题目描述

Farmer John is gathering the cows. His farm contains a network of N (1 <= N <= 100,000) fields conveniently numbered 1..N and connected by N-1 unidirectional paths that eventually lead to field 1. The fields and paths form a tree.

Each field i > 1 has a single one-way, exiting path to field P_i, and currently contains C_i cows (1 <= C_i <= 1,000,000,000). In each time unit, no more than M_i (0 <= M_i <= 1,000,000,000) cows can travel from field i to field P_i (1 <= P_i <= N) (i.e., only M_i cows can traverse the path).

Farmer John wants all the cows to congregate in field 1 (which has no limit on the number of cows it may have). Rules are as follows:

  • Time is considered in discrete units.

  • Any given cow might traverse multiple paths in the same time unit. However, no more than M_i total cows can leave field i (i.e., traverse its exit path) in the same time unit.

  • Cows never move *away* from field #1.

In other words, every time step, each cow has the choice either to

a) stay in its current field

b) move through one or more fields toward field #1, as long as the bottleneck constraints for each path are not violated

Farmer John wants to know how many cows can arrive in field 1 by certain times. In particular, he has a list of K (1 <= K <= 10,000) times T_i (1 <= T_i <= 1,000,000,000), and he wants to know, for each T_i in the list, the maximum number of cows that can arrive at field 1 by T_i if scheduled to optimize this quantity.

Consider an example where the tree is a straight line, and the T_i list contains only T_1=5, and cows are distibuted as shown:

Locn:      1---2---3---4      <-- Pasture ID numbers
C_i: 0 1 12 12 <-- Current number of cows
M_i: 5 8 3 <-- Limits on path traversal; field 1 has no limit since it has no exit
The solution is as follows; the goal is to move cows to field 1:

Tree: 1---2---3---4

t=0        0   1   12  12     <-- Initial state
t=1 5 4 7 9 <-- field 1 has cows from field 2 and 3 t=2 10 7 2 6
t=3 15 7 0 3
t=4 20 5 0 0
t=5 25 0 0 0
Thus, the answer is 25: all 25 cows can arrive at field 1 by time t=5.
Farmer John有一张N个农场构成的网络(1 <= N <= 100,000) ,我们将农场标记为1N。 农场被N-1条单向道路连接,保证从任何一个农场可以到达1号农场。FJ想让奶牛到1号农场集中 (P.S. 至于要做什么我也不知道)。 对于每个农场i > 1都有一条单独的单向道路通往P_i,并且这个农场里有C_i只奶牛 (1 <= C_i <= 1,000,000,000)。

在每个单位时间里,这条道路允许不超过M_i (0 <= M_i <= 1,000,000,000)只奶牛从农场i走到农场P_i (1 <= P_i <= N)。 Farmer John 想让所有的奶牛都集中在1号农场(农场容纳奶牛的数量是没有限制的)。 下面是奶牛集中到1号农场过程的规则:

  • 我们认为时间是离散的 * 任何奶牛都可以在一个单位时间里走过任意多条道路。但是,必须满足每条道路的上限M_i。

  • 奶牛从来不会离开1号农场。 换句话说,每一个单位时间,每只奶牛可以选择下面行动之一:

a) 留在当前的农场

b) 经过一条或者多条道路,向1号农场移动。

同样,需要满足每条道路的上限M_i。

FJ想知道有多少奶牛可以在某个特定的时刻到达1号农场。

特别的,他有一张列着K (1 <= K <= 10,000) 个时间T_i (1 <= T_i <= 1,000,000,000)的单子,他想知道对于每个T_i,如果采用最优 的策略在这个时刻结束时最多能有多少奶牛到达1号农场。

输入输出格式

输入格式:

  • Line 1: Two space-separated integers: N and K

  • Lines 2..N: Line i (not i+1) describes field i with three

space-separated integers: P_i, C_i, and M_i

  • Lines N+1..N+K: Line N+i contains a single integer: T_i

输出格式:

  • Lines 1..K: Line i contains a single integer that is the maximum number of cows that can arrive at field 1 by time T_i.

输入输出样例

输入样例#1:

4 1
1 1 5
2 12 7
3 12 3
5
输出样例#1:

25
 1 #include <algorithm>
2 #include <iostream>
3 #include <cstring>
4 #include <cstdio>
5 #include <queue>
6 typedef long long LL;
7 using namespace std;
8 const int N=100010;
9 struct Query{LL t,res;int id;}ask[N];
10 bool cmp1(Query x,Query y){return x.t<y.t;}
11 bool cmp2(Query x,Query y){return x.id<y.id;}
12 struct Node{
13 LL t;int x;Node(LL t_=0,int x_=0){t=t_;x=x_;}
14 friend bool operator<(Node a,Node b){return a.t>b.t;}
15 };
16 priority_queue<Node>q;
17 int fa[N],anc[N],lim[N];
18 LL cow[N],pass[N];
19 int Find(int x){
20 if(anc[x]==x)return x;
21 return anc[x]=Find(anc[x]);
22 }
23 int n,Q;
24 int main(){
25 freopen("bottleneck.in","r",stdin);
26 freopen("bottleneck.out","w",stdout);
27 scanf("%d%d",&n,&Q);
28 for(int i=1;i<=n;i++)
29 anc[i]=i;
30 for(int i=2;i<=n;i++){
31 scanf("%d",&fa[i]);
32 scanf("%lld",&cow[i]);
33 scanf("%lld",&lim[i]);
34 pass[fa[i]]-=lim[i];
35 pass[i]+=lim[i];
36 }
37 for(int i=1;i<=Q;i++){
38 scanf("%lld",&ask[i].t);
39 ask[i].id=i;
40 }sort(ask+1,ask+Q+1,cmp1);
41
42 for(int i=2;i<=n;i++)
43 if(pass[i]>0){
44 LL t=cow[i]/pass[i];
45 q.push(Node(t,i));
46 }
47
48 int p=1,x,tp;
49 while(!q.empty()&&p<=Q){
50 while(p<=Q&&ask[p].t<=q.top().t)
51 ask[p].res=cow[1]-pass[1]*ask[p].t,p++;
52 if(anc[q.top().x]!=q.top().x){q.pop();continue;}
53 x=q.top().x;tp=Find(fa[x]);cow[tp]+=cow[x];
54 pass[tp]+=pass[x];anc[x]=tp;
55 if(pass[tp]>0){
56 LL t=cow[tp]/pass[tp];
57 q.push(Node(t,tp));
58 }
59 q.pop();
60 }sort(ask+1,ask+Q+1,cmp2);
61 for(int i=1;i<=Q;i++)
62 printf("%lld\n",ask[i].res);
63 return 0;
64 }

模拟(堆):USACO Jan11 瓶颈的更多相关文章

  1. Java堆内存溢出模拟

    先了解一下Java堆: 关于Java内存区域的分配,可以查看Java运行时数据区域一篇文章. Java堆是虚拟机内存管理中最大的一块区域,该区域是线程共享的,某Java进程中所有的线程都可以访问该区域 ...

  2. Eclipse Memory Analysis进行堆转储文件分析

    生成堆转储文件 新建项目,设置Eclispe Java堆的大小: (1)限制Java堆大小:将最小值 -Xms参数与最大值-Xmx参数设置一样可避免堆的扩展         -Xmx20m -Xms2 ...

  3. USACO Section 1.1 Your Ride Is Here 解题报告

    题目 问题描述 将字符串转变为数字,字母A对应的值为1,依次对应,字母Z对应的值为26.现在有一个字符串,将其中的每个字符转变为数字之后进行累乘,最终的结果对47求余数. 题目给你两个字符串,其中的字 ...

  4. 基本数据结构——堆(Heap)的基本概念及其操作

    基本数据结构――堆的基本概念及其操作 小广告:福建安溪一中在线评测系统 Online Judge 在我刚听到堆这个名词的时候,我认为它是一堆东西的集合... 但其实吧它是利用完全二叉树的结构来维护一组 ...

  5. Java 伙伴系统(模拟)

    参考:https://labrick.cc/2015/10/12/buddy-system-algorithm/ 代码过烂 不宜参考. output: [operating.entity.Heap@4 ...

  6. 堆排序(大顶堆、小顶堆)----C语言

    堆排序 之前的随笔写了栈(顺序栈.链式栈).队列(循环队列.链式队列).链表.二叉树,这次随笔来写堆 1.什么是堆? 堆是一种非线性结构,(本篇随笔主要分析堆的数组实现)可以把堆看作一个数组,也可以被 ...

  7. Java伙伴系统(模拟)

    参考:https://labrick.cc/2015/10/12/buddy-system-algorithm/ 代码过烂 不宜参考. output: [operating.entity.Heap@4 ...

  8. hihocoder 1109 堆优化的Prim算法

    题目链接:http://hihocoder.com/problemset/problem/1109 , 最小生成树 + 堆优化(优先队列). 可以用优先队列,也可以自己手动模拟堆,为了练手,我两种都试 ...

  9. 数据结构中的堆(Heap)

    堆排序总结 这是排序,不是查找!!!查找去找二叉排序树等. 满二叉树一定是完全二叉树,但完全二叉树不一定是满二叉树. 构建顶堆: a.构造初始堆 b.从最后一层非叶节点开始调整,一直到根节点 c.如果 ...

随机推荐

  1. UVA 12097 LA 3635 Pie(二分法)

    Pie My birthday is coming up and traditionally I'm serving pie. Not just one pie, no, I have a numbe ...

  2. bzoj1901:Zju2112 Dynamic Rankings

    思路:树套树,我写了两种,一种是线段树套splay,线段树维护区间信息,splay维护第k大,一种是树状数组套权值线段树(并不是什么可持久化线段树,只不过是动态开点罢了,为什么网上一大堆题解都是可持久 ...

  3. 状态模式(State Pattern)

    状态模式:允许对象在内部状态改变时改变它的行为,对象看起来好像修改了它的类. 这个模式将状态封装成为独立的类,并将动作委托到代表当前对象的对象,这样行为就与拥有状态类解耦了. 从客户的角度来看,对象的 ...

  4. [zz]安装Ubuntu 15.10后要做的事

    很有用,收藏 http://blog.csdn.net/skykingf/article/details/45267517

  5. mysql主从之主键冲突

    收到短信报警,两台数据库都报slave同步失败了,先说明一下环境,架构:lvs+keepalived+amoeba+mysql,主主复制,单台写入, 主1:192.168.0.223(写) 主2:19 ...

  6. 用javascript操作xml(二)JavaScript 将XML转换成字符串(xml to string)

    function xmlToString(xmlData) { var xmlString; //IE if (window.ActiveXObject){ xmlString = xmlData.x ...

  7. onbeforeunload、beforeunload

    <script type="text/javascript"> function addOnBeforeUnload(e) {     var ev = e || ev ...

  8. python三级菜单的实现

    一.作业要求 1.使用字典实现三级菜单功能 2.直接输入前面数字进入下一级菜单 3.按B返回上一级,按Q退出 二.需要知识点 1.if循环 2.for循环,enumerate的用法 3.while循环 ...

  9. C语言-07其它相关

    预处理指令 /* 不带参数的宏定义 1.所有的预处理指令都是以#开头 2.预处理指令分3种 1> 宏定义 2> 条件编译 3> 文件包含 3.预处理指令在代码翻译成0和1之前执行 4 ...

  10. 贴板子系列_1-exgcd

    exgcd ll exgcd(ll a,ll b,ll &x,ll &y) { ) { x=;y=;return a; } ll r=exgcd(b,a%b,x,y); ll t=x; ...