Commit c42030eb authored by vaclav's avatar vaclav
Browse files

Harmonize ISM MD public structures.

parent e5f1900f

apps/renderer.c

+8 −9
Original line number Diff line number Diff line
@@ -100,7 +100,7 @@ typedef struct 
    uint16_t numObjects;

    IsmFileReader *ismReaders[RENDERER_MAX_ISM_INPUTS];

    uint32_t numPositions[RENDERER_MAX_ISM_INPUTS];

    IVAS_REND_AudioObjectPosition *positions[RENDERER_MAX_ISM_INPUTS]; /* size: [RENDERER_MAX_ISM_INPUTS][numPositions[object_index]] */

    IVAS_ISM_METADATA *positions[RENDERER_MAX_ISM_INPUTS]; /* size: [RENDERER_MAX_ISM_INPUTS][numPositions[object_index]] */

    uint16_t *positionDurations[RENDERER_MAX_ISM_INPUTS];  /* size: [RENDERER_MAX_ISM_INPUTS][numPositions[object_index]] */

    uint32_t currentPositionIdxs[RENDERER_MAX_ISM_INPUTS]; /* Index of current position as listed in the metadata file */

    uint16_t durationCounters[RENDERER_MAX_ISM_INPUTS];    /* Number of frames spent at current position */
@@ -123,7 +123,7 @@ typedef struct 


typedef struct

{

    IVAS_REND_AudioObjectPosition positions[RENDERER_MAX_ISM_INPUTS];

    IVAS_ISM_METADATA positions[RENDERER_MAX_ISM_INPUTS];

    int16_t numObjects;

} ObjectPositionBuffer;
@@ -437,7 +437,7 @@ static int8_t parseUint32( const char *line, uint32_t *ret ); 


static int8_t parseInt32( const char *line, int32_t *ret );



static void parseObjectPosition( char *line, IVAS_REND_AudioObjectPosition *position, uint16_t *positionDuration );

static void parseObjectPosition( char *line, IVAS_ISM_METADATA *position, uint16_t *positionDuration );



static void parseMetadata( char *metadataString, char *inDir, InputConfig *inConfig, IsmPositionProvider *positionProvider, MasaFileReader **masaReaders, LfeRoutingConfig **lfeRoutingConfigs );
@@ -3274,14 +3274,13 @@ static void parseOptionalInputValues( 


static void parseObjectPosition(

    char *line,

    IVAS_REND_AudioObjectPosition *position,

    IVAS_ISM_METADATA *position,

    uint16_t *positionDuration )

{

    char *endptr;

    int16_t read_values;

    float meta_prm[8] = { 0.0f, 0.0f, 1.0f, 0.0f, 1.0f, 0.0f, 0.0f, 0.0f };





    readNextMetadataChunk( line, "," );

    *positionDuration = (uint16_t) strtol( line, &endptr, 10 );

    readNextMetadataChunk( line, "\n" );
@@ -3323,7 +3322,7 @@ static void parseIsm( 
    if ( parseUint32( line, &numberOfObjectPositionsToRead ) == 0 )

    {

        positionProvider->numPositions[idx] = numberOfObjectPositionsToRead;

        positionProvider->positions[idx] = malloc( numberOfObjectPositionsToRead * sizeof( IVAS_REND_AudioObjectPosition ) );

        positionProvider->positions[idx] = malloc( numberOfObjectPositionsToRead * sizeof( IVAS_ISM_METADATA ) );

        positionProvider->positionDurations[idx] = malloc( numberOfObjectPositionsToRead * sizeof( uint16_t ) );



        for ( i = 0; i < numberOfObjectPositionsToRead; ++i )

lib_com/common_api_types.h

+106 −102
Original line number Diff line number Diff line
@@ -48,6 +48,7 @@ 
#define IVAS_MAX_FRAME_SIZE     ( 48000 / IVAS_NUM_FRAMES_PER_SEC )



#define IVAS_MAX_BITS_PER_FRAME ( 512000 / IVAS_NUM_FRAMES_PER_SEC )



#define IVAS_MAX_NUM_OBJECTS        4

#define IVAS_MAX_INPUT_CHANNELS     16

#define IVAS_MAX_OUTPUT_CHANNELS    16
@@ -55,14 +56,9 @@ 
#define IVAS_CLDFB_NO_CHANNELS_MAX  60

#define IVAS_MAX_INPUT_LFE_CHANNELS 4





#define IVAS_MAX_PARAM_SPATIAL_SUBFRAMES 4



#define IVAS_ROOM_ABS_COEFF 6

#ifdef SPLIT_REND_WITH_HEAD_ROT

#define IVAS_MAX_SPLIT_REND_BITRATE                   768000

#define IVAS_MAX_SPLIT_REND_BITS_BUFFER_SIZE_IN_BYTES ( ( ( (int32_t) IVAS_MAX_SPLIT_REND_BITRATE / IVAS_NUM_FRAMES_PER_SEC ) + 7 ) >> 3 )

#endif



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

 * Common API enum for output audio configurations
@@ -100,7 +96,7 @@ typedef enum _IVAS_AUDIO_CONFIG 




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

 * Common API structures

 * Common API structures and enums

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



typedef enum _IVAS_ENC_FEC_INDICATOR
@@ -108,6 +104,7 @@ typedef enum _IVAS_ENC_FEC_INDICATOR 
    IVAS_ENC_FEC_LO,

    IVAS_ENC_FEC_HI,

    IVAS_ENC_FEC_UNDEFINED = 0xffff



} IVAS_ENC_FEC_INDICATOR;



typedef struct _IVAS_ENC_CHANNEL_AWARE_CONFIG
@@ -115,6 +112,7 @@ typedef struct _IVAS_ENC_CHANNEL_AWARE_CONFIG 
    int16_t channelAwareModeEnabled;

    IVAS_ENC_FEC_INDICATOR fec_indicator;

    int16_t fec_offset;



} IVAS_ENC_CHANNEL_AWARE_CONFIG;



typedef struct _IVAS_ISM_METADATA
@@ -127,6 +125,7 @@ typedef struct _IVAS_ISM_METADATA 
    float yaw;

    float pitch;

    int16_t non_diegetic_flag;



} IVAS_ISM_METADATA;



typedef struct
@@ -135,7 +134,69 @@ typedef struct 


} IVAS_QUATERNION;



typedef struct

{

    float x, y, z;



} IVAS_VECTOR3;



typedef enum

{

    IVAS_HEAD_ORIENT_TRK_NONE,

    IVAS_HEAD_ORIENT_TRK_REF,

    IVAS_HEAD_ORIENT_TRK_AVG,

    IVAS_HEAD_ORIENT_TRK_REF_VEC,

    IVAS_HEAD_ORIENT_TRK_REF_VEC_LEV



} IVAS_HEAD_ORIENT_TRK_T;



typedef struct ivas_masa_metadata_frame_struct *IVAS_MASA_METADATA_HANDLE;

typedef struct ivas_masa_decoder_ext_out_meta_struct *IVAS_MASA_DECODER_EXT_OUT_META_HANDLE;



typedef struct TDREND_HRFILT_FiltSet_struct *IVAS_DEC_HRTF_HANDLE;

typedef struct ivas_hrtfs_crend_structure *IVAS_DEC_HRTF_CREND_HANDLE;

typedef struct ivas_hrtfs_fastconv_struct *IVAS_DEC_HRTF_FASTCONV_HANDLE;

typedef struct ivas_hrtfs_parambin_struct *IVAS_DEC_HRTF_PARAMBIN_HANDLE;

typedef struct cldfb_filter_bank_struct *IVAS_CLDFB_FILTER_BANK_HANDLE;



typedef struct ivas_LS_setup_custom *IVAS_LSSETUP_CUSTOM_HANDLE;

typedef struct ivas_LS_setup_custom IVAS_LSSETUP_CUSTOM_STRUCT;



typedef struct _IVAS_LS_CUSTOM_LAYOUT

{

    int16_t num_spk;

    float azimuth[IVAS_MAX_OUTPUT_CHANNELS];

    float elevation[IVAS_MAX_OUTPUT_CHANNELS];

    int16_t num_lfe;

    int16_t lfe_idx[IVAS_MAX_OUTPUT_CHANNELS];



} IVAS_CUSTOM_LS_DATA;



typedef struct _IVAS_JBM_TRACE_DATA

{

    uint32_t systemTimestamp_ms;

    uint16_t extBufferedSamples;

    uint16_t lastDecodedWasActive;

    int32_t output_Fs;

    int16_t dataUnit_flag;

    uint16_t sequenceNumber;

    uint32_t timeStamp;

    uint32_t rcvTime;



    int16_t partial_frame;

    int16_t partialCopyOffset;



} IVAS_JBM_TRACE_DATA;





#ifdef SPLIT_REND_WITH_HEAD_ROT

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

 * Split rendering API constants, structures, and enums

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



#define IVAS_MAX_SPLIT_REND_BITRATE                   768000

#define IVAS_MAX_SPLIT_REND_BITS_BUFFER_SIZE_IN_BYTES ( ( ( (int32_t) IVAS_MAX_SPLIT_REND_BITRATE / IVAS_NUM_FRAMES_PER_SEC ) + 7 ) >> 3 )



typedef enum

{

    DEFAULT_AXIS,
@@ -145,12 +206,14 @@ typedef enum 
    YAW_PITCH,

    YAW_ROLL,

    PITCH_ROLL



} IVAS_SPLIT_REND_ROT_AXIS;



typedef enum

{

    IVAS_SPLIT_REND_POSE_CORRECTION_MODE_NONE,

    IVAS_SPLIT_REND_POSE_CORRECTION_MODE_CLDFB,



} IVAS_SPLIT_REND_POSE_CORRECTION_MODE;



typedef enum
@@ -159,6 +222,7 @@ typedef enum 
    IVAS_SPLIT_REND_CODEC_LC3PLUS,

    IVAS_SPLIT_REND_CODEC_DEFAULT, /* Will use LCLD for CLDFB rendering paths and LC3plus for TD rendering paths */

    IVAS_SPLIT_REND_CODEC_NONE



} IVAS_SPLIT_REND_CODEC;



typedef enum
@@ -168,6 +232,7 @@ typedef enum 
    IVAS_SPLIT_REND_RENDERER_SELECTION_PARAMBIN,

    IVAS_SPLIT_REND_RENDERER_SELECTION_TDREND,

    IVAS_SPLIT_REND_RENDERER_SELECTION_DEFAULT,



} IVAS_SPLIT_REND_RENDERER_SELECTION;



typedef struct ivas_split_rend_bits_t
@@ -179,33 +244,32 @@ typedef struct ivas_split_rend_bits_t 
    int16_t codec_frame_size_ms;

    IVAS_SPLIT_REND_CODEC codec;

    IVAS_SPLIT_REND_POSE_CORRECTION_MODE pose_correction;

} ivas_split_rend_bits_t, IVAS_SPLIT_REND_BITS_DATA, *IVAS_SPLIT_REND_BITS_HANDLE;

#endif



typedef struct

{

    float x, y, z;

} IVAS_VECTOR3;

} ivas_split_rend_bits_t, IVAS_SPLIT_REND_BITS_DATA, *IVAS_SPLIT_REND_BITS_HANDLE;



typedef enum

typedef struct _IVAS_SPLIT_REND_CONFIG

{

    IVAS_HEAD_ORIENT_TRK_NONE,

    IVAS_HEAD_ORIENT_TRK_REF,

    IVAS_HEAD_ORIENT_TRK_AVG,

    IVAS_HEAD_ORIENT_TRK_REF_VEC,

    IVAS_HEAD_ORIENT_TRK_REF_VEC_LEV



} IVAS_HEAD_ORIENT_TRK_T;



typedef struct ivas_masa_metadata_frame_struct *IVAS_MASA_METADATA_HANDLE;

typedef struct ivas_masa_decoder_ext_out_meta_struct *IVAS_MASA_DECODER_EXT_OUT_META_HANDLE;

    int32_t splitRendBitRate;    /*Bit rate for split rendering mode, if "pcm_out" is set then "splitRendBitRate" is used as a limit for MD bitrate */

    int16_t hq_mode;             /*High quality 3DOF mode with additional side information. Requires more pre-renditions. */

    int16_t dof;                 /*flag to specify if pose correction is needed for 1, 2 or 3 degree of freedoms*/

                                 /*The axis can be set dynamically per frame based on a file input */

                                 /*possible values:

                                 1 - (1dof correction. By default YAW correction)

                                 2 - (2dof correction. By default YAW and PITCH correction)

                                 3 - (3dof correction. By default YAW, PITCH and ROLL correction)

                                 */

    int16_t codec_delay_ms;      /*PLACEHOLDER (currently being ignored) : look ahead delay of the codec that is used to code BIN signal output of pre-renderer*/

    int16_t codec_frame_size_ms; /*Codec frame size in milliseconds, only relevant with LC3plus */

    IVAS_SPLIT_REND_POSE_CORRECTION_MODE poseCorrectionMode;

    IVAS_SPLIT_REND_CODEC codec;

    IVAS_SPLIT_REND_RENDERER_SELECTION rendererSelection;



typedef struct TDREND_HRFILT_FiltSet_struct *IVAS_DEC_HRTF_HANDLE;

typedef struct ivas_hrtfs_crend_structure *IVAS_DEC_HRTF_CREND_HANDLE;

typedef struct ivas_hrtfs_fastconv_struct *IVAS_DEC_HRTF_FASTCONV_HANDLE;

typedef struct ivas_hrtfs_parambin_struct *IVAS_DEC_HRTF_PARAMBIN_HANDLE;

} IVAS_SPLIT_REND_CONFIG_DATA;

#endif



typedef struct cldfb_filter_bank_struct *IVAS_CLDFB_FILTER_BANK_HANDLE;

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

 * Renderer API structures and enums

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



#ifdef DEBUGGING

typedef enum
@@ -213,20 +277,10 @@ typedef enum 
    IVAS_RENDER_TYPE_OVERRIDE_NONE,

    IVAS_RENDER_TYPE_OVERRIDE_CREND,

    IVAS_RENDER_TYPE_OVERRIDE_FASTCONV



} IVAS_RENDER_TYPE_OVERRIDE;

#endif



typedef struct

{

    float azimuth;

    float elevation;

    float radius;

    float yaw;

    float pitch;

    int16_t non_diegetic_flag;



} IVAS_REND_AudioObjectPosition;



typedef struct _IVAS_ROOM_ACOUSTICS_CONFIG

{

    int16_t override;
@@ -246,27 +300,6 @@ typedef struct _IVAS_ROOM_ACOUSTICS_CONFIG 


} IVAS_ROOM_ACOUSTICS_CONFIG_DATA;



#ifdef SPLIT_REND_WITH_HEAD_ROT



typedef struct _IVAS_SPLIT_REND_CONFIG

{

    int32_t splitRendBitRate;    /*Bit rate for split rendering mode, if "pcm_out" is set then "splitRendBitRate" is used as a limit for MD bitrate */

    int16_t hq_mode;             /*High quality 3DOF mode with additional side information. Requires more pre-renditions. */

    int16_t dof;                 /*flag to specify if pose correction is needed for 1, 2 or 3 degree of freedoms*/

                                 /*The axis can be set dynamically per frame based on a file input */

                                 /*possible values:

                                 1 - (1dof correction. By default YAW correction)

                                 2 - (2dof correction. By default YAW and PITCH correction)

                                 3 - (3dof correction. By default YAW, PITCH and ROLL correction)

                                 */

    int16_t codec_delay_ms;      /*PLACEHOLDER (currently being ignored) : look ahead delay of the codec that is used to code BIN signal output of pre-renderer*/

    int16_t codec_frame_size_ms; /*Codec frame size in milliseconds, only relevant with LC3plus */

    IVAS_SPLIT_REND_POSE_CORRECTION_MODE poseCorrectionMode;

    IVAS_SPLIT_REND_CODEC codec;

    IVAS_SPLIT_REND_RENDERER_SELECTION rendererSelection;

} IVAS_SPLIT_REND_CONFIG_DATA;

#endif



typedef struct _IVAS_RENDER_CONFIG

{

#ifdef DEBUGGING
@@ -277,36 +310,7 @@ typedef struct _IVAS_RENDER_CONFIG 
    IVAS_SPLIT_REND_CONFIG_DATA split_rend_config;

#endif

    float directivity[IVAS_MAX_NUM_OBJECTS * 3];

} IVAS_RENDER_CONFIG_DATA, *IVAS_RENDER_CONFIG_HANDLE;



typedef struct _IVAS_LS_CUSTOM_LAYOUT

{

    int16_t num_spk;

    float azimuth[IVAS_MAX_OUTPUT_CHANNELS];

    float elevation[IVAS_MAX_OUTPUT_CHANNELS];

    int16_t num_lfe;

    int16_t lfe_idx[IVAS_MAX_OUTPUT_CHANNELS];



} IVAS_CUSTOM_LS_DATA;



typedef struct ivas_LS_setup_custom *IVAS_LSSETUP_CUSTOM_HANDLE;

typedef struct ivas_LS_setup_custom IVAS_LSSETUP_CUSTOM_STRUCT;





typedef struct _IVAS_JBM_TRACE_DATA

{

    uint32_t systemTimestamp_ms;

    uint16_t extBufferedSamples;

    uint16_t lastDecodedWasActive;

    int32_t output_Fs;

    int16_t dataUnit_flag;

    uint16_t sequenceNumber;

    uint32_t timeStamp;

    uint32_t rcvTime;



    int16_t partial_frame;

    int16_t partialCopyOffset;



} IVAS_JBM_TRACE_DATA;

} IVAS_RENDER_CONFIG_DATA, *IVAS_RENDER_CONFIG_HANDLE;



#endif /* COMMON_API_TYPES_H */

lib_com/ivas_error_utils.h

+3 −4
Original line number Diff line number Diff line
@@ -30,14 +30,13 @@ 


*******************************************************************************************************/



/* options.h needed for debugging/development features

 * It should be stripped for delivery along with debugging switches */

#include "options.h"



#include "ivas_error.h"

#include <assert.h>

#include <stdarg.h>

#include <stdint.h>



#include "ivas_error.h"



#ifdef DEBUGGING

#include <stdio.h>

#endif

lib_rend/ivas_objectRenderer.c

+1 −3
Original line number Diff line number Diff line
@@ -674,7 +674,7 @@ ivas_error ivas_td_binaural_renderer_ext( 
    const AUDIO_CONFIG inConfig,                                 /* i  : Input audio configuration               */

    const LSSETUP_CUSTOM_STRUCT *customLsInput,                  /* i  : Input custom loudspeaker layout         */

    const COMBINED_ORIENTATION_HANDLE *hCombinedOrientationData, /* i  : Combined head and external orientations */

    const IVAS_REND_AudioObjectPosition *currentPos,             /* i  : Object position                         */

    const IVAS_ISM_METADATA *currentPos,                         /* i  : Object position                         */

    const REVERB_HANDLE hReverb,                                 /* i  : Reverberator handle                     */

    const int16_t ism_md_subframe_update_ext,                    /* i  : Metadata Delay in subframes to sync with audio delay */

    const int32_t output_Fs,                                     /* i  : output sampling rate                                    */
@@ -699,12 +699,10 @@ ivas_error ivas_td_binaural_renderer_ext( 
    }



    push_wmops( "ivas_td_binaural_renderer_ext" );



    inConfigType = getAudioConfigType( inConfig );

    lfe_idx = LFE_CHANNEL;

    hIsmMetaData[0] = NULL;





    if ( inConfigType == IVAS_REND_AUDIO_CONFIG_TYPE_CHANNEL_BASED )

    {

        ivas_format = MC_FORMAT;

lib_rend/ivas_prot_rend.h

+1 −1
Original line number Diff line number Diff line
@@ -568,7 +568,7 @@ ivas_error ivas_td_binaural_renderer_ext( 
    const AUDIO_CONFIG inConfig,                                /* i  : Input audio configuration               */

    const LSSETUP_CUSTOM_STRUCT *customLsInput,                 /* i  : Input custom loudspeaker layout         */

    const COMBINED_ORIENTATION_HANDLE *hCombinedOrientationData,/* i  : Combined head and external orientations */

    const IVAS_REND_AudioObjectPosition *currentPos,            /* i  : Object position                         */

    const IVAS_ISM_METADATA *currentPos,            /* i  : Object position                         */

    const REVERB_HANDLE hReverb,                                /* i  : Reverberator handle                     */

    const int16_t ism_md_subframe_update_ext,                   /* i  : Metadata Delay in subframes to sync with audio delay */

    const int32_t output_Fs,                                    /* i  : output sampling rate                    */