Merge branch 'ivas_dirac_dec_fxd' into 'main' (526ae840) · Commits · SA4 / Audio / IVAS BASOP

lib_com/cnst.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -732,6 +732,7 @@ enum
		#define CLDFB_OVRLP_MIN_SLOTS 3 /* CLDFB resampling - minimize processing to minimum required for transition frame ACELP->TCX/HQ */
		#define INV_CLDFB_BANDWIDTH ( 1.f / 800.f )
		#define INV_CLDFB_BANDWIDTH_Q31 ( 2684355l )
		#define CLDFB_BANDWIDTH 800

		#define L_FILT_2OVER3 12
		#define L_FILT_2OVER3_LP 3

lib_com/ivas_dirac_com.c

+313 −17

Original line number	Diff line number	Diff line
		@@ -42,6 +42,7 @@
		#include "wmc_auto.h"
		#include "prot_fx1.h"
		#include "prot_fx2.h"
		#include "ivas_prot_fx.h"

		/-----------------------------------------------------------------------
		* Local function prototypes
		@@ -97,6 +98,135 @@ Word16 ivas_get_hodirac_flag_fx(
		*
		* DirAC Configuration function; used also in MASA decoder
		------------------------------------------------------------------------/
		#ifdef IVAS_FLOAT_FIXED
		ivas_error ivas_dirac_config_fx(
		void st_ivas, / i/o: IVAS encoder/decoder state structure */
		const Word16 enc_dec /* i : encoder or decoder flag */
		)
		{
		IVAS_FORMAT ivas_format;
		Word16 sba_order;
		Word16 *element_mode;
		Word32 ivas_total_brate;
		DIRAC_CONFIG_DATA_HANDLE hConfig;
		IVAS_QMETADATA_HANDLE hQMetaData;
		Word32 Fs;
		Word16 *band_grouping;
		ivas_error error;
		Word16 spar_dirac_split_band;
		IVAS_FB_MIXER_HANDLE hFbMdft;
		Word16 *dirac_to_spar_md_bands;

		error = IVAS_ERR_OK;

		IF ( EQ_16(enc_dec , ENC))
		{
		ivas_format = ((Encoder_Struct *)st_ivas)->hEncoderConfig->ivas_format;
		element_mode = &((Encoder_Struct *)st_ivas)->hEncoderConfig->element_mode_init;
		sba_order = ((Encoder_Struct *)st_ivas)->sba_analysis_order;
		ivas_total_brate = ((Encoder_Struct *)st_ivas)->hEncoderConfig->ivas_total_brate;
		Fs = ((Encoder_Struct *)st_ivas)->hEncoderConfig->input_Fs;
		band_grouping = ((Encoder_Struct *)st_ivas)->hDirAC->band_grouping;
		hConfig = ((Encoder_Struct *)st_ivas)->hDirAC->hConfig;
		hQMetaData = ((Encoder_Struct *)st_ivas)->hQMetaData;
		IF (((Encoder_Struct *)st_ivas)->hSpar != NULL)
		{
		hFbMdft = ((Encoder_Struct *)st_ivas)->hSpar->hFbMixer;
		dirac_to_spar_md_bands = ((Encoder_Struct *)st_ivas)->hSpar->dirac_to_spar_md_bands;
		}
		ELSE
		{
		hFbMdft = NULL;
		dirac_to_spar_md_bands = NULL;
		}
		}
		ELSE
		{
		ivas_format = ((Decoder_Struct *)st_ivas)->ivas_format;
		element_mode = &((Decoder_Struct *)st_ivas)->element_mode_init;
		sba_order = ((Decoder_Struct *)st_ivas)->sba_analysis_order;
		ivas_total_brate = ((Decoder_Struct *)st_ivas)->hDecoderConfig->ivas_total_brate;
		Fs = ((Decoder_Struct *)st_ivas)->hDecoderConfig->output_Fs;
		band_grouping = ((Decoder_Struct *)st_ivas)->hDirAC->band_grouping;
		hConfig = ((Decoder_Struct *)st_ivas)->hDirAC->hConfig;
		hQMetaData = ((Decoder_Struct *)st_ivas)->hQMetaData;
		IF (((Decoder_Struct *)st_ivas)->hSpar != NULL)
		{
		hFbMdft = ((Decoder_Struct *)st_ivas)->hSpar->hFbMixer;
		dirac_to_spar_md_bands = ((Decoder_Struct *)st_ivas)->hSpar->dirac_to_spar_md_bands;
		}
		ELSE
		{
		hFbMdft = NULL;
		dirac_to_spar_md_bands = NULL;
		}
		((Decoder_Struct *)st_ivas)->hDirAC->hFbMdft = hFbMdft;
		}

		IF (ivas_format == SBA_FORMAT \|\| ivas_format == SBA_ISM_FORMAT)
		{
		hConfig->nbands = IVAS_MAX_NUM_BANDS;

		spar_dirac_split_band = s_min(IVAS_MAX_NUM_BANDS, SPAR_DIRAC_SPLIT_START_BAND);

		IF (ivas_get_hodirac_flag(ivas_total_brate, sba_order))//add call after merge of 100861_dirac_dec
		{
		spar_dirac_split_band = 0;
		}
		}
		ELSE
		{
		hConfig->nbands = 5;
		spar_dirac_split_band = 0;
		}
		hConfig->enc_param_start_band = 0;
		hConfig->dec_param_estim = FALSE;
		IF (ivas_format == SBA_FORMAT \|\| ivas_format == SBA_ISM_FORMAT) /* skip for MASA decoder */
		{
		IF ((error = ivas_dirac_sba_config_fx(hQMetaData, element_mode, ivas_total_brate, sba_order, hConfig->nbands - spar_dirac_split_band,
		ivas_format)) != IVAS_ERR_OK)
		{
		return error;
		}

		IF (hQMetaData != NULL)
		{
		IF (enc_dec == ENC)
		{
		hConfig->nbands = hQMetaData->q_direction[0].cfg.nbands;
		}
		hConfig->enc_param_start_band = add(hQMetaData->q_direction[0].cfg.start_band , spar_dirac_split_band);
		}

		hConfig->dec_param_estim = TRUE;
		IF (hConfig->dec_param_estim == TRUE)
		{
		hConfig->enc_param_start_band = spar_dirac_split_band;
		}

		IF (ivas_get_hodirac_flag(ivas_total_brate, sba_order))
		{
		hConfig->dec_param_estim = FALSE;
		hConfig->enc_param_start_band = 0;

		set_c((int8_t *)hQMetaData->twoDirBands, (int8_t)1, hQMetaData->q_direction[0].cfg.nbands);
		hQMetaData->numTwoDirBands = (uint8_t)hQMetaData->q_direction[0].cfg.nbands;
		}
		}

		IF (ivas_format == SBA_FORMAT \|\| ivas_format == SBA_ISM_FORMAT)
		{
		//100861_dirac_dec
		ivas_dirac_config_bands(band_grouping, IVAS_MAX_NUM_BANDS, (Word16)(Fs * INV_CLDFB_BANDWIDTH + 0.5f), dirac_to_spar_md_bands, hQMetaData->useLowerBandRes, hConfig->enc_param_start_band, hFbMdft);
		}
		ELSE
		{
		ivas_dirac_config_bands(band_grouping, hConfig->nbands, (Word16)(Fs * INV_CLDFB_BANDWIDTH + 0.5f), NULL, 0, 0, hFbMdft);
		}

		return error;
		}
		#endif

		ivas_error ivas_dirac_config(
		void st_ivas, / i/o: IVAS encoder/decoder state structure */
		@@ -388,7 +518,7 @@ void ivas_dirac_config_bands_fx(
		{
		Word16 step = DIRAC_LOW_BANDRES_STEP;
		Word16 reduced_band;
		for ( band = enc_param_start_band + 2, reduced_band = enc_param_start_band + 1; band <= DIRAC_MAX_NBANDS; band += step, reduced_band++ )
		for ( band = add(enc_param_start_band , 2), reduced_band = add(enc_param_start_band , 1); band <= DIRAC_MAX_NBANDS; band = add(band, step), reduced_band++ )
		{
		band_grouping[reduced_band] = band_grouping[band];
		}
		@@ -433,6 +563,90 @@ void ivas_dirac_config_bands_fx(
		* Return maximum SBA DirAC metadata bit-budget and nominal bit-budget
		-------------------------------------------------------------------/

		#ifdef IVAS_FLOAT_FIXED
		void ivas_get_dirac_sba_max_md_bits_fx(
		const Word32 sba_total_brate,
		Word16 *bits_frame_nominal,
		Word16 *metadata_max_bits,
		Word16 *qmetadata_max_bit_req,
		const Word16 nbands,
		IVAS_FORMAT ivas_format )
		{
		IF ( sba_total_brate <= IVAS_13k2 )
		{
		*bits_frame_nominal = ACELP_9k60 / FRAMES_PER_SEC;
		*metadata_max_bits = 70;
		}
		ELSE IF ( sba_total_brate <= IVAS_16k4 )
		{
		*bits_frame_nominal = ACELP_13k20 / FRAMES_PER_SEC;
		*metadata_max_bits = 80;
		}
		ELSE IF ( sba_total_brate <= IVAS_24k4 )
		{
		*bits_frame_nominal = ACELP_16k40 / FRAMES_PER_SEC;
		*metadata_max_bits = 103;
		/* OSBA needs an additional 2-bits safety margin to avoid acelp crashes */
		IF ( ivas_format == SBA_ISM_FORMAT )
		{
		( metadata_max_bits ) = sub((metadata_max_bits), 3);
		}
		}
		ELSE IF ( sba_total_brate <= IVAS_32k )
		{
		*bits_frame_nominal = ACELP_32k / FRAMES_PER_SEC;
		*metadata_max_bits = 214;
		}
		ELSE IF ( sba_total_brate <= IVAS_48k )
		{
		*bits_frame_nominal = IVAS_48k / FRAMES_PER_SEC;
		*metadata_max_bits = 240;
		}
		ELSE IF ( sba_total_brate <= IVAS_64k )
		{
		*bits_frame_nominal = IVAS_64k / FRAMES_PER_SEC;
		*metadata_max_bits = 200;
		}
		ELSE IF ( sba_total_brate <= IVAS_80k )
		{
		*bits_frame_nominal = IVAS_80k / FRAMES_PER_SEC;
		*metadata_max_bits = 200;
		}
		ELSE IF ( sba_total_brate <= IVAS_96k )
		{
		*bits_frame_nominal = IVAS_96k / FRAMES_PER_SEC;
		*metadata_max_bits = 200;
		}
		ELSE IF ( sba_total_brate <= IVAS_128k )
		{
		*bits_frame_nominal = IVAS_128k / FRAMES_PER_SEC;
		*metadata_max_bits = 250;
		}
		ELSE
		{
		//*bits_frame_nominal = (int16_t) ( sba_total_brate / FRAMES_PER_SEC );
		Word16 tmp_exp = 0, tmp;
		tmp = BASOP_Util_Divide3232_Scale(sba_total_brate, FRAMES_PER_SEC, &tmp_exp);
		*bits_frame_nominal = shr(tmp, 15 - tmp_exp);
		metadata_max_bits = MAX16B; / no limit */
		}
		Word32 var1 = L_mult0(*metadata_max_bits, nbands);
		Word16 exp = 0;
		Word16 var2 = BASOP_Util_Divide3232_Scale(var1, 5, &exp);
		Word32 var2_32 = L_deposit_h(var2);
		Word32 var4 = var2_32;

		Word16 exp_res = 0;
		IF (var1 % 5 != 0) {
		var4 = BASOP_Util_Add_Mant32Exp(var2_32, exp, ONE_IN_Q30, 1, &exp_res);
		}
		*metadata_max_bits = extract_l(L_min(MAX16B, L_shr(var4, 14)));
		*qmetadata_max_bit_req = QMETADATA_MAXBIT_REQ_SBA >> 1;

		return;
		}
		#endif

		void ivas_get_dirac_sba_max_md_bits(
		const int32_t sba_total_brate,
		int16_t *bits_frame_nominal,
		@@ -502,12 +716,95 @@ void ivas_get_dirac_sba_max_md_bits(
		return;
		}


		/*-------------------------------------------------------------------------
		* ivas_dirac_sba_config()
		*
		* DirAC Configuration function for SBA
		------------------------------------------------------------------------/
		#ifdef IVAS_FLOAT_FIXED
		ivas_error ivas_dirac_sba_config_fx(
		IVAS_QMETADATA_HANDLE hQMetaData, /* i/o: q_metadata handle */
		Word16 element_mode, / i/o: element mode of the core coder */
		Word32 sba_total_brate, /* i : SBA total bitrate */
		const Word16 sba_order, /* i : Ambisonic (SBA) order */
		const Word16 nbands /* i : number of frequency bands */
		,
		IVAS_FORMAT ivas_format)
		{
		Word16 nbands_coded;
		Word16 hodirac_flag;
		ivas_error error;

		Word32 tmp1 = IVAS_192k;
		Word32 tmp2 = SPAR_DIRAC_SPLIT_START_BAND;
		Word16 exp = 0;
		Word16 tmp3 = BASOP_Util_Divide3232_Scale(tmp1, tmp2, &exp);
		Word32 res = L_shr(L_deposit_h(tmp3), 31 - exp);

		error = IVAS_ERR_OK;
		hQMetaData->is_masa_ivas_format = 0;
		hodirac_flag = ivas_get_hodirac_flag(sba_total_brate, sba_order);//implemented on 100861_dirac_dec

		/* map the bitrate for SID frame */
		IF (EQ_32(sba_total_brate , IVAS_SID_5k2))
		{
		IF (EQ_16(*element_mode , IVAS_SCE))
		{
		sba_total_brate = ACELP_24k40;
		}
		ELSE
		{
		sba_total_brate = ACELP_48k;
		}
		}

		ivas_set_qmetadata_maxbit_req(hQMetaData, SBA_FORMAT);

		IF (LE_32(sba_total_brate , IVAS_16k4))
		{
		hQMetaData->useLowerRes = 1;
		}
		ELSE
		{
		hQMetaData->useLowerRes = 0;
		}

		nbands_coded = nbands;

		IF (LE_32(sba_total_brate , res))
		{
		hQMetaData->useLowerBandRes = 1;
		Word16 tmp = nbands % 2;
		nbands_coded = add(shr(nbands, 1) , tmp);
		}
		ELSE
		{
		hQMetaData->useLowerBandRes = 0;
		IF (hodirac_flag == 0)
		{
		nbands_coded = nbands - 1; /* always combine the last two bands */
		}
		}

		{
		Word16 no_dirs = 1;
		IF (hodirac_flag)
		{
		no_dirs = 2;
		}

		IF ((error = ivas_qmetadata_allocate_memory(hQMetaData, nbands_coded, no_dirs, 0)) != IVAS_ERR_OK)//WIP
		{
		return error;
		}
		}

		ivas_get_dirac_sba_max_md_bits_fx(sba_total_brate, &hQMetaData->bits_frame_nominal, &hQMetaData->metadata_max_bits, &hQMetaData->qmetadata_max_bit_req, hQMetaData->q_direction[0].cfg.nbands,
		ivas_format);

		return error;
		}
		#endif

		ivas_error ivas_dirac_sba_config(
		IVAS_QMETADATA_HANDLE hQMetaData, /* i/o: q_metadata handle */
		@@ -584,7 +881,6 @@ ivas_error ivas_dirac_sba_config(
		return error;
		}


		/*-------------------------------------------------------------------------
		* computeDirectionVectors()
		*

lib_com/ivas_prot.h

+12 −0

Original line number	Diff line number	Diff line
		@@ -3634,6 +3634,14 @@ int16_t ivas_sba_get_order(
		const int16_t sba_planar /* i : SBA planar flag */
		);

		#ifdef IVAS_FLOAT_FIXED
		Word16 ivas_sba_get_order_fx(
		const Word16 nb_channels, /* i : Number of ambisonic channels */
		const Word16 sba_planar /* i : SBA planar flag */
		);

		#endif

		/! r: Ambisonic (SBA) order used for analysis and coding /
		int16_t ivas_sba_get_analysis_order(
		const int32_t ivas_total_brate, /* i : IVAS total bitrate */
		@@ -3849,6 +3857,10 @@ void generate_masking_noise_lb_dirac(
		);

		#ifdef IVAS_FLOAT_FIXED
		void ivas_dirac_dec_close_fx(
		DIRAC_DEC_HANDLE *hDirAC_out
		);

		void ivas_dirac_dec_set_md_map(
		Decoder_Struct st_ivas, / i/o: IVAS decoder structure */
		const Word16 nCldfbTs /* i : number of CLDFB time slots */

lib_com/ivas_prot_fx.h

+23 −0

Original line number	Diff line number	Diff line
		@@ -1090,4 +1090,27 @@ void stereo_dft_dmx_out_reset_fx(
		STEREO_DFT_DMX_DATA_HANDLE hStereoDftDmx /* i/o: DFT stereo DMX decoder */
		);

		void ivas_get_dirac_sba_max_md_bits_fx(
		const Word32 sba_total_brate,
		Word16 *bits_frame_nominal,
		Word16 *metadata_max_bits,
		Word16 *qmetadata_max_bit_req,
		const Word16 nbands,
		IVAS_FORMAT ivas_format);

		ivas_error ivas_dirac_sba_config_fx(
		IVAS_QMETADATA_HANDLE hQMetaData, /* i/o: q_metadata handle */
		Word16 element_mode, / i/o: element mode of the core coder */
		Word32 sba_total_brate, /* i : SBA total bitrate */
		const Word16 sba_order, /* i : Ambisonic (SBA) order */
		const Word16 nbands /* i : number of frequency bands */
		,
		IVAS_FORMAT ivas_format);

		ivas_error ivas_dirac_config_fx(
		void st_ivas, / i/o: IVAS encoder/decoder state structure */
		const Word16 enc_dec /* i : encoder or decoder flag */
		);
		#endif

lib_com/ivas_sba_config.c

+4 −5

Original line number	Diff line number	Diff line
		@@ -115,7 +115,6 @@ void ivas_sba_config(
		}

		#ifdef IVAS_FLOAT_FIXED
		/not tested/
		void ivas_sba_config_fx(
		const Word32 sba_total_brate, /* i : SBA total bitrate */
		Word16 sba_order, /* i : Ambisonic (SBA) order */
		@@ -133,11 +132,11 @@ void ivas_sba_config_fx(
		}
		ELSE IF( LT_16( sba_order, 0 ) )
		{
		sba_order = ivas_sba_get_order( nb_channels, sba_planar );
		sba_order = ivas_sba_get_order_fx( nb_channels, sba_planar );
		}
		ELSE IF( LT_16( nb_channels, 0 ) )
		{
		nb_channels = ivas_sba_get_nchan_fx( sba_order, sba_planar );
		nb_channels = (Word16)ivas_sba_get_nchan_fx( sba_order, sba_planar );
		}
		ELSE
		{
		@@ -301,7 +300,7 @@ Word16 ivas_sba_get_nchan_fx(
		}
		ELSE
		{
		nb_channels = mult(add(sba_order , 1) , (sba_order + 1));
		nb_channels = mult0(add(sba_order , 1) , add(sba_order , 1));
		}

		return (nb_channels);