],  qmf_c[k+448]) +
                MUL_F(pring_buffer_1[k+1024],       qmf_c[k+512]) +
                MUL_F(pring_buffer_1[k+(1024+192)], qmf_c[k+576]);
#endif // #ifdef PREFER_POINTERS
        }

        /* update ringbuffer index */
        qmfs->v_index -= 128;
        if (qmfs->v_index < 0)
            qmfs->v_index = (1280 - 128);
    }
}
#endif

#endif