void
fmpz_poly_evaluate_fmpz(fmpz_t res, const fmpz_poly_t f, const fmpz_t a)
{
if (res == a)
{
fmpz_t t;
fmpz_init(t);
_fmpz_poly_evaluate_fmpz(t, f->coeffs, f->length, a);
fmpz_swap(res, t);
fmpz_clear(t);
}
else
_fmpz_poly_evaluate_fmpz(res, f->coeffs, f->length, a);
}
void
_fmpz_poly_compose(fmpz * res, const fmpz * poly1, slong len1,
const fmpz * poly2, slong len2)
{
if (len1 == 1)
fmpz_set(res, poly1);
else if (len2 == 1)
_fmpz_poly_evaluate_fmpz(res, poly1, len1, poly2);
else if (len1 <= 4)
_fmpz_poly_compose_horner(res, poly1, len1, poly2, len2);
else
_fmpz_poly_compose_divconquer(res, poly1, len1, poly2, len2);
}
void
_fmpz_poly_compose_divconquer(fmpz * res, const fmpz * poly1, long len1,
const fmpz * poly2, long len2)
{
long i, j, k, n;
long *hlen, alloc, powlen;
fmpz *v, **h, *pow, *temp;
if (len1 == 1)
{
fmpz_set(res, poly1);
return;
}
if (len2 == 1)
{
_fmpz_poly_evaluate_fmpz(res, poly1, len1, poly2);
return;
}
if (len1 == 2)
{
_fmpz_poly_compose_horner(res, poly1, len1, poly2, len2);
return;
}
/* Initialisation */
hlen = (long *) malloc(((len1 + 1) / 2) * sizeof(long));
for (k = 1; (2 << k) < len1; k++) ;
hlen[0] = hlen[1] = ((1 << k) - 1) * (len2 - 1) + 1;
for (i = k - 1; i > 0; i--)
{
long hi = (len1 + (1 << i) - 1) / (1 << i);
for (n = (hi + 1) / 2; n < hi; n++)
hlen[n] = ((1 << i) - 1) * (len2 - 1) + 1;
}
powlen = (1 << k) * (len2 - 1) + 1;
alloc = 0;
for (i = 0; i < (len1 + 1) / 2; i++)
alloc += hlen[i];
v = _fmpz_vec_init(alloc + 2 * powlen);
h = (fmpz **) malloc(((len1 + 1) / 2) * sizeof(fmpz *));
h[0] = v;
for (i = 0; i < (len1 - 1) / 2; i++)
{
h[i + 1] = h[i] + hlen[i];
hlen[i] = 0;
}
hlen[(len1 - 1) / 2] = 0;
pow = v + alloc;
temp = pow + powlen;
/* Let's start the actual work */
for (i = 0, j = 0; i < len1 / 2; i++, j += 2)
{
if (poly1[j + 1] != 0L)
{
_fmpz_vec_scalar_mul_fmpz(h[i], poly2, len2, poly1 + j + 1);
fmpz_add(h[i], h[i], poly1 + j);
hlen[i] = len2;
}
else if (poly1[j] != 0L)
{
fmpz_set(h[i], poly1 + j);
hlen[i] = 1;
}
}
if ((len1 & 1L))
{
if (poly1[j] != 0L)
{
fmpz_set(h[i], poly1 + j);
hlen[i] = 1;
}
}
_fmpz_poly_mul(pow, poly2, len2, poly2, len2);
powlen = 2 * len2 - 1;
for (n = (len1 + 1) / 2; n > 2; n = (n + 1) / 2)
{
if (hlen[1] > 0)
{
long templen = powlen + hlen[1] - 1;
_fmpz_poly_mul(temp, pow, powlen, h[1], hlen[1]);
_fmpz_poly_add(h[0], temp, templen, h[0], hlen[0]);
hlen[0] = FLINT_MAX(hlen[0], templen);
}
for (i = 1; i < n / 2; i++)
{
if (hlen[2*i + 1] > 0)
{
_fmpz_poly_mul(h[i], pow, powlen, h[2*i + 1], hlen[2*i + 1]);
hlen[i] = hlen[2*i + 1] + powlen - 1;
} else
hlen[i] = 0;
_fmpz_poly_add(h[i], h[i], hlen[i], h[2*i], hlen[2*i]);
hlen[i] = FLINT_MAX(hlen[i], hlen[2*i]);
}
if ((n & 1L))
{
_fmpz_vec_set(h[i], h[2*i], hlen[2*i]);
hlen[i] = hlen[2*i];
}
_fmpz_poly_mul(temp, pow, powlen, pow, powlen);
powlen += powlen - 1;
{
fmpz * t = pow;
pow = temp;
temp = t;
}
}
_fmpz_poly_mul(res, pow, powlen, h[1], hlen[1]);
_fmpz_vec_add(res, res, h[0], hlen[0]);
_fmpz_vec_clear(v, alloc + 2 * powlen);
free(h);
free(hlen);
}