#pragma GCC target ("avx2")
#pragma GCC optimization ("O3")
#pragma GCC optimization ("unroll-loops")
#include <bits/stdc++.h>
using namespace std;
#define ll long long
#define db double
#define gcd(a,b) __gcd(a,b)
#define lcm(a,b) a*b / __gcd(a,b)
#define I first
#define II second
#define pb push_back
#define ii pair<int,int>
const int INF = 2e9;
const int N = 6e2 + 5;
const int MOD = 1e9 + 7;
int n, m, k;
mt19937 rd(chrono::steady_clock::now().time_since_epoch().count());
ll Rand(ll l,ll r)
{
ll k = rd() * rd();
if (k < 0) k = -k;
return l + k % (r - l + 1);
}
namespace sub3
{
int tg[N], num[N];
ll ans = 1e18;
ii p[N];
array<int, 3> cur[N], f[N][N]; // L, R, M
struct DQ
{
deque<ii> q;
inline void add(ii x)
{
while (q.size() && q.back().I <= x.I) q.pop_back();
q.pb(x);
}
inline void del(int& x)
{
while (q.size() && q.front().II <= x) q.pop_front();
}
inline int get()
{
if (q.empty()) return INF; else
return q.front().I;
}
} L, R, M;
inline void init()
{
for (int l = 1; l <= k; l++)
{
int sz = 0;
for (int r = l; r <= k; r++)
{
int mx = p[r].II;
for (int i = 1; i <= sz; i++)
if (mx < tg[i]) swap(tg[i], mx);
tg[++sz] = mx;
f[l][r][0] = tg[1] - 1;
f[l][r][1] = m - tg[sz];
for (int i = 1; i < sz; i++)
f[l][r][2] = max(f[l][r][2], tg[i + 1] - tg[i] - 1);
// cout << l << ' ' << r << ' ' << f[l][r][0] << ' ' << f[l][r][1] << ' ' << f[l][r][2] << '\n';
}
}
}
inline void solve(int le)
{
if (le >= ans) return;
int sz = 0;
for (int l = 1, r = 1; l <= k; l++)
{
while (r <= k && p[r].I <= p[l].I + le)
{
int val = p[r++].I;
if (val > num[sz]) num[++sz] = val;
}
if (p[l].I + le + 1 > num[sz]) num[++sz] = p[l].I + le + 1;
}
for (int i = 1, l = 1, r = 1; i <= sz; i++)
{
while (l <= k && !(p[l].I <= num[i] && num[i] <= p[l].I + le)) l++;
r = max(r, l);
while (r <= k && p[r].I <= num[i] && num[i] <= p[r].I + le) r++;
if (l > k) cur[i] = {INF, INF, INF}; else
cur[i] = f[l][r - 1];
}
for (int l = 1, r = 1; l <= sz; l++)
{
r = max(r, l);
while (r <= sz && num[r] < num[l] + n)
{
L.add({cur[r][0], r});
R.add({cur[r][1], r});
M.add({cur[r][2], r++});
}
// cout << le << ' ' << L.get() << ' ' << R.get() << ' ' << M.get() << '\n';
ans = min(ans, 1ll * le + 1ll * max(L.get() + R.get(), M.get()));
L.del(l);
R.del(l);
M.del(l);
}
}
int main()
{
for (int i = 1; i <= k; i++) cin >> p[i].I >> p[i].II;
vector<int> t;
sort(p + 1, p + k + 1);
for (int i = 1; i <= k; i++)
{
t.pb(p[i].I - 1);
t.pb(n - p[i].I);
for (int j = 1; j <= k; j++)
{
t.pb(max(0, p[j].I - p[i].I - 1));
t.pb(p[i].I - 1 + n - p[j].I);
}
}
sort(t.begin(), t.end());
t.resize(unique(t.begin(), t.end()) - t.begin());
// for (auto c : t) cout << c << ' ';cout << '\n';
int xx = t.size() - 1;
for (int i = 1; i <= 100000; i++)
{
int u = Rand(0, xx), v = Rand(0, xx);
swap(t[u], t[v]);
}
init();
for (auto l : t) solve(l);
cout << ans;
return 0;
}
}
int32_t main()
{
#define TASKNAME ""
ios_base::sync_with_stdio(0);
cin.tie(0);
if (fopen(TASKNAME".inp","r" ))
{
freopen(TASKNAME".inp","r",stdin);
freopen(TASKNAME".ans","w",stdout);
}
cin >> n >> m >> k;
// if (max(n, m) <= 40) sub1 :: main(); else
// if (k <= 25) sub2 :: main(); else
sub3 :: main();
return 0;
}
