#include <bits/stdc++.h>
#pragma GCC optimize ("O3")
#pragma GCC target ("sse4")
using namespace std;
#define int long long int
const int N = 1e3 + 1;
const int INF = 1e9;
const int md = 1e9 + 7;
int32_t main(int32_t argc, char *argv[]) {
ios::sync_with_stdio(false);
cin.tie(nullptr);
int T = 1;
// cin >> T;
while (T--) {
int n, m;
cin >> n >> m;
vector<vector<int>> a(n + 1, vector<int> (m + 1));
for (int i = 1; i <= n; i++)
for (int j = 1; j <= m; j++)
cin >> a[i][j];
vector<vector<int>> mx1(n + 1, vector<int> (m + 1, -INF)), mn1(n + 1, vector<int> (m + 1, INF));
vector<vector<int>> mx2(n + 1, vector<int> (m + 1, -INF)), mn2(n + 1, vector<int> (m + 1, INF));
for (int j = 1; j <= m; j++)
for (int i = 1; i <= n; i++) {
mx1[i][j] = max({mx1[i - 1][j], mx1[i][j - 1], (a[i][j] + i + j)});
mn1[i][j] = min({mn1[i - 1][j], mn1[i][j - 1], (a[i][j] - i - j)});
}
for (int j = m; j >= 1; j--)
for (int i = 1; i <= n; i++) {
mx2[i][j] = max({mx2[i - 1][j], mx2[i][j + 1], (a[i][j] + i - j)});
mn2[i][j] = min({mn2[i - 1][j], mn2[i][j + 1], (a[i][j] - i + j)});
}
int ans = -1;
for (int i = 1; i <= n; i++)
for (int j = 1; j <= m; j++) {
ans = max(ans, (a[i][j] - i - j) - mn1[i][j] - 1);
ans = max(ans, mx1[i][j] - (a[i][j] + i + j) - 1);
ans = max(ans, (a[i][j] - i - j) - mn2[i][j] - 1);
ans = max(ans, mx2[i][j] - (a[i][j] + i + j) - 1);
}
cout << ans << '\n';
}
return 0;
}
