Pollard_rho 因数分解

Int64以内Rabin-Miller强伪素数测试和Pollard 因数分解的算法实现

选取随机数(ａ) 随机数(ｂ)，检查(gcd(a - b, n))是否大于１，若大于１则(a - b)是(ｎ)的一个因数

实现１：floyd判环

利用多项式(f(x))迭代出({x_0, x_1, dots, x_k})

设定(x = y = x_0)的初始值，选用多项式进行迭代，每次：(x = f(x)), (y = f(f(y)))，即：(x = x_k, y = x_{2k})当(x == y)时出现循环

设(x = y = 2)，(f(n) = n^2 + a)

typedef long long ll;

ll mul_mod(ll a, ll b, ll m){
    ll ans = 0, exp = a;
    while(a >= m) a -= m;
    while(b){
        if(b & 1){
            ans += exp;
            while(ans >= m) ans -= m;
        }
        exp += exp;
        while(exp >= m) exp -= m;
        b >>= 1;
    }
    return ans;
}

ll pollard_rho(ll n, int a){
    ll x = 2, y = 2, d = 1;
    while(d == 1){
        x = mul_mod(x, x, n) + a;
        y = mul_mod(y, y, n) + a;
        y = mul_mod(y, y, n) + a;
        d = __gcd((x >= y ? x - y : y - x), n);
    }
    if(d == n) return pollard_rho(n, a + 1);
    return d;
}

实现２: brent判环（更高效）

不同于floyd每次计算(x_k, x_{2k})进行判断,brent每次只计算(x_k),当ｋ是2的方幂时,(y = x_k),每次计算(d = gcd(x_k - y, n))

typedef long long ll;

ll mul_mod(ll a, ll b, ll m){
    ll ans = 0, exp = a;
    while(a >= m) a -= m;
    while(b){
        if(b & 1){
            ans += exp;
            while(ans >= m) ans -= m;
        }
        exp += exp;
        while(exp >= m) exp -= m;
        b >>= 1;
    }
    return ans;
}

ll pollard_rho(ll n, int a){
    ll x = 2, y = 2, d = 1, k = 0, i = 1;
    while(d == 1){
        ++k;
        x = mul_mod(x, x, n) + a;
        d = __gcd(x >= y ? x - y : y - x, n);
        if(k == i){
            y = x;
            i <<= 1;
        }
    }
    if(d == n) return pollard_rho(n, a + 1);
    return d;
}