• Main Page
  • Related Pages
  • Modules
  • Data Structures
  • Files
  • File List
  • Globals

libavcodec/mpegaudioenc.c

Go to the documentation of this file.
00001 /*
00002  * The simplest mpeg audio layer 2 encoder
00003  * Copyright (c) 2000, 2001 Fabrice Bellard
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00027 #include "avcodec.h"
00028 #include "put_bits.h"
00029 
00030 #undef  CONFIG_MPEGAUDIO_HP
00031 #define CONFIG_MPEGAUDIO_HP 0
00032 #include "mpegaudio.h"
00033 
00034 /* currently, cannot change these constants (need to modify
00035    quantization stage) */
00036 #define MUL(a,b) (((int64_t)(a) * (int64_t)(b)) >> FRAC_BITS)
00037 
00038 #define SAMPLES_BUF_SIZE 4096
00039 
00040 typedef struct MpegAudioContext {
00041     PutBitContext pb;
00042     int nb_channels;
00043     int freq, bit_rate;
00044     int lsf;           /* 1 if mpeg2 low bitrate selected */
00045     int bitrate_index; /* bit rate */
00046     int freq_index;
00047     int frame_size; /* frame size, in bits, without padding */
00048     int64_t nb_samples; /* total number of samples encoded */
00049     /* padding computation */
00050     int frame_frac, frame_frac_incr, do_padding;
00051     short samples_buf[MPA_MAX_CHANNELS][SAMPLES_BUF_SIZE]; /* buffer for filter */
00052     int samples_offset[MPA_MAX_CHANNELS];       /* offset in samples_buf */
00053     int sb_samples[MPA_MAX_CHANNELS][3][12][SBLIMIT];
00054     unsigned char scale_factors[MPA_MAX_CHANNELS][SBLIMIT][3]; /* scale factors */
00055     /* code to group 3 scale factors */
00056     unsigned char scale_code[MPA_MAX_CHANNELS][SBLIMIT];
00057     int sblimit; /* number of used subbands */
00058     const unsigned char *alloc_table;
00059 } MpegAudioContext;
00060 
00061 /* define it to use floats in quantization (I don't like floats !) */
00062 #define USE_FLOATS
00063 
00064 #include "mpegaudiodata.h"
00065 #include "mpegaudiotab.h"
00066 
00067 static av_cold int MPA_encode_init(AVCodecContext *avctx)
00068 {
00069     MpegAudioContext *s = avctx->priv_data;
00070     int freq = avctx->sample_rate;
00071     int bitrate = avctx->bit_rate;
00072     int channels = avctx->channels;
00073     int i, v, table;
00074     float a;
00075 
00076     if (channels <= 0 || channels > 2){
00077         av_log(avctx, AV_LOG_ERROR, "encoding %d channel(s) is not allowed in mp2\n", channels);
00078         return -1;
00079     }
00080     bitrate = bitrate / 1000;
00081     s->nb_channels = channels;
00082     s->freq = freq;
00083     s->bit_rate = bitrate * 1000;
00084     avctx->frame_size = MPA_FRAME_SIZE;
00085 
00086     /* encoding freq */
00087     s->lsf = 0;
00088     for(i=0;i<3;i++) {
00089         if (ff_mpa_freq_tab[i] == freq)
00090             break;
00091         if ((ff_mpa_freq_tab[i] / 2) == freq) {
00092             s->lsf = 1;
00093             break;
00094         }
00095     }
00096     if (i == 3){
00097         av_log(avctx, AV_LOG_ERROR, "Sampling rate %d is not allowed in mp2\n", freq);
00098         return -1;
00099     }
00100     s->freq_index = i;
00101 
00102     /* encoding bitrate & frequency */
00103     for(i=0;i<15;i++) {
00104         if (ff_mpa_bitrate_tab[s->lsf][1][i] == bitrate)
00105             break;
00106     }
00107     if (i == 15){
00108         av_log(avctx, AV_LOG_ERROR, "bitrate %d is not allowed in mp2\n", bitrate);
00109         return -1;
00110     }
00111     s->bitrate_index = i;
00112 
00113     /* compute total header size & pad bit */
00114 
00115     a = (float)(bitrate * 1000 * MPA_FRAME_SIZE) / (freq * 8.0);
00116     s->frame_size = ((int)a) * 8;
00117 
00118     /* frame fractional size to compute padding */
00119     s->frame_frac = 0;
00120     s->frame_frac_incr = (int)((a - floor(a)) * 65536.0);
00121 
00122     /* select the right allocation table */
00123     table = ff_mpa_l2_select_table(bitrate, s->nb_channels, freq, s->lsf);
00124 
00125     /* number of used subbands */
00126     s->sblimit = ff_mpa_sblimit_table[table];
00127     s->alloc_table = ff_mpa_alloc_tables[table];
00128 
00129     dprintf(avctx, "%d kb/s, %d Hz, frame_size=%d bits, table=%d, padincr=%x\n",
00130             bitrate, freq, s->frame_size, table, s->frame_frac_incr);
00131 
00132     for(i=0;i<s->nb_channels;i++)
00133         s->samples_offset[i] = 0;
00134 
00135     for(i=0;i<257;i++) {
00136         int v;
00137         v = ff_mpa_enwindow[i];
00138 #if WFRAC_BITS != 16
00139         v = (v + (1 << (16 - WFRAC_BITS - 1))) >> (16 - WFRAC_BITS);
00140 #endif
00141         filter_bank[i] = v;
00142         if ((i & 63) != 0)
00143             v = -v;
00144         if (i != 0)
00145             filter_bank[512 - i] = v;
00146     }
00147 
00148     for(i=0;i<64;i++) {
00149         v = (int)(pow(2.0, (3 - i) / 3.0) * (1 << 20));
00150         if (v <= 0)
00151             v = 1;
00152         scale_factor_table[i] = v;
00153 #ifdef USE_FLOATS
00154         scale_factor_inv_table[i] = pow(2.0, -(3 - i) / 3.0) / (float)(1 << 20);
00155 #else
00156 #define P 15
00157         scale_factor_shift[i] = 21 - P - (i / 3);
00158         scale_factor_mult[i] = (1 << P) * pow(2.0, (i % 3) / 3.0);
00159 #endif
00160     }
00161     for(i=0;i<128;i++) {
00162         v = i - 64;
00163         if (v <= -3)
00164             v = 0;
00165         else if (v < 0)
00166             v = 1;
00167         else if (v == 0)
00168             v = 2;
00169         else if (v < 3)
00170             v = 3;
00171         else
00172             v = 4;
00173         scale_diff_table[i] = v;
00174     }
00175 
00176     for(i=0;i<17;i++) {
00177         v = ff_mpa_quant_bits[i];
00178         if (v < 0)
00179             v = -v;
00180         else
00181             v = v * 3;
00182         total_quant_bits[i] = 12 * v;
00183     }
00184 
00185     avctx->coded_frame= avcodec_alloc_frame();
00186     avctx->coded_frame->key_frame= 1;
00187 
00188     return 0;
00189 }
00190 
00191 /* 32 point floating point IDCT without 1/sqrt(2) coef zero scaling */
00192 static void idct32(int *out, int *tab)
00193 {
00194     int i, j;
00195     int *t, *t1, xr;
00196     const int *xp = costab32;
00197 
00198     for(j=31;j>=3;j-=2) tab[j] += tab[j - 2];
00199 
00200     t = tab + 30;
00201     t1 = tab + 2;
00202     do {
00203         t[0] += t[-4];
00204         t[1] += t[1 - 4];
00205         t -= 4;
00206     } while (t != t1);
00207 
00208     t = tab + 28;
00209     t1 = tab + 4;
00210     do {
00211         t[0] += t[-8];
00212         t[1] += t[1-8];
00213         t[2] += t[2-8];
00214         t[3] += t[3-8];
00215         t -= 8;
00216     } while (t != t1);
00217 
00218     t = tab;
00219     t1 = tab + 32;
00220     do {
00221         t[ 3] = -t[ 3];
00222         t[ 6] = -t[ 6];
00223 
00224         t[11] = -t[11];
00225         t[12] = -t[12];
00226         t[13] = -t[13];
00227         t[15] = -t[15];
00228         t += 16;
00229     } while (t != t1);
00230 
00231 
00232     t = tab;
00233     t1 = tab + 8;
00234     do {
00235         int x1, x2, x3, x4;
00236 
00237         x3 = MUL(t[16], FIX(SQRT2*0.5));
00238         x4 = t[0] - x3;
00239         x3 = t[0] + x3;
00240 
00241         x2 = MUL(-(t[24] + t[8]), FIX(SQRT2*0.5));
00242         x1 = MUL((t[8] - x2), xp[0]);
00243         x2 = MUL((t[8] + x2), xp[1]);
00244 
00245         t[ 0] = x3 + x1;
00246         t[ 8] = x4 - x2;
00247         t[16] = x4 + x2;
00248         t[24] = x3 - x1;
00249         t++;
00250     } while (t != t1);
00251 
00252     xp += 2;
00253     t = tab;
00254     t1 = tab + 4;
00255     do {
00256         xr = MUL(t[28],xp[0]);
00257         t[28] = (t[0] - xr);
00258         t[0] = (t[0] + xr);
00259 
00260         xr = MUL(t[4],xp[1]);
00261         t[ 4] = (t[24] - xr);
00262         t[24] = (t[24] + xr);
00263 
00264         xr = MUL(t[20],xp[2]);
00265         t[20] = (t[8] - xr);
00266         t[ 8] = (t[8] + xr);
00267 
00268         xr = MUL(t[12],xp[3]);
00269         t[12] = (t[16] - xr);
00270         t[16] = (t[16] + xr);
00271         t++;
00272     } while (t != t1);
00273     xp += 4;
00274 
00275     for (i = 0; i < 4; i++) {
00276         xr = MUL(tab[30-i*4],xp[0]);
00277         tab[30-i*4] = (tab[i*4] - xr);
00278         tab[   i*4] = (tab[i*4] + xr);
00279 
00280         xr = MUL(tab[ 2+i*4],xp[1]);
00281         tab[ 2+i*4] = (tab[28-i*4] - xr);
00282         tab[28-i*4] = (tab[28-i*4] + xr);
00283 
00284         xr = MUL(tab[31-i*4],xp[0]);
00285         tab[31-i*4] = (tab[1+i*4] - xr);
00286         tab[ 1+i*4] = (tab[1+i*4] + xr);
00287 
00288         xr = MUL(tab[ 3+i*4],xp[1]);
00289         tab[ 3+i*4] = (tab[29-i*4] - xr);
00290         tab[29-i*4] = (tab[29-i*4] + xr);
00291 
00292         xp += 2;
00293     }
00294 
00295     t = tab + 30;
00296     t1 = tab + 1;
00297     do {
00298         xr = MUL(t1[0], *xp);
00299         t1[0] = (t[0] - xr);
00300         t[0] = (t[0] + xr);
00301         t -= 2;
00302         t1 += 2;
00303         xp++;
00304     } while (t >= tab);
00305 
00306     for(i=0;i<32;i++) {
00307         out[i] = tab[bitinv32[i]];
00308     }
00309 }
00310 
00311 #define WSHIFT (WFRAC_BITS + 15 - FRAC_BITS)
00312 
00313 static void filter(MpegAudioContext *s, int ch, short *samples, int incr)
00314 {
00315     short *p, *q;
00316     int sum, offset, i, j;
00317     int tmp[64];
00318     int tmp1[32];
00319     int *out;
00320 
00321     //    print_pow1(samples, 1152);
00322 
00323     offset = s->samples_offset[ch];
00324     out = &s->sb_samples[ch][0][0][0];
00325     for(j=0;j<36;j++) {
00326         /* 32 samples at once */
00327         for(i=0;i<32;i++) {
00328             s->samples_buf[ch][offset + (31 - i)] = samples[0];
00329             samples += incr;
00330         }
00331 
00332         /* filter */
00333         p = s->samples_buf[ch] + offset;
00334         q = filter_bank;
00335         /* maxsum = 23169 */
00336         for(i=0;i<64;i++) {
00337             sum = p[0*64] * q[0*64];
00338             sum += p[1*64] * q[1*64];
00339             sum += p[2*64] * q[2*64];
00340             sum += p[3*64] * q[3*64];
00341             sum += p[4*64] * q[4*64];
00342             sum += p[5*64] * q[5*64];
00343             sum += p[6*64] * q[6*64];
00344             sum += p[7*64] * q[7*64];
00345             tmp[i] = sum;
00346             p++;
00347             q++;
00348         }
00349         tmp1[0] = tmp[16] >> WSHIFT;
00350         for( i=1; i<=16; i++ ) tmp1[i] = (tmp[i+16]+tmp[16-i]) >> WSHIFT;
00351         for( i=17; i<=31; i++ ) tmp1[i] = (tmp[i+16]-tmp[80-i]) >> WSHIFT;
00352 
00353         idct32(out, tmp1);
00354 
00355         /* advance of 32 samples */
00356         offset -= 32;
00357         out += 32;
00358         /* handle the wrap around */
00359         if (offset < 0) {
00360             memmove(s->samples_buf[ch] + SAMPLES_BUF_SIZE - (512 - 32),
00361                     s->samples_buf[ch], (512 - 32) * 2);
00362             offset = SAMPLES_BUF_SIZE - 512;
00363         }
00364     }
00365     s->samples_offset[ch] = offset;
00366 
00367     //    print_pow(s->sb_samples, 1152);
00368 }
00369 
00370 static void compute_scale_factors(unsigned char scale_code[SBLIMIT],
00371                                   unsigned char scale_factors[SBLIMIT][3],
00372                                   int sb_samples[3][12][SBLIMIT],
00373                                   int sblimit)
00374 {
00375     int *p, vmax, v, n, i, j, k, code;
00376     int index, d1, d2;
00377     unsigned char *sf = &scale_factors[0][0];
00378 
00379     for(j=0;j<sblimit;j++) {
00380         for(i=0;i<3;i++) {
00381             /* find the max absolute value */
00382             p = &sb_samples[i][0][j];
00383             vmax = abs(*p);
00384             for(k=1;k<12;k++) {
00385                 p += SBLIMIT;
00386                 v = abs(*p);
00387                 if (v > vmax)
00388                     vmax = v;
00389             }
00390             /* compute the scale factor index using log 2 computations */
00391             if (vmax > 1) {
00392                 n = av_log2(vmax);
00393                 /* n is the position of the MSB of vmax. now
00394                    use at most 2 compares to find the index */
00395                 index = (21 - n) * 3 - 3;
00396                 if (index >= 0) {
00397                     while (vmax <= scale_factor_table[index+1])
00398                         index++;
00399                 } else {
00400                     index = 0; /* very unlikely case of overflow */
00401                 }
00402             } else {
00403                 index = 62; /* value 63 is not allowed */
00404             }
00405 
00406 #if 0
00407             printf("%2d:%d in=%x %x %d\n",
00408                    j, i, vmax, scale_factor_table[index], index);
00409 #endif
00410             /* store the scale factor */
00411             assert(index >=0 && index <= 63);
00412             sf[i] = index;
00413         }
00414 
00415         /* compute the transmission factor : look if the scale factors
00416            are close enough to each other */
00417         d1 = scale_diff_table[sf[0] - sf[1] + 64];
00418         d2 = scale_diff_table[sf[1] - sf[2] + 64];
00419 
00420         /* handle the 25 cases */
00421         switch(d1 * 5 + d2) {
00422         case 0*5+0:
00423         case 0*5+4:
00424         case 3*5+4:
00425         case 4*5+0:
00426         case 4*5+4:
00427             code = 0;
00428             break;
00429         case 0*5+1:
00430         case 0*5+2:
00431         case 4*5+1:
00432         case 4*5+2:
00433             code = 3;
00434             sf[2] = sf[1];
00435             break;
00436         case 0*5+3:
00437         case 4*5+3:
00438             code = 3;
00439             sf[1] = sf[2];
00440             break;
00441         case 1*5+0:
00442         case 1*5+4:
00443         case 2*5+4:
00444             code = 1;
00445             sf[1] = sf[0];
00446             break;
00447         case 1*5+1:
00448         case 1*5+2:
00449         case 2*5+0:
00450         case 2*5+1:
00451         case 2*5+2:
00452             code = 2;
00453             sf[1] = sf[2] = sf[0];
00454             break;
00455         case 2*5+3:
00456         case 3*5+3:
00457             code = 2;
00458             sf[0] = sf[1] = sf[2];
00459             break;
00460         case 3*5+0:
00461         case 3*5+1:
00462         case 3*5+2:
00463             code = 2;
00464             sf[0] = sf[2] = sf[1];
00465             break;
00466         case 1*5+3:
00467             code = 2;
00468             if (sf[0] > sf[2])
00469               sf[0] = sf[2];
00470             sf[1] = sf[2] = sf[0];
00471             break;
00472         default:
00473             assert(0); //cannot happen
00474             code = 0;           /* kill warning */
00475         }
00476 
00477 #if 0
00478         printf("%d: %2d %2d %2d %d %d -> %d\n", j,
00479                sf[0], sf[1], sf[2], d1, d2, code);
00480 #endif
00481         scale_code[j] = code;
00482         sf += 3;
00483     }
00484 }
00485 
00486 /* The most important function : psycho acoustic module. In this
00487    encoder there is basically none, so this is the worst you can do,
00488    but also this is the simpler. */
00489 static void psycho_acoustic_model(MpegAudioContext *s, short smr[SBLIMIT])
00490 {
00491     int i;
00492 
00493     for(i=0;i<s->sblimit;i++) {
00494         smr[i] = (int)(fixed_smr[i] * 10);
00495     }
00496 }
00497 
00498 
00499 #define SB_NOTALLOCATED  0
00500 #define SB_ALLOCATED     1
00501 #define SB_NOMORE        2
00502 
00503 /* Try to maximize the smr while using a number of bits inferior to
00504    the frame size. I tried to make the code simpler, faster and
00505    smaller than other encoders :-) */
00506 static void compute_bit_allocation(MpegAudioContext *s,
00507                                    short smr1[MPA_MAX_CHANNELS][SBLIMIT],
00508                                    unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT],
00509                                    int *padding)
00510 {
00511     int i, ch, b, max_smr, max_ch, max_sb, current_frame_size, max_frame_size;
00512     int incr;
00513     short smr[MPA_MAX_CHANNELS][SBLIMIT];
00514     unsigned char subband_status[MPA_MAX_CHANNELS][SBLIMIT];
00515     const unsigned char *alloc;
00516 
00517     memcpy(smr, smr1, s->nb_channels * sizeof(short) * SBLIMIT);
00518     memset(subband_status, SB_NOTALLOCATED, s->nb_channels * SBLIMIT);
00519     memset(bit_alloc, 0, s->nb_channels * SBLIMIT);
00520 
00521     /* compute frame size and padding */
00522     max_frame_size = s->frame_size;
00523     s->frame_frac += s->frame_frac_incr;
00524     if (s->frame_frac >= 65536) {
00525         s->frame_frac -= 65536;
00526         s->do_padding = 1;
00527         max_frame_size += 8;
00528     } else {
00529         s->do_padding = 0;
00530     }
00531 
00532     /* compute the header + bit alloc size */
00533     current_frame_size = 32;
00534     alloc = s->alloc_table;
00535     for(i=0;i<s->sblimit;i++) {
00536         incr = alloc[0];
00537         current_frame_size += incr * s->nb_channels;
00538         alloc += 1 << incr;
00539     }
00540     for(;;) {
00541         /* look for the subband with the largest signal to mask ratio */
00542         max_sb = -1;
00543         max_ch = -1;
00544         max_smr = INT_MIN;
00545         for(ch=0;ch<s->nb_channels;ch++) {
00546             for(i=0;i<s->sblimit;i++) {
00547                 if (smr[ch][i] > max_smr && subband_status[ch][i] != SB_NOMORE) {
00548                     max_smr = smr[ch][i];
00549                     max_sb = i;
00550                     max_ch = ch;
00551                 }
00552             }
00553         }
00554 #if 0
00555         printf("current=%d max=%d max_sb=%d alloc=%d\n",
00556                current_frame_size, max_frame_size, max_sb,
00557                bit_alloc[max_sb]);
00558 #endif
00559         if (max_sb < 0)
00560             break;
00561 
00562         /* find alloc table entry (XXX: not optimal, should use
00563            pointer table) */
00564         alloc = s->alloc_table;
00565         for(i=0;i<max_sb;i++) {
00566             alloc += 1 << alloc[0];
00567         }
00568 
00569         if (subband_status[max_ch][max_sb] == SB_NOTALLOCATED) {
00570             /* nothing was coded for this band: add the necessary bits */
00571             incr = 2 + nb_scale_factors[s->scale_code[max_ch][max_sb]] * 6;
00572             incr += total_quant_bits[alloc[1]];
00573         } else {
00574             /* increments bit allocation */
00575             b = bit_alloc[max_ch][max_sb];
00576             incr = total_quant_bits[alloc[b + 1]] -
00577                 total_quant_bits[alloc[b]];
00578         }
00579 
00580         if (current_frame_size + incr <= max_frame_size) {
00581             /* can increase size */
00582             b = ++bit_alloc[max_ch][max_sb];
00583             current_frame_size += incr;
00584             /* decrease smr by the resolution we added */
00585             smr[max_ch][max_sb] = smr1[max_ch][max_sb] - quant_snr[alloc[b]];
00586             /* max allocation size reached ? */
00587             if (b == ((1 << alloc[0]) - 1))
00588                 subband_status[max_ch][max_sb] = SB_NOMORE;
00589             else
00590                 subband_status[max_ch][max_sb] = SB_ALLOCATED;
00591         } else {
00592             /* cannot increase the size of this subband */
00593             subband_status[max_ch][max_sb] = SB_NOMORE;
00594         }
00595     }
00596     *padding = max_frame_size - current_frame_size;
00597     assert(*padding >= 0);
00598 
00599 #if 0
00600     for(i=0;i<s->sblimit;i++) {
00601         printf("%d ", bit_alloc[i]);
00602     }
00603     printf("\n");
00604 #endif
00605 }
00606 
00607 /*
00608  * Output the mpeg audio layer 2 frame. Note how the code is small
00609  * compared to other encoders :-)
00610  */
00611 static void encode_frame(MpegAudioContext *s,
00612                          unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT],
00613                          int padding)
00614 {
00615     int i, j, k, l, bit_alloc_bits, b, ch;
00616     unsigned char *sf;
00617     int q[3];
00618     PutBitContext *p = &s->pb;
00619 
00620     /* header */
00621 
00622     put_bits(p, 12, 0xfff);
00623     put_bits(p, 1, 1 - s->lsf); /* 1 = mpeg1 ID, 0 = mpeg2 lsf ID */
00624     put_bits(p, 2, 4-2);  /* layer 2 */
00625     put_bits(p, 1, 1); /* no error protection */
00626     put_bits(p, 4, s->bitrate_index);
00627     put_bits(p, 2, s->freq_index);
00628     put_bits(p, 1, s->do_padding); /* use padding */
00629     put_bits(p, 1, 0);             /* private_bit */
00630     put_bits(p, 2, s->nb_channels == 2 ? MPA_STEREO : MPA_MONO);
00631     put_bits(p, 2, 0); /* mode_ext */
00632     put_bits(p, 1, 0); /* no copyright */
00633     put_bits(p, 1, 1); /* original */
00634     put_bits(p, 2, 0); /* no emphasis */
00635 
00636     /* bit allocation */
00637     j = 0;
00638     for(i=0;i<s->sblimit;i++) {
00639         bit_alloc_bits = s->alloc_table[j];
00640         for(ch=0;ch<s->nb_channels;ch++) {
00641             put_bits(p, bit_alloc_bits, bit_alloc[ch][i]);
00642         }
00643         j += 1 << bit_alloc_bits;
00644     }
00645 
00646     /* scale codes */
00647     for(i=0;i<s->sblimit;i++) {
00648         for(ch=0;ch<s->nb_channels;ch++) {
00649             if (bit_alloc[ch][i])
00650                 put_bits(p, 2, s->scale_code[ch][i]);
00651         }
00652     }
00653 
00654     /* scale factors */
00655     for(i=0;i<s->sblimit;i++) {
00656         for(ch=0;ch<s->nb_channels;ch++) {
00657             if (bit_alloc[ch][i]) {
00658                 sf = &s->scale_factors[ch][i][0];
00659                 switch(s->scale_code[ch][i]) {
00660                 case 0:
00661                     put_bits(p, 6, sf[0]);
00662                     put_bits(p, 6, sf[1]);
00663                     put_bits(p, 6, sf[2]);
00664                     break;
00665                 case 3:
00666                 case 1:
00667                     put_bits(p, 6, sf[0]);
00668                     put_bits(p, 6, sf[2]);
00669                     break;
00670                 case 2:
00671                     put_bits(p, 6, sf[0]);
00672                     break;
00673                 }
00674             }
00675         }
00676     }
00677 
00678     /* quantization & write sub band samples */
00679 
00680     for(k=0;k<3;k++) {
00681         for(l=0;l<12;l+=3) {
00682             j = 0;
00683             for(i=0;i<s->sblimit;i++) {
00684                 bit_alloc_bits = s->alloc_table[j];
00685                 for(ch=0;ch<s->nb_channels;ch++) {
00686                     b = bit_alloc[ch][i];
00687                     if (b) {
00688                         int qindex, steps, m, sample, bits;
00689                         /* we encode 3 sub band samples of the same sub band at a time */
00690                         qindex = s->alloc_table[j+b];
00691                         steps = ff_mpa_quant_steps[qindex];
00692                         for(m=0;m<3;m++) {
00693                             sample = s->sb_samples[ch][k][l + m][i];
00694                             /* divide by scale factor */
00695 #ifdef USE_FLOATS
00696                             {
00697                                 float a;
00698                                 a = (float)sample * scale_factor_inv_table[s->scale_factors[ch][i][k]];
00699                                 q[m] = (int)((a + 1.0) * steps * 0.5);
00700                             }
00701 #else
00702                             {
00703                                 int q1, e, shift, mult;
00704                                 e = s->scale_factors[ch][i][k];
00705                                 shift = scale_factor_shift[e];
00706                                 mult = scale_factor_mult[e];
00707 
00708                                 /* normalize to P bits */
00709                                 if (shift < 0)
00710                                     q1 = sample << (-shift);
00711                                 else
00712                                     q1 = sample >> shift;
00713                                 q1 = (q1 * mult) >> P;
00714                                 q[m] = ((q1 + (1 << P)) * steps) >> (P + 1);
00715                             }
00716 #endif
00717                             if (q[m] >= steps)
00718                                 q[m] = steps - 1;
00719                             assert(q[m] >= 0 && q[m] < steps);
00720                         }
00721                         bits = ff_mpa_quant_bits[qindex];
00722                         if (bits < 0) {
00723                             /* group the 3 values to save bits */
00724                             put_bits(p, -bits,
00725                                      q[0] + steps * (q[1] + steps * q[2]));
00726 #if 0
00727                             printf("%d: gr1 %d\n",
00728                                    i, q[0] + steps * (q[1] + steps * q[2]));
00729 #endif
00730                         } else {
00731 #if 0
00732                             printf("%d: gr3 %d %d %d\n",
00733                                    i, q[0], q[1], q[2]);
00734 #endif
00735                             put_bits(p, bits, q[0]);
00736                             put_bits(p, bits, q[1]);
00737                             put_bits(p, bits, q[2]);
00738                         }
00739                     }
00740                 }
00741                 /* next subband in alloc table */
00742                 j += 1 << bit_alloc_bits;
00743             }
00744         }
00745     }
00746 
00747     /* padding */
00748     for(i=0;i<padding;i++)
00749         put_bits(p, 1, 0);
00750 
00751     /* flush */
00752     flush_put_bits(p);
00753 }
00754 
00755 static int MPA_encode_frame(AVCodecContext *avctx,
00756                             unsigned char *frame, int buf_size, void *data)
00757 {
00758     MpegAudioContext *s = avctx->priv_data;
00759     short *samples = data;
00760     short smr[MPA_MAX_CHANNELS][SBLIMIT];
00761     unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT];
00762     int padding, i;
00763 
00764     for(i=0;i<s->nb_channels;i++) {
00765         filter(s, i, samples + i, s->nb_channels);
00766     }
00767 
00768     for(i=0;i<s->nb_channels;i++) {
00769         compute_scale_factors(s->scale_code[i], s->scale_factors[i],
00770                               s->sb_samples[i], s->sblimit);
00771     }
00772     for(i=0;i<s->nb_channels;i++) {
00773         psycho_acoustic_model(s, smr[i]);
00774     }
00775     compute_bit_allocation(s, smr, bit_alloc, &padding);
00776 
00777     init_put_bits(&s->pb, frame, MPA_MAX_CODED_FRAME_SIZE);
00778 
00779     encode_frame(s, bit_alloc, padding);
00780 
00781     s->nb_samples += MPA_FRAME_SIZE;
00782     return put_bits_ptr(&s->pb) - s->pb.buf;
00783 }
00784 
00785 static av_cold int MPA_encode_close(AVCodecContext *avctx)
00786 {
00787     av_freep(&avctx->coded_frame);
00788     return 0;
00789 }
00790 
00791 AVCodec mp2_encoder = {
00792     "mp2",
00793     AVMEDIA_TYPE_AUDIO,
00794     CODEC_ID_MP2,
00795     sizeof(MpegAudioContext),
00796     MPA_encode_init,
00797     MPA_encode_frame,
00798     MPA_encode_close,
00799     NULL,
00800     .sample_fmts = (const enum SampleFormat[]){SAMPLE_FMT_S16,SAMPLE_FMT_NONE},
00801     .supported_samplerates= (const int[]){44100, 48000,  32000, 22050, 24000, 16000, 0},
00802     .long_name = NULL_IF_CONFIG_SMALL("MP2 (MPEG audio layer 2)"),
00803 };
00804 
00805 #undef FIX

Generated on Fri Sep 16 2011 17:17:40 for FFmpeg by  doxygen 1.7.1