POJ - 2478 Eratosthenes筛法/线性筛求欧拉函数

题目链接：POJ-2478

The Farey Sequence Fn for any integer n with n >= 2 is the set of irreducible rational numbers a/b with 0 < a < b <= n and gcd(a,b) = 1 arranged in increasing order. The first few are 
F2 = {1/2} 
F3 = {1/3, 1/2, 2/3} 
F4 = {1/4, 1/3, 1/2, 2/3, 3/4} 
F5 = {1/5, 1/4, 1/3, 2/5, 1/2, 3/5, 2/3, 3/4, 4/5} 

You task is to calculate the number of terms in the Farey sequence Fn.

Input

There are several test cases. Each test case has only one line, which contains a positive integer n (2 <= n <= 10 6). There are no blank lines between cases. A line with a single 0 terminates the input.

Output

For each test case, you should output one line, which contains N(n) ---- the number of terms in the Farey sequence Fn. 

Sample Input

2
3
4
5
0

Sample Output

1
3
5
9

题意就是给定n，求出所有分母小于等于n的真分数个数。

思路很清晰，即求出小于等于n的数 的欧拉函数之和。

直接将n以内所有数字质因数分解后求欧拉函数是一定会超时的，所以我们要用更快速的方法求欧拉函数表。

法一：Eratosthenes筛法 复杂度 O(NlogN)，用时250ms

Eratosthenes的筛法基于 对任意x的整数倍数，2x，3x，… 都不是质数。从2开始，从小到大扫描每个数X，把它的倍数2x，3x，…,floor(N/x)*x标记为合数。当扫描到一个数时，如果它尚未标记，则它一定不能被2~x-1之间的任何数整除，该数就是质数。

又因为实际上小于等于x^2的x的倍数在扫描更小的数时就已经被标记过了，因此，我们可以对Eratosthens筛法进行优化。对于每个数x，我们只需要从x^2开始，把x^2，(x+1)*x，…，floor（N/x）*x标记为合数即可。

#include <cstdio>
#include <algorithm>
#include <cstring>
#include <cmath>
#include <iostream>
#include <map>
#include <vector>
#include <queue>
#include <set>
#define eps 1e-8
typedef long long ll;
const double PI = acos(-1.0);
const int maxn = 1e6;
const int INF = 1e9;
const ll linf = 0x3f3f3f3f3f3f3f3f;
using namespace std;
int n;
ll phi[maxn+5];
void euler(int n)
{
    for(int i = 2; i<=n; i++)
        phi[i] = i;
    for(int i = 2; i<=n; i++)
    {
        if(phi[i] == i)
        {
            for(int j = i*i; j<=n; j += i)
                phi[j] = phi[j]/i*(i-1);
        }
    }
}
int main()
{
    euler(maxn);
    for(int i = 3; i<=1e6; i++)
    {
        //cout<<phi[i]<<endl;
        phi[i] += phi[i-1];
    }

    while(scanf("%d",&n) != EOF && n)
    {
        cout<<phi[n]<<endl;
    }

    return 0;
}

 法二：线性筛 复杂度O（N），用时63ms

1.若p|n且p^2|n，则phi(n) = phi(n/p)*p

2.若p|n但p^2不能整除n，则phi(n) = phi(n/p)*(p-1)

在线性筛法中，每个合数n只会被它的最小质因子p筛一次。我们恰好可以在此时执行以上两条判断，从phi(n/p)递推到phi(n)

#include <cstdio>
#include <algorithm>
#include <cstring>
#include <cmath>
#include <iostream>
#include <map>
#include <vector>
#include <queue>
#include <set>
#define eps 1e-8
typedef long long ll;
const double PI = acos(-1.0);
const int maxn = 1e6;
const int INF = 1e9;
const ll linf = 0x3f3f3f3f3f3f3f3f;
using namespace std;
int n,cnt;
int v[maxn+5],prime[maxn+5];
ll phi[maxn+5];
void euler(int n)
{
    memset(v,0,sizeof v);
    cnt = 0;
    for(int i = 2; i<=n; i++)
    {
        //i是质数
        if(v[i] == 0)
        {
            v[i] = i;
            prime[cnt++] = i;
            phi[i] = i-1;
        }
        //给当前的数i乘上一个质因子
        for(int j = 0; j<cnt; j++)
        {
            //i有比prime[j]更小的质因子或超出n的范围
            if(prime[j]>v[i] || prime[j]>n/i)
                break;
            //prime[j]是i*prime[j]的最小质因子
            v[i*prime[j]] = prime[j];
            phi[i*prime[j]] = phi[i]*(i%prime[j] ? prime[j]-1:prime[j]);
        }
    }



}
int main()
{
    euler(maxn);
    for(int i = 3; i<=1e6; i++)
    {
        //cout<<phi[i]<<endl;
        phi[i] += phi[i-1];
    }

    while(scanf("%d",&n) != EOF && n)
    {
        cout<<phi[n]<<endl;
    }

    return 0;
}