Commit f908d5bf authored by Fabian Bauer's avatar Fabian Bauer
Browse files

fft_fx_evs.c : completed overflow op replacement

parent 5a23784c

lib_com/fft_fx_evs.c

+429 −1
Original line number Diff line number Diff line
@@ -1186,9 +1186,11 @@ static void fft5_32_16fx( 
    Word16 T1, To, T8, Tt, T9, Ts, Te, Tp, Th, Tn, T2, T3, T4, T5, T6, T7;

    Word16 i0, i1, i2, i3, i4;

    Word32 L_tmp;

#ifndef ISSUE_1836_replace_overflow_libcom

#ifdef BASOP_NOGLOB_DECLARE_LOCAL

    Flag Overflow = 0;

    move32();

#endif

#endif

    i0 = Idx[0];

    move16();
@@ -1216,9 +1218,15 @@ static void fft5_32_16fx( 
    move16();

    T6 = zRe[i3];

    move16();

#ifdef ISSUE_1836_replace_overflow_libcom

    T7 = add_sat( T5, T6 );

    T8 = add_sat( T4, T7 );

    Tt = sub_sat( T5, T6 );

#else

    T7 = add_o( T5, T6, &Overflow );

    T8 = add_o( T4, T7, &Overflow );

    Tt = sub_o( T5, T6, &Overflow );

#endif

    /* T9 = KP559016994 * (T4 - T7); */

    L_tmp = Mult_32_16( KP559016994_16FX, sub_sat( T4, T7 ) ); // Q(16 +x)

    T9 = round_fx_sat( L_tmp );                                // Qx
@@ -1240,11 +1248,17 @@ static void fft5_32_16fx( 
    L_tmp = Mult_32_16( KP559016994_16FX, sub_sat( T4, T7 ) ); // Q(16 +x)

    Tn = round_fx_sat( L_tmp );                                // Qx





#ifdef ISSUE_1836_replace_overflow_libcom

    zRe[i0] = add_sat( T1, T8 );

    move16();

    zIm[i0] = add_sat( To, Tp );

    move32();

#else

    zRe[i0] = add_o( T1, T8, &Overflow );

    move16();

    zIm[i0] = add_o( To, Tp, &Overflow );

    move32();

#endif



    /*T2 = KP951056516*Te + KP587785252*Th; */

    L_tmp = Mult_32_16( KP951056516_16FX, Te );        // Q(16 +x)
@@ -1261,6 +1275,16 @@ static void fft5_32_16fx( 
    T4 = add_sat( T9, T6 );

    T5 = sub_sat( T6, T9 );



#ifdef ISSUE_1836_replace_overflow_libcom

    zRe[i3] = sub_sat( T4, T2 );

    move32();

    zRe[i1] = add_sat( T5, T3 );

    move32();

    zRe[i2] = add_sat( T4, T2 );

    move32();

    zRe[i4] = sub_sat( T5, T3 );

    move32();

#else

    zRe[i3] = sub_o( T4, T2, &Overflow );

    move32();

    zRe[i1] = add_o( T5, T3, &Overflow );
@@ -1269,6 +1293,7 @@ static void fft5_32_16fx( 
    move32();

    zRe[i4] = sub_o( T5, T3, &Overflow );

    move32();

#endif



    /*    T2 = KP951056516 * Ts + KP587785252 * Tt; */

    L_tmp = Mult_32_16( KP951056516_16FX, Ts );        // Q(16 +x)
@@ -1675,9 +1700,11 @@ static void cftfsub_16fx( 
{

    Word16 j, j1, j2, j3, l;

    Word16 x0r, x0i, x1r, x1i, x2r, x2i, x3r, x3i;

#ifndef ISSUE_1836_replace_overflow_libcom

#ifdef BASOP_NOGLOB_DECLARE_LOCAL

    Flag Overflow = 0;

    move32();

#endif

#endif



    l = 2;
@@ -1731,6 +1758,15 @@ static void cftfsub_16fx( 
    {

        FOR( j = 0; j < l; j += 2 )

        {

#ifdef ISSUE_1836_replace_overflow_libcom

            j1 = add_sat( j, l );

            x0r = sub_sat( a[j], a[j1] );

            x0i = sub_sat( a[j + 1], a[j1 + 1] );

            a[j] = add_sat( a[j], a[j1] );

            move16();

            a[j + 1] = add_sat( a[j + 1], a[j1 + 1] );

            move16();

#else

            j1 = add_o( j, l, &Overflow );

            x0r = sub_o( a[j], a[j1], &Overflow );

            x0i = sub_o( a[j + 1], a[j1 + 1], &Overflow );
@@ -1738,6 +1774,7 @@ static void cftfsub_16fx( 
            move16();

            a[j + 1] = add_o( a[j + 1], a[j1 + 1], &Overflow );

            move16();

#endif

            a[j1] = x0r;

            move16();

            a[j1 + 1] = x0i;
@@ -1762,11 +1799,84 @@ static void cft1st_16fx( 
    Word16 x0r, x0i, x1r, x1i, x2r, x2i, x3r, x3i;

    Word16 tmp;

    Word32 L_tmp;

#ifndef ISSUE_1836_replace_overflow_libcom

#ifdef BASOP_NOGLOB_DECLARE_LOCAL

    Flag Overflow = 0;

    move32();

#endif

#endif



#ifdef ISSUE_1836_replace_overflow_libcom

    x0r = add_sat( a[0], a[2] );

    x0i = add_sat( a[1], a[3] );

    x1r = sub_sat( a[0], a[2] );

    x1i = sub_sat( a[1], a[3] );

    x2r = add_sat( a[4], a[6] );

    x2i = add_sat( a[5], a[7] );

    x3r = sub_sat( a[4], a[6] );

    x3i = sub_sat( a[5], a[7] );

    a[0] = add_sat( x0r, x2r );

    move16();

    a[1] = add_sat( x0i, x2i );

    move16();

    a[4] = sub_sat( x0r, x2r );

    move16();

    a[5] = sub_sat( x0i, x2i );

    move16();

    a[2] = sub_sat( x1r, x3i );

    move16();

    a[3] = add_sat( x1i, x3r );

    move16();

    a[6] = add_sat( x1r, x3i );

    move16();

    a[7] = sub_sat( x1i, x3r );

    wk1r = w[2];

    move32();



    x0r = add_sat( a[8], a[10] );

    x0i = add_sat( a[9], a[11] );

    x1r = sub_sat( a[8], a[10] );

    x1i = sub_sat( a[9], a[11] );

    x2r = add_sat( a[12], a[14] );

    x2i = add_sat( a[13], a[15] );

    x3r = sub_sat( a[12], a[14] );

    x3i = sub_sat( a[13], a[15] );

    a[8] = add_sat( x0r, x2r );

    move16();

    a[9] = add_sat( x0i, x2i );

    move16();

    a[12] = sub_sat( x2i, x0i );

    move16();

    a[13] = sub_sat( x0r, x2r );

    move16();



    x0r = sub_sat( x1r, x3i );

    x0i = add_sat( x1i, x3r );

    tmp = sub_sat( x0r, x0i );

    L_tmp = Mult_32_16( wk1r, tmp );                                 /*Q(15+Qx+Q_edct) */



    a[10] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

    move16();



    tmp = add_sat( x0r, x0i );

    L_tmp = Mult_32_16( wk1r, tmp );                                 /*Q(15+Qx+Q_edct) */

    a[11] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /* Q(Qx+Q_edct) */

    move16();



    x0r = add_sat( x3i, x1r );

    x0i = sub_sat( x3r, x1i );

    tmp = sub_sat( x0i, x0r );

    L_tmp = Mult_32_16( wk1r, tmp );                                 /*Q(15+Qx+Q_edct) */

    a[14] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

    move16();



    tmp = add_sat( x0i, x0r );

    L_tmp = Mult_32_16( wk1r, tmp );                                 /*Q(15+Qx+Q_edct) */

    a[15] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

    move16();

    k1 = 0;

    move16();

#else

    x0r = add_o( a[0], a[2], &Overflow );

    x0i = add_o( a[1], a[3], &Overflow );

    x1r = sub_o( a[0], a[2], &Overflow );
@@ -1836,6 +1946,7 @@ static void cft1st_16fx( 
    move16();

    k1 = 0;

    move16();

#endif



    FOR( j = 16; j < n; j += 16 )

    {
@@ -1856,6 +1967,118 @@ static void cft1st_16fx( 


        L_tmp = L_shl( Mult_32_32( wk2i, wk1r ), 1 ); /*Q29 */

        wk3i = L_sub( L_shl( L_tmp, 1 ), wk1i );      /*Q30 */

#ifdef ISSUE_1836_replace_overflow_libcom

        x0r = add_sat( a[j], a[j + 2] );

        x0i = add_sat( a[j + 1], a[j + 3] );

        x1r = sub_sat( a[j], a[j + 2] );

        x1i = sub_sat( a[j + 1], a[j + 3] );

        x2r = add_sat( a[j + 4], a[j + 6] );

        x2i = add_sat( a[j + 5], a[j + 7] );

        x3r = sub_sat( a[j + 4], a[j + 6] );

        x3i = sub_sat( a[j + 5], a[j + 7] );

        a[j] = add_sat( x0r, x2r );

        move16();

        a[j + 1] = add_sat( x0i, x2i );

        move16();



        x0r = sub_sat( x0r, x2r );

        x0i = sub_sat( x0i, x2i );

        L_tmp = Mult_32_16( wk2r, x0r );                                    /*Q(15+Qx+Q_edct) */

        L_tmp = Msub_32_16( L_tmp, wk2i, x0i );                             /*Q(15+Qx+Q_edct) */

        a[j + 4] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        L_tmp = Mult_32_16( wk2r, x0i );                                    /*Q(15+Qx+Q_edct) */

        L_tmp = Madd_32_16( L_tmp, wk2i, x0r );                             /*Q(15+Qx+Q_edct) */

        a[j + 5] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        x0r = sub_sat( x1r, x3i );

        x0i = add_sat( x1i, x3r );

        L_tmp = Mult_32_16( wk1r, x0r );                                    /*Q(15+Qx+Q_edct) */

        L_tmp = Msub_32_16( L_tmp, wk1i, x0i );                             /*Q(15+Qx+Q_edct) */

        a[j + 2] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        L_tmp = Mult_32_16( wk1r, x0i );                                    /*Q(15+Qx+Q_edct) */

        L_tmp = Madd_32_16( L_tmp, wk1i, x0r );                             /*Q(15+Qx+Q_edct) */

        a[j + 3] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        x0r = add_sat( x1r, x3i );

        x0i = sub_sat( x1i, x3r );

        L_tmp = Mult_32_16( wk3r, x0r );                                    /*Q(15+Qx+Q_edct) */

        L_tmp = Msub_32_16( L_tmp, wk3i, x0i );                             /*Q(15+Qx+Q_edct) */

        a[j + 6] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        L_tmp = Mult_32_16( wk3r, x0i );                                    /*Q(15+Qx+Q_edct) */

        L_tmp = Madd_32_16( L_tmp, wk3i, x0r );                             /*Q(15+Qx+Q_edct) */

        a[j + 7] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        wk1r = w[k2 + 2];

        move32();

        wk1i = w[k2 + 3];

        move32();

        L_tmp = L_shl( Mult_32_32( wk2r, wk1i ), 1 ); /*Q29 */

        wk3r = L_sub( wk1r, L_shl( L_tmp, 1 ) );      /*Q30  */



        L_tmp = L_shl( Mult_32_32( wk2r, wk1r ), 1 ); /*Q29 */

        wk3i = L_sub( L_shl( L_tmp, 1 ), wk1i );      /*Q30 */



        x0r = add_sat( a[j + 8], a[j + 10] );

        x0i = add_sat( a[j + 9], a[j + 11] );

        x1r = sub_sat( a[j + 8], a[j + 10] );

        x1i = sub_sat( a[j + 9], a[j + 11] );

        x2r = add_sat( a[j + 12], a[j + 14] );

        x2i = add_sat( a[j + 13], a[j + 15] );

        x3r = sub_sat( a[j + 12], a[j + 14] );

        x3i = sub_sat( a[j + 13], a[j + 15] );

        a[j + 8] = add_sat( x0r, x2r );

        move16();

        a[j + 9] = add_sat( x0i, x2i );

        move16();



        x0r = sub_sat( x0r, x2r );

        x0i = sub_sat( x0i, x2i );

        tmp = negate( x0r );

        L_tmp = Mult_32_16( wk2i, tmp );                                     /*Q(15+Qx+Q_edct) */

        L_tmp = Msub_32_16( L_tmp, wk2r, x0i );                              /*Q(15+Qx+Q_edct) */

        a[j + 12] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        tmp = negate( x0i );

        L_tmp = Mult_32_16( wk2i, tmp );                                     /*Q(15+Qx+Q_edct) */

        L_tmp = Madd_32_16( L_tmp, wk2r, x0r );                              /*Q(15+Qx+Q_edct) */

        a[j + 13] = round_fx_sat( L_shl_sat( L_tmp, 1 ) ); /*Q(Qx+Q_edct) */

        move16();



        x0r = sub_sat( x1r, x3i );

        x0i = add_sat( x1i, x3r );

        L_tmp = Mult_32_16( wk1r, x0r );                                     /*Q(15+Qx+Q_edct) */

        L_tmp = Msub_32_16( L_tmp, wk1i, x0i );                              /*Q(15+Qx+Q_edct) */

        a[j + 10] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        L_tmp = Mult_32_16( wk1r, x0i );                                     /*Q(15+Qx+Q_edct) */

        L_tmp = Madd_32_16( L_tmp, wk1i, x0r );                              /*Q(15+Qx+Q_edct) */

        a[j + 11] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        x0r = add_sat( x1r, x3i );

        x0i = sub_sat( x1i, x3r );



        L_tmp = Mult_32_16( wk3r, x0r );                                     /*Q(15+Qx+Q_edct) */

        L_tmp = Msub_32_16( L_tmp, wk3i, x0i );                              /*Q(15+Qx+Q_edct) */

        a[j + 14] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        L_tmp = Mult_32_16( wk3r, x0i );                                     /*Q(15+Qx+Q_edct) */

        L_tmp = Madd_32_16( L_tmp, wk3i, x0r );                              /*Q(15+Qx+Q_edct) */

        a[j + 15] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();

#else

        x0r = add_o( a[j], a[j + 2], &Overflow );

        x0i = add_o( a[j + 1], a[j + 3], &Overflow );

        x1r = sub_o( a[j], a[j + 2], &Overflow );
@@ -1966,6 +2189,7 @@ static void cft1st_16fx( 
        L_tmp = Madd_32_16( L_tmp, wk3i, x0r );                              /*Q(15+Qx+Q_edct) */

        a[j + 15] = round_fx_o( L_shl_o( L_tmp, 1, &Overflow ), &Overflow ); /*Q(Qx+Q_edct) */

        move16();

#endif

    }



    return;
@@ -1988,13 +2212,44 @@ static void cftmdl_16fx( 
    Word16 tmp, tmp2;

    Word32 L_tmp;

    Word32 L_x0r, L_x0i;

#ifndef ISSUE_1836_replace_overflow_libcom

#ifdef BASOP_NOGLOB_DECLARE_LOCAL

    Flag Overflow = 0;

    move32();

#endif

#endif

    m = shl( l, 2 );

    FOR( j = 0; j < l; j += 2 )

    {

#ifdef ISSUE_1836_replace_overflow_libcom

        j1 = add_sat( j, l );

        j2 = add_sat( j1, l );

        j3 = add_sat( j2, l );

        x0r = add_sat( a[j], a[j1] );

        x0i = add_sat( a[j + 1], a[j1 + 1] );

        x1r = sub_sat( a[j], a[j1] );

        x1i = sub_sat( a[j + 1], a[j1 + 1] );

        x2r = add_sat( a[j2], a[j3] );

        x2i = add_sat( a[j2 + 1], a[j3 + 1] );

        x3r = sub_sat( a[j2], a[j3] );

        x3i = sub_sat( a[j2 + 1], a[j3 + 1] );

        a[j] = add_sat( x0r, x2r );

        move16();

        a[j + 1] = add_sat( x0i, x2i );

        move16();

        a[j2] = sub_sat( x0r, x2r );

        move16();

        a[j2 + 1] = sub_sat( x0i, x2i );

        move16();

        a[j1] = sub_sat( x1r, x3i );

        move16();

        a[j1 + 1] = add_sat( x1i, x3r );

        move16();

        a[j3] = add_sat( x1r, x3i );

        move16();

        a[j3 + 1] = sub_sat( x1i, x3r );

        move16();

#else

        j1 = add_o( j, l, &Overflow );

        j2 = add_o( j1, l, &Overflow );

        j3 = add_o( j2, l, &Overflow );
@@ -2022,6 +2277,7 @@ static void cftmdl_16fx( 
        move16();

        a[j3 + 1] = sub_o( x1i, x3r, &Overflow );

        move16();

#endif

    }



    wk1r = w[2];
@@ -2029,6 +2285,51 @@ static void cftmdl_16fx( 
    tmp2 = add( l, m );

    FOR( j = m; j < tmp2; j += 2 )

    {

#ifdef ISSUE_1836_replace_overflow_libcom

        j1 = add_sat( j, l );

        j2 = add_sat( j1, l );

        j3 = add_sat( j2, l );

        x0r = add_sat( a[j], a[j1] );

        x0i = add_sat( a[j + 1], a[j1 + 1] );

        x1r = sub_sat( a[j], a[j1] );

        x1i = sub_sat( a[j + 1], a[j1 + 1] );

        x2r = add_sat( a[j2], a[j3] );

        x2i = add_sat( a[j2 + 1], a[j3 + 1] );

        x3r = sub_sat( a[j2], a[j3] );

        x3i = sub_sat( a[j2 + 1], a[j3 + 1] );

        a[j] = add_sat( x0r, x2r );

        move16();

        a[j + 1] = add_sat( x0i, x2i );

        move16();

        a[j2] = sub_sat( x2i, x0i );

        move16();

        a[j2 + 1] = sub_sat( x0r, x2r );

        move16();



        x0r = sub_sat( x1r, x3i );

        x0i = add_sat( x1i, x3r );

        tmp = sub_sat( x0r, x0i );

        L_tmp = Mult_32_16( wk1r, tmp );                                 /*Q(15+Qx+Q_edct) */

        a[j1] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        tmp = add_sat( x0r, x0i );

        L_tmp = Mult_32_16( wk1r, tmp );                                     /*Q(15+Qx+Q_edct) */

        a[j1 + 1] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        x0r = add_sat( x3i, x1r );

        x0i = sub_sat( x3r, x1i );

        tmp = sub_sat( x0i, x0r );

        L_tmp = Mult_32_16( wk1r, tmp );                                 /*Q(15+Qx+Q_edct) */

        a[j3] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();



        tmp = add_sat( x0i, x0r );

        L_tmp = Mult_32_16( wk1r, tmp );                                     /*Q(15+Qx+Q_edct) */

        a[j3 + 1] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

        move16();

#else

        j1 = add_o( j, l, &Overflow );

        j2 = add_o( j1, l, &Overflow );

        j3 = add_o( j2, l, &Overflow );
@@ -2072,6 +2373,7 @@ static void cftmdl_16fx( 
        L_tmp = Mult_32_16( wk1r, tmp );                                     /*Q(15+Qx+Q_edct) */

        a[j3 + 1] = round_fx_o( L_shl_o( L_tmp, 1, &Overflow ), &Overflow ); /*Q(Qx+Q_edct) */

        move16();

#endif

    }



    k1 = 0;
@@ -2098,6 +2400,63 @@ static void cftmdl_16fx( 
        tmp2 = add( l, k );

        FOR( j = k; j < tmp2; j += 2 )

        {

#ifdef ISSUE_1836_replace_overflow_libcom

            j1 = add_sat( j, l );

            j2 = add_sat( j1, l );

            j3 = add_sat( j2, l );

            x0r = add_sat( a[j], a[j1] );

            x0i = add_sat( a[j + 1], a[j1 + 1] );

            x1r = sub_sat( a[j], a[j1] );

            x1i = sub_sat( a[j + 1], a[j1 + 1] );

            x2r = add_sat( a[j2], a[j3] );

            x2i = add_sat( a[j2 + 1], a[j3 + 1] );

            x3r = sub_sat( a[j2], a[j3] );

            x3i = sub_sat( a[j2 + 1], a[j3 + 1] );

            a[j] = add_sat( x0r, x2r );

            move16();

            a[j + 1] = add_sat( x0i, x2i );

            move16();



            x0r = sub_sat( x0r, x2r );

            x0i = sub_sat( x0i, x2i );



            L_tmp = Mult_32_16( wk2r, x0r );                                 /*Q(15+Qx+Q_edct) */

            L_tmp = Msub_32_16( L_tmp, wk2i, x0i );                          /*Q(15+Qx+Q_edct) */

            a[j2] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

            move16();



            L_tmp = Mult_32_16( wk2r, x0i );                                     /*Q(15+Qx+Q_edct) */

            L_tmp = Madd_32_16( L_tmp, wk2i, x0r );                              /*Q(15+Qx+Q_edct) */

            a[j2 + 1] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

            move16();



            x0r = sub_sat( x1r, x3i );

            x0i = add_sat( x1i, x3r );



            L_tmp = Mult_32_16( wk1r, x0r );                                 /*Q(15+Qx+Q_edct) */

            L_tmp = Msub_32_16( L_tmp, wk1i, x0i );                          /*Q(15+Qx+Q_edct) */

            a[j1] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

            move16();



            L_tmp = Mult_32_16( wk1r, x0i );                                     /*Q(15+Qx+Q_edct) */

            L_tmp = Madd_32_16( L_tmp, wk1i, x0r );                              /*Q(15+Qx+Q_edct) */

            a[j1 + 1] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

            move16();



            L_x0r = L_add( (Word32) x1r, (Word32) x3i );

            L_x0i = L_sub( (Word32) x1i, (Word32) x3r );

            x0r = extract_l( L_x0r );

            x0i = extract_l( L_x0i );

            L_tmp = Mult_32_16( wk3r, x0r );                                 /*Q(15+Qx+Q_edct) */

            L_tmp = Msub_32_16( L_tmp, wk3i, x0i );                          /*Q(15+Qx+Q_edct) */

            a[j3] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

            move16();



            L_tmp = Mult_32_16( wk3r, x0i );                                     /*Q(15+Qx+Q_edct) */

            L_tmp = Madd_32_16( L_tmp, wk3i, x0r );                              /*Q(15+Qx+Q_edct) */

            a[j3 + 1] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

            move16();

#else

            j1 = add_o( j, l, &Overflow );

            j2 = add_o( j1, l, &Overflow );

            j3 = add_o( j2, l, &Overflow );
@@ -2153,20 +2512,88 @@ static void cftmdl_16fx( 
            L_tmp = Madd_32_16( L_tmp, wk3i, x0r );                              /*Q(15+Qx+Q_edct) */

            a[j3 + 1] = round_fx_o( L_shl_o( L_tmp, 1, &Overflow ), &Overflow ); /*Q(Qx+Q_edct) */

            move16();

#endif

        }



        wk1r = w[k2 + 2];

        move32();

        wk1i = w[k2 + 3];

        move32();

#ifdef ISSUE_1836_replace_overflow_libcom

        L_tmp = L_shl_sat( Mult_32_32( wk2r, wk1i ), 1 );                  /*Q29 */

        wk3r = L_sub_sat( wk1r, L_shl_sat( L_tmp, 1 ) );                   /*Q30  */



        L_tmp = L_shl_sat( Mult_32_32( wk2r, wk1r ), 1 );                  /*Q29 */

        wk3i = L_sub_sat( L_shl_sat( L_tmp, 1 ), wk1i );                   /*Q30 */

        tmp2 = add( l, add( k, m ) );

#else

        L_tmp = L_shl_o( Mult_32_32( wk2r, wk1i ), 1, &Overflow );         /*Q29 */

        wk3r = L_sub_o( wk1r, L_shl_o( L_tmp, 1, &Overflow ), &Overflow ); /*Q30  */



        L_tmp = L_shl_o( Mult_32_32( wk2r, wk1r ), 1, &Overflow );         /*Q29 */

        wk3i = L_sub_o( L_shl_o( L_tmp, 1, &Overflow ), wk1i, &Overflow ); /*Q30 */

        tmp2 = add( l, add( k, m ) );

#endif

        FOR( j = add( k, m ); j < tmp2; j += 2 )

        {

#ifdef ISSUE_1836_replace_overflow_libcom

            j1 = add_sat( j, l );

            j2 = add_sat( j1, l );

            j3 = add_sat( j2, l );

            x0r = add_sat( a[j], a[j1] );

            x0i = add_sat( a[j + 1], a[j1 + 1] );

            x1r = sub_sat( a[j], a[j1] );

            x1i = sub_sat( a[j + 1], a[j1 + 1] );

            x2r = add_sat( a[j2], a[j3] );

            x2i = add_sat( a[j2 + 1], a[j3 + 1] );

            x3r = sub_sat( a[j2], a[j3] );

            x3i = sub_sat( a[j2 + 1], a[j3 + 1] );

            a[j] = add_sat( x0r, x2r );

            move16();

            a[j + 1] = add_sat( x0i, x2i );

            move16();



            x0r = sub_sat( x0r, x2r );

            x0i = sub_sat( x0i, x2i );



            tmp = negate( x0r );

            L_tmp = Mult_32_16( wk2i, tmp );                                 /*Q(15+Qx+Q_edct) */

            L_tmp = Msub_32_16( L_tmp, wk2r, x0i );                          /*Q(15+Qx+Q_edct) */

            a[j2] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

            move16();



            tmp = negate( x0i );

            L_tmp = Mult_32_16( wk2i, tmp );                                     /*Q(15+Qx+Q_edct) */

            L_tmp = Madd_32_16( L_tmp, wk2r, x0r );                              /*Q(15+Qx+Q_edct) */

            a[j2 + 1] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

            move16();



            x0r = sub_sat( x1r, x3i );

            x0i = add_sat( x1i, x3r );



            L_tmp = Mult_32_16( wk1r, x0r );                                 /*Q(15+Qx+Q_edct) */

            L_tmp = Msub_32_16( L_tmp, wk1i, x0i );                          /*Q(15+Qx+Q_edct) */

            a[j1] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

            move16();



            L_tmp = Mult_32_16( wk1r, x0i );                                     /*Q(15+Qx+Q_edct) */

            L_tmp = Madd_32_16( L_tmp, wk1i, x0r );                              /*Q(15+Qx+Q_edct) */

            a[j1 + 1] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

            move16();



            x0r = add_sat( x1r, x3i );

            x0i = sub_sat( x1i, x3r );



            L_tmp = Mult_32_16( wk3r, x0r );                                 /*Q(15+Qx+Q_edct) */

            L_tmp = Msub_32_16( L_tmp, wk3i, x0i );                          /*Q(15+Qx+Q_edct) */

            a[j3] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

            move16();



            L_tmp = Mult_32_16( wk3r, x0i );                                     /*Q(15+Qx+Q_edct) */

            L_tmp = Madd_32_16( L_tmp, wk3i, x0r );                              /*Q(15+Qx+Q_edct) */

            a[j3 + 1] = round_fx_sat( L_shl_sat( L_tmp, 1 ) );                   /*Q(Qx+Q_edct) */

            move16();

#else

            j1 = add_o( j, l, &Overflow );

            j2 = add_o( j1, l, &Overflow );

            j3 = add_o( j2, l, &Overflow );
@@ -2223,6 +2650,7 @@ static void cftmdl_16fx( 
            L_tmp = Madd_32_16( L_tmp, wk3i, x0r );                              /*Q(15+Qx+Q_edct) */

            a[j3 + 1] = round_fx_o( L_shl_o( L_tmp, 1, &Overflow ), &Overflow ); /*Q(Qx+Q_edct) */

            move16();

#endif

        }

    }