this program isn't AC (WA test#10) because of the
precision of the answer, but Pascal solution using the
same ideas is AC. If you can help me to deal with the
precision I would be very grateful to you. Still i'd
recommend you to write the problem in Pascal to avoid all
potential precision problems :

#include <iostream>
#include <cmath>

using namespace std;

struct cross{int x,y;};

int main()
{
    int m,n,k,i;
    cross a[100];
    cin >> m >> n >> k;

    if (!k) {cout << 100*(m+n); return 0;}

    for (i=0;i<k;i++) cin >> a[i].x >> a[i].y;

    bool f=false;
    cross t;

    while (!f)
    {
        f=true;
        for (i=0;i<k-1;i++)
            if (a[i].x>a[i+1].x)
                {t=a[i]; a[i]=a[i+1]; a[i+1]=t; f=false;}
            else if (a[i].x==a[i+1].x && a[i].y>a[i+1].y)
                {t=a[i]; a[i]=a[i+1]; a[i+1]=t; f=false;}
    }

    int d[100];
    int mx,j;

    d[k-1]=1;
    for (i=k-2;i>=0;i--)
    {
        d[i]=1; mx=0;
        for (j=k-1;j>i;j--)
            if (a[i].x<a[j].x && a[i].y<a[j].y && d[j]>mx) mx=d[j];
        d[i]+=mx;
    }

    double ans;

    ans=(double)100*((double)m+(double)n-(double)2*d[0]+sqrt((double)2)*(double)d[0]);
    cout << floor(ans+(double).5);

    return 0;
}