acceptance of switches, step 2 (88f1eb3c) · Commits · SA4 / Audio / IVAS BASOP

apps/decoder.c

+0 −76

Original line number	Diff line number	Diff line
		@@ -45,9 +45,7 @@
		#include "masa_file_writer.h"
		#include "render_config_reader.h"
		#include "rotation_file_reader.h"
		#ifdef FIX_1053_REVERB_RECONFIGURATION
		#include "aeid_file_reader.h"
		#endif
		#include "split_render_file_read_write.h"
		#include "obj_edit_file_reader.h"
		#include "vector3_pair_file_reader.h"
		@@ -135,9 +133,7 @@ typedef struct
		bool applyPiData;
		char *piOutputFilename;
		bool rtpOutSR;
		#ifdef FIX_1419_MONO_STEREO_UMX
		bool evsMode;
		#endif
		IVAS_ROOM_SIZE_T roomSize;

		} DecArguments;
		@@ -888,9 +884,7 @@ static bool parseCmdlIVAS_dec(
		arg->output_Fs = IVAS_MAX_SAMPLING_RATE;
		arg->outputConfig = IVAS_AUDIO_CONFIG_MONO;
		arg->decMode = IVAS_DEC_MODE_IVAS;
		#ifdef FIX_1419_MONO_STEREO_UMX
		arg->evsMode = false;
		#endif
		arg->quietModeEnabled = false;
		arg->delayCompensationEnabled = true;
		arg->voipMode = false;
		@@ -1289,7 +1283,6 @@ static bool parseCmdlIVAS_dec(

		if ( !is_digits_only( argv[i] ) )
		{
		#ifdef FIX_1053_REVERB_RECONFIGURATION
		aeidFileReader *aeidReader = NULL;
		if ( aeidFileReader_open( argv[i], &aeidReader ) != IVAS_ERR_OK )
		{
		@@ -1305,67 +1298,6 @@ static bool parseCmdlIVAS_dec(
		}
		aeidFileReader_close( &aeidReader );
		i++;
		#else
		uint16_t k;
		char *s = argv[i];
		char *token = argv[i];

		for ( k = 0; s[k]; )
		{
		s[k] == ',' ? k++ : *s++;
		}
		k++;

		if ( k == 0 )
		{
		fprintf( stdout, "Error: Invalid acoustic environment sequence specified: %s\n\n", argv[i] );
		usage_dec();
		return false;
		}

		if ( NULL == ( arg->aeSequence.pID = malloc( sizeof( uint16_t ) * k ) ) \|\|
		NULL == ( arg->aeSequence.pValidity = malloc( sizeof( uint16_t ) * k ) ) )
		{
		fprintf( stdout, "Error: Unable to allocate memory for acoustic environment sequence: %s\n\n", argv[i] );
		usage_dec();
		return false;
		}

		arg->aeSequence.count = k;

		k = 0;
		token = strtok( argv[i++], ":" );

		while ( token != NULL )
		{
		if ( !is_number( token ) )
		{
		fprintf( stdout, "Error: Invalid token %s found in acoustic environment sequence: %s\n\n", token, argv[i] );
		usage_dec();
		return false;
		}
		arg->aeSequence.pID[k] = (uint16_t) atoi( token );

		token = strtok( NULL, "," );
		if ( !is_number( token ) )
		{
		fprintf( stdout, "Error: Invalid token %s found in acoustic environment sequence: %s\n\n", token, argv[i] );
		usage_dec();
		return false;
		}
		arg->aeSequence.pValidity[k] = (uint16_t) atoi( token );

		token = strtok( NULL, ":" );
		k++;
		}

		if ( k != arg->aeSequence.count )
		{
		fprintf( stdout, "Error while parsing acoustic environment sequence: %s\n\n", argv[i] );
		usage_dec();
		return false;
		}
		#endif
		}
		else
		{
		@@ -1444,14 +1376,12 @@ static bool parseCmdlIVAS_dec(
		}
		i++;
		}
		#ifdef FIX_1419_MONO_STEREO_UMX
		else if ( strcmp( argv_to_upper, "-EVS" ) == 0 )
		{
		arg->evsMode = true;
		arg->decMode = IVAS_DEC_MODE_EVS;
		i++;
		}
		#endif
		else if ( strcmp( argv_to_upper, "-ROOM_SIZE" ) == 0 )
		{
		i++;
		@@ -1539,7 +1469,6 @@ static bool parseCmdlIVAS_dec(
		usage_dec();
		return false;
		}
		#ifdef FIX_1419_MONO_STEREO_UMX
		else if ( arg->non_diegetic_pan_enabled && arg->outputConfig == IVAS_AUDIO_CONFIG_STEREO && arg->evsMode )
		{
		fprintf( stderr, "Error: Both non-diegetic panning and stereo output specified!\n\n" );
		@@ -1547,7 +1476,6 @@ static bool parseCmdlIVAS_dec(
		return false;
		}

		#endif
		if ( arg->outputMdFilename != NULL && arg->outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM )
		{
		fprintf( stderr, "Error: Output split rendering metadata file is supported for BINAURAL_SPLIT_PCM output config. only\n\n" );
		@@ -1639,9 +1567,7 @@ static bool parseCmdlIVAS_dec(
		static void usage_dec( void )
		{
		fprintf( stdout, "Usage for EVS: IVAS_dec.exe [Options] Fs bitstream_file output_file\n" );
		#ifdef FIX_1419_MONO_STEREO_UMX
		fprintf( stdout, " OR usage for IVAS (below) with -evs option and OutputConf\n" );
		#endif
		fprintf( stdout, "Usage for IVAS: IVAS_dec.exe [Options] OutputConf Fs bitstream_file output_file\n\n" );

		fprintf( stdout, "Mandatory parameters:\n" );
		@@ -1658,9 +1584,7 @@ static void usage_dec( void )

		fprintf( stdout, "Options:\n" );
		fprintf( stdout, "--------\n" );
		#ifdef FIX_1419_MONO_STEREO_UMX
		fprintf( stdout, "-evs : Specify that the supplied bitstream is an EVS bitstream\n" );
		#endif
		fprintf( stdout, "-VOIP : VoIP mode: RTP in G192\n" );
		fprintf( stdout, "-VOIP_hf_only=0 : VoIP mode: EVS RTP Payload Format hf_only=0 in rtpdump\n" );
		fprintf( stdout, "-VOIP_hf_only=1 : VoIP mode: EVS or IVAS RTP Payload Format hf_only=1 in rtpdump\n" );

apps/renderer.c

+4 −157

Original line number	Diff line number	Diff line
		@@ -46,9 +46,7 @@
		#include "masa_file_writer.h"
		#include "render_config_reader.h"
		#include "rotation_file_reader.h"
		#ifdef FIX_1053_REVERB_RECONFIGURATION
		#include "aeid_file_reader.h"
		#endif
		#include "split_render_file_read_write.h"
		#include "split_rend_bfi_file_reader.h"
		#include "vector3_pair_file_reader.h"
		@@ -394,11 +392,7 @@ static const CmdLnParser_Option cliOptions[] = {
		.id = CmdLnOptionId_acousticEnvironmentId,
		.match = "acoustic_environment_id",
		.matchShort = "aeid",
		#ifdef FIX_1053_REVERB_RECONFIGURATION
		.description = "Acoustic environment ID (number > 0) alternatively, it can be a text file where each line contains \"ID duration\" for BINAURAL_ROOM_REVERB output configuration.",
		#else
		.description = "Acoustic environment ID( number > 0 ) or a sequence thereof in the format [ID1:duration1,ID2:duration2...] without braces and spaces, with ':' character separating ID from duration and ',' separating ID and duration pairs, where duration is specified in frames for BINAURAL_ROOM_REVERB output configuration.",
		#endif
		},
		{
		.id = CmdLnOptionId_roomSize,
		@@ -810,9 +804,7 @@ int main(
		int32_t delayTimeScale = 0;
		int16_t i, numChannels;
		uint16_t aeID;
		#ifdef FIX_1053_REVERB_RECONFIGURATION
		IVAS_RENDER_CONFIG_DATA renderConfig;
		#endif
		ivas_error error = IVAS_ERR_OK;

		#ifdef WMOPS
		@@ -1175,11 +1167,6 @@ int main(

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

		#ifndef FIX_1053_REVERB_RECONFIGURATION
		IVAS_RENDER_CONFIG_DATA renderConfig;
		#endif

		/* sanity check */
		if ( ( args.outConfig.audioConfig != IVAS_AUDIO_CONFIG_BINAURAL ) &&
		( args.outConfig.audioConfig != IVAS_AUDIO_CONFIG_BINAURAL_ROOM_IR ) &&
		@@ -1542,16 +1529,7 @@ int main(
		audioWriter = NULL;
		}

		#ifdef FIX_1437_LC3PLUS_EXTREND_HIRES
		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 )
		#else
		if ( ( error = IVAS_REND_GetSplitRendBitstreamHeader( hIvasRend, &bitsBuffer.config.codec, &bitsBuffer.config.poseCorrection, &bitsBuffer.config.codec_frame_size_ms, &bitsBuffer.config.isar_frame_size_ms ) ) != IVAS_ERR_OK )
		#endif
		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;
		@@ -1676,9 +1654,6 @@ int main(
		{
		if ( ++args.aeSequence.frameCounter >= args.aeSequence.pValidity[args.aeSequence.selected] )
		{
		#ifndef FIX_1053_REVERB_RECONFIGURATION
		IVAS_RENDER_CONFIG_DATA renderConfig;
		#endif
		if ( ++args.aeSequence.selected >= args.aeSequence.count )
		{
		args.aeSequence.selected = 0;
		@@ -2654,9 +2629,10 @@ static bool parseAcousticEnvironmentIds(
		const char *value,
		AcousticEnvironmentSequence *aeSequence )
		{
		#ifdef FIX_1053_REVERB_RECONFIGURATION
		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;
		@@ -2672,72 +2648,6 @@ static bool parseAcousticEnvironmentIds(
		}
		aeidFileReader_close( &aeidReader );
		}
		#else
		uint16_t k;
		char config_string[RENDERER_MAX_METADATA_LINE_LENGTH];
		char *s;
		char *token;

		strncpy( config_string, value, RENDERER_MAX_METADATA_LINE_LENGTH );
		s = config_string;
		token = config_string;

		if ( !is_digits_only( config_string ) )
		{

		for ( k = 0; s[k]; )
		{
		s[k] == ',' ? k++ : *s++;
		}
		k++;

		if ( k == 0 )
		{
		fprintf( stdout, "Error: Invalid acoustic environment sequence specified: %s\n\n", config_string );
		return false;
		}

		if ( NULL == ( aeSequence->pID = malloc( sizeof( uint16_t ) * k ) ) \|\|
		NULL == ( aeSequence->pValidity = malloc( sizeof( uint16_t ) * k ) ) )
		{
		fprintf( stdout, "Error: Unable to allocate memory for acoustic environment sequence: %s\n\n", config_string );
		return false;
		}

		aeSequence->count = k;

		k = 0;

		token = strtok( config_string, ":" );

		while ( token != NULL )
		{
		if ( !is_number( token ) )
		{
		fprintf( stdout, "Error: Invalid token %s found in acoustic environment sequence: %s\n\n", token, config_string );
		return false;
		}
		aeSequence->pID[k] = (uint16_t) atoi( token );

		token = strtok( NULL, "," );
		if ( !is_number( token ) )
		{
		fprintf( stdout, "Error: Invalid token %s found in acoustic environment sequence: %s\n\n", token, config_string );
		return false;
		}
		aeSequence->pValidity[k] = (uint16_t) atoi( token );

		token = strtok( NULL, ":" );
		k++;
		}

		if ( k != aeSequence->count )
		{
		fprintf( stdout, "Error while parsing acoustic environment sequence: %s\n\n", config_string );
		return false;
		}
		}
		#endif
		else
		{
		/* A single acoustic environment */
		@@ -3191,25 +3101,12 @@ void getMetadataFromFileReader(
		fprintf( stderr, "\nError (%s) while reading ISM metadata from: %s\n\n", ivas_error_to_string( error ), IsmFileReader_getFilePath( ismReader ) );
		exit( -1 );
		}
		#ifdef FIX_2084_FLOATING_POINT_LEFTOVERS

		objectMetadataBuffer->positions[objIdx].azimuth_fx = ismMetadata.azimuth_fx;
		objectMetadataBuffer->positions[objIdx].elevation_fx = ismMetadata.elevation_fx;
		objectMetadataBuffer->positions[objIdx].radius_fx = ismMetadata.radius_fx;
		objectMetadataBuffer->positions[objIdx].yaw_fx = ismMetadata.yaw_fx;
		objectMetadataBuffer->positions[objIdx].pitch_fx = ismMetadata.pitch_fx;
		#else
		objectMetadataBuffer->positions[objIdx].azimuth_fx = (Word32) ( ismMetadata.azimuth * ( 1 << 22 ) );
		objectMetadataBuffer->positions[objIdx].elevation_fx = (Word32) ( ismMetadata.elevation * ( 1 << 22 ) );
		objectMetadataBuffer->positions[objIdx].radius_fx = (Word16) ( ismMetadata.radius * ( 1 << 9 ) );
		objectMetadataBuffer->positions[objIdx].yaw_fx = (Word32) ( ismMetadata.yaw * ( 1 << 22 ) );
		objectMetadataBuffer->positions[objIdx].pitch_fx = (Word32) ( ismMetadata.pitch * ( 1 << 22 ) );

		objectMetadataBuffer->positions[objIdx].azimuth = ismMetadata.azimuth;
		objectMetadataBuffer->positions[objIdx].elevation = ismMetadata.elevation;
		objectMetadataBuffer->positions[objIdx].radius = ismMetadata.radius;
		objectMetadataBuffer->positions[objIdx].yaw = ismMetadata.yaw;
		objectMetadataBuffer->positions[objIdx].pitch = ismMetadata.pitch;
		#endif
		objectMetadataBuffer->positions[objIdx].non_diegetic_flag = ismMetadata.non_diegetic_flag;

		return;
		@@ -3263,25 +3160,15 @@ static void IsmPositionProvider_getNextFrame(
		/* Otherwise fall back to default position */
		else
		{
		#ifdef FIX_2084_FLOATING_POINT_LEFTOVERS
		objectMetadataBuffer->positions[objIdx].azimuth_fx = 0;
		objectMetadataBuffer->positions[objIdx].elevation_fx = 0;
		objectMetadataBuffer->positions[objIdx].radius_fx = 512; // 1.f in Q9
		objectMetadataBuffer->positions[objIdx].yaw_fx = 0;
		objectMetadataBuffer->positions[objIdx].pitch_fx = 0;
		objectMetadataBuffer->positions[objIdx].non_diegetic_flag = 0;
		#else
		objectMetadataBuffer->positions[objIdx].azimuth = 0.0f;
		objectMetadataBuffer->positions[objIdx].elevation = 0.0f;
		objectMetadataBuffer->positions[objIdx].radius = 1.0f;
		objectMetadataBuffer->positions[objIdx].yaw = 0.0f;
		objectMetadataBuffer->positions[objIdx].pitch = 0.0f;
		objectMetadataBuffer->positions[objIdx].non_diegetic_flag = 0;
		#endif
		}

		/* Wrap azimuth to lie within (-180, 180] range */
		#ifdef FIX_2084_FLOATING_POINT_LEFTOVERS
		while ( LT_32( objectMetadataBuffer->positions[objIdx].azimuth_fx, 0 ) )
		{
		objectMetadataBuffer->positions[objIdx].azimuth_fx = L_add( objectMetadataBuffer->positions[objIdx].azimuth_fx, DEG_360_IN_Q22 );
		@@ -3306,35 +3193,6 @@ static void IsmPositionProvider_getNextFrame(

		/* Clamp pitch to lie within [-90, 90] range (can't be wrapped easily) */
		objectMetadataBuffer->positions[objIdx].pitch_fx = L_min( L_max( objectMetadataBuffer->positions[objIdx].pitch_fx, -DEG_90_IN_Q22 ), DEG_90_IN_Q22 );
		#else
		while ( objectMetadataBuffer->positions[objIdx].azimuth < 0.0f )
		{
		objectMetadataBuffer->positions[objIdx].azimuth += 360.0f;
		}
		while ( objectMetadataBuffer->positions[objIdx].azimuth >= 360.0f )
		{
		objectMetadataBuffer->positions[objIdx].azimuth -= 360.0f;
		}

		/* Clamp elevation to lie within [-90, 90] range (can't be wrapped easily) */
		objectMetadataBuffer->positions[objIdx].elevation = min( max( objectMetadataBuffer->positions[objIdx].elevation, -90 ), 90 );
		objectMetadataBuffer->positions[objIdx].azimuth_fx = (Word32) ( objectMetadataBuffer->positions[objIdx].azimuth * ( 1 << 22 ) );
		objectMetadataBuffer->positions[objIdx].elevation_fx = (Word32) ( objectMetadataBuffer->positions[objIdx].elevation * ( 1 << 22 ) );
		/* Wrap yaw to lie within (-180, 180] range */
		while ( objectMetadataBuffer->positions[objIdx].yaw < 0.0f )
		{
		objectMetadataBuffer->positions[objIdx].yaw += 360.0f;
		}
		while ( objectMetadataBuffer->positions[objIdx].yaw >= 360.0f )
		{
		objectMetadataBuffer->positions[objIdx].yaw -= 360.0f;
		}

		/* Clamp pitch to lie within [-90, 90] range (can't be wrapped easily) */
		objectMetadataBuffer->positions[objIdx].pitch = min( max( objectMetadataBuffer->positions[objIdx].pitch, -90 ), 90 );
		objectMetadataBuffer->positions[objIdx].yaw_fx = (Word32) ( ( objectMetadataBuffer->positions[objIdx].yaw ) * ( 1 << Q22 ) );
		objectMetadataBuffer->positions[objIdx].pitch_fx = (Word32) ( ( objectMetadataBuffer->positions[objIdx].pitch ) * ( 1 << Q22 ) );
		#endif
		}

		++positionProvider->frameCounter;
		@@ -3640,23 +3498,12 @@ static void parseObjectPosition(
		exit( -1 );
		}

		#ifdef FIX_2084_FLOATING_POINT_LEFTOVERS
		position->azimuth_fx = (Word32) ( meta_prm[0] * ( 1 << Q22 ) ); /* Q22 */
		position->elevation_fx = (Word32) ( meta_prm[1] * ( 1 << Q22 ) ); /* Q22 */
		position->radius_fx = (Word16) ( meta_prm[2] * ( 1 << Q9 ) ); /* Q9 */
		position->yaw_fx = (Word32) ( meta_prm[5] * ( 1 << Q22 ) ); /* Q22 */
		position->pitch_fx = (Word32) ( meta_prm[6] * ( 1 << Q22 ) ); /* Q22 */
		position->non_diegetic_flag = (Word16) meta_prm[7];
		#else
		position->azimuth = meta_prm[0];
		position->elevation = meta_prm[1];
		position->azimuth_fx = (Word32) ( meta_prm[0] * ( 1 << 22 ) );
		position->elevation_fx = (Word32) ( meta_prm[1] * ( 1 << 22 ) );
		position->radius = meta_prm[2];
		position->yaw = meta_prm[5];
		position->pitch = meta_prm[6];
		position->non_diegetic_flag = (int16_t) meta_prm[7];
		#endif

		return;
		}

lib_com/common_api_types.h

+0 −9

Original line number	Diff line number	Diff line
		@@ -133,15 +133,6 @@ typedef struct _IVAS_ISM_METADATA
		Word32 gainFactor_fx; /* Q29 */
		Word32 yaw_fx; /* Q22 */
		Word32 pitch_fx; /* Q22 */
		#ifndef FIX_2084_FLOATING_POINT_LEFTOVERS
		float azimuth;
		float elevation;
		float radius;
		float spread;
		float gainFactor;
		float yaw;
		float pitch;
		#endif
		Word16 non_diegetic_flag;
		Word32 gain_fx; /* Q29 */

lib_com/ivas_cnst.h

+0 −4

Original line number	Diff line number	Diff line
		@@ -1291,11 +1291,7 @@ typedef enum
		----------------------------------------------------------------------------------/

		#define MC_LS_SETUP_BITS 3 /* number of bits for writing the MC LS configuration */
		#ifdef FIX_1419_MONO_STEREO_UMX
		#define LS_SETUP_CONVERSION_NUM_MAPPINGS 41 /* number of mappings for LS setup conversion */
		#else
		#define LS_SETUP_CONVERSION_NUM_MAPPINGS 35 /* number of mappings for LS setup conversion */
		#endif

		typedef enum
		{

lib_com/options.h

+10 −16

Original line number	Diff line number	Diff line
		@@ -69,7 +69,10 @@
		/* #################### End DEBUGGING switches ############################ */


		/* ################### Start FIXES switches ########################### */
		/* ################### Start MAINTENANCE switches ########################### */

		/* ################### Start BE switches ################################# */
		/* only BE switches wrt wrt. TS 26.251 V3.0 */

		/#define FIX_I4_OL_PITCH/ /* fix open-loop pitch used for EVS core switching */
		#define FIX_1990_SANITIZER_IN_REVERB_LOAD /* Nokia: Fix issue part of issue 1990 by introducing missing free of structure - keep until #2059 is addressed */
		@@ -82,9 +85,7 @@


		#define REMOVE_BASOP_Util_Divide3232_Scale_cadence /* remove this division variant */
		#define TEMP_FIX_2088_MSAN_INIT_ERROR /* Eri: Temporary fix for Issue 2088 - MSAN error. Will come with later port of JBM+Split rendering update */
		#define FIX_2092_ASSERT_IN_OMASA_RENDER /* FhG, Nokia: Fix LTV crash due to overflow in OMASA EXT output */
		#define FIX_2084_FLOATING_POINT_LEFTOVERS /* FhG: convert floating-point leftovers in IVAS_ENC_FeedObjectMetadata() */
		#define FIX_2141_ASSERT_IN_OMASA_BITRATE_SWITSCHING /* FhG: Replace L_shl with L_shl_sat to prevent overflow when calculating scale factors for very small numbers in the logarithmic domain */
		#define FIX_APA_EXECS_SCALING /* VA: fix scaling of JBM APA buffer */
		#define FIX_2164_ASSERT_IN_OMASA_PREPROC_FOR_EDIT /* Nokia: Issue 2164: Prevent overflow when calculating equalization coefficient for editing before clamping to safe range */
		@@ -117,24 +118,17 @@



		/* ################### End FIXES switches ########################### */

		/* #################### Start BASOP porting switches ############################ */

		#define NONBE_1244_FIX_SWB_BWE_MEMORY /* VA: issue 1244: fix to SWB BWE memory in case of switching from FB coding - pending a review by Huawei */
		#define FIX_1053_REVERB_RECONFIGURATION
		#define FIX_1419_MONO_STEREO_UMX /* FhG: fix for issue 1419 : support upmix to all output formats for mono and stereo */
		#define FIX_1437_LC3PLUS_EXTREND_HIRES /* FhG: fix external renderer split bitstream header writing causing a crash for LC3plus High-res mode */


		/* #################### End BE switches ################################## */

		/* #################### Start NON-BE switches ############################ */

		/* any switch which is non-be wrt last submitted floating point code */
		/* any switch which is non-be wrt. TS 26.251 V3.0 */


		/* #################### End BASOP porting switches ############################ */

		/* ##################### End NON-BE switches ########################### */

		/* ################## End MAINTENANCE switches ######################### */

		/* clang-format on */

		#endif