harmonize encod_tran_fx(), under CLEANUP_ACELP_ENC (ff8fe054) · Commits · SA4 / Audio / IVAS BASOP

lib_com/ivas_prot_fx.h

+6 −6

Original line number	Diff line number	Diff line
		@@ -2996,7 +2996,7 @@ Word16 ivas_acelp_tcx20_switching_fx(
		const Word16 flag_16k_smc /* i : flag to indicate if the OL SMC is run at 16 kHz */
		);

		void encod_gen_2sbfr(
		void encod_gen_2sbfr_fx(
		Encoder_State st, / i/o: state structure */
		const Word16 speech[], /* i : input speech */
		const Word16 Aw[], /* i : weighted A(z) unquantized for subframes */
		@@ -3010,8 +3010,8 @@ void encod_gen_2sbfr(
		Word16 bwe_exc, / o : excitation for SWB TBE */
		const Word16 tdm_Pitch_reuse_flag, /* i : primary channel pitch reuse flag */
		const Word16 tdm_Pri_pitch_buf[], /* i : pitch values for primary channel */
		Word16 Q_new,
		Word16 shift
		const Word16 Q_new, /* i : Scaling factor */
		const Word16 shift /* i : Shift needed to obtain 12 bits vectors */
		);

		void acelp_fast_fx(
		@@ -4489,7 +4489,7 @@ ivas_error signaling_enc_secondary_fx(
		const Word16 tdm_Pitch_reuse_flag /* i : primary channel pitch reuse flag */
		);

		void tdm_low_rate_enc(
		void tdm_low_rate_enc_fx(
		Encoder_State st, / i/o: State structure */
		const Word16 Aq[], /* i : 12k8 Lp coefficient */
		const Word16 res, / i : residual signal Q_new*/
		@@ -4501,7 +4501,7 @@ void tdm_low_rate_enc(
		const Word16 attack_flag, /* i : attack flag */
		const Word16 lsf_new, / i : current frame ISF vector x2.56 */
		Word16 tmp_noise, / o : long-term noise energy Q8 */
		Word16 Q_new
		const Word16 Q_new /* i : Scaling factor */
		);

		void tdm_low_rate_dec_fx(

lib_enc/acelp_core_enc_fx.c

+7 −2

Original line number	Diff line number	Diff line
		@@ -499,8 +499,13 @@ ivas_error acelp_core_enc_fx(
		}
		ELSE IF( EQ_16( coder_type, TRANSITION ) )
		{
		#ifdef CLEANUP_ACELP_ENC
		tc_subfr_fx = encod_tran_ivas_fx( st_fx, inp_fx, Aw_fx, Aq_fx, Es_pred_fx, res_fx, syn_fx,
		exc_fx, exc2_fx, pitch_buf_fx, voice_factors_fx, bwe_exc_fx, tc_subfr_fx, position, unbits_fx, shift, Q_new ); /* Q0 */
		#else
		tc_subfr_fx = encod_tran_fx( st_fx, inp_fx, Aw_fx, Aq_fx, Es_pred_fx, res_fx, syn_fx,
		exc_fx, exc2_fx, pitch_buf_fx, voice_factors_fx, bwe_exc_fx, tc_subfr_fx, position, unbits_fx, shift, Q_new ); /* Q0 */
		#endif
		}
		ELSE IF( hSC_VBR->ppp_mode )
		{
		@@ -1344,11 +1349,11 @@ ivas_error acelp_core_enc_ivas_fx(
		{
		IF( LE_16( st->coder_type, UNVOICED ) )
		{
		tdm_low_rate_enc( st, Aq, res_fx, syn_fx, exc_fx, pitch_buf, voice_factors_fx, bwe_exc_fx, 0 /attack_flag/, lsf_new_fx, &tmp_noise_fx, Q_new );
		tdm_low_rate_enc_fx( st, Aq, res_fx, syn_fx, exc_fx, pitch_buf, voice_factors_fx, bwe_exc_fx, 0 /attack_flag/, lsf_new_fx, &tmp_noise_fx, Q_new );
		}
		ELSE /* GENERIC */
		{
		encod_gen_2sbfr( st, inp, Aw, Aq, res_fx, syn_fx, exc_fx, exc2_fx, pitch_buf, voice_factors_fx, bwe_exc_fx, tdm_Pitch_reuse_flag, tdm_Pri_pitch_buf_fx, Q_new, 0 );
		encod_gen_2sbfr_fx( st, inp, Aw, Aq, res_fx, syn_fx, exc_fx, exc2_fx, pitch_buf, voice_factors_fx, bwe_exc_fx, tdm_Pitch_reuse_flag, tdm_Pri_pitch_buf_fx, Q_new, 0 );
		}
		}
		ELSE IF( nelp_mode )

lib_enc/enc_gen_voic_fx.c

+7 −2

Original line number	Diff line number	Diff line
		@@ -466,8 +466,9 @@ void encod_gen_voic_ivas_fx(
		Word16 unbits_fx, / i/o: number of unused bits Q0*/
		const Word16 tdm_Pitch_reuse_flag, /* i : primary channel pitch reuse flag Q0*/
		const Word16 tdm_Pri_pitch_buf[], /* i : primary channel pitch buffer Q6*/
		Word16 shift,
		Word16 Q_new )
		const Word16 shift, /* i : Shift needed to obtain 12 bits vectors */
		const Word16 Q_new /* i : Scaling factor */
		)
		{
		Word16 xn_fx[L_SUBFR]; /* Target vector for pitch search */
		Word16 xn2_fx[L_SUBFR]; /* Target vector for codebook search */
		@@ -612,6 +613,7 @@ void encod_gen_voic_ivas_fx(

		Copy_Scale_sig( h1_fx, h2_fx, L_SUBFR, -2 ); /Q11/
		Scale_sig( h1_fx, L_SUBFR, add( 1, shift ) ); /* set h1[] in Q14 with scaling for convolution Q(14+shift)*/

		/* scaling of xn[] to limit dynamic at 12 bits */
		Scale_sig( xn_fx, L_SUBFR, shift );
		}
		@@ -624,8 +626,10 @@ void encod_gen_voic_ivas_fx(
		q_h1 = sub( 14, norm_s( h1_fx[0] ) );
		Copy_Scale_sig( h1_fx, h2_fx, L_SUBFR, sub( 11, q_h1 ) ); /Q11/
		Scale_sig( h1_fx, L_SUBFR, sub( 14, q_h1 ) ); /* set h1[] in Q14 with scaling for convolution Q14*/

		/* scaling of xn[] to limit dynamic at 12 bits */
		Scale_sig( xn_fx, L_SUBFR, shift );

		test();
		test();
		IF( Es_pred_fx < 0 && LE_16( i_subfr_fx, L_SUBFR ) && NE_16( st_fx->last_core, ACELP_CORE ) )
		@@ -908,6 +912,7 @@ void encod_gen_voic_ivas_fx(
		push_indice( hBstr, IND_UNUSED, 0, i );
		unbits_PI_fx -= i;
		}

		IF( st_fx->Opt_SC_VBR )
		{
		/* SC-VBR */

lib_enc/enc_ppp_fx.c

+7 −7

Original line number	Diff line number	Diff line
		/*====================================================================================
		EVS Codec 3GPP TS26.452 Nov 04, 2021. Version 16.4.0
		====================================================================================*/

		#include <stdint.h>
		#include "options.h" /* Compilation switches */
		#include "cnst.h" /* Common constants */
		//#include "prot_fx.h" /* Function prototypes */
		#include "rom_com.h" /* Common constants */
		#include "prot_fx.h" /* Function prototypes */
		#include "prot_fx_enc.h" /* Function prototypes */
		#include "basop_util.h" /* Function prototypes */

		/Temporary location to be move in prot when merge is done */
		void E_LPC_f_lsp_a_conversion( const Word16 lsp, Word16 a, const Word16 m );

		/=======================================================================================/
		/* FUNCTION : encod_ppp_fx() */
		@@ -51,6 +49,7 @@ void E_LPC_f_lsp_a_conversion( const Word16 lsp, Word16 a, const Word16 m );
		/---------------------------------------------------------------------------------------/
		/* CALLED FROM : TX */
		/=======================================================================================/

		ivas_error encod_ppp_fx(
		Encoder_State st_fx, / i/o: state structure */
		const Word16 speech_fx[], /* i : input speech Q_new*/
		@@ -63,8 +62,9 @@ ivas_error encod_ppp_fx(
		Word16 pitch_buf_fx, / o : floating pitch values for each subframe Q6*/
		Word16 voice_factors, / o : voicing factors Q15*/
		Word16 bwe_exc, / o : excitation for SWB TBE Q_new*/
		Word16 Q_new,
		Word16 shift )
		const Word16 Q_new, /* i : Scaling factor */
		const Word16 shift /* i : Shift needed to obtain 12 bits vectors */
		)
		{
		Word16 xn_fx[L_SUBFR]; /* Target vector for pitch search */
		Word16 h1_fx[L_SUBFR + ( M + 1 )]; /* Impulse response vector */

lib_enc/enc_tran_fx.c

+61 −19

Original line number	Diff line number	Diff line
		@@ -43,7 +43,7 @@
		/* RETURN ARGUMENTS : */
		/* _ None */
		/---------------------------------------------------------------------------------/

		#ifndef CLEANUP_ACELP_ENC
		Word16 encod_tran_fx(
		Encoder_State st_fx, / i/o: state structure */
		const Word16 speech_fx[], /* i : input speech Q0*/
		@@ -422,9 +422,13 @@ Word16 encod_tran_fx(

		return tc_subfr;
		}
		#endif


		#ifdef CLEANUP_ACELP_ENC
		Word16 encod_tran_ivas_fx(
		#else
		Word16 encod_tran_ivas_fx(
		#endif
		Encoder_State st_fx, / i/o: state structure */
		const Word16 speech_fx[], /* i : input speech Q0*/
		const Word16 Aw_fx[], /* i : weighted A(z) unquantized for subframes Q12*/
		@@ -485,8 +489,9 @@ Word16 encod_tran_ivas_fx(
		L_frame_fx = st_fx->L_frame;
		move16();

		#ifndef CLEANUP_ACELP_ENC
		set16_fx( h1, 0, L_SUBFR + ( M + 1 ) );

		#endif
		/------------------------------------------------------------------
		* Initializations
		------------------------------------------------------------------/
		@@ -528,7 +533,12 @@ Word16 encod_tran_ivas_fx(
		move16();
		set16_fx( code_preQ, 0, L_SUBFR );
		shift_wsp = add( Q_new, shift );
		#ifdef CLEANUP_ACELP_ENC
		test();
		if ( LT_32( st_fx->core_brate, MIN_BRATE_AVQ_EXC ) && st_fx->element_mode != EVS_MONO )
		#else
		if ( LT_32( st_fx->core_brate, MIN_BRATE_AVQ_EXC ) )
		#endif
		{
		shift_wsp = sub( shift_wsp, 1 );
		}
		@@ -547,7 +557,21 @@ Word16 encod_tran_ivas_fx(

		Copy( &res_fx[i_subfr], &exc_fx[i_subfr], L_SUBFR ); /* Q_new */

		#ifdef CLEANUP_ACELP_ENC
		IF( st_fx->element_mode == EVS_MONO )
		{
		find_targets_fx( speech_fx, hLPDmem->mem_syn, i_subfr, &hLPDmem->mem_w0, p_Aq,
		res_fx, L_SUBFR, p_Aw, st_fx->preemph_fac, xn, cn, h1 );

		Copy_Scale_sig( h1, h2_fx, L_SUBFR, -2 );
		Scale_sig( h1, L_SUBFR, add( 1, shift ) ); /* set h1[] in Q14 with scaling for convolution */

		/* scaling of xn[] to limit dynamic at 12 bits */
		Scale_sig( xn, L_SUBFR, shift );
		}
		ELSE
		#endif
		{
		find_targets_ivas_fx( speech_fx, hLPDmem->mem_syn, i_subfr, &hLPDmem->mem_w0, p_Aq,
		res_fx, L_SUBFR, p_Aw, st_fx->preemph_fac, xn, cn, h1 );

		@@ -561,8 +585,10 @@ Word16 encod_tran_ivas_fx(
		}
		Copy_Scale_sig( h1, h2_fx, L_SUBFR, sub( 11, q_h1 ) ); /Q11/
		Scale_sig( h1, L_SUBFR, add( sub( 14, q_h1 ), shift ) );

		/* scaling of xn[] to limit dynamic at 12 bits */
		Scale_sig( xn, L_SUBFR, shift );
		}

		/-----------------------------------------------------------------
		* TC: subframe determination &
		@@ -646,7 +672,14 @@ Word16 encod_tran_ivas_fx(
		gcode16 = round_fx_sat( Lgcode );

		#ifdef CLEANUP_ACELP_ENC
		IF( st_fx->element_mode == EVS_MONO )
		{
		hLPDmem->tilt_code = Est_tilt2( &exc_fx[i_subfr], gain_pit, code, Lgcode, &voice_fac, shift ); /* Q15 */
		}
		ELSE
		{
		hLPDmem->tilt_code = est_tilt_fx( &exc_fx[i_subfr], gain_pit, code, gain_code, &voice_fac, Q_new, L_SUBFR ); // Q15
		}
		move16();
		#else
		hLPDmem->tilt_code = est_tilt_ivas_fx( &exc_fx[i_subfr], gain_pit, code, gain_code, &voice_fac, Q_new, L_SUBFR, 0 ); // Q15
		@@ -657,7 +690,16 @@ Word16 encod_tran_ivas_fx(
		-----------------------------------------------------------------/

		/st->mem_w0 = xn[L_SUBFR-1] - (gain_pity1[L_SUBFR-1]) - (gain_codey2[L_SUBFR-1]);/
		#ifdef CLEANUP_ACELP_ENC
		IF( st_fx->element_mode == EVS_MONO )
		{
		L_tmp = L_mult( gcode16, y2[L_SUBFR - 1] ); /* Q10 + Q_new */
		}
		ELSE
		#endif
		{
		L_tmp = L_mult0( gcode16, y2[L_SUBFR - 1] );
		}
		L_tmp = L_shl( L_tmp, add( 5, shift ) ); // Q_new+14+shift
		L_tmp = L_negate( L_tmp );
		L_tmp = L_mac( L_tmp, xn[L_SUBFR - 1], 16384 ); // Q_new-1+15+shift
		@@ -680,7 +722,7 @@ Word16 encod_tran_ivas_fx(
		{
		FOR( i = 0; i < L_SUBFR; i++ )
		{
		exc2_fx[i + i_subfr] = round_fx_sat( L_shl_sat( L_mult_sat( gain_pit, exc_fx[i + i_subfr] ), 1 ) ); /* Q_exc */
		exc2_fx[i + i_subfr] = round_fx_sat( L_shl_sat( L_mult_sat( gain_pit, exc_fx[i + i_subfr] ), 1 ) ); /* Q_new / Q_exc */
		}
		}
		ELSE
		@@ -688,7 +730,7 @@ Word16 encod_tran_ivas_fx(
		Gain_pitX2 = shl( gain_pit, 1 );
		FOR( i = 0; i < L_SUBFR; i++ )
		{
		exc2_fx[i + i_subfr] = mult_r( Gain_pitX2, exc_fx[i + i_subfr] ); /* Q_exc */
		exc2_fx[i + i_subfr] = mult_r( Gain_pitX2, exc_fx[i + i_subfr] ); /* Q_new / Q_exc */
		}
		}