Ensure changes are cleanly wrapped in SPLIT_REND_WITH_HEAD_ROT (WIP) Part 2 (b1cea74c) · Commits · IVAS Codec Public Collaboration / IVAS Codec

apps/decoder.c

+31 −16

Original line number	Diff line number	Diff line
		@@ -33,7 +33,7 @@
		#include <stdio.h>
		#include <string.h>
		#include <stdlib.h>
		#include "assert.h"
		#include <assert.h>
		#include "options.h"
		#include "lib_dec.h"
		#include "cmdl_tools.h"
		@@ -74,14 +74,16 @@ static

		#define MIN_NUM_BITS_ACTIVE_FRAME 56
		#define NUM_BITS_SID_IVAS_5K2 104
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		#define NUM_FRAMES_PER_SEC 50
		#define MAX_FRAME_SIZE ( 48000 / NUM_FRAMES_PER_SEC )
		#else
		#define MAX_FRAME_SIZE ( 48000 / 50 )
		#endif
		#define MAX_NUM_OUTPUT_CHANNELS 16
		#define MAX_OUTPUT_PCM_BUFFER_SIZE ( MAX_NUM_OUTPUT_CHANNELS * MAX_FRAME_SIZE )
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		#define MAX_SPLIT_REND_BITRATE ( 1792000 ) /* TODO tmu: unify with SPLIT_REND_MAX_BRATE ?*/
		#endif
		#define MAX_SPLIT_REND_BITS_BUFFER_SIZE_IN_BYTES ( ( ( (int32_t) MAX_SPLIT_REND_BITRATE / NUM_FRAMES_PER_SEC ) + 7 ) >> 3 )
		#endif
		#define IVAS_PUBLIC_ORIENT_TRK_NONE ( 0 )
		@@ -154,6 +156,7 @@ typedef struct
		uint16_t frontendFetchSizeMs;
		#endif
		#endif

		} DecArguments;


		@@ -877,6 +880,7 @@ static IVAS_DEC_AUDIO_CONFIG cmdline2config(
		return output_config;
		}


		/---------------------------------------------------------------------
		* parseCmdlIVAS_dec()
		*
		@@ -1571,7 +1575,6 @@ static ivas_error initOnFirstGoodFrame(
		}
		*pRemainingDelayNumSamples = pFullDelayNumSamples[0];


		if ( ( error = IVAS_DEC_GetNumOutputChannels( hIvasDec, pNumOutChannels ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nError in IVAS_DEC_GetNumOutputChannels, code: %d\n", error );
		@@ -1601,9 +1604,7 @@ static ivas_error initOnFirstGoodFrame(
		*ppAfWriter = NULL;
		}
		else
		#endif
		{
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( arg.outputFormat == IVAS_DEC_OUTPUT_SPLIT_BINAURAL_PCM )
		{
		int16_t delayNumSamples_temp[3];
		@@ -1625,7 +1626,10 @@ static ivas_error initOnFirstGoodFrame(
		fprintf( stderr, "\nUnable to open output file %s\n", arg.outputWavFilename );
		return error;
		}
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		#endif


		int16_t zeroBuf = malloc( pcmFrameSize sizeof( int16_t ) );
		memset( zeroBuf, 0, pcmFrameSize * sizeof( int16_t ) );
		@@ -1646,8 +1650,8 @@ static ivas_error initOnFirstGoodFrame(
		}
		}
		else
		#endif
		{
		#endif
		if ( *pRemainingDelayNumSamples < numOutSamples )
		{
		if ( ( error = AudioFileWriter_write( ppAfWriter, zeroBuf, numOutSamples pNumOutChannels - ( pRemainingDelayNumSamples * *pNumOutChannels ) ) ) != IVAS_ERR_OK )
		@@ -1661,7 +1665,9 @@ static ivas_error initOnFirstGoodFrame(
		{
		*pRemainingDelayNumSamples -= numOutSamples;
		}
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		#endif
		}

		free( zeroBuf );
		@@ -1906,6 +1912,7 @@ static ivas_error decodeG192(
		if ( arg.enableHeadRotation )
		{
		IVAS_QUATERNION Quaternions[IVAS_MAX_PARAM_SPATIAL_SUBFRAMES];

		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( headRotReader == NULL )
		{
		@@ -1918,8 +1925,8 @@ static ivas_error decodeG192(
		}
		}
		else
		#endif
		{
		#endif
		for ( i = 0; i < IVAS_MAX_PARAM_SPATIAL_SUBFRAMES; i++ )
		{
		if ( ( error = HeadRotationFileReading( headRotReader, &Quaternions[i], &Pos[i] ) ) != IVAS_ERR_OK )
		@@ -1929,7 +1936,9 @@ static ivas_error decodeG192(
		goto cleanup;
		}
		}
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		#endif

		if ( ( error = IVAS_DEC_FeedHeadTrackData( hIvasDec, Quaternions, Pos
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		@@ -2037,9 +2046,7 @@ static ivas_error decodeG192(
		}
		}
		else
		#endif
		{
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( ( hSplitRendFileReadWrite != NULL ) && ( arg.outputFormat == IVAS_DEC_OUTPUT_SPLIT_BINAURAL_PCM ) )
		{
		if ( split_rend_write_bitstream_to_file( hSplitRendFileReadWrite, splitRendBits.bits_buf, &splitRendBits.bits_read, &splitRendBits.bits_written,
		@@ -2063,7 +2070,9 @@ static ivas_error decodeG192(
		{
		delayNumSamples -= nOutSamples;
		}
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		#endif
		}

		/* Write ISM metadata to external file(s) */
		@@ -2138,14 +2147,18 @@ static ivas_error decodeG192(

		memset( pcmBuf, 0, delayNumSamples_orig[0] * nOutChannels * sizeof( int16_t ) );

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

		/------------------------------------------------------------------------------------------
		* Printouts after decoding has finished
		@@ -2192,10 +2205,12 @@ cleanup:
		split_rend_reader_writer_close( &hSplitRendFileReadWrite );
		}
		if ( afWriter != NULL )
		#endif
		{
		#endif
		AudioFileWriter_close( &afWriter );
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		#endif
		MasaFileWriter_close( &masaWriter );
		for ( i = 0; i < IVAS_MAX_NUM_OBJECTS; i++ )
		{

apps/renderer.c

+60 −56

Original line number	Diff line number	Diff line
		@@ -932,24 +932,30 @@ int main(
		audioReader = NULL;
		}
		else
		#endif
		{
		#endif
		if ( AudioFileReader_open( &audioReader, audioFilePath ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "Error opening file: %s\n", audioFilePath );
		exit( -1 );
		}
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		#endif

		int32_t inFileSampleRate = 0;
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( audioReader != NULL )
		{
		#endif
		error = AudioFileReader_getSamplingRate( audioReader, &inFileSampleRate );
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		else
		{
		inFileSampleRate = args.sampleRate;
		}
		#endif
		switch ( error )
		{
		case IVAS_ERR_OK:
		@@ -978,23 +984,20 @@ int main(
		}

		int16_t inFileNumChannels = 0;
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( audioReader != NULL )
		{
		#endif
		error = AudioFileReader_getNumChannels( audioReader, &inFileNumChannels );
		if ( error != IVAS_ERR_OK && error != IVAS_ERR_NUM_CHANNELS_UNKNOWN )
		{
		fprintf( stderr, "Error: %s\n", ivas_error_to_string( error ) );
		exit( -1 );
		}
		}
		const int16_t frameSize_smpls = (int16_t) ( 20 * args.sampleRate / 1000 );

		#ifndef SPLIT_REND_WITH_HEAD_ROT
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		const int16_t CLDFBframeSize_smpls = (int16_t) frameSize_smpls * 2;
		cldfb_in = get_cldfb_in_flag( &args );
		#endif
		}
		#endif
		const int16_t frameSize_smpls = (int16_t) ( 20 * args.sampleRate / 1000 );

		IVAS_REND_InputId mcIds[RENDERER_MAX_MC_INPUTS] = { 0 };
		IVAS_REND_InputId ismIds[RENDERER_MAX_ISM_INPUTS] = { 0 };
		@@ -1044,7 +1047,11 @@ int main(

		if ( ( error = IVAS_REND_GetRenderConfig( hIvasRend, &renderConfig ) ) != IVAS_ERR_OK )
		{
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		fprintf( stderr, "\nIVAS_DEC_GetRenderConfig failed: %s\n", ivas_error_to_string( error ) );
		#else
		fprintf( stderr, "\nIVAS_DEC_GetRenderConfig failed\n" );
		#endif
		exit( -1 );
		}

		@@ -1061,7 +1068,11 @@ int main(

		if ( ( error = IVAS_REND_FeedRenderConfig( hIvasRend, renderConfig ) ) != IVAS_ERR_OK )
		{
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		fprintf( stderr, "\nIVAS_REND_FeedRenderConfig failed: %s\n", ivas_error_to_string( error ) );
		#else
		fprintf( stderr, "\nIVAS_DEC_FeedRenderConfig failed\n" );
		#endif
		exit( -1 );
		}
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		@@ -1296,9 +1307,7 @@ int main(
		{
		rend_openCldfb( cldfbAna, cldfbSyn, totalNumInChannels, numOutChannels, args.sampleRate );
		}
		#endif

		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( args.outConfig.audioConfig == IVAS_REND_AUDIO_CONFIG_BINAURAL_SPLIT_CODED )
		{
		IVAS_REND_GetDelay( hIvasRend, &delayNumSamples_temp, &delayTimeScale_temp );
		@@ -1331,20 +1340,8 @@ int main(
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		#endif

		inBufferSize = frameSize_smpls * totalNumInChannels;
		outBufferSize = frameSize_smpls * numOutChannels;
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( is_split_pre_rend_mode( &args ) \|\| is_split_post_rend_mode( &args ) )
		{
		bitsBufferSize = SPLIT_REND_BITS_BUFF_SIZE;
		}
		else
		{
		bitsBufferSize = 0;
		}
		#endif

		inpInt16Buffer = malloc( inBufferSize * sizeof( int16_t ) );
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( cldfb_in == 0 )
		@@ -1361,29 +1358,27 @@ int main(
		outFloatBuffer = malloc( CLDFBframeSize_smpls * totalNumInChannels * sizeof( float ) );
		outBuffer.config.numSamplesPerChannel = (int16_t) CLDFBframeSize_smpls;
		}
		#else
		inBuffer.config.numSamplesPerChannel = (int16_t) frameSize_smpls;
		inFloatBuffer = malloc( inBufferSize * sizeof( float ) );
		outFloatBuffer = malloc( outBufferSize * sizeof( float ) );
		outBuffer.config.numSamplesPerChannel = (int16_t) frameSize_smpls;
		#endif
		outInt16Buffer = malloc( outBufferSize * sizeof( int16_t ) );

		#ifdef SPLIT_REND_WITH_HEAD_ROT
		inBuffer.config.is_cldfb = cldfb_in;
		#endif
		inBuffer.config.numChannels = (int16_t) totalNumInChannels;
		inBuffer.data = inFloatBuffer;

		#ifdef SPLIT_REND_WITH_HEAD_ROT
		outBuffer.config.is_cldfb = cldfb_in;
		#endif
		outBuffer.config.numChannels = (int16_t) numOutChannels;
		outBuffer.data = outFloatBuffer;

		memset( outBuffer.data, 0, outBuffer.config.numSamplesPerChannel * outBuffer.config.numChannels * sizeof( float ) );

		if ( is_split_pre_rend_mode( &args ) \|\| is_split_post_rend_mode( &args ) )
		{
		bitsBufferSize = SPLIT_REND_BITS_BUFF_SIZE;
		}
		else
		{
		bitsBufferSize = 0;
		}

		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( bitsBufferSize > 0 )
		{
		bitsBufferData = malloc( bitsBufferSize * sizeof( uint8_t ) );
		@@ -1393,12 +1388,22 @@ int main(
		bitsBufferData = NULL;
		}

		#endif
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		bitsBuffer.bits = bitsBufferData;
		bitsBuffer.config.bitsRead = 0;
		bitsBuffer.config.bitsWritten = 0;
		bitsBuffer.config.bufLenInBytes = bitsBufferSize;
		#else
		inFloatBuffer = malloc( inBufferSize * sizeof( float ) );
		outInt16Buffer = malloc( outBufferSize * sizeof( int16_t ) );
		outFloatBuffer = malloc( outBufferSize * sizeof( float ) );

		inBuffer.config.numSamplesPerChannel = (int16_t) frameSize_smpls;
		inBuffer.config.numChannels = (int16_t) totalNumInChannels;
		inBuffer.data = inFloatBuffer;

		outBuffer.config.numSamplesPerChannel = (int16_t) frameSize_smpls;
		outBuffer.config.numChannels = (int16_t) numOutChannels;
		outBuffer.data = outFloatBuffer;
		#endif

		#ifdef WMOPS
		@@ -1443,15 +1448,17 @@ int main(
		}
		}
		if ( audioReader != NULL )
		#endif
		{
		#endif
		/* Read the input data */
		if ( ( error = AudioFileReader_read( audioReader, inpInt16Buffer, (int16_t) inBufferSize, &numSamplesRead ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nError reading from file %s\n", audioFilePath );
		exit( -1 );
		}
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		#endif

		if ( numSamplesRead == 0 )
		{
		@@ -1534,7 +1541,6 @@ int main(
		#endif
		);
		if ( error != IVAS_ERR_OK && error != IVAS_ERR_INVALID_OUTPUT_FORMAT )

		{
		fprintf( stderr, "Error setting Head Rotation: %s\n", ivas_error_to_string( error ) );
		exit( -1 );
		@@ -1706,7 +1712,11 @@ int main(
		#endif
		) ) != IVAS_ERR_OK )
		{
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		fprintf( stderr, "Error %s\n", ivas_error_to_string( error ) );
		#else
		fprintf( stderr, "Error in getting samples\n" );
		#endif
		exit( -1 );
		}

		@@ -1764,8 +1774,8 @@ int main(
		}
		}
		if ( audioWriter != NULL )
		#endif
		{
		#endif
		if ( delayNumSamples < outBufferSize )
		{
		if ( AudioFileWriter_write( audioWriter, &outInt16Buffer[delayNumSamples * num_out_channels], outBufferSize - ( delayNumSamples * num_out_channels ) ) != IVAS_ERR_OK )
		@@ -1779,8 +1789,8 @@ int main(
		{
		delayNumSamples -= (int16_t) outBufferSize;
		}
		}
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		bitsBuffer.config.bitsRead = 0;
		bitsBuffer.config.bitsWritten = 0;
		#endif
		@@ -1875,15 +1885,17 @@ int main(
		/* add zeros at the end to have equal length of synthesized signals */
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( audioWriter != NULL )
		#endif
		{
		#endif
		memset( outInt16Buffer, 0, zeroPad * outBuffer.config.numChannels * sizeof( int16_t ) );
		if ( ( error = AudioFileWriter_write( audioWriter, outInt16Buffer, zeroPad * outBuffer.config.numChannels ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\nOutput audio file writer error\n" );
		exit( -1 );
		}
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		#endif

		if ( !args.quietModeEnabled && args.delayCompensationEnabled )
		{
		@@ -1925,25 +1937,13 @@ int main(
		{
		rend_closeCldfb( cldfbAna, cldfbSyn );
		}
		#endif

		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( audioReader != NULL )
		#endif
		{
		AudioFileReader_close( &audioReader );
		}

		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( hSplitRendFileReadWrite != NULL )
		{
		split_rend_reader_writer_close( &hSplitRendFileReadWrite );
		}
		if ( audioWriter != NULL )
		#endif
		{
		AudioFileWriter_close( &audioWriter );
		}

		for ( i = 0; i < RENDERER_MAX_MC_INPUTS; ++i )
		{
		LfeRoutingConfig_close( lfeRoutingConfigs[i] );
		@@ -2021,6 +2021,7 @@ static bool parseInConfig(
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		inConfig->numBinBuses = 0;
		#endif

		/* First check if input is being set to scene description file - this is not covered by parseAudioConfig(). */
		strncpy( charBuf, inFormatStr, sizeof( charBuf ) - 1 );
		charBuf[sizeof( charBuf ) - 1] = '\0';
		@@ -2401,7 +2402,6 @@ static bool checkRequiredArgs(
		args.inConfig.numMultiChannelBuses != 0 \|\|
		args.inConfig.numMasaBuses != 0;
		#endif

		if ( !args.sceneDescriptionInput && !singleInputSpecified )
		{
		/* Neither scene description input nor single-type input was specified on command line */
		@@ -3649,8 +3649,8 @@ static void convertInputBuffer(
		}
		}
		else
		#endif
		{
		#endif
		for ( smpl = 0; smpl < numFloatSamplesPerChannel; ++smpl )
		{
		for ( chnl = 0; chnl < numChannels; ++chnl )
		@@ -3667,7 +3667,9 @@ static void convertInputBuffer(
		++i;
		}
		}
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		#endif

		return;
		}
		@@ -3742,8 +3744,8 @@ static void convertOutputBuffer(
		}
		}
		else
		#endif
		{
		#endif
		for ( smpl = 0; smpl < numSamplesPerChannel; ++smpl )
		{
		for ( chnl = 0; chnl < numChannels; ++chnl )
		@@ -3763,7 +3765,9 @@ static void convertOutputBuffer(
		++i;
		}
		}
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		#endif

		return;
		}

lib_com/common_api_types.h

+2 −4

Original line number	Diff line number	Diff line
		@@ -100,9 +100,7 @@ typedef enum
		YAW_ROLL,
		PITCH_ROLL
		} IVAS_SPLIT_REND_ROT_AXIS;
		#endif

		#ifdef SPLIT_REND_WITH_HEAD_ROT
		typedef enum
		{
		IVAS_SPLIT_REND_POSE_CORRECTION_MODE_NONE,
		@@ -149,7 +147,7 @@ typedef struct ivas_hrtfs_crend_structure *IVAS_DEC_HRTF_CREND_HANDLE;
		typedef struct ivas_hrtfs_fastconv_struct *IVAS_DEC_HRTF_FASTCONV_HANDLE;
		typedef struct ivas_hrtfs_parambin_struct *IVAS_DEC_HRTF_PARAMBIN_HANDLE;

		#if ( defined DEBUGGING ) \|\| ( defined SPLIT_REND_WITH_HEAD_ROT )
		#ifdef DEBUGGING
		typedef enum
		{
		IVAS_RENDER_TYPE_OVERRIDE_NONE,
		@@ -200,7 +198,7 @@ typedef struct _IVAS_SPLIT_REND_CONFIG

		typedef struct _IVAS_RENDER_CONFIG
		{
		#if ( defined DEBUGGING ) \|\| ( defined SPLIT_REND_WITH_HEAD_ROT )
		#ifdef DEBUGGING
		IVAS_RENDER_TYPE_OVERRIDE renderer_type_override;
		#endif
		IVAS_ROOM_ACOUSTICS_CONFIG_DATA room_acoustics;

lib_com/delay_comp.c

+3 −1

Original line number	Diff line number	Diff line
		@@ -104,14 +104,16 @@ int32_t get_delay(

		#ifdef SPLIT_REND_WITH_HEAD_ROT
		if ( output_config != AUDIO_CONFIG_BINAURAL_SPLIT_CODED )
		#endif
		{
		#endif
		if ( hCldfb != NULL )
		{
		/* compensate for filterbank delay */
		delay += IVAS_FB_DEC_DELAY_NS;
		}
		#ifdef SPLIT_REND_WITH_HEAD_ROT
		}
		#endif

		if ( ivas_format == MASA_FORMAT )
		{

lib_com/ivas_cnst.h

+2 −1

Original line number	Diff line number	Diff line
		@@ -117,7 +117,7 @@ typedef enum

		} AUDIO_CONFIG;

		#if ( defined DEBUGGING ) \|\| ( defined SPLIT_REND_WITH_HEAD_ROT )
		#ifdef DEBUGGING
		typedef enum
		{
		RENDER_TYPE_OVERRIDE_NONE,
		@@ -1804,6 +1804,7 @@ typedef enum
		#define SPLIT_REND_ADDITIONAL_BYTES_TO_READ ( 1 )

		#endif

		/----------------------------------------------------------------------------------
		* Limiter constants
		----------------------------------------------------------------------------------/