Merge branch 'main' into 1129-external-renderer-binaural-room-reverb-output-level-high (e4195c38) · Commits · IVAS Codec Public Collaboration / IVAS Codec

apps/decoder.c

+0 −47

Original line number	Diff line number	Diff line
		@@ -128,11 +128,7 @@ typedef struct
		bool customLsOutputEnabled;
		char *customLsSetupFilename;
		int16_t orientation_tracking;
		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		bool non_diegetic_pan_enabled;
		#else
		int16_t Opt_non_diegetic_pan;
		#endif
		float non_diegetic_pan_gain;
		bool renderConfigEnabled;
		char *renderConfigFilename;
		@@ -152,11 +148,7 @@ typedef struct
		#endif
		#endif
		AcousticEnvironmentSequence aeSequence;
		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		bool dpidEnabled;
		#else
		int16_t Opt_dpid_on;
		#endif
		uint16_t directivityPatternId[IVAS_MAX_NUM_OBJECTS];

		} DecArguments;
		@@ -415,12 +407,7 @@ int main(
		if ( arg.renderConfigEnabled )
		{
		/* sanity check */
		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		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 && arg.non_diegetic_pan_enabled == false )
		#else
		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 &&
		arg.Opt_non_diegetic_pan == 0 )
		#endif
		{
		fprintf( stderr, "\nError: Renderer configuration file cannot be used in this output configuration.\n\n" );
		goto cleanup;
		@@ -440,13 +427,8 @@ int main(
		asked_frame_size = arg.renderFramesize;
		uint16_t aeID = arg.aeSequence.count > 0 ? arg.aeSequence.pID[0] : 65535;

		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		if ( ( error = IVAS_DEC_Configure( hIvasDec, arg.output_Fs, arg.outputConfig, arg.tsmEnabled, arg.renderFramesize, arg.customLsOutputEnabled, arg.hrtfReaderEnabled, arg.enableHeadRotation, arg.enableExternalOrientation, arg.orientation_tracking, arg.renderConfigEnabled, arg.non_diegetic_pan_enabled, arg.non_diegetic_pan_gain,
		arg.dpidEnabled, aeID, arg.delayCompensationEnabled ) ) != IVAS_ERR_OK )
		#else
		if ( ( error = IVAS_DEC_Configure( hIvasDec, arg.output_Fs, arg.outputConfig, arg.tsmEnabled, arg.renderFramesize, arg.customLsOutputEnabled, arg.hrtfReaderEnabled, arg.enableHeadRotation, arg.enableExternalOrientation, arg.orientation_tracking, arg.renderConfigEnabled, arg.Opt_non_diegetic_pan, arg.non_diegetic_pan_gain,
		arg.Opt_dpid_on, aeID, arg.delayCompensationEnabled ) ) != IVAS_ERR_OK )
		#endif
		{
		fprintf( stderr, "\nConfigure failed: %s\n\n", IVAS_DEC_GetErrorMessage( error ) );
		goto cleanup;
		@@ -612,12 +594,7 @@ int main(

		/* 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 &&
		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED && arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM && arg.non_diegetic_pan_enabled == false )
		#else
		arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED && arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM &&
		arg.Opt_non_diegetic_pan == 0 )
		#endif
		{
		fprintf( stderr, "\nExternal Renderer Config is supported only when binaural output configurations is used as output OR when Split rendering mode is enabled. Exiting. \n" );
		goto cleanup;
		@@ -1088,20 +1065,12 @@ static bool parseCmdlIVAS_dec(

		arg->renderConfigEnabled = false;
		arg->renderConfigFilename = NULL;
		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		arg->dpidEnabled = false;
		#else
		arg->Opt_dpid_on = 0;
		#endif

		arg->outputMdFilename = NULL;

		arg->inputFormat = IVAS_DEC_INPUT_FORMAT_G192;
		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		arg->non_diegetic_pan_enabled = false;
		#else
		arg->Opt_non_diegetic_pan = 0;
		#endif
		arg->non_diegetic_pan_gain = 0.f;
		arg->tsmEnabled = false;
		arg->renderFramesize = IVAS_RENDER_FRAMESIZE_20MS;
		@@ -1440,11 +1409,7 @@ static bool parseCmdlIVAS_dec(
		else if ( strcmp( argv_to_upper, "-NON_DIEGETIC_PAN" ) == 0 )
		{
		i++;
		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		arg->non_diegetic_pan_enabled = true;
		#else
		arg->Opt_non_diegetic_pan = 1;
		#endif
		strncpy( argv_to_upper, argv[i], sizeof( argv_to_upper ) - 1 );
		argv_to_upper[sizeof( argv_to_upper ) - 1] = '\0';
		to_upper( argv_to_upper );
		@@ -1541,11 +1506,7 @@ static bool parseCmdlIVAS_dec(
		{
		int16_t id, tmp;

		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		arg->dpidEnabled = true;
		#else
		arg->Opt_dpid_on = 1;
		#endif
		++i;
		tmp = 0;
		while ( is_number( argv[i + tmp] ) && tmp < IVAS_MAX_NUM_OBJECTS )
		@@ -1626,11 +1587,7 @@ static bool parseCmdlIVAS_dec(
		}
		i++;

		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		if ( arg->non_diegetic_pan_enabled && arg->outputConfig != IVAS_AUDIO_CONFIG_STEREO )
		#else
		if ( ( arg->Opt_non_diegetic_pan ) && ( arg->outputConfig != IVAS_AUDIO_CONFIG_STEREO ) )
		#endif
		{
		fprintf( stderr, "Error: non-diegetic panning is supported in stereo only\n\n" );
		usage_dec();
		@@ -1649,11 +1606,7 @@ static bool parseCmdlIVAS_dec(
		arg->outputConfig = IVAS_AUDIO_CONFIG_MONO;
		arg->decMode = IVAS_DEC_MODE_EVS;

		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		if ( arg->non_diegetic_pan_enabled )
		#else
		if ( ( arg->Opt_non_diegetic_pan ) )
		#endif
		{
		arg->outputConfig = IVAS_AUDIO_CONFIG_STEREO;
		}

apps/encoder.c

+10 −1

Original line number	Diff line number	Diff line
		@@ -1519,6 +1519,7 @@ static bool parseCmdlIVAS_enc(
		{
		fprintf( stderr, "Error: Too low number of ISM channels specified!\n\n" );
		usage_enc();
		return false;
		}
		else
		{
		@@ -1530,6 +1531,7 @@ static bool parseCmdlIVAS_enc(
		{
		fprintf( stderr, "Error: Too high number of ISM channels!\n\n" );
		usage_enc();
		return false;
		}
		}
		}
		@@ -1537,6 +1539,7 @@ static bool parseCmdlIVAS_enc(
		{
		fprintf( stderr, "Error: Number of ISM channels not specified!\n\n" );
		usage_enc();
		return false;
		}
		if ( i < argc - 4 )
		{
		@@ -1556,7 +1559,7 @@ static bool parseCmdlIVAS_enc(
		default:
		fprintf( stderr, "Error: MASA channels must be 1 or 2.\n\n" );
		usage_enc();
		break;
		return false;
		}
		}

		@@ -1581,6 +1584,7 @@ static bool parseCmdlIVAS_enc(
		{
		fprintf( stderr, "Error: not enough arguments\n\n" );
		usage_enc();
		return false;
		}
		}

		@@ -1593,6 +1597,7 @@ static bool parseCmdlIVAS_enc(
		{
		fprintf( stderr, "Error: not enough MASA arguments\n\n" );
		usage_enc();
		return false;
		}
		}
		else if ( strcmp( to_upper( argv[i] ), "-ISM_SBA" ) == 0 )
		@@ -1611,6 +1616,7 @@ static bool parseCmdlIVAS_enc(
		{
		fprintf( stderr, "Error: Too low number of ISM channels specified!\n\n" );
		usage_enc();
		return false;
		}
		else
		{
		@@ -1622,6 +1628,7 @@ static bool parseCmdlIVAS_enc(
		{
		fprintf( stderr, "Error: Too high number of ISM channels!\n\n" );
		usage_enc();
		return false;
		}
		}
		}
		@@ -1629,6 +1636,7 @@ static bool parseCmdlIVAS_enc(
		{
		fprintf( stderr, "Error: Number of ISM channels not specified!\n\n" );
		usage_enc();
		return false;
		}

		if ( i < argc - 4 )
		@@ -1680,6 +1688,7 @@ static bool parseCmdlIVAS_enc(
		{
		fprintf( stderr, "Error: not enough arguments\n\n" );
		usage_enc();
		return false;
		}
		}
		}

ci/basop-pages/create_report_pages.py

+1 −0

Original line number	Diff line number	Diff line
		@@ -105,6 +105,7 @@ COLUMNS = [
		"MLD",
		"MAXIMUM ABS DIFF",
		"MIN_SSNR",
		"MIN_ODG",
		]
		COLUMNS_GLOBAL = COLUMNS[:1]
		COLUMNS_DIFFERENTIAL = COLUMNS[3:]

lib_com/gs_bitallocation.c

+77 −33

Original line number	Diff line number	Diff line
		@@ -46,14 +46,12 @@
		#include "ivas_prot.h"
		#include "wmc_auto.h"


		#ifdef NON_BE_FIX_1137_GSC_IVAS_FXFLT_DECODING
		/-------------------------------------------------------------------
		* Local function prototypes
		* Local constants
		-------------------------------------------------------------------/

		static float Find_bit_frac( const int16_t nb_band, const int16_t remaining_bits );

		static void reajust_bits( float *bits_per_bands, const int16_t st_band, const int16_t end_band, const int16_t sum_bit_in, const int16_t bit_bdgt_in );
		#ifdef NON_BE_FIX_1137_GSC_IVAS_FXFLT_DECODING
		#define Q15_0_33 10922 /* 0.33 */
		#define Q18_0_1 26214 /* 0.1 */
		#define Q18_0_50 131072 /* 0.50 */
		@@ -72,9 +70,24 @@ static void reajust_bits( float *bits_per_bands, const int16_t st_band, const in
		#define Q31_0_02 42949673 /* 0.02 */
		#define Q31_0_17 365072220 /* 0.17 */
		#define Q31_0_23 493921239 /* 0.23 */

		#endif

		/-------------------------------------------------------------------
		* Local function prototypes
		-------------------------------------------------------------------/

		static float Find_bit_frac( const int16_t nb_band, const int16_t remaining_bits );

		static void reajust_bits( float *bits_per_bands, const int16_t st_band, const int16_t end_band, const int16_t sum_bit_in, const int16_t bit_bdgt_in );

		#ifdef NON_BE_FIX_1137_GSC_IVAS_FXFLT_DECODING
		static Word16 Find_norm_inv( const Word32 ToDivide, Word16 *e_div );

		static Word16 Find_bit_alloc_IVAS_int( const Word32 core_brate, const Word16 GSC_IVAS_mode, const Word16 Diff_len, const Word16 nb_tot_bands, const Word16 L_frame, Word16 bit, Word16 max_ener_band, float ener_vec, float bits_per_bands );

		static Word16 maximum_fx( const Word16 vec_fx, const Word16 lvec_fx, Word16 max_fx );

		#endif

		/-------------------------------------------------------------------
		@@ -398,7 +411,7 @@ void bands_and_bit_alloc(
		}
		}
		#else
		nb_tot_bands = Find_bit_alloc_IVAS_int( core_brate, GSC_IVAS_mode, Diff_len, nb_tot_bands, L_frame, bit, max_ener_band, ener_vec, bits_per_bands );
		nb_tot_bands = (int16_t) Find_bit_alloc_IVAS_int( (Word32) core_brate, (Word16) GSC_IVAS_mode, (Word16) Diff_len, (Word16) nb_tot_bands, (Word16) L_frame, (Word16 ) bit, (Word16 ) max_ener_band, ener_vec, bits_per_bands );
		nb_bands = nb_tot_bands;
		#endif
		}
		@@ -962,9 +975,17 @@ static float Find_bit_frac(

		return ( var_out );
		}

		#ifdef NON_BE_FIX_1137_GSC_IVAS_FXFLT_DECODING

		static Word16 Find_bit_alloc_IVAS_int( /o: Number of band to encode /
		/-------------------------------------------------------------------
		* Find_bit_alloc_IVAS_int()
		*
		*
		-------------------------------------------------------------------/

		/! r: Number of bands to encode /
		static Word16 Find_bit_alloc_IVAS_int(
		const Word32 core_brate, /* i : core bit rate */
		const Word16 GSC_IVAS_mode, /* i : GSC IVAS mode */
		const Word16 Diff_len, /* i : Length of the difference signal (before pure spectral)*/
		@@ -978,27 +999,26 @@ static Word16 Find_bit_alloc_IVAS_int( /*o: Number
		{
		Word32 mp, mb, nb_bands_adj, bit_adj;
		Word16 nb_pulse_per_band[MBANDS_GN_BITALLOC16k];
		Word32 SWB_bit_budget; // Q0 -> Q18
		Word32 SWB_bit_budget; /* Q0 -> Q18 */
		Word16 i, j, nb_bands_max, st_band, nb_tot_bands_loc, etmp;
		Word32 sum_bit /Q18/, bit_fracf /Q18/;
		Word16 d_tmp, e_div, tmp16, ener_vec[MBANDS_GN_BITALLOC16k];
		Word32 Ltmp, etmp_32fx, bits_per_bands[MBANDS_GN_BITALLOC16k];


		SWB_bit_budget = *bit; // Q0
		SWB_bit_budget = bit; / Q0 */
		st_band = 5;
		nb_bands_max = nb_tot_bands_in;


		for ( i = 0; i < MBANDS_GN; i++ )
		{
		ener_vec[i] = (short) ( ener_vec_io[i] ); /* Q12 -> Q12 */
		ener_vec[i] = (Word16) ( ener_vec_io[i] ); /* Q12 -> Q12 */
		}

		if ( L_frame == L_FRAME16k )
		{
		for ( i = MBANDS_GN; i < MBANDS_GN_BITALLOC16k; i++ )
		{
		ener_vec[i] = (short) ( ener_vec_io[i] * 4096.0 + 0.5f ); /* Q0 -> Q12 */
		ener_vec[i] = (Word16) ( ener_vec_io[i] * 4096.0 + 0.5f ); /* Q0 -> Q12 */
		}
		}

		@@ -1020,12 +1040,14 @@ static Word16 Find_bit_alloc_IVAS_int( /*o: Number
		{
		nb_bands_max -= 6;
		}

		if ( GSC_IVAS_mode == 2 )
		{
		/* bit_fracf += 0.1f; */
		bit_fracf += Q18_0_1; /* Q18*/
		nb_bands_max -= 1;
		}

		if ( GSC_IVAS_mode == 3 )
		{
		/* bit_fracf -= 0.1f; */
		@@ -1041,12 +1063,12 @@ static Word16 Find_bit_alloc_IVAS_int( /*o: Number
		if ( GSC_IVAS_mode == 1 && core_brate < GSC_L_RATE_STG )
		{
		/* nb_bands_adj = 0.0125f * SWB_bit_budget - 0.75f;*/
		nb_bands_adj = L_sub( Mpy_32_32( Q31_0_0125, L_shl( SWB_bit_budget, 18 ) ), Q18_0_75 ); // Q18
		nb_bands_adj = L_sub( Mpy_32_32( Q31_0_0125, L_shl( SWB_bit_budget, 18 ) ), Q18_0_75 ); /* Q18 */
		}
		else if ( GSC_IVAS_mode != 2 && core_brate > GSC_H_RATE_STG )
		{
		/nb_bands_adj = 0.02f SWB_bit_budget - 1.2f;*/
		nb_bands_adj = L_sub( Mpy_32_32( Q31_0_02, L_shl( SWB_bit_budget, 18 ) ), Q18_1_2 ); // Q18
		nb_bands_adj = L_sub( Mpy_32_32( Q31_0_02, L_shl( SWB_bit_budget, 18 ) ), Q18_1_2 ); /* Q18 */
		}
		/nb_bands_max = (int16_t)(nb_bands_max nb_bands_adj + 0.5f);*/

		@@ -1135,10 +1157,10 @@ static Word16 Find_bit_alloc_IVAS_int( /*o: Number
		max_ener_band[j] = i;
		ener_vec[i] = MIN16B;
		}

		/* Recompute the final bit distribution for HF */
		IF( nb_tot_bands_loc > st_band )
		{
		/* This is not bit exact because of the precision lost */
		/* mb = ( SWB_bit_budget * 2 / ( nb_tot_bands_loc - st_band ) ) - mp; */
		d_tmp = Find_norm_inv( L_deposit_h( sub( nb_tot_bands_loc, st_band ) ), &e_div );
		mb = L_sub( L_shr( Mpy_32_16_1( L_shl( SWB_bit_budget, 1 ), d_tmp ), e_div ), mp ); /* Q18 */
		@@ -1155,9 +1177,10 @@ static Word16 Find_bit_alloc_IVAS_int( /*o: Number
		bits_per_bands[max_ener_band[j]] = mb;
		}
		mb -= bit_fracf;
		SWB_bit_budget -= bits_per_bands[max_ener_band[j]]; // Q18
		SWB_bit_budget -= bits_per_bands[max_ener_band[j]]; /* Q18 */
		}
		}

		/* Series of verification in case bit allocated != the budget */
		if ( SWB_bit_budget > 0 )
		{
		@@ -1191,7 +1214,6 @@ static Word16 Find_bit_alloc_IVAS_int( /*o: Number
		bits_per_bands[i] = Q18_112;
		j = add( j, add( i, 1 ) );
		}

		/* safety check for overage bit reallocation */
		/* else if (bits_per_bands[i] + sum_bit / 3 > 112) */
		else if ( L_add( bits_per_bands[i], Mpy_32_16_1( sum_bit, Q15_0_33 ) ) > Q18_112 )
		@@ -1210,24 +1232,46 @@ static Word16 Find_bit_alloc_IVAS_int( /*o: Number
		bits_per_bands[i] = L_add( bits_per_bands[i], sum_bit );
		}
		}

		for ( i = 0; i < MBANDS_GN_BITALLOC16k; i++ )
		{
		bits_per_bands_o[i] = (float) bits_per_bands[i] / 262144.0f; /* Q18 -> float */
		}

		return nb_tot_bands_loc;
		}

		static Word16 Find_norm_inv( const Word32 ToDivide, Word16 e_div ) / Find normalized 1 / ToDivide */

		/-------------------------------------------------------------------
		* Find_norm_inv()
		*
		* Find normalized 1 / ToDivide
		-------------------------------------------------------------------/

		static Word16 Find_norm_inv(
		const Word32 ToDivide,
		Word16 *e_div )
		{
		Word16 d_tmp, e_tmp;

		e_tmp = norm_l( ToDivide );
		d_tmp = round_fx( L_shl( ToDivide, e_tmp ) );
		d_tmp = div_s( 16384, d_tmp ); /* 1.0 in Q14, dividend is normalize so >= 16384 as required for the division */
		*e_div = sub( 14, e_tmp );
		move16();

		return d_tmp;
		}
		static Word16 maximum_fx( /* o : index of the maximum value in the input vector */


		/-------------------------------------------------------------------
		* maximum_fx()
		*
		*
		-------------------------------------------------------------------/

		/! r: index of the maximum value in the input vector /
		static Word16 maximum_fx(
		const Word16 vec_fx, / i : input vector */
		const Word16 lvec_fx, /* i : length of input vector */
		Word16 max_fx / o : maximum value in the input vector */
		@@ -1235,8 +1279,8 @@ static Word16 maximum_fx( /* o : index of the maximum val
		{
		Word16 j, ind;
		Word16 tmp;
		ind = 0;

		ind = 0;
		tmp = vec_fx[0];

		for ( j = 1; j < lvec_fx; j++ )

lib_com/options.h

+5 −6

Original line number	Diff line number	Diff line
		@@ -160,10 +160,11 @@
		/#define FIX_I4_OL_PITCH/ /* fix open-loop pitch used for EVS core switching */

		//#define FIX_CREND_FIX_POINT_HRTF_FILE_FORMAT /* Orange issue 1031 : new fix point hrtf binary file format */
		//#define FIX_CREND_SIMPLIFY_CODE /* Ora : simplify line code in crend */
		//#define FIX_CREND_SIMPLIFY_CODE /* Orange: simplify line code in Crend */
		#define FLOAT_FIX_POINT_HRTF_FILE_FORMAT /* allows reading floation or fix point hrtf binary file format */
		#define FIX_1138_SBA_EXT_ERROR_PRINTOUT /* VA: issue 1138: Fix SBA EXT output call of audioCfg2channels() */
		#define FIX_1117_ISAR_MALLOC /* VA: issue 1117: call ISAR malloc() only at the initialization. */
		#define FIX_1129_EXT_REND_OUTPUT_HIGH /* Philips: issue 1129: External renderer BINAURAL_ROOM_REVERB format output level too high compared to internal rendering output */
		#define FIX_745_FIX_DATA_TYPE_CONVERSION /* VA: issue 745: implicit data type conversion when calling IVAS_DEC_Configure() */

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

		@@ -171,10 +172,8 @@
		/* any switch which is non-be wrt selection floating point code */
		/* all switches in this category should start with "NONBE_" */

		#define NONBE_MDCT_ST_DTX_SKIP_DEWHITENING_OF_NOISE_SHAPES_ON_SID_FRAMES /* FhG: issue 1133: skip de-whitening of bg noise shape after frameloss period if the first good frame is an SID */
		#define NONBE_MDCT_ST_PLC_DO_NOT_SCALE_OLD_OUT_IF_FIRST_GOOD_IS_SID /* FhG: issue 1133: in TCX PLC, don't scale hHQ_core->old_out after applying fade to noise in burst frame error */
		#define NONBE_FIX_1130_DIV_ZERO_LEV_DUR /* VA: issue 1130: avoid div by zero in L-D by thresholding R[0] to a min value of 100.0 */
		#define NON_BE_FIX_1137_GSC_IVAS_FXFLT_DECODING /* VA: Add fix point bit allocation for special GSC mode such that float and fixed point have the same final bit allocation */
		#define NONBE_FIX_1132_THRESHOLD_POW_IN_SWB_TBE /* VA: issue 1132: prevent division by extremely low energy value in SWB TBE */

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