#include "tickets.h"
#include <vector>
#include <vector>
long long find_maximum(int k, std::vector<std::vector<int>> d);
void allocate_tickets( std::vector<std::vector<int>> _x);
#include <bits/stdc++.h>
using namespace std;
//#pragma GCC optimize ("O3")
//#pragma GCC target ("sse4")
#define endl "\n"
typedef long long ll;
typedef long double ld;
typedef unsigned long long ull;
template<class T, class T2> inline bool chkmax(T &x, const T2 &y) { return x < y ? x = y, 1 : 0; }
template<class T, class T2> inline bool chkmin(T &x, const T2 &y) { return x > y ? x = y, 1 : 0; }
const ll mod = 1e9 + 7;
template<class T> inline void fix(T &x) {if(labs(x) >= mod) {x %= mod;} if(x < 0) {x += mod;}}
#define out(x) cout << __LINE__ << ": " << (#x) << " = " << (x) << endl
long long find_maximum(int k, std::vector<std::vector<int>> x) {
int n = x.size();
int m = x[0].size();
std::vector<std::vector<int> > answer;
vector<vector<pair<int, int> > > pr;
priority_queue<pair<int, int> > pq;
vector<int> got;
got.resize(n);
int ans = 0;
for (int i = 0; i < n; i++) {
vector<int> row(m);
for (int j = 0; j < m; j++) {
row[j] = -1;
}
ans += x[i][m - 1];
pq.push({-x[i][0] - x[i][m - k], i});
answer.push_back(row);
}
int iter = n * k / 2;
for(int i = 0; i < iter; i ++) {
auto curr = pq.top(); pq.pop();
ans += curr.first;
got[curr.second] ++;
if(got[curr.second] < k) {
pq.push({-x[curr.second][got[curr.second]] - x[curr.second][m - k + got[curr.second]], curr.second});
}
}
int cnt = 0;
for(int i = 0; i < n; i ++) {
for(int j = 0; j < got[i]; j ++) {
answer[i][j] = (cnt + j) % k;
}
cnt += got[i];
int group = cnt;
for(int j = m - k + got[i]; j < m; j ++) {
answer[i][j] = group % k;
group ++;
}
}
allocate_tickets(answer);
return ans;
}
