Removed scalings before ProcessStereoIGF_fx in stereo_mdct_core_enc_fx and decoder MSAN fix (27e6f622) · Commits · SA4 / Audio / IVAS BASOP

lib_com/prot_fx.h

+6 −0

Original line number	Diff line number	Diff line
		@@ -10859,8 +10859,11 @@ void ProcessStereoIGF_fx(
		Word16 q_pITFMDCTSpectrum_1,
		Word16 q_pITFMDCTSpectrum_2,
		Word32 pPowerSpectrum_fx[CPE_CHANNELS], / i/o: MDCT^2 + MDST^2 spectrum, or estimate */
		Word16 exp_pPowerSpectrum_fx[CPE_CHANNELS], /* i/o: exp of pPowerSpectrum_fx */
		Word32 pPowerSpectrumMsInv_fx[CPE_CHANNELS][NB_DIV], / i : inverse power spectrum */
		Word16 exp_pPowerSpectrumMsInv_fx[CPE_CHANNELS], /* i/o: exp of pPowerSpectrumMsInv_fx */
		Word32 inv_spectrum_fx[CPE_CHANNELS][NB_DIV], / i : inverse spectrum */
		Word16 exp_inv_spectrum_fx[CPE_CHANNELS], /* i/o: exp of inv_spectrum_fx */
		const Word16 frameno, /* i : flag indicating index of current subfr. */
		const Word16 sp_aud_decision0, /* i : sp_aud_decision0 */
		const Word32 element_brate, /* i : element bitrate */
		@@ -10873,8 +10876,11 @@ void IGFEncApplyStereo_fx(
		const Word16 igfGridIdx, /* i : IGF grid index */
		Encoder_State sts[CPE_CHANNELS], / i : Encoder state */
		Word32 pPowerSpectrum_fx[CPE_CHANNELS], / i/o: MDCT^2 + MDST^2 spectrum, or estimate */
		Word16 exp_pPowerSpectrum_fx[CPE_CHANNELS], /* i/o: exp of pPowerSpectrum_fx */
		Word32 pPowerSpectrumMsInv_fx[CPE_CHANNELS][NB_DIV], / i/o: inverse power spectrum */
		Word16 exp_pPowerSpectrumMsInv_fx[CPE_CHANNELS], /* i/o: exp of pPowerSpectrumMsInv_fx */
		Word32 inv_spectrum_fx[CPE_CHANNELS][NB_DIV], / i : inverse spectrum */
		Word16 exp_inv_spectrum_fx[CPE_CHANNELS], /* i : exp of inverse spectrum */
		const Word16 frameno, /* i : flag indicating index of current subfr. */
		const Word16 sp_aud_decision0, /* i : sp_aud_decision0 */
		const Word32 element_brate, /* i : element bitrate */

lib_dec/ivas_osba_dec_fx.c

+5 −0

Original line number	Diff line number	Diff line
		@@ -138,6 +138,11 @@ ivas_error ivas_osba_dirac_td_binaural_jbm_fx(
		Word32 *p_sepobj_fx[BINAURAL_CHANNELS];
		Word16 channel_offset;

		FOR( Word16 i = 0; i < BINAURAL_CHANNELS; i++ )
		{
		set32_fx( output_separated_objects_fx[i], 0, L_FRAME48k );
		}

		FOR( n = 0; n < BINAURAL_CHANNELS; n++ )
		{
		p_sepobj_fx[n] = &output_separated_objects_fx[n][0];

lib_enc/igf_enc.c

+8 −3

Original line number	Diff line number	Diff line
		@@ -2626,8 +2626,11 @@ void IGFEncApplyStereo_fx(
		const Word16 igfGridIdx, /* i : IGF grid index */
		Encoder_State sts[CPE_CHANNELS], / i : Encoder state */
		Word32 pPowerSpectrum_fx[CPE_CHANNELS], / i/o: MDCT^2 + MDST^2 spectrum, or estimate */
		Word16 exp_pPowerSpectrum_fx[CPE_CHANNELS], /* i/o: exp of pPowerSpectrum_fx */
		Word32 pPowerSpectrumMsInv_fx[CPE_CHANNELS][NB_DIV], / i/o: inverse power spectrum */
		Word16 exp_pPowerSpectrumMsInv_fx[CPE_CHANNELS], /* i/o: exp of pPowerSpectrumMsInv_fx */
		Word32 inv_spectrum_fx[CPE_CHANNELS][NB_DIV], / i : inverse spectrum */
		Word16 exp_inv_spectrum_fx[CPE_CHANNELS], /* i : exp of inverse spectrum */
		const Word16 frameno, /* i : flag indicating index of current subfr. */
		const Word16 sp_aud_decision0, /* i : sp_aud_decision0 */
		const Word32 element_brate, /* i : element bitrate */
		@@ -2691,7 +2694,9 @@ void IGFEncApplyStereo_fx(
		last_core_acelp = extract_l( EQ_16( sts[ch]->last_core, ACELP_CORE ) );

		IGF_UpdateInfo( hIGFEnc[ch], igfGridIdx );
		IGF_CalculateStereoEnvelope_fx( hIGFEnc[ch], sts[ch]->hTcxEnc->spectrum_fx[frameno], sts[ch]->hTcxEnc->spectrum_e[frameno], inv_spectrum_fx[ch][frameno], sts[ch]->hTcxEnc->spectrum_e[frameno], pPowerSpectrumParameter_fx[ch], sts[ch]->hTcxEnc->spectrum_e[frameno], pPowerSpectrumParameterMsInv_fx[ch], sts[ch]->hTcxEnc->spectrum_e[frameno], igfGridIdx, coreMsMask, sts[ch]->hTranDet->transientDetector.bIsAttackPresent, last_core_acelp );
		IGF_CalculateStereoEnvelope_fx( hIGFEnc[ch], sts[ch]->hTcxEnc->spectrum_fx[frameno], sts[ch]->hTcxEnc->spectrum_e[frameno], inv_spectrum_fx[ch][frameno],
		exp_inv_spectrum_fx[ch], pPowerSpectrumParameter_fx[ch], exp_pPowerSpectrum_fx[ch], pPowerSpectrumParameterMsInv_fx[ch],
		exp_pPowerSpectrumMsInv_fx[ch], igfGridIdx, coreMsMask, sts[ch]->hTranDet->transientDetector.bIsAttackPresent, last_core_acelp );

		IF( EQ_16( sts[ch]->core, TCX_20_CORE ) )
		{
		@@ -2702,11 +2707,11 @@ void IGFEncApplyStereo_fx(
		pPowerSpectrumParameter_fx[ch] = NULL;
		}

		set16_fx( exp_pPowerSpectrum, sts[ch]->hTcxEnc->spectrum_e[frameno], L_FRAME48k );
		set16_fx( exp_pPowerSpectrum, exp_pPowerSpectrum_fx[ch], L_FRAME48k );

		IGF_Whitening_ivas_fx( hIGFEnc[ch], pPowerSpectrumParameter_fx[ch], &exp_pPowerSpectrum[0], igfGridIdx, sts[ch]->hTranDet->transientDetector.bIsAttackPresent, last_core_acelp, ( sts[0]->hTcxEnc->fUseTns[frameno] \|\| sts[1]->hTcxEnc->fUseTns[frameno] ), sp_aud_decision0, element_brate, sts[ch]->element_mode );

		IGF_ErodeSpectrum_ivas_fx( hIGFEnc[ch], sts[ch]->hTcxEnc->spectrum_fx[frameno], pPowerSpectrumParameter_fx[ch], sts[ch]->hTcxEnc->spectrum_e[frameno], igfGridIdx, mct_on );
		IGF_ErodeSpectrum_ivas_fx( hIGFEnc[ch], sts[ch]->hTcxEnc->spectrum_fx[frameno], pPowerSpectrumParameter_fx[ch], exp_pPowerSpectrum_fx[ch], igfGridIdx, mct_on );
		}
		return;
		}

lib_enc/ivas_mct_enc_mct_fx.c

+7 −5

Original line number	Diff line number	Diff line
		@@ -1072,7 +1072,7 @@ void mctStereoIGF_enc_fx(
		IF( NE_16( hMCT->hBlockData[b]->hStereoMdct->mdct_stereo_mode[n], hMCT->hBlockData[b]->hStereoMdct->IGFStereoMode[n] ) \|\|
		EQ_16( hMCT->hBlockData[b]->hStereoMdct->mdct_stereo_mode[n], SMDCT_BW_MS ) )
		{

		Word16 exp_powerSpec_tmp[CPE_CHANNELS];
		FOR( ch = 0; ch < CPE_CHANNELS; ch++ )
		{
		s = s_min( s, sub( 31, p_st[ch]->hTcxEnc->spectrum_e[n] ) );
		@@ -1095,10 +1095,12 @@ void mctStereoIGF_enc_fx(
		move16();
		move16();
		}

		ProcessStereoIGF_fx( hMCT->hBlockData[b]->hStereoMdct, p_st, hMCT->hBlockData[b]->mask,
		p_orig_spectrum_fx, q_origSpec, q_origSpec,
		p_powerSpec_fx, p_powerSpecMsInv_fx, p_inv_spectrum_fx, n, sp_aud_decision0[ch1], p_st[0]->total_brate, 1 );
		exp_powerSpec_tmp[0] = p_st[0]->hTcxEnc->spectrum_e[0];
		exp_powerSpec_tmp[1] = p_st[1]->hTcxEnc->spectrum_e[0];
		move16();
		move16();
		ProcessStereoIGF_fx( hMCT->hBlockData[b]->hStereoMdct, p_st, hMCT->hBlockData[b]->mask, p_orig_spectrum_fx, q_origSpec, q_origSpec, p_powerSpec_fx, exp_powerSpec_tmp,
		p_powerSpecMsInv_fx, exp_powerSpec_tmp, p_inv_spectrum_fx, exp_powerSpec_tmp, n, sp_aud_decision0[ch1], p_st[0]->total_brate, 1 );
		}
		ELSE
		{

lib_enc/ivas_stereo_mdct_core_enc_fx.c

+20 −48

Original line number	Diff line number	Diff line
		@@ -143,6 +143,7 @@ void stereo_mdct_core_enc_fx(
		Word32 *quantized_spectrum_fx[CPE_CHANNELS][NB_DIV];
		Word32 *inv_mdst_spectrum_fx[CPE_CHANNELS][NB_DIV];
		Word32 *inv_spectrum_fx[CPE_CHANNELS][NB_DIV];
		Word16 exp_inv_spectrum[CPE_CHANNELS];
		Word16 i, ch, nSubframes, L_subframeTCX;
		Word16 n, nAvailBits;
		Word16 tnsSize[CPE_CHANNELS][NB_DIV]; /* number of tns parameters put into prm */
		@@ -151,8 +152,8 @@ void stereo_mdct_core_enc_fx(
		Word32 *p_mdst_spectrum_long_fx[CPE_CHANNELS];
		Word32 mdst_spectrum_long_fx[CPE_CHANNELS][N_MAX];
		Word32 *mdst_spectrum_fx[CPE_CHANNELS][NB_DIV];
		Word16 q_powSpecMsInv[MCT_MAX_CHANNELS], q_spec, tmp_s;
		Word16 tmp_q_powSpec[N_MAX], tmp_q_powSpecInv[N_MAX], *tmp_q_psi[2];
		Word16 exp_powSpecMsInv[MCT_MAX_CHANNELS], q_spec, tmp_s;
		Word16 tmp_q_powSpecInv[N_MAX], *tmp_q_psi[2];
		Word64 W_tmp;
		Encoder_State st, *sts;
		STEREO_MDCT_ENC_DATA_HANDLE hStereoMdct;
		@@ -186,8 +187,6 @@ void stereo_mdct_core_enc_fx(
		}

		set16_fx( tmp_q_powSpecInv, 63, N_MAX );
		set16_fx( tmp_q_powSpec, 63, N_MAX );
		set16_fx( q_powSpecMsInv, 31, MCT_MAX_CHANNELS );

		tmp_q_psi[0] = tmp_q_powSpecInv;
		tmp_q_psi[1] = &tmp_q_powSpecInv[N_TCX10_MAX];
		@@ -398,6 +397,9 @@ void stereo_mdct_core_enc_fx(
		}
		}
		stereo_coder_tcx_fx( hStereoMdct, sts, ms_mask, mdst_spectrum_fx, inv_spectrum_fx, inv_mdst_spectrum_fx, 0, q_spec );
		exp_inv_spectrum[0] = exp_inv_spectrum[1] = sub( Q31, q_spec );
		move16();
		move16();
		}

		/--------------------------------------------------------------
		@@ -516,16 +518,15 @@ void stereo_mdct_core_enc_fx(

		/* Aligning the Q-factors */
		{
		q_powSpecMsInv[ch] = Q31;
		Word16 q_temp = Q31;
		move16();
		FOR( n = 0; n < nSubframes; n++ )
		{
		FOR( i = 0; i < L_subframeTCX; i++ )
		{
		IF( powerSpecMsInv_fx[ch][n][i] != 0 )
		if ( powerSpecMsInv_fx[ch][n][i] != 0 )
		{
		q_powSpecMsInv[ch] = s_min( q_powSpecMsInv[ch], tmp_q_psi[n][i] );
		move16();
		q_temp = s_min( q_temp, tmp_q_psi[n][i] );
		}
		}
		}
		@@ -533,10 +534,12 @@ void stereo_mdct_core_enc_fx(
		{
		FOR( i = 0; i < L_subframeTCX; i++ )
		{
		powerSpecMsInv_fx[ch][n][i] = L_shr_sat( powerSpecMsInv_fx[ch][n][i], sub( tmp_q_psi[n][i], q_powSpecMsInv[ch] ) );
		powerSpecMsInv_fx[ch][n][i] = L_shr_sat( powerSpecMsInv_fx[ch][n][i], sub( tmp_q_psi[n][i], q_temp ) );
		move32();
		}
		}
		exp_powSpecMsInv[ch] = sub( Q31, q_temp );
		move16();
		}
		}

		@@ -605,7 +608,7 @@ void stereo_mdct_core_enc_fx(
		{
		IGF_ENC_INSTANCE_HANDLE hIGFEnc[CPE_CHANNELS];
		Word32 powerSpec_fx[CPE_CHANNELS][N_MAX], *p_powerSpec_fx[CPE_CHANNELS];
		Word16 q_powSpec[CPE_CHANNELS];
		Word16 exp_powSpec[CPE_CHANNELS];
		p_powerSpec_fx[0] = powerSpec_fx[0];
		p_powerSpec_fx[1] = powerSpec_fx[1];
		{
		@@ -642,52 +645,21 @@ void stereo_mdct_core_enc_fx(
		}
		}
		set32_fx( powerSpec_fx[ch] + length, 0, sub( N_MAX, length ) );
		q_powSpec[ch] = sub( 31, exp ); // exp: exp
		exp_powSpec[ch] = exp; // exp: exp
		move16();
		}
		}

		hIGFEnc[0] = sts[0]->hIGFEnc;
		hIGFEnc[1] = sts[1]->hIGFEnc;

		FOR( ch = 0; ch < CPE_CHANNELS; ch++ )
		{
		q_spec = sub( Q31, sts[ch]->hTcxEnc->spectrum_e[n] );
		Word16 q_comm = s_min( sub( Q31, sts[ch]->hTcxEnc->spectrum_e[n] ), s_min( q_powSpec[ch], q_powSpecMsInv[ch] ) );
		Word16 n_sb = NB_DIV;
		move16();
		if ( EQ_16( sts[ch]->hTcxEnc->tcxMode, TCX_20 ) )
		{
		n_sb = 1;
		move16();
		}
		length = idiv1616( sts[ch]->hTcxEnc->L_frameTCX, n_sb ); /* Q0 */
		IF( NE_16( q_spec, q_comm ) )
		{
		Scale_sig32( sts[ch]->hTcxEnc->spectrum_fx[n], length, sub( q_comm, q_spec ) ); /* q_powSpec */
		Scale_sig32( inv_spectrum_fx[ch][n], length, sub( q_comm, q_spec ) ); /* q_powSpec */
		sts[ch]->hTcxEnc->spectrum_e[n] = sub( Q31, q_comm );
		move16();
		}
		IF( NE_16( q_powSpec[ch], q_comm ) )
		{
		Scale_sig32( &p_powerSpec_fx[ch][0], L_subframeTCX, sub( q_comm, q_powSpec[ch] ) ); /* q_spec */
		q_powSpec[ch] = q_comm;
		move16();
		}
		IF( NE_16( q_powSpecMsInv[ch], q_comm ) )
		{
		Scale_sig32( powerSpecMsInv_fx[ch][0], L_subframeTCX, sub( q_comm, q_powSpecMsInv[ch] ) ); /* q_spec */
		q_powSpecMsInv[ch] = q_comm;
		move16();
		}
		}
		hIGFEnc[0]->spec_be_igf_e = p_orig_spectrum_e[0];
		move16();
		hIGFEnc[1]->spec_be_igf_e = p_orig_spectrum_e[1];
		move16();
		ProcessStereoIGF_fx( hStereoMdct, sts, ms_mask,
		orig_spectrum_fx, sub( Q31, p_orig_spectrum_e[0] ), sub( Q31, p_orig_spectrum_e[1] ),
		p_powerSpec_fx, powerSpecMsInv_fx, inv_spectrum_fx, n, hCPE->hCoreCoder[0]->sp_aud_decision0, hCPE->hCoreCoder[0]->element_brate, 0 );
		move16();

		ProcessStereoIGF_fx( hStereoMdct, sts, ms_mask, orig_spectrum_fx, sub( Q31, p_orig_spectrum_e[0] ), sub( Q31, p_orig_spectrum_e[1] ),
		p_powerSpec_fx, exp_powSpec, powerSpecMsInv_fx, exp_powSpecMsInv, inv_spectrum_fx, exp_inv_spectrum,
		n, hCPE->hCoreCoder[0]->sp_aud_decision0, hCPE->hCoreCoder[0]->element_brate, 0 );
		}
		ELSE
		{