#include <iostream>
#include <iomanip>
#include <cassert>
#include <vector>
#include <algorithm>
#include <utility>
#include <numeric>
#include <tuple>
#include <ranges>
#include <random>
// #include "Src/Number/IntegerDivision.hpp"
// #include "Src/Utility/BinarySearch.hpp"
// #include "Src/Sequence/CompressedSequence.hpp"
// #include "Src/Sequence/RunLengthEncoding.hpp"
// #include "Src/Algebra/Group/AdditiveGroup.hpp"
// #include "Src/DataStructure/FenwickTree/FenwickTree.hpp"


#include <cstdint>
#include <cstddef>

namespace zawa {

using i16 = std::int16_t;
using i32 = std::int32_t;
using i64 = std::int64_t;
using i128 = __int128_t;

using u8 = std::uint8_t;
using u16 = std::uint16_t;
using u32 = std::uint32_t;
using u64 = std::uint64_t;

using usize = std::size_t;

} // namespace zawa


#include <concepts>

namespace zawa {

namespace concepts {

template <class T>
concept Semigroup = requires {
    typename T::Element;
    { T::operation(std::declval<typename T::Element>(), std::declval<typename T::Element>()) } -> std::same_as<typename T::Element>;
};

} // namespace concepts

} // namespace zawa


namespace zawa {

namespace concepts {

template <class T>
concept Identitiable = requires {
    typename T::Element;
    { T::identity() } -> std::same_as<typename T::Element>;
};

template <class T>
concept Monoid = Semigroup<T> and Identitiable<T>;

} // namespace

} // namespace zawa

#include <functional>
#include <type_traits>
#include <ostream>

namespace zawa {

template <concepts::Monoid Monoid>
class SegmentTree {
public:

    using VM = Monoid;

    using V = typename VM::Element;

    using OM = Monoid;

    using O = typename OM::Element;

    SegmentTree() = default;

    explicit SegmentTree(usize n) : m_n{ n }, m_dat(n << 1, VM::identity()) {}

    explicit SegmentTree(const std::vector<V>& dat) : m_n{ dat.size() }, m_dat(dat.size() << 1, VM::identity()) {
        for (usize i{} ; i < m_n ; i++) {
            m_dat[i + m_n] = dat[i];
        }
        for (usize i{m_n} ; i-- ; ) {
            m_dat[i] = VM::operation(m_dat[left(i)], m_dat[right(i)]);
        }
    }

    [[nodiscard]] inline usize size() const noexcept {
        return m_n;
    }

    [[nodiscard]] V get(usize i) const {
        assert(i < size());
        return m_dat[i + m_n];
    }

    [[nodiscard]] V operator[](usize i) const {
        assert(i < size());
        return m_dat[i + m_n];
    }

    void operation(usize i, const O& value) {
        assert(i < size());
        i += size();
        m_dat[i] = OM::operation(m_dat[i], value);
        while (i = parent(i), i) {
            m_dat[i] = VM::operation(m_dat[left(i)], m_dat[right(i)]);
        }
    }

    void assign(usize i, const V& value) {
        assert(i < size());
        i += size();
        m_dat[i] = value;
        while (i = parent(i), i) {
            m_dat[i] = VM::operation(m_dat[left(i)], m_dat[right(i)]);
        }
    }

    [[nodiscard]] V product(u32 l, u32 r) const {
        assert(l <= r and r <= size());
        V L{ VM::identity() }, R{ VM::identity() };
        for (l += size(), r += size() ; l < r ; l = parent(l), r = parent(r)) {
            if (l & 1) {
                L = VM::operation(L, m_dat[l++]);
            }
            if (r & 1) {
                R = VM::operation(m_dat[--r], R);
            }
        }
        return VM::operation(L, R);
    }

    template <class F>
    requires std::predicate<F, V>
    [[nodiscard]] usize maxRight(usize l, const F& f) {
        assert(l < size());
        static_assert(std::is_convertible_v<decltype(f), std::function<bool(V)>>, "maxRight's argument f must be function bool(T)");
        assert(f(VM::identity()));
        usize res{l}, width{1};
        V prod{ VM::identity() };
        for (l += size() ; res + width <= size() ; l = parent(l), width <<= 1) if (l & 1) {
            if (not f(VM::operation(prod, m_dat[l]))) break; 
            res += width;
            prod = VM::operation(prod, m_dat[l++]);
        }
        while (l = left(l), width >>= 1) {
            if (res + width <= size() and f(VM::operation(prod, m_dat[l]))) {
                res += width;
                prod = VM::operation(prod, m_dat[l++]);
            } 
        }
        return res;
    }

    template <class F>
    requires std::predicate<F, V>
    [[nodiscard]] usize minLeft(usize r, const F& f) const {
        assert(r <= size());
        static_assert(std::is_convertible_v<decltype(f), std::function<bool(V)>>, "minLeft's argument f must be function bool(T)");
        assert(f(VM::identity()));
        usize res{r}, width{1};
        V prod{ VM::identity() };
        for (r += size() ; res >= width ; r = parent(r), width <<= 1) if (r & 1) {
            if (not f(VM::operation(m_dat[r - 1], prod))) break;
            res -= width;
            prod = VM::operation(prod, m_dat[--r]);
        }
        while (r = left(r), width >>= 1) {
            if (res >= width and f(VM::operation(m_dat[r - 1], prod))) {
                res -= width;
                prod = VM::operation(m_dat[--r], prod);
            }
        }
        return res;
    }

    friend std::ostream& operator<<(std::ostream& os, const SegmentTree& st) {
        for (usize i{1} ; i < 2 * st.size() ; i++) {
            os << st.m_dat[i] << (i + 1 == 2 * st.size() ? "" : " ");
        }
        return os;
    }

private:

    constexpr u32 left(u32 v) const {
        return v << 1;
    }

    constexpr u32 right(u32 v) const {
        return v << 1 | 1;
    }

    constexpr u32 parent(u32 v) const {
        return v >> 1;
    }

    usize m_n;

    std::vector<V> m_dat;
};

} // namespace zawa
// #include "Src/DataStructure/DisjointSetUnion/DisjointSetUnion.hpp"
// #include "Src/DataStructure/Heap/BinaryHeap.hpp"
namespace zawa {}
using namespace zawa;
// #include "atcoder/modint"
// using mint = atcoder::modint998244353;
// #include <array>
// #include <bit>
// #include <bitset>
// #include <climits>
// #include <cmath>
// #include <set>
// #include <unordered_set>
// #include <map>
// #include <unordered_map>
// #include <optional>
// #include <queue>
// #include <stack>
// #include <deque>
// #pragma GCC target("avx2")
// #pragma GCC optimize("O3")
// #pragma GCC optimize("unroll-loops")
using namespace std;
template <class T, class U>
ostream& operator<<(ostream& os, const pair<T, U>& p) {
    os << '(' << p.first << ',' << p.second << ')';
    return os;
}
template <class T>
ostream& operator<<(ostream& os, const vector<T>& v) {
    for (int i = 0 ; i < ssize(v) ; i++)
        os << v[i] << (i + 1 == ssize(v) ? "" : " ");
    return os;
}
/*
 * 注: 誤読していたので、B日働いて、C日休む場合で解いている
 */
const long long INF=(long long)-1e18;
struct MAX {
   using Element = long long;
   static Element identity() {
       return INF;
   }
   static Element operation(Element L, Element R) {
       return max(L,R);
   }
};
long long solve(int N,int B,int C,vector<long long> A) {
    vector<long long> S(N+1);
    for (int i = 0 ; i < N ; i++)
        S[i+1]=S[i]+A[i];
    for (int i = 0 ; i < 10 ; i++)
        S.push_back(S.back());
    SegmentTree<MAX> dp(N),ep(N);
    for (int i = 0 ; i < B-1 ; i++) {
        long long kiyo=S[i+1];
        dp.assign(i,kiyo-S[i+2]);
        ep.assign(i,kiyo);
    }
    for (int i = B-1 ; i < N ; i++) {
        long long kiyo=INF;
        if (i <= B+C-2)
            kiyo=max(kiyo,S[i+1]-S[max(0,i-B+2)]);
        kiyo=max(kiyo,dp.product(max(i-B,0),max(i-1,0))+S[i+1]);
        kiyo=max(kiyo,ep.product(max(i-B-C+2,0),max(i-B,0))+S[i+1]-S[i-B+2]);
        dp.assign(i,kiyo-S[i+2]);
        ep.assign(i,kiyo);
    }
    long long ans=INF;
    for (int i = 0 ; i < min(N,C) ; i++)
        ans=max(ans,ep[N-i-1]);
    return ans;
}
long long naive(int N,int B,int C,vector<long long> A) {
    vector<long long> S(N+1);
    for (int i = 0 ; i < N ; i++)
        S[i+1]=S[i]+A[i];
    vector<long long> dp(N,INF);
    for (int i = 0 ; i < min(N,B) ; i++)
        dp[i]=S[i+1];
    for (int i = 0 ; i < N ; i++)
        for (int j = 1 ; j < C ; j++) {
            long long sum=0;
            for (int k = 1 ; k < B and i + j + k < N ; k++) {
                sum+=A[i+j+k];
                dp[i+j+k]=max(dp[i+j+k],dp[i]+sum);
            }
        }
    long long ans=INF;
    for (int i = 0 ; i < min(C,N) ; i++)
        ans=max(ans,dp[N-i-1]);
    return ans;
}
long long brute(int N,int B,int C,vector<long long> A) {
    vector<int> col(N);
    auto dfs=[&](auto dfs,int i)->long long {
        if (i==N) {
            bool ok=1;
            for (int l = 0, r = 0 ; l < N ; l = r) {
                while (r < N and col[l] == col[r])
                    r++;
                if (col[l] == 1 and r-l >= B)
                    ok=0;
                if (col[l] == 0 and r-l >= C)
                    ok=0;
            }
            if (!ok)
                return INF;
            long long res=0;
            for (int j = 0 ; j < N ; j++)
                res+=col[j]*A[j];
            return res;
        }
        long long ans=INF;
        col[i]=0;
        ans=max(ans,dfs(dfs,i+1));
        col[i]=1;
        ans=max(ans,dfs(dfs,i+1));
        return ans;
    };
    return dfs(dfs,0);
}
int main() {
    cin.tie(0);
    cout.tie(0);
    ios::sync_with_stdio(0);
    cout << fixed << setprecision(20);
#if !defined DEBUG
    int N,B,C;
    cin >> N >> B >> C;
    swap(B,C);
    vector<long long> A(N);
    for (auto& x : A)
        cin >> x;
    cout << solve(N,B,C,A) << '\n';
#else
    mt19937_64 mt{random_device{}()};
    for (int testcase = 0 ; ; ) {
        cerr << "----------" << ++testcase << "----------" << endl;
        int N=mt()%7+1;
        int B=1,C=1;
        while (B == 1 or C == 1) {
            B=mt()%(N+1)+1;
            C=mt()%(N+1)+1;
        }
        vector<long long> A(N);
        for (long long& a : A)
            a = mt()%5+1;
        auto a = solve(N,C,B,A), b = brute(N,C,B,A);
        if (a != b) {
            // print testcase
            cout << N << ' ' << B << ' ' << C << '\n';
            cout << A << endl;
            cerr << "you: " << a << endl;
            cout << "correct: " << b << endl;
            exit(0);
        }
    }
#endif
}