Merge branch '500-head-tracking-api-causes-extra-15ms-of-delay' of... (8d942cfa) · Commits · IVAS Codec Public Collaboration / IVAS Codec

Workspace_msvc/lib_rend.vcxproj

+2 −0

Original line number	Diff line number	Diff line
		@@ -239,6 +239,7 @@
		<ClCompile Include="..\lib_rend\ivas_omasa_ana.c" />
		<ClCompile Include="..\lib_rend\ivas_orient_trk.c" />
		<ClCompile Include="..\lib_rend\ivas_output_init.c" />
		<ClCompile Include="..\lib_rend\ivas_reflections.c" />
		<ClCompile Include="..\lib_rend\ivas_render_config.c" />
		<ClCompile Include="..\lib_rend\ivas_reverb.c" />
		<ClCompile Include="..\lib_rend\ivas_reverb_delay_line.c" />
		@@ -251,6 +252,7 @@
		<ClCompile Include="..\lib_rend\ivas_rom_binaural_crend_head.c" />
		<ClCompile Include="..\lib_rend\ivas_rotation.c" />
		<ClCompile Include="..\lib_rend\ivas_rom_rend.c" />
		<ClCompile Include="..\lib_rend\ivas_shoebox.c" />
		<ClCompile Include="..\lib_rend\lib_rend.c" />
		</ItemGroup>
		<ItemGroup>

apps/decoder.c

+78 −6

Original line number	Diff line number	Diff line
		@@ -168,6 +168,13 @@ typedef struct
		#endif
		#endif

		#ifdef CONTROL_METADATA_REVERB
		uint16_t acousticEnvironmentId;
		#ifdef CONTROL_METADATA_DIRECTIVITY
		uint16_t directivityPatternId[IVAS_MAX_NUM_OBJECTS];
		#endif
		#endif

		} DecArguments;


		@@ -464,7 +471,11 @@ int main(
		goto cleanup;
		}
		#endif
		if ( RenderConfigReader_read( renderConfigReader, &renderConfigSplit ) != IVAS_ERR_OK )
		if ( RenderConfigReader_read( renderConfigReader,
		#ifdef CONTROL_METADATA_REVERB
		arg.renderConfigFilename,
		#endif
		&renderConfigSplit ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "Failed to read renderer configuration from file %s\n\n", arg.renderConfigFilename );
		goto cleanup;
		@@ -700,12 +711,42 @@ int main(
		goto cleanup;
		}

		#ifdef CONTROL_METADATA_REVERB
		if ( RenderConfigReader_read( renderConfigReader, arg.renderConfigFilename, &renderConfig ) != IVAS_ERR_OK )
		#else
		if ( RenderConfigReader_read( renderConfigReader, &renderConfig ) != IVAS_ERR_OK )
		#endif
		{
		fprintf( stderr, "Failed to read renderer configuration from file %s\n\n", arg.renderConfigFilename );
		goto cleanup;
		}
		#ifdef CONTROL_METADATA_DIRECTIVITY
		if ( ( error = RenderConfigReader_getDirectivity( renderConfigReader, arg.directivityPatternId, renderConfig.directivity ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "Failed to get directivity for objects: %d %d %d %d\n\n", arg.directivityPatternId[0], arg.directivityPatternId[1], arg.directivityPatternId[2], arg.directivityPatternId[3] );
		goto cleanup;
		}
		#endif

		#ifdef CONTROL_METADATA_REVERB
		if ( arg.outputFormat == IVAS_DEC_OUTPUT_BINAURAL_ROOM_REVERB )
		{
		if ( ( error = RenderConfigReader_getAcousticEnvironment( renderConfigReader, arg.acousticEnvironmentId, &renderConfig.room_acoustics ) ) == IVAS_ERR_OK )
		{
		if ( RenderConfigReader_checkValues( &renderConfig ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "Invalid reverberation configuration parameters\n\n" );
		goto cleanup;
		}
		}
		else if ( error != IVAS_ERR_ACOUSTIC_ENVIRONMENT_MISSING )
		{
		fprintf( stderr, "Failed to get acoustic environment with ID: %d\n\n", arg.acousticEnvironmentId );
		goto cleanup;
		}
		renderConfig.room_acoustics.override = true;
		}
		#endif
		if ( ( error = IVAS_DEC_FeedRenderConfig( hIvasDec, renderConfig ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nIVAS_DEC_FeedRenderConfig failed: %s\n\n", IVAS_DEC_GetErrorMessage( error ) );
		@@ -1071,7 +1112,15 @@ static bool parseCmdlIVAS_dec(
		#endif
		#endif


		#ifdef CONTROL_METADATA_REVERB
		arg->acousticEnvironmentId = 0;
		#ifdef CONTROL_METADATA_DIRECTIVITY
		for ( i = 0; i < IVAS_MAX_NUM_OBJECTS; ++i )
		{
		arg->directivityPatternId[i] = 0;
		}
		#endif
		#endif
		/-----------------------------------------------------------------
		* Initialization
		-----------------------------------------------------------------/
		@@ -1442,8 +1491,27 @@ static bool parseCmdlIVAS_dec(
		fprintf( stdout, "Complexity levels 1 and 2 will be defined after characterisation - default to level 3 (full functionality).\n" );
		}
		}


		#ifdef CONTROL_METADATA_REVERB
		else if ( strcmp( argv_to_upper, "-AEID" ) == 0 )
		{
		++i;
		arg->acousticEnvironmentId = (int16_t) atoi( argv[i++] );
		}
		#ifdef CONTROL_METADATA_DIRECTIVITY
		else if ( strcmp( argv_to_upper, "-DPID" ) == 0 )
		{
		++i;
		int16_t tmp;
		tmp = 0;
		while ( is_number( argv[i + tmp] ) && tmp < IVAS_MAX_NUM_OBJECTS )
		{
		arg->directivityPatternId[tmp] = (int16_t) atoi( argv[i + tmp] );
		++tmp;
		}
		i += tmp;
		}
		#endif
		#endif
		/-----------------------------------------------------------------
		* Option not recognized
		-----------------------------------------------------------------/
		@@ -1628,6 +1696,12 @@ static void usage_dec( void )
		fprintf( stdout, " Currently, all values default to level 3 (full functionality).\n" );
		#endif
		fprintf( stdout, "-exof File : External orientation file for external orientation trajectory\n" );
		#ifdef CONTROL_METADATA_DIRECTIVITY
		fprintf( stdout, "-dpid ID : Directivity pattern ID(s) (space-separated list of up to 4 numbers can be specified) for binaural output configuration\n" );
		#endif
		#ifdef CONTROL_METADATA_REVERB
		fprintf( stdout, "-aeid ID : Acoustic environment ID (number >= 0) for BINAURAL_ROOM_REVERB output configuration\n" );
		#endif
		#ifdef DEBUG_MODE_INFO
		#ifdef DEBUG_MODE_INFO_TWEAK
		fprintf( stdout, "-info <folder> : specify subfolder name for debug output\n" );
		@@ -1760,8 +1834,6 @@ static ivas_error initOnFirstGoodFrame(
		if ( *hSplitRendFileReadWrite != NULL )
		{
		IVAS_SPLIT_REND_BITS splitRendBitsZero;
		splitRendBitsZero.bits_written = 0;
		splitRendBitsZero.bits_read = 0;
		splitRendBitsZero.bits_buf = NULL;
		splitRendBitsZero.bits_read = 0;
		splitRendBitsZero.bits_written = 0;

apps/renderer.c

+72 −0

Original line number	Diff line number	Diff line
		@@ -181,6 +181,12 @@ typedef struct
		#ifdef FIX_488_SYNC_DELAY
		float syncMdDelay;
		#endif
		#ifdef CONTROL_METADATA_DIRECTIVITY
		uint16_t directivityPatternId[RENDERER_MAX_ISM_INPUTS];
		#endif
		#ifdef CONTROL_METADATA_REVERB
		uint16_t acousticEnvironmentId;
		#endif
		} CmdlnArgs;

		typedef enum
		@@ -215,6 +221,12 @@ typedef enum
		#ifdef FIX_488_SYNC_DELAY
		CmdLnOptionId_syncMdDelay,
		#endif
		#ifdef CONTROL_METADATA_DIRECTIVITY
		CmdLnOptionId_directivityPatternId,
		#endif
		#ifdef CONTROL_METADATA_REVERB
		CmdLnOptionId_acousticEnvironmentId
		#endif
		} CmdLnOptionId;

		static const CmdLnParser_Option cliOptions[] = {
		@@ -370,6 +382,22 @@ static const CmdLnParser_Option cliOptions[] = {
		.description = "Metadata Synchronization Delay in ms, Default is 0. Quantized by 5ms subframes for TDRenderer (13ms -> 10ms -> 2subframes)",
		},
		#endif
		#ifdef CONTROL_METADATA_DIRECTIVITY
		{
		.id = CmdLnOptionId_directivityPatternId,
		.match = "ism_directivity_pattern_id",
		.matchShort = "dpid",
		.description = "Directivity pattern ID(s) (space-separated list of up to 4 numbers can be specified) for binaural output configuration",
		},
		#endif
		#ifdef CONTROL_METADATA_REVERB
		{
		.id = CmdLnOptionId_acousticEnvironmentId,
		.match = "acoustic_environment_id",
		.matchShort = "aeid",
		.description = "Acoustic environment ID (number >= 0) for BINAURAL_ROOM_REVERB output configuration",
		},
		#endif
		};


		@@ -1094,7 +1122,11 @@ int main(
		exit( -1 );
		}

		#ifdef CONTROL_METADATA_REVERB
		if ( RenderConfigReader_read( renderConfigReader, args.renderConfigFilePath, &renderConfig ) != IVAS_ERR_OK )
		#else
		if ( RenderConfigReader_read( renderConfigReader, &renderConfig ) != IVAS_ERR_OK )
		#endif
		{
		fprintf( stderr, "Failed to read renderer configuration from file %s\n", args.renderConfigFilePath );
		exit( -1 );
		@@ -1102,6 +1134,21 @@ int main(

		if ( args.outConfig.audioConfig == IVAS_REND_AUDIO_CONFIG_BINAURAL_ROOM_REVERB )
		{
		#ifdef CONTROL_METADATA_REVERB
		if ( ( error = RenderConfigReader_getAcousticEnvironment( renderConfigReader, args.acousticEnvironmentId, &renderConfig.room_acoustics ) ) == IVAS_ERR_OK )
		{
		if ( RenderConfigReader_checkValues( &renderConfig ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "Invalid room acoustics configuration parameters\n\n" );
		exit( -1 );
		}
		}
		else
		{
		fprintf( stderr, "Failed to get acoustic environment with ID: %d\n\n", args.acousticEnvironmentId );
		exit( -1 );
		}
		#endif
		renderConfig.room_acoustics.override = TRUE;
		}

		@@ -2711,6 +2758,17 @@ static CmdlnArgs defaultArgs(
		#ifdef FIX_488_SYNC_DELAY
		args.syncMdDelay = 0;
		#endif

		#ifdef CONTROL_METADATA_DIRECTIVITY
		for ( int32_t i = 0; i < RENDERER_MAX_ISM_INPUTS; ++i )
		{
		args.directivityPatternId[i] = 0;
		}
		#endif

		#ifdef CONTROL_METADATA_REVERB
		args.acousticEnvironmentId = 0;
		#endif
		return args;
		}

		@@ -2854,6 +2912,20 @@ static void parseOption(
		case CmdLnOptionId_framing5ms:
		assert( numOptionValues == 0 );
		args->framing_5ms = true;
		#endif
		#ifdef CONTROL_METADATA_DIRECTIVITY
		case CmdLnOptionId_directivityPatternId:
		assert( numOptionValues <= RENDERER_MAX_ISM_INPUTS );
		for ( int16_t i = 0; i < numOptionValues; ++i )
		{
		args->directivityPatternId[i] = (int16_t) strtol( optionValues[i], NULL, 10 );
		}
		break;
		#endif
		#ifdef CONTROL_METADATA_REVERB
		case CmdLnOptionId_acousticEnvironmentId:
		assert( numOptionValues == 1 );
		args->acousticEnvironmentId = (int16_t) strtol( optionValues[0], NULL, 10 );
		break;
		#endif
		#ifdef FIX_488_SYNC_DELAY

lib_com/common_api_types.h

+14 −0

Original line number	Diff line number	Diff line
		@@ -54,6 +54,9 @@
		#define RENDERER_HEAD_POSITIONS_PER_FRAME 4
		#endif

		#ifdef EARLY_REFLECTIONS
		#define QC_ABS_COEFF 6
		#endif

		/----------------------------------------------------------------------------------
		* Common API structures
		@@ -189,6 +192,13 @@ typedef struct _IVAS_ROOM_ACOUSTICS_CONFIG
		float pAcoustic_dsr[IVAS_CLDFB_NO_CHANNELS_MAX]; /* - The room's Diffuse to Source Ratio per center frequency */
		float acousticPreDelay; /* Time elapsed between input signal and late reverberation start, float, range [0.001..10] */
		float inputPreDelay; /* Offset in seconds from where DSR is computed in the RIR (0 = at source), float, range [0.001..10] */
		#ifdef EARLY_REFLECTIONS
		int16_t use_er; /* ER activation flag */
		int32_t lowComplexity; /* Low complexity ER flag */
		IVAS_VECTOR3 dimensions; /* Room dimensions [m] */
		float AbsCoeff[QC_ABS_COEFF]; /* Absorption coeffs */
		IVAS_VECTOR3 ListenerOrigin; /* Listener origin */
		#endif
		} IVAS_ROOM_ACOUSTICS_CONFIG_DATA;

		#ifdef SPLIT_REND_WITH_HEAD_ROT
		@@ -223,7 +233,11 @@ typedef struct _IVAS_RENDER_CONFIG
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		IVAS_SPLIT_REND_CONFIG_DATA split_rend_config;
		#endif
		#ifdef CONTROL_METADATA_DIRECTIVITY
		float directivity[IVAS_MAX_NUM_OBJECTS * 3];
		#else
		float directivity[3];
		#endif
		} IVAS_RENDER_CONFIG_DATA, *IVAS_RENDER_CONFIG_HANDLE;

		typedef struct _IVAS_LS_CUSTOM_LAYOUT

lib_com/ivas_cnst.h

+42 −0

Original line number	Diff line number	Diff line
		@@ -1028,6 +1028,13 @@ typedef enum
		#define DIRAC_SLOT_NS 1250000L /* time duration of a time slot, 1.25ms (==DELAY_RENERER_NS/MAX_PARAM_SPATIAL_SUBFRAMES) */
		#define DIRAC_SLOT_ENC_NS 5000000L

		#define DIRAC_MONO_THRESH_SILENCE 3e4f
		#define DIRAC_MONO_NORM_FACTOR 1e13f
		#define DIRAC_MONO_ONE_ON_NORM_FACTOR (1.f / 1e13f)
		#define DIRAC_MONO_MAX_THRESH 1e6f
		#define DIRAC_MONO_MIN_THRESH 1e2f
		#define DIRAC_MONO_FRAME_THRESH 15 /* 30ms */

		typedef enum
		{
		DIRAC_OPEN, /* initialize to default value */
		@@ -1066,6 +1073,11 @@ typedef enum
		#define IVAS_ACTIVEW_DM_F_SCALE_VLBR 0.25f
		#define IVAS_SPAR_FOA_DFLT_FREQ_PER_CHAN 24000

		#ifdef FIX_SBA_VANISHING_RESIDUAL
		#define IVAS_SPAR_DYN_ACTIVEW_THRESH (0.0039f)
		#define IVAS_SPAR_SIDE_CH_DYN_ACTIVEW_THRESH (32.0f)
		#endif

		#define MAX_QUANT_STRATS 3
		#define MAX_CODING_STRATS 3
		#define IVAS_NUM_PROB_MODELS 4
		@@ -1901,6 +1913,36 @@ typedef enum
		#define IVAS_LIMITER_THRESHOLD 32729 /* -0.01 dBFS */
		#define IVAS_LIMITER_ATTACK_SECONDS 0.005f


		#ifdef EARLY_REFLECTIONS
		/----------------------------------------------------------------------------------
		* Early Reflection constants
		----------------------------------------------------------------------------------/
		#define ER_ABS_COEFF 6
		#define ER_MAX_SOURCES 25
		#define ER_REF_ORDER 1
		#define ER_NUM_REF 6

		#define ER_AIR_COEFF (0.00137f)
		#define ER_SOUND_SPEED (343.0f)
		#define ER_MIN_WALL_DIST (0.1f)
		#define ER_EUCLIDEAN_SCALE (1.29246971E-26f)

		#define ER_DEFAULT_ROOM_L (3.0f)
		#define ER_DEFAULT_ROOM_W (4.0f)
		#define ER_DEFAULT_ROOM_H (5.0f)
		#define ER_RADIUS (1.0f)
		#define ER_LIST_ORIGIN_X (0.0f)
		#define ER_LIST_ORIGIN_Y (0.0f)
		#define ER_LIST_HEIGHT (1.6f)

		#define ER_MIN_ROOM_DIMENSION (1.0f)
		#define ER_MAX_ROOM_DIMENSION (999.0f)
		#define ER_MIN_ABS_COEFF (0.0f)
		#define ER_MAX_ABS_COEFF (1.0f)

		#endif

		/----------------------------------------------------------------------------------
		* Stereo downmix EVS constants
		----------------------------------------------------------------------------------/