Crash fixes for LTV encoder, ASAN/MSAN fixes (0e1ff48e) · Commits · SA4 / Audio / IVAS BASOP

lib_com/stat_com.h

+2 −1

Original line number	Diff line number	Diff line
		@@ -197,7 +197,8 @@ typedef struct TNS_filter_structure
		Word16 order; /* Filter order. */
		Word16 coefIndex[TNS_MAX_FILTER_ORDER]; /* Quantized filter coefficients. */
		Word16 predictionGain; /* Prediction gain. The ratio of a signal and TNS residual energy. E(PRED_GAIN_E), Q7 */
		Word32 predictionGain32; /* Prediction gain. The ratio of a signal and TNS residual energy. E(PRED_GAIN_E), Q23 */
		Word32 predictionGain32; /* Prediction gain. The ratio of a signal and TNS residual energy. predictionGain_e */
		Word16 predictionGain_e; /Exponent for predictionGain32 /
		Word16 avgSqrCoef; /* Average squared filter coefficient. E(0), Q15 */
		} STnsFilter;

lib_com/tns_base.c

+2 −0

Original line number	Diff line number	Diff line
		@@ -1204,6 +1204,8 @@ void ResetTnsData( STnsData *pTnsData )
		move16();
		pTnsFilter->predictionGain32 = ONE_IN_Q23; /Q23/
		move32();
		pTnsFilter->predictionGain_e = PRED_GAIN_E;
		move16();
		pTnsFilter->avgSqrCoef = 0;
		move16();
		pTnsFilter->filterType = TNS_FILTER_OFF; /Q0/

lib_enc/cod_tcx.c

+110 −17

Original line number	Diff line number	Diff line
		@@ -74,8 +74,11 @@ void TNSAnalysisStereo_fx(
		TCX_ENC_HANDLE hTcxEnc = NULL;
		Word16 individual_decision[NB_DIV];
		Word32 maxPredictionGain_fx = 0, meanPredictionGain_fx;
		move32();
		Word16 maxPredictionGain_e = Q31, meanPredictionGain_e;
		move16();
		Word16 sum_e = 0;
		move16();

		individual_decision[0] = 0;
		move16();
		individual_decision[1] = 0;
		@@ -210,7 +213,9 @@ void TNSAnalysisStereo_fx(
		test();
		IF( sts[0]->hTcxCfg->fIsTNSAllowed && NE_16( individual_decision[k], 1 ) && ( !bWhitenedDomain \|\| sts[0]->hTcxEnc->bTnsOnWhithenedSpectra[k] ) )
		{
		Word32 maxPredGain_fx = -ONE_IN_Q23;
		Word32 maxPredGain_fx = -ONE_IN_Q31;
		move32();
		Word16 maxPredGain_e = 0;
		move16();
		sts[0]->hTcxCfg->pCurrentTnsConfig = &sts[0]->hTcxCfg->tnsConfig[sts[0]->hTcxEnc->transform_type[k] == TCX_20][( k == 0 ) && ( sts[0]->last_core == ACELP_CORE )];
		sts[1]->hTcxCfg->pCurrentTnsConfig = &sts[1]->hTcxCfg->tnsConfig[sts[1]->hTcxEnc->transform_type[k] == TCX_20][( k == 0 ) && ( sts[1]->last_core == ACELP_CORE )];
		@@ -228,30 +233,74 @@ void TNSAnalysisStereo_fx(
		* both filters for the decision
		*/

		meanPredictionGain_fx = L_add( Mpy_32_16_1( pFilter[0]->predictionGain32, 16384 /0.5f Q15/ ), Mpy_32_16_1( pFilter[1]->predictionGain32, 16384 /0.5f Q15/ ) ); // Q23
		maxPredictionGain_fx = L_max( maxPredictionGain_fx, meanPredictionGain_fx ); // Q23

		meanPredictionGain_fx = BASOP_Util_Add_Mant32Exp( Mpy_32_16_1( pFilter[0]->predictionGain32, 16384 /0.5f Q15/ ), pFilter[0]->predictionGain_e, Mpy_32_16_1( pFilter[1]->predictionGain32, 16384 /0.5f Q15/ ), pFilter[1]->predictionGain_e, &meanPredictionGain_e ); // meanPredictionGain_e
		Word16 flag = BASOP_Util_Cmp_Mant32Exp( maxPredictionGain_fx, maxPredictionGain_e, meanPredictionGain_fx, meanPredictionGain_e );
		IF( flag < 0 )
		{
		maxPredictionGain_fx = meanPredictionGain_fx;
		maxPredictionGain_e = meanPredictionGain_e;
		move32();
		move16();
		}
		flag = BASOP_Util_Cmp_Mant32Exp( pFilter[0]->predictionGain32, pFilter[0]->predictionGain_e, L_deposit_h( pTnsParameters[0]->minPredictionGain ), PRED_GAIN_E );
		if ( flag < 0 )
		{
		flag = 0;
		move16();
		}
		Word16 flag_1 = BASOP_Util_Cmp_Mant32Exp( pFilter[1]->predictionGain32, pFilter[1]->predictionGain_e, L_deposit_h( pTnsParameters[1]->minPredictionGain ), PRED_GAIN_E );
		if ( flag_1 < 0 )
		{
		flag_1 = 0;
		move16();
		}
		test();
		test();
		test();
		IF( GT_32( pFilter[0]->predictionGain32, L_shl( pTnsParameters[0]->minPredictionGain, 16 ) ) && LT_32( sts[0]->element_brate, IVAS_80k ) &&
		GT_32( pFilter[1]->predictionGain32, L_shl( pTnsParameters[1]->minPredictionGain, 16 ) ) && EQ_16( sts[0]->hTcxEnc->tnsData[k].nFilters, sts[1]->hTcxEnc->tnsData[k].nFilters ) )
		IF( flag && LT_32( sts[0]->element_brate, IVAS_80k ) &&
		flag_1 && EQ_16( sts[0]->hTcxEnc->tnsData[k].nFilters, sts[1]->hTcxEnc->tnsData[k].nFilters ) )
		{
		pFilter[0]->predictionGain32 = pFilter[1]->predictionGain32 = meanPredictionGain_fx; /* more TNS filter sync at 48kbps */
		move32();
		move32();
		pFilter[0]->predictionGain_e = pFilter[1]->predictionGain_e = meanPredictionGain_e; /* more TNS filter sync at 48kbps */
		move16();
		move16();
		pFilter[0]->predictionGain = pFilter[1]->predictionGain = shl_sat( extract_h( meanPredictionGain_fx ), sub( meanPredictionGain_e, PRED_GAIN_E ) ); /* Q7 */
		move16();
		move16();
		}
		flag = BASOP_Util_Cmp_Mant32Exp( Mpy_32_16_1( meanPredictionGain_fx, SIMILAR_TNS_THRESHOLD_FX_IN_Q15 ), meanPredictionGain_e, L_abs( BASOP_Util_Add_Mant32Exp( pFilter[0]->predictionGain32, pFilter[0]->predictionGain_e, L_negate( pFilter[1]->predictionGain32 ), pFilter[1]->predictionGain_e, &sum_e ) ), sum_e );
		if ( flag < 0 )
		{
		flag = 0;
		move16();
		}
		test();
		IF( LT_32( L_abs( L_sub( pFilter[0]->predictionGain32, pFilter[1]->predictionGain32 ) ), Mpy_32_16_1( meanPredictionGain_fx, SIMILAR_TNS_THRESHOLD_FX_IN_Q15 ) ) &&
		IF( flag &&
		( EQ_16( sts[0]->hTcxEnc->tnsData[k].nFilters, sts[1]->hTcxEnc->tnsData[k].nFilters ) ) )
		{

		Word16 maxAvgSqrCoef_fx = s_max( pFilter[0]->avgSqrCoef, pFilter[1]->avgSqrCoef ); // Q15
		Word16 meanLtpGain_fx = add( shr( sts[0]->hTcxEnc->tcxltp_gain, 1 ), shr( sts[1]->hTcxEnc->tcxltp_gain, 1 ) );
		maxPredGain_fx = L_max( maxPredGain_fx, meanPredictionGain_fx );
		// maxPredGain_fx = L_max( maxPredGain_fx, meanPredictionGain_fx );
		flag = BASOP_Util_Cmp_Mant32Exp( maxPredGain_fx, maxPredGain_e, meanPredictionGain_fx, meanPredictionGain_e );
		IF( flag < 0 )
		{
		maxPredGain_fx = meanPredictionGain_fx;
		maxPredGain_e = meanPredictionGain_e;
		move32();
		move16();
		}
		flag = BASOP_Util_Cmp_Mant32Exp( meanPredictionGain_fx, meanPredictionGain_e, L_deposit_h( pTnsParameters[0]->minPredictionGain ), PRED_GAIN_E );
		if ( flag < 0 )
		{
		flag = 0;
		move16();
		}
		test();
		test();
		IF( GT_32( meanPredictionGain_fx, L_shl( pTnsParameters[0]->minPredictionGain, 16 ) ) \|\| GT_16( maxAvgSqrCoef_fx, pTnsParameters[0]->minAvgSqrCoef ) )
		IF( flag \|\| GT_16( maxAvgSqrCoef_fx, pTnsParameters[0]->minAvgSqrCoef ) )
		{
		test();
		test();
		@@ -454,10 +503,16 @@ void TNSAnalysisStereo_fx(
		move16();
		}
		}
		Word16 flag = BASOP_Util_Cmp_Mant32Exp( TNS_GAIN_THRESHOLD_FOR_WHITE_FX_IN_Q23, PRED_GAIN_E, maxPredGain_fx, maxPredGain_e );
		if ( flag < 0 )
		{
		flag = 0;
		move16();
		}
		test();
		test();
		test();
		IF( !bWhitenedDomain && individual_decision[k] == 0 && LT_32( maxPredGain_fx, TNS_GAIN_THRESHOLD_FOR_WHITE_FX_IN_Q23 ) && NE_16( sts[0]->hTcxEnc->transform_type[k], TCX_5 ) )
		IF( !bWhitenedDomain && individual_decision[k] == 0 && flag && NE_16( sts[0]->hTcxEnc->transform_type[k], TCX_5 ) )
		{
		sts[0]->hTcxEnc->bTnsOnWhithenedSpectra[k] = 1;
		move16();
		@@ -477,7 +532,15 @@ void TNSAnalysisStereo_fx(
		ClearTnsFilterCoefficients( sts[1]->hTcxEnc->tnsData[k].filter + iFilter );
		}
		}
		maxPredictionGain_fx = L_max( maxPredictionGain_fx, maxPredGain_fx );
		// maxPredictionGain_fx = L_max( maxPredictionGain_fx, maxPredGain_fx );
		flag = BASOP_Util_Cmp_Mant32Exp( maxPredictionGain_fx, maxPredictionGain_e, maxPredGain_fx, maxPredGain_e );
		IF( flag < 0 )
		{
		maxPredictionGain_fx = maxPredGain_fx;
		maxPredictionGain_e = maxPredGain_e;
		move32();
		move16();
		}
		}
		}
		}
		@@ -515,7 +578,9 @@ void TNSAnalysisStereo_fx(
		IF( sts[ch]->hTcxCfg->fIsTNSAllowed && ( individual_decision[k] \|\| mct_on ) &&
		( !bWhitenedDomain \|\| sts[ch]->hTcxEnc->bTnsOnWhithenedSpectra[k] ) )
		{
		Word32 maxPredGain_fx = -ONE_IN_Q23; // Q23
		Word32 maxPredGain_fx = -ONE_IN_Q31; // Q31
		move32();
		Word16 maxPredGain_e = 0;
		move16();
		sts[ch]->hTcxCfg->pCurrentTnsConfig = &sts[ch]->hTcxCfg->tnsConfig[sts[ch]->hTcxEnc->transform_type[k] == TCX_20][( k == 0 ) && ( sts[ch]->last_core == ACELP_CORE )];

		@@ -526,9 +591,23 @@ void TNSAnalysisStereo_fx(
		pFilter = sts[ch]->hTcxEnc->tnsData[k].filter + iFilter;
		pTnsParameters = sts[ch]->hTcxCfg->pCurrentTnsConfig->pTnsParameters + iFilter;

		maxPredGain_fx = L_max( maxPredGain_fx, pFilter->predictionGain32 );
		// maxPredGain_fx = L_max( maxPredGain_fx, pFilter->predictionGain32 );
		Word16 flag = BASOP_Util_Cmp_Mant32Exp( maxPredGain_fx, maxPredGain_e, pFilter->predictionGain32, pFilter->predictionGain_e );
		IF ( flag < 0 )
		{
		maxPredGain_fx = pFilter->predictionGain32;
		move32();
		maxPredGain_e = pFilter->predictionGain_e;
		move16();
		}
		flag = BASOP_Util_Cmp_Mant32Exp( pFilter->predictionGain32, pFilter->predictionGain_e, L_deposit_h( pTnsParameters->minPredictionGain ), PRED_GAIN_E );
		if ( flag < 0 )
		{
		flag = 0;
		move16();
		}
		test();
		IF( GT_32( pFilter->predictionGain32, L_shl( pTnsParameters->minPredictionGain, 16 ) ) \|\| GT_16( pFilter->avgSqrCoef, pTnsParameters->minAvgSqrCoef ) )
		IF( flag \|\| GT_16( pFilter->avgSqrCoef, pTnsParameters->minAvgSqrCoef ) )
		{
		test();
		test();
		@@ -600,9 +679,15 @@ void TNSAnalysisStereo_fx(
		sts[ch]->hTcxEnc->fUseTns[k] = 0;
		}
		move16();
		Word16 flag = BASOP_Util_Cmp_Mant32Exp( TNS_GAIN_THRESHOLD_FOR_WHITE_FX_IN_Q23, PRED_GAIN_E, maxPredGain_fx, maxPredGain_e );
		if ( flag < 0 )
		{
		flag = 0;
		move16();
		}
		test();
		test();
		IF( !bWhitenedDomain && LT_32( maxPredGain_fx, TNS_GAIN_THRESHOLD_FOR_WHITE_FX_IN_Q23 ) && NE_16( sts[ch]->hTcxEnc->transform_type[k], TCX_5 ) )
		IF( !bWhitenedDomain && flag && NE_16( sts[ch]->hTcxEnc->transform_type[k], TCX_5 ) )
		{
		sts[ch]->hTcxEnc->fUseTns[k] = 0;
		move16();
		@@ -617,7 +702,15 @@ void TNSAnalysisStereo_fx(
		move16();
		}
		}
		maxPredictionGain_fx = L_max( maxPredictionGain_fx, maxPredGain_fx );
		// maxPredictionGain_fx = L_max( maxPredictionGain_fx, maxPredGain_fx );
		flag = BASOP_Util_Cmp_Mant32Exp( maxPredictionGain_fx, maxPredictionGain_e, maxPredGain_fx, maxPredGain_e );
		IF( flag < 0 )
		{
		maxPredictionGain_fx = maxPredGain_fx;
		maxPredictionGain_e = maxPredGain_e;
		move32();
		move16();
		}
		}
		}
		}

lib_enc/ext_sig_ana_fx.c

+4 −0

Original line number	Diff line number	Diff line
		@@ -1372,7 +1372,11 @@ void core_signal_analysis_high_bitrate_ivas_fx(
		IF( st->igf )
		{
		Word16 q_spectrum = sub( Q31, hTcxEnc->spectrum_e[frameno] );
		#ifndef MSAN_FIX
		ProcessIGF_ivas_fx( st, hTcxEnc->spectrum_fx[frameno], hTcxEnc->spectrum_fx[frameno], &q_spectrum, powerSpec, powerSpec_e, transform_type[frameno] == TCX_20, frameno, 0, vad_hover_flag );
		#else
		ProcessIGF_ivas_fx( st, N_MAX + L_MDCT_OVLP_MAX, hTcxEnc->spectrum_fx[frameno], hTcxEnc->spectrum_fx[frameno], &q_spectrum, powerSpec, powerSpec_e, transform_type[frameno] == TCX_20, frameno, 0, vad_hover_flag );
		#endif
		}
		}
		}

lib_enc/igf_enc.c

+12 −1

Original line number	Diff line number	Diff line
		@@ -2578,6 +2578,9 @@ Word16 IGFEncWriteConcatenatedBitstream(

		void IGFEncApplyMono_ivas_fx(
		Encoder_State st, / i : Encoder state */
		#ifdef MSAN_FIX
		Word16 powerSpectrum_len, /* i: length of pPowerSpectrum_fx buffer */
		#endif
		const Word16 igfGridIdx, /* i : IGF grid index */
		Word32 pMDCTSpectrum_fx, / i/o: MDCT spectrum */
		Word16 e_mdct, /* i : exponent of pMDCTspectrum */
		@@ -2658,12 +2661,20 @@ void IGFEncApplyMono_ivas_fx(

		IF( pPowerSpectrumParameter_fx )
		{
		#ifndef MSAN_FIX
		FOR( Word16 i = 0; i < N_MAX + L_MDCT_OVLP_MAX; i++ )
		#else
		FOR( Word16 i = 0; i < powerSpectrum_len; i++ )
		#endif
		{
		common_pPowerSpectrum_exp = s_max( common_pPowerSpectrum_exp, pPowerSpectrumParameter_exp[i] );
		}

		#ifndef MSAN_FIX
		FOR( Word16 i = 0; i < N_MAX + L_MDCT_OVLP_MAX; i++ )
		#else
		FOR( Word16 i = 0; i < powerSpectrum_len; i++ )
		#endif
		{
		common_pPowerSpectrum_fx[i] = L_shl( pPowerSpectrumParameter_fx[i], sub( pPowerSpectrumParameter_exp[i], common_pPowerSpectrum_exp ) );
		move16();