#pragma GCC optimize("Ofast,unroll-loops")
#pragma GCC target("avx2,popcnt")
#include <bits/stdc++.h>
using namespace std;
const int P = 998244353;
int main() {
  ios::sync_with_stdio(false);
  cin.tie(nullptr);
  int N, M, K;
  cin >> N >> M >> K;
  vector<int> A(N);
  for (int i = 0; i < N; i++) {
    cin >> A[i];
  }
  int Q = N - M + 1;
  vector<int> ans(Q);
  vector<tuple<int, int, int>> querys;
  for (int i = 0; i < Q; i++) {
    querys.emplace_back(i, i + M, i);
  }
  vector<vector<int>> dp(N + 1, vector<int>(K));
  auto add = [&](vector<int> &f, int v) {
    vector<int> g = f;
    for (int i = 0; i < K; i++) {
      int j = (i + v < K ? i + v : i + v - K);
      g[j] += f[i];
      if (g[j] >= P) {
        g[j] -= P;
      }
    }
    return g;
  };
  auto dfs = [&](auto &dfs, int L, int R,
                 vector<tuple<int, int, int>> &query) {
    vector<tuple<int, int, int>> qL, qM, qR;
    int M = (L + R) / 2, lb = M, rb = M;
    for (auto &[l, r, i] : query) {
      if (l <= M && M <= r) {
        lb = min(lb, L);
        rb = max(rb, R);
        qM.emplace_back(l, r, i);
      } else if (r < M) {
        qL.emplace_back(l, r, i);
      } else {
        qR.emplace_back(l, r, i);
      }
    }
    for (int j = 0; j < K; j++) {
      dp[M][j] = (j == 0);
    }
    for (int i = M - 1; i >= lb; i--) {
      dp[i] = add(dp[i + 1], A[i]);
    }
    for (int i = M; i < rb; i++) {
      dp[i + 1] = add(dp[i], A[i]);
    }
    for (auto [l, r, i] : qM) {
      int res = P - 1;
      for (int j = 0; j < K; j++) {
        int k = (j == 0 ? 0 : K - j);
        res += 1ll * dp[l][j] * dp[r][k] % P;
        if (res >= P) {
          res -= P;
        }
      }
      ans[i] = res;
    }
    if (L + 1 == R) {
      return;
    } else {
      dfs(dfs, L, M, qL);
      dfs(dfs, M, R, qR);
    }
  };
  dfs(dfs, 0, N, querys);
  for (int i = 0; i < Q; i++) {
    cout << ans[i] << "\n";
  }
}