void M_4_byte_multiply(UCHAR *r, UCHAR *a, UCHAR *b)
{
int b0, jj;
UCHAR *cp1, *cp2, numdiv, numrem;
memset(r, 0, 8); /* zero out 8 byte result */
jj = 3;
/* loop for one number [b], un-roll the inner 'loop' [a] */
while (1)
{
b0 = (int)b[jj];
cp1 = r + (3 + jj);
cp2 = cp1 + 1;
M_get_div_rem((b0 * a[3]), &numdiv, &numrem);
*cp2 += numrem;
*cp1 += numdiv;
if (*cp2 >= 100)
{
*cp2 -= 100;
*cp1 += 1;
}
cp1--;
cp2--;
if (*cp2 >= 100)
{
*cp2 -= 100;
*cp1 += 1;
}
M_get_div_rem((b0 * a[2]), &numdiv, &numrem);
*cp2 += numrem;
*cp1 += numdiv;
if (*cp2 >= 100)
{
*cp2 -= 100;
*cp1 += 1;
}
cp1--;
cp2--;
if (*cp2 >= 100)
{
*cp2 -= 100;
*cp1 += 1;
}
M_get_div_rem((b0 * a[1]), &numdiv, &numrem);
*cp2 += numrem;
*cp1 += numdiv;
if (*cp2 >= 100)
{
*cp2 -= 100;
*cp1 += 1;
}
cp1--;
cp2--;
if (*cp2 >= 100)
{
*cp2 -= 100;
*cp1 += 1;
}
M_get_div_rem((b0 * a[0]), &numdiv, &numrem);
*cp2 += numrem;
*cp1 += numdiv;
if (*cp2 >= 100)
{
*cp2 -= 100;
*cp1 += 1;
}
if (jj-- == 0)
break;
}
}
void M_fast_mul_fft(UCHAR *ww, UCHAR *uu, UCHAR *vv, int nbytes)
{
int mflag, i, j, nn2, nn;
double carry, nnr, dtemp, *a, *b;
UCHAR *w0;
unsigned long ul;
if (M_size < 0) /* if first time in, setup working arrays */
{
if (M_get_sizeof_int() == 2) /* if still using 16 bit compilers */
M_size = 1030;
else
M_size = 8200;
M_aa_array = (double *)MAPM_MALLOC(M_size * sizeof(double));
M_bb_array = (double *)MAPM_MALLOC(M_size * sizeof(double));
if ((M_aa_array == NULL) || (M_bb_array == NULL))
{
/* fatal, this does not return */
M_apm_log_error_msg(M_APM_EXIT, "\'M_fast_mul_fft\', Out of memory");
}
}
nn = nbytes;
nn2 = nbytes >> 1;
if (nn > M_size)
{
mflag = TRUE;
a = (double *)MAPM_MALLOC((nn + 8) * sizeof(double));
b = (double *)MAPM_MALLOC((nn + 8) * sizeof(double));
if ((a == NULL) || (b == NULL))
{
/* fatal, this does not return */
M_apm_log_error_msg(M_APM_EXIT, "\'M_fast_mul_fft\', Out of memory");
}
}
else
{
mflag = FALSE;
a = M_aa_array;
b = M_bb_array;
}
/*
* convert normal base 100 MAPM numbers to base 10000
* for the FFT operation.
*/
i = 0;
for (j=0; j < nn2; j++)
{
a[j] = (double)((int)uu[i] * 100 + uu[i+1]);
b[j] = (double)((int)vv[i] * 100 + vv[i+1]);
i += 2;
}
/* zero fill the second half of the arrays */
for (j=nn2; j < nn; j++)
{
a[j] = 0.0;
b[j] = 0.0;
}
/* perform the forward Fourier transforms for both numbers */
M_rdft(nn, 1, a);
M_rdft(nn, 1, b);
/* perform the convolution ... */
b[0] *= a[0];
b[1] *= a[1];
for (j=3; j <= nn; j += 2)
{
dtemp = b[j-1];
b[j-1] = dtemp * a[j-1] - b[j] * a[j];
b[j] = dtemp * a[j] + b[j] * a[j-1];
}
/* perform the inverse transform on the result */
M_rdft(nn, -1, b);
/* perform a final pass to release all the carries */
/* we are still in base 10000 at this point */
carry = 0.0;
j = nn;
nnr = 2.0 / (double)nn;
while (1)
{
dtemp = b[--j] * nnr + carry + 0.5;
ul = (unsigned long)(dtemp * 1.0E-4);
carry = (double)ul;
b[j] = dtemp - carry * 10000.0;
if (j == 0)
break;
}
/* copy result to our destination after converting back to base 100 */
w0 = ww;
M_get_div_rem((int)ul, w0, (w0 + 1));
for (j=0; j <= (nn - 2); j++)
{
w0 += 2;
M_get_div_rem((int)b[j], w0, (w0 + 1));
}
if (mflag)
{
MAPM_FREE(b);
MAPM_FREE(a);
}
}