?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
man/hu/mplayer.1 synced with r19052
[mplayer/greg.git] / liba52 / downmix.c
blobfc28071618bab57dfe66e6f66d7bcd74c5051a72
1 /*
2 * downmix.c
3 * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
4 * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
5 *
6 * This file is part of a52dec, a free ATSC A-52 stream decoder.
7 * See http://liba52.sourceforge.net/ for updates.
8 *
9 * Modified for use with MPlayer, changes contained in liba52_changes.diff.
10 * detailed changelog at http://svn.mplayerhq.hu/mplayer/trunk/
11 * $Id$
12 *
13 * a52dec is free software; you can redistribute it and/or modify
14 * it under the terms of the GNU General Public License as published by
15 * the Free Software Foundation; either version 2 of the License, or
16 * (at your option) any later version.
17 *
18 * a52dec is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21 * GNU General Public License for more details.
22 *
23 * You should have received a copy of the GNU General Public License
24 * along with this program; if not, write to the Free Software
25 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
26 *
27 * SSE optimizations from Michael Niedermayer (michaelni@gmx.at)
28 */
29
30 #include "config.h"
31 #include "asmalign.h"
32
33 #include <string.h>
34 #include <inttypes.h>
35
36 #include "a52.h"
37 #include "a52_internal.h"
38 #include "mm_accel.h"
39
40 #define CONVERT(acmod,output) (((output) << 3) + (acmod))
41
42
43 void (*a52_downmix)(sample_t * samples, int acmod, int output, sample_t bias,
44 sample_t clev, sample_t slev)= NULL;
45 void (*a52_upmix)(sample_t * samples, int acmod, int output)= NULL;
46
47 static void downmix_SSE (sample_t * samples, int acmod, int output, sample_t bias,
48 sample_t clev, sample_t slev);
49 static void downmix_3dnow (sample_t * samples, int acmod, int output, sample_t bias,
50 sample_t clev, sample_t slev);
51 static void downmix_C (sample_t * samples, int acmod, int output, sample_t bias,
52 sample_t clev, sample_t slev);
53 static void upmix_MMX (sample_t * samples, int acmod, int output);
54 static void upmix_C (sample_t * samples, int acmod, int output);
55
56 void downmix_accel_init(uint32_t mm_accel)
57 {
58 a52_upmix= upmix_C;
59 a52_downmix= downmix_C;
60 #if defined(ARCH_X86) || defined(ARCH_X86_64)
61 if(mm_accel & MM_ACCEL_X86_MMX) a52_upmix= upmix_MMX;
62 if(mm_accel & MM_ACCEL_X86_SSE) a52_downmix= downmix_SSE;
63 if(mm_accel & MM_ACCEL_X86_3DNOW) a52_downmix= downmix_3dnow;
64 #endif
65 }
66
67 int a52_downmix_init (int input, int flags, sample_t * level,
68 sample_t clev, sample_t slev)
69 {
70 static uint8_t table[11][8] = {
71 {A52_CHANNEL, A52_DOLBY, A52_STEREO, A52_STEREO,
72 A52_STEREO, A52_STEREO, A52_STEREO, A52_STEREO},
73 {A52_MONO, A52_MONO, A52_MONO, A52_MONO,
74 A52_MONO, A52_MONO, A52_MONO, A52_MONO},
75 {A52_CHANNEL, A52_DOLBY, A52_STEREO, A52_STEREO,
76 A52_STEREO, A52_STEREO, A52_STEREO, A52_STEREO},
77 {A52_CHANNEL, A52_DOLBY, A52_STEREO, A52_3F,
78 A52_STEREO, A52_3F, A52_STEREO, A52_3F},
79 {A52_CHANNEL, A52_DOLBY, A52_STEREO, A52_STEREO,
80 A52_2F1R, A52_2F1R, A52_2F1R, A52_2F1R},
81 {A52_CHANNEL, A52_DOLBY, A52_STEREO, A52_STEREO,
82 A52_2F1R, A52_3F1R, A52_2F1R, A52_3F1R},
83 {A52_CHANNEL, A52_DOLBY, A52_STEREO, A52_3F,
84 A52_2F2R, A52_2F2R, A52_2F2R, A52_2F2R},
85 {A52_CHANNEL, A52_DOLBY, A52_STEREO, A52_3F,
86 A52_2F2R, A52_3F2R, A52_2F2R, A52_3F2R},
87 {A52_CHANNEL1, A52_MONO, A52_MONO, A52_MONO,
88 A52_MONO, A52_MONO, A52_MONO, A52_MONO},
89 {A52_CHANNEL2, A52_MONO, A52_MONO, A52_MONO,
90 A52_MONO, A52_MONO, A52_MONO, A52_MONO},
91 {A52_CHANNEL, A52_DOLBY, A52_STEREO, A52_DOLBY,
92 A52_DOLBY, A52_DOLBY, A52_DOLBY, A52_DOLBY}
93 };
94 int output;
95
96 output = flags & A52_CHANNEL_MASK;
97 if (output > A52_DOLBY)
98 return -1;
99
100 output = table[output][input & 7];
101
102 if ((output == A52_STEREO) &&
103 ((input == A52_DOLBY) || ((input == A52_3F) && (clev == LEVEL_3DB))))
104 output = A52_DOLBY;
105
106 if (flags & A52_ADJUST_LEVEL)
107 switch (CONVERT (input & 7, output)) {
108
109 case CONVERT (A52_3F, A52_MONO):
110 *level *= LEVEL_3DB / (1 + clev);
111 break;
112
113 case CONVERT (A52_STEREO, A52_MONO):
114 case CONVERT (A52_2F2R, A52_2F1R):
115 case CONVERT (A52_3F2R, A52_3F1R):
116 level_3db:
117 *level *= LEVEL_3DB;
118 break;
119
120 case CONVERT (A52_3F2R, A52_2F1R):
121 if (clev < LEVEL_PLUS3DB - 1)
122 goto level_3db;
123 /* break thru */
124 case CONVERT (A52_3F, A52_STEREO):
125 case CONVERT (A52_3F1R, A52_2F1R):
126 case CONVERT (A52_3F1R, A52_2F2R):
127 case CONVERT (A52_3F2R, A52_2F2R):
128 *level /= 1 + clev;
129 break;
130
131 case CONVERT (A52_2F1R, A52_MONO):
132 *level *= LEVEL_PLUS3DB / (2 + slev);
133 break;
134
135 case CONVERT (A52_2F1R, A52_STEREO):
136 case CONVERT (A52_3F1R, A52_3F):
137 *level /= 1 + slev * LEVEL_3DB;
138 break;
139
140 case CONVERT (A52_3F1R, A52_MONO):
141 *level *= LEVEL_3DB / (1 + clev + 0.5 * slev);
142 break;
143
144 case CONVERT (A52_3F1R, A52_STEREO):
145 *level /= 1 + clev + slev * LEVEL_3DB;
146 break;
147
148 case CONVERT (A52_2F2R, A52_MONO):
149 *level *= LEVEL_3DB / (1 + slev);
150 break;
151
152 case CONVERT (A52_2F2R, A52_STEREO):
153 case CONVERT (A52_3F2R, A52_3F):
154 *level /= 1 + slev;
155 break;
156
157 case CONVERT (A52_3F2R, A52_MONO):
158 *level *= LEVEL_3DB / (1 + clev + slev);
159 break;
160
161 case CONVERT (A52_3F2R, A52_STEREO):
162 *level /= 1 + clev + slev;
163 break;
164
165 case CONVERT (A52_MONO, A52_DOLBY):
166 *level *= LEVEL_PLUS3DB;
167 break;
168
169 case CONVERT (A52_3F, A52_DOLBY):
170 case CONVERT (A52_2F1R, A52_DOLBY):
171 *level *= 1 / (1 + LEVEL_3DB);
172 break;
173
174 case CONVERT (A52_3F1R, A52_DOLBY):
175 case CONVERT (A52_2F2R, A52_DOLBY):
176 *level *= 1 / (1 + 2 * LEVEL_3DB);
177 break;
178
179 case CONVERT (A52_3F2R, A52_DOLBY):
180 *level *= 1 / (1 + 3 * LEVEL_3DB);
181 break;
182 }
183
184 return output;
185 }
186
187 int a52_downmix_coeff (sample_t * coeff, int acmod, int output, sample_t level,
188 sample_t clev, sample_t slev)
189 {
190 switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
191
192 case CONVERT (A52_CHANNEL, A52_CHANNEL):
193 case CONVERT (A52_MONO, A52_MONO):
194 case CONVERT (A52_STEREO, A52_STEREO):
195 case CONVERT (A52_3F, A52_3F):
196 case CONVERT (A52_2F1R, A52_2F1R):
197 case CONVERT (A52_3F1R, A52_3F1R):
198 case CONVERT (A52_2F2R, A52_2F2R):
199 case CONVERT (A52_3F2R, A52_3F2R):
200 case CONVERT (A52_STEREO, A52_DOLBY):
201 coeff[0] = coeff[1] = coeff[2] = coeff[3] = coeff[4] = level;
202 return 0;
203
204 case CONVERT (A52_CHANNEL, A52_MONO):
205 coeff[0] = coeff[1] = level * LEVEL_6DB;
206 return 3;
207
208 case CONVERT (A52_STEREO, A52_MONO):
209 coeff[0] = coeff[1] = level * LEVEL_3DB;
210 return 3;
211
212 case CONVERT (A52_3F, A52_MONO):
213 coeff[0] = coeff[2] = level * LEVEL_3DB;
214 coeff[1] = level * clev * LEVEL_PLUS3DB;
215 return 7;
216
217 case CONVERT (A52_2F1R, A52_MONO):
218 coeff[0] = coeff[1] = level * LEVEL_3DB;
219 coeff[2] = level * slev * LEVEL_3DB;
220 return 7;
221
222 case CONVERT (A52_2F2R, A52_MONO):
223 coeff[0] = coeff[1] = level * LEVEL_3DB;
224 coeff[2] = coeff[3] = level * slev * LEVEL_3DB;
225 return 15;
226
227 case CONVERT (A52_3F1R, A52_MONO):
228 coeff[0] = coeff[2] = level * LEVEL_3DB;
229 coeff[1] = level * clev * LEVEL_PLUS3DB;
230 coeff[3] = level * slev * LEVEL_3DB;
231 return 15;
232
233 case CONVERT (A52_3F2R, A52_MONO):
234 coeff[0] = coeff[2] = level * LEVEL_3DB;
235 coeff[1] = level * clev * LEVEL_PLUS3DB;
236 coeff[3] = coeff[4] = level * slev * LEVEL_3DB;
237 return 31;
238
239 case CONVERT (A52_MONO, A52_DOLBY):
240 coeff[0] = level * LEVEL_3DB;
241 return 0;
242
243 case CONVERT (A52_3F, A52_DOLBY):
244 clev = LEVEL_3DB;
245 case CONVERT (A52_3F, A52_STEREO):
246 case CONVERT (A52_3F1R, A52_2F1R):
247 case CONVERT (A52_3F2R, A52_2F2R):
248 coeff[0] = coeff[2] = coeff[3] = coeff[4] = level;
249 coeff[1] = level * clev;
250 return 7;
251
252 case CONVERT (A52_2F1R, A52_DOLBY):
253 slev = 1;
254 case CONVERT (A52_2F1R, A52_STEREO):
255 coeff[0] = coeff[1] = level;
256 coeff[2] = level * slev * LEVEL_3DB;
257 return 7;
258
259 case CONVERT (A52_3F1R, A52_DOLBY):
260 clev = LEVEL_3DB;
261 slev = 1;
262 case CONVERT (A52_3F1R, A52_STEREO):
263 coeff[0] = coeff[2] = level;
264 coeff[1] = level * clev;
265 coeff[3] = level * slev * LEVEL_3DB;
266 return 15;
267
268 case CONVERT (A52_2F2R, A52_DOLBY):
269 slev = LEVEL_3DB;
270 case CONVERT (A52_2F2R, A52_STEREO):
271 coeff[0] = coeff[1] = level;
272 coeff[2] = coeff[3] = level * slev;
273 return 15;
274
275 case CONVERT (A52_3F2R, A52_DOLBY):
276 clev = LEVEL_3DB;
277 case CONVERT (A52_3F2R, A52_2F1R):
278 slev = LEVEL_3DB;
279 case CONVERT (A52_3F2R, A52_STEREO):
280 coeff[0] = coeff[2] = level;
281 coeff[1] = level * clev;
282 coeff[3] = coeff[4] = level * slev;
283 return 31;
284
285 case CONVERT (A52_3F1R, A52_3F):
286 coeff[0] = coeff[1] = coeff[2] = level;
287 coeff[3] = level * slev * LEVEL_3DB;
288 return 13;
289
290 case CONVERT (A52_3F2R, A52_3F):
291 coeff[0] = coeff[1] = coeff[2] = level;
292 coeff[3] = coeff[4] = level * slev;
293 return 29;
294
295 case CONVERT (A52_2F2R, A52_2F1R):
296 coeff[0] = coeff[1] = level;
297 coeff[2] = coeff[3] = level * LEVEL_3DB;
298 return 12;
299
300 case CONVERT (A52_3F2R, A52_3F1R):
301 coeff[0] = coeff[1] = coeff[2] = level;
302 coeff[3] = coeff[4] = level * LEVEL_3DB;
303 return 24;
304
305 case CONVERT (A52_2F1R, A52_2F2R):
306 coeff[0] = coeff[1] = level;
307 coeff[2] = level * LEVEL_3DB;
308 return 0;
309
310 case CONVERT (A52_3F1R, A52_2F2R):
311 coeff[0] = coeff[2] = level;
312 coeff[1] = level * clev;
313 coeff[3] = level * LEVEL_3DB;
314 return 7;
315
316 case CONVERT (A52_3F1R, A52_3F2R):
317 coeff[0] = coeff[1] = coeff[2] = level;
318 coeff[3] = level * LEVEL_3DB;
319 return 0;
320
321 case CONVERT (A52_CHANNEL, A52_CHANNEL1):
322 coeff[0] = level;
323 coeff[1] = 0;
324 return 0;
325
326 case CONVERT (A52_CHANNEL, A52_CHANNEL2):
327 coeff[0] = 0;
328 coeff[1] = level;
329 return 0;
330 }
331
332 return -1; /* NOTREACHED */
333 }
334
335 static void mix2to1 (sample_t * dest, sample_t * src, sample_t bias)
336 {
337 int i;
338
339 for (i = 0; i < 256; i++)
340 dest[i] += src[i] + bias;
341 }
342
343 static void mix3to1 (sample_t * samples, sample_t bias)
344 {
345 int i;
346
347 for (i = 0; i < 256; i++)
348 samples[i] += samples[i + 256] + samples[i + 512] + bias;
349 }
350
351 static void mix4to1 (sample_t * samples, sample_t bias)
352 {
353 int i;
354
355 for (i = 0; i < 256; i++)
356 samples[i] += (samples[i + 256] + samples[i + 512] +
357 samples[i + 768] + bias);
358 }
359
360 static void mix5to1 (sample_t * samples, sample_t bias)
361 {
362 int i;
363
364 for (i = 0; i < 256; i++)
365 samples[i] += (samples[i + 256] + samples[i + 512] +
366 samples[i + 768] + samples[i + 1024] + bias);
367 }
368
369 static void mix3to2 (sample_t * samples, sample_t bias)
370 {
371 int i;
372 sample_t common;
373
374 for (i = 0; i < 256; i++) {
375 common = samples[i + 256] + bias;
376 samples[i] += common;
377 samples[i + 256] = samples[i + 512] + common;
378 }
379 }
380
381 static void mix21to2 (sample_t * left, sample_t * right, sample_t bias)
382 {
383 int i;
384 sample_t common;
385
386 for (i = 0; i < 256; i++) {
387 common = right[i + 256] + bias;
388 left[i] += common;
389 right[i] += common;
390 }
391 }
392
393 static void mix21toS (sample_t * samples, sample_t bias)
394 {
395 int i;
396 sample_t surround;
397
398 for (i = 0; i < 256; i++) {
399 surround = samples[i + 512];
400 samples[i] += bias - surround;
401 samples[i + 256] += bias + surround;
402 }
403 }
404
405 static void mix31to2 (sample_t * samples, sample_t bias)
406 {
407 int i;
408 sample_t common;
409
410 for (i = 0; i < 256; i++) {
411 common = samples[i + 256] + samples[i + 768] + bias;
412 samples[i] += common;
413 samples[i + 256] = samples[i + 512] + common;
414 }
415 }
416
417 static void mix31toS (sample_t * samples, sample_t bias)
418 {
419 int i;
420 sample_t common, surround;
421
422 for (i = 0; i < 256; i++) {
423 common = samples[i + 256] + bias;
424 surround = samples[i + 768];
425 samples[i] += common - surround;
426 samples[i + 256] = samples[i + 512] + common + surround;
427 }
428 }
429
430 static void mix22toS (sample_t * samples, sample_t bias)
431 {
432 int i;
433 sample_t surround;
434
435 for (i = 0; i < 256; i++) {
436 surround = samples[i + 512] + samples[i + 768];
437 samples[i] += bias - surround;
438 samples[i + 256] += bias + surround;
439 }
440 }
441
442 static void mix32to2 (sample_t * samples, sample_t bias)
443 {
444 int i;
445 sample_t common;
446
447 for (i = 0; i < 256; i++) {
448 common = samples[i + 256] + bias;
449 samples[i] += common + samples[i + 768];
450 samples[i + 256] = common + samples[i + 512] + samples[i + 1024];
451 }
452 }
453
454 static void mix32toS (sample_t * samples, sample_t bias)
455 {
456 int i;
457 sample_t common, surround;
458
459 for (i = 0; i < 256; i++) {
460 common = samples[i + 256] + bias;
461 surround = samples[i + 768] + samples[i + 1024];
462 samples[i] += common - surround;
463 samples[i + 256] = samples[i + 512] + common + surround;
464 }
465 }
466
467 static void move2to1 (sample_t * src, sample_t * dest, sample_t bias)
468 {
469 int i;
470
471 for (i = 0; i < 256; i++)
472 dest[i] = src[i] + src[i + 256] + bias;
473 }
474
475 static void zero (sample_t * samples)
476 {
477 int i;
478
479 for (i = 0; i < 256; i++)
480 samples[i] = 0;
481 }
482
483 void downmix_C (sample_t * samples, int acmod, int output, sample_t bias,
484 sample_t clev, sample_t slev)
485 {
486 switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
487
488 case CONVERT (A52_CHANNEL, A52_CHANNEL2):
489 memcpy (samples, samples + 256, 256 * sizeof (sample_t));
490 break;
491
492 case CONVERT (A52_CHANNEL, A52_MONO):
493 case CONVERT (A52_STEREO, A52_MONO):
494 mix_2to1:
495 mix2to1 (samples, samples + 256, bias);
496 break;
497
498 case CONVERT (A52_2F1R, A52_MONO):
499 if (slev == 0)
500 goto mix_2to1;
501 case CONVERT (A52_3F, A52_MONO):
502 mix_3to1:
503 mix3to1 (samples, bias);
504 break;
505
506 case CONVERT (A52_3F1R, A52_MONO):
507 if (slev == 0)
508 goto mix_3to1;
509 case CONVERT (A52_2F2R, A52_MONO):
510 if (slev == 0)
511 goto mix_2to1;
512 mix4to1 (samples, bias);
513 break;
514
515 case CONVERT (A52_3F2R, A52_MONO):
516 if (slev == 0)
517 goto mix_3to1;
518 mix5to1 (samples, bias);
519 break;
520
521 case CONVERT (A52_MONO, A52_DOLBY):
522 memcpy (samples + 256, samples, 256 * sizeof (sample_t));
523 break;
524
525 case CONVERT (A52_3F, A52_STEREO):
526 case CONVERT (A52_3F, A52_DOLBY):
527 mix_3to2:
528 mix3to2 (samples, bias);
529 break;
530
531 case CONVERT (A52_2F1R, A52_STEREO):
532 if (slev == 0)
533 break;
534 mix21to2 (samples, samples + 256, bias);
535 break;
536
537 case CONVERT (A52_2F1R, A52_DOLBY):
538 mix21toS (samples, bias);
539 break;
540
541 case CONVERT (A52_3F1R, A52_STEREO):
542 if (slev == 0)
543 goto mix_3to2;
544 mix31to2 (samples, bias);
545 break;
546
547 case CONVERT (A52_3F1R, A52_DOLBY):
548 mix31toS (samples, bias);
549 break;
550
551 case CONVERT (A52_2F2R, A52_STEREO):
552 if (slev == 0)
553 break;
554 mix2to1 (samples, samples + 512, bias);
555 mix2to1 (samples + 256, samples + 768, bias);
556 break;
557
558 case CONVERT (A52_2F2R, A52_DOLBY):
559 mix22toS (samples, bias);
560 break;
561
562 case CONVERT (A52_3F2R, A52_STEREO):
563 if (slev == 0)
564 goto mix_3to2;
565 mix32to2 (samples, bias);
566 break;
567
568 case CONVERT (A52_3F2R, A52_DOLBY):
569 mix32toS (samples, bias);
570 break;
571
572 case CONVERT (A52_3F1R, A52_3F):
573 if (slev == 0)
574 break;
575 mix21to2 (samples, samples + 512, bias);
576 break;
577
578 case CONVERT (A52_3F2R, A52_3F):
579 if (slev == 0)
580 break;
581 mix2to1 (samples, samples + 768, bias);
582 mix2to1 (samples + 512, samples + 1024, bias);
583 break;
584
585 case CONVERT (A52_3F1R, A52_2F1R):
586 mix3to2 (samples, bias);
587 memcpy (samples + 512, samples + 768, 256 * sizeof (sample_t));
588 break;
589
590 case CONVERT (A52_2F2R, A52_2F1R):
591 mix2to1 (samples + 512, samples + 768, bias);
592 break;
593
594 case CONVERT (A52_3F2R, A52_2F1R):
595 mix3to2 (samples, bias); //FIXME possible bug? (output doesnt seem to be used)
596 move2to1 (samples + 768, samples + 512, bias);
597 break;
598
599 case CONVERT (A52_3F2R, A52_3F1R):
600 mix2to1 (samples + 768, samples + 1024, bias);
601 break;
602
603 case CONVERT (A52_2F1R, A52_2F2R):
604 memcpy (samples + 768, samples + 512, 256 * sizeof (sample_t));
605 break;
606
607 case CONVERT (A52_3F1R, A52_2F2R):
608 mix3to2 (samples, bias);
609 memcpy (samples + 512, samples + 768, 256 * sizeof (sample_t));
610 break;
611
612 case CONVERT (A52_3F2R, A52_2F2R):
613 mix3to2 (samples, bias);
614 memcpy (samples + 512, samples + 768, 256 * sizeof (sample_t));
615 memcpy (samples + 768, samples + 1024, 256 * sizeof (sample_t));
616 break;
617
618 case CONVERT (A52_3F1R, A52_3F2R):
619 memcpy (samples + 1024, samples + 768, 256 * sizeof (sample_t));
620 break;
621 }
622 }
623
624 void upmix_C (sample_t * samples, int acmod, int output)
625 {
626 switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
627
628 case CONVERT (A52_CHANNEL, A52_CHANNEL2):
629 memcpy (samples + 256, samples, 256 * sizeof (sample_t));
630 break;
631
632 case CONVERT (A52_3F2R, A52_MONO):
633 zero (samples + 1024);
634 case CONVERT (A52_3F1R, A52_MONO):
635 case CONVERT (A52_2F2R, A52_MONO):
636 zero (samples + 768);
637 case CONVERT (A52_3F, A52_MONO):
638 case CONVERT (A52_2F1R, A52_MONO):
639 zero (samples + 512);
640 case CONVERT (A52_CHANNEL, A52_MONO):
641 case CONVERT (A52_STEREO, A52_MONO):
642 zero (samples + 256);
643 break;
644
645 case CONVERT (A52_3F2R, A52_STEREO):
646 case CONVERT (A52_3F2R, A52_DOLBY):
647 zero (samples + 1024);
648 case CONVERT (A52_3F1R, A52_STEREO):
649 case CONVERT (A52_3F1R, A52_DOLBY):
650 zero (samples + 768);
651 case CONVERT (A52_3F, A52_STEREO):
652 case CONVERT (A52_3F, A52_DOLBY):
653 mix_3to2:
654 memcpy (samples + 512, samples + 256, 256 * sizeof (sample_t));
655 zero (samples + 256);
656 break;
657
658 case CONVERT (A52_2F2R, A52_STEREO):
659 case CONVERT (A52_2F2R, A52_DOLBY):
660 zero (samples + 768);
661 case CONVERT (A52_2F1R, A52_STEREO):
662 case CONVERT (A52_2F1R, A52_DOLBY):
663 zero (samples + 512);
664 break;
665
666 case CONVERT (A52_3F2R, A52_3F):
667 zero (samples + 1024);
668 case CONVERT (A52_3F1R, A52_3F):
669 case CONVERT (A52_2F2R, A52_2F1R):
670 zero (samples + 768);
671 break;
672
673 case CONVERT (A52_3F2R, A52_3F1R):
674 zero (samples + 1024);
675 break;
676
677 case CONVERT (A52_3F2R, A52_2F1R):
678 zero (samples + 1024);
679 case CONVERT (A52_3F1R, A52_2F1R):
680 mix_31to21:
681 memcpy (samples + 768, samples + 512, 256 * sizeof (sample_t));
682 goto mix_3to2;
683
684 case CONVERT (A52_3F2R, A52_2F2R):
685 memcpy (samples + 1024, samples + 768, 256 * sizeof (sample_t));
686 goto mix_31to21;
687 }
688 }
689
690 #if defined(ARCH_X86) || defined(ARCH_X86_64)
691 static void mix2to1_SSE (sample_t * dest, sample_t * src, sample_t bias)
692 {
693 asm volatile(
694 "movlps %2, %%xmm7 \n\t"
695 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
696 "mov $-1024, %%"REG_S" \n\t"
697 ASMALIGN16
698 "1: \n\t"
699 "movaps (%0, %%"REG_S"), %%xmm0 \n\t"
700 "movaps 16(%0, %%"REG_S"), %%xmm1\n\t"
701 "addps (%1, %%"REG_S"), %%xmm0 \n\t"
702 "addps 16(%1, %%"REG_S"), %%xmm1\n\t"
703 "addps %%xmm7, %%xmm0 \n\t"
704 "addps %%xmm7, %%xmm1 \n\t"
705 "movaps %%xmm0, (%1, %%"REG_S") \n\t"
706 "movaps %%xmm1, 16(%1, %%"REG_S")\n\t"
707 "add $32, %%"REG_S" \n\t"
708 " jnz 1b \n\t"
709 :: "r" (src+256), "r" (dest+256), "m" (bias)
710 : "%"REG_S
711 );
712 }
713
714 static void mix3to1_SSE (sample_t * samples, sample_t bias)
715 {
716 asm volatile(
717 "movlps %1, %%xmm7 \n\t"
718 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
719 "mov $-1024, %%"REG_S" \n\t"
720 ASMALIGN16
721 "1: \n\t"
722 "movaps (%0, %%"REG_S"), %%xmm0 \n\t"
723 "movaps 1024(%0, %%"REG_S"), %%xmm1\n\t"
724 "addps 2048(%0, %%"REG_S"), %%xmm0\n\t"
725 "addps %%xmm7, %%xmm1 \n\t"
726 "addps %%xmm1, %%xmm0 \n\t"
727 "movaps %%xmm0, (%0, %%"REG_S") \n\t"
728 "add $16, %%"REG_S" \n\t"
729 " jnz 1b \n\t"
730 :: "r" (samples+256), "m" (bias)
731 : "%"REG_S
732 );
733 }
734
735 static void mix4to1_SSE (sample_t * samples, sample_t bias)
736 {
737 asm volatile(
738 "movlps %1, %%xmm7 \n\t"
739 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
740 "mov $-1024, %%"REG_S" \n\t"
741 ASMALIGN16
742 "1: \n\t"
743 "movaps (%0, %%"REG_S"), %%xmm0 \n\t"
744 "movaps 1024(%0, %%"REG_S"), %%xmm1\n\t"
745 "addps 2048(%0, %%"REG_S"), %%xmm0\n\t"
746 "addps 3072(%0, %%"REG_S"), %%xmm1\n\t"
747 "addps %%xmm7, %%xmm0 \n\t"
748 "addps %%xmm1, %%xmm0 \n\t"
749 "movaps %%xmm0, (%0, %%"REG_S") \n\t"
750 "add $16, %%"REG_S" \n\t"
751 " jnz 1b \n\t"
752 :: "r" (samples+256), "m" (bias)
753 : "%"REG_S
754 );
755 }
756
757 static void mix5to1_SSE (sample_t * samples, sample_t bias)
758 {
759 asm volatile(
760 "movlps %1, %%xmm7 \n\t"
761 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
762 "mov $-1024, %%"REG_S" \n\t"
763 ASMALIGN16
764 "1: \n\t"
765 "movaps (%0, %%"REG_S"), %%xmm0 \n\t"
766 "movaps 1024(%0, %%"REG_S"), %%xmm1\n\t"
767 "addps 2048(%0, %%"REG_S"), %%xmm0\n\t"
768 "addps 3072(%0, %%"REG_S"), %%xmm1\n\t"
769 "addps %%xmm7, %%xmm0 \n\t"
770 "addps 4096(%0, %%"REG_S"), %%xmm1\n\t"
771 "addps %%xmm1, %%xmm0 \n\t"
772 "movaps %%xmm0, (%0, %%"REG_S") \n\t"
773 "add $16, %%"REG_S" \n\t"
774 " jnz 1b \n\t"
775 :: "r" (samples+256), "m" (bias)
776 : "%"REG_S
777 );
778 }
779
780 static void mix3to2_SSE (sample_t * samples, sample_t bias)
781 {
782 asm volatile(
783 "movlps %1, %%xmm7 \n\t"
784 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
785 "mov $-1024, %%"REG_S" \n\t"
786 ASMALIGN16
787 "1: \n\t"
788 "movaps 1024(%0, %%"REG_S"), %%xmm0\n\t"
789 "addps %%xmm7, %%xmm0 \n\t" //common
790 "movaps (%0, %%"REG_S"), %%xmm1 \n\t"
791 "movaps 2048(%0, %%"REG_S"), %%xmm2\n\t"
792 "addps %%xmm0, %%xmm1 \n\t"
793 "addps %%xmm0, %%xmm2 \n\t"
794 "movaps %%xmm1, (%0, %%"REG_S") \n\t"
795 "movaps %%xmm2, 1024(%0, %%"REG_S")\n\t"
796 "add $16, %%"REG_S" \n\t"
797 " jnz 1b \n\t"
798 :: "r" (samples+256), "m" (bias)
799 : "%"REG_S
800 );
801 }
802
803 static void mix21to2_SSE (sample_t * left, sample_t * right, sample_t bias)
804 {
805 asm volatile(
806 "movlps %2, %%xmm7 \n\t"
807 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
808 "mov $-1024, %%"REG_S" \n\t"
809 ASMALIGN16
810 "1: \n\t"
811 "movaps 1024(%1, %%"REG_S"), %%xmm0\n\t"
812 "addps %%xmm7, %%xmm0 \n\t" //common
813 "movaps (%0, %%"REG_S"), %%xmm1 \n\t"
814 "movaps (%1, %%"REG_S"), %%xmm2 \n\t"
815 "addps %%xmm0, %%xmm1 \n\t"
816 "addps %%xmm0, %%xmm2 \n\t"
817 "movaps %%xmm1, (%0, %%"REG_S") \n\t"
818 "movaps %%xmm2, (%1, %%"REG_S") \n\t"
819 "add $16, %%"REG_S" \n\t"
820 " jnz 1b \n\t"
821 :: "r" (left+256), "r" (right+256), "m" (bias)
822 : "%"REG_S
823 );
824 }
825
826 static void mix21toS_SSE (sample_t * samples, sample_t bias)
827 {
828 asm volatile(
829 "movlps %1, %%xmm7 \n\t"
830 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
831 "mov $-1024, %%"REG_S" \n\t"
832 ASMALIGN16
833 "1: \n\t"
834 "movaps 2048(%0, %%"REG_S"), %%xmm0\n\t" // surround
835 "movaps (%0, %%"REG_S"), %%xmm1 \n\t"
836 "movaps 1024(%0, %%"REG_S"), %%xmm2\n\t"
837 "addps %%xmm7, %%xmm1 \n\t"
838 "addps %%xmm7, %%xmm2 \n\t"
839 "subps %%xmm0, %%xmm1 \n\t"
840 "addps %%xmm0, %%xmm2 \n\t"
841 "movaps %%xmm1, (%0, %%"REG_S") \n\t"
842 "movaps %%xmm2, 1024(%0, %%"REG_S")\n\t"
843 "add $16, %%"REG_S" \n\t"
844 " jnz 1b \n\t"
845 :: "r" (samples+256), "m" (bias)
846 : "%"REG_S
847 );
848 }
849
850 static void mix31to2_SSE (sample_t * samples, sample_t bias)
851 {
852 asm volatile(
853 "movlps %1, %%xmm7 \n\t"
854 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
855 "mov $-1024, %%"REG_S" \n\t"
856 ASMALIGN16
857 "1: \n\t"
858 "movaps 1024(%0, %%"REG_S"), %%xmm0\n\t"
859 "addps 3072(%0, %%"REG_S"), %%xmm0\n\t"
860 "addps %%xmm7, %%xmm0 \n\t" // common
861 "movaps (%0, %%"REG_S"), %%xmm1 \n\t"
862 "movaps 2048(%0, %%"REG_S"), %%xmm2\n\t"
863 "addps %%xmm0, %%xmm1 \n\t"
864 "addps %%xmm0, %%xmm2 \n\t"
865 "movaps %%xmm1, (%0, %%"REG_S") \n\t"
866 "movaps %%xmm2, 1024(%0, %%"REG_S")\n\t"
867 "add $16, %%"REG_S" \n\t"
868 " jnz 1b \n\t"
869 :: "r" (samples+256), "m" (bias)
870 : "%"REG_S
871 );
872 }
873
874 static void mix31toS_SSE (sample_t * samples, sample_t bias)
875 {
876 asm volatile(
877 "movlps %1, %%xmm7 \n\t"
878 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
879 "mov $-1024, %%"REG_S" \n\t"
880 ASMALIGN16
881 "1: \n\t"
882 "movaps 1024(%0, %%"REG_S"), %%xmm0\n\t"
883 "movaps 3072(%0, %%"REG_S"), %%xmm3\n\t" // surround
884 "addps %%xmm7, %%xmm0 \n\t" // common
885 "movaps (%0, %%"REG_S"), %%xmm1 \n\t"
886 "movaps 2048(%0, %%"REG_S"), %%xmm2\n\t"
887 "addps %%xmm0, %%xmm1 \n\t"
888 "addps %%xmm0, %%xmm2 \n\t"
889 "subps %%xmm3, %%xmm1 \n\t"
890 "addps %%xmm3, %%xmm2 \n\t"
891 "movaps %%xmm1, (%0, %%"REG_S") \n\t"
892 "movaps %%xmm2, 1024(%0, %%"REG_S")\n\t"
893 "add $16, %%"REG_S" \n\t"
894 " jnz 1b \n\t"
895 :: "r" (samples+256), "m" (bias)
896 : "%"REG_S
897 );
898 }
899
900 static void mix22toS_SSE (sample_t * samples, sample_t bias)
901 {
902 asm volatile(
903 "movlps %1, %%xmm7 \n\t"
904 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
905 "mov $-1024, %%"REG_S" \n\t"
906 ASMALIGN16
907 "1: \n\t"
908 "movaps 2048(%0, %%"REG_S"), %%xmm0\n\t"
909 "addps 3072(%0, %%"REG_S"), %%xmm0\n\t" // surround
910 "movaps (%0, %%"REG_S"), %%xmm1 \n\t"
911 "movaps 1024(%0, %%"REG_S"), %%xmm2\n\t"
912 "addps %%xmm7, %%xmm1 \n\t"
913 "addps %%xmm7, %%xmm2 \n\t"
914 "subps %%xmm0, %%xmm1 \n\t"
915 "addps %%xmm0, %%xmm2 \n\t"
916 "movaps %%xmm1, (%0, %%"REG_S") \n\t"
917 "movaps %%xmm2, 1024(%0, %%"REG_S")\n\t"
918 "add $16, %%"REG_S" \n\t"
919 " jnz 1b \n\t"
920 :: "r" (samples+256), "m" (bias)
921 : "%"REG_S
922 );
923 }
924
925 static void mix32to2_SSE (sample_t * samples, sample_t bias)
926 {
927 asm volatile(
928 "movlps %1, %%xmm7 \n\t"
929 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
930 "mov $-1024, %%"REG_S" \n\t"
931 ASMALIGN16
932 "1: \n\t"
933 "movaps 1024(%0, %%"REG_S"), %%xmm0\n\t"
934 "addps %%xmm7, %%xmm0 \n\t" // common
935 "movaps %%xmm0, %%xmm1 \n\t" // common
936 "addps (%0, %%"REG_S"), %%xmm0 \n\t"
937 "addps 2048(%0, %%"REG_S"), %%xmm1\n\t"
938 "addps 3072(%0, %%"REG_S"), %%xmm0\n\t"
939 "addps 4096(%0, %%"REG_S"), %%xmm1\n\t"
940 "movaps %%xmm0, (%0, %%"REG_S") \n\t"
941 "movaps %%xmm1, 1024(%0, %%"REG_S")\n\t"
942 "add $16, %%"REG_S" \n\t"
943 " jnz 1b \n\t"
944 :: "r" (samples+256), "m" (bias)
945 : "%"REG_S
946 );
947 }
948
949 static void mix32toS_SSE (sample_t * samples, sample_t bias)
950 {
951 asm volatile(
952 "movlps %1, %%xmm7 \n\t"
953 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
954 "mov $-1024, %%"REG_S" \n\t"
955 ASMALIGN16
956 "1: \n\t"
957 "movaps 1024(%0, %%"REG_S"), %%xmm0\n\t"
958 "movaps 3072(%0, %%"REG_S"), %%xmm2\n\t"
959 "addps %%xmm7, %%xmm0 \n\t" // common
960 "addps 4096(%0, %%"REG_S"), %%xmm2\n\t" // surround
961 "movaps (%0, %%"REG_S"), %%xmm1 \n\t"
962 "movaps 2048(%0, %%"REG_S"), %%xmm3\n\t"
963 "subps %%xmm2, %%xmm1 \n\t"
964 "addps %%xmm2, %%xmm3 \n\t"
965 "addps %%xmm0, %%xmm1 \n\t"
966 "addps %%xmm0, %%xmm3 \n\t"
967 "movaps %%xmm1, (%0, %%"REG_S") \n\t"
968 "movaps %%xmm3, 1024(%0, %%"REG_S")\n\t"
969 "add $16, %%"REG_S" \n\t"
970 " jnz 1b \n\t"
971 :: "r" (samples+256), "m" (bias)
972 : "%"REG_S
973 );
974 }
975
976 static void move2to1_SSE (sample_t * src, sample_t * dest, sample_t bias)
977 {
978 asm volatile(
979 "movlps %2, %%xmm7 \n\t"
980 "shufps $0x00, %%xmm7, %%xmm7 \n\t"
981 "mov $-1024, %%"REG_S" \n\t"
982 ASMALIGN16
983 "1: \n\t"
984 "movaps (%0, %%"REG_S"), %%xmm0 \n\t"
985 "movaps 16(%0, %%"REG_S"), %%xmm1\n\t"
986 "addps 1024(%0, %%"REG_S"), %%xmm0\n\t"
987 "addps 1040(%0, %%"REG_S"), %%xmm1\n\t"
988 "addps %%xmm7, %%xmm0 \n\t"
989 "addps %%xmm7, %%xmm1 \n\t"
990 "movaps %%xmm0, (%1, %%"REG_S") \n\t"
991 "movaps %%xmm1, 16(%1, %%"REG_S")\n\t"
992 "add $32, %%"REG_S" \n\t"
993 " jnz 1b \n\t"
994 :: "r" (src+256), "r" (dest+256), "m" (bias)
995 : "%"REG_S
996 );
997 }
998
999 static void zero_MMX(sample_t * samples)
1000 {
1001 asm volatile(
1002 "mov $-1024, %%"REG_S" \n\t"
1003 "pxor %%mm0, %%mm0 \n\t"
1004 ASMALIGN16
1005 "1: \n\t"
1006 "movq %%mm0, (%0, %%"REG_S") \n\t"
1007 "movq %%mm0, 8(%0, %%"REG_S") \n\t"
1008 "movq %%mm0, 16(%0, %%"REG_S") \n\t"
1009 "movq %%mm0, 24(%0, %%"REG_S") \n\t"
1010 "add $32, %%"REG_S" \n\t"
1011 " jnz 1b \n\t"
1012 "emms"
1013 :: "r" (samples+256)
1014 : "%"REG_S
1015 );
1016 }
1017
1018 /*
1019 I hope dest and src will be at least 8 byte aligned and size
1020 will devide on 8 without remain
1021 Note: untested and unused.
1022 */
1023 static void copy_MMX(void *dest,const void *src,unsigned size)
1024 {
1025 unsigned i;
1026 size /= 64;
1027 for(i=0;i<size;i++)
1028 {
1029 __asm __volatile(
1030 "movq %0, %%mm0\n\t"
1031 "movq 8%0, %%mm1\n\t"
1032 "movq 16%0, %%mm2\n\t"
1033 "movq 24%0, %%mm3\n\t"
1034 "movq 32%0, %%mm4\n\t"
1035 "movq 40%0, %%mm5\n\t"
1036 "movq 48%0, %%mm6\n\t"
1037 "movq 56%0, %%mm7\n\t"
1038 "movq %%mm0, %1\n\t"
1039 "movq %%mm1, 8%1\n\t"
1040 "movq %%mm2, 16%1\n\t"
1041 "movq %%mm3, 24%1\n\t"
1042 "movq %%mm4, 32%1\n\t"
1043 "movq %%mm5, 40%1\n\t"
1044 "movq %%mm6, 48%1\n\t"
1045 "movq %%mm7, 56%1\n\t"
1046 :
1047 :"m"(src),"m"(dest));
1048 }
1049 }
1050
1051 static void downmix_SSE (sample_t * samples, int acmod, int output, sample_t bias,
1052 sample_t clev, sample_t slev)
1053 {
1054 switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
1055
1056 case CONVERT (A52_CHANNEL, A52_CHANNEL2):
1057 memcpy (samples, samples + 256, 256 * sizeof (sample_t));
1058 break;
1059
1060 case CONVERT (A52_CHANNEL, A52_MONO):
1061 case CONVERT (A52_STEREO, A52_MONO):
1062 mix_2to1_SSE:
1063 mix2to1_SSE (samples, samples + 256, bias);
1064 break;
1065
1066 case CONVERT (A52_2F1R, A52_MONO):
1067 if (slev == 0)
1068 goto mix_2to1_SSE;
1069 case CONVERT (A52_3F, A52_MONO):
1070 mix_3to1_SSE:
1071 mix3to1_SSE (samples, bias);
1072 break;
1073
1074 case CONVERT (A52_3F1R, A52_MONO):
1075 if (slev == 0)
1076 goto mix_3to1_SSE;
1077 case CONVERT (A52_2F2R, A52_MONO):
1078 if (slev == 0)
1079 goto mix_2to1_SSE;
1080 mix4to1_SSE (samples, bias);
1081 break;
1082
1083 case CONVERT (A52_3F2R, A52_MONO):
1084 if (slev == 0)
1085 goto mix_3to1_SSE;
1086 mix5to1_SSE (samples, bias);
1087 break;
1088
1089 case CONVERT (A52_MONO, A52_DOLBY):
1090 memcpy (samples + 256, samples, 256 * sizeof (sample_t));
1091 break;
1092
1093 case CONVERT (A52_3F, A52_STEREO):
1094 case CONVERT (A52_3F, A52_DOLBY):
1095 mix_3to2_SSE:
1096 mix3to2_SSE (samples, bias);
1097 break;
1098
1099 case CONVERT (A52_2F1R, A52_STEREO):
1100 if (slev == 0)
1101 break;
1102 mix21to2_SSE (samples, samples + 256, bias);
1103 break;
1104
1105 case CONVERT (A52_2F1R, A52_DOLBY):
1106 mix21toS_SSE (samples, bias);
1107 break;
1108
1109 case CONVERT (A52_3F1R, A52_STEREO):
1110 if (slev == 0)
1111 goto mix_3to2_SSE;
1112 mix31to2_SSE (samples, bias);
1113 break;
1114
1115 case CONVERT (A52_3F1R, A52_DOLBY):
1116 mix31toS_SSE (samples, bias);
1117 break;
1118
1119 case CONVERT (A52_2F2R, A52_STEREO):
1120 if (slev == 0)
1121 break;
1122 mix2to1_SSE (samples, samples + 512, bias);
1123 mix2to1_SSE (samples + 256, samples + 768, bias);
1124 break;
1125
1126 case CONVERT (A52_2F2R, A52_DOLBY):
1127 mix22toS_SSE (samples, bias);
1128 break;
1129
1130 case CONVERT (A52_3F2R, A52_STEREO):
1131 if (slev == 0)
1132 goto mix_3to2_SSE;
1133 mix32to2_SSE (samples, bias);
1134 break;
1135
1136 case CONVERT (A52_3F2R, A52_DOLBY):
1137 mix32toS_SSE (samples, bias);
1138 break;
1139
1140 case CONVERT (A52_3F1R, A52_3F):
1141 if (slev == 0)
1142 break;
1143 mix21to2_SSE (samples, samples + 512, bias);
1144 break;
1145
1146 case CONVERT (A52_3F2R, A52_3F):
1147 if (slev == 0)
1148 break;
1149 mix2to1_SSE (samples, samples + 768, bias);
1150 mix2to1_SSE (samples + 512, samples + 1024, bias);
1151 break;
1152
1153 case CONVERT (A52_3F1R, A52_2F1R):
1154 mix3to2_SSE (samples, bias);
1155 memcpy (samples + 512, samples + 768, 256 * sizeof (sample_t));
1156 break;
1157
1158 case CONVERT (A52_2F2R, A52_2F1R):
1159 mix2to1_SSE (samples + 512, samples + 768, bias);
1160 break;
1161
1162 case CONVERT (A52_3F2R, A52_2F1R):
1163 mix3to2_SSE (samples, bias); //FIXME possible bug? (output doesnt seem to be used)
1164 move2to1_SSE (samples + 768, samples + 512, bias);
1165 break;
1166
1167 case CONVERT (A52_3F2R, A52_3F1R):
1168 mix2to1_SSE (samples + 768, samples + 1024, bias);
1169 break;
1170
1171 case CONVERT (A52_2F1R, A52_2F2R):
1172 memcpy (samples + 768, samples + 512, 256 * sizeof (sample_t));
1173 break;
1174
1175 case CONVERT (A52_3F1R, A52_2F2R):
1176 mix3to2_SSE (samples, bias);
1177 memcpy (samples + 512, samples + 768, 256 * sizeof (sample_t));
1178 break;
1179
1180 case CONVERT (A52_3F2R, A52_2F2R):
1181 mix3to2_SSE (samples, bias);
1182 memcpy (samples + 512, samples + 768, 256 * sizeof (sample_t));
1183 memcpy (samples + 768, samples + 1024, 256 * sizeof (sample_t));
1184 break;
1185
1186 case CONVERT (A52_3F1R, A52_3F2R):
1187 memcpy (samples + 1024, samples + 768, 256 * sizeof (sample_t));
1188 break;
1189 }
1190 }
1191
1192 static void upmix_MMX (sample_t * samples, int acmod, int output)
1193 {
1194 switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
1195
1196 case CONVERT (A52_CHANNEL, A52_CHANNEL2):
1197 memcpy (samples + 256, samples, 256 * sizeof (sample_t));
1198 break;
1199
1200 case CONVERT (A52_3F2R, A52_MONO):
1201 zero_MMX (samples + 1024);
1202 case CONVERT (A52_3F1R, A52_MONO):
1203 case CONVERT (A52_2F2R, A52_MONO):
1204 zero_MMX (samples + 768);
1205 case CONVERT (A52_3F, A52_MONO):
1206 case CONVERT (A52_2F1R, A52_MONO):
1207 zero_MMX (samples + 512);
1208 case CONVERT (A52_CHANNEL, A52_MONO):
1209 case CONVERT (A52_STEREO, A52_MONO):
1210 zero_MMX (samples + 256);
1211 break;
1212
1213 case CONVERT (A52_3F2R, A52_STEREO):
1214 case CONVERT (A52_3F2R, A52_DOLBY):
1215 zero_MMX (samples + 1024);
1216 case CONVERT (A52_3F1R, A52_STEREO):
1217 case CONVERT (A52_3F1R, A52_DOLBY):
1218 zero_MMX (samples + 768);
1219 case CONVERT (A52_3F, A52_STEREO):
1220 case CONVERT (A52_3F, A52_DOLBY):
1221 mix_3to2_MMX:
1222 memcpy (samples + 512, samples + 256, 256 * sizeof (sample_t));
1223 zero_MMX (samples + 256);
1224 break;
1225
1226 case CONVERT (A52_2F2R, A52_STEREO):
1227 case CONVERT (A52_2F2R, A52_DOLBY):
1228 zero_MMX (samples + 768);
1229 case CONVERT (A52_2F1R, A52_STEREO):
1230 case CONVERT (A52_2F1R, A52_DOLBY):
1231 zero_MMX (samples + 512);
1232 break;
1233
1234 case CONVERT (A52_3F2R, A52_3F):
1235 zero_MMX (samples + 1024);
1236 case CONVERT (A52_3F1R, A52_3F):
1237 case CONVERT (A52_2F2R, A52_2F1R):
1238 zero_MMX (samples + 768);
1239 break;
1240
1241 case CONVERT (A52_3F2R, A52_3F1R):
1242 zero_MMX (samples + 1024);
1243 break;
1244
1245 case CONVERT (A52_3F2R, A52_2F1R):
1246 zero_MMX (samples + 1024);
1247 case CONVERT (A52_3F1R, A52_2F1R):
1248 mix_31to21_MMX:
1249 memcpy (samples + 768, samples + 512, 256 * sizeof (sample_t));
1250 goto mix_3to2_MMX;
1251
1252 case CONVERT (A52_3F2R, A52_2F2R):
1253 memcpy (samples + 1024, samples + 768, 256 * sizeof (sample_t));
1254 goto mix_31to21_MMX;
1255 }
1256 }
1257
1258 static void mix2to1_3dnow (sample_t * dest, sample_t * src, sample_t bias)
1259 {
1260 asm volatile(
1261 "movd %2, %%mm7 \n\t"
1262 "punpckldq %2, %%mm7 \n\t"
1263 "mov $-1024, %%"REG_S" \n\t"
1264 ASMALIGN16
1265 "1: \n\t"
1266 "movq (%0, %%"REG_S"), %%mm0 \n\t"
1267 "movq 8(%0, %%"REG_S"), %%mm1 \n\t"
1268 "movq 16(%0, %%"REG_S"), %%mm2 \n\t"
1269 "movq 24(%0, %%"REG_S"), %%mm3 \n\t"
1270 "pfadd (%1, %%"REG_S"), %%mm0 \n\t"
1271 "pfadd 8(%1, %%"REG_S"), %%mm1 \n\t"
1272 "pfadd 16(%1, %%"REG_S"), %%mm2 \n\t"
1273 "pfadd 24(%1, %%"REG_S"), %%mm3 \n\t"
1274 "pfadd %%mm7, %%mm0 \n\t"
1275 "pfadd %%mm7, %%mm1 \n\t"
1276 "pfadd %%mm7, %%mm2 \n\t"
1277 "pfadd %%mm7, %%mm3 \n\t"
1278 "movq %%mm0, (%1, %%"REG_S") \n\t"
1279 "movq %%mm1, 8(%1, %%"REG_S") \n\t"
1280 "movq %%mm2, 16(%1, %%"REG_S") \n\t"
1281 "movq %%mm3, 24(%1, %%"REG_S") \n\t"
1282 "add $32, %%"REG_S" \n\t"
1283 " jnz 1b \n\t"
1284 :: "r" (src+256), "r" (dest+256), "m" (bias)
1285 : "%"REG_S
1286 );
1287 }
1288
1289 static void mix3to1_3dnow (sample_t * samples, sample_t bias)
1290 {
1291 asm volatile(
1292 "movd %1, %%mm7 \n\t"
1293 "punpckldq %1, %%mm7 \n\t"
1294 "mov $-1024, %%"REG_S" \n\t"
1295 ASMALIGN16
1296 "1: \n\t"
1297 "movq (%0, %%"REG_S"), %%mm0 \n\t"
1298 "movq 8(%0, %%"REG_S"), %%mm1 \n\t"
1299 "movq 1024(%0, %%"REG_S"), %%mm2\n\t"
1300 "movq 1032(%0, %%"REG_S"), %%mm3\n\t"
1301 "pfadd 2048(%0, %%"REG_S"), %%mm0\n\t"
1302 "pfadd 2056(%0, %%"REG_S"), %%mm1\n\t"
1303 "pfadd %%mm7, %%mm0 \n\t"
1304 "pfadd %%mm7, %%mm1 \n\t"
1305 "pfadd %%mm2, %%mm0 \n\t"
1306 "pfadd %%mm3, %%mm1 \n\t"
1307 "movq %%mm0, (%0, %%"REG_S") \n\t"
1308 "movq %%mm1, 8(%0, %%"REG_S") \n\t"
1309 "add $16, %%"REG_S" \n\t"
1310 " jnz 1b \n\t"
1311 :: "r" (samples+256), "m" (bias)
1312 : "%"REG_S
1313 );
1314 }
1315
1316 static void mix4to1_3dnow (sample_t * samples, sample_t bias)
1317 {
1318 asm volatile(
1319 "movd %1, %%mm7 \n\t"
1320 "punpckldq %1, %%mm7 \n\t"
1321 "mov $-1024, %%"REG_S" \n\t"
1322 ASMALIGN16
1323 "1: \n\t"
1324 "movq (%0, %%"REG_S"), %%mm0 \n\t"
1325 "movq 8(%0, %%"REG_S"), %%mm1 \n\t"
1326 "movq 1024(%0, %%"REG_S"), %%mm2\n\t"
1327 "movq 1032(%0, %%"REG_S"), %%mm3\n\t"
1328 "pfadd 2048(%0, %%"REG_S"), %%mm0\n\t"
1329 "pfadd 2056(%0, %%"REG_S"), %%mm1\n\t"
1330 "pfadd 3072(%0, %%"REG_S"), %%mm2\n\t"
1331 "pfadd 3080(%0, %%"REG_S"), %%mm3\n\t"
1332 "pfadd %%mm7, %%mm0 \n\t"
1333 "pfadd %%mm7, %%mm1 \n\t"
1334 "pfadd %%mm2, %%mm0 \n\t"
1335 "pfadd %%mm3, %%mm1 \n\t"
1336 "movq %%mm0, (%0, %%"REG_S") \n\t"
1337 "movq %%mm1, 8(%0, %%"REG_S") \n\t"
1338 "add $16, %%"REG_S" \n\t"
1339 " jnz 1b \n\t"
1340 :: "r" (samples+256), "m" (bias)
1341 : "%"REG_S
1342 );
1343 }
1344
1345 static void mix5to1_3dnow (sample_t * samples, sample_t bias)
1346 {
1347 asm volatile(
1348 "movd %1, %%mm7 \n\t"
1349 "punpckldq %1, %%mm7 \n\t"
1350 "mov $-1024, %%"REG_S" \n\t"
1351 ASMALIGN16
1352 "1: \n\t"
1353 "movq (%0, %%"REG_S"), %%mm0 \n\t"
1354 "movq 8(%0, %%"REG_S"), %%mm1 \n\t"
1355 "movq 1024(%0, %%"REG_S"), %%mm2\n\t"
1356 "movq 1032(%0, %%"REG_S"), %%mm3\n\t"
1357 "pfadd 2048(%0, %%"REG_S"), %%mm0\n\t"
1358 "pfadd 2056(%0, %%"REG_S"), %%mm1\n\t"
1359 "pfadd 3072(%0, %%"REG_S"), %%mm2\n\t"
1360 "pfadd 3080(%0, %%"REG_S"), %%mm3\n\t"
1361 "pfadd %%mm7, %%mm0 \n\t"
1362 "pfadd %%mm7, %%mm1 \n\t"
1363 "pfadd 4096(%0, %%"REG_S"), %%mm2\n\t"
1364 "pfadd 4104(%0, %%"REG_S"), %%mm3\n\t"
1365 "pfadd %%mm2, %%mm0 \n\t"
1366 "pfadd %%mm3, %%mm1 \n\t"
1367 "movq %%mm0, (%0, %%"REG_S") \n\t"
1368 "movq %%mm1, 8(%0, %%"REG_S") \n\t"
1369 "add $16, %%"REG_S" \n\t"
1370 " jnz 1b \n\t"
1371 :: "r" (samples+256), "m" (bias)
1372 : "%"REG_S
1373 );
1374 }
1375
1376 static void mix3to2_3dnow (sample_t * samples, sample_t bias)
1377 {
1378 asm volatile(
1379 "movd %1, %%mm7 \n\t"
1380 "punpckldq %1, %%mm7 \n\t"
1381 "mov $-1024, %%"REG_S" \n\t"
1382 ASMALIGN16
1383 "1: \n\t"
1384 "movq 1024(%0, %%"REG_S"), %%mm0\n\t"
1385 "movq 1032(%0, %%"REG_S"), %%mm1\n\t"
1386 "pfadd %%mm7, %%mm0 \n\t" //common
1387 "pfadd %%mm7, %%mm1 \n\t" //common
1388 "movq (%0, %%"REG_S"), %%mm2 \n\t"
1389 "movq 8(%0, %%"REG_S"), %%mm3 \n\t"
1390 "movq 2048(%0, %%"REG_S"), %%mm4\n\t"
1391 "movq 2056(%0, %%"REG_S"), %%mm5\n\t"
1392 "pfadd %%mm0, %%mm2 \n\t"
1393 "pfadd %%mm1, %%mm3 \n\t"
1394 "pfadd %%mm0, %%mm4 \n\t"
1395 "pfadd %%mm1, %%mm5 \n\t"
1396 "movq %%mm2, (%0, %%"REG_S") \n\t"
1397 "movq %%mm3, 8(%0, %%"REG_S") \n\t"
1398 "movq %%mm4, 1024(%0, %%"REG_S")\n\t"
1399 "movq %%mm5, 1032(%0, %%"REG_S")\n\t"
1400 "add $16, %%"REG_S" \n\t"
1401 " jnz 1b \n\t"
1402 :: "r" (samples+256), "m" (bias)
1403 : "%"REG_S
1404 );
1405 }
1406
1407 static void mix21to2_3dnow (sample_t * left, sample_t * right, sample_t bias)
1408 {
1409 asm volatile(
1410 "movd %2, %%mm7 \n\t"
1411 "punpckldq %2, %%mm7 \n\t"
1412 "mov $-1024, %%"REG_S" \n\t"
1413 ASMALIGN16
1414 "1: \n\t"
1415 "movq 1024(%1, %%"REG_S"), %%mm0\n\t"
1416 "movq 1032(%1, %%"REG_S"), %%mm1\n\t"
1417 "pfadd %%mm7, %%mm0 \n\t" //common
1418 "pfadd %%mm7, %%mm1 \n\t" //common
1419 "movq (%0, %%"REG_S"), %%mm2 \n\t"
1420 "movq 8(%0, %%"REG_S"), %%mm3 \n\t"
1421 "movq (%1, %%"REG_S"), %%mm4 \n\t"
1422 "movq 8(%1, %%"REG_S"), %%mm5 \n\t"
1423 "pfadd %%mm0, %%mm2 \n\t"
1424 "pfadd %%mm1, %%mm3 \n\t"
1425 "pfadd %%mm0, %%mm4 \n\t"
1426 "pfadd %%mm1, %%mm5 \n\t"
1427 "movq %%mm2, (%0, %%"REG_S") \n\t"
1428 "movq %%mm3, 8(%0, %%"REG_S") \n\t"
1429 "movq %%mm4, (%1, %%"REG_S") \n\t"
1430 "movq %%mm5, 8(%1, %%"REG_S") \n\t"
1431 "add $16, %%"REG_S" \n\t"
1432 " jnz 1b \n\t"
1433 :: "r" (left+256), "r" (right+256), "m" (bias)
1434 : "%"REG_S
1435 );
1436 }
1437
1438 static void mix21toS_3dnow (sample_t * samples, sample_t bias)
1439 {
1440 asm volatile(
1441 "movd %1, %%mm7 \n\t"
1442 "punpckldq %1, %%mm7 \n\t"
1443 "mov $-1024, %%"REG_S" \n\t"
1444 ASMALIGN16
1445 "1: \n\t"
1446 "movq 2048(%0, %%"REG_S"), %%mm0\n\t" // surround
1447 "movq 2056(%0, %%"REG_S"), %%mm1\n\t" // surround
1448 "movq (%0, %%"REG_S"), %%mm2 \n\t"
1449 "movq 8(%0, %%"REG_S"), %%mm3 \n\t"
1450 "movq 1024(%0, %%"REG_S"), %%mm4\n\t"
1451 "movq 1032(%0, %%"REG_S"), %%mm5\n\t"
1452 "pfadd %%mm7, %%mm2 \n\t"
1453 "pfadd %%mm7, %%mm3 \n\t"
1454 "pfadd %%mm7, %%mm4 \n\t"
1455 "pfadd %%mm7, %%mm5 \n\t"
1456 "pfsub %%mm0, %%mm2 \n\t"
1457 "pfsub %%mm1, %%mm3 \n\t"
1458 "pfadd %%mm0, %%mm4 \n\t"
1459 "pfadd %%mm1, %%mm5 \n\t"
1460 "movq %%mm2, (%0, %%"REG_S") \n\t"
1461 "movq %%mm3, 8(%0, %%"REG_S") \n\t"
1462 "movq %%mm4, 1024(%0, %%"REG_S")\n\t"
1463 "movq %%mm5, 1032(%0, %%"REG_S")\n\t"
1464 "add $16, %%"REG_S" \n\t"
1465 " jnz 1b \n\t"
1466 :: "r" (samples+256), "m" (bias)
1467 : "%"REG_S
1468 );
1469 }
1470
1471 static void mix31to2_3dnow (sample_t * samples, sample_t bias)
1472 {
1473 asm volatile(
1474 "movd %1, %%mm7 \n\t"
1475 "punpckldq %1, %%mm7 \n\t"
1476 "mov $-1024, %%"REG_S" \n\t"
1477 ASMALIGN16
1478 "1: \n\t"
1479 "movq 1024(%0, %%"REG_S"), %%mm0\n\t"
1480 "movq 1032(%0, %%"REG_S"), %%mm1\n\t"
1481 "pfadd 3072(%0, %%"REG_S"), %%mm0\n\t"
1482 "pfadd 3080(%0, %%"REG_S"), %%mm1\n\t"
1483 "pfadd %%mm7, %%mm0 \n\t" // common
1484 "pfadd %%mm7, %%mm1 \n\t" // common
1485 "movq (%0, %%"REG_S"), %%mm2 \n\t"
1486 "movq 8(%0, %%"REG_S"), %%mm3 \n\t"
1487 "movq 2048(%0, %%"REG_S"), %%mm4\n\t"
1488 "movq 2056(%0, %%"REG_S"), %%mm5\n\t"
1489 "pfadd %%mm0, %%mm2 \n\t"
1490 "pfadd %%mm1, %%mm3 \n\t"
1491 "pfadd %%mm0, %%mm4 \n\t"
1492 "pfadd %%mm1, %%mm5 \n\t"
1493 "movq %%mm2, (%0, %%"REG_S") \n\t"
1494 "movq %%mm3, 8(%0, %%"REG_S") \n\t"
1495 "movq %%mm4, 1024(%0, %%"REG_S")\n\t"
1496 "movq %%mm5, 1032(%0, %%"REG_S")\n\t"
1497 "add $16, %%"REG_S" \n\t"
1498 " jnz 1b \n\t"
1499 :: "r" (samples+256), "m" (bias)
1500 : "%"REG_S
1501 );
1502 }
1503
1504 static void mix31toS_3dnow (sample_t * samples, sample_t bias)
1505 {
1506 asm volatile(
1507 "movd %1, %%mm7 \n\t"
1508 "punpckldq %1, %%mm7 \n\t"
1509 "mov $-1024, %%"REG_S" \n\t"
1510 ASMALIGN16
1511 "1: \n\t"
1512 "movq 1024(%0, %%"REG_S"), %%mm0\n\t"
1513 "movq 1032(%0, %%"REG_S"), %%mm1\n\t"
1514 "pfadd %%mm7, %%mm0 \n\t" // common
1515 "pfadd %%mm7, %%mm1 \n\t" // common
1516 "movq (%0, %%"REG_S"), %%mm2 \n\t"
1517 "movq 8(%0, %%"REG_S"), %%mm3 \n\t"
1518 "movq 2048(%0, %%"REG_S"), %%mm4\n\t"
1519 "movq 2056(%0, %%"REG_S"), %%mm5\n\t"
1520 "pfadd %%mm0, %%mm2 \n\t"
1521 "pfadd %%mm1, %%mm3 \n\t"
1522 "pfadd %%mm0, %%mm4 \n\t"
1523 "pfadd %%mm1, %%mm5 \n\t"
1524 "movq 3072(%0, %%"REG_S"), %%mm0\n\t" // surround
1525 "movq 3080(%0, %%"REG_S"), %%mm1\n\t" // surround
1526 "pfsub %%mm0, %%mm2 \n\t"
1527 "pfsub %%mm1, %%mm3 \n\t"
1528 "pfadd %%mm0, %%mm4 \n\t"
1529 "pfadd %%mm1, %%mm5 \n\t"
1530 "movq %%mm2, (%0, %%"REG_S") \n\t"
1531 "movq %%mm3, 8(%0, %%"REG_S") \n\t"
1532 "movq %%mm4, 1024(%0, %%"REG_S")\n\t"
1533 "movq %%mm5, 1032(%0, %%"REG_S")\n\t"
1534 "add $16, %%"REG_S" \n\t"
1535 " jnz 1b \n\t"
1536 :: "r" (samples+256), "m" (bias)
1537 : "%"REG_S
1538 );
1539 }
1540
1541 static void mix22toS_3dnow (sample_t * samples, sample_t bias)
1542 {
1543 asm volatile(
1544 "movd %1, %%mm7 \n\t"
1545 "punpckldq %1, %%mm7 \n\t"
1546 "mov $-1024, %%"REG_S" \n\t"
1547 ASMALIGN16
1548 "1: \n\t"
1549 "movq 2048(%0, %%"REG_S"), %%mm0\n\t"
1550 "movq 2056(%0, %%"REG_S"), %%mm1\n\t"
1551 "pfadd 3072(%0, %%"REG_S"), %%mm0\n\t" // surround
1552 "pfadd 3080(%0, %%"REG_S"), %%mm1\n\t" // surround
1553 "movq (%0, %%"REG_S"), %%mm2 \n\t"
1554 "movq 8(%0, %%"REG_S"), %%mm3 \n\t"
1555 "movq 1024(%0, %%"REG_S"), %%mm4\n\t"
1556 "movq 1032(%0, %%"REG_S"), %%mm5\n\t"
1557 "pfadd %%mm7, %%mm2 \n\t"
1558 "pfadd %%mm7, %%mm3 \n\t"
1559 "pfadd %%mm7, %%mm4 \n\t"
1560 "pfadd %%mm7, %%mm5 \n\t"
1561 "pfsub %%mm0, %%mm2 \n\t"
1562 "pfsub %%mm1, %%mm3 \n\t"
1563 "pfadd %%mm0, %%mm4 \n\t"
1564 "pfadd %%mm1, %%mm5 \n\t"
1565 "movq %%mm2, (%0, %%"REG_S") \n\t"
1566 "movq %%mm3, 8(%0, %%"REG_S") \n\t"
1567 "movq %%mm4, 1024(%0, %%"REG_S")\n\t"
1568 "movq %%mm5, 1032(%0, %%"REG_S")\n\t"
1569 "add $16, %%"REG_S" \n\t"
1570 " jnz 1b \n\t"
1571 :: "r" (samples+256), "m" (bias)
1572 : "%"REG_S
1573 );
1574 }
1575
1576 static void mix32to2_3dnow (sample_t * samples, sample_t bias)
1577 {
1578 asm volatile(
1579 "movd %1, %%mm7 \n\t"
1580 "punpckldq %1, %%mm7 \n\t"
1581 "mov $-1024, %%"REG_S" \n\t"
1582 ASMALIGN16
1583 "1: \n\t"
1584 "movq 1024(%0, %%"REG_S"), %%mm0\n\t"
1585 "movq 1032(%0, %%"REG_S"), %%mm1\n\t"
1586 "pfadd %%mm7, %%mm0 \n\t" // common
1587 "pfadd %%mm7, %%mm1 \n\t" // common
1588 "movq %%mm0, %%mm2 \n\t" // common
1589 "movq %%mm1, %%mm3 \n\t" // common
1590 "pfadd (%0, %%"REG_S"), %%mm0 \n\t"
1591 "pfadd 8(%0, %%"REG_S"), %%mm1 \n\t"
1592 "pfadd 2048(%0, %%"REG_S"), %%mm2\n\t"
1593 "pfadd 2056(%0, %%"REG_S"), %%mm3\n\t"
1594 "pfadd 3072(%0, %%"REG_S"), %%mm0\n\t"
1595 "pfadd 3080(%0, %%"REG_S"), %%mm1\n\t"
1596 "pfadd 4096(%0, %%"REG_S"), %%mm2\n\t"
1597 "pfadd 4104(%0, %%"REG_S"), %%mm3\n\t"
1598 "movq %%mm0, (%0, %%"REG_S") \n\t"
1599 "movq %%mm1, 8(%0, %%"REG_S") \n\t"
1600 "movq %%mm2, 1024(%0, %%"REG_S")\n\t"
1601 "movq %%mm3, 1032(%0, %%"REG_S")\n\t"
1602 "add $16, %%"REG_S" \n\t"
1603 " jnz 1b \n\t"
1604 :: "r" (samples+256), "m" (bias)
1605 : "%"REG_S
1606 );
1607 }
1608
1609 /* todo: should be optimized better */
1610 static void mix32toS_3dnow (sample_t * samples, sample_t bias)
1611 {
1612 asm volatile(
1613 "mov $-1024, %%"REG_S" \n\t"
1614 ASMALIGN16
1615 "1: \n\t"
1616 "movd %1, %%mm7 \n\t"
1617 "punpckldq %1, %%mm7 \n\t"
1618 "movq 1024(%0, %%"REG_S"), %%mm0\n\t"
1619 "movq 1032(%0, %%"REG_S"), %%mm1\n\t"
1620 "movq 3072(%0, %%"REG_S"), %%mm4\n\t"
1621 "movq 3080(%0, %%"REG_S"), %%mm5\n\t"
1622 "pfadd %%mm7, %%mm0 \n\t" // common
1623 "pfadd %%mm7, %%mm1 \n\t" // common
1624 "pfadd 4096(%0, %%"REG_S"), %%mm4\n\t" // surround
1625 "pfadd 4104(%0, %%"REG_S"), %%mm5\n\t" // surround
1626 "movq (%0, %%"REG_S"), %%mm2 \n\t"
1627 "movq 8(%0, %%"REG_S"), %%mm3 \n\t"
1628 "movq 2048(%0, %%"REG_S"), %%mm6\n\t"
1629 "movq 2056(%0, %%"REG_S"), %%mm7\n\t"
1630 "pfsub %%mm4, %%mm2 \n\t"
1631 "pfsub %%mm5, %%mm3 \n\t"
1632 "pfadd %%mm4, %%mm6 \n\t"
1633 "pfadd %%mm5, %%mm7 \n\t"
1634 "pfadd %%mm0, %%mm2 \n\t"
1635 "pfadd %%mm1, %%mm3 \n\t"
1636 "pfadd %%mm0, %%mm6 \n\t"
1637 "pfadd %%mm1, %%mm7 \n\t"
1638 "movq %%mm2, (%0, %%"REG_S") \n\t"
1639 "movq %%mm3, 8(%0, %%"REG_S") \n\t"
1640 "movq %%mm6, 1024(%0, %%"REG_S")\n\t"
1641 "movq %%mm7, 1032(%0, %%"REG_S")\n\t"
1642 "add $16, %%"REG_S" \n\t"
1643 " jnz 1b \n\t"
1644 :: "r" (samples+256), "m" (bias)
1645 : "%"REG_S
1646 );
1647 }
1648
1649 static void move2to1_3dnow (sample_t * src, sample_t * dest, sample_t bias)
1650 {
1651 asm volatile(
1652 "movd %2, %%mm7 \n\t"
1653 "punpckldq %2, %%mm7 \n\t"
1654 "mov $-1024, %%"REG_S" \n\t"
1655 ASMALIGN16
1656 "1: \n\t"
1657 "movq (%0, %%"REG_S"), %%mm0 \n\t"
1658 "movq 8(%0, %%"REG_S"), %%mm1 \n\t"
1659 "movq 16(%0, %%"REG_S"), %%mm2 \n\t"
1660 "movq 24(%0, %%"REG_S"), %%mm3 \n\t"
1661 "pfadd 1024(%0, %%"REG_S"), %%mm0\n\t"
1662 "pfadd 1032(%0, %%"REG_S"), %%mm1\n\t"
1663 "pfadd 1040(%0, %%"REG_S"), %%mm2\n\t"
1664 "pfadd 1048(%0, %%"REG_S"), %%mm3\n\t"
1665 "pfadd %%mm7, %%mm0 \n\t"
1666 "pfadd %%mm7, %%mm1 \n\t"
1667 "pfadd %%mm7, %%mm2 \n\t"
1668 "pfadd %%mm7, %%mm3 \n\t"
1669 "movq %%mm0, (%1, %%"REG_S") \n\t"
1670 "movq %%mm1, 8(%1, %%"REG_S") \n\t"
1671 "movq %%mm2, 16(%1, %%"REG_S") \n\t"
1672 "movq %%mm3, 24(%1, %%"REG_S") \n\t"
1673 "add $32, %%"REG_S" \n\t"
1674 " jnz 1b \n\t"
1675 :: "r" (src+256), "r" (dest+256), "m" (bias)
1676 : "%"REG_S
1677 );
1678 }
1679
1680 static void downmix_3dnow (sample_t * samples, int acmod, int output, sample_t bias,
1681 sample_t clev, sample_t slev)
1682 {
1683 switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
1684
1685 case CONVERT (A52_CHANNEL, A52_CHANNEL2):
1686 memcpy (samples, samples + 256, 256 * sizeof (sample_t));
1687 break;
1688
1689 case CONVERT (A52_CHANNEL, A52_MONO):
1690 case CONVERT (A52_STEREO, A52_MONO):
1691 mix_2to1_3dnow:
1692 mix2to1_3dnow (samples, samples + 256, bias);
1693 break;
1694
1695 case CONVERT (A52_2F1R, A52_MONO):
1696 if (slev == 0)
1697 goto mix_2to1_3dnow;
1698 case CONVERT (A52_3F, A52_MONO):
1699 mix_3to1_3dnow:
1700 mix3to1_3dnow (samples, bias);
1701 break;
1702
1703 case CONVERT (A52_3F1R, A52_MONO):
1704 if (slev == 0)
1705 goto mix_3to1_3dnow;
1706 case CONVERT (A52_2F2R, A52_MONO):
1707 if (slev == 0)
1708 goto mix_2to1_3dnow;
1709 mix4to1_3dnow (samples, bias);
1710 break;
1711
1712 case CONVERT (A52_3F2R, A52_MONO):
1713 if (slev == 0)
1714 goto mix_3to1_3dnow;
1715 mix5to1_3dnow (samples, bias);
1716 break;
1717
1718 case CONVERT (A52_MONO, A52_DOLBY):
1719 memcpy (samples + 256, samples, 256 * sizeof (sample_t));
1720 break;
1721
1722 case CONVERT (A52_3F, A52_STEREO):
1723 case CONVERT (A52_3F, A52_DOLBY):
1724 mix_3to2_3dnow:
1725 mix3to2_3dnow (samples, bias);
1726 break;
1727
1728 case CONVERT (A52_2F1R, A52_STEREO):
1729 if (slev == 0)
1730 break;
1731 mix21to2_3dnow (samples, samples + 256, bias);
1732 break;
1733
1734 case CONVERT (A52_2F1R, A52_DOLBY):
1735 mix21toS_3dnow (samples, bias);
1736 break;
1737
1738 case CONVERT (A52_3F1R, A52_STEREO):
1739 if (slev == 0)
1740 goto mix_3to2_3dnow;
1741 mix31to2_3dnow (samples, bias);
1742 break;
1743
1744 case CONVERT (A52_3F1R, A52_DOLBY):
1745 mix31toS_3dnow (samples, bias);
1746 break;
1747
1748 case CONVERT (A52_2F2R, A52_STEREO):
1749 if (slev == 0)
1750 break;
1751 mix2to1_3dnow (samples, samples + 512, bias);
1752 mix2to1_3dnow (samples + 256, samples + 768, bias);
1753 break;
1754
1755 case CONVERT (A52_2F2R, A52_DOLBY):
1756 mix22toS_3dnow (samples, bias);
1757 break;
1758
1759 case CONVERT (A52_3F2R, A52_STEREO):
1760 if (slev == 0)
1761 goto mix_3to2_3dnow;
1762 mix32to2_3dnow (samples, bias);
1763 break;
1764
1765 case CONVERT (A52_3F2R, A52_DOLBY):
1766 mix32toS_3dnow (samples, bias);
1767 break;
1768
1769 case CONVERT (A52_3F1R, A52_3F):
1770 if (slev == 0)
1771 break;
1772 mix21to2_3dnow (samples, samples + 512, bias);
1773 break;
1774
1775 case CONVERT (A52_3F2R, A52_3F):
1776 if (slev == 0)
1777 break;
1778 mix2to1_3dnow (samples, samples + 768, bias);
1779 mix2to1_3dnow (samples + 512, samples + 1024, bias);
1780 break;
1781
1782 case CONVERT (A52_3F1R, A52_2F1R):
1783 mix3to2_3dnow (samples, bias);
1784 memcpy (samples + 512, samples + 768, 256 * sizeof (sample_t));
1785 break;
1786
1787 case CONVERT (A52_2F2R, A52_2F1R):
1788 mix2to1_3dnow (samples + 512, samples + 768, bias);
1789 break;
1790
1791 case CONVERT (A52_3F2R, A52_2F1R):
1792 mix3to2_3dnow (samples, bias); //FIXME possible bug? (output doesnt seem to be used)
1793 move2to1_3dnow (samples + 768, samples + 512, bias);
1794 break;
1795
1796 case CONVERT (A52_3F2R, A52_3F1R):
1797 mix2to1_3dnow (samples + 768, samples + 1024, bias);
1798 break;
1799
1800 case CONVERT (A52_2F1R, A52_2F2R):
1801 memcpy (samples + 768, samples + 512, 256 * sizeof (sample_t));
1802 break;
1803
1804 case CONVERT (A52_3F1R, A52_2F2R):
1805 mix3to2_3dnow (samples, bias);
1806 memcpy (samples + 512, samples + 768, 256 * sizeof (sample_t));
1807 break;
1808
1809 case CONVERT (A52_3F2R, A52_2F2R):
1810 mix3to2_3dnow (samples, bias);
1811 memcpy (samples + 512, samples + 768, 256 * sizeof (sample_t));
1812 memcpy (samples + 768, samples + 1024, 256 * sizeof (sample_t));
1813 break;
1814
1815 case CONVERT (A52_3F1R, A52_3F2R):
1816 memcpy (samples + 1024, samples + 768, 256 * sizeof (sample_t));
1817 break;
1818 }
1819 __asm __volatile("femms":::"memory");
1820 }
1821
1822 #endif // ARCH_X86 || ARCH_X86_64