Commit 8ca59b2d authored by Sandesh Venkatesh's avatar Sandesh Venkatesh
Browse files

Merge branch 'basop_instrumentation_dec_renderer_6' into 'main' 

BASOP and instrumentation changes for lib_com, lib_rend, lib_dec

See merge request !436
parents e2806f2a ffbb3cc3

lib_com/fd_cng_com.c

+49 −31
Original line number Diff line number Diff line
@@ -961,56 +961,63 @@ void SynthesisSTFT_fx( 
    Word32 *timeDomainOutput,

    Word32 *olapBuffer,

    const Word16 *olapWin,

    const int16_t tcx_transition,

    const Word16 tcx_transition,

    HANDLE_FD_CNG_COM hFdCngCom, /* i/o: FD_CNG structure containing all buffers and variables */

    const int16_t element_mode,  /* i  : element mode */

    const int16_t nchan_out      /* i  : number of output channels */

    const Word16 element_mode,   /* i  : element mode */

    const Word16 nchan_out       /* i  : number of output channels */

)

{

    int16_t i;

    Word16 i;

    Word32 buf_fx[M + 1 + 320], tmp_fx;



    /* Perform IFFT */

    RFFTN_fx( fftBuffer, hFdCngCom->fftSineTab_fx, hFdCngCom->fftlen, 1 );



    /* Handle overlap in P/S domain for stereo */

    IF( ( element_mode == IVAS_CPE_TD || element_mode == IVAS_CPE_DFT ) && nchan_out == 2 )

    test();

    test();

    IF( ( EQ_16( element_mode, IVAS_CPE_TD ) || EQ_16( element_mode, IVAS_CPE_DFT ) ) && EQ_16( nchan_out, 2 ) )

    {

        mvl2l( olapBuffer + 3 * hFdCngCom->frameSize / 4 - ( M + 1 ), buf_fx, hFdCngCom->frameSize + M + 1 );

        set_l( olapBuffer, 0, hFdCngCom->fftlen );

        Copy32( olapBuffer + sub( i_mult( 3, shr( hFdCngCom->frameSize, 2 ) ), ( M + 1 ) ), buf_fx, add( hFdCngCom->frameSize, M + 1 ) );

        set32_fx( olapBuffer, 0, hFdCngCom->fftlen );

    }

    ELSE

    {

        mvl2l( olapBuffer + hFdCngCom->frameSize, olapBuffer, hFdCngCom->frameSize );

        set_l( olapBuffer + hFdCngCom->frameSize, 0, hFdCngCom->frameSize ); /*olapBuffer, fftBuffer, olapWin*/

        Copy32( olapBuffer + hFdCngCom->frameSize, olapBuffer, hFdCngCom->frameSize );

        set32_fx( olapBuffer + hFdCngCom->frameSize, 0, hFdCngCom->frameSize ); /*olapBuffer, fftBuffer, olapWin*/

    }



    IF( tcx_transition )

    {

        FOR( i = 0; i < 5 * hFdCngCom->frameSize / 4; i++ )

        FOR( i = 0; i < i_mult( 5, shr( hFdCngCom->frameSize, 2 ) ); i++ )

        {

            olapBuffer[i] = fftBuffer[i];

            move32();

        }

    }

    ELSE

    {

        FOR( i = hFdCngCom->frameSize / 4; i < 3 * hFdCngCom->frameSize / 4; i++ )

        FOR( i = hFdCngCom->frameSize / 4; i < i_mult( 3, shr( hFdCngCom->frameSize, 2 ) ); i++ )

        {

            olapBuffer[i] = L_add( olapBuffer[i], Mpy_32_16_1( fftBuffer[i], olapWin[i - hFdCngCom->frameSize / 4] ) );

            olapBuffer[i] = L_add( olapBuffer[i], Mpy_32_16_1( fftBuffer[i], olapWin[sub( i, shr( hFdCngCom->frameSize, 2 ) )] ) );

            move32();

        }

        FOR( ; i < 5 * hFdCngCom->frameSize / 4; i++ )

        FOR( ; i < i_mult( 5, shr( hFdCngCom->frameSize, 2 ) ); i++ )

        {

            olapBuffer[i] = fftBuffer[i];

            move32();

        }

    }

    FOR( ; i < 7 * hFdCngCom->frameSize / 4; i++ )

    FOR( ; i < i_mult( 7, shr( hFdCngCom->frameSize, 2 ) ); i++ )

    {

        olapBuffer[i] = Mpy_32_16_1( fftBuffer[i], olapWin[i - 3 * hFdCngCom->frameSize / 4] );

        olapBuffer[i] = Mpy_32_16_1( fftBuffer[i], olapWin[sub( i, i_mult( 3, shr( hFdCngCom->frameSize, 2 ) ) )] );

        move32();

    }



    FOR( ; i < hFdCngCom->fftlen; i++ )

    {

        olapBuffer[i] = 0;

        move32();

    }



    Word32 fftScale = 0;
@@ -1018,10 +1025,12 @@ void SynthesisSTFT_fx( 
    {

        case 640:

            fftScale = FFT_SCALING_640;

            break;

            move32();

            BREAK;

        case 512:

            fftScale = FFT_SCALING_512;

            break;

            move32();

            BREAK;

        default:

            assert( !"Not supported FFT length!" );

    }
@@ -1029,32 +1038,36 @@ void SynthesisSTFT_fx( 
    // v_multc(olapBuffer + hFdCngCom->frameSize / 4, (float)(hFdCngCom->fftlen / 2), timeDomainOutput, hFdCngCom->frameSize);

    v_multc_fixed( olapBuffer + hFdCngCom->frameSize / 4, fftScale, timeDomainOutput, hFdCngCom->frameSize ); // Q_in - 9

                                                                                                              /* Get excitation */

    IF( ( element_mode == IVAS_CPE_TD || element_mode == IVAS_CPE_DFT ) && nchan_out == 2 )

    test();

    test();

    IF( ( EQ_16( element_mode, IVAS_CPE_TD ) || EQ_16( element_mode, IVAS_CPE_DFT ) ) && EQ_16( nchan_out, 2 ) )

    {

        FOR( i = 0; i < hFdCngCom->frameSize / 2; i++ )

        {

            buf_fx[i + ( M + 1 )] = L_add( buf_fx[i + ( M + 1 )], olapBuffer[i + hFdCngCom->frameSize / 4] );

            buf_fx[i + ( M + 1 )] = L_add( buf_fx[i + ( M + 1 )], olapBuffer[add( i, shr( hFdCngCom->frameSize, 2 ) )] );

            move32();

        }

        // v_multc(buf, (float)(hFdCngCom->fftlen / 2), buf, M + 1 + hFdCngCom->frameSize);

        v_multc_fixed( buf_fx, fftScale, buf_fx, M + 1 + hFdCngCom->frameSize );

        v_multc_fixed( buf_fx, fftScale, buf_fx, add( M + 1, hFdCngCom->frameSize ) );

    }

    ELSE

    {

        // v_multc(olapBuffer + hFdCngCom->frameSize / 4 - (M + 1), (float)(hFdCngCom->fftlen / 2), buf, M + 1 + hFdCngCom->frameSize);

        v_multc_fixed( olapBuffer + ( hFdCngCom->frameSize / 4 ) - ( M + 1 ), fftScale, buf_fx, M + 1 + hFdCngCom->frameSize );

        v_multc_fixed( olapBuffer + sub( shr( hFdCngCom->frameSize, 2 ), ( M + 1 ) ), fftScale, buf_fx, add( M + 1, hFdCngCom->frameSize ) );

    }



    tmp_fx = buf_fx[0];

    move32();

    // preemph(buf + 1, PREEMPH_FAC_FLT, M + hFdCngCom->frameSize, &tmp);

    preemph_ivas_fx( buf_fx + 1, PREEMPH_FAC, M + hFdCngCom->frameSize, &tmp_fx );

    preemph_ivas_fx( buf_fx + 1, PREEMPH_FAC, add( M, hFdCngCom->frameSize ), &tmp_fx );

    // residu(hFdCngCom->A_cng_flt, M, buf + 1 + M, hFdCngCom->exc_cng_flt, hFdCngCom->frameSize);



    // floatToFixed_arr( hFdCngCom->A_cng_flt, hFdCngCom->A_cng, Q13, M + 1 );

    // residu_ivas_fx( hFdCngCom->A_cng, Q13, M, buf_fx + 1 + M, hFdCngCom->exc_cng_32fx, hFdCngCom->frameSize );

    residu_ivas_fx( hFdCngCom->A_cng, ( 15 - norm_s( hFdCngCom->A_cng[0] - 1 ) ), M, buf_fx + 1 + M, hFdCngCom->exc_cng_32fx, hFdCngCom->frameSize );

    residu_ivas_fx( hFdCngCom->A_cng, sub( 15, norm_s( hFdCngCom->A_cng[0] - 1 ) ), M, buf_fx + 1 + M, hFdCngCom->exc_cng_32fx, hFdCngCom->frameSize );

    for ( i = 0; i < hFdCngCom->frameSize; i++ )

    {

        hFdCngCom->exc_cng_flt[i] = fix_to_float( hFdCngCom->exc_cng_32fx[i], Q_in - 9 );

        hFdCngCom->exc_cng_flt[i] = fix_to_float( hFdCngCom->exc_cng_32fx[i], sub( Q_in, 9 ) );

    }



    return;
@@ -1149,7 +1162,8 @@ void SynthesisSTFT_dirac_fx( 


    FOR( i = shr( hFdCngCom->frameSize, 2 ); i < i_mult( 3, shr( hFdCngCom->frameSize, 2 ) ); i++ )

    {

        olapBuffer[i] = L_add( olapBuffer[i], Mpy_32_16_1( fftBuffer[i], olapWin[i - hFdCngCom->frameSize / 4] ) );

        olapBuffer[i] = L_add( olapBuffer[i], Mpy_32_16_1( fftBuffer[i], olapWin[sub( i, shr( hFdCngCom->frameSize, 2 ) )] ) );

        move32();

    }

    FOR( ; i < i_mult( 5, shr( hFdCngCom->frameSize, 2 ) ); i++ )

    {
@@ -1186,14 +1200,15 @@ void SynthesisSTFT_dirac_fx( 
    }



    /* Get time-domain signal */

    v_multc_fixed( olapBuffer + hFdCngCom->frameSize / 4, fftScale, timeDomainOutput, samples_out ); // Q_in - 9

    v_multc_fixed( olapBuffer + shr( hFdCngCom->frameSize, 2 ), fftScale, timeDomainOutput, samples_out ); // Q_in - 9



    /* Get excitation */

    v_multc_fixed( olapBuffer + ( hFdCngCom->frameSize / 4 ) - ( M + 1 ), fftScale, buf, M + 1 + hFdCngCom->frameSize );

    v_multc_fixed( olapBuffer + sub( shr( hFdCngCom->frameSize, 2 ), ( M + 1 ) ), fftScale, buf, add( M + 1, hFdCngCom->frameSize ) );

    tmp = buf[0];

    move32();

    preemph_ivas_fx( buf + 1, PREEMPH_FAC, M + hFdCngCom->frameSize, &tmp );

    // residu_ivas_fx( hFdCngCom->A_cng, Q13, M, buf + 1 + M, hFdCngCom->exc_cng_32fx, hFdCngCom->frameSize );

    residu_ivas_fx( hFdCngCom->A_cng, 15 - norm_s( hFdCngCom->A_cng[0] - 1 ), M, buf + 1 + M, hFdCngCom->exc_cng_32fx, hFdCngCom->frameSize );

    residu_ivas_fx( hFdCngCom->A_cng, sub( 15, norm_s( hFdCngCom->A_cng[0] - 1 ) ), M, buf + 1 + M, hFdCngCom->exc_cng_32fx, hFdCngCom->frameSize );



    /* update and window olapBuf if we have a output frame that is shorter than the default frame size...*/

    IF( LT_16( samples_out, hFdCngCom->frameSize ) )
@@ -1202,7 +1217,8 @@ void SynthesisSTFT_dirac_fx( 
    }

    FOR( i = i_mult( 5, shr( hFdCngCom->frameSize, 2 ) ); i < i_mult( 7, shr( hFdCngCom->frameSize, 2 ) ); i++ )

    {

        olapBuffer[i] = Mpy_32_16_1( olapBuffer[i], olapWin[i - 3 * hFdCngCom->frameSize / 4] );

        olapBuffer[i] = Mpy_32_16_1( olapBuffer[i], olapWin[sub( i, i_mult( 3, shr( hFdCngCom->frameSize, 2 ) ) )] );

        move32();

    }



    return;
@@ -1286,9 +1302,10 @@ Word32 rand_gauss_fx( 
    temp = own_random( seed );

    temp = L_add( temp, own_random( seed ) );

    temp = L_add( temp, own_random( seed ) );

    temp = L_shr( temp, 15 - q );

    temp = L_shr( temp, sub( 15, q ) );



    *x = temp;

    move32();



    return temp;

}
@@ -1311,6 +1328,7 @@ Word16 rand_gauss_fix( 
    temp = L_add( temp, shr( own_random( seed ), Q2 ) );



    *x = (Word16) temp;

    move32();



    return (Word16) temp;

}

lib_com/fd_cng_com_fx.c

+37 −3
Original line number Diff line number Diff line
@@ -395,10 +395,12 @@ void compress_range( 
            if ( in_s != 0 )

            {

                out[i] = extract_h( L_tmp );

                move16();

            }

            if ( out[i] == 0 )

            {

                out[i] = 1;

                move16();

            }

        }

    }
@@ -417,10 +419,12 @@ void compress_range( 
            if ( in[i] != 0 )

            {

                out[i] = extract_h( L_tmp );

                move16();

            }

            if ( out[i] == 0 )

            {

                out[i] = 1;

                move16();

            }

        }

    }
@@ -523,7 +527,7 @@ void expand_range_var_exp( 
        move32();



        Word32 tmp_low_lim = L_shr( low_lim, maxOutExp );

        IF( LT_32( out[i], tmp_low_lim ) )

        if ( LT_32( out[i], tmp_low_lim ) )

        {

            out[i] = tmp_low_lim;

            move32();
@@ -603,6 +607,7 @@ void minimum_statistics( 
    Word16 msNoiseFloor16;

#ifdef BASOP_NOGLOB_DECLARE_LOCAL

    Flag Overflow = 0;

    move32();

#endif
@@ -765,7 +770,7 @@ void minimum_statistics( 


            FOR( j = start; j < stop; j++ )

            {

                /* Compute optimal smoothing parameter for PSD estimation */ test();

                /* Compute optimal smoothing parameter for PSD estimation */

                test();

                IF( ( scalar == 0 ) || ( msNoiseFloor[j] == 0 ) )

                {
@@ -827,6 +832,7 @@ void minimum_statistics( 
                /* Compute the PSD (smoothed periodogram) in each band */

                msPsd[j] = round_fx( L_add( Mpy_32_16_1( msAlpha[j], msPsd[j] ),

                                            Mpy_32_16_1( L_sub( 2147483647l /*1.0 Q31*/, msAlpha[j] ), msPeriodog[j] ) ) );

                move16();

            }

            msPsdSum[cnt] = dotp_s_fx( msPsd + start, psize + start, current_len, CNG_HS );

            move32();
@@ -1077,6 +1083,7 @@ void minimum_statistics( 
                FOR( j = 0; j < len; j++ )

                {

                    msCurrentMinOut[j] = L_deposit_h( msPeriodog[j] );

                    move32();

                }

                set32_fx( hFdCngCom->msAlphaCor, 2147483647l /*1.0 Q31*/, cnt );

                set32_fx( msAlpha, 0l /*0.0 Q31*/, len );
@@ -1123,6 +1130,7 @@ void minimum_statistics( 
        FOR( j = 0; j < len; j++ )

        {

            msNoiseEst[j] = round_fx( L_mac( L_mult( 31130 /*0.95 Q15*/, msNoiseEst[j] ), 1638 /*0.05 Q15*/, msNoiseFloor[j] ) );

            move16();

        }

    }

#ifdef IVAS_CODE_CNG_COM
@@ -1221,6 +1229,7 @@ void minimum_statistics_fx( 
    Word16 msNoiseFloor16;

#ifdef BASOP_NOGLOB_DECLARE_LOCAL

    Flag Overflow = 0;

    move32();

#endif
@@ -1379,7 +1388,7 @@ void minimum_statistics_fx( 


            FOR( j = start; j < stop; j++ )

            {

                /* Compute optimal smoothing parameter for PSD estimation */ test();

                /* Compute optimal smoothing parameter for PSD estimation */

                test();

                IF( ( scalar == 0 ) || ( msNoiseFloor[j] == 0 ) )

                {
@@ -1441,6 +1450,7 @@ void minimum_statistics_fx( 
                /* Compute the PSD (smoothed periodogram) in each band */

                msPsd[j] = round_fx( L_add( Mpy_32_16_1( msAlpha[j], msPsd[j] ),

                                            Mpy_32_16_1( L_sub( 2147483647l /*1.0 Q31*/, msAlpha[j] ), msPeriodog[j] ) ) );

                move16();

            }

            msPsdSum[cnt] = dotp_s_fx( msPsd + start, psize + start, current_len, CNG_HS );

            move32();
@@ -1691,6 +1701,7 @@ void minimum_statistics_fx( 
                FOR( j = 0; j < len; j++ )

                {

                    msCurrentMinOut[j] = L_deposit_h( msPeriodog[j] );

                    move32();

                }

                set32_fx( hFdCngCom->msAlphaCor, 2147483647l /*1.0 Q31*/, cnt );

                set32_fx( msAlpha, 0l /*0.0 Q31*/, len );
@@ -1731,12 +1742,14 @@ void minimum_statistics_fx( 
        ELSE

        {

            hFdCngCom->msFrCnt = add( hFdCngCom->msFrCnt, 1 );

            move16();

        }



        /* Smooth noise estimate during CNG phases */

        FOR( j = 0; j < len; j++ )

        {

            msNoiseEst[j] = round_fx( L_mac( L_mult( 31130 /*0.95 Q15*/, msNoiseEst[j] ), 1638 /*0.05 Q15*/, msNoiseFloor[j] ) );

            move16();

        }

    }

    if ( enc_dec == DEC && element_mode == IVAS_CPE_TD )
@@ -1804,6 +1817,7 @@ void apply_scale( 


    {

        *scale = L_add( *scale, L_deposit_h( scaleTable[i].scale ) );

        move32();

    }

}
@@ -1837,6 +1851,7 @@ Word16 apply_scale_ind( 


    {

        *scale = L_add( *scale, L_deposit_h( scaleTable[i].scale ) );

        move32();

    }

    return i;

}
@@ -1864,7 +1879,9 @@ void apply_scale_ivas_fx( 
    assert( i < scaleTableSize );



    *scale = L_add( *scale, L_deposit_h( scaleTable[i].scale ) );

    move32();

    *index = i;

    move16();

}

#endif // IVAS_FLOAT_FIXED

/*-------------------------------------------------------------------
@@ -1961,6 +1978,7 @@ void scalebands( 
    delta = 0;

    move16();

    partpowLD64M1 = 0L; /* to avoid compilation warnings */

    move32();



    /* Interpolate the bin/band-wise levels from the partition levels */

    IF( EQ_16( nband, npart ) )
@@ -2102,6 +2120,7 @@ void scalebands_fx( 
    delta = 0;

    move16();

    partpowLD64M1 = 0L; /* to avoid compilation warnings */

    move32();



    /* Interpolate the bin/band-wise levels from the partition levels */

    IF( EQ_16( nband, npart ) )
@@ -2241,6 +2260,7 @@ static void getmidbands( 
    Word16 j, max_psize, shift;

#ifdef BASOP_NOGLOB_DECLARE_LOCAL

    Flag Overflow = 0;

    move32();

#endif
@@ -2248,6 +2268,7 @@ static void getmidbands( 
    move16();

    /* first half partition */ move16();

    midband[0] = part[0];

    move16();

    psize[0] = add( part[0], 1 );

    move16();

    psize_inv[0] = getNormReciprocalWord16( psize[0] );
@@ -2401,8 +2422,10 @@ void AnalysisSTFT_fx( 
    FOR( i = 0; i < hFdCngCom->fftlen; i++ )

    {

        fftBuffer[i] = L_shr( fftBuffer[i], 11 );

        move32();

    }

    *fftBuffer_exp = WORD16_BITS + 11;

    move16();



    /* Perform FFT */

    RFFTN_fx( fftBuffer, hFdCngCom->fftSineTab_fx, hFdCngCom->fftlen, -1 );
@@ -2439,9 +2462,11 @@ void SynthesisSTFT( 


    /* Perform IFFT */

    scale = 0;

    move16();

    BASOP_rfft( fftBuffer, hFdCngCom->fftlen, &scale, 1 );

    fftBufferExp = add( fftBufferExp, scale );

    hFdCngCom->fftBuffer_exp = fftBufferExp;

    move16();



    fftBufferExp = add( fftBufferExp, hFdCngCom->fftlenShift );
@@ -2466,6 +2491,7 @@ void SynthesisSTFT( 
        FOR( i = 0; i < len; i++ )

        {

            olapBuffer[i] = round_fx( L_shl( fftBuffer[i], fftBufferExp - 15 ) );

            move16();

        }

    }

    ELSE
@@ -2685,6 +2711,7 @@ void mhvals( 
        tmp2_m = BASOP_Util_Add_Mant32Exp( tmp2_m, tmp2_e, L_deposit_h( m_array[i] ), 0, &tmp2_e );

        assert( tmp2_e == 0 );

        *m = extract_h( tmp2_m );

        move32();

    }

}
@@ -2735,9 +2762,12 @@ void lpc_from_spectrum( 


    Word32 *powspec = hFdCngCom->cngNoiseLevel; /*i  : pointer to noise levels format Q5.27*/

    Word16 powspec_exp = hFdCngCom->cngNoiseLevelExp;

    move16();

    Word16 fftlen = hFdCngCom->fftlen; /*i  : size of fft*/

    Word16 *A = hFdCngCom->A_cng;      /*o  : lpc coefficients format Q3.12*/

    move16();

    Word16 lpcorder = M;

    move16();



    scale = 0;

    move16();
@@ -2767,6 +2797,7 @@ void lpc_from_spectrum( 
        *ptr = nf;

        move32();

        *pti = L_deposit_l( 0 );

        move32();

        ptr += 2;

        pti += 2;

    }
@@ -2776,6 +2807,7 @@ void lpc_from_spectrum( 
        *ptr = L_max( nf, L_shl( powspec[i - start], s1 ) );

        move32();

        *pti = L_deposit_l( 0 );

        move32();

        ptr += 2;

        pti += 2;

    }
@@ -2785,6 +2817,7 @@ void lpc_from_spectrum( 
        *ptr = nf;

        move32();

        *pti = L_deposit_l( 0 );

        move32();

        ptr += 2;

        pti += 2;

    }
@@ -2960,6 +2993,7 @@ void FdCng_exc( 
    Word16 i;



    *CNG_mode = -1;

    move16();



    FOR( i = 0; i < L_frame / L_SUBFR; i++ )

    {

lib_com/fft.c

+36 −36
Original line number Diff line number Diff line
@@ -6513,21 +6513,21 @@ static void BASOP_fft8( 
    move32();

    im[s * 6] = L_add( s05, s07 );

    move32();

    re[s * 3] = L_add( s08, s14 );

    re[i_mult( s, 3 )] = L_add( s08, s14 );

    move32();

    re[s * 7] = L_sub( s08, s14 );

    re[i_mult( s, 7 )] = L_sub( s08, s14 );

    move32();

    im[s * 3] = L_add( s09, s15 );

    im[i_mult( s, 3 )] = L_add( s09, s15 );

    move32();

    im[s * 7] = L_sub( s09, s15 );

    im[i_mult( s, 7 )] = L_sub( s09, s15 );

    move32();

    re[s * 1] = L_add( s10, s12 );

    move32();

    re[s * 5] = L_sub( s10, s12 );

    re[i_mult( s, 5 )] = L_sub( s10, s12 );

    move32();

    im[s * 1] = L_add( s11, s13 );

    move32();

    im[s * 5] = L_sub( s11, s13 );

    im[i_mult( s, 5 )] = L_sub( s11, s13 );

    move32();



    return;
@@ -6581,23 +6581,23 @@ static void BASOP_fftN2( 


        IF( i == 0 )

        {

            cplxMpy4_8_1( x02, x03, x[2 * i + 2 * 1 * dim1], x[2 * i + 2 * 1 * dim1 + 1] );

            cplxMpy4_8_1( x04, x05, x[2 * i + 2 * 2 * dim1], x[2 * i + 2 * 2 * dim1 + 1] );

            cplxMpy4_8_1( x06, x07, x[2 * i + 2 * 3 * dim1], x[2 * i + 2 * 3 * dim1 + 1] );

            cplxMpy4_8_1( x08, x09, x[2 * i + 2 * 4 * dim1], x[2 * i + 2 * 4 * dim1 + 1] );

            cplxMpy4_8_1( x10, x11, x[2 * i + 2 * 5 * dim1], x[2 * i + 2 * 5 * dim1 + 1] );

            cplxMpy4_8_1( x12, x13, x[2 * i + 2 * 6 * dim1], x[2 * i + 2 * 6 * dim1 + 1] );

            cplxMpy4_8_1( x14, x15, x[2 * i + 2 * 7 * dim1], x[2 * i + 2 * 7 * dim1 + 1] );

            cplxMpy4_8_1( x02, x03, x[add( shl( i, 1 ), i_mult( 2 * 1, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 1, dim1 ) ), 1 )] );

            cplxMpy4_8_1( x04, x05, x[add( shl( i, 1 ), i_mult( 2 * 2, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 2, dim1 ) ), 1 )] );

            cplxMpy4_8_1( x06, x07, x[add( shl( i, 1 ), i_mult( 2 * 3, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 3, dim1 ) ), 1 )] );

            cplxMpy4_8_1( x08, x09, x[add( shl( i, 1 ), i_mult( 2 * 4, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 4, dim1 ) ), 1 )] );

            cplxMpy4_8_1( x10, x11, x[add( shl( i, 1 ), i_mult( 2 * 5, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 5, dim1 ) ), 1 )] );

            cplxMpy4_8_1( x12, x13, x[add( shl( i, 1 ), i_mult( 2 * 6, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 6, dim1 ) ), 1 )] );

            cplxMpy4_8_1( x14, x15, x[add( shl( i, 1 ), i_mult( 2 * 7, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 7, dim1 ) ), 1 )] );

        }

        ELSE

        {

            cplxMpy4_8_0( x02, x03, x[2 * i + 2 * 1 * dim1], x[2 * i + 2 * 1 * dim1 + 1], W[sc * i + sc * 1 * dim1 - Woff], W[sc * i + sc * 1 * dim1 + 1 - Woff] );

            cplxMpy4_8_0( x04, x05, x[2 * i + 2 * 2 * dim1], x[2 * i + 2 * 2 * dim1 + 1], W[sc * i + sc * 2 * dim1 - Woff], W[sc * i + sc * 2 * dim1 + 1 - Woff] );

            cplxMpy4_8_0( x06, x07, x[2 * i + 2 * 3 * dim1], x[2 * i + 2 * 3 * dim1 + 1], W[sc * i + sc * 3 * dim1 - Woff], W[sc * i + sc * 3 * dim1 + 1 - Woff] );

            cplxMpy4_8_0( x08, x09, x[2 * i + 2 * 4 * dim1], x[2 * i + 2 * 4 * dim1 + 1], W[sc * i + sc * 4 * dim1 - Woff], W[sc * i + sc * 4 * dim1 + 1 - Woff] );

            cplxMpy4_8_0( x10, x11, x[2 * i + 2 * 5 * dim1], x[2 * i + 2 * 5 * dim1 + 1], W[sc * i + sc * 5 * dim1 - Woff], W[sc * i + sc * 5 * dim1 + 1 - Woff] );

            cplxMpy4_8_0( x12, x13, x[2 * i + 2 * 6 * dim1], x[2 * i + 2 * 6 * dim1 + 1], W[sc * i + sc * 6 * dim1 - Woff], W[sc * i + sc * 6 * dim1 + 1 - Woff] );

            cplxMpy4_8_0( x14, x15, x[2 * i + 2 * 7 * dim1], x[2 * i + 2 * 7 * dim1 + 1], W[sc * i + sc * 7 * dim1 - Woff], W[sc * i + sc * 7 * dim1 + 1 - Woff] );

            cplxMpy4_8_0( x02, x03, x[add( shl( i, 1 ), i_mult( 2 * 1, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 1, dim1 ) ), 1 )], W[sub( add( i_mult( sc, i ), i_mult( sc, i_mult( 1, dim1 ) ) ), Woff )], W[sub( add( add( i_mult( sc, i ), i_mult( sc, i_mult( 1, dim1 ) ) ), 1 ), Woff )] );

            cplxMpy4_8_0( x04, x05, x[add( shl( i, 1 ), i_mult( 2 * 2, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 2, dim1 ) ), 1 )], W[sub( add( i_mult( sc, i ), i_mult( sc, i_mult( 2, dim1 ) ) ), Woff )], W[sub( add( add( i_mult( sc, i ), i_mult( sc, i_mult( 2, dim1 ) ) ), 1 ), Woff )] );

            cplxMpy4_8_0( x06, x07, x[add( shl( i, 1 ), i_mult( 2 * 3, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 3, dim1 ) ), 1 )], W[sub( add( i_mult( sc, i ), i_mult( sc, i_mult( 3, dim1 ) ) ), Woff )], W[sub( add( add( i_mult( sc, i ), i_mult( sc, i_mult( 3, dim1 ) ) ), 1 ), Woff )] );

            cplxMpy4_8_0( x08, x09, x[add( shl( i, 1 ), i_mult( 2 * 4, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 4, dim1 ) ), 1 )], W[sub( add( i_mult( sc, i ), i_mult( sc, i_mult( 4, dim1 ) ) ), Woff )], W[sub( add( add( i_mult( sc, i ), i_mult( sc, i_mult( 4, dim1 ) ) ), 1 ), Woff )] );

            cplxMpy4_8_0( x10, x11, x[add( shl( i, 1 ), i_mult( 2 * 5, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 5, dim1 ) ), 1 )], W[sub( add( i_mult( sc, i ), i_mult( sc, i_mult( 5, dim1 ) ) ), Woff )], W[sub( add( add( i_mult( sc, i ), i_mult( sc, i_mult( 5, dim1 ) ) ), 1 ), Woff )] );

            cplxMpy4_8_0( x12, x13, x[add( shl( i, 1 ), i_mult( 2 * 6, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 6, dim1 ) ), 1 )], W[sub( add( i_mult( sc, i ), i_mult( sc, i_mult( 6, dim1 ) ) ), Woff )], W[sub( add( add( i_mult( sc, i ), i_mult( sc, i_mult( 6, dim1 ) ) ), 1 ), Woff )] );

            cplxMpy4_8_0( x14, x15, x[add( shl( i, 1 ), i_mult( 2 * 7, dim1 ) )], x[add( add( shl( i, 1 ), i_mult( 2 * 7, dim1 ) ), 1 )], W[sub( add( i_mult( sc, i ), i_mult( sc, i_mult( 7, dim1 ) ) ), Woff )], W[sub( add( add( i_mult( sc, i ), i_mult( sc, i_mult( 7, dim1 ) ) ), 1 ), Woff )] );

        }

        t00 = L_shr( L_add( x00, x08 ), SCALEFACTORN2 - 1 );

        t02 = L_shr( L_sub( x00, x08 ), SCALEFACTORN2 - 1 );
@@ -6639,37 +6639,37 @@ static void BASOP_fftN2( 
        s13 = Mpy_32_16_1( L_sub( t03, t01 ), C81_FX );

        s15 = Mpy_32_16_1( L_add( t01, t03 ), C82_FX );



        re[sx * i + sx * 0 * dim1] = L_add( s00, s02 );

        re[add( i_mult( sx, i ), i_mult( sx, i_mult( 0, dim1 ) ) )] = L_add( s00, s02 );

        move32();

        im[sx * i + sx * 0 * dim1] = L_add( s01, s03 );

        im[add( i_mult( sx, i ), i_mult( sx, i_mult( 0, dim1 ) ) )] = L_add( s01, s03 );

        move32();

        re[sx * i + sx * 1 * dim1] = L_add( s10, s12 );

        re[add( i_mult( sx, i ), i_mult( sx, i_mult( 1, dim1 ) ) )] = L_add( s10, s12 );

        move32();

        im[sx * i + sx * 1 * dim1] = L_add( s11, s13 );

        im[add( i_mult( sx, i ), i_mult( sx, i_mult( 1, dim1 ) ) )] = L_add( s11, s13 );

        move32();

        re[sx * i + sx * 2 * dim1] = L_sub( s04, s06 );

        re[add( i_mult( sx, i ), i_mult( sx, i_mult( 2, dim1 ) ) )] = L_sub( s04, s06 );

        move32();

        im[sx * i + sx * 2 * dim1] = L_sub( s05, s07 );

        im[add( i_mult( sx, i ), i_mult( sx, i_mult( 2, dim1 ) ) )] = L_sub( s05, s07 );

        move32();

        re[sx * i + sx * 3 * dim1] = L_add( s08, s14 );

        re[add( i_mult( sx, i ), i_mult( sx, i_mult( 3, dim1 ) ) )] = L_add( s08, s14 );

        move32();

        im[sx * i + sx * 3 * dim1] = L_add( s09, s15 );

        im[add( i_mult( sx, i ), i_mult( sx, i_mult( 3, dim1 ) ) )] = L_add( s09, s15 );

        move32();

        re[sx * i + sx * 4 * dim1] = L_sub( s00, s02 );

        re[add( i_mult( sx, i ), i_mult( sx, i_mult( 4, dim1 ) ) )] = L_sub( s00, s02 );

        move32();

        im[sx * i + sx * 4 * dim1] = L_sub( s01, s03 );

        im[add( i_mult( sx, i ), i_mult( sx, i_mult( 4, dim1 ) ) )] = L_sub( s01, s03 );

        move32();

        re[sx * i + sx * 5 * dim1] = L_sub( s10, s12 );

        re[add( i_mult( sx, i ), i_mult( sx, i_mult( 5, dim1 ) ) )] = L_sub( s10, s12 );

        move32();

        im[sx * i + sx * 5 * dim1] = L_sub( s11, s13 );

        im[add( i_mult( sx, i ), i_mult( sx, i_mult( 5, dim1 ) ) )] = L_sub( s11, s13 );

        move32();

        re[sx * i + sx * 6 * dim1] = L_add( s04, s06 );

        re[add( i_mult( sx, i ), i_mult( sx, i_mult( 6, dim1 ) ) )] = L_add( s04, s06 );

        move32();

        im[sx * i + sx * 6 * dim1] = L_add( s05, s07 );

        im[add( i_mult( sx, i ), i_mult( sx, i_mult( 6, dim1 ) ) )] = L_add( s05, s07 );

        move32();

        re[sx * i + sx * 7 * dim1] = L_sub( s08, s14 );

        re[add( i_mult( sx, i ), i_mult( sx, i_mult( 7, dim1 ) ) )] = L_sub( s08, s14 );

        move32();

        im[sx * i + sx * 7 * dim1] = L_sub( s09, s15 );

        im[add( i_mult( sx, i ), i_mult( sx, i_mult( 7, dim1 ) ) )] = L_sub( s09, s15 );

        move32();

    }

lib_com/fft_evs.c

+6 −0
Original line number Diff line number Diff line
@@ -555,12 +555,15 @@ void fft16( Word32 *re, Word32 *im, Word16 s, Word16 bScale ) 
        FOR( i = 0; i < 16; i++ )

        {

            inp_data[i] = CL_form( re[s * i], im[s * i] );

            move64();

        }

        fft16_with_cmplx_data( inp_data, bScale );

        FOR( i = 0; i < 16; i++ )

        {

            re[s * i] = CL_Extract_real( inp_data[i] );

            move32();

            im[s * i] = CL_Extract_imag( inp_data[i] );

            move32();

        }

    }

}
@@ -1927,6 +1930,7 @@ static void fftN2( 
            cmplx s0, s1, s2, s3, s4, s5, s6, s7;



            i = 0;

            move16();

            {

                y0 = CL_shr( x_cmplx[i + 0 * dim1], 1 );

                y1 = CL_shr( x_cmplx[i + 1 * dim1], 1 );
@@ -2463,7 +2467,9 @@ void BASOP_rfft( Word32 *x, Word16 sizeOfFft, Word16 *scale, Word16 isign ) 
                case 320:

                case 640:

                    c1 = FFTC( 0x66666680 );

                    move16();

                    c2 = FFTC( 0x99999980 );

                    move16();

                    FOR( i = 0; i < sizeOfFft2; i++ )

                    {

                        x[2 * i] = Mpy_32_xx( x[2 * i], c1 );

lib_com/fft_fx_evs.c

+77 −12

File changed.

Preview size limit exceeded, changes collapsed.