- Merge remote-tracking branch 'remotes/origin/main' into revision_of_constants (69b61a39) · Commits · SA4 / Audio / IVAS BASOP

apps/decoder.c

+20 −54

Original line number	Diff line number	Diff line
		@@ -38,6 +38,7 @@
		#include "bitstream_reader.h"
		#include "evs_rtp_payload.h"
		#include "ism_file_writer.h"
		#include "ivas_rtp_file.h"
		#include "jbm_file_writer.h"
		#include "hrtf_file_reader.h"
		#include "ls_custom_file_reader.h"
		@@ -50,7 +51,6 @@
		#include "vector3_pair_file_reader.h"
		#include "wmc_auto.h"
		#include "stl.h"
		#include "ivas_rtp_file.h"


		#define WMC_TOOL_SKIP
		@@ -133,7 +133,6 @@ typedef struct
		bool applyPiData;
		char *piOutputFilename;
		bool rtpOutSR;
		bool evsMode;
		IVAS_ROOM_SIZE_T roomSize;

		} DecArguments;
		@@ -164,19 +163,7 @@ typedef struct
		static bool parseCmdlIVAS_dec( int16_t argc, char *argv, DecArguments arg );
		static void usage_dec( void );
		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 *hHrtf,
		RotFileReader *headRotReader,
		RotFileReader *externalOrientationFileReader,
		RotFileReader *refRotReader,
		Vector3PairFileReader *referenceVectorReader,
		ObjectEditFileReader *objectEditFileReader,
		ISAR_SPLIT_REND_BITS_DATA *splitRendBits,
		IVAS_RENDER_CONFIG_DATA *renderConfig,
		IVAS_DEC_HANDLE *phIvasDec,
		int16_t *pcmBuf );
		static ivas_error decodeVoIP( DecArguments arg, BS_READER_HANDLE hBsReader, IVAS_DEC_HRTF_BINARY_WRAPPER hHrtf, RotFileReader headRotReader, RotFileReader externalOrientationFileReader, RotFileReader refRotReader, Vector3PairFileReader referenceVectorReader, ObjectEditFileReader objectEditFileReader, ISAR_SPLIT_REND_BITS_DATA splitRendBits, IVAS_RENDER_CONFIG_DATA renderConfig, IVAS_DEC_HANDLE phIvasDec, int16_t pcmBuf );
		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 );
		static void do_object_editing_fx( IVAS_EDITABLE_PARAMETERS editableParameters, ObjectEditFileReader objectEditFileReader );
		static ivas_error restartDecoder( IVAS_DEC_HANDLE phIvasDec, const IVAS_DEC_MODE decMode, DecArguments arg, IVAS_RENDER_CONFIG_DATA renderConfig, IVAS_CUSTOM_LS_DATA hLsCustomData );
		@@ -426,7 +413,6 @@ int main(

		asked_frame_size = arg.renderFramesize;
		uint16_t aeID = arg.aeSequence.count > 0 ? arg.aeSequence.pID[0] : IVAS_DEFAULT_AEID;

		arg.enableHeadRotation = arg.enableHeadRotation \|\| arg.outputConfig == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED \|\| arg.outputConfig == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM;

		if ( ( error = IVAS_DEC_Configure( hIvasDec, arg.output_Fs, arg.outputConfig, arg.renderFramesize, arg.customLsOutputEnabled, arg.hrtfReaderEnabled,
		@@ -497,7 +483,6 @@ int main(

		if ( arg.renderConfigEnabled )
		{

		/* sanity check */
		if ( arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL && arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_ROOM_IR && arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_ROOM_REVERB &&
		arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED && arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM &&
		@@ -564,6 +549,7 @@ int main(
		}
		}
		}

		if ( ( error = RenderConfigReader_getDirectivity( renderConfigReader, arg.directivityPatternId, renderConfig.directivity_fx ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "Failed to get directivity patterns for one or more of IDs: %d %d %d %d\n\n", arg.directivityPatternId[0], arg.directivityPatternId[1], arg.directivityPatternId[2], arg.directivityPatternId[3] );
		@@ -691,11 +677,7 @@ int main(

		if ( arg.voipMode )
		{
		error = decodeVoIP( arg, hBsReader, &hHrtfBinary, headRotReader, externalOrientationFileReader, refRotReader, referenceVectorReader, objectEditFileReader,
		&splitRendBits,
		&renderConfig,
		&hIvasDec,
		pcmBuf );
		error = decodeVoIP( arg, hBsReader, &hHrtfBinary, headRotReader, externalOrientationFileReader, refRotReader, referenceVectorReader, objectEditFileReader, &splitRendBits, &renderConfig, &hIvasDec, pcmBuf );
		}
		else
		{
		@@ -901,7 +883,6 @@ static bool parseCmdlIVAS_dec(
		arg->output_Fs = IVAS_MAX_SAMPLING_RATE;
		arg->outputConfig = IVAS_AUDIO_CONFIG_MONO;
		arg->decMode = IVAS_DEC_MODE_IVAS;
		arg->evsMode = false;
		arg->quietModeEnabled = false;
		arg->delayCompensationEnabled = true;
		arg->voipMode = false;
		@@ -1395,7 +1376,6 @@ static bool parseCmdlIVAS_dec(
		}
		else if ( strcmp( argv_to_upper, "-EVS" ) == 0 )
		{
		arg->evsMode = true;
		arg->decMode = IVAS_DEC_MODE_EVS;
		i++;
		}
		@@ -1486,7 +1466,7 @@ static bool parseCmdlIVAS_dec(
		usage_dec();
		return false;
		}
		else if ( arg->non_diegetic_pan_enabled && arg->outputConfig == IVAS_AUDIO_CONFIG_STEREO && arg->evsMode )
		else if ( arg->non_diegetic_pan_enabled && arg->outputConfig == IVAS_AUDIO_CONFIG_STEREO && arg->decMode == IVAS_DEC_MODE_EVS )
		{
		fprintf( stderr, "Error: Both non-diegetic panning and stereo output specified!\n\n" );
		usage_dec();
		@@ -2796,6 +2776,7 @@ static ivas_error decodeVoIP(
		int32_t delayTimeScale = -1;
		int16_t i;
		IVAS_DEC_HANDLE hIvasDec = *phIvasDec;

		bool restartNeeded;
		IVAS_RTP ivasRtp = { 0 };
		IVAS_RTP srRtp = { 0 };
		@@ -2876,7 +2857,6 @@ static ivas_error decodeVoIP(
		goto cleanup;
		}


		#ifdef SUPPORT_JBM_TRACEFILE
		if ( arg.jbmTraceFilename != NULL )
		{
		@@ -2908,7 +2888,9 @@ static ivas_error decodeVoIP(
		else
		{
		auPtr = au; /* might have been set to RTP packet in prev call */

		error = IVAS_RTP_ReadNextFrame( &ivasRtp, auPtr, &auSize, &rtpTimeStamp, &rtpSequenceNumber, &nextPacketRcvTime_ms, NULL, &qBit );

		initialTsOffsetSystemAndRTP = rtpTimeStamp - systemTime_ms * 16; /* For time mapping */

		/* EVS RTP payload format has timescale 16000, JBM uses 1000 internally */
		@@ -3105,6 +3087,7 @@ static ivas_error decodeVoIP(
		else
		{
		auPtr = au; /* might have been set to RTP packet in prev call */

		error = IVAS_RTP_ReadNextFrame( &ivasRtp, au, &auSize, &rtpTimeStamp, &rtpSequenceNumber, &nextPacketRcvTime_ms, NULL, &qBit );

		/* IVAS RTP payload format has timescale 16000, JBM uses 1000 internally */
		@@ -3294,22 +3277,8 @@ static ivas_error decodeVoIP(
		/* Once good frame decoded, catch up */
		if ( decodedGoodFrame )
		{
		if ( ( error = initOnFirstGoodFrame( hIvasDec,
		arg,
		numInitialBadFrames,
		&nOutSamples,
		&vec_pos_len,
		delayNumSamples_orig,
		&delayNumSamples,
		&delayTimeScale,
		&bsFormat,
		&afWriter,
		&masaWriter,
		ismWriters,
		&nOutChannels,
		&numObj,
		&srRtp,
		&splitRendWriter ) ) != IVAS_ERR_OK )
		if ( ( error = initOnFirstGoodFrame( hIvasDec, arg, numInitialBadFrames, &nOutSamples, &vec_pos_len, delayNumSamples_orig, &delayNumSamples, &delayTimeScale,
		&bsFormat, &afWriter, &masaWriter, ismWriters, &nOutChannels, &numObj, &srRtp, &splitRendWriter ) ) != IVAS_ERR_OK )
		{
		goto cleanup;
		}
		@@ -3411,6 +3380,7 @@ static ivas_error decodeVoIP(
		srInfo.bitrateKbps = splitRendBits->bits_written * 1000 / splitRendBits->codec_frame_size_ms;
		srInfo.codec = ( splitRendBits->codec == ISAR_SPLIT_REND_CODEC_LC3PLUS ) ? IVAS_SR_TRANSPORT_LC3PLUS : IVAS_SR_TRANSPORT_LCLD;
		srInfo.codecFrameSizeMs = (uint32_t) splitRendBits->codec_frame_size_ms;

		if ( ( error = IVAS_RTP_WriteNextFrame( &srRtp, splitRendBits->bits_buf, &srInfo, (int16_t) splitRendBits->bits_written, false, false ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nError %s while pushing SR audio bitstream to RTP pack\n", ivas_error_to_string( error ) );
		@@ -3661,7 +3631,8 @@ static void do_object_editing_fx(
		editableParameters->ism_metadata[obj_idx].elevation_fx = (Word32) ( readInfo->obj_ele[obj_idx] * 4194304 ); /* Q22 */
		}
		}
		/* object direction editing only for diegetic objects */

		/* object orientation editing only for diegetic objects */
		if ( readInfo->obj_yaw_edited[obj_idx] )
		{
		if ( readInfo->obj_yaw_relative[obj_idx] )
		@@ -3692,12 +3663,8 @@ static void do_object_editing_fx(
		if ( readInfo->obj_radius_relative[obj_idx] )
		{
		/* radius: apply relative edit + saturation */
		#define SQ 9

		#define SHIFT_DUE_TO_SQxSQ ( SQ - ( SQ + SQ + 1 - 16 ) ) /* ^shift by 6 == mult by 2^6 = 64.0 */

		Word32 L_tmp1 = L_mult( editableParameters->ism_metadata[obj_idx].radius_fx, (Word16) ( readInfo->obj_radius[obj_idx] * ( 1L << ( SQ ) ) ) ); /SQSQ*/
		Word32 L_tmp2 = L_shl_sat( L_tmp1, SHIFT_DUE_TO_SQxSQ ); /shift back to Q(9+16)/
		Word32 L_tmp1 = L_mult( editableParameters->ism_metadata[obj_idx].radius_fx, (Word16) ( readInfo->obj_radius[obj_idx] * ( 1L << ( 9 ) ) ) ); /SQSQ where SQ=9*/
		Word32 L_tmp2 = L_shl_sat( L_tmp1, ( 9 - ( 9 + 9 + 1 - 16 ) ) /* ^shift by 6 == mult by 2^6 = 64.0 / ); /shift back to Q(9+16)*/
		Word16 temp_radius = round_fx( L_tmp2 ); /* Q25 -> Q9 */
		temp_radius = s_max( s_min( temp_radius, OBJ_EDIT_RADIUS_MAX_FX ), 0 );

		@@ -3968,9 +3935,10 @@ static ivas_error restartDecoder(

		hIvasDec = *phIvasDec;

		uint16_t aeID = arg->aeSequence.count > 0 ? arg->aeSequence.pID[0] : 65535;
		uint16_t aeID = arg->aeSequence.count > 0 ? arg->aeSequence.pID[0] : IVAS_DEFAULT_AEID;

		IVAS_AUDIO_CONFIG outputConfig = ( decMode == IVAS_DEC_MODE_IVAS ) ? arg->outputConfig : IVAS_AUDIO_CONFIG_MONO;

		if ( ( error = IVAS_DEC_Configure( hIvasDec, arg->output_Fs, outputConfig, arg->renderFramesize, arg->customLsOutputEnabled, arg->hrtfReaderEnabled,
		arg->enableHeadRotation, arg->enableExternalOrientation, arg->orientation_tracking, arg->renderConfigEnabled, arg->roomSize, arg->non_diegetic_pan_enabled,
		arg->non_diegetic_pan_gain_fx, arg->dpidEnabled, aeID, arg->objEditEnabled, arg->delayCompensationEnabled ) ) != IVAS_ERR_OK )
		@@ -3985,7 +3953,6 @@ static ivas_error restartDecoder(
		goto cleanup;
		}


		if ( arg->voipMode )
		{
		if ( ( error = IVAS_DEC_EnableVoIP( hIvasDec, 60, arg->inputFormat ) ) != IVAS_ERR_OK )
		@@ -4034,5 +4001,4 @@ cleanup:
		return error;
		}


		#undef WMC_TOOL_SKIP

apps/encoder.c

+2 −2

Original line number	Diff line number	Diff line
		@@ -43,9 +43,10 @@
		#include "ism_file_reader.h"
		#include "jbm_file_reader.h"
		#include "masa_file_reader.h"
		#include "wmc_auto.h"
		#include "rotation_file_reader.h"
		#include "ivas_rtp_file.h"
		#include "wmc_auto.h"


		#ifdef DEBUG_FORCE_DIR
		/* Windows does not define the S_ISREG and S_ISDIR macros in stat.h, so we do.
		@@ -1867,7 +1868,6 @@ static bool parseCmdlIVAS_enc(
		i++;
		}


		/-----------------------------------------------------------------
		* Option not recognized
		-----------------------------------------------------------------/

apps/encoder_fmtsw.c

+4 −2

Original line number	Diff line number	Diff line
		@@ -43,9 +43,10 @@
		#include "ism_file_reader.h"
		#include "jbm_file_reader.h"
		#include "masa_file_reader.h"
		#include "wmc_auto.h"
		#include "rotation_file_reader.h"
		#include "ivas_rtp_file.h"
		#include "wmc_auto.h"


		#ifdef DEBUG_FORCE_DIR
		/* Windows does not define the S_ISREG and S_ISDIR macros in stat.h, so we do.
		@@ -185,6 +186,7 @@ static IVAS_ENC_FORCED_MODE parseForcedMode( char *forcedModeChar );
		static void str2arg( char str, int argc_local, char *argv_local[] );
		static int encoder_main( int argc, char argv[], IVAS_RTP ivasRtp, int init_RtpWriter );


		/------------------------------------------------------------------------------------------
		* main()
		*
		@@ -263,6 +265,7 @@ int main(
		cleanup:

		IVAS_RTP_Term( &ivasRtp );

		if ( FmtSWFile )
		{
		fclose( FmtSWFile );
		@@ -1968,7 +1971,6 @@ static bool parseCmdlIVAS_enc(
		i++;
		}


		/-----------------------------------------------------------------
		* Option not recognized
		-----------------------------------------------------------------/

apps/isar_post_rend.c

+5 −1

Original line number	Diff line number	Diff line
		@@ -229,6 +229,7 @@ static IVAS_AUDIO_CONFIG parseAudioConfig( const char configString, bool srRtp

		static void convertOutputBuffer( const Word32 fixedBuffer, Word16 q, const int16_t numSamplesPerChannel, const int16_t numChannels, int16_t intBuffer );


		/------------------------------------------------------------------------------------------
		* Local functions
		------------------------------------------------------------------------------------------/
		@@ -399,6 +400,7 @@ static IVAS_AUDIO_CONFIG parseAudioConfig(
		*srRtp = true;
		return IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED;
		}

		return IVAS_AUDIO_CONFIG_INVALID;
		}

		@@ -456,6 +458,7 @@ static CmdlnArgs defaultArgs(
		clearString( args.inputFilePath );
		clearString( args.outputFilePath );
		clearString( args.srParamsFilePath );

		args.sampleRate = 0;

		args.outConfig.audioConfig = IVAS_AUDIO_CONFIG_INVALID;
		@@ -561,7 +564,6 @@ static void parseOption(
		fprintf( stderr, "Unknown or invalid option for frame size: %s\n", optionValues[0] );
		exit( -1 );
		}

		break;
		case CmdLnOptionId_srParamsFile:
		assert( numOptionValues == 1 );
		@@ -695,6 +697,7 @@ static void convertOutputBuffer(
		return;
		}


		static void trim( char *str )
		{
		char c;
		@@ -712,6 +715,7 @@ static void trim( char *str )
		str[w] = 0;
		}


		static ivas_error parseSRParamsFile(
		const char *srParamsFilePath,
		const char *rtpFilePath,

apps/renderer.c

+7 −12

Original line number	Diff line number	Diff line
		@@ -230,7 +230,7 @@ typedef enum
		CmdLnOptionId_syncMdDelay,
		CmdLnOptionId_directivityPatternId,
		CmdLnOptionId_acousticEnvironmentId,
		CmdLnOptionId_roomSize,
		CmdLnOptionId_roomSize
		} CmdLnOptionId;

		static const CmdLnParser_Option cliOptions[] = {
		@@ -1169,8 +1169,6 @@ int main(

		if ( args.renderConfigFilePath[0] != '\0' )
		{


		/* sanity check */
		if ( ( args.outConfig.audioConfig != IVAS_AUDIO_CONFIG_BINAURAL ) &&
		( args.outConfig.audioConfig != IVAS_AUDIO_CONFIG_BINAURAL_ROOM_IR ) &&
		@@ -1205,7 +1203,7 @@ int main(
		}
		if ( args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_BINAURAL_ROOM_REVERB )
		{
		aeID = args.aeSequence.count > 0 ? args.aeSequence.pID[0] : 65535;
		aeID = args.aeSequence.count > 0 ? args.aeSequence.pID[0] : IVAS_DEFAULT_AEID;
		if ( ( error = RenderConfigReader_getAcousticEnvironment( renderConfigReader, aeID, &renderConfig.roomAcoustics ) ) == IVAS_ERR_OK )
		{
		if ( RenderConfigReader_checkValues( &renderConfig ) != IVAS_ERR_OK )
		@@ -1241,7 +1239,6 @@ int main(
		goto cleanup;
		}


		/* Set reverb room size if specified */
		if ( args.reverbRoomSize != IVAS_ROOM_SIZE_AUTO )
		{
		@@ -1534,12 +1531,7 @@ int main(
		audioWriter = NULL;
		}

		if ( ( error = IVAS_REND_GetSplitRendBitstreamHeader( hIvasRend,
		&bitsBuffer.config.codec,
		&bitsBuffer.config.poseCorrection,
		&bitsBuffer.config.codec_frame_size_ms,
		&bitsBuffer.config.isar_frame_size_ms,
		&bitsBuffer.config.lc3plus_highres ) ) != IVAS_ERR_OK )
		if ( ( error = IVAS_REND_GetSplitRendBitstreamHeader( hIvasRend, &bitsBuffer.config.codec, &bitsBuffer.config.poseCorrection, &bitsBuffer.config.codec_frame_size_ms, &bitsBuffer.config.isar_frame_size_ms, &bitsBuffer.config.lc3plus_highres ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nError in IVAS_REND_GetSplitRendBitstreamHeader(): %s!\n", ivas_error_to_string( error ) );
		goto cleanup;
		@@ -1650,7 +1642,7 @@ int main(

		ObjectPositionBuffer mtdBuffer;
		outBuffer.pq_fact = &outBuffer.q_factor;
		Word16 subframe_len = (Word16) ( args.sampleRate / ( 200 ) ); // sample rate /FRAMES_PER_SEC * MAX_PARAM_SPATIAL_SUBFRAMES
		Word16 subframe_len = (Word16) ( args.sampleRate / ( FRAMES_PER_SEC * MAX_PARAM_SPATIAL_SUBFRAMES ) );
		Word16 gd_bits = find_guard_bits( subframe_len );
		Word16 prev_q_fact = Q11;
		while ( 1 )
		@@ -2640,7 +2632,9 @@ static bool parseAcousticEnvironmentIds(
		AcousticEnvironmentSequence *aeSequence )
		{
		char config_string[RENDERER_MAX_METADATA_LINE_LENGTH];

		strncpy( config_string, value, RENDERER_MAX_METADATA_LINE_LENGTH );

		if ( !is_digits_only( config_string ) )
		{
		aeidFileReader *aeidReader = NULL;
		@@ -3109,6 +3103,7 @@ void getMetadataFromFileReader(
		fprintf( stderr, "\nError (%s) while reading ISM metadata from: %s\n\n", ivas_error_to_string( error ), IsmFileReader_getFilePath( ismReader ) );
		exit( -1 );
		}

		objectMetadataBuffer->positions[objIdx].azimuth_fx = ismMetadata.azimuth_fx;
		objectMetadataBuffer->positions[objIdx].elevation_fx = ismMetadata.elevation_fx;
		objectMetadataBuffer->positions[objIdx].radius_fx = ismMetadata.radius_fx;