Commit 4a1cd706 authored by vaillancour's avatar vaillancour
Browse files

more fixes for ffts

parent d1f30399

lib_com/edct_fx.c

+16 −3
Original line number Diff line number Diff line
@@ -314,13 +314,21 @@ void edct_16fx( 
        i2 = shl(i,1);



        L_tmp = L_mult(x[i2],edct_table[i]);/*Q(Qx+16) */

#ifdef BASOP_NOGLOB

        Lacc = L_mac_o(L_tmp,*px,*pt, &Overflow);/*Q(Qx+16) */

        re2[i] = round_fx_o(L_shl_o(Lacc, Q_edct, &Overflow), &Overflow); /* Q(Qx+Q_edct) */

#else

        Lacc = L_mac(L_tmp,*px,*pt);/*Q(Qx+16) */

        re2[i] = round_fx(L_shl(Lacc, Q_edct)); /* Q(Qx+Q_edct) */



#endif

        L_tmp = L_mult(*px,edct_table[i]);/*Q(Qx+16) */

#ifdef BASOP_NOGLOB

        Lacc = L_msu_o(L_tmp,x[i2],*pt, &Overflow);/*Q(Qx+16) */

        im2[i] = round_fx_o(L_shl_o(Lacc, Q_edct, &Overflow), &Overflow); /* Q(Qx+Q_edct) */

#else

        Lacc = L_msu(L_tmp,x[i2],*pt);/*Q(Qx+16) */

        im2[i] = round_fx(L_shl(Lacc, Q_edct)); /* Q(Qx+Q_edct) */



#endif

        px -= 2;

        pt--;

    }
@@ -352,7 +360,11 @@ void edct_16fx( 
#endif



        tmp = mult_r(re2[i],fact); /*Q(Qx+Q_edct) */

#ifdef BASOP_NOGLOB

        im[i] = add_o(im2[i],tmp, &Overflow); /*Q(Qx+Q_edct) */  move16();

#else

        im[i] = add(im2[i],tmp); /*Q(Qx+Q_edct) */  move16();

#endif

    }



    /* Post-rotate and obtain the output data */
@@ -363,10 +375,11 @@ void edct_16fx( 
        i2 = shl(i,1);



        L_tmp = L_mult(re[i],edct_table[i]);/*Q(Qx+Q_edct+16) */

        Lacc = L_mac(L_tmp,im[i],*pt);/*Q(Qx+Q_edct+16) */

#ifdef BASOP_NOGLOB

        Lacc = L_mac_o(L_tmp, im[i], *pt, &Overflow);/*Q(Qx+Q_edct+16) */

        y[i2] = round_fx_o(L_shr_o(Lacc,Q_edct, &Overflow), &Overflow); /* Q(Qx) */

#else /* BASOP_NOGLOB */

        Lacc = L_mac(L_tmp, im[i], *pt);/*Q(Qx+Q_edct+16) */

        y[i2] = round_fx(L_shr(Lacc,Q_edct)); /* Q(Qx) */

#endif /* BASOP_NOGLOB */

lib_com/fft_fx.c

+264 −41
Original line number Diff line number Diff line
@@ -1715,14 +1715,17 @@ static void cftfsub_16fx( 
#else

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

#endif

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

#ifdef BASOP_NOGLOB

            x0i = sub_o(a[j + 1], a[j1 + 1], &Overflow);

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

            move16();

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

#else

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

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

#endif

            move16();

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

#endif

            move16();

            a[j1] = x0r;

            move16();
@@ -1748,7 +1751,9 @@ static void cft1st_16fx( 
    Word16 x0r, x0i, x1r, x1i, x2r, x2i, x3r, x3i;

    Word16 tmp;

    Word32 L_tmp;



#ifdef BASOP_NOGLOB

    Flag Overflow = 0;

#endif

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

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

    x1r = sub(a[0],a[2]);
@@ -1834,6 +1839,23 @@ 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 BASOP_NOGLOB

        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);

        x1i = sub_o(a[j + 1],a[j + 3], &Overflow);

        x2r = add_o(a[j + 4],a[j + 6], &Overflow);

        x2i = add_o(a[j + 5],a[j + 7], &Overflow);

        x3r = sub_o(a[j + 4],a[j + 6], &Overflow);

        x3i = sub_o(a[j + 5],a[j + 7], &Overflow);

        a[j] = add_o(x0r,x2r, &Overflow);

        move16();

        a[j + 1] = add_o(x0i,x2i, &Overflow);

        move16();



        x0r = sub_o(x0r, x2r, &Overflow);

        x0i = sub_o(x0i, x2i, &Overflow);

#else

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

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

        x1r = sub(a[j], a[j + 2]);
@@ -1849,32 +1871,67 @@ static void cft1st_16fx( 


        x0r = sub(x0r, x2r);

        x0i = sub(x0i, x2i);

#endif



        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) */

#ifdef BASOP_NOGLOB

        a[j + 4] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */

#else

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



#endif

        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) */

#ifdef BASOP_NOGLOB

        a[j + 5] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */

        x0r = sub_o(x1r, x3i, &Overflow);

        x0i = add_o(x1i, x3r, &Overflow);

#else

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



        x0r = sub(x1r, x3i);

        x0i = add(x1i, x3r); 

#endif



        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) */

#ifdef BASOP_NOGLOB

        a[j + 2] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */

#else

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



#endif

        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) */

#ifdef BASOP_NOGLOB

        a[j + 3] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */



        x0r = add_o(x1r,x3i, &Overflow);

        x0i = sub_o(x1i,x3r, &Overflow);

#else

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



        x0r = add(x1r, x3i);

        x0i = sub(x1i, x3r);

#endif

        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) */

#ifdef BASOP_NOGLOB

        a[j + 6] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */

#else

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



#endif

        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) */

#ifdef BASOP_NOGLOB

        a[j + 7] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */



        wk1r = w[k2 + 2];	move32();

        wk1i = w[k2 + 3];	move32();

        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 */

#else

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



        wk1r = w[k2 + 2];	move32();
@@ -1884,7 +1941,22 @@ static void cft1st_16fx( 


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

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



#endif

#ifdef BASOP_NOGLOB

        x0r = add_o(a[j + 8],a[j + 10], &Overflow);

        x0i = add_o(a[j + 9],a[j + 11], &Overflow);

        x1r = sub_o(a[j + 8],a[j + 10], &Overflow);

        x1i = sub_o(a[j + 9],a[j + 11], &Overflow);

        x2r = add_o(a[j + 12],a[j + 14], &Overflow);

        x2i = add_o(a[j + 13],a[j + 15], &Overflow);

        x3r = sub_o(a[j + 12],a[j + 14], &Overflow);

        x3i = sub_o(a[j + 13],a[j + 15], &Overflow);

        a[j + 8] = add_o(x0r,x2r, &Overflow);

        move16();

        a[j + 9] = add_o(x0i,x2i, &Overflow);

        x0r = sub_o(x0r, x2r, &Overflow);

        x0i = sub_o(x0i, x2i, &Overflow);

#else

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

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

        x1r = sub(a[j + 8],a[j + 10]);
@@ -1897,39 +1969,66 @@ static void cft1st_16fx( 
        move16();

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

        move16();



        x0r = sub(x0r, x2r);

        x0i = sub(x0i, x2i); 

#endif



        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) */

#ifdef BASOP_NOGLOB

        a[j + 12] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */

#else

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



#endif

        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) */

#ifdef BASOP_NOGLOB

        a[j + 13] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */



        x0r = sub_o(x1r,x3i, &Overflow);

        x0i = add_o(x1i,x3r, &Overflow);

#else

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



        x0r = sub(x1r,x3i);

        x0i = add(x1i,x3r);

#endif

        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) */

#ifdef BASOP_NOGLOB

        a[j + 10] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */

#else

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



#endif

        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) */

#ifdef BASOP_NOGLOB

        a[j + 11] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */



        x0r = add_o(x1r,x3i, &Overflow);

        x0i = sub_o(x1i,x3r, &Overflow);

#else

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



        x0r = add(x1r,x3i);

        x0i = sub(x1i,x3r);



#endif

        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) */

#ifdef BASOP_NOGLOB

        a[j + 14] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */

#else

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



#endif

        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) */

#ifdef BASOP_NOGLOB

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

#else

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

#endif

    }



    return;
@@ -1952,7 +2051,9 @@ static void cftmdl_16fx( 
    Word16 tmp, tmp2;

    Word32 L_tmp;

    Word32 L_x0r, L_x0i;



#ifdef BASOP_NOGLOB

    Flag Overflow = 0;

#endif

    m = shl(l, 2);

    move16();

    FOR (j = 0; j < l; j += 2)
@@ -1960,6 +2061,32 @@ static void cftmdl_16fx( 
        j1 = add(j,l);

        j2 = add(j1,l);

        j3 = add(j2,l);

#ifdef BASOP_NOGLOB

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

        x0i = add_o(a[j + 1],a[j1 + 1], &Overflow);

        x1r = sub_o(a[j],a[j1], &Overflow);

        x1i = sub_o(a[j + 1],a[j1 + 1], &Overflow);

        x2r = add_o(a[j2],a[j3], &Overflow);

        x2i = add_o(a[j2 + 1],a[j3 + 1], &Overflow);

        x3r = sub_o(a[j2],a[j3], &Overflow);

        x3i = sub_o(a[j2 + 1],a[j3 + 1], &Overflow);

        a[j] = add_o(x0r,x2r, &Overflow);

        move16();

        a[j + 1] = add_o(x0i,x2i, &Overflow);

        move16();

        a[j2] = sub_o(x0r,x2r, &Overflow);

        move16();

        a[j2 + 1] = sub_o(x0i,x2i, &Overflow);

        move16();

        a[j1] = sub_o(x1r,x3i, &Overflow);

        move16();

        a[j1 + 1] = add_o(x1i,x3r, &Overflow);

        move16();

        a[j3] = add_o(x1r,x3i, &Overflow);

        move16();

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

        move16();

#else

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

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

        x1r = sub(a[j],a[j1]);
@@ -1984,6 +2111,7 @@ static void cftmdl_16fx( 
        move16();

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

        move16();

#endif

    }



    wk1r = w[2];
@@ -1994,6 +2122,28 @@ static void cftmdl_16fx( 
        j1 = add(j,l);

        j2 = add(j1,l);

        j3 = add(j2,l);

#ifdef BASOP_NOGLOB

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

        x0i = add_o(a[j + 1],a[j1 + 1], &Overflow);

        x1r = sub_o(a[j],a[j1], &Overflow);

        x1i = sub_o(a[j + 1],a[j1 + 1], &Overflow);

        x2r = add_o(a[j2],a[j3], &Overflow);

        x2i = add_o(a[j2 + 1],a[j3 + 1], &Overflow);

        x3r = sub_o(a[j2],a[j3], &Overflow);

        x3i = sub_o(a[j2 + 1],a[j3 + 1], &Overflow);

        a[j] = add_o(x0r,x2r, &Overflow);

        move16();

        a[j + 1] = add_o(x0i,x2i, &Overflow);

        move16();

        a[j2] = sub_o(x2i,x0i, &Overflow);

        move16();

        a[j2 + 1] = sub_o(x0r,x2r, &Overflow);

        move16();



        x0r = sub_o(x1r,x3i, &Overflow);

        x0i = add_o(x1i,x3r, &Overflow);

        tmp = sub_o(x0r,x0i, &Overflow);

#else

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

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

        x1r = sub(a[j],a[j1]);
@@ -2014,22 +2164,46 @@ static void cftmdl_16fx( 
        x0r = sub(x1r,x3i);

        x0i = add(x1i,x3r);

        tmp = sub(x0r,x0i);

#endif



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

#ifdef BASOP_NOGLOB

        a[j1] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */

        tmp = add_o(x0r, x0i, &Overflow);

#else

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



        tmp = add(x0r, x0i);

#endif

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

#ifdef BASOP_NOGLOB

        a[j1 + 1] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */



        x0r = add_o(x3i,x1r, &Overflow);

        x0i = sub_o(x3r,x1i, &Overflow);

        tmp = sub_o(x0i,x0r, &Overflow);

#else

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



        x0r = add(x3i,x1r);

        x0i = sub(x3r,x1i);

        tmp = sub(x0i,x0r);

#endif

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

#ifdef BASOP_NOGLOB

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



        tmp = add_o(x0i,x0r, &Overflow);

#else

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



        tmp = add(x0i,x0r);

#endif

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

#ifdef BASOP_NOGLOB

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

#else

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

#endif

    }



    k1 = 0;
@@ -2107,18 +2281,42 @@ static void cftmdl_16fx( 
        move32();

        wk1i = w[k2 + 3];

        move32();

#ifdef BASOP_NOGLOB

        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 */

#else

        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 */



#endif

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

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

        {

            j1 = add(j,l);

            j2 = add(j1,l);

            j3 = add(j2,l);

#ifdef BASOP_NOGLOB

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

            x0i = add_o(a[j + 1],a[j1 + 1], &Overflow);

            x1r = sub_o(a[j],a[j1], &Overflow);

            x1i = sub_o(a[j + 1],a[j1 + 1], &Overflow);

            x2r = add_o(a[j2],a[j3], &Overflow);

            x2i = add_o(a[j2 + 1],a[j3 + 1], &Overflow);

            x3r = sub_o(a[j2],a[j3], &Overflow);

            x3i = sub_o(a[j2 + 1],a[j3 + 1], &Overflow);

            a[j] = add_o(x0r,x2r, &Overflow);

            move16();

            a[j + 1] = add_o(x0i,x2i, &Overflow);

            move16();



            x0r = sub_o(x0r,x2r, &Overflow);

            x0i = sub_o(x0i,x2i, &Overflow);

#else

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

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

            x1r = sub(a[j],a[j1]);
@@ -2134,38 +2332,63 @@ static void cftmdl_16fx( 


            x0r = sub(x0r,x2r);

            x0i = sub(x0i,x2i);



#endif

            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) */

#ifdef BASOP_NOGLOB

            a[j2] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */

#else

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



#endif

            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) */

#ifdef BASOP_NOGLOB

            a[j2 + 1] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */



            x0r = sub_o(x1r,x3i, &Overflow);

            x0i = add_o(x1i,x3r, &Overflow);

#else

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



            x0r = sub(x1r,x3i);

            x0i = add(x1i,x3r);



#endif

            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) */

#ifdef BASOP_NOGLOB

            a[j1] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */

#else

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



#endif

            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) */

#ifdef BASOP_NOGLOB

            a[j1 + 1] = round_fx_o(L_shl_o(L_tmp,1, &Overflow), &Overflow); /*Q(Qx+Q_edct) */



            x0r = add_o(x1r,x3i, &Overflow);

            x0i = sub_o(x1i,x3r, &Overflow);

#else

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



            x0r = add(x1r,x3i);

            x0i = sub(x1i,x3r);



#endif

            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) */

#ifdef BASOP_NOGLOB

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

#else

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



#endif

            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) */

#ifdef BASOP_NOGLOB

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

#else

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

#endif

        }

    }