Mercurial > libavcodec.hg
comparison mpegaudiodec.c @ 12026:3f3d08bb5cf8 libavcodec
More mp{1,2,3} 32-point DCT transform to our common DCT framework.
Should allow for future SIMD optimizations.
author | vitor |
---|---|
date | Wed, 30 Jun 2010 20:11:27 +0000 |
parents | 11a8d4c1ee81 |
children | b4888704c11e |
comparison
equal
deleted
inserted
replaced
12025:2d730a4acc77 | 12026:3f3d08bb5cf8 |
---|---|
64 | 64 |
65 #define HEADER_SIZE 4 | 65 #define HEADER_SIZE 4 |
66 | 66 |
67 #include "mpegaudiodata.h" | 67 #include "mpegaudiodata.h" |
68 #include "mpegaudiodectab.h" | 68 #include "mpegaudiodectab.h" |
69 | |
70 #if CONFIG_FLOAT | |
71 # include "fft.h" | |
72 #else | |
73 # include "dct32.c" | |
74 #endif | |
69 | 75 |
70 static void compute_antialias_integer(MPADecodeContext *s, GranuleDef *g); | 76 static void compute_antialias_integer(MPADecodeContext *s, GranuleDef *g); |
71 static void compute_antialias_float(MPADecodeContext *s, GranuleDef *g); | 77 static void compute_antialias_float(MPADecodeContext *s, GranuleDef *g); |
72 static void apply_window_mp3_c(MPA_INT *synth_buf, MPA_INT *window, | 78 static void apply_window_mp3_c(MPA_INT *synth_buf, MPA_INT *window, |
73 int *dither_state, OUT_INT *samples, int incr); | 79 int *dither_state, OUT_INT *samples, int incr); |
347 scale_factor_mult[i][0], | 353 scale_factor_mult[i][0], |
348 scale_factor_mult[i][1], | 354 scale_factor_mult[i][1], |
349 scale_factor_mult[i][2]); | 355 scale_factor_mult[i][2]); |
350 } | 356 } |
351 | 357 |
358 #if CONFIG_FLOAT | |
359 ff_dct_init(&s->dct, 5, DCT_II); | |
360 #endif | |
352 RENAME(ff_mpa_synth_init)(RENAME(ff_mpa_synth_window)); | 361 RENAME(ff_mpa_synth_init)(RENAME(ff_mpa_synth_window)); |
353 | 362 |
354 /* huffman decode tables */ | 363 /* huffman decode tables */ |
355 offset = 0; | 364 offset = 0; |
356 for(i=1;i<16;i++) { | 365 for(i=1;i<16;i++) { |
510 if (avctx->codec_id == CODEC_ID_MP3ADU) | 519 if (avctx->codec_id == CODEC_ID_MP3ADU) |
511 s->adu_mode = 1; | 520 s->adu_mode = 1; |
512 return 0; | 521 return 0; |
513 } | 522 } |
514 | 523 |
515 /* tab[i][j] = 1.0 / (2.0 * cos(pi*(2*k+1) / 2^(6 - j))) */ | |
516 | |
517 /* cos(i*pi/64) */ | |
518 | |
519 #define COS0_0 FIXHR(0.50060299823519630134/2) | |
520 #define COS0_1 FIXHR(0.50547095989754365998/2) | |
521 #define COS0_2 FIXHR(0.51544730992262454697/2) | |
522 #define COS0_3 FIXHR(0.53104259108978417447/2) | |
523 #define COS0_4 FIXHR(0.55310389603444452782/2) | |
524 #define COS0_5 FIXHR(0.58293496820613387367/2) | |
525 #define COS0_6 FIXHR(0.62250412303566481615/2) | |
526 #define COS0_7 FIXHR(0.67480834145500574602/2) | |
527 #define COS0_8 FIXHR(0.74453627100229844977/2) | |
528 #define COS0_9 FIXHR(0.83934964541552703873/2) | |
529 #define COS0_10 FIXHR(0.97256823786196069369/2) | |
530 #define COS0_11 FIXHR(1.16943993343288495515/4) | |
531 #define COS0_12 FIXHR(1.48416461631416627724/4) | |
532 #define COS0_13 FIXHR(2.05778100995341155085/8) | |
533 #define COS0_14 FIXHR(3.40760841846871878570/8) | |
534 #define COS0_15 FIXHR(10.19000812354805681150/32) | |
535 | |
536 #define COS1_0 FIXHR(0.50241928618815570551/2) | |
537 #define COS1_1 FIXHR(0.52249861493968888062/2) | |
538 #define COS1_2 FIXHR(0.56694403481635770368/2) | |
539 #define COS1_3 FIXHR(0.64682178335999012954/2) | |
540 #define COS1_4 FIXHR(0.78815462345125022473/2) | |
541 #define COS1_5 FIXHR(1.06067768599034747134/4) | |
542 #define COS1_6 FIXHR(1.72244709823833392782/4) | |
543 #define COS1_7 FIXHR(5.10114861868916385802/16) | |
544 | |
545 #define COS2_0 FIXHR(0.50979557910415916894/2) | |
546 #define COS2_1 FIXHR(0.60134488693504528054/2) | |
547 #define COS2_2 FIXHR(0.89997622313641570463/2) | |
548 #define COS2_3 FIXHR(2.56291544774150617881/8) | |
549 | |
550 #define COS3_0 FIXHR(0.54119610014619698439/2) | |
551 #define COS3_1 FIXHR(1.30656296487637652785/4) | |
552 | |
553 #define COS4_0 FIXHR(0.70710678118654752439/2) | |
554 | |
555 /* butterfly operator */ | |
556 #define BF(a, b, c, s)\ | |
557 {\ | |
558 tmp0 = val##a + val##b;\ | |
559 tmp1 = val##a - val##b;\ | |
560 val##a = tmp0;\ | |
561 val##b = MULH3(tmp1, c, 1<<(s));\ | |
562 } | |
563 | |
564 #define BF0(a, b, c, s)\ | |
565 {\ | |
566 tmp0 = tab[a] + tab[b];\ | |
567 tmp1 = tab[a] - tab[b];\ | |
568 val##a = tmp0;\ | |
569 val##b = MULH3(tmp1, c, 1<<(s));\ | |
570 } | |
571 | |
572 #define BF1(a, b, c, d)\ | |
573 {\ | |
574 BF(a, b, COS4_0, 1);\ | |
575 BF(c, d,-COS4_0, 1);\ | |
576 val##c += val##d;\ | |
577 } | |
578 | |
579 #define BF2(a, b, c, d)\ | |
580 {\ | |
581 BF(a, b, COS4_0, 1);\ | |
582 BF(c, d,-COS4_0, 1);\ | |
583 val##c += val##d;\ | |
584 val##a += val##c;\ | |
585 val##c += val##b;\ | |
586 val##b += val##d;\ | |
587 } | |
588 | |
589 #define ADD(a, b) val##a += val##b | |
590 | |
591 /* DCT32 without 1/sqrt(2) coef zero scaling. */ | |
592 static void dct32(INTFLOAT *out, const INTFLOAT *tab) | |
593 { | |
594 INTFLOAT tmp0, tmp1; | |
595 | |
596 INTFLOAT val0 , val1 , val2 , val3 , val4 , val5 , val6 , val7 , | |
597 val8 , val9 , val10, val11, val12, val13, val14, val15, | |
598 val16, val17, val18, val19, val20, val21, val22, val23, | |
599 val24, val25, val26, val27, val28, val29, val30, val31; | |
600 | |
601 /* pass 1 */ | |
602 BF0( 0, 31, COS0_0 , 1); | |
603 BF0(15, 16, COS0_15, 5); | |
604 /* pass 2 */ | |
605 BF( 0, 15, COS1_0 , 1); | |
606 BF(16, 31,-COS1_0 , 1); | |
607 /* pass 1 */ | |
608 BF0( 7, 24, COS0_7 , 1); | |
609 BF0( 8, 23, COS0_8 , 1); | |
610 /* pass 2 */ | |
611 BF( 7, 8, COS1_7 , 4); | |
612 BF(23, 24,-COS1_7 , 4); | |
613 /* pass 3 */ | |
614 BF( 0, 7, COS2_0 , 1); | |
615 BF( 8, 15,-COS2_0 , 1); | |
616 BF(16, 23, COS2_0 , 1); | |
617 BF(24, 31,-COS2_0 , 1); | |
618 /* pass 1 */ | |
619 BF0( 3, 28, COS0_3 , 1); | |
620 BF0(12, 19, COS0_12, 2); | |
621 /* pass 2 */ | |
622 BF( 3, 12, COS1_3 , 1); | |
623 BF(19, 28,-COS1_3 , 1); | |
624 /* pass 1 */ | |
625 BF0( 4, 27, COS0_4 , 1); | |
626 BF0(11, 20, COS0_11, 2); | |
627 /* pass 2 */ | |
628 BF( 4, 11, COS1_4 , 1); | |
629 BF(20, 27,-COS1_4 , 1); | |
630 /* pass 3 */ | |
631 BF( 3, 4, COS2_3 , 3); | |
632 BF(11, 12,-COS2_3 , 3); | |
633 BF(19, 20, COS2_3 , 3); | |
634 BF(27, 28,-COS2_3 , 3); | |
635 /* pass 4 */ | |
636 BF( 0, 3, COS3_0 , 1); | |
637 BF( 4, 7,-COS3_0 , 1); | |
638 BF( 8, 11, COS3_0 , 1); | |
639 BF(12, 15,-COS3_0 , 1); | |
640 BF(16, 19, COS3_0 , 1); | |
641 BF(20, 23,-COS3_0 , 1); | |
642 BF(24, 27, COS3_0 , 1); | |
643 BF(28, 31,-COS3_0 , 1); | |
644 | |
645 | |
646 | |
647 /* pass 1 */ | |
648 BF0( 1, 30, COS0_1 , 1); | |
649 BF0(14, 17, COS0_14, 3); | |
650 /* pass 2 */ | |
651 BF( 1, 14, COS1_1 , 1); | |
652 BF(17, 30,-COS1_1 , 1); | |
653 /* pass 1 */ | |
654 BF0( 6, 25, COS0_6 , 1); | |
655 BF0( 9, 22, COS0_9 , 1); | |
656 /* pass 2 */ | |
657 BF( 6, 9, COS1_6 , 2); | |
658 BF(22, 25,-COS1_6 , 2); | |
659 /* pass 3 */ | |
660 BF( 1, 6, COS2_1 , 1); | |
661 BF( 9, 14,-COS2_1 , 1); | |
662 BF(17, 22, COS2_1 , 1); | |
663 BF(25, 30,-COS2_1 , 1); | |
664 | |
665 /* pass 1 */ | |
666 BF0( 2, 29, COS0_2 , 1); | |
667 BF0(13, 18, COS0_13, 3); | |
668 /* pass 2 */ | |
669 BF( 2, 13, COS1_2 , 1); | |
670 BF(18, 29,-COS1_2 , 1); | |
671 /* pass 1 */ | |
672 BF0( 5, 26, COS0_5 , 1); | |
673 BF0(10, 21, COS0_10, 1); | |
674 /* pass 2 */ | |
675 BF( 5, 10, COS1_5 , 2); | |
676 BF(21, 26,-COS1_5 , 2); | |
677 /* pass 3 */ | |
678 BF( 2, 5, COS2_2 , 1); | |
679 BF(10, 13,-COS2_2 , 1); | |
680 BF(18, 21, COS2_2 , 1); | |
681 BF(26, 29,-COS2_2 , 1); | |
682 /* pass 4 */ | |
683 BF( 1, 2, COS3_1 , 2); | |
684 BF( 5, 6,-COS3_1 , 2); | |
685 BF( 9, 10, COS3_1 , 2); | |
686 BF(13, 14,-COS3_1 , 2); | |
687 BF(17, 18, COS3_1 , 2); | |
688 BF(21, 22,-COS3_1 , 2); | |
689 BF(25, 26, COS3_1 , 2); | |
690 BF(29, 30,-COS3_1 , 2); | |
691 | |
692 /* pass 5 */ | |
693 BF1( 0, 1, 2, 3); | |
694 BF2( 4, 5, 6, 7); | |
695 BF1( 8, 9, 10, 11); | |
696 BF2(12, 13, 14, 15); | |
697 BF1(16, 17, 18, 19); | |
698 BF2(20, 21, 22, 23); | |
699 BF1(24, 25, 26, 27); | |
700 BF2(28, 29, 30, 31); | |
701 | |
702 /* pass 6 */ | |
703 | |
704 ADD( 8, 12); | |
705 ADD(12, 10); | |
706 ADD(10, 14); | |
707 ADD(14, 9); | |
708 ADD( 9, 13); | |
709 ADD(13, 11); | |
710 ADD(11, 15); | |
711 | |
712 out[ 0] = val0; | |
713 out[16] = val1; | |
714 out[ 8] = val2; | |
715 out[24] = val3; | |
716 out[ 4] = val4; | |
717 out[20] = val5; | |
718 out[12] = val6; | |
719 out[28] = val7; | |
720 out[ 2] = val8; | |
721 out[18] = val9; | |
722 out[10] = val10; | |
723 out[26] = val11; | |
724 out[ 6] = val12; | |
725 out[22] = val13; | |
726 out[14] = val14; | |
727 out[30] = val15; | |
728 | |
729 ADD(24, 28); | |
730 ADD(28, 26); | |
731 ADD(26, 30); | |
732 ADD(30, 25); | |
733 ADD(25, 29); | |
734 ADD(29, 27); | |
735 ADD(27, 31); | |
736 | |
737 out[ 1] = val16 + val24; | |
738 out[17] = val17 + val25; | |
739 out[ 9] = val18 + val26; | |
740 out[25] = val19 + val27; | |
741 out[ 5] = val20 + val28; | |
742 out[21] = val21 + val29; | |
743 out[13] = val22 + val30; | |
744 out[29] = val23 + val31; | |
745 out[ 3] = val24 + val20; | |
746 out[19] = val25 + val21; | |
747 out[11] = val26 + val22; | |
748 out[27] = val27 + val23; | |
749 out[ 7] = val28 + val18; | |
750 out[23] = val29 + val19; | |
751 out[15] = val30 + val17; | |
752 out[31] = val31; | |
753 } | |
754 | 524 |
755 #if CONFIG_FLOAT | 525 #if CONFIG_FLOAT |
756 static inline float round_sample(float *sum) | 526 static inline float round_sample(float *sum) |
757 { | 527 { |
758 float sum1=*sum; | 528 float sum1=*sum; |