Skip to content
Snippets Groups Projects
aac.c 65.1 KiB
Newer Older
  • Learn to ignore specific revisions
  • /*
     * AAC decoder
     * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
     * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
     *
     * 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 libavcodec/aac.c
    
     * 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
     * N (in progress)      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)
     * N (planned)          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 "aac.h"
    #include "aactab.h"
    
    #include "mpeg4audio.h"
    
    
    #include <assert.h>
    #include <errno.h>
    #include <math.h>
    #include <string.h>
    
    
    union float754 {
        float f;
        uint32_t i;
    };
    
    static VLC vlc_scalefactors;
    static VLC vlc_spectral[11];
    
    
    
    static ChannelElement *get_che(AACContext *ac, int type, int elem_id)
    {
    
        if (ac->tag_che_map[type][elem_id]) {
            return ac->tag_che_map[type][elem_id];
        }
        if (ac->tags_mapped >= tags_per_config[ac->m4ac.chan_config]) {
            return NULL;
        }
        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[0] LFE[0]. If we seem to have
               encountered such a stream, transfer the LFE[0] element to SCE[1] */
            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;
    
    /**
     * 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 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)
    {
    
        AVCodecContext *avctx = ac->avccontext;
        int i, type, channels = 0;
    
        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++) {
                const int id = aac_channel_layout_map[channel_config - 1][i][1];
                type         = aac_channel_layout_map[channel_config - 1][i][0];
    
                if (!ac->che[type][id] && !(ac->che[type][id] = av_mallocz(sizeof(ChannelElement))))
                    return AVERROR(ENOMEM);
    
                if (type != TYPE_CCE) {
                    ac->output_data[channels++] = ac->che[type][id]->ch[0].ret;
                    if (type == TYPE_CPE)
                        ac->output_data[channels++] = ac->che[type][id]->ch[1].ret;
                }
            }
    
            memset(ac->tag_che_map, 0,       4 * MAX_ELEM_ID * sizeof(ac->che[0][0]));
            ac->tags_mapped = 0;
    
            avctx->channel_layout = aac_channel_layout[channel_config - 1];
        } else {
    
        /* 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 (che_pos[type][i]) {
                    if (!ac->che[type][i] && !(ac->che[type][i] = av_mallocz(sizeof(ChannelElement))))
    
                        return AVERROR(ENOMEM);
    
                    if (type != TYPE_CCE) {
    
                        ac->output_data[channels++] = ac->che[type][i]->ch[0].ret;
    
                        if (type == TYPE_CPE) {
    
                            ac->output_data[channels++] = ac->che[type][i]->ch[1].ret;
                        }
                    }
                } else
                    av_freep(&ac->che[type][i]);
            }
        }
    
    
            memcpy(ac->tag_che_map, ac->che, 4 * MAX_ELEM_ID * sizeof(ac->che[0][0]));
    
            ac->tags_mapped = 4 * MAX_ELEM_ID;
    
        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(AACContext *ac, 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;
    
    
        skip_bits(gb, 2);  // object_type
    
    
        if (ac->m4ac.sampling_index != sampling_index)
            av_log(ac->avccontext, 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 */
        skip_bits_long(gb, 8 * get_bits(gb, 8));
    
    /**
     * 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 int set_default_channel_config(AACContext *ac,
                                          enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
                                          int channel_config)
    
        if (channel_config < 1 || channel_config > 7) {
    
            av_log(ac->avccontext, 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.
     *
     * @return  Returns error status. 0 - OK, !0 - error
     */
    
    static int decode_ga_specific_config(AACContext *ac, 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(ac->avccontext, "960/120 MDCT window is", 1);
    
            return -1;
        }
    
        if (get_bits1(gb))       // dependsOnCoreCoder
            skip_bits(gb, 14);   // coreCoderDelay
        extension_flag = get_bits1(gb);
    
    
        if (ac->m4ac.object_type == AOT_AAC_SCALABLE ||
            ac->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(ac, new_che_pos, gb)))
    
                return ret;
        } else {
    
            if ((ret = set_default_channel_config(ac, new_che_pos, channel_config)))
    
        if ((ret = output_configure(ac, ac->che_pos, new_che_pos, channel_config)))
    
            return ret;
    
        if (extension_flag) {
            switch (ac->m4ac.object_type) {
    
            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   data        pointer to AVCodecContext extradata
     * @param   data_size   size of AVCCodecContext extradata
     *
     * @return  Returns error status. 0 - OK, !0 - error
     */
    
    static int decode_audio_specific_config(AACContext *ac, void *data,
                                            int data_size)
    {
    
        GetBitContext gb;
        int i;
    
        init_get_bits(&gb, data, data_size * 8);
    
    
        if ((i = ff_mpeg4audio_get_config(&ac->m4ac, data, data_size)) < 0)
    
        if (ac->m4ac.sampling_index > 12) {
    
            av_log(ac->avccontext, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index);
            return -1;
        }
    
        skip_bits_long(&gb, i);
    
        switch (ac->m4ac.object_type) {
    
        case AOT_AAC_LC:
            if (decode_ga_specific_config(ac, &gb, ac->m4ac.chan_config))
                return -1;
            break;
        default:
            av_log(ac->avccontext, AV_LOG_ERROR, "Audio object type %s%d is not supported.\n",
                   ac->m4ac.sbr == 1? "SBR+" : "", ac->m4ac.object_type);
            return -1;
        }
        return 0;
    }
    
    
    /**
     * 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 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)
    
    static av_cold int aac_decode_init(AVCodecContext *avccontext)
    {
        AACContext *ac = avccontext->priv_data;
    
        int i;
    
        ac->avccontext = avccontext;
    
    
            if (decode_audio_specific_config(ac, avccontext->extradata, avccontext->extradata_size))
    
                return -1;
            avccontext->sample_rate = ac->m4ac.sample_rate;
        } else if (avccontext->channels > 0) {
            ac->m4ac.sample_rate = avccontext->sample_rate;
        }
    
        avccontext->sample_fmt = SAMPLE_FMT_S16;
        avccontext->frame_size = 1024;
    
    
        AAC_INIT_VLC_STATIC( 0, 144);
        AAC_INIT_VLC_STATIC( 1, 114);
        AAC_INIT_VLC_STATIC( 2, 188);
        AAC_INIT_VLC_STATIC( 3, 180);
        AAC_INIT_VLC_STATIC( 4, 172);
        AAC_INIT_VLC_STATIC( 5, 140);
        AAC_INIT_VLC_STATIC( 6, 168);
        AAC_INIT_VLC_STATIC( 7, 114);
        AAC_INIT_VLC_STATIC( 8, 262);
        AAC_INIT_VLC_STATIC( 9, 248);
        AAC_INIT_VLC_STATIC(10, 384);
    
        dsputil_init(&ac->dsp, avccontext);
    
    
        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 == ff_float_to_int16_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.;
    
    #if !CONFIG_HARDCODED_TABLES
    
            ff_aac_pow2sf_tab[i] = pow(2, (i - 200) / 4.);
    
    #endif /* CONFIG_HARDCODED_TABLES */
    
    
        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_sine_window_init(ff_sine_1024, 1024);
        ff_sine_window_init(ff_sine_128, 128);
    
    
    /**
     * Skip data_stream_element; reference: table 4.10.
     */
    
    static void skip_data_stream_element(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);
        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->avccontext, 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->avccontext, 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->avccontext, AV_LOG_ERROR, "Prediction is not allowed in AAC-LC.\n");
                    memset(ics, 0, sizeof(IndividualChannelStream));
                    return -1;
                } else {
    
                    av_log_missing_feature(ac->avccontext, "Predictor bit set but LTP is", 1);
    
                    memset(ics, 0, sizeof(IndividualChannelStream));
                    return -1;
    
        if (ics->max_sfb > ics->num_swb) {
    
            av_log(ac->avccontext, 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) {
                uint8_t sect_len = k;
                int sect_len_incr;
                int sect_band_type = get_bits(gb, 4);
                if (sect_band_type == 12) {
                    av_log(ac->avccontext, AV_LOG_ERROR, "invalid band type\n");
                    return -1;
                }
    
                while ((sect_len_incr = get_bits(gb, bits)) == (1 << bits) - 1)
    
                    sect_len += sect_len_incr;
                sect_len += sect_len_incr;
                if (sect_len > ics->max_sfb) {
                    av_log(ac->avccontext, AV_LOG_ERROR,
    
                           "Number of bands (%d) exceeds limit (%d).\n",
                           sect_len, ics->max_sfb);
    
                for (; k < sect_len; k++) {
                    band_type        [idx]   = sect_band_type;
                    band_type_run_end[idx++] = sect_len;
                }
    
    /**
     * 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) {
    
                                   "%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) {
    
                                   "%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) {
    
                                   "%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) {
    
    Robert Swain's avatar
    Robert Swain committed
                        av_log(ac->avccontext, AV_LOG_ERROR, "TNS filter order %d is greater than maximum %d.",
                               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]));
        }
    }
    
    /**
     * 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, 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;
    
        static const float sign_lookup[] = { 1.0f, -1.0f };
    
    
        for (g = 0; g < ics->num_windows; g++)
    
            memset(coef + g * 128 + offsets[ics->max_sfb], 0, sizeof(float) * (c - offsets[ics->max_sfb]));
    
    
        for (g = 0; g < ics->num_window_groups; g++) {
            for (i = 0; i < ics->max_sfb; i++, idx++) {
                const int cur_band_type = band_type[idx];
                const int dim = cur_band_type >= FIRST_PAIR_BT ? 2 : 4;
                const int is_cb_unsigned = IS_CODEBOOK_UNSIGNED(cur_band_type);
                int group;
    
                if (cur_band_type == ZERO_BT || cur_band_type == INTENSITY_BT2 || cur_band_type == INTENSITY_BT) {
    
                    for (group = 0; group < ics->group_len[g]; group++) {
    
                        memset(coef + group * 128 + offsets[i], 0, (offsets[i + 1] - offsets[i]) * sizeof(float));
    
                } else if (cur_band_type == NOISE_BT) {
    
                    for (group = 0; group < ics->group_len[g]; group++) {
    
                        float band_energy;
    
                        float *cf = coef + group * 128 + offsets[i];
    
                        int len = offsets[i+1] - offsets[i];
    
    
                        for (k = 0; k < len; k++) {
    
                            ac->random_state  = lcg_random(ac->random_state);
    
                            cf[k] = ac->random_state;
    
                        band_energy = ac->dsp.scalarproduct_float(cf, cf, len);
    
                        ac->dsp.vector_fmul_scalar(cf, cf, scale, len);
    
                } else {
    
                    for (group = 0; group < ics->group_len[g]; group++) {
    
                        const float *vq[96];
                        const float **vqp = vq;
                        float *cf = coef + (group << 7) + offsets[i];
                        int len = offsets[i + 1] - offsets[i];
    
    
                        for (k = offsets[i]; k < offsets[i + 1]; k += dim) {
    
                            const int index = get_vlc2(gb, vlc_spectral[cur_band_type - 1].table, 6, 3);
                            const int coef_tmp_idx = (group << 7) + k;
                            const float *vq_ptr;
                            int j;
    
                            if (index >= ff_aac_spectral_sizes[cur_band_type - 1]) {
    
                                av_log(ac->avccontext, AV_LOG_ERROR,
    
                                       "Read beyond end of ff_aac_codebook_vectors[%d][]. index %d >= %d\n",
                                       cur_band_type - 1, index, ff_aac_spectral_sizes[cur_band_type - 1]);
    
                                return -1;
                            }
                            vq_ptr = &ff_aac_codebook_vectors[cur_band_type - 1][index * dim];
    
                            *vqp++ = vq_ptr;
    
                            if (is_cb_unsigned) {
    
                                if (vq_ptr[0])
                                    coef[coef_tmp_idx    ] = sign_lookup[get_bits1(gb)];
                                if (vq_ptr[1])
                                    coef[coef_tmp_idx + 1] = sign_lookup[get_bits1(gb)];
    
                                    if (vq_ptr[2])
                                        coef[coef_tmp_idx + 2] = sign_lookup[get_bits1(gb)];
                                    if (vq_ptr[3])
                                        coef[coef_tmp_idx + 3] = sign_lookup[get_bits1(gb)];
    
                                if (cur_band_type == ESC_BT) {
                                    for (j = 0; j < 2; j++) {
                                        if (vq_ptr[j] == 64.0f) {
                                            int n = 4;
                                            /* The total length of escape_sequence must be < 22 bits according
                                               to the specification (i.e. max is 11111111110xxxxxxxxxx). */
                                            while (get_bits1(gb) && n < 15) n++;
    
                                            if (n == 15) {
    
                                                av_log(ac->avccontext, AV_LOG_ERROR, "error in spectral data, ESC overflow\n");
                                                return -1;
                                            }
    
                                            n = (1 << n) + get_bits(gb, n);
    
                                            coef[coef_tmp_idx + j] *= cbrtf(n) * n;
    
                                        } else
    
                                            coef[coef_tmp_idx + j] *= vq_ptr[j];
                                    }
    
    
                        if (is_cb_unsigned && cur_band_type != ESC_BT) {
                            ac->dsp.vector_fmul_sv_scalar[dim>>2](
                                cf, cf, vq, sf[idx], len);
    
                        } else if (cur_band_type == ESC_BT) {
    
                            ac->dsp.vector_fmul_scalar(cf, cf, sf[idx], len);
                        } else {    /* !is_cb_unsigned */
                            ac->dsp.sv_fmul_scalar[dim>>2](cf, vq, sf[idx], len);
                        }
    
            coef += ics->group_len[g] << 7;
    
        }
    
        if (pulse_present) {
    
            for (i = 0; i < pulse->num_pulse; i++) {
                float co = coef_base[ pulse->pos[i] ];
                while (offsets[idx + 1] <= pulse->pos[i])
    
                    idx++;
                if (band_type[idx] != NOISE_BT && sf[idx]) {
    
    Robert Swain's avatar
    Robert Swain committed
                    float ico = -pulse->amp[i];
                    if (co) {
                        co /= sf[idx];
                        ico = co / sqrtf(sqrtf(fabsf(co))) + (co > 0 ? -ico : ico);
                    }
                    coef_base[ pulse->pos[i] ] = cbrtf(fabsf(ico)) * ico * sf[idx];
    
    static av_always_inline float flt16_round(float pf)
    {
    
        union float754 tmp;
        tmp.f = pf;
        tmp.i = (tmp.i + 0x00008000U) & 0xFFFF0000U;
        return tmp.f;
    
    static av_always_inline float flt16_even(float pf)
    {
    
        tmp.i = (tmp.i + 0x00007FFFU + (tmp.i & 0x00010000U >> 16)) & 0xFFFF0000U;
    
    static av_always_inline float flt16_trunc(float pf)
    {
    
        union float754 pun;
        pun.f = pf;
        pun.i &= 0xFFFF0000U;
        return pun.f;