This was the first round when I had prepared some templates prior the contest.

    #include <iostream>
    #include <vector>
    
    #define FOR(it, seq) for(auto it : seq)
    #define FORI(it, beg, end) for(auto it = beg; it != end; ++it)
    
    #define DBG(x) cout << x
    #define DBGV(v) {cout << "["; FOR(i_, v) {DBG(i_); cout << ", ";} cout << "]" << endl; }
    
    using namespace std;
    
    typedef long long lint;
    typedef vector<int> VI;
    typedef vector<lint> VL;
    
    int T;
    
    int main(int argc, char *argv[])
    {
            cin >> T;
    
            FORI(t, 0, T) {
                    cout << "Test case #" << (t+1) << ": ";
    
                    // print answer
    
                    cout << endl;
            }
    
    
            return 0;
    }

My goal was to win the T-shirt, i.e. be in the first 1000 contestants. I was sligthly surprised that there were four tasks instead of ordinary three — I assigned uniformly 37 minutes per task.

Problem A (Pegman)

I mapped the problem to a graph problem, thought I’d implement it with some graph representation, then I realized I could use the existing grid and began implementing. I spent inappropriate amount of time on input parsing (skipping EOL characters using cint >> EOL, which was not necessary).

After submitting the problem A, I was somewhere around 1200th place and I was a bit behind my schedule (it took me ~50 minutes).

Problem B (Kiddie Pool)

First I didn’t like as I saw floating point numbers in the sample output, then I realized it wouldn’t be that bad. The general solution reminded me a sum of a subset problem (may be it is not), on the other hand small solution (with two pipes only) had a closed form solution. I had it almost right, except I forgot special case when there were two pipes with same temperature. That cost me few minutes.

After submitting the problem B, I was somewhere around 550th place. Simple solution took me ~25 minutes. I decided to skip problem C (it offered only a few points, I didn’t even read the task) and attempted to solve problem D.

Update 2015-08-02: Later I found out, it is actually a linear programming problem. How ignorant was I.

Problem D (Mathdrum)

I thought even small input would be impossible to bruteforce ($4^{36} \approx 10^{21}$, actually only $3^{36} \approx 10^{17}$). However, I thought some pruned backtracking could be feasible. I spent some time implementing, however I wasn’t able to debug it to working version (I didn’t reduce no. of unique solution due to cyclic shift and for more than 2 row it ran too long and probably not correctly). I didn’t submit anything.

When deadline passed, I was 1226th and when wrong solution of large input were reconciled, I moved to 1092nd place. That’s less than 100 place from my (pre)set goal. I was never so far in GCJ before, I can be delighted by making progress since first years I participated. On the other hand, there are hundreds of better programmers.