Merge branch '1955-basop-PortMr2077-from-float-revision-of-common-api-types' into 'main' (e51b4fce) · Commits · SA4 / Audio / IVAS BASOP

apps/decoder.c

+25 −30

Original line number	Diff line number	Diff line
		@@ -67,19 +67,10 @@ static
		#endif
		int32_t frame = 0; /* Counter of frames */

		#ifdef DEBUGGING
		#define MIN_NUM_BITS_ACTIVE_FRAME 56
		#define NUM_BITS_SID_IVAS_5K2 104
		#define MAX_OUTPUT_PCM_BUFFER_SIZE ( ( IVAS_MAX_OUTPUT_CHANNELS + IVAS_MAX_NUM_OBJECTS ) * IVAS_MAX_FRAME_SIZE )

		#define ORIENT_TRK_NONE ( 0 )
		#define ORIENT_TRK_REF ( 1 )
		#define ORIENT_TRK_AVG ( 2 )
		#define ORIENT_TRK_REF_VEC ( 3 )
		#define ORIENT_TRK_REF_VEC_LEV ( 4 )

		#define JBM_FRONTEND_FETCH_FRAMESIZE_MS 20
		#define HEADROTATION_FETCH_FRAMESIZE_MS 5
		#define DEFAULT_FETCH_FRAMESIZE_MS 20
		#endif


		/------------------------------------------------------------------------------------------
		@@ -125,7 +116,7 @@ typedef struct
		IVAS_DEC_INPUT_FORMAT inputFormat;
		bool customLsOutputEnabled;
		char *customLsSetupFilename;
		int16_t orientation_tracking;
		IVAS_HEAD_ORIENT_TRK_T orientation_tracking;
		bool non_diegetic_pan_enabled;
		float non_diegetic_pan_gain;
		Word16 non_diegetic_pan_gain_fx; /* Q15 */
		@@ -174,10 +165,10 @@ static bool parseCmdlIVAS_dec( int16_t argc, char *argv, DecArguments arg );
		static void usage_dec( void );
		#ifdef FIX_1217_OBJECT_EDIT_FILE_INTERFACE
		static ivas_error decodeG192( DecArguments arg, BS_READER_HANDLE hBsReader, IVAS_DEC_HRTF_BINARY_WRAPPER hHrtfBinary, RotFileReader headRotReader, RotFileReader externalOrientationFileReader, RotFileReader refRotReader, Vector3PairFileReader referenceVectorReader, ObjectEditFileReader objectEditFileReader, ISAR_SPLIT_REND_BITS_DATA splitRendBits, IVAS_DEC_HANDLE hIvasDec, int16_t pcmBuf );
		static ivas_error decodeVoIP( DecArguments arg, BS_READER_HANDLE hBsReader, IVAS_DEC_HRTF_BINARY_WRAPPER hHrtfBinary, RotFileReader headRotReader, RotFileReader externalOrientationFileReader, RotFileReader refRotReader, Vector3PairFileReader referenceVectorReader, ObjectEditFileReader objectEditFileReader, IVAS_DEC_HANDLE hIvasDec );
		static ivas_error decodeVoIP( DecArguments arg, BS_READER_HANDLE hBsReader, IVAS_DEC_HRTF_BINARY_WRAPPER hHrtfBinary, RotFileReader headRotReader, RotFileReader externalOrientationFileReader, RotFileReader refRotReader, Vector3PairFileReader referenceVectorReader, ObjectEditFileReader objectEditFileReader, IVAS_DEC_HANDLE hIvasDec, int16_t *pcmBuf );
		#else
		static ivas_error decodeG192( DecArguments arg, BS_READER_HANDLE hBsReader, IVAS_DEC_HRTF_BINARY_WRAPPER hHrtfBinary, RotFileReader headRotReader, RotFileReader externalOrientationFileReader, RotFileReader refRotReader, Vector3PairFileReader referenceVectorReader, ISAR_SPLIT_REND_BITS_DATA splitRendBits, IVAS_DEC_HANDLE hIvasDec, int16_t *pcmBuf );
		static ivas_error decodeVoIP( DecArguments arg, BS_READER_HANDLE hBsReader, IVAS_DEC_HRTF_BINARY_WRAPPER hHrtfBinary, RotFileReader headRotReader, RotFileReader externalOrientationFileReader, RotFileReader refRotReader, Vector3PairFileReader *referenceVectorReader, IVAS_DEC_HANDLE hIvasDec );
		static ivas_error decodeVoIP( DecArguments arg, BS_READER_HANDLE hBsReader, IVAS_DEC_HRTF_BINARY_WRAPPER hHrtfBinary, RotFileReader headRotReader, RotFileReader externalOrientationFileReader, RotFileReader refRotReader, Vector3PairFileReader referenceVectorReader, IVAS_DEC_HANDLE hIvasDec, int16_t pcmBuf );
		#endif
		static ivas_error load_hrtf_from_file( IVAS_DEC_HRTF_BINARY_WRAPPER *hHrtfBinary, IVAS_DEC_HANDLE hIvasDec, const IVAS_AUDIO_CONFIG OutputConfig, const int32_t output_Fs );
		#ifdef OBJ_EDITING_EXAMPLE
		@@ -348,7 +339,7 @@ int main(
		}

		/* sanity check */
		if ( arg.orientation_tracking != ORIENT_TRK_REF )
		if ( arg.orientation_tracking != IVAS_HEAD_ORIENT_TRK_REF )
		{
		fprintf( stderr, "\nError: Reference rotation file can be used in '-otr ref' mode only.\n\n" );
		goto cleanup;
		@@ -375,7 +366,7 @@ int main(
		}

		/* sanity check */
		if ( arg.orientation_tracking != ORIENT_TRK_REF_VEC && arg.orientation_tracking != ORIENT_TRK_REF_VEC_LEV )
		if ( arg.orientation_tracking != IVAS_HEAD_ORIENT_TRK_REF_VEC && arg.orientation_tracking != IVAS_HEAD_ORIENT_TRK_REF_VEC_LEV )
		{
		fprintf( stderr, "\nError: Reference trajectory file can be used in '-otr ref_vec' or '-otr ref_vec_lev' mode only.\n\n" );
		goto cleanup;
		@@ -674,9 +665,9 @@ int main(
		if ( arg.voipMode )
		{
		#ifdef FIX_1217_OBJECT_EDIT_FILE_INTERFACE
		error = decodeVoIP( arg, hBsReader, &hHrtfBinary, headRotReader, externalOrientationFileReader, refRotReader, referenceVectorReader, objectEditFileReader, hIvasDec );
		error = decodeVoIP( arg, hBsReader, &hHrtfBinary, headRotReader, externalOrientationFileReader, refRotReader, referenceVectorReader, objectEditFileReader, hIvasDec, pcmBuf );
		#else
		error = decodeVoIP( arg, hBsReader, &hHrtfBinary, headRotReader, externalOrientationFileReader, refRotReader, referenceVectorReader, hIvasDec );
		error = decodeVoIP( arg, hBsReader, &hHrtfBinary, headRotReader, externalOrientationFileReader, refRotReader, referenceVectorReader, hIvasDec, pcmBuf );
		#endif
		}
		else
		@@ -870,7 +861,11 @@ static bool parseCmdlIVAS_dec(
		int16_t i;
		char argv_to_upper[FILENAME_MAX];

		arg->output_Fs = 48000;
		/-----------------------------------------------------------------
		* Set default values
		-----------------------------------------------------------------/

		arg->output_Fs = IVAS_MAX_SAMPLING_RATE;
		arg->outputConfig = IVAS_AUDIO_CONFIG_MONO;
		arg->decMode = IVAS_DEC_MODE_IVAS;
		arg->quietModeEnabled = false;
		@@ -880,7 +875,7 @@ static bool parseCmdlIVAS_dec(

		arg->enableHeadRotation = false;
		arg->headrotTrajFileName = NULL;
		arg->orientation_tracking = ORIENT_TRK_NONE;
		arg->orientation_tracking = IVAS_HEAD_ORIENT_TRK_NONE;
		arg->enableReferenceRotation = false;
		arg->headrotTrajFileName = NULL;
		arg->enableReferenceVectorTracking = false;
		@@ -1097,23 +1092,23 @@ static bool parseCmdlIVAS_dec(

		if ( strcmp( argv_to_upper, "NONE" ) == 0 )
		{
		arg->orientation_tracking = ORIENT_TRK_NONE;
		arg->orientation_tracking = IVAS_HEAD_ORIENT_TRK_NONE;
		}
		else if ( strcmp( argv_to_upper, "REF" ) == 0 )
		{
		arg->orientation_tracking = ORIENT_TRK_REF;
		arg->orientation_tracking = IVAS_HEAD_ORIENT_TRK_REF;
		}
		else if ( strcmp( argv_to_upper, "AVG" ) == 0 )
		{
		arg->orientation_tracking = ORIENT_TRK_AVG;
		arg->orientation_tracking = IVAS_HEAD_ORIENT_TRK_AVG;
		}
		else if ( strcmp( argv_to_upper, "REF_VEC" ) == 0 )
		{
		arg->orientation_tracking = ORIENT_TRK_REF_VEC;
		arg->orientation_tracking = IVAS_HEAD_ORIENT_TRK_REF_VEC;
		}
		else if ( strcmp( argv_to_upper, "REF_VEC_LEV" ) == 0 )
		{
		arg->orientation_tracking = ORIENT_TRK_REF_VEC_LEV;
		arg->orientation_tracking = IVAS_HEAD_ORIENT_TRK_REF_VEC_LEV;
		}
		else
		{
		@@ -2765,14 +2760,15 @@ static ivas_error decodeVoIP(
		#ifdef FIX_1217_OBJECT_EDIT_FILE_INTERFACE
		ObjectEditFileReader *objectEditFileReader,
		#endif
		IVAS_DEC_HANDLE hIvasDec )
		IVAS_DEC_HANDLE hIvasDec,
		int16_t *pcmBuf )
		{
		bool decodingFailed = true; /* Assume failure until cleanup is reached without errors */
		ivas_error error = IVAS_ERR_OK;

		uint32_t nextPacketRcvTime_ms = 0;
		uint32_t systemTime_ms = 0;
		uint32_t systemTimeInc_ms = (uint32_t) JBM_FRONTEND_FETCH_FRAMESIZE_MS;
		uint32_t systemTimeInc_ms = (uint32_t) ( 1000 / IVAS_NUM_FRAMES_PER_SEC );
		int32_t nFramesFed = 0;

		uint8_t au[( IVAS_MAX_BITS_PER_FRAME + 7 ) >> 3];
		@@ -2786,7 +2782,6 @@ static ivas_error decodeVoIP(
		MasaFileWriter *masaWriter = NULL;
		uint16_t numObj = 0;

		int16_t pcmBuf[MAX_OUTPUT_PCM_BUFFER_SIZE];
		AudioFileWriter *afWriter = NULL;
		#ifdef SUPPORT_JBM_TRACEFILE
		JbmTraceFileWriter *jbmTraceWriter = NULL;

apps/encoder.c

+21 −25

Original line number	Diff line number	Diff line
		@@ -163,7 +163,6 @@ typedef struct
		* Local functions prototypes
		------------------------------------------------------------------------------------------/

		static void initArgStruct( EncArguments *arg );
		static bool parseCmdlIVAS_enc( int16_t argc, char argv[], EncArguments arg );
		static void usage_enc( void );
		static bool readBandwidth( FILE file, IVAS_ENC_BANDWIDTH bandwidth, int32_t *bandwidthFrameCounter );
		@@ -208,8 +207,6 @@ int main(
		reset_mem( USE_BYTES );
		#endif

		initArgStruct( &arg );

		/------------------------------------------------------------------------------------------
		* Parse command-line arguments
		------------------------------------------------------------------------------------------/
		@@ -853,11 +850,26 @@ cleanup:
		* Local functions
		-------------------------------------------------------------------/

		#define IVAS_DEFAULT_AGC ( 0 )
		/---------------------------------------------------------------------
		* parseCmdlIVAS_enc()
		*
		* Encoder command-line parsing
		---------------------------------------------------------------------/

		static void initArgStruct( EncArguments *arg )
		static bool parseCmdlIVAS_enc(
		int16_t argc,
		char *argv[],
		EncArguments *arg )
		{
		/* Set default values here */
		int16_t i, j;
		char argv_to_upper[FILENAME_MAX];
		char stmp[FILENAME_MAX];
		int32_t tmp;

		/-----------------------------------------------------------------
		* Set default values
		-----------------------------------------------------------------/

		arg->inputWavFilename = NULL;
		arg->outputBitstreamFilename = NULL;
		arg->inputFs = 0;
		@@ -887,25 +899,9 @@ static void initArgStruct( EncArguments *arg )
		#endif
		arg->pca = false;

		return;
		}


		/---------------------------------------------------------------------
		* parseCmdlIVAS_enc()
		*
		* Encoder command-line parsing
		---------------------------------------------------------------------/

		static bool parseCmdlIVAS_enc(
		int16_t argc,
		char *argv[],
		EncArguments *arg )
		{
		int16_t i, j;
		char argv_to_upper[FILENAME_MAX];
		char stmp[FILENAME_MAX];
		int32_t tmp;
		/-----------------------------------------------------------------
		* Initialization
		-----------------------------------------------------------------/

		if ( argc < 5 )
		{

apps/isar_post_rend.c

+1 −1

Original line number	Diff line number	Diff line
		@@ -952,7 +952,7 @@ int main(

		memset( outBuffer.data_fx, 0, outBuffer.config.numSamplesPerChannel * outBuffer.config.numChannels * sizeof( Word32 ) );

		bitsBufferSize = SPLIT_REND_BITS_BUFF_SIZE;
		bitsBufferSize = ISAR_SPLIT_REND_BITS_BUFF_SIZE;

		if ( bitsBufferSize > 0 )
		{

apps/renderer.c

+3 −3

Original line number	Diff line number	Diff line
		@@ -174,7 +174,7 @@ typedef struct
		char externalOrientationFilePath[RENDERER_MAX_CLI_ARG_LENGTH];
		char customHrtfFilePath[RENDERER_MAX_CLI_ARG_LENGTH];
		char renderConfigFilePath[RENDERER_MAX_CLI_ARG_LENGTH];
		int8_t orientation_tracking;
		IVAS_HEAD_ORIENT_TRK_T orientation_tracking;
		int16_t Opt_Headrotation;
		int16_t Opt_ExternalOrientation;
		int16_t nonDiegeticPan;
		@@ -1601,7 +1601,7 @@ int main(

		if ( is_split_pre_rend_mode( &args ) )
		{
		bitsBufferSize = SPLIT_REND_BITS_BUFF_SIZE;
		bitsBufferSize = ISAR_SPLIT_REND_BITS_BUFF_SIZE;
		}
		else
		{
		@@ -2431,7 +2431,7 @@ static bool parseRenderFramesize(

		static bool parseOrientationTracking(
		char *value,
		int8_t *orientation_tracking )
		IVAS_HEAD_ORIENT_TRK_T *orientation_tracking )
		{
		to_upper( value );

lib_com/common_api_types.h

+16 −11

Original line number	Diff line number	Diff line
		@@ -46,24 +46,28 @@
		* Common API constants
		----------------------------------------------------------------------------------/

		#define IVAS_NUM_FRAMES_PER_SEC 50
		#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_NUM_FRAMES_PER_SEC 50 /* number of frames per second */
		#define IVAS_MAX_SAMPLING_RATE 48000 /* maximum supported sampling rate in Hz */
		#define IVAS_MAX_FRAME_SIZE ( IVAS_MAX_SAMPLING_RATE / IVAS_NUM_FRAMES_PER_SEC ) /* maximum frame buffer length (per channel) in samples */
		#define IVAS_MAX_BITS_PER_FRAME ( 512000 / IVAS_NUM_FRAMES_PER_SEC ) /* maximum bits per frame; corresponds to maximum bitrate of 512 kbps */

		#define IVAS_MAX_NUM_OBJECTS 4
		#define IVAS_MAX_INPUT_CHANNELS 16
		#define IVAS_MAX_OUTPUT_CHANNELS 16
		#define IVAS_MAX_OUTPUT_CHANNELS 16 /* Note: there is an exception for OSBA and EXT otuput where it can be 20 (HOA3 + 4 ISM channels) */
		#define IVAS_CLDFB_NO_COL_MAX 16
		#define IVAS_CLDFB_NO_CHANNELS_MAX 60

		/* Renderer constants */
		#define IVAS_MAX_PARAM_SPATIAL_SUBFRAMES 4
		#define IVAS_ROOM_ABS_COEFF 6
		#define IVAS_REVERB_PREDELAY_MAX 20 /* Max input delay for reverb module */
		#define IVAS_ER_LIST_HEIGHT 1.6f
		#define IVAS_ER_LIST_HEIGHT_FX 6710886 /* 1.6f in Q.22 */

		/* Maximum buffer length (per channel) in samples */
		#define MAX_BUFFER_LENGTH_PER_CHANNEL ( L_FRAME48k )
		/* JBM constants for adaptive-playout */
		#define IVAS_TIME_SCALE_MIN 50 /* min. time-scaling [%] */
		#define IVAS_TIME_SCALE_MAX 150 /* max. time-scaling [%] */

		/* Frame size required when rendering to binaural */
		#define BINAURAL_RENDERING_FRAME_SIZE_MS 5

		/----------------------------------------------------------------------------------
		* Common API enum for output audio configurations
		@@ -196,7 +200,7 @@ typedef struct ivas_hrtf_crend_structure *IVAS_DEC_HRTF_CREND_HANDLE;
		typedef struct ivas_hrtf_fastconv_struct *IVAS_DEC_HRTF_FASTCONV_HANDLE;
		typedef struct ivas_hrtf_parambin_struct *IVAS_DEC_HRTF_PARAMBIN_HANDLE;
		typedef struct ivas_hrtf_statistics_struct *IVAS_DEC_HRTF_STATISTICS_HANDLE;
		typedef struct cldfb_filter_bank_struct *IVAS_CLDFB_FILTER_BANK_HANDLE;
		typedef struct ivas_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;
		@@ -247,7 +251,7 @@ typedef enum _ivas_binaural_renderer_type
		#define ISAR_MAX_SPLIT_REND_BITRATE 768000
		#define ISAR_MAX_SPLIT_REND_BITS_BUFFER_SIZE_IN_BYTES ( ( ( (Word32) ISAR_MAX_SPLIT_REND_BITRATE / IVAS_NUM_FRAMES_PER_SEC ) + 7 ) >> 3 )
		#define ISAR_SPLIT_REND_ADDITIONAL_BYTES_TO_READ 1
		#define SPLIT_REND_BITS_BUFF_SIZE ( ISAR_MAX_SPLIT_REND_BITS_BUFFER_SIZE_IN_BYTES + ISAR_SPLIT_REND_ADDITIONAL_BYTES_TO_READ )
		#define ISAR_SPLIT_REND_BITS_BUFF_SIZE ( ISAR_MAX_SPLIT_REND_BITS_BUFFER_SIZE_IN_BYTES + ISAR_SPLIT_REND_ADDITIONAL_BYTES_TO_READ )

		typedef enum
		{
		@@ -312,6 +316,7 @@ typedef struct _ISAR_SPLIT_REND_CONFIG

		} ISAR_SPLIT_REND_CONFIG_DATA, *ISAR_SPLIT_REND_CONFIG_HANDLE;


		/----------------------------------------------------------------------------------
		* Renderer API structures and enums
		----------------------------------------------------------------------------------/
		@@ -356,6 +361,7 @@ typedef struct _IVAS_RENDER_CONFIG
		Word16 directivity_fx[IVAS_MAX_NUM_OBJECTS * 3]; // has the following q-factor pattern: {6, 6, 15, 6, 6, 15, 6, 6, 15, 6, 6, 15}
		float distAtt[3];
		Word32 distAtt_fx[3]; /* {Q27, Q30, Q30} */

		} IVAS_RENDER_CONFIG_DATA, *IVAS_RENDER_CONFIG_HANDLE;

		typedef struct
		@@ -372,7 +378,6 @@ typedef struct
		Word16 q_factor;
		Word16 *pq_fact;
		Word32 *data_fx;
		// Word16 Q_data;

		} IVAS_REND_AudioBuffer;