#include <bits/stdc++.h>
using namespace std;
using namespace chrono;
#pragma GCC optimize("Ofast")
#pragma GCC target("tune=native")
#pragma GCC optimize("unroll-loops")
typedef long long ll;
typedef long double ld;
#define fs first
#define sc second
#define all(x) x.begin(), x.end()
#define rall(x) x.rbegin(), x.rend()
#define pii pair<int, int>
#define pll pair<ll, ll>
#define pb push_back
#define maxn 200000
#define BIG 1000000000000000000
mt19937 rnd(42);
int a[1002][1002], a1[4][1002][1002];
int main()
{
ios_base::sync_with_stdio(0);
cin.tie(0);
int n, m;
cin >> n >> m;
for(int i = 1; i <= n; i++){
for(int j = 1; j <= m; j++)
cin >> a[i][j];
}
for(int i = 1; i <= n; i++)
for(int j = 1; j <= m; j++)
a1[0][i][j] = max(a[i][j], max(a1[0][i-1][j], a1[0][i][j-1]));
for(int i = 1; i <= n; i++)
for(int j = m; j > 0; j--)
a1[1][i][j] = max(a[i][j], max(a1[1][i-1][j], a1[1][i][j+1]));
for(int i = n; i > 0; i--)
for(int j = 1; j <= m; j++)
a1[2][i][j] = max(a[i][j], max(a1[0][i+1][j], a1[0][i][j-1]));
for(int i = n; i > 0; i--)
for(int j = m; j > 0; j--)
a1[3][i][j] = max(a[i][j], max(a1[0][i+1][j], a1[0][i][j+1]));
int ans = -1;
for(int i = 1; i <= n; i++)
for(int j = 1; j <= m; j++){
int p = a[i][j];
for(int o = 0; o < 4; o++)
ans = max(ans, a1[o][i][j] - p - 1);
}
cout << ans;
}
/*
2 3
3 4 2
5 7 5
*/
