#![allow(unused_imports)]
#![allow(non_snake_case)]
use std::cmp::*;
use std::collections::*;
use std::io::Write;

#[allow(unused_macros)]
macro_rules! debug {
    ($($e:expr),*) => {
        #[cfg(debug_assertions)]
        $({
            let (e, mut err) = (stringify!($e), std::io::stderr());
            writeln!(err, "{} = {:?}", e, $e).unwrap()
        })*
    };
}

fn main() {
    let v = read_vec::<i64>();
    let (n, m) = (v[0], v[1] as usize);
    let mut edges = vec![];
    for i in 0..m {
        let v = read_vec::<i64>();
        let (b, c) = (v[0], v[1]);
        edges.push((b, c));
    }

    let mut used = edges.iter().map(|x| x.0).collect::<Vec<_>>();
    used.append(&mut edges.iter().map(|x| x.1).collect::<Vec<_>>());
    used.sort();
    used.dedup();

    let mut compress_dict = BTreeMap::new();
    for i in 0..used.len() {
        compress_dict.insert(used[i], i);
    }

    let mut g = vec![vec![]; compress_dict.len()];
    for &(b, c) in &edges {
        let from_idx = compress_dict[&b];
        let to_idx = compress_dict[&c];
        g[from_idx].push(to_idx);
    }

    let cmp = scc(&g);

    let mut indexes = (0..compress_dict.len()).collect::<Vec<_>>();
    indexes.sort_by_key(|&x| cmp[x]);
    indexes.reverse();

    let cmp_max = *cmp.iter().max().unwrap();
    let mut accum_max = vec![0; cmp_max + 1];

    for &idx in &indexes {
        accum_max[cmp[idx]] = max(used[idx], accum_max[cmp[idx]]);
    }
    for idx in indexes {
        for &to in &g[idx] {
            accum_max[cmp[idx]] = max(accum_max[cmp[to]], accum_max[cmp[idx]]);
        }
    }

    let mut ans = n * (n + 1) / 2;
    let mut used_b = HashSet::new();
    for (b, _) in edges {
        if used_b.contains(&b) {
            continue;
        }
        ans -= b;
        ans += accum_max[cmp[compress_dict[&b]]];
        used_b.insert(b);
    }
    println!("{}", ans);
}

fn scc(g: &[Vec<usize>]) -> Vec<usize> {
    fn dfs(v: usize, g: &[Vec<usize>], vs: &mut Vec<usize>, used: &mut [bool]) {
        used[v] = true;
        for &to in &g[v] {
            if !used[to] {
                dfs(to, g, vs, used);
            }
        }
        vs.push(v);
    }
    fn rdfs(v: usize, k: usize, g_rev: &[Vec<usize>], cmp: &mut [usize], used: &mut [bool]) {
        used[v] = true;
        cmp[v] = k;

        for &to in &g_rev[v] {
            if !used[to] {
                rdfs(to, k, g_rev, cmp, used);
            }
        }
    }

    let g_rev = {
        let mut g_rev = vec![Vec::new(); g.len()];
        for (i, es) in g.iter().enumerate() {
            for &to in es {
                g_rev[to].push(i);
            }
        }
        g_rev
    };

    let mut vs = Vec::new();
    let mut used = vec![false; g.len()];
    for v in 0..g.len() {
        if !used[v] {
            dfs(v, g, &mut vs, &mut used);
        }
    }
    let mut used = vec![false; g.len()];
    let mut cmp = vec![0; g.len()];
    let mut k = 0;
    for &v in vs.iter().rev() {
        if !used[v] {
            rdfs(v, k, &g_rev, &mut cmp, &mut used);
            k += 1;
        }
    }

    cmp
}

fn read<T: std::str::FromStr>() -> T {
    let mut s = String::new();
    std::io::stdin().read_line(&mut s).ok();
    s.trim().parse().ok().unwrap()
}

fn read_vec<T: std::str::FromStr>() -> Vec<T> {
    read::<String>()
        .split_whitespace()
        .map(|e| e.parse().ok().unwrap())
        .collect()
}