mpegaudioenc.c

ffmpeg的完整源代码和作者自己写的文档。不但有在Linux的工程哦

/*
 * The simplest mpeg audio layer 2 encoder
 * Copyright (c) 2000, 2001 Fabrice Bellard.
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg 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
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
 * @file mpegaudio.c
 * The simplest mpeg audio layer 2 encoder.
 */

#include "avcodec.h"
#include "bitstream.h"
#include "mpegaudio.h"

/* currently, cannot change these constants (need to modify
   quantization stage) */
#define MUL(a,b) (((int64_t)(a) * (int64_t)(b)) >> FRAC_BITS)

#define SAMPLES_BUF_SIZE 4096

typedef struct MpegAudioContext {
    PutBitContext pb;
    int nb_channels;
    int freq, bit_rate;
    int lsf;           /* 1 if mpeg2 low bitrate selected */
    int bitrate_index; /* bit rate */
    int freq_index;
    int frame_size; /* frame size, in bits, without padding */
    int64_t nb_samples; /* total number of samples encoded */
    /* padding computation */
    int frame_frac, frame_frac_incr, do_padding;
    short samples_buf[MPA_MAX_CHANNELS][SAMPLES_BUF_SIZE]; /* buffer for filter */
    int samples_offset[MPA_MAX_CHANNELS];       /* offset in samples_buf */
    int sb_samples[MPA_MAX_CHANNELS][3][12][SBLIMIT];
    unsigned char scale_factors[MPA_MAX_CHANNELS][SBLIMIT][3]; /* scale factors */
    /* code to group 3 scale factors */
    unsigned char scale_code[MPA_MAX_CHANNELS][SBLIMIT];
    int sblimit; /* number of used subbands */
    const unsigned char *alloc_table;
} MpegAudioContext;

/* define it to use floats in quantization (I don't like floats !) */
//#define USE_FLOATS

#include "mpegaudiodata.h"
#include "mpegaudiotab.h"

static int MPA_encode_init(AVCodecContext *avctx)
{
    MpegAudioContext *s = avctx->priv_data;
    int freq = avctx->sample_rate;
    int bitrate = avctx->bit_rate;
    int channels = avctx->channels;
    int i, v, table;
    float a;

    if (channels <= 0 || channels > 2){
        av_log(avctx, AV_LOG_ERROR, "encoding %d channel(s) is not allowed in mp2\n", channels);
        return -1;
    }
    bitrate = bitrate / 1000;
    s->nb_channels = channels;
    s->freq = freq;
    s->bit_rate = bitrate * 1000;
    avctx->frame_size = MPA_FRAME_SIZE;

    /* encoding freq */
    s->lsf = 0;
    for(i=0;i<3;i++) {
        if (ff_mpa_freq_tab[i] == freq)
            break;
        if ((ff_mpa_freq_tab[i] / 2) == freq) {
            s->lsf = 1;
            break;
        }
    }
    if (i == 3){
        av_log(avctx, AV_LOG_ERROR, "Sampling rate %d is not allowed in mp2\n", freq);
        return -1;
    }
    s->freq_index = i;

    /* encoding bitrate & frequency */
    for(i=0;i<15;i++) {
        if (ff_mpa_bitrate_tab[s->lsf][1][i] == bitrate)
            break;
    }
    if (i == 15){
        av_log(avctx, AV_LOG_ERROR, "bitrate %d is not allowed in mp2\n", bitrate);
        return -1;
    }
    s->bitrate_index = i;

    /* compute total header size & pad bit */

    a = (float)(bitrate * 1000 * MPA_FRAME_SIZE) / (freq * 8.0);
    s->frame_size = ((int)a) * 8;

    /* frame fractional size to compute padding */
    s->frame_frac = 0;
    s->frame_frac_incr = (int)((a - floor(a)) * 65536.0);

    /* select the right allocation table */
    table = ff_mpa_l2_select_table(bitrate, s->nb_channels, freq, s->lsf);

    /* number of used subbands */
    s->sblimit = ff_mpa_sblimit_table[table];
    s->alloc_table = ff_mpa_alloc_tables[table];

#ifdef DEBUG
    av_log(avctx, AV_LOG_DEBUG, "%d kb/s, %d Hz, frame_size=%d bits, table=%d, padincr=%x\n",
           bitrate, freq, s->frame_size, table, s->frame_frac_incr);
#endif

    for(i=0;i<s->nb_channels;i++)
        s->samples_offset[i] = 0;

    for(i=0;i<257;i++) {
        int v;
        v = ff_mpa_enwindow[i];
#if WFRAC_BITS != 16
        v = (v + (1 << (16 - WFRAC_BITS - 1))) >> (16 - WFRAC_BITS);
#endif
        filter_bank[i] = v;
        if ((i & 63) != 0)
            v = -v;
        if (i != 0)
            filter_bank[512 - i] = v;
    }

    for(i=0;i<64;i++) {
        v = (int)(pow(2.0, (3 - i) / 3.0) * (1 << 20));
        if (v <= 0)
            v = 1;
        scale_factor_table[i] = v;
#ifdef USE_FLOATS
        scale_factor_inv_table[i] = pow(2.0, -(3 - i) / 3.0) / (float)(1 << 20);
#else
#define P 15
        scale_factor_shift[i] = 21 - P - (i / 3);
        scale_factor_mult[i] = (1 << P) * pow(2.0, (i % 3) / 3.0);
#endif
    }
    for(i=0;i<128;i++) {
        v = i - 64;
        if (v <= -3)
            v = 0;
        else if (v < 0)
            v = 1;
        else if (v == 0)
            v = 2;
        else if (v < 3)
            v = 3;
        else
            v = 4;
        scale_diff_table[i] = v;
    }

    for(i=0;i<17;i++) {
        v = ff_mpa_quant_bits[i];
        if (v < 0)
            v = -v;
        else
            v = v * 3;
        total_quant_bits[i] = 12 * v;
    }

    avctx->coded_frame= avcodec_alloc_frame();
    avctx->coded_frame->key_frame= 1;

    return 0;
}

/* 32 point floating point IDCT without 1/sqrt(2) coef zero scaling */
static void idct32(int *out, int *tab)
{
    int i, j;
    int *t, *t1, xr;
    const int *xp = costab32;

    for(j=31;j>=3;j-=2) tab[j] += tab[j - 2];

    t = tab + 30;
    t1 = tab + 2;
    do {
        t[0] += t[-4];
        t[1] += t[1 - 4];
        t -= 4;
    } while (t != t1);

    t = tab + 28;
    t1 = tab + 4;
    do {
        t[0] += t[-8];
        t[1] += t[1-8];
        t[2] += t[2-8];
        t[3] += t[3-8];
        t -= 8;
    } while (t != t1);

    t = tab;
    t1 = tab + 32;
    do {
        t[ 3] = -t[ 3];
        t[ 6] = -t[ 6];

        t[11] = -t[11];
        t[12] = -t[12];
        t[13] = -t[13];
        t[15] = -t[15];
        t += 16;
    } while (t != t1);


    t = tab;
    t1 = tab + 8;
    do {
        int x1, x2, x3, x4;

        x3 = MUL(t[16], FIX(SQRT2*0.5));
        x4 = t[0] - x3;
        x3 = t[0] + x3;

        x2 = MUL(-(t[24] + t[8]), FIX(SQRT2*0.5));
        x1 = MUL((t[8] - x2), xp[0]);
        x2 = MUL((t[8] + x2), xp[1]);

        t[ 0] = x3 + x1;
        t[ 8] = x4 - x2;
        t[16] = x4 + x2;
        t[24] = x3 - x1;
        t++;
    } while (t != t1);

    xp += 2;
    t = tab;
    t1 = tab + 4;
    do {
        xr = MUL(t[28],xp[0]);
        t[28] = (t[0] - xr);
        t[0] = (t[0] + xr);

        xr = MUL(t[4],xp[1]);
        t[ 4] = (t[24] - xr);
        t[24] = (t[24] + xr);

        xr = MUL(t[20],xp[2]);
        t[20] = (t[8] - xr);
        t[ 8] = (t[8] + xr);

        xr = MUL(t[12],xp[3]);
        t[12] = (t[16] - xr);
        t[16] = (t[16] + xr);
        t++;
    } while (t != t1);
    xp += 4;

    for (i = 0; i < 4; i++) {
        xr = MUL(tab[30-i*4],xp[0]);
        tab[30-i*4] = (tab[i*4] - xr);
        tab[   i*4] = (tab[i*4] + xr);

        xr = MUL(tab[ 2+i*4],xp[1]);
        tab[ 2+i*4] = (tab[28-i*4] - xr);
        tab[28-i*4] = (tab[28-i*4] + xr);

        xr = MUL(tab[31-i*4],xp[0]);
        tab[31-i*4] = (tab[1+i*4] - xr);
        tab[ 1+i*4] = (tab[1+i*4] + xr);

        xr = MUL(tab[ 3+i*4],xp[1]);
        tab[ 3+i*4] = (tab[29-i*4] - xr);
        tab[29-i*4] = (tab[29-i*4] + xr);

        xp += 2;
    }

    t = tab + 30;
    t1 = tab + 1;
    do {
        xr = MUL(t1[0], *xp);
        t1[0] = (t[0] - xr);
        t[0] = (t[0] + xr);
        t -= 2;
        t1 += 2;
        xp++;
    } while (t >= tab);

    for(i=0;i<32;i++) {
        out[i] = tab[bitinv32[i]];
    }
}

#define WSHIFT (WFRAC_BITS + 15 - FRAC_BITS)

static void filter(MpegAudioContext *s, int ch, short *samples, int incr)
{
    short *p, *q;
    int sum, offset, i, j;
    int tmp[64];
    int tmp1[32];
    int *out;

    //    print_pow1(samples, 1152);

    offset = s->samples_offset[ch];
    out = &s->sb_samples[ch][0][0][0];
    for(j=0;j<36;j++) {
        /* 32 samples at once */
        for(i=0;i<32;i++) {
            s->samples_buf[ch][offset + (31 - i)] = samples[0];
            samples += incr;
        }

        /* filter */
        p = s->samples_buf[ch] + offset;
        q = filter_bank;
        /* maxsum = 23169 */
        for(i=0;i<64;i++) {
            sum = p[0*64] * q[0*64];
            sum += p[1*64] * q[1*64];
            sum += p[2*64] * q[2*64];
            sum += p[3*64] * q[3*64];
            sum += p[4*64] * q[4*64];
            sum += p[5*64] * q[5*64];
            sum += p[6*64] * q[6*64];
            sum += p[7*64] * q[7*64];
            tmp[i] = sum;
            p++;
            q++;
        }
        tmp1[0] = tmp[16] >> WSHIFT;
        for( i=1; i<=16; i++ ) tmp1[i] = (tmp[i+16]+tmp[16-i]) >> WSHIFT;
        for( i=17; i<=31; i++ ) tmp1[i] = (tmp[i+16]-tmp[80-i]) >> WSHIFT;

        idct32(out, tmp1);

        /* advance of 32 samples */
        offset -= 32;
        out += 32;
        /* handle the wrap around */
        if (offset < 0) {
            memmove(s->samples_buf[ch] + SAMPLES_BUF_SIZE - (512 - 32),
                    s->samples_buf[ch], (512 - 32) * 2);
            offset = SAMPLES_BUF_SIZE - 512;
        }
    }
    s->samples_offset[ch] = offset;

    //    print_pow(s->sb_samples, 1152);
}

static void compute_scale_factors(unsigned char scale_code[SBLIMIT],
                                  unsigned char scale_factors[SBLIMIT][3],
                                  int sb_samples[3][12][SBLIMIT],
                                  int sblimit)
{
    int *p, vmax, v, n, i, j, k, code;
    int index, d1, d2;
    unsigned char *sf = &scale_factors[0][0];

    for(j=0;j<sblimit;j++) {
        for(i=0;i<3;i++) {
            /* find the max absolute value */
            p = &sb_samples[i][0][j];
            vmax = abs(*p);
            for(k=1;k<12;k++) {
                p += SBLIMIT;
                v = abs(*p);
                if (v > vmax)
                    vmax = v;
            }
            /* compute the scale factor index using log 2 computations */
            if (vmax > 0) {
                n = av_log2(vmax);
                /* n is the position of the MSB of vmax. now
                   use at most 2 compares to find the index */
                index = (21 - n) * 3 - 3;
                if (index >= 0) {
                    while (vmax <= scale_factor_table[index+1])
                        index++;
                } else {
                    index = 0; /* very unlikely case of overflow */
                }
            } else {