#include <iostream>
#include <vector>
#include <atcoder/lazysegtree>

using namespace std;
using namespace atcoder;
typedef long long ll;
ll inf = 1000000000000000000;
struct S{ll x,sz;};
struct F{ll x;};
S op(S l,S r){return S{l.x + r.x,l.sz + r.sz};}
S e(){return S{0,0};}
S mapping(F l,S r){return S{r.x + l.x*r.sz,r.sz};}
F compo(F l,F r){return F{l.x + r.x};}
F id(){return F{0};}
ll a[400010],b[400010];
vector<ll> solve(vector<int> &v,int k){
    int i,m = v.size();
    vector<int> pos(m);
    for(i=0;i<m;i++) pos[v[i]] = i;
    vector<S> ve(m);
    for(i=0;i<m;i++) ve[i] = {0,1};
    lazy_segtree<S,op,e,F,mapping,compo,id> seg(ve),seg2(ve);
    vector<ll> ret(m);
    for(i=0;i<m;i++){
        ll inv = seg.prod(v[i] + 1,m).x;
        seg2.apply(i,i + 1,F{inv});
        seg.apply(v[i],v[i] + 1,F{1});
    }
    ret[0] = seg2.all_prod().x;
    for(i=1;i<m;i++){
        ll inv = seg2.get(pos[i - 1]).x;
        seg2.apply(pos[i - 1],pos[i - 1] + 1,F{-inv});
        seg2.apply(pos[i - 1] + 1,m,F{1});
        ret[i] = seg2.all_prod().x;
    }
    return ret;
}

int main(){
    int i,j,m,k; cin >> m >> k;
    for(i=0;i<m*k;i++) cin >> a[i];
    vector<vector<int>> ind(m);
    for(i=0;i<m*k;i++) ind[a[i]].push_back(i);
    for(i=0;i<m;i++){
        for(j=0;j<k;j++) b[ind[i][j]] = j;
    }
    /*for(i=0;i<m*k;i++) cout << b[i] << " ";
    cout << endl;*/
    ll ans = 0;
    vector<S> ve(k);
    for(i=0;i<k;i++) ve[i] = {0,1};
    lazy_segtree<S,op,e,F,mapping,compo,id> seg(ve);
    for(i=0;i<m*k;i++){
        ans += seg.prod(b[i] + 1,k).x;
        seg.apply(b[i],b[i] + 1,F{1});
    }
    vector<vector<int>> v(k);
    for(i=0;i<m*k;i++) v[b[i]].push_back(a[i]);
    vector<ll> sum(m);
    for(i=0;i<k;i++){
        vector<ll> nv = solve(v[i],k);
        /*for(j=0;j<m;j++) cout << nv[j] << " ";
        cout << endl;*/
        for(j=0;j<m;j++) sum[j] += nv[j];
    }
    ll mx = inf;
    for(i=0;i<m;i++) mx = min(mx,sum[i]);
    ans += mx;
    cout << ans << endl;
    return 0;
}