Commit c3694594 authored by sagnowski's avatar sagnowski
Browse files

Update external renderer 

Changes:
- Add clipping of samples before float to int conversion
- Minor improvements to command line parsing
- Merge recent main
parent 43bd3cf6

apps/renderer.c

+31 −9
Original line number Diff line number Diff line
@@ -607,7 +607,19 @@ int32_t main( int32_t argc, char **argv ) 
        fprintf( stdout, "\nRenderer delay:         %-5u [samples] - Timescale: %5u\n", delayNumSamples_orig, delayTimeScale );

    }



    fprintf( stdout, "\n\nRendering of %d frames finished\n", frame );

    fprintf( stdout, "\n\nRendering of %d frames finished\n\n", frame );



#ifdef DEBUGGING

    int32_t cnt_frames_limited, noClipping;

    if ( ( cnt_frames_limited = IVAS_REND_GetCntFramesLimited( hIvasRend ) ) > 0 )

    {

        fprintf( stdout, "Limiter applied in %d frames.\n\n", cnt_frames_limited );

    }

    if ( ( noClipping = IVAS_REND_GetNoCLipping( hIvasRend ) ) > 0 )

    {

        fprintf( stdout, "Clipping (saturation) detected: %d samples clipped!!!\n\n", noClipping );

    }

#endif



    /* === Close === */

    count_free( inpInt16Buffer );
@@ -1170,6 +1182,7 @@ typedef enum 
    CmdLnOptionId_neverDropLfe,

    CmdLnOptionId_noDelayCmp,

    CmdLnOptionId_quietModeEnabled,

    CmdLnOptionId_inputMetadata,

} CmdLnOptionId;



static void parseOption( int32_t optionId, char **optionValues, int16_t numOptionValues, void *pOutputStruct )
@@ -1260,15 +1273,22 @@ static CmdlnArgs parseCmdlnArgs( int32_t argc, char **argv ) 
            .match = "input_file",

            .matchShort = "i",

            .isRequired = 1,

            .description = "Path to the input file",

            .description = "Path to the input file (WAV, raw PCM or scene description file)",

        },

        {

            .id = CmdLnOptionId_inputFormat,

            .match = "input_format",

            .matchShort = "if",

            .isRequired = 1,

            // .description = "Audio format of input file (e.g. 5_1 or HOA3 or META)", /* TODO(sgi): Add additional flag for listing all available formats */

            .description = "Format of input file\nIn case of a metadata format this should be followed by a list of metadata file paths or NULL",

        },

        // { /* TODO(sgi): move metadata file paths from input_format to this separate flag */

        //     .id = CmdLnOptionId_inputMetadata,

        //     .match = "input_metadata",

        //     .matchShort = "im",

        //     .description = "Space-separated list of path to metadata files for ISM or MASA inputs",

        // },

        {

            .id = CmdLnOptionId_outputFile,

            .match = "output_file",
@@ -1287,40 +1307,42 @@ static CmdlnArgs parseCmdlnArgs( int32_t argc, char **argv ) 
            .id = CmdLnOptionId_sampleRate,

            .match = "sample_rate",

            .matchShort = "fs",

            .isRequired = 1,

            .description = "Input sampling rate in kHz",

            .isRequired = 1, /* TODO(sgi): Shouldn't be required */

            .description = "Input sampling rate in kHz (16, 32, 48)", /* TODO(sgi): Add sampling rate to scene description files */

        },

        {

            .id = CmdLnOptionId_trajFile,

            .match = "trajectory_file",

            .matchShort = "tf",

            .description = "Head rotation trajectory file",

            .description = "Head rotation trajectory file for simulation of head tracking (only for BINAURAL and BINAURAL_ROOM outputs)",

        },

        {

            .id = CmdLnOptionId_customHrtfFile,

            .match = "custom_hrtf",

            .matchShort = "hrtf",

            .description = "Custom HRTF file for binaural rendering",

            .description = "Custom HRTF file for binaural rendering (only for BINAURAL and BINAURAL_ROOM outputs)",

        },

        {

            .id = CmdLnOptionId_renderConfigFile,

            .match = "render_config",

            .matchShort = "rc",

            .description = "Renderer configuration file",

            .description = "Binaural renderer configuration file (only for BINAURAL and BINAURAL_ROOM outputs)",

        },

        {

            .id = CmdLnOptionId_noDiegeticPan,

            .match = "no_diegetic_pan",

            .matchShort = "ndp",

            .description = "Panning mono no dietic sound to stereo -1<= pan <= 1\nleft or l or 1->left, right or r or -1->right, center or c or 0 ->middle",

            .description = "Panning mono no diegetic sound to stereo -1<= pan <= 1\nleft or l or 1->left, right or r or -1->right, center or c or 0 ->middle\n(todo: implementation)",

        },

        {

            .id = CmdLnOptionId_orientationTracking,

            .match = "tracking_type",

            .matchShort = "otr",

            .description = "Head orientation tracking type: 'ref' or 'avg' (only for binaural rendering)",

            .description = "Head orientation tracking type: 'ref' or 'avg' (only for BINAURAL and BINAURAL_ROOM) (todo: check implementation)",

        },

        {

            /* TODO(sgi): Replace with more configurable input, e.g. ask for a list of triplets: (gain, azimuth, elevation) to place LFE signal */

            /* rename to "lfeHandling" */

            .id = CmdLnOptionId_neverDropLfe,

            .match = "neverDropLfe",

            .matchShort = "ndl",

lib_com/bitstream.c

+18 −0
Original line number Diff line number Diff line
@@ -1975,11 +1975,29 @@ ivas_error preview_indices( 
        }

        else if ( st_ivas->ivas_format == SBA_FORMAT )

        {



            /* Read SBA planar flag and SBA order */

            st_ivas->sba_planar = ( bit_stream[IVAS_FORMAT_SIGNALING_NBITS_SBA] == 1 );

#ifndef SBA_ORDER_BITSTREAM

            st_ivas->sba_order = ( bit_stream[IVAS_FORMAT_SIGNALING_NBITS_SBA + 2] == 1 );

            st_ivas->sba_order += 2 * ( bit_stream[IVAS_FORMAT_SIGNALING_NBITS_SBA + 1] == 1 );

#else

            st_ivas->hDecoderConfig->sba_order = ( bit_stream[IVAS_FORMAT_SIGNALING_NBITS_SBA + 2] == 1 );

            st_ivas->hDecoderConfig->sba_order += 2 * ( bit_stream[IVAS_FORMAT_SIGNALING_NBITS_SBA + 1] == 1 );

            st_ivas->sba_analysis_order = st_ivas->hDecoderConfig->sba_order;

#endif

#ifdef SBA_ORDER_BITSTREAM

            /*Hard coding the the sba_oder as 1 as higher not supported below 256k bitrate*/

            if ( total_brate < IVAS_256k )

            {

                st_ivas->sba_analysis_order = 1;

            }

#endif

#ifdef SBA_ORDER_BITSTREAM

            ivas_sba_config( total_brate, st_ivas->sba_analysis_order, -1, &( st_ivas->nchan_transport ), st_ivas->sba_planar, &( st_ivas->nSCE ), &( st_ivas->nCPE ), &( st_ivas->element_mode_init ), st_ivas->sba_mode );

#else

            ivas_sba_config( total_brate, st_ivas->sba_order, -1, &( st_ivas->nchan_transport ), st_ivas->sba_planar, &( st_ivas->nSCE ), &( st_ivas->nCPE ), &( st_ivas->element_mode_init ), st_ivas->sba_mode );

#endif

        }

    }

lib_com/ivas_dirac_com.c

+8 −0
Original line number Diff line number Diff line
@@ -82,7 +82,11 @@ ivas_error ivas_dirac_config( 
        nCPE = &( (Encoder_Struct *) st_ivas )->nCPE;

        element_mode = &( (Encoder_Struct *) st_ivas )->hEncoderConfig->element_mode_init;

        nchan_transport = &( (Encoder_Struct *) st_ivas )->nchan_transport;

#ifndef SBA_ORDER_BITSTREAM

        sba_order = ( (Encoder_Struct *) st_ivas )->hEncoderConfig->sba_order;

#else

        sba_order = ( (Encoder_Struct *) st_ivas )->sba_analysis_order;

#endif

        sba_planar = ( (Encoder_Struct *) st_ivas )->hEncoderConfig->sba_planar;

        ivas_total_brate = ( (Encoder_Struct *) st_ivas )->hEncoderConfig->ivas_total_brate;

        Fs = ( (Encoder_Struct *) st_ivas )->hEncoderConfig->input_Fs;
@@ -107,7 +111,11 @@ ivas_error ivas_dirac_config( 
        nCPE = &( (Decoder_Struct *) st_ivas )->nCPE;

        element_mode = &( (Decoder_Struct *) st_ivas )->element_mode_init;

        nchan_transport = &( (Decoder_Struct *) st_ivas )->nchan_transport;

#ifndef SBA_ORDER_BITSTREAM

        sba_order = ( (Decoder_Struct *) st_ivas )->sba_order;

#else

        sba_order = ( (Decoder_Struct *) st_ivas )->sba_analysis_order;

#endif

        sba_planar = ( (Decoder_Struct *) st_ivas )->sba_planar;

        ivas_total_brate = ( (Decoder_Struct *) st_ivas )->hDecoderConfig->ivas_total_brate;

        Fs = ( (Decoder_Struct *) st_ivas )->hDecoderConfig->output_Fs;

lib_com/ivas_prot.h

+8 −10
Original line number Diff line number Diff line
@@ -2155,9 +2155,7 @@ void stereo_mdct_core_dec( 
void splitAvailableBits(

    const int16_t total_bits,                                   /* i  : total available bits for TCX coding     */

    const int16_t split_ratio,                                  /* i  : split ratio                             */

#ifdef LBR_SBA_CORE_CODING_TUNING

    const int16_t isSBAStereoMode,                              /* i  : signal core coding for sba              */

#endif

    int16_t *bits_ch0,                                          /* o  : bits for channel 0                      */

    int16_t *bits_ch1                                           /* o  : bits for channel 1                      */

);
@@ -2175,9 +2173,7 @@ void parse_stereo_from_bitstream( 
    STEREO_MDCT_DEC_DATA_HANDLE hStereoMdct,                    /* i/o: MDCT stereo decoder structure           */

    Decoder_State **sts,                                        /* i/o: decoder state structure                 */

    const int16_t mct_on,                                       /* i  : flag mct block (1) or stereo (0)        */

#ifdef LBR_SBA_CORE_CODING_TUNING

    const int16_t isSBAStereoMode,                              /* i: flag core coding for sba         */

#endif

    Decoder_State *st0,                                         /* i/o: decoder state structure for Bstr        */

    int16_t ms_mask[NB_DIV][MAX_SFB]                            /* o  : bandwise MS mask                        */

);
@@ -2513,10 +2509,8 @@ ivas_error stereo_memory_enc( 
    const int16_t max_bwidth,                                   /* i  : maximum audio bandwidth                 */

    float *tdm_last_ratio,                                      /* o  : TD stereo last ratio                    */

    const IVAS_FORMAT ivas_format                               /* i  : IVAS format                             */

#ifdef LBR_SBA_CORE_CODING_TUNING

    ,

    const int16_t nchan_transport                               /* i  : number transport chans                  */

#endif



);
@@ -3013,11 +3007,7 @@ void ivas_dirac_param_est_enc( 
 *----------------------------------------------------------------------------------*/



/*! r: SBA format mode */

#ifdef LBR_SBA_CORE_CODING_TUNING

SBA_MODE ivas_sba_mode_select( 

#else

int16_t ivas_sba_mode_select( 

#endif

    const int32_t ivas_total_brate                              /* i  : IVAS total bitrate                      */

);
@@ -3876,6 +3866,10 @@ void ivas_spar_dec_upmixer( 
ivas_error ivas_spar_md_enc_open(

    ivas_spar_md_enc_state_t **hMdEnc,                          /* i/o: SPAR MD encoder handle                  */

    const ENCODER_CONFIG_HANDLE hEncoderConfig                  /* i  : configuration structure                 */

#ifdef SBA_ORDER_BITSTREAM

    ,

    int16_t sba_order

#endif

);



void ivas_spar_md_enc_close(
@@ -3888,6 +3882,10 @@ ivas_error ivas_spar_md_enc_process( 
    ivas_spar_md_enc_in_buf_t *pIn_buf,

    BSTR_ENC_HANDLE hMetaData,                                  /* i/o: MetaData handle                         */

    const int16_t dtx_silence_mode

#ifdef SBA_ORDER_BITSTREAM

    ,

    int16_t sba_order

#endif

);



void ivas_compute_spar_params(

lib_com/ivas_stereo_mdct_stereo_com.c

+0 −4
Original line number Diff line number Diff line
@@ -46,9 +46,7 @@ 
void splitAvailableBits(

    const int16_t total_bits,      /* i  : total available bits for TCX coding   */

    const int16_t split_ratio,     /* i  : split ratio                           */

#ifdef LBR_SBA_CORE_CODING_TUNING

    const int16_t isSBAStereoMode, /* i  : signal core coding for sba            */

#endif

    int16_t *bits_ch0,             /* o  : bits for channel 0                    */

    int16_t *bits_ch1              /* o  : bits for channel 1                    */

)
@@ -56,13 +54,11 @@ void splitAvailableBits( 
    assert( split_ratio >= 1 && split_ratio < SMDCT_BITRATE_RATIO_RANGE );



    *bits_ch0 = split_ratio * total_bits / SMDCT_BITRATE_RATIO_RANGE;

#ifdef LBR_SBA_CORE_CODING_TUNING

    /* for SBA mode bias the distribution towards the W channel */

    if ( split_ratio < 7 && isSBAStereoMode )

    {

        *bits_ch0 += (int16_t) ( 0.2 * *bits_ch0 );

    }

#endif

    *bits_ch1 = total_bits - *bits_ch0;



    return;