merge from main (2252d49c) · Commits · SA4 / Audio / IVAS BASOP

apps/decoder.c

+51 −0

Original line number	Diff line number	Diff line
		@@ -122,7 +122,11 @@ 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;
		Word16 non_diegetic_pan_gain_fx; /* Q15 */
		bool renderConfigEnabled;
		@@ -136,7 +140,11 @@ typedef struct
		#else
		uint16_t acousticEnvironmentId;
		#endif
		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		bool dpidEnabled;
		#else
		int16_t Opt_dpid_on;
		#endif
		uint16_t directivityPatternId[IVAS_MAX_NUM_OBJECTS];

		} DecArguments;
		@@ -258,7 +266,11 @@ int main(
		if ( arg.hrtfReaderEnabled )
		{
		/* sanity check */
		#ifdef NONBE_1293_SR_HRTF
		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 )
		#else
		if ( arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL && arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_ROOM_IR && arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_ROOM_REVERB )
		#endif
		{
		arg.hrtfReaderEnabled = false;
		fprintf( stderr, "\nError: HRTF binary file cannot be used in this output configuration.\n\n" );
		@@ -381,8 +393,12 @@ 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;
		@@ -403,8 +419,13 @@ int main(
		asked_frame_size = arg.renderFramesize;
		#ifdef FIX_1053_REVERB_RECONFIGURATION
		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_fx,
		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_fx,
		arg.Opt_dpid_on, aeID, arg.delayCompensationEnabled ) ) != IVAS_ERR_OK )
		#endif
		#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_fx,
		arg.Opt_dpid_on, arg.acousticEnvironmentId, arg.delayCompensationEnabled ) ) != IVAS_ERR_OK )
		@@ -484,7 +505,11 @@ 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 &&
		arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED && arg.outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM &&
		#ifdef FIX_745_FIX_DATA_TYPE_CONVERSION
		arg.non_diegetic_pan_enabled == false )
		#else
		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;
		@@ -968,12 +993,20 @@ 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;
		@@ -1240,7 +1273,11 @@ 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 );
		@@ -1396,7 +1433,11 @@ 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 )
		@@ -1476,7 +1517,12 @@ static bool parseCmdlIVAS_dec(
		arg->customLsSetupFilename = argv[i];
		}
		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();
		@@ -1493,7 +1539,12 @@ 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;
		}

lib_com/ivas_prot_fx.h

+48 −49

Original line number	Diff line number	Diff line
		@@ -6198,7 +6198,7 @@ ivas_error ivas_core_enc_fx(
		const Word16 n_CoreChannels, /* i : number of core channels to be coded Q0*/
		Word16 old_inp_12k8_fx[][L_INP_12k8], /* i : buffer of old input signal Q_new-1*/
		Word16 old_inp_16k_fx[][L_INP], /* i : buffer of old input signal Q_new-1*/
		Word16 Q_new[],
		Word16 Q_new[], /* i : Q factor of speech buffers */
		Word32 ener_fx[], /* i : residual energy from Levinson-Durbin epsP_fx_q*/
		Word16 A_fx[][NB_SUBFR16k * ( M + 1 )], /* i : A(z) unquantized for the 4 subframes Q12*/
		Word16 Aw_fx[][NB_SUBFR16k * ( M + 1 )], /* i : weighted A(z) unquantized for subframes Q12*/
		@@ -6210,9 +6210,9 @@ ivas_error ivas_core_enc_fx(
		Word16 attack_flag[], /* i : attack flag (GSC or TC) Q0*/
		Word32 realBuffer_fx[][CLDFB_NO_COL_MAX][CLDFB_NO_CHANNELS_MAX], /* i/o: real buffer q_re_im_buf*/
		Word32 imagBuffer_fx[][CLDFB_NO_COL_MAX][CLDFB_NO_CHANNELS_MAX], /* i/o: imag buffer q_re_im_buf*/
		Word16 *q_re_im_buf,
		Word16 q_re_im_buf, / i : Q factor of re/in CLDFB buffers */
		Word16 old_wsp_fx[][L_WSP], /* i : weighted input signal buffer e_old_wsp*/
		Word16 e_old_wsp[],
		Word16 e_old_wsp[], /* i : Q factor of old_wsp buffer */
		const Word16 loc_harm[], /* i : harmonicity flag Q0*/
		const Word16 cor_map_sum_fx[], /* i : speech/music clasif. parameter Q8*/
		const Word16 vad_flag_dtx[], /* i : HE-SAD flag with additional DTX HO Q0*/
		@@ -6308,12 +6308,11 @@ ivas_error ivas_ism_metadata_enc_create_fx(
		Word32 element_brate_tmp[] /* o : element bitrate per object */
		);


		/----------------------------------------------------------------------------------
		* Parametric ISM prototypes
		----------------------------------------------------------------------------------/

		/! r: ISM format mode /

		ivas_error ivas_param_ism_enc_open_fx(
		Encoder_Struct st_ivas / i/o: IVAS encoder structure */
		);
		@@ -6328,11 +6327,11 @@ void ivas_ism_metadata_close(
		const Word16 first_idx /* i : index of first handle to deallocate */
		);


		ivas_error ivas_ism_enc_config(
		Encoder_Struct st_ivas / i/o: IVAS encoder structure */
		);


		/----------------------------------------------------------------------------------
		* ISM DTX prototypes
		----------------------------------------------------------------------------------/

lib_com/options.h

+26 −2

Original line number	Diff line number	Diff line
		@@ -90,6 +90,24 @@

		#define FIX_1931_BIN_COHR_CROSS_MIX /* FhG: correct binauralCoherenceCrossmixGains_fx calculation */
		#define FIX_1939_REVERB_DMX_OUT_Q /Dlb: output correct q factor for the reverb dmx function/

		#define EVS_BE_REUSAGE
		#ifdef EVS_BE_REUSAGE
		#define REUSE_EVS_BE_ACELP_LP_FILT
		#define REUSE_EVS_BE_ACELP_4T64
		#define REUSE_EVS_BE_ACELP_2t32
		#define REUSE_EVS_BE_ACELP_1t64
		#define REUSE_EVS_BE_ACELP_AVQ
		#define REUSE_EVS_BE_GAUSS
		#define REUSE_EVS_BE_GAINQ
		#define REUSE_EVS_BE_ACELP_PITCH
		#define REUSE_EVS_BE_ACELP_PITCH_PIT_Q
		#define REUSE_EVS_BE_GAINQ_LBR // BE by adding one condition, but could be harmonized in a non-BE way. There is a small difference in how Etot is computed

		#endif
		#define FIX_ISSUE_1817_REPLACE_CARRY_OVERFLOW /* FhG: bit-exact, replace carry and overflow operations by 64-bit operations, MR 1931 */
		#define FIX_1844_MISSING_FREE /* FhG: add missing free in ivas_binRenderer_convModuleClose_fx() */

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

		#define FIX_1372_ISAR_POST_REND
		@@ -122,6 +140,7 @@
		#define NONBE_FIX_1128_OSBA_EXT_OUTPUT_ORDER /* FhG: issue 1128: set output ambisonics order to input order for EXT output */
		#define FIX_1138_SBA_EXT_ERROR_PRINTOUT /* VA: issue 1138: Fix SBA EXT output call of audioCfg2channels() */
		#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_745_FIX_DATA_TYPE_CONVERSION /* VA: issue 745: implicit data type conversion when calling IVAS_DEC_Configure() */
		#define NONBE_FIX_981_PARAMBIN_DEFAULT_EARLY_PART /* Nokia: Set default early part energy correction to unity for BINAURAL_ROOM_REVERB */
		#define NONBE_FIX_1174_MCMASA_LBR_LOOP_ERROR /* Nokia: Fix issue 1174 by removing the unnecessary inner loop causing problems. */
		#define FIX_587_DEFAULT_REVERB /* Philips: issue 587: inconsistent default reverb parameters across renderers */
		@@ -138,13 +157,18 @@
		#define NONBE_1229_FIX_ISM1_DPID /* Eri: issue 1229: fix bug causing ISM 1 to use default -dpid instead of the specified one */
		#define FIX_1135_EXT_RENDERER_HANDLES /* VA: issue 1135: Memory usage reduction in external renderer: Allocate only handles that are really needed. */
		#define FIX_1209_SID_SIGNALING /* VA: issue 1209: remove dead code in IVAS SID signaling */
		#define NONBE_1250_MCMASA_LS_OUTPUT /* VA: issue 1250: fix crash in McMASA to custom LS output decoding */
		#define NONBE_FIX_1176_OSBA_REVERB_JBM_ASAN_ERROR /* Ericsson: Issue 1176, fix in TDREND_firfilt for subframes shorter than the filter length */
		#define NONBE_1131_ACELP_OOB /* VA: issue 1131: fix division-by-zero in acelp gain decoding caused by wrong length of buffer update when switching from HQ core to ACELP core */
		#define NONBE_1240_FIX_CORE_SELECTION_ISM_SW /* VA: issue 1240: Remove the forcing of the TCX core in ISM when switching from a high bitarte to a low one */
		#define FIX_1101_CLEANING_JBM_CALL /* VA: issue 1101: remove obsolete call of ivas_jbm_dec_tc_buffer_open() */
		#define FIX_VOIP_FUNCTIONS /* VA: fix data type mismatch in IVAS_DEC_VoIP_SetScale() + add sanity checks to API functions */
		#define FIX_1298_MEMORY_OPT_IVAS_CORE_ENC /* VA: issue 1298: Memory saving in ivas_core_enc() */
		#define NONBE_FIX_1337_MISSING_DIRECTIVITY_DISTATT_EXTREND /* Eri: issue 1337: Missing directivity setting and distance attenuation in external renderer IVAS_rend */
		#define NONBE_1214_PLC_LSF_MEMORY /* VA: issue 1224: reset ACELP PLC FEC memory in case of switching from MDCT stereo to TD/DFT stereo */
		#define NONBE_1293_SR_HRTF /* VA: issue 1293: add support of external HRTFs in split rendering */
		#define NONBE_1220_OMASA_JBM_BRATE_SW_FLUSH /* VA: issue 1220: fix bug in renderer flush in OMASA 1ISM JBM bitrate switching */
		#define FIX_1139_REV_COLORATION_SHORT_T60 /* Nokia,FhG: Fix issue 1139, prevent sound coloration artefacts at very low reverberation times */

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

lib_dec/ivas_binRenderer_internal_fx.c

+4 −0

Original line number	Diff line number	Diff line
		@@ -1380,7 +1380,11 @@ ivas_error ivas_binRenderer_open_fx(
		{
		pRoomAcoustics = &( st_ivas->hRenderConfig->roomAcoustics );
		}
		#ifdef FIX_1139_REV_COLORATION_SHORT_T60
		IF( NE_32( ( error = ivas_binaural_reverb_init( &( hBinRenderer->hReverb ), st_ivas->hHrtfStatistics, hBinRenderer->conv_band, hBinRenderer->timeSlots, pRoomAcoustics, st_ivas->hDecoderConfig->output_Fs, st_ivas->hHrtfFastConv->fastconvReverberationTimes_fx, st_ivas->hHrtfFastConv->fastconvReverberationEneCorrections_fx, NULL ) ), IVAS_ERR_OK ) )
		#else
		IF( NE_32( ( error = ivas_binaural_reverb_init( &( hBinRenderer->hReverb ), st_ivas->hHrtfStatistics, hBinRenderer->conv_band, hBinRenderer->timeSlots, pRoomAcoustics, st_ivas->hDecoderConfig->output_Fs, st_ivas->hHrtfFastConv->fastconvReverberationTimes_fx, st_ivas->hHrtfFastConv->fastconvReverberationEneCorrections_fx ) ), IVAS_ERR_OK ) )
		#endif
		#else
		IF( NE_32( ( error = ivas_binaural_reverb_init( &( hBinRenderer->hReverb ), st_ivas->hHrtfStatistics, hBinRenderer->conv_band, hBinRenderer->timeSlots, &( st_ivas->hRenderConfig->roomAcoustics ), st_ivas->hDecoderConfig->output_Fs, st_ivas->hHrtfFastConv->fastconvReverberationTimes_fx, st_ivas->hHrtfFastConv->fastconvReverberationEneCorrections_fx ) ), IVAS_ERR_OK ) )
		#endif

lib_dec/ivas_dirac_dec_fx.c

+8 −0

Original line number	Diff line number	Diff line
		@@ -4006,7 +4006,15 @@ void ivas_dirac_dec_render_sf_fx(

		/* Move the separated and the LFE channels to temporary variables as spatial synthesis may overwrite current channels */
		Copy32( &( output_buf_fx[st_ivas->hOutSetup.separateChannelIndex][subframe_start_sample] ), tmp_separated_fx, num_samples_subframe );
		#ifdef NONBE_1250_MCMASA_LS_OUTPUT
		if ( hDirACRend->hOutSetup.num_lfe > 0 )
		{
		Copy32( &( output_buf_fx[LFE_CHANNEL][subframe_start_sample] ), tmp_lfe_fx, num_samples_subframe );
		}
		#else
		Copy32( &( output_buf_fx[LFE_CHANNEL][subframe_start_sample] ), tmp_lfe_fx, num_samples_subframe );
		#endif

		FOR( ch = 0; ch < outchannels; ch++ )
		{