Merge branch 'ivas_stereo_mdct_front_vad_spar_changes_enc' into 'main' (00fe242e) · Commits · SA4 / Audio / IVAS BASOP

lib_com/ACcontextMapping.c

+23 −0

Original line number	Diff line number	Diff line
		@@ -47,6 +47,7 @@
		-------------------------------------------------------------------/

		/! r: index of next coefficient /
		#ifndef IVAS_FLOAT_FIXED
		int16_t get_next_coeff_mapped_ivas(
		int16_t ii[2], /* i/o: coefficient indexes */
		int32_t pp, / o : peak(1)/hole(0) indicator */
		@@ -64,6 +65,28 @@ int16_t get_next_coeff_mapped_ivas(

		return hm_cfg->indexBuffer[*idx];
		}
		#else
		Word16 get_next_coeff_mapped_ivas(
		Word16 ii[2], /* i/o: coefficient indexes */
		Word32 pp, / o : peak(1)/hole(0) indicator */
		Word16 idx, / o : index in unmapped domain */
		CONTEXT_HM_CONFIG hm_cfg / i : HM configuration */
		)
		{
		UWord32 p;

		p = s_and( sub( ii[1], hm_cfg->numPeakIndices ), sub( hm_cfg->indexBuffer[ii[1]], hm_cfg->indexBuffer[ii[0]] ) );
		p >>= sub( sizeof( p ) * 8, 1 );
		*pp = p;
		move32();
		*idx = ii[p];
		move16();
		ii[p] = add( ii[p], 1 );
		move16();

		return hm_cfg->indexBuffer[*idx];
		}
		#endif


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

lib_com/cnst.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -1919,6 +1919,7 @@ typedef enum _DCTTYPE
		#define N_SMC_MIXTURES 6 /* number of mixtures */
		#define N_PCA_COEF 12 /* number of PCA components */
		#define SMC_ST_MEAN_FACT 0.5 /* forgetting factor of short-term IIR mean filter */
		#define SMC_ST_MEAN_RSHIFT_FACT_FX 1 /* SMC_ST_MEAN_FACT equivalent right shift factor */

		#define M_LSP_SPMUS 6 /* number of LSPs used in speech/music classifier */
		#define NB_BANDS_SPMUS 15

lib_com/ivas_prot.h

+53 −7

Original line number	Diff line number	Diff line
		@@ -720,7 +720,27 @@ int16_t ivas_smc_gmm(
		int16_t high_lpn_flag, / i/o: sp/mus LPN flag */
		const int16_t flag_spitch /* i : flag to indicate very short stable pitch */
		);

		#ifdef IVAS_FLOAT_FIXED
		/! r: S/M decision (0=speech or noise,1=unclear,2=music) /
		Word16 ivas_smc_gmm_fx(
		Encoder_State st, / i/o: state structure */
		STEREO_CLASSIF_HANDLE hStereoClassif, /* i/o: stereo classifier structure */
		const Word16 localVAD_HE_SAD, /* i : HE-SAD flag without hangover */
		const Word16 Etot_fx, /* i : total frame energy */
		Word16 lsp_new_fx[M], /* i : LSPs in current frame TODO:For now removing 'const' to avoid warning */
		Word16 cor_map_sum_fx, /* i : correlation map sum (from multi-harmonic anal.) */
		Word32 epsP_fx[M + 1], /* i : LP prediciton error TODO:For now removing 'const' to avoid warning */
		Word32 PS_fx[], /* i : energy spectrum TODO:For now removing 'const' to avoid warning */
		const Word16 non_sta_fx, /* i : unbound non-stationarity */
		const Word16 relE_fx, /* i : relative frame energy */
		Word16 high_lpn_flag, / i/o: sp/mus LPN flag */
		const Word16 flag_spitch /* i : flag to indicate very short stable pitch */
		,
		Word16 Qfact_PS,
		Word16 Q_esp,
		Word16 Qfact_PS_past
		);
		#endif
		void ivas_smc_mode_selection(
		Encoder_State st, / i/o: encoder state structure */
		const int32_t element_brate, /* i : element bitrate */
		@@ -2900,11 +2920,6 @@ ivas_error initMdctItdHandling(
		const int32_t input_Fs /* i : input sampling rate */
		);

		ivas_error initMdctItdHandling_fx(
		STEREO_MDCT_ENC_DATA hStereoMdct, / i/o: mdct stereo parameters structure */
		const Word32 input_Fs /* i : input sampling rate */
		);

		void stereo_mdct_enc_destroy(
		STEREO_MDCT_ENC_DATA_HANDLE hStereoMdct / i/o: encoder MDCT stereo handle */
		);
		@@ -3195,10 +3210,12 @@ void parse_stereo_from_bitstream(
		);
		#endif

		#ifndef IVAS_FLOAT_FIXED
		void FindSplitRatio(
		CPE_ENC_HANDLE hCPE, /* i/o: CPE encoder structure */
		Encoder_State *sts / i/o: Encoder state structure */
		);
		#endif

		void ComputeSpectrumNoiseMeasure(
		const float *powerSpec,
		@@ -3514,6 +3531,20 @@ void v_add_inc_fx(
		const Word16 y_inc, /* i : increment for vector y[] */
		const Word16 N /* i : Vector length */
		);
		Word32 logsumexp_fx(
		const Word32 x[], /* i : input array x */
		const Word16 x_e,
		const Word16 N /* i : number of elements in array x */
		);
		Word32 lin_interp32_fx(
		const Word32 x, /* i : the value to be mapped */
		const Word32 x1, /* i : source range interval: low end */
		const Word32 y1, /* i : source range interval: high end */
		const Word32 x2, /* i : target range interval: low */
		const Word32 y2, /* i : target range interval: high */
		const Word16 flag_sat, /* i : flag to indicate whether to apply saturation */
		Word16 Q_io /i/o : i/o Q factor of the output*/
		);
		#endif

		void v_mult_inc(
		@@ -3624,6 +3655,13 @@ Word32 dot_product_cholesky_fx(
		const Word32 A, / i : Cholesky matrix A */
		const Word16 N /* i : vector & matrix size */
		);
		Word32 dot_product_cholesky_fixed(
		const Word32 x, / i : vector x */
		const Word32 A, / i : Cholesky matrix A */
		const Word16 N, /* i : vector & matrix size */
		const Word16 exp_x,
		const Word16 exp_A,
		Word16 *exp_sum );
		#endif

		#ifdef IVAS_FLOAT_FIXED
		@@ -3637,6 +3675,14 @@ void v_mult_mat_fx(
		const Word16 N, /* i : number of rows */
		const Word16 C /* i : number of columns */
		);
		void v_mult_mat_fixed(
		Word32 y, / o : the product xA /
		const Word32 x, / i : vector x */
		const Word32 A, / i : matrix A */
		const Word16 Nr, /* i : number of rows */
		const Word16 Nc, /* i : number of columns */
		Word16 guardbits
		);
		#endif

		void v_mult_mat(

lib_com/ivas_prot_fx.h

+57 −0

Original line number	Diff line number	Diff line
		@@ -2445,4 +2445,61 @@ void ivas_ari_encode_14bits_ext_fx(
		Tastat *s,
		Word32 symbol,
		const UWord16 *cum_freq );

		void ms_inv_mask_processing_fx(
		STEREO_MDCT_ENC_DATA_HANDLE hStereoMdct, /* i/o: MDCT encoder structure */
		Encoder_State *sts, / i/o: Encoder state structure */
		int16_t ms_mask[NB_DIV][MAX_SFB], /* i : bandwise MS mask */
		const int16_t iSubframe, /* i : subframe number */
		const Word32 x_0_fx[], /* i : spectrum 1 */
		const Word32 x_1_fx[], /* i : spectrum 2 */
		Word32 x_inv_0_fx[], /* o : inverse spectrum 1 */
		Word32 x_inv_1_fx[], /* o : inverse spectrum 2 */
		int16_t maxSfb /* i : number of stereo frequency bands */
		);

		void ms_processing_fx(
		STEREO_MDCT_ENC_DATA_HANDLE hStereoMdct, /* i/o: Stereo MDCT encoder structure */
		Encoder_State *sts, / i/o: Encoder state structure */
		Word16 ms_mask[NB_DIV][MAX_SFB], /* i : bandwise MS mask */
		const int16_t iSubframe, /* i : subframe number */
		Word32 x_0_fx[], /* i/o: spectrum 1 */
		Word32 x_1_fx[], /* i/o: spectrum 1 */
		Word16 maxSfb /* i : number of stereo frequency bands*/
		);

		void convertToMS_fx(
		const int16_t L_frame, /* i : frame length */
		Word32 x0[], /* i/o: mid/left channel coefficients */
		Word32 x1[], /* i/o: side/right channel coefficients */
		const Word32 norm_fac /* i : normalization factor */
		);

		void FindSplitRatio_fx(
		CPE_ENC_HANDLE hCPE, /* i/o: CPE encoder structure */
		Encoder_State *sts / i/o: Encoder state structure */
		);

		void IGFEncStereoEncoder_fx(
		STEREO_MDCT_BAND_PARAMETERS sfbParam, / i/o: sfb parameters for the right channel */
		const IGF_ENC_INSTANCE_HANDLE hIGFEnc, /* i : IGF handle */
		const Word32 mdctSpectrumL_fx, / i : left spectrum */
		const Word32 mdctSpectrumR_fx, / i : right spectrum */
		Word16 q_mdctSpectrum,
		Word16 msMask, / i/o: MS mask */
		Word16 igfStereoMode, / o : IGF stereo mode */
		const Word16 mdct_stereo_mode, /* i : MDCT stereo mode */
		const Word16 isTCX20, /* i : flag for indicating TCX20 */
		const Word16 isTransition /* i : flag for transtition */
		);

		void stereo_coder_tcx_fx(
		STEREO_MDCT_ENC_DATA_HANDLE hStereoMdct, /* i/o: Stereo MDCT encoder structure */
		Encoder_State *sts, / i/o: encoder state structure */
		Word16 ms_mask[NB_DIV][MAX_SFB], /* i : bandwise MS mask */
		Word32 mdst_spectrum_fx[CPE_CHANNELS][NB_DIV], / i/o: MDST spectrum */
		Word32 inv_spectrum_fx[CPE_CHANNELS][NB_DIV], / i/o: inverse spectrum */
		Word32 inv_mdst_spectrum_fx[CPE_CHANNELS][NB_DIV], / i/o: inverse MDST spectrum */
		const Word16 mct_on, /* i : flag mct block (1) or stereo (0) */
		Word16 q_spec );
		#endif

lib_com/ivas_tools.c

+172 −1

Original line number	Diff line number	Diff line
		@@ -809,6 +809,27 @@ void v_sub_s16_fx(
		return;
		}
		#endif // IVAS_FLOAT_FIXED


		#ifdef IVAS_FLOAT_FIXED
		void v_sub32_fx(
		const Word32 x1[], /* i : Input vector 1 */
		const Word32 x2[], /* i : Input vector 2 */
		Word32 y[], /* o : Output vector that contains vector 1 - vector 2 */
		const Word16 N /* i : Vector length */
		)
		{
		Word16 i;

		FOR( i = 0; LT_16( i, N ); i++ )
		{
		y[i] = L_sub( x1[i], x2[i] );
		}

		return;
		}
		#endif // IVAS_FLOAT_FIXED

		void v_sub_s(
		const int16_t x1[], /* i : Input vector 1 */
		const int16_t x2[], /* i : Input vector 2 */
		@@ -855,7 +876,8 @@ float dot_product_cholesky(
		pt_x = x;
		for ( j = 0; j <= i; j++ )
		{
		tmp_sum += pt_x++ *pt_A++;
		float mul = pt_x++ *pt_A++;
		tmp_sum += mul;
		}

		suma += tmp_sum * tmp_sum;
		@@ -863,8 +885,73 @@ float dot_product_cholesky(

		return suma;
		}
		#ifdef IVAS_FLOAT_FIXED
		Word32 dot_product_cholesky_fixed(
		const Word32 x, / i : vector x */
		const Word32 A, / i : Cholesky matrix A */
		const Word16 N, /* i : vector & matrix size */
		const Word16 exp_x,
		const Word16 exp_A,
		Word16 *exp_sum )
		{
		Word16 i, j;
		Word32 suma, tmp_sum, mul;
		const Word32 pt_x, pt_A;
		Word16 mul_exp, tmp_sum_exp;
		mul_exp = add( exp_x, exp_A );
		pt_A = A;
		suma = 0;
		move32();
		FOR( i = 0; i < N; i++ )
		{
		tmp_sum = 0;
		move32();
		tmp_sum_exp = 0;
		move16();
		pt_x = x;

		FOR( j = 0; j <= i; j++ )
		{
		mul = Mpy_32_32( pt_x++, pt_A++ );
		tmp_sum = BASOP_Util_Add_Mant32Exp( tmp_sum, tmp_sum_exp, mul, mul_exp, &tmp_sum_exp ); // exp_x+exp_A
		}

		suma = BASOP_Util_Add_Mant32Exp( suma, *exp_sum, Mpy_32_32( tmp_sum, tmp_sum ), shl( tmp_sum_exp, 1 ), exp_sum );
		}

		return suma;
		}
		#endif
		#ifdef IVAS_FLOAT_FIXED
		void v_mult_mat_fixed(
		Word32 y, / o : the product xA /
		const Word32 x, / i : vector x */
		const Word32 A, / i : matrix A */
		const Word16 Nr, /* i : number of rows */
		const Word16 Nc, /* i : number of columns */
		Word16 guardbits )
		{
		Word16 i, j;
		const Word32 pt_x, pt_A;
		Word32 tmp_y[MAX_V_MULT_MAT];
		Word32 *pt_y;

		pt_y = tmp_y;
		pt_A = A;

		FOR( i = 0; i < Nc; i++ )
		{
		pt_x = x;
		*pt_y = 0;
		FOR( j = 0; j < Nr; j++ )
		{
		pt_y = L_add( pt_y, L_shr( Mpy_32_32( ( pt_x++ ), ( pt_A++ ) ), guardbits ) );
		}
		pt_y++;
		}

		mvr2r_Word32( tmp_y, y, Nc );
		}
		Word32 dot_product_cholesky_fx(
		const Word32 x, / i : vector x */
		const Word32 A, / i : Cholesky matrix A */
		@@ -1045,14 +1132,98 @@ float logsumexp(

		return logf( sum ) + max_exp;
		}
		#ifdef IVAS_FLOAT_FIXED
		Word32 logsumexp_fx(
		const Word32 x[], /* i : input array x */
		const Word16 x_e,
		const Word16 N /* i : number of elements in array x */
		)
		{
		Word32 max_exp, temp32_sub;
		Word32 sum, temp32, pow_temp;
		Word32 log2_e_fx = 1549082005; // Q30 of log2(e);
		Word16 log2_e_fx_e = 1;
		move16();
		move16();
		Word16 i;
		Word16 pow_e, sum_e = 0;
		move16();
		max_exp = x[0];
		move32();
		sum = 0;
		move32();
		FOR( i = 1; i < N; i++ )
		{
		IF( GT_32( x[i], max_exp ) )
		{
		max_exp = x[i];
		move32();
		}
		}

		FOR( i = 0; i < N; i++ )
		{
		temp32_sub = L_sub( x[i], max_exp );
		pow_e = 0;
		move16();
		temp32 = Mpy_32_32( log2_e_fx, temp32_sub );
		pow_temp = BASOP_util_Pow2( temp32, add( x_e, log2_e_fx_e ), &pow_e );
		sum = BASOP_Util_Add_Mant32Exp( sum, sum_e, pow_temp, pow_e, &sum_e );
		}
		temp32 = L_add( BASOP_Util_Log2( sum ), L_shl( sum_e, Q25 ) );
		temp32 = Mpy_32_32( temp32, 1488522239 ); /logf(x) = log2(x)logf(2)*/
		temp32 = L_add( L_shr( temp32, sub( x_e, 6 ) ), max_exp ); // q = 31-x_e
		return temp32;
		}
		#endif

		/---------------------------------------------------------------------
		* lin_interp()
		*
		* Linearly maps x from source range <x1, x2> to the target range <y1, y2>
		---------------------------------------------------------------------/
		#ifdef IVAS_FLOAT_FIXED
		/! r: mapped output value /
		Word32 lin_interp32_fx(
		const Word32 x, /* i : the value to be mapped */
		const Word32 x1, /* i : source range interval: low end */
		const Word32 y1, /* i : source range interval: high end */
		const Word32 x2, /* i : target range interval: low */
		const Word32 y2, /* i : target range interval: high */
		const Word16 flag_sat, /* i : flag to indicate whether to apply saturation */
		Word16 Q_io /i/o : i/o Q factor of the output*/
		)
		{
		Word32 temp32;
		Word32 temp_div;
		Word16 temp_e = 0;
		Word16 exp_out = 0;
		move16();
		move16();
		IF( L_sub( x2, x1 ) == 0 )
		{
		return y1;
		}
		ELSE IF( flag_sat )
		{
		IF( GE_32( x, L_max( x1, x2 ) ) )
		{
		return GT_32( x1, x2 ) ? y1 : y2;
		}
		ELSE IF( LE_32( x, L_min( x1, x2 ) ) )
		{
		return LT_32( x1, x2 ) ? y1 : y2;
		}
		}

		temp32 = Mpy_32_32( L_sub( x, x1 ), L_sub( y2, y1 ) ); // Qin*2 -31
		temp_div = L_deposit_h( BASOP_Util_Divide3232_Scale( temp32, L_sub( x2, x1 ), &temp_e ) ); // 31-temp_e + 2*Qin -31 - Qin = Qin-temp_e
		temp32 = BASOP_Util_Add_Mant32Exp( y1, sub( 31, Q_io ), temp_div, sub( 31, sub( Q_io, temp_e ) ), &exp_out ); // Qin-temp_e
		*Q_io = sub( 31, exp_out );
		return temp32;
		}

		#endif
		/! r: mapped output value /
		float lin_interp(
		const float x, /* i : the value to be mapped */