#include <iostream>
#include <string>
#include <vector>
#include <algorithm>
#include <map>
#include <atcoder/modint>
using namespace std;
using i32 = int32_t;
using u32 = uint32_t;
using i64 = int64_t;
using u64 = uint64_t;
#define rep(i,n) for(int i=0; i<(int)(n); i++)


const i64 INF = 1001001001001001001;
using modint = atcoder::static_modint<1000000007>;

int GCD(i64 a, i64 b){ return b ? GCD(b, a%b) : a; }


int main(){
    i64 N,M,L; cin >> N >> M >> L;
    vector<i64> A(N);
    rep(i,N) cin >> A[i];
    i64 B = GCD(N, L);
    i64 NbyB = N/B;

    i64 full_length = N * M;

    i64 ans = -INF * 2;

    vector<i64> permA(NbyB);
    vector<i64> indexPermA(NbyB);
    vector<i64> sum_permA(NbyB + 1);

    auto sum_permA_wide = [&](i64 l, i64 r) -> i64 {
        i64 lf = l % NbyB, rf = r % NbyB;
        i64 lH = l / NbyB, rH = r / NbyB;
        return sum_permA[rf] - sum_permA[lf] + (rH - lH) * sum_permA[NbyB];
    };

    rep(t,NbyB) indexPermA[t] = L%N*t%N;
    rep(s,B){
        rep(t,NbyB) permA[t] = A[s + indexPermA[t]];
        rep(t,NbyB) sum_permA[t+1] = sum_permA[t] + permA[t];
        
        rep(t,NbyB){
            i64 s_idx = s + indexPermA[t];
            if(L <= s) continue;
            i64 min_l_idx = s_idx;
            i64 max_l_idx = (L - s_idx - 1) / N * N + s_idx;
            if(min_l_idx > max_l_idx) continue;
            ans = max(ans, sum_permA_wide(t, t + (full_length - 1 - min_l_idx) / L + 1));
            ans = max(ans, sum_permA_wide(t, t + (full_length - 1 - max_l_idx) / L + 1));
        }
    }

    cout << ans << '\n';

    return 0;
}


struct ios_do_not_sync{
    ios_do_not_sync(){
        std::ios::sync_with_stdio(false);
        std::cin.tie(nullptr);
    }
} ios_do_not_sync_instance;