view tta.c @ 4704:0cb02a723a62 libavcodec

utilize multi-stage AC-3 bit allocation. speeds up encoding by 25-30%

/*
 * TTA (The Lossless True Audio) decoder
 * Copyright (c) 2006 Alex Beregszaszi
 *
 * 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 St, Fifth Floor, Boston, MA  02110-1301  USA
 */

/**
 * @file tta.c
 * TTA (The Lossless True Audio) decoder
 * (www.true-audio.com or tta.corecodec.org)
 * @author Alex Beregszaszi
 *
 */

#define ALT_BITSTREAM_READER_LE
//#define DEBUG
#include <limits.h>
#include "avcodec.h"
#include "bitstream.h"

#define FORMAT_INT 1
#define FORMAT_FLOAT 3

typedef struct TTAContext {
    AVCodecContext *avctx;
    GetBitContext gb;

    int flags, channels, bps, is_float, data_length;
    int frame_length, last_frame_length, total_frames;

    int32_t *decode_buffer;
} TTAContext;

#if 0
static inline int shift_1(int i)
{
    if (i < 32)
        return 1 << i;
    else
        return 0x80000000; // 16 << 31
}

static inline int shift_16(int i)
{
    if (i < 28)
        return 16 << i;
    else
        return 0x80000000; // 16 << 27
}
#else
static const uint32_t shift_1[] = {
    0x00000001, 0x00000002, 0x00000004, 0x00000008,
    0x00000010, 0x00000020, 0x00000040, 0x00000080,
    0x00000100, 0x00000200, 0x00000400, 0x00000800,
    0x00001000, 0x00002000, 0x00004000, 0x00008000,
    0x00010000, 0x00020000, 0x00040000, 0x00080000,
    0x00100000, 0x00200000, 0x00400000, 0x00800000,
    0x01000000, 0x02000000, 0x04000000, 0x08000000,
    0x10000000, 0x20000000, 0x40000000, 0x80000000,
    0x80000000, 0x80000000, 0x80000000, 0x80000000,
    0x80000000, 0x80000000, 0x80000000, 0x80000000
};

static const uint32_t *shift_16 = shift_1 + 4;
#endif

#define MAX_ORDER 16
typedef struct TTAFilter {
    int32_t shift, round, error, mode;
    int32_t qm[MAX_ORDER];
    int32_t dx[MAX_ORDER];
    int32_t dl[MAX_ORDER];
} TTAFilter;

static int32_t ttafilter_configs[4][2] = {
    {10, 1},
    {9, 1},
    {10, 1},
    {12, 0}
};

static void ttafilter_init(TTAFilter *c, int32_t shift, int32_t mode) {
    memset(c, 0, sizeof(TTAFilter));
    c->shift = shift;
   c->round = shift_1[shift-1];
//    c->round = 1 << (shift - 1);
    c->mode = mode;
}

// FIXME: copy paste from original
static inline void memshl(register int32_t *a, register int32_t *b) {
    *a++ = *b++;
    *a++ = *b++;
    *a++ = *b++;
    *a++ = *b++;
    *a++ = *b++;
    *a++ = *b++;
    *a++ = *b++;
    *a = *b;
}

// FIXME: copy paste from original
// mode=1 encoder, mode=0 decoder
static inline void ttafilter_process(TTAFilter *c, int32_t *in, int32_t mode) {
    register int32_t *dl = c->dl, *qm = c->qm, *dx = c->dx, sum = c->round;

    if (!c->error) {
        sum += *dl++ * *qm, qm++;
        sum += *dl++ * *qm, qm++;
        sum += *dl++ * *qm, qm++;
        sum += *dl++ * *qm, qm++;
        sum += *dl++ * *qm, qm++;
        sum += *dl++ * *qm, qm++;
        sum += *dl++ * *qm, qm++;
        sum += *dl++ * *qm, qm++;
        dx += 8;
    } else if(c->error < 0) {
        sum += *dl++ * (*qm -= *dx++), qm++;
        sum += *dl++ * (*qm -= *dx++), qm++;
        sum += *dl++ * (*qm -= *dx++), qm++;
        sum += *dl++ * (*qm -= *dx++), qm++;
        sum += *dl++ * (*qm -= *dx++), qm++;
        sum += *dl++ * (*qm -= *dx++), qm++;
        sum += *dl++ * (*qm -= *dx++), qm++;
        sum += *dl++ * (*qm -= *dx++), qm++;
    } else {
        sum += *dl++ * (*qm += *dx++), qm++;
        sum += *dl++ * (*qm += *dx++), qm++;
        sum += *dl++ * (*qm += *dx++), qm++;
        sum += *dl++ * (*qm += *dx++), qm++;
        sum += *dl++ * (*qm += *dx++), qm++;
        sum += *dl++ * (*qm += *dx++), qm++;
        sum += *dl++ * (*qm += *dx++), qm++;
        sum += *dl++ * (*qm += *dx++), qm++;
    }

    *(dx-0) = ((*(dl-1) >> 30) | 1) << 2;
    *(dx-1) = ((*(dl-2) >> 30) | 1) << 1;
    *(dx-2) = ((*(dl-3) >> 30) | 1) << 1;
    *(dx-3) = ((*(dl-4) >> 30) | 1);

    // compress
    if (mode) {
        *dl = *in;
        *in -= (sum >> c->shift);
        c->error = *in;
    } else {
        c->error = *in;
        *in += (sum >> c->shift);
        *dl = *in;
    }

    if (c->mode) {
        *(dl-1) = *dl - *(dl-1);
        *(dl-2) = *(dl-1) - *(dl-2);
        *(dl-3) = *(dl-2) - *(dl-3);
    }

    memshl(c->dl, c->dl + 1);
    memshl(c->dx, c->dx + 1);
}

typedef struct TTARice {
    uint32_t k0, k1, sum0, sum1;
} TTARice;

static void rice_init(TTARice *c, uint32_t k0, uint32_t k1)
{
    c->k0 = k0;
    c->k1 = k1;
    c->sum0 = shift_16[k0];
    c->sum1 = shift_16[k1];
}

static int tta_get_unary(GetBitContext *gb)
{
    int ret = 0;

    // count ones
    while(get_bits1(gb))
        ret++;
    return ret;
}

static int tta_decode_init(AVCodecContext * avctx)
{
    TTAContext *s = avctx->priv_data;
    int i;

    s->avctx = avctx;

    // 30bytes includes a seektable with one frame
    if (avctx->extradata_size < 30)
        return -1;

    init_get_bits(&s->gb, avctx->extradata, avctx->extradata_size);
    if (show_bits_long(&s->gb, 32) == ff_get_fourcc("TTA1"))
    {
        /* signature */
        skip_bits(&s->gb, 32);
//        if (get_bits_long(&s->gb, 32) != bswap_32(ff_get_fourcc("TTA1"))) {
//            av_log(s->avctx, AV_LOG_ERROR, "Missing magic\n");
//            return -1;
//        }

        s->flags = get_bits(&s->gb, 16);
        if (s->flags != 1 && s->flags != 3)
        {
            av_log(s->avctx, AV_LOG_ERROR, "Invalid flags\n");
            return -1;
        }
        s->is_float = (s->flags == FORMAT_FLOAT);
        avctx->channels = s->channels = get_bits(&s->gb, 16);
        avctx->bits_per_sample = get_bits(&s->gb, 16);
        s->bps = (avctx->bits_per_sample + 7) / 8;
        avctx->sample_rate = get_bits_long(&s->gb, 32);
        if(avctx->sample_rate > 1000000){ //prevent FRAME_TIME * avctx->sample_rate from overflowing and sanity check
            av_log(avctx, AV_LOG_ERROR, "sample_rate too large\n");
            return -1;
        }
        s->data_length = get_bits_long(&s->gb, 32);
        skip_bits(&s->gb, 32); // CRC32 of header

        if (s->is_float)
        {
            avctx->sample_fmt = SAMPLE_FMT_FLT;
            av_log(s->avctx, AV_LOG_ERROR, "Unsupported sample format. Please contact the developers.\n");
            return -1;
        }
        else switch(s->bps) {
//            case 1: avctx->sample_fmt = SAMPLE_FMT_U8; break;
            case 2: avctx->sample_fmt = SAMPLE_FMT_S16; break;
//            case 3: avctx->sample_fmt = SAMPLE_FMT_S24; break;
            case 4: avctx->sample_fmt = SAMPLE_FMT_S32; break;
            default:
                av_log(s->avctx, AV_LOG_ERROR, "Invalid/unsupported sample format. Please contact the developers.\n");
                return -1;
        }

        // FIXME: horribly broken, but directly from reference source
#define FRAME_TIME 1.04489795918367346939
        s->frame_length = (int)(FRAME_TIME * avctx->sample_rate);

        s->last_frame_length = s->data_length % s->frame_length;
        s->total_frames = s->data_length / s->frame_length +
                        (s->last_frame_length ? 1 : 0);

        av_log(s->avctx, AV_LOG_DEBUG, "flags: %x chans: %d bps: %d rate: %d block: %d\n",
            s->flags, avctx->channels, avctx->bits_per_sample, avctx->sample_rate,
            avctx->block_align);
        av_log(s->avctx, AV_LOG_DEBUG, "data_length: %d frame_length: %d last: %d total: %d\n",
            s->data_length, s->frame_length, s->last_frame_length, s->total_frames);

        // FIXME: seek table
        for (i = 0; i < s->total_frames; i++)
            skip_bits(&s->gb, 32);
        skip_bits(&s->gb, 32); // CRC32 of seektable

        if(s->frame_length >= UINT_MAX / (s->channels * sizeof(int32_t))){
            av_log(avctx, AV_LOG_ERROR, "frame_length too large\n");
            return -1;
        }

        s->decode_buffer = av_mallocz(sizeof(int32_t)*s->frame_length*s->channels);
    } else {
        av_log(avctx, AV_LOG_ERROR, "Wrong extradata present\n");
        return -1;
    }

    return 0;
}

static int tta_decode_frame(AVCodecContext *avctx,
        void *data, int *data_size,
        uint8_t *buf, int buf_size)
{
    TTAContext *s = avctx->priv_data;
    int i;

    init_get_bits(&s->gb, buf, buf_size*8);
    {
        int32_t predictors[s->channels];
        TTAFilter filters[s->channels];
        TTARice rices[s->channels];
        int cur_chan = 0, framelen = s->frame_length;
        int32_t *p;

        // FIXME: seeking
        s->total_frames--;
        if (!s->total_frames && s->last_frame_length)
            framelen = s->last_frame_length;

        // init per channel states
        for (i = 0; i < s->channels; i++) {
            predictors[i] = 0;
            ttafilter_init(&(filters[i]), ttafilter_configs[s->bps-1][0], ttafilter_configs[s->bps-1][1]);
            rice_init(&(rices[i]), 10, 10);
        }

        for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++) {
            int32_t *predictor = &(predictors[cur_chan]);
            TTAFilter *filter = &(filters[cur_chan]);
            TTARice *rice = &(rices[cur_chan]);
            uint32_t unary, depth, k;
            int32_t value;

            unary = tta_get_unary(&s->gb);

            if (unary == 0) {
                depth = 0;
                k = rice->k0;
            } else {
                depth = 1;
                k = rice->k1;
                unary--;
            }

            if (k)
                value = (unary << k) + get_bits(&s->gb, k);
            else
                value = unary;

            // FIXME: copy paste from original
            switch (depth) {
            case 1:
                rice->sum1 += value - (rice->sum1 >> 4);
                if (rice->k1 > 0 && rice->sum1 < shift_16[rice->k1])
                    rice->k1--;
                else if(rice->sum1 > shift_16[rice->k1 + 1])
                    rice->k1++;
                value += shift_1[rice->k0];
            default:
                rice->sum0 += value - (rice->sum0 >> 4);
                if (rice->k0 > 0 && rice->sum0 < shift_16[rice->k0])
                    rice->k0--;
                else if(rice->sum0 > shift_16[rice->k0 + 1])
                    rice->k0++;
            }

            // extract coded value
#define UNFOLD(x) (((x)&1) ? (++(x)>>1) : (-(x)>>1))
            *p = UNFOLD(value);

            // run hybrid filter
            ttafilter_process(filter, p, 0);

            // fixed order prediction
#define PRED(x, k) (int32_t)((((uint64_t)x << k) - x) >> k)
            switch (s->bps) {
                case 1: *p += PRED(*predictor, 4); break;
                case 2:
                case 3: *p += PRED(*predictor, 5); break;
                case 4: *p += *predictor; break;
            }
            *predictor = *p;

#if 0
            // extract 32bit float from last two int samples
            if (s->is_float && ((p - data) & 1)) {
                uint32_t neg = *p & 0x80000000;
                uint32_t hi = *(p - 1);
                uint32_t lo = abs(*p) - 1;

                hi += (hi || lo) ? 0x3f80 : 0;
                // SWAP16: swap all the 16 bits
                *(p - 1) = (hi << 16) | SWAP16(lo) | neg;
            }
#endif

            /*if ((get_bits_count(&s->gb)+7)/8 > buf_size)
            {
                av_log(NULL, AV_LOG_INFO, "overread!!\n");
                break;
            }*/

            // flip channels
            if (cur_chan < (s->channels-1))
                cur_chan++;
            else {
                // decorrelate in case of stereo integer
                if (!s->is_float && (s->channels > 1)) {
                    int32_t *r = p - 1;
                    for (*p += *r / 2; r > p - s->channels; r--)
                        *r = *(r + 1) - *r;
                }
                cur_chan = 0;
            }
        }

        skip_bits(&s->gb, 32); // frame crc

        // convert to output buffer
        switch(s->bps) {
            case 2: {
                uint16_t *samples = data;
                for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++) {
//                    *samples++ = (unsigned char)*p;
//                    *samples++ = (unsigned char)(*p >> 8);
                    *samples++ = *p;
                }
                *data_size = (uint8_t *)samples - (uint8_t *)data;
                break;
            }
            default:
                av_log(s->avctx, AV_LOG_ERROR, "Error, only 16bit samples supported!\n");
        }
    }

//    return get_bits_count(&s->gb)+7)/8;
    return buf_size;
}

static int tta_decode_close(AVCodecContext *avctx) {
    TTAContext *s = avctx->priv_data;

    if (s->decode_buffer)
        av_free(s->decode_buffer);

    return 0;
}

AVCodec tta_decoder = {
    "tta",
    CODEC_TYPE_AUDIO,
    CODEC_ID_TTA,
    sizeof(TTAContext),
    tta_decode_init,
    NULL,
    tta_decode_close,
    tta_decode_frame,
};