/src/opus/silk/dec_API.c

Source (jump to first uncovered line)
/***********************************************************************
Copyright (c) 2006-2011, Skype Limited. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of Internet Society, IETF or IETF Trust, nor the
names of specific contributors, may be used to endorse or promote
products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
***********************************************************************/

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include "API.h"
#include "main.h"
#include "stack_alloc.h"
#include "os_support.h"

#ifdef ENABLE_OSCE
#include "osce.h"
#include "osce_structs.h"
#endif

/************************/
/* Decoder Super Struct */
/************************/
typedef struct {
    silk_decoder_state          channel_state[ DECODER_NUM_CHANNELS ];
    stereo_dec_state                sStereo;
    opus_int                         nChannelsAPI;
    opus_int                         nChannelsInternal;
    opus_int                         prev_decode_only_middle;
#ifdef ENABLE_OSCE
    OSCEModel                        osce_model;
#endif
} silk_decoder;

/*********************/
/* Decoder functions */
/*********************/



opus_int silk_LoadOSCEModels(void *decState, const unsigned char *data, int len)
{
#ifdef ENABLE_OSCE
    opus_int ret = SILK_NO_ERROR;

    ret = osce_load_models(&((silk_decoder *)decState)->osce_model, data, len);
    ((silk_decoder *)decState)->osce_model.loaded = (ret == 0);
    return ret;
#else
    (void) decState;
    (void) data;
    (void) len;
    return SILK_NO_ERROR;
#endif
}

opus_int silk_Get_Decoder_Size(                         /* O    Returns error code                              */
    opus_int                        *decSizeBytes       /* O    Number of bytes in SILK decoder state           */
)
{
    opus_int ret = SILK_NO_ERROR;

    *decSizeBytes = sizeof( silk_decoder );

    return ret;
}

/* Reset decoder state */
opus_int silk_ResetDecoder(                              /* O    Returns error code                              */
    void                            *decState           /* I/O  State                                           */
)
{
    opus_int n, ret = SILK_NO_ERROR;
    silk_decoder_state *channel_state = ((silk_decoder *)decState)->channel_state;

    for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) {
        ret  = silk_reset_decoder( &channel_state[ n ] );
    }
    silk_memset(&((silk_decoder *)decState)->sStereo, 0, sizeof(((silk_decoder *)decState)->sStereo));
    /* Not strictly needed, but it's cleaner that way */
    ((silk_decoder *)decState)->prev_decode_only_middle = 0;

    return ret;
}


opus_int silk_InitDecoder(                              /* O    Returns error code                              */
    void                            *decState           /* I/O  State                                           */
)
{
    opus_int n, ret = SILK_NO_ERROR;
    silk_decoder_state *channel_state = ((silk_decoder *)decState)->channel_state;
#ifdef ENABLE_OSCE
    ((silk_decoder *)decState)->osce_model.loaded = 0;
#endif
#ifndef USE_WEIGHTS_FILE
    /* load osce models */
    silk_LoadOSCEModels(decState, NULL, 0);
#endif

    for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) {
        ret  = silk_init_decoder( &channel_state[ n ] );
    }
    silk_memset(&((silk_decoder *)decState)->sStereo, 0, sizeof(((silk_decoder *)decState)->sStereo));
    /* Not strictly needed, but it's cleaner that way */
    ((silk_decoder *)decState)->prev_decode_only_middle = 0;

    return ret;
}

/* Decode a frame */
opus_int silk_Decode(                                   /* O    Returns error code                              */
    void*                           decState,           /* I/O  State                                           */
    silk_DecControlStruct*          decControl,         /* I/O  Control Structure                               */
    opus_int                        lostFlag,           /* I    0: no loss, 1 loss, 2 decode fec                */
    opus_int                        newPacketFlag,      /* I    Indicates first decoder call for this packet    */
    ec_dec                          *psRangeDec,        /* I/O  Compressor data structure                       */
    opus_int16                      *samplesOut,        /* O    Decoded output speech vector                    */
    opus_int32                      *nSamplesOut,       /* O    Number of samples decoded                       */
#ifdef ENABLE_DEEP_PLC
    LPCNetPLCState                  *lpcnet,
#endif
    int                             arch                /* I    Run-time architecture                           */
)
{
    opus_int   i, n, decode_only_middle = 0, ret = SILK_NO_ERROR;
    opus_int32 nSamplesOutDec, LBRR_symbol;
    opus_int16 *samplesOut1_tmp[ 2 ];
    VARDECL( opus_int16, samplesOut1_tmp_storage1 );
    VARDECL( opus_int16, samplesOut1_tmp_storage2 );
    VARDECL( opus_int16, samplesOut2_tmp );
    opus_int32 MS_pred_Q13[ 2 ] = { 0 };
    opus_int16 *resample_out_ptr;
    silk_decoder *psDec = ( silk_decoder * )decState;
    silk_decoder_state *channel_state = psDec->channel_state;
    opus_int has_side;
    opus_int stereo_to_mono;
    int delay_stack_alloc;
    SAVE_STACK;

    celt_assert( decControl->nChannelsInternal == 1 || decControl->nChannelsInternal == 2 );

    /**********************************/
    /* Test if first frame in payload */
    /**********************************/
    if( newPacketFlag ) {
        for( n = 0; n < decControl->nChannelsInternal; n++ ) {
            channel_state[ n ].nFramesDecoded = 0;  /* Used to count frames in packet */
        }
    }

    /* If Mono -> Stereo transition in bitstream: init state of second channel */
    if( decControl->nChannelsInternal > psDec->nChannelsInternal ) {
        ret += silk_init_decoder( &channel_state[ 1 ] );
    }

    stereo_to_mono = decControl->nChannelsInternal == 1 && psDec->nChannelsInternal == 2 &&
                     ( decControl->internalSampleRate == 1000*channel_state[ 0 ].fs_kHz );

    if( channel_state[ 0 ].nFramesDecoded == 0 ) {
        for( n = 0; n < decControl->nChannelsInternal; n++ ) {
            opus_int fs_kHz_dec;
            if( decControl->payloadSize_ms == 0 ) {
                /* Assuming packet loss, use 10 ms */
                channel_state[ n ].nFramesPerPacket = 1;
                channel_state[ n ].nb_subfr = 2;
            } else if( decControl->payloadSize_ms == 10 ) {
                channel_state[ n ].nFramesPerPacket = 1;
                channel_state[ n ].nb_subfr = 2;
            } else if( decControl->payloadSize_ms == 20 ) {
                channel_state[ n ].nFramesPerPacket = 1;
                channel_state[ n ].nb_subfr = 4;
            } else if( decControl->payloadSize_ms == 40 ) {
                channel_state[ n ].nFramesPerPacket = 2;
                channel_state[ n ].nb_subfr = 4;
            } else if( decControl->payloadSize_ms == 60 ) {
                channel_state[ n ].nFramesPerPacket = 3;
                channel_state[ n ].nb_subfr = 4;
            } else {
                celt_assert( 0 );
                RESTORE_STACK;
                return SILK_DEC_INVALID_FRAME_SIZE;
            }
            fs_kHz_dec = ( decControl->internalSampleRate >> 10 ) + 1;
            if( fs_kHz_dec != 8 && fs_kHz_dec != 12 && fs_kHz_dec != 16 ) {
                celt_assert( 0 );
                RESTORE_STACK;
                return SILK_DEC_INVALID_SAMPLING_FREQUENCY;
            }
            ret += silk_decoder_set_fs( &channel_state[ n ], fs_kHz_dec, decControl->API_sampleRate );
        }
    }

    if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 && ( psDec->nChannelsAPI == 1 || psDec->nChannelsInternal == 1 ) ) {
        silk_memset( psDec->sStereo.pred_prev_Q13, 0, sizeof( psDec->sStereo.pred_prev_Q13 ) );
        silk_memset( psDec->sStereo.sSide, 0, sizeof( psDec->sStereo.sSide ) );
        silk_memcpy( &channel_state[ 1 ].resampler_state, &channel_state[ 0 ].resampler_state, sizeof( silk_resampler_state_struct ) );
    }
    psDec->nChannelsAPI      = decControl->nChannelsAPI;
    psDec->nChannelsInternal = decControl->nChannelsInternal;

    if( decControl->API_sampleRate > (opus_int32)MAX_API_FS_KHZ * 1000 || decControl->API_sampleRate < 8000 ) {
        ret = SILK_DEC_INVALID_SAMPLING_FREQUENCY;
        RESTORE_STACK;
        return( ret );
    }

    if( lostFlag != FLAG_PACKET_LOST && channel_state[ 0 ].nFramesDecoded == 0 ) {
        /* First decoder call for this payload */
        /* Decode VAD flags and LBRR flag */
        for( n = 0; n < decControl->nChannelsInternal; n++ ) {
            for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) {
                channel_state[ n ].VAD_flags[ i ] = ec_dec_bit_logp(psRangeDec, 1);
            }
            channel_state[ n ].LBRR_flag = ec_dec_bit_logp(psRangeDec, 1);
        }
        /* Decode LBRR flags */
        for( n = 0; n < decControl->nChannelsInternal; n++ ) {
            silk_memset( channel_state[ n ].LBRR_flags, 0, sizeof( channel_state[ n ].LBRR_flags ) );
            if( channel_state[ n ].LBRR_flag ) {
                if( channel_state[ n ].nFramesPerPacket == 1 ) {
                    channel_state[ n ].LBRR_flags[ 0 ] = 1;
                } else {
                    LBRR_symbol = ec_dec_icdf( psRangeDec, silk_LBRR_flags_iCDF_ptr[ channel_state[ n ].nFramesPerPacket - 2 ], 8 ) + 1;
                    for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) {
                        channel_state[ n ].LBRR_flags[ i ] = silk_RSHIFT( LBRR_symbol, i ) & 1;
                    }
                }
            }
        }

        if( lostFlag == FLAG_DECODE_NORMAL ) {
            /* Regular decoding: skip all LBRR data */
            for( i = 0; i < channel_state[ 0 ].nFramesPerPacket; i++ ) {
                for( n = 0; n < decControl->nChannelsInternal; n++ ) {
                    if( channel_state[ n ].LBRR_flags[ i ] ) {
                        opus_int16 pulses[ MAX_FRAME_LENGTH ];
                        opus_int condCoding;

                        if( decControl->nChannelsInternal == 2 && n == 0 ) {
                            silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 );
                            if( channel_state[ 1 ].LBRR_flags[ i ] == 0 ) {
                                silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle );
                            }
                        }
                        /* Use conditional coding if previous frame available */
                        if( i > 0 && channel_state[ n ].LBRR_flags[ i - 1 ] ) {
                            condCoding = CODE_CONDITIONALLY;
                        } else {
                            condCoding = CODE_INDEPENDENTLY;
                        }
                        silk_decode_indices( &channel_state[ n ], psRangeDec, i, 1, condCoding );
                        silk_decode_pulses( psRangeDec, pulses, channel_state[ n ].indices.signalType,
                            channel_state[ n ].indices.quantOffsetType, channel_state[ n ].frame_length );
                    }
                }
            }
        }
    }

    /* Get MS predictor index */
    if( decControl->nChannelsInternal == 2 ) {
        if(   lostFlag == FLAG_DECODE_NORMAL ||
            ( lostFlag == FLAG_DECODE_LBRR && channel_state[ 0 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 1 ) )
        {
            silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 );
            /* For LBRR data, decode mid-only flag only if side-channel's LBRR flag is false */
            if( ( lostFlag == FLAG_DECODE_NORMAL && channel_state[ 1 ].VAD_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) ||
                ( lostFlag == FLAG_DECODE_LBRR && channel_state[ 1 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) )
            {
                silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle );
            } else {
                decode_only_middle = 0;
            }
        } else {
            for( n = 0; n < 2; n++ ) {
                MS_pred_Q13[ n ] = psDec->sStereo.pred_prev_Q13[ n ];
            }
        }
    }

    /* Reset side channel decoder prediction memory for first frame with side coding */
    if( decControl->nChannelsInternal == 2 && decode_only_middle == 0 && psDec->prev_decode_only_middle == 1 ) {
        silk_memset( psDec->channel_state[ 1 ].outBuf, 0, sizeof(psDec->channel_state[ 1 ].outBuf) );
        silk_memset( psDec->channel_state[ 1 ].sLPC_Q14_buf, 0, sizeof(psDec->channel_state[ 1 ].sLPC_Q14_buf) );
        psDec->channel_state[ 1 ].lagPrev        = 100;
        psDec->channel_state[ 1 ].LastGainIndex  = 10;
        psDec->channel_state[ 1 ].prevSignalType = TYPE_NO_VOICE_ACTIVITY;
        psDec->channel_state[ 1 ].first_frame_after_reset = 1;
    }

    /* Check if the temp buffer fits into the output PCM buffer. If it fits,
       we can delay allocating the temp buffer until after the SILK peak stack
       usage. We need to use a < and not a <= because of the two extra samples. */
    delay_stack_alloc = decControl->internalSampleRate*decControl->nChannelsInternal
          < decControl->API_sampleRate*decControl->nChannelsAPI;
    ALLOC( samplesOut1_tmp_storage1, delay_stack_alloc ? ALLOC_NONE
           : decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2 ),
           opus_int16 );
    if ( delay_stack_alloc )
    {
       samplesOut1_tmp[ 0 ] = samplesOut;
       samplesOut1_tmp[ 1 ] = samplesOut + channel_state[ 0 ].frame_length + 2;
    } else {
       samplesOut1_tmp[ 0 ] = samplesOut1_tmp_storage1;
       samplesOut1_tmp[ 1 ] = samplesOut1_tmp_storage1 + channel_state[ 0 ].frame_length + 2;
    }

    if( lostFlag == FLAG_DECODE_NORMAL ) {
        has_side = !decode_only_middle;
    } else {
        has_side = !psDec->prev_decode_only_middle
              || (decControl->nChannelsInternal == 2 && lostFlag == FLAG_DECODE_LBRR && channel_state[1].LBRR_flags[ channel_state[1].nFramesDecoded ] == 1 );
    }
    channel_state[ 0 ].sPLC.enable_deep_plc = decControl->enable_deep_plc;
    /* Call decoder for one frame */
    for( n = 0; n < decControl->nChannelsInternal; n++ ) {
        if( n == 0 || has_side ) {
            opus_int FrameIndex;
            opus_int condCoding;

            FrameIndex = channel_state[ 0 ].nFramesDecoded - n;
            /* Use independent coding if no previous frame available */
            if( FrameIndex <= 0 ) {
                condCoding = CODE_INDEPENDENTLY;
            } else if( lostFlag == FLAG_DECODE_LBRR ) {
                condCoding = channel_state[ n ].LBRR_flags[ FrameIndex - 1 ] ? CODE_CONDITIONALLY : CODE_INDEPENDENTLY;
            } else if( n > 0 && psDec->prev_decode_only_middle ) {
                /* If we skipped a side frame in this packet, we don't
                   need LTP scaling; the LTP state is well-defined. */
                condCoding = CODE_INDEPENDENTLY_NO_LTP_SCALING;
            } else {
                condCoding = CODE_CONDITIONALLY;
            }
#ifdef ENABLE_OSCE
            if ( channel_state[n].osce.method != decControl->osce_method ) {
                osce_reset( &channel_state[n].osce, decControl->osce_method );
            }
#endif
            ret += silk_decode_frame( &channel_state[ n ], psRangeDec, &samplesOut1_tmp[ n ][ 2 ], &nSamplesOutDec, lostFlag, condCoding,
#ifdef ENABLE_DEEP_PLC
                n == 0 ? lpcnet : NULL,
#endif
#ifdef ENABLE_OSCE
                &psDec->osce_model,
#endif
                arch);
        } else {
            silk_memset( &samplesOut1_tmp[ n ][ 2 ], 0, nSamplesOutDec * sizeof( opus_int16 ) );
        }
        channel_state[ n ].nFramesDecoded++;
    }

    if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 ) {
        /* Convert Mid/Side to Left/Right */
        silk_stereo_MS_to_LR( &psDec->sStereo, samplesOut1_tmp[ 0 ], samplesOut1_tmp[ 1 ], MS_pred_Q13, channel_state[ 0 ].fs_kHz, nSamplesOutDec );
    } else {
        /* Buffering */
        silk_memcpy( samplesOut1_tmp[ 0 ], psDec->sStereo.sMid, 2 * sizeof( opus_int16 ) );
        silk_memcpy( psDec->sStereo.sMid, &samplesOut1_tmp[ 0 ][ nSamplesOutDec ], 2 * sizeof( opus_int16 ) );
    }

    /* Number of output samples */
    *nSamplesOut = silk_DIV32( nSamplesOutDec * decControl->API_sampleRate, silk_SMULBB( channel_state[ 0 ].fs_kHz, 1000 ) );

    /* Set up pointers to temp buffers */
    ALLOC( samplesOut2_tmp,
           decControl->nChannelsAPI == 2 ? *nSamplesOut : ALLOC_NONE, opus_int16 );
    if( decControl->nChannelsAPI == 2 ) {
        resample_out_ptr = samplesOut2_tmp;
    } else {
        resample_out_ptr = samplesOut;
    }

    ALLOC( samplesOut1_tmp_storage2, delay_stack_alloc
           ? decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2 )
           : ALLOC_NONE,
           opus_int16 );
    if ( delay_stack_alloc ) {
       OPUS_COPY(samplesOut1_tmp_storage2, samplesOut, decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2));
       samplesOut1_tmp[ 0 ] = samplesOut1_tmp_storage2;
       samplesOut1_tmp[ 1 ] = samplesOut1_tmp_storage2 + channel_state[ 0 ].frame_length + 2;
    }
    for( n = 0; n < silk_min( decControl->nChannelsAPI, decControl->nChannelsInternal ); n++ ) {

        /* Resample decoded signal to API_sampleRate */
        ret += silk_resampler( &channel_state[ n ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec );

        /* Interleave if stereo output and stereo stream */
        if( decControl->nChannelsAPI == 2 ) {
            for( i = 0; i < *nSamplesOut; i++ ) {
                samplesOut[ n + 2 * i ] = resample_out_ptr[ i ];
            }
        }
    }

    /* Create two channel output from mono stream */
    if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 1 ) {
        if ( stereo_to_mono ){
            /* Resample right channel for newly collapsed stereo just in case
               we weren't doing collapsing when switching to mono */
            ret += silk_resampler( &channel_state[ 1 ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ 0 ][ 1 ], nSamplesOutDec );

            for( i = 0; i < *nSamplesOut; i++ ) {
                samplesOut[ 1 + 2 * i ] = resample_out_ptr[ i ];
            }
        } else {
            for( i = 0; i < *nSamplesOut; i++ ) {
                samplesOut[ 1 + 2 * i ] = samplesOut[ 0 + 2 * i ];
            }
        }
    }

    /* Export pitch lag, measured at 48 kHz sampling rate */
    if( channel_state[ 0 ].prevSignalType == TYPE_VOICED ) {
        int mult_tab[ 3 ] = { 6, 4, 3 };
        decControl->prevPitchLag = channel_state[ 0 ].lagPrev * mult_tab[ ( channel_state[ 0 ].fs_kHz - 8 ) >> 2 ];
    } else {
        decControl->prevPitchLag = 0;
    }

    if( lostFlag == FLAG_PACKET_LOST ) {
       /* On packet loss, remove the gain clamping to prevent having the energy "bounce back"
          if we lose packets when the energy is going down */
       for ( i = 0; i < psDec->nChannelsInternal; i++ )
          psDec->channel_state[ i ].LastGainIndex = 10;
    } else {
       psDec->prev_decode_only_middle = decode_only_middle;
    }
    RESTORE_STACK;
    return ret;
}

#if 0
/* Getting table of contents for a packet */
opus_int silk_get_TOC(
    const opus_uint8                *payload,           /* I    Payload data                                */
    const opus_int                  nBytesIn,           /* I    Number of input bytes                       */
    const opus_int                  nFramesPerPayload,  /* I    Number of SILK frames per payload           */
    silk_TOC_struct                 *Silk_TOC           /* O    Type of content                             */
)
{
    opus_int i, flags, ret = SILK_NO_ERROR;

    if( nBytesIn < 1 ) {
        return -1;
    }
    if( nFramesPerPayload < 0 || nFramesPerPayload > 3 ) {
        return -1;
    }

    silk_memset( Silk_TOC, 0, sizeof( *Silk_TOC ) );

    /* For stereo, extract the flags for the mid channel */
    flags = silk_RSHIFT( payload[ 0 ], 7 - nFramesPerPayload ) & ( silk_LSHIFT( 1, nFramesPerPayload + 1 ) - 1 );

    Silk_TOC->inbandFECFlag = flags & 1;
    for( i = nFramesPerPayload - 1; i >= 0 ; i-- ) {
        flags = silk_RSHIFT( flags, 1 );
        Silk_TOC->VADFlags[ i ] = flags & 1;
        Silk_TOC->VADFlag |= flags & 1;
    }

    return ret;
}
#endif

Coverage Report

Created: 2024-09-06 07:53

Line	Count	Source (jump to first uncovered line)
1		/***********************************************************************
2		Copyright (c) 2006-2011, Skype Limited. All rights reserved.
3		Redistribution and use in source and binary forms, with or without
4		modification, are permitted provided that the following conditions
5		are met:
6		- Redistributions of source code must retain the above copyright notice,
7		this list of conditions and the following disclaimer.
8		- Redistributions in binary form must reproduce the above copyright
9		notice, this list of conditions and the following disclaimer in the
10		documentation and/or other materials provided with the distribution.
11		- Neither the name of Internet Society, IETF or IETF Trust, nor the
12		names of specific contributors, may be used to endorse or promote
13		products derived from this software without specific prior written
14		permission.
15		THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
16		AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17		IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18		ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
19		LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
20		CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
21		SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
22		INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
23		CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
24		ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
25		POSSIBILITY OF SUCH DAMAGE.
26		***********************************************************************/
27
28		#ifdef HAVE_CONFIG_H
29		#include "config.h"
30		#endif
31		#include "API.h"
32		#include "main.h"
33		#include "stack_alloc.h"
34		#include "os_support.h"
35
36		#ifdef ENABLE_OSCE
37		#include "osce.h"
38		#include "osce_structs.h"
39		#endif
40
41		/************************/
42		/* Decoder Super Struct */
43		/************************/
44		typedef struct {
45		silk_decoder_state channel_state[ DECODER_NUM_CHANNELS ];
46		stereo_dec_state sStereo;
47		opus_int nChannelsAPI;
48		opus_int nChannelsInternal;
49		opus_int prev_decode_only_middle;
50		#ifdef ENABLE_OSCE
51		OSCEModel osce_model;
52		#endif
53		} silk_decoder;
54
55		/*********************/
56		/* Decoder functions */
57		/*********************/
58
59
60
61		opus_int silk_LoadOSCEModels(void decState, const unsigned char data, int len)
62	0	{
63		#ifdef ENABLE_OSCE
64		opus_int ret = SILK_NO_ERROR;
65
66		ret = osce_load_models(&((silk_decoder *)decState)->osce_model, data, len);
67		((silk_decoder *)decState)->osce_model.loaded = (ret == 0);
68		return ret;
69		#else
70	0	(void) decState;
71	0	(void) data;
72	0	(void) len;
73	0	return SILK_NO_ERROR;
74	0	#endif
75	0	}
76
77		opus_int silk_Get_Decoder_Size( /* O Returns error code */
78		opus_int decSizeBytes / O Number of bytes in SILK decoder state */
79		)
80	0	{
81	0	opus_int ret = SILK_NO_ERROR;
82
83	0	*decSizeBytes = sizeof( silk_decoder );
84
85	0	return ret;
86	0	}
87
88		/* Reset decoder state */
89		opus_int silk_ResetDecoder( /* O Returns error code */
90		void decState / I/O State */
91		)
92	0	{
93	0	opus_int n, ret = SILK_NO_ERROR;
94	0	silk_decoder_state channel_state = ((silk_decoder )decState)->channel_state;
95
96	0	for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) {
97	0	ret = silk_reset_decoder( &channel_state[ n ] );
98	0	}
99	0	silk_memset(&((silk_decoder )decState)->sStereo, 0, sizeof(((silk_decoder )decState)->sStereo));
100		/* Not strictly needed, but it's cleaner that way */
101	0	((silk_decoder *)decState)->prev_decode_only_middle = 0;
102
103	0	return ret;
104	0	}
105
106
107		opus_int silk_InitDecoder( /* O Returns error code */
108		void decState / I/O State */
109		)
110	0	{
111	0	opus_int n, ret = SILK_NO_ERROR;
112	0	silk_decoder_state channel_state = ((silk_decoder )decState)->channel_state;
113		#ifdef ENABLE_OSCE
114		((silk_decoder *)decState)->osce_model.loaded = 0;
115		#endif
116	0	#ifndef USE_WEIGHTS_FILE
117		/* load osce models */
118	0	silk_LoadOSCEModels(decState, NULL, 0);
119	0	#endif
120
121	0	for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) {
122	0	ret = silk_init_decoder( &channel_state[ n ] );
123	0	}
124	0	silk_memset(&((silk_decoder )decState)->sStereo, 0, sizeof(((silk_decoder )decState)->sStereo));
125		/* Not strictly needed, but it's cleaner that way */
126	0	((silk_decoder *)decState)->prev_decode_only_middle = 0;
127
128	0	return ret;
129	0	}
130
131		/* Decode a frame */
132		opus_int silk_Decode( /* O Returns error code */
133		void* decState, /* I/O State */
134		silk_DecControlStruct* decControl, /* I/O Control Structure */
135		opus_int lostFlag, /* I 0: no loss, 1 loss, 2 decode fec */
136		opus_int newPacketFlag, /* I Indicates first decoder call for this packet */
137		ec_dec psRangeDec, / I/O Compressor data structure */
138		opus_int16 samplesOut, / O Decoded output speech vector */
139		opus_int32 nSamplesOut, / O Number of samples decoded */
140		#ifdef ENABLE_DEEP_PLC
141		LPCNetPLCState *lpcnet,
142		#endif
143		int arch /* I Run-time architecture */
144		)
145	0	{
146	0	opus_int i, n, decode_only_middle = 0, ret = SILK_NO_ERROR;
147	0	opus_int32 nSamplesOutDec, LBRR_symbol;
148	0	opus_int16 *samplesOut1_tmp[ 2 ];
149	0	VARDECL( opus_int16, samplesOut1_tmp_storage1 );
150	0	VARDECL( opus_int16, samplesOut1_tmp_storage2 );
151	0	VARDECL( opus_int16, samplesOut2_tmp );
152	0	opus_int32 MS_pred_Q13[ 2 ] = { 0 };
153	0	opus_int16 *resample_out_ptr;
154	0	silk_decoder psDec = ( silk_decoder )decState;
155	0	silk_decoder_state *channel_state = psDec->channel_state;
156	0	opus_int has_side;
157	0	opus_int stereo_to_mono;
158	0	int delay_stack_alloc;
159	0	SAVE_STACK;
160
161	0	celt_assert( decControl->nChannelsInternal == 1 \|\| decControl->nChannelsInternal == 2 );
162
163		/**********************************/
164		/* Test if first frame in payload */
165		/**********************************/
166	0	if( newPacketFlag ) {
167	0	for( n = 0; n < decControl->nChannelsInternal; n++ ) {
168	0	channel_state[ n ].nFramesDecoded = 0; /* Used to count frames in packet */
169	0	}
170	0	}
171
172		/* If Mono -> Stereo transition in bitstream: init state of second channel */
173	0	if( decControl->nChannelsInternal > psDec->nChannelsInternal ) {
174	0	ret += silk_init_decoder( &channel_state[ 1 ] );
175	0	}
176
177	0	stereo_to_mono = decControl->nChannelsInternal == 1 && psDec->nChannelsInternal == 2 &&
178	0	( decControl->internalSampleRate == 1000*channel_state[ 0 ].fs_kHz );
179
180	0	if( channel_state[ 0 ].nFramesDecoded == 0 ) {
181	0	for( n = 0; n < decControl->nChannelsInternal; n++ ) {
182	0	opus_int fs_kHz_dec;
183	0	if( decControl->payloadSize_ms == 0 ) {
184		/* Assuming packet loss, use 10 ms */
185	0	channel_state[ n ].nFramesPerPacket = 1;
186	0	channel_state[ n ].nb_subfr = 2;
187	0	} else if( decControl->payloadSize_ms == 10 ) {
188	0	channel_state[ n ].nFramesPerPacket = 1;
189	0	channel_state[ n ].nb_subfr = 2;
190	0	} else if( decControl->payloadSize_ms == 20 ) {
191	0	channel_state[ n ].nFramesPerPacket = 1;
192	0	channel_state[ n ].nb_subfr = 4;
193	0	} else if( decControl->payloadSize_ms == 40 ) {
194	0	channel_state[ n ].nFramesPerPacket = 2;
195	0	channel_state[ n ].nb_subfr = 4;
196	0	} else if( decControl->payloadSize_ms == 60 ) {
197	0	channel_state[ n ].nFramesPerPacket = 3;
198	0	channel_state[ n ].nb_subfr = 4;
199	0	} else {
200	0	celt_assert( 0 );
201	0	RESTORE_STACK;
202	0	return SILK_DEC_INVALID_FRAME_SIZE;
203	0	}
204	0	fs_kHz_dec = ( decControl->internalSampleRate >> 10 ) + 1;
205	0	if( fs_kHz_dec != 8 && fs_kHz_dec != 12 && fs_kHz_dec != 16 ) {
206	0	celt_assert( 0 );
207	0	RESTORE_STACK;
208	0	return SILK_DEC_INVALID_SAMPLING_FREQUENCY;
209	0	}
210	0	ret += silk_decoder_set_fs( &channel_state[ n ], fs_kHz_dec, decControl->API_sampleRate );
211	0	}
212	0	}
213
214	0	if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 && ( psDec->nChannelsAPI == 1 \|\| psDec->nChannelsInternal == 1 ) ) {
215	0	silk_memset( psDec->sStereo.pred_prev_Q13, 0, sizeof( psDec->sStereo.pred_prev_Q13 ) );
216	0	silk_memset( psDec->sStereo.sSide, 0, sizeof( psDec->sStereo.sSide ) );
217	0	silk_memcpy( &channel_state[ 1 ].resampler_state, &channel_state[ 0 ].resampler_state, sizeof( silk_resampler_state_struct ) );
218	0	}
219	0	psDec->nChannelsAPI = decControl->nChannelsAPI;
220	0	psDec->nChannelsInternal = decControl->nChannelsInternal;
221
222	0	if( decControl->API_sampleRate > (opus_int32)MAX_API_FS_KHZ * 1000 \|\| decControl->API_sampleRate < 8000 ) {
223	0	ret = SILK_DEC_INVALID_SAMPLING_FREQUENCY;
224	0	RESTORE_STACK;
225	0	return( ret );
226	0	}
227
228	0	if( lostFlag != FLAG_PACKET_LOST && channel_state[ 0 ].nFramesDecoded == 0 ) {
229		/* First decoder call for this payload */
230		/* Decode VAD flags and LBRR flag */
231	0	for( n = 0; n < decControl->nChannelsInternal; n++ ) {
232	0	for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) {
233	0	channel_state[ n ].VAD_flags[ i ] = ec_dec_bit_logp(psRangeDec, 1);
234	0	}
235	0	channel_state[ n ].LBRR_flag = ec_dec_bit_logp(psRangeDec, 1);
236	0	}
237		/* Decode LBRR flags */
238	0	for( n = 0; n < decControl->nChannelsInternal; n++ ) {
239	0	silk_memset( channel_state[ n ].LBRR_flags, 0, sizeof( channel_state[ n ].LBRR_flags ) );
240	0	if( channel_state[ n ].LBRR_flag ) {
241	0	if( channel_state[ n ].nFramesPerPacket == 1 ) {
242	0	channel_state[ n ].LBRR_flags[ 0 ] = 1;
243	0	} else {
244	0	LBRR_symbol = ec_dec_icdf( psRangeDec, silk_LBRR_flags_iCDF_ptr[ channel_state[ n ].nFramesPerPacket - 2 ], 8 ) + 1;
245	0	for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) {
246	0	channel_state[ n ].LBRR_flags[ i ] = silk_RSHIFT( LBRR_symbol, i ) & 1;
247	0	}
248	0	}
249	0	}
250	0	}
251
252	0	if( lostFlag == FLAG_DECODE_NORMAL ) {
253		/* Regular decoding: skip all LBRR data */
254	0	for( i = 0; i < channel_state[ 0 ].nFramesPerPacket; i++ ) {
255	0	for( n = 0; n < decControl->nChannelsInternal; n++ ) {
256	0	if( channel_state[ n ].LBRR_flags[ i ] ) {
257	0	opus_int16 pulses[ MAX_FRAME_LENGTH ];
258	0	opus_int condCoding;
259
260	0	if( decControl->nChannelsInternal == 2 && n == 0 ) {
261	0	silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 );
262	0	if( channel_state[ 1 ].LBRR_flags[ i ] == 0 ) {
263	0	silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle );
264	0	}
265	0	}
266		/* Use conditional coding if previous frame available */
267	0	if( i > 0 && channel_state[ n ].LBRR_flags[ i - 1 ] ) {
268	0	condCoding = CODE_CONDITIONALLY;
269	0	} else {
270	0	condCoding = CODE_INDEPENDENTLY;
271	0	}
272	0	silk_decode_indices( &channel_state[ n ], psRangeDec, i, 1, condCoding );
273	0	silk_decode_pulses( psRangeDec, pulses, channel_state[ n ].indices.signalType,
274	0	channel_state[ n ].indices.quantOffsetType, channel_state[ n ].frame_length );
275	0	}
276	0	}
277	0	}
278	0	}
279	0	}
280
281		/* Get MS predictor index */
282	0	if( decControl->nChannelsInternal == 2 ) {
283	0	if( lostFlag == FLAG_DECODE_NORMAL \|\|
284	0	( lostFlag == FLAG_DECODE_LBRR && channel_state[ 0 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 1 ) )
285	0	{
286	0	silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 );
287		/* For LBRR data, decode mid-only flag only if side-channel's LBRR flag is false */
288	0	if( ( lostFlag == FLAG_DECODE_NORMAL && channel_state[ 1 ].VAD_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) \|\|
289	0	( lostFlag == FLAG_DECODE_LBRR && channel_state[ 1 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) )
290	0	{
291	0	silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle );
292	0	} else {
293	0	decode_only_middle = 0;
294	0	}
295	0	} else {
296	0	for( n = 0; n < 2; n++ ) {
297	0	MS_pred_Q13[ n ] = psDec->sStereo.pred_prev_Q13[ n ];
298	0	}
299	0	}
300	0	}
301
302		/* Reset side channel decoder prediction memory for first frame with side coding */
303	0	if( decControl->nChannelsInternal == 2 && decode_only_middle == 0 && psDec->prev_decode_only_middle == 1 ) {
304	0	silk_memset( psDec->channel_state[ 1 ].outBuf, 0, sizeof(psDec->channel_state[ 1 ].outBuf) );
305	0	silk_memset( psDec->channel_state[ 1 ].sLPC_Q14_buf, 0, sizeof(psDec->channel_state[ 1 ].sLPC_Q14_buf) );
306	0	psDec->channel_state[ 1 ].lagPrev = 100;
307	0	psDec->channel_state[ 1 ].LastGainIndex = 10;
308	0	psDec->channel_state[ 1 ].prevSignalType = TYPE_NO_VOICE_ACTIVITY;
309	0	psDec->channel_state[ 1 ].first_frame_after_reset = 1;
310	0	}
311
312		/* Check if the temp buffer fits into the output PCM buffer. If it fits,
313		we can delay allocating the temp buffer until after the SILK peak stack
314		usage. We need to use a < and not a <= because of the two extra samples. */
315	0	delay_stack_alloc = decControl->internalSampleRate*decControl->nChannelsInternal
316	0	< decControl->API_sampleRate*decControl->nChannelsAPI;
317	0	ALLOC( samplesOut1_tmp_storage1, delay_stack_alloc ? ALLOC_NONE
318	0	: decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2 ),
319	0	opus_int16 );
320	0	if ( delay_stack_alloc )
321	0	{
322	0	samplesOut1_tmp[ 0 ] = samplesOut;
323	0	samplesOut1_tmp[ 1 ] = samplesOut + channel_state[ 0 ].frame_length + 2;
324	0	} else {
325	0	samplesOut1_tmp[ 0 ] = samplesOut1_tmp_storage1;
326	0	samplesOut1_tmp[ 1 ] = samplesOut1_tmp_storage1 + channel_state[ 0 ].frame_length + 2;
327	0	}
328
329	0	if( lostFlag == FLAG_DECODE_NORMAL ) {
330	0	has_side = !decode_only_middle;
331	0	} else {
332	0	has_side = !psDec->prev_decode_only_middle
333	0	\|\| (decControl->nChannelsInternal == 2 && lostFlag == FLAG_DECODE_LBRR && channel_state[1].LBRR_flags[ channel_state[1].nFramesDecoded ] == 1 );
334	0	}
335	0	channel_state[ 0 ].sPLC.enable_deep_plc = decControl->enable_deep_plc;
336		/* Call decoder for one frame */
337	0	for( n = 0; n < decControl->nChannelsInternal; n++ ) {
338	0	if( n == 0 \|\| has_side ) {
339	0	opus_int FrameIndex;
340	0	opus_int condCoding;
341
342	0	FrameIndex = channel_state[ 0 ].nFramesDecoded - n;
343		/* Use independent coding if no previous frame available */
344	0	if( FrameIndex <= 0 ) {
345	0	condCoding = CODE_INDEPENDENTLY;
346	0	} else if( lostFlag == FLAG_DECODE_LBRR ) {
347	0	condCoding = channel_state[ n ].LBRR_flags[ FrameIndex - 1 ] ? CODE_CONDITIONALLY : CODE_INDEPENDENTLY;
348	0	} else if( n > 0 && psDec->prev_decode_only_middle ) {
349		/* If we skipped a side frame in this packet, we don't
350		need LTP scaling; the LTP state is well-defined. */
351	0	condCoding = CODE_INDEPENDENTLY_NO_LTP_SCALING;
352	0	} else {
353	0	condCoding = CODE_CONDITIONALLY;
354	0	}
355		#ifdef ENABLE_OSCE
356		if ( channel_state[n].osce.method != decControl->osce_method ) {
357		osce_reset( &channel_state[n].osce, decControl->osce_method );
358		}
359		#endif
360	0	ret += silk_decode_frame( &channel_state[ n ], psRangeDec, &samplesOut1_tmp[ n ][ 2 ], &nSamplesOutDec, lostFlag, condCoding,
361		#ifdef ENABLE_DEEP_PLC
362		n == 0 ? lpcnet : NULL,
363		#endif
364		#ifdef ENABLE_OSCE
365		&psDec->osce_model,
366		#endif
367	0	arch);
368	0	} else {
369	0	silk_memset( &samplesOut1_tmp[ n ][ 2 ], 0, nSamplesOutDec * sizeof( opus_int16 ) );
370	0	}
371	0	channel_state[ n ].nFramesDecoded++;
372	0	}
373
374	0	if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 ) {
375		/* Convert Mid/Side to Left/Right */
376	0	silk_stereo_MS_to_LR( &psDec->sStereo, samplesOut1_tmp[ 0 ], samplesOut1_tmp[ 1 ], MS_pred_Q13, channel_state[ 0 ].fs_kHz, nSamplesOutDec );
377	0	} else {
378		/* Buffering */
379	0	silk_memcpy( samplesOut1_tmp[ 0 ], psDec->sStereo.sMid, 2 * sizeof( opus_int16 ) );
380	0	silk_memcpy( psDec->sStereo.sMid, &samplesOut1_tmp[ 0 ][ nSamplesOutDec ], 2 * sizeof( opus_int16 ) );
381	0	}
382
383		/* Number of output samples */
384	0	nSamplesOut = silk_DIV32( nSamplesOutDec decControl->API_sampleRate, silk_SMULBB( channel_state[ 0 ].fs_kHz, 1000 ) );
385
386		/* Set up pointers to temp buffers */
387	0	ALLOC( samplesOut2_tmp,
388	0	decControl->nChannelsAPI == 2 ? *nSamplesOut : ALLOC_NONE, opus_int16 );
389	0	if( decControl->nChannelsAPI == 2 ) {
390	0	resample_out_ptr = samplesOut2_tmp;
391	0	} else {
392	0	resample_out_ptr = samplesOut;
393	0	}
394
395	0	ALLOC( samplesOut1_tmp_storage2, delay_stack_alloc
396	0	? decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2 )
397	0	: ALLOC_NONE,
398	0	opus_int16 );
399	0	if ( delay_stack_alloc ) {
400	0	OPUS_COPY(samplesOut1_tmp_storage2, samplesOut, decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2));
401	0	samplesOut1_tmp[ 0 ] = samplesOut1_tmp_storage2;
402	0	samplesOut1_tmp[ 1 ] = samplesOut1_tmp_storage2 + channel_state[ 0 ].frame_length + 2;
403	0	}
404	0	for( n = 0; n < silk_min( decControl->nChannelsAPI, decControl->nChannelsInternal ); n++ ) {
405
406		/* Resample decoded signal to API_sampleRate */
407	0	ret += silk_resampler( &channel_state[ n ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec );
408
409		/* Interleave if stereo output and stereo stream */
410	0	if( decControl->nChannelsAPI == 2 ) {
411	0	for( i = 0; i < *nSamplesOut; i++ ) {
412	0	samplesOut[ n + 2 * i ] = resample_out_ptr[ i ];
413	0	}
414	0	}
415	0	}
416
417		/* Create two channel output from mono stream */
418	0	if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 1 ) {
419	0	if ( stereo_to_mono ){
420		/* Resample right channel for newly collapsed stereo just in case
421		we weren't doing collapsing when switching to mono */
422	0	ret += silk_resampler( &channel_state[ 1 ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ 0 ][ 1 ], nSamplesOutDec );
423
424	0	for( i = 0; i < *nSamplesOut; i++ ) {
425	0	samplesOut[ 1 + 2 * i ] = resample_out_ptr[ i ];
426	0	}
427	0	} else {
428	0	for( i = 0; i < *nSamplesOut; i++ ) {
429	0	samplesOut[ 1 + 2 * i ] = samplesOut[ 0 + 2 * i ];
430	0	}
431	0	}
432	0	}
433
434		/* Export pitch lag, measured at 48 kHz sampling rate */
435	0	if( channel_state[ 0 ].prevSignalType == TYPE_VOICED ) {
436	0	int mult_tab[ 3 ] = { 6, 4, 3 };
437	0	decControl->prevPitchLag = channel_state[ 0 ].lagPrev * mult_tab[ ( channel_state[ 0 ].fs_kHz - 8 ) >> 2 ];
438	0	} else {
439	0	decControl->prevPitchLag = 0;
440	0	}
441
442	0	if( lostFlag == FLAG_PACKET_LOST ) {
443		/* On packet loss, remove the gain clamping to prevent having the energy "bounce back"
444		if we lose packets when the energy is going down */
445	0	for ( i = 0; i < psDec->nChannelsInternal; i++ )
446	0	psDec->channel_state[ i ].LastGainIndex = 10;
447	0	} else {
448	0	psDec->prev_decode_only_middle = decode_only_middle;
449	0	}
450	0	RESTORE_STACK;
451	0	return ret;
452	0	}
453
454		#if 0
455		/* Getting table of contents for a packet */
456		opus_int silk_get_TOC(
457		const opus_uint8 payload, / I Payload data */
458		const opus_int nBytesIn, /* I Number of input bytes */
459		const opus_int nFramesPerPayload, /* I Number of SILK frames per payload */
460		silk_TOC_struct Silk_TOC / O Type of content */
461		)
462		{
463		opus_int i, flags, ret = SILK_NO_ERROR;
464
465		if( nBytesIn < 1 ) {
466		return -1;
467		}
468		if( nFramesPerPayload < 0 \|\| nFramesPerPayload > 3 ) {
469		return -1;
470		}
471
472		silk_memset( Silk_TOC, 0, sizeof( *Silk_TOC ) );
473
474		/* For stereo, extract the flags for the mid channel */
475		flags = silk_RSHIFT( payload[ 0 ], 7 - nFramesPerPayload ) & ( silk_LSHIFT( 1, nFramesPerPayload + 1 ) - 1 );
476
477		Silk_TOC->inbandFECFlag = flags & 1;
478		for( i = nFramesPerPayload - 1; i >= 0 ; i-- ) {
479		flags = silk_RSHIFT( flags, 1 );
480		Silk_TOC->VADFlags[ i ] = flags & 1;
481		Silk_TOC->VADFlag \|= flags & 1;
482		}
483
484		return ret;
485		}
486		#endif