308 lines
7.6 KiB
C
308 lines
7.6 KiB
C
|
/*=============================================================================
|
||
|
|
|
||
|
|
This file is part of FLINT.
|
||
|
|
|
||
|
|
FLINT is free software; you can redistribute it and/or modify
|
||
|
|
it under the terms of the GNU General Public License as published by
|
||
|
|
the Free Software Foundation; either version 2 of the License, or
|
||
|
|
(at your option) any later version.
|
||
|
|
|
||
|
|
FLINT is distributed in the hope that it will be useful,
|
||
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||
|
|
GNU General Public License for more details.
|
||
|
|
|
||
|
|
You should have received a copy of the GNU General Public License
|
||
|
|
along with FLINT; if not, write to the Free Software
|
||
|
|
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
|
||
|
|
|
||
|
|
=============================================================================*/
|
||
|
|
/******************************************************************************
|
||
|
|
|
||
|
|
Copyright (C) 2011 Fredrik Johansson
|
||
|
|
|
||
|
|
******************************************************************************/
|
||
|
|
|
||
|
|
#include "arith.h"
|
||
|
|
|
||
|
|
static const int mod4_tab[8] = { 2, 1, 3, 0, 0, 3, 1, 2 };
|
||
|
|
|
||
|
|
static const int gcd24_tab[24] = {
|
||
|
|
24, 1, 2, 3, 4, 1, 6, 1, 8, 3, 2, 1,
|
||
|
|
12, 1, 2, 3, 8, 1, 6, 1, 4, 3, 2, 1
|
||
|
|
};
|
||
|
|
|
||
|
|
static mp_limb_t
|
||
|
|
n_sqrtmod_2exp(mp_limb_t a, int k)
|
||
|
|
{
|
||
|
|
mp_limb_t x;
|
||
|
|
int i;
|
||
|
|
|
||
|
|
if (a == 0 || k == 0)
|
||
|
|
return 0;
|
||
|
|
|
||
|
|
if (k == 1)
|
||
|
|
return 1;
|
||
|
|
|
||
|
|
if (k == 2)
|
||
|
|
{
|
||
|
|
if (a == 1)
|
||
|
|
return 1;
|
||
|
|
return 0;
|
||
|
|
}
|
||
|
|
|
||
|
|
x = 1;
|
||
|
|
for (i = 3; i < k; i++)
|
||
|
|
x += (a - x * x) / 2;
|
||
|
|
|
||
|
|
if (k < FLINT_BITS)
|
||
|
|
x &= ((UWORD(1) << k) - 1);
|
||
|
|
|
||
|
|
return x;
|
||
|
|
}
|
||
|
|
|
||
|
|
static mp_limb_t
|
||
|
|
n_sqrtmod_ppow(mp_limb_t a, mp_limb_t p, int k, mp_limb_t pk, mp_limb_t pkinv)
|
||
|
|
{
|
||
|
|
mp_limb_t r, t;
|
||
|
|
int i;
|
||
|
|
|
||
|
|
r = n_sqrtmod(a, p);
|
||
|
|
if (r == 0)
|
||
|
|
return r;
|
||
|
|
|
||
|
|
i = 1;
|
||
|
|
while (i < k)
|
||
|
|
{
|
||
|
|
t = n_mulmod2_preinv(r, r, pk, pkinv);
|
||
|
|
t = n_submod(t, a, pk);
|
||
|
|
t = n_mulmod2_preinv(t, n_invmod(n_addmod(r, r, pk), pk), pk, pkinv);
|
||
|
|
r = n_submod(r, t, pk);
|
||
|
|
i *= 2;
|
||
|
|
}
|
||
|
|
|
||
|
|
return r;
|
||
|
|
}
|
||
|
|
|
||
|
|
void
|
||
|
|
trigprod_mul_prime_power(trig_prod_t prod, mp_limb_t k, mp_limb_t n,
|
||
|
|
mp_limb_t p, int exp)
|
||
|
|
{
|
||
|
|
mp_limb_t m, mod, inv;
|
||
|
|
|
||
|
|
if (k <= 3)
|
||
|
|
{
|
||
|
|
if (k == 0)
|
||
|
|
{
|
||
|
|
prod->prefactor = 0;
|
||
|
|
}
|
||
|
|
else if (k == 2 && (n % 2 == 1))
|
||
|
|
{
|
||
|
|
prod->prefactor *= -1;
|
||
|
|
}
|
||
|
|
else if (k == 3)
|
||
|
|
{
|
||
|
|
switch (n % 3)
|
||
|
|
{
|
||
|
|
case 0:
|
||
|
|
prod->prefactor *= 2;
|
||
|
|
prod->cos_p[prod->n] = 1;
|
||
|
|
prod->cos_q[prod->n] = 18;
|
||
|
|
break;
|
||
|
|
case 1:
|
||
|
|
prod->prefactor *= -2;
|
||
|
|
prod->cos_p[prod->n] = 7;
|
||
|
|
prod->cos_q[prod->n] = 18;
|
||
|
|
break;
|
||
|
|
case 2:
|
||
|
|
prod->prefactor *= -2;
|
||
|
|
prod->cos_p[prod->n] = 5;
|
||
|
|
prod->cos_q[prod->n] = 18;
|
||
|
|
break;
|
||
|
|
}
|
||
|
|
prod->n++;
|
||
|
|
}
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
|
||
|
|
/* Power of 2 */
|
||
|
|
if (p == 2)
|
||
|
|
{
|
||
|
|
mod = 8 * k;
|
||
|
|
inv = n_preinvert_limb(mod);
|
||
|
|
|
||
|
|
m = n_submod(1, n_mod2_preinv(24 * n, mod, inv), mod);
|
||
|
|
m = n_sqrtmod_2exp(m, exp + 3);
|
||
|
|
m = n_mulmod2_preinv(m, n_invmod(3, mod), mod, inv);
|
||
|
|
|
||
|
|
prod->prefactor *= n_jacobi(-1, m);
|
||
|
|
if (exp % 2 == 1)
|
||
|
|
prod->prefactor *= -1;
|
||
|
|
prod->sqrt_p *= k;
|
||
|
|
prod->cos_p[prod->n] = (mp_limb_signed_t)(k - m);
|
||
|
|
prod->cos_q[prod->n] = 2 * k;
|
||
|
|
prod->n++;
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
|
||
|
|
/* Power of 3 */
|
||
|
|
if (p == 3)
|
||
|
|
{
|
||
|
|
mod = 3 * k;
|
||
|
|
inv = n_preinvert_limb(mod);
|
||
|
|
|
||
|
|
m = n_submod(1, n_mod2_preinv(24 * n, mod, inv), mod);
|
||
|
|
m = n_sqrtmod_ppow(m, p, exp + 1, mod, inv);
|
||
|
|
m = n_mulmod2_preinv(m, n_invmod(8, mod), mod, inv);
|
||
|
|
|
||
|
|
prod->prefactor *= (2 * n_jacobi_unsigned(m, 3));
|
||
|
|
if (exp % 2 == 0)
|
||
|
|
prod->prefactor *= -1;
|
||
|
|
prod->sqrt_p *= k;
|
||
|
|
prod->sqrt_q *= 3;
|
||
|
|
prod->cos_p[prod->n] = (mp_limb_signed_t)(3 * k - 8 * m);
|
||
|
|
prod->cos_q[prod->n] = 6 * k;
|
||
|
|
prod->n++;
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
|
||
|
|
/* Power of prime greater than 3 */
|
||
|
|
inv = n_preinvert_limb(k);
|
||
|
|
m = n_submod(1, n_mod2_preinv(24 * n, k, inv), k);
|
||
|
|
|
||
|
|
if (m % p == 0)
|
||
|
|
{
|
||
|
|
if (exp == 1)
|
||
|
|
{
|
||
|
|
prod->prefactor *= n_jacobi(3, k);
|
||
|
|
prod->sqrt_p *= k;
|
||
|
|
}
|
||
|
|
else
|
||
|
|
prod->prefactor = 0;
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
|
||
|
|
m = n_sqrtmod_ppow(m, p, exp, k, inv);
|
||
|
|
|
||
|
|
if (m == 0)
|
||
|
|
{
|
||
|
|
prod->prefactor = 0;
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
|
||
|
|
prod->prefactor *= 2;
|
||
|
|
prod->prefactor *= n_jacobi(3, k);
|
||
|
|
prod->sqrt_p *= k;
|
||
|
|
prod->cos_p[prod->n] = 4 * n_mulmod2_preinv(m, n_invmod(24, k), k, inv);
|
||
|
|
prod->cos_q[prod->n] = k;
|
||
|
|
prod->n++;
|
||
|
|
}
|
||
|
|
|
||
|
|
/*
|
||
|
|
Solve (k2^2 * d2 * e) * n1 = (d2 * e * n + (k2^2 - 1) / d1) mod k2
|
||
|
|
|
||
|
|
TODO: test this on 32 bit
|
||
|
|
*/
|
||
|
|
static mp_limb_t
|
||
|
|
solve_n1(mp_limb_t n, mp_limb_t k1, mp_limb_t k2,
|
||
|
|
mp_limb_t d1, mp_limb_t d2, mp_limb_t e)
|
||
|
|
{
|
||
|
|
mp_limb_t inv, n1, u, t[2];
|
||
|
|
|
||
|
|
inv = n_preinvert_limb(k1);
|
||
|
|
|
||
|
|
umul_ppmm(t[1], t[0], k2, k2);
|
||
|
|
sub_ddmmss(t[1], t[0], t[1], t[0], UWORD(0), UWORD(1));
|
||
|
|
mpn_divrem_1(t, 0, t, 2, d1);
|
||
|
|
|
||
|
|
n1 = n_ll_mod_preinv(t[1], t[0], k1, inv);
|
||
|
|
n1 = n_mod2_preinv(n1 + d2*e*n, k1, inv);
|
||
|
|
|
||
|
|
u = n_mulmod2_preinv(k2, k2, k1, inv);
|
||
|
|
u = n_invmod(u * d2 * e, k1);
|
||
|
|
n1 = n_mulmod2_preinv(n1, u, k1, inv);
|
||
|
|
|
||
|
|
return n1;
|
||
|
|
}
|
||
|
|
|
||
|
|
|
||
|
|
void
|
||
|
|
arith_hrr_expsum_factored(trig_prod_t prod, mp_limb_t k, mp_limb_t n)
|
||
|
|
{
|
||
|
|
n_factor_t fac;
|
||
|
|
int i;
|
||
|
|
|
||
|
|
if (k <= 1)
|
||
|
|
{
|
||
|
|
prod->prefactor = k;
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
|
||
|
|
n_factor_init(&fac);
|
||
|
|
n_factor(&fac, k, 0);
|
||
|
|
|
||
|
|
/* Repeatedly factor A_k(n) into A_k1(n1)*A_k2(n2) with k1, k2 coprime */
|
||
|
|
for (i = 0; i + 1 < fac.num && prod->prefactor != 0; i++)
|
||
|
|
{
|
||
|
|
mp_limb_t p, k1, k2, inv, n1, n2;
|
||
|
|
|
||
|
|
p = fac.p[i];
|
||
|
|
|
||
|
|
/* k = 2 * k1 with k1 odd */
|
||
|
|
if (p == UWORD(2) && fac.exp[i] == 1)
|
||
|
|
{
|
||
|
|
k2 = k / 2;
|
||
|
|
inv = n_preinvert_limb(k2);
|
||
|
|
|
||
|
|
n2 = n_invmod(32, k2);
|
||
|
|
n2 = n_mulmod2_preinv(n2,
|
||
|
|
n_mod2_preinv(8*n + 1, k2, inv), k2, inv);
|
||
|
|
n1 = ((k2 % 8 == 3) || (k2 % 8 == 5)) ^ (n & 1);
|
||
|
|
|
||
|
|
trigprod_mul_prime_power(prod, 2, n1, 2, 1);
|
||
|
|
k = k2;
|
||
|
|
n = n2;
|
||
|
|
}
|
||
|
|
/* k = 4 * k1 with k1 odd */
|
||
|
|
else if (p == UWORD(2) && fac.exp[i] == 2)
|
||
|
|
{
|
||
|
|
k2 = k / 4;
|
||
|
|
inv = n_preinvert_limb(k2);
|
||
|
|
|
||
|
|
n2 = n_invmod(128, k2);
|
||
|
|
n2 = n_mulmod2_preinv(n2,
|
||
|
|
n_mod2_preinv(8*n + 5, k2, inv), k2, inv);
|
||
|
|
n1 = (n + mod4_tab[(k2 / 2) % 8]) % 4;
|
||
|
|
|
||
|
|
trigprod_mul_prime_power(prod, 4, n1, 2, 2);
|
||
|
|
prod->prefactor *= -1;
|
||
|
|
k = k2;
|
||
|
|
n = n2;
|
||
|
|
}
|
||
|
|
/* k = k1 * k2 with k1 odd or divisible by 8 */
|
||
|
|
else
|
||
|
|
{
|
||
|
|
mp_limb_t d1, d2, e;
|
||
|
|
|
||
|
|
k1 = n_pow(fac.p[i], fac.exp[i]);
|
||
|
|
k2 = k / k1;
|
||
|
|
|
||
|
|
d1 = gcd24_tab[k1 % 24];
|
||
|
|
d2 = gcd24_tab[k2 % 24];
|
||
|
|
e = 24 / (d1 * d2);
|
||
|
|
|
||
|
|
n1 = solve_n1(n, k1, k2, d1, d2, e);
|
||
|
|
n2 = solve_n1(n, k2, k1, d2, d1, e);
|
||
|
|
|
||
|
|
trigprod_mul_prime_power(prod, k1, n1, fac.p[i], fac.exp[i]);
|
||
|
|
k = k2;
|
||
|
|
n = n2;
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
if (fac.num != 0 && prod->prefactor != 0)
|
||
|
|
trigprod_mul_prime_power(prod, k, n,
|
||
|
|
fac.p[fac.num - 1], fac.exp[fac.num - 1]);
|
||
|
|
|
||
|
|
}
|