Commit 3bdd36d2 authored by sagnowski's avatar sagnowski
Browse files

Keep head poses in ring buffer to persist data between calls to... 

Keep head poses in ring buffer to persist data between calls to IVAS_DEC_VoIP_GetSplitBinauralBitstream
parent d66368c9

lib_dec/ivas_stat_dec.h

+1 −0
Original line number Diff line number Diff line
@@ -849,6 +849,7 @@ typedef struct 
typedef struct

{

#ifdef FIX_1119_SPLIT_RENDERING_VOIP

    TD_RINGBUF_HANDLE hMultiBinTdData;

    CLDFB_RINGBUF_HANDLE hMultiBinCldfbData[MAX_HEAD_ROT_POSES * BINAURAL_CHANNELS];

#else

    ISAR_DEC_SPLIT_REND_MULTI_BIN_CLDFB_DATA_HANDLE hMultiBinCldfbData; /*scratch buffer for frame by frame processing*/

lib_dec/lib_dec.c

+61 −22
Original line number Diff line number Diff line
@@ -129,6 +129,10 @@ static ivas_error ivas_dec_init_split_rend( Decoder_Struct *st_ivas ); 
static ivas_error ivas_create_handle_isar( ISAR_DEC_SPLIT_REND_WRAPPER_HANDLE *hSplitBinRend_out );

static void ivas_destroy_handle_isar( ISAR_DEC_SPLIT_REND_WRAPPER_HANDLE *hSplitBinRend_out );

static int16_t get_render_frame_size_ms( IVAS_RENDER_FRAMESIZE render_framesize );

#ifdef FIX_1119_SPLIT_RENDERING_VOIP

static int16_t get_render_frame_size_samples( Decoder_Struct *st_ivas );

static int16_t ivas_dec_split_rend_cldfb_in( Decoder_Struct* st_ivas );

#endif

static void update_voip_rendered20ms( IVAS_DEC_HANDLE hIvasDec, const int16_t nSamplesRendered );
@@ -637,6 +641,13 @@ ivas_error IVAS_DEC_GetRenderFramesize( 
    return IVAS_ERR_OK;

}



#ifdef FIX_1119_SPLIT_RENDERING_VOIP

static int16_t get_render_frame_size_samples( Decoder_Struct *st_ivas )

{

    return (int16_t) ( st_ivas->hDecoderConfig->output_Fs * st_ivas->hDecoderConfig->render_framesize / ( FRAMES_PER_SEC * IVAS_MAX_PARAM_SPATIAL_SUBFRAMES ) );

}

#endif



/*---------------------------------------------------------------------*

 * IVAS_DEC_GetGetRenderFramesizeSamples( )

 *
@@ -653,7 +664,11 @@ ivas_error IVAS_DEC_GetRenderFramesizeSamples( 
        return IVAS_ERR_UNEXPECTED_NULL_POINTER;

    }



#ifdef FIX_1119_SPLIT_RENDERING_VOIP

    *render_framesize = get_render_frame_size_samples( hIvasDec->st_ivas );

#else

    *render_framesize = (int16_t) ( hIvasDec->st_ivas->hDecoderConfig->output_Fs * hIvasDec->st_ivas->hDecoderConfig->render_framesize / ( FRAMES_PER_SEC * IVAS_MAX_PARAM_SPATIAL_SUBFRAMES ) );

#endif



    return IVAS_ERR_OK;

}
@@ -1866,14 +1881,12 @@ static int16_t isar_get_frame_size( 
static ivas_error isar_render_poses(

    IVAS_DEC_HANDLE hIvasDec,    /* i/o: IVAS decoder handle                                                     */

    const int16_t nSamplesAsked, /* i  : number of samples wanted by the caller                                  */

    float **p_head_pose_buf,

    int16_t *nOutSamples, /* o  : number of samples per channel written to output buffer                  */

    bool *needNewFrame    /* o  : indication that the decoder needs a new frame                           */

)

{

    Decoder_Struct *st_ivas;

    float pcmBuf[BINAURAL_CHANNELS * MAX_HEAD_ROT_POSES * L_FRAME48k];

    int16_t i, j;

    ivas_error error;

    ISAR_DEC_SPLIT_REND_WRAPPER_HANDLE hSplitBinRend;

    int16_t numPoses;
@@ -1907,13 +1920,9 @@ static ivas_error isar_render_poses( 
        return IVAS_ERR_OK;

    }



    /* change buffer layout */

    for ( i = 0; i < *nOutSamples; ++i )

    {

        for ( j = 0; j < BINAURAL_CHANNELS * numPoses; ++j )

    if ( !ivas_dec_split_rend_cldfb_in( st_ivas ) )

    {

            p_head_pose_buf[j][i] = pcmBuf[i * BINAURAL_CHANNELS * numPoses + j];

        }

        ivas_TD_RINGBUF_PushInterleaved( hIvasDec->st_ivas->hSplitBinRend->hMultiBinTdData, pcmBuf, *nOutSamples );

    }



    return error;
@@ -2057,12 +2066,7 @@ ivas_error IVAS_DEC_GetSplitBinauralBitstream( 
    pcm_out_flag = ( st_ivas->hDecoderConfig->output_config == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM ) ? 1 : 0;

    numSamplesPerChannelToOutput = isar_get_frame_size( st_ivas );



    for ( i = 0; i < BINAURAL_CHANNELS * MAX_HEAD_ROT_POSES; ++i )

    {

        p_head_pose_buf[i] = head_pose_buf[i];

    }



    error = isar_render_poses( hIvasDec, numSamplesPerChannelToOutput, p_head_pose_buf, nOutSamples, needNewFrame );

    error = isar_render_poses( hIvasDec, numSamplesPerChannelToOutput, nOutSamples, needNewFrame );

    if ( error != IVAS_ERR_OK )

    {

        return error;
@@ -2072,6 +2076,16 @@ ivas_error IVAS_DEC_GetSplitBinauralBitstream( 
        return IVAS_ERR_OK;

    }



    for ( i = 0; i < BINAURAL_CHANNELS * MAX_HEAD_ROT_POSES; ++i )

    {

        p_head_pose_buf[i] = head_pose_buf[i];

    }



    if ( !ivas_dec_split_rend_cldfb_in( st_ivas ) )

    {

        ivas_TD_RINGBUF_PopChannels( hIvasDec->st_ivas->hSplitBinRend->hMultiBinTdData, p_head_pose_buf, *nOutSamples );

    }



    error = isar_generate_metadata_and_bitstream( hIvasDec, p_head_pose_buf, splitRendBits );

    if ( error != IVAS_ERR_OK )

    {
@@ -3932,14 +3946,8 @@ ivas_error IVAS_DEC_VoIP_GetSamples 
#ifdef FIX_1119_SPLIT_RENDERING_VOIP

            if ( hIvasDec->st_ivas->hDecoderConfig->output_config == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED || hIvasDec->st_ivas->hDecoderConfig->output_config == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM )

            {

                /* Move output pointers forward */

                for ( i = 0; i < BINAURAL_CHANNELS * MAX_HEAD_ROT_POSES; ++i )

                {

                    p_head_pose_buf[i] = &head_pose_buf[i][*nSamplesRendered];

                }



                /* Render head poses from time-scaled transport channels */

                if ( ( error = isar_render_poses( hIvasDec, nSamplesToRender, p_head_pose_buf, &nSamplesRendered_loop, &tmp ) ) != IVAS_ERR_OK )

                if ( ( error = isar_render_poses( hIvasDec, nSamplesToRender, &nSamplesRendered_loop, &tmp ) ) != IVAS_ERR_OK )

                {

                    return error;

                }
@@ -3966,12 +3974,16 @@ ivas_error IVAS_DEC_VoIP_GetSamples 
         ( hIvasDec->st_ivas->hDecoderConfig->output_config == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED ||

           hIvasDec->st_ivas->hDecoderConfig->output_config == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM ) )

    {

        /* Set pointers to beginning of head pose buffers */

        for ( i = 0; i < BINAURAL_CHANNELS * MAX_HEAD_ROT_POSES; ++i )

        {

            p_head_pose_buf[i] = head_pose_buf[i];

        }



        if ( !ivas_dec_split_rend_cldfb_in( st_ivas ) )

        {

            ivas_TD_RINGBUF_PopChannels( hIvasDec->st_ivas->hSplitBinRend->hMultiBinTdData, p_head_pose_buf, *nSamplesRendered );

        }



        /* Analyse head poses over entire frame, generate ISAR metadata and maybe encode if split coded */

        error = isar_generate_metadata_and_bitstream( hIvasDec, p_head_pose_buf, splitRendBits );

        if ( error != IVAS_ERR_OK )
@@ -5186,6 +5198,7 @@ static ivas_error ivas_create_handle_isar( 
    isar_init_split_rend_handles( &hSplitBinRend->splitrend );



#ifdef FIX_1119_SPLIT_RENDERING_VOIP

    hSplitBinRend->hMultiBinTdData = NULL;

    for ( i = 0; i < (int16_t) ( MAX_HEAD_ROT_POSES * BINAURAL_CHANNELS ); ++i )

    {

        hSplitBinRend->hMultiBinCldfbData[i] = NULL;
@@ -5219,6 +5232,10 @@ static void ivas_destroy_handle_isar( 
    if ( *hSplitBinRend != NULL )

    {

#ifdef FIX_1119_SPLIT_RENDERING_VOIP

        if ( ( *hSplitBinRend )->hMultiBinTdData != NULL )

        {

            ivas_TD_RINGBUF_Close( &( *hSplitBinRend )->hMultiBinTdData );

        }

        for ( i = 0; i < (int16_t) ( MAX_HEAD_ROT_POSES * BINAURAL_CHANNELS ); ++i )

        {

            if ( ( *hSplitBinRend )->hMultiBinCldfbData[i] != NULL )
@@ -5403,6 +5420,16 @@ static ivas_error ivas_dec_reconfig_split_rend( 
    return IVAS_ERR_OK;

}



#ifdef FIX_1119_SPLIT_RENDERING_VOIP

static int16_t ivas_dec_split_rend_cldfb_in( Decoder_Struct* st_ivas )

{



    return st_ivas->renderer_type == RENDERER_BINAURAL_FASTCONV ||

        st_ivas->renderer_type == RENDERER_BINAURAL_FASTCONV_ROOM ||

        st_ivas->renderer_type == RENDERER_BINAURAL_PARAMETRIC ||

        st_ivas->renderer_type == RENDERER_BINAURAL_PARAMETRIC_ROOM;

}

#endif



/*-------------------------------------------------------------------*

 * ivas_dec_init_split_rend()
@@ -5424,6 +5451,9 @@ static ivas_error ivas_dec_init_split_rend( 
    pcm_out_flag = ( st_ivas->hDecoderConfig->output_config == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM ) ? 1 : 0;

    cldfb_in_flag = 0;



#ifdef FIX_1119_SPLIT_RENDERING_VOIP

    cldfb_in_flag = ivas_dec_split_rend_cldfb_in( st_ivas );

#else

    if ( st_ivas->renderer_type == RENDERER_BINAURAL_FASTCONV ||

         st_ivas->renderer_type == RENDERER_BINAURAL_FASTCONV_ROOM ||

         st_ivas->renderer_type == RENDERER_BINAURAL_PARAMETRIC ||
@@ -5431,6 +5461,7 @@ static ivas_error ivas_dec_init_split_rend( 
    {

        cldfb_in_flag = 1;

    }

#endif



#ifdef FIX_1119_SPLIT_RENDERING_VOIP

    ISAR_PRE_REND_GetMultiBinPoseData( &st_ivas->hRenderConfig->split_rend_config, &st_ivas->hSplitBinRend->splitrend.multiBinPoseData, ( st_ivas->hHeadTrackData != NULL ) ? st_ivas->hHeadTrackData->sr_pose_pred_axis : DEFAULT_AXIS );
@@ -5449,6 +5480,14 @@ static ivas_error ivas_dec_init_split_rend( 
            }

        }

    }

    else

    {

        error = ivas_TD_RINGBUF_Open( &st_ivas->hSplitBinRend->hMultiBinTdData, get_render_frame_size_samples( st_ivas ), num_poses * BINAURAL_CHANNELS );

        if ( error != IVAS_ERR_OK )

        {

            return IVAS_ERROR( IVAS_ERR_FAILED_ALLOC, "Cannot allocate memory for split rendering structure" );

        }

    }

#else

    /* note: this is intra-frame heap memory */

    if ( ( st_ivas->hSplitBinRend->hMultiBinCldfbData = (ISAR_DEC_SPLIT_REND_MULTI_BIN_CLDFB_DATA_HANDLE) malloc( sizeof( ISAR_DEC_SPLIT_REND_MULTI_BIN_CLDFB_DATA ) ) ) == NULL )