/*
#pragma GCC optimize("Ofast,unroll-loops")
#pragma GCC target("avx2,fma,bmi,bmi2,sse4.2,popcnt,lzcnt")
*/
#include <bits/stdc++.h>
#define taskname ""
#define all(x) x.begin(), x.end()
#define rall(x) x.rbegin(), x.rend()
#define i64 long long
#define isz(x) (int)x.size()
using namespace std;
void solve() {
int n, m;
cin >> n >> m;
vector<vector<int>> a(n, vector<int>(m));
vector<vector<int>> b(n, vector<int>(m));
for (int i = 0; i < n; ++i) for (int j = 0; j < m; ++j) {
cin >> a[i][j];
}
for (int i = 0; i < n; ++i) for (int j = 0; j < m; ++j) {
cin >> b[i][j];
}
int res = 0;
vector<int> process;
vector<int> ptr(m), val(m), cnt(n);
vector<vector<int>> idx(m, vector<int>(n));
auto update = [&]() -> void {
for (int j = 0; j < m; ++j) {
while (ptr[j] < n and a[idx[j][ptr[j]]][j] <= val[j]) {
if (++cnt[idx[j][ptr[j]]] == m) {
process.emplace_back(idx[j][ptr[j]]);
}
++ptr[j];
}
}
};
for (int j = 0; j < m; ++j) {
iota(all(idx[j]), 0);
sort(all(idx[j]), [&](int x, int y) {
return a[x][j] < a[y][j];
});
}
do {
while (not process.empty()) {
int idx = process.back();
process.pop_back();
// cout << "+ " << idx << endl;
for (int j = 0; j < m; ++j) val[j] += b[idx][j];
++res;
}
update();
// cout << endl;
} while (not process.empty());
cout << res << endl;
}
signed main() {
#ifndef CDuongg
if (fopen(taskname".inp", "r"))
assert(freopen(taskname".inp", "r", stdin)), assert(freopen(taskname".out", "w", stdout));
#else
freopen("bai3.inp", "r", stdin);
freopen("bai3.out", "w", stdout);
auto start = chrono::high_resolution_clock::now();
#endif
ios_base::sync_with_stdio(false);
cin.tie(nullptr);
int t = 1; //cin >> t;
while(t--) solve();
#ifdef CDuongg
auto end = chrono::high_resolution_clock::now();
cout << "\n"; for(int i = 1; i <= 100; ++i) cout << '=';
cout << "\nExecution time: " << chrono::duration_cast<chrono::milliseconds> (end - start).count() << "[ms]" << endl;
#endif
}
