1603C - Extreme Extension - CodeForces Solution

#include <algorithm>
#include <array>
#include <bitset>
#include <cassert>
#include <chrono>
#include <cstring>
#include <functional>
#include <iomanip>
#include <iostream>
#include <map>
#include <numeric>
#include <queue>
#include <random>
#include <set>
#include <stack>
#include <vector>

using namespace std;

// BEGIN NO SAD
#define rep(i, a, b) for(int i = a; i < (b); ++i)
#define trav(a, x) for(auto& a : x)
#define all(x) x.begin(), x.end()
#define sz(x) (int)(x).size()
#define mp make_pair
#define pb push_back
#define eb emplace_back
#define lb lower_bound
#define ub upper_bound
typedef vector<int> vi;
#define f first
#define s second
#define derr if(1) cerr
// END NO SAD

template<class Fun>
class y_combinator_result {
  Fun fun_;
public:
  template<class T>
  explicit y_combinator_result(T &&fun): fun_(std::forward<T>(fun)) {}

  template<class ...Args>
  decltype(auto) operator()(Args &&...args) {
    return fun_(std::ref(*this), std::forward<Args>(args)...);
  }
};

template<class Fun>
decltype(auto) y_combinator(Fun &&fun) {
  return y_combinator_result<std::decay_t<Fun>>(std::forward<Fun>(fun));
}

template<class T>
bool updmin(T& a, T b) {
  if(b < a) {
    a = b;
    return true;
  }
  return false;
}
template<class T>
bool updmax(T& a, T b) {
  if(b > a) {
    a = b;
    return true;
  }
  return false;
}
typedef int64_t ll;

struct barrett_reduction {
    unsigned mod;
    uint64_t div;
 
    barrett_reduction(unsigned m) : mod(m), div(-1LLU / m) {}
 
    unsigned operator()(uint64_t a) const {
#ifdef __SIZEOF_INT128__
        uint64_t q = uint64_t(__uint128_t(div) * a >> 64);
        uint64_t r = a - q * mod;
        return unsigned(r < mod ? r : r - mod);
#endif
        return unsigned(a % mod);
    }
};
 
template<const int &MOD, const barrett_reduction &barrett>
struct _b_int {
    int val;
 
    _b_int(int64_t v = 0) {
        if (v < 0) v = v % MOD + MOD;
        if (v >= MOD) v %= MOD;
        val = int(v);
    }
 
    _b_int(uint64_t v) {
        if (v >= uint64_t(MOD)) v %= MOD;
        val = int(v);
    }
 
    _b_int(int v) : _b_int(int64_t(v)) {}
    _b_int(unsigned v) : _b_int(uint64_t(v)) {}
 
    static int inv_mod(int a, int m = MOD) {
        // https://en.wikipedia.org/wiki/Extended_Euclidean_algorithm#Example
        int g = m, r = a, x = 0, y = 1;
 
        while (r != 0) {
            int q = g / r;
            g %= r; swap(g, r);
            x -= q * y; swap(x, y);
        }
 
        return x < 0 ? x + m : x;
    }
 
    explicit operator int() const { return val; }
    explicit operator unsigned() const { return val; }
    explicit operator int64_t() const { return val; }
    explicit operator uint64_t() const { return val; }
    explicit operator double() const { return val; }
    explicit operator long double() const { return val; }
 
    _b_int& operator+=(const _b_int &other) {
        val -= MOD - other.val;
        if (val < 0) val += MOD;
        return *this;
    }
 
    _b_int& operator-=(const _b_int &other) {
        val -= other.val;
        if (val < 0) val += MOD;
        return *this;
    }
 
    static unsigned fast_mod(uint64_t x) {
#if !defined(_WIN32) || defined(_WIN64)
        return barrett(x);
#endif
        // Optimized mod for Codeforces 32-bit machines.
        // x must be less than 2^32 * MOD for this to work, so that x / MOD fits in an unsigned 32-bit int.
        unsigned x_high = unsigned(x >> 32), x_low = unsigned(x);
        unsigned quot, rem;
        asm("divl %4\n"
            : "=a" (quot), "=d" (rem)
            : "d" (x_high), "a" (x_low), "r" (MOD));
        return rem;
    }
 
    _b_int& operator*=(const _b_int &other) {
        val = fast_mod(uint64_t(val) * other.val);
        return *this;
    }
 
    _b_int& operator/=(const _b_int &other) {
        return *this *= other.inv();
    }
 
    friend _b_int operator+(const _b_int &a, const _b_int &b) { return _b_int(a) += b; }
    friend _b_int operator-(const _b_int &a, const _b_int &b) { return _b_int(a) -= b; }
    friend _b_int operator*(const _b_int &a, const _b_int &b) { return _b_int(a) *= b; }
    friend _b_int operator/(const _b_int &a, const _b_int &b) { return _b_int(a) /= b; }
 
    _b_int& operator++() {
        val = val == MOD - 1 ? 0 : val + 1;
        return *this;
    }
 
    _b_int& operator--() {
        val = val == 0 ? MOD - 1 : val - 1;
        return *this;
    }
 
    _b_int operator++(int) { _b_int before = *this; ++*this; return before; }
    _b_int operator--(int) { _b_int before = *this; --*this; return before; }
 
    _b_int operator-() const {
        return val == 0 ? 0 : MOD - val;
    }
 
    friend bool operator==(const _b_int &a, const _b_int &b) { return a.val == b.val; }
    friend bool operator!=(const _b_int &a, const _b_int &b) { return a.val != b.val; }
    friend bool operator<(const _b_int &a, const _b_int &b) { return a.val < b.val; }
    friend bool operator>(const _b_int &a, const _b_int &b) { return a.val > b.val; }
    friend bool operator<=(const _b_int &a, const _b_int &b) { return a.val <= b.val; }
    friend bool operator>=(const _b_int &a, const _b_int &b) { return a.val >= b.val; }
 
    _b_int inv() const {
        return inv_mod(val);
    }
 
    _b_int pow(int64_t p) const {
        if (p < 0)
            return inv().pow(-p);
 
        _b_int a = *this, result = 1;
 
        while (p > 0) {
            if (p & 1)
                result *= a;
 
            p >>= 1;
 
            if (p > 0)
                a *= a;
        }
 
        return result;
    }
 
    friend ostream& operator<<(ostream &os, const _b_int &m) {
        return os << m.val;
    }
 
    friend istream& operator>>(istream &is, _b_int &m) {
        int64_t x;
        is >> x;
        m = x;
        return is;
    }
};
 
int MOD = 998244353;
barrett_reduction barrett(MOD);
using mnum = _b_int<MOD, barrett>;

void rsolve() {
  int n;
  cin >> n;
  vector<int> v(n);
  for(auto& x: v) cin >> x;
  vector<mnum> dp(100005);
  vector<mnum> ndp(100005);
  vector<int> lastupd(100005, -1);
  vector<int> vals;
  dp[v[n-1]] = 1;
  vals.pb(v[n-1]);
  lastupd[v[n-1]] = n-1;
  mnum ret = 0;
  for(int i = n-2; i >= 0; i--) {
    ndp[v[i]] = 1;
    lastupd[v[i]] = i;
    vector<int> nvals;
    nvals.pb(v[i]);
    for(auto key: vals) {
      assert(lastupd[key] <= i+1);
      auto val = dp[key];
      if(v[i] <= key) {
        ndp[v[i]] += val;
        continue;
      }
      int need = (v[i] + key - 1) / key;
      ret += (need-1) * val * (i+1);
      if(lastupd[v[i]/need] != i) {
        lastupd[v[i]/need] = i;
        ndp[v[i]/need] = 0;
        nvals.pb(v[i]/need);
      }
      ndp[v[i] / need] += val;
    }
    vals.swap(nvals);
    dp.swap(ndp);
  }
  cout << ret << "\n";
}

void solve() {
  int t;
  cin >> t;
  while(t--) rsolve();
}

// what would chika do
// are there edge cases (N=1?)
// are array sizes proper (scaled by proper constant, for example 2* for koosaga tree)
// integer overflow?
// DS reset properly between test cases
// are you doing geometry in floating points
// are you not using modint when you should

int main() {
  ios_base::sync_with_stdio(false);
  cin.tie(NULL);
  solve();
}