/src/mozilla-central/media/libopus/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"

/************************/
/* 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;
} silk_decoder;

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

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_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;

    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                       */
    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 );
    }
    /* 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;
            }
            ret += silk_decode_frame( &channel_state[ n ], psRangeDec, &samplesOut1_tmp[ n ][ 2 ], &nSamplesOutDec, lostFlag, condCoding, 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: 2018-09-25 14:53

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		/************************/
37		/* Decoder Super Struct */
38		/************************/
39		typedef struct {
40		silk_decoder_state channel_state[ DECODER_NUM_CHANNELS ];
41		stereo_dec_state sStereo;
42		opus_int nChannelsAPI;
43		opus_int nChannelsInternal;
44		opus_int prev_decode_only_middle;
45		} silk_decoder;
46
47		/*********************/
48		/* Decoder functions */
49		/*********************/
50
51		opus_int silk_Get_Decoder_Size( /* O Returns error code */
52		opus_int decSizeBytes / O Number of bytes in SILK decoder state */
53		)
54	0	{
55	0	opus_int ret = SILK_NO_ERROR;
56	0
57	0	*decSizeBytes = sizeof( silk_decoder );
58	0
59	0	return ret;
60	0	}
61
62		/* Reset decoder state */
63		opus_int silk_InitDecoder( /* O Returns error code */
64		void decState / I/O State */
65		)
66	0	{
67	0	opus_int n, ret = SILK_NO_ERROR;
68	0	silk_decoder_state channel_state = ((silk_decoder )decState)->channel_state;
69	0
70	0	for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) {
71	0	ret = silk_init_decoder( &channel_state[ n ] );
72	0	}
73	0	silk_memset(&((silk_decoder )decState)->sStereo, 0, sizeof(((silk_decoder )decState)->sStereo));
74	0	/* Not strictly needed, but it's cleaner that way */
75	0	((silk_decoder *)decState)->prev_decode_only_middle = 0;
76	0
77	0	return ret;
78	0	}
79
80		/* Decode a frame */
81		opus_int silk_Decode( /* O Returns error code */
82		void* decState, /* I/O State */
83		silk_DecControlStruct* decControl, /* I/O Control Structure */
84		opus_int lostFlag, /* I 0: no loss, 1 loss, 2 decode fec */
85		opus_int newPacketFlag, /* I Indicates first decoder call for this packet */
86		ec_dec psRangeDec, / I/O Compressor data structure */
87		opus_int16 samplesOut, / O Decoded output speech vector */
88		opus_int32 nSamplesOut, / O Number of samples decoded */
89		int arch /* I Run-time architecture */
90		)
91	0	{
92	0	opus_int i, n, decode_only_middle = 0, ret = SILK_NO_ERROR;
93	0	opus_int32 nSamplesOutDec, LBRR_symbol;
94	0	opus_int16 *samplesOut1_tmp[ 2 ];
95	0	VARDECL( opus_int16, samplesOut1_tmp_storage1 );
96	0	VARDECL( opus_int16, samplesOut1_tmp_storage2 );
97	0	VARDECL( opus_int16, samplesOut2_tmp );
98	0	opus_int32 MS_pred_Q13[ 2 ] = { 0 };
99	0	opus_int16 *resample_out_ptr;
100	0	silk_decoder psDec = ( silk_decoder )decState;
101	0	silk_decoder_state *channel_state = psDec->channel_state;
102	0	opus_int has_side;
103	0	opus_int stereo_to_mono;
104	0	int delay_stack_alloc;
105	0	SAVE_STACK;
106	0
107	0	celt_assert( decControl->nChannelsInternal == 1 \|\| decControl->nChannelsInternal == 2 );
108	0
109	0	/**********************************/
110	0	/* Test if first frame in payload */
111	0	/**********************************/
112	0	if( newPacketFlag ) {
113	0	for( n = 0; n < decControl->nChannelsInternal; n++ ) {
114	0	channel_state[ n ].nFramesDecoded = 0; /* Used to count frames in packet */
115	0	}
116	0	}
117	0
118	0	/* If Mono -> Stereo transition in bitstream: init state of second channel */
119	0	if( decControl->nChannelsInternal > psDec->nChannelsInternal ) {
120	0	ret += silk_init_decoder( &channel_state[ 1 ] );
121	0	}
122	0
123	0	stereo_to_mono = decControl->nChannelsInternal == 1 && psDec->nChannelsInternal == 2 &&
124	0	( decControl->internalSampleRate == 1000*channel_state[ 0 ].fs_kHz );
125	0
126	0	if( channel_state[ 0 ].nFramesDecoded == 0 ) {
127	0	for( n = 0; n < decControl->nChannelsInternal; n++ ) {
128	0	opus_int fs_kHz_dec;
129	0	if( decControl->payloadSize_ms == 0 ) {
130	0	/* Assuming packet loss, use 10 ms */
131	0	channel_state[ n ].nFramesPerPacket = 1;
132	0	channel_state[ n ].nb_subfr = 2;
133	0	} else if( decControl->payloadSize_ms == 10 ) {
134	0	channel_state[ n ].nFramesPerPacket = 1;
135	0	channel_state[ n ].nb_subfr = 2;
136	0	} else if( decControl->payloadSize_ms == 20 ) {
137	0	channel_state[ n ].nFramesPerPacket = 1;
138	0	channel_state[ n ].nb_subfr = 4;
139	0	} else if( decControl->payloadSize_ms == 40 ) {
140	0	channel_state[ n ].nFramesPerPacket = 2;
141	0	channel_state[ n ].nb_subfr = 4;
142	0	} else if( decControl->payloadSize_ms == 60 ) {
143	0	channel_state[ n ].nFramesPerPacket = 3;
144	0	channel_state[ n ].nb_subfr = 4;
145	0	} else {
146	0	celt_assert( 0 );
147	0	RESTORE_STACK;
148	0	return SILK_DEC_INVALID_FRAME_SIZE;
149	0	}
150	0	fs_kHz_dec = ( decControl->internalSampleRate >> 10 ) + 1;
151	0	if( fs_kHz_dec != 8 && fs_kHz_dec != 12 && fs_kHz_dec != 16 ) {
152	0	celt_assert( 0 );
153	0	RESTORE_STACK;
154	0	return SILK_DEC_INVALID_SAMPLING_FREQUENCY;
155	0	}
156	0	ret += silk_decoder_set_fs( &channel_state[ n ], fs_kHz_dec, decControl->API_sampleRate );
157	0	}
158	0	}
159	0
160	0	if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 && ( psDec->nChannelsAPI == 1 \|\| psDec->nChannelsInternal == 1 ) ) {
161	0	silk_memset( psDec->sStereo.pred_prev_Q13, 0, sizeof( psDec->sStereo.pred_prev_Q13 ) );
162	0	silk_memset( psDec->sStereo.sSide, 0, sizeof( psDec->sStereo.sSide ) );
163	0	silk_memcpy( &channel_state[ 1 ].resampler_state, &channel_state[ 0 ].resampler_state, sizeof( silk_resampler_state_struct ) );
164	0	}
165	0	psDec->nChannelsAPI = decControl->nChannelsAPI;
166	0	psDec->nChannelsInternal = decControl->nChannelsInternal;
167	0
168	0	if( decControl->API_sampleRate > (opus_int32)MAX_API_FS_KHZ * 1000 \|\| decControl->API_sampleRate < 8000 ) {
169	0	ret = SILK_DEC_INVALID_SAMPLING_FREQUENCY;
170	0	RESTORE_STACK;
171	0	return( ret );
172	0	}
173	0
174	0	if( lostFlag != FLAG_PACKET_LOST && channel_state[ 0 ].nFramesDecoded == 0 ) {
175	0	/* First decoder call for this payload */
176	0	/* Decode VAD flags and LBRR flag */
177	0	for( n = 0; n < decControl->nChannelsInternal; n++ ) {
178	0	for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) {
179	0	channel_state[ n ].VAD_flags[ i ] = ec_dec_bit_logp(psRangeDec, 1);
180	0	}
181	0	channel_state[ n ].LBRR_flag = ec_dec_bit_logp(psRangeDec, 1);
182	0	}
183	0	/* Decode LBRR flags */
184	0	for( n = 0; n < decControl->nChannelsInternal; n++ ) {
185	0	silk_memset( channel_state[ n ].LBRR_flags, 0, sizeof( channel_state[ n ].LBRR_flags ) );
186	0	if( channel_state[ n ].LBRR_flag ) {
187	0	if( channel_state[ n ].nFramesPerPacket == 1 ) {
188	0	channel_state[ n ].LBRR_flags[ 0 ] = 1;
189	0	} else {
190	0	LBRR_symbol = ec_dec_icdf( psRangeDec, silk_LBRR_flags_iCDF_ptr[ channel_state[ n ].nFramesPerPacket - 2 ], 8 ) + 1;
191	0	for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) {
192	0	channel_state[ n ].LBRR_flags[ i ] = silk_RSHIFT( LBRR_symbol, i ) & 1;
193	0	}
194	0	}
195	0	}
196	0	}
197	0
198	0	if( lostFlag == FLAG_DECODE_NORMAL ) {
199	0	/* Regular decoding: skip all LBRR data */
200	0	for( i = 0; i < channel_state[ 0 ].nFramesPerPacket; i++ ) {
201	0	for( n = 0; n < decControl->nChannelsInternal; n++ ) {
202	0	if( channel_state[ n ].LBRR_flags[ i ] ) {
203	0	opus_int16 pulses[ MAX_FRAME_LENGTH ];
204	0	opus_int condCoding;
205	0
206	0	if( decControl->nChannelsInternal == 2 && n == 0 ) {
207	0	silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 );
208	0	if( channel_state[ 1 ].LBRR_flags[ i ] == 0 ) {
209	0	silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle );
210	0	}
211	0	}
212	0	/* Use conditional coding if previous frame available */
213	0	if( i > 0 && channel_state[ n ].LBRR_flags[ i - 1 ] ) {
214	0	condCoding = CODE_CONDITIONALLY;
215	0	} else {
216	0	condCoding = CODE_INDEPENDENTLY;
217	0	}
218	0	silk_decode_indices( &channel_state[ n ], psRangeDec, i, 1, condCoding );
219	0	silk_decode_pulses( psRangeDec, pulses, channel_state[ n ].indices.signalType,
220	0	channel_state[ n ].indices.quantOffsetType, channel_state[ n ].frame_length );
221	0	}
222	0	}
223	0	}
224	0	}
225	0	}
226	0
227	0	/* Get MS predictor index */
228	0	if( decControl->nChannelsInternal == 2 ) {
229	0	if( lostFlag == FLAG_DECODE_NORMAL \|\|
230	0	( lostFlag == FLAG_DECODE_LBRR && channel_state[ 0 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 1 ) )
231	0	{
232	0	silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 );
233	0	/* For LBRR data, decode mid-only flag only if side-channel's LBRR flag is false */
234	0	if( ( lostFlag == FLAG_DECODE_NORMAL && channel_state[ 1 ].VAD_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) \|\|
235	0	( lostFlag == FLAG_DECODE_LBRR && channel_state[ 1 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) )
236	0	{
237	0	silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle );
238	0	} else {
239	0	decode_only_middle = 0;
240	0	}
241	0	} else {
242	0	for( n = 0; n < 2; n++ ) {
243	0	MS_pred_Q13[ n ] = psDec->sStereo.pred_prev_Q13[ n ];
244	0	}
245	0	}
246	0	}
247	0
248	0	/* Reset side channel decoder prediction memory for first frame with side coding */
249	0	if( decControl->nChannelsInternal == 2 && decode_only_middle == 0 && psDec->prev_decode_only_middle == 1 ) {
250	0	silk_memset( psDec->channel_state[ 1 ].outBuf, 0, sizeof(psDec->channel_state[ 1 ].outBuf) );
251	0	silk_memset( psDec->channel_state[ 1 ].sLPC_Q14_buf, 0, sizeof(psDec->channel_state[ 1 ].sLPC_Q14_buf) );
252	0	psDec->channel_state[ 1 ].lagPrev = 100;
253	0	psDec->channel_state[ 1 ].LastGainIndex = 10;
254	0	psDec->channel_state[ 1 ].prevSignalType = TYPE_NO_VOICE_ACTIVITY;
255	0	psDec->channel_state[ 1 ].first_frame_after_reset = 1;
256	0	}
257	0
258	0	/* Check if the temp buffer fits into the output PCM buffer. If it fits,
259	0	we can delay allocating the temp buffer until after the SILK peak stack
260	0	usage. We need to use a < and not a <= because of the two extra samples. */
261	0	delay_stack_alloc = decControl->internalSampleRate*decControl->nChannelsInternal
262	0	< decControl->API_sampleRate*decControl->nChannelsAPI;
263	0	ALLOC( samplesOut1_tmp_storage1, delay_stack_alloc ? ALLOC_NONE
264	0	: decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2 ),
265	0	opus_int16 );
266	0	if ( delay_stack_alloc )
267	0	{
268	0	samplesOut1_tmp[ 0 ] = samplesOut;
269	0	samplesOut1_tmp[ 1 ] = samplesOut + channel_state[ 0 ].frame_length + 2;
270	0	} else {
271	0	samplesOut1_tmp[ 0 ] = samplesOut1_tmp_storage1;
272	0	samplesOut1_tmp[ 1 ] = samplesOut1_tmp_storage1 + channel_state[ 0 ].frame_length + 2;
273	0	}
274	0
275	0	if( lostFlag == FLAG_DECODE_NORMAL ) {
276	0	has_side = !decode_only_middle;
277	0	} else {
278	0	has_side = !psDec->prev_decode_only_middle
279	0	\|\| (decControl->nChannelsInternal == 2 && lostFlag == FLAG_DECODE_LBRR && channel_state[1].LBRR_flags[ channel_state[1].nFramesDecoded ] == 1 );
280	0	}
281	0	/* Call decoder for one frame */
282	0	for( n = 0; n < decControl->nChannelsInternal; n++ ) {
283	0	if( n == 0 \|\| has_side ) {
284	0	opus_int FrameIndex;
285	0	opus_int condCoding;
286	0
287	0	FrameIndex = channel_state[ 0 ].nFramesDecoded - n;
288	0	/* Use independent coding if no previous frame available */
289	0	if( FrameIndex <= 0 ) {
290	0	condCoding = CODE_INDEPENDENTLY;
291	0	} else if( lostFlag == FLAG_DECODE_LBRR ) {
292	0	condCoding = channel_state[ n ].LBRR_flags[ FrameIndex - 1 ] ? CODE_CONDITIONALLY : CODE_INDEPENDENTLY;
293	0	} else if( n > 0 && psDec->prev_decode_only_middle ) {
294	0	/* If we skipped a side frame in this packet, we don't
295	0	need LTP scaling; the LTP state is well-defined. */
296	0	condCoding = CODE_INDEPENDENTLY_NO_LTP_SCALING;
297	0	} else {
298	0	condCoding = CODE_CONDITIONALLY;
299	0	}
300	0	ret += silk_decode_frame( &channel_state[ n ], psRangeDec, &samplesOut1_tmp[ n ][ 2 ], &nSamplesOutDec, lostFlag, condCoding, arch);
301	0	} else {
302	0	silk_memset( &samplesOut1_tmp[ n ][ 2 ], 0, nSamplesOutDec * sizeof( opus_int16 ) );
303	0	}
304	0	channel_state[ n ].nFramesDecoded++;
305	0	}
306	0
307	0	if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 ) {
308	0	/* Convert Mid/Side to Left/Right */
309	0	silk_stereo_MS_to_LR( &psDec->sStereo, samplesOut1_tmp[ 0 ], samplesOut1_tmp[ 1 ], MS_pred_Q13, channel_state[ 0 ].fs_kHz, nSamplesOutDec );
310	0	} else {
311	0	/* Buffering */
312	0	silk_memcpy( samplesOut1_tmp[ 0 ], psDec->sStereo.sMid, 2 * sizeof( opus_int16 ) );
313	0	silk_memcpy( psDec->sStereo.sMid, &samplesOut1_tmp[ 0 ][ nSamplesOutDec ], 2 * sizeof( opus_int16 ) );
314	0	}
315	0
316	0	/* Number of output samples */
317	0	nSamplesOut = silk_DIV32( nSamplesOutDec decControl->API_sampleRate, silk_SMULBB( channel_state[ 0 ].fs_kHz, 1000 ) );
318	0
319	0	/* Set up pointers to temp buffers */
320	0	ALLOC( samplesOut2_tmp,
321	0	decControl->nChannelsAPI == 2 ? *nSamplesOut : ALLOC_NONE, opus_int16 );
322	0	if( decControl->nChannelsAPI == 2 ) {
323	0	resample_out_ptr = samplesOut2_tmp;
324	0	} else {
325	0	resample_out_ptr = samplesOut;
326	0	}
327	0
328	0	ALLOC( samplesOut1_tmp_storage2, delay_stack_alloc
329	0	? decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2 )
330	0	: ALLOC_NONE,
331	0	opus_int16 );
332	0	if ( delay_stack_alloc ) {
333	0	OPUS_COPY(samplesOut1_tmp_storage2, samplesOut, decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2));
334	0	samplesOut1_tmp[ 0 ] = samplesOut1_tmp_storage2;
335	0	samplesOut1_tmp[ 1 ] = samplesOut1_tmp_storage2 + channel_state[ 0 ].frame_length + 2;
336	0	}
337	0	for( n = 0; n < silk_min( decControl->nChannelsAPI, decControl->nChannelsInternal ); n++ ) {
338	0
339	0	/* Resample decoded signal to API_sampleRate */
340	0	ret += silk_resampler( &channel_state[ n ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec );
341	0
342	0	/* Interleave if stereo output and stereo stream */
343	0	if( decControl->nChannelsAPI == 2 ) {
344	0	for( i = 0; i < *nSamplesOut; i++ ) {
345	0	samplesOut[ n + 2 * i ] = resample_out_ptr[ i ];
346	0	}
347	0	}
348	0	}
349	0
350	0	/* Create two channel output from mono stream */
351	0	if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 1 ) {
352	0	if ( stereo_to_mono ){
353	0	/* Resample right channel for newly collapsed stereo just in case
354	0	we weren't doing collapsing when switching to mono */
355	0	ret += silk_resampler( &channel_state[ 1 ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ 0 ][ 1 ], nSamplesOutDec );
356	0
357	0	for( i = 0; i < *nSamplesOut; i++ ) {
358	0	samplesOut[ 1 + 2 * i ] = resample_out_ptr[ i ];
359	0	}
360	0	} else {
361	0	for( i = 0; i < *nSamplesOut; i++ ) {
362	0	samplesOut[ 1 + 2 * i ] = samplesOut[ 0 + 2 * i ];
363	0	}
364	0	}
365	0	}
366	0
367	0	/* Export pitch lag, measured at 48 kHz sampling rate */
368	0	if( channel_state[ 0 ].prevSignalType == TYPE_VOICED ) {
369	0	int mult_tab[ 3 ] = { 6, 4, 3 };
370	0	decControl->prevPitchLag = channel_state[ 0 ].lagPrev * mult_tab[ ( channel_state[ 0 ].fs_kHz - 8 ) >> 2 ];
371	0	} else {
372	0	decControl->prevPitchLag = 0;
373	0	}
374	0
375	0	if( lostFlag == FLAG_PACKET_LOST ) {
376	0	/* On packet loss, remove the gain clamping to prevent having the energy "bounce back"
377	0	if we lose packets when the energy is going down */
378	0	for ( i = 0; i < psDec->nChannelsInternal; i++ )
379	0	psDec->channel_state[ i ].LastGainIndex = 10;
380	0	} else {
381	0	psDec->prev_decode_only_middle = decode_only_middle;
382	0	}
383	0	RESTORE_STACK;
384	0	return ret;
385	0	}
386
387		#if 0
388		/* Getting table of contents for a packet */
389		opus_int silk_get_TOC(
390		const opus_uint8 payload, / I Payload data */
391		const opus_int nBytesIn, /* I Number of input bytes */
392		const opus_int nFramesPerPayload, /* I Number of SILK frames per payload */
393		silk_TOC_struct Silk_TOC / O Type of content */
394		)
395		{
396		opus_int i, flags, ret = SILK_NO_ERROR;
397
398		if( nBytesIn < 1 ) {
399		return -1;
400		}
401		if( nFramesPerPayload < 0 \|\| nFramesPerPayload > 3 ) {
402		return -1;
403		}
404
405		silk_memset( Silk_TOC, 0, sizeof( *Silk_TOC ) );
406
407		/* For stereo, extract the flags for the mid channel */
408		flags = silk_RSHIFT( payload[ 0 ], 7 - nFramesPerPayload ) & ( silk_LSHIFT( 1, nFramesPerPayload + 1 ) - 1 );
409
410		Silk_TOC->inbandFECFlag = flags & 1;
411		for( i = nFramesPerPayload - 1; i >= 0 ; i-- ) {
412		flags = silk_RSHIFT( flags, 1 );
413		Silk_TOC->VADFlags[ i ] = flags & 1;
414		Silk_TOC->VADFlag \|= flags & 1;
415		}
416
417		return ret;
418		}
419		#endif