#pragma GCC target("avx2")
#pragma GCC optimize("Ofast,unroll-loops")
#include <bits/stdc++.h>
#include <atcoder/all>
#define rep(i, n) for (int i = 0; i < n; i++)
#define per(i, n) for (int i = n - 1; i >= 0; i--)
#define ALL(a) a.begin(), a.end()
#undef long
#define long long long
#define ll long
#define vec vector
using namespace std;
using mint = atcoder::modint;
ostream &operator<<(ostream &os, mint a)
{
	return os << a.val();
}
template <typename T>
ostream &operator<<(ostream &os, vector<T> &a)
{
	const int n = a.size();
	rep(i, n)
	{
		os << a[i];
		if (i + 1 != n)
			os << " ";
	}
	return os;
}
template <typename T, size_t n>
ostream &operator<<(ostream &os, array<T, n> &a)
{
	rep(i, n) os << a[i] << " \n"[i + 1 == n];
	return os;
}
template <typename T>
istream &operator>>(istream &is, vector<T> &a)
{
	for (T &i : a)
		is >> i;
	return is;
}
template <typename T>
bool chmin(T &x, T y)
{
	if (x > y)
	{
		x = y;
		return true;
	}
	return false;
}
template <typename T>
bool chmax(T &x, T y)
{
	if (x < y)
	{
		x = y;
		return true;
	}
	return false;
}
void solve()
{
	int h, w;
	cin >> h >> w;
	vec<vec<int>> a(h, vec<int>(w));
	rep(i, h) rep(j, w) cin >> a[i][j];
	vec<vec<int>> g(h * w);
	constexpr int dx[] = {0, 1, 0, -1};
	constexpr int dy[] = {1, 0, -1, 0};
	rep(i, h) rep(j, w)
	{
		rep(k, 4)
		{
			int nx = i + dx[k];
			int ny = j + dy[k];
			if (nx < 0 || nx >= h || ny < 0 || ny >= w)
				continue;
			if (a[i][j] < a[nx][ny])
				g[i * w + j].push_back(nx * w + ny);
		}
	}
	queue<int> q;
	vec<int> d(h * w);
	rep(i, h * w) for (int j : g[i]) d[j]++;
	rep(i, h * w) if (d[i] == 0) q.push(i);
	vec<int> dp(h * w);
	while (!q.empty())
	{
		const int v = q.front();
		q.pop();
		for (int u : g[v])
		{
			chmax(dp[u], dp[v] + 1);
			if (--d[u] == 0)
				q.push(u);
		}
	}
	cout << *max_element(ALL(dp)) + 1 << endl;
}
int main()
{
	// srand((unsigned)time(NULL));
	cin.tie(nullptr);
	ios::sync_with_stdio(false);
	// cout << fixed << setprecision(40);
	int t = 1;
	// cin >> t;
	while (t--)
		solve();
	return 0;
}