summaryrefslogtreecommitdiff
path: root/ac3dec/coeff.c
diff options
context:
space:
mode:
Diffstat (limited to 'ac3dec/coeff.c')
-rw-r--r--ac3dec/coeff.c431
1 files changed, 0 insertions, 431 deletions
diff --git a/ac3dec/coeff.c b/ac3dec/coeff.c
deleted file mode 100644
index 7d5579c..0000000
--- a/ac3dec/coeff.c
+++ /dev/null
@@ -1,431 +0,0 @@
-/*
- * coeff.c
- *
- * Copyright (C) Aaron Holtzman - May 1999
- *
- * This file is part of ac3dec, a free Dolby AC-3 stream decoder.
- *
- * ac3dec 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, or (at your option)
- * any later version.
- *
- * ac3dec 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 GNU Make; see the file COPYING. If not, write to
- * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
- *
- */
-
-
-#include <stdlib.h>
-#include <stdio.h>
-#include "ac3.h"
-#include "ac3_internal.h"
-
-
-#include "bitstream.h"
-#include "dither.h"
-#include "coeff.h"
-
-
-//
-//Lookup tables of 0.15 two's complement quantization values
-//
-#define Q0 ((-2 << 15) / 3.0)
-#define Q1 (0)
-#define Q2 ((2 << 15) / 3.0)
-
-static const float q_1_0[ 32 ] =
-{
- Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,
- Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,
- Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,
- 0,0,0,0,0
-};
-
-static const float q_1_1[ 32 ] =
-{
- Q0,Q0,Q0,Q1,Q1,Q1,Q2,Q2,Q2,
- Q0,Q0,Q0,Q1,Q1,Q1,Q2,Q2,Q2,
- Q0,Q0,Q0,Q1,Q1,Q1,Q2,Q2,Q2,
- 0,0,0,0,0
-};
-
-static const float q_1_2[ 32 ] =
-{
- Q0,Q1,Q2,Q0,Q1,Q2,Q0,Q1,Q2,
- Q0,Q1,Q2,Q0,Q1,Q2,Q0,Q1,Q2,
- Q0,Q1,Q2,Q0,Q1,Q2,Q0,Q1,Q2,
- 0,0,0,0,0
-};
-
-#undef Q0
-#undef Q1
-#undef Q2
-
-#define Q0 ((-4 << 15) / 5.0)
-#define Q1 ((-2 << 15) / 5.0)
-#define Q2 (0)
-#define Q3 ((2 << 15) / 5.0)
-#define Q4 ((4 << 15) / 5.0)
-
-static const float q_2_0[ 128 ] =
-{
- Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,
- Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,Q0,
- Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,
- Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,Q1,
- Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,
- Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,Q2,
- Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,
- Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,Q3,
- Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,
- Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,Q4,
- 0,0,0
-};
-
-static const float q_2_1[ 128 ] =
-{
- Q0,Q0,Q0,Q0,Q0,Q1,Q1,Q1,Q1,Q1,
- Q2,Q2,Q2,Q2,Q2,Q3,Q3,Q3,Q3,Q3,
- Q4,Q4,Q4,Q4,Q4,Q0,Q0,Q0,Q0,Q0,
- Q1,Q1,Q1,Q1,Q1,Q2,Q2,Q2,Q2,Q2,
- Q3,Q3,Q3,Q3,Q3,Q4,Q4,Q4,Q4,Q4,
- Q0,Q0,Q0,Q0,Q0,Q1,Q1,Q1,Q1,Q1,
- Q2,Q2,Q2,Q2,Q2,Q3,Q3,Q3,Q3,Q3,
- Q4,Q4,Q4,Q4,Q4,Q0,Q0,Q0,Q0,Q0,
- Q1,Q1,Q1,Q1,Q1,Q2,Q2,Q2,Q2,Q2,
- Q3,Q3,Q3,Q3,Q3,Q4,Q4,Q4,Q4,Q4,
- Q0,Q0,Q0,Q0,Q0,Q1,Q1,Q1,Q1,Q1,
- Q2,Q2,Q2,Q2,Q2,Q3,Q3,Q3,Q3,Q3,
- Q4,Q4,Q4,Q4,Q4,0,0,0
- };
-
-static const float q_2_2[ 128 ] =
- {
- Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
- Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
- Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
- Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
- Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
- Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
- Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
- Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
- Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
- Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
- Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
- Q0,Q1,Q2,Q3,Q4,Q0,Q1,Q2,Q3,Q4,
- Q0,Q1,Q2,Q3,Q4,0,0,0
-};
-
-#undef Q0
-#undef Q1
-#undef Q2
-#undef Q3
-#undef Q4
-
-static const float q_3[7] =
-{
- (-6 << 15)/7.0, (-4 << 15)/7.0, (-2 << 15)/7.0, 0.0,
- ( 2 << 15)/7.0, ( 4 << 15)/7.0, ( 6 << 15)/7.0
-};
-
-#define Q0 ((-10 << 15) / 11.0)
-#define Q1 ((-8 << 15) / 11.0)
-#define Q2 ((-6 << 15) / 11.0)
-#define Q3 ((-4 << 15) / 11.0)
-#define Q4 ((-2 << 15) / 11.0)
-#define Q5 (0)
-#define Q6 ((2 << 15) / 11.0)
-#define Q7 ((4 << 15) / 11.0)
-#define Q8 ((6 << 15) / 11.0)
-#define Q9 ((8 << 15) / 11.0)
-#define QA ((10 << 15) / 11.0)
-
-static const float q_4_0[ 128 ] =
-{
- Q0, Q0, Q0, Q0, Q0, Q0, Q0, Q0, Q0, Q0, Q0,
- Q1, Q1, Q1, Q1, Q1, Q1, Q1, Q1, Q1, Q1, Q1,
- Q2, Q2, Q2, Q2, Q2, Q2, Q2, Q2, Q2, Q2, Q2,
- Q3, Q3, Q3, Q3, Q3, Q3, Q3, Q3, Q3, Q3, Q3,
- Q4, Q4, Q4, Q4, Q4, Q4, Q4, Q4, Q4, Q4, Q4,
- Q5, Q5, Q5, Q5, Q5, Q5, Q5, Q5, Q5, Q5, Q5,
- Q6, Q6, Q6, Q6, Q6, Q6, Q6, Q6, Q6, Q6, Q6,
- Q7, Q7, Q7, Q7, Q7, Q7, Q7, Q7, Q7, Q7, Q7,
- Q8, Q8, Q8, Q8, Q8, Q8, Q8, Q8, Q8, Q8, Q8,
- Q9, Q9, Q9, Q9, Q9, Q9, Q9, Q9, Q9, Q9, Q9,
- QA, QA, QA, QA, QA, QA, QA, QA, QA, QA, QA,
- 0, 0, 0, 0, 0, 0, 0
- };
-
-static const float q_4_1[ 128 ] =
-{
- Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
- Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
- Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
- Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
- Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
- Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
- Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
- Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
- Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
- Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
- Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7, Q8, Q9, QA,
- 0, 0, 0, 0, 0, 0, 0
-};
-
-#undef Q0
-#undef Q1
-#undef Q2
-#undef Q3
-#undef Q4
-#undef Q5
-#undef Q6
-#undef Q7
-#undef Q8
-#undef Q9
-#undef QA
-
-static const float q_5[15] =
-{
- (-14 << 15)/15.0,(-12 << 15)/15.0,(-10 << 15)/15.0,
- ( -8 << 15)/15.0,( -6 << 15)/15.0,( -4 << 15)/15.0,
- ( -2 << 15)/15.0, 0.0 ,( 2 << 15)/15.0,
- ( 4 << 15)/15.0,( 6 << 15)/15.0,( 8 << 15)/15.0,
- ( 10 << 15)/15.0,( 12 << 15)/15.0,( 14 << 15)/15.0
-};
-
-//
-// Scale factors for convert_to_float
-//
-
-static const uint32_t u32_scale_factors[25] =
-{
- 0x38000000, //2 ^ -(0 + 15)
- 0x37800000, //2 ^ -(1 + 15)
- 0x37000000, //2 ^ -(2 + 15)
- 0x36800000, //2 ^ -(3 + 15)
- 0x36000000, //2 ^ -(4 + 15)
- 0x35800000, //2 ^ -(5 + 15)
- 0x35000000, //2 ^ -(6 + 15)
- 0x34800000, //2 ^ -(7 + 15)
- 0x34000000, //2 ^ -(8 + 15)
- 0x33800000, //2 ^ -(9 + 15)
- 0x33000000, //2 ^ -(10 + 15)
- 0x32800000, //2 ^ -(11 + 15)
- 0x32000000, //2 ^ -(12 + 15)
- 0x31800000, //2 ^ -(13 + 15)
- 0x31000000, //2 ^ -(14 + 15)
- 0x30800000, //2 ^ -(15 + 15)
- 0x30000000, //2 ^ -(16 + 15)
- 0x2f800000, //2 ^ -(17 + 15)
- 0x2f000000, //2 ^ -(18 + 15)
- 0x2e800000, //2 ^ -(19 + 15)
- 0x2e000000, //2 ^ -(20 + 15)
- 0x2d800000, //2 ^ -(21 + 15)
- 0x2d000000, //2 ^ -(22 + 15)
- 0x2c800000, //2 ^ -(23 + 15)
- 0x2c000000 //2 ^ -(24 + 15)
-};
-
-static float *scale_factor = (float*)u32_scale_factors;
-
-//These store the persistent state of the packed mantissas
-static float q_1[2];
-static float q_2[2];
-static float q_4[1];
-static int32_t q_1_pointer;
-static int32_t q_2_pointer;
-static int32_t q_4_pointer;
-static float __inline__
-coeff_get_float(uint16_t bap, uint16_t dithflag, uint16_t exp);
-
-//Conversion from bap to number of bits in the mantissas
-//zeros account for cases 0,1,2,4 which are special cased
-static uint16_t qnttztab[16] =
-{
- 0, 0, 0, 3,
- 0, 4, 5, 6,
- 7, 8, 9, 10,
- 11, 12, 14, 16
-};
-
-static void coeff_reset(void);
-static float coeff_get_float(uint16_t bap, uint16_t dithflag, uint16_t exp);
-static void coeff_uncouple_ch(float samples[],bsi_t *bsi,audblk_t *audblk,uint32_t ch);
-
-void coeff_unpack(bsi_t *bsi, audblk_t *audblk, stream_samples_t samples)
-{
- uint16_t i,j;
- uint32_t done_cpl = 0;
-
- coeff_reset();
-
- for(i=0; i< bsi->nfchans; i++) {
- for(j=0; j < audblk->endmant[i]; j++)
- samples[i][j] = coeff_get_float(audblk->fbw_bap[i][j], audblk->dithflag[i], audblk->fbw_exp[i][j]);
-
- if(audblk->cplinu && audblk->chincpl[i] && !(done_cpl)) {
- // ncplmant is equal to 12 * ncplsubnd
- // Don't dither coupling channel until channel
- // separation so that interchannel noise is uncorrelated
- for(j=audblk->cplstrtmant; j < audblk->cplendmant; j++)
- audblk->cpl_flt[j] = coeff_get_float(audblk->cpl_bap[j],0, audblk->cpl_exp[j]);
- done_cpl = 1;
- }
- }
-
- //uncouple the channel if necessary
- if(audblk->cplinu) {
- for(i=0; i< bsi->nfchans; i++) {
- if(audblk->chincpl[i])
- coeff_uncouple_ch(samples[i],bsi,audblk,i);
- }
-
- }
-
- if(bsi->lfeon) {
- // There are always 7 mantissas for lfe, no dither for lfe
- for(j=0; j < 7 ; j++)
- samples[5][j] = coeff_get_float(audblk->lfe_bap[j], 0, audblk->lfe_exp[j]);
- }
-}
-
-
-/**
- * Fetch a float from the bitstream
- **/
-
-static float inline coeff_get_float (uint16_t bap, uint16_t dithflag, uint16_t exp)
-{
- uint16_t dummy = 0;
-
- //If the bap is 0-5 then we have special cases to take care of
- switch(bap) {
- case 0:
- if(dithflag)
- return (dither_gen() * scale_factor[exp]);
-
- return 0.0;
-
- case 1:
- if (q_1_pointer >= 0)
- return(q_1[q_1_pointer--] * scale_factor[exp]);
-
- if ((dummy = bitstream_get (5)) > 26)
- goto error;
-
- q_1[1] = q_1_1[dummy];
- q_1[0] = q_1_2[dummy];
- q_1_pointer = 1;
-
- return (q_1_0[dummy] * scale_factor[exp]);
-
- case 2:
- if(q_2_pointer >= 0)
- return (q_2[q_2_pointer--] * scale_factor[exp]);
-
- if ((dummy = bitstream_get (7)) > 124)
- goto error;
-
- q_2[1] = q_2_1[dummy];
- q_2[0] = q_2_2[dummy];
- q_2_pointer = 1;
-
- return (q_2_0[dummy] * scale_factor[exp]);
-
- case 3:
- if ((dummy = bitstream_get (3)) > 6)
- goto error;
-
- return (q_3[dummy] * scale_factor[exp]);
-
- case 4:
- if(q_4_pointer >= 0)
- return (q_4[q_4_pointer--] * scale_factor[exp]);
-
- if ((dummy = bitstream_get (7)) > 120)
- goto error;
-
- q_4[0] = q_4_1[dummy];
- q_4_pointer = 0;
-
- return (q_4_0[dummy] * scale_factor[exp]);
-
- case 5:
- if ((dummy = bitstream_get (4)) > 14)
- goto error;
-
- return (q_5[dummy] * scale_factor[exp]);
-
- default:
- dummy = bitstream_get(qnttztab[bap]);
- dummy <<= 16 - qnttztab[bap];
- return ((int16_t)dummy * scale_factor[exp]);
- }
-
-error:
-#ifdef DEBUG
- fprintf(stderr,"** Invalid mantissa - skipping frame **\n");
-#endif
- HANDLE_ERROR();
-}
-
-
-/**
- * Reset the mantissa state
- **/
-
-static void coeff_reset(void)
-{
- q_1_pointer = q_2_pointer = q_4_pointer = -1;
-}
-
-
-/**
- * Uncouple the coupling channel into a fbw channel
- **/
-
-static void coeff_uncouple_ch (float samples[],bsi_t *bsi,audblk_t *audblk,uint32_t ch)
-{
- uint32_t bnd = 0;
- uint32_t sub_bnd = 0;
- uint32_t i,j;
- float cpl_coord = 1.0;
- uint32_t cpl_exp_tmp;
- uint32_t cpl_mant_tmp;
-
- for (i=audblk->cplstrtmant;i<audblk->cplendmant;) {
- if (!audblk->cplbndstrc[sub_bnd++]) {
- cpl_exp_tmp = audblk->cplcoexp[ch][bnd] + 3 * audblk->mstrcplco[ch];
- if (audblk->cplcoexp[ch][bnd] == 15)
- cpl_mant_tmp = (audblk->cplcomant[ch][bnd]) << 11;
- else
- cpl_mant_tmp = ((0x10) | audblk->cplcomant[ch][bnd]) << 10;
-
- cpl_coord = (cpl_mant_tmp * scale_factor[cpl_exp_tmp]) * 8.0f;
-
- //Invert the phase for the right channel if necessary
- if(bsi->acmod == 0x2 && audblk->phsflginu && ch == 1 && audblk->phsflg[bnd])
- cpl_coord *= -1;
-
- bnd++;
- }
-
- for(j=0;j < 12; j++) {
- // Get new dither values for each channel if necessary,
- // so the channels are uncorrelated
- if(audblk->dithflag[ch] && !audblk->cpl_bap[i])
- samples[i] = cpl_coord * (dither_gen() * scale_factor[audblk->cpl_exp[i]]);
- else
- samples[i] = cpl_coord * audblk->cpl_flt[i];
-
- i++;
- }
- }
-}