Submission #1061174

# Submission time Handle Problem Language Result Execution time Memory
1061174 2024-08-16T06:59:03 Z hashiryo Potatoes and fertilizers (LMIO19_bulves) C++17
30 / 100
121 ms 14932 KB
// #define _GLIBCXX_DEBUG
#include <bits/stdc++.h>
// clang-format off
std::ostream&operator<<(std::ostream&os,std::int8_t x){return os<<(int)x;}
std::ostream&operator<<(std::ostream&os,std::uint8_t x){return os<<(int)x;}
std::ostream&operator<<(std::ostream&os,const __int128_t &u){if(!u)os<<"0";__int128_t tmp=u<0?(os<<"-",-u):u;std::string s;while(tmp)s+='0'+(tmp%10),tmp/=10;return std::reverse(s.begin(),s.end()),os<<s;}
std::ostream&operator<<(std::ostream&os,const __uint128_t &u){if(!u)os<<"0";__uint128_t tmp=u;std::string s;while(tmp)s+='0'+(tmp%10),tmp/=10;return std::reverse(s.begin(),s.end()),os<<s;}
#define checkpoint() (void(0))
#define debug(...) (void(0))
#define debugArray(x,n) (void(0))
#define debugMatrix(x,h,w) (void(0))
// clang-format on
// clang-format off
template<class T>struct make_long{using type= T;};
template<>struct make_long<int8_t>{using type= int16_t;};
template<>struct make_long<uint8_t>{using type= uint16_t;};
template<>struct make_long<int16_t>{using type= int32_t;};
template<>struct make_long<uint16_t>{using type= uint32_t;};
template<>struct make_long<int32_t>{using type= int64_t;};
template<>struct make_long<uint32_t>{using type= uint64_t;};
template<>struct make_long<int64_t>{using type= __int128_t;};
template<>struct make_long<uint64_t>{using type= __uint128_t;};
template<>struct make_long<float>{using type= double;};
template<>struct make_long<double>{using type= long double;};
template<class T> using make_long_t= typename make_long<T>::type;
// clang-format on
template <class T, bool persistent= false, size_t NODE_SIZE= 1 << 22> class PiecewiseLinearConvex {
 using D= make_long_t<T>;
 struct Node {
  int ch[2];
  T z, x, d, a;
  D s;
  size_t sz;
  friend std::ostream &operator<<(std::ostream &os, const Node &t) { return os << "{z:" << t.z << ",x:" << t.x << ",d:" << t.d << ",a:" << t.a << ",s:" << t.s << ",sz:" << t.sz << ",ch:(" << t.ch[0] << "," << t.ch[1] << ")}"; }
 };
 static inline size_t ni= 1;
 static inline Node *n= new Node[NODE_SIZE];
 static inline void info(int t, int d, std::stringstream &ss) {
  if (!t) return;
  info(n[t].ch[0], d + 1, ss);
  for (int i= 0; i < d; ++i) ss << "   ";
  ss << " ■ " << n[t] << '\n', info(n[t].ch[1], d + 1, ss);
 }
 static inline void dump_xs(int t, std::vector<T> &xs) {
  if (t) push(t), dump_xs(n[t].ch[0], xs), xs.push_back(n[t].x), dump_xs(n[t].ch[1], xs);
 }
 static inline void dump_slopes_l(int t, T ofs, std::vector<T> &as) {
  if (t) push(t), dump_slopes_l(n[t].ch[1], ofs, as), ofs+= n[n[t].ch[1]].a + n[t].d, as.push_back(-ofs), dump_slopes_l(n[t].ch[0], ofs, as);
 }
 static inline void dump_slopes_r(int t, T ofs, std::vector<T> &as) {
  if (t) push(t), dump_slopes_r(n[t].ch[0], ofs, as), ofs+= n[n[t].ch[0]].a + n[t].d, as.push_back(ofs), dump_slopes_r(n[t].ch[1], ofs, as);
 }
 static inline int create(T d, T x) { return n[ni].d= d, n[ni].x= x, n[ni].z= 0, ni++; }
 template <class Iter> static inline int build(Iter bg, Iter ed) {
  if (bg == ed) return 0;
  auto md= bg + (ed - bg) / 2;
  int t= create(md->first, md->second);
  return n[t].ch[0]= build(bg, md), n[t].ch[1]= build(md + 1, ed), update(t), t;
 }
 template <class Iter> static inline void dump(Iter itr, int t) {
  if (!t) return;
  push(t);
  size_t sz= n[n[t].ch[0]].sz;
  dump(itr, n[t].ch[0]), *(itr + sz)= {n[t].d, n[t].x}, dump(itr + sz + 1, n[t].ch[1]);
 }
 static inline void update(int t) {
  int l= n[t].ch[0], r= n[t].ch[1];
  n[t].sz= 1 + n[l].sz + n[r].sz, n[t].a= n[t].d + n[l].a + n[r].a, n[t].s= D(n[t].x) * n[t].d + n[l].s + n[r].s;
 }
 template <bool b= 1> static inline void prop(int &t, T v) {
  if constexpr (persistent && b) {
   if (!t) return;
   n[ni]= n[t], t= ni++;
  }
  n[t].z+= v, n[t].s+= D(v) * n[t].a, n[t].x+= v;
 }
 static inline void push(int t) {
  if (n[t].z != 0) prop(n[t].ch[0], n[t].z), prop(n[t].ch[1], n[t].z), n[t].z= 0;
 }
 template <bool r> static inline int join_(int t, int a, int b) {
  push(a);
  if constexpr (r) b= join<0>(b, t, n[a].ch[0]);
  else b= join<0>(n[a].ch[1], t, b);
  if constexpr (persistent) n[ni]= n[a], a= ni++;
  if (n[n[a].ch[r]].sz * 4 >= n[b].sz) return n[a].ch[!r]= b, update(a), a;
  return n[a].ch[!r]= n[b].ch[r], update(a), n[b].ch[r]= a, update(b), b;
 }
 template <bool b= 1> static inline int join(int l, int t, int r) {
  if constexpr (persistent && b) n[ni]= n[t], t= ni++;
  if (n[l].sz > n[r].sz * 4) return join_<0>(t, l, r);
  if (n[r].sz > n[l].sz * 4) return join_<1>(t, r, l);
  return n[t].ch[0]= l, n[t].ch[1]= r, update(t), t;
 }
 static inline std::array<int, 3> split(int t, T x) {
  if (!t) return {0, 0, 0};
  push(t);
  if (n[t].x < x) {
   auto [a, b, c]= split(n[t].ch[1], x);
   return {join(n[t].ch[0], t, a), b, c};
  } else if (x < n[t].x) {
   auto [a, b, c]= split(n[t].ch[0], x);
   return {a, b, join(c, t, n[t].ch[1])};
  }
  return {n[t].ch[0], t, n[t].ch[1]};
 }
 static inline int unite(int l, int r) {
  if (!l) return r;
  if (!r) return l;
  push(l);
  if constexpr (persistent) n[ni]= n[l], l= ni++;
  auto [a, b, c]= split(r, n[l].x);
  return n[l].d+= n[b].d, join<0>(unite(a, n[l].ch[0]), l, unite(n[l].ch[1], c));
 }
 static inline int insert(int t, T x, T d) {
  if (!t) return n[ni]= Node{{0, 0}, 0, x, d, d, D(x) * d, 1}, ni++;
  push(t);
  if constexpr (persistent) n[ni]= n[t], t= ni++;
  if (n[t].x == x) return n[t].d+= d, update(t), t;
  return x < n[t].x ? join<0>(insert(n[t].ch[0], x, d), t, n[t].ch[1]) : join<0>(n[t].ch[0], t, insert(n[t].ch[1], x, d));
 }
 template <bool r> static inline std::pair<int, int> pop(int t) {
  if (push(t); !n[t].ch[r]) return {n[t].ch[!r], t};
  auto [a, s]= pop<r>(n[t].ch[r]);
  if constexpr (r) return {join(n[t].ch[!r], t, a), s};
  else return {join(a, t, n[t].ch[!r]), s};
 }
 template <bool r> static inline bool lt(T a, T b) {
  if constexpr (r) return b < a;
  else return a < b;
 }
 template <bool r> static inline int cut(int t, T x) {
  if (!t) return t;
  if (push(t); n[t].x == x) return n[t].ch[!r];
  if (lt<r>(n[t].x, x)) return cut<r>(n[t].ch[!r], x);
  if constexpr (r) return join(n[t].ch[0], t, cut<1>(n[t].ch[1], x));
  else return join(cut<0>(n[t].ch[0], x), t, n[t].ch[1]);
 }
 template <bool r> static inline D calc_y(int t, T x, T ol, D ou) {
  for (; t;) {
   if (push(t); lt<r>(n[t].x, x)) t= n[t].ch[!r];
   else {
    if (ol+= n[n[t].ch[!r]].a, ou+= n[n[t].ch[!r]].s; n[t].x == x) break;
    ol+= n[t].d, ou+= D(n[t].x) * n[t].d, t= n[t].ch[r];
   }
  }
  return D(x) * ol - ou;
 }
 template <bool r> static inline std::array<int, 3> split(int t, T p, T &ol, D &ou) {
  push(t);
  T s= ol + n[n[t].ch[!r]].a;
  if (p < s) {
   auto [a, b, c]= split<r>(n[t].ch[!r], p, ol, ou);
   if constexpr (r) return {a, b, join(c, t, n[t].ch[r])};
   else return {join(n[t].ch[r], t, a), b, c};
  }
  ol= s + n[t].d;
  if (ol < p) {
   ou+= n[n[t].ch[!r]].s + D(n[t].x) * n[t].d;
   auto [a, b, c]= split<r>(n[t].ch[r], p, ol, ou);
   if constexpr (r) return {join(n[t].ch[!r], t, a), b, c};
   else return {a, b, join(c, t, n[t].ch[!r])};
  }
  ou+= n[n[t].ch[!r]].s;
  return {n[t].ch[0], t, n[t].ch[1]};
 }
 template <bool l> static inline bool lte(T a, T b) {
  if constexpr (l) return a < b;
  else return a <= b;
 }
 template <bool l, bool r> static inline std::pair<int, int> split_cum(int t, T p, T &ol, D &ou) {
  push(t);
  T s= ol + n[n[t].ch[!r]].a;
  if (lte<l>(p, s)) {
   auto [c, b]= split_cum<l, r>(n[t].ch[!r], p, ol, ou);
   if constexpr (l) {
    if constexpr (r) return {join(c, t, n[t].ch[r]), b};
    else return {join(n[t].ch[r], t, c), b};
   } else return {c, b};
  }
  ol= s + n[t].d;
  if (lte<!l>(ol, p)) {
   ou+= n[n[t].ch[!r]].s + D(n[t].x) * n[t].d;
   auto [a, b]= split_cum<l, r>(n[t].ch[r], p, ol, ou);
   if constexpr (l) return {a, b};
   else {
    if constexpr (r) return {join(n[t].ch[!r], t, a), b};
    else return {join(a, t, n[t].ch[!r]), b};
   }
  }
  ou+= n[n[t].ch[!r]].s;
  return {n[t].ch[!r ^ l], t};
 }
 int mn, lr[2];
 bool bf[2];
 T o[2], rem, bx[2];
 D y;
 inline D calc_y(T x) {
  if (!mn) return 0;
  if (n[mn].x == x) return 0;
  return x < n[mn].x ? -calc_y<0>(lr[0], x, o[0], D(n[mn].x) * o[0]) : calc_y<1>(lr[1], x, o[1], D(n[mn].x) * o[1]);
 }
 inline void slope_eval(bool neg) {
  T p= neg ? -rem : rem, ol= o[neg];
  if (p <= ol) o[neg]-= p, o[!neg]+= p, y+= D(n[mn].x) * rem;
  else {
   D ou= D(n[mn].x) * ol;
   auto [a, b, c]= neg ? split<1>(lr[neg], p, ol, ou) : split<0>(lr[neg], p, ol, ou);
   o[neg]= ol - p, ol-= n[b].d, ou+= D(n[b].x) * (o[!neg]= p - ol);
   if (neg) y-= ou, lr[!neg]= join(lr[!neg], mn, a), lr[neg]= c;
   else y+= ou, lr[!neg]= join(c, mn, lr[!neg]), lr[neg]= a;
   mn= b;
  }
  rem= 0;
 }
 template <bool l, bool neg> inline void slope_eval_cum() {
  T p= neg ? -rem : rem, ol= o[neg];
  if (lte<l>(p, ol)) o[neg]-= p, o[!neg]+= p, y+= D(n[mn].x) * rem;
  else {
   D ou= D(n[mn].x) * ol;
   auto [a, b]= split_cum<l, neg>(lr[neg], p, ol, ou);
   o[neg]= ol - p, ol-= n[b].d, ou+= D(n[b].x) * (o[!neg]= p - ol);
   if constexpr (l) lr[neg]= a;
   else {
    if constexpr (neg) lr[!neg]= join(lr[!neg], mn, a);
    else lr[!neg]= join(a, mn, lr[!neg]);
   }
   if constexpr (neg) y-= ou;
   else y+= ou;
   mn= b;
  }
  rem= 0;
 }
 template <bool r> void add_inf(T x0) {
  if (bf[r] && !lt<r>(bx[r], x0)) return;
  if (assert(!bf[!r] || !lt<r>(bx[!r], x0)), bf[r]= true, bx[r]= x0; !mn) return;
  if (lt<r>(x0, n[mn].x)) return lr[r]= cut<r>(lr[r], x0), void();
  D q= n[lr[!r]].s + D(n[mn].x) * o[!r];
  T v= o[!r] + n[lr[!r]].a;
  lr[!r]= cut<r>(lr[!r], x0);
  if (!r) y-= q, rem+= v;
  else y+= q, rem-= v;
  if (lr[!r]) std::tie(lr[r], mn)= pop<!r>(lr[!r]), lr[!r]= 0;
  else mn= lr[r]= 0;
  o[r]= n[mn].d, o[!r]= 0;
 }
 inline void prop(T x) {
  if constexpr (persistent) mn= create(n[mn].d, n[mn].x);
  n[mn].x+= x;
 }
public:
 // f(x) := 0
 PiecewiseLinearConvex(): mn(0), lr{0, 0}, bf{0, 0}, o{0, 0}, rem(0), bx{0, 0}, y(0) {}
 //  f(x) := sum max(0, a(x-x0))
 PiecewiseLinearConvex(const std::vector<std::pair<T, T>> &ramps): PiecewiseLinearConvex() {
  int m= ramps.size();
  if (!m) return;
  std::vector<std::pair<T, T>> w(m);
  int s= 0, t= 0;
  for (auto [d, x]: ramps) {
   if (d == 0) continue;
   if (d < 0) y-= D(d) * x, rem+= d, d= -d;
   w[s++]= {d, x};
  }
  std::sort(w.begin(), w.begin() + s, [](auto a, auto b) { return a.second < b.second; });
  for (int i= 0; i < s; ++i) {
   if (t && w[t - 1].second == w[i].second) w[t - 1].first+= w[i].first;
   else w[t++]= w[i];
  }
  mn= create(w[0].first, w[0].second), o[1]= n[mn].d, lr[1]= build(w.begin() + 1, w.begin() + t);
 }
 std::string info() {
  std::stringstream ss;
  if (ss << "\n rem:" << rem << ", y:" << y << ", mn:" << mn << ", lr:{" << lr[0] << ", " << lr[1] << "}\n bf[0]:" << bf[0] << ", bf[1]:" << bf[1] << ", bx[0]:" << bx[0] << ", bx[1]:" << bx[1] << "\n " << "o[0]:" << o[0] << ", o[1]:" << o[1] << "\n"; mn) {
   if (lr[0]) info(lr[0], 1, ss);
   ss << " ■ " << n[mn] << '\n';
   if (lr[1]) info(lr[1], 1, ss);
  }
  return ss.str();
 }
 template <class... Args> static inline void rebuild(Args &...plc) {
  static_assert(std::conjunction_v<std::is_same<PiecewiseLinearConvex, Args>...>);
  constexpr size_t m= sizeof...(Args);
  std::array<std::vector<std::pair<T, T>>, m> ls, rs;
  std::array<std::pair<T, T>, m> mns;
  int i= 0;
  (void)(int[]){(mns[i]= {n[plc.mn].d, n[plc.mn].x}, ls[i].resize(n[plc.lr[0]].sz), rs[i].resize(n[plc.lr[1]].sz), dump(ls[i].begin(), plc.lr[0]), dump(rs[i].begin(), plc.lr[1]), ++i)...};
  ni= 1, i= 0;
  (void)(int[]){((plc.mn ? (plc.mn= create(mns[i].first, mns[i].second)) : 0), plc.lr[0]= build(ls[i].begin(), ls[i].end()), plc.lr[1]= build(rs[i].begin(), rs[i].end()), ++i)...};
 }
 static inline void rebuild(std::vector<PiecewiseLinearConvex> &plcs) {
  size_t m= plcs.size();
  std::vector<std::vector<std::pair<T, T>>> ls(m), rs(m);
  std::vector<std::pair<T, T>> mns(m);
  for (int i= m; i--;) mns[i]= {n[plcs[i].mn].d, n[plcs[i].mn].x}, ls[i].resize(n[plcs[i].lr[0]].sz), rs[i].resize(n[plcs[i].lr[1]].sz), dump(ls[i].begin(), plcs[i].lr[0]), dump(rs[i].begin(), plcs[i].lr[1]);
  ni= 1;
  for (int i= m; i--;) (plcs[i].mn ? (plcs[i].mn= create(mns[i].first, mns[i].second)) : 0), plcs[i].lr[0]= build(ls[i].begin(), ls[i].end()), plcs[i].lr[1]= build(rs[i].begin(), rs[i].end());
 }
 static void reset() { ni= 1; }
 static bool pool_empty() {
  if constexpr (persistent) return ni >= NODE_SIZE * 0.8;
  else return ni + 1000 >= NODE_SIZE;
 }
 // f(x) += c
 void add_const(D c) { y+= c; }
 // f(x) += ax, /
 void add_linear(T a) { rem+= a; }
 //  f(x) += max(a(x-x0),b(x-x0)), (a < b)
 void add_max(T a, T b, T x0) {
  assert(a < b);
  if (bf[0] && x0 <= bx[0]) y-= D(b) * x0, rem+= b;
  else if (bf[1] && bx[1] <= x0) y-= D(a) * x0, rem+= a;
  else if (T c= b - a; mn) {
   if (n[mn].x == x0) {
    if constexpr (persistent) mn= create(n[mn].d, n[mn].x);
    n[mn].d+= c, o[1]+= c, y-= D(a) * x0, rem+= a;
   } else {
    if (n[mn].x < x0) lr[1]= insert(lr[1], x0, c), y-= D(a) * x0, rem+= a;
    else lr[0]= insert(lr[0], x0, c), y-= D(b) * x0, rem+= b;
   }
  } else mn= create(c, x0), y-= D(a) * x0, rem+= a, o[0]= 0, o[1]= c;
 }
 // f(x) +=  max(0, a(x-x0))
 void add_ramp(T a, T x0) {
  if (a != 0) a > 0 ? add_max(0, a, x0) : add_max(a, 0, x0);
 }
 // f(x) += a|x-x0|, \/
 void add_abs(T a, T x0) {
  if (assert(a >= 0); a != 0) add_max(-a, a, x0);
 }
 // right=false : f(x) +=  inf  (x < x_0), right=true: f(x) += inf  (x_0 < x)
 void add_inf(bool right= false, T x0= 0) { return right ? add_inf<1>(x0) : add_inf<0>(x0); }
 // f(x) <- f(x-x0)
 void shift(T x0) {
  if (bx[0]+= x0, bx[1]+= x0, y-= D(rem) * x0; mn) prop(x0), prop(lr[0], x0), prop(lr[1], x0);
 }
 // rev=false: f(x) <- min_{y<=x} f(y), rev=true : f(x) <- min_{x<=y} f(y)
 void chmin_cum(bool rev= false) {
  if (bf[0] && bf[1] && bx[0] == bx[1]) y+= D(rem) * bx[0], rem= 0;
  else if (rem != 0) {
   bool r= rem < 0;
   T u= (r ? -rem : rem) - o[r] - n[lr[r]].a;
   if (0 <= u) {
    if (r ^ rev) {
     if (u > 0 && bf[r]) {
      D q= n[lr[r]].s + D(n[mn].x) * o[r] + D(u) * bx[r];
      if (r ? y-= q : y+= q; mn) lr[!r]= join(lr[0], mn, lr[1]);
      o[!r]= u, rem= 0, mn= create(u, bx[r]);
     }
    } else {
     assert(bf[r]);
     D q= n[lr[r]].s + D(n[mn].x) * o[r] + D(u) * bx[r];
     (r ? y-= q : y+= q), rem= 0, mn= lr[r]= 0, o[r]= 0;
    }
    bf[!rev]= false;
    return;
   }
   if ((r ^ rev)) r ? slope_eval_cum<0, 1>() : slope_eval_cum<0, 0>();
   else r ? slope_eval_cum<1, 1>() : slope_eval_cum<1, 0>();
   if constexpr (persistent) mn= create(o[rev], n[mn].x);
   else n[mn].d= o[rev];
  } else if (mn) {
   if (o[rev] == 0) {
    if (lr[rev]) std::tie(lr[rev], mn)= rev ? pop<0>(lr[rev]) : pop<1>(lr[rev]), o[rev]= n[mn].d;
    else mn= 0;
   } else {
    if constexpr (persistent) mn= create(o[rev], n[mn].x);
    else n[mn].d= o[rev];
   }
  }
  bf[!rev]= false, lr[!rev]= 0, o[!rev]= 0;
 }
 //  f(x) <- min_{lb<=y<=ub} f(x-y). (lb <= ub), \_/ -> \__/
 void chmin_slide_win(T lb, T ub) {
  assert(lb <= ub);
  if (bf[0] && bf[1] && bx[0] == bx[1]) y+= D(rem) * bx[0], rem= 0;
  else {
   if (rem != 0) {
    bool r= rem < 0;
    T u= (r ? -rem : rem) - o[r] - n[lr[r]].a;
    if (0 < u) {
     T b[2]= {lb, ub};
     if (bf[r]) {
      D q= n[lr[r]].s + D(n[mn].x) * o[r] + D(u) * bx[r];
      if (r ? y-= q : y+= q; mn) lr[!r]= join(lr[0], mn, lr[1]), prop<0>(lr[!r], b[!r]);
      lr[r]= 0, rem= 0, o[!r]= u, o[r]= 0, mn= create(u, bx[r] + b[!r]);
     } else {
      y-= D(rem) * b[!r];
      if (mn) prop(b[!r]), prop(lr[0], b[!r]), prop(lr[1], b[!r]);
     }
     bx[0]+= lb, bx[1]+= ub;
     return;
    }
    slope_eval(r);
   }
   if (mn) {
    if (o[0] == 0) prop(ub);
    else if (o[1] == 0) prop(lb);
    else lr[1]= join<0>(0, create(o[1], n[mn].x), lr[1]), prop(lb), n[mn].d= o[0], o[1]= 0;
    prop(lr[0], lb), prop(lr[1], ub);
   }
  }
  bx[0]+= lb, bx[1]+= ub;
 }
 D operator()(T x) { return assert(!bf[0] || bx[0] <= x), assert(!bf[1] || x <= bx[1]), calc_y(x) + D(rem) * x + y; }
 D min() {
  if (rem == 0) return y;
  bool r= rem < 0;
  T u= (r ? -rem : rem) - o[r] - n[lr[r]].a;
  if (0 < u) {
   assert(bf[r]);
   D q= n[lr[r]].s + D(n[mn].x) * o[r] + D(u) * bx[r];
   return r ? y - q : y + q;
  }
  return slope_eval(r), y;
 }
 std::array<T, 2> argmin() {
  if (rem != 0) {
   bool r= rem < 0;
   if (o[r] + n[lr[r]].a < (r ? -rem : rem)) {
    assert(bf[r]);
    return {bx[r], bx[r]};
   }
   slope_eval(r);
  }
  std::array<T, 2> ret= {bx[0], bx[1]};
  int t= mn;
  if (!t) return ret;
  bool r= o[0] == 0;
  if (!r && o[1] != 0) ret[0]= ret[1]= n[t].x;
  else if (ret[r]= n[t].x, t= lr[!r]; t) {
   for (; push(t), n[t].ch[r];) t= n[t].ch[r];
   ret[!r]= n[t].x;
  } else assert(bf[!r]);
  return ret;
 }
 size_t size() { return n[lr[0]].sz + n[lr[1]].sz + !!mn; }
 PiecewiseLinearConvex &operator+=(const PiecewiseLinearConvex &g) { return *this= *this + g; }
 PiecewiseLinearConvex operator+(PiecewiseLinearConvex g) const {
  PiecewiseLinearConvex ret= *this;
  if (g.bf[0]) ret.add_inf(false, g.bx[0]);
  if (g.bf[1]) ret.add_inf(true, g.bx[1]);
  if (bf[0]) g.add_inf(false, bx[0]);
  if (bf[1]) g.add_inf(true, bx[1]);
  ret.y+= g.y, ret.rem+= g.rem;
  if (!g.mn) return ret;
  if (!ret.mn) return ret.mn= g.mn, ret.lr[0]= g.lr[0], ret.lr[1]= g.lr[1], ret.o[0]= g.o[0], ret.o[1]= g.o[1], ret;
  ret.y+= n[ret.lr[0]].s + D(n[ret.mn].x) * ret.o[0] + n[g.lr[0]].s + D(n[g.mn].x) * g.o[0], ret.rem-= ret.o[0] + n[ret.lr[0]].a + g.o[0] + n[g.lr[0]].a;
  int t= unite(join(ret.lr[0], ret.mn, ret.lr[1]), join(g.lr[0], g.mn, g.lr[1]));
  return std::tie(ret.lr[1], ret.mn)= pop<0>(t), ret.lr[0]= 0, ret.o[0]= 0, ret.o[1]= n[ret.mn].d, ret;
 }
 std::vector<T> dump_xs() {
  std::vector<T> xs;
  if (bf[0]) xs.push_back(bx[0]);
  dump_xs(lr[0], xs);
  if (mn) xs.push_back(n[mn].x);
  dump_xs(lr[1], xs);
  if (bf[1]) xs.push_back(bx[1]);
  return xs;
 }
 std::vector<std::pair<T, D>> dump_xys() {
  auto xs= dump_xs();
  std::vector<std::pair<T, D>> xys(xs.size());
  for (int i= xs.size(); i--;) xys[i]= {xs[i], operator()(xs[i])};
  return xys;
 }
 std::vector<T> dump_slopes() {
  std::vector<T> as;
  if (mn) as.push_back(-o[0]), dump_slopes_l(lr[0], o[0], as), std::reverse(as.begin(), as.end()), as.push_back(o[1]), dump_slopes_r(lr[1], o[1], as);
  else as.push_back(0);
  for (auto &a: as) a+= rem;
  return as;
 }
};
using namespace std;
signed main() {
 cin.tie(0);
 ios::sync_with_stdio(0);
 int N;
 cin >> N;
 PiecewiseLinearConvex<int> f;
 f.add_inf(true);
 for (int i= 0; i < N; ++i) {
  int a, b;
  cin >> a >> b;
  f.chmin_cum(true);
  f.shift(a - b);
  f.add_abs(1, 0);
 }
 cout << f(0) << '\n';
 return 0;
}
# Verdict Execution time Memory Grader output
1 Correct 0 ms 348 KB Output is correct
2 Correct 0 ms 348 KB Output is correct
3 Correct 1 ms 348 KB Output is correct
4 Correct 14 ms 3084 KB Output is correct
5 Correct 33 ms 5612 KB Output is correct
6 Correct 83 ms 9300 KB Output is correct
7 Correct 72 ms 9044 KB Output is correct
8 Incorrect 121 ms 14932 KB Output isn't correct
9 Halted 0 ms 0 KB -
# Verdict Execution time Memory Grader output
1 Correct 0 ms 348 KB Output is correct
2 Correct 0 ms 348 KB Output is correct
3 Correct 1 ms 348 KB Output is correct
4 Correct 14 ms 3084 KB Output is correct
5 Correct 33 ms 5612 KB Output is correct
6 Correct 83 ms 9300 KB Output is correct
7 Correct 72 ms 9044 KB Output is correct
8 Incorrect 121 ms 14932 KB Output isn't correct
9 Halted 0 ms 0 KB -
# Verdict Execution time Memory Grader output
1 Correct 0 ms 348 KB Output is correct
2 Correct 0 ms 348 KB Output is correct
3 Correct 0 ms 348 KB Output is correct
4 Correct 1 ms 472 KB Output is correct
5 Correct 1 ms 348 KB Output is correct
6 Correct 1 ms 348 KB Output is correct
7 Correct 1 ms 348 KB Output is correct
8 Correct 1 ms 348 KB Output is correct
9 Correct 0 ms 348 KB Output is correct
10 Correct 1 ms 348 KB Output is correct
# Verdict Execution time Memory Grader output
1 Correct 0 ms 348 KB Output is correct
2 Correct 0 ms 348 KB Output is correct
3 Correct 1 ms 348 KB Output is correct
4 Correct 0 ms 348 KB Output is correct
5 Correct 1 ms 472 KB Output is correct
6 Correct 1 ms 348 KB Output is correct
7 Correct 1 ms 348 KB Output is correct
8 Correct 1 ms 348 KB Output is correct
9 Correct 1 ms 348 KB Output is correct
10 Correct 0 ms 348 KB Output is correct
11 Correct 1 ms 348 KB Output is correct
12 Correct 1 ms 348 KB Output is correct
13 Correct 1 ms 604 KB Output is correct
14 Correct 1 ms 564 KB Output is correct
15 Correct 1 ms 2396 KB Output is correct
16 Correct 1 ms 348 KB Output is correct
17 Correct 1 ms 348 KB Output is correct
18 Correct 1 ms 348 KB Output is correct
# Verdict Execution time Memory Grader output
1 Correct 0 ms 348 KB Output is correct
2 Correct 0 ms 348 KB Output is correct
3 Correct 1 ms 348 KB Output is correct
4 Correct 0 ms 348 KB Output is correct
5 Correct 1 ms 472 KB Output is correct
6 Correct 1 ms 348 KB Output is correct
7 Correct 1 ms 348 KB Output is correct
8 Correct 1 ms 348 KB Output is correct
9 Correct 1 ms 348 KB Output is correct
10 Correct 0 ms 348 KB Output is correct
11 Correct 14 ms 3084 KB Output is correct
12 Correct 33 ms 5612 KB Output is correct
13 Correct 83 ms 9300 KB Output is correct
14 Correct 72 ms 9044 KB Output is correct
15 Incorrect 121 ms 14932 KB Output isn't correct
16 Halted 0 ms 0 KB -