Merge branch '255-binauralization-delay-report' into 'main' (b0284f1e) · Commits · IVAS Codec Public Collaboration / IVAS Codec

apps/decoder.c

+77 −9

Original line number	Diff line number	Diff line
		@@ -1154,9 +1154,17 @@ static ivas_error initOnFirstGoodFrame(
		}
		else
		{
		#ifdef BINAURALIZATION_DELAY_REPORT
		pFullDelayNumSamples[0] = 0;
		#else
		*pFullDelayNumSamples = 0;
		#endif
		}
		#ifdef BINAURALIZATION_DELAY_REPORT
		*pRemainingDelayNumSamples = pFullDelayNumSamples[0];
		#else
		pRemainingDelayNumSamples = pFullDelayNumSamples;
		#endif

		if ( ( error = IVAS_DEC_GetNumOutputChannels( hIvasDec, pNumOutChannels ) ) != IVAS_ERR_OK )
		{
		@@ -1307,7 +1315,11 @@ static ivas_error decodeG192(
		int16_t numInitialBadFrames = 0; /* Number of bad frames received until first good frame is decoded */
		int16_t nOutChannels = 0;
		int16_t delayNumSamples = -1;
		#ifdef BINAURALIZATION_DELAY_REPORT
		int16_t delayNumSamples_orig[3]; /* stores: overall delay, dec+rend delay, and binauralization delay */
		#else
		int16_t delayNumSamples_orig = 0;
		#endif
		int16_t nOutSamples = 0;
		int32_t delayTimeScale = 0;
		ivas_error error = IVAS_ERR_UNKNOWN;
		@@ -1330,6 +1342,10 @@ static ivas_error decodeG192(
		fprintf( stdout, "\n-- Start the decoder (quiet mode) --\n\n" );
		}

		#ifdef BINAURALIZATION_DELAY_REPORT
		delayNumSamples_orig[0] = -1;
		#endif

		#ifdef WMOPS
		reset_stack();
		reset_wmops();
		@@ -1420,7 +1436,11 @@ static ivas_error decodeG192(
		arg,
		numInitialBadFrames,
		nOutSamples,
		#ifdef BINAURALIZATION_DELAY_REPORT
		delayNumSamples_orig,
		#else
		&delayNumSamples_orig,
		#endif
		&delayNumSamples,
		&delayTimeScale,
		&bsFormat,
		@@ -1521,13 +1541,39 @@ static ivas_error decodeG192(
		#endif
		}

		/------------------------------------------------------------------------------------------
		* Add zeros at the end to have equal length of synthesized signals
		------------------------------------------------------------------------------------------/

		#ifdef BINAURALIZATION_DELAY_REPORT
		memset( pcmBuf, 0, delayNumSamples_orig[0] * nOutChannels * sizeof( int16_t ) );
		if ( ( error = AudioFileWriter_write( afWriter, pcmBuf, delayNumSamples_orig[0] * nOutChannels ) ) != IVAS_ERR_OK )
		#else
		memset( pcmBuf, 0, delayNumSamples_orig * nOutChannels * sizeof( int16_t ) );
		if ( ( error = AudioFileWriter_write( afWriter, pcmBuf, delayNumSamples_orig * nOutChannels ) ) != IVAS_ERR_OK )
		#endif
		{
		fprintf( stderr, "\nError writing output file: %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		}

		/------------------------------------------------------------------------------------------
		* Printouts after decoding has finished
		------------------------------------------------------------------------------------------/

		if ( !arg.quietModeEnabled )
		{
		#ifdef BINAURALIZATION_DELAY_REPORT
		printf( "\n\nDecoder+renderer delay: %4.2f ms (%3u samples at timescale %5u)\n", 1000.f * delayNumSamples_orig[1] / (float) delayTimeScale, delayNumSamples_orig[1], delayTimeScale );

		if ( delayNumSamples_orig[2] > 0 )
		{
		printf( "HRIR/BRIR delay: %4.2f ms (%3u samples at timescale %5u)\n", 1000.f * delayNumSamples_orig[2] / (float) delayTimeScale, delayNumSamples_orig[2], delayTimeScale );
		printf( "Total delay: %4.2f ms (%3u samples at timescale %5u)\n", 1000.f * delayNumSamples_orig[0] / (float) delayTimeScale, delayNumSamples_orig[0], delayTimeScale );
		}
		#else
		fprintf( stdout, "\nDecoder delay: %-5u [samples] - Timescale: %5u\n", delayNumSamples_orig, delayTimeScale );
		#endif
		}

		/* Print output metadata file name(s) */
		@@ -1547,14 +1593,6 @@ static ivas_error decodeG192(
		}
		}

		/* add zeros at the end to have equal length of synthesized signals */
		memset( pcmBuf, 0, delayNumSamples_orig * nOutChannels * sizeof( int16_t ) );
		if ( ( error = AudioFileWriter_write( afWriter, pcmBuf, delayNumSamples_orig * nOutChannels ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nError writing output file: %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		}

		/------------------------------------------------------------------------------------------
		* Close files and deallocate resources
		------------------------------------------------------------------------------------------/
		@@ -1724,7 +1762,11 @@ static ivas_error decodeVoIP(
		#endif
		JbmOffsetFileWriter *jbmOffsetWriter = NULL;

		#ifdef BINAURALIZATION_DELAY_REPORT
		int16_t delayNumSamples_orig[3]; /* stores: overall delay, dec+rend delay, and binauralization delay */
		#else
		int16_t delayNumSamples_orig = -1;
		#endif
		int16_t delayNumSamples = -1;
		int32_t delayTimeScale = -1;

		@@ -1738,11 +1780,15 @@ static ivas_error decodeVoIP(

		IVAS_DEC_BS_FORMAT bsFormat = IVAS_DEC_BS_UNKOWN;
		IsmFileWriter *ismWriters[IVAS_MAX_NUM_OBJECTS];

		for ( int16_t i = 0; i < IVAS_MAX_NUM_OBJECTS; ++i )
		{
		ismWriters[i] = NULL;
		}

		#ifdef BINAURALIZATION_DELAY_REPORT
		delayNumSamples_orig[0] = -1;
		#endif

		rtpdumpDepacker.rtpdump = NULL;
		switch ( arg.inputFormat )
		@@ -1939,7 +1985,11 @@ static ivas_error decodeVoIP(
		arg,
		numInitialBadFrames,
		nOutSamples,
		#ifdef BINAURALIZATION_DELAY_REPORT
		delayNumSamples_orig,
		#else
		&delayNumSamples_orig,
		#endif
		&delayNumSamples,
		&delayTimeScale,
		&bsFormat,
		@@ -1992,9 +2042,17 @@ static ivas_error decodeVoIP(
		#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
		------------------------------------------------------------------------------------------/

		#ifdef BINAURALIZATION_DELAY_REPORT
		memset( pcmBuf, 0, delayNumSamples_orig[0] * nOutChannels * sizeof( int16_t ) );
		if ( ( error = AudioFileWriter_write( afWriter, pcmBuf, delayNumSamples_orig[0] * nOutChannels ) ) != IVAS_ERR_OK )
		#else
		memset( pcmBuf, 0, delayNumSamples_orig * nOutChannels * sizeof( int16_t ) );
		if ( ( error = AudioFileWriter_write( afWriter, pcmBuf, delayNumSamples_orig * nOutChannels ) ) != IVAS_ERR_OK )
		#endif
		{
		fprintf( stderr, "\nError writing output file: %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		@@ -2006,7 +2064,17 @@ static ivas_error decodeVoIP(

		if ( !arg.quietModeEnabled )
		{
		#ifdef BINAURALIZATION_DELAY_REPORT
		printf( "\n\nDecoder+renderer delay: %4.2f ms (%3u samples at timescale %5u)\n", 1000.f * delayNumSamples_orig[1] / (float) delayTimeScale, delayNumSamples_orig[1], delayTimeScale );

		if ( delayNumSamples_orig[2] > 0 )
		{
		printf( "HRIR/BRIR delay: %4.2f ms (%3u samples at timescale %5u)\n", 1000.f * delayNumSamples_orig[2] / (float) delayTimeScale, delayNumSamples_orig[2], delayTimeScale );
		printf( "Total delay: %4.2f ms (%3u samples at timescale %5u)\n", 1000.f * delayNumSamples_orig[0] / (float) delayTimeScale, delayNumSamples_orig[0], delayTimeScale );
		}
		#else
		printf( "\nDecoder delay: %5u [samples] - Timescale: %5u\n", delayNumSamples_orig, delayTimeScale );
		#endif
		}

		/------------------------------------------------------------------------------------------

lib_com/delay_comp.c

+9 −4

Original line number	Diff line number	Diff line
		@@ -54,8 +54,11 @@ int32_t get_delay(
		const int16_t enc_dec, /* i : encoder/decoder flag */
		const int32_t io_fs, /* i : input/output sampling frequency */
		const IVAS_FORMAT ivas_format, /* i : IVAS format */
		HANDLE_CLDFB_FILTER_BANK hCldfb, /* i : Handle of Cldfb analysis */
		HANDLE_CLDFB_FILTER_BANK hCldfb /* i : Handle of Cldfb analysis */
		#ifndef BINAURALIZATION_DELAY_REPORT
		,
		const int32_t binaural_latency_ns /* i : binaural renderer HRTF delay in ns */
		#endif
		)
		{
		int32_t delay = 0;
		@@ -100,8 +103,10 @@ int32_t get_delay(
		delay += IVAS_FB_DEC_DELAY_NS;
		}

		#ifndef BINAURALIZATION_DELAY_REPORT
		/* compensate for binauralization delay */
		delay += binaural_latency_ns;
		#endif
		}
		}

lib_com/options.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -160,6 +160,7 @@
		#define SMOOTH_WITH_TRANS_DET
		#endif

		#define BINAURALIZATION_DELAY_REPORT /* VA: Issue 255 - Changes the way the decoder delay is reported */


		/* ################## End DEVELOPMENT switches ######################### */

lib_com/prot.h

+7 −4

Original line number	Diff line number	Diff line
		@@ -680,8 +680,11 @@ int32_t get_delay(
		const int16_t enc_dec, /* i : encoder/decoder flag */
		const int32_t io_fs, /* i : input/output sampling frequency */
		const IVAS_FORMAT ivas_format, /* i : IVAS format */
		HANDLE_CLDFB_FILTER_BANK hCldfb, /* i : Handle of Cldfb analysis */
		HANDLE_CLDFB_FILTER_BANK hCldfb /* i : Handle of Cldfb analysis */
		#ifndef BINAURALIZATION_DELAY_REPORT
		,
		const int32_t binaural_latency_ns /* i : binauralization delay in ns */
		#endif
		);

		void decision_matrix_enc(

lib_dec/lib_dec.c

+6 −0

Original line number	Diff line number	Diff line
		@@ -1183,7 +1183,13 @@ ivas_error IVAS_DEC_GetDelay(
		st_ivas = hIvasDec->st_ivas;
		hDecoderConfig = st_ivas->hDecoderConfig;

		#ifdef BINAURALIZATION_DELAY_REPORT
		nSamples[1] = NS2SA( hDecoderConfig->output_Fs, get_delay( DEC, hDecoderConfig->output_Fs, st_ivas->ivas_format, st_ivas->cldfbAnaDec[0] ) );
		nSamples[2] = (int16_t) roundf( (float) st_ivas->binaural_latency_ns * hDecoderConfig->output_Fs / 1000000000.f );
		nSamples[0] = nSamples[1] + nSamples[2];
		#else
		nSamples = (int16_t) roundf( (float) get_delay( DEC, hDecoderConfig->output_Fs, st_ivas->ivas_format, st_ivas->cldfbAnaDec[0], st_ivas->binaural_latency_ns ) hDecoderConfig->output_Fs / 1000000000.f );
		#endif

		*timeScale = hDecoderConfig->output_Fs;