Merge branch 'main' into 2142-implement-basop-w_min-and-w_max-functions (4048efa5) · Commits · SA4 / Audio / IVAS BASOP

apps/renderer.c

+70 −56

Original line number	Diff line number	Diff line
		@@ -31,6 +31,7 @@
		*******************************************************************************************************/

		#include "lib_rend.h"
		#include "typedef.h"
		#include <assert.h>
		#include <math.h>
		#include <string.h>
		@@ -195,7 +196,7 @@ typedef struct
		float lfeConfigElevation;
		bool lfeCustomRoutingEnabled;
		char inLfePanningMatrixFile[RENDERER_MAX_CLI_ARG_LENGTH];
		float syncMdDelay;
		int16_t syncMdDelay;
		IVAS_RENDER_FRAMESIZE render_framesize;
		uint16_t directivityPatternId[RENDERER_MAX_ISM_INPUTS];
		AcousticEnvironmentSequence aeSequence;
		@@ -802,6 +803,7 @@ int main(
		SplitFileReadWrite *hSplitRendFileReadWrite;
		int16_t delayNumSamples_temp;
		int32_t delayTimeScale_temp;
		bool flushRendererLastFrame = false;
		int16_t numSamplesRead;
		int16_t delayNumSamples = -1;
		int16_t delayNumSamples_orig = 0;
		@@ -847,8 +849,11 @@ int main(

		CmdlnArgs args = parseCmdlnArgs( argc, argv );

		if ( args.nonDiegeticPan && !( ( args.inConfig.numAudioObjects == 0 && args.inConfig.multiChannelBuses[0].audioConfig == IVAS_AUDIO_CONFIG_MONO ) \|\|
		( args.inConfig.numAudioObjects > 0 && args.inConfig.audioObjects[0].audioConfig == IVAS_AUDIO_CONFIG_OBA && args.inConfig.numAudioObjects == 1 ) ) )
		if ( args.nonDiegeticPan &&
		!( ( args.inConfig.numAudioObjects == 0 &&
		args.inConfig.multiChannelBuses[0].audioConfig == IVAS_AUDIO_CONFIG_MONO ) \|\|
		( args.inConfig.numAudioObjects > 0 &&
		args.inConfig.audioObjects[0].audioConfig == IVAS_AUDIO_CONFIG_OBA && args.inConfig.numAudioObjects == 1 ) ) )
		{
		fprintf( stderr, "\ninvalid configuration - non-diegetic panning requires mono or ISM1 input\n" );
		goto cleanup;
		@@ -1178,7 +1183,10 @@ int main(
		#endif

		/* sanity check */
		if ( ( args.outConfig.audioConfig != IVAS_AUDIO_CONFIG_BINAURAL ) && ( args.outConfig.audioConfig != IVAS_AUDIO_CONFIG_BINAURAL_ROOM_IR ) && ( args.outConfig.audioConfig != IVAS_AUDIO_CONFIG_BINAURAL_ROOM_REVERB ) && !is_split_pre_rend_mode( &args ) )
		if ( ( args.outConfig.audioConfig != IVAS_AUDIO_CONFIG_BINAURAL ) &&
		( args.outConfig.audioConfig != IVAS_AUDIO_CONFIG_BINAURAL_ROOM_IR ) &&
		( args.outConfig.audioConfig != IVAS_AUDIO_CONFIG_BINAURAL_ROOM_REVERB ) &&
		!is_split_pre_rend_mode( &args ) )
		{
		fprintf( stderr, "\nExternal Renderer Config is supported only when binaural output configurations is used as output OR when Split pre-rendering mode is enabled. Exiting. \n" );
		goto cleanup;
		@@ -1284,8 +1292,7 @@ int main(
		#ifdef NONBE_1359_FIX_IVASREND_OMASA_BINAURAL_LOUDNESS
		/* Set the metadata delay for objects */
		#endif
		Word32 var1 = (Word32) ( args.syncMdDelay );
		IF( ( error = IVAS_REND_SetIsmMetadataDelay( hIvasRend, var1 ) ) != IVAS_ERR_OK )
		IF( ( error = IVAS_REND_SetIsmMetadataDelay( hIvasRend, args.syncMdDelay ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nError in IVAS_REND_SetIsmMetadataDelay(): %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		@@ -1516,7 +1523,7 @@ int main(
		}

		Word16 numOutChannels;
		if ( ( error = IVAS_REND_NumOutChannels( hIvasRend, &numOutChannels ) ) != IVAS_ERR_OK )
		if ( ( error = IVAS_REND_GetNumOutChannels( hIvasRend, &numOutChannels ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nError in IVAS_REND_NumOutChannels(): %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		@@ -1531,31 +1538,16 @@ int main(
		}
		}

		if ( args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED )
		{
		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 )
		{
		fprintf( stderr, "\nError in IVAS_REND_GetSplitRendBitstreamHeader(): %s!\n", ivas_error_to_string( error ) );
		goto cleanup;
		}

		if ( IVAS_REND_GetDelay_fx( hIvasRend, &delayNumSamples_temp, &delayTimeScale_temp ) != IVAS_ERR_OK )
		if ( args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED \|\| args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM )
		{
		fprintf( stderr, "\nUnable to get delay of renderer!\n" );
		goto cleanup;
		}
		char *outFile = args.outMetadataFilePath;

		if ( ( error = split_rend_writer_open( &hSplitRendFileReadWrite, args.outputFilePath, delayNumSamples_temp, delayTimeScale_temp, bitsBuffer.config.codec, bitsBuffer.config.poseCorrection, bitsBuffer.config.codec_frame_size_ms, bitsBuffer.config.isar_frame_size_ms, args.sampleRate, bitsBuffer.config.lc3plus_highres ) ) != IVAS_ERR_OK )
		if ( args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED )
		{
		fprintf( stderr, "\nCould not open split rend metadata file %s\n", args.outputFilePath );
		goto cleanup;
		}
		outFile = args.outputFilePath;
		audioWriter = NULL;
		}
		else
		{
		if ( args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM )
		{

		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 )
		{
		fprintf( stderr, "\nError in IVAS_REND_GetSplitRendBitstreamHeader(): %s!\n", ivas_error_to_string( error ) );
		@@ -1568,13 +1560,15 @@ int main(
		goto cleanup;
		}

		if ( ( error = split_rend_writer_open( &hSplitRendFileReadWrite, args.outMetadataFilePath, delayNumSamples_temp, delayTimeScale_temp, bitsBuffer.config.codec, bitsBuffer.config.poseCorrection, bitsBuffer.config.codec_frame_size_ms, bitsBuffer.config.isar_frame_size_ms, args.sampleRate, bitsBuffer.config.lc3plus_highres ) ) != IVAS_ERR_OK )
		if ( ( error = split_rend_writer_open( &hSplitRendFileReadWrite, outFile, delayNumSamples_temp, delayTimeScale_temp, bitsBuffer.config.codec, bitsBuffer.config.poseCorrection, bitsBuffer.config.codec_frame_size_ms, bitsBuffer.config.isar_frame_size_ms, args.sampleRate, bitsBuffer.config.lc3plus_highres ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nCould not open split rend metadata file %s\n", args.outMetadataFilePath );
		fprintf( stderr, "\nCould not open split rend metadata file %s\n", outFile );
		goto cleanup;
		}
		}

		if ( args.outConfig.audioConfig != IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED )
		{
		if ( AudioFileWriter_open( &audioWriter, args.outputFilePath, args.sampleRate, numOutChannels ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nFailed to open file: %s\n", args.outputFilePath );
		@@ -1667,6 +1661,7 @@ int main(
		outBuffer.pq_fact = &outBuffer.q_factor;
		Word16 subframe_len = (Word16) ( args.sampleRate / ( 200 ) ); // sample rate /FRAMES_PER_SEC * MAX_PARAM_SPATIAL_SUBFRAMES
		Word16 gd_bits = find_guard_bits( subframe_len );
		Word16 prev_q_fact = Q11;
		while ( 1 )
		{
		int16_t num_in_channels;
		@@ -1719,19 +1714,28 @@ int main(
		if ( numSamplesRead == 0 )
		{
		/* end of input data */
		break;
		flushRendererLastFrame = true;
		}

		/* Convert from int to float and from interleaved to packed */
		if ( !flushRendererLastFrame )
		{
		Word16 Q_out;
		*outBuffer.pq_fact = 16 - ( gd_bits );
		convertInputBuffer_fx( inpInt16Buffer, numSamplesRead, inBuffer.config.numSamplesPerChannel, num_in_channels, inInt32Buffer, *outBuffer.pq_fact, inBuffer.config.is_cldfb, cldfbAna, &Q_out );
		*outBuffer.pq_fact = Q_out;
		prev_q_fact = Q_out;
		}
		else
		{
		memset( inBuffer.data_fx, 0, inBuffer.config.numChannels * inBuffer.config.numSamplesPerChannel * sizeof( Word32 ) );
		*outBuffer.pq_fact = prev_q_fact;
		}

		int16_t num_subframes, sf_idx;
		num_subframes = (int16_t) args.render_framesize;

		if ( isCurrentFrameMultipleOf20ms )
		if ( isCurrentFrameMultipleOf20ms && !flushRendererLastFrame )
		{
		IsmPositionProvider_getNextFrame( positionProvider, &mtdBuffer );

		@@ -1843,7 +1847,7 @@ int main(
		}

		IVAS_REND_ReadOnlyAudioBuffer tmpBuffer = getReadOnlySubBuffer( inBuffer, (int16_t) args.inConfig.multiChannelBuses[i].inputChannelIndex, numChannels );
		if ( ( error = IVAS_REND_FeedInputAudio_fx( hIvasRend, mcIds[i], tmpBuffer ) ) != IVAS_ERR_OK )
		if ( ( error = IVAS_REND_FeedInputAudio_fx( hIvasRend, mcIds[i], tmpBuffer, flushRendererLastFrame ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nIVAS_REND_FeedInputAudio failed: %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		@@ -1857,7 +1861,7 @@ int main(
		if ( i == 0 )
		{
		IVAS_REND_ReadOnlyAudioBuffer tmpBuffer = getReadOnlySubBuffer( inBuffer, (int16_t) args.inConfig.audioObjects[i].inputChannelIndex, args.inConfig.numAudioObjects );
		if ( ( error = IVAS_REND_FeedInputAudio_fx( hIvasRend, ismIds[i], tmpBuffer ) ) != IVAS_ERR_OK )
		if ( ( error = IVAS_REND_FeedInputAudio_fx( hIvasRend, ismIds[i], tmpBuffer, flushRendererLastFrame ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nIVAS_REND_FeedInputAudio failed: %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		@@ -1873,7 +1877,7 @@ int main(
		else
		{
		IVAS_REND_ReadOnlyAudioBuffer tmpBuffer = getReadOnlySubBuffer( inBuffer, (int16_t) args.inConfig.audioObjects[i].inputChannelIndex, 1 );
		if ( ( error = IVAS_REND_FeedInputAudio_fx( hIvasRend, ismIds[i], tmpBuffer ) ) != IVAS_ERR_OK )
		if ( ( error = IVAS_REND_FeedInputAudio_fx( hIvasRend, ismIds[i], tmpBuffer, flushRendererLastFrame ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nIVAS_REND_FeedInputAudio failed: %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		@@ -1896,7 +1900,7 @@ int main(
		}

		IVAS_REND_ReadOnlyAudioBuffer tmpBuffer = getReadOnlySubBuffer( inBuffer, (int16_t) args.inConfig.ambisonicsBuses[i].inputChannelIndex, numChannels );
		if ( ( error = IVAS_REND_FeedInputAudio_fx( hIvasRend, sbaIds[i], tmpBuffer ) ) != IVAS_ERR_OK )
		if ( ( error = IVAS_REND_FeedInputAudio_fx( hIvasRend, sbaIds[i], tmpBuffer, flushRendererLastFrame ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nIVAS_REND_FeedInputAudio failed: %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		@@ -1913,13 +1917,13 @@ int main(

		IVAS_REND_ReadOnlyAudioBuffer tmpBuffer = getReadOnlySubBuffer( inBuffer, (int16_t) args.inConfig.masaBuses[i].inputChannelIndex, numChannels );

		IF( ( error = IVAS_REND_FeedInputAudio_fx( hIvasRend, masaIds[i], tmpBuffer ) ) != IVAS_ERR_OK )
		IF( ( error = IVAS_REND_FeedInputAudio_fx( hIvasRend, masaIds[i], tmpBuffer, flushRendererLastFrame ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nIVAS_REND_FeedInputAudio failed: %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		}

		if ( isCurrentFrameMultipleOf20ms )
		if ( isCurrentFrameMultipleOf20ms && !flushRendererLastFrame )
		{
		if ( masaReaders[i] != NULL )
		{
		@@ -1982,7 +1986,7 @@ int main(
		zeroPad = delayNumSamples;
		}

		if ( is_split_pre_rend_mode( &args ) )
		if ( is_split_pre_rend_mode( &args ) && !flushRendererLastFrame )
		{
		if ( split_rend_write_bitstream_to_file( hSplitRendFileReadWrite, bitsBuffer.bits, &bitsBuffer.config.bitsRead,
		&bitsBuffer.config.bitsWritten ) != IVAS_ERR_OK )
		@@ -1992,7 +1996,7 @@ int main(
		}
		}

		if ( audioWriter != NULL )
		if ( audioWriter != NULL && !flushRendererLastFrame )
		{
		if ( delayNumSamples * num_out_channels < outBufferSize )
		{
		@@ -2013,7 +2017,7 @@ int main(
		bitsBuffer.config.bitsWritten = 0;

		/* Write MASA metadata for MASA outputs */
		if ( args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_MASA1 \|\| args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_MASA2 )
		if ( !flushRendererLastFrame && ( args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_MASA1 \|\| args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_MASA2 ) )
		{
		IVAS_REND_AudioConfigType inputType1;
		IVAS_REND_AudioConfigType inputType2;
		@@ -2085,7 +2089,8 @@ int main(
		}
		}

		if ( ( args.inConfig.numAmbisonicsBuses > 0 \|\| args.inConfig.numMultiChannelBuses > 0 \|\| args.inConfig.numMasaBuses > 0 ) && args.inConfig.numAudioObjects > 0 )
		if ( ( args.inConfig.numAmbisonicsBuses > 0 \|\| args.inConfig.numMultiChannelBuses > 0 \|\| args.inConfig.numMasaBuses > 0 ) &&
		args.inConfig.numAudioObjects > 0 )
		{
		inputType2 = IVAS_REND_AUDIO_CONFIG_TYPE_OBJECT_BASED;
		if ( ( error = IVAS_REND_MergeMasaMetadata( hIvasRend, &hMetaOutput, inputType1, inputType2 ) ) != IVAS_ERR_OK )
		@@ -2101,6 +2106,13 @@ int main(
		}
		}

		/* no new input was actually read, only delay buffers were flushed
		* therefore this is not a real frame */
		if ( flushRendererLastFrame )
		{
		break;
		}

		frame++;
		if ( !args.quietModeEnabled )
		{
		@@ -2113,12 +2125,13 @@ int main(
		#endif
		}

		/* add zeros at the end to have equal length of synthesized signals */
		/* add zeros at the end to have equal length of synthesized signals
		* the output buffer will contain either leftover input samples from delay aligned inputs
		* or zeros for padding */
		if ( audioWriter != NULL )
		{
		for ( zeroPadToWrite = zeroPad; zeroPadToWrite > frameSize_smpls; zeroPadToWrite -= frameSize_smpls )
		{
		memset( outInt16Buffer, 0, outBufferSize * sizeof( int16_t ) );
		if ( ( error = AudioFileWriter_write( audioWriter, outInt16Buffer, outBufferSize ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nOutput audio file writer error\n" );
		@@ -2126,7 +2139,6 @@ int main(
		}
		}

		memset( outInt16Buffer, 0, zeroPadToWrite * outBuffer.config.numChannels * sizeof( int16_t ) );
		if ( ( error = AudioFileWriter_write( audioWriter, outInt16Buffer, zeroPadToWrite * outBuffer.config.numChannels ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nOutput audio file writer error\n" );
		@@ -2136,9 +2148,10 @@ int main(
		}


		if ( args.inConfig.numAudioObjects != 0 && ( args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_BINAURAL \|\| args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_BINAURAL_ROOM_REVERB ) )
		if ( args.inConfig.numAudioObjects != 0 &&
		( args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_BINAURAL \|\| args.outConfig.audioConfig == IVAS_AUDIO_CONFIG_BINAURAL_ROOM_REVERB ) )
		{
		fprintf( stdout, "\n\nMetadata delayed %d subframes\n\n", (int16_t) round( args.syncMdDelay / ( 1000 / IVAS_NUM_FRAMES_PER_SEC / IVAS_MAX_PARAM_SPATIAL_SUBFRAMES ) ) );
		fprintf( stdout, "\n\nMetadata delayed %d subframes\n\n", (int16_t) round( args.syncMdDelay / ( (float) BINAURAL_RENDERING_FRAME_SIZE_MS ) ) );
		}

		if ( !args.quietModeEnabled && args.delayCompensationEnabled )
		@@ -3038,7 +3051,7 @@ static void parseOption(
		case CmdLnOptionId_syncMdDelay:
		assert( numOptionValues == 1 );
		/* Metadata Delay to sync with audio delay in ms */
		args->syncMdDelay = strtof( optionValues[0], NULL );
		args->syncMdDelay = (int16_t) strtol( optionValues[0], NULL, 10 );
		break;
		default:
		assert( 0 && "This should be unreachable - all command line options should be explicitly handled." );
		@@ -3574,7 +3587,8 @@ static void parseObjectPosition(
		*positionDuration = (uint16_t) strtol( line, &endptr, 10 );
		readNextMetadataChunk( line, "\n" );

		read_values = (int16_t) sscanf( line, "%f,%f,%f,%f,%f,%f,%f,%f", &meta_prm[0], &meta_prm[1], &meta_prm[2], &meta_prm[3], &meta_prm[4], &meta_prm[5], &meta_prm[6], &meta_prm[7] );
		read_values = (int16_t) sscanf( line, "%f,%f,%f,%f,%f,%f,%f,%f",
		&meta_prm[0], &meta_prm[1], &meta_prm[2], &meta_prm[3], &meta_prm[4], &meta_prm[5], &meta_prm[6], &meta_prm[7] );

		if ( read_values < 2 )
		{

lib_com/bitstream_fx.c

+2 −1

Original line number	Diff line number	Diff line
		@@ -5235,7 +5235,7 @@ void evs_dec_previewFrame(
		return;
		}


		#ifndef NONBE_MDCT_ST_DTX_FIX_SUBOPT_SPATIAL_CNG
		void dtx_read_padding_bits_fx(
		DEC_CORE_HANDLE st,
		const Word16 num_bits )
		@@ -5253,3 +5253,4 @@ void dtx_read_padding_bits_fx(

		return;
		}
		#endif

lib_com/fd_cng_com_fx.c

+5 −0

Original line number	Diff line number	Diff line
		@@ -2,6 +2,7 @@
		EVS Codec 3GPP TS26.452 Aug 12, 2021. Version 16.3.0
		====================================================================================*/

		#include "ivas_cnst.h"
		#include <assert.h>
		#include <stdint.h>
		#include "options.h"
		@@ -161,8 +162,12 @@ void ivas_initFdCngCom_fx( HANDLE_FD_CNG_COM hFdCngCom, Word16 scale )
		hFdCngCom->exp_cldfb_periodog = 0;
		move16();

		#ifdef NONBE_MDCT_ST_DTX_FIX_SUBOPT_SPATIAL_CNG
		set16_fx( hFdCngCom->coherence_fx, 16384 /* 0.5 in Q15 */, MDCT_ST_DTX_NUM_COHERENCE_BANDS );
		#else
		hFdCngCom->coherence_fx = 16384; /* 0.5 in Q15 */
		move16();
		#endif

		set32_fx( hFdCngCom->olapBufferSynth_fx, 0, FFTLEN );
		set32_fx( hFdCngCom->olapBufferSynth2_fx, 0, FFTLEN );

lib_com/hp50_fx.c

+5 −1

Original line number	Diff line number	Diff line
		@@ -316,8 +316,12 @@ void hp20_fx_32_opt(
		prescale = s_min( prescale, diff );

		prescale = sub( 1 + HP20_FX_COEFF_SCALE, prescale );

		#ifdef FIX_2086_ENABLE_HP20_OPT_FOR_ENC
		if ( EQ_16( prescale_current_frame, 1 + HP20_FX_COEFF_SCALE - 31 ) \|\| // signal_fx buffer contains only zeros, so use the mem_fx scale_factor instead
		LT_16( prescale_current_frame, prescale ) ) // To avoid overflow in the subsequent shr() scaling for W_y1, W_y2, x2, and x1 calculations before the for loop.
		#else
		if ( EQ_16( prescale_current_frame, 1 + HP20_FX_COEFF_SCALE - 31 ) ) // signal_fx buffer contains only zeros, so use the mem_fx scale_factor instead
		#endif
		{
		prescale_current_frame = prescale;
		}

lib_com/ivas_cnst.h

+5 −0

Original line number	Diff line number	Diff line
		@@ -889,6 +889,11 @@ enum fea_names
		#define MDCT_ST_PLC_FADEOUT_TO_ZERO_LEN 20
		#define MDCT_ST_PLC_FADEOUT_DELAY_4_LSP_FADE 3

		#ifdef NONBE_MDCT_ST_DTX_FIX_SUBOPT_SPATIAL_CNG
		#define MDCT_ST_DTX_NUM_COHERENCE_BANDS 5
		#define MDCT_ST_DTX_FIRST_BAND_OFFSET 2
		#endif

		typedef enum {
		NOISE_GEN_MODE_UNDEF = -1,
		EQUAL_CORES = 0,