modify renderer zero padding to flush any delayed input samples from the delay buffers (74c1a873) · Commits · IVAS Codec Public Collaboration / IVAS Codec

apps/renderer.c

+30 −21

Original line number	Diff line number	Diff line
		@@ -698,6 +698,7 @@ 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;
		@@ -1576,10 +1577,12 @@ int main(
		if ( numSamplesRead == 0 )
		{
		/* end of input data */
		break;
		flushRendererLastFrame = true;
		}

		/* Convert from int to float and from interleaved to packed */
		if ( !flushRendererLastFrame )
		{
		convertInputBuffer( inpInt16Buffer,
		numSamplesRead,
		inBuffer.config.numSamplesPerChannel,
		@@ -1587,11 +1590,12 @@ int main(
		inFloatBuffer,
		inBuffer.config.is_cldfb,
		cldfbAna );
		}

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

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

		@@ -1716,7 +1720,7 @@ int main(
		(int16_t) args.inConfig.multiChannelBuses[i].inputChannelIndex,
		numChannels );

		if ( ( error = IVAS_REND_FeedInputAudio( hIvasRend, mcIds[i], tmpBuffer ) ) != IVAS_ERR_OK )
		if ( ( error = IVAS_REND_FeedInputAudio( hIvasRend, mcIds[i], tmpBuffer, flushRendererLastFrame ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nIVAS_REND_FeedInputAudio failed: %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		@@ -1733,7 +1737,7 @@ int main(
		(int16_t) args.inConfig.audioObjects[i].inputChannelIndex,
		args.inConfig.numAudioObjects );

		if ( ( error = IVAS_REND_FeedInputAudio( hIvasRend, ismIds[i], tmpBuffer ) ) != IVAS_ERR_OK )
		if ( ( error = IVAS_REND_FeedInputAudio( hIvasRend, ismIds[i], tmpBuffer, flushRendererLastFrame ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nIVAS_REND_FeedInputAudio failed: %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		@@ -1752,7 +1756,7 @@ int main(
		(int16_t) args.inConfig.audioObjects[i].inputChannelIndex,
		1 );

		if ( ( error = IVAS_REND_FeedInputAudio( hIvasRend, ismIds[i], tmpBuffer ) ) != IVAS_ERR_OK )
		if ( ( error = IVAS_REND_FeedInputAudio( hIvasRend, ismIds[i], tmpBuffer, flushRendererLastFrame ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nIVAS_REND_FeedInputAudio failed: %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		@@ -1777,7 +1781,7 @@ int main(
		(int16_t) args.inConfig.ambisonicsBuses[i].inputChannelIndex,
		numChannels );

		if ( ( error = IVAS_REND_FeedInputAudio( hIvasRend, sbaIds[i], tmpBuffer ) ) != IVAS_ERR_OK )
		if ( ( error = IVAS_REND_FeedInputAudio( hIvasRend, sbaIds[i], tmpBuffer, flushRendererLastFrame ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nIVAS_REND_FeedInputAudio failed: %s\n", ivas_error_to_string( error ) );
		goto cleanup;
		@@ -1795,13 +1799,13 @@ int main(
		(int16_t) args.inConfig.masaBuses[i].inputChannelIndex,
		numChannels );

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

		if ( isCurrentFrameMultipleOf20ms )
		if ( isCurrentFrameMultipleOf20ms && !flushRendererLastFrame )
		{
		if ( masaReaders[i] != NULL )
		{
		@@ -1879,7 +1883,7 @@ int main(
		}
		}

		if ( audioWriter != NULL )
		if ( audioWriter != NULL && !flushRendererLastFrame )
		{
		if ( delayNumSamples * num_out_channels < outBufferSize )
		{
		@@ -1992,6 +1996,13 @@ 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 )
		{
		@@ -2004,14 +2015,13 @@ int main(
		#endif
		}

		/* TODO a flush method is needed to empty the delay buffers */

		/* 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" );
		@@ -2019,7 +2029,6 @@ 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" );

lib_rend/lib_rend.c

+11 −8

Original line number	Diff line number	Diff line
		@@ -1348,7 +1348,8 @@ static ivas_error alignInputDelay(
		const IVAS_REND_ReadOnlyAudioBuffer inputAudio,
		const int32_t maxGlobalDelayNs,
		const int32_t sampleRateOut,
		const int16_t cldfb2tdSampleFact )
		const int16_t cldfb2tdSampleFact,
		const bool flushInputs )
		{
		ivas_error error;
		input_ism *inputIsm;
		@@ -1381,7 +1382,7 @@ static ivas_error alignInputDelay(
		if ( getAudioConfigType( inputBase->inConfig ) == IVAS_REND_AUDIO_CONFIG_TYPE_OBJECT_BASED )
		{
		inputIsm = (input_ism *) inputBase;
		inputIsm->ism_metadata_delay_ms = maxGlobalDelayNs * 1000;
		inputIsm->ism_metadata_delay_ms = maxGlobalDelayNs * 1000.f;
		}
		}

		@@ -1391,7 +1392,7 @@ static ivas_error alignInputDelay(
		inputAudio.config.numSamplesPerChannel );
		TD_RINGBUF_Pop( inputBase->delayBuffer,
		inputBase->inputBuffer.data,
		inputAudio.config.numSamplesPerChannel );
		flushInputs ? TD_RINGBUF_Size( inputBase->delayBuffer ) : inputAudio.config.numSamplesPerChannel );

		return IVAS_ERR_OK;
		}
		@@ -4134,7 +4135,8 @@ ivas_error IVAS_REND_GetDelay(
		ivas_error IVAS_REND_FeedInputAudio(
		IVAS_REND_HANDLE hIvasRend, /* i/o: Renderer handle */
		const IVAS_REND_InputId inputId, /* i : ID of the input */
		const IVAS_REND_ReadOnlyAudioBuffer inputAudio /* i : buffer with input audio */
		const IVAS_REND_ReadOnlyAudioBuffer inputAudio, /* i : buffer with input audio */
		const bool flushInputs /* i : flush input audio */
		)
		{
		ivas_error error;
		@@ -4195,7 +4197,8 @@ ivas_error IVAS_REND_FeedInputAudio(
		inputAudio,
		hIvasRend->maxGlobalDelayNs,
		hIvasRend->sampleRateOut,
		cldfb2tdSampleFact ) ) != IVAS_ERR_OK )
		cldfb2tdSampleFact,
		flushInputs ) ) != IVAS_ERR_OK )
		{
		return error;
		}

lib_rend/lib_rend.h

+16 −15

Original line number	Diff line number	Diff line
		@@ -221,7 +221,8 @@ ivas_error IVAS_REND_GetHrtfStatisticsHandle(
		ivas_error IVAS_REND_FeedInputAudio(
		IVAS_REND_HANDLE hIvasRend, /* i/o: Renderer handle */
		const IVAS_REND_InputId inputId, /* i : ID of the input */
		const IVAS_REND_ReadOnlyAudioBuffer inputAudio /* i : buffer with input audio */
		const IVAS_REND_ReadOnlyAudioBuffer inputAudio, /* i : buffer with input audio */
		const bool flushInputs /* i : flush input audio */
		);

		ivas_error IVAS_REND_FeedInputObjectMetadata(