/* Native 63-bit Pollard-Rho-Brent for x86-64. */
#include <gmp.h>
#include "ptypes.h"
#include "pbrent63.h"
#define FUNC_gcd_ui 1
#include "utility.h"
#if BITS_PER_WORD == 64 && HAVE_STD_U64 && defined(__GNUC__) && defined(__x86_64__)
static INLINE UV mpz_getuv(mpz_t n) {
UV v = mpz_getlimbn(n,0);
if (GMP_LIMB_BITS < 64 || sizeof(mp_limb_t) < sizeof(UV))
v |= ((UV)mpz_getlimbn(n,1)) << 32;
return v;
}
int pbrent63(mpz_t n, mpz_t f, UV rounds) {
UV facs[2];
int nfactors;
if (mpz_sizeinbase(n,2) > 63) return 0;
nfactors = uvpbrent63(mpz_getuv(n), facs, rounds, 1);
if (nfactors < 2) return 0;
/* Smallest factor of 64-bit n always fits in 32-bit */
mpz_set_ui(f, (facs[0] < facs[1]) ? facs[0] : facs[1]);
return 1;
}
/* Trimmed out all the extra stuff and the 64-bit variation */
#define mont_get1(n) _u64div(1,n)
/* Must have npi = mont_inverse(n), mont1 = mont_get1(n) */
#define mont_geta(a,n) mulmod(a,mont1,n)
#define mont_mulmod(a,b,n) _mulredc63(a,b,n,npi)
static INLINE uint64_t mont_inverse(const uint64_t n) {
uint64_t ret = (3*n) ^ 2;
ret *= (uint64_t)2 - n * ret;
ret *= (uint64_t)2 - n * ret;
ret *= (uint64_t)2 - n * ret;
ret *= (uint64_t)2 - n * ret;
return (uint64_t)0 - ret;
}
/* MULREDC asm from Ben Buhrow */
static INLINE uint64_t _mulredc63(uint64_t a, uint64_t b, uint64_t n, uint64_t npi) {
asm("mulq %2 \n\t"
"movq %%rax, %%r10 \n\t"
"movq %%rdx, %%r11 \n\t"
"mulq %3 \n\t"
"mulq %4 \n\t"
"addq %%r10, %%rax \n\t"
"adcq %%r11, %%rdx \n\t"
"xorq %%rax, %%rax \n\t"
"subq %4, %%rdx \n\t"
"cmovc %4, %%rax \n\t"
"addq %%rdx, %%rax \n\t"
: "=a"(a)
: "0"(a), "r"(b), "r"(npi), "r"(n)
: "rdx", "r10", "r11", "cc");
return a;
}
static INLINE uint64_t _u64div(uint64_t c, uint64_t n) {
asm("divq %4"
: "=a"(c), "=d"(n)
: "1"(c), "0"(0), "r"(n));
return n;
}
static INLINE UV mulmod(UV a, UV b, UV n) {
UV d, dummy; /* d will get a*b mod c */
asm ("mulq %3\n\t" /* mul a*b -> rdx:rax */
"divq %4\n\t" /* (a*b)/c -> quot in rax remainder in rdx */
:"=a"(dummy), "=&d"(d) /* output */
:"a"(a), "r"(b), "r"(n) /* input */
:"cc" /* mulq and divq can set conditions */
);
return d;
}
static INLINE UV addmod(UV a, UV b, UV n) {
UV t = a-n;
a += b;
asm ("add %2, %1\n\t" /* t := t + b */
"cmovc %1, %0\n\t" /* if (carry) a := t */
:"+r" (a), "+&r" (t)
:"r" (b)
:"cc"
);
return a;
}
#define ABSDIFF(x,y) (x>y) ? x-y : y-x
/* Brent's modifications to Pollard's Rho. */
int uvpbrent63(UV n, UV *factors, UV rounds, UV a)
{
UV const nbits = BITS_PER_WORD - __builtin_ctzll(n);
const UV inner = (nbits <= 31) ? 32 : (nbits <= 35) ? 64 : (nbits <= 40) ? 160 : (nbits <= 52) ? 256 : 320;
UV f, m, r, rleft, Xi, Xm, Xs;
int irounds, fails = 6;
const uint64_t npi = mont_inverse(n), mont1 = mont_get1(n);
if (n <= 3) { factors[0] = n; return 1; }
if (!(n&1)) { factors[0] = 2; factors[1] = n/2; return 2; }
r = f = 1;
Xi = Xm = Xs = mont1;
a = mont_geta(a,n);
while (rounds > 0) {
rleft = (r > rounds) ? rounds : r;
Xm = Xi;
/* Do rleft rounds, inner at a time */
while (rleft > 0) {
irounds = (rleft > (UV)inner) ? inner : rleft;
rleft -= irounds;
rounds -= irounds;
Xs = Xi;
Xi = mont_mulmod(Xi,Xi+a,n);
m = ABSDIFF(Xi,Xm);
while (--irounds > 0) {
Xi = mont_mulmod(Xi,Xi+a,n);
f = ABSDIFF(Xi,Xm);
m = mont_mulmod(m, f, n);
}
f = gcd_ui(m, n);
if (f != 1)
break;
}
/* If f == 1, then we didn't find a factor. Move on. */
if (f == 1) {
r *= 2;
continue;
}
if (f == n) { /* back up, with safety */
Xi = Xs;
do {
Xi = mont_mulmod(Xi,Xi+a,n);
m = ABSDIFF(Xi,Xm);
f = gcd_ui(m, n);
} while (f == 1 && r-- != 0);
}
if (f == 0 || f == n) {
if (fails-- <= 0) break;
Xi = Xm = mont1;
a = addmod(a, mont_geta(11,n), n);
continue;
}
factors[0] = f;
factors[1] = n / f;
MPUassert( factors[0] * factors[1] == n , "incorrect factoring");
return 2;
}
factors[0] = n;
return 1;
}
#else /* no 64-bit gcc x86-64 */
int pbrent63(mpz_t n, mpz_t f, UV rounds) { return 0; }
int uvpbrent63(UV n, UV *factors, UV rounds, UV a) { factors[0] = n; return 1; }
#endif