Skip to content
Snippets Groups Projects
aacdec.c 82.6 KiB
Newer Older
/*
 * AAC decoder
 * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
 * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
 *
 * AAC LATM decoder
 * Copyright (c) 2008-2010 Paul Kendall <paul@kcbbs.gen.nz>
 * Copyright (c) 2010      Janne Grunau <janne-ffmpeg@jannau.net>
 *
 * 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
 */

/**
 * AAC decoder
 * @author Oded Shimon  ( ods15 ods15 dyndns org )
 * @author Maxim Gavrilov ( maxim.gavrilov gmail com )
 */

/*
 * supported tools
 *
 * Support?             Name
 * N (code in SoC repo) gain control
 * Y                    block switching
 * Y                    window shapes - standard
 * N                    window shapes - Low Delay
 * Y                    filterbank - standard
 * N (code in SoC repo) filterbank - Scalable Sample Rate
 * Y                    Temporal Noise Shaping
 * N (code in SoC repo) Long Term Prediction
 * Y                    intensity stereo
 * Y                    channel coupling
 * Y                    frequency domain prediction
 * Y                    Perceptual Noise Substitution
 * Y                    Mid/Side stereo
 * N                    Scalable Inverse AAC Quantization
 * N                    Frequency Selective Switch
 * N                    upsampling filter
 * Y                    quantization & coding - AAC
 * N                    quantization & coding - TwinVQ
 * N                    quantization & coding - BSAC
 * N                    AAC Error Resilience tools
 * N                    Error Resilience payload syntax
 * N                    Error Protection tool
 * N                    CELP
 * N                    Silence Compression
 * N                    HVXC
 * N                    HVXC 4kbits/s VR
 * N                    Structured Audio tools
 * N                    Structured Audio Sample Bank Format
 * N                    MIDI
 * N                    Harmonic and Individual Lines plus Noise
 * N                    Text-To-Speech Interface
Alex Converse's avatar
Alex Converse committed
 * Y                    Spectral Band Replication
 * Y (not in this code) Layer-1
 * Y (not in this code) Layer-2
 * Y (not in this code) Layer-3
 * N                    SinuSoidal Coding (Transient, Sinusoid, Noise)
 * Y                    Parametric Stereo
 * N                    Direct Stream Transfer
 *
 * Note: - HE AAC v1 comprises LC AAC with Spectral Band Replication.
 *       - HE AAC v2 comprises LC AAC with Spectral Band Replication and
           Parametric Stereo.
 */


#include "avcodec.h"
#include "get_bits.h"
#include "dsputil.h"
#include "fft.h"

#include "aac.h"
#include "aactab.h"
#include "cbrt_tablegen.h"
Alex Converse's avatar
Alex Converse committed
#include "sbr.h"
#include "aacsbr.h"
#include "mpeg4audio.h"
#include "aacadtsdec.h"

#include <assert.h>
#include <errno.h>
#include <math.h>
#include <string.h>

#if ARCH_ARM
#   include "arm/aac.h"
#endif

union float754 {
    float f;
    uint32_t i;
};
static VLC vlc_scalefactors;
static VLC vlc_spectral[11];

static const char overread_err[] = "Input buffer exhausted before END element found\n";

static ChannelElement *get_che(AACContext *ac, int type, int elem_id)
{
    // For PCE based channel configurations map the channels solely based on tags.
    if (!ac->m4ac.chan_config) {
        return ac->tag_che_map[type][elem_id];
    }
    // For indexed channel configurations map the channels solely based on position.
    switch (ac->m4ac.chan_config) {
    case 7:
        if (ac->tags_mapped == 3 && type == TYPE_CPE) {
            ac->tags_mapped++;
            return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][2];
        }
    case 6:
        /* Some streams incorrectly code 5.1 audio as SCE[0] CPE[0] CPE[1] SCE[1]
           instead of SCE[0] CPE[0] CPE[1] LFE[0]. If we seem to have
           encountered such a stream, transfer the LFE[0] element to the SCE[1]'s mapping */
        if (ac->tags_mapped == tags_per_config[ac->m4ac.chan_config] - 1 && (type == TYPE_LFE || type == TYPE_SCE)) {
            ac->tags_mapped++;
            return ac->tag_che_map[type][elem_id] = ac->che[TYPE_LFE][0];
        }
    case 5:
        if (ac->tags_mapped == 2 && type == TYPE_CPE) {
            ac->tags_mapped++;
            return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][1];
        }
    case 4:
        if (ac->tags_mapped == 2 && ac->m4ac.chan_config == 4 && type == TYPE_SCE) {
            ac->tags_mapped++;
            return ac->tag_che_map[TYPE_SCE][elem_id] = ac->che[TYPE_SCE][1];
        }
    case 3:
    case 2:
        if (ac->tags_mapped == (ac->m4ac.chan_config != 2) && type == TYPE_CPE) {
            ac->tags_mapped++;
            return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][0];
        } else if (ac->m4ac.chan_config == 2) {
        }
    case 1:
        if (!ac->tags_mapped && type == TYPE_SCE) {
            ac->tags_mapped++;
            return ac->tag_che_map[TYPE_SCE][elem_id] = ac->che[TYPE_SCE][0];
        }
    default:
        return NULL;
/**
 * Check for the channel element in the current channel position configuration.
 * If it exists, make sure the appropriate element is allocated and map the
 * channel order to match the internal FFmpeg channel layout.
 *
 * @param   che_pos current channel position configuration
 * @param   type channel element type
 * @param   id channel element id
 * @param   channels count of the number of channels in the configuration
 *
 * @return  Returns error status. 0 - OK, !0 - error
 */
static av_cold int che_configure(AACContext *ac,
                         enum ChannelPosition che_pos[4][MAX_ELEM_ID],
                         int type, int id,
                         int *channels)
{
    if (che_pos[type][id]) {
        if (!ac->che[type][id] && !(ac->che[type][id] = av_mallocz(sizeof(ChannelElement))))
            return AVERROR(ENOMEM);
Alex Converse's avatar
Alex Converse committed
        ff_aac_sbr_ctx_init(&ac->che[type][id]->sbr);
        if (type != TYPE_CCE) {
            ac->output_data[(*channels)++] = ac->che[type][id]->ch[0].ret;
            if (type == TYPE_CPE ||
                (type == TYPE_SCE && ac->m4ac.ps == 1)) {
                ac->output_data[(*channels)++] = ac->che[type][id]->ch[1].ret;
            }
        }
Alex Converse's avatar
Alex Converse committed
    } else {
        if (ac->che[type][id])
            ff_aac_sbr_ctx_close(&ac->che[type][id]->sbr);
/**
 * Configure output channel order based on the current program configuration element.
 *
 * @param   che_pos current channel position configuration
 * @param   new_che_pos New channel position configuration - we only do something if it differs from the current one.
 *
 * @return  Returns error status. 0 - OK, !0 - error
 */
static av_cold int output_configure(AACContext *ac,
                            enum ChannelPosition che_pos[4][MAX_ELEM_ID],
                            enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
                            int channel_config, enum OCStatus oc_type)
    AVCodecContext *avctx = ac->avctx;
    if (new_che_pos != che_pos)
    memcpy(che_pos, new_che_pos, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));

    if (channel_config) {
        for (i = 0; i < tags_per_config[channel_config]; i++) {
            if ((ret = che_configure(ac, che_pos,
                                     aac_channel_layout_map[channel_config - 1][i][0],
                                     aac_channel_layout_map[channel_config - 1][i][1],
                                     &channels)))
                return ret;
        }

        memset(ac->tag_che_map, 0,       4 * MAX_ELEM_ID * sizeof(ac->che[0][0]));

        avctx->channel_layout = aac_channel_layout[channel_config - 1];
    } else {
Robert Swain's avatar
Robert Swain committed
        /* Allocate or free elements depending on if they are in the
         * current program configuration.
         *
         * Set up default 1:1 output mapping.
         *
         * For a 5.1 stream the output order will be:
         *    [ Center ] [ Front Left ] [ Front Right ] [ LFE ] [ Surround Left ] [ Surround Right ]
         */

        for (i = 0; i < MAX_ELEM_ID; i++) {
            for (type = 0; type < 4; type++) {
                if ((ret = che_configure(ac, che_pos, type, i, &channels)))
                    return ret;
        memcpy(ac->tag_che_map, ac->che, 4 * MAX_ELEM_ID * sizeof(ac->che[0][0]));
    avctx->channels = channels;
/**
 * Decode an array of 4 bit element IDs, optionally interleaved with a stereo/mono switching bit.
 *
 * @param cpe_map Stereo (Channel Pair Element) map, NULL if stereo bit is not present.
 * @param sce_map mono (Single Channel Element) map
 * @param type speaker type/position for these channels
 */
static void decode_channel_map(enum ChannelPosition *cpe_map,
                               enum ChannelPosition *sce_map,
                               enum ChannelPosition type,
                               GetBitContext *gb, int n)
{
    while (n--) {
        enum ChannelPosition *map = cpe_map && get_bits1(gb) ? cpe_map : sce_map; // stereo or mono map
        map[get_bits(gb, 4)] = type;
    }
}

/**
 * Decode program configuration element; reference: table 4.2.
 *
 * @param   new_che_pos New channel position configuration - we only do something if it differs from the current one.
 *
 * @return  Returns error status. 0 - OK, !0 - error
 */
static int decode_pce(AVCodecContext *avctx, MPEG4AudioConfig *m4ac,
                      enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
                      GetBitContext *gb)
{
    int num_front, num_side, num_back, num_lfe, num_assoc_data, num_cc, sampling_index;
    int comment_len;

    skip_bits(gb, 2);  // object_type

    if (m4ac->sampling_index != sampling_index)
        av_log(avctx, AV_LOG_WARNING, "Sample rate index in program config element does not match the sample rate index configured by the container.\n");
    num_front       = get_bits(gb, 4);
    num_side        = get_bits(gb, 4);
    num_back        = get_bits(gb, 4);
    num_lfe         = get_bits(gb, 2);
    num_assoc_data  = get_bits(gb, 3);
    num_cc          = get_bits(gb, 4);

    if (get_bits1(gb))
        skip_bits(gb, 4); // mono_mixdown_tag
    if (get_bits1(gb))
        skip_bits(gb, 4); // stereo_mixdown_tag
    if (get_bits1(gb))
        skip_bits(gb, 3); // mixdown_coeff_index and pseudo_surround
    decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_FRONT, gb, num_front);
    decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_SIDE,  gb, num_side );
    decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_BACK,  gb, num_back );
    decode_channel_map(NULL,                  new_che_pos[TYPE_LFE], AAC_CHANNEL_LFE,   gb, num_lfe  );

    skip_bits_long(gb, 4 * num_assoc_data);

    decode_channel_map(new_che_pos[TYPE_CCE], new_che_pos[TYPE_CCE], AAC_CHANNEL_CC,    gb, num_cc   );

    align_get_bits(gb);

    /* comment field, first byte is length */
    comment_len = get_bits(gb, 8) * 8;
    if (get_bits_left(gb) < comment_len) {
        av_log(avctx, AV_LOG_ERROR, overread_err);
        return -1;
    }
    skip_bits_long(gb, comment_len);
/**
 * Set up channel positions based on a default channel configuration
 * as specified in table 1.17.
 *
 * @param   new_che_pos New channel position configuration - we only do something if it differs from the current one.
 *
 * @return  Returns error status. 0 - OK, !0 - error
 */
static av_cold int set_default_channel_config(AVCodecContext *avctx,
                                      enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
                                      int channel_config)
    if (channel_config < 1 || channel_config > 7) {
        av_log(avctx, AV_LOG_ERROR, "invalid default channel configuration (%d)\n",
               channel_config);
        return -1;
    }

    /* default channel configurations:
     *
     * 1ch : front center (mono)
     * 2ch : L + R (stereo)
     * 3ch : front center + L + R
     * 4ch : front center + L + R + back center
     * 5ch : front center + L + R + back stereo
     * 6ch : front center + L + R + back stereo + LFE
     * 7ch : front center + L + R + outer front left + outer front right + back stereo + LFE
     */

    if (channel_config != 2)
        new_che_pos[TYPE_SCE][0] = AAC_CHANNEL_FRONT; // front center (or mono)
    if (channel_config > 1)
        new_che_pos[TYPE_CPE][0] = AAC_CHANNEL_FRONT; // L + R (or stereo)
    if (channel_config == 4)
        new_che_pos[TYPE_SCE][1] = AAC_CHANNEL_BACK;  // back center
    if (channel_config > 4)
        new_che_pos[TYPE_CPE][(channel_config == 7) + 1]
        = AAC_CHANNEL_BACK;  // back stereo
    if (channel_config > 5)
        new_che_pos[TYPE_LFE][0] = AAC_CHANNEL_LFE;   // LFE
    if (channel_config == 7)
        new_che_pos[TYPE_CPE][1] = AAC_CHANNEL_FRONT; // outer front left + outer front right

    return 0;
}

/**
 * Decode GA "General Audio" specific configuration; reference: table 4.1.
 *
 * @param   ac          pointer to AACContext, may be null
 * @param   avctx       pointer to AVCCodecContext, used for logging
 *
 * @return  Returns error status. 0 - OK, !0 - error
 */
static int decode_ga_specific_config(AACContext *ac, AVCodecContext *avctx,
                                     GetBitContext *gb,
                                     int channel_config)
{
    enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
    int extension_flag, ret;

    if (get_bits1(gb)) { // frameLengthFlag
        av_log_missing_feature(avctx, "960/120 MDCT window is", 1);
        return -1;
    }

    if (get_bits1(gb))       // dependsOnCoreCoder
        skip_bits(gb, 14);   // coreCoderDelay
    extension_flag = get_bits1(gb);

    if (m4ac->object_type == AOT_AAC_SCALABLE ||
        m4ac->object_type == AOT_ER_AAC_SCALABLE)
        skip_bits(gb, 3);     // layerNr

    memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
    if (channel_config == 0) {
        skip_bits(gb, 4);  // element_instance_tag
        if ((ret = decode_pce(avctx, m4ac, new_che_pos, gb)))
            return ret;
    } else {
        if ((ret = set_default_channel_config(avctx, new_che_pos, channel_config)))
    if (ac && (ret = output_configure(ac, ac->che_pos, new_che_pos, channel_config, OC_GLOBAL_HDR)))
        return ret;

    if (extension_flag) {
        case AOT_ER_BSAC:
            skip_bits(gb, 5);    // numOfSubFrame
            skip_bits(gb, 11);   // layer_length
            break;
        case AOT_ER_AAC_LC:
        case AOT_ER_AAC_LTP:
        case AOT_ER_AAC_SCALABLE:
        case AOT_ER_AAC_LD:
            skip_bits(gb, 3);  /* aacSectionDataResilienceFlag
                                    * aacScalefactorDataResilienceFlag
                                    * aacSpectralDataResilienceFlag
                                    */
            break;
        }
        skip_bits1(gb);    // extensionFlag3 (TBD in version 3)
    }
    return 0;
}

/**
 * Decode audio specific configuration; reference: table 1.13.
 *
 * @param   ac          pointer to AACContext, may be null
 * @param   avctx       pointer to AVCCodecContext, used for logging
 * @param   m4ac        pointer to MPEG4AudioConfig, used for parsing
 * @param   data        pointer to AVCodecContext extradata
 * @param   data_size   size of AVCCodecContext extradata
 *
 * @return  Returns error status or number of consumed bits. <0 - error
static int decode_audio_specific_config(AACContext *ac,
                                        AVCodecContext *avctx,
                                        MPEG4AudioConfig *m4ac,
                                        const uint8_t *data, int data_size)
    GetBitContext gb;
    int i;

    init_get_bits(&gb, data, data_size * 8);

    if ((i = ff_mpeg4audio_get_config(m4ac, data, data_size)) < 0)
        av_log(avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", m4ac->sampling_index);
    if (m4ac->sbr == 1 && m4ac->ps == -1)
        m4ac->ps = 1;

    skip_bits_long(&gb, i);

    case AOT_AAC_LC:
        if (decode_ga_specific_config(ac, avctx, &gb, m4ac, m4ac->chan_config))
            return -1;
        break;
    default:
        av_log(avctx, AV_LOG_ERROR, "Audio object type %s%d is not supported.\n",
               m4ac->sbr == 1? "SBR+" : "", m4ac->object_type);
    return get_bits_count(&gb);
/**
 * linear congruential pseudorandom number generator
 *
 * @param   previous_val    pointer to the current state of the generator
 *
 * @return  Returns a 32-bit pseudorandom integer
 */
static av_always_inline int lcg_random(int previous_val)
{
    return previous_val * 1664525 + 1013904223;
}

static av_always_inline void reset_predict_state(PredictorState *ps)
{
    ps->r0   = 0.0f;
    ps->r1   = 0.0f;
    ps->cor0 = 0.0f;
    ps->cor1 = 0.0f;
    ps->var0 = 1.0f;
    ps->var1 = 1.0f;
}

static void reset_all_predictors(PredictorState *ps)
{
    int i;
    for (i = 0; i < MAX_PREDICTORS; i++)
        reset_predict_state(&ps[i]);
}

static void reset_predictor_group(PredictorState *ps, int group_num)
{
    for (i = group_num - 1; i < MAX_PREDICTORS; i += 30)
#define AAC_INIT_VLC_STATIC(num, size) \
    INIT_VLC_STATIC(&vlc_spectral[num], 8, ff_aac_spectral_sizes[num], \
         ff_aac_spectral_bits[num], sizeof( ff_aac_spectral_bits[num][0]), sizeof( ff_aac_spectral_bits[num][0]), \
        ff_aac_spectral_codes[num], sizeof(ff_aac_spectral_codes[num][0]), sizeof(ff_aac_spectral_codes[num][0]), \
        size);

static av_cold int aac_decode_init(AVCodecContext *avctx)
    AACContext *ac = avctx->priv_data;
    ac->avctx = avctx;
    ac->m4ac.sample_rate = avctx->sample_rate;
    if (avctx->extradata_size > 0) {
        if (decode_audio_specific_config(ac, ac->avctx, &ac->m4ac,
                                         avctx->extradata,
                                         avctx->extradata_size) < 0)
    avctx->sample_fmt = SAMPLE_FMT_S16;
    AAC_INIT_VLC_STATIC( 0, 304);
    AAC_INIT_VLC_STATIC( 1, 270);
    AAC_INIT_VLC_STATIC( 2, 550);
    AAC_INIT_VLC_STATIC( 3, 300);
    AAC_INIT_VLC_STATIC( 4, 328);
    AAC_INIT_VLC_STATIC( 5, 294);
    AAC_INIT_VLC_STATIC( 6, 306);
    AAC_INIT_VLC_STATIC( 7, 268);
    AAC_INIT_VLC_STATIC( 8, 510);
    AAC_INIT_VLC_STATIC( 9, 366);
    AAC_INIT_VLC_STATIC(10, 462);
Alex Converse's avatar
Alex Converse committed
    ff_aac_sbr_init();

    dsputil_init(&ac->dsp, avctx);
    ac->random_state = 0x1f2e3d4c;

    // -1024 - Compensate wrong IMDCT method.
    // 32768 - Required to scale values to the correct range for the bias method
    //         for float to int16 conversion.

    if (ac->dsp.float_to_int16_interleave == ff_float_to_int16_interleave_c) {
        ac->add_bias  = 385.0f;
        ac->sf_scale  = 1. / (-1024. * 32768.);
        ac->sf_offset = 0;
    } else {
        ac->add_bias  = 0.0f;
        ac->sf_scale  = 1. / -1024.;
    ff_aac_tableinit();
    INIT_VLC_STATIC(&vlc_scalefactors,7,FF_ARRAY_ELEMS(ff_aac_scalefactor_code),
                    ff_aac_scalefactor_bits, sizeof(ff_aac_scalefactor_bits[0]), sizeof(ff_aac_scalefactor_bits[0]),
                    ff_aac_scalefactor_code, sizeof(ff_aac_scalefactor_code[0]), sizeof(ff_aac_scalefactor_code[0]),
                    352);
    ff_mdct_init(&ac->mdct, 11, 1, 1.0);
    ff_mdct_init(&ac->mdct_small, 8, 1, 1.0);
    // window initialization
    ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
    ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
    ff_init_ff_sine_windows(10);
    ff_init_ff_sine_windows( 7);
/**
 * Skip data_stream_element; reference: table 4.10.
 */
static int skip_data_stream_element(AACContext *ac, GetBitContext *gb)
    int byte_align = get_bits1(gb);
    int count = get_bits(gb, 8);
    if (count == 255)
        count += get_bits(gb, 8);
    if (byte_align)
        align_get_bits(gb);

    if (get_bits_left(gb) < 8 * count) {
        av_log(ac->avctx, AV_LOG_ERROR, overread_err);
    skip_bits_long(gb, 8 * count);
static int decode_prediction(AACContext *ac, IndividualChannelStream *ics,
                             GetBitContext *gb)
{
    int sfb;
    if (get_bits1(gb)) {
        ics->predictor_reset_group = get_bits(gb, 5);
        if (ics->predictor_reset_group == 0 || ics->predictor_reset_group > 30) {
            av_log(ac->avctx, AV_LOG_ERROR, "Invalid Predictor Reset Group.\n");
            return -1;
        }
    }
    for (sfb = 0; sfb < FFMIN(ics->max_sfb, ff_aac_pred_sfb_max[ac->m4ac.sampling_index]); sfb++) {
        ics->prediction_used[sfb] = get_bits1(gb);
    }
    return 0;
}

/**
 * Decode Individual Channel Stream info; reference: table 4.6.
 *
 * @param   common_window   Channels have independent [0], or shared [1], Individual Channel Stream information.
 */
static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics,
                           GetBitContext *gb, int common_window)
{
    if (get_bits1(gb)) {
        av_log(ac->avctx, AV_LOG_ERROR, "Reserved bit set.\n");
        memset(ics, 0, sizeof(IndividualChannelStream));
        return -1;
    }
    ics->window_sequence[1] = ics->window_sequence[0];
    ics->window_sequence[0] = get_bits(gb, 2);
    ics->use_kb_window[1]   = ics->use_kb_window[0];
    ics->use_kb_window[0]   = get_bits1(gb);
    ics->num_window_groups  = 1;
    ics->group_len[0]       = 1;
    if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
        int i;
        ics->max_sfb = get_bits(gb, 4);
        for (i = 0; i < 7; i++) {
            if (get_bits1(gb)) {
                ics->group_len[ics->num_window_groups - 1]++;
            } else {
                ics->num_window_groups++;
                ics->group_len[ics->num_window_groups - 1] = 1;
        ics->num_windows       = 8;
        ics->swb_offset        =    ff_swb_offset_128[ac->m4ac.sampling_index];
        ics->num_swb           =   ff_aac_num_swb_128[ac->m4ac.sampling_index];
        ics->tns_max_bands     = ff_tns_max_bands_128[ac->m4ac.sampling_index];
    } else {
        ics->max_sfb               = get_bits(gb, 6);
        ics->num_windows           = 1;
        ics->swb_offset            =    ff_swb_offset_1024[ac->m4ac.sampling_index];
        ics->num_swb               =   ff_aac_num_swb_1024[ac->m4ac.sampling_index];
        ics->tns_max_bands         = ff_tns_max_bands_1024[ac->m4ac.sampling_index];
        ics->predictor_present     = get_bits1(gb);
        ics->predictor_reset_group = 0;
        if (ics->predictor_present) {
            if (ac->m4ac.object_type == AOT_AAC_MAIN) {
                if (decode_prediction(ac, ics, gb)) {
                    memset(ics, 0, sizeof(IndividualChannelStream));
                    return -1;
                }
            } else if (ac->m4ac.object_type == AOT_AAC_LC) {
                av_log(ac->avctx, AV_LOG_ERROR, "Prediction is not allowed in AAC-LC.\n");
                memset(ics, 0, sizeof(IndividualChannelStream));
                return -1;
            } else {
                av_log_missing_feature(ac->avctx, "Predictor bit set but LTP is", 1);
                memset(ics, 0, sizeof(IndividualChannelStream));
                return -1;
    if (ics->max_sfb > ics->num_swb) {
        av_log(ac->avctx, AV_LOG_ERROR,
               "Number of scalefactor bands in group (%d) exceeds limit (%d).\n",
               ics->max_sfb, ics->num_swb);
        memset(ics, 0, sizeof(IndividualChannelStream));
        return -1;
    }

    return 0;
}

/**
 * Decode band types (section_data payload); reference: table 4.46.
 *
 * @param   band_type           array of the used band type
 * @param   band_type_run_end   array of the last scalefactor band of a band type run
 *
 * @return  Returns error status. 0 - OK, !0 - error
 */
static int decode_band_types(AACContext *ac, enum BandType band_type[120],
                             int band_type_run_end[120], GetBitContext *gb,
                             IndividualChannelStream *ics)
{
    int g, idx = 0;
    const int bits = (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) ? 3 : 5;
    for (g = 0; g < ics->num_window_groups; g++) {
        int k = 0;
        while (k < ics->max_sfb) {
            int sect_len_incr;
            int sect_band_type = get_bits(gb, 4);
            if (sect_band_type == 12) {
                av_log(ac->avctx, AV_LOG_ERROR, "invalid band type\n");
            while ((sect_len_incr = get_bits(gb, bits)) == (1 << bits) - 1)
                sect_end += sect_len_incr;
            sect_end += sect_len_incr;
                av_log(ac->avctx, AV_LOG_ERROR, overread_err);
                av_log(ac->avctx, AV_LOG_ERROR,
                       "Number of bands (%d) exceeds limit (%d).\n",
                band_type        [idx]   = sect_band_type;
/**
 * Decode scalefactors; reference: table 4.47.
 *
 * @param   global_gain         first scalefactor value as scalefactors are differentially coded
 * @param   band_type           array of the used band type
 * @param   band_type_run_end   array of the last scalefactor band of a band type run
 * @param   sf                  array of scalefactors or intensity stereo positions
 *
 * @return  Returns error status. 0 - OK, !0 - error
 */
static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb,
                               unsigned int global_gain,
                               IndividualChannelStream *ics,
                               enum BandType band_type[120],
                               int band_type_run_end[120])
{
    const int sf_offset = ac->sf_offset + (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE ? 12 : 0);
    int g, i, idx = 0;
    int offset[3] = { global_gain, global_gain - 90, 100 };
    int noise_flag = 1;
    static const char *sf_str[3] = { "Global gain", "Noise gain", "Intensity stereo position" };
    for (g = 0; g < ics->num_window_groups; g++) {
        for (i = 0; i < ics->max_sfb;) {
            int run_end = band_type_run_end[idx];
            if (band_type[idx] == ZERO_BT) {
                for (; i < run_end; i++, idx++)
            } else if ((band_type[idx] == INTENSITY_BT) || (band_type[idx] == INTENSITY_BT2)) {
                for (; i < run_end; i++, idx++) {
                    offset[2] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
                    if (offset[2] > 255U) {
                        av_log(ac->avctx, AV_LOG_ERROR,
                               "%s (%d) out of range.\n", sf_str[2], offset[2]);
                    sf[idx] = ff_aac_pow2sf_tab[-offset[2] + 300];
            } else if (band_type[idx] == NOISE_BT) {
                for (; i < run_end; i++, idx++) {
                    if (noise_flag-- > 0)
                        offset[1] += get_bits(gb, 9) - 256;
                    else
                        offset[1] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
                    if (offset[1] > 255U) {
                        av_log(ac->avctx, AV_LOG_ERROR,
                               "%s (%d) out of range.\n", sf_str[1], offset[1]);
                    sf[idx] = -ff_aac_pow2sf_tab[offset[1] + sf_offset + 100];
            } else {
                for (; i < run_end; i++, idx++) {
                    offset[0] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
                    if (offset[0] > 255U) {
                        av_log(ac->avctx, AV_LOG_ERROR,
                               "%s (%d) out of range.\n", sf_str[0], offset[0]);
                        return -1;
                    }
                    sf[idx] = -ff_aac_pow2sf_tab[ offset[0] + sf_offset];
                }
            }
        }
    }
    return 0;
}

/**
 * Decode pulse data; reference: table 4.7.
 */
static int decode_pulses(Pulse *pulse, GetBitContext *gb,
                         const uint16_t *swb_offset, int num_swb)
{
    pulse->num_pulse = get_bits(gb, 2) + 1;
    pulse_swb        = get_bits(gb, 6);
    if (pulse_swb >= num_swb)
        return -1;
    pulse->pos[0]    = swb_offset[pulse_swb];
    pulse->amp[0]    = get_bits(gb, 4);
    for (i = 1; i < pulse->num_pulse; i++) {
        pulse->pos[i] = get_bits(gb, 5) + pulse->pos[i - 1];
        pulse->amp[i] = get_bits(gb, 4);
/**
 * Decode Temporal Noise Shaping data; reference: table 4.48.
 *
 * @return  Returns error status. 0 - OK, !0 - error
 */
static int decode_tns(AACContext *ac, TemporalNoiseShaping *tns,
                      GetBitContext *gb, const IndividualChannelStream *ics)
{
    int w, filt, i, coef_len, coef_res, coef_compress;
    const int is8 = ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE;
    const int tns_max_order = is8 ? 7 : ac->m4ac.object_type == AOT_AAC_MAIN ? 20 : 12;
    for (w = 0; w < ics->num_windows; w++) {
        if ((tns->n_filt[w] = get_bits(gb, 2 - is8))) {
            coef_res = get_bits1(gb);

Robert Swain's avatar
Robert Swain committed
            for (filt = 0; filt < tns->n_filt[w]; filt++) {
                int tmp2_idx;
                tns->length[w][filt] = get_bits(gb, 6 - 2 * is8);
                if ((tns->order[w][filt] = get_bits(gb, 5 - 2 * is8)) > tns_max_order) {
                    av_log(ac->avctx, AV_LOG_ERROR, "TNS filter order %d is greater than maximum %d.\n",
Robert Swain's avatar
Robert Swain committed
                           tns->order[w][filt], tns_max_order);
                    tns->order[w][filt] = 0;
                    return -1;
                }
Robert Swain's avatar
Robert Swain committed
                    tns->direction[w][filt] = get_bits1(gb);
                    coef_compress = get_bits1(gb);
                    coef_len = coef_res + 3 - coef_compress;
                    tmp2_idx = 2 * coef_compress + coef_res;
Robert Swain's avatar
Robert Swain committed
                    for (i = 0; i < tns->order[w][filt]; i++)
                        tns->coef[w][filt][i] = tns_tmp2_map[tmp2_idx][get_bits(gb, coef_len)];
/**
 * Decode Mid/Side data; reference: table 4.54.
 *
 * @param   ms_present  Indicates mid/side stereo presence. [0] mask is all 0s;
 *                      [1] mask is decoded from bitstream; [2] mask is all 1s;
 *                      [3] reserved for scalable AAC
 */
static void decode_mid_side_stereo(ChannelElement *cpe, GetBitContext *gb,
                                   int ms_present)
{
    int idx;
    if (ms_present == 1) {
        for (idx = 0; idx < cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb; idx++)
            cpe->ms_mask[idx] = get_bits1(gb);
    } else if (ms_present == 2) {
        memset(cpe->ms_mask, 1, cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb * sizeof(cpe->ms_mask[0]));
    }
}
static inline float *VMUL2(float *dst, const float *v, unsigned idx,
                           const float *scale)
{
    float s = *scale;
    *dst++ = v[idx    & 15] * s;
    *dst++ = v[idx>>4 & 15] * s;
    return dst;
}
static inline float *VMUL4(float *dst, const float *v, unsigned idx,
                           const float *scale)
{
    float s = *scale;
    *dst++ = v[idx    & 3] * s;
    *dst++ = v[idx>>2 & 3] * s;
    *dst++ = v[idx>>4 & 3] * s;
    *dst++ = v[idx>>6 & 3] * s;
    return dst;
}
#ifndef VMUL2S
static inline float *VMUL2S(float *dst, const float *v, unsigned idx,
                            unsigned sign, const float *scale)
{
    union float754 s0, s1;

    s0.f = s1.f = *scale;
    s0.i ^= sign >> 1 << 31;
    s1.i ^= sign      << 31;

    *dst++ = v[idx    & 15] * s0.f;
    *dst++ = v[idx>>4 & 15] * s1.f;

    return dst;
}
#ifndef VMUL4S
static inline float *VMUL4S(float *dst, const float *v, unsigned idx,
                            unsigned sign, const float *scale)
{
    unsigned nz = idx >> 12;
    union float754 s = { .f = *scale };
    union float754 t;

    t.i = s.i ^ (sign & 1<<31);
    *dst++ = v[idx    & 3] * t.f;

    sign <<= nz & 1; nz >>= 1;
    t.i = s.i ^ (sign & 1<<31);
    *dst++ = v[idx>>2 & 3] * t.f;

    sign <<= nz & 1; nz >>= 1;
    t.i = s.i ^ (sign & 1<<31);
    *dst++ = v[idx>>4 & 3] * t.f;

    sign <<= nz & 1; nz >>= 1;
    t.i = s.i ^ (sign & 1<<31);
    *dst++ = v[idx>>6 & 3] * t.f;

    return dst;
}
/**
 * Decode spectral data; reference: table 4.50.
 * Dequantize and scale spectral data; reference: 4.6.3.3.
 *
 * @param   coef            array of dequantized, scaled spectral data
 * @param   sf              array of scalefactors or intensity stereo positions
 * @param   pulse_present   set if pulses are present
 * @param   pulse           pointer to pulse data struct
 * @param   band_type       array of the used band type
 *
 * @return  Returns error status. 0 - OK, !0 - error
 */
static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024],
                                       GetBitContext *gb, const float sf[120],
                                       int pulse_present, const Pulse *pulse,
                                       const IndividualChannelStream *ics,
                                       enum BandType band_type[120])
{
    int i, k, g, idx = 0;
    const int c = 1024 / ics->num_windows;
    const uint16_t *offsets = ics->swb_offset;
    float *coef_base = coef;

    for (g = 0; g < ics->num_windows; g++)
        memset(coef + g * 128 + offsets[ics->max_sfb], 0, sizeof(float) * (c - offsets[ics->max_sfb]));