983 lines
21 KiB
C
983 lines
21 KiB
C
/*
|
|
Copyright 2009 Jason Moxham
|
|
Copyright (C) 2008 Peter Shrimpton
|
|
Copyright (C) 2008, 2009 William Hart
|
|
|
|
This file is part of the MPIR Library.
|
|
|
|
The MPIR Library is free software; you can redistribute it and/or modify
|
|
it under the terms of the GNU Lesser General Public License as published
|
|
by the Free Software Foundation; either version 2.1 of the License, or (at
|
|
your option) any later version.
|
|
|
|
The MPIR Library 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 Lesser General Public
|
|
License for more details.
|
|
|
|
You should have received a copy of the GNU Lesser General Public License
|
|
along with the MPIR Library; see the file COPYING.LIB. If not, write
|
|
to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
|
|
Boston, MA 02110-1301, USA.
|
|
*/
|
|
|
|
#include "mpir.h"
|
|
#include "gmp-impl.h"
|
|
#include "longlong.h"
|
|
|
|
#if GMP_LIMB_BITS == 32 || GMP_LIMB_BITS == 64
|
|
|
|
#if GMP_LIMB_BITS == 32
|
|
#define D_BITS 31
|
|
#else
|
|
#define D_BITS 53
|
|
#endif
|
|
|
|
typedef struct pair_s
|
|
{
|
|
mp_limb_t x, y;
|
|
} n_pair_t;
|
|
|
|
#define r_shift(in, shift) \
|
|
((shift == GMP_LIMB_BITS) ? CNST_LIMB(0) : ((in)>>(shift)))
|
|
|
|
mp_limb_t mpir_sqrt(mp_limb_t r)
|
|
{
|
|
mp_limb_t res, is;
|
|
|
|
#if GMP_LIMB_BITS == 32
|
|
float x, z;
|
|
union {
|
|
float f;
|
|
mp_limb_t l;
|
|
} temp;
|
|
|
|
mp_limb_t bits32 = (r & GMP_LIMB_HIGHBIT);
|
|
mp_limb_t r2;
|
|
/* algorithm can't handle 32 bits */
|
|
if (bits32)
|
|
{
|
|
r2 = r;
|
|
r >>= 2;
|
|
}
|
|
|
|
temp.f = (float) r;
|
|
temp.l = (CNST_LIMB(0xbe6ec85e) - temp.l)>>1; // estimate of 1/sqrt(y)
|
|
x = temp.f;
|
|
z = (float) r*0.5;
|
|
x = (1.5*x) - (x*x)*(x*z);
|
|
x = (1.5*x) - (x*x)*(x*z);
|
|
x = (1.5*x) - (x*x)*(x*z);
|
|
x = (1.5*x) - (x*x)*(x*z);
|
|
is = (mp_limb_t) (x*(double) r);
|
|
res = is + ((is+1)*(is+1) <= r);
|
|
if (!bits32) return res - (res*res > r);
|
|
else
|
|
{
|
|
mp_limb_t sq;
|
|
res = res - (res*res > r);
|
|
res <<= 1;
|
|
sq = res*res;
|
|
res = res - ((sq > r2) || ((sq ^ r2) & GMP_LIMB_HIGHBIT));
|
|
sq = (res + 1)*(res + 1);
|
|
res = res + ((sq <= r2) && !((sq ^ r2) & GMP_LIMB_HIGHBIT));
|
|
return res;
|
|
}
|
|
#else
|
|
|
|
double x, z;
|
|
union {
|
|
double f;
|
|
mp_limb_t l;
|
|
} temp;
|
|
|
|
mp_limb_t bits64 = (r & GMP_LIMB_HIGHBIT);
|
|
mp_limb_t r2;
|
|
/* algorithm can't handle 64 bits */
|
|
if (bits64)
|
|
{
|
|
r2 = r;
|
|
r >>= 2;
|
|
}
|
|
|
|
temp.f = (double) r;
|
|
temp.l = (CNST_LIMB(0xbfcdd90a00000000) - temp.l)>>1; /* estimate of 1/sqrt(y) */
|
|
x = temp.f;
|
|
z = (double) r*0.5;
|
|
x = (1.5*x) - (x*x)*(x*z);
|
|
x = (1.5*x) - (x*x)*(x*z);
|
|
x = (1.5*x) - (x*x)*(x*z);
|
|
x = (1.5*x) - (x*x)*(x*z);
|
|
x = (1.5*x) - (x*x)*(x*z);
|
|
is = (mp_limb_t) (x*(double) r);
|
|
res = is + ((is+1)*(is+1) <= r);
|
|
if (!bits64) return res - (res*res > r);
|
|
else
|
|
{
|
|
mp_limb_t sq;
|
|
res = res - (res*res > r);
|
|
res <<= 1;
|
|
sq = res*res;
|
|
res = res - ((sq > r2) || ((sq ^ r2) & GMP_LIMB_HIGHBIT));
|
|
sq = (res + 1)*(res + 1);
|
|
res = res + ((sq <= r2) && !((sq ^ r2) & GMP_LIMB_HIGHBIT));
|
|
return res;
|
|
}
|
|
#endif
|
|
}
|
|
|
|
static
|
|
double n_precompute_inverse(mp_limb_t n)
|
|
{
|
|
return (double) 1 / (double) n;
|
|
}
|
|
|
|
static
|
|
unsigned int BIT_COUNT(mp_limb_t x)
|
|
{
|
|
unsigned int zeros = GMP_LIMB_BITS;
|
|
if (x) count_leading_zeros(zeros, x);
|
|
return GMP_LIMB_BITS - zeros;
|
|
}
|
|
|
|
static
|
|
int mod64[64] = {1,1,0,0,1,0,0,0,0,1,0,0,0,0,0,0,1,1,0,0,0,
|
|
0,0,0,0,1,0,0,0,0,0,0,0,1,0,0,1,0,0,0,0,1,
|
|
0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,1,0,0,0,0,0,0};
|
|
|
|
static
|
|
int mod65[65] = {1,1,0,0,1,0,0,0,0,1,1,0,0,0,1,0,1,0,0,0,0,0,
|
|
0,0,0,1,1,0,0,1,1,0,0,0,0,1,1,0,0,1,1,0,0,0,
|
|
0,0,0,0,0,1,0,1,0,0,0,1,1,0,0,0,0,1,0,0,1};
|
|
|
|
static
|
|
int mod63[63] = {1,1,0,0,1,0,0,1,0,1,0,0,0,0,1,0,1,0,1,0,0,
|
|
0,1,0,0,1,0,0,1,0,0,0,0,0,0,1,1,1,0,0,0,0,
|
|
0,1,0,0,1,0,0,1,0,0,0,0,0,0,1,0,1,0,0,0,0};
|
|
|
|
static
|
|
int n_is_square(mp_limb_t x)
|
|
{
|
|
mp_limb_t sq;
|
|
if (!mod64[x%CNST_LIMB(64)]) return 0;
|
|
if (!mod63[x%CNST_LIMB(63)]) return 0;
|
|
if (!mod65[x%CNST_LIMB(65)]) return 0;
|
|
|
|
sq = mpir_sqrt(x);
|
|
|
|
return (x == sq*sq);
|
|
}
|
|
|
|
static
|
|
mp_limb_t n_preinvert_limb(mp_limb_t n)
|
|
{
|
|
mp_limb_t norm, ninv;
|
|
|
|
count_leading_zeros(norm, n);
|
|
invert_limb(ninv, n<<norm);
|
|
|
|
return ninv;
|
|
}
|
|
|
|
static
|
|
mp_limb_t n_addmod(mp_limb_t x, mp_limb_t y, mp_limb_t n)
|
|
{
|
|
if (n - y > x) return x + y;
|
|
else return x + y - n;
|
|
}
|
|
|
|
static inline
|
|
mp_limb_t n_submod(mp_limb_t x, mp_limb_t y, mp_limb_t n)
|
|
{
|
|
if (y > x) return x - y + n;
|
|
else return x - y;
|
|
}
|
|
|
|
static
|
|
mp_limb_t n_mod2_preinv(mp_limb_t a, mp_limb_t n, mp_limb_t ninv)
|
|
{
|
|
unsigned int norm;
|
|
mp_limb_t q, r;
|
|
|
|
count_leading_zeros(norm, n);
|
|
udiv_qrnnd_preinv(q, r, r_shift(a, GMP_LIMB_BITS-norm), a<<norm, n<<norm, ninv);
|
|
return (r>>norm);
|
|
}
|
|
|
|
static
|
|
mp_limb_t n_ll_mod_preinv(mp_limb_t a_hi, mp_limb_t a_lo,
|
|
mp_limb_t n, mp_limb_t ninv)
|
|
{
|
|
mp_limb_t q, r, norm;
|
|
|
|
if (a_hi > n) a_hi = n_mod2_preinv(a_hi, n, ninv);
|
|
|
|
count_leading_zeros(norm, n);
|
|
|
|
udiv_qrnnd_preinv(q, r, (a_hi<<norm) +
|
|
r_shift(a_lo, GMP_LIMB_BITS-norm), a_lo<<norm, n<<norm, ninv);
|
|
|
|
return (r>>norm);
|
|
}
|
|
|
|
static
|
|
mp_limb_t n_mulmod_precomp(mp_limb_t a, mp_limb_t b, mp_limb_t n, double npre)
|
|
{
|
|
mp_limb_t quot = (mp_limb_t) ((double) a * (double) b * npre);
|
|
mp_limb_signed_t rem = a*b - quot*n;
|
|
if (rem < 0)
|
|
{
|
|
rem += n;
|
|
if (rem < 0) return rem + n;
|
|
} else if (rem >= n) return rem - n;
|
|
return rem;
|
|
}
|
|
|
|
static
|
|
mp_limb_t n_mulmod2_preinv(mp_limb_t a, mp_limb_t b, mp_limb_t n, mp_limb_t ninv)
|
|
{
|
|
mp_limb_t p1, p2;
|
|
|
|
umul_ppmm(p1, p2, a, b);
|
|
|
|
return n_ll_mod_preinv(p1, p2, n, ninv);
|
|
}
|
|
|
|
static
|
|
mp_limb_t n_powmod_precomp(mp_limb_t a, mp_limb_t exp, mp_limb_t n, double npre)
|
|
{
|
|
|
|
mp_limb_t x, y;
|
|
mp_limb_t e;
|
|
|
|
if (n == CNST_LIMB(1)) return 0L;
|
|
e = exp;
|
|
|
|
x = CNST_LIMB(1);
|
|
y = a;
|
|
|
|
while (e)
|
|
{
|
|
if (e & 1L) x = n_mulmod_precomp(x, y, n, npre);
|
|
e >>= 1;
|
|
if (e) y = n_mulmod_precomp(y, y, n, npre);
|
|
}
|
|
|
|
return x;
|
|
}
|
|
|
|
static
|
|
mp_limb_t n_powmod2_preinv(mp_limb_t a, mp_limb_t exp, mp_limb_t n, mp_limb_t ninv)
|
|
{
|
|
mp_limb_t x, y;
|
|
mp_limb_t e;
|
|
|
|
if (n == CNST_LIMB(1)) return CNST_LIMB(0);
|
|
e = exp;
|
|
|
|
x = CNST_LIMB(1);
|
|
y = a;
|
|
while (e)
|
|
{
|
|
if (e & 1) x = n_mulmod2_preinv(x, y, n, ninv);
|
|
e = e >> 1;
|
|
if (e) y = n_mulmod2_preinv(y, y, n, ninv);
|
|
}
|
|
|
|
return x;
|
|
}
|
|
|
|
static
|
|
mp_limb_t n_powmod(mp_limb_t a, mp_limb_signed_t exp, mp_limb_t n)
|
|
{
|
|
double npre = n_precompute_inverse(n);
|
|
|
|
return n_powmod_precomp(a, exp, n, npre);
|
|
}
|
|
|
|
static
|
|
mp_limb_t n_powmod2(mp_limb_t a, mp_limb_signed_t exp, mp_limb_t n)
|
|
{
|
|
mp_limb_t ninv = n_preinvert_limb(n);
|
|
|
|
return n_powmod2_preinv(a, exp, n, ninv);
|
|
}
|
|
|
|
static
|
|
mp_limb_t n_gcd(mp_limb_t x, mp_limb_t y)
|
|
{
|
|
mp_limb_t u3, v3;
|
|
mp_limb_t quot, rem;
|
|
|
|
u3 = x; v3 = y;
|
|
|
|
if ((mp_limb_signed_t) (x & y) < 0L) /* x and y both have top bit set */
|
|
{
|
|
quot=u3-v3;
|
|
u3 = v3;
|
|
v3 = quot;
|
|
}
|
|
|
|
while ((mp_limb_signed_t) (v3<<1) < 0L) /* second value has second msb set */
|
|
{
|
|
quot=u3-v3;
|
|
if (quot < v3)
|
|
{
|
|
u3 = v3;
|
|
v3 = quot;
|
|
} else if (quot < (v3<<1))
|
|
{
|
|
u3 = v3;
|
|
v3 = quot-u3;
|
|
} else
|
|
{
|
|
u3 = v3;
|
|
v3 = quot-(u3<<1);
|
|
}
|
|
}
|
|
|
|
while (v3) {
|
|
quot=u3-v3;
|
|
if (u3 < (v3<<2)) /* overflow not possible due to top 2 bits of v3 not being set */
|
|
{
|
|
if (quot < v3)
|
|
{
|
|
u3 = v3;
|
|
v3 = quot;
|
|
} else if (quot < (v3<<1))
|
|
{
|
|
u3 = v3;
|
|
v3 = quot-u3;
|
|
} else
|
|
{
|
|
u3 = v3;
|
|
v3 = quot-(u3<<1);
|
|
}
|
|
} else
|
|
{
|
|
quot=u3/v3;
|
|
rem = u3 - v3*quot;
|
|
u3 = v3;
|
|
v3 = rem;
|
|
}
|
|
}
|
|
|
|
return u3;
|
|
}
|
|
|
|
static
|
|
mp_limb_t n_invmod(mp_limb_t x, mp_limb_t y)
|
|
{
|
|
mp_limb_signed_t v1 = CNST_LIMB(0);
|
|
mp_limb_signed_t v2 = CNST_LIMB(1);
|
|
mp_limb_signed_t t2;
|
|
mp_limb_t u3, v3;
|
|
mp_limb_t quot, rem;
|
|
|
|
u3 = y, v3 = x;
|
|
|
|
if (v3 > u3)
|
|
{
|
|
rem = u3;
|
|
u3 = v3;
|
|
t2 = v2; v2 = v1; v1 = t2; v3 = rem;
|
|
}
|
|
|
|
if ((mp_limb_signed_t) (y & x) < 0L) /* y and x both have top bit set */
|
|
{
|
|
quot=u3-v3;
|
|
t2 = v2;
|
|
u3 = v3;
|
|
v2 = v1 - v2; v1 = t2; v3 = quot;
|
|
}
|
|
|
|
while ((mp_limb_signed_t) (v3<<1) < 0L) /* second value has second msb set */
|
|
{
|
|
quot=u3-v3;
|
|
if (quot < v3)
|
|
{
|
|
t2 = v2;
|
|
u3 = v3;
|
|
v2 = v1 - v2; v1 = t2; v3 = quot;
|
|
} else if (quot < (v3<<1))
|
|
{
|
|
u3 = v3;
|
|
t2 = v2; v2 = v1 - (v2<<1); v1 = t2; v3 = quot-u3;
|
|
} else
|
|
{
|
|
u3 = v3;
|
|
t2 = v2; v2 = v1 - 3*v2; v1 = t2; v3 = quot-(u3<<1);
|
|
}
|
|
}
|
|
|
|
while (v3) {
|
|
quot=u3-v3;
|
|
if (u3 < (v3<<2)) /* overflow not possible due to top 2 bits of v3 not being set */
|
|
{
|
|
if (quot < v3)
|
|
{
|
|
t2 = v2;
|
|
u3 = v3;
|
|
v2 = v1 - v2; v1 = t2; v3 = quot;
|
|
} else if (quot < (v3<<1))
|
|
{
|
|
u3 = v3;
|
|
t2 = v2; v2 = v1 - (v2<<1); v1 = t2; v3 = quot-u3;
|
|
} else
|
|
{
|
|
u3 = v3;
|
|
t2 = v2; v2 = v1 - 3*v2; v1 = t2; v3 = quot-(u3<<1);
|
|
}
|
|
} else
|
|
{
|
|
quot=u3/v3;
|
|
rem = u3 - v3*quot;
|
|
u3 = v3;
|
|
t2 = v2; v2 = v1 - quot*v2; v1 = t2; v3 = rem;
|
|
}
|
|
}
|
|
|
|
if (v1 < 0L) v1 += y;
|
|
|
|
return v1;
|
|
}
|
|
|
|
static
|
|
int n_jacobi(mp_limb_signed_t x, mp_limb_t y)
|
|
{
|
|
mp_limb_t a, b, temp;
|
|
int s, exp;
|
|
|
|
a = x;
|
|
b = y;
|
|
s = 1;
|
|
|
|
if (x < 0L)
|
|
{
|
|
if (((b - 1)/2)%2 == CNST_LIMB(1))
|
|
s = -s;
|
|
a = -x;
|
|
}
|
|
|
|
if ((a < b) && (b != CNST_LIMB(1)))
|
|
{
|
|
if (a == CNST_LIMB(0)) return 0;
|
|
|
|
temp = a;
|
|
a = b;
|
|
b = temp;
|
|
|
|
count_trailing_zeros(exp, b);
|
|
b>>=exp;
|
|
|
|
if (((exp*(a*a - 1))/8)%2 == CNST_LIMB(1)) /* only want values mod 8, */
|
|
s = -s; /* so overflows don't matter here */
|
|
|
|
if ((((a - 1)*(b - 1))/4)%2 == CNST_LIMB(1)) /* only want values mod 4, */
|
|
s = -s; /* so overflows don't matter here */
|
|
}
|
|
|
|
while (b != CNST_LIMB(1))
|
|
{
|
|
if ((a>>2) < b)
|
|
{
|
|
temp = a - b;
|
|
a = b;
|
|
if (temp < b)
|
|
b = temp;
|
|
else if (temp < (b<<1))
|
|
b = temp - a;
|
|
else
|
|
b = temp - (a<<1);
|
|
} else
|
|
{
|
|
temp = (a%b);
|
|
a = b;
|
|
b = temp;
|
|
}
|
|
|
|
if (b == CNST_LIMB(0)) return 0;
|
|
|
|
count_trailing_zeros(exp, b);
|
|
b>>=exp;
|
|
|
|
if (((exp*(a*a - 1))/8)%2 == CNST_LIMB(1)) /* only want values mod 8, */
|
|
s = -s; /* so overflows don't matter here */
|
|
|
|
if ((((a - 1)*(b - 1))/4)%2 == CNST_LIMB(1)) /* only want values mod 4, */
|
|
s = -s; /* so overflows don't matter here */
|
|
}
|
|
|
|
return s;
|
|
}
|
|
|
|
static
|
|
int n_is_pseudoprime_fermat(mp_limb_t n, mp_limb_t i)
|
|
{
|
|
if (BIT_COUNT(n) <= D_BITS) return (n_powmod(i, n - 1, n) == CNST_LIMB(1));
|
|
else
|
|
{
|
|
if ((mp_limb_signed_t) (n - 1) < 0L)
|
|
{
|
|
mp_limb_t temp = n_powmod2(i, (n - 1)/2, n);
|
|
return (n_powmod2(temp, 2, n) == CNST_LIMB(1));
|
|
} else
|
|
{
|
|
return (n_powmod2(i, n - 1, n) == CNST_LIMB(1));
|
|
}
|
|
}
|
|
}
|
|
|
|
static
|
|
int n_is_strong_pseudoprime_precomp(mp_limb_t n, double npre, mp_limb_t a, mp_limb_t d)
|
|
{
|
|
mp_limb_t t = d;
|
|
mp_limb_t y;
|
|
|
|
y = n_powmod_precomp(a, t, n, npre);
|
|
|
|
if (y == CNST_LIMB(1)) return 1;
|
|
t <<= 1;
|
|
|
|
while ((t != n - 1) && (y != n - 1))
|
|
{
|
|
y = n_mulmod_precomp(y, y, n, npre);
|
|
t <<= 1;
|
|
}
|
|
|
|
return (y == n - 1);
|
|
}
|
|
|
|
static
|
|
int n_is_strong_pseudoprime2_preinv(mp_limb_t n, mp_limb_t ninv, mp_limb_t a, mp_limb_t d)
|
|
{
|
|
mp_limb_t t = d;
|
|
mp_limb_t y;
|
|
|
|
y = n_powmod2_preinv(a, t, n, ninv);
|
|
|
|
if (y == CNST_LIMB(1)) return 1;
|
|
t <<= 1;
|
|
|
|
while ((t != n - 1) && (y != n - 1))
|
|
{
|
|
y = n_mulmod2_preinv(y, y, n, ninv);
|
|
t <<= 1;
|
|
}
|
|
|
|
return (y == n - 1);
|
|
}
|
|
|
|
static
|
|
n_pair_t fchain_precomp(mp_limb_t m, mp_limb_t n, double npre)
|
|
{
|
|
n_pair_t current, old;
|
|
int length;
|
|
mp_limb_t power, xy, xx, yy;
|
|
|
|
old.x = CNST_LIMB(2);
|
|
old.y = n - CNST_LIMB(3);
|
|
|
|
length = BIT_COUNT(m);
|
|
power = (CNST_LIMB(1)<<(length-1));
|
|
|
|
for ( ; length > 0; length--)
|
|
{
|
|
xy = n_mulmod_precomp(old.x, old.y, n, npre);
|
|
|
|
xy = n_addmod(xy, CNST_LIMB(3), n);
|
|
|
|
if (m & power)
|
|
{
|
|
current.y = n_submod(n_mulmod_precomp(old.y, old.y, n, npre), CNST_LIMB(2), n);
|
|
current.x = xy;
|
|
} else
|
|
{
|
|
current.x = n_submod(n_mulmod_precomp(old.x, old.x, n, npre), CNST_LIMB(2), n);
|
|
current.y = xy;
|
|
}
|
|
|
|
power >>= 1;
|
|
old = current;
|
|
}
|
|
|
|
return current;
|
|
}
|
|
|
|
static
|
|
n_pair_t fchain2_preinv(mp_limb_t m, mp_limb_t n, mp_limb_t ninv)
|
|
{
|
|
n_pair_t current, old;
|
|
int length;
|
|
mp_limb_t power, xy, xx, yy;
|
|
|
|
old.x = CNST_LIMB(2);
|
|
old.y = n - CNST_LIMB(3);
|
|
|
|
length = BIT_COUNT(m);
|
|
power = (CNST_LIMB(1)<<(length-1));
|
|
|
|
for ( ; length > 0; length--)
|
|
{
|
|
xy = n_mulmod2_preinv(old.x, old.y, n, ninv);
|
|
|
|
xy = n_addmod(xy, CNST_LIMB(3), n);
|
|
|
|
if (m & power)
|
|
{
|
|
current.y = n_submod(n_mulmod2_preinv(old.y, old.y, n, ninv), CNST_LIMB(2), n);
|
|
current.x = xy;
|
|
} else
|
|
{
|
|
current.x = n_submod(n_mulmod2_preinv(old.x, old.x, n, ninv), CNST_LIMB(2), n);
|
|
current.y = xy;
|
|
}
|
|
|
|
power >>= 1;
|
|
old = current;
|
|
}
|
|
|
|
return current;
|
|
}
|
|
|
|
static
|
|
int n_is_pseudoprime_fibonacci(mp_limb_t n)
|
|
{
|
|
mp_limb_t m, left, right;
|
|
n_pair_t V;
|
|
|
|
if (ABS((mp_limb_signed_t) n) <= CNST_LIMB(3))
|
|
{
|
|
if (n >= CNST_LIMB(2)) return 1;
|
|
return 0;
|
|
}
|
|
|
|
m = (n - n_jacobi(CNST_LIMB(5), n))/2; /* no overflow as (5/n) = 0 for n = 2^64 - 1 */
|
|
|
|
if (BIT_COUNT(n) <= D_BITS)
|
|
{
|
|
double npre = n_precompute_inverse(n);
|
|
|
|
V = fchain_precomp(m, n, npre);
|
|
return (n_mulmod_precomp(n - CNST_LIMB(3), V.x, n, npre)
|
|
== n_mulmod_precomp(CNST_LIMB(2), V.y, n, npre));
|
|
} else
|
|
{
|
|
mp_limb_t ninv = n_preinvert_limb(n);
|
|
|
|
V = fchain2_preinv(m, n, ninv);
|
|
return (n_mulmod2_preinv(n - CNST_LIMB(3), V.x, n, ninv)
|
|
== n_mulmod2_preinv(CNST_LIMB(2), V.y, n, ninv));
|
|
}
|
|
}
|
|
|
|
static
|
|
n_pair_t lchain_precomp(mp_limb_t m, mp_limb_t a, mp_limb_t n, double npre)
|
|
{
|
|
n_pair_t current, old;
|
|
int length, i;
|
|
mp_limb_t power, xy, xx, yy;
|
|
|
|
old.x = CNST_LIMB(2);
|
|
old.y = a;
|
|
|
|
length = BIT_COUNT(m);
|
|
power = (CNST_LIMB(1)<<(length - 1));
|
|
|
|
for (i = 0; i < length; i++)
|
|
{
|
|
xy = n_submod(n_mulmod_precomp(old.x, old.y, n, npre), a, n);
|
|
|
|
if (m & power)
|
|
{
|
|
yy = n_submod(n_mulmod_precomp(old.y, old.y, n, npre), CNST_LIMB(2), n);
|
|
current.x = xy;
|
|
current.y = yy;
|
|
} else
|
|
{
|
|
xx = n_submod(n_mulmod_precomp(old.x, old.x, n, npre), CNST_LIMB(2), n);
|
|
current.x = xx;
|
|
current.y = xy;
|
|
}
|
|
|
|
power >>= 1;
|
|
old = current;
|
|
}
|
|
|
|
return current;
|
|
}
|
|
|
|
static
|
|
n_pair_t lchain2_preinv(mp_limb_t m, mp_limb_t a, mp_limb_t n, mp_limb_t ninv)
|
|
{
|
|
n_pair_t current, old;
|
|
int length, i;
|
|
mp_limb_t power, xy, xx, yy;
|
|
|
|
old.x = CNST_LIMB(2);
|
|
old.y = a;
|
|
|
|
length = BIT_COUNT(m);
|
|
power = (CNST_LIMB(1)<<(length - 1));
|
|
|
|
for (i = 0; i < length; i++)
|
|
{
|
|
xy = n_submod(n_mulmod2_preinv(old.x, old.y, n, ninv), a, n);
|
|
|
|
if (m & power)
|
|
{
|
|
yy = n_submod(n_mulmod2_preinv(old.y, old.y, n, ninv), CNST_LIMB(2), n);
|
|
current.x = xy;
|
|
current.y = yy;
|
|
} else
|
|
{
|
|
xx = n_submod(n_mulmod2_preinv(old.x, old.x, n, ninv), CNST_LIMB(2), n);
|
|
current.x = xx;
|
|
current.y = xy;
|
|
}
|
|
|
|
power >>= 1;
|
|
old = current;
|
|
}
|
|
|
|
return current;
|
|
}
|
|
|
|
static
|
|
int n_is_pseudoprime_lucas(mp_limb_t n)
|
|
{
|
|
int i, D, Q;
|
|
mp_limb_t A;
|
|
mp_limb_t left, right;
|
|
n_pair_t V;
|
|
|
|
D = 0;
|
|
Q = 0;
|
|
|
|
if (((n % 2) == 0) || (ABS((mp_limb_signed_t) n) <= 2))
|
|
{
|
|
if (n == CNST_LIMB(2)) return 1;
|
|
else return 0;
|
|
}
|
|
|
|
for (i = 0; i < 100; i++)
|
|
{
|
|
D = 5 + 2*i;
|
|
if (n_gcd(D, n%D) != CNST_LIMB(1))
|
|
{
|
|
if (n == D) continue;
|
|
else return 0;
|
|
}
|
|
if (i % 2 == 1) D = -D;
|
|
if (n_jacobi(D, n) == -1) break;
|
|
}
|
|
|
|
if (i == 100)
|
|
{
|
|
if (n_is_square(n)) return -1;
|
|
else return 1;
|
|
}
|
|
|
|
Q = (1 - D)/4;
|
|
if (Q < 0)
|
|
{
|
|
if (n < CNST_LIMB(52))
|
|
{
|
|
while (Q < 0) Q += n;
|
|
A = n_submod(n_invmod(Q, n), CNST_LIMB(2), n);
|
|
} else
|
|
A = n_submod(n_invmod(Q + n, n), CNST_LIMB(2), n);
|
|
} else
|
|
{
|
|
if (n < CNST_LIMB(52))
|
|
{
|
|
while (Q >= n) Q -= n;
|
|
A = n_submod(n_invmod(Q, n), CNST_LIMB(2), n);
|
|
} else
|
|
A = n_submod(n_invmod(Q, n), CNST_LIMB(2), n);
|
|
}
|
|
|
|
if (BIT_COUNT(n) <= D_BITS)
|
|
{
|
|
double npre = n_precompute_inverse(n);
|
|
V = lchain_precomp(n + 1, A, n, npre);
|
|
|
|
left = n_mulmod_precomp(A, V.x, n, npre);
|
|
right = n_mulmod_precomp(2, V.y, n, npre);
|
|
} else
|
|
{
|
|
mp_limb_t ninv = n_preinvert_limb(n);
|
|
V = lchain2_preinv(n + 1, A, n, ninv);
|
|
|
|
left = n_mulmod_precomp(A, V.x, n, ninv);
|
|
right = n_mulmod_precomp(2, V.y, n, ninv);
|
|
}
|
|
|
|
return (left == right);
|
|
}
|
|
|
|
int mpir_is_likely_prime_BPSW(mp_limb_t n)
|
|
{
|
|
if (n <= CNST_LIMB(1)) return 0;
|
|
|
|
if ((n & CNST_LIMB(1)) == CNST_LIMB(0))
|
|
{
|
|
if (n == CNST_LIMB(2)) return 1;
|
|
return 0;
|
|
}
|
|
|
|
if (((n % CNST_LIMB(10)) == CNST_LIMB(3)) || ((n % CNST_LIMB(10)) == CNST_LIMB(7)))
|
|
{
|
|
if (n_is_pseudoprime_fermat(n, 2) == 0) return 0;
|
|
|
|
return n_is_pseudoprime_fibonacci(n);
|
|
} else
|
|
{
|
|
mp_limb_t d;
|
|
|
|
d = n - CNST_LIMB(1);
|
|
while ((d & CNST_LIMB(1)) == CNST_LIMB(0)) d >>= 1;
|
|
|
|
if (BIT_COUNT(n) <= D_BITS)
|
|
{
|
|
double npre = n_precompute_inverse(n);
|
|
if (n_is_strong_pseudoprime_precomp(n, npre, 2L, d) == 0) return 0;
|
|
} else
|
|
{
|
|
mp_limb_t ninv = n_preinvert_limb(n);
|
|
if (n_is_strong_pseudoprime2_preinv(n, ninv, 2L, d) == 0) return 0;
|
|
}
|
|
|
|
return (n_is_pseudoprime_lucas(n) == 1);
|
|
}
|
|
}
|
|
|
|
#endif /* GMP_LIMB_BITS */
|
|
|
|
/*
|
|
Could have another parameter to specify what "likely" means
|
|
i.e. for factoring, for RSA or to state that we have already done
|
|
trial div
|
|
*/
|
|
|
|
/*
|
|
could call it mpz_likely_composite_p then when true return more info,
|
|
i.e. a factor
|
|
*/
|
|
int
|
|
mpz_likely_prime_p (mpz_srcptr N, gmp_randstate_t STATE, mpir_ui td)
|
|
{
|
|
int d, t, r;
|
|
unsigned long tdlim, i;
|
|
mpz_t base, nm1, x, e, n;
|
|
|
|
ALLOC (n) = ALLOC(N);
|
|
SIZ (n) = ABSIZ(N);
|
|
PTR (n) = PTR(N); /* fake up an absolute value that we don't have de-allocate */
|
|
|
|
/* algorithm does not handle small values, get rid of them here */
|
|
if (mpz_cmp_ui (n, 2) == 0 || mpz_cmp_ui (n, 3) == 0)
|
|
return 1;
|
|
|
|
if (mpz_cmp_ui (n, 5) < 0 || mpz_even_p (n))
|
|
return 0;
|
|
|
|
#if GMP_LIMB_BITS == 64 || GMP_LIMB_BITS == 32
|
|
if (SIZ(n) == 1)
|
|
{
|
|
return mpir_is_likely_prime_BPSW(PTR(n)[0]);
|
|
}
|
|
#endif
|
|
|
|
/*
|
|
For factoring purposes we assume we know nothing about N i.e. it is
|
|
a random integer. Therefore it has a good chance of factoring by small
|
|
divisiors. So try trial division as its fast and it checks small
|
|
divisors. Checking for other divisors is not worth it even if the test
|
|
is fast as we have random integer so only small divisors are common
|
|
enough. Remember this is not exact so it doesn't matter if we miss a
|
|
few divisors
|
|
*/
|
|
tdlim = mpz_sizeinbase(n, 2);
|
|
tdlim = MAX(1000, tdlim);
|
|
|
|
d = mpz_trial_division(n, 3, tdlim);
|
|
|
|
if (d != 0)
|
|
{
|
|
if (mpz_cmp_ui(n, d) == 0)
|
|
return 1;
|
|
|
|
return 0;
|
|
}
|
|
|
|
if (mpz_cmp_ui (n, tdlim * tdlim) < 0)
|
|
return 1; /* if tdlim*tdlim overflows, n is not a single limb so can't be true */
|
|
|
|
ASSERT (mpz_odd_p(n));
|
|
ASSERT (mpz_cmp_ui (n, 5) >= 0); /* so we can choose a base */
|
|
|
|
/* now do strong pseudoprime test */
|
|
/* get random base, for now choose any size, later choose a small one */
|
|
mpz_init (base);
|
|
mpz_init_set (nm1, n);
|
|
mpz_sub_ui (nm1, nm1, 1);
|
|
|
|
mpz_init (e);
|
|
mpz_init (x);
|
|
|
|
r = 1;
|
|
|
|
for (i = 0; i < 10; i++) /* try LP_ITERS random bases */
|
|
{
|
|
do
|
|
{
|
|
mpz_urandomm(base, STATE, nm1);
|
|
} while (mpz_cmp_ui(base, 1) <= 0);
|
|
|
|
/*
|
|
Base is 2 to n - 2 which implies n >= 4. Only really want a
|
|
small base, and ignore the rare base = n - 1 condition etc.
|
|
*/
|
|
t = mpz_scan1(nm1, 0); /* 2^t divides nm1 */
|
|
|
|
ASSERT(t > 0);
|
|
|
|
mpz_tdiv_q_2exp(e, nm1, t); /* e = nm1/2^t */
|
|
mpz_powm(x, base, e, n); /* x = base^e mod n */
|
|
|
|
if (mpz_cmp_ui(x, 1) == 0)
|
|
continue;
|
|
|
|
if (mpz_cmp(x, nm1) == 0)
|
|
continue;
|
|
|
|
for (r = 0, t = t - 1; t > 0; t--)
|
|
{
|
|
mpz_mul(x, x, x);
|
|
mpz_mod(x, x, n);
|
|
|
|
if (mpz_cmp(x, nm1) == 0)
|
|
{
|
|
r = 1;
|
|
break;
|
|
}
|
|
|
|
if (mpz_cmp_ui (x, 1) == 0)
|
|
break;
|
|
}
|
|
|
|
if (r == 1)
|
|
continue;
|
|
|
|
break;
|
|
}
|
|
|
|
mpz_clear (e);
|
|
mpz_clear (base);
|
|
mpz_clear (nm1);
|
|
mpz_clear (x);
|
|
|
|
return r;
|
|
}
|